diff --git a/@file_array/cat.m b/@file_array/cat.m
index 7671e34..df738ef 100644
--- a/@file_array/cat.m
+++ b/@file_array/cat.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: cat.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: cat.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 if dr>32 || dr<0, error('Unknown command option.'); end;
@@ -36,5 +36,5 @@
     error('All matrices on a row in the bracketed expression must have the same number of rows.');
 else
     o = vertcat(tmp{:});
-    o = class(o,'file_array');
+    o = file_array(o);
 end;
diff --git a/@file_array/disp.m b/@file_array/disp.m
index 49fc219..5a81693 100644
--- a/@file_array/disp.m
+++ b/@file_array/disp.m
@@ -4,7 +4,7 @@ function disp(obj)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: disp.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: disp.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 if numel(struct(obj))>1,
@@ -19,7 +19,7 @@ function disp(obj)
         fprintf('%d\n',sz(end));
     end;
 else
-    display(mystruct(obj))
+    disp(mystruct(obj))
 end;
 return;
 %=======================================================================
diff --git a/@file_array/file_array.m b/@file_array/file_array.m
index 34c9e1c..59819a4 100644
--- a/@file_array/file_array.m
+++ b/@file_array/file_array.m
@@ -13,7 +13,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: file_array.m 1340 2008-04-09 17:11:23Z john $
+% $Id: file_array.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 if nargin==1
@@ -38,5 +38,5 @@
 if nargin>=7, a = permission(a,varargin{7}); end;
 
 a.pos = ones(size(a.dim));
-a     = class(a,'file_array');
+a     = file_array(a);
 
diff --git a/@file_array/private/datatypes.m b/@file_array/private/datatypes.m
index e3e8c45..0e2ea0b 100644
--- a/@file_array/private/datatypes.m
+++ b/@file_array/private/datatypes.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: datatypes.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: datatypes.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 persistent dtype
@@ -24,10 +24,10 @@
         1280,'UINT64'    ,'uint64'  ,@uint64 ,1,8  ,t,t,f
         16  ,'FLOAT32'   ,'float32' ,@single ,1,4  ,f,f,t
         64  ,'FLOAT64'   ,'double'  ,@double ,1,8  ,f,f,t
-        1536,'FLOAT128'  ,'float128',@crash  ,1,16 ,f,f,f
+        1536,'FLOAT128'  ,'float128',@error  ,1,16 ,f,f,f
         32  ,'COMPLEX64' ,'float32' ,@single ,2,4  ,f,f,f
         1792,'COMPLEX128','double'  ,@double ,2,8  ,f,f,f
-        2048,'COMPLEX256','float128',@crash  ,2,16 ,f,f,f
+        2048,'COMPLEX256','float128',@error  ,2,16 ,f,f,f
         128 ,'RGB24'     ,'uint8'   ,@uint8  ,3,1  ,t,t,f};
     dtype = struct(...
         'code'     ,table(:,1),...
diff --git a/@file_array/subsasgn.m b/@file_array/subsasgn.m
index a47c540..6c2f7bc 100644
--- a/@file_array/subsasgn.m
+++ b/@file_array/subsasgn.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: subsasgn.m 1355 2008-04-10 10:52:35Z vladimir $
+% $Id: subsasgn.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 if isempty(subs)
@@ -47,7 +47,7 @@
 end;
 
 dm   = [dm ones(1,16)];
-do   = ones(1,16);
+di   = ones(1,16);
 args = {};
 for i=1:length(subs.subs),
     if ischar(subs.subs{i}),
@@ -56,7 +56,7 @@
     else
         args{i} = int32(subs.subs{i});
     end;
-    do(i) = length(args{i});
+    di(i) = length(args{i});
 end;
 for j=1:length(sobj),
     if strcmp(sobj(j).permission,'ro'),
@@ -69,7 +69,7 @@
     if numel(dat)~=1,
         subfun(sobj,double(dat),args{:});
     else
-        dat1 = double(dat) + zeros(do);
+        dat1 = double(dat) + zeros(di);
         subfun(sobj,dat1,args{:});
     end;
 else
diff --git a/@file_array/subsref.m b/@file_array/subsref.m
index e59bcdd..0f952ff 100644
--- a/@file_array/subsref.m
+++ b/@file_array/subsref.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: subsref.m 3958 2010-06-30 16:24:46Z guillaume $
+% $Id: subsref.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 if isempty(subs), return; end
@@ -39,7 +39,7 @@
     end
 end
 
-do   = ones(16,1);
+di   = ones(16,1);
 args = cell(1,length(subs.subs));
 for i=1:length(subs.subs)
     if ischar(subs.subs{i})
@@ -58,7 +58,7 @@
     else
         args{i} = subs.subs{i};
     end
-    do(i) = length(args{i});
+    di(i) = length(args{i});
 end
 
 if length(sobj)==1
@@ -66,7 +66,7 @@
 else
     dt  = datatypes;
     dt  = dt([dt.code]==sobj(1).dtype); % assuming identical datatypes
-    t = zeros(do',func2str(dt.conv));
+    t = zeros(di',func2str(dt.conv));
     for j=1:length(sobj)
         ps = [sobj(j).pos ones(1,length(args))];
         dm = [sobj(j).dim ones(1,length(args))];
diff --git a/@gifti/display.m b/@gifti/display.m
index 9f151fe..ce56cd0 100644
--- a/@gifti/display.m
+++ b/@gifti/display.m
@@ -6,7 +6,7 @@ function display(this)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: display.m 2076 2008-09-10 12:34:08Z guillaume $
+% $Id: display.m 4182 2011-02-01 12:29:09Z guillaume $
 
 display_name = inputname(1);
 if isempty(display_name)
@@ -17,6 +17,9 @@ function display(this)
     eval([display_name ' = struct(this);']);
     eval(['display(' display_name ');']);
 else
+    disp(' ')
+    disp([display_name ' =']);
+    disp(' ');
     eval([display_name ' = this;']);
-    eval(['builtin(''display'',' display_name ');']);
+    eval(['disp(' display_name ');']);
 end
\ No newline at end of file
diff --git a/@gifti/private/read_gifti_file.m b/@gifti/private/read_gifti_file.m
index 74861b3..a5641fa 100644
--- a/@gifti/private/read_gifti_file.m
+++ b/@gifti/private/read_gifti_file.m
@@ -7,7 +7,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: read_gifti_file.m 3999 2010-07-19 10:54:18Z guillaume $
+% $Id: read_gifti_file.m 4013 2010-07-22 17:12:45Z guillaume $
 
 % Import XML-based GIfTI file
 %--------------------------------------------------------------------------
@@ -23,7 +23,7 @@
     error('[GIFTI] %s is not a GIFTI 1.0 file.', filename);
 end
 attr = cell2mat(attributes(t,'get',root(t)));
-attr = cell2struct({attr.val},strrep({attr.key},':','_'),2);
+attr = cell2struct({attr.val},strrep({attr.key},':','___'),2);
 if ~all(ismember({'Version','NumberOfDataArrays'},fieldnames(attr)))
     error('[GIFTI] Missing mandatory attributes for GIFTI root element.');
 end
diff --git a/@gifti/save.m b/@gifti/save.m
index 4ecffb7..5694757 100644
--- a/@gifti/save.m
+++ b/@gifti/save.m
@@ -10,7 +10,7 @@ function save(this,filename,encoding)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: save.m 3999 2010-07-19 10:54:18Z guillaume $
+% $Id: save.m 4022 2010-07-28 12:50:20Z guillaume $
 
 error(nargchk(1,3,nargin));
 
@@ -66,9 +66,9 @@ function save(this,filename,encoding)
 % Defaults for DataArray's attributes
 %--------------------------------------------------------------------------
 [unused,unused,mach]   = fopen(fid);
-if ~isempty(strmatch('ieee-be',mach))
+if strncmp('ieee-be',mach,7)
     def.Endian         = 'BigEndian';
-elseif ~isempty(strmatch('ieee-le',mach))
+elseif strncmp('ieee-le',mach,7)
     def.Endian         = 'LittleEndian';
 else
     error('[GIFTI] Unknown byte order "%s".',mach);
diff --git a/@gifti/subsref.m b/@gifti/subsref.m
index 7671693..19af133 100644
--- a/@gifti/subsref.m
+++ b/@gifti/subsref.m
@@ -4,13 +4,14 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: subsref.m 3999 2010-07-19 10:54:18Z guillaume $
+% $Id: subsref.m 4136 2010-12-09 22:22:28Z guillaume $
 
-if length(this) > 1
+if length(this) > 1 && ~strcmp(subs(1).type,'()')
     warning('Not implemented.');
     for i=1:numel(this)
         varargout{i} = subsref(this(i),subs);
     end
+    return;
 end
 
 switch subs(1).type
diff --git a/@meeg/badchannels.m b/@meeg/badchannels.m
index f1b9388..ac9d713 100644
--- a/@meeg/badchannels.m
+++ b/@meeg/badchannels.m
@@ -5,18 +5,19 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: badchannels.m 3942 2010-06-21 14:03:28Z vladimir $
+% $Id: badchannels.m 4040 2010-08-20 23:25:46Z vladimir $
 
 
-if length(varargin) == 2
+
+if length(varargin) == 2 && ~isempty(varargin{1})
     % make sure that the two inputs for set are the same length
     if ~(length(varargin{2}) == 1 | (length(varargin{1}) == length(varargin{2})))
         error('Use either same vector length or scalar for value');
     end
 end
 
-if numel(varargin) >= 1    
-    if ~(varargin{1} >= 1 & varargin{1} <= nchannels(this))
+if numel(varargin) >= 1  && ~isempty(varargin{1})  
+    if ~(all(varargin{1} >= 1) && all(varargin{1} <= nchannels(this)))
         error('Channel number of out range.');
     end
 end
diff --git a/@meeg/clone.m b/@meeg/clone.m
index d138cbf..541aee2 100644
--- a/@meeg/clone.m
+++ b/@meeg/clone.m
@@ -11,7 +11,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel, Vladimir Litvak
-% $Id: clone.m 3807 2010-04-06 19:29:08Z vladimir $
+% $Id: clone.m 4207 2011-02-22 10:48:36Z christophe $
 
 if nargin < 4
     reset = 0;
@@ -32,11 +32,26 @@
 if isempty(pth)
     pth = this.path;
 end
-newFileName = [fullfile(pth,fname),ext];
+newFileName = [fullfile(pth,fname),'.dat'];
 % copy the file_array
 d = this.data.y; % 
 d.fname = newFileName;
-d.dim = dim; 
+dim_o = d.dim;
+
+% This takes care of an issue specific to data files with a scaling factor
+% which are not officially supported in SPM8 (float without scaling).
+% Also assuming scaling is the *same* for all channels...
+if dim(1)>dim_o(1) && length(d.scl_slope)>1
+    % adding channel to montage and scl_slope defined for old montage
+    %       -> need to increase size of scl_slope
+    v_slope = mode(d.scl_slope);
+    if length(v_slope)>1
+        warning(['Trying to guess the scaling factor for new channels.',...
+            ' This factor might be wrong now.']);        
+    end
+    d.scl_slope = [d.scl_slope' ones(1,dim(1)-dim_o(1))*v_slope]';
+end
+d.dim = dim;
 
 % physically initialise file
 if length(dim) == 3
diff --git a/@meeg/frequencies.m b/@meeg/frequencies.m
index 2b95477..9fd8f01 100644
--- a/@meeg/frequencies.m
+++ b/@meeg/frequencies.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: frequencies.m 3579 2009-11-18 16:21:55Z vladimir $
+% $Id: frequencies.m 4015 2010-07-23 16:35:03Z vladimir $
 
 if nargin < 3
     if strncmpi(transformtype(this), 'TF',2)
@@ -34,7 +34,11 @@
           error('Wrong frequency axis or indices'); 
     end
 
-    this.transform.frequencies(ind) = f;
+    if length(ind) == size(this.data.y, 2)
+        this.transform.frequencies = f;
+    else
+        this.transform.frequencies(ind) = f;
+    end
     
     res = this;    
 end
diff --git a/@meeg/private/checkmeeg.m b/@meeg/private/checkmeeg.m
index 46386d4..0ee19f9 100644
--- a/@meeg/private/checkmeeg.m
+++ b/@meeg/private/checkmeeg.m
@@ -9,7 +9,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: checkmeeg.m 3978 2010-07-08 14:26:39Z vladimir $
+% $Id: checkmeeg.m 4207 2011-02-22 10:48:36Z christophe $
 
 if nargin==1
     option = 'basic';
@@ -64,10 +64,24 @@
     end
     
     for k = 1:Ntrials
-        if isnumeric(meegstruct.trials(k).label)
-            meegstruct.trials(k).label = num2str(meegstruct.trials(k).label);
+        
+        label = meegstruct.trials(k).label;
+        
+        if iscell(label) && numel(label) == 1
+            label = label{1};
         end
         
+        if isnumeric(label)
+            label = num2str(label);
+        end
+        
+        if isa(label, 'char')
+           meegstruct.trials(k).label = label;
+        else
+           meegstruct.trials(k).label = 'Unknown';
+           disp('checkmeeg: some trial labels were not strings, changing back to ''Unknown''');
+        end                        
+        
         if  length(meegstruct.trials(k).bad)>1 || ~ismember(meegstruct.trials(k).bad, [0, 1])
             disp(['checkmeeg: illegal value for bad flag in trial ' num2str(k) ', setting to zero.']);
             meegstruct.trials(k).bad = 0;
@@ -135,6 +149,11 @@
     else
         [meegstruct.channels(find(cellfun('isempty', {meegstruct.channels.bad}))).bad] = deal(0);
     end
+    
+    for i = 1:Nchannels
+        meegstruct.channels(i).bad = double(~~meegstruct.channels(i).bad);
+    end            
+    
     if ~isfield(meegstruct.channels, 'type')
         disp('checkmeeg: no channel type, assigning default');
         [meegstruct.channels.type] = deal('Other');
@@ -199,12 +218,20 @@
     end
     
     if isa(meegstruct.data.y, 'file_array')
+        % catching up (unlikely case) where filearray.fname is
+        % different from data.fnamedat -> set data.fnamedat
+        [junk, yfname, yext] = fileparts(meegstruct.data.y.fname);
+        [junk, dfname, dext] = fileparts(meegstruct.data.fnamedat);
+        if ~strcmp([yfname yext],[dfname dext])
+            meegstruct.data.fnamedat = [yfname yext];
+        end
+        % save original file_array scale & offset, just in case
+        sav_sc = meegstruct.data.y.scl_slope;
+        sav_os = meegstruct.data.y.offset;
         try
             % Try reading data, i.e. check if it's a "good" filearray
             meegstruct.data.y(1, 1, 1);
         catch
-            % save original file_array scale, just in case
-            sav_sc = meegstruct.data.y.scl_slope;
             meegstruct.data.y = [];
         end
     end
@@ -235,9 +262,14 @@
                 error('Unknown transform type');
         end
         % and restore original file_array scale, if available (exist) & useful (~=[])
-        if exist('sav_sc','var') && ~isempty(sav_sc)
+        if exist('sav_sc','var') && ~isempty(sav_sc) && ...
+                   size(meegstruct.data.y, 1) == length(sav_sc)            
             meegstruct.data.y.scl_slope = sav_sc;
         end
+        % and restore original file_array offset, if available (exist) & useful (~=0)
+        if exist('sav_os','var') && sav_os
+            meegstruct.data.y.offset = sav_os;
+        end
         
     end
     
diff --git a/@nifti/disp.m b/@nifti/disp.m
index 3f2dea3..68e0efb 100644
--- a/@nifti/disp.m
+++ b/@nifti/disp.m
@@ -4,7 +4,7 @@ function disp(obj)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: disp.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: disp.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 sz = size(obj);
@@ -18,6 +18,6 @@ function disp(obj)
     fprintf('%d\n',sz(end));
 end;
 if prod(sz)==1,
-    display(structn(obj))
+    disp(structn(obj))
 end;
 return;
diff --git a/@nifti/nifti.m b/@nifti/nifti.m
index dbc0987..76c3e63 100644
--- a/@nifti/nifti.m
+++ b/@nifti/nifti.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: nifti.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: nifti.m 4270 2011-03-29 16:26:26Z john $
 
 
 switch nargin
@@ -59,7 +59,7 @@
         dim   = double(vol.hdr.dim);
         dim   = dim(2:(dim(1)+1));
         dt    = double(vol.hdr.datatype);
-        offs  = double(vol.hdr.vox_offset);
+        offs  = max(double(vol.hdr.vox_offset),0);
 
         if ~vol.hdr.scl_slope && ~vol.hdr.scl_inter,
             vol.hdr.scl_slope = 1;
diff --git a/@nifti/subsasgn.m b/@nifti/subsasgn.m
index 5d92060..440202d 100644
--- a/@nifti/subsasgn.m
+++ b/@nifti/subsasgn.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: subsasgn.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: subsasgn.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 switch subs(1).type,
@@ -16,11 +16,11 @@
     objs = struct(obj);
     for i=1:length(varargin),
         val     = varargin{i};
-        obji    = class(objs(i),'nifti');
+        obji    = nifti(objs(i));
         obji    = fun(obji,subs,val);
         objs(i) = struct(obji);
     end;
-    obj = class(objs,'nifti');
+    obj = nifti(objs);
 
 case {'()'},
     objs = struct(obj);
@@ -33,7 +33,7 @@
         %end;
         for i=1:numel(t),
             val  = varargin{1};
-            obji = class(t(i),'nifti');
+            obji = nifti(t(i));
             obji = subsasgn(obji,subs(2:end),val);
             t(i) = struct(obji);
         end;
@@ -51,7 +51,7 @@
             error('Assignment between unlike types is not allowed.');
         end;
     end;
-    obj = class(objs,'nifti');
+    obj = nifti(objs);
 
 otherwise
     error('Cell contents reference from a non-cell array object.');
@@ -82,7 +82,7 @@
         if ~isfield(obj.hdr,'magic'), error('Not a NIFTI-1 header'); end;
 
         if length(subs)>1, % && ~strcmpi(subs(1).subs,{'raw','dat'}),
-            val0 = subsref(class(obj,'nifti'),subs(1));
+            val0 = subsref(nifti(obj),subs(1));
             val1 = subsasgn(val0,subs(2:end),val);
         else
             val1 = val;
@@ -333,7 +333,7 @@
 
         objs(ii) = obj;
     end
-    obj = class(objs,'nifti');
+    obj = nifti(objs);
 
 otherwise
     error('This should not happen.');
@@ -345,11 +345,11 @@
 function obj = assigndat(obj,val)
 if isa(val,'file_array'),
     sz = size(val);
-    if numel(sz)>8,
+    if numel(sz)>7,
         error('Too many dimensions in data.');
     end;
-    sz = [sz 1 1 1 1 1 1 1 1];
-    sz = sz(1:8);
+    sz = [sz 1 1 1 1 1 1 1];
+    sz = sz(1:7);
     sval = struct(val);
     d    = findindict(sval.dtype,'dtype');
     if isempty(d)
@@ -359,10 +359,10 @@
     [pth,nam,suf]    = fileparts(sval.fname);
     if any(strcmp(suf,{'.img','.IMG'}))
         val.offset = max(sval.offset,0);
-        obj.hdr.magic = 'ni1';
+        obj.hdr.magic = ['ni1' char(0)];
     elseif any(strcmp(suf,{'.nii','.NII'}))
         val.offset = max(sval.offset,352);
-        obj.hdr.magic = 'n+1';
+        obj.hdr.magic = ['n+1' char(0)];
     else
         error(['Unknown filename extension (' suf ').']);
     end;
diff --git a/@nifti/subsref.m b/@nifti/subsref.m
index 62878d8..b55620c 100644
--- a/@nifti/subsref.m
+++ b/@nifti/subsref.m
@@ -39,7 +39,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 %
-% $Id: subsref.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: subsref.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 varargout = rec(opt,subs);
@@ -51,7 +51,7 @@
     c = {};
     opts = struct(opt);
     for ii=1:numel(opts)
-        opt = class(opts(ii),'nifti');
+        opt = nifti(opts(ii));
         %if ~isstruct(opt)
         %    error('Attempt to reference field of non-structure array.');
         %end;
@@ -228,12 +228,12 @@
     if length(subs)>1
         c = {};
         for i=1:numel(t),
-            ti = class(t(i),'nifti');
+            ti = nifti(t(i));
             ti = rec(ti,subs(2:end));
             c  = {c{:}, ti{:}};
         end;
     else
-        c    = {class(t,'nifti')};
+        c    = {nifti(t)};
     end;
 
 otherwise
diff --git a/@xmltree/private/xml_findstr.c b/@xmltree/private/xml_findstr.c
index 6c3c978..7a19e57 100644
--- a/@xmltree/private/xml_findstr.c
+++ b/@xmltree/private/xml_findstr.c
@@ -1,7 +1,7 @@
 #include "mex.h"
 
 /*
- * $Id: xml_findstr.c 1460 2008-04-21 17:43:18Z guillaume $
+ * $Id: xml_findstr.c 4151 2011-01-07 18:09:25Z guillaume $
  * Guillaume Flandin <guillaume@artefact.tk>
  */
 
@@ -10,16 +10,18 @@
         - allows to search only the n first occurences of a pattern
         - allows to search only in a substring (given an index of the beginning)
    
-    Matlab hack:
+    MATLAB hack:
         - doesn't use mxGetString to prevent a copy of the string.
-        - assumes Matlab stores strings as unsigned short (Unicode 16 bits)
+        - assumes MATLAB stores strings as unsigned short (Unicode 16 bits)
           matrix.h: typedef uint16_T mxChar;
-          (that's the case for Matlab 5.* and 6.* but Matlab 4.* stores strings
-           as double)
+          (that's the case for MATLAB 5.*, 6.* and 7.* but MATLAB 4.* stores strings
+           as double and Octave as char, see src/mxarray.h)
 */
 
 /* Comment the following line to use standard mxGetString (slower) */
+#if !defined (HAVE_OCTAVE)
 #define __HACK_MXCHAR__
+#endif
 
 void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
 
diff --git a/@xmltree/private/xml_parser.m b/@xmltree/private/xml_parser.m
index f05f9ca..0ec5da3 100644
--- a/@xmltree/private/xml_parser.m
+++ b/@xmltree/private/xml_parser.m
@@ -16,7 +16,7 @@
 % Copyright (C) 2002-2008  http://www.artefact.tk/
 
 % Guillaume Flandin <guillaume@artefact.tk>
-% $Id: xml_parser.m 1460 2008-04-21 17:43:18Z guillaume $
+% $Id: xml_parser.m 4013 2010-07-22 17:12:45Z guillaume $
 
 % XML Processor for MATLAB (The Mathworks, Inc.).
 % Copyright (C) 2002-2008 Guillaume Flandin <Guillaume@artefact.tk>
@@ -308,8 +308,8 @@
     while 1,
         eq = xml_findstr(str,'=',1,1);
         if isempty(str) || isempty(eq), return; end
-        id = xml_findstr(str,'"',1,1);       % should also look for ''''
-        nextid = xml_findstr(str,'"',id+1,1);% rather than only '"'
+        id = sort([xml_findstr(str,'"',1,1),xml_findstr(str,'''',1,1)]); id=id(1);
+        nextid = sort([xml_findstr(str,'"',id+1,1),xml_findstr(str,'''',id+1,1)]);nextid=nextid(1);
         nbattr = nbattr + 1;
         all{nbattr}.key = strip(str(1:(eq-1)));
         all{nbattr}.val = entity(str((id+1):(nextid-1)));
diff --git a/Contents.m b/Contents.m
index dde85d3..5e9e743 100644
--- a/Contents.m
+++ b/Contents.m
@@ -1,5 +1,5 @@
 % Statistical Parametric Mapping
-% Version 4010 (SPM8) 21-Jul-10
+% Version 4290 (SPM8) 04-Apr-11
 %__________________________________________________________________________
 %  ___  ____  __  __
 % / __)(  _ \(  \/  )  
@@ -40,9 +40,9 @@
 % You should  have received a copy of the  GNU General Public License along
 % with SPM, in spm_LICENCE.man. If not, see <http://www.gnu.org/licenses/>.
 %__________________________________________________________________________
-% Copyright (C) 1991,1994-2003,2005-2010 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1991,1994-2011 Wellcome Trust Centre for Neuroimaging
 
-% $Id: Contents.m 4008 2010-07-21 12:57:02Z guillaume $
+% $Id: Contents.m 4289 2011-04-04 14:47:35Z guillaume $
 
 %==========================================================================
 % PROGRAMMERS NOTE:
diff --git a/config/spm_cfg.m b/config/spm_cfg.m
index 8142f93..09bbf11 100644
--- a/config/spm_cfg.m
+++ b/config/spm_cfg.m
@@ -3,7 +3,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg.m 3881 2010-05-07 21:02:57Z vladimir $
+% $Id: spm_cfg.m 4257 2011-03-18 15:28:29Z vladimir $
 
 %_______________________________________________________________________
 % temporal Temporal
@@ -52,7 +52,7 @@
 source.tag    = 'source';
 source.name   = 'M/EEG Source reconstruction';
 source.help   = {'M/EEG source reconstruction.'};
-source.values = {spm_cfg_eeg_inv_headmodel, spm_cfg_eeg_inv_invert, spm_cfg_eeg_inv_results}; 
+source.values = {spm_cfg_eeg_inv_headmodel, spm_cfg_eeg_inv_invert, spm_cfg_eeg_inv_results, spm_cfg_eeg_inv_extract}; 
 % ---------------------------------------------------------------------
 % meeg Meeg
 % ---------------------------------------------------------------------
diff --git a/config/spm_cfg_checkreg.m b/config/spm_cfg_checkreg.m
index 1fb473b..37cf0e6 100644
--- a/config/spm_cfg_checkreg.m
+++ b/config/spm_cfg_checkreg.m
@@ -1,35 +1,38 @@
 function checkreg = spm_cfg_checkreg
-% SPM Configuration file
-% automatically generated by the MATLABBATCH utility function GENCODE
-%_______________________________________________________________________
+% SPM Configuration file for Check Reg
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_checkreg.m 1295 2008-04-02 14:31:24Z volkmar $
+% $Id: spm_cfg_checkreg.m 4205 2011-02-21 15:39:08Z guillaume $
 
-rev = '$Rev: 1295 $';
-% ---------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % data Images to Display
-% ---------------------------------------------------------------------
+%--------------------------------------------------------------------------
 data         = cfg_files;
 data.tag     = 'data';
 data.name    = 'Images to Display';
 data.help    = {'Images to display.'};
-data.filter = 'image';
+data.filter  = 'image';
 data.ufilter = '.*';
 data.num     = [1 15];
-% ---------------------------------------------------------------------
+
+%--------------------------------------------------------------------------
 % checkreg Check Registration
-% ---------------------------------------------------------------------
-checkreg         = cfg_exbranch;
-checkreg.tag     = 'checkreg';
-checkreg.name    = 'Check Registration';
-checkreg.val     = {data };
-checkreg.help    = {
-                    'Orthogonal views of one or more images are displayed.  Clicking in any image moves the centre of the orthogonal views.  Images are shown in orientations relative to that of the first selected image. The first specified image is shown at the top-left, and the last at the bottom right.  The fastest increment is in the left-to-right direction (the same as you are reading this).'
-                    ''
-                    'If you have put your images in the correct file format, then (possibly after specifying some rigid-body rotations):'
-                    '    The top-left image is coronal with the top (superior) of the head displayed at the top and the left shown on the left. This is as if the subject is viewed from behind.'
-                    '    The bottom-left image is axial with the front (anterior) of the head at the top and the left shown on the left. This is as if the subject is viewed from above.'
-                    '    The top-right image is sagittal with the front (anterior) of the head at the left and the top of the head shown at the top. This is as if the subject is viewed from the left.'
+%--------------------------------------------------------------------------
+checkreg      = cfg_exbranch;
+checkreg.tag  = 'checkreg';
+checkreg.name = 'Check Registration';
+checkreg.val  = {data };
+checkreg.help = {
+                 'Orthogonal views of one or more images are displayed.  Clicking in any image moves the centre of the orthogonal views.  Images are shown in orientations relative to that of the first selected image. The first specified image is shown at the top-left, and the last at the bottom right.  The fastest increment is in the left-to-right direction (the same as you are reading this).'
+                 ''
+                 'If you have put your images in the correct file format, then (possibly after specifying some rigid-body rotations):'
+                 '    The top-left image is coronal with the top (superior) of the head displayed at the top and the left shown on the left. This is as if the subject is viewed from behind.'
+                 '    The bottom-left image is axial with the front (anterior) of the head at the top and the left shown on the left. This is as if the subject is viewed from above.'
+                 '    The top-right image is sagittal with the front (anterior) of the head at the left and the top of the head shown at the top. This is as if the subject is viewed from the left.'
 }';
-checkreg.prog = @spm_check_registration;
+checkreg.prog = @check_reg;
+
+%==========================================================================
+function check_reg(job)
+spm_check_registration(char(job.data));
\ No newline at end of file
diff --git a/config/spm_cfg_coreg.m b/config/spm_cfg_coreg.m
index 892bc2e..792068b 100644
--- a/config/spm_cfg_coreg.m
+++ b/config/spm_cfg_coreg.m
@@ -3,7 +3,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_coreg.m 3596 2009-11-25 19:17:07Z guillaume $
+% $Id: spm_cfg_coreg.m 4269 2011-03-29 16:03:43Z guillaume $
 
 % ---------------------------------------------------------------------
 % ref Reference Image
@@ -99,7 +99,7 @@
 % ---------------------------------------------------------------------
 estimate         = cfg_exbranch;
 estimate.tag     = 'estimate';
-estimate.name    = 'Coreg: Estimate';
+estimate.name    = 'Coregister: Estimate';
 estimate.val     = {ref source other eoptions };
 estimate.help    = {
                     'The registration method used here is based on work by Collignon et al/* \cite{collignon95}*/. The original interpolation method described in this paper has been changed in order to give a smoother cost function.  The images are also smoothed slightly, as is the histogram.  This is all in order to make the cost function as smooth as possible, to give faster convergence and less chance of local minima.'
@@ -209,7 +209,7 @@
 % ---------------------------------------------------------------------
 write         = cfg_exbranch;
 write.tag     = 'write';
-write.name    = 'Coreg: Reslice';
+write.name    = 'Coregister: Reslice';
 write.val     = {refwrite source roptions };
 write.help    = {'Reslice images to match voxel-for-voxel with an image defining some space. The resliced images are named the same as the originals except that they are prefixed by ''r''.'};
 write.prog = @spm_run_coreg_reslice;
@@ -229,7 +229,7 @@
 % ---------------------------------------------------------------------
 estwrite         = cfg_exbranch;
 estwrite.tag     = 'estwrite';
-estwrite.name    = 'Coreg: Estimate & Reslice';
+estwrite.name    = 'Coregister: Estimate & Reslice';
 estwrite.val     = {ref source other eoptions roptions };
 estwrite.help    = {
                     'The registration method used here is based on work by Collignon et al/* \cite{collignon95}*/. The original interpolation method described in this paper has been changed in order to give a smoother cost function.  The images are also smoothed slightly, as is the histogram.  This is all in order to make the cost function as smooth as possible, to give faster convergence and less chance of local minima.'
@@ -245,7 +245,7 @@
 % ---------------------------------------------------------------------
 coreg         = cfg_choice;
 coreg.tag     = 'coreg';
-coreg.name    = 'Coreg';
+coreg.name    = 'Coregister';
 coreg.help    = {
                  'Within-subject registration using a rigid-body model. A rigid-body transformation (in 3D) can be parameterised by three translations and three rotations about the different axes.'
                  ''
diff --git a/config/spm_cfg_defs.m b/config/spm_cfg_defs.m
index a360b5a..de8db61 100644
--- a/config/spm_cfg_defs.m
+++ b/config/spm_cfg_defs.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_cfg_defs.m 3804 2010-03-31 16:16:21Z ged $
+% $Id: spm_cfg_defs.m 4136 2010-12-09 22:22:28Z guillaume $
 
 hsummary = {[...
 'This is a utility for working with deformation fields. ',...
@@ -81,9 +81,9 @@
 'Specify the image file on which to base the dimensions, orientation etc.'};
 
 hid = {[...
-'This option generates an identity transform, but this can be useful for '...
-'changing the dimensions of the resulting deformation (and any images that '...
-'are generated from it).  Dimensions, orientation etc are derived from '...
+'This option generates an identity transform, but this can be useful for ',...
+'changing the dimensions of the resulting deformation (and any images that ',...
+'are generated from it).  Dimensions, orientation etc are derived from ',...
 'an image.']};
 
 def          = files('Deformation Field','def','.*y_.*\.nii$',[1 1]);
diff --git a/config/spm_cfg_dicom.m b/config/spm_cfg_dicom.m
index 6c6f876..5eebbbc 100644
--- a/config/spm_cfg_dicom.m
+++ b/config/spm_cfg_dicom.m
@@ -3,7 +3,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_dicom.m 3697 2010-01-22 17:07:18Z guillaume $
+% $Id: spm_cfg_dicom.m 4012 2010-07-22 12:45:53Z volkmar $
 
 % ---------------------------------------------------------------------
 % data DICOM files
@@ -29,15 +29,17 @@
                 ''
                 '* Output directory: ./<PatientName>: Convert into the output directory, starting with a PatientName subdirectory.'
                 '* No directory hierarchy: Convert all files into the output directory, without sequence/series subdirectories'}';
-root.labels  = {'Output directory: ./<StudyDate-StudyTime>'
-                'Output directory: ./<PatientID>'
-                'Output directory: ./<PatientID>/<StudyDate-StudyTime>'
-                'Output directory: ./<PatientName>'
+root.labels  = {'Output directory: ./<StudyDate-StudyTime>/<ProtocolName>'
+                'Output directory: ./<PatientID>/<ProtocolName>'
+                'Output directory: ./<PatientID>/<StudyDate-StudyTime>/<ProtocolName>'
+                'Output directory: ./<PatientName>/<ProtocolName>'
+                'Output directory: ./<ProtocolName>'
                 'No directory hierarchy'}';
 root.values  = {'date_time'
                 'patid'
                 'patid_date'
                 'patname'
+                'series'
                 'flat'}';
 root.def     = @(val)spm_get_defaults('dicom.root', val{:});
 % ---------------------------------------------------------------------
diff --git a/config/spm_cfg_disp.m b/config/spm_cfg_disp.m
index 06d9cd9..fc709dc 100644
--- a/config/spm_cfg_disp.m
+++ b/config/spm_cfg_disp.m
@@ -1,29 +1,28 @@
 function disp = spm_cfg_disp
-% SPM Configuration file
-% automatically generated by the MATLABBATCH utility function GENCODE
-%_______________________________________________________________________
+% SPM Configuration file for Image Display
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_disp.m 1295 2008-04-02 14:31:24Z volkmar $
+% $Id: spm_cfg_disp.m 4205 2011-02-21 15:39:08Z guillaume $
 
-rev = '$Rev: 1295 $';
-% ---------------------------------------------------------------------
+%-------------------------------------------------------------------------
 % data Image to Display
-% ---------------------------------------------------------------------
+%-------------------------------------------------------------------------
 data         = cfg_files;
 data.tag     = 'data';
 data.name    = 'Image to Display';
 data.help    = {'Image to display.'};
-data.filter = 'image';
+data.filter  = 'image';
 data.ufilter = '.*';
 data.num     = [1 1];
-% ---------------------------------------------------------------------
+
+%--------------------------------------------------------------------------
 % disp Display Image
-% ---------------------------------------------------------------------
+%--------------------------------------------------------------------------
 disp         = cfg_exbranch;
 disp.tag     = 'disp';
 disp.name    = 'Display Image';
-disp.val     = {data };
+disp.val     = {data};
 disp.help    = {
                 'This is an interactive facility that allows orthogonal sections from an image volume to be displayed.  Clicking the cursor on either of the three images moves the point around which the orthogonal sections are viewed.  The co-ordinates of the cursor are shown both in voxel co-ordinates and millimetres within some fixed framework. The intensity at that point in the image (sampled using the current interpolation scheme) is also given. The position of the cross-hairs can also be moved by specifying the co-ordinates in millimetres to which they should be moved.  Clicking on the horizontal bar above these boxes will move the cursor back to the origin  (analogous to setting the cross-hair position (in mm) to [0 0 0]).'
                 ''
@@ -36,9 +35,9 @@
                 '   Datatype   - the computer representation of each voxel.'
                 '   Intensity  - scale-factors and possibly a DC offset.'
                 '   Miscellaneous other information about the image.'
-                '   Vox size   - the distance (in mm) between the centres of                 neighbouring voxels.'
+                '   Vox size   - the distance (in mm) between the centres of neighbouring voxels.'
                 '   Origin     - the voxel at the origin of the co-ordinate system'
-                '   DIr Cos    - Direction cosines.  This is a widely used                representation of the orientation of an image.'
+                '   DIr Cos    - Direction cosines.  This is a widely used representation of the orientation of an image.'
                 ''
                 'There are also a few options for different resampling modes, zooms etc.  You can also flip between voxel space (as would be displayed by Analyze) or world space (the orientation that SPM considers the image to be in).  If you are re-orienting the images, make sure that world space is specified.  Blobs (from activation studies) can be superimposed on the images and the intensity windowing can also be changed.'
                 ''
@@ -49,4 +48,8 @@
                 '    The top-right image is sagittal with the front (anterior) of the head at the left and the top of the head shown at the top. This is as if the subject is viewed from the left.'
                 '/*\begin{figure} \begin{center} \includegraphics[width=150mm]{images/disp1} \end{center} \caption{The Display routine. \label{disp1}}\end{figure} */'
 }';
-disp.prog = @spm_image;
+disp.prog = @disp_image;
+
+%==========================================================================
+function disp_image(job)
+spm_image('Init', job.data{1});
diff --git a/config/spm_cfg_eeg_filter.m b/config/spm_cfg_eeg_filter.m
index 1069a64..e32a309 100644
--- a/config/spm_cfg_eeg_filter.m
+++ b/config/spm_cfg_eeg_filter.m
@@ -4,9 +4,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: spm_cfg_eeg_filter.m 3881 2010-05-07 21:02:57Z vladimir $
+% $Id: spm_cfg_eeg_filter.m 4212 2011-02-23 17:50:55Z vladimir $
 
-rev = '$Rev: 3881 $';
+rev = '$Rev: 4212 $';
 D = cfg_files;
 D.tag = 'D';
 D.name = 'File Name';
@@ -48,10 +48,10 @@
 order = cfg_entry;
 order.tag = 'order';
 order.name = 'Filter order';
-order.val = {[]};
+order.val = {5};
 order.strtype = 'n';
-order.num = [0 1];
-order.help = {'Enter the filter order (leave empty for default)'};
+order.num = [1 1];
+order.help = {'Enter the filter order'};
 
 flt = cfg_branch;
 flt.tag = 'filter';
diff --git a/config/spm_cfg_eeg_inv_extract.m b/config/spm_cfg_eeg_inv_extract.m
new file mode 100644
index 0000000..cd95298
--- /dev/null
+++ b/config/spm_cfg_eeg_inv_extract.m
@@ -0,0 +1,121 @@
+function extract = spm_cfg_eeg_inv_extract
+% configuration file for extracting source data from imaging source
+% reconstruction
+%_______________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+
+% Vladimir Litvak
+% $Id: spm_cfg_eeg_inv_extract.m 4257 2011-03-18 15:28:29Z vladimir $
+
+D = cfg_files;
+D.tag = 'D';
+D.name = 'M/EEG datasets';
+D.filter = 'mat';
+D.num = [1 Inf];
+D.help = {'Select the M/EEG mat files.'};
+
+val = cfg_entry;
+val.tag = 'val';
+val.name = 'Inversion index';
+val.strtype = 'n';
+val.help = {'Index of the cell in D.inv where the inversion results are be stored.'};
+val.val = {1};
+
+xyz = cfg_entry;
+xyz.tag = 'xyz';
+xyz.name = 'Source location';
+xyz.strtype = 'r';
+xyz.num = [1 3];
+xyz.help = {'Source location (in MNI coordinates)'};
+
+label = cfg_entry;
+label.tag = 'label';
+label.name = 'Source label';
+label.strtype = 's';
+label.num = [1 Inf];
+label.help = {'Label for the source channel in the output file'};
+
+source = cfg_branch;
+source.tag = 'source';
+source.name = 'Source';
+source.val = {label, xyz};
+
+sources = cfg_repeat;
+sources.tag = 'sources';
+sources.name = 'Sources to extract';
+sources.values  = {source};
+sources.num     = [1 Inf];
+sources.help = {'Specify sources to extract data from'};
+
+rad = cfg_entry;
+rad.tag = 'rad';
+rad.name = 'VOI radius';
+rad.strtype = 'r';
+rad.num = [1 1];
+rad.val = {5};
+woi.help = {'Radius around each location to extract an eigenvariate from (mm).'};
+
+type = cfg_menu;
+type.tag = 'type';
+type.name = 'What to extract';
+type.help = {'What to extract: evoked activity or single trials.'};
+type.labels = {'Evoked', 'Single trials'};
+type.values = {'evoked', 'trials'};
+type.val = {'trials'};
+
+fname = cfg_entry;
+fname.tag = 'fname';
+fname.name = 'Output dataset name';
+fname.strtype = 's';
+fname.num = [0 Inf];
+fname.val = {''};
+fname.help = {'Output file name (empty for default)'};
+
+extract = cfg_exbranch;
+extract.tag = 'extract';
+extract.name = 'M/EEG source extraction';
+extract.val = {D, val, sources, rad, type, fname};
+extract.help = {'Extract source data from the results of inverse source reconstruction'};
+extract.prog = @run_extract;
+extract.vout = @vout_extract;
+extract.modality = {'EEG'};
+
+function  out = run_extract(job)
+
+source       = [];
+source.XYZ   = cat(1, job.source.xyz);
+source.label = {job.source.label};
+source.rad   = job.rad;
+source.type  = job.type;
+
+if ~isempty(job.fname)
+    source.fname = job.fname;
+end
+
+out.D = {};
+
+for i = 1:numel(job.D)
+    D = spm_eeg_load(job.D{i});
+    
+    D.val = job.val;      
+            
+    if ~isfield(D.inv{D.val}, 'inverse') || ~isfield(D.inv{D.val}.inverse, 'J')
+        error(sprintf('Imaging source reconstruction is missing for subject %d', i));
+    end       
+    
+    D.inv{D.val}.source = source;
+    
+    Ds = spm_eeg_inv_extract(D);
+    
+    out.D{i, 1} = fullfile(Ds.path, Ds.fname);      
+end
+
+function dep = vout_extract(job)
+% Output is always in field "D", no matter how job is structured
+% Output is always in field "D", no matter how job is structured
+dep = cfg_dep;
+dep.sname = 'M/EEG dataset(s) extracted source data';
+% reference field "D" from output
+dep.src_output = substruct('.','D');
+% this can be entered into any evaluated input
+dep.tgt_spec   = cfg_findspec({{'filter','mat'}});
diff --git a/config/spm_cfg_eeg_inv_headmodel.m b/config/spm_cfg_eeg_inv_headmodel.m
index fb6b2ac..217e3df 100644
--- a/config/spm_cfg_eeg_inv_headmodel.m
+++ b/config/spm_cfg_eeg_inv_headmodel.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_cfg_eeg_inv_headmodel.m 3760 2010-03-08 17:00:30Z vladimir $
+% $Id: spm_cfg_eeg_inv_headmodel.m 4118 2010-11-10 14:48:16Z vladimir $
 
 D = cfg_files;
 D.tag = 'D';
@@ -211,7 +211,7 @@
         meegfid.fid.label = meegfid.fid.label(selection);
         
         mrifid = [];
-        mrifid.pnt = [];
+        mrifid.pnt = D.inv{val}.mesh.fid.pnt;
         mrifid.fid.pnt = [];
         mrifid.fid.label = {job.coregistration.coregspecify.fiducial.fidname}';
         
diff --git a/config/spm_cfg_eeg_montage.m b/config/spm_cfg_eeg_montage.m
index 94af3c3..9c56d9c 100644
--- a/config/spm_cfg_eeg_montage.m
+++ b/config/spm_cfg_eeg_montage.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: spm_cfg_eeg_montage.m 3881 2010-05-07 21:02:57Z vladimir $
+% $Id: spm_cfg_eeg_montage.m 4212 2011-02-23 17:50:55Z vladimir $
 
 D = cfg_files;
 D.tag = 'D';
@@ -20,21 +20,12 @@
 montage.num = [1 1];
 montage.help = {'Select a montage file.'};
 
-yes = cfg_const;
-yes.tag = 'yes';
-yes.name = 'Keep the other channels';
-yes.val = {1};
-
-no = cfg_const;
-no.tag = 'no';
-no.name = 'Remove the other channels';
-no.val = {1};
-
-keepothers = cfg_choice;
+keepothers = cfg_menu;
 keepothers.tag = 'keepothers';
 keepothers.name = 'Keep other channels';
-keepothers.values = {yes,no};
-keepothers.val = {yes};
+keepothers.labels = {'Yes', 'No'};
+keepothers.values = {'yes', 'no'};
+keepothers.val = {'no'};
 keepothers.help = {'Specify whether you want to keep channels that are not contributing to the new channels'};
 
 S = cfg_exbranch;
@@ -50,12 +41,7 @@
 % construct the S struct
 S.D = job.D{1};
 S.montage = job.montage{1};
-if isfield(job.keepothers, 'yes')
-    S.keepothers = 'yes';
-else
-    S.keepothers = 'no';
-end
-
+S.keepothers = job.keepothers;
 out.D = spm_eeg_montage(S);
 out.Dfname = {fullfile(out.D.path, out.D.fname)};
 
diff --git a/config/spm_cfg_eeg_tf.m b/config/spm_cfg_eeg_tf.m
index e12f75d..8d8a7f6 100644
--- a/config/spm_cfg_eeg_tf.m
+++ b/config/spm_cfg_eeg_tf.m
@@ -4,9 +4,9 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_cfg_eeg_tf.m 3818 2010-04-13 14:36:31Z vladimir $
+% $Id: spm_cfg_eeg_tf.m 4257 2011-03-18 15:28:29Z vladimir $
 
-rev = '$Rev: 3818 $';
+rev = '$Rev: 4257 $';
 
 D = cfg_files;
 D.tag = 'D';
@@ -29,7 +29,7 @@
 frequencies.strtype = 'r';
 frequencies.num = [0 Inf];
 frequencies.val = {[]};
-frequencies.help = {'Frequencies of interest (as a vector), if empty 0-48 with optimal frequency bins ~1 Hz or above resolution'};
+frequencies.help = {'Frequencies of interest (as a vector), if empty 1-48 with optimal frequency bins ~1 Hz or above resolution'};
 
 specest_funs = dir(fullfile(spm('dir'), 'spm_eeg_specest_*.m'));
 specest_funs = {specest_funs(:).name};
diff --git a/config/spm_cfg_eeg_tf_rescale.m b/config/spm_cfg_eeg_tf_rescale.m
index a494105..647dfc6 100644
--- a/config/spm_cfg_eeg_tf_rescale.m
+++ b/config/spm_cfg_eeg_tf_rescale.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_cfg_eeg_tf_rescale.m 3931 2010-06-16 15:28:54Z vladimir $
+% $Id: spm_cfg_eeg_tf_rescale.m 4287 2011-04-04 13:55:54Z vladimir $
 
 %--------------------------------------------------------------------------
 % D
@@ -73,6 +73,15 @@
 method_rel.val   = {baseline};
 method_rel.help  = {'Relative.'};
 
+%--------------------------------------------------------------------------
+% method_zscore
+%--------------------------------------------------------------------------
+method_zscore       = cfg_branch;
+method_zscore.tag   = 'Zscore';
+method_zscore.name  = 'Zscore';
+method_zscore.val   = {baseline};
+method_zscore.help  = {'Z score'};
+
 %--------------------------------------------------------------------------
 % method_log
 %--------------------------------------------------------------------------
@@ -99,7 +108,7 @@
 method.name   = 'Rescale method';
 method.val    = {method_logr};
 method.help   = {'Select the rescale method.'};
-method.values = {method_logr method_diff method_rel method_log method_sqrt};
+method.values = {method_logr method_diff method_rel method_zscore method_log method_sqrt};
 
 %--------------------------------------------------------------------------
 % S
@@ -126,7 +135,7 @@
 S.tf.method    = fieldnames(job.method);
 S.tf.method    = S.tf.method{1};
 switch lower(S.tf.method)
-    case {'logr','diff', 'rel'}
+    case {'logr','diff', 'rel', 'zscore'}
         S.tf.Sbaseline = 1e-3*job.method.(S.tf.method).baseline.Sbaseline;
         if ~(isempty(job.method.(S.tf.method).baseline.Db) || isequal(job.method.(S.tf.method).baseline.Db, {''}))
             S.tf.Db = job.method.(S.tf.method).baseline.Db{1};
diff --git a/config/spm_cfg_mfx.m b/config/spm_cfg_mfx.m
new file mode 100644
index 0000000..f727bc6
--- /dev/null
+++ b/config/spm_cfg_mfx.m
@@ -0,0 +1,176 @@
+function mfx = spm_cfg_mfx
+% SPM Configuration file for MFX
+%______________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_cfg_mfx.m 4023 2010-07-28 18:41:36Z guillaume $
+
+% ---------------------------------------------------------------------
+% dir Directory
+% ---------------------------------------------------------------------
+dir         = cfg_files;
+dir.tag     = 'dir';
+dir.name    = 'Directory';
+dir.help    = {'Select a directory where the SPM.mat file containing the specified design matrix will be written.'};
+dir.filter = 'dir';
+dir.ufilter = '.*';
+dir.num     = [1 1];
+
+% ---------------------------------------------------------------------
+% spmmat Select SPM.mat
+% ---------------------------------------------------------------------
+spmmat         = cfg_files;
+spmmat.tag     = 'spmmat';
+spmmat.name    = 'Select SPM.mat files';
+spmmat.help    = {...
+    'Select the SPM.mat files that contains first-level designs.'
+    'They must have the same number of parameters for each session.'
+    ['These are assumed to represent session-specific realisations of ' ...
+    '2nd-level effects.']};
+spmmat.filter  = 'mat';
+spmmat.ufilter = '^SPM\.mat$';
+spmmat.num     = [1 Inf];
+
+% ---------------------------------------------------------------------
+% ffx Create first-level design
+% ---------------------------------------------------------------------
+ffx       = cfg_exbranch;
+ffx.tag   = 'ffx';
+ffx.name  = 'Create repeated-measure design';
+ffx.val   = {dir spmmat};
+ffx.help  = {'Create repeated-measure multi-session first-level design'};
+ffx.prog  = @spm_local_ffx;
+ffx.vout  = @vout_ffx;
+
+% ---------------------------------------------------------------------
+% spmmat Select SPM.mat
+% ---------------------------------------------------------------------
+spmmat         = cfg_files;
+spmmat.tag     = 'spmmat';
+spmmat.name    = 'Select SPM.mat';
+spmmat.help    = {...
+    'Design and estimation structure after a 1st-level analysis'};
+spmmat.filter  = 'mat';
+spmmat.ufilter = '^SPM\.mat$';
+spmmat.num     = [1 1];
+
+% ---------------------------------------------------------------------
+% spec MFX Specification
+% ---------------------------------------------------------------------
+spec       = cfg_exbranch;
+spec.tag   = 'spec';
+spec.name  = 'MFX Specification';
+spec.val   = {spmmat};
+spec.help  = {'MFX Specification'};
+spec.prog  = @spm_local_mfx;
+spec.vout  = @vout_mfx;
+
+% ---------------------------------------------------------------------
+% mfx Mixed-effects (MFX) analysis
+% ---------------------------------------------------------------------
+mfx         = cfg_choice;
+mfx.tag     = 'mfx';
+mfx.name    = 'Mixed-effects (MFX) analysis';
+mfx.help    = {'Mixed-effects (MFX) analysis'};
+mfx.values  = {ffx spec};
+
+% =====================================================================
+function out = spm_local_ffx(job)
+spmmat = job.spmmat;
+SPMS = cell(size(spmmat));
+for i=1:numel(spmmat)
+    load(spmmat{i},'SPM');
+    SPMS{i} = SPM;
+end
+
+matlabbatch{1}.spm.stats.fmri_spec.dir            = cellstr(job.dir);
+matlabbatch{1}.spm.stats.fmri_spec.timing.units   = SPMS{1}.xBF.UNITS;
+matlabbatch{1}.spm.stats.fmri_spec.timing.RT      = SPMS{1}.xY.RT;
+matlabbatch{1}.spm.stats.fmri_spec.timing.fmri_t  = SPMS{1}.xBF.T;
+matlabbatch{1}.spm.stats.fmri_spec.timing.fmri_t0 = SPMS{1}.xBF.T0;
+switch SPMS{1}.xBF.name
+    case 'hrf'
+        matlabbatch{1}.spm.stats.fmri_spec.hrf.derivs = [0 0];
+    case 'hrf (with time derivative)'
+        matlabbatch{1}.spm.stats.fmri_spec.hrf.derivs = [1 0];
+    case 'hrf (with time and dispersion derivatives)'
+        matlabbatch{1}.spm.stats.fmri_spec.hrf.derivs = [1 1];
+    case 'Fourier set'
+        matlabbatch{1}.spm.stats.fmri_spec.bases.fourier.length = SPMS{1}.xBF.length;
+        matlabbatch{1}.spm.stats.fmri_spec.bases.fourier.order  = SPMS{1}.xBF.order;
+    case 'Fourier set (Hanning)'
+        matlabbatch{1}.spm.stats.fmri_spec.bases.fourier_han.length = SPMS{1}.xBF.length;
+        matlabbatch{1}.spm.stats.fmri_spec.bases.fourier_han.order  = SPMS{1}.xBF.order;
+    case 'Gamma functions'
+        matlabbatch{1}.spm.stats.fmri_spec.bases.gamma.length = SPMS{1}.xBF.length;
+        matlabbatch{1}.spm.stats.fmri_spec.bases.gamma.order  = SPMS{1}.xBF.order;
+    case 'Finite Impulse Response'
+        matlabbatch{1}.spm.stats.fmri_spec.bases.fir.length = SPMS{1}.xBF.length;
+        matlabbatch{1}.spm.stats.fmri_spec.bases.fir.order  = SPMS{1}.xBF.order;
+end
+matlabbatch{1}.spm.stats.fmri_spec.volt   = SPMS{1}.xBF.Volterra;
+matlabbatch{1}.spm.stats.fmri_spec.global = SPMS{1}.xGX.iGXcalc;
+if ~isempty(SPMS{1}.xM.VM)
+    matlabbatch{1}.spm.stats.fmri_spec.mask = cellstr(SPMS{1}.xM.VM.fname); % can be intersection
+end
+if strncmp('AR',SPMS{1}.xVi.form,2)
+    matlabbatch{1}.spm.stats.fmri_spec.cvi  = 'AR(1)';
+else
+    matlabbatch{1}.spm.stats.fmri_spec.cvi  = 'none';
+end
+
+k = 1;
+for i=1:numel(SPMS)
+    n = cumsum([1 SPMS{i}.nscan]);
+    for j=1:numel(SPMS{i}.Sess)
+        matlabbatch{1}.spm.stats.fmri_spec.sess(k).scans = cellstr(SPMS{i}.xY.P(n(j):n(j+1)-1,:));
+        for l=1:numel(SPMS{i}.Sess(j).U)
+            matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).name = SPMS{i}.Sess(j).U(l).name{1};
+            matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).onset = SPMS{i}.Sess(j).U(l).ons;
+            matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).duration = SPMS{i}.Sess(j).U(l).dur;
+            o = 1;
+            for m=1:numel(SPMS{i}.Sess(j).U(l).P)
+                switch SPMS{i}.Sess(j).U(l).P(m).name
+                    case 'time'
+                        matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).tmod = SPMS{i}.Sess(j).U(l).P(m).h;
+                    case 'none'
+                    otherwise
+                        matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).pmod(o).name  = SPMS{i}.Sess(j).U(l).P(m).name;
+                        matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).pmod(o).param = SPMS{i}.Sess(j).U(l).P(m).P;
+                        matlabbatch{1}.spm.stats.fmri_spec.sess(k).cond(l).pmod(o).poly  = SPMS{i}.Sess(j).U(l).P(m).h;
+                        o = o + 1;
+                end
+            end
+        end
+        for l=1:numel(SPMS{i}.Sess(j).C.name)
+            matlabbatch{1}.spm.stats.fmri_spec.sess(k).regress(l).name = SPMS{i}.Sess(j).C.name{l};
+            matlabbatch{1}.spm.stats.fmri_spec.sess(k).regress(l).val  = SPMS{i}.Sess(j).C.C(:,l);
+        end
+        matlabbatch{1}.spm.stats.fmri_spec.sess(k).hpf = SPMS{i}.xX.K(j).HParam;
+        k = k + 1;
+    end
+end
+spm_jobman('run',matlabbatch);
+
+out.spmmat{1} = fullfile(job.dir{1},'SPM.mat');
+
+% =====================================================================
+function dep = vout_ffx(job)
+dep = cfg_dep;
+dep.sname      = 'SPM.mat File';
+dep.src_output = substruct('.','spmmat');
+dep.tgt_spec   = cfg_findspec({{'filter','mat','strtype','e'}});
+
+% =====================================================================
+function out = spm_local_mfx(job)
+load(job.spmmat{1},'SPM');
+spm_mfx(SPM);
+out.spmmat{1} = fullfile(fileparts(job.spmmat{1}),'mfx','SPM.mat');
+
+% =====================================================================
+function dep = vout_mfx(job)
+dep = cfg_dep;
+dep.sname      = 'SPM.mat File';
+dep.src_output = substruct('.','spmmat');
+dep.tgt_spec   = cfg_findspec({{'filter','mat','strtype','e'}});
diff --git a/config/spm_cfg_ppi.m b/config/spm_cfg_ppi.m
index 5c0927e..68d9908 100644
--- a/config/spm_cfg_ppi.m
+++ b/config/spm_cfg_ppi.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_cfg_ppi.m 3553 2009-11-11 12:34:37Z guillaume $
+% $Id: spm_cfg_ppi.m 4136 2010-12-09 22:22:28Z guillaume $
 
 % ---------------------------------------------------------------------
 % spmmat Select SPM.mat
@@ -45,11 +45,11 @@
 con         = cfg_entry;
 con.tag     = 'u';
 con.name    = ' Input variables and contrast weights';
-con.help    = {['Matrix of input variables and contrast weights.'... 
-'This is an [n x 3] matrix. The first column indexes SPM.Sess.U(i). ' ...
-'The second column indexes the name of the input or cause, see ' ...
-'SPM.Sess.U(i).name{j}. The third column is the contrast weight. ' ...
-' Unless there are parametric effects the second column will generally ' ...
+con.help    = {['Matrix of input variables and contrast weights.',... 
+'This is an [n x 3] matrix. The first column indexes SPM.Sess.U(i). ',...
+'The second column indexes the name of the input or cause, see ',...
+'SPM.Sess.U(i).name{j}. The third column is the contrast weight. ',...
+' Unless there are parametric effects the second column will generally ',...
 'be a 1.']};
 con.strtype = 'e';
 con.num     = [Inf 3];
@@ -96,7 +96,7 @@
 name         = cfg_entry;
 name.tag     = 'name';
 name.name    = 'Name of PPI';
-name.help    = {['Name of the PPI mat file that will be saved in the ' ...
+name.help    = {['Name of the PPI mat file that will be saved in the ',...
 'same directory than the SPM.mat file. A ''PPI_'' prefix will be added.']};
 name.strtype = 's';
 name.num     = [1 Inf];
diff --git a/config/spm_cfg_print.m b/config/spm_cfg_print.m
index 36c1aef..6c68fe2 100644
--- a/config/spm_cfg_print.m
+++ b/config/spm_cfg_print.m
@@ -4,7 +4,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_print.m 1775 2008-06-02 09:18:18Z volkmar $
+% $Id: spm_cfg_print.m 4046 2010-08-26 15:12:57Z volkmar $
  
 % ---------------------------------------------------------------------
 % print Printing
@@ -113,19 +113,41 @@
 opts.values{18}.ext = '.tif';
 opts.def = @(val)spm_get_defaults('ui.print', val{:});
 
-fname = cfg_entry;
-fname.tag  = 'fname';
-fname.name = 'Print Filename';
+fname         = cfg_entry;
+fname.tag     = 'fname';
+fname.name    = 'Print Filename';
 fname.strtype = 's';
-fname.val  = {}; % explicitly unset this val
-fname.help = {['Filename to print to. If this is set as a default - even ' ...
-               'if it is set to an empty string - spm_print(''fname'') ' ...
-               'will not print to file ''fname'', but to the standard SPM ' ...
-               'print file.']};
+fname.val     = {}; % explicitly unset this val
+fname.help    = {['Filename to print to. If this is set as a default - even ' ...
+                  'if it is set to an empty string - spm_print(''fname'') ' ...
+                  'will not print to file ''fname'', but to the standard SPM ' ...
+                  'print file.']};
 
-cfg_print = cfg_exbranch;
-cfg_print.tag = 'print';
+figname         = cfg_entry;
+figname.tag     = 'figname';
+figname.name    = 'Figure Name';
+figname.strtype = 's';
+figname.val     = {'Graphics'};
+figname.help    = {['Figure to print. The figure will be determined by ' ...
+                    'calling spm_figure(''FindWin'', ...).']};
+
+fighandle         = cfg_entry;
+fighandle.tag     = 'fighandle';
+fighandle.name    = 'Figure Handle';
+fighandle.strtype = 'r';
+fighandle.val     = {}; % explicitly unset this val 
+fighandle.help    = {['Figure to print. The value entered here must be a ' ...
+                    'figure handle. If it is a non-finite value (Inf/NaN), ' ...
+                    'the SPM Graphics/Help window is printed.']};
+
+fig        = cfg_choice;
+fig.tag    = 'fig';
+fig.name   = 'Figure to print';
+fig.values = {figname, fighandle};
+
+cfg_print      = cfg_exbranch;
+cfg_print.tag  = 'print';
 cfg_print.name = 'Print';
-cfg_print.val = {fname, opts};
+cfg_print.val  = {fname, fig, opts};
 cfg_print.prog = @spm_print;
 cfg_print.help = {'Print the Graphics/Help window.'};
diff --git a/config/spm_cfg_realign.m b/config/spm_cfg_realign.m
index 3ec1905..c7594a1 100644
--- a/config/spm_cfg_realign.m
+++ b/config/spm_cfg_realign.m
@@ -4,9 +4,9 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_realign.m 3804 2010-03-31 16:16:21Z ged $
+% $Id: spm_cfg_realign.m 4152 2011-01-11 14:13:35Z volkmar $
 
-rev = '$Rev: 3804 $';
+rev = '$Rev: 4152 $';
 % ---------------------------------------------------------------------
 % data Session
 % ---------------------------------------------------------------------
@@ -130,11 +130,11 @@
 weight         = cfg_files;
 weight.tag     = 'weight';
 weight.name    = 'Weighting';
+weight.val     = {''};
 weight.help    = {'The option of providing a weighting image to weight each voxel of the reference image differently when estimating the realignment parameters.  The weights are proportional to the inverses of the standard deviations. This would be used, for example, when there is a lot of extra-brain motion - e.g., during speech, or when there are serious artifacts in a particular region of the images.'};
 weight.filter  = 'image';
 weight.ufilter = '.*';
 weight.num     = [0 1];
-weight.def     = @(val)spm_get_defaults('realign.estimate.weight', val{:});
 % ---------------------------------------------------------------------
 % eoptions Estimation Options
 % ---------------------------------------------------------------------
diff --git a/config/spm_cfg_realignunwarp.m b/config/spm_cfg_realignunwarp.m
index 6e2d1df..3d7c3ba 100644
--- a/config/spm_cfg_realignunwarp.m
+++ b/config/spm_cfg_realignunwarp.m
@@ -4,9 +4,9 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_realignunwarp.m 3804 2010-03-31 16:16:21Z ged $
+% $Id: spm_cfg_realignunwarp.m 4152 2011-01-11 14:13:35Z volkmar $
 
-rev = '$Rev: 3804 $';
+rev = '$Rev: 4152 $';
 % ---------------------------------------------------------------------
 % scans Images
 % ---------------------------------------------------------------------
@@ -159,11 +159,11 @@
 weight         = cfg_files;
 weight.tag     = 'weight';
 weight.name    = 'Weighting';
+weight.val     = {''};
 weight.help    = {'The option of providing a weighting image to weight each voxel of the reference image differently when estimating the realignment parameters.  The weights are proportional to the inverses of the standard deviations. For example, when there is a lot of extra-brain motion - e.g., during speech, or when there are serious artifacts in a particular region of the images.'};
 weight.filter  = 'image';
 weight.ufilter = '.*';
 weight.num     = [0 1];
-weight.def     = @(val)spm_get_defaults('realign.estimate.weight', val{:});
 % ---------------------------------------------------------------------
 % eoptions Estimation Options
 % ---------------------------------------------------------------------
diff --git a/config/spm_cfg_st.m b/config/spm_cfg_st.m
index 3b57c0e..81c27b0 100644
--- a/config/spm_cfg_st.m
+++ b/config/spm_cfg_st.m
@@ -3,7 +3,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_cfg_st.m 4005 2010-07-21 10:54:55Z guillaume $
+% $Id: spm_cfg_st.m 4269 2011-03-29 16:03:43Z guillaume $
 
 % ---------------------------------------------------------------------
 % scans Session
@@ -11,7 +11,7 @@
 scans         = cfg_files;
 scans.tag     = 'scans';
 scans.name    = 'Session';
-scans.help    = {'Select images to acquisition correct.'};
+scans.help    = {'Select images to slice-time correct.'};
 scans.filter = 'image';
 scans.ufilter = '.*';
 scans.num     = [2 Inf];
@@ -30,7 +30,7 @@
 nslices         = cfg_entry;
 nslices.tag     = 'nslices';
 nslices.name    = 'Number of Slices';
-nslices.help    = {'Enter the number of slices'};
+nslices.help    = {'Enter the number of slices.'};
 nslices.strtype = 'n';
 nslices.num     = [1 1];
 % ---------------------------------------------------------------------
@@ -39,7 +39,7 @@
 tr         = cfg_entry;
 tr.tag     = 'tr';
 tr.name    = 'TR';
-tr.help    = {'Enter the TR in seconds'};
+tr.help    = {'Enter the TR (in seconds).'};
 tr.strtype = 'r';
 tr.num     = [1 1];
 % ---------------------------------------------------------------------
@@ -48,7 +48,7 @@
 ta         = cfg_entry;
 ta.tag     = 'ta';
 ta.name    = 'TA';
-ta.help    = {'The TA (in seconds) must be entered by the user. It is usually calculated as TR-(TR/nslices). You can simply enter this equation with the variables replaced by appropriate numbers.'};
+ta.help    = {'Enter the TA (in seconds). It is usually calculated as TR-(TR/nslices). You can simply enter this equation with the variables replaced by appropriate numbers.'};
 ta.strtype = 'e';
 ta.num     = [1 1];
 % ---------------------------------------------------------------------
@@ -90,7 +90,7 @@
 prefix         = cfg_entry;
 prefix.tag     = 'prefix';
 prefix.name    = 'Filename Prefix';
-prefix.help    = {'Specify the string to be prepended to the filenames of the smoothed image file(s). Default prefix is ''a''.'};
+prefix.help    = {'Specify the string to be prepended to the filenames of the slice-time corrected image file(s). Default prefix is ''a''.'};
 prefix.strtype = 's';
 prefix.num     = [1 Inf];
 prefix.def     = @(val)spm_get_defaults('slicetiming.prefix', val{:});
@@ -114,9 +114,11 @@
               ''
               'This correction assumes that the data are band-limited (i.e. there is no meaningful information present in the data at a frequency higher than that of the Nyquist). This assumption is support by the study of Josephs et al (1997, NeuroImage) that obtained event-related data at an effective TR of 166 msecs. No physio-logical signal change was present at frequencies higher than our typical Nyquist (0.25 HZ).'
               ''
-              'Written by Darren Gitelman at Northwestern U., 1998.  Based (in large part) on ACQCORRECT.PRO from Geoff Aguirre and Eric Zarahn at U. Penn.'
+              'When using the slice timing correction it is very important that you input the correct slice order, and if there is any uncertainty then users are encouraged to work with their physicist to determine the actual slice acquisition order.'
+              ''
+              'One can also consider augmenting the model by including the temporal derivative in the informed basis set instead of slice timing, which can account for +/- 1 second of changes in timing.'
               ''
-              'Note that the authors of SPM do not generally suggest that this correction should be used, but the option is still retained for the few people who like to use it.'
+              'Written by Darren Gitelman at Northwestern U., 1998.  Based (in large part) on ACQCORRECT.PRO from Geoff Aguirre and Eric Zarahn at U. Penn.'
               }';
 st.prog = @spm_run_st;
 st.vout = @vout;
diff --git a/config/spm_cfg_voi.m b/config/spm_cfg_voi.m
index 62863dc..e468062 100644
--- a/config/spm_cfg_voi.m
+++ b/config/spm_cfg_voi.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_cfg_voi.m 3919 2010-06-04 16:45:16Z guillaume $
+% $Id: spm_cfg_voi.m 4136 2010-12-09 22:22:28Z guillaume $
 
 % -------------------------------------------------------------------------
 % spmmat Select SPM.mat
@@ -315,7 +315,7 @@
 expression         = cfg_entry;
 expression.tag     = 'expression';
 expression.name    = 'Expression';
-expression.help    = {['Expression to be evaluated, using i1, i2, ...' ...
+expression.help    = {['Expression to be evaluated, using i1, i2, ...',...
     'and logical operators (&, |, ~)']};
 expression.strtype = 's';
 expression.num     = [2 Inf];
@@ -358,8 +358,8 @@
 name.tag     = 'name';
 name.name    = 'Name of VOI';
 name.help    = {['Name of the VOI mat file that will be saved in the ' ...
-'same directory than the SPM.mat file. A ''VOI_'' prefix will be added. '...
-''...
+'same directory than the SPM.mat file. A ''VOI_'' prefix will be added. ' ...
+'' ...
 'A binary NIfTI image of the VOI will also be saved.']};
 name.strtype = 's';
 name.num     = [1 Inf];
diff --git a/config/spm_make_standalone.m b/config/spm_make_standalone.m
new file mode 100644
index 0000000..a1d5f6a
--- /dev/null
+++ b/config/spm_make_standalone.m
@@ -0,0 +1,104 @@
+function spm_make_standalone(outdir)
+% Compile SPM as a standalone executable using the MATLAB compiler
+%   http://www.mathworks.com/products/compiler/
+%
+% This will generate a standalone program, which can be run
+% outside MATLAB, and therefore does not use up a MATLAB licence.
+%
+% On Windows:
+%   spm8_wxx.exe <modality>
+%   spm8_wxx.exe run <batch.m(at)>
+%
+% On Linux/Mac:
+%   ./run_spm8.sh <MCRroot> <modality>
+%   ./run_spm8.sh <MCRroot> run <batch.m(at)>
+%
+% The first command starts SPM in interactive mode with GUI. The second
+% executes a batch file or starts the Batch Editor if empty.
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_make_standalone.m 4017 2010-07-26 17:14:31Z guillaume $
+
+%-Care of startup.m
+%--------------------------------------------------------------------------
+% see http://www.mathworks.com/support/solutions/data/1-QXFMQ.html?1-QXFMQ
+if exist('startup','file')
+    warning('A startup.m has been detected in %s.\n',...
+        fileparts(which('startup')));
+end
+
+if ~nargin, outdir = fullfile(spm('dir'),'..','spm_exec'); mkdir(outdir); end
+
+%==========================================================================
+%-Static listing of SPM toolboxes
+%==========================================================================
+fid = fopen(fullfile(spm('dir'),'config','spm_cfg_static_tools.m'),'wt');
+fprintf(fid,'function values = spm_cfg_static_tools\n');
+fprintf(fid,...
+    '%% Static listing of all batch configuration files in the SPM toolbox folder\n');
+% create code to insert toolbox config
+%-Toolbox autodetection
+%-Get the list of toolbox directories
+tbxdir = fullfile(spm('Dir'),'toolbox');
+d  = dir(tbxdir); d = {d([d.isdir]).name};
+dd = regexp(d,'^\.');
+%(Beware, regexp returns an array if input cell array is of dim 0 or 1)
+if ~iscell(dd), dd = {dd}; end
+d  = {'' d{cellfun('isempty',dd)}};
+ft = {};
+ftc = {};
+%-Look for '*_cfg_*.m' or '*_config_*.m' files in these directories
+for i=1:length(d)
+    d2 = fullfile(tbxdir,d{i});
+    di = dir(d2); di = {di(~[di.isdir]).name};
+    f2 = regexp(di,'.*_cfg_.*\.m$');
+    if ~iscell(f2), f2 = {f2}; end
+    fi = {di{~cellfun('isempty',f2)}};
+    if ~isempty(fi)
+        ft = [ft(:); fi(:)];
+    else
+        % try *_config_*.m files, if toolbox does not have '*_cfg_*.m' files
+        f2 = regexp(di,'.*_config_.*\.m$');
+        if ~iscell(f2), f2 = {f2}; end
+        fi = {di{~cellfun('isempty',f2)}};
+        ftc = [ftc(:); fi(:)];
+    end;
+end
+if ~isempty(ft)||~isempty(ftc)
+    if isempty(ft)
+        ftstr = '';
+    else
+        ft = cellfun(@(cft)strtok(cft,'.'),ft,'UniformOutput',false);
+        ftstr  = sprintf('%s ', ft{:});
+    end
+    if isempty(ftc)
+        ftcstr = '';
+    else
+        ftc = cellfun(@(cftc)strtok(cftc,'.'),ftc,'UniformOutput',false);
+        ftcstr = sprintf('cfg_struct2cfg(%s) ', ftc{:});
+    end
+    fprintf(fid,'values = {%s %s};\n', ftstr, ftcstr);
+end
+fclose(fid);
+
+%==========================================================================
+%-Static listing of batch application initialisation files
+%==========================================================================
+cfg_util('dumpcfg');
+
+%==========================================================================
+%-Duplicate Contents.m in Contents.txt for use in spm('Ver')
+%==========================================================================
+sts = copyfile(fullfile(spm('Dir'),'Contents.m'),...
+               fullfile(spm('Dir'),'Contents.txt'));
+if ~sts, warning('Copy of Contents.m failed.'); end
+
+%==========================================================================
+%-Compilation
+%==========================================================================
+mcc('-m', '-C', '-v', '-o',lower(spm('Ver')), 'spm_standalone.m',...
+    '-d',outdir,...
+    '-N','-p',fullfile(matlabroot,'toolbox','signal'),...
+    '-a',spm('Dir'))
diff --git a/config/spm_run_bms_dcm.m b/config/spm_run_bms_dcm.m
index 0c26527..c6876f9 100644
--- a/config/spm_run_bms_dcm.m
+++ b/config/spm_run_bms_dcm.m
@@ -17,7 +17,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % CC Chen & Maria Joao Rosa
-% $Id: spm_run_bms_dcm.m 3955 2010-06-29 17:26:29Z maria $
+% $Id: spm_run_bms_dcm.m 4229 2011-03-04 15:01:34Z maria $
 
 % input
 % -------------------------------------------------------------------------
@@ -53,7 +53,8 @@
         if data_se
             load(job.sess_dcm{1}(1).mod_dcm{1})
         else
-            error('Plase specify DCM.mat files to do BMA!')
+            spm('alert','Please specify DCM.mat files or model_space.mat to do BMA!','Error');
+            return
         end
     end
     
@@ -74,7 +75,8 @@
         disp('Loading model space')
         load(job.model_sp{1});
         if ~exist('subj','var')
-            error('Incorrect model space file! File must contain ''subj'' structure.')
+            spm('alert','Incorrect model space file! File must contain ''subj'' structure.','Error');
+            return
         end
     end
 end
@@ -110,7 +112,7 @@
         ns        = size(job.sess_dcm,2);                 % No of Subjects
         nsess     = size(job.sess_dcm{1},2);              % No of sessions
         nm        = size(job.sess_dcm{1}(1).mod_dcm,1);   % No of Models
-        fname_msp = [job.dir{1} 'model_space.mat'];
+        fname_msp = fullfile(job.dir{1},'model_space.mat');
     end
     
     F       = zeros(ns,nm);
@@ -118,7 +120,7 @@
     
     % Check if No of models > 2
     if nm < 2
-        msgbox('Please select more than one file')
+        spm('alert','Please select more than one file','Error');
         return
     end
     
@@ -215,16 +217,16 @@
                 failind = find(max(abs(diff(ID)),[],1) > eps);
                 if ~isempty(failind)
                     out.files{1} = [];
-                    msgbox(['Error: the models for subject ' num2str(k) ...
-                        ' session(s) ' num2str(failind) ' were not fitted to the same data.']);
+                    str_error = ['The models for subject ' num2str(k) ...
+                        ' session(s) ' num2str(failind) ' were not fitted to the same data.'];
+                    spm('alert',str_error,'Error');
                     return
                 end
             end
         else
             out.files{1} = [];
-            msgbox('Error: the number of sessions/models should be the same for all subjects!')
-            return
-            
+            spm('alert','The number of sessions/models should be the same for all subjects!','Error');
+            return         
         end
         
     end
@@ -253,8 +255,9 @@
         m_indx  = 1:nm;
         
         if nfam ~= npart || npart == 1 || maxpart > npart
-            error('Invalid family file!')
+            spm('alert','Invalid family file!','Error');
             out.files{1} = [];
+            return
         end
     else
         do_family = 0;
@@ -285,7 +288,7 @@
         nmodfam = length(m_indx);
         
         if nfam ~= npart || npart == 1 || maxpart > npart || nmodfam > nm
-            error('Invalid family!')
+            spm('alert','Invalid family!','Error');
             out.files{1} = [];
             return
         end
@@ -325,7 +328,8 @@
                         indx  = find(family.partition==bma_fam);
                         post  = model.post(indx);
                     else
-                        error('Incorrect family for BMA!');
+                        spm('alert','Incorrect family for BMA!','Error');
+                        return
                     end
                 end
             end
@@ -349,8 +353,8 @@
     if exist(fullfile(job.dir{1},'BMS.mat'),'file')
         load(fname);
         if  isfield(BMS,'DCM') && isfield(BMS.DCM,'ffx')
-            str = { 'Warning: existing BMS.mat file has been over-written!'};
-            msgbox(str)
+            spm('alert','Existing BMS.mat file has been over-written!',...
+                'Warning');
         end
     end
     BMS.DCM.ffx.data    = fname_msp;
@@ -362,7 +366,7 @@
     BMS.DCM.ffx.bma     = bma;
     
     disp('Saving BMS.mat file...')
-    if spm_matlab_version_chk('7') >= 0
+    if spm_check_version('matlab','7') >= 0
         save(fname,'-V6','BMS');
     else
         save(fname,'BMS');
@@ -412,7 +416,8 @@
                         indx = find(family.partition==bma_fam);
                         post = model.g_post(:,indx);
                     else
-                        error('Incorrect family for BMA!');
+                        spm('alert','Incorrect family for BMA!','Error');
+                        return
                     end
                 end
             end
@@ -435,8 +440,8 @@
     if exist(fullfile(job.dir{1},'BMS.mat'),'file')
         load(fname);
         if  isfield(BMS,'DCM') && isfield(BMS.DCM,'rfx')
-            str = { 'Warning:  existing BMS.mat file has been over-written!'};
-            msgbox(str)
+            spm('alert','Existing BMS.mat file has been over-written!',...
+                'Warning');
         end
     end
     BMS.DCM.rfx.data    = fname_msp;
@@ -448,7 +453,7 @@
     BMS.DCM.rfx.bma     = bma;
     
     disp('Saving BMS.mat file...')
-    if spm_matlab_version_chk('7') >= 0
+    if spm_check_version('matlab','7') >= 0
         save(fname,'-V6','BMS');
     else
         save(fname,'BMS');
@@ -461,7 +466,7 @@
 % -------------------------------------------------------------------------
 if ~ld_msp && data_se && ~ld_f
     disp('Saving model space...')
-    if spm_matlab_version_chk('7') >= 0
+    if spm_check_version('matlab','7') >= 0
         save(fname_msp,'-V6','subj');
     else
         save(fname_msp,'subj');
diff --git a/config/spm_run_bms_dcm_vis.m b/config/spm_run_bms_dcm_vis.m
index 1b84ec2..2cde943 100644
--- a/config/spm_run_bms_dcm_vis.m
+++ b/config/spm_run_bms_dcm_vis.m
@@ -4,10 +4,10 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Chun-Chuan Chen and Maria Joao Rosa
-% $Id: spm_run_bms_dcm_vis.m 3465 2009-10-14 15:14:29Z guillaume $
+% $Id: spm_run_bms_dcm_vis.m 4136 2010-12-09 22:22:28Z guillaume $
 
-if isempty(job.file{1})==1
-    fname = spm_select([1 1],'^BMS.mat$','select BMS.mat file');
+if ~exist('job','var') || isempty(job.file{1})
+    fname = spm_select([1 1],'^BMS\.mat$','select BMS.mat file');
     load(fname);
 else
     load(job.file{1});
diff --git a/config/spm_run_coreg_estimate.m b/config/spm_run_coreg_estimate.m
index e6f2908..141f14b 100644
--- a/config/spm_run_coreg_estimate.m
+++ b/config/spm_run_coreg_estimate.m
@@ -6,24 +6,26 @@
 % job    - harvested job data structure (see matlabbatch help)
 % Output:
 % out    - computation results, usually a struct variable.
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_coreg_estimate.m 2187 2008-09-25 11:50:50Z volkmar $
+% $Id: spm_run_coreg_estimate.m 4161 2011-01-13 13:44:37Z volkmar $
 
 job = varargin{1};
-%disp(job);
-%disp(job.eoptions);
+if isempty(job.other{1})
+    job.other = {};
+end
 
-x  = spm_coreg(strvcat(job.ref), strvcat(job.source),job.eoptions);
-M  = inv(spm_matrix(x));
-PO = strvcat(strvcat(job.source),strvcat(job.other));
-MM = zeros(4,4,size(PO,1));
-for j=1:size(PO,1),
-    MM(:,:,j) = spm_get_space(deblank(PO(j,:)));
-end;
-for j=1:size(PO,1),
-    spm_get_space(deblank(PO(j,:)), M*MM(:,:,j));
-end;
-out.cfiles = cellstr(PO);
-return;
+x  = spm_coreg(char(job.ref), char(job.source), job.eoptions);
+
+M  = spm_matrix(x);
+PO = [job.source(:); job.other(:)];
+MM = zeros(4,4,numel(PO));
+for j=1:numel(PO)
+    MM(:,:,j) = spm_get_space(PO{j});
+end
+for j=1:numel(PO)
+    spm_get_space(PO{j}, M\MM(:,:,j));
+end
+
+out.cfiles = PO;
diff --git a/config/spm_run_coreg_estwrite.m b/config/spm_run_coreg_estwrite.m
index ae36ccb..7ce4845 100644
--- a/config/spm_run_coreg_estwrite.m
+++ b/config/spm_run_coreg_estwrite.m
@@ -9,7 +9,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_coreg_estwrite.m 2485 2008-11-21 13:31:24Z john $
+% $Id: spm_run_coreg_estwrite.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job = varargin{1};
 if isempty(job.other{1})
@@ -20,30 +20,31 @@
     error('This is not supposed to work.');
 end
 
-x  = spm_coreg(strvcat(job.ref), strvcat(job.source),job.eoptions);
+x  = spm_coreg(char(job.ref), char(job.source),job.eoptions);
 
 if numel(job.other)>0,
-    M  = inv(spm_matrix(x));
-    PO = {job.source{:} job.other{:}};
+    M  = spm_matrix(x);
+    PO = [job.source(:); job.other(:)];
     MM = zeros(4,4,numel(PO));
     for j=1:numel(PO),
         MM(:,:,j) = spm_get_space(PO{j});
     end
     for j=1:numel(PO),
-        spm_get_space(PO{j}, M*MM(:,:,j));
+        spm_get_space(PO{j}, M\MM(:,:,j));
     end
 else
     PO = job.source;
+    MM = zeros(4,4,numel(PO));
     for j=1:numel(PO)
         MM(:,:,j) = spm_get_space(PO{j});
     end
     for j=1:numel(PO),
-        M  = inv(spm_matrix(x(j,:)));
-        spm_get_space(PO{j}, M*MM(:,:,j));
+        M  = spm_matrix(x(j,:));
+        spm_get_space(PO{j}, M\MM(:,:,j));
     end
 end
 
-P            = strvcat(job.ref{:},job.source{:},job.other{:});
+P            = char(job.ref{:},job.source{:},job.other{:});
 flags.mask   = job.roptions.mask;
 flags.mean   = 0;
 flags.interp = job.roptions.interp;
@@ -53,7 +54,7 @@
 
 spm_reslice(P,flags);
 
-out.cfiles = {job.source{:} job.other{:}};
+out.cfiles = [job.source(:); job.other(:)];
 out.rfiles = cell(size(out.cfiles));
 for i=1:numel(out.cfiles),
     [pth,nam,ext,num] = spm_fileparts(out.cfiles{i});
diff --git a/config/spm_run_coreg_reslice.m b/config/spm_run_coreg_reslice.m
index c14121b..b7c1339 100644
--- a/config/spm_run_coreg_reslice.m
+++ b/config/spm_run_coreg_reslice.m
@@ -9,11 +9,11 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_coreg_reslice.m 2312 2008-10-07 17:02:46Z volkmar $
+% $Id: spm_run_coreg_reslice.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job = varargin{1};
 
-P            = strvcat(strvcat(job.ref),strvcat(job.source));
+P            = char([job.ref(:);job.source(:)]);
 flags.mask   = job.roptions.mask;
 flags.mean   = 0;
 flags.interp = job.roptions.interp;
diff --git a/config/spm_run_factorial_design.m b/config/spm_run_factorial_design.m
index 4d31a81..1a7e965 100644
--- a/config/spm_run_factorial_design.m
+++ b/config/spm_run_factorial_design.m
@@ -8,7 +8,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_run_factorial_design.m 3860 2010-05-04 15:59:25Z guillaume $
+% $Id: spm_run_factorial_design.m 4185 2011-02-01 18:46:18Z guillaume $
 
 %--------------------------------------------------------------------------
 % This function configures the design matrix (describing the general
@@ -357,8 +357,8 @@
         I(:,3)  = kron(ones(Npairs,1),[1 2]');
         I(:,4)  = I(:,1);
 
-        [H,Hnames] = spm_DesMtx(I(:,2),'-','Subject');
-        [B,Bnames] = spm_DesMtx(I(:,3),'-','Condition');
+        [B,Bnames] = spm_DesMtx(I(:,2),'-','Subject');
+        [H,Hnames] = spm_DesMtx(I(:,3),'-','Condition');
 
         % Names and levels
         SPM.factor(1).name     = 'Subject';
@@ -397,11 +397,10 @@
         SPM.factor(1).variance = 0;
         SPM.factor(1).dept     = 0;
 
-        H = []; Hnames = [];
         if job.des.mreg.incint==0
-            B = []; Bnames = '';
+            H = []; Hnames = '';
         else
-            [B,Bnames] = spm_DesMtx(I(:,2),'-','mean');
+            [H,Hnames] = spm_DesMtx(I(:,2),'-','mean');
         end
 
         for i=1:length(job.des.mreg.mcov)
@@ -446,33 +445,33 @@
         
         DesName = 'ANOVA - within subject';
         
-        anovaw=job.des.anovaw;
-        anovaw.fac(1).name='Subject';
-        anovaw.fac(1).dept=0;
-        anovaw.fac(1).variance=0;
-        anovaw.fac(1).gmsca=0;
-        anovaw.fac(1).ancova=0;
+        anovaw  = job.des.anovaw;
+        anovaw.fac(1).name      = 'Subject';
+        anovaw.fac(1).dept      = 0;
+        anovaw.fac(1).variance  = 0;
+        anovaw.fac(1).gmsca     = 0;
+        anovaw.fac(1).ancova    = 0;
         
-        anovaw.fac(2).name='Groups';
-        anovaw.fac(2).dept=job.des.anovaw.dept;
-        anovaw.fac(2).variance=job.des.anovaw.variance;
-        anovaw.fac(2).gmsca=job.des.anovaw.gmsca;
-        anovaw.fac(2).ancova=job.des.anovaw.ancova;
+        anovaw.fac(2).name      = 'Groups';
+        anovaw.fac(2).dept      = job.des.anovaw.dept;
+        anovaw.fac(2).variance  = job.des.anovaw.variance;
+        anovaw.fac(2).gmsca     = job.des.anovaw.gmsca;
+        anovaw.fac(2).ancova    = job.des.anovaw.ancova;
         
-        anovaw.fsuball.fsubject=anovaw.fsubject;
+        anovaw.fsuball.fsubject = anovaw.fsubject;
         
         % Main effect of subject and group
-        anovaw.maininters{1}.fmain.fnum=1;
-        anovaw.maininters{2}.fmain.fnum=2;
+        anovaw.maininters{1}.fmain.fnum = 1;
+        anovaw.maininters{2}.fmain.fnum = 2;
         
         [I,P,job.cov] = spm_design_within_subject(anovaw,job.cov);
             
-        [H,Hnames] = spm_design_flexible(anovaw,I);
+        [H,Hnames,B,Bnames]  = spm_design_flexible(anovaw,I);
         
-        for i=1:2,
-            SPM.factor(i)=anovaw.fac(i);
+        for i=1:2
+            SPM.factor(i)    = anovaw.fac(i);
         end
-        SPM.factor(1).levels   = length(job.des.anovaw.fsubject);
+        SPM.factor(1).levels = length(job.des.anovaw.fsubject);
         
     %-Full Factorial Design
     %======================================================================
@@ -526,7 +525,7 @@
             
         end
 
-        [H,Hnames] = spm_design_flexible(job.des.fblock,I);
+        [H,Hnames,B,Bnames] = spm_design_flexible(job.des.fblock,I);
         
         for i=1:nf
             % Names and levels
@@ -1009,10 +1008,10 @@
 %--------------------------------------------------------------------------
 fprintf('%-40s: ','Saving SPM configuration')                           %-#
 
-if spm_matlab_version_chk('7') >= 0
-    save('SPM', 'SPM', '-V6');
+if spm_check_version('matlab','7') >= 0
+    save('SPM.mat', 'SPM', '-V6');
 else
-    save('SPM', 'SPM');
+    save('SPM.mat', 'SPM');
 end
 fprintf('%30s\n','...SPM.mat saved')                                    %-#
 
diff --git a/config/spm_run_fmri_data.m b/config/spm_run_fmri_data.m
index 0ee139c..d128820 100644
--- a/config/spm_run_fmri_data.m
+++ b/config/spm_run_fmri_data.m
@@ -10,7 +10,7 @@
 %_________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_fmri_data.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: spm_run_fmri_data.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 spm('defaults','FMRI');
@@ -36,11 +36,11 @@
 %-Save SPM.mat
 %-------------------------------------------------------------------------
 fprintf('%-40s: ','Saving SPM configuration')                          %-#
-if spm_matlab_version_chk('7') >= 0
-    save('SPM','-V6','SPM');
+if spm_check_version('matlab','7') >= 0
+    save('SPM.mat','-V6','SPM');
 else
-    save('SPM','SPM');
-end;
+    save('SPM.mat','SPM');
+end
 
 fprintf('%30s\n','...SPM.mat saved')                                   %-#
 
diff --git a/config/spm_run_fmri_design.m b/config/spm_run_fmri_design.m
index 737ca59..c924f39 100644
--- a/config/spm_run_fmri_design.m
+++ b/config/spm_run_fmri_design.m
@@ -10,7 +10,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_fmri_design.m 3692 2010-01-21 21:43:31Z guillaume $
+% $Id: spm_run_fmri_design.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 original_dir = pwd;
@@ -302,11 +302,11 @@
 %-Save SPM.mat
 %-------------------------------------------------------------------------
 fprintf('%-40s: ','Saving SPM configuration')                          %-#
-if spm_matlab_version_chk('7') >= 0
-    save('SPM','-V6','SPM');
+if spm_check_version('matlab','7') >= 0
+    save('SPM.mat','-V6','SPM');
 else
-    save('SPM','SPM');
-end;
+    save('SPM.mat','SPM');
+end
 
 fprintf('%30s\n','...SPM.mat saved')                                   %-#
 
diff --git a/config/spm_run_fmri_est.m b/config/spm_run_fmri_est.m
index 5e617f1..41a1389 100644
--- a/config/spm_run_fmri_est.m
+++ b/config/spm_run_fmri_est.m
@@ -10,7 +10,7 @@
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_fmri_est.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_run_fmri_est.m 4022 2010-07-28 12:50:20Z guillaume $
 
 %-Load SPM.mat file
 %--------------------------------------------------------------------------
@@ -70,7 +70,7 @@
             for r=1:size(Fcon,1)
                 con       = Fcon(r,:);
                 str       = cons(i).name;
-                if ~isempty(strmatch('Interaction',str))
+                if strncmp('Interaction',str,11)
                     name  = ['Positive ',str,'_',int2str(r)];
                 else
                     sp1   = find(isspace(str), 1);
diff --git a/config/spm_run_fmri_spec.m b/config/spm_run_fmri_spec.m
index 22efede..5760ee6 100644
--- a/config/spm_run_fmri_spec.m
+++ b/config/spm_run_fmri_spec.m
@@ -10,7 +10,7 @@
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_fmri_spec.m 3757 2010-03-08 11:41:53Z guillaume $
+% $Id: spm_run_fmri_spec.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 original_dir = pwd;
@@ -308,10 +308,10 @@
 %-Save SPM.mat
 %--------------------------------------------------------------------------
 fprintf('%-40s: ','Saving SPM configuration')                           %-#
-if spm_matlab_version_chk('7') >= 0
-    save('SPM','-V6','SPM');
+if spm_check_version('matlab','7') >= 0
+    save('SPM.mat','-V6','SPM');
 else
-    save('SPM','SPM');
+    save('SPM.mat','SPM');
 end;
 
 fprintf('%30s\n','...SPM.mat saved')                                    %-#
diff --git a/config/spm_run_minc.m b/config/spm_run_minc.m
index f0bd0b6..4fd14ed 100644
--- a/config/spm_run_minc.m
+++ b/config/spm_run_minc.m
@@ -1,4 +1,4 @@
-function out = convert_minc(job)
+function out = spm_run_minc(job)
 % SPM job execution function
 % takes a harvested job data structure and call SPM functions to perform
 % computations on the data.
@@ -9,7 +9,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_minc.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_run_minc.m 4136 2010-12-09 22:22:28Z guillaume $
 
 if job.opts.ext(1) ~= '.', job.opts.ext = ['.' job.opts.ext]; end
 out.files = cell(size(job.data));
diff --git a/config/spm_run_normalise_estimate.m b/config/spm_run_normalise_estimate.m
index b39e701..3ddc5bd 100644
--- a/config/spm_run_normalise_estimate.m
+++ b/config/spm_run_normalise_estimate.m
@@ -9,23 +9,17 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_normalise_estimate.m 2335 2008-10-13 18:49:25Z christophe $
+% $Id: spm_run_normalise_estimate.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job    = varargin{1};
-o      = job.eoptions;
-eflags = struct(...
-    'smosrc', o.smosrc,...
-    'smoref', o.smoref,...
-    'regtype',o.regtype,...
-    'cutoff', o.cutoff,...
-    'nits',   o.nits,...
-    'reg',    o.reg);
+eflags = rmfield(job.eoptions,{'template','weight'});
 
+out = repmat(struct('params',{''}),size(job.subj));
 for i=1:numel(job.subj),
-    [pth,nam,ext,ind] = spm_fileparts(strvcat(job.subj(i).source{:}));
-    out(i).params{1}  = fullfile(pth,[nam,'_sn.mat']);
-    spm_normalise(strvcat(job.eoptions.template{:}),...
-        strvcat(job.subj(i).source{:}), out(i).params{1},...
-        strvcat(job.eoptions.weight), strvcat(job.subj(i).wtsrc), eflags);
+    [pth,nam] = spm_fileparts(char(job.subj(i).source));
+    out(i).params  = {fullfile(pth,[nam,'_sn.mat'])};
+    spm_normalise(char(job.eoptions.template),...
+        char(job.subj(i).source), out(i).params{1},...
+        char(job.eoptions.weight), char(job.subj(i).wtsrc), eflags);
 end;
 return;
diff --git a/config/spm_run_normalise_estwrite.m b/config/spm_run_normalise_estwrite.m
index 04358bb..9188115 100644
--- a/config/spm_run_normalise_estwrite.m
+++ b/config/spm_run_normalise_estwrite.m
@@ -9,33 +9,20 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_normalise_estwrite.m 2312 2008-10-07 17:02:46Z volkmar $
+% $Id: spm_run_normalise_estwrite.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job    = varargin{1};
-o      = job.eoptions;
-eflags = struct(...
-    'smosrc', o.smosrc,...
-    'smoref', o.smoref,...
-    'regtype',o.regtype,...
-    'cutoff', o.cutoff,...
-    'nits',   o.nits,...
-    'reg',    o.reg);
-o      = job.roptions;
-rflags = struct(...
-    'preserve',o.preserve,...
-    'bb',      o.bb,...
-    'vox',     o.vox,...
-    'interp',  o.interp,...
-    'wrap',    o.wrap,...
-        'prefix',  o.prefix);
+eflags = rmfield(job.eoptions,{'weight','template'});
+rflags = job.roptions;
 
+out = repmat(struct('params',{''},'files',{''}),size(job.subj));
 for i=1:numel(job.subj),
-    [ pth,nam ] = spm_fileparts(deblank(job.subj(i).source{:}));
-    out(i).params{1}  = fullfile(pth,[nam,'_sn.mat']);
-    spm_normalise(strvcat(job.eoptions.template{:}),...
-        strvcat(job.subj(i).source{:}), out(i).params{1},...
-        strvcat(job.eoptions.weight), strvcat(job.subj(i).wtsrc), eflags);
-    spm_write_sn(strvcat(job.subj(i).resample{:}), out(i).params{1}, rflags);
+    [ pth,nam ] = spm_fileparts(char(job.subj(i).source));
+    out(i).params  = {fullfile(pth,[nam,'_sn.mat'])};
+    spm_normalise(char(job.eoptions.template),...
+        char(job.subj(i).source), out(i).params{1},...
+        char(job.eoptions.weight), char(job.subj(i).wtsrc), eflags);
+    spm_write_sn(char(job.subj(i).resample), out(i).params{1}, rflags);
     out(i).files = cell(size(job.subj(i).resample));
     for j=1:numel(job.subj(i).resample),
         [pth,nam,ext,num] = spm_fileparts(job.subj(i).resample{j});
diff --git a/config/spm_run_normalise_write.m b/config/spm_run_normalise_write.m
index c4e94c3..8f08321 100644
--- a/config/spm_run_normalise_write.m
+++ b/config/spm_run_normalise_write.m
@@ -9,21 +9,15 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_normalise_write.m 2312 2008-10-07 17:02:46Z volkmar $
+% $Id: spm_run_normalise_write.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job    = varargin{1};
-o      = job.roptions;
-rflags = struct(...
-    'preserve',o.preserve,...
-    'bb',      o.bb,...
-    'vox',     o.vox,...
-    'interp',  o.interp,...
-    'wrap',    o.wrap,...
-        'prefix',  o.prefix);
+rflags = job.roptions;
 
+out = repmat(struct('files',{''}),size(job.subj));
 for i=1:numel(job.subj),
-    spm_write_sn(strvcat(job.subj(i).resample{:}),...
-        strvcat(job.subj(i).matname{:}),rflags);
+    spm_write_sn(char(job.subj(i).resample),...
+        char(job.subj(i).matname),rflags);
     res = job.subj(i).resample;
     out(i).files = cell(size(res));
     for j=1:numel(res),
diff --git a/config/spm_run_preproc.m b/config/spm_run_preproc.m
index f25e0d9..2b812b2 100644
--- a/config/spm_run_preproc.m
+++ b/config/spm_run_preproc.m
@@ -6,14 +6,14 @@
 % job    - harvested job data structure (see matlabbatch help)
 % Output:
 % out    - computation results, usually a struct variable.
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_preproc.m 2312 2008-10-07 17:02:46Z volkmar $
+% $Id: spm_run_preproc.m 4185 2011-02-01 18:46:18Z guillaume $
 
-job.opts.tpm = strvcat(job.opts.tpm{:});
+job.opts.tpm = char(job.opts.tpm);
 if isfield(job.opts,'msk'),
-    job.opts.msk = strvcat(job.opts.msk{:});
+    job.opts.msk = char(job.opts.msk);
 end;
 for i=1:numel(job.data),
     res           = spm_preproc(job.data{i},job.opts);
@@ -50,20 +50,19 @@
     end;
 end;
 return;
-%------------------------------------------------------------------------
 
-%------------------------------------------------------------------------
+%==========================================================================
 function savefields(fnam,p)
-if length(p)>1, error('Can''t save fields.'); end;
+if length(p)>1, error('Can''t save fields.'); end
 fn = fieldnames(p);
-if numel(fn)==0, return; end;
-for i=1:length(fn),
+if numel(fn)==0, return; end
+for i=1:length(fn)
     eval([fn{i} '= p.' fn{i} ';']);
-end;
-if spm_matlab_version_chk('7') >= 0
+end
+if spm_check_version('matlab','7') >= 0
     save(fnam,'-V6',fn{:});
 else
     save(fnam,fn{:});
-end;
+end
 
 return;
diff --git a/config/spm_run_realign_estimate.m b/config/spm_run_realign_estimate.m
index 59a674b..dcdfbff 100644
--- a/config/spm_run_realign_estimate.m
+++ b/config/spm_run_realign_estimate.m
@@ -9,18 +9,18 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_realign_estimate.m 1185 2008-03-04 16:31:21Z volkmar $
+% $Id: spm_run_realign_estimate.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job           = varargin{1};
-P             = {};
+P             = cell(size(job.data));
 for i=1:length(job.data),
-    P{i}  = strvcat(job.data{i});
+    P{i}  = char(job.data{i});
 end;
 flags.quality = job.eoptions.quality;
 flags.fwhm    = job.eoptions.fwhm;
 flags.sep     = job.eoptions.sep;
 flags.rtm     = job.eoptions.rtm;
-flags.PW      = strvcat(job.eoptions.weight);
+flags.PW      = char(job.eoptions.weight);
 flags.interp  = job.eoptions.interp;
 flags.wrap    = job.eoptions.wrap;
 spm_realign(P,flags);
diff --git a/config/spm_run_realign_estwrite.m b/config/spm_run_realign_estwrite.m
index 1741ea7..43bcc2b 100644
--- a/config/spm_run_realign_estwrite.m
+++ b/config/spm_run_realign_estwrite.m
@@ -9,18 +9,18 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_realign_estwrite.m 1185 2008-03-04 16:31:21Z volkmar $
+% $Id: spm_run_realign_estwrite.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job           = varargin{1};
-P             = {};
-for i=1:length(job.data),
-    P{i} = strvcat(job.data{i});
+P             = cell(size(job.data));
+for i=1:numel(job.data),
+    P{i} = char(job.data{i});
 end;
 flags.quality = job.eoptions.quality;
 flags.fwhm    = job.eoptions.fwhm;
 flags.sep     = job.eoptions.sep;
 flags.rtm     = job.eoptions.rtm;
-flags.PW      = strvcat(job.eoptions.weight);
+flags.PW      = char(job.eoptions.weight);
 flags.interp  = job.eoptions.interp;
 flags.wrap    = job.eoptions.wrap;
 spm_realign(P,flags);
@@ -30,11 +30,10 @@
     out.sess(i).rpfile{1} =  fullfile(pth, sprintf('rp_%s.txt', nam));
 end;
 
-P            = strvcat(P);
+P            = char(P);
 flags.mask   = job.roptions.mask;
-flags.mean   = job.roptions.which(2);
 flags.interp = job.roptions.interp;
-flags.which  = job.roptions.which(1);
+flags.which  = job.roptions.which;
 flags.wrap   = job.roptions.wrap;
 flags.prefix = job.roptions.prefix;
 spm_reslice(P,flags);
diff --git a/config/spm_run_realign_reslice.m b/config/spm_run_realign_reslice.m
index f28f478..0ec2c7c 100644
--- a/config/spm_run_realign_reslice.m
+++ b/config/spm_run_realign_reslice.m
@@ -9,14 +9,13 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_realign_reslice.m 1185 2008-03-04 16:31:21Z volkmar $
+% $Id: spm_run_realign_reslice.m 4152 2011-01-11 14:13:35Z volkmar $
 
 job          = varargin{1};
-P            = strvcat(job.data);
+P            = char(job.data);
 flags.mask   = job.roptions.mask;
-flags.mean   = job.roptions.which(2);
 flags.interp = job.roptions.interp;
-flags.which  = job.roptions.which(1);
+flags.which  = job.roptions.which;
 flags.wrap   = job.roptions.wrap;
 flags.prefix = job.roptions.prefix;
 spm_reslice(P,flags);
diff --git a/config/spm_run_realignunwarp.m b/config/spm_run_realignunwarp.m
index 2194a35..875dc96 100644
--- a/config/spm_run_realignunwarp.m
+++ b/config/spm_run_realignunwarp.m
@@ -10,7 +10,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Darren R. Gitelman
-% $Id: spm_run_realignunwarp.m 2337 2008-10-14 14:16:54Z volkmar $
+% $Id: spm_run_realignunwarp.m 4185 2011-02-01 18:46:18Z guillaume $
 
 job = varargin{1};
 
@@ -21,7 +21,7 @@
 flags.fwhm    = job.eoptions.fwhm;
 flags.sep     = job.eoptions.sep;
 flags.rtm     = job.eoptions.rtm;
-flags.PW      = strvcat(job.eoptions.weight);
+flags.PW      = char(job.eoptions.weight);
 flags.interp  = job.eoptions.einterp;
 flags.wrap    = job.eoptions.ewrap;
 
@@ -65,9 +65,10 @@
 
 % assemble files
 %-----------------------------------------------------------------------
-P = {};
+P   = cell(size(job.data));
+sfP = cell(size(job.data));
 for i = 1:numel(job.data)
-    P{i} = strvcat(job.data(i).scans{:});
+    P{i} = char(job.data(i).scans{:});
     if ~isempty(job.data(i).pmscan)
         sfP{i} = job.data(i).pmscan{1};
     else
@@ -91,7 +92,7 @@
     [path,name] = fileparts(P{i}(1,:));
     out.sess(i).dsfile{1} =  fullfile(path,[name '_uw.mat']);
 
-    if spm_matlab_version_chk('7') >= 0
+    if spm_check_version('matlab','7') >= 0
         save(out.sess(i).dsfile{1},'-V6','ds');
     else
         save(out.sess(i).dsfile{1},'ds');
diff --git a/config/spm_run_reorient.m b/config/spm_run_reorient.m
index 4551263..fcaa2d2 100644
--- a/config/spm_run_reorient.m
+++ b/config/spm_run_reorient.m
@@ -9,7 +9,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: spm_run_reorient.m 3130 2009-05-18 14:41:31Z volkmar $
+% $Id: spm_run_reorient.m 4177 2011-01-27 14:35:32Z volkmar $
 
 job = varargin{1};
 if isfield(job.transform,'transprm')
@@ -28,7 +28,7 @@
     for k = 1:numel(job.srcfiles)
         V = spm_vol(job.srcfiles{k});
         X = spm_read_vols(V);
-        [p n e v] = spm_fileparts(V.fname)
+        [p n e v] = spm_fileparts(V.fname);
         V.mat = job.transform.transM*V.mat;
         V.fname = fullfile(p, [job.prefix n e v]);
         spm_write_vol(V,X);
diff --git a/config/spm_run_voi.m b/config/spm_run_voi.m
index f7cbe36..3d84563 100644
--- a/config/spm_run_voi.m
+++ b/config/spm_run_voi.m
@@ -10,7 +10,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_run_voi.m 3780 2010-03-15 17:15:00Z guillaume $
+% $Id: spm_run_voi.m 4269 2011-03-29 16:03:43Z guillaume $
 
 fprintf('## Note: this VOI facility is in a beta version.      ##\n');
 fprintf('## Interface and features might change in the future. ##\n');
@@ -65,7 +65,9 @@
 xSPM.XYZ   = XYZ;
 xSPM.M     = SPM.xVol.M; % irrelevant here
 
+if ~isempty(xY.Ic), cwd = pwd; cd(SPM.swd); end % to find beta images
 [Y,xY]     = spm_regions(xSPM,SPM,[],xY);
+if  ~isempty(xY.Ic), cd(cwd); end
 
 %-Export results
 %--------------------------------------------------------------------------
@@ -185,7 +187,8 @@
     error('The SPM index does not correspond to a Thresholded SPM ROI.');
 end
 [mySPM, xSPM] = getSPM(job.roi{m}.spm);
-XYZ           = SPM.xVol.iM(1:3,:)*[xSPM.XYZmm;ones(1,size(xSPM.XYZmm,2))];
+XYZmm         = xSPM.XYZmm;
+XYZ           = SPM.xVol.iM(1:3,:)*[XYZmm;ones(1,size(XYZmm,2))];
 Z             = xSPM.Z;
 if ~isempty(e)
     R         = spm_sample_vol(uint8(roi_eval(voi,e)), ...
@@ -194,7 +197,7 @@
     XYZmm     = xSPM.XYZmm(:,R);
     Z         = xSPM.Z(R);
 end
-[N Z M A]     = spm_max(Z,XYZ);
+[N, Z, M]     = spm_max(Z,XYZ);
 if isempty(Z)
     warning('No voxel survived. Default to user-specified centre.');
     return
@@ -207,8 +210,8 @@
         nc    = SPM.xVol.M(1:3,:)*[M(:,j);1];
         str   = sprintf(['centre moved to global maxima ' str],nc);
     case 'local'
-        XYZ   = SPM.xVol.M(1:3,:)*[M;ones(1,size(M,2))];
-        nc    = spm_XYZreg('NearestXYZ',c,XYZ);
+        XYZmm = SPM.xVol.M(1:3,:)*[M;ones(1,size(M,2))];
+        nc    = spm_XYZreg('NearestXYZ',c,XYZmm);
         str   = sprintf(['centre moved from ' str ' to ' str],c,nc);
     case 'supra'
         nc    = spm_XYZreg('NearestXYZ',c,XYZmm);
diff --git a/man/FieldMap/fieldmap.tex b/man/FieldMap/fieldmap.tex
index e5f7793..ba3e28c 100644
--- a/man/FieldMap/fieldmap.tex
+++ b/man/FieldMap/fieldmap.tex
@@ -7,18 +7,310 @@ \section{Introduction}
 \url{http://www.fil.ion.ucl.ac.uk/spm/software/spm5/}\\
 FieldMap Version 1.1 for SPM2 can be downloaded from
 \url{http://www.fil.ion.ucl.ac.uk/spm/toolbox/fieldmap/}}
- for creating unwrapped field maps that can be used to do geometric distortion correction of EPI images \cite{jezzard95,jenkinson03,chloe_distortion,chloe_distortion2,ja_geometric}. The toolbox can be used in an interactive mode so that the user can see the effect of applying different field maps and unwarping parameters to EPI images. However, once a set of parameters has been established for a specific scanning protocol, the toolbox routines can be scripted or batched using the SPM8 user-interface. The toolbox creates a voxel-displacement map that can be used with Realign \& Unwarp for doing a combined static and dynamic distortion correction. FieldMap and Unwarp are designed only to work with images collected with the phase-encode direction aligned with the anterior-posterior direction.
+ for creating unwrapped field maps that can be used to do geometric distortion correction of EPI images \cite{chloe_distortion,chloe_distortion2,ja_geometric}. The methods are based on earlier work by Jezzard et al.,\cite{jezzard95} and a phase-unwrapping algorithm by Jenkinson \cite{jenkinson03}. The toolbox can be used via the SPM batch editor or in an interactive mode so that the user can see the effect of applying different field maps and unwarping parameters to EPI images. A voxel displacement map (VDM) is created that can be used with Realign \& Unwarp for doing a combined static and dynamic distortion correction or with an Apply VDM function for doing a static distortion correction on a set of realigned images. Realign \& Unwarp is designed to work only with images acquired with the phase-encode direction aligned with the anterior-posterior axis. Images acquired with phase-encode directions aligned with other axes can be distortion corrected using the FieldMap toolbox and Apply VDM utility.
+
+\begin{figure}
+\begin{center}
+\includegraphics[width=70mm]{FieldMap/fieldmap_taskmgr}
+\end{center}
+\caption{FieldMap using the SPM8 User Interface.\label{FM1}}
+\end{figure}
+
+\section{Presubtracted Phase and Magnitude Data}
+Calculate a voxel displacement map (VDM) from presubtracted phase and magnitude field map data (Figure \ref{FM1}). This option expects a single magnitude image and a single phase image resulting from the subtraction of two phase images (where the subtraction is usually done automatically by the scanner software). The phase image will be scaled between +/- PI.
+
+
+\subsection{Data}
+Subjects or sessions for which individual field map data has been acquired.
+
+
+\subsubsection{Subject}
+Data for this subject or field map session.
+
+
+\paragraph{Phase Image}
+Select a single phase image. This should be the result from the subtraction of two phase images (where the subtraction is usually done automatically by the scanner software). The phase image will be scaled between +/- PI.
+
+
+\paragraph{Magnitude Image}
+Select a single magnitude image. This is used for masking the phase information and coregistration with the EPI data. If two magnitude images are available, select the one acquired at the shorter echo time because it will have greater signal
+
+
+\paragraph{FieldMap defaults}
+FieldMap default values can be entered as a file or set of values.
+
+
+\subparagraph{Defaults values}
+Defaults values
+
+
+\textbf{Echo times [short TE long TE]}
+Enter the short and long echo times (in ms) of the data used to acquire the field map.
+
+
+\textbf{Mask brain}
+Select masking or no masking of the brain. If masking is selected, the magnitude image is used to generate a mask of the brain.
+
+
+\textbf{Blip direction}
+Enter the blip direction. This is the polarity of the phase-encode blips describing the direction in which k-space is traversed along the y-axis during EPI acquisition with respect to the coordinate system used in SPM. In this coordinate system, the phase encode direction corresponds with the y-direction and is defined as positive from the posterior to the anterior of the head.
+
+The convention used to describe the direction of the k-space traversal is based on the coordinate system used by SPM. In this coordinate system, the phase encode direction corresponds with the y-direction and is defined as positive from the posterior to the anterior of the head. The x-direction is defined as positive from left to right and the z-direction is defined as positive from foot to head. The polarity of the phase-encode blips describes in which direction k-space is traversed along the y-axis with respect to the coordinate system described here. 
+
+\textbf{Total EPI readout time}
+Enter the total EPI readout time (in ms). This is the time taken to acquire all of the phase encode steps required to cover k-space (ie one image slice). For example, if the EPI sequence has 64 phase encode steps, the total readout time is the time taken to acquire 64 echoes, e.g. total readout time = number of echos x echo spacing. This time does not include i) the duration of the excitation, ii) the delay between, the excitation and the start of the acquisition or iii) time for fat saturation etc.
+
+
+\textbf{EPI-based field map?}
+Select non-EPI or EPI based field map. The field map data may be acquired using a non-EPI sequence (typically a gradient echo sequence) or an EPI sequence. The processing will be slightly different for the two cases. If using an EPI-based field map, the resulting Voxel Displacement Map will be inverted since the field map was acquired in distorted space.
+
+
+\textbf{Jacobian modulation?}
+Select whether or not to use Jacobian modulation. This will adjust the intensities of voxels that have been stretched or compressed but in general is not recommended for EPI distortion correction
+
+
+\textbf{uflags}
+Different options for phase unwrapping and field map processing
+
+
+\textsc{Unwrapping method}
+Select method for phase unwrapping
+
+
+\textsc{FWHM}
+FWHM of Gaussian filter used to implement weighted smoothing of unwrapped maps.
+
+
+\textsc{pad}
+Size of padding kernel if required.
+
+
+\textsc{Weighted smoothing}
+Select normal or weighted smoothing.
+
+
+\textbf{mflags}
+Different options used for the segmentation and creation of the brain mask.
+
+
+\textsc{Template image for brain masking}
+Select template file for segmentation to create brain mask
+
+
+\textsc{FWHM}
+FWHM of Gaussian filter for smoothing brain mask.
+
+
+\textsc{Number of erosions}
+Number of erosions used to create brain mask.
+
+
+\textsc{Number of dilations}
+Number of dilations used to create brain mask.
+
+
+\textsc{Threshold}
+Threshold used to create brain mask from segmented data.
+
+
+\textsc{Regularization}
+Regularization value used in the segmentation. A larger value helps the segmentation to converge.
+
+
+\subparagraph{Defaults File \label{PMDEF}}
+Select the 'pm\_defaults*.m' file containing the parameters for the field map data. Please make sure that the parameters defined in the defaults file are correct for your field map and EPI sequence. To create your own customised defaults file, either edit the distributed version and/or save it with the name 'pm\_defaults\_yourname.m'.
+
+
+\paragraph{EPI Sessions}
+If a single set of field map data will be used for multiple EPI runs/sessions, select the first EPI in each run/session. A VDM file will created for each run/session, matched to the first EPI in each run/session and saved with a unique name extension.
+
+
+\subparagraph{Session}
+Data for this session.
+
+
+\textbf{Select EPI to Unwarp}
+Select a single image to distortion correct. The corrected image will be saved with the prefix u. Note that this option is mainly for quality control of correction so that the original and distortion corrected images can be displayed for comparison. To unwarp multiple images please use either Realign \& Unwarp or Apply VDM.
+
+
+\paragraph{Match VDM to EPI?}
+Match VDM file to EPI image. This will coregister the field map data to the selected EPI for each run/session.
+
+In general, the field map data should be acquired so that it is as closely registered with the EPI data as possible but matching can be selected if required. If a precalculated field map was loaded then the user is prompted to select a magnitude image in the same space as the field map. If real and imaginary images were selected, the toolbox automatically creates a magnitude image from these images and saves it with the name mag\_NAME-OF-FIRST-INPUT-IMAGE.img. 
+
+\paragraph{Name extension for run/session specific VDM file}
+This will be the name extension followed by an incremented integer for run/session specific VDM files.
+
+
+\paragraph{Write unwarped EPI?}
+Write out distortion corrected EPI image. The image is saved with the prefix u. Note that this option is mainly for quality control of correction so that the original and distortion corrected images can be displayed for comparison. To unwarp multiple images please use either Realign \& Unwarp or Apply VDM.
+
+
+\paragraph{Select anatomical image for comparison}
+Select an anatomical image for comparison with the distortion corrected EPI or leave empty. Note that this option is mainly for quality control of correction.
+
+
+\paragraph{Match anatomical image to EPI?}
+Match the anatomical image to the distortion corrected EPI. Note that this option is mainly for quality control of correction allowing for visual inspection and comparison of the distortion corrected EPI.
+
+
+\section{Real and Imaginary Data}
+Calculate a voxel displacement map (VDM) from real and imaginary field map data. This option expects two real and imaginary pairs of data of two different echo times. The phase images will be scaled between +/- PI.
+
+
+\subsection{Data}
+Subjects or sessions for which individual field map data has been acquired.
+
+
+\subsubsection{Subject}
+Data for this subject or field map session.
+
+
+\paragraph{Short Echo Real Image}
+Select short echo real image
+
+
+\paragraph{Short Echo Imaginary Image}
+Select short echo imaginary image
+
+
+\paragraph{Long Echo Real Image}
+Select long echo real image
+
+
+\paragraph{Long Echo Imaginary Image}
+Select long echo imaginary image
+
+\paragraph{Other inputs}
+As for Presubtracted Phase and Magnitude Data.
+
+\section{Phase and Magnitude Data}
+Calculate a voxel displacement map (VDM) from double phase and magnitude field map data. This option expects two phase and magnitude pairs of data of two different echo times.
+
+
+\subsection{Data}
+Subjects or sessions for which individual field map data has been acquired.
+
+
+\subsubsection{Subject}
+Data for this subject or field map session.
+
+
+\paragraph{Short Echo Phase Image}
+Select short echo phase image
+
+
+\paragraph{Short Echo Magnitude Image}
+Select short echo magnitude image
+
+
+\paragraph{Long Echo Phase Image}
+Select long echo phase image
+
+
+\paragraph{Long Echo Magnitude Image}
+Select long echo magnitude image
+
+\paragraph{Other inputs}
+As for Presubtracted Phase and Magnitude Data.
+
+\section{Precalculated FieldMap (in Hz)}
+Calculate a voxel displacement map (VDM) from a precalculated field map. This option expects a processed field map (ie phase unwrapped, masked if necessary and scaled to Hz). Precalculated field maps can be generated by the FieldMap toolbox and stored as fpm\_* files.
+
+
+\subsection{Data}
+Subjects or sessions for which individual field map data has been acquired.
+
+
+\subsubsection{Subject}
+Data for this subject or field map session.
+
+
+\paragraph{Precalculated field map}
+Select a precalculated field map. This should be a processed field map (ie phase unwrapped, masked if necessary and scaled to Hz) , for example as generated by the FieldMap toolbox and are stored with fpm\_* prefix.
+
+
+\paragraph{Select magnitude image in same space as fieldmap}
+Select magnitude image which is in the same space as the field map to do matching to EPI.
+
+\paragraph{Other inputs}
+As for Presubtracted Phase and Magnitude Data.
+
+\section{Apply VDM }
+Apply VDM (voxel displacement map) to resample voxel values in selected image(s). This allows a VDM to be applied to any images which are assumed to be already realigned (e.g. including EPI fMRI time series and DTI data).
+
+The VDM can be been created from a field map acquisition using the FieldMap toolbox and comprises voxel shift values which describe geometric distortions occuring as a result of magnetic susceptbility artefacts. Distortions along any single dimension can be corrected therefore input data may have been acquired with phase encode directions in X, Y (most typical) and Z.
+
+The selected images are assumed to be realigned to the first in the time series (e.g. using Realign: Estimate) but do not need to be resliced. The VDM is assumed to be in alignment with the images selected for resampling (note this can be achieved via the FieldMap toolbox). The resampled images are written to the input subdirectory with the same (prefixed) filename.
+
+e.g. The typical processing steps for fMRI time series would be 1) Realign: Estimate, 2) FieldMap to create VDM, 3) Apply VDM.
+
+Note that this routine is a general alternative to using the VDM in combination with Realign \& Unwarp which estimates and corrects for the combined effects of static and movement-related susceptibility induced distortions. Apply VDM can be used when dynamic distortions are not (well) modelled by Realign \& Unwarp (e.g. for fMRI data acquired with R-$>$L phase-encoding direction, high field fMRI data or DTI data).
+
+
+\subsection{Data}
+Subjects or sessions for which VDM file is being applied to images.
+
+
+\subsubsection{Session}
+Data for this session.
+
+
+\paragraph{Images}
+Select scans for this session. These are assumed to be realigned to the first in the time series (e.g. using Realign: Estimate) but do not need to be resliced
+
+
+\paragraph{Fieldmap (vdm* file)}
+Select VDM (voxel displacement map) for this session (e.g. created via FieldMap toolbox). This is assumed to be in alignment with the images selected for resampling (note this can be achieved via the FieldMap toolbox).
+
+
+\subsection{Reslice Options}
+Apply VDM reslice options
+
+
+\subsubsection{Distortion direction}
+In which direction are the distortions? Any single dimension can be corrected therefore input data may have been acquired with phase encode directions in Y (most typical), X or Z
+
+
+\subsubsection{Reslice which images ?}
+All Images (1..n) 
+
+  This applies the VDM and reslices all the images. 
+
+All Images + Mean Image 
+
+   This applies the VDM reslices all the images and creates a mean of the resliced images.
+
+
+\subsubsection{Interpolation}
+The method by which the images are sampled when being written in a different space. Nearest Neighbour is fastest, but not recommended for image realignment. Bilinear Interpolation is probably OK for PET, but not so suitable for fMRI because higher degree interpolation generally gives better results \cite{thevenaz00a,unser93a,unser93b}. Although higher degree methods provide better interpolation, but they are slower because they use more neighbouring voxels.
+
+
+\subsubsection{Wrapping}
+This indicates which directions in the volumes the values should wrap around in.  For example, in MRI scans, the images wrap around in the phase encode direction, so (e.g.) the subject's nose may poke into the back of the subject's head. These are typically:
+
+    No wrapping - for PET or images that have already                   been spatially transformed. Also the recommended option if                   you are not really sure.
+
+    Wrap in  Y  - for (un-resliced) MRI where phase encoding                   is in the Y direction (voxel space) etc.
+
+
+\subsubsection{Masking}
+Because of subject motion, different images are likely to have different patterns of zeros from where it was not possible to sample data. With masking enabled, the program searches through the whole time series looking for voxels which need to be sampled from outside the original images. Where this occurs, that voxel is set to zero for the whole set of images (unless the image format can represent NaN, in which case NaNs are used where possible).
+
+
+\subsubsection{Filename Prefix}
+Specify the string to be prepended to the filenames of the distortion corrected image file(s). Default prefix is 'u'.
 
 \begin{figure}
 \begin{center}
 \includegraphics[width=70mm]{FieldMap/fieldmap_gui1}
 \includegraphics[width=70mm]{FieldMap/fieldmap_results1}
 \end{center}
-\caption{FieldMap GUI and Results. \label{FM1}}
+\caption{FieldMap GUI and Results. \label{FM2}}
 \end{figure}
 
 \section{Creating Field Maps Using the FieldMap GUI}
-The FieldMap Toolbox GUI is shown on the left Figure \ref{FM1}. It is divided into two parts. The top part deals with creating the field map in Hz and the bottom part deals with creating the voxel displacement map (VDM) and unwarping the EPI. The toolbox can be used by working through the different inputs in the following order:
+The FieldMap Toolbox GUI is shown on the left Figure \ref{FM2}. It is divided into two parts. The top part deals with creating the field map in Hz and the bottom part deals with creating the voxel displacement map (VDM) and unwarping the EPI. The toolbox can be used by working through the different inputs in the following order:
 
 \subsection{Create field map in Hz}
 
@@ -117,21 +409,6 @@ \subsubsection{Help}
 \subsubsection{Quit}
 Quit the toolbox and closes all windows associated with it.
 
-\section{Using the pm\_defaults file \label{PMDEF}}
-Input parameters and the mode in which the toolbox works can be customised in the defaults file called pm\_defaults.m. Different defaults files can be selected. This means that it is possible to have more than one set of default parameters, e.g. to accommodate different scanners or sequences. To be recognised by the FieldMap toolbox, these files must be named pm\_defaults\_NAME.m and have the same format as the existing pm\_defaults.m. These defaults files can be loaded using the FieldMap GUI (as described in section \ref{loaddefs}) or selected when filling in the job fields using the SPM User-Interface (as described in section 6).
-
-\begin{figure}
-\begin{center}
-\includegraphics[width=70mm]{FieldMap/fieldmap_taskmgr}
-\end{center}
-\caption{FieldMap using the SPM8 User Interface.\label{FM2}}
-\end{figure}
-
-\section{Using the SPM8 User Interface \label{SPM8UI}}
-FieldMap version 2.1 for SPM8 can be run using the SPM8 UI (Figure \ref{FM2}). To do this, from the top menu on the SPM8 Graphics window, select TASKS $\rightarrow$  SPM(interactive) $\rightarrow$ Tools $\rightarrow$  FieldMap. Select the type of FieldMap job to run from 1) Presubtracted Phase and Magntiude Data (this means a single phase and magnitude pair), 2) Real and Imaginary Data, 3) Phase and Magnitude Data (this means a double phase and magnitude pair) and 4) Precalculated FieldMap. In the Batch Editor, select New ``Subject'' in the lower panel, then select each item in the upper panel to fill in the different job fields. All fields marked with an X must be filled in. The other fields can be optionally filled or contain default values which can be changed if required.
-
-\section{Using FieldMap with multiple sessions or runs}
-If multiple EPI sessions or runs have been acquired, but with only a single set of fieldmap data, it is possible to automatically create a vdm file for each session. To do this, for each new each session under the input field EPI sessions, select the first EPI in each session. A copy of the VDM file will be matched to the first EPI in each session and saved with a separate name.
 
 \section{Using the FieldMap in Batch scripts}
 \texttt{FieldMap\_preprocess.m} which calls \texttt{FieldMap\_create.m} gives an example of how to run the FieldMap toolbox without using the GUI. To run the script, make sure your \matlab\ path includes the directory where the FieldMap toolbox is installed. This can be done using the Set Path option under File in the \matlab\ windows manager or using the command: 
@@ -163,7 +440,6 @@ \section{Using the FieldMap in Batch scripts}
 
 \emph{match} match vdm file to first EPI in run (1/0).
 
-
 NB: FieldMap will match the field map to the first epi image in the time series (after removing the dummy scans). Therefore, epi\_dir must be the directory that contains the epi run that all other images will be realigned to.
 
 The script will create an fpm* file, a vdm5\_* file and an unwarped version of the EPI saved with the prescript ``u''.
diff --git a/man/FieldMap/fieldmap_gui1.png b/man/FieldMap/fieldmap_gui1.png
index 11fa4e0..e0e3d47 100644
Binary files a/man/FieldMap/fieldmap_gui1.png and b/man/FieldMap/fieldmap_gui1.png differ
diff --git a/man/FieldMap/fieldmap_results1.png b/man/FieldMap/fieldmap_results1.png
index df2268e..62de716 100644
Binary files a/man/FieldMap/fieldmap_results1.png and b/man/FieldMap/fieldmap_results1.png differ
diff --git a/man/FieldMap/fieldmap_taskmgr.png b/man/FieldMap/fieldmap_taskmgr.png
index d6798f4..425d8c1 100644
Binary files a/man/FieldMap/fieldmap_taskmgr.png and b/man/FieldMap/fieldmap_taskmgr.png differ
diff --git a/man/auditory/aud_bayes.png b/man/auditory/aud_bayes.png
index b31b211..daaebb4 100644
Binary files a/man/auditory/aud_bayes.png and b/man/auditory/aud_bayes.png differ
diff --git a/man/auditory/aud_bayes2.png b/man/auditory/aud_bayes2.png
index dbd6da2..1ce4d8d 100644
Binary files a/man/auditory/aud_bayes2.png and b/man/auditory/aud_bayes2.png differ
diff --git a/man/auditory/aud_orth.png b/man/auditory/aud_orth.png
index 13bdc63..2195857 100644
Binary files a/man/auditory/aud_orth.png and b/man/auditory/aud_orth.png differ
diff --git a/man/auditory/auditory.tex b/man/auditory/auditory.tex
index 4741d6b..d391bd7 100644
--- a/man/auditory/auditory.tex
+++ b/man/auditory/auditory.tex
@@ -19,7 +19,7 @@ \section{Spatial pre-processing}
 
 \subsection{Realignment}
 
-Under the spatial pre-processing section of the SPM base window select textsc{Realign (Est \& Res)} from the \textsc{Realign} pulldown menu. This will call up a realignment job specification in the batch editor. Then
+Under the spatial pre-processing section of the SPM base window select \textsc{Realign (Est \& Res)} from the \textsc{Realign} pulldown menu. This will call up a realignment job specification in the batch editor. Then
 \begin{itemize}
 \item Highlight data, select ``New Session'', then highlight the newly created ``Session'' option.
 \item Select ''Specify Files'' and use the SPM file selector to choose all of your functional images eg. ``\texttt{fM000*.img}''. There should be 96 files.
@@ -410,7 +410,7 @@ \subsection{Overlays}
 \end{center}
 \end{figure}
 
-It is also possible to project and display the results on a surface mesh, we are going to use here one of the canonical mesh distributed with SPM (in MNI space). Press ``Overlays'' and choose ``Render'', then go in the \texttt{canonical} folder of your SPM installation and select file \texttt{cortex\_20484.gii} (this is a surface mesh stored using the GIfTI format) and you will obtain a figure similar to \ref{aud_render2}.
+It is also possible to project and display the results on a surface mesh, we are going to use here one of the canonical mesh distributed with SPM (in MNI space). Press ``Overlays'' and choose ``Render'', then go in the \texttt{canonical} folder of your SPM installation and select file \texttt{cortex\_20484.surf.gii} (this is a surface mesh stored using the GIfTI format) and you will obtain a figure similar to \ref{aud_render2}.
 
 \begin{figure}
 \begin{center}
diff --git a/man/auditory/bias.png b/man/auditory/bias.png
index 7edee40..a69a90e 100644
Binary files a/man/auditory/bias.png and b/man/auditory/bias.png differ
diff --git a/man/auditory/checkreg.png b/man/auditory/checkreg.png
index 655f556..2213117 100644
Binary files a/man/auditory/checkreg.png and b/man/auditory/checkreg.png differ
diff --git a/man/auditory/command.png b/man/auditory/command.png
index b5078a6..bf0f4a9 100644
Binary files a/man/auditory/command.png and b/man/auditory/command.png differ
diff --git a/man/auditory/con_man.png b/man/auditory/con_man.png
index 04f6e85..acf62e8 100644
Binary files a/man/auditory/con_man.png and b/man/auditory/con_man.png differ
diff --git a/man/auditory/con_man2.png b/man/auditory/con_man2.png
index 6b637be..340a809 100644
Binary files a/man/auditory/con_man2.png and b/man/auditory/con_man2.png differ
diff --git a/man/auditory/con_man3.png b/man/auditory/con_man3.png
index a91ff2f..bde0f39 100644
Binary files a/man/auditory/con_man3.png and b/man/auditory/con_man3.png differ
diff --git a/man/auditory/contrast.png b/man/auditory/contrast.png
index 3e05e78..bce6bac 100644
Binary files a/man/auditory/contrast.png and b/man/auditory/contrast.png differ
diff --git a/man/auditory/coreg.png b/man/auditory/coreg.png
index b1e851c..5171ace 100644
Binary files a/man/auditory/coreg.png and b/man/auditory/coreg.png differ
diff --git a/man/auditory/design.png b/man/auditory/design.png
index bfd9614..85e6d36 100644
Binary files a/man/auditory/design.png and b/man/auditory/design.png differ
diff --git a/man/auditory/explore.png b/man/auditory/explore.png
index ed298c8..385c2b6 100644
Binary files a/man/auditory/explore.png and b/man/auditory/explore.png differ
diff --git a/man/auditory/fitted.png b/man/auditory/fitted.png
index 4aba37d..038bcea 100644
Binary files a/man/auditory/fitted.png and b/man/auditory/fitted.png differ
diff --git a/man/auditory/gray.png b/man/auditory/gray.png
index 9a2aea8..b6af345 100644
Binary files a/man/auditory/gray.png and b/man/auditory/gray.png differ
diff --git a/man/auditory/interactive.png b/man/auditory/interactive.png
index 6113a18..f8e6493 100644
Binary files a/man/auditory/interactive.png and b/man/auditory/interactive.png differ
diff --git a/man/auditory/realign.png b/man/auditory/realign.png
index 9447642..2979e9e 100644
Binary files a/man/auditory/realign.png and b/man/auditory/realign.png differ
diff --git a/man/auditory/render.png b/man/auditory/render.png
index f0b9b50..8eb45ba 100644
Binary files a/man/auditory/render.png and b/man/auditory/render.png differ
diff --git a/man/auditory/sections.png b/man/auditory/sections.png
index ebe374d..e4e783c 100644
Binary files a/man/auditory/sections.png and b/man/auditory/sections.png differ
diff --git a/man/auditory/slices.png b/man/auditory/slices.png
index cc5bce3..f91295f 100644
Binary files a/man/auditory/slices.png and b/man/auditory/slices.png differ
diff --git a/man/auditory/smooth.png b/man/auditory/smooth.png
index aa6e659..6b21129 100644
Binary files a/man/auditory/smooth.png and b/man/auditory/smooth.png differ
diff --git a/man/auditory/spm1.png b/man/auditory/spm1.png
index fa2c4bc..288441c 100644
Binary files a/man/auditory/spm1.png and b/man/auditory/spm1.png differ
diff --git a/man/auditory/volume.png b/man/auditory/volume.png
index 2538b1b..a20d989 100644
Binary files a/man/auditory/volume.png and b/man/auditory/volume.png differ
diff --git a/man/batch/batch.tex b/man/batch/batch.tex
index 1744e3a..c4cfa49 100644
--- a/man/batch/batch.tex
+++ b/man/batch/batch.tex
@@ -1,11 +1,12 @@
-\chapter{Batch interface}
-\label{Chap:batch_interface}
+\chapter{Batch tutorial}
+\label{Chap:batch:tutorial}
 
 Details about the algorithms used for data processing are given in the other
 sections of this manual. This section explains how a sequence of processing
 steps can be run at once without MATLAB programming. SPM8 includes
-\texttt{matlabbatch}\footnote{\url{http://sourceforge.net/projects/matlabbatch}} which has been derived from the SPM5
-batch system, but is also available as a separate package.
+\texttt{matlabbatch}\footnote{\url{http://sourceforge.net/projects/matlabbatch}}
+which has been derived from the SPM5 batch system, but is also available as a
+separate package.
 
 In \texttt{matlabbatch}, each data processing step is called ``module''. There
 are e.g.\ modules for spatial processing of MRI data (realignment,
@@ -31,7 +32,7 @@ \chapter{Batch interface}
 \item[Multiple batches] Multiple batches can be loaded and executed together.
 \end{description}
 
-\section{Batch tutorial - single subject}
+\section{Single subject}
 
 In this tutorial we will develop a batch for spatial processing and fMRI
 statistics of a single subject of the ``Face'' example dataset (see
@@ -92,18 +93,10 @@ \subsubsection{SPM processing}
 
 \subsection{Add modules to the batch}
 
-\begin{figure}[htbp]
-  \centering
-  \includegraphics{batch/tasks}
-  \caption{TASKS menu in ``Graphics'' window}
-  \label{fig:tasks}
-\end{figure}
 The helper modules and the SPM processing modules can be assembled using the
-GUI. Click the ``BATCH'' button in the SPM Menu window or locate the Graphics
-window and open the batch editor by selecting ``BATCH'' from the ``TASKS''
-menu (fig.~\ref{fig:tasks}). First, add the helper modules, followed by the
-SPM modules in the order listed above. Do not configure any details until you
-have selected all modules.
+GUI. Click the ``BATCH'' button in the SPM Menu window. First, add the helper
+modules, followed by the SPM modules in the order listed above. Do not configure
+any details until you have selected all modules.
 
 \begin{figure}
   \centering
@@ -532,12 +525,8 @@ \subsubsection{Complete and run a pre-specified job}
     % fMRI model specification: Multiple conditions - cfg_files
     inputs{4, crun} = MATLAB_CODE_TO_FILL_INPUT; 
 end
-job_id = cfg_util('initjob', jobs);
-sts    = cfg_util('filljob', job_id, inputs{:});
-if sts
-    cfg_util('run', job_id);
-end
-cfg_util('deljob', job_id);
+spm('defaults','fmri');
+spm_jobman('serial',jobs,'',inputs{:});
 \end{verbatim}
 
 The skeleton needs to be adapted to the actual data layout by adding
@@ -556,193 +545,3 @@ \subsection{Modifying a saved job}
 \verb|.m| files. These files contain a number of MATLAB commands,
 which will create a variable \verb|matlabbatch|. The commands can be
 modified to set different values, add or remove options. 
- 
-\section{SPM5 to matlabbatch transition guide}
-\label{sec:overview}
-
-This is a short overview to describe code organisation and interfaces between
-SPM and the batch system.
-
-\subsection{Code Reorganisation}
-\label{sec:overcode}
-
-The following paths have changed:
-\begin{itemize}
-\item \verb|fullfile(spm('dir'),'matlabbatch')| Core batch system.
-\item \verb|fullfile(spm('dir'),'config')| New SPM config files.
-\item \verb|fullfile(spm('dir'),'oldconfig')| Old SPM config files (unused)
-\item \verb|spm_jobman.m| and \verb|spm_select.m| replaced with compatibility
-  code
-\item \verb|spm_Menu.fig| Callbacks adapted
-\end{itemize}
-
-Configuration code has been generated automatically from the existing SPM
-configuration using \verb|cfg_struct2cfg| and \verb|gencode|. This sometimes
-results in redundant/duplicate code. Also, conditional constructs like
-\verb|if|, \verb|case| may not have been considered. 
-
-Some assignments to configuration items are guarded by validity
-checks. Usually, there will be a warning issued if a wrong value is
-supplied. Special care needs to be taken for \verb|.prog|, \verb|.vfiles|,
-\verb|.vout|, \verb|.check| functions or function handles. The functions
-referenced here must be on MATLAB path before they are assigned to one of
-these fields. For toolboxes, this implies that toolbox paths must be added at
-the top of the configuration file.
-
-For details, see section~\ref{sec:code}. 
-
-\subsection{Interfaces between SPM and Matlabbatch}
-\label{sec:overinter}
-
-\begin{description}
-\item[Unchanged] harvested job structure.
-\item[Changed] Top-level node in SPM config now called \verb|spmjobs| instead
-  of \verb|jobs|. New overall top-level node
-  \verb|matlabbatch|. \verb|spm_jobman| will convert and load SPM5 style batch
-  jobs into the new batch system.
-\item[Changed] Configuration file syntax - instead of structs, configuration
-  items are now objects. Structs of type \verb|<type>| are now represented as
-  objects of class \verb|cfg_<type>|. Existing SPM5 configuration can
-  be imported using \verb|cfg_struct2cfg|. There is a new class
-  \verb|cfg_exbranch| which is used for branches that have a \verb|.prog|
-  field.
-\item[Deprecated] Virtual files have been replaced by dependencies. These
-  require computations to return a single output argument (e.g. a cell,
-  struct). Parts of this output argument can be passed on to new inputs at
-  run-time. Virtual files are treated as a special output argument.
-\item[Added] Interface to the batch system
-  \begin{itemize}
-  \item \verb|cfg_util| Configuration management, job management, job
-    execution
-  \item \verb|cfg_serial| A utility to fill missing inputs and run a job
-    (optionally with a GUI input function)
-  \item \verb|cfg_ui| GUI - inspired by \verb|spm_jobman|, but modified to
-    work around some MATLAB GUI ``features'' (like input widgets loosing focus
-    before editing has finished).
-  \end{itemize}
-
-\end{description}
-
-\section{Configuration Code Details}
-\label{sec:code}
-
-Configuration code has been split into two files per configuration:
-\begin{description}
-\item[spm\_cfg\_*.m] Configuration classes, \verb|.check|, \verb|.vout|
-  subfunctions
-\item[spm\_run\_*.m] Run-time code, takes job structure as input and returns
-  output structure as specified in \verb|.vout|.
-\end{description}
-In a few cases (where there was no additional magic in the code), run-time
-code has been integrated into the main SPM code. This may be useful to run
-test batches without using the configuration/batch system.
-
-\subsection{Virtual Outputs}
-
-Virtual outputs are described by arrays of \verb|cfg_dep| objects. These
-objects contain a ``source'' and a ``target'' part. Functions may have more
-than one virtual output (e.g. one output per session, a collection of results
-variables). One \verb|cfg_dep| object has to be created for each output.
-
-Only two fields in the ``source'' part need to be set in a \verb|.vout|
-callback: 
-\begin{description}
-\item[sname] A display name for this output. This will appear in the
-  dependencies list and should describe the contents of this dependency.
-\item[src\_output] A subscript reference that can be used to address this
-  output in the variable returned at run-time.
-\item[tgt\_spec (optional)] A description on what kind of inputs this output
-  should be displayed as dependency. This is not very convenient yet, the
-  \verb|match| and \verb|cfg_findspec| methods are very restrictive in the
-  kind of expressions that are allowed.
-\end{description}
-
-The \verb|.vout| callback will be evaluated once the configuration system
-thinks that enough information about the \emph{structure} of the outputs is
-available. This condition is met, once all in-tree nodes
-\verb|cfg_(ex)branch|, \verb|cfg_choice|, \verb|cfg_repeat| have the required
-number of child nodes. 
-
-The \verb|.vout| callback is called with a job structure as input, but its
-code \emph{should not rely} on the evaluation of any contents of this
-structure (or at least provide a fallback). The contents of the leaf nodes may
-not be set or may contain a dependency object instead of a value during
-evalution of \verb|.vout|.
-
-The ``target'' part will be filled by the configuration classes, the
-\verb|src_exbranch| field is set in \verb|cfg_util|.
-
-\subsection{SPM Startup}
-
-The top level configuration file for SPM is \verb|spm_cfg.m|. It collects SPM
-core configuration files and does toolbox autodetection. If a toolbox
-directory contains \verb|*_cfg_*.m| files, they will be loaded. Otherwise, if
-there are only SPM5-style \verb|*_config_*.m| files, the configuration will be
-converted at run-time using \verb|cfg_struct2cfg|.
-
-\subsection{Defaults Settings}
-
-In Matlabbatch, there are different ways to set defaults:
-\begin{enumerate}
-\item \label{def:cf}in the configuration file itself,
-\item \label{def:cd}in a defaults file, which has a structure similar to a harvested
-  job,
-\item \label{def:def}using a \verb|.def| field for leaf items.
-\end{enumerate}
-Defaults set using option \ref{def:cf} or \ref{def:cd} will only be
-updated at SPM/matlabbatch startup. Defaults set using option
-\ref{def:def} will be set once a new job is started. These defaults take
-precedence over the other defaults.
-
-In core SPM, these defaults refer to \verb|spm_get_defaults|, which accesses
-\verb|spm_defaults|. Toolboxes may use their own callback functions.
-
-\subsection{Toolbox Migration}
-
-In the \verb|fullfile(spm('dir'),'toolbox')| folder there exists a migration
-utility \verb|spm_tbx_config2cfg.m|. This utility will create a
-\verb|*_cfg_*.m| and a \verb|*_def_*.m| file based on the configuration tree
-given as input argument.
-
-Toolboxes should set their defaults using the \verb|.def| fields,
-using a mechanism similar to \verb|spm_get_defaults|. This allows for
-flexibility without interfering with SPMs own defaults.
-
-\section{Utilities}
-
-\subsection{Batch Utilities}
-
-Matlabbatch is designed to support multiple applications. A standard
-application ``BasicIO'' is enabled by default. Among other options, it
-contains file/file selection manipulation utilities which can be used as as
-dependency source if multiple functions require the same set of files as input
-argument. For debugging purposes, ``Pass Output to Workspace'' can be used to
-assign outputs of a computation to a workspace variable.
-
-The \verb|cfg_confgui| folder contains an application which describes all
-configuration items in terms of configuration items. It is not enabled by
-default, but can be added to the batch system using
-\verb|cfg_util('addapp'...)|. This utility can be used generate a batch
-configuration file with the batch system itself.
-
-\subsection{Matlab Code Generation}
-
-The \verb|gencode| utility generates MATLAB \verb|.m| file code for any kind
-of MATLAB variable. This is used to save batch files as well as to generate
-configuration code.
-
-\subsection{Configuration Management}
-
-The backend utility to manage the configuration data is \verb|cfg_util|. It
-provides callbacks to add application configurations, and to load, modify,
-save or run jobs. These callbacks are used by two frontends: \verb|cfg_ui| is
-a MATLAB GUI, while \verb|cfg_serial| can be used both as a GUI and in script
-mode. In script mode, it will fill in job inputs from an argument list. This
-allows to run predefined jobs with e.g. subject dependent inputs without
-knowing the exact details of the job structure.
-
-
-%%% Local Variables: 
-%%% mode: latex
-%%% TeX-master: ../manual.tex
-%%% End: 
diff --git a/man/batch/batch_basicio.png b/man/batch/batch_basicio.png
index ba53d68..5b92361 100644
Binary files a/man/batch/batch_basicio.png and b/man/batch/batch_basicio.png differ
diff --git a/man/batch/batch_dependencies.png b/man/batch/batch_dependencies.png
index 11eb0af..29f84d0 100644
Binary files a/man/batch/batch_dependencies.png and b/man/batch/batch_dependencies.png differ
diff --git a/man/batch/batch_multi_subject_template.png b/man/batch/batch_multi_subject_template.png
index 4313191..f6c1c6f 100644
Binary files a/man/batch/batch_multi_subject_template.png and b/man/batch/batch_multi_subject_template.png differ
diff --git a/man/batch/batch_single_subject.png b/man/batch/batch_single_subject.png
index 8b47cbe..52b7e2c 100644
Binary files a/man/batch/batch_single_subject.png and b/man/batch/batch_single_subject.png differ
diff --git a/man/batch/batch_single_subject_template.png b/man/batch/batch_single_subject_template.png
index f720a6d..0c1ed8f 100644
Binary files a/man/batch/batch_single_subject_template.png and b/man/batch/batch_single_subject_template.png differ
diff --git a/man/batch/batch_single_subject_template_nodeps.png b/man/batch/batch_single_subject_template_nodeps.png
index 87a2f4e..487c2af 100644
Binary files a/man/batch/batch_single_subject_template_nodeps.png and b/man/batch/batch_single_subject_template_nodeps.png differ
diff --git a/man/batch/batch_spm.png b/man/batch/batch_spm.png
index 778391a..46d24dd 100644
Binary files a/man/batch/batch_spm.png and b/man/batch/batch_spm.png differ
diff --git a/man/batch/devel.tex b/man/batch/devel.tex
new file mode 100644
index 0000000..350f631
--- /dev/null
+++ b/man/batch/devel.tex
@@ -0,0 +1,180 @@
+\chapter{Developer's guide}
+\label{Chap:batch:devel}
+
+\section{SPM5 to Matlabbatch transition guide}
+
+This is a short overview to describe code organisation and interfaces between
+SPM and the batch system.
+
+\subsection{Code Reorganisation}
+
+The following paths have changed:
+\begin{itemize}
+\item \verb|fullfile(spm('dir'),'matlabbatch')| Core batch system.
+\item \verb|fullfile(spm('dir'),'config')| New SPM config files.
+\item \verb|fullfile(spm('dir'),'oldconfig')| Old SPM config files (unused)
+\item \verb|spm_jobman.m| and \verb|spm_select.m| replaced with compatibility
+  code
+\item \verb|spm_Menu.fig| Callbacks adapted
+\end{itemize}
+
+Configuration code has been generated automatically from the existing SPM
+configuration using \verb|cfg_struct2cfg| and \verb|gencode|. This sometimes
+results in redundant/duplicate code. Also, conditional constructs like
+\verb|if|, \verb|case| may not have been considered. 
+
+Some assignments to configuration items are guarded by validity
+checks. Usually, there will be a warning issued if a wrong value is
+supplied. Special care needs to be taken for \verb|.prog|, \verb|.vfiles|,
+\verb|.vout|, \verb|.check| functions or function handles. The functions
+referenced here must be on MATLAB path before they are assigned to one of
+these fields. For toolboxes, this implies that toolbox paths must be added at
+the top of the configuration file.
+
+\subsection{Interfaces between SPM and Matlabbatch}
+
+\begin{description}
+\item[Unchanged] harvested job structure.
+\item[Changed] Top-level node in SPM config now called \verb|spmjobs| instead
+  of \verb|jobs|. New overall top-level node
+  \verb|matlabbatch|. \verb|spm_jobman| will convert and load SPM5 style batch
+  jobs into the new batch system.
+\item[Changed] Configuration file syntax - instead of structs, configuration
+  items are now objects. Structs of type \verb|<type>| are now represented as
+  objects of class \verb|cfg_<type>|. Existing SPM5 configuration can
+  be imported using \verb|cfg_struct2cfg|. There is a new class
+  \verb|cfg_exbranch| which is used for branches that have a \verb|.prog|
+  field.
+\item[Deprecated] Virtual files have been replaced by dependencies. These
+  require computations to return a single output argument (e.g. a cell,
+  struct). Parts of this output argument can be passed on to new inputs at
+  run-time. Virtual files are treated as a special output argument.
+\item[Added] Interface to the batch system
+  \begin{itemize}
+  \item \verb|cfg_util| Configuration management, job management, job
+    execution
+  \item \verb|cfg_serial| A utility to fill missing inputs and run a job
+    (optionally with a GUI input function)
+  \item \verb|cfg_ui| GUI - inspired by \verb|spm_jobman|, but modified to
+    work around some MATLAB GUI ``features'' (like input widgets loosing focus
+    before editing has finished).
+  \end{itemize}
+
+\end{description}
+
+\section{Configuration Code Details}
+
+Configuration code has been split into two files per configuration:
+\begin{description}
+\item[spm\_cfg\_*.m] Configuration classes, \verb|.check|, \verb|.vout|
+  subfunctions
+\item[spm\_run\_*.m] Run-time code, takes job structure as input and returns
+  output structure as specified in \verb|.vout|.
+\end{description}
+In a few cases (where there was no additional magic in the code), run-time
+code has been integrated into the main SPM code. This may be useful to run
+test batches without using the configuration/batch system.
+
+\subsection{Virtual Outputs}
+
+Virtual outputs are described by arrays of \verb|cfg_dep| objects. These
+objects contain a ``source'' and a ``target'' part. Functions may have more
+than one virtual output (e.g. one output per session, a collection of results
+variables). One \verb|cfg_dep| object has to be created for each output.
+
+Only two fields in the ``source'' part need to be set in a \verb|.vout|
+callback: 
+\begin{description}
+\item[sname] A display name for this output. This will appear in the
+  dependencies list and should describe the contents of this dependency.
+\item[src\_output] A subscript reference that can be used to address this
+  output in the variable returned at run-time.
+\item[tgt\_spec (optional)] A description on what kind of inputs this output
+  should be displayed as dependency. This is not very convenient yet, the
+  \verb|match| and \verb|cfg_findspec| methods are very restrictive in the
+  kind of expressions that are allowed.
+\end{description}
+
+The \verb|.vout| callback will be evaluated once the configuration system
+thinks that enough information about the \emph{structure} of the outputs is
+available. This condition is met, once all in-tree nodes
+\verb|cfg_(ex)branch|, \verb|cfg_choice|, \verb|cfg_repeat| have the required
+number of child nodes. 
+
+The \verb|.vout| callback is called with a job structure as input, but its
+code \emph{should not rely} on the evaluation of any contents of this
+structure (or at least provide a fallback). The contents of the leaf nodes may
+not be set or may contain a dependency object instead of a value during
+evalution of \verb|.vout|.
+
+The ``target'' part will be filled by the configuration classes, the
+\verb|src_exbranch| field is set in \verb|cfg_util|.
+
+\subsection{SPM Startup}
+
+The top level configuration file for SPM is \verb|spm_cfg.m|. It collects SPM
+core configuration files and does toolbox autodetection. If a toolbox
+directory contains \verb|*_cfg_*.m| files, they will be loaded. Otherwise, if
+there are only SPM5-style \verb|*_config_*.m| files, the configuration will be
+converted at run-time using \verb|cfg_struct2cfg|.
+
+\subsection{Defaults Settings}
+
+In Matlabbatch, there are different ways to set defaults:
+\begin{enumerate}
+\item \label{def:cf}in the configuration file itself,
+\item \label{def:cd}in a defaults file, which has a structure similar to a harvested
+  job,
+\item \label{def:def}using a \verb|.def| field for leaf items.
+\end{enumerate}
+Defaults set using option \ref{def:cf} or \ref{def:cd} will only be
+updated at SPM/matlabbatch startup. Defaults set using option
+\ref{def:def} will be set once a new job is started. These defaults take
+precedence over the other defaults.
+
+In core SPM, these defaults refer to \verb|spm_get_defaults|, which accesses
+\verb|spm_defaults|. Toolboxes may use their own callback functions.
+
+\subsection{Toolbox Migration}
+
+In the \verb|fullfile(spm('dir'),'toolbox')| folder there exists a migration
+utility \verb|spm_tbx_config2cfg.m|. This utility will create a
+\verb|*_cfg_*.m| and a \verb|*_def_*.m| file based on the configuration tree
+given as input argument.
+
+Toolboxes should set their defaults using the \verb|.def| fields,
+using a mechanism similar to \verb|spm_get_defaults|. This allows for
+flexibility without interfering with SPMs own defaults.
+
+\section{Utilities}
+
+\subsection{Batch Utilities}
+
+Matlabbatch is designed to support multiple applications. A standard
+application ``BasicIO'' is enabled by default. Among other options, it
+contains file/file selection manipulation utilities which can be used as as
+dependency source if multiple functions require the same set of files as input
+argument. For debugging purposes, ``Pass Output to Workspace'' can be used to
+assign outputs of a computation to a workspace variable.
+
+The \verb|cfg_confgui| folder contains an application which describes all
+configuration items in terms of configuration items. It is not enabled by
+default, but can be added to the batch system using
+\verb|cfg_util('addapp'...)|. This utility can be used generate a batch
+configuration file with the batch system itself.
+
+\subsection{MATLAB Code Generation}
+
+The \verb|gencode| utility generates MATLAB \verb|.m| file code for any kind
+of MATLAB variable. This is used to save batch files as well as to generate
+configuration code.
+
+\subsection{Configuration Management}
+
+The backend utility to manage the configuration data is \verb|cfg_util|. It
+provides callbacks to add application configurations, and to load, modify,
+save or run jobs. These callbacks are used by two frontends: \verb|cfg_ui| is
+a MATLAB GUI, while \verb|cfg_serial| can be used both as a GUI and in script
+mode. In script mode, it will fill in job inputs from an argument list. This
+allows to run predefined jobs with e.g. subject dependent inputs without
+knowing the exact details of the job structure.
diff --git a/man/batch/face_single_subject_script.m b/man/batch/face_single_subject_script.m
index 954aad1..a1cc968 100644
--- a/man/batch/face_single_subject_script.m
+++ b/man/batch/face_single_subject_script.m
@@ -13,9 +13,5 @@
     inputs{3, crun} = MATLAB_CODE_TO_FILL_INPUT; % Coreg: Estimate: Source Image - cfg_files
     inputs{4, crun} = MATLAB_CODE_TO_FILL_INPUT; % fMRI model specification: Multiple conditions - cfg_files
 end
-job_id = cfg_util('initjob', jobs);
-sts    = cfg_util('filljob', job_id, inputs{:});
-if sts
-    cfg_util('run', job_id);
-end
-cfg_util('deljob', job_id);
+spm('defaults','fmri');
+spm_jobman('serial',jobs,'',inputs{:});
diff --git a/man/batch/flow.png b/man/batch/flow.png
index 3861137..3c51e83 100644
Binary files a/man/batch/flow.png and b/man/batch/flow.png differ
diff --git a/man/batch/multimodal_fmri_script.m b/man/batch/multimodal_fmri_script.m
index ea14d58..d0facae 100644
--- a/man/batch/multimodal_fmri_script.m
+++ b/man/batch/multimodal_fmri_script.m
@@ -24,13 +24,13 @@
 inputs    = cell(6,numel(subjdirs));
 for sub = 1:numel(subjdirs)
     inputs{1,sub} = {fullfile(studydir,subjdirs{sub})};
-    inputs{2,sub} = cfg_getfile('FPList',fullfile(inputs{1,sub}{1}, ...
-                                                  'fMRI','Session1'),'^f.*\.img');
-    inputs{3,sub} = cfg_getfile('FPList',fullfile(inputs{1,sub}{1}, ...
-                                                  'fMRI','Session2'),'^f.*\.img');
+    inputs{2,sub} = cellstr(spm_select('FPList',fullfile(inputs{1,sub}{1}, ...
+                                                  'fMRI','Session1'),'^f.*\.img'));
+    inputs{3,sub} = cellstr(spm_select('FPList',fullfile(inputs{1,sub}{1}, ...
+                                                  'fMRI','Session2'),'^f.*\.img'));
     inputs{4,sub} = {fullfile(inputs{1,sub}{1},'sMRI','smri.img')};
     inputs{5,sub} = {fullfile(inputs{1,sub}{1},'fMRI','trials_ses1.mat')};
     inputs{6,sub} = {fullfile(inputs{1,sub}{1},'fMRI','trials_ses2.mat')};
 end
 spm('defaults','fmri');
-spm_jobman('serial',jobs,'',inputs{:});
\ No newline at end of file
+spm_jobman('serial',jobs,'',inputs{:});
diff --git a/man/biblio/methods_group.bib b/man/biblio/methods_group.bib
index 42de7e4..ea16f61 100644
--- a/man/biblio/methods_group.bib
+++ b/man/biblio/methods_group.bib
@@ -2510,30 +2510,6 @@ @Article{cc_induced
   keyword   = {DCM}
 }
 
-@Article{karl_filtering,
-  author    = karl,
-  title     = {Variational filtering},
-  journal   = ni,
-  year      = 2008,
-  volume    = 41,
-  number    = {3},
-  pages     = {747--766},
-  doi       = {10.1016/j.neuroimage.2008.03.017},
-  keyword   = {Bayesian}
-}
-
-@Article{karl_DEM,
-  author    = karl # and # nelson # and # jean,
-  title     = {{DEM}: A variational treatment of dynamic systems},
-  journal   = ni,
-  year      = 2008,
-  volume    = 41,
-  number    = {3},
-  pages     = {849--885},
-  doi       = {10.1016/j.neuroimage.2008.02.054},
-  keyword   = {DEM}
-}
-
 @Article{rm_spectralnmm,
   author    = rm # and # sjk # and # {N. Rombach} # and # {W.T. O'Connor} # and # {K.J. Murphy} # and # {R.B. Reilly} # and # karl,
   title     = {Bayesian estimation of synaptic physiology from the spectral responses of neural masses},
@@ -3094,10 +3070,11 @@ @Article{rnah_fmripriors
 title       = {A Parametric Empirical Bayesian framework for f{MRI}-constrained {MEG}/{EEG} source reconstruction},
 journal     = hbm,
 year        = {2010},
-volume      = {in press},
-number      = {},
-pages       = {},
-doi         = {}
+volume      = {31},
+number      = {10},
+pages       = {1512--1531},
+doi         = {10.1002/hbm.20956},
+pdf         = {/spm/doc/papers/rnah_HBM_10_fMRIPriors_EMEG.pdf}
 } 
 
 @Article{vl_optibf,
@@ -3227,6 +3204,147 @@ @Article{klaas_effcon_fmri
 keyword     = {DCM, fMRI}
 } 
 
+@Article{karl_network_discovery,
+author      = karl # and # {B. Li} # and # jean # and # klaas,
+title       = {Network discovery with DCM},
+journal     = ni,
+year        = {2010},
+volume      = {in press},
+number      = {},
+pages       = {},
+doi         = {10.1016/j.neuroimage.2010.12.039},
+keyword     = {DCM}
+} 
+
+@Article{meeg_spm8,
+author      = vl # and # jm # and # sjk # and # cp # and # rnah # and # jk # and # gb # and # {R. Oostenveld} # and # jean # and # gf #and # will # and # karl,
+title       = {EEG and MEG data analysis in SPM8},
+journal     = {Comput. Intell. Neurosci.},
+year        = {2010},
+volume      = {in press},
+number      = {},
+pages       = {},
+doi         = {},
+keyword     = {introduction, EEG, MEG}
+} 
+
+@Article{jean_oto1,
+author      = jean # and # {H.E.M den Ouden} # and # {M. Pessiglione} # and # klaas # and # sjk # and # karl,
+title       = {Observing the observer (I): meta-Bayesian models of learning and decision-making},
+journal     = {Plos One},
+year        = {2010},
+volume      = {5},
+number      = {12},
+pages       = {e15554},
+doi         = {10.1371/journal.pone.0015554},
+pdf         = {/spm/doc/papers/Daunizeau_PLOSOne_5_e15554.pdf},
+keyword     = {Bayesian}
+} 
+
+@Article{jean_oto2,
+author      = jean # and # {H.E.M den Ouden} # and # {M. Pessiglione} # and # sjk # and # karl # and # klaas,
+title       = {Observing the observer (II): deciding when to decide},
+journal     = {Plos One},
+year        = {2010},
+volume      = {5},
+number      = {12},
+pages       = {e15555},
+doi         = {10.1371/journal.pone.0015555},
+pdf         = {/spm/doc/papers/Daunizeau_PLOSOne_5_e15555.pdf},
+keyword     = {Bayesian}
+} 
+
+@Article{maria_eeg_fmri_integ,
+author      =  mjr # and # jean # and # karl,
+title       = {EEG-fMRI integration : a critical review of biophysical modelling and data analysis approaches},
+journal     = {J. Integrative Neurosci.},
+year        = {2010},
+volume      = {9},
+number      = {4},
+pages       = {453--476},
+doi         = {},
+keyword     = {}
+} 
+
+@Article{james_topological_inference,
+author      =  jk # and # karl,
+title       = {Topological Inference for {EEG} and {MEG}},
+journal     = {The Annals of Applied Statistics},
+year        = {2010},
+volume      = {4},
+number      = {3},
+pages       = {1272--1290},
+doi         = {10.1214/10-AOAS337},
+pdf         = {/spm/doc/papers/1011.2901v1.pdf},
+keyword     = {RFT, topological, EEG, MEG}
+} 
+
+
+@Article{karl_action_behaviour,
+author      = karl # and # jean # and # jk # and # sjk,
+title       = {Action and behavior: a free-energy formulation},
+journal     = {Biol. Cybern.},
+year        = {2010},
+volume      = {102},
+number      = {3},
+pages       = {227--260},
+doi         = {10.1007/s00422-010-0364-z},
+pdf         = {/spm/doc/papers/Action_and_behavior_A_free-energy_formulation.pdf},
+keyword     = {DEM}
+} 
+
+@Article{harriet_attention_uncertainty,
+author      = {Harriet Feldman} # and # karl,
+title       = {Attention, uncertainty, and free-energy},
+journal     = {Frontiers in Human Neuroscience},
+year        = {2010},
+volume      = {4},
+number      = {215},
+pages       = {},
+doi         = {10.3389/fnhum.2010.00215},
+pdf         = {/spm/doc/papers/Attention_uncertainty_and_free-energy.pdf},
+keyword     = {DEM}
+} 
+
+@Article{karl_cortical_circuits,
+author      = karl # and # sjk,
+title       = {Cortical circuits for perceptual inference},
+journal     = {Neural Networks},
+year        = {2009},
+volume      = {22},
+number      = {},
+pages       = {1093--1104},
+doi         = {10.1016/j.neunet.2009.07.023},
+pdf         = {/spm/doc/papers/Cortical_circuits_for_perceptual_inference.pdf},
+keyword     = {DEM}
+} 
+
+@Article{karl_variational_filtering,
+  author    = karl,
+  title     = {Variational filtering},
+  journal   = ni,
+  year      = 2008,
+  volume    = 41,
+  number    = {3},
+  pages     = {747--766},
+  doi       = {10.1016/j.neuroimage.2008.03.017},
+  pdf       = {/spm/doc/papers/Variational_filtering.pdf},
+  keyword   = {DEM}
+}
+
+@Article{karl_DEM,
+  author    = karl # and # nelson # and # jean,
+  title     = {{DEM}: A variational treatment of dynamic systems},
+  journal   = ni,
+  year      = 2008,
+  volume    = 41,
+  number    = {3},
+  pages     = {849--885},
+  doi       = {10.1016/j.neuroimage.2008.02.054},
+  pdf       = {/spm/doc/papers/DEM_A_variational_treatment_of_dynamic_systems.pdf},
+  keyword   = {DEM}
+}
+
 @Article{karl_generalised_filtering,
 author      = karl # and # klaas # and # bl # and # jean,
 title       = {Generalised Filtering},
@@ -3237,5 +3355,32 @@ @Article{karl_generalised_filtering
 pages       = {},
 doi         = {10.1155/2010/621670},
 pdf         = {/spm/doc/papers/Friston_MathProblEngin_2010_621670.pdf},
-keyword     = {}
+keyword     = {DEM}
+} 
+
+
+@Article{karl_hierarchical_models,
+author      = karl,
+title       = {Hierarchical Models in the Brain},
+journal     = {PLOS Computational Biology},
+year        = {2008},
+volume      = {4},
+number      = {11},
+pages       = {e1000211},
+doi         = {10.1371/journal.pcbi.1000211},
+pdf         = {/spm/doc/papers/Hierarchical_Models_in_the_Brain.pdf},
+keyword     = {DEM}
+} 
+
+@Article{karl_reinforcement_learning,
+author      =  karl # and # jean # and # sjk,
+title       = {Reinforcement Learning or Active Inference?},
+journal     = {PLOS One},
+year        = {2009},
+volume      = {4},
+number      = {7},
+pages       = {e6421},
+doi         = {10.1371/journal.pone.0006421},
+pdf         = {/spm/doc/papers/Reinforcement_Learning_or_Active_Inference.pdf},
+keyword     = {DEM}
 } 
diff --git a/man/dcm/Fig9.png b/man/dcm/Fig9.png
index 79bbfde..1f5114b 100644
Binary files a/man/dcm/Fig9.png and b/man/dcm/Fig9.png differ
diff --git a/man/dcm/dcm.tex b/man/dcm/dcm.tex
index bc7e354..09dc14e 100644
--- a/man/dcm/dcm.tex
+++ b/man/dcm/dcm.tex
@@ -288,8 +288,13 @@ \subsection{Comparing models}
 \begin{enumerate}
  \item Press the ``DCM'' button.
  \item Choose \textsc{Compare}.
- \item Number of models to compare: 2
- \item Select the two models, e.g. in the order \texttt{DCM\_mod\_bwd.mat} and \texttt{DCM\_mod\_fwd.mat}.
+ \item In the Batch Editor window that opened, fill in the ``BMS: DCM'' module:
+ \begin{enumerate}
+   \item \texttt{Directory}: choose current directory,
+   \item \texttt{Data}: add a New Subject with a New Session and select the two models, e.g. in the order \texttt{DCM\_mod\_bwd.mat} and \texttt{DCM\_mod\_fwd.mat},
+   \item \texttt{Inference method}: choose ``Fixed effects (FFX)''.
+ \end{enumerate}
+ \item Press Run (the green triangle in the Batch Editor).
 \end{enumerate}
 
 The Graphics window, Fig.~\ref{dcm_fig9}, now shows a bar plot of the model evidence. You can see that our second model is better than the first one.
diff --git a/man/dcm_ssr/fig1.png b/man/dcm_ssr/fig1.png
index 36116ff..aa92e34 100644
Binary files a/man/dcm_ssr/fig1.png and b/man/dcm_ssr/fig1.png differ
diff --git a/man/dcm_ssr/fig2.png b/man/dcm_ssr/fig2.png
index cef3d19..a511e6f 100644
Binary files a/man/dcm_ssr/fig2.png and b/man/dcm_ssr/fig2.png differ
diff --git a/man/dcm_ssr/fig3.png b/man/dcm_ssr/fig3.png
index 42976ad..d104b83 100644
Binary files a/man/dcm_ssr/fig3.png and b/man/dcm_ssr/fig3.png differ
diff --git a/man/example_scripts/simmegdip_for_faces.m b/man/example_scripts/simmegdip_for_faces.m
index 2eb6bfe..825d334 100644
--- a/man/example_scripts/simmegdip_for_faces.m
+++ b/man/example_scripts/simmegdip_for_faces.m
@@ -37,8 +37,8 @@
 allmeshvert_mni=D.inv{1}.mesh.tess_mni.vert;
 allmeshvert_ctf=D.inv{1}.forward.mesh.vert;
 allmeshfaces=D.inv{1}.forward.mesh.face;
-allmeshnorms_ctf=spm_eeg_inv_normals(allmeshvert_ctf,allmeshfaces);
-allmeshnorms_mni=spm_eeg_inv_normals(allmeshvert_mni,allmeshfaces);
+allmeshnorms_ctf=spm_mesh_normals(struct('faces',allmeshfaces,'vertices',allmeshvert_ctf),true);
+allmeshnorms_mni=spm_mesh_normals(struct('faces',allmeshfaces,'vertices',allmeshvert_mni),true);
 
 %%% FORCE  A SINGE SPHERE HEAD MODEL- simpler to compare inversions
 headmodels = {'Single Sphere', 'MEG Local Spheres', 'Single Shell'};
diff --git a/man/example_scripts/simmegdip_mult_subj.m b/man/example_scripts/simmegdip_mult_subj.m
index 18fdf76..90feb65 100644
--- a/man/example_scripts/simmegdip_mult_subj.m
+++ b/man/example_scripts/simmegdip_mult_subj.m
@@ -2,17 +2,17 @@
 %% template, 
 %% can vary source std in amplitude and/or location across subjects
 
-cd C:\users\gbarnes\documents\mv_multsubj
+cd('D:\SPM MEEG 2010\Preprocessing demo\MEG');
 
 Nsubj=10;
 %% ORIGINAL FACES DATA FILE
 origspmfilename='cdbespm8_SPM_CTF_MEG_example_faces1_3D.mat'
-outpath='C:\users\gbarnes\documents\simdata_vlad';
+outpath='D:\Data\Scaling\simulated';
 %% FOR SIMULATED NOISE- recording BW of 80Hz, white noise level 10ft/rtHz
-noiselevel=10*1e-15;
+noiselevel=50*1e-15;
 BW=80; 
-cross_subj_disp=0; %% sd of positions across subjects mm
-cross_subj_sdamp_pc=0; %% %sd of amplitude variation across subjects
+cross_subj_disp=5; %% sd of positions across subjects mm
+cross_subj_sdamp_pc=100; %% %sd of amplitude variation across subjects
 
 %% set frequency and amplitude of the two dipoles defined above 
 %% for each condition (faces/scrambled) separately
@@ -35,7 +35,7 @@
 
 
 for s=1:Nsubj,
-    dipamp_subj=dipamp+dipamp.*randn(size(dipamp))*cross_subj_sdamp_pc./100;;
+    dipamp_subj=dipamp+dipamp.*randn(size(dipamp))*cross_subj_sdamp_pc./100;
 
 %% OUTPUT SIMULATED DATA FILE
 spmfilename=[outpath filesep sprintf('%s%d',substr,s)];
@@ -60,8 +60,8 @@
 allmeshvert_mni=D.inv{1}.mesh.tess_mni.vert;
 allmeshvert_ctf=D.inv{1}.forward.mesh.vert;
 allmeshfaces=D.inv{1}.forward.mesh.face;
-allmeshnorms_ctf=spm_eeg_inv_normals(allmeshvert_ctf,allmeshfaces);
-allmeshnorms_mni=spm_eeg_inv_normals(allmeshvert_mni,allmeshfaces);
+allmeshnorms_ctf=spm_mesh_normals(struct('faces',allmeshfaces,'vertices',allmeshvert_ctf),true);
+allmeshnorms_mni=spm_mesh_normals(struct('faces',allmeshfaces,'vertices',allmeshvert_mni),true);
 
 %%% FORCE  A SINGE SPHERE HEAD MODEL- simpler to compare inversions
 headmodels = {'Single Sphere', 'MEG Local Spheres', 'Single Shell'};
diff --git a/man/faces/cat3_contrast.png b/man/faces/cat3_contrast.png
index dfeaa62..d419010 100644
Binary files a/man/faces/cat3_contrast.png and b/man/faces/cat3_contrast.png differ
diff --git a/man/faces/cat3_psth.png b/man/faces/cat3_psth.png
index 1043ed1..808007d 100644
Binary files a/man/faces/cat3_psth.png and b/man/faces/cat3_psth.png differ
diff --git a/man/faces/cat5_volume.png b/man/faces/cat5_volume.png
index c32dd60..dfee307 100644
Binary files a/man/faces/cat5_volume.png and b/man/faces/cat5_volume.png differ
diff --git a/man/faces/cat_contrasts.png b/man/faces/cat_contrasts.png
index 43a9476..95d2449 100644
Binary files a/man/faces/cat_contrasts.png and b/man/faces/cat_contrasts.png differ
diff --git a/man/faces/cat_design.png b/man/faces/cat_design.png
index 87df04e..0d2a48b 100644
Binary files a/man/faces/cat_design.png and b/man/faces/cat_design.png differ
diff --git a/man/faces/cat_explore.png b/man/faces/cat_explore.png
index 6a4656a..1318564 100644
Binary files a/man/faces/cat_explore.png and b/man/faces/cat_explore.png differ
diff --git a/man/faces/command.png b/man/faces/command.png
index b5078a6..bf0f4a9 100644
Binary files a/man/faces/command.png and b/man/faces/command.png differ
diff --git a/man/faces/coreg.png b/man/faces/coreg.png
index 9137389..7ef0ce4 100644
Binary files a/man/faces/coreg.png and b/man/faces/coreg.png differ
diff --git a/man/faces/dropout.png b/man/faces/dropout.png
index 9a33350..22fbc5c 100644
Binary files a/man/faces/dropout.png and b/man/faces/dropout.png differ
diff --git a/man/faces/face_ar1.png b/man/faces/face_ar1.png
index 05e16c6..3fadc02 100644
Binary files a/man/faces/face_ar1.png and b/man/faces/face_ar1.png differ
diff --git a/man/faces/face_bayes.png b/man/faces/face_bayes.png
index 8195fdb..9c560d2 100644
Binary files a/man/faces/face_bayes.png and b/man/faces/face_bayes.png differ
diff --git a/man/faces/face_stim.png b/man/faces/face_stim.png
index 65ab89f..15938e9 100644
Binary files a/man/faces/face_stim.png and b/man/faces/face_stim.png differ
diff --git a/man/faces/face_timing.png b/man/faces/face_timing.png
index f491370..ba7e742 100644
Binary files a/man/faces/face_timing.png and b/man/faces/face_timing.png differ
diff --git a/man/faces/famous_lag_mip.png b/man/faces/famous_lag_mip.png
index 12bbb3d..e50c5f6 100644
Binary files a/man/faces/famous_lag_mip.png and b/man/faces/famous_lag_mip.png differ
diff --git a/man/faces/gray.png b/man/faces/gray.png
index b1c26fe..12aed78 100644
Binary files a/man/faces/gray.png and b/man/faces/gray.png differ
diff --git a/man/faces/movements.png b/man/faces/movements.png
index 2792100..44e40ef 100644
Binary files a/man/faces/movements.png and b/man/faces/movements.png differ
diff --git a/man/faces/par_design.png b/man/faces/par_design.png
index a9832c5..5c0094d 100644
Binary files a/man/faces/par_design.png and b/man/faces/par_design.png differ
diff --git a/man/faces/realign.png b/man/faces/realign.png
index 590e760..42b0ca4 100644
Binary files a/man/faces/realign.png and b/man/faces/realign.png differ
diff --git a/man/faces/smooth.png b/man/faces/smooth.png
index b859700..2dce12a 100644
Binary files a/man/faces/smooth.png and b/man/faces/smooth.png differ
diff --git a/man/faces_group/Fdisp.png b/man/faces_group/Fdisp.png
index 9c6091d..46965db 100644
Binary files a/man/faces_group/Fdisp.png and b/man/faces_group/Fdisp.png differ
diff --git a/man/faces_group/Ftemp.png b/man/faces_group/Ftemp.png
index 179322b..e68b3b9 100644
Binary files a/man/faces_group/Ftemp.png and b/man/faces_group/Ftemp.png differ
diff --git a/man/faces_group/can_weighted_fir.png b/man/faces_group/can_weighted_fir.png
index 29ad3c9..0970dd4 100644
Binary files a/man/faces_group/can_weighted_fir.png and b/man/faces_group/can_weighted_fir.png differ
diff --git a/man/faces_group/f1_res.png b/man/faces_group/f1_res.png
index e8f1550..06dbdd5 100644
Binary files a/man/faces_group/f1_res.png and b/man/faces_group/f1_res.png differ
diff --git a/man/faces_group/fir_covariance.png b/man/faces_group/fir_covariance.png
index 8195025..7629eb0 100644
Binary files a/man/faces_group/fir_covariance.png and b/man/faces_group/fir_covariance.png differ
diff --git a/man/faces_group/fir_design.png b/man/faces_group/fir_design.png
index 9aa9daf..59a485e 100644
Binary files a/man/faces_group/fir_design.png and b/man/faces_group/fir_design.png differ
diff --git a/man/faces_group/fir_results.png b/man/faces_group/fir_results.png
index 39b6e67..a8d8d62 100644
Binary files a/man/faces_group/fir_results.png and b/man/faces_group/fir_results.png differ
diff --git a/man/faces_group/informed_covariance.png b/man/faces_group/informed_covariance.png
index 068547a..fbc45cd 100644
Binary files a/man/faces_group/informed_covariance.png and b/man/faces_group/informed_covariance.png differ
diff --git a/man/faces_group/informed_design.png b/man/faces_group/informed_design.png
index 07ac9c6..14cb578 100644
Binary files a/man/faces_group/informed_design.png and b/man/faces_group/informed_design.png differ
diff --git a/man/faces_group/informed_plot.png b/man/faces_group/informed_plot.png
index 039ae65..b15bd32 100644
Binary files a/man/faces_group/informed_plot.png and b/man/faces_group/informed_plot.png differ
diff --git a/man/faces_group/informed_results.png b/man/faces_group/informed_results.png
index dfe6d21..92c1a32 100644
Binary files a/man/faces_group/informed_results.png and b/man/faces_group/informed_results.png differ
diff --git a/man/faces_group/informed_t.png b/man/faces_group/informed_t.png
index e84c926..cfd34a2 100644
Binary files a/man/faces_group/informed_t.png and b/man/faces_group/informed_t.png differ
diff --git a/man/faces_group/nullall.png b/man/faces_group/nullall.png
index 5fa9cea..ff3a4ff 100644
Binary files a/man/faces_group/nullall.png and b/man/faces_group/nullall.png differ
diff --git a/man/faces_group/nullcan.png b/man/faces_group/nullcan.png
index 5a5f5c3..1347d62 100644
Binary files a/man/faces_group/nullcan.png and b/man/faces_group/nullcan.png differ
diff --git a/man/faces_group/t1.png b/man/faces_group/t1.png
index f23221d..83af4ba 100644
Binary files a/man/faces_group/t1.png and b/man/faces_group/t1.png differ
diff --git a/man/fmri_est/basis.png b/man/fmri_est/basis.png
index dde8478..c4f61e9 100644
Binary files a/man/fmri_est/basis.png and b/man/fmri_est/basis.png differ
diff --git a/man/fmri_est/bayes_options.png b/man/fmri_est/bayes_options.png
index 56b1375..093b006 100644
Binary files a/man/fmri_est/bayes_options.png and b/man/fmri_est/bayes_options.png differ
diff --git a/man/fmri_est/est_method.png b/man/fmri_est/est_method.png
index 4a105c5..aa5076e 100644
Binary files a/man/fmri_est/est_method.png and b/man/fmri_est/est_method.png differ
diff --git a/man/fmri_spec/design.png b/man/fmri_spec/design.png
index a42aa71..518d905 100644
Binary files a/man/fmri_spec/design.png and b/man/fmri_spec/design.png differ
diff --git a/man/fmri_spec/fmri_model.png b/man/fmri_spec/fmri_model.png
index 57188a5..6d4d97e 100644
Binary files a/man/fmri_spec/fmri_model.png and b/man/fmri_spec/fmri_model.png differ
diff --git a/man/fmri_spec/reg4.png b/man/fmri_spec/reg4.png
index 46a44a2..4b7ddae 100644
Binary files a/man/fmri_spec/reg4.png and b/man/fmri_spec/reg4.png differ
diff --git a/man/manual.tex b/man/manual.tex
index a3d84a5..ef553d4 100644
--- a/man/manual.tex
+++ b/man/manual.tex
@@ -157,7 +157,6 @@ \part{Data sets and examples}
 \include{auditory/auditory}
 \include{faces/face}
 \include{faces_group/face_group}
-%\include{fiac}
 \include{pet/pet}
 \include{dcm/dcm}
 \include{ppi/ppi}
@@ -173,9 +172,12 @@ \part{Data sets and examples}
 %%%%%% Appendix %%%%%%%
 
 \part{Batch Interface}
+\label{Chap:batch_interface}
 
 \include{batch/batch}
 
+\include{batch/devel}
+
 %%%%%% BIBLIOGRAPHY %%%%%%
 
 \part{Bibliography}
diff --git a/man/meeg/eeg_review_buttons.png b/man/meeg/eeg_review_buttons.png
index 9910e6c..9f6af83 100644
Binary files a/man/meeg/eeg_review_buttons.png and b/man/meeg/eeg_review_buttons.png differ
diff --git a/man/mmn/3DSPM.png b/man/mmn/3DSPM.png
index 21f317b..d174c8f 100644
Binary files a/man/mmn/3DSPM.png and b/man/mmn/3DSPM.png differ
diff --git a/man/mmn/coreg.png b/man/mmn/coreg.png
index d120e2d..e155e87 100644
Binary files a/man/mmn/coreg.png and b/man/mmn/coreg.png differ
diff --git a/man/mmn/data_and_design.png b/man/mmn/data_and_design.png
index d516663..9545c35 100644
Binary files a/man/mmn/data_and_design.png and b/man/mmn/data_and_design.png differ
diff --git a/man/mmn/dcm_gui.png b/man/mmn/dcm_gui.png
index 367dded..b119a6b 100644
Binary files a/man/mmn/dcm_gui.png and b/man/mmn/dcm_gui.png differ
diff --git a/man/mmn/electro_model.png b/man/mmn/electro_model.png
index 036b9b0..6a1eaea 100644
Binary files a/man/mmn/electro_model.png and b/man/mmn/electro_model.png differ
diff --git a/man/mmn/erp_c23.png b/man/mmn/erp_c23.png
index be35554..991cfd6 100644
Binary files a/man/mmn/erp_c23.png and b/man/mmn/erp_c23.png differ
diff --git a/man/mmn/forward.png b/man/mmn/forward.png
index 43fc0e6..63ff5fe 100644
Binary files a/man/mmn/forward.png and b/man/mmn/forward.png differ
diff --git a/man/mmn/invert.png b/man/mmn/invert.png
index b0a4eb1..1f78e81 100644
Binary files a/man/mmn/invert.png and b/man/mmn/invert.png differ
diff --git a/man/mmn/mesh.png b/man/mmn/mesh.png
index 8a29393..1453e39 100644
Binary files a/man/mmn/mesh.png and b/man/mmn/mesh.png differ
diff --git a/man/mmn/neuronal_model.png b/man/mmn/neuronal_model.png
index a719516..a9679a7 100644
Binary files a/man/mmn/neuronal_model.png and b/man/mmn/neuronal_model.png differ
diff --git a/man/mmn/source_erps.png b/man/mmn/source_erps.png
index 1d05c77..30ef4ef 100644
Binary files a/man/mmn/source_erps.png and b/man/mmn/source_erps.png differ
diff --git a/man/mmn/source_gui.png b/man/mmn/source_gui.png
index 63d0244..c1a419b 100644
Binary files a/man/mmn/source_gui.png and b/man/mmn/source_gui.png differ
diff --git a/man/mmn/topo1.png b/man/mmn/topo1.png
index 17cec84..773aac1 100644
Binary files a/man/mmn/topo1.png and b/man/mmn/topo1.png differ
diff --git a/man/pet/all.png b/man/pet/all.png
index 91bad3e..d9336e4 100644
Binary files a/man/pet/all.png and b/man/pet/all.png differ
diff --git a/man/pet/all_results.png b/man/pet/all_results.png
index a2bc372..d800a39 100644
Binary files a/man/pet/all_results.png and b/man/pet/all_results.png differ
diff --git a/man/pet/cluster.png b/man/pet/cluster.png
index 48b5db2..9d44ed8 100644
Binary files a/man/pet/cluster.png and b/man/pet/cluster.png differ
diff --git a/man/pet/conj.png b/man/pet/conj.png
index e94c8a7..7f802dd 100644
Binary files a/man/pet/conj.png and b/man/pet/conj.png differ
diff --git a/man/pet/interactive.png b/man/pet/interactive.png
index e58a16a..9089f59 100644
Binary files a/man/pet/interactive.png and b/man/pet/interactive.png differ
diff --git a/man/pet/masked_con.png b/man/pet/masked_con.png
index 70c1601..fd741a9 100644
Binary files a/man/pet/masked_con.png and b/man/pet/masked_con.png differ
diff --git a/man/pet/masked_res.png b/man/pet/masked_res.png
index 5c3c5d5..d101432 100644
Binary files a/man/pet/masked_res.png and b/man/pet/masked_res.png differ
diff --git a/man/pet/multi_design.png b/man/pet/multi_design.png
index cc242a8..b3a6f52 100644
Binary files a/man/pet/multi_design.png and b/man/pet/multi_design.png differ
diff --git a/man/pet/sc_design.png b/man/pet/sc_design.png
index 387de64..603f020 100644
Binary files a/man/pet/sc_design.png and b/man/pet/sc_design.png differ
diff --git a/man/pet/single_design.png b/man/pet/single_design.png
index f7c810c..8b93c9f 100644
Binary files a/man/pet/single_design.png and b/man/pet/single_design.png differ
diff --git a/man/pet/st_design.png b/man/pet/st_design.png
index 81f0df3..d18960b 100644
Binary files a/man/pet/st_design.png and b/man/pet/st_design.png differ
diff --git a/man/pet/subject2.png b/man/pet/subject2.png
index 0dd7bdb..de59364 100644
Binary files a/man/pet/subject2.png and b/man/pet/subject2.png differ
diff --git a/man/pet/svc.png b/man/pet/svc.png
index ebfdb3c..30e890b 100644
Binary files a/man/pet/svc.png and b/man/pet/svc.png differ
diff --git a/man/pet/table.png b/man/pet/table.png
index 2512947..2f881b9 100644
Binary files a/man/pet/table.png and b/man/pet/table.png differ
diff --git a/man/pet/voi.png b/man/pet/voi.png
index 872f0e6..9c36cc0 100644
Binary files a/man/pet/voi.png and b/man/pet/voi.png differ
diff --git a/matlabbatch/@cfg_branch/cfg_branch.m b/matlabbatch/@cfg_branch/cfg_branch.m
index e1a28ab..205f031 100644
--- a/matlabbatch/@cfg_branch/cfg_branch.m
+++ b/matlabbatch/@cfg_branch/cfg_branch.m
@@ -50,9 +50,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_branch.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_branch.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -63,7 +63,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_intree'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_intree'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item','cfg_intree'});
             item  = class(sitem, myclass, gitem, cfg_intree);
diff --git a/matlabbatch/@cfg_branch/subsasgn.m b/matlabbatch/@cfg_branch/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_branch/subsasgn.m
+++ b/matlabbatch/@cfg_branch/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_branch/subsref.m b/matlabbatch/@cfg_branch/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_branch/subsref.m
+++ b/matlabbatch/@cfg_branch/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_choice/cfg_choice.m b/matlabbatch/@cfg_choice/cfg_choice.m
index 5a74817..0a7a012 100644
--- a/matlabbatch/@cfg_choice/cfg_choice.m
+++ b/matlabbatch/@cfg_choice/cfg_choice.m
@@ -46,9 +46,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_choice.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_choice.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -59,7 +59,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_intree'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_intree'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item','cfg_intree'});
             item  = class(sitem, myclass, gitem, cfg_intree);
diff --git a/matlabbatch/@cfg_choice/subsasgn.m b/matlabbatch/@cfg_choice/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_choice/subsasgn.m
+++ b/matlabbatch/@cfg_choice/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_choice/subsref.m b/matlabbatch/@cfg_choice/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_choice/subsref.m
+++ b/matlabbatch/@cfg_choice/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_const/cfg_const.m b/matlabbatch/@cfg_const/cfg_const.m
index 2198cb6..301bb8a 100644
--- a/matlabbatch/@cfg_const/cfg_const.m
+++ b/matlabbatch/@cfg_const/cfg_const.m
@@ -46,9 +46,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_const.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_const.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -59,7 +59,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_leaf'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_leaf'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item', 'cfg_leaf'});
             item  = class(sitem, myclass, gitem, cfg_leaf);
diff --git a/matlabbatch/@cfg_const/subsasgn.m b/matlabbatch/@cfg_const/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_const/subsasgn.m
+++ b/matlabbatch/@cfg_const/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_const/subsref.m b/matlabbatch/@cfg_const/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_const/subsref.m
+++ b/matlabbatch/@cfg_const/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_dep/add_to_source.m b/matlabbatch/@cfg_dep/add_to_source.m
index ecd8e2f..171fae3 100644
--- a/matlabbatch/@cfg_dep/add_to_source.m
+++ b/matlabbatch/@cfg_dep/add_to_source.m
@@ -10,16 +10,16 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: add_to_source.m 1716 2008-05-23 08:18:45Z volkmar $
+% $Id: add_to_source.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 1716 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 for k = 1:numel(tdeps)
     sitem = subsref(cj, tdeps(k).src_exbranch); % Source item to deal with
     if isempty(sitem.sdeps)
         sitem.sdeps = tdeps(k);
     else
-        sitem.sdeps = [sitem.sdeps(:) tdeps(k)];
+        sitem.sdeps = [sitem.sdeps(:)' tdeps(k)];
     end;
     cj = subsasgn(cj, tdeps(k).src_exbranch,sitem);
 end;
diff --git a/matlabbatch/@cfg_dep/subsasgn.m b/matlabbatch/@cfg_dep/subsasgn.m
index 3b46b57..c8c9cda 100644
--- a/matlabbatch/@cfg_dep/subsasgn.m
+++ b/matlabbatch/@cfg_dep/subsasgn.m
@@ -20,112 +20,109 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3792 2010-03-22 13:11:36Z volkmar $
+% $Id: subsasgn.m 4075 2010-10-04 07:36:23Z volkmar $
 
-rev = '$Rev: 3792 $'; %#ok
+rev = '$Rev: 4075 $'; %#ok
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(dep) == nargin-2
-                for k = 1:numel(dep)
-                    % input checks
-                    ok = true;
-                    switch subs(1).subs
-                        case {'tname','sname'}
-                            if ~ischar(varargin{k})
-                                cfg_message('matlabbatch:subsasgn:name', 'Value for field ''%s'' must be a string.', ...
-                                    subs(1).subs);
-                                ok = false;
-                            end
-                        case {'tgt_spec'}
-                            if isempty(varargin{k})
-                                varargin{k} = cfg_findspec;
-                            else
-                                ok = iscell(varargin{k});
-                                if ok
-                                    for l = 1:numel(varargin{k})
-                                        ok = isstruct(varargin{k}{l}) && numel(fieldnames(varargin{k}{l}))==2 ...
-                                              && all(isfield(varargin{k}{l},{'name', ...
-                                                            'value'}));
-                                        if ~ok
-                                            break;
-                                        end
-                                    end
-                                end
-                                if ~ok
-                                    cfg_message('matlabbatch:ok_subsasgn:tgt_spec', ...
-                                            'Target specification must be a cfg_findspec.');
-                                end
-                            end
-                        case subs_fields(dep),
-                            if isempty(varargin{k})
-                                varargin{k} = struct('type',{}, 'subs',{});
-                            elseif ~(isstruct(varargin{k}) && isfield(varargin{k},'type') && isfield(varargin{k},'subs'))
-                                cfg_message('matlabbatch:subsasgn:subs', ['Value for field ''%s'' must be a struct with' ...
-                                    ' fields ''type'' and ''subs''.'], subs(1).subs);
-                                ok = false;
-                            end
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn:unknownfield', 'Reference to unknown field ''%s''.', subs(1).subs);
-                    end
-                    if ok,
-                        dep(k).(subs(1).subs) = varargin{k};
-                    end
-                end
-            else
-                cfg_message('matlabbatch:subsasgn:numel', 'In an assignment  A.X = B, the number of elements in A and B must be the same.');
-            end
-        case {'()'},
-            if isempty(dep)
-                dep = cfg_dep;
-            end
-            dep(subs(1).subs{:}) = varargin{:};
-        case {'{}'},
-            cfg_message('matlabbatch:subsasgn:notcell', 'Cell content reference from non cell-array object.');
-        otherwise
-            cfg_message('matlabbatch:subsref:unknowntype', 'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+persistent my_cfg_dep
+sflag = false;
+if strcmpi(subs(1).type, '()')
+    % check array boundaries, extend if necessary
+    if (numel(subs(1).subs) == 1 && numel(dep) < max(subs(1).subs{1})) || ...
+            (numel(subs(1).subs) > ndims(dep)) || ...
+            (numel(subs(1).subs) == ndims(dep) && any(cellfun(@max, subs(1).subs) > size(dep)))
+        if isempty(my_cfg_dep)
+            my_cfg_dep = cfg_dep;
+        end
+        if isempty(dep)
+            dep = my_cfg_dep;
+        end
+        [dep(subs(1).subs{:})] = deal(my_cfg_dep);
+    end
+    if numel(subs) == 1
+        return;
+    else
+        % select referenced depects from input array, run subsasgn on them an
+        % put results back into input array
+        odep  = dep;
+        osubs = subs(1);
+        dep   = dep(subs(1).subs{:});
+        subs  = subs(2:end);
+        sflag = true;
     end
-    return;
 end
-
-%% Canonicalise and check multi-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) >= 2 && strcmp(subs(1).type,'.')
-    % Canonicalise subs, so that there are two levels within this class
-    subs = [struct('type','()','subs',{{':'}}) subs];
+if  strcmpi(subs(1).type, '.')
+    % field assignment
+    if numel(subs) > 1
+        % Only part of field value(s) assigned. Get old field value(s),
+        % assign values to it.
+        val = {dep.(subs(1).subs)};
+        if nargin == 3
+            % Same value to assign to all field values.
+            val = cellfun(@(cval)subsasgn(cval,subs(2:end),varargin{1}), val, 'UniformOutput', false);
+        else
+            % Different values to assign to each field value.
+            val = cellfun(@(cval,cin)subsasgn(cval,subs(2:end),cin), val, varargin, 'UniformOutput', false);
+        end
+    else
+        % Field values to assign
+        val = varargin;
+    end
+    % Assign to fields
+    [ok val] = valcheck(subs(1).subs, val);
+    if ok
+        if isempty(dep)
+            dep = repmat(cfg_dep,size(val));
+        end
+        [dep.(subs(1).subs)] = deal(val{:});
+    end
+else
+    cfg_message('matlabbatch:subsref:unknowntype', 'Bad subscript type ''%s''.',subs(1).type);
 end
-
-if ~strcmp(subs(1).type,'()')
-    cfg_message('matlabbatch:subsasgn:wrongsubs', 'Wrong subscript reference.');
+if sflag
+    odep(osubs.subs{:}) = dep;
+    dep = odep;
 end
 
-item1 = subsref(dep, subs(1));
-if nargin-2 ~= 1 && nargin-2 ~= numel(item1)
-    cfg_message('matlabbatch:subsasgn:numel', 'In an assignment  A.X = B, the number of elements in A and B must be the same.');
+function [ok val] = valcheck(subs, val)
+persistent local_mysubs_fields;
+if ~iscell(local_mysubs_fields)
+    local_mysubs_fields = mysubs_fields;
 end
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-for k = 1:numel(item1)
-    if numel(subs) == 2
-        if nargin-2 == 1
-            item1(k) = subsasgn(item1(k),subs(2),varargin{1});
-        else
-            item1(k) = subsasgn(item1(k),subs(2),varargin{k});
+ok = true;
+switch subs
+    case {'tname','sname'}
+        if ~all(cellfun(@ischar,val))
+            cfg_message('matlabbatch:subsasgn:name', 'Value for field ''%s'' must be a string.', ...
+                subs);
+            ok = false;
         end
-    else
-        val = subsref(item1(k), subs(2));
-        % explicitly use builtin subsasgn for subscripts into field values
-        if nargin-2 == 1
-            val = builtin('subsasgn',val,subs(3:end),varargin{1});
-        else
-            val = builtin('subsasgn',val,subs(3:end),varargin{k});
+    case {'tgt_spec'}
+        sel = cellfun(@isempty,val);
+        if any(sel)
+            [val{sel}] = deal(cfg_findspec);
         end
-        item1(k) = subsasgn(item1(k), subs(2), val);
-    end
+        ok = all(cellfun(@is_cfg_findspec,val));
+    case local_mysubs_fields,
+        sel = cellfun(@isempty,val);
+        if any(sel)
+            [val{sel}] = deal(struct('type',{}, 'subs',{}));
+        end
+        ok = all(cellfun(@(cval)(isstruct(cval) && all(isfield(cval,{'type','subs'}))),val));
+        if ~ok
+            cfg_message('matlabbatch:subsasgn:subs', ['Value for field ''%s'' must be a struct with' ...
+                ' fields ''type'' and ''subs''.'], subs);
+        end
+    otherwise
+        cfg_message('matlabbatch:subsasgn:unknownfield', 'Reference to unknown field ''%s''.', subs);
 end
-dep = subsasgn(dep, subs(1), item1);
-return;
+
+function ok = is_cfg_findspec(fs)
+ok = iscell(fs) && ...
+    all(cellfun(@(cs)(isstruct(cs) && ...
+    ((numel(fieldnames(cs))==1 && any(isfield(cs,{'name','value'}))) || ...
+    (numel(fieldnames(cs))==2 && all(isfield(cs,{'name','value'}))))),fs));
+if ~ok
+    cfg_message('matlabbatch:ok_subsasgn:tgt_spec', ...
+        'Target specification must be a cfg_findspec.');
+end
\ No newline at end of file
diff --git a/matlabbatch/@cfg_dep/subsref.m b/matlabbatch/@cfg_dep/subsref.m
index 89febf1..8b2e45e 100644
--- a/matlabbatch/@cfg_dep/subsref.m
+++ b/matlabbatch/@cfg_dep/subsref.m
@@ -20,43 +20,31 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 2512 2008-12-01 13:21:29Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 2512 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
-switch subs(1).type,
-    case {'.'},
-        if numel(subs) > 1 && numel(dep) > 1
-            cfg_message('matlabbatch:subsref:multiref', 'Field reference for multiple structure elements that is followed by more reference blocks is an error.');
-        end;
-        switch subs(1).subs
-            case subs_fields(dep),
-                val = cell(size(dep));
-                for k = 1:numel(dep)
-                    val{k} = dep(k).(subs(1).subs);
-                end;
-            otherwise
-                cfg_message('matlabbatch:subsref:unknownfield', 'Reference to unknown field ''%s''.', subs(1).subs);
-        end;
-    case {'()'},
-        if numel(subs(1).subs) == 1 % vectorise output
-            szi = numel(dep);
-        else % sub-index output
-            szi = size(dep);
-        end;
-        for k = 1:numel(szi)
-            if ischar(subs(1).subs{k}) && strcmp(subs(1).subs{k},':')
-                subs(1).subs{k} = 1:szi(k);
-            end;
-        end;        
-        val = {dep(subs(1).subs{:})};
-    case {'{}'}
-        cfg_message('matlabbatch:subsref:notcell', 'Cell content reference from non cell-array object.');
-    otherwise
-        cfg_message('matlabbatch:subsref:unknowntype', 'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+if strcmpi(subs(1).type, '()')
+    % select referenced objects from input array
+    dep = dep(subs(1).subs{:});
+    if numel(subs) == 1
+        % done, return selected objects
+        varargout{1} = dep;
+        return;
+    else
+        % continue, avoid recursion
+        subs = subs(2:end);
+    end
 end
-if numel(subs) > 1 % in this case, val has only one element, and subs(2:end) are indices into val{1}
-    val = {builtin('subsref', val{1}, subs(2:end))};
-end;
 
-varargout = val;
\ No newline at end of file
+if strcmpi(subs(1).type, '.')
+    % field reference
+    val = {dep.(subs(1).subs)};
+    if numel(subs) > 1
+        varargout = cellfun(@(cval)subsref(cval,subs(2:end)), val, 'UniformOutput', false);
+    else
+        varargout = val;
+    end
+else
+    cfg_message('matlabbatch:subsref:unknowntype', 'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end
\ No newline at end of file
diff --git a/matlabbatch/@cfg_entry/cfg_entry.m b/matlabbatch/@cfg_entry/cfg_entry.m
index ff5a7d9..9d8f0ea 100644
--- a/matlabbatch/@cfg_entry/cfg_entry.m
+++ b/matlabbatch/@cfg_entry/cfg_entry.m
@@ -90,9 +90,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_entry.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_entry.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -103,7 +103,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_leaf'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_leaf'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item', 'cfg_leaf'});
             item  = class(sitem, myclass, gitem, cfg_leaf);
diff --git a/matlabbatch/@cfg_entry/subsasgn.m b/matlabbatch/@cfg_entry/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_entry/subsasgn.m
+++ b/matlabbatch/@cfg_entry/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_entry/subsasgn_check.m b/matlabbatch/@cfg_entry/subsasgn_check.m
index 07757a5..33aa0cb 100644
--- a/matlabbatch/@cfg_entry/subsasgn_check.m
+++ b/matlabbatch/@cfg_entry/subsasgn_check.m
@@ -11,9 +11,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn_check.m 2512 2008-12-01 13:21:29Z volkmar $
+% $Id: subsasgn_check.m 4089 2010-10-11 11:53:34Z volkmar $
 
-rev = '$Rev: 2512 $'; %#ok
+rev = '$Rev: 4089 $'; %#ok
 
 sts = true;
 switch subs(1).subs
@@ -66,13 +66,10 @@
 end
 if isa(val,'cfg_dep')
     % Check dependency match
-    sts2 = false(size(val));
-    for k = 1:numel(val)
-        sts2(k) = match(item, val(k).tgt_spec);
-        if ~sts2(k)
-            cfg_message('matlabbatch:checkval', ...
-                        '%s: Dependency does not match.', subsasgn_checkstr(item,subs));
-        end
+    sts2 = cellfun(@(cspec)match(item,cspec),{val.tgt_spec});
+    if ~all(sts2)
+        cfg_message('matlabbatch:checkval', ...
+            '%s: Dependency does not match.', subsasgn_checkstr(item,subs));
     end
     val = val(sts2);
     sts = any(sts2);
diff --git a/matlabbatch/@cfg_entry/subsref.m b/matlabbatch/@cfg_entry/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_entry/subsref.m
+++ b/matlabbatch/@cfg_entry/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_exbranch/display.m b/matlabbatch/@cfg_exbranch/display.m
new file mode 100644
index 0000000..8461583
--- /dev/null
+++ b/matlabbatch/@cfg_exbranch/display.m
@@ -0,0 +1,21 @@
+function display(item)
+
+% function display(item)
+% Display a configuration object
+%
+% This code is part of a batch job configuration system for MATLAB. See 
+%      help matlabbatch
+% for a general overview.
+%_______________________________________________________________________
+% Copyright (C) 2007 Freiburg Brain Imaging
+
+% Volkmar Glauche
+% $Id: display.m 4166 2011-01-17 15:06:41Z volkmar $
+
+rev = '$Rev: 4166 $'; %#ok
+
+disp(' ');
+disp([inputname(1),' = ']);
+disp(' ');
+disp(item);
+disp(' ');
diff --git a/matlabbatch/@cfg_exbranch/subsasgn.m b/matlabbatch/@cfg_exbranch/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_exbranch/subsasgn.m
+++ b/matlabbatch/@cfg_exbranch/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_exbranch/subsref.m b/matlabbatch/@cfg_exbranch/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_exbranch/subsref.m
+++ b/matlabbatch/@cfg_exbranch/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_files/cfg_files.m b/matlabbatch/@cfg_files/cfg_files.m
index 1095681..7c9a1bf 100644
--- a/matlabbatch/@cfg_files/cfg_files.m
+++ b/matlabbatch/@cfg_files/cfg_files.m
@@ -69,9 +69,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_files.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_files.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -82,7 +82,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_leaf'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_leaf'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item', 'cfg_leaf'});
             item  = class(sitem, myclass, gitem, cfg_leaf);
diff --git a/matlabbatch/@cfg_files/subsasgn.m b/matlabbatch/@cfg_files/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_files/subsasgn.m
+++ b/matlabbatch/@cfg_files/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_files/subsasgn_check.m b/matlabbatch/@cfg_files/subsasgn_check.m
index fb68df7..2f931c1 100644
--- a/matlabbatch/@cfg_files/subsasgn_check.m
+++ b/matlabbatch/@cfg_files/subsasgn_check.m
@@ -12,9 +12,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn_check.m 2539 2008-12-09 11:19:55Z volkmar $
+% $Id: subsasgn_check.m 4089 2010-10-11 11:53:34Z volkmar $
 
-rev = '$Rev: 2539 $'; %#ok
+rev = '$Rev: 4089 $'; %#ok
 
 sts = true;
 switch subs(1).subs
@@ -62,14 +62,13 @@
                 end
                 val = {val1};
             elseif isa(val{1}, 'cfg_dep')
+                % {val{1}.tgt_spec} does not seem to work properly
+                cval = val{1};
                 % Check dependency match
-                sts2 = false(size(val{1}));
-                for k = 1:numel(val{1})
-                    sts2(k) = match(item, val{1}(k).tgt_spec);
-                    if ~sts2(k)
-                        cfg_message('matlabbatch:checkval', ...
-                                    '%s: Dependency does not match.', subsasgn_checkstr(item,subs));
-                    end
+                sts2 = cellfun(@(cspec)match(item,cspec),{cval.tgt_spec});
+                if ~all(sts2)
+                    cfg_message('matlabbatch:checkval', ...
+                        '%s: Dependency does not match.', subsasgn_checkstr(item,subs));
                 end
                 val{1} = val{1}(sts2);
                 sts = any(sts2);
diff --git a/matlabbatch/@cfg_files/subsref.m b/matlabbatch/@cfg_files/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_files/subsref.m
+++ b/matlabbatch/@cfg_files/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_item/subsasgn.m b/matlabbatch/@cfg_item/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_item/subsasgn.m
+++ b/matlabbatch/@cfg_item/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_item/subsref.m b/matlabbatch/@cfg_item/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_item/subsref.m
+++ b/matlabbatch/@cfg_item/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_menu/cfg_menu.m b/matlabbatch/@cfg_menu/cfg_menu.m
index 9802deb..010a47e 100644
--- a/matlabbatch/@cfg_menu/cfg_menu.m
+++ b/matlabbatch/@cfg_menu/cfg_menu.m
@@ -46,9 +46,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_menu.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_menu.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -59,7 +59,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_leaf'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_leaf'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item', 'cfg_leaf'});
             item  = class(sitem, myclass, gitem, cfg_leaf);
diff --git a/matlabbatch/@cfg_menu/subsasgn.m b/matlabbatch/@cfg_menu/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_menu/subsasgn.m
+++ b/matlabbatch/@cfg_menu/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_menu/subsref.m b/matlabbatch/@cfg_menu/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_menu/subsref.m
+++ b/matlabbatch/@cfg_menu/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/@cfg_repeat/cfg_repeat.m b/matlabbatch/@cfg_repeat/cfg_repeat.m
index 9b3ad4c..25923c7 100644
--- a/matlabbatch/@cfg_repeat/cfg_repeat.m
+++ b/matlabbatch/@cfg_repeat/cfg_repeat.m
@@ -59,9 +59,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_repeat.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_repeat.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
 myclass = mfilename;
 % Get local fields and defaults from private/mysubs_fields
@@ -72,7 +72,7 @@
         % assume input is a struct to be converted back into a class
         % return with error if this does not work
         if numel(fieldnames(varargin{1})) == numel(fn)+2 && ...
-                all(isfield(varargin{1}, [fn(:)' {'cfg_item' 'cfg_intree'}]))
+                all(isfield(varargin{1}, [fn(:)', {'cfg_item' 'cfg_intree'}]))
             gitem = varargin{1}.cfg_item;
             sitem = rmfield(varargin{1},{'cfg_item','cfg_intree'});
             item  = class(sitem, myclass, gitem, cfg_intree);
diff --git a/matlabbatch/@cfg_repeat/initialise.m b/matlabbatch/@cfg_repeat/initialise.m
index c84fc69..9c6c005 100644
--- a/matlabbatch/@cfg_repeat/initialise.m
+++ b/matlabbatch/@cfg_repeat/initialise.m
@@ -21,9 +21,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: initialise.m 3591 2009-11-23 10:19:57Z volkmar $
+% $Id: initialise.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3591 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 if strcmp(val,'<DEFAULTS>')
     item = initialise_def(item, val, dflag);
@@ -51,9 +51,11 @@
         if dflag
             item.values{1} = initialise(item.values{1}, val, dflag);
         else
+            citem = cell(1,numel(val));
             for k = 1:numel(val)
-                item.cfg_item.val{k} = initialise(item.values{1}, val(k), dflag);
+                citem{k} = initialise(item.values{1}, val(k), dflag);
             end;
+            item.cfg_item.val = citem;
         end;
     else
         cfg_message('matlabbatch:initialise', ...
@@ -99,13 +101,14 @@
             item.values{1} = initialise(item.values{1}, val{1}, dflag);
         end;
     else
+        citem = cell(1,numel(val));
         if numel(item.values) > 1 || item.forcestruct
             for l = 1:numel(val)
                 % val{l} should be a struct with a single field
                 vtag = fieldnames(val{l});
                 for k = 1:numel(item.values)
                     if strcmp(gettag(item.values{k}), vtag{1})
-                        item.cfg_item.val{l} = initialise(item.values{k}, ...
+                        citem{l} = initialise(item.values{k}, ...
                             val{l}.(vtag{1}), ...
                             dflag);
                     end;
@@ -113,10 +116,11 @@
             end;
         else
             for l = 1:numel(val)
-                item.cfg_item.val{l} = initialise(item.values{1}, ...
+                citem{l} = initialise(item.values{1}, ...
                     val{l}, dflag);
             end;
         end;
+        item.cfg_item.val = citem;
     end;
 end;
 
diff --git a/matlabbatch/@cfg_repeat/subsasgn.m b/matlabbatch/@cfg_repeat/subsasgn.m
index 4424009..fdf523d 100644
--- a/matlabbatch/@cfg_repeat/subsasgn.m
+++ b/matlabbatch/@cfg_repeat/subsasgn.m
@@ -32,9 +32,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsasgn.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsasgn.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -61,80 +61,48 @@
           'Arrays of cfg_item objects not supported.');
 end;
 
-%% One-level subscripts
-%--------------------------------------------------------------------------
-if numel(subs) == 1
-    switch subs(1).type,
-        case {'.'},
-            if numel(item) == 1 && nargin == 3
-                [ok val] = subsasgn_check(item,subs,varargin{1});
-                if ok,
-                    switch subs(1).subs
-                        case par_fields,
-                            item.(par_class) = subsasgn(item.(par_class), subs, val);
-                        case local_mysubs_fields,
-                            item.(subs(1).subs) = val;
-                        otherwise
-                            cfg_message('matlabbatch:subsasgn', ...
-                                  ['Reference to unknown field ''%s''.\n' ...
-                                   'To assign to a field in the job structure, use a reference like ' ...
-                                   '''(x).%s''.'], subs(1).subs, subs(1).subs);
-                    end;
-                end;
+switch subs(1).type,
+    case {'.'},
+        if numel(item) == 1 && nargin == 3
+            if numel(subs) > 1
+                % Only part of field value(s) assigned. Get old field value(s),
+                % assign values to it.
+                if any(strcmp(subs(1).subs,par_fields))
+                    val0 = item.(par_class).(subs(1).subs);
+                else
+                    val0 = item.(subs(1).subs);
+                end
+                val1 = cfg_callbuiltin('subsasgn',val0,subs(2:end),varargin{1});
             else
-                cfg_message('matlabbatch:subsasgn', ...
-                      ['Array assignments not supported for fields of cfg_item objects.\n' ...
-                       'To assign to a field in the job structure, use a reference like ' ...
-                       '''(x).%s''.'], subs(1).subs);
-
+                % Field values to assign
+                val1 = varargin{1};
+            end
+            [ok val] = subsasgn_check(item,subs,val1);
+            if ok,
+                switch subs(1).subs
+                    case par_fields,
+                        item.(par_class).(subs(1).subs) = val;
+                    case local_mysubs_fields,
+                        item.(subs(1).subs) = val;
+                    otherwise
+                        cfg_message('matlabbatch:subsasgn', ...
+                                    ['Reference to unknown field ''%s''.\n' ...
+                                     'To assign to a field in the job structure, use a reference like ' ...
+                                     '''(x).%s''.'], subs(1).subs, subs(1).subs);
+                end;
             end;
-        case {'()','{}'},
+        else
             cfg_message('matlabbatch:subsasgn', ...
-                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
-        otherwise
-            cfg_message('matlabbatch:subsasgn', ...
-                  'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
-    end;
-    return;
-end;
-
-%% Two- and multi-level subscripts
-%--------------------------------------------------------------------------
-item = subsasgn_rec(item, subs, varargin{1});
-return;
+                        ['Array assignments not supported for fields of cfg_item objects.\n' ...
+                         'To assign to a field in the job structure, use a reference like ' ...
+                         '''(x).%s''.'], subs(1).subs);
 
-function it = subsasgn_rec(it, subs, val)
-if numel(subs) == 1
-    % final subscript
-    try
-        it = builtin('subsasgn', it, subs, val);
-    catch
-        % there may be an object method that is missed by builtin
-        it = subsasgn(it, subs, val);
-    end;
-else
-    try
-        it1 = subsref(it, subs(1));
-    catch
-        % it does not yet contain contents at this subscript
-        % builtin subsasgn will be clever enough to create the necessary
-        % struct/fields that are referenced
-        it = builtin('subsasgn', it, subs, val);
-        return;
-    end;
-    it1 = subsasgn_rec(it1, subs(2:end), val);
-    %    try
-        % do we need this?
-        %    it = builtin('subsasgn', it, subs(1), it1);
-        %catch
-    if isobject(it)
-        try
-            % try to use class subsasgn
-            it = subsasgn(it,subs(1), it1);
-        catch
-            it = builtin('subsasgn', it, subs(1), it1);
         end;
-    else
-        it = builtin('subsasgn', it, subs(1), it1);
-    end;
-end;
\ No newline at end of file
+    case {'()','{}'},
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Subscript type ''%s'' reserved for future use.', subs(1).type);
+    otherwise
+        cfg_message('matlabbatch:subsasgn', ...
+                    'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
+end;
+return;
\ No newline at end of file
diff --git a/matlabbatch/@cfg_repeat/subsref.m b/matlabbatch/@cfg_repeat/subsref.m
index 02c14bc..945a1be 100644
--- a/matlabbatch/@cfg_repeat/subsref.m
+++ b/matlabbatch/@cfg_repeat/subsref.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: subsref.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: subsref.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 persistent local_mysubs_fields;
 persistent par_class;
@@ -55,30 +55,23 @@
         end;
         switch subs(1).subs
             case local_mysubs_fields,
-                val = {item.(subs(1).subs)};
+                val = item.(subs(1).subs);
             case par_fields,
-                val = {item.(par_class).(subs(1).subs)};
+                val = item.(par_class).(subs(1).subs);
             otherwise
                 cfg_message('matlabbatch:subsref:unknownfield', ...
                       ['Reference to unknown field ''%s''.\nTo reference ' ...
                        'a field in the job structure, use a reference like ' ...
                        '''(x).%s'''], subs(1).subs, subs(1).subs);
         end;
-        if numel(subs) > 1 
-            % in this case, val has only one element, and
-            % subs(2:end) are indices into val{1} 
-            %    val = {builtin('subsref', val{1}, subs(2:end))};
-            % The line above does not seem to work, as MATLAB is
-            % not able to figure out which subsref to call and when.
-            for k = 2:numel(subs)
-                val = {subsref(val{1}, subs(k))};
-            end;
-        end;
-    case {'()','{}'},
-        val = subsref_job(item, subs, false);
+        if numel(subs) > 1
+            val = cfg_callbuiltin('subsref',val,subs(2:end));
+        end
+%    case {'()','{}'},
+%        val = subsref_job(item, subs, false);
     otherwise
         cfg_message('matlabbatch:subsref:unknowntype', ...
               'Unknown subsref type: ''%s''. This should not happen.', subs(1).type);
 end
 
-varargout = val;
\ No newline at end of file
+varargout = {val};
\ No newline at end of file
diff --git a/matlabbatch/cfg_basicio/cfg_run_file_move.m b/matlabbatch/cfg_basicio/cfg_run_file_move.m
index 4f7e8af..a5ff642 100644
--- a/matlabbatch/cfg_basicio/cfg_run_file_move.m
+++ b/matlabbatch/cfg_basicio/cfg_run_file_move.m
@@ -11,23 +11,30 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_run_file_move.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_run_file_move.m 4102 2010-10-28 10:47:15Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4102 $'; %#ok
 
 action = fieldnames(job.action);
 action = action{1};
 if strcmp(action, 'delete')
+    todelete = {};
     for k = 1:numel(job.files)
         [p n e] = fileparts(job.files{k});
         if numel(e)>=4 && any(strcmp(e(1:4), {'.nii','.img'}))
             try
                 [p n e v] = spm_fileparts(job.files{k});
-                delete(fullfile(p,[n '.hdr']));
-                delete(fullfile(p,[n '.mat']));
+                todelete{end+1} = fullfile(p,[n '.hdr']);
+                todelete{end+1} = fullfile(p,[n '.mat']);
             end
         end
-        delete(fullfile(p, [n e]));
+        todelete{end+1} = fullfile(p, [n e]);
+    end
+    if ~isempty(todelete)
+        ws = warning;
+        warning off;
+        delete(todelete{:});
+        warning(ws);
     end
     out = [];
 else
diff --git a/matlabbatch/cfg_callbuiltin.m b/matlabbatch/cfg_callbuiltin.m
new file mode 100644
index 0000000..3e8545c
--- /dev/null
+++ b/matlabbatch/cfg_callbuiltin.m
@@ -0,0 +1,11 @@
+function varargout = cfg_callbuiltin(varargin)
+if nargout == 0
+    try
+        varargout{1} = builtin(varargin{:});
+    catch
+        builtin(varargin{:});
+    end
+else
+    varargout = cell(1,nargout);
+    [varargout{:}] = builtin(varargin{:});
+end
diff --git a/matlabbatch/cfg_getfile.m b/matlabbatch/cfg_getfile.m
index bc9a898..3d14d91 100644
--- a/matlabbatch/cfg_getfile.m
+++ b/matlabbatch/cfg_getfile.m
@@ -85,7 +85,7 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % John Ashburner and Volkmar Glauche
-% $Id: cfg_getfile.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_getfile.m 4176 2011-01-27 14:34:44Z volkmar $
 
 t = {};
 sts = false;
@@ -255,9 +255,8 @@
 if ~isempty(fg)
     delete(fg);
 end
-% create temporary figure to work out position etc.
-% seems to be necessary 
-fgtmp = figure('IntegerHandle','off',...
+% create figure
+fg = figure('IntegerHandle','off',...
     'Tag',mfilename,...
     'Name',mesg,...
     'NumberTitle','off',...
@@ -267,30 +266,8 @@
     'DefaultUicontrolInterruptible','on',...
     'Visible','off');
     
-Rect = get(fgtmp,'Position');
-delete(fgtmp);
-%S0 = spm('WinSize','0',1);
-S0   = get(0,'MonitorPosition');
-if size(S0,1) > 1 % Multiple Monitors
-    %-Use Monitor containing the Pointer
-    pl = get(0,'PointerLocation');
-    w  = find(pl(1)>=S0(:,1) & pl(1)<S0(:,1)+S0(:,3)-1 &...
-            pl(2)>=S0(:,2) & pl(2)<S0(:,2)+S0(:,4));
-    if numel(w)~=1, w = 1; end
-    S0 = S0(w,:);
-end
-Rect(1) = S0(1) + (S0(3) - Rect(3))/2;
-Rect(2) = S0(2) + (S0(4) - Rect(4))/2;
-% create final figure
-fg = figure('IntegerHandle','off',...
-    'Tag',mfilename,...
-    'Name',mesg,...
-    'NumberTitle','off',...
-    'Units','Pixels',...
-    'MenuBar','none',...
-    'DefaultTextInterpreter','none',...
-    'DefaultUicontrolInterruptible','on',...
-    'Position',Rect);
+cfg_onscreen(fg);
+set(fg,'Visible','on');
 
 sellines = min([max([n(2) numel(already)]), 4]);
 [pselp pcntp pfdp pdirp] = panelpositions(fg, sellines+1);
@@ -568,9 +545,8 @@
     t  = get(sel,'String');
     if isempty(t)
         t = {''};
-    end
-    if sfilt.code == -1 && ~isempty(t)
-        % don't canonicalise empty selection
+    elseif sfilt.code == -1
+        % canonicalise non-empty folder selection
         t = cellfun(@(t1)cpath(t1, pwd), t, 'UniformOutput',false);
     end;
     ok = 1;
@@ -949,8 +925,12 @@ function unselect_all(ob,varargin)
 
 %=======================================================================
 function [f,d] = listfiles(dr,filt)
-ob = sib(gco,'msg');
-domsg = ~isempty(ob);
+try
+    ob = sib(gco,'msg');
+    domsg = ~isempty(ob);
+catch
+    domsg = false;
+end
 if domsg
     omsg = msg(ob,'Listing directory...');
 end
@@ -1094,7 +1074,7 @@ function heelp(ob,varargin)
 'You can use the right mouse button to un-select everything.'],...
 '',[...
 'Directories are navigated by editing the name of the current directory (where it says "Dir"), ',...
-'by going to one of the previously entered directories ("Prev"), '...
+'by going to one of the previously entered directories ("Prev"), ',...
 'by going to one of the parent directories ("Up") or by navigating around ',...
 'the parent or subdirectories listed in the left side panel.'],...
 '',[...
diff --git a/matlabbatch/cfg_load_jobs.m b/matlabbatch/cfg_load_jobs.m
index 72386b0..d06bb0b 100644
--- a/matlabbatch/cfg_load_jobs.m
+++ b/matlabbatch/cfg_load_jobs.m
@@ -1,9 +1,8 @@
-function newjobs = cfg_load_jobs(job)
+function [newjobs uind] = cfg_load_jobs(job)
 
 % function newjobs = cfg_load_jobs(job)
 %
-% Load a list of possible job files, return a cell list of jobs. If a job
-% file failed to load, an empty cell is returned in the list.
+% Load a list of possible job files, return a cell list of jobs.
 %
 % This code is part of a batch job configuration system for MATLAB. See 
 %      help matlabbatch
@@ -12,49 +11,59 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_load_jobs.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_load_jobs.m 4073 2010-09-24 12:07:57Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4073 $'; %#ok
 
 if ischar(job)
     filenames = cellstr(job);
 else
     filenames = job;
 end;
-newjobs = {};
-for cf = 1:numel(filenames)
-    [p,nam,ext] = fileparts(filenames{cf});
-    switch ext
-        case '.xml',
-            try
-                loadxml(filenames{cf},'matlabbatch');
-            catch
-                cfg_message('matlabbatch:initialise:xml','LoadXML failed: ''%s''',filenames{cf});
-            end;
-        case '.mat'
-            try
-                S=load(filenames{cf});
-                matlabbatch = S.matlabbatch;
-            catch
-                cfg_message('matlabbatch:initialise:mat','Load failed: ''%s''',filenames{cf});
-            end;
-        case '.m'
-            try
-                fid = fopen(filenames{cf},'rt');
-                str = fread(fid,'*char');
-                fclose(fid);
-                eval(str);
-            catch
-                cfg_message('matlabbatch:initialise:m','Eval failed: ''%s''',filenames{cf});
-            end;
-            if ~exist('matlabbatch','var')
-                cfg_message('matlabbatch:initialise:m','No matlabbatch job found in ''%s''', filenames{cf});
-            end;
-        otherwise
-            cfg_message('matlabbatch:initialise:unknown','Unknown extension: ''%s''', filenames{cf});
-    end;
-    if exist('matlabbatch','var')
-        newjobs = [newjobs(:); {matlabbatch}];
-        clear matlabbatch;
-    end;
+[ufilenames unused uind] = unique(filenames);
+ujobs = cell(size(ufilenames));
+usts  = false(size(ufilenames));
+for cf = 1:numel(ufilenames)
+    [ujobs{cf} usts(cf)] = load_single_job(ufilenames{cf});
+end
+sts   = usts(uind);
+uind  = uind(sts);
+newjobs = ujobs(uind);
+
+function [matlabbatch sts] = load_single_job(filename)
+[p,nam,ext] = fileparts(filename);
+switch ext
+    case '.xml',
+        try
+            loadxml(filename,'matlabbatch');
+        catch
+            cfg_message('matlabbatch:initialise:xml','LoadXML failed: ''%s''',filename);
+        end;
+    case '.mat'
+        try
+            S=load(filename);
+            matlabbatch = S.matlabbatch;
+        catch
+            cfg_message('matlabbatch:initialise:mat','Load failed: ''%s''',filename);
+        end;
+    case '.m'
+        try
+            fid = fopen(filename,'rt');
+            str = fread(fid,'*char');
+            fclose(fid);
+            eval(str);
+        catch
+            cfg_message('matlabbatch:initialise:m','Eval failed: ''%s''',filename);
+        end;
+        if ~exist('matlabbatch','var')
+            cfg_message('matlabbatch:initialise:m','No matlabbatch job found in ''%s''', filename);
+        end;
+    otherwise
+        cfg_message('matlabbatch:initialise:unknown','Unknown extension: ''%s''', filename);
+end;
+if exist('matlabbatch','var')
+    sts = true;
+else
+    sts = false;
+    matlabbatch = [];
 end;
diff --git a/matlabbatch/cfg_tropts.m b/matlabbatch/cfg_tropts.m
index 9e0af81..a597765 100644
--- a/matlabbatch/cfg_tropts.m
+++ b/matlabbatch/cfg_tropts.m
@@ -1,6 +1,6 @@
 function tropts = cfg_tropts(stopspec, clvl, mlvl, cnt, mcnt, dflag)
 
-% function tropts = cfg_tropts(stopspec, clvl, mlvl, cnt, mcnt)
+% function tropts = cfg_tropts(stopspec, clvl, mlvl, cnt, mcnt, dflag)
 % This function is a shorthand that generates a traversal options structure
 % from the following items:
 % stopspec -   a find spec shorthand as input to cfg_findspec (see
@@ -16,9 +16,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_tropts.m 1716 2008-05-23 08:18:45Z volkmar $
+% $Id: cfg_tropts.m 4102 2010-10-28 10:47:15Z volkmar $
 
-rev = '$Rev: 1716 $'; %#ok
+rev = '$Rev: 4102 $'; %#ok
 
 if nargin == 0
     tropts = struct('stopspec', {}, 'clvl', {}, 'mlvl', {}, 'cnt', {}, 'mcnt', {}, 'dflag', {});
diff --git a/matlabbatch/cfg_ui.m b/matlabbatch/cfg_ui.m
index 6d641cd..c3071da 100644
--- a/matlabbatch/cfg_ui.m
+++ b/matlabbatch/cfg_ui.m
@@ -27,9 +27,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_ui.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_ui.m 4255 2011-03-18 13:11:03Z volkmar $
 
-rev = '$Rev: 3944 $'; %#ok
+rev = '$Rev: 4255 $'; %#ok
 
 % edit the above text to modify the response to help cfg_ui
 
@@ -254,7 +254,7 @@ function local_editdefs(varargin)
 udmodlist = get(handles.modlist, 'userdata');
 udmodlist(1).defid = id;
 udmodlist(1).cmod  = 1;
-set(handles.modlist, 'String',val{1}, 'Value',1, 'Userdata',udmodlist);
+set(handles.modlist, 'Value',1, 'ListboxTop',1, 'Userdata',udmodlist, 'String',val{1});
 local_showmod(gcbo);
 % --------------------------------------------------------------------
 function local_editdefsquit(varargin)
@@ -282,6 +282,8 @@ function local_showjob(obj,cjob)
     % set cjob, if supplied
     udmodlist = local_init_udmodlist;
     udmodlist(1).cjob = cjob;
+    % move figure onscreen
+    cfg_onscreen(obj);
     set(obj,'Visible','on');
 end;
 [id str sts dep sout] = cfg_util('showjob',cjob);
@@ -309,7 +311,8 @@ function local_showjob(obj,cjob)
     [mrk{~dep & sts}] = deal('');
     str = cfg_textfill(handles.modlist, str, mrk, false);
 end;
-set(handles.modlist, 'string', str, 'userdata',udmodlist, 'value', cmod);
+ltop = local_getListboxTop(handles.modlist, cmod, numel(str));
+set(handles.modlist, 'userdata',udmodlist, 'value', cmod, 'ListboxTop', ltop, 'string', str);
 if ~isempty(sts) && all(sts)
     set(findobj(handles.cfg_ui,'-regexp', 'Tag','.*File(Run)|(RunSerial)$'),'Enable','on');
 else
@@ -373,7 +376,7 @@ function local_showmod(obj)
                     case 'cfg_menu',
                         datastr{k} = 'Unknown selection';
                         for l = 1:numel(contents{4}{k})
-                            if isequal(contents{2}{k}{1}, contents{4}{k}{l})
+                            if isequalwithequalnans(contents{2}{k}{1}, contents{4}{k}{l})
                                 datastr{k} = contents{3}{k}{l};
                                 break;
                             end;
@@ -441,13 +444,14 @@ function local_showmod(obj)
     udmodule.contents = contents;
     udmodule.id = id;
     udmodule.oldvalue = citem;
-    set(handles.module, 'String', str, 'Value', citem, 'userdata', udmodule);
+    ltop = local_getListboxTop(handles.module, citem, numel(str));
+    set(handles.module, 'Value', citem, 'ListboxTop', ltop, 'userdata', udmodule, 'String', str);
     udmodlist(1).cmod = cmod;
     set(handles.modlist, 'userdata', udmodlist);
     local_showvaledit(obj);
     uicontrol(handles.module);
 else
-    set(handles.module, 'String','No Module selected', 'Value',1,'Userdata',[]);
+    set(handles.module, 'Value',1,'ListboxTop',1,'Userdata',[], 'String',{'No Module selected'});
     set(handles.moduleHead,'String','No Current Module');
     set(findobj(handles.cfg_ui,'-regexp','Tag','^Btn.*'), 'Visible', 'off');
     set(findobj(handles.cfg_ui,'-regexp','Tag','^MenuEditVal.*'), 'Enable', 'off');
@@ -455,7 +459,7 @@ function local_showmod(obj)
     set(handles.valshowLabel, 'Visible','off');
     % set help box to matlabbatch top node help
     [id stop help] = cfg_util('listcfgall', [], cfg_findspec({{'tag','matlabbatch'}}), {'showdoc'});
-    set(handles.helpbox, 'String',cfg_justify(handles.helpbox, help{1}{1}), 'Value',1);
+    set(handles.helpbox, 'Value',1, 'ListboxTop',1, 'String',cfg_justify(handles.helpbox, help{1}{1}));
 end;
 
 %% Show Item
@@ -498,7 +502,7 @@ function local_showvaledit(obj)
         else
             str = '';
         end;
-        set(handles.valshow,'String', str, 'Visible','on', 'Value', 1);
+        set(handles.valshow, 'Visible','on', 'Value',1, 'ListboxTop',1,'String', str);
         set(handles.valshowLabel, 'Visible','on');
         if ~isfield(udmodlist, 'defid')
             sout = local_showvaledit_deps(obj);
@@ -551,7 +555,8 @@ function local_showvaledit(obj)
             if cval == -1
                 cval = 1;
             end;
-            set(handles.valshow, 'String',str, 'Visible','on', 'Value',cval, ...
+            ltop = local_getListboxTop(handles.valshow, cval, numel(str));
+            set(handles.valshow, 'Visible','on', 'Value',cval, 'ListboxTop',ltop, 'String',str, ...
                               'Callback',@local_valedit_list, ...
                               'Keypressfcn',@local_valedit_key, ...
                               'Userdata',udvalshow);
@@ -636,7 +641,7 @@ function local_showvaledit(obj)
             end
             str = [str1(:); str2(:); str3(:)];
             udvalshow.cmd = [cmd1(:); cmd2(:); cmd3(:)];
-            set(handles.valshow,'String', str, 'Visible','on', 'Value',1, ...
+            set(handles.valshow, 'Visible','on', 'Value',1, 'ListboxTop',1, 'String', str, ...
                 'Callback',@local_valedit_repeat, ...
                 'KeyPressFcn', @local_valedit_key, ...
                 'Userdata',udvalshow);
@@ -658,7 +663,7 @@ function local_showvaledit(obj)
 end;
 [id stop help] = cfg_util('listmod', cmid{:}, udmodule.id{citem}, cfg_findspec, ...
                           cfg_tropts(cfg_findspec,1,1,1,1,false), {'showdoc'});
-set(handles.helpbox, 'string',cfg_justify(handles.helpbox, help{1}{1}), 'Value',1);
+set(handles.helpbox, 'Value',1, 'ListboxTop',1, 'string',cfg_justify(handles.helpbox, help{1}{1}));
 drawnow;
 
 %% List matching dependencies
@@ -1030,6 +1035,9 @@ function cfg_ui_OpeningFcn(hObject, eventdata, handles, varargin)
 % handles    structure with handles and user data (see GUIDATA)
 % varargin   command line arguments to cfg_ui (see VARARGIN)
 
+% move figure onscreen
+cfg_onscreen(hObject);
+
 % Add configuration specific menu items
 local_setmenu(handles.cfg_ui, [], @local_addtojob, true);
 
@@ -1523,6 +1531,15 @@ function BtnValAddDep_Callback(hObject, eventdata, handles)
 % Initialise udvalshow to empty struct
 udvalshow = struct('cval',[],'en',[],'key',[]);
 
+% --------------------------------------------------------------------
+function ltop = local_getListboxTop(obj, val, maxval)
+% Get a safe value for ListboxTop property while keeping previous settings
+% if possible.
+% obj     handle of Listbox object
+% val     new Value property
+% maxval  new number of lines in obj
+oltop = get(obj, 'ListboxTop');
+ltop  = min([max(oltop,1), max(val-1,1), maxval]);
 
 % --- Executes when user attempts to close cfg_ui.
 function cfg_ui_CloseRequestFcn(hObject, eventdata, handles)
@@ -1672,7 +1689,8 @@ function MenuViewShowCode_Callback(hObject, eventdata, handles)
 um1 = uimenu('Label','Copy', 'Callback',@(ob,ev)ShowCode_Copy(ob,ev,ctxt), 'Parent',um);
 um1 = uimenu('Label','Select all', 'Callback',@(ob,ev)ShowCode_SelAll(ob,ev,ctxt), 'Parent',um);
 um1 = uimenu('Label','Unselect all', 'Callback',@(ob,ev)ShowCode_UnSelAll(ob,ev,ctxt), 'Parent',um);
-set(ctxt, 'Max',numel(str), 'String',str, 'UIContextMenu',um);
+set(ctxt, 'Max',numel(str), 'UIContextMenu',um, 'Value',[], 'ListboxTop',1);
+set(ctxt, 'String',str);
 
 function ShowCode_Copy(ob, ev, ctxt)
 str = get(ctxt,'String');
diff --git a/matlabbatch/cfg_util.m b/matlabbatch/cfg_util.m
index d5afb38..04e4984 100644
--- a/matlabbatch/cfg_util.m
+++ b/matlabbatch/cfg_util.m
@@ -102,12 +102,14 @@
 % outputs - cell array with module outputs. If a module has not yet been
 %           run, a cfg_inv_out object is returned.
 %
-%  voutputs = cfg_util('getAllVOutputs', job_id)
+%  voutputs = cfg_util('getAllVOutputs', job_id[, mod_job_id])
 % 
 % voutputs - cell array with virtual output descriptions (cfg_dep objects).
 %            These describe the structure of the job outputs. To create
 %            dependencies, they can be entered into matching input objects
 %            in subsequent modules of the same job.
+%            If mod_job_id is supplied, only virtual output descriptions of
+%            the referenced module are returned.
 %
 %  [tag, val] = cfg_util('harvest', job_id[, mod_job_id])
 %
@@ -174,11 +176,11 @@
 %
 %  [job_id, mod_job_idlist] = cfg_util('initjob'[, job])
 %
-% Initialise a new job. 
+% Initialise a new job. If no further input arguments are provided, a new
+% job without modules will be created.
 % If job is given as input argument, the job tree structure will be
 % loaded with data from the struct/cell array job and a cell list of job
-% ids will be returned. Otherwise, a new job without modules will be
-% created.
+% ids will be returned.
 % The new job will be appended to an internal list of jobs. It must
 % always be referenced by its job_id.
 %
@@ -373,9 +375,9 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_util.m 3944 2010-06-23 08:53:40Z volkmar $
+% $Id: cfg_util.m 4252 2011-03-15 17:49:51Z volkmar $
 
-rev = '$Rev: 3944 $';
+rev = '$Rev: 4252 $';
 
 %% Initialisation of cfg variables
 % load persistent configuration data, initialise if necessary
@@ -434,6 +436,7 @@
                 jobs(varargin{1}).cj = c0;
                 jobs(varargin{1}).cjid2subs = {};
                 jobs(varargin{1}).cjrun = [];
+                jobs(varargin{1}).cjid2subsrun = {};
             end
         end
     case 'dumpcfg'
@@ -558,10 +561,14 @@
             varargout{1} = {};
         end
     case 'getallvoutputs',
-        cjob = varargin{1};
-        if cfg_util('isjob_id', cjob)
+        if nargin == 2 && cfg_util('isjob_id', varargin{1})
+            cjob = varargin{1};
             vmods = ~cellfun('isempty',jobs(cjob).cjid2subs); % Ignore deleted modules
             varargout{1} = cellfun(@(cid)subsref(jobs(cjob).cj, [cid substruct('.','sout')]), jobs(cjob).cjid2subs(vmods), 'UniformOutput',false);
+        elseif nargin > 2 && cfg_util('ismod_job_id',varargin{1},varargin{2})
+            cjob = varargin{1};
+            cmod = varargin{2};
+            varargout{1} = {subsref(jobs(cjob).cj, [jobs(cjob).cjid2subs{cmod} substruct('.','sout')])};
         else
             varargout{1} = {};
         end
@@ -630,13 +637,15 @@
             varargout{2} = {};
             return;
         elseif ischar(varargin{1}) || iscellstr(varargin{1})
-            job = cfg_load_jobs(varargin{1});
+            [job jobdedup] = cfg_load_jobs(varargin{1});
         elseif iscell(varargin{1}) && iscell(varargin{1}{1})
             % try to initialise cell array of jobs
             job = varargin{1};
+            jobdedup = NaN; % Postpone duplicate detection
         else
             % try to initialise single job
             job{1} = varargin{1};
+            jobdedup = 1;
         end
         % job should be a cell array of job structures
         isjob = true(size(job));
@@ -647,8 +656,26 @@
         if isempty(job)
             cfg_message('matlabbatch:initialise:invalid','No valid job.');
         else
+            if any(isnan(jobdedup))
+                % build up list of unique jobs
+                jobdedup = NaN*ones(size(job));
+                cu = 0;
+                for k = 1:numel(job)
+                    if isnan(jobdedup(k))
+                        % found new candidate
+                        cu = cu+1;
+                        jobdedup(k) = cu;
+                        % look for similar jobs under remaining candidates
+                        csel = find(isnan(jobdedup));
+                        eqind = cellfun(@(cjob)isequalwithequalnans(cjob,job{k}),job(csel));
+                        jobdedup(csel(eqind)) = cu;
+                    end
+                end
+            else
+                jobdedup = jobdedup(isjob);
+            end
             jobs(cjob).c0 = c0;
-            [jobs(cjob) mod_job_idlist] = local_initjob(jobs(cjob), job);
+            [jobs(cjob) mod_job_idlist] = local_initjob(jobs(cjob), job, jobdedup);
             varargout{1} = cjob;
             varargout{2} = mod_job_idlist;
         end
@@ -786,8 +813,11 @@
                                         varargin{3}], cm);
                 end
             end
+            % clear run configuration
+            jobs(cjob).cjrun = [];
+            jobs(cjob).cjid2subsrun = {};
         end
-    case {'run','runserial'}
+    case {'run','runserial','cont','contserial'}
         if cfg_util('isjob_id',varargin{1})
             cjob = varargin{1};
             dflag = false;
@@ -795,12 +825,9 @@
             cjob = cfg_util('initjob',varargin{1});
             dflag = true;
         end
-        if strcmpi(cmd, 'run')
-            pflag = cfg_get_defaults([mfilename '.runparallel']);
-        else
-            pflag = false;
-        end
-        [jobs(cjob) err] = local_runcj(jobs(cjob), cjob, pflag);
+        pflag = any(strcmpi(cmd, {'run','cont'})) && cfg_get_defaults([mfilename '.runparallel']);
+        cflag = any(strcmpi(cmd, {'cont','contserial'}));
+        [jobs(cjob) err] = local_runcj(jobs(cjob), cjob, pflag, cflag);
         if dflag
             cfg_util('deljob', cjob);
         end
@@ -849,9 +876,10 @@
             cm = setval(cm, varargin{4}, false);
             jobs(cjob).cj = subsasgn(jobs(cjob).cj, [jobs(cjob).cjid2subs{mod_job_id}, item_mod_id], cm);
             varargout{1} = all_set_item(cm);
+            % clear run configuration
+            jobs(cjob).cjrun = [];
+            jobs(cjob).cjid2subsrun = {};
         end
-        % clear run configuration
-        jobs(cjob).cjrun = [];
     case 'showdoc',
         if nargin == 2
             % get item from defaults tree
@@ -955,6 +983,7 @@
     jobs(k).c0.values{end+1} = c1;
     % clear run configuration
     jobs(k).cjrun = [];
+    jobs(k).cjid2subsrun = {};
 end
 cfg_message('matlabbatch:cfg_util:addapp:done', 'Added application ''%s''\n', c1.name);
 %-----------------------------------------------------------------------
@@ -994,6 +1023,7 @@
 job.cjid2subs{id} = cjsubs;
 % clear run configuration
 job.cjrun = [];
+job.cjid2subsrun = {};
 %-----------------------------------------------------------------------
 
 %-----------------------------------------------------------------------
@@ -1001,7 +1031,8 @@
 % Try to work around some unexpected behaviour of MATLAB's cd command
 if ~isempty(pth)
     if ischar(pth)
-        wd = cd(pth);
+        wd = pwd;
+        cd(pth);
     else
         cfg_message('matlabbatch:usage','CD: path must be a string.');
     end
@@ -1049,6 +1080,9 @@
 if any(cid)
     job.cj = update_deps(job.cj, oid(cid), job.cjid2subs(cid));
 end
+% clear run configuration
+job.cjrun = [];
+job.cjid2subsrun = {};
 %-----------------------------------------------------------------------
 
 %-----------------------------------------------------------------------
@@ -1082,6 +1116,7 @@
 job.cjid2subs{id} = struct([]);
 % clear run configuration
 job.cjrun = [];
+job.cjid2subsrun = {};
 %-----------------------------------------------------------------------
 
 %-----------------------------------------------------------------------
@@ -1314,65 +1349,59 @@ function local_gencode(c0, fname, tropts, preamble)
 %-----------------------------------------------------------------------
 
 %-----------------------------------------------------------------------
-function [cjob, mod_job_idlist] = local_initjob(cjob, job)
+function [cjob, mod_job_idlist] = local_initjob(cjob, job, jobdedup)
 % Initialise a cell array of jobs
-for n = 1:numel(job)
-    % update application defaults - not only in .values, but also in
-    % pre-configured .val items
-    cj1 = initialise(cjob.c0, '<DEFAULTS>', false);
-    % init job
-    cj1 = initialise(cj1, job{n}, false);
-    % canonicalise (this may break dependencies, see comment in
-    % local_getcjid2subs)
-    [cj1 cjid2subs1] = local_getcjid2subs(cj1);
-    % harvest, keeping dependencies
-    [u1 u2 u3 u4 u5 cj1] = harvest(cj1, cj1, false, false);
-    if n == 1
-        cjob.cj = cj1;
-        cjob.cjid2subs = cjid2subs1;
-    else
-        cjidoffset = numel(cjob.cjid2subs);
-        cjid2subs2 = cjid2subs1;
-        idsubs = substruct('.','id');
-        sdsubs = substruct('.','sdeps');
-        v1subs = substruct('.','val','{}',{1});
-        for k = 1:numel(cjid2subs2)
-            % update id subscripts
-            cjid2subs2{k}(2).subs{1} = cjid2subs2{k}(2).subs{1} + cjidoffset;
-            cj1 = subsasgn(cj1, [cjid2subs1{k}, idsubs], ...
-                           cjid2subs2{k});
-            % update src_exbranch in dependent cfg_items
-            sdeps = subsref(cj1, [cjid2subs1{k}, sdsubs]);
-            for l = 1:numel(sdeps)
-                % dependent module
-                dm = subsref(cj1, sdeps(l).tgt_exbranch);
-                % delete old tdeps - needs to be updated by harvest
-                dm.tdeps = [];
-                % dependencies in dependent item
-                ideps = subsref(dm, ...
-                                [sdeps(l).tgt_input v1subs]);
-                for m = 1:numel(ideps)
-                    % find reference that matches old source id
-                    if isequal(ideps(m).src_exbranch, cjid2subs1{k})
-                        ideps(m).src_exbranch = cjid2subs2{k};
-                    end
-                end
-                % save updated item
-                dm = subsasgn(dm, ...
-                              [sdeps(l).tgt_input v1subs], ...
-                              ideps);
-                % save updated module
-                cj1 = subsasgn(cj1, sdeps(l).tgt_exbranch, dm);
-            end
-            % done with sdeps - clear
-            cj1 = subsasgn(cj1, [cjid2subs1{k}, sdsubs], []);
-        end
-        % concatenate configs
-        cjob.cjid2subs = [cjob.cjid2subs cjid2subs2];
-        for k = 1:numel(cj1.val)
-            cjob.cj.val{end+1} = cj1.val{k};
+idsubs = substruct('.','id');
+sdsubs = substruct('.','sdeps');
+tdsubs = substruct('.','tdeps');
+v1subs = substruct('.','val','{}',{1});
+% Update application defaults - not only in .values, but also in
+% pre-configured .val items. Use this as template for all jobs.
+cjd = initialise(cjob.c0, '<DEFAULTS>', false);
+% Prepare all jobs - initialise similar jobs only once.
+[ujobdedup ui uj] = unique(jobdedup);
+% Initialise jobs
+ucj = cellfun(@(ucjob)initialise(cjd,ucjob,false), job(ui), 'UniformOutput', false);
+% Canonicalise (this may break dependencies, see comment in
+% local_getcjid2subs)
+[ucj ucjid2subs] = cellfun(@local_getcjid2subs, ucj, 'UniformOutput', false);
+% Harvest, keeping dependencies
+[u1 u2 u3 u4 u5 ucj] = cellfun(@(cucj)harvest(cucj, cucj, false, false), ucj, 'UniformOutput', false);
+% Deduplicate, concatenate
+cjob.cj = ucj{uj(1)};
+cjob.cjid2subs = ucjid2subs{uj(1)};
+for n = 2:numel(uj)
+    cjidoffset = numel(cjob.cjid2subs);
+    cj1        = ucj{uj(n)};
+    cjid2subs1 = ucjid2subs{uj(n)};
+    cjid2subs2 = ucjid2subs{uj(n)};
+    for k = 1:numel(cjid2subs2)
+        % update id subscripts
+        cjid2subs2{k}(2).subs{1} = cjid2subs2{k}(2).subs{1} + cjidoffset;
+        cj1 = subsasgn(cj1, [cjid2subs1{k}, idsubs], ...
+            cjid2subs2{k});
+        % update src_exbranch in dependent cfg_items
+        sdeps = subsref(cj1, [cjid2subs1{k}, sdsubs]);
+        for l = 1:numel(sdeps)
+            csdep = sdeps(l);
+            tgt_exbranch = csdep.tgt_exbranch;
+            tgt_input    = [csdep.tgt_input v1subs];
+            % dependencies in dependent item
+            ideps = subsref(cj1, [tgt_exbranch tgt_input]);
+            isel = cellfun(@(iid)isequal(iid,cjid2subs1{k}),{ideps.src_exbranch});
+            ideps(isel).src_exbranch = cjid2subs2{k};
+            % save updated item & module
+            cj1 = subsasgn(cj1, [tgt_exbranch tgt_input], ...
+                ideps);
+            % delete old tdeps - needs to be updated by harvest
+            cj1 = subsasgn(cj1, [tgt_exbranch tdsubs], []);
         end
+        % done with sdeps - clear
+        cj1 = subsasgn(cj1, [cjid2subs1{k}, sdsubs], []);
     end
+    % concatenate configs
+    cjob.cjid2subs = [cjob.cjid2subs cjid2subs2];
+    cjob.cj.val    = [cjob.cj.val cj1.val];
 end
 % harvest, update dependencies
 [u1 u2 u3 u4 u5 cjob.cj] = harvest(cjob.cj, cjob.cj, false, false);
@@ -1407,10 +1436,11 @@ function local_gencode(c0, fname, tropts, preamble)
 job.cjid2subs{id} = rcjsubs;
 % clear run configuration
 job.cjrun = [];
+job.cjid2subsrun = {};
 %-----------------------------------------------------------------------
 
 %-----------------------------------------------------------------------
-function [job err] = local_runcj(job, cjob, pflag)
+function [job err] = local_runcj(job, cjob, pflag, cflag)
 % Matlab uses a copy-on-write policy with very high granularity - if
 % modified, only parts of a struct or cell array are copied.
 % However, forward resolution may lead to high memory consumption if
@@ -1423,19 +1453,22 @@ function local_gencode(c0, fname, tropts, preamble)
 % If pflag is true, then modules will be executed in parallel if they are
 % independent. Setting pflag to false forces serial execution of modules
 % even if they seem to be independent.
-% If a job with pre-set module outputs .jout is passed in cj, the
-% corresponding modules will not be run again. This feature is currently unused.
+% If cflag is true, and a job with pre-set module outputs .jout is passed
+% in job.cjrun, the corresponding modules will not be run again.
 
 cfg_message('matlabbatch:run:jobstart', ...
             ['\n\n------------------------------------------------------------------------\n',...
              'Running job #%d\n', ...
              '------------------------------------------------------------------------'], cjob);
-
-job1 = local_compactjob(job);
+if cflag && ~isempty(job.cjrun)
+    [u1 mlbch] = harvest(job.cjrun, job.cjrun, false, true);
+else
+    job1 = local_compactjob(job);
+    job.cjid2subsrun = job1.cjid2subs;
+    [u1 mlbch u3 u4 u5 job.cjrun] = harvest(job1.cj, job1.cj, false, true);
+end
 % copy cjid2subs, it will be modified for each module that is run
-cjid2subs = job1.cjid2subs;
-job.cjid2subsrun = job1.cjid2subs;
-[u1 mlbch u3 u4 u5 job.cjrun] = harvest(job1.cj, job1.cj, false, true);
+cjid2subs = job.cjid2subsrun;
 cjid2subsfailed = {};
 cjid2subsskipped = {};
 tdsubs = substruct('.','tdeps');
@@ -1529,6 +1562,11 @@ function local_gencode(c0, fname, tropts, preamble)
         str{numel(cjid2subsfailed)+k+1} = sprintf('Skipped: %s', ...
                                                   subsref(job.cjrun, [cjid2subsskipped{k} substruct('.','name')]));
     end
+% Commented out for SPM public release
+%    str{end+1} = sprintf(['If the problem can be fixed without modifying ' ...
+%                        'the job, the computation can be resumed by ' ...
+%                        'running\n   cfg_util(''cont'',%d)\nfrom the ' ...
+%                        'MATLAB command line.'],cjob);
     cfg_message('matlabbatch:run:jobfailed', '%s\n', str{:});
     err.identifier = 'matlabbatch:run:jobfailederr';
     err.message    = sprintf(['Job execution failed. The full log of this run can ' ...
diff --git a/matlabbatch/private/cfg_load_jobs.m b/matlabbatch/private/cfg_load_jobs.m
new file mode 100644
index 0000000..74f2cec
--- /dev/null
+++ b/matlabbatch/private/cfg_load_jobs.m
@@ -0,0 +1,69 @@
+function [newjobs uind] = cfg_load_jobs(job)
+
+% function newjobs = cfg_load_jobs(job)
+%
+% Load a list of possible job files, return a cell list of jobs.
+%
+% This code is part of a batch job configuration system for MATLAB. See 
+%      help matlabbatch
+% for a general overview.
+%_______________________________________________________________________
+% Copyright (C) 2007 Freiburg Brain Imaging
+
+% Volkmar Glauche
+% $Id: cfg_load_jobs.m 4150 2011-01-07 14:41:56Z volkmar $
+
+rev = '$Rev: 4150 $'; %#ok
+
+if ischar(job)
+    filenames = cellstr(job);
+else
+    filenames = job;
+end;
+[ufilenames unused uind] = unique(filenames);
+ujobs = cell(size(ufilenames));
+usts  = false(size(ufilenames));
+for cf = 1:numel(ufilenames)
+    [ujobs{cf} usts(cf)] = load_single_job(ufilenames{cf});
+end
+sts   = usts(uind);
+uind  = uind(sts);
+newjobs = ujobs(uind);
+
+function [matlabbatch sts] = load_single_job(filename)
+[p,nam,ext] = fileparts(filename);
+switch ext
+    case '.xml',
+        try
+            loadxml(filename,'matlabbatch');
+        catch
+            cfg_message('matlabbatch:initialise:xml','LoadXML failed: ''%s''',filename);
+        end;
+    case '.mat'
+        try
+            S=load(filename);
+            matlabbatch = S.matlabbatch;
+        catch
+            cfg_message('matlabbatch:initialise:mat','Load failed: ''%s''',filename);
+        end;
+    case '.m'
+        try
+            fid = fopen(filename,'rt');
+            str = fread(fid,'*char');
+            fclose(fid);
+            eval(str);
+        catch
+            cfg_message('matlabbatch:initialise:m','Eval failed: ''%s''',filename);
+        end;
+        if ~exist('matlabbatch','var')
+            cfg_message('matlabbatch:initialise:m','No matlabbatch job found in ''%s''', filename);
+        end;
+    otherwise
+        cfg_message('matlabbatch:initialise:unknown','Unknown extension: ''%s''', filename);
+end;
+if exist('matlabbatch','var')
+    sts = true;
+else
+    sts = false;
+    matlabbatch = [];
+end;
diff --git a/matlabbatch/private/cfg_mlbatch_defaults.m b/matlabbatch/private/cfg_mlbatch_defaults.m
index 38107ca..f728d6d 100644
--- a/matlabbatch/private/cfg_mlbatch_defaults.m
+++ b/matlabbatch/private/cfg_mlbatch_defaults.m
@@ -11,26 +11,30 @@
 % Copyright (C) 2007 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: cfg_mlbatch_defaults.m 3792 2010-03-22 13:11:36Z volkmar $
+% $Id: cfg_mlbatch_defaults.m 4166 2011-01-17 15:06:41Z volkmar $
 
-rev = '$Rev: 3792 $'; %#ok
+rev = '$Rev: 4166 $'; %#ok
 
-% Font definition for cfg_ui user interface
-% cfg_defaults.cfg_ui.Xfont is a font struct as returned by uisetfont
-% lfont: used in lists, value edit dialogues etc.
-cfg_defaults.cfg_ui.lfont.FontAngle  = 'normal';
-cfg_defaults.cfg_ui.lfont.FontName   = get(0,'factoryTextFontName');
-cfg_defaults.cfg_ui.lfont.FontSize   = 12;
-cfg_defaults.cfg_ui.lfont.FontUnits  = 'points';
-cfg_defaults.cfg_ui.lfont.FontWeight = 'normal';
-% bfont: used for buttons
-cfg_defaults.cfg_ui.bfont.FontAngle  = get(0, 'factoryUicontrolFontAngle');
-cfg_defaults.cfg_ui.bfont.FontName   = get(0,'factoryUicontrolFontName');
-cfg_defaults.cfg_ui.bfont.FontSize   = get(0, 'factoryUicontrolFontSize');
-cfg_defaults.cfg_ui.bfont.FontUnits  = get(0, 'factoryUicontrolFontUnits');
-cfg_defaults.cfg_ui.bfont.FontWeight = get(0, 'factoryUicontrolFontWeight');
-% Toggle ExpertEdit mode. Value can be 'on' or 'off'
-cfg_defaults.cfg_ui.ExpertEdit = 'off';
+try
+    % Font definition for cfg_ui user interface
+    % cfg_defaults.cfg_ui.Xfont is a font struct as returned by uisetfont
+    % lfont: used in lists, value edit dialogues etc.
+    cfg_defaults.cfg_ui.lfont.FontAngle  = 'normal';
+    cfg_defaults.cfg_ui.lfont.FontName   = get(0,'factoryTextFontName');
+    cfg_defaults.cfg_ui.lfont.FontSize   = 12;
+    cfg_defaults.cfg_ui.lfont.FontUnits  = 'points';
+    cfg_defaults.cfg_ui.lfont.FontWeight = 'normal';
+    % bfont: used for buttons
+    cfg_defaults.cfg_ui.bfont.FontAngle  = get(0, 'factoryUicontrolFontAngle');
+    cfg_defaults.cfg_ui.bfont.FontName   = get(0,'factoryUicontrolFontName');
+    cfg_defaults.cfg_ui.bfont.FontSize   = get(0, 'factoryUicontrolFontSize');
+    cfg_defaults.cfg_ui.bfont.FontUnits  = get(0, 'factoryUicontrolFontUnits');
+    cfg_defaults.cfg_ui.bfont.FontWeight = get(0, 'factoryUicontrolFontWeight');
+    % Toggle ExpertEdit mode. Value can be 'on' or 'off'
+    cfg_defaults.cfg_ui.ExpertEdit = 'off';
+catch
+    cfg_defaults.cfg_ui = false;
+end
 
 % cfg_util
 % Parallel execution of independent modules
@@ -120,4 +124,4 @@
 % configuration has been initialised to speed up job
 % initialisation/harvest/run - set this to true if you want to debug some
 % configuration or the batch system itself
-cfg_defaults.cfg_item.checkval = false;
\ No newline at end of file
+cfg_defaults.cfg_item.checkval = false;
diff --git a/matlabbatch/private/cfg_onscreen.m b/matlabbatch/private/cfg_onscreen.m
new file mode 100644
index 0000000..5a34690
--- /dev/null
+++ b/matlabbatch/private/cfg_onscreen.m
@@ -0,0 +1,37 @@
+function varargout = cfg_onscreen(fg)
+% Move figure on the screen containing the mouse
+%    cfg_onscreen(fg) - move figure fg on the screen containing the mouse
+%    pos = cfg_onscreen(fg) - compute position of figure, do not move it
+%
+% This code is part of a batch job configuration system for MATLAB. See 
+%      help matlabbatch
+% for a general overview.
+%_______________________________________________________________________
+% Copyright (C) 2007 Freiburg Brain Imaging
+
+% Volkmar Glauche
+% $Id: cfg_onscreen.m 4033 2010-08-04 15:53:35Z volkmar $
+
+rev = '$Rev: 4033 $'; 
+
+% save figure units - use pixels here
+units = get(fg,'Units');
+set(fg,'Units','pixels');
+Rect = get(fg,'Position');
+S0   = get(0,'MonitorPosition');
+if size(S0,1) > 1 % Multiple Monitors
+    %-Use Monitor containing the Pointer
+    pl = get(0,'PointerLocation');
+    w  = find(pl(1)>=S0(:,1) & pl(1)<S0(:,1)+S0(:,3)-1 &...
+            pl(2)>=S0(:,2) & pl(2)<S0(:,2)+S0(:,4));
+    if numel(w)~=1, w = 1; end
+    S0 = S0(w,:);
+end
+Rect(1) = S0(1) + (S0(3) - Rect(3))/2;
+Rect(2) = S0(2) + (S0(4) - Rect(4))/2;
+if nargout == 0
+    set(fg, 'Position',Rect);
+else
+    varargout{1} = Rect;
+end
+set(fg,'Units',units);
diff --git a/spm.m b/spm.m
index 9828cc1..e76a190 100644
--- a/spm.m
+++ b/spm.m
@@ -63,7 +63,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm.m 3958 2010-06-30 16:24:46Z guillaume $
+% $Id: spm.m 4178 2011-01-27 15:12:53Z guillaume $
 
 
 %=======================================================================
@@ -434,7 +434,7 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
 %-----------------------------------------------------------------------
 if strcmpi(Modality,'EEG') && ~isdeployed
     addpath(fullfile(spm('Dir'),'external','fieldtrip'));
-    fieldtripdefs;
+    ft_defaults;
     addpath(fullfile(spm('Dir'),'external','bemcp'));
     addpath(fullfile(spm('Dir'),'external','ctf'));
     addpath(fullfile(spm('Dir'),'external','eeprobe'));
@@ -446,6 +446,16 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
     addpath(fullfile(spm('Dir'),'toolbox', 'MEEGtools'));
 end
 
+%-Turn output pagination off in Octave
+%-----------------------------------------------------------------------
+if strcmpi(spm_check_version,'octave')
+    try
+        more('off');
+        page_screen_output(false);
+        page_output_immediately(true);
+    end
+end
+
 %-Return defaults variable if asked
 %-----------------------------------------------------------------------
 if nargout, varargout = {spm_get_defaults}; end
@@ -532,14 +542,32 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
 %-Close any existing 'Interactive' 'Tag'ged windows
 %-----------------------------------------------------------------------
 delete(spm_figure('FindWin','Interactive'))
-Finter = openfig(fullfile(spm('Dir'),'spm_Interactive.fig'),'new','invisible');
-set(Finter,'name',spm('Ver'));
-S0 = spm('WinSize','0',1);
-SI = spm('WinSize','I');
-set(Finter,'Units','pixels', 'Position',[S0(1) S0(2) 0 0] +  SI);
-set(Finter,'Visible',Vis);
-varargout = {Finter};
 
+%-Create SPM Interactive window
+%-----------------------------------------------------------------------
+FS     = spm('FontSizes');
+PF     = spm_platform('fonts');
+S0     = spm('WinSize','0',1);
+SI     = spm('WinSize','I');
+Finter = figure('IntegerHandle','off',...
+    'Tag','Interactive',...
+    'Name',spm('Ver'),...
+    'NumberTitle','off',...
+    'Units','pixels',...
+    'Position',[S0(1) S0(2) 0 0] +  SI,...
+    'Resize','on',...
+    'Color',[1 1 1]*.8,...
+    'MenuBar','none',...
+    'DefaultTextFontName',PF.helvetica,...
+    'DefaultTextFontSize',FS(10),...
+    'DefaultAxesFontName',PF.helvetica,...
+    'DefaultUicontrolBackgroundColor',[1 1 1]*.7,...
+    'DefaultUicontrolFontName',PF.helvetica,...
+    'DefaultUicontrolFontSize',FS(10),...
+    'DefaultUicontrolInterruptible','on',...
+    'Renderer','painters',...
+    'Visible',Vis);
+varargout = {Finter};
 
 %=======================================================================
 case 'fnuisetup'                %-Robust UI setup for main SPM functions
@@ -793,15 +821,7 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
         error(['Can''t find ',Mfile,' on MATLABPATH']);
     end
 end
-SPMdir = fileparts(SPMdir);
-
-% if isdeployed
-%     ind = strfind(SPMdir,'_mcr')-1;
-%     if ~isempty(ind)
-%         % MATLAB 2008a/2009a doesn't need this
-%         SPMdir = fileparts(SPMdir(1:ind(1)));
-%     end
-% end
+SPMdir    = fileparts(SPMdir);
 varargout = {SPMdir};
 
 
@@ -938,14 +958,15 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
 %-----------------------------------------------------------------------
 if nargin<2, CmdLine=[]; else CmdLine=varargin{2}; end
 if isempty(CmdLine)
-    defaults = spm('getglobal','defaults');
-    if isfield(defaults,'cmdline')
-        CmdLine = defaults.cmdline;
-    else
+    try
+        CmdLine = spm_get_defaults('cmdline');
+    catch
         CmdLine = 0;
     end
 end
-varargout = {CmdLine || (get(0,'ScreenDepth')==0)};
+varargout = { CmdLine | ...
+              (get(0,'ScreenDepth')==0) | ...
+              strcmpi(spm_check_version,'octave') };
 
 
 %=======================================================================
@@ -1006,7 +1027,7 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
 % spm('Pointer',Pointer)
 %-----------------------------------------------------------------------
 if nargin<2, Pointer='Arrow'; else  Pointer=varargin{2}; end
-set(get(0,'Children'),'Pointer',Pointer)
+set(get(0,'Children'),'Pointer',lower(Pointer))
 
 
 %=======================================================================
@@ -1148,7 +1169,14 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
 if isempty(SPM_VER) || (nargin > 0 && ReDo)
     v = struct('Name','','Version','','Release','','Date','');
     try
-        fid = fopen(fullfile(spm('Dir'),'Contents.m'),'rt');
+        if isdeployed
+            % in deployed mode, M-files are encrypted
+            % (even if first two lines of Contents.m "should" be preserved)
+            vfile = fullfile(spm('Dir'),'Contents.txt');
+        else
+            vfile = fullfile(spm('Dir'),'Contents.m');
+        end
+        fid = fopen(vfile,'rt');
         if fid == -1, error(str); end
         l1 = fgetl(fid); l2 = fgetl(fid);
         fclose(fid);
@@ -1157,16 +1185,7 @@ case lower(Modalities)       %-Initialise SPM in PET, fMRI, EEG modality
         v.Name = l1; v.Date = t{4};
         v.Version = t{2}; v.Release = t{3}(2:end-1);
     catch
-        if isdeployed
-            % in deployed mode, M-files are encrypted
-            % (but first two lines of Contents.m should be preserved)
-            v.Name    = 'Statistical Parametric Mapping';
-            v.Version = '8';
-            v.Release = 'SPM8';
-            v.Date    = date;
-        else
-            error('Can''t obtain SPM Revision information.');
-        end
+        error('Can''t obtain SPM Revision information.');
     end
     SPM_VER = v;
 end
diff --git a/spm_ADEM.m b/spm_ADEM.m
index 2a7365c..3a5b182 100644
--- a/spm_ADEM.m
+++ b/spm_ADEM.m
@@ -38,6 +38,8 @@
 %   M(i).V  = fixed precision (input noise)
 %   M(i).W  = fixed precision (state noise)
 %   M(i).xP = precision (states)
+%   
+%   M(1).Ra = indices of prediction errors driving action
 %
 %   M(i).m  = number of inputs v(i + 1);
 %   M(i).n  = number of states x(i)
@@ -112,7 +114,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_ADEM.m 3901 2010-05-27 16:14:36Z karl $
+% $Id: spm_ADEM.m 4230 2011-03-07 20:58:38Z karl $
  
 % check model, data, priors and unpack
 %--------------------------------------------------------------------------
@@ -567,7 +569,7 @@
         for i = 1:nh
             iS = iS + Q{i}*exp(qh.h(i));
         end
-        S     = inv(iS);
+        S     = spm_inv(iS);
         dS    = ECE + EE - S*nY;
          
         % 1st-order derivatives: dFdh = dF/dh
@@ -637,8 +639,10 @@
             z     = spm_unvec(pU(t).z{1},{G.v});
             w     = spm_unvec(pU(t).w{1},{G.x});
             for i = 1:nl
-                PU.v{i}(:,t)     = spm_vec(v{i});
-                PU.z{i}(:,t)     = spm_vec(z{i});
+                try
+                    PU.v{i}(:,t) = spm_vec(v{i});
+                    PU.z{i}(:,t) = spm_vec(z{i});
+                end
                 try
                     PU.x{i}(:,t) = spm_vec(x{i});
                     PU.w{i}(:,t) = spm_vec(w{i});
diff --git a/spm_Bcdf.m b/spm_Bcdf.m
index 3be87b4..a170bb5 100644
--- a/spm_Bcdf.m
+++ b/spm_Bcdf.m
@@ -12,7 +12,7 @@
 % distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Beta distribution has two shape parameters, v and w, and is
 % defined for v>0 & w>0 and for x in [0,1] (See Evans et al., Ch5).
 % The Cumulative Distribution Function (CDF) F(x) is the probability
@@ -22,15 +22,15 @@
 % definitions of the incomplete beta function.
 %
 % Variate relationships:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Many: See Evans et al., Ch5
 %
 % Algorithm:
-%-----------------------------------------------------------------------
-% Using Matlab's implementation of the incomplete beta finction (betainc).
+%--------------------------------------------------------------------------
+% Using MATLAB's implementation of the incomplete beta finction (betainc).
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -43,37 +43,39 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Bcdf.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_Bcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(v);ndims(w)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];...
-    [size(w),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];...
+      [size(w),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 F = zeros(rs);
 
 %-Only defined for x in [0,1] & strictly positive v & w.
 % Return NaN if undefined.
 md = ( x>=0  &  x<=1  &  v>0  &  w>0 );
-if any(~md(:)), F(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    F(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special cases: F=1 when x=1
 F(md & x==1) = 1;
diff --git a/spm_Bpdf.m b/spm_Bpdf.m
index 810f5b3..af0e4ec 100644
--- a/spm_Bpdf.m
+++ b/spm_Bpdf.m
@@ -11,7 +11,7 @@
 % spm_Bpdf implements the Probability Density Function for Beta distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The PDF of the Beta distribution shape parameters v & w, defined
 % for positive integer degrees of freedom v>0 & w>0, and for x in
 % [0,1] is given by: (See Evans et al., Ch5)
@@ -21,16 +21,16 @@
 %                 beta(v,w)
 %
 % Variate relationships:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Many: See Evans et al., Ch5
 %
 % Algorithm:
-%-----------------------------------------------------------------------
-% Direct computation using logs and Matlab's implementation of the log
+%--------------------------------------------------------------------------
+% Direct computation using logs and MATLAB's implementation of the log
 % beta function (betaln).
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -43,37 +43,39 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Bpdf.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_Bpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(v);ndims(w)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];...
-    [size(w),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];...
+      [size(w),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for x in [0,1] & strictly positive v & w.
 % Return NaN if undefined.
 md = ( x>=0  &  x<=1  &  v>0  &  w>0 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special cases: x=0 & x=1
 f(md & x==0 & v<1) = Inf;
diff --git a/spm_DEM.m b/spm_DEM.m
index 9960879..6d174ec 100644
--- a/spm_DEM.m
+++ b/spm_DEM.m
@@ -42,7 +42,7 @@
 %--------------------------------------------------------------------------
 %   qP.P    = Conditional expectation
 %   qP.C    = Conditional covariance
-%  
+%
 % conditional moments of hyper-parameters (log-transformed) - q(h)
 %--------------------------------------------------------------------------
 %   qH.h    = Conditional expectation (cause noise)
@@ -63,9 +63,9 @@
 %                E: qp.p = max <L>q(u,h)
 %                M: qh.h = max <L>q(u,p)
 %
-% where qu.u corresponds to the conditional expectation of hidden states x 
-% and causal states v and so on.  L is the ln p(y,u,p,h|M) under the model 
-% M. The conditional covariances obtain analytically from the curvature of 
+% where qu.u corresponds to the conditional expectation of hidden states x
+% and causal states v and so on.  L is the ln p(y,u,p,h|M) under the model
+% M. The conditional covariances obtain analytically from the curvature of
 % L with respect to u, p and h.
 %
 % The D-step is embedded in the E-step because q(u) changes with each
@@ -73,13 +73,13 @@
 % model using temporal derivatives at each time point.  Continuity of the
 % conditional trajectories q(u,t) is assured by a continuous ascent of F(t)
 % in generalised co-ordinates.  This means DEM can deconvolve online and can
-% represents an alternative to Kalman filtering or alternative Bayesian 
+% represents an alternative to Kalman filtering or alternative Bayesian
 % update procedures.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_DEM.m 3977 2010-07-08 14:14:35Z karl $
+% $Id: spm_DEM.m 4146 2010-12-23 21:01:39Z karl $
  
 % check model, data, priors and confounds and unpack
 %--------------------------------------------------------------------------
@@ -91,11 +91,7 @@
  
 % tolerance for changes in norm
 %--------------------------------------------------------------------------
-TOL  = 1e-3;
-
-% enable mean field effects
-%--------------------------------------------------------------------------
-mf   = 1;
+TOL  = exp(-4);
  
 % order parameters (d = n = 1 for static models) and checks
 %==========================================================================
@@ -114,12 +110,11 @@
  
 % number of iterations
 %--------------------------------------------------------------------------
-try nD = M(1).E.nD; catch nD = 1;  end
-try nE = M(1).E.nE; catch nE = 1;  end
-try nM = M(1).E.nM; catch nM = 8;  end
-try nN = M(1).E.nN; catch nN = 16; end
-try K  = M(1).E.K;  catch K  = 1;  end
-
+try nD = M(1).E.nD; catch nD = 1; end
+try nE = M(1).E.nE; catch nE = 8; end
+try nM = M(1).E.nM; catch nM = 8; end
+try K  = M(1).E.K;  catch K  = 1; end
+ 
  
 % initialise regularisation parameters
 %--------------------------------------------------------------------------
@@ -128,11 +123,12 @@
 else
     td = {8};
 end
-if M(1).E.linear
-    te = 8;                            % integration time for E-Step
+if M(1).E.linear == 1
+    te = 4;                            % integration time for E-Step
 else
-    te = 4;
+    te = 0;
 end
+tm     = 4;                             % integration time for M-Step
  
 % precision components Q{} requiring [Re]ML estimators (M-Step)
 %==========================================================================
@@ -145,12 +141,12 @@
 W0    = kron(sparse(n,n),spm_cat(w0));
 Qp    = blkdiag(V0,W0);
 for i = 1:nl
- 
+    
     % precision (R) and covariance of generalised errors
     %----------------------------------------------------------------------
     iVv   = spm_DEM_R(n,M(i).sv);
     iVw   = spm_DEM_R(n,M(i).sw);
- 
+    
     % noise on causal states (Q)
     %----------------------------------------------------------------------
     for j = 1:length(M(i).Q)
@@ -158,13 +154,13 @@
         q{i,i}     = M(i).Q{j};
         Q{end + 1} = blkdiag(kron(iVv,spm_cat(q)),W0);
     end
- 
+    
     % and fixed components (V)
     %----------------------------------------------------------------------
     q      = v0;
     q{i,i} = M(i).V;
     Qp     = Qp + blkdiag(kron(iVv,spm_cat(q)),W0);
- 
+    
     % noise on hidden states (R)
     %----------------------------------------------------------------------
     for j = 1:length(M(i).R)
@@ -172,7 +168,7 @@
         q{i,i}     = M(i).R{j};
         Q{end + 1} = blkdiag(V0,kron(iVw,spm_cat(q)));
     end
- 
+    
     % and fixed components (W)
     %----------------------------------------------------------------------
     q      = w0;
@@ -200,21 +196,21 @@
 ph.c  = spm_cat(spm_diag({M.hC M.gC}));    % prior covariances of h
 qh.h  = ph.h;                              % conditional expectation
 qh.c  = ph.c;                              % conditional covariance
-ph.ic = spm_pinv(ph.c);                    % prior precision 
+ph.ic = spm_pinv(ph.c);                    % prior precision
  
 % priors on parameters (in reduced parameter space)
 %==========================================================================
 pp.c  = cell(nl,nl);
 qp.p  = cell(nl,1);
 for i = 1:(nl - 1)
- 
+    
     % eigenvector reduction: p <- pE + qp.u*qp.p
     %----------------------------------------------------------------------
     qp.u{i}   = spm_svd(M(i).pC,exp(-32));           % basis for parameters
     M(i).p    = size(qp.u{i},2);                     % number of qp.p
     qp.p{i}   = sparse(M(i).p,1);                    % initial qp.p
     pp.c{i,i} = qp.u{i}'*M(i).pC*qp.u{i};            % prior covariance
- 
+    
 end
 Up    = spm_cat(spm_diag(qp.u));
  
@@ -224,25 +220,26 @@
 nb    = size(X,1);                          % number of confounds
 nn    = nb*ny;                              % number of nuisance parameters
 nf    = np + nn;                            % number of free parameters
-ip    = [1:np];
-ib    = [1:nn] + np;
+ip    = (1:np);
+ib    = (1:nn) + np;
 pp.c  = spm_cat(pp.c);
 pp.ic = spm_inv(pp.c);
+pp.p  = spm_vec(qp.p);
  
-% initialise conditional density q(p) (for D-Step)
+% initialise conditional density q(p) := qp.e (for D-Step)
 %--------------------------------------------------------------------------
 for i = 1:(nl - 1)
     try
         qp.e{i} = qp.p{i} + qp.u{i}'*(spm_vec(M(i).P) - spm_vec(M(i).pE));
     catch
-        qp.e{i} = qp.p{i};                           % initial qp.e
+        qp.e{i} = qp.p{i};
     end
-end   
+end
 qp.e  = spm_vec(qp.e);
 qp.c  = sparse(nf,nf);
 qp.b  = sparse(ny,nb);
-
-
+ 
+ 
 % initialise dedb
 %--------------------------------------------------------------------------
 for i = 1:nl
@@ -285,11 +282,11 @@
 Dy    = kron(spm_speye(n,n,1),spm_speye(ny,ny,0));
 Dc    = kron(spm_speye(d,d,1),spm_speye(nc,nc,0));
 D     = spm_cat(spm_diag({Dx,Dv,Dy,Dc}));
-              
+ 
 % and null blocks
 %--------------------------------------------------------------------------
 dVdy  = sparse(n*ny,1);
-dVdc  = sparse(d*nc,1);              
+dVdc  = sparse(d*nc,1);
 dVdyy = sparse(n*ny,n*ny);
 dVdcc = sparse(d*nc,d*nc);
  
@@ -300,323 +297,257 @@
 dWduu = sparse(nu,nu);
 dWdpp = sparse(nf,nf);
  
-% preclude unneceassry iterations
+% preclude unnecessary iterations
 %--------------------------------------------------------------------------
 if ~nh,        nM = 1; end
-if ~nf,        nE = 1; end
-if ~nf && ~nh, nN = 1; end
+if ~nf && ~nh, nE = 1; end
+ 
  
+% preclude very precise states from entering free-energy/action
+%--------------------------------------------------------------------------
+ix     = (1:(nx*n)) + ny*n + nv*n;
+iv     = (1:(nv*d)) + ny*n;
+je     = diag(Qp) < exp(16);
+ju     = [je(ix); je(iv)];
  
-% Iterate DEM
+ 
+% E-Step: (with embedded D and M-Steps)
 %==========================================================================
 Fi     = -Inf;
-for iN = 1:nN
+for iE = 1:nE
     
     % get time and celar persistent variables in evaluation routines
     %----------------------------------------------------------------------
-    tic; clear spm_DEM_eval
- 
-    % E-Step: (with embedded D-Step)
+    tic;  clear spm_DEM_eval
+    
+    % [re-]set accumulators for E-Step
+    %----------------------------------------------------------------------
+    dFdh  = sparse(nh,1);                 % gradient   (hyperparamteres)
+    dFdhh = sparse(nh,nh);                % curvatiure (hyperparamteres)
+    dFdp  = sparse(nf,1);                 % gradient   (paramteres)
+    dFdpp = sparse(nf,nf);                % curvatiure (paramteres)
+    qp.ic = sparse(0);                    % conditional precision (p)
+    iqu.c = sparse(0);                    % conditional information (p)
+    EE    = sparse(0);
+    ECE   = sparse(0);
+    
+    % [re-]set precisions using ReML hyperparameter estimates
+    %----------------------------------------------------------------------
+    iS    = Qp;
+    for i = 1:nh
+        iS = iS + Q{i}*exp(qh.h(i));
+    end
+    
+    % [re-]adjust for confounds
+    %----------------------------------------------------------------------
+    Y     = Y - qp.b*X;
+    
+    % [re-]set states & their derivatives
+    %----------------------------------------------------------------------
+    try, qu = qU(1); end
+    
+    % D-Step: (nD D-Steps for each sample)
     %======================================================================
-    mp     = 0;                                % cumulator for changes in p
-    Fe     = -exp(64);                         % Free energy (E-stpep)
-    for iE = 1:nE
- 
- 
-        % [re-]set accumulators for E-Step
-        %------------------------------------------------------------------
-        spm_logdet_qu_c = 0;
-        dFdp  = zeros(nf,1);
-        dFdpp = zeros(nf,nf);
-        EE    = sparse(0);
-        ECE   = sparse(0);
-        qp.ic = sparse(0);
-       
-        % [re-]set precisions using ReML hyperparameter estimates
-        %------------------------------------------------------------------
-        iS    = Qp;
-        for i = 1:nh
-           iS = iS + Q{i}*exp(qh.h(i));
-        end
+    for iY = 1:nY
         
-        % [re-]adjust for confounds
+        % [re-]set states for static systems
         %------------------------------------------------------------------
-        Y     = Y - qp.b*X;
+        if ~nx, try, qu = qU(iY); end, end
         
-        % [re-]set states & their derivatives
-        %------------------------------------------------------------------
-        try
-            qu = qU(1);
-        end
-        
-        % D-Step: (nD D-Steps for each sample)
+        % D-Step: until convergence for static systems
         %==================================================================
-        for iY = 1:nY
- 
-            % [re-]set states for static systems
+        Fd     = -exp(64);
+        for iD = 1:nD
+            
+            % sampling time
             %--------------------------------------------------------------
-            if ~nx
-                try, qu = qU(iY); end
+            ts = iY + (iD - 1)/nD;
+            
+            % derivatives of responses and inputs
+            %--------------------------------------------------------------
+            try
+                qu.y(1:n) = spm_DEM_embed(Y,n,ts,1,M(1).delays);
+                qu.u(1:d) = spm_DEM_embed(U,d,ts);
+            catch
+                qu.y(1:n) = spm_DEM_embed(Y,n,ts);
+                qu.u(1:d) = spm_DEM_embed(U,d,ts);
             end
- 
-            % D-Step: until convergence for static systems
+            
+            % compute dEdb (derivatives of confounds)
+            %--------------------------------------------------------------
+            b     = spm_DEM_embed(X,n,ts);
+            for i = 1:n
+                dedbi{1}  = -kron(b{i}',speye(ny,ny));
+                dEdb{i,1} =  spm_cat(dedbi);
+            end
+            
+            % evaluate functions:
+            % E = v - g(x,v) and derivatives dE.dx, ...
             %==============================================================
-            Fd     = -exp(64);
-            for iD = 1:nD
-                
-                % sampling time
-                %----------------------------------------------------------
-                ts        = iY + (iD - 1)/nD;
-                
-                % derivatives of responses and inputs
-                %----------------------------------------------------------
-                try
-                    qu.y(1:n) = spm_DEM_embed(Y,n,ts,1,M(1).delays);
-                    qu.u(1:d) = spm_DEM_embed(U,d,ts);
-                catch
-                    qu.y(1:n) = spm_DEM_embed(Y,n,ts);
-                    qu.u(1:d) = spm_DEM_embed(U,d,ts);
-                end
-
-                % compute dEdb (derivatives of confounds)
-                %----------------------------------------------------------
-                b     = spm_DEM_embed(X,n,ts);
-                for i = 1:n
-                    dedbi{1}  = -kron(b{i}',speye(ny,ny));
-                    dEdb{i,1} =  spm_cat(dedbi);
-                end
- 
-                % evaluate functions:
-                % E = v - g(x,v) and derivatives dE.dx, ...
-                %==========================================================       
-                [E dE] = spm_DEM_eval(M,qu,qp);
+            [E dE] = spm_DEM_eval(M,qu,qp);
+            
+            % conditional covariance [of states {u}]
+            %--------------------------------------------------------------
+            qu.p   = dE.du'*iS*dE.du + Pu;
+            qu.c   = diag(ju)*spm_inv(qu.p)*diag(ju);
+            iqu.c  = iqu.c + spm_logdet(qu.c);
+            
+            % and conditional covariance [of parameters {P}]
+            %--------------------------------------------------------------
+            dE.dP  = spm_cat({dE.dp dEdb});
+            ECEu   = dE.du*qu.c*dE.du';
+            ECEp   = dE.dP*qp.c*dE.dP';
+            
+            if ~nx
                 
-                % conditional covariance [of states {u}]
-                %----------------------------------------------------------
-                qu.p   = dE.du'*iS*dE.du + Pu;
-                qu.c   = spm_inv(qu.p);
-                spm_logdet_qu_c = spm_logdet_qu_c + spm_logdet(qu.c);
-
-                % and conditional covariance [of parameters {P}]
+                % Evaluate objective function L(t) (for static models)
                 %----------------------------------------------------------
-                dE.dP  = spm_cat({dE.dp dEdb});
-                ECEu   = dE.du*qu.c*dE.du';
-                ECEp   = dE.dP*qp.c*dE.dP';
+                L = - trace(E'*iS*E)/2 ...           % states (u)
+                    - trace(iS*ECEp)/2;              % expectation q(p)
                 
-                % Evaluate objective function L(t) (for static models)
+                % if F is increasing, save expansion point
                 %----------------------------------------------------------
-                if ~nx
+                if L > Fd
+                    td     = {min(td{1} + 1, 8)};
+                    Fd     = L;
+                    B.qu   = qu;
+                    B.E    = E;
+                    B.dE   = dE;
+                    B.ECEp = ECEp;
                     
-                    L = - trace(E'*iS*E)/2 ...           % states (u)
-                        - trace(iS*ECEp)/2;              % expectation q(p)
- 
-                    % if F is increasing, save expansion point
+                else
+                    
+                    % otherwise, return to previous expansion point
                     %------------------------------------------------------
-                    if L > Fd
-                        td     = {min(td{1} + 1,+8)};
-                        Fd     = L;
-                        B.qu   = qu;
-                        B.E    = E;
-                        B.dE   = dE;
-                        B.ECEp = ECEp;
-                        
-                    else
-                        
-                        % otherwise, return to previous expansion point
-                        %--------------------------------------------------
-                        qu     = B.qu;
-                        E      = B.E;
-                        dE     = B.dE;
-                        ECEp   = B.ECEp;
-                        td     = {min(td{1} - 2,-4)};
-                    end
-                end
- 
-                % save states at qu(t)
-                %----------------------------------------------------------
-                if iD == 1
-                    qE{iY} = E;
-                    qU(iY) = qu;
-                end
-                
-                
-                % uncertainty about parameters dWdv, ... ; W = ln(|qp.c|)
-                %==========================================================
-                if np && mf
-                    for i = 1:nu
-                        CJp(:,i)   = spm_vec(qp.c(ip,ip)*dE.dpu{i}'*iS);
-                        dEdpu(:,i) = spm_vec(dE.dpu{i}');
-                    end
-                    dWdu   = CJp'*spm_vec(dE.dp');
-                    dWduu  = CJp'*dEdpu;
-                end
- 
- 
-                % D-step update: of causes v{i}, and hidden states x(i)
-                %==========================================================
-                
-                % conditional modes
-                %----------------------------------------------------------
-                q     = {qu.x{1:n} qu.v{1:d} qu.y{1:n} qu.u{1:d}};
-                u     = spm_vec(q);
-                
-                % first-order derivatives
-                %----------------------------------------------------------             
-                dVdu  = -dE.du'*iS*E     - dWdu/2  - Pu*u(1:nu); 
-                
-                % and second-order derivatives
-                %----------------------------------------------------------
-                dVduu = -dE.du'*iS*dE.du - dWduu/2 - Pu;
-                dVduy = -dE.du'*iS*dE.dy;
-                dVduc = -dE.du'*iS*dE.dc;
-                
-                % gradient
-                %----------------------------------------------------------
-                dFdu  = spm_vec({dVdu;  dVdy;  dVdc });         
-                                 
-                % Jacobian (variational flow)
-                %----------------------------------------------------------
-                dFduu = spm_cat({dVduu  dVduy  dVduc  ;
-                                 []     dVdyy  []     ;
-                                 []     []     dVdcc});
-                             
-                             
-                % update conditional modes of states
-                %==========================================================
-                f     = K*dFdu  + D*u;
-                dfdu  = K*dFduu + D;
-                du    = spm_dx(dfdu,f,td);
-                q     = spm_unvec(u + du,q);
-                
-                % and save them
-                %----------------------------------------------------------
-                qu.x(1:n) = q([1:n]);
-                qu.v(1:d) = q([1:d] + n);
-                
-        
-                % D-Step: break if convergence (for static models)
-                %----------------------------------------------------------
-                if ~nx
-                    qU(iY) = qu; 
-                end
-                if ~nx && ((dFdu'*du < 1e-2) || (norm(du,1) < TOL))
-                    break
-                end
-                
-               % report (D-Steps)
-               %-----------------------------------------------------------
-                if ~nx && nY < 8
-                    str{1} = sprintf('D-Step: %i (%i)',iD,iY);
-                    str{2} = sprintf('I:%.6e',full(Fd));
-                    str{3} = sprintf('u:%.2e',full(du'*du));
-                    fprintf('%-16s%-24s%-16s\n',str{1:3})
+                    qu     = B.qu;
+                    E      = B.E;
+                    dE     = B.dE;
+                    ECEp   = B.ECEp;
+                    td     = {min(td{1} - 2,-4)};
                 end
- 
-            end % D-Step
+            end
             
-            % Gradients and curvatures for E-Step: W = tr(C*J'*iS*J)
+            % save states at qu(t)
+            %--------------------------------------------------------------
+            if iD == 1
+                qE{iY} = E;
+                qU(iY) = qu;
+            end
+            
+            
+            % uncertainty about parameters dWdv, ... ; W = ln(|qp.c|)
             %==============================================================
-            if np && mf
-                for i = ip
-                    CJu(:,i)   = spm_vec(qu.c*dE.dup{i}'*iS);
-                    dEdup(:,i) = spm_vec(dE.dup{i}');
+            if np
+                for i = 1:nu
+                    CJp(:,i)   = spm_vec(qp.c(ip,ip)*dE.dpu{i}'*iS);
+                    dEdpu(:,i) = spm_vec(dE.dpu{i}');
                 end
-                dWdp(ip)       = CJu'*spm_vec(dE.du');
-                dWdpp(ip,ip)   = CJu'*dEdup;
+                dWdu   = CJp'*spm_vec(dE.dp');
+                dWduu  = CJp'*dEdpu;
             end
- 
- 
-            % Accumulate; dF/dP = <dL/dp>, dF/dpp = ...
+            
+            
+            % D-step update: of causes v{i}, and hidden states x(i)
+            %==============================================================
+            
+            % conditional modes
             %--------------------------------------------------------------
-            dFdp  = dFdp  - dWdp/2  - dE.dP'*iS*E;
-            dFdpp = dFdpp - dWdpp/2 - dE.dP'*iS*dE.dP;
-            qp.ic = qp.ic           + dE.dP'*iS*dE.dP;
- 
-            % and quantities for M-Step
+            q     = {qu.x{1:n} qu.v{1:d} qu.y{1:n} qu.u{1:d}};
+            u     = spm_vec(q);
+            
+            % first-order derivatives
             %--------------------------------------------------------------
-            EE    = E*E'+ EE;
-            if mf
-                ECE = ECE + ECEu + ECEp;
-            end
+            dVdu  = -dE.du'*iS*E     - dWdu/2  - Pu*u(1:nu);
             
-        end % sequence (nY)
- 
-        % augment with priors
-        %------------------------------------------------------------------
-        dFdp(ip)     = dFdp(ip)     - pp.ic*qp.e;
-        dFdpp(ip,ip) = dFdpp(ip,ip) - pp.ic;
-        qp.ic(ip,ip) = qp.ic(ip,ip) + pp.ic;
-        qp.c         = spm_inv(qp.ic);
-             
-        % evaluate objective function <L(t)>
-        %==================================================================
-        L = - trace(iS*EE)/2  ...                    % states (u)
-            - trace(qp.e'*pp.ic*qp.e)/2;             % parameters (p)
- 
- 
-        % if F is increasing, save expansion point and derivatives
-        %------------------------------------------------------------------
-        if L > Fe || iN == 1
+            % and second-order derivatives
+            %--------------------------------------------------------------
+            dVduu = -dE.du'*iS*dE.du - dWduu/2 - Pu;
+            dVduy = -dE.du'*iS*dE.dy;
+            dVduc = -dE.du'*iS*dE.dc;
+            
+            % gradient
+            %--------------------------------------------------------------
+            dFdu  = spm_vec({dVdu;  dVdy;  dVdc });
             
-            Fe      = L;
-            te      = min(te + 1,+8);
-            B.dFdp  = dFdp;
-            B.dFdpp = dFdpp;
-            B.qp    = qp;
-            B.mp    = mp;
+            % Jacobian (variational flow)
+            %--------------------------------------------------------------
+            dFduu = spm_cat({dVduu  dVduy  dVduc  ;
+                             []     dVdyy  []     ;
+                             []     []     dVdcc});
             
-        else
             
-            % otherwise, return to previous expansion point
+            % update conditional modes of states
+            %==============================================================
+            f     = K*dFdu  + D*u;
+            dfdu  = K*dFduu + D;
+            du    = spm_dx(dfdu,f,td);
+            q     = spm_unvec(u + du,q);
+            
+            % and save them
             %--------------------------------------------------------------
-            dFdp    = B.dFdp;
-            dFdpp   = B.dFdpp;
-            qp      = B.qp;
-            mp      = B.mp;
-            te      = min(te - 2,-4);
-        end
- 
-        % E-step: update expectation (p)
+            qu.x(1:n) = q([1:n]);
+            qu.v(1:d) = q([1:d] + n);
+            
+            
+            % D-Step: break if convergence (for static models)
+            %--------------------------------------------------------------
+            if ~nx
+                qU(iY) = qu;
+            end
+            if ~nx && ((dFdu'*du < TOL) || (norm(du,1) < TOL))
+                break
+            end
+            
+        end % D-Step
+        
+        % Gradients and curvatures for E-Step: W = tr(C*J'*iS*J)
         %==================================================================
- 
-        % update conditional expectation
+        if np
+            for i = ip
+                CJu(:,i)   = spm_vec(qu.c*dE.dup{i}'*iS);
+                dEdup(:,i) = spm_vec(dE.dup{i}');
+            end
+            dWdp(ip)       = CJu'*spm_vec(dE.du');
+            dWdpp(ip,ip)   = CJu'*dEdup;
+        end
+        
+        
+        % Accumulate; dF/dP = <dL/dp>, dF/dpp = ...
         %------------------------------------------------------------------
-        dp   = spm_dx(dFdpp,dFdp,{te});
-        qp.e = qp.e + dp(ip);
-        qp.p = spm_unvec(qp.e,qp.p);
-        qp.b = spm_unvec(dp(ib),qp.b);
-        mp   = mp + dp;
- 
-        % convergence (E-Step)
+        dFdp  = dFdp  - dWdp/2  - dE.dP'*iS*E;
+        dFdpp = dFdpp - dWdpp/2 - dE.dP'*iS*dE.dP;
+        qp.ic = qp.ic           + dE.dP'*iS*dE.dP;
+        
+        % and quantities for M-Step
         %------------------------------------------------------------------
-        if (dFdp'*dp < 1e-2) || (norm(dp,1) < TOL), break, end
+        EE    = E*E'+ EE;
+        ECE   = ECE + ECEu + ECEp;
+
         
-    end % E-Step
+    end % sequence (nY)
     
     
-    % M-step - hyperparameters (h = exp(l))
+    % M-step - optimise hyperparameters (mh = total update)
     %======================================================================
-    mh     = zeros(nh,1);
-    dFdh   = zeros(nh,1);
-    dFdhh  = zeros(nh,nh);
+    mh     = 0;                      
     for iM = 1:nM
- 
+        
         % [re-]set precisions using ReML hyperparameter estimates
         %------------------------------------------------------------------
         iS    = Qp;
         for i = 1:nh
-           iS = iS + Q{i}*exp(qh.h(i));
+            iS = iS + Q{i}*exp(qh.h(i));
         end
         S     = spm_inv(iS);
         dS    = ECE + EE - S*nY;
-         
+        
         % 1st-order derivatives: dFdh = dF/dh
         %------------------------------------------------------------------
         for i = 1:nh
-            dPdh{i}   = Q{i}*exp(qh.h(i));
+            dPdh{i}   =    Q{i}*exp(qh.h(i));
             dFdh(i,1) = -trace(dPdh{i}*dS)/2;
         end
- 
+        
         % 2nd-order derivatives: dFdhh
         %------------------------------------------------------------------
         for i = 1:nh
@@ -633,123 +564,157 @@
         
         % update ReML estimate of parameters
         %------------------------------------------------------------------
-        dh    = spm_dx(dFdhh,dFdh,{4});
+        dh    = spm_dx(dFdhh,dFdh,{tm});
+        dh    = max(min(dh,2),-2);
         qh.h  = qh.h + dh;
         mh    = mh   + dh;
         
         % conditional covariance of hyperparameters
         %------------------------------------------------------------------
         qh.c  = -spm_inv(dFdhh);
- 
+        
         % convergence (M-Step)
         %------------------------------------------------------------------
-        if (dFdh'*dh < 1e-2) || (norm(dh,1) < TOL), break, end
+        if (dFdh'*dh < TOL) || (norm(dh,1) < TOL), break, end
         
     end % M-Step
+    
  
+    % conditional precision of parameters
+    %------------------------------------------------------------------
+    qp.ic(ip,ip) = qp.ic(ip,ip) + pp.ic;
+    qp.c         = spm_inv(qp.ic);
+    
     % evaluate objective function (F)
     %======================================================================
-    Lu  = - trace(iS*EE)/2 - n*ny*log(2*pi)*nY/2 ... % states (u)
-          + spm_logdet(iS)*nY/2 ...                  % entropy - error
-          + spm_logdet_qu_c/(2*nD);                  % entropy q(u)
-          
-    Lp  = - trace(qp.e'*pp.ic*qp.e)/2  ...           % parameters (p)
-          - trace(qh.e'*ph.ic*qh.e)/2  ...           % hyperparameters (h)
-          + spm_logdet(qp.c)/2  ...                  % entropy q(p)
-          + spm_logdet(qh.c)/2  ...                  % entropy q(h)
-          - spm_logdet(pp.c)/2  ...                  % entropy - prior p
-          - spm_logdet(ph.c)/2;                      % entropy - prior h
-      
-    La  = - trace(qp.e'*pp.ic*qp.e)*nY/2  ...        % parameters (p)
-          - trace(qh.e'*ph.ic*qh.e)*nY/2  ...        % hyperparameters (h)
-          + spm_logdet(qp.c*nY)*nY/2 ...             % entropy q(p)
-          + spm_logdet(qh.c*nY)*nY/2 ...             % entropy q(h)
-          - spm_logdet(pp.c)*nY/2    ...             % entropy - prior p
-          - spm_logdet(ph.c)*nY/2;                   % entropy - prior h
-
-    Li  = Lu + Lp;                                   % free-energy
-    Ai  = Lu + La;                                   % free-action
     
-    % if F is increasing, save expansion point and derivatives
+    % free-energy and action
     %----------------------------------------------------------------------
-    if Li > Fi || iN < 3
-
- 
-        Fi    = Li;
-        F(iN) = Fi;
-        A(iN) = Ai;
- 
-        % save model-states (for each time point)
-        %==================================================================
-        for t = 1:length(qU)
-            v     = spm_unvec(qU(t).v{1},v);
-            x     = spm_unvec(qU(t).x{1},x);
-            z     = spm_unvec(qE{t}(1:(ny + nv)),{M.v});
-            w     = spm_unvec(qE{t}([1:nx] + (ny + nv)*n),{M.x});
-            for i = 1:(nl - 1)
-                if M(i).m, QU.v{i + 1}(:,t) = spm_vec(v{i}); end
-                if M(i).n, QU.x{i}(:,t)     = spm_vec(x{i}); end
-                if M(i).n, QU.w{i}(:,t)     = spm_vec(w{i}); end
-                if M(i).l, QU.z{i}(:,t)     = spm_vec(z{i}); end
-            end
-            QU.v{1}(:,t)  = spm_vec(qU(t).y{1}) - spm_vec(z{1});
-            if M(nl).l, QU.z{nl}(:,t) = spm_vec(z{nl});      end
- 
-            % and conditional covariances
-            %--------------------------------------------------------------
-            i       = [1:nx];
-            QU.S{t} = qU(t).c(i,i);
-            i       = [1:nv] + nx*n;
-            QU.C{t} = qU(t).c(i,i);
-        end
+    Lu  = - trace(iS(je,je)*EE(je,je))/2 ...          % states (u)
+          - n*ny*log(2*pi)*nY/2 ...                   % constant
+          + spm_logdet(iS(je,je))*nY/2 ...            % entropy - error
+          + iqu.c/(2*nD);                             % entropy q(u)
+    
+    Lp  = - trace(qp.e'*pp.ic*qp.e)/2  ...            % parameters (p)
+          - trace(qh.e'*ph.ic*qh.e)/2  ...            % hyperparameters (h)
+          + spm_logdet(qp.c(ip,ip)*pp.ic)/2   ...     % entropy q(p)
+          + spm_logdet(qh.c*ph.ic)/2;                 % entropy q(h)
+    
+    La  = - trace(qp.e'*pp.ic*qp.e)*nY/2   ...        % parameters (p)
+          - trace(qh.e'*ph.ic*qh.e)*nY/2   ...        % hyperparameters (h)
+          + spm_logdet(qp.c(ip,ip)*pp.ic*nY)*nY/2 ... % entropy q(p)
+          + spm_logdet(qh.c*ph.ic*nY)*nY/2;           % entropy q(h)
+    
+    Li  = Lu + Lp;                                    % free-energy
+    Ai  = Lu + La;                                    % free-action
+    
+    
+    % if F is increasing, save expansion point and derivatives
+    %------------------------------------------------------------------
+    if Li > Fi || iE < 2
+        
         
-        % save condotional densities
+        % Accept free-energy and save current parameter estimates
         %------------------------------------------------------------------
-        B.QU   = QU;
-        B.qp   = qp;
-        B.qh   = qh;
- 
-        % report and break if convergence
+        Fi      = Li;
+        te      = min(te + 1/2,4);
+        tm      = min(tm + 1/2,4);
+        B.qp    = qp;
+        B.qh    = qh;
+        B.pp    = pp;
+        
+        % E-step: update expectation (p)
+        %==================================================================
+        
+        % gradients and curvatures
         %------------------------------------------------------------------
-        figure(Fdem)
-        spm_DEM_qU(QU)
-        if np
-            subplot(2*nl,2,4*nl)
-            bar(full(Up*qp.e))
-            xlabel({'parameters {minus prior}'})
-        end
-        if nh
-            subplot(2*nl,4,8*nl - 4)
-            bar(full(qh.h))
-            title({'log-precision'})
-        end
-        if length(F) > 2
-            subplot(2*nl,4,8*nl - 5)
-            plot(F - F(1))
-            xlabel('updates')
-            title('free-energy')
-        end
-        drawnow
- 
-        % report (EM-Steps)
+        dFdp(ip)     = dFdp(ip)     - pp.ic*(qp.e - pp.p);
+        dFdpp(ip,ip) = dFdpp(ip,ip) - pp.ic;
+        
+        % update conditional expectation
         %------------------------------------------------------------------
-        str{1} = sprintf('DEM: %i (%i:%i:%i)',iN,iD,iE,iM);
-        str{2} = sprintf('F:%.4e',full(F(iN) - F(1)));
-        str{3} = sprintf('p:%.2e',full(mp'*mp));
-        str{4} = sprintf('h:%.2e',full(mh'*mh));
-        str{5} = sprintf('(%.2e sec)',full(toc));
-        fprintf('%-16s%-16s%-14s%-14s%-16s\n',str{:})
- 
+        dp      = spm_dx(dFdpp,dFdp,{te});
+        qp.e    = qp.e + dp(ip);
+        qp.p    = spm_unvec(qp.e,qp.p);
+        qp.b    = spm_unvec(dp(ib),qp.b);
+
     else
- 
-        % otherwise, return to previous expansion point and break
+        
+        % otherwise, return to previous expansion point
         %------------------------------------------------------------------
-        QU   = B.QU;
-        qp   = B.qp;
-        qh   = B.qh;
-        break
+        nM      = 1;
+        qp      = B.qp;
+        pp      = B.pp;
+        qh      = B.qh;
+        te      = min(te - 2, -2);
+        tm      = min(tm - 2, -2); 
         
     end
+    F(iE)   = Fi;
+    A(iE)   = Ai;
+        
+    
+    % save model-states (for each time point)
+    %==================================================================
+    for t = 1:length(qU)
+        v     = spm_unvec(qU(t).v{1},v);
+        x     = spm_unvec(qU(t).x{1},x);
+        z     = spm_unvec(qE{t}(1:(ny + nv)),{M.v});
+        w     = spm_unvec(qE{t}([1:nx] + (ny + nv)*n),{M.x});
+        for i = 1:(nl - 1)
+            if M(i).m, QU.v{i + 1}(:,t) = spm_vec(v{i}); end
+            if M(i).n, QU.x{i}(:,t)     = spm_vec(x{i}); end
+            if M(i).n, QU.w{i}(:,t)     = spm_vec(w{i}); end
+            if M(i).l, QU.z{i}(:,t)     = spm_vec(z{i}); end
+        end
+        QU.v{1}(:,t)  = spm_vec(qU(t).y{1}) - spm_vec(z{1});
+        if M(nl).l, QU.z{nl}(:,t) = spm_vec(z{nl});      end
+        
+        % and conditional covariances
+        %--------------------------------------------------------------
+        i       = (1:nx);
+        QU.S{t} = qU(t).c(i,i);
+        i       = (1:nv) + nx*n;
+        QU.C{t} = qU(t).c(i,i);
+    end
+    
+    % report and break if convergence
+    %------------------------------------------------------------------
+    figure(Fdem)
+    spm_DEM_qU(QU)
+    if np
+        subplot(2*nl,2,4*nl)
+        bar(full(Up*qp.e))
+        xlabel({'parameters {minus prior}'})
+    end
+    if nh
+        subplot(2*nl,4,8*nl - 4)
+        bar(full(qh.h))
+        title({'log-precision'})
+    end
+    if length(F) > 2
+        subplot(2*nl,4,8*nl - 5)
+        plot(F - F(1))
+        xlabel('updates')
+        title('free-energy')
+    end
+    drawnow
+    
+    % report (EM-Steps)
+    %------------------------------------------------------------------
+    str{1} = sprintf('DEM: %i (%i:%i)',iE,iD,iM);
+    str{2} = sprintf('F:%.4e',full(F(iE) - F(1)));
+    str{3} = sprintf('p:%.2e',full(norm(dp,1)));
+    str{4} = sprintf('h:%.2e',full(norm(mh,1)));
+    str{5} = sprintf('(%.2e sec)',full(toc));
+    fprintf('%-16s%-16s%-14s%-14s%-16s\n',str{:})
+    
+    
+    % Convergence
+    %------------------------------------------------------------------
+    if (norm(dp,1) < TOL*norm(spm_vec(qp.p),1)) && (norm(mh,1) < TOL), break, end
+    if te < -8, break, end
+    
 end
  
 % Assemble output arguments
@@ -757,9 +722,11 @@
  
 % conditional moments of model-parameters (rotated into original space)
 %--------------------------------------------------------------------------
-qP.P   = spm_unvec(Up*qp.e + spm_vec(M.pE),M.pE);
-qP.C   = Up*qp.c(ip,ip)*Up';
-qP.V   = spm_unvec(diag(qP.C),M.pE);
+qP.P     = spm_unvec(Up*qp.e + spm_vec(M.pE),M.pE);
+qP.C     = Up*qp.c(ip,ip)*Up';
+qP.V     = spm_unvec(diag(qP.C),M.pE);
+qP.dFdp  = Up*dFdp(ip);
+qP.dFdpp = Up*dFdpp(ip,ip)*Up';
  
 % conditional moments of hyper-parameters (log-transformed)
 %--------------------------------------------------------------------------
diff --git a/spm_DEM_F.m b/spm_DEM_F.m
new file mode 100644
index 0000000..f3a4c9a
--- /dev/null
+++ b/spm_DEM_F.m
@@ -0,0 +1,70 @@
+function [F,p] = spm_DEM_F(DEM,ip)
+% Free-energy as a function of conditional parameters
+% [F,P1,P2] = spm_DEM_F(DEM))
+%
+% DEM    - hierarchical model
+%
+% F(i) - free-energy at <q(P(ip))> = p(i)
+%
+% where p(i) is the ip-th free-parameter. This is a bound on 
+% the log-likehood (log-evidence) conditioned on the expeceted parameters
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_DEM_F.m 4146 2010-12-23 21:01:39Z karl $
+ 
+ 
+% Find paramter ranges (using prior covariance)
+%--------------------------------------------------------------------------
+pE   = spm_vec(DEM.M(1).pE);
+p    = linspace(-6,6,16);
+dp   = sqrt(DEM.M(1).pC(ip,ip))*p;
+p    = dp + pE(ip);
+ 
+% get F
+%==========================================================================
+DEM.M(1).E.nE = 1;
+DEM.M(1).E.nN = 1;
+
+for i = 1:length(p)
+    
+    % adjust paramter (through prio expecatation)
+    %----------------------------------------------------------------------
+    P          = pE;
+    P(ip)      = p(i);
+    DEM.M(1).P = spm_unvec(P,DEM.M(1).pE);
+    
+    % comute free-energy
+    %----------------------------------------------------------------------
+    DEM  = spm_DEM(DEM);
+    F(i) = DEM.F(end);
+    
+end
+
+
+% predicted F under the Laplace assumption
+%==========================================================================
+DEM.M(1).P = DEM.M(1).pE;
+DEM        = spm_DEM(DEM);
+
+% compute free-energy
+%--------------------------------------------------------------------------
+dFdp   = DEM.qP.dFdp(ip);
+dFdpp  = DEM.qP.dFdpp(ip,ip);
+FP     = dFdp*dp + (dFdpp*dp.^2)/2;
+FP     = FP - max(FP);
+    
+
+% plot
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Free-energy');
+F    = F - max(F);
+subplot(2,1,1)
+plot(p,F,p,FP,':'), hold on
+plot(pE(ip)*[1 1],[min(F) 0],':'),  hold on
+xlabel('Parameter expecatation','FontSize',12)
+ylabel('Free-energy','FontSize',12)
+title('Free-energy profile','FontSize',16)
+axis square, box off
+
diff --git a/spm_DEM_M_set.m b/spm_DEM_M_set.m
index 19d4942..6445975 100644
--- a/spm_DEM_M_set.m
+++ b/spm_DEM_M_set.m
@@ -48,7 +48,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_DEM_M_set.m 3733 2010-02-18 17:43:18Z karl $
+% $Id: spm_DEM_M_set.m 4146 2010-12-23 21:01:39Z karl $
 
 % order
 %--------------------------------------------------------------------------
@@ -61,7 +61,6 @@
 %--------------------------------------------------------------------------
 if isfield(M,'f') && ~isfield(M,'n') && ~isfield(M,'x')
     msgbox('please specify hidden states or their number')
-    error(' ')
 end
  
  
@@ -151,7 +150,6 @@
     %----------------------------------------------------------------------
     if length(M(i).pC) ~= np
         warndlg(sprintf('please check: M(%i).pC',i))
-        who,keyboard
     end
  
 end
@@ -199,12 +197,10 @@
         f       = feval(M(i).f,x,v,M(i).pE);
         if length(spm_vec(x)) ~= length(spm_vec(f))
             warndlg(sprintf('please check: M(%i).f(x,v,P)',i));
-            who,keyboard
         end
  
     catch
         warndlg(sprintf('evaluation failure: M(%i).f(x,v,P)',i))
-        who,keyboard
     end
     try M(i).fx = fcnchk(M(i).fx,'x','v','P'); end
     try M(i).fv = fcnchk(M(i).fv,'x','v','P'); end
@@ -227,7 +223,6 @@
  
     catch
         warndlg(sprintf('evaluation failure: M(%i).g(x,v,P)',i))
-        who,keyboard
     end
     try M(i).gx = fcnchk(M(i).gx,'x','v','P'); end
     try M(i).gv = fcnchk(M(i).gv,'x','v','P'); end
@@ -249,7 +244,6 @@
         M(i).xP = sparse(M(i).n,M(i).n);
     elseif any(size(M(i).xP) ~= [M(i).n M(i).n]);
         warndlg(sprintf('please Check: M(%i).xP',i))
-        who,keyboard
     end
 end
 
@@ -271,6 +265,7 @@
 
 % check hyperpriors hE - [log]hyper-parameters and components
 %--------------------------------------------------------------------------
+pP    = 1;               % prior precision on log-precisions
 for i = 1:g
     
     
@@ -294,11 +289,11 @@
     
     % check hyperpriors (covariances)
     %----------------------------------------------------------------------
-    try, M(i).hC*M(i).hE; catch, M(i).hC = speye(length(M(i).hE)); end
-    try, M(i).gC*M(i).gE; catch, M(i).gC = speye(length(M(i).gE)); end
+    try, M(i).hC*M(i).hE; catch, M(i).hC = speye(length(M(i).hE))/pP; end
+    try, M(i).gC*M(i).gE; catch, M(i).gC = speye(length(M(i).gE))/pP; end
     
-    if isempty(M(i).hC), M(i).hC = speye(length(M(i).hE)); end
-    if isempty(M(i).gC), M(i).gC = speye(length(M(i).gE)); end
+    if isempty(M(i).hC), M(i).hC = speye(length(M(i).hE))/pP; end
+    if isempty(M(i).gC), M(i).gC = speye(length(M(i).gE))/pP; end
     
     % check Q and R (precision components)
     %======================================================================
@@ -332,7 +327,6 @@
     for j = 1:length(M(i).Q)
         if length(M(i).Q{j}) ~= M(i).l
             warndlg(sprintf('wrong size; M(%d).Q{%d}',i,j))
-            who,keyboard
         end
     end
     
@@ -341,7 +335,6 @@
     for j = 1:length(M(i).R)
         if length(M(i).R{j}) ~= M(i).n
             warndlg(sprintf('wrong size; M(%d).R{%d}',i,j))
-            who,keyboard
         end
     end
     
@@ -397,17 +390,17 @@
  
 % embedding orders
 %--------------------------------------------------------------------------
-try M(1).E.d;  catch, if nx, M(1).E.d = 2;  else M(1).E.d = 0;  end, end
-try M(1).E.n;  catch, if nx, M(1).E.n = 6;  else M(1).E.n = 0;  end, end
+try M(1).E.d;  catch, if nx, M(1).E.d = 2; else M(1).E.d = 0;  end, end
+try M(1).E.n;  catch, if nx, M(1).E.n = 6; else M(1).E.n = 0;  end, end
 
 M(1).E.d = min(M(1).E.d,M(1).E.n);
  
 % number of iterations
 %--------------------------------------------------------------------------
-try M(1).E.nD; catch, if nx, M(1).E.nD = 1;  else M(1).E.nD = 8; end, end
-try M(1).E.nE; catch,        M(1).E.nE = 1;  end
-try M(1).E.nM; catch,        M(1).E.nM = 8;  end
-try M(1).E.nN; catch,        M(1).E.nN = 16; end
+try M(1).E.nD; catch, if nx, M(1).E.nD = 1; else M(1).E.nD = 8; end, end
+try M(1).E.nE; catch,        M(1).E.nE = 8; end
+try M(1).E.nM; catch,        M(1).E.nM = 8; end
+try M(1).E.nN; catch,        M(1).E.nN = 8; end
  
 % checks on smoothness hyperparameter
 %==========================================================================
diff --git a/spm_DEM_R.m b/spm_DEM_R.m
index 4b3a7ef..173a89d 100644
--- a/spm_DEM_R.m
+++ b/spm_DEM_R.m
@@ -1,6 +1,6 @@
-function [R,V,D] = spm_DEM_R(n,s,form)
+function [R,V] = spm_DEM_R(n,s,form)
 % returns the precision of the temporal derivatives of a Gaussian process
-% FORMAT [R,V,D] = spm_DEM_R(n,s,form)
+% FORMAT [R,V] = spm_DEM_R(n,s,form)
 %__________________________________________________________________________
 % n    - truncation order
 % s    - temporal smoothness - s.d. of kernel {bins}
@@ -14,12 +14,11 @@
 %
 % R    - (n x n)     E*V*E: precision of n derivatives
 % V    - (n x n)     V:    covariance of n derivatives
-% D    - (n x n)     D:    projector:   Y(t) <- D*y[:]
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_DEM_R.m 3878 2010-05-07 19:53:54Z karl $
+% $Id: spm_DEM_R.m 4278 2011-03-31 11:48:00Z karl $
 
 % if no serial dependencies 
 %--------------------------------------------------------------------------
@@ -31,7 +30,7 @@
 
 % temporal correlations (assuming known form) - V
 %--------------------------------------------------------------------------
-try, form;, catch, form = 'Gaussian'; end
+try, form; catch, form = 'Gaussian'; end
 
 switch form
 
@@ -97,7 +96,7 @@
 
 % NB: temporal correlations (assuming stationary Gaussian form)
 %--------------------------------------------------------------------------
-t     = ([1:n] - 1)*dt;
+t     = ((1:n) - 1)*dt;
 v     = 2*(s^2);
 V     = exp(-t.^2/(2*v));
 V     = toeplitz(V);
diff --git a/spm_DEM_diff.m b/spm_DEM_diff.m
index 2eab3b8..a6e4617 100644
--- a/spm_DEM_diff.m
+++ b/spm_DEM_diff.m
@@ -17,7 +17,7 @@
 % Copyright (C) 2005 Wellcome Department of Imaging Neuroscience
  
 % Karl Friston
-% $Id: spm_DEM_diff.m 3655 2009-12-23 20:15:34Z karl $
+% $Id: spm_DEM_diff.m 4247 2011-03-14 18:16:50Z karl $
 
 % Check for action (ADEM)
 %==========================================================================
diff --git a/spm_DEM_qH.m b/spm_DEM_qH.m
index e34e85b..7c7e7b0 100644
--- a/spm_DEM_qH.m
+++ b/spm_DEM_qH.m
@@ -15,14 +15,13 @@ function spm_DEM_qH(qH,pH)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_DEM_qH.m 3878 2010-05-07 19:53:54Z karl $
+% $Id: spm_DEM_qH.m 4052 2010-08-27 19:22:44Z karl $
  
 % unpack conditional covariances
 %--------------------------------------------------------------------------
 try, qH = qH.qH; end
 try, pH = pH.pH; end
 
-
 % [Re]ML estimates - h
 %==========================================================================
 ci = spm_invNcdf(1 - 0.05);
@@ -64,35 +63,39 @@ function spm_DEM_qH(qH,pH)
 h = spm_vec(qH.g);
 c = spm_vec(qH.W);
 c = sqrt(c)*spm_invNcdf(1 - 0.05);
-subplot(2,2,2)
-bar(full(h),'Edgecolor',[1 1 1]/2,'Facecolor',[1 1 1]*.8)
-title({'log-precision';'states'},'FontSize',16);
-axis square
-set(gca,'XLim',[0 length(c) + 1])
- 
-% conditional covariances
-%--------------------------------------------------------------------------
-for i = 1:length(c)
-    line([i i],[-1 1]*c(i) + h(i),'LineWidth',4,'Color','r');
-end
-
-% prior or true means
-%--------------------------------------------------------------------------
-try
-    p     = spm_vec(pH.g);
-    for i = 1:length(h)
-        line([-1 1]/2 + i,[0 0] + p(i),'LineWidth',4,'Color','k');
+if h
+    subplot(2,2,2)
+    bar(full(h),'Edgecolor',[1 1 1]/2,'Facecolor',[1 1 1]*.8)
+    title({'log-precision';'states'},'FontSize',16);
+    axis square
+    set(gca,'XLim',[0 length(c) + 1])
+    
+    % conditional covariances
+    %----------------------------------------------------------------------
+    for i = 1:length(c)
+        line([i i],[-1 1]*c(i) + h(i),'LineWidth',4,'Color','r');
     end
-end
-
-glabel = {};
-for i = 1:length(qH.h)
-    for j = 1:length(qH.h{i})
-        glabel{end + 1} = sprintf('g:level %i',i);
+    
+    % prior or true means
+    %----------------------------------------------------------------------
+    try
+        p     = spm_vec(pH.g);
+        for i = 1:length(h)
+            line([-1 1]/2 + i,[0 0] + p(i),'LineWidth',4,'Color','k');
+        end
     end
+    
+    glabel = {};
+    for i = 1:length(qH.h)
+        for j = 1:length(qH.h{i})
+            glabel{end + 1} = sprintf('g:level %i',i);
+        end
+    end
+    set(gca,'XTickLabel',glabel)
+else
+    glabel = {};
 end
-set(gca,'XTickLabel',glabel)
- 
+
 % conditional covariance and correlations
 %--------------------------------------------------------------------------
 subplot(2,2,3)
diff --git a/spm_DEM_qU.m b/spm_DEM_qU.m
index 526d053..3ce9c4e 100644
--- a/spm_DEM_qU.m
+++ b/spm_DEM_qU.m
@@ -13,7 +13,7 @@ function spm_DEM_qU(qU,pU)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_DEM_qU.m 3878 2010-05-07 19:53:54Z karl $
+% $Id: spm_DEM_qU.m 4247 2011-03-14 18:16:50Z karl $
  
 % unpack
 %--------------------------------------------------------------------------
@@ -31,17 +31,23 @@ function spm_DEM_qU(qU,pU)
     pV = pU.v;
     pX = pU.x;
 end
+ 
+% order of hierarchy
+%--------------------------------------------------------------------------
 try
-    pA = qU.a;
+    g = length(X) + 1;                           
+    if issempty(X{end})
+        g = g - 1;
+    end
+catch
+    g = length(V);
 end
- 
+
 % time-series specification
 %--------------------------------------------------------------------------
-g     = length(V);          % order of hierarchy
-N     = size(V{1},2);       % length of data sequence
-dt    = 1;                  % time step
-t     = [1:N]*dt;           % time
- 
+N     = size(V{1},2);                            % length of data sequence
+dt    = 1;                                       % time step
+t     = [1:N]*dt;                                % time
  
 % unpack conditional covariances
 %--------------------------------------------------------------------------
@@ -61,7 +67,7 @@ function spm_DEM_qU(qU,pU)
  
     if N == 1
  
-        % causal states and error - single observation
+        % hidden causes and error - single observation
         %------------------------------------------------------------------
         subplot(g,2,2*i - 1)
         t = 1:size(V{i},1);
@@ -77,7 +83,7 @@ function spm_DEM_qU(qU,pU)
             y      = ci*c(j,:);
             c(j,:) = [];
             fill([t fliplr(t)],[full(V{i} + y)' fliplr(full(V{i} - y)')],...
-                 [1 1 1]*.8,'EdgeColor',[1 1 1]*.6)
+                 [1 1 1]*.8,'EdgeColor',[1 1 1]*.8)
             plot(t,full(E{i}),'r:',t,full(V{i}))
             hold off
         end
@@ -91,17 +97,20 @@ function spm_DEM_qU(qU,pU)
  
     else
  
-        % causal states and error - time series
+        % hidden causes and error - time series
         %------------------------------------------------------------------
         subplot(g,2,2*i - 1)
         try
-            plot(t,pV{i},':k','linewidth',1),box off
+            plot(t,pV{i},':k','linewidth',1)
         end
         hold on
         try
-            plot(t,full(E{i}(:,1:N)),'r:',t,full(V{i})),box off
+            plot(t,full(V{i}))
         end
-        hold off
+        try
+            plot(t,full(E{i}),'r:')
+        end
+        box off, hold off
         set(gca,'XLim',[t(1) t(end)])
         a   = axis;
  
@@ -109,15 +118,18 @@ function spm_DEM_qU(qU,pU)
         %------------------------------------------------------------------
         if i > 1 && size(c,1)
             hold on
-            j      = [1:size(V{i},1)];
+            j      = (1:size(V{i},1));
             y      = ci*c(j,:);
             c(j,:) = [];
             fill([t fliplr(t)],[full(V{i} + y) fliplr(full(V{i} - y))],...
-                        [1 1 1]*.8,'EdgeColor',[1 1 1]*.6)
+                        [1 1 1]*.8,'EdgeColor',[1 1 1]*.8)
             try 
-                plot(t,pV{i},':k','linewidth',1),box off
+                plot(t,pV{i},':k','linewidth',1)
+            end
+            try
+                plot(t,full(E{i}),'r:')
             end
-            plot(t,full(E{i}(:,1:N)),'r:',t,full(V{i}))
+            plot(t,full(V{i})),box off
             hold off
         end
  
@@ -125,6 +137,8 @@ function spm_DEM_qU(qU,pU)
         %------------------------------------------------------------------
         if i == 1
             title('prediction and error','FontSize',16);
+        elseif length(V) < i
+            title('no causes','FontSize',16);
         elseif ~size(V{i},1)
             title('no causes','FontSize',16);
         else
@@ -134,11 +148,6 @@ function spm_DEM_qU(qU,pU)
                 plot(t,pV{i},':k','linewidth',1),box off
             end
             hold off
-            try
-                hold on
-                plot(t,pA{i - 1},'linewidth',1,'color',[1 0 0]),box off
-            end
-            hold off
         end
         xlabel('time','FontSize',14)
         axis square
@@ -189,14 +198,16 @@ function spm_DEM_qU(qU,pU)
 %--------------------------------------------------------------------------
 if isfield(qU,'a')
     subplot(g,2,2*g)
-    plot(t,qU.a{2});
+    plot(t,qU.a{end});
+    str = 'action';
     try
         hold on
         plot(t,pU.v{2},':b','Linewidth',2),box off
+        str = 'perturbation and action';
     end
     hold off
     xlabel('time','Fontsize',14)
-    title('perturbation and action','Fontsize',16)
+    title(str,'Fontsize',16)
     axis square
     set(gca,'XLim',[t(1) t(end)])
     box off
diff --git a/spm_DesMtx.m b/spm_DesMtx.m
index d224f79..1e15ab1 100644
--- a/spm_DesMtx.m
+++ b/spm_DesMtx.m
@@ -1,4 +1,4 @@
-function [X,Pnames,Index,idx,jdx,kdx]=spm_DesMtx(varargin);
+function [X,Pnames,Index,idx,jdx,kdx]=spm_DesMtx(varargin)
 % Design matrix construction from factor level and covariate vectors
 % FORMAT [X,Pnames] = spm_DesMtx(<FCLevels-Constraint-FCnames> list)
 % FORMAT [X,Pnames,Index,idx,jdx,kdx] = spm_DesMtx(FCLevels,Constraint,FCnames)
@@ -250,7 +250,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_DesMtx.m 3934 2010-06-17 14:58:25Z guillaume $
+% $Id: spm_DesMtx.m 4137 2010-12-15 17:18:32Z guillaume $
 
 
 
@@ -261,12 +261,12 @@
 if ischar(varargin{1})
     %-Non-recursive action string usage
     Constraint=varargin{1};
-elseif nargin>=2 & ~(ischar(varargin{2}) | iscell(varargin{2}))
+elseif nargin>=2 && ~(ischar(varargin{2}) || iscell(varargin{2}))
     [X1,Pnames1]=spm_DesMtx(varargin{1});
     [X2,Pnames2]=spm_DesMtx(varargin{2:end});
     X=[X1,X2]; Pnames=[Pnames1;Pnames2];
     return
-elseif nargin>=3 & ~(ischar(varargin{3}) | iscell(varargin{3}))
+elseif nargin>=3 && ~(ischar(varargin{3}) || iscell(varargin{3}))
     [X1,Pnames1]=spm_DesMtx(varargin{1:2});
     [X2,Pnames2]=spm_DesMtx(varargin{3:end});
     X=[X1,X2]; Pnames=[Pnames1;Pnames2];
@@ -280,9 +280,9 @@
     %-If I is a vector, make it a column vector
     I=varargin{1}; if size(I,1)==1, I=I'; end
     %-Sort out constraint and Factor/Covariate name parameters
-    if nargin<2, Constraint='-'; else, Constraint=varargin{2}; end
+    if nargin<2, Constraint='-'; else Constraint=varargin{2}; end
     if isempty(I), Constraint='mt'; end
-    if nargin<3, FCnames={}; else, FCnames=varargin{3}; end
+    if nargin<3, FCnames={}; else FCnames=varargin{3}; end
     if char(FCnames), FCnames=cellstr(FCnames); end
 end
 
@@ -304,10 +304,10 @@
 %-Construct parameter name index
 %-----------------------------------------------------------------------
 if isempty(FCnames)
-    if strcmp(Constraint,'C'), FCnames={'<Cov>'}; else, FCnames={'<X>'}; end
+    if strcmp(Constraint,'C'), FCnames={'<Cov>'}; else FCnames={'<X>'}; end
 end
 
-if length(FCnames)==1 & size(X,2)>1
+if length(FCnames)==1 && size(X,2)>1
     Pnames = cell(size(X,2),1);
     for i=1:size(X,2)
         Pnames{i} = sprintf('%s [%d]',FCnames{1},i);
@@ -443,7 +443,7 @@
 %-Set up design matrix X & names matrix - ignore zero levels if '~FxC' use
 %-----------------------------------------------------------------------
 if Constraint(1)~='~',  [X,Pnames,Index] = spm_DesMtx(F,'-',Fnames);
-    else,       [X,Pnames,Index] = spm_DesMtx(F,'~',Fnames); end
+    else       [X,Pnames,Index] = spm_DesMtx(F,'~',Fnames); end
 X = X.*(C*ones(1,size(X,2)));
 Pnames = cellstr([repmat([Cnames,'@'],size(Index,2),1),char(Pnames)]);
 
@@ -463,7 +463,7 @@
 %-'.0' corner point constraint is applied to zero factor level only
 nXcols = size(X,2);
 zCol   = find(Index==0);
-if nXcols==1 & ~strcmp(Constraint,'.0')
+if nXcols==1 && ~strcmp(Constraint,'.0')
     error('only one level: can''t constrain')
 elseif strcmp(Constraint,'.')
     X(:,nXcols)=[]; Pnames(nXcols)=[]; Index(nXcols)=[];
@@ -572,7 +572,7 @@
 
     if CornerPointI %-impose CornerPointI constraints
         if Constraint(end)~='0',    i = max(Index(1,:));
-            else,           i = 0; end
+            else           i = 0; end
         cols=find(Index(1,:)==i); %-columns to delete
         if isempty(cols)
             warning('no zero i level to constrain')
@@ -585,7 +585,7 @@
 
     if CornerPointJ %-impose CornerPointJ constraints
         if Constraint(end)~='0',    j = max(Index(2,:));
-            else,           j = 0; end
+            else           j = 0; end
         cols=find(Index(2,:)==j);
         if isempty(cols)
             warning('no zero j level to constrain')
@@ -600,9 +600,9 @@
 case {'PDS','pds'}                          %-Patterned data set utility
 %=======================================================================
 % i = spm_DesMtx('pds',v,m,n)
-if nargin<4, n=1; else, n=varargin{4}; end
-if nargin<3, m=1; else, m=varargin{3}; end
-if nargin<2, varargout={[]}, return, else, v=varargin{2}; end
+if nargin<4, n=1; else n=varargin{4}; end
+if nargin<3, m=1; else m=varargin{3}; end
+if nargin<2, varargout={[]}; return, else v=varargin{2}; end
 if any([size(n),size(m)])>1, error('n & m must be scalars'), end
 if any(([m,n]~=floor([m,n]))|([m,n]<1))
     error('n & m must be natural numbers'), end
@@ -623,8 +623,8 @@
 %-----------------------------------------------------------------------
 while(Carg <= nargin)
     rX = varargin{Carg}; Carg=Carg+1;
-    if Carg<=nargin & ~isempty(varargin{Carg}) & ...
-            (ischar(varargin{Carg}) | iscellstr(varargin{Carg}))
+    if Carg<=nargin && ~isempty(varargin{Carg}) && ...
+            (ischar(varargin{Carg}) || iscellstr(varargin{Carg}))
     rPnames = char(varargin{Carg}); Carg=Carg+1;
     else    %-No names to work out blocks from - normalise by column
     rPnames = repmat('<UnSpec>',size(rX,2),1);
@@ -634,7 +634,7 @@
 
 
     while(~isempty(rX))
-    if size(rX,2)>1 & max([1,find(rPnames(1,:)=='(')]) < ...
+    if size(rX,2)>1 && max([1,find(rPnames(1,:)=='(')]) < ...
                     max([0,find(rPnames(1,:)==')')])
     %-Non-specific block: find the rest & column normalise round zero
     %===============================================================
@@ -658,7 +658,7 @@
         rX(:,1:t) = []; rPnames(1:t,:)=[];
 
 
-    elseif size(rX,2)>1 & max([1,find(rPnames(1,:)=='[')]) < ...
+    elseif size(rX,2)>1 && max([1,find(rPnames(1,:)=='[')]) < ...
                     max([0,find(rPnames(1,:)==']')])
     %-Block: find the rest & normalise together
     %===============================================================
@@ -674,7 +674,7 @@
         rX(:,1:t) = []; rPnames(1:t,:)=[];
 
 
-    elseif size(rX,2)>1 & max([1,strfind(rPnames(1,:),'_{')]) < ...
+    elseif size(rX,2)>1 && max([1,strfind(rPnames(1,:),'_{')]) < ...
                     max([0,find(rPnames(1,:)=='}')])
     %-Factor, interaction of factors, or FxC: find the rest...
     %===============================================================
@@ -754,9 +754,9 @@
     '\rho',     'covint';...
     '\zeta',    'global';...
     '\epsilon', 'error'};
-if nargin<2, aMap={}; else, aMap = varargin{2}; end
+if nargin<2, aMap={}; else aMap = varargin{2}; end
 if isempty(aMap), X=Map; return, end
-if ~(iscellstr(aMap) & ndims(aMap)==2), error('aMap must be an nx2 cellstr'), end
+if ~(iscellstr(aMap) && ndims(aMap)==2), error('aMap must be an nx2 cellstr'), end
 for i=1:size(aMap,1)
     j = find(strcmp(aMap{i,1},Map(:,1)));
     if isempty(j)
@@ -772,7 +772,7 @@
 %=======================================================================
 % EPnames = spm_DesMtx('TeXnames',Pnames,aMap)
 if nargin<2, varargout={''}; return, end
-if nargin<3, aMap={}; else, aMap = varargin{3}; end
+if nargin<3, aMap={}; else aMap = varargin{3}; end
 Map = spm_DesMtx('ParMap',aMap);
 EPnames = varargin{2};
 for i=1:size(Map,1)
diff --git a/spm_DesRep.m b/spm_DesRep.m
index 43a230f..c618ec0 100644
--- a/spm_DesRep.m
+++ b/spm_DesRep.m
@@ -128,7 +128,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_DesRep.m 1157 2008-02-19 11:34:58Z guillaume $
+% $Id: spm_DesRep.m 4209 2011-02-22 20:11:05Z guillaume $
 
 
 
@@ -262,7 +262,13 @@
 
 %-Format arguments
 %-----------------------------------------------------------------------
-if nargin==0, spm_DesRep('DesRepUI'); return, end
+if nargin==0
+    hC = spm_DesRep('DesRepUI'); 
+    SPM = get(hC,'UserData');
+    spm_DesRep('DesMtx',SPM.xX,...
+            reshape({SPM.xY.VY.fname},size(SPM.xY.VY)),SPM.xsDes);
+    return;
+end
 
 switch lower(varargin{1})
 
@@ -293,10 +299,10 @@
 %=======================================================================
 %-Work out where design configuration has come from!
 if ~isfield(SPM,'cfg')
-    if     isfield(SPM.xX,'V'),     cfg = 'SPMest';
-    elseif isfield(SPM.xY,'VY'),        cfg = 'SPMdata';
-    elseif isfield(SPM,'Sess'),     cfg = 'SPMcfg';
-    else, error('Can''t fathom origin!')
+    if     isfield(SPM.xX,'V'),  cfg = 'SPMest';
+    elseif isfield(SPM.xY,'VY'), cfg = 'SPMdata';
+    elseif isfield(SPM,'Sess'),  cfg = 'SPMcfg';
+    else   error('Can''t fathom origin!')
     end
     SPM.cfg = cfg;
 end
@@ -397,8 +403,10 @@
 
 hxvi = uimenu(hExplore, 'Label','Covariance structure', ...
     'Callback',[cb, 'spm_DesRep(''xVi'', tmp.xVi);'], 'UserData',hC, 'HandleVisibility','off');
-if ~isfield(SPM,'xVi') | (isfield(SPM.xVi,'iid') & SPM.xVi.iid) | ...
-    ~(isfield(SPM.xVi,'V')|isfield(SPM.xVi,'Vi')), set(hxvi,'Enable','off'); end;
+if ~isfield(SPM,'xVi') || (isfield(SPM.xVi,'iid') && SPM.xVi.iid) || ...
+    ~(isfield(SPM.xVi,'V') || isfield(SPM.xVi,'Vi'))
+    set(hxvi,'Enable','off');
+end
 
 %-Clear, Quit, Help
 %-----------------------------------------------------------------------
@@ -427,7 +435,7 @@
 I       = varargin{3};
 xC      = varargin{4};
 sF      = varargin{5};
-if nargin<6, xs=[]; else, xs = varargin{6}; end %-Structure of strings
+if nargin<6, xs=[]; else xs = varargin{6}; end %-Structure of strings
 
 [fnames,CPath] = spm_str_manip(fnames,'c'); %-extract common path component
 nScan          = size(I,1);         %-#images
@@ -460,7 +468,7 @@
 
 for j = 1:length(xC)
     n = size(xC(j).rc,2);
-    if n>1, tmp=xC(j).cname; else, tmp={xC(j).rcname}; end
+    if n>1, tmp=xC(j).cname; else tmp={xC(j).rcname}; end
     for k=1:n
         x=x+dx2;
         text(x,.1,tmp{k},'Rotation',90,'Interpreter','TeX')
@@ -532,9 +540,9 @@
         text(0.3,y,[strrep(sf{1},'_',' '),' :'],...
             'HorizontalAlignment','Right','FontWeight','Bold',...
             'FontSize',FS(9))
-        s = getfield(xs,sf{1});
+        s = xs.(sf{1});
         if ~iscellstr(s), s={s}; end
-        for i=1:prod(size(s))
+        for i=1:numel(s)
             text(0.31,y,s{i},'FontSize',FS(9))
             y=y-dy;
         end
@@ -601,23 +609,23 @@
     str = sprintf('Covariance structure V: %s',varargin{2}.form);
   else
     str = 'Covariance structure V';
-  end;
-  if isfield(varargin{2},'var') & isfield(varargin{2},'dep') & ...
-      isfield(varargin{2},'sF') & isfield(varargin{2},'I')
+  end
+  if isfield(varargin{2},'var') && isfield(varargin{2},'dep') && ...
+     isfield(varargin{2},'sF') && isfield(varargin{2},'I')
     r     = find((max(varargin{2}.I) > 1) & ~varargin{2}.var & ~varargin{2}.dep);
     if any(varargin{2}.dep)
       cmstr = 'yes';
     else
       cmstr = 'no';
-    end;
+    end
     str = sprintf(['Covariance structure V - replications over ''%s''\n'...
         'Correlated repeated measures: %s'], ...
       varargin{2}.sF{r},cmstr);
-  end;
+  end
   xlabel(str);
 else
   text(.5,.5, 'Covariance not (yet) estimated.', 'HorizontalAlignment','center');
-end;
+end
 
 if isfield(varargin{2},'h')
   hPEstAx   = axes('Position',[.07 .315 .6 .025],...
@@ -660,30 +668,30 @@
 % spm_DesRep('DesOrth',xX)
 if nargin<2, error('insufficient arguments'), end
 if ~isstruct(varargin{2}), error('design matrix structure required'), end
-if nargin<3, fnames={}; else, fnames=varargin{3}; end
-if nargin<4, xs=[]; else, xs=varargin{4}; end
+if nargin<3, fnames={}; else fnames=varargin{3}; end
+if nargin<4, xs=[]; else xs=varargin{4}; end
 
-desmtx = strcmp(lower(varargin{1}),'desmtx');
+desmtx = strcmpi(varargin{1},'desmtx');
 
 
 %-Locate DesMtx (X), scaled DesMtx (nX) & get parameter names (Xnames)
 %-----------------------------------------------------------------------
-if isfield(varargin{2},'xKXs') & ...
-        ~isempty(varargin{2}.xKXs) & isstruct(varargin{2}.xKXs)
+if isfield(varargin{2},'xKXs') && ...
+        ~isempty(varargin{2}.xKXs) && isstruct(varargin{2}.xKXs)
     iX = 1;
     [nScan,nPar] = size(varargin{2}.xKXs.X);
-elseif isfield(varargin{2},'X') & ~isempty(varargin{2}.X)
+elseif isfield(varargin{2},'X') && ~isempty(varargin{2}.X)
     iX = 0;
     [nScan,nPar] = size(varargin{2}.X);
 else
     error('Can''t find DesMtx in this structure!')
 end
 
-if isfield(varargin{2},'nKX') & ~isempty(varargin{2}.nKX)
-    inX = 1; else, inX = 0; end
+if isfield(varargin{2},'nKX') && ~isempty(varargin{2}.nKX)
+    inX = 1; else inX = 0; end
 
-if isfield(varargin{2},'name') & ~isempty(varargin{2}.name)
-    Xnames = varargin{2}.name; else, Xnames = {}; end
+if isfield(varargin{2},'name') && ~isempty(varargin{2}.name)
+    Xnames = varargin{2}.name; else Xnames = {}; end
 
 
 %-Compute design orthogonality matrix if DesOrth
@@ -767,7 +775,7 @@
 %-Filenames
 % ( Show at most 32, showing every 2nd/3rd/4th/... as necessary to pair )
 % ( down to <32 items. Always show last item so #images is indicated.   )     
-if desmtx & ~isempty(fnames)
+if desmtx && ~isempty(fnames)
     axes('Position',[.68 .4 .3 .4],'Visible','off',...
         'DefaultTextFontSize',FS(8),...
         'YLim',[0,nScan]+0.5,'YDir','Reverse')
@@ -785,7 +793,7 @@
 %-----------------------------------------------------------------------
 if iX,  set(hDesMtxIm,'UserData',...
     struct('X',varargin{2}.xKXs.X,'Xnames',{Xnames},'fnames',{fnames}))
-else,   set(hDesMtxIm,'UserData',...
+else   set(hDesMtxIm,'UserData',...
     struct('X',varargin{2}.X,     'Xnames',{Xnames},'fnames',{fnames}))
 end
 set(hDesMtxIm,'ButtonDownFcn','spm_DesRep(''SurfDesMtx_CB'')')
@@ -797,7 +805,7 @@
     hPEstAx   = axes('Position',[.07 .315 .6 .025],...
             'DefaultTextInterpreter','TeX');
     if iX,  est = spm_SpUtil('IsCon',varargin{2}.xKXs);
-    else,   est = spm_SpUtil('IsCon',varargin{2}.X); end
+    else    est = spm_SpUtil('IsCon',varargin{2}.X); end
     hParEstIm = image((est+1)*32);
     set(hPEstAx,...
         'XLim',[0,nPar]+.5,'XTick',[1:nPar-1]+.5,'XTickLabel','',...
@@ -879,9 +887,9 @@
         text(0.3,y,[strrep(sf{1},'_',' '),' :'],...
             'HorizontalAlignment','Right','FontWeight','Bold',...
             'FontSize',FS(9))
-        s = getfield(xs,sf{1});
+        s = xs.(sf{1});
         if ~iscellstr(s), s={s}; end
-        for i=1:prod(size(s))
+        for i=1:numel(s)
             text(0.31,y,s{i},'FontSize',FS(9))
             y=y-dy;
         end
@@ -901,7 +909,7 @@
 xC     = varargin{3};   %-Struct array of covariate information
 if ~isstruct(varargin{2}), error('design matrix structure required'), end
 
-if ~length(xC), spm('alert!','No covariates!',mfilename), return, end
+if isempty(xC), spm('alert!','No covariates!',mfilename), return, end
 
 %-Get graphics window & window scaling
 Fgraph = spm_figure('GetWin','Graphics');
@@ -929,14 +937,14 @@
 %-Design matrix (as underlay for plots) and parameter names
 %-----------------------------------------------------------------------
 [nScan,nPar]   = size(varargin{2}.X);
-if isfield(varargin{2},'name') & ~isempty(varargin{2}.name)
-    Xnames = varargin{2}.name; else, Xnames = {}; end
+if isfield(varargin{2},'name') && ~isempty(varargin{2}.name)
+    Xnames = varargin{2}.name; else Xnames = {}; end
 
 %-Design matrix
 hDesMtx = axes('Position',[.1 .5 .7 .3]);
-if isfield(varargin{2},'nKX') & ~isempty(varargin{2}.nKX)
+if isfield(varargin{2},'nKX') && ~isempty(varargin{2}.nKX)
     image(varargin{2}.nKX'*32+32)
-elseif isfield(varargin{2},'xKXs') & ~isempty(varargin{2}.xKXs)
+elseif isfield(varargin{2},'xKXs') && ~isempty(varargin{2}.xKXs)
     image(spm_DesMtx('sca',varargin{2}.xKXs.X,Xnames)*32+32)
 else
     image(spm_DesMtx('sca',varargin{2}.X,Xnames)*32+32)
@@ -990,7 +998,7 @@
         'FontWeight','Bold','FontSize',FS(9))
     s = xC(i).descrip;
     if ~iscellstr(s), s={s}; end
-    for j=1:prod(size(s))
+    for j=1:numel(s)
         text(0.31,y,s{j},'FontSize',FS(9))
         y=y-dy;
     end
@@ -1067,7 +1075,7 @@
 % spm_DesRep('ScanTick',nScan,lim)
 % ( Show at most 32, showing every 2nd/3rd/4th/... as necessary to pair )
 % ( down to <32 items. Always show last item so #images is indicated.    )     
-if nargin<3, lim=32; else, lim=varargin{3}; end
+if nargin<3, lim=32; else lim=varargin{3}; end
 if nargin<2, error('insufficient arguments'), end
 nScan = varargin{2};
 
@@ -1086,7 +1094,7 @@
 
 h    = get(gca,'Xlabel');
 
-if strcmp(lower(varargin{1}),'surfdesmtxup_cb')
+if strcmpi(varargin{1},'surfdesmtxup_cb')
     UD = get(h,'UserData');
     set(h,'String',UD.String,'Interpreter',UD.Interpreter,...
         'UserData',UD.UserData)
@@ -1095,13 +1103,13 @@
 end
 
 
-if strcmp(lower(varargin{1}),'surfdesmtx_cb')
-    UD = struct(    'String',   get(h,'String'),...
-            'Interpreter',  get(h,'Interpreter'),...
-            'UserData', get(h,'UserData'));
+if strcmpi(varargin{1},'surfdesmtx_cb')
+    UD = struct('String',      get(h,'String'),...
+                'Interpreter', get(h,'Interpreter'),...
+                'UserData',    get(h,'UserData'));
     set(h,'UserData',UD)
     set(gcbf,'WindowButtonMotionFcn','spm_DesRep(''SurfDesMtxMo_CB'')',...
-         'WindowButtonUpFcn',    'spm_DesRep(''SurfDesMtxUp_CB'')')
+         'WindowButtonUpFcn','spm_DesRep(''SurfDesMtxUp_CB'')')
 end
 
 mm  = [get(gca,'YLim')',get(gca,'XLim')']+[.5,.5;-.5,-.5];
@@ -1149,7 +1157,7 @@
 
 h    = get(gca,'Xlabel');
 
-if strcmp(lower(varargin{1}),'surfestimup_cb')
+if strcmpi(varargin{1},'surfestimup_cb')
     UD = get(h,'UserData');
     set(h,'String',UD.String,'Interpreter',UD.Interpreter,...
         'UserData',UD.UserData)
@@ -1157,7 +1165,7 @@
     return
 end
 
-if strcmp(lower(varargin{1}),'surfestim_cb')
+if strcmpi(varargin{1},'surfestim_cb')
     UD = struct(    'String',   get(h,'String'),...
             'Interpreter',  get(h,'Interpreter'),...
             'UserData', get(h,'UserData'));
@@ -1208,7 +1216,7 @@
 
 h    = get(gca,'Xlabel');
 
-if strcmp(lower(varargin{1}),'surfdesoup_cb')
+if strcmpi(varargin{1},'surfdesoup_cb')
     UD = get(h,'UserData');
     set(h,'String',UD.String,'Interpreter',UD.Interpreter,...
         'UserData',UD.UserData)
@@ -1216,7 +1224,7 @@
     return
 end
 
-if strcmp(lower(varargin{1}),'surfdeso_cb')
+if strcmpi(varargin{1},'surfdeso_cb')
     UD = struct(    'String',   get(h,'String'),...
             'Interpreter',  get(h,'Interpreter'),...
             'UserData', get(h,'UserData'));
@@ -1243,11 +1251,12 @@
             str = '{\bf not orthogonal}';
         end
         if ~diff(ij), str=[str,' {\it(same column)}']; end
-        if UD.bC(ij(1),ij(2)), tmp=' ={\it r}'; else, tmp=''; end
-        str = { sprintf('{\\bf %s} (col %d) & {\\bf %s} (col %d)',...
-                UD.name{ij(1)},ij(1),...
-                UD.name{ij(2)},ij(2)),...
-            sprintf('cos(\\theta)%s = %1.2f',...
+        if UD.bC(ij(1),ij(2)), tmp=' ={\it r}'; else tmp=''; end
+        str = { sprintf('{\\bf %s} (col %d)',...
+                UD.Xnames{ij(1)},ij(1)),...
+                sprintf('& {\\bf %s} (col %d)',...
+                UD.Xnames{ij(2)},ij(2)),...
+                sprintf('cos(\\theta)%s = %1.2f',...
                 tmp,UD.O(ij(1),ij(2))),...
             ['\rightarrow ',str]};
         istr = 'tex';
@@ -1274,12 +1283,12 @@
 % spm_DesRep('SurfConOUp_CB')
 
 cUD = get(gco,'UserData');
-if ~isstruct(cUD) | ~isfield(cUD,'h')
+if ~isstruct(cUD) || ~isfield(cUD,'h')
     warning('contrast GUI objects setup incorrectly'), return
 end
 h    = cUD.h;
 
-if strcmp(lower(varargin{1}),'surfconup_cb')
+if strcmpi(varargin{1},'surfconup_cb')
     UD = get(h,'UserData');
     set(h,'String',UD.String,'Interpreter',UD.Interpreter,...
         'UserData',UD.UserData)
@@ -1287,7 +1296,7 @@
     return
 end
 
-if strcmp(lower(varargin{1}),'surfcon_cb')
+if strcmpi(varargin{1},'surfcon_cb')
     UD = struct(    'String',   get(h,'String'),...
             'Interpreter',  get(h,'Interpreter'),...
             'UserData', get(h,'UserData'));
@@ -1304,7 +1313,7 @@
 switch get(gcbf,'SelectionType')
 case 'normal'
     try
-        if cUD.i>0, str = sprintf('%d',cUD.i); else, str = ''; end
+        if cUD.i>0, str = sprintf('%d',cUD.i); else str = ''; end
         switch get(gco,'Type')
         case 'image'
             str = sprintf('%s\\{F\\}: {\\bf%s} (%d,%d) = %.2f',...
@@ -1339,7 +1348,7 @@
 
 h    = get(gca,'Xlabel');
 
-if strcmp(lower(varargin{1}),'surfxviup_cb')
+if strcmpi(varargin{1},'surfxviup_cb')
       UD = get(h,'UserData');
       set(h,'String',UD.String,'Interpreter',UD.Interpreter,...
               'UserData',UD.UserData)
@@ -1348,7 +1357,7 @@
 end
 
 
-if strcmp(lower(varargin{1}),'surfxvi_cb')
+if strcmpi(varargin{1},'surfxvi_cb')
       UD = struct(    'String',       get(h,'String'),...
                       'Interpreter',  get(h,'Interpreter'),...
                       'UserData',     get(h,'UserData'));
@@ -1366,8 +1375,8 @@
 case 'normal'
     try
         str = sprintf('V(%d,%d) = %g',ij(1),ij(2),...
-              subsref(get(gco,'UserData'),...
-              struct('type',{'.','()'},'subs',{'V',{ij(1),ij(2)}})));
+              full(subsref(get(gco,'UserData'),...
+              struct('type',{'.','()'},'subs',{'V',{ij(1),ij(2)}}))));
     catch
         str = '(no cached covariance matrix to surf)';
     end
@@ -1375,7 +1384,7 @@
     try
         ind = 1:length(subsref(get(gco,'Userdata'),...
             struct('type','.','subs','h')));
-        isel = logical(zeros(size(ind)));
+        isel = false(size(ind));
         for k = 1:length(ind)
             isel(k) = subsref(get(gco,'UserData'),...
                 struct('type',{'.','{}','()'},'subs',{'Vi',{k},{ij(1),ij(2)}})) ~= 0;
@@ -1412,7 +1421,7 @@
 
 h    = get(gca,'Xlabel');
 
-if strcmp(lower(varargin{1}),'surfhpestimup_cb')
+if strcmpi(varargin{1},'surfhpestimup_cb')
       UD = get(h,'UserData');
       set(h,'String',UD.String,'Interpreter',UD.Interpreter,...
               'UserData',UD.UserData)
@@ -1420,7 +1429,7 @@
       return
 end
 
-if strcmp(lower(varargin{1}),'surfhpestim_cb')
+if strcmpi(varargin{1},'surfhpestim_cb')
       UD = struct(    'String',       get(h,'String'),...
                       'Interpreter',  get(h,'Interpreter'),...
                       'UserData',     get(h,'UserData'));
diff --git a/spm_Dpdf.m b/spm_Dpdf.m
index 2f455c0..91c2661 100644
--- a/spm_Dpdf.m
+++ b/spm_Dpdf.m
@@ -16,13 +16,13 @@
 %
 % Algorithm:
 %--------------------------------------------------------------------------
-% Direct computation using logs and Matlab's implementation of the log of 
+% Direct computation using logs and MATLAB's implementation of the log of 
 % the gamma function (gammaln).
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 2008-2011 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_Dpdf.m 2636 2009-01-21 09:57:20Z guillaume $
+% $Id: spm_Dpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 %-Check enough arguments
 %--------------------------------------------------------------------------
diff --git a/spm_FcUtil.m b/spm_FcUtil.m
index 7d665de..d49c36d 100644
--- a/spm_FcUtil.m
+++ b/spm_FcUtil.m
@@ -127,13 +127,13 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jean-Baptiste Poline
-% $Id: spm_FcUtil.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_FcUtil.m 4137 2010-12-15 17:18:32Z guillaume $
 
 
 %-Format arguments
 %-----------------------------------------------------------------------
 if nargin==0, error('do what? no arguments given...')
-else, action = varargin{1}; end
+else action = varargin{1}; end
 
 
 switch lower(action),
@@ -142,10 +142,10 @@
 %=======================================================================
 % Fc = spm_FcUtil('FconFields')
 
-if nargout > 1, error('Too many output arguments: FconFields'), end;
-if nargin  > 1, error('Too many input arguments:  FconFields'), end;
+if nargout > 1, error('Too many output arguments: FconFields'), end
+if nargin  > 1, error('Too many input arguments:  FconFields'), end
 
-varargout = {sf_FconFields;}
+varargout = {sf_FconFields};
 
 case {'set','v1set'}                %- Create an F contrast
 %=======================================================================
@@ -167,20 +167,20 @@
 
 %-check # arguments...
 %--------------------------------------------------------------------------
-if nargin<6, error('insufficient arguments'), end;
-if nargout > 1, error('Too many output arguments Set'), end;
+if nargin<6, error('insufficient arguments'), end
+if nargout > 1, error('Too many output arguments Set'), end
 
 %-check arguments...
 %--------------------------------------------------------------------------
-if ~ischar(varargin{2}), error('~ischar(name)'), end;
-if ~(varargin{3}=='F'|varargin{3}=='T'|varargin{3}=='P'), 
-    error('~(STAT==F|STAT==T|STAT==P)'), end;
+if ~ischar(varargin{2}), error('~ischar(name)'), end
+if ~(varargin{3}=='F'||varargin{3}=='T'||varargin{3}=='P'), 
+    error('~(STAT==F|STAT==T|STAT==P)'), end
 if ~ischar(varargin{4}), error('~ischar(varargin{4})'); 
-else set_action = varargin{4}; end;
+else set_action = varargin{4}; end
 
 sX = varargin{6};
-if ~spm_sp('isspc',sX), sX = spm_sp('set',sX);  end;
-if isempty(sX.X), error('Empty space X in Set'); end;
+if ~spm_sp('isspc',sX), sX = spm_sp('set',sX);  end
+if isempty(sX.X), error('Empty space X in Set'); end
 
 Fc = sf_FconFields;
 %- use the name as a flag to insure that F-contrast has been 
@@ -188,7 +188,7 @@
 Fc.name = varargin{2};
 
 Fc.STAT = varargin{3};
-if Fc.STAT=='T' &  ~(any(strcmp(set_action,{'c+','c'}))) 
+if Fc.STAT=='T' &&  ~(any(strcmp(set_action,{'c+','c'}))) 
    warning('enter T stat with contrast - here no check rank == 1');
 end
 
@@ -198,80 +198,80 @@
 %- NO. v1 = strcmp(action,'v1set');
 
 switch set_action,
-        case {'c','c+'}
-       Fc.iX0  = set_action;
-           c = spm_sp(':', sX, varargin{5});
-       if isempty(c)
-              [Fc.X1o.ukX1o Fc.X0.ukX0] = spm_SpUtil('+c->Tsp',sX,[]);
-          %- v1 [Fc.X1o Fc.X0] = spm_SpUtil('c->Tsp',sX,[]);
-              Fc.c    = c;
-       elseif size(c,1) ~= sL, 
-          error(['not contrast dim. in ' mfilename ' ' set_action]); 
-       else 
-              if strcmp(set_action,'c+')  
-                 if ~spm_sp('isinspp',sX,c), c = spm_sp('oPp:',sX,c); end;
-          end;
-          if Fc.STAT=='T' &  ~sf_is_T(sX,c)
-                 %- Could be make more self-correcting by giving back an F
-         error('trying to define a t that looks like an F'); 
-          end
-          Fc.c   = c;
-              [Fc.X1o.ukX1o Fc.X0.ukX0] = spm_SpUtil('+c->Tsp',sX,c);
-          %- v1 [Fc.X1o Fc.X0] = spm_SpUtil('c->Tsp',sX,c);
-       end;
+    case {'c','c+'}
+        Fc.iX0  = set_action;
+        c = spm_sp(':', sX, varargin{5});
+        if isempty(c)
+            [Fc.X1o.ukX1o Fc.X0.ukX0] = spm_SpUtil('+c->Tsp',sX,[]);
+            %- v1 [Fc.X1o Fc.X0] = spm_SpUtil('c->Tsp',sX,[]);
+            Fc.c    = c;
+        elseif size(c,1) ~= sL,
+            error(['not contrast dim. in ' mfilename ' ' set_action]);
+        else
+            if strcmp(set_action,'c+')
+                if ~spm_sp('isinspp',sX,c), c = spm_sp('oPp:',sX,c); end
+            end;
+            if Fc.STAT=='T' &&  ~sf_is_T(sX,c)
+                %- Could be make more self-correcting by giving back an F
+                error('trying to define a t that looks like an F');
+            end
+            Fc.c   = c;
+            [Fc.X1o.ukX1o Fc.X0.ukX0] = spm_SpUtil('+c->Tsp',sX,c);
+            %- v1 [Fc.X1o Fc.X0] = spm_SpUtil('c->Tsp',sX,c);
+        end
 
 %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 % 'option given for completeness - not for SPM use'
 
     case {'X0'}
-           warning(['option given for completeness - not for SPM use']);
-       Fc.iX0  = set_action;
-       X0 = spm_sp(':', sX, varargin{5});
-       if isempty(X0),
-              Fc.c         = spm_sp('xpx',sX); 
-              Fc.X1o.ukX1o = spm_sp('cukx',sX);
-              Fc.X0.ukX0   = [];    
-       elseif size(X0,1) ~= sC, 
-              error('dimension of X0 wrong in Set');
-       else 
-              Fc.c         = spm_SpUtil('X0->c',sX,X0);
-          Fc.X0.ukX0   = spm_sp('ox',sX)'*X0;   
-              Fc.X1o.ukX1o = spm_SpUtil('+c->Tsp',sX,Fc.c);
-       end
-
-        case 'ukX0' 
-           warning(['option given for completeness - not for SPM use']);
-       Fc.iX0  = set_action;
-       if isempty(ukX0), 
-              Fc.c         = spm_sp('xpx',sX); 
-              Fc.X1o.ukX1o = spm_sp('cukx',sX);
-              Fc.X0.ukX0   = [];    
-       elseif size(ukX0,1) ~= spm_sp('rk',sX), 
-              error('dimension of cukX0 wrong in Set');
-       else 
-              Fc.c         = spm_SpUtil('+X0->c',sX,ukX0);
-          Fc.X0.ukX0   = ukX0;  
-              Fc.X1o.ukX1o = spm_SpUtil('+c->Tsp',sX,Fc.c);
-       end 
+        warning('option given for completeness - not for SPM use');
+        Fc.iX0  = set_action;
+        X0 = spm_sp(':', sX, varargin{5});
+        if isempty(X0),
+            Fc.c         = spm_sp('xpx',sX);
+            Fc.X1o.ukX1o = spm_sp('cukx',sX);
+            Fc.X0.ukX0   = [];
+        elseif size(X0,1) ~= sC,
+            error('dimension of X0 wrong in Set');
+        else
+            Fc.c         = spm_SpUtil('X0->c',sX,X0);
+            Fc.X0.ukX0   = spm_sp('ox',sX)'*X0;
+            Fc.X1o.ukX1o = spm_SpUtil('+c->Tsp',sX,Fc.c);
+        end
+        
+    case 'ukX0'
+        warning('option given for completeness - not for SPM use');
+        Fc.iX0  = set_action;
+        if isempty(ukX0),
+            Fc.c         = spm_sp('xpx',sX);
+            Fc.X1o.ukX1o = spm_sp('cukx',sX);
+            Fc.X0.ukX0   = [];
+        elseif size(ukX0,1) ~= spm_sp('rk',sX),
+            error('dimension of cukX0 wrong in Set');
+        else
+            Fc.c         = spm_SpUtil('+X0->c',sX,ukX0);
+            Fc.X0.ukX0   = ukX0;
+            Fc.X1o.ukX1o = spm_SpUtil('+c->Tsp',sX,Fc.c);
+        end
 %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
     case 'iX0'
-       iX0        = varargin{5};
-       iX0        = spm_SpUtil('iX0check',iX0,sL);
-       Fc.iX0     = iX0;
-           Fc.X0.ukX0 = spm_sp('ox',sX)' * spm_sp('Xi',sX,iX0);
-           if isempty(iX0), 
-          Fc.c          = spm_sp('xpx',sX); 
-          Fc.X1o.ukX1o  = spm_sp('cukx',sX);  
-       else             
-          Fc.c          = spm_SpUtil('i0->c',sX,iX0);
-          Fc.X1o.ukX1o  = spm_SpUtil('+c->Tsp',sX,Fc.c);
-       end;
-
-    otherwise 
-       error('wrong action in Set '); 
-
-end;
+        iX0        = varargin{5};
+        iX0        = spm_SpUtil('iX0check',iX0,sL);
+        Fc.iX0     = iX0;
+        Fc.X0.ukX0 = spm_sp('ox',sX)' * spm_sp('Xi',sX,iX0);
+        if isempty(iX0),
+            Fc.c          = spm_sp('xpx',sX);
+            Fc.X1o.ukX1o  = spm_sp('cukx',sX);
+        else
+            Fc.c          = spm_SpUtil('i0->c',sX,iX0);
+            Fc.X1o.ukX1o  = spm_SpUtil('+c->Tsp',sX,Fc.c);
+        end
+        
+    otherwise
+        error('wrong action in Set ');
+        
+end
 varargout = {Fc};
 
 case 'x0' % spm_FcUtil('X0',Fc,sX)
@@ -283,7 +283,7 @@
 end
 if nargout ~= 1, error('too few/many output arguments - need 1'), end
 if ~sf_IsFcon(Fc), error('argument is not a contrast struct'), end
-if ~spm_sp('isspc',sX), sX = spm_sp('set',sX); end;
+if ~spm_sp('isspc',sX), sX = spm_sp('set',sX); end
 
 varargout = {sf_X0(Fc,sX)};
 
@@ -296,7 +296,7 @@
 end
 if nargout ~= 1, error('too few/many output arguments - need 1'), end
 if ~sf_IsFcon(Fc), error('argument is not a contrast struct'), end
-if ~spm_sp('isspc',sX), sX = spm_sp('set',sX); end;
+if ~spm_sp('isspc',sX), sX = spm_sp('set',sX); end
 
 varargout = {sf_X1o(Fc,sX)};
 
@@ -307,7 +307,7 @@
 
 if nargin~=2, error('too few/many input arguments - need 2'), end
 if ~isstruct(varargin{2}), varargout={0}; 
-else, varargout = {sf_IsFcon(varargin{2})};
+else varargout = {sf_IsFcon(varargin{2})};
 end
 
 
@@ -319,11 +319,11 @@
 if nargout >= 3, error('Too many output argument.'), end
 Fc = varargin{2};
 sX = varargin{3};
-if nargin == 4, V = varargin{4}; else V = []; end;
+if nargin == 4, V = varargin{4}; else V = []; end
 
 if ~sf_IsFcon(Fc), error('Fc must be Fcon'), end
 if ~spm_sp('isspc',sX)
-    sX = spm_sp('set',sX);  end;
+    sX = spm_sp('set',sX);  end
 
 if ~sf_isempty_X1o(Fc)
    [trMV, trMVMV] = spm_SpUtil('trMV',sf_X1o(Fc,sX),V);
@@ -332,18 +332,18 @@
    trMVMV = 0;
 end 
 if ~trMVMV, edf_tsp = 0; warning('edf_tsp = 0'), 
-else,  edf_tsp = trMV^2/trMVMV; end;    
+else  edf_tsp = trMV^2/trMVMV; end;    
 
 if nargout == 2
 
    [trRV, trRVRV] = spm_SpUtil('trRV',sX,V);
-   if ~trRVRV, edf_Xsp = 0, warning('edf_Xsp = 0'),
-   else,  edf_Xsp = trRV^2/trRVRV; end;
+   if ~trRVRV, edf_Xsp = 0; warning('edf_Xsp = 0'),
+   else  edf_Xsp = trRV^2/trRVRV; end;
 
    varargout = {edf_tsp, edf_Xsp};
 else    
    varargout = {edf_tsp};
-end;
+end
 
 
 
@@ -511,11 +511,11 @@
 str  = Fc2(1).name; for i=2:L2 str = [str ' ' Fc2(i).name]; end;
 Fc2  = spm_FcUtil('Set',str,'F','c+',cat(2,Fc2(:).c),sX);
 
-if sf_isempty_X1o(Fc2) | sf_isnull(Fc2,sX)
+if sf_isempty_X1o(Fc2) || sf_isnull(Fc2,sX)
     varargout = {Fc1};
 else
     for i=1:L1
-        if sf_isempty_X1o(Fc1(i)) | sf_isnull(Fc1(i),sX)
+        if sf_isempty_X1o(Fc1(i)) || sf_isnull(Fc1(i),sX)
             %- Fc1(i) is an [] or 0 contrast : ortho to anything; 
             out(i) = Fc1(i);
         else
@@ -531,7 +531,7 @@
 
 if nargin < 3, error('Insufficient arguments'), end
 Fc1 = varargin{2}; sX = varargin{3};
-if nargin > 3, Fc2 = varargin{4}; else, Fc2 = []; end;
+if nargin > 3, Fc2 = varargin{4}; else Fc2 = []; end;
 if isempty(Fc1), error('give at least one non empty contrast'), end;
 
 if ~spm_sp('isspc',sX), sX = spm_sp('set',sX);  end;
@@ -563,7 +563,7 @@
 L1  = length(Fc1);
 if  ~L1, warning('no contrast given to in'); 
      if nargout == 2, varargout = {[] []}; 
-     else, varargout = {[]}; end;
+     else varargout = {[]}; end;
      return; 
 end
 for i=1:L1
@@ -634,7 +634,7 @@
 
 idx = [];
 for i=1:L1
-    if sf_isempty_X1o(Fc(i)) | sf_isnull(Fc(i),sX), idx = [idx i]; end
+    if sf_isempty_X1o(Fc(i)) || sf_isnull(Fc(i),sX), idx = [idx i]; end
 end
 if isempty(idx) 
     varargout = {0};
@@ -666,12 +666,12 @@
     'name',     '',...
     'STAT',     '',...
     'c',        [],...
-    'X0',       struct('ukX0',[]),... %!15/10
+    'X0',       struct('ukX0',[]),...
     'iX0',      [],...
-    'X1o',      struct('ukX1o',[]),...  %!15/10
+    'X1o',      struct('ukX1o',[]),...
     'eidf',     [],...
     'Vcon',     [],...
-    'Vspm',     []  );
+    'Vspm',     []);
  
 %=======================================================================
 % used internally. Minimum contrast structure
@@ -682,9 +682,8 @@
     'name',     '',...
     'STAT',     '',...
     'c',        [],...
-    'X0',           [],...
-    'X1o',          []...
-    );
+    'X0',       [],...
+    'X1o',      []);
 
 %=======================================================================
 % yes_no = spm_FcUtil('IsFcon',Fc)
@@ -727,7 +726,7 @@
    if b ~= isempty(Fc.c), 
       Fc.c, Fc.X1o.ukX1o, error('Contrast internally not consistent');
    end
-else, 
+else
    b = isempty(Fc.X1o);
    %- consistency check
    if b ~= isempty(Fc.c), 
@@ -741,7 +740,7 @@
 
 if sf_ver(Fc) > 1, 
    b = spm_sp('ox',sX)*Fc.X1o.ukX1o; 
-else, 
+else
    b = Fc.X1o;
 end
 
@@ -752,7 +751,7 @@
 
 if sf_ver(Fc) > 1, 
    b = spm_sp('ox',sX)*Fc.X0.ukX0; 
-else, 
+else
    b = Fc.X0;
 end
 
@@ -763,7 +762,7 @@
 
 if sf_ver(Fc) > 1, 
    b = isempty(Fc.X0.ukX0); 
-else, 
+else
    b = isempty(Fc.X0);
 end
 
@@ -789,7 +788,7 @@
 %-**** fprintf('v%d\n',sf_ver(Fc));
 if sf_ver(Fc) > 1,    
    hsqr = spm_sp('ox',spm_sp('set',Fc.X1o.ukX1o))' * spm_sp('cukx',sX);
-else, 
+else
    hsqr = spm_sp('ox',spm_sp('set',Fc.X1o))'*spm_sp('x',sX);
 end
 %=======================================================================
@@ -845,7 +844,7 @@
 %-                          Fc1.STAT, 'c',c,sX);
 
 %=======================================================================
-function b = sf_Rortho(Fc1,sX,Fc2);
+function b = sf_Rortho(Fc1,sX,Fc2)
 
 if isempty(Fc2)
    if length(Fc1) <= 1, b = 0; 
@@ -881,13 +880,10 @@
 
 boul = 1;
 if ~spm_sp('isinspp',sX,c), c = spm_sp('oPp:',sX,c); end;
-if rank(c) > 1 | any(any(c'*c < 0)), boul = 0; end;
-
-
-
+if rank(c) > 1 || any(any(c'*c < 0)), boul = 0; end;
 
 %=======================================================================
-function [idxFc2, idxFc1] =  sf_in(Fc1, sX, Fc2);
+function [idxFc2, idxFc1] =  sf_in(Fc1, sX, Fc2)
 
 L2 = length(Fc2);
 L1 = length(Fc1);
@@ -897,14 +893,14 @@
    %- project Fc1(j).c if not estimable
    if ~spm_sp('isinspp',sX,Fc1(j).c), %- warning ? 
       c1  = spm_sp('oPp:',sX,Fc1(j).c); 
-   else, 
+   else
       c1 = Fc1(j).c; 
    end
    sc1 = spm_sp('Set',c1); 
    S   = Fc1(j).STAT;
 
-   boul = 0; i = 1;
-   for i =1:L2,
+   boul = 0;
+   for i =1:L2
       if Fc2(i).STAT == S
          %- the same statistics. else just go on to the next contrast
          boul = spm_sp('==',sc1,spm_sp('oPp',sX,Fc2(i).c));
@@ -912,14 +908,14 @@
          %- if they are the same space and T stat (same direction),
          %- then check wether they are in the same ORIENTATION
          %- works because size(X1o,2) == 1, else .* with (Fc1(j).c'*Fc2(i).c)
-         if boul & S == 'T'
+         if boul && S == 'T'
             atmp = sf_X1o(Fc1(j),sX);
             btmp = sf_X1o(Fc2(i),sX);
             boul = ~any(any( (atmp' * btmp) < 0 ));
          end
          %- note the indices
-         if boul, idxFc1 = [idxFc1 j]; idxFc2 = [idxFc2 i]; end;
-      end;
+         if boul, idxFc1 = [idxFc1 j]; idxFc2 = [idxFc2 i]; end
+      end
    end
 end %- for j=1:L1
 
diff --git a/spm_Fcdf.m b/spm_Fcdf.m
index e34b4bb..7029bfb 100644
--- a/spm_Fcdf.m
+++ b/spm_Fcdf.m
@@ -9,12 +9,12 @@
 % v  - Shape parameter 1 /   numerator degrees of freedom (v>0)
 % w  - Shape parameter 2 / denominator degrees of freedom (w>0)
 % F  - CDF of F-distribution with [v,w] degrees of freedom at points x
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Fcdf implements the Cumulative Distribution Function of the F-distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF F(x) of the F distribution with degrees of freedom v & w,
 % defined for positive integer degrees of freedom v & w, is the
 % probability that a realisation of an F random variable X has value
@@ -22,7 +22,7 @@
 % for v>0 & w>0, and for x in [0,Inf) (See Evans et al., Ch16).
 %
 % Variate relationships: (Evans et al., Ch16 & 37)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The square of a Student's t variate with w degrees of freedom is
 % distributed as an F-distribution with [1,w] degrees of freedom.
 %
@@ -31,7 +31,7 @@
 % w/2 & v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Using the relationship with the Beta distribution: The CDF of the
 % F-distribution with v,w degrees of freedom is related to the
 % incomplete beta function by:
@@ -41,11 +41,11 @@
 % easily verified by substituting for w/(w+v*x^2) in the integral of the
 % incomplete beta function.
 %
-% MatLab's implementation of the incomplete beta function is used.
+% MATLAB's implementation of the incomplete beta function is used.
 %
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -59,14 +59,14 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1992-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Fcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Fcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 %-Unpack degrees of freedom v & w from single df parameter (v)
@@ -88,23 +88,25 @@
 %-Check argument sizes
 ad = [ndims(x);ndims(v);ndims(w)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];...
-    [size(w),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];...
+      [size(w),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
     error('non-scalar args must match in size'), end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 F = zeros(rs);
 
 %-Only defined for strictly positive v & w. Return NaN if undefined.
 md = ( ones(size(x))  &  v>0  &  w>0 );
-if any(~md(:)), F(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    F(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero where defined and x>0
 Q  = find( md  &  x>0 );
diff --git a/spm_Fpdf.m b/spm_Fpdf.m
index 64b4a90..fea2578 100644
--- a/spm_Fpdf.m
+++ b/spm_Fpdf.m
@@ -9,12 +9,12 @@
 % v  - Shape parameter 1 /   numerator degrees of freedom (v>0)
 % w  - Shape parameter 2 / denominator degrees of freedom (w>0)
 % f  - PDF of F-distribution with [v,w] degrees of freedom at points x
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Fpdf implements the Probability Density Function of the F-distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The PDF of the F-distribution with degrees of freedom v & w, defined
 % for positive integer degrees of freedom v>0 & w>0, and for x in
 % [0,Inf) by: (See Evans et al., Ch16)
@@ -24,7 +24,7 @@
 %           gamma(v/2)*gamma(w/2) * (1+(v/w)x)^((v+w)/2)
 %
 % Variate relationships: (Evans et al., Ch16 & 37)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The square of a Student's t variate with w degrees of freedom is
 % distributed as an F-distribution with [1,w] degrees of freedom.
 %
@@ -33,12 +33,12 @@
 % w/2 & v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Direct computation using the beta function for
 %       gamma(v/2)*gamma(w/2) / gamma((v+w)/2)  =  beta(v/2,w/2)
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -51,15 +51,15 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1994-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Fpdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Fpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 %-Unpack degrees of freedom v & w from single df parameter (v)
@@ -81,23 +81,26 @@
 %-Check argument sizes
 ad = [ndims(x);ndims(v);ndims(w)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];...
-    [size(w),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];...
+      [size(w),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for strictly positive v & w. Return NaN if undefined.
 md = ( ones(size(x))  &  v>0  &  w>0 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero where defined and x>0
 Q  = find( md  &  x>0 );
diff --git a/spm_Gcdf.m b/spm_Gcdf.m
index c5afbf7..c7b37da 100644
--- a/spm_Gcdf.m
+++ b/spm_Gcdf.m
@@ -11,7 +11,7 @@
 % spm_Gcdf implements the Cumulative Distribution of the Gamma-distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF F(x) of the Gamma distribution with shape parameter h and
 % scale l is the probability that a realisation of a Gamma random
 % variable X has value less than x F(x)=Pr{X<x} for X~G(h,l). The Gamma
@@ -20,7 +20,7 @@
 % parameterisation of the Gamma with scale parameter c=1/l)
 %
 % Variate relationships: (Evans et al., Ch18 & Ch8)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % For natural (strictly +ve integer) shape h this is an Erlang distribution.
 %
 % The Standard Gamma distribution has a single parameter, the shape h.
@@ -30,7 +30,7 @@
 % to the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF of the Gamma distribution with scale parameter l and shape h
 % is related to the incomplete Gamma function by
 %
@@ -40,10 +40,10 @@
 % of the incomplete Gamma function. The relationship is easily verified
 % by substituting for t/c in the integral, where c=1/l.
 %
-% MatLab's implementation of the incomplete gamma function is used.
+% MATLAB's implementation of the incomplete gamma function is used.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -56,14 +56,14 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1992-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Gcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Gcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(h);ndims(l)];
@@ -73,18 +73,21 @@
     [size(l),ones(1,rd-ad(3))]     ];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 F = zeros(rs);
 
 %-Only defined for strictly positive h & l. Return NaN if undefined.
 md = ( ones(size(x))  &  h>0  &  l>0 );
-if any(~md(:)), F(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    F(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero where defined and x>0
 Q  = find( md  &  x>0 );
diff --git a/spm_Gpdf.m b/spm_Gpdf.m
index 9a7c848..309fd9f 100644
--- a/spm_Gpdf.m
+++ b/spm_Gpdf.m
@@ -1,29 +1,29 @@
 function f = spm_Gpdf(x,h,l)
 % Probability Density Function (PDF) of Gamma distribution
-% FORMAT f = spm_Gpdf(g,h,l)
+% FORMAT f = spm_Gpdf(x,h,l)
 %
 % x - Gamma-variate   (Gamma has range [0,Inf) )
 % h - Shape parameter (h>0)
 % l - Scale parameter (l>0)
 % f - PDF of Gamma-distribution with shape & scale parameters h & l
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Gpdf implements the Probability Density Function of the Gamma
 % distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The PDF of the Gamma distribution with shape parameter h and scale l
 % is defined for h>0 & l>0 and for x in [0,Inf) by: (See Evans et al.,
 % Ch18, but note that this reference uses the alternative
 % parameterisation of the Gamma with scale parameter c=1/l)
 %
 %           l^h * x^(h-1) exp(-lx)
-%    f(x) = ---------------------
-%                gamma(h)
+%    f(x) = ----------------------
+%                   gamma(h)
 %
 % Variate relationships: (Evans et al., Ch18 & Ch8)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % For natural (strictly +ve integer) shape h this is an Erlang distribution.
 %
 % The Standard Gamma distribution has a single parameter, the shape h.
@@ -33,11 +33,11 @@
 % to the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Direct computation using logs to avoid roundoff errors.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -49,36 +49,39 @@
 % Press WH, Teukolsky SA, Vetterling AT, Flannery BP (1992)
 %       "Numerical Recipes in C"
 %        Cambridge
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1993-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Gpdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Gpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(h);ndims(l)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(h),ones(1,rd-ad(2))];...
-    [size(l),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(h),ones(1,rd-ad(2))];...
+      [size(l),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for strictly positive h & l. Return NaN if undefined.
 md = ( ones(size(x))  &  h>0  &  l>0 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Degenerate cases at x==0: h<1 => f=Inf; h==1 => f=l; h>1 => f=0
 ml = ( md  &  x==0  &  h<1 );
diff --git a/spm_Icdf.m b/spm_Icdf.m
index 7ae699e..f335286 100644
--- a/spm_Icdf.m
+++ b/spm_Icdf.m
@@ -6,13 +6,13 @@
 % n - Binomial n
 % p - Binomial p [Defaults to 0.5]
 % F - CDF
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Icdf returns the Cumulative Distribution Function for the
 % Binomial family of distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Bin(n,p) distribution is the distribution of the number of
 % successes from n identical independent Bernoulli trials each with
 % success probability p. If random variable X is the number of
@@ -31,7 +31,7 @@
 % where nCx is the Binomial coefficient "n-choose-x", given by n!/(x!(n-x)!)
 %
 % Normal approximation:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % For (npq>5 & 0.1<=p<=0.9) | min(np,nq)>10 | npq>25 the Normal
 % approximation to the Binomial may be used:
 %       X~Bin(n,p),  X~:~N(np,npq)              ( ~:~ -> approx. distributed as)
@@ -41,7 +41,7 @@
 %       \Phi(x) = 0.5+0.5*erf(x/sqrt(2))
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % F(x), the CDF of the Binomial distribution, for X~Bin(n,p), is related
 % to the incomplete beta function, by:
 %
@@ -51,7 +51,7 @@
 % further details.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -64,16 +64,16 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Icdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Icdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, p=0.5; end
 if nargin<2, error('Insufficient arguments'), end
 ad = [ndims(x);ndims(n);ndims(p)];
@@ -83,18 +83,21 @@
     [size(p),ones(1,rd-ad(3))]     ];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 F = zeros(rs);
 
 %-Only defined for whole n, and for p in [0,1]. Return NaN if undefined.
 md = ( ones(size(x))  &  n==floor(n)  &  n>=0  &  p>=0  &  p<=1 );
-if any(~md(:)), F(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    F(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-F is 1 where x>=n, or (p=0 & x>=0) (where betainc involves log of zero)
 m1 = ( x>=n  |  (p==0 & x>=0) );
diff --git a/spm_Ipdf.m b/spm_Ipdf.m
index efbfaf4..5457a31 100644
--- a/spm_Ipdf.m
+++ b/spm_Ipdf.m
@@ -6,13 +6,13 @@
 % n - Binomial n
 % p - Binomial p [Defaults to 0.5]
 % f - PDF
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Ipdf returns the Probability (Distribution) Function (PDF) for
 % the Binomial family of distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Bin(n,p) distribution is the distribution of the number of
 % successes from n identical independent Bernoulli trials each with
 % success probability p. If random variable X is the number of
@@ -27,7 +27,7 @@
 % where nCx is the Binomial coefficient "n-choose-x", given by n!/(x!(n-x)!).
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % For vary small n, nCx can be computed naively as the ratio of
 % factorials, using gamma(n+1) to return n!. For moderately sized n, n!
 % (& x! &/or (n-x)!) become very large, and naive computation isn't
@@ -49,7 +49,7 @@
 % of n & x. See Press et al., Sec6.1 for further details.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -62,37 +62,39 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Ipdf.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_Ipdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, p=0.5; end
 if nargin<2, error('Insufficient arguments'), end
 ad = [ndims(x);ndims(n);ndims(p)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(n),ones(1,rd-ad(2))];...
-    [size(p),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(n),ones(1,rd-ad(2))];...
+      [size(p),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for whole n, and for p in [0,1]. Return NaN if undefined.
 md = ( ones(size(x))  &  n==floor(n)  &  n>=0  &  p>=0  &  p<=1 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero only where defined and x is whole with 0<=x<=n
 Q  = find( md  &  x==floor(x)  &  n>=x  &  x>=0 );
@@ -106,23 +108,23 @@
 f(Q) = f(Q).* p(Qp).^x(Qx) .* (1-p(Qp)).^(n(Qn)-x(Qx));
 
 %-Return
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 return
 
 
 
-%=======================================================================
+%==========================================================================
 %-Direct computation method: (For interest)
-%=======================================================================
+%==========================================================================
 % The following cunning direct computation is faster than using log
 % gammas, but is rather difficult to vectorise.
 %q=1-p;
 %if r<n/2
 %   %-better for small r (less terms) / small p (smaller numbers)
-%   %---------------------------------------------------------------
+%   %----------------------------------------------------------------------
 %   f=prod([[n:-1:n-r+1]*p,1]./[r:-1:1,1])*q^(n-r);
 %else
 %   %-better for large r (less terms) / small q (smaller numbers)
-%   %---------------------------------------------------------------
+%   %----------------------------------------------------------------------
 %   f=prod([[n:-1:r+1]*q,1]./[n-r:-1:1,1])*p^r;
 %end
diff --git a/spm_LAP.m b/spm_LAP.m
index b0494a4..15b9732 100644
--- a/spm_LAP.m
+++ b/spm_LAP.m
@@ -67,9 +67,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_LAP.m 3977 2010-07-08 14:14:35Z karl $
-
-
+% $Id: spm_LAP.m 4146 2010-12-23 21:01:39Z karl $
+ 
+ 
 % find or create a DEM figure
 %--------------------------------------------------------------------------
 try
@@ -80,23 +80,33 @@
 if ~DEM.M(1).nograph
     Fdem = spm_figure('GetWin','DEM');
 end
-
-
+ 
+ 
 % check model, data and priors
 %==========================================================================
 [M Y U] = spm_DEM_set(DEM);
 
 
+% set regularisation
+%--------------------------------------------------------------------------
+try
+    dt = DEM.M(1).E.v;
+catch
+    dt = 0;
+    DEM.M(1).E.v = dt;
+end
+
+ 
 % number of iterations
 %--------------------------------------------------------------------------
 try, nD = M(1).E.nD; catch, nD = 1;   end
 try, nN = M(1).E.nN; catch, nN = 16;  end
-
-
+ 
+ 
 % ensure integration scheme evaluates gradients at each time-step
 %--------------------------------------------------------------------------
 M(1).E.linear = 4;
-
+ 
 % assume precisions are a function of, and only of, hyperparameters
 %--------------------------------------------------------------------------
 try
@@ -111,9 +121,9 @@
 try method.g; catch, method.g = 0; end
 try method.x; catch, method.x = 0; end
 try method.v; catch, method.v = 0; end
-
+ 
 M(1).E.method = method;
-
+ 
 % assume precisions are a function of, and only of, hyperparameters
 %--------------------------------------------------------------------------
 try
@@ -121,7 +131,7 @@
 catch
     form = 'Gaussian';
 end
-
+ 
 % checks for Laplace models (precision functions; ph and pg)
 %--------------------------------------------------------------------------
 for i  = 1:length(M)
@@ -136,7 +146,7 @@
         M(i).pg = inline('spm_LAP_pg(x,v,h,M)','x','v','h','M');
     end
 end
-
+ 
  
 % order parameters (d = n = 1 for static models) and checks
 %==========================================================================
@@ -154,17 +164,17 @@
 nu  = nv*d + nx*n;                           % number of generalised states
 ne  = nv*n + nx*n + ny*n;                    % number of generalised errors
  
-
+ 
 % precision (R) of generalised errors and null matrices for concatenation
 %==========================================================================
 s     = M(1).E.s;
 Rh    = spm_DEM_R(n,s,form);
 Rg    = spm_DEM_R(n,s,form);
-
+ 
 W     = sparse(nx*n,nx*n);
 V     = sparse((ny + nv)*n,(ny + nv)*n);
-
-
+ 
+ 
 % fixed priors on states (u)
 %--------------------------------------------------------------------------
 Px    = kron(sparse(1,1,1,n,n),spm_cat(spm_diag({M.xP})));
@@ -175,15 +185,15 @@
 %--------------------------------------------------------------------------
 ph.h  = spm_vec({M.hE M.gE});                % prior expectation of h,g
 ph.c  = spm_cat(spm_diag({M.hC M.gC}));      % prior covariances of h,g
-ph.ic = spm_pinv(ph.c);                      % prior precision of h,g
+ph.ic = spm_inv(ph.c);                       % prior precision of h,g
  
 qh.h  = {M.hE};                              % conditional expectation h
 qh.g  = {M.gE};                              % conditional expectation g
 nh    = length(spm_vec(qh.h));               % number of hyperparameters h
 ng    = length(spm_vec(qh.g));               % number of hyperparameters g
-nb    = nh + ng;                             % number of hyerparameters
-
-
+nb    = nh + ng;                             % number of hyperparameters
+ 
+ 
 % priors on parameters (in reduced parameter space)
 %==========================================================================
 pp.c  = cell(nl,nl);
@@ -206,6 +216,7 @@
 pp.c  = spm_cat(pp.c);
 pp.ic = spm_inv(pp.c);
  
+
 % initialise conditional density q(p)
 %--------------------------------------------------------------------------
 for i = 1:(nl - 1)
@@ -214,8 +225,8 @@
     end
 end
 np    = size(Up,2);
-
-
+ 
+ 
 % initialise cell arrays for D-Step; e{i + 1} = (d/dt)^i[e] = e[i]
 %==========================================================================
 qu.x      = cell(n,1);
@@ -241,12 +252,10 @@
 Dy     = kron(spm_speye(n,n,1),spm_speye(ny,ny));
 Dc     = kron(spm_speye(d,d,1),spm_speye(nc,nc));
 Du     = spm_cat(spm_diag({Dx,Dv}));
-Ip     = spm_speye(np,np);
-Ih     = spm_speye(nb,nb);
-qp.dp  = sparse(np,1);                   % conditional expectation of dp/dt
-qh.dp  = sparse(nb,1);                   % conditional expectation of dh/dt
-
-
+Ib     = spm_speye(np + nb,np + nb);
+dbdt   = sparse(np + nb,1);
+ 
+ 
 % gradients of generalised weighted errors
 %--------------------------------------------------------------------------
 dedh   = sparse(nh,ne);
@@ -258,32 +267,44 @@
             
 % curvatures of Gibb's energy w.r.t. hyperparameters
 %--------------------------------------------------------------------------
-dHdh   = sparse(nh,  1);
-dHdg   = sparse(ng,  1);
-dHdp   = sparse(np,  1);
-dHdx   = sparse(nx*n,1);
-dHdv   = sparse(nv*d,1);
+dHdh   = sparse(nh,1);
+dHdg   = sparse(ng,1);
+dHdp   = sparse(np,1);
+dHdu   = sparse(nu,1);
 
-% preclude unnecessary iterations and set switchs
+ 
+% preclude unnecessary iterations and set switches
 %--------------------------------------------------------------------------
 if ~np && ~nh && ~ng, nN = 1; end
-mnx = nx*~~method.x;
-mnv = nv*~~method.v;
-
-
-% preclude very precise states from entering free-energy
+mnx    = nx*~~method.x;
+mnv    = nv*~~method.v;
+ 
+% preclude very precise states from entering free-energy/action
 %--------------------------------------------------------------------------
-[p dp] = spm_LAP_eval(M,qu,qh);
+p      = spm_LAP_eval(M,qu,qh);
 ih     = p.h < 16;
 ig     = p.g < 16;
-iv     = kron(ones(d,1),ih([1:nv] + ny));
+ie     = kron(ones(n,1),ih);
 ix     = kron(ones(n,1),ig);
-iu     = logical([ix; iv]);
-iup    = logical([iu; ones(np,1)]);
-
-% Iterate Lapalace scheme
+iv     = kron(ones(d,1),ih((1:nv) + ny));
+je     = find([ie; ix]); ix(1:nx) = 1;
+ju     = find([ix; iv]);
+jub    = find([ix; iv; ones(np + nb,1)]);
+ 
+% and other useful indices
+%--------------------------------------------------------------------------
+ix     = (1:nx);
+ih     = (1:nb);
+iv     = (1:nv) + nx*n;
+ip     = (1:np) + nu;
+iup    = (1:(nu + np));
+ib     = (1:(np + nb)) + nu;
+ 
+ 
+ 
+% Iterate Laplace scheme
 %==========================================================================
-Fa     = -Inf;
+F      = -Inf;
 for iN = 1:nN
  
     % get time and clear persistent variables in evaluation routines
@@ -323,38 +344,35 @@
             
             % prediction errors (E) and precision vectors (p)
             %--------------------------------------------------------------
-            [E dE]  = spm_DEM_eval(M,qu,qp);
-            [p dp]  = spm_LAP_eval(M,qu,qh);
+            [E dE] = spm_DEM_eval(M,qu,qp);
+            [p dp] = spm_LAP_eval(M,qu,qh);
             
  
             % gradients of log(det(iS)) dDd...
             %==============================================================
             
-            % get precision matrices (iSS is conditioned)
+            % get precision matrices
             %--------------------------------------------------------------
-            iSh     = diag(exp(p.h));
-            iSg     = diag(exp(p.g));
-            iS      = blkdiag(kron(Rh,iSh),kron(Rg,iSg));
-            iSh     = diag(exp(p.h).*ih);
-            iSg     = diag(exp(p.g).*ig);
-            iSS     = blkdiag(kron(Rh,iSh),kron(Rg,iSg));
-            
+            iSh   = diag(exp(p.h));
+            iSg   = diag(exp(p.g));
+            iS    = blkdiag(kron(Rh,iSh),kron(Rg,iSg));
+ 
             
             % gradients of trace(diag(p)) = sum(p); p = precision vector
             %--------------------------------------------------------------
-            dpdx    = n*sum(spm_cat({dp.h.dx; dp.g.dx}));
-            dpdv    = n*sum(spm_cat({dp.h.dv; dp.g.dv}));
-            dpdh    = n*sum(dp.h.dh);
-            dpdg    = n*sum(dp.g.dg);
-            dpdx    = kron(sparse(1,1,1,1,n),dpdx);
-            dpdv    = kron(sparse(1,1,1,1,d),dpdv);
-            dDdu    = [dpdx dpdv]';
-            dDdh    = [dpdh dpdg]';
+            dpdx  = n*sum(spm_cat({dp.h.dx; dp.g.dx}));
+            dpdv  = n*sum(spm_cat({dp.h.dv; dp.g.dv}));
+            dpdh  = n*sum(dp.h.dh);
+            dpdg  = n*sum(dp.g.dg);
+            dpdx  = kron(sparse(1,1,1,1,n),dpdx);
+            dpdv  = kron(sparse(1,1,1,1,d),dpdv);
+            dDdu  = [dpdx dpdv]';
+            dDdh  = [dpdh dpdg]';
  
             
             % gradients precision-weighted generalised error dSd..
             %==============================================================
-
+ 
             % gradients w.r.t. hyperparameters
             %--------------------------------------------------------------
             for i = 1:nh
@@ -367,7 +385,7 @@
                 diSdg{i}  = blkdiag(V,kron(Rg,diS));
                 dedg(i,:) = E'*diSdg{i};
             end
-
+ 
             % gradients w.r.t. hidden states
             %--------------------------------------------------------------
             for i = 1:mnx
@@ -385,7 +403,7 @@
                 diSdv{i}  = blkdiag(kron(Rh,diV),kron(Rg,diW));
                 dedv(i,:) = E'*diSdv{i};
             end
-
+ 
             dSdx  = kron(sparse(1,1,1,n,1),dedx);
             dSdv  = kron(sparse(1,1,1,d,1),dedv);
             dSdu  = [dSdx; dSdv];
@@ -414,8 +432,8 @@
             
             % combined curvature
             %--------------------------------------------------------------
-            dSdhh = spm_cat({dedhh  []     ;
-                             []     dedgg});
+            dSdhh = spm_cat({dedhh  [] ;
+                             [] dedgg});
                  
             
             % errors (from prior expectations) (NB pp.p = 0)
@@ -424,14 +442,14 @@
             Ep    = spm_vec(qp.p);
             Eh    = spm_vec(qh.h,qh.g) - ph.h;
             
-
+ 
             % first-order derivatives of Gibb's Energy
             %==============================================================
             dLdu  = dEdu*E + dSdu*E/2 - dDdu/2 + pu.ic*Eu;
             dLdh  = dEdh*E/2          - dDdh/2 + ph.ic*Eh;
             dLdp  = dEdp*E                     + pp.ic*Ep;
             
- 
+
             % and second-order derivatives of Gibb's Energy
             %--------------------------------------------------------------
             dLduu = dEdu*dE.du + pu.ic;
@@ -441,20 +459,22 @@
             dLdhu = dEdh*dE.du;
             dLduy = dEdu*dE.dy;
             dLduc = dEdu*dE.dc;
-            dLdpy = dEdp*dE.dy;
-            dLdpc = dEdp*dE.dc;
-            dLdhy = dEdh*dE.dy;
-            dLdhc = dEdh*dE.dc;
+            dLdhp = dEdh*dE.dp;
             dLdpu = dLdup';
-
-            
+            dLduh = dLdhu';
+            dLdph = dLdhp';
+ 
+ 
             % precision and covariances
-            %--------------------------------------------------------------                        
-            iC    = spm_cat({dLduu dLdup;
-                             dLdpu dLdpp});
+            %--------------------------------------------------------------                      
+            iC    = spm_cat({dLduu dLdup dLduh  ;
+                             dLdpu dLdpp dLdph  ;
+                             dLdhu dLdhp dLdhh});
+            
+            dLdbb = spm_sqrtm(iC(ib,ib)*iC(ib,ib)');
+            Cup   = spm_inv(iC(iup,iup));
+            Chh   = spm_inv(dLdhh);
             
-            C     = spm_inv(iC);
-            iC    = iC(iup,iup);
             
             % first-order derivatives of Entropy term
             %==============================================================
@@ -467,7 +487,7 @@
                 Lppb    = dE.dp'*diSdh{i}*dE.dp;
                 diCdh   = spm_cat({Luub Lpub';
                                    Lpub Lppb});
-                dHdh(i) = sum(sum(diCdh.*C))/2;
+                dHdh(i) = sum(sum(diCdh.*Cup))/2;
             end
             for i = 1:ng
                 Luub    = dE.du'*diSdg{i}*dE.du;
@@ -475,7 +495,7 @@
                 Lppb    = dE.dp'*diSdg{i}*dE.dp;
                 diCdg   = spm_cat({Luub Lpub';
                                    Lpub Lppb});
-                dHdg(i) = sum(sum(diCdg.*C))/2;
+                dHdg(i) = sum(sum(diCdg.*Cup))/2;
             end
             
             % parameters
@@ -486,162 +506,130 @@
                 Luup    = Luup + Luup';
                 diCdp   = spm_cat({Luup Lpup';
                                    Lpup [] });
-                dHdp(i) = sum(sum(diCdp.*C))/2;
+                dHdp(i) = sum(sum(diCdp.*Cup))/2;
             end
-
+            
+            % hidden states and causes (disabled for stability)
+            %--------------------------------------------------------------
+            for i = 1:(nu - nu)
+                Lppu    = dE.dpu{i}'*dEdp';
+                Lupu    = dEdu*dE.dpu{i};
+                Lppu    = Lppu + Lppu';
+                diCdu   = spm_cat({[]    Lupu;
+                                   Lupu' Lppu});
+                dHdu(i) = sum(sum(diCdu.*Cup))/2;
+            end
+ 
+            % and concatenate
+            %--------------------------------------------------------------
             dHdb  = [dHdh; dHdg];
-            dHdu  = [dHdx; dHdv];
-
+            dHdb  = [dHdp; dHdb];
+            dLdb  = [dLdp; dLdh];
 
+                        
             % save conditional moments (and prediction error) at Q{t}
             %==============================================================
             if iD == 1
                 
                 % save means
                 %----------------------------------------------------------
-                Q(is).e = E;
-                Q(is).E = iS*E;
-                Q(is).u = qu;
-                Q(is).p = qp;
-                Q(is).h = qh;
-                                
+                Q(is).e   = E;
+                Q(is).E   = iS*E;
+                Q(is).u   = qu;
+                Q(is).p   = qp;
+                Q(is).h   = qh;
+                
                 % and conditional covariances
                 %----------------------------------------------------------
-                Q(is).u.s = C((1:nx),(1:nx));
-                Q(is).u.c = C((1:nv) + nx*n, (1:nv) + nx*n);
-                Q(is).p.c = C((1:np) + nu,   (1:np) + nu);
-                Q(is).h.c = spm_inv(dLdhh);
-                Cu        = C(iu,iu);
-                
+                Q(is).u.s = Cup(ix,ix);
+                Q(is).u.c = Cup(iv,iv);
+                Q(is).p.c = Cup(ip,ip);
+                Q(is).h.c = Chh(ih,ih);              
                 
                 % Free-energy (components)
                 %----------------------------------------------------------
-%                 Fc(is,1)  = - E'*iSS*E/2;
-%                 Fc(is,2)  = spm_logdet(iSS)/2;
-%                 Fc(is,3)  = - Eu'*pu.ic*Eu/2;
-%                 Fc(is,4)  = spm_logdet(pp.ic)/2;
-%                 Fc(is,5)  = - Ep'*pp.ic*Ep/2;
-%                 Fc(is,6)  = spm_logdet(pp.ic)/2;
-%                 Fc(is,7)  = - Eh'*ph.ic*Eh/2;
-%                 Fc(is,8)  = spm_logdet(ph.ic)/2;
-%                 Fc(is,9)  = - spm_logdet(iC)/2;
-%                 Fc(is,10) = - spm_logdet(dLdhh)/2;
-                
+                Fc(is,1)  = - E(je)'*iS(je,je)*E(je)/2;
+                Fc(is,2)  = - Eu(ju)'*pu.ic(ju,ju)*Eu(ju)/2;
+                Fc(is,3)  = - n*ny*log(2*pi)/2;
+                Fc(is,4)  = spm_logdet(iS(je,je))/2;
+                Fc(is,5)  = spm_logdet(pu.ic(ju,ju)*Cup(ju,ju))/2;
 
-                % Free-energy
-                %----------------------------------------------------------
-                L(is) = - E'*iSS*E/2       + spm_logdet(iSS)/2   ...          
-                        - Eu'*pu.ic*Eu/2   + spm_logdet(pu.ic)/2 ...
-                        - n*ny*log(2*pi)/2;
-                    
-                % Free-energy (states and parameters)
+                                
+                % Free-action (states and parameters)
                 %----------------------------------------------------------
-                A(is) = L(is)                                    ...
-                        - Ep'*pp.ic*Ep/2   + spm_logdet(pp.ic)/2 ...
-                        - Eh'*ph.ic*Eh/2   + spm_logdet(ph.ic)/2 ...
-                        - spm_logdet(iC)/2 - spm_logdet(dLdhh)/2;
-                    
-                % Free-energy (states)
-                %----------------------------------------------------------                
-                L(is) = L(is) + spm_logdet(Cu)/2;
-                
+                AC(is)    = sum(Fc(is,:))       ...
+                          - Ep'*pp.ic*Ep/2      ...
+                          - Eh'*ph.ic*Eh/2      ...
+                          + spm_logdet(pp.ic)/2 ...
+                          + spm_logdet(ph.ic)/2 ...
+                          - spm_logdet(dLdbb)/2;
+  
             end
  
             % update conditional moments
             %==============================================================
             
-            % precision of fluctuations on parameters of hyperparameters
+            % prior precision of fluctuations on [hyper] parameters
             %--------------------------------------------------------------
-            Kp     = ns*Ip;
-            Kh     = ns*Ih*2;
+            Kb    = ns*Ib;
             
-            % uopdate curvatures of [hyper]paramters
+            % accumulate curvatures of [hyper] parameters
             %--------------------------------------------------------------
-            nk     = 128;
             try
-                dLdPP = dLdPP*(1 - 1/nk) + dLdpp/nk;
-                dLdHH = dLdHH*(1 - 1/nk) + dLdhh/nk;
+                dLdBB = dLdBB*(1 - 1/ns) + dLdbb/ns;
+
             catch
-                dLdPP = dLdpp*16;
-                dLdHH = dLdhh*16;
+                dLdBB = dLdbb + Ib*32;
             end
 
-            % rotate and scale gradient (and curvatures)
+            % whiten gradient (and curvatures) with regularised precision
             %--------------------------------------------------------------
-            [Vp Sp] = spm_svd(dLdPP,0);
-            [Vh Sh] = spm_svd(dLdHH,0);
-            Sp      = diag(1./(diag(Sp)));
-            Sh      = diag(1./(diag(Sh)));
-            
-            dLdp  = Sp*Vp'*dLdp;
-            dHdp  = Sp*Vp'*dHdp;
-            dLdpy = Sp*Vp'*dLdpy;
-            dLdpu = Sp*Vp'*dLdpu;
-            dLdpc = Sp*Vp'*dLdpc;
-
-            dLdh  = Sh*Vh'*dLdh;
-            dHdb  = Sh*Vh'*dHdb;
-            dLdhy = Sh*Vh'*dLdhy;
-            dLdhu = Sh*Vh'*dLdhu;
-            dLdhc = Sh*Vh'*dLdhc;
- 
-
+            IB    = diag(diag(dLdBB));
+            Cb    = spm_inv(dLdBB + IB*exp(dt));
+            dLdb  = Cb*dLdb;
+            dHdb  = Cb*dHdb;
             
             % assemble conditional means
             %--------------------------------------------------------------
-            q_p   = spm_unvec(Vp'*spm_vec(qp.p),qp.p);
-            q_b   = spm_unvec(Vh'*spm_vec({qh.h qh.g}),{qh.h qh.g});
-            q{1}  = qu.y(1:n);
-            q{2}  = qu.x(1:n);
-            q{3}  = qu.v(1:d);
-            q{4}  = qu.u(1:d);
-            q{5}  = q_p;
-            q{6}  = q_b{1};
-            q{7}  = q_b{2};
-            q{8}  = Vp'*qp.dp;
-            q{9}  = Vh'*qh.dp;
-            
-
-
+            q.y  = qu.y(1:n);
+            q.x  = qu.x(1:n);
+            q.v  = qu.v(1:d);
+            q.c  = qu.u(1:d);
+            q.p  = qp.p;
+            q.h  = qh.h;
+            q.g  = qh.g;
+            q.d  = dbdt;
+                        
             % flow
             %--------------------------------------------------------------
-            f{1}  =  Dy*spm_vec(q{1});
-            f{2}  =  Du*spm_vec(q{2:3}) - dLdu - dHdu;
-            f{3}  =  Dc*spm_vec(q{4});
-            f{4}  =     spm_vec(q{8});
-            f{5}  =     spm_vec(q{9});
-            f{6}  = -Kp*spm_vec(q{8})   - dLdp - dHdp;
-            f{7}  = -Kh*spm_vec(q{9})   - dLdh - dHdb;
-            
- 
+            f.y  =  Dy*spm_vec(q.y)                  ;
+            f.u  =  Du*spm_vec(q.x,q.v) - dLdu - dHdu;
+            f.c  =  Dc*spm_vec(q.c)                  ;
+            f.b  =     spm_vec(q.d)                  ;
+            f.d  = -Kb*spm_vec(q.d)     - dLdb - dHdb;
+
             % and Jacobian
             %--------------------------------------------------------------
-            dfdq  = spm_cat({Dy      []       []     []   []   []   [];
-                            -dLduy  Du-dLduu -dLduc  []   []   []   [];
-                             []      []       Dc     []   []   []   [];
-                             []      []       []     []   []   Ip   [];
-                             []      []       []     []   []   []   Ih;
-                            -dLdpy  -dLdpu   -dLdpc -Ip   []  -Kp   [];
-                            -dLdhy  -dLdhu   -dLdhc  []  -Ih   []  -Kh});
- 
+            dfdq = {Dy     []        []     []    [] ;
+                   -dLduy  Du-dLduu -dLduc  []    [] ;
+                    []     []        Dc     []    [] ;
+                    []     []        []     []    Ib ;
+                    []     []        []    -Ib   -Kb};
+          
  
             % update conditional modes of states
             %==============================================================
-            dq    = spm_dx(dfdq, spm_vec(f), 1/nD);
-            q     = spm_unvec(spm_vec(q) + dq,q);
+            dq   = spm_dx(spm_cat(dfdq),spm_vec(f),1/nD);
+            q    = spm_unvec(spm_vec(q) + dq,q);
             
             % unpack conditional means
             %--------------------------------------------------------------
-            q_p       = spm_unvec(Vp*spm_vec(q{5}),qp.p);
-            q_b       = spm_unvec(Vh*spm_vec(q{6:7}),{qh.h qh.g});
-            qu.x(1:n) = q{2};
-            qu.v(1:d) = q{3};
-            qp.p      = q_p;
-            qh.h      = q_b{1};
-            qh.g      = q_b{2};
-            qp.dp     = Vp*q{8};
-            qh.dp     = Vh*q{9};
-
+            qu.x(1:n) = q.x;
+            qu.v(1:d) = q.v;
+            qp.p      = q.p;
+            qh.h      = q.h;
+            qh.g      = q.g;
+            dbdt      = q.d;
  
         end % D-Step
  
@@ -650,55 +638,110 @@
     
     % Bayesian parameter averaging
     %======================================================================
-
+ 
     % Conditional moments of time-averaged parameters
     %----------------------------------------------------------------------
-    Pp  = 0;
-    Ep  = 0;
+    Ep    = 0;
+    Pp    = 0;
     for i = 1:ns
         P   = spm_inv(Q(i).p.c);
-        Ep  = Ep + P*spm_vec(Q(i).p.p);
+        Ep  = Ep  + P*spm_vec(Q(i).p.p);
         Pp  = Pp + P;       
     end
-    Cp  = spm_inv(Pp);
-    Ep  = Cp*Ep;
-
+    Ep    = spm_inv(Pp)*Ep;
+    Cp    = spm_inv(Pp + (1 - ns)*pp.ic);
+ 
     % conditional moments of hyper-parameters
     %----------------------------------------------------------------------
-    Ph  = 0;
-    Eh  = 0;
+    Eh    = 0;
+    Ph    = 0;
     for i = 1:ns
         P   = spm_inv(Q(i).h.c);
+        Eh  = Eh  + P*spm_vec({Q(i).h.h Q(i).h.g});
         Ph  = Ph + P;
-        Eh  = Eh + P*spm_vec({Q(i).h.h Q(i).h.g});
     end
-    Ch  = spm_inv(Ph);
-    Eh  = Ch*Eh - ph.h;
-
+    Eh    = spm_inv(Ph)*Eh - ph.h;
+    Ch    = spm_inv(Ph + (1 - ns)*ph.ic);
+ 
     % Free-action of states plus free-energy of parameters
     %======================================================================
-    Fs  = sum(A);
-    Fi  = sum(L) ...
-          - Ep'*pp.ic*Ep/2 + spm_logdet(pp.ic)/2 - spm_logdet(Pp)/2 ...
-          - Eh'*ph.ic*Eh/2 + spm_logdet(ph.ic)/2 - spm_logdet(Ph)/2;
-      
-    try, disp(full(sum(Fc))), end
-
-    % if F is increasing terminate
+    FC(1) = sum(Fc(:,1));       % - E'*iS*E/2;
+    FC(2) = sum(Fc(:,2));       % - Eu'*pu.ic*Eu/2;
+    FC(3) = sum(Fc(:,3));       % - n*ny*log(2*pi)/2;
+    FC(4) = sum(Fc(:,4));       %   spm_logdet(iS)/2;
+    FC(5) = sum(Fc(:,5));       %   spm_logdet(pu.ic*Cu)/2;
+    FC(6) = -Ep'*pp.ic*Ep/2;
+    FC(7) = -Eh'*ph.ic*Eh/2;
+    FC(8) = spm_logdet(pp.ic*Cp)/2;
+    FC(9) = spm_logdet(ph.ic*Ch)/2;
+    
+    CC(iN,:) = FC;
+    S(iN)    = sum(AC);
+    Fe       = sum(FC);
+ 
+    % if F is decreasing, revert [hyper] parameters and slow down
     %----------------------------------------------------------------------
-    if Fi < Fa && iN > 2
-        break
+    if Fe < F(iN)
+        
+        % start again if F never increased
+        %------------------------------------------------------------------
+        if iN == 2 && dt < 8
+            DEM.M(1).E.v = DEM.M(1).E.v + 2;
+            DEM = spm_LAP(DEM);
+            return
+        end
+                
+        % save free-energy
+        %------------------------------------------------------------------
+        F(iN + 1) = F(iN);
+               
+        % load current MAP estimates
+        %------------------------------------------------------------------
+        qp = PQ.qp;
+        qh = PQ.qh;
+        
+        % decrease update time
+        %------------------------------------------------------------------
+        dt = max(dt + 2,2);
+        
+        % convergence
+        %------------------------------------------------------------------
+        if dt > 6; convergence = 1; end
+        
     else
-        Fa    = Fi;
-        F(iN) = Fi;
-        S(iN) = Fs;
+        
+        % convergence
+        %------------------------------------------------------------------
+        if Fe - F(iN) < 1e-2 && iN > 4
+            convergence = 1; 
+        else
+            convergence = 0; 
+        end
+            
+        % save free-energy
+        %------------------------------------------------------------------
+        F(iN)     = Fe;
+        F(iN + 1) = Fe;
+        
+        % save current MAP estimates
+        %------------------------------------------------------------------
+        PQ.qp = qp;
+        PQ.qh = qh;
+        
+        % increase update time
+        %------------------------------------------------------------------
+        dt    = max(dt - 1,-8);
+        
     end
  
+    % Convergence
+    %======================================================================
+    if convergence; break, end
+    
     % otherwise save conditional moments (for each time point)
     %======================================================================
     for t = 1:length(Q)
  
- 
         % states and predictions
         %------------------------------------------------------------------
         v     = spm_unvec(Q(t).u.v{1},v);
@@ -717,7 +760,7 @@
         end
         if    M(nl).l, qU.z{nl}(:,t)    = spm_vec(z{nl}); end
         if    M(nl).l, qU.Z{nl}(:,t)    = spm_vec(Z{nl}); end
-
+ 
         qU.v{1}(:,t)  = spm_vec(Q(t).u.y{1}) - spm_vec(z{1});
  
         % and conditional covariances
@@ -766,14 +809,14 @@
         plot(1:ns,spm_cat(qP.p))
         set(gca,'XLim',[1 ns])
         title('parameters (modes)','FontSize',16)
-
+ 
     end
     drawnow
  
     % report (EM-Steps)
     %----------------------------------------------------------------------
     try
-        dF = F(end) - F(end - 1);
+        dF = F(iN) - F(iN - 1);
     catch
         dF = 0;
     end
@@ -802,12 +845,11 @@
     Pp = Pp + P;
  
 end
-Cp     = spm_inv(Pp);
-Ep     = Cp*Ep;
-P      = {M.pE};
-qP.P   = spm_unvec(Up*Ep + pp.p,P);
+Ep     = spm_inv(Pp)*Ep;
+Cp     = spm_inv(Pp + (1 - ns)*pp.ic);
+qP.P   = spm_unvec(Up*Ep + pp.p,{M.pE});
 qP.C   = Up*Cp*Up';
-qP.V   = spm_unvec(diag(qP.C),P);
+qP.V   = spm_unvec(diag(qP.C),{M.pE});
 qP.U   = Up;
  
 % conditional moments of hyper-parameters
@@ -819,21 +861,20 @@
     % weight in proportion to precisions
     %----------------------------------------------------------------------
     P  = spm_inv(qH.c{i});
-    Ph = Ph + P;
     Eh = Eh + P*qH.p{i};
+    Ph = Ph + P;
  
 end
-Ch     = spm_inv(Ph);
-Eh     = Ch*Eh;
-P      = {qh.h qh.g};
-P      = spm_unvec(Eh,P);
+Eh     = spm_inv(Ph)*Eh;
+Ch     = spm_inv(Ph + (1 - ns)*ph.ic);
+P      = spm_unvec(Eh,{qh.h qh.g});
 qH.h   = P{1};
 qH.g   = P{2};
 qH.C   = Ch;
 P      = spm_unvec(diag(qH.C),P);
 qH.V   = P{1};
 qH.W   = P{2};
-
+ 
  
  
 % assign output variables
@@ -845,15 +886,12 @@
 DEM.qP = qP;                  % conditional moments of model-parameters
 DEM.qH = qH;                  % conditional moments of hyper-parameters
  
-DEM.F  = F;                   % [-ve] Free energy
-DEM.S  = S;                   % [-ve] Free action
+DEM.F  = F(1:iN);             % [-ve] Free-energy
+DEM.S  = S(1:iN);             % [-ve] Free-action
+DEM.FC = FC;                  % Free-energy components
+DEM.CC = CC;                  % over iterations
 
 return
 
-
-
-
-
-
-
-
+ 
+ 
diff --git a/spm_Menu.fig b/spm_Menu.fig
index 6b19620..9d2402a 100644
Binary files a/spm_Menu.fig and b/spm_Menu.fig differ
diff --git a/spm_Ncdf.m b/spm_Ncdf.m
index 29952b1..b37b920 100644
--- a/spm_Ncdf.m
+++ b/spm_Ncdf.m
@@ -48,10 +48,10 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1995-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Ncdf.m 1154 2008-02-15 16:08:15Z guillaume $
+% $Id: spm_Ncdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
@@ -61,13 +61,14 @@
 if nargin<1, F=[]; return, end
 ad = [ndims(x);ndims(u);ndims(v)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(u),ones(1,rd-ad(2))];...
-    [size(v),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(u),ones(1,rd-ad(2))];...
+      [size(v),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
 %--------------------------------------------------------------------------
diff --git a/spm_Npdf.m b/spm_Npdf.m
index 5e61e78..68cc332 100644
--- a/spm_Npdf.m
+++ b/spm_Npdf.m
@@ -6,13 +6,13 @@
 % u - mean              [Defaults to 0]
 % v - variance  (v>0)   [Defaults to 1]
 % f - pdf of N(u,v) at x
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Npdf returns the Probability Density Function (PDF) for the
 % univariate Normal (Gaussian) family of distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Let random variable X have a Normal distribution with mean u and
 % variance v, then Z~N(u,v). The Probability Density Function (PDF) of
 % the Normal (sometimes called Gaussian) family is f(x), defined on all
@@ -26,7 +26,7 @@
 % variance, Z~N(0,1), is commonly referred to as \phi(z).
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -39,38 +39,40 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1994-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Npdf.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_Npdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, v=1; end
 if nargin<2, u=0; end
 if nargin<1, f=[]; return, end
 ad = [ndims(x);ndims(u);ndims(v)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(u),ones(1,rd-ad(2))];...
-    [size(v),ones(1,rd-ad(3))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(u),ones(1,rd-ad(2))];...
+      [size(v),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for strictly positive variance v. Return NaN if undefined.
 md = ( ones(size(x))  &  ones(size(u))  &  v>0 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero where defined
 Q  = find( md );
diff --git a/spm_PEB.m b/spm_PEB.m
index 4bd5d8d..e154523 100644
--- a/spm_PEB.m
+++ b/spm_PEB.m
@@ -55,7 +55,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_PEB.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_PEB.m 4283 2011-04-01 16:53:36Z karl $
 
 % set default
 %--------------------------------------------------------------------------
@@ -76,7 +76,7 @@
     if ~isfield(P{i},'C')
         [n m] = size(P{i}.X);
         if i == 1
-            P{i}.C            = {speye(n,n)};
+            P{i}.C  = {speye(n,n)};
         else
             for j = 1:m
                 k         = find(P{i}.X(:,j));
@@ -107,8 +107,8 @@
     % indices for ith level parameters
     %----------------------------------------------------------------------
     [n m] = size(P{i}.X);
-    I{i}  = [1:n] + I{end}(end);
-    J{i}  = [1:m] + J{end}(end);
+    I{i}  = (1:n) + I{end}(end);
+    J{i}  = (1:m) + J{end}(end);
 
 end
 
@@ -241,7 +241,7 @@
             Ce = Ce + Q{i}*h(i);
         end
     end
-    iC    = spm_inv(Ce);
+    iC    = spm_inv(Ce,exp(-16));
 
     % E-step: conditional mean E{B|y} and covariance cov(B|y)
     %======================================================================
@@ -298,7 +298,8 @@
     % Convergence
     %======================================================================
     w     = norm(dh,1);
-%     fprintf('%-30s: %i %30s%e\n','  PEB Iteration',k,'...',full(w));
+    
+    % fprintf('%-30s: %i %30s%e\n','  PEB Iteration',k,'...',full(w));
     
     % if dF < 0.01
     %----------------------------------------------------------------------
@@ -362,11 +363,3 @@
 %--------------------------------------------------------------------------
 if k == M, warning('maximum number of iterations exceeded'), end
 
-return
-
-function [C] = spm_inv(C);
-% inversion of sparse matrices
-%__________________________________________________________________________
-
-C  = inv(C + speye(length(C))*exp(-32));
-
diff --git a/spm_P_FDR.m b/spm_P_FDR.m
index 0f15cbb..e968128 100644
--- a/spm_P_FDR.m
+++ b/spm_P_FDR.m
@@ -58,8 +58,8 @@
 % Ser B.  57:289-300.
 %
 % Benjamini & Yekutieli (2001), "The Control of the false discovery rate
-% in multiple testing under dependency". To appear, Annals of Statistics.
-% Available at http://www.math.tau.ac.il/~benja 
+% in multiple testing under dependency". Annals of Statistics, 
+% 29(4):1165-1188.
 %
 % Yekutieli & Benjamini (1999). "Resampling-based false discovery rate
 % controlling multiple test procedures for correlated test
@@ -68,7 +68,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Thomas Nichols
-% $Id: spm_P_FDR.m 3982 2010-07-09 13:36:53Z guillaume $
+% $Id: spm_P_FDR.m 4013 2010-07-22 17:12:45Z guillaume $
 
 
 %-Set Benjamini & Yeuketeli cV for independence/PosRegDep case
diff --git a/spm_P_RF.m b/spm_P_RF.m
index 2a6f32a..1832e9c 100644
--- a/spm_P_RF.m
+++ b/spm_P_RF.m
@@ -49,7 +49,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_P_RF.m 3994 2010-07-13 15:53:30Z guillaume $
+% $Id: spm_P_RF.m 4225 2011-03-02 15:53:05Z guillaume $
 
 
 % get expectations
@@ -112,9 +112,6 @@
 
 % set P and p = [] for non-implemented cases
 %++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
-if k > 0 && n > 1
-    P    = []; p = [];
-end
 if k > 0 && (STAT == 'X' || STAT == 'F')
     P    = []; p = [];
 end
diff --git a/spm_Pcdf.m b/spm_Pcdf.m
index b7e9acc..2e191f5 100644
--- a/spm_Pcdf.m
+++ b/spm_Pcdf.m
@@ -56,10 +56,10 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1996-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Pcdf.m 2975 2009-03-26 21:43:31Z guillaume $
+% $Id: spm_Pcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
@@ -69,12 +69,13 @@
 
 ad = [ndims(x);ndims(l)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(l),ones(1,rd-ad(2))];    ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(l),ones(1,rd-ad(2))];];
 rs = max(as);
 xa = prod(as,2)>1;
-if all(xa) & any(diff(as(xa,:)))
-    error('non-scalar args must match in size'), end
+if all(xa) && any(diff(as(xa,:)))
+    error('non-scalar args must match in size');
+end
 
 
 %-Computation
@@ -84,9 +85,11 @@
 
 %-Only defined for l>0. Return NaN if undefined.
 md = ( ones(size(x))  &  l>0 );
-if any(~md(:)), F(~md) = NaN;
+if any(~md(:))
+    F(~md) = NaN;
     warning('SPM:outOfRangePoisson',...
-        'Returning NaN for out of range arguments'), end
+        'Returning NaN for out of range arguments');
+end
 
 %-Non-zero only where defined and x>=0
 Q  = find( md  &  x>=0 );
diff --git a/spm_Ppdf.m b/spm_Ppdf.m
index 5bf16e2..dc8538e 100644
--- a/spm_Ppdf.m
+++ b/spm_Ppdf.m
@@ -5,13 +5,13 @@
 % x - ordinates
 % l - Poisson mean parameter (lambda l>0) [Defaults to 1]
 % f - Poisson PDF
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Ppdf implements the Probaility Distribution Function of the
 % Poisson distribution.
 % 
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Poisson Po(l) distribution is the distribution of the number of
 % events in unit time for a stationary Poisson process with mean
 % parameter lambda=1, or equivalently rate 1/l. If random variable X is
@@ -25,19 +25,19 @@
 %           {  0                          otherwise
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % To avoid roundoff errors for large x (in x! & l^x) & l (in l^x),
 % computation is done in logs.
 %
 % Normal approximation:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % For large lambda the normal approximation Y~:~N(l,l) may be used.
 % With continuity correction this gives
 % f(x) ~=~ Phi((x+.5-l)/sqrt(l)) -Phi((x-.5-l)/sqrt(l));
 % where Phi is the standard normal CDF, and ~=~ means "appox. =".
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -50,37 +50,40 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1996-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Ppdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Ppdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, l=1; end
 if nargin<1, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(l)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(l),ones(1,rd-ad(2))];    ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(l),ones(1,rd-ad(2))];];
 rs = max(as);
 xa = prod(as,2)>1;
-if all(xa) & any(diff(as(xa,:)))
-    error('non-scalar args must match in size'), end
+if all(xa) && any(diff(as(xa,:)))
+    error('non-scalar args must match in size');
+end
 
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for l>0. Return NaN if undefined.
 md = ( ones(size(x))  &  l>0 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero only where defined and x is whole
 Q  = find( md  &  x>=0  &  x==floor(x) );
diff --git a/spm_Q.m b/spm_Q.m
index 929216c..13c0aed 100644
--- a/spm_Q.m
+++ b/spm_Q.m
@@ -2,7 +2,7 @@
 % returns an (n x n) autocorrelation matrix for an AR(p) process
 % FORMAT [Q] = spm_Q(a,n,q)
 %
-% A  - vector pf p AR coeficients
+% A  - vector pf p AR coefficients
 % n  - size of Q
 % q  - swtich to return precision [default q = 0]
 %__________________________________________________________________________
@@ -15,44 +15,35 @@
 % 
 % cov(y) = <K*z*z'*K> = K*K'
 % 
-% Critically, this is not the correlation because if cov(z) = eye(n) 
-% then trace(cov(y)) ~= n.
-% The reason the diagonals of corr(y) are not constant is that we 
-% are modeling finite length AR sequences, which incur boundary effects 
-% at the beginning and end of the sequence.  These are resolved with a
-% Toeplitz form
+% If called with q ~= 0, a first order process is assumed when evaluating
+% the precision (inverse covariance) matrix; i.e., a = a(1)
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
-
+ 
 % Karl Friston
-% $Id: spm_Q.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_Q.m 4230 2011-03-07 20:58:38Z karl $
+ 
 % default
 %--------------------------------------------------------------------------
 try, q; catch, q = 0; end
-
-
-% compute P
-%--------------------------------------------------------------------------
-p    = length(a);
+ 
 if q
-
-    % compute P
+ 
+    % compute P (precision)
     %----------------------------------------------------------------------
     A    = [-a(1) (1 + a(1)^2) -a(1)];
-    Q    = spdiags(ones(n,1)*A,[-p:p],n,n);
-
+    Q    = spdiags(ones(n,1)*A,[-1:1],n,n);
+ 
 else
-
-    % compute Q
+ 
+    % compute Q (covariance)
     %----------------------------------------------------------------------
+    p    = length(a);
     A    = [1 -a(:)'];
     P    = spdiags(ones(n,1)*A,-[0:p],n,n);
     K    = inv(P);
     K    = K.*(abs(K) > 1e-4);
     Q    = K*K';
     Q    = toeplitz(Q(:,1));
-
-end
-
-
+ 
+end
\ No newline at end of file
diff --git a/spm_SpUtil.m b/spm_SpUtil.m
index 489cef4..0014932 100644
--- a/spm_SpUtil.m
+++ b/spm_SpUtil.m
@@ -256,14 +256,14 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes Jean-Baptiste Poline
-% $Id: spm_SpUtil.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_SpUtil.m 4137 2010-12-15 17:18:32Z guillaume $
 
 % (frobenius norm trick by S. Rouquette)
 
 %-Format arguments
 %-----------------------------------------------------------------------
 if nargin==0, error('do what? no arguments given...')
-    else, action = varargin{1}; end
+else action = varargin{1}; end
 
 
 switch lower(action), 
@@ -271,14 +271,14 @@
 case {'iscon','allcon','conr','cono'}
 %=======================================================================
 % i = spm_SpUtil('isCon',x,c)
-if nargin==0, varargout={[]}, error('isCon : no argument specified'), end;
+if nargin==0, varargout={[]}; error('isCon : no argument specified'), end;
 if nargin==1, 
    varargout={[]}; warning('isCon : no contrast specified'); return; 
 end;
 if ~spm_sp('isspc',varargin{2})
-    sX = spm_sp('Set',varargin{2});
-else,   sX = varargin{2}; end
-if nargin==2, c=eye(spm_sp('size',sX,2)); else, c=varargin{3}; end;
+     sX = spm_sp('Set',varargin{2});
+else sX = varargin{2}; end
+if nargin==2, c=eye(spm_sp('size',sX,2)); else c=varargin{3}; end;
 if isempty(c), varargout={[]}; return, end
 
 switch lower(action)
@@ -312,18 +312,18 @@
 
 %--------- begin argument check ------------------------------
 if nargin ~= 3, error(['Wrong number of arguments in ' action])
-else,   sX = varargin{2}; c = varargin{3}; end;
+else sX = varargin{2}; c = varargin{3}; end;
 if nargout > 2, error(['Too many output arguments in ' action]), end;
 if ~spm_sp('isspc',sX),  sX = spm_sp('set',varargin{2}); end;
 if sX.rk == 0, error('c->Tsp null rank sX == 0'); end;
-if ~isempty(c) & spm_sp('size',sX,2) ~= size(c,1), 
+if ~isempty(c) && spm_sp('size',sX,2) ~= size(c,1), 
    error(' c->TSp matrix & contrast dimensions don''t match');   
 end
 %--------- end argument check --------------------------------- 
 
 %- project c onto the space of  X' if needed 
 %-------------------------------------------
-if ~isempty(c) & ~spm_sp('isinspp',sX,c),
+if ~isempty(c) && ~spm_sp('isinspp',sX,c),
    warning([sprintf('\n') 'c is not a proper contrast in ' action ...
             ' in ' mfilename sprintf('\n') '!!! projecting...' ]);
    disp('from'), c, disp('to'), c = spm_sp('oPp:',sX,c)
@@ -335,40 +335,40 @@
 % case 0
 %   warning(['no output demanded in ' mfilename ' ' action])
 case {0,1}
-   if ~isempty(c) & any(any(c))                 %- c not empty & not null
+   if ~isempty(c) && any(any(c))               %- c not empty & not null
       if cukFlag, varargout = { spm_sp('cukxp-:',sX,c) };
       else        varargout = { spm_sp('xp-:',sX,c) };
       end
-   else if isempty(c), varargout = { [] };      %- c empty
-        else,                                   %- c null
+   else if isempty(c), varargout = { [] };     %- c empty
+       else                                    %- c null
            if cukFlag, varargout = { spm_sp('cukx',sX,c) }; 
-           else,       varargout = { spm_sp('x',sX)*c };
+           else        varargout = { spm_sp('x',sX)*c };
            end
-        end;
+       end
    end
 
 case 2
-   if ~isempty(c) & any(any(c))         %- not empty and not null
+   if ~isempty(c) && any(any(c))         %- not empty and not null
       if cukFlag,
          varargout = {  
-        spm_sp('cukxp-:',sX,c), ...             %- X1o
-        spm_sp('cukx',sX,spm_sp('r',spm_sp('set',c))) };    %- X0
+        spm_sp('cukxp-:',sX,c), ...      %- X1o
+        spm_sp('cukx',sX,spm_sp('r',spm_sp('set',c))) };      %- X0
       else
          varargout = {  
-        spm_sp(':',sX, spm_sp('xp-:',sX,c)), ...            %- X1o
+        spm_sp(':',sX, spm_sp('xp-:',sX,c)), ...              %- X1o
         spm_sp(':',sX, ...
-               spm_sp('x',sX)*spm_sp('r',spm_sp('set',c))) };   %- X0
+               spm_sp('x',sX)*spm_sp('r',spm_sp('set',c))) }; %- X0
       end
    else            
-      if isempty(c),                    %- empty
+      if isempty(c),                     %- empty
          if cukFlag, varargout = { [], spm_sp('cukx',sX) }; 
-         else,       varargout = { [], spm_sp('x',sX) };
+         else        varargout = { [], spm_sp('x',sX) };
          end
-      else,                             %- null
+      else                               %- null
          if cukFlag, 
             varargout = { spm_sp(':',sX,spm_sp('cukx',sX,c)), ...
                           spm_sp(':',sX,spm_sp('cukx',sX)) }; 
-         else,
+         else
             varargout = { spm_sp('x',sX)*c, spm_sp('x',sX)};
          end
       end;
@@ -386,7 +386,7 @@
 % arguments are checked in calls to spm_Util
 %--------------------------------------------
 if nargin<3, error('Insufficient arguments'),  
-else, sX = varargin{2}; i0 = varargin{3}; end;
+else sX = varargin{2}; i0 = varargin{3}; end;
 
 cukFlag = strcmp(lower(action),'+i0->x1o');
 
@@ -416,7 +416,7 @@
 
 %--------- begin argument check --------------------------------
 if nargin<3, error('Insufficient arguments'),  
-else, sX = varargin{2}; i0 = varargin{3}; end;
+else sX = varargin{2}; i0 = varargin{3}; end;
 if ~spm_sp('isspc',sX),  sX = spm_sp('set',varargin{2}); end;
 if spm_sp('rk',sX) == 0, error('i0->c null rank sX == 0'); end;
 sL  = spm_sp('size',sX,2);
@@ -461,25 +461,25 @@
 
 %--------- begin argument check --------- 
 if nargin<3, error('Insufficient arguments'),  
-else, 
+else
    sX = varargin{2}; 
-   if cukFlag, cukX0 = varargin{3}; else, X0 = varargin{3}; end 
-end;
-if ~spm_sp('isspc',sX),  sX = spm_sp('set',varargin{2}); end;
-if spm_sp('rk',sX) == 0, error(['null rank sX == 0 in ' action]); end;
+   if cukFlag, cukX0 = varargin{3}; else X0 = varargin{3}; end 
+end
+if ~spm_sp('isspc',sX),  sX = spm_sp('set',varargin{2}); end
+if spm_sp('rk',sX) == 0, error(['null rank sX == 0 in ' action]); end
 if cukFlag
-   if ~isempty(cukX0) & spm_sp('rk',sX) ~= size(cukX0,1),
+   if ~isempty(cukX0) && spm_sp('rk',sX) ~= size(cukX0,1),
       cukX0, spm_sp('rk',sX),
-      error(['cukX0 of wrong size ' mfilename ' ' action]), end;
+      error(['cukX0 of wrong size ' mfilename ' ' action]), end
 else
-   if ~isempty(X0) & spm_sp('size',sX,1) ~= size(X0,1),
+   if ~isempty(X0) && spm_sp('size',sX,1) ~= size(X0,1),
       X0, spm_sp('size',sX,1),
-      error(['X0 of wrong size ' mfilename ' ' action]),X0, end;
+      error(['X0 of wrong size ' mfilename ' ' action]),X0, end
 end
 %--------- end argument check --------- 
 
 if cukFlag
-   if isempty(cukX0), X0 = []; else, X0 = spm_sp('ox',sX)*cukX0; end
+   if isempty(cukX0), X0 = []; else X0 = spm_sp('ox',sX)*cukX0; end
 end
 
 varargout = { sf_X0_2_c(X0,sX) };
@@ -520,22 +520,22 @@
     warning('space with no dimension ');    
     if nargout==1, varargout = {[]};
     else varargout = {[], []}; end
-else, 
+else
 
-   if nargin > 2 & ~isempty(varargin{3})    
+   if nargin > 2 && ~isempty(varargin{3})    
 
     V = varargin{3};
     u = sX.u(:,1:rk);
     clear sX;
     if nargout==1
         %-only trRV needed
-        if rk==0 | isempty(rk),  trMV = 0; 
-        else,   trMV = sum(sum( u .* (V*u) ));
-        end;
+        if rk==0 || isempty(rk),  trMV = 0; 
+        else trMV = sum(sum( u .* (V*u) ));
+        end
         varargout = { trace(V) - trMV};
     else
         %-trRVRV is needed as well
-        if rk==0 | isempty(rk),  
+        if rk==0 || isempty(rk),  
             trMV = 0; 
             trRVRV = (norm(V,'fro'))^2;
          trV = trace(V);
@@ -556,13 +556,13 @@
    else  %- nargin == 2 | isempty(varargin{3})
 
     if nargout==1
-        if rk==0 | isempty(rk), varargout = {sL}; 
-        else, varargout = {sL - rk}; 
-        end;
+        if rk==0 || isempty(rk), varargout = {sL}; 
+        else varargout = {sL - rk}; 
+        end
     else
-        if rk==0 | isempty(rk),  varargout = {sL,sL};
-        else, varargout = {sL - rk, sL - rk};
-        end;    
+        if rk==0 || isempty(rk),  varargout = {sL,sL};
+        else varargout = {sL - rk, sL - rk};
+        end    
     end
 
     end
@@ -592,7 +592,7 @@
     return; 
 end;
 
-if nargin > 2 & ~isempty(varargin{3}) %- V provided, and assumed correct !
+if nargin > 2 && ~isempty(varargin{3}) %- V provided, and assumed correct !
 
     V = varargin{3};
     u = sX.u(:,1:rk);
@@ -628,10 +628,10 @@
 %-----------------------------------------------------------------------
 %--------- begin argument check ---------------------------------------- 
 if nargin<3, error('insufficient arguments'),
-else, i0 = varargin{3}; sX = varargin{2}; end
+else i0 = varargin{3}; sX = varargin{2}; end
 if ~spm_sp('isspc',sX), sX = spm_sp('Set',sX); end;
 i0  = sf_check_i0(i0,spm_sp('size',sX,2));
-if nargin == 4, V=varargin{4}; else, V = eye(spm_sp('size',sX,1)); end;
+if nargin == 4, V=varargin{4}; else V = eye(spm_sp('size',sX,1)); end;
 if nargin>4, error('Too many input arguments'), end;
 %--------- end argument check ------------------------------------------ 
 
@@ -653,19 +653,19 @@
 %=======================================================================
 % sz = spm_SpUtil('size',x,dim)
 
-if nargin<3, dim=[]; else, dim = varargin{3}; end
+if nargin<3, dim=[]; else dim = varargin{3}; end
 if nargin<2, error('insufficient arguments'), end
 
 if isstruct(varargin{2})
    if isfield(varargin{2},'X')
     sz = size(varargin{2}.X);
-   else, error('no X field'); end;  
+   else error('no X field'); end;  
 else
     sz = size(varargin{2});
 end
 
 if ~isempty(dim)
-    if dim>length(sz), sz = 1; else, sz = sz(dim); end
+    if dim>length(sz), sz = 1; else sz = sz(dim); end
     varargout = {sz};
 elseif nargout>1
     varargout = cell(1,min(nargout,length(sz)));
@@ -680,7 +680,7 @@
 % i0c = spm_SpUtil('iX0check',i0,sL)
 
 if nargin<3, error('insufficient arguments'),
-else, i0 = varargin{2}; sL = varargin{3}; end;
+else i0 = varargin{2}; sL = varargin{3}; end;
 
 varargout = {sf_check_i0(i0,sL)};
 
@@ -700,10 +700,10 @@
 % NB : [] = sf_check_i0([],SL);
 %
 
-if all(ismember(i0,[0,1])) & length(i0(:))==sL, i0c=find(i0); 
-elseif ~isempty(i0) & any(floor(i0)~=i0) | any(i0<1) | any(i0>sL)
+if all(ismember(i0,[0,1])) && length(i0(:))==sL, i0c=find(i0); 
+elseif ~isempty(i0) && any(floor(i0)~=i0) || any(i0<1) || any(i0>sL)
     error('logical mask or vector of column indices required')
-else, i0c = i0; end
+else i0c = i0; end
 
 %=======================================================================
 function c = sf_X0_2_c(X0,sX)  
@@ -726,7 +726,7 @@
    sL = spm_sp('size',sX,2);
 
    %- why the   "& size(X0,2) ~= sL"   !!!?
-   if isempty(c) & size(X0,2) ~= sL
+   if isempty(c) && size(X0,2) ~= sL
       c = zeros(sL,1);
    end
 
@@ -734,17 +734,4 @@
    c = spm_sp('xpx',sX);
 end
 
-%
 %- c = spm_sp('r',sc0,spm_sp('oxp',sX)); would also works.
-
-
-
-
-
-
-
-
-
-
-
-
diff --git a/spm_Tcdf.m b/spm_Tcdf.m
index 1d01b28..f90db00 100644
--- a/spm_Tcdf.m
+++ b/spm_Tcdf.m
@@ -5,12 +5,12 @@
 % x - T-variate (Student's t has range (-Inf,Inf)
 % v - degrees of freedom (v>0, non-integer d.f. accepted)
 % F - CDF of Student's t-distribution with v degrees of freedom at points x
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Tcdf implements the Cumulative Distribution of the Students t-distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF F(x) of the Student's t-distribution with v degrees of
 % freedom is the probability that a realisation of a t random variable
 % X has value less than x; F(x)=Pr{X<x} for X~G(h,c). Student's
@@ -21,12 +21,12 @@
 % v, rather it will compute for any df v>0.
 %
 % Variate relationships: (Evans et al., Ch37 & 7)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Student's t distribution with 1 degree of freedom is the Standard
 % Cauchy distribution, which has a simple closed form CDF.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF of the Student's t-distribution with v degrees of freedom
 % is related to the incomplete beta function by:
 %       Pr(|X|<x) = betainc(v/(v+x^2),v/2,1/2)
@@ -40,11 +40,11 @@
 % easily verified by substituting for v/(v+x^2) in the integral of the
 % incomplete beta function.
 %
-% MatLab's implementation of the incomplete beta function is used.
+% MATLAB's implementation of the incomplete beta function is used.
 %
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -57,42 +57,45 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1992-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Tcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Tcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(v)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))]     ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if all(xa) & any(diff(as(xa,:)))
-    error('non-scalar args must match in size'), end
+if all(xa) && any(diff(as(xa,:)))
+    error('non-scalar args must match in size');
+end
 
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 F = zeros(rs);
 
 %-Only defined for strictly positive v. Return NaN if undefined.
 md = ( ones(size(x))  &  v>0 );
-if any(~md(:)), F(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    F(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special case: f is 0.5 when x=0 (where betainc involves log of zero)
 F( md  &  x==0 ) = 0.5;
 
 %-Special case: Standard Cauchy distribution when v=1
-ml = ( md  &  v==1 ); if xa(1), mlx=ml; else, mlx=1; end
+ml = ( md  &  v==1 ); if xa(1), mlx=ml; else mlx=1; end
 F(ml) = 0.5 + atan(x(mlx))/pi;
 
 %-Compute where defined & not special cases
diff --git a/spm_Tpdf.m b/spm_Tpdf.m
index 07f4a0c..f8480b8 100644
--- a/spm_Tpdf.m
+++ b/spm_Tpdf.m
@@ -5,13 +5,13 @@
 % x - t-ordinates
 % v - degrees of freedom (v>0, non-integer d.f. accepted)
 % f - PDF of t-distribution with v degrees of freedom (df) at points t
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Tpdf implements the Probability Density Function of Students 
 % t-distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Student's t-distribution with v degrees of freedom is defined for
 % positive integer v and x in (-Inf,Inf), and has Probability Distribution
 % Function (PDF) f(x) given by: (See Evans et al., Ch37)
@@ -24,12 +24,12 @@
 % v, rather it will compute for any df v>0.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Direct computation using the beta function for
 %       sqrt(pi)*gamma(v/2) / gamma((v+1)/2)  =  beta(v/2,1/2)
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -42,36 +42,39 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1993-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Tpdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Tpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 ad = [ndims(x);ndims(v)];
 rd = max(ad);
-as = [  [size(x),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];    ];
+as = [[size(x),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];];
 rs = max(as);
 xa = prod(as,2)>1;
-if all(xa) & any(diff(as(xa,:)))
-    error('non-scalar args must match in size'), end
+if all(xa) && any(diff(as(xa,:)))
+    error('non-scalar args must match in size');
+end
 
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 f = zeros(rs);
 
 %-Only defined for v>0. Return NaN if undefined.
 md = ( ones(size(x))  &  v>0 );
-if any(~md(:)), f(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    f(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Compute where defined
 Q  = find( md );
diff --git a/spm_Welcome.fig b/spm_Welcome.fig
index 419e41c..dee85d2 100644
Binary files a/spm_Welcome.fig and b/spm_Welcome.fig differ
diff --git a/spm_Xcdf.m b/spm_Xcdf.m
index c74457e..a704892 100644
--- a/spm_Xcdf.m
+++ b/spm_Xcdf.m
@@ -14,7 +14,7 @@
 % degrees of freedom is less than x. F(x) = Pr{X<x} for X~\Chi^2(v)
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Chi-squared distribution with v degrees of freedom is defined for
 % positive integer v and x in [0,Inf). The Cumulative Distribution
 % Function (CDF) F(x) is the probability that a realisation of a
@@ -22,16 +22,16 @@
 % (See Evans et al., Ch8)
 %
 % Variate relationships: (Evans et al., Ch8 & Ch18)
-%-----------------------------------------------------------------------
-% The Chi-squared distribution with v degrees of freedom is equivalent
-% to the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
+%--------------------------------------------------------------------------
+% The Chi-squared distribution with v degrees of freedom is equivalent to
+% the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Using routine spm_Gcdf for Gamma distribution, with appropriate parameters.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -45,16 +45,16 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1992-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Xcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Xcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Check enough arguments
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 %-Computation
-%---------------------------------------------------------------------------
+%--------------------------------------------------------------------------
 F = spm_Gcdf(x,v/2,1/2);
diff --git a/spm_Xpdf.m b/spm_Xpdf.m
index c1e1c85..7eab9ba 100644
--- a/spm_Xpdf.m
+++ b/spm_Xpdf.m
@@ -11,7 +11,7 @@
 % Chi-squared distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Chi-squared distribution with v degrees of freedom is defined for
 % positive integer v and x in [0,Inf), and has Probability Distribution
 % Function (PDF) f(x) given by: (See Evans et al., Ch8)
@@ -21,16 +21,16 @@
 %           2^(v/2) * gamma(v/2)
 %
 % Variate relationships: (Evans et al., Ch8 & Ch18)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Chi-squared distribution with v degrees of freedom is equivalent
 % to the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Using routine spm_Gpdf for Gamma distribution, with appropriate parameters.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -44,16 +44,16 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1993-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_Xpdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_Xpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Check enough arguments
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 %-Computation
-%---------------------------------------------------------------------------
+%--------------------------------------------------------------------------
 f = spm_Gpdf(x,v/2,1/2);
diff --git a/spm_adjmean_fmri_ui.m b/spm_adjmean_fmri_ui.m
index 5a5ad56..a87beef 100644
--- a/spm_adjmean_fmri_ui.m
+++ b/spm_adjmean_fmri_ui.m
@@ -192,7 +192,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_adjmean_fmri_ui.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_adjmean_fmri_ui.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 
@@ -595,11 +595,11 @@
         'aPMap',    [],...
         'EXnames',  EXnames,...
         'iX',       iX      );
-if spm_matlab_version_chk('7') >= 0
-    save('SPMadj','-V6',...
+if spm_check_version('matlab','7') >= 0
+    save('SPMadj.mat','-V6',...
         'SPMid','Des','V','c','cNames','W','Fnames','HPFc','HPF','sHPF','rX','nrX','rGX','GX');
 else
-    save('SPMadj',...
+    save('SPMadj.mat',...
         'SPMid','Des','V','c','cNames','W','Fnames','HPFc','HPF','sHPF','rX','nrX','rGX','GX');
 end;
 
diff --git a/spm_adjmean_ui.m b/spm_adjmean_ui.m
index 93cfacb..972626b 100644
--- a/spm_adjmean_ui.m
+++ b/spm_adjmean_ui.m
@@ -178,7 +178,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_adjmean_ui.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_adjmean_ui.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 
@@ -641,10 +641,10 @@
         'aPMap',    aPMap,...
         'EXnames',  EXnames,...
         'iX',       iX      );
-if spm_matlab_version_chk('7') >= 0
-    save('SPMadj','-V6','SPMid','DesDef','Des','V','c','cNames','W','Fnames','rGX','GX');
+if spm_check_version('matlab','7') >= 0
+    save('SPMadj.mat','-V6','SPMid','DesDef','Des','V','c','cNames','W','Fnames','rGX','GX');
 else
-    save('SPMadj','SPMid','DesDef','Des','V','c','cNames','W','Fnames','rGX','GX');
+    save('SPMadj.mat','SPMid','DesDef','Des','V','c','cNames','W','Fnames','rGX','GX');
 end;
 
 
diff --git a/spm_affine_priors.m b/spm_affine_priors.m
new file mode 100644
index 0000000..a7b293a
--- /dev/null
+++ b/spm_affine_priors.m
@@ -0,0 +1,86 @@
+function [mu,isig] = spm_affine_priors(typ)
+% Distribution of the priors used in affine registration
+%
+% The parameters for this distribution were derived empirically from 227
+% scans, that were matched to the ICBM space.
+%_______________________________________________________________________
+% Copyright (C) 2003-2011 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_affine_priors.m 4155 2011-01-11 15:22:39Z guillaume $
+
+%% Values can be derived by...
+%sn = spm_select(Inf,'.*seg_inv_sn.mat$');
+%X  = zeros(size(sn,1),12);
+%for i=1:size(sn,1),
+%    p  = load(deblank(sn(i,:)));
+%    M  = p.VF(1).mat*p.Affine/p.VG(1).mat;
+%    J  = M(1:3,1:3);
+%    V  = sqrtm(J*J');
+%    R  = V\J;
+%    lV      =  logm(V);
+%    lR      = -logm(R);
+%    P       = zeros(12,1);
+%    P(1:3)  = M(1:3,4);
+%    P(4:6)  = lR([2 3 6]);
+%    P(7:12) = lV([1 2 3 5 6 9]);
+%    X(i,:)  = P';
+%end;
+%mu   = mean(X(:,7:12));
+%XR   = X(:,7:12) - repmat(mu,[size(X,1),1]);
+%isig = inv(XR'*XR/(size(X,1)-1))
+
+
+switch deblank(lower(typ))
+
+case 'mni', % For registering with MNI templates...
+    mu   = [0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
+    isig = 1e4 * [
+        0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
+       -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
+       -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
+       -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
+       -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
+       -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
+
+case 'imni', % For registering with MNI templates...
+    mu   = -[0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
+    isig = 1e4 * [
+        0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
+       -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
+       -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
+       -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
+       -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
+       -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
+
+case 'rigid', % Constrained to be almost rigid...
+    mu   = zeros(6,1);
+    isig = eye(6)*1e8; % spm_affreg used 1e9
+
+case 'subj', % For inter-subject registration...
+    mu   = zeros(6,1);
+    isig = 1e3 * [
+        0.8876    0.0784    0.0784   -0.1749    0.0784   -0.1749
+        0.0784    5.3894    0.2655    0.0784    0.2655    0.0784
+        0.0784    0.2655    5.3894    0.0784    0.2655    0.0784
+       -0.1749    0.0784    0.0784    0.8876    0.0784   -0.1749
+        0.0784    0.2655    0.2655    0.0784    5.3894    0.0784
+       -0.1749    0.0784    0.0784   -0.1749    0.0784    0.8876];
+
+case 'eastern', % For East Asian brains to MNI...
+    mu   = [0.0719   -0.0040   -0.0032    0.1416    0.0601    0.2578]';
+    isig = 1e4 * [
+        0.0757    0.0220   -0.0224   -0.0049    0.0304   -0.0327
+        0.0220    0.3125   -0.1555    0.0280   -0.0012   -0.0284
+       -0.0224   -0.1555    1.9727    0.0196   -0.0019    0.0122
+       -0.0049    0.0280    0.0196    0.0576   -0.0282   -0.0200
+        0.0304   -0.0012   -0.0019   -0.0282    0.2128   -0.0275
+       -0.0327   -0.0284    0.0122   -0.0200   -0.0275    0.0511];
+
+case 'none', % No regularisation...
+    mu   = zeros(6,1);
+    isig = zeros(6);
+
+otherwise
+    error(['"' typ '" not recognised as type of regularisation.']);
+end
diff --git a/spm_affreg.m b/spm_affreg.m
index 9186ec8..742e0c0 100644
--- a/spm_affreg.m
+++ b/spm_affreg.m
@@ -38,7 +38,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_affreg.m 1966 2008-07-29 15:40:11Z john $
+% $Id: spm_affreg.m 4152 2011-01-11 14:13:35Z volkmar $
 
 
 if nargin<5, scal = ones(length(VG),1); end;
@@ -315,7 +315,7 @@
     % and try again.  Repeat a few times.
     % ---------------------------------------------------------------
     ss = ss/n;
-    if iter>1, spm_chi2_plot('Set',ss); end;
+    if iter>1, spm_plot_convergence('Set',ss); end;
     if (pss-ss)/pss < 1e-6,
         est_smo = 1;
     end;
@@ -351,7 +351,7 @@
 if nargin<3, typ = 'subj'; end;
 if nargin<2, n   = 13;     end;
 
-[mu,isig] = priors(typ);
+[mu,isig] = spm_affine_priors(typ);
 ds  = 0.000001;
 d1  = zeros(n,1);
 d2  = zeros(n);
@@ -396,55 +396,6 @@
 return;
 %_______________________________________________________________________
 
-%_______________________________________________________________________
-function [mu,isig] = priors(typ)
-% The parameters for this distribution were derived empirically from 227
-% scans, that were matched to the ICBM space.
-%_______________________________________________________________________
-
-mu   = zeros(6,1);
-isig = zeros(6);
-switch deblank(lower(typ)),
-
-case 'mni', % For registering with MNI templates...
-    mu   = [0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
-    isig = 1e4 * [
-        0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
-       -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
-       -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
-       -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
-       -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
-       -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
-
-case 'rigid', % Constrained to be almost rigid...
-    mu   = zeros(6,1);
-    isig = eye(6)*1e9;
-
-case 'isochoric', % Volume preserving...
-    error('Not implemented');
-
-case 'isotropic', % Isotropic zoom in all directions...
-    error('Not implemented');
-
-case 'subj', % For inter-subject registration...
-    mu   = zeros(6,1);
-    isig = 1e3 * [
-        0.8876    0.0784    0.0784   -0.1749    0.0784   -0.1749
-        0.0784    5.3894    0.2655    0.0784    0.2655    0.0784
-        0.0784    0.2655    5.3894    0.0784    0.2655    0.0784
-       -0.1749    0.0784    0.0784    0.8876    0.0784   -0.1749
-        0.0784    0.2655    0.2655    0.0784    5.3894    0.0784
-       -0.1749    0.0784    0.0784   -0.1749    0.0784    0.8876];
-
-case 'none', % No regularisation...
-    mu   = zeros(6,1);
-    isig = zeros(6);
-
-otherwise,
-    error(['"' typ '" not recognised as type of regularisation.']);
-end;
-return;
-
 %_______________________________________________________________________
 function [y1,y2,y3]=coords(M,x1,x2,x3)
 % Affine transformation of a set of coordinates.
diff --git a/spm_argmax.m b/spm_argmax.m
new file mode 100644
index 0000000..cb71164
--- /dev/null
+++ b/spm_argmax.m
@@ -0,0 +1,97 @@
+function [p f] = spm_argmax(fun,varargin)
+% function minimisation using Guass-Newton
+% FORMAT [P f] = spm_argmax('fun',varargin,i)
+%
+% fun      - inline function f - fun(P,varargin)
+% varargin - function arguments
+% i        - argument to minimise: varargin{i}
+%
+% P   - optimised argument
+% f   - optimised value of fun(P)
+%
+%--------------------------------------------------------------------------
+% spm_argmax uses numerical derivatives and a and adaptive Gauss-Newton 
+% scheme: see also spm_argmin.
+%__________________________________________________________________________
+% Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_argmax.m 4060 2010-09-01 17:17:36Z karl $
+ 
+ 
+% arguments
+%--------------------------------------------------------------------------
+x     = varargin(1:end - 1);
+i     = varargin{end};
+
+% Gauss-Newton scheme
+%======================================================================
+v     = -8;
+for j = 1:64
+    
+    % time
+    %----------------------------------------------------------------------
+    tic
+    
+    % get derivatives of function w.r.t. argument
+    %----------------------------------------------------------------------
+    [dfdpp dfdp f] = spm_diff(fun,x{:},[i i]);
+    dfdp           = spm_vec(dfdp);
+    dfdpp          = spm_cat(dfdpp');
+    dfdpp          = -spm_sqrtm(dfdpp'*dfdpp);
+    
+    % report change in f
+    %----------------------------------------------------------------------
+    try
+        fprintf(' actual: %6.2e (%.2f sec)\n',full(f - P.f),toc)
+    catch
+        f0   = f;
+        p    = spm_vec(x{i});
+        P.f  = -Inf;
+    end
+    
+    % If f is decreasing accept current estimates decrease regularization
+    %----------------------------------------------------------------------
+    if f > P.f
+        P.x  = x;
+        P.f  = f;
+        v    = min(v + 1/8,4);
+        v    = max(v,-8);
+        str  = 'GN:(+)';
+ 
+    else
+        x    = P.x;
+        v    = min(v - 2, -4);
+        str  = 'EM:(-)';
+    end
+
+    % Gradient ascent
+    %======================================================================
+    dp   = spm_dx(dfdpp,dfdp,{v});
+    df   = dfdp'*dp;
+      
+    % check stability
+    %----------------------------------------------------------------------
+    while df > 1e4
+        v    = v - log(abs(df));
+        dp   = spm_dx(dfdpp,dfdp,{v});
+        df   = dfdp'*dp;
+    end
+    
+    % update
+    %---------------------------------------------------------------------- 
+    p    = p + dp;
+    x{i} = spm_unvec(p,x{i});
+    
+    % convergence
+    %----------------------------------------------------------------------
+    fprintf('%-6s: %i %6s %-6.2e %6s %-6.2e',str,j,'F:',full(P.f - f0),'dF predicted:',full(df))
+    if j > 2 && df < 1e-6
+        fprintf(' convergence\n')
+        break
+    end
+end
+
+% output
+%----------------------------------------------------------------------
+p = x{i};
diff --git a/spm_authors.m b/spm_authors.m
new file mode 100644
index 0000000..551f087
--- /dev/null
+++ b/spm_authors.m
@@ -0,0 +1,48 @@
+function [current, previous] = spm_authors
+% Return list of SPM coauthors
+% FORMAT [current, previous] = spm_authors
+% current  - cell array of coauthors of the current SPM release
+% previous - cell array of former SPM coauthors
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% SPM
+% $Id: spm_authors.m 4025 2010-07-29 11:10:15Z guillaume $
+
+
+current = {...
+'John Ashburner'
+'Gareth Barnes'
+'Chun-Chuan Chen'
+'Justin Chumbley'
+'Jean Daunizeau'
+'Guillaume Flandin'
+'Karl Friston'
+'Darren Gitelman'
+'Volkmar Glauche'
+'Lee Harrison'
+'Rik Henson'
+'Chloe Hutton'
+'Maria Joao Rosa'
+'Stefan Kiebel'
+'James Kilner'
+'Vladimir Litvak'
+'Rosalyn Moran'
+'Tom Nichols'
+'Robert Oostenveld'
+'Will Penny'
+'Christophe Phillips'
+'Ged Ridgway'
+'Klaas Enno Stephan'
+};
+
+previous = {...
+'Jesper Andersson'
+'Matthew Brett'
+'Christian Buechel'
+'Jon Heather'
+'Andrew Holmes'
+'Jeremie Mattout'
+'Jean-Baptiste Poline'
+'Keith Worsley'
+};
diff --git a/spm_axis.m b/spm_axis.m
new file mode 100644
index 0000000..fb32b9d
--- /dev/null
+++ b/spm_axis.m
@@ -0,0 +1,23 @@
+function varargout = spm_axis(varargin)
+% AXIS  Control axis scaling and appearance.
+
+if nargout
+    [varargout{1:nargout}] = axis(varargin{:});
+else
+    axis(varargin{:});
+end
+
+if nargin ==1 && strcmpi(varargin{1},'tight')
+    spm_axis(gca,'tight');
+elseif nargin == 2 && allAxes(varargin{1}) && strcmpi(varargin{2},'tight')
+    for i=1:numel(varargin{1})
+        lm = get(varargin{1}(i),'ylim');
+        set(varargin{1}(i),'ylim',lm + [-1 1]*diff(lm)/16);
+    end
+end
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+function result = allAxes(h)
+
+result = all(ishghandle(h)) && ...
+         length(findobj(h,'type','axes','-depth',0)) == length(h);
diff --git a/spm_bayes2_x2.m b/spm_bayes2_x2.m
new file mode 100644
index 0000000..e02b701
--- /dev/null
+++ b/spm_bayes2_x2.m
@@ -0,0 +1,94 @@
+function [xCon,SPM]= spm_bayes2_x2(SPM,XYZ,xCon,ic)
+% Compute and write Chi^2 image for 2nd level Bayes
+% FORMAT [xCon,SPM]= spm_bayes2_x2(SPM,XYZ,xCon,ic)
+%
+% SPM  - SPM data structure
+% XYZ  - voxel list
+% xCon - contrast info
+% ic   - contrast number
+%_______________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Will Penny 
+% $$
+        
+
+%- Compound Contrast
+%-----------------------------------------------------------------------
+c  = xCon(ic).c;
+kc = size(c,2);
+
+%-Get posterior beta's
+%-----------------------------------------------------------------------
+Nk = size(SPM.xX.X,2);
+
+for k=1:Nk,
+    beta(k,:) = spm_get_data(SPM.VCbeta(k),XYZ);
+end
+
+% Get noise hyperparameters
+Nh=length(SPM.PPM.l);
+for jj = 1:Nh,
+    hyper(jj).l   = spm_get_data(SPM.VHp(jj),XYZ);
+end
+ 
+%-Get posterior SD beta's
+%-----------------------------------------------------------------------
+Nk = size(SPM.xX.X,2);
+
+%-Loop over voxels
+%=======================================================================
+Nvoxels = size(XYZ,2);
+D       = repmat(NaN,reshape(SPM.xVol.DIM(1:3),1,[]));
+
+spm_progress_bar('Init',100,'Estimating posterior contrast covariance','');
+
+for v=1:Nvoxels,
+    
+    % Reconstruct approximation to voxel wise covariance matrix
+    Sigma_post   = SPM.PPM.Cby;
+    for jj = 1:Nh,
+        % Taylor approximation to posterior covariance
+        Sigma_post = Sigma_post + SPM.PPM.dC{jj}*(hyper(jj).l(v) - SPM.PPM.l(jj));
+    end
+    
+    % Get posterior covariance of contrast
+    V  =  c' * Sigma_post * c;
+   
+    % Get posterior mean of contrast
+    m = c'*beta(:,v);
+    
+    % Get Chi^2 value
+    d = m'*inv(V)*m;
+    
+    D(XYZ(1,v),XYZ(2,v),XYZ(3,v)) = d;
+    if rem(v,100)==0
+        % update progress bar every 100th voxel
+        spm_progress_bar('Set',100*v/Nvoxels);
+    end
+    
+end
+
+xCon(ic).eidf=rank(V);
+
+spm_progress_bar('Clear');   
+
+% Create handle
+%-----------------------------------------------------------------------
+Vhandle = struct(...
+    'fname',  sprintf('x2_%04d.img',ic),...
+    'dim',    SPM.xVol.DIM',...
+    'dt',     [spm_type('float32') spm_platform('bigend')],... 
+    'mat',    SPM.xVol.M,...
+    'pinfo',  [1,0,0]',...
+    'descrip',sprintf('Chi^2 stat for Bayes multivar con %d: %s',ic,xCon(ic).name));
+
+%-Write image
+%-----------------------------------------------------------------------
+Vhandle = spm_create_vol(Vhandle);
+Vhandle = spm_write_vol(Vhandle,D);
+
+xCon(ic).Vcon = Vhandle;
+
+fprintf('%s%30s\n',repmat(sprintf('\b'),1,30),...
+    sprintf('...written %s',spm_str_manip(Vhandle.fname,'t')));            %-#
diff --git a/spm_bireduce.m b/spm_bireduce.m
index 00c9482..db8686b 100644
--- a/spm_bireduce.m
+++ b/spm_bireduce.m
@@ -14,7 +14,7 @@
 %   M.u   - (m x 1) = u    = expansion point: defaults to u = 0;
 %
 % P   - model parameters
-% D   - delay matrix (default D = 1)
+% D   - delay operator df/dx -> D*df/dx (default D = 1)
 %
 % A Bilinear approximation is returned where the states are
 %
@@ -36,7 +36,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_bireduce.m 3696 2010-01-22 14:22:31Z karl $
+% $Id: spm_bireduce.m 4261 2011-03-24 16:39:42Z karl $
 
 
 % set up
@@ -114,7 +114,7 @@
     fung = fcnchk(M.g,'x','u','P','M');
 catch
     M.g  = inline('spm_vec(x)','x','u','P','M');
-    M.l  = M.n;
+    M.l  = n;
     fung = fcnchk(M.g,'x','u','P','M');
 end
 
diff --git a/spm_changepath.m b/spm_changepath.m
index 23c888b..2add519 100644
--- a/spm_changepath.m
+++ b/spm_changepath.m
@@ -2,10 +2,13 @@
 % Recursively replace all occurences of a text pattern in a MATLAB variable.
 % FORMAT S = spm_changepath(Sf, oldp, newp)
 %
-% Sf       - MATLAB variable to fix, or char array of MAT filenames
+% Sf       - MATLAB variable to fix, or char array of MAT filenames,
+%            or directory name (all found MAT files will be analysed)
 % oldp     - old string to replace
 % newp     - new string replacing oldp
 %
+% S        - updated MATLAB variable (only if Sf is one)
+%
 % If the pattern is found in a string, any occurence of an invalid file
 % separator is replaced to match that of the current system.
 %
@@ -15,51 +18,50 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_changepath.m 3277 2009-07-15 11:47:40Z guillaume $
+% $Id: spm_changepath.m 4078 2010-10-06 17:41:26Z guillaume $
 
 
+%-Input arguments
+%--------------------------------------------------------------------------
 if ~nargin
     Sf = spm_select(Inf,'mat','Select MAT files to fix');
 end
-if ischar(Sf)
-    S = cell(1,size(Sf,1));
-    for i=1:size(Sf,1)
-        try
-            S{i} = load(deblank(Sf(i,:)));
-        catch
-            error(sprintf('Cannot load %s.',deblank(Sf(i,:))));
-        end
-    end
-else
-    S = {Sf};
-end
 
 if nargin <= 1
     oldp = spm_input('Old pattern','+1','s');
 end
+
 if nargin <= 2
     newp = spm_input('New pattern','+1','s');
 end
 
-for i=1:numel(S)
-    S{i} = changepath(S{i},oldp,newp);
-    if ischar(Sf)
-        f   = deblank(Sf(i,:));
-        tmp = S{i};
-        [sts, msg] = movefile(f,[f '.old']);
-        if ~sts, error(msg); end
-        save(f ,'-struct','tmp');
+%-Replace pattern in given MAT-files
+%--------------------------------------------------------------------------
+if ischar(Sf)
+    if nargout
+        error('Output argument only valid for MATLAB variable input');
+    end
+    if exist(Sf,'dir')
+        Sf = spm_select('FPList',Sf,'^.*\.mat$'); % FPListRec for recursive
+    end
+    if isempty(Sf), Sf = {}; else Sf = cellstr(Sf); end
+    for i=1:numel(Sf)
+        f = Sf{i};
+        try
+            S = load(f);
+        catch
+            error(sprintf('Cannot load %s.',f));
+        end
+        tmp = changepath(S,oldp,newp);
+        if ~isequalwithequalnans(tmp,S)
+            fprintf('=> Fixing %s\n',f);
+            [sts, msg] = movefile(f,[f '.old']);
+            if ~sts, error(msg); end
+            save(f ,'-struct','tmp','-V6');
+        end
     end
-end
-
-if numel(S) == 1
-    S = S{1};
-end
-
-if nargout
-    varargout = {S};
 else
-    varargout = {};
+    varargout = { changepath(Sf,oldp,newp) };
 end
 
 %==========================================================================
diff --git a/spm_check_installation.m b/spm_check_installation.m
index 531584b..1a080a0 100644
--- a/spm_check_installation.m
+++ b/spm_check_installation.m
@@ -13,7 +13,7 @@ function spm_check_installation(action)
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_check_installation.m 3913 2010-06-02 15:25:23Z guillaume $
+% $Id: spm_check_installation.m 4288 2011-04-04 14:45:17Z guillaume $
 
 if isdeployed, return; end
 
@@ -39,9 +39,9 @@ function spm_check_installation(action)
 %-Minimal MATLAB version required
 %--------------------------------------------------------------------------
 try
-    v = spm_matlab_version_chk('7.1');
+    v = spm_check_version('matlab','7.1');
 catch
-    error('Where is spm_matlab_version_chk.m?');
+    error('Where is spm_check_version.m?');
 end
 if v < 0
     error([...
@@ -72,7 +72,7 @@ function spm_check_installation(action)
 
 %-Ensure that the original release - as well as the updates - was installed
 %--------------------------------------------------------------------------
-if ~exist(fullfile(d,'spm_Npdf.m'),'file') % File that should not have changed
+if ~exist(fullfile(d,'@nifti','create.m'),'file') % File that should not have changed
     if isunix
         error(sprintf([...
             'There appears to be some problem with the installation.\n'...
@@ -80,14 +80,12 @@ function spm_check_installation(action)
             'and the updates installed on top of this. Unix commands\n'...
             'to do this are:\n'...
             '   unzip spm8.zip\n'...
-            '   unzip -o spm8_updates_r????.zip -d spm8\n'...
-            'You may need help from your local network administrator.']));
+            '   unzip -o spm8_updates_r????.zip -d spm8']));
     else
         error(sprintf([...
             'There appears to be some problem with the installation.\n'...
             'The original spm8.zip distribution should be installed\n'...
-            'and the updates installed on top of this. If in doubt,\n'...
-            'consult your local network administrator.']));
+            'and the updates installed on top of this.']));
     end
 end
 
@@ -190,11 +188,11 @@ function spm_check_installation(action)
 %-Detect SPM toolboxes
 %--------------------------------------------------------------------------
 officials = {'Beamforming', 'DARTEL', 'dcm_meeg', 'DEM', 'FieldMap', ...
-    'HDW', 'MEEGtools', 'mixture', 'Neural_Models', 'Seg', 'spectral', ...
-    'SRender'};
+    'HDW', 'MEEGtools', 'mixture', 'Neural_Models', 'Seg', 'Shoot', ...
+    'spectral', 'SRender'};
 dd = dir(fullfile(SPMdir,'toolbox'));
 dd = {dd([dd.isdir]).name};
-dd(strmatch('.',dd)) = [];
+dd(strncmp('.',dd,1)) = [];
 dd = setdiff(dd,officials);
 fprintf('SPM toolboxes:');
 for i=1:length(dd)
@@ -491,7 +489,7 @@ function compare_versions(l,r)
             dispw = true;
             fprintf('File %s appears %d times in your MATLAB path:\n',f{i},numel(w));
             for j=1:numel(w)
-                if j==1 && isempty(strmatch(d,w{1}))
+                if j==1 && ~strncmp(d,w{1},length(d))
                     fprintf('  %s (SHADOWING)\n',w{1});
                 else
                     fprintf('  %s\n',w{j});
@@ -505,7 +503,7 @@ function compare_versions(l,r)
 %-Recursively extract subdirectories
 %--------------------------------------------------------------------------
 dd = {dd([dd.isdir]).name};
-dd(strmatch('.',dd)) = [];
+dd(strncmp('.',dd,1)) = [];
 for i=1:length(dd)
     l = generate_listing(d,fullfile(r,dd{i}),l);
 end
@@ -514,7 +512,7 @@ function compare_versions(l,r)
 % FUNCTION extract_info
 %==========================================================================
 function svnprops = extract_info(f)
-%Extract Subversion properties ($Id: spm_check_installation.m 3913 2010-06-02 15:25:23Z guillaume $ tag)
+%Extract Subversion properties (Id tag)
 
 svnprops = struct('file',f, 'id',[], 'date','', 'md5','');
 
@@ -533,8 +531,8 @@ function compare_versions(l,r)
     svnprops.id   = str2num(r(1).id);
     svnprops.date = r(1).date;
     [p,name,ext]  = fileparts(f);
-    if ~strmatch(svnprops.file,[name ext],'exact')
-        warning('SVN Id for %s does not match filename.',f);
+    if ~strcmp(svnprops.file,[name ext])
+        fprintf('\nSVN Id does not match filename for file:\n  %s\n',f);
     end
 end
 if numel(r) > 1
diff --git a/spm_check_orientations.m b/spm_check_orientations.m
index 05cded7..2df7f6a 100644
--- a/spm_check_orientations.m
+++ b/spm_check_orientations.m
@@ -1,18 +1,24 @@
-function spm_check_orientations(V)
+function sts = spm_check_orientations(V)
 % Check the dimensions and orientations of the images
-% FORMAT spm_check_orientations(V)
+% FORMAT sts = spm_check_orientations(V)
 %
-% V - a structure as returned by spm_vol.
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% V    - a struct array as returned by spm_vol
+% sts  - status (true means OK)
+%
+% When used without LHS, this function throws an error accordingly.
+%__________________________________________________________________________
+% Copyright (C) 2005-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_check_orientations.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_check_orientations.m 4199 2011-02-10 20:07:17Z guillaume $
+
+sts = true;
 
-if numel(V)<=1, return; end;
+if numel(V) <= 1 || isa(V,'gifti'), return; end
 
 dims = cat(1,V.dim);
-if any(any(diff(dims,1,1),1)),
+if any(any(diff(dims,1,1),1))
+    sts = false;
     fprintf('\n    ** The images do not all have the same dimensions. **\n');
     fprintf('The function assumes that a voxel in one image corresponds with\n');
     fprintf('the same  voxel in another.   This  is not a safe assumption if\n');
@@ -21,15 +27,16 @@ function spm_check_orientations(V)
     fprintf('normalisation bounding-boxes, voxel-sizes etc).\n');
     fprintf('Here are the dimensions of the image volumes.  This list can be\n');
     fprintf('used to determine which file(s) are causing the problem.\n\n');
-    for i=1:numel(V),
+    for i=1:numel(V)
         fprintf('[%d %d %d]  %s\n',V(i).dim, V(i).fname);
-    end;
+    end
     fprintf('\n');
-    error('The dimensions must be identical for this procedure.');
-end;
+    if ~nargout, error('The dimensions must be identical for this procedure.'); end
+end
 
 matx = reshape(cat(3,V.mat),[16,numel(V)]);
-if any(any(abs(diff(matx,1,2))>1e-4)),
+if any(any(abs(diff(matx,1,2))>1e-4))
+    sts = false;
     fprintf('\n** The images do not all have same orientation and/or voxel sizes. **\n');
     fprintf('The function assumes that a voxel in one image  corresponds exactly\n');
     fprintf('with  the same voxel in another.   This is not a safe assumption if\n');
@@ -39,11 +46,10 @@ function spm_check_orientations(V)
     fprintf('but not actually resliced them to be in voxel-wise alignment.\n');
     fprintf('Here are the orientation matrices of the image volumes.   This list\n');
     fprintf('can be used to determine which file(s) are causing the problem.\n\n');
-    for i=1:numel(V),
+    for i=1:numel(V)
         fprintf('[%g %g %g %g; %g %g %g %g; %g %g %g %g]  %s\n',...
                 V(i).mat(1:3,:)', V(i).fname);
-    end;
+    end
     fprintf('\n');
-    error('The orientations etc must be identical for this procedure.');
-end;
-
+    if ~nargout, error('The orientations etc must be identical for this procedure.'); end
+end
diff --git a/spm_check_registration.m b/spm_check_registration.m
index 8111f90..4556419 100644
--- a/spm_check_registration.m
+++ b/spm_check_registration.m
@@ -1,51 +1,41 @@
 function spm_check_registration(images)
 % A visual check of image registration quality.
 % FORMAT spm_check_registration
-% Orthogonal views of one or more images are displayed.  Clicking in
-% any image moves the centre of the orthogonal views.  Images are
+% Orthogonal views of one or more images are displayed. Clicking in
+% any image moves the centre of the orthogonal views. Images are
 % shown in orientations relative to that of the first selected image.
 % The first specified image is shown at the top-left, and the last at
-% the bottom right.  The fastest increment is in the left-to-right
+% the bottom right. The fastest increment is in the left-to-right
 % direction (the same as you are reading this).
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1997-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_check_registration.m 1185 2008-03-04 16:31:21Z volkmar $
+% $Id: spm_check_registration.m 4205 2011-02-21 15:39:08Z guillaume $
 
 
-if nargin==0,
-    images = spm_select([1 15],'image','Select images');
-    if size(images,1)<1, return; end;
-    spm_check_registration(images);
-elseif nargin==1,
-    fg = spm_figure('Findwin','Graphics');
-    if isempty(fg),
-        fg=spm_figure('Create','Graphics');
-        if isempty(fg),
-            error('Cant create graphics window');
-        end;
-    else
-        spm_figure('Clear','Graphics');
-    end;
-    if isstruct(images) && isfield(images,'data')
-        images = spm_vol(strvcat(images.data));
-    elseif ischar(images), images=spm_vol(images); end;
-    spm_orthviews('Reset');
-    mn = length(images);
-    n  = round(mn^0.4);
-    m  = ceil(mn/n);
-    w  = 1/n;
-    h  = 1/m;
-    ds = (w+h)*0.02;
-    for ij=1:mn,
-        i  = 1-h*(floor((ij-1)/n)+1);
-        j  = w*rem(ij-1,n);
-        handle(ij) = spm_orthviews('Image', images(ij),...
-            [j+ds/2 i+ds/2 w-ds h-ds]);
-        if ij==1, spm_orthviews('Space'); end;
-        spm_orthviews('AddContext',handle(ij));
-    end;
-else
-    error('Incorrect Usage');
-end;
+if ~nargin
+    [images, sts] = spm_select([1 15],'image','Select images');
+    if ~sts, return; end
+end
+
+if ischar(images), images = spm_vol(images); end
+
+spm_figure('GetWin','Graphics');
+spm_figure('Clear','Graphics');
+spm_orthviews('Reset');
+
+mn = length(images);
+n  = round(mn^0.4);
+m  = ceil(mn/n);
+w  = 1/n;
+h  = 1/m;
+ds = (w+h)*0.02;
+for ij=1:mn
+    i = 1-h*(floor((ij-1)/n)+1);
+    j = w*rem(ij-1,n);
+    handle = spm_orthviews('Image', images(ij),...
+        [j+ds/2 i+ds/2 w-ds h-ds]);
+    if ij==1, spm_orthviews('Space'); end
+    spm_orthviews('AddContext',handle);
+end
diff --git a/spm_check_version.m b/spm_check_version.m
new file mode 100644
index 0000000..4be4cff
--- /dev/null
+++ b/spm_check_version.m
@@ -0,0 +1,193 @@
+function varargout = spm_check_version(tbx,chk)
+% Check a version number against a Toolbox version
+%
+% FORMAT tbx = spm_check_version
+% tbx    - toolbox name {'matlab','octave',...}
+%
+% FORMAT v = spm_check_version(tbx)
+% tbx    - toolbox name {'matlab','octave','signal',...}
+%
+% v      - version number {string}
+%
+% FORMAT status = spm_check_version(tbx,chk)
+% tbx    - toolbox name {'matlab','octave','signal',...}
+% chk    - version number to be checked {string}
+%
+% status - outcome of the comparison:
+%          -1: Toolbox version is earlier than the user supplied version
+%           0: Toolbox and user versions are the same
+%           1: Toolbox version is later than the user supplied version
+%              Think of it this way, the sign of status is determined by
+%              MATLAB_TOOLBOX_VERSION - USER_VERSION (i.e., THE VERSION YOU
+%              INPUT).
+%__________________________________________________________________________
+%
+% This function checks if a user supplied version number is less than,
+% equal to or greater than the version number of specified toolbox. If no
+% toolbox is specified the function checks the version of MATLAB. User
+% defined toolboxes can be checked but the Contents.m file must conform
+% to the specification defined in ver.m
+%
+% This function assumes that the version number is really a text string
+% with fields major.minor.revision.build. Other formats are not supported.
+% Checking is done to the level specified by the input version. Thus an
+% input of '7' will be rated as the same version as 7.1, but 7.0 would be
+% rated as earlier than 7.1.
+%__________________________________________________________________________
+%
+% EXAMPLES:
+%
+% If the MATLAB version is 7.1.0.83, and the user supplied version is '7':
+% status = spm_check_version('matlab','7');
+% returns status == 0   : major revision numbers are the same.
+%
+% If the MATLAB version is 7.1.0.0, and the user supplied version is '7.1':
+% status = spm_check_version('matlab','7');
+% returns status == 0   : major and minor revision numbers are the same.
+%
+% If the MATLAB version is 7.1.0.83, and the user supplied version is '7.2':
+% status = spm_check_version('matlab','7.2');
+% returns status == -1   : major + minor revision is earlier for MATLAB.
+%
+% If the MATLAB version is 6.5.1, and the user supplied version is '6.5.0'.
+% status = spm_check_version('matlab','6.5.0');
+% returns status == 1     : major + minor + release revision is later
+%                           for MATLAB
+% The statement ( spm_check_version('matlab','6.5.0') > 0 ) is true for
+% all MATLAB Toolbox versions after 6.5.0.
+%__________________________________________________________________________
+%
+% See also VERSION, VER.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Darren Gitelman
+% $Id: spm_check_version.m 4145 2010-12-23 15:18:30Z guillaume $
+
+%-Detect software used
+%==========================================================================
+if ~nargin || isempty(tbx)
+    if exist('OCTAVE_VERSION')
+        tbx = 'octave';
+    else
+        tbx = 'matlab';
+    end
+    if ~nargin, varargout = {tbx}; return; end
+end
+
+
+%-Get the requested toolbox version
+%==========================================================================
+if strcmpi(tbx,'matlab')
+    % MATLAB is a special case. The ver command does not report the
+    % entire version string.
+    tbxVer = strtok(version);
+else
+    tbxStruct = ver(tbx);
+    if isempty(tbxStruct)
+        error('Cannot find given toolbox.');
+    elseif numel(tbxStruct) > 1
+        error('Too many toolboxes found for given toolbox name.')
+    end
+    tbxVer = tbxStruct.Version;
+end
+
+if nargin == 1, varargout = {tbxVer}; return; end
+
+
+%-Parse user supplied version number
+%==========================================================================
+
+if strcmpi(tbx,'matlab') && strcmpi(spm_check_version,'octave')
+    varargout = {1}; % hack
+    return;
+end
+
+% If a number is supplied then convert to text
+%--------------------------------------------------------------------------
+if isnumeric(chk)
+    chk = num2str(chk);
+end
+
+% If too many fields in input then error
+%--------------------------------------------------------------------------
+if numel(strfind(chk,'.')) > 3
+    error(['Input string has too many fields. ',...
+        'Only major.minor.release.build fields are supported.']);
+end
+
+% parse the input string into a cell array of strings
+%--------------------------------------------------------------------------
+vCHK = textscan(chk,'%s %s %s %s','delimiter','.');
+
+% find empty cells. These will be set to 0 in both cell arrays so that the
+% same revision fields are compared
+%--------------------------------------------------------------------------
+emptyCHK = cellfun('isempty',vCHK);
+
+% parse the requested toolbox version string into a cell array of strings.
+%--------------------------------------------------------------------------
+vTBX = textscan(tbxVer,'%s %s %s %s','delimiter','.');
+
+% find empty cells. These will be removed from the cell array so that the
+% same revision fields are compared
+%--------------------------------------------------------------------------
+emptyTBX = cellfun('isempty',vTBX);
+
+% combine empty cell indices
+%--------------------------------------------------------------------------
+emptyCells = find(emptyCHK | emptyTBX);
+
+% and remove them.
+%--------------------------------------------------------------------------
+vCHK(emptyCells) = [];
+vTBX(emptyCells) = [];
+
+% final version fields to compare are converted to numbers
+%--------------------------------------------------------------------------
+vCHKMat = str2num(strvcat([vCHK{:}]))';
+vTBXMat = str2num(strvcat([vTBX{:}]))';
+
+
+%-Compare versions
+%==========================================================================
+% This array will be used to decide which version is the later one. The
+% differences between versions are computed on each element of the version
+% number: major.minor.release.build and then multiplied by the base2array
+% and summed. This is equivalent to logically combining the base 2 place
+% values and ensures that the fields are evaluated by their significance.
+% The sign of the summed values gives the result. Positive values mean the
+% toolbox version is later, negative values mean the user supplied version,
+% is later, and 0 means they are equal.
+%--------------------------------------------------------------------------
+base2array             = [8 4 2 1];
+base2array(emptyCells) = [];
+
+% take the difference between version elements.
+%--------------------------------------------------------------------------
+vDiff = vTBXMat - vCHKMat;
+
+% get the sign of the differences
+%--------------------------------------------------------------------------
+signDiff = sign(vDiff);
+
+% multiply by the base2array
+%--------------------------------------------------------------------------
+vVal = signDiff.*base2array;
+
+% If the sign of the sum is positive then toolbox version is later; if 0
+% they are the same, and if negative the user supplied version is later.
+% Think of it this way: MATLAB_TOOLBOX_VERSION - USER_VERSION.
+% Remember the comparison is made with respect to the FIELDS SUPPLED IN THE
+% INPUT
+%--------------------------------------------------------------------------
+switch sign(sum(vVal))
+    case -1     % MATLB_TOOLBOX_VERSION is EARLIER than USER_VERSION
+        status = -1;
+    case 0      % MATLB_TOOLBOX_VERSION = EQUALS = USER_VERSION
+        status =  0;
+    case 1      % MATLB_TOOLBOX_VERSION is LATER than USER_VERSION
+        status =  1;
+end
+
+varargout = {status};
diff --git a/spm_chi2_plot.m b/spm_chi2_plot.m
index 424eef2..fc62521 100644
--- a/spm_chi2_plot.m
+++ b/spm_chi2_plot.m
@@ -1,88 +1,28 @@
-function data = spm_chi2_plot(action,arg1,arg2,arg3)
-% Display a plot showing convergence of an optimization routine.
-% FORMAT spm_chi2_plot('Init',title,xlabel,ylabel)
-% Initialises the plot in the 'Interactive' window.
+function spm_chi2_plot(action,varargin)
+% Display a plot showing convergence of an optimisation routine.
+% FORMAT spm_chi2_plot('Init',title,ylabel,xlabel)
+% Initialise the plot in the 'Interactive' window.
 %
 % FORMAT spm_chi2_plot('Set',value)
-% Updates the plot.
+% Update the plot.
 %
 % FORMAT spm_chi2_plot('Clear')
-% Clears the 'Interactive' window.
-%_______________________________________________________________________
+% Clear the 'Interactive' window.
+%__________________________________________________________________________
+%
+% This function is deprecated, use SPM_PLOT_CONVERGENCE instead.
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_chi2_plot.m 1143 2008-02-07 19:33:33Z spm $
-
-
-global pb_pointer pb_name ax
-%-----------------------------------------------------------------------
-if (nargin == 0)
-    spm_chi2_plot('Init');
-else
-    % initialize
-    %---------------------------------------------------------------
-    if (strcmpi(action,'init'))
-        if (nargin<4)
-            arg3 = 'Iteration #';
-            if (nargin<3)
-                arg2 = 'Chi-squared';
-                if (nargin<2)
-                    arg1 = 'Optimizing';
-                end
-            end
-        end
-        fg = spm_figure('FindWin','Interactive');
-        if ~isempty(fg)
-            pb_pointer = get(fg,'Pointer');
-            pb_name    = get(fg,'Name');
-            spm_chi2_plot('Clear');
-            set(fg,'Pointer','watch');
-            set(fg,'Name',pb_name);
-            ax = axes('Position', [0.15 0.1 0.8 0.75],...
-                'Box', 'on','Parent',fg);
-            lab = get(ax,'Xlabel');
-            set(lab,'string',arg3,'FontSize',10);
-            lab = get(ax,'Ylabel');
-            set(lab,'string',arg2,'FontSize',10);
-            lab = get(ax,'Title');
-            set(lab,'string',arg1);
-            line('Xdata',[], 'Ydata',[],...
-                'LineWidth',2,'Tag','Chi2Plot','Parent',ax);
-            drawnow;
-        end
+% $Id: spm_chi2_plot.m 4148 2011-01-04 16:49:23Z guillaume $
 
-    % reset
-    %---------------------------------------------------------------
-    elseif (strcmpi(action,'set'))
-        if (nargin<2)
-            arg1 = 0;
-        end
-        F = spm_figure('FindWin','Interactive');
-        br = findobj(F,'Tag','Chi2Plot');
-        if (~isempty(br))
-            xd = get(br,'Xdata');
-            yd = [get(br,'Ydata') arg1];
-            xd = [xd (length(xd)+1)];
-            set(br,'Ydata',yd,'Xdata',xd);
-            drawnow;
-        end
-
-    % clear
-    %---------------------------------------------------------------
-    elseif (strcmpi(action,'clear'))
-        fg = spm_figure('FindWin','Interactive');
-        spm_figure('Clear',fg);
-        set(fg,'Pointer',pb_pointer);
-        set(fg,'Name',pb_name);
-        drawnow;
-    elseif (strcmpi(action,'data'))
-        F = spm_figure('FindWin','Interactive');
-        br = findobj(F,'Tag','Chi2Plot');
-        if (~isempty(br))
-            data = get(br,'Ydata');
-        else
-            data = [];
-        end
-    end
+persistent runonce
+if isempty(runonce)
+    warning(['spm_chi2_plot is deprecated. ',...
+        'Use spm_plot_convergence instead.']);
+    runonce = 1;
 end
+
+if ~nargin, action = 'Init'; end
+spm_plot_convergence(action,varargin{:});
diff --git a/spm_cond_units.m b/spm_cond_units.m
index 2473681..2f0677d 100644
--- a/spm_cond_units.m
+++ b/spm_cond_units.m
@@ -7,15 +7,34 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_cond_units.m 3205 2009-06-16 10:15:00Z vladimir $
+% $Id: spm_cond_units.m 4052 2010-08-27 19:22:44Z karl $
  
 % default n = 3
 %--------------------------------------------------------------------------
 try, n; catch, n = 3; end
 
-% rescale
-%--------------------------------------------------------------------------
-d           = spm_vec(y);
-scalefactor = norm(d(~isnan(d)),1);
-scalefactor = (10^n)^-round(log10(scalefactor)/n);
-y           = spm_unvec(d*scalefactor,y);
+switch lower(n)
+    
+    case{'csd'}
+        
+        % normalise to total power
+        %------------------------------------------------------------------
+        g     = 0;
+        for i = 1:length(y)
+            csd = y{i};
+            for i = 1:size(csd,2);
+                g = g + sum(csd(:,i,i));
+            end
+        end
+        scalefactor = (length(y)*size(csd,1)*size(csd,2))/g;
+        y           = spm_unvec(spm_vec(y)*scalefactor,y);
+        
+    otherwise
+        
+        % rescale
+        %------------------------------------------------------------------
+        d           = spm_vec(y);
+        scalefactor = norm(d(~isnan(d)),1);
+        scalefactor = (10^n)^-round(log10(scalefactor)/n);
+        y           = spm_unvec(d*scalefactor,y);
+end
diff --git a/spm_conman.m b/spm_conman.m
index d07ea81..09f4549 100644
--- a/spm_conman.m
+++ b/spm_conman.m
@@ -244,7 +244,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_conman.m 3933 2010-06-17 14:19:08Z guillaume $
+% $Id: spm_conman.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 %==========================================================================
@@ -492,10 +492,10 @@
     
     %-Save SPM.mat only if SPM structure as changed
     if ~isequal(tmpSPM,SPM)
-        if spm_matlab_version_chk('7') >=0
-            save('SPM', 'SPM', '-V6');
+        if spm_check_version('matlab','7') >=0
+            save('SPM.mat', 'SPM', '-V6');
         else
-            save('SPM', 'SPM');
+            save('SPM.mat', 'SPM');
         end
     end
 
diff --git a/spm_contrasts.m b/spm_contrasts.m
index 20a71fd..cacb9f1 100644
--- a/spm_contrasts.m
+++ b/spm_contrasts.m
@@ -8,7 +8,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes, Karl Friston & Jean-Baptiste Poline
-% $Id: spm_contrasts.m 3995 2010-07-13 17:19:49Z karl $
+% $Id: spm_contrasts.m 4185 2011-02-01 18:46:18Z guillaume $
 
 % Temporary SPM variable to check for any changes to SPM. We want to avoid
 % always having to save SPM.mat unless it has changed, because this is
@@ -86,8 +86,13 @@
                     fprintf('\t%-32s: %30s',sprintf('X2 image %2d',ic),...
                         '...computing');                                %-#
                     
-                    xCon = spm_vb_x2(SPM,XYZ,xCon,ic);
-                    
+                    if isfield(SPM.PPM,'VB')
+                        % First level Bayes
+                        xCon = spm_vb_x2(SPM,XYZ,xCon,ic);
+                    else
+                        % Second level Bayes
+                        xCon = spm_bayes2_x2(SPM,XYZ,xCon,ic);
+                    end
                 else
                     %-Implement contrast as sum of scaled beta images
                     %------------------------------------------------------
@@ -175,7 +180,7 @@
                 l     = spm_get_data(VHp,XYZ);       % get hyperparamters
                 Vc    = xCon(ic).c'*SPM.xX.Bcov*xCon(ic).c;
                 SE    = sqrt(l*Vc);                  % and standard error
-                Z     = cB./(SE + exp(-8)*max(SE));
+                Z     = cB./SE;
                 str   = sprintf('[%.1f]',SPM.xX.erdf);
                 
                 
@@ -275,9 +280,9 @@
 % Check if SPM has changed. Save only if it has.
 %--------------------------------------------------------------------------
 if ~isequal(tmpSPM,SPM)
-    if spm_matlab_version_chk('7') >=0
-        save('SPM', 'SPM', '-V6');
+    if spm_check_version('matlab','7') >=0
+        save('SPM.mat', 'SPM', '-V6');
     else
-        save('SPM', 'SPM');
+        save('SPM.mat', 'SPM');
     end
 end
diff --git a/spm_conv.m b/spm_conv.m
index 4a6b656..0200fdb 100644
--- a/spm_conv.m
+++ b/spm_conv.m
@@ -16,7 +16,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_conv.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_conv.m 4195 2011-02-05 18:39:13Z ged $
 
 
 % assume isomorphic smoothing
@@ -35,11 +35,11 @@
 
 % kernels
 %---------------------------------------------------------------------------
-Ex    = min([ceil(3*sx) lx]);
+Ex    = min([round(6*sx) lx]);
 x     = [-Ex:Ex];
 kx    = exp(-x.^2/(2*sx^2));
 kx    = kx/sum(kx);
-Ey    = min([ceil(3*sy) ly]);
+Ey    = min([round(6*sy) ly]);
 y     = [-Ey:Ey];
 ky    = exp(-y.^2/(2*sy^2));
 ky    = ky/sum(ky);
diff --git a/spm_coreg.m b/spm_coreg.m
index 70aacce..48d3252 100644
--- a/spm_coreg.m
+++ b/spm_coreg.m
@@ -1,26 +1,35 @@
 function x = spm_coreg(varargin)
 % Between modality coregistration using information theory
-% FORMAT x = spm_coreg(VG,VF,params)
-% VG - handle for first image (see spm_vol).
-% VF - handle for second image.
-% x - the parameters describing the rigid body rotation.
-%     such that a mapping from voxels in G to voxels in F
-%     is attained by:  VF.mat\spm_matrix(x(:)')*VG.mat
+% FORMAT x = spm_coreg(VG,VF,flags)
+% VG    - handle for reference image (see spm_vol).
+% VF    - handle for source (moved) image.
 % flags - a structure containing the following elements:
 %          sep      - optimisation sampling steps (mm)
 %                     default: [4 2]
 %          params   - starting estimates (6 elements)
 %                     default: [0 0 0  0 0 0]
 %          cost_fun - cost function string:
-%                      'mi'  - Mutual Information
-%                      'nmi' - Normalised Mutual Information
-%                      'ecc' - Entropy Correlation Coefficient
-%                      'ncc' - Normalised Cross Correlation
-%                      default: 'nmi'
+%                       'mi'  - Mutual Information
+%                       'nmi' - Normalised Mutual Information
+%                       'ecc' - Entropy Correlation Coefficient
+%                       'ncc' - Normalised Cross Correlation
+%                     default: 'nmi'
 %          tol      - tolerences for accuracy of each param
 %                     default: [0.02 0.02 0.02 0.001 0.001 0.001]
 %          fwhm     - smoothing to apply to 256x256 joint histogram
 %                     default: [7 7]
+%          graphics - display coregistration outputs
+%                     default: ~spm('CmdLine')
+%
+% x     - the parameters describing the rigid body rotation, such that a
+%         mapping from voxels in G to voxels in F is attained by:
+%         VF.mat\spm_matrix(x(:)')*VG.mat
+%
+% At the end, the voxel-to-voxel affine transformation matrix is
+% displayed, along with the histograms for the images in the original
+% orientations, and the final orientations. The registered images are
+% displayed at the bottom.
+%__________________________________________________________________________
 %
 % The registration method used here is based on the work described in:
 % A Collignon, F Maes, D Delaere, D Vandermeulen, P Suetens & G Marchal
@@ -32,13 +41,20 @@
 % The original interpolation method described in this paper has been
 % changed in order to give a smoother cost function.  The images are
 % also smoothed slightly, as is the histogram.  This is all in order to
-% make the cost function as smooth as possible, to give faster
-% convergence and less chance of local minima.
-%
+% make the cost function as smooth as possible, to give faster convergence
+% and less chance of local minima.
+%__________________________________________________________________________
+% Copyright (C) 1994-2011 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_coreg.m 4156 2011-01-11 19:03:31Z guillaume $
+
+%==========================================================================
 % References
-% ==========
+%==========================================================================
+%
 % Mutual Information
-% ------------------
+% -------------------------------------------------------------------------
 % Collignon, Maes, Delaere, Vandermeulen, Suetens & Marchal (1995).
 % "Automated multi-modality image registration based on information theory".
 % In Bizais, Barillot & Di Paola, editors, Proc. Information Processing
@@ -50,103 +66,97 @@
 % Medical Image Analysis, 1(1):35-51, 1996. 
 %
 % Entropy Correlation Coefficient
-% -------------------------------
-% F Maes, A Collignon, D Vandermeulen, G Marchal & P Suetens (1997).
+% -------------------------------------------------------------------------
+% Maes, Collignon, Vandermeulen, Marchal & Suetens (1997).
 % "Multimodality image registration by maximisation of mutual
 % information". IEEE Transactions on Medical Imaging 16(2):187-198
 %
 % Normalised Mutual Information
-% -----------------------------
-% Studholme,  Hill & Hawkes (1998).
+% -------------------------------------------------------------------------
+% Studholme, Hill & Hawkes (1998).
 % "A normalized entropy measure of 3-D medical image alignment".
 % in Proc. Medical Imaging 1998, vol. 3338, San Diego, CA, pp. 132-143.             
 %
 % Optimisation
-% ------------
+% -------------------------------------------------------------------------
 % Press, Teukolsky, Vetterling & Flannery (1992).
 % "Numerical Recipes in C (Second Edition)".
 % Published by Cambridge.
-%
-% At the end, the voxel-to-voxel affine transformation matrix is
-% displayed, along with the histograms for the images in the original
-% orientations, and the final orientations.  The registered images are
-% displayed at the bottom.
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%==========================================================================
 
-% John Ashburner
-% $Id: spm_coreg.m 1143 2008-02-07 19:33:33Z spm $
-
-
-if nargin>=4,
+if nargin >= 4
     x = optfun(varargin{:});
     return;
-end;
+end
 
-def_flags = struct('sep',[4 2],'params',[0 0 0  0 0 0], 'cost_fun','nmi','fwhm',[7 7],...
-    'tol',[0.02 0.02 0.02 0.001 0.001 0.001 0.01 0.01 0.01 0.001 0.001 0.001],'graphics',1);
-if nargin < 3,
+def_flags          = spm_get_defaults('coreg.estimate');
+def_flags.params   = [0 0 0  0 0 0];
+def_flags.graphics = ~spm('CmdLine');
+if nargin < 3
     flags = def_flags;
 else
     flags = varargin{3};
     fnms  = fieldnames(def_flags);
-    for i=1:length(fnms),
-        if ~isfield(flags,fnms{i}), flags.(fnms{i}) = def_flags.(fnms{i}); end;
-    end;
-end;
-%disp(flags)
+    for i=1:length(fnms)
+        if ~isfield(flags,fnms{i})
+            flags.(fnms{i}) = def_flags.(fnms{i});
+        end
+    end
+end
 
-if nargin < 1,
+if nargin < 1
     VG = spm_vol(spm_select(1,'image','Select reference image'));
 else
     VG = varargin{1};
-    if ischar(VG), VG = spm_vol(VG); end;
-end;
-if nargin < 2,
+    if ischar(VG), VG = spm_vol(VG); end
+end
+if nargin < 2
     VF = spm_vol(spm_select(Inf,'image','Select moved image(s)'));
 else
     VF = varargin{2};
-    if ischar(VF) || iscellstr(VF), VF = spm_vol(strvcat(VF)); end;
-end;
+    if ischar(VF) || iscellstr(VF), VF = spm_vol(char(VF)); end;
+end
 
-if ~isfield(VG, 'uint8'),
+if ~isfield(VG, 'uint8')
     VG.uint8 = loaduint8(VG);
     vxg      = sqrt(sum(VG.mat(1:3,1:3).^2));
     fwhmg    = sqrt(max([1 1 1]*flags.sep(end)^2 - vxg.^2, [0 0 0]))./vxg;
     VG       = smooth_uint8(VG,fwhmg); % Note side effects
-end;
+end
 
 sc = flags.tol(:)'; % Required accuracy
 sc = sc(1:length(flags.params));
 xi = diag(sc*20);
+x = zeros(numel(VF),numel(flags.params));
 
-for k=1:numel(VF),
+for k=1:numel(VF)
     VFk = VF(k);
-    if ~isfield(VFk, 'uint8'),
+    if ~isfield(VFk, 'uint8')
         VFk.uint8 = loaduint8(VFk);
         vxf       = sqrt(sum(VFk.mat(1:3,1:3).^2));
         fwhmf     = sqrt(max([1 1 1]*flags.sep(end)^2 - vxf.^2, [0 0 0]))./vxf;
         VFk       = smooth_uint8(VFk,fwhmf); % Note side effects
-    end;
+    end
 
     xk  = flags.params(:);
-    for samp=flags.sep(:)',
+    for samp=flags.sep(:)'
         xk     = spm_powell(xk(:), xi,sc,mfilename,VG,VFk,samp,flags.cost_fun,flags.fwhm);
         x(k,:) = xk(:)';
-    end;
-    if flags.graphics,
+    end
+    if flags.graphics
         display_results(VG(1),VFk(1),xk(:)',flags);
-    end;
-end;
-return;
-%_______________________________________________________________________
+    end
+end
 
-%_______________________________________________________________________
+
+%==========================================================================
+% function o = optfun(x,VG,VF,s,cf,fwhm)
+%==========================================================================
 function o = optfun(x,VG,VF,s,cf,fwhm)
 % The function that is minimised.
-if nargin<6, fwhm = [7 7];   end;
-if nargin<5, cf   = 'mi';    end;
-if nargin<4, s    = [1 1 1]; end;
+if nargin<6, fwhm = [7 7];   end
+if nargin<5, cf   = 'mi';    end
+if nargin<4, s    = [1 1 1]; end
 
 % Voxel sizes
 vxg = sqrt(sum(VG.mat(1:3,1:3).^2));sg = s./vxg;
@@ -167,12 +177,12 @@
 s2 = sum(H,2);
 
 switch lower(cf)
-    case 'mi',
+    case 'mi'
         % Mutual Information:
         H   = H.*log2(H./(s2*s1));
         mi  = sum(H(:));
         o   = -mi;
-    case 'ecc',
+    case 'ecc'
         % Entropy Correlation Coefficient of:
         % Maes, Collignon, Vandermeulen, Marchal & Suetens (1997).
         % "Multimodality image registration by maximisation of mutual
@@ -181,14 +191,14 @@
         mi  = sum(H(:));
         ecc = -2*mi/(sum(s1.*log2(s1))+sum(s2.*log2(s2)));
         o   = -ecc;
-    case 'nmi',
+    case 'nmi'
         % Normalised Mutual Information of:
         % Studholme,  Hill & Hawkes (1998).
         % "A normalized entropy measure of 3-D medical image alignment".
         % in Proc. Medical Imaging 1998, vol. 3338, San Diego, CA, pp. 132-143.
         nmi = (sum(s1.*log2(s1))+sum(s2.*log2(s2)))/sum(sum(H.*log2(H)));
         o   = -nmi;
-    case 'ncc',
+    case 'ncc'
         % Normalised Cross Correlation
         i     = 1:size(H,1);
         j     = 1:size(H,2);
@@ -199,17 +209,17 @@
         [i,j] = ndgrid(i-m1,j-m2);
         ncc   = sum(sum(H.*i.*j))/(sig1*sig2);
         o     = -ncc;
-    otherwise,
+    otherwise
         error('Invalid cost function specified');
-end;
+end
 
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+%==========================================================================
+% function udat = loaduint8(V)
+%==========================================================================
 function udat = loaduint8(V)
 % Load data from file indicated by V into an array of unsigned bytes.
-if size(V.pinfo,2)==1 && V.pinfo(1) == 2,
+if size(V.pinfo,2)==1 && V.pinfo(1) == 2
     mx = 255*V.pinfo(1) + V.pinfo(2);
     mn = V.pinfo(2);
 else
@@ -217,13 +227,13 @@
         ['Computing max/min of ' spm_str_manip(V.fname,'t')],...
         'Planes complete');
     mx = -Inf; mn =  Inf;
-    for p=1:V.dim(3),
+    for p=1:V.dim(3)
         img = spm_slice_vol(V,spm_matrix([0 0 p]),V.dim(1:2),1);
         mx  = max([max(img(:))+paccuracy(V,p) mx]);
         mn  = min([min(img(:)) mn]);
         spm_progress_bar('Set',p);
-    end;
-end;
+    end
+end
 
 % Another pass to find a maximum that allows a few hot-spots in the data.
 spm_progress_bar('Init',V.dim(3),...
@@ -231,57 +241,57 @@
         'Planes complete');
 nh = 2048;
 h  = zeros(nh,1);
-for p=1:V.dim(3),
+for p=1:V.dim(3)
     img = spm_slice_vol(V,spm_matrix([0 0 p]),V.dim(1:2),1);
     img = img(isfinite(img));
     img = round((img+((mx-mn)/(nh-1)-mn))*((nh-1)/(mx-mn)));
-    if spm_matlab_version_chk('7.0')>=0,
-        h = h + accumarray(img,1,[nh 1]);
-    else
-        h = h + full(sparse(img,1,1,nh,1));
-    end
+    h   = h + accumarray(img,1,[nh 1]);
     spm_progress_bar('Set',p);
-end;
+end
 tmp = [find(cumsum(h)/sum(h)>0.9999); nh];
 mx  = (mn*nh-mx+tmp(1)*(mx-mn))/(nh-1);
 
+% Load data from file indicated by V into an array of unsigned bytes.
 spm_progress_bar('Init',V.dim(3),...
     ['Loading ' spm_str_manip(V.fname,'t')],...
     'Planes loaded');
-
-%udat = zeros(V.dim,'uint8'); Needs MATLAB 7 onwards
-udat = uint8(0);
-udat(V.dim(1),V.dim(2),V.dim(3)) = 0;
-
-rand('state',100);
-for p=1:V.dim(3),
+udat = zeros(V.dim,'uint8');
+st = rand('state'); % st = rng;
+rand('state',100); % rng(100,'v5uniform'); % rng('defaults');
+for p=1:V.dim(3)
     img = spm_slice_vol(V,spm_matrix([0 0 p]),V.dim(1:2),1);
     acc = paccuracy(V,p);
-    if acc==0,
+    if acc==0
         udat(:,:,p) = uint8(max(min(round((img-mn)*(255/(mx-mn))),255),0));
     else
         % Add random numbers before rounding to reduce aliasing artifact
         r = rand(size(img))*acc;
         udat(:,:,p) = uint8(max(min(round((img+r-mn)*(255/(mx-mn))),255),0));
-    end;
+    end
     spm_progress_bar('Set',p);
-end;
+end
 spm_progress_bar('Clear');
-return;
+rand('state',st); % rng(st);
 
+
+%==========================================================================
+% function acc = paccuracy(V,p)
+%==========================================================================
 function acc = paccuracy(V,p)
-if ~spm_type(V.dt(1),'intt'),
+if ~spm_type(V.dt(1),'intt')
     acc = 0;
 else
-    if size(V.pinfo,2)==1,
+    if size(V.pinfo,2)==1
         acc = abs(V.pinfo(1,1));
     else
         acc = abs(V.pinfo(1,p));
-    end;
-end;
-%_______________________________________________________________________
+    end
+end
 
-%_______________________________________________________________________
+
+%==========================================================================
+% function V = smooth_uint8(V,fwhm)
+%==========================================================================
 function V = smooth_uint8(V,fwhm)
 % Convolve the volume in memory (fwhm in voxels).
 lim = ceil(2*fwhm);
@@ -292,10 +302,11 @@
 j  = (length(y) - 1)/2;
 k  = (length(z) - 1)/2;
 spm_conv_vol(V.uint8,V.uint8,x,y,z,-[i j k]);
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+
+%==========================================================================
+% function krn = smoothing_kernel(fwhm,x)
+%==========================================================================
 function krn = smoothing_kernel(fwhm,x)
 
 % Variance from FWHM
@@ -326,10 +337,11 @@
       +w3*(exp(w2*(x+1).^2)     + exp(w2*(x-1).^2)     - 2*exp(w2*x.^2));
 
 krn(krn<0) = 0;
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+
+%==========================================================================
+% function display_results(VG,VF,x,flags)
+%==========================================================================
 function display_results(VG,VF,x,flags)
 fig = spm_figure('FindWin','Graphics');
 if isempty(fig), return; end;
@@ -343,10 +355,10 @@ function display_results(VG,VF,x,flags)
     case 'nmi', txt = 'Normalised Mutual Information Coregistration';
     case 'ncc', txt = 'Normalised Cross Correlation';
     otherwise, error('Invalid cost function specified');
-end;
+end
 
 % Display text
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 ax = axes('Position',[0.1 0.8 0.8 0.15],'Visible','off','Parent',fig);
 text(0.5,0.7, txt,'FontSize',16,...
     'FontWeight','Bold','HorizontalAlignment','center','Parent',ax);
@@ -357,7 +369,7 @@ function display_results(VG,VF,x,flags)
 text(0,0.1, sprintf('Z1 = %0.3f*X %+0.3f*Y %+0.3f*Z %+0.3f',Q(3,:)),'Parent',ax);
 
 % Display joint histograms
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 ax  = axes('Position',[0.1 0.5 0.35 0.3],'Visible','off','Parent',fig);
 H   = spm_hist2(VG.uint8,VF.uint8,VF.mat\VG.mat,[1 1 1]);
 tmp = log(H+1);
@@ -381,7 +393,7 @@ function display_results(VG,VF,x,flags)
 ylabel(spm_str_manip(VF.fname,'k22'),'Parent',ax);
 
 % Display ortho-views
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 spm_orthviews('Reset');
      spm_orthviews('Image',VG,[0.01 0.01 .48 .49]);
 h2 = spm_orthviews('Image',VF,[.51 0.01 .48 .49]);
@@ -389,6 +401,4 @@ function display_results(VG,VF,x,flags)
 st.vols{h2}.premul = inv(spm_matrix(x(:)'));
 spm_orthviews('Space');
 
-spm_print
-return;
-
+spm_print;
diff --git a/spm_cva.m b/spm_cva.m
index 0f2bbd1..3342217 100644
--- a/spm_cva.m
+++ b/spm_cva.m
@@ -81,7 +81,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_cva.m 3710 2010-02-03 19:11:26Z guillaume $
+% $Id: spm_cva.m 4185 2011-02-01 18:46:18Z guillaume $
  
 
 % get figure handles
@@ -120,7 +120,7 @@
     name = spm_input('name','-8','s',con);
 end
 name   = strrep(name,' ','_');
-name   = ['CVA_' name];
+name   = ['CVA_' name '.mat'];
  
 % get current location {mm}
 %--------------------------------------------------------------------------
@@ -262,7 +262,7 @@
  
 % save
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
+if spm_check_version('matlab','7') >= 0
     save(fullfile(SPM.swd,name),'-V6','CVA')
 else
     save(fullfile(SPM.swd,name),'CVA')
diff --git a/spm_dcm_U.m b/spm_dcm_U.m
index 1140191..2fec7cf 100644
--- a/spm_dcm_U.m
+++ b/spm_dcm_U.m
@@ -2,112 +2,96 @@ function spm_dcm_U(DCM_filename,SPM_filename,session,input_nos)
 % Insert new inputs into a DCM model
 % FORMAT spm_dcm_U(DCM_filename,SPM_filename,session,input_nos)
 %
-% DCM_filename      Name of DCM file
-% SPM_filename      Name of SPM file (eg. 'SPM')
-% session           Session number (eg. 1)
-% input_nos         Inputs to include and their parametric modulations (PMs).
-%                   Example for a case without PM: {1, 0, 1} includes inputs 1 and 3.
-%                   Example for a case with PM:    {1,0,[0 0 1],[0 1]} includes the
-%                   non-modulated first input, the second PM of the third
-%                   input and the first PM of the fourth input.  Note that
-%                   this cell array only has to be specified up to the last
-%                   input that is replaced.
+% DCM_filename  - Name of DCM file  (char array)
+% SPM_filename  - Name of SPM file  (char array)
+% session       - Session number    (integer)
+% input_nos     - Inputs to include (cell array)
 %
-% This function can be used, for example, to replace subject X's inputs by subject Y's.
-% The model can then be re-estimated without having to go through
-% model specification again.
+% Examples of specification of parameter 'input_nos':
+% * without parametric modulations:
+%   {1, 0, 1} includes inputs 1 and 3.
+% * with parametric modulations:
+%   {1,0,[0 0 1],[0 1]} includes the non-modulated first input, the second
+%   PM of the third input and the first PM of the fourth input.
+% Note that this cell array only has to be specified up to the last input
+% that is replaced.
+%
+% This function can be used, for example, to replace subject X's inputs by
+% subject Y's. The model can then be re-estimated without having to go
+% through model specification again.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny & Klaas Enno Stephan
-% $Id: spm_dcm_U.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_dcm_U.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
+%-Load DCM and SPM files
+%--------------------------------------------------------------------------
 load(DCM_filename);
 load(SPM_filename);
 
-if session>length(SPM.Sess)
-    disp(sprintf('Error in spm_dcm_U: SPM file does not have %d sessions',session));
-    return
-end
-Sess   = SPM.Sess(session);
-
 
-% Check numbers of inputs match
+%-Get session
 %--------------------------------------------------------------------------
-% Number of selected inputs from SPM file
-m_sel  = length(find(cell2mat(input_nos)));
-
-% Number of inputs in DCM file
-m      = size(DCM.c,2);
-
-if ~(m_sel==m)
-    msg = sprintf('Error in spm_dcm_U: %d inputs expected, but only %d passed on',m,m_sel);
-    disp(msg)
-    error('spm_dcm_U aborted')
-    return
+try
+    Sess = SPM.Sess(session);
+catch
+    error('SPM file does not have a session %d.',session);
 end
 
 
-% Replace inputs
+%-Check numbers of inputs
 %--------------------------------------------------------------------------
-% Number of inputs in SPM file
-u      = length(Sess.U);
-
-% Last relevant input
-u_last = max(size(input_nos));
-if u_last > u 
-    msg = sprintf('Error in spm_dcm_U: more inputs specified than exist in SPM.mat');
-    disp(msg)
-    error('spm_dcm_U aborted')
-    return
+if size(DCM.c,2) ~= sum(cellfun(@nnz,input_nos))
+    error('Number of specified inputs does not match DCM.');
+end
+if length(input_nos) > length(Sess.U)
+    error('More inputs specified than exist in SPM.mat.');
 end
 
-% Loop through all specified inputs and store them in U
+
+%-Replace inputs
+%--------------------------------------------------------------------------
 U.name = {};
 U.u    = [];
 U.dt   = DCM.U.dt;
-for k = 1:u_last,
-    if find(input_nos{k})
-        mo = find(input_nos{k});
-        if (length(mo)-1) > length(Sess.U(k).P)
-            msg = ['Error in spm_dcm_U: more parametric modulations specified than exist for input ' Sess.U(k).name{1} 'in SPM file.'];
-            disp(msg)
-            error('spm_dcm_U aborted')
-            return
+for i  = 1:length(input_nos)
+    if any(input_nos{i})
+        mo = find(input_nos{i});
+        if (length(mo)-1) > length(Sess.U(i).P)
+            error(['More parametric modulations specified than exist ' ...
+                'for input %s in SPM.mat.'],Sess.U(i).name{1});
         end
-        for j=mo,
-            U.u             = [U.u Sess.U(k).u(33:end,j)];
-            U.name{end + 1} = Sess.U(k).name{j};
+        for j=mo
+            U.u             = [U.u Sess.U(i).u(33:end,j)];
+            U.name{end + 1} = Sess.U(i).name{j};
         end
     end
 end
 DCM.U = U;
 
 
-% Check inputs and outputs match up (to the nearest DCM.U.dt)
+%-Check inputs and outputs match up (to the nearest DCM.U.dt)
 %--------------------------------------------------------------------------
-% Get input sampling rate from first input
 DCM.U.dt      = Sess.U(1).dt;
-
-% Use the TR from the SPM data structure
 DCM.Y.dt      = SPM.xY.RT;
 
 num_inputs    = size(DCM.U.u,1);
 input_period  = DCM.U.dt*num_inputs;
 output_period = DCM.v*DCM.Y.dt;
-if ~(round(DCM.v*DCM.Y.dt/DCM.U.dt) == num_inputs)
-    disp('Error in spm_dcm_U: input period and output period do not match');
-    disp(sprintf('Number of inputs=%d, input dt=%1.2f, input period=%1.2f',num_inputs,DCM.U.dt,input_period));
-    disp(sprintf('Number of outputs=%d, output dt=%1.2f, output period=%1.2f',DCM.v,DCM.Y.dt,output_period));
-    error('spm_dcm_U aborted')
-    return
+if round(DCM.v*DCM.Y.dt/DCM.U.dt) ~= num_inputs
+    error(sprintf(['Input period and output period do not match.\n'...
+      sprintf('Number of inputs=%d, input dt=%1.2f, input period=%1.2f\n',...
+        num_inputs,DCM.U.dt,input_period) ...
+      sprintf('Number of outputs=%d, output dt=%1.2f, output period=%1.2f\n',...
+        DCM.v,DCM.Y.dt,output_period)]));
 end
 
 
 % Save DCM with replaced inputs
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
+if spm_check_version('matlab','7') >= 0
     save(DCM_filename, 'DCM','-V6');
 else
     save(DCM_filename, 'DCM');
diff --git a/spm_dcm_average.m b/spm_dcm_average.m
index 0deed0b..afc14d1 100644
--- a/spm_dcm_average.m
+++ b/spm_dcm_average.m
@@ -27,7 +27,7 @@ function spm_dcm_average (P,name)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny & Klaas Enno Stephan
-% $Id: spm_dcm_average.m 3723 2010-02-12 15:15:18Z guillaume $
+% $Id: spm_dcm_average.m 4185 2011-02-01 18:46:18Z guillaume $
 
 try
     P;
@@ -115,10 +115,10 @@ function spm_dcm_average (P,name)
 %-Save new DCM
 %==========================================================================
 DCM.name = [name ' (Bayesian FFX average)'];
-if spm_matlab_version_chk('7') >= 0
-    save(['DCM_avg_' name], 'DCM', '-V6');
+if spm_check_version('matlab','7') >= 0
+    save(['DCM_avg_' name '.mat'], 'DCM', '-V6');
 else
-    save(['DCM_avg_' name], 'DCM');
+    save(['DCM_avg_' name '.mat'], 'DCM');
 end
 
 % Warn the user how this average DCM should NOT be used
diff --git a/spm_dcm_bma.m b/spm_dcm_bma.m
index 19ef94e..33b14b9 100644
--- a/spm_dcm_bma.m
+++ b/spm_dcm_bma.m
@@ -1,16 +1,16 @@
 function bma = spm_dcm_bma(post,post_indx,subj,Nsamp,oddsr)
-% Model-independent samples from DCM posterior  
+% Model-independent samples from DCM posterior
 % FORMAT bma = spm_dcm_bma (post,post_indx,subj,Nsamp,oddsr)
 %
 % post      [Ni x M] vector of posterior model probabilities
-%           If Ni>1 then inference is based on subject-specific RFX posterior 
+%           If Ni>1 then inference is based on subject-specific RFX posterior
 % post_indx models to use in BMA (position of models in subj sctructure)
 % subj      subj(n).sess(s).model(m).fname: DCM filename
 % Nsamp     Number of samples (default = 1e3)
 % oddsr     posterior odds ratio for defining Occam's window (default=0, ie
 %           all models used in average)
 %
-% Ep        [Np x Nsamp] posterior density matrix. Parameter vector is of 
+% Ep        [Np x Nsamp] posterior density matrix. Parameter vector is of
 %           dimension Np and there are Nsamp samples
 % Nocc      Number of models in Occam's window
 %
@@ -22,8 +22,8 @@
 %__________________________________________________________________________
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
-% Will Penny 
-% $Id: spm_dcm_bma.m 3966 2010-07-02 11:54:42Z maria $
+% Will Penny
+% $Id: spm_dcm_bma.m 4067 2010-09-03 16:59:35Z maria $
 
 if nargin < 4 || isempty(Nsamp)
     Nsamp = 1e3;
@@ -41,66 +41,75 @@
     dcm_fmri = 1;
     nreg = DCM.n;
     min  = DCM.M.m;
+    dimD = 0;
 else
     dcm_fmri = 0;
 end
 
+firstsub  = 1;
+firstmod  = 1;
+
 Ep  = [];
 [Ni M] = size(post);
 
-if Ni > 1 
+if Ni > 1
     rfx = 1;
 else
     rfx = 0;
 end
 
 if rfx
-    
+
     for i = 1:Ni,
-        
+
         mp          = max(post(i,:));
         post_ind{i} = find(post(i,:)>mp*oddsr);
         Nocc(i)     = length(post_ind{i});
         disp(' ');
         disp(sprintf('Subject %d has %d models in Occams window',i,Nocc(i)));
-        
+
         if Nocc(i) == 0,
             return;
         end
-        
+
         for occ = 1:Nocc(i),
             m = post_ind{i}(occ);
             disp(sprintf('Model %d, <p(m|Y>=%1.2f',m,post(i,m)));
         end
-        
+
         % Renormalise post prob to Occam group
         renorm(i).post = post(i,post_ind{i});
         sp             = sum(renorm(i).post,2);
         renorm(i).post = renorm(i).post./(sp*ones(1,Nocc(i)));
-        
+
         % Load DCM posteriors for models in Occam's window
         for kk = 1:Nocc(i),
-            
+
             sel     = post_indx(post_ind{i}(kk));
-            
+
             params(i).model(kk).Ep  = subj(i).sess(1).model(sel).Ep;
             params(i).model(kk).vEp = spm_vec(params(i).model(kk).Ep);
             params(i).model(kk).Cp  = full(subj(i).sess(1).model(sel).Cp);
-              
+
+            if dcm_fmri
+                dimDtmp = size(params(i).model(kk).Ep.D,3);
+                if dimDtmp ~= 0, dimD = dimDtmp; firstsub = i; firstmod = kk;end
+            end
+            
             % Average sessions
             if Nses > 1
-                
+
                 clear miCp mEp
                 disp('Averaging sessions...')
-                
-                for ss = 1:Nses 
-                    
+
+                for ss = 1:Nses
+
                     % Only parameters with non-zero prior variance
                     %------------------------------------------------------
                     sess_model.Cp = full(subj(i).sess(ss).model(sel).Cp);
                     pCdiag        = diag(full(sess_model.Cp));
                     wsel          = find(pCdiag);
-                    
+
                     if ss == 1
                         wsel_first = wsel;
                     else
@@ -113,20 +122,20 @@
                             return
                         end
                     end
- 
+
                     % Get posterior precision matrix and mean
                     %------------------------------------------------------
                     Cp           = sess_model.Cp;
                     Ep           = spm_vec(subj(i).sess(ss).model(sel).Ep);
                     miCp(:,:,ss) = inv(full(Cp(wsel,wsel)));
                     mEp(:,ss)    = full(Ep(wsel));
- 
+
                 end
-                
+
                 % Average models using Bayesian fixed-effects analysis
                 %==========================================================
                 Cp(wsel,wsel) = inv(sum(miCp,3));
-                
+
                 pE            = subj(i).sess(ss).model(sel).Ep;
                 weighted_Ep   = 0;
                 for s = 1:Nses
@@ -135,65 +144,71 @@
                 Ep(wsel)    = Cp(wsel,wsel)*weighted_Ep;
                 vEp         = Ep;
                 Ep          = spm_unvec(Ep,pE);
-                
+
                 params(i).model(kk).Ep  = Ep;
                 params(i).model(kk).vEp = vEp;
                 params(i).model(kk).Cp  = Cp;
-            
+
             end
-            
+
             [evec, eval] = eig(params(i).model(kk).Cp);
             deig         = diag(eval);
-            
+
             params(i).model(kk).dCp = deig;
             params(i).model(kk).vCp = evec;
-            
-        end
+
+        end   
     end
-else
+    
+else % Use an FFX
     % Find models in Occam's window
     mp       = max(post);
     post_ind = find(post>mp*oddsr);
     Nocc     = length(post_ind);
     disp(' ');
     disp(sprintf('%d models in Occams window',Nocc));
-    
+
     if Nocc == 0, return; end
-    
+
     for occ = 1:Nocc,
         m = post_ind(occ);
         disp(sprintf('Model %d, p(m|Y)=%1.2f',m,post(m)));
     end
-    
+
     % Renormalise post prob to Occam group
     post=post(post_ind);
     post=post/sum(post);
-    
+
     % Load DCM posteriors for models in Occam's window
     for n=1:Nsub,
-        
+
         for kk=1:Nocc,
-            
+
             sel = post_indx(post_ind(kk));
-            
+
             params(n).model(kk).Ep  = subj(n).sess(1).model(sel).Ep;
             params(n).model(kk).vEp = spm_vec(params(n).model(kk).Ep);
             params(n).model(kk).Cp  = full(subj(n).sess(1).model(sel).Cp);
             
+            if dcm_fmri
+                dimDtmp = size(params(n).model(kk).Ep.D,3);
+                if dimDtmp ~= 0, dimD = dimDtmp; firstsub = n; firstmod = kk; end
+            end
+
             if Nses > 1
-                
+
                 clear miCp mEp
                 disp('Averaging sessions...')
-                
+
                 % Average sessions
                 for ss = 1:Nses
-                    
+
                     % Only parameters with non-zero prior variance
                     %------------------------------------------------------
                     sess_model.Cp = full(subj(n).sess(ss).model(sel).Cp);
                     pCdiag        = diag(full(sess_model.Cp));
                     wsel          = find(pCdiag);
-                    
+
                     if ss == 1
                         wsel_first = wsel;
                     else
@@ -206,20 +221,20 @@
                             return
                         end
                     end
-                    
+
                     % Get posterior precision matrix and mean
                     %------------------------------------------------------
                     Cp           = sess_model.Cp;
                     Ep           = spm_vec(subj(n).sess(ss).model(sel).Ep);
                     miCp(:,:,ss) = inv(full(Cp(wsel,wsel)));
                     mEp(:,ss)    = full(Ep(wsel));
-                    
+
                 end
-                
+
                 % Average models using Bayesian fixed-effects analysis
                 %==========================================================
                 Cp(wsel,wsel) = inv(sum(miCp,3));
-                
+
                 pE            = subj(n).sess(ss).model(sel).Ep;
                 weighted_Ep   = 0;
                 for s = 1:Nses
@@ -228,47 +243,50 @@
                 Ep(wsel)    = Cp(wsel,wsel)*weighted_Ep;
                 vEp         = Ep;
                 Ep          = spm_unvec(Ep,pE);
-                
+
                 params(n).model(kk).Ep  = Ep;
                 params(n).model(kk).vEp = vEp;
                 params(n).model(kk).Cp  = Cp;
-                
+
             end
-            
+
             [evec, eval] = eig(params(n).model(kk).Cp);
             deig         = diag(eval);
-            
+
             params(n).model(kk).dCp = deig;
             params(n).model(kk).vCp = evec;
         end
     end
-    
+
 end
 
 % Pre-allocate sample arrays
-Np = length(params(1).model(1).vEp);
+Np = length(params(firstsub).model(firstmod).vEp);
 
 % get dimensions of a b c d parameters
 if dcm_fmri
-    
+
     Nr      = nreg*nreg;
-    dimd    = size(params(1).model(1).Ep.D,3);
-    Np      = Nr + Nr*min + nreg*min + Nr*dimd;
-    
+
     Etmp.A  = zeros(nreg,nreg,Nsamp);
     Etmp.B  = zeros(nreg,nreg,min,Nsamp);
     Etmp.C  = zeros(nreg,min,Nsamp);
-    Etmp.D  = zeros(nreg,nreg,dimd,Nsamp);
-    
+    Etmp.D  = zeros(nreg,nreg,dimD,Nsamp);
+
     dima    = Nr;
     dimb    = Nr+Nr*min;
     dimc    = Nr+Nr*min+nreg*min;
-    
+
 end
 
 clear Ep
 disp('')
 disp('Averaging models in Occams window...')
+
+Ep_all = zeros(Np,Nsub);
+Ep_sbj = zeros(Np,Nsub,Nsamp);
+Ep     = zeros(Np,Nsamp);
+
 for i=1:Nsamp
     % Pick a model
     if ~rfx
@@ -276,24 +294,28 @@
     end
     % Pick parameters from model for each subject
     for n=1:Nsub
+
+        clear mu dsig vsig
         
         if rfx
             m = spm_multrnd(renorm(n).post,1);
         end
-        
+
         mu                    = params(n).model(m).vEp;
-        dsig                  = params(n).model(m).dCp(:,1);
-        vsig(:,:)             = params(n).model(m).vCp(:,:);
-        
+        nmu                   = length(mu);
+        dsig                  = params(n).model(m).dCp(1:nmu,1);
+        vsig(:,:)             = params(n).model(m).vCp(1:nmu,1:nmu);
+
         tmp                   = spm_normrnd(mu,{dsig,vsig},1);
-        Ep_all(:,n)           = tmp(:);
-        Ep_sbj(:,n,i)         = Ep_all(:,n);
-                   
+        
+        Ep_all(1:nmu,n)      = tmp(:);
+        Ep_sbj(1:nmu,n,i)    = Ep_all(1:nmu,n); 
+
     end
     
     % Average over subjects
-    Ep(:,i)       = mean(Ep_all,2);
-        
+    Ep(:,i) = mean(Ep_all,2);
+
 end
 
 % save mean parameters
@@ -316,7 +338,11 @@
     bma.a  = spm_unvec(Ep(1:dima,:),Etmp.A);
     bma.b  = spm_unvec(Ep(dima+1:dimb,:),Etmp.B);
     bma.c  = spm_unvec(Ep(dimb+1:dimc,:),Etmp.C);
-    bma.d  = spm_unvec(Ep(dimc+1:Np,:),Etmp.D);
+    if dimD ~=0
+        bma.d  = spm_unvec(Ep(dimc+1:dimc+Nr*dimD,:),Etmp.D);
+    else
+        bma.d  = Etmp.D;
+    end
 end
 
 % storing parameters
diff --git a/spm_dcm_bma_results.m b/spm_dcm_bma_results.m
index d3ffcd1..282cff9 100644
--- a/spm_dcm_bma_results.m
+++ b/spm_dcm_bma_results.m
@@ -9,7 +9,7 @@ function spm_dcm_bma_results(BMS,method)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Maria Joao
-% $Id: spm_dcm_bma_results.m 3853 2010-04-29 14:54:43Z christophe $
+% $Id: spm_dcm_bma_results.m 4022 2010-07-28 12:50:20Z guillaume $
 
 if nargin < 1
     fname       = spm_select(1,'^BMS.mat$','select BMS.mat file');
@@ -234,8 +234,8 @@ function plot_a(F)
 
 % spot the bmod input index from the fig name
 ht = intersect(findobj('style','pushbutton'),get(hf,'children'));
-it = strmatch('bold',get(ht,'Fontweight'));
-t_str = get(ht(it),'string');
+ht = findobj(ht,'flat','Fontweight','bold');
+t_str = get(ht,'string');
 b_ind = str2num(t_str(strfind(t_str,'#')+1:end));
 i_mod = find(ud.mod_input==b_ind);
 
@@ -279,8 +279,8 @@ function plot_a(F)
 
 % spot the c_drv input index from the fig name
 ht = intersect(findobj('style','pushbutton'),get(hf,'children'));
-it = strmatch('bold',get(ht,'Fontweight'));
-t_str = get(ht(it),'string');
+ht = findobj(ht,'flat','Fontweight','bold');
+t_str = get(ht,'string');
 c_ind = str2num(t_str(strfind(t_str,'#')+1:end));
 i_drv = find(ud.drive_input==c_ind);
 
@@ -321,8 +321,8 @@ function plot_a(F)
 
 % spot the d_reg input index from the fig name
 ht = intersect(findobj('style','pushbutton'),get(hf,'children'));
-it = strmatch('bold',get(ht,'Fontweight'));
-t_str = get(ht(it),'string');
+ht = findobj(ht,'flat','Fontweight','bold');
+t_str = get(ht,'string');
 d_ind = str2num(t_str(strfind(t_str,'#')+1:end));
 i_mreg = find(ud.mod_reg==d_ind);
 
diff --git a/spm_dcm_create.m b/spm_dcm_create.m
index a5805c2..4ea7da0 100644
--- a/spm_dcm_create.m
+++ b/spm_dcm_create.m
@@ -23,7 +23,7 @@ function spm_dcm_create(syn_model, source_model, SNR)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny & Klaas Enno Stephan
-% $Id: spm_dcm_create.m 3741 2010-03-01 15:35:30Z guillaume $
+% $Id: spm_dcm_create.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 Finter = spm_figure('GetWin','Interactive');
@@ -48,6 +48,7 @@ function spm_dcm_create(syn_model, source_model, SNR)
 switch upper(source_model)
     
     case 'GUI'
+        
         % Define model by GUI
         %==================================================================
 
@@ -249,12 +250,12 @@ function spm_dcm_create(syn_model, source_model, SNR)
 
         
     case 'IMPORT'
+        
         % Import existing model - prompt user to choose it
         %==================================================================
         P     = spm_select(1,'^DCM.*\.mat$','Select source DCM_???.mat');
         load(P)
 
-        
     otherwise
         % Import existing model (directly specified by directory & name)
         %==================================================================
@@ -290,14 +291,14 @@ function spm_dcm_create(syn_model, source_model, SNR)
 
 %-Save and reset title
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
-    save(['DCM_',syn_model],'-V6','DCM');
+if spm_check_version('matlab','7') >= 0
+    save(['DCM_' syn_model '.mat'],'-V6','DCM');
 else
-    save(['DCM_',syn_model],'DCM');
+    save(['DCM_' syn_model '.mat'],'DCM');
 end
 
 % Now generate synthetic output data
 %==========================================================================
-spm_dcm_generate(['DCM_' syn_model],source_model,SNR);
+spm_dcm_generate(['DCM_' syn_model '.mat'],source_model,SNR);
 
 spm('FigName',header);
diff --git a/spm_dcm_delay.m b/spm_dcm_delay.m
new file mode 100644
index 0000000..b1598dc
--- /dev/null
+++ b/spm_dcm_delay.m
@@ -0,0 +1,70 @@
+function [Q,J] = spm_dcm_delay(M,P,D)
+% returns the delay operator for flow and Jacobians of dynamical systems
+% FORMAT [Q,J] = spm_dcm_delay(M,P,D)
+%
+% M   - model specification structure
+% Required fields:
+%   M.f - dx/dt    = f(x,u,P,M)            {function string or m-file}
+%   M.m - m inputs
+%   M.n - n states
+%   M.x - (n x 1) = x(0) = expansion point: defaults to x = 0;
+%   M.u - (m x 1) = u    = expansion point: defaults to u = 0;
+%
+% P     - model parameters
+% D     - delay matrix (among hidden states): default D = zeros(n,n)
+%
+% return the delay operator for Jacobians of dynamical systems where the
+% states are
+%
+% f     - dx(t)/dt  = f(x(t))
+% Q     - delay operator dx(t)/dt = f(x(t - d))
+%                                 = Q(d)*f(x(t))
+% J     - Jacobian  = df/dt = (where delayed Jacobian = Q*J)
+%
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_delay.m 4261 2011-03-24 16:39:42Z karl $
+ 
+% create inline functions
+%--------------------------------------------------------------------------
+try
+    funx = fcnchk(M.f,'x','u','P','M');
+catch
+    M.f  = inline('sparse(0,1)','x','u','P','M');
+    M.n  = 0;
+    M.x  = sparse(0,0);
+    funx = fcnchk(M.f,'x','u','P','M');
+end
+ 
+% expansion point
+%--------------------------------------------------------------------------
+try, x = spm_vec(M.x); catch,  x = sparse(M.n,1); end
+try, u = spm_vec(M.u); catch,  u = sparse(M.m,1); end
+ 
+% Jacobian and delay operator
+%==========================================================================
+ 
+% derivatives
+%--------------------------------------------------------------------------
+J     = spm_diff(funx,x,u,P,M,1);
+ 
+% delay operator
+%--------------------------------------------------------------------------
+D     = -D;
+QJ    = (speye(length(J)) - D.*J)\J;
+Q     = J;
+for n = 1:128
+    
+    % n-th order Taylor term
+    %----------------------------------------------------------------------
+    dQ = ((D.^n).*J)*(QJ^n)/factorial(n);
+    Q  = Q + dQ;
+    
+    % break if convergence
+    %----------------------------------------------------------------------
+    if norm(dQ,'inf') < 1e-6;     break, end
+
+end
+Q      = Q*spm_inv(J);
diff --git a/spm_dcm_diagnose.m b/spm_dcm_diagnose.m
new file mode 100644
index 0000000..41c2451
--- /dev/null
+++ b/spm_dcm_diagnose.m
@@ -0,0 +1,247 @@
+function DCM = spm_dcm_diagnose(DCM,varargin)
+% Post hoc diagnosis of DCMs (under the Laplace approximation)
+% FORMAT spm_dcm_diagnose(DCM,’field,’field’,…)
+%
+% DCM        - DCM stricture (inverted)
+% field      - field name(s) of parameters to consider
+%
+%--------------------------------------------------------------------------
+% This routine searches over all possible reduced models of a full model
+% (DCM) and uses post hoc model selection to select the best. Reduced
+% models mean all permutations of free parameter sets (parameters with non-
+% zero prior covariance), where models are defined in terms of their prior
+% covariance. The full model should be inverted prior to post hoc
+% optimization. If there are more than 8 free-parameter sets, this routine
+% will implement a greedy search: This entails searching over all
+% permutations of the 8 sets whose removal (shrinking the prior
+% variance to zero) produces the smallest reduction (greatest increase)
+% in model evidence. This procedure is repeated until all sets
+% are retained in the best model or there are no more parameters to
+% consider.
+%
+% A parameter set is specified implicitly by the structure (DCM.Ep). Each
+% set corresponds to a column of (the cell arrays or matrix) each field of
+% DCM.Ep.
+%
+% if only one field is specified the log-evidence is computed as a function
+% of the scaled prior variance. Redundant parameters have a log-evidence 
+% that keeps increasing as the prior variance shrinks.
+%
+% The outputs of this routine are graphics reporting the model reduction
+% (optimisation). Red means weak evidence; blue strong evidence (> 3) and 
+% cyan very strong evidence (> 5)
+%__________________________________________________________________________
+% Copyright (C) 2008-2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_diagnose.m 4261 2011-03-24 16:39:42Z karl $
+ 
+ 
+ 
+ 
+%-Get priors and posteriors
+%==========================================================================
+qE      = DCM.Ep;
+qC      = DCM.Cp;
+pE      = DCM.M.pE;
+pC      = DCM.M.pC;
+ 
+if isstruct(pC);
+    pV  = pC;
+    pC  = diag(spm_vec(pC));
+end
+ 
+ 
+ 
+% find the partition of parameters
+%--------------------------------------------------------------------------
+str   = {};
+ind   = {};
+if isstruct(qE)
+    
+    % get fields (parameters) to consider
+    %----------------------------------------------------------------------
+    if ~isempty(varargin)
+        field = varargin;
+    else
+        field = fieldnames(qE);
+    end
+    for i = 1:length(field)
+        q = getfield(qE,field{i});
+        k = spm_fieldindices(qE,field{i});
+        for j = 1:size(q,2)
+            u            = 1:length(spm_vec(q(:,j)));
+            str{end + 1} = [field{i} '(' num2str(j) ')'];
+            ind{end + 1} = k(u); k(u) = [];
+        end
+    end
+    
+else
+    field = {};
+    k     = 1:length(spm_vec(qE));
+    for j = 1:size(qE,2)
+        u            = 1:length(spm_vec(qE(:,j)));
+        str{end + 1} = ['qE(' num2str(j) ')'];
+        ind{end + 1} = k(u); k(u) = [];
+    end
+end
+ 
+% find free parameters
+%--------------------------------------------------------------------------
+ip    = {};
+sp    = {};
+for i = 1:length(str)
+    j = find(diag(pC(ind{i},ind{i})));
+    if ~isempty(j)
+        ip{end + 1} = ind{i}(j);
+        sp{end + 1} = str{i};
+    end
+end
+ 
+ 
+% model search over shrinking priors
+% -------------------------------------------------------------------------
+n     = length(pC);
+if length(field) == 1
+    
+    r     = (0:64)/32;
+    for j = 1:length(r)
+        for i = 1:length(ip)
+            R      = speye(n,n) - sparse(ip{i},ip{i},1,n,n)*(1 - r(j));
+            Z(i,j) = spm_log_evidence(qE,qC,pE,pC,pE,R*pC*R);
+            
+        end
+    end
+    
+    subplot(2,1,1)
+    plot(r,Z)
+    title(['log-evidence for ' field{1}],'FontSize',16);
+    xlabel('proportion of full prior')
+    ylabel('negative free-energy')
+    return
+    
+end
+ 
+ 
+ 
+% model search over new prior without the i-th parameter set
+% -------------------------------------------------------------------------
+for i = 1:length(ip)
+    R    = speye(n,n) - sparse(ip{i},ip{i},1,n,n);
+    Z(i) = spm_log_evidence(qE,qC,pE,pC,pE,R*pC*R);
+end
+ 
+% find parameters with the least evidence
+%--------------------------------------------------------------------------
+[Z i] = sort(-Z);
+ip    = ip(i);
+sp    = sp(i);
+ 
+% If there are too many subsets search use those with the least evidence
+%--------------------------------------------------------------------------
+if length(ip) > 8
+    
+    % flag a greedy search
+    %----------------------------------------------------------------------
+    ip     = ip(1:8);
+    repeat = 1;
+    
+else
+    repeat = 0;
+end
+ 
+% Create model space in terms of free parameter indices
+%--------------------------------------------------------------------------
+K     = spm_perm_mtx(length(ip));
+ 
+% and get log-evidences
+%==========================================================================
+for i = 1:size(K,1)
+    k    = spm_vec(ip(find(K(i,:))));
+    R    = speye(n,n) - sparse(k,k,1,n,n);
+    S(i) = spm_log_evidence(qE,qC,pE,pC,pE,R*pC*R);
+end
+ 
+ 
+% Find a good model with the most parameters
+% =========================================================================
+[i k]  = max(S);
+k      = find(K(k,:));
+ 
+nelim  = length(k);
+repeat = repeat & nelim;
+ 
+% Continue greedy search if any parameters have been eliminated
+%--------------------------------------------------------------------------
+if repeat
+    
+    fprintf('removing %i parameter set: ',nelim)
+    fprintf('%s, ',sp{k});fprintf('\n')
+    
+    % Optimise selected model (prior covariance)
+    %----------------------------------------------------------------------
+    k         = spm_vec(ip(k));
+    R         = speye(n,n) - sparse(k,k,1,n,n);
+    rC        = R*pC*R;
+    
+    % Get posterior of selected model (rC)
+    % ---------------------------------------------------------------------
+    [F Ep Cp] = spm_log_evidence(qE,qC,pE,pC,pE,rC);
+    
+    DCM.M.pC  = rC;
+    DCM.Ep    = Ep;
+    DCM.Cp    = Cp;
+    DCM.F     = F;
+    DCM       = spm_dcm_diagnose(DCM,field{:});
+    return
+    
+else
+    
+    % model posterior
+    % ---------------------------------------------------------------------
+    S     = S - S(end);
+    p     = exp(S - max(S));
+    p     = p/sum(p);
+    
+    fprintf('\n%i Irrelevant parameter sets: ',length(k))
+    fprintf('%s, ',sp{k});fprintf('\n')
+    
+    % Organise field names and show results
+    % ---------------------------------------------------------------------
+    spm_figure('Getwin','Graphics');
+    
+    
+    for i = 1:4
+        j = (1:8) + (i - 1)*8;
+        try
+            tsr{i,1} = sprintf('%s,',sp{j});
+        catch
+            tsr{i,1} = sprintf('%s,',sp{j(1):end});
+        end
+    end
+       
+    subplot(2,1,1)
+    bar(Z,'c'),hold on
+    bar(Z(Z < 5),'b'),hold on
+    bar(Z(Z < 3),'r'),hold off
+    title('relative log evidence (with vs. without)','FontSize',16)
+    xlabel(tsr,'FontSize',16)
+    ylabel('negative free-energy','FontSize',12)
+    
+    subplot(2,2,3)
+    if size(K,1) > 32, plot(S,'k'), else, bar(S,'c'), end
+    title('log-posterior','FontSize',16)
+    xlabel('model','FontSize',12)
+    ylabel('log-probability','FontSize',12)
+    axis square
+    
+    subplot(2,2,4)
+    if size(K,1) > 32, plot(p,'k'), else, bar(p,'r'), end
+    title('model posterior','FontSize',16)
+    xlabel('model','FontSize',12)
+    ylabel('probability','FontSize',12)
+    axis square
+    
+end
+
+
diff --git a/spm_dcm_estimate.m b/spm_dcm_estimate.m
index 3160575..d6a18ab 100644
--- a/spm_dcm_estimate.m
+++ b/spm_dcm_estimate.m
@@ -1,18 +1,19 @@
 function [DCM] = spm_dcm_estimate(P)
 % Estimates parameters of a DCM (bilinear or nonlinear) for fMRI data
 % FORMAT [DCM] = spm_dcm_estimate(DCM)
-%   DCM - the DCM or its filename
+%   DCM - DCM structure or its filename
 %
 % Expects
 %--------------------------------------------------------------------------
-% DCM.a;                             % switch on endogenous connections
-% DCM.b;                             % switch on bilinear modulations
-% DCM.c;                             % switch on exogenous connections
-% DCM.U;                             % exogenous inputs
-% DCM.Y;                             % responses
-% DCM.Y.X0;                          % confounds
-% DCM.n;                             % number of regions
-% DCM.v;                             % number of scans
+% DCM.a                              % switch on endogenous connections
+% DCM.b                              % switch on bilinear modulations
+% DCM.c                              % switch on exogenous connections
+% DCM.d                              % switch on nonlinear modulations
+% DCM.U                              % exogenous inputs
+% DCM.Y                              % responses
+% DCM.Y.X0                           % confounds
+% DCM.n                              % number of regions
+% DCM.v                              % number of scans
 %
 % Options
 %--------------------------------------------------------------------------
@@ -20,10 +21,8 @@
 % DCM.options.stochastic             % fluctuations on hidden states
 % DCM.options.nonlinear              % interactions among hidden states
 % DCM.options.nograph                % graphical display
+% DCM.options.centre                 % mean-centre inputs
 % DCM.options.P                      % Starting estimates for parameters
-% DCM.options.W                      % noise log-precicion (hidden-states)
-% DCM.options.V                      % noise log-precicion (hidden-causes)
-% DCM.options.s                      % smoothness of random fluctuations
 %
 % Evaluates:
 %--------------------------------------------------------------------------
@@ -49,19 +48,19 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_dcm_estimate.m 3888 2010-05-15 18:49:56Z karl $
+% $Id: spm_dcm_estimate.m 4185 2011-02-01 18:46:18Z guillaume $
  
  
-% load DCM structure
+%-Load DCM structure
 %--------------------------------------------------------------------------
 if ~nargin
  
-    %-display model details
+    %-Display model details
     %----------------------------------------------------------------------
     Finter = spm_figure('GetWin','Interactive');
-    set(Finter,'name','Dynamic Causal Modeling')
+    set(Finter,'name','Dynamic Causal Modelling')
  
-    %-get DCM
+    %-Get DCM
     %----------------------------------------------------------------------
     [P, sts] = spm_select(1,'^DCM.*\.mat$','select DCM_???.mat');
     if ~sts, DCM = []; return; end
@@ -71,21 +70,17 @@
 end
 if isstruct(P)
     DCM = P;
-    P   = ['DCM-' date];
+    P   = ['DCM-' date '.mat'];
 else
     load(P)
 end
- 
+
 % check options
 %==========================================================================
 try, DCM.options.two_state;  catch, DCM.options.two_state  = 0;     end
 try, DCM.options.stochastic; catch, DCM.options.stochastic = 0;     end
 try, DCM.options.nonlinear;  catch, DCM.options.nonlinear  = 0;     end
-try, DCM.options.form;       catch, DCM.options.form  = 'Gaussian'; end
-try, DCM.options.W;          catch, DCM.options.W = 9;              end
-try, DCM.options.V;          catch, DCM.options.V = 4;              end
-try, DCM.options.s;          catch, DCM.options.s = 1/4;            end
-
+try, DCM.options.centre;     catch, DCM.options.centre     = 1;     end
 
 try, M.nograph = DCM.options.nograph; catch, M.nograph = spm('CmdLine');end
 try, M.P       = DCM.options.P ;end
@@ -97,7 +92,13 @@
 n  = DCM.n;                             % number of regions
 v  = DCM.v;                             % number of scans
  
- 
+% detrend outputs (and inputs)  
+%--------------------------------------------------------------------------
+Y.y = spm_detrend(Y.y);
+if DCM.options.centre
+    U.u = spm_detrend(U.u);
+end
+
 % check scaling of Y (enforcing a maximum change of 4%
 %--------------------------------------------------------------------------
 scale   = max(max((Y.y))) - min(min((Y.y)));
@@ -105,10 +106,18 @@
 Y.y     = Y.y*scale;
 Y.scale = scale;
 
-% normalise inputs
+% check confounds (add constant if necessary)
 %--------------------------------------------------------------------------
-U.u  = spm_detrend(U.u);
- 
+if ~size(Y.X0,2), Y.X0 = ones(v,1); end
+
+% fMRI slice time sampling
+%--------------------------------------------------------------------------
+try, M.delays = DCM.delays; end
+try, M.TE     = DCM.TE;     end
+
+% create priors
+%==========================================================================
+
 % check DCM.d (for nonlinear DCMs)
 %--------------------------------------------------------------------------
 try
@@ -118,11 +127,6 @@
     DCM.options.nonlinear = 0;
 end
  
-% check confounds (add constant if necessary)
-%--------------------------------------------------------------------------
-if ~size(Y.X0,2), Y.X0 = ones(v,1); end
- 
- 
 % specify parameters for spm_int_D (ensuring updates every second or so)
 %--------------------------------------------------------------------------
 if DCM.options.nonlinear
@@ -133,22 +137,29 @@
     M.IS     = 'spm_int';
 end
  
-% priors
+% check DCM.c (for endogenous DCMs)
 %--------------------------------------------------------------------------
-if DCM.options.two_state
-    [pE,pC] = spm_dcm_fmri_priors(DCM.a,DCM.b,DCM.c,DCM.d,'2s');
-    x       = sparse(n,6);
+if isempty(DCM.c)
+    DCM.c  = zeros(n,1);
+    DCM.b  = zeros(n,n,1);
+    U.u    = zeros(v,1);
+    U.name = {'null'};
+    DCM.options.endogenous = 1;
+end
+if ~any(spm_vec(DCM.c))
+    DCM.options.endogenous = 1;
 else
-    [pE,pC] = spm_dcm_fmri_priors(DCM.a,DCM.b,DCM.c,DCM.d);
-    x       = sparse(n,5);
+    DCM.options.endogenous = 0;
 end
- 
-% fMRI slice time sampling
+if  DCM.options.endogenous
+    DCM.options.two_state  = 0;
+    DCM.options.stochastic = 1;
+end
+
+% priors (and initial states)
 %--------------------------------------------------------------------------
-try, M.delays = DCM.delays; end
-try, M.TE     = DCM.TE;     end
- 
- 
+[pE,pC,x] = spm_dcm_fmri_priors(DCM.a,DCM.b,DCM.c,DCM.d,DCM.options);
+
 % complete model specification
 %--------------------------------------------------------------------------
 M.f   = 'spm_fx_fmri';
@@ -169,8 +180,14 @@
  
 % nonlinear system identification (Variational EM) - deterministic DCM
 %--------------------------------------------------------------------------
-[Ep,Cp,Eh,F] = spm_nlsi_GN(M,U,Y);
- 
+if ~DCM.options.endogenous
+    [Ep,Cp,Eh,F] = spm_nlsi_GN(M,U,Y);
+else
+    Ep = pE;
+    Eh = 3;
+    F  = 0;
+end
+
 % proceed to stochastic (initialising with deterministic estimates)
 %--------------------------------------------------------------------------
 if DCM.options.stochastic
@@ -200,48 +217,59 @@
  
     % set inversion parameters
     % ---------------------------------------------------------------------
-    form    = DCM.options.form;
-    s       = DCM.options.s;
-    DEM.M(1).E.form  = form;       % form of random fluctuations
-    DEM.M(1).E.s     = s;          % smoothness of random fluctuations
-    DEM.M(1).E.d     = 2;          % embedding dimension 
-    DEM.M(1).E.n     = 4;          % embedding dimension
-    DEM.M(1).E.nN    = 8;          % maximum number of DEM iterations
+    DEM.M(1).E.form = 'Gaussian';         % form of random fluctuations
+    DEM.M(1).E.s    = 1/2;                % smoothness of fluctuations
+    DEM.M(1).E.d    = 2;                  % embedding dimension 
+    DEM.M(1).E.n    = 6;                  % embedding dimension
+    DEM.M(1).E.nN   = 32;                 % maximum number of iterations
 
  
     % adjust M.f (DEM works in time bins not seconds) and initialize M.P
     % ---------------------------------------------------------------------
+    DEM.M(1).delays = M.delays/Y.dt;
     DEM.M(1).f  = inline([M.f '(x,v,P)*' num2str(Y.dt)],'x','v','P');
     DEM.M(1).P  = Ep;
-    DEM.M(1).pC = pC;
     
-    % delays (DEM works in time bins not seconds)
-    % ---------------------------------------------------------------------
-    DEM.M(1).delays = M.delays/Y.dt;
-    
-    % Precision on hidden states (s.d. of about 10%)
-    % ---------------------------------------------------------------------
-    DEM.M(1).xP = 128;
     
-    % Specify hyper-priors on precisions
+    % Specify hyper-priors on (log-precision of) observation noise
     % ---------------------------------------------------------------------
     DEM.M(1).Q  = spm_Ce(ones(1,n));
-    DEM.M(1).hE = Eh + 1;      % prior expectation of log-precision (noise)
-    DEM.M(1).hC = 1/8;         % prior covariance  of log-precision (noise)
-
-    DEM.M(1).W  = exp(DCM.options.W);      % fixed precision (hidden-state)
-    DEM.M(2).V  = exp(DCM.options.V);      % fixed precision (hidden-cause)
+    DEM.M(1).hE = Eh + 1;                 % prior expectation
+    DEM.M(1).hC = 1/16;                   % prior covariance
     
     
+    % Specify hyper-priors, allowing hidden causes to fluctuate
+    % ---------------------------------------------------------------------
+    DEM.M(1).xP = 128;                    % s.d. ~ 10%      (hidden-state)
+    DEM.M(1).W  = exp(9);                 % fixed precision (hidden-motion)
+    DEM.M(2).V  = exp(4);                 % fixed precision (hidden-cause)
+        
+    % for endgenous DCMs, allow neuronal hidden states to fluctuate
+    % ---------------------------------------------------------------------
+    if DCM.options.endogenous
+        W           = exp(spm_vec(sparse(1:n,1,(6 - 16),n,5) + 16));
+        DEM.M(1).xP = exp(6);             % fixed precision (hidden-state)
+        DEM.M(1).W  = diag(W);            % fixed precision (hidden-motion)
+        DEM.M(2).V  = exp(16);            % fixed precision (hidden-cause)
+    end
+
+
     % Generalised filtering (under the Laplace assumption)
     % =====================================================================
     if DCM.options.stochastic == 1
-        
-            DEM  = spm_LAP(DEM);           % no mean-field assumption
+            DEM  = spm_LAP(DEM);          % no mean-field assumption
     else
-            DEM  = spm_DEM(DEM);           % mean-field assumption (DEM)
+            DEM  = spm_DEM(DEM);          % mean-field assumption (DEM)
+    end
+    
+    % Helpful suggestion
+    %----------------------------------------------------------------------
+    if length(DEM.F) < 4
+        disp('This inversion may have been unstable;')
+        disp('try reducing DCM.options.s to 1/4')
     end
     
+    
     % Save DEM estimates
     %----------------------------------------------------------------------
     DCM.qU = DEM.qU;
@@ -259,7 +287,7 @@
     %----------------------------------------------------------------------
     y      = DEM.qU.v{1}';
     R      = DEM.qU.z{1}';
-    R      = R - Y.X0*inv(Y.X0'*Y.X0)*(Y.X0'*R);
+    R      = R - Y.X0*spm_inv(Y.X0'*Y.X0)*(Y.X0'*R);
     Ce     = exp(-DEM.qH.h{1});
  
 else
@@ -268,7 +296,7 @@
     %----------------------------------------------------------------------
     y      = feval(M.IS,Ep,M,U);
     R      = Y.y - y;
-    R      = R - Y.X0*inv(Y.X0'*Y.X0)*(Y.X0'*R);
+    R      = R - Y.X0*spm_inv(Y.X0'*Y.X0)*(Y.X0'*R);
     Ce     = exp(-Eh);
     
 end
@@ -334,12 +362,14 @@
  
 %-Save DCM
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
-    save(P,'-V6','DCM','F','Ep','Cp');
-else
-    save(P,'DCM','F','Ep','Cp');
+if ~isstruct(P)
+    if spm_check_version('matlab','7') >= 0
+        save(P,'-V6','DCM','F','Ep','Cp');
+    else
+        save(P,'DCM','F','Ep','Cp');
+    end
 end
- 
+
 if ~nargin
     spm('Pointer','Arrow');
     spm('FigName','Done');
diff --git a/spm_dcm_evidence.m b/spm_dcm_evidence.m
index cd0250a..4d9ae5e 100644
--- a/spm_dcm_evidence.m
+++ b/spm_dcm_evidence.m
@@ -16,7 +16,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_dcm_evidence.m 3741 2010-03-01 15:35:30Z guillaume $
+% $Id: spm_dcm_evidence.m 4281 2011-03-31 19:49:57Z karl $
 
 
 % Only look at those parameters with non-zero prior covariance
@@ -29,9 +29,13 @@
 %--------------------------------------------------------------------------
 for i = 1:n
     try
-        lambda_i = DCM.Ce(i*v,i*v);              % old format
+        lambda_i = DCM.Ce(i*v,i*v);              % Ce is error covariance
     catch
-        lambda_i = DCM.Ce(i);                    % new format
+        try
+            lambda_i = DCM.Ce(i);                % Ce is a hyperparameter
+        catch
+            lambda_i = DCM.Ce;                   % Ce is the hyperparameter
+        end
     end
     evidence.region_cost(i) = -0.5*v*log(lambda_i) ...
         - 0.5*DCM.R(:,i)'*(1/lambda_i)*eye(v)*DCM.R(:,i);
diff --git a/spm_dcm_fmri_priors.m b/spm_dcm_fmri_priors.m
index 56b29e5..8c20f0e 100644
--- a/spm_dcm_fmri_priors.m
+++ b/spm_dcm_fmri_priors.m
@@ -1,9 +1,9 @@
-function [pE,pC] = spm_dcm_fmri_priors(A,B,C,varargin)
+function [pE,pC,x] = spm_dcm_fmri_priors(A,B,C,D,options)
 % Returns the priors for a two-state DCM for fMRI.
-% FORMAT:
-%    for bi-linear DCM: [pE,pC] = spm_dcm_fmri_priors(A,B,C)
-%    for nonlinear DCM: [pE,pC] = spm_dcm_fmri_priors(A,B,C,D)
-%    for two-state DCM: [pE,pC] = spm_dcm_fmri_priors(A,B,C,D,'2s')
+% FORMAT:[pE,pC,x] = spm_dcm_fmri_priors(A,B,C,D,options)
+%
+%   options.two_state:  (0 or 1) one or two states per region
+%   options.endogenous: (0 or 1) exogenous or endogenous fluctuations
 %
 % INPUT:
 %    A,B,C,D - constraints on connections (1 - present, 0 - absent)
@@ -11,6 +11,7 @@
 % OUTPUT:
 %    pE     - prior expectations (connections and hemodynamic)
 %    pC     - prior covariances  (connections and hemodynamic)
+%    x      - prior (initial) states
 %__________________________________________________________________________
 %
 % References for state equations:
@@ -25,23 +26,28 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_fmri_priors.m 3888 2010-05-15 18:49:56Z karl $
-
-
+% $Id: spm_dcm_fmri_priors.m 4146 2010-12-23 21:01:39Z karl $
 
 % number of regions
 %--------------------------------------------------------------------------
-n     = length(A);       % number of regions
+n = length(A);
 
-% varargin (D for nonlinear coupling)
+% check options and D (for nonlinear coupling)
 %--------------------------------------------------------------------------
-if nargin > 3, D = varargin{1}; else, D = zeros(n,n,0); end
-if nargin > 4, two_states = 1;  else, two_states = 0;   end
+try, options.two_state;  catch, options.two_state  = 0; end
+try, options.endogenous; catch, options.endogenous = 0; end
+try, D;                  catch, D = zeros(n,n,0);       end
+
+
+% prior (initial) states and shrinkage priors on A for endogenous DCMs
+%--------------------------------------------------------------------------
+if options.two_state,  x = sparse(n,6); else, x = sparse(n,5); end
+if options.endogenous, a = 128;         else, a = 8;           end
+
 
 % connectivity priors
 %==========================================================================
-
-if two_states
+if options.two_state
     
     % enforce optimisation of intrinsic (I to E) connections
     %----------------------------------------------------------------------
@@ -70,17 +76,17 @@
 
     % prior expectations
     %----------------------------------------------------------------------
-    pE.A  =  A/64 - eye(n,n);
+    pE.A  =  A/(64*n) - eye(n,n)/2;
     pE.B  =  B*0;
     pE.C  =  C*0;
     pE.D  =  D*0;
     
     % prior covariances
     %----------------------------------------------------------------------
-    pC.A  =  A/4 + eye(n,n)/16;
-    pC.B  =  B*4;
-    pC.C  =  C*4;
-    pC.D  =  D*4;
+    pC.A  =  A*a/n + eye(n,n)/(8*n);
+    pC.B  =  B;
+    pC.C  =  C;
+    pC.D  =  D;
 
 end
 
@@ -90,9 +96,9 @@
 pE.decay   = sparse(n,1);
 pE.epsilon = sparse(1,1);
 
-pC.transit = sparse(n,1) + exp(-4);
-pC.decay   = sparse(n,1) + exp(-4);
-pC.epsilon = sparse(1,1) + exp(-4);
+pC.transit = sparse(n,1) + exp(-6);
+pC.decay   = sparse(n,1) + exp(-6);
+pC.epsilon = sparse(1,1) + exp(-6);
 
 pC         = diag(spm_vec(pC));
 
diff --git a/spm_dcm_generate.m b/spm_dcm_generate.m
index 36b0991..b537085 100644
--- a/spm_dcm_generate.m
+++ b/spm_dcm_generate.m
@@ -15,13 +15,13 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny & Klaas Enno Stephan
-% $Id: spm_dcm_generate.m 3899 2010-05-25 15:36:40Z guillaume $
+% $Id: spm_dcm_generate.m 4185 2011-02-01 18:46:18Z guillaume $
 
 % Check parameters and load specified DCM
 %--------------------------------------------------------------------------
 if isstruct(syn_model)
     DCM       = syn_model;
-    syn_model = ['DCM-' date];
+    syn_model = ['DCM-' date '.mat'];
 else
     load(syn_model)
 end
@@ -115,7 +115,7 @@
 DCM.Y  = Y;                                    % simulated data
 DCM.y  = y;                                    % simulated signal
 DCM.M  = M;                                    % model
-if spm_matlab_version_chk('7') >= 0
+if spm_check_version('matlab','7') >= 0
     save(syn_model, 'DCM', '-V6');
 else
     save(syn_model, 'DCM');
diff --git a/spm_dcm_graph.m b/spm_dcm_graph.m
new file mode 100644
index 0000000..8c1cc1d
--- /dev/null
+++ b/spm_dcm_graph.m
@@ -0,0 +1,115 @@
+function spm_dcm_graph(xY,A)
+% Region and anatomical graph display
+% FORMAT spm_dcm_graph(xY,A)
+% xY    - cell of region structures (see spm_regions)
+% A     - connections of weighted directed graph
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_graph.m 4099 2010-10-22 19:47:37Z karl $
+ 
+ 
+% display parameters
+%--------------------------------------------------------------------------
+col   = {'b','g','r','c','m','y','k','w'};
+ 
+% get dimensions, locations and names
+%--------------------------------------------------------------------------
+m     = size(xY,2);
+L     = [];
+S     = [];
+for i = 1:m
+    L       = [L xY(i).xyz];
+    name{i} = xY(i).name(1:min(end,3));
+    S       = [S xY(i).spec];
+end
+
+
+% Render graph in anatomical space
+%==========================================================================
+subplot(2,1,1);cla
+set(gca,'position',[0 .5 1 .5])
+options.query = [];
+options.hfig  = gcf;
+options.ParentAxes = gca;
+options.markersize = 32;
+options.meshsurf   = fullfile(spm('Dir'),'canonical','cortex_5124.surf.gii');
+h  = spm_eeg_displayECD(L,[],8,name,options);
+for i = 1:m
+    set(h.handles.ht(i),'FontWeight','bold')
+end
+set(h.handles.mesh,'FaceAlpha',1/16);
+if nargin < 2, return, end
+ 
+ 
+% Connections
+%--------------------------------------------------------------------------
+W     = max(abs(A),abs(A'));
+W     = W - diag(diag(W));
+W     = 3*W/max(W(:));
+W     = W.*(W > 1/128);
+for i = 1:length(A)
+    for j = (i + 1):length(A)
+        if W(i,j)
+            
+            % associate colour with the strongest influence
+            %--------------------------------------------------------------
+            if abs(A(i,j)) > abs(A(j,i)), c = j; else, c = i; end
+            line(L(1,[i j]),L(2,[i j]),L(3,[i j]),'Color',col{c},...
+                'LineStyle','-',...
+                'LineWidth',W(i,j));
+        end
+    end
+end
+
+
+% Render graph in functional space
+%==========================================================================
+ 
+% Multidimensional scaling (with the Weighted Graph Laplacian)
+%--------------------------------------------------------------------------
+D      = diag(sum(W));
+G      = D - W;
+[U V]  = eig(pinv(G));
+U      = U*sqrt(V);
+[V i]  = sort(-diag(V));
+U      = U(:,i(1:3))';
+ 
+% Procrustean transform to align with anatomy (not currently used)
+%--------------------------------------------------------------------------
+U       = spm_detrend(U')';
+L       = spm_detrend(L')';
+% [R V S] = spm_svd(U*L');
+% R       = R*S;
+% U       = R'*U;
+U       = diag(sign(diag(U*L')))*U;
+U       = real(U*80/max(abs(U(:))));
+ 
+ 
+subplot(2,1,2);cla
+set(gca,'position',[0 0 1 .5])
+options.ParentAxes = gca;
+g      = spm_eeg_displayECD(U,[],16,name,options);
+delete(g.handles.mesh)
+delete(findobj(get(gcf,'Children'),'Type','uicontrol'))
+for i = 1:m
+    set(g.handles.ht(i),'FontWeight','bold')
+end
+ 
+% Connections
+%--------------------------------------------------------------------------
+for i = 1:length(A)
+    for j = (i + 1):length(A)
+        if W(i,j)
+            
+            % associate colour with the strongest influence
+            %--------------------------------------------------------------
+            if abs(A(i,j)) > abs(A(j,i)), c = j; else, c = i; end
+            line(U(1,[i j]),U(2,[i j]),U(3,[i j]),'Color',col{c},...
+                'LineStyle','-',...
+                'LineWidth',W(i,j));
+        end
+    end
+end
+
diff --git a/spm_dcm_optimise.m b/spm_dcm_optimise.m
new file mode 100644
index 0000000..8ce1514
--- /dev/null
+++ b/spm_dcm_optimise.m
@@ -0,0 +1,45 @@
+function [varargout] = spm_dcm_optimise(qE,qC,pE,pC,priorfun,varargin)
+% Optimises the priors of a model (under Laplace approximation)
+% FORMAT [rE,rC] = spm_dcm_optimise(qE,qC,pE,pC,priorfun,varargin)
+%
+% qE,qC    - posterior expectation and covariance of model
+% pE,pC    - prior expectation and covariance of model
+% priorfun - inline function that returns priors
+%            {rE rC} = priorfun(varargin{:})
+%
+% rE,rC    - optimal priors defining a reduced model
+%
+%--------------------------------------------------------------------------
+% This routine optimizes the prior covariance on the free parameters of any 
+% model (DCM) under the Laplace approximation. In other words, it assumes 
+% that the prior means are fixed and will maximise model evidence with 
+% respect to the hyperparameters of a function that returns the prior 
+% covariance. This optimization uses the reduced free-energy, based upon 
+% the posterior and prior densities of the full model supplied. If the 
+% prior covariance function is not specified, this routine will assume a 
+% simple diagonal form with a single hyperparameter. In principle, this 
+% routine can be used in a flexible and powerful way to emulate 
+% hierarchical modeling by using suitable prior covariance functions with 
+% unknown hyperparameters. The outputs are the prior moments (mean and 
+% covariance) of the optimum model.
+%__________________________________________________________________________
+% Copyright (C) 2008-2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_optimise.m 4261 2011-03-24 16:39:42Z karl $
+ 
+% Compute reduced log-evidence
+%==========================================================================
+
+% default prior function
+%--------------------------------------------------------------------------
+if nargin == 4
+    priorfun    = inline('{v2, diag(exp(v1))}','v1','v2');
+    varargin{1} = log(diag(pC));
+    varargin{2} = pE;
+end
+
+varargout   = spm_argmax('spm_log_evidence',qE,qC,pE,pC,priorfun,varargin{:},6);
+varargin{1} = varargout;
+varargout   = priorfun(varargin{:});
+
diff --git a/spm_dcm_post_hoc.m b/spm_dcm_post_hoc.m
new file mode 100644
index 0000000..76ea0b2
--- /dev/null
+++ b/spm_dcm_post_hoc.m
@@ -0,0 +1,302 @@
+function spm_dcm_post_hoc(P)
+% Post hoc optimisation of DCMs (under Laplace approximation)
+% FORMAT spm_dcm_post_hoc(P)
+%
+% P         - character/cell array of DCM filenames
+%           - or cell array of DCM structures
+%
+%--------------------------------------------------------------------------
+% This routine searches over all possible reduced models of a full model 
+% (DCM) and uses post hoc model selection to select the best. Reduced 
+% models mean all permutations of free parameters (parameters with a non-
+% zero prior covariance), where models are defined in terms of their prior 
+% covariance. The full model should be inverted prior to post hoc 
+% optimization. If there are more than 16 free-parameters, this routine 
+% will implement a greedy search: This entails searching over all 
+% permutations of the 8 parameters whose removal (shrinking the prior 
+% variance to zero) produces the smallest reduction (greatest increase) 
+% in model evidence. This procedure is repeated until all 8 parameters 
+% are retained in the best model or there are no more parameters to 
+% consider. When several DCMs are optimized together (as in group studies), 
+% they are checked to ensure the same free parameters have been specified 
+% and the log-evidences are pooled in a fixed effects fashion.
+% 
+% This application of post hoc optimization assumes the DCMs that are 
+% optimized are the same model of different data. Normally, this would be 
+% a full model, in the sense of having the maximum number of free 
+% parameters, such that the set of reduced models is as large as possible. 
+% In contrast spm_dcm_search operates on different DCMs of the same data 
+% to identify the best model
+% 
+% The outputs of this routine are graphics reporting the model reduction 
+% (optimisation) and a DCM_opt_??? for every input DCM that contains the 
+% reduced conditional parameters estimates (for simplicity, the original 
+% kernels and predicted states are retained). The structural and function 
+% (spectral embedding) graphs are based on Bayesian parameter averages 
+% over multiple DCMs.
+%__________________________________________________________________________
+% Copyright (C) 2008-2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_post_hoc.m 4261 2011-03-24 16:39:42Z karl $
+ 
+% get filenames
+%--------------------------------------------------------------------------
+try
+    P;
+catch
+    [P, sts] = spm_select([1 Inf],'^DCM.*\.mat$','Select DCM*.mat files');
+    if ~sts, return; end
+end
+ 
+if ischar(P),   P = cellstr(P); end
+if isstruct(P), P = {P}; end
+N = numel(P);
+ 
+%-Check models are compatible in terms of their prior variances
+%==========================================================================
+for j = 1:N
+    
+    % get prior covariances
+    %----------------------------------------------------------------------
+    try, load(P{j}); catch, DCM = P{j}; end
+    pC   = diag(DCM.M.pC);
+    
+    
+    % and compare it with the first model
+    %----------------------------------------------------------------------
+    if j == 1
+        C = pC;
+    else
+        if any(~pC - ~C)
+            fprintf('Please check model %i for compatibility',j)
+            return
+        end
+    end
+end
+ 
+% find free coupling parameters
+%--------------------------------------------------------------------------
+k     = spm_fieldindices(DCM.Ep,'A','B','D');
+k     = k(~~C(k));
+n     = length(C);
+
+
+% If there are too many find those with the least evidence
+%--------------------------------------------------------------------------
+if length(k) > 16
+    
+    %-Loop through DCMs and free parameters and get log-evidences
+    %----------------------------------------------------------------------
+    for j = 1:N
+        
+        try, load(P{j}); catch, DCM = P{j}; end
+        
+        % Get priors and posteriors
+        % -----------------------------------------------------------------
+        qE    = DCM.Ep;
+        qC    = DCM.Cp;
+        pE    = DCM.M.pE;
+        pC    = DCM.M.pC;
+        
+        % model search over new prior without the i-th parameter
+        % -----------------------------------------------------------------
+        for i = 1:length(k)
+            R      = speye(n,n) - sparse(k(i),k(i),1,n,n);
+            Z(i,j) = spm_log_evidence(qE,qC,pE,pC,pE,R*pC*R);
+        end
+    end
+    
+    % find parameters with the least evidence
+    %----------------------------------------------------------------------
+    Z      = sum(Z,2);
+    [Z i]  = sort(-Z);
+    k      = k(i(1:8));
+    
+    % flag a greedy search
+    %----------------------------------------------------------------------
+    repeat = 1;
+    
+else
+    repeat = 0;
+end
+
+% Create model space in terms of free parameter indices
+%--------------------------------------------------------------------------
+K     = spm_perm_mtx(length(k));
+ 
+%-Loop through DCMs and models and get log-evidences
+%==========================================================================
+for j = 1:N
+    
+    try, load(P{j}); catch, DCM = P{j}; end
+    fprintf('\nsearching (%i): 00%%',j)
+    
+    % Get priors and posteriors
+    % ---------------------------------------------------------------------
+    qE    = DCM.Ep;
+    qC    = DCM.Cp;
+    pE    = DCM.M.pE;
+    pC    = DCM.M.pC;
+    
+    % model search over new prior (covariance)
+    % ---------------------------------------------------------------------
+    for i = 1:length(K)
+        R      = speye(n,n) - sparse(k,k,K(i,:),n,n);
+        G(i,j) = spm_log_evidence(qE,qC,pE,pC,pE,R*pC*R);
+    
+        fprintf('\b\b\b\b')
+        fprintf('%-3.0f%%',i*100/length(K))
+    end 
+end
+fprintf('\n')
+ 
+% Pooled model evidence
+% =========================================================================
+S     = sum(G,2);
+S     = S - S(end);
+p     = exp(S - max(S));
+p     = p/sum(p);
+ 
+% Get selected model (prior covariance)
+%--------------------------------------------------------------------------
+[q i] = max(p);
+R     = speye(n,n) - sparse(k,k,K(i,:),n,n);
+rC    = R*pC*R;
+ 
+% Continue greedy search if any parameters have been eliminated
+%--------------------------------------------------------------------------
+nelim  = full(sum(K(i,:)));
+repeat = repeat & nelim;
+if repeat
+    fprintf('%i paramters eliminated in this step\n',nelim)
+end
+ 
+% Show results
+% -------------------------------------------------------------------------
+spm_figure('Getwin','Graphics');
+ 
+subplot(2,2,1)
+if length(K) > 32, plot(S,'k'), else, bar(S,'c'), end
+title('log-posterior','FontSize',16)
+xlabel('model','FontSize',12)
+ylabel('log-probability','FontSize',12)
+axis square
+ 
+subplot(2,2,2)
+if length(K) > 32, plot(p,'k'), else, bar(p,'r'), end
+title('model posterior','FontSize',16)
+xlabel('model','FontSize',12)
+ylabel('probability','FontSize',12)
+axis square
+ 
+% conditional estimates of selected model
+% =========================================================================
+Eq    = 0;
+Pq    = 0;
+for j = 1:N
+    
+    % Get priors and posteriors
+    % ---------------------------------------------------------------------
+    try, load(P{j}); catch, DCM = P{j}; end
+    
+    qE   = DCM.Ep;
+    qC   = DCM.Cp;
+    pE   = DCM.M.pE;
+    pC   = DCM.M.pC;
+    
+    % Get posterior of selected model - rC
+    % ---------------------------------------------------------------------
+    [F Ep Cp] = spm_log_evidence(qE,qC,pE,pC,pE,rC);
+    
+    % Bayesian parameter average
+    %----------------------------------------------------------------------
+    Pp   = spm_inv(Cp);
+    Eq   = Eq + Pp*spm_vec(Ep);
+    Pq   = Pq + Pp;
+ 
+ 
+    % Put reduced conditional estimates in DCM
+    % =====================================================================
+ 
+    % Bayesian inference and variance
+    %----------------------------------------------------------------------
+    sw   = warning('off','SPM:negativeVariance');
+    Pp   = spm_unvec(1 - spm_Ncdf(DCM.T,abs(spm_vec(Ep)),diag(Cp)),Ep);
+    Vp   = spm_unvec(diag(Cp),Ep);
+    warning(sw);
+    
+    
+    % Store parameter estimates
+    %----------------------------------------------------------------------
+    DCM.M.pC = rC;
+    DCM.Ep   = Ep;
+    DCM.Cp   = Cp;
+    DCM.Pp   = Pp;
+    DCM.Vp   = Vp;   
+    
+    % Save approximations to model evidence: negative free energy, AIC, BIC
+    %----------------------------------------------------------------------
+    evidence = spm_dcm_evidence(DCM);
+    DCM.F    = F;
+    DCM.AIC  = evidence.aic_overall;
+    DCM.BIC  = evidence.bic_overall;
+    
+    
+    %-Save optimised DCM
+    %======================================================================
+    try
+        [pth, name] = fileparts(P{j});
+        P{j} = fullfile(pth,['DCM_opt_' name(4:end) '.mat']);
+    catch
+        P{j} = fullfile(pwd,['DCM_opt_' date '.mat']);
+    end
+    if spm_check_version('matlab','7') >= 0
+        save(P{j},'-V6','DCM','F','Ep','Cp');
+    else
+        save(P{j},'DCM','F','Ep','Cp');
+    end
+    
+end
+ 
+% Show full and reduced conditional estimates (for last DCM)
+%--------------------------------------------------------------------------
+spm_figure('Getwin','Graphics');
+ 
+i   = spm_fieldindices(DCM.Ep,'A','B','C','D');
+qE  = spm_vec(qE);
+Ep  = spm_vec(Ep);
+ 
+subplot(2,2,3)
+spm_plot_ci(qE(i),qC(i,i)), hold on
+title('MAP connections (full)','FontSize',16)
+axis square
+a   = axis;
+ 
+subplot(2,2,4)
+spm_plot_ci(Ep(i),Cp(i,i)), hold off
+title('MAP connections (reduced)','FontSize',16)
+axis square
+axis(a)
+drawnow
+ 
+% repeat optimisation (greedy search) if necessary
+%--------------------------------------------------------------------------
+if repeat
+    spm_dcm_post_hoc(P);
+    return
+end
+ 
+ 
+% Show structural and functional graphs
+%--------------------------------------------------------------------------
+spm_figure('Getwin','Graph'); clf
+ 
+% Bayesian parameter average
+%--------------------------------------------------------------------------
+Cq  = spm_inv(Pq);
+Eq  = Cq*Eq;
+Eq  = spm_unvec(Eq,pE);
+A   = Eq.A + sum(Eq.B,3) + sum(Eq.D,3);
+ 
+spm_dcm_graph(DCM.xY,A)
diff --git a/spm_dcm_review.m b/spm_dcm_review.m
index 262bd9a..2e5ba5e 100644
--- a/spm_dcm_review.m
+++ b/spm_dcm_review.m
@@ -20,7 +20,7 @@ function spm_dcm_review(DCM,action)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_review.m 3888 2010-05-15 18:49:56Z karl $
+% $Id: spm_dcm_review.m 4057 2010-08-31 15:23:39Z guillaume $
 
 
 %-Get DCM structure
@@ -50,12 +50,12 @@ function spm_dcm_review(DCM,action)
 % experimental input specific reports
 %----------------------------------------------------------------------
 for i = 1:m
-    str{i} = ['    effects of ' DCM.U.name{i}];
+    inputstr{i} = ['    effects of ' DCM.U.name{i}];
 end
 
 % connectivity and kernels
 %----------------------------------------------------------------------
-str{end + 1} = 'fixed connections';
+str = [inputstr,{'fixed connections'}];
 
 if isfield(DCM,'averaged')
     str    = {str{:},'contrast of connections'};
@@ -75,11 +75,18 @@ function spm_dcm_review(DCM,action)
         % region specific reports
         %------------------------------------------------------------------
         for i = 1:l
-            str{end + 1} = ['   hidden states: ' DCM.Y.name{i}];
+            hiddenstr{i} = ['   hidden states: ' DCM.Y.name{i}];
         end
+        str = [str,hiddenstr];
+    else
+        hiddenstr = {};
     end
 end
-str{end + 1} = 'quit';
+
+if ~isfield(DCM,'M')
+    str = {'location of regions','inputs','outputs'};
+end
+str = [str,{'quit'}];
 
 %-Get action
 %==========================================================================
@@ -89,12 +96,8 @@ function spm_dcm_review(DCM,action)
     
     %-Get action
     %----------------------------------------------------------------------
-    if isfield(DCM,'M')
-        action = str{spm_input('review',1,'m',str)};
-    else
-        str2   = str([m+3 m+4 end]);
-        action = str2{spm_input('review',1,'m',str2)};
-    end
+    action = str{spm_input('review',1,'m',str)};
+
 end
 
 
@@ -117,10 +120,9 @@ function spm_dcm_review(DCM,action)
 
 %-Set up Graphics window
 %--------------------------------------------------------------------------
-if nargin < 2
-    Fgraph = spm_figure('GetWin','Graphics');
-    figure(Fgraph); spm_figure('Clear',Fgraph);
-end
+Fgraph = spm_figure('GetWin','Graphics');
+spm_figure('Clear',Fgraph);
+set(Fgraph,'Renderer','zbuffer')
 
 switch action
 
@@ -128,7 +130,7 @@ function spm_dcm_review(DCM,action)
     %======================================================================
     % Inputs
     %======================================================================
-    case str(1:m)
+    case inputstr
 
         %-input effects
         %------------------------------------------------------------------
@@ -291,10 +293,9 @@ case str(1:m)
         % transverse
         %------------------------------------------------------------------
         subplot(2,1,1)
-        spm_dcm_display(DCM.xY);
+        spm_dcm_graph(DCM.xY);
         title('Regional locations','FontSize',16)
 
-
         % table
         %------------------------------------------------------------------
         subplot(2,1,2)
@@ -337,19 +338,19 @@ case str(1:m)
 
             % posteriors (if stochastic DCM)
             %--------------------------------------------------------------
-            if DCM.options.stochastic
+            try
                 hold on
                 plot(t,DCM.qU.v{2}(i,:),':')
-                hold off
+                
             end
-
+            hold off
             title(['Input - ' DCM.U.name{i}],'FontSize',16)
             ylabel('event density {Hz}')
-            a = axis;
-            axis([0 max(x) (a(3) - 1/8) (a(4) + 1/8)]);
+            axis tight
+            lm = get(gca,'ylim'); set(gca,'ylim',lm + [-1 1]*diff(lm)/16);
         end
         xlabel('time {seconds}')
-        if DCM.options.stochastic
+        if DCM.options.stochastic && isfield(DCM,'M')
             legend({'prior','posterior'})
         end
 
@@ -361,14 +362,20 @@ case str(1:m)
 
         % graph
         %------------------------------------------------------------------
-        x  = [1:length(DCM.y)]*DCM.Y.dt;
+        x     = [1:DCM.v]*DCM.Y.dt;
         for i = 1:l
             subplot(l,1,i);
-            plot(x,DCM.y(:,i),x,DCM.y(:,i) + DCM.R(:,i),':');
-            title([DCM.Y.name{i} ': responses and predictions' ],'FontSize',16);
+            try
+                plot(x,DCM.y(:,i),x,DCM.y(:,i) + DCM.R(:,i),':');
+                title([DCM.Y.name{i} ': responses and predictions' ],'FontSize',16);
+                xlabel('time {seconds}');
+                legend('predicted', 'observed')
+            catch
+                plot(x,DCM.Y.y(:,i));
+                title([DCM.Y.name{i} ': responses' ],'FontSize',16);
+                xlabel('time {seconds}');
+            end
         end
-        xlabel('time {seconds}');
-        legend('predicted', 'observed')
 
 
     %======================================================================
@@ -427,12 +434,12 @@ case str(1:m)
     case {'estimates of precisions'}
         spm_DEM_qH(DCM.qH)
 
-    case {str{end - l:end - 1}}
+    case hiddenstr
         
         
         % get region
         %------------------------------------------------------------------
-        i = find(strcmp(action,str)) + m - length(str) + 1;
+        i = find(strcmp(action,str)) - find(strcmp('estimates of precisions',str));
         t = [1:length(DCM.Y.y)]*DCM.Y.dt;
 
 
diff --git a/spm_dcm_search.m b/spm_dcm_search.m
new file mode 100644
index 0000000..25ba74f
--- /dev/null
+++ b/spm_dcm_search.m
@@ -0,0 +1,258 @@
+function spm_dcm_search(P)
+% Post hoc optimisation of DCMs (under Laplace approximation)
+% FORMAT spm_dcm_search(P)
+%
+% P         -  character/cell array of DCM filenames
+%
+%--------------------------------------------------------------------------
+% spm_dcm_search operates on different DCMs of the same data to identify
+% the best model. It will invert the full model whose free-parameters are
+% the union (superset) of all free parameters in each model specified. The
+% routine then uses a post hoc selection procedure to evaluate the log-
+% evidence and conditional density over free-parameters of each model
+% specified.
+%
+% The DCM specified does not need to be estimated. spm_dcm_search will 
+% invert the requisite (full DCM) automatically.
+%
+% The outputs of this routine are graphics reporting the model space search
+% (optimisation) and a DCM_optimum (in the first DCMs directory) for the
+% best DCM. The structural and function (spectral embedding) graphs are
+% based on this DCM.
+%
+% DCM_optimum  contains the fields:
+%        DCM.P   - character/cell array of DCM filenames
+%        DCM.PF  - their associated free energies
+%        DCM.PP  - and posterior (model) probabilities
+%
+% In addition, the free energies and posterior estimates of each DCM in P 
+% are saved for subsequent searches over different partitions of model 
+% space.
+%
+%__________________________________________________________________________
+% Copyright (C) 2008-2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_search.m 4261 2011-03-24 16:39:42Z karl $
+ 
+% get filenames
+%--------------------------------------------------------------------------
+try
+    P;
+catch
+    [P, sts] = spm_select([2 Inf],'^DCM.*\.mat$','Select DCM*.mat files');
+    if ~sts, return; end
+end
+ 
+if ischar(P), P = cellstr(P); end
+N = numel(P);
+ 
+%-Check models are compatible in terms of their data
+%==========================================================================
+for j = 1:N
+    
+    % get prior covariances
+    %----------------------------------------------------------------------
+    load(P{j});
+    
+    % and compare it with the first model
+    %----------------------------------------------------------------------
+    if j == 1
+        Y = DCM.Y.y;
+    else
+        try
+            if any(any(Y - DCM.Y.y))
+                fprintf('Please check model %i for compatibility',j)
+                return
+            end
+        catch
+            fprintf('Please check model %i for compatibility',j)
+            return
+        end
+    end
+    
+    % accumate model in terms of which parameters are free
+    %----------------------------------------------------------------------
+    A{j} = DCM.a;
+    B{j} = DCM.b;
+    C{j} = DCM.c;
+    D{j} = DCM.d;
+    
+    % Get full models free parameters
+    %----------------------------------------------------------------------
+    if j == 1
+        a    = A{j};
+        b    = B{j};
+        c    = C{j};
+        d    = D{j};
+    else
+        a    = a | A{j};
+        b    = b | B{j};
+        c    = c | C{j};
+        d    = d | D{j};
+    end
+end
+ 
+%-Estimate full model
+%==========================================================================
+DCM.a = a;
+DCM.b = b;
+DCM.c = c;
+DCM.d = d;
+ 
+DCM.name = 'optimum';
+ 
+% Get full priors and posteriors
+% -------------------------------------------------------------------------
+FUL   = spm_dcm_estimate(DCM);
+ 
+qE    = FUL.Ep;
+qC    = FUL.Cp;
+pE    = FUL.M.pE;
+pC    = FUL.M.pC;
+ 
+%-Loop through models and get log-evidences
+%==========================================================================
+for j = 1:N
+    
+    load(P{j});
+    
+    % Get model (priors) and evaluate (reduced) free-energy and posteriors
+    % ---------------------------------------------------------------------
+    [rE rC]   = spm_dcm_fmri_priors(DCM.a,DCM.b,DCM.c,DCM.d,DCM.options);
+    [F Ep Cp] = spm_log_evidence(qE,qC,pE,pC,rE,rC);
+    
+    % Put reduced conditional estimates in DCM
+    % =====================================================================
+    
+    % Bayesian inference and variance
+    %----------------------------------------------------------------------
+    sw       = warning('off','SPM:negativeVariance');
+    Pp       = spm_unvec(1 - spm_Ncdf(0,abs(spm_vec(Ep)),diag(Cp)),Ep);
+    Vp       = spm_unvec(diag(Cp),Ep);
+    warning(sw);
+    
+    
+    % Store parameter estimates
+    %----------------------------------------------------------------------
+    DCM.M.pC = rC;
+    DCM.Ep   = Ep;
+    DCM.Cp   = Cp;
+    DCM.Pp   = Pp;
+    DCM.Vp   = Vp;
+    DCM.T    = 0;
+    
+    % Store predictions of states from full model for simplicity
+    %----------------------------------------------------------------------
+    DCM.Ce   = FUL.Ce;
+    DCM.H1   = FUL.H1;
+    DCM.K1   = FUL.K1;
+    DCM.R    = FUL.R;
+    DCM.y    = FUL.y;
+    
+    % Save approximations to model evidence: negative free energy, AIC, BIC
+    %----------------------------------------------------------------------
+    evidence = spm_dcm_evidence(DCM);
+    DCM.F    = F;
+    DCM.AIC  = evidence.aic_overall;
+    DCM.BIC  = evidence.bic_overall;
+    
+    % Save DCM
+    %======================================================================
+    if spm_check_version('matlab','7') >= 0
+        save(P{j},'-V6','DCM','F','Ep','Cp');
+    else
+        save(P{j},'DCM','F','Ep','Cp');
+    end
+    
+    % Record free-energy
+    %----------------------------------------------------------------------
+    G(j) = F;
+    
+end
+ 
+% Model evidences and best model
+% =========================================================================
+G     = G - min(G);
+p     = exp(G - max(G));
+p     = p/sum(p);
+ 
+% Get selected model
+%--------------------------------------------------------------------------
+[q j] = max(p);
+load(P{j});
+ 
+i   = spm_fieldindices(DCM.Ep,'A','B','C','D');
+qE  = spm_vec(FUL.Ep);
+Ep  = spm_vec(DCM.Ep);
+qC  = DCM.Cp;
+Cp  = DCM.Cp;
+F   = DCM.F;
+ 
+% Show results
+% -------------------------------------------------------------------------
+spm_figure('Getwin','Graphics'); clf
+ 
+subplot(2,2,1)
+if length(P) > 32, plot(G,'k'), else, bar(G,'c'), end
+title('log-posterior','FontSize',16)
+xlabel('model','FontSize',12)
+ylabel('log-probability','FontSize',12)
+axis square
+ 
+subplot(2,2,2)
+if length(P) > 32, plot(p,'k'), else, bar(p,'r'), end
+title('model posterior','FontSize',16)
+xlabel('model','FontSize',12)
+ylabel('probability','FontSize',12)
+axis square
+ 
+% Show full and reduced conditional estimates (for optimum DCM)
+%--------------------------------------------------------------------------
+subplot(2,2,3)
+spm_plot_ci(qE(i),qC(i,i))
+title('MAP connections (full)','FontSize',16)
+axis square
+a   = axis;
+ 
+subplot(2,2,4)
+spm_plot_ci(Ep(i),Cp(i,i))
+title('MAP connections (optimum)','FontSize',16)
+axis square
+axis(a)
+ 
+% Show structural and functional graphs
+%--------------------------------------------------------------------------
+spm_figure('Getwin','Graph'); clf
+ 
+spm_dcm_graph(DCM.xY,DCM.Ep.A)
+ 
+ 
+%-Save optimum and full DCM
+%==========================================================================
+DCM.P  = P;
+DCM.PF = G;
+DCM.PP = p;
+ 
+% Reduced model (optimum)
+%--------------------------------------------------------------------------
+pth      = fileparts(P{1});
+filename = fullfile(pth,'DCM_optimum.mat');
+if spm_check_version('matlab','7') >= 0
+    save(filename,'-V6','DCM','F','Ep','Cp');
+else
+    save(filename,'DCM','F','Ep','Cp');
+end
+ 
+% Full model
+%--------------------------------------------------------------------------
+DCM      = FUL;
+Ep       = FUL.Ep;
+Cp       = FUL.Cp;
+F        = FUL.F;
+filename = fullfile(pth,'DCM_full');
+if spm_check_version('matlab','7') >= 0
+    save(filename,'-V6','DCM','F','Ep','Cp');
+else
+    save(filename,'DCM','F','Ep','Cp');
+end
diff --git a/spm_dcm_sparse_priors.m b/spm_dcm_sparse_priors.m
new file mode 100644
index 0000000..a633f67
--- /dev/null
+++ b/spm_dcm_sparse_priors.m
@@ -0,0 +1,53 @@
+function [A,K,k] = spm_dcm_sparse_priors(n)
+% Returns Adjacency matrices for bidirectional coupling
+%
+% INPUT:
+%    n         - number of nodes
+%
+% OUTPUT:
+%    A{:}      - adjacency matrices
+%    K{1:K}{:} - adjacency matrices (for k - 1 edges)
+%    k         - row vector of edge numbers (size)
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_dcm_sparse_priors.m 4146 2010-12-23 21:01:39Z karl $
+
+
+%--------------------------------------------------------------------------
+if nargin < 2; k = n - 1; end
+
+% get permutations of connections
+%==========================================================================
+N     = (n*n - n)/2;                % number of (bidirectional) connections
+np    = [1; 0];
+for i = 1:N
+    p     = np;
+    np    = [p; p];
+    for j = 1:i
+       np(:,j)    = kron(p(:,j),[1; 1]);
+    end
+    np(:,(i + 1)) = kron(ones(2^i,1),[1; 0]);
+end
+
+
+% get permutations of K connections (in A{:})
+%==========================================================================
+[I J] = find(triu(ones(n,n),1));
+K     = cell(N + 1,1);
+A     = {};
+for i = 1:2^N
+    
+    % adjacency matrix
+    %----------------------------------------------------------------------
+    a = full(sparse(I,J,p(i,:),n,n));
+    a = a + a' + eye(n,n);
+    
+    % save in arrays
+    %----------------------------------------------------------------------
+    k(i,1)               = sum(p(i,:));
+    K{1 + k(i)}{end + 1} = a;
+    A{i}                 = a;
+    
+end
diff --git a/spm_dcm_specify.m b/spm_dcm_specify.m
index 7a01a39..c58ebea 100644
--- a/spm_dcm_specify.m
+++ b/spm_dcm_specify.m
@@ -7,7 +7,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_specify.m 3710 2010-02-03 19:11:26Z guillaume $
+% $Id: spm_dcm_specify.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 %-Interactive window
@@ -44,7 +44,7 @@
 
 %-Get cell array of region structures
 %--------------------------------------------------------------------------
-P     = cellstr(spm_select([2 8],'^VOI.*\.mat$',{'select VOIs'},'',swd));
+P     = cellstr(spm_select([1 8],'^VOI.*\.mat$',{'select VOIs'},'',swd));
 m     = numel(P);
 for i = 1:m
     p     = load(P{i},'xY');
@@ -56,25 +56,32 @@
 % Inputs
 %==========================================================================
 
-%-Get 'causes' or inputs U
+%-Get (nc) 'causes' or inputs U
 %--------------------------------------------------------------------------
 spm_input('Input specification:...  ',1,'d');
 Sess   = SPM.Sess(xY(1).Sess);
-U.dt   = Sess.U(1).dt;
-u      = length(Sess.U);
-U.name = {};
-U.u    = [];
-for  i = 1:u
-    for j = 1:length(Sess.U(i).name)
-        str = ['include ' Sess.U(i).name{j} '?'];
-        if spm_input(str,'+1','y/n',[1 0],1)
-            U.u             = [U.u Sess.U(i).u(33:end,j)];
-            U.name{end + 1} = Sess.U(i).name{j};
+if isempty(Sess.U)
+    % spontaneous activity, i.e. no stimuli
+    nc = 0;
+    U = [];
+else
+    % with stimuli
+    U.dt   = Sess.U(1).dt;
+    u      = length(Sess.U);
+    U.name = {};
+    U.u    = [];
+    for  i = 1:u
+        for j = 1:length(Sess.U(i).name)
+            str = ['include ' Sess.U(i).name{j} '?'];
+            if spm_input(str,'+1','y/n',[1 0],1)
+                U.u             = [U.u Sess.U(i).u(33:end,j)];
+                U.name{end + 1} = Sess.U(i).name{j};
+            end
         end
     end
+    nc     = size(U.u,2);
 end
 
-
 %==========================================================================
 % Timings
 %==========================================================================
@@ -105,19 +112,32 @@
 %==========================================================================
 % Model options
 %==========================================================================
-spm_input('Model options:...  ',-1,'d');
-options.nonlinear  = spm_input('modulatory effects','+1','b',{'bilinear','nonlinear'},[0 1],1);
-options.two_state  = spm_input('states per region', '+1','b',{'one','two'},[0 1],1);
-options.stochastic = spm_input('stochastic effects','+1','b',{'no','yes'},[0 1],1);
-
+if nc                                                     % there are inputs
+    spm_input('Model options:...  ',-1,'d');
+    options.nonlinear  = spm_input('modulatory effects','+1','b',{'bilinear','nonlinear'},[0 1],1);
+    options.two_state  = spm_input('states per region', '+1','b',{'one','two'},[0 1],1);
+    options.stochastic = spm_input('stochastic effects','+1','b',{'no','yes'},[0 1],1);
+    options.centre     = spm_input('centre input',      '+1','b',{'no','yes'},[0 1],1);
+    options.endogenous = 0;
+else
+    options.nonlinear  = 0;
+    options.two_state  = 0;
+    options.stochastic = 1;
+    options.centre     = 1;
+    options.endogenous = 1;
+end
 
 %==========================================================================
 % Graph connections
 %==========================================================================
-n     = size(U.u,2);
 a     = zeros(m,m);
-b     = zeros(m,m,n);
-c     = zeros(m,n);
+if options.endogenous
+    b     = zeros(m,m,1);
+    c     = zeros(m,1);
+else
+    b     = zeros(m,m,nc);
+    c     = zeros(m,nc);
+end
 d     = zeros(m,m,0);
 
 %-Intrinsic connections (A matrix)
@@ -151,10 +171,14 @@
             set(h3(i,j),'Value',1,...
                 'enable','off');
         else
-            set(h3(i,j),'TooltipString', ...
+            set(h3(i,j),'enable','on','TooltipString', ...
                 sprintf('from %s to %s',xY(j).name,xY(i).name));
         end
-
+        if nc && i~=j
+            set(h3(i,j),'Value',0);
+        else
+            set(h3(i,j),'Value',1);
+        end
     end
 end
 uicontrol(Finter,'String','done','Position', [300 100 060 020].*WS,...
@@ -176,7 +200,7 @@
 %==========================================================================
 uicontrol(Finter,'String','done','Position', [300 100 060 020].*WS,...
     'Callback', 'uiresume(gcbf)');
-for k = 1:n
+for k = 1:nc
 
     %-Buttons and labels
     %----------------------------------------------------------------------
@@ -198,25 +222,27 @@
     end
     for i = 1:m
         for j = 1:m
-            if a(i,j)==1
-                % If there is an intrinsic connection
-                % allow it to be modulated
+            if a(i,j) == 1
+
+                % Allow modulation of intrinsic connections
+                %----------------------------------------------------------
                 h3(i,j) = uicontrol(Finter,...
                     'Position',[220+dx*j 360-dx*i 020 020].*WS,...
                     'BackgroundColor',bcolor,...
                     'Style','radiobutton');
                 set(h3(i,j),'TooltipString', ...
                     sprintf('from %s to %s',xY(j).name,xY(i).name));
+
             end
         end
     end
-    
+
     uiwait(Finter);
-    
+
     %-Get c
     %----------------------------------------------------------------------
     for i = 1:m
-        c(i,k)   = get(h2(i)  ,'Value');
+        c(i,k)   = get(h2(i),'Value');
     end
 
     %-Get b allowing any 2nd order effects
@@ -229,7 +255,7 @@
         end
     end
     delete([h1(:); h2(:); h3(a==1)])
-    
+
 end
 delete(findobj(get(Finter,'Children'),'flat'));
 
@@ -249,8 +275,9 @@
         for i = 1:m
             for j = 1:m
                 if a(i,j)==1
-                    % If there is an intrinsic connection
-                    % allow it to be modulated
+
+                    % Allow modulation of intrinsic connections
+                    %------------------------------------------------------
                     h4(i,j) = uicontrol(Finter,...
                         'Position',[220+dx*j 360-dx*i 020 020].*WS,...
                         'BackgroundColor',bcolor,...
@@ -303,6 +330,12 @@
 % DCM structure
 %==========================================================================
 
+% Endogenous input specification
+if isempty(U)
+    U.u    = zeros(v,1);
+    U.name = {'null'};
+end
+
 %-Store all variables in DCM structure
 %--------------------------------------------------------------------------
 DCM.a       = a;
@@ -320,8 +353,8 @@
 
 %-Save
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
-    save(fullfile(swd,['DCM_' name]),'-V6','DCM');
+if spm_check_version('matlab','7') >= 0
+    save(fullfile(swd,['DCM_' name '.mat']),'-V6','DCM');
 else
-    save(fullfile(swd,['DCM_' name]),'DCM');
+    save(fullfile(swd,['DCM_' name '.mat']),'DCM');
 end
diff --git a/spm_dcm_ui.m b/spm_dcm_ui.m
index f9f414c..9f89098 100644
--- a/spm_dcm_ui.m
+++ b/spm_dcm_ui.m
@@ -65,7 +65,7 @@ function spm_dcm_ui(Action)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_ui.m 3958 2010-06-30 16:24:46Z guillaume $
+% $Id: spm_dcm_ui.m 4267 2011-03-29 15:45:26Z guillaume $
 
 
 % Get figure handles
@@ -83,8 +83,13 @@ function spm_dcm_ui(Action)
 % Options, using pull-down menu
 %--------------------------------------------------------------------------
 if ~nargin
-    str       = 'Action: ';
-    Actions   = {'specify','estimate','review','compare','average','quit'};
+    str       =  'Action: ';
+    Actions   = {'specify',  ...
+                 'estimate', ...
+                 'review',   ...
+                 'compare',  ...
+                 'average',  ...
+                 'quit'};
     selected = spm_input(str,1,'m',Actions);
     Action   = Actions{selected};
 end
@@ -135,8 +140,8 @@ function spm_dcm_ui(Action)
     spm('FnBanner','spm_dcm_review');
     
     spm_dcm_review;
-
-
+    
+    
 %==========================================================================
 % Compare different models
 %==========================================================================
diff --git a/spm_dcm_voi.m b/spm_dcm_voi.m
index d574707..da71956 100644
--- a/spm_dcm_voi.m
+++ b/spm_dcm_voi.m
@@ -15,7 +15,7 @@ function spm_dcm_voi (DCM_filename,voi_filenames)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_dcm_voi.m 3521 2009-10-29 18:37:41Z guillaume $
+% $Id: spm_dcm_voi.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 load(DCM_filename);
@@ -44,7 +44,7 @@ function spm_dcm_voi (DCM_filename,voi_filenames)
 
 % Save (overwrite) new DCM file
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
+if spm_check_version('matlab','7') >= 0
     save(DCM_filename, 'DCM', '-V6');
 else
     save(DCM_filename, 'DCM');
diff --git a/spm_defaults.m b/spm_defaults.m
index cb7b979..a7e491a 100644
--- a/spm_defaults.m
+++ b/spm_defaults.m
@@ -17,7 +17,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % 
-% $Id: spm_defaults.m 3953 2010-06-28 16:58:48Z guillaume $
+% $Id: spm_defaults.m 4152 2011-01-11 14:13:35Z volkmar $
 
 %-Prevent users from making direct calls to this function
 %-----------------------------------------------------------------------
@@ -60,13 +60,13 @@
 
 % Stats defaults
 %=======================================================================
-defaults.stats.maxmem      = 2^26;
-defaults.stats.maxres      = 64;
-defaults.stats.fmri.ufp    = 0.001;  % Upper tail F-probability
-defaults.stats.pet.ufp     = 0.05;
-defaults.stats.eeg.ufp     = 1;
-defaults.stats.topoFDR     = 1;
-defaults.stats.rft.nonstat = 0;
+defaults.stats.maxmem       = 2^26;
+defaults.stats.maxres       = 64;
+defaults.stats.fmri.ufp     = 0.001;  % Upper tail F-probability
+defaults.stats.pet.ufp      = 0.05;
+defaults.stats.eeg.ufp      = 1;
+defaults.stats.topoFDR      = 1;
+defaults.stats.rft.nonstat  = 0;
 
 % Mask defaults
 %=======================================================================
@@ -91,7 +91,6 @@
 % Realignment defaults
 %=======================================================================
 defaults.realign.estimate.quality = 0.9;
-defaults.realign.estimate.weight  = {''};
 defaults.realign.estimate.interp  = 2;
 defaults.realign.estimate.wrap    = [0 0 0];
 defaults.realign.estimate.sep     = 4;
@@ -139,7 +138,6 @@
 defaults.normalise.estimate.cutoff  = 25;
 defaults.normalise.estimate.nits    = 16;
 defaults.normalise.estimate.reg     = 1;
-defaults.normalise.estimate.wtsrc   = 0;
 defaults.normalise.write.preserve   = 0;
 defaults.normalise.write.bb         = [[-78 -112 -50];[78 76 85]];
 defaults.normalise.write.vox        = [2 2 2];
@@ -158,6 +156,7 @@
 defaults.preproc.biasreg        = 0.0001;     % Bias regularisation
 defaults.preproc.biasfwhm       = 60;         % Bias FWHM
 defaults.preproc.regtype        = 'mni';      % Affine Regularisation
+defaults.preproc.fudge          = 5;          % Fudge factor, can not be set in batch GUI
 defaults.preproc.samp           = 3;          % Sampling distance
 defaults.preproc.output.GM      = [0 0 1];
 defaults.preproc.output.WM      = [0 0 1];
diff --git a/spm_defs.m b/spm_defs.m
index df0da03..134ef21 100644
--- a/spm_defs.m
+++ b/spm_defs.m
@@ -11,7 +11,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_defs.m 3755 2010-03-05 14:14:36Z volkmar $
+% $Id: spm_defs.m 4194 2011-02-05 18:08:06Z ged $
 
 [Def,mat] = get_comp(job.comp);
 [dpath ipath] = get_paths(job);
@@ -81,7 +81,7 @@
 sn  = load(job.matname{1});
 
 if any(isfinite(bb(:))) || any(isfinite(vox)),
-    [bb0,vox0] = bbvox_from_V(sn.VG(1));
+    [bb0,vox0] = spm_get_bbox(sn.VG(1), 'old');
 
     if any(~isfinite(vox)), vox = vox0; end;
     if any(~isfinite(bb)),  bb  = bb0;  end;
@@ -165,17 +165,6 @@
 end;
 %_______________________________________________________________________
 
-%_______________________________________________________________________
-function [bb,vx] = bbvox_from_V(V)
-% Return the default bounding box for an image volume
-
-vx = sqrt(sum(V.mat(1:3,1:3).^2));
-o  = V.mat\[0 0 0 1]';
-o  = o(1:3)';
-bb = [-vx.*(o-1) ; vx.*(V.dim(1:3)-o)];
-return;
-%_______________________________________________________________________
-
 %_______________________________________________________________________
 function [Def,mat] = get_def(job)
 % Load a deformation field saved as an image
@@ -301,7 +290,7 @@
 
 fname = {fullfile(odir,['y_' ofname '.nii'])};
 dim   = [size(Def{1},1) size(Def{1},2) size(Def{1},3) 1 3];
-dtype = 'FLOAT32-LE';
+dtype = 'FLOAT32';
 off   = 0;
 scale = 1;
 inter = 0;
@@ -333,7 +322,7 @@
 for i=1:size(fnames,1),
     V = spm_vol(fnames(i,:));
     M = inv(V.mat);
-    [pth,nam,ext] = spm_fileparts(fnames(i,:));
+    [pth,nam,ext,num] = spm_fileparts(deblank(fnames(i,:)));
     if isempty(odir)
         % use same path as source image
         opth = pth;
@@ -349,6 +338,7 @@
                 'mat',mat,...
                 'n',V.n,...
                 'descrip',V.descrip);
+    ofnames{i} = [ofnames{i} num];
     C  = spm_bsplinc(V,intrp);
     Vo = spm_create_vol(Vo);
     for j=1:size(Def{1},3)
diff --git a/spm_design_flexible.m b/spm_design_flexible.m
index 7fe0331..a685815 100644
--- a/spm_design_flexible.m
+++ b/spm_design_flexible.m
@@ -1,4 +1,4 @@
-function [H,Hnames] = spm_design_flexible(fblock,I)
+function [H,Hnames,B,Bnames] = spm_design_flexible(fblock,I)
 % Create H partition of design matrix
 % FORMAT [H,Hnames] = spm_design_flexible(fblock,I)
 %
@@ -7,45 +7,52 @@
 %
 % H        - Component of design matrix describing conditions
 % Hnames   - Condition names
+% B        - Component of design matrix describing blocks
+% Bnames   - Block names
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_design_flexible.m 3860 2010-05-04 15:59:25Z guillaume $
+% $Id: spm_design_flexible.m 4111 2010-11-05 12:59:10Z guillaume $
 
-% Sort main effects and interactions
+%-Sort main effects and interactions
 %--------------------------------------------------------------------------
 fmain = struct('fnum',{});
 inter = struct('fnums',{});
 for k=1:numel(fblock.maininters)
     if isfield(fblock.maininters{k},'fmain')
-        fmain(end+1)=fblock.maininters{k}.fmain;
+        fmain(end+1) = fblock.maininters{k}.fmain;
     elseif isfield(fblock.maininters{k},'inter')
-        inter(end+1)=fblock.maininters{k}.inter;
+        inter(end+1) = fblock.maininters{k}.inter;
     end
 end
 
-% Create main effects
+H = []; Hnames = [];
+B = []; Bnames = [];
+
+%-Create main effects
 %--------------------------------------------------------------------------
-H=[];Hnames=[];
-nmain=length(fmain);
-for f=1:nmain
-    fcol=fmain(f).fnum;
-    fname=fblock.fac(fcol).name;
+for f=1:length(fmain)
+    fcol  = fmain(f).fnum;
+    fname = fblock.fac(fcol).name;
     
     % Augment H partition - explicit factor numbers are 1 lower than in I matrix
-    [Hf,Hfnames]=spm_DesMtx(I(:,fcol+1),'-',fname);
-    H=[H,Hf];
-    Hnames=[Hnames;Hfnames];
+    [Hf,Hfnames] = spm_DesMtx(I(:,fcol+1),'-',fname);
+    if strcmpi(fname,'subject')
+        B        = [B,Hf];
+        Bnames   = [Bnames; Hfnames];
+    else
+        H        = [H,Hf];
+        Hnames   = [Hnames; Hfnames];
+    end
 end
 
-% Create interactions
+%-Create interactions
 %--------------------------------------------------------------------------
-ni=length(inter);
-for i=1:ni
+for i=1:length(inter)
     % Get the two factors for this interaction
-    fnums=inter(i).fnums;
-    f1=fnums(1);f2=fnums(2);
+    f1 = inter(i).fnums(1);
+    f2 = inter(i).fnums(2);
     
     % Names
     iname{1}     = fblock.fac(f1).name;
@@ -55,10 +62,9 @@
     Isub         = [I(:,f1+1),I(:,f2+1)];
     [Hf,Hfnames] = spm_DesMtx(Isub,'-',iname);
     H            = [H,Hf];
-    Hnames       = [Hnames;Hfnames];
-    
+    Hnames       = [Hnames; Hfnames];
 end
 
-if nmain==0 && ni==0
+if isempty(H)
     error('You have not specified any main effects or interactions');
 end
diff --git a/spm_design_within_subject.m b/spm_design_within_subject.m
index d75fcce..8773ce8 100644
--- a/spm_design_within_subject.m
+++ b/spm_design_within_subject.m
@@ -8,25 +8,23 @@
 % I        - Nscan x 4 factor matrix
 % P        - List of scans
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_design_within_subject.m 3860 2010-05-04 15:59:25Z guillaume $
+% $Id: spm_design_within_subject.m 4178 2011-01-27 15:12:53Z guillaume $
 
-nsub=length(fblock.fsuball.fsubject);
-% Specify design subject-by-subject
-P=[];I=[];
-subj=[];
-for s=1:nsub,
+%-Specify design subject-by-subject
+%--------------------------------------------------------------------------
+[P, I, subj] = deal([]);
+for s = 1:numel(fblock.fsuball.fsubject)
     P  = [P; fblock.fsuball.fsubject(s).scans];
     ns = length(fblock.fsuball.fsubject(s).scans);
     cc = fblock.fsuball.fsubject(s).conds;
     
     [ccr,ccc] = size(cc);
-    if ~(ccr==ns) && ~(ccc==ns)
-        disp(sprintf('Error for subject %d: conditions not specified for each scan',s));
-        return
-    elseif ~(ccr==ccc) && (ccc==ns)
+    if ccr~=ns && ccc~=ns
+        error('Error for subject %d: conditions not specified for each scan',s);
+    elseif ccr~=ccc && ccc==ns
         %warning('spm:transposingConditions',['Condition matrix ',...
         %    'appears to be transposed. Transposing back to fix.\n',...
         %    'Alert developers if it is not actually transposed.'])
@@ -38,40 +36,36 @@
     repl = zeros(ns, 1);
     for k=1:max(ccj)
         repl(ccj==k) = 1:sum(ccj==k);
-    end;
+    end
     I = [I; [repl cc]];
 end
 
-nf=length(fblock.fac);
-subject_factor=0;
-for i=1:nf,
+%-Handle keyword factors 'repl' and 'subject'
+%--------------------------------------------------------------------------
+nf = length(fblock.fac);
+subject_factor = false;
+for i=1:nf
     if strcmpi(fblock.fac(i).name,'repl')
-        % Copy `replications' column to create explicit `replications' factor
-        nI=I(:,1:i);
-        nI=[nI,I(:,1)];
-        nI=[nI,I(:,i+1:end)];
-        I=nI;
+        % Copy 'replications' column to create explicit 'replications' factor
+        I = [I(:,1:i) I(:,1) I(:,i+1:end)];
     end
     if strcmpi(fblock.fac(i).name,'subject')
-        % Create explicit `subject' factor
-        nI=I(:,1:i);
-        nI=[nI,subj];
-        nI=[nI,I(:,i+1:end)];
-        I=nI;
-        subject_factor=1;
-    end
-    
+        % Create explicit 'subject' factor
+        I = [I(:,1:i) subj I(:,i+1:end)];
+        subject_factor = true;
+    end  
 end
 
-% Re-order scans conditions and covariates into standard format
+%-Re-order scans conditions and covariates into standard format
 % This is to ensure compatibility with how variance components are created
+%--------------------------------------------------------------------------
 if subject_factor
     U=unique(I(:,2:nf+1),'rows');
     Un=length(U);
     Uc=zeros(Un,1);
     r=1;rj=[];
-    for k=1:Un,
-        for j=1:size(I,1),
+    for k=1:Un
+        for j=1:size(I,1)
             match=sum(I(j,2:nf+1)==U(k,:))==nf;
             if match
                 Uc(k)=Uc(k)+1;
@@ -85,11 +79,9 @@
     I=Ir;    % -conditions
     for k=1:numel(cov) % -covariates
         cov(k).c = cov(k).c(rj);
-    end;
+    end
 end
 
-% Pad out factorial matrix to cover the four canonical factors
-[ns,nI]=size(I);
-if nI < 4
-    I = [I, ones(ns,4-nI)];
-end
+%-Pad out factorial matrix to cover the four canonical factors
+%--------------------------------------------------------------------------
+I = [I ones(size(I,1),4-size(I,2))];
diff --git a/spm_detrend.m b/spm_detrend.m
index 132a516..963b074 100644
--- a/spm_detrend.m
+++ b/spm_detrend.m
@@ -1,19 +1,19 @@
-function [y] = spm_detrend(x,p)
-% polynomial detrending over columns
-% FORMAT [y] = spm_detrend(x,p)
-%___________________________________________________________________________
-%
-% spm_detrend removes linear and nonlinear trends from column-wise data
-% matrices
+function y = spm_detrend(x,p)
+% Polynomial detrending over columns
+% FORMAT y = spm_detrend(x,p)
 % x   - data matrix
-% p   - order of polynomial
+% p   - order of polynomial [default: 0]
+% 
 % y   - detrended data matrix
+%__________________________________________________________________________
 %
+% spm_detrend removes linear and nonlinear trends from column-wise data
+% matrices.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_detrend.m 3717 2010-02-08 16:44:42Z guillaume $
+% $Id: spm_detrend.m 4299 2011-04-08 12:13:18Z guillaume $
 
 
 % defaults
diff --git a/spm_dftmtx.m b/spm_dftmtx.m
new file mode 100644
index 0000000..05181ee
--- /dev/null
+++ b/spm_dftmtx.m
@@ -0,0 +1,30 @@
+function C = spm_dftmtx(N,K,a)
+% Creates basis functions for Discrete Cosine Transform.
+% FORMAT C = spm_dftmtx(N,K,a)
+%
+% N - dimension
+% K - order
+% a - number of (1/2)5Hz frequency steps (default a = 2)
+%__________________________________________________________________________
+% spm_dftmtx creates a matrix for the first few basis functions of a one
+% dimensional discrete Fourier transform.
+%
+% See:    Fundamentals of Digital Image Processing (p 150-154).
+%         Anil K. Jain 1989.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dftmtx.m 4170 2011-01-24 18:37:42Z karl $
+ 
+ 
+% initialise
+%--------------------------------------------------------------------------
+try, a; catch, a = 2; end
+ 
+C     = ones(N,1);
+x     = (0:(N - 1))/N;
+for k = 1:K
+    C(:,end + 1) = sin(a*pi*k*x);
+    C(:,end + 1) = cos(a*pi*k*x);
+end
diff --git a/spm_dicom_convert.m b/spm_dicom_convert.m
index c2a9e31..6dd85e4 100644
--- a/spm_dicom_convert.m
+++ b/spm_dicom_convert.m
@@ -7,9 +7,10 @@
 %        'all'      - all DICOM files [default]
 %        'mosaic'   - the mosaic images
 %        'standard' - standard DICOM files
-%        'spect'    - SIEMENS Spectroscopy DICOMs (position only)
-%                     This will write out a mask image volume with 1's
-%                     set at the position of spectroscopy voxel(s).
+%        'spect'    - SIEMENS Spectroscopy DICOMs (some formats only)
+%                     This will write out a 5D NIFTI containing real and
+%                     imaginary part of the spectroscopy time points at the
+%                     position of spectroscopy voxel(s).
 %        'raw'      - convert raw FIDs (not implemented)
 % root_dir - 'flat' - do not produce file tree [default]
 %            With all other options, files will be sorted into
@@ -17,7 +18,9 @@
 %            'date_time'  - Place files under ./<StudyDate-StudyTime>
 %            'patid'      - Place files under ./<PatID>
 %            'patid_date' - Place files under ./<PatID-StudyDate>
-%            'name'       - Place files under ./<PatName>
+%            'patname'    - Place files under ./<PatName>
+%            'series'     - Place files in series folders, without
+%                           creating patient folders
 % format - output format
 %          'img' Two file (hdr+img) NIfTI format [default]
 %          'nii' Single file NIfTI format
@@ -31,7 +34,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner & Jesper Andersson
-% $Id: spm_dicom_convert.m 3934 2010-06-17 14:58:25Z guillaume $
+% $Id: spm_dicom_convert.m 4213 2011-02-23 19:18:30Z john $
 
 
 if nargin<2, opts     = 'all'; end
@@ -55,7 +58,7 @@
     fspe = convert_spectroscopy(spect,root_dir,format);
 end;
 
-out.files = {fmos{:} fstd{:} fspe{:}};
+out.files = [fmos fstd fspe];
 if isempty(out.files)
     out.files = {''};
 end;
@@ -108,7 +111,7 @@
         volume(:,:,j) = img;
     end;
     dim = size(volume);
-    dt  = [spm_type('int16') spm_platform('bigend')];
+    dt  = determine_datatype(hdr{1});
 
     % Orientation information
     %-------------------------------------------------------------------
@@ -484,7 +487,7 @@
 nr = hdr{1}.Rows;
 
 dim    = [nc nr length(hdr)];
-dt     = [spm_type('int16') spm_platform('bigend')];
+dt     = determine_datatype(hdr{1});
 
 % Orientation information
 %-------------------------------------------------------------------
@@ -551,29 +554,64 @@
 %-------------------------------------------------------------------
 spm_progress_bar('Init',length(hdr),['Writing ' fname], 'Planes written');
 N      = nifti;
-pinfo  = [1 0];
-if isfield(hdr{1},'RescaleSlope'),      pinfo(1) = hdr{1}.RescaleSlope;     end;
-if isfield(hdr{1},'RescaleIntercept'),  pinfo(2) = hdr{1}.RescaleIntercept; end;
-N.dat  = file_array(fname,dim,dt,0,pinfo(1),pinfo(2));
-N.mat  = mat;
-N.mat0 = mat;
-N.mat_intent  = 'Scanner';
-N.mat0_intent = 'Scanner';
-N.descrip     = descrip;
-create(N);
-volume = zeros(dim);
+pinfos = [ones(length(hdr),1) zeros(length(hdr),1)];
+for i=1:length(hdr)
+    if isfield(hdr{i},'RescaleSlope'),     pinfos(i,1) = hdr{i}.RescaleSlope;     end 
+    if isfield(hdr{i},'RescaleIntercept'), pinfos(i,2) = hdr{i}.RescaleIntercept; end
+end
 
+if any(any(diff(pinfos,1))),
+    % Ensure random numbers are reproducible (see later)
+    % when intensities are dithered to prevent aliasing effects.
+    rand('state',0);
+end
+
+volume = zeros(dim);
 for i=1:length(hdr),
     plane = read_image_data(hdr{i});
 
-    if pinfo(1)~=1, plane = plane*pinfo(1); end;
-    if pinfo(2)~=0, plane = plane+pinfo(2); end;
+    if any(any(diff(pinfos,1))),
+        % This is to prevent aliasing effects in any subsequent histograms
+        % of the data (eg for mutual information coregistration).
+        % It's a bit inelegant, but probably necessary for when slices are
+        % individually rescaled.
+        plane = double(plane) + rand(size(plane)) - 0.5;
+    end
+
+    if pinfos(i,1)~=1, plane = plane*pinfos(i,1); end;
+    if pinfos(i,2)~=0, plane = plane+pinfos(i,2); end;
 
     plane = fliplr(plane);
     if ~true, plane = flipud(plane); end; % LEFT-HANDED STORAGE
     volume(:,:,i) = plane;
     spm_progress_bar('Set',i);
 end;
+
+if ~any(any(diff(pinfos,1))),
+    % Same slopes and intercepts for all slices
+    pinfo = pinfos(1,:);
+else
+    % Variable slopes and intercept (maybe PET/SPECT)
+    mx = max(volume(:));
+    mn = min(volume(:));
+
+    %%  Slope and Intercept
+    %%  32767*pinfo(1) + pinfo(2) = mx
+    %% -32768*pinfo(1) + pinfo(2) = mn
+    % pinfo = ([32767 1; -32768 1]\[mx; mn])';
+
+    % Slope only
+    dt    = 'int16-be';
+    pinfo = [max(mx/32767,-mn/32768) 0];
+end
+
+N.dat  = file_array(fname,dim,dt,0,pinfo(1),pinfo(2));
+N.mat  = mat;
+N.mat0 = mat;
+N.mat_intent  = 'Scanner';
+N.mat0_intent = 'Scanner';
+N.descrip     = descrip;
+create(N);
 N.dat(:,:,:) = volume;
 spm_progress_bar('Clear');
 return;
@@ -609,9 +647,12 @@
 %-------------------------------------------------------------------
 nc = get_numaris4_numval(privdat,'Columns');
 nr = get_numaris4_numval(privdat,'Rows');
+% Guess number of timepoints in file - don't know whether this should be
+% 'DataPointRows'-by-'DataPointColumns' or 'SpectroscopyAcquisitionDataColumns'
+ntp = get_numaris4_numval(privdat,'DataPointRows')*get_numaris4_numval(privdat,'DataPointColumns');
 
-dim    = [nc nr numel(hdr)];
-dt     = [spm_type('int16') spm_platform('bigend')];
+dim    = [nc nr numel(hdr) 2 ntp];
+dt     = spm_type('float32'); % Fixed datatype
 
 % Orientation information
 %-------------------------------------------------------------------
@@ -635,18 +676,38 @@
 shift_vx         = [eye(4,3) [.5; .5; 0; 1]];
 
 orient           = reshape(get_numaris4_numval(privdat,...
-    'ImageOrientationPatient'),[3 2]);
-try
-    ps(1) = get_numaris4_numval(privdat,...
-        'VoiReadoutFoV');
-    ps(2) = get_numaris4_numval(privdat,...
-        'VoiPhaseFoV');
-catch
-    ps  = get_numaris4_numval(privdat,'PixelSpacing');
+                                               'ImageOrientationPatient'),[3 2]);
+ps               = get_numaris4_numval(privdat,'PixelSpacing');
+if nc*nr == 1
+    % Single Voxel Spectroscopy (based on the following information from SIEMENS)
+    %---------------------------------------------------------------
+    % NOTE: Internally the position vector of the CSI matrix shows to the outer border
+    % of the first voxel. Therefore the position vector has to be corrected.
+    % (Note: The convention of Siemens spectroscopy raw data is in contrast to the
+    %  DICOM standard where the position vector points to the center of the first voxel.)
+    %---------------------------------------------------------------
+    % SIEMENS decides which definition to use based on the contents of the
+    % 'PixelSpacing' internal header field. If it has non-zero values,
+    % assume DICOM convention. If any value is zero, assume SIEMENS
+    % internal convention for this direction.
+    % Note that in SIEMENS code, there is a shift when PixelSpacing is
+    % zero. Here, the shift seems to be necessary when PixelSpacing is
+    % non-zero. This may indicate more fundamental problems with
+    % orientation decoding.
+    if ps(1) == 0 % row
+        ps(1) = get_numaris4_numval(privdat,...
+                                    'VoiPhaseFoV');
+        shift_vx(1,4) = 0;
+    end
+    if ps(2) == 0 % col
+        ps(2) = get_numaris4_numval(privdat,...
+                                    'VoiReadoutFoV');
+        shift_vx(2,4) = 0;
+    end
 end
 pos = get_numaris4_numval(privdat,'ImagePositionPatient');
-
-R  = [orient*diag(ps); 0 0];
+% for some reason, pixel spacing needs to be swapped
+R  = [orient*diag(ps([2 1])); 0 0];
 x1 = [1;1;1;1];
 y1 = [pos; 1];
 
@@ -671,7 +732,6 @@
     y2 = [orient*[0;0;z];0];
 end
 dicom_to_patient = [y1 y2 R]/[x1 x2 eye(4,2)];
-warning('Don''t know exactly what positions in spectroscopy files should be - just guessing!')
 mat              = patient_to_tal*dicom_to_patient*shift_vx*analyze_to_dicom;
 
 % Possibly useful information
@@ -695,7 +755,6 @@
 
 % Write the image volume
 %-------------------------------------------------------------------
-spm_progress_bar('Init',length(hdr),['Writing ' fname], 'Planes written');
 N      = nifti;
 pinfo  = [1 0];
 if isfield(hdr{1},'RescaleSlope'),      pinfo(1) = hdr{1}.RescaleSlope;     end;
@@ -706,20 +765,18 @@
 N.mat_intent  = 'Scanner';
 N.mat0_intent = 'Scanner';
 N.descrip     = descrip;
+N.extras      = struct('MagneticFieldStrength',...
+                       get_numaris4_numval(privdat,'MagneticFieldStrength'),...
+                       'TransmitterReferenceAmplitude',...
+                       get_numaris4_numval(privdat,'TransmitterReferenceAmplitude'));
 create(N);
-volume = zeros(dim);
 
-for i=1:length(hdr),
-    plane = read_spect_data(hdr{i},privdat);
-    if pinfo(1)~=1, plane = plane*pinfo(1); end;
-    if pinfo(2)~=0, plane = plane+pinfo(2); end;
-    plane = fliplr(plane);
-    if ~true, plane = flipud(plane); end; % LEFT-HANDED STORAGE
-    volume(:,:,i) = plane;
-    spm_progress_bar('Set',i);
-end;
-N.dat(:,:,:) = volume;
-spm_progress_bar('Clear');
+% Read data, swap dimensions
+data = permute(reshape(read_spect_data(hdr{1},privdat),dim([4 5 1 2 3])), ...
+                [3 4 5 1 2]);
+% plane = fliplr(plane);
+
+N.dat(:,:,:,:,:) = data;
 return;
 %_______________________________________________________________________
 
@@ -730,7 +787,11 @@
 for i=1:length(hdr),
     if ~checkfields(hdr{i},'Modality') || ~(strcmp(hdr{i}.Modality,'MR') ||...
             strcmp(hdr{i}.Modality,'PT') || strcmp(hdr{i}.Modality,'CT'))
-        disp(['Cant find appropriate modality information for "' hdr{i}.Filename '".']);
+        if checkfields(hdr{i},'Modality'),
+            fprintf('File "%s" can not be converted because it is of type "%s", which is not MRI, CT or PET.\n', hdr{i}.Filename, hdr{i}.Modality);
+        else
+            fprintf('File "%s" can not be converted because it does not encode an image.\n', hdr{i}.Filename);
+        end
         guff = [guff(:)',hdr(i)];
     elseif ~checkfields(hdr{i},'StartOfPixelData','SamplesperPixel',...
             'Rows','Columns','BitsAllocated','BitsStored','HighBit','PixelRepresentation'),
@@ -742,7 +803,7 @@
    %    % No documentation about this private field is yet available.
    %    disp('Cant yet convert Phillips Intera DICOM.');
    %    guff = {guff{:},hdr{i}};
-    elseif ~(checkfields(hdr{i},'PixelSpacing','ImagePositionPatient','ImageOrientationPatient')||isfield(hdr{i},'Private_0029_1210')),
+    elseif ~(checkfields(hdr{i},'PixelSpacing','ImagePositionPatient','ImageOrientationPatient')||isfield(hdr{i},'Private_0029_1110')||isfield(hdr{i},'Private_0029_1210')),
         disp(['Cant find "Image Plane" information for "' hdr{i}.Filename '".']);
         guff = [guff(:)',hdr(i)];
     elseif ~checkfields(hdr{i},'PatientID','SeriesNumber','AcquisitionNumber','InstanceNumber'),
@@ -782,7 +843,7 @@
         % NumberOfImagesInMosaic seems to be set to zero for pseudo images
         % containing e.g. online-fMRI design matrices, don't treat them as
         % mosaics
-        standard = {standard{:},hdr{i}};
+        standard = {standard{:}, hdr{i}};
     else
         mosaic = {mosaic{:},hdr{i}};
     end;
@@ -889,11 +950,19 @@
 
 %_______________________________________________________________________
 function img = read_spect_data(hdr,privdat)
-% Image dimensions
-%-------------------------------------------------------------------
-nc = get_numaris4_numval(privdat,'Columns');
-nr = get_numaris4_numval(privdat,'Rows');
-img = ones(nr,nc);
+% Guess number of timepoints in file - don't know whether this should be
+% 'DataPointRows'-by-'DataPointColumns' or 'SpectroscopyAcquisitionDataColumns'
+ntp = get_numaris4_numval(privdat,'DataPointRows')*get_numaris4_numval(privdat,'DataPointColumns');
+% Data is stored as complex float32 values, timepoint by timepoint, voxel
+% by voxel. Reshaping is done in write_spectroscopy_volume.
+if ntp*2*4 ~= hdr.SizeOfCSAData
+    warning([hdr.Filename,': Data size mismatch.']);
+end
+fp = fopen(hdr.Filename,'r','ieee-le');
+fseek(fp,hdr.StartOfCSAData,'bof');
+img = fread(fp,2*ntp,'float32');
+fclose(fp);
+return;
 %_______________________________________________________________________
 
 %_______________________________________________________________________
@@ -1016,7 +1085,7 @@
 studydate = sprintf('%s_%s-%s', datestr(hdr.StudyDate,'yyyy-mm-dd'), ...
     h,m);
 switch root_dir
-    case 'date_time',
+    case {'date_time','series'}
     id = studydate;
     case {'patid', 'patid_date', 'patname'},
     id = strip_unwanted(hdr.PatientID);
@@ -1035,6 +1104,8 @@
     case 'patname',
         dname = fullfile(pwd, strip_unwanted(hdr.PatientsName), ...
             id, protname);
+    case 'series',
+        dname = fullfile(pwd, protname);
     otherwise
         error('unknown file root specification');
 end;
@@ -1081,7 +1152,9 @@
 % ### ASCCONV BEGIN ###
 % and ends with
 % ### ASCCONV END ###
-% It is read by spm_dicom_headers into an entry 'MrProtocol' in CSASeriesHeaderInfo
+% It is read by spm_dicom_headers into an entry 'MrProtocol' in
+% CSASeriesHeaderInfo or into an entry 'MrPhoenixProtocol' in
+% Private_0029_1110 or Private_0029_1120.
 % The additional items are assignments in C syntax, here they are just
 % translated according to
 % [] -> ()
@@ -1091,8 +1164,10 @@
 ret=struct;
 
 % get ascconv data
-if strcmp(hdr.CSAImageHeaderVersion, 'syngo MR B13')
-    X = hdr.Private_0029_1120;
+if isfield(hdr, 'Private_0029_1110')
+    X = get_numaris4_val(hdr.Private_0029_1110,'MrPhoenixProtocol');
+elseif isfield(hdr, 'Private_0029_1120')
+    X = get_numaris4_val(hdr.Private_0029_1120,'MrPhoenixProtocol');
 else
     X=get_numaris4_val(hdr.CSASeriesHeaderInfo,'MrProtocol');
 end
@@ -1103,10 +1178,10 @@
 if ~isempty(ascstart) && ~isempty(ascend)
     tokens = textscan(char(X((ascstart+22):(ascend-1))),'%s', ...
         'delimiter',char(10));
+    tokens{1}=regexprep(tokens{1},{'\[([0-9]*)\]','"(.*)"','0x([0-9a-fA-F]*)'},{'($1+1)','''$1''','hex2dec(''$1'')'});
+    % If everything would evaluate correctly, we could use
+    % eval(sprintf('ret.%s;\n',tokens{1}{:}));
     for k = 1:numel(tokens{1})
-        tokens{1}{k}=regexprep(tokens{1}{k},'\[([0-9]*)\]','($1+1)');
-        tokens{1}{k}=regexprep(tokens{1}{k},'"(.*)"','''$1''');
-        tokens{1}{k}=regexprep(tokens{1}{k},'0x([0-9a-fA-F]*)','hex2dec(''$1'')');
         try
             eval(['ret.' tokens{1}{k} ';']);
         catch
@@ -1115,3 +1190,22 @@
     end;
 end;
 %_______________________________________________________________________
+
+%_______________________________________________________________________
+function dt = determine_datatype(hdr)
+% Determine what datatype to use for NIfTI images
+be = spm_platform('bigend');
+if hdr.HighBit>16
+    if hdr.PixelRepresentation
+        dt  = [spm_type( 'int32') be];
+    else
+        dt  = [spm_type('uint32') be];
+    end
+else
+    if hdr.PixelRepresentation
+        dt  = [spm_type( 'int16') be];
+    else
+        dt  = [spm_type('uint16') be];
+    end
+end
+
diff --git a/spm_dicom_essentials.m b/spm_dicom_essentials.m
index ee700c3..ff6c174 100644
--- a/spm_dicom_essentials.m
+++ b/spm_dicom_essentials.m
@@ -12,7 +12,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_dicom_essentials.m 2268 2008-09-30 21:15:32Z guillaume $
+% $Id: spm_dicom_essentials.m 4044 2010-08-25 14:03:48Z volkmar $
 
 used_fields = {...
     'AcquisitionDate',...
@@ -38,6 +38,7 @@
     'PatientsName',...
     'PixelRepresentation',...
     'PixelSpacing',...
+    'Private_0029_1110',...
     'Private_0029_1210',...
     'ProtocolName',...
     'RepetitionTime',...
@@ -61,26 +62,34 @@
     'VROfPixelData'};
 
 fnames = fieldnames(hdr0);
-for i=1:numel(used_fields),
-    if any(strmatch(used_fields{i},fnames,'exact')),
+for i=1:numel(used_fields)
+    if ismember(used_fields{i},fnames)
        hdr1.(used_fields{i}) = hdr0.(used_fields{i});
     end
 end
 
-if isfield(hdr1,'Private_0029_1210'),
-    Private_0029_1210_fields = {...
-        'Columns',...
-        'Rows',...
-        'ImageOrientationPatient',...
-        'ImagePositionPatient',...
-        'SliceThickness',...
-        'PixelSpacing'};
+Private_spectroscopy_fields = {...
+    'Columns',...
+    'Rows',...
+    'ImageOrientationPatient',...
+    'ImagePositionPatient',...
+    'SliceThickness',...
+    'PixelSpacing',...
+    'VoiPhaseFoV',...
+    'VoiReadoutFoV',...
+    'VoiThickness'};
+if isfield(hdr1,'Private_0029_1110')
+    hdr1.Private_0029_1110 = ...
+        getfields(hdr1.Private_0029_1110,...
+                       Private_spectroscopy_fields);
+end
+if isfield(hdr1,'Private_0029_1210')
     hdr1.Private_0029_1210 = ...
         getfields(hdr1.Private_0029_1210,...
-                       Private_0029_1210_fields);
+                       Private_spectroscopy_fields);
 end
  
-if isfield(hdr1,'CSAImageHeaderInfo'), 
+if isfield(hdr1,'CSAImageHeaderInfo')
     CSAImageHeaderInfo_fields = {...
         'SliceNormalVector',...
         'NumberOfImagesInMosaic',...
@@ -91,7 +100,7 @@
                        CSAImageHeaderInfo_fields);
 end
 
-if isfield(hdr1,'CSASeriesHeaderInfo'),    
+if isfield(hdr1,'CSASeriesHeaderInfo')
     CSASeriesHeaderInfo_fields = {};
     hdr1.CSASeriesHeaderInfo = ...
         getfields(hdr1.CSASeriesHeaderInfo,...
@@ -102,7 +111,7 @@
 function str1 = getfields(str0,names)
 str1 = [];
 for i=1:numel(names)
-    for j=1:numel(str0),
+    for j=1:numel(str0)
         if strcmp(str0(j).name,names{i})
             str1 = [str1,str0(j)];
         end
diff --git a/spm_diff.m b/spm_diff.m
index a47c53e..b1ad837 100644
--- a/spm_diff.m
+++ b/spm_diff.m
@@ -24,7 +24,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_diff.m 3696 2010-01-22 14:22:31Z karl $
+% $Id: spm_diff.m 4060 2010-09-01 17:17:36Z karl $
  
 % create inline object
 %--------------------------------------------------------------------------
@@ -136,7 +136,7 @@
         fi    = spm_diff(f,xi{:},n(1:end - 1),V);
         J{i}  = spm_dfdx(fi,f0{1},dx);
     end
-    varargout = {J f0{:}};
+    varargout = [{J} f0];
 end
  
  
diff --git a/spm_dwtmtx.m b/spm_dwtmtx.m
new file mode 100644
index 0000000..54c2720
--- /dev/null
+++ b/spm_dwtmtx.m
@@ -0,0 +1,76 @@
+function H = spm_dwtmtx(N,K,T)
+% Creates basis functions for Discrete (Haar) Wavelet Transform.
+% FORMAT H = spm_dwtmtx(N,K,T)
+%
+%   N - dimension
+%   K - order: number of basis functions = N/K
+%
+%   T - option flag for thinning eccentric wavelets
+%__________________________________________________________________________
+% spm_dctmtx creates a matrix for the first few basis functions of a one
+% dimensional Haar Discrete Wavelet transform.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dwtmtx.m 4170 2011-01-24 18:37:42Z karl $
+ 
+ 
+% Create basis set
+%==========================================================================
+try, T; catch, T = 0; end
+ 
+K     = max(K,1); 
+H     = ones(N,1);
+I     = 1;
+while ~isempty(I)
+    
+    % Find non-zero elements and create new bases
+    %----------------------------------------------------------------------
+    j      = find(H(:,I(1)) > 0);
+    k      = find(H(:,I(1)) < 0);
+    nj     = length(j);
+    nk     = length(k);
+    hj     = sparse(1:fix(nj/2),1,2,nj,1) - 1;
+    hk     = sparse(1:fix(nk/2),1,2,nk,1) - 1;
+    
+    % Add new columns if there are enough elements
+    %----------------------------------------------------------------------
+    if nj > K
+        H(j,end + 1) = hj;
+        I(end + 1)   = size(H,2);
+    end
+    if nk > K
+        H(k,end + 1) = hk;
+        I(end + 1)   = size(H,2);
+    end
+ 
+    % This column has now been expanded
+    %----------------------------------------------------------------------
+    I(1) = [];
+    
+end
+ 
+% Thin eccentric basis functions
+%==========================================================================
+if T
+    
+    % indices of column (bases) to retain
+    %----------------------------------------------------------------------
+    j = logical(H(:,1));
+    for i = 1:length(j);
+        k = find(H(:,i));
+        l = length(k);
+        if any(k < (N/2 - 2*l)) || any(k > (N/2 + 2*l))
+            j(i) = 0;
+        end
+    end
+    
+    % thin
+    %----------------------------------------------------------------------
+    H  = H(:,j);
+end
+
+
+
+
diff --git a/spm_eeg_convert2scalp.m b/spm_eeg_convert2scalp.m
index d5229ea..774d626 100644
--- a/spm_eeg_convert2scalp.m
+++ b/spm_eeg_convert2scalp.m
@@ -29,9 +29,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: spm_eeg_convert2scalp.m 3538 2009-11-05 22:33:42Z vladimir $
+% $Id: spm_eeg_convert2scalp.m 4164 2011-01-14 14:42:06Z vladimir $
 
-SVNrev = '$Rev: 3538 $';
+SVNrev = '$Rev: 4164 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -130,10 +130,16 @@
 %-Make output directory for each dataset
 %--------------------------------------------------------------------------
 [P, F] = fileparts(Fname);
-if isempty(P), P = pwd; end
+%if isempty(P), P = pwd; end
 [sts, msg] = mkdir(P, F);
 if ~sts, error(msg); end
 P  = fullfile(P, F);
+if isfield(S, 'images') && isfield(S.images,'region_no')
+    F = sprintf('%dROI', S.images.region_no);
+    [sts, msg] = mkdir(P, F);
+    if ~sts, error(msg); end
+    P  = fullfile(P, F);
+end
 
 %-Loop over conditions
 %--------------------------------------------------------------------------
diff --git a/spm_eeg_crop.m b/spm_eeg_crop.m
new file mode 100644
index 0000000..ed08dd3
--- /dev/null
+++ b/spm_eeg_crop.m
@@ -0,0 +1,137 @@
+function D = spm_eeg_crop(S)
+% Reduce the data size by cutting in time and frequency.
+% FORMAT D = spm_eeg_crop(S)
+%
+% S        - optional input struct
+% (optional) fields of S:
+% D        - MEEG object or filename of M/EEG mat-file with epoched data
+% timewin  - time window to retain (in PST ms)
+% freqwin  - frequency window to retain
+% channels - cell array of channel labels or 'all'.
+%
+% Output:
+% D        - MEEG object (also written on disk)
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Vladimir Litvak
+% $Id: spm_eeg_crop.m 4202 2011-02-16 14:59:58Z vladimir $
+
+SVNrev = '$Rev: 4202 $';
+
+%-Startup
+%--------------------------------------------------------------------------
+spm('FnBanner', mfilename, SVNrev);
+spm('FigName','Crop M/EEG data'); spm('Pointer','Watch');
+
+%-Get MEEG object
+%--------------------------------------------------------------------------
+try
+    D = S.D;
+catch
+    [D, sts] = spm_select(1, 'mat', 'Select M/EEG mat file');
+    if ~sts, D = []; return; end
+    S.D = D;
+end
+
+D = spm_eeg_load(D);
+
+isTF = strncmpi(D.transformtype,'TF',2);
+
+if ~isfield(S, 'timewin')
+    S.timewin = spm_input('Time window (ms)', '+1', 'r', num2str(1000*[D.time(1) D.time(end)]), 2);
+end
+
+if  isTF && ~isfield(S, 'freqwin')
+    S.freqwin = spm_input('Frequency window (Hz)', '+1', 'r', num2str([D.frequencies(1) D.frequencies(end)]), 2);
+end
+
+timeind = D.indsample(1e-3*(min(S.timewin))):D.indsample(1e-3*(max(S.timewin)));
+if isempty(timeind) || any(isnan(timeind))
+    error('Selected time window is invalid.');
+end
+
+if isTF
+    freqind = D.indfrequency(min(S.freqwin)):D.indfrequency(max(S.freqwin));
+    if isempty(freqind) || any(isnan(freqind))
+        error('Selected frequency window is invalid.');
+    end
+end
+
+if D.nchannels > 1
+    if ~isfield(S, 'channels')
+        [selection, ok]= listdlg('ListString', D.chanlabels, 'SelectionMode', 'multiple' ,'Name', 'Select channels' , 'ListSize', [400 300]);
+        if ~ok
+            return;
+        end
+        
+        S.channels = D.chanlabels(selection);
+    end
+    
+    if isequal(S.channels, 'all')
+        chanind = 1:D.nchannels;
+    else
+        chanind = spm_match_str(D.chanlabels, S.channels);
+    end
+else
+    chanind = 1;
+end
+
+
+%-Generate new MEEG object with new files
+%--------------------------------------------------------------------------
+if isTF
+    Dnew = clone(D, ['p' fnamedat(D)], [length(chanind) length(freqind) length(timeind) D.ntrials]);
+    Dnew = frequencies(Dnew, [], D.frequencies(freqind));
+else
+    Dnew = clone(D, ['p' fnamedat(D)], [length(chanind) length(timeind) D.ntrials]);
+end
+
+Dnew = timeonset(Dnew, D.time(timeind(1)));
+
+Dnew = chanlabels(Dnew, [], D.chanlabels(chanind));
+Dnew = badchannels(Dnew, [], badchannels(D, chanind));
+Dnew = chantype(Dnew, [], chantype(D, chanind));
+Dnew = units(Dnew, [], units(D, chanind));
+Dnew = coor2D(Dnew, [], coor2D(D, chanind));
+
+if isequal(Dnew.type, 'continuous')
+   ev = Dnew.events;
+   ev = ev([ev.time]>=Dnew.time(1) & [ev.time]<=Dnew.time(end));
+   Dnew   = events(Dnew, ev);
+end
+
+%-Copy data
+%--------------------------------------------------------------------------
+spm_progress_bar('Init', D.ntrials, 'Trials copied');
+if D.ntrials > 100, Ibar = floor(linspace(1, D.ntrials, 100));
+else Ibar = 1:D.ntrials; end
+
+for i = 1:D.ntrials
+    
+    if isTF
+        Dnew(:, :, :, i) =  D(chanind, freqind, timeind, i);
+    else
+        Dnew(:, :, i) =  D(chanind, timeind, i);
+    end
+    
+    if D.trialonset(i) ~= 0
+        Dnew = trialonset(Dnew, i,  D.trialonset(i)+ D.time(timeind(1))-D.time(1));
+    end
+    
+    if ismember(i, Ibar), spm_progress_bar('Set', i); end
+end  %
+
+spm_progress_bar('Clear');
+
+%-Save the new M/EEG dataset
+%--------------------------------------------------------------------------
+Dnew = Dnew.history(mfilename, S);
+save(Dnew);
+
+D = Dnew;
+
+%-Cleanup
+%--------------------------------------------------------------------------
+spm('FigName','Crop M/EEG data: done'); spm('Pointer','Arrow');
diff --git a/spm_eeg_displayECD.m b/spm_eeg_displayECD.m
index 71d6f55..cdbf38b 100644
--- a/spm_eeg_displayECD.m
+++ b/spm_eeg_displayECD.m
@@ -20,19 +20,22 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jean Daunizeau
-% $Id: spm_eeg_displayECD.m 3682 2010-01-14 15:38:48Z guillaume $
+% $Id: spm_eeg_displayECD.m 4054 2010-08-27 19:27:09Z karl $
 
-hfig = [];
+
+% checks and defaults
+%--------------------------------------------------------------------------
+hfig       = [];
 ParentAxes = [];
-tag = '';
-query = [];
-handles = [];
-try, options; catch, options=[];end
-try, hfig = options.hfig;end
-try, tag = options.tag;end
-try, ParentAxes  = options.ParentAxes;end
-try, query = options.query;end
-try, handles = options.handles; end
+query      = [];
+handles    = [];
+tag        = '';
+try, options; catch, options = [];      end
+try, hfig        = options.hfig;        end
+try, tag         = options.tag;         end
+try, ParentAxes  = options.ParentAxes;  end
+try, query       = options.query;       end
+try, handles     = options.handles;     end
 try
     figure(hfig);
 catch
@@ -40,10 +43,21 @@
     spm_figure('Clear',hfig);
     ParentAxes = axes('parent',hfig);
 end
+try
+    markersize = options.markersize;
+catch
+    markersize = 20;
+end
+try
+    meshsurf = options.meshsurf;
+catch
+    meshsurf = fullfile(spm('Dir'),'canonical','cortex_5124.surf.gii');
+end
 
-try, Pos{1}; catch, Pos={Pos};end
-try, Orient{1}; catch, Orient={Orient};end
-try, Var{1}; catch, Var = {Var};end
+if isscalar(Var), Var = Pos*0 + Var^2;   end
+try, Pos{1};    catch, Pos = {Pos};      end
+try, Orient{1}; catch, Orient = {Orient};end
+try, Var{1};    catch, Var = {Var};      end
 
 ndip = size(Pos{1},2);
 if ~exist('Names','var') || isempty(Names)
@@ -52,24 +66,20 @@
     end
 end
 
+
 col = ['b','g','r','c','m','y','k','w'];
 tmp = ceil(ndip./numel(col));
 col = repmat(col,1,tmp);
+pa  = get(ParentAxes,'position');
 
 if ndip > 0
     
     if isempty(query)
-        m = fullfile(spm('Dir'),'canonical','cortex_5124.surf.gii');
         opt.hfig = hfig;
         opt.ParentAxes = ParentAxes;
         opt.visible = 'off';
-        try
-            pa = get(ParentAxes,'position');
-            pos2 = [pa(1),pa(2)+0.25*pa(4),0.03,0.5*pa(4)];
-        catch
-            pos2 = [0.1,0.55,0.03,0.4];
-        end
-        [out] = spm_eeg_render(m,opt);
+        pos2 = [pa(1),pa(2)+0.25*pa(4),0.03,0.5*pa(4)];
+        out  = spm_eeg_render(meshsurf,opt);
         handles.mesh = out.handles.p;
         handles.BUTTONS.transp = out.handles.transp;
         handles.hfig = out.handles.fi;
@@ -79,7 +89,7 @@
             'visible','on')
         set(handles.BUTTONS.transp,...
             'value',0.1,...
-           'position',pos2,...
+            'position',pos2,...
             'visible','on')
     end
     
@@ -95,33 +105,35 @@
                 handles.hp(j,i) = plot3(handles.ParentAxes,...
                     Pos{j}(1,i),Pos{j}(2,i),Pos{j}(3,i),...
                     [col(i),'.'],...
-                    'markerSize',20,...
+                    'markerSize',markersize,...
                     'visible','off');
             end
-            no = sqrt(sum(Orient{j}(:,i).^2));
-            if no > 0
-                Oi = 1e1.*Orient{j}(:,i)./no;
-            else
-                Oi = 1e-5*ones(3,1);
-            end
             try
-                set(handles.hq(j,i),...
-                    'xdata',Pos{j}(1,i),...
-                    'ydata',Pos{j}(2,i),...
-                    'zdata',Pos{j}(3,i),...
-                    'udata',Oi(1),...
-                    'vdata',Oi(2),...
-                    'wdata',Oi(3))
-            catch
-                handles.hq(j,i) = quiver3(handles.ParentAxes,...
-                    Pos{j}(1,i),Pos{j}(2,i),Pos{j}(3,i),...
-                    Oi(1),Oi(2),Oi(3),col(i),...
-                    'lineWidth',2,'visible','off');
-            end
-            if isequal(query,'add')
-                set(handles.hq(j,i),...
-                    'LineStyle','--',...
-                    'lineWidth',1)
+                no = sqrt(sum(Orient{j}(:,i).^2));
+                if no > 0
+                    Oi = 1e1.*Orient{j}(:,i)./no;
+                else
+                    Oi = 1e-5*ones(3,1);
+                end
+                try
+                    set(handles.hq(j,i),...
+                        'xdata',Pos{j}(1,i),...
+                        'ydata',Pos{j}(2,i),...
+                        'zdata',Pos{j}(3,i),...
+                        'udata',Oi(1),...
+                        'vdata',Oi(2),...
+                        'wdata',Oi(3))
+                catch
+                    handles.hq(j,i) = quiver3(handles.ParentAxes,...
+                        Pos{j}(1,i),Pos{j}(2,i),Pos{j}(3,i),...
+                        Oi(1),Oi(2),Oi(3),col(i),...
+                        'lineWidth',2,'visible','off');
+                end
+                if isequal(query,'add')
+                    set(handles.hq(j,i),...
+                        'LineStyle','--',...
+                        'lineWidth',1)
+                end
             end
             [x,y,z]= ellipsoid(Pos{j}(1,i),Pos{j}(2,i),Pos{j}(3,i),...
                 1.*sqrt(Var{j}(1,i)),1.*sqrt(Var{j}(2,i)),1.*sqrt(Var{j}(1,i)),20);
diff --git a/spm_eeg_downsample.m b/spm_eeg_downsample.m
index 06a55a8..8ebb2ea 100644
--- a/spm_eeg_downsample.m
+++ b/spm_eeg_downsample.m
@@ -6,6 +6,7 @@
 % (optional) fields of S:
 %   S.D           - MEEG object or filename of M/EEG mat-file
 %   S.fsample_new - new sampling rate, must be lower than the original one
+%   S.prefix      - prefix of generated file
 %
 % D               - MEEG object (also written on disk)
 %__________________________________________________________________________
@@ -17,9 +18,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: spm_eeg_downsample.m 3958 2010-06-30 16:24:46Z guillaume $
+% $Id: spm_eeg_downsample.m 4127 2010-11-19 18:05:18Z christophe $
 
-SVNrev = '$Rev: 3958 $';
+SVNrev = '$Rev: 4127 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -64,6 +65,12 @@
     S.fsample_new = fsample_new;
 end
 
+try
+    prefix = S.prefix;
+catch
+    prefix = 'd';
+end
+
 % This is to handle non-integer sampling rates up to a reasonable precision
 P = round(10*fsample_new);
 Q = round(10*D.fsample);
@@ -83,7 +90,7 @@
 
 %-Generate new meeg object with new filenames
 %--------------------------------------------------------------------------
-Dnew = clone(D, ['d' fnamedat(D)], [D.nchannels nsamples_new D.ntrials]);
+Dnew = clone(D, [prefix fnamedat(D)], [D.nchannels nsamples_new D.ntrials]);
 t0 = clock;
 
 %-Second pass: resample all
@@ -99,6 +106,7 @@
     datasz = nchannels(D)*nsamples(D)*8; % datapoints x 8 bytes per double value
     blknum = ceil(datasz/memsz);
     blksz  = ceil(nchannels(D)/blknum);
+    blknum = ceil(nchannels(D)/blksz);
     
     % now downsample blocks of channels
     chncnt=1;
diff --git a/spm_eeg_epochs.m b/spm_eeg_epochs.m
index be820df..8487233 100644
--- a/spm_eeg_epochs.m
+++ b/spm_eeg_epochs.m
@@ -39,9 +39,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: spm_eeg_epochs.m 3660 2010-01-04 19:11:24Z guillaume $
+% $Id: spm_eeg_epochs.m 4244 2011-03-14 13:33:01Z vladimir $
 
-SVNrev = '$Rev: 3660 $';
+SVNrev = '$Rev: 4244 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -134,6 +134,10 @@
 trl = epochinfo.trl;
 conditionlabels = epochinfo.conditionlabels;
 
+if numel(conditionlabels) == 1
+   conditionlabels = repmat(conditionlabels, 1, size(trl, 1));
+end
+
 try
     epochinfo.padding = S.epochinfo.padding;
 catch
diff --git a/spm_eeg_filter.m b/spm_eeg_filter.m
index 7c299ff..e7553ed 100644
--- a/spm_eeg_filter.m
+++ b/spm_eeg_filter.m
@@ -22,9 +22,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel
-% $Id: spm_eeg_filter.m 3958 2010-06-30 16:24:46Z guillaume $
+% $Id: spm_eeg_filter.m 4127 2010-11-19 18:05:18Z christophe $
 
-SVNrev = '$Rev: 3958 $';
+SVNrev = '$Rev: 4127 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -136,6 +136,7 @@
     datasz = nchannels(D)*nsamples(D)*8; % datapoints x 8 bytes per double value
     blknum = ceil(datasz/memsz);
     blksz  = ceil(nchannels(D)/blknum);
+    blknum = ceil(nchannels(D)/blksz);
 
     % now filter blocks of channels
     chncnt=1;
diff --git a/spm_eeg_ft2spm.m b/spm_eeg_ft2spm.m
index 5cd0c04..c2380ec 100644
--- a/spm_eeg_ft2spm.m
+++ b/spm_eeg_ft2spm.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_eeg_ft2spm.m 3833 2010-04-22 14:49:48Z vladimir $
+% $Id: spm_eeg_ft2spm.m 4038 2010-08-10 10:39:00Z vladimir $
 
 isTF = 0;
 
@@ -137,6 +137,20 @@
 
 D = meeg(D);
 
+if  isfield(ftdata, 'hdr')
+    % Uses fileio function to get the information about channel types stored in
+    % the original header. This is now mainly useful for Neuromag support but might
+    % have other functions in the future.
+    origchantypes = ft_chantype(ftdata.hdr);
+    [sel1, sel2] = spm_match_str(D.chanlabels, ftdata.hdr.label);
+    origchantypes = origchantypes(sel2);
+    if length(strmatch('unknown', origchantypes, 'exact')) ~= numel(origchantypes)
+        D.origchantypes = struct([]);
+        D.origchantypes(1).label = ftdata.hdr.label(sel2);
+        D.origchantypes(1).type = origchantypes;
+    end
+end
+
 % Set channel types to default
 S1 = [];
 S1.task = 'defaulttype';
diff --git a/spm_eeg_inv_Mesh2Voxels.m b/spm_eeg_inv_Mesh2Voxels.m
index 51b300c..951964e 100644
--- a/spm_eeg_inv_Mesh2Voxels.m
+++ b/spm_eeg_inv_Mesh2Voxels.m
@@ -1,133 +1,91 @@
 function [D] = spm_eeg_inv_Mesh2Voxels(varargin)
-% converts a mesh-representation of M/EEG power into a smoothed image
-% FORMAT [D] = spm_eeg_inv_Mesh2Voxels(D,val)
+% Convert a mesh representation of M/EEG power into a smoothed image
+% FORMAT [D] = spm_eeg_inv_Mesh2Voxels(D,[val])
 % Input:
-% D        - input data struct (optional)
+% D        - MEEG object or filename of M/EEG mat-file (optional)
 %
-%     D.inv{val}.contrast.display = display (spm_image) flag [0]
+%     D.inv{val}.contrast.display:   display image at the end {true, [false]}
+%     D.inv{val}.contrast.space:     native [0] or MNI {1} output image space
+%     D.inv{val}.contrast.smoothing: # iterations for cortical smoothing
 %
 % Output:
-% D        - data struct including the new files and parameters and
+% D        - MEEG object containing the new image filenames in fields:
+%     D.inv{val}.contrast.Vout
+%     D.inv{val}.contrast.fname
+%__________________________________________________________________________
 %
-%--------------------------------------------------------------------------
 % Non-linear interpolation of a Mesh contrast into MNI Voxel space
 % This routine is used to produce a 3D image canonical sMRI
 % space (in voxel coordinates) from a cortical mesh (3D surface).
-% This yields a .nifti image of the summary statistics of the cortical
+% This yields a NIfTI image of the summary statistics of the cortical
 % activity for the effect of interest. This image can then enter the
 % classical SPM routines for statistical testing.
 % The [non-negative] mean square contrast is smoothed both on the mesh
-% (using a graph Laplacian and then in voxel-space using a conventional
+% (using a graph Laplacian) and then in voxel-space using a conventional
 % Gaussian filter.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_eeg_inv_Mesh2Voxels.m 3996 2010-07-13 22:20:40Z vladimir $
+% $Id: spm_eeg_inv_Mesh2Voxels.m 4086 2010-10-08 12:35:36Z vladimir $
+
+
+SVNrev = '$Rev: 4086 $';
 
-% checks
+%-Startup
 %--------------------------------------------------------------------------
-[D,val]  = spm_eeg_inv_check(varargin{:});
+spm('FnBanner', mfilename, SVNrev);
 
-% Switches
+%-Parse input arguments
 %--------------------------------------------------------------------------
-try, Disp   = D.inv{val}.contrast.display;   catch, Disp   = 0;  end
-try, space  = D.inv{val}.contrast.space;     catch, space  = 1;  end
-try, smooth = D.inv{val}.contrast.smoothing; catch, smooth = 8; end
+[D,val]      = spm_eeg_inv_check(varargin{:});
 
+%-Get options
+%--------------------------------------------------------------------------
+try, Disp    = D.inv{val}.contrast.display;   catch, Disp   = 0; end
+try, space   = D.inv{val}.contrast.space;     catch, space  = 1; end
+try, smooth  = D.inv{val}.contrast.smoothing; catch, smooth = 8; end
 
-% smoothing FWHM (mm)
+%-Time and Frequency windows of interest
 %--------------------------------------------------------------------------
-fprintf('writing and smoothing image - please wait\n');                 %-#
+woi          = D.inv{val}.contrast.woi;
+foi          = D.inv{val}.contrast.fboi;
+Nw           = size(woi,1);
 
-% Get volume (MNI [1] or native [0] output image space)
+%-Get output image field of view and resolution
 %--------------------------------------------------------------------------
 if space
     sMRIfile = fullfile(spm('dir'),'templates','T2.nii');
 else
     sMRIfile = D.inv{val}.mesh.sMRI;
 end
-Vin   = spm_vol(sMRIfile);
-
-% Tag (to identify particular contrast settings)
-%--------------------------------------------------------------------------
-woi   = D.inv{val}.contrast.woi;
-foi   = D.inv{val}.contrast.fboi;
-Nw    = size(woi,1);
-for i = 1:Nw
-    tag{i} = ['t' sprintf('%d_', woi(i,:)) 'f' sprintf('%d_', foi)];
-end
+Vin          = spm_vol(sMRIfile);
 
-% Get mesh
+%-Get surface mesh
 %--------------------------------------------------------------------------
 if space
-    m = D.inv{val}.mesh.tess_mni;
+    m        = export(gifti(D.inv{val}.mesh.tess_mni),'patch');
 else
-    m = export(gifti(D.inv{val}.mesh.tess_ctx),'spm');
+    m        = export(gifti(D.inv{val}.mesh.tess_ctx),'patch');
 end
-vert  = m.vert;
-face  = m.face;
-nd    = D.inv{val}.inverse.Nd;
-nv    = size(vert,1);
-nf    = size(face,1);
-
-% Compute a densely sampled triangular mask
-%--------------------------------------------------------------------------
-[tx ty] = meshgrid(0.05:0.1:0.95, 0.05:0.1:0.95);
-tx    = tx(:);
-ty    = ty(:);
-ti    = find(sum([tx ty],2) <= 0.9);
-t     = [tx(ti) ty(ti)];
-np    = length(t);
-
-% Map the template (square) triangle onto each face of the Cortical Mesh
-%--------------------------------------------------------------------------
-P1    = vert(face(:,1),:);
-P2    = vert(face(:,2),:);
-P3    = vert(face(:,3),:);
-
-If    = speye(nf);
-alpha = t(:,1);
-beta  = t(:,2);
-teta  = ones(np,1) - alpha - beta;
-clear t tx ty ti
+nd           = D.inv{val}.inverse.Nd;
 
-DenseCortex = kron(If,alpha)*P2 + kron(If,beta)*P3 + kron(If,teta)*P1;
-clear P1 P2 P3
 
-% Transform the sampling point coordinates from mm to voxels
-%--------------------------------------------------------------------------
-VoxelCoord = Vin.mat\[DenseCortex';ones(1,size(DenseCortex,1))];
-VoxelCoord = round(VoxelCoord(1:3,:)');
-clear DenseCortex
-
-% Get Graph Laplacian for smoothing on the cortical surface
-%--------------------------------------------------------------------------
-A     = spm_mesh_distmtx(struct('vertices',vert,'faces',face),0);
-GL    = speye(nd,nd) + (A - spdiags(sum(A,2),0,nd,nd))/16;
-
-
-% normalize and embed in 3-space
+%-Accumulate mean of log-contrasts (over trials)
 %==========================================================================
-[PTH,NAME,EXT] = fileparts(D.fname);
-
+GL      = spm_mesh_smooth(m);
 
-% accumulate mean of log-contrasts (over trials)
-%--------------------------------------------------------------------------
 GW      = D.inv{val}.contrast.GW;
 
 bytrial = iscell(GW{1});
 
-Ne = [];
 if bytrial
-    for c = 1:numel(GW)
-        Ne(c) = numel(GW{c});
-    end
+    Ne  = cellfun(@numel,GW);
 else
-    Ne = ones(1, numel(GW));
+    Ne  = ones(1, numel(GW));
 end
 
-Nj = numel(GW)/Nw;
+Nj      = numel(GW)/Nw;
 
 k  = 1;
 iw = [];
@@ -138,98 +96,88 @@
     else
         cGW = GW(c);
     end
-    
+
     for t = 1:Ne(c)
-        % Smooth on the cortical surface
-        %----------------------------------------------------------------------
+        %-Smooth on the cortical surface
+        %------------------------------------------------------------------
         ssq{k} = full(sparse(D.inv{val}.inverse.Is,1,cGW{t},nd,1));
-        for i = 1:smooth
-            ssq{k} = GL*ssq{k};
-        end
-        
-        % compute (truncated) moment
-        %----------------------------------------------------------------------
+        ssq{k} = spm_mesh_smooth(GL,ssq{k},smooth);
+
+        %-Compute (truncated) moment
+        %------------------------------------------------------------------
         lss        = log(ssq{k} + eps);
         i          = lss > (max(lss) - log(32));
         meanlss(k) = mean(lss(i));
-        
+
         iw(k) = c;
         ie(k) = t;
-        
-        k = k+1;
+
+        k = k + 1;
     end
 end
 
 scale = exp(mean(meanlss));
+
+%-Normalise and embed in 3D-space
+%==========================================================================
+fprintf('%-40s: %30s','Writing images','...please wait');               %-#
+
+[pth,name] = fileparts(D.fname);
+tag = cell(1,Nw);
+for i = 1:Nw
+    tag{i} = ['t' sprintf('%d_', woi(i,:)) 'f' sprintf('%d_', foi)];
+end
+
+spm_progress_bar('Init',numel(ssq),'Interpolating images','');
+
 for c = 1:numel(ssq)
     
-    % Normalise
-    %----------------------------------------------------------------------
-    Contrast   = ssq{c}/scale;
-    
+    fprintf('%s%30s',repmat(sprintf('\b'),1,30),...
+        sprintf('...image %d/%d',c,numel(ssq)));                        %-#
     
-    % Initialise image
+    %-Initialise image
     %----------------------------------------------------------------------
     con       = mod(iw(c) - 1, Nj) + 1;
-    str       = tag{ceil(iw(c)/Nj)};    
-    if bytrial
-        fname     = fullfile(D.path,sprintf('%s_%.0f_%s%.0f_%.0f.nii',NAME,val,str,con, ie(c)));
-    else
-        fname     = fullfile(D.path,sprintf('%s_%.0f_%s%.0f.nii',NAME,val,str,con));
-    end
-    Vout           = struct(...
-        'fname',     fname,...
-        'dim',       Vin.dim,...
-        'dt',        [spm_type('float32') spm_platform('bigend')],...
-        'mat',       Vin.mat,...
-        'pinfo',     [1 0 0]',...
-        'descrip',   '');
-    
-    InterpOp  = [teta alpha beta];
-    SPvalues  = zeros(nf*np,1);
-    RECimage  = zeros(Vout.dim);
-    
-    
-    % And interpolate those values into voxels
+    str       = tag{ceil(iw(c)/Nj)};
+    if bytrial, bt = sprintf('_%.0f',ie(c)); else bt = ''; end
+    fname     = fullfile(D.path,...
+        sprintf('%s_%.0f_%s%.0f%s.nii', name, val, str, con, bt));
+    Vout      = struct(...
+        'fname',   fname,...
+        'dim',     Vin.dim,...
+        'dt',      [spm_type('float32') spm_platform('bigend')],...
+        'mat',     Vin.mat,...
+        'pinfo',   [1 0 0]',...
+        'descrip', '');
+
+    %-Normalise
     %----------------------------------------------------------------------
-    for i = 1:nf
-        TextVal = Contrast(face(i,:));
-        if any(TextVal)
-            ValTemp        = InterpOp*TextVal;
-            SPvalues((i-1)*np+1:i*np) = sum(TextVal)*(ValTemp/sum(ValTemp));
-            Vox            = VoxelCoord( (i-1)*np+1 : i*np , : );
-            Val            = SPvalues( (i-1)*np+1 : i*np );
-            [UnVox,I,J]    = unique(Vox,'rows');
-            IndV           = sub2ind(Vin.dim,UnVox(:,1),UnVox(:,2),UnVox(:,3));
-            K              = 1:length(J);
-            MatV           = zeros(max(J),length(J));
-            IndK           = sub2ind(size(MatV),J,K');
-            MatV(IndK)     = Val;
-            SV             = sum(MatV,2);
-            RECimage(IndV) = RECimage(IndV) + SV;
-        end
-    end
+    Contrast = ssq{c} / scale;
+
+    %-Interpolate those values into voxels
+    %----------------------------------------------------------------------
+    RECimage = spm_mesh_to_grid(m, Vin, Contrast);
     
-    % 3D smoothing and thresholding
+    %-3D smoothing and thresholding
     %----------------------------------------------------------------------
-    spm_smooth(RECimage,RECimage,1);
+    spm_smooth(RECimage, RECimage, 1);
     RECimage = RECimage.*(RECimage > exp(-8));
-    
-    % Write (smoothed and scaled) image
+
+    %-Write (smoothed and scaled) image
     %----------------------------------------------------------------------
-    Vout  = spm_write_vol(Vout,RECimage);
-    
-    
-    % save and report
+    Vout     = spm_write_vol(Vout, RECimage);
+
+    %-Store filename
     %----------------------------------------------------------------------
-    str = 'Summary-statistic image written:\n %s\n';
-    fprintf(str,fname);
-    
-    D.inv{val}.contrast.Vout{c}  = Vout;
     D.inv{val}.contrast.fname{c} = fname;
+
+    spm_progress_bar('Set', c);
     
 end
 
-% display
+spm_progress_bar('Clear');
+fprintf('%s%30s\n',repmat(sprintf('\b'),1,30),'...done');               %-#
+
+%-Display
 %==========================================================================
 if Disp, spm_eeg_inv_image_display(D); end
diff --git a/spm_eeg_inv_custom_ui.m b/spm_eeg_inv_custom_ui.m
index d582216..65a7005 100644
--- a/spm_eeg_inv_custom_ui.m
+++ b/spm_eeg_inv_custom_ui.m
@@ -21,7 +21,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_eeg_inv_custom_ui.m 3903 2010-05-28 09:39:13Z vladimir $
+% $Id: spm_eeg_inv_custom_ui.m 4032 2010-08-04 10:15:35Z vladimir $
  
 % defaults from D is specified
 %==========================================================================
@@ -98,11 +98,7 @@
     
     % Source space restictions
     %----------------------------------------------------------------------
-    if spm_input('Restrict solutions','+1','no|yes',[0 1],1)
-        
-        warndlg(['Please note that model comparison of restricted solutions is presently not valid. ' ...
-            'The problem is known and it will be fixed in future releases.']);
-
+    if spm_input('Restrict solutions','+1','no|yes',[0 1],1)        
         [P, sts] = spm_select(1, 'mat', 'Select source (n x 3) location file');
         if sts
             xyz         = load(P);
diff --git a/spm_eeg_inv_datareg_ui.m b/spm_eeg_inv_datareg_ui.m
index 6c0023a..ca0e82a 100644
--- a/spm_eeg_inv_datareg_ui.m
+++ b/spm_eeg_inv_datareg_ui.m
@@ -16,7 +16,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_eeg_inv_datareg_ui.m 3833 2010-04-22 14:49:48Z vladimir $
+% $Id: spm_eeg_inv_datareg_ui.m 4116 2010-11-10 14:23:54Z vladimir $
 
 % initialise
 %--------------------------------------------------------------------------
@@ -135,8 +135,10 @@
     S.targetfid = newmrifid;    
 end
 
-if ~isempty(S.sourcefid.pnt) && ~isfield(S, 'useheadshape')
-    S.useheadshape = spm_input('Use headshape points?' , '+1','yes|no', [1,0], 1);
+if ~isempty(S.sourcefid.pnt)
+    if ~isfield(S, 'useheadshape')
+        S.useheadshape = spm_input('Use headshape points?' , '+1','yes|no', [1,0], 1);
+    end
 else
     S.useheadshape = 0;
 end
@@ -144,7 +146,9 @@
 ind = 1;
 D.inv{val}.datareg = struct([]);
 
-if ~isempty(D.sensors('EEG'))
+[junk, modalities] = modality(D);
+
+if ismember('EEG', modalities) && ~isempty(D.sensors('EEG'))
     if isempty(M1)
         S.template = (D.inv{val}.mesh.template | S.useheadshape);
         M1 = spm_eeg_inv_datareg(S);
@@ -159,7 +163,7 @@
     ind = ind+1;
 end
 
-if ~isempty(D.sensors('MEG'))
+if ismember('MEG', modalities) && ~isempty(D.sensors('MEG'))
     if  D.inv{val}.mesh.template
         S.template = 2;
     else
diff --git a/spm_eeg_inv_extract.m b/spm_eeg_inv_extract.m
new file mode 100644
index 0000000..ec021a2
--- /dev/null
+++ b/spm_eeg_inv_extract.m
@@ -0,0 +1,227 @@
+function [Ds, D] = spm_eeg_inv_extract(D)
+% Exports source activity using the MAP projector
+% FORMAT [Ds] = spm_eeg_inv_extract(D)
+% Requires:
+%
+%     D.inv{i}.source.XYZ   - (n x 3) matrix of MNI coordinates
+%
+% Optional:
+%
+%     D.inv{i}.source.rad   - radius (mm) of VOIs (default 5 mm)
+%     D.inv{i}.source.label - label(s) for sources (cell array)
+%     D.inv{i}.source.fname - output file name
+%     D.inv{i}.source.type  - output type ('evoked'/'trials')
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Vladimir Litvak, Laurence Hunt, Karl Friston
+% $Id: spm_eeg_inv_extract.m 4260 2011-03-23 13:42:21Z vladimir $
+ 
+% SPM data structure
+%==========================================================================
+try
+    inv = D.inv{D.val};
+catch
+    inv = D.inv{end};
+    D.val = numel(D.inv);
+end
+ 
+Ds = [];
+ 
+% defaults
+%--------------------------------------------------------------------------
+try, XYZ   = inv.source.XYZ;   catch,  return;                          end
+try, rad   = inv.source.rad;   catch,  rad   = 5;                       end
+try, label = inv.source.label; catch,  label = {};                      end
+try
+    fname  = inv.source.fname; 
+catch
+    fname  = fullfile(D.path, ['i' D.fname]); 
+end
+try
+    type   =  inv.source.type;
+catch
+    if isequal(D.type, 'evoked')
+        type = 'evoked';
+    else
+        type = 'trials';
+    end
+end
+ 
+if numel(label)<size(XYZ, 1)
+    for i = (numel(label)+1):size(XYZ, 1)
+        label{i} = ['Source ' num2str(i)];
+    end
+end
+ 
+% find relevant vertices
+%==========================================================================
+vert  = inv.mesh.tess_mni.vert;              % vertices
+Ns    = size(XYZ, 1);                        % number of sources
+svert = {};
+for i = 1:Ns
+    dist = sqrt(sum([vert(:,1) - XYZ(i,1), ...
+                     vert(:,2) - XYZ(i,2), ...
+                     vert(:,3) - XYZ(i,3)].^2, 2));
+    if rad > 0
+        svert{i} = find(dist < rad);
+    else
+        [junk svert{i}] = min(dist);
+        XYZ = vert(svert{i}, :);
+    end
+end
+ 
+
+% report
+%--------------------------------------------------------------------------
+Iy = find(cellfun('isempty', svert));
+if ~isempty(Iy)
+    disp(['No proximal vertices found for source(s) ' num2str(Iy)]);
+    disp('These sources will be discarded');
+    
+    svert(Iy) = [];
+    label(Iy) = [];
+    Ns        = numel(label);
+    
+    if Ns == 0
+        disp('No valid sources - Exiting');
+        return;
+    end
+end
+ 
+js      = spm_vec(svert);
+ 
+% get info from inv
+%--------------------------------------------------------------------------
+scale   = inv.inverse.scale;
+J       = inv.inverse.J;  
+U       = inv.inverse.U; % spatial projector (contains multiple cells if multimodal)
+T       = inv.inverse.T; % S in Friston 2008 - temporal projector
+TT      = T*T';
+M       = inv.inverse.M(js, :);
+Ic      = inv.inverse.Ic;
+It      = inv.inverse.It;
+Np      = length(It);
+ 
+try
+    trial = model.inverse.trials;
+catch
+    trial = D.condlist;
+end
+ 
+ 
+% get source data
+%==========================================================================
+MYi    = 1;
+switch(type)
+    case 'evoked'
+        Ne     = length(trial);
+        MY     = zeros(size(M,1), Ne*Np);   
+        
+        for i = 1:Ne
+            MY(:, MYi:(MYi+Np-1)) = J{i}(js,:)*T';
+            MYi                   = MYi + Np;
+        end
+        
+        clabel = trial;
+    case 'trials'
+        Ne     = length(D.pickconditions(trial));
+        MY     = zeros(size(M,1), Ne*Np);
+        clabel = {};
+        
+        for i = 1:numel(trial)
+            
+            c      = D.pickconditions(trial{i});
+            clabel = [clabel D.conditions(c)];
+            
+            % conditional expectation of contrast (J*W) and its energy
+            %--------------------------------------------------------------
+            Nt    = length(c);
+            spm_progress_bar('Init',Nt,sprintf('extracting data condition %d',i),'trials');
+            
+            for j = 1:Nt
+                
+                if ~strcmp(D.modality(1,1), 'Multimodal')
+                    
+                    % unimodal data
+                    %------------------------------------------------------
+                    Y     = D(Ic{1},It,c(j))*TT;
+                    Y     = U{1}*Y*scale;
+                    
+                else
+                    
+                    % multimodal data
+                    %------------------------------------------------------
+                    for k = 1:length(U)
+                        Y       = D(Ic{k},It,c(j))*TT;
+                        UY{k,1} = U{k}*Y*scale(k);
+                    end
+                    Y = spm_cat(UY);
+                end
+                
+                MY(:, MYi:(MYi+Np-1)) = M*Y;
+                MYi                   = MYi+Np;
+                
+                spm_progress_bar('Set',j)
+            end
+            spm_progress_bar('Clear')
+        end
+end
+ 
+% compute regional response in terms of first eigenvariate
+%==========================================================================
+iS    = [1 cumsum(cellfun('length', svert))];
+Y     = zeros(Ns, size(MY, 2));
+ 
+spm_progress_bar('Init', Ns, 'extracting eigenvariates', 'sources');
+ 
+for i = 1:Ns    
+    y     = MY(iS(i):iS(i+1),:);
+    [m n] = size(y);
+    if m > n && n > 1
+        [v s v] = svd(y'*y);
+        Y(i, :) = v(:,1);
+    elseif m>1
+        [u s u] = svd(y*y');
+        u       = u(:,1);
+        Y(i, :) = y'*u;
+    else
+        Y(i, :) = y;
+    end
+    spm_progress_bar('Set',i);
+end
+spm_progress_bar('Clear')
+ 
+% create source dataset
+%-----------------------------------------------------------------------
+Ds = clone(D, fname, [Ns Np Ne], 3);
+Ds = chanlabels(Ds, 1:Ns, label);
+Ds = timeonset(Ds, D.time(It(1)));
+Ds = chantype(Ds, 1:Ns, 'LFP');
+Ds = conditions(Ds, 1:Ne, clabel);
+Ds = sensors(Ds, 'MEG', []);
+Ds = sensors(Ds, 'EEG', []);
+Ds = fiducials(Ds, []);
+Ds = rmfield(Ds, fieldnames(Ds));
+Ds = condlist(Ds, trial);
+ 
+spm_progress_bar('Init', Ne, 'writing out data', 'trials');
+for i = 1:Ne
+    Ds(:,:,i) = Y(:,((i - 1)*Np + 1):i*Np);
+    spm_progress_bar('Set',i);
+end
+spm_progress_bar('Clear');
+ 
+% store results in the original inv
+%-----------------------------------------------------------------------
+D.inv{D.val}.source.type  = type;
+D.inv{D.val}.source.XYZ   = XYZ;
+D.inv{D.val}.source.label = label;
+D.inv{D.val}.source.rad   = rad;
+D.inv{D.val}.source.fname = fname;
+ 
+% save headers
+%-----------------------------------------------------------------------
+save(Ds);
+
+
diff --git a/spm_eeg_inv_extract_ui.m b/spm_eeg_inv_extract_ui.m
new file mode 100644
index 0000000..c172bf7
--- /dev/null
+++ b/spm_eeg_inv_extract_ui.m
@@ -0,0 +1,90 @@
+function [D] = spm_eeg_inv_extract_ui(varargin)
+% GUI for contrast of evoked responses and power for an MEG-EEG model
+% FORMAT [D] = spm_eeg_inv_extract_ui(D, val, PST)
+% Sets:
+%
+%     D.contrast.woi   - time (ms) window of interest
+%     D.contrast.fboi  - freq (Hz) window of interest
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+
+% Vladimir Litvak
+% $Id: spm_eeg_inv_extract_ui.m 4260 2011-03-23 13:42:21Z vladimir $
+
+% initialise
+%--------------------------------------------------------------------------
+[D,val] = spm_eeg_inv_check(varargin{1:2});
+
+D.inv{val}.source = [];
+
+try
+    D.inv{val}.source.XYZ = varargin{3};
+catch
+    warndlg('Position in MNI coordinates should be provided');
+    return
+end
+
+% get source label
+%--------------------------------------------------------------------------
+D.inv{val}.source.label{1} = spm_input('Source label','+1','s');
+
+% get VOI radius
+%--------------------------------------------------------------------------
+D.inv{val}.source.rad = spm_input('VOI radius','+1','r', 5, [1 1]);
+
+% trials or or evoked?
+%--------------------------------------------------------------------------
+if strcmp(D.type, 'single')
+    str  = {'evoked','trials'};
+    type = spm_input('What to save?','+1','b',str,[],1);
+else
+    type = 'evoked';
+end
+D.inv{val}.source.type = type;
+
+% output file name
+%--------------------------------------------------------------------------
+fname = spm_input('File name','+1','s', ['i' spm_str_manip(D.fname, 'r')]);
+D.inv{val}.source.fname = fullfile(D.path, [fname '.mat']);
+
+% extract
+%--------------------------------------------------------------------------
+[Ds, D] = spm_eeg_inv_extract(D);
+
+% display
+%==========================================================================
+if ~isempty(Ds)
+    
+    try, con = D.con; catch, con = 1; end
+    
+    if con == 0
+        con = 1;
+    end
+    
+    Fgraph = spm_figure('GetWin','Graphics');
+    clf(Fgraph)
+    figure(Fgraph)
+    
+    cl = D.condlist;
+    
+    for i = 1:numel(cl)
+        if i == con
+            Color = [1 0 0];
+        else
+            Color = [1 1 1]*(1 - 1/4);
+        end
+        
+        plot(time(Ds, [], 'ms'), squeeze(Ds(1, :, Ds.pickconditions(cl{i}))), 'Color', Color);                
+        
+        hold on
+    end
+    
+    axis square
+    
+    title({sprintf('Extracted data from source %s', Ds.chanlabels{1}),...
+        sprintf('at %d %d %d mm', round(D.inv{val}.source.XYZ))});
+    
+    xlabel('PST {ms}')
+    ylabel('source amplitude')
+    drawnow
+end
\ No newline at end of file
diff --git a/spm_eeg_inv_fmripriors.m b/spm_eeg_inv_fmripriors.m
index fb925aa..b63eedb 100644
--- a/spm_eeg_inv_fmripriors.m
+++ b/spm_eeg_inv_fmripriors.m
@@ -30,7 +30,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin and Rik Henson
-% $Id: spm_eeg_inv_fmripriors.m 3586 2009-11-19 16:56:58Z rik $
+% $Id: spm_eeg_inv_fmripriors.m 4185 2011-02-01 18:46:18Z guillaume $
 
 %-Get MEEG object
 %--------------------------------------------------------------------------
@@ -216,7 +216,7 @@
     pQ = pQ*pQ';
 end
 
-if spm_matlab_version_chk('7') >=0
+if spm_check_version('matlab','7') >=0
     save(D.inv{val}.inverse.fmri.priors,'pQ','-V6');
 else
     save(D.inv{val}.inverse.fmri.priors,'pQ');
diff --git a/spm_eeg_inv_imag_api.fig b/spm_eeg_inv_imag_api.fig
index a3272c9..9315654 100644
Binary files a/spm_eeg_inv_imag_api.fig and b/spm_eeg_inv_imag_api.fig differ
diff --git a/spm_eeg_inv_imag_api.m b/spm_eeg_inv_imag_api.m
index 997746f..b0db398 100644
--- a/spm_eeg_inv_imag_api.m
+++ b/spm_eeg_inv_imag_api.m
@@ -7,7 +7,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jeremie Mattout
-% $Id: spm_eeg_inv_imag_api.m 3996 2010-07-13 22:20:40Z vladimir $
+% $Id: spm_eeg_inv_imag_api.m 4260 2011-03-23 13:42:21Z vladimir $
 
 spm('Clear');
 
@@ -205,7 +205,7 @@ function clear_Callback(hObject, eventdata, handles)
 % --- Executes on button press in delete.
 %--------------------------------------------------------------------------
 function delete_Callback(hObject, eventdata, handles)
-if length(handles.D.inv)
+if ~isempty(handles.D.inv)
     try
         str = handles.D.inv{handles.D.val}.comment;
         warndlg({'you are about to delete:',str{1}});
@@ -231,7 +231,7 @@ function Reset(hObject, eventdata, handles)
     new_Callback(hObject, eventdata, handles)
     return
 end
-if ~length(handles.D.inv)
+if isempty(handles.D.inv)
     new_Callback(hObject, eventdata, handles)
     return
 end
@@ -259,7 +259,6 @@ function Reset(hObject, eventdata, handles)
 end
 % =========================================================================
 
-
 set(handles.new,      'enable','on','value',0)
 set(handles.clear,    'enable','on','value',0)
 set(handles.delete,   'enable','on','value',0)
@@ -418,6 +417,9 @@ function CheckInverse_Callback(hObject, eventdata, handles)
 if strcmp(handles.D.inv{handles.D.val}.method,'Imaging')
     PST    = str2num(get(handles.PST,'String'));
     spm_eeg_invert_display(handles.D,PST);
+    if length(PST) == 3 && get(handles.extract, 'Value')
+        handles.D = spm_eeg_inv_extract_ui(handles.D, handles.D.val, PST);        
+    end
 elseif strcmp(handles.D.inv{handles.D.val}.method, 'vbecd')
     spm_eeg_inv_vbecd_disp('init',handles.D);
 end
diff --git a/spm_eeg_inv_mesh_ui.m b/spm_eeg_inv_mesh_ui.m
index 36103a4..2e8ee6e 100644
--- a/spm_eeg_inv_mesh_ui.m
+++ b/spm_eeg_inv_mesh_ui.m
@@ -16,9 +16,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jeremie Mattout & Christophe Phillips
-% $Id: spm_eeg_inv_mesh_ui.m 3731 2010-02-17 14:45:18Z vladimir $
+% $Id: spm_eeg_inv_mesh_ui.m 4027 2010-07-31 12:49:19Z vladimir $
 
-SVNrev = '$Rev: 3731 $';
+SVNrev = '$Rev: 4027 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -55,7 +55,7 @@
 elseif ~ischar(sMRI)
     template = sMRI; % for backward compatibility
 else
-    template = 1;
+    template = 0;
 end
 
 if template
diff --git a/spm_eeg_inv_rigidreg.m b/spm_eeg_inv_rigidreg.m
index c465dce..0d3128f 100644
--- a/spm_eeg_inv_rigidreg.m
+++ b/spm_eeg_inv_rigidreg.m
@@ -10,7 +10,9 @@
 %            second 3D surface to be registered 
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
-%$Id: spm_eeg_inv_rigidreg.m 1523 2008-04-30 17:33:04Z vladimir $
+
+% Vladimir Litvak
+% $Id: spm_eeg_inv_rigidreg.m 4082 2010-10-07 16:11:48Z guillaume $
  
 M       = spm_detrend(data1');
 S       = spm_detrend(data2');
@@ -23,5 +25,3 @@
 end
 t1 = mean(data1,2) - R1*mean(data2,2);
 M1 = [R1 t1; 0 0 0 1];
-
-return
diff --git a/spm_eeg_inv_spatnorm.m b/spm_eeg_inv_spatnorm.m
index b6f43c9..f13c9be 100644
--- a/spm_eeg_inv_spatnorm.m
+++ b/spm_eeg_inv_spatnorm.m
@@ -13,7 +13,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jeremie Mattout
-% $Id: spm_eeg_inv_spatnorm.m 3117 2009-05-12 15:50:00Z john $
+% $Id: spm_eeg_inv_spatnorm.m 4136 2010-12-09 22:22:28Z guillaume $
 
 spm('Pointer','Watch');
 
@@ -36,7 +36,7 @@
         'write',    [0 0]),...
         'tissue', struct('tpm',   {{[T,'1']},{[T,'2']},{[T,'3']},{[T,'4']},{[T,'5']},{[T,'6']}},...
         'ngaus', {2,2,2,3,4,2},...
-        'native',{[0 0],[0 0],[0 0],[0 0],[0 0],[0 0]},....
+        'native',{[0 0],[0 0],[0 0],[0 0],[0 0],[0 0]},...
         'warped',{[0 0],[0 0],[0 0],[0 0],[0 0],[0 0]}),...
         'warp',   struct('reg',4, 'affreg', 'mni', 'samp', 3, 'write', [0 1]));
 
diff --git a/spm_eeg_inv_transform_mesh.m b/spm_eeg_inv_transform_mesh.m
index 3266de7..18aa420 100644
--- a/spm_eeg_inv_transform_mesh.m
+++ b/spm_eeg_inv_transform_mesh.m
@@ -1,22 +1,21 @@
 function mesh = spm_eeg_inv_transform_mesh(M, mesh)
-% Applies homogenous transformation to cortical mesh
+% Applies affine transformation to surface mesh
 % FORMAT mesh = spm_eeg_inv_transform_mesh(M, mesh)
 %
-% M1          - affine transformation matrix [4 x 4]
-% mesh        - 
+% M           - affine transformation matrix [4 x 4]
+% mesh        - patch structure
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_eeg_inv_transform_mesh.m 2863 2009-03-11 20:25:33Z guillaume $
+% $Id: spm_eeg_inv_transform_mesh.m 4082 2010-10-07 16:11:48Z guillaume $
 
 fn = fieldnames(mesh);
 
-tess_ind = strmatch('tess_', fn);
-tess_ind = setdiff(tess_ind, strmatch('tess_mni', fn, 'exact'));
+tess_ind = find(strncmp(fn,'tess_',5) & ~strcmp(fn,'tess_mni'));
 
 for i = 1:length(tess_ind)
-    cmesh =  export(gifti(getfield(mesh, fn{tess_ind(i)})), 'spm');
+    cmesh = export(gifti(mesh.(fn{tess_ind(i)})),'spm');
     cmesh.vert = spm_eeg_inv_transform_points(M,cmesh.vert);
-    mesh = setfield(mesh, fn{tess_ind(i)}, cmesh);
+    mesh.(fn{tess_ind(i)}) = cmesh;
 end
diff --git a/spm_eeg_inv_vbecd.m b/spm_eeg_inv_vbecd.m
index aacc50c..b7d563e 100644
--- a/spm_eeg_inv_vbecd.m
+++ b/spm_eeg_inv_vbecd.m
@@ -31,7 +31,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Gareth Barnes
-% $Id: spm_eeg_inv_vbecd.m 3906 2010-06-01 10:48:32Z gareth $
+% $Id: spm_eeg_inv_vbecd.m 4281 2011-03-31 19:49:57Z karl $
 
 
 
@@ -100,10 +100,10 @@
   %% startguess=[-0.3553  -69.8440    1.0484    0.2545    0.3428    1.8526]'
     
     [starty]=spm_eeg_wrap_dipfit_vbecd(startguess,M,U);
-        [Ep,Cp,S,F] = spm_nlsi_GN(M,U,Y);
+        [Ep,Cp,Eh,F] = spm_nlsi_GN(M,U,Y);
     P.Ep = Ep;
     P.Cp = Cp;
-    P.S  = S;
+    P.Eh = Eh;
     P.F  = F;
     [P.ypost,outsideflag]=spm_eeg_wrap_dipfit_vbecd(P.Ep,M,U);
     P.ypost = P.ypost./sc_y; %%  scale it back
diff --git a/spm_eeg_inv_vbecd_gui.m b/spm_eeg_inv_vbecd_gui.m
index 8628c8e..aae819e 100644
--- a/spm_eeg_inv_vbecd_gui.m
+++ b/spm_eeg_inv_vbecd_gui.m
@@ -7,7 +7,7 @@
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 % 
-% $Id: spm_eeg_inv_vbecd_gui.m 3951 2010-06-28 15:09:36Z gareth $
+% $Id: spm_eeg_inv_vbecd_gui.m 4210 2011-02-23 11:02:32Z gareth $
 
 %%
 % Load data, if necessary
@@ -221,7 +221,7 @@
 
 priorlocvardefault=[100, 100, 100]; %% location variance default in mm
 nopriorlocvardefault=[80*80, 80*80, 80*80];
-nopriormomvardefault=[10, 10, 10]; %% moment variance in nAM
+nopriormomvardefault=[10, 10, 10]*100; %% moment variance in nAM
 priormomvardefault=[1, 1, 1]; %% 
 
 
@@ -262,19 +262,25 @@
                 
                 str2='Prior location variance (mm2)';
                 diags_s0_mni = spm_input(str2, 1+tr_q+dip_q+2,'e',priorlocvardefault)';
-                diags_s0=(orM1*sqrt(diags_s0_mni)).^2;
+                  
+          
+                S_s0_ctf=orM1*diag(diags_s0_mni)*orM1'; %% transform covariance
+            
+                
+                %% need to leave diags(S0) free
               
                 if all(~outside), break, end
                     str = 'Prior location must be inside head';
                   end
             dip_pr(dip_q).mu_s0 = s0;    
-        else
+           else
             % no location  prior
             dip_pr(dip_q).mu_s0 = zeros(3,1);
-            diags_s0= nopriorlocvardefault';
-        end
-        %% prior cov matrix for single dipole location (i.e. no crosstalk)
-        dip_pr(dip_q).S_s0=eye(length(diags_s0)).*repmat(diags_s0,1,length(diags_s0)); 
+            diags_s0_mni= nopriorlocvardefault';
+            S_s0_ctf=diag(diags_s0_mni);
+            end
+        
+        dip_pr(dip_q).S_s0=S_s0_ctf; %
         
         % Moment prior
         wpr_q = spm_input('Moment prior ?',1+tr_q+dip_q+spr_q+2,'b', ...
@@ -286,16 +292,17 @@
             str2='Prior moment variance (nAm2)';
             diags_w0_mni = spm_input(str2, 1+tr_q+dip_q+2,'e',priormomvardefault)';
             dip_pr(dip_q).mu_w0 =orM1*w0_mni;
-            diags_w0=(orM1*sqrt(diags_w0_mni)).^2;
+            S_w0_ctf=orM1*w0_mni*orM1';
+            
              
         else
             % no location  prior
             dip_pr(dip_q).mu_w0 = zeros(3,1);
-            diags_w0= nopriormomvardefault';
+            S_w0_ctf= diag(nopriormomvardefault);
         end
         %% set up covariance matrix for orientation with no crosstalk terms (for single
         %% dip)
-        dip_pr(dip_q).S_w0=eye(length(diags_w0)).*repmat(diags_w0,1,length(diags_w0));
+        dip_pr(dip_q).S_w0=S_w0_ctf;
         dip_c = dip_c+1;
     else
     % add a pair of symmetric dipoles to the model
@@ -323,6 +330,7 @@
                 str2='Prior location variance (mm2)';
                 tmp_diags_s0_mni = spm_input(str2, 1+tr_q+dip_q+2,'e',priorlocvardefault)';
                 tmp_diags_s0_mni= [tmp_diags_s0_mni ; tmp_diags_s0_mni ];
+                
               
                 if all(~outside), break, end
                     str = 'Prior location must be inside head';
@@ -333,6 +341,7 @@
             % no location  prior
             dip_pr(dip_q).mu_s0 = zeros(6,1);
             tmp_diags_s0 = [nopriorlocvardefault';nopriorlocvardefault'];
+            tmp_diags_s0_mni=[nopriorlocvardefault';nopriorlocvardefault'];
         end %% end of if informative prior
         %% setting up a covariance matrix where there is covariance between
         %% the x parameters negatively coupled, y,z positively.
@@ -346,8 +355,16 @@
          mni_dip_pr(dip_q).S_s0(3,6)=mni_dip_pr(dip_q).S_s0(6,3);
          %% transform to MEG space   
          
-         dip_pr(dip_q).S_s0(:,1:3)=mni_dip_pr(dip_q).S_s0(:,1:3)*orM1; %% NEED TO LOOK AT THIS
-         dip_pr(dip_q).S_s0(:,4:6)=mni_dip_pr(dip_q).S_s0(:,4:6)*orM1;
+         %dip_pr(dip_q).S_s0(:,1:3)=orM1*mni_dip_pr(dip_q).S_s0(:,1:3)*orM1'; %% NEED TO LOOK AT THIS
+         %dip_pr(dip_q).S_s0(:,4:6)=orM1*mni_dip_pr(dip_q).S_s0(:,4:6)*orM1';
+         tmp1=orM1*mni_dip_pr(dip_q).S_s0(1:3,1:3)*orM1'; %% NEED TO LOOK AT THIS
+         tmp2=orM1*mni_dip_pr(dip_q).S_s0(1:3,4:6)*orM1';
+         tmp3=orM1*mni_dip_pr(dip_q).S_s0(4:6,4:6)*orM1';
+         tmp4=orM1*mni_dip_pr(dip_q).S_s0(4:6,1:3)*orM1';
+         dip_pr(dip_q).S_s0(1:3,1:3)=tmp1;
+         dip_pr(dip_q).S_s0(1:3,4:6)=tmp2;
+         dip_pr(dip_q).S_s0(4:6,4:6)=tmp3;
+         dip_pr(dip_q).S_s0(4:6,1:3)=tmp4;
          
         % Moment prior
         wpr_q = spm_input('Moment prior ?',1+tr_q+dip_q+spr_q+2,'b', ...
@@ -370,15 +387,23 @@
         end
         %dip_pr(dip_q).S_w0=eye(length(diags_w0)).*repmat(diags_w0,1,length(diags_w0));
         %% couple all orientations, except x, positively or leave for now...
+          
                 mni_dip_pr(dip_q).S_w0 = eye(length(tmp_diags_w0)).*repmat(tmp_diags_w0,1,length(tmp_diags_w0));
-            mni_dip_pr(dip_q).S_w0(4,1)=-mni_dip_pr(dip_q).S_w0(4,4); %
+                
+            mni_dip_pr(dip_q).S_w0(4,1)=-mni_dip_pr(dip_q).S_w0(4,4); % reflect x orientation
             mni_dip_pr(dip_q).S_w0(5,2)=mni_dip_pr(dip_q).S_w0(5,5); %
             mni_dip_pr(dip_q).S_w0(6,3)=mni_dip_pr(dip_q).S_w0(6,6); %
             mni_dip_pr(dip_q).S_w0(1,4)=-mni_dip_pr(dip_q).S_w0(4,1); %
             mni_dip_pr(dip_q).S_w0(2,5)=mni_dip_pr(dip_q).S_w0(5,2); %
             mni_dip_pr(dip_q).S_w0(3,6)=mni_dip_pr(dip_q).S_w0(6,3); %
-          dip_pr(dip_q).S_w0(:,1:3)=mni_dip_pr(dip_q).S_w0(:,1:3)*orM1;
-         dip_pr(dip_q).S_w0(:,4:6)=mni_dip_pr(dip_q).S_w0(:,4:6)*orM1;
+            tmp1=orM1*mni_dip_pr(dip_q).S_w0(1:3,1:3)*orM1'; 
+            tmp2=orM1*mni_dip_pr(dip_q).S_w0(1:3,4:6)*orM1';
+            tmp3=orM1*mni_dip_pr(dip_q).S_w0(4:6,4:6)*orM1';
+            tmp4=orM1*mni_dip_pr(dip_q).S_w0(4:6,1:3)*orM1';
+            dip_pr(dip_q).S_w0(1:3,1:3)=tmp1;
+            dip_pr(dip_q).S_w0(1:3,4:6)=tmp2;
+            dip_pr(dip_q).S_w0(4:6,4:6)=tmp3;
+            dip_pr(dip_q).S_w0(4:6,1:3)=tmp4;
          
         
         
@@ -386,10 +411,12 @@
     end
 end
 
-str2='Data SNR (amp)';
- SNRamp = spm_input(str2, 1+tr_q+dip_q+2+1,'e',5)';
+%str2='Data SNR (amp)';
+% SNRamp = spm_input(str2, 1+tr_q+dip_q+2+1,'e',5)';
+SNRamp=3;
  hE=log(SNRamp^2); %% expected log precision of data
- hC=0.0000001; % Have to tie the expected precision down for the moment (nlsi_gn can become unstable)
+ hC=1; % variability of the above precision
+ 
  
 str2='Number of iterations';
  Niter = spm_input(str2, 1+tr_q+dip_q+2+2,'e',10)';
@@ -467,6 +494,8 @@
     inverse.F(ii) = P.F; % free energy
     
      megloc=reshape(P.post_mu_s,3,length(P.post_mu_s)/3); % loc of dip (3 x n_dip)
+     meg_w=reshape(P.post_mu_w,3,length(P.post_mu_w)/3); % moments of dip (3 x n_dip)
+     mni_w=orM1*meg_w; %% orientation in mni space
      mniloc=D.inv{val}.datareg.toMNI*[megloc;ones(1,size(megloc,2))]; %% actual MNI location (with scaling)
      inverse.mniloc{ii}=mniloc(1:3,:);
     inverse.loc{ii} = megloc;
@@ -475,7 +504,7 @@
     
     
     inverse.j{ii} = P.post_mu_w; % dipole(s) orient/ampl, in 1 column in meg space
-    inverse.jmni{ii} = orM1*P.post_mu_w; % dipole(s) orient/ampl in mni space
+    inverse.jmni{ii} = reshape(mni_w,1,prod(size(mni_w)))'; % dipole(s) orient/ampl in mni space
     inverse.cov_loc{ii} = P.post_S_s; % cov matrix of source location
     inverse.cov_j{ii} = P.post_S_w; % cov matrix of source orient/ampl
     inverse.exitflag(ii) = 1; % Converged (1) or not (0)
@@ -487,12 +516,19 @@
     spm_clf(P.handles.hfig)
      
      megmom=reshape(Pout(maxind).post_mu_w,3,length(Pout(maxind).post_mu_w)/3); % moments of dip (3 x n_dip)
-     megposvar=reshape(diag(Pout(maxind).post_S_s),3,length(Pout(maxind).post_mu_s)/3); %% estimate of positional uncertainty in three principal axes
+     %megposvar=reshape(diag(Pout(maxind).post_S_s),3,length(Pout(maxind).post_mu_s)/3); %% estimate of positional uncertainty in three principal axes
      mnimom=orM1*megmom; %% convert moments into mni coordinates through a rotation (no scaling or translation)
-     mniposvar=(orM1*sqrt(megposvar)).^2; %% convert pos variance into approx mni space by switching axes
      
      
-     displayVBupdate2(Pout(j).y,pov,allF,Niter,dip_amp,mnimom,mniloc(1:3,:),mniposvar,P,j,[],Pout(j).F,Pout(j).ypost,maxind);
+     longorM1=zeros(size(Pout(maxind).post_S_s,1));
+     for k1=1:length(Pout(maxind).post_S_s)/3;
+        longorM1((k1-1)*3+1:k1*3,(k1-1)*3+1:k1*3)=orM1;
+     end; % for k1
+     S0_mni=longorM1*Pout(maxind).post_S_s*longorM1';
+     mniposvar=diag(S0_mni); %% convert pos variance into approx mni space by switching axes
+     mniposvar=reshape(mniposvar,3,length(Pout(maxind).post_S_s)/3);
+     
+     displayVBupdate2(Pout(maxind).y,pov,allF,Niter,dip_amp,mnimom,mniloc(1:3,:),mniposvar,P,j,[],Pout(maxind).F,Pout(maxind).ypost,maxind);
      
     %displayVBupdate2(Pout(maxind).y,pov,allF,Niter,dip_amp,mniloc,Pout(maxind).post_mu_s,Pout(maxind).post_S_s,P,j,[],Pout(maxind).F,Pout(maxind).ypost,maxind,D);
   % 
diff --git a/spm_eeg_inv_visu3D_api.fig b/spm_eeg_inv_visu3D_api.fig
index 82fd85d..27728ad 100644
Binary files a/spm_eeg_inv_visu3D_api.fig and b/spm_eeg_inv_visu3D_api.fig differ
diff --git a/spm_eeg_inv_visu3D_api.m b/spm_eeg_inv_visu3D_api.m
index 31c152b..ce37ee5 100644
--- a/spm_eeg_inv_visu3D_api.m
+++ b/spm_eeg_inv_visu3D_api.m
@@ -5,12 +5,12 @@
 % - SPM_EEG_INV_VISU3D_API(filename) where filename is the eeg/meg .mat file
 % - SPM_EEG_INV_VISU3D_API('callback_name', ...) invoke the named callback.
 %
-% Last Modified by GUIDE v2.5 01-Feb-2007 20:16:13
+% Last Modified by GUIDE v2.5 18-Feb-2011 14:23:27
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jeremie Mattout
-% $Id: spm_eeg_inv_visu3D_api.m 4004 2010-07-20 15:45:39Z vladimir $
+% $Id: spm_eeg_inv_visu3D_api.m 4211 2011-02-23 16:00:02Z vladimir $
 
 % INITIALISATION CODE
 %--------------------------------------------------------------------------
@@ -71,7 +71,7 @@ function spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
 end
 
 if ~isfield(D,'inv')
-    error(sprintf('Please specify and invert a forward model\n'));
+    error('Please specify and invert a forward model\n');
 end
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % GET RESULTS (default: current or last analysis)
@@ -129,13 +129,18 @@ function spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
     handles.pst       = D.inv{val}.inverse.pst(Ts);
     handles.srcs_data = J;
     handles.Nmax      = max(abs(J(:)));
+    handles.Is        = Is;
     
     % sensor data
-    %----------------------------------------------------------------------
-    A                 = pinv(full(spm_cat(spm_diag(U))));
+    %----------------------------------------------------------------------        
+    if ~iscell(U)
+        U = {U'};
+    end
+    
+    A             = spm_pinv(spm_cat(spm_diag(U))')';    
+    
     handles.sens_data = A*Y*T(Ts,:)';
     handles.pred_data = A*L*J(Is,:);
-    
 catch
     warndlg({'Please invert your model';'inverse solution not valid'});
     return
@@ -185,7 +190,7 @@ function spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
 
 handles.vert  = vert;
 handles.face  = face;
-handles.grayc = sqrt(sum((vert.^2)')); handles.grayc = handles.grayc'/max(handles.grayc);
+handles.grayc = sqrt(sum((vert.^2),2)); handles.grayc = handles.grayc'/max(handles.grayc);
 clear vert face
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
@@ -228,9 +233,24 @@ function spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % LOAD SENSOR FILE
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-coor = D.coor2D;
-xp = coor(1,:)';
-yp = coor(2,:)';
+Ic = {};
+if iscell(D.inv{val}.inverse.Ic)
+    for i = 1:numel(D.inv{val}.inverse.Ic)
+        if i == 1
+            Ic{i} = 1:length(D.inv{val}.inverse.Ic{i});
+        else
+            Ic{i} = Ic{i-1}(end)+(1:length(D.inv{val}.inverse.Ic{i}));
+        end
+    end
+else
+   Ic{1} = 1:length(D.inv{val}.inverse.Ic); 
+end
+
+handles.Ic = Ic;
+
+coor = D.coor2D(full(spm_cat(D.inv{val}.inverse.Ic)));
+xp   = coor(1,:)';
+yp   = coor(2,:)';
 
 x        = linspace(min(xp),max(xp),64);
 y        = linspace(min(yp),max(yp),64);
@@ -242,20 +262,29 @@ function spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % INITIAL SENSOR LEVEL DISPLAY
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+
+if isfield(D.inv{val}.inverse, 'modality')
+  set(handles.modality, 'String', D.inv{val}.inverse.modality);
+else
+   set(handles.modality, 'String', 'MEEG');  % This is for backward compatibility with old DCM-IMG
+end
+
 figure(handles.fig)
 axes(handles.sensors_axes);
 cla; axis off
 
-ic       = 1:length(handles.sens_coord);
-disp     = full(handles.sens_data(ic,1));
-imagesc(x,y,griddata(xp,yp,disp,xm,ym));
+im       = get(handles.modality, 'Value');
+
+ic         = handles.Ic{im};
+disp       = full(handles.sens_data(ic,1));
+imagesc(x,y,griddata(xp(ic),yp(ic),disp,xm,ym));
 axis image xy off
 handles.sens_coord_x   = x;
 handles.sens_coord_y   = y;
 handles.sens_coord2D_X = xm;
 handles.sens_coord2D_Y = ym;
 hold on
-handles.sensor_loc = plot(handles.sens_coord(:,1),handles.sens_coord(:,2),'o','MarkerFaceColor',[1 1 1]/2,'MarkerSize',6);
+handles.sensor_loc = plot(handles.sens_coord(ic,1),handles.sens_coord(ic,2),'o','MarkerFaceColor',[1 1 1]/2,'MarkerSize',6);
 set(handles.checkbox_sensloc,'Value',1);
 hold off
 
@@ -264,7 +293,7 @@ function spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 axes(handles.pred_axes); cla;
 disp      = full(handles.pred_data(ic,1));
-imagesc(x,y,griddata(xp,yp,disp,xm,ym));
+imagesc(x,y,griddata(xp(ic),yp(ic),disp,xm,ym));
 axis image xy off
 drawnow
 
@@ -346,7 +375,9 @@ function UpDate_Display_SRCS(hObject,handles)
 function UpDate_Display_SENS(hObject,handles)
 TypOfDisp = get(handles.sens_display,'Value');
 ActToDisp = get(handles.Activity,'Value');
-ic        = 1:length(handles.sens_coord);
+
+im       = get(handles.modality, 'Value');
+ic       = handles.Ic{im};
 
 
 % topography
@@ -376,7 +407,7 @@ function UpDate_Display_SENS(hObject,handles)
     end
     
     axes(handles.sensors_axes);
-    disp = griddata(handles.sens_coord(:,1),handles.sens_coord(:,2),full(sens_disp),handles.sens_coord2D_X,handles.sens_coord2D_Y);
+    disp = griddata(handles.sens_coord(ic,1),handles.sens_coord(ic,2),full(sens_disp),handles.sens_coord2D_X,handles.sens_coord2D_Y);
     imagesc(handles.sens_coord_x,handles.sens_coord_y,disp);
     axis image xy off
     
@@ -385,11 +416,11 @@ function UpDate_Display_SENS(hObject,handles)
     try, delete(handles.sensor_loc); end
     hold(handles.sensors_axes, 'on');
     handles.sensor_loc = plot(handles.sensors_axes,...
-        handles.sens_coord(:,1),handles.sens_coord(:,2),'o','MarkerFaceColor',[1 1 1]/2,'MarkerSize',6);
+        handles.sens_coord(ic,1),handles.sens_coord(ic,2),'o','MarkerFaceColor',[1 1 1]/2,'MarkerSize',6);
     hold(handles.sensors_axes, 'off');
     
     axes(handles.pred_axes);
-    disp = griddata(handles.sens_coord(:,1),handles.sens_coord(:,2),full(pred_disp),handles.sens_coord2D_X,handles.sens_coord2D_Y);
+    disp = griddata(handles.sens_coord(ic,1),handles.sens_coord(ic,2),full(pred_disp),handles.sens_coord2D_X,handles.sens_coord2D_Y);
     imagesc(handles.sens_coord_x,handles.sens_coord_y,disp);
     axis image xy off;
     
@@ -401,7 +432,7 @@ function UpDate_Display_SENS(hObject,handles)
     axes(handles.sensors_axes)
     daspect('auto')
     handles.fig2 = ...
-        plot(handles.pst,handles.sens_data,'b-.',handles.pst,handles.pred_data,'r:');
+        plot(handles.pst,handles.sens_data(ic, :),'b-.',handles.pst,handles.pred_data(ic, :),'r:');
     if ActToDisp > 1
         hold on
         Scal = norm(handles.sens_data,1)/norm(handles.W,1);
@@ -468,7 +499,8 @@ function Activity_Callback(hObject, eventdata, handles)
 function Individual_Callback(hObject, eventdata, handles)
 set(handles.Template,'Value',0);
 try
-    handles.vert = handles.D.inv{handles.D.val}.mesh.tess_ctx.vert;
+    tess_ctx = gifti(handles.D.inv{handles.D.val}.mesh.tess_ctx);
+    handles.vert = tess_ctx.vertices;
     set(handles.Template,  'Value',0);
     set(handles.Individual,'Value',1);
 end
@@ -722,21 +754,39 @@ function con_Callback(hObject, eventdata, handles)
 set(handles.con,'String',sprintf('condition %d',handles.D.con),'Value',0);
 spm_eeg_inv_visu3D_api_OpeningFcn(hObject, eventdata, handles)
 
-
 % --- Executes on button press in VDE.
 %--------------------------------------------------------------------------
 function Velec_Callback(hObject, eventdata, handles)
 axes(handles.sources_axes);
-try
-    vde = getCursorInfo(handles.location);
-    spm_eeg_invert_display(handles.D,vde.Position)
-    datacursormode(handles.fig, 'off');
-catch
-    vde = [];
-    handles.location = datacursormode(handles.fig, 'on');
-    set(handles.location,'Enable','on','DisplayStyle','datatip','SnapToDataVertex','on', 'UpdateFcn', {@CursorUpdate_Callback, handles});
+rotate3d off;
+datacursormode off;
+
+if isfield(handles, 'velec')
+    delete(handles.velec);
+    handles = rmfield(handles, 'velec');
 end
 
+set(handles.fig1, 'ButtonDownFcn', @Velec_ButtonDown)
+
+guidata(hObject,handles);
+
+function Velec_ButtonDown(hObject, eventdata)
+handles     = guidata(hObject);
+vert        = handles.vert(handles.Is, :);
+coord       = get(handles.sources_axes, 'CurrentPoint');
+dist        = sum((vert - repmat(coord(1, :), size(vert, 1), 1)).^2, 2);
+[junk, ind] = min(dist);
+coord       = vert(ind, :);
+
+axes(handles.sources_axes);
+hold on
+handles.velec = plot3(coord(1), coord(2), coord(3), 'rv', 'MarkerSize', 10);
+
+spm_eeg_invert_display(handles.D, coord);
+
+set(handles.fig1, 'ButtonDownFcn', '');
+guidata(hObject,handles);
+
 
 % --- Executes on button press in Rot.
 function Rot_Callback(hObject, eventdata, handles)
@@ -745,7 +795,12 @@ function Rot_Callback(hObject, eventdata, handles)
 return
 
 
-function text = CursorUpdate_Callback(hObject, eventdata, handles)
-pos = get(eventdata,'Position');
-text = num2str(pos);
-spm_eeg_invert_display(handles.D, get(eventdata, 'Position'));
+% --- Executes on selection change in modality.
+function modality_Callback(hObject, eventdata, handles)
+% hObject    handle to modality (see GCBO)
+% eventdata  reserved - to be defined in a future version of MATLAB
+% handles    structure with handles and user data (see GUIDATA)
+
+% Hints: contents = cellstr(get(hObject,'String')) returns modality contents as cell array
+%        contents{get(hObject,'Value')} returns selected item from modality
+UpDate_Display_SENS(hObject,handles)
diff --git a/spm_eeg_invert.m b/spm_eeg_invert.m
index 3f3171a..c629d8a 100644
--- a/spm_eeg_invert.m
+++ b/spm_eeg_invert.m
@@ -14,7 +14,7 @@
 %                      'ARD' ARD search on MSPs
 %                      'MSP' GS and ARD multiple sparse priors
 %                      'LOR' LORETA-like model
-%                      'IID' LORETA and minimum norm
+%                      'IID' minimum norm
 %     inverse.woi    - time window of interest ([start stop] in ms)
 %     inverse.lpf    - band-pass filter - low frequency cutoff (Hz)
 %     inverse.hpf    - band-pass filter - high frequency cutoff (Hz)
@@ -120,7 +120,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_eeg_invert.m 3983 2010-07-09 13:56:01Z vladimir $
+% $Id: spm_eeg_invert.m 4254 2011-03-16 13:15:33Z vladimir $
 
 % check whether this is a group inversion for (Nl) number of subjects
 %--------------------------------------------------------------------------
diff --git a/spm_eeg_invert_display.m b/spm_eeg_invert_display.m
index 50f9276..43ec6e1 100644
--- a/spm_eeg_invert_display.m
+++ b/spm_eeg_invert_display.m
@@ -12,7 +12,7 @@ function spm_eeg_invert_display(D,PST,Ndip)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_eeg_invert_display.m 3813 2010-04-07 19:21:49Z karl $
+% $Id: spm_eeg_invert_display.m 4169 2011-01-24 18:34:20Z karl $
  
 % Number of dipoles to display
 %==========================================================================
@@ -47,11 +47,15 @@ function spm_eeg_invert_display(D,PST,Ndip)
 Is     = model.inverse.Is;
 pst    = model.inverse.pst;
 R2     = model.inverse.R2;
-VE     = model.inverse.VE;
 F      = model.inverse.F;
+Nd     = model.inverse.Nd;
 
-Nd     = length(Is);
-Ndip   = min(Ndip,Nd);
+Ndip   = min(Ndip,length(Is));
+try
+    VE = model.inverse.VE;
+catch
+    VE = R2;
+end
 
  
 % - project J onto pst
diff --git a/spm_eeg_lgainmat.m b/spm_eeg_lgainmat.m
index 13b0907..d6b1904 100644
--- a/spm_eeg_lgainmat.m
+++ b/spm_eeg_lgainmat.m
@@ -1,5 +1,5 @@
 function [L,D] = spm_eeg_lgainmat(D,Is, channels)
-% loads (memory maps) or computes if necessary a gain matrix
+% loads or computes if necessary a gain matrix
 % FORMAT [L,D] = spm_eeg_lgainmat(D,Is)
 % D    - Data structure
 % Is   - indices of vertices
@@ -9,10 +9,10 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_eeg_lgainmat.m 3877 2010-05-07 19:49:35Z karl $
+% $Id: spm_eeg_lgainmat.m 4082 2010-10-07 16:11:48Z guillaume $
 
 
-% get gain or lead-feild matrix
+% get gain or lead-field matrix
 %--------------------------------------------------------------------------
 val = D.val;
 
@@ -22,7 +22,7 @@
     modality = forward(ind).modality;
     
     % channels
-    %--------------------------------------------------------------------------
+    %----------------------------------------------------------------------
     chanind = strmatch(modality, D.chantype);
     chanind = setdiff(chanind, D.badchannels);
 
@@ -43,24 +43,23 @@
     
     label = G.label;
     G     = G.G;
-    if numel(label) ~= size(G, 1) ||...
-            ~all(ismember(channels, label))
+    if numel(label) ~= size(G, 1) || ~all(ismember(channels, label))
         error('Gain matrix has an incorrect number of channels');
     end
 catch
     label = {};
     for ind = 1:numel(forward)
         % create a new lead-field matrix
-        %----------------------------------------------------------------------
+        %------------------------------------------------------------------
 
         % Head Geometry (create tesselation file)
-        %--------------------------------------------------------------------------
+        %------------------------------------------------------------------
         vert = forward(ind).mesh.vert;
         face = forward(ind).mesh.face;
 
         % normals
-        %--------------------------------------------------------------------------
-        norm = spm_eeg_inv_normals(vert,face);
+        %------------------------------------------------------------------
+        norm = spm_mesh_normals(struct('faces',face,'vertices',vert),true);
 
         vol  = forward(ind).vol;
         
@@ -72,7 +71,7 @@
         modality = forward(ind).modality;     
 
         % Forward computation
-        %--------------------------------------------------------------------------
+        %------------------------------------------------------------------
         [vol, sens] = ft_prepare_vol_sens(vol, sens, 'channel', forward(ind).channels);
         nvert = size(vert, 1);
 
@@ -121,7 +120,7 @@
     end
     
     % Save
-    %--------------------------------------------------------------------------
+    %----------------------------------------------------------------------
     D.inv{val}.gainmat = ['SPMgainmatrix_' spm_str_manip(D.fname, 'tr') '_' num2str(val) '.mat'];
     save(fullfile(D.path, D.inv{val}.gainmat), 'G', 'label');
     save(D);
diff --git a/spm_eeg_merge.m b/spm_eeg_merge.m
index 547c6de..196ba64 100644
--- a/spm_eeg_merge.m
+++ b/spm_eeg_merge.m
@@ -56,9 +56,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 %
 % Stefan Kiebel, Vladimir Litvak, Doris Eckstein, Rik Henson
-% $Id: spm_eeg_merge.m 3563 2009-11-12 15:06:05Z vladimir $
+% $Id: spm_eeg_merge.m 4223 2011-03-02 10:30:45Z vladimir $
 
-SVNrev = '$Rev: 3563 $';
+SVNrev = '$Rev: 4223 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -95,8 +95,11 @@
 
 %-Check input and determine number of new number of trial types
 %--------------------------------------------------------------------------
-Ntrials = 0;
-isTF   =  strncmpi(D{1}.transformtype,'TF',2); % TF and TFphase
+Ntrials = [];
+megsens = [];
+eegsens = [];
+fid     = [];
+isTF    =  strncmpi(D{1}.transformtype,'TF',2); % TF and TFphase
 
 for i = 1:Nfiles
     if ~isequal(D{i}.transformtype, D{1}.transformtype)
@@ -129,10 +132,21 @@
                D{1}.fname, D{i}.fname);
     end
 
+    if ~isempty(D{i}.sensors('MEG'))
+        megsens = [megsens D{i}.sensors('MEG')];
+    end
+    
+    if ~isempty(D{i}.sensors('EEG'))
+        eegsens = [eegsens D{i}.sensors('EEG')];
+    end
+    
+    if ~isempty(megsens) || ~isempty(eegsens)
+        fid = [fid D{i}.fiducials];
+    end
+    
     Ntrials = [Ntrials D{i}.ntrials];
 end
 
-
 %-Prepare some useful lists
 %--------------------------------------------------------------------------
 F        = {};
@@ -276,6 +290,32 @@
     end
 end
             
+%-Average sensor locations
+%--------------------------------------------------------------------------
+if ~isempty(megsens)
+    spm_figure('GetWin','Graphics');clf;
+    if ~isempty(eegsens)
+        h = subplot(2, 1, 1);
+        aeegsens = ft_average_sens(eegsens, 'weights', Ntrials, 'feedback', h);
+        Dout = sensors(Dout, 'EEG', aeegsens);
+        
+        h = subplot(2, 1, 2);
+    else
+        h = axes;
+    end
+    
+    [amegsens afid] = ft_average_sens(megsens, 'fiducials', fid, 'weights', Ntrials, 'feedback', h);
+    Dout = sensors(Dout, 'MEG', amegsens);
+    Dout = fiducials(Dout, afid);
+elseif ~isempty(eegsens)
+    spm_figure('GetWin','Graphics');clf;
+    h = axes;
+    [aeegsens afid] = ft_average_sens(eegsens, 'fiducials', fid, 'weights', Ntrials, 'feedback', h);
+    Dout = sensors(Dout, 'EEG', aeegsens);
+    Dout = fiducials(Dout, afid);
+end
+
+
 
 %-Write files
 %--------------------------------------------------------------------------
diff --git a/spm_eeg_review_switchDisplay.m b/spm_eeg_review_switchDisplay.m
index 7b968b0..940ca9d 100644
--- a/spm_eeg_review_switchDisplay.m
+++ b/spm_eeg_review_switchDisplay.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jean Daunizeau
-% $Id: spm_eeg_review_switchDisplay.m 3571 2009-11-13 16:16:48Z vladimir $
+% $Id: spm_eeg_review_switchDisplay.m 4136 2010-12-09 22:22:28Z guillaume $
 
 try % only if already displayed stuffs
     handles = rmfield(D.PSD.handles,'PLOT');
@@ -348,7 +348,7 @@
         'xdata',[gridTime;gridTime],...
         'ydata',[miJ;maJ],...
         'visible','off');
-    set(D.PSD.handles.axes2,....
+    set(D.PSD.handles.axes2,...
         'ylim',[miJ;maJ]);  
 
     % create buttons
diff --git a/spm_eeg_specest_mtmspec.m b/spm_eeg_specest_mtmspec.m
index 26ffe8b..3e3c8bd 100644
--- a/spm_eeg_specest_mtmspec.m
+++ b/spm_eeg_specest_mtmspec.m
@@ -4,10 +4,8 @@
 %
 % S                     - input structure
 % fields of S:
-%    S.taper       - taper to use ('hanning', 'rectwin', 'dpss', 'sine' or
-%                    other possible inputs of 'window'
-%    S.freqres     - frequency resolutions (plus-minus for each frequency, can
-%                    be a vector with a value per frequency)
+%    S.bandwidth   - time bandwidth parameter determining the degree of
+%                    spectral smoothing (typically 3 or 4).
 %    S.frequencies - vector of frequencies
 %    S.timeres     - time resolution in ms (length of the sliding time-window)
 %    S.timestep    - time step (in ms) to slide the time-window by.
@@ -26,7 +24,7 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak based on the code contributed by Krish Singh
-% $Id: spm_eeg_specest_mtmspec.m 3876 2010-05-07 18:51:03Z vladimir $
+% $Id: spm_eeg_specest_mtmspec.m 4021 2010-07-28 12:43:16Z vladimir $
 
 
 %-This part if for creating a config branch that plugs into spm_cfg_eeg_tf
diff --git a/spm_eeg_tf.m b/spm_eeg_tf.m
index 2f34ec9..a7ec8d6 100644
--- a/spm_eeg_tf.m
+++ b/spm_eeg_tf.m
@@ -42,9 +42,9 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_eeg_tf.m 3842 2010-04-27 15:09:48Z vladimir $
+% $Id: spm_eeg_tf.m 4021 2010-07-28 12:43:16Z vladimir $
 
-SVNrev = '$Rev: 3842 $';
+SVNrev = '$Rev: 4021 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -87,7 +87,7 @@
     error('No channels selected.');
 end
 
-if ~isfield(S, 'frequencies')
+if ~isfield(S, 'frequencies') || isempty(S.frequencies)
     S.frequencies = 1:48;
 end
 
diff --git a/spm_eeg_tf_rescale.m b/spm_eeg_tf_rescale.m
index 46e65f1..8528507 100644
--- a/spm_eeg_tf_rescale.m
+++ b/spm_eeg_tf_rescale.m
@@ -23,9 +23,9 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_eeg_tf_rescale.m 3938 2010-06-18 15:03:28Z vladimir $
+% $Id: spm_eeg_tf_rescale.m 4287 2011-04-04 13:55:54Z vladimir $
 
-SVNrev = '$Rev: 3938 $';
+SVNrev = '$Rev: 4287 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -45,8 +45,8 @@
 try
     S.tf.method;
 catch
-    str  = {'LogR','Diff', 'Rel', 'Log','Sqrt'};
-    S.tf.method = spm_input('Rescale method','+1','b',str,[],1);
+    str  = {'LogR','Diff', 'Rel', 'Log', 'Sqrt', 'Zscore'};
+    S.tf.method = spm_input('Rescale method','+1','m',str,char(str),1);
 end
 
 Din  = spm_eeg_load(D);
@@ -57,7 +57,7 @@
 
 switch lower(S.tf.method)
     
-    case {'logr','diff', 'rel'}
+    case {'logr','diff', 'rel', 'zscore'}
         try
             S.tf.Sbaseline;
         catch            
@@ -91,17 +91,21 @@
                 xbase=spm_squeeze(Db(:,:,:,c), 4);
             else
                 xbase=spm_squeeze(Db(:,:,:,1), 4);
-            end
-            xbase=mean(xbase(:,:,inds),3);
+            end            
             switch lower(S.tf.method)
                 case 'logr'
-                    x=log10(x);
-                    xbase=mean(x(:,:,inds),3);
+                    xbase=mean(log10(xbase(:,:,inds)),3);
                     D(:,:,:,c)= 10*(x - repmat(xbase,[1 1 D.nsamples 1]));
                     D = units(D, [], 'dB');
                 case 'diff'
+                    xbase=mean(xbase(:,:,inds),3);
                     D(:,:,:,c)= (x - repmat(xbase,[1 1 D.nsamples 1]));
+                case 'zscore'
+                    stdev = std(xbase(:,:,inds), [], 3);
+                    xbase= mean(xbase(:,:,inds),3);                    
+                    D(:,:,:,c)= (x - repmat(xbase,[1 1 D.nsamples 1]))./repmat(stdev,[1 1 D.nsamples 1]);
                 case 'rel'
+                    xbase=mean(xbase(:,:,inds),3);
                     D(:,:,:,c)= 100*((x./repmat(xbase,[1 1 D.nsamples 1]) - 1));
                     D = units(D, [], '%');
             end
diff --git a/spm_eeg_wrap_dipfit_vbecd.m b/spm_eeg_wrap_dipfit_vbecd.m
index 044eaaa..8019864 100644
--- a/spm_eeg_wrap_dipfit_vbecd.m
+++ b/spm_eeg_wrap_dipfit_vbecd.m
@@ -13,7 +13,7 @@
 
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 %
-% $Id: spm_eeg_wrap_dipfit_vbecd.m 3908 2010-06-01 11:06:53Z gareth $
+% $Id: spm_eeg_wrap_dipfit_vbecd.m 4154 2011-01-11 14:39:36Z gareth $
 
 x=U.u; %% input , unused
 
@@ -47,11 +47,11 @@
         tmp = tmp - repmat(mean(tmp), size(tmp,1), 1); %% should this be here ?
     else %% reduce rank of leadfield for MEG- assume one direction (radial) is silent
         [tmp] = ft_compute_leadfield(pos, sens, vol,'reducerank',MEGRANK);
-        if isfield(vol, 'type') && strcmp(vol.type,'nolte'),
-            %% this is a temp fix to make up for scaling changes in nolte
-            %% model
-            tmp=tmp.*1e-10;
-        end;
+%         if isfield(vol, 'type') && strcmp(vol.type,'nolte'), %% 
+%             %% this is a temp fix to make up for scaling changes in nolte
+%             %% model
+%             tmp=tmp.*1e-10;
+%         end;
     end
     gmn=tmp;
     y=y+gmn*mom';
diff --git a/spm_est_smoothness.m b/spm_est_smoothness.m
index aa47255..28906e5 100644
--- a/spm_est_smoothness.m
+++ b/spm_est_smoothness.m
@@ -53,7 +53,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Stefan Kiebel, Tom Nichols
-% $Id: spm_est_smoothness.m 3960 2010-06-30 17:41:24Z ged $
+% $Id: spm_est_smoothness.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 %-Assign input arguments
@@ -187,7 +187,7 @@
 end
 if D == 2
     resel_xyz = [L(:,1,1) L(:,2,2)];
-    resel_img = L(:,1,1).*L(:,2,2) - ......
+    resel_img = L(:,1,1).*L(:,2,2) - ...
                 L(:,1,2).*L(:,1,2);
 end
 if D == 3
diff --git a/spm_fMRI_design.m b/spm_fMRI_design.m
index 3b49dc6..5139ada 100644
--- a/spm_fMRI_design.m
+++ b/spm_fMRI_design.m
@@ -164,10 +164,10 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_fMRI_design.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_fMRI_design.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
-SVNid = '$Rev: 3691 $';
+SVNid = '$Rev: 4185 $';
 
 %-GUI setup
 %--------------------------------------------------------------------------
@@ -365,10 +365,10 @@
     %-End: Save SPM.mat
     %----------------------------------------------------------------------
     fprintf('%-40s: ','Saving fMRI design')                             %-#
-    if spm_matlab_version_chk('7') >= 0,
-        save('SPM', 'SPM', '-V6');
+    if spm_check_version('matlab','7') >= 0,
+        save('SPM.mat', 'SPM', '-V6');
     else
-        save('SPM', 'SPM');
+        save('SPM.mat', 'SPM');
     end
     fprintf('%30s\n','...SPM.mat saved');
 end
diff --git a/spm_fieldindices.m b/spm_fieldindices.m
new file mode 100644
index 0000000..b5db593
--- /dev/null
+++ b/spm_fieldindices.m
@@ -0,0 +1,37 @@
+function [i] = spm_fieldindices(X,varargin)
+% returns the indices of fields in a structure
+% FORMAT [i] = spm_fieldindices(X,feild1,feild2,...)
+%
+% X         - structure
+% feild1,.. – fields
+%
+% i         - vector of indices
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fieldindices.m 4100 2010-10-22 19:49:17Z karl $
+ 
+ 
+% create structure of zeros
+%--------------------------------------------------------------------------
+X0    = spm_vec(X)*0;
+ix    = X0;
+X0    = spm_unvec(X0,X);
+ 
+% and add one to specified fields
+%--------------------------------------------------------------------------
+for i = 1:length(varargin)
+    
+    x  = X0;
+    f  = getfield(x,varargin{i});
+    f  = spm_unvec(spm_vec(f) + 1,f);
+    x  = setfield(x,varargin{i},f);
+    ix = ix + spm_vec(x);
+    
+end
+ 
+% find indices
+%--------------------------------------------------------------------------
+i = find(ix);
diff --git a/spm_figure.m b/spm_figure.m
index 834cd6e..106a601 100644
--- a/spm_figure.m
+++ b/spm_figure.m
@@ -55,7 +55,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_figure.m 3953 2010-06-28 16:58:48Z guillaume $
+% $Id: spm_figure.m 4272 2011-03-29 17:39:38Z guillaume $
 
 
 %==========================================================================
@@ -155,7 +155,10 @@
 % Creates toolbar in figure F (defaults to gcf). F can be a 'Tag'
 %
 % FORMAT spm_figure('ColorMap')
-% Callback for "ColorMap" buttons
+% Callback for "ColorMap" menu
+%
+% FORMAT spm_figure('FontSize')
+% Callback for "FontSize" menu
 %__________________________________________________________________________
 
 
@@ -296,7 +299,7 @@
     end 
 end
 set(F,'Pointer','Arrow')
-if ~isdocked, movegui(F); end
+if ~isdocked && ~spm('CmdLine'), movegui(F); end
 
 %==========================================================================
 case 'close'
@@ -359,25 +362,26 @@
 
 %-Print
 if ~iPaged
-    spm_print(fname)
+    spm_print(fname,F);
 else
     hPg    = get(hNextPage,'UserData');
     Cpage  = get(hPageNo,  'UserData');
     nPages = size(hPg,1);
 
-    set([hNextPage,hPrevPage,hPageNo],'Visible','off')
+    set([hNextPage,hPrevPage,hPageNo],'Visible','off');
     if Cpage~=1
-        set(hPg{Cpage,1},'Visible','off'), end
+        set(hPg{Cpage,1},'Visible','off');
+    end
     for p = 1:nPages
         set(hPg{p,1},'Visible','on');
-        spm_print(fname);
-        set(hPg{p,1},'Visible','off')
+        spm_print(fname,F);
+        set(hPg{p,1},'Visible','off');
     end
-    set(hPg{Cpage,1},'Visible','on')
-    set([hNextPage,hPrevPage,hPageNo],'Visible','on')
+    set(hPg{Cpage,1},'Visible','on');
+    set([hNextPage,hPrevPage,hPageNo],'Visible','on');
 end
-if ~isempty(H), set(H,{'Units'},un); end;
-set(0,'CurrentFigure',cF)
+if ~isempty(H), set(H,{'Units'},un); end
+set(0,'CurrentFigure',cF);
 
 %==========================================================================
 case 'printto'
@@ -731,14 +735,18 @@
 
 %-Copy Figure
 if ispc
-    uimenu(t0, 'Label','&Copy Figure', 'HandleVisibility','off',...
+    uimenu(t0, 'Label','Co&py Figure', 'HandleVisibility','off',...
        'CallBack','editmenufcn(gcbf,''EditCopyFigure'')');
 end
 
 %-Clear Menu
-uimenu(t0,    'Label','C&lear Figure', 'HandleVisibility','off', ...
+uimenu(t0,    'Label','&Clear Figure', 'HandleVisibility','off', ...
     'CallBack','spm_figure(''Clear'',gcbf)');
 
+%-Close non-SPM figures
+uimenu(t0,    'Label','C&lose non-SPM Figures', 'HandleVisibility','off', ...
+    'CallBack',@myclosefig);
+
 %-Colour Menu
 t1=uimenu(t0, 'Label','C&olours',  'HandleVisibility','off','Separator','on');
 t2=uimenu(t1, 'Label','Colormap');
@@ -755,6 +763,11 @@
 uimenu(t2,    'Label','Brighten',  'CallBack','spm_figure(''ColorMap'',''brighten'')');
 uimenu(t2,    'Label','Darken',    'CallBack','spm_figure(''ColorMap'',''darken'')');
 
+%-Font Size Menu
+t1=uimenu(t0, 'Label','&Font Size', 'HandleVisibility','off');
+uimenu(t1, 'Label','&Increase', 'CallBack','spm_figure(''FontSize'',1)');
+uimenu(t1, 'Label','&Decrease', 'CallBack','spm_figure(''FontSize'',-1)');
+
 %-Satellite Table
 uimenu(t0,    'Label','&Results Table', 'HandleVisibility','off', ...
     'Separator','on', 'Callback',@mysatfig);
@@ -817,6 +830,18 @@
     error('Illegal ColAction specification');
 end
 
+%==========================================================================
+case 'fontsize'
+%==========================================================================
+% spm_figure('FontSize',sz)
+
+if nargin<2, sz=0; else sz=varargin{2}; end
+
+h  = [get(0,'CurrentFigure') spm_figure('FindWin','Satellite')];
+h  = findall(h,'type','text');
+fs = get(h,'fontsize');
+set(h,{'fontsize'},cellfun(@(x) x+sz,fs,'UniformOutput',false)) ;
+
 %==========================================================================
 otherwise
 %==========================================================================
@@ -910,6 +935,16 @@ function mydockspm(obj,evt)
 set(h,'WindowStyle','docked');
 try, pause(0.5), desktop.setGroupDocked(group,false); end
 
+%==========================================================================
+function myclosefig(obj,evt)
+%==========================================================================
+hMenu = spm_figure('FindWin','Menu');
+hInt  = spm_figure('FindWin','Interactive');
+hGra  = spm_figure('FindWin','Graphics');
+h     = setdiff(findobj(get(0,'children'),'flat','visible','on'), ...
+    [hMenu hInt hGra gcf]);
+close(h,'force');
+
 %==========================================================================
 function copy_menu(F,G)
 %==========================================================================
@@ -982,7 +1017,7 @@ function spm_about(obj,evt)
     'Color',[1 1 1]);
 text(0.5,0.30,'Wellcome Trust Centre for Neuroimaging','Parent',a,...
     'HorizontalAlignment','center','Color',[1 1 1],'FontWeight','Bold');
-text(0.5,0.25,['Copyright (C) 1991,1994-2003, 2005-' datestr(now,'yyyy')],...
+text(0.5,0.25,['Copyright (C) 1991,1994-' datestr(now,'yyyy')],...
     'Parent',a,'HorizontalAlignment','center','Color',[1 1 1]);
 text(0.5,0.20,'http://www.fil.ion.ucl.ac.uk/spm/','Parent',a,...
     'HorizontalAlignment','center','Color',[1 1 1],...
@@ -998,7 +1033,10 @@ function myscroll(obj,evt)
 %==========================================================================
 ax = findobj(gcf,'Tag','textcont');
 cla(ax);
-authors = spm_authors;
+[current, previous] = spm_authors;
+authors = {'*SPM8*' current{:} '' ...
+           '*Previous versions*' previous{:} '' ...
+           '*Thanks to the SPM community*'};
 x = 0.2;
 h = [];
 for i=1:numel(authors)
@@ -1033,43 +1071,3 @@ function myscroll(obj,evt)
     c = [1 1 1] - 6*abs(0.4-x);
 end
 c(c<0) = 0; c(c>1) = 1;
-
-%==========================================================================
-function authors = spm_authors
-%==========================================================================
-authors = {...
-'*SPM8*'
-'John Ashburner'
-'Gareth Barnes'
-'Chun-Chuan Chen'
-'Justin Chumbley'
-'Jean Daunizeau'
-'Guillaume Flandin'
-'Karl Friston'
-'Darren Gitelman'
-'Volkmar Glauche'
-'Lee Harrison'
-'Rik Henson'
-'Chloe Hutton'
-'Maria Joao Rosa'
-'Stefan Kiebel'
-'James Kilner'
-'Vladimir Litvak'
-'Rosalyn Moran'
-'Tom Nichols'
-'Robert Oostenveld'
-'Will Penny'
-'Christophe Phillips'
-'Klaas Enno Stephan'
-''
-'*Previous versions*'
-'Matthew Brett'
-'Christian Buechel'
-'Jon Heather'
-'Andrew Holmes'
-'Jeremie Mattout'
-'Jean-Baptiste Poline'
-'Keith Worsley'
-''
-'*Thanks to the SPM community*'
-};
diff --git a/spm_fileparts.m b/spm_fileparts.m
index cbd22e1..286ab1f 100644
--- a/spm_fileparts.m
+++ b/spm_fileparts.m
@@ -6,19 +6,17 @@
 % nam   - filename
 % ext   - extension
 % num   - comma separated list of values
-%
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_fileparts.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_fileparts.m 4205 2011-02-21 15:39:08Z guillaume $
 
 
 num = '';
 [pth,nam,ext] = fileparts(fname);
 ind = find(ext==',');
-if ~isempty(ind),
+if ~isempty(ind)
     num = ext(ind(1):end);
     ext = ext(1:(ind(1)-1));
-end;
-
+end
diff --git a/spm_filter.m b/spm_filter.m
index f524b0e..46220ae 100644
--- a/spm_filter.m
+++ b/spm_filter.m
@@ -1,4 +1,4 @@
-function [argout] = spm_filter(K,Y)
+function varargout = spm_filter(K,Y)
 % Removes low frequency confounds X0
 % FORMAT [Y] = spm_filter(K,Y)
 % FORMAT [K] = spm_filter(K)
@@ -11,7 +11,7 @@
 % K(s).HParam - cut-off period in seconds
 %
 % K(s).X0     - low frequencies to be removed (DCT)
-% 
+%
 % Y           - data matrix
 %
 % K           - filter structure
@@ -20,69 +20,75 @@
 %
 % spm_filter implements high-pass filtering in an efficient way by
 % using the residual forming matrix of X0 - low frequency confounds
-%.spm_filter also configures the filter structure in accord with the 
+%.spm_filter also configures the filter structure in accord with the
 % specification fields if called with one argument
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_filter.m 3303 2009-08-03 15:26:49Z guillaume $
+% $Id: spm_filter.m 4191 2011-02-03 13:30:02Z guillaume $
 
 
-% set or apply
-%--------------------------------------------------------------------------
+%-Configure filter
+%==========================================================================
 if nargin == 1 && isstruct(K)
-
+    
     % set K.X0
     %----------------------------------------------------------------------
     for s = 1:length(K)
-
+        
         % make high pass filter
         %------------------------------------------------------------------
         k       = length(K(s).row);
         n       = fix(2*(k*K(s).RT)/K(s).HParam + 1);
         X0      = spm_dctmtx(k,n);
         K(s).X0 = X0(:,2:end);
+        
     end
-
-    % return structure
+    
+    % return filter structure
     %----------------------------------------------------------------------
-    argout = K;
-
+    varargout = { K };
+    
+%-Apply filter
+%==========================================================================
 else
-    % apply
+    
+    % K is a filter structure
     %----------------------------------------------------------------------
     if isstruct(K)
-
+        
         % ensure requisite fields are present
         %------------------------------------------------------------------
         if ~isfield(K(1),'X0')
             K = spm_filter(K);
         end
-
+        
         for s = 1:length(K)
-
+            
             % select data
             %--------------------------------------------------------------
             y = Y(K(s).row,:);
-
+            
             % apply high pass filter
             %--------------------------------------------------------------
             y = y - K(s).X0*(K(s).X0'*y);
-
+            
             % reset filtered data in Y
             %--------------------------------------------------------------
             Y(K(s).row,:) = y;
-
+            
         end
-
+        
     % K is simply a filter matrix
     %----------------------------------------------------------------------
     else
-        Y = K*Y;
+        
+        Y = K * Y;
+        
     end
-
+    
     % return filtered data
     %----------------------------------------------------------------------
-    argout = Y;
+    varargout = { Y };
 end
diff --git a/spm_fmri_spm_ui.m b/spm_fmri_spm_ui.m
index 9c9917c..6c177e2 100644
--- a/spm_fmri_spm_ui.m
+++ b/spm_fmri_spm_ui.m
@@ -173,9 +173,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston, Jean-Baptiste Poline & Christian Buchel
-% $Id: spm_fmri_spm_ui.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_fmri_spm_ui.m 4178 2011-01-27 15:12:53Z guillaume $
 
-SVNid  = '$Rev: 3691 $';
+SVNid  = '$Rev: 4178 $';
 
 %-GUI setup
 %--------------------------------------------------------------------------
@@ -444,9 +444,9 @@
 %==========================================================================
 fprintf('%-40s: ','Saving SPM configuration')                           %-#
 if spm_matlab_version_chk('7') >= 0,
-    save('SPM', 'SPM', '-V6');
+    save('SPM.mat', 'SPM', '-V6');
 else
-    save('SPM', 'SPM');
+    save('SPM.mat', 'SPM');
 end
 fprintf('%30s\n','...SPM.mat saved')                                    %-#
 
diff --git a/spm_fp_fmin.m b/spm_fp_fmin.m
index 70a96df..459b110 100644
--- a/spm_fp_fmin.m
+++ b/spm_fp_fmin.m
@@ -15,7 +15,7 @@
 % Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_fp_fmin.m 2030 2008-09-02 18:28:40Z karl $
+% $Id: spm_fp_fmin.m 4136 2010-12-09 22:22:28Z guillaume $
  
  
 % specify function returning the flow-dependent part of dp/dt 'spm_fp_fun'
diff --git a/spm_fx_fmri.m b/spm_fx_fmri.m
index 12432bf..f5d4cc5 100644
--- a/spm_fx_fmri.m
+++ b/spm_fx_fmri.m
@@ -33,7 +33,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston & Klaas Enno Stephan
-% $Id: spm_fx_fmri.m 3888 2010-05-15 18:49:56Z karl $
+% $Id: spm_fx_fmri.m 4052 2010-08-27 19:22:44Z karl $
 
 
 % Neuronal motion
@@ -57,7 +57,7 @@
 % implement differential state equation y = dx/dt (neuronal)
 %--------------------------------------------------------------------------
 y    = x;
-if size(x,2) == 5
+if size(x,2) == 5  
     
     % one neuronal state per region
     %----------------------------------------------------------------------
@@ -97,7 +97,6 @@
 H        = [0.65 0.41 2.00 0.32 0.34];
 H        = [0.64 0.32 2.00 0.32 0.32];
 
-
 % exponentiation of hemodynamic state variables
 %--------------------------------------------------------------------------
 x(:,3:5) = exp(x(:,3:5));
diff --git a/spm_fx_fmri_linear.m b/spm_fx_fmri_linear.m
new file mode 100644
index 0000000..784f3c0
--- /dev/null
+++ b/spm_fx_fmri_linear.m
@@ -0,0 +1,122 @@
+function [y] = spm_fx_fmri_linear(x,u,P,M)
+% state equation for a dynamic model of fMRI (linear version)
+% responses
+% FORMAT [y] = spm_fx_fmri_linear(x,u,P,M)
+% x      - state vector
+%   x(:,1) - excitatory neuronal activity             ue
+%   x(:,2) - vascular signal                          s
+%   x(:,3) - rCBF                                  ln(f)
+%   x(:,4) - venous volume                         ln(v)
+%   x(:,5) - deoyxHb                               ln(q)
+%  [x(:,6) - inhibitory neuronal activity             ui]
+%
+% y      - dx/dt
+%
+%___________________________________________________________________________
+%
+% References for hemodynamic & neuronal state equations:
+% 1. Buxton RB, Wong EC & Frank LR. Dynamics of blood flow and oxygenation
+%    changes during brain activation: The Balloon model. MRM 39:855-864,
+%    1998.
+% 2. Friston KJ, Mechelli A, Turner R, Price CJ. Nonlinear responses in
+%    fMRI: the Balloon model, Volterra kernels, and other hemodynamics.
+%    Neuroimage 12:466-477, 2000.
+% 3. Stephan KE, Kasper L, Harrison LM, Daunizeau J, den Ouden HE,
+%    Breakspear M, Friston KJ. Nonlinear dynamic causal models for fMRI.
+%    Neuroimage 42:649-662, 2008.
+% 4. Marreiros AC, Kiebel SJ, Friston KJ. Dynamic causal modelling for
+%    fMRI: a two-state model.
+%    Neuroimage. 2008 Jan 1;39(1):269-78.
+%__________________________________________________________________________
+
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston & Klaas Enno Stephan
+% $Id: spm_fx_fmri_linear.m 4052 2010-08-27 19:22:44Z karl $
+
+
+% Neuronal motion
+%==========================================================================
+P.B   = full(P.B);                       % bi-linear parameters
+P.C   = P.C/16;                          % exogenous parameters
+P.D   = full(P.D);                       % nonlinear parameters
+
+% excitatory connections
+%--------------------------------------------------------------------------
+for i = 1:size(P.B,3)
+    P.A = P.A + u(i)*P.B(:,:,i);
+end
+
+% and nonlinear (state) terms
+%--------------------------------------------------------------------------
+for i = 1:size(P.D,3)
+    P.A = P.A + x(i,1)*P.D(:,:,i);
+end
+
+% implement differential state equation y = dx/dt (neuronal)
+%--------------------------------------------------------------------------
+y    = x;
+if size(x,2) == 5  
+    
+    % one neuronal state per region
+    %----------------------------------------------------------------------
+    y(:,1) = P.A*x(:,1) + P.C*u(:);
+
+else
+
+    % extrinsic (two neuronal states)
+    %----------------------------------------------------------------------
+    A      = exp(P.A)/8;             % enforce positivity
+    IE     = diag(diag(A));          % inhibitory to excitatory
+    EE     = A - IE;                 % excitatory to excitatory
+    EI     = 1;                      % excitatory to inhibitory
+    SE     = 1;                      % self-inhibition (excitatory)
+    SI     = 2;                      % self-inhibition (inhibitory)
+
+    % motion - excitatory and inhibitory: y = dx/dt
+    %----------------------------------------------------------------------
+    y(:,1) = EE*x(:,1) - SE*x(:,1) - IE*x(:,6) + P.C*u(:);
+    y(:,6) = EI*x(:,1) - SI*x(:,6);
+
+end
+
+% Hemodynamic motion
+%==========================================================================
+
+% hemodynamic parameters
+%--------------------------------------------------------------------------
+%   H(1) - signal decay                                   d(ds/dt)/ds)
+%   H(2) - autoregulation                                 d(ds/dt)/df)
+%   H(3) - transit time                                   (t0)
+%   H(4) - exponent for Fout(v)                           (alpha)
+%   H(5) - resting oxygen extraction                      (E0)
+%   H(6) - ratio of intra- to extra-vascular components   (epsilon)
+%          of the gradient echo signal
+%--------------------------------------------------------------------------
+H      = [0.65 0.41 2.00 0.32 0.34];
+H      = [0.64 0.32 2.00 0.32 0.32];
+
+% signal decay
+%--------------------------------------------------------------------------
+sd     = H(1)*exp(P.decay);
+
+% transit time
+%--------------------------------------------------------------------------
+tt     = H(3)*exp(P.transit);
+
+% Fout = f(v) - outflow
+%--------------------------------------------------------------------------
+fv     = 1 + x(:,4)/H(4);
+
+
+% e = f(f) - oxygen extraction x rCBF
+%--------------------------------------------------------------------------
+ff     = (1 - log(H(5))).*x(:,3);
+
+% implement differential state equation y = dx/dt (hemodynamic)
+%--------------------------------------------------------------------------
+y(:,2) = x(:,1) - sd.*x(:,2) - H(2)*x(:,3);
+y(:,3) = x(:,2);
+y(:,4) = (1  + x(:,3) - fv)./tt;
+y(:,5) = (ff - x(:,5) + x(:,4))./tt;
+y      = y(:);
diff --git a/spm_getSPM.m b/spm_getSPM.m
index d2019a7..e661739 100644
--- a/spm_getSPM.m
+++ b/spm_getSPM.m
@@ -182,7 +182,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes, Karl Friston & Jean-Baptiste Poline
-% $Id: spm_getSPM.m 3980 2010-07-08 15:55:37Z karl $
+% $Id: spm_getSPM.m 4225 2011-03-02 15:53:05Z guillaume $
 
 
 %-GUI setup
@@ -224,7 +224,7 @@
 %-Check the model has been estimated
 %--------------------------------------------------------------------------
 try
-    SPM.xVol.XYZ;
+    SPM.xVol.S;
 catch
     
     %-Check the model has been estimated
@@ -246,15 +246,6 @@
 M    = SPM.xVol.M(1:3,1:3);         %-voxels to mm matrix
 VOX  = sqrt(diag(M'*M))';           %-voxel dimensions
 
-% check the data and other files have valid filenames
-%--------------------------------------------------------------------------
-try, SPM.xY.VY     = spm_check_filename(SPM.xY.VY);     end
-try, SPM.xVol.VRpv = spm_check_filename(SPM.xVol.VRpv); end
-try, SPM.Vbeta     = spm_check_filename(SPM.Vbeta);     end
-try, SPM.VResMS    = spm_check_filename(SPM.VResMS);    end
-try, SPM.VM        = spm_check_filename(SPM.VM);        end
-
-
 %==========================================================================
 % - C O N T R A S T S ,   S P M    C O M P U T A T I O N ,    M A S K I N G
 %==========================================================================
@@ -497,15 +488,11 @@
     
     % Set Bayesian con type
     %----------------------------------------------------------------------
-    if length(SPM.PPM.xCon) < Ic
-        SPM.PPM.xCon(Ic).PSTAT = xCon(Ic).STAT;
-    end
+    SPM.PPM.xCon(Ic).PSTAT = xCon(Ic).STAT;
     
-    % Make this one a Bayesian contrast - if first level
+    % Make all new contrasts Bayesian contrasts 
     %----------------------------------------------------------------------
-    if isfield(SPM.PPM,'VB')
-        [xCon(Ic).STAT] = deal('P');
-    end
+    [xCon(Ic).STAT] = deal('P');
     
     if all(strcmp([SPM.PPM.xCon(Ic).PSTAT],'T'))
         
@@ -773,15 +760,15 @@
 XYZ    = XYZ(:,Q);
 if isempty(Q)
     fprintf('\n');                                                      %-#
-    warning('SPM:NoVoxels','No voxels survive masking at p=%4.2f',pm);
+    warning('SPM:NoVoxels','No voxels survive height threshold at u=%0.2g',u);
 end
 
 
-%-Extent threshold (disallowed for conjunctions)
+%-Extent threshold
 %--------------------------------------------------------------------------
-if ~isempty(XYZ) && nc == 1
+if ~isempty(XYZ)
     
-    fprintf('%s%30s',repmat(sprintf('\b'),1,30),'...extent threshold')  %-#
+    fprintf('%s%30s',repmat(sprintf('\b'),1,30),'...extent threshold'); %-#
     
     %-Get extent threshold [default = 0]
     %----------------------------------------------------------------------
@@ -797,7 +784,7 @@
     Q     = [];
     for i = 1:max(A)
         j = find(A == i);
-        if length(j) >= k; Q = [Q j]; end
+        if length(j) >= k, Q = [Q j]; end
     end
     
     % ...eliminate voxels
@@ -806,7 +793,7 @@
     XYZ   = XYZ(:,Q);
     if isempty(Q)
         fprintf('\n');                                                  %-#
-        warning('SPM:NoVoxels','No voxels survive masking at p=%4.2f',pm);
+        warning('SPM:NoVoxels','No voxels survive extent threshold at k=%0.2g',k);
     end
     
 else
diff --git a/spm_get_bbox.m b/spm_get_bbox.m
new file mode 100644
index 0000000..5e4bb5b
--- /dev/null
+++ b/spm_get_bbox.m
@@ -0,0 +1,107 @@
+function [BB vx] = spm_get_bbox(V, thr, premul)
+% Compute volume's bounding box, for full field of view or object bounds
+% FORMAT [BB vx] = spm_get_bbox(V, thr)
+% V   - mapped image volume(s) (from spm_vol) or filename (empty for GUI)
+% thr - threshold, such that BB contains voxels with intensities > thr
+%       or strings 'nz', 'nn', fv', for non-zero, non-NaN, or field of view
+%       where 'fv' (the default) uses only the image's header information.
+%
+% BB  - a [2 x 3] array of the min and max X, Y, and Z coordinates {mm},
+%       i.e. BB = [minX minY minZ; maxX maxY maxZ].
+% vx  - a [1 x 3] vector of voxel dimensions {mm}.
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+
+% Ged Ridgway
+% $Id: spm_get_bbox.m 4205 2011-02-21 15:39:08Z guillaume $
+
+% Undocumented expert options:
+% V           - can be a 4D @nifti object (but not 5D), image-based BBs
+%               will be computed using "all" along the 4th dimension.
+% thr = 'old' - reproduce spm_write_sn/bbvox_from_V (and elsewhere)
+% premul      - a matrix that premultiplies V.mat, as used in spm_orthviews
+
+%-Get an SPM volume structure
+%--------------------------------------------------------------------------
+if nargin < 1 || isempty(V)
+    [V, sts] = spm_select(1, 'image', 'Select Image');
+    if ~sts, error('Must select an image'), end
+end
+if ischar(V), V = spm_vol(V); end
+
+%-Get volume structure from @nifti object if given
+%--------------------------------------------------------------------------
+if isa(V, 'nifti')
+    V = spm_vol(V.dat.fname); % (potentially a struct array of volumes)
+end
+
+%-Compute voxel dimensions (for compatibility with bbvox_from_V)
+%--------------------------------------------------------------------------
+P = spm_imatrix(V(1).mat);
+vx = P(7:9);
+% the above agrees with sqrt(sum(V.mat(1:3,1:3).^2)) for simple rotations,
+% and seems more appropriate if there are reflections and/or skews.
+% Note that spm_imatrix(diag([-1 1 1 1])) is [-1 1 1] as expected.
+
+%-Compute bounding box
+%--------------------------------------------------------------------------
+if nargin < 2 || isempty(thr) || strcmpi(thr, 'fv')
+    % overall field-of-view bounding box from header information
+    d = V(1).dim;
+    corners = [
+        1    1    1    1
+        1    1    d(3) 1
+        1    d(2) 1    1
+        1    d(2) d(3) 1
+        d(1) 1    1    1
+        d(1) 1    d(3) 1
+        d(1) d(2) 1    1
+        d(1) d(2) d(3) 1
+        ]';
+    XYZ = V(1).mat(1:3, :) * corners;
+elseif strcmpi(thr, 'old')
+    % code from spm_write_sn/bbvox_from_V (and other places)
+    % NB: main difference is that vx(1)<0 gives descending BB(:,1),
+    % shouldn't be used if V.mat contains rotations or skews.
+    o  = V(1).mat\[0 0 0 1]';
+    o  = o(1:3)';
+    BB = [-vx.*(o-1) ; vx.*(V(1).dim(1:3)-o)];
+    if exist('premul', 'var')
+        warning('spm_get_bbox:old_and_premul', 'old method ignores premul')
+    end
+else
+    % image-based bounding box using voxel intensities
+    [img XYZ] = spm_read_vols(V);
+    if ischar(thr)
+        switch lower(thr)
+            case 'nn'  % non-NaN, though include +/- Inf in computation
+                img = ~isnan(img);
+            case 'nz'  % special case of non-zero (rather than > 0)
+                img = ~isnan(img) & img ~= 0;
+            otherwise
+                error('Unknown threshold type %s', thr)
+        end
+    else
+        % treat thr as numeric threshold
+        img = img > thr;
+    end
+    if ndims(img) == 4
+        img = all(img, 4);
+    end
+    if nnz(img) == 0
+        warning('spm_get_bbox:nothing', ...
+            'Threshold leaves no voxels, returning full field of view');
+    else
+        XYZ = XYZ(:, img); % keep only coords that satisfy condition
+    end
+end
+
+if ~exist('BB', 'var') % exists already if 'old' case chosen above
+    if exist('premul', 'var')
+        XYZ = premul(1:3, :) * [XYZ; ones(1, size(XYZ, 2))];
+    end
+    BB = [
+        min(XYZ, [], 2)'
+        max(XYZ, [], 2)'
+        ];
+end
diff --git a/spm_get_ons.m b/spm_get_ons.m
index f867116..366475e 100644
--- a/spm_get_ons.m
+++ b/spm_get_ons.m
@@ -1,8 +1,9 @@
 function [U] = spm_get_ons(SPM,s)
-% returns input [designed effects] structures
+% Returns input [designed effects] structures
 % FORMAT [U] = spm_get_ons(SPM,s)
 %
-% s  - session number (used by batch system)
+% SPM   - SPM structure (see spm_fMRI_design.m)
+% s     - session number
 %
 % U     - (1 x n)   struct array of (n) trial-specific structures
 %
@@ -19,30 +20,29 @@
 %       U(i).P(p).i    - sub-indices of u pertaining to P
 %__________________________________________________________________________
 %
-%
 % SLICE TIMING
 %
 % With longs TRs you may want to shift the regressors so that they are
-% aligned to a particular slice.  This is effected by resetting the
+% aligned to a particular slice. This is effected by resetting the
 % values of defaults.stats.fmri.t and defaults.stats.fmri.t0 in
-% spm_defaults. defaults.stats.fmri.t is the number of time-bins per
-% scan used when building regressors.  Onsets are defined
-% in temporal units of scans starting at 0.  defaults.stats.fmri.t0 is
-% the first time-bin at which the regressors are resampled to coincide
-% with data acquisition.  If defaults.stats.fmri.t0 = 1 then the
-% regressors will be appropriate for the first slice.  If you want to
-% temporally realign the regressors so that they match responses in the
-% middle slice then make defaults.stats.fmri.t0 =
+% spm_defaults. 
+% defaults.stats.fmri.t is the number of time-bins per scan used when
+% building regressors. Onsets are defined in temporal units of scans
+% starting at 0.
+% defaults.stats.fmri.t0 is the first time-bin at which the regressors are
+% resampled to coincide with data acquisition. If defaults.stats.fmri.t0
+% is set to 1 then the regressors will be appropriate for the first slice.
+% If you want to temporally realign the regressors so that they match
+% responses in the middle slice then make defaults.stats.fmri.t0 =
 % defaults.stats.fmri.t/2 (assuming there is a negligible gap between
-% volume acquisitions. Default values are defaults.stats.fmri.t = 16
-% and defaults.stats.fmri.t0 = 1.
-%
+% volume acquisitions).
+% Default values are defaults.stats.fmri.t=16 and defaults.stats.fmri.t0=1.
 %
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_get_ons.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_get_ons.m 4083 2010-10-08 10:31:55Z guillaume $
 
 
 % time units
@@ -70,7 +70,7 @@
         error('Unknown unit "%s".',UNITS);
 end
 
-% get inputs and names (try SPM.Sess(s).U first)
+% get inputs and names
 %==========================================================================
 try
     U   = SPM.Sess(s).U;
diff --git a/spm_get_space.m b/spm_get_space.m
index d727575..bbf0023 100644
--- a/spm_get_space.m
+++ b/spm_get_space.m
@@ -4,60 +4,54 @@
 %            spm_get_space(P,M)
 % M - voxel-to-world mapping
 % P - image filename
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_get_space.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_get_space.m 4208 2011-02-22 12:04:18Z guillaume $
 
 
-[pth,nam,ext] = fileparts(P);
-t = find(ext==',');
-n = [1 1];
-if ~isempty(t),
-    n   = str2num(ext((t(1)+1):end));
-    ext = ext(1:(t-1));
-    P   = fullfile(pth,[nam ext]);
-end;
+[pth,nam,ext,num] = spm_fileparts(P);
+if ~isempty(num), n = str2num(num(2:end)); else n = [1 1]; end
+P = fullfile(pth,[nam ext]);
 
 N = nifti(P);
-if nargin==2,
+if nargin==2
     N.mat_intent = 'Aligned';
-    if n(1)==1,
+    if n(1)==1
 
         % Ensure volumes 2..N have the original matrix
-        if size(N.dat,4)>1 && sum(sum((N.mat-M).^2))>1e-8,
+        if size(N.dat,4)>1 && sum(sum((N.mat-M).^2))>1e-8
             M0 = N.mat;
-            if ~isfield(N.extras,'mat'),
+            if ~isfield(N.extras,'mat')
                 N.extras.mat = zeros([4 4 size(N.dat,4)]);
             else
-                if size(N.extras.mat,4)<size(N.dat,4),
+                if size(N.extras.mat,4)<size(N.dat,4)
                     N.extras.mat(:,:,size(N.dat,4)) = zeros(4);
-                end;
-            end;
-            for i=2:size(N.dat,4),
+                end
+            end
+            for i=2:size(N.dat,4)
                 if sum(sum(N.extras.mat(:,:,i).^2))==0,
                     N.extras.mat(:,:,i) = M0;
-                end;
-            end;
-        end;
+                end
+            end
+        end
 
-        N.mat        = M;
-        if strcmp(N.mat0_intent,'Aligned'), N.mat0 = M; end;
+        N.mat = M;
+        if strcmp(N.mat0_intent,'Aligned'), N.mat0 = M; end
         if ~isempty(N.extras) && isstruct(N.extras) && isfield(N.extras,'mat') &&...
-            size(N.extras.mat,3)>=1,
+            size(N.extras.mat,3)>=1
             N.extras.mat(:,:,n(1)) = M;
-        end;
+        end
     else
         N.extras.mat(:,:,n(1)) = M;
-    end;
+    end
     create(N);
 else
     if ~isempty(N.extras) && isstruct(N.extras) && isfield(N.extras,'mat') &&...
-        size(N.extras.mat,3)>=n(1) && sum(sum(N.extras.mat(:,:,n(1)).^2)),
+        size(N.extras.mat,3)>=n(1) && sum(sum(N.extras.mat(:,:,n(1)).^2))
         M = N.extras.mat(:,:,n(1));
     else
         M = N.mat;
-    end;
-end;
-
+    end
+end
diff --git a/spm_global.m b/spm_global.m
index 78cd233..69f442b 100644
--- a/spm_global.m
+++ b/spm_global.m
@@ -1,22 +1,23 @@
 function GX = spm_global(V)
-% returns the global mean for a memory mapped volume image - a compiled routine
+% returns the global mean for a volume image - a compiled routine
 % FORMAT GX = spm_global(V)
 % V   - memory mapped volume
 % GX  - mean global activity
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_global returns the mean counts integrated over all the  
-% slices from the volume
+% slices from the volume.
 %
 % The mean is estimated after discounting voxels outside the object
-% using a criteria of greater than > (global mean)/8
+% using a criteria of greater than > (global mean)/8.
 %
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Anonymous
-% $Id: spm_global.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_global.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-This is merely the help file for the compiled routine
 error('spm_global.c not compiled - see Makefile')
+
diff --git a/spm_graph.m b/spm_graph.m
index 5a9530c..ee4b426 100644
--- a/spm_graph.m
+++ b/spm_graph.m
@@ -5,14 +5,14 @@
 % xSPM   - structure containing SPM, distributional & filtering details
 %          about the excursion set
 % SPM    - structure containing generic details about the analysis
-% hReg   - handle of MIP register
+% hReg   - handle of MIP register or [x y z] coordinates
 %
 % Y      - fitted   data for the selected voxel
 % y      - adjusted data for the selected voxel
 % beta   - parameter estimates (ML or MAP)
 % Bcov   - Covariance of parameter estimates (ML or conditional)
 %
-% see spm_getSPM for details
+% See spm_getSPM for details.
 %__________________________________________________________________________
 %
 % spm_graph is a Callback script that uses the structures above to:  (i)
@@ -22,40 +22,39 @@
 % a) Contrasts of parameter estimates (e.g. activations) and their
 % standard error.
 %
-% b) Fitted and adjusted responses that can be plotted against time,
-% scan, or an indicator variable in the design matrix.
+% b) Fitted and adjusted responses that can be plotted against time, scan,
+% or an indicator variable in the design matrix.
 %
 % c) (fMRI only).  Evoked responses using the basis functions to give
-% impulse responses that would have been seen in the absence of other effects.
-% The PSTH (peristimulus-time histogram) option provides a finite impulse
-% response (FIR) estimate of the trial-specific evoked response as a function
-% of peristimulus time.  This is estimated by refitting a convolution model
-% to the selected voxel using an FIR basis set.  This is simply a set of
-% small boxes covering successive time bins after trial onset.  The width
-% of each bin is usually the TR.  This option provides a more time-resolved
-% quantitative characterisation of the evoked hemodynamic response.  However,
-% it should not be over-interpreted because inference is usually made using a
-% simpler and more efficient basis set (e.g., canonical hrf, or canonical
-% plus time derivative).
+% impulse responses that would have been seen in the absence of other
+% effects. The PSTH (peristimulus-time histogram) option provides a finite
+% impulse response (FIR) estimate of the trial-specific evoked response as
+% a function of peristimulus time.  This is estimated by refitting a
+% convolution model to the selected voxel using an FIR basis set.  This is
+% simply a set of small boxes covering successive time bins after trial
+% onset.  The width of each bin is usually the TR.  This option provides a
+% more time-resolved quantitative characterisation of the evoked
+% hemodynamic response.  However, it should not be over-interpreted because
+% inference is usually made using a simpler and more efficient basis set
+% (e.g., canonical hrf, or canonical plus time derivative).
 %
 % Getting adjusted data:
 % Ensuring the data are adjusted properly can be important (e.g. in
 % constructing explanatory variables such as in a psychophysiological
 % interaction). To remove or correct for specific effects, specify an
 % appropriate F contrast and simply plot the fitted (and adjusted)
-% responses after selecting that F contrast.  The vectors Y (fitted)
-% and y (adjusted) in the workspace will now be corrected for the
-% effects in the reduced design matrix (X0) specified in the contrast
-% manager with the column indices (iX0) of the confounds in this
-% adjustment.
+% responses after selecting that F contrast. The vectors Y (fitted) and y
+% (adjusted) in the workspace will now be corrected for the effects in the
+% reduced design matrix (X0) specified in the contrast manager with the
+% column indices (iX0) of the confounds in this adjustment.
 %
 % Plotting data:
 % All data and graphics use filtered/whitened data and residuals. In PET
 % studies the parameter estimates and the fitted data are often the same
 % because the explanatory variables are simply indicator variables taking
-% the value of one.  Only contrasts previously defined can be plotted.
-% This ensures that the parameters plotted are meaningful even when there
-% is collinearity among the design matrix subpartitions.
+% the value of one.  Only contrasts previously defined can be plotted. This
+% ensures that the parameters plotted are meaningful even when there is
+% collinearity among the design matrix subpartitions.
 %
 % Selecting contrasts used for PPMs will automatically give plots
 % based on conditonal estimates.
@@ -64,13 +63,12 @@
 %                   contrast.standarderror = SE;
 %                   contrast.interval      = 2*CI;
 %
-% is assigned in base workspace for plots of contrasts and their error
+% is assigned in base workspace for plots of contrasts and their error.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_graph.m 3598 2009-11-26 11:48:59Z guillaume $
-
+% $Id: spm_graph.m 4262 2011-03-25 13:39:20Z guillaume $
 
 
 %-Get Graphics figure handle
@@ -85,22 +83,25 @@
 
 %-Find nearest voxel [Euclidean distance] in point list & update GUI
 %--------------------------------------------------------------------------
-if ~length(xSPM.XYZmm)
+if isempty(xSPM.XYZmm)
     spm('alert!','No suprathreshold voxels!',mfilename,0);
 
     Y = []; y = []; beta = []; Bcov = [];
     return
 end
 
-[xyz,i] = spm_XYZreg('NearestXYZ',spm_XYZreg('GetCoords',hReg),xSPM.XYZmm);
-spm_XYZreg('SetCoords',xyz,hReg);
-XYZ     = xSPM.XYZ(:,i);        % coordinates
+if numel(hReg) == 1
+    xyz = spm_XYZreg('GetCoords',hReg);
+else
+    xyz = hReg;
+end
+[xyz,i] = spm_XYZreg('NearestXYZ',xyz,xSPM.XYZmm);
+if numel(hReg) == 1, spm_XYZreg('SetCoords',xyz,hReg); end
+XYZ     = xSPM.XYZ(:,i);
 
-%-Plot
-%==========================================================================
 
-% find out what to plot
-%--------------------------------------------------------------------------
+%-Find out what to plot
+%==========================================================================
 Cplot = {   'Contrast estimates and 90% C.I.',...
             'Fitted responses',...
             'Event-related responses',...
@@ -167,30 +168,53 @@
 
 end
 
-spm('pointer','watch');
+spm('Pointer','Watch');
 
 %-Extract filtered and whitened data from files
 %==========================================================================
 try
     y = spm_get_data(SPM.xY.VY,XYZ);
-    y = spm_filter(SPM.xX.K,SPM.xX.W*y);
 catch
     try
         % remap files in SPM.xY.P if SPM.xY.VY is no longer valid
         %------------------------------------------------------------------
         SPM.xY.VY = spm_vol(SPM.xY.P);
         y = spm_get_data(SPM.xY.VY,XYZ);
-        y = spm_filter(SPM.xX.K,SPM.xX.W*y);
-
+        
     catch
         % data has been moved or renamed
         %------------------------------------------------------------------
-        y = [];
-        spm('alert!',{'Original data have been moved or renamed',...
-            'Recomendation: please update SPM.xY.P'},...
-            mfilename,0);
+        choice = questdlg({'Original data have been moved or renamed',...
+            'How to proceed next?'},...
+            [mfilename ': data files missing...'],...
+            'Specify','Search','Ignore','Ignore');
+        
+        switch choice
+            case 'Specify'
+                [SPM.xY.P,sts] = ...
+                    spm_select(numel(SPM.xY.VY),'image','Select images');
+                if ~sts
+                    [Y,y,beta,Bcov] = deal([]);
+                    spm('Pointer','Arrow');
+                    return;
+                end
+                SPM.xY.VY = spm_vol(SPM.xY.P);
+                for i = 1:numel(SPM.xY.VY)
+                    SPM.xY.VY(i).pinfo(1:2,:) = ...
+                        SPM.xY.VY(i).pinfo(1:2,:)*SPM.xGX.gSF(i);
+                end
+                y = spm_get_data(SPM.xY.VY,XYZ);
+            case 'Search'
+                SPM.xY.VY = spm_check_filename(SPM.xY.VY);
+                y = spm_get_data(SPM.xY.VY,XYZ);
+            otherwise
+                y = [];
+        end
     end
 end
+
+if ~isempty(y), y = spm_filter(SPM.xX.K,SPM.xX.W*y); end
+
 XYZstr = sprintf(' at [%g, %g, %g]',xyz);
 
 
@@ -200,9 +224,10 @@
 
     % make R = NaN so it will not be plotted
     %----------------------------------------------------------------------
-    R   = NaN*ones(size(SPM.xX.X,1),1);
+    R   = NaN(size(SPM.xX.X,1),1);
 
 else
+    
     % residuals (non-whitened)
     %----------------------------------------------------------------------
     R   = spm_sp('r',SPM.xX.xKXs,y);
@@ -273,9 +298,8 @@
         end
 
     else
-        Bcov  = SPM.PPM.Cby;
+        Bcov     = SPM.PPM.Cby;
         for j = 1:length(SPM.PPM.l)
-
             l    = spm_get_data(SPM.VHp(j),XYZ);
             Bcov = Bcov + SPM.PPM.dC{j}*(l - SPM.PPM.l(j));
         end
@@ -283,7 +307,10 @@
 end
 CI    = 1.6449;                 % = spm_invNcdf(1 - 0.05);
 
-spm('pointer','arrow');
+spm('Pointer','Arrow');
+
+%-Plot
+%==========================================================================
 
 %-Colour specifications and index;
 %--------------------------------------------------------------------------
@@ -291,8 +318,8 @@
 
 switch Cplot
 
-    % plot parameter estimates
-    %----------------------------------------------------------------------
+    %-Plot parameter estimates
+    %======================================================================
     case 'Contrast estimates and 90% C.I.'
 
         % compute contrast of parameter estimates and 90% C.I.
@@ -335,8 +362,9 @@
         %------------------------------------------------------------------
         Y = [];
 
-        % all fitted effects or selected effects
-        %------------------------------------------------------------------
+        
+    %-All fitted effects or selected effects
+    %======================================================================
     case 'Fitted responses'
 
         % predicted or adjusted response
@@ -375,8 +403,8 @@
             x    = SPM.xX.xKXs.X(:,i);
             XLAB = SPM.xX.name{i};
 
-            % scan or time
-            %--------------------------------------------------------------
+        % scan or time
+        %------------------------------------------------------------------
         elseif Cx == 2
 
             if isfield(SPM.xY,'RT')
@@ -387,8 +415,8 @@
                 XLAB = 'scan number';
             end
 
-            % user specified
-            %--------------------------------------------------------------
+        % user specified
+        %------------------------------------------------------------------
         elseif Cx == 3
 
             x    = spm_input('enter ordinate','!+1','e','',size(Y,1));
@@ -421,13 +449,14 @@
         ylabel(['response',XYZstr])
         legend('fitted','plus error')
         hold off
-
-        % modeling evoked responses based on Sess
-        %------------------------------------------------------------------
+        
+        
+    %-Modeling evoked responses based on Sess
+    %======================================================================
     case 'Event-related responses'
 
         % get plot type
-        %--------------------------------------------------------------
+        %------------------------------------------------------------------
         Rplot   = { 'fitted response and PSTH',...
             'fitted response and 90% C.I.',...
             'fitted response and adjusted data'};
@@ -442,8 +471,6 @@
         %------------------------------------------------------------------
         switch TITLE
             case 'fitted response and PSTH'
-
-
                 % build a simple FIR model subpartition (X); bin size = TR
                 %----------------------------------------------------------
                 BIN         = SPM.xY.RT;
@@ -484,8 +511,6 @@
                 PCI         = CI*sqrt(diag(bcov(1:j,(1:j))))/dt;
         end
 
-
-
         % basis functions and parameters
         %------------------------------------------------------------------
         X     = SPM.xBF.bf/dt;
@@ -507,8 +532,6 @@
         pst   = pst(q);
         y     = y(q) + Y(bin(q) + 1);
 
-
-
         % plot
         %------------------------------------------------------------------
         figure(Fgraph)
@@ -534,7 +557,6 @@
 
         end
 
-
         % label
         %------------------------------------------------------------------
         [i j] = max(Y);
@@ -545,11 +567,10 @@
         hold off
 
 
-        % modeling evoked responses based on Sess
-        %------------------------------------------------------------------
+    %-Modeling evoked responses based on Sess
+    %======================================================================
     case 'Parametric responses'
 
-
         % return gracefully if no parameters
         %------------------------------------------------------------------
         if ~Sess(s).U(u).P(1).h, return, end
@@ -570,10 +591,9 @@
         end
         q     = spm_orth(q);
 
-
         % parameter estimates for this effect
         %------------------------------------------------------------------
-        B     = beta(Sess(s).Fc(u).i);
+        B     = beta(Sess(s).col(Sess(s).Fc(u).i));
 
         % reconstruct trial-specific responses
         %------------------------------------------------------------------
@@ -611,8 +631,8 @@
         grid on
 
 
-        % modeling evoked responses based on Sess
-        %------------------------------------------------------------------
+    %-Modeling evoked responses based on Sess
+    %======================================================================
     case 'Volterra Kernels'
 
         % Parameter estimates and basis functions
@@ -626,7 +646,7 @@
 
             % Parameter estimates and kernel
             %--------------------------------------------------------------
-            B     = beta(Sess(s).Fc(u).i);
+            B     = beta(Sess(s).col(Sess(s).Fc(u).i));
             i     = 1;
             Y     = 0;
             for p = 1:size(bf,2)
@@ -657,11 +677,11 @@
             xlabel('peristimulus time {secs}')
 
 
-            % first  order kernel
-            %--------------------------------------------------------------
+        % first  order kernel
+        %------------------------------------------------------------------
         else
-            B     = beta(Sess(s).Fc(u).i(1:size(bf,2)));
-            Y     = bf*B;
+            B = beta(Sess(s).col(Sess(s).Fc(u).i(1:size(bf,2))));
+            Y = bf*B;
 
             % plot
             %--------------------------------------------------------------
diff --git a/spm_gx_fmri_linear.m b/spm_gx_fmri_linear.m
new file mode 100644
index 0000000..f7ef89a
--- /dev/null
+++ b/spm_gx_fmri_linear.m
@@ -0,0 +1,62 @@
+function [y] = spm_gx_fmri_linear(x,u,P,M)
+% Simulated BOLD response to input (linear version)
+% FORMAT [y] = spm_gx_fmri_linear(x,u,P,M)
+% y          - BOLD response (%)
+% x          - state vector     (see spm_fx_dcm)
+% P          - Parameter vector (see spm_fx_dcm)
+% M          - model specification structure (see spm_nlsi)
+%__________________________________________________________________________
+%
+% This function implements the BOLD signal model described in: 
+%
+% Stephan KE, Weiskopf N, Drysdale PM, Robinson PA, Friston KJ (2007)
+% Comparing hemodynamic models with DCM. NeuroImage 38: 387-401.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston & Klaas Enno Stephan
+% $Id: spm_gx_fmri_linear.m 4052 2010-08-27 19:22:44Z karl $
+ 
+ 
+% Biophysical constants for 1.5T
+%==========================================================================
+ 
+% time to echo (TE) (default 0.04 sec)
+%--------------------------------------------------------------------------
+try, TE = M.TE; catch, TE = 0.04; end
+ 
+% resting venous volume (%)
+%--------------------------------------------------------------------------
+V0  = 4;
+
+% estimated region-specific ratios of intra- to extra-vascular signal 
+%--------------------------------------------------------------------------
+ep  = 1*exp(P.epsilon);
+ 
+% slope r0 of intravascular relaxation rate R_iv as a function of oxygen 
+% saturation S:  R_iv = r0*[(1 - S)-(1 - S0)] (Hz)
+%--------------------------------------------------------------------------
+r0  = 25;
+ 
+% frequency offset at the outer surface of magnetized vessels (Hz)
+%--------------------------------------------------------------------------
+nu0 = 40.3; 
+ 
+% resting oxygen extraction fraction
+%--------------------------------------------------------------------------
+E0  = 0.4;
+ 
+%-Coefficients in BOLD signal model
+%==========================================================================
+k1  = 4.3*nu0*E0*TE;
+k2  = ep*r0*E0*TE;
+k3  = 1 - ep;
+ 
+%-Output equation of BOLD signal model
+%==========================================================================
+v   = x(:,4) + 1;
+q   = x(:,5) + 1;
+y   = V0*((k1 + k2).*(1 - q) + (k3 - k2).*(1 - v));
+
+
+
diff --git a/spm_hdm_priors.m b/spm_hdm_priors.m
index 5cc53ca..dd540df 100644
--- a/spm_hdm_priors.m
+++ b/spm_hdm_priors.m
@@ -24,7 +24,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_hdm_priors.m 3888 2010-05-15 18:49:56Z karl $
+% $Id: spm_hdm_priors.m 4052 2010-08-27 19:22:44Z karl $
 
 
 
@@ -53,7 +53,7 @@
 %---------------------------------------------------------------------------
 i     = (h + 1):5;
 e(i)  = 0;
-pC    = v*diag(e)*v';
+pC    = v*diag(e)*v'/32;
 
 % append scaling parameter for epsilon: prior variance = 1/32,
 %---------------------------------------------------------------------------
diff --git a/spm_hdm_ui.m b/spm_hdm_ui.m
index b931427..2d858a0 100644
--- a/spm_hdm_ui.m
+++ b/spm_hdm_ui.m
@@ -15,7 +15,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_hdm_ui.m 3812 2010-04-07 16:52:05Z karl $
+% $Id: spm_hdm_ui.m 4048 2010-08-26 16:29:36Z guillaume $
 
 
 % get figure handles
@@ -209,7 +209,7 @@
 title({['1st order kernels for ' U.name{j}];...
        'state variables'},'FontSize',9)
 ylabel('normalized values')
-legend('s','f','q','v',0);
+legend('s','f','v','q',0);
 grid on
 
 
diff --git a/spm_help.m b/spm_help.m
index 6aac766..c77f585 100644
--- a/spm_help.m
+++ b/spm_help.m
@@ -118,7 +118,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes, Karl Friston
-% $Id: spm_help.m 3934 2010-06-17 14:58:25Z guillaume $
+% $Id: spm_help.m 4078 2010-10-06 17:41:26Z guillaume $
 
 
 %=======================================================================
@@ -707,7 +707,7 @@
 if isempty(Fhelp)
     Fhelp = spm_help('!CreateHelpWin');
 else
-    set(Fhelp,'Visible','on');
+    %set(Fhelp,'Visible','on');
 end
 FS = spm('FontSizes');
 PF = spm_platform('fonts');
diff --git a/spm_hilbert.m b/spm_hilbert.m
new file mode 100644
index 0000000..bcffbb8
--- /dev/null
+++ b/spm_hilbert.m
@@ -0,0 +1,35 @@
+function [x] = spm_hilbert(xr)
+% Computes analytic signal
+% FORMAT [x] = spm_hilbert(xr)
+%
+% Returns analytic signal x = xr + i*xi such that 
+% xi is the Hilbert transform of real vector xr.
+%__________________________________________________________________________
+% Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
+ 
+% Will Penny
+% $Id: spm_hilbert.m 4024 2010-07-28 18:43:19Z guillaume $
+
+if ~isreal(xr)
+  xr = real(xr);
+end
+
+% Work along the first nonsingleton dimension
+[xr,nshifts] = shiftdim(xr);
+
+n = size(xr,1);
+x = fft(xr,n,1); % n-point FFT over columns.
+h  = zeros(n,~isempty(x)); % nx1 for nonempty. 0x0 for empty.
+if n > 0 && 2*fix(n/2) == n
+  % even and nonempty
+  h([1 n/2+1]) = 1;
+  h(2:n/2) = 2;
+elseif n>0
+  % odd and nonempty
+  h(1) = 1;
+  h(2:(n+1)/2) = 2;
+end
+x = ifft(x.*h(:,ones(1,size(x,2))));
+
+% Convert back to the original shape.
+x = shiftdim(x,-nshifts);
diff --git a/spm_hist2.m b/spm_hist2.m
index e9bba5d..0c77d8d 100644
--- a/spm_hist2.m
+++ b/spm_hist2.m
@@ -10,17 +10,16 @@
 %      sampled, whereas [4 4 4] means that only 1 in 64 voxels are
 %      sampled.
 %
-% This function is called by spm_mireg for rapidly computing
-% joint histograms for mutual information image registration.
+% This function is called by spm_coreg for rapidly computing joint
+% histograms for mutual information image registration.
 %
 % Note that the function jitters the sampling of the data to reduce
 % interpolation artifacts.
-%
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 2001-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_hist2.m 2696 2009-02-05 20:29:48Z guillaume $
+% $Id: spm_hist2.m 4156 2011-01-11 19:03:31Z guillaume $
 
 %-This is merely the help file for the compiled routine
 error('spm_hist2.c not compiled - see Makefile')
\ No newline at end of file
diff --git a/spm_image.m b/spm_image.m
index 2492bc8..dd2db54 100644
--- a/spm_image.m
+++ b/spm_image.m
@@ -1,7 +1,7 @@
-function spm_image(op,varargin)
-% image and header display
+function spm_image(action,varargin)
+% Image and header display
 % FORMAT spm_image
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_image is an interactive facility that allows orthogonal sections
 % from an image volume to be displayed.  Clicking the cursor on either
@@ -17,12 +17,12 @@ function spm_image(op,varargin)
 %
 % The images can be re-oriented by entering appropriate translations,
 % rotations and zooms into the panel on the left.  The transformations
-% can then be saved by hitting the ``Reorient images...'' button.  The
-% transformations that were applied to the image are saved to the
-% ``.mat'' files of the selected images.  The transformations are
-% considered to be relative to any existing transformations that may be
-% stored in the ``.mat'' files.  Note that the order that the
-% transformations are applied in is the same as in ``spm_matrix.m''.
+% can then be saved by hitting the "Reorient images..." button.  The
+% transformations that were applied to the image are saved to the header
+% information of the selected images.  The transformations are considered
+% to be relative to any existing transformations that may be stored.
+% Note that the order that the transformations are applied in is the
+% same as in spm_matrix.m.
 %
 % The ``Reset...'' button next to it is for setting the orientation of
 % images back to transverse.  It retains the current voxel sizes,
@@ -38,217 +38,198 @@ function spm_image(op,varargin)
 %   Vox size   - the distance (in mm) between the centres of
 %                neighbouring voxels.
 %   Origin     - the voxel at the origin of the co-ordinate system
-%   DIr Cos    - Direction cosines.  This is a widely used
+%   Dir Cos    - Direction cosines.  This is a widely used
 %                representation of the orientation of an image.
 %
 % There are also a few options for different resampling modes, zooms
-% etc.  You can also flip between voxel space (as would be displayed
-% by Analyze) or world space (the orientation that SPM considers the
-% image to be in).  If you are re-orienting the images, make sure that
-% world space is specified.  Blobs (from activation studies) can be
-% superimposed on the images and the intensity windowing can also be
-% changed.
-%
-%_______________________________________________________________________
+% etc.  You can also flip between voxel space or world space.  If you
+% are re-orienting the images, make sure that world space is specified.
+% Blobs (from activation studies) can be superimposed on the images and 
+% the intensity windowing can also be changed.
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_image.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_image.m 4205 2011-02-21 15:39:08Z guillaume $
 
 
 global st
 
-if nargin == 0,
-    spm('FnUIsetup','Display',0);
-    spm('FnBanner',mfilename,'$Rev: 3691 $');
-
-    % get the image's filename {P}
-    %----------------------------------------------------------------------
-    P      = spm_select(1,'image','Select image');
-    spm_image('init',P);
-    return;
-end;
+if ~nargin, action = 'Init'; end
 
-if isstruct(op)
-    % job data structure
-    spm_image('init', op.data{1});
+if ~any(strcmpi(action,{'init','reset'})) && ...
+        (isempty(st) || ~isfield(st,'vols') || isempty(st.vols{1}))
+    spm_image('Reset');
+    warning('Lost all the image information');
     return;
-end;
-
-try
-    if ~strcmp(op,'init') && ~strcmp(op,'reset') && isempty(st.vols{1})
-        my_reset; warning('Lost all the image information');
-        return;
-    end;
-catch
 end
 
-if strcmp(op,'repos'),
-    % The widgets for translation rotation or zooms have been modified.
-    %-----------------------------------------------------------------------
-    fg      = spm_figure('Findwin','Graphics');
-    set(fg,'Pointer','watch');
-    i       = varargin{1};
-    st.B(i) = eval(get(gco,'String'),num2str(st.B(i)));
-    set(gco,'String',st.B(i));
+switch lower(action)
+    
+    case {'init','display'}
+    % Display image
+    %----------------------------------------------------------------------
+    if isempty(varargin)
+        [P, sts] = spm_select(1,'image','Select image');
+        if ~sts, return; end
+    else
+        P = varargin{1};
+    end
+    if ischar(P), P = spm_vol(P); end
+    P = P(1);
+
+    init_display(P);
+    
+    case 'repos'
+    % The widgets for translation rotation or zooms have been modified
+    %----------------------------------------------------------------------
+    trz = varargin{1};
+    try, st.B(trz) = eval(get(gco,'String')); end
+    set(gco,'String',st.B(trz));
     st.vols{1}.premul = spm_matrix(st.B);
     % spm_orthviews('MaxBB');
-    spm_image('zoom_in');
-    spm_image('update_info');
-    set(fg,'Pointer','arrow');
-    return;
-end;
-
-if strcmp(op,'shopos'),
-    % The position of the crosshairs has been moved.
-    %-----------------------------------------------------------------------
-    if isfield(st,'mp'),
-        fg  = spm_figure('Findwin','Graphics');
-        if any(findobj(fg) == st.mp),
-            set(st.mp,'String',sprintf('%.1f %.1f %.1f',spm_orthviews('pos')));
-            pos = spm_orthviews('pos',1);
+    spm_image('Zoom');
+    spm_image('Update');
+    
+    case 'shopos'
+    % The position of the crosshairs has been moved
+    %----------------------------------------------------------------------
+    if isfield(st,'mp')
+        fg = spm_figure('Findwin','Graphics');
+        if any(findobj(fg) == st.mp)
+            set(st.mp,'String',sprintf('%.1f %.1f %.1f',spm_orthviews('Pos')));
+            pos = spm_orthviews('Pos',1);
             set(st.vp,'String',sprintf('%.1f %.1f %.1f',pos));
             set(st.in,'String',sprintf('%g',spm_sample_vol(st.vols{1},pos(1),pos(2),pos(3),st.hld)));
         else
             st.Callback = ';';
             st = rmfield(st,{'mp','vp','in'});
-        end;
+        end
     else
         st.Callback = ';';
-    end;
-    return;
-end;
-
-if strcmp(op,'setposmm'),
-    % Move the crosshairs to the specified position
-    %-----------------------------------------------------------------------
-    if isfield(st,'mp'),
+    end
+ 
+    case 'setposmm'
+    % Move the crosshairs to the specified position {mm}
+    %----------------------------------------------------------------------
+    if isfield(st,'mp')
         fg = spm_figure('Findwin','Graphics');
-        if any(findobj(fg) == st.mp),
+        if any(findobj(fg) == st.mp)
             pos = sscanf(get(st.mp,'String'), '%g %g %g');
-            if length(pos)~=3,
-                pos = spm_orthviews('pos');
-            end;
+            if length(pos)~=3
+                pos = spm_orthviews('Pos');
+            end
             spm_orthviews('Reposition',pos);
-        end;
-    end;
-    return;
-end;
-
-if strcmp(op,'setposvx'),
-    % Move the crosshairs to the specified position
-    %-----------------------------------------------------------------------
-    if isfield(st,'mp'),
+        end
+    end
+    
+    case 'setposvx'
+    % Move the crosshairs to the specified position {vx}
+    %----------------------------------------------------------------------
+    if isfield(st,'mp')
         fg = spm_figure('Findwin','Graphics');
-        if any(findobj(fg) == st.vp),
+        if any(findobj(fg) == st.vp)
             pos = sscanf(get(st.vp,'String'), '%g %g %g');
-            if length(pos)~=3,
+            if length(pos)~=3
                 pos = spm_orthviews('pos',1);
-            end;
+            end
             tmp = st.vols{1}.premul*st.vols{1}.mat;
             pos = tmp(1:3,:)*[pos ; 1];
             spm_orthviews('Reposition',pos);
-        end;
-    end;
-    return;
-end;
-
+        end
+    end
 
-if strcmp(op,'addblobs'),
+    case 'addblobs'
     % Add blobs to the image - in full colour
+    %----------------------------------------------------------------------
     spm_figure('Clear','Interactive');
     nblobs = spm_input('Number of sets of blobs',1,'1|2|3|4|5|6',[1 2 3 4 5 6],1);
-    for i=1:nblobs,
-        [SPM,VOL] = spm_getSPM;
+    for i=1:nblobs
+        [SPM,xSPM] = spm_getSPM;
         c = spm_input('Colour','+1','m','Red blobs|Yellow blobs|Green blobs|Cyan blobs|Blue blobs|Magenta blobs',[1 2 3 4 5 6],1);
         colours = [1 0 0;1 1 0;0 1 0;0 1 1;0 0 1;1 0 1];
-        spm_orthviews('addcolouredblobs',1,VOL.XYZ,VOL.Z,VOL.M,colours(c,:));
-        set(st.blobber,'String','Remove Blobs','Callback','spm_image(''rmblobs'');');
-    end;
-    spm_orthviews('addcontext',1);
+        spm_orthviews('AddColouredBlobs',1,xSPM.XYZ,xSPM.Z,xSPM.M,colours(c,:));
+        set(st.blobber,'String','Remove Blobs','Callback','spm_image(''RemoveBlobs'');');
+    end
+    spm_orthviews('AddContext',1);
     spm_orthviews('Redraw');
-end;
 
-if strcmp(op,'rmblobs'),
+    case {'removeblobs','rmblobs'}
     % Remove all blobs from the images
-    spm_orthviews('rmblobs',1);
-    set(st.blobber,'String','Add Blobs','Callback','spm_image(''addblobs'');');
-    spm_orthviews('rmcontext',1); 
+    %----------------------------------------------------------------------
+    spm_orthviews('RemoveBlobs',1);
+    set(st.blobber,'String','Add Blobs','Callback','spm_image(''AddBlobs'');');
+    spm_orthviews('RemoveContext',1); 
     spm_orthviews('Redraw');
-end;
 
-if strcmp(op,'window'),
+    case 'window'
+    % Window
+    %----------------------------------------------------------------------
     op = get(st.win,'Value');
-    if op == 1,
-        spm_orthviews('window',1);
+    if op == 1
+        spm_orthviews('Window',1); % automatic
     else
-        spm_orthviews('window',1,spm_input('Range','+1','e','',2));
-    end;
-end;
-
-
-if strcmp(op,'reorient'),
-    % Time to modify the ``.mat'' files for the images.
-    % I hope that giving people this facility is the right thing to do....
-    %-----------------------------------------------------------------------
+        spm_orthviews('Window',1,spm_input('Range','+1','e','',2));
+    end
+    
+    case 'reorient'
+    % Reorient images
+    %----------------------------------------------------------------------
     mat = spm_matrix(st.B);
     if det(mat)<=0
         spm('alert!','This will flip the images',mfilename,0,1);
-    end;
-    P = spm_select(Inf, 'image','Images to reorient');
-    Mats = zeros(4,4,size(P,1));
-    spm_progress_bar('Init',size(P,1),'Reading current orientations',...
+    end
+    [P, sts] = spm_select([1 Inf], 'image','Images to reorient');
+    if ~sts, return; else P = cellstr(P); end
+    Mats = zeros(4,4,numel(P));
+    spm_progress_bar('Init',numel(P),'Reading current orientations',...
         'Images Complete');
-    for i=1:size(P,1),
-        Mats(:,:,i) = spm_get_space(P(i,:));
+    for i=1:numel(P)
+        Mats(:,:,i) = spm_get_space(P{i});
         spm_progress_bar('Set',i);
-    end;
-    spm_progress_bar('Init',size(P,1),'Reorienting images',...
+    end
+    spm_progress_bar('Init',numel(P),'Reorienting images',...
         'Images Complete');
-    for i=1:size(P,1),
-        spm_get_space(P(i,:),mat*Mats(:,:,i));
+    for i=1:numel(P)
+        spm_get_space(P{i},mat*Mats(:,:,i));
         spm_progress_bar('Set',i);
-    end;
+    end
     spm_progress_bar('Clear');
     tmp = spm_get_space([st.vols{1}.fname ',' num2str(st.vols{1}.n)]);
-    if sum((tmp(:)-st.vols{1}.mat(:)).^2) > 1e-8,
-        spm_image('init',st.vols{1}.fname);
-    end;
-    return;
-end;
-
-if strcmp(op,'resetorient'),
-    % Time to modify the ``.mat'' files for the images.
-    % I hope that giving people this facility is the right thing to do....
-    %-----------------------------------------------------------------------
-    P = spm_select(Inf, 'image','Images to reset orientation of');
-    spm_progress_bar('Init',size(P,1),'Resetting orientations',...
+    if sum((tmp(:)-st.vols{1}.mat(:)).^2) > 1e-8
+        spm_image('Init',st.vols{1}.fname);
+    end
+
+    case 'resetorient'
+    % Reset orientation of images
+    %----------------------------------------------------------------------
+    [P,sts] = spm_select([1 Inf], 'image','Images to reset orientation of');
+    if ~sts, return; else P = cellstr(P); end
+    spm_progress_bar('Init',numel(P),'Resetting orientations',...
         'Images Complete');
-    for i=1:size(P,1),
-        V    = spm_vol(deblank(P(i,:)));
+    for i=1:numel(P)
+        V    = spm_vol(P{i});
         M    = V.mat;
         vox  = sqrt(sum(M(1:3,1:3).^2));
-        if det(M(1:3,1:3))<0, vox(1) = -vox(1); end;
+        if det(M(1:3,1:3))<0, vox(1) = -vox(1); end
         orig = (V.dim(1:3)+1)/2;
-                off  = -vox.*orig;
-                M    = [vox(1) 0      0      off(1)
+        off  = -vox.*orig;
+        M    = [vox(1) 0      0      off(1)
                 0      vox(2) 0      off(2)
                 0      0      vox(3) off(3)
                 0      0      0      1];
-        spm_get_space(P(i,:),M);
+        spm_get_space(P{i},M);
         spm_progress_bar('Set',i);
-    end;
+    end
     spm_progress_bar('Clear');
     tmp = spm_get_space([st.vols{1}.fname ',' num2str(st.vols{1}.n)]);
-    if sum((tmp(:)-st.vols{1}.mat(:)).^2) > 1e-8,
-        spm_image('init',st.vols{1}.fname);
-    end;
-    return;
-end;
+    if sum((tmp(:)-st.vols{1}.mat(:)).^2) > 1e-8
+        spm_image('Init',st.vols{1}.fname);
+    end
 
-if strcmp(op,'update_info'),
-    % Modify the positional information in the right hand panel.
-    %-----------------------------------------------------------------------
+    case 'update'
+    % Modify the positional information in the right hand panel
+    %----------------------------------------------------------------------
     mat = st.vols{1}.premul*st.vols{1}.mat;
     Z = spm_imatrix(mat);
     Z = Z(7:9);
@@ -271,64 +252,48 @@ function spm_image(op,varargin)
 
     tmp = [[R zeros(3,1)] ; 0 0 0 1]*diag([Z 1])*spm_matrix(-O) - mat;
 
-    if sum(tmp(:).^2)>1e-8,
+    if sum(tmp(:).^2)>1e-5
         set(st.posinf.w, 'String', 'Warning: shears involved');
     else
         set(st.posinf.w, 'String', '');
-    end;
+    end
 
-    return;
-end;
+    case 'zoom'
+    % Zoom in
+    %----------------------------------------------------------------------
+    [zl rl] = spm_orthviews('ZoomMenu');
+    % Values are listed in reverse order
+    cz = numel(zl)-get(st.zoomer,'Value')+1;
+    spm_orthviews('Zoom',zl(cz),rl(cz));
 
-if strcmp(op,'reset'),
-    my_reset;
-end;
+    case 'reset'
+    % Reset
+    %----------------------------------------------------------------------
+    spm_orthviews('Reset');
+    spm_figure('Clear','Graphics');
 
-if strcmp(op,'zoom_in'),
-    op = get(st.zoomer,'Value');
-    if op==1,
-        spm_orthviews('resolution',1);
-        spm_orthviews('MaxBB');
-    else
-        vx = sqrt(sum(st.Space(1:3,1:3).^2));
-        vx = vx.^(-1);
-        pos = spm_orthviews('pos');
-        pos = st.Space\[pos ; 1];
-        pos = pos(1:3)';
-        if     op == 2, st.bb = [pos-80*vx ; pos+80*vx] ; spm_orthviews('resolution',1);
-        elseif op == 3, st.bb = [pos-40*vx ; pos+40*vx] ; spm_orthviews('resolution',.5);
-        elseif op == 4, st.bb = [pos-20*vx ; pos+20*vx] ; spm_orthviews('resolution',.25);
-        elseif op == 5, st.bb = [pos-10*vx ; pos+10*vx] ; spm_orthviews('resolution',.125);
-        else            st.bb = [pos- 5*vx ; pos+ 5*vx] ; spm_orthviews('resolution',.125);
-        end;
-    end;
-    return;
-end;
+end
 
-if strcmp(op,'init'),
-fg = spm_figure('GetWin','Graphics');
-if isempty(fg), error('Can''t create graphics window'); end
-spm_figure('Clear','Graphics');
 
-P = varargin{1};
-if ischar(P), P = spm_vol(P); end;
-P = P(1);
+%==========================================================================
+function init_display(P)
+
+global st
 
-spm_orthviews('Reset');
+fg = spm_figure('GetWin','Graphics');
+spm_image('Reset');
 spm_orthviews('Image', P, [0.0 0.45 1 0.55]);
-if isempty(st.vols{1}), return; end;
+if isempty(st.vols{1}), return; end
 
 spm_orthviews('MaxBB');
 st.callback = 'spm_image(''shopos'');';
 
 st.B = [0 0 0  0 0 0  1 1 1  0 0 0];
 
-% locate Graphics window and clear it
-%-----------------------------------------------------------------------
+% Widgets for re-orienting images.
+%--------------------------------------------------------------------------
 WS = spm('WinScale');
 
-% Widgets for re-orienting images.
-%-----------------------------------------------------------------------
 uicontrol(fg,'Style','Frame','Position',[60 25 200 325].*WS,'DeleteFcn','spm_image(''reset'');');
 uicontrol(fg,'Style','Text', 'Position',[75 220 100 016].*WS,'String','right  {mm}');
 uicontrol(fg,'Style','Text', 'Position',[75 200 100 016].*WS,'String','forward  {mm}');
@@ -357,7 +322,7 @@ function spm_image(op,varargin)
          'Position',[195 35 55 020].*WS,'ToolTipString','reset orientations of selected images');
 
 % Crosshair position
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 uicontrol(fg,'Style','Frame','Position',[70 250 180 90].*WS);
 uicontrol(fg,'Style','Text', 'Position',[75 320 170 016].*WS,'String','Crosshair Position');
 uicontrol(fg,'Style','PushButton', 'Position',[75 316 170 006].*WS,...
@@ -373,7 +338,7 @@ function spm_image(op,varargin)
 st.in = uicontrol(fg,'Style','Text', 'Position',[140 255  85 020].*WS,'String','');
 
 % General information
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 uicontrol(fg,'Style','Frame','Position',[305  25 280 325].*WS);
 uicontrol(fg,'Style','Text','Position' ,[310 330 50 016].*WS,...
     'HorizontalAlignment','right', 'String', 'File:');
@@ -390,24 +355,23 @@ function spm_image(op,varargin)
 uicontrol(fg,'Style','Text','Position' ,[310 270 100 016].*WS,...
     'HorizontalAlignment','right', 'String', 'Intensity:');
 str = 'varied';
-if size(st.vols{1}.pinfo,2) == 1,
-    if st.vols{1}.pinfo(2),
+if size(st.vols{1}.pinfo,2) == 1
+    if st.vols{1}.pinfo(2)
         str = sprintf('Y = %g X + %g', st.vols{1}.pinfo(1:2)');
     else
         str = sprintf('Y = %g X', st.vols{1}.pinfo(1)');
-    end;
-end;
+    end
+end
 uicontrol(fg,'Style','Text','Position' ,[410 270 160 016].*WS,...
     'HorizontalAlignment','left', 'String', str,'FontWeight','bold');
 
-if isfield(st.vols{1}, 'descrip'),
+if isfield(st.vols{1}, 'descrip')
     uicontrol(fg,'Style','Text','Position' ,[310 250 260 016].*WS,...
     'HorizontalAlignment','center', 'String', st.vols{1}.descrip,'FontWeight','bold');
-end;
-
+end
 
 % Positional information
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 mat = st.vols{1}.premul*st.vols{1}.mat;
 Z = spm_imatrix(mat);
 Z = Z(7:9);
@@ -441,17 +405,32 @@ function spm_image(op,varargin)
 tmp = [[R zeros(3,1)] ; 0 0 0 1]*diag([Z 1])*spm_matrix(-O) - mat;
 st.posinf.w = uicontrol(fg,'Style','Text','Position' ,[310 110 260 016].*WS,...
     'HorizontalAlignment','center', 'String', '','FontWeight','bold');
-if sum(tmp(:).^2)>1e-8,
+if sum(tmp(:).^2)>1e-8
     set(st.posinf.w, 'String', 'Warning: shears involved');
-end;
+end
 
-% Assorted other buttons.
-%-----------------------------------------------------------------------
+% Assorted other buttons
+%--------------------------------------------------------------------------
 uicontrol(fg,'Style','Frame','Position',[310 30 270 70].*WS);
+zl = spm_orthviews('ZoomMenu');
+czlabel = cell(size(zl));
+% List zoom steps in reverse order
+zl = zl(end:-1:1);
+for cz = 1:numel(zl)
+    if isinf(zl(cz))
+        czlabel{cz} = 'Full Volume';
+    elseif isnan(zl(cz))
+        czlabel{cz} = 'BBox (Y > ...)';
+    elseif zl(cz) == 0
+        czlabel{cz} = 'BBox (nonzero)';
+    else
+        czlabel{cz} = sprintf('%dx%dx%dmm', 2*zl(cz), 2*zl(cz), 2*zl(cz));
+    end
+end
 st.zoomer = uicontrol(fg,'Style','popupmenu' ,'Position',[315 75 125 20].*WS,...
-    'String',char('Full Volume','160x160x160mm','80x80x80mm','40x40x40mm','20x20x20mm','10x10x10mm'),...
-    'Callback','spm_image(''zoom_in'')','ToolTipString','zoom in by different amounts');
-c = 'if get(gco,''Value'')==1, spm_orthviews(''Space''), else, spm_orthviews(''Space'', 1);end;spm_image(''zoom_in'')';
+    'String',czlabel,...
+    'Callback','spm_image(''zoom'')','ToolTipString','zoom in by different amounts');
+c = 'if get(gco,''Value'')==1, spm_orthviews(''Space''), else, spm_orthviews(''Space'', 1);end;spm_image(''zoom'')';
 uicontrol(fg,'Style','popupmenu' ,'Position',[315 55 125 20].*WS,...
     'String',char('World Space','Voxel Space'),...
     'Callback',c,'ToolTipString','display in aquired/world orientation');
@@ -468,11 +447,3 @@ function spm_image(op,varargin)
 %   'String','Window','Callback','spm_image(''window'');','ToolTipString','range of voxel intensities % displayed');
 st.blobber = uicontrol(fg,'Style','pushbutton','Position',[450 35 125 20].*WS,...
     'String','Add Blobs','Callback','spm_image(''addblobs'');','ToolTipString','superimpose activations');
-end;
-return;
-
-
-function my_reset
-spm_orthviews('reset');
-spm_figure('Clear','Graphics');
-return;
diff --git a/spm_input.m b/spm_input.m
index 8d4fbf5..22a132e 100644
--- a/spm_input.m
+++ b/spm_input.m
@@ -151,7 +151,7 @@
 % edit widget.
 %
 %
-% - Comand line -
+% - Command line -
 % If YPos is 0 or global CMDLINE is true, then the command line is used.
 % Negative YPos overrides CMDLINE, ensuring the GUI is used, at
 % YPos=abs(YPos). Similarly relative YPos beginning with '!'
@@ -171,7 +171,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_input.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: spm_input.m 4143 2010-12-22 11:55:43Z guillaume $
 
 
 %=======================================================================
@@ -622,15 +622,15 @@
 
 %-Condition arguments
 %=======================================================================
-if nargin<1|isempty(varargin{1}), Prompt=''; else, Prompt=varargin{1}; end
+if nargin<1||isempty(varargin{1}), Prompt=''; else Prompt=varargin{1}; end
 
-if ~isempty(Prompt) & ischar(Prompt) & Prompt(1)=='!'
+if ~isempty(Prompt) && ischar(Prompt) && Prompt(1)=='!'
     %-Utility functions have Prompt string starting with '!'
     Type = Prompt;
 else            %-Should be an input request: get Type & YPos
-    if nargin<3|isempty(varargin{3}), Type='e';  else, Type=varargin{3}; end
+    if nargin<3||isempty(varargin{3}), Type='e';  else Type=varargin{3}; end
     if any(Type=='|'), Type='b|'; end
-    if nargin<2|isempty(varargin{2}), YPos='+1'; else, YPos=varargin{2}; end
+    if nargin<2||isempty(varargin{2}), YPos='+1'; else YPos=varargin{2}; end
 
     [CmdLine,YPos] = spm_input('!CmdLine',YPos);
 
@@ -652,13 +652,13 @@
 case {'s','s+','e','n','w','i','r','c','x','p'}  %-String and evaluated input
 %=======================================================================
 %-Condition arguments
-if nargin<6|isempty(varargin{6}), m=[]; else, m=varargin{6}; end
-if nargin<5|isempty(varargin{5}), n=[]; else, n=varargin{5}; end
-if nargin<4, DefStr=''; else, DefStr=varargin{4}; end
-if strcmp(lower(Type),'s+')
+if nargin<6||isempty(varargin{6}), m=[]; else m=varargin{6}; end
+if nargin<5||isempty(varargin{5}), n=[]; else n=varargin{5}; end
+if nargin<4, DefStr=''; else DefStr=varargin{4}; end
+if strcmpi(Type,'s+')
     %-DefStr should be a cellstr for 's+' type.
     if isempty(DefStr), DefStr = {};
-        else, DefStr = cellstr(DefStr); end
+    else DefStr = cellstr(DefStr); end
     DefStr = DefStr(:);
 else
     %-DefStr needs to be a string
@@ -679,18 +679,20 @@
 case 'r', TTstr='enter expression - real number(s)';
     if ~isempty(m), TTstr=[TTstr,sprintf(' in [%g,%g]',min(m),max(m))]; end
 case 'c', TTstr='enter indicator vector e.g. 0101...  or abab...';
-    if ~isempty(m) & isfinite(m), strM=sprintf(' (%d)',m); end
+    if ~isempty(m) && isfinite(m), strM=sprintf(' (%d)',m); end
 case 'x', TTstr='enter contrast matrix';
 case 'p',
-    if isempty(n), error('permutation of what?'), else, P=n(:)'; end
+    if isempty(n), error('permutation of what?'), else P=n(:)'; end
     if isempty(m), n = [1,length(P)]; end
     m = P;
-    if ~length(setxor(m,[1:max(m)]))
+    if isempty(setxor(m,[1:max(m)]))
         TTstr=['enter permutation of [1:',num2str(max(m)),']'];
     else
         TTstr=['enter permutation of [',num2str(m),']'];
     end
-otherwise, TTstr='enter expression'; end
+otherwise
+    TTstr='enter expression';
+end
 
 strN = sf_SzStr(n);
 
@@ -725,7 +727,7 @@
         fprintf('\n')
 
         str = input('l001 : ','s');
-        while (isempty(str) | strcmp(str,'.')) & isempty(DefStr)
+        while (isempty(str) || strcmp(str,'.')) && isempty(DefStr)
             spm('Beep')
             fprintf('! %s : enter something!\n',mfilename)
             str = input('l001 : ','s');
@@ -774,7 +776,7 @@
         'BackgroundColor',COLOUR,...
         'HorizontalAlignment','Right',...
         'Position',PRec);
-
+    if TTips, set(hPrmpt,'ToolTipString',[strN,Prompt,strM]); end
 
 
     %-Default button surrounding edit widget (if a DefStr given)
@@ -786,7 +788,7 @@
             'get(gcbo,''String''))'];
     if ~isempty(DefStr)
         if iscellstr(DefStr), str=[DefStr{1},'...'];
-            else, str=DefStr; end
+        else str=DefStr; end
         hDef = uicontrol(Finter,'Style','PushButton',...
             'String',DefStr,...
             'ToolTipString',...
@@ -804,7 +806,7 @@
     cb = 'set(get(gcbo,''UserData''),''UserData'',get(gcbo,''String''))';
     h = uicontrol(Finter,'Style','Edit',...
         'String',DefStr,...
-        'Max',strcmp(lower(Type),'s+')+1,...
+        'Max',strcmpi(Type,'s+')+1,...
         'Tag',['GUIinput_',int2str(YPos)],...
         'UserData',hPrmpt,...
         'CallBack',cb,...
@@ -886,40 +888,40 @@
 %-Condition arguments
 switch lower(Type), case {'b','be1','bi1','br1','m'}
     m = []; Title = '';
-    if nargin<6, DefItem=[];  else, DefItem=varargin{6}; end
-    if nargin<5, Values=[];   else, Values =varargin{5}; end
-    if nargin<4, Labels='';   else, Labels =varargin{4}; end
+    if nargin<6, DefItem=[];  else DefItem=varargin{6}; end
+    if nargin<5, Values=[];   else Values =varargin{5}; end
+    if nargin<4, Labels='';   else Labels =varargin{4}; end
 case 'bd'
-    if nargin<7, Title='';    else, Title  =varargin{7}; end
-    if nargin<6, DefItem=[];  else, DefItem=varargin{6}; end
-    if nargin<5, Values=[];   else, Values =varargin{5}; end
-    if nargin<4, Labels='';   else, Labels =varargin{4}; end
+    if nargin<7, Title='';    else Title  =varargin{7}; end
+    if nargin<6, DefItem=[];  else DefItem=varargin{6}; end
+    if nargin<5, Values=[];   else Values =varargin{5}; end
+    if nargin<4, Labels='';   else Labels =varargin{4}; end
 case 'y/n'
     Title = '';
-    if nargin<5, DefItem=[];  else, DefItem=varargin{5}; end
-    if nargin<4, Values=[];   else, Values =varargin{4}; end
+    if nargin<5, DefItem=[];  else DefItem=varargin{5}; end
+    if nargin<4, Values=[];   else Values =varargin{4}; end
     if isempty(Values), Values='yn'; end
     Labels = {'yes','no'};
 case 'b|'
     Title = '';
-    if nargin<5, DefItem=[];  else, DefItem=varargin{5}; end
-    if nargin<4, Values=[];   else, Values =varargin{4}; end
+    if nargin<5, DefItem=[];  else DefItem=varargin{5}; end
+    if nargin<4, Values=[];   else Values =varargin{4}; end
     Labels = varargin{3};
 case 'bn1'
-    if nargin<7, m=[];        else, m=varargin{7};       end
-    if nargin<6, DefItem=[];  else, DefItem=varargin{6}; end
-    if nargin<5, Values=[];   else, Values =varargin{5}; end
+    if nargin<7, m=[];        else m=varargin{7};       end
+    if nargin<6, DefItem=[];  else DefItem=varargin{6}; end
+    if nargin<5, Values=[];   else Values =varargin{5}; end
     if nargin<4, Labels=[1:5]'; Values=[1:5]; Type='-n1';
-        else, Labels=varargin{4}; end
+    else Labels=varargin{4}; end
 case 'bw1'
-    if nargin<7, m=[];        else, m=varargin{7};       end
-    if nargin<6, DefItem=[];  else, DefItem=varargin{6}; end
-    if nargin<5, Values=[];   else, Values =varargin{5}; end
+    if nargin<7, m=[];        else m=varargin{7};       end
+    if nargin<6, DefItem=[];  else DefItem=varargin{6}; end
+    if nargin<5, Values=[];   else Values =varargin{5}; end
     if nargin<4, Labels=[0:4]'; Values=[0:4]; Type='-w1';
-        else, Labels=varargin{4}; end
+    else Labels=varargin{4}; end
 case {'-n1','n1','-w1','w1'}
-    if nargin<5, m=[];        else, m=varargin{5};       end
-    if nargin<4, DefItem=[];  else, DefItem=varargin{4}; end
+    if nargin<5, m=[];        else m=varargin{5};       end
+    if nargin<4, DefItem=[];  else DefItem=varargin{4}; end
     switch lower(Type)
     case {'n1','-n1'}, Labels=[1:min([5,m])]'; Values=Labels'; Type='-n1';
     case {'w1','-w1'}, Labels=[0:min([4,m])]'; Values=Labels'; Type='-w1';
@@ -933,7 +935,7 @@
 if iscellstr(Labels), Labels=char(Labels); end
 
 %-Convert Labels "option" string to string matrix if required
-if ischar(Labels) & any(Labels=='|')
+if ischar(Labels) && any(Labels(:)=='|')
     OptStr=Labels;
     BarPos=find([OptStr=='|',1]);
     Labels=OptStr(1:BarPos(1)-1);
@@ -945,14 +947,14 @@
 
 %-Set default Values for the Labels
 if isempty(Values)
-    if strcmp(lower(Type),'m')
+    if strcmpi(Type,'m')
         Values=[1:size(Labels,1)]';
     else
         Values=Labels;
     end
 else
     %-Make sure Values are in rows
-    if size(Labels,1)>1 & size(Values,1)==1, Values = Values'; end
+    if size(Labels,1)>1 && size(Values,1)==1, Values = Values'; end
     %-Check numbers of Labels and Values match
     if (size(Labels,1)~=size(Values,1))
         error('Labels & Values incompatible sizes'), end
@@ -962,7 +964,7 @@
 %-Numeric Labels to strings
 if isnumeric(Labels)
     tmp = Labels; Labels = cell(size(tmp,1),1);
-    for i=1:prod(size(tmp)), Labels{i}=num2str(tmp(i,:)); end
+    for i=1:numel(tmp), Labels{i}=num2str(tmp(i,:)); end
     Labels=char(Labels);
 end
 
@@ -975,7 +977,7 @@
     nLabels     = size(Labels,1);
     [Keys,Labs] = sf_labkeys(Labels);
 
-    if ~isempty(DefItem) & any(DefItem==[1:nLabels])
+    if ~isempty(DefItem) && any(DefItem==[1:nLabels])
         DefKey = Keys(DefItem);
     else
         DefItem = 0;
@@ -1009,7 +1011,7 @@
         else
             str = input([Prmpt,'? '],'s');
             if isempty(str), str=DefKey; end
-            while isempty(str) | ~any(lower(Keys)==lower(str(1)))
+            while isempty(str) || ~any(lower(Keys)==lower(str(1)))
                 if ~isempty(str),fprintf('%c\t!Out of range\n',7),end
                 str = input([Prmpt,'? '],'s');
                 if isempty(str), str=DefKey; end
@@ -1020,9 +1022,9 @@
 
         p = Values(k,:); if ischar(p), p=deblank(p); end
 
-    elseif strcmp(lower(Type),'bd')
+    elseif strcmpi(Type,'bd')
 
-        if nLabels>3, error('at most 3 labels for GUI ''db'' type'), end
+        if nLabels>3, error('at most 3 labels for GUI ''bd'' type'), end
 
         tmp = cellstr(Labels);
         if DefItem
@@ -1054,6 +1056,7 @@
             'BackgroundColor',COLOUR,...
             'HorizontalAlignment','Right',...
             'Position',PRec);
+        if TTips, set(hPrmpt,'ToolTipString',Prompt); end
     
         if nLabels==1
             %-Only one choice - auto-pick
@@ -1064,7 +1067,7 @@
             
             if TTips, str = ['select with mouse or use kbd: ',...
                 sprintf('%c/',Keys(1:end-1)),Keys(end)];
-            else, str=''; end
+            else str=''; end
         
             %-Store button # in buttons 'UserData' property
             %-Store handle of prompt string in buttons 'Max' property
@@ -1089,7 +1092,7 @@
                 end
                 h = uicontrol(Finter,'Style','Pushbutton',...
                     'String',deblank(Labels(i,:)),...
-                    'ToolTipString',str,...
+                    'ToolTipString',sprintf('%s\n%s',deblank(Labels(i,:)),str),...
                     'Tag',Tag,...
                     'Max',hPrmpt,...
                     'UserData',i,...
@@ -1147,8 +1150,8 @@
 case {'be1','bn1','bw1','bi1','br1','-n1','-w1'}
                                       %-Process button/entry combo types
 %=======================================================================
-if ischar(DefItem), DefStr=DefItem; else, DefStr=num2str(DefItem); end
-if isempty(m), strM=''; else, strM=sprintf(' (<=%d)',m); end
+if ischar(DefItem), DefStr=DefItem; else DefStr=num2str(DefItem); end
+if isempty(m), strM=''; else strM=sprintf(' (<=%d)',m); end
 
 if CmdLine
 
@@ -1169,7 +1172,7 @@
     nLabels     = size(Labels,1);
     [Keys,Labs] = sf_labkeys(Labels);
 
-    if ~isempty(DefItem), DefKey = Keys(DefItem); else, DefKey = ''; end
+    if ~isempty(DefItem), DefKey = Keys(DefItem); else DefKey = ''; end
 
     %-Print banner prompt
     %---------------------------------------------------------------
@@ -1204,7 +1207,7 @@
         else
             str = input([Prmpt,'? '],'s');
             if isempty(str), str=DefKey; end
-            while isempty(str) | ~any(lower(Keys)==lower(str(1)))
+            while isempty(str) || ~any(lower(Keys)==lower(str(1)))
                 if ~isempty(str),fprintf('%c\t!Invalid response\n',7),end
                 str = input([Prmpt,'? '],'s');
                 if isempty(str), str=DefKey; end
@@ -1225,10 +1228,14 @@
         %-"specify option chosen: ask user to specify
         %-------------------------------------------------------
         switch lower(Type(2))
-                case 's', tstr=' string';        case 'e', tstr='n expression';
-                case 'n', tstr=' natural number';case 'w', tstr=' whole number';
-                case 'i', tstr='n integer';      case 'r', tstr=' real number';
-                otherwise, tstr=''; end
+            case 's',  tstr=' string';
+            case 'e',  tstr='n expression';
+            case 'n',  tstr=' natural number';
+            case 'w',  tstr=' whole number';
+            case 'i',  tstr='n integer';
+            case 'r',  tstr=' real number';
+            otherwise, tstr='';
+        end
         
         Prompt = sprintf('%s (a%s%s)',Prompt,tstr,strM);
         if ~isempty(DefStr)
@@ -1263,6 +1270,7 @@
         'BackgroundColor',COLOUR,...
         'HorizontalAlignment','Right',...
         'Position',PRec);
+    if TTips, set(hPrmpt,'ToolTipString',[Prompt,strM]); end
 
     %-Draw buttons & entry widget, & process response
     dX = RRec(3)*(2/3)/nLabels;
@@ -1272,13 +1280,13 @@
     %-Callback sets UserData of prompt string to button number.
     cb = ['set(get(gcbo,''Max''),''UserData'',get(gcbo,''UserData''))'];
     if TTips, str=sprintf('select by mouse or enter value in text widget');
-        else, str=''; end
+    else str=''; end
     H = [];
     for i=1:nLabels
         h = uicontrol(Finter,'Style','Pushbutton',...
             'String',deblank(Labels(i,:)),...
             'Max',hPrmpt,...
-            'ToolTipString',str,...
+            'ToolTipString',sprintf('%s\n%s',deblank(Labels(i,:)),str),...
             'Tag',Tag,...
             'UserData',i,...
             'BackgroundColor',COLOUR,...
@@ -1390,7 +1398,7 @@
 case 'm'                                             %-Process menu type
 %=======================================================================
     nLabels = size(Labels,1);
-    if ~isempty(DefItem) & ~any(DefItem==[1:nLabels]), DefItem=[]; end
+    if ~isempty(DefItem) && ~any(DefItem==[1:nLabels]), DefItem=[]; end
     %-Process pull down menu type
     if CmdLine
         spm_input('!PrntPrmpt',Prompt)
@@ -1409,11 +1417,11 @@
             fprintf('Menu choice: 1 - %s\t(only option)',Labels)
         else
             k = input([Prmpt,' ? ']);
-            if DefItem & isempty(k), k=DefItem; end
-            while isempty(k) | ~any([1:nLabels]==k)
+            if DefItem && isempty(k), k=DefItem; end
+            while isempty(k) || ~any([1:nLabels]==k)
                 if ~isempty(k),fprintf('%c\t!Out of range\n',7),end
                 k = input([Prmpt,' ? ']);
-                if DefItem & isempty(k), k=DefItem; end
+                if DefItem && isempty(k), k=DefItem; end
             end
         end
         fprintf('\n')
@@ -1448,9 +1456,15 @@
                 'UserData',DefItem,...
                 'CallBack',cb,...
                 'Position',QRec);
-            if TTips, set(hPopUp,'ToolTipString',['select with ',...
+            if TTips
+                cLabs  = cellstr(Labels);
+                cInd   = num2cell(1:nLabels);
+                scLabs = [cInd; cLabs'];
+                scLabs = sprintf('%d: %s\n',scLabs{:});
+                set(hPopUp,'ToolTipString',sprintf(['select with ',...
                 'mouse or type option number (1-',...
-                num2str(nLabels),') & press return']), end
+                num2str(nLabels),') & press return\n%s'],scLabs));
+            end
     
             %-Figure ContextMenu for shortcuts
             hM = spm_input('!InptConMen',Finter,[hPopUp,H]);
@@ -1506,15 +1520,15 @@
 %=======================================================================
 % H = spm_input(Prompt,YPos,'p',Labels,cb,UD,XCB)
 %-Condition arguments
-if nargin<7, XCB    = 0;  else, XCB    = varargin{7}; end
-if nargin<6, UD     = []; else, UD     = varargin{6}; end
-if nargin<5, cb     = ''; else, cb     = varargin{5}; end
-if nargin<4, Labels = []; else, Labels = varargin{4}; end
+if nargin<7, XCB    = 0;  else XCB    = varargin{7}; end
+if nargin<6, UD     = []; else UD     = varargin{6}; end
+if nargin<5, cb     = ''; else cb     = varargin{5}; end
+if nargin<4, Labels = []; else Labels = varargin{4}; end
 
 if CmdLine, error('Can''t do CmdLine GUI utilities!'), end
 if isempty(cb), cb = 'disp(''(CallBack not set)'')'; end
 if ischar(cb), cb = cellstr(cb); end
-if length(cb)>1 & strcmp(lower(Type),'m!'), XCB=1; end
+if length(cb)>1 && strcmpi(Type,'m!'), XCB=1; end
 
 if iscellstr(Labels), Labels=char(Labels); end
 %-Convert Labels "option" string to string matrix if required
@@ -1528,7 +1542,7 @@
 end
 
 %-Check #CallBacks
-if ~( length(cb)==1 | (length(cb)==size(Labels,1)) )
+if ~( length(cb)==1 || (length(cb)==size(Labels,1)) )
     error('Labels & Callbacks size mismatch'), end
 
 
@@ -1574,7 +1588,7 @@
 
     H = [];
     for i=1:nLabels
-        if length(cb)>1, tcb=cb(i); else, tcb=cb; end
+        if length(cb)>1, tcb=cb(i); else tcb=cb; end
         if XCB, UD.cb=tcb; tcb = {'spm_input(''!m_cb'')'}; end
         h = uicontrol(Finter,'Style','Pushbutton',...
             'String',deblank(Labels(i,:)),...
@@ -1602,12 +1616,12 @@
 case {'d','d!'}                                        %-Display message
 %=======================================================================
 %-Condition arguments
-if nargin<4, Label=''; else, Label=varargin{4}; end
+if nargin<4, Label=''; else Label=varargin{4}; end
 
-if CmdLine & strcmp(lower(Type),'d')
+if CmdLine && strcmpi(Type,'d')
     fprintf('\n     +-%s%s+',Label,repmat('-',1,57-length(Label)))
     Prompt = [Prompt,' '];
-    while length(Prompt)>0
+    while ~isempty(Prompt)
         tmp = length(Prompt);
         if tmp>56, tmp=min([max(find(Prompt(1:56)==' ')),56]); end
         fprintf('\n     | %s%s |',Prompt(1:tmp),repmat(' ',1,56-tmp))
@@ -1663,19 +1677,19 @@
 % [iCond,msg] = spm_input('!iCond',str,n,m)
 % Parse condition indicator spec strings:
 %   '2 3 2 3', '0 1 0 1', '2323', '0101', 'abab', 'R A R A'
-if nargin<4, m=Inf; else, m=varargin{4}; end
-if nargin<3, n=NaN; else, n=varargin{3}; end
+if nargin<4, m=Inf; else m=varargin{4}; end
+if nargin<3, n=NaN; else n=varargin{3}; end
 if any(isnan(n(:)))
     n=Inf;
-elseif (length(n(:))==2 & ~any(n==1)) | length(n(:))>2
+elseif (length(n(:))==2 && ~any(n==1)) || length(n(:))>2
     error('condition input can only do vectors')
 end
-if nargin<2, i=''; else, i=varargin{2}; end
+if nargin<2, i=''; else i=varargin{2}; end
 if isempty(i), varargout={[],'empty input'}; return, end
 msg = ''; i=i(:)';
 
 if ischar(i)
-    if i(1)=='0' & all(ismember(unique(i(:)),char(abs('0'):abs('9'))))
+    if i(1)=='0' && all(ismember(unique(i(:)),char(abs('0'):abs('9'))))
         %-Leading zeros in a digit list
         msg = sprintf('%s expanded',i);
         z = min(find([diff(i=='0'),1]));
@@ -1701,7 +1715,7 @@
     end
 elseif ~all(floor(i(:))==i(:))
     i = '!'; msg = 'must be integers';
-elseif length(i)==1 & prod(n)>1
+elseif length(i)==1 && prod(n)>1
     msg = sprintf('%d expanded',i);
     i = floor(i./10.^[floor(log10(i)+eps):-1:0]);
     i = i-[0,10*i(1:end-1)];
@@ -1709,7 +1723,7 @@
 
 %-Check size of i & #conditions
 if ~ischar(i), [i,msg] = sf_SzChk(i,n,msg); end
-if ~ischar(i) & isfinite(m) & length(unique(i))~=m
+if ~ischar(i) && isfinite(m) && length(unique(i))~=m
     i = '!'; msg = sprintf('%d conditions required',m);
 end
 
@@ -1719,8 +1733,8 @@
 case '!inptconmen'
 %=======================================================================
 % hM = spm_input('!InptConMen',Finter,H)
-if nargin<3, H=[]; else, H=varargin{3}; end
-if nargin<2, varargout={[]}; else, Finter=varargin{2}; end
+if nargin<3, H=[]; else H=varargin{3}; end
+if nargin<2, varargout={[]}; else Finter=varargin{2}; end
 hM = uicontextmenu('Parent',Finter);
 uimenu(hM,'Label','help on spm_input',...
     'CallBack','spm_help(''spm_input.m'')')
@@ -1739,7 +1753,7 @@
 %=======================================================================
 % [CmdLine,YPos] = spm_input('!CmdLine',YPos)
 %-Sorts out whether to use CmdLine or not & canonicalises YPos
-if nargin<2, YPos=''; else, YPos=varargin{2}; end
+if nargin<2, YPos=''; else YPos=varargin{2}; end
 if isempty(YPos), YPos='+1'; end
 
 CmdLine = [];
@@ -1764,13 +1778,13 @@
 %=======================================================================
 % Finter = spm_input('!GetWin',F)
 %-Locate (or create) figure to work in (Don't use 'Tag'ged figs)
-if nargin<2, F='Interactive'; else, F=varargin{2}; end
+if nargin<2, F='Interactive'; else F=varargin{2}; end
 Finter = spm_figure('FindWin',F);
 if isempty(Finter)
     if any(get(0,'Children'))
         if isempty(get(gcf,'Tag')), Finter = gcf;
-        else, Finter = spm('CreateIntWin'); end
-    else, Finter = spm('CreateIntWin'); end
+        else Finter = spm('CreateIntWin'); end
+    else Finter = spm('CreateIntWin'); end
 end
 varargout = {Finter};
 
@@ -1779,9 +1793,9 @@
 %=======================================================================
 % [PLoc,cF] = spm_input('!PointerJump',RRec,F,XDisp)
 %-Raise window & jump pointer over question
-if nargin<4, XDisp=[]; else, XDisp=varargin{4}; end
-if nargin<3, F='Interactive'; else, F=varargin{3}; end
-if nargin<2, error('Insufficient arguments'), else, RRec=varargin{2}; end
+if nargin<4, XDisp=[]; else XDisp=varargin{4}; end
+if nargin<3, F='Interactive'; else F=varargin{3}; end
+if nargin<2, error('Insufficient arguments'), else RRec=varargin{2}; end
 F = spm_figure('FindWin',F);
 PLoc = get(0,'PointerLocation');
 cF   = get(0,'CurrentFigure');
@@ -1800,8 +1814,8 @@
 %=======================================================================
 % spm_input('!PointerJumpBack',PLoc,cF)
 %-Replace pointer and reset CurrentFigure back
-if nargin<4, cF=[]; else, F=varargin{3}; end
-if nargin<2, error('Insufficient arguments'), else, PLoc=varargin{2}; end
+if nargin<4, cF=[]; else F=varargin{3}; end
+if nargin<2, error('Insufficient arguments'), else PLoc=varargin{2}; end
 if PJump, set(0,'PointerLocation',PLoc), end
 cF = spm_figure('FindWin',cF);
 if ~isempty(cF), set(0,'CurrentFigure',cF); end
@@ -1811,9 +1825,9 @@
 %=======================================================================
 % spm_input('!PrntPrmpt',Prompt,TipStr,Title)
 %-Print prompt for CmdLine questioning
-if nargin<4, Title  = ''; else, Title  = varargin{4}; end
-if nargin<3, TipStr = ''; else, TipStr = varargin{3}; end
-if nargin<2, Prompt = ''; else, Prompt = varargin{2}; end
+if nargin<4, Title  = ''; else Title  = varargin{4}; end
+if nargin<3, TipStr = ''; else TipStr = varargin{3}; end
+if nargin<2, Prompt = ''; else Prompt = varargin{2}; end
 if isempty(Prompt), Prompt='Enter an expression'; end
 
 Prompt = cellstr(Prompt);
@@ -1833,16 +1847,16 @@
     fprintf('\n= %s %s\n',Title,repmat('~',1,72-length(Title)-3))
 end
 fprintf('\t%s',Prompt{1})
-for i=2:prod(size(Prompt)), fprintf('\n\t%s',Prompt{i}), end
+for i=2:numel(Prompt), fprintf('\n\t%s',Prompt{i}), end
 fprintf('%s\n%s\n',TipStr,repmat('~',1,72))
 
 
 case '!inputrects'
 %=======================================================================
 % [Frec,QRec,PRec,RRec,Sz,Se] = spm_input('!InputRects',YPos,rec,F)
-if nargin<4, F='Interactive'; else, F=varargin{4}; end
-if nargin<3, rec=''; else, rec=varargin{3}; end
-if nargin<2, YPos=1; else, YPos=varargin{2}; end
+if nargin<4, F='Interactive'; else F=varargin{4}; end
+if nargin<3, rec=''; else rec=varargin{3}; end
+if nargin<2, YPos=1; else YPos=varargin{2}; end
 F = spm_figure('FindWin',F);
 if isempty(F), error('Figure not found'), end
 
@@ -1876,7 +1890,7 @@
 case '!deleteinputobj'
 %=======================================================================
 % spm_input('!DeleteInputObj',F)
-if nargin<2, F='Interactive'; else, F=varargin{2}; end
+if nargin<2, F='Interactive'; else F=varargin{2}; end
 h = spm_input('!FindInputObj',F);
 delete(h(h>0))
 
@@ -1886,7 +1900,7 @@
 % [CPos,hCPos] = spm_input('!CurrentPos',F)
 % h            = spm_input('!FindInputObj',F)
 % hPos contains handles: Columns contain handles corresponding to Pos
-if nargin<2, F='Interactive'; else, F=varargin{2}; end
+if nargin<2, F='Interactive'; else F=varargin{2}; end
 F = spm_figure('FindWin',F);
 
 %-Find tags and YPos positions of 'GUIinput_' 'Tag'ged objects
@@ -1927,10 +1941,10 @@
 %=======================================================================
 % [NPos,CPos,hCPos] = spm_input('!NextPos',YPos,F,CmdLine)
 %-Return next position to use
-if nargin<3, F='Interactive'; else, F=varargin{3}; end
-if nargin<2, YPos='+1'; else, YPos=varargin{2}; end
+if nargin<3, F='Interactive'; else F=varargin{3}; end
+if nargin<2, YPos='+1'; else YPos=varargin{2}; end
 if nargin<4, [CmdLine,YPos]=spm_input('!CmdLine',YPos);
-    else, CmdLine=varargin{4}; end
+else CmdLine=varargin{4}; end
 
 F = spm_figure('FindWin',F);
 
@@ -1968,10 +1982,10 @@
 %=======================================================================
 % NPos = spm_input('!SetNextPos',YPos,F,CmdLine)
 %-Set next position to use
-if nargin<3, F='Interactive'; else, F=varargin{3}; end
-if nargin<2, YPos='+1'; else, YPos=varargin{2}; end
+if nargin<3, F='Interactive'; else F=varargin{3}; end
+if nargin<2, YPos='+1'; else YPos=varargin{2}; end
 if nargin<4, [CmdLine,YPos]=spm_input('!CmdLine',YPos);
-    else, CmdLine=varargin{4}; end
+else CmdLine=varargin{4}; end
 
 %-Find out which Y-position to use
 [NPos,CPos,hCPos] = spm_input('!NextPos',YPos,F,CmdLine);
@@ -1986,7 +2000,7 @@
 % MPos = spm_input('!MaxPos',F,FRec3)
 %
 if nargin<3
-    if nargin<2, F='Interactive'; else, F=varargin{2}; end
+    if nargin<2, F='Interactive'; else F=varargin{2}; end
     F = spm_figure('FindWin',F);
     if isempty(F)
         FRec3=spm('WinSize','Interactive')*[0;0;0;1];
@@ -2008,14 +2022,14 @@
 case '!editablekeypressfcn'
 %=======================================================================
 % spm_input('!EditableKeyPressFcn',h,ch,hPrmpt)
-if nargin<2, error('Insufficient arguments'), else, h=varargin{2}; end
+if nargin<2, error('Insufficient arguments'), else h=varargin{2}; end
 if isempty(h), set(gcbf,'KeyPressFcn','','UserData',[]), return, end
-if nargin<3, ch=get(get(h,'Parent'),'CurrentCharacter'); else, ch=varargin{3};end
-if nargin<4, hPrmpt=get(h,'UserData'); else, hPrmpt=varargin{4}; end
+if nargin<3, ch=get(get(h,'Parent'),'CurrentCharacter'); else ch=varargin{3};end
+if nargin<4, hPrmpt=get(h,'UserData'); else hPrmpt=varargin{4}; end
 
 tmp = get(h,'String');
 if isempty(tmp), tmp=''; end
-if iscellstr(tmp) & length(tmp)==1; tmp=tmp{:}; end
+if iscellstr(tmp) && length(tmp)==1; tmp=tmp{:}; end
 
 if isempty(ch)                  %- shift / control / &c. pressed
     return
@@ -2026,7 +2040,7 @@
     tmp = '';
 elseif any(abs(ch)==[8,127])            %-BackSpace or Delete
     if iscellstr(tmp), return, end
-    if length(tmp), tmp(length(tmp))=''; end
+    if ~isempty(tmp), tmp(length(tmp))=''; end
 elseif abs(ch)==13              %-Return pressed
     if ~isempty(tmp)
         set(hPrmpt,'UserData',get(h,'String'))
@@ -2046,16 +2060,16 @@
 % DefItem if (DefItem~=0) & return (ASCII-13) is pressed
 
 %-Condition arguments
-if nargin<2, error('Insufficient arguments'), else, h=varargin{2}; end
+if nargin<2, error('Insufficient arguments'), else h=varargin{2}; end
 if isempty(h), set(gcf,'KeyPressFcn','','UserData',[]), return, end
-if nargin<3, error('Insufficient arguments'); else, Keys=varargin{3}; end
-if nargin<4, DefItem=0; else, DefItem=varargin{4}; end
-if nargin<5, ch=get(gcf,'CurrentCharacter'); else, ch=varargin{5}; end
+if nargin<3, error('Insufficient arguments'); else Keys=varargin{3}; end
+if nargin<4, DefItem=0; else DefItem=varargin{4}; end
+if nargin<5, ch=get(gcf,'CurrentCharacter'); else ch=varargin{5}; end
 
 if isempty(ch)
     %- shift / control / &c. pressed
     return
-elseif (DefItem & ch==13)
+elseif (DefItem && ch==13)
     But = DefItem;
 else
     But = find(lower(ch)==lower(Keys));
@@ -2066,10 +2080,10 @@
 case '!pulldownkeypressfcn'
 %=======================================================================
 % spm_input('!PullDownKeyPressFcn',h,ch,DefItem)
-if nargin<2, error('Insufficient arguments'), else, h=varargin{2}; end
+if nargin<2, error('Insufficient arguments'), else h=varargin{2}; end
 if isempty(h), set(gcf,'KeyPressFcn',''), return, end
-if nargin<3, ch=get(get(h,'Parent'),'CurrentCharacter'); else, ch=varargin{3};end
-if nargin<4, DefItem=get(h,'UserData'); else, ch=varargin{4}; end
+if nargin<3, ch=get(get(h,'Parent'),'CurrentCharacter'); else ch=varargin{3};end
+if nargin<4, DefItem=get(h,'UserData'); else ch=varargin{4}; end
 
 Pmax = get(h,'Max');
 Pval = get(h,'Value');
@@ -2109,7 +2123,7 @@
 
 %-Get PopUp's UserData, check cb and UD fields exist, extract cb & UD
 tmp = get(h,'UserData');
-if ~(isfield(tmp,'cb') & isfield(tmp,'UD'))
+if ~(isfield(tmp,'cb') && isfield(tmp,'UD'))
     error('Invalid UserData structure for spm_input extended callback')
 end
 cb  = tmp.cb;
@@ -2118,15 +2132,15 @@
 %-Evaluate appropriate CallBack string (ignoring any return arguments)
 % NB: Using varargout={eval(cb{n})}; gives an error if the CallBack 
 % has no return arguments!
-if length(cb)==1, eval(char(cb)); else, eval(cb{n}); end
+if length(cb)==1, eval(char(cb)); else eval(cb{n}); end
 
 
 case '!dscroll'
 %=======================================================================
 % spm_input('!dScroll',h,Prompt)
 %-Scroll text in object h
-if nargin<2, return, else, h=varargin{2}; end
-if nargin<3, Prompt = get(h,'UserData'); else, Prompt=varargin{3}; end
+if nargin<2, return, else h=varargin{2}; end
+if nargin<3, Prompt = get(h,'UserData'); else Prompt=varargin{3}; end
 tmp = Prompt;
 if length(Prompt)>56
     while length(tmp)>56
@@ -2197,9 +2211,9 @@
     p = evalin('base',['[',str,']'],'''!''');
     if ischar(p)
         msg = 'evaluation error';
-    elseif any(floor(p(:))~=p(:)|p(:)<1)|~isreal(p)
+    elseif any(floor(p(:))~=p(:)|p(:)<1)||~isreal(p)
         p='!'; msg='natural number(s) required';
-    elseif ~isempty(m) & any(p(:)>m)
+    elseif ~isempty(m) && any(p(:)>m)
         p='!'; msg=['max value is ',num2str(m)];
     else
         [p,msg] = sf_SzChk(p,n);
@@ -2208,9 +2222,9 @@
     p = evalin('base',['[',str,']'],'''!''');
     if ischar(p)
         msg = 'evaluation error';
-    elseif any(floor(p(:))~=p(:)|p(:)<0)|~isreal(p)
+    elseif any(floor(p(:))~=p(:)|p(:)<0)||~isreal(p)
         p='!'; msg='whole number(s) required';
-    elseif ~isempty(m) & any(p(:)>m)
+    elseif ~isempty(m) && any(p(:)>m)
         p='!'; msg=['max value is ',num2str(m)];
     else
         [p,msg] = sf_SzChk(p,n);
@@ -2219,7 +2233,7 @@
     p = evalin('base',['[',str,']'],'''!''');
     if ischar(p)
         msg = 'evaluation error';
-    elseif any(floor(p(:))~=p(:))|~isreal(p)
+    elseif any(floor(p(:))~=p(:))||~isreal(p)
         p='!'; msg='integer(s) required';
     else
         [p,msg] = sf_SzChk(p,n);
@@ -2228,7 +2242,7 @@
     p = evalin('base',['[',str,']'],'''!''');
     if ischar(p)
         msg = 'evaluation error';
-    elseif length(setxor(p(:)',m))
+    elseif ~isempty(setxor(p(:)',m))
         p='!'; msg='invalid permutation';
     else
         [p,msg] = sf_SzChk(p,n);
@@ -2239,7 +2253,7 @@
         msg = 'evaluation error';
     elseif ~isreal(p)
         p='!'; msg='real number(s) required';
-    elseif ~isempty(m) & ( max(p)>max(m) | min(p)<min(m) )
+    elseif ~isempty(m) && ( max(p)>max(m) || min(p)<min(m) )
         p='!'; msg=sprintf('real(s) in [%g,%g] required',min(m),max(m));
     else
         [p,msg] = sf_SzChk(p,n);
@@ -2252,7 +2266,7 @@
     if isempty(n), n=1; end
 
     %-Sort out contrast matrix dimensions (contrast vectors in rows)
-    if length(n)==1, n=[n,Inf]; else, n=reshape(n(1:2),1,2); end
+    if length(n)==1, n=[n,Inf]; else n=reshape(n(1:2),1,2); end
     if ~isempty(X)      % - override n(2) w/ design column dimension
         n(2) = spm_SpUtil('size',X,2);
     end
@@ -2261,10 +2275,10 @@
     if ischar(p)
         msg = 'evaluation error';
     else
-        if isfinite(n(2)) & size(p,2)<n(2)
+        if isfinite(n(2)) && size(p,2)<n(2)
             tmp = n(2) -size(p,2);
             p   = [p, zeros(size(p,1),tmp)];
-            if size(p,1)>1, str=' columns'; else, str='s'; end
+            if size(p,1)>1, str=' columns'; else str='s'; end
             msg = sprintf('right padded with %d zero%s',tmp,str);
         else
             msg = '';
@@ -2272,11 +2286,11 @@
 
         if size(p,2)>n(2)
             p='!'; msg=sprintf('too long - only %d prams',n(2));
-        elseif isfinite(n(1)) & size(p,1)~=n(1)
+        elseif isfinite(n(1)) && size(p,1)~=n(1)
             p='!';
             if n(1)==1, msg='vector required';
-                else, msg=sprintf('%d contrasts required',n(1)); end
-        elseif ~isempty(X) & ~spm_SpUtil('allCon',X,p')
+            else msg=sprintf('%d contrasts required',n(1)); end
+        elseif ~isempty(X) && ~spm_SpUtil('allCon',X,p')
             p='!'; msg='invalid contrast';
         end
     end
@@ -2288,60 +2302,61 @@
 
 function str = sf_SzStr(n,l)
 %=======================================================================
-%-Size info string constuction
-if nargin<2, l=0; else, l=1; end
+%-Size info string construction
+if nargin<2, l=0; else l=1; end
 if nargin<1, error('insufficient arguments'), end
 if isempty(n), n=NaN; end
 n=n(:); if length(n)==1, n=[n,1]; end, dn=length(n);
-if any(isnan(n)) | (prod(n)==1 & dn<=2) | (dn==2 & min(n)==1 & isinf(max(n)))
+if any(isnan(n)) || (prod(n)==1 && dn<=2) || (dn==2 && min(n)==1 && isinf(max(n)))
     str = ''; lstr = '';
-elseif dn==2 & min(n)==1
+elseif dn==2 && min(n)==1
     str = sprintf('[%d]',max(n));   lstr = [str,'-vector'];
-elseif dn==2 & sum(isinf(n))==1
+elseif dn==2 && sum(isinf(n))==1
     str = sprintf('[%d]',min(n));   lstr = [str,'-vector(s)'];
 else
-    str=''; for i = 1:dn
+    str='';
+    for i = 1:dn
         if isfinite(n(i)), str = sprintf('%s,%d',str,n(i));
-        else, str = sprintf('%s,*',str); end
+        else str = sprintf('%s,*',str); end
     end
     str = ['[',str(2:end),']']; lstr = [str,'-matrix'];
 end
-if l, str=sprintf('\t%s',lstr); else, str=[str,' ']; end
+if l, str=sprintf('\t%s',lstr); else str=[str,' ']; end
 
 
 function [p,msg] = sf_SzChk(p,n,msg)
 %=======================================================================
 %-Size checking
 if nargin<3, msg=''; end
-if nargin<2, n=[]; end, if isempty(n), n=NaN; else, n=n(:)'; end
+if nargin<2, n=[]; end, if isempty(n), n=NaN; else n=n(:)'; end
 if nargin<1, error('insufficient arguments'), end
 
-if ischar(p) | any(isnan(n(:))), return, end
+if ischar(p) || any(isnan(n(:))), return, end
 if length(n)==1, n=[n,1]; end
 
 dn = length(n);
 sp = size(p);
 dp = ndims(p);
 
-if dn==2 & min(n)==1
+if dn==2 && min(n)==1
     %-[1,1], [1,n], [n,1], [1,Inf], [Inf,1] - vector - allow transpose
     %---------------------------------------------------------------
     i = min(find(n==max(n)));
-    if n(i)==1 & max(sp)>1
+    if n(i)==1 && max(sp)>1
         p='!'; msg='scalar required';
-    elseif ndims(p)~=2 | ~any(sp==1) | ( isfinite(n(i)) & max(sp)~=n(i) )
+    elseif ndims(p)~=2 || ~any(sp==1) || ( isfinite(n(i)) && max(sp)~=n(i) )
         %-error: Not2D | not vector | not right length
-        if isfinite(n(i)), str=sprintf('%d-',n(i)); else, str=''; end
+        if isfinite(n(i)), str=sprintf('%d-',n(i)); else str=''; end
         p='!'; msg=[str,'vector required'];
-    elseif sp(i)==1 & n(i)~=1
+    elseif sp(i)==1 && n(i)~=1
         p=p'; msg=[msg,' (input transposed)'];
     end
 
-elseif dn==2 & sum(isinf(n))==1
+elseif dn==2 && sum(isinf(n))==1
     %-[n,Inf], [Inf,n] - n vector(s) required - allow transposing
     %---------------------------------------------------------------
     i = find(isfinite(n));
-    if ndims(p)~=2 | ~any(sp==n(i))
+    if ndims(p)~=2 || ~any(sp==n(i))
         p='!'; msg=sprintf('%d-vector(s) required',min(n));
     elseif sp(i)~=n
         p=p'; msg=[msg,' (input transposed)'];
@@ -2350,11 +2365,11 @@
 else
     %-multi-dimensional matrix required - check dimensions
     %---------------------------------------------------------------
-    if ndims(p)~=dn | ~all( size(p)==n | isinf(n) )
+    if ndims(p)~=dn || ~all( size(p)==n | isinf(n) )
         p = '!'; msg='';
         for i = 1:dn
             if isfinite(n(i)), msg = sprintf('%s,%d',msg,n(i));
-            else, msg = sprintf('%s,*',msg); end
+            else msg = sprintf('%s,*',msg); end
         end
         msg = ['[',msg(2:end),']-matrix required'];
     end
diff --git a/spm_int.m b/spm_int.m
index d0c7902..80f25f7 100644
--- a/spm_int.m
+++ b/spm_int.m
@@ -55,7 +55,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_int.m 3812 2010-04-07 16:52:05Z karl $
+% $Id: spm_int.m 4121 2010-11-17 16:16:18Z karl $
  
  
 % convert U to U.u if necessary
@@ -63,7 +63,12 @@
 if ~isstruct(U), u.u = U; U = u; end
 try, dt = U.dt; catch, U.dt = 1; end
  
- 
+% number of times to sample (v) and number of microtime bins (u)
+%--------------------------------------------------------------------------
+u       = size(U.u,1);
+try,  v = M.ns;  catch, v = u;   end
+
+
 % get expansion point
 %--------------------------------------------------------------------------
 x = [1; spm_vec(M.x)];
@@ -76,14 +81,7 @@
     M.x = sparse(0,0);
 end
  
-% number of times to sample (v) and number of microtime bins (u)
-%--------------------------------------------------------------------------
-u = size(U.u,1);
-try
-    v = M.ns;
-catch
-    v = size(U.u,1);
-end
+
  
 % output nonlinearity, if specified
 %--------------------------------------------------------------------------
@@ -102,7 +100,7 @@
 % delays
 %--------------------------------------------------------------------------
 try
-    D  = round(M.delays/U.dt);
+    D  = max(round(M.delays/U.dt),1);
 catch
     D  = ones(M.l,1)*round(u/v);
 end
@@ -121,7 +119,7 @@
 s     = ceil([0:v - 1]*u/v);
 for j = 1:M.l
     i       = s + D(j);
-    sy(j,:) = sparse(1,i,[1:v],1,u);
+    sy(j,:) = sparse(1,i,1:v,1,u);
 end
  
 % time in seconds
diff --git a/spm_int_B.m b/spm_int_B.m
index 44a68ff..c68fccc 100644
--- a/spm_int_B.m
+++ b/spm_int_B.m
@@ -59,15 +59,13 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_int_B.m 1188 2008-03-05 17:14:43Z karl $
+% $Id: spm_int_B.m 4121 2010-11-17 16:16:18Z karl $
  
  
 % convert U to U.u if necessary
 %--------------------------------------------------------------------------
 if ~isstruct(U), U.u = U; end
 try, dt = U.dt; catch, dt = 1; end
-try, ns = M.ns; catch, ns = length(U.u); end
-
  
 % state equation; add [0] states if not specified
 %--------------------------------------------------------------------------
@@ -111,7 +109,7 @@
 % integrate
 %==========================================================================
 x0    = spm_vec(M.x);
-for i = 1:ns
+for i = 1:size(U.u,1)
  
     % input
     %----------------------------------------------------------------------
diff --git a/spm_int_D.m b/spm_int_D.m
index c54d15b..c146140 100644
--- a/spm_int_D.m
+++ b/spm_int_D.m
@@ -57,7 +57,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_int_D.m 3739 2010-02-26 13:12:44Z karl $
+% $Id: spm_int_D.m 4121 2010-11-17 16:16:18Z karl $
  
  
 % convert U to U.u if necessary
@@ -65,7 +65,12 @@
 if ~isstruct(U), u.u = U; U = u; end
 try, dt = U.dt; catch, U.dt = 1; end
  
- 
+ % number of times to sample (v) and number of microtime bins (u)
+%--------------------------------------------------------------------------
+u      = size(U.u,1);
+try, v = M.ns;  catch, v = u;    end
+
+
 % get expansion point
 %--------------------------------------------------------------------------
 x = [1; spm_vec(M.x)];
@@ -78,14 +83,7 @@
     M.x = sparse(0,0);
 end
  
-% number of times to sample (v) and number of microtime bins (u)
-%--------------------------------------------------------------------------
-u  = size(U.u,1);
-try
-    v = M.ns;
-catch
-    v = size(U.u,1);
-end
+
  
 % output nonlinearity, if specified
 %--------------------------------------------------------------------------
@@ -106,9 +104,9 @@
 % delays
 %--------------------------------------------------------------------------
 try
-    D  = round(M.delays/U.dt);
+    D  = max(round(M.delays/U.dt),1);
 catch
-    D  = ones(l,1)*round(u/v);
+    D  = ones(M.l,1)*round(u/v);
 end
  
 % state-dependent effects to include during integration
@@ -117,7 +115,7 @@
     M2 = M2(M.states);
     n  = length(M2);
 end
- 
+
 % decrease integration time steps
 %--------------------------------------------------------------------------
 try
@@ -125,8 +123,8 @@
 catch
     N  = 1;
 end
- 
- 
+
+
 % Evaluation times (t) and indicator array for inputs (su) and output (sy)
 %==========================================================================
  
@@ -140,22 +138,21 @@
 s     = ceil([0:v - 1]*u/v);
 for j = 1:M.l
     i       = s + D(j);
-    sy(j,:) = sparse(1,i,[1:v],1,u);
+    sy(j,:) = sparse(1,i,1:v,1,u);
 end
  
 % get (N) intervening times to evaluate
 %--------------------------------------------------------------------------
-i     = ceil([0:(v - 1)*N]*u/v/N) + D(1);
+i     = ceil((0:(v - 1)*N)*u/v/N) + D(1);
 sx    = sparse(1,i,1,1,u);
  
 % time in seconds
 %--------------------------------------------------------------------------
-t     = find(su | any(sy) | sx);
+t     = find(su | any(sy,1) | sx);
 su    = full(su(:,t));
 sy    = full(sy(:,t));
 dt    = [diff(t) 0]*U.dt;
  
- 
 % Integrate
 %--------------------------------------------------------------------------
 y     = zeros(M.l,v);
diff --git a/spm_int_J.m b/spm_int_J.m
index 90cd54e..9b51aec 100644
--- a/spm_int_J.m
+++ b/spm_int_J.m
@@ -61,15 +61,14 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_int_J.m 3657 2009-12-23 20:22:10Z karl $
+% $Id: spm_int_J.m 4121 2010-11-17 16:16:18Z karl $
 
 
 % convert U to U.u if necessary and M(1) to M
 %--------------------------------------------------------------------------
 if ~isstruct(U), u.u = U; U = u; end
-M       = M(1);
 try, dt = U.dt;  catch, dt = 1;  end
-try, ns = M.ns;  catch, ns = length(U.u); end
+M       = M(1);
 
 % state equation; add [0] states if not specified
 %--------------------------------------------------------------------------
@@ -127,7 +126,7 @@
 
 % integrate
 %==========================================================================
-for i = 1:ns
+for i = 1:size(U.u,1)
 
     % input
     %----------------------------------------------------------------------
diff --git a/spm_int_L.m b/spm_int_L.m
index 5a5cf65..79b00c1 100644
--- a/spm_int_L.m
+++ b/spm_int_L.m
@@ -58,15 +58,13 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_int_L.m 3705 2010-02-01 20:51:28Z karl $
+% $Id: spm_int_L.m 4281 2011-03-31 19:49:57Z karl $
  
  
 % convert U to U.u if necessary
 %--------------------------------------------------------------------------
 if ~isstruct(U), u.u = U; U = u;         end
 try, dt = U.dt;        catch, dt = 1;    end
-try, ns = size(U.u,1); catch, ns = M.ns; end
-
  
 % state equation; add [0] states if not specified
 %--------------------------------------------------------------------------
@@ -116,11 +114,11 @@
 p     = max(abs(real(eig(dfdx))));
 N     = ceil(max(1,dt*p*2));
 n     = length(spm_vec(x));
-Q     = (spm_expm(dt*D*dfdx/N) - speye(n,n))*pinv(dfdx);
+Q     = (spm_expm(dt*D*dfdx/N) - speye(n,n))*spm_inv(dfdx);
  
 % integrate
 %==========================================================================
-for i = 1:ns
+for i = 1:size(U.u,1)
  
     % input
     %----------------------------------------------------------------------
diff --git a/spm_int_ode.m b/spm_int_ode.m
index a1d90e4..3b1611a 100644
--- a/spm_int_ode.m
+++ b/spm_int_ode.m
@@ -60,28 +60,18 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_int_ode.m 2707 2009-02-06 19:51:34Z karl $
+% $Id: spm_int_ode.m 4121 2010-11-17 16:16:18Z karl $
 
 
 % convert U to U.u if necessary
 %--------------------------------------------------------------------------
-if ~isstruct(U)
-    U.u = U;
-end
-try
-    dt = U.dt;
-catch
-    dt = 1;
-end
-try
-    ns = length(U.u);
-catch
-    ns = M.ns;
-end
- 
+if ~isstruct(U), U.u = U;   end
+try, U.dt; catch, U.dt = 1; end
+
 % sample times
 %--------------------------------------------------------------------------
-tspan = [1:ns]*dt;
+ns      = size(U.u,1);
+tspan   = (1:ns)*U.dt;
  
 % state equation; add [0] states if not specified
 %--------------------------------------------------------------------------
@@ -111,12 +101,21 @@
 end
  
 % ODE45 functional form (note, the third argument of the ODE function is
-% used by ode15i (i.e, yp0).
+% used by ode15i (i.e, OPTIONS).
 %--------------------------------------------------------------------------
 ode   = inline('spm_vec(f(spm_unvec(x,M.x),U.u(ceil(t/U.dt),:),P,M))',...
-               't','x','yp0','P','M','U','f');
+               't','x','OPTIONS','P','M','U','f');
            
-[t x] = ode113(ode,tspan,spm_vec(x),[],P,M,U,f);
+
+OPTIONS = odeset;
+[t x]   = ode113(ode,tspan,spm_vec(M.x),OPTIONS,P,M,U,f);
+
+% Give the integrator something to work with for 'flat' inputs:
+%--------------------------------------------------------------------------
+if norm(x,1) < exp(-16)
+   U.u   = spm_conv(U.u,8,0); 
+   [t x] = ode113(ode,tspan,spm_vec(M.x),OPTIONS,P,M,U,f);
+end
  
  
 % output
@@ -125,7 +124,7 @@
  
     % output - implement g(x)
     %----------------------------------------------------------------------
-    if length(g)
+    if ~isempty(g)
         y(:,i) = spm_vec(g(spm_unvec(x(i,:),M.x),U.u(i,:),P,M));
     else
         y(:,i) = spm_vec(spm_unvec(x(i,:),M.x));
diff --git a/spm_int_sde.m b/spm_int_sde.m
index 9d380c5..6e12adf 100644
--- a/spm_int_sde.m
+++ b/spm_int_sde.m
@@ -61,15 +61,14 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_int_sde.m 2032 2008-09-02 18:31:16Z karl $
+% $Id: spm_int_sde.m 4121 2010-11-17 16:16:18Z karl $
 
 
 % convert U to U.u if necessary and M(1) to M
 %--------------------------------------------------------------------------
 if ~isstruct(U), u.u = U; U = u; end
-M       = M(1);
 try, dt = U.dt;  catch, dt = 1;  end
-try, ns = M.ns;  catch, ns = length(U.u); end
+M       = M(1);
 
 % state equation; add [0] states if not specified
 %--------------------------------------------------------------------------
@@ -86,7 +85,7 @@
 try
     g   = fcnchk(M.g,'x','u','P','M');
 catch
-    g   = [];
+    g   = inline('x','x','u','P','M');
 end
 
 % Initial states and inputs
@@ -134,7 +133,7 @@
 
 % integrate
 %==========================================================================
-for i = 1:ns
+for i = 1:size(U.u,1)
 
     % input
     %----------------------------------------------------------------------
@@ -158,11 +157,7 @@
 
     % output - implement g(x)
     %----------------------------------------------------------------------
-    if length(g)
-        y(:,i) = spm_vec(g(x,u,P,M));
-    else
-        y(:,i) = spm_vec(x);
-    end
+    y(:,i) = spm_vec(g(x,u,P,M));
 
 end
 
diff --git a/spm_inv.m b/spm_inv.m
index 96ba287..a298e97 100644
--- a/spm_inv.m
+++ b/spm_inv.m
@@ -1,22 +1,30 @@
-function X = spm_inv(A)
+function X = spm_inv(A,TOL)
 % inverse for ill-conditioned matrices
-% FORMAT X = spm_inv(A)
+% FORMAT X = spm_inv(A,TOL)
 %
-% X   - matrix
+% A   - matrix
+% X   - inverse
 %
-% This routine simply adds a small identity matrix to A and calls inv.m
+% TOL - tolerance: default = max(eps(norm(A,'inf'))*max(m,n),exp(-32))
+%
+% This routine simply adds a small diagonal matrix to A and calls inv.m
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_inv.m 3877 2010-05-07 19:49:35Z karl $
+% $Id: spm_inv.m 4281 2011-03-31 19:49:57Z karl $
  
 % check A 
 %--------------------------------------------------------------------------
 [m,n] = size(A);
 if isempty(A), X = sparse(n,m); return, end
  
+% tolerance
+%--------------------------------------------------------------------------
+if nargin == 1
+    TOL  = max(eps(norm(A,'inf'))*max(m,n),exp(-32)); 
+end
+
 % inverse
 %--------------------------------------------------------------------------
-TOL = eps(norm(A,'inf'))*max(m,n);
-X   = inv(A + speye(m,n)*TOL);
+X     = inv(sparse(A) + speye(m,n)*TOL);
diff --git a/spm_invBcdf.m b/spm_invBcdf.m
index d00e1b0..e657437 100644
--- a/spm_invBcdf.m
+++ b/spm_invBcdf.m
@@ -16,7 +16,7 @@
 % random variable with shape parameters v and w.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Beta distribution has two shape parameters, v and w, and is
 % defined for v>0 & w>0 and for x in [0,1] (See Evans et al., Ch5).
 % The Cumulative Distribution Function (CDF) F(x) is the probability
@@ -26,15 +26,15 @@
 % definitions of the incomplete beta function.
 %
 % Variate relationships:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Many: See Evans et al., Ch5
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Interval bisection to find the inverse CDF to within tol.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -47,65 +47,66 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invBcdf.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_invBcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Parameters
-%-----------------------------------------------------------------------
-Dtol   = 10^-8;
+%--------------------------------------------------------------------------
+Dtol  = 10^-8;
 maxIt = 10000;
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<4, tol = Dtol; end
 if nargin<3, error('Insufficient arguments'), end
 
 ad = [ndims(F);ndims(v);ndims(w)];
 rd = max(ad);
-as = [  [size(F),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];...
-    [size(w),ones(1,rd-ad(3))]     ];
+as = [[size(F),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];...
+      [size(w),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation - Undefined and special cases
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 x = zeros(rs);
 
 %-Only defined for F in [0,1] & strictly positive v & w.
 % Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  v>0  &  w>0 );
-if any(~md(:)), x(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    x(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special cases: x=0 when F=0, x=1 when F=1
 x(md & F==1) = 1;
 
 %-Compute where defined & not special case
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 Q  = find( md  &  F>0  &  F<1 );
 if isempty(Q), return, end
-if xa(1), FQ=F(Q); FQ=FQ(:); else, FQ=F*ones(length(Q),1); end
-if xa(2), vQ=v(Q); vQ=vQ(:); else, vQ=v*ones(length(Q),1); end
-if xa(3), wQ=w(Q); wQ=wQ(:); else, wQ=w*ones(length(Q),1); end
-%-?Q=?Q(:) stuff is to avoid discrepant orientations of vector arguments!
+if xa(1), FQ=F(Q); FQ=FQ(:); else FQ=F*ones(length(Q),1); end
+if xa(2), vQ=v(Q); vQ=vQ(:); else vQ=v*ones(length(Q),1); end
+if xa(3), wQ=w(Q); wQ=wQ(:); else wQ=w*ones(length(Q),1); end
 
 %-Interval bisection
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 a  = zeros(length(Q),1); fa = a-FQ;
 b  =  ones(length(Q),1); fb = b-FQ;
 i  = 0;
 xQ = a+1/2;
 QQ = 1:length(Q);
 
-while length(QQ) &  i<maxIt
+while ~isempty(QQ) &&  i<maxIt
     fxQQ        = betainc(xQ(QQ),vQ(QQ),wQ(QQ))-FQ(QQ);
     mQQ         = fa(QQ).*fxQQ > 0;
     a(QQ(mQQ))  = xQ(QQ(mQQ));   fa(QQ(mQQ))  = fxQQ(mQQ);
@@ -115,6 +116,8 @@
     i           = i+1;
 end
 
-if i==maxIt, warning('convergence criteria not reached - maxIt reached'), end
+if i==maxIt
+    warning('convergence criteria not reached - maxIt reached');
+end
 
 x(Q) = xQ;
diff --git a/spm_invFcdf.m b/spm_invFcdf.m
index 4dda7dd..b135d04 100644
--- a/spm_invFcdf.m
+++ b/spm_invFcdf.m
@@ -9,13 +9,13 @@
 % v  - Shape parameter 1 /   numerator degrees of freedom (v>0)
 % w  - Shape parameter 2 / denominator degrees of freedom (w>0)
 % x  - F-variate   (F has range [0,Inf) )
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Fcdf implements the inverse Cumulative Distribution Function
 % for the F-distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF F(x) of the F distribution with degrees of freedom v & w,
 % defined for positive integer degrees of freedom v & w, is the
 % probability that a realisation of an F random variable X has value
@@ -23,7 +23,7 @@
 % for v>0 & w>0, and for x in [0,Inf) (See Evans et al., Ch16).
 %
 % Variate relationships: (Evans et al., Ch16 & 37)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The square of a Student's t variate with w degrees of freedom is
 % distributed as an F-distribution with [1,w] degrees of freedom.
 %
@@ -32,7 +32,7 @@
 % w/2 & v/2, as described below.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Using the routine spm_invBcdf for the Beta distribution, with
 % appropriate parameters:  The CDF of the F-distribution with v,w
 % degrees of freedom is related to the incomplete beta function by:
@@ -44,7 +44,7 @@
 %
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -57,15 +57,15 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1993-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invFcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_invFcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 %-Unpack degrees of freedom v & w from single df parameter (v)
@@ -87,24 +87,27 @@
 %-Check argument sizes
 ad = [ndims(F);ndims(v);ndims(w)];
 rd = max(ad);
-as = [  [size(F),ones(1,rd-ad(1))];...
-    [size(v),ones(1,rd-ad(2))];...
-    [size(w),ones(1,rd-ad(3))]     ];
+as = [[size(F),ones(1,rd-ad(1))];...
+      [size(v),ones(1,rd-ad(2))];...
+      [size(w),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 x = zeros(rs);
 
 %-Only defined for F in [0,1] & strictly positive v & w.
 % Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  v>0  &  w>0 );
-if any(~md(:)), x(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    x(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special cases: x=0 when F=0, x=Inf when F=1
 x(md & F==1) = Inf;
diff --git a/spm_invGcdf.m b/spm_invGcdf.m
index cebd1c4..79a7ad6 100644
--- a/spm_invGcdf.m
+++ b/spm_invGcdf.m
@@ -13,7 +13,7 @@
 % for Gamma distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Gamma distribution has shape parameter h and scale l, and is
 % defined for h>0 & l>0 and for x in [0,Inf) (See Evans et al., Ch18,
 % but note that this reference uses the alternative parameterisation of
@@ -24,7 +24,7 @@
 % x. F(x)=Pr{X<x}:  (See Evans et al., Ch18)
 %
 % Variate relationships: (Evans et al., Ch8 & Ch18)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % For natural (strictly +ve integer) shape h this is an Erlang distribution.
 %
 % The Standard Gamma distribution has a single parameter, the shape h.
@@ -34,11 +34,11 @@
 % to the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Interval bisection to find the inverse CDF to within tol.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -51,75 +51,78 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1992-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invGcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_invGcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 
 %-Parameters
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 Dtol   = 10^-8;
 maxIt = 10000;
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<4, tol = Dtol; end
 if nargin<3, error('Insufficient arguments'), end
 
 ad = [ndims(F);ndims(h);ndims(l)];
 rd = max(ad);
-as = [  [size(F),ones(1,rd-ad(1))];...
-    [size(h),ones(1,rd-ad(2))];...
-    [size(l),ones(1,rd-ad(3))]     ];
+as = [[size(F),ones(1,rd-ad(1))];...
+      [size(h),ones(1,rd-ad(2))];...
+      [size(l),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation - Undefined and special cases
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 x = zeros(rs);
 
 %-Only defined for F in [0,1] & strictly positive h & l.
 % Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  h>0  &  l>0 );
-if any(~md(:)), x(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    x(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special cases: x=0 when F=0, x=Inf when F=1
 x(md & F==1) = Inf;
 
 %-Compute where defined & not special case
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 Q  = find( md  &  F>0  &  F<1 );
 if isempty(Q), return, end
-if xa(1), FQ=F(Q); FQ=FQ(:); else, FQ=F*ones(length(Q),1); end
-if xa(2), hQ=h(Q); hQ=hQ(:); else, hQ=h*ones(length(Q),1); end
-if xa(3), lQ=l(Q); lQ=lQ(:); else, lQ=l*ones(length(Q),1); end
+if xa(1), FQ=F(Q); FQ=FQ(:); else FQ=F*ones(length(Q),1); end
+if xa(2), hQ=h(Q); hQ=hQ(:); else hQ=h*ones(length(Q),1); end
+if xa(3), lQ=l(Q); lQ=lQ(:); else lQ=l*ones(length(Q),1); end
 %-?Q=?Q(:) stuff is to avoid discrepant orientations of vector arguments!
 
 %-Compute initial interval estimates [0,mean] & Grow to encompass F(QF);
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 a  = zeros(length(Q),1);
 b  = hQ./(2*lQ);
 QQ = 1:length(Q);
-while length(QQ)
+while ~isempty(QQ)
     b(QQ) = 2*b(QQ);
     QQ    = find(spm_Gcdf(b,hQ,lQ) < FQ);
 end
 
 %-Interval bisection
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 fa = spm_Gcdf(a,hQ,lQ)-FQ;
 fb = spm_Gcdf(b,hQ,lQ)-FQ;
 i  = 0;
 xQ = a+(b-a)/2;
 QQ = 1:length(Q);
 
-while length(QQ) &  i<maxIt
+while ~isempty(QQ) &&  i<maxIt
     fxQQ        = spm_Gcdf(xQ(QQ),hQ(QQ),lQ(QQ))-FQ(QQ);
     mQQ         = fa(QQ).*fxQQ > 0;
     a(QQ(mQQ))  = xQ(QQ(mQQ));   fa(QQ(mQQ))  = fxQQ(mQQ);
@@ -129,6 +132,8 @@
     i           = i+1;
 end
 
-if i==maxIt, warning('convergence criteria not reached - maxIt reached'), end
+if i==maxIt
+    warning('convergence criteria not reached - maxIt reached');
+end
 
 x(Q) = xQ;
diff --git a/spm_invIcdf.m b/spm_invIcdf.m
index 9be46e9..9295bd9 100644
--- a/spm_invIcdf.m
+++ b/spm_invIcdf.m
@@ -6,13 +6,13 @@
 % n - Binomial n
 % p - Binomial p [Defaults to 0.5]
 % r - ordinate
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_invIcdf returns the inverse Cumulative Distribution Function for
 % the Binomial family of distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Bin(n,p) distribution is the distribution of the number of
 % successes from n identical independent Bernoulli trials each with
 % success probability p. If random variable R is the number of
@@ -26,11 +26,11 @@
 % function.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % r is found by direct summation of the Binomial PDFs until F is exceeded.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -43,37 +43,39 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invIcdf.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_invIcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, p=0.5; end
 if nargin<2, error('Insufficient arguments'), end
 ad = [ndims(F);ndims(n);ndims(p)];
 rd = max(ad);
-as = [  [size(F),ones(1,rd-ad(1))];...
-    [size(n),ones(1,rd-ad(2))];...
-    [size(p),ones(1,rd-ad(3))]     ];
+as = [[size(F),ones(1,rd-ad(1))];...
+      [size(n),ones(1,rd-ad(2))];...
+      [size(p),ones(1,rd-ad(3))]];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 r = zeros(rs);
 
 %-Only defined for whole n, p&F in [0,1]. Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  n==floor(n)  &  n>=0  &  p>=0  &  p<=1 );
-if any(~md(:)), r(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    r(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Non-zero only where defined, & F>0
 Q  = find( md  &  F>0 );
@@ -85,7 +87,7 @@
 %-Compute by directly summing Bin PDF's for successive r & comparing with F
 tr  = 0;
 Ftr = spm_Ipdf(tr,n(Qn),p(Qp));
-while any(F(QF)>Ftr) & any(n(Qn)>tr)
+while any(F(QF)>Ftr) && any(n(Qn)>tr)
     tr      = tr+1;
     i       = find(Ftr<F(QF));
     r(Q(i)) = r(Q(i)) + 1;
diff --git a/spm_invNcdf.m b/spm_invNcdf.m
index d7c75a3..479b8a9 100644
--- a/spm_invNcdf.m
+++ b/spm_invNcdf.m
@@ -6,7 +6,7 @@
 % u - mean              [Defaults to 0]
 % v - variance  (v>0)   [Defaults to 1]
 % x - ordinates of N(u,v) at which CDF F(x)=F
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_invNcdf implements the inverse of the Cumulative Distribution
 % Function (CDF) for the Normal (Gaussian) family of distributions.
@@ -16,7 +16,7 @@
 % variance v.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF F(x) of a Normal distribution with mean u and variance v is
 % the probability that a random realisation X from this distribution
 % will be less than x. F(x)=Pr(X<=x) for X~N(u,v). The inverse CDF
@@ -28,7 +28,7 @@
 % its inverse as \Phi^{-1}(F).
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The CDF for a standard N(0,1) Normal distribution, \Phi(z), is
 % related to the error function by: (Abramowitz & Stegun, 26.2.29)
 %
@@ -39,13 +39,13 @@
 %
 % where erfinv(.) is the inverse error function.
 %
-% MatLab's implementation of the inverse error function is used for
+% MATLAB's implementation of the inverse error function is used for
 % computation of z=\Phi^{-1}(F), the corresponding standard normal
 % variate, which converted to a variate x from a N(u,v) distribution by:
 %       x = u+z*sqrt(v)
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -58,16 +58,15 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1992-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invNcdf.m 3450 2009-10-08 16:11:13Z guillaume $
-
+% $Id: spm_invNcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<3, v=1; end
 if nargin<2, u=0; end
 if nargin<1, x=[]; return, end
@@ -78,19 +77,22 @@
     [size(v),ones(1,rd-ad(3))]     ];
 rs = max(as);
 xa = prod(as,2)>1;
-if sum(xa)>1 & any(any(diff(as(xa,:)),1))
-    error('non-scalar args must match in size'), end
+if sum(xa)>1 && any(any(diff(as(xa,:)),1))
+    error('non-scalar args must match in size');
+end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 x = zeros(rs);
 
 %-Only defined for F in [0,1], & strictly positive variance v.
 % Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  ones(size(u))  &  v>0 );
-if any(~md(:)), x(~md) = NaN;
-    warning('SPM:outOfRangeNormal','Returning NaN for out of range arguments'), end
+if any(~md(:))
+    x(~md) = NaN;
+    warning('SPM:outOfRangeNormal','Returning NaN for out of range arguments');
+end
 
 %-Compute where defined
 Q  = find( md );
diff --git a/spm_invPcdf.m b/spm_invPcdf.m
index 1ddc371..be96432 100644
--- a/spm_invPcdf.m
+++ b/spm_invPcdf.m
@@ -5,13 +5,13 @@
 % F - CDF (lower tail p-value)
 % x - ordinates
 % l - Poisson mean parameter (lambda l>0) [Defaults to 1]
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_invPcdf returns the inverse Cumulative Distribution Function for
 % the Poisson family of distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Poisson Po(l) distribution is the distribution of the number of
 % events in unit time for a stationary Poisson process with mean
 % parameter lambda=1, or equivalently rate 1/l. If random variable X is
@@ -25,11 +25,11 @@
 % function.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % x is found by direct summation of the Poisson PDFs until F is exceeded.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -42,38 +42,41 @@
 %       "Numerical Recipes in C"
 %        Cambridge
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1996-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invPcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_invPcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, l=1; end
 if nargin<1, error('Insufficient arguments'), end
 
 ad = [ndims(F);ndims(l)];
 rd = max(ad);
-as = [  [size(F),ones(1,rd-ad(1))];...
-    [size(l),ones(1,rd-ad(2))];    ];
+as = [[size(F),ones(1,rd-ad(1))];...
+      [size(l),ones(1,rd-ad(2))];];
 rs = max(as);
 xa = prod(as,2)>1;
-if all(xa) & any(diff(as(xa,:)))
-    error('non-scalar args must match in size'), end
+if all(xa) && any(diff(as(xa,:)))
+    error('non-scalar args must match in size');
+end
 
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 x = zeros(rs);
 
 %-Only defined for F in [0,1] & l>0 & . Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  l>0 );
-if any(~md(:)), x(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    x(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Infinite where defined but F=1
 mi = ( F==1 ); x(md&mi) = Inf;
diff --git a/spm_invTcdf.m b/spm_invTcdf.m
index 3cb592d..970c7fb 100644
--- a/spm_invTcdf.m
+++ b/spm_invTcdf.m
@@ -5,13 +5,13 @@
 % F  - CDF (lower tail p-value)
 % v - degrees of freedom (v>0, non-integer d.f. accepted)
 % x - T-variate (Student's t has range (-Inf,Inf))
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_invTcdf implements the inverse Cumulative Distribution of Students
 % t-distributions.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Student's t-distribution with v degrees of freedom is defined for
 % positive integer v and real x. The Cumulative Distribution
 % Function (CDF) F(x) is the probability that a realisation of a
@@ -22,7 +22,7 @@
 % v, rather it will compute for any df v>0.
 %
 % Variate relationships: (Evans et al., Ch37 & 7)
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Student's t distribution with 1 degree of freedom is the Standard
 % Cauchy distribution, which has a simple closed form CDF.
 %
@@ -31,7 +31,7 @@
 % w/2 & 1/2, as described below.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Using the routine spm_invBcdf for the Beta distribution, with
 % appropriate parameters:  The CDF of the Student's t-distribution with
 % v degrees of freedom is related to the incomplete beta function by:
@@ -47,7 +47,7 @@
 % incomplete beta function.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -61,14 +61,14 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1993-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invTcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_invTcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Format arguments, note & check sizes
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 ad = [ndims(F);ndims(v)];
@@ -77,27 +77,30 @@
     [size(v),ones(1,rd-ad(2))]     ];
 rs = max(as);
 xa = prod(as,2)>1;
-if all(xa) & any(diff(as(xa,:)))
-    error('non-scalar args must match in size'), end
+if all(xa) && any(diff(as(xa,:)))
+    error('non-scalar args must match in size');
+end
 
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 %-Initialise result to zeros
 x = zeros(rs);
 
 %-Only defined for F in [0,1] & strictly positive v.
 % Return NaN if undefined.
 md = ( F>=0  &  F<=1  &  v>0 );
-if any(~md(:)), x(~md) = NaN;
-    warning('Returning NaN for out of range arguments'), end
+if any(~md(:))
+    x(~md) = NaN;
+    warning('Returning NaN for out of range arguments');
+end
 
 %-Special case: x is 0 when F=0.5, -Inf when F=0, +Inf when F=1
 x(md & F==0) = -Inf;
 x(md & F==1) = +Inf;
 
 %-Special case: Standard Cauchy distribution when v=1
-ml = ( md  &  v==1 ); if xa(1), mlF=ml; else, mlF=1; end
+ml = ( md  &  v==1 ); if xa(1), mlF=ml; else mlF=1; end
 x(ml) = tan(pi*(F(mlF)-0.5));
 
 %-Compute where defined & not special cases
diff --git a/spm_invXcdf.m b/spm_invXcdf.m
index 146f8cd..61f518a 100644
--- a/spm_invXcdf.m
+++ b/spm_invXcdf.m
@@ -11,7 +11,7 @@
 % Chi-squared distribution.
 %
 % Definition:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % The Chi-squared distribution with v degrees of freedom is defined for
 % positive integer v and x in [0,Inf). The Cumulative Distribution
 % Function (CDF) F(x) is the probability that a realisation of a
@@ -19,16 +19,17 @@
 % (See Evans et al., Ch8)
 %
 % Variate relationships: (Evans et al., Ch8 & Ch18)
-%-----------------------------------------------------------------------
-% The Chi-squared distribution with v degrees of freedom is equivalent
-% to the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
+%--------------------------------------------------------------------------
+% The Chi-squared distribution with v degrees of freedom is equivalent to
+% the Gamma distribution with scale parameter 1/2 and shape parameter v/2.
 %
 % Algorithm:
-%-----------------------------------------------------------------------
-% Using routine spm_invGcdf for Gamma distribution, with appropriate parameters.
+%--------------------------------------------------------------------------
+% Using routine spm_invGcdf for Gamma distribution, with appropriate
+% parameters.
 %
 % References:
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 % Evans M, Hastings N, Peacock B (1993)
 %       "Statistical Distributions"
 %        2nd Ed. Wiley, New York
@@ -42,16 +43,16 @@
 %        Cambridge
 %
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1993-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_invXcdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_invXcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Check enough arguments
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<2, error('Insufficient arguments'), end
 
 %-Computation
-%---------------------------------------------------------------------------
+%--------------------------------------------------------------------------
 x = spm_invGcdf(F,v/2,1/2);
diff --git a/spm_jobman.m b/spm_jobman.m
index 80f6896..1a3e7fd 100644
--- a/spm_jobman.m
+++ b/spm_jobman.m
@@ -4,51 +4,52 @@
 % It translates spm_jobman callbacks into matlabbatch callbacks and allows
 % to edit and run SPM5 style batch jobs.
 %
+% FORMAT spm_jobman('initcfg')
+% Initialise jobs configuration and set MATLAB path accordingly.
+%
+% FORMAT spm_jobman('run',job)
+% FORMAT output_list = spm_jobman('run',job)
+% Run specified job
+% job         - filename of a job (.m, .mat or .xml), or
+%               cell array of filenames, or
+%               'jobs'/'matlabbatch' variable, or
+%               cell array of 'jobs'/'matlabbatch' variables.
+% output_list - cell array containing the output arguments from each
+%               module in the job. The format and contents of these
+%               outputs is defined in the configuration of each module
+%               (.prog and .vout callbacks).
+%
 % FORMAT job_id = spm_jobman
 %        job_id = spm_jobman('interactive')
 %        job_id = spm_jobman('interactive',job)
 %        job_id = spm_jobman('interactive',job,node)
 %        job_id = spm_jobman('interactive','',node)
-% Runs the user interface in interactive mode. The job_id can be used to
-% manipulate this job in cfg_util. Note that changes to the job in cfg_util
-% will not show up in cfg_ui unless 'Update View' is called.
+% Run the user interface in interactive mode.
+% node        - indicate which part of the configuration is to be used.
+%               For example, it could be 'spm.spatial.coreg.estimate'.
+% job_id      - can be used to manipulate this job in cfg_util. Note that
+%               changes to the job in cfg_util will not show up in cfg_ui
+%               unless 'Update View' is called.
+%__________________________________________________________________________
+%
+% Programmers help:
 %
 % FORMAT output_list = spm_jobman('serial')
 %        output_list = spm_jobman('serial',job[,'',   input1,...inputN])
 %        output_list = spm_jobman('serial',job ,node[,input1,...inputN])
 %        output_list = spm_jobman('serial',''  ,node[,input1,...inputN])
-% Runs the user interface in serial mode. I job is not empty, then node
-% is silently ignored. Inputs can be a list of arguments. These are
-% passed on to the open inputs of the specified job/node. Each input should
-% be suitable to be assigned to item.val{1}. For cfg_repeat/cfg_choice
-% items, input should be a cell list of indices input{1}...input{k} into
+% Run the user interface in serial mode. If job is not empty, then node
+% is silently ignored. Inputs can be a list of arguments. These are passed
+% on to the open inputs of the specified job/node. Each input should be
+% suitable to be assigned to item.val{1}. For cfg_repeat/cfg_choice items,
+% input should be a cell list of indices input{1}...input{k} into
 % item.value. See cfg_util('filljob',...) for details.
-%
-% FORMAT output_list = spm_jobman('run',job)
-%        output_list = spm_jobman('run_nogui',job)
-% Runs a job without X11 (as long as there is no graphics output from the
-% job itself). The matlabbatch system does not need graphics output to run
-% a job.
-%
-%     node - indicates which part of the configuration is to be used.
-%            For example, it could be 'jobs.spatial.coreg'.
-%
-%     job  - can be the name of a jobfile (as a .m, .mat or a .xml), a
-%            cellstr of filenames, a 'jobs'/'matlabbatch' variable or a
-%            cell of 'jobs'/'matlabbatch' variables loaded from a jobfile.
-%
-% Output_list is a cell array and contains the output arguments from each
-% module in the job. The format and contents of these outputs is defined in
-% the configuration of each module (.prog and .vout callbacks).
-%
-% FORMAT spm_jobman('initcfg')
-% Initialise cfg_util configuration and set path accordingly.
-%
+% 
 % FORMAT jobs = spm_jobman('spm5tospm8',jobs)
-% Takes a cell list of SPM5 job structures and returns SPM8 compatible versions.
+% Take a cell list of SPM5 job structures and returns SPM8 compatible versions.
 %
 % FORMAT job = spm_jobman('spm5tospm8bulk',jobfiles)
-% Takes a cell string with SPM5 job filenames and saves them in SPM8
+% Take a cell string with SPM5 job filenames and saves them in SPM8
 % compatible format. The new job files will be MATLAB .m files. Their
 % filenames will be derived from the input filenames. To make sure they are
 % valid MATLAB script names they will be processed with
@@ -57,13 +58,10 @@
 %
 % FORMAT spm_jobman('help',node)
 %        spm_jobman('help',node,width)
-% Creates a cell array containing help information.  This is justified
+% Create a cell array containing help information.  This is justified
 % to be 'width' characters wide. e.g.
-%     h = spm_jobman('help','jobs.spatial.coreg.estimate');
-%     for i=1:numel(h),fprintf('%s\n',h{i}); end;
-%
-% not implemented: FORMAT spm_jobman('defaults')
-% Runs the interactive defaults editor.
+%     h = spm_jobman('help','spm.spatial.coreg.estimate');
+%     for i=1:numel(h), fprintf('%s\n',h{i}); end
 %
 % FORMAT [tag, job] = spm_jobman('harvest', job_id|cfg_item|cfg_struct)
 % Take the job with id job_id in cfg_util and extract what is
@@ -73,191 +71,217 @@
 % job - harvested data from the current job/cfg_item tree
 %
 % FORMAT spm_jobman('pulldown')
-% Creates a pulldown 'TASKS' menu in the Graphics window.
-%
+% Create a pulldown 'TASKS' menu in the Graphics window.
+%__________________________________________________________________________
+% 
 % not implemented: FORMAT spm_jobman('jobhelp')
-% Creates a cell array containing help information specific for a certain
+% Create a cell array containing help information specific for a certain
 % job. Help is only printed for items where job specific help is
 % present. This can be used together with spm_jobman('help') to create a
-% job specific manual. This feature is available only on MATLAB R14SP2
-% and higher.
+% job specific manual.
 %
 % not implemented: FORMAT spm_jobman('chmod')
-% Changes the modality for the TASKS pulldown.
+% Change the modality for the TASKS pulldown.
+%
+% not implemented: FORMAT spm_jobman('defaults')
+% Run the interactive defaults editor.
 %
+% not implemented: FORMAT output_list = spm_jobman('run_nogui',job)
+% Run a job without X11 (as long as there is no graphics output from the
+% job itself). The matlabbatch system does not need graphics output to run
+% a job.
+%__________________________________________________________________________
+% 
 % This code is based on earlier versions by John Ashburner, Philippe 
 % Ciuciu and Guillaume Flandin.
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 % Copyright (C) 2008 Freiburg Brain Imaging
 
 % Volkmar Glauche
-% $Id: spm_jobman.m 3916 2010-06-03 11:12:34Z guillaume $
+% $Id: spm_jobman.m 4242 2011-03-11 15:12:04Z guillaume $
 
 
-if nargin==0
+persistent isInitCfg;
+if isempty(isInitCfg) &&  ~(nargin == 1 && strcmpi(varargin{1},'initcfg'))
+    warning('spm:spm_jobman:NotInitialised',...
+        'Run spm_jobman(''initcfg''); beforehand');
+    spm_jobman('initcfg');
+end
+isInitCfg = true;
+
+if ~nargin
     h = cfg_ui;
     if nargout > 0, varargout = {h}; end
-else
-    cmd = lower(varargin{1});
-    if any(strcmp(cmd, {'serial','interactive','run','run_nogui'}))
-        if nargin > 1
-            % sort out job/node arguments for interactive, serial, run cmds
-            if nargin>=2 && ~isempty(varargin{2})
-                % do not consider node if job is given
-                if ischar(varargin{2}) || iscellstr(varargin{2})
-                    jobs = load_jobs(varargin{2});
-                elseif iscell(varargin{2})
-                    if iscell(varargin{2}{1})
-                        % assume varargin{2} is a cell of jobs
-                        jobs = varargin{2};
-                    else
-                        % assume varargin{2} is a single job
-                        jobs{1} = varargin{2};
-                    end;
-                end;
-                [mljob comp] = canonicalise_job(jobs);
-            elseif any(strcmp(cmd, {'interactive','serial'})) && nargin>=3 && isempty(varargin{2})
-                % Node spec only allowed for 'interactive', 'serial'
-                arg3       = regexprep(varargin{3},'^spmjobs\.','spm.');
-                mod_cfg_id = cfg_util('tag2mod_cfg_id',arg3);
-            else
-                error('spm:spm_jobman:WrongUI', ...
-                    'Don''t know how to handle this ''%s'' call.', lower(varargin{1}));
-            end;
-        end
-    end;
-    switch cmd
-        case 'help'
-            if (nargin < 2) || isempty(varargin{2})
-                node = 'spm';
-            else
-                node = regexprep(varargin{2},'^spmjobs\.','spm.');
-            end;
-            if nargin < 3
-                width = 60;
-            else
-                width = varargin{3};
-            end;
-            varargout{1} = cfg_util('showdocwidth', width, node);
-            
-        case 'initcfg'
-            if ~isdeployed
-                addpath(fullfile(spm('Dir'),'matlabbatch'));
-                addpath(fullfile(spm('Dir'),'config'));
-            end
-            cfg_get_defaults('cfg_util.genscript_run', @genscript_run);
-            cfg_util('initcfg'); % This must be the first call to cfg_util
-            if ~spm('cmdline')
-                f = cfg_ui('Visible','off'); % Create invisible batch ui
-                f0 = findobj(f, 'Tag','MenuFile'); % Add entries to file menu
-                f2 = uimenu(f0,'Label','Load SPM5 job', 'Callback',@load_job, ...
-                    'HandleVisibility','off', 'tag','jobs', ...
-                    'Separator','on');
-                f3 = uimenu(f0,'Label','Bulk Convert SPM5 job(s)', ...
-                    'Callback',@conv_jobs, ...
-                    'HandleVisibility','off', 'tag','jobs');
-            end
-        case 'interactive',
-            if exist('mljob', 'var')
-                cjob = cfg_util('initjob', mljob);
-            elseif exist('mod_cfg_id', 'var')
-                if isempty(mod_cfg_id)
-                    arg3 = regexprep(varargin{3},'^spmjobs\.','spm.');
-                    warning('spm:spm_jobman:NodeNotFound', ...
-                        ['Can not find executable node ''%s'' - running '...
-                        'matlabbatch without default node.'], arg3);
-                    cjob = cfg_util('initjob');
+    return;
+end
+
+cmd = lower(varargin{1});
+if strcmp(cmd,'run_nogui')
+    warning('spm:spm_jobman:NotImplemented', ...
+            'Callback ''%s'' not implemented.', cmd);
+    cmd = 'run';
+end
+if any(strcmp(cmd, {'serial','interactive','run'}))
+    if nargin > 1
+        % sort out job/node arguments for interactive, serial, run cmds
+        if nargin>=2 && ~isempty(varargin{2})
+            % do not consider node if job is given
+            if ischar(varargin{2}) || iscellstr(varargin{2})
+                jobs = load_jobs(varargin{2});
+            elseif iscell(varargin{2})
+                if iscell(varargin{2}{1})
+                    % assume varargin{2} is a cell of jobs
+                    jobs = varargin{2};
                 else
-                    cjob = cfg_util('initjob');
-                    mod_job_id = cfg_util('addtojob', cjob, mod_cfg_id);
-                    cfg_util('harvest', cjob, mod_job_id);
-                end;
-            else
+                    % assume varargin{2} is a single job
+                    jobs{1} = varargin{2};
+                end
+            end
+            mljob = canonicalise_job(jobs);
+        elseif any(strcmp(cmd, {'interactive','serial'})) && nargin>=3 && isempty(varargin{2})
+            % Node spec only allowed for 'interactive', 'serial'
+            arg3       = regexprep(varargin{3},'^spmjobs\.','spm.');
+            mod_cfg_id = cfg_util('tag2mod_cfg_id',arg3);
+        else
+            error('spm:spm_jobman:WrongUI', ...
+                'Don''t know how to handle this ''%s'' call.', lower(varargin{1}));
+        end
+    end
+end
+
+switch cmd
+    case {'initcfg'}
+        if ~isdeployed
+            addpath(fullfile(spm('Dir'),'matlabbatch'));
+            addpath(fullfile(spm('Dir'),'config'));
+        end
+        cfg_get_defaults('cfg_util.genscript_run', @genscript_run);
+        cfg_util('initcfg'); % This must be the first call to cfg_util
+        if ~spm('cmdline')
+            f = cfg_ui('Visible','off'); % Create invisible batch ui
+            f0 = findobj(f, 'Tag','MenuFile'); % Add entries to file menu
+            f2 = uimenu(f0,'Label','Load SPM5 job', 'Callback',@load_job, ...
+                'HandleVisibility','off', 'tag','jobs', ...
+                'Separator','on');
+            f3 = uimenu(f0,'Label','Bulk Convert SPM5 job(s)', ...
+                'Callback',@conv_jobs, ...
+                'HandleVisibility','off', 'tag','jobs');
+        end
+        
+    case {'interactive'}
+        if exist('mljob', 'var')
+            cjob = cfg_util('initjob', mljob);
+        elseif exist('mod_cfg_id', 'var')
+            if isempty(mod_cfg_id)
+                arg3 = regexprep(varargin{3},'^spmjobs\.','spm.');
+                warning('spm:spm_jobman:NodeNotFound', ...
+                    ['Can not find executable node ''%s'' - running '...
+                    'matlabbatch without default node.'], arg3);
                 cjob = cfg_util('initjob');
-            end;
-            cfg_ui('local_showjob', findobj(0,'tag','cfg_ui'), cjob);
-            if nargout > 0
-                varargout{1} = cjob;
-            end;
-            
-        case 'serial',
-            if exist('mljob', 'var')
-                cjob = cfg_util('initjob', mljob);
             else
                 cjob = cfg_util('initjob');
-                if nargin > 2
-                    arg3 = regexprep(varargin{3},'^spmjobs\.','spm.');
-                    [mod_cfg_id, item_mod_id] = cfg_util('tag2cfg_id', lower(arg3));
-                    cfg_util('addtojob', cjob, mod_cfg_id);
-                end;
-            end;
-            sts  = cfg_util('filljobui', cjob, @serial_ui, varargin{4:end});
-            if sts
-                cfg_util('run', cjob);
-                if nargout > 0
-                    varargout{1} = cfg_util('getalloutputs', cjob);
-                end
-            end;
-            cfg_util('deljob', cjob);
-
-        case {'run','run_nogui'}
+                mod_job_id = cfg_util('addtojob', cjob, mod_cfg_id);
+                cfg_util('harvest', cjob, mod_job_id);
+            end
+        else
+            cjob = cfg_util('initjob');
+        end
+        cfg_ui('local_showjob', findobj(0,'tag','cfg_ui'), cjob);
+        if nargout > 0
+            varargout{1} = cjob;
+        end
+        
+    case {'serial'}
+        if exist('mljob', 'var')
             cjob = cfg_util('initjob', mljob);
+        else
+            cjob = cfg_util('initjob');
+            if nargin > 2
+                arg3 = regexprep(varargin{3},'^spmjobs\.','spm.');
+                [mod_cfg_id, item_mod_id] = cfg_util('tag2cfg_id', lower(arg3));
+                cfg_util('addtojob', cjob, mod_cfg_id);
+            end
+        end
+        sts  = cfg_util('filljobui', cjob, @serial_ui, varargin{4:end});
+        if sts
             cfg_util('run', cjob);
             if nargout > 0
                 varargout{1} = cfg_util('getalloutputs', cjob);
             end
-            cfg_util('deljob', cjob);
-            
-        case {'spm5tospm8'}
-            varargout{1} = canonicalise_job(varargin{2});
-
-        case {'spm5tospm8bulk'}
-            conv_jobs(varargin{2});
-            
-        case {'defaults'},
-            warning('spm:spm_jobman:NotImplemented', 'Not yet implemented.');
-            
-        case {'pulldown'}
-            pulldown;
-
-        case {'chmod'}
-            warning('spm:spm_jobman:NotImplemented', 'Callback ''%s'' not implemented.', varargin{1});
-
-        case {'help'}
-            warning('spm:spm_jobman:NotImplemented', 'Not yet implemented.');
-
-        case {'jobhelp'}
-            warning('spm:spm_jobman:NotImplemented', 'Callback ''%s'' not implemented.', varargin{1});
-
-        case {'harvest'}
-            if nargin == 1
-                error('spm:spm_jobman:CantHarvest', ...
-                        ['Can not harvest job without job_id. Please use ' ...
-                         'spm_jobman(''harvest'', job_id).']);
-            elseif cfg_util('isjob_id', varargin{2})
-                [tag job] = cfg_util('harvest', varargin{2});
-            elseif isa(varargin{2}, 'cfg_item')
-                [tag job] = harvest(varargin{2}, varargin{2}, false, false);
-            elseif isstruct(varargin{2})
-                % try to convert into class before harvesting
-                c = cfg_struct2cfg(varargin{2});
-                [tag job] = harvest(c,c,false,false);
-            else
-                error('spm:spm_jobman:CantHarvestThis', ['Can not harvest ' ...
-                                    'this argument.']);
-            end;
-            varargout{1} = tag;
-            varargout{2} = job;
+        end
+        cfg_util('deljob', cjob);
+        
+    case {'run'}
+        cjob = cfg_util('initjob', mljob);
+        cfg_util('run', cjob);
+        if nargout > 0
+            varargout{1} = cfg_util('getalloutputs', cjob);
+        end
+        cfg_util('deljob', cjob);
+        
+    case {'spm5tospm8'}
+        varargout{1} = canonicalise_job(varargin{2});
+        
+    case {'spm5tospm8bulk'}
+        conv_jobs(varargin{2});
+        
+    case {'harvest'}
+        if nargin == 1
+            error('spm:spm_jobman:CantHarvest', ...
+                ['Can not harvest job without job_id. Please use ' ...
+                'spm_jobman(''harvest'', job_id).']);
+        elseif cfg_util('isjob_id', varargin{2})
+            [tag job] = cfg_util('harvest', varargin{2});
+        elseif isa(varargin{2}, 'cfg_item')
+            [tag job] = harvest(varargin{2}, varargin{2}, false, false);
+        elseif isstruct(varargin{2})
+            % try to convert into class before harvesting
+            c = cfg_struct2cfg(varargin{2});
+            [tag job] = harvest(c,c,false,false);
+        else
+            error('spm:spm_jobman:CantHarvestThis', ...
+                'Can not harvest this argument.');
+        end
+        varargout{1} = tag;
+        varargout{2} = job;
+        
+    case {'help'}
+        if (nargin < 2) || isempty(varargin{2})
+            node = 'spm';
+        else
+            node = regexprep(varargin{2},'^spmjobs\.','spm.');
+        end
+        if nargin < 3
+            width = 60;
+        else
+            width = varargin{3};
+        end
+        varargout{1} = cfg_util('showdocwidth', width, node);
+        
+    case {'pulldown'}
+        pulldown;
+        
+    case {'defaults'}
+        warning('spm:spm_jobman:NotImplemented', ...
+            'Callback ''%s'' not implemented.', varargin{1});
+        
+    case {'chmod'}
+        warning('spm:spm_jobman:NotImplemented', ...
+            'Callback ''%s'' not implemented.', varargin{1});
+        
+    case {'jobhelp'}
+        warning('spm:spm_jobman:NotImplemented', ...
+            'Callback ''%s'' not implemented.', varargin{1});
+        
+    otherwise
+        error(['"' varargin{1} '" - unknown option']);
+end
 
-        otherwise
-            error(['"' varargin{1} '" - unknown option']);
-    end;
-end;
-return;
-%------------------------------------------------------------------------
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function [mljob, comp] = canonicalise_job(job)
+%==========================================================================
 function [mljob, comp] = canonicalise_job(job)
 % job: a cell list of job data structures.
 % Check whether job is a SPM5 or matlabbatch job. In the first case, all
@@ -274,8 +298,8 @@
             {'temporal', 'spatial', 'stats', 'tools', 'util'}));
         if ~comp(cj)
             break;
-        end;
-    end;
+        end
+    end
     if comp(cj)
         tmp = convert_jobs(job{cj});
         for i=1:numel(tmp),
@@ -283,27 +307,28 @@
         end
     else
         mljob{cj} = job{cj};
-    end;
-end;
-%------------------------------------------------------------------------
+    end
+end
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function conv_jobs(varargin)
+%==========================================================================
 function conv_jobs(varargin)
 % Select a list of jobs, canonicalise each of it and save as a .m file
 % using gencode.
-spm('pointer','watch');
+spm('Pointer','Watch');
 if nargin == 0 || ~iscellstr(varargin{1})
     [fname sts] = spm_select([1 Inf], 'batch', 'Select job file(s)');
     fname = cellstr(fname);
-    if ~sts, return; end;
+    if ~sts, return; end
 else
     fname = varargin{1};
-end;
+end
 
 joblist = load_jobs(fname);
 for k = 1:numel(fname)
     if ~isempty(joblist{k})
-        [p n e v] = spm_fileparts(fname{k});
+        [p n] = spm_fileparts(fname{k});
         % Save new job as genvarname(*_spm8).m
         newfname = fullfile(p, sprintf('%s.m', ...
             genvarname(sprintf('%s_spm8', n))));
@@ -311,27 +336,28 @@ function conv_jobs(varargin)
         cjob = cfg_util('initjob', canonicalise_job(joblist(k)));
         cfg_util('savejob', cjob, newfname);
         cfg_util('deljob', cjob);
-    end;
-end;
-spm('pointer','arrow');
-%------------------------------------------------------------------------
+    end
+end
+spm('Pointer','Arrow');
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function load_job(varargin)
+%==========================================================================
 function load_job(varargin)
 % Select a single job file, canonicalise it and display it in GUI
 [fname sts] = spm_select([1 Inf], 'batch', 'Select job file');
-if ~sts, return; end;
+if ~sts, return; end
 
-spm('pointer','watch');
+spm('Pointer','Watch');
 joblist = load_jobs(fname);
 if ~isempty(joblist{1})
     spm_jobman('interactive',joblist{1});
-end;
-spm('pointer','arrow');
-return;
-%------------------------------------------------------------------------
+end
+spm('Pointer','Arrow');
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function newjobs = load_jobs(job)
+%==========================================================================
 function newjobs = load_jobs(job)
 % Load a list of possible job files, return a cell list of jobs. Jobs can
 % be either SPM5 (i.e. containing a 'jobs' variable) or SPM8/matlabbatch
@@ -341,12 +367,12 @@ function load_job(varargin)
     filenames = cellstr(job);
 else
     filenames = job;
-end;
+end
 newjobs = {};
 for cf = 1:numel(filenames)
     [p,nam,ext] = fileparts(filenames{cf});
     switch ext
-        case '.xml',
+        case '.xml'
             spm('Pointer','Watch');
             try
                 loadxml(filenames{cf},'jobs');
@@ -355,9 +381,9 @@ function load_job(varargin)
                     loadxml(filenames{cf},'matlabbatch');
                 catch
                     warning('spm:spm_jobman:LoadFailed','LoadXML failed: ''%s''',filenames{cf});
-                end;
-            end;
-            spm('Pointer');
+                end
+            end
+            spm('Pointer','Arrow');
         case '.mat'
             try
                 S=load(filenames{cf});
@@ -370,7 +396,7 @@ function load_job(varargin)
                 end
             catch
                 warning('spm:spm_jobman:LoadFailed','Load failed: ''%s''',filenames{cf});
-            end;
+            end
         case '.m'
             try
                 fid = fopen(filenames{cf},'rt');
@@ -379,25 +405,25 @@ function load_job(varargin)
                 eval(str);
             catch
                 warning('spm:spm_jobman:LoadFailed','Load failed: ''%s''',filenames{cf});
-            end;
+            end
             if ~(exist('jobs','var') || exist('matlabbatch','var'))
                 warning('spm:spm_jobman:JobNotFound','No SPM5/SPM8 job found in ''%s''', filenames{cf});
-            end;
+            end
         otherwise
             warning('Unknown extension: ''%s''', filenames{cf});
-    end;
+    end
     if exist('jobs','var')
         newjobs = [newjobs(:); {jobs}];
         clear jobs;
     elseif exist('matlabbatch','var')
         newjobs = [newjobs(:); {matlabbatch}];
         clear matlabbatch;
-    end;
-end;
-return;
-%------------------------------------------------------------------------
+    end
+end
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function njobs = convert_jobs(jobs)
+%==========================================================================
 function njobs = convert_jobs(jobs)
 decel    = struct('spatial',struct('realign',[],'coreg',[],'normalise',[]),...
                  'temporal',[],...
@@ -406,14 +432,14 @@ function load_job(varargin)
                  'util',[],...
                  'tools',struct('dartel',[]));
 njobs  = {};
-for i0 = 1:numel(jobs),
+for i0 = 1:numel(jobs)
     tmp0  = fieldnames(jobs{i0});
     tmp0  = tmp0{1};
-    if any(strcmp(tmp0,fieldnames(decel))),
-        for i1=1:numel(jobs{i0}.(tmp0)),
+    if any(strcmp(tmp0,fieldnames(decel)))
+        for i1=1:numel(jobs{i0}.(tmp0))
             tmp1  = fieldnames(jobs{i0}.(tmp0){i1});
             tmp1  = tmp1{1};
-            if ~isempty(decel.(tmp0)),
+            if ~isempty(decel.(tmp0))
                 if any(strcmp(tmp1,fieldnames(decel.(tmp0)))),
                     for i2=1:numel(jobs{i0}.(tmp0){i1}.(tmp1)),
                         njobs{end+1} = struct(tmp0,struct(tmp1,jobs{i0}.(tmp0){i1}.(tmp1){i2}));
@@ -429,9 +455,10 @@ function load_job(varargin)
         njobs{end+1} = jobs{i0};
     end
 end
-%------------------------------------------------------------------------
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function pulldown
+%==========================================================================
 function pulldown
 fg = spm_figure('findwin','Graphics');
 if isempty(fg), return; end;
@@ -448,17 +475,19 @@ function load_job(varargin)
             'HandleVisibility','off', 'tag','jobs');
 cfg_ui('local_setmenu', f5, cfg_util('tag2cfg_id', 'spm'), ...
        @local_init_serial, false);
-%------------------------------------------------------------------------
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function local_init_interactive(varargin)
+%==========================================================================
 function local_init_interactive(varargin)
 cjob = cfg_util('initjob');
 mod_cfg_id = get(gcbo,'userdata');
 cfg_util('addtojob', cjob, mod_cfg_id);
 cfg_ui('local_showjob', findobj(0,'tag','cfg_ui'), cjob);
-%------------------------------------------------------------------------
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function local_init_serial(varargin)
+%==========================================================================
 function local_init_serial(varargin)
 mod_cfg_id = get(gcbo,'userdata');
 cjob = cfg_util('initjob');
@@ -466,39 +495,40 @@ function local_init_serial(varargin)
 sts = cfg_util('filljobui', cjob, @serial_ui);
 if sts
     cfg_util('run', cjob);
-end;
+end
 cfg_util('deljob', cjob);
-%------------------------------------------------------------------------
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function [val sts] = serial_ui(item)
+%==========================================================================
 function [val sts] = serial_ui(item)
 % wrapper function to translate cfg_util('filljobui'... input requests into
 % spm_input/cfg_select calls.
 sts = true;
-switch class(item),
-    case 'cfg_choice',
+switch class(item)
+    case 'cfg_choice'
         labels = cell(size(item.values));
         values = cell(size(item.values));
         for k = 1:numel(item.values)
             labels{k} = item.values{k}.name;
             values{k} = k;
-        end;
+        end
         val = spm_input(item.name, 1, 'm', labels, values);
-    case 'cfg_menu',
+    case 'cfg_menu'
         val = spm_input(item.name, 1, 'm', item.labels, item.values);
         val = val{1};
-    case 'cfg_repeat',
+    case 'cfg_repeat'
         labels = cell(size(item.values));
         values = cell(size(item.values));
         for k = 1:numel(item.values)
             labels{k} = item.values{k}.name;
             values{k} = k;
-        end;
+        end
         % enter at least item.num(1) values
         for k = 1:item.num(1)
             val(k) = spm_input(sprintf('%s(%d)', item.name, k), 1, 'm', ...
                                labels, values);
-        end;
+        end
         % enter more (up to varargin{3}(2) values
         labels = {labels{:} 'Done'};
         % values is a cell list of natural numbers, use -1 for Done
@@ -510,12 +540,12 @@ function local_init_serial(varargin)
                 break;
             else
                 val(end+1) = val1;
-            end;
-        end;
-    case 'cfg_entry',
+            end
+        end
+    case 'cfg_entry'
         val = spm_input(item.name, 1, item.strtype, '', item.num, ...
                         item.extras);
-    case 'cfg_files',
+    case 'cfg_files'
         [t,sts] = cfg_getfile(item.num, item.filter, item.name, '', ...
                               item.dir, item.ufilter);
         if sts
@@ -523,11 +553,12 @@ function local_init_serial(varargin)
         else
             val = {};
             error('File selector was closed.');
-        end;
-end;
-%------------------------------------------------------------------------
+        end
+end
 
-%------------------------------------------------------------------------
+%==========================================================================
+% function [code cont] = genscript_run
+%==========================================================================
 function [code cont] = genscript_run
 % Return code snippet to initialise SPM defaults and run a job generated by
 % cfg_util('genscript',...) through spm_jobman.
diff --git a/spm_load_priors.m b/spm_load_priors.m
index caf5e0a..31e6ca8 100644
--- a/spm_load_priors.m
+++ b/spm_load_priors.m
@@ -3,21 +3,21 @@
 % FORMAT b0 = spm_load_priors(B)
 % B  - structures of image volume information (or filenames)
 % b0 - a cell array of tissue probabilities
-%____________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 2005-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_load_priors.m 1902 2008-07-09 19:00:37Z john $
+% $Id: spm_load_priors.m 4199 2011-02-10 20:07:17Z guillaume $
 
 
 % deg = 3;
-lm  = 0;
-if ~isstruct(B), B  = spm_vol(B); end;
+lm = 0;
+if ~isstruct(B), B  = spm_vol(B); end
 Kb = length(B);
 b0 = cell(Kb,1);
-for k1=1:(Kb),
+for k1=1:(Kb)
     b0{k1} = zeros(B(1).dim(1:3));
-end;
+end
 
 spm_progress_bar('Init',B(1).dim(3),'Loading priors','Planes loaded');
 for i=1:B(1).dim(3)
@@ -27,22 +27,20 @@
         tmp           = spm_slice_vol(B(k1),M,B(1).dim(1:2),0)*(1-lm*2)+lm;
         b0{k1}(:,:,i) = max(min(tmp,1),0);
         s             = s + tmp;
-    end;
+    end
     t = s>1;
-    if any(any(t)),
+    if any(any(t))
         for k1=1:Kb
             tmp           = b0{k1}(:,:,i);
             tmp(t)        = tmp(t)./s(t);
             b0{k1}(:,:,i) = tmp;
-        end;
-    end;
+        end
+    end
     s(t) = 1;
     b0{Kb+1}(:,:,i) = max(min(1-s,1),0);
     spm_progress_bar('Set',i);
-end;
-%for k1=1:Kb+1,
+end
+%for k1=1:Kb+1
 %    b0{k1} = spm_bsplinc(log(b0{k1}),[deg deg deg  0 0 0]);
-%end;
+%end
 spm_progress_bar('Clear');
-return;
-
diff --git a/spm_log_evidence.m b/spm_log_evidence.m
new file mode 100644
index 0000000..d2d6781
--- /dev/null
+++ b/spm_log_evidence.m
@@ -0,0 +1,112 @@
+function [F,sE,sC] = spm_log_evidence(varargin)
+% Return the log-evidence of a reduced model (under Laplace approximation)
+% FORMAT [F,sE,sC] = spm_log_evidence(qE,qC,pE,pC,rE,rC)
+% FORMAT [F,sE,sC] = spm_log_evidence(qE,qC,pE,pC,priorfun,varargin)
+% FORMAT [F,sE,sC] = spm_log_evidence(qE,qC,pE,pC)
+%
+% qE,qC    - posterior expectation and covariance of full model
+% pE,pC    - prior expectation and covariance of full model
+% rE,rC    - prior expectation and covariance of reduced model
+% or 
+% priorfun - inline function that returns prior moments
+%            {rE rC} = priorfun(varargin{:})
+%
+% or (if omitted) rE = 0 and rC = 0;
+%
+% F        - reduced log-evidence: ln p(y|reduced model) - ln p(y|full model)
+% [sE,sC]  - posterior expectation and covariance of reduced model
+%
+%--------------------------------------------------------------------------
+% This routine assumes the reduced model is nested within a full model and
+% that the posteriors (and priors) are Gaussian. Nested here means that the
+% prior precision of the reduced model, minus the prior precision of the
+% full model is positive definite. We additionally assume that the prior
+% means are unchanged. The two input argument formats are for use with
+% spm_argmax.
+%__________________________________________________________________________
+% Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_log_evidence.m 4281 2011-03-31 19:49:57Z karl $
+ 
+% Compute reduced log-evidence
+%==========================================================================
+ 
+% check to see if priors are specified by a function
+%--------------------------------------------------------------------------
+qE = varargin{1};
+qC = varargin{2};
+pE = varargin{3};
+pC = varargin{4};
+try
+    priors = varargin{5}(varargin{6:end});
+    rE     = priors{1};
+    rC     = priors{2};
+catch
+    try
+        rE = varargin{5};
+        rC = varargin{6};
+    catch
+        n  = size(qC,1);
+        rE = sparse(n,1);
+        rC = sparse(n,n);
+    end
+end
+ 
+% reduced subspace 
+%--------------------------------------------------------------------------
+qE  = spm_vec(qE);
+pE  = spm_vec(pE);
+rE  = spm_vec(rE);
+ 
+if nargout < 2
+    dE  = pE - rE;
+    dC  = pC - rC;
+    k   = find(dE | any(dC,2));
+    if ~isempty(k)
+        qE  = qE(k);
+        pE  = pE(k);
+        rE  = rE(k);
+        qC  = qC(k,k);
+        pC  = pC(k,k);
+        rC  = rC(k,k);
+    else
+        
+        % the reduced and full models are the same
+        %------------------------------------------------------------------
+        F   = 0;
+        return
+    end
+end
+
+% fix tolerance for matrix inversions
+%--------------------------------------------------------------------------
+TOL   = exp(-16);
+
+% remove fixed parameters under full model
+%--------------------------------------------------------------------------
+i     = find(diag(pC));
+
+% preliminaries
+%--------------------------------------------------------------------------
+qP    = spm_inv(qC(i,i),TOL);
+pP    = spm_inv(pC(i,i),TOL);
+rP    = spm_inv(rC(i,i),TOL);
+sP    = qP + rP - pP;
+sC    = spm_inv(sP,TOL);
+sE    = qP*qE(i) + rP*rE(i) - pP*pE(i);
+
+% log-evidence
+%--------------------------------------------------------------------------
+F     = spm_logdet(rP*qP*sC*pC(i,i)) ...
+      - (qE(i)'*qP*qE(i) + rE(i)'*rP*rE(i) - pE(i)'*pP*pE(i) - sE'*sC*sE);
+F     = F/2;
+    
+% restore full conditional density
+%--------------------------------------------------------------------------
+if nargout > 1
+    rE(i)   = sC*sE;
+    rC(i,i) = sC;
+    sE      = spm_unvec(rE,varargin{1});
+    sC      = rC;
+end
diff --git a/spm_logdet.m b/spm_logdet.m
index aece139..729ae05 100644
--- a/spm_logdet.m
+++ b/spm_logdet.m
@@ -9,13 +9,13 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_logdet.m 3707 2010-02-02 19:01:59Z guillaume $
+% $Id: spm_logdet.m 4278 2011-03-31 11:48:00Z karl $
 
 sw    = warning('off','MATLAB:log:logOfZero');
 
 % assume diagonal form
 %--------------------------------------------------------------------------
-TOL   = 1e-16;                                        % c.f. n*max(s)*eps
+TOL   = 1e-16;
 n     = length(C);
 s     = diag(C);
 i     = find(s > TOL & s < 1/TOL);
@@ -31,11 +31,11 @@
       H = H + log(det(C/a));           
 end
 
-% invoke svd is rank deficient
+% invoke svd if rank deficient
 %--------------------------------------------------------------------------
 if ~isreal(H) || isinf(H)
     s  = svd(full(C));
     H  = sum(log(s(s > TOL & s < 1/TOL)));
 end
-
+H     = full(H);
 warning(sw)
diff --git a/spm_lotka_volterra.m b/spm_lotka_volterra.m
index 83be306..3a1f036 100644
--- a/spm_lotka_volterra.m
+++ b/spm_lotka_volterra.m
@@ -1,11 +1,12 @@
 function [f] = spm_lotka_volterra(x,v,P)
 % equations of motion for Lotka-Volterra dynamics
 % FORMAT [f] = spm_lotka_volterra(x,v,P)
+% FORMAT [f] = spm_lotka_volterra(n)
 %
-% x   - hidden states
-% v   - exogenous inputs
-% P.f - lateral connectivity
-% P.k - rate [default 1]
+% [x.]x - hidden states
+% [x.]v - exogenous inputs
+% P.f   - lateral connectivity
+% P.k   - rate [default 1]
 %
 % returns f = dx/dt = P.f*S(x) - x/8 + 1;
 %              S(x) = 1./(1 + exp(-x))
@@ -17,32 +18,48 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_lotka_volterra.m 3113 2009-05-11 15:25:13Z karl $
-
+% $Id: spm_lotka_volterra.m 4146 2010-12-23 21:01:39Z karl $
 
 
 % intialise
 %==========================================================================
-try, P.k; catch, P.k = 1; end
-try, P.l; catch, P.l = 1; end
+try, k = P.k; catch, k = 1; end
+try, l = P.l; catch, l = 1; end
+
+% check for hidden cause
+%--------------------------------------------------------------------------
+if isempty(v), v = 1;       end
+
+% check for parameters of succession
+%--------------------------------------------------------------------------
+try
+    P.f;
+catch
+    n   = length(x);
+    P.f = spm_speye(n,n,-1) - spm_speye(n,n,1); P.f(n,1) = -1; P.f(1,n) = 1;
+    P.f = v*P.f + speye(n,n) - 1;
+end
 
+
+% flow
+%==========================================================================
 try
     
-    % SHC states 
+    % SHC states
     %----------------------------------------------------------------------
     f  = P.f*(1./(1 + exp(-x))) - x/8 + 1;
-    f  = P.k*f;
-
+    f  = k*f;
+    
 catch
-
-    % SHC states and flow to point attractors in P.g
+    
+    % SHC states (x.x and x.v) and flow to point attractors in P.g
     %----------------------------------------------------------------------
     x.e  = exp(x.x);
     f.x  = P.f*(1./(1 + exp(-x.x))) - x.x/8 + 1;
     f.v  = P.g*x.e - x.v*sum(x.e);
-
-    f.x  = P.k*f.x;
-    f.v  = P.l*f.v;
-
+    
+    f.x  = k*f.x;
+    f.v  = l*f.v;
+    
 end
 
diff --git a/spm_maff.m b/spm_maff.m
index d24a140..c41f1c3 100644
--- a/spm_maff.m
+++ b/spm_maff.m
@@ -1,4 +1,4 @@
-function [M,h] = spm_maff(varargin)
+function M = spm_maff(varargin)
 % Affine registration to MNI space using mutual information
 % FORMAT M = spm_maff(P,samp,x,b0,MF,M,regtyp,ff)
 % P       - filename or structure handle of image
@@ -21,7 +21,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_maff.m 1265 2008-03-28 11:45:04Z john $
+% $Id: spm_maff.m 4178 2011-01-27 15:12:53Z guillaume $
 
 [buf,MG] = loadbuf(varargin{1:2});
 M        = affreg(buf, MG, varargin{2:end});
@@ -68,7 +68,9 @@
 x1 = x{1};
 x2 = x{2};
 x3 = x{3};
-[mu,isig] = priors(regtyp);
+[mu,isig] = spm_affine_priors(regtyp);
+mu   = [zeros(6,1) ; mu];
+isig = [zeros(6,12) ; zeros(6,6) isig];
 isig      =  isig*ff;
 Alpha0    =  isig;
 Beta0     = -isig*mu;
@@ -78,7 +80,7 @@
 ll   = -Inf;
 nsmp = sum(cat(1,buf.nm));
 pr   = struct('b',[],'db1',[],'db2',[],'db3',[]);
-spm_chi2_plot('Init','Registering','Log-likelihood','Iteration');
+spm_plot_convergence('Init','Registering','Log-likelihood','Iteration');
 
 for iter=1:200
     penalty = (sol1-mu)'*isig*(sol1-mu);
@@ -105,7 +107,7 @@
     h1    = h1/ssh;
     h2    = log2(h1./(sum(h1,2)*sum(h1,1)));
     ll1   = sum(sum(h0.*h2))/ssh - penalty/ssh;
-    spm_chi2_plot('Set',ll1);
+    spm_plot_convergence('Set',ll1);
     if ll1-ll<1e-5, break; end;
     ll    = ll1;
     sol   = sol1;
@@ -148,7 +150,7 @@
     sol1  = (Alpha+Alpha0)\(Alpha*sol - Beta - Beta0);
 end;
 
-spm_chi2_plot('Clear');
+spm_plot_convergence('Clear');
 M = P2M(sol);
 return;
 %_______________________________________________________________________
@@ -162,7 +164,11 @@
 
 lV = logm(V);
 lR = -logm(R);
-if sum(sum(imag(lR).^2))>1e-6, error('Rotations by pi are still a problem.'); end;
+if sum(sum(imag(lR).^2))>1e-6
+    error('Rotations by pi are still a problem.');
+else
+    lR = real(lR);
+end
 P       = zeros(12,1);
 P(1:3)  = M(1:3,4);
 P(4:6)  = lR([2 3 6]);
@@ -211,92 +217,6 @@
 return;
 %_______________________________________________________________________
 %_______________________________________________________________________
-function [mu,isig] = priors(typ)
-% The parameters for this distribution were derived empirically from 227
-% scans, that were matched to the ICBM space.
-
-%% Values can be derived by...
-%sn = spm_select(Inf,'.*seg_inv_sn.mat$');
-%X  = zeros(size(sn,1),12);
-%for i=1:size(sn,1),
-%    p  = load(deblank(sn(i,:)));
-%    M  = p.VF(1).mat*p.Affine/p.VG(1).mat;
-%    J  = M(1:3,1:3);
-%    V  = sqrtm(J*J');
-%    R  = V\J;
-%    lV      =  logm(V);
-%    lR      = -logm(R);
-%    P       = zeros(12,1);
-%    P(1:3)  = M(1:3,4);
-%    P(4:6)  = lR([2 3 6]);
-%    P(7:12) = lV([1 2 3 5 6 9]);
-%    X(i,:)  = P';
-%end;
-%mu   = mean(X(:,7:12));
-%XR   = X(:,7:12) - repmat(mu,[size(X,1),1]);
-%isig = inv(XR'*XR/(size(X,1)-1))
-
-
-mu   = zeros(6,1);
-isig = zeros(6);
-switch deblank(lower(typ))
-
-case 'mni', % For registering with MNI templates...
-        mu   = [0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
-        isig = 1e4 * [
-            0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
-           -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
-           -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
-           -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
-           -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
-           -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
-
-case 'imni', % For registering with MNI templates...
-        mu   = -[0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
-        isig = 1e4 * [
-            0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
-           -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
-           -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
-           -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
-           -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
-           -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
-
-case 'rigid', % Constrained to be almost rigid...
-        mu   = zeros(6,1);
-        isig = eye(6)*1e8;
-
-case 'subj', % For inter-subject registration...
-        mu   = zeros(6,1);
-        isig = 1e3 * [
-            0.8876    0.0784    0.0784   -0.1749    0.0784   -0.1749
-            0.0784    5.3894    0.2655    0.0784    0.2655    0.0784
-            0.0784    0.2655    5.3894    0.0784    0.2655    0.0784
-           -0.1749    0.0784    0.0784    0.8876    0.0784   -0.1749
-            0.0784    0.2655    0.2655    0.0784    5.3894    0.0784
-           -0.1749    0.0784    0.0784   -0.1749    0.0784    0.8876];
-
-case 'eastern', % For East Asian brains to MNI...
-    mu   = [0.0719   -0.0040   -0.0032    0.1416    0.0601    0.2578]';
-    isig = 1e4 * [
-        0.0757    0.0220   -0.0224   -0.0049    0.0304   -0.0327
-        0.0220    0.3125   -0.1555    0.0280   -0.0012   -0.0284
-       -0.0224   -0.1555    1.9727    0.0196   -0.0019    0.0122
-       -0.0049    0.0280    0.0196    0.0576   -0.0282   -0.0200
-        0.0304   -0.0012   -0.0019   -0.0282    0.2128   -0.0275
-       -0.0327   -0.0284    0.0122   -0.0200   -0.0275    0.0511];
-
-case 'none', % No regularisation...
-        mu   = zeros(6,1);
-        isig = zeros(6);
-
-otherwise
-        error(['"' typ '" not recognised as type of regularisation.']);
-end;
-mu   = [zeros(6,1) ; mu];
-isig = [zeros(6,12) ; zeros(6,6) isig];
-return;
-%_______________________________________________________________________
-%_______________________________________________________________________
 function [h0,d1] = reg_unused(M)
 % Try to analytically compute the first and second derivatives of a
 % penalty function w.r.t. changes in parameters.  It works for first
diff --git a/spm_matlab_version_chk.m b/spm_matlab_version_chk.m
index cdfc952..483abf4 100644
--- a/spm_matlab_version_chk.m
+++ b/spm_matlab_version_chk.m
@@ -1,186 +1,36 @@
 function [status, fieldsUsed] = spm_matlab_version_chk(chk,tbx)
-% Check a version number against a Toolbox version.
-%    FORMAT [STATUS FIELDSUSED] = SPM_MATLAB_VERSION(CHK,TBX)
-%
-%    This function checks if a user supplied version number is less than, 
-%    equal to or greater than the version number of specified toolbox. If no
-%    toolbox is specified the function checks the version of Matlab. User
-%    defined toolboxes can be checked but the Contents.m file must conform
-%    to the specification defined in ver.m
-%
-%    This function assumes that the version number is really a text string
-%    with fields major.minor.revision.build. Other formats are not supported.
-%    Checking is done to the level specified by the input version. Thus an
-%    input of '7' will be rated as the same version as 7.1, but 7.0 would be
-%    rated as earlier than 7.1.
-%
-%    INPUTS
-%       chk         Version number to be checked [string].
-%       tbx         Name of toolbox to check. Defaults to 'Matlab'.
-%
-%    OUPUTS
-%       This is based on the fields the user supplies in the input.
-%
-%       status      Defines the outcome of the comparison
-%            -1       Toolbox version is earlier than the user supplied version.
-%             0       Toolbox and user versions are the same
-%             1       Toolbox version is later than the user supplied version.
-%                   Think of it this way, the sign of status is determined by
-%                   MATLAB_TOOLBOX_VERSION - USER_VERSION (i.e., THE VERSION YOU
-%                   INPUT).
-%       fieldsUsed  Returns the version fields that were actually compared.
-%
-%    EXAMPLES
-%       If the Matlab version is 7.1.0.83, and the user supplied version is '7':
-%       status = spm_matlab_version_chk('7');
-%       returns status == 0   : major revision numbers are the same.
-%
-%       If the Matlab version is 7.1.0.0, and the user supplied version is '7.1':
-%       status = spm_matlab_version_chk('7');
-%       returns status == 0   : major and minor revision numbers are the same.
-%
-%       If the Matlab version is 7.1.0.83, and the user supplied version is '7.2':
-%       status = spm_matlab_version_chk('7.2');
-%       returns status == -1   : major + minor revision is earlier for Matlab.
-%
-%       If the Matlab version is 6.5.1, and the user supplied version is '6.5.0'.
-%       status = spm_matlab_version_chk('6.5.0');
-%       returns status == 1     : major + minor + release revision is later
-%                                 for matlab
-%       The statement ( spm_matlab_version_chk('6.5.0') > 0 ) is true for
-%       all Matlab Toolbox versions after 6.5.0.
+% Check a version number against a Toolbox version
+% FORMAT [status, fieldsUsed] = spm_matlab_version_chk(chk,tbx)
+% chk        - Version number to be checked {string}
+% tbx        - Name of toolbox to check [Default: 'MATLAB']
+%
+% status     - Defines the outcome of the comparison
+%              -1: Toolbox version is earlier than the user supplied version
+%               0: Toolbox and user versions are the same
+%               1: Toolbox version is later than the user supplied version
+%                  Think of it this way, the sign of status is determined
+%                  by MATLAB_TOOLBOX_VERSION - USER_VERSION (i.e., THE 
+%                  VERSION YOU INPUT).
+% fieldsUsed - deprecated [Returns {}]
+%__________________________________________________________________________
 %
-%    See also  VERSION, VER.
-
+% This function is deprecated, use SPM_CHECK_VERSION instead.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Darren Gitelman
-% $Id: spm_matlab_version_chk.m 3934 2010-06-17 14:58:25Z guillaume $
-
-% output variable
-%--------------------------------------------------------------------------
-status = [];
-fieldsUsed = [];
-
-%  Check inputs
-%--------------------------------------------------------------------------
-% If no inputs then error; if 1 then set tbx to Matlab
-%-------------------------------------------------------
-switch nargin
-    case 0
-        error('Please provide a version number to be checked.');
-    case 1
-        tbx = 'Matlab';
-end
-
-% If a number is supplied then convert to text
-%-------------------------------------------------------
-if isnumeric(chk)
-    chk = num2str(chk);
-end
-
-% If too many fields in input then error
-%-------------------------------------------------------
-if numel(strfind(chk,'.')) > 3
-    error('Input string has too many fields. Only major.minor.release.build fields are supported.');
-end
-
-% parse the input string into a cell array of strings
-%--------------------------------------------------------------------------
-vCHK = textscan(chk,'%s %s %s %s','delimiter','.');
-
-% find empty cells. These will be set to 0 in both cell arrays so that the
-% same revision fields are compared
-%--------------------------------------------------------------------------
-emptyCHK = cellfun('isempty',vCHK);
-
-% Get the requested toolbox version
-%--------------------------------------------------------------------------
-if strcmpi(tbx,'Matlab')
-    % Matlab is a special case. The ver command does not report the
-    % entire version string.
-    [ tbxVer tmp ] = strtok(version);
-else
-    tbxStruct = ver(tbx);
-    if numel(tbxStruct) >1
-        error('Too many toolboxes found for given toolbox name.')
-    end
-    tbxVer = tbxStruct.Version;
-end
-
-% parse the requested toolbox version string into a cell array of strings.
-%--------------------------------------------------------------------------
-vTBX = textscan(tbxVer,'%s %s %s %s','delimiter','.');
-
-% find empty cells. These will be removed from the cell array so that the
-% same revision fields are compared
-%--------------------------------------------------------------------------
-emptyTBX = cellfun('isempty',vTBX);
-
-
-% combine empty cell indices. notEmptyCells tell us which version
-% fields are compared.
-%--------------------------------------------------------------------------
-emptyCellLogical = emptyCHK | emptyTBX;
-emptyCells       = find(emptyCellLogical);
-notEmptyCells    = find(~emptyCellLogical);
-
-% and remove them. 
-%--------------------------------------------------------------------------
- vCHK(emptyCells) = [];
- vTBX(emptyCells) = [];
- 
- 
-% final version fields to compare are converted to numbers
-%--------------------------------------------------------------------------
-vCHKMat = str2num(strvcat([vCHK{:}]))';
-vTBXMat = str2num(strvcat([vTBX{:}]))';
-
-
-% compare versions
-%--------------------------------------------------------------------------
-% This array will be used to decide which version is the later one. The
-% differences between versions are computed on each element of the version
-% number: major.minor.release.build and then multiplied by the base2array
-% and summed. This is equivalent to logically combining the base 2 place
-% values and ensures that the fields are evaluated by their significance.
-% The sign of the summed values gives the result. Positive values mean the
-% toolbox version is later, negative values mean the user supplied version,
-% is later, and 0 means they are equal.
-%--------------------------------------------------------------------------
-base2array             = [8 4 2 1];
-base2array(emptyCells) = [];
-
-% take the difference between version elements.
-%--------------------------------------------------------------------------
-vDiff = vTBXMat - vCHKMat;
+% $Id: spm_matlab_version_chk.m 4145 2010-12-23 15:18:30Z guillaume $
 
-% get the sign of the differences
-%--------------------------------------------------------------------------
-signDiff = sign(vDiff);
+%persistent runonce
+%if isempty(runonce)
+%    warning(['spm_matlab_version_check is deprecated. ',...
+%        'Use spm_check_version instead.']);
+%    runonce = 1;
+%end
 
-% multiply by the base2array
-%--------------------------------------------------------------------------
-vVal = signDiff.*base2array;
+if nargin < 1, error('Please provide a version number to be checked.'); end
+if nargin < 2, tbx = 'MATLAB'; end
 
-% If the sign of the sum is positive then toolbox version is later; if 0
-% they are the same, and if negative the user supplied version is later.
-% Think of it this way: MATLAB_TOOLBOX_VERSION - USER_VERSION.
-% Remember the comparison is made with respect to the FIELDS SUPPLED IN THE
-% INPUT
-%--------------------------------------------------------------------------
-switch sign(sum(vVal))
-    case -1     % MATLB_TOOLBOX_VERSION is EARLIER than USER_VERSION
-        status = -1;
-    case 0      % MATLB_TOOLBOX_VERSION = EQUALS = USER_VERSION
-        status =  0;
-    case 1      % MATLB_TOOLBOX_VERSION is LATER than USER_VERSION
-        status =  1;
-end
+status = spm_check_version(tbx,chk);
 
-% also return the fields that were compared.
-%--------------------------------------------------------------------------
-fieldsUsed             = {'major','minor','revision','build'}';
-fieldsUsed(emptyCells) = [];
-return
+if nargout > 1, fieldsUsed = {}; end
diff --git a/spm_mesh_adjacency.m b/spm_mesh_adjacency.m
index a2541a4..ea2d9f3 100644
--- a/spm_mesh_adjacency.m
+++ b/spm_mesh_adjacency.m
@@ -1,21 +1,25 @@
 function A = spm_mesh_adjacency(F)
 % Compute the adjacency matrix of a triangle mesh
 % FORMAT A = spm_mesh_adjacency(F)
-% F        - a [nx3] faces array or a patch structure
+% F        - a [fx3] faces array or a patch structure
 % 
-% A        - adjacency matrix as a sparse [nxn] array
+% A        - adjacency matrix as a sparse [vxv] array
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_mesh_adjacency.m 3135 2009-05-19 14:49:42Z guillaume $
+% $Id: spm_mesh_adjacency.m 4035 2010-08-05 18:54:32Z guillaume $
 
-if ~isnumeric(F)
-    F = F.faces;
-end
-
-F = double(F);
-
-A = sparse([F(:,1); F(:,1); F(:,2); F(:,2); F(:,3); F(:,3)], ...
+if ~isnumeric(F) && isfield(F,'vertices')
+    N = size(F.vertices,1);
+    F = double(F.faces);
+    A = sparse([F(:,1); F(:,1); F(:,2); F(:,2); F(:,3); F(:,3)], ...
+           [F(:,2); F(:,3); F(:,1); F(:,3); F(:,1); F(:,2)], 1, N, N);
+else
+    if isstruct(F), F = F.faces; end
+    F = double(F);
+    A = sparse([F(:,1); F(:,1); F(:,2); F(:,2); F(:,3); F(:,3)], ...
            [F(:,2); F(:,3); F(:,1); F(:,3); F(:,1); F(:,2)], 1);
-A = double(A > 0);
\ No newline at end of file
+end
+       
+A = double(A > 0);
diff --git a/spm_mesh_area.m b/spm_mesh_area.m
new file mode 100644
index 0000000..047dd83
--- /dev/null
+++ b/spm_mesh_area.m
@@ -0,0 +1,34 @@
+function A = spm_mesh_area(M,PF)
+% Compute the surface area of a triangle mesh
+% M        - patch structure: vertices and faces must be mx3 and nx3 arrays
+%            or 3xm array of edge distances
+% PF       - logical indicating whether to return surface area as a whole
+%            or per face [default: false]
+%
+% A        - surface area
+%__________________________________________________________________________
+%
+% Computed using numerically stable version of Heron's formula:
+% See http://en.wikipedia.org/wiki/Heron%27s_formula
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_area.m 4139 2010-12-15 18:31:49Z guillaume $
+
+if isnumeric(M)
+    A = M;
+else
+    A = M.vertices(M.faces',:);
+    A = permute(reshape(A',3,3,[]),[2 1 3]);
+    A = squeeze(sqrt(sum((A([1 2 3],:,:) - A([2 3 1],:,:)).^2,2)));
+end
+A = sort(A,1,'descend');
+A = 1/4 * sqrt(( A(1,:) + ( A(2,:) + A(3,:) ) ) .* ...
+               ( A(3,:) - ( A(1,:) - A(2,:) ) ) .* ...
+               ( A(3,:) + ( A(1,:) - A(2,:) ) ) .* ...
+               ( A(1,:) + ( A(2,:) - A(3,:) ) ));
+
+if nargin < 2 || ~PF
+    A = sum(A);
+end
diff --git a/spm_mesh_clusters.m b/spm_mesh_clusters.m
index 105b494..e2a6ccd 100644
--- a/spm_mesh_clusters.m
+++ b/spm_mesh_clusters.m
@@ -10,37 +10,29 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_mesh_clusters.m 4003 2010-07-19 18:22:38Z guillaume $
+% $Id: spm_mesh_clusters.m 4035 2010-08-05 18:54:32Z guillaume $
 
 
 %-Input parameters
 %--------------------------------------------------------------------------
-if isnumeric(M)
-    F   = M;
-else
-    F   = M.faces;
+if ~islogical(T)
+    T   = ~isnan(T);
 end
 
 %-Compute a reduced mesh corresponding to the data
 %--------------------------------------------------------------------------
-B       = NaN(size(T));
-if islogical(T)
-    i   = find(T);
-else
-    i   = find(~isnan(T));
-end
-B(i)    = 1:length(i);
-F       = B(F);
-F(any(isnan(F),2),:) = [];
+F       = spm_mesh_split(M, T);
 
 %-Label connected components
 %--------------------------------------------------------------------------
+% will it find two connected vertices that do not form a triangle?
+% -> maybe use spm_mesh_neighbours
 [CC,N]  = spm_mesh_label(F, 'vertices');
 C       = NaN(numel(T),1);
-C(i)    = CC;
+C(T)    = CC;
 
 %-Sort connected component labels according to their size
 %--------------------------------------------------------------------------
 [N,ni]  = sort(N(:), 1, 'descend');
 [ni,ni] = sort(ni);
-C(i)    = ni(C(i));
+C(T)    = ni(C(T));
diff --git a/spm_mesh_distmtx.m b/spm_mesh_distmtx.m
index 35d6cbb..5683506 100644
--- a/spm_mesh_distmtx.m
+++ b/spm_mesh_distmtx.m
@@ -11,7 +11,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_mesh_distmtx.m 3409 2009-09-18 15:40:25Z guillaume $
+% $Id: spm_mesh_distmtx.m 4035 2010-08-05 18:54:32Z guillaume $
 
 if nargin < 2, order = 1; end
 
@@ -26,7 +26,12 @@
 
 %-First order distance matrix
 %--------------------------------------------------------------------------
-M.faces = double(M.faces);
+if isstruct(M) && ~isa(M.faces,'double')
+    M.faces = double(M.faces);
+elseif isa(M,'gifti')
+    M = export(M,'patch');
+    M.faces = double(M.faces);
+end
 d = M.vertices(M.faces(:,[1 2 3]),:) - M.vertices(M.faces(:,[2 3 1]),:);
 
 D = sparse([M.faces(:,1); M.faces(:,2); M.faces(:,3)], ...
diff --git a/spm_mesh_edges.m b/spm_mesh_edges.m
new file mode 100644
index 0000000..abb8e5e
--- /dev/null
+++ b/spm_mesh_edges.m
@@ -0,0 +1,25 @@
+function E = spm_mesh_edges(M)
+% Return edges of a surface mesh
+% FORMAT E = spm_mesh_edges(M)
+% M        - a [nx3] faces array or a patch handle/structure
+%
+% E        - a [mx2] edges array 
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_edges.m 4018 2010-07-27 18:22:42Z guillaume $
+
+
+%-Parse input arguments
+%--------------------------------------------------------------------------
+if ishandle(M)
+    M = get(M,'Faces');
+elseif ~isnumeric(M)
+    M = M.faces;
+end
+
+%-Compute edges
+%--------------------------------------------------------------------------
+M = sort(M,2);
+E = unique([M(:,[1 2]);M(:,[2 3]);M(:,[1 3])],'rows');
diff --git a/spm_mesh_geodesic.m b/spm_mesh_geodesic.m
new file mode 100644
index 0000000..fcb4e38
--- /dev/null
+++ b/spm_mesh_geodesic.m
@@ -0,0 +1,19 @@
+function D = spm_mesh_geodesic(M,i,order)
+% Approximate geodesic distances on a mesh using Dijkstra algorith
+% FORMAT D = spm_mesh_geodesic(M,i,order)
+% M        - a patch structure or an adjacency matrix
+% i        - indices of starting points
+% order    - type of distance for 1st order neighbours {0, [1]}
+%
+% D        - a [nx1] vector of geodesic distances from i
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_geodesic.m 4081 2010-10-07 14:04:44Z guillaume $
+
+if nargin < 3, order = 1; end
+
+[N, D] = spm_mesh_neighbours(M,order);
+
+D      = spm_mesh_utils('dijkstra',N,D,i,Inf);
diff --git a/spm_mesh_label.m b/spm_mesh_label.m
index 3ab5746..5802c66 100644
--- a/spm_mesh_label.m
+++ b/spm_mesh_label.m
@@ -12,15 +12,15 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_mesh_label.m 3991 2010-07-13 11:01:17Z guillaume $
+% $Id: spm_mesh_label.m 4035 2010-08-05 18:54:32Z guillaume $
 
 
 %-Parse input arguments
 %--------------------------------------------------------------------------
 if ishandle(M)
-    M = get(M,'Faces');
-elseif ~isnumeric(M)
-    M = M.faces;
+    M = struct('faces',get(M,'Faces'), 'vertices',get(M,'Vertices'));
+elseif isnumeric(M)
+    M = struct('faces',M);
 end
 
 if nargin < 2, opt = 'faces'; end
@@ -38,9 +38,9 @@
     case 'faces'
         %-Label faces
         %------------------------------------------------------------------
-        C = zeros(size(M,1),1);
+        C = zeros(size(M.faces,1),1);
         for i=1:length(r)-1
-            C(any(ismember(M,p(r(i):r(i+1)-1)),2)) = i;
+            C(any(ismember(M.faces,p(r(i):r(i+1)-1)),2)) = i;
         end
 
     case 'vertices'
diff --git a/spm_mesh_neighbours.m b/spm_mesh_neighbours.m
new file mode 100644
index 0000000..0b110cc
--- /dev/null
+++ b/spm_mesh_neighbours.m
@@ -0,0 +1,43 @@
+function [N,D] = spm_mesh_neighbours(M,order)
+% Return first-order neighbours of all vertices of a surface mesh
+% FORMAT N = spm_mesh_neighbours(M,order)
+% M        - a patch structure or an adjacency matrix
+% order    - ordinal or euclidean distance for 1st order neighbours {[0],1}
+%
+% N        - a [nxp] neighbours array (n = #vertices, p = # max neighbours)
+% D        - a [nxp] distance array to neighbours
+% N & D contain 0 when number of neighbours is smaller than p.
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_neighbours.m 4081 2010-10-07 14:04:44Z guillaume $
+
+if nargin < 2, order = 0; end
+
+%-Compute adjacency matrix
+%--------------------------------------------------------------------------
+if issparse(M) && size(M,1) == size(M,2)
+    A = M;
+else
+    A = spm_mesh_distmtx(M,order);
+end
+
+%-Build neighbours array
+%--------------------------------------------------------------------------
+if nargout > 1
+    [N, D] = spm_mesh_utils('neighbours',A);
+else
+    N = spm_mesh_utils('neighbours',A);
+end
+
+% [i, j, s] = find(A);
+% p = max(sum(A~=0,2));
+% N = zeros(size(A,1),p);
+% if nargout > 1, D = zeros(size(N)); end
+% l = 1;
+% for k=1:length(j)
+%     N(j(k),l) = i(k);
+%     if nargout > 1, D(j(k),l) = s(k); end
+%     try, if j(k+1) ~= j(k), l = 1; else l = l + 1; end; end
+% end
diff --git a/spm_mesh_project.m b/spm_mesh_project.m
index e359306..9ef77ef 100644
--- a/spm_mesh_project.m
+++ b/spm_mesh_project.m
@@ -15,11 +15,11 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_mesh_project.m 3409 2009-09-18 15:40:25Z guillaume $
+% $Id: spm_mesh_project.m 4018 2010-07-27 18:22:42Z guillaume $
 
 if ishandle(M)
     V = get(M,'Vertices');
-elseif isstruct(M)
+elseif isstruct(M) || isa(M,'gifti')
     V = M.vertices;
 else
     V = M;
@@ -41,8 +41,13 @@
         Y      = dat(i).dat;
         mat    = dat(i).mat;
     else
+        if isfield(dat,'Z') %-xSPM structure
+           dat = struct('dim',dat.DIM,'XYZ',dat.XYZ,'t',dat.Z,'mat',dat.M);
+        end
         Y      = zeros(dat(i).dim(1:3)');
-        OFF    = dat(i).XYZ(1,:) + dat(i).dim(1)*(dat(i).XYZ(2,:)-1 + dat(i).dim(2)*(dat(i).XYZ(3,:)-1));
+        OFF    = dat(i).XYZ(1,:) + ...
+                 dat(i).dim(1)*(dat(i).XYZ(2,:)-1 + ...
+                 dat(i).dim(2)*(dat(i).XYZ(3,:)-1));
         Y(OFF) = dat(i).t .* (dat(i).t > 0);
         mat    = dat(i).mat;
     end
diff --git a/spm_mesh_render.m b/spm_mesh_render.m
index 8cf2ec5..94cae97 100644
--- a/spm_mesh_render.m
+++ b/spm_mesh_render.m
@@ -1,26 +1,41 @@
 function varargout = spm_mesh_render(action,varargin)
 % Display a surface mesh & various utilities
-% FORMAT H = spm_mesh_render('Disp',M)
+% FORMAT H = spm_mesh_render('Disp',M,'PropertyName',propertyvalue)
 % M        - a GIfTI filename/object or patch structure
 % H        - structure containing handles of various objects
+% Opens a new figure unless a 'parent' Property is provided with an axis
+% handle.
 %
 % FORMAT H = spm_mesh_render(M)
-% Shortcut to previous call format
+% Shortcut to previous call format.
 %
 % FORMAT H = spm_mesh_render('ContextMenu',AX)
-% AX       - axis handle or structure given by spm_mesh_render('Disp',...)
+% AX       - axis handle or structure returned by spm_mesh_render('Disp',...)
 %
-% FORMAT H = spm_mesh_render('AddOverlay',AX,P)
+% FORMAT H = spm_mesh_render('Overlay',AX,P)
 % AX       - axis handle or structure given by spm_mesh_render('Disp',...)
 % P        - data to be overlayed on mesh (see spm_mesh_project)
 %
-% FORMAT H = spm_mesh_render('AddColourBar',AX)
-% AX       - axis handle or structure given by spm_mesh_render('Disp',...)
+% FORMAT H = spm_mesh_render('ColourBar',AX,MODE)
+% AX       - axis handle or structure returned by spm_mesh_render('Disp',...)
+% MODE     - {['on'],'off'}
+%
+% FORMAT H = spm_mesh_render('ColourMap',AX,MAP)
+% AX       - axis handle or structure returned by spm_mesh_render('Disp',...)
+% MAP      - a colour map matrix
+%
+% FORMAT MAP = spm_mesh_render('ColourMap',AX)
+% Retrieves the current colourmap.
+%
+% FORMAT spm_mesh_render('Register',AX,hReg)
+% AX       - axis handle or structure returned by spm_mesh_render('Disp',...)
+% hReg     - Handle of HandleGraphics object to build registry in.
+% See spm_XYZreg for more information.
 %__________________________________________________________________________
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_mesh_render.m 4003 2010-07-19 18:22:38Z guillaume $
+% $Id: spm_mesh_render.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 %-Input parameters
@@ -74,6 +89,7 @@
             'AmbientStrength',  0.1,...
             'DiffuseStrength',  0.7,...
             'SpecularExponent', 10,...
+            'Clipping',         'off',...
             'DeleteFcn',        {@myDeleteFcn, renderer},...
             'Visible',          'off',...
             'Tag',              'SPMMeshRender',...
@@ -126,85 +142,179 @@
         H = getHandles(varargin{1});
         if ~isempty(get(H.patch,'UIContextMenu')), return; end
         
-        cmenu = uicontextmenu('Callback',{@myMenuCallback});
+        cmenu = uicontextmenu('Callback',{@myMenuCallback, H});
         
-        c1 = uimenu(cmenu, 'Label','Inflate', 'Interruptible','off', ...
+        uimenu(cmenu, 'Label','Inflate', 'Interruptible','off', ...
             'Callback',{@myInflate, H});
         
-        c1 = uimenu(cmenu, 'Label','Overlay...', 'Interruptible','off', ...
+        uimenu(cmenu, 'Label','Overlay...', 'Interruptible','off', ...
             'Callback',{@myOverlay, H});
         
-        c2 = uimenu(cmenu, 'Label', 'Connected Components', 'Interruptible','off');
-        C=getappdata(H.patch,'cclabel');
+        uimenu(cmenu, 'Label','Image Sections...', 'Interruptible','off', ...
+            'Callback',{@myImageSections, H});
+        
+        c = uimenu(cmenu, 'Label', 'Connected Components', 'Interruptible','off');
+        C = getappdata(H.patch,'cclabel');
         for i=1:length(unique(C))
-            uimenu(c2, 'Label',sprintf('Component %d',i), 'Checked','on', ...
+            uimenu(c, 'Label',sprintf('Component %d',i), 'Checked','on', ...
                 'Callback',{@myCCLabel, H});
         end
         
-        c3 = uimenu(cmenu, 'Label','Rotate', 'Checked','on', 'Separator','on', ...
+        uimenu(cmenu, 'Label','Rotate', 'Checked','on', 'Separator','on', ...
             'Callback',{@mySwitchRotate, H});
         
-        c3 = uimenu(cmenu, 'Label','Synchronise Views', 'Visible','off', ...
+        uimenu(cmenu, 'Label','Synchronise Views', 'Visible','off', ...
             'Checked','off', 'Tag','SynchroMenu', 'Callback',{@mySynchroniseViews, H});
         
-        c4 = uimenu(cmenu, 'Label','View');
-        uimenu(c4, 'Label','Go to Y-Z view (right)',  'Callback', {@myView, H, [90 0]});
-        uimenu(c4, 'Label','Go to Y-Z view (left)',   'Callback', {@myView, H, [-90 0]});
-        uimenu(c4, 'Label','Go to X-Y view (top)',    'Callback', {@myView, H, [0 90]});
-        uimenu(c4, 'Label','Go to X-Y view (bottom)', 'Callback', {@myView, H, [-180 -90]});
-        uimenu(c4, 'Label','Go to X-Z view (front)',  'Callback', {@myView, H, [-180 0]});
-        uimenu(c4, 'Label','Go to X-Z view (back)',   'Callback', {@myView, H, [0 0]});
+        c = uimenu(cmenu, 'Label','View');
+        uimenu(c, 'Label','Go to Y-Z view (right)',  'Callback', {@myView, H, [90 0]});
+        uimenu(c, 'Label','Go to Y-Z view (left)',   'Callback', {@myView, H, [-90 0]});
+        uimenu(c, 'Label','Go to X-Y view (top)',    'Callback', {@myView, H, [0 90]});
+        uimenu(c, 'Label','Go to X-Y view (bottom)', 'Callback', {@myView, H, [-180 -90]});
+        uimenu(c, 'Label','Go to X-Z view (front)',  'Callback', {@myView, H, [-180 0]});
+        uimenu(c, 'Label','Go to X-Z view (back)',   'Callback', {@myView, H, [0 0]});
+        
+        uimenu(cmenu, 'Label','Colorbar', 'Callback', {@myColourbar, H});
         
-        c5 = uimenu(cmenu, 'Label','Transparency');
-        uimenu(c5, 'Label','0%',  'Checked','on',  'Callback', {@myTransparency, H});
-        uimenu(c5, 'Label','20%', 'Checked','off', 'Callback', {@myTransparency, H});
-        uimenu(c5, 'Label','40%', 'Checked','off', 'Callback', {@myTransparency, H});
-        uimenu(c5, 'Label','60%', 'Checked','off', 'Callback', {@myTransparency, H});
-        uimenu(c5, 'Label','80%', 'Checked','off', 'Callback', {@myTransparency, H});
+        c = uimenu(cmenu, 'Label','Colormap');
+        clrmp = {'hot' 'jet' 'gray' 'hsv' 'bone' 'copper' 'pink' 'white' ...
+            'flag' 'lines' 'colorcube' 'prism' 'cool' 'autumn' ...
+             'spring' 'winter' 'summer'};
+        for i=1:numel(clrmp)
+            uimenu(c, 'Label', clrmp{i}, 'Callback', {@myColourmap, H});
+        end
+        
+        c = uimenu(cmenu, 'Label','Transparency');
+        uimenu(c, 'Label','0%',  'Checked','on',  'Callback', {@myTransparency, H});
+        uimenu(c, 'Label','20%', 'Checked','off', 'Callback', {@myTransparency, H});
+        uimenu(c, 'Label','40%', 'Checked','off', 'Callback', {@myTransparency, H});
+        uimenu(c, 'Label','60%', 'Checked','off', 'Callback', {@myTransparency, H});
+        uimenu(c, 'Label','80%', 'Checked','off', 'Callback', {@myTransparency, H});
         
-        c6 = uimenu(cmenu, 'Label','Background Color');
-        uimenu(c6, 'Label','White',     'Callback', {@myBackgroundColor, H, [1 1 1]});
-        uimenu(c6, 'Label','Black',     'Callback', {@myBackgroundColor, H, [0 0 0]});
-        uimenu(c6, 'Label','Custom...', 'Callback', {@myBackgroundColor, H, []});
+        uimenu(cmenu, 'Label','Data Cursor', 'Callback', {@myDataCursor, H});
         
-        c7 = uimenu(cmenu, 'Label','Save As...', 'Separator', 'on', ...
+        c = uimenu(cmenu, 'Label','Background Color');
+        uimenu(c, 'Label','White',     'Callback', {@myBackgroundColor, H, [1 1 1]});
+        uimenu(c, 'Label','Black',     'Callback', {@myBackgroundColor, H, [0 0 0]});
+        uimenu(c, 'Label','Custom...', 'Callback', {@myBackgroundColor, H, []});
+        
+        uimenu(cmenu, 'Label','Save As...', 'Separator', 'on', ...
             'Callback', {@mySave, H});
         
+        set(H.rotate3d,'enable','off');
         try, set(H.rotate3d,'uicontextmenu',cmenu); end
         try, set(H.patch,   'uicontextmenu',cmenu); end
+        set(H.rotate3d,'enable','on');
+        
+        dcm_obj = datacursormode(H.figure);
+        set(dcm_obj, 'Enable','off', 'SnapToDataVertex','on', ...
+            'DisplayStyle','Window', 'Updatefcn',{@myDataCursorUpdate, H});
         
-    %-AddOverlay
+    %-Overlay
     %======================================================================
-    case 'addoverlay'
+    case 'overlay'
         if isempty(varargin), varargin{1} = gca; end
         H = getHandles(varargin{1});
         if nargin < 3, varargin{2} = []; end
-        if nargin < 4, varargin{3} = []; end
-        updateTexture(H,varargin{2},varargin{3});
+        updateTexture(H,varargin{2:end});
         
-    %-AddColourBar
+    %-Slices
     %======================================================================
-    case 'addcolourbar'
+    case 'slices'
         if isempty(varargin), varargin{1} = gca; end
         H = getHandles(varargin{1});
-        d = getappdata(H.patch,'data');
+        if nargin < 3, varargin{2} = []; end
+        renderSlices(H,varargin{2:end});
+    
+    %-ColourBar
+    %======================================================================
+    case {'colourbar', 'colorbar'}
+        if isempty(varargin), varargin{1} = gca; end
+        if length(varargin) == 1, varargin{2} = 'on'; end
+        H   = getHandles(varargin{1});
+        d   = getappdata(H.patch,'data');
         col = getappdata(H.patch,'colourmap');
-        if isempty(d) || ~any(d(:)), return; end
+        if strcmpi(varargin{2},'off')
+            if isfield(H,'colourbar') && ishandle(H.colourbar)
+                delete(H.colourbar);
+                H = rmfield(H,'colourbar');
+                setappdata(H.axis,'handles',H);
+            end
+            return;
+        end
+        if isempty(d) || ~any(d(:)), varargout = {H}; return; end
         if isempty(col), col = hot(256); end
-        H.colourbar = colorbar('peer',H.axis);
+        if ~isfield(H,'colourbar') || ~ishandle(H.colourbar)
+            H.colourbar = colorbar('peer',H.axis);
+            set(H.colourbar,'Tag','');
+            set(get(H.colourbar,'Children'),'Tag','');
+        end
         c(1:size(col,1),1,1:size(col,2)) = col;
-        set(get(H.colourbar,'child'),'CData',c);
-        set(get(H.colourbar,'child'),'YData',[min(d) max(d)]);
-        set(H.colourbar,'YLim',[min(d) max(d)]);
-        set(H.colourbar,'Tag','');
-        set(get(H.colourbar,'child'),'Tag','');
+        ic = findobj(H.colourbar,'Type','image');
+        if size(d,1) > 1
+            set(ic,'CData',c(1:size(d,1),:,:));
+            set(ic,'YData',[1 size(d,1)]);
+            set(H.colourbar,'YLim',[1 size(d,1)]);
+            set(H.colourbar,'YTickLabel',[]);
+        else
+            set(ic,'CData',c);
+            clim = getappdata(H.patch,'clim');
+            if isempty(clim), clim = [false min(d) max(d)]; end
+            set(ic,'YData',clim(2:3));
+            set(H.colourbar,'YLim',clim(2:3));
+        end
         setappdata(H.axis,'handles',H);
         
+    %-ColourMap
+    %======================================================================
+    case {'colourmap', 'colormap'}
+        if isempty(varargin), varargin{1} = gca; end
+        H = getHandles(varargin{1});
+        if length(varargin) == 1
+            varargout = { getappdata(H.patch,'colourmap') };
+            return;
+        else
+            setappdata(H.patch,'colourmap',varargin{2});
+            d = getappdata(H.patch,'data');
+            updateTexture(H,d);
+        end
+    
+    %-CLim
+    %======================================================================
+    case 'clim'
+        if isempty(varargin), varargin{1} = gca; end
+        H = getHandles(varargin{1});
+        if length(varargin) == 1
+            c = getappdata(H.patch,'clim');
+            if ~isempty(c), c = c(2:3); end
+            varargout = { c };
+            return;
+        else
+            if isempty(varargin{2}) || any(~isfinite(varargin{2}))
+                setappdata(H.patch,'clim',[false NaN NaN]);
+            else
+                setappdata(H.patch,'clim',[true varargin{2}]);
+            end
+            d = getappdata(H.patch,'data');
+            updateTexture(H,d);
+        end
+        
+    %-Register
+    %======================================================================
+    case 'register'
+        if isempty(varargin), varargin{1} = gca; end
+        H = getHandles(varargin{1});
+        hReg = varargin{2};
+        xyz  = spm_XYZreg('GetCoords',hReg);
+        hs   = myCrossBar('Create',H,xyz);
+        set(hs,'UserData',hReg);
+        spm_XYZreg('Add2Reg',hReg,hs,@myCrossBar);
+        
     %-Otherwise...
     %======================================================================
     otherwise
         try
-            spm_mesh_render('Disp',action,varargin{:});
+            H = spm_mesh_render('Disp',action,varargin{:});
         catch
             error('Unknown action.');
         end
@@ -232,8 +342,10 @@
 
 %==========================================================================
 function H = getHandles(H)
+if ~nargin || isempty(H), H = gca; end
 if ishandle(H) && ~isappdata(H,'handles')
-    H.axis     = H;
+    a = H; clear H;
+    H.axis     = a;
     H.figure   = ancestor(H.axis,'figure');
     H.patch    = findobj(H.axis,'type','patch');
     H.light    = findobj(H.axis,'type','light');
@@ -242,9 +354,40 @@
 elseif ishandle(H)
     H = getappdata(H,'handles');
 else
-    %H = H;
+    H = getappdata(H.axis,'handles');
+end
+
+%==========================================================================
+function myMenuCallback(obj,evt,H)
+H = getHandles(H);
+
+h = findobj(obj,'Label','Rotate');
+if strcmpi(get(H.rotate3d,'Enable'),'on')
+    set(h,'Checked','on');
+else
+    set(h,'Checked','off');
+end
+
+if numel(findobj('Tag','SPMMeshRender','Type','Patch')) > 1
+    h = findobj(obj,'Tag','SynchroMenu');
+    set(h,'Visible','on');
 end
 
+h = findobj(obj,'Label','Colorbar');
+d = getappdata(H.patch,'data');
+if isempty(d) || ~any(d(:)), set(h,'Enable','off'); else set(h,'Enable','on'); end
+if isfield(H,'colourbar')
+    if ishandle(H.colourbar)
+        set(h,'Checked','on');
+    else
+        H = rmfield(H,'colourbar');
+        set(h,'Checked','off');
+    end
+else
+    set(h,'Checked','off');
+end
+setappdata(H.axis,'handles',H);
+
 %==========================================================================
 function myPostCallback(obj,evt,H)
 P = findobj('Tag','SPMMeshRender','Type','Patch');
@@ -256,14 +399,36 @@ function myPostCallback(obj,evt,H)
         camlight(H.light);
     end
 end
-    
+
 %==========================================================================
-function myMenuCallback(obj,evt)
-if numel(findobj('Tag','SPMMeshRender','Type','Patch')) > 1
-    h = findobj(obj,'Tag','SynchroMenu');
-    if ~isempty(h)
-        set(h,'Visible','on');
-    end
+function varargout = myCrossBar(varargin)
+
+switch lower(varargin{1})
+
+    case 'create'
+    %----------------------------------------------------------------------
+    % hMe = myCrossBar('Create',H,xyz)
+    H  = varargin{2};
+    xyz = varargin{3};
+    hold(H.axis,'on');
+    hs = plot3(xyz(1),xyz(2),xyz(3),'Marker','+','MarkerSize',40,...
+        'parent',H.axis,'Color',[1 1 1],'Tag','CrossBar','ButtonDownFcn',{});
+    varargout = {hs};
+    
+    case 'setcoords'
+    %----------------------------------------------------------------------
+    % [xyz,d] = myCrossBar('SetCoords',xyz,hMe)
+    hMe  = varargin{3};
+    xyz  = varargin{2};
+    set(hMe,'XData',xyz(1));
+    set(hMe,'YData',xyz(2));
+    set(hMe,'ZData',xyz(3));
+    varargout = {xyz,[]};
+    
+    otherwise
+    %----------------------------------------------------------------------
+    error('Unknown action string')
+
 end
 
 %==========================================================================
@@ -304,6 +469,13 @@ function myCCLabel(obj,evt,H)
     set(H.patch, 'FaceAlpha', 'interp');
 end
 
+%==========================================================================
+function myTransparency(obj,evt,H)
+t = 1 - sscanf(get(obj,'Label'),'%d%%') / 100;
+set(H.patch,'FaceAlpha',t);
+set(get(get(obj,'parent'),'children'),'Checked','off');
+set(obj,'Checked','on');
+
 %==========================================================================
 function mySwitchRotate(obj,evt,H)
 if strcmpi(get(H.rotate3d,'enable'),'on')
@@ -321,7 +493,16 @@ function myView(obj,evt,H,varargin)
 camlight(H.light);
 
 %==========================================================================
-function mySynchroniseViews(obj,evt,H,varargin)
+function myColourbar(obj,evt,H)
+y = {'on','off'}; toggle = @(x) y{1+strcmpi(x,'on')};
+spm_mesh_render('Colourbar',H,toggle(get(obj,'Checked')));
+
+%==========================================================================
+function myColourmap(obj,evt,H)
+spm_mesh_render('Colourmap',H,feval(get(obj,'Label'),256));
+
+%==========================================================================
+function mySynchroniseViews(obj,evt,H)
 P = findobj('Tag','SPMMeshRender','Type','Patch');
 v = get(H.axis,'cameraposition');
 for i=1:numel(P)
@@ -332,11 +513,29 @@ function mySynchroniseViews(obj,evt,H,varargin)
 end
 
 %==========================================================================
-function myTransparency(obj,evt,H)
-t = 1 - sscanf(get(obj,'Label'),'%d%%') / 100;
-set(H.patch,'FaceAlpha',t);
-set(get(get(obj,'parent'),'children'),'Checked','off');
-set(obj,'Checked','on');
+function myDataCursor(obj,evt,H)
+dcm_obj = datacursormode(H.figure);
+set(dcm_obj, 'Enable','on', 'SnapToDataVertex','on', ...
+    'DisplayStyle','Window', 'Updatefcn',{@myDataCursorUpdate, H});
+
+%==========================================================================
+function txt = myDataCursorUpdate(obj,evt,H)
+pos = get(evt,'Position');
+txt = {['X: ',num2str(pos(1))],...
+       ['Y: ',num2str(pos(2))],...
+       ['Z: ',num2str(pos(3))]};
+i = ismember(get(H.patch,'vertices'),pos,'rows');
+txt = {['Node: ' num2str(find(i))] txt{:}};
+d = getappdata(H.patch,'data');
+if ~isempty(d) && any(d(:))
+    if any(i), txt = {txt{:} ['T: ',num2str(d(i))]}; end
+end
+hMe = findobj(H.axis,'Tag','CrossBar');
+if ~isempty(hMe)
+    ws = warning('off');
+    spm_XYZreg('SetCoords',pos,get(hMe,'UserData'));
+    warning(ws);
+end
 
 %==========================================================================
 function myBackgroundColor(obj,evt,H,varargin)
@@ -384,15 +583,35 @@ function mySave(obj,evt,H)
     end
     switch filterindex
         case 1
-            save(gifti(H.patch),fullfile(pathname, filename));
+            G = gifti(H.patch);
+            [p,n,e] = fileparts(filename);
+            [p,n,e] = fileparts(n);
+            switch lower(e)
+                case '.func'
+                    save(gifti(getappdata(H.patch,'data')),...
+                        fullfile(pathname, filename));
+                case '.surf'
+                    save(gifti(struct('vertices',G.vertices,'faces',G.faces)),...
+                        fullfile(pathname, filename));
+                case '.rgba'
+                    save(gifti(G.cdata),fullfile(pathname, filename));
+                otherwise
+                    save(G,fullfile(pathname, filename));
+            end
         case 2
-            u = get(H.axis,'units');
+            u  = get(H.axis,'units');
             set(H.axis,'units','pixels');
-            p = get(H.axis,'Position');
-            r = get(getappdata(obj,'fig'),'Renderer');
+            p  = get(H.axis,'Position');
+            r  = get(H.figure,'Renderer');
+            hc = findobj(H.figure,'Tag','SPMMeshRenderBackground');
+            if isempty(hc)
+                c = get(H.figure,'Color');
+            else
+                c = get(hc,'Color');
+            end
             h = figure('Position',p+[0 0 10 10], ...
                 'InvertHardcopy','off', ...
-                'Color',get(H.figure,'Color'), ...
+                'Color',c, ...
                 'Renderer',r);
             copyobj(H.axis,h);
             set(H.axis,'units',u);
@@ -420,14 +639,44 @@ function myDeleteFcn(obj,evt,renderer)
 
 %==========================================================================
 function myOverlay(obj,evt,H)
-[P, sts] = spm_select(1,'image','Select image to overlay');
+[P, sts] = spm_select(1,'\.img|\.nii|\.gii|\.mat','Select file to overlay');
 if ~sts, return; end
-spm_mesh_render('AddOverlay',H,P);
+spm_mesh_render('Overlay',H,P);
+
+%==========================================================================
+function myImageSections(obj,evt,H)
+[P, sts] = spm_select(1,'image','Select image to render');
+if ~sts, return; end
+renderSlices(H,P);
+
+%==========================================================================
+function renderSlices(H,P,pls)
+if nargin <3
+    pls = 0.05:0.2:0.9;
+end
+N   = nifti(P);
+d   = size(N.dat);
+pls = round(pls.*d(3));
+hold(H.axis,'on');
+for i=1:numel(pls)
+    [x,y,z] = ndgrid(1:d(1),1:d(2),pls(i));
+    f  = N.dat(:,:,pls(i));
+    x1 = N.mat(1,1)*x + N.mat(1,2)*y + N.mat(1,3)*z + N.mat(1,4);
+    y1 = N.mat(2,1)*x + N.mat(2,2)*y + N.mat(2,3)*z + N.mat(2,4);
+    z1 = N.mat(3,1)*x + N.mat(3,2)*y + N.mat(3,3)*z + N.mat(3,4);
+    surf(x1,y1,z1, repmat(f,[1 1 3]), 'EdgeColor','none', ...
+        'Clipping','off', 'Parent',H.axis);
+end
+hold(H.axis,'off');
+axis(H.axis,'image');
 
 %==========================================================================
 function C = updateTexture(H,v,col)
 
-if nargin <3 || isempty(col), col = hot(256); end
+%-Get colourmap
+%--------------------------------------------------------------------------
+if nargin<3, col = getappdata(H.patch,'colourmap'); end
+if isempty(col), col = hot(256); end
 setappdata(H.patch,'colourmap',col);
 
 %-Get curvature
@@ -439,24 +688,45 @@ function myOverlay(obj,evt,H)
     
 %-Project data onto surface mesh
 %--------------------------------------------------------------------------
-if isstruct(v) && isfield(v,'FWHM') %-xSPM structure
-    v = struct('XYZ',v.XYZ, 't',v.Z, 'mat',v.M, 'dim',v.DIM);
+if nargin < 2, v = []; end
+if ischar(v)
+    [p,n,e] = fileparts(v);
+    if strcmp([n e],'SPM.mat')
+        swd = pwd;
+        spm_figure('GetWin','Interactive');
+        [SPM,v] = spm_getSPM(struct('swd',p));
+        cd(swd);
+    else
+        try, spm_vol(v); catch, v = gifti(v); end;
+    end
 end
+if isa(v,'gifti'), v = v.cdata; end
 if isempty(v)
     v = zeros(size(curv))';
 elseif ischar(v) || iscellstr(v) || isstruct(v)
     v = spm_mesh_project(H.patch,v);
+elseif isnumeric(v) || islogical(v)
+    if size(v,2) == 1
+        v = v';
+    end
+else
+    error('Unknown data type.');
 end
+v(isinf(v)) = NaN;
 
 setappdata(H.patch,'data',v);
 
-%-Build texture by merging curvature and data
+%-Create RGB representation of data according to colourmap
 %--------------------------------------------------------------------------
 C = zeros(size(v,2),3);
+clim = getappdata(H.patch, 'clim');
+if isempty(clim), clim = [false NaN NaN]; end
+mi = clim(2); ma = clim(3);
 if any(v(:))
     if size(col,1)>3
         if size(v,1) == 1
-            C = squeeze(ind2rgb(floor(v(:)/max(v(:))*size(col,1)),col));
+            if ~clim(1), mi = min(v(:)); ma = max(v(:)); end
+            C = squeeze(ind2rgb(floor(((v(:)-mi)/(ma-mi))*size(col,1)),col));
         else
             C = v; v = v';
         end
@@ -466,15 +736,18 @@ function myOverlay(obj,evt,H)
             C = C + v(i,:)'/m * col(i,:);
         end
     end
+else
 end
+setappdata(H.patch, 'clim', [false mi ma]);
+
+%-Build texture by merging curvature and data
+%--------------------------------------------------------------------------
 C = repmat(~any(v,1),3,1)' .* curv + repmat(any(v,1),3,1)' .* C;
 
 set(H.patch, 'FaceVertexCData',C, 'FaceColor','interp');
 
-% if isfield(M,'cdata') && ~isempty(M.cdata)
-%     if any(size(M.cdata)==1), M.cdata = M.cdata(:); end
-%     if size(M.cdata,1)==3,    M.cdata = M.cdata';   end
-%     set(H.patch, 'FaceVertexCData',M.cdata, 'FaceColor','interp');
-% else
-%     set(H.patch, 'FaceVertexCData',curv, 'FaceColor','interp');
-% end
+%-Update the colourbar
+%--------------------------------------------------------------------------
+if isfield(H,'colourbar')
+    spm_mesh_render('Colourbar',H);
+end
diff --git a/spm_mesh_resels.m b/spm_mesh_resels.m
new file mode 100644
index 0000000..600053d
--- /dev/null
+++ b/spm_mesh_resels.m
@@ -0,0 +1,99 @@
+function [R, RPV] = spm_mesh_resels(M,T,S)
+% Returns the RESEL counts of a search volume on a surface mesh
+% FORMAT R = spm_mesh_resels(M,T,[S])
+% M        - a patch structure or [nx3] faces array (#faces = n)
+% T        - a [mx1] logical vector (#vertices = m) defining search volume
+% S        - a [mxp] array of standardised residuals [optional]
+%
+% R        - a [1xD] array of RESEL counts {adimensional}
+% RPV      - a [mx1] vector of RESELs per vertex
+%__________________________________________________________________________
+%
+% References:
+%
+% [1] Detecting Sparse Signals in Random Fields, With an Application to
+% Brain Imaging, J.E. Taylor and K.J. Worsley, Journal of the American
+% Statistical Association, 102(479):913-928, 2007.
+%
+% [2] SurfStat: http://www.math.mcgill.ca/keith/surfstat/, K.J. Worsley.
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_resels.m 4139 2010-12-15 18:31:49Z guillaume $
+
+%-Parse input arguments
+%--------------------------------------------------------------------------
+if nargin < 3
+    if ~isnumeric(M)
+        S = M.vertices;
+    else
+        S = ones(max(M(:)),2);
+    end
+end
+
+if ~isnumeric(M), M = M.faces; end
+
+if nargin < 2
+    T = true(size(S,1),1);
+end
+if any(isnan(T)), T = ~isnan(T); end
+
+D = 2;
+
+%-Compute edges
+%--------------------------------------------------------------------------
+M      = sort(M,2); % so that ismember(M...,E,'rows') works later
+E      = spm_mesh_edges(M);
+
+%-Compute (standardised) residual sum of squares of differences along edges
+%--------------------------------------------------------------------------
+SSR    = S(E',:);
+SSR    = reshape(SSR',size(SSR,2),2,[]);
+SSR    = sum(squeeze((SSR(:,1,:) - SSR(:,2,:)).^2),1)';
+SSR    = sqrt(SSR);
+
+%-Lipschitz-Killing Curvature (LKC)
+%==========================================================================
+LKC    = zeros(1,3);
+
+%-Vertices contribution
+%--------------------------------------------------------------------------
+LKC(1) = LKC(1) + sum(T);
+
+%-Edges contribution
+%--------------------------------------------------------------------------
+TE     = all(T(E),2);
+
+LKC(1) = LKC(1) - sum(TE);
+LKC(2) = LKC(2) + sum(SSR(TE));
+
+%-Triangles contribution
+%--------------------------------------------------------------------------
+TM     = all(T(M),2);
+% [E,I,J] = unique([M(:,[1 2]);M(:,[2 3]);M(:,[1 3])],'rows');
+% SSR = reshape(SSR(J),[],3); SSR = SSR(TM,:); A = spm_mesh_area(SSR',true);
+[is,l] = ismember(M(TM,[1 2]),E,'rows'); l12 = SSR(l);
+[is,l] = ismember(M(TM,[1 3]),E,'rows'); l13 = SSR(l);
+[is,l] = ismember(M(TM,[2 3]),E,'rows'); l23 = SSR(l);
+A      = spm_mesh_area([l12 l13 l23]',true);
+
+LKC(1) = LKC(1) + sum(TM);
+LKC(2) = LKC(2) - sum(l12 + l13 + l23) / 2;
+LKC(3) = LKC(3) + sum(A);
+
+%-RESELS per vertex
+%==========================================================================
+if nargout > 1
+    RPV = zeros(size(T));
+    for i=1:3
+        RPV = RPV + accumarray(M(TM,i),A,[size(T,1) 1]);
+    end
+    RPV = RPV / 3; % each triangle contributes a third to each of its vertices 
+                   % (sum(A)==sum(RPV))
+    RPV = RPV / sqrt((4*log(2))^D);
+end
+
+%-RESEL Counts
+%==========================================================================
+R      = LKC ./ sqrt((4*log(2)).^(0:D));
diff --git a/spm_mesh_smooth.m b/spm_mesh_smooth.m
new file mode 100644
index 0000000..eca68f9
--- /dev/null
+++ b/spm_mesh_smooth.m
@@ -0,0 +1,31 @@
+function T = spm_mesh_smooth(M, T, S)
+% Perform Gaussian smoothing on data lying on a surface mesh
+% FORMAT GL = spm_mesh_smooth(M)
+% M        - a patch structure or a handle to a patch
+% GL       - graph Laplacian
+%
+% FORMAT T = spm_mesh_smooth(M, T, S)
+% FORMAT T = spm_mesh_smooth(GL, T, S)
+% T        - [vx1] data vector
+% S        - smoothing parameter (number of iterations)
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston, Guillaume Flandin
+% $Id: spm_mesh_smooth.m 4079 2010-10-07 11:41:54Z guillaume $
+
+if isstruct(M) || numel(M) == 1
+    A  = spm_mesh_distmtx(M,0);
+    N  = size(A,1);
+    GL = speye(N,N) + (A - spdiags(sum(A,2),0,N,N))/16;
+else
+    GL = M;
+end
+
+if nargin == 1, T = GL; return; end
+
+for i=1:S
+    T = GL * T;
+end
+
+return;
diff --git a/spm_mesh_split.m b/spm_mesh_split.m
new file mode 100644
index 0000000..9e0c71b
--- /dev/null
+++ b/spm_mesh_split.m
@@ -0,0 +1,39 @@
+function MS = spm_mesh_split(M, C)
+% Split a surface mesh into its connected components 
+% FUNCTION [MS] = spm_mesh_split(M, C)
+% M        - a [nx3] faces array or a patch structure
+% C        - a [nx1] vector containing labels for the connected components
+%            or a logical vector indicating vertices to keep
+%
+% MS       - a patch structure array
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_split.m 4035 2010-08-05 18:54:32Z guillaume $
+
+if isnumeric(M), M = struct('faces',M); end
+
+if nargin < 2
+    [C, N] = spm_mesh_label(M,'vertices');
+else
+    if islogical(C)
+        N  = nnz(C);
+        %[C, N] = spm_mesh_clusters(M, C);
+    else
+        N  = histc(C,1:max(C));
+    end
+end
+
+for i=1:numel(N)
+    j                    = C == i;
+    F                    = M.faces;
+    B                    = NaN(size(C));
+    B(j)                 = 1:N(i);
+    F                    = B(F);
+    F(any(isnan(F),2),:) = [];
+    MS(i).faces          = F;
+    try, MS(i).mat       = M.mat; end
+    try, MS(i).cdata     = M.cdata(j,:); end
+    try, MS(i).vertices  = M.vertices(j,:); end
+end
diff --git a/spm_mesh_to_grid.m b/spm_mesh_to_grid.m
new file mode 100644
index 0000000..1e5192c
--- /dev/null
+++ b/spm_mesh_to_grid.m
@@ -0,0 +1,94 @@
+function R = spm_mesh_to_grid(M, V, T)
+% Non-linear interpolation of surface-based data onto a regular grid
+% FORMAT R = spm_mesh_to_grid(M, V, T)
+% M        - a patch structure with fields 'faces' and 'vertices'
+% V        - an spm_vol structure with fields 'dim' and 'mat'
+% T        - array of data to be interpolated
+%
+% R        - interpolated data on grid defined by V
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston, Guillaume Flandin
+% $Id: spm_mesh_to_grid.m 4079 2010-10-07 11:41:54Z guillaume $
+
+
+%-Precompute interpolation kernel and voxel grid
+%==========================================================================
+
+%-Compute a densely sampled triangular mask
+%--------------------------------------------------------------------------
+[tx, ty]  = meshgrid(0.05:0.1:0.95, 0.05:0.1:0.95);
+tx        = tx(:);
+ty        = ty(:);
+ti        = find(sum([tx ty],2) <= 0.9);
+tx        = tx(ti); 
+ty        = ty(ti); 
+tz        = 1 - tx - ty;
+
+%-Map the dense template triangle onto each face of the surface mesh
+%--------------------------------------------------------------------------
+P1        = M.vertices(M.faces(:,1),:);
+P2        = M.vertices(M.faces(:,2),:);
+P3        = M.vertices(M.faces(:,3),:);
+
+If        = speye(size(M.faces,1));
+
+vertDense = kron(If,tx)*P2 + kron(If,ty)*P3 + kron(If,tz)*P1;
+kernel    = [tz tx ty];
+
+%-Get voxel indices of all vertices from dense mesh within the image
+%--------------------------------------------------------------------------
+voxDense  = V.mat\[vertDense';ones(1,size(vertDense,1))];
+voxDense  = round(voxDense(1:3,:)');
+voxInd    = sub2ind(V.dim,voxDense(:,1),voxDense(:,2),voxDense(:,3));
+voxInd    = reshape(voxInd,size(kernel,1),[]);
+
+
+%-Interpolation
+%==========================================================================
+integralConservation = true;
+
+%-Normalise vertex data by the number of faces they are in
+%--------------------------------------------------------------------------
+if integralConservation
+    A = spm_mesh_adjacency(M);
+    T = spdiags(1./sum(A,2),0,size(A,1),size(A,1)) * T;
+end
+    
+%-Interpolate each dense face data in the voxel grid
+%--------------------------------------------------------------------------
+R = zeros([V.dim size(T,2)]);
+if ~integralConservation, countR = zeros(size(R)); end
+
+for i = 1:size(M.faces,1)
+    %- Values at the vertices of the face
+    faceVal      = T(M.faces(i,:),:);
+    isData       = any(faceVal);
+    if any(isData)
+        %- Voxels corresponding to those vertices
+        [iV,I,J] = unique(voxInd(:,i));
+        for j=find(isData)
+            %- Values at the vertices of the dense face
+            faceValDense = kernel*faceVal(:,j);
+            if integralConservation
+                faceValDense = faceValDense * sum(faceVal(:,j))/sum(faceValDense);
+            end
+            %- Integrating values within voxels
+            k = (j-1) * prod(V.dim);
+            if integralConservation
+                R(iV+k) = R(iV+k) + accumarray(J,faceValDense);
+            else
+                R(iV+k) = R(iV+k) + accumarray(J,faceValDense,[],@mean);
+                countR(iV+k) = countR(iV+k) + 1;
+            end
+        end
+    end
+end
+
+if ~integralConservation
+    countR(~countR) = 1;
+    R = R ./ countR;
+end
+
+if size(T,2) == 1, R = R(:,:,:,1); end
diff --git a/spm_mesh_transform.m b/spm_mesh_transform.m
new file mode 100644
index 0000000..62529f2
--- /dev/null
+++ b/spm_mesh_transform.m
@@ -0,0 +1,25 @@
+function M = spm_mesh_transform(M,T,def)
+% Apply a spatial transformation to vertices of a surface mesh
+% FORMAT M = spm_mesh_transform(M,T,def)
+% M        - a patch structure or a gifti object
+% T        - a [4 x 4] transformation matrix [default: identity]
+% def      - a deformation field (nifti object or filename) [default: none]
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_transform.m 4082 2010-10-07 16:11:48Z guillaume $
+
+C = class(M);
+
+if nargin < 2 || isempty(T)
+    T = eye(4);
+end
+
+if nargin < 3
+    M.vertices = [M.vertices ones(size(M.vertices,1),1)] * T(1:3,:)';
+else
+    M = spm_swarp(M,def,T);
+end
+
+M = feval(C, M);
diff --git a/spm_mesh_utils.m b/spm_mesh_utils.m
new file mode 100644
index 0000000..7e596ce
--- /dev/null
+++ b/spm_mesh_utils.m
@@ -0,0 +1,19 @@
+function varargout = spm_mesh_utils(action,varargin)
+% A gateway function for surface mesh-related compiled algorithms 
+%
+% FORMAT [N, D] = spm_mesh_utils('neighbours',A)
+% Return an array of first-order neighbours given an adjacency matrix
+%
+% FORMAT D = spm_mesh_utils('dijkstra',N,D,i,dmax)
+% Compute geodesic distance on a triangular mesh using Dijkstra algorith
+%
+% FORMAT V = spm_mesh_utils('volume',M)
+% Compute the volume of a closed surface mesh
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_mesh_utils.m 4081 2010-10-07 14:04:44Z guillaume $
+
+%-This is merely the help file for the compiled routine
+error('spm_mesh_utils.c not compiled - see Makefile')
\ No newline at end of file
diff --git a/spm_mfx.m b/spm_mfx.m
index 2573aa6..7496018 100644
--- a/spm_mfx.m
+++ b/spm_mfx.m
@@ -71,9 +71,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_mfx.m 3665 2010-01-08 13:07:41Z guillaume $
+% $Id: spm_mfx.m 4185 2011-02-01 18:46:18Z guillaume $
 
-SVNid = '$Rev: 3665 $';
+SVNid = '$Rev: 4185 $';
 
 %-Say hello
 %--------------------------------------------------------------------------
@@ -90,7 +90,14 @@
 catch
     error('This model has not been estimated.');
 end
-    
+
+%-Change to SPM.swd
+%--------------------------------------------------------------------------
+try
+    cd(swd);
+end
+ 
+
 %-Check this is a repeated measures design
 %--------------------------------------------------------------------------
 n     = length(SPM.Sess);            % number of sessions
@@ -160,7 +167,7 @@
 spm('FigName','Stats: MFX-ReML',Finter); spm('Pointer','Watch')
 
 xsDes.Design = '2nd-level MFX analysis';
-xsDes.Name   = xCon.name;
+xsDes.Name   = xCon(I).name;
 S.xsDes      = xsDes;       % description
 
 
@@ -256,13 +263,13 @@
 %-Change to SPM.swd/mfx and save analysis parameters in SPM.mat file
 %--------------------------------------------------------------------------
 SPM       = S;
-[st, me]  = mkdir('mfx');
+SPM.swd   = fullfile(swd,'mfx');
+[st, me]  = mkdir(SPM.swd);
 if st
-    SPM.swd = fullfile(swd,'mfx');
-    if spm_matlab_version_chk('7') >= 0,
-        save(fullfile(SPM.swd,'SPM'), 'SPM', '-V6');
+    if spm_check_version('matlab','7') >= 0,
+        save(fullfile(SPM.swd,'SPM.mat'), 'SPM', '-V6');
     else
-        save(fullfile(SPM.swd,'SPM'), 'SPM');
+        save(fullfile(SPM.swd,'SPM.mat'), 'SPM');
     end
 else
     error('Could not save SPM.mat in mfx: %s', me)
diff --git a/spm_mip.m b/spm_mip.m
index dcb6fa2..35d5aa1 100644
--- a/spm_mip.m
+++ b/spm_mip.m
@@ -17,9 +17,9 @@ function spm_mip(Z,XYZ,M,units)
 % defined in the atlas of Talairach and Tournoux (1988); unless the third
 % dimension is time.
 %
-% This routine loads a mip putline from MIP.mat. This is an image with
+% This routine loads a mip outline from MIP.mat. This is an image with
 % contours and grids defining the space of Talairach & Tournoux (1988).
-% mip05 corresponds to the Talairach atlas, mip96 to the MNI templates.
+% mip95 corresponds to the Talairach atlas, mip96 to the MNI templates.
 % The outline and grid are superimposed at intensity 0.4.
 %
 % A default colormap of 64 levels is assumed. The pointlist image is
@@ -33,7 +33,7 @@ function spm_mip(Z,XYZ,M,units)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston et al.
-% $Id: spm_mip.m 3691 2010-01-20 17:08:30Z guillaume $
+% $Id: spm_mip.m 4093 2010-10-15 12:57:53Z volkmar $
 
 %-Get units and grid scaling
 %--------------------------------------------------------------------------
diff --git a/spm_mip_ui.m b/spm_mip_ui.m
index 501421d..47e1395 100644
--- a/spm_mip_ui.m
+++ b/spm_mip_ui.m
@@ -69,7 +69,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_mip_ui.m 3535 2009-11-05 14:32:51Z guillaume $
+% $Id: spm_mip_ui.m 4136 2010-12-09 22:22:28Z guillaume $
 
 
 %==========================================================================
@@ -278,7 +278,7 @@
         'Color','r','Fontsize',16,...
         'Tag','hX2r',...
         'ButtonDownFcn','spm_mip_ui(''MoveStart'')');
-    hX3r  = text(Po(4)+xyz(1),Po(3)-xyz(3),'<',....
+    hX3r  = text(Po(4)+xyz(1),Po(3)-xyz(3),'<',...
         'Color','r','Fontsize',16,...
         'Tag','hX3r',...
         'ButtonDownFcn','spm_mip_ui(''MoveStart'')');
diff --git a/spm_motd.man b/spm_motd.man
index 459cbd4..dae7fb0 100644
--- a/spm_motd.man
+++ b/spm_motd.man
@@ -35,7 +35,7 @@
 % or at your option, any later version). Further details on "copyleft"
 % can be found at http://www.gnu.org/copyleft/.
 %_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1991,1994-2011 Wellcome Trust Centre for Neuroimaging
 
 % The FIL Methods Group
-% $Id: spm_motd.man 3028 2009-04-01 12:02:59Z guillaume $
+% $Id: spm_motd.man 4288 2011-04-04 14:45:17Z guillaume $
diff --git a/spm_mvNpdf.m b/spm_mvNpdf.m
index a981702..f0e111d 100644
--- a/spm_mvNpdf.m
+++ b/spm_mvNpdf.m
@@ -1,11 +1,11 @@
-function pdf = spm_Npdf(z,Mu,V)
+function pdf = spm_mvNpdf(z,Mu,V)
 % Probability Density Function (PDF) of multivariate Normal distribution
 % FORMAT pdf = spm_Npdf(z,Mu,V)
 %
 % z  - ordinates
 % Mu - mean (a d-vector)
 % V  - d x d variance-covariance matrix
-%_______________________________________________________________________
+%__________________________________________________________________________
 %
 % spm_Npdf returns the Probability Density Function (PDF) for the
 % multivariate Normal (Gaussian) family of distributions.
@@ -20,38 +20,37 @@
 % For multivarate PDFs, the ordinates must be in the columns of z, so
 % z must have column dimension d. Multiple columns can be entered. 
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1998-2011 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_mvNpdf.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_mvNpdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 %-Condition arguments
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if nargin<1,   pdf=[]; return, end
 if isempty(z), pdf=[]; return, end
 if nargin<2,   Mu=0;           end
 
 %-Check Mu, make a column vector, get dimension
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if min(size(Mu)) > 1, error('Mu must be a vector'); end
 Mu = Mu(:)';
 d  = length(Mu);
-n  = size(z,2);
 
 if nargin<3, V=eye(d); end
 
 %-Size & range checks
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if any(any(V~=V')),     error('V must be symmetric'); end
-if any(size(V)~=[d,d]), error('V wrong dimension'),   end
+if any(size(V)~=[d,d]), error('V wrong dimension');   end
 
 %-Computation
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 if d==1
     %-Simpler computation for univariate normal
-    %---------------------------------------------------------------
+    %----------------------------------------------------------------------
     pdf = exp(-(z - Mu).^2/(2*V))./sqrt(2*pi*V);
 else
     if size(z,1) ~= d, error('z wrong dimension'), end
@@ -62,7 +61,7 @@
 return
 
 %-Notes
-%=======================================================================
+%==========================================================================
 %-The following line computes the PDF in one go for all the ordinates,
 % The diag()'s allow for the multiplicity.
 % This way is inefficient for large numbers of ordinates.
diff --git a/spm_mvb_ui.m b/spm_mvb_ui.m
index 811cb85..e5e6d9f 100644
--- a/spm_mvb_ui.m
+++ b/spm_mvb_ui.m
@@ -77,7 +77,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_mvb_ui.m 3656 2009-12-23 20:17:30Z karl $
+% $Id: spm_mvb_ui.m 4185 2011-02-01 18:46:18Z guillaume $
  
  
 %-Get figure handles and set title
@@ -222,10 +222,10 @@
  
 % save
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >= 0
-    save(fullfile(SPM.swd,name),'-V6','MVB')
+if spm_check_version('matlab','7') >= 0
+    save(fullfile(SPM.swd,[name '.mat']),'-V6','MVB')
 else
-    save(fullfile(SPM.swd,name),'MVB')
+    save(fullfile(SPM.swd,[name '.mat']),'MVB')
 end
 assignin('base','MVB',MVB)
 
diff --git a/spm_ncTcdf.m b/spm_ncTcdf.m
index 8d379d8..12e6bb2 100644
--- a/spm_ncTcdf.m
+++ b/spm_ncTcdf.m
@@ -2,7 +2,7 @@
 % Cumulative Distribution Function (CDF) of non-central t-distribution
 % FORMAT [p] = spm_ncTcdf(x,v,d)
 % x -  T-variate (Student's t has range (-Inf,Inf)
-% v - degrees of freedom (v>0, non-integer d.f. accepted
+% v - degrees of freedom (v>0, non-integer d.f. accepted)
 % F - CDF of non-central t-distribution with v d.f. at points x
 %
 % see:
@@ -11,10 +11,11 @@
 % Russell V. Lenth
 % Applied Statistics, Vol. 38, No. 1 (1989), pp. 185-189
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 2009-2011 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_ncTcdf.m 2807 2009-03-02 12:08:44Z karl $
+% $Id: spm_ncTcdf.m 4182 2011-02-01 12:29:09Z guillaume $
+
 
 % CDF
 %--------------------------------------------------------------------------
diff --git a/spm_nlsi_GN.m b/spm_nlsi_GN.m
index b457d9f..73e8655 100644
--- a/spm_nlsi_GN.m
+++ b/spm_nlsi_GN.m
@@ -73,12 +73,26 @@
 % invoked.  The M-Step estimates the precision components of e, in terms
 % of [Re]ML point estimators of the log-precisions.
 %
-% An optional feature selection can be specified with parameters M.FS
+% An optional feature selection can be specified with parameters M.FS.
+%
+% For generic aspects of the scheme see:
+% 
+% Friston K, Mattout J, Trujillo-Barreto N, Ashburner J, Penny W. 
+% Variational free energy and the Laplace approximation. 
+% NeuroImage. 2007 Jan 1;34(1):220-34.
+% 
+% This scheme handels complex data along the lines originally described in:
+% 
+% Sehpard RJ, Lordan BP, and Grant EH. 
+% Least squares analysis of complex data with applications to permittivity
+% measurements.
+% J. Phys. D. Appl. Phys 1970 3:1759-1764.
+%
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_nlsi_GN.m 3812 2010-04-07 16:52:05Z karl $
+% $Id: spm_nlsi_GN.m 4261 2011-03-24 16:39:42Z karl $
  
 % figure (unless disabled)
 %--------------------------------------------------------------------------
@@ -101,24 +115,24 @@
  
 % composition of feature selection and prediction (usually an integrator)
 %--------------------------------------------------------------------------
-if isfield(M,'FS')
+try
     
-    % FS(y,M)
+    % try FS(y,M)
     %----------------------------------------------------------------------
     try
         y  = feval(M.FS,Y.y,M);
         IS = inline([M.FS '(' M.IS '(P,M,U),M)'],'P','M','U');
         
-    % FS(y,M)
+    % try FS(y)
     %----------------------------------------------------------------------
     catch
         y  = feval(M.FS,Y.y);
         IS = inline([M.FS '(' M.IS '(P,M,U))'],'P','M','U');
- 
     end
-else
     
-    % FS(y) = y
+catch
+    
+    % otherwise FS(y) = y
     %----------------------------------------------------------------------
     y   = Y.y;
     IS  = inline([M.IS '(P,M,U)'],'P','M','U');
@@ -179,7 +193,7 @@
 nh    = length(Q);                  % number of precision components
 nt    = length(Q{1});               % number of time bins
 nq    = nr*ns/nt;                   % for compact Kronecker form of M-step
-h     = zeros(nh,1);                % initialise hyperparameters
+h     = sparse(nh,1);                % initialise hyperparameters
  
 % prior moments
 %--------------------------------------------------------------------------
@@ -200,25 +214,38 @@
 %--------------------------------------------------------------------------
 try
     hE = M.hE;
+    if length(hE) ~= nh
+        hE = hE(1) + sparse(nh,1);
+    end
 catch
     hE = sparse(nh,1);
 end
+h      = hE;
  
 % hyperpriors - covariance
 %--------------------------------------------------------------------------
 try
     ihC = inv(M.hC);
+    if length(ihC) ~= nh
+        ihC = ihC*speye(nh,nh);
+    end
 catch
-    ihC = eye(nh,nh);
+    ihC = speye(nh,nh);
 end
  
+% unpack covariance
+%--------------------------------------------------------------------------
+if isstruct(pC);
+    pC = spm_diag(spm_vec(pC));
+end
+
 % dimension reduction of parameter space
 %--------------------------------------------------------------------------
 V     = spm_svd(pC,exp(-32));
 nu    = size(dfdu,2);                 % number of parameters (confounds)
 np    = size(V,2);                    % number of parameters (effective)
-ip    = [1:np]';
-iu    = [1:nu]' + np;
+ip    = (1:np)';
+iu    = (1:nu)' + np;
  
 % second-order moments (in reduced space)
 %--------------------------------------------------------------------------
@@ -235,15 +262,17 @@
  
 % EM
 %==========================================================================
+criterion = [0 0 0 0];
+
 C.F   = -Inf;                                   % free energy
-v     = 0;                                      % log ascent rate
+v     = -2;                                     % log ascent rate
 dFdh  = zeros(nh,1);
 dFdhh = zeros(nh,nh);
-for k = 1:64
+for k = 1:128
     
     % time
     %----------------------------------------------------------------------  
-    tic;
+    tStart = tic;
  
     % M-Step: ReML estimator of variance components:  h = max{F(p,h)}
     %======================================================================
@@ -255,8 +284,8 @@
        
     % prediction error and full gradients
     %----------------------------------------------------------------------
-    e     = spm_vec(y) - spm_vec(f) - dfdu*p(iu);
-    dfdp  = reshape(spm_vec(dfdp),ns*nr,np);
+    e     =  spm_vec(y) - spm_vec(f) - dfdu*p(iu);
+    dfdp  =  reshape(spm_vec(dfdp),ns*nr,np);
     J     = -[dfdp dfdu];
     
  
@@ -270,45 +299,50 @@
         
         % precision and conditional covariance
         %------------------------------------------------------------------
-        iS    = 0;
+        iS    = sparse(0);
         for i = 1:nh
-            iS = iS + Q{i}*exp(h(i));
+            iS = iS + Q{i}*(exp(-32) + exp(h(i)));
         end
-        S     = inv(iS);
+        S     = spm_inv(iS);
         iS    = kron(speye(nq),iS);
-        Cp    = spm_inv(J'*iS*J + ipC);
-        
+        Pp    = real(J)'*iS*real(J) + imag(J)'*iS*imag(J);
+        Cp    = spm_inv(Pp + ipC);
+        if any(isnan(Cp(:))) || rcond(full(Cp)) < exp(-32), break, end
  
         % precision operators for M-Step
         %------------------------------------------------------------------
         for i = 1:nh
-            P{i}  = Q{i}*exp(h(i));
-            PS{i} = P{i}*S;
-            P{i}  = kron(speye(nq),P{i});
+            P{i}   = Q{i}*exp(h(i));
+            PS{i}  = P{i}*S;
+            P{i}   = kron(speye(nq),P{i});
+            JPJ{i} = real(J)'*P{i}*real(J) + ...
+                     imag(J)'*P{i}*imag(J);
         end
-            
-            
+                    
         % derivatives: dLdh = dL/dh,...
         %------------------------------------------------------------------
         for i = 1:nh
-            dFdh(i,1)      =  trace(PS{i})*nq/2 - e'*P{i}*e/2 ...
-                             -sum(sum(Cp.*(J'*P{i}*J)))/2;
+            dFdh(i,1)      =   trace(PS{i})*nq/2 ...
+                             - real(e)'*P{i}*real(e)/2 ...
+                             - imag(e)'*P{i}*imag(e)/2 ...
+                             - sum(sum(Cp.*JPJ{i}))/2;
             for j = i:nh
-                dFdhh(i,j) = -sum(sum(PS{i}.*PS{j}))*nq/2;
-                dFdhh(j,i) =  dFdhh(i,j);
+                dFdhh(i,j) = - sum(sum(PS{i}.*PS{j}))*nq/2;
+                dFdhh(j,i) =   dFdhh(i,j);
             end
         end
         
         % add hyperpriors
         %------------------------------------------------------------------
-        d     = h - hE;
+        d     = h     - hE;
         dFdh  = dFdh  - ihC*d;
         dFdhh = dFdhh - ihC;
-        Ch    = inv(-dFdhh); 
+        Ch    = spm_inv(-dFdhh); 
         
         % update ReML estimate
         %------------------------------------------------------------------
-        dh    = spm_dx(dFdhh,dFdh,{8});
+        dh    = spm_dx(dFdhh,dFdh,{4});
+        dh    = min(max(dh,-1),1);
         h     = h  + dh;
  
         % convergence
@@ -317,13 +351,15 @@
         if dF < 1e-2, break, end
  
     end
- 
+
+    
     % E-Step with Levenberg-Marquardt regularization
     %======================================================================
  
     % objective function: F(p) (= log evidence - divergence)
     %----------------------------------------------------------------------
-    F = - e'*iS*e/2 ...
+    F = - real(e)'*iS*real(e)/2 ...
+        - imag(e)'*iS*imag(e)/2 ...
         - p'*ipC*p/2 ...
         - d'*ihC*d/2 ...
         - ns*nr*log(8*atan(1))/2 ...
@@ -335,7 +371,7 @@
     %----------------------------------------------------------------------
     try
         F0; 
-        fprintf(' actual: %.3e (%.2f sec)\n',full(F - C.F),toc)
+        fprintf(' actual: %.3e (%.2f sec)\n',full(F - C.F),toc(tStart))
     catch
         F0 = F;
     end
@@ -353,8 +389,10 @@
         
         % E-Step: Conditional update of gradients and curvature
         %------------------------------------------------------------------
-        dFdp  = -J'*iS*e - ipC*p;
-        dFdpp = -J'*iS*J - ipC;
+        dFdp  = -real(J)'*iS*real(e) ...
+                -imag(J)'*iS*imag(e) - ipC*p;
+        dFdpp = -real(J)'*iS*real(J) ...
+                -imag(J)'*iS*imag(J) - ipC;
         
         % decrease regularization
         %------------------------------------------------------------------
@@ -393,7 +431,7 @@
         % subplot prediction
         %------------------------------------------------------------------
         figure(Fsi)
-        x    = [1:ns]*Y.dt;
+        x    = (1:ns)*Y.dt;
         xLab = 'time (seconds)';
         try
             if length(M.Hz) == ns
@@ -402,13 +440,35 @@
             end
         end
         
-        subplot(2,1,1)
-        plot(x,f), hold on
-        plot(x,f + spm_unvec(e,f),':'), hold off
-        xlabel(xLab)
-        title(sprintf('%s: %i','prediction and response: E-Step',k))
-        grid on
- 
+        if isreal(f)
+            
+            subplot(2,1,1)
+            plot(x,f), hold on
+            plot(x,f + spm_unvec(e,f),':'), hold off
+            xlabel(xLab)
+            title(sprintf('%s: %i','prediction and response: E-Step',k))
+            grid on
+            
+        else
+            
+            subplot(2,2,1)
+            plot(x,real(f)), hold on
+            plot(x,real(f + spm_unvec(e,f)),':'), hold off
+            xlabel(xLab)
+            ylabel('real')
+            title(sprintf('%s: %i','prediction and response: E-Step',k))
+            grid on
+            
+            subplot(2,2,2)
+            plot(x,imag(f)), hold on
+            plot(x,imag(f + spm_unvec(e,f)),':'), hold off
+            xlabel(xLab)
+            ylabel('imaginary')
+            title(sprintf('%s: %i','prediction and response: E-Step',k))
+            grid on
+            
+        end
+        
         % subplot parameters
         %------------------------------------------------------------------
         subplot(2,1,2)
@@ -424,10 +484,9 @@
     %----------------------------------------------------------------------
     dF  = dFdp'*dp;
     fprintf('%-6s: %i %6s %-6.3e %6s %.3e ',str,k,'F:',full(C.F - F0),'dF predicted:',full(dF))
-    if k > 2 && dF < 1e-2
-        fprintf(' convergence\n')
-        break
-    end
+    
+    criterion = [(dF < 1e-2) criterion(1:end - 1)];
+    if all(criterion), fprintf(' convergence\n'), break, end
  
 end
  
diff --git a/spm_nlsi_LS.m b/spm_nlsi_LS.m
new file mode 100644
index 0000000..5563931
--- /dev/null
+++ b/spm_nlsi_LS.m
@@ -0,0 +1,381 @@
+function [Ep,qC,qh,F] = spm_nlsi_LS(M,U,Y)
+% Bayesian inversion of a nonlinear model using (Laplacian) sampling
+% FORMAT [Ep,Cp,Eh,F] = spm_nlsi_LS(M,U,Y)
+%
+% Dynamical MIMO models
+%__________________________________________________________________________
+%
+% M.IS - function name f(P,M,U) - generative model
+%        This function specifies the nonlinear model:
+%        y = Y.y = IS(P,M,U) + X0*P0 + e
+%        were e ~ N(0,C).  For dynamic systems this would be an integration
+%        scheme (e.g. spm_int). spm_int expects the following:
+%
+%     M.f  - f(x,u,P,M)
+%     M.g  - g(x,u,P,M)
+%       x  - state variables
+%       u  - inputs or causes
+%       P  - free parameters
+%       M  - fixed functional forms and parameters in M
+%
+% M.FS - function name f(y,M)   - feature selection
+%        This [optional] function performs feature selection assuming the
+%        generalized model y = FS(y,M) = FS(IS(P,M,U),M) + X0*P0 + e
+%
+% M.P  - starting estimates for model parameters [optional]
+%
+% M.pE - prior expectation  - E{P}   of model parameters
+% M.pC - prior covariance   - Cov{P} of model parameters
+%
+% M.hE - prior expectation  - E{h}   of log-precision parameters
+% M.hC - prior covariance   - Cov{h} of log-precision parameters
+%
+% U.u  - inputs
+% U.dt - sampling interval
+%
+% Y.y  - outputs
+% Y.dt - sampling interval for outputs
+% Y.X0 - Confounds or null space      (over size(y,1) bins or all vec(y))
+% Y.Q  - q error precision components (over size(y,1) bins or all vec(y))
+%
+%
+% Parameter estimates
+%--------------------------------------------------------------------------
+% Ep  - (p x 1)         conditional expectation    E{P|y}
+% Cp  - (p x p)         conditional covariance     Cov{P|y}
+% Eh  - (q x 1)         conditional log-precisions E{h|y}
+%
+% log evidence
+%--------------------------------------------------------------------------
+% F   - [-ve] free energy F = log evidence = p(y|f,g,pE,pC) = p(y|m)
+%
+%__________________________________________________________________________
+% Returns the moments of the posterior p.d.f. of the parameters of a
+% nonlinear model specified by IS(P,M,U) under Gaussian assumptions.
+% Usually, IS is an integrator of a dynamic MIMO input-state-output model
+%
+%              dx/dt = f(x,u,P)
+%              y     = g(x,u,P)  + X0*P0 + e
+%
+% A static nonlinear observation model with fixed input or causes u
+% obtains when x = []. i.e.
+%
+%              y     = g([],u,P) + X0*P0e + e
+%
+% but static nonlinear models are specified more simply using
+%
+%              y     = IS(P,M,U) + X0*P0 + e
+%
+% Priors on the free parameters P are specified in terms of expectation pE
+% and covariance pC.
+%
+% For generic aspects of the scheme see:
+%
+% Friston K, Mattout J, Trujillo-Barreto N, Ashburner J, Penny W.
+% Variational free energy and the Laplace approximation.
+% NeuroImage. 2007 Jan 1;34(1):220-34.
+%
+% This scheme handels complex data along the lines originally described in:
+%
+% Sehpard RJ, Lordan BP, and Grant EH.
+% Least squares analysis of complex data with applications to permittivity
+% measurements.
+% J. Phys. D. Appl. Phys 1970 3:1759-1764.
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_nlsi_LS.m 4098 2010-10-22 19:46:28Z karl $
+
+% figure (unless disabled)
+%--------------------------------------------------------------------------
+try
+    M.nograph;
+catch
+    M.nograph = 0;
+end
+if ~M.nograph
+    Fsi = spm_figure('GetWin','SI');
+end
+
+% check integrator
+%--------------------------------------------------------------------------
+try
+    M.IS;
+catch
+    M.IS = 'spm_int';
+end
+
+% composition of feature selection and prediction (usually an integrator)
+%--------------------------------------------------------------------------
+if isfield(M,'FS')
+    
+    % FS(y,M)
+    %----------------------------------------------------------------------
+    try
+        y  = feval(M.FS,Y.y,M);
+        IS = inline([M.FS '(' M.IS '(P,M,U),M)'],'P','M','U');
+        
+        % FS(y,M)
+        %----------------------------------------------------------------------
+    catch
+        y  = feval(M.FS,Y.y);
+        IS = inline([M.FS '(' M.IS '(P,M,U))'],'P','M','U');
+        
+    end
+else
+    
+    % FS(y) = y
+    %----------------------------------------------------------------------
+    y   = Y.y;
+    IS  = inline([M.IS '(P,M,U)'],'P','M','U');
+end
+
+% size of data (usually samples x channels)
+%--------------------------------------------------------------------------
+if iscell(y)
+    ns = size(y{1},1);
+else
+    ns = size(y,1);
+end
+nr   = length(spm_vec(y))/ns;       % number of samples and responses
+M.ns = ns;                          % store in M.ns for integrator
+
+% initial states
+%--------------------------------------------------------------------------
+try
+    M.x;
+catch
+    if ~isfield(M,'n'), M.n = 0;    end
+    M.x = sparse(M.n,1);
+end
+
+% input
+%--------------------------------------------------------------------------
+try
+    U;
+catch
+    U = [];
+end
+
+% initial parameters
+%--------------------------------------------------------------------------
+try
+    spm_vec(M.P) - spm_vec(M.pE);
+    fprintf('\nParameter initialisation successful\n')
+catch
+    M.P = M.pE;
+end
+
+% time-step
+%--------------------------------------------------------------------------
+try
+    Y.dt;
+catch
+    Y.dt = 1;
+end
+
+% precision components Q
+%--------------------------------------------------------------------------
+try
+    Q = Y.Q;
+    if isnumeric(Q), Q = {Q}; end
+catch
+    Q = spm_Ce(ns*ones(1,nr));
+end
+nh    = length(Q);                  % number of precision components
+nt    = length(Q{1});               % number of time bins
+nq    = nr*ns/nt;                   % for compact Kronecker form of M-step
+h     = zeros(nh,1);                % initialise hyperparameters
+
+
+
+% confounds (if specified)
+%--------------------------------------------------------------------------
+try
+    nb   = size(Y.X0,1);            % number of bins
+    nx   = nr*ns/nb;                % number of blocks
+    dfdu = kron(speye(nx,nx),Y.X0);
+catch
+    dfdu = sparse(ns*nr,0);
+end
+
+% hyperpriors - expectation
+%--------------------------------------------------------------------------
+try
+    hE = M.hE;
+    if length(hE) ~= nh
+        hE = hE*sparse(nh,1);
+    end
+catch
+    hE = sparse(nh,1);
+end
+
+% prior moments
+%--------------------------------------------------------------------------
+pE    = M.pE;
+pC    = M.pC;
+nu    = size(dfdu,2);                 % number of parameters (confounds)
+np    = size(pC,2);                   % number of parameters (effective)
+
+% second-order moments (in reduced space)
+%--------------------------------------------------------------------------
+ipC   = spm_inv(pC);
+
+
+% initialize conditional density
+%--------------------------------------------------------------------------
+Eu    = spm_pinv(dfdu)*spm_vec(y);
+Ep    = pE;
+
+% precision and conditional covariance
+%------------------------------------------------------------------
+iS    = sparse(0);
+for i = 1:nh
+    iS = iS + Q{i}*(exp(-16) + exp(hE(i)));
+end
+S     = spm_inv(iS);
+iS    = kron(speye(nq),iS);
+qS    = spm_sqrtm(pC/32);
+qE    = spm_vec(pE);
+pE    = spm_vec(pE);
+y     = spm_vec(y);
+np    = length(qE);
+
+
+% Sampling
+%==========================================================================
+Gmax  = -Inf;
+for k = 1:64
+    
+    % time
+    %----------------------------------------------------------------------
+    tic;
+    
+    % Gibb's sampling
+    %======================================================================
+    for i = 1:128
+        
+        % prediction
+        %------------------------------------------------------------------
+        P(:,i) = qE + qS*randn(np,1);
+        R(:,i) = spm_vec(feval(IS,spm_unvec(P(:,i),M.pE),M,U));
+
+        % prediction error
+        %------------------------------------------------------------------
+        ey     = R(:,i) - y;
+        ep     = P(:,i) - pE;
+        
+        % Gibb's energy
+        %------------------------------------------------------------------
+        qh     = real(ey')*iS*real(ey) + imag(ey)'*iS*imag(ey);
+        G(i,1) = - ns*log(qh)/2 - ep'*ipC*ep/2;
+        
+        
+        % conditional mode
+        %----------------------------------------------------------------------
+        [maxG j] = max(G);
+        if maxG  > Gmax
+            qE   = P(:,j);
+            f    = R(:,j);
+            Gmax = maxG;
+        end
+        pE       = qE;
+        
+        disp(i)
+    end
+    
+    
+    % conditional dispersion
+    %----------------------------------------------------------------------
+    q     = exp((G - maxG));
+    q     = q/sum(q);
+    for i = 1:np
+        for j = 1:np
+            qC(i,j) = ((P(i,:) - qE(i)).*(P(j,:) - qE(j)))*q;
+        end
+    end
+    qS    = spm_sqrtm(qC);
+      
+    
+    % objective function:
+    %======================================================================
+    F     = Gmax + spm_logdet(ipC*qC)/2;
+    F     = Gmax;
+    
+    
+    % graphics
+    %----------------------------------------------------------------------
+    try
+        
+        % reshape prediction if necessary
+        %------------------------------------------------------------------
+        f  = reshape(f,ns,nr);
+        d  = reshape(y,ns,nr);
+        
+        % subplot prediction
+        %------------------------------------------------------------------
+        figure(Fsi)
+        x    = (1:ns)*Y.dt;
+        xLab = 'time (seconds)';
+        try
+            if length(M.Hz) == ns
+                x    = Y.Hz;
+                xLab = 'Frequency (Hz)';
+            end
+        end
+        
+        if isreal(f)
+            subplot(2,1,1)
+            plot(x,f,x,d,':')
+            xlabel(xLab)
+            title(sprintf('%s: %i','prediction and response: E-Step',k))
+            grid on
+            
+        else
+            subplot(2,2,1)
+            plot(x,real(f),x,real(d),':')
+            xlabel(xLab)
+            ylabel('real')
+            title(sprintf('%s: %i','prediction and response: E-Step',k))
+            grid on
+            
+            subplot(2,2,2)
+            plot(x,imag(f),x,imag(d),':')
+            xlabel(xLab)
+            ylabel('imaginary')
+            title(sprintf('%s: %i','prediction and response: E-Step',k))
+            grid on
+        end
+        
+        % subplot Gibb's smapling
+        %------------------------------------------------------------------
+        subplot(2,2,3)
+        plot(G)
+        xlabel('smaple')
+        title('Gibbs energy')
+    
+        % subplot parameters
+        %------------------------------------------------------------------
+        subplot(2,2,4)
+        bar(full(qE - spm_vec(M.pE)))
+        xlabel('parameter')
+        title('conditional expectation')
+        grid on
+        drawnow
+        
+    end
+    
+    % convergence
+    %----------------------------------------------------------------------
+    try, dF = F - Fk; catch, dF = 0; end
+    Fk      = F;
+    fprintf('%-6s: %i %6s %-6.3e %6s %.3e (%.2f)\n','LS',k,'F:',full(F),'dF:',full(dF),toc)
+    if k > 4 && dF < 1e-4
+        break
+    end  
+end
+
+
+
diff --git a/spm_nlsi_N.m b/spm_nlsi_N.m
index bc9ab95..af2d269 100644
--- a/spm_nlsi_N.m
+++ b/spm_nlsi_N.m
@@ -83,7 +83,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_nlsi_N.m 3696 2010-01-22 14:22:31Z karl $
+% $Id: spm_nlsi_N.m 4281 2011-03-31 19:49:57Z karl $
  
 % figure (unless disabled)
 %--------------------------------------------------------------------------
@@ -201,7 +201,7 @@
 nh    = length(Q);          % number of precision components
 nt    = length(Q{1});       % number of time bins
 nq    = nr*ns/nt;           % for compact Kronecker form of M-step
-h     = zeros(nh,1);        % initialize hyperparameters
+
  
  
 % confounds (if specified)
@@ -217,17 +217,23 @@
 try
     hE  = M.hE;
 catch
-    hE  = sparse(nh,1) - log(mean(var(y)));
+    hE  = sparse(nh,1) - log(var(spm_vec(y))) + 3;
 end
+h       =  hE;              % initialize hyperparameters
  
 % hyperpriors - covariance
 %--------------------------------------------------------------------------
 try
-    ihC = inv(M.hC);
+    ihC = spm_inv(M.hC);
 catch
-    ihC = speye(nh,nh)*8;
+    ihC = speye(nh,nh)*exp(8);
 end
- 
+
+% unpack prior covariances
+%--------------------------------------------------------------------------
+if isstruct(M.pC); M.pC = spm_diag(spm_vec(M.pC)); end
+if isstruct(M.gC); M.gC = spm_diag(spm_vec(M.gC)); end
+
 % dimension reduction of parameter space
 %--------------------------------------------------------------------------
 Vp    = spm_svd(M.pC,exp(-32));
@@ -248,9 +254,9 @@
 pC    = Vp'*M.pC*Vp;
 gC    = Vg'*M.gC*Vg;
 uC    = speye(nu,nu)*exp(32);
-ipC   = inv(pC);                               % p - state parameters
-igC   = inv(gC);                               % g - observer parameters
-iuC   = inv(uC);                               % u - fixed parameters
+ipC   = spm_inv(pC);                               % p - state parameters
+igC   = spm_inv(gC);                               % g - observer parameters
+iuC   = spm_inv(uC);                               % u - fixed parameters
 ibC   = spm_cat(spm_diag({ipC,igC,iuC}));      % all parameters
  
 % initialize conditional density
@@ -330,7 +336,7 @@
         % Optimize F(h): parameters of iS(h)
         %==================================================================
         dgdb   = [dgdp dgdg dgdu];           
-        for ih = 1:8
+        for ih = 1:16
  
             % precision
             %--------------------------------------------------------------
@@ -338,10 +344,10 @@
             for i = 1:nh
                 iS = iS + Q{i}*exp(h(i));
             end
-            S     = inv(iS);
+            S     = spm_inv(iS);
             iS    = kron(speye(nq),iS);
             dFdbb = dgdb'*iS*dgdb + ibC;
-            Cb    = inv(dFdbb);
+            Cb    = spm_inv(dFdbb);
             
             % precision operators for M-Step
             %--------------------------------------------------------------
@@ -367,7 +373,7 @@
             eh    = h     - hE;
             dFdh  = dFdh  - ihC*eh;
             dFdhh = dFdhh - ihC;
-            Ch    = inv(-dFdhh);
+            Ch    = spm_inv(-dFdhh);
             
             % M-Step: update ReML estimate of h
             %--------------------------------------------------------------
@@ -402,7 +408,7 @@
         %------------------------------------------------------------------
         dg    = spm_dx(dFdgg,dFdg,{8});
         Eg    = spm_unvec(spm_vec(Eg) + Vg*dg,Eg);
- 
+         
         % convergence
         %------------------------------------------------------------------
         dG    = dFdg'*dg;
@@ -443,7 +449,7 @@
  
         % decrease regularization
         %------------------------------------------------------------------
-        v     = min(v + 1/2,4);
+        v     = min(v + 1/2,8);
         str   = 'EM(+)';
  
         % accept current estimates
@@ -468,7 +474,7 @@
  
         % and increase regularization
         %------------------------------------------------------------------
-        v     = min(v - 2,-4);
+        v     = min(v - 2,0);
         str   = 'EM(-)';
  
     end
diff --git a/spm_normalise.m b/spm_normalise.m
index 1013955..835cef1 100644
--- a/spm_normalise.m
+++ b/spm_normalise.m
@@ -120,7 +120,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_normalise.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: spm_normalise.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 if nargin<2, error('Incorrect usage.'); end;
@@ -140,15 +140,15 @@
 if ischar(VWF), VWF=spm_vol(VWF); end;                                                                     
 
 
-def_flags = struct('smosrc',8,'smoref',0,'regtype','mni',...
-    'cutoff',30,'nits',16,'reg',0.1,'graphics',1);
+def_flags          = spm_get_defaults('normalise.estimate');
+def_flags.graphics = 1;
 if nargin < 6,
     flags = def_flags;
 else
     fnms  = fieldnames(def_flags);
     for i=1:length(fnms),
         if ~isfield(flags,fnms{i}),
-            flags = setfield(flags,fnms{i},getfield(def_flags,fnms{i}));
+            flags.(fnms{i}) = def_flags.(fnms{i});
         end;
     end;
 end;
@@ -172,7 +172,7 @@
 aflags.sep = max(aflags.sep,max(sqrt(sum(VF(1).mat(1:3,1:3).^2))));
 
 M         = eye(4); %spm_matrix(prms');
-spm_chi2_plot('Init','Affine Registration','Mean squared difference','Iteration');
+spm_plot_convergence('Init','Affine Registration','Mean squared difference','Iteration');
 [M,scal]  = spm_affreg(VG1, VF1, aflags, M);
  
 fprintf('Fine Affine Registration..\n');
@@ -181,7 +181,7 @@
 aflags.sep = aflags.sep/2;
 [M,scal]   = spm_affreg(VG1, VF1, aflags, M,scal);
 Affine     = inv(VG(1).mat\M*VF1(1).mat);
-spm_chi2_plot('Clear');
+spm_plot_convergence('Clear');
 
 % Basis function Normalisation
 %-----------------------------------------------------------------------
@@ -211,7 +211,7 @@
 if isfield(VG,'dat'), VG = rmfield(VG,'dat'); end;
 if ~isempty(matname),
     fprintf('Saving Parameters..\n');
-    if spm_matlab_version_chk('7') >= 0,
+    if spm_check_version('matlab','7') >= 0,
         save(matname,'-V6','Affine','Tr','VF','VG','flags');
     else
         save(matname,'Affine','Tr','VF','VG','flags');
diff --git a/spm_orthviews.m b/spm_orthviews.m
index 1527357..228f82c 100644
--- a/spm_orthviews.m
+++ b/spm_orthviews.m
@@ -1,17 +1,13 @@
 function varargout = spm_orthviews(action,varargin)
-% Display Orthogonal Views of a Normalized Image
+% Display orthogonal views of a set of images
 % FORMAT H = spm_orthviews('Image',filename[,position])
 % filename - name of image to display
-% area     - position of image
+% area     - position of image {relative}
 %            -  area(1) - position x
 %            -  area(2) - position y
 %            -  area(3) - size x
 %            -  area(4) - size y
 % H        - handle for ortho sections
-% FORMAT spm_orthviews('BB',bb)
-% bb       - bounding box
-%            [loX loY loZ
-%             hiX hiY hiZ]
 %
 % FORMAT spm_orthviews('Redraw')
 % Redraws the images
@@ -25,11 +21,33 @@
 %            of a view)
 % with no arguments - puts things into mm space
 %
+% FORMAT spm_orthviews('BB',bb)
+% bb       - bounding box
+%            [loX loY loZ
+%             hiX hiY hiZ]
+%
 % FORMAT spm_orthviews('MaxBB')
 % sets the bounding box big enough display the whole of all images
 %
-% FORMAT spm_orthviews('Resolution',res)
+% FORMAT spm_orthviews('Resolution'[,res])
+% res      - resolution (mm)
+% sets the sampling resolution for all images. The effective resolution
+% will be the minimum of res and the voxel sizes of all images. If no
+% resolution is specified, the minimum of 1mm and the voxel sizes of the
+% images is used.
+%
+% FORMAT spm_orthviews('Zoom'[,fov[,res]])
+% fov      - half width of field of view (mm)
 % res      - resolution (mm)
+% sets the displayed part and sampling resolution for all images. The
+% image display will be centered at the current crosshair position. The
+% image region [xhairs-fov xhairs+fov] will be shown.
+% If no argument is given or fov == Inf, the image display will be reset to
+% "Full Volume". If fov == 0, the image will be zoomed to the bounding box
+% from spm_get_bbox for the non-zero voxels of the image. If fov is NaN,
+% then a threshold can be entered, and spm_get_bbox will be used to derive
+% the bounding box of the voxels above this threshold.
+% Optionally, the display resolution can be set as well.
 %
 % FORMAT spm_orthviews('Delete', handle)
 % handle   - image number to delete
@@ -54,7 +72,7 @@
 % FORMAT spm_orthviews('Interp',hld)
 % sets the hold value to hld (see spm_slice_vol).
 %
-% FORMAT spm_orthviews('AddBlobs',handle,XYZ,Z,mat)
+% FORMAT spm_orthviews('AddBlobs',handle,XYZ,Z,mat,name)
 % Adds blobs from a pointlist to the image specified by the handle(s).
 % handle   - image number to add blobs to
 % XYZ      - blob voxel locations
@@ -77,44 +95,54 @@
 % blobs print well.
 %
 % FORMAT spm_orthviews('AddColourBar',handle,blobno)
-% Adds colourbar for a specified blob set
+% Adds colourbar for a specified blob set.
 % handle   - image number
 % blobno   - blob number
 %
 % FORMAT spm_orthviews('RemoveBlobs',handle)
 % Removes all blobs from the image specified by the handle(s).
 %
+% FORMAT spm_orthviews('Addtruecolourimage',handle,filename,colourmap,prop,mx,mn)
+% Adds blobs from an image in true colour.
+% handle   - image number to add blobs to [default: 1]
+% filename - image containing blob data [default: request via GUI]
+% colourmap - colormap to display blobs in [default: GUI input]
+% prop     - intensity proportion of activation cf grayscale [default: 0.4]
+% mx       - maximum intensity to scale to [maximum value in activation image]
+% mn       - minimum intensity to scale to [minimum value in activation image]
+%
 % FORMAT spm_orthviews('Register',hReg)
-% hReg      - Handle of HandleGraphics object to build registry in.
+% hReg     - Handle of HandleGraphics object to build registry in.
 % See spm_XYZreg for more information.
 %
-% spm_orthviews('AddContext',handle)
+% FORMAT spm_orthviews('AddContext',handle)
 % handle   - image number to add context menu to
 %
-% spm_orthviews('RemoveContext',handle)
+% FORMAT spm_orthviews('RemoveContext',handle)
 % handle   - image number to remove context menu from
 %
-% CONTEXT MENU
-% spm_orthviews offers many of its features in a context menu, which is
-% accessible via the right mouse button in each displayed image.
+% FORMAT spm_orthviews('ZoomMenu',zoom,res)
+% FORMAT [zoom, res] = spm_orthviews('ZoomMenu')
+% zoom     - A list of predefined zoom values
+% res      - A list of predefined resolutions
+% This list is used by spm_image and spm_orthviews('addcontext',...) to
+% create the 'Zoom' menu. The values can be retrieved by calling
+% spm_orthviews('ZoomMenu') with 2 output arguments. Values of 0, NaN and
+% Inf are treated specially, see the help for spm_orthviews('Zoom' ...).
+%__________________________________________________________________________
 %
 % PLUGINS
-% The display capabilities of spm_orthviews can be extended with
-% plugins. These are located in the spm_orthviews subdirectory of the SPM
-% distribution. Currently there are 3 plugins available:
-% quiver    Add Quiver plots to a displayed image
-% quiver3d  Add 3D Quiver plots to a displayed image
-% roi       ROI creation and modification
+% The display capabilities of spm_orthviews can be extended with plugins.
+% These are located in the spm_orthviews subdirectory of the SPM
+% distribution.
 % The functionality of plugins can be accessed via calls to
 % spm_orthviews('plugin_name', plugin_arguments). For detailed descriptions
 % of each plugin see help spm_orthviews/spm_ov_'plugin_name'.
-%
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% John Ashburner, Matthew Brett, Tom Nichols and Volkmar Glauche
-% $Id: spm_orthviews.m 3756 2010-03-05 18:43:37Z guillaume $
-
+% John Ashburner et al
+% $Id: spm_orthviews.m 4276 2011-03-31 11:25:34Z spm $
 
 
 % The basic fields of st are:
@@ -142,39 +170,39 @@
 %
 % When a new image is displayed, the cell entry contains the information
 % returned by spm_vol (type help spm_vol for more info).  In addition,
-% there are a few other fields, some of which I will document here:
+% there are a few other fields, some of which are documented here:
 %
-%         premul - a matrix to premultiply the .mat field by.  Useful
-%                  for re-orienting images.
-%         window - either 'auto' or an intensity range to display the
-%                  image with.
-%         mapping- Mapping of image intensities to grey values. Currently
-%                  one of 'linear', 'histeq', loghisteq',
-%                  'quadhisteq'. Default is 'linear'.
-%                  Histogram equalisation depends on the image toolbox
-%                  and is only available if there is a license available
-%                  for it.
-%         ax     - a cell array containing an element for the three
-%                  views.  The fields of each element are handles for
-%                  the axis, image and crosshairs.
+%         premul  - a matrix to premultiply the .mat field by.  Useful
+%                   for re-orienting images.
+%         window  - either 'auto' or an intensity range to display the
+%                   image with.
+%         mapping - Mapping of image intensities to grey values. Currently
+%                   one of 'linear', 'histeq', loghisteq',
+%                   'quadhisteq'. Default is 'linear'.
+%                   Histogram equalisation depends on the image toolbox
+%                   and is only available if there is a license available
+%                   for it.
+%         ax      - a cell array containing an element for the three
+%                   views.  The fields of each element are handles for
+%                   the axis, image and crosshairs.
 %
-%         blobs - optional.  Is there for using to superimpose blobs.
-%                 vol     - 3D array of image data
-%                 mat     - a mapping from vox-to-mm (see spm_vol, or
-%                           help on image formats).
-%                 max     - maximum intensity for scaling to.  If it
-%                           does not exist, then images are auto-scaled.
+%         blobs   - optional.  Is there for using to superimpose blobs.
+%                   vol     - 3D array of image data
+%                   mat     - a mapping from vox-to-mm (see spm_vol, or
+%                             help on image formats).
+%                   max     - maximum intensity for scaling to.  If it
+%                             does not exist, then images are auto-scaled.
 %
-%                 There are two colouring modes: full colour, and split
-%                 colour.  When using full colour, there should be a
-%                 'colour' field for each cell element.  When using
-%                 split colourscale, there is a handle for the colorbar
-%                 axis.
+%                   There are two colouring modes: full colour, and split
+%                   colour.  When using full colour, there should be a
+%                   'colour' field for each cell element.  When using
+%                   split colourscale, there is a handle for the colorbar
+%                   axis.
 %
-%                 colour  - if it exists it contains the
-%                           red,green,blue that the blobs should be
-%                           displayed in.
-%                 cbar    - handle for colorbar (for split colourscale).
+%                   colour  - if it exists it contains the
+%                             red,green,blue that the blobs should be
+%                             displayed in.
+%                   cbar    - handle for colorbar (for split colourscale).
 %
 % PLUGINS
 % The plugin concept has been developed to extend the display capabilities
@@ -188,7 +216,7 @@
 %                st.plugins{:}.
 %                The subfunction 'add_context' calls each plugin with
 %                feval(['spm_ov_', st.plugins{k}], ...
-%             'context_menu', i, parent_menu)
+%                  'context_menu', i, parent_menu)
 %                Each plugin may add its own submenu to the context
 %                menu.
 % b) at redraw:  After images and blobs of st.vols{i} are drawn, the
@@ -196,7 +224,7 @@
 %                the plugin list st.plugins{:}. For each matching entry, the
 %                corresponding plugin is called with the command 'redraw':
 %                feval(['spm_ov_', st.plugins{k}], ...
-%             'redraw', i, TM0, TD, CM0, CD, SM0, SD);
+%                  'redraw', i, TM0, TD, CM0, CD, SM0, SD);
 %                The values of TM0, TD, CM0, CD, SM0, SD are defined in the
 %                same way as in the redraw subfunction of spm_orthviews.
 %                It is up to the plugin to do all necessary redraw
@@ -208,29 +236,38 @@
 
 global st;
 
-if isempty(st), reset_st; end;
+persistent zoomlist;
+persistent reslist;
 
-spm('Pointer','watch');
+if isempty(st), reset_st; end
 
-if nargin == 0, action = ''; end;
-action = lower(action);
+if nargin == 0, action = ''; end
 
-switch lower(action),
+if ~any(strcmpi(action,{'reposition','pos'}))
+    spm('Pointer','Watch');
+end
+    
+switch lower(action)
     case 'image',
         H = specify_image(varargin{1});
         if ~isempty(H)
-            st.vols{H}.area = [0 0 1 1];
-            if numel(varargin)>=2, st.vols{H}.area = varargin{2}; end;
-            if isempty(st.bb), st.bb = maxbb; end;
+            if numel(varargin)>=2
+                st.vols{H}.area = varargin{2};
+            else
+                st.vols{H}.area = [0 0 1 1];
+            end
+            if isempty(st.bb), st.bb = maxbb; end
+            resolution;
             bbox;
             cm_pos;
-        end;
+        end
         varargout{1} = H;
-        st.centre    = mean(maxbb);
+        mmcentre     = mean(st.Space*[maxbb';1 1],2)';
+        st.centre    = mmcentre(1:3);
         redraw_all
         
     case 'bb',
-        if ~isempty(varargin) && all(size(varargin{1})==[2 3]), st.bb = varargin{1}; end;
+        if ~isempty(varargin) && all(size(varargin{1})==[2 3]), st.bb = varargin{1}; end
         bbox;
         redraw_all;
         
@@ -239,11 +276,11 @@
         eval(st.callback);
         if isfield(st,'registry'),
             spm_XYZreg('SetCoords',st.centre,st.registry.hReg,st.registry.hMe);
-        end;
+        end
         
     case 'reposition',
         if isempty(varargin), tmp = findcent;
-        else tmp = varargin{1}; end;
+        else tmp = varargin{1}; end
         if numel(tmp) == 3
             h = valid_handles(st.snap);
             if ~isempty(h)
@@ -254,7 +291,7 @@
         end
         redraw_all;
         eval(st.callback);
-        if isfield(st,'registry'),
+        if isfield(st,'registry')
             spm_XYZreg('SetCoords',st.centre,st.registry.hReg,st.registry.hMe);
         end
         cm_pos;
@@ -267,16 +304,18 @@
         cm_pos;
         
     case 'space',
-        if numel(varargin)<1,
+        if numel(varargin)<1
             st.Space = eye(4);
             st.bb = maxbb;
+            resolution;
             bbox;
             redraw_all;
         else
             space(varargin{:});
+            resolution;
             bbox;
             redraw_all;
-        end;
+        end
         
     case 'maxbb',
         st.bb = maxbb;
@@ -284,19 +323,19 @@
         redraw_all;
         
     case 'resolution',
-        resolution(varargin{1});
+        resolution(varargin{:});
         bbox;
         redraw_all;
         
     case 'window',
-        if numel(varargin)<2,
+        if numel(varargin)<2
             win = 'auto';
-        elseif numel(varargin{2})==2,
+        elseif numel(varargin{2})==2
             win = varargin{2};
-        end;
-        for i=valid_handles(varargin{1}),
+        end
+        for i=valid_handles(varargin{1})
             st.vols{i}.window = win;
-        end;
+        end
         redraw(varargin{1});
         
     case 'delete',
@@ -310,11 +349,11 @@
         my_reset;
         
     case 'pos',
-        if isempty(varargin),
+        if isempty(varargin)
             H = st.centre(:);
         else
             H = pos(varargin{1});
-        end;
+        end
         varargout{1} = H;
         
     case 'interp',
@@ -344,15 +383,6 @@
         % redraw(varargin{1});
         
     case 'addtruecolourimage',
-        % spm_orthviews('Addtruecolourimage',handle,filename,colourmap,prop,mx,mn)
-        % Adds blobs from an image in true colour
-        % handle   - image number to add blobs to [default 1]
-        % filename of image containing blob data [default - request via GUI]
-        % colourmap - colormap to display blobs in [GUI input]
-        % prop - intensity proportion of activation cf grayscale [0.4]
-        % mx   - maximum intensity to scale to [maximum value in activation image]
-        % mn   - minimum intensity to scale to [minimum value in activation image]
-        %
         if nargin < 2
             varargin(1) = {1};
         end
@@ -384,24 +414,24 @@
     case 'addcolourbar',
         addcolourbar(varargin{1}, varargin{2});
         
-    case 'rmblobs',
+    case {'removeblobs','rmblobs'},
         rmblobs(varargin{1});
         redraw(varargin{1});
         
     case 'addcontext',
-        if nargin == 1,
+        if nargin == 1
             handles = 1:24;
         else
             handles = varargin{1};
-        end;
+        end
         addcontexts(handles);
         
-    case 'rmcontext',
-        if nargin == 1,
+    case {'removecontext','rmcontext'},
+        if nargin == 1
             handles = 1:24;
         else
             handles = varargin{1};
-        end;
+        end
         rmcontexts(handles);
         
     case 'context_menu',
@@ -412,17 +442,42 @@
             handles = 1:24;
         else
             handles = varargin{1};
-        end;
+        end
         varargout{1} = valid_handles(handles);
+
+    case 'zoom',
+        zoom_op(varargin{:});
+        
+    case 'zoommenu',
+        if isempty(zoomlist)
+            zoomlist = [NaN 0 5    10  20 40 80 Inf];
+            reslist  = [1   1 .125 .25 .5 .5 1  1  ];
+        end
+        if nargin >= 3
+            if all(cellfun(@isnumeric,varargin(1:2))) && ...
+                    numel(varargin{1})==numel(varargin{2})
+                zoomlist = varargin{1}(:);
+                reslist  = varargin{2}(:);
+            else
+                warning('spm_orthviews:zoom',...
+                        'Invalid zoom or resolution list.')
+            end
+        end
+        if nargout > 0
+            varargout{1} = zoomlist;
+        end
+        if nargout > 1
+            varargout{2} = reslist;
+        end
         
     otherwise,
-        addonaction = strcmp(st.plugins,action);
+        addonaction = strcmpi(st.plugins,action);
         if any(addonaction)
             feval(['spm_ov_' st.plugins{addonaction}],varargin{:});
-        end;
-end;
+        end
+end
 
-spm('Pointer');
+spm('Pointer','Arrow');
 return;
 
 
@@ -649,7 +704,16 @@ function my_delete(arg1)
 %_______________________________________________________________________
 function resolution(arg1)
 global st
-res      = arg1/mean(svd(st.Space(1:3,1:3)));
+if nargin == 0
+    res = 1; % Default minimum resolution 1mm
+else
+    res = arg1;
+end
+for i=valid_handles(1:24)
+    % adapt resolution to smallest voxel size of displayed images
+    res = min([res,sqrt(sum((st.vols{i}.mat(1:3,1:3)).^2))]);
+end
+res      = res/mean(svd(st.Space(1:3,1:3)));
 Mat      = diag([res res res 1]);
 st.Space = st.Space*Mat;
 st.bb    = st.bb/res;
@@ -670,20 +734,11 @@ function move(handle,pos)
 global st
 mn = [Inf Inf Inf];
 mx = -mn;
-for i=valid_handles(1:24),
-    bb = [[1 1 1];st.vols{i}.dim(1:3)];
-    c = [   bb(1,1) bb(1,2) bb(1,3) 1
-        bb(1,1) bb(1,2) bb(2,3) 1
-        bb(1,1) bb(2,2) bb(1,3) 1
-        bb(1,1) bb(2,2) bb(2,3) 1
-        bb(2,1) bb(1,2) bb(1,3) 1
-        bb(2,1) bb(1,2) bb(2,3) 1
-        bb(2,1) bb(2,2) bb(1,3) 1
-        bb(2,1) bb(2,2) bb(2,3) 1]';
-    tc = st.Space\(st.vols{i}.premul*st.vols{i}.mat)*c;
-    tc = tc(1:3,:)';
-    mx = max([tc ; mx]);
-    mn = min([tc ; mn]);
+for i=valid_handles(1:24)
+    premul = st.Space \ st.vols{i}.premul;
+    bb = spm_get_bbox(st.vols{i}, 'fv', premul);
+    mx = max([bb ; mx]);
+    mn = min([bb ; mn]);
 end;
 bb = [mn ; mx];
 return;
@@ -713,6 +768,51 @@ function space(arg1,M,dim)
 return;
 %_______________________________________________________________________
 %_______________________________________________________________________
+function zoom_op(varargin)
+global st
+if nargin > 0
+    fov = varargin{1};
+else
+    fov = Inf;
+end
+if nargin > 1
+    res = varargin{2};
+else
+    res = Inf;
+end
+if isinf(fov)
+    st.bb = maxbb;
+elseif isnan(fov) || fov == 0
+    current_handle = valid_handles(1:24);
+    if numel(current_handle) > 1 % called from check reg context menu
+        current_handle = get_current_handle;
+    end
+    if fov == 0
+        % zoom to bounding box of current image ~= 0
+        thr = 'nz';
+    else
+        % zoom to bounding box of current image > chosen threshold
+        thr = spm_input('Threshold (Y > ...)', '+1', 'r', '0', 1);
+    end
+    premul = st.Space \ st.vols{current_handle}.premul;
+    st.bb = spm_get_bbox(st.vols{current_handle}, thr, premul);
+else
+    vx = sqrt(sum(st.Space(1:3,1:3).^2));
+    vx = vx.^(-1);
+    pos = spm_orthviews('pos');
+    pos = st.Space\[pos ; 1];
+    pos = pos(1:3)';
+    st.bb = [pos-fov*vx; pos+fov*vx];
+end;
+resolution(res);
+bbox;
+redraw_all;
+if isfield(st.vols{1},'sdip')
+    spm_eeg_inv_vbecd_disp('RedrawDip');
+end
+return;
+%_______________________________________________________________________
+%_______________________________________________________________________
 function H = specify_image(arg1)
 global st
 H=[];
@@ -726,6 +826,7 @@ function space(arg1,M,dim)
         return;
     end;
 end;
+if numel(V)>1, V=V(1); end
 
 ii = 1;
 while ~isempty(st.vols{ii}), ii = ii + 1; end;
@@ -857,6 +958,7 @@ function rmcontexts(handles)
 redraw(1:24);
 return;
 %_______________________________________________________________________
+%_______________________________________________________________________
 function mx = maxval(vol)
 if isstruct(vol),
     mx = -Inf;
@@ -869,6 +971,7 @@ function rmcontexts(handles)
     mx = max(vol(isfinite(vol)));
 end;
 %_______________________________________________________________________
+%_______________________________________________________________________
 function mn = minval(vol)
 if isstruct(vol),
     mn = Inf;
@@ -891,19 +994,19 @@ function redraw(arg1)
 cent = is(1:3,1:3)*st.centre(:) + is(1:3,4);
 
 for i = valid_handles(arg1),
-    M = st.vols{i}.premul*st.vols{i}.mat;
+    M = st.Space\st.vols{i}.premul*st.vols{i}.mat;
     TM0 = [ 1 0 0 -bb(1,1)+1
         0 1 0 -bb(1,2)+1
         0 0 1 -cent(3)
         0 0 0 1];
-    TM = inv(TM0*(st.Space\M));
+    TM = inv(TM0*M);
     TD = Dims([1 2]);
     
     CM0 = [ 1 0 0 -bb(1,1)+1
         0 0 1 -bb(1,3)+1
         0 1 0 -cent(2)
         0 0 0 1];
-    CM = inv(CM0*(st.Space\M));
+    CM = inv(CM0*M);
     CD = Dims([1 3]);
     
     if st.mode ==0,
@@ -911,13 +1014,14 @@ function redraw(arg1)
             0 1 0 -bb(1,2)+1
             1 0 0 -cent(1)
             0 0 0 1];
-        SM = inv(SM0*(st.Space\M)); SD = Dims([3 2]);
+        SM = inv(SM0*M); 
+        SD = Dims([3 2]);
     else
         SM0 = [ 0 -1 0 +bb(2,2)+1
             0  0 1 -bb(1,3)+1
             1  0 0 -cent(1)
             0  0 0 1];
-        SM = inv(SM0*(st.Space\M));
+        SM = inv(SM0*M);
         SD = Dims([2 3]);
     end;
     
@@ -1035,10 +1139,10 @@ function redraw(arg1)
                 end;
                 
                 vol  = st.vols{i}.blobs{1}.vol;
-                M    = st.vols{i}.premul*st.vols{i}.blobs{1}.mat;
-                tmpt = spm_slice_vol(vol,inv(TM0*(st.Space\M)),TD,[0 NaN])';
-                tmpc = spm_slice_vol(vol,inv(CM0*(st.Space\M)),CD,[0 NaN])';
-                tmps = spm_slice_vol(vol,inv(SM0*(st.Space\M)),SD,[0 NaN])';
+                M    = st.Space\st.vols{i}.premul*st.vols{i}.blobs{1}.mat;
+                tmpt = spm_slice_vol(vol,inv(TM0*M),TD,[0 NaN])';
+                tmpc = spm_slice_vol(vol,inv(CM0*M),CD,[0 NaN])';
+                tmps = spm_slice_vol(vol,inv(SM0*M),SD,[0 NaN])';
                 
                 %tmpt_z = find(tmpt==0);tmpt(tmpt_z) = NaN;
                 %tmpc_z = find(tmpc==0);tmpc(tmpc_z) = NaN;
@@ -1087,10 +1191,10 @@ function redraw(arg1)
                 
                 % get blob data
                 vol  = st.vols{i}.blobs{1}.vol;
-                M    = st.vols{i}.premul*st.vols{i}.blobs{1}.mat;
-                tmpt = spm_slice_vol(vol,inv(TM0*(st.Space\M)),TD,[0 NaN])';
-                tmpc = spm_slice_vol(vol,inv(CM0*(st.Space\M)),CD,[0 NaN])';
-                tmps = spm_slice_vol(vol,inv(SM0*(st.Space\M)),SD,[0 NaN])';
+                M    = st.Space\st.vols{i}.premul*st.vols{i}.blobs{1}.mat;
+                tmpt = spm_slice_vol(vol,inv(TM0*M),TD,[0 NaN])';
+                tmpc = spm_slice_vol(vol,inv(CM0*M),CD,[0 NaN])';
+                tmps = spm_slice_vol(vol,inv(SM0*M),SD,[0 NaN])';
                 
                 % actimg scaled round 0, black NaNs
                 topc = size(actc,1)+1;
@@ -1387,20 +1491,33 @@ function redraw_colourbar(vh,bh,interval,cdata)
 item_parent = uicontextmenu;
 
 %contextsubmenu 0
-item00  = uimenu(item_parent, 'Label','unknown image', 'Separator','on');
+item00  = uimenu(item_parent, 'Label','unknown image');
 spm_orthviews('context_menu','image_info',item00,volhandle);
 item0a    = uimenu(item_parent, 'UserData','pos_mm',     'Callback','spm_orthviews(''context_menu'',''repos_mm'');','Separator','on');
 item0b    = uimenu(item_parent, 'UserData','pos_vx',     'Callback','spm_orthviews(''context_menu'',''repos_vx'');');
 item0c    = uimenu(item_parent, 'UserData','v_value');
 
 %contextsubmenu 1
-item1     = uimenu(item_parent,'Label','Zoom');
-item1_1   = uimenu(item1,      'Label','Full Volume',   'Callback','spm_orthviews(''context_menu'',''zoom'',6);', 'Checked','on');
-item1_2   = uimenu(item1,      'Label','160x160x160mm', 'Callback','spm_orthviews(''context_menu'',''zoom'',5);');
-item1_3   = uimenu(item1,      'Label','80x80x80mm',    'Callback','spm_orthviews(''context_menu'',''zoom'',4);');
-item1_4   = uimenu(item1,      'Label','40x40x40mm',    'Callback','spm_orthviews(''context_menu'',''zoom'',3);');
-item1_5   = uimenu(item1,      'Label','20x20x20mm',    'Callback','spm_orthviews(''context_menu'',''zoom'',2);');
-item1_6   = uimenu(item1,      'Label','10x10x10mm',    'Callback','spm_orthviews(''context_menu'',''zoom'',1);');
+item1     = uimenu(item_parent,'Label','Zoom','Separator','on');
+[zl, rl]  = spm_orthviews('ZoomMenu');
+for cz = numel(zl):-1:1
+    if isinf(zl(cz))
+        czlabel = 'Full Volume';
+    elseif isnan(zl(cz))
+        czlabel = 'BBox, this image > ...';
+    elseif zl(cz) == 0
+        czlabel = 'BBox, this image nonzero';
+    else
+        czlabel = sprintf('%dx%d mm', 2*zl(cz), 2*zl(cz));
+    end
+    item1_x = uimenu(item1, 'Label',czlabel,...
+        'Callback', sprintf(...
+        'spm_orthviews(''context_menu'',''zoom'',%d,%d)',...
+        zl(cz),rl(cz)));
+    if isinf(zl(cz)) % default display is Full Volume
+        set(item1_x, 'Checked','on');
+    end
+end
 
 %contextsubmenu 2
 checked={'off','off'};
@@ -1504,11 +1621,14 @@ function redraw_colourbar(vh,bh,interval,cdata)
     st.vols{volhandle}.ax{i}.cm = item_parent;
 end;
 
-if ~isempty(st.plugins) % process any plugins
-    for k = 1:numel(st.plugins),
-        feval(['spm_ov_', st.plugins{k}], ...
-            'context_menu', volhandle, item_parent);
-    end;
+% process any plugins
+for k = 1:numel(st.plugins),
+    feval(['spm_ov_', st.plugins{k}], ...
+        'context_menu', volhandle, item_parent);
+    if k==1
+        h = get(item_parent,'Children');
+        set(h(1),'Separator','on'); 
+    end
 end;
 return;
 %_______________________________________________________________________
@@ -1582,7 +1702,7 @@ function c_menu(varargin)
         spm_orthviews('reposition',newpos_mm(1:3));
         
     case 'zoom'
-        zoom_all(varargin{2});
+        zoom_all(varargin{2:end});
         bbox;
         redraw_all;
         
@@ -1754,10 +1874,10 @@ function c_menu(varargin)
         cm_handles = valid_handles(1:24);
         if varargin{2} == 2, cm_handles = get_current_handle; end;
         spm_figure('Clear','Interactive');
-        [SPM,VOL] = spm_getSPM;
+        [SPM,xSPM] = spm_getSPM;
         if ~isempty(SPM)
             for i = 1:numel(cm_handles),
-                addblobs(cm_handles(i),VOL.XYZ,VOL.Z,VOL.M);
+                addblobs(cm_handles(i),xSPM.XYZ,xSPM.Z,xSPM.M);
                 c_handle = findobj(findobj(st.vols{cm_handles(i)}.ax{1}.cm,'label','Blobs'),'Label','Remove blobs');
                 set(c_handle,'Visible','on');
                 delete(get(c_handle,'Children'));
@@ -1805,15 +1925,15 @@ function c_menu(varargin)
         cm_handles = valid_handles(1:24);
         if varargin{2} == 2, cm_handles = get_current_handle; end;
         spm_figure('Clear','Interactive');
-        [SPM,VOL] = spm_getSPM;
+        [SPM,xSPM] = spm_getSPM;
         if ~isempty(SPM)
             c         = spm_input('Colour','+1','m',...
                 'Red blobs|Yellow blobs|Green blobs|Cyan blobs|Blue blobs|Magenta blobs',[1 2 3 4 5 6],1);
             colours   = [1 0 0;1 1 0;0 1 0;0 1 1;0 0 1;1 0 1];
             c_names   = {'red';'yellow';'green';'cyan';'blue';'magenta'};
-            hlabel = sprintf('%s (%s)',VOL.title,c_names{c});
+            hlabel = sprintf('%s (%s)',xSPM.title,c_names{c});
             for i = 1:numel(cm_handles),
-                addcolouredblobs(cm_handles(i),VOL.XYZ,VOL.Z,VOL.M,colours(c,:),VOL.title);
+                addcolouredblobs(cm_handles(i),xSPM.XYZ,xSPM.Z,xSPM.M,colours(c,:),xSPM.title);
                 addcolourbar(cm_handles(i),numel(st.vols{cm_handles(i)}.blobs));
                 c_handle    = findobj(findobj(st.vols{cm_handles(i)}.ax{1}.cm,'label','Blobs'),'Label','Remove colored blobs');
                 ch_c_handle = get(c_handle,'Children');
@@ -1860,14 +1980,14 @@ function c_menu(varargin)
         cm_handles = valid_handles(1:24);
         if varargin{2} == 2, cm_handles = get_current_handle;end;
         spm_figure('Clear','Interactive');
-        fname   = spm_select(1,'image','select image');
-        if ~isempty(fname)
-            c       = spm_input('Colour','+1','m','Red blobs|Yellow blobs|Green blobs|Cyan blobs|Blue blobs|Magenta blobs',[1 2 3 4 5 6],1);
+        fname   = spm_select([1 Inf],'image','select image(s)');
+        for k = 1:size(fname,1)
+            c       = spm_input(sprintf('Image %d: Colour',k),'+1','m','Red blobs|Yellow blobs|Green blobs|Cyan blobs|Blue blobs|Magenta blobs',[1 2 3 4 5 6],1);
             colours = [1 0 0;1 1 0;0 1 0;0 1 1;0 0 1;1 0 1];
             c_names = {'red';'yellow';'green';'cyan';'blue';'magenta'};
-            hlabel = sprintf('%s (%s)',fname,c_names{c});
+            hlabel = sprintf('%s (%s)',fname(k,:),c_names{c});
             for i = 1:numel(cm_handles),
-                addcolouredimage(cm_handles(i),fname,colours(c,:));
+                addcolouredimage(cm_handles(i),fname(k,:),colours(c,:));
                 addcolourbar(cm_handles(i),numel(st.vols{cm_handles(i)}.blobs));
                 c_handle    = findobj(findobj(st.vols{cm_handles(i)}.ax{1}.cm,'label','Blobs'),'Label','Remove colored blobs');
                 ch_c_handle = get(c_handle,'Children');
@@ -1918,35 +2038,17 @@ function c_menu(varargin)
 return;
 %_______________________________________________________________________
 %_______________________________________________________________________
-function zoom_all(op)
+function zoom_all(zoom,res)
 global st
 cm_handles = get_cm_handles;
-res = [.125 .125 .25 .5 1 1];
-if op==6,
-    st.bb = maxbb;
-else
-    vx = sqrt(sum(st.Space(1:3,1:3).^2));
-    vx = vx.^(-1);
-    pos = spm_orthviews('pos');
-    pos = st.Space\[pos ; 1];
-    pos = pos(1:3)';
-    if     op == 5, st.bb = [pos-80*vx ; pos+80*vx] ;
-    elseif op == 4, st.bb = [pos-40*vx ; pos+40*vx] ;
-    elseif op == 3, st.bb = [pos-20*vx ; pos+20*vx] ;
-    elseif op == 2, st.bb = [pos-10*vx ; pos+10*vx] ;
-    elseif op == 1; st.bb = [pos- 5*vx ; pos+ 5*vx] ;
-    else disp('no Zoom possible');
-    end;
-end
-resolution(res(op));
-redraw_all;
+zoom_op(zoom,res);
 for i = 1:numel(cm_handles)
     z_handle = get(findobj(cm_handles(i),'label','Zoom'),'Children');
     set(z_handle,'Checked','off');
-    set(z_handle(op),'Checked','on');
-end
-
-if isfield(st.vols{1},'sdip')
-    spm_eeg_inv_vbecd_disp('RedrawDip');
+    if isinf(zoom)
+        set(findobj(z_handle,'Label','Full Volume'),'Checked','on');
+    elseif zoom > 0
+        set(findobj(z_handle,'Label',sprintf('%dx%d mm', 2*zoom, 2*zoom)),'Checked','on');
+    end % leave all unchecked if either bounding box option was chosen
 end
 return;
diff --git a/spm_orthviews/spm_ov_reorient.m b/spm_orthviews/spm_ov_reorient.m
index 31b9549..0ae1f73 100644
--- a/spm_orthviews/spm_ov_reorient.m
+++ b/spm_orthviews/spm_ov_reorient.m
@@ -14,9 +14,9 @@
 %             help spm_orthviews
 % at the matlab prompt.
 %_____________________________________________________________________________
-% $Id: spm_ov_reorient.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: spm_ov_reorient.m 4205 2011-02-21 15:39:08Z guillaume $
 
-rev = '$Revision: 3756 $';
+rev = '$Revision: 4205 $';
 
 global st;
 if isempty(st)
@@ -54,8 +54,8 @@
         ret = item0;
 
     case 'context_init'
-        Finter = spm_figure('FindWin', 'Interactive');
-        Fgraph = spm_figure('FindWin', 'Graphics');
+        Finter = spm_figure('GetWin', 'Interactive');
+        Fgraph = spm_figure('GetWin', 'Graphics');
         figure(Finter);
         spm_input('!DeleteInputObj',Finter);
         handles = spm_orthviews('valid_handles');
@@ -78,7 +78,7 @@
             labels{end+1} = '#contour lines';
         end;
         st.vols{volhandle(1)}.reorient.order = uicontrol(...
-            Finter, 'Style','PopupMenu', 'Position', [75 60 330 025], ...
+            Finter, 'Style','PopupMenu', 'Position', [75 60 250 025], ...
             'String',{'Translation(1) Rotation(2) Zoom(3)', ...
             'Zoom(1) Translation(2) Rotation(3)', ...
             'Zoom(1) Rotation(2) Translation(3)'},...
@@ -244,8 +244,8 @@
             if ncl > 0
                 todraw=spm_orthviews('valid_handles');
                 for d = 1:3
-                    CData = sqrt(sum(get(st.vols{volhandle}.ax{d}.d,'CData').^2, ...
-                        3));
+                    CData = sqrt(sum(get(st.vols{volhandle}.ax{d}.d,'CData').^2, 3));
+                    CData(isinf(CData)) = NaN;
                     for h = todraw
                         if h ~= volhandle
                             axes(st.vols{h}.ax{d}.ax);
diff --git a/spm_peb_ppi.m b/spm_peb_ppi.m
index ce8cc36..7a47881 100644
--- a/spm_peb_ppi.m
+++ b/spm_peb_ppi.m
@@ -47,7 +47,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Darren Gitelman
-% $Id: spm_peb_ppi.m 3678 2010-01-14 04:54:38Z Darren $
+% $Id: spm_peb_ppi.m 4185 2011-02-01 18:46:18Z guillaume $
 
 % SETTING UP A PPI THAT ACCOUNTS FOR THE HRF
 % =========================================================================
@@ -583,9 +583,9 @@
 %--------------------------------------------------------------------------
 PPI.xY = xY;
 PPI.dt = dt;
-str    = ['PPI_' PPI.name];
+str    = ['PPI_' PPI.name '.mat'];
 
-if spm_matlab_version_chk('7') >= 0,
+if spm_check_version('matlab','7') >= 0,
     save(fullfile(SPM.swd,str),'-V6','PPI')
 else
     save(fullfile(SPM.swd,str),'PPI')
diff --git a/spm_percentile.m b/spm_percentile.m
index bd12347..f9fb035 100644
--- a/spm_percentile.m
+++ b/spm_percentile.m
@@ -12,27 +12,27 @@
 %
 % Example:
 %  spm_summarise(vols, 'all', @spm_percentile) % quartiles of images
-%
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% Ged Ridgway
+% $Id: spm_percentile.m 4013 2010-07-22 17:12:45Z guillaume $
+
 % The algorithm used is that described by NIST, with x = k+d = 1+p(N-1)/100
 % http://www.itl.nist.gov/div898/handbook/prc/section2/prc252.htm
 %
 % This choice apparently matches Excel, but not MATLAB's prctile, though 
 % the differences are typically small (and are zero for min, median, max).
-
-% For the curious, I [Ged] chose this algorithm because it requires no
-% special handling of 0 or 100, and lets 0 and 1 percentiles differ even
-% with less than 100 samples. It also has the appealing (though perhaps
-% totally inconsequential!) property of returning uniformly spaced values
+%
+% This algorithm was chosen because it requires no special handling of 0 or
+% 100, and lets 0 and 1 percentiles differ even with less than 100 samples.
+% It also has the appealing property of returning uniformly spaced values
 % for uniformly spaced percentiles of uniformly spaced data. For example:
 %  x = 1:11;
 %  p = 0:25:100;
 %  diff([prctile(x, p(:)) spm_percentile(x, p) spm_percentile_nist(x, p)])
 % gives constant differences only for spm_percentile.
 
-% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
-
-% Ged Ridgway
-% $Id: spm_percentile.m 3960 2010-06-30 17:41:24Z ged $
 
 if nargin < 2
     p = 0:25:100;
@@ -41,15 +41,15 @@
     error('Percentage values outside 0 <= p <= 100 are not allowed')
 end
 if all(p < 1) && any(p > 0)
-    warning('spm_percentile:fractions', ...
+    warning('SPM:invalidPercentageValues', ...
         'Percentage values are below 1, but values up to 100 are expected')
 end
 
-p = p(:);
+p    = p(:);
 data = data(:);
 
-data = sort(data(~isnan(data)));
-N = length(data);
+data        = sort(data(~isnan(data)));
+N           = length(data);
 data(end+1) = data(end); % to handle special case of 100th percentile below
 
 x = 1 + p * (N - 1) / 100;
@@ -57,4 +57,3 @@
 d = x - k;
 
 y = (1 - d) .* data(k) + d .* data(k + 1);
-
diff --git a/spm_perm_mtx.m b/spm_perm_mtx.m
new file mode 100644
index 0000000..6066a4d
--- /dev/null
+++ b/spm_perm_mtx.m
@@ -0,0 +1,48 @@
+function [K] = spm_perm_mtx(n)
+% Returns a matrix of indices permuted over n
+% FORMAT [K] = spm_perm_mtx(n)
+%    n   - (scalar) number of indices
+%    K   - (n^2 x n) permutation matrix
+%    n   - (vector) indices
+%    K   - (length(n)! x n) permutation matrix
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_perm_mtx.m 4100 2010-10-22 19:49:17Z karl $
+ 
+% get permutations
+%==========================================================================
+
+% permute zeros and ones
+%--------------------------------------------------------------------------
+if isscalar(n)
+    
+    N  = 2^n;
+    K  = sparse(N,n);
+    x  = sparse(1,1,1,2,1);
+    for i = 1:n
+        y      = ones(N/length(x),1);
+        K(:,i) = kron(x,y);
+        x      = [x;x];
+    end
+    
+% permute indices
+%--------------------------------------------------------------------------
+elseif isvector(n)
+    
+    n  = n(:);
+    K  = n;
+    while size(K,2) < length(n)
+        x     = [];
+        for i = 1:size(K,1)
+            d = K(i,:);
+            r = n;
+            for j = 1:length(d)
+                r(r == d(j)) = [];
+            end
+            x = [x; [kron(ones(length(r),1),d) r]];
+        end
+        K = x;
+    end
+end
\ No newline at end of file
diff --git a/spm_platform.m b/spm_platform.m
index fe0324b..1fbfa3f 100644
--- a/spm_platform.m
+++ b/spm_platform.m
@@ -53,7 +53,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Matthew Brett
-% $Id: spm_platform.m 2820 2009-03-03 18:59:39Z guillaume $
+% $Id: spm_platform.m 4137 2010-12-15 17:18:32Z guillaume $
 
 
 %-Initialise
@@ -139,7 +139,24 @@
 
 function PLATFORM = init_platform(comp)     %-Initialise platform variables
 %==========================================================================
-if nargin<1, comp=computer; end
+if nargin<1
+    comp = computer;
+    
+    if any(comp=='-') % Octave
+        if isunix
+            switch uname.machine
+                case 'x86_64'
+                    comp = 'GLNXA64';
+                case {'i586','i686'}
+                    comp = 'GLNX86';
+            end
+        elseif ispc
+            comp = 'PCWIN';
+        elseif ismac
+            comp = 'MACI';
+        end
+    end
+end
 
 %-Platform definitions
 %--------------------------------------------------------------------------
@@ -155,7 +172,6 @@
 
 PDefs = cell2struct(PDefs,{'computer','filesys','endian'},2);
 
-
 %-Which computer?
 %--------------------------------------------------------------------------
 [issup, ci] = ismember(comp,{PDefs.computer});
diff --git a/spm_plot_ci.m b/spm_plot_ci.m
index 37ad813..32f8fdf 100644
--- a/spm_plot_ci.m
+++ b/spm_plot_ci.m
@@ -14,7 +14,7 @@ function spm_plot_ci(varargin)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_plot_ci.m 3739 2010-02-26 13:12:44Z karl $
+% $Id: spm_plot_ci.m 4169 2011-01-24 18:34:20Z karl $
 
 % unpack
 %--------------------------------------------------------------------------
@@ -43,7 +43,6 @@ function spm_plot_ci(varargin)
 if ~exist('t','var'), t = 1:size(E,2); end
 if ~exist('s','var'), s = '';          end
 
-
 % order and length of sequence
 %--------------------------------------------------------------------------
 E     = E(j,:);
@@ -58,13 +57,23 @@ function spm_plot_ci(varargin)
     end
 else
     if isvector(C)
-        c = ci*sqrt(C(j));
+        c = ci*sqrt(C);
     else
         C = diag(C);
-        c = ci*sqrt(C(j));
+        c = ci*sqrt(C(j,:));
     end
 end
 
+% set plot parameters
+%--------------------------------------------------------------------------
+switch get(gca,'NextPlot')
+    case{lower('add')}
+        col   = [1 1/4 1/4];
+        width = .9;
+    otherwise
+        col   = [1 3/4 3/4];
+        width = .8;
+end
 
 % conditional covariances
 %--------------------------------------------------------------------------
@@ -73,7 +82,7 @@ function spm_plot_ci(varargin)
     % time-series plot
     %======================================================================
     fill([t fliplr(t)],[full(E + c) fliplr(full(E - c))],...
-        [1 1 1]*.8,'EdgeColor',[1 1 1]*.5),hold on
+         [1 1 1]*.8,'EdgeColor',[1 1 1]*.5),hold on
     plot(t,E,s)
     
 elseif n == 2
@@ -90,16 +99,16 @@ function spm_plot_ci(varargin)
     %======================================================================
     
     % conditional means
-    %------------------------------------------------------------------
-    bar(E,'Edgecolor',[1 1 1]/2,'Facecolor',[1 1 1]*.8), hold on
-    axis square
+    %----------------------------------------------------------------------
+    bar(E,width,'Edgecolor',[1 1 1]/2,'Facecolor',[1 1 1]*.8), hold on
     box off
     set(gca,'XLim',[0 n + 1])
     
     % conditional variances
-    %------------------------------------------------------------------
+    %----------------------------------------------------------------------
+
     for k = 1:n
-        line([k k], [-1 1]*c(k) + E(k),'LineWidth',4,'Color','r');
+        line([k k], [-1 1]*c(k) + E(k),'LineWidth',4,'Color',col);
     end
 end
 hold off
diff --git a/spm_plot_convergence.m b/spm_plot_convergence.m
new file mode 100644
index 0000000..065dc15
--- /dev/null
+++ b/spm_plot_convergence.m
@@ -0,0 +1,80 @@
+function spm_plot_convergence(action,varargin)
+% Display a plot showing convergence of an optimisation routine.
+% FORMAT spm_plot_convergence('Init',title,ylabel,xlabel)
+% Initialise the plot in the 'Interactive' window.
+%
+% FORMAT spm_plot_convergence('Set',value)
+% Update the plot.
+%
+% FORMAT spm_plot_convergence('Clear')
+% Clear the 'Interactive' window.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_plot_convergence.m 4147 2010-12-24 13:50:06Z guillaume $
+
+if ~nargin, action = 'Init'; end
+
+% Find the interactive window and exit if not
+%--------------------------------------------------------------------------
+Finter = spm_figure('FindWin','Interactive');
+if isempty(Finter), return; end
+
+switch lower(action)
+    % Initialise
+    %----------------------------------------------------------------------
+    case 'init'
+        if nargin > 1, arg1 = varargin{1}; else arg1 = 'Optimising';  end
+        if nargin > 2, arg2 = varargin{2}; else arg2 = 'Chi-squared'; end
+        if nargin > 3, arg3 = varargin{3}; else arg3 = 'Iteration';   end
+        pb = struct('pointer',get(Finter,'Pointer'),...
+                    'name',   get(Finter,'Name'),...
+                    'buffer', get(Finter,'DoubleBuffer'));
+        spm_plot_convergence('Clear');
+        set(Finter,'Pointer','watch');
+        set(Finter,'Name',pb.name);
+        set(Finter,'DoubleBuffer','on');
+        pb.ax = axes('Position',[0.15 0.1 0.8 0.75],...
+                     'Box',     'on',...
+                     'Parent',  Finter);
+        set(get(pb.ax,'Xlabel'), 'string',arg3, 'FontSize',10);
+        set(get(pb.ax,'Ylabel'), 'string',arg2, 'FontSize',10);
+        set(get(pb.ax,'Title'),  'string',arg1);
+        l = line('Xdata',     [],...
+                 'Ydata',     [],...
+                 'LineWidth', 2,...
+                 'Tag',       'PlotConvOptim',...
+                 'Parent',    pb.ax);
+        set(l,'UserData',pb);
+        
+    % Set
+    %----------------------------------------------------------------------
+    case 'set'
+        if nargin > 1, arg1 = varargin{1}; else arg1 = 0; end
+        br = findobj(Finter,'Tag','PlotConvOptim');
+        if ~isempty(br)
+            xd = get(br,'Xdata');
+            yd = [get(br,'Ydata') arg1];
+            xd = [xd (length(xd)+1)];
+            set(br,'Ydata',yd, 'Xdata',xd);
+        end
+        
+    % Clear
+    %----------------------------------------------------------------------
+    case 'clear'
+        pb = get(findobj(Finter,'Tag','PlotConvOptim'),'UserData');
+        spm_figure('Clear',Finter);
+        if isstruct(pb)
+            set(Finter,'Pointer',     pb.pointer);
+            set(Finter,'Name',        pb.name);
+            set(Finter,'DoubleBuffer',pb.buffer);
+        end
+        
+    % Error
+    %----------------------------------------------------------------------
+    otherwise
+        error('Unknown action string');
+end
+
+drawnow;
diff --git a/spm_powell.m b/spm_powell.m
index 5a02660..788ef74 100644
--- a/spm_powell.m
+++ b/spm_powell.m
@@ -20,7 +20,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_powell.m 3283 2009-07-23 17:37:36Z john $
+% $Id: spm_powell.m 4156 2011-01-11 19:03:31Z guillaume $
 
 
 p     = p(:);
@@ -125,7 +125,7 @@
             d = (1+gold)*(t(3).p-t(2).p);
         end;
         u.p  = t(2).p+d;
-    else,
+    else
         % sign of pol(3) (the 2nd deriv) is not +ve
         % so extend out by golden ratio instead
         u.p  = t(3).p+gold*(t(3).p-t(2).p);
@@ -196,11 +196,11 @@
     % that the displaced point is between the brackets
     % and that the solution is a minimum rather than a maximum
     eps2 = 2*eps*abs(t(1).p)+eps;
-    if abs(d) > abs(ppd)/2 | u.p < brk(1)+eps2 | u.p > brk(2)-eps2 | pol(3)<=0,
+    if abs(d) > abs(ppd)/2 || u.p < brk(1)+eps2 || u.p > brk(2)-eps2 || pol(3)<=0,
         % if criteria are not met, then golden search into the larger part
         if t(1).p >= 0.5*(brk(1)+brk(2)),
             d = gold1*(brk(1)-t(1).p);
-        else,
+        else
             d = gold1*(brk(2)-t(1).p);
         end;
         u.p = t(1).p+d;
@@ -212,12 +212,12 @@
     % Insert the new point into the appropriate position and update
     % the brackets if necessary
     if u.f <= t(1).f,
-        if u.p >= t(1).p, brk(1)=t(1).p; else, brk(2)=t(1).p; end;
+        if u.p >= t(1).p, brk(1)=t(1).p; else brk(2)=t(1).p; end;
         t(3) = t(2);
         t(2) = t(1);
         t(1) = u;
-    else,
-        if u.p < t(1).p, brk(1)=u.p; else, brk(2)=u.p; end;
+    else
+        if u.p < t(1).p, brk(1)=u.p; else brk(2)=u.p; end;
         if u.f <= t(2).f,
             t(3) = t(2);
             t(2) = u;
@@ -239,7 +239,7 @@ function min1d_plot(action,arg1,arg2,arg3,arg4)
 else
     % initialize
     %---------------------------------------------------------------
-    if (strcmp(lower(action),'init'))
+    if strcmpi(action,'init')
         if (nargin<4)
             arg3 = 'Function';
             if (nargin<3)
@@ -272,7 +272,7 @@ function min1d_plot(action,arg1,arg2,arg3,arg4)
 
     % reset
     %---------------------------------------------------------------
-    elseif (strcmp(lower(action),'set'))
+    elseif strcmpi(action,'set')
         F = spm_figure('FindWin','Interactive');
         br = findobj(F,'Tag','LinMinPlot');
         if (~isempty(br))
@@ -285,15 +285,15 @@ function min1d_plot(action,arg1,arg2,arg3,arg4)
 
     % clear
     %---------------------------------------------------------------
-    elseif (strcmp(lower(action),'clear'))
+    elseif strcmpi(action,'clear')
         fg = spm_figure('FindWin','Interactive');
         if isstruct(min1dplot),
-            if ishandle(min1dplot.ax), delete(min1dplot.ax); end;
+            if ishandle(min1dplot.ax), delete(min1dplot.ax); end
             set(fg,'Pointer',min1dplot.pointer);
             set(fg,'Name',min1dplot.name);
-        end;
+        end
         spm_figure('Clear',fg);
         drawnow;
-    end;
+    end
 end
 %_______________________________________________________________________
diff --git a/spm_prep2sn.m b/spm_prep2sn.m
index f3c1637..7307bcb 100644
--- a/spm_prep2sn.m
+++ b/spm_prep2sn.m
@@ -1,25 +1,26 @@
 function [po,pin] = spm_prep2sn(p)
-% Convert the output from spm_preproc into an sn.mat
+% Convert the output from spm_preproc into an sn.mat file
 % FORMAT [po,pin] = spm_prep2sn(p)
 % p   - the results of spm_preproc
-% po  - the output in a form that can be used by
-%       spm_write_sn.
-% pin - the inverse transform in a form that can be
-%       used by spm_write_sn.
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%
+% po  - the output in a form that can be used by spm_write_sn
+% pin - the inverse transform in a form that can be used by spm_write_sn
+%
+% The outputs are saved in sn.mat files only if they are not requested LHS.
+%__________________________________________________________________________
+% Copyright (C) 2005-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_prep2sn.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_prep2sn.m 4276 2011-03-31 11:25:34Z spm $
 
 
-if ischar(p), p = load(p); end;
+if ischar(p), p = load(p); end
 
 VG          = p.tpm;
 VF          = p.image;
 [Y1,Y2,Y3]  = create_def(p.Twarp,VF,VG(1),p.Affine);
-[Y1,Y2,Y3]  =     spm_invdef(Y1,Y2,Y3,VG(1).dim(1:3),eye(4),eye(4));
-MT          =     procrustes(Y1,Y2,Y3,VG(1),VF.mat);
+[Y1,Y2,Y3]  = spm_invdef(Y1,Y2,Y3,VG(1).dim(1:3),eye(4),eye(4));
+MT          = procrustes(Y1,Y2,Y3,VG(1),VF.mat);
 d           = size(p.Twarp);
 [Affine,Tr] = reparameterise(Y1,Y2,Y3,VG(1),VF.mat,MT,max(d(1:3)+2,[8 8 8]));
 flags       = struct(...
@@ -40,40 +41,39 @@
     'Tbias',    p.Tbias,...
     'thresh',   p.thresh);
 
-if nargout==0,
-    [pth,nam,ext] = fileparts(VF.fname);
-    fnam          = fullfile(pth,[nam '_seg_sn.mat']);
-    if spm_matlab_version_chk('7') >= 0,
-        save(fnam,'-V6','VG','VF','Tr','Affine','flags');
+% Parameterisation for the forward transform
+po          = struct(...
+    'VG',       VG,...
+    'VF',       VF,...
+    'Tr',       Tr,...
+    'Affine',   Affine,...
+    'flags',    flags);
+
+% Parameterisation for the inverse transform
+pin         = struct(...
+    'VG',       p.image,...
+    'VF',       p.tpm(1),...
+    'Tr',       p.Twarp,...
+    'Affine',   p.tpm(1).mat\p.Affine*p.image.mat,...
+    'flags',    flags);
+
+if ~nargout
+    [pth,nam] = spm_fileparts(VF.fname);
+    
+    fnam_out  = fullfile(pth,[nam '_seg_sn.mat']);
+    fnam_inv  = fullfile(pth,[nam '_seg_inv_sn.mat']);
+    if spm_check_version('matlab','7') >= 0
+        save(fnam_out,'-V6','-struct','po');
+        save(fnam_inv,'-V6','-struct','pin');
     else
-        save(fnam,'VG','VF','Tr','Affine','flags');
-    end;
-else
-    po = struct(...
-        'VG',     VG,...
-        'VF',     VF,...
-        'Tr',     Tr,...
-        'Affine', Affine,...
-        'flags',  flags);
-end;
-if nargout>=2,
-   % Parameterisation for the inverse
-    pin = struct(...
-        'VG',     p.image,...
-        'VF',     p.tpm(1),...
-        'Tr',     p.Twarp,...
-        'Affine', p.tpm(1).mat\p.Affine*p.image.mat,...
-        'flags',  flags);
-    % VG = p.image;
-    % VF = p.tpm(1);
-    % Tr = p.Twarp;
-    % Affine = p.tpm(1).mat\p.Affine*p.image.mat;
-    % save('junk_sn.mat','VG','VF','Tr','Affine');
-end;
+        save(fnam_out,'-struct','po');
+        save(fnam_inv,'-struct','pin');
+    end
+end
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function [Affine,Tr] = reparameterise(Y1,Y2,Y3,B,M2,MT,d2)
 % Take a deformation field and reparameterise in the same form
 % as used by the spatial normalisation routines of SPM
@@ -89,9 +89,8 @@
 pd  = prod(d2(1:3));
 AA  = eye(pd)*0.01;
 Ab  = zeros(pd,3);
-spm_progress_bar('init',length(x3),['Reparameterising'],'Planes completed');
-mx = [0 0 0];
-for z=1:length(x3),
+spm_progress_bar('init',length(x3),'Reparameterising','Planes completed');
+for z=1:length(x3)
     y1       = double(Y1(:,:,z));
     y2       = double(Y2(:,:,z));
     y3       = double(Y3(:,:,z));
@@ -109,13 +108,13 @@
     Ab(:,3)  = Ab(:,3) + kron(b3,spm_krutil(z3,B1,B2,0));
     AA       = AA  + kron(b3*b3',spm_krutil(w, B1,B2,1));
     spm_progress_bar('set',z);
-end;
+end
 spm_progress_bar('clear');
-Tr  = reshape(AA\Ab,[d2(1:3) 3]); drawnow;
+Tr  = reshape(AA\Ab,[d2(1:3) 3]);
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function MT = procrustes(Y1,Y2,Y3,B,M2)
 % Take a deformation field and determine the closest rigid-body
 % transform to match it, with weighing.
@@ -125,52 +124,52 @@
 % Landmarks: Morphometrics of Group Differences in Outline Shape"
 % Medical Image Analysis 1(3):225-243
 
-M1         = B.mat;
-d          = B.dim(1:3);
-[x1,x2,o]  = ndgrid(1:d(1),1:d(2),1);
-x3         = 1:d(3);
-c1  = [0 0 0];
-c2  = [0 0 0];
-sw  =  0;
-spm_progress_bar('init',length(x3),['Procrustes (1)'],'Planes completed');
-for z=1:length(x3),
-    y1         = double(Y1(:,:,z));
-    y2         = double(Y2(:,:,z));
-    y3         = double(Y3(:,:,z));
-    msk        = find(isfinite(y1));
-    w          = spm_sample_vol(B(1),x1(msk),x2(msk),o(msk)*z,0);
-    swz        = sum(w(:));
-    sw         = sw+swz;
-    c1         = c1 + [w'*[x1(msk) x2(msk)] swz*z ];
-    c2         = c2 +  w'*[y1(msk) y2(msk) y3(msk)];
+M1        = B.mat;
+d         = B.dim(1:3);
+[x1,x2,o] = ndgrid(1:d(1),1:d(2),1);
+x3        = 1:d(3);
+c1        = [0 0 0];
+c2        = [0 0 0];
+sw        =  0;
+spm_progress_bar('init',length(x3),'Procrustes (1)','Planes completed');
+for z=1:length(x3)
+    y1    = double(Y1(:,:,z));
+    y2    = double(Y2(:,:,z));
+    y3    = double(Y3(:,:,z));
+    msk   = find(isfinite(y1));
+    w     = spm_sample_vol(B(1),x1(msk),x2(msk),o(msk)*z,0);
+    swz   = sum(w(:));
+    sw    = sw+swz;
+    c1    = c1 + [w'*[x1(msk) x2(msk)] swz*z ];
+    c2    = c2 +  w'*[y1(msk) y2(msk) y3(msk)];
     spm_progress_bar('set',z);
-end;
+end
 spm_progress_bar('clear');
-c1 = c1/sw;
-c2 = c2/sw;
-T1 = [eye(4,3) M1*[c1 1]'];
-T2 = [eye(4,3) M2*[c2 1]'];
-C  = zeros(3);
-spm_progress_bar('init',length(x3),['Procrustes (2)'],'Planes completed');
-for z=1:length(x3),
-    y1         = double(Y1(:,:,z));
-    y2         = double(Y2(:,:,z));
-    y3         = double(Y3(:,:,z));
-    msk        = find(isfinite(y1));
-    w          = spm_sample_vol(B(1),x1(msk),x2(msk),o(msk)*z,0);
+c1        = c1/sw;
+c2        = c2/sw;
+T1        = [eye(4,3) M1*[c1 1]'];
+T2        = [eye(4,3) M2*[c2 1]'];
+C         = zeros(3);
+spm_progress_bar('init',length(x3),'Procrustes (2)','Planes completed');
+for z=1:length(x3)
+    y1    = double(Y1(:,:,z));
+    y2    = double(Y2(:,:,z));
+    y3    = double(Y3(:,:,z));
+    msk   = find(isfinite(y1));
+    w     = spm_sample_vol(B(1),x1(msk),x2(msk),o(msk)*z,0);
     C = C + [(x1(msk)-c1(1)).*w (x2(msk)-c1(2)).*w (    z-c1(3))*w ]' * ...
             [(y1(msk)-c2(1))    (y2(msk)-c2(2))    (y3(msk)-c2(3)) ];
     spm_progress_bar('set',z);
-end;
+end
 spm_progress_bar('clear');
 [u,s,v]    = svd(M1(1:3,1:3)*C*M2(1:3,1:3)');
 R          = eye(4);
 R(1:3,1:3) = v*u';
 MT         = T2*R*inv(T1);
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function [Y1,Y2,Y3] = create_def(T,VG,VF,Affine)
 % Generate a deformation field from its parameterisation.
 d2   = size(T);
@@ -181,22 +180,22 @@
 B1   = spm_dctmtx(d(1),d2(1));
 B2   = spm_dctmtx(d(2),d2(2));
 B3   = spm_dctmtx(d(3),d2(3));
-[pth,nam,ext] = fileparts(VG.fname);
+[pth,nam] = spm_fileparts(VG.fname);
 spm_progress_bar('init',length(x3),['Creating Def: ' nam],'Planes completed');
-for z=1:length(x3),
+for z=1:length(x3)
     [y1,y2,y3] = defs(T,z,B1,B2,B3,x1,x2,x3,M);
     Y1(:,:,z)  = single(y1);
     Y2(:,:,z)  = single(y2);
     Y3(:,:,z)  = single(y3);
     spm_progress_bar('set',z);
-end;
+end
 spm_progress_bar('clear');
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function [x1,y1,z1] = defs(sol,z,B1,B2,B3,x0,y0,z0,M)
-if ~isempty(sol),
+if ~isempty(sol)
     x1a = x0    + transf(B1,B2,B3(z,:),sol(:,:,:,1));
     y1a = y0    + transf(B1,B2,B3(z,:),sol(:,:,:,2));
     z1a = z0(z) + transf(B1,B2,B3(z,:),sol(:,:,:,3));
@@ -204,18 +203,18 @@
     x1a = x0;
     y1a = y0;
     z1a = z0;
-end;
+end
 x1  = M(1,1)*x1a + M(1,2)*y1a + M(1,3)*z1a + M(1,4);
 y1  = M(2,1)*x1a + M(2,2)*y1a + M(2,3)*z1a + M(2,4);
 z1  = M(3,1)*x1a + M(3,2)*y1a + M(3,3)*z1a + M(3,4);
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function t = transf(B1,B2,B3,T)
 d2 = [size(T) 1];
-t1 = reshape(T, d2(1)*d2(2),d2(3)); drawnow;
-t1 = reshape(t1*B3', d2(1), d2(2)); drawnow;
+t1 = reshape(T, d2(1)*d2(2),d2(3));
+t1 = reshape(t1*B3', d2(1), d2(2));
 t  = B1*t1*B2';
 return;
-%=======================================================================
+%==========================================================================
diff --git a/spm_preproc.m b/spm_preproc.m
index 5773cd0..0cc78ca 100644
--- a/spm_preproc.m
+++ b/spm_preproc.m
@@ -13,35 +13,23 @@
 %  opts.regtype  - regularisation for affine part
 %  opts.fudge    - a fudge factor
 %  opts.msk      - unused
-%
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 2005-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_preproc.m 1143 2008-02-07 19:33:33Z spm $
-
+% $Id: spm_preproc.m 4178 2011-01-27 15:12:53Z guillaume $
 
-[dir,nam,ext]  = fileparts(which(mfilename));
-opts0.tpm   = char(...
-               fullfile(dir,'tpm','grey.nii'),...
-               fullfile(dir,'tpm','white.nii'),...
-               fullfile(dir,'tpm','csf.nii'));
-opts0.ngaus    = [2 2 2 4];
-opts0.warpreg  = 1;
-opts0.warpco   = 25;
-opts0.biasreg  = 0.0001;
-opts0.biasfwhm = 75;
-opts0.regtype  = 'mni';
-opts0.fudge    = 5;
-opts0.samp     = 3;
-opts0.msk      = '';
+opts0     = spm_get_defaults('preproc');
+opts0     = rmfield(opts0,'output');
+opts0.tpm = char(opts0.tpm); % In defaults, tpms are stored as cellstr
+opts0.msk = '';
 
-if nargin==0
+if ~nargin
     V = spm_select(1,'image');
 else
     V = varargin{1};
-end;
-if ischar(V), V = spm_vol(V); end;
+end
+if ischar(V), V = spm_vol(V); end
 
 if nargin < 2
     opts = opts0;
@@ -49,19 +37,19 @@
     opts = varargin{2};
     fnms = fieldnames(opts0);
     for i=1:length(fnms)
-        if ~isfield(opts,fnms{i}), opts.(fnms{i}) = opts0.(fnms{i}); end;
-    end;
-end;
+        if ~isfield(opts,fnms{i}), opts.(fnms{i}) = opts0.(fnms{i}); end
+    end
+end
 
-if length(opts.ngaus)~= size(opts.tpm,1)+1,
+if length(opts.ngaus) ~= size(opts.tpm,1)+1
     error('Number of Gaussians per class is not compatible with number of classes');
-end;
+end
 K   = sum(opts.ngaus);
 Kb  = length(opts.ngaus);
 lkp = [];
-for k=1:Kb,
+for k=1:Kb
     lkp = [lkp ones(1,opts.ngaus(k))*k];
-end;
+end
 
 B         = spm_vol(opts.tpm);
 b0        = spm_load_priors(B);
@@ -84,11 +72,11 @@
 ky     = (pi*((1:d2(2))'-1)/d(2)/vx(2)).^2;
 kz     = (pi*((1:d2(3))'-1)/d(3)/vx(3)).^2;
 Cwarp  = (1*kron(kz.^2,kron(ky.^0,kx.^0)) +...
-        1*kron(kz.^0,kron(ky.^2,kx.^0)) +...
-        1*kron(kz.^0,kron(ky.^0,kx.^2)) +...
-        2*kron(kz.^1,kron(ky.^1,kx.^0)) +...
-        2*kron(kz.^1,kron(ky.^0,kx.^1)) +...
-        2*kron(kz.^0,kron(ky.^1,kx.^1)) );
+    1*kron(kz.^0,kron(ky.^2,kx.^0)) +...
+    1*kron(kz.^0,kron(ky.^0,kx.^2)) +...
+    2*kron(kz.^1,kron(ky.^1,kx.^0)) +...
+    2*kron(kz.^1,kron(ky.^0,kx.^1)) +...
+    2*kron(kz.^0,kron(ky.^1,kx.^1)) );
 Cwarp  = Cwarp*opts.warpreg;
 Cwarp  = [Cwarp*vx(1)^4 ; Cwarp*vx(2)^4 ; Cwarp*vx(3)^4];
 Cwarp  = sparse(1:length(Cwarp),1:length(Cwarp),Cwarp,length(Cwarp),length(Cwarp));
@@ -99,7 +87,7 @@
 Twarp  = zeros([d2 3]);
 
 % GAUSSIAN REGULARISATION for bias correction
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 fwhm    = opts.biasfwhm;
 sd      = vx(1)*V(1).dim(1)/fwhm; d3(1) = ceil(sd*2); krn_x   = exp(-(0:(d3(1)-1)).^2/sd.^2)/sqrt(vx(1));
 sd      = vx(2)*V(1).dim(2)/fwhm; d3(2) = ceil(sd*2); krn_y   = exp(-(0:(d3(2)-1)).^2/sd.^2)/sqrt(vx(2));
@@ -113,8 +101,7 @@
 Tbias   = zeros(d3);
 
 
-% Fudge Factor - to (approximately) account for
-% non-independence of voxels
+% Fudge Factor - to (approximately) account for non-independence of voxels
 ff     = opts.fudge;
 ff     = max(1,ff^3/prod(sk)/abs(det(V.mat(1:3,1:3))));
 Cwarp  = Cwarp*ff;
@@ -123,39 +110,42 @@
 ll     = -Inf;
 llr    = 0;
 llrb   = 0;
-tol1   = 1e-4; % Stopping criterion.  For more accuracy, use a smaller value
+tol1   = 1e-4; % Stopping criterion. For more accuracy, use a smaller value
 d      = [size(x0) length(z0)];
 
 f = zeros(d);
-for z=1:length(z0),
+for z=1:length(z0)
     f(:,:,z) = spm_sample_vol(V,x0,y0,o*z0(z),0);
-end;
+end
 [thresh,mx]  = spm_minmax(f);
 mn           = zeros(K,1);
-rand('state',0); % give same results each time
-for k1=1:Kb,
+% give same results each time
+st = rand('state'); % st = rng;
+rand('state',0); % rng(0,'v5uniform'); % rng('defaults');
+for k1=1:Kb
     kk = sum(lkp==k1);
     mn(lkp==k1) = rand(kk,1)*mx;
-end;
+end
+rand('state',st); % rng(st);
 vr           = ones(K,1)*mx^2;
 mg           = ones(K,1)/K;
 
-if ~isempty(opts.msk),
+if ~isempty(opts.msk)
     VM = spm_vol(opts.msk);
-    if sum(sum((VM.mat-V(1).mat).^2)) > 1e-6 || any(VM.dim(1:3) ~= V(1).dim(1:3)),
+    if sum(sum((VM.mat-V(1).mat).^2)) > 1e-6 || any(VM.dim(1:3) ~= V(1).dim(1:3))
         error('Mask must have the same dimensions and orientation as the image.');
-    end;
-end;
+    end
+end
 
 Affine  = eye(4);
-if ~isempty(opts.regtype),
+if ~isempty(opts.regtype)
     Affine  = spm_maff(V,{x0,y0,z0},b0,B(1).mat,Affine,opts.regtype,ff*100);
     Affine  = spm_maff(V,{x0,y0,z0},b0,B(1).mat,Affine,opts.regtype,ff);
-end;
+end
 M       = B(1).mat\Affine*V(1).mat;
 
 nm      = 0;
-for z=1:length(z0),
+for z=1:length(z0)
     x1  = M(1,1)*x0 + M(1,2)*y0 + (M(1,3)*z0(z) + M(1,4));
     y1  = M(2,1)*x0 + M(2,2)*y0 + (M(2,3)*z0(z) + M(2,4));
     z1  = M(3,1)*x0 + M(3,2)*y0 + (M(3,3)*z0(z) + M(3,4));
@@ -163,11 +153,11 @@
     fz         = f(:,:,z);
     %buf(z).msk = fz>thresh;
     buf(z).msk = buf(z).msk & isfinite(fz) & (fz~=0);
-
+    
     if ~isempty(opts.msk),
         msk = spm_sample_vol(VM,x0,y0,o*z0(z),0);
         buf(z).msk = buf(z).msk & msk;
-    end;
+    end
     buf(z).nm  = sum(buf(z).msk(:));
     buf(z).f   = fz(buf(z).msk);
     nm         = nm + buf(z).nm;
@@ -175,233 +165,230 @@
     buf(z).dat = single(0);
     if buf(z).nm,
         buf(z).dat(buf(z).nm,Kb) = single(0);
-    end;
-end;
+    end
+end
 clear f
 
 finalit = 0;
 
-spm_chi2_plot('Init','Processing','Log-likelihood','Iteration');
-for iter=1:100,
-
-    if finalit,
+spm_plot_convergence('Init','Processing','Log-likelihood','Iteration');
+for iter=1:100
+    
+    if finalit
         % THIS CODE MAY BE USED IN FUTURE
-
+        
         % Reload the data for the final iteration.  This iteration
         % does not do any registration, so there is no need to
         % mask out the background voxels.
-        %------------------------------------------------------------
+        %------------------------------------------------------------------
         llrb  = -0.5*Tbias(:)'*Cbias*Tbias(:);
-        for z=1:length(z0),
+        for z=1:length(z0)
             fz         = spm_sample_vol(V,x0,y0,o*z0(z),0);
             buf(z).msk = fz~=0;
-            if ~isempty(opts.msk),
+            if ~isempty(opts.msk)
                 msk = spm_sample_vol(VM,x0,y0,o*z0(z),0);
                 buf(z).msk = buf(z).msk & msk;
-            end;
+            end
             buf(z).nm  = sum(buf(z).msk(:));
             buf(z).f   = fz(buf(z).msk);
             nm         = nm + buf(z).nm;
             buf(z).bf(1:buf(z).nm,1) = single(1);
             buf(z).dat = single(0);
-            if buf(z).nm,
+            if buf(z).nm
                 buf(z).dat(buf(z).nm,Kb) = single(0);
-            end;
-            if buf(z).nm, 
+            end
+            if buf(z).nm
                 bf        = transf(B1bias,B2bias,B3bias(z,:),Tbias);
                 tmp       = bf(buf(z).msk);
                 llrb      = llrb + sum(tmp);
                 buf(z).bf = single(exp(tmp));
-            end;
-        end;
-
+            end
+        end
+        
         % The background won't fit well any more, so increase the
         % variances of these Gaussians in order to give it a chance
         vr(lkp(K)) = vr(lkp(K))*8;
-
-        spm_chi2_plot('Init','Processing','Log-likelihood','Iteration');
-    end;
-
+        
+        spm_plot_convergence('Init','Processing','Log-likelihood','Iteration');
+    end
+    
     % Load the warped prior probability images into the buffer
-    %------------------------------------------------------------
-    for z=1:length(z0),
-        if ~buf(z).nm, continue; end;
+    %----------------------------------------------------------------------
+    for z=1:length(z0)
+        if ~buf(z).nm, continue; end
         [x1,y1,z1] = defs(Twarp,z,B1warp,B2warp,B3warp,x0,y0,z0,M,buf(z).msk);
-        for k1=1:Kb,
+        for k1=1:Kb
             tmp              = spm_sample_priors(b0{k1},x1,y1,z1,k1==Kb);
             buf(z).dat(:,k1) = single(tmp);
-        end;
-    end;
-
-    for iter1=1:10,
-
-        %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+        end
+    end
+    
+    for iter1=1:10
+        
         % Estimate cluster parameters
-        %------------------------------------------------------------
-        for subit=1:40,
+        %==================================================================
+        for subit=1:40
             oll  = ll;
             mom0 = zeros(K,1)+tiny;
             mom1 = zeros(K,1);
             mom2 = zeros(K,1);
             mgm  = zeros(Kb,1);
             ll   = llr+llrb;
-            for z=1:length(z0),
-                if ~buf(z).nm, continue; end;
+            for z=1:length(z0)
+                if ~buf(z).nm, continue; end
                 bf  = double(buf(z).bf);
                 cr  = double(buf(z).f).*bf;
                 q   =  zeros(buf(z).nm,K);
                 b   =  zeros(buf(z).nm,Kb);
                 s   =  zeros(buf(z).nm,1)+tiny;
-                for k1=1:Kb,
+                for k1=1:Kb
                     pr      = double(buf(z).dat(:,k1));
                     b(:,k1) = pr;
                     s       = s + pr*sum(mg(lkp==k1));
-                end;
-                for k1=1:Kb,
+                end
+                for k1=1:Kb
                     b(:,k1) = b(:,k1)./s;
-                end;
+                end
                 mgm = mgm + sum(b,1)';
-                for k=1:K,
+                for k=1:K
                     q(:,k) = mg(k)*b(:,lkp(k)) .* exp((cr-mn(k)).^2/(-2*vr(k)))/sqrt(2*pi*vr(k));
-                end;
+                end
                 sq = sum(q,2)+tiny;
                 ll = ll + sum(log(sq));
-                for k=1:K, % Moments
+                for k=1:K % Moments
                     p1      = q(:,k)./sq; mom0(k) = mom0(k) + sum(p1(:));
                     p1      = p1.*cr;     mom1(k) = mom1(k) + sum(p1(:));
                     p1      = p1.*cr;     mom2(k) = mom2(k) + sum(p1(:));
-                end;
-            end;
-
+                end
+            end
+            
             % Mixing proportions, Means and Variances
-            for k=1:K,
+            for k=1:K
                 mg(k) = (mom0(k)+eps)/(mgm(lkp(k))+eps);
                 mn(k) = mom1(k)/(mom0(k)+eps);
                 vr(k) =(mom2(k)-mom1(k)*mom1(k)/mom0(k)+1e6*eps)/(mom0(k)+eps);
                 vr(k) = max(vr(k),eps);
-            end;
-
+            end
+            
             if subit>1 || (iter>1 && ~finalit),
-                spm_chi2_plot('Set',ll);
+                spm_plot_convergence('Set',ll);
             end;
-if finalit, fprintf('Mix: %g\n',ll); end;
-            if subit == 1,
+            if finalit, fprintf('Mix: %g\n',ll); end;
+            if subit == 1
                 ooll = ll;
-            elseif (ll-oll)<tol1*nm,
+            elseif (ll-oll)<tol1*nm
                 % Improvement is small, so go to next step
                 break;
-            end;
-        end;
-
-        %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+            end
+        end
+        
         % Estimate bias
-        %------------------------------------------------------------
-        if prod(d3)>0,
-            for subit=1:40,
+        %==================================================================
+        if prod(d3)>0
+            for subit=1:40
                 % Compute objective function and its 1st and second derivatives
                 Alpha = zeros(prod(d3),prod(d3)); % Second derivatives
                 Beta  = zeros(prod(d3),1); % First derivatives
                 ollrb = llrb;
                 oll   = ll;
                 ll    = llr+llrb;
-                for z=1:length(z0),
-                    if ~buf(z).nm, continue; end;
+                for z=1:length(z0)
+                    if ~buf(z).nm, continue; end
                     bf  = double(buf(z).bf);
                     cr  = double(buf(z).f).*bf;
                     q   = zeros(buf(z).nm,K);
-                    for k=1:K,
+                    for k=1:K
                         q(:,k) = double(buf(z).dat(:,lkp(k)))*mg(k);
-                    end;
+                    end
                     s = sum(q,2)+tiny;
-                    for k=1:K,
+                    for k=1:K
                         q(:,k) = q(:,k)./s .* exp((cr-mn(k)).^2/(-2*vr(k)))/sqrt(2*pi*vr(k));
-                    end;
+                    end
                     sq    = sum(q,2)+tiny;
                     ll    = ll   + sum(log(sq));
-
+                    
                     w1 = zeros(buf(z).nm,1);
                     w2 = zeros(buf(z).nm,1);
-                    for k=1:K,
+                    for k=1:K
                         tmp = q(:,k)./sq/vr(k);
                         w1  = w1 + tmp.*(mn(k) - cr);
                         w2  = w2 + tmp;
-                    end;
+                    end
                     wt1   = zeros(d(1:2)); wt1(buf(z).msk) = 1 + cr.*w1;
                     wt2   = zeros(d(1:2)); wt2(buf(z).msk) = cr.*(cr.*w2 - w1);
                     b3    = B3bias(z,:)';
                     Beta  = Beta  + kron(b3,spm_krutil(wt1,B1bias,B2bias,0));
                     Alpha = Alpha + kron(b3*b3',spm_krutil(wt2,B1bias,B2bias,1));
                     clear w1 w2 wt1 wt2 b3
-                end;
-                if finalit, fprintf('Bia: %g\n',ll); end;
-                if subit > 1 && ~(ll>oll),
+                end
+                if finalit, fprintf('Bia: %g\n',ll); end
+                if subit > 1 && ~(ll>oll)
                     % Hasn't improved, so go back to previous solution
                     Tbias = oTbias;
                     llrb  = ollrb;
-                    for z=1:length(z0),
-                        if ~buf(z).nm, continue; end;
+                    for z=1:length(z0)
+                        if ~buf(z).nm, continue; end
                         bf        = transf(B1bias,B2bias,B3bias(z,:),Tbias);
                         buf(z).bf = single(exp(bf(buf(z).msk)));
-                    end;
+                    end
                     break;
                 else
                     % Accept new solution
-                    spm_chi2_plot('Set',ll);
+                    spm_plot_convergence('Set',ll);
                     oTbias = Tbias;
-                    if subit > 1 && ~((ll-oll)>tol1*nm),
+                    if subit > 1 && ~((ll-oll)>tol1*nm)
                         % Improvement is only small, so go to next step
                         break;
                     else
                         % Use new solution and continue the Levenberg-Marquardt iterations
                         Tbias  = reshape((Alpha + Cbias + lmRb)\((Alpha+lmRb)*Tbias(:) + Beta),d3);
                         llrb  = -0.5*Tbias(:)'*Cbias*Tbias(:);
-                        for z=1:length(z0),
-                            if ~buf(z).nm, continue; end;
+                        for z=1:length(z0)
+                            if ~buf(z).nm, continue; end
                             bf        = transf(B1bias,B2bias,B3bias(z,:),Tbias);
                             tmp       = bf(buf(z).msk);
                             llrb      = llrb + sum(tmp);
                             buf(z).bf = single(exp(tmp));
-                        end;
-                    end;
-                end;
-            end;
-            if ~((ll-ooll)>tol1*nm), break; end;
-        end;
-    end;
+                        end
+                    end
+                end
+            end
+            if ~((ll-ooll)>tol1*nm), break; end
+        end
+    end
+    
+    if finalit, break; end
     
-    if finalit, break; end;
-
-    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     % Estimate deformations
-    %------------------------------------------------------------
+    %======================================================================
     mg1 = full(sparse(lkp,1,mg));
     ll  = llr+llrb;
-    for z=1:length(z0),
-        if ~buf(z).nm, continue; end;
+    for z=1:length(z0)
+        if ~buf(z).nm, continue; end
         bf  = double(buf(z).bf);
         cr  = double(buf(z).f).*bf;
         q   =  zeros(buf(z).nm,Kb);
         tmp =  zeros(buf(z).nm,1)+tiny;
         s   =  zeros(buf(z).nm,1)+tiny;
-        for k1=1:Kb,
+        for k1=1:Kb
             s = s + mg1(k1)*double(buf(z).dat(:,k1));
-        end;
-        for k1=1:Kb,
+        end
+        for k1=1:Kb
             kk = find(lkp==k1);
             pp = zeros(buf(z).nm,1);
-            for k=kk,
+            for k=kk
                 pp = pp + exp((cr-mn(k)).^2/(-2*vr(k)))/sqrt(2*pi*vr(k))*mg(k);
-            end;
+            end
             q(:,k1) = pp;
             tmp     = tmp+pp.*double(buf(z).dat(:,k1))./s;
-        end;
+        end
         ll = ll + sum(log(tmp));
-        for k1=1:Kb,
+        for k1=1:Kb
             buf(z).dat(:,k1) = single(q(:,k1));
-        end;
-    end;
-
-    for subit=1:20,
+        end
+    end
+    
+    for subit=1:20
         oll    = ll;
         A      = cell(3,3);
         A{1,1} = zeros(prod(d2));
@@ -411,9 +398,9 @@
         A{2,3} = zeros(prod(d2));
         A{3,3} = zeros(prod(d2));
         Beta   = zeros(prod(d2)*3,1);
-
-        for z=1:length(z0),
-            if ~buf(z).nm, continue; end;
+        
+        for z=1:length(z0)
+            if ~buf(z).nm, continue; end
             [x1,y1,z1] = defs(Twarp,z,B1warp,B2warp,B3warp,x0,y0,z0,M,buf(z).msk);
             b   = zeros(buf(z).nm,Kb);
             db1 = zeros(buf(z).nm,Kb);
@@ -427,40 +414,40 @@
             dp1 = zeros(buf(z).nm,1);
             dp2 = zeros(buf(z).nm,1);
             dp3 = zeros(buf(z).nm,1);
-
-            for k1=1:Kb,
+            
+            for k1=1:Kb
                 [b(:,k1),db1(:,k1),db2(:,k1),db3(:,k1)] = spm_sample_priors(b0{k1},x1,y1,z1,k1==Kb);
                 s   = s   + mg1(k1)*  b(:,k1);
                 ds1 = ds1 + mg1(k1)*db1(:,k1);
                 ds2 = ds2 + mg1(k1)*db2(:,k1);
                 ds3 = ds3 + mg1(k1)*db3(:,k1);
-            end;
-            for k1=1:Kb,
+            end
+            for k1=1:Kb
                 b(:,k1)   = b(:,k1)./s;
                 db1(:,k1) = (db1(:,k1)-b(:,k1).*ds1)./s;
                 db2(:,k1) = (db2(:,k1)-b(:,k1).*ds2)./s;
                 db3(:,k1) = (db3(:,k1)-b(:,k1).*ds3)./s;
-
+                
                 pp  = double(buf(z).dat(:,k1));
                 p   = p   + pp.*b(:,k1);
                 dp1 = dp1 + pp.*(M(1,1)*db1(:,k1) + M(2,1)*db2(:,k1) + M(3,1)*db3(:,k1));
                 dp2 = dp2 + pp.*(M(1,2)*db1(:,k1) + M(2,2)*db2(:,k1) + M(3,2)*db3(:,k1));
                 dp3 = dp3 + pp.*(M(1,3)*db1(:,k1) + M(2,3)*db2(:,k1) + M(3,3)*db3(:,k1));
-            end;
-
+            end
+            
             clear x1 y1 z1 b db1 db2 db3 s ds1 ds2 ds3
-
+            
             tmp             = zeros(d(1:2));
             tmp(buf(z).msk) = dp1./p; dp1 = tmp;
             tmp(buf(z).msk) = dp2./p; dp2 = tmp;
             tmp(buf(z).msk) = dp3./p; dp3 = tmp;
-
+            
             b3     = B3warp(z,:)';
             Beta   = Beta - [...
                 kron(b3,spm_krutil(dp1,B1warp,B2warp,0))
                 kron(b3,spm_krutil(dp2,B1warp,B2warp,0))
                 kron(b3,spm_krutil(dp3,B1warp,B2warp,0))];
-
+            
             b3b3   = b3*b3';
             A{1,1} = A{1,1} +  kron(b3b3,spm_krutil(dp1.*dp1,B1warp,B2warp,1));
             A{1,2} = A{1,2} +  kron(b3b3,spm_krutil(dp1.*dp2,B1warp,B2warp,1));
@@ -468,40 +455,40 @@
             A{2,2} = A{2,2} +  kron(b3b3,spm_krutil(dp2.*dp2,B1warp,B2warp,1));
             A{2,3} = A{2,3} +  kron(b3b3,spm_krutil(dp2.*dp3,B1warp,B2warp,1));
             A{3,3} = A{3,3} +  kron(b3b3,spm_krutil(dp3.*dp3,B1warp,B2warp,1));
-
+            
             clear b3 b3b3 tmp p dp1 dp2 dp3
-        end;
-
+        end
+        
         Alpha = [A{1,1} A{1,2} A{1,3} ; A{1,2} A{2,2} A{2,3}; A{1,3} A{2,3} A{3,3}];
         clear A
-
-        for subit1 = 1:3,
+        
+        for subit1 = 1:3
             if iter==1,
                 nTwarp = (Alpha+lmR*lam + 10*Cwarp)\((Alpha+lmR*lam)*Twarp(:) - Beta);
             else
                 nTwarp = (Alpha+lmR*lam +    Cwarp)\((Alpha+lmR*lam)*Twarp(:) - Beta);
-            end;
+            end
             nTwarp = reshape(nTwarp,[d2 3]);
             nllr   = -0.5*nTwarp(:)'*Cwarp*nTwarp(:);
             nll    = nllr+llrb;
-            for z=1:length(z0),
-                if ~buf(z).nm, continue; end;
+            for z=1:length(z0)
+                if ~buf(z).nm, continue; end
                 [x1,y1,z1] = defs(nTwarp,z,B1warp,B2warp,B3warp,x0,y0,z0,M,buf(z).msk);
                 sq = zeros(buf(z).nm,1) + tiny;
                 b  = zeros(buf(z).nm,Kb);
                 s  = zeros(buf(z).nm,1)+tiny;
-                for k1=1:Kb,
+                for k1=1:Kb
                     b(:,k1) = spm_sample_priors(b0{k1},x1,y1,z1,k1==Kb);
                     s       = s + mg1(k1)*b(:,k1);
-                end;
-                for k1=1:Kb,
+                end
+                for k1=1:Kb
                     sq   = sq + double(buf(z).dat(:,k1)).*b(:,k1)./s;
-                end;
+                end
                 clear b
                 nll = nll + sum(log(sq));
                 clear sq x1 y1 z1
-            end;
-            if nll<ll,
+            end
+            if nll<ll
                 % Worse solution, so use old solution and increase regularisation
                 lam = lam*10;
             else
@@ -510,20 +497,20 @@
                 llr    = nllr;
                 Twarp  = nTwarp;
                 lam    = lam*0.5;
-                break;
-            end;
-        end;
-
-        spm_chi2_plot('Set',ll);
-        if (ll-oll)<tol1*nm, break; end;
-    end;
-
-    if ~((ll-ooll)>tol1*nm),
+                break
+            end
+        end
+        
+        spm_plot_convergence('Set',ll);
+        if (ll-oll)<tol1*nm, break; end
+    end
+    
+    if ~((ll-ooll)>tol1*nm)
         finalit = 1;
         break; % This can be commented out.
-    end;
-end;
-spm_chi2_plot('Clear');
+    end
+end
+spm_plot_convergence('Clear');
 
 results        = opts;
 results.image  = V;
@@ -536,7 +523,7 @@
 results.vr     = vr;
 results.thresh = 0; %thresh;
 results.ll     = ll;
-return;
+return
 %=======================================================================
 
 %=======================================================================
diff --git a/spm_preproc_write.m b/spm_preproc_write.m
index 7c4e9f3..0d88387 100644
--- a/spm_preproc_write.m
+++ b/spm_preproc_write.m
@@ -7,26 +7,27 @@ function spm_preproc_write(p,opts)
 %        GM      - flags for which images should be written
 %        WM      - similar to GM
 %        CSF     - similar to GM
-%____________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 2005-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_preproc_write.m 1586 2008-05-08 20:45:20Z john $
+% $Id: spm_preproc_write.m 4199 2011-02-10 20:07:17Z guillaume $
 
+if ischar(p), p = load(p); end
 
-if nargin==1,
-    opts = struct('biascor',0,'GM',[0 0 1],'WM',[0 0 1],'CSF',[0 0 0],'cleanup',0);
-end;
-if numel(p)>0,
+if nargin==1
+    opts = spm_get_defaults('preproc.output');
+end
+if numel(p)>0
     b0  = spm_load_priors(p(1).VG);
-end;
-for i=1:numel(p),
+end
+for i=1:numel(p)
     preproc_apply(p(i),opts,b0);
-end;
+end
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function preproc_apply(p,opts,b0)
 
 %sopts = [opts.GM ; opts.WM ; opts.CSF];
@@ -35,11 +36,11 @@ function preproc_apply(p,opts,b0)
     case 3
         sopts = [opts.GM ; opts.WM ; opts.CSF];
     case 4
-       sopts = [opts.GM ; opts.WM ; opts.CSF ; opts.EXTRA1];
+        sopts = [opts.GM ; opts.WM ; opts.CSF ; opts.EXTRA1];
     case 5
-       sopts = [opts.GM ; opts.WM ; opts.CSF ; opts.EXTRA1 ; opts.EXTRA2];
+        sopts = [opts.GM ; opts.WM ; opts.CSF ; opts.EXTRA1 ; opts.EXTRA2];
     otherwise
-        error('######## unsupported number of classes....!!!')
+        error('Unsupported number of classes!')
 end
 
 [pth,nam,ext]=fileparts(p.VF.fname);
@@ -175,9 +176,9 @@ function preproc_apply(p,opts,b0)
     end;
 end;
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function [x1,y1,z1] = defs(sol,z,B1,B2,B3,x0,y0,z0,M)
 x1a = x0    + transf(B1,B2,B3(z,:),sol(:,:,:,1));
 y1a = y0    + transf(B1,B2,B3(z,:),sol(:,:,:,2));
@@ -186,9 +187,9 @@ function preproc_apply(p,opts,b0)
 y1  = M(2,1)*x1a + M(2,2)*y1a + M(2,3)*z1a + M(2,4);
 z1  = M(3,1)*x1a + M(3,2)*y1a + M(3,3)*z1a + M(3,4);
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function t = transf(B1,B2,B3,T)
 if ~isempty(T)
     d2 = [size(T) 1];
@@ -198,9 +199,9 @@ function preproc_apply(p,opts,b0)
     t = zeros(size(B1,1),size(B2,1),size(B3,1));
 end;
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function dat = decimate(dat,fwhm)
 % Convolve the volume in memory (fwhm in voxels).
 lim = ceil(2*fwhm);
@@ -212,9 +213,9 @@ function preproc_apply(p,opts,b0)
 k  = (length(z) - 1)/2;
 spm_conv_vol(dat,dat,x,y,z,-[i j k]);
 return;
-%=======================================================================
+%==========================================================================
 
-%=======================================================================
+%==========================================================================
 function [g,w,c] = clean_gwc(g,w,c, level)
 if nargin<4, level = 1; end;
 
@@ -222,7 +223,7 @@ function preproc_apply(p,opts,b0)
 b(1) = w(1);
 
 % Build a 3x3x3 seperable smoothing kernel
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 kx=[0.75 1 0.75];
 ky=[0.75 1 0.75];
 kz=[0.75 1 0.75];
@@ -232,7 +233,7 @@ function preproc_apply(p,opts,b0)
 th1 = 0.15;
 if level==2, th1 = 0.2; end;
 % Erosions and conditional dilations
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 niter = 32;
 spm_progress_bar('Init',niter,'Extracting Brain','Iterations completed');
 for j=1:niter,
@@ -260,7 +261,4 @@ function preproc_apply(p,opts,b0)
 end;
 spm_progress_bar('Clear');
 return;
-%=======================================================================
-
-%=======================================================================
-
+%==========================================================================
diff --git a/spm_print.m b/spm_print.m
index df7e9d5..e6cd9e1 100644
--- a/spm_print.m
+++ b/spm_print.m
@@ -1,23 +1,44 @@
-function spm_print(job)
+function spm_print(varargin)
 % Print the Graphics window
+% FORMAT spm_print
+% Print the default window to the default output file.
+% FORMAT spm_print(filename)
+% Print the default window to the specified output file.
+% FORMAT spm_print('',figurename)
+% Print the window with the specified name to the default output file.
+% FORMAT spm_print(filename,figurename)
+% Print the window with the specified name to the specified output file.
+% FORMAT spm_print('',figurehandle)
+% Print the figure with the specified handle to the default output file.
+% FORMAT spm_print(filename,figurehandle)
+% Print the figure with the specified handle to the specified output file.
+% FORMAT spm_print(job)
+% Run a batch print job, print the specified window to the specified
+% output file.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_print.m 3933 2010-06-17 14:19:08Z guillaume $
+% $Id: spm_print.m 4050 2010-08-26 18:09:18Z guillaume $
 
 %-Run spm_print through the Batch System to get configured print options
 %==========================================================================
 if ~nargin
-    spm_jobman('serial','','spm.util.print','');
+    spm_jobman('serial','','spm.util.print','',{2},NaN);
     return;
-elseif ischar(job)
-    spm_jobman('serial','','spm.util.print',job);
+elseif ischar(varargin{1})
+    if nargin == 1
+        spm_jobman('serial','','spm.util.print',varargin{1},{2},NaN);
+    else
+        spm_jobman('serial','','spm.util.print',varargin{1},{2}, ...
+                   varargin{2});
+    end
     return;
 end
 
 %-Print the Graphics window
 %==========================================================================
+job = varargin{1};
 try
     %-Get output filename
     %----------------------------------------------------------------------
@@ -42,13 +63,23 @@ function spm_print(job)
     
     %-Get figure handle
     %----------------------------------------------------------------------
-    if strcmp(get(gcf,'Tag'),'Help')
-        fg = gcf;
+    if isfield(job.fig,'fighandle')
+        if ishandle(job.fig.fighandle) && ...
+                strcmpi(get(job.fig.fighandle,'type'),'figure')
+            fg = job.fig.fighandle;
+        elseif ~isfinite(job.fig.fighandle)
+            fg = get(0,'CurrentFigure');
+            if ~strcmp(get(fg,'Tag'),'Help')
+                fg = spm_figure('FindWin','Graphics');
+            end
+        else
+            fg = [];
+        end
     else
-        fg = spm_figure('FindWin','Graphics');
+        fg = spm_figure('FindWin',job.fig.figname);
     end
     if isempty(fg)
-        fprintf('\nGraphics window not found: nothing has been printed.\n');
+        fprintf('\nFigure not found: nothing has been printed.\n');
         return;
     end
     
@@ -63,9 +94,9 @@ function spm_print(job)
     %-Report
     %----------------------------------------------------------------------
     if isempty(strfind(nam1,filesep))
-        fprintf('\nPrinting Graphics Windows to\n%s%s%s\n',pwd,filesep,nam1);
+        fprintf('\nPrinting to\n%s%s%s\n',pwd,filesep,nam1);
     else
-        fprintf('\nPrinting Graphics Windows to\n%s\n',nam1);
+        fprintf('\nPrinting to\n%s\n',nam1);
     end
     
 %-Report errors
diff --git a/spm_read_netcdf.m b/spm_read_netcdf.m
index 8030e16..c57d529 100644
--- a/spm_read_netcdf.m
+++ b/spm_read_netcdf.m
@@ -5,73 +5,81 @@
 % cdf   - data structure
 %
 % See: http://www.unidata.ucar.edu/packages/netcdf/
-% _______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% _________________________________________________________________________
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_read_netcdf.m 3014 2009-03-31 14:22:42Z john $
+% $Id: spm_read_netcdf.m 4182 2011-02-01 12:29:09Z guillaume $
 
 
 dsiz     = [1 1 2 4 4 8];
 fp=fopen(fname,'r','ieee-be');
-if fp==-1,
+if fp==-1
     cdf = [];
     return;
-end;
+end
 
 % Return null if not a CDF file.
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 mgc = fread(fp,4,'uchar')';
-if ~all(['CDF' 1] == mgc),
+if ~all(['CDF' 1] == mgc)
     cdf = [];
     fclose(fp);
-    if all(mgc==[137,72,68,70]), fprintf('"%s" appears to be based around HDF.\nThis is a newer version of MINC that SPM can not yet read.\n', fname); end
+    if all(mgc==[137,72,68,70])
+        fprintf(['"%s" appears to be based around HDF.\n',...
+            'This is a newer version of MINC that SPM can not yet read.\n'],...
+            fname);
+    end
     return;
 end
 
 % I've no idea what this is for
 numrecs = fread(fp,1,'uint32');
 
-cdf = struct('numrecs',numrecs,'dim_array',[], 'gatt_array',[], 'var_array', []);
+cdf = struct('numrecs',    numrecs,...
+             'dim_array',  [], ...
+             'gatt_array', [], ...
+             'var_array',  []);
 
 dt = fread(fp,1,'uint32');
-if dt == 10,
+if dt == 10
     % Dimensions
     nelem = fread(fp,1,'uint32');
-    for j=1:nelem,
-        str   = readname(fp);
+    for j=1:nelem
+        str = readname(fp);
         dim_length = fread(fp,1,'uint32');
         cdf.dim_array(j).name       = str;
         cdf.dim_array(j).dim_length = dim_length;
-    end;
-    dt   = fread(fp,1,'uint32');
+    end
+    dt = fread(fp,1,'uint32');
 end
 
-while ~dt, dt   = fread(fp,1,'uint32'); end;
+while ~dt, dt = fread(fp,1,'uint32'); end
 
-if dt == 12,
+if dt == 12
     % Attributes
     nelem = fread(fp,1,'uint32');
-    for j=1:nelem,
-        str    = readname(fp);
-        nc_type= fread(fp,1,'uint32');
-        nnelem = fread(fp,1,'uint32');
-        val    = fread(fp,nnelem,dtypestr(nc_type));
+    for j=1:nelem
+        str     = readname(fp);
+        nc_type = fread(fp,1,'uint32');
+        nnelem  = fread(fp,1,'uint32');
+        val     = fread(fp,nnelem,dtypestr(nc_type));
         if nc_type == 2, val = deblank([val' ' ']); end
-        padding= fread(fp,ceil(nnelem*dsiz(nc_type)/4)*4-nnelem*dsiz(nc_type),'uchar');
+        padding= fread(fp,...
+            ceil(nnelem*dsiz(nc_type)/4)*4-nnelem*dsiz(nc_type),'uchar');
         cdf.gatt_array(j).name    = str;
         cdf.gatt_array(j).nc_type = nc_type;
         cdf.gatt_array(j).val     = val;
-    end;
-    dt   = fread(fp,1,'uint32');
+    end
+    dt = fread(fp,1,'uint32');
 end
 
-while ~dt, dt   = fread(fp,1,'uint32'); end;
+while ~dt, dt = fread(fp,1,'uint32'); end
 
-if dt == 11,
+if dt == 11
     % Variables
     nelem = fread(fp,1,'uint32');
-    for j=1:nelem,
+    for j=1:nelem
         str    = readname(fp);
         nnelem = fread(fp,1,'uint32');
         val    = fread(fp,nnelem,'uint32');
@@ -81,9 +89,9 @@
         cdf.var_array(j).vsize   = 0;
         cdf.var_array(j).begin   = 0;
         dt0    = fread(fp,1,'uint32');
-        if dt0 == 12,
+        if dt0 == 12
             nelem0 = fread(fp,1,'uint32');
-            for jj=1:nelem0,
+            for jj=1:nelem0
                 str    = readname(fp);
                 nc_type= fread(fp,1,'uint32');
                 nnelem = fread(fp,1,'uint32');
@@ -94,34 +102,31 @@
                 cdf.var_array(j).vatt_array(jj).name    = str;
                 cdf.var_array(j).vatt_array(jj).nc_type = nc_type;
                 cdf.var_array(j).vatt_array(jj).val     = val;
-            end;
+            end
             dt0    = fread(fp,1,'uint32');
-        end;
-        cdf.var_array(j).nc_type  = dt0;
-        cdf.var_array(j).vsize = fread(fp,1,'uint32');
-        cdf.var_array(j).begin = fread(fp,1,'uint32');
-    end;
-    dt   = fread(fp,1,'uint32');
-end;
+        end
+        cdf.var_array(j).nc_type = dt0;
+        cdf.var_array(j).vsize   = fread(fp,1,'uint32');
+        cdf.var_array(j).begin   = fread(fp,1,'uint32');
+    end
+    dt = fread(fp,1,'uint32');
+end
 
 fclose(fp);
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+%==========================================================================
+% function str = dtypestr(i)
+%==========================================================================
 function str = dtypestr(i)
 % Returns a string appropriate for reading or writing the CDF data-type.
 types = char('uint8','uint8','int16','int32','float32','float64');
 str   = deblank(types(i,:));
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+%==========================================================================
+% function name = readname(fp)
+%==========================================================================
 function name = readname(fp)
-% Extracts a name from a CDF file pointed to at the right location by
-% fp.
-stlen  = fread(fp,1,'uint32');
-name   = deblank([fread(fp,stlen,'uchar')' ' ']);
-padding= fread(fp,ceil(stlen/4)*4-stlen,'uchar');
-return;
-%_______________________________________________________________________
+% Extracts a name from a CDF file pointed to at the right location by fp.
+stlen   = fread(fp,1,'uint32');
+name    = deblank([fread(fp,stlen,'uchar')' ' ']);
+padding = fread(fp,ceil(stlen/4)*4-stlen,'uchar');
diff --git a/spm_realign.m b/spm_realign.m
index b9a0091..70267fe 100644
--- a/spm_realign.m
+++ b/spm_realign.m
@@ -80,12 +80,15 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_realign.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: spm_realign.m 4152 2011-01-11 14:13:35Z volkmar $
 
 
 if nargin==0, return; end;
 
-def_flags = struct('quality',1,'fwhm',5,'sep',4,'interp',2,'wrap',[0 0 0],'rtm',0,'PW','','graphics',1,'lkp',1:6);
+def_flags          = spm_get_defaults('realign.estimate');
+def_flags.PW       = '';
+def_flags.graphics = 1;
+def_flags.lkp      = 1:6;
 if nargin < 2,
     flags = def_flags;
 else
@@ -217,13 +220,13 @@
     % now. It basically involves removing the voxels that contribute
     % least to the determinant of the inverse covariance matrix.
 
-    spm_chi2_plot('Init','Eliminating Unimportant Voxels',...
+    spm_plot_convergence('Init','Eliminating Unimportant Voxels',...
               'Relative quality','Iteration');
     Alpha = [A0 b];
     Alpha = Alpha'*Alpha;
     det0  = det(Alpha);
     det1  = det0;
-    spm_chi2_plot('Set',det1/det0);
+    spm_plot_convergence('Set',det1/det0);
     while det1/det0 > flags.quality,
         dets  = zeros(size(A0,1),1);
         for i=1:size(A0,1),
@@ -240,9 +243,9 @@
         Alpha = [A0 b];
         Alpha = Alpha'*Alpha;
         det1  = det(Alpha);
-        spm_chi2_plot('Set',single(det1/det0));
+        spm_plot_convergence('Set',single(det1/det0));
     end;
-    spm_chi2_plot('Clear');
+    spm_plot_convergence('Clear');
 end;
 %-----------------------------------------------------------------------
 
diff --git a/spm_regions.m b/spm_regions.m
index a6ff2aa..b938409 100644
--- a/spm_regions.m
+++ b/spm_regions.m
@@ -42,10 +42,10 @@
 % be extracted from xY.y, and will be the same as the [adjusted] data 
 % returned by the plotting routine (spm_graph.m) for the same contrast.
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1999-2011 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_regions.m 3812 2010-04-07 16:52:05Z karl $
+% $Id: spm_regions.m 4185 2011-02-01 18:46:18Z guillaume $
 
 if nargin < 4, xY = []; end
 
@@ -122,7 +122,11 @@
 try, xY = rmfield(xY,'M'); end
 try, xY = rmfield(xY,'rej'); end
 
-if isempty(xY.XYZmm), error('Empty region.'); end
+if isempty(xY.XYZmm)
+    warning('Empty region.');
+    Y = [];
+    return;
+end
 
 
 %-Extract required data from results files
@@ -241,18 +245,17 @@
 
 %-Save
 %==========================================================================
-str = ['VOI_' xY.name];
+str = ['VOI_' xY.name '.mat'];
 if isfield(xY,'Sess') && isfield(SPM,'Sess')
-    str = sprintf('VOI_%s_%i',xY.name,xY.Sess);
+    str = sprintf('VOI_%s_%i.mat',xY.name,xY.Sess);
 end
-if spm_matlab_version_chk('7') >= 0
+if spm_check_version('matlab','7') >= 0
     save(fullfile(SPM.swd,str),'-V6','Y','xY')
 else
     save(fullfile(SPM.swd,str),'Y','xY')
 end
 
-fprintf('   VOI saved as %s\n', ...
-    spm_str_manip(fullfile(SPM.swd,[str '.mat']),'k55'));               %-#
+fprintf('   VOI saved as %s\n',spm_str_manip(fullfile(SPM.swd,str),'k55'));
 
 %-Reset title
 %--------------------------------------------------------------------------
diff --git a/spm_reml_A.m b/spm_reml_A.m
new file mode 100644
index 0000000..8db18c5
--- /dev/null
+++ b/spm_reml_A.m
@@ -0,0 +1,191 @@
+function  [C,h,Ph,F,Fa,Fc] = spm_reml_A(YY,X,Q,N,hE,hC,V)
+% ReML estimation of covariance components from y*y' - factored components
+% FORMAT [C,h,Ph,F,Fa,Fc] = spm_reml_A(YY,X,Q,N,[hE,hC,V]);
+%
+% YY  - (m x m) sample covariance matrix Y*Y'  {Y = (m x N) data matrix}
+% X   - (m x p) design matrix
+% Q   - {1 x q} covariance components (factors)
+% N   - number of samples
+%
+% hE  - hyperprior expectation [default = 0]
+% hC  - hyperprior covariance  [default = 256]
+% V   - fixed covariance component
+%
+% C   - (m x m) estimated errors: C = A*A': A = h(1)*Q{1} + h(2)*Q{2} + ...
+% h   - (q x 1) ReML hyperparameters h
+% Ph  - (q x q) conditional precision of h
+%
+% F   - [-ve] free energy F = log evidence = p(Y|X,Q) = ReML objective
+%
+% Fa  - accuracy
+% Fc  - complexity (F = Fa - Fc)
+%
+% Performs a Fisher-Scoring ascent on F to find MAP variance parameter
+% estimates. NB: uses weakly informative normal hyperpriors on the factors.
+%
+%__________________________________________________________________________
+%
+% SPM ReML routines:
+%
+%      spm_reml:    no positivity constraints on covariance parameters
+%      spm_reml_sc: positivity constraints on covariance parameters
+%      spm_sp_reml: for sparse patterns (c.f., ARD)
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_reml_A.m 4195 2011-02-05 18:39:13Z ged $
+
+
+% assume a single sample if not specified
+%--------------------------------------------------------------------------
+try, N; catch, N  = 1;      end
+try, V; catch, V  = 1e-16;  end
+
+% initialise h
+%--------------------------------------------------------------------------
+n     = length(Q{1});
+m     = length(Q);
+h     = zeros(m,1) + exp(-4);
+dFdh  = zeros(m,1);
+dFdhh = zeros(m,m);
+
+% ortho-normalise X
+%--------------------------------------------------------------------------
+if isempty(X)
+    X = sparse(n,0);
+else
+    X = spm_svd(X,0);
+end
+
+% check fixed component
+%--------------------------------------------------------------------------
+if length(V) == 1
+    V = V*speye(n,n);
+end
+
+
+% initialise and specify hyperpriors
+%==========================================================================
+
+% hyperpriors
+%--------------------------------------------------------------------------
+try, hE = hE(:);        catch, hE = 0;    end
+try, hP = spm_inv(hC);  catch, hP = 1/256; end
+
+% check sise
+%--------------------------------------------------------------------------
+if length(hE) < m, hE = hE(1)*ones(m,1);   end
+if length(hP) < m, hP = hP(1)*speye(m,m);  end
+
+% intialise h: so that sum(exp(h)) = 1
+%--------------------------------------------------------------------------
+if any(diag(hP) > exp(16))
+    h = hE;
+end
+
+% ReML (EM/VB)
+%--------------------------------------------------------------------------
+dF    = Inf;
+t     = 4;
+for k = 1:32
+    
+    % compute current estimate of covariance
+    %----------------------------------------------------------------------
+    A     = 0;
+    for i = 1:m
+        A = A + Q{i}*h(i);
+    end
+    C     = V + A*A';
+    iC    = spm_inv(C);
+    
+    % E-step: conditional covariance cov(B|y) {Cq}
+    %======================================================================
+    iCX    = iC*X;
+    if ~isempty(X)
+        Cq = inv(X'*iCX);
+    else
+        Cq = sparse(0);
+    end
+    
+    % M-step: ReML estimate of hyperparameters
+    %======================================================================
+    
+    % Gradient dF/dh (first derivatives)
+    %----------------------------------------------------------------------
+    P     = iC - iCX*Cq*iCX';
+    U     = speye(n) - P*YY/N;
+    PQ = cell(m, 1);
+    for i = 1:m
+        
+        % dF/dh
+        %------------------------------------------------------------------
+        PQ{i}   = P*(A*Q{i}' + Q{i}'*A);
+        dFdh(i) = -sum(sum(PQ{i}.*U'))*N/2; % = -trace(PQ{i}*U)*N/2;
+        
+    end
+    
+    % Expected curvature E{dF/dhh} (second derivatives)
+    % dF/dhh = -trace{P*Q{i}*P*Q{j}} NB trace(A*B)==sum(sum(A.*B'))
+    %----------------------------------------------------------------------
+    for i = 1:m
+        dFdhh(i,i) = -sum(sum(PQ{i}.*PQ{i}'))*N/2;
+        for j = i+1:m
+            dFdhh(i,j) = -sum(sum(PQ{i}.*PQ{j}'))*N/2;
+            dFdhh(j,i) =  dFdhh(i,j);
+        end
+    end
+    
+    % add hyperpriors
+    %----------------------------------------------------------------------
+    e     = h     - hE;
+    dFdh  = dFdh  - hP*e;
+    dFdhh = dFdhh - hP;
+    
+    % Fisher scoring: update dh = -inv(ddF/dhh)*dF/dh
+    %----------------------------------------------------------------------
+    dh    = spm_dx(dFdhh,dFdh,{t});
+    h     = h + dh;
+    
+    
+    % predicted change in F - increase regularisation if increasing
+    %----------------------------------------------------------------------
+    pF    = dFdh'*dh;
+    if pF > dF
+        t = t - 1;
+    else
+        t = t + 1/8;
+    end
+    dF    = pF;
+    
+    % convergence
+    %----------------------------------------------------------------------
+    fprintf('%s %-23d: %10s%e [%+3.2f]\n','  ReML Iteration',k,'...',full(dF),t);
+    if dF < 1e-2
+        break
+    end
+end
+
+% log evidence = ln p(y|X,Q) = ReML objective = F = trace(R'*iC*R*YY)/2 ...
+%--------------------------------------------------------------------------
+Ph    = -dFdhh;
+if nargout > 3
+    
+    % tr(hP*inv(Ph)) - nh + tr(pP*inv(Pp)) - np (pP = 0)
+    %----------------------------------------------------------------------
+    Ft = trace(hP/Ph) - length(Ph) - length(Cq);
+    
+    % complexity - KL(Ph,hP)
+    %----------------------------------------------------------------------
+    Fc = Ft/2 + e'*hP*e/2 + spm_logdet(Ph/hP)/2 - N*spm_logdet(Cq)/2;
+    
+    % Accuracy - ln p(Y|h)
+    %----------------------------------------------------------------------
+    Fa = Ft/2 - trace(C*P*YY*P)/2 - N*n*log(2*pi)/2  - N*spm_logdet(C)/2;
+    
+    % Free-energy
+    %----------------------------------------------------------------------
+    F  = Fa - Fc;
+    
+end
diff --git a/spm_render.m b/spm_render.m
index a5152dc..063cf8d 100644
--- a/spm_render.m
+++ b/spm_render.m
@@ -33,15 +33,13 @@ function spm_render(dat,brt,rendfile)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_render.m 3975 2010-07-08 11:31:35Z guillaume $
+% $Id: spm_render.m 4018 2010-07-27 18:22:42Z guillaume $
 
-SVNrev = '$Rev: 3975 $';
+SVNrev = '$Rev: 4018 $';
 
 global prevrend
 if ~isstruct(prevrend)
-    prevrend = struct('rendfile','',...
-                      'brt',[],...
-                      'col',[]);
+    prevrend = struct('rendfile','', 'brt',[], 'col',[]);
 end
 
 %-Parse arguments, get data if not passed as parameters
@@ -204,7 +202,7 @@ function spm_render(dat,brt,rendfile)
             z1  = dep(round(xyz(1,:))+round(xyz(2,:)-1)*size(dep,1));
 
             if ~isfinite(brt), msk = find(xyz(3,:) < (z1+20) & xyz(3,:) > (z1-5));
-            else, msk = find(xyz(3,:) < (z1+60) & xyz(3,:) > (z1-5)); end
+            else msk = find(xyz(3,:) < (z1+60) & xyz(3,:) > (z1-5)); end
         else
             msk = [];
         end
@@ -215,7 +213,8 @@ function spm_render(dat,brt,rendfile)
             %--------------------------------------------------------------
             xyz = xyz(:,msk);
             if ~isfinite(brt), t0  = t(msk);
-            else,   dst = xyz(3,:) - z1(msk);
+            else
+                dst = xyz(3,:) - z1(msk);
                 dst = max(dst,0);
                 t0  = t(msk).*exp((log(0.5)/10)*dst)';
             end
@@ -249,7 +248,7 @@ function spm_render(dat,brt,rendfile)
 spm_results_ui('Clear',Fgraph);
 
 nrow = ceil(length(rend)/2);
-if showbar, hght = 0.95; else, hght = 0.5; end
+if showbar, hght = 0.95; else hght = 0.5; end
 % subplot('Position',[0, 0, 1, hght]);
 ax=axes('Parent',Fgraph,'units','normalized','Position',[0, 0, 1, hght],'Visible','off');
 image(0,'Parent',ax);
@@ -317,114 +316,27 @@ function surf_rend(dat,rend,col)
 %--------------------------------------------------------------------------
 Fgraph = spm_figure('GetWin','Graphics');
 spm_results_ui('Clear',Fgraph);
-rdr = get(Fgraph,'Renderer');
-set(Fgraph,'Renderer','OpenGL');
 
 ax0 = axes(...
-    'Parent',Fgraph,...
-    'Units','normalized',...
-    'Color', [1 1 1],...
-    'XTick',[],...
-    'YTick',[],...
-    'Position',[-0.05, -0.05, 1.05, 0.555]);
+    'Tag',      'SPMMeshRenderBackground',...
+    'Parent',   Fgraph,...
+    'Units',    'normalized',...
+    'Color',    [1 1 1],...
+    'XTick',    [],...
+    'YTick',    [],...
+    'Position', [-0.05, -0.05, 1.05, 0.555]);
 
 ax = axes(...
-    'Parent',Fgraph,...
-    'Units','normalized',...
-    'Position',[0.05, 0.05, 0.9, 0.4],...
-    'Visible','off');
-
-%-Project data onto surface mesh
-%--------------------------------------------------------------------------
-v = spm_mesh_project(rend, dat);
+    'Parent',   Fgraph,...
+    'Units',    'normalized',...
+    'Position', [0.05, 0.05, 0.9, 0.4],...
+    'Visible',  'off');
 
-%-Compute mesh curvature texture
-%--------------------------------------------------------------------------
-curv = spm_mesh_curvature(rend) > 0;
-curv = 0.5 * repmat(curv,1,3) + 0.3 * repmat(~curv,1,3);
+H = spm_mesh_render('Disp',rend,struct('parent',ax));
+spm_mesh_render('Overlay',H,dat,col);
 
-%-Combine projected data and mesh curvature
-%--------------------------------------------------------------------------
-cdat = zeros(size(v,2),3);
-if any(v(:))
-    if size(col,1)>3
-        cdat = squeeze(ind2rgb(floor(v(:)/max(v(:))*size(col,1)),col));
-    else
-        m = max(v(:));
-        for i=1:size(v,1)
-            cdat = cdat + v(i,:)'/m * col(i,:);
-        end
-    end
-end
-cdat = repmat(~any(v,1),3,1)' .* curv + repmat(any(v,1),3,1)' .* cdat;
-
-%-Display the surface mesh with texture
-%--------------------------------------------------------------------------
-hp = patch(rend, 'Parent',ax,...
-    'FaceVertexCData',cdat, ...
-    'FaceColor', 'interp', ...
-    'EdgeColor', 'none',...
-    'FaceLighting', 'phong',...
-    'SpecularStrength' ,0.7, 'AmbientStrength', 0.1,...
-    'DiffuseStrength', 0.7, 'SpecularExponent', 10,...
-    'DeleteFcn', {@mydeletefcn,Fgraph,rdr});
-
-set(Fgraph,'CurrentAxes',ax);
-view(ax,[-90 0]);
-axis(ax,'image');
-
-l = camlight; set(l,'Parent',ax);
-material(Fgraph,'dull');
-setappdata(ax,'camlight',l);
-
-%-Setup context menu
-%--------------------------------------------------------------------------
-r = rotate3d(ax);
-set(r,'enable','off');
-cmenu = uicontextmenu;
-c1 = uimenu(cmenu, 'Label', 'Inflate', 'Interruptible','off', 'Callback', @myinflate);
-setappdata(c1,'patch',hp);
-setappdata(c1,'axis',ax);
-c2 = uimenu(cmenu, 'Label', 'Connected Components', 'Interruptible','off');
-C = spm_mesh_label(hp);
-setappdata(c2,'patch',hp);
-setappdata(c2,'cclabel',C);
-for i=1:length(unique(C))
-    uimenu(c2,'Label',sprintf('Component %d',i), 'Checked','on', 'Callback', @mycclabel);
-end
-c3 = uimenu(cmenu, 'Label', 'Rotate', 'Checked','on','Separator','on','Callback', @myswitchrotate);
-setappdata(c3,'rotate3d',r);
-c4 = uimenu(cmenu, 'Label', 'View');
-setappdata(c4,'axis',ax);
-uimenu(c4,'Label','Go to Y-Z view (right)',  'Callback', {@myview, [90 0]});
-uimenu(c4,'Label','Go to Y-Z view (left)',   'Callback', {@myview, [-90 0]});
-uimenu(c4,'Label','Go to X-Y view (top)',    'Callback', {@myview, [0 90]});
-uimenu(c4,'Label','Go to X-Y view (bottom)', 'Callback', {@myview, [-180 -90]});
-uimenu(c4,'Label','Go to X-Z view (front)',  'Callback', {@myview, [-180 0]});
-uimenu(c4,'Label','Go to X-Z view (back)',   'Callback', {@myview, [0 0]});
-c5 = uimenu(cmenu, 'Label', 'Transparency');
-setappdata(c5,'patch',hp);
-uimenu(c5,'Label','0%',  'Checked','on',  'Callback', @mytransparency);
-uimenu(c5,'Label','20%', 'Checked','off', 'Callback', @mytransparency);
-uimenu(c5,'Label','40%', 'Checked','off', 'Callback', @mytransparency);
-uimenu(c5,'Label','60%', 'Checked','off', 'Callback', @mytransparency);
-uimenu(c5,'Label','80%', 'Checked','off', 'Callback', @mytransparency);
-c6 = uimenu(cmenu, 'Label', 'Background Color');
-setappdata(c6,'fig',ax0);
-uimenu(c6,'Label','White', 'Callback', {@mybackgroundcolor, [1 1 1]});
-uimenu(c6,'Label','Black', 'Callback', {@mybackgroundcolor, [0 0 0]});
-uimenu(c6,'Label','Custom...', 'Callback', {@mybackgroundcolor, []});
-c7 = uimenu(cmenu, 'Label', 'Save As...','Separator','on','Callback', @mysave);
-setappdata(c7,'patch',hp);
-setappdata(c7,'fig',Fgraph);
-setappdata(c7,'axis',ax);
-setappdata(c7,'ax0',ax0);
-try, set(r,'uicontextmenu',cmenu); end
-try, set(hp,'uicontextmenu',cmenu); end
-set(r,'enable','on');
-set(r,'ActionPostCallback',@mypostcallback);
 try
-    setAllowAxesRotate(r, setxor(findobj(Fgraph,'Type','axes'),ax), false);
+    setAllowAxesRotate(H.rotate3d, setxor(findobj(Fgraph,'Type','axes'),ax), false);
 end
     
 %-Register with MIP
@@ -435,159 +347,7 @@ function surf_rend(dat,rend,col)
     hs   = mydispcursor('Create',ax,dat.mat,xyz);
     spm_XYZreg('Add2Reg',hReg,hs,@mydispcursor);
 end
-    
-%==========================================================================
-function myinflate(obj,evd)
-spm_mesh_inflate(getappdata(obj,'patch'),Inf,1);
-axis(getappdata(obj,'axis'),'image');
-
-%==========================================================================
-function mycclabel(obj,evd)
-C = getappdata(get(obj,'parent'),'cclabel');
-F = get(getappdata(get(obj,'parent'),'patch'),'Faces');
-ind = sscanf(get(obj,'Label'),'Component %d');
-V = get(getappdata(get(obj,'parent'),'patch'),'FaceVertexAlphaData');
-Fa = get(getappdata(get(obj,'parent'),'patch'),'FaceAlpha');
-if ~isnumeric(Fa)
-    if ~isempty(V), Fa = max(V); else Fa = 1; end
-    if Fa == 0, Fa = 1; end
-end
-if isempty(V) || numel(V) == 1
-    Ve = get(getappdata(get(obj,'parent'),'patch'),'Vertices');
-    if isempty(V) || V == 1
-        V = Fa * ones(size(Ve,1),1);
-    else
-        V = zeros(size(Ve,1),1);
-    end
-end
-if strcmpi(get(obj,'Checked'),'on')
-    V(reshape(F(C==ind,:),[],1)) = 0;
-    set(obj,'Checked','off');
-else
-    V(reshape(F(C==ind,:),[],1)) = Fa;
-    set(obj,'Checked','on');
-end
-set(getappdata(get(obj,'parent'),'patch'), 'FaceVertexAlphaData', V);
-if all(V)
-    set(getappdata(get(obj,'parent'),'patch'), 'FaceAlpha', Fa);
-else
-    set(getappdata(get(obj,'parent'),'patch'), 'FaceAlpha', 'interp');
-end
-    
-%==========================================================================
-function myswitchrotate(obj,evd)
-if strcmpi(get(getappdata(obj,'rotate3d'),'enable'),'on')
-    set(getappdata(obj,'rotate3d'),'enable','off');
-    set(obj,'Checked','off');
-else
-    set(getappdata(obj,'rotate3d'),'enable','on');
-    set(obj,'Checked','on');
-end
-
-%==========================================================================
-function myview(obj,evd,varargin)
-view(getappdata(get(obj,'parent'),'axis'),varargin{1});
-axis(getappdata(get(obj,'parent'),'axis'),'image');
-camlight(getappdata(getappdata(get(obj,'parent'),'axis'),'camlight'));
-
-%==========================================================================
-function mytransparency(obj,evd)
-t = 1 - sscanf(get(obj,'Label'),'%d%%') / 100;
-set(getappdata(get(obj,'parent'),'patch'),'FaceAlpha',t);
-set(get(get(obj,'parent'),'children'),'Checked','off');
-set(obj,'Checked','on');
-
-%==========================================================================
-function mybackgroundcolor(obj,evd,varargin)
-if isempty(varargin{1})
-    c = uisetcolor(getappdata(get(obj,'parent'),'fig'), ...
-        'Pick a background color...');
-    if numel(c) == 1, return; end
-else
-    c = varargin{1};
-end
-set(getappdata(get(obj,'parent'),'fig'),'Color',c);
-
-%==========================================================================
-function mysave(obj,evd)
-[filename, pathname, filterindex] = uiputfile({...
-    '*.gii' 'GIfTI files (*.gii)'; ...
-    '*.png' 'PNG files (*.png)';...
-    '*.dae' 'Collada files (*.dae)';...
-    '*.idtf' 'IDTF files (*.idtf)'}, 'Save as');
-if ~isequal(filename,0) && ~isequal(pathname,0)
-    [pth,nam,ext] = fileparts(filename);
-    switch ext
-        case '.gii'
-            filterindex = 1;
-        case '.png'
-            filterindex = 2;
-        case '.dae'
-            filterindex = 3;
-        case '.idtf'
-            filterindex = 4;
-        otherwise
-            switch filterindex
-                case 1
-                    filename = [filename '.gii'];
-                case 2
-                    filename = [filename '.png'];
-                case 3
-                    filename = [filename '.dae'];
-            end
-    end
-    switch filterindex
-        case 1
-            g = gifti;
-            g.vertices = get(getappdata(obj,'patch'),'Vertices');
-            g.faces = get(getappdata(obj,'patch'),'Faces');
-            g.cdata = get(getappdata(obj,'patch'),'FaceVertexCData');
-            save(g,fullfile(pathname, filename));
-        case 2
-            ax = getappdata(obj,'axis');
-            u = get(ax,'units');
-            set(ax,'units','pixels');
-            p = get(ax,'Position');
-            r = get(getappdata(obj,'fig'),'Renderer');
-            h = figure('Position',p+[0 0 10 10], ...
-                'InvertHardcopy','off', ...
-                'Color',get(getappdata(obj,'ax0'),'Color'), ...
-                'Renderer',r);
-            copyobj(getappdata(obj,'axis'),h);
-            set(ax,'units',u);
-            set(get(h,'children'),'visible','off');
-            %a = get(h,'children');
-            %set(a,'Position',get(a,'Position').*[0 0 1 1]+[10 10 0 0]);       
-            if isdeployed
-                deployprint(h, '-dpng', '-opengl', fullfile(pathname, filename));
-            else
-                print(h, '-dpng', '-opengl', fullfile(pathname, filename));
-            end
-            close(h);
-            set(getappdata(obj,'fig'),'renderer',r);
-        case 3
-            g = gifti;
-            g.vertices = get(getappdata(obj,'patch'),'Vertices');
-            g.faces = get(getappdata(obj,'patch'),'Faces');
-            g.cdata = get(getappdata(obj,'patch'),'FaceVertexCData');
-            save(g,fullfile(pathname, filename),'collada');
-        case 4
-            g = gifti;
-            g.vertices = get(getappdata(obj,'patch'),'Vertices');
-            g.faces = get(getappdata(obj,'patch'),'Faces');
-            g.cdata = get(getappdata(obj,'patch'),'FaceVertexCData');
-            save(g,fullfile(pathname, filename),'idtf');
-    end
-end
-
-%==========================================================================
-function mypostcallback(obj,evd)
-try, camlight(getappdata(evd.Axes,'camlight')); end
-
-%==========================================================================
-function mydeletefcn(obj,evd,varargin)
-try, rotate3d(get(obj,'parent'),'off'); end
-set(varargin{1},'Renderer',varargin{2});
+  
 
 %==========================================================================
 function varargout = mydispcursor(varargin)
diff --git a/spm_reslice.m b/spm_reslice.m
index f46e55c..c8d5133 100644
--- a/spm_reslice.m
+++ b/spm_reslice.m
@@ -2,66 +2,73 @@ function spm_reslice(P,flags)
 % Rigid body reslicing of images
 % FORMAT spm_reslice(P,flags)
 %
-% P     - matrix of filenames {one string per row}
-%         All operations are performed relative to the first image.
-%         ie. Coregistration is to the first image, and resampling
-%         of images is into the space of the first image.
+% P      - matrix or cell array of filenames {one string per row}
+%          All operations are performed relative to the first image.
+%          ie. Coregistration is to the first image, and resampling
+%          of images is into the space of the first image.
 %
-% flags    - a structure containing various options.  The fields are:
+% flags  - a structure containing various options.  The fields are:
 %
-%         mask   - mask output images (1 for yes, 0 for no)
-%                  To avoid artifactual movement-related variance the realigned
-%                  set of images can be internally masked, within the set (i.e.
-%                  if any image has a zero value at a voxel than all images have
-%                  zero values at that voxel).  Zero values occur when regions
-%                  'outside' the image are moved 'inside' the image during
-%                  realignment.
+%         mask   - mask output images (true/false) [default: true]
+%                  To avoid artifactual movement-related variance the
+%                  realigned set of images can be internally masked, within
+%                  the set (i.e. if any image has a zero value at a voxel
+%                  than all images have zero values at that voxel). Zero
+%                  values occur when regions 'outside' the image are moved
+%                  'inside' the image during realignment.
 %
-%         mean   - write mean image (1 for yes, 0 for no)
+%         mean   - write mean image (true/false) [default: true]
 %                  The average of all the realigned scans is written to
-%                  mean*.img.
+%                  an image file with 'mean' prefix.
 %
-%         interp - the B-spline interpolation method. 
-%                  Non-finite values result in Fourier interpolation.  Note that
-%                  Fourier interpolation only works for purely rigid body
-%                  transformations.  Voxel sizes must all be identical and
-%                  isotropic.
+%         interp - the B-spline interpolation method [default: 1]
+%                  Non-finite values result in Fourier interpolation. Note
+%                  that Fourier interpolation only works for purely rigid
+%                  body transformations. Voxel sizes must all be identical
+%                  and isotropic.
 %
-%         which   - Values of 0, 1 or 2 are allowed.
+%         which  - values of 0, 1 or 2 are allowed [default: 2]
 %                  0   - don't create any resliced images.
 %                        Useful if you only want a mean resliced image.
 %                  1   - don't reslice the first image.
-%                        The first image is not actually moved, so it may not be
-%                        necessary to resample it.
+%                        The first image is not actually moved, so it may
+%                        not be necessary to resample it.
 %                  2   - reslice all the images.
+%                  If which is a 2-element vector, flags.mean will be set
+%                  to flags.which(2).
 %
-%         prefix - prefix for resliced images. Defaults to 'r'.
+%         wrap   - three values of either 0 or 1, representing wrapping in
+%                  each of the dimensions. For fMRI, [1 1 0] would be used.
+%                  For PET, it would be [0 0 0]. [default: [0 0 0]]
+%
+%         prefix - prefix for resliced images [default: 'r']
 %
-%             The spatially realigned images are written to the orginal
-%             subdirectory with the same filename but prefixed with an 'r'.
-%             They are all aligned with the first.
 %__________________________________________________________________________
 %
-% Inputs
-% A series of *.img conforming to SPM data format (see 'Data Format').  The 
-% relative displacement of the images is stored in their ".mat" files.
+% The spatially realigned images are written to the original subdirectory
+% with the same (prefixed) filename. They are all aligned with the first.
+%
+% Inputs:
+% A series of images conforming to SPM data format (see 'Data Format'). The
+% relative displacement of the images is stored in their header.
 %
-% Outputs
-% The routine uses information in these ".mat" files and writes the
-% realigned *.img files to the same subdirectory prefixed with an 'r'
-% (i.e. r*.img).
+% Outputs:
+% The routine uses information in their headers and writes the realigned 
+% image files to the same subdirectory with a prefix.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_reslice.m 4179 2011-01-28 13:57:20Z volkmar $
+
 %__________________________________________________________________________
 %
-% The `.mat' files.
+% The headers of the images contain a 4x4 affine transformation matrix 'M',
+% usually affected bu the `realignment' and `coregistration' modules.
+% What these matrices contain is a mapping from the voxel coordinates
+% (x0,y0,z0) (where the first voxel is at coordinate (1,1,1)), to 
+% coordinates in millimeters (x1,y1,z1).
 %
-% This simply contains a 4x4 affine transformation matrix in a variable `M'.
-% These files are normally generated by the `realignment' and
-% `coregistration' modules.  What these matrixes contain is a mapping from
-% the voxel coordinates (x0,y0,z0) (where the first voxel is at coordinate
-% (1,1,1)), to coordinates in millimeters (x1,y1,z1).  By default, the
-% the new coordinate system is derived from the `origin' and `vox' fields
-% of the image header.
-%  
 % x1 = M(1,1)*x0 + M(1,2)*y0 + M(1,3)*z0 + M(1,4)
 % y1 = M(2,1)*x0 + M(2,2)*y0 + M(2,3)*z0 + M(2,4)
 % z1 = M(3,1)*x0 + M(3,2)*y0 + M(3,3)*z0 + M(3,4)
@@ -71,10 +78,10 @@ function spm_reslice(P,flags)
 % (ie. from the coordinate system of image1 into millimeters, followed
 % by a mapping from millimeters into the space of image2).
 %
-% These `.mat' files allow several realignment or coregistration steps to be
-% combined into a single operation (without the necessity of resampling the
-% images several times).  The `.mat' files are also used by the spatial
-% normalisation module.
+% Several spatial transformations (realignment, coregistration,
+% normalisation) can be combined into a single operation (without the
+% necessity of resampling the images several times).
+%
 %__________________________________________________________________________
 % Refs:
 %
@@ -85,160 +92,133 @@ function spm_reslice(P,flags)
 % Registration by Using Fourier Interpolation. Mag. Res. Med. 36(6):923-931
 %
 % R. W. Cox and A. Jesmanowicz (1999)  Real-Time 3D Image Registration
-% for Functional MRI.  Submitted to MRM (April 1999) and avaliable from:
-% http://varda.biophysics.mcw.edu/~cox/index.html.
-%
+% for Functional MRI. Mag. Res. Med. 42(6):1014-1018
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
-
-% John Ashburner
-% $Id: spm_reslice.m 3756 2010-03-05 18:43:37Z guillaume $
-
 
 
-def_flags = struct('interp',1,'mask',1,'mean',1,'which',2,'wrap',[0 0 0]',...
-                   'prefix','r');
-if nargin < 2,
+def_flags        = spm_get_defaults('realign.write');
+def_flags.prefix = 'r';
+if nargin < 2
     flags = def_flags;
-else,
+else
     fnms = fieldnames(def_flags);
-    for i=1:length(fnms),
-        if ~isfield(flags,fnms{i}),
-            flags = setfield(flags,fnms{i},getfield(def_flags,fnms{i}));
-        end;
-    end;
-end;
-
-if iscell(P), P = strvcat(P{:}); end;
-if ischar(P), P = spm_vol(P); end;
+    for i=1:length(fnms)
+        if ~isfield(flags,fnms{i})
+            flags.(fnms{i}) = def_flags.(fnms{i});
+        end
+    end
+end
+
+if numel(flags.which) == 2
+    flags.mean  = flags.which(2);
+    flags.which = flags.which(1);
+elseif ~isfield(flags,'mean')
+    flags.mean  = 1; 
+end
+
+if ~nargin || isempty(P), P = spm_select([2 Inf],'image'); end
+if iscellstr(P), P = char(P);    end;
+if ischar(P),    P = spm_vol(P); end;
 reslice_images(P,flags);
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+
+%==========================================================================
 function reslice_images(P,flags)
 % Reslices images volume by volume
 % FORMAT reslice_images(P,flags)
-%
-% P        - matrix of image handles from spm_vol.
-%            All operations are performed relative to the first image.
-%            ie. resampling of images is into the space of the first image.
-%
-% flags    - a structure containing various options.  The fields are:
-%
-%         mask - mask output images (1 for yes, 0 for no)
-%                To avoid artifactual movement-related variance the realigned
-%                set of images can be internally masked, within the set (i.e.
-%                if any image has a zero value at a voxel than all images have
-%                zero values at that voxel).  Zero values occur when regions
-%                'outside' the image are moved 'inside' the image during
-%                realignment.
-%
-%         mean - write mean image
-%                The average of all the realigned scans is written to
-%                mean*.img.
-%
-%         interp - the B-spline interpolation method (see spm_bsplinc and spm_bsplins).
-%                Non-finite values result in Fourier interpolation
-%
-%         which - Values of 0, 1 or 2 are allowed.
-%                0   - don't create any resliced images.
-%                      Useful if you only want a mean resliced image.
-%                1   - don't reslice the first image.
-%                      The first image is not actually moved, so it may not be
-%                      necessary to resample it.
-%                2   - reslice all the images.
-%         wrap - three values of either 0 or 1, representing wrapping in each of
-%                the dimensions.  For fMRI, [1 1 0] would be used.  For PET, it would
-%                be [0 0 0].
-%
-%         prefix - prefix for resliced images. Defaults to 'r'.
-%
-%             The spatially realigned images are written to the orginal
-%             subdirectory with the same filename but prefixed with an 'r'.
-%             They are all aligned with the first.
+% See main function for a description of the input parameters
 
 if ~isfinite(flags.interp), % Use Fourier method
     % Check for non-rigid transformations in the matrixes
-    for i=1:prod(size(P)),
+    for i=1:numel(P)
         pp = P(1).mat\P(i).mat;
-        if any(abs(svd(pp(1:3,1:3))-1)>1e-7),
+        if any(abs(svd(pp(1:3,1:3))-1)>1e-7)
             fprintf('\n  Zooms  or shears  appear to  be needed');
             fprintf('\n  (probably due to non-isotropic voxels).');
             fprintf('\n  These  can not yet be  done  using  the');
             fprintf('\n  Fourier reslicing method.  Switching to');
             fprintf('\n  7th degree B-spline interpolation instead.\n\n');
             flags.interp = 7;
-            break;
-        end;
-    end;
-end;
+            break
+        end
+    end
+end
 
-if flags.mask | flags.mean,
+if flags.mask || flags.mean
     spm_progress_bar('Init',P(1).dim(3),'Computing available voxels','planes completed');
     x1    = repmat((1:P(1).dim(1))',1,P(1).dim(2));
     x2    = repmat( 1:P(1).dim(2)  ,P(1).dim(1),1);
-    if flags.mean,
+    if flags.mean
         Count    = zeros(P(1).dim(1:3));
         Integral = zeros(P(1).dim(1:3));
-    end;
+    end
     if flags.mask, msk = cell(P(1).dim(3),1);  end;
-    for x3 = 1:P(1).dim(3),
+    for x3 = 1:P(1).dim(3)
         tmp = zeros(P(1).dim(1:2));
-        for i = 1:prod(size(P)),
+        for i = 1:numel(P)
             tmp = tmp + getmask(inv(P(1).mat\P(i).mat),x1,x2,x3,P(i).dim(1:3),flags.wrap);
-        end;
-        if flags.mask, msk{x3} = find(tmp ~= prod(size(P))); end;
+        end
+        if flags.mask, msk{x3} = find(tmp ~= numel(P)); end;
         if flags.mean, Count(:,:,x3) = tmp; end;
         spm_progress_bar('Set',x3);
-    end;
-end;
+    end
+end
 
-nread = prod(size(P));
-if ~flags.mean,
+nread = numel(P);
+if ~flags.mean
     if flags.which == 1, nread = nread - 1; end;
     if flags.which == 0, nread = 0; end;
-end;
+end
 spm_progress_bar('Init',nread,'Reslicing','volumes completed');
 
-tiny = 5e-2; % From spm_vol_utils.c
-
 [x1,x2] = ndgrid(1:P(1).dim(1),1:P(1).dim(2));
-
-PO    = P;
-nread = 0;
-d     = [flags.interp*[1 1 1]' flags.wrap(:)];
-
-for i = 1:prod(size(P)),
-
-    if (i>1 & flags.which==1) | flags.which==2, write_vol = 1; else, write_vol = 0; end;
-    if write_vol | flags.mean,                   read_vol = 1; else   read_vol = 0; end;
-
-    if read_vol,
-        if ~isfinite(flags.interp),
+nread   = 0;
+d       = [flags.interp*[1 1 1]' flags.wrap(:)];
+
+for i = 1:numel(P)
+
+    if (i>1 && flags.which==1) || flags.which==2
+        write_vol = 1;
+    else
+        write_vol = 0;
+    end
+    if write_vol || flags.mean
+        read_vol = 1;
+    else
+        read_vol = 0;
+    end
+
+    if read_vol
+        if ~isfinite(flags.interp)
             v = abs(kspace3d(spm_bsplinc(P(i),[0 0 0 ; 0 0 0]'),P(1).mat\P(i).mat));
-            for x3 = 1:P(1).dim(3),
-                if flags.mean,
-                    Integral(:,:,x3) = Integral(:,:,x3) + ...
-                        nan2zero(v(:,:,x3).*getmask(inv(P(1).mat\P(i).mat),x1,x2,x3,P(i).dim(1:3),flags.wrap));
-                end;
-                if flags.mask, tmp = v(:,:,x3); tmp(msk{x3}) = NaN; v(:,:,x3) = tmp; end;
-            end;
-        else,
+            for x3 = 1:P(1).dim(3)
+                if flags.mean
+                    Integral(:,:,x3) = ...
+                        Integral(:,:,x3) + ...
+                        nan2zero(v(:,:,x3) .* ...
+                        getmask(inv(P(1).mat\P(i).mat),x1,x2,x3,P(i).dim(1:3),flags.wrap));
+                end
+                if flags.mask
+                    tmp = v(:,:,x3); tmp(msk{x3}) = NaN; v(:,:,x3) = tmp;
+                end
+            end
+        else
             C = spm_bsplinc(P(i), d);
             v = zeros(P(1).dim);
-            for x3 = 1:P(1).dim(3),
-
+            for x3 = 1:P(1).dim(3)
                 [tmp,y1,y2,y3] = getmask(inv(P(1).mat\P(i).mat),x1,x2,x3,P(i).dim(1:3),flags.wrap);
                 v(:,:,x3)      = spm_bsplins(C, y1,y2,y3, d);
-                % v(~tmp)        = 0;
-
-                if flags.mean, Integral(:,:,x3) = Integral(:,:,x3) + nan2zero(v(:,:,x3)); end;
-
-                if flags.mask, tmp = v(:,:,x3); tmp(msk{x3}) = NaN; v(:,:,x3) = tmp; end;
-            end;
-        end;
-        if write_vol,
+                % v(~tmp)      = 0;
+
+                if flags.mean
+                    Integral(:,:,x3) = Integral(:,:,x3) + nan2zero(v(:,:,x3));
+                end
+                if flags.mask
+                    tmp = v(:,:,x3); tmp(msk{x3}) = NaN; v(:,:,x3) = tmp;
+                end
+            end
+        end
+        if write_vol
             VO         = P(i);
             [pth,nm,xt,vr] = spm_fileparts(deblank(P(i).fname));
             VO.fname   = fullfile(pth,[flags.prefix nm xt vr]);
@@ -248,32 +228,31 @@ function reslice_images(P,flags)
             VO.mat     = P(1).mat;
             VO.descrip = 'spm - realigned';
             VO = spm_write_vol(VO,v);
-        end;
+        end
 
         nread = nread + 1;
-    end;
+    end
     spm_progress_bar('Set',nread);
-end;
+end
 
 if flags.mean
     % Write integral image (16 bit signed)
-    %-----------------------------------------------------------
-    Integral   = Integral./Count;
-    PO         = P(1);
-    PO         = rmfield(PO,'pinfo');
-    [pth,nm,xt,vr] = spm_fileparts(deblank(P(1).fname));
-    PO.fname   = fullfile(pth,['mean' nm xt]);
-    PO.pinfo   = [max(max(max(Integral)))/32767 0 0]';
-    PO.descrip = 'spm - mean image';
-    PO.dt      = [4 spm_platform('bigend')];
+    %----------------------------------------------------------------------
+    Integral    = Integral./Count;
+    PO          = P(1);
+    PO          = rmfield(PO,'pinfo');
+    [pth,nm,xt] = spm_fileparts(deblank(P(1).fname));
+    PO.fname    = fullfile(pth,['mean' nm xt]);
+    PO.pinfo    = [max(max(max(Integral)))/32767 0 0]';
+    PO.descrip  = 'spm - mean image';
+    PO.dt       = [spm_type('int16') spm_platform('bigend')];
     spm_write_vol(PO,Integral);
 end
 
 spm_figure('Clear','Interactive');
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+
+%==========================================================================
 function v = kspace3d(v,M)
 % 3D rigid body transformation performed as shears in 1D Fourier space.
 % FORMAT v1 = kspace3d(v,M)
@@ -286,70 +265,67 @@ function reslice_images(P,flags)
 % The routine is based on the excellent papers:
 % R. W. Cox and A. Jesmanowicz (1999)
 % Real-Time 3D Image Registration for Functional MRI
-% Submitted to MRM (April 1999) and avaliable from:
-% http://varda.biophysics.mcw.edu/~cox/index.html.
-% and:
+% Magnetic Resonance in Medicine 42(6):1014-1018
+%
 % W. F. Eddy, M. Fitzgerald and D. C. Noll (1996)
 % Improved Image Registration by Using Fourier Interpolation
 % Magnetic Resonance in Medicine 36(6):923-931
-%_______________________________________________________________________
+%__________________________________________________________________________
 
 [S0,S1,S2,S3] = shear_decomp(M);
 
 d  = [size(v) 1 1 1];
 g = 2.^ceil(log2(d));
-if any(g~=d),
+if any(g~=d)
     tmp = v;
     v   = zeros(g);
     v(1:d(1),1:d(2),1:d(3)) = tmp;
     clear tmp;
-end;
+end
 
 % XY-shear
 tmp1 = -sqrt(-1)*2*pi*([0:((g(3)-1)/2) 0 (-g(3)/2+1):-1])/g(3);
-for j=1:g(2),
+for j=1:g(2)
     t        = reshape( exp((j*S3(3,2) + S3(3,1)*(1:g(1)) + S3(3,4)).'*tmp1) ,[g(1) 1 g(3)]);
     v(:,j,:) = real(ifft(fft(v(:,j,:),[],3).*t,[],3));
-end;
+end
 
 % XZ-shear
 tmp1 = -sqrt(-1)*2*pi*([0:((g(2)-1)/2) 0 (-g(2)/2+1):-1])/g(2);
-for k=1:g(3),
+for k=1:g(3)
     t        = exp( (k*S2(2,3) + S2(2,1)*(1:g(1)) + S2(2,4)).'*tmp1);
     v(:,:,k) = real(ifft(fft(v(:,:,k),[],2).*t,[],2));
-end;
+end
 
 % YZ-shear
 tmp1 = -sqrt(-1)*2*pi*([0:((g(1)-1)/2) 0 (-g(1)/2+1):-1])/g(1);
-for k=1:g(3),
+for k=1:g(3)
     t        = exp( tmp1.'*(k*S1(1,3) + S1(1,2)*(1:g(2)) + S1(1,4)));
     v(:,:,k) = real(ifft(fft(v(:,:,k),[],1).*t,[],1));
-end;
+end
 
 % XY-shear
 tmp1 = -sqrt(-1)*2*pi*([0:((g(3)-1)/2) 0 (-g(3)/2+1):-1])/g(3);
-for j=1:g(2),
+for j=1:g(2)
     t        = reshape( exp( (j*S0(3,2) + S0(3,1)*(1:g(1)) + S0(3,4)).'*tmp1) ,[g(1) 1 g(3)]);
     v(:,j,:) = real(ifft(fft(v(:,j,:),[],3).*t,[],3));
-end;
+end
+
+if any(g~=d), v = v(1:d(1),1:d(2),1:d(3)); end
 
-if any(g~=d), v = v(1:d(1),1:d(2),1:d(3)); end;
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+%==========================================================================
 function [S0,S1,S2,S3] = shear_decomp(A)
 % Decompose rotation and translation matrix A into shears S0, S1, S2 and
-% S3, such that A = S0*S1*S2*S3.  The original procedure is documented
-% in:
+% S3, such that A = S0*S1*S2*S3. The original procedure is documented in:
 % R. W. Cox and A. Jesmanowicz (1999)
 % Real-Time 3D Image Registration for Functional MRI
+% Magnetic Resonance in Medicine 42(6):1014-1018
 
 A0 = A(1:3,1:3);
-if any(abs(svd(A0)-1)>1e-7), error('Can''t decompose matrix'); end;
+if any(abs(svd(A0)-1)>1e-7), error('Can''t decompose matrix'); end
 
-
-t  = A0(2,3); if t==0, t=eps; end;
+t  = A0(2,3); if t==0, t=eps; end
 a0 = pinv(A0([1 2],[2 3])')*[(A0(3,2)-(A0(2,2)-1)/t) (A0(3,3)-1)]';
 S0 = [1 0 0; 0 1 0; a0(1) a0(2) 1];
 A1 = S0\A0;  a1 = pinv(A1([2 3],[2 3])')*A1(1,[2 3])';  S1 = [1 a1(1) a1(2); 0 1 0; 0 0 1];
@@ -363,28 +339,21 @@ function reslice_images(P,flags)
 S1 = [[S1 [s1 0  0]'];[0 0 0 1]];
 S2 = [[S2 [0 s2  0]'];[0 0 0 1]];
 S3 = [[S3 [0  0  0]'];[0 0 0 1]];
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+
+%==========================================================================
 function [Mask,y1,y2,y3] = getmask(M,x1,x2,x3,dim,wrp)
 tiny = 5e-2; % From spm_vol_utils.c
 y1   = M(1,1)*x1+M(1,2)*x2+(M(1,3)*x3+M(1,4));
 y2   = M(2,1)*x1+M(2,2)*x2+(M(2,3)*x3+M(2,4));
 y3   = M(3,1)*x1+M(3,2)*x2+(M(3,3)*x3+M(3,4));
-Mask = logical(ones(size(y1)));
-if ~wrp(1), Mask = Mask & (y1 >= (1-tiny) & y1 <= (dim(1)+tiny)); end;
-if ~wrp(2), Mask = Mask & (y2 >= (1-tiny) & y2 <= (dim(2)+tiny)); end;
-if ~wrp(3), Mask = Mask & (y3 >= (1-tiny) & y3 <= (dim(3)+tiny)); end;
-return;
-%_______________________________________________________________________
-
-%_______________________________________________________________________
+Mask = true(size(y1));
+if ~wrp(1), Mask = Mask & (y1 >= (1-tiny) & y1 <= (dim(1)+tiny)); end
+if ~wrp(2), Mask = Mask & (y2 >= (1-tiny) & y2 <= (dim(2)+tiny)); end
+if ~wrp(3), Mask = Mask & (y3 >= (1-tiny) & y3 <= (dim(3)+tiny)); end
+
+
+%==========================================================================
 function vo = nan2zero(vi)
 vo = vi;
 vo(~isfinite(vo)) = 0;
-return;
-%_______________________________________________________________________
-
-%_______________________________________________________________________
-
diff --git a/spm_results_ui.m b/spm_results_ui.m
index 4d0f75f..1168403 100644
--- a/spm_results_ui.m
+++ b/spm_results_ui.m
@@ -125,7 +125,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston & Andrew Holmes
-% $Id: spm_results_ui.m 3928 2010-06-16 12:09:22Z guillaume $
+% $Id: spm_results_ui.m 4209 2011-02-22 20:11:05Z guillaume $
  
  
 %==========================================================================
@@ -237,7 +237,7 @@
 % warning statements from MATLAB.
 %__________________________________________________________________________
  
-SVNid = '$Rev: 3928 $'; 
+SVNid = '$Rev: 4209 $'; 
 
 %-Condition arguments
 %--------------------------------------------------------------------------
@@ -477,7 +477,7 @@
                 'Box','on','TickDir','out',...
                 'XTick',spm_DesRep('ScanTick',nPar,10),'XTickLabel','',...
                 'XLim', [0,nPar]+0.5,...
-                'YTick',[0:size(SPM.xCon(xSPM.Ic(ii)).c,2)]+0.5,....
+                'YTick',[0:size(SPM.xCon(xSPM.Ic(ii)).c,2)]+0.5,...
                 'YTickLabel','',...
                 'YLim', [0,size(xCon(xSPM.Ic(ii)).c,2)]+0.5 )
  
@@ -710,9 +710,7 @@
         hExit  = uicontrol(Finter,'Style','PushButton','String','exit',...
             'ToolTipString','exit the results section',...
             'FontSize',FS(9),'ForegroundColor','r',...
-            'Callback',['spm_clf(''Interactive''), spm_clf(''Graphics''),'...
-              'close(spm_figure(''FindWin'',''Satellite'')),'...
-              'clear'],...
+            'Callback','spm_results_ui(''close'')',...
             'Interruptible','on','Enable','on',...
             'Position',[325 055 035 018].*WS);
  
@@ -1130,7 +1128,16 @@
             set(HR,{'Visible'},HRv)
         end
  
- 
+        
+    %======================================================================
+    case 'close'                                            %-Close Results
+    %======================================================================
+        spm_clf('Interactive');
+        spm_clf('Graphics');
+        close(spm_figure('FindWin','Satellite'));
+        evalin('base','clear');
+    
+    
     %======================================================================
     case 'launchmp'                            %-Launch multiplanar toolbox
     %======================================================================
diff --git a/spm_rmpath.m b/spm_rmpath.m
index 46db594..f1837a9 100644
--- a/spm_rmpath.m
+++ b/spm_rmpath.m
@@ -17,7 +17,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Darren Gitelman & Guillaume Flandin
-% $Id: spm_rmpath.m 3756 2010-03-05 18:43:37Z guillaume $ 
+% $Id: spm_rmpath.m 4025 2010-07-29 11:10:15Z guillaume $ 
 
 varargout = {};
 if ~nargin
@@ -28,7 +28,7 @@
 
 % Recursively remove directories in the MATLAB path
 p = textscan(path,'%s','delimiter',pathsep); p = p{1};
-i = strmatch(d,p); P = p(i); p(i) = [];
+i = strncmp(d,p,length(d)); P = p(i); p(i) = [];
 if ~nargin && ~isempty(P)
     fprintf('Removed %s paths starting from base path: "%s"\n',spm('ver','',1),d);
 elseif ~isempty(P)
@@ -45,4 +45,4 @@
 % Return the cleaned path if requested
 if nargout
     varargout{1} = p;
-end
\ No newline at end of file
+end
diff --git a/spm_sections.m b/spm_sections.m
index 6c624fb..d419523 100644
--- a/spm_sections.m
+++ b/spm_sections.m
@@ -1,40 +1,40 @@
-function spm_sections(SPM,hReg,spms)
-% Rendering of regional effects [SPM{Z}] on orthogonal sections
-% FORMAT spm_sections(SPM,hReg)
+function spm_sections(xSPM,hReg,img)
+% Rendering of regional effects [SPM{.}] on orthogonal sections
+% FORMAT spm_sections(xSPM,hReg,img)
 %
-% SPM  - xSPM structure containing details of excursion set
-% hReg - handle of MIP register
-%
-% see spm_getSPM for details
+% xSPM  - structure containing details of excursion set (see spm_getSPM)
+% hReg  - handle of MIP register
+% img   - filename of background image
 %__________________________________________________________________________
 %
-% spm_sections is called by spm_results and uses variables in SPM and
-% VOL to create three orthogonal sections though a background image.
-% Regional foci from the selected SPM are rendered on this image.
-%
+% spm_sections is called by spm_results_ui and uses variable img to
+% create three orthogonal sections through a background image.
+% Regional foci from the selected xSPM are rendered on this image.
 %__________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 1994-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_sections.m 3081 2009-04-22 20:15:38Z guillaume $
+% $Id: spm_sections.m 4199 2011-02-10 20:07:17Z guillaume $
 
 
-if nargin < 3 || isempty(spms)
-    spms   = spm_select(1,'image','select image for rendering on');
+if ~nargin, [SPM,xSPM] = spm_getSPM; end
+if nargin < 2, hReg = []; end
+if nargin < 3 || isempty(img)
+    [img, sts] = spm_select(1,'image','Select image for rendering on');
+    if ~sts, return; end
 end
 
-Fgraph = spm_figure('FindWin','Graphics');
+Fgraph = spm_figure('GetWin','Graphics');
 spm_results_ui('Clear',Fgraph);
 spm_orthviews('Reset');
-global st
-st.Space = spm_matrix([0 0 0  0 0 -pi/2])*st.Space;
-h = spm_orthviews('Image',spms,[0.05 0.05 0.9 0.45]);
-spm_orthviews('AddContext',h); 
-spm_orthviews MaxBB;
-spm_orthviews('register',hReg);
-spm_orthviews('addblobs',1,SPM.XYZ,SPM.Z,SPM.M);
-spm_orthviews('Redraw');
 
-global prevsect
-prevsect = spms;
+global st prevsect
+st.Space = spm_matrix([0 0 0  0 0 -pi/2]) * st.Space;
+prevsect = img;
 
+h = spm_orthviews('Image', img, [0.05 0.05 0.9 0.45]);
+spm_orthviews('AddContext', h); 
+spm_orthviews('MaxBB');
+if ~isempty(hReg), spm_orthviews('Register', hReg); end
+spm_orthviews('AddBlobs', h, xSPM.XYZ, xSPM.Z, xSPM.M);
+spm_orthviews('Redraw');
diff --git a/spm_select.m b/spm_select.m
index 86f02ae..41d6154 100644
--- a/spm_select.m
+++ b/spm_select.m
@@ -72,15 +72,15 @@
 % Add directory dir to list of previous directories.
 % FORMAT dirs = spm_select('prevdirs')
 % Retrieve list of previous directories.
-%____________________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 2005-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_select.m 3948 2010-06-25 09:48:03Z volkmar $
+% $Id: spm_select.m 4185 2011-02-01 18:46:18Z guillaume $
 
 if ~isdeployed && ~exist('cfg_getfile','file')
     addpath(fullfile(spm('dir'),'matlabbatch'));
-end;
+end
 % cfg_getfile expects and returns cellstr arguments for multi-line strings
 if nargin > 0 && ischar(varargin{1}) && strcmpi(varargin{1},'filter') && ischar(varargin{2})
     varargin{2} = cellstr(varargin{2});
diff --git a/spm_smooth.m b/spm_smooth.m
index d6386db..1299aa6 100644
--- a/spm_smooth.m
+++ b/spm_smooth.m
@@ -23,11 +23,11 @@ function spm_smooth(P,Q,s,dtype)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner & Tom Nichols
-% $Id: spm_smooth.m 3976 2010-07-08 14:12:31Z karl $
+% $Id: spm_smooth.m 4172 2011-01-26 12:13:29Z guillaume $
 
 
 %-----------------------------------------------------------------------
-if length(s) == 1; s = [s s s]; end
+if numel(s) == 1, s = [s s s]; end
 if nargin < 4, dtype = 0; end;
 
 if ischar(P), P = spm_vol(P); end;
diff --git a/spm_softmax.m b/spm_softmax.m
index 31b1044..da9bb18 100644
--- a/spm_softmax.m
+++ b/spm_softmax.m
@@ -1,6 +1,6 @@
 function [y] = spm_softmax(x,k)
 % softmax (neural transfer)  function
-% FORMAT [y] = spm_softmax(x)
+% FORMAT [y] = spm_softmax(x,k)
 %
 % x - vector of activity
 % k - temperature or inverse sensitivity parameter (default k = 1)
@@ -11,7 +11,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_softmax.m 3757 2010-03-08 11:41:53Z guillaume $
+% $Id: spm_softmax.m 4187 2011-02-01 20:13:57Z karl $
  
 % apply
 %--------------------------------------------------------------------------
diff --git a/spm_spm.m b/spm_spm.m
index ef2ef02..f6f812f 100644
--- a/spm_spm.m
+++ b/spm_spm.m
@@ -281,9 +281,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Andrew Holmes, Jean-Baptiste Poline & Karl Friston
-% $Id: spm_spm.m 3960 2010-06-30 17:41:24Z ged $
+% $Id: spm_spm.m 4191 2011-02-03 13:30:02Z guillaume $
  
-SVNid     = '$Rev: 3960 $';
+SVNid     = '$Rev: 4191 $';
  
 %-Say hello
 %--------------------------------------------------------------------------
@@ -333,7 +333,9 @@
         return
     else
         warning('Overwriting old results\n\t (pwd = %s) ',SPM.swd);
-        try, SPM = rmfield(SPM,'xVol'); end
+        try, SPM     = rmfield(SPM,     'xVol'); end
+        try, SPM.xX  = rmfield(SPM.xX,  'W');    end
+        try, SPM.xVi = rmfield(SPM.xVi, 'V');    end
     end
 end
  
@@ -441,28 +443,32 @@
 xX.pKX    = spm_sp('x-',xX.xKXs);                        % projector
 erdf      = spm_SpUtil('trRV',xX.xKXs);                  % Working error df
  
-%-If xVi.V is not defined compute Hsqr and F-threshold under i.i.d.
+%-Compute Hsqr and F-threshold under i.i.d.
 %--------------------------------------------------------------------------
-if ~isfield(xVi,'V')
-    
+if isfield(xVi,'Fcontrast')
+    Fcname = 'User-specified contrast';
+    xCon   = spm_FcUtil('Set',Fcname,'F','c',xVi.Fcontrast,xX.xKXs);
+else
     Fcname = 'effects of interest';
     iX0    = [SPM.xX.iB SPM.xX.iG];
     xCon   = spm_FcUtil('Set',Fcname,'F','iX0',iX0,xX.xKXs);
-    X1o    = spm_FcUtil('X1o', xCon(1),xX.xKXs);
-    Hsqr   = spm_FcUtil('Hsqr',xCon(1),xX.xKXs);
-    trRV   = spm_SpUtil('trRV',xX.xKXs);
-    trMV   = spm_SpUtil('trMV',X1o);
- 
-    % Threshold for voxels entering non-sphericity estimates
-    %----------------------------------------------------------------------
-    try
-        modality = lower(spm_get_defaults('modality'));
-        UFp      = spm_get_defaults(['stats.' modality '.ufp']);
-    catch
-        UFp      = 0.001;
-    end
-    UF           = spm_invFcdf(1 - UFp,[trMV,trRV]);
 end
+
+X1o    = spm_FcUtil('X1o', xCon(1),xX.xKXs);
+Hsqr   = spm_FcUtil('Hsqr',xCon(1),xX.xKXs);
+trRV   = spm_SpUtil('trRV',xX.xKXs);
+trMV   = spm_SpUtil('trMV',X1o);
+
+% Threshold for voxels entering non-sphericity estimates and prior for PPM
+%----------------------------------------------------------------------
+try
+    modality = lower(spm_get_defaults('modality'));
+    UFp      = spm_get_defaults(['stats.' modality '.ufp']);
+catch
+    UFp      = 0.001;
+end
+xVi.UFp      = UFp;
+UF           = spm_invFcdf(1 - UFp,[trMV,trRV]);
  
 %-Image dimensions and data
 %==========================================================================
@@ -713,17 +719,23 @@
             %--------------------------------------------------------------
             if isfield(xX,'W')
  
+                j   = sum((Hsqr*beta).^2,1)/trMV > UF*ResSS/trRV;
+                j   = find(j);
+                
                 %-sample covariance and mean of Y (all voxels)
                 %----------------------------------------------------------
-                CY         = CY + Y*Y';
-                EY         = EY + sum(Y,2);
- 
+                if ~isempty(j)
+                    Y          = Y(:,j);
+                    CY         = CY + Y*Y';
+                    EY         = EY + sum(Y,2);
+                end
+                
                 %-Save betas etc. for current plane as we go along
                 %----------------------------------------------------------
                 CrBl       = [CrBl,    beta];
                 CrResI     = [CrResI,  res(i_res,:)];
                 CrResSS    = [CrResSS, ResSS];
- 
+                
             end % (xX,'W')
             clear Y                         %-Clear to save memory
  
@@ -802,18 +814,14 @@
     if s == 0
         spm('FigName','Stats: no significant voxels',Finter); 
         spm('Pointer','Arrow');
-        if isfield(SPM.xGX,'rg')&&~isempty(SPM.xGX.rg)
+        if isfield(SPM.xGX,'rg') && ~isempty(SPM.xGX.rg) && ~spm('CmdLine')
             figure(Finter);
             plot(SPM.xGX.rg);
-            spm('alert*',{'Please check your data'; ...
-                'There are no significant voxels';...
-                'The globals are plotted for diagnosis'});
-        else
-            spm('alert*',{'Please check your data'; ...
-                'There are no significant voxels'});
+            title('global mean per volume image');
         end
-        warning('Please check your data: There are no significant voxels.');
-        return
+        spm('alert*',{'Please check your data'; ...
+            'There are no significant voxels'});
+        error('Please check your data: There are no significant voxels.');
     end
  
     %-ReML estimate of residual correlations through hyperparameters (h)
@@ -871,13 +879,13 @@
     % If xX.W is not specified use W*W' = inv(V) to give ML estimators
     %----------------------------------------------------------------------
     if ~isfield(xX,'W')
-        if spm_matlab_version_chk('7') >=0
-            save('SPM','SPM','-V6');
+        if spm_check_version('matlab','7') >=0
+            save('SPM.mat','SPM','-V6');
         else
-            save('SPM','SPM');
+            save('SPM.mat','SPM');
         end
         clear
-        load SPM
+        load('SPM.mat');
         SPM = spm_spm(SPM);
         return
     end
@@ -898,7 +906,18 @@
 %--------------------------------------------------------------------------
 VResMS.pinfo(1) = 1/xX.trRV;
 VResMS          = spm_create_vol(VResMS);
- 
+
+%-Modify ResMS (a form of shrinkage) to avoid problems of very low variance
+%--------------------------------------------------------------------------
+try
+    if ~strcmpi(spm_get_defaults('modality'),'fmri')
+        ResMS = spm_read_vols(VResMS);
+        ResMS = ResMS + 1e-3 * max(ResMS(isfinite(ResMS)));
+        spm_write_vol(VResMS, ResMS);
+        clear ResMS
+    end
+end
+
 %-Smoothness estimates of component fields and RESEL counts for volume
 %==========================================================================
 try
@@ -956,10 +975,10 @@
  
 %-Save analysis parameters in SPM.mat file
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7') >=0
-    save('SPM','SPM','-V6');
+if spm_check_version('matlab','7') >=0
+    save('SPM.mat','SPM','-V6');
 else
-    save('SPM','SPM');
+    save('SPM.mat','SPM');
 end
  
 %==========================================================================
diff --git a/spm_spm_Bayes.m b/spm_spm_Bayes.m
index e227ba3..717908f 100644
--- a/spm_spm_Bayes.m
+++ b/spm_spm_Bayes.m
@@ -68,7 +68,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_spm_Bayes.m 3692 2010-01-21 21:43:31Z guillaume $
+% $Id: spm_spm_Bayes.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 %-Say hello
@@ -106,7 +106,7 @@
 %=======================================================================
 fprintf('%-40s: %30s','Output images','...initialising')             %-#
 
-%-Initialise oonditional estimate image files
+%-Initialise conditional estimate image files
 %-----------------------------------------------------------------------
 xX             = SPM.xX;
 [nScan nBeta]  = size(xX.X);
@@ -147,7 +147,7 @@
 end
 VHp   = spm_create_vol(VHp);
 
-fprintf('%s%30s\n',repmat(sprintf('\b'),1,30),'...initialised')      %-#
+fprintf('%s%30s',repmat(sprintf('\b'),1,30),'...initialised')        %-#
 
 
 %=======================================================================
@@ -157,7 +157,7 @@
 
 % get row u{i} and column v{i}/v0{i} indices for separable designs
 %----------------------------------------------------------------------
-s      = nHp;
+s = nHp;
 if isfield(SPM,'Sess')
     for i = 1:s
          u{i} = SPM.Sess(i).row;
@@ -176,7 +176,7 @@
 
     % Get design X and confounds X0
     %---------------------------------------------------------------
-    fprintf('%-30s- %i\n','  ReML Session',i);                   %-#
+    fprintf('%-30s\n',sprintf('  ReML Session %i',i));           %-#
     X     = xX.X(u{i}, v{i});
     X0    = xX.X(u{i},v0{i});
     [m n] = size(X);
@@ -245,9 +245,9 @@
     %-------------------------------------------------------------------
     U       = find(XYZ(3,:) == z);
     nbch    = ceil(length(U)/blksz);
-    CrBl    = zeros(nBeta,length(U));   %-conditional parameter estimates
-    CrHp    = zeros(nHp,  length(U));   %-ReML hyperparameter estimates
-    for bch = 1:nbch            %-loop over bunches of lines (planks)
+    CrBl    = zeros(nBeta,length(U)); %-conditional parameter estimates
+    CrHp    = zeros(nHp,  length(U)); %-ReML hyperparameter estimates
+    for bch = 1:nbch                  %-loop over bunches of lines (planks)
 
     %-construct list of voxels in this block
     %---------------------------------------------------------------
@@ -367,11 +367,11 @@
 SPM.VHp    = VHp;           % Filenames - hyperparameters
 SPM.PPM    = PPM;           % PPM structure
 
-if spm_matlab_version_chk('7') >=0
-    save('SPM', 'SPM', '-V6');
+if spm_check_version('matlab','7') >=0
+    save('SPM.mat', 'SPM', '-V6');
 else
-    save('SPM', 'SPM');
-end;
+    save('SPM.mat', 'SPM');
+end
 
 fprintf('%s%30s\n',repmat(sprintf('\b'),1,30),'...done')             %-#
 
diff --git a/spm_spm_ui.m b/spm_spm_ui.m
index 7e55f14..7800e98 100644
--- a/spm_spm_ui.m
+++ b/spm_spm_ui.m
@@ -482,9 +482,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Andrew Holmes
-% $Id: spm_spm_ui.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_spm_ui.m 4185 2011-02-01 18:46:18Z guillaume $
 
-SCCSid  = '$Rev: 1143 $';
+SCCSid  = '$Rev: 4185 $';
 
 %=======================================================================
 % - FORMAT specifications for programers
@@ -1485,10 +1485,10 @@
     %-------------------------------------------------------------------
     fprintf('%-40s: ','Saving SPM configuration')                    %-#
 
-    if spm_matlab_version_chk('7') >=0
-        save('SPM', 'SPM', '-V6');
+    if spm_check_version('matlab','7') >=0
+        save('SPM.mat', 'SPM', '-V6');
     else
-        save('SPM', 'SPM');
+        save('SPM.mat', 'SPM');
     end;
     fprintf('%30s\n','...SPM.mat saved')                             %-#
     varargout = {SPM};
diff --git a/spm_spm_vb.m b/spm_spm_vb.m
index 98a28d7..ba7d3cf 100644
--- a/spm_spm_vb.m
+++ b/spm_spm_vb.m
@@ -156,7 +156,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny, Nelson Trujillo-Barreto and Lee Harrison
-% $Id: spm_spm_vb.m 3055 2009-04-09 06:17:29Z volkmar $
+% $Id: spm_spm_vb.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 %-Get SPM.mat if necessary
@@ -1093,11 +1093,11 @@
 
 %-Save analysis parameters in SPM.mat file
 %-----------------------------------------------------------------------
-if spm_matlab_version_chk('7') >=0
-    save('SPM', 'SPM', '-V6');
+if spm_check_version('matlab','7') >= 0
+    save('SPM.mat', 'SPM', '-V6');
 else
-    save('SPM', 'SPM');
-end;
+    save('SPM.mat', 'SPM');
+end
 
 fprintf('%s%30s\n',repmat(sprintf('\b'),1,30),'...done')                %-#
 
diff --git a/spm_sqrtm.m b/spm_sqrtm.m
index 8cd7db6..a984127 100644
--- a/spm_sqrtm.m
+++ b/spm_sqrtm.m
@@ -1,19 +1,18 @@
 function [K] = spm_sqrtm(V)
-% Matrix square root (sqrtm) for sparse matrices
+% Matrix square root for sparse symmetric positive semi-definite matrices
 % FORMAT [K] = spm_sqrtm(V)
 %
 % This routine covers and extends sqrtm functionality by using a
 % computationally expedient approximation that can handle sparse
-% matrices
- %___________________________________________________________________________
+% symmetric positive semi-definite matrices.
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_sqrtm.m 2696 2009-02-05 20:29:48Z guillaume $
+% $Id: spm_sqrtm.m 4124 2010-11-18 16:56:53Z karl $
 
 %--------------------------------------------------------------------------
-n     = length(V);
-[u s] = spm_svd(V,1e-16);
+[u s] = spm_svd(V,0,1e-16);
 s     = sqrt(diag(s));
 m     = length(s);
 s     = sparse(1:m,1:m,s);
diff --git a/spm_standalone.m b/spm_standalone.m
new file mode 100644
index 0000000..3e4637a
--- /dev/null
+++ b/spm_standalone.m
@@ -0,0 +1,60 @@
+function spm_standalone(varargin)
+% A function to be compiled, which will run a standalone SPM.
+%
+% See MATLAB Compiler: http://www.mathworks.com/products/compiler/
+% See also config/spm_make_standalone.m
+%__________________________________________________________________________
+% Copyright (C) 2010-2011 Wellcome Trust Centre for Neuroimaging
+
+% Guillaume Flandin
+% $Id: spm_standalone.m 4237 2011-03-10 19:46:33Z guillaume $ 
+
+[v,r] = spm('Ver');
+fprintf('%s (%s): %s\n',v,r,spm('Dir'));
+
+if ~nargin, action = ''; else action = varargin{1}; end
+
+switch lower(action)
+    
+    case {'batch', 'run'}
+    %----------------------------------------------------------------------
+        spm('asciiwelcome');
+        %spm('defaults','fmri');
+        spm_jobman('initcfg');
+        if nargin == 1
+            h = spm_jobman;
+            waitfor(h,'Visible','off');
+        else
+            %spm_get_defaults('cmdline',true);
+            for i=2:nargin
+                try
+                    spm_jobman('run',varargin{i});
+                catch
+                    fprintf('Execution failed: %s', varargin{i});
+                end
+            end
+        end
+        spm('Quit');
+        
+    case 'script'
+    %----------------------------------------------------------------------
+        spm('asciiwelcome');
+        if nargin > 1
+            inputs = varargin(3:end);
+            fid = fopen(varargin{2});
+            if fid == -1, error('Cannot open %s',varargin{2}); end
+            S = fscanf(fid,'%c');
+            fclose(fid);
+            try
+                eval(S);
+            catch
+                fprintf('Execution failed: %s\n',varargin{2});
+                rethrow(lasterror);
+            end
+        end
+        
+    otherwise
+    %----------------------------------------------------------------------
+        spm(varargin{:});
+        
+end
diff --git a/spm_str_manip.m b/spm_str_manip.m
index 922237a..d43236d 100644
--- a/spm_str_manip.m
+++ b/spm_str_manip.m
@@ -3,7 +3,7 @@
 % FORMAT string_out = spm_str_manip(string_in,options)
 % string_in    - input string, string matrix, or cell array of strings
 % options      - a string of options flags, see below
-
+%
 % string_out   - output string, string matrix, or cell array of strings
 % R2           - extra output for 'c' and 'C' options
 %__________________________________________________________________________
@@ -46,7 +46,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_str_manip.m 3950 2010-06-28 10:44:02Z guillaume $
+% $Id: spm_str_manip.m 4042 2010-08-25 11:18:59Z christophe $
 
 
 if nargin<2, options=''; end
diff --git a/spm_summarise.m b/spm_summarise.m
index 2f1f29f..9e52c1c 100644
--- a/spm_summarise.m
+++ b/spm_summarise.m
@@ -8,7 +8,7 @@
 %           Or a [3 x m] matrix of voxel coordinates {mm}
 % fhandle - function handle to be applied on image data within VOI
 %           Must transform a [1 x m] array into a [1 x p] array
-%           Default is Identity (returns raw data, vectorised intro rows)
+%           Default is Identity (returns raw data, vectorised into rows)
 %
 % Y       - [n x p] data summary
 % xY      - (updated) VOI structure
@@ -16,7 +16,7 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin, Ged Ridgway
-% $Id: spm_summarise.m 3960 2010-06-30 17:41:24Z ged $
+% $Id: spm_summarise.m 4013 2010-07-22 17:12:45Z guillaume $
 
 %-Argument checks
 %--------------------------------------------------------------------------
diff --git a/spm_surf.m b/spm_surf.m
index 7475034..9b1ae0a 100644
--- a/spm_surf.m
+++ b/spm_surf.m
@@ -53,9 +53,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_surf.m 3081 2009-04-22 20:15:38Z guillaume $
+% $Id: spm_surf.m 4185 2011-02-01 18:46:18Z guillaume $
 
-SVNrev = '$Rev: 3081 $';
+SVNrev = '$Rev: 4185 $';
 
 spm('FnBanner',mfilename,SVNrev);
 spm('FigName','Surface');
@@ -194,7 +194,7 @@ function renviews(V,oname)
 linfun('Rendering: Coronal 2..');    rend{6} = make_struct(V,[pi/2 pi/2 pi]);
 
 linfun('Rendering: Save..');
-if spm_matlab_version_chk('7') >=0
+if spm_check_version('matlab','7') >= 0
     save(oname,'-V6','rend');
 else
     save(oname,'rend');
diff --git a/spm_svd.m b/spm_svd.m
index afc13cc..f2c0c00 100644
--- a/spm_svd.m
+++ b/spm_svd.m
@@ -12,19 +12,14 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_svd.m 1265 2008-03-28 11:45:04Z john $
+% $Id: spm_svd.m 4124 2010-11-18 16:56:53Z karl $
 
 
 
 % default thresholds
 %---------------------------------------------------------------------------
-if nargin < 2
-    U = 1e-6;
-end
-
-if nargin < 3
-    T = 0;
-end
+if nargin < 2, U = 1e-6; end
+if nargin < 3, T = 0;    end
 
 % deal with sparse matrices
 %---------------------------------------------------------------------------
diff --git a/spm_type.m b/spm_type.m
index bb6b272..c56b99c 100644
--- a/spm_type.m
+++ b/spm_type.m
@@ -1,72 +1,65 @@
 function T = spm_type(x, arg)
-% translates data type specifiers between SPM & Matlab representations
+% Translates data type specifiers between SPM & MATLAB representations
 % FORMAT T = spm_type(x, arg)
 % x    - specifier
 % T    - type
-% arg  - optional string argument, can be
-%    - 'maxval'  - return maximum allowed value.
-%    - 'minval'  - return minimum allowed value.
-%    - 'nanrep'  - return 1 if there is a NaN representation.
-%    - 'bits'    - return the number of bits per voxel.
-%    - 'intt'    - return 1 if values rounded to nearest integer.
-%_______________________________________________________________________
+% arg  - optional string argument, can be:
+%         - 'maxval'  - return maximum allowed value.
+%         - 'minval'  - return minimum allowed value.
+%         - 'nanrep'  - return 1 if there is a NaN representation.
+%         - 'bits'    - return the number of bits per voxel.
+%         - 'intt'    - return 1 if values rounded to nearest integer.
+%__________________________________________________________________________
 %
-% Format specifiers are based on NIFTI-1.  If the input is
-% a number then the corresponding matlab string is returned by default.
+% Format specifiers are based on NIFTI-1.
+% If the input is a number then the corresponding MATLAB string is
+% returned by default.
 % If the input is a string then the appropriate TYPE is returned.
 % However, if the optional arg argument is supplied then other
 % information will be returned instead.
 %
 % With no arguments, a list of data types is returned.
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1996-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner & Andrew Holmes
-% $Id: spm_type.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_type.m 4156 2011-01-11 19:03:31Z guillaume $
 
+prec   = {'uint8','int16','int32','float32','float64','int8','uint16','uint32'};
+types  = [    2      4      8   16   64   256    512    768];
+maxval = [2^8-1 2^15-1 2^31-1  Inf  Inf 2^7-1 2^16-1 2^32-1];
+minval = [    0  -2^15  -2^31 -Inf -Inf  -2^7      0      0];
+nanrep = [    0      0      0    1    1     0      0      0];
+bits   = [    8     16     32   32   64     8     16     32];
+intt   = [    1      1      1    0    0     1      1      1];
 
-
-prec = char('uint8','int16','int32','float32','float64','int8','uint16','uint32');
-types   = [    2      4      8   16   64   256    512    768];
-maxval  = [2^8-1 2^15-1 2^31-1  Inf  Inf 2^7-1 2^16-1 2^32-1];
-minval  = [    0  -2^15  -2^31 -Inf -Inf  -2^7      0      0];
-nanrep  = [    0      0      0    1    1     0      0      0];
-bits    = [    8     16     32   32   64     8     16     32];
-intt    = [    1      1      1    0    0     1      1      1];
-
-if nargin==0,
-    T=types;
+if ~nargin
+    T = types;
     return;
-end;
+end
 
-if ischar(x),
-    sel = [];
-    for i=1:numel(types),
-        if strcmpi(deblank(prec(i,:)),deblank(x)), 
-            sel = i;
-            break;
-        end;
-    end;
-else,
+if ischar(x)
+    sel = find(strcmpi(prec,deblank(x)));
+else
     sel = find(types == x);
-end;
-if nargin == 1,
-    if ischar(x),
+end
+if nargin == 1
+    if ischar(x)
         if isempty(sel), T = NaN;
-        else, T = types(sel); end;
-    else,
+        else T = types(sel); end
+    else
         if isempty(sel), T = 'unknown';
-        else, T = deblank(prec(sel,:)); end;
-    end;
-elseif isempty(sel),
+        else T = prec{sel}; end
+    end
+elseif isempty(sel)
     T = NaN;
-else,
+else
     switch lower(arg)
-    case 'maxval',  T = maxval(sel);
-    case 'minval',  T = minval(sel);
-    case 'nanrep',  T = nanrep(sel);
-    case 'bits',    T = bits(sel);
-    case 'intt',    T = intt(sel);
-    otherwise,      T = NaN;
-    end;
-end;
+        case 'maxval', T = maxval(sel);
+        case 'minval', T = minval(sel);
+        case 'nanrep', T = nanrep(sel);
+        case 'bits',   T = bits(sel);
+        case 'intt',   T = intt(sel);
+        otherwise,     T = NaN;
+    end
+end
diff --git a/spm_uitable.m b/spm_uitable.m
index eb3831c..3a7c3fe 100644
--- a/spm_uitable.m
+++ b/spm_uitable.m
@@ -65,16 +65,16 @@
 %     See also AWTCREATE, AWTINVOKE, JAVACOMPONENT, UITREE, UITREENODE
 
 %   Copyright 2002-2006 The MathWorks, Inc.
-%   $Revision: 2925 $  $Date: 2006/11/29 21:53:13 $
+%   $Revision: 4185 $  $Date: 2006/11/29 21:53:13 $
 
 %   Release: R14. This feature will not work in previous versions of MATLAB.
 
-% $Id: spm_uitable.m 2925 2009-03-23 20:49:24Z jean $
+% $Id: spm_uitable.m 4185 2011-02-01 18:46:18Z guillaume $
 
 % Setup and P-V parsing
 
 if isempty(varargin)
-    if ~isempty(javachk('awt')) || spm_matlab_version_chk('7.3') <= 0
+    if ~isempty(javachk('awt')) || spm_check_version('matlab','7.3') <= 0
         varargout{1} = 'off';
     else
         varargout{1} = 'on';
@@ -82,7 +82,7 @@
     return
 end
 
-if ~isempty(javachk('awt')) || spm_matlab_version_chk('7.3') <= 0
+if ~isempty(javachk('awt')) || spm_check_version('matlab','7.3') <= 0
     varargout{1} = [];
     varargout{2} = [];
     return;
diff --git a/spm_update.m b/spm_update.m
index 4b597fc..98a4dc1 100644
--- a/spm_update.m
+++ b/spm_update.m
@@ -9,10 +9,10 @@ function spm_update(update)
 % Invoking this function with any input parameter will do the same as
 % above but will also attempt to download and install the updates.
 %__________________________________________________________________________
-% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 2010-2011 Wellcome Trust Centre for Neuroimaging
 
 % Guillaume Flandin
-% $Id: spm_update.m 4009 2010-07-21 13:36:20Z guillaume $
+% $Id: spm_update.m 4289 2011-04-04 14:47:35Z guillaume $
 
 url = 'ftp://ftp.fil.ion.ucl.ac.uk/spm/spm8_updates/';
 
@@ -38,7 +38,7 @@ function spm_update(update)
     error('SPM cannot be found in MATLAB path.');
 end
 if ~strcmp(v,'SPM8'), error('Your SPM version is %s and not SPM8',v); end
-rs = [3042 3164 3408 3684 4010];
+rs = [3042 3164 3408 3684 4010 4290];
 if isnan(r), r = rs(1); end 
 if floor(r) == 8
     try
diff --git a/spm_uw_apply.m b/spm_uw_apply.m
index c07f2d6..a69790a 100644
--- a/spm_uw_apply.m
+++ b/spm_uw_apply.m
@@ -120,32 +120,16 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jesper Andersson
-% $Id: spm_uw_apply.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: spm_uw_apply.m 4152 2011-01-11 14:13:35Z volkmar $
 
 tiny = 5e-2;
 
-def_flags = struct('mask',       1,...
-                   'mean',       1,...
-                   'interp',     4,...
-                   'wrap',       [0 1 0],...
-                   'which',      1,...
-                   'udc',        1,...
-                   'prefix',     'u');
+def_flags        = spm_get_defaults('realign.write');
+def_flags.udc    = 1;
+def_flags.prefix = 'u';
 
 defnames = fieldnames(def_flags);
 
-%
-% Replace hardcoded defaults with spm_defaults
-% when exist and defined.
-%
-defaults = spm_get_defaults;
-if isfield(defaults,'realign') && isfield(defaults.realign,'write')
-   wd = defaults.realign.write;
-   if isfield(wd,'interp'),    def_flags.interp = wd.interp; end
-   if isfield(wd,'wrap'),      def_flags.wrap = wd.wrap; end
-   if isfield(wd,'mask'),      def_flags.mask = wd.mask; end
-end
-
 if nargin < 1 || isempty(ds)
     ds = load(spm_select(1,'.*uw\.mat$','Select Unwarp result file'),'ds');
     ds = ds.ds;
@@ -161,23 +145,19 @@
 
 %
 % Replace defaults with user supplied values for all fields
-% defined by user. Also, warn user of any invalid fields,
-% probably reflecting misspellings.
+% defined by user. 
 %
 if nargin < 2 || isempty(flags)
    flags = def_flags;
 end
 for i=1:length(defnames)
    if ~isfield(flags,defnames{i})
-      %flags = setfield(flags,defnames{i},getfield(def_flags,defnames{i}));
       flags.(defnames{i}) = def_flags.(defnames{i});
    end
 end
-flagnames = fieldnames(flags);
-for i=1:length(flagnames)
-   if ~isfield(def_flags,flagnames{i})
-      warning('Warning, unknown flag field %s',flagnames{i});
-   end
+if numel(flags.which) == 2
+    flags.mean  = flags.which(2);
+    flags.which = flags.which(1);
 end
 
 ntot = 0;
diff --git a/spm_uw_estimate.m b/spm_uw_estimate.m
index f4e598d..a831ca4 100644
--- a/spm_uw_estimate.m
+++ b/spm_uw_estimate.m
@@ -169,46 +169,35 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jesper Andersson
-% $Id: spm_uw_estimate.m 3770 2010-03-10 10:36:55Z chloe $
+% $Id: spm_uw_estimate.m 4152 2011-01-11 14:13:35Z volkmar $
 
-if nargin < 1 | isempty(P), P = spm_select(Inf,'image'); end
+if nargin < 1 || isempty(P), P = spm_select(Inf,'image'); end
 if ~isstruct(P), P = spm_vol(P); end
 
 %
 % Hardcoded default input parameters.
 %
-defpar = struct('order',           [12 12],...
-                'sfP',             [],...
+defpar = struct('sfP',             [],...
                 'M',               P(1).mat,...
-                'regorder',        1,...
-                'lambda',          1e5,...
-                'jm',              0,...
                 'fot',             [4 5],...
                 'sot',             [],...
-                'fwhm',            4,...
-                'rem',             1,...
-                'exp_round',       'Average',...
-                'noi',             5,...
                 'hold',            [1 1 1 0 1 0]);
 
-defnames = fieldnames(defpar);
-
 %
-% Replace hardcoded defaults with spm_defaults
-% when exist and defined.
-%
-defaults = spm_get_defaults;
-if isfield(defaults,'unwarp') && isfield(defaults.unwarp,'estimate')
-   ud = defaults.unwarp.estimate;
-   if isfield(ud,'basfcn'),    defpar.order = ud.basfcn; end
-   if isfield(ud,'regorder'),  defpar.regorder = ud.regorder; end
-   if isfield(ud,'regwgt'),    defpar.lambda = ud.regwgt; end
-   if isfield(ud,'jm'),        defpar.jm = ud.jm; end
-   if isfield(ud,'fwhm'),      defpar.fwhm = ud.fwhm; end
-   if isfield(ud,'rem'),       defpar.rem = ud.rem; end
-   if isfield(ud,'noi'),       defpar.noi = ud.noi; end
-   if isfield(ud,'expround'),  defpar.exp_round = ud.expround; end
-end
+% Merge hardcoded defaults and spm_defaults. Translate spm_defaults
+% settings to internal defaults.
+%
+ud = spm_get_defaults('unwarp.estimate');
+if isfield(ud,'basfcn'),    defpar.order = ud.basfcn; end
+if isfield(ud,'regorder'),  defpar.regorder = ud.regorder; end
+if isfield(ud,'regwgt'),    defpar.lambda = ud.regwgt; end
+if isfield(ud,'jm'),        defpar.jm = ud.jm; end
+if isfield(ud,'fwhm'),      defpar.fwhm = ud.fwhm; end
+if isfield(ud,'rem'),       defpar.rem = ud.rem; end
+if isfield(ud,'noi'),       defpar.noi = ud.noi; end
+if isfield(ud,'expround'),  defpar.exp_round = ud.expround; end
+
+defnames = fieldnames(defpar);
 
 %
 % Go through input parameters, chosing the default
@@ -217,21 +206,21 @@
 % reflecting a misspelling).
 %
 
-if nargin < 2 | isempty(par)
+if nargin < 2 || isempty(par)
    par = defpar;
 end
 ds = [];
 for i=1:length(defnames)
-   if isfield(par,defnames{i}) & ~isempty(getfield(par,defnames{i}))
-      ds = setfield(ds,defnames{i},getfield(par,defnames{i}));
+   if isfield(par,defnames{i}) && ~isempty(par.(defnames{i}))
+      ds.(defnames{i}) = par.(defnames{i});
    else
-      ds = setfield(ds,defnames{i},getfield(defpar,defnames{i}));
+      ds.(defnames{i}) = defpar.(defnames{i});
    end
 end
 parnames = fieldnames(par);
 for i=1:length(parnames)
    if ~isfield(defpar,parnames{i})
-      warning(sprintf('Unknown par field %s',parnames{i}));
+      warning('Unknown par field %s',parnames{i});
    end
 end
 
diff --git a/spm_vol.m b/spm_vol.m
index b5e01e3..8c45ec9 100644
--- a/spm_vol.m
+++ b/spm_vol.m
@@ -1,5 +1,5 @@
 function V = spm_vol(P)
-% Get header information etc for images.
+% Get header information for images.
 % FORMAT V = spm_vol(P)
 % P - a matrix of filenames.
 % V - a vector of structures containing image volume information.
@@ -30,63 +30,57 @@
 % Note that spm_vol can also be applied to the filename(s) of 4-dim
 % volumes. In that case, the elements of V will point to a series of 3-dim
 % images.
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_vol.m 2237 2008-09-29 17:39:53Z guillaume $
-
-if nargin==0,
-    V   = struct('fname', {},...
-                 'dim',   {},...
-                 'dt',    {},...
-                 'pinfo', {},...
-                 'mat',   {},...
-                 'n',     {},...
-                 'descrip', {},...
-                 'private',{});
+% $Id: spm_vol.m 4045 2010-08-26 15:10:46Z guillaume $
+
+if ~nargin
+    V = struct('fname',   {},...
+               'dim',     {},...
+               'dt',      {},...
+               'pinfo',   {},...
+               'mat',     {},...
+               'n',       {},...
+               'descrip', {},...
+               'private', {});
     return;
-end;
+end
 
-% If is already a vol structure then just return;
-if isstruct(P), V = P; return; end;
+% If is already a vol structure then just return
+if isstruct(P), V = P; return; end
 
 V = subfunc2(P);
-return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+
+%==========================================================================
 function V = subfunc2(P)
-if iscell(P),
+if iscell(P)
     V = cell(size(P));
-    for j=1:numel(P),
-        if iscell(P{j}),
+    for j=1:numel(P)
+        if iscell(P{j})
             V{j} = subfunc2(P{j});
         else
             V{j} = subfunc1(P{j});
-        end;
-    end;
+        end
+    end
 else
     V = subfunc1(P);
-end;
-return;
-%_______________________________________________________________________
+end
 
-%_______________________________________________________________________
+%==========================================================================
 function V = subfunc1(P)
-if isempty(P),
-    V = [];
-    return;
-end;
-
+if isempty(P), V = []; return; end
 counter = 0;
-for i=1:size(P,1),
+for i=1:size(P,1)
     v = subfunc(P(i,:));
     [V(counter+1:counter+size(v, 2),1).fname] = deal('');
-    [V(counter+1:counter+size(v, 2),1).mat] = deal([0 0 0 0]);
-    [V(counter+1:counter+size(v, 2),1).mat] = deal(eye(4));
-    [V(counter+1:counter+size(v, 2),1).mat] = deal([1 0 0]');
-    if isempty(v),
+    [V(counter+1:counter+size(v, 2),1).dim]   = deal([0 0 0 0]);
+    [V(counter+1:counter+size(v, 2),1).mat]   = deal(eye(4));
+    [V(counter+1:counter+size(v, 2),1).pinfo] = deal([1 0 0]');
+    [V(counter+1:counter+size(v, 2),1).dt]    = deal([0 0]);
+    if isempty(v)
         hread_error_message(P(i,:));
         error(['Can''t get volume information for ''' P(i,:) '''']);
     end
@@ -98,45 +92,36 @@
     counter = counter + size(v,2);
 end
 
-return;
-%_______________________________________________________________________
-
-%_______________________________________________________________________
+%==========================================================================
 function V = subfunc(p)
 [pth,nam,ext,n1] = spm_fileparts(deblank(p));
 p = fullfile(pth,[nam ext]);
 n = str2num(n1);
-if ~spm_existfile(p),
-    %existance_error_message(p);
+if ~spm_existfile(p)
     error('File "%s" does not exist.', p);
 end
-switch ext,
-    case {'.nii','.NII'},
+switch ext
+    case {'.nii','.NII'}
         % Do nothing
 
-    case {'.img','.IMG'},
-        if ~spm_existfile(fullfile(pth,[nam '.hdr'])) && ~spm_existfile(fullfile(pth,[nam '.HDR'])),
-            %existance_error_message(fullfile(pth,[nam '.hdr'])),
+    case {'.img','.IMG'}
+        if ~spm_existfile(fullfile(pth,[nam '.hdr'])) && ...
+           ~spm_existfile(fullfile(pth,[nam '.HDR']))
             error('File "%s" does not exist.', fullfile(pth,[nam '.hdr']));
         end
 
-    case {'.hdr','.HDR'},
+    case {'.hdr','.HDR'}
         ext = '.img';
         p   = fullfile(pth,[nam ext]);
-        if ~spm_existfile(p),
-            %existance_error_message(p),
+        if ~spm_existfile(p)
             error('File "%s" does not exist.', p);
         end
 
-    otherwise,
+    otherwise
         error('File "%s" is not of a recognised type.', p);
 end
 
-if isempty(n),
-    V = spm_vol_nifti(p);
-else
-    V = spm_vol_nifti(p,n);
-end;
+V = spm_vol_nifti(p,n);
     
 if isempty(n) && length(V.private.dat.dim) > 3
     V0(1) = V;
@@ -145,11 +130,10 @@
     end
     V = V0;
 end
-if ~isempty(V), return; end;
+if ~isempty(V), return; end
 return;
-%_______________________________________________________________________
 
-%_______________________________________________________________________
+%==========================================================================
 function hread_error_message(q)
 str = {...
     'Error reading information on:',...
@@ -158,15 +142,3 @@ function hread_error_message(q)
     'Please check that it is in the correct format.'};
 spm('alert*',str,mfilename,sqrt(-1));
 return;
-%_______________________________________________________________________
-
-%_______________________________________________________________________
-function existance_error_message(q)
-str = {...
-        'Unable to find file:',...
-        ['        ',spm_str_manip(q,'k40d')],...
-        ' ',...
-        'Please check that it exists.'};
-spm('alert*',str,mfilename,sqrt(-1));
-return;
-
diff --git a/spm_write_filtered.m b/spm_write_filtered.m
index 18e55e3..89d9bba 100644
--- a/spm_write_filtered.m
+++ b/spm_write_filtered.m
@@ -5,57 +5,58 @@
 % Z       - {1 x ?} vector point list of SPM values for MIP
 % XYZ     - {3 x ?} matrix of coordinates of points (voxel coordinates)
 % DIM     - image dimensions {voxels}
-% M       - voxels - > mm matrix (used for header file)
-%           [default spm_matrix(-(DIM+1)/2) ]
-% descrip - description string [default 'SPM-filtered']
+% M       - voxels -> mm matrix [default: spm_matrix(-(DIM+1)/2)]
+% descrip - description string [default: 'SPM-filtered']
+% F       - output file's basename [default: user query]
 %
-%-----------------------------------------------------------------------
+% Vo      - output image volume information
+%__________________________________________________________________________
 %
-% spm_write_filtered takes a pointlist image (parallel matrixes of
+% spm_write_filtered takes a pointlist image (parallel matrices of
 % co-ordinates and voxel intensities), and writes it out into an image
 % file.
 %
 % It is intended for writing out filtered SPM's from the results
 % section of SPM, but can be used freestanding.
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1996-2011 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_write_filtered.m 3100 2009-05-06 19:00:39Z guillaume $
+% $Id: spm_write_filtered.m 4178 2011-01-27 15:12:53Z guillaume $
 
 
 %-Parse arguments
-%-----------------------------------------------------------------------
-if nargin<3, error('Insufficient arguments'), end
+%--------------------------------------------------------------------------
+if nargin<3, error('Insufficient arguments'); end
 if nargin<4, M = spm_matrix(-(DIM+1)/2); end
-if nargin<5, descrip='SPM-filtered'; end
+if nargin<5, descrip = 'SPM-filtered'; end
 if nargin<6, F = spm_input('Output filename',1,'s'); end
 
 %-Get filename
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 Q = [spm_str_manip(F,'sd'), '.img'];
 spm('Pointer','Watch')
 
 %-Set up header information
-%-----------------------------------------------------------------------
-Vo      = struct(...
+%--------------------------------------------------------------------------
+Vo     = struct(...
         'fname',    Q,...
         'dim',      DIM',...
-        'dt',       [spm_type('uint8') spm_platform('bigend')],...
+        'dt',       [spm_type('float32') spm_platform('bigend')],...
         'mat',      M,...
         'descrip',  descrip);
 
 %-Reconstruct (filtered) image from XYZ & Z pointlist
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 Y      = zeros(DIM(1:3)');
 OFF    = XYZ(1,:) + DIM(1)*(XYZ(2,:)-1 + DIM(2)*(XYZ(3,:)-1));
 Y(OFF) = Z.*(Z > 0);
 
 %-Write the reconstructed volume
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 Vo = spm_write_vol(Vo,Y);
 spm('alert"',{'Written:',['    ',spm_select('CPath',Q)]},mfilename,1);
 
 %-End
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 spm('Pointer','Arrow');
diff --git a/spm_write_sn.m b/spm_write_sn.m
index fad7b1a..92b716b 100644
--- a/spm_write_sn.m
+++ b/spm_write_sn.m
@@ -1,8 +1,8 @@
 function VO = spm_write_sn(V,prm,flags,extras)
-% Write Out Warped Images.
-% FORMAT VO = spm_write_sn(V,matname,flags,msk)
+% Write out warped images
+% FORMAT VO = spm_write_sn(V,prm,flags,msk)
 % V         - Images to transform (filenames or volume structure).
-% matname   - Transformation information (filename or structure).
+% prm       - Transformation information (filename or structure).
 % flags     - flags structure, with fields...
 %           interp   - interpolation method (0-7)
 %           wrap     - wrap edges (e.g., [1 1 0] for 2D MRI sequences)
@@ -25,22 +25,22 @@
 %
 % Don't use interpolation methods greater than one for data containing
 % NaNs.
-% _______________________________________________________________________
+%__________________________________________________________________________
 %
-% FORMAT msk = spm_write_sn(V,matname,flags,'mask')
-% V         - Images to transform (filenames or volume structure).
-% matname   - Transformation information (filename or structure).
-% flags     - flags structure, with fields...
-%           wrap     - wrap edges (e.g., [1 1 0] for 2D MRI sequences)
-%           vox      - voxel sizes (3 element vector - in mm)
-%                      Non-finite values mean use template vox.
-%           bb       - bounding box (2x3 matrix - in mm)
-%                      Non-finite values mean use template bb.
-% msk       - a cell array for masking a series of spatially normalised
-%             images.
+% FORMAT msk = spm_write_sn(V,prm,flags,'mask')
+% V          - Images to transform (filenames or volume structure).
+% prm        - Transformation information (filename or structure).
+% flags      - flags structure, with fields...
+%            wrap - wrap edges (e.g., [1 1 0] for 2D MRI sequences)
+%            vox  - voxel sizes (3 element vector - in mm)
+%                   Non-finite values mean use template vox.
+%            bb   - bounding box (2x3 matrix - in mm)
+%                   Non-finite values mean use template bb.
+% msk        - a cell array for masking a series of spatially normalised
+%              images.
 %
 %
-% _______________________________________________________________________
+%_________________________________________________________________________
 %
 % FORMAT VO = spm_write_sn(V,prm,'modulate')
 % V         - Spatially normalised images to modulate (filenames or
@@ -58,11 +58,11 @@
 %  the total counts from any structure are preserved.  It was written
 %  as an optional step in performing voxel based morphometry.
 %
-%_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+%__________________________________________________________________________
+% Copyright (C) 1996-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_write_sn.m 2534 2008-12-08 10:16:46Z christophe $
+% $Id: spm_write_sn.m 4201 2011-02-15 10:52:00Z ged $
 
 
 if isempty(V), return; end;
@@ -79,8 +79,8 @@
     return;
 end;
 
-def_flags = struct('interp',1,'vox',NaN,'bb',NaN,'wrap',[0 0 0],'preserve',0,...
-                   'prefix','w');
+def_flags        = spm_get_defaults('normalise.write');
+def_flags.prefix = 'w';
 
 if nargin < 3,
     flags = def_flags;
@@ -128,9 +128,9 @@
 end;
 
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function VO = affine_transform(V,prm,x,y,z,mat,flags,msk)
 
 [X,Y] = ndgrid(x,y);
@@ -172,9 +172,9 @@
 end;
 spm_progress_bar('Clear');
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function VO = nonlin_transform(V,prm,x,y,z,mat,flags,msk)
 
 [X,Y] = ndgrid(x,y);
@@ -206,7 +206,7 @@
 
     for j=1:length(z),   % Cycle over planes
         % Nonlinear deformations
-        %----------------------------------------------------------------------------
+        %------------------------------------------------------------------
         tx = get_2Dtrans(Tr(:,:,:,1),BZ,j);
         ty = get_2Dtrans(Tr(:,:,:,2),BZ,j);
         tz = get_2Dtrans(Tr(:,:,:,3),BZ,j);
@@ -225,8 +225,8 @@
             j21 = DX*ty*BY';     j22 = BX*ty*DY' + 1; j23 = BX*get_2Dtrans(Tr(:,:,:,2),DZ,j)*BY';
             j31 = DX*tz*BY';     j32 = BX*tz*DY';     j33 = BX*get_2Dtrans(Tr(:,:,:,3),DZ,j)*BY' + 1;
 
-            % The determinant of the Jacobian reflects relative volume changes.
-            %------------------------------------------------------------------
+            % The determinant of the Jacobian reflects relative volume changes
+            %-----------------------------------------------------------------
             dat       = dat .* (j11.*(j22.*j33-j23.*j32) - j21.*(j12.*j33-j13.*j32) + j31.*(j12.*j23-j13.*j22)) * detAff;
             Dat(:,:,j) = single(dat);
         end;
@@ -244,9 +244,9 @@
 end;
 spm_progress_bar('Clear');
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function VO = modulate(V,prm)
 
 spm_progress_bar('Init',numel(V),'Modulating','volumes completed');
@@ -288,8 +288,8 @@
             j21 = DX*ty*BY';     j22 = BX*ty*DY' + 1; j23 = BX*get_2Dtrans(Tr(:,:,:,2),DZ,j)*BY';
             j31 = DX*tz*BY';     j32 = BX*tz*DY';     j33 = BX*get_2Dtrans(Tr(:,:,:,3),DZ,j)*BY' + 1;
 
-            % The determinant of the Jacobian reflects relative volume changes.
-            %------------------------------------------------------------------
+            % The determinant of the Jacobian reflects relative volume changes
+            %-----------------------------------------------------------------
             dat        = spm_slice_vol(V(i),spm_matrix([0 0 j]),V(i).dim(1:2),0);
             dat        = dat .* (j11.*(j22.*j33-j23.*j32) - j21.*(j12.*j33-j13.*j32) + j31.*(j12.*j23-j13.*j22)) * detAff;
             Dat(:,:,j) = single(dat);
@@ -309,9 +309,9 @@
 end;
 spm_progress_bar('Clear');
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function VO   = make_hdr_struct(V,x,y,z,mat,prefix)
 VO            = V;
 VO.fname      = prepend(V.fname,prefix);
@@ -320,24 +320,24 @@
 VO.pinfo      = V.pinfo;
 VO.descrip    = 'spm - 3D normalized';
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function T2 = get_2Dtrans(T3,B,j)
 d   = [size(T3) 1 1 1];
 tmp = reshape(T3,d(1)*d(2),d(3));
 T2  = reshape(tmp*B(j,:)',d(1),d(2));
 return;
-%_______________________________________________________________________
+%==========================================================================
 
 %_______________________________________________________________________
 function PO = prepend(PI,pre)
 [pth,nm,xt,vr] = spm_fileparts(deblank(PI));
 PO             = fullfile(pth,[pre nm xt vr]);
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function Mask = getmask(X,Y,Z,dim,wrp)
 % Find range of slice
 tiny = 5e-2;
@@ -346,9 +346,9 @@
 if ~wrp(2), Mask = Mask & (Y >= (1-tiny) & Y <= (dim(2)+tiny)); end;
 if ~wrp(3), Mask = Mask & (Z >= (1-tiny) & Z <= (dim(3)+tiny)); end;
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function [X2,Y2,Z2] = mmult(X1,Y1,Z1,Mult)
 if length(Z1) == 1,
     X2= Mult(1,1)*X1 + Mult(1,2)*Y1 + (Mult(1,3)*Z1 + Mult(1,4));
@@ -360,12 +360,12 @@
     Z2= Mult(3,1)*X1 + Mult(3,2)*Y1 + Mult(3,3)*Z1 + Mult(3,4);
 end;
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function msk = get_snmask(V,prm,x,y,z,wrap)
 % Generate a mask for where there is data for all images
-%-----------------------------------------------------------------------
+%--------------------------------------------------------------------------
 msk = cell(length(z),1);
 t1 = cat(3,V.mat);
 t2 = cat(1,V.dim);
@@ -383,20 +383,20 @@
         Count = zeros(length(x),length(y));
         if isempty(Tr),
             % Generate a mask for where there is data for all images
-            %----------------------------------------------------------------------------
+            %--------------------------------------------------------------
             for i=1:numel(V),
                 [X2,Y2,Z2] = mmult(X,Y,z(j),V(i).mat\prm.VF(1).mat*prm.Affine);
                 Count      = Count + getmask(X2,Y2,Z2,V(i).dim(1:3),wrap);
             end;
         else
             % Nonlinear deformations
-            %----------------------------------------------------------------------------
+            %--------------------------------------------------------------
             X1 = X    + BX*get_2Dtrans(Tr(:,:,:,1),BZ,j)*BY';
             Y1 = Y    + BX*get_2Dtrans(Tr(:,:,:,2),BZ,j)*BY';
             Z1 = z(j) + BX*get_2Dtrans(Tr(:,:,:,3),BZ,j)*BY';
 
             % Generate a mask for where there is data for all images
-            %----------------------------------------------------------------------------
+            %--------------------------------------------------------------
             for i=1:numel(V),
                 [X2,Y2,Z2] = mmult(X1,Y1,Z1,V(i).mat\prm.VF(1).mat*prm.Affine);
                 Count      = Count + getmask(X2,Y2,Z2,V(i).dim(1:3),wrap);
@@ -410,9 +410,9 @@
     for j=1:length(z), msk{j} = uint32([]); end;
 end;
 return;
-%_______________________________________________________________________
+%==========================================================================
 
-%_______________________________________________________________________
+%==========================================================================
 function [x,y,z,mat] = get_xyzmat(prm,bb,vox,VG)
 % The old voxel size and origin notation is used here.
 % This requires that the position and orientation
@@ -438,7 +438,7 @@
     end
 end
 
-[bb0,vox0] = bbvox_from_V(VG);
+[bb0 vox0] = spm_get_bbox(VG, 'old');
 if ~all(isfinite(vox(:))), vox = vox0; end;
 if ~all(isfinite(bb(:))),  bb  = bb0;  end;
 
@@ -480,20 +480,9 @@
     x   = flipud(x(:))';
 end;
 return;
-%_______________________________________________________________________
-
-%_______________________________________________________________________
-function [bb,vx] = bbvox_from_V(V)
-vx = sqrt(sum(V.mat(1:3,1:3).^2));
-if det(V.mat(1:3,1:3))<0, vx(1) = -vx(1); end;
+%==========================================================================
 
-o  = V.mat\[0 0 0 1]';
-o  = o(1:3)';
-bb = [-vx.*(o-1) ; vx.*(V.dim(1:3)-o)];
-return;
-%_______________________________________________________________________
-
-%_______________________________________________________________________
+%==========================================================================
 function VO = write_dets(P,bb,vox)
 if nargin==1,
     job = P;
@@ -534,7 +523,7 @@
         ty = get_2Dtrans(Tr(:,:,:,2),BZ,j);
         tz = get_2Dtrans(Tr(:,:,:,3),BZ,j);
 
-        %----------------------------------------------------------------------------
+        %------------------------------------------------------------------
         j11 = DX*tx*BY' + 1; j12 = BX*tx*DY';     j13 = BX*get_2Dtrans(Tr(:,:,:,1),DZ,j)*BY';
         j21 = DX*ty*BY';     j22 = BX*ty*DY' + 1; j23 = BX*get_2Dtrans(Tr(:,:,:,2),DZ,j)*BY';
         j31 = DX*tz*BY';     j32 = BX*tz*DY';     j33 = BX*get_2Dtrans(Tr(:,:,:,3),DZ,j)*BY' + 1;
@@ -549,4 +538,4 @@
 end;
 spm_progress_bar('Clear');
 return;
-%_______________________________________________________________________
+%==========================================================================
diff --git a/src/Makefile b/src/Makefile
index b195c53..9ced6df 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -3,7 +3,7 @@
 #
 # Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 #
-# $Id: Makefile 3251 2009-07-06 17:29:44Z guillaume $
+# $Id: Makefile 4199 2011-02-10 20:07:17Z guillaume $
 #
 ###############################################################################
 #
@@ -65,6 +65,9 @@ all: verb.$(SUF) $(SPMMEX) verb.end
 
 clean: verb.clean
 	$(DEL) $(OBS)
+ifeq (mex,$(SUF))
+	$(DEL) $(subst .$(SUF),.o,$(SPMMEX)) spm_matfuns.o
+endif
 
 distclean: clean verb.distclean
 	$(DEL) $(SPMMEX) spm_vol_utils.$(SUF).a
@@ -387,6 +390,13 @@ verb.mexs64:
 	@ echo "_____________________________________________________________"
 	@ echo ""
 
+verb.mex:
+	@ echo "_____________________________________________________________"
+	@ echo ""
+	@ echo "        ${PLATFORM} compilation (`octave -v | head -n 1`)"
+	@ echo "_____________________________________________________________"
+	@ echo ""
+
 verb.external:
 	@ echo "_____________________________________________________________"
 	@ echo ""
diff --git a/src/Makefile.var b/src/Makefile.var
index 6886805..391eac6 100644
--- a/src/Makefile.var
+++ b/src/Makefile.var
@@ -2,7 +2,7 @@
 #
 # Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 #
-# $Id: Makefile.var 3775 2010-03-11 19:26:20Z guillaume $
+# $Id: Makefile.var 4136 2010-12-09 22:22:28Z guillaume $
 #
 ###############################################################################
 #
@@ -125,6 +125,13 @@ ifndef SUF
       $(error Unknowm platform $(PLATFORM)-$(SUNBIT))
     endif
   endif
+  #### Octave ####
+  ifeq (octave,$(PLATFORM))
+    MEXBIN     = mkoctfile
+    MEXOPTS    = --mex
+    SUF        = mex
+    override PLATFORM = $(shell $(UNAME))
+  endif
   ifndef SUF
     $(error Unknowm platform $(PLATFORM))
   endif
diff --git a/src/file2mat.c b/src/file2mat.c
index 027e1b7..7347d4c 100644
--- a/src/file2mat.c
+++ b/src/file2mat.c
@@ -1,5 +1,5 @@
 /*
- * $Id: file2mat.c 3392 2009-09-11 14:13:38Z guillaume $
+ * $Id: file2mat.c 4136 2010-12-09 22:22:28Z guillaume $
  * John Ashburner
  */
 
@@ -16,6 +16,7 @@ Memory mapping is used by this module. For more information on this, see:
 #include <stdlib.h>
 #include <sys/types.h>
 #include <string.h>
+#include <stdio.h>
 #include "mex.h"
 
 #ifdef SPM_WIN32
diff --git a/src/spm_bwlabel.c b/src/spm_bwlabel.c
index c6876c7..d3415cd 100644
--- a/src/spm_bwlabel.c
+++ b/src/spm_bwlabel.c
@@ -1,5 +1,5 @@
 /*
- * $Id: spm_bwlabel.c 1790 2008-06-05 11:27:02Z spm $
+ * $Id: spm_bwlabel.c 4178 2011-01-27 15:12:53Z guillaume $
  * Jesper Andersson
  */
 
@@ -23,8 +23,9 @@
  ***************************************************************/
 
 #include <math.h>
-#include "mex.h"
+#include <stdio.h>
 #include <limits.h>
+#include "mex.h"
 
 /* Silly little macros. */
 
@@ -361,7 +362,7 @@ void mexFunction(int             nlhs,      /* No. of output arguments */
 
    nl = translate_labels(il,dim,tt,ttn,l);
 
-   plhs[1] = mxCreateScalarDouble(nl);
+   plhs[1] = mxCreateDoubleScalar(nl);
 
    /* Clean up a bit. */
 
diff --git a/src/spm_existfile.c b/src/spm_existfile.c
index 0ae1859..887e476 100644
--- a/src/spm_existfile.c
+++ b/src/spm_existfile.c
@@ -1,15 +1,16 @@
 /*
- * $Id: spm_existfile.c 2267 2008-09-30 21:13:55Z guillaume $
+ * $Id: spm_existfile.c 4145 2010-12-23 15:18:30Z guillaume $
  * Guillaume Flandin
  */
  
 #define _FILE_OFFSET_BITS 64
 
 #include "mex.h"
+#include <stdio.h>
 
 void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
 {
-    bool status    = false;
+    int status     = 0;
     char *filename = NULL;
     FILE *fid      = NULL;
     
@@ -23,7 +24,7 @@ void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
         fid = fopen(filename,"r");
         if (fid != NULL)
         {
-            status = true;
+            status = 1;
             fclose(fid);
         }
         mxFree(filename);
diff --git a/src/spm_mapping.c b/src/spm_mapping.c
index b6f968b..0abf87d 100644
--- a/src/spm_mapping.c
+++ b/src/spm_mapping.c
@@ -1,10 +1,11 @@
 /*
- * $Id: spm_mapping.c 2026 2008-08-29 10:41:22Z john $
+ * $Id: spm_mapping.c 4136 2010-12-09 22:22:28Z guillaume $
  * John Ashburner
  */
 
 /* Matlab dependent high level data access and map manipulation routines */
 
+#include <stdio.h>
 #include <math.h>
 #include <fcntl.h>
 #include <sys/stat.h>
diff --git a/src/spm_mesh_utils.c b/src/spm_mesh_utils.c
new file mode 100644
index 0000000..3483193
--- /dev/null
+++ b/src/spm_mesh_utils.c
@@ -0,0 +1,179 @@
+/*
+ * $Id: spm_mesh_utils.c 4081 2010-10-07 14:04:44Z guillaume $
+ * Guillaume Flandin
+ */
+
+#include <string.h>
+#include "mex.h"
+
+/* --- Compute geodesic distance using Dijkstra algorithm --- */
+void dijkstra(double *nghbr, double *dnghbr, int nv, int nb,
+              int *source, int ns,
+              double maxdist,
+              double *dist) {
+    int i, nQ, u = 0, v, Qu = 0;
+    double dmin, alt;
+    int *Q = mxMalloc(nv*sizeof(int));
+    
+    for (i=0;i<nv;i++) {
+        dist[i] = mxGetInf();
+        Q[i] = i;
+    }
+    for (i=0;i<ns;i++)
+        dist[source[i]] = 0.0;
+    nQ = nv;
+    
+    while(nQ) {
+        for (i=0,dmin=mxGetInf();i<nQ;i++) {
+            v = Q[i];
+            if (dist[v] < dmin) {
+                dmin = dist[v];
+                Qu = i;
+                u = v;
+            }
+        }
+        if ((mxIsInf(dmin)) || dmin > maxdist) break;
+        Q[Qu] = Q[--nQ];
+        for (i=0;i<nb;i++) {
+            v = (int)nghbr[u+nv*i] - 1;
+            if (v < 0) break;
+            alt = dmin + dnghbr[u+nv*i];
+            if (alt < dist[v])
+                dist[v] = alt;
+        }
+    }
+    mxFree(Q);
+}
+
+/* Gateway Function for Volume */
+void mexFunctionVolume(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
+{
+    double vol, x1, x2, x3, y1, y2, y3, z1, z2, z3; 
+    double *f, *v;
+    mxArray *array;
+    int i, nv, nf;
+    
+    if (nrhs == 0) mexErrMsgTxt("Not enough input arguments.");
+    if (nrhs > 1) mexErrMsgTxt("Too many input arguments.");
+    if ((!mxIsStruct(prhs[0])) || (mxIsClass(prhs[0],"gifti")))
+        mexErrMsgTxt("First argument must be a patch structure.");
+
+    array = mxGetField(prhs[0], 0, "vertices");
+    if (!mxIsDouble(array))
+        mexErrMsgTxt("Vertices have to be stored as double.");
+    nv    = mxGetM(array);
+    v     = mxGetPr(array);
+    array = mxGetField(prhs[0], 0, "faces");
+    if (!mxIsDouble(array))
+        mexErrMsgTxt("Faces have to be stored as double.");
+    nf    = mxGetM(array);
+    f     = mxGetPr(array);
+    
+    for (i=0,vol=0;i<nf;i++) {
+        x1 = v[(int)f[i]];      y1 = v[(int)f[i]+nv];      z1 = v[(int)f[i]+2*nv];
+        x2 = v[(int)f[i+nf]];   y2 = v[(int)f[i+nf]+nv];   z2 = v[(int)f[i+nf]+2*nv];
+        x3 = v[(int)f[i+2*nf]]; y3 = v[(int)f[i+2*nf]+nv]; z3 = v[(int)f[i+2*nf]+2*nv];
+        vol += 1.0/6.0 * (-x3*y2*z1 + x2*y3*z1 + x3*y1*z2 - x1*y3*z2 - x2*y1*z3 + x1*y2*z3);
+    }
+    vol = (vol < 0)? -vol: vol;
+    plhs[0] = mxCreateDoubleScalar(vol);
+}
+
+/* Gateway Function for Neighbours */
+void mexFunctionNeighbours(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
+{
+    mwSize i, j, k, n, d = 0;
+    mwIndex *Ir, *Jc;
+    double *N = NULL, *D = NULL, *pr = NULL;
+    
+    if (nrhs == 0) mexErrMsgTxt("Not enough input arguments.");
+    if (nrhs > 1) mexErrMsgTxt("Too many input arguments.");
+    if (!mxIsSparse(prhs[0])) mexErrMsgTxt("First argument must be a sparse array.");
+
+    plhs[0] = mxCreateDoubleMatrix(mxGetM(prhs[0]), d, mxREAL);
+    N = mxGetPr(plhs[0]);
+    if (nlhs > 1) {
+        plhs[1] = mxCreateDoubleMatrix(mxGetM(prhs[0]), d, mxREAL);
+        D = mxGetPr(plhs[1]);
+    }
+    
+    n  = mxGetM(prhs[0]);
+    Ir = mxGetIr(prhs[0]);
+    Jc = mxGetJc(prhs[0]);
+    pr = mxGetPr(prhs[0]);
+    
+    for (i=0;i<n;i++) {
+        k = Jc[i+1]-Jc[i];
+        if (k > d) {
+            d = k; N = mxRealloc(N, mxGetM(prhs[0])*d*sizeof(double));
+            if (nlhs > 1) D = mxRealloc(D, mxGetM(prhs[0])*d*sizeof(double));
+        }
+        for (j=0;j<k;j++) N[i+n*j] = 1 + *Ir++;
+        if (nlhs > 1)
+            for (j=0;j<k;j++) D[i+n*j] = *pr++;
+    }
+    
+    mxSetPr(plhs[0],N); mxSetN(plhs[0],d);
+    if (nlhs > 1) { mxSetPr(plhs[1],D); mxSetN(plhs[1],d); }
+}
+
+/* Gateway Function for Dijkstra */
+void mexFunctionDijkstra(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
+{
+    int i, nv, nb, ns, *source;
+    double distmax;
+    
+    if (nrhs < 4) mexErrMsgTxt("Not enough input arguments.");
+    if (nrhs > 4) mexErrMsgTxt("Too many input arguments.");
+    if (!mxIsNumeric(prhs[0]))
+        mexErrMsgTxt("First argument must be a neighbour array.");
+    if (!mxIsNumeric(prhs[1]))
+        mexErrMsgTxt("Second argument must be a distance array.");
+    
+    nv = mxGetM(prhs[0]);
+    nb = mxGetN(prhs[0]);
+    
+    ns = mxGetNumberOfElements(prhs[2]);
+    source = mxMalloc(ns*sizeof(double));
+    for (i=0;i<ns;i++) {
+        source[i] = (int)mxGetPr(prhs[2])[i];
+        if ((source[i]<0) || (source[i]>=nv)) mexErrMsgTxt("Invalid vertex index.");
+    }
+    
+    distmax = mxGetScalar(prhs[3]);
+    
+    plhs[0] = mxCreateDoubleMatrix(nv, 1, mxREAL);
+    
+    dijkstra(mxGetPr(prhs[0]), mxGetPr(prhs[1]), nv, nb, source, ns, distmax, mxGetPr(plhs[0]));
+    
+    mxFree(source);
+}
+
+/* Main Gateway Function */
+void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
+{
+    char *action = NULL;
+    mwSize strlgh;
+    
+    if (nrhs < 2) mexErrMsgTxt("Not enough input arguments.");
+    if (!mxIsChar(prhs[0])) mexErrMsgTxt("First argument must be an action string.");
+    
+    strlgh = (mxGetM(prhs[0]) * mxGetN(prhs[0]) * sizeof(char)) + 1;
+    action = mxCalloc(strlgh, sizeof(char));
+    mxGetString(prhs[0], action, strlgh);
+    
+    if (!strcmp(action,"dijkstra")) {
+        mexFunctionDijkstra(nlhs,plhs,nrhs-1,&prhs[1]);
+    }
+    else if (!strcmp(action,"neighbours")) {
+        mexFunctionNeighbours(nlhs,plhs,nrhs-1,&prhs[1]);
+    }
+    else if (!strcmp(action,"volume")) {
+        mexFunctionVolume(nlhs,plhs,nrhs-1,&prhs[1]);    
+    }
+    else {
+        mexErrMsgTxt("Unknown action.");
+    }
+    
+    mxFree(action);
+}
diff --git a/toolbox/Beamforming/spm_eeg_ft_beamformer_cva.m b/toolbox/Beamforming/spm_eeg_ft_beamformer_cva.m
new file mode 100644
index 0000000..efe2082
--- /dev/null
+++ b/toolbox/Beamforming/spm_eeg_ft_beamformer_cva.m
@@ -0,0 +1,1055 @@
+function [stats,talpositions,gridpositions,grid,fftnewdata,alllf,allepochdata]=spm_eeg_ft_beamformer_cva(S)
+% Computes power-based beamformer image
+% FORMAT [outfilenames,ctf_inside,ctf_weights]=spm_eeg_ft_beamformer_cva (S)
+%
+% S         MEEG object where coregistration has been performed.
+%
+% 
+% Outputs (1) normalised power, (2) t-stat and (3) multivarariate
+% (Hotellings)images. Uses Sekihara eigenval approach to choose optimal
+% direction.
+%
+% outfilenames       Output filenames (for 1,2,3)
+%
+%                    The following fields are returned if you set 
+%                    S.return_weights=1:
+%
+% ctf_inside         CTF locations inside head
+% ctf_weights        Corresponding beamformer weight matrices
+%                   fftnewdata is fft of data in trial order
+% _______________________________________________________________________
+% Copyright (C) 2009 Institute of Neurology, UCL
+
+% Gareth Barnes
+% $Id: spm_eeg_ft_beamformer_cva.m 4217 2011-02-25 14:04:49Z gareth $
+
+[Finter,Fgraph] = spm('FnUIsetup','Multivariate LCMV beamformer for power', 0);
+%%
+
+%% ============ Load SPM EEG file and verify consistency
+if nargin == 0
+    S = [];
+end
+
+try
+    D = S.D;
+catch
+    D = spm_select(1, '\.mat$', 'Select EEG mat file');
+    S.D = D;
+end
+
+if ischar(D)
+    try
+        D = spm_eeg_load(D);
+    catch
+        error(sprintf('Trouble reading file %s', D));
+    end
+end
+
+[ok, D] = check(D, 'sensfid');
+
+if ~ok
+    if check(D, 'basic')
+        errordlg(['The requested file is not ready for source reconstruction.'...
+            'Use prep to specify sensors and fiducials.']);
+    else
+        errordlg('The meeg file is corrupt or incomplete');
+    end
+    return
+end
+
+modality = spm_eeg_modality_ui(D, 1, 1);
+%try
+    channel_labels = D.chanlabels(strmatch(modality, D.chantype))';
+%catch
+%    warning('ASSUMING ALL CHANNELS ARE MEG CHANNELS');
+%    channel_labels=D.chanlabels;
+%end;
+
+if isfield(S, 'refchan') && ~isempty(S.refchan)
+    refchan = S.refchan;
+else
+    refchan = [];
+end
+
+%% ============ Find or prepare head model
+
+if ~isfield(D, 'val')
+    D.val = 1;
+end
+
+   if ~isfield(S,'filenamestr'),
+       S.filenamestr=[];
+   end;%
+%D.inv{1}.forward(1).voltype
+
+ try
+     vol = D.inv{D.val}.forward.vol;
+     datareg = D.inv{D.val}.datareg;
+ catch
+     D = spm_eeg_inv_mesh_ui(D, D.val, [], 1);
+     D = spm_eeg_inv_datareg_ui(D, D.val);
+     datareg = D.inv{D.val}.datareg;
+ end
+ 
+for m = 1:numel(D.inv{D.val}.forward)
+    if strncmp(modality, D.inv{D.val}.forward(m).modality, 3)
+        vol  = D.inv{D.val}.forward(m).vol;
+        if isa(vol, 'char')
+            vol = fileio_read_vol(vol);
+        end
+        datareg  = D.inv{D.val}.datareg(m); %%  NEED TO KNOW HOW TO HANDLE DATAREG
+        disp('USING LAST FORWARD MODEL !');
+    end
+end
+
+
+
+% Return beamformer weights
+if ~isfield(S,'return_weights')
+    ctf_weights=[];
+    S.return_weights=0;
+    alllf=[];
+end
+
+if ~isfield(S,'Niter')
+    S.Niter=[];
+end; % if 
+
+if isempty(S.Niter),
+    S.Niter=1;
+end; % if
+
+
+if ~isfield(S,'design'),
+   error('Design matrix required');
+end; % if
+
+X=S.design.X;
+c=S.design.contrast; %% c is contrast eg [ 0 1 -1] compare columns 2,3 of X
+
+if ~isfield(S,'ttest'),
+    S.ttest=[];
+    end; % if
+
+if S.ttest,
+    TWOSAMPLETEST=1
+else
+    TWOSAMPLETEST=0
+    end;
+    
+try S.design.X(:,1)-S.design.Xtrials-S.design.Xstartlatencies;
+    catch
+    error('Design windows missepcified');
+end;
+
+X0  = X - X*c*pinv(c);  %% make sure X0 is orthogonal to X
+Xdesign   = full(X*c);
+X0  = spm_svd(X0); %% X0 is null space i.e. everything that is happening in other columns of X
+
+
+outfilenames='';
+
+
+
+
+
+
+freqbands=[];
+if ~isfield(S, 'freqbands')
+    for i = 1:spm_input('Number of frequency bands:', '+1', 'r', '1', 1)
+        outstr=sprintf('Band %d [st end] in Hz ',i);
+        S.freqbands{i} = spm_input(outstr, '+1', 'r', '', 2);
+        freqbands =[freqbands;S.freqbands{i}'];
+    end
+else 
+    freqbands=cell2mat(S.freqbands');
+end
+
+if ~isfield(S,'testbands'),
+    S.testbands=[];
+end;
+
+if isempty(S.testbands),
+    S.testbands=S.freqbands; %% bands to do the test on
+end; % if
+    
+    
+Nbands=numel(S.freqbands);
+
+if ~isfield(S,'weightspect'), 
+    S.weightspect=[];
+end; 
+
+if ~isfield(S,'weightttest'), 
+    S.weightttest=[];
+end; 
+
+
+
+
+% if ~isempty(S.weightspect),
+%         if Nbands>1,
+%             error('cannot weight spectra for more than one freqband');
+%         end; % if 
+% end; % if
+
+if ~isfield(S,'gridpos'),
+    if ~isfield(S,'gridstep');
+    S.gridstep = spm_input('Grid step (mm):', '+1', 'r', '5');
+    end; 
+end; % if
+
+if ~isfield(S,'rankflag'),
+    S.rankflag=[];
+end; % if
+if isempty(S.rankflag),
+    S.rankflag=0;
+end; % if
+
+if ~isfield(S,'detrend'),
+    S.detrend=[];
+    end;
+    
+ if isempty(S.detrend),
+     disp('detrending data by default');
+     S.detrend=1;
+ end; % 
+
+ if ~isfield(S,'hanning'),
+    S.hanning=[];
+    end;
+    
+ 
+ if isempty(S.hanning),
+     disp('windowing data by default');
+     S.hanning=1;
+ end; % 
+
+if ~isfield(S,'logflag'),
+    S.logflag=[];
+end; % if
+if isempty(S.logflag),
+    S.logflag=0;
+end; % if
+%  
+
+if ~isfield(S,'regpc'),
+    S.regpc=[];
+end; % if
+if isempty(S.regpc),
+    S.regpc=0;
+end; % if
+
+%% READ IN JUST THE DATA WE need
+%% ASSUME THAT INPUT IS AS FOLLOWS
+%% a list of trial types and latency ranges for 2 conditions/periods (active and
+%% rest for now)
+%% each condition has an associated time window which must of equal sizes
+%% SO will have
+%% a list of trial indices and latencies, a list of trial types of the same
+%% length
+
+try
+    data = D.ftraw; %% convert to field trip format- direct memory map 
+catch
+    disp('failed to read data directly.. going slowly');
+   data = D.ftraw(0); %% convert to field trip format - file pointers
+end; 
+%% Check latencies are same here
+
+%% now read in the first trial of data just to get sizes of variables right
+cfg=[];
+cfg.keeptrials='no';
+cfg.trials=1; 
+
+
+Ntrials=size(S.design.X,1);
+cfg.latency=[S.design.Xstartlatencies(1) S.design.Xstartlatencies(1)+S.design.Xwindowduration];
+
+subdata=ft_timelockanalysis(cfg,data); 
+Nsamples=length(subdata.time);
+%channel_labels = D.chanlabels(D.meegchannels(modality));
+Nchans=length(channel_labels);
+dfe=Ntrials-rank(X);
+
+if S.hanning,
+    fftwindow=hamming(Nsamples);
+else
+    disp('not windowing');
+    fftwindow=ones(Nsamples,1);
+end; 
+allfftwindow=repmat(fftwindow,1,Nchans);
+NumUniquePts = ceil((Nsamples+1)/2); %% data is real so fft is symmetric
+
+fftnewdata=zeros(Ntrials,NumUniquePts,Nchans);
+allepochdata=zeros(Ntrials,Nchans,Nsamples); %% for loading in data quickly
+
+%fftnewdata=zeros(Ntrials,Nsamples,Nchans);
+
+fHz = (0:NumUniquePts-1)*D.fsample/Nsamples;
+
+if ~isfield(S,'Nfeatures'),
+     S.Nfeatures=[];
+end;
+
+
+if isempty(S.Nfeatures),
+    Nfeatures=floor(Ntrials/3);
+else
+    Nfeatures=S.Nfeatures;
+end;
+
+if ~isfield(S,'write_epochs'),
+    S.write_epochs=[];
+end;
+
+
+%% now read in all trialtype and hold them as windowed fourier transforms
+[uniquewindows]=unique(S.design.Xstartlatencies);
+Nwindows=length(uniquewindows);
+    
+    
+for i=1:Nwindows,     %% puts trials into epoch data according to order of design.X structures
+    winstart=uniquewindows(i); %% window start
+    cfg=[];
+    cfg.keeptrials='yes';
+    cfg.channel=channel_labels;
+    cfg.feedback='off';
+    useind=find(winstart==S.design.Xstartlatencies); %% indices into design.X structures
+    cfg.trials=S.design.Xtrials(useind); %% trials starting at these times
+    cfg.latency=[winstart winstart+S.design.Xwindowduration];
+    subdata=ft_timelockanalysis(cfg,data); % subset of original data
+    allepochdata(useind,:,:)=squeeze(subdata.trial); %% get an epoch of data with channels in columns
+end; % for i
+% 
+% for i=1:Nwindows,
+%     winstart=uniquewindows(i); %% window start
+%     cfg=[];
+%     cfg.keeptrials='yes';
+%     cfg.channel=channel_labels;
+%     cfg.feedback='off';
+%     cfg.trials=S.design.Xtrials(find(winstart==S.design.Xstartlatencies)); %% trials starting at these times
+%     cfg.latency=[winstart winstart+S.design.Xwindowduration];
+%     subdata=ft_timelockanalysis(cfg,data); % subset of original data
+%     allepochdata(cfg.trials,:,:)=squeeze(subdata.trial); %% get an epoch of data with channels in columns
+% end; % for i
+
+
+for i=1:Ntrials, %% read in all individual trial types
+    epochdata=squeeze(allepochdata(i,:,:))'; %% get an epoch of data with channels in columns
+    if S.detrend==1
+        dtepochdata=detrend(epochdata); %% detrend epoch data, this includes removind dc level. NB. This will have an effect on specific evoked response components !
+        else 
+        dtepochdata=epochdata; %% no dc removal, no detrend : this will have effect on accuracy of fourier estimate at non dc bins
+        end; % detrend 
+    wdtepochfft=dtepochdata.*allfftwindow; %% windowed
+    
+    epochfft=fft(wdtepochfft);
+    fftnewdata(i,:,:)=epochfft(1:NumUniquePts,:); % .*filtervect';    
+    
+end; 
+if size(fftnewdata,3)~=Nchans,
+    size(fftnewdata)
+    error('Data dimension mismatch');
+end;
+
+
+
+%% now have an fft for each channel in each condition
+
+    
+% %%
+cfg                       = [];
+if strcmp('EEG', modality)
+    cfg.elec = D.inv{D.val}.datareg.sensors;
+    cfg.reducerank=3;
+else
+    cfg.grad = D.sensors('MEG');
+    cfg.reducerank=2;
+    disp('Reducing possible source orientations to a tangential plane for MEG');
+end
+
+
+cfg.channel = channel_labels;
+cfg.vol                   = vol;
+
+if ~isfield(S,'gridpos'),
+    S.gridpos=[];
+    end;
+if ~isfield(S,'gridori'),
+    S.gridori=[];
+end;
+
+ if ~isfield(S,'maskgrid'),
+    S.maskgrid=[];
+    end;
+
+    
+ if ~isfield(S,'dimauto'),
+    S.dimauto=[];
+    end;
+ 
+    if ~isfield(S,'maskfile'),
+        S.maskfile=[];
+    end;
+    
+    if ~isfield(S,'randomweight'),
+        S.randomweight=0;
+    end;
+
+ if ~isfield(S,'fixedweights'),
+        S.fixedweights=[];
+    end;
+
+if  isempty(S.gridpos),
+    cfg.resolution            = S.gridstep;
+else
+    disp('USING pre-specified gridpoints');
+    cfg.grid.pos=S.gridpos; %% predefined grid
+    cfg.grid.inside=[1:size(S.gridpos,1)]; %% assume all in head
+    cfg.grid.outside=[];
+    end;
+
+    
+if ~isfield(S,'bootstrap'),
+    S.bootstrap=[];
+else
+    Nboot=S.bootstrap;
+    end;
+if isempty(S.bootstrap),
+    S.bootstrap=0;
+    Nboot=1;
+    end;
+    
+cfg.feedback='off';
+cfg.inwardshift           = 0; % mm
+
+if ~isfield(S,'grid'),
+    disp('preparing leadfield');
+    grid                      = ft_prepare_leadfield(cfg);
+else
+    disp('Using precomputed leadfield');
+    grid=S.grid;
+end; % if 
+
+
+maskedgrid_inside_ind=[1:length(grid.inside)];
+if ~isempty(S.maskgrid),
+    if length(S.maskgrid)~=length(grid.inside),
+        error('mask and grid points must be of same size');
+        end;
+    maskedgrid_inside_ind=find(S.maskgrid==1); %% indices into grid.inside 
+    end; 
+
+ if ~isempty(S.maskfile),
+     if ~isempty(S.maskgrid),
+         error('cannot have two masks defined');
+        end;
+     alltalpositions = spm_eeg_inv_transform_points(datareg.toMNI, grid.pos(grid.inside,:));
+      V_aal = spm_vol(S.maskfile);
+     [Y_aal,XYZ]=spm_read_vols(V_aal);
+      mask_coords=XYZ(:,find(Y_aal>0));
+      
+    
+    %% now express mni grid and mesh grid at same coarse scale to get
+    %% intersection
+      coarse_mm=10;
+      mask_coarse=unique(round(mask_coords'/coarse_mm)*coarse_mm,'rows');
+      grid_coarse= round(alltalpositions/coarse_mm)*coarse_mm;
+      [overlap_pos,maskedgrid_inside_ind]=intersect(grid_coarse,mask_coarse,'rows'); %% mesh_vert_ind are mesh points in this mask
+      maskedgrid_inside_ind=sort(maskedgrid_inside_ind);
+ end; % if 
+
+  if cfg.reducerank, %% follow up rank reduction and remove redundant dimension from lead fields
+    for i=1:length(maskedgrid_inside_ind), %% 81
+        lf1=cell2mat(grid.leadfield(grid.inside(maskedgrid_inside_ind(i))));
+        [u1,s1,v1]=svd(lf1'*lf1);
+        grid.leadfield(grid.inside(maskedgrid_inside_ind(i)))={lf1*u1(:,1:cfg.reducerank)};
+        %normlf(i)=std(dot(lfnew',lfnew'));
+    end;
+ end; % if reduce rank
+
+if ~isempty(S.fixedweights),
+    disp('NB USING SUPPLIED FIXED WEIGHTS');
+    if size(S.fixedweights,1)~=length(maskedgrid_inside_ind),
+        error('Suppied weights are the wrong size for grid');
+        end;
+end;
+
+%% prepare to write images
+talpositions = spm_eeg_inv_transform_points(datareg.toMNI, grid.pos(grid.inside,:)); %% MADE DATAREG STAND ALONE
+sMRI = fullfile(spm('dir'), 'canonical', 'single_subj_T1.nii');    
+ if S.write_epochs,
+     if (S.Nfeatures>1) || (S.Niter>1),
+         error('cannot write epochs for more than one permutation or one feature');
+     end;
+     disp('storing all data in memory');
+     allY=zeros(length(grid.inside),Ntrials,2);
+    epochdirname='Epoch_cvaBf_images';
+    if S.logflag,
+        epochdirname=[epochdirname '_log'];
+    end;
+    res = mkdir(D.path, epochdirname);
+ end;
+
+%% Now have all lead fields and all data
+%% Now do actual beamforming
+%% decide on the covariance matrix we need
+%% construct covariance matrix within frequency range of interest
+
+
+disp('now running through freq bands and constructing t stat images');
+
+origfftnewdata=fftnewdata;
+for boot=1:Nboot,
+    
+    bttrials=randi(Ntrials,Ntrials,1);
+    if boot==1,
+        bttrials=1:Ntrials;
+    else
+        disp(['Bootstrap run' num2str(boot)]);
+    end; % if boot
+    fftnewdata=origfftnewdata(bttrials,:,:);
+for fband=1:Nbands,
+    freqrange=freqbands(fband,:);
+    freq_ind=intersect(find(fHz>=freqrange(1)),find(fHz<freqrange(2)));
+    freqrangetest=cell2mat(S.testbands(fband));
+    Ntestbands=length(freqrangetest)/2;
+    if floor(Ntestbands)~=Ntestbands,
+        error('Need pairs of values in testband');
+        end;
+    freq_indtest=[];
+    freq_teststr='';
+    for k=1:Ntestbands
+        newtestind=intersect(find(fHz>=freqrangetest(k*2-1)),find(fHz<freqrangetest(k*2)));
+        freq_indtest=[freq_indtest newtestind];    
+        freq_teststr=sprintf('%s %3.2f-%3.2fHz,',freq_teststr,fHz(min(newtestind)),fHz(max(newtestind)));
+        end; % for k
+    if length(setdiff(freq_indtest,freq_ind))>0,
+            error('Bands in test band are not within covariance band')
+     end;
+    covtrial=zeros(Nchans,Nchans);
+   
+    if ~isempty(S.weightspect),
+        [fweighted,weightspectind]=intersect(fHz,S.weightspect(fband).fHz);
+        if abs(max(fHz(freq_ind)-S.weightspect(fband).fHz))>0,
+            error('weight spect vector wrong length');
+            end; % 
+        filtervect=S.weightspect(fband).fAmp; %% arb filter
+    else
+        filtervect=ones(length(freq_ind),1);
+    end;  % weightspect
+    
+    if ~isempty(S.weightttest),
+        [fweightedt,weightspectindt]=intersect(fHz,S.weightttest(fband).fHz);
+        if abs(max(fHz(freq_ind)-S.weightttest(fband).fHz))>0,
+            error('weight ttest vector wrong length');
+            end; % 
+        tfiltervect=S.weightttest(fband).vect; %% weighted by previous mv analysis
+        else
+        tfiltervect=ones(length(freq_indtest),1);
+        end;  % weightspect
+        
+    Allfiltervect=repmat(filtervect,1,Nchans);
+
+for i=1:Ntrials, %% read in all individual trial types 
+     ffttrial=squeeze(fftnewdata(i,freq_ind,:)).*Allfiltervect;
+     covtrial=covtrial+real(cov(ffttrial));
+end; % for i
+  covtrial=covtrial/Ntrials;
+  allsvd = svd(covtrial);
+  cumpower=cumsum(allsvd)./sum(allsvd);
+  nmodes99=min(find(cumpower>0.99));
+  disp(sprintf('99 percent of the power in this data is in the first %d principal components of the cov matrix',nmodes99));
+  disp(sprintf('largest/smallest eigenvalue=%3.2f',allsvd(1)/allsvd(end)));
+  disp(sprintf('\nFrequency resolution %3.2fHz',mean(diff(fHz))));
+  noise = allsvd(end); %% use smallest eigenvalue
+  redNfeatures=Nfeatures;
+  
+  if S.dimauto, %% only do this flag set
+    if nmodes99<redNfeatures,
+        redNfeatures=nmodes99+1;
+        disp(sprintf('reducing number of features to those describing 99 percent of data (from %d to %d)',Nfeatures,redNfeatures));
+    end; % 
+   end; % if S.dimauto
+    
+    if length(freq_indtest)<redNfeatures,    
+        disp(sprintf('reducing number of features to match bandwidth (from %d to %d) !!',redNfeatures,length(freq_indtest)));
+        redNfeatures=length(freq_indtest);
+    end;
+  
+    
+  disp(sprintf('covariance band from %3.2f to %3.2fHz (%d bins), test band %s (%d bins)',fHz(freq_ind(1)),fHz(freq_ind(end)),length(freq_ind),freq_teststr,length(freq_indtest)))
+  disp(sprintf('Using %d features',redNfeatures));
+    
+  
+  
+  lambda = (S.regpc/100) * sum(allsvd)/size(covtrial,1); %% scale lambda relative mean eigenvalue
+  disp(sprintf('regularisation =%3.2f percent',S.regpc));
+  cinv=pinv(covtrial+eye(size(covtrial,1))*lambda); %% get inverse - if N features has been reduced should maybe think of sorting this out too.
+  
+      
+
+CVA_maxstat=zeros(length(grid.inside),2,S.Niter);
+roymax=zeros(size(CVA_maxstat));
+
+ tstat=zeros(length(grid.inside),2,S.Niter);
+maxcva=zeros(2,S.Niter);
+power_trial=zeros(Ntrials,length(freq_indtest));
+evoked_trial=zeros(Ntrials,length(freq_indtest));
+
+TrueIter=1; %% no permutation for this iteration
+for j=1:S.Niter, %% set up permutations in advance- so perms across grid points are identical
+    randind(j,:)=randperm(Ntrials);
+    if j==TrueIter,
+        randind(j,:)=1:Ntrials; % don't permute first run
+        end;
+    end;
+    
+  
+  
+  
+  
+    for i=1:length(maskedgrid_inside_ind), %% 81
+        %disp('gridpoint')
+        %i
+        %length(maskedgrid_inside_ind)
+        lf=cell2mat(grid.leadfield(grid.inside(maskedgrid_inside_ind(i))));
+        
+        %% get optimal orientation- direct copy from Robert's beamformer_lcmv.m
+        projpower_vect=pinv(lf'*cinv*lf);
+        if isempty(S.gridori),
+            [u, s, v] = svd(real(projpower_vect)); %% MAY NOT BE POWER WE NEED TO OPTIMISE
+            eta = u(:,1);
+        else
+            eta=S.gridori(grid.inside(maskedgrid_inside_ind(i)),:)';
+        end;
+%         if i==80,
+%             disp('FIXING ORIENTATION AT SOURCE LOCATION !!')
+%             eta=[0.9513    0.1621         0]';
+%             end; % if i
+        lf  = lf * eta; %% now have got the lead field at this voxel, compute some contrast
+       
+        weights=(lf'*cinv*lf)\lf'*cinv; %% no regularisation for now
+        normweights=sqrt(weights*weights');
+        if S.fixedweights,
+            weights=S.fixedweights(i,:);
+            end;
+        if S.randomweight,
+            disp('RANDOM WEIGHT !');
+            weights=randn(size(weights));
+        end;
+        
+        stats(fband).ctf_weights(maskedgrid_inside_ind(i),:)=weights;
+        alllf(maskedgrid_inside_ind(i),:)=lf;
+        
+        
+        for j=1:Ntrials, %% this non-linear step (power estimation) has to be done at each location
+            fdata=squeeze(fftnewdata(j,freq_indtest,:));
+              if length(freq_indtest)==1,
+                  fdata=fdata';
+                  end; % if length
+             fdatatrial=fdata*weights';
+            evoked_trial(j,:)=fdatatrial;
+            if S.logflag,
+                power_trial(j,:)=log(fdatatrial.*conj(fdatatrial));
+            else
+                power_trial(j,:)=fdatatrial.*conj(fdatatrial); %%
+            end; % i 
+
+        end; % for j
+        
+        
+
+        for power_flag=0:1,
+            if power_flag,
+                Yfull=power_trial;
+            else
+                Yfull=[real(evoked_trial), imag(evoked_trial)]; %% split into sin and cos parts 
+            end; % if power_fla
+        Y     = Yfull - X0*(X0'*Yfull); %% eg remove DC level or drift terms from all of Y
+        [u s v] = spm_svd(Y,-1,-1); % ,1/4);              %% get largest variance in Y   -1s make sure nothing is removed
+            %% reduce dimensions to Nfeatures
+        thresh=s(redNfeatures,redNfeatures);
+        if redNfeatures==length(freq_indtest),
+            thresh=0;
+            end;
+      
+          [u,s,v]=svd(Y'*Y); %% reduce dimensions of useful signal
+       %   U=u(1:redNfeatur es,:);
+          U=u(:,1:redNfeatures); %% IMPORTANT ?
+          Y   = Y*U;
+          if S.write_epochs,
+            nfact=normweights.^(power_flag+1); %% normalise for amplitude or power
+            allY(i,:,power_flag+1)=Y./nfact; %% normalise the Ys by the depth
+            
+            end;
+        
+  %% NOTE TSTAT COMPUTED Y BEFORE DIMENSION REDUCTION- COMPARISON ONLY VALID WHEN Y
+  %% and YFULL have same dimensionality
+  %% compare with hotellings
+        
+       %% Now permute the rows of X if necessary
+        for iter=1:S.Niter,
+        
+            
+        X=Xdesign(randind(iter,:),:); %% randind(1,:)=1, i.e. unpermuted
+        
+        if boot>1, %% have to also shuffle design matrix with data in bootstrap
+            tmp=X;
+            X=tmp(bttrials,:);
+            end;
+            
+  
+              %-Canonical Variates Analysis
+    %   ==========================================================================
+    % remove null space of contrast
+     %--------------------------------------------------------------------------
+            Y     = Y - X0*(X0'*Y); %% eg remove DC level or drift terms from all of Y
+            X     = X - X0*(X0'*X);
+            P     = pinv(X);
+  
+        if TWOSAMPLETEST,
+            Ym=mean(Y,2);
+            B  = pinv(X)*Ym;
+            RSS   = sum((Ym - X*B).^2);
+            MRSS  = RSS / dfe;
+            SE    = sqrt(MRSS*(pinv(X'*X)));
+            tstat(maskedgrid_inside_ind(i),power_flag+1,iter)=B./SE;
+            Ftstat=(B./SE).^2;
+        end; % if TWOSAMPLETEST
+    
+        
+        [n,b] = size(X);                       
+        [n,m] = size(Y); %% n is number of epochs, m is number of features
+        b     = rank(X);
+        h     = min(b,m); %% either number of features or rank of X
+        f     = n - b - size(X0,2); %% number of epochs- rank(X) - number of confounds
+ 
+ 
+        % generalised eigensolution for treatment and residual sum of squares
+        %--------------------------------------------------------------------------
+        T     = X*(P*Y); %% projection of X onto Y ?
+        SST   = T'*T;
+        SSR   = Y - T; %% residuals in Y (unexplained by X)
+        SSR   = SSR'*SSR;
+        [v,d] = eig(SSR\SST); %% explained/unexplained variance
+        [q,r] = sort(-real(diag(d)));
+        r     = r(1:h);
+        d     = real(d(r,r));
+        v     = real(v(:,r));
+        V     = U*v;                          % canonical vectors  (data)
+        v     = Y*v;                          % canonical variates (data)
+        W     = P*v;                          % canonical vectors  (design)
+        w     = X*W;                          % canonical variates (design)
+        C     = c*W;                          % canonical contrast (design)
+        
+        
+        
+ 
+% inference on dimensionality - p(i) test of D >= i; Wilk's Lambda := p(1)
+%--------------------------------------------------------------------------
+        roymax(maskedgrid_inside_ind(i),power_flag+1,iter)=d(1,1); %% maximum root is the first one
+        Roy2F=(f+m-max(b,m))/max(b,m); %% conversion from Roy's max root to F
+        
+
+        cval  = log(diag(d) + 1);
+        for i1 = 1:h
+            chi(i1) = (f - (m - b + 1)/2)*sum(cval(i1:h));
+            df(i1)  = (m - i1 + 1)*(b - i1 + 1); % m is the number of features, b is the rank of the design matrix
+            p(i1)   = 1 - spm_Xcdf(chi(i1),df(i1));
+            p05thresh(i1) = spm_invXcdf(1-0.05,df(i1));
+        end
+        
+       
+        CVA_maxstat(maskedgrid_inside_ind(i),power_flag+1,iter)=chi(1); %% get wilk's lambda stat
+        CVA_otherdim(maskedgrid_inside_ind(i),power_flag+1,iter,1:h-1)=chi(2:h); %% tests on other dimensions
+        
+        allw(maskedgrid_inside_ind(i),power_flag+1,iter,:,:)=W;
+        
+        
+        if CVA_maxstat(maskedgrid_inside_ind(i),power_flag+1,iter)>maxcva(power_flag+1,iter),
+            stats(fband).bestchi(power_flag+1,1:h,iter)=chi;
+            %if length(V)>1,
+                if power_flag,
+                 stats(fband).bestVpw(1:h,iter,:)=V';
+                 stats(fband).bestvpw(1:h,iter,:)=v';
+                 stats(fband).bestUpw=U;
+                else
+                 stats(fband).bestVev(1:h,iter,:)=complex(V(1:size(evoked_trial,2),:),V(size(evoked_trial,2)+1:end,:))';
+                 stats(fband).bestvev(1:h,iter,:)=v';
+                 stats(fband).bestUev=U;
+                end;
+            %end;% 
+           
+            stats(fband).pval(power_flag+1,1:h,iter)=p;
+            stats(fband).df(power_flag+1,1:h,iter)=df;
+            stats(fband).p05alyt(power_flag+1,1:h)=p05thresh;
+            stats(fband).maxind(power_flag+1,iter)=maskedgrid_inside_ind(i);
+            stats(fband).freqHz=fHz(freq_indtest);
+            stats(fband).freq_indtest=freq_indtest;
+            stats(fband).freq_ind=freq_ind;
+            stats(fband).maxw=w;
+            stats(fband).maxW=W;
+            maxcva(power_flag+1,iter)=CVA_maxstat(maskedgrid_inside_ind(i),power_flag+1,iter);
+            end; % if CVA_Stat
+        if (iter==TrueIter) && (length(V)>1),
+            if power_flag,
+                stats(fband).allVpw(1:h,i,:)=V';
+                else
+                stats(fband).allVev(1:h,i,:)=complex(V(1:size(evoked_trial,2),:),V(size(evoked_trial,2)+1:end,:))';
+                end;
+            end;% if iter
+        
+        end; % for Niter
+        
+         
+        
+        end; % for power_flag  
+     
+        if i/100==floor(i/100)
+            disp(sprintf('done CVA stats for %3.2f percent of freq band %d of %d, log=%d, rank=%d',100*i/length(maskedgrid_inside_ind),fband,Nbands,S.logflag,S.rankflag));
+        end; % if
+    
+    
+  
+end; % for grid points
+
+
+      
+stats(fband).maskedgrid_inside_ind=maskedgrid_inside_ind;
+stats(fband).CVAmax=CVA_maxstat;
+stats(fband).CVAotherdim=CVA_otherdim;
+stats(fband).roymax=roymax;
+stats(fband).Roy2F=Roy2F;
+stats(fband).allw=allw;
+stats(fband).tstat=tstat.^2;
+stats(fband).fHz=fHz;
+stats(fband).dfmisc=[b h m];
+dispthresh_mv=max(stats(fband).CVAmax)/2; % default display thresholds
+max_mv=max(stats(fband).CVAmax); % default display thresholds
+dispthresh_uv=max(stats(fband).tstat)/2;
+max_uv=max(stats(fband).CVAmax); % default display thresholds
+if S.Niter>1,
+    %% get corrected p values for CVA
+    allglobalmax=squeeze(max(stats(fband).CVAmax(:,:,1:end)));
+    [sortglobalmax,sortglobalmaxind]=sort(allglobalmax','descend');
+    allglobalmax_roy=squeeze(max(stats(fband).roymax(:,:,1:end)));
+    [sortglobalmax_roy,sortglobalmaxind_roy]=sort(allglobalmax_roy','descend');
+    
+    for k=1:2,
+        stats(fband).corrpmax_cva(k)=find(sortglobalmaxind(:,k)==TrueIter)/length(sortglobalmaxind);
+        stats(fband).corrpmax_roy(k)=find(sortglobalmaxind_roy(:,k)==TrueIter)/length(sortglobalmaxind_roy);
+        end;
+     stats(fband).thresh05globalmax_cva=sortglobalmax(round(length(sortglobalmaxind)*5/100),:);
+     stats(fband).thresh05globalmax_roy=sortglobalmax_roy(round(length(sortglobalmaxind)*5/100),:);
+    dispthresh_mv=stats(fband).thresh05globalmax_cva; % display only significant effects
+    if TWOSAMPLETEST
+    %% get corrected p values to t
+        allglobalmax=squeeze(max(abs(stats(fband).tstat(:,:,1:end))));
+        [sortglobalmax,sortglobalmaxind]=sort(allglobalmax','descend');
+        for k=1:2,
+            stats(fband).corrpmax_tstat(k)=find(sortglobalmaxind(:,k)==TrueIter)/length(sortglobalmaxind); 
+            end;
+        stats(fband).thresh05globalmax_tstat=sortglobalmax(round(length(sortglobalmaxind)*5/100),:);
+        dispthresh_uv=stats(fband).thresh05globalmax_tstat; % display only significant effects
+    end; % if TWOSAMPLETEST
+end; % if
+  
+    
+    
+    csource=grid; %% only plot and write out unpermuted iteration
+    gridpositions=csource.pos(csource.inside,:);
+    csource.pow_maxchi(csource.inside) = CVA_maxstat(:,2,TrueIter);  %power
+    csource.evoked_maxchi(csource.inside) = CVA_maxstat(:,1,TrueIter);
+    csource.pow_tstat(csource.inside) = tstat(:,2,TrueIter).^2;
+    
+    csource.pow_maxchi(csource.outside)=0;
+    csource.evoked_maxchi(csource.outside)=0;
+    csource.pow_tstat(csource.outside)=0;
+    
+    csource.pos = spm_eeg_inv_transform_points(datareg.toMNI, csource.pos);
+    
+    
+    
+    if isempty(S.gridpos), %% only write images if they use whole volume
+        
+    
+    cfg1 = [];
+    cfg1.sourceunits   = 'mm';
+    cfg1.parameter = 'pow_maxchi';
+    cfg1.downsample = 1;
+    sourceint_pow_maxchi = ft_sourceinterpolate(cfg1, csource, sMRI);
+    
+    cfg1 = [];
+    cfg1.sourceunits   = 'mm';
+    cfg1.parameter = 'evoked_maxchi';
+    cfg1.downsample = 1;
+    sourceint_evoked_maxchi = ft_sourceinterpolate(cfg1, csource, sMRI);
+    
+    if TWOSAMPLETEST
+    cfg1 = [];
+    cfg1.sourceunits   = 'mm';
+    cfg1.parameter = 'pow_tstat';
+    cfg1.downsample = 1;
+    sourceint_pow_tstat = ft_sourceinterpolate(cfg1, csource, sMRI);
+    end; % IF TWOSAMPLETEST
+    
+    
+    %[maxvals,maxind]=max(stats(fband).ctf_weights');
+    % [minvals,minind]=max(-stats(fband).ctf_weights');
+    [maxvals,maxind]=max(alllf');
+    [minvals,minind]=max(-alllf');
+     
+       weight_extrema=[maxind' minind'];
+       weight_extrema=sort(weight_extrema')'; %% doesn't matter if extrema are reversed
+       u_extrema=unique(weight_extrema,'rows');
+       stats(fband).Nu_extrema=size(u_extrema,1);
+       alpha=0.05;
+       stats(fband).alyt_thresh_Chi_05=spm_invXcdf(1-alpha/(stats(fband).Nu_extrema),Nfeatures*b); %% estimate of volumetric threshold
+    
+    if isfield(S,'S.dispthresh_chi'),
+        dispthresh_mv(2)=S.dispthresh_chi;
+        end;
+    
+        
+   
+    
+    
+    %% else %% write out the data sets
+    disp('writing images');
+    
+    cfg = [];
+    cfg.sourceunits   = 'mm';
+    cfg.parameter = 'pow';
+    cfg.downsample = 1;
+    % write t stat
+    dirname='cvaBf_images';
+    if S.logflag,
+        dirname=[dirname '_log'];
+    end;
+
+    res = mkdir(D.path, dirname);
+    outvol = spm_vol(sMRI);
+    outvol.dt(1) = spm_type('float32');
+    featurestr=[S.filenamestr 'Nf' num2str(redNfeatures)] ;
+    if S.bootstrap,
+        featurestr=sprintf('%s_bt%03d_',featurestr,boot);
+        end; % if
+    outvol.fname= fullfile(D.path, dirname, ['chi_pw_'  spm_str_manip(D.fname, 'r') '_' num2str(freqbands(fband,1)) '-' num2str(freqbands(fband,2)) 'Hz' featurestr '.nii']);
+    stats(fband).outfile_chi_pw=outvol.fname;
+    outvol = spm_create_vol(outvol);
+    spm_write_vol(outvol, sourceint_pow_maxchi.pow_maxchi);
+    cmap=colormap;
+    jetmap=colormap('jet');
+    if (isfield(S, 'preview') && S.preview)
+            spm_check_registration(sMRI)
+            prop=0.4;
+            colourmap=jetmap;
+            spm_orthviews('Addtruecolourimage',1,outvol.fname,colourmap,prop,max_mv(2),dispthresh_mv(2));
+            
+            disp(sprintf('Chi pw image. Est thresh for p<0.05 (corr) is %3.2f. Press any key to continue..',stats(fband).alyt_thresh_Chi_05));
+            pause;
+        end; % if preview
+    colormap(cmap);
+    outvol.fname= fullfile(D.path, dirname, ['chi_ev_' spm_str_manip(D.fname, 'r') '_' num2str(freqbands(fband,1)) '-' num2str(freqbands(fband,2)) 'Hz' featurestr '.nii']);
+    stats(fband).outfile_chi_ev=outvol.fname;
+    outvol = spm_create_vol(outvol);
+    spm_write_vol(outvol, sourceint_evoked_maxchi.evoked_maxchi);
+    
+        if (isfield(S, 'preview') && S.preview)
+            spm_check_registration(sMRI)
+            prop=0.4;
+            colourmap=jetmap;
+            spm_orthviews('Addtruecolourimage',1,outvol.fname,colourmap,prop,max_mv(1),dispthresh_mv(1));
+            disp(sprintf('Chi ev image. Est thresh for p<0.05 (corr) is %3.2f. Press any key to continue..',stats(fband).alyt_thresh_Chi_05));
+            pause;
+        end; % if preview
+        
+        outvals=CVA_maxstat(:,1,TrueIter);
+      %  prefix='testrun';
+     %[outvol]=write_trial_image(0,0,talpositions,datareg,sMRI,D,dirname,csource,outvals,prefix,freqbands(fband,:))
+    colormap(cmap);
+    if TWOSAMPLETEST
+        outvol.fname= fullfile(D.path, dirname, ['tstat_pow_' spm_str_manip(D.fname, 'r') '_' num2str(freqbands(fband,1)) '-' num2str(freqbands(fband,2)) 'Hz' S.filenamestr '.nii']);
+        stats(fband).outfile_pow_tstat=outvol.fname;
+        outvol = spm_create_vol(outvol);
+        spm_write_vol(outvol, sourceint_pow_tstat.pow_tstat);
+    end; % if twosample
+    
+% %% now write all trials of data out
+    if S.write_epochs,
+     permnum=1;
+     outfilenames_pw=[];
+     outfilenames_ev=[];
+     for trialnum=1:Ntrials,
+        prefix='ev';outvals=allY(:,trialnum,1); %% evoked first
+        [outvolev]=write_trial_image(trialnum,permnum,talpositions,datareg,sMRI,D,epochdirname,csource,outvals,prefix,freqbands(fband,:));
+        outfilenames_ev=strvcat(outfilenames_ev,outvolev.fname);
+        prefix='pow';outvals=allY(:,trialnum,2); %% then power
+        [outvolpw]=write_trial_image(trialnum,permnum,talpositions,datareg,sMRI,D,epochdirname,csource,outvals,prefix,freqbands(fband,:));
+        outfilenames_pw=strvcat(outfilenames_pw,outvolpw.fname);
+     end; % for trialnum
+     stats(fband).outfile_pw_epochs=outfilenames_pw;
+     stats(fband).outfile_ev_epochs=outfilenames_ev;
+    end;
+    
+   
+    end; % if ~S.gridpos
+    
+    
+       
+    
+end; % for fband=1:Nbands
+end; %% bootstrap    
+bootlist= fullfile(D.path, dirname, ['bootlist_'  spm_str_manip(D.fname, 'r') '_' num2str(freqbands(fband,1)) '-' num2str(freqbands(fband,2)) 'Hz' featurestr '.mat']);
+save(bootlist,'bttrials');
+
+end % function
+
+
+
+
+function [soutvol]=write_trial_image(trialnum,permnum,talpositions,datareg,sMRI,D,dirname,csource,outvals,prefix,freqband)
+    
+    
+    %outvals=outvals./1e12;
+    %outvals=outvals.*0; 
+    %outvals(715)=trialnum;
+    csource.pow_maxchi(csource.inside) = outvals;
+    
+    csource.pow_maxchi(csource.outside)=0;
+    
+  
+        
+    cfg1 = [];
+    cfg1.sourceunits   = 'mm';
+    cfg1.parameter = 'pow_maxchi';
+    cfg1.downsample = 1;
+    sourceint_pow_maxchi = ft_sourceinterpolate(cfg1, csource, sMRI);
+    
+    
+    cfg = [];
+    cfg.sourceunits   = 'mm';
+    cfg.parameter = 'pow';
+    cfg.downsample = 1;
+    
+    
+    outvol = spm_vol(sMRI);
+    outvol.dt(1) = spm_type('float32');
+    featurestr=['Nf1'] ; %% only for 1 feature
+    outvol.fname= fullfile(D.path, dirname, [sprintf('Trial%04d_Perm%04d_%s',trialnum,permnum,prefix)  spm_str_manip(D.fname, 'r') '_' num2str(freqband(1)) '-' num2str(freqband(2)) 'Hz' featurestr '.nii']);
+    
+    outvol = spm_create_vol(outvol);
+    
+    spm_write_vol(outvol, sourceint_pow_maxchi.pow_maxchi);
+    soutvol=outvol;
+    mmsampling=min(abs(diag(soutvol.mat(1:3,1:3)))); %% get sampling of the output image
+    Sfwhm=mmsampling*3; %% smooth by 3* this to get sufficiently sampled output image
+    soutvol.fname= fullfile(D.path, dirname, [sprintf('S%dmmTrial%04d_Perm%04d_%s',Sfwhm,trialnum,permnum,prefix)  spm_str_manip(D.fname, 'r') '_' num2str(freqband(1)) '-' num2str(freqband(2)) 'Hz' featurestr '.nii']);
+    
+    spm_smooth(outvol,soutvol,Sfwhm);
+    
+      
+end
diff --git a/toolbox/Beamforming/spm_eeg_ft_beamformer_lcmv.m b/toolbox/Beamforming/spm_eeg_ft_beamformer_lcmv.m
index b86d599..2cc64fa 100644
--- a/toolbox/Beamforming/spm_eeg_ft_beamformer_lcmv.m
+++ b/toolbox/Beamforming/spm_eeg_ft_beamformer_lcmv.m
@@ -9,7 +9,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Gareth Barnes
-% $Id: spm_eeg_ft_beamformer_lcmv.m 3969 2010-07-05 09:26:46Z gareth $
+% $Id: spm_eeg_ft_beamformer_lcmv.m 4218 2011-03-01 11:58:01Z gareth $
 
 [Finter,Fgraph] = spm('FnUIsetup','univariate LCMV beamformer for power', 0);
 %%
@@ -97,6 +97,33 @@
     S.regpc=0;
 end; % if
 
+if ~isfield(S,'suffix_str'),
+    S.suffix_str=[];
+end; % if
+
+
+    
+if ~isfield(S,'bootstrap'),
+    S.bootstrap=[];
+else
+    Nboot=S.bootstrap;
+    end;
+    
+if isempty(S.bootstrap),
+    S.bootstrap=0;
+    Nboot=1;
+    end;
+    
+if ~isfield(S,'components'),
+    S.components=[];
+end; % if
+
+if isempty(S.components),
+    compind=1;
+else
+    compind=S.components;
+end; % if
+
 
 try
     D = S.D;
@@ -302,6 +329,10 @@
     cfg.trials=S.design.Xtrials(useind); %% trials starting at these times
     cfg.latency=[winstart winstart+S.design.Xwindowduration];
     subdata=ft_timelockanalysis(cfg,data); % subset of original data
+    if length(subdata.time)~=Nsamples,
+        error('Check the window specified is within epoch');
+        
+        end;
     allepochdata(useind,:,:)=squeeze(subdata.trial); %% get an epoch of data with channels in columns
 end; % for i
 
@@ -322,6 +353,10 @@
     error('Data dimension mismatch');
 end;
 
+if isfield(S,'return_data')
+    stats.allepochdata=allepochdata;
+    end;
+    
 clear allepochdata; %% no longer needed
 
 
@@ -372,7 +407,17 @@
     maskedgrid_inside_ind=find(S.maskgrid==1); %% indices into grid.inside 
     end; 
 
-
+ if cfg.reducerank, %% follow up rank reduction and remove redundant dimension from lead fields
+    for i=1:length(maskedgrid_inside_ind), %% 81
+        lf1=cell2mat(grid.leadfield(grid.inside(maskedgrid_inside_ind(i))));
+        [u1,s1,v1]=svd(lf1'*lf1);
+        grid.leadfield(grid.inside(maskedgrid_inside_ind(i)))={lf1*u1(:,1:cfg.reducerank)};
+        %normlf(i)=std(dot(lfnew',lfnew'));
+    end;
+ end; % if reduce rank
+ 
+ %[a,b]=min(normlf) 
+ %origin=grid.pos(grid.inside(maskedgrid_inside_ind(b)),:)
 
     
 %% Now have all lead fields and all data
@@ -381,6 +426,17 @@
 %% construct covariance matrix within frequency range of interest
 
 disp('now running through freq bands and constructing t stat images');
+origfftnewdata=fftnewdata;
+for boot=1:Nboot,
+    
+    bttrials=randi(Ntrials,Ntrials,1);
+    if boot==1,
+        bttrials=1:Ntrials;
+    else
+        disp(['Bootstrap run' num2str(boot)]);
+    end; % if boot
+    fftnewdata=origfftnewdata(bttrials,:,:);
+    
 for fband=1:Nbands,
     freqrange=S.freqbands{fband};
     freq_ind=intersect(find(fHz>=freqrange(1)),find(fHz<freqrange(2)));
@@ -463,13 +519,19 @@
       
  dfe=Ntrials-rank(Xdesign);  % df test
  
+
+    
     for i=1:length(maskedgrid_inside_ind), %% 81
         lf=cell2mat(grid.leadfield(grid.inside(maskedgrid_inside_ind(i))));
         
         %% get optimal orientation- direct copy from Robert's beamformer_lcmv.m
         projpower_vect=pinv(lf'*cinv*lf);
+        [u1,s1,v1]=svd(lf'*cinv*lf);
+        usecomp=length(s1);
+        
         [u, s, v] = svd(real(projpower_vect));
-        eta = u(:,1);
+        eta = u(:,compind);
+        
         lf  = lf * eta; %% now have got the lead field at this voxel, compute some contrast
         weights=lf'*cinv/(lf'*cinv*lf); %% CORRECT WEIGHTS CALC
         
@@ -502,23 +564,13 @@
        %% Now permute the rows of X if necessary
         for iter=1:S.Niter,
         
-%             
+          
+            
              X=Xdesign(randind(iter,:),:); %% randind(1,:)=1, i.e. unpermuted
-%             cond1_ind=find(X(:,1)>0);
-%             cond2_ind=find(X(:,1)<=0);
-%             nx=length(cond1_ind);
-%             ny=length(cond2_ind);
-%             dfe = nx + ny - 2;
-%             xba_epochs=Yfull(cond1_ind,:)*tfiltervect; %% for univariate tfiltervect is all ones (i.e. sum)
-%             yba_epochs=Yfull(cond2_ind,:)*tfiltervect; %%
-%             pdiff=mean(xba_epochs)-mean(yba_epochs);
-%             s2x=var(xba_epochs);
-%             s2y=var(yba_epochs);
-%             sPooled = sqrt(((nx-1) .* s2x + (ny-1) .* s2y) ./ dfe);
-%             se = sPooled .* sqrt(1./nx + 1./ny);
-%             tstat(maskedgrid_inside_ind(i),iter) = pdiff ./ se; %
-%             normdiff(maskedgrid_inside_ind(i),iter)=pdiff/(weights*noise_id*weights');
-%             
+             if boot>1, %% have to also shuffle design matrix with data in bootstrap
+                tmp=X;
+                X=tmp(bttrials,:);
+                end;
             
             % Contrast
             
@@ -534,7 +586,7 @@
             
             tstat(maskedgrid_inside_ind(i),iter)=c*B./SE;
             normdiff(maskedgrid_inside_ind(i),iter)=c*B/(weights*noise_id*weights'); %% maybe a factor missing here
-
+            %allB(maskedgrid_inside_ind(i))=B;
     
         end; % for Niter
         
@@ -624,6 +676,11 @@
     cfg.sourceunits   = 'mm';
     cfg.parameter = 'pow';
     cfg.downsample = 1;
+    
+     featurestr='';
+    if S.bootstrap,
+        featurestr=sprintf('%s_bt%03d_',featurestr,boot);
+        end; % if
     % write t stat
     dirname='tstatBf_images';
     if S.logflag,
@@ -633,7 +690,8 @@
     res = mkdir(D.path, dirname);
     outvol = spm_vol(sMRI);
     outvol.dt(1) = spm_type('float32');
-        outvol.fname= fullfile(D.path, dirname, ['spmT_' spm_str_manip(D.fname, 'r') '_' num2str(S.freqbands{fband}(1)) '-' num2str(S.freqbands{fband}(2)) 'Hz' S.filenamestr '.nii']);
+    
+        outvol.fname= fullfile(D.path, dirname, ['spmT_' spm_str_manip(D.fname, 'r') '_' num2str(S.freqbands{fband}(1)) '-' num2str(S.freqbands{fband}(2)) 'Hz' S.filenamestr S.suffix_str featurestr '.nii']);
         
         stats(fband).outfile_pow_tstat=outvol.fname;
         outvol = spm_create_vol(outvol);
@@ -648,7 +706,7 @@
             disp('Press any key to continue');
             pause;
         end; % if preview
-        outvol.fname= fullfile(D.path, dirname, ['spmNdiff_' spm_str_manip(D.fname, 'r') '_' num2str(S.freqbands{fband}(1)) '-' num2str(S.freqbands{fband}(2)) 'Hz' S.filenamestr '.nii']);
+        outvol.fname= fullfile(D.path, dirname, ['spmNdiff_' spm_str_manip(D.fname, 'r') '_' num2str(S.freqbands{fband}(1)) '-' num2str(S.freqbands{fband}(2)) 'Hz' S.filenamestr S.suffix_str featurestr '.nii']);
         
          stats(fband).outfile_normdiff=outvol.fname;
          outvol = spm_create_vol(outvol);
@@ -659,8 +717,11 @@
     
 end; % for fband=1:Nbands
 
-   
-     
+end; % for boot   
+
+bootlist= fullfile(D.path, dirname, ['bootlist_'  spm_str_manip(D.fname, 'r') '_' num2str(S.freqbands{fband}(1)) '-' num2str(S.freqbands{fband}(2)) 'Hz' featurestr '.mat']);
+save(bootlist,'bttrials');
+
      
 end % function
 
diff --git a/toolbox/Beamforming/spm_eeg_ft_beamformer_source.m b/toolbox/Beamforming/spm_eeg_ft_beamformer_source.m
index e834611..d443980 100644
--- a/toolbox/Beamforming/spm_eeg_ft_beamformer_source.m
+++ b/toolbox/Beamforming/spm_eeg_ft_beamformer_source.m
@@ -23,7 +23,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak, Robert Oostenveld
-% $Id: spm_eeg_ft_beamformer_source.m 3833 2010-04-22 14:49:48Z vladimir $
+% $Id: spm_eeg_ft_beamformer_source.m 4238 2011-03-10 19:48:37Z vladimir $
 
 [Finter,Fgraph,CmdLine] = spm('FnUIsetup', 'Beamformer source activity extraction',0);
 
@@ -92,7 +92,14 @@
     S.outfile = spm_input('Output file name', '+1', 's', ['B' D.fname]);
 end
 
-modality = spm_eeg_modality_ui(D, 1, 1);
+modality = D.modality(1, 1);
+if isequal(modality, 'Multimodal')
+    if isfield(S, 'modality') && ~isempty(S.modality)
+        modality = S.modality;
+    else
+        modality = spm_eeg_modality_ui(D, 1, 1);
+    end
+end
 
 %% ============ Select the data and convert to Fieldtrip struct
 if ~isfield(S, 'conditions')
@@ -168,6 +175,8 @@
 vol = ft_transform_vol(M1, vol);
 sens = ft_transform_sens(M1, sens);
 
+pos = spm_eeg_inv_transform_points(M1*datareg.fromMNI, S.sources.pos);
+
 channel = D.chanlabels(setdiff(meegchannels(D, modality), D.badchannels));
 
 [vol, sens] = ft_prepare_vol_sens(vol, sens, 'channel', channel);
@@ -181,9 +190,8 @@
 data.time = data.time(trialind);
 
 cfg = [];
-cfg.channel = modality;
+cfg.channel = D.chanlabels(setdiff(D.meegchannels(modality), D.badchannels))';
 cfg.covariance = 'yes';
-cfg.covariancewindow = 'maxperlength';
 cfg.keeptrials = 'no';
 timelock1 = ft_timelockanalysis(cfg, data);
 cfg.keeptrials = 'yes';
@@ -200,7 +208,7 @@
 
 if ~isfield(S, 'voi') || isequal(S.voi, 'no')
     nvoi = 0;
-    cfg.grid.pos     = S.sources.pos;
+    cfg.grid.pos     = pos;
     if isfield(S.sources, 'ori')
         cfg.grid.mom    = S.sources.ori;
     else
@@ -214,8 +222,8 @@
     sphere(sqrt(X(:).^2 + Y(:).^2 + Z(:).^2)>S.voi.radius, :) = [];
     nvoi = size(sphere, 1);
     cfg.grid.pos = [];
-    for s = 1:size(S.sources.pos, 1)
-        cfg.grid.pos = [cfg.grid.pos; sphere+repmat(S.sources.pos(s, :), nvoi, 1)];
+    for s = 1:size(pos, 1)
+        cfg.grid.pos = [cfg.grid.pos; sphere+repmat(pos(s, :), nvoi, 1)];
     end
 end
 
@@ -229,11 +237,60 @@
 cfg.lambda =  S.lambda;
 source1 = ft_sourceanalysis(cfg, timelock1);
 
+if isfield(S, 'makecorrimage') &&  S.makecorrimage
+    if size(pos, 1) ~= numel(source1.avg.filter)
+        error('Can only make correlation images for point sources');
+    end
+    
+    res = mkdir(pwd, 'corrimages');
+         
+    cfg.grid = [];
+    cfg.grid.xgrid = -90:10:90;
+    cfg.grid.ygrid = -120:10:100;
+    cfg.grid.zgrid = -70:10:110;
+    cfg.inwardshift = -10;
+    
+    fsource = ft_sourceanalysis(cfg, timelock1);
+    
+    
+    sMRI = fullfile(spm('dir'), 'canonical', 'single_subj_T1.nii');
+    
+    pow = nan(size(fsource.avg.pow));
+    for p = 1:numel(source1.avg.filter)
+        for q = 1:length(fsource.inside);
+            cc = corrcoef(source1.avg.filter{p}, fsource.avg.filter{fsource.inside(q)});
+            pow(fsource.inside(q)) = (cc(1,2))^2;
+            fprintf('Correlation image %d/%d\n', q, length(fsource.inside));
+        end
+        
+        fsource.pow = pow;
+        
+        cfg1 = [];
+        cfg1.sourceunits   = 'mm';
+        cfg1.parameter = 'pow';
+        cfg1.downsample = 1;
+        sourceint = ft_sourceinterpolate(cfg1, fsource, sMRI);
+        %%           
+        outvol = spm_vol(sMRI);
+        outvol.dt(1) = spm_type('float32');
+        
+        if length(trialind) == 1
+            suff = num2str(trialind);
+        else
+            suff = '';
+        end
+        
+        outvol.fname= fullfile(pwd, 'corrimages', ['corrimg_' spm_str_manip(D.fname, 'r') '_' S.sources.label{p} '_' suff '.nii']);
+        outvol = spm_create_vol(outvol);
+        spm_write_vol(outvol, sourceint.pow);
+    end
+end
+
 cfg = [];
 cfg.inwardshift = -30;
 cfg.vol = vol;
 cfg.grad = sens;
-cfg.grid = source2grid(source1);
+cfg.grid = ft_source2grid(source1);
 cfg.channel = modality;
 cfg.lambda =  S.lambda;
 cfg.rawtrial = 'yes';
@@ -278,22 +335,25 @@
     disp(['Extracting source data trial ' num2str(i) '/'  num2str(length(source2.trial))]);
     for j = 1:nsources
         if nvoi>0
-            y = cat(1, source2.trial(i).mom{(j-1)*nvoi+[1:nvoi]});
-            
-            % compute regional response in terms of first eigenvariate
-            %-----------------------------------------------------------------------
-            [m n]   = size(y);
-            if m > n
-                [v s v] = svd(y'*y);
-                Y       = v(:,1);
-            else
-                [u s u] = svd(y*y');
-                u       = u(:,1);
-                Y       = y'*u;
-            end
+            y = cat(1, source2.trial(i).mom{(j-1)*nvoi+[1:nvoi]});            
+        else
+            y       = source2.trial(i).mom{j};
+        end
+        
+        % compute regional response in terms of first eigenvariate
+        %-----------------------------------------------------------------------
+        [m n]   = size(y);
+        if m > n && n>1
+            [v s v] = svd(y'*y);
+            Y       = v(:,1);
+        elseif m>1
+            [u s u] = svd(y*y');
+            u       = u(:,1);
+            Y       = y'*u;
         else
-            Y       = source2.trial(i).mom{j};
+            Y = y;
         end
+        
         sourcedata.trial(i, j, :)= Y;
     end
 end
@@ -302,7 +362,6 @@
 sourcedata.time = timelock2.time;
 sourcedata.dimord = 'rpt_chan_time';
 sourcedata.label = S.sources.label;
-sourcedata.fsample = timelock2.fsample;
 sourcedata.avg = [];
 %%
 if ~isempty(S.appendchannels)
diff --git a/toolbox/DARTEL/Makefile b/toolbox/DARTEL/Makefile
index f2f0466..495c445 100644
--- a/toolbox/DARTEL/Makefile
+++ b/toolbox/DARTEL/Makefile
@@ -3,7 +3,7 @@
 #
 # Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 #
-# $Id: Makefile 2457 2008-11-12 09:24:30Z guillaume $
+# $Id: Makefile 4136 2010-12-09 22:22:28Z guillaume $
 
 include ../../src/Makefile.var
 
@@ -36,4 +36,8 @@ optimN.$(SUF): optimN.c optimN_mex.c
 	$(MEX) optimN.c optimN_mex.c $(MEXEND)
 
 optimNn.$(SUF): optimN.c optimN_mex.c
+ifneq (mex,$(SUF))
 	$(MEX) optimN.c optimN_mex.c -DNEUMANN -output optimNn $(MEXEND)
+else
+	$(MEX) optimN.c optimN_mex.c -DNEUMANN --output optimNn.$(SUF) $(MEXEND)
+endif
diff --git a/toolbox/DARTEL/optimizer2d.c b/toolbox/DARTEL/optimizer2d.c
index d195347..4ec0641 100644
--- a/toolbox/DARTEL/optimizer2d.c
+++ b/toolbox/DARTEL/optimizer2d.c
@@ -1,4 +1,4 @@
-/* $Id: optimizer2d.c 2644 2009-01-23 13:01:50Z john $ */
+/* $Id: optimizer2d.c 4016 2010-07-26 13:12:40Z john $ */
 /* (c) John Ashburner (2007) */
 
 #include<mex.h>
@@ -150,17 +150,15 @@ static void relax_le(int dm[], double a[], double b[], double s[], int nit, doub
     printf("%dx%d: ", dm[0],dm[1]);
 #endif
 
-    for(it=0; it<2*nit; it++)
+    for(it=0; it<4*nit; it++)
     {
-        int j, jstart;
-        int i, istart;
+        int i, j;
 
 #ifdef VERBOSE
         printf(" %g", sumsq_le(dm, a, b, s, u));
 #endif
 
-        jstart = it%2;
-        for(j=0; j<dm[1]; j++)
+        for(j=(it>>1)&1; j<dm[1]; j+=2)
         {
             double *pux, *puy, *pbx, *pby, *paxx, *paxy, *payy;
             int jm1,jp1, im1,ip1;
@@ -176,9 +174,7 @@ static void relax_le(int dm[], double a[], double b[], double s[], int nit, doub
             jm1 = (BOUND(j-1,dm[1])-j)*dm[0];
             jp1 = (BOUND(j+1,dm[1])-j)*dm[0];
 
-            istart = (jstart == (j%2));
-
-            for(i=istart; i<dm[0]; i+=2)
+            for(i=(it>>0)&1; i<dm[0]; i+=2)
             {
                 double sux, suy, axx, ayy, axy, idt;
                 double *px = pux+i, *py = puy+i;
diff --git a/toolbox/DARTEL/optimizer3d.c b/toolbox/DARTEL/optimizer3d.c
index 560e45a..adad112 100644
--- a/toolbox/DARTEL/optimizer3d.c
+++ b/toolbox/DARTEL/optimizer3d.c
@@ -1,4 +1,4 @@
-/* $Id: optimizer3d.c 2644 2009-01-23 13:01:50Z john $ */
+/* $Id: optimizer3d.c 4016 2010-07-26 13:12:40Z john $ */
 /* (c) John Ashburner (2007) */
 
 #include<mex.h>
@@ -151,22 +151,20 @@ static void relax_le_noa(int dm[], float b[], double s[], int nit, float u[])
         printf("%dx%dx%d (%g,%g,%g): ", dm[0],dm[1],dm[2], regx,regy,regz);
 #   endif
 
-    for(it=0; it<2*nit; it++)
+    for(it=0; it<8*nit; it++)
     {
-        int k, kstart;
+        int k;
         /* double ss = 0.0; */
-        kstart = it%2;
-        for(k=0; k<dm[2]; k++)
+        for(k=it&1; k<dm[2]; k+=2)
         {
-            int j, jstart, km1, kp1;
+            int j, km1, kp1;
             km1 = (BOUND(k-1,dm[2])-k)*dm[0]*dm[1];
             kp1 = (BOUND(k+1,dm[2])-k)*dm[0]*dm[1];
 
-            jstart = (kstart==(k%2));
-            for(j=0; j<dm[1]; j++)
+            for(j=(it>>1)&1; j<dm[1]; j+=2)
             {
                 float *pux, *puy, *puz, *pbx, *pby, *pbz;
-                int i, istart, jm1,jp1;
+                int i, jm1,jp1;
 
                 pux  = u+dm[0]*(j+dm[1]* k);
                 puy  = u+dm[0]*(j+dm[1]*(k+dm[2]));
@@ -178,8 +176,7 @@ static void relax_le_noa(int dm[], float b[], double s[], int nit, float u[])
                 jm1 = (BOUND(j-1,dm[1])-j)*dm[0];
                 jp1 = (BOUND(j+1,dm[1])-j)*dm[0];
 
-                istart = (jstart==(j%2));
-                for(i=istart; i<dm[0]; i+=2)
+                for(i=(it>>2)&1; i<dm[0]; i+=2)
                 {
                     int im1,ip1;
                     float *px = pux+i, *py = puy+i, *pz = puz+i;
@@ -448,22 +445,20 @@ static void relax_le(int dm[], float a[], float b[], double s[], int nit, float
         printf("%dx%dx%d (%g,%g,%g): ", dm[0],dm[1],dm[2], regx,regy,regz);
 #   endif
 
-    for(it=0; it<2*nit; it++)
+    for(it=0; it<8*nit; it++)
     {
-        int k, kstart;
+        int k;
         /* double ss = 0.0; */
-        kstart = it%2;
-        for(k=0; k<dm[2]; k++)
+        for(k=it&1; k<dm[2]; k+=2)
         {
-            int j, jstart, km1, kp1;
+            int j, km1, kp1;
             km1 = (BOUND(k-1,dm[2])-k)*dm[0]*dm[1];
             kp1 = (BOUND(k+1,dm[2])-k)*dm[0]*dm[1];
 
-            jstart = (kstart==(k%2));
-            for(j=0; j<dm[1]; j++)
+            for(j=(it>>1)&1; j<dm[1]; j+=2)
             {
                 float *pux, *puy, *puz, *pbx, *pby, *pbz, *paxx, *payy, *pazz, *paxy, *paxz, *payz;
-                int i, istart, jm1,jp1;
+                int i, jm1,jp1;
 
                 pux  = u+dm[0]*(j+dm[1]* k);
                 puy  = u+dm[0]*(j+dm[1]*(k+dm[2]));
@@ -481,8 +476,7 @@ static void relax_le(int dm[], float a[], float b[], double s[], int nit, float
                 jm1 = (BOUND(j-1,dm[1])-j)*dm[0];
                 jp1 = (BOUND(j+1,dm[1])-j)*dm[0];
 
-                istart = (jstart==(j%2));
-                for(i=istart; i<dm[0]; i+=2)
+                for(i=(it>>2)&1; i<dm[0]; i+=2)
                 {
                     int im1,ip1;
                     double sux, suy, suz, axx, ayy, azz, axy, axz, ayz, idt;
diff --git a/toolbox/DARTEL/spm_dartel_import.m b/toolbox/DARTEL/spm_dartel_import.m
index 6511d63..39bb981 100644
--- a/toolbox/DARTEL/spm_dartel_import.m
+++ b/toolbox/DARTEL/spm_dartel_import.m
@@ -15,7 +15,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_dartel_import.m 2340 2008-10-15 10:17:27Z john $
+% $Id: spm_dartel_import.m 4194 2011-02-05 18:08:06Z ged $
 
 matnames = job.matnames;
 for i=1:numel(matnames),
@@ -174,7 +174,7 @@ function preproc_apply(p,odir,b0,bb,vx,iopt,opt,matname)
 end;
 
 % Sort out bounding box etc
-[bb1,vx1]       = bbvox_from_V(p.VG(1));
+[bb1,vx1] = spm_get_bbox(p.VG(1), 'old');
 bb(~isfinite(bb)) = bb1(~isfinite(bb));
 if ~isfinite(vx), vx = abs(prod(vx1))^(1/3); end;
 bb(1,:) = vx*ceil(bb(1,:)/vx);
@@ -353,14 +353,3 @@ function preproc_apply(p,odir,b0,bb,vx,iopt,opt,matname)
         return;
     end;
 end;
-%=======================================================================
-
-%=======================================================================
-function [bb,vx] = bbvox_from_V(V)
-vx = sqrt(sum(V(1).mat(1:3,1:3).^2));
-if det(V(1).mat(1:3,1:3))<0, vx(1) = -vx(1); end;
-
-o  = V(1).mat\[0 0 0 1]';
-o  = o(1:3)';
-bb = [-vx.*(o-1) ; vx.*(V(1).dim(1:3)-o)];
-return;
diff --git a/toolbox/DARTEL/spm_dartel_norm_fun.m b/toolbox/DARTEL/spm_dartel_norm_fun.m
index 234b2fa..d019c35 100644
--- a/toolbox/DARTEL/spm_dartel_norm_fun.m
+++ b/toolbox/DARTEL/spm_dartel_norm_fun.m
@@ -35,7 +35,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_dartel_norm_fun.m 3728 2010-02-17 12:18:34Z john $
+% $Id: spm_dartel_norm_fun.m 4194 2011-02-05 18:08:06Z ged $
 
 % Hard coded stuff, that should maybe be customisable
 K    = 6;
@@ -59,8 +59,8 @@
 dimt = size(Nt.dat);
 
 if any(isfinite(bb(:))) || any(isfinite(vox)),
-    [bb0,vox0] = bbvox(Mt,dimt);
-
+    [bb0 vox0] = spm_get_bbox(Nt, 'old');
+    
     msk = ~isfinite(vox); vox(msk) = vox0(msk);
     msk = ~isfinite(bb);   bb(msk) =  bb0(msk);
 
@@ -109,7 +109,7 @@
         % Re-order data
         %--------------------------------------------------------------------------
         subjs = job.data.subjs;
-        subj  = struct('flowfield',cell(numel(subjs.flowfields),1),....
+        subj  = struct('flowfield',cell(numel(subjs.flowfields),1),...
                        'images',   cell(numel(subjs.flowfields),1));
         for i=1:numel(subj)
             subj(i).flowfield = {subjs.flowfields{i}};
@@ -173,19 +173,31 @@
     NI = nifti(fullfile(pth,[nam ext]));
     NO = NI;
     if jactransf,
-        NO.dat.fname=fullfile(pth,['smw' nam ext]);
+        if fwhm==0,
+            NO.dat.fname=fullfile(pth,['mw' nam ext]);
+        else
+            NO.dat.fname=fullfile(pth,['smw' nam ext]);
+        end
         NO.dat.scl_slope = 1.0;
         NO.dat.scl_inter = 0.0;
         NO.dat.dtype     = 'float32-le';
     else
-        NO.dat.fname=fullfile(pth,['sw' nam ext]);
+        if fwhm==0,
+            NO.dat.fname=fullfile(pth,['w' nam ext]);
+        else
+            NO.dat.fname=fullfile(pth,['sw' nam ext]);
+        end
     end
     NO.dat.dim = [dim NI.dat.dim(4:end)];
     NO.mat  = mat;
     NO.mat0 = mat;
     NO.mat_intent  = mat_intent;
     NO.mat0_intent = mat_intent;
-    NO.descrip = sprintf('Smoothed (%gx%gx%g) DARTEL normed',fwhm);
+    if fwhm==0,
+        NO.descrip = 'DARTEL normed';
+    else
+        NO.descrip = sprintf('Smoothed (%gx%gx%g) DARTEL normed',fwhm);
+    end
     out{m} = NO.dat.fname;
     NO.extras = [];
     create(NO);
@@ -245,7 +257,7 @@
                     [f,c] = dartel3('push',f,y,dim);
                     spm_smooth(f,f,krn); % Side effects
                     spm_smooth(c,c,krn); % Side effects
-                    f = f./(c+eps);
+                    f = f./(c+0.001); % I don't like it, but it may stop a few emails.
                 else
                     % Modulated, by pushing
                     scal = abs(det(NI.mat(1:3,1:3))/det(NO.mat(1:3,1:3))); % Account for vox sizes
@@ -260,13 +272,3 @@
     fprintf('\n'); drawnow;
 end
 %__________________________________________________________________________
-
-%__________________________________________________________________________
-function [bb,vx] = bbvox(M,dim)
-vx = sqrt(sum(M(1:3,1:3).^2));
-if det(M(1:3,1:3))<0, vx(1) = -vx(1); end;
-o  = M\[0 0 0 1]';
-o  = o(1:3)';
-bb = [-vx.*(o-1) ; vx.*(dim(1:3)-o)];
-return;
- 
diff --git a/toolbox/DARTEL/spm_dartel_template.m b/toolbox/DARTEL/spm_dartel_template.m
index fac6abf..03bcfdf 100644
--- a/toolbox/DARTEL/spm_dartel_template.m
+++ b/toolbox/DARTEL/spm_dartel_template.m
@@ -7,7 +7,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_dartel_template.m 3172 2009-06-02 11:40:42Z john $
+% $Id: spm_dartel_template.m 4064 2010-09-03 12:57:10Z john $
 
 code = 2;
 st = job.settings;
@@ -60,6 +60,7 @@
    %    end;
    %end;
     if exist(NU(i).dat.fname,'file'),
+        fprintf('Continuing registration from pre-existing parameters (%s)\n', NU(i).dat.fname);
         u = NU(i).dat(:,:,:,1,:);
         u = single(squeeze(u));
         y = dartel3('Exp',u,[K 1 1]);
diff --git a/toolbox/DARTEL/spm_dartel_warp.m b/toolbox/DARTEL/spm_dartel_warp.m
index b72e5f3..5fe5ff1 100644
--- a/toolbox/DARTEL/spm_dartel_warp.m
+++ b/toolbox/DARTEL/spm_dartel_warp.m
@@ -7,7 +7,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_dartel_warp.m 1381 2008-04-11 19:10:56Z john $
+% $Id: spm_dartel_warp.m 4064 2010-09-03 12:57:10Z john $
 
 code = 2;
 st = job.settings;
@@ -58,7 +58,9 @@
     NU.dat.scl_inter = 0;
     NU.descrip = 'Flow Field';
     if exist(NU.dat.fname,'file'),
-        u = NU(i).dat(:,:,:,1,:);
+        fprintf('Continuing registration from pre-existing parameters (%s)\n', NU.dat.fname);
+
+        u = NU.dat(:,:,:,1,:);
         u = single(squeeze(u));
     else
         u = zeros([dm,3],'single');
diff --git a/toolbox/DARTEL/spm_klaff.m b/toolbox/DARTEL/spm_klaff.m
index b3f6728..5b24b1d 100644
--- a/toolbox/DARTEL/spm_klaff.m
+++ b/toolbox/DARTEL/spm_klaff.m
@@ -18,7 +18,7 @@
 % (c) Wellcome Trust Centre for NeuroImaging (2009)
 
 % John Ashburner
-% $Id: spm_klaff.m 3150 2009-05-26 18:11:14Z john $
+% $Id: spm_klaff.m 4148 2011-01-04 16:49:23Z guillaume $
 
 if nargin<2,   Ng = fullfile(spm('Dir'),'toolbox','Seg','TPM.nii'); end
 if ischar(Nf), Nf = nifti(Nf); end
@@ -85,7 +85,7 @@
 
 M = Nf.mat\Ng.mat;
 
-spm_chi2_plot('Init','KL Divergence Affine Registration',...
+spm_plot_convergence('Init','KL Divergence Affine Registration',...
               'RMS Change', 'Iteration');
 
 for it=1:64,
@@ -202,10 +202,10 @@
    %M = M*(eye(4) + dM);  % Another possibility
     M = M*real(expm(dM)); % Forces the transform to have non-neg Jacobian
     fprintf('%3d  %15g  %15g\n', it, sqrt(sum(dM(:).^2)), kl/nv);
-    spm_chi2_plot('Set', sqrt(sum(dM(:).^2)));
+    spm_plot_convergence('Set', sqrt(sum(dM(:).^2)));
     if sum(dM(:).^2) < 1e-7, break; end
 end
-spm_chi2_plot('Clear');
+spm_plot_convergence('Clear');
 %________________________________________________________
 
 %________________________________________________________
diff --git a/toolbox/DARTEL/tbx_cfg_dartel.m b/toolbox/DARTEL/tbx_cfg_dartel.m
index d5ca878..43f7e33 100644
--- a/toolbox/DARTEL/tbx_cfg_dartel.m
+++ b/toolbox/DARTEL/tbx_cfg_dartel.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: tbx_cfg_dartel.m 3804 2010-03-31 16:16:21Z ged $
+% $Id: tbx_cfg_dartel.m 4197 2011-02-08 18:57:11Z ged $
 
 if ~isdeployed, addpath(fullfile(spm('dir'),'toolbox','DARTEL')); end
 
@@ -614,7 +614,7 @@
 many_subj.name = 'Many Subjects';
 many_subj.val  = {flowfields,images};
 many_subj.help = {[...
-'Select this option if you have many subjects to spatially normalise, '...
+'Select this option if you have many subjects to spatially normalise, ',...
 'but there are a small and fixed number of scans for each subject.']};
 % ---------------------------------------------------------------------
 % jactransf Modulation
@@ -658,19 +658,19 @@
 interp.name    = 'Interpolation';
 interp.val     = {1};
 interp.help    = {
-                  ['The method by which the images are sampled when ' ...
-                  'being written in a different space. ' ...
-                  '(Note that Inf or NaN values are treated as zero, ' ...
+                  ['The method by which the images are sampled when ',...
+                  'being written in a different space. ',...
+                  '(Note that Inf or NaN values are treated as zero, ',...
                   'rather than as missing data)']
                   '    Nearest Neighbour:'
                   '      - Fastest, but not normally recommended.'
                   '    Bilinear Interpolation:'
                   '      - OK for PET, realigned fMRI, or segmentations'
                   '    B-spline Interpolation:'
-                  ['      - Better quality (but slower) interpolation' ...
-                  '/* \cite{thevenaz00a}*/, especially with higher ' ...
-                  'degree splines. Can produce values outside the ' ...
-                  'original range (e.g. small negative values from an ' ...
+                  ['      - Better quality (but slower) interpolation',...
+                  '/* \cite{thevenaz00a}*/, especially with higher ',...
+                  'degree splines. Can produce values outside the ',...
+                  'original range (e.g. small negative values from an ',...
                   'originally all positive image).']
 }';
 interp.labels  = {
@@ -822,7 +822,7 @@
 few_subj.name   = 'Few Subjects';
 few_subj.values = {subj};
 few_subj.help   = {[...
-'Select this option if there are only a few subjects, each with many or '...
+'Select this option if there are only a few subjects, each with many or ' ...
 'a variable number of scans each. You will then need to specify a series of subjects, and the flow field and images of each of them.']};
 % ---------------------------------------------------------------------
 %
@@ -868,13 +868,13 @@
 preserve.tag     = 'preserve';
 preserve.name    = 'Preserve';
 preserve.help    = {
-'Preserve Concentrations: Smoothed spatially normalised images (sw*) represent weighted averages of the signal under the smoothing kernel, approximately preserving the intensities of the original images. This option is currently suggested for eg fMRI.'
+'Preserve Concentrations (no "modulation"): Smoothed spatially normalised images (sw*) represent weighted averages of the signal under the smoothing kernel, approximately preserving the intensities of the original images. This option is currently suggested for eg fMRI.'
 ''
-'Preserve Total: Smoothed and spatially normalised images preserve the total amount of signal from each region in the images (smw*). Areas that are expanded during warping are correspondingly reduced in intensity. This option is suggested for VBM.'
+'Preserve Amount ("modulation"): Smoothed and spatially normalised images preserve the total amount of signal from each region in the images (smw*). Areas that are expanded during warping are correspondingly reduced in intensity. This option is suggested for VBM.'
 }';
 preserve.labels = {
-                   'Preserve Concentrations (no "modulation")'
-                   'Preserve Amount ("modulation")'
+                   'Preserve Concentrations'
+                   'Preserve Amount'
 }';
 preserve.values = {0 1};
 preserve.val    = {0};
@@ -915,20 +915,20 @@
 nrm.vout  = @vout_norm_fun;
 nrm.check = @check_norm_fun;
 nrm.help  = {[...
-'Normally, DARTEL generates warped images that align with the average-shaped template. '...
-'This routine includes an initial affine regisration of the template (the final one '...
+'Normally, DARTEL generates warped images that align with the average-shaped template. ',...
+'This routine includes an initial affine regisration of the template (the final one ',...
 'generated by DARTEL), with the TPM data released with SPM.'],[...
-'``Smoothed'''' (blurred) spatially normalised images are generated in such a '...
-'way that the original signal is preserved. Normalised images are '...
-'generated by a ``pushing'''' rather than a ``pulling'''' (the usual) procedure. '...
-'Note that trilinear interpolation is used, and no masking is done.  It '...
-'is therefore essential that the images are realigned and resliced '...
-'before they are spatially normalised.  Alternatively, contrast images '...
-'generated from unsmoothed native-space fMRI/PET data can be spatially '...
+'``Smoothed'''' (blurred) spatially normalised images are generated in such a ',...
+'way that the original signal is preserved. Normalised images are ',...
+'generated by a ``pushing'''' rather than a ``pulling'''' (the usual) procedure. ',...
+'Note that a procedure related to trilinear interpolation is used, and no masking is done.  It ',...
+'is therefore recommended that the images are realigned and resliced ',...
+'before they are spatially normalised, in order to benefit from motion correction using higher order interpolation.  Alternatively, contrast images ',...
+'generated from unsmoothed native-space fMRI/PET data can be spatially ',...
 'normalised for a 2nd level analysis.'],[...
-'Two ``preserve'''' options are provided.  One of them should do the '...
-'equavalent of generating smoothed ``modulated'''' spatially normalised '...
-'images.  The other does the equivalent of smoothing the modulated '...
+'Two ``preserve'''' options are provided.  One of them should do the ',...
+'equavalent of generating smoothed ``modulated'''' spatially normalised ',...
+'images.  The other does the equivalent of smoothing the modulated ',...
 'normalised fMRI/PET, and dividing by the smoothed Jacobian determinants.']};
 % ---------------------------------------------------------------------
 % images Images
diff --git a/toolbox/DEM/ADEM_cued_response.m b/toolbox/DEM/ADEM_cued_response.m
new file mode 100644
index 0000000..67c5995
--- /dev/null
+++ b/toolbox/DEM/ADEM_cued_response.m
@@ -0,0 +1,234 @@
+% Cued responses under active inference: 
+%__________________________________________________________________________
+% This demo illustrates cued sequential movements. It uses active inference
+% under a hierarchal generative model of sequential cues and consequent
+% movements. The agent has a (second level) model of (two) contingencies or
+% contexts; these correspond to the sequential appearance of targets in a
+% clockwise direction. The other context has no sequential aspect. The
+% first level model is contextually modulated to produce the appropriate
+% sequence of (location – specific) affordances, which predict both
+% visual and proprioceptive consequences. This is sufficient to engender
+% cued reaching movements, which are slightly anticipatory if the agent
+% infers the correct probabilistic context. However, if we reverse the
+% order of the stimuli there is an accuracy and reaction time cost, due to
+% the fact that the sequence is unpredictable.  Furthermore, there is a
+% set switching cost as the hidden states at the second (contextual) level
+% are inferred. This provides a simple but very rich model of cued reaching
+% movements and set switching that is consistent with notions of salience
+% and affordance. Furthermore, we can simulate Parkinsonic lesions by
+% reducing the precision of affordance – based cues. These are the visual
+% attributes that confer saliency on the current target. Reducing this
+% precision (for example, dopamine) delays and can even preclude set
+% switching, with associated costs in pointing accuracy. By completely
+% removing the precision of the salience or affordance cues, we obtain
+% autonomous behavior that is prescribed by the itinerant expectations of
+% the agent. This can be regarded as perseveration in a pathological
+% setting or the emission of autonomous behavior in the absence of any
+% precise sensory information
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: ADEM_cued_response.m 4230 2011-03-07 20:58:38Z karl $
+ 
+ 
+% hidden causes and states
+%==========================================================================
+% x    - hidden states:
+%   x.o  - motor angle (intrinsic frame of reference)
+%   x.a  - target salience
+%
+% v    - hidden causes
+%
+% P    - parameters
+%
+% g    - sensations:
+%   g.o  - motor angle (proprioception)
+%   g.p  - finger location (visual)
+%   g.c  - target contrast (visual)
+%--------------------------------------------------------------------------
+ 
+clear
+ 
+% parameters mapping from (unstable) point attractors to visual space
+%--------------------------------------------------------------------------
+P.x    = [-1  1  1 -1;                       % target locations (extrinsic)
+          -1 -1  1  1];
+ 
+P.s    = 1;                                  % softmax parameter
+n      = size(P.x,2);                        % number of attractors
+N      = 128;                                % length of data sequence
+ 
+% hidden states (M)
+%--------------------------------------------------------------------------
+x.o    = sparse(2,1);                        % finger location
+x.a    = sparse(n,1) - 4;                    % attractor (SHC) states
+x.a(4) = 0;
+ 
+ 
+% Recognition model
+%==========================================================================
+M(1).E.s = 1;                                % smoothness
+M(1).E.n = 3;                                % order of 
+M(1).E.d = 2;                                % generalised motion
+ 
+ 
+% precisions: sensory input
+%--------------------------------------------------------------------------
+V.o = exp(4) + sparse(2,1);                  % motor (proprioceptive)
+V.p = exp(4) + sparse(2,1);                  % target locations (visual)
+V.c = exp(4) + sparse(n,1);                  % target contrast (visual)
+ 
+ 
+% level 1: Displacement dynamics and mapping to sensory/proprioception
+%--------------------------------------------------------------------------
+M(1).f  = 'spm_fx_dem_cue';                  % plant dynamics
+M(1).g  = 'spm_gx_dem_cue';                  % prediction
+M(1).pE = P;                                 % target locations
+M(1).x  = x;                                 % hidden states
+M(1).V  = diag(spm_vec(V));                  % error precision
+M(1).W  = exp(4);                            % error precision
+ 
+% level 2
+%--------------------------------------------------------------------------
+M(2).f  = inline('spm_lotka_volterra(x,1/16)', 'x', 'v', 'P');
+M(2).g  = inline('spm_softmax(x)',             'x', 'v', 'P');
+ 
+M(2).x  = [2; -2];                           % hidden states
+M(2).v  = [1;  0];                           % hidden causes
+M(2).V  = exp(4);                            % error precision
+M(2).W  = exp(4);                            % error precision
+ 
+% generative model
+%==========================================================================
+ 
+% hidden states (G)
+%--------------------------------------------------------------------------
+x.o = [0;0];                                 % finger location
+x.a = sparse(n,1);                           % attractor (SHC) states
+ 
+% first level
+%--------------------------------------------------------------------------
+G(1).f  = 'spm_fx_adem_cue';
+G(1).g  = 'spm_gx_adem_cue';
+G(1).pE = P.x;
+G(1).x  = x;                                 % hidden states
+G(1).V  = exp(16);                           % error precision
+G(1).W  = exp(16);                           % error precision
+ 
+ 
+% second level
+%--------------------------------------------------------------------------
+G(2).v  = sparse(n,1);                       % exogenous cues
+G(2).a  = [0; 0];                            % action forces
+G(2).V  = exp(16);
+ 
+% Create cues
+%==========================================================================
+dt    = 64;                                  % bin size (ms)
+isi   = 12;                                  % inter stimulus interval
+dur   = 4;                                   % stimulus duration
+j     = [1 2 3 4 1 4 3 2 1 4 3 2 1];         % stimulus order
+ 
+C     = sparse(n,N);
+for i = 1:length(j)
+    C(j(i),:) = C(j(i),:) + exp(-([1:N] - i*isi).^2/(2*dur^2));
+end
+ 
+% generate and invert of over different DA levels (D = V.c)
+%==========================================================================
+DEM.G = G;
+DEM.M = M;
+DEM.C = C;
+ 
+D     = 2 + (1:6)/2;            
+for i = 1:length(D)
+ 
+ 
+    % set precision of salience or affordance - V.c
+    %----------------------------------------------------------------------
+    ADEM{i}        = DEM;
+    V.c            = exp(D(i)) + sparse(n,1);
+    ADEM{i}.M(1).V = diag(spm_vec(V));
+    
+    % solve
+    %----------------------------------------------------------------------
+    ADEM{i}        = spm_ADEM(ADEM{i});
+    
+end
+ 
+ 
+ 
+% dynamics and movies
+%==========================================================================
+ 
+% normal
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 1');
+spm_DEM_qU(ADEM{6}.qU)
+ 
+subplot(3,2,5)
+spm_dem_cue_movie(ADEM{6})
+ 
+% delayed set switching
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 2');
+spm_DEM_qU(ADEM{3}.qU)
+ 
+subplot(3,2,5)
+spm_dem_cue_movie(ADEM{3})
+ 
+% failure to switch set
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 3');
+spm_DEM_qU(ADEM{2}.qU)
+ 
+subplot(3,2,5)
+spm_dem_cue_movie(ADEM{2})
+ 
+ 
+% reaction times and accuracy
+%==========================================================================
+spm_figure('GetWin','Figure 4'); clf
+ 
+C     = {'r','g','b','c','m'};
+for i = 1:length(C)
+ 
+    % get behavioral performance
+    %----------------------------------------------------------------------
+    [on,rt,ac] = spm_ADEM_cue_rt(ADEM{i});
+ 
+    % graphics
+    %----------------------------------------------------------------------
+    subplot(2,1,1), hold on
+    plot(on,rt,[C{i} ':'],on,rt,[C{i} '.'],'MarkerSize',24)
+    xlabel('cue onset (sec)','FontSize',12)
+    ylabel('milliseconds','FontSize',12)
+    title('reaction times','FontSize',16)
+    axis square, box off, hold off
+ 
+    subplot(2,1,2), hold on
+    plot(on,ac,[C{i} ':'],on,ac,[C{i} '.'],'MarkerSize',24)
+    xlabel('cue onset (sec)','FontSize',12)
+    ylabel('inverse distance','FontSize',12)
+    title('spatial error','FontSize',16)
+    axis square, box off, hold off
+    drawnow
+ 
+end
+ 
+ 
+% demonstrate autonomous movement (by removing salient information)
+%==========================================================================
+DEM        = DEM;
+V.c        = exp(-2) + sparse(n,1);
+DEM.M(1).V = diag(spm_vec(V));
+DEM        = spm_ADEM(DEM);
+ 
+% perseverative dynamics and trajectories
+%--------------------------------------------------------------------------
+spm_figure('GetWin','DEM');
+spm_DEM_qU(DEM.qU)
+ 
+subplot(3,2,5)
+spm_dem_cue_movie(DEM)
diff --git a/toolbox/DEM/ADEM_lorenz.m b/toolbox/DEM/ADEM_lorenz.m
index ea334ca..5253492 100644
--- a/toolbox/DEM/ADEM_lorenz.m
+++ b/toolbox/DEM/ADEM_lorenz.m
@@ -9,7 +9,7 @@
 % generative model
 %==========================================================================
 clear
-DEMO     = 0;                          % switch for demo
+DEMO     = 1;                          % switch for demo
 G(1).E.s = 1/4;                        % smoothness
 G(1).E.n = 6;                          % smoothness
 G(1).E.d = 2;                          % smoothness
@@ -143,7 +143,7 @@
     save DEM_lorenz G DEM
 else
     try
-    load DEM_lorenz
+        load DEM_lorenz
     catch
         disp({'sorry; the previously prepared .mat file is not compatible';
               'with your version of matlab - run this routine with DEMO = 1'})
@@ -159,7 +159,7 @@
 % replace priors with learned conditional expectation and plot
 %--------------------------------------------------------------------------
 M             = DEM.M;
-M(1).E.linear = 3;                     % nonlinear
+M(1).E.linear = 3;
 
 spm_figure('GetWin','Graphics');
 subplot(3,2,5)
diff --git a/toolbox/DEM/ADEM_lorenz_entropy.m b/toolbox/DEM/ADEM_lorenz_entropy.m
index a814a5d..e59cf96 100644
--- a/toolbox/DEM/ADEM_lorenz_entropy.m
+++ b/toolbox/DEM/ADEM_lorenz_entropy.m
@@ -12,7 +12,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: ADEM_lorenz_entropy.m 3977 2010-07-08 14:14:35Z karl $
+% $Id: ADEM_lorenz_entropy.m 4146 2010-12-23 21:01:39Z karl $
  
 % generative process (environment)
 %==========================================================================
@@ -99,7 +99,6 @@
  
 M(1).E.nE = 1;
 M(1).E.nM = 8;
-M(1).E.nN = 1;
  
 DEM{1}.M = M;
 DEM{1}.G = G;
@@ -128,7 +127,7 @@
         % display
         %------------------------------------------------------------------
         spm_DEM_qU(DEM{i + 1}.qU,DEM{i + 1}.pU)
-        DEM{i + 1}.F
+        
     end
     
     save DEM_lorenz_entropy DEM
diff --git a/toolbox/DEM/ADEM_observe.m b/toolbox/DEM/ADEM_observe.m
index ea6f0d9..febab13 100644
--- a/toolbox/DEM/ADEM_observe.m
+++ b/toolbox/DEM/ADEM_observe.m
@@ -1,4 +1,4 @@
-% This demo illustrates action-observation using synthetics writing under 
+% This demo illustrates action-observation using synthetic writing under 
 % active inference. It shows how expectations about hidden states can be 
 % both cause and consequence of observed action (of self and others 
 % respectively). We first illustrate the generation of behaviour using a 
@@ -12,7 +12,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: ADEM_observe.m 3977 2010-07-08 14:14:35Z karl $
+% $Id: ADEM_observe.m 4170 2011-01-24 18:37:42Z karl $
  
  
 % hidden causes and states
@@ -174,7 +174,7 @@
 
 % coherence analysis
 %==========================================================================
-spm_figure('GetWin','Figure 4')
+spm_figure('GetWin','Figure 4');
 
 T  = 16;
 xe = DEM.qU.w{1}(5:end,T:end);      % prediction error on attractor states
diff --git a/toolbox/DEM/ADEM_pursuit.m b/toolbox/DEM/ADEM_pursuit.m
new file mode 100644
index 0000000..de9cc63
--- /dev/null
+++ b/toolbox/DEM/ADEM_pursuit.m
@@ -0,0 +1,116 @@
+% Slow pursuit under active inference: 
+%__________________________________________________________________________
+% This demo illustrates slow pursuit eye movements under active inference. 
+% Its focus is on frames of references and the entrainment of gaze-
+% direction by the motion of a visual target. The generative process (and 
+% model) is based upon the itinerant trajectory of a target (in Cartesian 
+% coordinates) produced by Lotka-Volterra dynamics. The agent expects its 
+% sampling (in polar coordinates) to be centred on the target. Here, the 
+% agent is equipped with a model of the trajectory and the oculomotor 
+% plant. This means it represents both the location of the target and the 
+% mapping from target location (in relation to a fixation point) to 
+% egocentric polar coordinates.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: ADEM_pursuit.m 4170 2011-01-24 18:37:42Z karl $
+ 
+ 
+% hidden causes and states
+%==========================================================================
+% x    - hidden states:
+%   x.o(1) - oculomotor angle
+%   x.o(2) - oculomotor angle
+%   x.x(1) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.x(2) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.a(:) - attractor (SHC) states
+%
+% v    - causal states
+%   v(1) - not used
+%
+% g    - sensations:
+%   g(1) - oculomotor angle (proprioception)
+%   g(2) - oculomotor angle (proprioception)
+%   g(3) - target location (visual) - intrinsic coordinates (polar)
+%   g(4) - target location (visual) - intrinsic coordinates (polar)
+%--------------------------------------------------------------------------
+
+
+% parameters mapping from (unstable) point attractors to visual space
+%--------------------------------------------------------------------------
+P   = [-1 -1;
+       -1  1;
+        1  1;
+        1 -1]';
+n   = size(P,2);                              % number of attractors
+x.o = [0;0];                                  % oculomotor angle
+x.x = [0;0];                                  % target location
+x.a = sparse(1,1,4,n,1) - 6;                  % attractor (SHC) states
+ 
+
+% Recognition model
+%==========================================================================
+M(1).E.s = 1/2;                               % smoothness
+M(1).E.n = 4;                                 % order of 
+M(1).E.d = 2;                                 % generalised motion
+ 
+% level 1: Displacement dynamics and mapping to sensory/proprioception
+%--------------------------------------------------------------------------
+M(1).f   = 'spm_fx_dem_pursuit';              % plant dynamics
+M(1).g   = 'spm_gx_dem_pursuit';              % prediction
+ 
+M(1).x   = x;                                 % hidden states
+M(1).V   = exp(4);                            % error precision
+M(1).W   = exp(8);                            % error precision
+M(1).pE  = P;
+ 
+ 
+% level 2: not used
+%--------------------------------------------------------------------------
+M(2).v  = 0;                                  % inputs
+M(2).V  = exp(0);
+ 
+
+% generative model
+%==========================================================================
+ 
+% first level
+%--------------------------------------------------------------------------
+G(1).f  = 'spm_fx_adem_pursuit';
+G(1).g  = 'spm_gx_adem_pursuit';
+G(1).x  = x;                                  % hidden states
+G(1).V  = exp(16);                            % error precision
+G(1).W  = exp(16);                            % error precision
+G(1).pE = P;
+ 
+% second level
+%--------------------------------------------------------------------------
+G(2).v  = 0;                                  % exogenous forces
+G(2).a  = [0; 0];                             % action forces
+G(2).V  = exp(16);
+ 
+ 
+% generate and invert
+%==========================================================================
+N       = 128;                                % length of data sequence
+t       = [1:N]*8;
+DEM.G   = G;
+DEM.M   = M;
+DEM.C   = sparse(1,N) + 1/2;
+DEM.U   = sparse(1,N) + 1/2;
+DEM     = spm_ADEM(DEM);
+
+ 
+% overlay true values
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 1');
+spm_DEM_qU(DEM.qU,DEM.pU)
+ 
+spm_figure('GetWin','Figure 2');
+spm_dem_pursuit_movie(DEM)
+
+
+ 
+ 
+
diff --git a/toolbox/DEM/ADEM_response_set.m b/toolbox/DEM/ADEM_response_set.m
new file mode 100644
index 0000000..8e438d6
--- /dev/null
+++ b/toolbox/DEM/ADEM_response_set.m
@@ -0,0 +1,248 @@
+% Cued responses under active inference: 
+%__________________________________________________________________________
+% This demo illustrates cued sequential movements. It uses active inference
+% under a hierarchal generative model of sequential cues and consequent
+% movements. The agent has a (second level) model of (two) contingencies or
+% contexts; these correspond to the sequential appearance of targets in a
+% clockwise direction. The other context has no sequential aspect. The
+% first level model is contextually modulated to produce the appropriate
+% sequence of (location – specific) affordances, which predict both
+% visual and proprioceptive consequences. This is sufficient to engender
+% cued reaching movements, which are slightly anticipatory if the agent
+% infers the correct probabilistic context. However, if we reverse the
+% order of the stimuli there is an accuracy and reaction time cost, due to
+% the fact that the sequence is unpredictable.  Furthermore, there is a
+% set switching cost as the hidden states at the second (contextual) level
+% are inferred. This provides a simple but very rich model of cued reaching
+% movements and set switching that is consistent with notions of salience
+% and affordance. Furthermore, we can simulate Parkinsonic lesions by
+% reducing the precision of affordance – based cues. These are the visual
+% attributes that confer saliency on the current target. Reducing this
+% precision (for example, dopamine) delays and can even preclude set
+% switching, with associated costs in pointing accuracy. By completely
+% removing the precision of the salience or affordance cues, we obtain
+% autonomous behavior that is prescribed by the itinerant expectations of
+% the agent. This can be regarded as perseveration in a pathological
+% setting or the emission of autonomous behavior in the absence of any
+% precise sensory information
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: ADEM_response_set.m 4230 2011-03-07 20:58:38Z karl $
+ 
+ 
+% hidden causes and states
+%==========================================================================
+% x    - hidden states:
+%   x.o  - motor angle (intrinsic frame of reference)
+%   x.a  - target salience
+%
+% v    - hidden causes
+%
+% P    - parameters
+%
+% g    - sensations:
+%   g.o  - motor angle (proprioception)
+%   g.p  - target locations (visual) - intrinsic coordinates (polar)
+%   g.c  - target contrast  (visual)
+%--------------------------------------------------------------------------
+ 
+clear
+ 
+% parameters mapping from (unstable) point attractors to visual space
+%--------------------------------------------------------------------------
+P      = [-1  1  1 -1;                       % target locations (extrinsic)
+          -1 -1  1  1];
+ 
+n      = size(P,2);                          % number of attractors
+N      = 128;                                % length of data sequence
+ 
+% Recognition model
+%==========================================================================
+M(1).E.s = 1;                                % smoothness
+M(1).E.n = 3;                                % order of 
+M(1).E.d = 2;                                % generalised motion
+
+% precisions (sensory)
+%--------------------------------------------------------------------------
+V.o     = exp(8)  + sparse(2,1);
+V.p     = exp(8)  + sparse(2,1);
+V.c     = exp(8)  + sparse(n,1);
+
+
+% level 1: Displacement dynamics and mapping to sensory/proprioception
+%--------------------------------------------------------------------------
+M(1).f  = inline('(P*v - tan(x))/4', 'x', 'v', 'P');
+M(1).g  = inline('[x; tan(x); exp(-sum((P - tan(x)*ones(1,size(P,2))).^2)/2)'']', 'x', 'v', 'P');
+M(1).pE = P;                                 % target locations
+M(1).x  = [0; 0];                            % hidden states (movement)
+M(1).V  = diag(spm_vec(V));                  % error precision
+M(1).W  = exp(8);                            % error precision
+
+
+% level 2: Displacement dynamics and mapping to sensory/proprioception
+%--------------------------------------------------------------------------
+M(2).f  = inline('spm_lotka_volterra(x,v(1)/2)', 'x', 'v', 'P');
+M(2).g  = inline('spm_softmax(x,2)',             'x', 'v', 'P');
+M(2).pE = P;                                 % target locations
+M(2).x  = [0; -1; -6; 2];                    % hidden states (affordance)
+M(2).V  = exp(6);                            % error precision
+M(2).W  = exp(4);                            % error precision
+M(2).xP = exp(0);
+
+ 
+% level 3
+%--------------------------------------------------------------------------
+M(3).f  = inline('spm_lotka_volterra(x,1/64)', 'x', 'v', 'P');
+M(3).g  = inline('spm_softmax(x)',             'x', 'v', 'P');
+ 
+M(3).x  = [2; -2];                           % hidden states (set)
+M(3).V  = exp(2);                            % error precision
+M(3).W  = exp(0);                            % error precision
+
+ 
+% generative model
+%==========================================================================
+
+% first level
+%--------------------------------------------------------------------------
+G(1).f  = inline('tanh(a) - x/8',  'x', 'v', 'a', 'P');
+G(1).g  = inline('[x; tan(x); v]', 'x', 'v', 'a', 'P');
+G(1).pE = P;
+G(1).x  = [0; 0];                            % hidden states (movement)
+G(1).V  = exp(16);                           % error precision
+G(1).W  = exp(16);                           % error precision
+ 
+ 
+% second level
+%--------------------------------------------------------------------------
+G(2).v  = sparse(n,1);                       % exogenous cues (salience)
+G(2).a  = [0; 0];                            % action
+G(2).V  = exp(16);
+ 
+% Create cues
+%==========================================================================
+dt    = 64;                                  % bin size (ms)
+isi   = 12;                                  % inter stimulus interval
+dur   = 4;                                   % stimulus duration
+j     = [1 2 3 4 1 4 3 2 1 4 3 2 1];         % stimulus order
+ 
+C     = sparse(n,N);
+for i = 1:length(j)
+    C(j(i),:) = C(j(i),:) + exp(-([1:N] - i*isi).^2/(2*dur^2));
+end
+ 
+% generate and invert of over different DA levels (D = V.c)
+%==========================================================================
+DEM.G = G;
+DEM.M = M;
+DEM.C = C;
+
+
+DEM   = spm_ADEM(DEM)
+subplot(4,2,7)
+%spm_dem_set_movie(DEM)
+return
+
+
+% graphics
+%----------------------------------------------------------------------
+[on,rt,ac] = spm_ADEM_set_rt(DEM);
+
+subplot(2,1,1), hold on
+plot(on,rt,'r:',on,rt,'r.','MarkerSize',24)
+xlabel('cue onset (sec)','FontSize',12)
+ylabel('milliseconds','FontSize',12)
+title('reaction times','FontSize',16)
+axis square, box off, hold off
+
+
+
+
+% set precision of salience or affordance
+%--------------------------------------------------------------------------
+D     = (-3:3)/2;
+D     = -2:2; 
+for i = 1:length(D)
+ 
+    ADEM{i}        = DEM;
+    ADEM{i}.M(3).V = ADEM{i}.M(3).V*exp(D(i));
+    ADEM{i}        = spm_ADEM(ADEM{i});   
+    
+end
+
+ 
+% reaction times and accuracy
+%==========================================================================
+spm_figure('GetWin','Figure 4'); clf
+ 
+C     = {'r','g','b','c','m'};
+for i = 1:length(C)
+ 
+    % get behavioral performance
+    %----------------------------------------------------------------------
+    [on,rt,ac] = spm_ADEM_set_rt(ADEM{i});
+ 
+    % graphics
+    %----------------------------------------------------------------------
+    subplot(2,1,1), hold on
+    plot(on,rt,[C{i} ':'],on,rt,[C{i} '.'],'MarkerSize',24)
+    xlabel('cue onset (sec)','FontSize',12)
+    ylabel('milliseconds','FontSize',12)
+    title('reaction times','FontSize',16)
+    axis square, box off, hold off
+ 
+    subplot(2,1,2), hold on
+    plot(on,ac,[C{i} ':'],on,ac,[C{i} '.'],'MarkerSize',24)
+    xlabel('cue onset (sec)','FontSize',12)
+    ylabel('inverse distance','FontSize',12)
+    title('spatial error','FontSize',16)
+    axis square, box off, , hold off, drawnow
+ 
+end
+
+return
+ 
+% dynamics and movies
+%==========================================================================
+ 
+% normal
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 1');
+spm_DEM_qU(ADEM{6}.qU)
+ 
+subplot(3,2,5)
+spm_dem_set_movie(ADEM{6})
+ 
+% delayed set switching
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 2');
+spm_DEM_qU(ADEM{3}.qU)
+ 
+subplot(3,2,5)
+spm_dem_set_movie(ADEM{3})
+ 
+% failure to switch set
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 3');
+spm_DEM_qU(ADEM{2}.qU)
+ 
+subplot(3,2,5)
+spm_dem_set_movie(ADEM{2})
+
+
+% demonstrate autonomous movement (by removing salient information)
+%==========================================================================
+DEM        = DEM;
+V.c        = exp(0) + sparse(n,1);
+DEM.M(1).V = diag(spm_vec(V));
+DEM        = spm_ADEM(DEM);
+ 
+% perseverative dynamics and trajectories
+%--------------------------------------------------------------------------
+spm_figure('GetWin','DEM');
+spm_DEM_qU(DEM.qU)
+ 
+subplot(3,2,5)
+spm_dem_set_movie(DEM)
diff --git a/toolbox/DEM/DCM_HDM.mat b/toolbox/DEM/DCM_HDM.mat
new file mode 100644
index 0000000..54e9920
Binary files /dev/null and b/toolbox/DEM/DCM_HDM.mat differ
diff --git a/toolbox/DEM/DEM.m b/toolbox/DEM/DEM.m
new file mode 100644
index 0000000..1a7c015
--- /dev/null
+++ b/toolbox/DEM/DEM.m
@@ -0,0 +1,3 @@
+% Wrapper for SPM Toolbox menu
+
+DEM_demo;
diff --git a/toolbox/DEM/DEM_Dendrites.mat b/toolbox/DEM/DEM_Dendrites.mat
new file mode 100644
index 0000000..896501a
Binary files /dev/null and b/toolbox/DEM/DEM_Dendrites.mat differ
diff --git a/toolbox/DEM/DEM_birdsong.m b/toolbox/DEM/DEM_birdsong.m
index 59795f0..d98b8b9 100644
--- a/toolbox/DEM/DEM_birdsong.m
+++ b/toolbox/DEM/DEM_birdsong.m
@@ -16,7 +16,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_birdsong.m 1380 2008-04-11 18:55:18Z karl $ 
+% $Id: DEM_birdsong.m 4136 2010-12-09 22:22:28Z guillaume $ 
  
 % load bird song
 %==========================================================================
diff --git a/toolbox/DEM/DEM_demo.fig b/toolbox/DEM/DEM_demo.fig
index 85be7ca..cd8a0c2 100644
Binary files a/toolbox/DEM/DEM_demo.fig and b/toolbox/DEM/DEM_demo.fig differ
diff --git a/toolbox/DEM/DEM_demo.m b/toolbox/DEM/DEM_demo.m
index 377fe1a..64dd146 100644
--- a/toolbox/DEM/DEM_demo.m
+++ b/toolbox/DEM/DEM_demo.m
@@ -22,7 +22,7 @@
 
 % Edit the above text to modify the response to help DEM_demo
 
-% Last Modified by GUIDE v2.5 26-May-2010 19:20:30
+% Last Modified by GUIDE v2.5 04-Mar-2011 21:40:03
 
 % Begin initialization code - DO NOT EDIT
 gui_Singleton = 1;
@@ -248,3 +248,108 @@ function pushbutton95_Callback(hObject, eventdata, handles)
 function pushbutton96_Callback(hObject, eventdata, handles)
 run_demo_Callback(hObject, handles, 'ADEM_observe')
 
+% --- Executes on button press in pushbutton97.
+function pushbutton97_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_DCM_LAP')
+
+% --- Executes on button press in pushbutton98.
+function pushbutton98_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_convolution_LAP')
+
+% --- Executes on button press in pushbutton99.
+function pushbutton99_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_lorenz_LAP')
+
+% --- Executes on button press in pushbutton100.
+function pushbutton100_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_doublewell_LAP')
+
+% --- Executes on button press in pushbutton101.
+function pushbutton101_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_hdm_SCK')
+
+% --- Executes on button press in pushbutton102.
+function pushbutton102_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'ADEM_writing')
+
+% --- Executes on button press in pushbutton103.
+function pushbutton103_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_dendrite')
+
+% --- Executes on button press in pushbutton104.
+function pushbutton104_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_Cornsweet')
+
+% --- Executes on button press in pushbutton105.
+function pushbutton105_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'ADEM_pursuit')
+
+% --- Executes on button press in pushbutton117.
+function pushbutton117_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'ADEM_cued_response')
+
+% --- Executes on button press in pushbutton118.
+function pushbutton118_Callback(hObject, eventdata, handles)
+run_demo_Callback(hObject, handles, 'DEM_demo_MMN_deviance')
+
+% --- Executes on button press in pushbutton119.
+function pushbutton119_Callback(hObject, eventdata, handles)
+% hObject    handle to pushbutton119 (see GCBO)
+run_demo_Callback(hObject, handles, 'DEM_demo_contact_lens')
+
+
+
+
+% PDF callbacks
+%--------------------------------------------------------------------------
+
+% --- Executes on button press in pushbutton107.
+function pushbutton107_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Action_and_behavior_A_free-energy_formulation.pdf']);
+
+% --- Executes on button press in pushbutton108.
+function pushbutton108_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Variational_free_energy_and_the_Laplace_approximation.pdf']);
+
+% --- Executes on button press in pushbutton109.
+function pushbutton109_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Hierarchical_Models_in_the_Brain.pdf']);
+
+% --- Executes on button press in pushbutton110.
+function pushbutton110_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Friston_MathProblEngin_2010_621670.pdf']);
+
+% --- Executes on button press in pushbutton111.
+function pushbutton111_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Variational_filtering.pdf']);
+
+% --- Executes on button press in pushbutton112.
+function pushbutton112_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Cortical_circuits_for_perceptual_inference.pdf']);
+
+% --- Executes on button press in pushbutton113.
+function pushbutton113_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Reinforcement_Learning_or_Active_Inference.pdf']);
+
+% --- Executes on button press in pushbutton114.
+function pushbutton114_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'karl_nonlinear.pdf']);
+
+% --- Executes on button press in pushbutton115.
+function pushbutton115_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+    'Attention_uncertainty_and_free-energy.pdf']);
+
+% --- Executes on button press in pushbutton116.
+function pushbutton116_Callback(hObject, eventdata, handles)
+web(['http://www.fil.ion.ucl.ac.uk/spm/doc/papers/', ...
+'DEM_A_variational_treatment_of_dynamic_systems.pdf']);
+
diff --git a/toolbox/DEM/DEM_demo_Cornsweet.m b/toolbox/DEM/DEM_demo_Cornsweet.m
new file mode 100644
index 0000000..cb14835
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_Cornsweet.m
@@ -0,0 +1,388 @@
+% The Cornsweet effect: This demo illustrates the inference underlying the
+% Cornsweet effect or illusion. It exploits formal priors on the spatial
+% contiguity of the illuminant and reflectance; where the illuminant does not
+% have edges, but the reflectance can. This is implemented using a
+% discrete cosine set (DCT) as the spatial basis for the illuminant and a 
+% (Haar) Discrete Wavelet transform (DWT) for the reflectance. Appropriate
+% shrinkage priors on the (implicit) transform coefficients ensure that the
+% explanation for visual input (reflectance times illuminant) assigns edges
+% to the reflectance; thereby producing the Cornsweet effect.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_Cornsweet.m 4281 2011-03-31 19:49:57Z karl $
+ 
+ 
+% Illustrate the Cornsweet effect
+%==========================================================================
+clear
+spm_figure('GetWin','Figure 1');
+colormap([1:255]'*[1 1 1]/255)
+ 
+% basic profile
+%--------------------------------------------------------------------------
+nx    = 64;
+YX    = [zeros(1,nx*3/8) 1:(nx/8) -(nx/8):-1 zeros(1,nx*3/8)]*8/nx;
+YT    = ones(32,1);
+ 
+% present with different contrasts (C)
+%--------------------------------------------------------------------------
+C     = [1 4 32 64];
+nc    = length(C);
+for i = 1:length(C)
+ 
+    subplot(nc,2,2*i - 1)
+    image(YT*YX*C(i) + 128)
+    title('Cornsweet effect','FontSize',16)
+    xlabel('eccentricity ','FontSize',12)
+    axis square
+ 
+    subplot(nc,2,2*i)
+    plot(YX*C(i))
+    title('reflectance ','FontSize',16)
+    xlabel('eccentricity ','FontSize',12)
+    axis square, axis([1 nx -64 64]);
+    
+end
+ 
+ 
+% Simulations
+%==========================================================================
+ 
+ 
+% Basis functions of illuminant and reflectance
+%--------------------------------------------------------------------------
+nx      = 32;                               % number of pixels
+P.R     = spm_dwtmtx(nx,1,1);               % DWT for hidden causes
+P.I     = spm_dctmtx(nx,3);                 % DCT for hidden causes
+P.I     = P.I/diag(max(P.I));
+nr      = size(P.R,2);
+ni      = size(P.I,2);
+W       = log2(nx./sum(P.R > 0))*3;         % scale of reflectance
+W(1)    = 16;
+
+ 
+% initial hidden states and causes
+%--------------------------------------------------------------------------
+x       = sparse(nr,1);
+v       = sparse(ni + nr,1);
+v(1)    = -4;
+ii      = (1:ni);
+ir      = (1:nr) + ni;
+ 
+% level 1
+%--------------------------------------------------------------------------
+M(1).f  = inline('(v(4:end) - x)/16','x','v','P');
+M(1).g  = inline('exp(P.R*x + P.I*v(1:3))','x','v','P');
+M(1).pE = P;                                % The prior expectation
+M(1).x  = x;                                % The prior expectation
+M(1).V  = exp(6);                           % error precision (data)
+M(1).W  = exp(12);                           % error precision (motion)
+M(1).xP = diag(exp(W));                     % error precision (motion)
+ 
+% level 2
+%--------------------------------------------------------------------------
+M(2).v  = v;                                % hidden causes 
+M(2).V  = exp(0);                           % error precision (cause)
+ 
+% Create stimulus
+%==========================================================================
+ 
+% Create stimulus
+%--------------------------------------------------------------------------
+M(1).E.n = 1;
+N        = 64;                             % length of sequence
+t        = ([1:N] - N/4)*8;                % time (ms)
+YX       = [zeros(1,nx*3/8) 1:(nx/8) -(nx/8):-1 zeros(1,nx*3/8)]*8/nx;
+YX       = YX'/4 + 1;
+YX       = exp(P.R*pinv(P.R)*log(YX));
+YT       = exp((tanh(([1:N] - N/3)/8) - 1)*4);
+YT       = exp(-([1:N] - N/2).^2/(2*(N/8)^2));
+Y        = YX*YT;
+ 
+% invert
+%==========================================================================
+DEM.M    = M;
+DEM.Y    = Y;
+DEM      = spm_DEM(DEM);
+ 
+ 
+% render true and perceived stimuli
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 2'); clf
+ 
+subplot(2,2,1)
+PI  = P.I*DEM.qU.v{2}(ii,:);
+imagesc(exp(PI))
+title('perceived illunimant','FontSize',16)
+xlabel('time (bins)','FontSize',12)
+ylabel('eccentricity ','FontSize',12)
+axis square
+ 
+subplot(2,2,2)
+PR = P.R*DEM.qU.x{1};
+imagesc(exp(PR))
+title('perceived reflectance','FontSize',16)
+xlabel('time (bins)','FontSize',12)
+ylabel('eccentricity ','FontSize',12)
+axis square
+ 
+subplot(2,2,3)
+imagesc(DEM.qU.v{1})
+title('predicted stimulus','FontSize',16)
+xlabel('time (bins)','FontSize',12)
+ylabel('eccentricity ','FontSize',12)
+axis square
+ 
+ 
+% first order effect
+%--------------------------------------------------------------------------
+C     = sparse(1,nx);
+i     = nx/2 - nx/4 - 1;
+C(i)  =  1;
+i     = nx/2 + nx/4 + 2;
+C(i)  = -1;
+R1    = C*P.R;
+ 
+% second order effect
+%--------------------------------------------------------------------------
+i     = nx/2 - nx/8 - 1;
+C(i)  = -1;
+i     = nx/2 + nx/8 + 2;
+C(i)  =  1;
+R2    = C*P.R;
+ 
+D1    = R1*DEM.qU.x{1};
+D2    = R2*DEM.qU.x{1};
+for i = 1:N
+    V1(i) = R1*DEM.qU.S{i}*R1';
+    V2(i) = R2*DEM.qU.S{i}*R2';
+end
+ 
+ 
+subplot(2,2,4)
+spm_plot_ci(t,D1,V1), hold on
+spm_plot_ci(t,D2,V2), hold off
+title('perceived reflectance','FontSize',16)
+xlabel('time (ms)','FontSize',12)
+ylabel('eccentricity ','FontSize',12)
+axis square
+spm_axis tight
+drawnow
+ 
+ 
+% Cycle over different levels of visual precision (cf contrast)
+%==========================================================================
+C   = -2:2:16;
+for i = 1:length(C)
+    
+    % Change precision (contrast) and invert
+    %----------------------------------------------------------------------
+    SIM{i}        = DEM;
+    SIM{i}.M(1).V = exp(C(i));
+    SIM{i}        = spm_DEM(SIM{i});
+ 
+    % Record conditional peripheral reflectance difference
+    %----------------------------------------------------------------------
+    d1(i) = R1*SIM{i}.qU.x{1}(:,N/2);
+    v1(i) = R1*SIM{i}.qU.S{N/2}*R1';
+    d2(i) = R2*SIM{i}.qU.x{1}(:,N/2);
+    v2(i) = R2*SIM{i}.qU.S{N/2}*R2';
+ 
+end
+ 
+ 
+ 
+% show Cornsweet effect as a function of precision (contrast)
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 3'); clf
+ 
+subplot(3,1,1)
+spm_plot_ci(C,d1,v1),                hold on
+plot(C,d1,'ob',C,C*0,'LineWidth',2), hold off
+title('conditional difference (Cornsweet)','FontSize',16)
+xlabel('log-precision (contrast)','FontSize',12)
+ylabel('reflectance difference','FontSize',12)
+spm_axis tight square
+ 
+subplot(3,1,2)
+spm_plot_ci(C,d2,v2),                hold on
+plot(C,d2,'or',C,C*0,'LineWidth',2), hold off
+title('conditional difference (Mach Band)','FontSize',16)
+xlabel('log-precision (contrast)','FontSize',12)
+ylabel('reflectance difference','FontSize',12)
+spm_axis tight square
+ 
+% and associated percepts
+%--------------------------------------------------------------------------
+colormap([1:255]'*[1 1 1]/255)
+j     = [1 3 6];
+nj    = length(j);
+for i = 1:nj
+   
+    % predictions
+    %----------------------------------------------------------------------
+    PR    = P.R*SIM{j(i)}.qU.x{1};
+    PR    = PR*256 + 128;
+    
+    subplot(3,nj,i + 2*nj)
+    image(PR)
+    ylabel('eccentricity','FontSize',12)
+    axis square
+ 
+end
+ 
+% and associated prediction errors (cf ERPs)
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 4'); clf
+as    = 0;
+ax    = 0;
+av    = 0;
+for i = 1:nj
+    
+    
+    % predictions and errors
+    %----------------------------------------------------------------------
+    Ev = SIM{j(i)}.qU.z{2}(ir,:);
+    Es = SIM{j(i)}.qU.z{1};
+    av = max(max(abs(Ev(:))),av);
+    as = max(max(abs(Es(:))),as);
+    
+    subplot(3,nj,i)
+    plot(t,Es,'r')
+    title('error (sensory)','FontSize',14)
+    xlabel('time (ms)','FontSize',12)
+    ylabel('error (state)','FontSize',12)
+ 
+    subplot(3,nj,i + nj)
+    plot(t,Ev,'r')
+    title('error (hidden states)','FontSize',14)
+    xlabel('time (ms)','FontSize',12)
+    ylabel('error (state)','FontSize',12)
+    
+ 
+end
+
+
+% adjust axes
+%--------------------------------------------------------------------------
+for i = 1:nj
+    
+    subplot(3,nj,i)
+    axis square, axis([t(1) t(end) -as*1.2 as*1.2])
+    
+    subplot(3,nj,i + nj)
+    axis square, axis([t(1) t(end) -av*1.2 av*1.2])
+ 
+end
+ 
+ 
+% redraw inference for effect
+%--------------------------------------------------------------------------
+spm_figure('GetWin','DEM');
+spm_DEM_qU(SIM{7}.qU)
+spm_figure('GetWin','Figure 3');
+
+
+% Fitting behavioural data
+%==========================================================================
+
+
+ 
+ 
+return
+ 
+ 
+% Ancillary code for producing figures:
+%==========================================================================
+spm_figure('GetWin','paper'); clf
+ 
+n = 128;
+a = 4;
+i = tanh(([1:n] - n/2)/(n/8))/a;
+r = kron([1 -1],ones(1,n/2))/a;
+s = exp(i + r);
+c = zeros(1,n);
+ 
+subplot(5,1,1), plot(i),     spm_axis tight, title('illuminant')
+subplot(5,1,2), plot(r),     spm_axis tight, title('reflectant')
+subplot(5,1,3), plot(s),     spm_axis tight, title('stimulus')
+subplot(5,1,4), plot(i + r), spm_axis tight, title('reflectant')
+subplot(5,1,5), plot(c),     spm_axis tight, title('illuminant')
+ 
+ 
+% Illustrate the generative model (for 1D)
+%==========================================================================
+spm_figure('GetWin','paper'); clf
+ 
+DEM      = spm_DEM_generate(M,64,P,{16 0},{16});
+ 
+subplot(3,2,1)
+PI  = P.I*DEM.pU.v{2}(ii,:);
+imagesc(exp(PI))
+title('illunimant','FontSize',16)
+xlabel('time (bins)','FontSize',12)
+ylabel('location ','FontSize',12)
+axis square
+ 
+subplot(3,2,3)
+PR = P.R*DEM.pU.x{1};
+imagesc(exp(PR))
+title('reflectant','FontSize',16)
+xlabel('time (bins)','FontSize',12)
+ylabel('location ','FontSize',12)
+axis square
+ 
+subplot(3,2,5)
+imagesc(DEM.pU.v{1})
+title('stimulus','FontSize',16)
+xlabel('time (bins)','FontSize',12)
+ylabel('location ','FontSize',12)
+axis square
+ 
+ 
+subplot(3,2,2), imagesc(P.I), axis image off
+subplot(3,2,4), imagesc(P.R), axis image off
+ 
+ 
+% Illustrate the generative model (for 2D)
+%==========================================================================
+nx    = 128;
+P.R   = spm_dwtmtx(nx,2);                 % DWT for hidden causes
+P.I   = spm_dctmtx(nx,3);                 % DCT for hidden causes
+I     = 0;
+R     = 0;
+for i = 1:size(P.I,2)
+    for j = 1:size(P.I,2)
+        I = I + (P.I(:,i)*P.I(:,j)')*randn;
+    end
+end
+for i = 1:size(P.R,2)
+    for j = 1:size(P.R,2)
+        v = sum(P.R(:,i) > 0)*sum(P.R(:,j) > 0);
+        R = R + (P.R(:,i)*P.R(:,j)')*randn*sqrt(v);
+    end
+end
+ 
+R   = R/std(R(:))/2;
+I   = I/std(I(:))/2;
+ 
+% Plot
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Paper');
+ 
+subplot(2,2,1)
+imagesc(exp(I))
+title('illuminant ','FontSize',16)
+axis square
+ 
+subplot(2,2,2)
+imagesc(exp(R))
+title('reflectance ','FontSize',16)
+axis square
+ 
+subplot(2,1,2)
+imagesc(exp(I + R))
+title('image','FontSize',16)
+axis square, drawnow
diff --git a/toolbox/DEM/DEM_demo_DCM_LAP.m b/toolbox/DEM/DEM_demo_DCM_LAP.m
new file mode 100644
index 0000000..aaceb98
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_DCM_LAP.m
@@ -0,0 +1,205 @@
+% Demo applying the Laplace scheme to DCM with hidden states
+%__________________________________________________________________________
+% This routine demonstrates Generalised filtering for a DCM (Dynamic Causal
+% Model) of fMRI respsosnes using simulated data. This is an endogenous 
+% DCM in that there are no exogenous inputs. The demonstration specifies 
+% and inverts a full connectivity model and then illustrates post-hoc model
+% optimisation to recover (discover) the true architecture. It concludes 
+% with an automatic model optimization in terms of the prior variances over
+% coupling parameters.
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_DCM_LAP.m 4170 2011-01-24 18:37:42Z karl $
+ 
+% Specify a DCM to generte synthetic data
+%==========================================================================
+clear
+ 
+% DEM Structure: create inputs
+% -------------------------------------------------------------------------
+T  = 256;
+TR = 3.22;
+t  = (1:T)*TR;
+n  = 4;
+U  = spm_conv(randn(n,T),0,2)/exp(4/2);
+ 
+% set inversion parameters
+% -------------------------------------------------------------------------
+M(1).E.s  = 1/2;         % smoothness of random fluctuations
+M(1).E.d  = 2;           % embedding dimension
+M(1).E.n  = 6;           % embedding dimension
+M(1).E.nE = 32;          % maximum number of DEM iterations
+ 
+ 
+% priors
+% -------------------------------------------------------------------------
+A  = ones(n,n);
+B  = zeros(n,n,0);
+C  = zeros(n,n);
+D  = zeros(n,n,0);
+ 
+[pE,pC] = spm_dcm_fmri_priors(A,B,C,D);
+ 
+ 
+% adjust M.f (GF works in time bins not seconds) and initialize M.P
+% -------------------------------------------------------------------------
+M(1).f  = inline(['spm_fx_fmri(x,v,P)*' num2str(TR)],'x','v','P');
+M(1).g  = 'spm_gx_fmri';
+M(1).x  = sparse(n,5);
+M(1).pE = pE;
+M(1).pC = pC;
+ 
+M(2).v  = sparse(n,1);
+ 
+ 
+% simulate endogenous dynamics
+%==========================================================================
+ 
+% true parameters (stochastic estimates)
+% -------------------------------------------------------------------------
+pP   = pE;
+a    =  0.3;
+b    = -0.3;
+c    = -0.5;
+pP.A = [c  a  0  0  0;
+        a  c  b  0  0;
+        0  b  c  a  0;
+        0  0  a  c  b;
+        0  0  0  a  c];
+     
+pP.A = pP.A(1:n,1:n);
+pP.C = eye(n,n);
+SIM  = spm_DEM_generate(M,U,{pP},{6,16},{16});
+ 
+ 
+% Show simulated response
+%--------------------------------------------------------------------------
+spm_figure('Getwin','Figure 1');
+spm_DEM_qU(SIM.pU)
+ 
+ 
+% Specify generative model for inversion (DCM)
+% =========================================================================
+ 
+% set inversion parameters
+% -------------------------------------------------------------------------
+DCM.M       = M;
+DCM.M(2).v  = 0;
+ 
+% Specify hyper-priors on precisions
+% -------------------------------------------------------------------------
+W           = exp(spm_vec(sparse(1:n,1,(6 - 16),n,5) + 16));
+DCM.M(1).xP = exp(6);
+DCM.M(1).V  = exp(6);        % prior log precision (noise)
+DCM.M(1).W  = diag(W);       % fixed precision (hidden-state)
+DCM.M(2).V  = exp(16);       % fixed precision (hidden-cause)
+ 
+% Add data
+% -------------------------------------------------------------------------
+DCM.Y       = SIM.Y;
+ 
+% Full connectivity inversion
+% =========================================================================
+F  = ones(n,n);
+B  = zeros(n,n,0);
+C  = zeros(n,1);
+D  = zeros(n,n,0);
+ 
+[pE pC]     = spm_dcm_fmri_priors(F,B,C,D);
+DCM.M(1).pE = pE;
+DCM.M(1).pC = pC;
+FULL        = spm_LAP(DCM);
+ 
+ 
+% Search model space with Savage-Dickey density ratio
+% =========================================================================
+[A K Nk]  = spm_dcm_sparse_priors(n);
+ 
+% find candidate models based on full-connectivity
+% -------------------------------------------------------------------------
+pE    = FULL.M(1).pE.A;
+qE    = FULL.qP.P{1}.A;
+qC    = FULL.qP.C(1:n*n,1:n*n);
+pC    = FULL.M(1).pC(1:n*n,1:n*n);
+for i = 1:length(A)
+    k       = find(~A{i});       
+    rE      = pE;
+    rC      = pC;
+    rE(k)   = 0;
+    rC(k,k) = 0;
+    P(i,1)  = spm_log_evidence(qE,qC,pE,pC,rE,rC);
+end
+ 
+% find true model
+% -------------------------------------------------------------------------
+for i = 1:length(A);
+    if ~any(spm_vec(~~SIM.pP.P{1}.A - A{i}))
+        tA = i; break
+    end
+end
+ 
+% log-posterior (model)
+% -------------------------------------------------------------------------
+PP    = exp(P - max(P));
+PP    = PP/sum(PP);
+ 
+% Graphics (density on parameter and model space)
+% -------------------------------------------------------------------------
+spm_figure('Getwin','Figure 2'); clf
+ 
+subplot(2,2,1)
+spm_plot_ci(spm_vec(qE),qC),     hold on
+bar(spm_vec(pP.A),1/2), hold off
+title('true and MAP connections','FontSize',16)
+axis square
+ 
+subplot(2,2,2)
+bar(P)
+title('log-posterior','FontSize',16)
+xlabel('model','FontSize',12)
+ylabel('log-probability','FontSize',12)
+axis square
+ 
+subplot(2,2,3)
+plot(Nk,    P,    '.k','MarkerSize',16), hold on
+plot(Nk(tA),P(tA),'.r','MarkerSize',32), hold off
+title('log-evidence','FontSize',16)
+xlabel('graph size','FontSize',12)
+ylabel('log-probability','FontSize',12)
+axis square
+ 
+subplot(2,2,4)
+bar(PP)
+title('posterior','FontSize',16)
+xlabel('model','FontSize',12)
+ylabel('probability','FontSize',12)
+axis square
+ 
+ 
+% Compare true and AMS adjacency
+% -------------------------------------------------------------------------
+spm_figure('Getwin','Figure 3'); clf
+ 
+% Automatic model selection (optimising prior variances on parameters)
+% =========================================================================
+[rE,rC]   = spm_dcm_optimise(qE,qC,pE,pC);
+rA        = spm_unvec(diag(rC),pE);
+ 
+subplot(2,2,1)
+imagesc(full(A{tA} - diag(diag(A{tA}))))
+title('true adjacency','FontSize',16)
+xlabel('source','FontSize',12)
+ylabel('target','FontSize',12)
+axis square
+ 
+subplot(2,2,2)
+imagesc(full(rA))
+title('optmised priors','FontSize',16)
+xlabel('source','FontSize',12)
+ylabel('target','FontSize',12)
+axis square
+ 
+
+
diff --git a/toolbox/DEM/DEM_demo_DEM.m b/toolbox/DEM/DEM_demo_DEM.m
index 841db5c..22e72f5 100644
--- a/toolbox/DEM/DEM_demo_DEM.m
+++ b/toolbox/DEM/DEM_demo_DEM.m
@@ -27,7 +27,7 @@
  
 % generate data and invert
 %==========================================================================
-M(1).E.nN = 16;                                % DEM-steps
+M(1).E.nE = 16;                                % DEM-steps
 N         = 32;                                % length of data sequence
 U         = exp(-([1:N] - 12).^2/(2.^2));      % this is the Gaussian cause
 DEM       = spm_DEM_generate(M,U,{P},{8,32},{32});
diff --git a/toolbox/DEM/DEM_demo_DFP.m b/toolbox/DEM/DEM_demo_DFP.m
index 6e83ab4..0eb97dc 100644
--- a/toolbox/DEM/DEM_demo_DFP.m
+++ b/toolbox/DEM/DEM_demo_DFP.m
@@ -36,12 +36,9 @@
  
 % overlay true values
 %--------------------------------------------------------------------------
-spm_figure('GetWin','Graphics')
+spm_figure('GetWin','Graphics');
 spm_DEM_qU(DEM.qU,DEM.pU)
 
-if ~strcmp(questdlg('proceed with DEM using the same model and data'),'Yes')
-    return
-end
  
 % invert model - DEM
 %==========================================================================
diff --git a/toolbox/DEM/DEM_demo_EM.m b/toolbox/DEM/DEM_demo_EM.m
index ef730a9..2babc6a 100644
--- a/toolbox/DEM/DEM_demo_EM.m
+++ b/toolbox/DEM/DEM_demo_EM.m
@@ -86,7 +86,7 @@
  
 % EM with a Gauss-Newton-like optimization of free energy
 %==========================================================================
-[Ep,Cp,S,F] = spm_nlsi_GN(G,GU,GY);
+[Ep,Cp,Eh,F] = spm_nlsi_GN(G,GU,GY);
  
 % parameters
 %--------------------------------------------------------------------------
@@ -135,7 +135,7 @@
     % EM
     %----------------------------------------------------------------------
     GY.y  = DEM.Y';
-    [Ep,Cp,S,F] = spm_nlsi_GN(G,GU,GY);
+    [Ep,Cp,Eh,F] = spm_nlsi_GN(G,GU,GY);
  
     % retain parameter estimates
     %----------------------------------------------------------------------
@@ -148,7 +148,7 @@
     EP(:,i) = eP;
  
     QH(i) = DEM.qH.h{1}(1);
-    EH(i) = -exp(S(1));
+    EH(i) = Eh(1);
 end
  
 clf
diff --git a/toolbox/DEM/DEM_demo_GLM.m b/toolbox/DEM/DEM_demo_GLM.m
index 8d10ae8..06d3ce1 100644
--- a/toolbox/DEM/DEM_demo_GLM.m
+++ b/toolbox/DEM/DEM_demo_GLM.m
@@ -11,12 +11,12 @@
  
 % specify parameters
 %--------------------------------------------------------------------------
-N     = 4;                                        % length of data sequence
+N     = 16;                                        % length of data sequence
 X     = randn(8,2);
  
 % generate data
 %--------------------------------------------------------------------------
-DEM   = spm_DEM_generate(spm_DEM_M('GLM',X),N,{},{1})
+DEM   = spm_DEM_generate(spm_DEM_M('GLM',X),N,{},{1});
  
 % DEM estimation
 %==========================================================================
@@ -27,7 +27,7 @@
  
 % Classical (OLS) estimation of states
 %==========================================================================
-[Ce,h,W,L] = spm_reml_sc(DEM.Y*DEM.Y',X,DEM.M(1).Q,N);
+[Ce,h,W,L] = spm_reml(DEM.Y*DEM.Y',X,DEM.M(1).Q,N);
 V          = pinv(X)*DEM.Y;
 C          = h*pinv(X)*DEM.M(1).Q{1}*pinv(X)';
  
diff --git a/toolbox/DEM/DEM_demo_LAP.m b/toolbox/DEM/DEM_demo_LAP.m
index 4b9a26f..41217b0 100644
--- a/toolbox/DEM/DEM_demo_LAP.m
+++ b/toolbox/DEM/DEM_demo_LAP.m
@@ -11,7 +11,7 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_demo_LAP.m 3878 2010-05-07 19:53:54Z karl $
+% $Id: DEM_demo_LAP.m 4146 2010-12-23 21:01:39Z karl $
  
 % get basic convolution model
 %==========================================================================
@@ -32,28 +32,27 @@
 np      = length(pE);
 pE(ip)  = 0;
 pE      = spm_unvec(pE,P);
-pC      = sparse(ip,ip,32,np,np);
+pC      = sparse(ip,ip,exp(8),np,np);
 M(1).pE = pE;
 M(1).pC = pC;
  
-% free hyperparameters
+% free hyperparameters+
 %--------------------------------------------------------------------------
 M(1).Q  = {speye(M(1).l,M(1).l)};
 M(1).R  = {speye(M(1).n,M(1).n)};
-M(1).hE = 4;
-M(1).gE = 4;
-M(1).hC = 1/2;
-M(1).gC = 1/2;
-M(1).V  = 0;
-M(1).W  = 0;
+M(1).hE = 6;
+M(1).gE = 6;
+M(1).hC = 1/4;
+M(1).gC = 1/4;
  
 % generate data and invert
 %==========================================================================
-M(1).E.nN = 24;                                % number of time steps
+M(1).E.nN = 16;                                % number of E steps
+M(1).E.nE = 16;                                % number of E steps
 M(1).E.nD = 1;                                 % number of time steps
-M(1).E.s  = 1/4;                               % smoothness
+M(1).E.s  = 1;                                 % smoothness
 M(1).E.d  = 2;                                 % order
-M(1).E.n  = 4;                                 % order
+M(1).E.n  = 6;                                 % order
  
 N         = 32;                                % length of data sequence
 U         = exp(-([1:N] - 12).^2/(2.^2));      % this is the Gaussian cause
diff --git a/toolbox/DEM/DEM_demo_MMN.m b/toolbox/DEM/DEM_demo_MMN.m
index 6146736..bd60fec 100644
--- a/toolbox/DEM/DEM_demo_MMN.m
+++ b/toolbox/DEM/DEM_demo_MMN.m
@@ -13,7 +13,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_demo_MMN.m 3740 2010-02-26 13:13:14Z karl $
+% $Id: DEM_demo_MMN.m 4146 2010-12-23 21:01:39Z karl $
  
 
 % level 1
@@ -71,9 +71,8 @@
 M(1).E.s  = 1;                    % temporal smoothness
  
 M(1).E.nD = 1;                    % D-steps per time bin (1 for dynamic systems)
-M(1).E.nE = 2;                    % E-steps per iteration
+M(1).E.nE = 1;                    % E-steps per iteration
 M(1).E.nM = 8;                    % M-steps per iteration
-M(1).E.nN = 1;                    % number of iterations
  
  
 % Because we want to record the response of the model over time, to each
@@ -176,19 +175,21 @@
  
 % ERPs (precision weighted) prediction error
 %--------------------------------------------------------------------------
-spm_figure('GetWin','Figure 2'); 
+spm_figure('GetWin','Figure 2');
+a     = [t(1) t(end) -30 30];
+b     = [t(1) t(end) -1 1];
 for i = 1:n
     
     subplot(6,n,i + 2*n)
     plot(t,qR{i + 1}{1},'r')
     axis square tight
-    if i == 1, a = axis; else, axis(a); end
+    axis(a)
     box off
         
     subplot(6,n,i + 3*n)
     plot(t,qR{i + 1}{2},'r')
     axis square tight
-    if i == 1, b = axis; else, axis(b); end
+    axis(b)
     box off
     
 end
diff --git a/toolbox/DEM/DEM_demo_MMN_deviance.m b/toolbox/DEM/DEM_demo_MMN_deviance.m
new file mode 100644
index 0000000..e88bc94
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_MMN_deviance.m
@@ -0,0 +1,175 @@
+% This Demo is a more refined illustration of the mismatch negativity that
+% uses a plausible (nonlinear) mapping from hidden states to sensory
+% samples; namely, a discrete spectral density (with a fixed frequency or
+% pitch tuning). This is modelled using a radial basis function over
+% frequencies, whose location parameter is a hidden state encoding pitch and
+% whose amplitude is modulated dynamically by hidden states encoding
+% amplitude modulation (the sum of two squared hidden states showing damped
+% linear oscillation when perturbed by a hidden cause). The recognition
+% dynamics illustrate a dissociation between the effects of changing the
+% (i) the level of deviancy of a stimulus (modelled here as a deviation of 
+% the hidden state (cause) encoding pitch from zero) and (ii) the
+% probability of encountering a pitch deviant. This is modelled by changing
+% the precision on the hidden (pitch) state. Crucially, the nonlinearities
+% in this more plausible generative model of pure tone stimuli induce a
+% latency difference in the mismatch response and increase the amplitude
+% of the mismatch (i.e., MMN). Conversely, changing the precision only
+% affects the amplitude. In these simulations the MMN is modelled simply as
+% the difference between  prediction errors evoked by an expected stimulus
+% (the standard) and a deviant stimulus. Prior expectations are encoded
+% in terms of hidden causes, where the onset of the stimulus is known.
+% This means the agent has correct prior expectations about amplitude
+% modulation (i.e., knows when to expect a stimulus) but can have
+% incorrect expectations about its pitch (of varying confidence or
+% precision).
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_MMN_deviance.m 4230 2011-03-07 20:58:38Z karl $
+ 
+ 
+% Create the generative model
+%==========================================================================
+clear
+ 
+% level 1
+%--------------------------------------------------------------------------
+P       = [-1  4;                            % parameters of motion or
+           -2 -1]/16;                        % Jacobian
+ 
+M(1).f  = inline('[P*x(1:2); (v(2) - x(3))/32] + [v(1); 0; 0]','x','v','P');
+M(1).g  = inline('sum(x(1:2).^2)*exp(-([-2;-1;0;1;2] - x(3)*2).^2/2)/32','x','v','P');
+M(1).pE = P;                                 % The prior expectation
+M(1).x  = zeros(3,1);                        % hidden states
+M(1).V  = exp(4);                            % precision (data)
+M(1).W  = exp(4);                            % precision (motion)
+M(1).xP = exp(-4);                           % precision on states
+ 
+% level 2
+%--------------------------------------------------------------------------
+M(2).v  = zeros(2,1);                        % hidden causes
+M(2).V  = diag(exp([2,0]));                  % precision
+ 
+ 
+% Create hidden causes that determine amplitude and pitch
+%--------------------------------------------------------------------------
+N       = 64;                                % length of data sequence
+dt      = 0.005;                             % time bin (sec)
+T       = N*dt;                              % duration of chirp (sec)
+pst     = [1:N]*dt*1000;                     % time bin (ms)
+C       = exp(-([1:N] - 20).^2/(8.^2));      % amplitude
+U       = [C; (zeros(1,N))];                 % pitch
+ 
+ 
+ 
+% Difference waveforms under different levels of deviance (from zero)
+%==========================================================================
+spm_figure('GetWin','Figure 1');
+M(1).E.s = 1;                                % temporal smoothness
+ 
+DEM   = {};
+D     = [0 0.75 1 1.25 1.5];
+n     = length(D);
+for i = 1:n
+    
+    % generate stimulus
+    %----------------------------------------------------------------------
+    C(1,:)   = U(1,:);
+    C(2,:)   = U(2,:) + D(i);
+    DEM{i}   = spm_DEM_generate(M,C,P,{16 16},{16});
+    
+    % solve (under prior expectations about a standard stimulus)
+    %----------------------------------------------------------------------
+    spm_figure('GetWin','DEM');
+    
+    DEM{i}.U = U;
+    DEM{i}   = spm_DEM(DEM{i});
+    
+end
+ 
+% graphics (ERPs at different levels)
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 1');
+for i = 1:n
+    
+    subplot(n,2,(i - 1)*2 + 1)
+    qx{i} = spm_DEM_MEG(DEM{i},dt,1,1);
+    str = sprintf('(N1/P1) deviance: %.2f',D(i));
+    title(str,'FontSize',12)
+    
+    subplot(n,2,(i - 1)*2 + 2)
+    qv{i} = spm_DEM_MEG(DEM{i},dt,2,1);
+    title('secondary level (MMN)','FontSize',12)
+    drawnow
+end
+ 
+ 
+% graphics (difference waveforms)
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 2');
+for i = 1:n
+    MMN(i,:) = sum(qx{i}{1}) - sum(qx{1}{1});
+    leg{i}   = sprintf('deviance: %.2f',D(i));
+end
+ 
+subplot(2,1,1), plot(pst,MMN), legend(leg), drawnow
+title('Difference waveforms under different levels of deviance','FontSize',16)
+xlabel('prestimulus time (ms)','FontSize',12)
+ 
+ 
+ 
+ 
+% Difference waveforms under different levels of confidence (D)
+%==========================================================================
+spm_figure('GetWin','DEM');
+    
+DEM   = {};
+D     = [0:4] - 4;
+n     = length(D);
+for i = 1:n
+    
+    % change probabilistic context - log-precision of hidden pitch
+    %----------------------------------------------------------------------
+    M(1).xP   = diag(exp([-8 -8 D(i)]));
+    
+    % generate standard
+    %----------------------------------------------------------------------
+    C(1,:)   = U(1,:);
+    C(2,:)   = U(2,:);
+    DEM{i}   = spm_DEM_generate(M,C,P,{16 16},{16});
+    
+    
+    % solve and record sensory prediction errors
+    %----------------------------------------------------------------------
+    DEM{i}.U = U;
+    DEM{i}   = spm_DEM(DEM{i});
+    qs{i}    = spm_DEM_MEG(DEM{i},dt,1);
+    
+    % generate deviant
+    %----------------------------------------------------------------------
+    C(1,:)   = U(1,:);
+    C(2,:)   = U(2,:) + 2;
+    DEM{i}   = spm_DEM_generate(M,C,P,{16 16},{16});
+    
+    
+    % solve and record sensory prediction errors
+    %----------------------------------------------------------------------
+    DEM{i}.U = U;
+    DEM{i}   = spm_DEM(DEM{i});
+    qd{i}    = spm_DEM_MEG(DEM{i},dt,1);
+    
+end
+ 
+ 
+% graphics (differences)
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Figure 2');
+for i = 1:n
+    MMN(i,:) = sum(qd{i}{1}) - sum(qs{1}{1});
+    leg{i}   = sprintf('log-precision: %.2f',D(i));
+end
+
+subplot(2,1,2), plot(pst,MMN), legend(leg)
+title('Difference waveforms under different levels of confidence','FontSize',16)
+xlabel('prestimulus time (ms)','FontSize',12)
diff --git a/toolbox/DEM/DEM_demo_MMN_gen.m b/toolbox/DEM/DEM_demo_MMN_gen.m
index 17aeca8..8410128 100644
--- a/toolbox/DEM/DEM_demo_MMN_gen.m
+++ b/toolbox/DEM/DEM_demo_MMN_gen.m
@@ -14,7 +14,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_demo_MMN_gen.m 2907 2009-03-20 13:02:24Z karl $
+% $Id: DEM_demo_MMN_gen.m 4146 2010-12-23 21:01:39Z karl $
  
  
  
@@ -68,9 +68,8 @@
 % number of steps
 %--------------------------------------------------------------------------
 M(1).E.nD = 1;
-M(1).E.nE = 2;
+M(1).E.nE = 1;
 M(1).E.nM = 8;
-M(1).E.nN = 1;
  
 % parameterise input or cause
 %==========================================================================
diff --git a/toolbox/DEM/DEM_demo_PEB.m b/toolbox/DEM/DEM_demo_PEB.m
index 2cc6203..4b8053b 100644
--- a/toolbox/DEM/DEM_demo_PEB.m
+++ b/toolbox/DEM/DEM_demo_PEB.m
@@ -20,6 +20,8 @@
  
 % DEM estimation
 %==========================================================================
+DEM.M(1).E.nE = 16;
+
 [DEM] = spm_DEM(DEM);
 qU    = DEM.qU;
 qP    = DEM.qP;
diff --git a/toolbox/DEM/DEM_demo_contact_lens.m b/toolbox/DEM/DEM_demo_contact_lens.m
new file mode 100644
index 0000000..cba6916
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_contact_lens.m
@@ -0,0 +1,143 @@
+% This demo illustrates tracking under the contact lens problem:
+% The contact lens refers to the non-Gaussian uncertainty induced by
+% nonlinear measurements. Here it is illustrated in terms of tracking the
+% motion of a target in Cartesian coordinates, given the distance to target
+% (range) and direction as measurements. The problem is to accumulate
+% information over time about the target location under random fluctuations
+% on the velocity (technically this is a constant acceleration model).
+% Comparative evaluations are made with Extended Kalman filtering.
+%
+% See: X. Tian, Y. Bar-Shalom, Coordinate Conversion and Tracking for 
+% Very Long Range Radars. IEEE Transactions on Aerospace and Electronic
+% Systems, AES-45(3):1073–1088, July 2009.
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_contact_lens.m 4279 2011-03-31 11:50:19Z karl $
+ 
+ 
+% non-linear generative model
+%==========================================================================
+clear
+reset(RandStream.getDefaultStream);
+ 
+% The problem: states = [x(1) x(2)]; causes = [v(1) v(2)]
+%--------------------------------------------------------------------------
+x        = [1e4 1e4 -8 -2]';   % initial states: location = 10km,  10km
+                               %                 velocity = -8m/s, -2m/s
+V        = [1e-2 4];           % observation precision (inverse variance) 
+                               % standard deviation of range: 1/sqrt(V(1)) = 10m
+                               % standard deviation of angle: 1/sqrt(V(2)) = .5 mrad
+s        = 1;                  % smoothness of fluctuations
+w        = 8;                  % precision of fluctuations in motion
+W        = [128 128 w w];      % standard deviation of velocity: 1/sqrt(w) = .3536 m/s^2
+
+% precision of fluctuations on hidden states
+%--------------------------------------------------------------------------
+
+
+                               
+ % preliminaries
+%--------------------------------------------------------------------------
+N        = 256;                % length of sequence
+t        = 1:N;                % time (seconds)
+M(1).E.s = s;                  % smoothness of fluctuations
+M(1).E.n = 4;                  % order of generalised coordinates
+M(1).E.K = 128;                % rate of generalized gradient descent
+ 
+% Level 1: Hidden states = [x(1) x(2) v(1) v(2)];
+%--------------------------------------------------------------------------
+f        = '[x(3); x(4); 0; 0]';
+g        = '[sqrt(x(1)^2 + x(2)^2); 1000*atan(x(2)/x(1))]';
+M(1).f   = inline(f,'x','v','P');
+M(1).g   = inline(g,'x','v','P');
+M(1).x   = x;
+M(1).V   = diag(V);            % precision of observation noise
+M(1).W   = diag(W);            % precision of fluctuations on hidden states
+ 
+   
+% create data
+%==========================================================================
+DEM    = spm_DEM_generate(M,N);
+ 
+spm_figure('Getwin','DEM');
+spm_DEM_qU(DEM.pU)
+ 
+ 
+% Comparative inversions (variants of generalised filtering)
+%==========================================================================
+
+% reset initial position and bearing
+%--------------------------------------------------------------------------
+DEM.M(1).x = [1e3; 1e3; 0; 0];
+
+% DEM and EKF
+%--------------------------------------------------------------------------
+DEM     = spm_DEM(DEM);
+[EKF S] = spm_ekf(DEM.M,DEM.Y);
+ 
+spm_figure('Getwin','DEM');
+spm_DEM_qU(DEM.qU,DEM.pU)
+ 
+ 
+% show prediction errors
+%==========================================================================
+spm_figure('Getwin','Graphics');
+ 
+D(1,:) = sqrt(sum((DEM.pU.x{1}([1 2],:) - DEM.qU.x{1}([1 2],:)).^2));
+D(2,:) = sqrt(sum((DEM.pU.x{1}([1 2],:) - EKF([1 2],:)).^2));
+D(3,:) = sqrt(sum((DEM.pU.x{1}([3 4],:) - DEM.qU.x{1}([3 4],:)).^2));
+D(4,:) = sqrt(sum((DEM.pU.x{1}([3 4],:) - EKF([3 4],:)).^2));
+
+for i = 1:N
+    E         = DEM.pU.x{1}(:,i) - DEM.qU.x{1}(:,i);
+    NEES(1,i) = E'*spm_inv(DEM.qU.S{i})*E;
+    E         = DEM.pU.x{1}(:,i) - EKF(:,i);
+    NEES(2,i) = E'*spm_inv(S{i})*E;
+end
+
+
+% plot errors
+%--------------------------------------------------------------------------
+subplot(2,2,1)
+plot(t,log(D(1:2,:)))
+title('log(location error (m))','Fontsize',16)
+xlabel('time (secs)')
+ylabel('log distance from target')
+axis square
+legend('DEM','EKF')
+ 
+subplot(2,2,2)
+plot(t,log(D(3:4,:)))
+title('log(speed error (m/s))','Fontsize',16)
+xlabel('time (secs)')
+ylabel('log distance from target')
+axis square
+legend('DEM','EKF')
+
+
+% plot trajectories
+%--------------------------------------------------------------------------
+subplot(2,2,3)
+plot(DEM.pU.x{1}(1,:),DEM.pU.x{1}(2,:)     ),hold on
+plot(DEM.qU.x{1}(1,:),DEM.qU.x{1}(2,:),'-.'),hold on
+plot(EKF(1,:),EKF(2,:),':'),hold off
+title('location (m)','Fontsize',16)
+axis([8 12 8 12]*1000)
+xlabel('x')
+ylabel('y')
+axis square
+legend('true','DEM','EKF')
+
+
+% plot NEES
+%--------------------------------------------------------------------------
+subplot(2,2,4)
+plot(t,log(NEES))
+title('log(NEES)','Fontsize',16)
+xlabel('time (secs)')
+axis square
+legend('DEM','EKF')
+drawnow
+
diff --git a/toolbox/DEM/DEM_demo_convolution.m b/toolbox/DEM/DEM_demo_convolution.m
index b2246d1..edc05ce 100644
--- a/toolbox/DEM/DEM_demo_convolution.m
+++ b/toolbox/DEM/DEM_demo_convolution.m
@@ -10,14 +10,14 @@
 f       = spm_figure('GetWin','Graphics');
 M       = spm_DEM_M('convolution model');
 
-M(1).V  = exp(8);                           % error precision
-M(1).W  = exp(16);                          % error precision
+M(1).V  = exp(8);                             % error precision
+M(1).W  = exp(16);                             % error precision
  
 % and generate data
 %==========================================================================
 N       = 32;                                 % length of data sequence
 U       = exp(-([1:N] - 12).^2/(2.^2));       % this is the Gaussian cause;
-DEM     = spm_DEM_generate(M,U,{},{[] 16});
+DEM     = spm_DEM_generate(M,U,{},{[] 16},{16});
  
 % display
 %--------------------------------------------------------------------------
diff --git a/toolbox/DEM/DEM_demo_convolution_LAP.m b/toolbox/DEM/DEM_demo_convolution_LAP.m
new file mode 100644
index 0000000..73fcb55
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_convolution_LAP.m
@@ -0,0 +1,111 @@
+% Linear convolution revisited: A dual estimation problem
+%__________________________________________________________________________
+% This demonstration compares generalised filtering and a state-of-the-art 
+% Bayesian smoother (SCKS) in the context of dual estimation. Note that the
+% parameter estimates are smaller then the true values for generalised 
+% schemes (LAP and DEM). This is largely due to the shrinkage priors and 
+% optimisation of model evidence (marginal likelihood), as opposed to the
+% likelihood optimised by the Square-root Cubature Kalman Smoother (SCKS).
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_convolution_LAP.m 4146 2010-12-23 21:01:39Z karl $
+ 
+ 
+% create model and data
+%==========================================================================
+M       = spm_DEM_M('convolution model');
+ 
+% SCK needs to be able to evaluate model equations in matrix form, i.e. for all
+% cubature point at once! (in order to avoid FOR loops and speed up the process)
+%--------------------------------------------------------------------------
+M(1).f  = inline('[P(1,:).*x(1,:)+P(3,:).*x(2,:)+P(13,:).*v(:,:);P(2,:).*x(1,:)+P(4,:).*x(2,:)+P(14,:).*v(:,:);]','x','v','P');
+M(1).g  = inline('[P(5,:).*x(1,:)+P(9,:).*x(2,:);P(6,:).*x(1,:)+P(10,:).*x(2,:);P(7,:).*x(1,:)+P(11,:).*x(2,:);P(8,:).*x(1,:)+P(12,:).*x(2,:);]','x','v','P');
+M(1).V  = exp(8);                              % error precision
+M(1).W  = exp(6);                              % error precision
+M(2).V  = exp(8);
+ 
+% free parameters
+%--------------------------------------------------------------------------
+P       = spm_vec(M(1).pE);                    % true parameters
+ip      = [2 4 5 9];                           % free parameters
+pE      = P;
+np      = length(pE);
+pE(ip)  = 0;
+pC      = sparse(ip,ip,32,np,np);
+M(1).pE = pE;
+M(1).pC = pC;
+ 
+% generate data
+%==========================================================================
+M(1).E.nE  = 32;                                % DEM-steps
+M(1).E.nN  = 32;                                % DEM-steps
+N          = 32;                                % length of data sequence
+U          = exp(-((1:N) - 12).^2/(2.^2));      % this is the Gaussian cause
+DEM        = spm_DEM_generate(M,U,{P});
+DEM.M(2).V = exp(0);
+ 
+% plot data
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Graphics'); spm_DEM_qU(DEM.pU);
+ 
+% Initialization of SCKS
+%==========================================================================
+SCK         = DEM;
+SCK.M(1).cb = [];               % option to specify constrain on parameters values [min max]
+SCK.M(1).ip = ip;               % indices of model parameters to be estimated
+SCK.M(2).v  = 0;                % input initial condition
+SCK.M(2).V  = 10;               % input noise precision (fixed)
+SCK.M(1).xP = eye(2)*1e-1^2;    % state error covariance matrix
+SCK.M(1).uP = eye(1)*1e-1^2;    % input error covariance matrix
+SCK.M(1).wP = sparse(ip,ip,10e-10^2,np,np); % parameter error covariance matrix
+SCK.M(1).pC = sparse(ip,ip,1e-3,np,np);  % covariance matrix of parameter noise
+SCK.M(1).Q  = {speye(M(1).l,M(1).l)};    % if Q is specified then algorithm performs
+                                         % estimation of measurement noise covariance 
+SCK.M(1).Qf      = 'auto';      % form of estimation of measurement noise covariance 
+                                % (after online VB estimation); options:
+                                % [auto,all,min,mean]
+SCK.M(1).E.nN    = 50;          % max number of iteration of SCKF-SCK algorithm
+SCK.M(1).E.Itol  = 1e-3;        % convergence tolerance value for SCKF_SCK algorithm
+SCK.M(1).E.RM    = [5e2 1e6];   % scaling parameter for Robbins-Monro approximation of 
+                                % parameter noise covariance [scaling
+                                % parameter, max-limit]
+SCK.M(1).VB.N    = 10;          % max number of VB iteration during one SCKF-SCK run
+SCK.M(1).VB.Itol = 1e-6;        % convergence tolerance value for VB algorithm
+SCK.M(1).VB.l    = 1 - exp(-2); % scaling parameter for VB algorithm, 
+                                % controls dynamics
+%--------------------------------------------------------------------------
+ 
+ 
+% Inversion:
+%==========================================================================
+DEM     = spm_DEM(DEM);
+LAP     = spm_LAP(DEM);
+SCK     = spm_SCK(SCK);
+ 
+% Show estimates of states
+%--------------------------------------------------------------------------
+spm_figure('GetWin','spm_DEM'); spm_DEM_qU(DEM.qU,LAP.pU)
+spm_figure('GetWin','spm_SCK'); spm_DEM_qU(SCK.qU,LAP.pU)
+spm_figure('GetWin','spm_LAP'); spm_DEM_qU(LAP.qU,LAP.pU)
+ 
+ 
+% and parameters
+%--------------------------------------------------------------------------
+qP1    = spm_vec(LAP.pP.P);
+qP2    = spm_vec(LAP.qP.P);
+qP3    = spm_vec(DEM.qP.P);
+qP4    = spm_vec(SCK.qP.P);
+qP1    = qP1(ip);
+qP2    = qP2(ip);
+qP3    = qP3(ip);
+qP4    = qP4(ip);
+ 
+subplot(2,2,4)
+bar([qP1 qP2 qP3 qP4])
+axis square
+legend('True','LAP','DEM','SCK')
+title('parameters','FontSize',16)
+
+
diff --git a/toolbox/DEM/DEM_demo_dendrite.m b/toolbox/DEM/DEM_demo_dendrite.m
new file mode 100644
index 0000000..8a68f8a
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_dendrite.m
@@ -0,0 +1,351 @@
+% Free-energy and the single neuron:
+%__________________________________________________________________________
+% This demo illustrates the use of Lotka-Volterra form SHCs (Stable
+% heteroclinic channels) to prescribe active sampling (inference). In this
+% example, we assume that neurons self-organise to minimise a free energy
+% bound on the informational surprise in the pre-synaptic inputs that are 
+% sampled.  We model this as a selective pruning of post-synaptic spines 
+% that are expressed on the dendritic tree.  This pruning occurs when the 
+% (optimised) post-synaptic gain falls to small values.  Crucially, post-
+% synaptic gain (encoding the precision of the neuron’s prediction errors 
+% about its pre-synaptic inputs) is itself optimised with respect to free-
+% energy.  Furthermore, the pruning itself suppresses free-energy as the 
+% neuron selects post-synaptic specialisations that conform to its 
+% expectations.  This provide a principled account of how neurons organise 
+% and selectively sample the myriad of potential pre-synaptic inputs they 
+% are exposed to, but it also connects elemental neuronal (dendritic) 
+% processing to generic schemes in statistics and machine learning, such 
+% as Bayesian model selection and automatic relevance determination.  
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_dendrite.m 4170 2011-01-24 18:37:42Z karl $
+ 
+% preliminaries
+%==========================================================================
+DemoMode = 1;
+spm_figure('Getwin','DEM');
+ 
+% generative model - Stable heteroclinic channel of pre-synaptic neurons
+%==========================================================================
+fx    = inline('spm_lotka_volterra(x,exp(v),P)','x','v','P');
+gx    = inline('x(P.w)','x','v','P');
+ 
+% connection weights: NB synapses are changed by switching Y (not P.w)
+%--------------------------------------------------------------------------
+np    = 5;                            % number of pre-synaptic neurons
+ns    = 4;                            % number of dendritic segments
+ny    = ns*np;                        % number of synapses
+ 
+w     = kron([1:np],ones(1,ns));      % prior contacts
+g     = rem(randperm(ns*np),np) + 1;  % initial contacts
+ 
+P.w   = w';                           % synaptic weights 
+Q.w   = g';                           % synaptic weights
+ 
+ 
+% level 1
+%--------------------------------------------------------------------------
+M(1).x  = sparse(1,1,np,np,1) - np;
+M(1).f  = fx;
+M(1).g  = gx;
+M(1).pE = P;
+M(1).Q  = spm_Ce(ones(1,ny));          % error components
+M(1).hE = -2;                          % log-precision prior mean
+M(1).hC = 2;                           % log-precision prior covariance
+M(1).W  = exp(16);                     % error precision
+ 
+% level 2
+%--------------------------------------------------------------------------
+M(2).v  = 0;
+M(2).V  = exp(16);
+ 
+ 
+% generate presynaptic inputs
+%--------------------------------------------------------------------------
+M(1).E.nE = 4;
+M(1).E.s  = 1;
+ 
+N       = 128;
+U       = log(sparse(1,N) + 1/2);
+DEM     = spm_DEM_generate(M,U,Q,{2 16},{8});
+DEM.U   = U;
+ 
+spm_DEM_qU(DEM.pU)
+ 
+ 
+% Synaptic pruning
+%==========================================================================
+if DemoMode
+    load DEM_Dendrites
+else
+    
+    % Synaptic pruning
+    %======================================================================
+    for i = 1:64
+        
+        % integrate
+        %------------------------------------------------------------------
+        DEM   = spm_DEM(DEM);
+        
+        % Get priors and posteriors
+        % -----------------------------------------------------------------
+        qE    = DEM.qH.h{1};
+        qC    = DEM.qH.C;
+        pE    = DEM.M(1).hE;
+        pC    = DEM.M(1).hC;
+                
+        % record
+        %------------------------------------------------------------------
+        G(:,i) = Q.w;
+        F(1,i) = DEM.F(end);
+        H(:,i) = qE;
+        
+        % model search over new prior without the i-th parameter
+        % -----------------------------------------------------------------
+        for k = 1:ny
+            R     = speye(ny,ny) - sparse(k,k,1,ny,ny);
+            rE    = pE;
+            rE(k) = 4;
+            rC    = R*pC*R;
+            Z(k)  = spm_log_evidence(qE,qC,pE,pC,rE,pC);
+        end
+        
+        % change synaptic locations based on optimized precision
+        %------------------------------------------------------------------
+        j     = Z < 0;
+        
+        % and generate new presynaptic inputs
+        %------------------------------------------------------------------
+        r     = rem(randperm(ns*np),np) + 1;
+        g(j)  = r(j);
+        Q.w   = g';
+        GEN   = spm_DEM_generate(DEM.M,U,Q,{2 16},{8});
+        DEM.Y = GEN.Y;
+        
+        
+        % report
+        %==================================================================
+        spm_figure('Getwin','Figure 1');
+        
+        D     = sparse(1:ny,G(:,i),exp(H(:,i)),ny,np);
+        subplot(2,1,1)
+        imagesc(-D)
+        axis image, axis off
+        title('synaptic connections','FontSize',16)
+        
+        subplot(2,2,3)
+        plot(F - F(1))
+        xlabel('iterations')
+        axis square, box off
+        title('Free-energy','FontSize',16)
+        
+        subplot(2,2,4)
+        imagesc(-exp(H))
+        xlabel('iterations')
+        ylabel('synaptic strength')
+        axis square
+        title('Precision','FontSize',16)
+        drawnow
+        
+    end
+    
+    % save 
+    %----------------------------------------------------------------------
+    save DEM_Dendrites
+ 
+end
+ 
+ 
+% Illustrate selection of synaptic connections
+%==========================================================================
+spm_figure('Getwin','Figure 1');
+ 
+subplot(2,2,1)
+imagesc(-exp(H))
+xlabel('cycles')
+ylabel('synaptic strength')
+axis square
+title('Precision','FontSize',16)
+ 
+subplot(2,2,2)
+plot(F - F(1))
+xlabel('cycles')
+axis square, box off
+title('Free-energy','FontSize',16)
+spm_axis tight
+ 
+% show connectivity at selected time points
+%--------------------------------------------------------------------------
+T     = [1 2 32 48];
+for i = 1:length(T)
+    subplot(2,4,4 + i)
+    imagesc(-sparse(1:ny,G(:,T(i)),exp(H(:,T(i))),ny,np))
+    axis image
+    title(sprintf('Cycle %i',T(i)),'FontSize',16)
+end
+    
+ 
+% Illustrate synaptic selection function (cf post hoc model selection)
+%==========================================================================
+spm_figure('Getwin','Figure 2');
+ 
+% Get priors and posteriors
+% -------------------------------------------------------------------------
+qE    = DEM.qH.h{1};
+qC    = DEM.qH.C;
+pE    = DEM.M(1).hE;
+pC    = DEM.M(1).hC;
+ 
+% model search over new prior without the i-th parameter
+% -------------------------------------------------------------------------
+QE    = linspace(-2,4,32);
+for k = 1:length(QE)
+    R     = speye(ny,ny) - sparse(1,1,1,ny,ny);
+    rE    = pE;
+    rE(1) = 4;
+    qE(1) = QE(k);
+    rC    = R*pC*R;
+    L     = spm_log_evidence(qE,qC,pE,pC,rE,pC);
+    p(k)  = exp(L)/(exp(L) + 1);
+end
+ 
+spm_figure('Getwin','Figure 2');
+subplot(2,1,1)
+plot(QE,p,QE,QE*0 + 1/2,':');
+ylabel('conditional probability')
+xlabel('synaptic strength (log-precision)')
+title('Synaptic selection function','FontSize',16)
+axis square
+ 
+ 
+% make move of synaptic re-organization
+%--------------------------------------------------------------------------
+subplot(2,1,2),cla
+for i = 1:size(G,2)
+    imagesc(-sparse(1:ny,G(:,i),exp(H(:,i)),ny,np))
+    axis image
+    MM(i) = getframe(gca);
+end
+ 
+title('synaptic connections','FontSize',16)
+xlabel('(left) click image for movie','FontSize',16)
+ 
+% set ButtonDownFcn
+%--------------------------------------------------------------------------
+h = findobj(gca,'type','image');
+set(h(1),'Userdata',{MM,16})
+set(h(1),'ButtonDownFcn','spm_DEM_ButtonDownFcn')
+ 
+
+% Illustrate sequence specificity
+%==========================================================================
+SIM   = DEM;
+U     = -((1:N) - N/2).^2/(2*(N/8)^2);
+SIM.M(1).E.nE =  1;
+SIM.M(1).W    =  8;
+SIM.M(1).hE   =  2;
+SIM.M(2).v    = -4;
+SIM.M(2).V    =  8;
+n     = 4;
+t     = 1:N;
+j     = 1:np;
+Q     = P;
+for i = 1:n
+    
+    
+    % Integrate
+    %----------------------------------------------------------------------
+    spm_figure('Getwin','DEM');
+    
+    SIM   = spm_DEM_generate(SIM.M,U,Q,{2 16},{8});
+    SIM   = spm_DEM(SIM);
+    
+    % plot
+    %----------------------------------------------------------------------
+    spm_figure('Getwin','Figure 3');
+    
+    subplot(n,2,(i - 1)*2 + 1)
+    imagesc(SIM.Y)
+    xlabel('time','FontSize',12)
+    ylabel('synapse','FontSize',12)
+    title('presynaptic input','FontSize',16)
+    
+    subplot(n,2,(i - 1)*2 + 2)
+   
+    m    = SIM.qU.v{2};
+    ci   = 1.64* sqrt(spm_vec(SIM.qU.C)');
+    cu   = exp(m + ci);
+    cl   = exp(m - ci);
+    m    = exp(m);
+    fill([t fliplr(t)],[full(cu) fliplr(full(cl))],...
+         [1 1 1]*.8,'EdgeColor',[1 1 1]*.8),hold on
+    plot(t,m), hold off
+    xlabel('time','FontSize',12)
+    ylabel('activity','FontSize',12)
+    title('postsynaptic response','FontSize',16)
+    axis([1 N 0 2])
+    
+    % change presynaptic firing order
+    %----------------------------------------------------------------------
+    j     = randperm(np);
+    Q.w   = kron(j,ones(1,ns));
+    
+end
+ 
+% show response
+%==========================================================================
+spm_figure('Getwin','DEM');
+spm_DEM_qU(SIM.qU,SIM.pU)
+ 
+ 
+ 
+% Changing the speed (u) using IN and OUT sequences
+%==========================================================================
+ 
+SIM = DEM;
+ 
+SIM.M(1).E.nE =  1;
+SIM.M(1).W    =  8;
+SIM.M(1).hE   =  2;
+SIM.M(2).v    = -4;
+SIM.M(2).V    =  8;
+ 
+IN    = 1:np;
+OUT   = fliplr(IN);
+u     = [1 2 4 6];
+U     = -((1:N) - N/2).^2/(2*(N/8)^2);
+ 
+for i = 1:length(u)
+    
+    % Speed
+    %----------------------------------------------------------------------
+    U      = log(zeros(1,round(64/u(i))) + u(i));
+        
+    % Integrate IN
+    %---------------------------------------------------------------------- 
+    Q.w    = kron(IN,ones(1,ns));
+    SIM    = spm_DEM_generate(SIM.M,U,Q,{2 16},{8});
+    SIM    = spm_DEM(SIM);
+    V(1,i) = mean(exp(SIM.qU.v{2}));
+    
+    % Integrate OUT
+    %---------------------------------------------------------------------- 
+    Q.w    = kron(OUT,ones(1,ns));
+    SIM    = spm_DEM_generate(SIM.M,U,Q,{2 16},{8});
+    SIM    = spm_DEM(SIM);
+    V(2,i) = mean(exp(SIM.qU.v{2}));
+    
+end
+
+% plot direction-selectivity
+%--------------------------------------------------------------------------
+spm_figure('Getwin','Figure 4');
+
+subplot(2,1,1)
+plot(u,V(1,:),'r',u,V(1,:),'or',u,V(2,:),'b',u,V(2,:),'ob')
+xlabel('input velocity','FontSize',12)
+ylabel('mean response','FontSize',12)
+title('velocity-dependent responses','FontSize',16)
+axis square, box off
diff --git a/toolbox/DEM/DEM_demo_doublewell_LAP.m b/toolbox/DEM/DEM_demo_doublewell_LAP.m
new file mode 100644
index 0000000..16fe41e
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_doublewell_LAP.m
@@ -0,0 +1,68 @@
+% The double-well revisited:
+%__________________________________________________________________________
+% This demonstration compares generalised filtering and a state-of-the-art 
+% Bayesian smoother (SCKS) in the context of a double-well system. Here the
+% Cubature filtering outperforms generalised schemes that are confounded by 
+% the failure of the Laplace assumption. Note that generalised filtering
+% and DEM give the same conditional estimates of states because there are 
+% no free parameters or hyperparameters and the mean-field assumption 
+% implcit in DEM is irrelevant.
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_doublewell_LAP.m 4146 2010-12-23 21:01:39Z karl $
+ 
+ 
+% Create model and data
+%==========================================================================
+M      = spm_DEM_M('double well');
+ 
+% generate data (output)
+%--------------------------------------------------------------------------
+N         = 120;
+U         = 8*sin(pi*[1:N]/16);     
+M(1).E.s  = 2;
+M(1).E.n  = 3;
+M(1).E.nD = 4;
+M(1).V    = exp(6);
+M(1).W    = exp(6);
+DEM       = spm_DEM_generate(M,U);
+ 
+% show data (output)
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Simulated'); spm_DEM_qU(DEM.pU);
+ 
+% Initialization of SCKS
+%==========================================================================
+SCK         = DEM;
+SCK.M(1).cb = [];          % option to specify constrain on parameters values [min max]
+SCK.M(1).ip = [];          % indices of model parameters to be estimated
+SCK.M(2).v  = 0;           % input initial condition
+SCK.M(2).V  = 0.5;         % input noise precision (fixed)
+SCK.M(1).xP = eye(1)*1e-1^2;          % state error covariance matrix
+SCK.M(1).uP = eye(1)*1e0^2;           % input error covariance matrix
+SCK.M(1).Q  = {speye(M(1).l,M(1).l)}; % if Q is specified then algorithm performs
+                                      % estimation of measurement noise covariance
+SCK.M(1).Qf      = 'auto'; % form of estimation of measurement noise covariance 
+                           % (after online VB estimation); options: [auto,all,min,mean]        
+SCK.M(1).E.nN    = 50;     % max number of iterations of SCKF-SCK algorithm
+SCK.M(1).E.Itol  = 1e-3;   % convergence tolerance value for SCKF_SCK algorithm
+SCK.M(1).VB.N    = 2;      % max number of VB iteration during one SCKF-SCK run
+SCK.M(1).VB.Itol = 1e-2;   % convergence tolerance value for VB algorithm
+SCK.M(1).VB.l    = 1 - exp(-1.5);     % scaling parameter for VB algorithm, 
+                                      % controls dynamics
+%--------------------------------------------------------------------------
+ 
+ 
+% Deconvolution
+%==========================================================================
+SCK     = spm_SCK(SCK);
+DEM     = spm_DEM(DEM);
+LAP     = spm_LAP(DEM);
+  
+% Shoe estimates of states
+%--------------------------------------------------------------------------
+spm_figure('GetWin','spm_DEM'); spm_DEM_qU(DEM.qU,LAP.pU)
+spm_figure('GetWin','spm_LAP'); spm_DEM_qU(LAP.qU,LAP.pU)
+spm_figure('GetWin','spm_SCK'); spm_DEM_qU(SCK.qU,LAP.pU)
diff --git a/toolbox/DEM/DEM_demo_factor_analysis.m b/toolbox/DEM/DEM_demo_factor_analysis.m
index 4150a5c..b84cae3 100644
--- a/toolbox/DEM/DEM_demo_factor_analysis.m
+++ b/toolbox/DEM/DEM_demo_factor_analysis.m
@@ -21,6 +21,7 @@
  
 % DEM estimation
 %==========================================================================
+DEM.M(1).E.nE = 16;
 DEM   = spm_DEM(DEM);
  
 % compare real and estimated factor and causes
diff --git a/toolbox/DEM/DEM_demo_hdm_LAP.m b/toolbox/DEM/DEM_demo_hdm_LAP.m
index 06e5709..b052a2c 100644
--- a/toolbox/DEM/DEM_demo_hdm_LAP.m
+++ b/toolbox/DEM/DEM_demo_hdm_LAP.m
@@ -11,12 +11,13 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_demo_hdm_LAP.m 3901 2010-05-27 16:14:36Z karl $
+% $Id: DEM_demo_hdm_LAP.m 4146 2010-12-23 21:01:39Z karl $
 
 % set-up
 %--------------------------------------------------------------------------
 clear
 DEMO = 1;
+global dt
 
 if DEMO
     
@@ -30,19 +31,16 @@
     %======================================================================
     dt      = Y.dt;
     T       = 1:256;
-
+    
     % level 1
     %----------------------------------------------------------------------
-    [pE pC] = spm_hdm_priors(3,1);
-    P       = pE;
-    P(7:9)  = 1/8;
+    [pE pC] = spm_hdm_priors(3,2);
     M(1).x  = [0 0 0 0]';
     M(1).g  = 'spm_gx_hdm';
     M(1).f  = 'spm_fx_hdm';
     M(1).pE = pE;
     M(1).pC = pC;
-    M(1).gE = 7;
-    M(1).gC = 1/2;
+    M(1).W  = exp(8);
     M(1).V  = exp(4);
 
     % level 2
@@ -53,19 +51,24 @@
     M(1).E.linear = 1;
     M(1).E.n  = 4;
     M(1).E.nD = 1;
-    M(1).E.nN = 8;
-    M(1).E.s  = 1/4;
+    M(1).E.nE = 16;
+    M(1).E.nN = 16;
+    M(1).E.s  = 1/2;
+
+    % Decimate U.u from micro-time
+    % ---------------------------------------------------------------------
+    Dy  = spm_dctmtx(size(Y.y,1),size(Y.y,1));
+    Du  = spm_dctmtx(size(U.u,1),size(Y.y,1));
+    Dy  = Dy*sqrt(size(Y.y,1)/size(U.u,1));
+    u   = Dy*(Du'*U.u);
+    U   = spm_detrend(u(T,:))';
 
-    % get causes (i.e. experimental inputs)
-    %----------------------------------------------------------------------
-    t     = fix(linspace(1,length(U.u),360));
-    U     = U.u(t(T),:)';
 
     % Emprical data
     %----------------------------------------------------------------------
     DEM.M = M;
     DEM.U = U;
-    DEM.Y = 1/2 + Y.y(T,:)'/8;
+    DEM.Y = Y.y(T,:)'/4;
 
     % DEM estimation
     %======================================================================
diff --git a/toolbox/DEM/DEM_demo_hdm_SCK.m b/toolbox/DEM/DEM_demo_hdm_SCK.m
new file mode 100644
index 0000000..0c87193
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_hdm_SCK.m
@@ -0,0 +1,146 @@
+% Demo for Hemodynamic deconvolution: Cross-validation of Laplace scheme
+%__________________________________________________________________________
+% This demonstration compares generalised filtering and SCKS in the context 
+% of a nonlinear convolution model using synthetic data. Here, we look at
+% estimating three of the hemodynamic parameters. This is a particularly
+% difficult (almost impossible) problem, given their distance from the data
+% and the conditional dependencies with the hidden states. Furthermore, 
+% this is an unrealistic simulation, because we assume the data are almost 
+% noiseless. The key thing to focus on is the comparative performance in
+% recovering the hidden states and causes.
+%__________________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_hdm_SCK.m 4146 2010-12-23 21:01:39Z karl $
+ 
+ 
+% create model and generate data
+%==========================================================================
+ 
+% set options
+%--------------------------------------------------------------------------
+clear M
+M(1).E.linear = 0;                          % linear model
+M(1).E.s      = 1;                          % smoothness
+ 
+% level 1
+%--------------------------------------------------------------------------
+ip      = [1 2 5]';
+pE      = spm_hdm_priors(1,5);              % parameters
+np      = length(pE);
+pC      = sparse(ip,ip,exp(-3),np,np);
+pE(6)   = 0.02;
+pE(7)   = 0.5;
+ 
+M(1).n  = 4;
+M(1).f  = 'spm_fx_hdm';
+M(1).g  = 'spm_gx_hdm';
+M(1).pE = pE;                               % prior expectation
+M(1).pC = pC;                               % prior expectation
+M(1).xP = exp(4);                           % prior expectation
+M(1).V  = exp(8);                           % error precision
+M(1).W  = exp(12);                          % error precision
+M(1).ip = ip;
+ 
+% level 2
+%--------------------------------------------------------------------------
+M(2).l  = 1;                                % inputs
+M(2).V  = exp(0);                           % with shrinkage priors
+ 
+% true parameters
+%--------------------------------------------------------------------------
+P       = M(1).pE;
+P(ip)   = P(ip) - P(ip)/8;
+ 
+ 
+% generate data
+%==========================================================================
+N         = 64;                             % length of data sequence
+U         = exp(-([1:11] - 6).^2/(2.^2))/8; % this is the Gaussian cause
+input     = zeros(1,N);
+input([5 10 20 34 43 50]) = [1 0.8 1 0.2 .9 0.4];
+U         = conv(U,input);
+U         = U(1:N);
+DEM       = spm_DEM_generate(M,U,{P},{8,16},{16});
+ 
+spm_figure('GetWin','Simulated');
+spm_DEM_qU(DEM.pU)
+ 
+% Initialise SCK
+%==========================================================================
+DEM.M(1).E.nN = 32;
+DEM.M(1).E.nN = 32;
+SCK           = DEM;
+ 
+% option to specify constrains on parameters (example, but not used here)
+%-------------------------------------------------------------------------
+cb(1,:) = [0.5,0.8];              % low and high bound
+cb(2,:) = [0.3,0.5];              % low and high bound
+cb(3,:) = [0.7,1.3];              % low and high bound
+cb(4,:) = [0.29,0.35];            % low and high bound
+cb(5,:) = [0.32,.4];              % low and high bound
+cb(6,:) = [0.01,0.04];            % low and high bound
+cb(7,:) = [0.5,0.6];              % low and high bound
+ 
+SCK.M(1).ip = ip;  % indices of model parameters to be estimated
+SCK.M(1).cb = [];  % option to specify constrain on parameters values [min max]
+SCK.M(2).v  = 0;   % input initial condition
+SCK.M(2).V  = 5;   % input noise precision (fixed)
+SCK.M(1).xP = eye(4)*1e-1^2;      % state error covariance matrix
+SCK.M(1).uP = eye(1)*1e-2^2;      % input error covariance matrix
+SCK.M(1).wP = eye(np)*1e-10^2;    % parameter error covariance matrix
+SCK.M(1).pC = eye(np)*1e-10^4;    % parameter error covariance matrix
+SCK.M(1).f  = 'spm_fx_hdm_sck';   % state equations rewritten for matrix operations
+SCK.M(1).g  = 'spm_gx_hdm_sck';   % observation equations rewritten for matrix operations
+SCK.M(1).Q  = {}; 
+SCK.M(1).E.nN    = 32;            % max number of iteration of SCKF-SCK algorithm
+SCK.M(1).E.nD    = 4;             % time-steps
+SCK.M(1).E.Itol  = 1e-3;          % convergence tolerance value for SCKF_SCK algorithm
+SCK.M(1).E.RM    = [5e3 1e6];     % scaling parameter for Robbins-Monro approximation 
+
+
+
+% % Free-energy landscape 
+% %========================================================================
+% spm_dem_F(DEM,ip(1));
+
+
+% Inversion 
+%==========================================================================
+SCK     = spm_SCK(SCK);
+LAP     = spm_LAP(DEM);
+DEM     = spm_DEM(DEM);
+ 
+ 
+% Show estimates of states
+%--------------------------------------------------------------------------
+spm_figure('GetWin','spm_DEM'); spm_DEM_qU(DEM.qU,LAP.pU)
+spm_figure('GetWin','spm_LAP'); spm_DEM_qU(LAP.qU,LAP.pU)
+spm_figure('GetWin','spm_SCK'); spm_DEM_qU(SCK.qU,LAP.pU)
+ 
+ 
+% and parameters
+%--------------------------------------------------------------------------
+pP1    = spm_vec(LAP.M(1).pE);
+qP1    = spm_vec(LAP.pP.P);
+qP2    = spm_vec(LAP.qP.P);
+qP3    = spm_vec(DEM.qP.P);
+qP4    = spm_vec(SCK.qP.P);
+
+ 
+spm_figure('GetWin','spm_DEM'); subplot(2,2,4)
+bar([pP1 qP1 qP2 qP3 qP4])
+axis square
+legend('Prior','True','LAP','DEM','SCK')
+title('parameters (minus true)','FontSize',16)
+
+
+spm_figure('GetWin','spm_SCK'); subplot(2,1,2)
+t  = 1:N;
+plot(t,LAP.pU.v{2},t,LAP.qU.v{2},t,DEM.qU.v{2},t,SCK.qU.v{2})
+box off
+legend('True','LAP','DEM','SCK')
+title('Hidden causes','FontSize',16)
+xlabel('time','FontSize',12)
+ 
diff --git a/toolbox/DEM/DEM_demo_lorenz_LAP.m b/toolbox/DEM/DEM_demo_lorenz_LAP.m
new file mode 100644
index 0000000..7b048db
--- /dev/null
+++ b/toolbox/DEM/DEM_demo_lorenz_LAP.m
@@ -0,0 +1,128 @@
+% Dual estimation of the Lorenz system: Cross-validation of Laplace schemes
+%__________________________________________________________________________
+% Inversion of the Lorenz attractor with DEM, LAP and SCKS schemes: This
+% demo tackles the difficult problem of deconvolving (chaotic) hidden states
+% from a single response variable, while estimating the parameters of the
+% underlying equations of motion. It calls generalised filtering, DEM and
+% a state-of-the-art Bayesian smoother (SCKS).  This example is chosen to
+% show that it is, in principle, possible to perform dual estimation in the
+% context of chaotic dynamics (although small variations in this problem
+% will cause the schemes to fail due it its inherently nonlinear nature and
+% non-identifiability); however, the results are imperfect.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: DEM_demo_lorenz_LAP.m 4146 2010-12-23 21:01:39Z karl $
+ 
+ 
+% get model
+%==========================================================================
+M         = spm_DEM_M('Lorenz');
+ 
+% create data
+%==========================================================================
+ 
+% create innovations & add causes
+%--------------------------------------------------------------------------
+N         = 128;
+U         = sparse(1,N);
+ 
+% specify precisions
+%--------------------------------------------------------------------------
+M(1).E.nN = 32;
+M(1).E.nE = 32;
+M(1).E.nD = 2;
+M(1).E.n  = 4;
+M(1).E.d  = 2;
+M(1).E.s  = 1/16;
+M(1).V    = exp(1);
+M(1).W    = exp(16);
+DEM       = spm_DEM_generate(M,U);
+ 
+% show data
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Graphics'); spm_DEM_qU(DEM.pU);
+ 
+% initialization of parameters (True values: pE =[18;-4;46.92])
+%--------------------------------------------------------------------------
+pE  = M(1).pE + [0 1 -1]';        % Modified prior parameters
+pC  = diag(exp(-[32 2 2]));       % prior covariance
+DEM.M(1).pE = pE;
+DEM.M(1).pC = pC;
+ 
+ 
+% Initialization of SCKS
+%--------------------------------------------------------------------------
+np          = length(pE);
+ip          = 1:np; 
+SCK         = DEM;
+SCK.M(1).x  = [rand(2,1)*16;25];  % partly random initialization of the states
+SCK.M(1).x  = [0.9;0.8;30];       % True state values: 
+          
+ 
+% option to specify constrains on parameters (example, not used further)
+%--------------------------------------------------------------------------
+cb = [0, 30;
+    -10, 10;
+     40, 50;];                    % [min max]
+ 
+SCK.M(1).cb = [];                 % option to specify constrain on parameters values [min max]
+SCK.M(1).ip = ip;                 % indices of model parameters to be estimated
+SCK.M(2).v  = [];                 % input initial condition
+SCK.M(2).V  = [];                 % input noise precision (fixed)
+SCK.M(1).xP = eye(3)*1e-1^2;      % state error covariance matrix
+SCK.M(1).uP = eye(1)*1e-10^2;     % input error covariance matrix
+SCK.M(1).wP = pC*pC;              % parameter error covariance matrix
+SCK.M(1).Q  = {speye(M(1).l,M(1).l)};     % if Q is specified then algorithm performs
+                                          % estimation of measurement noise covariance 
+SCK.M(1).Qf      = 'auto';        % form of estimation of measurement noise covariance 
+                                  % (after online VB estimating); options:
+                                  % [auto,all,min,mean]
+SCK.M(1).E.nN    = 32;            % max number of iteration of SCKF-SCK algorithm
+SCK.M(1).E.Itol  = 1e-3;          % convergence tolerance value for SCKF_SCK algorithm
+SCK.M(1).E.RM    = [5e3 1e6];     % scaling parameter for Robbins-Monro approximation of 
+                                  % parameter noise covariance [scaling parameter, max-limit]
+SCK.M(1).VB.N    = 3;             % max number of VB iteration during one SCKF-SCK run
+SCK.M(1).VB.Itol = 1e-2;          % convergence tolerance value for VB algorithm
+SCK.M(1).VB.l    = 1 - exp(-2);   % scaling parameter for VB algorithm, 
+                                  
+ 
+% Inversion: 
+%==========================================================================
+LAP     = spm_LAP(DEM);
+DEM     = spm_DEM(DEM);
+SCK     = spm_SCK(SCK);
+ 
+% Show estimates of states
+%--------------------------------------------------------------------------
+spm_figure('GetWin','spm_DEM'); spm_DEM_qU(DEM.qU,LAP.pU)
+spm_figure('GetWin','spm_SCK'); spm_DEM_qU(SCK.qU,LAP.pU)
+spm_figure('GetWin','spm_LAP'); spm_DEM_qU(LAP.qU,LAP.pU)
+ 
+ 
+% and parameters
+%--------------------------------------------------------------------------
+pP1    = spm_vec(LAP.pP.P);
+qP2    = spm_vec(LAP.qP.P);
+qP3    = spm_vec(DEM.qP.P);
+qP4    = spm_vec(SCK.qP.P);
+dP1    = pE(ip)  - pP1(ip);
+dP2    = qP2(ip) - pP1(ip);
+dP3    = qP3(ip) - pP1(ip);
+dP4    = qP4(ip) - pP1(ip);
+ 
+ 
+subplot(2,2,3)
+bar([pP1 qP2 qP3 qP4])
+axis square
+legend('True','LAP','DEM','SCK')
+title('parameters (minus true)','FontSize',16)
+ 
+subplot(2,2,4)
+bar([dP1 dP2 dP3 dP4])
+axis square
+legend('Prior','LAP','DEM','SCK')
+title('difference from true value','FontSize',16)
+
+
diff --git a/toolbox/DEM/DEM_demo_song_inference.m b/toolbox/DEM/DEM_demo_song_inference.m
index aa01e55..429a395 100644
--- a/toolbox/DEM/DEM_demo_song_inference.m
+++ b/toolbox/DEM/DEM_demo_song_inference.m
@@ -1,22 +1,25 @@
-% Demo for a bird songs: In this example, we show that DEM can not only
-% estimate the hidden states of an autonomous system but can also
-% deconvolve dynamics changes in its control parameters.  We illustrate
-% this using a slow Lorentz attractor to drive a fast one; both  showing
-% deterministic chaos.  We endow the simulations with a little ethological
-% validity by using the states of the fast Lorentz attractor as control
-% variables in the syrinx of a song bird (usually these would control a van
-% der Pol oscillator model). We will look at the true and inferred songs
-% with and without the last part missing.  When sonograms are displayed the
-% song can be played by a mouse click on the image. The final plots show
-% simulated event related potential to show that there is a marked
-% responses (prediction error) of the system when an expected ‘syllable’ is
-% omitted. This demonstrates the implicit sequence-decoding of input
-% streams, using uncontrollable state-space models
+% Perceptual categorisation of bird songs: The generative model of 
+% birdsong used in this simulation comprises a Lorenz attractor with two 
+% control parameters (or hidden causes), which, in turn, delivers two 
+% control parameters to a synthetic syrinx to produce ‘chirps’ that are 
+% modulated in amplitude and frequency.  The chirps were then presented 
+% as a stimulus to a synthetic bird to see if it could infer the 
+% underlying causal states and thereby categorise the song. This entails 
+% minimising free energy by changing the internal representation of the 
+% control parameters. Each simulated song comprises a series of chirps 
+% whose frequency and number fall progressively from song a to song c, 
+% as a causal state (known as the Raleigh number) is decreased.  The 
+% simulations show that the causes are identified after about 600 
+% milliseconds with high conditional precision (90% confidence intervals 
+% are shown in grey). These simulations illustrate the nature of 
+% perceptual categorisation under generalised predictive coding: Here, 
+% recognition corresponds to mapping from a continuously changing and 
+% chaotic sensory input to a fixed point in perceptual space.
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_demo_song_inference.m 3655 2009-12-23 20:15:34Z karl $
+% $Id: DEM_demo_song_inference.m 4170 2011-01-24 18:37:42Z karl $
  
  
 % Hierarchical non-linear generative model (dynamic & chaotic)
@@ -37,9 +40,9 @@
  
 % level 1
 %--------------------------------------------------------------------------
-% P(1): Prandtl number
-% P(2): 8/3
-% P(3): Rayleigh number
+% Prandtl number  = 10
+% P(2)            = v(2)
+% Rayleigh number = v(1) - 4: 
  
 x       = [0.9; 0.8; 2];
 M(1).f  = ' [-10 10 0; (v(1) - 4 - x(3)) -1 0; x(2) 0 -v(2)]*x/16;';
diff --git a/toolbox/DEM/DEM_demo_song_omission.m b/toolbox/DEM/DEM_demo_song_omission.m
index cbe766c..469d5f8 100644
--- a/toolbox/DEM/DEM_demo_song_omission.m
+++ b/toolbox/DEM/DEM_demo_song_omission.m
@@ -1,6 +1,6 @@
-% Demo for a bird songs: In this example, we show that DEM can not only
+% Demo for bird songs: In this example, we show that DEM can not only
 % estimate the hidden states of an autonomous system but can also
-% deconvolve dynamics changes in its control parameters.  We illustrate
+% deconvolve dynamic changes in its control parameters.  We illustrate
 % this using a slow Lorentz attractor to drive a fast one; both  showing
 % deterministic chaos.  We endow the simulations with a little ethological
 % validity by using the states of the fast Lorentz attractor as control
@@ -16,7 +16,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: DEM_demo_song_omission.m 3054 2009-04-07 19:22:49Z karl $
+% $Id: DEM_demo_song_omission.m 4170 2011-01-24 18:37:42Z karl $
  
  
 % Hierarchical non-linear generative model (dynamic & chaotic)
diff --git a/toolbox/DEM/DEM_lorenz.mat b/toolbox/DEM/DEM_lorenz.mat
index da188a4..bc70234 100644
Binary files a/toolbox/DEM/DEM_lorenz.mat and b/toolbox/DEM/DEM_lorenz.mat differ
diff --git a/toolbox/DEM/DFP_demo_double_well.m b/toolbox/DEM/DFP_demo_double_well.m
index 2596630..fbd2c6f 100644
--- a/toolbox/DEM/DFP_demo_double_well.m
+++ b/toolbox/DEM/DFP_demo_double_well.m
@@ -28,7 +28,6 @@
  
 % get nonlinear state-space model
 %==========================================================================
-if ~strcmp(questdlg('proceed with demo'),'Yes'), return, end
 M        = spm_DEM_M('ssm');
 M(1).E.N = 64;
  
@@ -66,8 +65,7 @@
 
 % Graphical comparison of DFP and PF
 %==========================================================================
-if ~strcmp(questdlg('compare with particle filter'),'Yes'), return, end
-figure(Fgraph)
+spm_figure('GetWin','Figure 1');
 clf
  
 % show density on state (PF)
@@ -100,8 +98,6 @@
  
 % show density on casue
 %==========================================================================
-if ~strcmp(questdlg('show density on cause'),'Yes'), return, end
- 
 for i = 1:T
     x = {DFP.QU{i}.v};
     for j = 1:length(x)
diff --git a/toolbox/DEM/LAP_HDM.mat b/toolbox/DEM/LAP_HDM.mat
index 0f28ee7..0b90c9a 100644
Binary files a/toolbox/DEM/LAP_HDM.mat and b/toolbox/DEM/LAP_HDM.mat differ
diff --git a/toolbox/DEM/cornsweet_data.mat b/toolbox/DEM/cornsweet_data.mat
new file mode 100644
index 0000000..d3ae20d
Binary files /dev/null and b/toolbox/DEM/cornsweet_data.mat differ
diff --git a/toolbox/DEM/expmall.m b/toolbox/DEM/expmall.m
new file mode 100644
index 0000000..9f01083
--- /dev/null
+++ b/toolbox/DEM/expmall.m
@@ -0,0 +1,8 @@
+function dx = expmall(J,f,t,EP)
+
+EP(EP==1) = J;
+EP(EP==2) = f;
+
+
+dx = expm(EP*t);
+
diff --git a/toolbox/DEM/pointfinger.jpg b/toolbox/DEM/pointfinger.jpg
new file mode 100644
index 0000000..7744ac1
Binary files /dev/null and b/toolbox/DEM/pointfinger.jpg differ
diff --git a/toolbox/DEM/spm_ADEM_cue_rt.m b/toolbox/DEM/spm_ADEM_cue_rt.m
new file mode 100644
index 0000000..198cb89
--- /dev/null
+++ b/toolbox/DEM/spm_ADEM_cue_rt.m
@@ -0,0 +1,79 @@
+function [on,rt,ac] = spm_ADEM_cue_rt(DEM)
+% returns reaction times and accuracy for ADEM_cued_response demo
+% FORMAT [on,rt,ac] = spm_ADEM_cue_rt(DEM)
+%
+% DEM - DEM structure from ADEM_cued_response.m
+%
+% on  - cue onset
+% ac  - accuracy
+% rt  - reaction time
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_ADEM_cue_rt.m 4231 2011-03-07 21:00:02Z karl $
+ 
+% distance from target and cue contrast
+%--------------------------------------------------------------------------
+n   = length(DEM.M(1).x.a);                   % number of targets
+F   = DEM.pU.v{1}((1:2) + 2,:);               % location of finger
+L   = DEM.pP.P{1};                            % location of targets
+for i = 1:n
+    D(i,:) = (F(1,:) - L(1,i)).^2;
+    D(i,:) = (F(2,:) - L(2,i)).^2 + D(i,:);   % distance from targets
+end        
+C   = DEM.pU.v{1}((1:n) + 4,:);               % contrast of targets
+ 
+r   = 1/32;                                   % radius of proximity
+c   = diff(C > 1,1,2) > 0;                    % target onset
+on  = {};                                     % cue onset
+ac  = {};                                     % accuracy
+rt  = {};                                     % reaction time
+ 
+% get performance
+%--------------------------------------------------------------------------
+for i = 1:n
+    
+    on{i} = find(c(i,:));
+    for j = 1:length(on{i})
+        try
+            
+            % minimum distance
+            %--------------------------------------------------------------
+            d        = D(i,(1:8) + on{i}(j))';
+            ac{i}(j) = sqrt(min(d));
+            
+            % estimated reaction time
+            %--------------------------------------------------------------
+            X        = (1:length(d))' - 1;
+            B        = pinv([X.^0 X.^1])*log(d);
+            rt{i}(j) = (log(r) - B(1))/B(2);
+            
+        catch
+            ac{i}(j) = NaN;
+            rt{i}(j) = NaN;
+        end
+    end
+end
+ 
+% sort trials (over all targets
+%--------------------------------------------------------------------------
+on = spm_vec(on); [i j] = sort(on,1,'ascend'); on = on(j);
+ac = spm_vec(ac); ac = ac(j);
+rt = spm_vec(rt); rt = rt(j);
+ 
+% remove first trial
+%--------------------------------------------------------------------------
+on(1) = [];
+rt(1) = [];
+ac(1) = [];
+ 
+% convert spatial error to accuracy
+%--------------------------------------------------------------------------
+ac    = 1./ac;
+ 
+% convert reaction time to ms
+%--------------------------------------------------------------------------
+dt    = 64/1000;
+rt    = rt*1000*dt;
+on    = on*dt;
diff --git a/toolbox/DEM/spm_ADEM_set_rt.m b/toolbox/DEM/spm_ADEM_set_rt.m
new file mode 100644
index 0000000..be353a2
--- /dev/null
+++ b/toolbox/DEM/spm_ADEM_set_rt.m
@@ -0,0 +1,83 @@
+function [on,rt,ac] = spm_ADEM_cue_rt(DEM)
+% returns reaction times and accuracy for ADEM_cued_response demo
+% FORMAT [on,rt,ac] = spm_ADEM_cue_rt(DEM)
+%
+% DEM - DEM structure from ADEM_cued_response.m
+%
+% on  - cue onset
+% ac  - accuracy
+% rt  - reaction time
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_ADEM_set_rt.m 4231 2011-03-07 21:00:02Z karl $
+ 
+% distance from target and cue contrast
+%--------------------------------------------------------------------------
+P   = DEM.pP.P{1};                            % location of targets
+L   = DEM.pU.v{1}((1:2) + 2,:);               % location of finger
+n   = length(P);                              % number of targets
+N   = length(L);                              % number of time bins
+D   = sparse(n,N);
+for j = 1:N
+    for i = 1:n
+        D(i,j) = D(i,j) + sum((L(:,j) - P(:,i)).^2);
+    end
+end
+D   = sqrt(D);                                % distance from targets
+C   = DEM.pU.v{1}((1:n) + 4,:);               % contrast of targets
+ 
+r   = 1/32;                                   % radius of proximity
+c   = diff(C > 1/2,1,2) > 0;                  % target onset
+on  = {};                                     % cue onset
+ac  = {};                                     % accuracy
+rt  = {};                                     % reaction time
+ 
+% get performance
+%--------------------------------------------------------------------------
+for i = 1:n
+    
+    on{i} = find(c(i,:));
+    for j = 1:length(on{i})
+        try
+            
+            % minimum distance
+            %--------------------------------------------------------------
+            d        = D(i,(1:8) + on{i}(j))';
+            ac{i}(j) = sqrt(min(d));
+            
+            % estimated reaction time
+            %--------------------------------------------------------------
+            X        = (1:length(d))' - 1;
+            B        = pinv([X.^0 X.^1])*log(d);
+            rt{i}(j) = (log(r) - B(1))/B(2);
+            
+        catch
+            ac{i}(j) = NaN;
+            rt{i}(j) = NaN;
+        end
+    end
+end
+ 
+% sort trials (over all targets
+%--------------------------------------------------------------------------
+on = spm_vec(on); [i j] = sort(on,1,'ascend'); on = on(j);
+ac = spm_vec(ac); ac = ac(j);
+rt = spm_vec(rt); rt = rt(j);
+ 
+% remove first trial
+%--------------------------------------------------------------------------
+on(1) = [];
+rt(1) = [];
+ac(1) = [];
+ 
+% convert spatial error to accuracy
+%--------------------------------------------------------------------------
+ac    = 1./ac;
+ 
+% convert reaction time to ms
+%--------------------------------------------------------------------------
+dt    = 64/1000;
+rt    = rt*1000*dt;
+on    = on*dt;
diff --git a/toolbox/DEM/spm_DEM_ButtonDownFcn.m b/toolbox/DEM/spm_DEM_ButtonDownFcn.m
index e853310..a724f36 100644
--- a/toolbox/DEM/spm_DEM_ButtonDownFcn.m
+++ b/toolbox/DEM/spm_DEM_ButtonDownFcn.m
@@ -1,14 +1,14 @@
 function spm_DEM_ButtonDownFcn
-% ButtonDownFcn to play a movie or sound on button press
+% ButtonDownFcn to play (or save) a movie or sound on button press
 % FORMAT spm_DEM_ButtonDownFcn
 %
-% Require gcbo to have appropriate UserData; see spm_DEM_movie and
+% Requires gcbo to have appropriate UserData; see spm_DEM_movie and
 % spm_DEM_play_song
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_DEM_ButtonDownFcn.m 3774 2010-03-11 19:20:11Z karl $
+% $Id: spm_DEM_ButtonDownFcn.m 4170 2011-01-24 18:37:42Z karl $
  
 % default
 %--------------------------------------------------------------------------
diff --git a/toolbox/DEM/spm_DEM_EEG.m b/toolbox/DEM/spm_DEM_EEG.m
index a81f4c1..1ac166c 100644
--- a/toolbox/DEM/spm_DEM_EEG.m
+++ b/toolbox/DEM/spm_DEM_EEG.m
@@ -17,7 +17,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_DEM_EEG.m 3054 2009-04-07 19:22:49Z karl $
+% $Id: spm_DEM_EEG.m 4187 2011-02-01 20:13:57Z karl $
 
 % defaults
 %--------------------------------------------------------------------------
@@ -36,7 +36,7 @@
 
 % loop over hierarchical (cortical) levels
 %--------------------------------------------------------------------------
-if graphics, cla; hold on; end
+if graphics, hold off; cla; hold on; end
 
 z     = DEM.qU.z;
 pst   = [1:size(z{1},2)]*dt*1000;
diff --git a/toolbox/DEM/spm_DEM_ERP_demo.m b/toolbox/DEM/spm_DEM_ERP_demo.m
index 9db6bf3..957d22b 100644
--- a/toolbox/DEM/spm_DEM_ERP_demo.m
+++ b/toolbox/DEM/spm_DEM_ERP_demo.m
@@ -94,7 +94,6 @@
 M(1).E.nD = 1;                % D-steps per time bin (1 for dynamic systems)
 M(1).E.nE = 1;                % E-steps per iteration
 M(1).E.nM = 1;                % M-steps per iteration
-M(1).E.nN = 1;                % number of iterations
 DEM.M     = M;
  
 % Because we want to record the response of the model over time, to each
diff --git a/toolbox/DEM/spm_DEM_M.m b/toolbox/DEM/spm_DEM_M.m
index 9289733..d5e2c79 100644
--- a/toolbox/DEM/spm_DEM_M.m
+++ b/toolbox/DEM/spm_DEM_M.m
@@ -72,6 +72,7 @@
             M(i).g  = inline('P*v','x','v','P');
             M(i).pE = pE{i};
             M(i).hE = 0;
+            M(i).hC = 16;
 
         end
 
@@ -237,8 +238,8 @@
 
         % model specification - 1st level
         %------------------------------------------------------------------
-        f      = '(16*x/(1 + x^2) - x/2 + v*2)/8';
-        g      = '(x^2)/16';
+        f      = '(16*x./(1 + x.^2) - x/2 + v*2)/8';
+        g      = '(x.^2)/16';
         M(1).x = 1;
         M(1).f = inline(f,'x','v','P');
         M(1).g = inline(g,'x','v','P');
diff --git a/toolbox/DEM/spm_DEM_MEG.m b/toolbox/DEM/spm_DEM_MEG.m
index c4a5ec4..62a3f31 100644
--- a/toolbox/DEM/spm_DEM_MEG.m
+++ b/toolbox/DEM/spm_DEM_MEG.m
@@ -17,7 +17,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_DEM_MEG.m 2907 2009-03-20 13:02:24Z karl $
+% $Id: spm_DEM_MEG.m 4187 2011-02-01 20:13:57Z karl $
 
 % defaults
 %--------------------------------------------------------------------------
@@ -36,7 +36,7 @@
 
 % loop over hierarchical (cortical) levels
 %--------------------------------------------------------------------------
-if graphics, cla; hold on; end
+if graphics, hold off; end
 
 z     = DEM.qU.z;
 w     = DEM.qU.w;
@@ -71,7 +71,7 @@
     end
 
     if graphics
-        plot(pst,R{k,1},'r')
+        plot(pst,R{k,1},'r'), hold on
         try
             plot(pst,R{k,2},'r:')
         end
diff --git a/toolbox/DEM/spm_SCK.m b/toolbox/DEM/spm_SCK.m
new file mode 100644
index 0000000..1a64b24
--- /dev/null
+++ b/toolbox/DEM/spm_SCK.m
@@ -0,0 +1,743 @@
+function SCKS = spm_SCK(SCKS)
+% FORMAT SCKS = spm_SCK(SCKS)
+%__________________________________________________________________________
+% Square-root Cubature Kalman Filters [2] & Square-root Rauch-Tang-Striebel
+% Smoother (SCKF-SCKS [1]).
+%==========================================================================
+% This function performs joint estimation of the states, input and parameters
+% of a model that is described as a stochastic continuous-discrete 
+% state-space in terms of nonlinear blind deconvolution. The state equations
+% must have the form of ordinary differential equations, where the 
+% discretization is performed through local-linearization scheme [3]. 
+% Additionally, the parameter noise covariance is estimated online via 
+% stochastic Robbins-Monro approximation method [4], and the measurement noise 
+% covariance is estimated using a combined variational Bayesian (VB) 
+% approach with a nonlinear filter/smoother [5].
+%__________________________________________________________________________
+%
+% SCKS.M  - model structure (based on DEM [6] in SPM8 toolbox)
+% SCKS.Y  - response variable, output or data
+%__________________________________________________________________________
+%
+% generative model:
+%--------------------------------------------------------------------------
+%   M(1).f  = dx/dt = f(x,v,P)    {inline function, string or m-file}
+%   M(1).g  = y(t)  = g(x,v,P)    {inline function, string or m-file}
+%   
+%   M(1).xP = state error covariance matrix
+%   M(1).uP = input error variance
+%   M(1).wP = parameter error covariance matrix
+%
+%   M(1).pE = prior expectation of p model-parameters
+%   M(1).pC = prior covariances of p model-parameters
+%   M(1).ip = parameter indices
+%   M(1).cb = constrain on parameters [lower, upper];
+%
+%   M(1).Q  = precision components on observation noise
+%   M(1).V  = fixed precision (input noise)
+%   M(1).W  = precision on state noise (approximated by annealing)
+%
+%   M(i).m  = number of inputs v(i + 1);
+%   M(1).n  = number of states x(i);
+%   M(1).l  = number of output v(i);
+%
+%   M(1).Qf      = form of measurement noise cov estimate:
+%                  'auto'[default],'min','mean'
+%   M(1).E.nN    = number of SCKF-SCKS algorithm iterations
+%   M(1).E.Itol  = tolerance value for SCKF-SCKS convergence 
+%   M(1).E.nD    = number of integration step between observations
+%   M(1).VB.N    = number of VB algorithm iterations
+%   M(1).VB.Itol = tolerance value for VB convergence 
+%   M(1).VB.l    = VB scaling factor;
+%
+%   conditional moments of model-states - q(u)
+%--------------------------------------------------------------------------
+%   qU.x  = Conditional expectation of hidden states (backward estimate)
+%   qU.v  = Conditional expectation of input (backward estimate)
+%   qU.z  = Conditional prediction error 
+%   qU.S  = Conditional covariance: cov(x) (states - backward estimate)
+%   qU.C  = Conditional covariance: cov(u) (input - backward estimate)
+%
+% conditional moments of model-parameters - q(p)
+%--------------------------------------------------------------------------
+%   qP.P    = Conditional expectation
+%   qP.C    = Conditional covariance
+%
+%      F    = negative log-likelihood
+%__________________________________________________________________________
+% Copyright (c) Brno University of Technology (2010)...
+% Martin Havlicek 05-12-2010
+% 
+% References:
+% [1] Havlicek M et al (2011)
+% [2] Arasaratnam, I., Haykin, S. (2009) Cubature Kalman Filters. IEEE
+%     Transactions on Automatic Control 54, 1254-1269.
+% [3] Jimenez, J.C. (2002) A simple algebraic expression to evaluate the
+%     local linearization schemes for stochastic differential equations* 
+%     1. Applied Mathematics Letters 15, 775-780.
+% [4] Van der Merwe, R., 2004. Sigma-point Kalman filters for probabilistic
+%     inference in dynamic state-space models. Ph.D.thesis, Oregon Graduate 
+%     Institute of Science and Technology.
+% [5] Sarkka, S., Hartikainen, J. (2011?) Extension of VB-AKF to Estimation
+%     of Full Covariance and Non-Linear Systems. In Press.
+% [6] Friston, K.J., et al. (2008) DEM: a variational treatment of dynamic
+%     systems. Neuroimage 41, 849-885.
+%__________________________________________________________________________
+% Copyright (C) - Martin Havlicek
+ 
+% Martin Havlicek
+% $Id: spm_SCK.m 4146 2010-12-23 21:01:39Z karl $
+% check model specification
+%--------------------------------------------------------------------------
+M  = SCKS.M;
+M  = spm_DEM_M_set(M);
+ 
+% get integration step dt:
+%--------------------------------------------------------------------------
+nD = M(1).E.nD;    
+dt = 1/nD;            
+
+% INITIALISATION:
+% =========================================================================
+
+% interpolate observation according to integration step
+%--------------------------------------------------------------------------
+y     = SCKS.Y;            % observations
+if size(y,1) > size(y,2)   % check the dimensions
+    y = y';
+end
+
+% interpolate if dt < 1:
+%--------------------------------------------------------------------------
+y    = spm_interp(y',1/dt)';
+if size(y,1) > size(y,2)   % check dimensions again
+    y = y';
+end
+T    = size(y,2);          % number of time points 
+ 
+% initial condition:
+%--------------------------------------------------------------------------
+x     = M(1).x;            % states
+u     = M(2).v;            % input
+pE    = spm_vec(M(1).pE);  % all model parameter
+ip    = M(1).ip;           % parameter indices to be estimated
+theta = pE(ip);            % selected parameters
+ 
+try cb  = M(1).cb;               catch, cb = []; end; % parameter constraints  
+try tE  = spm_vec(SCKS.pP.P{1}); catch, tE = []; end; % true parameters for display (if available)
+                            
+% covariances (square-roots)
+%--------------------------------------------------------------------------
+sR      = cell(1,T);
+[sR{:}] = deal(sparse(real(chol(inv(M(1).V)))*dt));   % observation noise variance
+sQ      = sparse(real(chol(inv(M(1).W)))*dt);         % hidden state noise variance
+if ~isempty(M(2).v)
+    sV  = sparse(real(chol(inv(M(2).V)))*dt);         % input noise variance
+else
+    sV  = [];
+end
+                                                           
+% process error covariances (square-roots)
+%--------------------------------------------------------------------------
+Sx = sparse(real(chol(M(1).xP))*dt);
+if ~isempty(M(2).v)
+    Su = sparse(real(chol(M(1).uP))*dt);
+else
+    Su = [];
+end
+if ~isempty(ip)
+    Sw = sparse(real(chol(M(1).wP(ip,ip)))*dt);
+    sW = sparse(real(chol(M(1).pC(ip,ip)))*dt);  % parameter noise variance
+    dv = diag(sW); 
+else
+    Sw = [];
+    sW = [];
+end
+ 
+% number of states, inputs and parameters:
+%--------------------------------------------------------------------------
+nx      = size(Sx,1);         % number of states
+nu      = size(sV,1);         % number of states
+nw      = size(sW,1);         % number of parameters
+no      = size(sR{1},1);      % number of observations
+noises  = nx + nu + nw + no;  % number of noise components
+ 
+% concatenate state vector and square-root error covariance:
+%--------------------------------------------------------------------------
+xc      = [x(:); u(:); theta(:)];
+xx      = zeros(nx+nu+nw,T);
+xx(:,1) = xc;
+Sc      = cell(1,T);
+[Sc{:}] = deal(sparse(nx+nu+nw,nx+nu+nw));
+Sc{1}   = blkdiag(Sx,Su,Sw);
+ 
+% get vector indices for components of concatenated state vector
+%--------------------------------------------------------------------------
+xmask   = [ones(1,nx),ones(1,nu)*2,ones(1,nw)*3,ones(1,no)*4];
+xind    = find(xmask==1);
+uind    = find(xmask==2);
+wind    = find(xmask==3);
+clear xmask;
+ 
+% setting for VB: observation noise estimation:
+%--------------------------------------------------------------------------
+if ~isempty(M(1).Q)
+    try, iter0  = M(1).VB.N;   catch, iter0 = 3;          end
+    try, lambda = M(1).VB.l;   catch, lambda = 1-exp(-2); end
+    NU      = 6;
+    V       = diag(repmat(1e-4,1,no));
+    [sR{:}] = deal(sqrt(1./(NU-no-1).*V));
+    B       = sqrt(lambda)*eye(no);
+    k       = size(sR{1},1);
+    iter    = iter0;
+    MSE0    = zeros(no,1);
+    RR0     = zeros(no,T);
+    VBrun   = [];
+else
+    iter0   = 1;
+    iter    = iter0;
+    RR      = [];
+    VBrun   = [];
+end
+ 
+% Pre-calculate cubature points: 
+%--------------------------------------------------------------------------
+n          = nx + nu + nw + noises;       % total state vector dimension
+nPts       = 2*n;                         % number of cubature points
+CubPtArray = sqrt(n)*[eye(n) -eye(n)];    % cubature points array
+ 
+% augment paramter matrix by number of cubature points:
+%--------------------------------------------------------------------------
+pE   = pE(:,ones(1,nPts));
+ 
+% prepare matrix template for integration by Local linearization scheme:
+%--------------------------------------------------------------------------
+EXPm = repmat({[ones(nx),2*ones(nx,1);zeros(1,nx+1)]},1,nPts);
+EXPm = sparse(blkdiag(EXPm{:}));
+xt   = repmat([zeros(1,nx) 1],1,nPts)';
+ 
+ 
+% =========================================================================
+% Iteration scheme:
+% =========================================================================
+% get maximum number of iterations and tolerance:
+%--------------------------------------------------------------------------
+try, ItolVB = M(1).VB.Itol;  catch,  ItolVB = 1e-4;      end
+try, Itol   = M(1).E.Itol;   catch,  Itol   = 1e-3;      end
+try, RUN    = M(1).E.nN;     catch,  RUN    = 60;        end
+try, ap     = M(1).E.RM;     catch,  ap     = [1e3 1e6]; end   % Robins-Monro approximation parameters
+ 
+MLdiff0  = 1e-4;
+mloglik0 = 0;
+ML       = [];
+VBrun    = RUN;
+t0       = tic;
+% =========================================================================
+% Iteration loop (until convergence)
+% =========================================================================
+ 
+for run = 1:RUN
+    t1  = tic;
+    mloglik = -log(2.*pi).*(T/dt);
+    % =====================================================================
+    %   Forward pass:
+    % =====================================================================
+     for t = 2:T
+         
+        sQ = diag(diag((1/sqrt(0.9995)-1)*Sc{t-1}(xind,xind)));
+        Sa = blkdiag(Sc{t-1},sQ,sV,sW,sR{t-1});
+        xa = [xc;zeros(noises,1)];
+        Xi =  xa(:,ones(1,nPts)) + Sa*CubPtArray;
+        
+        %------------------------------------------------------------------
+        % PREDICTION STEP:
+        %------------------------------------------------------------------
+        xPred(uind,:) = Xi(uind,:) + Xi(uind+nx+nu+nw,:);  % add input noise
+        xPred(wind,:) = Xi(wind,:) + Xi(wind+nx+nu+nw,:);  % add parameter noise
+        
+        % parameter constraint:
+        %------------------------------------------------------------------
+        if ~isempty(cb) && ~isempty(ip)
+            xPred(wind,:) = min(cb(:,2*ones(1,nPts)),xPred(wind,:)); % upper constrain
+            xPred(wind,:) = max(cb(:,1*ones(1,nPts)),xPred(wind,:)); % lower constrain
+        end
+        pE(ip,:) = xPred(wind,:);
+        
+        % propagation of cubature points through nonlinear function:
+        %------------------------------------------------------------------
+        f             = M(1).f(Xi(xind,:),xPred(uind,:),pE);
+        
+        % integration by local-linearization scheme:
+        %------------------------------------------------------------------
+        dfdx          = spm_diff_all(M(1).f,Xi(xind,:),xPred(uind,:),pE,1);
+        dx            = expmall(dfdx,f,dt,EXPm)*xt;
+        xPred(xind,:) = Xi(xind,:) + reshape(dx(~xt),nx,nPts) + Xi(xind+nx+nu+nw,:);
+        
+        % mean prediction:
+        %------------------------------------------------------------------
+        x1      = sum(xPred,2)/nPts;
+        X       = (xPred-x1(:,ones(1,nPts)))/sqrt(nPts);
+        [foo,S] = qr([X]',0);
+        S       = S';
+        
+        % in the case of VB observation noise estimation:
+        %------------------------------------------------------------------
+        if ~isempty(M(1).Q) && iter~=1
+            NU   = lambda.*(NU-k-1)+k+1;
+            V    = B*V*B';
+            NU   = NU + 1;
+            V0   = V;
+        end
+        x10    = x1;
+        S0     = S;
+        yPred0 = M(1).g(xPred(xind,:),xPred(uind,:),pE);
+        
+        %------------------------------------------------------------------
+        % UPDATE STEP:
+        %------------------------------------------------------------------
+        
+        % VB estimation of sR (iteratively)
+        %------------------------------------------------------------------
+        for it = 1:iter
+            
+           % VB part - update of square-root measurement noise cov
+           %---------------------------------------------------------------
+            if ~isempty(M(1).Q) && iter ~= 1
+                sR{t} = real(chol(1/(NU - no - 1)*V)); 
+            end
+            
+            % propagate cubature points through observation function:
+            %--------------------------------------------------------------
+            yPred    = yPred0 + sR{t}*CubPtArray(end-no+1:end,:);
+            y1       = sum(yPred,2)/nPts;
+            
+            Y        = (yPred-y1(:,ones(1,nPts)))/sqrt(nPts);
+            [foo,Sy] = qr([Y]',0);
+            Sy       = Sy';
+            
+            Pxy      = X*Y';                % cross covariance
+            K        = (Pxy/Sy')/Sy;        % Kalman gain
+            resid    =  y(:,t) - y1;        % innovations
+            
+            % state (input,parameter) estimates:
+            %--------------------------------------------------------------
+            xc = x10 + K*(resid);
+            
+            % check parameter constraints: 
+            %--------------------------------------------------------------
+            if ~isempty(cb) && ~isempty(ip)
+                xc(wind) = min(cb(:,2),xc(wind)); % upper constrain
+                xc(wind) = max(cb(:,1),xc(wind)); % lower constrain
+            end
+            % estimate of process error covarinace:
+            %--------------------------------------------------------------
+            [foo,S] = qr([(X - K*Y)]',0);
+            S       = S';
+            
+            % VB part
+            %--------------------------------------------------------------
+            if ~isempty(M(1).Q) && iter~=1
+                Sa     = blkdiag(S,sQ,sV,sW,sR{t});
+                Xi     = repmat([xc;zeros(noises,1)],1,nPts) + Sa*CubPtArray;
+                yPreds = M(1).g(Xi(xind,:),Xi(uind,:),pE);  % no additive noise!
+                D      = repmat(y(:,t),1,nPts)-yPreds;
+                D      = D*D'/nPts;
+                V      = V0 + D;
+            end
+         
+        end
+        Sc{t}   = S;
+        xx(:,t) = xc;
+ 
+        % Maximum log-Likelihood
+        %------------------------------------------------------------------
+         mloglik = mloglik - log(det(Sy*Sy')) - resid'/(Sy*Sy')*resid;
+       
+        % Robins-Monro stochastic approximation for of parameters noise cov
+        %------------------------------------------------------------------
+         if ~isempty(ip)
+             subKG = K(wind,:);
+             dv    = sqrt((1-1/ap(1))*(dv.^2) + 1/ap(1)*diag(subKG*(subKG*resid*resid')'));
+             sW    = diag(dv);
+             ap(1) = min(ap(1)+1,ap(2));
+         end
+         if ~isempty(M(1).Q) && iter~=1
+             RR(:,t) = diag(sR{t});
+         end
+    end
+    xxf = xx;
+    Sf  = Sc;
+    %----------------------------------------------------------------------
+    % END of forward pass
+    % ---------------------------------------------------------------------
+  
+    % =====================================================================
+    %   Backward pass:
+    % =====================================================================
+    for t = T-1:-1:1
+        % VB part:
+        %------------------------------------------------------------------
+        if ~isempty(M(1).Q) && iter~=1
+            NU    = lambda.*(NU-k-1)+k+1;
+            V     = B*V*B';
+            NU    = NU + 1;
+            V0    = V;
+            sR{t} = real(chol(1/(NU-no-1)*V)); 
+        end
+        
+        % Square-root Cubature Rauch-Tung-Striebel smoother
+        %------------------------------------------------------------------
+        
+        % evaluate cubature points:
+        %------------------------------------------------------------------
+        Sa = blkdiag(Sc{t},sQ,sV,sW,sR{t});
+        xa = [xx(:,t);zeros(noises,1)];
+        Xi = xa(:,ones(1,nPts)) + Sa*CubPtArray;
+      
+        xPred(uind,:) = Xi(uind,:) + Xi(uind+nx+nu+nw,:);
+        xPred(wind,:) = Xi(wind,:) + Xi(wind+nx+nu+nw,:);
+        
+        % check parameter constraints:
+        %------------------------------------------------------------------
+        if ~isempty(cb) && ~isempty(ip)
+            xPred(wind,:) = min(cb(:,2*ones(1,nPts)),xPred(wind,:)); % upper constrain
+            xPred(wind,:) = max(cb(:,1*ones(1,nPts)),xPred(wind,:)); % lower constrain
+        end
+        
+        pE(ip,:)      = xPred(wind,:);
+        % propagate cubature points through nonlinear function:
+        %------------------------------------------------------------------
+        f             = M(1).f(Xi(xind,:),xPred(uind,:),pE);
+        dfdx          = spm_diff_all(M(1).f,Xi(xind,:),xPred(uind,:),pE,1);
+        dx            = expmall(dfdx,f,dt,EXPm)*xt;
+        xPred(xind,:) = Xi(xind,:) + reshape(dx(~xt),nx,nPts) + Xi(xind+nx+nu+nw,:) ;
+    
+        x1       = sum(xPred,2)/nPts;
+        X        = (xPred-x1(:,ones(1,nPts)))/sqrt(nPts) + eps;
+        x01      = xx(:,t);
+        X01      = (Xi([xind,uind,wind],:) - repmat(x01,1,nPts))/sqrt(nPts);
+        [foo,S]  = qr([X]',0);
+        S        = S';
+        
+        Pxy      = X01*X';      % cross covariance
+        K        = (Pxy/S')/S;  % Kalman gain
+        
+        % smoothed estimate of the states (input, parameters)
+        % and process error covariance:
+        %------------------------------------------------------------------
+        xx(:,t)  = xx(:,t) + K*(xx(:,t+1) - x1);
+        [foo,S]  = qr([X01 - K*X, K*Sc{t+1}]',0);
+        S        = S';
+        Sc{t}    = S;
+        
+        % check parameter constraint:
+        %------------------------------------------------------------------
+        if ~isempty(cb) && ~isempty(ip)
+            xx(wind,t) = min(cb(:,2),xx(wind,t)); % upper constrain
+            xx(wind,t) = max(cb(:,1),xx(wind,t));
+        end
+        
+        % VB part (smoothing):
+        %------------------------------------------------------------------
+        if ~isempty(M(1).Q) && iter~=1
+            Sa     = blkdiag(S,sQ,sV,sW,sR{t});
+            Xi     = repmat([xx(:,t);zeros(noises,1)],1,nPts) + Sa*CubPtArray;
+            yPreds = M(1).g(Xi(xind,:),Xi(uind,:),pE);  % no additive noise!
+            D      = repmat(y(:,t),1,nPts)-yPreds;
+            D      = D*D'/nPts;
+            V      = V0 + D;
+        end
+        
+        if ~isempty(M(1).Q) && iter~=1
+            RR(:,t) = diag(sR{t});
+        end
+     
+    end
+    xxb = xx;
+    Sb  = Sc;
+    %----------------------------------------------------------------------
+    % END of backward pass
+    %----------------------------------------------------------------------
+ 
+    str{1} = sprintf('SCKS: %i (1:%i)',run,iter);
+ 
+    % iteration condition for measurement noise estimate:
+    % iterate until stabilization of sR estimate
+    %----------------------------------------------------------------------
+    if ~isempty(M(1).Q) && iter0 ~= 1
+        
+        MSE     = mean((RR -(RR0)).^2,2);
+        RR0     = RR;
+        MSEdiff = abs(MSE - MSE0);
+        MSE0    = MSE;
+ 
+        if MSEdiff < ItolVB  % (till it gets stable)
+            switch(lower(M(1).Qf))
+                case('all')
+                    % take all
+                case('mean')
+                    sR      = cell(1,T);
+                    [sR{:}] = deal(diag(mean(RR,2)));
+                case('min')
+                    RRs     = sort(RR,2,'descend');
+                    sR      = cell(1,T);
+                    [sR{:}] = deal(diag(mean(RRs(:,round(T*0.90):end),2)));
+                case('auto')
+                    dlim    = min(RR,[],2);
+                    ulim    = max(RR,[],2);
+                    if all(ulim./dlim<4)
+                        % take all
+                    else
+                        RRs     = sort(RR,2,'descend');
+                        sR      = cell(1,T);
+                        [sR{:}] = deal(diag(mean(RRs(:,round(T*0.90):end),2)));
+                    end
+            end
+            iter0    = 1;
+            iter     = iter0;
+            mloglik0 = 0;
+            VBrun    = run;
+        end
+    end
+    
+    
+    % log-likelihood difference:
+    %----------------------------------------------------------------------
+    MLdiff(run) = abs(mloglik-mloglik0);
+    ML(run)     = mloglik;
+    if mloglik > 0
+        mloglik     = mloglik - 5000;
+        MLdiff(run) = abs(mloglik-mloglik0);
+        ML(run)     = mloglik;
+    end
+    
+    timed  = toc(t1);
+    str{2} = sprintf('F:%.4e',ML(end));
+    str{3} = sprintf('dF:%.4e',MLdiff(end));
+    str{4} = sprintf('(%.2e sec)',timed);
+    fprintf('%-16s%-16s%-16s%-16s\n',str{:})
+ 
+    % plot estimates:
+    %----------------------------------------------------------------------
+    doplotting(M,xxf,xxb,Sf,Sb,ML,T,wind,ip,run,RR,VBrun);
+ 
+    
+    % stopping condition:
+    %----------------------------------------------------------------------
+    if RUN > 1 && (~isempty(ip) || ~isempty(M(1).Q))
+        if run == 2
+            MLdiff0 = MLdiff(run);
+        elseif run > 2
+            if MLdiff0 < MLdiff(run),
+                MLdiff0 = MLdiff(run);
+            end
+        end
+        if (((MLdiff(run)/MLdiff0)<Itol || run==RUN) || (isempty(ip)&&MLdiff(run)<Itol)) && iter0 == 1,
+ 
+            yy       = M(1).g(xx(xind,:),xx(uind,:),pE(:,1));
+            res      = y - yy;
+ 
+            try SCKS = rmfield(SCKS,'qU'); end
+            try SCKS = rmfield(SCKS,'qP'); end
+            try SCKS = rmfield(SCKS,'qH'); end
+            
+            % save results in structure:
+            %--------------------------------------------------------------
+            SCKS.qU.v{1} = yy(:,1:nD:end);
+            SCKS.qU.x{1} = xxb(xind,1:nD:end);
+            SCKS.qU.v{2} = xxb(uind,1:nD:end);
+            SCKS.qU.z{1} = res(:,1:nD:end);
+            if ~isempty(ip)
+                qP           = SCKS.M(1).pE;
+                qP(ip)       = mean(xx(wind,:),2);
+                SCKS.qP.P{1} = qP;
+                SCKS.qP.P{2} = [];
+            end
+            SCKS.F       = ML;
+            
+            for i = 1:nD:T
+                j = 1 + (i - 1)/nD;
+                SCKS.qU.S{j} = Sb{i}(xind,xind);
+                SCKS.qU.C{j} = Sb{i}(uind,uind);
+                if ~isempty(ip)
+                    SCKS.qP.p{j} = xxb(wind,:);
+                    SCKS.qP.c{j} = Sb{i}(wind,wind);
+                    SCKS.qP.ip   = ip;
+                end
+            end
+            return
+        end
+ 
+        mloglik0 = mloglik;
+        xc       = [xx([xind,uind],1); mean(xx(wind,:),2)];
+        xx(:,1)  = xc;
+ 
+    else
+        
+        pE(ip,1) = mean(xx(wind,:),2);
+        yy       = M(1).g(xx(xind,:),xx(uind,:),pE(:,1));
+        res      = y - yy;
+ 
+        try SCKS = rmfield(SCKS,'qU'); end
+        try SCKS = rmfield(SCKS,'qP'); end
+        try SCKS = rmfield(SCKS,'qH'); end
+        
+        % save results in structure:
+        %------------------------------------------------------------------
+        SCKS.qU.v{1} = yy(:,1:nD:end);
+        SCKS.qU.x{1} = xxb(xind,1:nD:end);
+        SCKS.qU.v{2} = xxb(uind,1:nD:end);
+        SCKS.qU.z{1} = res(:,1:nD:end);
+        if ~isempty(ip)
+            qP           = SCKS.M(1).pE;
+            qP(ip)       = mean(xx(wind,:),2);
+            SCKS.qP.P{1} = qP;
+            SCKS.qP.P{2} = [];
+        end
+        SCKS.F       = ML;
+        
+        for i = 1:nD:T
+            j = 1 + (i - 1)/nD;
+            SCKS.qU.S{j} = Sb{i}(xind,xind);
+            SCKS.qU.C{j} = Sb{i}(uind,uind);
+            if ~isempty(ip)
+                SCKS.qP.p{j} = xxb(wind,:);
+                SCKS.qP.c{j} = Sb{i}(wind,wind);
+                SCKS.qP.ip   = ip;
+            end
+        end
+        return
+    end
+end
+ 
+%==========================================================================
+%==========================================================================
+ 
+ 
+%--------------------------------------------------------------------------
+% Plot estimates at each iteration:
+%--------------------------------------------------------------------------
+function doplotting(M,xxf,xxb,Sf,Sb,ML,T,wind,ip,run,RR,VBrun)
+ 
+% Initialize display:
+%--------------------------------------------------------------------------
+spm_figure('GetWin','SCKF-SCKS estimates');
+set(gcf,'Renderer','painter'); clf
+ 
+% Hidden states
+%--------------------------------------------------------------------------
+for p=1:2
+    subplot(3,3,[1:3]+3*(p-1)),
+    hax = gca;
+    si    = spm_invNcdf(1 - 0.05);
+    s     = [];
+    if p == 1,
+        xxfig = xxf;
+        Sfig  = Sf;
+        tit   = 'SCKF - forward pass';
+    else
+        xxfig = xxb;
+        Sfig  = Sb;
+        tit   = 'SCKS - backward pass';
+    end
+    for i = 1:T
+        s = [s abs(diag(Sfig{i}))];
+    end
+ 
+    % conditional covariances
+    %----------------------------------------------------------------------
+    j           = [1:size(xxfig(:,:),1)];
+    ss          = si*s(j,:);
+    [ill indss] = sort(full(mean(ss,2)),'descend');
+ 
+    pf = plot(1:T,xxfig,'linewidth',1.5);
+    set(hax,'xlim',[1,T],'nextplot','add')
+    box(hax,'on')
+    for ic = 1:size(xxfig,1)
+        col0 = get(pf(indss(ic)),'color');
+        col  = (ones(1,3)-col0)*0.65 + col0;
+        fill([(1:T) fliplr(1:T)],[(xxfig(indss(ic),:) + ss(indss(ic),:)) fliplr((xxfig(indss(ic),:) - ss(indss(ic),:)))],...
+            'r',...
+            'FaceColor',col,...
+            'EdgeColor',col);
+        hold on;
+        COL{ic} = col0;
+    end
+    for ic = 1:size(xxfig,1)
+        plot(xxfig(indss(ic),:),'color',COL{ic},'linewidth',0.75);
+    end
+    title(tit);
+    grid(hax,'on');
+    axis(hax,'tight');
+end
+ 
+ 
+% Parameter estimates
+%--------------------------------------------------------------------------
+subplot(3,3,7)
+h = plot([1:length(ML)],ML);
+if ~isempty(M(1).Q)
+    AYlim = get(gca,'Ylim');
+    if VBrun>=run
+        bkg = ones(1,run)*max(AYlim);
+    else
+        bkg = [ones(1,VBrun)*max(AYlim),ones(1,abs(VBrun-run))*min(AYlim)];
+    end
+    a = area(bkg,min(AYlim));
+    axis([1 run+1 AYlim(1) AYlim(2)]); hold on;
+    set(a(1),'FaceColor',ones(1,3)*0.6,'EdgeColor',ones(1,3)*0.6)
+    h = plot([1:length(ML)],ML);
+    axis([1 run+1 AYlim(1) AYlim(2)]); hold on;
+    if VBrun>=run
+        text(run/2+1,mean(AYlim),'VB-SCKS','HorizontalAlignment','center','VerticalAlignment','top');
+    else
+        text(VBrun/2+1,mean(AYlim),'VB-SCKS','HorizontalAlignment','center','VerticalAlignment','top');
+        text(VBrun+(abs(VBrun-run)/2)+1,mean(AYlim),'SCKS','HorizontalAlignment','center','VerticalAlignment','top');
+    end
+    set(gca,'Layer','top')
+    hold off
+end
+set(h,'color','k','Marker','o','Markersize',4,'MarkerFaceColor','k','linewidth',1);
+title('Log-Likelihood');
+ 
+if ~isempty(ip)
+    subplot(3,3,8)
+    b1 = bar(ip,mean(xxb(wind,:),2)','FaceColor','k');
+    set(b1,'BarWidth',0.5); hold on;
+    title('Parameters');
+end
+ 
+% Standard deviation of residuals
+%--------------------------------------------------------------------------
+if ~isempty(M(1).Q)
+    subplot(3,3,9)
+    if VBrun >= run
+        plot(RR'); hold on;
+    else
+        plot(RR'); hold on;
+        switch(lower(M(1).Qf))
+            case('all')
+                sR  = RR';
+            case('mean')
+                sR  = repmat(mean(RR,2),1,T)';
+            case('min')
+                RRs = sort(RR,2,'descend');
+                sR  = repmat(mean(RRs(:,round(T*0.90):end),2),1,T)';
+            case('auto')
+                dlim    = min(RR,[],2);
+                ulim    = max(RR,[],2);
+                if all(ulim./dlim<4)
+                    sR  = RR';
+                else
+                    RRs  = sort(RR,2,'descend');
+                    sR   = repmat(mean(RRs(:,round(T*0.90):end),2),1,T)';
+                end
+        end
+        plot([1:T],sR,'r','linewidth',2);
+    end
+    axis(gca,'tight');
+    title('Standard deviation of residuals');
+    hold off;
+end
+drawnow
+
diff --git a/toolbox/DEM/spm_dem_cue_movie.m b/toolbox/DEM/spm_dem_cue_movie.m
new file mode 100644
index 0000000..dc4b74f
--- /dev/null
+++ b/toolbox/DEM/spm_dem_cue_movie.m
@@ -0,0 +1,81 @@
+function spm_dem_cue_movie(DEM,q)
+% creates a movie of cued pointing
+% FORMAT spm_dem_cue_movie(DEM,q)
+%
+% DEM - DEM structure from reaching simulations
+% q   - flag switching from true to perceived reaching
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_dem_cue_movie.m 4230 2011-03-07 20:58:38Z karl $
+
+% Dimensions
+%--------------------------------------------------------------------------
+N    = size(DEM.pU.x{1},2);
+n    = size(DEM.pP.P{1},2);
+
+% evaluate true location (targets)
+%----------------------------------------------------------------------
+for i = 1:N
+    L(:,:,i) = DEM.pP.P{1};
+end
+
+if nargin > 1
+    
+    % evaluate perceived positions (motor plant)
+    %----------------------------------------------------------------------
+    x    = tan(DEM.qU.x{1}(1:2,:));                  % finger location
+    c    = DEM.qU.x{1}(2 + (1:n),:);                 % target contrast
+    
+else
+    
+    % evaluate true positions (motor plant)
+    %----------------------------------------------------------------------
+    x    = tan(DEM.pU.x{1}(1:2,:));                  % finger location
+    c    = DEM.pU.x{1}(2 + (1:n),:);                 % target contrast
+    
+end
+c    = c - min(c(:)) + 1/32;
+c    = c/max(c(:));
+fin  = imread('pointfinger.jpg');
+
+
+% movie
+%--------------------------------------------------------------------------
+s     = 2;
+for i = 1:N
+    
+    cla
+    axis image ij
+    hold on
+    
+    % finger
+    %----------------------------------------------------------------------
+    imagesc(([-1 0] + .68)*s + x(1,i),([-1 0] + .96)*s + x(2,i),fin);
+    hold on
+    
+    % trajectory
+    %----------------------------------------------------------------------
+    plot(x(1,1:i),x(2,1:i),'k:')
+    
+    % targets
+    %----------------------------------------------------------------------
+    for j = 1:n
+        plot(L(1,j,i),L(2,j,i),'.','MarkerSize',64,'color',[c(j,i) (1 - c(j,i)) 0])
+    end
+    axis([-1 1 -1 1]*2)
+    hold off
+    drawnow
+    
+    % save
+    %----------------------------------------------------------------------
+    M(i) = getframe(gca);
+    
+end
+
+% set ButtonDownFcn
+%--------------------------------------------------------------------------
+h = findobj(gca,'type','image');
+set(h(1),'Userdata',{M,16})
+set(h(1),'ButtonDownFcn','spm_DEM_ButtonDownFcn')
diff --git a/toolbox/DEM/spm_dem_pursuit_movie.m b/toolbox/DEM/spm_dem_pursuit_movie.m
new file mode 100644
index 0000000..287d2d7
--- /dev/null
+++ b/toolbox/DEM/spm_dem_pursuit_movie.m
@@ -0,0 +1,104 @@
+function spm_dem_pursuit_movie(DEM)
+% creates a movie of visual prusuit in extrinsic and intrinsic coordinates
+% FORMAT spm_dem_pursuit_movie(DEM)
+%
+% DEM - DEM structure from reaching simulations
+%
+% hidden causes and states
+%--------------------------------------------------------------------------
+% x    - hidden states:
+%   x.o(1) - oculomotor angle
+%   x.o(2) - oculomotor angle
+%   x.x(1) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.x(2) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.a(:) - attractor (SHC) states
+%
+% v    - causal states
+%   v(1) - not used
+%
+% g    - sensations:
+%   g(1) - oculomotor angle (proprioception)
+%   g(2) - oculomotor angle (proprioception)
+%   g(3) - target location (visual) - intrinsic coordinates (polar)
+%   g(4) - target location (visual) - intrinsic coordinates (polar)
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dem_pursuit_movie.m 4170 2011-01-24 18:37:42Z karl $
+ 
+ 
+% movie
+%--------------------------------------------------------------------------
+N      = length(DEM.pU.v{2});
+p      = -pi:pi/64:pi;
+circle = [sin(p);cos(p)];
+ 
+for i = 1:N
+ 
+    
+    % stimulus and tracking in extrinsic coordinates
+    %======================================================================
+    subplot(2,2,1), cla
+    axis image ij
+    hold on
+    px  = spm_unvec(DEM.pU.x{1}(:,i),DEM.M(1).x);   % extrinsic target
+    qx  = spm_unvec(DEM.qU.x{1}(:,i),DEM.M(1).x);   % conditional estimate
+    
+ 
+    % target location
+    %----------------------------------------------------------------------
+    plot(0,0,'+','MarkerSize',32,'Color',[1 1 1]*(1 - 1/8))
+    plot(px.x(1),px.x(2),'.','MarkerSize',32,'Color',[1 .0 .0])
+    plot(qx.x(1),px.x(2),'.','MarkerSize',32,'Color',[1 .4 .4])
+    plot(px.o(1),px.o(2),'+','MarkerSize',32,'Color',[1 .4 .4])
+    axis image, grid on, box off
+    axis([-1 1 -1 1]*(1 + 1/8))
+    drawnow
+    
+    % save
+    %----------------------------------------------------------------------
+    Me(i) = getframe(gca);
+    
+    
+    % stimulus and tracking in intrinsic coordinates
+    %======================================================================
+    subplot(2,2,2), cla
+    hold on
+    pv  = DEM.pU.v{1}(:,i);   % intrinsic stimulus
+    
+    % target location
+    %----------------------------------------------------------------------
+    for j = 1:3
+       plot(circle(1,:)*j/2,circle(2,:)*j/2,':k');
+    end
+    plot(p,p*0,':k');plot(p*0,p,':k');
+    plot(pv(3),pv(4),'.','MarkerSize',32,'Color',[1 .0 .0]);     
+    plot(-tan(pv(1)),-tan(pv(2)),'+','MarkerSize',32,'Color',[1 1 1]*(1 - 1/8));
+    plot(0,0,'+','MarkerSize',32,'Color',[1 .4 .4]);
+    axis image, box off
+    axis([-1 1 -1 1]*pi/2)
+    drawnow
+    
+    % save
+    %----------------------------------------------------------------------
+    Mi(i) = getframe(gca);
+ 
+end
+ 
+ 
+% set ButtonDownFcn
+%--------------------------------------------------------------------------
+subplot(2,2,1)
+set(gca,'Userdata',{Me,16})
+set(gca,'ButtonDownFcn','spm_DEM_ButtonDownFcn')
+title('left click for movie','FontSize',16)
+xlabel('displacement (cm)')
+hold off
+ 
+subplot(2,2,2)
+set(gca,'Userdata',{Mi,16})
+set(gca,'ButtonDownFcn','spm_DEM_ButtonDownFcn')
+title('left click for movie','FontSize',16)
+xlabel('displacement (radians)')
+hold off
diff --git a/toolbox/DEM/spm_dem_set_movie.m b/toolbox/DEM/spm_dem_set_movie.m
new file mode 100644
index 0000000..36f735b
--- /dev/null
+++ b/toolbox/DEM/spm_dem_set_movie.m
@@ -0,0 +1,81 @@
+function spm_dem_cue_movie(DEM,q)
+% creates a movie of cued pointing
+% FORMAT spm_dem_cue_movie(DEM,q)
+%
+% DEM - DEM structure from reaching simulations
+% q   - flag switching from true to perceived reaching
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_dem_set_movie.m 4231 2011-03-07 21:00:02Z karl $
+
+% Dimensions
+%--------------------------------------------------------------------------
+N    = size(DEM.pU.v{1},2);
+n    = size(DEM.pP.P{1},2);
+
+% evaluate true location (targets)
+%----------------------------------------------------------------------
+for i = 1:N
+    L(:,:,i) = DEM.pP.P{1};
+end
+
+if nargin > 1
+    
+    % evaluate perceived positions (motor plant)
+    %----------------------------------------------------------------------
+    x    = tan(DEM.qU.v{1}(1:2,:));                  % finger location
+    c    = DEM.qU.v{1}(4 + (1:n),:);                 % target contrast
+    
+else
+    
+    % evaluate true positions (motor plant)
+    %----------------------------------------------------------------------
+    x    = tan(DEM.pU.v{1}(1:2,:));                  % finger location
+    c    = DEM.pU.v{1}(4 + (1:n),:);                 % target contrast
+    
+end
+c    = c - min(c(:)) + 1/32;
+c    = c/max(c(:));
+fin  = imread('pointfinger.jpg');
+
+
+% movie
+%--------------------------------------------------------------------------
+s     = 2;
+for i = 1:N
+    
+    cla
+    axis image ij
+    hold on
+    
+    % finger
+    %----------------------------------------------------------------------
+    imagesc(([-1 0] + .68)*s + x(1,i),([-1 0] + .96)*s + x(2,i),fin);
+    hold on
+    
+    % trajectory
+    %----------------------------------------------------------------------
+    plot(x(1,1:i),x(2,1:i),'k:')
+    
+    % targets
+    %----------------------------------------------------------------------
+    for j = 1:n
+        plot(L(1,j,i),L(2,j,i),'.','MarkerSize',64,'color',[c(j,i) (1 - c(j,i)) 0])
+    end
+    axis([-1 1 -1 1]*2)
+    hold off
+    drawnow
+    
+    % save
+    %----------------------------------------------------------------------
+    M(i) = getframe(gca);
+    
+end
+
+% set ButtonDownFcn
+%--------------------------------------------------------------------------
+h = findobj(gca,'type','image');
+set(h(1),'Userdata',{M,16})
+set(h(1),'ButtonDownFcn','spm_DEM_ButtonDownFcn')
diff --git a/toolbox/DEM/spm_diff_all.m b/toolbox/DEM/spm_diff_all.m
new file mode 100644
index 0000000..89089b0
--- /dev/null
+++ b/toolbox/DEM/spm_diff_all.m
@@ -0,0 +1,144 @@
+function [varargout] = spm_diff_all(varargin)
+% matrix high-order numerical differentiation
+% FORMAT [dfdx] = spm_diff(f,x,...,n)
+% FORMAT [dfdx] = spm_diff(f,x,...,n,V)
+% FORMAT [dfdx] = spm_diff(f,x,...,n,'q')
+%
+% f      - [inline] function f(x{1},...)
+% x      - input argument[s]
+% n      - arguments to differentiate w.r.t.
+%
+% V      - cell array of matrices that allow for differentiation w.r.t.
+% to a linear transformation of the parameters: i.e., returns
+% 
+% df/dy{i};    x = V{i}y{i};    V = dx(i)/dy(i)
+%
+% q      - flag to preclude default concatenation of dfdx
+%
+% dfdx          - df/dx{i}                     ; n =  i
+% dfdx{p}...{q} - df/dx{i}dx{j}(q)...dx{k}(p)  ; n = [i j ... k]
+%
+%
+% - a cunning recursive routine
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_diff_all.m 4146 2010-12-23 21:01:39Z karl $
+ 
+% create inline object
+%--------------------------------------------------------------------------
+f     = varargin{1};
+ 
+% parse input arguments
+%--------------------------------------------------------------------------
+if iscell(varargin{end})
+    x = varargin(2:(end - 2));
+    n = varargin{end - 1};
+    V = varargin{end};
+    q = 1;
+elseif isnumeric(varargin{end})
+    x = varargin(2:(end - 1));
+    n = varargin{end};
+    V = cell(1,length(x));
+    q = 1;
+elseif ischar(varargin{end})
+    x = varargin(2:(end - 2));
+    n = varargin{end - 1};
+    V = cell(1,length(x));
+    q = 0;
+else
+    error('improper call')
+end
+ 
+% check transform matrices V = dxdy
+%--------------------------------------------------------------------------
+for i = 1:length(x(:))
+    try
+        V{i};
+    catch
+        V{i} = [];
+    end
+    if isempty(V{i}) && any(n == i);
+      nx = size(x{1},2);
+      V{i} = speye(length(spm_vec(x{i}))/nx);
+    end
+end
+ 
+% initialise
+%--------------------------------------------------------------------------
+m     = n(end);
+xm    = x{m};
+dx    = exp(-8);
+J0    = cell(1,size(V{m},2));
+J     = sparse(size(V{m},2)*size(V{m},2),nx);
+% proceed to derivatives
+%==========================================================================
+if length(n) == 1
+ 
+    % dfdx
+    %----------------------------------------------------------------------
+    f0    = feval(f,x{:});
+    for i = 1:length(J0)
+        xi    = x;
+        xmi   = xm + V{m}(:,i*ones(1,nx))*dx;
+        xi{m} = spm_unvec(xmi,x{m});
+        fi    = feval(f,xi{:});
+        J0{i}  = spm_dfdx(fi,f0,dx);
+    end
+
+    J = sparse(cell2mat(J0'));
+    % return numeric array for first-order derivatives
+    %======================================================================
+ 
+    % vectorise f
+    %----------------------------------------------------------------------
+    f  = spm_vec(f0);
+   
+    
+    % assign output argument and return
+    %----------------------------------------------------------------------
+    varargout{1} = J;
+    varargout{2} = f0;
+ 
+else
+ 
+    % dfdxdxdx....
+    %----------------------------------------------------------------------
+    f0        = cell(1,length(n));
+    [f0{:}]   = spm_diff(f,x{:},n(1:end - 1),V);
+    
+    for i = 1:length(J)
+        xi    = x;
+        xmi   = xm + V{m}(:,i)*dx;
+        xi{m} = spm_unvec(xmi,x{m});
+        fi    = spm_diff(f,xi{:},n(1:end - 1),V);
+        J{i}  = spm_dfdx(fi,f0{1},dx);
+    end
+    varargout = {J f0{:}};
+end
+ 
+ 
+function dfdx = spm_dfdx(f,f0,dx)
+% cell subtraction
+%--------------------------------------------------------------------------
+if iscell(f)
+    dfdx  = f;
+    for i = 1:length(f(:))
+        dfdx{i} = spm_dfdx(f{i},f0{i},dx);
+    end
+elseif isstruct(f)
+    dfdx  = (spm_vec(f) - spm_vec(f0))/dx;
+else
+    dfdx  = (f - f0)/dx;
+end
+ 
+ 
+function is = isvec(v)
+% isvector(v) returns true if v is 1-by-n or n-by-1 where n>=0
+%__________________________________________________________________________
+ 
+% vec if just two dimensions, and one (or both) unity
+%--------------------------------------------------------------------------
+is = length(size(v)) == 2 && isnumeric(v);
+is = is && (size(v,1) == 1 || size(v,2) == 1);
diff --git a/toolbox/DEM/spm_fx_adem_cue.m b/toolbox/DEM/spm_fx_adem_cue.m
new file mode 100644
index 0000000..1d46b8c
--- /dev/null
+++ b/toolbox/DEM/spm_fx_adem_cue.m
@@ -0,0 +1,26 @@
+function [f]= spm_fx_adem_cue(x,v,a,P)
+% returns the flow for cued response (with action)
+% FORMAT [f]= spm_fx_adem_cue(x,v,a,P)
+%
+% x    - hidden states:
+%   x.o  - motor angle
+%
+% v    - hidden causes
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fx_adem_cue.m 4230 2011-03-07 20:58:38Z karl $
+ 
+% intisaise flow (to ensure fields are aligned)
+%--------------------------------------------------------------------------
+f    = x;
+
+% motion of oculomotor angles (driven by bounded action) with decay to 0
+%==========================================================================
+f.o  = tanh(a) - x.o/8;
+
+% motion of target contrast
+%==========================================================================
+f.a  = v - x.a;
\ No newline at end of file
diff --git a/toolbox/DEM/spm_fx_adem_pursuit.m b/toolbox/DEM/spm_fx_adem_pursuit.m
new file mode 100644
index 0000000..4cb7fb6
--- /dev/null
+++ b/toolbox/DEM/spm_fx_adem_pursuit.m
@@ -0,0 +1,42 @@
+function [f]= spm_fx_adem_pursuit(x,v,a,P)
+% returns the flow for a two-joint arm (with action)
+% FORMAT [f]= spm_fx_adem_pursuit(x,v,a,P)
+%
+% x    - hidden states:
+%   x.o(1) - oculomotor angle
+%   x.o(2) - oculomotor angle
+%   x.x(1) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.x(2) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.a(:) - attractor (SHC) states
+%
+% v    - hidden cause (speed)
+% P    - parameters
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fx_adem_pursuit.m 4170 2011-01-24 18:37:42Z karl $
+ 
+% intisaise flow (to ensure fields are aligned)
+%--------------------------------------------------------------------------
+f    = x;
+ 
+% motion of attractor states
+%==========================================================================
+f.a  = spm_lotka_volterra(x.a,v);
+ 
+ 
+% motion of target states
+%==========================================================================
+ 
+% target location is determined by the attractor state x.a
+%--------------------------------------------------------------------------
+n    = length(x.a);
+p    = exp(2*x.a);
+t    = P*p/sum(p);
+f.x  = (t - x.x)/8;
+ 
+ 
+% motion of oculomotor angles (driven by bounded action)
+%==========================================================================
+f.o  = tanh(a) - x.o/64;
diff --git a/toolbox/DEM/spm_fx_dem_cue.m b/toolbox/DEM/spm_fx_dem_cue.m
new file mode 100644
index 0000000..76d6ffc
--- /dev/null
+++ b/toolbox/DEM/spm_fx_dem_cue.m
@@ -0,0 +1,34 @@
+function [f]= spm_fx_dem_cue(x,v,P)
+% returns the flow for cued response
+% FORMAT [f]= spm_fx_dem_cue(x,v,P)
+%
+% x    - hidden states:
+%   x.o  - intrinsic motor state (proprioceptive)
+%   x.a  - target salience (attractiveness)
+%
+% v    - hidden causes
+%
+% P.x  - target locations (visual) - extrinsic coordinates (Cartesian)
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fx_dem_cue.m 4230 2011-03-07 20:58:38Z karl $
+ 
+% intisaise flow (to ensure fields are aligned)
+%--------------------------------------------------------------------------
+f    = x;
+
+% motion of oculomotor angles (attracted to target)
+%==========================================================================
+
+% target location is determined by the attractor state softmax(x.a)
+%--------------------------------------------------------------------------
+t    = P.x*spm_softmax(x.a,P.s);
+f.o  = (t - tan(x.o))/2;
+
+% motion of attractor states
+%==========================================================================
+f.a  = spm_lotka_volterra(x.a,1/2*v(1));
+
+
diff --git a/toolbox/DEM/spm_fx_dem_pursuit.m b/toolbox/DEM/spm_fx_dem_pursuit.m
new file mode 100644
index 0000000..6dcbefd
--- /dev/null
+++ b/toolbox/DEM/spm_fx_dem_pursuit.m
@@ -0,0 +1,49 @@
+function [f]= spm_fx_dem_pursuit(x,v,P)
+% returns the flow for visual pursuit demo
+% FORMAT [f]= spm_fx_dem_pursuit(x,v,P)
+%
+% x    - hidden states:
+%   x.o(1) - oculomotor angle
+%   x.o(2) - oculomotor angle
+%   x.x(1) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.x(2) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x.a(:) - attractor (SHC) states
+%
+% v    - hidden causes
+% P    - parameters
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fx_dem_pursuit.m 4187 2011-02-01 20:13:57Z karl $
+ 
+% intisaise flow (to ensure fields are aligned)
+%--------------------------------------------------------------------------
+f    = x;
+ 
+% motion of attractor states
+%==========================================================================
+f.a  = spm_lotka_volterra(x.a,v);
+ 
+ 
+% motion of target states
+%==========================================================================
+ 
+% target location is determined by the attractor state softmax(x.a)
+%--------------------------------------------------------------------------
+n    = length(x.a);
+p    = exp(2*x.a);
+t    = P*p/sum(p);
+f.x  = (t - x.x)/8;
+
+ 
+% motion of oculomotor angles (attracted to target)
+%==========================================================================
+t    = atan(x.x);
+f.o  = (t - x.o)/2;
+
+
+
+
+
+
diff --git a/toolbox/DEM/spm_fx_hdm_sck.m b/toolbox/DEM/spm_fx_hdm_sck.m
new file mode 100644
index 0000000..f14fe6e
--- /dev/null
+++ b/toolbox/DEM/spm_fx_hdm_sck.m
@@ -0,0 +1,57 @@
+function [f] = spm_fx_hdm_sck(x,u,P,M)
+% state equation for the hemodynamic model
+% FORMAT [f] = spm_fx_hdm_sck(x,u,P,M)
+% x      - state vector
+%   x(1) - vascular signal                                    s
+%   x(2) - rCBF                                           log(f)
+%   x(3) - venous volume                                  log(v)
+%   x(4) - dHb                                            log(q)
+% u      - input (neuronal activity)                      (u)
+% P      - free parameter vector
+%   P(1) - signal decay                                   d(ds/dt)/ds)
+%   P(2) - autoregulation                                 d(ds/dt)/df)
+%   P(3) - transit time                                   (t0)
+%   P(4) - exponent for Fout(v)                           (alpha)
+%   P(5) - resting oxygen extraction                      (E0)
+%   P(6) - ratio of intra- to extra-vascular components   (epsilon)
+%          of the gradient echo signal   
+%
+%   P(6 + 1:m)   - input efficacies                       d(ds/dt)/du)
+%
+% y      - dx/dt
+%__________________________________________________________________________
+%
+% Ref Buxton RB, Wong EC & Frank LR. Dynamics of blood flow and oxygenation
+% changes during brain activation: The Balloon model. MRM 39:855-864 (1998)
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_fx_hdm_sck.m 4146 2010-12-23 21:01:39Z karl $
+
+% exponentiation of hemodynamic state variables
+%--------------------------------------------------------------------------
+x(2:end,:) = exp(x(2:end,:)); 
+
+% Fout = f(v) - outflow
+%--------------------------------------------------------------------------
+fv(:,:)       = x(3,:).^(1./P(4,:));
+
+% e = f(f) - oxygen extraction
+%--------------------------------------------------------------------------
+ff(:,:)       = (1 - (1 - P(5,:)).^(1./x(2,:)))./P(5,:);
+
+% implement differential state equations
+%--------------------------------------------------------------------------
+f(1,:)     = sum(P(7:end,:),1).*u(:,:) - P(1,:).*x(1,:) - P(2,:).*(x(2,:) - 1);
+f(2,:)     = x(1,:)./x(2,:);
+f(3,:)     = (x(2,:) - fv)./(P(3,:).*x(3,:));
+f(4,:)     = (ff.*x(2,:) - fv.*x(4,:)./x(3,:))./(P(3,:).*x(4,:));
+%f        = f(:);
+% f(2:end,:)=exp(f(2:end,:));
+% f(3:end,:) = f(3:end,:)+1;
+% adjust motion for DEM (that uses time-bins as units of time)
+%--------------------------------------------------------------------------
+try, global dt, f  = f*dt; end
+
+return
\ No newline at end of file
diff --git a/toolbox/DEM/spm_gx_adem_cue.m b/toolbox/DEM/spm_gx_adem_cue.m
new file mode 100644
index 0000000..5920f20
--- /dev/null
+++ b/toolbox/DEM/spm_gx_adem_cue.m
@@ -0,0 +1,27 @@
+function [g]= spm_gx_adem_cue(x,v,a,P)
+% returns the prediction for cued responses (proprioception and vision)
+% FORMAT [g]= spm_gx_adem_cue(x,v,a,P)
+%
+% x    - hidden states:
+%   x.o  - intrinsic motor state (proprioceptive)
+%
+% v    - hidden causes
+%
+% P    - target locations (visual) - extrinsic coordinates (Cartesian)
+%
+% g    - sensations:
+%   g.o  - motor angle (proprioception)
+%   g.p  - finger location (visual)
+%   g.c  - target contrast (visual)
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_gx_adem_cue.m 4230 2011-03-07 20:58:38Z karl $
+ 
+% evaluate positions in intrinsic (polar) coordinates
+%--------------------------------------------------------------------------
+g.o = x.o;
+g.p = tan(x.o);
+g.c = x.a*4;
+
diff --git a/toolbox/DEM/spm_gx_adem_pursuit.m b/toolbox/DEM/spm_gx_adem_pursuit.m
new file mode 100644
index 0000000..a341ceb
--- /dev/null
+++ b/toolbox/DEM/spm_gx_adem_pursuit.m
@@ -0,0 +1,31 @@
+function [g]= spm_gx_adem_pursuit(x,v,a,P)
+% returns the prediction for pursuit model (proprioception and vision)
+% FORMAT [g]= spm_gx_adem_pursuit(x,v,a,P)
+%
+% x    - hidden states:
+%   o(1) - oculomotor angle
+%   o(2) - oculomotor angle
+%   x(1) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x(2) - target location (visual) - extrinsic coordinates (Cartesian)
+%
+% v    - hidden causes
+% P    - parameters
+%
+% g    - sensations:
+%   g(1) - oculomotor angle (proprioception)
+%   g(2) - oculomotor angle (proprioception)
+%   g(3) - target location (visual) - intrinsic coordinates (polar)
+%   g(4) - target location (visual) - intrinsic coordinates (polar)
+% 
+% As for spm_dem_reach but with no visual target
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_gx_adem_pursuit.m 4187 2011-02-01 20:13:57Z karl $
+ 
+% evaluate positions in intrinsic (polar) coordinates
+%--------------------------------------------------------------------------
+x.p = atan(x.x) - x.o;
+g   = [x.o; x.p];
+
diff --git a/toolbox/DEM/spm_gx_dem_cue.m b/toolbox/DEM/spm_gx_dem_cue.m
new file mode 100644
index 0000000..9ceebb3
--- /dev/null
+++ b/toolbox/DEM/spm_gx_dem_cue.m
@@ -0,0 +1,28 @@
+function [g]= spm_gx_dem_cue(x,v,P)
+% returns the prediction for cued responses (proprioception and vision)
+% FORMAT [g]= spm_gx_dem_cue(x,v,P)
+%
+% x    - hidden states:
+%   x.o  - intrinsic motor state (proprioceptive)
+%   x.a  - target salience (attractiveness)
+%
+% v    - hidden causes
+%
+% P.x  - target locations (visual) - extrinsic coordinates (Cartesian)
+%
+% g    - sensations:
+%   g.o  - motor angle (proprioception)
+%   g.p  - finger locations (visual)
+%   g.c  - target contrast  (visual)
+% 
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_gx_dem_cue.m 4230 2011-03-07 20:58:38Z karl $
+ 
+% evaluate positions in intrinsic (polar) coordinates
+%--------------------------------------------------------------------------
+g.o = x.o;
+g.p = tan(x.o);
+g.c = exp(x.a/2);
diff --git a/toolbox/DEM/spm_gx_dem_pursuit.m b/toolbox/DEM/spm_gx_dem_pursuit.m
new file mode 100644
index 0000000..8e0bda8
--- /dev/null
+++ b/toolbox/DEM/spm_gx_dem_pursuit.m
@@ -0,0 +1,33 @@
+function [g] = spm_gx_dem_pursuit(x,v,P)
+% returns the prediction for visual pursuit
+% FORMAT [g] = spm_gx_dem_pursuit(x,v,P)
+%
+% x    - hidden states:
+%   o(1) - oculomotor angle
+%   o(2) - oculomotor angle
+%   x(1) - target location (visual) - extrinsic coordinates (Cartesian)
+%   x(2) - target location (visual) - extrinsic coordinates (Cartesian)
+%   a(:) - attractor (SHC) states
+%
+% v    - hidden causes
+% P    - parameters
+%
+% g    - sensations:
+%   g(1) - oculomotor angle (proprioception)
+%   g(2) - oculomotor angle (proprioception)
+%   g(3) - target location (visual) - intrinsic coordinates (polar)
+%   g(4) - target location (visual) - intrinsic coordinates (polar)
+% 
+% As for spm_dem_reach but with no visual target
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_gx_dem_pursuit.m 4170 2011-01-24 18:37:42Z karl $
+ 
+% evaluate positions in intrinsic (polar) coordinates
+%--------------------------------------------------------------------------
+x.p = atan(x.x) - x.o;
+g   = [x.o; x.p];
+
+
diff --git a/toolbox/DEM/spm_gx_hdm_sck.m b/toolbox/DEM/spm_gx_hdm_sck.m
new file mode 100644
index 0000000..5d67d66
--- /dev/null
+++ b/toolbox/DEM/spm_gx_hdm_sck.m
@@ -0,0 +1,68 @@
+function [y] = spm_gx_hdm_sck(x,u,P,M)
+% Simulated BOLD response to input.
+% FORMAT [y] = spm_gx_hdm_sck(x,u,P,M)
+% y    - BOLD response (%)
+% x    - state vector     (see spm_fx_dcm)
+% P    - Parameter vector (see spm_fx_dcm)
+%__________________________________________________________________________
+%
+% This function implements the BOLD signal model described in: 
+%
+% Stephan KE, Weiskopf N, Drysdale PM, Robinson PA, Friston KJ (2007)
+% Comparing hemodynamic models with DCM. NeuroImage 38: 387-401.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston & Klaas Enno Stephan
+% $Id: spm_gx_hdm_sck.m 4146 2010-12-23 21:01:39Z karl $
+
+
+% biophysical constants for 1.5 T: 
+%==========================================================================
+
+% echo time (seconds)
+%--------------------------------------------------------------------------
+try
+    TE = M(1).TE;
+catch
+    TE = 0.04;
+end
+
+% resting venous volume
+%--------------------------------------------------------------------------
+V0    = 100*0.08;                                
+
+% slope r0 of intravascular relaxation rate R_iv as a function of oxygen 
+% saturation Y:  R_iv = r0*[(1-Y)-(1-Y0)]
+%--------------------------------------------------------------------------
+r0    = 25;
+
+% frequency offset at the outer surface of magnetized vessels
+%--------------------------------------------------------------------------
+nu0   = 40.3;
+
+% region-specific resting oxygen extraction fractions
+%-------------------------------------------------------------------------- 
+E0    = P(5,:); 
+
+% region-specific ratios of intra- to extravascular components of
+% the gradient echo signal (prior mean = 1, log-normally distributed 
+% scaling factor)
+%--------------------------------------------------------------------------
+epsi  = exp(P(6,:));
+ 
+% coefficients in BOLD signal model
+%--------------------------------------------------------------------------
+k1    = 4.3.*nu0.*E0.*TE;
+k2    = epsi.*r0.*E0.*TE;
+k3    = 1 - epsi;
+ 
+% exponentiation of hemodynamic state variables
+%--------------------------------------------------------------------------
+x     = exp(x); 
+
+% BOLD signal
+%--------------------------------------------------------------------------
+v     = x(3,:);
+q     = x(4,:);
+y     = V0.*(k1.*(1 - q) + k2.*(1 - (q./v)) + k3.*(1 - v)); 
diff --git a/toolbox/FieldMap/FieldMap.m b/toolbox/FieldMap/FieldMap.m
index 4a5d47c..9ff2ccb 100644
--- a/toolbox/FieldMap/FieldMap.m
+++ b/toolbox/FieldMap/FieldMap.m
@@ -114,7 +114,7 @@
 % Copyright (C) 2006 Wellcome Department of Imaging Neuroscience
 
 % Jesper Andersson and Chloe Hutton 
-% $Id: FieldMap.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: FieldMap.m 4259 2011-03-22 15:48:59Z chloe $
 %_______________________________________________________________________
 
 persistent PF FS WS PM   % GUI related constants
@@ -1480,6 +1480,7 @@
    mn=min(vol(:));
    mx=max(vol(:));
    vol=-pi+(vol-mn)*2*pi/(mx-mn);
+   V.dt(1)=4;   
    varargout{1} = FieldMap('Write',V,vol,'sc',V.dt(1),V.descrip);
 
 %=======================================================================
diff --git a/toolbox/FieldMap/FieldMap_applyvdm.m b/toolbox/FieldMap/FieldMap_applyvdm.m
new file mode 100644
index 0000000..38f714a
--- /dev/null
+++ b/toolbox/FieldMap/FieldMap_applyvdm.m
@@ -0,0 +1,232 @@
+function out = FieldMap_applyvdm(job)
+% Apply VDM and reslice images 
+% FORMAT FieldMap_applyvdm(job)
+% job.data(sessnum).scans   - images for session/run sessnum
+% job.data(sessnum).vdmfile - VDM file for session/run sessnum      
+% job.roptions.rinterp - interpolation method
+% job.roptions.wrap    - perform warp around in specified dimensions
+% job.roptions.mask    - perform masking
+% job.roptions.which(1) - reslice images in time series only
+% job.roptions.which(2) - reslice images in time series and mean
+% job.roptions.prefix   - prefix for vdm applied files
+% job.roptions.pedir    - phase encode direction (i.e. aplly vdm file along
+% this dimension
+%__________________________________________________________________________
+%
+% A VDM (voxel displacement map) created using the FieldMap toolbox 
+% can be used to resample and reslice realigned images to the original 
+% subdirectory with the same (prefixed) filename. 
+%
+% Voxels in the images will be shifted according to the values in the VDM 
+% file along the direction specified by job.roptions.pedir (i.e. this is
+% usually the phase encode direction) and resliced to the space of the 
+% first image in the time series.
+%
+% Inputs:
+% A job structure containing fields for the input data and the processing
+% options. The input data contains the series of images conforming to 
+% SPM data format (see 'Data Format'), the relative displacement of the images 
+% is stored in their header and a VDM which has (probably) been created 
+% using the FieldMap toolbox and matched to the first image in the time 
+% series (this can also be done via the FieldMap toolbox).
+%
+% Outputs:
+% The resmapled and resliced images resliced to the same subdirectory with a prefix.
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+
+% Chloe Hutton
+% $Id: FieldMap_applyvdm.m 4228 2011-03-04 15:00:15Z chloe $
+
+tiny = 5e-2;
+
+% assemble roptions
+%-----------------------------------------------------------------------
+flags.interp    = job.roptions.rinterp;
+flags.wrap      = job.roptions.wrap;
+flags.mask      = job.roptions.mask;
+flags.which     = job.roptions.which(1);
+flags.mean      = job.roptions.which(2);
+flags.prefix    = job.roptions.prefix;
+flags.pedir     = job.roptions.pedir;
+hold = [repmat(flags.interp,1,3) flags.wrap];
+
+% Determine dimension along which to apply vdm
+applydim=flags.pedir;
+if applydim~=1 &  applydim~=2 & applydim~=3
+    applydim=2;
+end
+
+% Gather up data into ds structure which holds images and vdm file
+%-----------------------------------------------------------------------
+P = {};
+for i = 1:numel(job.data)
+    P{i} = strvcat(job.data(i).scans{:});
+    ds(i).P=spm_vol(P{i});
+    if ~isempty(job.data(i).vdmfile)
+        sfP{i} = job.data(i).vdmfile{1};
+        ds(i).sfP=spm_vol(sfP{i}); 
+    else
+        sfP{i} = [];       
+    end  
+    ds(i).hold = [1 1 1 0 1 0];
+end
+
+ntot = 0;
+for i=1:length(ds)
+   ntot = ntot + length(ds(i).P);
+end
+
+% Set up  x y z for resampling
+%-----------------------------------------------------------------------
+[x,y,z] = ndgrid(1:ds(1).P(1).dim(1),1:ds(1).P(1).dim(2),1:ds(1).P(1).dim(3));
+xyz = [x(:) y(:) z(:) ones(prod(ds(1).P(1).dim(1:3)),1)]; clear x y z;
+
+% Create mask if required (usually default and required to create mean)
+%-----------------------------------------------------------------------
+if flags.mask || flags.mean,
+    spm_progress_bar('Init',ntot,'Computing available voxels',...
+        'volumes completed');
+    
+    if flags.mean
+        Count    = zeros(prod(ds(1).P(1).dim(1:3)),1);
+        Integral = zeros(prod(ds(1).P(1).dim(1:3)),1);
+    end
+    
+    % if flags.mask
+    msk = zeros(prod(ds(1).P(1).dim(1:3)),1);
+    % end
+    
+    % To create mean, read each session specific vdmfile in
+    % to the space of the first image of first session
+    tv = 1;
+    for s=1:length(ds)
+        T = ds(s).sfP.mat\ds(1).P(1).mat;
+        txyz = xyz * T';
+        c = spm_bsplinc(ds(s).sfP,ds(1).hold);
+        ds(s).sfield = spm_bsplins(c,txyz(:,1),txyz(:,2),txyz(:,3),ds(1).hold);
+        ds(s).sfield = ds(s).sfield(:);
+        clear c txyz;
+        sess_msk = zeros(prod(ds(s).P(1).dim(1:3)),1);
+        % Read in each images in space of first image of first session
+        for i = 1:numel(ds(s).P)
+            T = inv(ds(s).P(i).mat) * ds(1).P(1).mat;
+            txyz = xyz * T';
+            txyz(:,applydim) = txyz(:,applydim) + ds(s).sfield;
+            tmp       = false(size(txyz,1),1);
+            if ~flags.wrap(1), tmp = tmp | txyz(:,1) < (1-tiny) | txyz(:,1) > (ds(s).P(i).dim(1)+tiny); end
+            if ~flags.wrap(2), tmp = tmp | txyz(:,2) < (1-tiny) | txyz(:,2) > (ds(s).P(i).dim(2)+tiny); end
+            if ~flags.wrap(3), tmp = tmp | txyz(:,3) < (1-tiny) | txyz(:,3) > (ds(s).P(i).dim(3)+tiny); end
+            sess_msk = sess_msk + real(tmp);
+            spm_progress_bar('Set',tv);
+            tv = tv+1;
+        end
+        msk = msk + sess_msk;
+        if flags.mean, Count = Count + repmat(length(ds(s).P),prod(ds(s).P(1).dim(1:3)),1) - sess_msk; end
+        
+        %
+        % Include static field in estmation of mask.
+        %
+        if isfield(ds(s),'sfP') && ~isempty(ds(s).sfP)
+            T = inv(ds(1).sfP.mat) * ds(1).P(1).mat;
+            txyz = xyz * T';
+            tmp  = false(size(txyz,1),1);
+            if ~flags.wrap(1), tmp = tmp | txyz(:,1) < (1-tiny) | txyz(:,1) > (ds(1).sfP.dim(1)+tiny); end
+            if ~flags.wrap(2), tmp = tmp | txyz(:,2) < (1-tiny) | txyz(:,2) > (ds(1).sfP.dim(2)+tiny); end
+            if ~flags.wrap(3), tmp = tmp | txyz(:,3) < (1-tiny) | txyz(:,3) > (ds(1).sfP.dim(3)+tiny); end
+            msk = msk + real(tmp);
+        end
+
+        if isfield(ds(1),'sfield') && ~isempty(ds(1).sfield)
+            ds(1).sfield = [];
+        end
+    end
+    if flags.mask, msk = find(msk ~= 0); end
+end
+
+% Apply fieldmap to all files, looping through sessions
+%-----------------------------------------------------------------------
+spm_progress_bar('Init',ntot,'Reslicing','volumes completed');
+tv = 1;
+for s = 1:numel(ds)
+    
+    % Get transformation between distortion field and first image
+    T = ds(s).sfP.mat\ds(1).P(1).mat;
+    txyz = xyz * T';
+    c = spm_bsplinc(ds(s).sfP,ds(1).hold);
+    ds(s).sfield = spm_bsplins(c,txyz(:,1),txyz(:,2),txyz(:,3),ds(1).hold);
+    ds(s).sfield = ds(s).sfield(:);
+    
+    % Read in each images in space of first image of first session
+    for i = 1:numel(ds(s).P)
+        T = inv(ds(s).P(i).mat) * ds(1).P(1).mat;
+        txyz = xyz * T';
+        txyz(:,applydim) = txyz(:,applydim) + ds(s).sfield;       
+        c = spm_bsplinc(ds(s).P(i),hold);
+        ima = spm_bsplins(c,txyz(:,1),txyz(:,2),txyz(:,3),hold);
+        %
+        % Write out resliced images
+        %
+        if flags.which
+            PO         = ds(s).P(i);
+            [pth,nm,xt,vr] = spm_fileparts(deblank(PO.fname));
+            PO.fname       = fullfile(pth,[flags.prefix nm xt vr]);
+            PO.mat     = ds(1).P(1).mat;
+            PO.descrip = sprintf('spm - applied vdm');
+            ivol       = ima;
+            if flags.mask
+                ivol(msk) = NaN;
+            end
+            ivol = reshape(ivol,PO.dim(1:3));
+            PO   = spm_create_vol(PO);
+            spm_write_vol(PO,ivol);
+            if nargout > 0
+                  out.sess(s).rfiles{i} = PO.fname;   
+            end
+        end
+        %
+        % Build up mean image if so required.
+        %
+        if flags.mean
+            Integral = Integral + nan2zero(ima);
+        end
+        spm_progress_bar('Set',tv);
+        tv = tv+1;
+    end
+    if isfield(ds(s),'sfield') && ~isempty(ds(s).sfield)
+        ds(s).sfield = [];
+    end
+end
+
+% Write mean image
+%-----------------------------------------------------------
+if flags.mean
+   % Write integral image (16 bit signed)
+   %-----------------------------------------------------------
+   sw = warning('off','MATLAB:divideByZero'); 
+   Integral   = Integral./Count;
+   warning(sw);
+   PO         = ds(1).P(1);
+   [pth,nm,xt,vr] = spm_fileparts(deblank(ds(1).P(1).fname));
+   PO.fname       = fullfile(pth,['mean' flags.prefix nm xt vr]);
+   PO.pinfo   = [max(max(max(Integral)))/32767 0 0]';
+   PO.descrip = 'spm - mean applied vdm image';
+   PO.dt      = [spm_type('int16') spm_platform('bigend')];
+   ivol = reshape(Integral,PO.dim);
+   spm_write_vol(PO,ivol);
+end
+
+if nargout > 0
+   out.rmean{1} = PO.fname;
+end
+
+spm_figure('Clear','Interactive');
+
+%_______________________________________________________________________
+function vo = nan2zero(vi)
+vo = vi;
+vo(~isfinite(vo)) = 0;
+return;
+%_______________________________________________________________________
+
+
diff --git a/toolbox/FieldMap/pm_create_connectogram_dtj.c b/toolbox/FieldMap/pm_create_connectogram_dtj.c
index 04e0980..86d26a0 100644
--- a/toolbox/FieldMap/pm_create_connectogram_dtj.c
+++ b/toolbox/FieldMap/pm_create_connectogram_dtj.c
@@ -1,5 +1,4 @@
 #include "mex.h"
-#include "matrix.h"
 #include <stdlib.h>
 #include <math.h>
 #include <limits.h>
diff --git a/toolbox/FieldMap/pm_defaults.m b/toolbox/FieldMap/pm_defaults.m
index 9a380dc..12bebb1 100644
--- a/toolbox/FieldMap/pm_defaults.m
+++ b/toolbox/FieldMap/pm_defaults.m
@@ -1,23 +1,24 @@
-% SPM5 UPDATE 16/12/06
-% UPDATE 27/01/05
 % Sets the default values for the FieldMap toolbox
 %
 % FORMAT pm_defaults
-%_______________________________________________________________________
+%
+% SPM5 UPDATE 16/12/06
+% UPDATE 27/01/05
+%__________________________________________________________________________
 %
 % This file is intended for site and/or scanner and/or
 % sequence-specific customisations.
-%_______________________________________________________________________
+%__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Chloe Hutton and Jesper Andersson
-% $Id: pm_defaults.m 2616 2009-01-19 16:49:42Z chloe $
+% $Id: pm_defaults.m 4178 2011-01-27 15:12:53Z guillaume $
 
 global pm_def
 
 % Defaults for creating field map. (See pm_make_fieldmap.m and 
 %                                   FieldMap.man for more info.)
-%=======================================================================
+%==========================================================================
 pm_def.INPUT_DATA_FORMAT = 'PM';      % 'RI' = load two real and 
                                       % imaginary image pairs
                                       % 'PM' = load one or two
@@ -26,11 +27,11 @@
 pm_def.SHORT_ECHO_TIME = 10.0;        % Short echo time in ms for Allegra
 pm_def.LONG_ECHO_TIME = 12.46;        % Long echo time in ms for Allegra
 pm_def.MASKBRAIN = 0;                 % Do brain masking (1 or 0,
-                      % 0 for EPI fieldmaps)
+                                      % 0 for EPI fieldmaps)
 
 % Defaults for unwrapping options. (See pm_make_fieldmap.m and 
 %                                   FieldMap.man for more info.)
-%=======================================================================
+%==========================================================================
 pm_def.UNWRAPPING_METHOD = 'Mark3D';  % Unwrapping options are:
                                       % 'Huttonish', 'Mark3D' or 'Mark2D'
 pm_def.FWHM = 10;                     % FWHM of Gaussian filter used to 
@@ -40,71 +41,67 @@
 pm_def.WS = 1;                        % Weighted or normal smoothing.
 
 % Flags for brain extraction
-%=======================================================================
+%==========================================================================
 % Default T1 template for segmentation
 pm_def.MFLAGS.TEMPLATE = fullfile(spm('Dir'),'templates','T1.nii'); 
-pm_def.MFLAGS.FWHM = 5; % In mm - used for smoothing to fill holes in brain mask
-pm_def.MFLAGS.NERODE = 2;% In voxels - used for erosion to remove scalp from brain mask
-pm_def.MFLAGS.NDILATE = 4; % In voxels - used for dilaton to condition scalp removal
+pm_def.MFLAGS.FWHM = 5;     % {mm} used for smoothing to fill holes in brain mask
+pm_def.MFLAGS.NERODE = 2;   % {vox} used for erosion to remove scalp from brain mask
+pm_def.MFLAGS.NDILATE = 4;  % {vox} used for dilation to condition scalp removal
 pm_def.MFLAGS.THRESH = 0.5; % Intensity thresholding for filling holes
-pm_def.MFLAGS.REG = 0.02; % A larger value helps segmentation to converge
+pm_def.MFLAGS.REG = 0.02;   % A larger value helps segmentation to converge
 pm_def.MFLAGS.GRAPHICS = 0; % Don't display segmentation results
 
 % Defaults for converting field map to voxel displacement map.
-%=======================================================================
-pm_def.EPI_BASED_FIELDMAPS = 0;        % EPI=1, other=0.
+%==========================================================================
+pm_def.EPI_BASED_FIELDMAPS = 0;         % EPI=1, other=0.
 pm_def.K_SPACE_TRAVERSAL_BLIP_DIR = -1; % +ve k-space = 1, -ve = -1.
-pm_def.TOTAL_EPI_READOUT_TIME = 21.1;    % Allegra EPI RO time (330E-6*64)
+pm_def.TOTAL_EPI_READOUT_TIME = 21.1;   % Allegra EPI RO time (330E-6*64)
 
 % Defaults for Unwarping.
-%=======================================================================
-pm_def.DO_JACOBIAN_MODULATION = 0;    % Do jacobian modulation to adjust 
-                                      % for compression or stretching
-                                      % No = 0, Yes = 1
-
-                      
-return;
+%==========================================================================
+pm_def.DO_JACOBIAN_MODULATION = 0;      % Do jacobian modulation to adjust 
+                                        % for compression or stretching
+                                        % No = 0, Yes = 1
 
+                                      
 % The rest is dead code that I leave in for "documentation"
 % purposes.  JeAn 161206
 
-
 % FIL specific additions 
-%=======================================================================
-
-
-global SCANNER
-global SEQUENCE
-
-if findstr(SCANNER,'Sonata') & findstr(SEQUENCE,'Siemens')
-   pm_def.TOTAL_EPI_READOUT_TIME = 32;    % Sonata EPI RO time (500E-6*64)
-   pm_def.K_SPACE_TRAVERSAL_BLIP_DIR = -1;
-   if findstr(SEQUENCE,'Siemens')
-      disp('Using Sonata Siemens parameters'); 
-      pm_def.INPUT_DATA_FORMAT = 'PM'; 
-      pm_def.SHORT_ECHO_TIME = 10.0; 
-      pm_def.LONG_ECHO_TIME = 14.76;
-      pm_def.EPI_BASED_FIELDMAPS = 0;
-   else
-      disp('Using Sonata EPI parameters');       
-      pm_def.SHORT_ECHO_TIME = 25;
-      pm_def.LONG_ECHO_TIME = 34.5;
-      pm_def.EPI_BASED_FIELDMAPS = 1;
-   end
-elseif findstr(SCANNER, 'Allegra') 
-   pm_def.TOTAL_EPI_READOUT_TIME = 21.1;  % Allegra EPI RO time (330E-6*64)
-   pm_def.K_SPACE_TRAVERSAL_BLIP_DIR = -1;
-   if findstr(SEQUENCE,'Siemens')
-      disp('Using Allegra Siemens parameters');
-      pm_def.INPUT_DATA_FORMAT = 'PM';
-      pm_def.SHORT_ECHO_TIME = 10.0; % 
-      pm_def.LONG_ECHO_TIME = 12.46;
-      pm_def.EPI_BASED_FIELDMAPS = 0;
-   else
-      disp('Using Allegra EPI parameters');   
-      pm_def.INPUT_DATA_FORMAT = 'RI';
-      pm_def.SHORT_ECHO_TIME = 19;
-      pm_def.LONG_ECHO_TIME = 29;
-      pm_def.EPI_BASED_FIELDMAPS = 1;
-   end
-end
+%==========================================================================
+%
+% global SCANNER
+% global SEQUENCE
+% 
+% if findstr(SCANNER,'Sonata') & findstr(SEQUENCE,'Siemens')
+%    pm_def.TOTAL_EPI_READOUT_TIME = 32;    % Sonata EPI RO time (500E-6*64)
+%    pm_def.K_SPACE_TRAVERSAL_BLIP_DIR = -1;
+%    if findstr(SEQUENCE,'Siemens')
+%       disp('Using Sonata Siemens parameters'); 
+%       pm_def.INPUT_DATA_FORMAT = 'PM'; 
+%       pm_def.SHORT_ECHO_TIME = 10.0; 
+%       pm_def.LONG_ECHO_TIME = 14.76;
+%       pm_def.EPI_BASED_FIELDMAPS = 0;
+%    else
+%       disp('Using Sonata EPI parameters');       
+%       pm_def.SHORT_ECHO_TIME = 25;
+%       pm_def.LONG_ECHO_TIME = 34.5;
+%       pm_def.EPI_BASED_FIELDMAPS = 1;
+%    end
+% elseif findstr(SCANNER, 'Allegra') 
+%    pm_def.TOTAL_EPI_READOUT_TIME = 21.1;  % Allegra EPI RO time (330E-6*64)
+%    pm_def.K_SPACE_TRAVERSAL_BLIP_DIR = -1;
+%    if findstr(SEQUENCE,'Siemens')
+%       disp('Using Allegra Siemens parameters');
+%       pm_def.INPUT_DATA_FORMAT = 'PM';
+%       pm_def.SHORT_ECHO_TIME = 10.0; % 
+%       pm_def.LONG_ECHO_TIME = 12.46;
+%       pm_def.EPI_BASED_FIELDMAPS = 0;
+%    else
+%       disp('Using Allegra EPI parameters');   
+%       pm_def.INPUT_DATA_FORMAT = 'RI';
+%       pm_def.SHORT_ECHO_TIME = 19;
+%       pm_def.LONG_ECHO_TIME = 29;
+%       pm_def.EPI_BASED_FIELDMAPS = 1;
+%    end
+% end
diff --git a/toolbox/FieldMap/pm_defaults_Trio_al_128.m b/toolbox/FieldMap/pm_defaults_Trio_al_128.m
new file mode 100644
index 0000000..ea2b668
--- /dev/null
+++ b/toolbox/FieldMap/pm_defaults_Trio_al_128.m
@@ -0,0 +1,65 @@
+% SPM5 UPDATE 23/11/07
+% Sets the default values for the FieldMap toolbox
+%
+% FORMAT pm_defaults_Trio_eFoV
+%_______________________________________________________________________
+%
+% This file is intended for use with the Siemens fieldmap sequence
+% on the Trio scanner at the AMRIG/FIL and the al_128 EPI sequences with
+% PE blips=-1:
+% 
+%_______________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Chloe Hutton and Jesper Andersson
+% $Id: pm_defaults_Trio_al_128.m 4088 2010-10-11 10:37:42Z chloe $
+
+global pm_def
+
+% Defaults for creating field map. (See pm_make_fieldmap.m and 
+%                                   FieldMap.man for more info.)
+%=======================================================================
+pm_def.INPUT_DATA_FORMAT = 'PM';      % 'RI' = load two real and 
+                                      % imaginary image pairs
+                                      % 'PM' = load one or two
+                                      % phase and magnitude image
+                                      % pairs.
+pm_def.SHORT_ECHO_TIME = 5.0;        % Short echo time in ms for Trio
+pm_def.LONG_ECHO_TIME = 7.46;        % Long echo time in ms for Trio
+pm_def.MASKBRAIN = 1;                 % Do brain masking (1 or 0,
+                      % 0 for EPI fieldmaps)
+
+% Defaults for unwrapping options. (See pm_make_fieldmap.m and 
+%                                   FieldMap.man for more info.)
+%=======================================================================
+pm_def.UNWRAPPING_METHOD = 'Mark3D';  % Unwrapping options are:
+                                      % 'Huttonish', 'Mark3D' or 'Mark2D'
+pm_def.FWHM = 10;                     % FWHM of Gaussian filter used to 
+                                      % implement weighted smoothing of
+                                      % unwrapped maps.
+pm_def.PAD = 0;                       % Size of padding kernel if required.
+pm_def.WS = 1;                        % Weighted or normal smoothing.
+
+% Flags for brain extraction
+%=======================================================================
+pm_def.MFLAGS.TEMPLATE = fullfile(spm('Dir'),'templates','T1.nii');
+pm_def.MFLAGS.FWHM = 5;     % In mm
+pm_def.MFLAGS.NERODE = 2;   % In voxels
+pm_def.MFLAGS.NDILATE = 4;  % In voxels
+pm_def.MFLAGS.THRESH = 0.5;
+pm_def.MFLAGS.REG = 0.02;   % A larger value helps segmentation to converge
+pm_def.MFLAGS.GRAPHICS = 0; % A larger value helps segmentation to converge
+
+% Defaults for converting field map to voxel displacement map.
+%=======================================================================
+pm_def.EPI_BASED_FIELDMAPS = 0;         % EPI=1, other=0.
+pm_def.K_SPACE_TRAVERSAL_BLIP_DIR = -1; % +ve k-space = 1, -ve = -1.
+pm_def.TOTAL_EPI_READOUT_TIME = 51.2;   % Trio al hires EPI 
+
+% Defaults for Unwarping.
+%=======================================================================
+pm_def.DO_JACOBIAN_MODULATION = 0;    % Do jacobian modulation to adjust 
+                                      % for compression or stretching
+                                      % No = 0, Yes = 1
+                      
+                      
diff --git a/toolbox/FieldMap/pm_estimate_ramp.c b/toolbox/FieldMap/pm_estimate_ramp.c
index 29c09ad..800166e 100644
--- a/toolbox/FieldMap/pm_estimate_ramp.c
+++ b/toolbox/FieldMap/pm_estimate_ramp.c
@@ -1,5 +1,4 @@
 #include "mex.h"
-#include "matrix.h"
 #include <stdlib.h>
 #include <math.h>
 #include <limits.h>
diff --git a/toolbox/FieldMap/pm_get_defaults.m b/toolbox/FieldMap/pm_get_defaults.m
index c3d638d..9f80c0c 100644
--- a/toolbox/FieldMap/pm_get_defaults.m
+++ b/toolbox/FieldMap/pm_get_defaults.m
@@ -16,12 +16,13 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Chloe Hutton
-% $Id: pm_get_defaults.m 3449 2009-10-08 15:09:33Z chloe $
+% $Id: pm_get_defaults.m 4228 2011-03-04 15:00:15Z chloe $
 % ---------------------------------------------------------------------
  global pm_def
  if isempty(pm_def)
       pm_defaults;
       pm_def.sessname='session';
+      pm_def.pedir=2;
  end
 
 [default_file_path, tmpname] = fileparts(mfilename('fullpath'));
diff --git a/toolbox/FieldMap/pm_merge_regions.c b/toolbox/FieldMap/pm_merge_regions.c
index c2f8ffe..d77f8e9 100644
--- a/toolbox/FieldMap/pm_merge_regions.c
+++ b/toolbox/FieldMap/pm_merge_regions.c
@@ -1,5 +1,4 @@
 #include "mex.h"
-#include "matrix.h"
 #include <stdlib.h>
 #include <math.h>
 #include <limits.h>
diff --git a/toolbox/FieldMap/pm_pad.c b/toolbox/FieldMap/pm_pad.c
index 740c921..8d7a7b1 100644
--- a/toolbox/FieldMap/pm_pad.c
+++ b/toolbox/FieldMap/pm_pad.c
@@ -1,5 +1,4 @@
 #include "mex.h"
-#include "matrix.h"
 #include <stdlib.h>
 #include <math.h>
 #include <limits.h>
diff --git a/toolbox/FieldMap/pm_restore_ramp.c b/toolbox/FieldMap/pm_restore_ramp.c
index 5ab2ec0..176ed89 100644
--- a/toolbox/FieldMap/pm_restore_ramp.c
+++ b/toolbox/FieldMap/pm_restore_ramp.c
@@ -1,5 +1,4 @@
 #include "mex.h"
-#include "matrix.h"
 #include <stdlib.h>
 #include <math.h>
 #include <limits.h>
diff --git a/toolbox/FieldMap/pm_scale_phase.m b/toolbox/FieldMap/pm_scale_phase.m
index 84103e5..dcf9b24 100644
--- a/toolbox/FieldMap/pm_scale_phase.m
+++ b/toolbox/FieldMap/pm_scale_phase.m
@@ -6,7 +6,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Chloe Hutton
-% $Id: pm_scale_phase.m 1358 2008-04-10 11:20:26Z guillaume $
+% $Id: pm_scale_phase.m 4259 2011-03-22 15:48:59Z chloe $
 
 spm('defaults','FMRI');
 
@@ -17,7 +17,12 @@
 mx=max(vol(:));
 svol=-pi+(vol-mn)*2*pi/(mx-mn);
 
-oV=V;
-name='sc';
-oV.fname=[spm_str_manip(V.fname, 'h') ['/' name] deblank(spm_str_manip(V.fname,'t'))];
+
+% Output image struct
+oV=struct('dim', V.dim(1:3),...
+          'dt',  [4 spm_platform('bigend')],...
+          'mat',     V.mat);
+name='sc';       
+oV.fname=fullfile(spm_str_manip(V.fname, 'h'),[name deblank(spm_str_manip(V.fname,'t'))]);
+oV.descrip='Scaled phase';
 spm_write_vol(oV,svol);
diff --git a/toolbox/FieldMap/pm_seed.m b/toolbox/FieldMap/pm_seed.m
index c29d59b..2eff58d 100644
--- a/toolbox/FieldMap/pm_seed.m
+++ b/toolbox/FieldMap/pm_seed.m
@@ -32,7 +32,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jesper Andersson 
-% $Id: pm_seed.m 1317 2008-04-08 16:16:38Z chloe $
+% $Id: pm_seed.m 4243 2011-03-11 15:38:50Z chloe $
 
 if nargin < 3
    mask = ones(size(angvar));
@@ -54,7 +54,7 @@
            'mat',     M);
 P.dat = double(angvar<(pi^2)/12);
 svol = zeros(size(angvar));
-spm_smooth(P,svol,50);
+spm_smooth(P.dat,svol,50);
 
 %
 % A high value in svol "probably" indicates a
diff --git a/toolbox/FieldMap/pm_segment.m b/toolbox/FieldMap/pm_segment.m
index 233b65d..c49e3ab 100644
--- a/toolbox/FieldMap/pm_segment.m
+++ b/toolbox/FieldMap/pm_segment.m
@@ -52,7 +52,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: pm_segment.m 3756 2010-03-05 18:43:37Z guillaume $
+% $Id: pm_segment.m 4148 2011-01-04 16:49:23Z guillaume $
 
 % Create some suitable default values
 %-----------------------------------------------------------------------
@@ -268,7 +268,7 @@ function display_graphics(VF,VS,mn,cv,mg)
     end;
     VFS(1).pinfo(1:2,:) = VFS(1).pinfo(1:2,:)/spm_global(VFS(1));
 
-    spm_chi2_plot('Init','Affine Registration','Mean squared difference','Iteration');
+    spm_plot_convergence('Init','Affine Registration','Mean squared difference','Iteration');
     flags     = struct('sep',aflags.smosrc, 'regtype',aflags.regtype,'WG',[],'globnorm',0,'debug',0);
     M         = eye(4);
     [M,scal]  = spm_affreg(VG, VFS, flags, M);
@@ -277,7 +277,7 @@ function display_graphics(VF,VS,mn,cv,mg)
 
     flags.sep = aflags.smosrc/2;
     M         = spm_affreg(VG, VFS, flags, M,scal);
-    spm_chi2_plot('Clear');
+    spm_plot_convergence('Clear');
 
 elseif all(size(VG) == [4 4])
     % Assume that second argument is a matrix that will do the job
@@ -351,7 +351,7 @@ function display_graphics(VF,VS,mn,cv,mg)
 %=======================================================================
 function [CP,BP,SP] = run_segment(CP,BP,SP,VF,sums,x1,x2,x3)
 oll = -Inf;
-spm_chi2_plot('Init','Segmenting','Log-likelihood','Iteration #');
+spm_plot_convergence('Init','Segmenting','Log-likelihood','Iteration #');
 
 for iter = 1:64,
     ll= 0;
@@ -367,7 +367,7 @@ function display_graphics(VF,VS,mn,cv,mg)
     end;
 
     BP = update_bp(BP);
-    if iter>1, spm_chi2_plot('Set',ll); end;
+    if iter>1, spm_plot_convergence('Set',ll); end;
     %fprintf('\t%g\n', ll);
 
     % Stopping criterion
@@ -379,7 +379,7 @@ function display_graphics(VF,VS,mn,cv,mg)
     end;
     oll = ll;
 end;
-spm_chi2_plot('Clear');
+spm_plot_convergence('Clear');
 return;
 %=======================================================================
 
diff --git a/toolbox/FieldMap/pm_smooth_phasemap_dtj.c b/toolbox/FieldMap/pm_smooth_phasemap_dtj.c
index 7ef4d3f..9f1c4a3 100644
--- a/toolbox/FieldMap/pm_smooth_phasemap_dtj.c
+++ b/toolbox/FieldMap/pm_smooth_phasemap_dtj.c
@@ -1,5 +1,4 @@
 #include "mex.h"
-#include "matrix.h"
 #include <stdlib.h>
 #include <math.h>
 #include <limits.h>
diff --git a/toolbox/FieldMap/pm_unwrap.m b/toolbox/FieldMap/pm_unwrap.m
index 804539f..a952e04 100644
--- a/toolbox/FieldMap/pm_unwrap.m
+++ b/toolbox/FieldMap/pm_unwrap.m
@@ -79,7 +79,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Jesper Andersson 
-% $Id: pm_unwrap.m 1533 2008-05-01 14:29:03Z spm $
+% $Id: pm_unwrap.m 4175 2011-01-27 12:42:36Z chloe $
 
 %
 % The following are a set of parameters that
@@ -212,6 +212,12 @@
       %
       [irima,cn] = pm_initial_regions(opm,mask,nstep);
       %
+      % Added this little bug fix which prevents pm_merge_regions crashing
+      % because it has too many regions to merge. 
+      while cn>1800 & nstep > 2
+          nstep=nstep-1;
+          [irima,cn] = pm_initial_regions(opm,mask,nstep);
+      end
       % Get connectogram
       %
       [ii,jj,nn,pp] = pm_create_connectogram(irima,opm);
diff --git a/toolbox/FieldMap/tbx_cfg_fieldmap.m b/toolbox/FieldMap/tbx_cfg_fieldmap.m
index c03f3b6..3131afd 100644
--- a/toolbox/FieldMap/tbx_cfg_fieldmap.m
+++ b/toolbox/FieldMap/tbx_cfg_fieldmap.m
@@ -3,7 +3,7 @@
 %_______________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% $Id: tbx_cfg_fieldmap.m 3536 2009-11-05 15:00:32Z chloe $
+% $Id: tbx_cfg_fieldmap.m 4228 2011-03-04 15:00:15Z chloe $
 
 addpath(fullfile(spm('dir'),'toolbox','FieldMap'));
 
@@ -237,7 +237,7 @@
 defaultsfile         = cfg_files;
 defaultsfile.tag     = 'defaultsfile';
 defaultsfile.name    = 'Defaults File';
-defaultsfile.help    = {'Select the ''pm_defaults*.m'' file containing the parameters for the field map data. Please make sure that the parameters defined in the defaults file are correct for your field map and EPI sequence. To create your own defaults file, either edit the distributed version and/or save it with the name ''pm_defaults_yourname.m''.'};
+defaultsfile.help    = {'Select the ''pm_defaults*.m'' file containing the parameters for the field map data. Please make sure that the parameters defined in the defaults file are correct for your field map and EPI sequence. To create your own customised defaults file, either edit the distributed version and/or save it with the name ''pm_defaults_yourname.m''.'};
 defaultsfile.filter  = 'm';
 [deffilepath, tmp]   = fileparts(mfilename('fullpath'));
 defaultsfile.dir     = deffilepath;
@@ -258,7 +258,7 @@
 epi         = cfg_files;
 epi.tag     = 'epi';
 epi.name    = 'Select EPI to Unwarp';
-epi.help    = {'Select an image to distortion correct. The corrected image will be saved with the prefix u. The original and the distortion corrected images can be displayed for comparison.'};
+epi.help    = {'Select a single image to distortion correct. The corrected image will be saved with the prefix u. Note that this option is mainly for quality control of correction so that the original and distortion corrected images can be displayed for comparison. To unwarp multiple images please use either Realign & Unwarp or Apply VDM.'};
 epi.filter = 'image';
 epi.ufilter = '.*';
 epi.num     = [1 1];
@@ -274,9 +274,9 @@
 % generic EPI Sessions
 % ---------------------------------------------------------------------
 generic1         = cfg_repeat;
-generic1.tag     = 'generic';
+generic1.tag     = 'generic1';
 generic1.name    = 'EPI Sessions';
-generic1.help    = {'If a single VDM file will be used for multiple sessions, select the first EPI in each session. A copy of the VDM file will be matched to the first EPI in each session and save with a seprate name.'};
+generic1.help    = {'If a single set of field map data will be used for multiple EPI runs/sessions, select the first EPI in each run/session. A VDM file will created for each run/session, matched to the first EPI in each run/session and saved with a unique name extension.'};
 generic1.values  = {session };
 generic1.num     = [1 Inf];
 % ---------------------------------------------------------------------
@@ -285,9 +285,9 @@
 matchvdm         = cfg_menu;
 matchvdm.tag     = 'matchvdm';
 matchvdm.name    = 'Match VDM to EPI?';
-matchvdm.help    = {'Match VDM file to EPI image. This option will coregister the field map data to the selected EPI before doing distortion correction.'};
+matchvdm.help    = {'Match VDM file to EPI image. This will coregister the field map data to the selected EPI for each run/session.'};
 matchvdm.labels = {
-                   'match vdm'
+                   'match VDM'
                    'none'
 }';
 matchvdm.values{1} = 1;
@@ -297,8 +297,8 @@
 % ---------------------------------------------------------------------
 sessname         = cfg_entry;
 sessname.tag     = 'sessname';
-sessname.name    = 'Name extension for session specific vdm files';
-sessname.help    = {'This will be the name extension followed by an incremented integer for session specific vdm files.'};
+sessname.name    = 'Name extension for run/session specific VDM file';
+sessname.help    = {'This will be the name extension followed by an incremented integer for run/session specific VDM files.'};
 sessname.strtype = 's';
 sessname.num     = [1  Inf];
 sessname.def     = @(val)pm_get_defaults('sessname', val{:});
@@ -308,7 +308,7 @@
 writeunwarped         = cfg_menu;
 writeunwarped.tag     = 'writeunwarped';
 writeunwarped.name    = 'Write unwarped EPI?';
-writeunwarped.help    = {'Write out distortion corrected EPI image. The image is saved with the prefix u.'};
+writeunwarped.help    = {'Write out distortion corrected EPI image. The image is saved with the prefix u. Note that this option is mainly for quality control of correction so that the original and distortion corrected images can be displayed for comparison. To unwarp multiple images please use either Realign & Unwarp or Apply VDM.'};
 writeunwarped.labels = {
                         'write unwarped EPI'
                         'none'
@@ -321,30 +321,30 @@
 anat         = cfg_files;
 anat.tag     = 'anat';
 anat.name    = 'Select anatomical image for comparison';
-anat.help    = {'Select an anatomical image for comparison with the distortion corrected EPI.'};
+anat.help    = {'Select an anatomical image for comparison with the distortion corrected EPI or leave empty. Note that this option is mainly for quality control of correction.'};
 anat.filter = 'image';
 anat.ufilter = '.*';
 anat.num     = [0 1];
+anat.val     = {''};
 % ---------------------------------------------------------------------
 % matchanat Match anatomical image to EPI?
 % ---------------------------------------------------------------------
 matchanat         = cfg_menu;
 matchanat.tag     = 'matchanat';
 matchanat.name    = 'Match anatomical image to EPI?';
-matchanat.help    = {'Match the anatomical image to the distortion corrected EPI. This can be done to allow for visual inspection and comparison of the distortion corrected EPI.'};
+matchanat.help    = {'Match the anatomical image to the distortion corrected EPI. Note that this option is mainly for quality control of correction allowing for visual inspection and comparison of the distortion corrected EPI.'};
 matchanat.labels = {
-                    'match anat'
                     'none'
+                    'match anat'
 }';
-matchanat.values{1} = 1;
-matchanat.values{2} = 0;
-
+matchanat.values{1} = 0;
+matchanat.values{2} = 1;
 % ---------------------------------------------------------------------
 % ---------------------------------------------------------------------
 % Different kinds of input jobs
 % ---------------------------------------------------------------------
 % ---------------------------------------------------------------------
-% Presubtracted phase and magnitude
+% 1) Presubtracted phase and magnitude data
 % ---------------------------------------------------------------------
 % phase Phase Image
 % ---------------------------------------------------------------------
@@ -361,11 +361,10 @@
 magnitude         = cfg_files;
 magnitude.tag     = 'magnitude';
 magnitude.name    = 'Magnitude Image';
-magnitude.help    = {'Select a single magnitude image'};
+magnitude.help    = {'Select a single magnitude image. This is used for masking the phase information and coregistration with the EPI data. If two magnitude images are available, select the one acquired at the shorter echo time because it will have greater signal'};
 magnitude.filter = 'image';
 magnitude.ufilter = '.*';
 magnitude.num     = [1 1];
-
 % ---------------------------------------------------------------------
 % subj Subject
 % ---------------------------------------------------------------------
@@ -395,7 +394,7 @@
 presubphasemag.vout = @vout;
 
 % ---------------------------------------------------------------------
-% Real and imaginary
+% 2) Real and imaginary data
 % ---------------------------------------------------------------------
 % shortreal Short Echo Real Image
 % ---------------------------------------------------------------------
@@ -460,12 +459,12 @@
 realimag.tag     = 'realimag';
 realimag.name    = 'Real and Imaginary Data';
 realimag.val     = {generic };
-realimag.help    = {'Calculate a voxel displacement map (VDM) from real and imaginary field map data. This option expects two real and imaginary pairs of data of two different echo times.'};
+realimag.help    = {'Calculate a voxel displacement map (VDM) from real and imaginary field map data. This option expects two real and imaginary pairs of data of two different echo times. The phase images will be scaled between +/- PI.'};
 realimag.prog = @fieldmap_realimag;
 realimag.vout = @vout;
 
 % ---------------------------------------------------------------------
-% Phase and magnitude (4 files)
+% 3) Phase and magnitude (4 files)
 % ---------------------------------------------------------------------
 % shortphase Short Echo Phase Image
 % ---------------------------------------------------------------------
@@ -535,14 +534,14 @@
 phasemag.vout = @vout;
 
 % ---------------------------------------------------------------------
-% Precalculated fieldmap
+% 4) Precalculated fieldmap
 % ---------------------------------------------------------------------
 % precalcfieldmap Precalculated field map
 % ---------------------------------------------------------------------
 precalcfieldmap1         = cfg_files;
 precalcfieldmap1.tag     = 'precalcfieldmap';
 precalcfieldmap1.name    = 'Precalculated field map';
-precalcfieldmap1.help    = {'Select a precalculated field map. This should be a processed field map (ie phase unwrapped, masked if necessary and scaled to Hz) , for example as generated by the FieldMap toolbox and stored as an fpm_* file.'};
+precalcfieldmap1.help    = {'Select a precalculated field map. This should be a processed field map (ie phase unwrapped, masked if necessary and scaled to Hz) , for example as generated by the FieldMap toolbox and are stored with fpm_* prefix.'};
 precalcfieldmap1.filter = 'image';
 precalcfieldmap1.ufilter = '.*';
 precalcfieldmap1.num     = [1 1];
@@ -556,7 +555,6 @@
 magfieldmap.filter = 'image';
 magfieldmap.ufilter = '.*';
 magfieldmap.num     = [0 1];
-
 % ---------------------------------------------------------------------
 % subj Subject
 % ---------------------------------------------------------------------
@@ -585,6 +583,172 @@
 precalcfieldmap.prog = @fieldmap_precalcfieldmap;
 precalcfieldmap.vout = @vout;
 
+% ---------------------------------------------------------------------
+% 5) Apply vdm* file
+% ---------------------------------------------------------------------
+% scans Images
+% ---------------------------------------------------------------------
+scans         = cfg_files;
+scans.tag     = 'scans';
+scans.name    = 'Images';
+scans.help    = {
+                 'Select scans for this session. These are assumed to be realigned to the first in the time series (e.g. using Realign: Estimate) but do not need to be resliced'
+                 };
+scans.filter = 'image';
+scans.ufilter = '.*';
+scans.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% vdmfile selected vdm files
+% ---------------------------------------------------------------------
+vdmfile         = cfg_files;
+vdmfile.tag     = 'vdmfile';
+vdmfile.name    = 'Fieldmap (vdm* file)';
+vdmfile.help    = {'Select VDM (voxel displacement map) for this session (e.g. created via FieldMap toolbox). This is assumed to be in alignment with the images selected for resampling (note this can be achieved via the FieldMap toolbox).'};
+vdmfile.filter = 'image';
+vdmfile.ufilter = '.*';
+vdmfile.num     = [1 1];
+% ---------------------------------------------------------------------
+% vdmapply session Session
+% ---------------------------------------------------------------------
+data         = cfg_branch;
+data.tag     = 'data';
+data.name    = 'Session';
+data.val     = {scans vdmfile};
+data.help    = {'Data for this session.'};  
+% ---------------------------------------------------------------------
+% generic Data
+% ---------------------------------------------------------------------
+generic2         = cfg_repeat;
+generic2.tag     = 'generic2';
+generic2.name    = 'Data';
+generic2.help    = {'Subjects or sessions for which VDM file is being applied to images.'};
+generic2.values  = {data};
+generic2.num     = [1 Inf];
+
+% ---------------------------------------------------------------------
+% Reslice options for applyvdm
+% ---------------------------------------------------------------------
+% ---------------------------------------------------------------------
+% wrap Wrapping
+% ---------------------------------------------------------------------
+pedir         = cfg_menu;
+pedir.tag     = 'pedir';
+pedir.name    = 'Distortion direction';
+pedir.help    = {
+                 'In which direction are the distortions? Any single dimension can be corrected therefore input data may have been acquired with phase encode directions in Y (most typical), X or Z'};
+pedir.labels = {
+               'Posterior-Anterior (Y)'
+               'Right-Left (X)'
+               'Foot-Head (Z)'}';
+pedir.values = {[2] [1] [3]};
+pedir.def   = @(val)pm_get_defaults('pedir', val{:});
+% ---------------------------------------------------------------------
+%  Which images to reslice?
+% ---------------------------------------------------------------------
+applyvdmwhich         = cfg_menu;
+applyvdmwhich.tag     = 'which';
+applyvdmwhich.name    = 'Reslice which images ?';
+applyvdmwhich.help    = {
+                   'All Images (1..n) '
+                   '  This applies the VDM and reslices all the images. '
+                   'All Images + Mean Image '
+                   '   This applies the VDM reslices all the images and creates a mean of the resliced images.'
+}';
+applyvdmwhich.labels = {
+                  ' All Images (1..n)'
+                  ' All Images + Mean Image'
+}';
+applyvdmwhich.values = {[2 0] [2 1]};
+applyvdmwhich.def    = @(val)spm_get_defaults('realign.write.which', val{:});
+% ---------------------------------------------------------------------
+% rinterp Interpolation
+% ---------------------------------------------------------------------
+rinterp         = cfg_menu;
+rinterp.tag     = 'rinterp';
+rinterp.name    = 'Interpolation';
+rinterp.help    = {'The method by which the images are sampled when being written in a different space. Nearest Neighbour is fastest, but not recommended for image realignment. Bilinear Interpolation is probably OK for PET, but not so suitable for fMRI because higher degree interpolation generally gives better results/* \cite{thevenaz00a,unser93a,unser93b}*/. Although higher degree methods provide better interpolation, but they are slower because they use more neighbouring voxels.'};
+rinterp.labels = {
+                  'Nearest neighbour'
+                  'Trilinear'
+                  '2nd Degree B-spline '
+                  '3rd Degree B-Spline'
+                  '4th Degree B-Spline'
+                  '5th Degree B-Spline '
+                  '6th Degree B-Spline'
+                  '7th Degree B-Spline'
+}';
+rinterp.values = {0 1 2 3 4 5 6 7};
+rinterp.def    = @(val)spm_get_defaults('realign.write.interp', val{:});
+% ---------------------------------------------------------------------
+% wrap Wrapping
+% ---------------------------------------------------------------------
+wrap         = cfg_menu;
+wrap.tag     = 'wrap';
+wrap.name    = 'Wrapping';
+wrap.help    = {
+                'This indicates which directions in the volumes the values should wrap around in.  For example, in MRI scans, the images wrap around in the phase encode direction, so (e.g.) the subject''s nose may poke into the back of the subject''s head. These are typically:'
+                '    No wrapping - for PET or images that have already                   been spatially transformed. Also the recommended option if                   you are not really sure.'
+                '    Wrap in  Y  - for (un-resliced) MRI where phase encoding                   is in the Y direction (voxel space) etc.'
+}';
+wrap.labels = {
+               'No wrap'
+               'Wrap X'
+               'Wrap Y'
+               'Wrap X & Y'
+               'Wrap Z '
+               'Wrap X & Z'
+               'Wrap Y & Z'
+               'Wrap X, Y & Z'
+}';
+wrap.values = {[0 0 0] [1 0 0] [0 1 0] [1 1 0] [0 0 1] [1 0 1] [0 1 1]...
+               [1 1 1]};
+wrap.def    = @(val)spm_get_defaults('realign.write.wrap', val{:});
+% ---------------------------------------------------------------------
+% mask Masking
+% ---------------------------------------------------------------------
+mask         = cfg_menu;
+mask.tag     = 'mask';
+mask.name    = 'Masking';
+mask.help    = {'Because of subject motion, different images are likely to have different patterns of zeros from where it was not possible to sample data. With masking enabled, the program searches through the whole time series looking for voxels which need to be sampled from outside the original images. Where this occurs, that voxel is set to zero for the whole set of images (unless the image format can represent NaN, in which case NaNs are used where possible).'};
+mask.labels = {
+               'Mask images'
+               'Dont mask images'
+}';
+mask.values = {1 0};
+mask.def    = @(val)spm_get_defaults('realign.write.mask', val{:});
+% ---------------------------------------------------------------------
+% prefix Filename Prefix
+% ---------------------------------------------------------------------
+prefix         = cfg_entry;
+prefix.tag     = 'prefix';
+prefix.name    = 'Filename Prefix';
+prefix.help    = {'Specify the string to be prepended to the filenames of the distortion corrected image file(s). Default prefix is ''u''.'};
+prefix.strtype = 's';
+prefix.num     = [1 Inf];
+prefix.def     = @(val)spm_get_defaults('unwarp.write.prefix', val{:});
+% ---------------------------------------------------------------------
+% applyvdmroptions Reslicing Options
+% ---------------------------------------------------------------------
+applyvdmroptions         = cfg_branch;
+applyvdmroptions.tag     = 'roptions';
+applyvdmroptions.name    = 'Reslice Options';
+applyvdmroptions.val     = {pedir applyvdmwhich rinterp wrap mask prefix };
+applyvdmroptions.help    = {'Apply VDM reslice options'};
+% ---------------------------------------------------------------------
+% applyvdm Apply vdm* file to EPI files
+% ---------------------------------------------------------------------
+applyvdm         = cfg_exbranch;
+applyvdm.tag     = 'applyvdm';
+applyvdm.name    = 'Apply VDM ';
+applyvdm.val     = {generic2 applyvdmroptions};
+applyvdm.help = {'Apply VDM (voxel displacement map) to resample voxel values in selected image(s). This allows a VDM to be applied to any images which are assumed to be already realigned (e.g. including EPI fMRI time series and DTI data).' 
+                 'The VDM can be been created from a field map acquisition using the FieldMap toolbox and comprises voxel shift values which describe geometric distortions occuring as a result of magnetic susceptbility artefacts. Distortions along any single dimension can be corrected therefore input data may have been acquired with phase encode directions in X, Y (most typical) and Z.'                      
+                 'The selected images are assumed to be realigned to the first in the time series (e.g. using Realign: Estimate) but do not need to be resliced. The VDM is assumed to be in alignment with the images selected for resampling (note this can be achieved via the FieldMap toolbox). The resampled images are written to the input subdirectory with the same (prefixed) filename.'
+                 'e.g. The typical processing steps for fMRI time series would be 1) Realign: Estimate, 2) FieldMap to create VDM, 3) Apply VDM.'
+                 'Note that this routine is a general alternative to using the VDM in combination with Realign & Unwarp which estimates and corrects for the combined effects of static and movement-related susceptibility induced distortions. Apply VDM can be used when dynamic distortions are not (well) modelled by Realign & Unwarp (e.g. for fMRI data acquired with R->L phase-encoding direction, high field fMRI data or DTI data).'
+                 }';                
+applyvdm.prog = @fieldmap_applyvdm;
+applyvdm.vout = @vout_applyvdm;
 % ---------------------------------------------------------------------
 % fieldmap FieldMap
 % ---------------------------------------------------------------------
@@ -592,11 +756,10 @@
 fieldmap.tag     = 'fieldmap';
 fieldmap.name    = 'FieldMap';
 fieldmap.help    = {
-                    'The FieldMap toolbox generates unwrapped field maps which are converted to voxel displacement maps (VDM) that can be used to unwarp geometrically distorted EPI images. For references and an explantion of the theory behind the field map based unwarping, see FieldMap_principles.man. The resulting VDM files are saved with the prefix vdm and can be used in combination with Realign & Unwarp to calculate and correct for the combined effects of static and movement-related susceptibility induced distortions.'
+                    'The FieldMap toolbox generates unwrapped field maps which are converted to voxel displacement maps (VDM) that can be used to unwarp geometrically distorted EPI images. For references and an explantion of the theory behind the field map based unwarping, see FieldMap_principles.man. The resulting VDM files are saved with the prefix vdm and can be applied to images using Apply VDM or in combination with Realign & Unwarp to calculate and correct for the combined effects of static and movement-related susceptibility induced distortions.'
                     ''
 }';
-fieldmap.values  = {presubphasemag realimag phasemag precalcfieldmap };
-
+fieldmap.values  = {presubphasemag realimag phasemag precalcfieldmap applyvdm};
 %------------------------------------------------------------------------
 function out=fieldmap_presubphasemag(job)
 for i=1:numel(job.subj),
@@ -617,7 +780,9 @@
 for i=1:numel(job.subj),
    out(i)=FieldMap_Run(job.subj(i));
 end
-
+%------------------------------------------------------------------------
+function out=fieldmap_applyvdm(job)
+   out=FieldMap_applyvdm(job);
 %------------------------------------------------------------------------
 function dep = vout(job)
 
@@ -631,6 +796,31 @@
    depnum=depnum+1;
  end
 end;
+%------------------------------------------------------------------------
+function dep = vout_applyvdm(job)
+for k=1:numel(job.data)   
+    if job.roptions.which(1) > 0
+        cdep(1)            = cfg_dep;
+        cdep(1).sname      = sprintf('VDM corrected images (Sess %d)', k);
+        cdep(1).src_output = substruct('.','sess', '()',{k}, '.','rfiles');
+        cdep(1).tgt_spec   = cfg_findspec({{'filter','image','strtype','e'}});
+    end;
+    if k == 1
+        dep = cdep;
+    else
+        dep = [dep cdep];
+    end;
+end
+if ~strcmp(job.roptions.which,'<UNDEFINED>') && job.roptions.which(2),
+    if exist('dep','var')
+        dep(end+1) = cfg_dep;
+    else
+        dep = cfg_dep;
+    end;
+    dep(end).sname      = 'Mean Image';
+    dep(end).src_output = substruct('.','rmean');
+    dep(end).tgt_spec   = cfg_findspec({{'filter','image','strtype','e'}});
+end;
 
 
 
diff --git a/toolbox/HDW/spm_hdw.m b/toolbox/HDW/spm_hdw.m
index 9b5868a..923bf94 100644
--- a/toolbox/HDW/spm_hdw.m
+++ b/toolbox/HDW/spm_hdw.m
@@ -8,7 +8,7 @@ function spm_hdw(job)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_hdw.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_hdw.m 4148 2011-01-04 16:49:23Z guillaume $
 
 for i=1:numel(job.data),
     run_warping(job.data(i).mov{1},job.data(i).ref{1},job.warp_opts,job.bias_opts);
@@ -143,7 +143,7 @@ function run_warping(PF,PG,warp_opts,bias_opts)
 Tbias   = zeros(d3);
 
 ll = Inf;
-spm_chi2_plot('Init','Bias Correction','- Log-likelihood','Iteration');
+spm_plot_convergence('Init','Bias Correction','- Log-likelihood','Iteration');
 for subit=1:nits,
 
     % Compute objective function and its 1st and second derivatives
@@ -193,7 +193,7 @@ function run_warping(PF,PG,warp_opts,bias_opts)
         Beta  = Beta  + kron(b3,spm_krutil(wt1,B1bias,B2bias,0));
         Alpha = Alpha + kron(b3*b3',spm_krutil(wt2,B1bias,B2bias,1));
     end;
-    spm_chi2_plot('Set',ll/prod(d));
+    spm_plot_convergence('Set',ll/prod(d));
 
 
     if subit > 1 && ll>oll,
@@ -227,7 +227,7 @@ function save_def(Def,mat,fname)
 % Save a deformation field as an image
 
 dim   = [size(Def{1},1) size(Def{1},2) size(Def{1},3) 1 3];
-dtype = 'FLOAT32-BE';
+dtype = 'FLOAT32';
 off   = 0;
 scale = 1;
 inter = 0;
diff --git a/toolbox/MEEGtools/spm_MEEGtools.m b/toolbox/MEEGtools/spm_MEEGtools.m
index 84761f3..3f13c30 100644
--- a/toolbox/MEEGtools/spm_MEEGtools.m
+++ b/toolbox/MEEGtools/spm_MEEGtools.m
@@ -4,11 +4,12 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_MEEGtools.m 3767 2010-03-09 22:49:30Z vladimir $
+% $Id: spm_MEEGtools.m 4286 2011-04-04 13:35:04Z vladimir $
 
 
 funlist = {
     'Copy MEG sensors', 'spm_eeg_copygrad';
+    'Transform EEG cap', 'spm_eeg_transform_cap';
     'Re-reference EEG', 'spm_eeg_reref_eeg';
     'Split conditions into separate datasets', 'spm_eeg_split_conditions';
     'Fieldtrip interactive plotting', 'spm_eeg_plot_interactive';
@@ -18,6 +19,7 @@
     'Define spatial confounds' , 'spm_eeg_spatial_confounds'
     'Correct sensor data',        'spm_eeg_correct_sensor_data'
     'Use CTF head localization' , 'spm_eeg_megheadloc'
+    'Fix CTF head position data' ,'spm_eeg_fix_ctf_headloc'
     'Fieldtrip manual coregistration' , 'spm_eeg_ft_datareg_manual'
     'Remove spikes from EEG' , 'spm_eeg_remove_spikes'
     'Reduce jumps in MEG data' , 'spm_eeg_remove_jumps'
@@ -31,6 +33,8 @@
     'Detect eyeblinks',  'spm_eeg_detect_eyeblinks'
     'Relabel trials for epoched CTF datasets', 'spm_eeg_recode_epoched_ctf'
     'Correct TMS artefact', 'spm_eeg_tms_correct'
+    'Plot scalp maps from M/EEG image', 'spm_eeg_img2maps'
+    'Continuous data power', 'spm_eeg_cont_power'
     };
 
 str = sprintf('%s|', funlist{:, 1});
diff --git a/toolbox/MEEGtools/spm_eeg_cont_power.m b/toolbox/MEEGtools/spm_eeg_cont_power.m
new file mode 100644
index 0000000..e8afb74
--- /dev/null
+++ b/toolbox/MEEGtools/spm_eeg_cont_power.m
@@ -0,0 +1,105 @@
+function D = spm_eeg_cont_power(S)
+% Compute power of continuous M/EEG data
+% FORMAT D = spm_eeg_cont_power(S)
+%
+% This function computes power from band-pass filtered data using hilbert
+% transform. Can also be used as a template fof any kind of computation on
+% continuous data.
+%
+% S           - input structure (optional)
+% (optional) fields of S:
+%   S.D       - MEEG object or filename of M/EEG mat-file
+%
+% D           - MEEG object (also written to disk)
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+
+% Vladimir Litvak
+% $Id: spm_eeg_cont_power.m 4286 2011-04-04 13:35:04Z vladimir $
+
+SVNrev = '$Rev: 4286 $';
+
+%-Startup
+%--------------------------------------------------------------------------
+spm('FnBanner', mfilename, SVNrev);
+spm('FigName','M/EEG power'); spm('Pointer', 'Watch');
+
+if nargin == 0
+    S = [];
+end
+
+%-Get MEEG object
+%--------------------------------------------------------------------------
+try
+    D = S.D;
+catch
+    [D, sts] = spm_select(1, 'mat', 'Select M/EEG mat file');
+    if ~sts, D = []; return; end
+    S.D = D;
+end
+
+D = spm_eeg_load(D);
+
+%
+%--------------------------------------------------------------------------
+
+% generate new meeg object with new filenames
+Dnew = clone(D, ['h' fnamedat(D)], [D.nchannels D.nsamples D.ntrials]);
+
+% determine channels to process
+chanind = D.meegchannels;
+
+
+% continuous data
+spm_progress_bar('Init', nchannels(D), 'Channels processed'); drawnow;
+if nchannels(D) > 100, Ibar = floor(linspace(1, nchannels(D),100));
+else Ibar = [1:nchannels(D)]; end
+
+% work on blocks of channels
+% determine blocksize
+% determine block size, dependent on memory
+memsz  = spm('Memory');
+datasz = nchannels(D)*nsamples(D)*8; % datapoints x 8 bytes per double value
+blknum = ceil(datasz/memsz);
+blksz  = ceil(nchannels(D)/blknum);
+blknum = ceil(nchannels(D)/blksz);
+
+% now filter blocks of channels
+chncnt=1;
+for blk=1:blknum
+    % load old meeg object blockwise into workspace
+    blkchan=chncnt:(min(nchannels(D), chncnt+blksz-1));
+    if isempty(blkchan), break, end
+    tempdata=D(blkchan,:,1);
+    chncnt=chncnt+blksz;
+    %loop through channels
+    for j = 1:numel(blkchan)
+        
+        if ismember(blkchan(j), chanind)
+            % ********** this is actually the processing part
+            tempdata(j, :) = abs(spm_hilbert(tempdata(j,:)));
+            % **********
+        end
+        
+        if ismember(j, Ibar), spm_progress_bar('Set', blkchan(j)); end
+        
+    end
+    
+    % write tempdata to Dnew
+    Dnew(blkchan,:,1)=tempdata;
+    clear tempdata;   
+end;
+
+
+spm_progress_bar('Clear');
+
+%-Save new evoked M/EEG dataset
+%--------------------------------------------------------------------------
+D = Dnew;
+D = D.history(mfilename, S);
+save(D);
+
+%-Cleanup
+%--------------------------------------------------------------------------
+spm('FigName','M/EEG power: done'); spm('Pointer', 'Arrow');
+
diff --git a/toolbox/MEEGtools/spm_eeg_detect_eyeblinks.m b/toolbox/MEEGtools/spm_eeg_detect_eyeblinks.m
index 7e80793..d6ae1ab 100644
--- a/toolbox/MEEGtools/spm_eeg_detect_eyeblinks.m
+++ b/toolbox/MEEGtools/spm_eeg_detect_eyeblinks.m
@@ -15,9 +15,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Laurence Hunt
-% $Id: spm_eeg_detect_eyeblinks.m 3858 2010-04-30 20:56:20Z vladimir $
+% $Id: spm_eeg_detect_eyeblinks.m 4265 2011-03-28 13:18:31Z vladimir $
 
-SVNrev = '$Rev: 3858 $';
+SVNrev = '$Rev: 4265 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -25,10 +25,11 @@
 spm('FigName','Eyeblink detect'); spm('Pointer','Watch');
 
 
-%-Test for the presence of required Matlab toolbox
+%-Test for the presence of required Matlab toolbox - not needed anymore, 
+% using ft_preproc_bandpassfilter instead
 %--------------------------------------------------------------------------
 if ~license('test','signal_toolbox')
-    error('Signal Processing Toolbox is required for eyeblink detection.');
+%    error('Signal Processing Toolbox is required for eyeblink detection.');
 end
 
 %-Get MEEG object
@@ -73,7 +74,7 @@
 try 
     stdthresh = S.stdthresh;
 catch
-    stdthresh = 3;
+    stdthresh = 4;
 end
 
 if ~isfield(S, 'overwrite')
@@ -88,11 +89,11 @@
 eog_data = D(eogchan,:,:);
 
 %% filter data at 1-15Hz (eyeblink duration typically 100-300ms) and demean
-eog_filt = detrend(ft_preproc_bandpassfilter(eog_data, D.fsample, [1 15], 1001, 'fir'), 'constant');
+eog_filt = detrend(ft_preproc_bandpassfilter(eog_data, D.fsample, [1 15], 4, 'but'), 'constant');
 
 %% find eye-movements
 
-sd_eeg = std(eog_filt);
+sd_eeg=(spm_percentile(eog_filt,85)-spm_percentile(eog_filt,15))/2; %robust estimate of standard deviation, suggested by Mark Woolrich
 em_thresh = stdthresh*sd_eeg;
 
 %% find 'spikes' (putative eyeblinks):
@@ -115,17 +116,26 @@
     spikemat(:,i) = eog_filt(spikes(i)-eblength+1:spikes(i)+eblength);
 end
 
-%reject spikes whose peak is not within 0.5 s.d. of the mean (gets rid of most artefacts
+%reject spikes whose peak is not within 1 s.d. of the mean (gets rid of most artefacts
 %    etc. not removed by filtering):
 mn_spike = mean(spikemat(eblength,:));
-sd_spike = 0.5*std(spikemat(eblength,:));
+sd_spike = std(spikemat(eblength,:));
 spikes(spikemat(eblength,:)>mn_spike+sd_spike | ...
        spikemat(eblength,:)<mn_spike-sd_spike) = [];
 spikemat(:,find(spikemat(eblength,:)>mn_spike+sd_spike | ...
        spikemat(eblength,:)<mn_spike-sd_spike)) = [];
 
 disp(['Number of putative eyeblinks detected: ' num2str(length(spikes))]);
-          
+       
+num_eb_per_min=60*length(spikes)/(D.time(end)-D.time(1));
+disp([num2str(num_eb_per_min) ' eye-blinks per minute '])
+if (num_eb_per_min<0.5)
+    error(['Only ' num2str(num_eb_per_min) ' eye-blinks per minute detected by algorithm. Try a lower threshold.'])
+end
+if (num_eb_per_min>60)
+    error(['As many as ' num2str(num_eb_per_min) ' eye-blinks per minute detected by algorithm. Try a higher threshold.'])
+end
+
 % plot
 %----------------------------------------------------------------------
 Fgraph = spm_figure('GetWin','Graphics');
diff --git a/toolbox/MEEGtools/spm_eeg_fix_ctf_headloc.m b/toolbox/MEEGtools/spm_eeg_fix_ctf_headloc.m
new file mode 100644
index 0000000..50571df
--- /dev/null
+++ b/toolbox/MEEGtools/spm_eeg_fix_ctf_headloc.m
@@ -0,0 +1,500 @@
+function D = spm_eeg_fix_ctf_headloc(S)
+% Fix head localization data in a continuous CTF dataset with continuous
+% head localization. The tracking has to be valid at least some of the time
+%
+% The functionality requires the original CTF header (read with CTF toolbox)
+% to be present (set S.saveorigheader = 1 at conversion).
+%
+% FORMAT D = spm_eeg_fix_ctf_headloc(S)
+%
+% S         - struct (optional)
+% (optional) fields of S:
+% S.D - meeg object or filename
+%
+%
+% Output:
+% D         - MEEG data struct or cell array of MEEG objects with the
+%             rejected trials set to bad and sensors corrected (if
+%             requested).
+%
+% Disclaimer: this code is provided as an example and is not guaranteed to work
+% with data on which it was not tested. If it does not work for you, feel
+% free to improve it and contribute your improvements to the MEEGtools toolbox
+% in SPM (http://www.fil.ion.ucl.ac.uk/spm)
+%
+% _______________________________________________________________________
+% Copyright (C) 2008 Institute of Neurology, UCL
+
+% Vladimir Litvak, Robert Oostenveld
+% $Id: spm_eeg_fix_ctf_headloc.m 4250 2011-03-15 14:50:04Z vladimir $
+
+
+[Finter,Fgraph,CmdLine] = spm('FnUIsetup','Fix CTF head locations',0);
+
+if nargin == 0
+    S = [];
+end
+
+try
+    D = S.D;
+catch
+    D = spm_select(1, '\.mat$', 'Select M/EEG mat file');
+    S.D = D;
+end
+
+D = spm_eeg_load(S.D);
+
+
+%read HLC-channels
+%HLC0011 HLC0012 HLC0013 x, y, z coordinates of nasion-coil in m.
+%HLC0021 HLC0022 HLC0023 x, y, z coordinates of lpa-coil in m.
+%HLC0031 HLC0032 HLC0033 x, y, z coordinates of rpa-coil in m.
+hlc_chan_label = {'HLC0011' 'HLC0012' 'HLC0013'...
+    'HLC0021' 'HLC0022' 'HLC0023'...
+    'HLC0031' 'HLC0032' 'HLC0033'};
+
+if ~all(ismember(hlc_chan_label, D.chanlabels))
+    error('Head localization is not supported for this data')
+end
+
+if ~isfield(D, 'origheader')
+    error('Original CTF header needs to be present')
+end
+
+if ~isfield(S, 'quickfix')
+    
+    %read HLC-channels
+    %HLC0011 HLC0012 HLC0013 x, y, z coordinates of nasion-coil in m.
+    %HLC0021 HLC0022 HLC0023 x, y, z coordinates of lpa-coil in m.
+    %HLC0031 HLC0032 HLC0033 x, y, z coordinates of rpa-coil in m.
+    hlc_chan_label = {'HLC0011' 'HLC0012' 'HLC0013'...
+        'HLC0021' 'HLC0022' 'HLC0023'...
+        'HLC0031' 'HLC0032' 'HLC0033'};
+    
+    
+    tmpdat  = D(D.indchannel(hlc_chan_label), :);
+    
+    cont_fid  = permute(reshape(tmpdat', [], 3, 3), [1 3 2]);
+    
+    if isfield(S, 'valid_fid')
+        valid_fid = S.valid_fid;
+        
+        dist_dev = [
+            (squeeze(sqrt(sum((cont_fid(:, 1, :) - cont_fid(:, 2, :)).^2, 3))) - norm(valid_fid(1,:) - valid_fid(2,:)))';...
+            (squeeze(sqrt(sum((cont_fid(:, 2, :) - cont_fid(:, 3, :)).^2, 3))) - norm(valid_fid(2,:) - valid_fid(3,:)))';...
+            (squeeze(sqrt(sum((cont_fid(:, 3, :) - cont_fid(:, 1, :)).^2, 3))) - norm(valid_fid(3,:) - valid_fid(1,:)))'];
+        
+        W = abs(dist_dev) < 0.01;
+    else
+        %%
+        dist = [
+            squeeze(sqrt(sum((cont_fid(:, 1, :) - cont_fid(:, 2, :)).^2, 3)))';...
+            squeeze(sqrt(sum((cont_fid(:, 2, :) - cont_fid(:, 3, :)).^2, 3)))';...
+            squeeze(sqrt(sum((cont_fid(:, 3, :) - cont_fid(:, 1, :)).^2, 3)))' ];
+        %%
+        [y, W]= spm_robust_average(dist, 2, 3);
+        %%
+        
+        W = W>0.9;
+    end
+    
+    % Theoretically one should use 'or' here and not 'and' but I found it
+    % leaves too much rubbish in
+    nasOK = W(1, :) & W(3, :);
+    leOK =  W(1, :) & W(2, :);
+    reOK =  W(2, :) & W(3, :);
+    
+    OK = [nasOK;nasOK;nasOK; leOK; leOK;leOK; reOK;reOK;reOK];
+    
+    if min(sum(OK, 2))<2
+        error('Not enough valid head localization data');
+    end
+    %%
+    for i = 1:size(tmpdat, 1)
+        if any(~OK(i, :))
+            tmpdat(i, :) = interp1(D.time(OK(i, :)), tmpdat(i, OK(i, :)),  D.time, 'linear', 'extrap');
+        end
+    end
+    D(D.indchannel(hlc_chan_label), :) = tmpdat;
+    
+    dewarfid = 100*reshape(median(tmpdat, 2), 3, 3);
+    %%
+    D.origheader.hc.dewar = dewarfid;
+    
+    spm_figure('GetWin','Graphics');clf;
+    subplot(2, 1, 1);
+    plot(D.time, tmpdat);
+    subplot(2, 1, 2);
+else
+    spm_figure('GetWin','Graphics');clf;
+    dewarfid = D.origheader.hc.dewar;
+end
+
+dewargrad = ctf2grad(D.origheader, 1);
+
+M = spm_eeg_inv_headcoordinates(dewarfid(:, 1), dewarfid(:, 2), dewarfid(:, 3));
+
+grad = ft_convert_units(ft_transform_sens(M, dewargrad), 'mm');
+
+D = sensors(D, 'MEG', grad);
+
+fid = D.fiducials;
+fid.pnt = [];
+fid.fid.pnt = dewarfid';
+fid.unit = 'cm';
+
+fid = ft_convert_units(ft_transform_headshape(M, fid), 'mm');
+
+D = fiducials(D, fid);
+
+save(D);
+
+plot3(grad.pnt(:, 1), grad.pnt(:, 2), grad.pnt(:, 3), 'k.');
+axis equal off
+hold on
+plot3(fid.fid.pnt(:, 1), fid.fid.pnt(:, 2), fid.fid.pnt(:, 3), 'r.', 'MarkerSize', 15);
+
+
+function [grad] = ctf2grad(hdr, dewar)
+
+% CTF2GRAD converts a CTF header to a gradiometer structure that can be
+% understood by FieldTrip and Robert Oostenveld's low-level forward and
+% inverse routines. The fieldtrip/fileio read_header function can use three
+% different implementations of the low-level code for CTF data. Each of
+% these implementations is dealt with here.
+%
+% See also READ_HEADER, FIF2GRAD, BTI2GRAD, YOKOGAWA2GRAD
+
+% undocumented option: it will return the gradiometer information in dewar
+% coordinates if second argument is present and non-zero
+
+% Copyright (C) 2004, Robert Oostenveld
+%
+% This file is part of FieldTrip, see http://www.ru.nl/neuroimaging/fieldtrip
+% for the documentation and details.
+%
+%    FieldTrip is free software: you can redistribute it and/or modify
+%    it under the terms of the GNU General Public License as published by
+%    the Free Software Foundation, either version 3 of the License, or
+%    (at your option) any later version.
+%
+%    FieldTrip is distributed in the hope that it will be useful,
+%    but WITHOUT ANY WARRANTY; without even the implied warranty of
+%    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+%    GNU General Public License for more details.
+%
+%    You should have received a copy of the GNU General Public License
+%    along with FieldTrip. If not, see <http://www.gnu.org/licenses/>.
+%
+% $Id: spm_eeg_fix_ctf_headloc.m 4250 2011-03-15 14:50:04Z vladimir $
+
+% My preferred ordering in the grad structure is:
+%   1st 151 coils are bottom coils of MEG channels
+%   2nd 151 are the top coils of MEG channels
+%   following coils belong to reference channels
+
+if nargin<2 || isempty(dewar)
+    dewar = 0;
+end
+
+if isfield(hdr, 'orig')
+    hdr = hdr.orig; % use the original CTF header, not the FieldTrip header
+end
+
+% start with empty gradiometer
+grad = [];
+grad.pnt = [];
+grad.ori = [];
+grad.tra = [];
+
+% meg channels are 5, refmag 0, refgrad 1, ADCs 18
+% UPPT001 is 11
+% UTRG001 is 11
+% SCLK01  is 17
+% STIM    is 11
+% SCLK01  is 17
+% EEG057  is 9
+% ADC06   is 18
+% ADC07   is 18
+% ADC16   is 18
+% V0      is 15
+
+if isfield(hdr, 'res4') && isfield(hdr.res4, 'senres')
+    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+    % the header was read using the CTF p-files, i.e. readCTFds
+    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+    
+    sensType  = [hdr.res4.senres.sensorTypeIndex];
+    selMEG    = find(sensType==5);
+    selREF    = find(sensType==0 | sensType==1);
+    selMEG    = selMEG(:)';
+    selREF    = selREF(:)';
+    numMEG    = length(selMEG);
+    numREF    = length(selREF);
+    
+    % determine the number of channels and coils
+    coilcount = 0;
+    coilcount = coilcount + sum([hdr.res4.senres(selREF).numCoils]);
+    coilcount = coilcount + sum([hdr.res4.senres(selMEG).numCoils]);
+    chancount = numMEG + numREF;
+    % preallocate the memory
+    grad.pnt = zeros(coilcount, 3);         % this will hold the position of each coil
+    grad.ori = zeros(coilcount, 3);         % this will hold the orientation of each coil
+    grad.tra = zeros(chancount, coilcount); % this describes how each coil contributes to each channel
+    
+    % combine the bottom and top coil of each MEG channel
+    for i=1:numMEG
+        n = selMEG(i);
+        % get coil positions and orientations of this channel (max. 8)
+        if dewar
+            pos = hdr.res4.senres(n).pos0';
+            ori = hdr.res4.senres(n).ori0';
+        else
+            pos = hdr.res4.senres(n).pos';
+            ori = hdr.res4.senres(n).ori';
+        end
+        if hdr.res4.senres(n).numCoils~=2
+            error('unexpected number of coils in MEG channel');
+        end
+        % add the coils of this channel to the gradiometer array
+        grad.pnt(i       ,:) = pos(1,:);
+        grad.pnt(i+numMEG,:) = pos(2,:);
+        grad.ori(i       ,:) = ori(1,:) .* -sign(hdr.res4.senres(n).properGain);
+        grad.ori(i+numMEG,:) = ori(2,:) .* -sign(hdr.res4.senres(n).properGain);
+        grad.tra(i,i       ) = 1;
+        grad.tra(i,i+numMEG) = 1;
+    end
+    
+    % the MEG channels always have 2 coils, the reference channels vary in the number of coils
+    chancount = 1*numMEG;
+    coilcount = 2*numMEG;
+    
+    % combine the coils of each reference channel
+    for i=1:numREF
+        n = selREF(i);
+        % get coil positions and orientations of this channel (max. 8)
+        if dewar
+            pos = hdr.res4.senres(n).pos0';
+            ori = hdr.res4.senres(n).ori0';
+        else
+            pos = hdr.res4.senres(n).pos';
+            ori = hdr.res4.senres(n).ori';
+        end
+        % determine the number of coils for this channel
+        numcoils = hdr.res4.senres(n).numCoils;
+        % add the coils of this channel to the gradiometer array
+        chancount = chancount+1;
+        for j=1:numcoils
+            coilcount = coilcount+1;
+            grad.pnt(coilcount, :)         = pos(j,:);
+            grad.ori(coilcount, :)         = ori(j,:) .* -sign(hdr.res4.senres(n).properGain);
+            grad.tra(chancount, coilcount) = 1;
+        end
+    end
+    
+    label = cellstr(hdr.res4.chanNames);
+    for i=1:numel(label)
+        % remove the site-specific numbers from each channel name, e.g. 'MZC01-1706' becomes 'MZC01'
+        label{i} = strtok(label{i}, '-');
+    end
+    
+    grad.label = label([selMEG selREF]);
+    grad.unit  = 'cm';
+    
+    % convert the balancing coefficients into a montage that can be used with the ft_apply_montage function
+    if isfield(hdr.BalanceCoefs, 'G1BR')
+        meglabel          = label(hdr.BalanceCoefs.G1BR.MEGlist);
+        reflabel          = label(hdr.BalanceCoefs.G1BR.Refindex);
+        nmeg              = length(meglabel);
+        nref              = length(reflabel);
+        montage.labelorg  = cat(1, meglabel, reflabel);
+        montage.labelnew  = cat(1, meglabel, reflabel);
+        montage.tra       = [eye(nmeg, nmeg), -hdr.BalanceCoefs.G1BR.alphaMEG'; zeros(nref, nmeg), eye(nref, nref)];
+        grad.balance.G1BR = montage;
+    end
+    
+    if isfield(hdr.BalanceCoefs, 'G2BR')
+        meglabel          = label(hdr.BalanceCoefs.G2BR.MEGlist);
+        reflabel          = label(hdr.BalanceCoefs.G2BR.Refindex);
+        nmeg              = length(meglabel);
+        nref              = length(reflabel);
+        montage.labelorg  = cat(1, meglabel, reflabel);
+        montage.labelnew  = cat(1, meglabel, reflabel);
+        montage.tra       = [eye(nmeg, nmeg), -hdr.BalanceCoefs.G2BR.alphaMEG'; zeros(nref, nmeg), eye(nref, nref)];
+        grad.balance.G2BR = montage;
+    end
+    
+    if isfield(hdr.BalanceCoefs, 'G3BR')
+        meglabel          = label(hdr.BalanceCoefs.G3BR.MEGlist);
+        reflabel          = label(hdr.BalanceCoefs.G3BR.Refindex);
+        nmeg              = length(meglabel);
+        nref              = length(reflabel);
+        montage.labelorg  = cat(1, meglabel, reflabel);
+        montage.labelnew  = cat(1, meglabel, reflabel);
+        montage.tra       = [eye(nmeg, nmeg), -hdr.BalanceCoefs.G3BR.alphaMEG'; zeros(nref, nmeg), eye(nref, nref)];
+        grad.balance.G3BR = montage;
+    end
+    
+    if isfield(hdr.BalanceCoefs, 'G3AR')
+        meglabel          = label(hdr.BalanceCoefs.G3AR.MEGlist);
+        reflabel          = label(hdr.BalanceCoefs.G3AR.Refindex);
+        nmeg              = length(meglabel);
+        nref              = length(reflabel);
+        montage.labelorg  = cat(1, meglabel, reflabel);
+        montage.labelnew  = cat(1, meglabel, reflabel);
+        montage.tra       = [eye(nmeg, nmeg), -hdr.BalanceCoefs.G3AR.alphaMEG'; zeros(nref, nmeg), eye(nref, nref)];
+        grad.balance.G3AR = montage;
+    end
+    
+    if     all([hdr.res4.senres(selMEG).grad_order_no]==0)
+        grad.balance.current = 'none';
+    elseif all([hdr.res4.senres(selMEG).grad_order_no]==1)
+        grad.balance.current = 'G1BR';
+    elseif all([hdr.res4.senres(selMEG).grad_order_no]==2)
+        grad.balance.current = 'G2BR';
+    elseif all([hdr.res4.senres(selMEG).grad_order_no]==3)
+        grad.balance.current = 'G3BR';
+    elseif all([hdr.res4.senres(selMEG).grad_order_no]==13)
+        grad.balance.current = 'G3AR';
+    else
+        warning('cannot determine balancing of CTF gradiometers');
+        grad = rmfield(grad, 'balance');
+    end
+    
+    % sofar the gradiometer definition was the ideal, non-balenced one
+    if isfield(grad, 'balance') && ~strcmp(grad.balance.current, 'none')
+        % apply the current balancing parameters to the gradiometer definition
+        grad = ft_apply_montage(grad, getfield(grad.balance, grad.balance.current));
+    end
+    
+    
+elseif isfield(hdr, 'sensType') && isfield(hdr, 'Chan')
+    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+    % the header was read using the open-source matlab code that originates from CTF and that was modified by the FCDC
+    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+    
+    selMEG = find(hdr.sensType==5);
+    selREF = find(hdr.sensType==0 | hdr.sensType==1);
+    selMEG = selMEG(:)';
+    selREF = selREF(:)';
+    numMEG = length(selMEG);
+    numREF = length(selREF);
+    
+    % combine the bottom and top coil of each MEG channel
+    for i=1:numMEG
+        n = selMEG(i);
+        % get coil positions and orientations of this channel (max. 8)
+        if dewar
+            pos = cell2mat({hdr.Chan(n).coil.pos}');
+            ori = cell2mat({hdr.Chan(n).coil.ori}');
+        else
+            pos = cell2mat({hdr.Chan(n).coilHC.pos}');
+            ori = cell2mat({hdr.Chan(n).coilHC.ori}');
+        end
+        % determine the number of coils for this channel
+        numcoils = sum(sum(pos.^2, 2)~=0);
+        if numcoils~=2
+            error('unexpected number of coils in MEG channel');
+        end
+        % add the coils of this channel to the gradiometer array
+        grad.pnt(i       ,:) = pos(1,:);
+        grad.pnt(i+numMEG,:) = pos(2,:);
+        grad.ori(i       ,:) = ori(1,:) .* -sign(hdr.gainV(n));
+        grad.ori(i+numMEG,:) = ori(2,:) .* -sign(hdr.gainV(n));
+        grad.tra(i,i       ) = 1;
+        grad.tra(i,i+numMEG) = 1;
+    end
+    
+    % combine the coils of each reference channel
+    for i=1:numREF
+        n = selREF(i);
+        % get coil positions and orientations of this channel (max. 8)
+        if dewar
+            pos = cell2mat({hdr.Chan(n).coil.pos}');
+            ori = cell2mat({hdr.Chan(n).coil.ori}');
+        else
+            pos = cell2mat({hdr.Chan(n).coilHC.pos}');
+            ori = cell2mat({hdr.Chan(n).coilHC.ori}');
+        end
+        % determine the number of coils for this channel
+        numcoils = sum(sum(pos.^2, 2)~=0);
+        % add the coils of this channel to the gradiometer array
+        for j=1:numcoils
+            grad.pnt(numMEG+i, :)     = pos(j,:);
+            grad.ori(numMEG+i, :)     = ori(j,:) .* -sign(hdr.gainV(n));
+            grad.tra(numMEG+i, 2*numMEG+i) = 1;
+        end
+    end
+    
+    grad.label = hdr.label([selMEG selREF]);
+    grad.unit  = 'cm';
+    
+    
+elseif isfield(hdr, 'sensor') && isfield(hdr.sensor, 'info')
+    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+    % the header was read using the CTF importer from the NIH and Daren Weber
+    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+    
+    if dewar
+        % this does not work for Daren Webers implementation
+        error('cannot return the gradiometer definition in dewar coordinates');
+    end
+    
+    % only work on the MEG channels
+    if isfield(hdr.sensor.index, 'meg')
+        sel = hdr.sensor.index.meg;
+    else
+        sel = hdr.sensor.index.meg_sens;
+    end
+    
+    for i=1:length(sel)
+        pnt = hdr.sensor.info(sel(i)).location';
+        ori = hdr.sensor.info(sel(i)).orientation';
+        numcoils(i) = size(pnt,1);
+        if size(ori,1)==1 && size(pnt,1)==1
+            % one coil position with one orientation: magnetometer
+            ori = ori;
+        elseif size(ori,1)==1 && size(pnt,1)==2
+            % two coil positions with one orientation: first order gradiometer
+            % assume that the orientation of the upper coil is opposite to the lower coil
+            ori = [ori; -ori];
+        else
+            error('do not know how to deal with higher order gradiometer hardware')
+        end
+        
+        % add this channels coil positions and orientations
+        grad.pnt = [grad.pnt; pnt];
+        grad.ori = [grad.ori; ori];
+        grad.label{i} = hdr.sensor.info(sel(i)).label;
+        
+        % determine the contribution of each coil to each channel's output signal
+        if size(pnt,1)==1
+            % one coil, assume that the orientation is correct, i.e. the weight is +1
+            grad.tra(i,end+1) = 1;
+        elseif size(pnt,1)==2
+            % two coils, assume that the orientation for each coil is correct,
+            % i.e. the weights are +1 and +1
+            grad.tra(i,end+1) = 1;
+            grad.tra(i,end+1) = 1;
+        else
+            error('do not know how to deal with higher order gradiometer hardware')
+        end
+    end
+    
+    % prefer to have the labels in a column vector
+    grad.label = grad.label(:);
+    
+    % reorder the coils, such that the bottom coils are at the first N
+    % locations and the top coils at the last N positions. This makes it
+    % easier to use a selection of the coils for topographic plotting
+    if all(numcoils==2)
+        bot = 1:2:sum(numcoils);
+        top = 2:2:sum(numcoils);
+        grad.pnt = grad.pnt([bot top], :);
+        grad.ori = grad.ori([bot top], :);
+        grad.tra = grad.tra(:, [bot top]);
+    end
+    
+else
+    error('unknown header to contruct gradiometer definition');
+end
diff --git a/toolbox/MEEGtools/spm_eeg_img2maps.m b/toolbox/MEEGtools/spm_eeg_img2maps.m
new file mode 100644
index 0000000..66bb95b
--- /dev/null
+++ b/toolbox/MEEGtools/spm_eeg_img2maps.m
@@ -0,0 +1,117 @@
+function spm_eeg_img2maps(S)
+% Make a series of scalp maps from data in an image
+% FORMAT  spm_eeg_img2maps(S)
+%
+% S         - input structure (optional)
+% (optional) fields of S:
+%    D              - M/EEG dataset containing the sensor locations
+%    image          - file name of an image containing M/EEG data in voxel-space
+%    window         - start and end of a window in peri-stimulus time [ms]
+%    clim           - color limits of the plot
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Vladimir Litvak
+% $Id: spm_eeg_img2maps.m 4204 2011-02-21 12:36:08Z vladimir $
+
+SVNrev = '$Rev: 4204 $';
+
+%-Startup
+%--------------------------------------------------------------------------
+spm('FnBanner', mfilename, SVNrev);
+spm('FigName','Plot scalp maps');
+
+if nargin == 0
+    S = [];
+end
+
+%% -Input parameters
+%--------------------------------------------------------------------------
+if ~isfield(S, 'image')
+    [S.image, sts] = spm_select(1, 'image', 'Select an M/EEG image (in voxel-space)');
+    if ~sts, return; end
+end
+
+%-Get MEEG object
+%--------------------------------------------------------------------------
+try
+    D = S.D;
+catch
+    [D, sts] = spm_select(1, 'mat', 'Select M/EEG mat file');
+    if ~sts, D = []; return; end
+    S.D = D;
+end
+
+D = spm_eeg_load(D);
+
+if ~isfield(S, 'window')
+    S.window = spm_input('start and end of window [ms or Hz]', '+1', 'r', '', 2);
+end
+
+V = spm_vol(S.image);
+Y = spm_read_vols(V);
+
+Nt = size(Y, 3);
+
+begsample = inv(V.mat)*[0 0 S.window(1) 1]';
+begsample = begsample(3);
+
+endsample = inv(V.mat)*[0 0 S.window(2) 1]';
+endsample = endsample(3);
+
+if any([begsample endsample] < 0) || ...
+        any([begsample endsample] > Nt)
+    error('The window is out of limits for the image.');
+end
+
+[junk begsample] = min(abs(begsample-[1:Nt]));
+[junk endsample] = min(abs(endsample-[1:Nt]));
+
+Y = squeeze(mean(Y(: , :, begsample:endsample), 3));
+
+if any(diff(size(Y)))
+    error('The image should be square');
+else
+    n = size(Y, 1);
+end
+
+%-Get channel indices and coordinates
+%--------------------------------------------------------------------------
+[Cel, Cind] = spm_eeg_locate_channels(D, n, 1);
+
+modality    = spm_eeg_modality_ui(D, 1, 1);
+
+goodchan    = find(ismember(Cind, setdiff(D.meegchannels(modality), D.badchannels)));
+
+Cel         = Cel(goodchan, :);
+Cind        = Cind(goodchan);
+
+if isempty(Cind)
+    error('No good channels to plot');
+end
+
+Y = Y(sub2ind(size(Y), Cel(:, 1), Cel(:, 2)));
+
+
+%%
+% SPM graphics figure
+%--------------------------------------------------------------------------
+Fgraph  = spm_figure('GetWin','Graphics'); spm_figure('Clear',Fgraph)
+
+if ~isfield(S, 'clim')
+    in.max = max(abs(Y));
+    in.min = -in.max;
+else
+    in.min = S.clim(1);
+    in.max = S.clim(2);
+end
+
+in.type = modality;
+in.f = Fgraph;
+in.ParentAxes = axes;
+
+spm_eeg_plotScalpData(Y, D.coor2D(Cind), D.chanlabels(Cind), in);
+%%
+%-Cleanup
+%--------------------------------------------------------------------------
+spm('FigName','Plot scalp maps: done');
diff --git a/toolbox/MEEGtools/spm_eeg_reref_eeg.m b/toolbox/MEEGtools/spm_eeg_reref_eeg.m
index 7e34411..978fa43 100644
--- a/toolbox/MEEGtools/spm_eeg_reref_eeg.m
+++ b/toolbox/MEEGtools/spm_eeg_reref_eeg.m
@@ -18,10 +18,10 @@
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
-% Rik Henson
-% $Id: spm_eeg_reref_eeg.m 3701 2010-01-28 13:01:40Z rik $
+% Rik Henson (updated so that bad channels included in tra but not used)
+% $Id: spm_eeg_reref_eeg.m 4119 2010-11-11 12:12:00Z vladimir $
 
-SVNrev = '$Rev: 3701 $';
+SVNrev = '$Rev: 4119 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -42,43 +42,47 @@
 
 % Get indices for just EEG channels and remove any bad channels
 %--------------------------------------------------------------------------
-eegchan = setdiff(D.meegchannels('EEG'), D.badchannels);
+eegchan  = D.meegchannels('EEG');
+goodchan = setdiff(eegchan, D.badchannels);
 
-if isempty(eegchan)
+if isempty(goodchan)
     error('No EEG channels that are not marked as bad...!?')
 end
 
 %-Get reference channel indices 
 %--------------------------------------------------------------------------
 if ~isfield(S, 'refchan')
-      [selection, ok]= listdlg('ListString', D.chanlabels(eegchan), 'SelectionMode', 'multiple' ,'Name', 'Select reference channels' , 'ListSize', [400 300]);
+      [selection, ok]= listdlg('ListString', D.chanlabels(goodchan), 'SelectionMode', 'multiple' ,'Name', 'Select reference channels' , 'ListSize', [400 300]);
     if ~ok
         return;
     end
     
-    if length(selection) == length(eegchan)
+    if length(selection) == length(goodchan)
         S.refchan = 'average';
     else
-        S.refchan = D.chanlabels(eegchan(selection));
+        S.refchan = D.chanlabels(goodchan(selection));
     end
 end
 
 if strcmpi(S.refchan,'average')
-    refchan = eegchan;
+    refchan = goodchan;
 elseif iscell(S.refchan)
     refchan = D.indchannel(S.refchan);
 elseif isnumeric(S.refchan)
     refchan = S.refchan;
 end
 
-refind = find(ismember(eegchan, refchan));
+refind  = find(ismember(eegchan, refchan));
+goodind = find(ismember(eegchan, goodchan));
+badind  = find(ismember(eegchan, D.badchannels));
 
 if length(refind) ~= length(refchan)
     error('Not all S.refchan are valid indices for (non-bad) EEG channels')
 end
 
-tra           = eye(length(eegchan));
-tra(:,refind) = tra(:,refind) - 1/length(refchan);
+tra                 = eye(length(eegchan));
+tra(goodind,refind) = tra(goodind,refind) - 1/length(refchan);
+tra(badind,refind)  = tra(badind,refind)  - 1/length(refchan);
 
 S1=[];
 S1.D = D;
diff --git a/toolbox/MEEGtools/spm_eeg_spatial_confounds.m b/toolbox/MEEGtools/spm_eeg_spatial_confounds.m
index 3509c9e..890e421 100644
--- a/toolbox/MEEGtools/spm_eeg_spatial_confounds.m
+++ b/toolbox/MEEGtools/spm_eeg_spatial_confounds.m
@@ -10,10 +10,10 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Vladimir Litvak
-% $Id: spm_eeg_spatial_confounds.m 3833 2010-04-22 14:49:48Z vladimir $
+% $Id: spm_eeg_spatial_confounds.m 4132 2010-12-02 17:32:37Z vladimir $
 
 
-SVNrev = '$Rev: 3833 $';
+SVNrev = '$Rev: 4132 $';
 
 %-Startup
 %--------------------------------------------------------------------------
@@ -163,7 +163,7 @@
             sconf = [];
             sconf.label = D.chanlabels(D.meegchannels);
             sconf.coeff = nan(length(sconf.label), S.ncomp);
-            sconf.coeff(sel1, :) = U(sel2, 1:S.ncomp);
+            sconf.coeff(sel1, :) = U(sel2, 1:min(S.ncomp, size(U, 2)));
             sconf.bad = ones(length(sconf.label), 1);
             sconf.bad(sel1, :) = 0;
             D = sconfounds(D, sconf);
diff --git a/toolbox/MEEGtools/spm_eeg_transform_cap.m b/toolbox/MEEGtools/spm_eeg_transform_cap.m
new file mode 100644
index 0000000..8a4b06e
--- /dev/null
+++ b/toolbox/MEEGtools/spm_eeg_transform_cap.m
@@ -0,0 +1,99 @@
+function elec = spm_eeg_transform_cap(S)
+% Transform an electrode cap to match the subject's headshape
+% FORMAT shape = spm_eeg_transform_cap(S)
+%
+% S                    - input structure (optional)
+% (optional) fields of S:
+%   S.standard         - headshape (file) with the standard locations
+%   S.custom           - headshape (file) with individually measured locations
+%   S.outfile          - file name to save the output                     
+%
+% Output:
+%   sens               - transformed sensors
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Vladimir Litvak
+% $Id: spm_eeg_transform_cap.m 4020 2010-07-28 12:41:23Z vladimir $
+
+SVNrev = '$Rev: 4020 $';
+
+%-Startup
+%--------------------------------------------------------------------------
+spm('FnBanner', mfilename, SVNrev);
+spm('FigName','Transform cap'); spm('Pointer','Watch');
+
+
+%-Input
+%--------------------------------------------------------------------------
+if nargin == 0
+    S = [];
+end
+
+if ~isfield(S, 'standard')
+    [S.standard, sts] = spm_select(1, '.*', 'Select standard locations file');
+    if ~sts, return; end
+end
+
+if ischar(S.standard)
+    S.standard = ft_read_sens(S.standard);
+end
+
+if ~isfield(S, 'custom')
+    [S.custom, sts] = spm_select(1, '.*', 'Select individual locations file');
+    if ~sts, return; end
+end
+
+if ischar(S.custom)
+    S.custom = ft_read_sens(S.custom);
+end
+
+S.standard = ft_convert_units(S.standard, 'mm');
+S.custom = ft_convert_units(S.custom, 'mm');
+
+%-Compute
+%--------------------------------------------------------------------------
+S1 = [];
+S1.targetfid.pnt = [];
+S1.targetfid.fid = S.custom;
+
+S1.sourcefid.pnt = [];
+S1.sourcefid.fid = S.standard;
+
+S1.template = 1;
+
+M1 = spm_eeg_inv_datareg(S1);
+
+elec = ft_transform_sens(M1, S.standard);
+
+%-Change SPM fiducial labels if present to prevent recognition as template.
+%--------------------------------------------------------------------------
+[sel1, sel2] = spm_match_str(elec.label, {'spmnas', 'spmlpa', 'spmrpa'});
+fidlabels = {'nas', 'lpa', 'rpa'};
+elec.label(sel1) = fidlabels(sel2);
+
+%-Save
+%--------------------------------------------------------------------------
+if ~isfield(S, 'outfile')
+    [f, p] = uiputfile( ...
+        {'*.mat', 'MATLAB File (*.mat)'}, 'Save custom sensors as');
+    S.outfile = fullfile(p, f);
+end
+
+save(S.outfile, 'elec');
+
+%-Plot
+%--------------------------------------------------------------------------
+spm_figure('GetWin','Graphics');clf
+ind = spm_match_str(elec.label, S.custom.label);
+plot3(S.custom.pnt(:,1), S.custom.pnt(:,2), S.custom.pnt(:,3), '.r', 'MarkerSize', 20);
+hold on
+plot3(elec.pnt(:,1), elec.pnt(:,2), elec.pnt(:,3), '.k');
+plot3(elec.pnt(ind, 1), elec.pnt(ind ,2), elec.pnt(ind ,3), '.g', 'MarkerSize', 20);
+axis equal off
+rotate3d on
+
+%-Cleanup
+%--------------------------------------------------------------------------
+spm_progress_bar('Clear');
+spm('FigName','Transform cap: done'); spm('Pointer','Arrow');
diff --git a/toolbox/Neural_Models/spm_csd_demo.m b/toolbox/Neural_Models/spm_csd_demo.m
index 35580df..e4d6f19 100644
--- a/toolbox/Neural_Models/spm_csd_demo.m
+++ b/toolbox/Neural_Models/spm_csd_demo.m
@@ -48,7 +48,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_csd_demo.m 2374 2008-10-21 18:52:29Z karl $
+% $Id: spm_csd_demo.m 4281 2011-03-31 19:49:57Z karl $
  
 clear global
 clear
@@ -146,7 +146,7 @@
  
 % invert
 %--------------------------------------------------------------------------
-[Ep,Cp,S,F] = spm_nlsi_GN(M,[],Y);
+[Ep,Cp,Eh,F] = spm_nlsi_GN(M,[],Y);
  
 
 spm_demo_proceed
diff --git a/toolbox/Neural_Models/spm_demo_proceed.m b/toolbox/Neural_Models/spm_demo_proceed.m
index 1f996c9..cbcc52b 100644
--- a/toolbox/Neural_Models/spm_demo_proceed.m
+++ b/toolbox/Neural_Models/spm_demo_proceed.m
@@ -8,7 +8,7 @@ function spm_demo_proceed(tag,str)
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_demo_proceed.m 1380 2008-04-11 18:55:18Z karl $
+% $Id: spm_demo_proceed.m 4136 2010-12-09 22:22:28Z guillaume $
 
 % get figure
 %--------------------------------------------------------------------------
diff --git a/toolbox/Neural_Models/spm_fx_lfp.m b/toolbox/Neural_Models/spm_fx_lfp.m
index 23d8c3d..b8d328b 100644
--- a/toolbox/Neural_Models/spm_fx_lfp.m
+++ b/toolbox/Neural_Models/spm_fx_lfp.m
@@ -39,7 +39,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_fx_lfp.m 1207 2008-03-13 20:57:56Z karl $
+% $Id: spm_fx_lfp.m 4096 2010-10-22 19:40:34Z karl $
 
 % check if intrinsic connections are free parameters
 %--------------------------------------------------------------------------
@@ -60,7 +60,16 @@
 H    = [8 32];                 % receptor densities (excitatory, inhibitory)
 T    = [4 16];                 % synaptic constants (excitatory, inhibitory)
 R    = [1 2];                  % parameters of static nonlinearity
- 
+
+% [specified] fixed parameters
+%--------------------------------------------------------------------------
+try, E  = M.pF.E; end
+try, G  = M.pF.G; end
+try, D  = M.pF.D; end
+try, H  = M.pF.H; end
+try, T  = M.pF.T; end
+try, R  = M.pF.R; end
+
 % exponential transform to ensure positivity constraints
 %--------------------------------------------------------------------------
 A{1} = exp(P.A{1})*E(1);
diff --git a/toolbox/Neural_Models/spm_lfp_mtf.m b/toolbox/Neural_Models/spm_lfp_mtf.m
index 19d2cf2..eea0aa6 100644
--- a/toolbox/Neural_Models/spm_lfp_mtf.m
+++ b/toolbox/Neural_Models/spm_lfp_mtf.m
@@ -14,7 +14,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_lfp_mtf.m 3119 2009-05-13 10:53:31Z rosalyn $
+% $Id: spm_lfp_mtf.m 4096 2010-10-22 19:40:34Z karl $
 
 
 % compute log-spectral density
@@ -58,13 +58,12 @@
 
 % spectrum of innovations (Gu)
 %--------------------------------------------------------------------------
-Gu   = exp(P.a)*f.^(-1)*2;                % spectral density of (AR) input
-Gu   = Gu + exp(P.b);                     % spectral density of IID input
+AR   = f.^-1;                                     % pink component
+ID   = f.^0;                                      % white component
+Gu   = AR*exp(P.a(1))   + ID*exp(P.a(2))/4;       % neuronal innovations
+Gn   = AR*exp(P.b(1))   + ID*exp(P.b(2))/8;       % channel noise (non-specific)
+Gs   = AR*exp(P.c(1,:)) + ID*exp(P.c(2,:))/8;     % channel noise (specific)
 
-% channel noise (specific and non-specific)
-%--------------------------------------------------------------------------
-Gs   = (exp(P.c(1))*f.^(-1) + exp(P.d(1)))/8;
-Gn   = (exp(P.c(2))*f.^(-1) + exp(P.d(2)))/16;
 
 
 % trial-specific effects
@@ -74,6 +73,7 @@
 
 % cycle over trials
 %--------------------------------------------------------------------------
+GS     = 0;
 for  c = 1:size(X,1)
 
     % basline parameters
@@ -87,7 +87,7 @@
         Q.A{2} = Q.A{2} + X(c,i)*P.B{i};         % backward  connections
         Q.A{3} = Q.A{3} + X(c,i)*P.B{i};         % lateral   connections
        try
-            Q.H = Q.H + X(c,i)*diag(P.B{i});      % intrinsic connections
+            Q.H = Q.H + X(c,i)*diag(P.B{i});     % intrinsic connections
         catch
             Q.G = Q.G + X(c,i)*diag(P.B{i});
         end
@@ -121,27 +121,49 @@
                 Si       = fft(K1(:,i,k));
                 Sj       = fft(K1(:,j,k));
                 Gij      = Si.*conj(Sj);
-                Gij      = abs(Gij([1:N/2] + 1)).*Gu;
+                Gij      = Gij([1:N/2] + 1).*Gu;
                 G(:,i,j) = G(:,i,j) + Gij;
             end
 
-            % cross-spectral density from channel noise
+        end        
+    end
+
+    % save trial-specific frequencies of interest
+    %----------------------------------------------------------------------
+    y{c} = G(If,:,:);
+
+end
+
+% and add channel noise
+%--------------------------------------------------------------------------
+for c = 1:length(y)
+    G     = y{c};
+    for i = 1:nc
+        for j = i:nc
+
+            % cross-spectral density from common channel noise
+            %--------------------------------------------------------------
+            G(:,i,j) = G(:,i,j) + Gn(If);
+            
+            % and channel specific noise
             %--------------------------------------------------------------
-            G(:,i,j) = G(:,i,j) + Gn;            % common noise
             if i == j
-                G(:,i,i) = G(:,i,i) + Gs;        % and channel specific
-
+                try
+                    G(:,i,i) = G(:,i,i) + Gs(If,i);
+                catch
+                    G(:,i,i) = G(:,i,i) + Gs(If,1);
+                end
             else
+                
                 % fill in lower half of CSD matrix
-                %------------------------------------------------------
+                %----------------------------------------------------------
                 G(:,j,i) = G(:,i,j);
+                
             end
         end
     end
-
-    % save frequencies of interest
-    %----------------------------------------------------------------------
-    y{c} = G(If,:,:);
-
+    y{c} = abs(G);
 end
+    
+
 
diff --git a/toolbox/Neural_Models/spm_lfp_priors.m b/toolbox/Neural_Models/spm_lfp_priors.m
index c92542a..256d5a5 100644
--- a/toolbox/Neural_Models/spm_lfp_priors.m
+++ b/toolbox/Neural_Models/spm_lfp_priors.m
@@ -1,4 +1,4 @@
-function [pE,pC] = spm_lfp_priors(A,B,C)
+function [E,V] = spm_lfp_priors(A,B,C)
 % prior moments for a neural mass model of ERPs
 % FORMAT [pE,pC] = spm_lfp_priors(A,B,C)
 %
@@ -23,7 +23,7 @@
 %
 %--------------------------------------------------------------------------
 %
-% pC - prior covariances: cov(spm_vec(pE))
+% pC - prior (co)variances
 %
 % Because priors are specified under log normal assumptions, most
 % parameters are simply scaling coefficients with a prior expectation
@@ -38,7 +38,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_lfp_priors.m 2374 2008-10-21 18:52:29Z karl $
+% $Id: spm_lfp_priors.m 4096 2010-10-22 19:40:34Z karl $
  
 % defaults
 %--------------------------------------------------------------------------
@@ -94,14 +94,6 @@
 %--------------------------------------------------------------------------
 E.D    = sparse(n,n);     V.D = Q/16;              % extrinsic delays
 E.I    = 0;               V.I = 1/32;              % intrinsic delays
- 
-% prior covariances
-%--------------------------------------------------------------------------
-pE     = E;
-pC     = diag(sparse(spm_vec(V)));
- 
-% prior moments if two arguments
-%--------------------------------------------------------------------------
 warning('on','MATLAB:log:logOfZero');
 
 
diff --git a/toolbox/Neural_Models/spm_mtf_demo.m b/toolbox/Neural_Models/spm_mtf_demo.m
index 04e6bda..5a12a84 100644
--- a/toolbox/Neural_Models/spm_mtf_demo.m
+++ b/toolbox/Neural_Models/spm_mtf_demo.m
@@ -40,7 +40,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_mtf_demo.m 2374 2008-10-21 18:52:29Z karl $
+% $Id: spm_mtf_demo.m 4281 2011-03-31 19:49:57Z karl $
  
 clear global
 clear
@@ -109,7 +109,7 @@
  
 % invert
 %--------------------------------------------------------------------------
-[Ep,Cp,S,F] = spm_nlsi_GN(M,[],Y);
+[Ep,Cp,Eh,F] = spm_nlsi_GN(M,[],Y);
  
 % plot spectral density 
 %--------------------------------------------------------------------------
diff --git a/toolbox/Seg/Makefile b/toolbox/Seg/Makefile
new file mode 100644
index 0000000..1377294
--- /dev/null
+++ b/toolbox/Seg/Makefile
@@ -0,0 +1,25 @@
+#!/usr/bin/env make -f
+# Seg Makefile called by {SPM}/src/Makefile
+#
+# Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+#
+# $Id: Makefile 4037 2010-08-06 14:52:27Z guillaume $
+
+include ../../src/Makefile.var
+
+SPMMEX = spm_mrf.$(SUF)
+
+all: $(SPMMEX)
+
+clean:
+
+distclean: clean
+	$(DEL) $(SPMMEX)
+
+install:
+
+tarball: all
+	$(TAR) cf spm_mex.tar $(SPMMEX)
+
+%.$(SUF) : %.c
+	$(MEX) $< $(MEXEND)
diff --git a/toolbox/Seg/spm_maff8.m b/toolbox/Seg/spm_maff8.m
index 3085238..fc952ed 100644
--- a/toolbox/Seg/spm_maff8.m
+++ b/toolbox/Seg/spm_maff8.m
@@ -21,7 +21,7 @@
 % Copyright (C) 2008 Wellcome Department of Imaging Neuroscience
 
 % John Ashburner
-% $Id: spm_maff8.m 3217 2009-06-22 10:50:46Z john $
+% $Id: spm_maff8.m 4152 2011-01-11 14:13:35Z volkmar $
 
 [buf,MG,x,ff] = loadbuf(varargin{1:3});
 [M,h]         = affreg(buf, MG, x, ff, varargin{4:end});
@@ -102,7 +102,9 @@
 x1 = x{1};
 x2 = x{2};
 x3 = x{3};
-[mu,isig] = priors(regtyp);
+[mu,isig] = spm_affine_priors(regtyp);
+mu   = [zeros(6,1) ; mu];
+isig = [zeros(6,12) ; zeros(6,6) isig];
 isig      =  isig*ff;
 Alpha0    =  isig;
 
@@ -111,7 +113,7 @@
 ll   = -Inf;
 krn  = spm_smoothkern(2,(-256:256)',0);
 
-spm_chi2_plot('Init','Registering','Log-likelihood','Iteration');
+spm_plot_convergence('Init','Registering','Log-likelihood','Iteration');
 
 h1 = ones(256,numel(tpm.dat));
 for iter=1:200
@@ -179,7 +181,7 @@
     ssh   = sum(h0(:));
     ll1   = sum(sum(h0.*log2(h1)))/ssh - penalty/ssh;
    %fprintf('%g\t%g\n', sum(sum(h0.*log2(h1)))/ssh, -penalty/ssh);
-    spm_chi2_plot('Set',ll1);
+    spm_plot_convergence('Set',ll1);
     if (ll1-ll)<1e-4, break; end;
     ll    = ll1;
     sol   = sol1;
@@ -236,7 +238,7 @@
     sol1  = sol - (Alpha+Alpha0)\(Beta+isig*(sol-mu));
 end;
 
-spm_chi2_plot('Clear');
+spm_plot_convergence('Clear');
 M = P2M(sol);
 return;
 %_______________________________________________________________________
@@ -300,88 +302,4 @@
 return;
 %_______________________________________________________________________
 %_______________________________________________________________________
-function [mu,isig] = priors(typ)
-% The parameters for this distribution were derived empirically from 227
-% scans, that were matched to the ICBM space.
-
-%% Values can be derived by...
-%sn = spm_select(Inf,'.*seg_inv_sn.mat$');
-%X  = zeros(size(sn,1),12);
-%for i=1:size(sn,1),
-%    p  = load(deblank(sn(i,:)));
-%    M  = p.VF(1).mat*p.Affine/p.VG(1).mat;
-%    J  = M(1:3,1:3);
-%    V  = sqrtm(J*J');
-%    R  = V\J;
-%    lV      =  logm(V);
-%    lR      = -logm(R);
-%    P       = zeros(12,1);
-%    P(1:3)  = M(1:3,4);
-%    P(4:6)  = lR([2 3 6]);
-%    P(7:12) = lV([1 2 3 5 6 9]);
-%    X(i,:)  = P';
-%end;
-%mu   = mean(X(:,7:12));
-%XR   = X(:,7:12) - repmat(mu,[size(X,1),1]);
-%isig = inv(XR'*XR/(size(X,1)-1))
-
-
-switch deblank(lower(typ))
-
-case 'mni', % For registering with MNI templates...
-        mu   = [0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
-        isig = 1e4 * [
-            0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
-           -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
-           -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
-           -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
-           -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
-           -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
-
-case 'imni', % For registering with MNI templates...
-        mu   = -[0.0667 0.0039 0.0008 0.0333 0.0071 0.1071]';
-        isig = 1e4 * [
-            0.0902   -0.0345   -0.0106   -0.0025   -0.0005   -0.0163
-           -0.0345    0.7901    0.3883    0.0041   -0.0103   -0.0116
-           -0.0106    0.3883    2.2599    0.0113    0.0396   -0.0060
-           -0.0025    0.0041    0.0113    0.0925    0.0471   -0.0440
-           -0.0005   -0.0103    0.0396    0.0471    0.2964   -0.0062
-           -0.0163   -0.0116   -0.0060   -0.0440   -0.0062    0.1144];
-
-case 'rigid', % Constrained to be almost rigid...
-        mu   = zeros(6,1);
-        isig = eye(6)*1e8;
-
-case 'subj', % For inter-subject registration...
-        mu   = zeros(6,1);
-        isig = 1e3 * [
-            0.8876    0.0784    0.0784   -0.1749    0.0784   -0.1749
-            0.0784    5.3894    0.2655    0.0784    0.2655    0.0784
-            0.0784    0.2655    5.3894    0.0784    0.2655    0.0784
-           -0.1749    0.0784    0.0784    0.8876    0.0784   -0.1749
-            0.0784    0.2655    0.2655    0.0784    5.3894    0.0784
-           -0.1749    0.0784    0.0784   -0.1749    0.0784    0.8876];
-
-case 'eastern', % For East Asian brains to MNI...
-    mu   = [0.0719   -0.0040   -0.0032    0.1416    0.0601    0.2578]';
-    isig = 1e4 * [
-        0.0757    0.0220   -0.0224   -0.0049    0.0304   -0.0327
-        0.0220    0.3125   -0.1555    0.0280   -0.0012   -0.0284
-       -0.0224   -0.1555    1.9727    0.0196   -0.0019    0.0122
-       -0.0049    0.0280    0.0196    0.0576   -0.0282   -0.0200
-        0.0304   -0.0012   -0.0019   -0.0282    0.2128   -0.0275
-       -0.0327   -0.0284    0.0122   -0.0200   -0.0275    0.0511];
-
-case 'none', % No regularisation...
-        mu   = zeros(6,1);
-        isig = zeros(6);
-
-otherwise
-        error(['"' typ '" not recognised as type of regularisation.']);
-end;
-mu   = [zeros(6,1) ; mu];
-isig = [zeros(6,12) ; zeros(6,6) isig];
-return;
-%_______________________________________________________________________
-%_______________________________________________________________________
 
diff --git a/toolbox/Seg/spm_mrf.c b/toolbox/Seg/spm_mrf.c
new file mode 100644
index 0000000..b3a807b
--- /dev/null
+++ b/toolbox/Seg/spm_mrf.c
@@ -0,0 +1,333 @@
+/* $Id: spm_mrf.c 4036 2010-08-06 14:45:17Z john $ */
+/* (c) John Ashburner (2010) */
+
+#include "mex.h"
+#include <math.h>
+#define MAXCLASSES 1024
+
+static void mrf1(mwSize dm[], unsigned char q[], float p[], float G[], float w[], int code)
+{
+    mwSize i0, i1, i2, k, m, n;
+    float a[MAXCLASSES], e[MAXCLASSES], *p0, *p1;
+    unsigned char *q0, *q1;
+    int it;
+
+    m = dm[0]*dm[1]*dm[2];
+
+    /* Use a red-black scheme, so the updates are for
+       alternating voxels.  Then do another pass to
+       update the other half.
+       A B A B A B
+       B A B A B A
+       A B A B A B
+       B A B A B A
+
+       Updates involve computing the number of neighbours
+       of each type (stored in vector a), and using the
+       connectivity matrix (G) to update with:
+           q = (p.*exp(G'*a))/sum((p.*exp(G'*a))
+    */
+    for(it=0; it<2; it++) 
+    {
+        mwSize i2start = it%2;
+        for(i2=0; i2<dm[2]; i2++) /* Inferior -> Superior */
+        {
+            mwSize i1start = (i2start == (i2%2));
+            for(i1=0; i1<dm[1]; i1++) /* Posterior -> Anterior */
+            {
+                mwSize i0start = (i1start == (i1%2));
+                p1 = p + dm[0]*(i1+dm[1]*i2);
+                q1 = q + dm[0]*(i1+dm[1]*i2);
+
+                for(i0=i0start; i0<dm[0]; i0+=2) /* Left -> Right */
+                {
+                    float se;
+                    unsigned char *qq;
+
+                    /* Pointers to current voxel in first volume */
+                    p0 = p1 + i0;
+                    q0 = q1 + i0;
+
+                    /* Initialise neighbour counts to zero */
+                    for(k=0; k<dm[3]; k++) a[k] = 0.0;
+
+                    /* Count neighbours of each class */
+                    if(i2>0)       /* Inferior */
+                    {
+
+                        qq = q0 - dm[0]*dm[1];
+                        for(k=0; k<dm[3]; k++) a[k] += qq[k*m]*w[2];
+                    }
+
+                    if(i2<dm[2]-1) /* Superior */
+                    {
+                        qq = q0 + dm[0]*dm[1];
+                        for(k=0; k<dm[3]; k++) a[k] += qq[k*m]*w[2];
+                    }
+
+                    if(i1>0)       /* Posterior */
+                    {
+                        qq = q0 - dm[0];
+                        for(k=0; k<dm[3]; k++) a[k] += qq[k*m]*w[1];
+                    }
+
+                    if(i1<dm[1]-1) /* Anterior */
+                    {
+                        qq = q0 + dm[0];
+                        for(k=0; k<dm[3]; k++) a[k] += qq[k*m]*w[1];
+                    }
+
+                    if(i0>0)       /* Left */
+                    {
+                        qq = q0 - 1;
+                        for(k=0; k<dm[3]; k++) a[k] += qq[k*m]*w[0];
+                    }
+
+                    if(i0<dm[0]-1) /* Right */
+                    {
+                        qq = q0 + 1;
+                        for(k=0; k<dm[3]; k++) a[k] += qq[k*m]*w[0];
+                    }
+
+                    /* Responsibility data is uint8, so correct scaling.
+                       Note also that data is divided by 6 (the number
+                       of neighbours examined). */
+                    for(k=0; k<dm[3]; k++)
+                        a[k]/=(255.0*6.0);
+
+                    if (code == 1) 
+                    {
+                        /* Weights are in the form of a matrix,
+                           shared among all voxels. */
+                        float *g;
+                        se = 0.0;
+                        for(k=0, g=G; k<dm[3]; k++)
+                        {
+                            e[k] = 0;
+                            for(n=0; n<dm[3]; n++, g++)
+                                e[k] += (*g)*a[n];
+                            e[k] = exp((double)e[k])*p0[k*m];
+                            se  += e[k];
+                        }
+                    }
+                    else if (code == 2)
+                    {
+                        /* Weights are assumed to be a diagonal matrix,
+                           so only the diagonal elements are passed. */
+                        se = 0.0;
+                        for(k=0; k<dm[3]; k++)
+                        {
+                            e[k] = exp((double)(G[k]*a[k]))*p0[k*m];
+                            se  += e[k];
+                        }
+                    }
+                    else if (code == 3)
+                    {
+                        /* Separate weights for each voxel, in the form of
+                           the full matrix (loads of memory). */
+                        float *g;
+                        se = 0.0;
+                        g = G + i0+dm[0]*(i1+dm[1]*i2);
+                        for(k=0; k<dm[3]; k++)
+                        {
+                            e[k] = 0.0;
+                            for(n=0; n<dm[3]; n++, g+=m)
+                                e[k] += (*g)*a[n];
+                            e[k] = exp((double)e[k])*p0[k*m];
+                            se  += e[k];
+                        }
+                    }
+                    else if (code == 4)
+                    {
+                        /* Separate weight matrices for each voxel,
+                           where the matrices are assumed to be symmetric
+                           with zeros on the diagonal. For a 4x4
+                           matrix, the elements are ordered as
+                           (2,1), (3,1), (4,1), (3,2), (4,2), (4,3).
+                         */
+                        float *g;
+                        g = G + i0+dm[0]*(i1+dm[1]*i2);
+                        for(k=0; k<dm[3]; k++) e[k] = 0.0;
+
+                        for(k=0; k<dm[3]; k++)
+                        {
+                            for(n=k+1; n<dm[3]; n++, g+=m)
+                            {
+                                e[k] += (*g)*a[n];
+                                e[n] += (*g)*a[k];
+                            }
+                        }
+                        se = 0.0;
+                        for(k=0; k<dm[3]; k++)
+                        {
+                            e[k] = exp((double)e[k])*p0[k*m];
+                            se  += e[k];
+                        }
+                    }
+                    else if (code == 5)
+                    {
+                        /* Separate weight matrices for each voxel,
+                           where the matrices are assumed to be symmetric
+                           with zeros on the diagonal. For a 4x4
+                           matrix, the elements are ordered as
+                           (2,1), (3,1), (4,1), (3,2), (4,2), (4,3).
+                           
+                           The weight matrices are encoded as uint8, and
+                           their values need to be scaled by -0.0625 to
+                           bring them into a reasonable range.
+                         */
+                        unsigned char *g;
+                        g = (unsigned char *)G + i0+dm[0]*(i1+dm[1]*i2);
+                        for(k=0; k<dm[3]; k++) e[k] = 0.0;
+
+                        for(k=0; k<dm[3]; k++)
+                        {
+                            for(n=k+1; n<dm[3]; n++, g+=m)
+                            {
+                                e[k] += ((float)(*g))*a[n];
+                                e[n] += ((float)(*g))*a[k];
+                            }
+                        }
+                        se = 0.0;
+                        for(k=0; k<dm[3]; k++)
+                        {
+                            e[k] = exp(-0.0625*e[k])*p0[k*m];
+                            se  += e[k];
+                        }
+                    }
+
+
+                    /* Normalise responsibilities to sum to 1
+                       and rescale for saving as uint8 data. */
+                    se = 255.0/se;
+                    for(k=0; k<dm[3]; k++)
+                        q0[k*m] = (unsigned char)(e[k]*se+0.5);
+
+                }
+            }
+        }
+    }
+}
+
+void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
+{
+    mwSize i;
+    mwSize dm[4];
+    float *p, w[3];
+    unsigned char *q;
+    float *G;
+    int code=0;
+
+    if (nrhs<3 || nrhs>4 || nlhs>1)
+        mexErrMsgTxt("Incorrect usage");
+
+    if (!mxIsNumeric(prhs[0]) || mxIsComplex(prhs[0]) || mxIsSparse(prhs[0]) || !mxIsUint8(prhs[0]))
+        mexErrMsgTxt("First arg must be numeric, real, full and uint8.");
+
+    if (!mxIsNumeric(prhs[1]) || mxIsComplex(prhs[1]) || mxIsSparse(prhs[1]) || !mxIsSingle(prhs[1]))
+        mexErrMsgTxt("Second arg must be numeric, real, full and single.");
+
+    if (!mxIsNumeric(prhs[2]) || mxIsComplex(prhs[2]) || mxIsSparse(prhs[2]))
+        mexErrMsgTxt("Third arg must be numeric, real and full.");
+
+    if (mxGetNumberOfDimensions(prhs[0])!= mxGetNumberOfDimensions(prhs[1]) ||
+        mxGetNumberOfDimensions(prhs[0])>4)
+        mexErrMsgTxt("First or second args have wrong number of dimensions.");
+
+    for(i=0; i<mxGetNumberOfDimensions(prhs[0]); i++)
+        dm[i] = mxGetDimensions(prhs[0])[i];
+
+    for(i=mxGetNumberOfDimensions(prhs[0]); i<4; i++)
+        dm[i] = 1;
+
+    if (dm[3]>MAXCLASSES) mexErrMsgTxt("Too many classes.");
+
+    for(i=0; i<4; i++)
+        if (mxGetDimensions(prhs[1])[i] != dm[i])
+            mexErrMsgTxt("First and second args have incompatible dimensions.");
+
+    if (mxGetDimensions(prhs[2])[1] == 1)
+    {
+        code = 2;
+        if (mxGetDimensions(prhs[2])[0] != dm[3])
+            mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+        if (!mxIsSingle(prhs[2])) mexErrMsgTxt("Third arg must be single.");
+    }
+    else if (mxGetNumberOfDimensions(prhs[2])==2)
+    {
+        code = 1;
+        if (mxGetDimensions(prhs[2])[0] != dm[3] || mxGetDimensions(prhs[2])[1] != dm[3])
+            mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+        if (!mxIsSingle(prhs[2])) mexErrMsgTxt("Third arg must be single.");
+    }
+    else if (mxGetNumberOfDimensions(prhs[2])==5)
+    {
+        code = 3;
+        for(i=0; i<4; i++)
+            if (mxGetDimensions(prhs[2])[i] != dm[i])
+                mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+        if (mxGetDimensions(prhs[2])[4] != dm[3])
+            mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+        if (!mxIsSingle(prhs[2])) mexErrMsgTxt("Third arg must be single.");
+    }
+    else if (mxGetNumberOfDimensions(prhs[2])==4)
+    {
+        for(i=0; i<3; i++)
+            if (mxGetDimensions(prhs[2])[i] != dm[i])
+                mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+        if (mxGetDimensions(prhs[2])[3] != (dm[3]*(dm[3]-1))/2)
+            mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+        if (mxIsSingle(prhs[2]))
+            code = 4;
+        else if (mxIsUint8(prhs[0]))
+            code = 5;
+        else
+            mexErrMsgTxt("Third arg must be either single or uint8.");
+    }
+    else
+        mexErrMsgTxt("Third arg has incompatible dimensions.");
+
+
+    p = (float *)mxGetData(prhs[1]);
+    G = (float *)mxGetData(prhs[2]);
+
+    if (nrhs>=4)
+    {
+        /* Adjustment for anisotropic voxel sizes.  w should contain
+           the square of each voxel size. */
+        if (!mxIsNumeric(prhs[3]) || mxIsComplex(prhs[3]) || mxIsSparse(prhs[3]) || !mxIsSingle(prhs[3]))
+            mexErrMsgTxt("Fourth arg must be numeric, real, full and single.");
+
+        if (mxGetNumberOfElements(prhs[3]) != 3)
+            mexErrMsgTxt("Fourth arg must contain three elements.");
+
+        for(i=0; i<3; i++) w[i] = ((float *)mxGetData(prhs[3]))[i];
+    }
+    else
+    {
+        for(i=0; i<3; i++) w[i] = 1.0;
+    }
+
+    if (nlhs>0)
+    {
+        /* Copy input to output */
+        unsigned char *q0;
+        plhs[0]  = mxCreateNumericArray(4,dm, mxUINT8_CLASS, mxREAL);
+        q0 = (unsigned char *)mxGetData(prhs[0]);
+        q  = (unsigned char *)mxGetData(plhs[0]);
+
+        for(i=0; i<dm[0]*dm[1]*dm[2]*dm[3]; i++)
+            q[i] = q0[i];
+    }
+    else /* Note the nasty side effects - but it does save memory */
+        q = (unsigned char *)mxGetData(prhs[0]);
+
+    mrf1(dm, q,p,G,w,code);
+}
+
diff --git a/toolbox/Seg/spm_mrf.m b/toolbox/Seg/spm_mrf.m
new file mode 100644
index 0000000..71460af
--- /dev/null
+++ b/toolbox/Seg/spm_mrf.m
@@ -0,0 +1,51 @@
+function varargout = spm_mrf(varargin)
+%  Markov Random Field Code
+%
+%_______________________________________________________________________
+%
+% FORMAT q1 = spm_mrf(q0,p,G,w)
+% q0 - Original responsibilities.
+%      Values are stored as uint8 and converted to responsibilities by
+%      rescaling by 1/255. Dimensions are dim1 x dim2 x dim3 x K
+% p  - Probabilities.
+%      Values are as single precision floating point.  This array must
+%      have the same dimensions as q0.
+% G  - Matrix used to encode neighbourhood information.
+%      May be of a number of types.
+%        i) K x K matrix (single precision), where K is the 4th
+%           dimension of q0 and p.  This matrix is shared by all voxels.
+%       ii) K x 1 vector (single precision), encoding the diagonal of
+%           a matrix.
+%      iii) dim1 x dim2 x dim3 x K x K (single precision).  Encodes a
+%           different matrix at each voxel, and is very memory hungry.
+%       iv) dim1 x dim2 x dim3 x (K*(K-1)/2) (single precision).
+%           Encodes a symmetric matrix, with zeros on the diagonal, at
+%           each voxel.
+%        v) dim1 x dim2 x dim3 x (K*(K-1)/2) (uint8).
+%           Encodes a symmetric matrix, with zeros on the diagonal, at
+%           each voxel. Saves more memory by using uint8.  Note that
+%           when used, the uint8 values are rescaled by -1/(2^4).
+% w  - A vector of three weights, which normally encode the square of
+%      the voxel sizes.  This is for dealing with anisotropic voxels.
+%      If this argument is not supplied, then [1 1 1] is assumed.
+% q1 - Output responsibilities.
+%
+% FORMAT spm_mrf(q,p,G,w)
+% This is the dodgy way of using the function, as it changes the RHS
+% argument (q) and can lead to some strange side effects.  This approach
+% would not be ebdorsed by the MathWorks, but it does save a bit of memory.
+%
+%
+% The MRF updates are done using a red-black checkerboard scheme.  Each
+% voxel is updated by q = (exp(G'*a).*p)/sum(exp(G'*a).*p), where
+% vector a is computed from the number of neighbours of each type
+% (divided by 6). The contribution of each neighbour is scaled by w.
+%
+%_______________________________________________________________________
+% Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_mrf.m 4036 2010-08-06 14:45:17Z john $
+
+error('Not compiled for %s in MATLAB %s.\n\nThis could easily be done by typing:\n    mex spm_mrf.c -O\n', computer, version);
+
diff --git a/toolbox/Seg/spm_preproc8.m b/toolbox/Seg/spm_preproc8.m
index 0847901..babf646 100644
--- a/toolbox/Seg/spm_preproc8.m
+++ b/toolbox/Seg/spm_preproc8.m
@@ -72,7 +72,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_preproc8.m 3864 2010-05-05 17:21:20Z john $
+% $Id: spm_preproc8.m 4148 2011-01-04 16:49:23Z guillaume $
 
 Affine    = obj.Affine;
 tpm       = obj.tpm;
@@ -256,7 +256,7 @@
     clear B1 B2 B3 T C
 end
 
-spm_chi2_plot('Init','Initialising','Log-likelihood','Iteration');
+spm_plot_convergence('Init','Initialising','Log-likelihood','Iteration');
 for iter=1:20,
 
     % Load the warped prior probability images into the buffer
@@ -424,7 +424,7 @@
                 end
 
                 if subit>1 || iter>1,
-                    spm_chi2_plot('Set',ll);
+                    spm_plot_convergence('Set',ll);
                 end
                 if ll-oll<tol1*nm,
                     % Improvement is small, so go to next step
@@ -491,7 +491,7 @@
                %fprintf('Hist:\t%g\t%g\t%g\n', ll,llr,llrb);
 
                 if subit>1 || iter>1,
-                    spm_chi2_plot('Set',ll);
+                    spm_plot_convergence('Set',ll);
                 end
                 if ll-oll<tol1*nm,
                     % Improvement is small, so go to next step
@@ -651,7 +651,7 @@
                             for n1=1:N, llrb = llrb + chan(n1).ll; end
                         else
                             % Accept new solution
-                            spm_chi2_plot('Set',ll);
+                            spm_plot_convergence('Set',ll);
                            %fprintf('Bias-%d:\t%g\t%g\t%g :o)\n', n, ll, llr,llrb);
                             if oll-ll<0,
                                 chan(n).lmreg = chan(n).lmreg*0.5;
@@ -673,8 +673,8 @@
         if iter==1 && iter1==1,
             % Most of the log-likelihood improvements are in the first iteration.
             % Show only improvements after this, as they are more clearly visible.
-            spm_chi2_plot('Clear');
-            spm_chi2_plot('Init','Processing','Log-likelihood','Iteration');
+            spm_plot_convergence('Clear');
+            spm_plot_convergence('Init','Processing','Log-likelihood','Iteration');
 
            if use_mog && numel(obj.lkp) ~= numel(lkp),
                 mn1 = mn;
@@ -842,7 +842,7 @@
                 lam   = lam*8;
                %fprintf('Warp:\t%g\t%g\t%g :o(\n', ll1, llr1,llrb);
             else
-                spm_chi2_plot('Set',ll1);
+                spm_plot_convergence('Set',ll1);
                 lam   = lam*0.5;
                 ll    = ll1;
                 llr   = llr1;
@@ -863,7 +863,7 @@
         break
     end
 end
-% spm_chi2_plot('Clear');
+% spm_plot_convergence('Clear');
 
 results.image  = obj.image;
 results.tpm    = tpm.V;
diff --git a/toolbox/Seg/spm_preproc_run.m b/toolbox/Seg/spm_preproc_run.m
index 9c0af3b..e275aca 100644
--- a/toolbox/Seg/spm_preproc_run.m
+++ b/toolbox/Seg/spm_preproc_run.m
@@ -18,21 +18,24 @@
 %
 % See the user interface for a description of the fields.
 %_______________________________________________________________________
-% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+% Copyright (C) 2008-2011 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: spm_preproc_run.m 3859 2010-05-04 13:43:41Z john $
+% $Id: spm_preproc_run.m 4221 2011-03-01 20:22:52Z guillaume $
 
-if nargin==1,
-    varargout{:} = run_job(job);
-elseif strcmpi(arg,'check'),
-    varargout{:} = check_job(job);
-elseif strcmpi(arg,'vfiles'),
-    varargout{:} = vfiles_job(job);
-elseif strcmpi(arg,'vout'),
-    varargout{:} = vout_job(job);
-else
-    error('Unknown argument ("%s").', arg);
+if nargin == 1, arg = 'run'; end
+
+switch lower(arg)
+    case 'run'
+        varargout{1} = run_job(job);
+    case 'check'
+        varargout{1} = check_job(job);
+    case 'vfiles'
+        varargout{1} = vfiles_job(job);
+    case 'vout'
+        varargout{1} = vout_job(job);
+    otherwise
+        error('Unknown argument ("%s").', arg);
 end
 return
 %_______________________________________________________________________
@@ -162,17 +165,17 @@
 
 %_______________________________________________________________________
 function savefields(fnam,p)
-if length(p)>1, error('Can''t save fields.'); end;
+if length(p)>1, error('Can''t save fields.'); end
 fn = fieldnames(p);
-if numel(fn)==0, return; end;
-for i=1:length(fn),
+if numel(fn)==0, return; end
+for i=1:length(fn)
     eval([fn{i} '= p.' fn{i} ';']);
-end;
-if spm_matlab_version_chk('7') >= 0
+end
+if spm_check_version('matlab','7') >= 0
     save(fnam,'-V6',fn{:});
 else
     save(fnam,fn{:});
-end;
+end
 
 return;
 %_______________________________________________________________________
diff --git a/toolbox/Seg/spm_preproc_write8.m b/toolbox/Seg/spm_preproc_write8.m
index 48c947a..714c6ac 100644
--- a/toolbox/Seg/spm_preproc_write8.m
+++ b/toolbox/Seg/spm_preproc_write8.m
@@ -5,7 +5,7 @@
 % Copyright (C) 2008 Wellcome Department of Imaging Neuroscience
 
 % John Ashburner
-% $Id: spm_preproc_write8.m 3172 2009-06-02 11:40:42Z john $
+% $Id: spm_preproc_write8.m 4194 2011-02-05 18:08:06Z ged $
 
 % Read essentials from tpm (it will be cleared later)
 tpm = res.tpm;
@@ -15,8 +15,7 @@
 d1        = size(tpm.dat{1});
 d1        = d1(1:3);
 M1        = tpm.M;
-[bb1,vx1] = bbvox_from_V(tpm.V(1));
-
+[bb1 vx1] = spm_get_bbox(tpm.V(1), 'old');
 
 if isfield(res,'mg'),
     lkp = res.lkp;
@@ -27,7 +26,7 @@
 
 N   = numel(res.image);
 if nargin<2, tc = true(Kb,4); end % native, import, warped, warped-mod
-if nargin<3, bf = true(N,2);  end % corrected, field
+if nargin<3, bf = true(N,2);  end % field, corrected
 if nargin<4, df = true(1,2);  end % inverse, forward
 
 [pth,nam]=fileparts(res.image(1).fname);
@@ -321,7 +320,7 @@
         if tc(k1,3),
             N      = nifti;
             N.dat  = file_array(fullfile(pth,['wc', num2str(k1), nam, '.nii']),...
-                                d1,'uint8-be',0,1/255,0);
+                                d1,'uint8',0,1/255,0);
             N.mat  = M1;
             N.mat0 = M1;
             N.descrip = ['Warped tissue class ' num2str(k1)];
@@ -338,7 +337,7 @@
     y         = spm_invert_def(y,M1,d1,M0,[1 0]);
     N         = nifti;
     N.dat     = file_array(fullfile(pth,['y_', nam1, '.nii']),...
-                           [d1,1,3],'float32-be',0,1,0);
+                           [d1,1,3],'float32',0,1,0);
     N.mat     = M1;
     N.mat0    = M1;
     N.descrip = 'Deformation';
@@ -411,17 +410,6 @@
 return;
 %=======================================================================
 
-%=======================================================================
-function [bb,vx] = bbvox_from_V(V)
-vx = sqrt(sum(V(1).mat(1:3,1:3).^2));
-if det(V(1).mat(1:3,1:3))<0, vx(1) = -vx(1); end;
-
-o  = V(1).mat\[0 0 0 1]';
-o  = o(1:3)';
-bb = [-vx.*(o-1) ; vx.*(V(1).dim(1:3)-o)];
-return;
-%=======================================================================
-
 %=======================================================================
 function y1 = affind(y0,M)
 y1 = zeros(size(y0),'single');
diff --git a/toolbox/Seg/tbx_cfg_preproc8.m b/toolbox/Seg/tbx_cfg_preproc8.m
index 401fde7..f178734 100644
--- a/toolbox/Seg/tbx_cfg_preproc8.m
+++ b/toolbox/Seg/tbx_cfg_preproc8.m
@@ -4,7 +4,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % John Ashburner
-% $Id: tbx_cfg_preproc8.m 3910 2010-06-01 14:58:20Z guillaume $
+% $Id: tbx_cfg_preproc8.m 4221 2011-03-01 20:22:52Z guillaume $
 
 if ~isdeployed, addpath(fullfile(spm('Dir'),'toolbox','Seg')); end
 
@@ -350,7 +350,7 @@
 
 %======================================================================
 function varargout = spm_local_preproc_run(job)
-varargout{:} = spm_preproc_run(job);
+varargout{1} = spm_preproc_run(job);
 
 %======================================================================
 function dep = vout(job)
diff --git a/toolbox/Shoot/spm_shoot3d.m b/toolbox/Shoot/spm_shoot3d.m
new file mode 100644
index 0000000..c007442
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot3d.m
@@ -0,0 +1,174 @@
+function varargout = spm_shoot3d(v0,prm,args)
+% Geodesic shooting
+% FORMAT [phi,Jphi,v1,theta,Jtheta] = spm_shoot3d(v0,prm,args)
+% v0   - Initial velocity field n1*n2*n3*3 (single prec. float)
+% prm  - Differential operator parameters
+% prm  - 7 parameters (settings)
+%        - [1] Regularisation type (ie the form of the differential
+%          operator), can take values of
+%           - 0 Linear elasticity
+%           - 1 Membrane energy
+%           - 2 Bending energy
+%        - [2][3][4] Voxel sizes
+%        - [5][6][7] Regularisation parameters
+%           - For "membrane energy", the parameters are
+%             lambda, unused and id.
+%           - For "linear elasticity", the parameters are
+%             mu, lambda, and id
+%           - For "bending energy", the parameters are
+%             lambda, id1 and id2.
+% args - Integration parameters
+%        - [1] Num time steps
+%        - [2][3] Multigrid parameters (cycles and iterations)
+%          for generating velocities from momentum
+%
+% phi  - Forward deformation field n1*n2*n3*3 (single prec. float)
+% Jphi - Forward Jacobian tensors n1*n2*n3 (single prec. float)
+% v1   - Final velocity field n1*n2*n3*3 (single prec. float)
+% theta - Inverse deformation field n1*n2*n3*3 (single prec. float)
+% Jtheta   - Inverse Jacobian tensors n1*n2*n3 (single prec. float)
+%
+% This code generates deformations and their Jacobian determinans from
+% initial velocity fields by gedesic shooting.  See the work of Miller,
+% Younes and others.
+%
+% LDDMM (Beg et al) uses the following evolution equation:
+%     d\phi/dt = v_t(\phi_t)
+% where a variational procedure is used to find the stationary solution
+% for the time varying velocity field.
+% In principle though, once the initial velocity is known, then the
+% velocity at subsequent time points can be computed.  This requires
+% initial momentum (m_0), computed (using differential operator L) by:
+%     m_0 = L v_0
+% Then (Ad_{\phi_t})^* m_0 is computed:
+%     m_t = |d \phi_t| (d\phi_t)^T m_0(\phi_t)
+% The velocity field at this time point is then obtained by using
+% multigrid to solve:
+%     v_t = L^{-1} m_t
+%
+% These equations can be found in:
+% Younes (2007). "Jacobi fields in groups of diffeomorphisms and
+% applications". Quarterly of Applied Mathematics, vol LXV,
+% number 1, pages 113-134 (2007).
+%
+% Note that in practice, (Ad_{\phi_t})^* m_0 is computed differently,
+% by multiplying the initial momentum by the inverse of the Jacobian
+% matrices of the inverse warp, and pushing the values to their new
+% location by the inverse warp (see the "pushg" code of dartel3).
+% Multigrid is currently used to obtain v_t = L^{-1} m_t, but
+% this could also be done by convolution with the Greens function
+% K = L^{-1} (see e.g. Bro-Nielson).
+%
+%________________________________________________________
+% (c) Wellcome Trust Centre for NeuroImaging (2009)
+
+% John Ashburner
+% $Id: spm_shoot3d.m 4026 2010-07-29 13:45:50Z john $
+
+args0 = [8 4 4];
+if nargin<3,
+    args = args0;
+else
+    if numel(args)<numel(args0),
+        args = [args args0((numel(args)+1):end)];
+    end
+end
+verb     = false;
+N        = args(1);   % # Time steps
+fmg_args = args(2:3); % Multigrid params
+d        = size(v0);
+d        = d(1:3);
+vt       = v0;
+
+if ~isfinite(N),
+    % Number of time steps from an educated guess about how far to move
+    N = double(floor(sqrt(max(max(max(v0(:,:,:,1).^2+v0(:,:,:,2).^2+v0(:,:,:,3).^2)))))+1);
+end
+
+if verb, fprintf('N=%g:', N); end
+
+m0       = dartel3('vel2mom',v0,prm); % Initial momentum (m_0 = L v_0)
+if verb, fprintf('\t%g', 0.5*sum(v0(:).*m0(:))/prod(d)); end
+
+% Compute initial small deformation and Jacobian matrices from the velocity.
+% The overall deformation and its Jacobian tensor field is generated by
+% composing a series of small deformations.
+[ phi, Jphi]     = dartel3('smalldef', vt,1/N);
+
+% If required, also compute the forward and possibly also its Jacobian
+% tensor field. Note that the order of the compositions will be the
+% reverse of those for building the forward deformation.
+if nargout>=5,
+    [theta,Jtheta] = dartel3('smalldef', vt,-1/N);
+elseif nargout>=4,
+    theta          = dartel3('smalldef', vt,-1/N);
+end
+
+for t=2:abs(N),
+    mt             = dartel3('pushg',m0,phi,Jphi);
+    vt             = mom2vel(mt,prm,fmg_args,vt);
+    if verb, fprintf('\t%g', 0.5*sum(vt(:).*mt(:))/prod(d)); end
+
+    [  dp,   dJ]   = dartel3('smalldef',  vt,1/N);       % Small deformation
+    [ phi, Jphi]   = dartel3('comp', dp, phi, dJ, Jphi); % Build up large def. from small defs
+
+    clear dp dJ
+
+    % If required, build up forward deformation and its Jacobian tensor field from
+    % small deformations
+    if nargout>=5,
+        [   dp,    dJ] = dartel3('smalldef',  vt,-1/N);    % Small deformation
+        [theta,Jtheta] = dartel3('comp', theta,dp, Jtheta,dJ);  % Build up large def. from small defs
+        clear dp dJ
+    elseif nargout>=4,
+        dp             = dartel3('smalldef',  vt,-1/N);    % Small deformation
+        theta          = dartel3('comp', theta, dp);       % Build up large def. from small defs
+        clear dp
+    end
+    drawnow
+end
+if verb, fprintf('\n'); end
+
+
+varargout{1} = phi;
+varargout{2} = Jphi;
+if nargout>=3,
+    mt           = dartel3('pushg',m0,phi,Jphi);
+    varargout{3} = mom2vel(mt,prm,fmg_args,vt);
+end
+if nargout>=4, varargout{4} = theta; end
+if nargout>=5, varargout{5} = Jtheta;   end
+%__________________________________________________________________________________
+
+%__________________________________________________________________________________
+function vt = mom2vel(mt,prm,fmg_args,vt)
+% L^{-1} m_t
+
+r   = dartel3('vel2mom',vt,prm);
+if prm(1) == 0,
+    % This option has been coded up in C
+    vt = dartel3('mom2vel', mt-r, [prm fmg_args])+vt;
+else
+    % This option has not been coded in C yet, so needs to be
+    % done in a less efficient way
+    dm = size(mt);
+    H  = zeros([dm(1:3),6],'single');
+    vt = dartel3('fmg',H,mt-r, [prm fmg_args])+vt;
+end
+
+if false,
+    % Go for machine precision
+    r  = dartel3('vel2mom',vt,prm);
+    ss = sum(sum(sum(sum((mt-r).^2))));
+    for i=1:8,
+        oss = ss;
+        vt  = dartel3('mom2vel', mt-r, [prm fmg_args])+vt;
+        r   = dartel3('vel2mom',vt,prm);
+        ss  = sum(sum(sum(sum((mt-r).^2))));
+        if ss/oss>0.9, break; end
+    end
+end
+%__________________________________________________________________________________
+
+%__________________________________________________________________________________
+
diff --git a/toolbox/Shoot/spm_shoot3di.m b/toolbox/Shoot/spm_shoot3di.m
new file mode 100644
index 0000000..0ee3a9d
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot3di.m
@@ -0,0 +1,179 @@
+function varargout = spm_shoot3di(v0,prm,args)
+% Geodesic shooting
+% FORMAT [theta,Jtheta,v1,phi,Jphi] = spm_shoot3di(v0,prm,args)
+% v0     - Initial velocity field n1*n2*n3*3 (single prec. float)
+% prm    - 7 Differential operator parameter settings
+%          - [1] Regularisation type (ie the form of the differential
+%            operator), can take values of
+%             - 0 Linear elasticity
+%             - 1 Membrane energy
+%             - 2 Bending energy
+%          - [2][3][4] Voxel sizes
+%          - [5][6][7] Regularisation parameters
+%             - For "membrane energy", the parameters are
+%               lambda, unused and id.
+%             - For "linear elasticity", the parameters are
+%               mu, lambda, and id
+%             - For "bending energy", the parameters are
+%               lambda, id1 and id2.
+% args   - Integration parameters
+%          - [1] Num time steps
+%          - [2][3] Multigrid parameters (cycles and iterations)
+%            for generating velocities from momentum
+%
+% theta  - Inverse deformation field n1*n2*n3*3 (single prec. float)
+% Jtheta - Inverse Jacobian tensors n1*n2*n3 (single prec. float)
+% v1     - Final velocity field n1*n2*n3*3 (single prec. float)
+% phi    - Forward deformation field n1*n2*n3*3 (single prec. float)
+% Jphi   - Forward Jacobian tensors n1*n2*n3 (single prec. float)
+%
+% This code generates deformations and their Jacobian determinans from
+% initial velocity fields by gedesic shooting.  See the work of Miller,
+% Younes and others.
+%
+% LDDMM (Beg et al) uses the following evolution equation:
+%     d\phi/dt = v_t(\phi_t)
+% where a variational procedure is used to find the stationary solution
+% for the time varying velocity field.
+% In principle though, once the initial velocity is known, then the
+% velocity at subsequent time points can be computed.  This requires
+% initial momentum (m_0), computed (using differential operator L) by:
+%     m_0 = L v_0
+% Then (Ad_{\phi_t})^* m_0 is computed:
+%     m_t = |d \phi_t| (d\phi_t)^T m_0(\phi_t)
+% The velocity field at this time point is then obtained by using
+% multigrid to solve:
+%     v_t = L^{-1} m_t
+%
+% These equations can be found in:
+% Younes (2007). "Jacobi fields in groups of diffeomorphisms and
+% applications". Quarterly of Applied Mathematics, vol LXV,
+% number 1, pages 113-134 (2007).
+%
+% Multigrid is currently used to obtain v_t = L^{-1} m_t, but
+% this could also be done by convolution with the Greens function
+% N = L^{-1} (see e.g. Bro-Nielson).
+%
+%________________________________________________________
+% (c) Wellcome Trust Centre for NeuroImaging (2009)
+
+% John Ashburner
+% $Id: spm_shoot3di.m 4026 2010-07-29 13:45:50Z john $
+
+args0 = [8 4 4];
+if nargin<3,
+    args = args0;
+else
+    if numel(args)<numel(args0),
+        args = [args args0((numel(args)+1):end)];
+    end
+end
+verb     = false;
+N        = args(1);   % # Time steps
+fmg_args = args(2:3); % Multigrid params
+d        = size(v0);
+d        = d(1:3);
+vt       = v0;
+
+if ~isfinite(N),
+    % Number of time steps from an educated guess about how far to move
+    N = double(floor(sqrt(max(max(max(v0(:,:,:,1).^2+v0(:,:,:,2).^2+v0(:,:,:,3).^2)))))+1);
+end
+
+if verb, fprintf('N=%g:', N); end
+
+m0    = dartel3('vel2mom',v0,prm); % Initial momentum (m_0 = L v_0)
+if verb, fprintf('\t%g', 0.5*sum(v0(:).*m0(:))/prod(d)); end
+
+% Compute initial small deformation and Jacobian matrices from the velocity.
+% The overall deformation and its Jacobian tensor field is generated by
+% composing a series of small deformations.
+[theta,Jtheta] = dartel3('smalldef', vt,-1/N);
+
+% If required, also compute the forward and possibly also its Jacobian
+% tensor field. Note that the order of the compositions will be the
+% reverse of those for building the forward deformation.
+if nargout>=5,
+    [ phi, Jphi]  = dartel3('smalldef', vt,1/N);
+elseif nargout>=4,
+    phi           = dartel3('smalldef', vt,1/N);
+end
+
+for t=2:abs(N),
+    mt        = pullg(m0,theta,Jtheta);
+    vt        = mom2vel(mt,prm,fmg_args,vt);
+    if verb, fprintf('\t%g', 0.5*sum(vt(:).*mt(:))/prod(d)); end
+
+    [   dp,    dJ] = dartel3('smalldef',  vt,-1/N);    % Small deformation
+    [theta,Jtheta] = dartel3('comp', theta,dp, Jtheta,dJ); % Build up large def. from small defs
+    clear dp dJ
+
+    % If required, build up forward deformation and its Jacobian tensor field from
+    % small deformations
+    if nargout>=5,
+        [ dp,dJ]    = dartel3('smalldef',  vt,1/N);    % Small deformation
+        [phi, Jphi] = dartel3('comp', dp, phi, dJ, Jphi); % Build up large def. from small defs
+        clear dp dJ
+    elseif nargout>=4,
+        dp          = dartel3('smalldef',  vt,1/N);    % Small deformation
+        phi         = dartel3('comp', dp, phi);        % Build up large def. from small defs
+        clear dp
+    end
+    drawnow
+end
+if verb, fprintf('\n'); end
+
+
+varargout{1} = theta;
+varargout{2} = Jtheta;
+if nargout>=3,
+    mt           = pullg(m0,theta,Jtheta);
+    varargout{3} = mom2vel(mt,prm,fmg_args,vt);
+end
+if nargout>=4, varargout{4} = phi;  end
+if nargout>=5, varargout{5} = Jphi; end
+%__________________________________________________________________________________
+
+%__________________________________________________________________________________
+function vt = mom2vel(mt,prm,fmg_args,vt)
+% L^{-1} m_t
+
+r   = dartel3('vel2mom',vt,prm);
+vt  = dartel3('mom2vel', mt-r, [prm fmg_args])+vt;
+
+if false,
+    % Go for machine precision
+    r  = dartel3('vel2mom',vt,prm);
+    ss = sum(sum(sum(sum((mt-r).^2))));
+    for i=1:8,
+        oss = ss;
+        vt  = dartel3('mom2vel', mt-r, [prm fmg_args])+vt;
+        r   = dartel3('vel2mom',vt,prm);
+        ss  = sum(sum(sum(sum((mt-r).^2))));
+        if ss/oss>0.9, break; end
+    end
+end
+%__________________________________________________________________________________
+
+%__________________________________________________________________________________
+function mt = pullg(m0,phi,J)
+% The Ad* operation
+mt = zeros(size(m0),'single');
+for i=1:size(m0,3),
+    phii        = phi(:,:,i,:);
+    mr1         = dartel3('samp',m0(:,:,:,1), phii);
+    mr2         = dartel3('samp',m0(:,:,:,2), phii);
+    mr3         = dartel3('samp',m0(:,:,:,3), phii);
+
+    dj          = J(:,:,i,1,1).*(J(:,:,i,2,2).*J(:,:,i,3,3)-J(:,:,i,2,3).*J(:,:,i,3,2))...
+                 +J(:,:,i,2,1).*(J(:,:,i,1,3).*J(:,:,i,3,2)-J(:,:,i,1,2).*J(:,:,i,3,3))...
+                 +J(:,:,i,3,1).*(J(:,:,i,1,2).*J(:,:,i,2,3)-J(:,:,i,1,3).*J(:,:,i,2,2));
+
+    mt(:,:,i,1) = (J(:,:,i,1,1).*mr1 + J(:,:,i,2,1).*mr2 + J(:,:,i,3,1).*mr3).*dj;
+    mt(:,:,i,2) = (J(:,:,i,1,2).*mr1 + J(:,:,i,2,2).*mr2 + J(:,:,i,3,2).*mr3).*dj;
+    mt(:,:,i,3) = (J(:,:,i,1,3).*mr1 + J(:,:,i,2,3).*mr2 + J(:,:,i,3,3).*mr3).*dj;
+end
+%__________________________________________________________________________________
+
+%__________________________________________________________________________________
+
diff --git a/toolbox/Shoot/spm_shoot_blur.m b/toolbox/Shoot/spm_shoot_blur.m
new file mode 100644
index 0000000..b0881d4
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot_blur.m
@@ -0,0 +1,191 @@
+function [sig,a] = spm_shoot_blur(t,prm,its,sig)
+% A function for blurring ("smoothing") tissue probability maps
+% FORMAT [sig,a_new] = spm_shoot_blur(t,prm,its,sig)
+%     t   - sufficient statistics
+%     prm - regularisation parameters (2, 1,1,1, 1,0.01,0.01)
+%     its - max no. iterations (12)
+%     sig - optional starting estimates
+%
+%     sig - "smoothed" average
+%     a   - parameters
+%
+% The core of this procedure is described in:
+%     John Ashburner & Karl J. Friston.
+%     "Computing Average Shaped Tissue Probability Templates"
+%     NeuroImage, In Press, Accepted Manuscript, Available online 24 December 2008
+%
+% However, there is an additional modification such that the the null space
+% of the parameters is rotated out.
+%________________________________________________________
+% (c) Wellcome Trust Centre for NeuroImaging (2009)
+
+% John Ashburner
+% $Id: spm_shoot_blur.m 4174 2011-01-26 13:33:13Z john $
+
+d   = [size(t),1,1,1];
+if nargin<3, its = 16;                            end; % Maximum no. iterations
+if nargin<2, prm = [2, 1,1,1, 1,0.01,0.01];       end; % Default regularisation
+rits = [1 1]; % No. cycles and no. relaxation iterations
+
+W    = zeros([d(1:3) round(((d(4)-1)*d(4))/2)],'single'); % 2nd derivatives
+gr   = zeros([d(1:3),d(4)-1],'single');                   % 1st derivatives
+
+% Re-organise sufficient statistics to a form that is easier to work with
+t    = max(t,eps('single')*1000);
+s    = sum(t,4);
+for k=1:d(4),
+    t(:,:,:,k) = t(:,:,:,k)./s;
+end
+maxs     = max(s(:)); % Used for scaling the regularisation
+prm(end) = prm(end)+maxs*d(4)*1e-3;
+
+% Only d(4)-1 fields need to be estimated because sum(a,4) = 0.  This matrix
+% is used to rotate out the null space
+R     = null(ones(1,d(4)));
+
+% Initial starting estimates (if sig is passed)
+a     = zeros([d(1:3),d(4)-1],'single');
+if nargin>=4,
+    for z=1:d(3), % Loop over planes
+        sz = sig(:,:,z,:);
+        sz = min(max(sig,0),1);
+        sz(~isfinite(sz)) = 1/d(4);
+        sz = squeeze(log(double(sz*(1-d(4)*1e-3)+1e-3)));
+        for j1=1:(d(4)-1),
+            az = zeros(d(1:2));
+            for j2=1:d(4),
+                az = az + R(j2,j1)*sz(:,:,j2); % Note the rotation
+            end
+            a(:,:,z,j1) = az;
+        end
+        clear sz az
+    end
+end
+
+for i=1:its,
+
+    ll  = 0;
+    for z=1:d(3), % Loop over planes
+
+        % Compute softmax for this plane
+        sig = double(reshape(sftmax(a(:,:,z,:),R),[d(1:2),d(4)]));
+
+        % -ve log likelihood of the likelihood
+        ll  = ll - sum(sum(sum(log(sig).*reshape(t(:,:,z,:),[d(1:2),d(4)]),3).*s(:,:,z)));
+
+        % Compute first derivatives (d(4)-1) x 1 
+        grz = sig - double(reshape(t(:,:,z,:),[d(1:2),d(4)]));
+        for j1=1:(d(4)-1),
+            gr(:,:,z,j1) = 0;
+            for j2=1:d(4),
+                gr(:,:,z,j1) = gr(:,:,z,j1) + R(j2,j1)*grz(:,:,j2); % Note the rotation
+            end
+            gr(:,:,z,j1) = gr(:,:,z,j1).*s(:,:,z);
+        end
+
+        % Compute d(4) x d(4) matrix of second derivatives at each voxel.
+        % These should be positive definate, but rounding errors may prevent this.
+        % Regularisation is included to enforce +ve definateness.
+        wz = zeros([d(1:2),d(4),d(4)]);
+        for j1=1:d(4),
+            wz(:,:,j1,j1) =   (1-sig(:,:,j1)).*sig(:,:,j1).*s(:,:,z);
+            for j2=1:(j1-1),
+                wz(:,:,j1,j2) = -sig(:,:,j1) .*sig(:,:,j2).*s(:,:,z);
+                wz(:,:,j2,j1) = wz(:,:,j1,j2);
+            end
+        end
+
+        % First step of rotating 2nd derivatives to (d(4)-1) x (d(4)-1)
+        % by R'*W*R
+        wz1 = zeros([d(1:2),d(4),d(4)-1]);
+        for j1=1:d(4),
+            for j2=1:(d(4)-1),
+                tmp = zeros(d(1:2));
+                for j3=1:d(4),
+                    tmp = tmp + wz(:,:,j1,j3)*R(j3,j2);
+                end
+                wz1(:,:,j1,j2) = tmp;
+            end
+        end
+
+        % Second step of rotating 2nd derivatives to (d(4)-1) x (d(4)-1)
+        % by R'*W*R
+        wz = zeros([d(1:2),d(4)-1,d(4)-1]);
+        for j1=1:(d(4)-1),
+            for j2=1:(d(4)-1),
+                tmp = zeros(d(1:2));
+                for j3=1:d(4),
+                    tmp = tmp + R(j3,j1)*wz1(:,:,j3,j2);
+                end
+                wz(:,:,j1,j2) = tmp;
+            end
+        end
+
+        % First pull out the diagonal of the 2nd derivs
+        for j1=1:d(4)-1,
+            W(:,:,z,j1) = wz(:,:,j1,j1);% + maxs*sqrt(eps('single'))*d(4)^2;
+        end
+
+        % Then pull out the off diagonal parts (note that matrices are symmetric)
+        jj = d(4);
+        for j1=1:d(4)-1,
+           for j2=(j1+1):(d(4)-1),
+               W(:,:,z,jj) = wz(:,:,j2,j1);
+               jj = jj+1;
+           end
+        end
+    end
+
+    % ss1 and ss2 are for examining how close the 1st derivatives are to zero.
+    % At convergence, the derivatives from the likelihood term should match those
+    % from the prior (regularisation) term.
+    ss1 = sum(sum(sum(sum(gr.^2))));
+    gr1 = optimN('vel2mom',a,prm);        % 1st derivative of the prior term
+    ll1 = 0.5*sum(sum(sum(sum(gr1.*a)))); % -ve log probability of the prior term
+    gr  = gr + gr1;                       % Combine the derivatives of the two terms
+    ss2 = sum(sum(sum(sum(gr.^2))));      % This should approach zero at convergence
+    mx  = max(max(max(sum(gr.^2,4))));
+
+    fprintf('%d\t%g\t%g\t%g\t%g\t%g\t%g\n', i, ll/prod(d(1:3)),ll1/prod(d(1:3)), (ll+ll1)/prod(d(1:3)), ss2/ss1, mx, ss2/prod(d(1:3)));
+
+    reg = double(0.1*sqrt(mx)*d(4));
+   %reg = double(0.1*sqrt(ss2/prod(d(1:3))));
+    a   = a - optimN(W,gr,[prm(1:6) prm(7)+reg rits]); % Gauss-Newton update
+
+    if ss2/ss1<1e-4, break; end        % Converged?
+end
+
+sig = sftmax(a,R);
+%________________________________________________________
+
+%________________________________________________________
+function sig = sftmax(a,R)
+% Softmax function
+
+d     = [size(a) 1 1 1];
+sig   = zeros([d(1:3),d(4)+1],'single');
+trunc = log(realmax('single')*(1-eps('single'))/(d(4)+1));
+
+for j=1:size(a,3), % Loop over planes
+
+    % Rotate the null-space back in to the data
+    aj  = double(reshape(a(:,:,j,:),[d(1:2),d(4)]));
+    sj  = zeros([d(1:2),d(4)+1]);
+    for j1=1:d(4)+1,
+        sj(:,:,j1) = 0;
+        for j2=1:d(4),
+            sj(:,:,j1) = sj(:,:,j1) + R(j1,j2)*aj(:,:,j2);
+        end
+    end
+
+    % Compute softmax
+    sj = min(max(sj,-trunc),trunc);
+    sj = exp(sj)+eps('single')*(d(4)+1);
+    s  = sum(sj,3);
+    for i=1:d(4)+1,
+        sig(:,:,j,i) = single(sj(:,:,i)./s);
+    end
+
+end
+%________________________________________________________
+
diff --git a/toolbox/Shoot/spm_shoot_defaults.m b/toolbox/Shoot/spm_shoot_defaults.m
new file mode 100644
index 0000000..f7f91e3
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot_defaults.m
@@ -0,0 +1,46 @@
+function d = spm_shoot_defaults
+% Defaults file
+% FORMAT d = spm_shoot_defaults
+% This file contains settings that are intended to be customised
+% according to taste.  Some of them will influence the speed/accuracy
+% tradeoff, whereas others are various regularisation settings
+% (registration and template blurring)...
+
+%_______________________________________________________________________
+% Copyright (C) Wellcome Trust Centre for Neuroimaging (2009)
+
+% John Ashburner
+% $Id: spm_shoot_defaults.m 4174 2011-01-26 13:33:13Z john $
+
+
+%_______________________________________________________________________
+% The following settings are intended to be customised according to
+% taste.  Some of them will influence the speed/accuracy tradeoff,
+% whereas others are various regularisation settings (registration
+% and template blurring)...
+
+d.tname   = 'Template'; % Base file name for templates
+d.issym   = false;      % Use a symmetric template?
+
+d.cyc_its = [3 3];      % No. multigrid cycles and iterations
+
+nits      = 24;         % No. iterations of Gauss-Newton
+
+% Schedule for coarse to fine
+lam     = 0.5;          % Decay of coarse to fine schedule
+inter   = 64;           % Scaling of parameters at first iteration
+d.sched = (inter-1)*exp(-lam*((1:(nits+1))-1))+1;
+d.sched = d.sched/d.sched(end);
+
+maxoil    = 8;                 % Maximum number of time steps for integration
+d.eul_its = round((0:(nits-1))*(maxoil-0.5001)/(nits-1)+1); % Start with fewer steps
+
+%d.rparam  = [0 1.0 0.5  1e-3]; % Regularisation parameters for deformation - used for evaluations
+%d.rparam  = [0 0.5 1.0  1e-3]; % Regularisation parameters for deformation - used for evaluations
+d.rparam  = [0 0.5 1.0  1e-4]; % Regularisation parameters for deformation
+d.sparam  = [2 2.0 0.05 0.01]; % Regularisation parameters for blurring
+d.smits   = 16;                % No. smoothing iterations
+
+d.bs_args = [2 2 2  1 1 1];    % B-spline settings for interpolation
+%_______________________________________________________________________
+
diff --git a/toolbox/Shoot/spm_shoot_kernel.m b/toolbox/Shoot/spm_shoot_kernel.m
new file mode 100644
index 0000000..4d0bd40
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot_kernel.m
@@ -0,0 +1,57 @@
+function out = spm_shoot_kernel(job)
+% Generate kernel matrix from initial velocity fields
+% FORMAT spm_shoot_kernel(job)
+% job.velocities - Initial velocity fields
+% job.dotprod    - Part of filename for results
+%
+% k(x_1,x_2) = <x_1,L x_2> = <L x_1, x_2>
+%
+% This is very slow, and is not in a form that would be
+% suited to weighting according to location in the image.
+% For this, the "square root" of L would need to be used
+% in order to convert the flow fields into (e.g.) their
+% Jacobian tensor fields.  For linear elasticity, this
+% field would be decomposed by J = (J+J')/2 + (J-J')/2.
+% The elements of the symetric part (along with its trace)
+% would then be used to generate the kernel.
+%_______________________________________________________________________
+% Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
+
+% John Ashburner
+% $Id: spm_shoot_kernel.m 4026 2010-07-29 13:45:50Z john $
+
+defs   = spm_shoot_defaults;
+rparam = defs.rparam;
+P      = strvcat(job.velocities);
+[pth,nam,ext] = fileparts(job.dotprod);
+ofname = fullfile(pwd,['dp_' nam '.mat']);
+out.fname = {ofname};
+
+N   = nifti(P);
+dm  = size(N(1).dat);
+vx  = sqrt(sum(N(1).mat(1:3,1:3).^2));
+prm = [rparam(1) vx rparam(2:end)];
+n   = numel(N);
+K   = zeros(n,n);
+spm_progress_bar('Init',n*n,'Generating matrix','Elements done');
+for i=1:n,
+    x1 = single(squeeze(N(i).dat(:,:,:,end,:)));
+    x1 = dartel3('vel2mom',x1,prm);
+    for j=i:n,
+        x2       =squeeze(N(j).dat(:,:,:,end,:,:));
+        d        = x1(:)'*x2(:);
+        K(i,j)   = d;
+        K(j,i)   = d;
+
+        if ~rem(j,8)
+            spm_progress_bar('Set',...
+                n*n-(n+1-i)*(n+1-i)+(j-i)*2+1);
+        end
+    end
+    input        = job;
+    input.rparam = rparam;
+    typ          = 'initvel';
+    save(ofname,'K','input','typ');
+end
+spm_progress_bar('Clear');
+
diff --git a/toolbox/Shoot/spm_shoot_template.m b/toolbox/Shoot/spm_shoot_template.m
new file mode 100644
index 0000000..f66b026
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot_template.m
@@ -0,0 +1,377 @@
+function out = spm_shoot_template(job)
+% Iteratively compute a template with mean shape and intensities
+% format spm_shoot_template(job)
+% Fields of job:
+%     job.images{1} first set of images (eg rc1*.nii)
+%     job.images{2} second set of images (eg rc2*.nii)
+%     etc
+%
+% Other settings are defined in spm_shoot_defaults.m
+%
+% The outputs are flow fields (v*.nii), deformation fields (y*.nii)
+% and a series of Template images.
+%_______________________________________________________________________
+% Copyright (C) Wellcome Trust Centre for Neuroimaging (2009)
+
+% John Ashburner
+% $Id: spm_shoot_template.m 4264 2011-03-25 19:18:33Z john $
+
+%_______________________________________________________________________
+d       = spm_shoot_defaults;
+tname   = d.tname;   % Base file name for templates
+issym   = d.issym;   % Use a symmetric template
+
+cyc_its = d.cyc_its; % No. multigrid cycles and inerations
+smits   = d.smits;   % No. smoothing iterations
+sched   = d.sched;   % Schedule for coarse to fine
+nits    = numel(sched)-1;
+rparam  = d.rparam;  % Regularisation parameters for deformation
+sparam  = d.sparam;  % Regularisation parameters for blurring
+eul_its = d.eul_its; % Start with fewer steps
+
+bs_args = d.bs_args; % B-spline settings for interpolation
+%_______________________________________________________________________
+
+
+% Sort out handles to images
+n1 = numel(job.images);
+n2 = numel(job.images{1});
+NF = struct('NI',[],'vn',[1 1]);
+NF(n1,n2) = struct('NI',[],'vn',[1 1]);
+
+% Pick out individual volumes within NIfTI files
+for i=1:n1,
+    if numel(job.images{i}) ~= n2,
+        error('Incompatible number of images');
+    end;
+    for j=1:n2,
+        [pth,nam,ext,num] = spm_fileparts(job.images{i}{j});
+        NF(i,j).NI        = nifti(fullfile(pth,[nam ext]));
+        num               = [str2num(num) 1 1];
+        NF(i,j).vn        = num(1:2);
+    end;
+end;
+
+spm_progress_bar('Init',n2,'Initial mean','Subjects done');
+dm = [size(NF(1,1).NI.dat) 1];
+dm = dm(1:3);
+
+NU     = nifti;
+NU(n2) = nifti;
+NY     = nifti;
+NY(n2) = nifti;
+NJ     = nifti;
+NJ(n2) = nifti;
+
+t  = zeros([dm n1+1],'single');
+for i=1:n2,
+    % Generate files for flow fields, deformations and Jacobian determinants.
+    [pth,nam,ext]   = fileparts(NF(1,i).NI.dat.fname);
+
+    NU(i) = nifti;
+    NY(i) = nifti;
+    NJ(i) = nifti;
+
+    offs  = 352;
+    if ~isempty(tname),
+        NU(i).dat = file_array(fullfile(pth,['v_' nam '_' tname '.nii']),...
+                               [dm 1 3], 'float32-le', 352, 1, 0);
+        NY(i).dat = file_array(fullfile(pth,['y_' nam '_' tname '.nii']),...
+                               [dm 1 3], 'float32-le', 352, 1, 0);
+        NJ(i).dat = file_array(fullfile(pth,['j_' nam '_' tname '.nii']),...
+                               [dm 1 1], 'float32-le', 352, 1, 0);
+    else
+        NU(i).dat = file_array(fullfile(pth,['v_' nam '.nii']),...
+                               [dm 1 3], 'float32-le', 352, 1, 0);
+        NY(i).dat = file_array(fullfile(pth,['y_' nam '.nii']),...
+                               [dm 1 3], 'float32-le', 352, 1, 0);
+        NJ(i).dat = file_array(fullfile(pth,['y_' nam '.nii']),...
+                               [dm 1 1], 'float32-le', 352, 1, 0);
+    end
+
+    NU(i).descrip = sprintf('Velocity (%d %.4g %.4g %.4g)', rparam(1), rparam(2), rparam(3), rparam(4));
+    NU(i).mat     = NF(1,i).NI.mat;
+    NU(i).mat0    = NF(1,i).NI.mat0;
+
+    NY(i).descrip = 'Deformation (templ. to. ind.)';
+    NY(i).mat     = NF(1,i).NI.mat;
+
+    NJ(i).descrip = 'Jacobian det (templ. to. ind.)';
+    NJ(i).mat     = NF(1,i).NI.mat;
+
+
+    create(NU(i)); NU(i).dat(:,:,:,:,:) = 0;
+    create(NY(i)); NY(i).dat(:,:,:,:,:) = reshape(affind(dartel3('Exp',zeros([dm,3],'single'),[0 1]),NU(i).mat0),[dm,1,3]);
+    create(NJ(i)); NJ(i).dat(:,:,:)     = 1;
+
+    % Add to sufficient statistics for generating initial template
+    for j=1:n1,
+        vn         = NF(j,i).vn;
+        dat        = NF(j,i).NI.dat(:,:,:,vn(1),vn(2));
+        msk        = isfinite(dat);
+        dat(~msk)  = 0;
+        t(:,:,:,j) = t(:,:,:,j) + dat;
+    end;
+    t(:,:,:,end) = t(:,:,:,end) + msk;
+    clear tmp msk
+    spm_progress_bar('Set',i);
+end;
+spm_progress_bar('Clear');
+
+% Make symmetric (if necessary)
+if issym, t = t + t(end:-1:1,:,:,:); end
+
+% Generate template from sufficient statistics
+tmp = t(:,:,:,end);
+msk = tmp<=0;
+for j=1:n1,
+    tmp = tmp - t(:,:,:,j);
+end
+tmp(msk) = 0.01; % Most NaNs are likely to be background
+t(:,:,:,end) = tmp;
+clear tmp msk
+M  = NF(1,1).NI.mat;
+vx = sqrt(sum(M(1:3,1:3).^2));
+t  = max(t,0);
+g  = cell(n1+1,1);
+
+%%%%%%%%%%%%%%%
+% save SuffStats0.mat t sparam vx sched smits
+%%%%%%%%%%%%%%%
+
+if ~isempty(sparam),
+    g0 = spm_shoot_blur(t,[sparam(1), vx, sched(1)*sparam(2) sparam(3:4)],smits);
+    for j=1:n1+1,
+        g{j} = max(g0(:,:,:,j),1e-4);
+    end
+    clear g0
+else
+    sumt = max(sum(t,4),0)+eps;
+    for j=1:n1+1,
+        g{j} = (t(:,:,:,j)+0.1)./(sumt+0.1*(n1+1));
+    end
+    clear sumt
+end
+
+if ~isempty(tname),
+    % Write template
+    NG = NF(1,1).NI;
+    NG.descrip       = sprintf('Avg of %d', n2);
+    [tdir,nam,ext]   = fileparts(job.images{1}{1});
+    NG.dat.fname     = fullfile(tdir,[tname, '_0.nii']);
+    NG.dat.dim       = [dm n1+1];
+    NG.dat.dtype     = 'float32-le';
+    NG.dat.scl_slope = 1;
+    NG.dat.scl_inter = 0;
+    NG.mat0          = NG.mat;
+    create(NG);
+    for j=1:n1+1,
+        NG.dat(:,:,:,j)  = g{j};
+    end
+end
+for j=1:n1+1, g{j} = spm_bsplinc(log(g{j}), bs_args); end
+
+vx = sqrt(sum(NG.mat(1:3,1:3).^2));
+
+ok = true(n2,1);
+
+% The actual work
+for it=1:nits,
+
+    % More regularisation in the early iterations, as well as a
+    % a less accurate approximation in the integration.
+    prm      = [rparam(1), vx, rparam(2:3)*sched(it+1), rparam(4)];
+    int_args = [eul_its(it), cyc_its];
+    drawnow
+
+    t   = zeros([dm n1+1],'single');
+    su  = zeros([dm 3]);
+
+    % Update velocities
+    spm_progress_bar('Init',n2,sprintf('Update velocities (%d)',it),'Subjects done');
+    for i=1:n2, % Loop over subjects
+
+        if ok(i)
+            fprintf(' %-3d %-5d\t| ',it,i);
+
+            % Load image data for this subject
+            f = loadimage(NF(:,i));
+
+            % Load this subject's flow field and deformation
+            u  = squeeze(single(NU(i).dat(:,:,:,:,:)));
+            y  = affind(squeeze(single(NY(i).dat(:,:,:,:,:))),inv(NU(i).mat0));
+            dt = squeeze(single(NJ(i).dat(:,:,:)));
+            drawnow
+
+            % Gauss-Newton iteration to re-estimate deformations for this subject
+            u = spm_shoot_update(g,f,u,y,dt,prm,int_args,bs_args);
+            clear f y
+
+            drawnow
+            NU(i).dat(:,:,:,:,:) = reshape(u,[dm 1 3]);
+            su = su + u;
+            clear u
+            spm_progress_bar('Set',i);
+        end
+
+    end
+    spm_progress_bar('Clear');
+
+    if issym
+        su(:,:,:,1)   = (su(:,:,:,1)   - su(end:-1:1,:,:,1)  )/(sum(ok)*2);
+        su(:,:,:,2:3) = (su(:,:,:,2:3) + su(end:-1:1,:,:,2:3))/(sum(ok)*2);
+    else
+        su = su/sum(ok);
+    end
+    
+    % Update template sufficient statistics
+    spm_progress_bar('Init',n2,sprintf('Update deformations and template (%d)',it),'Subjects done');
+    for i=1:n2, % Loop over subjects
+
+        if ok(i)
+            % Load velocity, mean adjust and re-save
+            u  = squeeze(single(NU(i).dat(:,:,:,:,:)));
+            u  = u - su; % Subtract mean
+            NU(i).dat(:,:,:,:,:) = reshape(u,[dm 1 3]);
+
+            % Generate inverse deformation and save
+            [y,J] = spm_shoot3d(u,prm,int_args);
+            clear u
+            dt    = dartel3('det',J); clear J
+
+            if any(~isfinite(dt(:)) | dt(:)>100 | dt(:)<1/100)
+                ok(i) = false;
+                fprintf('Problem with %s (dets: %g .. %g)\n', NU(i).dat.fname, min(dt(:)), max(dt(:)));
+                clear dt
+            end
+        end
+
+        if ok(i)
+            NY(i).dat(:,:,:,:,:) = reshape(affind(y,NU(i).mat0),[dm 1 3]);
+            NJ(i).dat(:,:,:)     = dt;
+            drawnow;
+
+            % Load image data for this subject
+            f = loadimage(NF(:,i));
+
+            % Increment sufficient statistic for template
+            for j=1:n1+1,
+                tmp        = dartel3('samp',f{j},y);
+                t(:,:,:,j) = t(:,:,:,j) + tmp.*dt;
+                drawnow
+            end;
+
+            clear f y dt
+            fprintf('.');
+            spm_progress_bar('Set',i);
+        end
+
+    end
+    clear su
+    fprintf('\n');
+    spm_progress_bar('Clear');
+
+    % Make symmetric (if necessary)
+    if issym, t = t + t(end:-1:1,:,:,:); end
+
+    %%%%%%%%%%%%%%%
+    % save(['SuffStats' num2str(it) '.mat'],'t', 'sparam', 'vx', 'sched', 'smits', 'g');
+    %%%%%%%%%%%%%%%
+
+    % Re-generate template data from sufficient statistics
+    if ~isempty(sparam),
+        g0 = reconv(g,bs_args);
+        g0 = spm_shoot_blur(t,[sparam(1), vx, sched(it+1)*sparam(2) sparam(3:4)],smits,g0);
+        g  = cell(n1+1,1);
+        for j=1:n1+1,
+            g{j} = max(g0(:,:,:,j),1e-4);
+        end
+        clear g0
+    else
+        sumt = max(sum(t,4),0)+eps;
+        for j=1:n1+1,
+            g{j} = (t(:,:,:,j)+0.1)./(sumt+0.1*(n1+1));
+        end
+        clear sumt
+    end
+    clear t
+
+    % Write template
+    if ~isempty(tname),
+        NG.dat.fname    = fullfile(tdir,[tname '_' num2str(ceil(it/3)) '.nii']);
+        create(NG);
+        for j=1:n1+1,
+            NG.dat(:,:,:,j) = g{j};
+        end
+    end
+
+    % Compute template's B-spline coefficients
+    for j=1:n1+1, g{j} = spm_bsplinc(log(g{j}), bs_args); end
+    drawnow
+end
+
+if any(~ok)
+    fprintf('Problems with:\n');
+    for i=find(~ok)',
+        fprintf('\t%s\n', NU(i).dat.fname);
+    end
+end
+
+% Finish off
+[tdir,nam,ext] = fileparts(job.images{1}{1});
+out.template = cell(floor(nits/3)+1,1);
+if ~isempty(tname),
+    for it=0:floor(nits/3),
+        fname    = fullfile(tdir,[tname '_' num2str(it) '.nii']);
+        out.template{it+1} = fname;
+    end
+end
+out.vel = cell(n2,1);
+out.def = cell(n2,1);
+out.jac = cell(n2,1);
+for i=1:n2,
+    out.vel{i} = NU(i).dat.fname;
+    out.def{i} = NY(i).dat.fname;
+    out.jac{i} = NJ(i).dat.fname;
+end
+%=======================================================================
+
+%=======================================================================
+function y1 = affind(y0,M)
+% Affine transform of deformation
+y1 = zeros(size(y0),'single');
+for d=1:3,
+    y1(:,:,:,d) = y0(:,:,:,1)*M(d,1) + y0(:,:,:,2)*M(d,2) + y0(:,:,:,3)*M(d,3) + M(d,4);
+end
+%=======================================================================
+
+%=======================================================================
+function g0 = reconv(g,bs_args)
+d = [size(g{1}), 1];
+[i1,i2,i3]=ndgrid(1:d(1),1:d(2),1:d(3));
+g0 = zeros([d,numel(g)],'single');
+for k=1:numel(g),
+    g0(:,:,:,k) = max(exp(spm_bsplins(g{k},i1,i2,i3,bs_args)),1e-4);
+end
+%=======================================================================
+
+%=======================================================================
+function f = loadimage(NF)
+n1      = size(NF,1);
+f       = cell(n1+1,1);
+dm      = [NF(1).NI.dat.dim 1 1 1];
+dm      = dm(1:3);
+f{n1+1} = ones(dm,'single');
+for j=1:n1,
+    vn      = NF(j,1).vn;
+    f{j}    = single(NF(j,1).NI.dat(:,:,:,vn(1),vn(2)));
+    msk     = ~isfinite(f{j});
+    f{j}(msk) = 0;
+    f{n1+1} = f{n1+1} - f{j};
+    drawnow
+end
+f{n1+1}(msk) = 0.00001;
+%=======================================================================
+
+%=======================================================================
+
diff --git a/toolbox/Shoot/spm_shoot_update.m b/toolbox/Shoot/spm_shoot_update.m
new file mode 100644
index 0000000..412041c
--- /dev/null
+++ b/toolbox/Shoot/spm_shoot_update.m
@@ -0,0 +1,126 @@
+function [u0,ll1, ll2,grad_norm] = spm_shoot_update(g,f,u0,phi,dt,prm,int_args, bs_args,scale)
+% Shooting Of Diffeomorphisms (Spawn Of Dartel).
+% FORMAT u0 = spm_shoot_update(g,f,u0,phi,dt,prm,int_args, bs_args)
+% g        - template
+% f        - individual
+% u0       - initial velocity
+% phi      - deformation
+% dt       - Jacobian determinants
+% prm      - Parameters of differential operator
+% int_args - integration settings
+% bs_args  - interpolation settings
+% scale    - scaling of the update step
+%
+% u0       - updated initial velocity
+% ll1      - matching part of objective function
+% ll2      - regularisation part of objective function
+% grad_norm - Norm of the 1st derivatives
+%
+% The easiest way to figure out what this function does is to read the code.
+%________________________________________________________
+% (c) Wellcome Trust Centre for NeuroImaging (2009)
+
+% John Ashburner
+% $Id: spm_shoot_update.m 4103 2010-10-28 15:43:16Z john $
+
+if nargin<9, scale = 1.0; end
+scale = max(min(scale,1.0),0.0);
+
+d = [size(g{1}),1,1];
+d = d(1:3);
+
+if isempty(u0)
+    u0   = zeros([d,3],'single');
+end
+
+[ll1,b,A] = mnom_derivs(g,f,phi,dt, bs_args);
+
+m0      = dartel3('vel2mom',u0,prm);
+ll2     = 0.5*sum(sum(sum(sum(m0.*u0))));
+var1    = sum(sum(sum(sum(b.^2))));
+b       = b + m0;
+var2    = sum(sum(sum(sum(b.^2))));
+grad_norm = sqrt(var2/prod(d));
+fprintf('%-10.5g %-10.5g %-10.5g %-10.5g %-10.5g\n',...
+                            ll1/prod(d), ll2/prod(d), (ll1+ll2)/prod(d),...
+                            var2/(var1+eps), grad_norm);
+u0      = u0 - scale*dartel3('fmg',A, b, [prm 3 2]);
+clear A b
+%=======================================================================
+
+%=======================================================================
+function [ll,b,A] = mnom_derivs(g,f,phi,dt, bs_args)
+% Compute log-likelihood, first and second derivatives for multi-nomial matching
+% FORMAT [ll,b,A] = mnom_derivs(g,f,phi,dt,bs_args)
+% g       - cell array of template B-spline coefficients
+% f       - cell array of individual subject data
+% phi     - deformation field
+% dt       - Jacobian determinants
+% bs_args - B-spline arguments for sampling g.  Defaults to [2 2 2  1 1 1] if
+%           not supplied.
+%
+% ll      - log-likelihood
+% b       - first derivatives
+% A       - (approx) second derivatives (Fisher information)
+%
+%_______________________________________________________________________
+% Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
+
+if nargin<4,
+    bs_args = [2 2 2  1 1 1];
+end
+
+d  = [size(g{1}),1,1];
+d  = d(1:3);
+
+ll = 0;
+b  = zeros([d,3],'single');
+A  = zeros([d,6],'single');
+
+[id{1},id{2},id{3}] = ndgrid(1:d(1),1:d(2),ceil(bs_args(3)/2)+1);
+
+for z=1:d(3),
+    ind = z+(-ceil(bs_args(3)/2):ceil(bs_args(3)/2));
+    if bs_args(6), ind = rem(ind-1+d(3),d(3))+1; end
+
+    f1  = cell(size(f));
+    for k=1:numel(g),
+        % Note the fudge with the indexing because spm_bsplins only works for double.
+        [g1,d1,d2,d3] = spm_bsplins(double(g{k}(:,:,ind)),id{:},bs_args);
+        slice(k) = struct('mu',exp(g1),'d1',d1,'d2',d2,'d3',d3);
+        if isempty(phi)
+            f1{k} = f{k}(:,:,z);
+        else
+            f1{k} = dartel3('samp',f{k},phi(:,:,z,:));
+        end
+    end
+    s = zeros(d(1:2));
+    for k=1:numel(g), s           = s + slice(k).mu; end
+    for k=1:numel(g), slice(k).mu = slice(k).mu./s;  end
+
+    b(:,:,z,:)  = 0;
+    A(:,:,z,:)  = 0;
+    for k=1:numel(g),
+        tmp      = f1{k}.*log(slice(k).mu);
+        ll       = ll - sum(tmp(:));
+        tmp      = f1{k} - slice(k).mu.*dt(:,:,z);
+
+        b(:,:,z,1) = b(:,:,z,1) + tmp.*slice(k).d1;
+        b(:,:,z,2) = b(:,:,z,2) + tmp.*slice(k).d2;
+        b(:,:,z,3) = b(:,:,z,3) + tmp.*slice(k).d3;
+        for k1=1:numel(g),
+            if k1~=k,
+                tmp = -slice(k).mu.*slice(k1).mu.*dt(:,:,z);
+            else
+                tmp = max(slice(k).mu.*(1-slice(k1).mu),0).*dt(:,:,z);
+            end
+            A(:,:,z,1) = A(:,:,z,1) + tmp.*slice(k).d1.*slice(k1).d1;
+            A(:,:,z,2) = A(:,:,z,2) + tmp.*slice(k).d2.*slice(k1).d2;
+            A(:,:,z,3) = A(:,:,z,3) + tmp.*slice(k).d3.*slice(k1).d3;
+            A(:,:,z,4) = A(:,:,z,4) + tmp.*slice(k).d1.*slice(k1).d2;
+            A(:,:,z,5) = A(:,:,z,5) + tmp.*slice(k).d1.*slice(k1).d3;
+            A(:,:,z,6) = A(:,:,z,6) + tmp.*slice(k).d2.*slice(k1).d3;
+        end
+    end
+end
+
diff --git a/toolbox/Shoot/tbx_cfg_shoot.m b/toolbox/Shoot/tbx_cfg_shoot.m
new file mode 100644
index 0000000..9e1b47d
--- /dev/null
+++ b/toolbox/Shoot/tbx_cfg_shoot.m
@@ -0,0 +1,760 @@
+function shoot = tbx_cfg_shoot
+% MATLABBATCH Configuration file for toolbox 'Shoot Tools'
+
+% $Id: tbx_cfg_shoot.m 4136 2010-12-09 22:22:28Z guillaume $
+addpath(fullfile(spm('dir'),'toolbox','DARTEL'));
+addpath(fullfile(spm('dir'),'toolbox','Shoot'));
+
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images1         = cfg_files;
+images1.tag     = 'images';
+images1.name    = 'Images';
+images1.help    = {'Select a set of imported images of the same type to be registered by minimising a measure of difference from the template.'};
+images1.filter = 'image';
+images1.ufilter = '^r.*';
+images1.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images         = cfg_repeat;
+images.tag     = 'images';
+images.name    = 'Images';
+images.help    = {'Select the images to be warped together. Multiple sets of images can be simultaneously registered. For example, the first set may be a bunch of grey matter images, and the second set may be the white matter images of the same subjects.'};
+images.values  = {images1};
+images.val     = {images1};
+images.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% warp Run Shoot (create Templates)
+% ---------------------------------------------------------------------
+warp         = cfg_exbranch;
+warp.tag     = 'warp';
+warp.name    = 'Run Shooting (create Templates)';
+warp.val     = {images };
+warp.help    = {'Run the geodesic shooting nonlinear image registration procedure. This involves iteratively matching all the selected images to a template generated from their own mean. A series of Template*.nii files are generated, which become increasingly crisp as the registration proceeds.'};
+warp.prog = @spm_shoot_template;
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images1         = cfg_files;
+images1.tag     = 'images';
+images1.name    = 'Images';
+images1.help    = {'Select a set of imported images of the same type to be registered by minimising a measure of difference from the template.'};
+images1.filter = 'image';
+images1.ufilter = '^r.*';
+images1.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images         = cfg_repeat;
+images.tag     = 'images';
+images.name    = 'Images';
+images.help    = {'Select the images to be warped together. Multiple sets of images can be simultaneously registered. For example, the first set may be a bunch of grey matter images, and the second set may be the white matter images of the same subjects.'};
+images.values  = {images1 };
+images.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% rform Regularisation Form
+% ---------------------------------------------------------------------
+rform         = cfg_menu;
+rform.tag     = 'rform';
+rform.name    = 'Regularisation Form';
+rform.val     = {0};
+rform.help    = {'The registration is penalised by some ``energy'''' term.  Here, the form of this energy term is specified. Three different forms of regularisation can currently be used.'};
+rform.labels  = {
+                'Linear Elastic Energy'
+                'Membrane Energy'
+                'Bending Energy'
+}';
+rform.values  = {0 1 2};
+% ---------------------------------------------------------------------
+% template Template
+% ---------------------------------------------------------------------
+template         = cfg_files;
+template.tag     = 'templates';
+template.name    = 'Templates';
+template.help    = {'Select templates. Smoother templates should be used for the early iterations. Note that the template should be a 4D file, with the 4th dimension equal to the number of sets of images.'};
+template.filter = 'nifti';
+template.ufilter = '.*';
+template.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% warp1 Run Shooting (existing Templates)
+% ---------------------------------------------------------------------
+warp1         = cfg_exbranch;
+warp1.tag     = 'warp1';
+warp1.name    = 'Run Shoot (existing Templates)';
+warp1.val     = {images template };
+warp1.check   = @check_shoot_template;
+warp1.help    = {'Run the Shoot nonlinear image registration procedure to match individual images to pre-existing template data. Start out with smooth templates, and select crisp templates for the later iterations.'};
+%warp1.prog = @spm_shoot_warp;
+%warp1.vout = @vout_shoot_warp;
+% ---------------------------------------------------------------------
+% velocities Velocity fields
+% ---------------------------------------------------------------------
+velocities         = cfg_files;
+velocities.tag     = 'velocities';
+velocities.name    = 'Velocity fields';
+velocities.help    = {'The velocity fields store the deformation information. The same fields can be used for both forward or backward deformations (or even, in principle, half way or exaggerated deformations).'};
+velocities.filter = 'nifti';
+velocities.ufilter = '^v_.*';
+velocities.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images1         = cfg_files;
+images1.tag     = 'images';
+images1.name    = 'Images';
+images1.help    = {'Select images to be warped. Note that there should be the same number of images as there are velocity fields, such that each velocity field warps one image.'};
+images1.filter = 'nifti';
+images1.ufilter = '.*';
+images1.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images         = cfg_repeat;
+images.tag     = 'images';
+images.name    = 'Images';
+images.help    = {'The velocity field deformations can be applied to multiple images. At this point, you are choosing how many images each velocity field should be applied to.'};
+images.values  = {images1 };
+images.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% See later
+% ---------------------------------------------------------------------
+many_subj = cfg_branch;
+many_subj.tag = 'subjs';
+many_subj.name = 'Many Subjects';
+many_subj.val  = {velocities,images};
+many_subj.help = {[...
+'Select this option if you have many subjects to spatially normalise, ',...
+'but there are a small and fixed number of scans for each subject.']};
+% ---------------------------------------------------------------------
+% jactransf Modulation
+% ---------------------------------------------------------------------
+jactransf         = cfg_menu;
+jactransf.tag     = 'jactransf';
+jactransf.name    = 'Modulation';
+jactransf.val     = {0};
+jactransf.help    = {'This allows the spatially normalised images to be rescaled by the Jacobian determinants of the deformations. Note that the rescaling is only approximate for deformations generated using smaller numbers of time steps.'};
+jactransf.labels  = {
+                    'Pres. Concentration (No "modulation")'
+                    'Pres. Amount ("Modulation")'
+}';
+jactransf.values  = {0 1};
+% ---------------------------------------------------------------------
+% interp Interpolation
+% ---------------------------------------------------------------------
+interp         = cfg_menu;
+interp.tag     = 'interp';
+interp.name    = 'Interpolation';
+interp.val     = {1};
+interp.help    = {
+                  'The method by which the images are sampled when being written in a different space.'
+                  '    Nearest Neighbour:     - Fastest, but not normally recommended.'
+                  '    Bilinear Interpolation:     - OK for PET, or realigned fMRI.'
+                  '    B-spline Interpolation:     - Better quality (but slower) interpolation/* \cite{thevenaz00a}*/, especially       with higher degree splines.  Do not use B-splines when       there is any region of NaN or Inf in the images. '
+}';
+interp.labels  = {
+                 'Nearest neighbour'
+                 'Trilinear'
+                 '2nd Degree B-spline'
+                 '3rd Degree B-Spline '
+                 '4th Degree B-Spline '
+                 '5th Degree B-Spline'
+                 '6th Degree B-Spline'
+                 '7th Degree B-Spline'
+}';
+interp.values  = {0 1 2 3 4 5 6 7};
+% ---------------------------------------------------------------------
+% crt_warped Create Warped
+% ---------------------------------------------------------------------
+crt_warped         = cfg_exbranch;
+crt_warped.tag     = 'crt_warped';
+crt_warped.name    = 'Create Warped';
+crt_warped.val     = {velocities images jactransf interp };
+crt_warped.check   = @check_norm;
+crt_warped.help    = {'This allows spatially normalised images to be generated. Note that voxel sizes and bounding boxes can not be adjusted, and that there may be strange effects due to the boundary conditions used by the warping. Also note that the warped images are not in Talairach or MNI space. The coordinate system is that of the average shape and size of the subjects to which Shoot was applied. In order to have MNI-space normalised images, then the Deformations Utility can be used to compose the individual Shoot warps, with a deformation field that matches (e.g.) the Template grey matter generated by Shoot, with one of the grey matter volumes released with SPM.'};
+%crt_warped.prog = @spm_shoot_norm;
+%crt_warped.vout = @vout_norm;
+% ---------------------------------------------------------------------
+% velocities Velocity fields
+% ---------------------------------------------------------------------
+velocities         = cfg_files;
+velocities.tag     = 'velocities';
+velocities.name    = 'Velocity fields';
+velocities.help    = {'The velocity fields store the deformation information. The same fields can be used for both forward or backward deformations (or even, in principle, half way or exaggerated deformations).'};
+velocities.filter = 'nifti';
+velocities.ufilter = '^v_.*';
+velocities.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% jacdet Jacobian determinants
+% ---------------------------------------------------------------------
+jacdet         = cfg_exbranch;
+jacdet.tag     = 'jacdet';
+jacdet.name    = 'Jacobian determinants';
+jacdet.val     = {velocities};
+jacdet.help    = {'Create Jacobian determinant fields from velocities.'};
+%jacdet.prog = @spm_shoot_jacobian;
+%jacdet.vout = @vout_jacdet;
+% ---------------------------------------------------------------------
+% velocities Velocity fields
+% ---------------------------------------------------------------------
+velocities         = cfg_files;
+velocities.tag     = 'velocities';
+velocities.name    = 'Velocity fields';
+velocities.help    = {'The velocity fields store the deformation information. The same fields can be used for both forward or backward deformations (or even, in principle, half way or exaggerated deformations).'};
+velocities.filter = 'nifti';
+velocities.ufilter = '^v_.*';
+velocities.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images         = cfg_files;
+images.tag     = 'images';
+images.name    = 'Images';
+images.help    = {'Select the image(s) to be inverse normalised.  These should be in alignment with the template image generated by the warping procedure.'};
+images.filter = 'nifti';
+images.ufilter = '.*';
+images.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% interp Interpolation
+% ---------------------------------------------------------------------
+interp         = cfg_menu;
+interp.tag     = 'interp';
+interp.name    = 'Interpolation';
+interp.val{1} = double(1);
+interp.help    = {
+                  'The method by which the images are sampled when being written in a different space.'
+                  '    Nearest Neighbour:     - Fastest, but not normally recommended.'
+                  '    Bilinear Interpolation:     - OK for PET, or realigned fMRI.'
+                  '    B-spline Interpolation:     - Better quality (but slower) interpolation/* \cite{thevenaz00a}*/, especially       with higher degree splines.  Do not use B-splines when       there is any region of NaN or Inf in the images. '
+}';
+interp.labels = {
+                 'Nearest neighbour'
+                 'Trilinear'
+                 '2nd Degree B-spline'
+                 '3rd Degree B-Spline '
+                 '4th Degree B-Spline '
+                 '5th Degree B-Spline'
+                 '6th Degree B-Spline'
+                 '7th Degree B-Spline'
+}';
+interp.values  = {0 1 2 3 4 5 6 7};
+% ---------------------------------------------------------------------
+% crt_iwarped Create Inverse Warped
+% ---------------------------------------------------------------------
+crt_iwarped         = cfg_exbranch;
+crt_iwarped.tag     = 'crt_iwarped';
+crt_iwarped.name    = 'Create Inverse Warped';
+crt_iwarped.val     = {velocities images interp };
+crt_iwarped.help    = {'Create inverse normalised versions of some image(s). The image that is inverse-normalised should be in alignment with the template (generated during the warping procedure). Note that the results have the same dimensions as the ``velocity fields'''', but are mapped to the original images via the affine transformations in their headers.'};
+%crt_iwarped.prog = @spm_shoot_invnorm;
+%crt_iwarped.vout = @vout_invnorm;
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+velocityfield         = cfg_files;
+velocityfield.tag     =  'velocityfield';
+velocityfield.name    = 'Flow Field';
+velocityfield.filter  = 'nifti';
+velocityfield.ufilter = '^v_.*\.nii$';
+velocityfield.num     = [1 1];
+velocityfield.help    = {'Shoot velocity field for this subject.'};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+images        = cfg_files;
+images.tag    = 'images';
+images.name   = 'Images';
+images.filter = 'nifti';
+images.num    = [1 Inf];
+images.help   = {'Images for this subject to spatially normalise.'};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+subj      = cfg_branch;
+subj.tag  = 'subj';
+subj.name = 'Subject';
+subj.val  = {velocityfield,images};
+subj.help = {'Subject to be spatially normalized.'};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+few_subj        = cfg_repeat;
+few_subj.tag    = 'few_subj';
+few_subj.name   = 'Few Subjects';
+few_subj.values = {subj};
+few_subj.help   = {[...
+'Select this option if there are only a few subjects, each with many or ',...
+'a variable number of scans each. You will then need to specify a series of subjects, and the velocity field and images of each of them.']};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+way        = cfg_choice;
+way.tag    = 'data';
+way.name   = 'Select according to';
+way.values = {few_subj,many_subj};
+way.help   = {...
+['You may wish to spatially normalise only a few subjects, '...
+ 'but have many scans per subject (eg for fMRI), '...
+ 'or you may have lots of subjects, but with a small and fixed number '...
+ 'of scans for each of them (eg for VBM).  The idea is to chose the way of '...
+ 'selecting files that is easier.']};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+template        = cfg_files;
+template.tag    = 'template';
+template.name   = 'Shoot Template';
+template.filter = 'nifti';
+template.num    = [0 1];
+template.help   = {...
+['Select the final Template file generated by Shoot. This will be affine '...
+ 'registered with a TPM file, such that the resulting spatially normalised '...
+ 'images are closer aligned to MNI space. Leave empty if you do not wish to '...
+ 'incorporate a transform to MNI space '...
+ '(ie just click ``done'' on the file selector, without selecting any images).']};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+fwhm         = cfg_entry;
+fwhm.tag     = 'fwhm';
+fwhm.name    = 'Gaussian FWHM';
+fwhm.val     = {[8 8 8]};
+fwhm.strtype = 'e';
+fwhm.num     = [1 3];
+fwhm.help    = {'Specify the full-width at half maximum (FWHM) of the Gaussian blurring kernel in mm. Three values should be entered, denoting the FWHM in the x, y and z directions. Note that you can also specify [0 0 0], but any ``modulated'' data will show aliasing (see eg Wikipedia), which occurs because of the way the warped images are generated.'};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+preserve         = cfg_menu;
+preserve.tag     = 'preserve';
+preserve.name    = 'Preserve';
+preserve.help    = {
+'Preserve Concentrations: Smoothed spatially normalised images (sw*) represent weighted averages of the signal under the smoothing kernel, approximately preserving the intensities of the original images. This option is currently suggested for eg fMRI.'
+''
+'Preserve Total: Smoothed and spatially normalised images preserve the total amount of signal from each region in the images (smw*). Areas that are expanded during warping are correspondingly reduced in intensity. This option is suggested for VBM.'
+}';
+preserve.labels = {
+                   'Preserve Concentrations (no "modulation")'
+                   'Preserve Amount ("modulation")'
+}';
+preserve.values = {0 1};
+preserve.val    = {0};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+vox          = cfg_entry;
+vox.tag      = 'vox';
+vox.name     = 'Voxel sizes';
+vox.num      = [1 3];
+vox.strtype  = 'e';
+vox.def      = @(val)spm_get_defaults('defs.vox',val{:});
+vox.help     = {[...
+'Specify the voxel sizes of the deformation field to be produced. ',...
+'Non-finite values will default to the voxel sizes of the template image',...
+'that was originally used to estimate the deformation.']};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+bb           = cfg_entry;
+bb.tag       = 'bb';
+bb.name      = 'Bounding box';
+bb.strtype   = 'e';
+bb.num       = [2 3];
+bb.def       = @(val)spm_get_defaults('defs.bb',val{:});
+bb.help      = {[...
+'Specify the bounding box of the deformation field to be produced. ',...
+'Non-finite values will default to the bounding box of the template image',...
+'that was originally used to estimate the deformation.']};
+% ---------------------------------------------------------------------
+%
+% ---------------------------------------------------------------------
+nrm       = cfg_exbranch;
+nrm.tag   = 'mni_norm';
+nrm.name  = 'Normalise to MNI Space';
+nrm.val   = {template,way,vox,bb,preserve,fwhm};
+%nrm.prog  = @spm_shoot_norm_fun;
+%nrm.vout  = @vout_norm_fun;
+%nrm.check = @check_norm_fun;
+nrm.help  = {[...
+'Normally, Shoot generates warped images that align with the average-shaped template. ',...
+'This routine includes an initial affine regisration of the template (the final one ',...
+'generated by Shoot), with the TPM data released with SPM.'],[...
+'``Smoothed'''' (blurred) spatially normalised images are generated in such a ',...
+'way that the original signal is preserved. Normalised images are ',...
+'generated by a ``pushing'''' rather than a ``pulling'''' (the usual) procedure. ',...
+'Note that trilinear interpolation is used, and no masking is done.  It ',...
+'is therefore essential that the images are realigned and resliced ',...
+'before they are spatially normalised.  Alternatively, contrast images ',...
+'generated from unsmoothed native-space fMRI/PET data can be spatially ',...
+'normalised for a 2nd level analysis.'],[...
+'Two ``preserve'''' options are provided.  One of them should do the ',...
+'equavalent of generating smoothed ``modulated'''' spatially normalised ',...
+'images.  The other does the equivalent of smoothing the modulated ',...
+'normalised fMRI/PET, and dividing by the smoothed Jacobian determinants.']};
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images1         = cfg_files;
+images1.tag     = 'images';
+images1.name    = 'Images';
+images1.help    = {'Select tissue class images (one per subject).'};
+images1.filter = 'nifti';
+images1.ufilter = '^r.*c';
+images1.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% images Images
+% ---------------------------------------------------------------------
+images         = cfg_repeat;
+images.tag     = 'images';
+images.name    = 'Images';
+images.help    = {'Multiple sets of images are used here. For example, the first set may be a bunch of grey matter images, and the second set may be the white matter images of the same subjects.  The number of sets of images must be the same as was used to generate the template.'};
+images.values  = {images1 };
+images.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% velocities Velocity fields
+% ---------------------------------------------------------------------
+velocities         = cfg_files;
+velocities.tag     = 'velocities';
+velocities.name    = 'Velocity fields';
+velocities.help    = {'Select the velocity fields for each subject.'};
+velocities.filter = 'nifti';
+velocities.ufilter = '^v_.*';
+velocities.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% template Template
+% ---------------------------------------------------------------------
+template         = cfg_files;
+template.tag     = 'template';
+template.name    = 'Template';
+template.help    = {'Residual differences are computed between the warped images and template.'};
+template.filter = 'nifti';
+template.ufilter = '^Template.*';
+template.num     = [0 1];
+% ---------------------------------------------------------------------
+% fwhm Smoothing
+% ---------------------------------------------------------------------
+fwhm         = cfg_menu;
+fwhm.tag     = 'fwhm';
+fwhm.name    = 'Smoothing';
+fwhm.val     = {4};
+fwhm.help    = {'The residuals can be smoothed with a Gaussian to reduce dimensionality. More smoothing is recommended if there are fewer training images.'};
+fwhm.labels  = {
+               'None'
+               ' 2mm'
+               ' 4mm'
+               ' 6mm'
+               ' 8mm'
+               '10mm'
+               '12mm'
+               '14mm'
+               '16mm'
+}';
+fwhm.values  = {0 2 4 6 8 10 12 14 16};
+% ---------------------------------------------------------------------
+% resids Generate Residuals
+% ---------------------------------------------------------------------
+resids         = cfg_exbranch;
+resids.tag     = 'resids';
+resids.name    = 'Generate Residuals';
+resids.val     = {images velocities template fwhm };
+resids.check   = @check_resids;
+resids.help    = {'Generate residual images in a form suitable for computing a Fisher kernel. In principle, a Gaussian Process model can be used to determine the optimal (positive) linear combination of kernel matrices.  The idea would be to combine the kernel from the residuals, with a kernel derived from the velocity-fields. Such a combined kernel should then encode more relevant information than the individual kernels alone.'};
+%resids.prog = @spm_shoot_resids;
+%resids.vout = @vout_resids;
+% ---------------------------------------------------------------------
+% images Data
+% ---------------------------------------------------------------------
+images         = cfg_files;
+images.tag     = 'images';
+images.name    = 'Data';
+images.help    = {'Select images to generate dot-products from.'};
+images.filter = 'nifti';
+images.ufilter = '.*';
+images.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% weight Weighting image
+% ---------------------------------------------------------------------
+weight         = cfg_files;
+weight.tag     = 'weight';
+weight.name    = 'Weighting image';
+weight.val = {{}};
+weight.help    = {'The kernel can be generated so that some voxels contribute to the similarity measures more than others.  This is achieved by supplying a weighting image, which each of the component images are multiplied before the dot-products are computed. This image needs to have the same dimensions as the component images, but orientation information (encoded by matrices in the headers) is ignored. If left empty, then all voxels are weighted equally.'};
+weight.filter = 'image';
+weight.ufilter = '.*';
+weight.num     = [0 1];
+% ---------------------------------------------------------------------
+% dotprod Dot-product Filename
+% ---------------------------------------------------------------------
+dotprod         = cfg_entry;
+dotprod.tag     = 'dotprod';
+dotprod.name    = 'Dot-product Filename';
+dotprod.help    = {'Enter a filename for results (it will be prefixed by ``dp_'''' and saved in the current directory).'};
+dotprod.strtype = 's';
+dotprod.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% reskern Kernel from Resids
+% ---------------------------------------------------------------------
+reskern         = cfg_exbranch;
+reskern.tag     = 'reskern';
+reskern.name    = 'Kernel from Images';
+reskern.val     = {images weight dotprod };
+reskern.help    = {'Generate a kernel matrix from images. In principle, this same function could be used for generating kernels from any image data (e.g. ``modulated'''' grey matter). If there is prior knowledge about some region providing more predictive information (e.g. the hippocampi for AD), then it is possible to weight the generation of the kernel accordingly. The matrix of dot-products is saved in a variable ``Phi'''', which can be loaded from the dp_*.mat file. The ``kernel trick'''' can be used to convert these dot-products into distance measures for e.g. radial basis-function approaches.'};
+%reskern.prog = @spm_shoot_dotprods;
+% ---------------------------------------------------------------------
+% velocities Velocity fields
+% ---------------------------------------------------------------------
+velocities         = cfg_files;
+velocities.tag     = 'velocities';
+velocities.name    = 'Velocity fields';
+velocities.help    = {'Select the velocity fields for each subject.'};
+velocities.filter = 'nifti';
+velocities.ufilter = '^v_.*';
+velocities.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% dotprod Dot-product Filename
+% ---------------------------------------------------------------------
+dotprod         = cfg_entry;
+dotprod.tag     = 'dotprod';
+dotprod.name    = 'Dot-product Filename';
+dotprod.help    = {'Enter a filename for results (it will be prefixed by ``dp_'''' and saved in the current directory.'};
+dotprod.strtype = 's';
+dotprod.num     = [1 Inf];
+% ---------------------------------------------------------------------
+% velkern Kernel from Flows
+% ---------------------------------------------------------------------
+velkern         = cfg_exbranch;
+velkern.tag     = 'velkern';
+velkern.name    = 'Kernel from velocities';
+velkern.val     = {velocities dotprod};
+velkern.help    = {'Generate a kernel from velocity fields. The dot-products are saved in a variable ``Phi'''' in the resulting dp_*.mat file.'};
+velkern.prog = @spm_shoot_kernel;
+% ---------------------------------------------------------------------
+% kernfun Kernel Utilities
+% ---------------------------------------------------------------------
+kernfun         = cfg_choice;
+kernfun.tag     = 'kernfun';
+kernfun.name    = 'Kernel Utilities';
+kernfun.help    = {
+                   'Shoot can be used for generating matrices of dot-products for various kernel pattern-recognition procedures.'
+                   'The idea of applying pattern-recognition procedures is to obtain a multi-variate characterisation of the anatomical differences among groups of subjects. These characterisations can then be used to separate (eg) healthy individuals from particular patient populations. There is still a great deal of methodological work to be done, so the types of kernel that can be generated here are unlikely to be the definitive ways of proceeding.  They are only just a few ideas that may be worth trying out. The idea is simply to attempt a vaguely principled way to combine generative models with discriminative models (see the ``Pattern Recognition and Machine Learning'''' book by Chris Bishop for more ideas). Better ways (higher predictive accuracy) will eventually emerge.'
+                   'Various pattern recognition algorithms are available freely over the Internet. Possible approaches include Support-Vector Machines, Relevance-Vector machines and Gaussian Process Models. Gaussian Process Models probably give the most accurate probabilistic predictions, and allow kernels generated from different pieces of data to be most easily combined.'
+}';
+kernfun.values  = {velkern};
+% ---------------------------------------------------------------------
+% shoot Shoot Tools
+% ---------------------------------------------------------------------
+shoot         = cfg_choice;
+shoot.tag     = 'shoot';
+shoot.name    = 'Shoot Tools';
+shoot.help    = {
+                  'This toolbox is based around the ``Diffeomorphic Registration using Geodesic Shooting and Gauss-Newton Optimisation'''' paper, which has been submitted to NeuroImage. The idea is to register images by estimating an initial velocity field, which can then be integrated to generate both forward and backward deformations.  Currently, the software only works with images that have isotropic voxels, identical dimensions and which are in approximate alignment with each other. One of the reasons for this is that the approach assumes circulant boundary conditions, which makes modelling global rotations impossible. Because of these limitations, the registration should be based on images that have first been ``imported'''' via the New Segment toolbox.'
+                  'The next step is the registration itself, which involves the simultaneous registration of e.g. GM with GM, WM with WM and 1-(GM+WM) with 1-(GM+WM) (when needed, the 1-(GM+WM) class is generated implicitly, so there is no need to include this class yourself). This procedure begins by creating a mean of all the images, which is used as an initial template. Deformations from this template to each of the individual images are computed, and the template is then re-generated by applying the inverses of the deformations to the images and averaging. This procedure is repeated a number of times.'
+                  ''
+                  'This toolbox should be considered as only a beta (trial) version, and will include a number of (as yet unspecified) extensions in future updates.  Please report any bugs or problems to the SPM mailing list.'
+}';
+shoot.values  = {warp kernfun};
+
+%_______________________________________________________________________
+%
+%_______________________________________________________________________
+
+function dep = vout_initial_import(job)
+cls = {'GM', 'WM', 'CSF'};
+kk = 1;
+for k=1:3,
+    if isnumeric(job.(cls{k})) && job.(cls{k})
+        dep(kk)            = cfg_dep;
+        dep(kk).sname      = sprintf('Imported Tissue (%s)', cls{k});
+        dep(kk).src_output = substruct('.','cfiles','()',{':',k});
+        dep(kk).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+        kk = kk + 1;
+    end
+end
+if isnumeric(job.image) && job.image
+    dep(kk)            = cfg_dep;
+    dep(kk).sname      = sprintf('Resliced Original Images');
+    dep(kk).src_output = substruct('.','files');
+    dep(kk).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function chk = check_shoot_template(job)
+n1 = numel(job.images);
+n2 = numel(job.images{1});
+chk = '';
+for i=1:n1,
+    if numel(job.images{i}) ~= n2,
+        chk = 'Incompatible number of images';
+        break;
+    end;
+end;
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_shoot_template(job)
+
+if isa(job.template,'cfg_dep') || ~ ...
+        isempty(deblank(job.template))
+    for it=0:numel(job.param),
+        tdep(it+1)            = cfg_dep;
+        tdep(it+1).sname      = sprintf('Template (Iteration %d)', it);
+        tdep(it+1).src_output = substruct('.','template','()',{it+1});
+        tdep(it+1).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+    end
+else
+    tdep = cfg_dep;
+    tdep = tdep(false);
+end
+fdep            = cfg_dep;
+fdep.sname      = 'Flow Fields';
+fdep.src_output = substruct('.','files','()',{':'});
+fdep.tgt_spec   = cfg_findspec({{'filter','nifti'}});
+dep = [tdep fdep];
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_shoot_warp(job)
+dep            = cfg_dep;
+dep.sname      = 'Flow Fields';
+dep.src_output = substruct('.','files','()',{':'});
+dep.tgt_spec   = cfg_findspec({{'filter','nifti'}});
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function chk = check_norm(job)
+chk = '';
+PU = job.velocities;
+PI = job.images;
+n1 = numel(PU);
+for i=1:numel(PI),
+    if numel(PI{i}) ~= n1,
+        chk = 'Incompatible number of images';
+        break;
+    end
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function chk = check_norm_fun(job)
+chk = '';
+if isfield(job.data,'subjs')
+    PU = job.data.subjs.velocities;
+    PI = job.data.subjs.images;
+    n1 = numel(PU);
+    for i=1:numel(PI),
+        if numel(PI{i}) ~= n1,
+            chk = 'Incompatible number of images';
+            break;
+        end
+    end
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_norm(job)
+if job.jactransf,
+    sname = 'Warped Images - Jacobian Transformed';
+else
+    sname = 'Warped Images';
+end
+PU    = job.velocities;
+PI    = job.images;
+for m=1:numel(PI),
+    dep(m)            = cfg_dep;
+    dep(m).sname      = sprintf('%s (Image %d)',sname,m);
+    dep(m).src_output = substruct('.','files','()',{':',m});
+    dep(m).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+end
+n = numel(PI);
+for m=1:numel(PU),
+    dep(m+n)            = cfg_dep;
+    dep(m+n).sname      = sprintf('%s (Deformation %d)',sname,m);
+    dep(m+n).src_output = substruct('.','files','()',{m,':'});
+    dep(m+n).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_invnorm(job)
+PU    = job.velocities;
+PI    = job.images;
+
+for m=1:numel(PI),
+    dep(m)            = cfg_dep;
+    dep(m).sname      = sprintf('Inverse Warped Images (Image %d)',m);
+    dep(m).src_output = substruct('.','files','()',{':',m});
+    dep(m).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+end
+n = numel(PI);
+for m=1:numel(PU),
+    dep(m+n)            = cfg_dep;
+    dep(m+n).sname      = sprintf('Inverse Warped Images (Deformation %d)',m);
+    dep(m+n).src_output = substruct('.','files','()',{m,':'});
+    dep(m+n).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_jacdet(job)
+dep            = cfg_dep;
+dep.sname      = 'Jacobian Determinant Fields';
+dep.src_output = substruct('.','files','()',{':'});
+dep.tgt_spec   = cfg_findspec({{'filter','nifti'}});
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_norm_fun(job)
+if job.preserve,
+    sname = 'MNI Smo. Warped - Amount';
+else
+    sname = 'MNI Smo. Warped - Concentrations';
+end
+
+dep = cfg_dep;
+
+if isfield(job.data,'subj')
+    for m=1:numel(job.data.subj)
+        dep(m)            = cfg_dep;
+        dep(m).sname      = sprintf('%s (Deformation %d)',sname,m);
+        dep(m).src_output = substruct('{}',{m},'()',{':'});
+        dep(m).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+    end
+end
+
+if isfield(job.data,'subjs')
+    for m=1:numel(job.data.subjs.images),
+        dep(m)            = cfg_dep;
+        dep(m).sname      = sprintf('%s (Image %d)',sname,m);
+        dep(m).src_output = substruct('()',{':',m});
+        dep(m).tgt_spec   = cfg_findspec({{'filter','nifti'}});
+    end
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function chk = check_resids(job)
+chk = '';
+PU = job.velocities;
+PI = job.images;
+n1 = numel(PU);
+for i=1:numel(PI),
+    if numel(PI{i}) ~= n1,
+        chk = 'Incompatible number of images';
+        break;
+    end
+end
+%_______________________________________________________________________
+
+%_______________________________________________________________________
+function dep = vout_resids(job)
+dep = cfg_dep;
+dep.sname      = 'Residual Files';
+dep.src_output = substruct('.','files','()',{':'});
+dep.tgt_spec   = cfg_findspec({{'filter','nifti'}});
+%_______________________________________________________________________
+
+
+
+
diff --git a/toolbox/dcm_meeg/spm_L_priors.m b/toolbox/dcm_meeg/spm_L_priors.m
index b00f1f3..3d7addd 100644
--- a/toolbox/dcm_meeg/spm_L_priors.m
+++ b/toolbox/dcm_meeg/spm_L_priors.m
@@ -31,7 +31,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_L_priors.m 3738 2010-02-25 13:19:07Z vladimir $
+% $Id: spm_L_priors.m 4281 2011-03-31 19:49:57Z karl $
 
 % defaults
 %--------------------------------------------------------------------------
@@ -62,19 +62,19 @@
     
     case{'ECD'} % mean           and variance
         %------------------------------------------------------------------
-        pE.Lpos = dipfit.Lpos;   Lpos = V*eye(3*n);     % positions
-        pE.L    = sparse(3,n);   L    =   eye(3*n);     % orientations
+        pE.Lpos = dipfit.Lpos;   pC.Lpos = ones(3,n)*V;   % positions
+        pE.L    = sparse(3,n);   pC.L    = ones(3,n)*16;  % orientations
         
     case{'IMG'}
         %------------------------------------------------------------------
-        m       = dipfit.Nm;                            % number of modes
-        pE.Lpos = sparse(3,0);   Lpos =   eye(0,0);     % positions
-        pE.L    = sparse(m,n);   L    =   eye(m*n);     % modes
+        m       = dipfit.Nm;                              % number modes
+        pE.Lpos = sparse(3,0);   pC.Lpos = sparse(3,0);   % positions
+        pE.L    = sparse(m,n);   pC.L    = ones(m,n)*16;  % modes
         
     case{'LFP'}
         %------------------------------------------------------------------
-        pE.Lpos = sparse(3,0);   Lpos =   eye(0,0);     % positions
-        pE.L    = ones(1,m);     L    =   eye(m,m);     % gains
+        pE.Lpos = sparse(3,0);   pC.Lpos = sparse(3,0);   % positions
+        pE.L    = sparse(1,m);   pC.L    = ones(1,m)/16;  % gains
         
     otherwise
         warndlg('Unknown spatial model')
@@ -87,70 +87,35 @@
     
     case{'ERP','SEP'}
         %------------------------------------------------------------------
-        pE.J = sparse(1,[1 7 9],[0.2 0.2 0.6],1,9);       % 9 states
-        J    = diag(pE.J/64);
+        pE.J = sparse(1,[1 7 9],[0.2 0.8 0.6],1,9);       % 9 states
+        pC.J = pE.J/16;
+        
+    case{'CMC'}
+        %------------------------------------------------------------------
+        pE.J = sparse(1,[1 3 7],[0.2 0.8 0.2],1,8);       % 8 states
+        pC.J = sparse(1,[3 7],1,1,8)/16;
         
     case{'LFP'}
         %------------------------------------------------------------------
         pE.J = sparse(1,[1 7 9],[0.2 0.2 0.6],1,13);      % 13 states
-        J    = diag(pE.J/64);
+        pC.J = pE.J/16;
         
     case{'NMM'}
         %------------------------------------------------------------------
-        pE.J = sparse(1,[1,2,3],[0.1 0.1 1],1,9);          % 9 states
-        J    = sparse([1,2],[1,2],[1/128 1/128],9,9);
+        pE.J = sparse(1,[1,2,3],[0.1 0.1 1],1,9);         % 9 states
+        pC.J = sparse(1,[1,2],[1/32 1/128],1,9);
         
     case{'MFM'}
         %------------------------------------------------------------------
-        pE.J = sparse(1,[1,2,3],[0.1 0.1 1],1,36);         % 9 states
-        J    = sparse([1,2],[1,2],[1/128 1/128],36,36);    % 27 covariances
+        pE.J = sparse(1,[1,2,3],[0.1 0.1 1],1,36);        % 36 states =
+        pC.J = sparse(1,[1,2],[1/64 1/128],1,36);         % 9 1st + 27 2nd
         
     case{'DEM'}
         %------------------------------------------------------------------
-        pE.J = [];                                         % null
-        J    = [];
+        pE.J = [];                                        % null
+        pC.J = [];
         
     otherwise
         warndlg('Unknown neural model')
         
 end
-
-
-% Distance between homolgous sources (16mm)
-%--------------------------------------------------------------------------
-if symmetry, V = 16; else V = 0; end
-
-% symmetry constraints (based on Euclidean distance from mirror image)
-%==========================================================================
-switch type
-    
-    case{'ECD'}
-        if symmetry
-            
-            % correlation
-            %------------------------------------------------------------------
-            Mpos  = [-pE.Lpos(1,:); pE.Lpos(2,:); pE.Lpos(3,:)];
-            D = Inf*ones(n);
-            for i = 1:n
-                for j = 1:n
-                    if sign(pE.Lpos(1,i)) == sign(Mpos(1, j))
-                        D(i,j) = sqrt(sum(pE.Lpos(:,i) - Mpos(:,j)).^2);
-                    end
-                end
-            end
-            D     = (D + D')/2;
-            DD = zeros(n);
-            for i = 1:n
-                [M, I] = min(D(i, :));
-                if M<V                   
-                    DD(i, I) = 1;
-                end
-            end
-            
-            L     = L - kron(DD,diag([1 0 0])) + kron(DD,diag([0 1 1]));
-        end
-end
-
-% prior covariance
-%--------------------------------------------------------------------------
-pC  = spm_cat(spm_diag({pC, Lpos, exp(8)*L, J}));
diff --git a/toolbox/dcm_meeg/spm_api_erp.fig b/toolbox/dcm_meeg/spm_api_erp.fig
index 7830da9..cced324 100644
Binary files a/toolbox/dcm_meeg/spm_api_erp.fig and b/toolbox/dcm_meeg/spm_api_erp.fig differ
diff --git a/toolbox/dcm_meeg/spm_api_erp.m b/toolbox/dcm_meeg/spm_api_erp.m
index 90f23c8..b4e0ab0 100644
--- a/toolbox/dcm_meeg/spm_api_erp.m
+++ b/toolbox/dcm_meeg/spm_api_erp.m
@@ -6,7 +6,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_api_erp.m 3846 2010-04-27 17:25:13Z vladimir $
+% $Id: spm_api_erp.m 4232 2011-03-07 21:01:16Z karl $
  
 if nargin == 0 || nargin == 1  % LAUNCH GUI
  
@@ -68,6 +68,7 @@
 % Type of analysis
 %--------------------------------------------------------------------------
 % 'ERP'    - Event related responses
+% 'CSD'    - Cross-spectral density
 % 'SSR'    - Steady-state responses
 % 'IND'    - Induced responses
 % 'PHA'    - (for phase coupling)
@@ -81,6 +82,7 @@
 end
 switch model
     case{'ERP'}, set(handles.ERP,'Value',1);
+    case{'CSD'}, set(handles.ERP,'Value',2);
     case{'SSR'}, set(handles.ERP,'Value',2);
     case{'IND'}, set(handles.ERP,'Value',3);
     case{'PHA'}, set(handles.ERP,'Value',4);
@@ -93,6 +95,7 @@
 %--------------------------------------------------------------------------
 % 'ERP'    - (linear second order NMM slow)
 % 'SEP'    - (linear second order NMM fast)
+% 'CMC'    - (linear second order NMM Canonical microcircuit)
 % 'LFP'    - (linear second order NMM self-inhibition)
 % 'NMM'    - (nonlinear second order NMM first-order moments)
 % 'MFM'    - (nonlinear second order NMM second-order moments)
@@ -107,10 +110,10 @@
 switch model
     case{'ERP'}, set(handles.model,'Value',1);
     case{'SEP'}, set(handles.model,'Value',2);
-    case{'LFP'}, set(handles.model,'Value',3);
-    case{'NMM'}, set(handles.model,'Value',4);
-    case{'MFM'}, set(handles.model,'Value',5);
-    case{'DEM'}, set(handles.model,'Value',6);
+    case{'CMC'}, set(handles.model,'Value',3);
+    case{'LFP'}, set(handles.model,'Value',4);
+    case{'NMM'}, set(handles.model,'Value',5);
+    case{'MFM'}, set(handles.model,'Value',6);
     otherwise
 end
  
@@ -531,7 +534,7 @@ function design_Callback(hObject, eventdata, handles)
 set(handles.spatial_ok,    'Enable', 'on');
 
 switch handles.DCM.options.analysis
-    case{'SSR'}
+    case{'SSR','CSD'}
         set(handles.onset, 'Enable', 'off');
     otherwise
         set(handles.onset, 'Enable', 'on');
@@ -565,7 +568,7 @@ function design_Callback(hObject, eventdata, handles)
         %------------------------------------------------------------------
         Slocation = zeros(Nareas, 3);
         tmp       = get(handles.Slocation, 'String');
-        if ~isempty(tmp) & size(tmp,1) == Nareas
+        if ~isempty(tmp) && size(tmp,1) == Nareas
             for i = 1:Nareas
                 tmp2 = str2num(tmp(i, :));
                 if length(tmp2) ~= 3
@@ -600,7 +603,7 @@ function design_Callback(hObject, eventdata, handles)
         %------------------------------------------------------------------
         DCM.Lpos = zeros(3,0);
         try
-            set(handles.Slocation, 'String', {'assuming LFP in', Sname{1:Nchannels}});  
+            set(handles.Slocation, 'String', Sname(1:Nchannels));  
             set(handles.plot_dipoles,'enable','off')
         catch
             warndlg('There are more LFP channels than sources')
@@ -679,6 +682,7 @@ function spatial_back_Callback(hObject, eventdata, handles)
 function plot_dipoles_Callback(hObject, eventdata, handles)
  
 % read location coordinates
+%--------------------------------------------------------------------------
 tmp       = get(handles.Slocation, 'String');
 Slocation = [];
 if ~isempty(tmp) 
@@ -712,7 +716,7 @@ function plot_dipoles_Callback(hObject, eventdata, handles)
 n       = length(DCM.Sname);              % number of sources
 m       = size(DCM.xU.X,2);               % number of experimental inputs
 switch DCM.options.analysis
-    case{'SSR'}                           % for Steady-state responses
+    case{'SSR','CSD'}                     % for Steady-state responses
         l = n;                            % number of endogenous inputs
     otherwise
         l = length(DCM.options.onset);    % number of peristimulus inputs
@@ -951,7 +955,7 @@ function connectivity_back_Callback(hObject, eventdata, handles)
 set(handles.spatial_back,      'Enable', 'on');
 
 switch handles.DCM.options.analysis
-    case{'SSR'}
+    case{'SSR','CSD'}
         set(handles.onset,     'Enable', 'off');
     otherwise
         set(handles.onset,     'Enable', 'on');
@@ -1033,6 +1037,11 @@ function connectivity_back_Callback(hObject, eventdata, handles)
             otherwise
                 handles.DCM = spm_dcm_erp(handles.DCM);
         end
+        
+    % cross-spectral density model (complex)
+    %----------------------------------------------------------------------
+    case{'CSD'}
+        handles.DCM = spm_dcm_csd(handles.DCM);
 
     % cross-spectral density model (steady-state responses)
     %----------------------------------------------------------------------
@@ -1078,6 +1087,11 @@ function connectivity_back_Callback(hObject, eventdata, handles)
     %----------------------------------------------------------------------
     case{'ERP'}
         spm_dcm_erp_results(handles.DCM, Action);
+        
+    % Cross-spectral density model (complex)
+    %----------------------------------------------------------------------
+    case{'CSD'}
+        spm_dcm_csd_results(handles.DCM, Action);
  
     % Cross-spectral density model (steady-state responses)
     %----------------------------------------------------------------------
@@ -1176,6 +1190,36 @@ function BMS_Callback(hObject, eventdata, handles)
         set(handles.Spatial,    'String',{'IMG','ECD','LFP'});
         set(handles.Wavelet,    'Enable','off','String','-');
         set(handles.onset,      'Enable','on');
+        
+    % Cross-spectral density model (complex)
+    %----------------------------------------------------------------------
+    case{'CSD'}
+        Action = {
+              'spectral data',...
+              'Coupling (A)',...
+              'Coupling (B)',...
+              'Coupling (C)',...
+              'trial-specific effects',...
+              'Input',...
+              'Transfer functions',...
+              'Cross-spectra (sources)',...
+              'Cross-spectra (channels)',...
+              'Coherence (sources)',...
+              'Coherence (channels)',...
+              'Covariance (sources)',...
+              'Covariance (channels)',...
+              'Dipoles'};
+        try
+            set(handles.Nmodes, 'Value', handles.DCM.options.Nmodes);
+        catch
+            set(handles.Nmodes, 'Value', 4);
+        end
+        
+        set(handles.text20, 'String', 'modes');
+        set(handles.model,  'Enable','on');              
+        set(handles.Spatial,'String',{'IMG','ECD','LFP'});
+        set(handles.Wavelet,'Enable','on','String','Spectral density');
+        set(handles.onset,  'Enable','off');
 
  
     % Cross-spectral density model (steady-state responses)
@@ -1281,6 +1325,19 @@ function Wavelet_Callback(hObject, eventdata, handles)
 handles.DCM = spm_dcm_erp_dipfit(handles.DCM, 1);
 
 switch handles.DCM.options.analysis
+    
+    case{'CSD'}
+        
+        % cross-spectral density (if DCM.M.U (eigen-space) exists
+        %------------------------------------------------------------------
+        try
+            handles.DCM = spm_dcm_csd_data(handles.DCM);
+        end
+ 
+        % and display
+        %------------------------------------------------------------------
+        spm_dcm_csd_results(handles.DCM,'spectral data');
+        
  
     case{'SSR'}
         
@@ -1292,7 +1349,7 @@ function Wavelet_Callback(hObject, eventdata, handles)
  
         % and display
         %------------------------------------------------------------------
-        spm_dcm_ssr_results(handles.DCM,'Data');
+        spm_dcm_ssr_results(handles.DCM,'spectral data');
  
  
     case{'IND'}
@@ -1316,5 +1373,3 @@ function Wavelet_Callback(hObject, eventdata, handles)
 handles = reset_Callback(hObject, eventdata, handles);
 spm_api_nmm(handles.DCM)
 
-
-
diff --git a/toolbox/dcm_meeg/spm_api_nmm.fig b/toolbox/dcm_meeg/spm_api_nmm.fig
index 2b4d17b..bff1649 100644
Binary files a/toolbox/dcm_meeg/spm_api_nmm.fig and b/toolbox/dcm_meeg/spm_api_nmm.fig differ
diff --git a/toolbox/dcm_meeg/spm_api_nmm.m b/toolbox/dcm_meeg/spm_api_nmm.m
index 0f87036..8007f15 100644
--- a/toolbox/dcm_meeg/spm_api_nmm.m
+++ b/toolbox/dcm_meeg/spm_api_nmm.m
@@ -141,8 +141,9 @@ function kernels_Callback(hObject, eventdata, handles)
 M.f   = f;
 M.m   = size(M.pE.C,2);
 M.g   = {};
+M.ons = 32*ones(M.m,1);
 
-try, M.ons; catch M.ons = 32; end
+    
  
 % Volterra Kernels
 %==========================================================================
@@ -226,6 +227,7 @@ function kernels_Callback(hObject, eventdata, handles)
 %--------------------------------------------------------------------------
 % 'ERP'    - (linear second order NMM slow)
 % 'SEP'    - (linear second order NMM fast)
+% 'CMC'    - (linear second order NMM Canonical microcircuit)
 % 'LFP'    - (linear second order NMM self-inhibition)
 % 'NMM'    - (nonlinear second order NMM first-order moments)
 % 'MFM'    - (nonlinear second order NMM second-order moments)
@@ -239,9 +241,10 @@ function kernels_Callback(hObject, eventdata, handles)
 switch model
     case{'ERP'}, set(handles.model,'Value',1);
     case{'SEP'}, set(handles.model,'Value',2);
-    case{'LFP'}, set(handles.model,'Value',3);
-    case{'NMM'}, set(handles.model,'Value',4);
-    case{'MFM'}, set(handles.model,'Value',5);
+    case{'CMC'}, set(handles.model,'Value',3);
+    case{'LFP'}, set(handles.model,'Value',4);
+    case{'NMM'}, set(handles.model,'Value',5);
+    case{'MFM'}, set(handles.model,'Value',6);
     otherwise
 end
 
@@ -254,7 +257,9 @@ function kernels_Callback(hObject, eventdata, handles)
 end
 pE = handles.DCM.M.pE;
 pC = handles.DCM.M.pC;
-pC = spm_unvec(diag(pC),pE);
+try
+    pC = spm_unvec(diag(pC),pE);
+end
 
 % display connection switches later
 %--------------------------------------------------------------------------
@@ -322,7 +327,7 @@ function kernels_Callback(hObject, eventdata, handles)
                 'Units','Normalized',...
                 'Position',[x y sx sy],...
                 'Style','edit',...
-                'String',sprintf('%-4.2f',P(i)),...
+                'String',sprintf('%-4.2f',full(P(i))),...
                 'enable','on',...
                 'Tag','tmp',...
                 'BackgroundColor',color{1 + ~eval(sprintf('pC.%s(%i)',F{f},i))},...
@@ -342,7 +347,7 @@ function kernels_Callback(hObject, eventdata, handles)
                     'Units','Normalized',...
                     'Position',[x y sx sy],...
                     'Style','edit',...
-                    'String',sprintf('%-4.2f',P(i,j)),...
+                    'String',sprintf('%-4.2f',full(P(i,j))),...
                     'enable','on',...
                     'Tag','tmp',...
                     'BackgroundColor',color{1 + ~eval(sprintf('pC.%s(%i,%i)',F{f},i,j))},...
diff --git a/toolbox/dcm_meeg/spm_cmc_priors.m b/toolbox/dcm_meeg/spm_cmc_priors.m
new file mode 100644
index 0000000..dfd3fec
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_cmc_priors.m
@@ -0,0 +1,128 @@
+function [E,V] = spm_cmc_priors(A,B,C)
+% prior moments for a canonical microcircuit model
+% FORMAT [pE,pC] = spm_cmc_priors(A,B,C)
+%
+% A{3},B{m},C  - binary constraints on extrinsic connections
+%
+% pE - prior expectation - f(x,u,P,M)
+%
+% synaptic parameters
+%--------------------------------------------------------------------------
+%    pE.T - syaptic time constants
+%    pE.H - syaptic densities
+%    pE.S - activation function parameters
+%    pE.G - intrinsic connection strengths
+%
+% connectivity parameters
+%--------------------------------------------------------------------------
+%    pE.A - extrinsic
+%    pE.B - trial-dependent
+%    pE.C - stimulus input
+%    pE.D - delays
+%
+% stimulus and noise parameters
+%--------------------------------------------------------------------------
+%    pE.R - onset and dispersion
+%
+% pC - prior (co)variances
+%
+% Because priors are specified under log normal assumptions, most
+% parameters are simply scaling coefficients with a prior expectation
+% and variance of one.  After log transform this renders pE = 0 and
+% pC = 1;  The prior expectations of what they scale are specified in
+% spm_erp_fx
+%__________________________________________________________________________
+%
+% David O, Friston KJ (2003) A neural mass model for MEG/EEG: coupling and
+% neuronal dynamics. NeuroImage 20: 1743-1755
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_cmc_priors.m 4281 2011-03-31 19:49:57Z karl $
+ 
+% default: a single source model
+%--------------------------------------------------------------------------
+if nargin < 3
+    A   = {0 0 0};
+    B   = {};
+    C   = 1;
+end
+ 
+% disable log zero warning
+%--------------------------------------------------------------------------
+warning('off','MATLAB:log:logOfZero');
+n     = size(C,1);                                % number of sources
+u     = size(C,2);                                % number of inputs
+ 
+% parameters for neural-mass forward model
+%==========================================================================
+ 
+% restructure adjacency matrices
+%--------------------------------------------------------------------------
+D{1}  = A{1} | A{3};                              % forward
+D{2}  = A{1} | A{3};                              % forward
+D{3}  = A{2} | A{3};                              % backward
+D{4}  = A{2} | A{3};                              % backward
+A     = D;
+ 
+% extrinsic connectivity
+%--------------------------------------------------------------------------
+Q     = sparse(n,n);
+for i = 1:length(A)
+      A{i} = ~~A{i};
+    E.A{i} = A{i}*32 - 32;
+    V.A{i} = A{i}/16;
+    Q      = Q | A{i};                            % and lateral connections
+end
+ 
+% extrinsic connectivity - input-dependent scaling
+%--------------------------------------------------------------------------
+for i = 1:length(B)
+      B{i} = ~~B{i};
+    E.B{i} = 0*B{i};
+    V.B{i} = B{i}/8;
+    Q      = Q | B{i};
+end
+ 
+% extrinsic connectivity - where inputs enter
+%--------------------------------------------------------------------------
+C      = ~~C;
+E.C    = C*32 - 32;
+V.C    = C - C;
+ 
+% synaptic parameters
+%--------------------------------------------------------------------------
+m    = 4;                                         % number of intrinsic
+Q    = Q + speye(n,n);                            % allow intrinsic delays
+E.T  = sparse(n,4);   V.T  = sparse(n,4) + 1/16;  % time constants
+E.H  = sparse(n,1);   V.H  = sparse(n,1) + 1/16;  % synaptic density
+E.G  = sparse(n,m);   V.G  = sparse(n,m) + 1/16;  % intrinsic connectivity
+E.D  = sparse(n,n);   V.D  = Q/32;                % delay
+E.S  = 0;             V.S  = 1/32;                % slope of sigmoid
+
+ 
+% set stimulus parameters: onset and dispersion
+%--------------------------------------------------------------------------
+E.R    = sparse(u,2);  V.R   = ones(u,1)*[1/16 1/16];
+warning('on','MATLAB:log:logOfZero');
+ 
+return
+ 
+ 
+ 
+% demo for log-normal pdf
+%==========================================================================
+x  = (1:64)/16;
+for i = [2 16 32 128]
+    v = 1/i;
+    p = 1./x.*exp(-log(x).^2/(2*v))/sqrt(2*pi*v);
+    plot(x,p)
+    text(x(16),p(16),sprintf('variance = 1/%i',1/v))
+    hold on
+end
+xlabel('scaling')
+ylabel('density')
+grid on
+hold off
+axis square
diff --git a/toolbox/dcm_meeg/spm_csd_chf.m b/toolbox/dcm_meeg/spm_csd_chf.m
new file mode 100644
index 0000000..26b1248
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_csd_chf.m
@@ -0,0 +1,39 @@
+function [m,v] = spm_csd_chf(P,M,U)
+% Characteristic (expected) frequency of a NMM
+% FORMAT [G,w] = spm_csd_mtf(P,M,U)
+%
+% P - parameters
+% M - neural mass model structure
+% U - trial-specific effects
+%
+% m - expected frequency
+% v - dispersion
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_csd_chf.m 4095 2010-10-22 19:37:51Z karl $
+ 
+ 
+% compute log-spectral density (without noise)
+%==========================================================================
+P.a(1) = -0;
+P.a(2) = -32;
+P.b    = P.b - 32;
+P.c    = P.c - 32;
+ 
+[G,w]  = spm_csd_mtf(P,M,U);
+ 
+% compute moments (treating abs(G{i}) as a density)
+%--------------------------------------------------------------------------
+for i = 1:length(G)
+    p     = abs(G{i});
+    p     = p/sum(p);
+    m(i)  = w'*p;
+    v(i)  = ((w - m(i)).^2)'*p;
+end
+ 
+% average over trials
+%--------------------------------------------------------------------------
+m    = mean(m);
+v    = mean(v);
diff --git a/toolbox/dcm_meeg/spm_csd_mtf.m b/toolbox/dcm_meeg/spm_csd_mtf.m
new file mode 100644
index 0000000..010d72a
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_csd_mtf.m
@@ -0,0 +1,185 @@
+function [y,w,s] = spm_csd_mtf(P,M,U)
+% Spectral response of a NMM (transfer function x noise spectrum)
+% FORMAT [y,w,s] = spm_csd_mtf(P,M,U)
+%
+% P - parameters
+% M - neural mass model structure
+% U - trial-specific effects
+%
+% G - {G(N,nc,nc}} - cross-spectral density for nc channels {trials}
+%                  - for N frequencies in M.Hz [default 1:64Hz]
+% w - frequencies
+% s - directed transfer functions (complex)
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_csd_mtf.m 4281 2011-03-31 19:49:57Z karl $
+
+
+% compute log-spectral density
+%==========================================================================
+
+% frequencies of interest
+%--------------------------------------------------------------------------
+try
+    dt = 1/(2*round(M.Hz(end)));
+    N  = 1/dt;
+    If = round(linspace(M.Hz(1),M.Hz(end),length(M.Hz)));
+catch
+    N  = 128;
+    dt = 1/N;
+    If = 1:N/2;
+end
+f    = [1:N/2]';                         % frequencies
+w    = f(If);                            % frequencies selected
+
+% number of channels and exogenous (neuronal) inputs
+%--------------------------------------------------------------------------
+nc   = M.l;
+nu   = M.m;
+nx   = size(M.x,2);
+M.u  = sparse(nu,1);
+
+
+% solve for fixed point (with 64 ms burn in)
+%--------------------------------------------------------------------------
+S    = M;
+S.g  = {};
+V.u  = sparse(8,M.m);
+V.dt = 8/1000;
+x    = spm_int_L(P,S,V);
+x    = spm_unvec(x(end,:),S.x);
+M.x  = x;
+
+% spectrum of innovations (Gu) and noise (Gs and Gn)
+%--------------------------------------------------------------------------
+[Gu,Gs,Gn] = spm_csd_mtf_gu(P,M);
+
+% get prior means (delays)
+%--------------------------------------------------------------------------
+try
+    di = M.pF.D(1);                    % intrinsic delays
+    de = M.pF.D(2);                    % extrinsic delays
+catch
+    de = 16;
+    di = 1;
+end
+
+
+% trial-specific effects
+%==========================================================================
+try, X = U.X; catch, X = sparse(1,0); end
+
+
+% cycle over trials
+%--------------------------------------------------------------------------
+for  c = 1:size(X,1)
+    
+    % baseline parameters
+    %----------------------------------------------------------------------
+    Q  = P;
+    
+    % trial-specific effective connectivity
+    %----------------------------------------------------------------------
+    for i = 1:size(X,2)
+        
+        
+        % extrinsic connections
+        %------------------------------------------------------------------
+        for j = 1:length(A)
+            Q.A{j} = Q.A{j} + X(c,i)*P.B{i};
+        end
+        
+        % intrinsic connections
+        %----------------------------------------------------------------------
+        try
+            Q.H(:,1) = Q.H(:,1) + X(c,i)*diag(P.B{i});
+        catch
+            Q.G(:,1) = Q.G(:,1) + X(c,i)*diag(P.B{i});
+        end
+        
+    end
+    
+    
+    % delays
+    %----------------------------------------------------------------------
+    try
+        
+        % evaluate delay matrix
+        %------------------------------------------------------------------
+        De = exp(P.D);
+        Di = diag(diag(De));
+        De = De - Di;
+        De = De*de/1000;
+        Di = Di*di/1000;
+        De = kron(ones(nx,nx),De);
+        Di = kron(ones(nx,nx) - speye(nx,nx),Di);
+        D  = Di + De;
+        
+        % get delay operator
+        %------------------------------------------------------------------
+        D  = spm_dcm_delay(M,Q,D);
+        
+    catch
+        D  = 1;
+    end
+    
+    
+    % augment and bi-linearise (with intrinsic delays)
+    %----------------------------------------------------------------------
+    [M0,M1,L] = spm_bireduce(M,Q,D);
+    
+    % project onto spatial modes
+    %----------------------------------------------------------------------
+    try,    L = M.U'*L;   end
+    
+    % kernels
+    %----------------------------------------------------------------------
+    [K0,K1]   = spm_kernels(M0,M1,L,N,dt);
+    
+    % Transfer functions (FFT of kernel)
+    %----------------------------------------------------------------------
+    S     = fft(K1);
+    
+    % [cross]-spectral density from neuronal innovations
+    %----------------------------------------------------------------------
+    G     = zeros(N/2,nc,nc);
+    for i = 1:nc
+        for j = 1:nc
+            for k = 1:nu
+                Gij      = S(:,i,k).*conj(S(:,j,k));
+                Gij      = Gij((1:N/2) + 1).*Gu(:,k);
+                G(:,i,j) = G(:,i,j) + Gij;
+            end
+        end
+    end
+    
+    % save trial-specific frequencies of interest
+    %----------------------------------------------------------------------
+    y{c} = G(If,:,:);
+    s{c} = S(If,:,:);
+    
+end
+
+% and add channel noise
+%--------------------------------------------------------------------------
+for c = 1:length(y)
+    
+    G     = y{c};
+    for i = 1:nc
+        
+        % channel specific noise
+        %------------------------------------------------------------------
+        G(:,i,i) = G(:,i,i) + Gs(If,i);
+        
+        % and cross-spectral density from common channel noise
+        %------------------------------------------------------------------
+        for j = 1:nc
+            G(:,i,j) = G(:,i,j) + Gn(If);
+        end
+    end
+    y{c} = G;
+end
+
diff --git a/toolbox/dcm_meeg/spm_csd_mtf_gu.m b/toolbox/dcm_meeg/spm_csd_mtf_gu.m
new file mode 100644
index 0000000..4357852
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_csd_mtf_gu.m
@@ -0,0 +1,56 @@
+function [Gu,Gs,Gn,f] = spm_csd_mtf_gu(P,M)
+% Spectral desnities of innovations and noise for DCM for CSD
+% FORMAT [Gu,Gs,Gn,f] = spm_csd_mtf_gu(P,M)
+%
+% P   - parameters
+% M   - neural mass model structure
+%
+% Gu  - neuronal innovations
+% Gn  - channel noise (non-specific)
+% Gs  - channel noise (specific)
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_csd_mtf_gu.m 4281 2011-03-31 19:49:57Z karl $
+
+ 
+% frequencies of interest
+%--------------------------------------------------------------------------
+try
+    dt = 1/(2*round(M.Hz(end)));
+    N  = 1/dt;
+catch
+    N  = 128;
+end
+f     = [1:N/2]';
+
+
+% spectrum of innovations (Gu)
+%--------------------------------------------------------------------------
+for i = 1:size(P.a,2)
+    Gu(:,i) = exp(P.a(1,i))*f.^(-exp(P.a(2,i)));
+end
+
+% spectrum of channel noise (non-specific)
+%--------------------------------------------------------------------------
+Gn          = exp(P.b(1))*f.^(-exp(P.b(2)))*4; 
+
+
+% spectrum of channel noise (non-specific)
+%--------------------------------------------------------------------------
+for i = 1:size(P.c,2)
+    Gs(:,i) = exp(P.c(1,i))*f.^(-exp(P.c(2,i)))*4;
+end
+ 
+% add structured innovations - a discrete cosine set of order length(P.d)
+%--------------------------------------------------------------------------
+try
+    X  = spm_dctmtx(N/2,size(P.d,1) + 1);
+    Gu = Gu.*exp(X(:,2:end)*P.d);
+end
+if size(Gu,2) == 1, Gu = Gu*ones(1,M.m); end
+if size(Gs,2) == 1, Gs = Gs*ones(1,M.l); end
+
+
diff --git a/toolbox/dcm_meeg/spm_dcm_csd.m b/toolbox/dcm_meeg/spm_dcm_csd.m
new file mode 100644
index 0000000..1e7fcde
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd.m
@@ -0,0 +1,239 @@
+function DCM = spm_dcm_csd(DCM)
+% Estimate parameters of a DCM of (complex) cross-spectral density
+% FORMAT DCM = spm_dcm_csd(DCM)
+%
+% DCM
+%    name: name string
+%       M:  Forward model
+%              M.dipfit - lead-field specification
+%       xY: data   [1x1 struct]
+%       xU: design [1x1 struct]
+%
+%   Sname: cell of source name strings
+%       A: {[nr x nr double]  [nr x nr double]  [nr x nr double]}
+%       B: {[nr x nr double], ...}   Connection constraints
+%       C: [nr x 1 double]
+%
+%   options.Nmodes       - number of spatial modes
+%   options.Tdcm         - [start end] time window in ms
+%   options.Fdcm         - [start end] Frequency window in Hz
+%   options.D            - time bin decimation       (usually 1 or 2)
+%   options.spatial      - 'ECD', 'LFP' or 'IMG'     (see spm_erp_L)
+%   options.model        - 'ERP', 'SEP', 'CMC', 'LFP', 'NMM' or 'MFM'
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_csd.m 4281 2011-03-31 19:49:57Z karl $
+ 
+ 
+% check options
+%==========================================================================
+drawnow
+clear spm_erp_L
+name = sprintf('DCM_%s',date);
+ 
+% Filename and options
+%--------------------------------------------------------------------------
+try, DCM.name;                      catch, DCM.name = name;      end
+try, DCM.name;                      catch, DCM.name = 'DCM_SSR'; end
+try, model   = DCM.options.model;   catch, model    = 'NMM';     end
+try, spatial = DCM.options.spatial; catch, spatial  = 'LFP';     end
+try, Nm      = DCM.options.Nmodes;  catch, Nm = 8;               end
+ 
+% Spatial model
+%==========================================================================
+DCM.options.Nmodes = Nm;
+ 
+DCM  = spm_dcm_erp_dipfit(DCM, 1);
+Ns   = size(DCM.C,1);                                   % number of sources
+Nc   = DCM.M.dipfit.Nc;
+DCM  = spm_dcm_erp_data(DCM);
+DCM.M.dipfit.model = model;
+ 
+ 
+% Neural mass model
+%==========================================================================
+ 
+% prior moments on parameters
+%--------------------------------------------------------------------------
+[pE,pC]  = spm_dcm_neural_priors(DCM.A,DCM.B,DCM.C,model);
+ 
+% reduce prior conduction delays for LFP spatial models
+%--------------------------------------------------------------------------
+if strcmpi(spatial,'LFP')
+    DCM.M.Pf.D = [2 16];
+end
+ 
+% check to see if neuronal priors have already been specified
+%--------------------------------------------------------------------------
+try
+    if length(spm_vec(DCM.M.pE)) == length(spm_vec(pE));
+        pE = DCM.M.pE;
+        fprintf('Using existing priors\n')
+    end
+end
+ 
+% augment with priors on spatial model
+%--------------------------------------------------------------------------
+[pE,pC]  = spm_L_priors(DCM.M.dipfit,pE,pC);
+ 
+% augment with priors on endogenous inputs (neuronal) and noise
+%--------------------------------------------------------------------------
+[pE,pC]  = spm_ssr_priors(pE,pC);
+ 
+% intial states and equations of motion
+%--------------------------------------------------------------------------
+[x,f]    = spm_dcm_x_neural(pE,model);
+ 
+% create DCM
+%--------------------------------------------------------------------------
+DCM.M.IS = 'spm_csd_mtf';
+DCM.M.FS = 'spm_fs_csd';
+DCM.M.g  = 'spm_gx_erp';
+DCM.M.f  = f;
+DCM.M.x  = x;
+DCM.M.n  = length(spm_vec(x));
+DCM.M.pE = pE;
+DCM.M.pC = pC;
+DCM.M.hE = 8;
+DCM.M.hC = exp(-8);
+DCM.M.m  = Ns;
+DCM.M.u  = sparse(Ns,1);
+ 
+%-Feature selection using principal components (U) of lead-field
+%==========================================================================
+ 
+% Spatial modes
+%--------------------------------------------------------------------------
+if Nc < Nm
+    U     = speye(Nc,Nc);
+    DCM.M.U = U;
+else
+    dGdg  = spm_diff('spm_lx_erp',pE,DCM.M,1);
+    L     = spm_cat(dGdg);
+    U     = spm_svd(L*L',exp(-8));
+    try
+        U = U(:,1:Nm);
+    end
+    DCM.M.U = U;
+end
+ 
+% get data-features (in reduced eigenspace)
+%--------------------------------------------------------------------------
+DCM       = spm_dcm_csd_data(DCM);
+ 
+ 
+% complete model specification and invert
+%==========================================================================
+Nm        = size(DCM.M.U,2);                    % number of spatial modes
+Nt        = size(DCM.xY.y,1);                   % number of trials
+Nf        = size(DCM.xY.y{1},1);                % number of frequency bins
+DCM.M.l   = Nm;
+DCM.M.Hz  = DCM.xY.Hz;
+ 
+% precision of noise: AR(1/2)
+%--------------------------------------------------------------------------
+DCM.xY.Q    = {spm_Q(1/2,Nf,1)*diag(DCM.M.Hz)*spm_Q(1/2,Nf,1)};
+DCM.xY.X0   = sparse(Nf,0);
+
+
+% adjust gain to accommodate scaling differences among models and data
+%==========================================================================
+
+% cross-spectral data
+%--------------------------------------------------------------------------
+y        = spm_vec(DCM.xY.y);
+scale    = mean(abs(y));
+DCM.xY.y = spm_unvec(y/scale,DCM.xY.y);
+
+% check neural activity (without sensor noise) and extrinic coupling
+%--------------------------------------------------------------------------
+pE.b     = pE.b - 32;
+pE.c     = pE.c - 32;
+scale    = mean(abs(spm_vec(feval(DCM.M.IS,pE,DCM.M,DCM.xU))));
+while scale > 32
+    pE.A  = spm_unvec(spm_vec(pE.A) - 1/8,pE.A);
+    scale = mean(abs(spm_vec(feval(DCM.M.IS,pE,DCM.M,DCM.xU))));
+end
+DCM.M.pE.A = pE.A;
+DCM.M.U    = DCM.M.U/sqrt(scale)/2;
+
+
+% Variational Laplace: model inversion
+%==========================================================================
+[Qp,Cp,Eh,F] = spm_nlsi_GN(DCM.M,DCM.xU,DCM.xY);
+Ce           = exp(-Eh);
+
+ 
+% Data ID
+%--------------------------------------------------------------------------
+try
+    try
+        ID  = spm_data_id(feval(DCM.M.FS,DCM.xY.y,DCM.M));
+    catch
+        ID  = spm_data_id(feval(DCM.M.FS,DCM.xY.y));
+    end
+catch
+    ID  = spm_data_id(DCM.xY.y);
+end
+ 
+ 
+% Bayesian inference {threshold = prior} NB Prior on A,B and C = exp(0) = 1
+%==========================================================================
+warning('off','SPM:negativeVariance');
+dp  = spm_vec(Qp) - spm_vec(pE);
+Pp  = spm_unvec(1 - spm_Ncdf(0,abs(dp),diag(Cp)),Qp);
+warning('on', 'SPM:negativeVariance');
+ 
+ 
+% predictions (csd) and error (sensor space)
+%--------------------------------------------------------------------------
+Hc  = spm_csd_mtf(Qp,DCM.M,DCM.xU);                      % prediction
+Ec  = spm_unvec(spm_vec(DCM.xY.y) - spm_vec(Hc),Hc);     % prediction error
+ 
+ 
+% predictions (source space - cf, a LFP from virtual electrode)
+%--------------------------------------------------------------------------
+M             = DCM.M;    
+M.dipfit.type = 'LFP';
+ 
+qp        = Qp;
+qp.L      = sparse(1,Ns);           % set virtual electrode gain to unity
+qp.b      = qp.b - 32;              % and suppress non-specific and
+qp.c      = qp.c - 32;              % specific channel noise
+
+[Hs Hz dtf] = spm_csd_mtf(qp,M,DCM.xU);
+[ccf pst]   = spm_csd2ccf(Hs,DCM.M.Hz);
+[coh fsd]   = spm_csd2coh(Hs,DCM.M.Hz);
+DCM.dtf  = dtf;
+DCM.ccf  = ccf;
+DCM.coh  = coh;
+DCM.fsd  = fsd;
+DCM.pst  = pst;
+DCM.Hz   = Hz;
+
+ 
+% store estimates in DCM
+%--------------------------------------------------------------------------
+DCM.Ep = Qp;                   % conditional expectation
+DCM.Cp = Cp;                   % conditional covariance
+DCM.Pp = Pp;                   % conditional probability
+DCM.Hc = Hc;                   % conditional responses (y), channel space
+DCM.Rc = Ec;                   % conditional residuals (y), channel space
+DCM.Hs = Hs;                   % conditional responses (y), source space
+DCM.Ce = Ce;                   % ReML error covariance
+DCM.F  = F;                    % Laplace log evidence
+DCM.ID = ID;                   % data ID
+ 
+% and save
+%--------------------------------------------------------------------------
+DCM.options.Nmodes = Nm;
+ 
+if spm_check_version('matlab','7') >= 0
+    save(DCM.name, '-V6', 'DCM');
+else
+    save(DCM.name, 'DCM');
+end
+assignin('base','DCM',DCM)
+return
diff --git a/toolbox/dcm_meeg/spm_dcm_csd_data.m b/toolbox/dcm_meeg/spm_dcm_csd_data.m
new file mode 100644
index 0000000..6b25fc4
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd_data.m
@@ -0,0 +1,225 @@
+function DCM = spm_dcm_ssr_data(DCM)
+% gets cross-spectral density data-features using a VAR model
+% FORMAT DCM = spm_dcm_ssr_data(DCM)
+% DCM    -  DCM structure
+% requires
+%
+%    DCM.xY.Dfile        - name of data file
+%    DCM.M.U             - channel subspace
+%    DCM.options.trials  - trial to evaluate
+%    DCM.options.Tdcm    - time limits
+%    DCM.options.Fdcm    - frequency limits
+%    DCM.options.D       - Down-sampling
+%
+% sets
+%
+%    DCM.xY.pst     - Peristimulus Time [ms] sampled
+%    DCM.xY.dt      - sampling in seconds [s] (down-sampled)
+%    DCM.xY.U       - channel subspace
+%    DCM.xY.y       - cross spectral density over sources
+%    DCM.xY.csd     - cross spectral density over sources
+%    DCM.xY.It      - Indices of time bins
+%    DCM.xY.Ic      - Indices of good channels
+%    DCM.xY.Hz      - Frequency bins
+%    DCM.xY.code    - trial codes evaluated
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_csd_data.m 4095 2010-10-22 19:37:51Z karl $
+ 
+% Set defaults and Get D filename
+%-------------------------------------------------------------------------
+try
+    Dfile = DCM.xY.Dfile;
+catch
+    errordlg('Please specify data and trials');
+    error('')
+end
+
+% ensure spatial modes have been computed (see spm_dcm_ssr)
+%-------------------------------------------------------------------------
+try
+    DCM.M.U;
+catch
+    errordlg('Please estimate this model first');
+    error('')
+end
+
+% load D
+%--------------------------------------------------------------------------
+try
+    D = spm_eeg_load(Dfile);
+catch
+    try
+        [p,f]        = fileparts(Dfile);
+        D            = spm_eeg_load(f);
+        DCM.xY.Dfile = fullfile(pwd,f);
+    catch
+        try
+            [f,p]        = uigetfile('*.mat','please select data file');
+            name         = fullfile(p,f);
+            D            = spm_eeg_load(name);
+            DCM.xY.Dfile = fullfile(name);
+        catch
+            warndlg([Dfile ' could not be found'])
+            return
+        end
+    end
+end
+
+ 
+% indices of EEG channel (excluding bad channels)
+%--------------------------------------------------------------------------
+if ~isfield(DCM.xY, 'modality')
+    [mod, list] = modality(D, 0, 1);
+
+    if isequal(mod, 'Multimodal')
+        qstr = 'Only one modality can be modelled at a time. Please select.';
+        if numel(list) < 4
+            % Nice looking dialog. Will usually be OK
+            options = [];
+            options.Default = list{1};
+            options.Interpreter = 'none';
+            DCM.xY.modality = questdlg(qstr, 'Select modality', list{:}, options);
+        else
+            % Ugly but can accomodate more buttons
+            ind = menu(qstr, list);
+            DCM.xY.modality = list{ind};
+        end
+    else
+        DCM.xY.modality = mod;
+    end
+end
+
+
+
+if ~isfield(DCM.xY, 'Ic')
+    DCM.xY.Ic  = setdiff(D.meegchannels(DCM.xY.modality), D.badchannels);
+end
+
+Ic        = DCM.xY.Ic;
+Nc        = length(Ic);
+Nm        = size(DCM.M.U,2);
+DCM.xY.Ic = Ic;
+
+% options
+%--------------------------------------------------------------------------
+try
+    DT    = DCM.options.D;
+catch
+    DT    = 1;
+end
+try
+    trial = DCM.options.trials;
+catch
+    trial = D.nconditions;
+end
+ 
+% check data are not oversampled (< 4ms)
+%--------------------------------------------------------------------------
+if DT/D.fsample < 0.004
+    DT            = ceil(0.004*D.fsample);
+    DCM.options.D = DT;
+end
+ 
+ 
+% get peristimulus times
+%--------------------------------------------------------------------------
+try
+    
+    % time window and bins for modelling
+    %----------------------------------------------------------------------
+    DCM.xY.Time = 1000*D.time; % ms
+    T1          = DCM.options.Tdcm(1);
+    T2          = DCM.options.Tdcm(2);
+    [i, T1]     = min(abs(DCM.xY.Time - T1));
+    [i, T2]     = min(abs(DCM.xY.Time - T2));
+    
+    % Time [ms] of down-sampled data
+    %----------------------------------------------------------------------
+    It          = [T1:DT:T2]';               % indices - bins
+    DCM.xY.pst  = DCM.xY.Time(It);           % PST
+    DCM.xY.It   = It;                        % Indices of time bins
+    DCM.xY.dt   = DT/D.fsample;              % sampling in seconds
+    Nb          = length(It);                % number of bins
+    
+catch
+    errordlg('Please specify time window');
+    error('')
+end
+ 
+% get frequency range
+%--------------------------------------------------------------------------
+try
+    Hz1     = DCM.options.Fdcm(1);          % lower frequency
+    Hz2     = DCM.options.Fdcm(2);          % upper frequency
+catch
+    pst     = DCM.xY.pst(end) - DCM.xY.pst(1);
+    Hz1     = max(ceil(2*1000/pst),4);
+    if Hz1 < 8;
+        Hz2 = 48;
+    else
+        Hz2 = 128;
+    end
+end
+
+ 
+% Frequencies
+%--------------------------------------------------------------------------
+DCM.xY.Hz  = fix(Hz1:Hz2);             % Frequencies
+Nf         = length(DCM.xY.Hz);        % number of frequencies
+Ne         = length(trial);            % number of ERPs
+ 
+% get induced responses (use previous CSD results if possible) 
+%==========================================================================
+try
+    if size(DCM.xY.csd,2) == Ne;
+        if size(DCM.xY.csd{1},1) == Nf;
+            if size(DCM.xY.csd{1},2) == Nm;
+                DCM.xY.y  = spm_cond_units(DCM.xY.csd,'csd');
+                return
+            end
+        end
+    end
+end
+ 
+% Cross spectral density for each trial type
+%==========================================================================
+condlabels = D.condlist;
+for i = 1:Ne;
+   
+    % trial indices
+    %----------------------------------------------------------------------
+    c = D.pickconditions(condlabels{trial(i)});
+    
+    % use only the first 512 trial
+    %----------------------------------------------------------------------
+    try c = c(1:512); end
+    Nt    = length(c);
+
+    
+    % Get data
+    %----------------------------------------------------------------------
+    P     = zeros(Nf,Nm,Nm);
+    for j = 1:Nt
+        
+        fprintf('\nevaluating condition %i (trial %i)',i,j)
+        Y   = full(double(D(Ic,It,c(j))'*DCM.M.U));
+        mar = spm_mar(Y,8);
+        mar = spm_mar_spectra(mar,DCM.xY.Hz,1/DCM.xY.dt);
+        P   = P + mar.P;
+    end
+  
+    % store
+    %----------------------------------------------------------------------
+    DCM.xY.csd{i} = P/Nt;
+   
+end
+ 
+% place cross-spectral density in xY.y
+%==========================================================================
+DCM.xY.y    = spm_cond_units(DCM.xY.csd,'csd'); 
+DCM.xY.U    = DCM.M.U;
+DCM.xY.code = condlabels(trial);
+
diff --git a/toolbox/dcm_meeg/spm_dcm_csd_plot.m b/toolbox/dcm_meeg/spm_dcm_csd_plot.m
new file mode 100644
index 0000000..667c605
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd_plot.m
@@ -0,0 +1,126 @@
+function spm_dcm_csd_plot(DCM,i1,i2,C)
+% Plots the conditional density of coherence etc for a given connection
+% FORMAT spm_dcm_csd_plot(DCM,i,j,C)
+%
+% DCM - inverted DCM structure for CSD models
+% i   - target source (or channel mode)
+% j   - source source (or channel mode)
+% C   - flag for channels (as opposed to sources
+%
+% This routine is a graphics routine that plots the Bayesian confidence 
+% tubes around cross-covariance, coherence and phase delays as functions 
+% of lag and frequency. It also plots the conditional density over the 
+% delay. The confidence tubes (Bayesian confidence intervals) are 
+% approximated by sampling the underlying parameters from the 
+% [approximate] conditional density.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_csd_plot.m 4095 2010-10-22 19:37:51Z karl $
+ 
+ 
+% check options
+%==========================================================================
+try, C; catch, C = 0; end
+    
+ 
+% Compute conditional density from N samples (cf, numerical integration)
+%--------------------------------------------------------------------------
+N     = 32;
+D     = exp(DCM.Ep.D(i1,i2))*16;
+Cp    = spm_sqrtm(DCM.Cp);
+Qp    = spm_vec(DCM.Ep);
+Vp    = spm_unvec(diag(DCM.Cp),DCM.Ep);
+
+for i = 1:N
+    
+    % predictions (csd) and error (sensor space)
+    %----------------------------------------------------------------------
+    qp  = Qp + Cp*randn(size(Qp));
+    qp  = spm_unvec(qp,DCM.Ep);
+
+    % if source space is required
+    %----------------------------------------------------------------------
+    if ~C
+        DCM.M.dipfit.type = 'LFP';
+        qp.L  = qp.L - qp.L;
+        qp.b  = qp.b - 32;
+        qp.c  = qp.c - 32;
+    end
+    hc        = spm_csd_mtf(qp,DCM.M,DCM.xU);
+    
+    % functions
+    %----------------------------------------------------------------------
+    [ccf pst] = spm_csd2ccf(hc,DCM.M.Hz);
+    [coh fsd] = spm_csd2coh(hc,DCM.M.Hz);
+    
+    CCF(:,i)  = ccf{1}(:,i1,i2);
+    COH(:,i)  = coh{1}(:,i1,i2);
+    FSD(:,i)  = fsd{1}(:,i1,i2);
+    
+    fprintf('sample: %i(%i)\n',i,N);
+end
+ 
+ 
+% Plot
+%==========================================================================
+HZ  = DCM.Hz(:);
+PST = DCM.pst(:)*1000;
+ 
+
+% Covariance function
+%-------------------------------------------------------------------------- 
+subplot(2,2,1); hold off
+i   = abs(PST) < 128;
+CI  = 1.6449*std(CCF')';
+CC  = mean(CCF')';
+CI  = CI(i);
+CC  = CC(i);
+PST = PST(i);
+ 
+fill([PST; flipud(PST)],[CC + CI; flipud(CC - CI)],[1 1 1]*.8), hold on
+plot(PST,CC), hold on
+plot([0 0],[min(CC - CI) max(CC + CI)],'-.'), hold on
+plot([D D],[min(CC - CI) max(CC + CI)]),      hold on
+title('Covariance','FontSize',16)
+xlabel('lag (ms)')
+ylabel('covariance')
+axis square
+
+% Coherence function
+%-------------------------------------------------------------------------- 
+subplot(2,2,2); hold off
+CI  = 1.6449*std(COH')';
+CC  = mean(COH')';
+fill([HZ; flipud(HZ)],[CC + CI; flipud(CC - CI)],[1 1 1]*.8), hold on
+plot(HZ,CC), hold on
+title('Coherence','FontSize',16)
+xlabel('a.u.')
+ylabel('covariance')
+axis square
+ 
+% Phase delay function
+%--------------------------------------------------------------------------
+subplot(2,2,3); hold off
+CI  = 1000*1.6449*std(FSD')';
+CC  = 1000*mean(FSD')';
+fill([HZ; flipud(HZ)],[CC + CI; flipud(CC - CI)],[1 1 1]*.8), hold on
+plot(HZ,CC,[min(HZ) max(HZ)],[D D],'-.'), hold on
+title('Phase delay','FontSize',16)
+xlabel('delay (ms)')
+ylabel('covariance')
+axis square
+ 
+% Conditional density over delay
+%--------------------------------------------------------------------------
+x  = linspace(1/128,32,128);
+qp = spm_Npdf(log(x/16),DCM.Ep.D(i1,i2),Vp.D(i1,i2));
+ 
+subplot(2,2,4),hold off
+plot(x,qp,[D D],[0 max(qp)],'-.')
+title('Conduction delay','FontSize',16)
+xlabel('delay (ms)')
+ylabel('conditional density')
+axis square
+
diff --git a/toolbox/dcm_meeg/spm_dcm_csd_priors.m b/toolbox/dcm_meeg/spm_dcm_csd_priors.m
new file mode 100644
index 0000000..b2b8a02
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd_priors.m
@@ -0,0 +1,185 @@
+function [pE] = spm_dcm_csd_priors(M,U,Y,k)
+% Optimisation of priors
+% FORMAT [pE] = spm_dcm_csd_priors(M,U,Y,k)
+%__________________________________________________________________________
+%
+% M.IS - function name f(P,M,U) - generative model
+%        This function specifies the nonlinear model: 
+%        y = Y.y = IS(P,M,U) + X0*P0 + e
+%        were e ~ N(0,C).  For dynamic systems this would be an integration
+%        scheme (e.g. spm_int). spm_int expects the following:
+%
+%     M.f  - f(x,u,P,M)
+%     M.g  - g(x,u,P,M)
+%       x  - state variables
+%       u  - inputs or causes
+%       P  - free parameters
+%       M  - fixed functional forms and parameters in M
+%
+% M.FS - function name f(y,M)   - feature selection
+%        This [optional] function performs feature selection assuming the
+%        generalized model y = FS(y,M) = FS(IS(P,M,U),M) + X0*P0 + e
+%
+% M.P  - starting estimates for model parameters [optional]
+%
+% M.pE - prior expectation  - E{P}   of model parameters
+% M.pC - prior covariance   - Cov{P} of model parameters
+%
+% M.hE - prior expectation  - E{h}   of log-precision parameters
+% M.hC - prior covariance   - Cov{h} of log-precision parameters
+%
+% U.u  - inputs
+% U.dt - sampling interval
+%
+% Y.y  - outputs
+% Y.dt - sampling interval for outputs
+% Y.X0 - Confounds or null space      (over size(y,1) bins or all vec(y))
+% Y.Q  - q error precision components (over size(y,1) bins or all vec(y))
+%
+% k    - indices of parameter vector to search over
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_csd_priors.m 4095 2010-10-22 19:37:51Z karl $
+ 
+% reduce system to one source
+%==========================================================================
+ 
+% reduce data to s single channel
+%--------------------------------------------------------------------------
+y     = 0;
+for i = 1:length(Y.y)
+    for j = 1:size(Y.y{i},2)
+        y = y + Y.y{i}(:,j,j);
+    end
+end
+y     = real(y);
+ 
+% reduce model to a single source
+%--------------------------------------------------------------------------
+S.dipfit    = M.dipfit;
+S.dipfit.Ns = 1;
+S.dipfit.Nc = 1;
+model       = S.dipfit.model;
+[pE,pC]     = spm_dcm_neural_priors({1,1,1},[],1,model);
+[pE,pC]     = spm_L_priors(S.dipfit,pE,pC);
+[pE,pC]     = spm_ssr_priors(pE,pC);
+[x,f]       = spm_dcm_x_neural(pE,model);
+
+% single source model
+%--------------------------------------------------------------------------
+S.IS = 'spm_csd_mtf';
+S.g  = 'spm_gx_erp';
+S.f  = f;
+S.x  = x;
+S.n  = length(spm_vec(x));
+S.pE = pE;
+S.m  = 1;
+S.Hz = M.Hz;
+
+% unpack covariance
+%--------------------------------------------------------------------------
+if isstruct(pC);
+    pC = spm_diag(spm_vec(pC));
+end
+ 
+% find parameters that most affect expected frequency
+%==========================================================================
+try
+    k;
+catch
+    dfdp  = spm_diff('spm_csd_chf',S.pE,S,U,1);
+    dfdp  = pC*spm_vec(dfdp);
+    [q k] = sort(-abs(dfdp));
+end
+ 
+% and compute csd over two of them
+%--------------------------------------------------------------------------
+s     = linspace(-3,3,8);
+P     = [];
+X     = [];
+for i = 1:length(s)
+    for j = 1:length(s);
+        P(:,end + 1) = spm_vec(S.pE);
+        P(k(1),end)  = P(k(1),end) + sqrt(pC(k(1),k(1)))*s(i);
+        P(k(2),end)  = P(k(2),end) + sqrt(pC(k(2),k(2)))*s(j);
+        [G w]        = spm_csd_mtf(spm_unvec(P(:,end),S.pE),S,U);
+        X(:,end + 1) = spm_vec(G);
+    end
+end
+ 
+% find the best match with average response
+%==========================================================================
+for i = 1:size(P,2)
+    L      = [w.^0 1./w X(:,i)/norm(X(:,i))];
+    b(:,i) = (L\y);
+    g(:,i) = L*b(:,i);
+    ssq(i) = sum((y - g(:,i)).^2);
+end
+[q i] = min(ssq);
+P     = P(:,i);
+G     = g(:,i);
+B     = b(:,i);
+ 
+subplot(2,1,1)
+plot(w,y,w,G,':')
+title('optimised priors')
+xlabel('frequency')
+ylabel('power')
+axis square
+drawnow
+ 
+% adjust original (neuronal) priors
+%--------------------------------------------------------------------------
+pE     = M.pE;
+pS     = spm_unvec(P,S.pE);
+feilds = fieldnames(pS);
+for  i = 1:length(feilds)
+    try
+        pF = getfield(pS,feilds{i});
+        pM = getfield(pE,feilds{i});
+        if isvector(pM)
+            if any(pM - pF)
+                pF = pM - pM + pF;
+                pE = setfield(pE,feilds{i},pF);
+                fprintf('paramters %s optimised\n',feilds{i})
+            end
+        end
+    end
+end
+
+ 
+return
+ 
+ 
+% show expected frequency as a function of two parameters
+%==========================================================================
+dfdp  = spm_diff('spm_csd_chf',S.pE,S,U,1);
+dfdp  = pC*spm_vec(dfdp);
+[q k] = sort(-abs(dfdp));
+s     = linspace(-4,4,64);
+for i = 1:length(s)
+    for j = 1:length(s);
+        P        = spm_vec(S.pE);
+        P(k(1))  = P(k(1)) + sqrt(pC(k(1),k(1)))*s(i);
+        P(k(2))  = P(k(2)) + sqrt(pC(k(2),k(2)))*s(j);
+        P        = spm_unvec(P,S.pE);
+        Ew(i,j)  = spm_csd_chf(P,S,U);
+    end
+end
+ 
+subplot(2,2,1)
+imagesc(s,s,Ew)
+title('expected frequency')
+xlabel(sprintf('parameter %i',k(2)))
+ylabel(sprintf('parameter %i',k(1)))
+axis square
+ 
+subplot(2,2,2)
+plot(s,Ew)
+title('expected frequency')
+xlabel(sprintf('parameter %i',k(2)))
+ylabel(sprintf('parameter %i',k(1)))
+axis square
diff --git a/toolbox/dcm_meeg/spm_dcm_csd_results.m b/toolbox/dcm_meeg/spm_dcm_csd_results.m
new file mode 100644
index 0000000..0ce6c8f
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd_results.m
@@ -0,0 +1,745 @@
+function [DCM] = spm_dcm_csd_results(DCM,Action)
+% Results for ERP Dynamic Causal Modeling (DCM)
+% FORMAT spm_dcm_csd_results(DCM,'spectral data');
+% FORMAT spm_dcm_csd_results(DCM,'Coupling (A)');
+% FORMAT spm_dcm_csd_results(DCM,'Coupling (B)');
+% FORMAT spm_dcm_csd_results(DCM,'Coupling (C)');
+% FORMAT spm_dcm_csd_results(DCM,'trial-specific effects');
+% FORMAT spm_dcm_csd_results(DCM,'Input');
+% FORMAT spm_dcm_csd_results(DCM,'Transfer functions');
+% FORMAT spm_dcm_csd_results(DCM,'Cross-spectra (sources)')
+% FORMAT spm_dcm_csd_results(DCM,'Cross-spectra (channels)')
+% FORMAT spm_dcm_csd_results(DCM,'Coherence (sources)')
+% FORMAT spm_dcm_csd_results(DCM,'Coherence (channels)')
+% FORMAT spm_dcm_csd_results(DCM,'Covariance (sources)')
+% FORMAT spm_dcm_csd_results(DCM,'Covariance (channels)')
+% FORMAT spm_dcm_csd_results(DCM,'Dipoles');
+%                
+%___________________________________________________________________________
+%
+% DCM is a causal modelling procedure for dynamical systems in which
+% causality is inherent in the differential equations that specify the model.
+% The basic idea is to treat the system of interest, in this case the brain,
+% as an input-state-output system.  By perturbing the system with known
+% inputs, measured responses are used to estimate various parameters that
+% govern the evolution of brain states.  Although there are no restrictions
+% on the parameterisation of the model, a bilinear approximation affords a
+% simple re-parameterisation in terms of effective connectivity.  This
+% effective connectivity can be latent or intrinsic or, through bilinear
+% terms, model input-dependent changes in effective connectivity.  Parameter
+% estimation proceeds using fairly standard approaches to system
+% identification that rest upon Bayesian inference.
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_csd_results.m 4281 2011-03-31 19:49:57Z karl $
+ 
+ 
+% get figure handle
+%--------------------------------------------------------------------------
+Fgraph = spm_figure('GetWin','Graphics');
+colormap(gray)
+figure(Fgraph)
+clf
+ 
+% get action if neccessary
+%--------------------------------------------------------------------------
+if nargin == 1
+    str    = {'spectral data',...
+              'Coupling (A)',...
+              'Coupling (B)',...
+              'Coupling (C)',...
+              'trial-specific effects',...
+              'Input',...
+              'Transfer functions',...
+              'Cross-spectra (sources)',...
+              'Cross-spectra (channels)',...
+              'Coherence (sources)',...
+              'Coherence (channels)',...
+              'Covariance (sources)',...
+              'Covariance (channels)',...
+              'Dipoles'};
+          
+    s      = listdlg('PromptString','Select an option:',...
+                     'SelectionMode','single',...
+                     'ListString',str);
+    Action = str{s};
+end
+ 
+ 
+% place spectral features in xY.y
+%--------------------------------------------------------------------------
+DCM.xY.y  = spm_cond_units(DCM.xY.csd,'csd');
+ 
+% trial data
+%--------------------------------------------------------------------------
+xY  = DCM.xY;                   % data
+nt  = length(xY.y);             % Nr trial types
+nf  = size(xY.y{1},1);          % Nr frequency bins
+nm  = size(xY.y{1},2);          % Nr spatial modes
+Hz  = xY.Hz;                    % PST
+ 
+% switch
+%--------------------------------------------------------------------------
+switch(lower(Action))    
+    
+case{lower('spectral data')}
+    
+    % spm_dcm_ssr_results(DCM,'Data');
+    %----------------------------------------------------------------------
+    co    = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    Hz    = xY.Hz;
+    nm    = min(nm,4);
+    name  = DCM.xY.name;
+    
+    for k = 1:nt
+        str{k} = sprintf('trial %i',k);
+    end
+    
+    for i = 1:nm
+        for j = i:nm
+ 
+            % for each trial type
+            %--------------------------------------------------------------
+            subplot(nm,nm,(i - 1)*nm + j),cla
+            for k = 1:nt
+                plot(Hz,real(xY.y{k}(:,i,j)),'color',co{k}), hold on
+                plot(Hz,imag(xY.y{k}(:,i,j)),':','color',co{k}), hold on
+                title(sprintf('%s to %s',name{j},name{i}))
+                axis square, spm_axis tight
+            end
+        end
+ 
+        % spectral density
+        %------------------------------------------------------------------
+        subplot(2,2,3)
+        for k = 1:nt
+            plot(Hz,abs(xY.y{k}(:,i,i)),'color',co{i}), hold on
+            axis square, spm_axis tight
+        end
+    end
+    
+    title('Spectral density over modes')
+    xlabel('Frequency (Hz)')
+    ylabel('CSD')
+    axis square, spm_axis tight
+    legend(name)
+    
+    subplot(nm,nm,nm)
+    legend('real','imag')
+    return
+    
+end
+ 
+% post inversion parameters
+%--------------------------------------------------------------------------
+nu  = length(DCM.B);          % Nr experimental inputs
+ns  = size(DCM.A{1},2);       % Nr of sources
+ 
+ 
+% switch
+%--------------------------------------------------------------------------
+switch(lower(Action))    
+    
+case{lower('Coupling (A)')}
+    
+    % spm_dcm_ssr_results(DCM,'coupling (A)');
+    %----------------------------------------------------------------------
+    str = {'Forward (i)','Forward (ii)','Backward (i)','Backward (ii)'};
+    m   = length(DCM.Ep.A);
+    for  i = 1:m
+        
+        % images
+        %------------------------------------------------------------------
+        subplot(4,m,i)
+        imagesc(exp(DCM.Ep.A{i}))
+        title(str{i},'FontSize',10)
+        set(gca,'YTick',1:ns,'YTickLabel',DCM.Sname,'FontSize',8)
+        set(gca,'XTick',[])
+        xlabel('from','FontSize',8)
+        ylabel('to','FontSize',8)
+        axis square
+    
+        % table
+        %------------------------------------------------------------------
+        subplot(4,m,i + m)
+        text(0,1/2,num2str(full(exp(DCM.Ep.A{i})),' %.2f'),'FontSize',8)
+        axis off,axis square
+ 
+    
+        % PPM
+        %------------------------------------------------------------------
+        subplot(4,m,i + m + m)
+        image(64*DCM.Pp.A{i})
+        set(gca,'YTick',[1:ns],'YTickLabel',DCM.Sname,'FontSize',8)
+        set(gca,'XTick',[])
+        title('PPM')
+        axis square
+    
+        % table
+        %------------------------------------------------------------------
+        subplot(4,m,i + m + m + m)
+        text(0,1/2,num2str(DCM.Pp.A{i},' %.2f'),'FontSize',8)
+        axis off, axis square
+        
+    end
+    
+case{lower('Coupling (C)')}
+    
+    % spm_dcm_ssr_results(DCM,'coupling (C)');
+    %----------------------------------------------------------------------
+    
+    % images
+    %----------------------------------------------------------------------
+    subplot(2,4,1)
+    imagesc(exp(DCM.Ep.C))
+    title('Factors','FontSize',10)
+    set(gca,'XTick',1:nu,'XTickLabel','Input','FontSize',8)
+    set(gca,'YTick',1:ns,'YTickLabel',DCM.Sname, 'FontSize',8)
+    axis square
+    
+    % PPM
+    %----------------------------------------------------------------------
+    subplot(2,4,3)
+    image(64*DCM.Pp.C)
+    title('Factors','FontSize',10)
+    set(gca,'XTick',1:nu,'XTickLabel','Input','FontSize',8)
+    set(gca,'YTick',1:ns,'YTickLabel',DCM.Sname, 'FontSize',8)
+    axis square
+    title('PPM')
+    
+    % table
+    %----------------------------------------------------------------------
+    subplot(2,4,2)
+    text(0,1/2,num2str(full(exp(DCM.Ep.C)),' %.2f'),'FontSize',8)
+    axis off
+ 
+    % table
+    %----------------------------------------------------------------------
+    subplot(2,4,4)
+    text(0,1/2,num2str(DCM.Pp.C,' %.2f'),'FontSize',8)
+    axis off
+ 
+ 
+case{lower('Coupling (B)')}
+    
+    % spm_dcm_ssr_results(DCM,'coupling (B)');
+    %----------------------------------------------------------------------
+    for i = 1:nu
+        
+        % images
+        %------------------------------------------------------------------
+        subplot(4,nu,i)
+        imagesc(exp(DCM.Ep.B{i}))
+        title(DCM.xU.name{i},'FontSize',10)
+        set(gca,'YTick',1:ns,'YTickLabel',DCM.Sname,'FontSize',8)
+        set(gca,'XTick',[])
+        xlabel('from','FontSize',8)
+        ylabel('to','FontSize',8)
+        axis square
+ 
+        % tables
+        %------------------------------------------------------------------
+        subplot(4,nu,i + nu)
+        text(0,1/2,num2str(full(exp(DCM.Ep.B{i})),' %.2f'),'FontSize',8)
+        axis off
+        axis square
+        
+        % PPM
+        %------------------------------------------------------------------
+        subplot(4,nu,i + 2*nu)
+        image(64*DCM.Pp.B{i})
+        set(gca,'YTick',[1:ns],'YTickLabel',DCM.Sname,'FontSize',8)
+        set(gca,'XTick',[])
+        title('PPM')
+        axis square
+ 
+        % tables
+        %------------------------------------------------------------------
+        subplot(4,nu,i + 3*nu)
+        text(0,1/2,num2str(DCM.Pp.B{i},' %.2f'),'FontSize',8)
+        axis off
+        axis square
+        
+    end
+    
+case{lower('trial-specific effects')}
+    
+    % spm_dcm_ssr_results(DCM,'trial-specific effects');
+    %----------------------------------------------------------------------
+    for i = 1:ns
+        for j = 1:ns
+ 
+            % ensure connection is enabled
+            %--------------------------------------------------------------
+            q     = 0;
+            for k = 1:nu
+                q = q | DCM.B{k}(i,j);
+            end
+ 
+            % plot trial-specific effects
+            %--------------------------------------------------------------
+            if q
+                B     = zeros(nt,1);
+                for k = 1:nu
+                    B = B + DCM.xU.X(:,k)*DCM.Ep.B{k}(i,j);
+                end
+                
+                subplot(ns,ns,(i - 1)*ns + j)
+                bar(exp(B)*100,'c')
+                title([DCM.Sname{j}, ' to ' DCM.Sname{i}],'FontSize',10)
+                xlabel('trial',  'FontSize',8)
+                ylabel('strength (%)','FontSize',8)
+                set(gca,'XLim',[0 nt + 1])
+                axis square
+ 
+            end
+        end
+    end
+    
+case{lower('Input')}
+    
+    % spectrum of innovations or noise (Gu)
+    %----------------------------------------------------------------------
+    [Gu,Gs,Gn,Hz] = spm_csd_mtf_gu(DCM.Ep,DCM.M);
+ 
+    
+    % plot spectral density of innovations
+    % ---------------------------------------------------------------------
+    subplot(2,1,1)
+    plot(Hz,Gu)
+    xlabel('frquency (Hz)')
+    title('Spectrum of innovations','FontSize',16)
+    axis square, grid on
+    legend(DCM.Sname)
+
+    % plot spectral density of noise
+    % ---------------------------------------------------------------------
+    subplot(2,2,3)
+    plot(Hz,Gs)
+    xlabel('frquency (Hz)')
+    title('Channel-specifc noise')
+    axis square, grid on
+    legend(xY.name)
+
+    % plot spectral density of noise
+    % ---------------------------------------------------------------------
+    subplot(2,2,4)
+    plot(Hz,Gn)
+    xlabel('frquency (Hz)')
+    title('Non-specifc noise')
+    axis square, grid on
+    
+    
+    
+case{lower('Transfer functions')}
+    
+    % spm_dcm_ssr_results(DCM,'Cross-spectral density');
+    %----------------------------------------------------------------------
+    co   = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    Hz   = DCM.Hz;
+    name = DCM.Sname;
+    nm   = length(name);
+    
+    
+    % spectrum of innovations or noise (Gu)
+    %----------------------------------------------------------------------
+    Gu   = spm_csd_mtf_gu(DCM.Ep,DCM.M);
+    
+    
+    tstr = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('trial %i',k);
+    end
+    
+    for i = 1:nm
+        for j = 1:nm
+ 
+            % for each trial type
+            %--------------------------------------------------------------
+            subplot(nm,nm,(i - 1)*nm + j),cla
+            for k = 1:nt
+                
+                dtf = abs(DCM.dtf{k}(:,i,j));
+                stf = dtf.*Gu(Hz,j);
+                dtf = dtf/max(dtf);
+                stf = stf/max(stf);
+                
+                plot(Hz,dtf,   'color',co{k}), hold on
+                plot(Hz,stf,':','color',co{k}), hold off
+                title(sprintf('Spectral transfer: %s to %s',name{j},name{i}))
+                xlabel('frequency Hz')
+                axis square, spm_axis tight
+            end
+        end
+ 
+        % legend
+        %------------------------------------------------------------------      
+        if i == nm && j == nm
+            legend(tstr)
+        end
+    end
+   
+    
+case{lower('Cross-spectra (sources)')}
+    
+    % spm_dcm_ssr_results(DCM,'Cross-spectral density');
+    %----------------------------------------------------------------------
+    co   = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    Hz   = DCM.Hz;
+    name = DCM.Sname;
+    nm   = length(name);
+    
+    tstr = {};
+    mstr = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('predicted: trial %i',k);
+    end
+    for k = 1:nm
+        mstr{end + 1} = sprintf('predicted: %s',name{k});
+    end
+    
+    for i = 1:nm
+        for j = i:nm
+ 
+            % for each trial type
+            %--------------------------------------------------------------
+            subplot(nm,nm,(i - 1)*nm + j),cla
+            for k = 1:nt
+                plot(Hz,abs(DCM.Hs{k}(:,i,j)),'color',co{k}), hold on
+                title(sprintf('CSD: %s to %s',name{j},name{i}))
+                xlabel('frequency Hz')
+                axis square, spm_axis tight
+            end
+        end
+ 
+        % legend
+        %------------------------------------------------------------------      
+        if i == nm && j == nm
+            legend(tstr)
+        end
+        
+        % spectral density
+        %------------------------------------------------------------------
+        subplot(2,2,3)
+        for k = 1:nt
+            plot(Hz,abs(DCM.Hs{k}(:,i,i)),'color',co{i}), hold on
+            axis square, spm_axis tight
+        end
+    end
+   
+    title({'Spectral density over sources';'(in source-space)'},'FontSize',16)
+    xlabel('frequency (Hz)')
+    ylabel('abs(CSD)')
+    axis square, spm_axis tight
+    legend(mstr)
+    
+    
+case{lower('Cross-spectra (channels)')}
+    
+    % spm_dcm_ssr_results(DCM,'Cross-spectral density');
+    %----------------------------------------------------------------------
+    co   = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    Hz   = DCM.Hz;
+    nm   = min(nm,4);
+    
+    tstr = {};
+    mstr = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('predicted: trial %i',k);
+        tstr{end + 1} = sprintf('observed: trial %i',k);
+    end
+    for k = 1:nm
+        mstr{end + 1} = sprintf('predicted: mode %i',k);
+        mstr{end + 1} = sprintf('observed: mode %i',k);
+    end
+    
+    for i = 1:nm
+        for j = i:nm
+ 
+            % for each trial type
+            %--------------------------------------------------------------
+            subplot(nm,nm,(i - 1)*nm + j),cla
+            for k = 1:nt
+                plot(Hz,abs(DCM.Hc{k}(:,i,j)),'color',co{k}), hold on
+                plot(Hz,abs(DCM.Hc{k}(:,i,j) + DCM.Rc{k}(:,i,j)),':','color',co{k})
+                title(sprintf('mode %i to %i',j,i))
+                xlabel('frequency Hz')
+                axis square, spm_axis tight
+            end
+        end
+ 
+        % legend
+        %------------------------------------------------------------------      
+        if i == nm && j == nm
+            legend(tstr)
+        end
+        
+        % spectral density
+        %------------------------------------------------------------------
+        subplot(2,2,3)
+        for k = 1:nt
+            plot(Hz,abs(DCM.Hc{k}(:,i,i)),'color',co{i}), hold on
+            plot(Hz,abs(DCM.Hc{k}(:,i,i) + DCM.Rc{k}(:,i,i)),':','color',co{i})
+            axis square, spm_axis tight
+        end
+    end
+   
+    title({'Spectral density over modes';'(in channel-space)'},'FontSize',16)
+    xlabel('frequency (Hz)')
+    ylabel('abs(CSD)')
+    legend(mstr)
+    
+    
+case{lower('Coherence (sources)')}
+    
+    % get coherence
+    %----------------------------------------------------------------------
+    [coh fsd] = spm_csd2coh(DCM.Hs,DCM.Hz);
+    
+    % spm_dcm_ssr_results(DCM, 'Coherence (sources)’);
+    %----------------------------------------------------------------------
+    co    = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    Hz    = DCM.Hz;
+    name  = DCM.Sname;
+    nm    = length(name);
+    tstr  = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('trial %i',k);
+    end
+    
+    for i = 1:nm
+        for j = 1:nm
+ 
+            if j > i
+                
+                % for each trial type - coherence
+                %----------------------------------------------------------
+                subplot(nm,nm,(i - 1)*nm + j),cla
+                for k = 1:nt
+                    plot(Hz,coh{k}(:,i,j),'color',co{k}), hold on
+                    title(sprintf('Coh: %s to %s',name{j},name{i}))
+                    xlabel('frequency Hz')
+                    axis square, spm_axis tight
+                end
+            
+            end
+            
+            if j < i
+                % for each trial type - delay
+                %--------------------------------------------------------------
+                subplot(nm,nm,(i - 1)*nm + j),cla
+                for k = 1:nt
+                    plot(Hz,1000*fsd{k}(:,i,j),'color',co{k}), hold on
+                    title(sprintf('Delay (ms) %s to %s',name{j},name{i}))
+                    xlabel('Frequency Hz')
+                    axis square, spm_axis tight
+                end  
+            end
+           
+        end
+    end
+    legend(tstr)
+    
+    
+case{lower('Coherence (channels)')}
+     
+    % get coherence
+    %----------------------------------------------------------------------
+    [coh fsd] = spm_csd2coh(DCM.Hc,DCM.Hz);
+    [COH FSD] = spm_csd2coh(DCM.xY.y,DCM.Hz);
+ 
+    
+    % spm_dcm_ssr_results(DCM,'Coherence (channels)’);
+    %----------------------------------------------------------------------
+    co    = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    nm    = min(nm,4);
+    tstr  = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('predicted: trial %i',k);
+        tstr{end + 1} = sprintf('observed: trial %i',k);
+    end
+ 
+    
+    for i = 1:nm
+        for j = 1:nm
+ 
+            if j > i
+                
+                % for each trial type - coherence
+                %----------------------------------------------------------
+                subplot(nm,nm,(i - 1)*nm + j),cla
+                for k = 1:nt
+                    plot(Hz,coh{k}(:,i,j),'color',co{k}), hold on
+                    plot(Hz,COH{k}(:,i,j),':','color',co{k}), hold on
+                    title(sprintf('Coh: %i to %i',j,i))
+                    xlabel('frequency Hz')
+                    axis square, spm_axis tight
+                end
+            
+            end
+            
+            if j < i
+                % for each trial type - delay
+                %----------------------------------------------------------
+                subplot(nm,nm,(i - 1)*nm + j),cla
+                for k = 1:nt
+                    plot(Hz,1000*fsd{k}(:,i,j),'color',co{k}), hold on
+                    plot(Hz,1000*FSD{k}(:,i,j),':','color',co{k}), hold on
+                    title(sprintf('Delay (ms) %i to %i',j,i))
+                    xlabel('frequency Hz')
+                    axis square, spm_axis tight
+                end  
+            end
+           
+        end
+    end
+    legend(tstr)
+    
+    
+case{lower('Covariance (sources)')}
+    
+    % get covariances
+    %----------------------------------------------------------------------
+    [ccf pst] = spm_csd2ccf(DCM.Hs,DCM.Hz);
+    
+    
+    % spm_dcm_ssr_results(DCM,'Cross-covariance (sources)');
+    %----------------------------------------------------------------------
+    co   = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    pst  = 1000*pst;
+    name = DCM.Sname;
+    nm   = length(name);
+    q    = max(spm_vec(ccf));
+    p    = min(spm_vec(ccf));
+    tstr = {};
+    mstr = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('trial %i',k);
+    end
+    for k = 1:nm
+        mstr{end + 1} = sprintf('source %i',k);
+    end
+    
+    for i = 1:nm
+        for j = i:nm
+ 
+            % for each trial type
+            %--------------------------------------------------------------
+            subplot(nm,nm,(i - 1)*nm + j),cla
+            plot([0 0],[p q],':'), hold on
+            for k = 1:nt
+                plot(pst,ccf{k}(:,i,j),'color',co{k}), hold on
+                title(sprintf('%s to %s',name{j},name{i}))
+                xlabel('lag (ms)')
+                axis tight square, set(gca,'XLim',[-128 128],'YLim',[p q])
+            end
+        end
+ 
+        % legend
+        %------------------------------------------------------------------      
+        if i == nm && j == nm
+            legend(tstr)
+        end
+        
+        % spectral density
+        %------------------------------------------------------------------
+        subplot(2,2,3)
+        for k = 1:nt
+            plot(pst,ccf{k}(:,i,i),'color',co{i}), hold on
+            axis tight square, set(gca,'XLim',[-128 128],'YLim',[p q])
+        end
+    end
+   
+    title({'Auto-covariance';'(in source-space)'},'FontSize',16)
+    xlabel('lag (ms)')
+    ylabel('auto-covariance')
+    axis square
+    legend(mstr)
+    
+    
+case{lower('Covariance (channels)')}
+    
+    % get covariances
+    %----------------------------------------------------------------------
+    [ccf pst] = spm_csd2ccf(DCM.Hc,DCM.Hz);
+    
+    
+    % spm_dcm_ssr_results(DCM,'Cross-covariance (sources)');
+    %----------------------------------------------------------------------
+    co   = {'b', 'r', 'g', 'm', 'y', 'k', 'c'};
+    pst  = 1000*pst;
+    nm   = min(nm,4);
+    q    = max(spm_vec(ccf));
+    p    = min(spm_vec(ccf));
+    tstr = {};
+    mstr = {};
+    for k = 1:nt
+        tstr{end + 1} = sprintf('trial %i',k);
+    end
+    for k = 1:nm
+        mstr{end + 1} = sprintf('channel %i',k);
+    end
+    
+    for i = 1:nm
+        for j = i:nm
+ 
+            % for each trial type
+            %--------------------------------------------------------------
+            subplot(nm,nm,(i - 1)*nm + j),cla
+            plot([0 0],[p q],':'), hold on
+            for k = 1:nt
+                plot(pst,ccf{k}(:,i,j),'color',co{k}), hold on
+                title(sprintf('mode %i to %i',j,i))
+                xlabel('lag (ms)')
+                axis tight square, set(gca,'XLim',[-128 128],'YLim',[p q])
+            end
+        end
+ 
+        % legend
+        %------------------------------------------------------------------      
+        if i == nm && j == nm
+            legend(tstr)
+        end
+        
+        % spectral density
+        %------------------------------------------------------------------
+        subplot(2,2,3)
+        for k = 1:nt
+            plot(pst,ccf{k}(:,i,i),'color',co{i}), hold on
+            axis tight square, set(gca,'XLim',[-128 128],'YLim',[p q])
+        end
+    end
+   
+    title({'Auto-covariance';'(in channel-space)'},'FontSize',16)
+    xlabel('Lag (ms)')
+    ylabel('auto-covariance')
+    axis square
+    legend(mstr)
+    
+    
+case{lower('Dipoles')}
+    
+    % return if LFP
+    % ---------------------------------------------------------------------
+    if strcmpi(DCM.xY.modality,'lfp')
+        warndlg('There are no ECDs for these LFP data')
+        return
+    end
+    
+    % plot dipoles
+    % ---------------------------------------------------------------------
+    try
+        P            = DCM.Ep;   
+        np           = size(P.L,2)/size(P.Lpos,2);
+        sdip.n_seeds = 1;
+        sdip.n_dip   = np*ns;
+        sdip.Mtb     = 1;
+        sdip.j{1}    = full(P.L);
+        sdip.j{1}    = sdip.j{1}(:);
+        sdip.loc{1}  = kron(ones(1,np),full(P.Lpos));
+        spm_eeg_inv_ecd_DrawDip('Init', sdip)
+    end
+ 
+end
diff --git a/toolbox/dcm_meeg/spm_dcm_csd_source_optimise.m b/toolbox/dcm_meeg/spm_dcm_csd_source_optimise.m
new file mode 100644
index 0000000..2a3ef13
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd_source_optimise.m
@@ -0,0 +1,124 @@
+function [PF] = spm_dcm_csd_source_optimise
+% Stochastic optimisation of single source neural mass model
+% FORMAT [PF] = spm_dcm_csd_source_optimise
+%
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_csd_source_optimise.m 4261 2011-03-24 16:39:42Z karl $
+ 
+ 
+% Initaislie fixed paramters
+%==========================================================================
+% PF.G  = [4 4 4 4 4 2 4 4 2 1]*200;
+% PF.T  = [2 2 28 24];
+% PF.D  = [1 16];
+
+PF.G  = [1 1 1 1/2]*64;         % intrinsic rates (g1 g2 g3 g4)
+PF.H  = [4 64];                 % receptor densities (excitatory, inhibitory)
+PF.T  = [4 8];                  % synaptic constants (excitatory, inhibitory)
+ 
+
+HZ    = [60 16]; model = 'CMC'; s = [3 7]; Hz = [1:128]';
+HZ    = [20 16]; model = 'SEP'; s = [7 9]; Hz = [1:64]';
+
+N     = 512;
+n     = length(spm_vec(PF));
+sC    = speye(n,n)/2;
+for k = 1:4
+    for i = 1:N
+        
+        % parameters
+        %----------------------------------------------------------------------
+        pF     = spm_vec(PF);
+        pF     = pF.*exp(sC*randn(n,1)/k);
+        P(:,i) = pF;
+        pF     = spm_unvec(pF,PF);
+                
+        % CSD
+        %------------------------------------------------------------------
+        G      = spm_dcm_csd_source_plot(model,s,pF,2*Hz(end));
+        R(:,i) = spm_vec(G);
+        
+        % score - mean and variance over Hz
+        %------------------------------------------------------------------
+        for j = 1:length(s)
+            p     = G(:,j,j);
+            S(j)  = norm(p,1);
+            p     = p/sum(p);
+            E1(j) = sum(Hz.*p);
+            E2(j) = sum((Hz - E1(j)).^2.*p);
+            E3(j) = sum((Hz - E1(j)).^3.*p)/(E2(j)^(3/2));
+            E4(j) = sum((Hz - E1(j)).^4.*p)/(E2(j)^2) - 3;
+        end
+        
+        % entropy over sources
+        %------------------------------------------------------------------
+        S      = S/sum(S);
+        S      = S*log(S');
+        
+        % cost function
+        %------------------------------------------------------------------
+        c(1,i) = sum((E1 - HZ).^2);
+        c(2,i) = sum((sqrt(E2) - sqrt([64 64])).^2);
+        c(3,i) = sum((E3 - [0 0]).^2);
+        c(4,i) = sum((E4 - [0 0]).^2);
+        c(5,i) = S*32;
+        
+    end
+    
+    % total cost and minimise
+    %----------------------------------------------------------------------
+    F     = sum(c)';
+    i     = isfinite(F);
+    F     = F(i); P = P(:,i); R = R(:,i);
+    i     = find(F < (min(F) + 8*std(F)));
+    F     = F(i); P = P(:,i); R = R(:,i);
+
+    
+    % Laplace approximation: p(P)
+    %======================================================================
+
+    % temperature
+    %----------------------------------------------------------------------
+    T     = std(F)/128;
+
+    % mean
+    %----------------------------------------------------------------------
+    q     = exp(-(F - mean(F))/T);
+    q     = q/sum(q);
+    Lq    = P*q;
+
+    % dispersion
+    %----------------------------------------------------------------------
+    for i = 1:n
+        for j = 1:n
+            C(i,j) = ((P(i,:) - Lq(i)).*(P(j,:) - Lq(j)))*q;
+        end
+    end   
+    PF    = spm_unvec(Lq,PF);
+    
+    spm_dcm_csd_source_plot(model,s,PF,2*Hz(end));
+    disp(PF)
+    
+end
+
+return
+
+% SEP models
+%--------------------------------------------------------------------------
+% E  = [32 16 4];              % extrinsic rates (forward, backward, lateral)
+% G  = [1 1 1/4 1/4]*128;      % intrinsic rates (g1 g2 g3 g4)
+% D  = [2 32];                 % propagation delays (intrinsic, extrinsic)
+% H  = [4 32];                 % receptor densities (excitatory, inhibitory)
+% T  = [4 8];                  % synaptic constants (excitatory, inhibitory)
+% R  = [1 0];                  % parameters of static nonlinearity
+
+
+% SVD of mapping from parameters to CSD
+%--------------------------------------------------------------------------
+X       = log(abs(R'));
+X       = X(:,sum(X) > -8);
+[U S V] = spm_svd(X'*P',0);
+
diff --git a/toolbox/dcm_meeg/spm_dcm_csd_source_plot.m b/toolbox/dcm_meeg/spm_dcm_csd_source_plot.m
new file mode 100644
index 0000000..986c733
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_csd_source_plot.m
@@ -0,0 +1,128 @@
+function [G] = spm_dcm_csd_source_plot(model,s,pF,N)
+% Spectral response (G) of a single source neural mass model
+% FORMAT [G] = spm_dcm_csd_source_plot(model,s)
+%
+% model - 'ERP', 'SEP', 'CMC', 'LFP', 'NMM' or 'MFM'
+% s     - indices of hidden neuronal states to plot
+% pF    - fixed parameters
+% N     - twice the maximum frequency
+%
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_dcm_csd_source_plot.m 4261 2011-03-24 16:39:42Z karl $
+
+
+% Create model
+%==========================================================================
+try, s; catch, s = [1 2 3 7]; end
+try, N; catch, N = 256;       end
+
+% prior moments on parameters
+%--------------------------------------------------------------------------
+pE = spm_dcm_neural_priors({0,0,0},{},1,model);
+
+
+% intial states and equations of motion
+%--------------------------------------------------------------------------
+[x,f] = spm_dcm_x_neural(pE,model);
+
+% create DCM
+%--------------------------------------------------------------------------
+M.f     = f;
+M.x     = x;
+M.n     = length(spm_vec(x));
+[ns nx] = size(x);
+
+
+% compute spectral density
+%==========================================================================
+
+% frequencies of interest
+%--------------------------------------------------------------------------
+dt   = 1/N;
+If   = 1:N/2;
+f    = [1:N/2]';
+w    = f(If);
+
+% exogenous (neuronal) inputs
+%--------------------------------------------------------------------------
+M.u  = 0;
+
+% fixed parameters
+%--------------------------------------------------------------------------
+try, M.pF  = pF; end
+
+% get prior means (delays)
+%--------------------------------------------------------------------------
+try
+    di = M.pF.D(1);                    % intrinsic delays
+    de = M.pF.D(2);                    % extrinsic delays
+catch
+    de = 16;
+    di = 1;
+end
+
+
+% spectrum of innovations (Gu)
+%--------------------------------------------------------------------------
+Gu   = f.^(-1)*8;
+
+% get delay matrix
+%--------------------------------------------------------------------------
+De = exp(pE.D);
+Di = diag(diag(De));
+De = De - Di;
+De = De*de/1000;
+Di = Di*di/1000;
+De = kron(ones(nx,nx),De);
+Di = kron(ones(nx,nx) - speye(nx,nx),Di);
+D  = Di + De;
+
+% get delay operator
+%--------------------------------------------------------------------------
+D  = spm_dcm_delay(M,pE,D);
+
+
+% augment and bi-linearise (with delays)
+%--------------------------------------------------------------------------
+[M0,M1,L] = spm_bireduce(M,pE,D);
+
+% compute modulation transfer function using FFT of the kernels
+%--------------------------------------------------------------------------
+[K0,K1]   = spm_kernels(M0,M1,L,N,dt);
+[N,nc,nu] = size(K1);
+
+
+% [cross]-spectral density
+%--------------------------------------------------------------------------
+ns    = length(s);
+G     = zeros(N/2,ns,ns);
+for i = 1:ns
+    for j = i:ns
+        for k = 1:nu
+            Si       = fft(K1(:,s(i),k));
+            Sj       = fft(K1(:,s(j),k));
+            Gij      = Si.*conj(Sj);
+            Gij      = Gij([1:N/2] + 1).*Gu;
+            G(:,i,j) = G(:,i,j) + Gij;
+        end
+        
+        % Graphics
+        %==================================================================
+        subplot(ns,ns,(i - 1)*ns + j)
+        plot(f,abs(G(:,i,j)))
+        xlabel('frequency')
+        axis square
+        
+    end
+end
+drawnow
+
+
+
+
+
+
+
diff --git a/toolbox/dcm_meeg/spm_dcm_dem.m b/toolbox/dcm_meeg/spm_dcm_dem.m
index a7ea258..82a0f94 100644
--- a/toolbox/dcm_meeg/spm_dcm_dem.m
+++ b/toolbox/dcm_meeg/spm_dcm_dem.m
@@ -22,7 +22,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_dem.m 3517 2009-10-29 15:11:56Z guillaume $
+% $Id: spm_dcm_dem.m 4281 2011-03-31 19:49:57Z karl $
 
 % check options
 %==========================================================================
@@ -37,7 +37,6 @@
 try, Nm    = DCM.options.Nmodes; catch, Nm        = 8;         end
 try, onset = DCM.options.onset;  catch, onset     = 60;        end
 try, model = DCM.options.model;  catch, model     = 'DEM';     end
-try, lock  = DCM.options.lock;   catch, lock      = 0;         end
 
 
 % Data and spatial model (use h only for de-trending data)
@@ -197,6 +196,7 @@
 DCM.Cp  = Cp;                   % conditional covariances G(g)
 DCM.Eg  = Qg;                   % conditional expectation
 DCM.Cg  = Cg;                   % conditional covariances
+DCM.Ce  = Ce;                   % conditional error covariance
 DCM.Pp  = Pp;                   % conditional probability
 DCM.H   = y;                    % conditional responses (y), projected space
 DCM.K   = x;                    % conditional responses (x)
@@ -209,7 +209,6 @@
 DCM.options.Nmodes = Nm;
 DCM.options.onset  = onset;
 DCM.options.model  = model;
-DCM.options.lock   = lock;
 
 % store estimates in D
 %--------------------------------------------------------------------------
@@ -293,7 +292,7 @@
 
 % and save
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7.1') >= 0
+if spm_check_version('matlab','7') >= 0
     save(DCM.name, '-V6', 'DCM');
 else
     save(DCM.name, 'DCM');
diff --git a/toolbox/dcm_meeg/spm_dcm_erp.m b/toolbox/dcm_meeg/spm_dcm_erp.m
index 5044ec5..6f3bac6 100644
--- a/toolbox/dcm_meeg/spm_dcm_erp.m
+++ b/toolbox/dcm_meeg/spm_dcm_erp.m
@@ -20,14 +20,14 @@
 %   options.h            - number of DCT drift terms (usually 1 or 2)
 %   options.Nmodes       - number of spatial models to invert
 %   options.analysis     - 'ERP', 'SSR' or 'IND'
-%   options.model        - 'ERP', 'SEP', NMM or 'MFM'
+%   options.model        - 'ERP', 'SEP', 'CMC', 'NMM' or 'MFM'
 %   options.spatial      - 'ERP', 'LFP' or 'IMG'
 %   options.onset        - stimulus onset (ms)
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_erp.m 3653 2009-12-23 20:06:48Z karl $
+% $Id: spm_dcm_erp.m 4281 2011-03-31 19:49:57Z karl $
 
 % check options
 %==========================================================================
@@ -37,15 +37,17 @@
 
 % Filename and options
 %--------------------------------------------------------------------------
-try, DCM.name;                   catch, DCM.name  = name;        end
-try, DCM.xU;                     catch, DCM.xU.X  = sparse(1,0); end
-try, h     = DCM.options.h;      catch, h         = 1;           end
-try, Nm    = DCM.options.Nmodes; catch, Nm        = 8;           end
-try, onset = DCM.options.onset;  catch, onset     = 60;          end
-try, model = DCM.options.model;  catch, model     = 'NMM';       end
-try, lock  = DCM.options.lock;   catch, lock      = 0;           end
-
+try, DCM.name;                      catch, DCM.name  = name;        end
+try, DCM.xU;                        catch, DCM.xU.X  = sparse(1,0); end
+try, h     = DCM.options.h;         catch, h         = 1;           end
+try, Nm    = DCM.options.Nmodes;    catch, Nm        = 8;           end
+try, onset = DCM.options.onset;     catch, onset     = 60;          end
+try, model = DCM.options.model;     catch, model     = 'NMM';       end
+try, lock  = DCM.options.lock;      catch, lock      = 0;           end
+try, symm  = DCM.options.symmetry;  catch, symm      = 0;           end
 
+if ~strcmp(DCM.options.spatial,'ECD'), symm = 0; end
+    
 
 
 % Data and spatial model (use h only for de-trending data)
@@ -132,33 +134,29 @@
     end
 end
 
-% lock experimental effects by introducing prior correlations
-%--------------------------------------------------------------------------
-if lock
-    pV    = spm_unvec(diag(pC),pE);
-    for i = 1:Nx
-        pB      = pV;
-        pB.B{i} = pB.B{i} - pB.B{i};
-        pB      = spm_vec(pV)  - spm_vec(pB);
-        pB      = sqrt(pB*pB') - diag(pB);
-        pC      = pC + pB;
-    end
-end
-
 % priors on spatial model
 %--------------------------------------------------------------------------
 M.dipfit.model = model;
 [gE,gC] = spm_L_priors(M.dipfit);
 
+% Set prior correlations (locking trial effects and dipole orientations
+%--------------------------------------------------------------------------
+if lock, pC = spm_dcm_lock(pC);    end
+if symm, gC = spm_dcm_symm(gC,gE); end
+
 
 % intial states and equations of motion
 %--------------------------------------------------------------------------
 [x,f] = spm_dcm_x_neural(pE,model);
 
+% hyperpriors (assuming about 99% signal to noise)
+%--------------------------------------------------------------------------
+hE    = 4 - log(var(spm_vec(xY.y)));
+hC    = exp(-8);
+
 
 % likelihood model
 %--------------------------------------------------------------------------
-M.IS  = 'spm_gen_erp_dfdp';                            % for fast inversion
 M.IS  = 'spm_gen_erp';
 M.FS  = 'spm_fy_erp';
 M.G   = 'spm_lx_erp';
@@ -168,6 +166,8 @@
 M.pC  = pC;
 M.gE  = gE;
 M.gC  = gC;
+M.hE  = hE;
+M.hC  = hC;
 M.m   = Nu;
 M.n   = length(spm_vec(M.x));
 M.l   = Nc;
@@ -192,7 +192,6 @@
 end
 Nm    = size(U,2);
 
-
 % EM: inversion
 %==========================================================================
 [Qp,Qg,Cp,Cg,Ce,F,LE] = spm_nlsi_N(M,xU,xY);
@@ -246,6 +245,7 @@
 DCM.Cp = Cp;                   % conditional covariances G(g)
 DCM.Eg = Qg;                   % conditional expectation
 DCM.Cg = Cg;                   % conditional covariances
+DCM.Ce = Ce;                   % conditional error
 DCM.Pp = Pp;                   % conditional probability
 DCM.H  = y;                    % conditional responses (y), projected space
 DCM.K  = x;                    % conditional responses (x)
@@ -260,6 +260,8 @@
 DCM.options.onset  = onset;
 DCM.options.model  = model;
 DCM.options.lock   = lock;
+DCM.options.symm   = symm;
+
 
 % store estimates in D
 %--------------------------------------------------------------------------
@@ -318,23 +320,24 @@
 
     % fill in fields of inverse structure
     %----------------------------------------------------------------------
-    inverse.trials = DCM.options.trials;   % trial or condition
-    inverse.type   = 'DCM';                % inverse model
-    inverse.J      = J;                    % Conditional expectation
-    inverse.L      = L;                    % Lead field (reduced)
-    inverse.R      = speye(Nc,Nc);         % Re-referencing matrix
-    inverse.T      = T0;                   % temporal subspace
-    inverse.U      = U;                    % spatial subspace
-    inverse.Is     = Is;                   % Indices of active dipoles
-    inverse.It     = DCM.xY.It;            % Indices of time bins
-    inverse.Ic     = DCM.xY.Ic;            % Indices of good channels
-    inverse.Y      = Y;                    % reduced data
-    inverse.Nd     = Nd;                   % number of dipoles
-    inverse.Nt     = Nt;                   % number of trials
-    inverse.pst    = xY.pst;               % peri-stimulus time
-    inverse.F      = DCM.F;                % log-evidence
-    inverse.R2     = R2;                   % variance accounted for (%)
-    inverse.dipfit = M.dipfit;             % forward model for DCM
+    inverse.trials   = DCM.options.trials;   % trial or condition
+    inverse.modality = {DCM.xY.modality};    % modality
+    inverse.type     = 'DCM';                % inverse model
+    inverse.J        = J;                    % Conditional expectation
+    inverse.L        = L;                    % Lead field (reduced)
+    inverse.R        = speye(Nc,Nc);         % Re-referencing matrix
+    inverse.T        = T0;                   % temporal subspace
+    inverse.U        = U;                    % spatial subspace
+    inverse.Is       = Is;                   % Indices of active dipoles
+    inverse.It       = DCM.xY.It;            % Indices of time bins
+    inverse.Ic       = DCM.xY.Ic;            % Indices of good channels
+    inverse.Y        = Y;                    % reduced data
+    inverse.Nd       = Nd;                   % number of dipoles
+    inverse.Nt       = Nt;                   % number of trials
+    inverse.pst      = xY.pst;               % peri-stimulus time
+    inverse.F        = DCM.F;                % log-evidence
+    inverse.R2       = R2;                   % variance accounted for (%)
+    inverse.dipfit   = M.dipfit;             % forward model for DCM
 
     % append DCM results and save in structure
     %----------------------------------------------------------------------
@@ -353,6 +356,10 @@
 
 % and save
 %--------------------------------------------------------------------------
-save(DCM.name,  'DCM');
-assignin('base','DCM',DCM)
+try
+    save(DCM.name,'DCM');
+catch
+    save(name,    'DCM');
+end
+assignin('base',  'DCM',DCM)
 return
diff --git a/toolbox/dcm_meeg/spm_dcm_erp_plot.m b/toolbox/dcm_meeg/spm_dcm_erp_plot.m
new file mode 100644
index 0000000..564ac19
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_erp_plot.m
@@ -0,0 +1,79 @@
+function x = spm_dcm_erp_plot(DCM)
+% plots predicted source activity
+% FORMAT x = spm_dcm_erp_plot(DCM)
+%
+% DCM - DCM structure:
+% store estimates in DCM
+%--------------------------------------------------------------------------
+% DCM.M  - model specification
+% DCM.xY - data structure
+% DCM.xU - input structure
+% DCM.Ep - conditional expectation f(x,u,p)
+% DCM.Cp - conditional covariances G(g)
+% DCM.Eg - conditional expectation
+% DCM.Cg - conditional covariances
+% DCM.Pp - conditional probability
+% DCM.H  - conditional responses (y), projected space
+% DCM.K  - conditional responses (x)
+% DCM.R  - conditional residuals (y)
+% DCM.F  - Laplace log evidence
+% DCM.L  - Laplace log evidence components
+% DCM.ID - data ID
+% 
+% 
+% DCM.options.h
+% DCM.options.Nmodes
+% DCM.options.onset
+% DCM.options.model
+% DCM.options.lock
+% DCM.options.symm
+%
+% x{i}   - source activity contributing sources {trial i}
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_dcm_erp_plot.m 4186 2011-02-01 20:11:32Z karl $
+
+% check options
+%==========================================================================
+
+% unpack
+%--------------------------------------------------------------------------
+xY     = DCM.xY;
+xU     = DCM.xU;
+M      = DCM.M;
+
+% dimensions
+%--------------------------------------------------------------------------
+Nr     = size(DCM.C,1);                 % number of sources
+Nu     = size(DCM.C,2);                 % number of exogenous inputs
+Ns     = size(xY.xy{1},1);              % number of time bins
+
+% confounds and parameters
+%--------------------------------------------------------------------------
+T0     = speye(Ns) - xY.X0*((xY.X0'*xY.X0)\xY.X0');
+Qp     = DCM.Ep;
+
+% neuronal responses
+%--------------------------------------------------------------------------
+x     = feval(M.IS,Qp,M,xU);
+
+% trial-specific responses (in source space)
+%--------------------------------------------------------------------------
+x0    = ones(Ns,1)*spm_vec(M.x)';       % expansion point for states
+j     = kron(DCM.Eg.J,ones(1,Nr));      % Indices of contributing states
+j     = logical(j);
+for i = 1:numel(x)
+    x{i} = x{i} - x0;                   % centre on expansion point
+    x{i} = x{i}(:,j);                   % Depolarization in sources
+end
+
+
+% store estimates in DCM
+%--------------------------------------------------------------------------
+DCM.K  = x;                             % conditional sensitivity
+
+% display
+%--------------------------------------------------------------------------
+spm_dcm_erp_results(DCM,'ERPs (sources)');
diff --git a/toolbox/dcm_meeg/spm_dcm_erp_results.m b/toolbox/dcm_meeg/spm_dcm_erp_results.m
index 9fc78a9..19e7d86 100644
--- a/toolbox/dcm_meeg/spm_dcm_erp_results.m
+++ b/toolbox/dcm_meeg/spm_dcm_erp_results.m
@@ -8,6 +8,7 @@
 % FORMAT spm_dcm_erp_results(DCM,'trial-specific effects');
 % FORMAT spm_dcm_erp_results(DCM,'Input');
 % FORMAT spm_dcm_erp_results(DCM,'Response');
+% FORMAT spm_dcm_erp_results(DCM,'Response (image)');
 % FORMAT spm_dcm_erp_results(DCM,'Data');
 %                
 %___________________________________________________________________________
@@ -28,22 +29,29 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_dcm_erp_results.m 3653 2009-12-23 20:06:48Z karl $
+% $Id: spm_dcm_erp_results.m 4281 2011-03-31 19:49:57Z karl $
 
 
 % get Action if necessary
 %--------------------------------------------------------------------------
 if nargin < 2
     
-    str{1} = 'ERPs (sources)';
-    str{2} = 'Coupling (A)';
-    str{3} = 'Coupling (B)';
-    str{4} = 'Coupling (C)';
-    str{5} = 'trial-specific effects';
-    str{6} = 'Input';
-    str{7} = 'Response';
-    str{8} = 'Data';
-    Action = str{spm_input('Action',0,'m',str)};
+    str{1} = 'ERPs (mode)';
+    str{2} = 'ERPs (sources)';
+    str{3} = 'Coupling (A)';
+    str{4} = 'Coupling (B)';
+    str{5} = 'Coupling (C)';
+    str{6} = 'trial-specific effects';
+    str{7} = 'Input';
+    str{8} = 'Response';
+    str{9} = 'Response (image)';
+    str{10} = 'Data';
+    
+    s      = listdlg('PromptString','Select an option:',...
+                     'SelectionMode','single',...
+                     'ListString',str);
+                 
+    Action = str{s};
     
 end
     
@@ -434,7 +442,8 @@
 
             subplot(nt,2,2*i - 0)
             plot(t,DCM.H{i}*U)
-            xlabel('time (ms)')
+            xlabel('channels')
+            ylabel('time (ms)')
             title('Predicted')
             A(end + 1,:) = axis;
         end
@@ -479,7 +488,8 @@
 
             subplot(nt,2,2*i - 0)
             imagesc([1:ne],t,DCM.H{i}*U)
-            xlabel('time (ms)')
+            xlabel('channels')
+            ylabel('time (ms)')
             title('Predicted')
             axis(A); axis square, grid on
         end
diff --git a/toolbox/dcm_meeg/spm_dcm_erp_sensitivity.m b/toolbox/dcm_meeg/spm_dcm_erp_sensitivity.m
new file mode 100644
index 0000000..06e6788
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_erp_sensitivity.m
@@ -0,0 +1,97 @@
+function x = spm_dcm_erp_sensitivity(DCM,C)
+% plots change in source activity with a contrast of parameters
+% FORMAT x = spm_dcm_erp_sensitivity(DCM,C)
+%
+% DCM - DCM structure:
+% store estimates in DCM
+%--------------------------------------------------------------------------
+% DCM.M  - model specification
+% DCM.xY - data structure
+% DCM.xU - input structure
+% DCM.Ep - conditional expectation f(x,u,p)
+% DCM.Cp - conditional covariances G(g)
+% DCM.Eg - conditional expectation
+% DCM.Cg - conditional covariances
+% DCM.Pp - conditional probability
+% DCM.H  - conditional responses (y), projected space
+% DCM.K  - conditional responses (x)
+% DCM.R  - conditional residuals (y)
+% DCM.F  - Laplace log evidence
+% DCM.L  - Laplace log evidence components
+% DCM.ID - data ID
+% 
+% 
+% DCM.options.h
+% DCM.options.Nmodes
+% DCM.options.onset
+% DCM.options.model
+% DCM.options.lock
+% DCM.options.symm
+%
+% C      - contrast (in the form of DCM.pE)
+%        - or string identifying a parameter: e.g. 'A{2}(3,1)' 
+%
+% x{i}   - dx/dC for contributing sources {trial i}
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+
+% Karl Friston
+% $Id: spm_dcm_erp_sensitivity.m 4186 2011-02-01 20:11:32Z karl $
+
+% check options
+%==========================================================================
+
+% unpack
+%--------------------------------------------------------------------------
+xY     = DCM.xY;
+xU     = DCM.xU;
+M      = DCM.M;
+
+% contrast
+%--------------------------------------------------------------------------
+if ischar(C)
+    Cp = DCM.Ep;
+    Cp = spm_unvec(spm_vec(Cp)*0,Cp);
+    
+    eval(['Cp.' C ' = 1;']);
+    C  = Cp;
+end
+
+% dimensions
+%--------------------------------------------------------------------------
+Nr     = size(DCM.C,1);                 % number of sources
+Nu     = size(DCM.C,2);                 % number of exogenous inputs
+Ns     = size(xY.xy{1},1);              % number of time bins
+
+% confounds and parameters
+%--------------------------------------------------------------------------
+T0     = speye(Ns) - xY.X0*((xY.X0'*xY.X0)\xY.X0');
+Qp     = DCM.Ep;
+
+% neuronal responses
+%--------------------------------------------------------------------------
+dp    = exp(-6);
+Cp    = spm_unvec(spm_vec(Qp) + dp*spm_vec(C),Qp);
+x     = feval(M.IS,Qp,M,xU);
+dx    = feval(M.IS,Cp,M,xU);
+dx    = spm_vec(dx) - spm_vec(x);
+x     = spm_unvec(dx/dp,x);
+
+% trial-specific responses (in mode, channel and source space)
+%--------------------------------------------------------------------------
+x0    = ones(Ns,1)*spm_vec(M.x)';       % expansion point for states
+j     = kron(DCM.Eg.J,ones(1,Nr));      % Indices of contributing states
+j     = logical(j);
+for i = 1:numel(x)
+    x{i} = x{i} - x0;                   % centre on expansion point
+    x{i} = x{i}(:,j);                   % Depolarization in sources
+end
+
+
+% store estimates in DCM
+%--------------------------------------------------------------------------
+DCM.K  = x;                             % conditional sensitivity
+
+% display
+%--------------------------------------------------------------------------
+spm_dcm_erp_results(DCM,'ERPs (sources)');
diff --git a/toolbox/dcm_meeg/spm_dcm_ind.m b/toolbox/dcm_meeg/spm_dcm_ind.m
index d1bcf10..dd0fe05 100644
--- a/toolbox/dcm_meeg/spm_dcm_ind.m
+++ b/toolbox/dcm_meeg/spm_dcm_ind.m
@@ -23,7 +23,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_dcm_ind.m 3177 2009-06-03 08:47:41Z vladimir $
+% $Id: spm_dcm_ind.m 4185 2011-02-01 18:46:18Z guillaume $
 
 
 % check options 
@@ -176,7 +176,7 @@
 
 % and save
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7.1') >= 0
+if spm_check_version('matlab','7') >= 0
     save(DCM.name, '-V6', 'DCM');
 else
     save(DCM.name, 'DCM');
diff --git a/toolbox/dcm_meeg/spm_dcm_lock.m b/toolbox/dcm_meeg/spm_dcm_lock.m
new file mode 100644
index 0000000..82bf229
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_lock.m
@@ -0,0 +1,23 @@
+function [pC] = spm_dcm_lock(pV)
+% locks experimental effects by introducing prior correlations
+% FORMAT [pC] = spm_dcm_lock(pV)
+%__________________________________________________________________________
+%
+% pV   - prior variance
+% pC   - prior covariance
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_lock.m 4095 2010-10-22 19:37:51Z karl $
+
+% lock experimental effects by introducing prior correlations
+%==========================================================================
+pC    = spm_diag(spm_vec(pV));
+for i = 1:length(pV.B)
+    pB      = pV;
+    pB.B{i} = pB.B{i} - pB.B{i};
+    pB      = spm_vec(pV)  - spm_vec(pB);
+    pB      = sqrt(pB*pB') - diag(pB);
+    pC      = pC + pB;
+end
diff --git a/toolbox/dcm_meeg/spm_dcm_neural_priors.m b/toolbox/dcm_meeg/spm_dcm_neural_priors.m
index babad83..69e4aa5 100644
--- a/toolbox/dcm_meeg/spm_dcm_neural_priors.m
+++ b/toolbox/dcm_meeg/spm_dcm_neural_priors.m
@@ -1,16 +1,16 @@
 function [pE,pC] = spm_dcm_neural_priors(A,B,C,model)
-% Prepares the priros on the paramters of neural mass models
+% Prepares the priors on the parameters of neural mass models
 % FORMAT [pE,pC] = spm_dcm_neural_priors(A,B,C,'model'))
 %
 % A,B{m},C  - binary constraints on extrinsic connections for m conditions
-% 'model'   - 'ERP','SEP','LFP','NNM' or 'MFM'
+% 'model'   - 'ERP','SEP','CMC','LFP','NNM' or 'MFM'
 %
 % pE - prior expectation - f(x,u,P,M)
 %
-% synaptic parameters
+% synaptic parameters (for NMN and MFM)
 %--------------------------------------------------------------------------
-%    pE.T - syaptic time constants
-%    pE.H - syaptic densities
+%    pE.T - synaptic time constants
+%    pE.H - synaptic densities
 %    pE.S - activation function parameters
 %
 % connectivity parameters
@@ -30,7 +30,7 @@
 %    pE.R - onset and dispersion
 %    pE.X - endogenous activity
 %
-% pC - prior covariances: cov(spm_vec(pE))
+% pC - prior (co)variances
 %
 % Because priors are specified under log normal assumptions, most
 % parameters are simply scaling coefficients with a prior expectation
@@ -38,54 +38,52 @@
 % pC = 1;
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
-
+ 
 % Karl Friston
-% $Id: spm_dcm_neural_priors.m 2374 2008-10-21 18:52:29Z karl $
-
+% $Id: spm_dcm_neural_priors.m 4281 2011-03-31 19:49:57Z karl $
+ 
 % check options
 %==========================================================================
-
+ 
 % get priors on neural model
 %--------------------------------------------------------------------------
 switch lower(model)
-
+ 
     % linear David et al model (linear in states)
     %======================================================================
-    case{'erp'}
-
+    case{'erp','sep'}
+ 
         % prior moments on parameters
         %------------------------------------------------------------------
         [pE,pC] = spm_erp_priors(A,B,C);
-
-
-    % linear David et al model (linear in states) - fast version for SEPs
-    %======================================================================
-    case{'sep'}
-
+        
+        
+    % linear David et al model (Canonical microcircuit)
+    %======================================================================        
+    case{'cmc'}
+ 
         % prior moments on parameters
         %------------------------------------------------------------------
-        [pE,pC]   = spm_erp_priors(A,B,C);
-        pE.T      = pE.T - 1;
-        pE.R(:,2) = pE.R(:,2) - 2;
-        
+        [pE,pC] = spm_cmc_priors(A,B,C);
         
+ 
     % linear David et al model (linear in states)  - with self-inhibition
     %======================================================================
     case{'lfp'}
-
+ 
         % prior moments on parameters
         %------------------------------------------------------------------
         [pE,pC] = spm_lfp_priors(A,B,C);
-
-
+ 
+ 
     % Neural mass model (nonlinear in states)
     %======================================================================
     case{'nmm','mfm'}
-
+ 
         % prior moments on parameters
         %------------------------------------------------------------------
         [pE,pC] = spm_nmm_priors(A,B,C);
-
+ 
     otherwise
         warndlg('Unknown model')
 end
diff --git a/toolbox/dcm_meeg/spm_dcm_phase.m b/toolbox/dcm_meeg/spm_dcm_phase.m
index edd679b..d3ddc0b 100644
--- a/toolbox/dcm_meeg/spm_dcm_phase.m
+++ b/toolbox/dcm_meeg/spm_dcm_phase.m
@@ -28,7 +28,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 %
 % Will Penny
-% $Id: spm_dcm_phase.m 3497 2009-10-21 21:54:28Z vladimir $
+% $Id: spm_dcm_phase.m 4281 2011-03-31 19:49:57Z karl $
 
 
 % check options 
@@ -76,7 +76,7 @@
 
 % Set error covariance components - one per region
 %--------------------------------------------------------------------------
-xY.Q = spm_Ce(Ns*Nt*ones(1,Nr));
+xY.Q   = spm_Ce(Ns*Nt*ones(1,Nr));
 
 % Offsets
 xY.X0  = sparse(Ns,0);
@@ -140,7 +140,7 @@
 % Set up initial phases 
 for n=1:length(DCM.xY.y)
     xx=DCM.xY.y{n}(1,:);
-    M.trial{n}.x=double(xx(:)');
+    M.trial{n}.x = double(xx(:)');
 end
 
 
@@ -149,7 +149,7 @@
 
 % EM: inversion
 %--------------------------------------------------------------------------
-[Qp,Qg,Cp,Cg,Ce,F] = spm_nlsi_N (M,xU,xY);
+[Qp,Qg,Cp,Cg,Ce,F] = spm_nlsi_N(M,xU,xY);
 
 
 % Data ID
@@ -175,17 +175,15 @@
 
 % neuronal and sensor responses (x and y)
 %--------------------------------------------------------------------------
-
 x   = feval(M.IS,Qp,M,xU);        % prediction (source space)
-
 L   = feval(M.G, Qg,M);           % get gain matrix
 
 
 % trial-specific responses (in mode, channel and source space)
 %--------------------------------------------------------------------------
 for i = 1:Nt
-    s  = x{i};                   % prediction (source space)
-    y{i}  = s*L';                   % prediction (sensor space)
+    s    = x{i};                  % prediction (source space)
+    y{i} = s*L';                  % prediction (sensor space)
     %r  = xY.y{i} - y;            % residuals  (sensor space)
     
 end
@@ -203,14 +201,14 @@
 DCM.Eg = Qg;                   % conditional expectation
 DCM.Cg = Cg;                   % conditional covariances
 DCM.Pp = Pp;                   % conditional probability
-DCM.Ce = Ce;                   % ReML error covariance
+DCM.Ce = Ce;                   % conditional error covariance
 DCM.F  = F;                    % Laplace log evidence
 DCM.ID = ID;                   % data ID
-DCM.y = y;                     % Model predictions
+DCM.y  = y;                    % Model predictions
 
 % and save
 %--------------------------------------------------------------------------
-if spm_matlab_version_chk('7.1') >= 0
+if spm_check_version('matlab','7') >= 0
     save(DCM.name, '-V6', 'DCM');
 else
     save(DCM.name, 'DCM');
diff --git a/toolbox/dcm_meeg/spm_dcm_phase_data.m b/toolbox/dcm_meeg/spm_dcm_phase_data.m
index a92e93c..57e4472 100644
--- a/toolbox/dcm_meeg/spm_dcm_phase_data.m
+++ b/toolbox/dcm_meeg/spm_dcm_phase_data.m
@@ -26,7 +26,7 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny
-% $Id: spm_dcm_phase_data.m 3847 2010-04-28 11:27:20Z vladimir $
+% $Id: spm_dcm_phase_data.m 4266 2011-03-29 11:16:06Z will $
 
 % Get data filename
 %-------------------------------------------------------------------------
@@ -152,7 +152,7 @@
         xr = ft_preproc_bandpassfilter(xr, fsample(D), DCM.options.Fdcm, 5);
 
         hx=spm_hilbert(xr);
-        DCM.xY.y{n}(:,c)=double(angle(hx));
+        DCM.xY.y{n}(:,c)=unwrap(double(angle(hx)));
     end
 end
 disp('Source extraction complete ...');
diff --git a/toolbox/dcm_meeg/spm_dcm_ssr.m b/toolbox/dcm_meeg/spm_dcm_ssr.m
index 17c7611..03bc74b 100644
--- a/toolbox/dcm_meeg/spm_dcm_ssr.m
+++ b/toolbox/dcm_meeg/spm_dcm_ssr.m
@@ -18,13 +18,13 @@
 %   options.Tdcm         - [start end] time window in ms
 %   options.Fdcm         - [start end] Frequency window in Hz
 %   options.D            - time bin decimation       (usually 1 or 2)
-%   options.type         - 'ECD', ‘LFP’ or ‘IMG’     (see spm_erp_L)
-%   options.model        - 'ECD', ‘SEP’, 'NMM' or ‘MFM’
+%   options.type         - 'ECD', 'LFP' or 'IMG'     (see spm_erp_L)
+%   options.model        - 'ECD', 'SEP', 'CMC', 'NMM' or 'MFM'
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_dcm_ssr.m 3497 2009-10-21 21:54:28Z vladimir $
+% $Id: spm_dcm_ssr.m 4281 2011-03-31 19:49:57Z karl $
 
 
 % check options
@@ -57,8 +57,10 @@
 [pE,pC]  = spm_dcm_neural_priors(DCM.A,DCM.B,DCM.C,model);
 
 try
-    pE   = DCM.M.pE;
-    fprintf('Using existing priors\n')
+    if length(spm_vec(DCM.M.pE)) == length(spm_vec(pE));
+        pE = DCM.M.pE;
+        fprintf('Using existing priors\n')
+    end
 end
 
 % augment with priors on spatial model
@@ -83,6 +85,8 @@
 DCM.M.n  = length(spm_vec(x));
 DCM.M.pE = pE;
 DCM.M.pC = pC;
+DCM.M.hE = 6;
+DCM.M.hC = 1/32;
 DCM.M.m  = Ns;
 
 %-Feature selection using principal components (U) of lead-field
@@ -105,25 +109,33 @@
 
 % get data-features (in reduced eigen-space)
 %--------------------------------------------------------------------------
-DCM      = spm_dcm_ssr_data(DCM);
+DCM        = spm_dcm_ssr_data(DCM);
+DCM.xY.y   = spm_unvec(abs(spm_vec(DCM.xY.y)),DCM.xY.y);
 
 
 % complete model specification and invert
 %==========================================================================
-Nm       = size(DCM.M.U,2);               % number of spatial modes
-Nt       = size(DCM.xY.y,1);              % number of trials
-Nf       = size(DCM.xY.y{1},1);           % number of frequency bins
-DCM.M.l  = Nm;
-DCM.M.Hz = DCM.xY.Hz;
+Nm         = size(DCM.M.U,2);                    % number of spatial modes
+Nt         = size(DCM.xY.y,1);                   % number of trials
+Nf         = size(DCM.xY.y{1},1);                % number of frequency bins
+DCM.M.l    = Nm;
+DCM.M.Hz   = DCM.xY.Hz;
 
 % precision of noise: AR(1/2)
 %--------------------------------------------------------------------------
-DCM.xY.Q  = spm_Q(1/2,Nf,1);
-DCM.xY.X0 = sparse(Nf,0);
+DCM.xY.Q   = spm_Q(1/2,Nf,1);
+DCM.xY.X0  = sparse(Nf,0);
+
+% adjust priors on gain to accomodate scaling differences among models
+%--------------------------------------------------------------------------
+Hc         = feval(DCM.M.IS,DCM.M.pE,DCM.M,DCM.xU);
+DCM.M.U        = U*sqrt(norm(spm_vec(DCM.xY.y),'inf')/norm(spm_vec(Hc,'inf')));
+
 
 % EM: inversion
 %--------------------------------------------------------------------------
-[Qp,Cp,Ce,F] = spm_nlsi_GN(DCM.M,DCM.xU,DCM.xY);
+[Qp,Cp,Eh,F] = spm_nlsi_GN(DCM.M,DCM.xU,DCM.xY);
+Ce           = exp(-Eh);
 
 % Data ID
 %==========================================================================
@@ -167,7 +179,7 @@
 %--------------------------------------------------------------------------
 DCM.options.Nmodes = Nm;
 
-if spm_matlab_version_chk('7.1') >= 0
+if spm_check_version('matlab','7') >= 0
     save(DCM.name, '-V6', 'DCM');
 else
     save(DCM.name, 'DCM');
diff --git a/toolbox/dcm_meeg/spm_dcm_ssr_data.m b/toolbox/dcm_meeg/spm_dcm_ssr_data.m
index 0b57956..929e090 100644
--- a/toolbox/dcm_meeg/spm_dcm_ssr_data.m
+++ b/toolbox/dcm_meeg/spm_dcm_ssr_data.m
@@ -26,7 +26,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_dcm_ssr_data.m 2773 2009-02-23 13:08:55Z vladimir $
+% $Id: spm_dcm_ssr_data.m 4096 2010-10-22 19:40:34Z karl $
  
 % Set defaults and Get D filename
 %-------------------------------------------------------------------------
@@ -135,6 +135,7 @@
     T2          = DCM.options.Tdcm(2);
     [i, T1]     = min(abs(DCM.xY.Time - T1));
     [i, T2]     = min(abs(DCM.xY.Time - T2));
+    
     % Time [ms] of down-sampled data
     %----------------------------------------------------------------------
     It          = [T1:DT:T2]';               % indices - bins
@@ -176,7 +177,7 @@
     if size(DCM.xY.csd,2) == Ne;
         if size(DCM.xY.csd{1},1) == Nf;
             if size(DCM.xY.csd{1},2) == Nm;
-                DCM.xY.y  = spm_cond_units(DCM.xY.csd);
+                DCM.xY.y  = spm_cond_units(DCM.xY.csd,'csd');
                 return
             end
         end
@@ -214,11 +215,10 @@
     %----------------------------------------------------------------------
     DCM.xY.csd{i} = P/Nt;
    
-    
 end
  
 % place cross-spectral density in xY.y
 %==========================================================================
-DCM.xY.y    = spm_cond_units(DCM.xY.csd); 
+DCM.xY.y    = spm_cond_units(DCM.xY.csd,'csd'); 
 DCM.xY.U    = DCM.M.U;
 DCM.xY.code = condlabels(trial);
\ No newline at end of file
diff --git a/toolbox/dcm_meeg/spm_dcm_ssr_results.m b/toolbox/dcm_meeg/spm_dcm_ssr_results.m
index 2c2c47a..5c2e1b6 100644
--- a/toolbox/dcm_meeg/spm_dcm_ssr_results.m
+++ b/toolbox/dcm_meeg/spm_dcm_ssr_results.m
@@ -27,7 +27,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_dcm_ssr_results.m 2374 2008-10-21 18:52:29Z karl $
+% $Id: spm_dcm_ssr_results.m 4096 2010-10-22 19:40:34Z karl $
  
  
 % get figure handle
@@ -39,7 +39,7 @@
 
 % placespectral features in xY.y
 %--------------------------------------------------------------------------
-DCM.xY.y  = spm_cond_units(DCM.xY.csd);
+DCM.xY.y  = spm_cond_units(DCM.xY.csd,'csd');
 
 % trial data
 %--------------------------------------------------------------------------
@@ -264,9 +264,14 @@
     
     % spectrum of innovations or noise (Gu)
     %----------------------------------------------------------------------
-    Gu   = exp(DCM.Ep.a)*xY.Hz.^(-1)*2;    % spectral density of (AR) input
-    Gu   = Gu + exp(DCM.Ep.b);             % spectral density of IID input
- 
+    try
+        Gu   = exp(DCM.Ep.a)*xY.Hz.^(-1)*2;    % spectral density of (AR) input
+        Gu   = Gu + exp(DCM.Ep.b);             % spectral density of IID input
+    catch
+        Gu   = exp(DCM.Ep.a(1))*xY.Hz.^(-1);    % spectral density of (AR) input
+        Gu   = Gu + exp(DCM.Ep.a(2));           % spectral density of IID input
+    end
+    
     % plot spectral density of innovations
     % ---------------------------------------------------------------------
     subplot(2,1,1)
diff --git a/toolbox/dcm_meeg/spm_dcm_symm.m b/toolbox/dcm_meeg/spm_dcm_symm.m
new file mode 100644
index 0000000..4cea094
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_dcm_symm.m
@@ -0,0 +1,64 @@
+function [pC] = spm_dcm_symm(pV,pE)
+% locks ECD orientations by introducing prior correlations
+% FORMAT [pC] = spm_dcm_symm(pV,pE)
+%__________________________________________________________________________
+%
+% pE   - prior expectation
+% pV   - prior variance
+% pC   - prior covariance
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_dcm_symm.m 4095 2010-10-22 19:37:51Z karl $
+
+% Distance between homolgous sources (16mm)
+%--------------------------------------------------------------------------
+V     = 16;
+
+% symmetry constraints (based on Euclidean distance from mirror image)
+%==========================================================================
+
+% diagonalise feilds
+%--------------------------------------------------------------------------
+feilds = fieldnames(pV);
+for  i = 1:length(feilds)
+    pF = getfield(pV,feilds{i});    
+    pV = setfield(pV,feilds{i},spm_diag(spm_vec(pF)));
+end
+
+% impose correlations between orientations (L)
+%--------------------------------------------------------------------------
+n     = size(pE.Lpos,2);
+Mpos  = [-pE.Lpos(1,:); pE.Lpos(2,:); pE.Lpos(3,:)];
+D     = Inf*ones(n);
+for i = 1:n
+    for j = 1:n
+        if sign(pE.Lpos(1,i)) == sign(Mpos(1, j))
+            D(i,j) = sqrt(sum(pE.Lpos(:,i) - Mpos(:,j)).^2);
+        end
+    end
+end
+D     = (D + D')/2;
+DD    = zeros(n);
+for i = 1:n
+    [M, I] = min(D(i,:));
+    if M < V
+        DD(i,I) = 1;
+    end
+end
+try
+    pV.L = pV.L - kron(DD,diag([1 0 0])) + kron(DD,diag([0 1 1]));
+end
+
+% and concatenate
+%--------------------------------------------------------------------------
+for  i = 1:length(feilds)
+    pF      = getfield(pV,feilds{i});
+    if ~isempty(pF)
+        pC{i,i} = pF;  
+    end
+end
+pC    = spm_cat(pC);
+
+
diff --git a/toolbox/dcm_meeg/spm_dcm_x_neural.m b/toolbox/dcm_meeg/spm_dcm_x_neural.m
index 0d5958f..74ccb90 100644
--- a/toolbox/dcm_meeg/spm_dcm_x_neural.m
+++ b/toolbox/dcm_meeg/spm_dcm_x_neural.m
@@ -3,70 +3,79 @@
 % FORMAT [x,f] = spm_dcm_x_neural(P,'model')
 %
 %  P      - parameter structure
-% 'model' - 'ERP','SEP','LFP','NNM' or 'MFM'
+% 'model'   - 'ERP','SEP','CMC','LFP','NNM' or 'MFM'
 %
 % x   - initial states
 % f   - state euquation
 %__________________________________________________________________________
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
-
+ 
 % Karl Friston
-% $Id: spm_dcm_x_neural.m 2374 2008-10-21 18:52:29Z karl $
-
-
+% $Id: spm_dcm_x_neural.m 4232 2011-03-07 21:01:16Z karl $
+ 
+ 
 % initial state and equation
 %--------------------------------------------------------------------------
 switch lower(model)
-
+ 
     % linear David et al model (linear in states)
     %======================================================================
     case{'erp'}
-
+ 
         % inital states and equations of motion
         %------------------------------------------------------------------
         x  =  spm_x_erp(P);
         f  = 'spm_fx_erp';
-
-
+ 
+ 
     % linear David et al model (linear in states) - fast version for SEPs
     %======================================================================
     case{'sep'}
-
+ 
         % inital states
         %------------------------------------------------------------------
         x  =  spm_x_erp(P);
-        f  = 'spm_fx_erp';
+        f  = 'spm_fx_sep';
+        
+    % Linear in states – canonical microcircuit
+    %======================================================================
+    case{'cmc'}
+ 
+        % inital states
+        %------------------------------------------------------------------
+        x  =  spm_x_cmc(P);
+        f  = 'spm_fx_cmc';
         
     % linear David et al model (linear in states) - with self-inhibition
     %======================================================================
     case{'lfp'}
-
+ 
         % inital states
         %------------------------------------------------------------------
         x  =  spm_x_lfp(P);
         f  = 'spm_fx_lfp';
-
-
+ 
+ 
     % Neural mass model (nonlinear in states)
     %======================================================================
     case{'nmm'}
-
+ 
         % inital states and model
         %------------------------------------------------------------------
         x  = spm_x_nmm(P);
         f  = 'spm_fx_mfm';
-
-
+ 
+ 
     % Mean field model (nonlinear in states) - with covariance
     %======================================================================
     case{'mfm'}
-
+ 
         % inital states and model
         %------------------------------------------------------------------
         x  = spm_x_mfm(P);
         f  = 'spm_fx_mfm';
         
-
+ 
     otherwise
         warndlg('Unknown model')
 end
diff --git a/toolbox/dcm_meeg/spm_erp_L.m b/toolbox/dcm_meeg/spm_erp_L.m
index 106a7f0..7e47d09 100644
--- a/toolbox/dcm_meeg/spm_erp_L.m
+++ b/toolbox/dcm_meeg/spm_erp_L.m
@@ -25,7 +25,7 @@
 % Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_erp_L.m 3833 2010-04-22 14:49:48Z vladimir $
+% $Id: spm_erp_L.m 4200 2011-02-14 18:19:17Z vladimir $
 
 % Create a persient variable that rembers the last locations
 %--------------------------------------------------------------------------
@@ -58,7 +58,11 @@
         %----------------------------------------------------------
         LastLpos = P.Lpos;
         for i  = Id
-            Lf = ft_compute_leadfield(transform_points(M.dipfit.datareg.fromMNI, P.Lpos(:,i)'), M.dipfit.sens, M.dipfit.vol);
+            if any(P.Lpos(:,i)>=200)
+                Lf = zeros(M.dipfit.Nc, 3);
+            else
+                Lf = ft_compute_leadfield(transform_points(M.dipfit.datareg.fromMNI, P.Lpos(:,i)'), M.dipfit.sens, M.dipfit.vol);
+            end
             LastL(:,:,i) = Lf;
         end
         G     = spm_cond_units(LastL);
@@ -99,7 +103,7 @@
         catch
             n = m;
         end
-        L     = sparse(1:m,1:m,P.L,m,n);
+        L     = sparse(1:m,1:m,exp(P.L),m,n);
 
     otherwise
         warndlg('unknown spatial model')
diff --git a/toolbox/dcm_meeg/spm_erp_priors.m b/toolbox/dcm_meeg/spm_erp_priors.m
index 9adf860..d1b1539 100644
--- a/toolbox/dcm_meeg/spm_erp_priors.m
+++ b/toolbox/dcm_meeg/spm_erp_priors.m
@@ -1,4 +1,4 @@
-function [pE,pC] = spm_erp_priors(A,B,C)
+function [E,V] = spm_erp_priors(A,B,C)
 % prior moments for a neural-mass model of ERPs
 % FORMAT [pE,pC] = spm_erp_priors(A,B,C)
 %
@@ -11,6 +11,7 @@
 %    pE.T - syaptic time constants
 %    pE.H - syaptic densities
 %    pE.S - activation function parameters
+%    pE.G - intrinsic connection strengths
 %
 % connectivity parameters
 %--------------------------------------------------------------------------
@@ -23,7 +24,7 @@
 %--------------------------------------------------------------------------
 %    pE.R - onset and dispersion
 %
-% pC - prior covariances: cov(spm_vec(pE))
+% pC - prior (co)variances
 %
 % Because priors are specified under log normal assumptions, most
 % parameters are simply scaling coefficients with a prior expectation
@@ -38,7 +39,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_erp_priors.m 2374 2008-10-21 18:52:29Z karl $
+% $Id: spm_erp_priors.m 4281 2011-03-31 19:49:57Z karl $
  
 % default: a single source model
 %--------------------------------------------------------------------------
@@ -51,21 +52,20 @@
 % disable log zero warning
 %--------------------------------------------------------------------------
 warning('off','MATLAB:log:logOfZero');
-n     = size(C,1);                                 % number of sources
-u     = size(C,2);                                 % number of inputs
+n     = size(C,1);                                % number of sources
+u     = size(C,2);                                % number of inputs
 
 % parameters for neural-mass forward model
 %==========================================================================
-n1    = ones(n,1);
  
 % set intrinsic [excitatory] time constants
 %--------------------------------------------------------------------------
-E.T   = log(n1);        V.T = n1/8;                % time constants
-E.H   = log(n1);        V.H = n1/32;               % synaptic density
- 
+E.T   = sparse(n,2);  V.T = sparse(n,2) + 1/16;   % time constants
+E.H   = sparse(n,2);  V.H = sparse(n,2) + 1/16;   % synaptic density
+
 % set parameter of activation function
 %--------------------------------------------------------------------------
-E.S   = [0 0];          V.S = [1 1]/8;             % dispersion & threshold
+E.S   = [0 0];        V.S = [1 1]/16;             % dispersion & threshold
  
  
 % set extrinsic connectivity
@@ -73,43 +73,43 @@
 Q     = sparse(n,n);
 for i = 1:length(A)
       A{i} = ~~A{i};
-    E.A{i} = A{i}*32 - 32;                         % forward
-    V.A{i} = A{i}/16;                              % backward
-    Q      = Q | A{i};                             % and lateral connections
+    E.A{i} = A{i}*32 - 32;                        % forward
+    V.A{i} = A{i}/16;                             % backward
+    Q      = Q | A{i};                            % and lateral connections
 end
  
 for i = 1:length(B)
       B{i} = ~~B{i};
-    E.B{i} = 0*B{i};                               % input-dependent scaling
-    V.B{i} = B{i}/16;
+    E.B{i} = 0*B{i};                              % input-dependent scaling
+    V.B{i} = B{i}/8;
     Q      = Q | B{i};
 end
 C      = ~~C;
-E.C    = C*32 - 32;                                % where inputs enter
+E.C    = C*32 - 32;                               % where inputs enter
 V.C    = C/32;
  
-% set delay (enforcing symmetric delays)
+% set intrinsic connectivity
+%--------------------------------------------------------------------------
+E.G    = sparse(1,4);
+V.G    = sparse(1,4) + 1/16;
+
+% set delay
 %--------------------------------------------------------------------------
-E.D        = sparse(n,n);
-V.D        = Q/16;
+E.D    = sparse(n,n);
+V.D    = Q/16;
  
 % set stimulus parameters: onset and dispersion
 %--------------------------------------------------------------------------
-E.R        = sparse(u,2);  V.R   = ones(u,1)*[1/16 1/16];
+E.R    = sparse(u,2);  V.R   = ones(u,1)*[1/16 1/16];
 warning('on','MATLAB:log:logOfZero');
 
- 
-% covariance of neural parameters
-%==========================================================================
-pE    = E;
-pC    = diag(sparse(spm_vec(V)));
+return
 
 
-return
  
 % demo for log-normal pdf
 %==========================================================================
-x    = [1:64]/16;
+x  = (1:64)/16;
 for i = [2 16 128]
     v = 1/i;
     p = 1./x.*exp(-log(x).^2/(2*v))/sqrt(2*pi*v);
diff --git a/toolbox/dcm_meeg/spm_fs_csd.m b/toolbox/dcm_meeg/spm_fs_csd.m
new file mode 100644
index 0000000..a644f10
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_fs_csd.m
@@ -0,0 +1,32 @@
+function [y] = spm_fs_csd(y,M)
+% spectral feature selection for a CSD DCM
+% FORMAT [y] = spm_fs_csd(y,M)
+% y      - CSD
+% M      - model structure
+%__________________________________________________________________________
+%
+% This simply log-transforms the (real) auto-spectra
+%
+% David O, Friston KJ (2003) A neural mass model for MEG/EEG: coupling and
+% neuronal dynamics. NeuroImage 20: 1743-1755
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fs_csd.m 4281 2011-03-31 19:49:57Z karl $
+
+
+% just return auto-spectra at present
+%--------------------------------------------------------------------------
+y     = y;
+
+return
+
+% log-transform (real) auto-spectra
+%--------------------------------------------------------------------------
+for i = 1:length(y);
+    for j = 1:size(y{i},2)
+        y{i}(:,j,j) = log(y{i}(:,j,j));
+    end
+end
+
diff --git a/toolbox/dcm_meeg/spm_fx_cmc.m b/toolbox/dcm_meeg/spm_fx_cmc.m
new file mode 100644
index 0000000..c572d73
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_fx_cmc.m
@@ -0,0 +1,189 @@
+function [f,J,Q] = spm_fx_cmc(x,u,P,M)
+% state equations for a neural mass model of erps
+% FORMAT [f,J,D] = spm_fx_cmc(x,u,P,M)
+% FORMAT [f,J]   = spm_fx_cmc(x,u,P,M)
+% FORMAT [f]     = spm_fx_cmc(x,u,P,M)
+% x      - state vector
+%   x(:,1) - voltage     (spiny stellate cells)
+%   x(:,2) - conductance (spiny stellate cells)
+%   x(:,3) - voltage     (superficial pyramidal cells)
+%   x(:,4) - conductance (superficial pyramidal cells)
+%   x(:,5) - current     (inhibitory interneurons)
+%   x(:,6) - conductance (inhibitory interneurons)
+%   x(:,7) - voltage     (deep pyramidal cells)
+%   x(:,8) - conductance (deep pyramidal cells)
+%
+% f        - dx(t)/dt  = f(x(t))
+% J        - df(t)/dx(t)
+% D        - delay operator dx(t)/dt = f(x(t - d))
+%                                    = D(d)*f(x(t))
+%
+% Prior fixed parameter scaling [Defaults]
+%
+% E  = (forward, backward, lateral) extrinsic rates 
+% G  = intrinsic rates
+% D  = propagation delays (intrinsic, extrinsic)
+% H  = overall synaptic kinetics
+% T  = synaptic time constants
+% R  = slope of sigmoid activation function
+%
+%__________________________________________________________________________
+% David O, Friston KJ (2003) A neural mass model for MEG/EEG: coupling and
+% neuronal dynamics. NeuroImage 20: 1743-1755
+%___________________________________________________________________________
+% Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fx_cmc.m 4261 2011-03-24 16:39:42Z karl $
+ 
+ 
+% get dimensions and configure state variables
+%--------------------------------------------------------------------------
+M.u   = u;                          % place inputs in M
+x     = spm_unvec(x,M.x);           % neuronal states
+[n m] = size(x);                    % number of sources and states  
+
+ 
+% [default] fixed parameters
+%--------------------------------------------------------------------------
+E  = [1 1 1 1/4]*200;                   % extrinsic (forward and backward)  
+G  = [4 4 4 4 4 2 4 4 2 1]*200;     % intrinsic connections
+D  = [1 16];                        % delays (intrinsic, extrinsic)
+T  = [2 2 28 28];                   % synaptic time constants
+R  = 1;                             % slope of sigmoid activation function
+ 
+
+% [specified] fixed parameters
+%--------------------------------------------------------------------------
+try, E = M.pF.E; end
+try, G = M.pF.G; end
+try, D = M.pF.D; end
+try, H = M.pF.H; end
+try, T = M.pF.T; end
+try, R = M.pF.R; end
+ 
+ 
+% Extrinsic connections
+%--------------------------------------------------------------------------
+A{1} = exp(P.A{1})*E(1);
+A{2} = exp(P.A{2})*E(2);
+A{3} = exp(P.A{3})*E(3);
+A{4} = exp(P.A{4})*E(4);
+C    = exp(P.C);
+ 
+% pre-synaptic inputs: s(V)
+%--------------------------------------------------------------------------
+R    = R.*exp(P.S);
+S    = 1./(1 + exp(-R*x)) - 1/2;
+ 
+% exogenous input
+%--------------------------------------------------------------------------
+U    = C*u(:);
+ 
+ 
+% time constants and intrinsic connections
+%==========================================================================
+T    = ones(n,1)*T/1000;
+G    = ones(n,1)*G;
+
+% free parameters on time constants and intrinsic connections
+%--------------------------------------------------------------------------
+% G(:,1)  ss -> ss
+% G(:,2)  sp -> ss
+% G(:,3)  ii -> ss
+% G(:,4)  ii -> ii
+% G(:,5)  ss -> ii
+% G(:,6)  dp -> ii
+% G(:,7)  sp -> sp
+% G(:,8)  ss -> sp
+% G(:,9)  ii -> dp
+% G(:,10) dp -> dp
+%--------------------------------------------------------------------------
+for i = 1:size(P.T,2)
+    T(:,i) = T(:,i).*exp(P.T(:,i));
+end
+for i = 1:size(P.G,2)
+    G(:,i) = G(:,i).*exp(P.G(:,i));
+end
+
+
+% Motion of states: f(x)
+%--------------------------------------------------------------------------
+ 
+% Conductance
+%==========================================================================
+ 
+% Granular layer (excitatory interneurons): spiny stellate: Hidden causes
+%--------------------------------------------------------------------------
+u      =   A{1}*S(:,3) + U;
+u      = - G(:,1).*S(:,1) - G(:,3).*S(:,5) - G(:,2).*S(:,3) + u;
+f(:,2) = (u - 2*x(:,2) - x(:,1)./T(:,1))./T(:,1);
+ 
+% Supra-granular layer (superficial pyramidal cells): Hidden causes - error
+%--------------------------------------------------------------------------
+u      = - A{3}*S(:,7);
+u      =   G(:,8).*S(:,1) - G(:,7).*S(:,3) + u;
+f(:,4) = (u - 2*x(:,4) - x(:,3)./T(:,2))./T(:,2);
+ 
+% Supra-granular layer (inhibitory interneurons): Hidden states - error
+%--------------------------------------------------------------------------
+u      = - A{4}*S(:,3);
+u      =   G(:,5).*S(:,1) + G(:,6).*S(:,7) - G(:,4).*S(:,5) + u;
+f(:,6) = (u - 2*x(:,6) - x(:,5)./T(:,3))./T(:,3);
+ 
+% Infra-granular layer (deep pyramidal cells): Hidden states
+%--------------------------------------------------------------------------
+u      =   A{2}*S(:,3);
+u      = - G(:,10).*S(:,7) - G(:,9).*S(:,5) + u;
+f(:,8) = (u - 2*x(:,8) - x(:,7)./T(:,4))./T(:,4);
+ 
+% Voltage
+%==========================================================================
+f(:,1) = x(:,2);
+f(:,3) = x(:,4);
+f(:,5) = x(:,6);
+f(:,7) = x(:,8);
+f      = diag(exp(P.H))*f;
+f      = spm_vec(f);
+ 
+ 
+if nargout == 1; return, end
+ 
+ 
+% delays
+%==========================================================================
+% Delay differential equations can be integrated efficiently (but
+% approximately) by absorbing the delay operator into the Jacobian
+%
+%    dx(t)/dt     = f(x(t - d))
+%                 = Q(d)f(x(t))
+%
+%    J(d)         = Q(d)df/dx
+%--------------------------------------------------------------------------
+De = exp(P.D);
+Di = diag(diag(De));
+De = De - Di;
+De = De*D(2)/1000;
+Di = Di*D(1)/1000;
+De = kron(ones(m,m),De);
+Di = kron(ones(m,m) - speye(m,m),Di);
+D  = Di + De;
+ 
+% Implement: dx(t)/dt = f(x(t - d)) = inv(1 + D.*dfdx)*f(x(t))
+%                     = Q*f = Q*J*x(t)
+%--------------------------------------------------------------------------
+[Q,J] = spm_dcm_delay(M,P,D);
+ 
+ 
+return
+ 
+% notes and alpha function (kernels)
+%==========================================================================
+% x   = t*exp(k*t)
+% x'  = exp(k*t) + k*t*exp(k*t)
+%     = exp(k*t) + k*x
+% x'' = 2*k*exp(k*t) + k^2*t*exp(k*t)
+%     = 2*k*(x' - k*x) + k^2*x
+%     = 2*k*x' - k^2*x
+
+
diff --git a/toolbox/dcm_meeg/spm_fx_erp.m b/toolbox/dcm_meeg/spm_fx_erp.m
index a0c2340..51de2e8 100644
--- a/toolbox/dcm_meeg/spm_fx_erp.m
+++ b/toolbox/dcm_meeg/spm_fx_erp.m
@@ -26,7 +26,7 @@
 %  M.pF.D = [2 16];              % propogation delays (intrinsic, extrinsic)
 %  M.pF.H = [4 32];              % receptor densities (excitatory, inhibitory)
 %  M.pF.T = [8 16];              % synaptic constants (excitatory, inhibitory)
-%  M.pF.R = 0.56;                % parameter of static nonlinearity
+%  M.pF.R = [1 1/2];             % parameter of static nonlinearity
 %
 %__________________________________________________________________________
 % David O, Friston KJ (2003) A neural mass model for MEG/EEG: coupling and
@@ -35,41 +35,39 @@
 % Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_fx_erp.m 2374 2008-10-21 18:52:29Z karl $
-
-
+% $Id: spm_fx_erp.m 4281 2011-03-31 19:49:57Z karl $
 
 
 % get dimensions and configure state variables
 %--------------------------------------------------------------------------
-n  = length(P.A{1});             % number of sources
-x  = spm_unvec(x,M.x);           % neuronal states
+n = length(P.A{1});         % number of sources
+x = spm_unvec(x,M.x);       % neuronal states
 
 % [default] fixed parameters
 %--------------------------------------------------------------------------
-try
-    E  = M.pF.E;                % extrinsic rates (forward, backward, lateral)
-    G  = M.pF.G;                % intrinsic rates (g1, g2 g3, g4)
-    D  = M.pF.D;                % propogation delays (intrinsic, extrinsic)
-    H  = M.pF.H;                % receptor densities (excitatory, inhibitory)
-    T  = M.pF.T;                % synaptic constants (excitatory, inhibitory)
-    R  = M.pF.R;                % parameters of static nonlinearity
-catch
-    E = [32 16 4];              % extrinsic rates (forward, backward, lateral)
-    G = [1 4/5 1/4 1/4]*128;    % intrinsic rates (g1, g2 g3, g4)
-    D = [2 32];                 % propogation delays (intrinsic, extrinsic)
-    H = [4 32];                 % receptor densities (excitatory, inhibitory)
-    T = [8 16];                 % synaptic constants (excitatory, inhibitory)
-    R = [2 1]/3;                % parameters of static nonlinearity
-end
+E = [1 1/2 1/8]*32;         % extrinsic rates (forward, backward, lateral)
+G = [1 4/5 1/4 1/4]*128;    % intrinsic rates (g1 g2 g3 g4)
+D = [2 16];                 % propogation delays (intrinsic, extrinsic)
+H = [4 32];                 % receptor densities (excitatory, inhibitory)
+T = [8 16];                 % synaptic constants (excitatory, inhibitory)
+R = [2 1]/3;                % parameters of static nonlinearity
+
+% [specified] fixed parameters
+%--------------------------------------------------------------------------
+try, E = M.pF.E; end
+try, G = M.pF.G; end
+try, D = M.pF.D; end
+try, H = M.pF.H; end
+try, T = M.pF.T; end
+try, R = M.pF.R; end
 
 
 % test for free parameters on intrinsic connections
 %--------------------------------------------------------------------------
-G     = ones(n,1)*G;
 try
     G = G.*exp(P.G);
 end
+G     = ones(n,1)*G;
 
 % exponential transform to ensure positivity constraints
 %--------------------------------------------------------------------------
@@ -80,10 +78,10 @@
 
 % intrinsic connectivity and parameters
 %--------------------------------------------------------------------------
-Te    = T(1)/1000*exp(P.T);           % excitatory time constants
-Ti    = T(2)/1000;                    % inhibitory time constants
-Hi    = H(2);                         % inhibitory receptor density
-He    = H(1)*exp(P.H);                % excitatory receptor density
+Te    = T(1)/1000*exp(P.T(:,1));         % excitatory time constants
+Ti    = T(2)/1000*exp(P.T(:,2));         % inhibitory time constants
+He    = H(1)*exp(P.H(:,1));              % excitatory receptor density
+Hi    = H(2)*exp(P.H(:,2));              % inhibitory receptor density
 
 % pre-synaptic inputs: s(V)
 %--------------------------------------------------------------------------
@@ -117,67 +115,18 @@
 % Infra-granular layer (pyramidal cells): hyperpolarizing current
 %--------------------------------------------------------------------------
 f(:,3) = x(:,6);
-f(:,6) = (Hi*G(:,4).*S(:,7) - 2*x(:,6) - x(:,3)/Ti)/Ti;
+f(:,6) = (Hi.*G(:,4).*S(:,7) - 2*x(:,6) - x(:,3)./Ti)./Ti;
 
 % Infra-granular layer (pyramidal cells): Voltage
 %--------------------------------------------------------------------------
 f(:,9) = x(:,5) - x(:,6);
-
-f      = f(:);
+f      = spm_vec(f);
 
 if nargout == 1; return, end
 
-
-
-% Jacobian: J = df(x)/dx
-%===========================================================================
-I  = speye(n,n);
-J  = kron(sparse(9,9),sparse(n,n));
-
-% changes in voltage with current
-%--------------------------------------------------------------------------
-S  = sparse([7 2 1 3 9],[8 5 4 6 5],1,9,9);  J = J + kron(S,I);
-S  = sparse(9,6,-1,9,9);                     J = J + kron(S,I);
-
-% synaptic kernel
-%--------------------------------------------------------------------------
-S  = sparse([8 4 5],[8 4 5],1,9,9); J = J - kron(S,diag(2./Te));
-S  = sparse(6,6,1,9,9);             J = J - kron(S,I)*2/Ti;
-S  = sparse([8 4 5],[7 1 2],1,9,9); J = J - kron(S,diag(1./(Te.*Te)));
-S  = sparse(6,3,1,9,9);             J = J - kron(S,I)/(Ti*Ti);
-
-% Supragranular layer (inhibitory interneurons)
-%--------------------------------------------------------------------------
-E  = (A{2} + A{3})*diag(dSdx(:,9)) + diag(G(:,3).*dSdx(:,9));
-E  = diag(He./Te)*E;
-S  = sparse(8,9,1,9,9); J = J + kron(S,E);
-
-% Granular layer (spiny stellate cells)
-%--------------------------------------------------------------------------
-E  = (A{1} + A{3})*diag(dSdx(:,9)) + diag(G(:,1).*dSdx(:,9));
-E  = diag(He./Te)*E;
-S  = sparse(4,9,1,9,9); J = J + kron(S,E);
-
-% Infra-granular layer (pyramidal cells)
-%--------------------------------------------------------------------------
-E  = (A{2} + A{3})*diag(dSdx(:,9));
-E  = diag(He./Te)*E;
-S  = sparse(5,9,1,9,9); J = J + kron(S,E);
-
-E  = diag(G(:,2).*dSdx(:,1));
-E  = diag(He./Te)*E;
-S  = sparse(5,1,1,9,9); J = J + kron(S,E);
-
-% Infra-granular layer (pyramidal cells)
-%--------------------------------------------------------------------------
-E  = diag(G(:,4).*dSdx(:,7));
-E  = Hi*E/Ti;
-S  = sparse(6,7,1,9,9); J = J + kron(S,E);
-
-if nargout == 2; return, end
-
-
-
+% Jacobian
+%==========================================================================
+J     = spm_diff(M.f,x,u,P,M,1);
 
 % delays
 %==========================================================================
@@ -189,8 +138,6 @@
 %
 %    J(d)         = Q(d)df/dx
 %--------------------------------------------------------------------------
-
-
 De = D(2).*exp(P.D)/1000;
 Di = D(1)/1000;
 De = (1 - speye(n,n)).*De;
diff --git a/toolbox/dcm_meeg/spm_fx_mfm.m b/toolbox/dcm_meeg/spm_fx_mfm.m
index 0d39666..5c64f7b 100644
--- a/toolbox/dcm_meeg/spm_fx_mfm.m
+++ b/toolbox/dcm_meeg/spm_fx_mfm.m
@@ -46,7 +46,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_fx_mfm.m 3653 2009-12-23 20:06:48Z karl $
+% $Id: spm_fx_mfm.m 4096 2010-10-22 19:40:34Z karl $
  
 % get dimensions and configure state variables
 %--------------------------------------------------------------------------
@@ -266,7 +266,15 @@
 %
 %    J(d)         = Q(d)df/dx
 %--------------------------------------------------------------------------
-d  = -[2 16].*exp(P.D)/1000;
+% [specified] fixed parameters
+%--------------------------------------------------------------------------
+try
+    D  = M.pF.D;
+catch
+    D  = [2 16];
+end
+
+d  = -D.*exp(P.D)/1000;
 nk = 3;                                               % number of states
 Sp = kron(ones(nk,nk),kron( eye(np,np),eye(ns,ns)));  % states: same pop.
 Ss = kron(ones(nk,nk),kron(ones(np,np),eye(ns,ns)));  % states: same source
diff --git a/toolbox/dcm_meeg/spm_fx_sep.m b/toolbox/dcm_meeg/spm_fx_sep.m
new file mode 100644
index 0000000..7e309ce
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_fx_sep.m
@@ -0,0 +1,168 @@
+function [f,J,Q] = spm_fx_sep(x,u,P,M)
+% state equations for a neural mass model of erps
+% FORMAT [f,J,D] = spm_fx_sep(x,u,P,M)
+% FORMAT [f,J]   = spm_fx_sep(x,u,P,M)
+% FORMAT [f]     = spm_fx_sep(x,u,P,M)
+% x      - state vector
+%   x(:,1) - voltage (spiny stellate cells)
+%   x(:,2) - voltage (pyramidal cells) +ve
+%   x(:,3) - voltage (pyramidal cells) -ve
+%   x(:,4) - current (spiny stellate cells)    depolarizing
+%   x(:,5) - current (pyramidal cells)         depolarizing
+%   x(:,6) - current (pyramidal cells)         hyperpolarizing
+%   x(:,7) - voltage (inhibitory interneurons)
+%   x(:,8) - current (inhibitory interneurons) depolarizing
+%   x(:,9) - voltage (pyramidal cells)
+%
+% f        - dx(t)/dt  = f(x(t))
+% J        - df(t)/dx(t)
+% D        - delay operator dx(t)/dt = f(x(t - d))
+%                                    = D(d)*f(x(t))
+%
+% Prior fixed parameter scaling [Defaults]
+%
+%  M.pF.E = [32 16 4];           % extrinsic rates (forward, backward, lateral)
+%  M.pF.G = [1 1 1 1/4]*128;     % intrinsic rates (g1, g2 g3, g4)
+%  M.pF.D = [1 16];              % propagation delays (intrinsic, extrinsic)
+%  M.pF.H = [4 64];              % receptor densities (excitatory, inhibitory)
+%  M.pF.T = [4 8];               % synaptic constants (excitatory, inhibitory)
+%  M.pF.R = [1 0];               % parameter of static nonlinearity
+%
+% This is just a faster version of spm_fx_erp
+%
+%__________________________________________________________________________
+% David O, Friston KJ (2003) A neural mass model for MEG/EEG: coupling and
+% neuronal dynamics. NeuroImage 20: 1743-1755
+%___________________________________________________________________________
+% Copyright (C) 2005 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_fx_sep.m 4281 2011-03-31 19:49:57Z karl $
+ 
+ 
+% get dimensions and configure state variables
+%--------------------------------------------------------------------------
+n  = length(P.A{1});         % number of sources
+x  = spm_unvec(x,M.x);       % neuronal states
+ 
+% [default] fixed parameters
+%--------------------------------------------------------------------------
+E  = [32 16 4];              % extrinsic rates (forward, backward, lateral)
+G  = [1 1 1 1/2]*64;         % intrinsic rates (g1 g2 g3 g4)
+D  = [1 16];                 % propagation delays (intrinsic, extrinsic)
+H  = [4 64];                 % receptor densities (excitatory, inhibitory)
+T  = [4 8];                  % synaptic constants (excitatory, inhibitory)
+R  = [1 0];                  % parameters of static nonlinearity
+   
+
+
+% [specified] fixed parameters
+%--------------------------------------------------------------------------
+try, E = M.pF.E; end
+try, G = M.pF.G; end
+try, D = M.pF.D; end
+try, H = M.pF.H; end
+try, T = M.pF.T; end
+try, R = M.pF.R; end
+ 
+ 
+% test for free parameters on intrinsic connections
+%--------------------------------------------------------------------------
+try
+    G = G.*exp(P.G);
+end
+G     = ones(n,1)*G;
+ 
+% exponential transform to ensure positivity constraints
+%--------------------------------------------------------------------------
+A{1}  = exp(P.A{1})*E(1);
+A{2}  = exp(P.A{2})*E(2);
+A{3}  = exp(P.A{3})*E(3);
+C     = exp(P.C);
+ 
+% intrinsic connectivity and parameters
+%--------------------------------------------------------------------------
+Te    = T(1)/1000*exp(P.T(:,1));         % excitatory time constants
+Ti    = T(2)/1000*exp(P.T(:,2));         % inhibitory time constants
+He    = H(1)*exp(P.H(:,1));              % excitatory receptor density
+Hi    = H(2)*exp(P.H(:,2));              % inhibitory receptor density
+
+ 
+% pre-synaptic inputs: s(V)
+%--------------------------------------------------------------------------
+R(1)  = R(1).*exp(P.S(1));
+R(2)  = R(2)  + P.S(2);
+S     = 1./(1 + exp(-R(1)*(x - R(2)))) - 1./(1 + exp(R(1)*R(2)));
+ 
+% exogenous input
+%==========================================================================
+U     = C*u(:);
+ 
+ 
+% State: f(x)
+%==========================================================================
+ 
+% Supra-granular layer (inhibitory interneurons): Voltage & depolarizing current
+%--------------------------------------------------------------------------
+f(:,7) = x(:,8);
+f(:,8) = (He.*((A{3} - A{2})*S(:,9) + G(:,3).*S(:,9)) - 2*x(:,8) - x(:,7)./Te)./Te;
+ 
+% Granular layer (spiny stellate cells): Voltage & depolarizing current
+%--------------------------------------------------------------------------
+f(:,1) = x(:,4);
+f(:,4) = (He.*((A{1} - A{3})*S(:,9) + G(:,1).*S(:,9) + U) - 2*x(:,4) - x(:,1)./Te)./Te;
+ 
+% Infra-granular layer (pyramidal cells): depolarizing current
+%--------------------------------------------------------------------------
+f(:,2) = x(:,5);
+f(:,5) = (He.*((A{3} - A{2})*S(:,9) + G(:,2).*S(:,1)) - 2*x(:,5) - x(:,2)./Te)./Te;
+ 
+% Infra-granular layer (pyramidal cells): hyperpolarizing current
+%--------------------------------------------------------------------------
+f(:,3) = x(:,6);
+f(:,6) = (Hi.*(G(:,4).*S(:,7)) - 2*x(:,6) - x(:,3)./Ti)./Ti;
+ 
+% Infra-granular layer (pyramidal cells): Voltage
+%--------------------------------------------------------------------------
+f(:,9) = x(:,5) - x(:,6);
+f      = spm_vec(f);
+ 
+if nargout == 1; return, end
+ 
+ 
+% delays
+%==========================================================================
+% Delay differential equations can be integrated efficiently (but
+% approximately) by absorbing the delay operator into the Jacobian
+%
+%    dx(t)/dt     = f(x(t - d))
+%                 = Q(d)f(x(t))
+%
+%    J(d)         = Q(d)df/dx
+%--------------------------------------------------------------------------
+m  = size(x,2);
+De = exp(P.D);
+Di = diag(diag(De));
+De = De - Di;
+De = De*D(2)/1000;
+Di = Di*D(1)/1000;
+De = kron(ones(m,m),De);
+Di = kron(ones(m,m) - speye(m,m),Di);
+D  = Di + De;
+ 
+% Implement: dx(t)/dt = f(x(t - d)) = inv(1 + D.*dfdx)*f(x(t))
+%                     = Q*f = Q*J*x(t)
+%--------------------------------------------------------------------------
+[Q,J] = spm_dcm_delay(M,P,D);
+
+
+return
+ 
+% notes and alpha function (kernels)
+%==========================================================================
+% x   = t*exp(k*t)
+% x'  = exp(k*t) + k*t*exp(k*t)
+%     = exp(k*t) + k*x
+% x'' = 2*k*exp(k*t) + k^2*t*exp(k*t)
+%     = 2*k*(x' - k*x) + k^2*x
+%     = 2*k*x' - k^2*x
diff --git a/toolbox/dcm_meeg/spm_gen_erp.m b/toolbox/dcm_meeg/spm_gen_erp.m
index 37b55b9..1d47b41 100644
--- a/toolbox/dcm_meeg/spm_gen_erp.m
+++ b/toolbox/dcm_meeg/spm_gen_erp.m
@@ -15,7 +15,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Karl Friston
-% $Id: spm_gen_erp.m 3653 2009-12-23 20:06:48Z karl $
+% $Id: spm_gen_erp.m 4281 2011-03-31 19:49:57Z karl $
 
 % within-trial inputs
 %==========================================================================
@@ -64,15 +64,18 @@
     %----------------------------------------------------------------------
     for i = 1:size(X,2)
 
-        Q.A{1}  = Q.A{1} + X(c,i)*P.B{i};         % forward   connections
-        Q.A{2}  = Q.A{2} + X(c,i)*P.B{i};         % backward  connections
-        Q.A{3}  = Q.A{3} + X(c,i)*P.B{i};         % lateral   connections
+        Q.A{1}  = Q.A{1} + X(c,i)*P.B{i};         % forward connections
+        Q.A{2}  = Q.A{2} + X(c,i)*P.B{i};         % backward connections
+        Q.A{3}  = Q.A{3} + X(c,i)*P.B{i};         % lateral connections
 
+        % intrinsic connections
+        %----------------------------------------------------------------------
         try
-            Q.H = Q.H + X(c,i)*diag(P.B{i});      % intrinsic connections
+            Q.H(:,1) = Q.H(:,1) + X(c,i)*diag(P.B{i});
         catch
-            Q.G = Q.G + X(c,i)*diag(P.B{i});
+            Q.G(:,1) = Q.G(:,1) + X(c,i)*diag(P.B{i});
         end
+        
     end
 
     % integrate DCM for this trial
diff --git a/toolbox/dcm_meeg/spm_nmm_priors.m b/toolbox/dcm_meeg/spm_nmm_priors.m
index 4386226..7aa55a8 100644
--- a/toolbox/dcm_meeg/spm_nmm_priors.m
+++ b/toolbox/dcm_meeg/spm_nmm_priors.m
@@ -31,7 +31,7 @@
 %    pE.D    - delays
 %    pE.X    - exogenous background activity
 %
-% pC - prior covariances: cov(spm_vec(pE))
+% pC - prior (co)variances
 %
 % Because priors are specified under log normal assumptions, most
 % parameters are simply scaling coefficients with a prior expectation
@@ -46,7 +46,7 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_nmm_priors.m 3653 2009-12-23 20:06:48Z karl $
+% $Id: spm_nmm_priors.m 4096 2010-10-22 19:40:34Z karl $
  
  
 % disable log zero warning
@@ -60,12 +60,12 @@
  
 % population variance
 %--------------------------------------------------------------------------
-pE.S   = 0;                 pC.S = 1/64;
+pE.S   = 0;                 pC.S = 1/16;
  
 % set intrinic [excitatory] time constants
 %--------------------------------------------------------------------------
-pE.T   = log(ones(n,1));    pC.T = ones(n,1)/32;    % excitatory constants
-pE.G   = log(ones(n,1));    pC.G = ones(n,1)/32;    % intrinsic connections
+pE.T   = log(ones(n,1));    pC.T = ones(n,1)/16;    % excitatory constants
+pE.G   = log(ones(n,1));    pC.G = ones(n,1)/16;    % intrinsic connections
  
 % set extrinsic connectivity
 %--------------------------------------------------------------------------
@@ -120,15 +120,10 @@
 
 % Exogenous background activity
 %--------------------------------------------------------------------------
-pE.X   = 0;              pC.X  = 1/64;
+pE.X   = 0;              pC.X  = 1/16;
 
 % Capacitance
 %--------------------------------------------------------------------------
-pE.CV  = 0;              pC.CV = 1/32;
+pE.CV  = 0;              pC.CV = 1/16;
 
 warning('on','MATLAB:log:logOfZero');
- 
-
-% covariance of neural parameters
-%==========================================================================
-pC    = diag(sparse(spm_vec(pC)));
diff --git a/toolbox/dcm_meeg/spm_ssr_priors.m b/toolbox/dcm_meeg/spm_ssr_priors.m
index c8ed6d0..0c513aa 100644
--- a/toolbox/dcm_meeg/spm_ssr_priors.m
+++ b/toolbox/dcm_meeg/spm_ssr_priors.m
@@ -15,7 +15,7 @@
 %
 %--------------------------------------------------------------------------
 %
-% pC - prior covariances: cov(spm_vec(pE))
+% pC - prior (co)variances
 %
 % Because priors are specified under log normal assumptions, most
 % parameters are simply scaling coefficients with a prior expectation
@@ -30,17 +30,27 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
  
 % Karl Friston
-% $Id: spm_ssr_priors.m 3517 2009-10-29 15:11:56Z guillaume $
+% $Id: spm_ssr_priors.m 4281 2011-03-31 19:49:57Z karl $
  
 
-% add prior on endogenous inputs (neuronal) and noise
+% number of LFP channels
 %--------------------------------------------------------------------------
-pE.a  = 0;               pV.a = 1/16;              % amplitude input AR
-pE.b  = 0;               pV.b = 0;                 % amplitude input IID
-pE.c  = [0 0];           pV.c = [1/16 1/16];       % amplitude noise AR
-pE.d  = [0 0];           pV.d = [1/16 1/16];       % amplitude noise IID
+if size(pE.L,1) == 1, n = size(pE.L,2); else, n = 1; end
+if size(pE.C,1),      m = size(pE.C,2); else, m = 1; end
 
-% and augment prior covariance
+% add prior on spectral density of innovations (pink and white coeficients)
 %--------------------------------------------------------------------------
-pC    = spm_cat(spm_diag({pC, diag(spm_vec(pV))}));
+pE.a = sparse(2,m); pC.a = sparse(2,m) + 1/32; % neuronal innovations
+pE.b = sparse(2,1); pC.b = sparse(2,1) + 1/32; % channel noise non-specific
+pE.c = sparse(2,n); pC.c = sparse(2,n) + 1/32; % channel noise specific
+
+
+% neuronal innovations
+%--------------------------------------------------------------------------
+pE.d = sparse(8,m); pC.d = sparse(8,m) + 1/16; 
+
+
+
+
+
  
diff --git a/toolbox/dcm_meeg/spm_x_cmc.m b/toolbox/dcm_meeg/spm_x_cmc.m
new file mode 100644
index 0000000..3b61d75
--- /dev/null
+++ b/toolbox/dcm_meeg/spm_x_cmc.m
@@ -0,0 +1,21 @@
+function [x] = spm_x_cmc(P)
+% returns the initial state of a canonical microcircuit model
+% FORMAT [x] = spm_x_cmc(P)
+% P - parameters
+%
+% x        - x(0)
+%__________________________________________________________________________
+% David O, Friston KJ (2003) A neural mass model for MEG/EEG: coupling and
+% neuronal dynamics. NeuroImage 20: 1743-1755
+%__________________________________________________________________________
+% Copyright (C) 2011 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_x_cmc.m 4232 2011-03-07 21:01:16Z karl $
+ 
+% array of states
+%--------------------------------------------------------------------------
+n  = length(P.A{1});                          % number of sources
+m  = 8;                                       % number of states
+x  = sparse(n,m);
+
diff --git a/toolbox/mixture/spm_mix_plot1d.m b/toolbox/mixture/spm_mix_plot1d.m
index 4c7c0a6..5c004a7 100644
--- a/toolbox/mixture/spm_mix_plot1d.m
+++ b/toolbox/mixture/spm_mix_plot1d.m
@@ -10,7 +10,7 @@
 % Copyright (C) 2010 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny & Ged Ridgway
-% $Id: spm_mix_plot1d.m 3997 2010-07-15 12:38:24Z ged $
+% $Id: spm_mix_plot1d.m 4195 2011-02-05 18:39:13Z ged $
 
 if nargin < 4 || isempty(nPoints), nPoints = 100; end
 if nargin < 3 || isempty(rng)
@@ -46,6 +46,7 @@
 legend(legs, 'Location', 'BestOutside');
 
 if ~isempty(data)
+    xl = xlim;
     subplot(2,1,2)
     [c b] = hist(data, round(sqrt(numel(data))));
     bar(b, c / sum(c) / mean(diff(b)), 'k');
@@ -55,5 +56,6 @@
         plot(x, f, '--', 'Color', mixcol);
         legend('Histogram', 'Parzen PDF', 'Location', 'BestOutside');
     end
+    xlim(xl);
     subplot(2,1,1); % (so any subsequent title command goes over this plot)
 end
diff --git a/toolbox/spectral/spm_csd2ccf.m b/toolbox/spectral/spm_csd2ccf.m
new file mode 100644
index 0000000..022f254
--- /dev/null
+++ b/toolbox/spectral/spm_csd2ccf.m
@@ -0,0 +1,46 @@
+function [ccf,pst] = spm_csd2ccf(csd,Hz)
+% Converts cross spectral density to cross covariance function
+% FORMAT [ccf,pst] = spm_csd2ccf(csd,Hz)
+%
+% csd  (Hz,:,:)         - cross spectral density (cf, mar.P)
+% Hz   (n x 1)          - vector of frequencies (Hertz)
+%
+% ccf                   - cross covariance functions
+% pst  (n + n + 1 x 1)  - vector of lags for evaluation (seconds)
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_csd2ccf.m 4095 2010-10-22 19:37:51Z karl $
+ 
+% unpack cells
+%--------------------------------------------------------------------------
+if iscell(csd)
+    for i = 1:length(csd)
+       [ccfi,pst] = spm_csd2ccf(csd{i},Hz);
+       ccf{i}     = ccfi;
+    end
+    return
+end
+ 
+% Nyquist
+%--------------------------------------------------------------------------
+Ny    = 256;
+g     = zeros(Ny,1);
+Hz    = fix(Hz);
+ 
+% Fourier transform cross-spectral density
+%==========================================================================
+for i = 1:size(csd,2)
+    for j = 1:size(csd,3)
+        g(Hz)      = csd(:,i,j);
+        f          = ifft([0; g; flipud(conj(g))]);
+        ccf(:,i,j) = real(fftshift(f));
+    end
+end
+ 
+ 
+% Compute time bins if necessary
+%--------------------------------------------------------------------------
+N   = Ny;
+pst = (-N:N)/N/2;
diff --git a/toolbox/spectral/spm_csd2coh.m b/toolbox/spectral/spm_csd2coh.m
new file mode 100644
index 0000000..3768da0
--- /dev/null
+++ b/toolbox/spectral/spm_csd2coh.m
@@ -0,0 +1,41 @@
+function [coh,fsd] = spm_csd2coh(csd,Hz)
+% Converts cross spectral density to coherence and (phase) delay
+% FORMAT [coh,fsd] = spm_csd2coh(csd,Hz)
+%
+% csd  (Hz,:,:) - cross spectral density (cf, mar.P)
+% Hz   (n x 1)  - vector of frequencies
+%
+% coh           - coherence
+% fsd           - frequency specific delay (seconds) 
+%               - phase-delay/radial frequnecy
+%__________________________________________________________________________
+% Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
+ 
+% Karl Friston
+% $Id: spm_csd2coh.m 4095 2010-10-22 19:37:51Z karl $
+ 
+
+% unpack cells
+%--------------------------------------------------------------------------
+if iscell(csd)
+    for i = 1:length(csd)
+       [cohi,fsdi] = spm_csd2coh(csd{i},Hz);
+       coh{i}     = cohi;
+       fsd{i}     = fsdi;
+    end
+    return
+end
+
+% preliminaries
+%--------------------------------------------------------------------------
+Hz    = spm_vec(Hz);
+ 
+% compute coherence and delay
+%==========================================================================
+for i = 1:size(csd,2)
+    for j = 1:size(csd,2)
+        coh(:,i,j) = abs(csd(:,i,j).*conj(csd(:,i,j)))./abs(csd(:,i,i).*csd(:,j,j));
+      % coh(:,i,j) = abs(csd(:,i,j));
+        fsd(:,i,j) = unwrap(angle(csd(:,i,j)))./(2*pi*Hz);
+    end
+end
diff --git a/toolbox/spectral/spm_mar.m b/toolbox/spectral/spm_mar.m
index 8867867..44a185d 100644
--- a/toolbox/spectral/spm_mar.m
+++ b/toolbox/spectral/spm_mar.m
@@ -22,9 +22,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny 
-% $Id: spm_mar.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_mar.m 4096 2010-10-22 19:40:34Z karl $
 
-if nargin < 4 | isempty(verbose)
+if nargin < 4 || isempty(verbose)
     verbose=0;
 end
 
@@ -60,7 +60,7 @@
 y=X([p+1:1:N],:);
 
 k=p*d*d;
-if nargin < 3 | isempty(prior)
+if nargin < 3 || isempty(prior)
   prior.type='global';
   prior.groups=1;
   prior.group(1).index=ones(1,k);
@@ -149,7 +149,7 @@
     else
         change=norm(w(:)-old_w(:))/k;
         if verbose
-            disp(sprintf('Iteration %d Delta_w=%1.6f',it,change));
+            fprintf('Iteration %d Delta_w=%1.6f',it,change);
         end
         if change < tol
             break;
@@ -180,7 +180,7 @@
 acc=-0.5*N*log(det(B));
 f_m=acc-kl_weights-kl_alpha+lga;
 if verbose
-    disp(sprintf('Iteration %d Acc=%1.3f KL_w=%1.3f KL_alpha=%1.3f LogGena=%1.3f Fm=%1.3f',it,acc,kl_weights,kl_alpha,lga,f_m));
+    fprintf('Iteration %d Acc=%1.3f KL_w=%1.3f KL_alpha=%1.3f LogGena=%1.3f Fm=%1.3f',it,acc,kl_weights,kl_alpha,lga,f_m);
 end
 
 % Load up returning data structure
diff --git a/toolbox/spectral/spm_mar_spectra.m b/toolbox/spectral/spm_mar_spectra.m
index 063029e..6eeb7a2 100644
--- a/toolbox/spectral/spm_mar_spectra.m
+++ b/toolbox/spectral/spm_mar_spectra.m
@@ -38,9 +38,9 @@
 % Copyright (C) 2008 Wellcome Trust Centre for Neuroimaging
 
 % Will Penny 
-% $Id: spm_mar_spectra.m 1143 2008-02-07 19:33:33Z spm $
+% $Id: spm_mar_spectra.m 4096 2010-10-22 19:40:34Z karl $
 
-if nargin < 4  | isempty(show)
+if nargin < 4  || isempty(show)
     show=0;
 end
 
@@ -56,7 +56,7 @@
 for ff=1:Nf,
   af_tmp=eye(d);
   for k=1:p,
-    af_tmp=af_tmp+mar.lag(k).a*exp(-i*w(ff)*k);
+    af_tmp=af_tmp+mar.lag(k).a*exp(-1i*w(ff)*k);
   end
   iaf_tmp=inv(af_tmp);
   H(ff,:,:)=iaf_tmp;  % transfer function
diff --git a/toolbox/spectral/spm_mmtspec.m b/toolbox/spectral/spm_mmtspec.m
index 9ad2c2c..d63fcc1 100644
--- a/toolbox/spectral/spm_mmtspec.m
+++ b/toolbox/spectral/spm_mmtspec.m
@@ -19,15 +19,13 @@
 % Copyright (C) 2009 Wellcome Trust Centre for Neuroimaging
 
 % Partha Mitra, Ken Harris and Will Penny
-% $Id: spm_mmtspec.m 3994 2010-07-13 15:53:30Z guillaume $
+% $Id: spm_mmtspec.m 4021 2010-07-28 12:43:16Z vladimir $
 
 nChannels = size(x, 2);
 nSamples = size(x,1);
 
 if (nargin<4 || isempty(timeres)) 
-    C=16;
-else
-    C=round(nSamples/(Fs*timeres));
+    timeres = nSamples/(16*Fs);
 end
 
 if (nargin<5 || isempty(timestep))
@@ -47,7 +45,7 @@
 df=freqs(2)-freqs(1);
 nFFT=round(Fs/df);
 
-WinLength=round(nSamples/C);
+WinLength=round(Fs*timeres);
 nOverlap=ceil(percent_overlap*WinLength); 
 
 nTapers = 2*NW -1;