Hi Patrick, Then, given you're not using them for proper inference, it probably doesn't matter too much - it's just that the separation in two hemispheres is an artefact of the preprocessing pipeline (eg you would rarely do that with volumetric data). To combine the surface meshes, you can do something like this: M = gifti({'L.surf.gii','R.surf.gii'}); M = spm_mesh_join(M); save(gifti(M),'mesh.surf.gii') and for the functional data it would be something like: gl = gifti('L.func.gii'); gr = gifti('R.func.gii'); for i=1:size(gr.cdata,2) gg = gifti([gl.cdata(:,i);gr.cdata(:,i)]); gg.private.metadata = struct('name','SurfaceID','value','mesh.surf.gii'); save(gg,spm_file('mesh.func.gii','suffix',sprintf('_%05d',i)),'ExternalFileBinary'); end As with the previous snippet, I haven't tried it so it might require adjustments... Best regards, Guillaume. On 13/10/17 15:39, Patrick Sadil wrote: > Hello Guillaume, > I see. We had only been planning on using these data as a really rough localizer, grabbing voxels that were active as determined by an uncorrected p-value threshold. But, it sounds like even that would be improved by analyzing the entire hemisphere together. Sorry, to keep asking questions about this, but do you know of a reliable way to combine these hemispheres? > > Best, > Patrick > -- Guillaume Flandin, PhD Wellcome Trust Centre for Neuroimaging University College London 12 Queen Square London WC1N 3BG