I read from the freesurfer discussion archives that when combining FSL FEAT with freesurfer, one should definitely NOT do spatial smoothing in the preprocessing when preparing to examine cortical data.

However I can see that in the FSFast pipeline, 2D and 3D smoothing *is* included in the FSFast preprocessing.

So:

1. I am guessing that the 2D smoothing applies only to the surface and is smoothing on the surface space. Is this correct?
2. I'm also guessing that the 3D smoothing applies only to the subcortical regions. Is all this also correct?
3. The freesurfer.feat.ppt guide does NOT contain any instructions for doing any smoothing on data derived from FSL. But I assume that, if smoothing is done during FSFast preprocessing but is not done when preprocessing in FSL, then smoothing will still need to be done at a subsequent point with the FSL data. But I don't see instructions on that in the Freesurfer FEAT documentation on freesurfer.feat.ppt. If I do FSL preprocessing+1st-level analysis with no spatial smoothing, and then proceed with the analysis, at which point do I apply preprocessing? Is there a guide specifically describing how to apply smoothing to data originating from FSL?