I've tried both relatively small and large size of smoothing kernel (i.e., 4mm vs. 10mm) on the surface EPI data using mri_surf2surf function (e.g., mri_surf2surf --hemi lh --s fsaverage_sym --fwhm )

Attached are displays of raw EPI, 4mm, and 10 mm FWHM. 

As can be seen, the areas being smoothed by 10mm FWHM is quite off from those in the raw 2D EPI whereas 4mm result looks reasonable. 

If anyone has some intuition about why this happened, please help me. 

-Glen 

 

ps) the area with colors show vertices with non-zeros values (also it was reconstructed from slap EPI)