is multiple comparison correction relevant for searchlights

[bpinsard]

Hi, this is not really an "issue".

I am interested by "crossnobis" searchlight technique that I have seen applied in few paper but I am concerned about multiple multivariate testing (same voxel used in multiple searchlights).
Having let say one distance map (one pair distance or averages of pairs distance) per subject and performing a 1-sample t-test vs. 0 , is it necessary to perform multiple comparison correction or the cross-validated unbiased distance is enough?
One option would be to run some sort of fsl randomize for non-parametric testing: repeated sign permutations of distance with 1-sample t-test to create voxelwise t-distribution for thresholding followed by cluster size correction.
lmnwyt

Thanks