Dear Kasia,

First thank you for your script!

I am not exactly sure how CONN generates the p-FDR maps at 1st-level, I tried with my own scripts and even p-uncorrected threshold is too high to keep any significant voxel. I am probably missing something, as my script works correctly on 2nd-level results. You can try it here:

https://github.com/lrq3000/neuro_experim...

About p-FDR maps generated by CONN, I think they are already thresholded at 0.05, so it is probably unnecessary to threshold them any more. So you can probably take them as-is.

Also if you want to show the average p-FDR corrected 1st-level map of 2 or more seeds, you can probably just calculate with SPM imcalc the min-T of both seeds' p-FDR maps, as according to Nichols et al the min-T conjunction of two voxel-wise corrected maps is equivalent to the voxel-wise correction of the conjunction of unthresholded maps.