Hi Douglas,
I am getting somewhat confused in switching from fs3 to fs4. I pre-prosessed the same (functional) session with freesurfer 3.0.5 (mc-sess, inorm-sess, spatialsmoothing-sess, mkanalysis-sess.new (3.0.5), selxavg-sess, mkcontrasts-sess, stx-grinfer-sess) and freesurfer 4.5.0 (mc-sess, inorm-sess, spatialsmoothing-sess, mkanalysis-sess (with the -fsv3 switch to be sure they do the same whitening)), unfortunately the results for the exact same contrast look strikingly different. If I extract a ROI from the fs3 analysis (averaged volume) and apply it to the fs4 averaged volume the extracted time-course (from all the face patches looks almost identical, with 4 nice dips for the face epochs), so the pre processing seems to be almost identical for fs3 and fs4, but the fs4 analysis does not show any face patch at all (see attached image). Note how even the hot-colored object activation is weaker in the upper row (fs4) compared to the lower row (fs3) (both with identical threshold and saturation values). So something is going on here and I have no clue. I used MION contrast agent so I expect signal extinction for activated epochs and it is non-human primates, but I assume that does not make a big difference (and I accounted for the different shape in the IRF in both analyses). Now the mkcontrast-sess I used invocation is identical between the two Fs-versions. mkanalysis-sess does differ somewhat: FS4: mkanalysis-sess -fsd bold -funcstem fmcs2 -analysis a1 -paradigm pardigm.para -dt blocked -sf sf -df df -force -inorm -timewindow 80 -polyfit 2 -motioncor -TR 3.000 -gammafit 0 8 -gammaexp 0.3 -nconditions 7 -runlistfile rlf1 -mask brain_fmc -taumax 40 -noautostimdur -acfbins 10 -fsv3 -mcextreg
FS3: mkanalysis-sess.new -fsd bold -funcstem fmcs2 -analysis a1 -paradigm pardigm.para -dt blocked -sf sf -df df -force -inorm -polyfit 2 -motioncor -TR 3.000 -gammafit 0 8 -gammaexp 0.3 -nconditions 7 -runlistfile rlf1 -mask brain_fmc -taumax 40 -mcextreg
the paradigm file is identical, I tried "-acfbins 10 -fix-acf" instead of "-fsv3" before with the identical unexpected low p result, I did not yet try no whitening as the fs3 analysis used whitening successfully.
Do you have any idea what I must have done wrong to get there? Can it be the -timewindow 80 in the new analysis compared to the traditional 40 in the old analysis?
Thanks for your insight Sebastian Moeller