My TR is 3.0 seconds. There are 47 slices of 60 ms each, leaving an acquisition gap of 200 ms. The SPM manual states direly that the slice timing correction will be removed and ought not be used. I would very much like to follow this advice, but my statistical results appear to be significantly better with the slice timing correction (sometimes a difference between no result without and some activation with it -- and since this is a multimodal study, the location of activation is confirmed by other modalities). That brings about a number of questions: 1- If I don't use slice timing correction, what can I do beyond using hrf+derivatives? Add other hrfs to the code, peaking at different times? What do people do in practice? 2- If I do use slice timing correction, then... a) Looking at the code, one can see that it is indeed deprecated. What is my TA? Intuitively I guess it is 2.8 ms, and not what is suggested, i.e. TA = TR*(1-1/47). But the code then defines parameters timing(1) and timing(2), and to get sensible values for these I would have to backward-calculate TA: the code is written as if the difference between TA and TR is an additional slice. Anyhow that's just a few milliseconds difference here and there, and it shouldn't matter anyway. b) is it OK to still use hrf+derivatives when using slice timing correction, or should use hrf only? (for aliasing or other reasons) I also seem to note that many people are still using the slice timing correction... so there appears to be a disconnect between the SPM Manual recommendation, and what is done in practice. Thanks, Philippe