afni.nimh.nih.gov

Hello Lukas:

I don't know of an easy way to do that with the current software. Perhaps
the best alternative would be to calculate the contrast within subject
using 3dDeconvolve, and then perform a t-test across subjects. This may
even be a preferable approach, since more of the calculation is performed
within-subject, prior to the across-subject analysis.

Doug Ward