afni.nimh.nih.gov

I assume that you want to model Decision and Outcome. You may try different HDR basis functions on one or a few subjects, and see how they compare to each other. My limited experience shows that the model-free approach (TENT or CSPLIN in 3dDeconvolve) achieves the highest detection power at the group level; in addition, you would be able to characterize the detailed HDR shape for each condition.

Gang