Thanks, Rick. One more thing, though. Is the RETROICOR step you refer to only generating the regressors or actually applying the regression as well? If the latter, I wonder why it would be advisable to apply regression twice (once for physio, then again including the motion parameters from 3dvolreg)? I would have assumed a single regression step would be ideal. Is the assumption that physio and motion parameters are orthogonal? Or do we assume noise is noise, and it doesn't really matter to which noise regressors it is assigned?