But if the pure RT is put in there, this would seem to pose a problem. The RT regressor is meant to sop up the variance induced by the neurons firing for longer in condition A than in condition B (if A's RT > B's). But, if the RT is 500ms in A and 550ms in B, we know full well that the neurons in any one region were not firing for 500ms in A and 550ms in B. In truth, in some brain region, it might be 50ms and 100ms respectively. Thus, the scaling of activity resulting from the RT difference is far greater than would be estimated by passing in 500 and 550ms, as this voxels "trial length" is, in fact, twice as long in B than in A not 10% longer.
Craig