Another suggestion: a colleague of mine takes the first correct response and treats it as incorrect. You should also be able to define regressors with a negative onset, e.g. -50 s. Even if it's just -1 s this event seems not to be convolved with the HRF though, meaning it does not appear in the design matrix. But check yourself, especially if this negative onset is the only onset for that condition. And the timecourse might be different after high-pass filtering.
In general, however, I think it is nicer not to go with any arbitray dummy regressor at all. If the subject does not make any mistakes, well, then that's it.
|