Well, yes, as always, it is more complicated than I let on… I plan to concatenate across 4 scans, and, from a previous post, by Doug Ward (2-27-03), he suggested that it might be easier to “normalize” the individual runs, concatenate, and then run 3dDeconvolve- instead of using the 4 separate baseline regressors that would come out of 3dDeconvolve to normalize the betas of each scan separately.
I realize that you can do it either way, but one of the main reasons I wanted to go ahead and “normalize” (i.e. convert to % signal change) across the scan, was because I want to do some single trial averaging to look at the actual shape of the response to compare them across scans and subjects - and normalizing (and detrending) the scan time series seemed to me to be the most appropriate way to obtain “good” single-trial averages for comparison. So, since I had already done that, I just used these percent-signal-change time series as input into 3dDeconvolve. That is an "ok" way to do this, right?

So, back to my question… Given that the time series are in percent signal change when entered into 3dDeconvolve, and I’ve used the –peak 1 option to produce the gamma wave IRF model, but the actual peak = 1.604 (b/c of the block-type design of the study); do the beta’s need to be converted (divided by 1.604) to be true reflections of % signal change, or are the beta values = % signal change without conversion?