> I read that the stim_maxlag defines the desired length of the impulse
> response function. The first six regressors are experimental conditions
> (combination of two factors: Factor A = 3 levels, Factor B = 2 levels).
First of all, option stim_maxlag is really obsolete now since there is a better approach to handling response function modeling. Secondly, I'm still not so sure what kind of data in fmrigroup1.txt: stimulus timing coded in real time (for example, seconds) or 0's and 1's, OR real regressors? stim_maxlag would be appropriate only if you have stimulus timing in 0's and 1's inside fmrigroup1.txt and you want to run deconvolution analysis.
> If I got you right, I should not specify a stim_maxlag for the motion
> parameters and the seed and the interaction regressor, right? And why not for
> the motion parameters?
No, there is no point adding -stim_maxlag for these regressors. You may want to read the 3dDeconvolve manual for better understanding about the option. However, I would rather suggest you abandon using this option in your future usage of 3dDeconvolve.
Gang