Hello Arika:

The -maxlag option is to allow for modeling of the signal delays due to the
hemodynamic response (IRF). This, in principle, is independent of the length
of the stimulus presentation. The system response is modeled as the convolution
of the input stimulus function with the IRF. Therefore, the modeled response
can be of arbitrary length, even though the IRF itself is short. It's probably
better to use the same -maxlag value for each of the conditions, unless there
is some reason for doing otherwise.

If the input stimulus function represents the extent of the stimulus duration,
i.e., 1's at each time point during stimulus presentation, then the modeled
response will be of similar (though somewhat longer) duration. You did not
specify the TR, but suppose TR = 2 sec. And suppose maxlag = 4 for each of the
conditions. Then the movie duration would be represented by 6 consecutive 1's.
The modeled response could be of duration (6+4)*2 = 20 seconds, even though
the estimated IRF is only of duration (4+1)*2 = 10 seconds.

As always, use the Deconvolution plugin to visually verify that the model is
adequate.

Doug Ward