Well, actually you just have take the block length and add about 30 s, this would be a reasonable window length then. Assume the canonical HRF and a brief stimulus, then it takes about 30 s till baseline has reached again. The same holds for longer durations.
If you want to check and look at the expected time course, specify a model with the canonical HRF as basis funtion, click on "Review", select the SPM.mat, go on Design in the lower left window, then Explore, Session 1, select the condition of interest. The upper left plot shows you the expected overall time course for that particular condition (note, this ALWAYS corresponds to the unfiltered time course, so no high-pass filter applied). If trials are relatively close together, it might be difficult to detect when the expected time course reaches baseline again though. Then simply set up an arbitrary model with a single event/block and the corresponding block length.
|