How does the hemodynamic response change with stimulus length?
In very nonlinear ways. See Glover (1999) in PercentSignalChangePapers (and Logothetis & Wandell, 2004, HrfPapers). You’re on very shaky ground if you attempt to model stimuli longer than 6 sec by convolving a standard HRF with a boxcar. You may be better off using stimuli separated by longer times or shorter stimuli.