+ Letting HRF shape varies allows for subject and regional variability in
hemodynamics
+ Can test HRF estimate for different shapes; e.g., are later time points more
ÒactiveÓ than earlier?
Ð Need to estimate more parameters for each stimulus class than a
fixed-shape model (e.g., 4-15 vs. 1 parameter=amplitude of HRF)
Ð Which means you need more data to get the same statistical power
(assuming that the fixed-shape model you would
otherwise use was in fact ÒcorrectÓ)
Ð Freedom to get any shape in HRF results can give weird shapes that are
difficult to interpret