Hello Elizabeth:

For an accurate estimate of the HRF, you should use an event-related design.
With a block design, the estimated HRF will be very noisy. However, the fit
to the data will still be good (as you mentioned). Moreover, the AUC is still
physically meaningful: The sum of the IRF coefficients represents the steady-
state amplitude of the response to a step input. That is, if the input is
in the "on" state for a long time (such as with a long block design), then the
fitted response approaches the sum of the IRF coefficients (plus baseline).
You can verify this for yourself, using the Deconvolution plugin. Therefore,
the AUC can be used as a measure of the magnitude of the response, whether
you are using an event-related design or a block design. (You may wish to
convert this to % change relative to baseline, using 3dcalc).

You can sum up the IRF coefficients using the -glt option of 3dDeconvolve.
Just set up each GLT matrix as a single row of 0's and 1's, with one entry
for each estimated parameter. Place a 1 in each of the (4) positions
corresponding to that stimulus, and 0's elsewhere. Do this for each of
the 3 stimuli. For more about the -glt option, see Sections 1.2.11 and 1.4.4
of the documentation in file 3dDeconvolve.ps.

Doug Ward