Thanks for the response.

About the sub-brick index, the problem exists when I have only 2 functional data sets.
The index is not confusing me because I know all the volumes shown for sub-brick
indexes that are out of range are actually the last sub-brick, but it is a little annoying
when I try to see how good the sub-bricks are registered.

About the number of simuli, the real situation is usually complicated and some
simplification is always needed. What I am thinking is that the more stimuli we can
identify, the more accurate we can fit the response. (But as pointed out by Bob, then
we will need more time points in the functional scan to have enough degree of freedom
for parameter estimation.) Am I correct?

Another question, for 3dDeconvol, is it better to use boxcar as the stimuli rather than
to use waver beforehand if we don't know what parameter to use with waver?

Thanks
george