Hmmm.
Well, you answered some of my questions and raised some new ones.

First, I should tell you a bit more about my experimental design. I have 10 subjects who each do 3 different tasks within each scan (block design). In my deconvolution, I have each of the 3 tasks plus the 6 motion parameters as regressors (no lags yet). I also have a glt for each pair-wise combination of tasks. When I did 3dttest, I was intersted first in just creating one group map for each of the 3 comparisons, so I used the regression coefficient for condition 1 as -set2 in 3dtest, and either 0 or the baseline map as -set1 (both gave wierd maps).

So, based on your answer, my new questions are:
1. My understanding of 3dANOVA was that it was for inter-group analyses, but I believe you are suggesting that I use it instead of 3dttest for intra-group analysis, is that correct? (Stats has never been my strong suit, so please excuse my ignorance.)

2. You suggested using condition 1 as -set1 and condition 2 as -set2 in order to see the difference between these 2 conditions. How is that different than using the results of the glt for these 2 conditions? (is one better than the other?).

3. You suggest using 3dcalc to create a conjunction map. Although the resulting mask will show the regions of activation within the group, how do I then determine group statistics for the ROIs? (p-values and percent change?). Is this the only way to create group maps for each condition?

4. You suggest "creating a glt to calculate area under the IRF for each regressor". How do I do this? (0 0 1 0 0 0 0 0 0 0 0 ?) How is this different than the regression coefficient that is automatically generated for each regressor?

Thanks!!
Sara