Based on the glt matrix, it seems that you run 3dDeconvolve with maximum lag of 2 (e.g., -stim_minlag k 0 -stim_maxlag k 2) instead of 3, correct?

Assume the hemodynamic response coefficients corresponding to lags of 0, 1, and 2 are h_i(0), h_i(1) and h_i(2) respectively, here i = 1, 2, 3 indicates three different stimuli.

(1) 0 0 1 1 1 1 1 1 0 0 0 would test whether [h_1(0) + h_1(1) + h_1(2)] + [h_2(0) + h_2(1) + h_2(2)] is significantly different from 0. In other words, it would tell whether the area under the curve (AUC) for both stimuli of 1 and 2 combined would be negligible.

By the same token, if you want to compare the difference between stimulus 1 and stimulus 2, the corresponding vector would be

0 0 1 1 1 -1 -1 -1 0 0 0

(2) With a glt matrix of

0 0 1 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 1 0 0 0

you are testing

H_0: h_1(0) = h_1(1) = h_1(2) = h_2(0) = h_2(1) = h_2(2) = 0

versus

H_1: not all the above 6 numbers are 0, or at least one number is not 0

Gang