Hi AFNI experts,

I just had a puzzling result for a few subjects when using 3dDeconvolve and wanted to check my understanding.

I have a breath hold task for which I have 3 conditions - breathing normally, cued breathing, and breath hold. In my 3dDeconvolve command I specified 3 contrasts, one for comparing the hold to each breathing condition, and one to compare it to both breathing conditions. The afni output correctly specified the matrices like this:
1 -1 -1
1 -1 0
1 0 -1

these were specified like:
-gltsym 'SYM: hold -instructed -normal' -glt_label 1 hold_gt_breathing
-gltsym 'SYM: hold -instructed' -glt_label 2 hold_gt_instructed
-gltsym 'SYM: hold -normal' -glt_label 3 hold_gt_normal


Somehow it appears that in many voxels, the sign is flipped for the third contrast as opposed to the first two (ie the coefficients for 1 and 2 will be positive, and 3 is negative). Does this make sense and could someone explain under what conditions this would happen if so?

Thank you!
James