Thanks for your quick answer

afni -ver:
Version AFNI_2011_12_21_1014
[[Precompiled binary macosx_10.7_Intel_64: Sep 30 2013]]

Yes I used "-oob_value 0"
In fact I followed the Afni_tuto_data6 (with the subject FT) apart that I added the standarization of the surfaces.

Why I see that ther is some mis-registration:
1: the task of the subjects was just to make saccade. So I know that there is some visual activity and broadly in the parietal and the frontal (FEF). In fact I was seeing these activations before the mis-registration...
2: I opened at the same time a first afni viewer and suma with the glm result made in surface domain AND a second viewer of afni with the result of the same glm but conducted in volume domain (completely separate analyses). Every windows were linked : activations were not at the same place. They were correct in the volume space (visual, parietal, frontal) but not in the surface space (suma).

Actually the mis-resgistration is maybe not here for all my subjects (from my last checking 2 were wrong and one was good, I have 12 subjects in total). But for the one with the brain of which I tried to see the result ofthe ttest (see my first message), I'm sure there is a misregistration now...