Okay, so it sounds like when you do a 3dANOVA only considering the voxels in a mask, only the cluster threshold should change? Correct? I had been hearing the opposite. In this analysis, I am looking at connectivity differences in different drug conditions within a mask (all within subject). Here's the script, it's in bash

3dANOVA3 -DAFNI_FLOATIZE=YES -type 4 \
-alevels 2 \
-blevels 2 \
-clevels 25 \
-dset .... #### 100 images here
-mask $maskimg \
-fa FSalKet \
-fb FDay12 \
-fab FAB \
-amean 1 SalMean \
-amean 2 KetMean \
-bmean 1 plbMean \
-bmean 2 guaMean \
-abmean 1 1 SalPlbMean \
-abmean 1 2 SalGuaMean \
-abmean 2 1 KetPlbMean \
-abmean 2 2 KetGuaMean \
-adiff 1 2 SalvsKet \
-bdiff 1 2 PlbvsGua \
-aBdiff 1 2 : 1 salPlbvsKetPlb \
-aBdiff 1 2 : 2 SalGuavsKetGua \
-Abdiff 1 : 1 2 SalPlbvsSalGua \
-Abdiff 2 : 1 2 KetPlbvsKetGua \
-bucket $tdir/anova2x2_${seed}_cope_z_${msk}