We tested group differences in neural response using second-level random-effects factorial models in SPM. We examined functional connectivity using interaction (PPI) analyses in SPM8, with the bilateral nucleus accumbens, defined by the PickAtlas anatomical region, as the seed region. The psychological variable for PPI was win > loss, which serves to isolate response to reward from response to feedback generally. Because the contrast for PPI involved blocks for which BOLD response would not overlap, our task timing was appropriate for PPI analyses. Sex, race, and daily smoking quantity were included as covariates in analyses for group differences. We investigated brain-behavior analyses focusing on drinking behavior by conducting regressions in SPM within the alcohol dependent group. These regression analyses included age as a covariate. There were no evident sex differences in drinking behaviors, and accordingly we did not include sex as a covariate in these regression analyses. All analyses initially used a voxel-wise threshold of p<.05 and a minimum extent of 10 contiguous voxels. We then used simulations in AlphaSim based on the five-ROI mask to compute minimum cluster sizes in
