Social skills training (SST) has been shown to produce significant increases in skills immediately after training. However, persistence of skills has not been demonstrated during the follow-up period. The present study investigated this issue using SST and cognitive restructuring (CR) methods. Twenty-four hospitalized "alcoholics" were allocated to one of the following three group: (i) SST, (ii) CR, and (iii) traditional supportive therapy (TST). Twelve 2-hour sessions were conducted with each group. Before, following, and three months after treatment (i) behaviour ratings through structured interview, (ii) nurses' ratings, (iii) self-reported ratings of social skills and (iv) alcohol intake were recorded. Analysis of variance indicated SST to be significantly superior on all measures compared to CR, which was superior to TST immediately after training. However, 3 months after treatment, the subjects in the CR group indicated greater skill increments than the SST group subjects who, in turn, showed greater improvement than subjects in the TST group. These results led to the conclusion that CR was more effective than SST and TST in producing long-term skill increments. The significance of these findings in providing support for the social learning model of alcohol dependence and cognitive restructuring theory of social skills modification is discussed.