We evaluated whether alcohol outcome expectancies moderate the association between measures of anxiety and alcohol use. Student subjects completed questionnaires related to their level of anxiety, recent alcohol-use patterns, and outcome expectancies for alcohol to be tension reducing. Interviews were used to determine the presence or absence of alcohol dependence in subjects and in their first- and second-degree relatives. Consistent with predictions, male subjects with high tension-reduction alcohol outcome expectancies showed a stronger positive correlation between measures of anxiety and drinking behavior than did male subjects with low tension-reduction outcome expectancies. However, this effect was not found for female subjects. We note past studies showing similar gender effects, and relate the overall study findings to the tension-reduction hypothesis of stress-induced drinking.