Forty male subjects were randomly assigned to one of two conditions designed to induce either positive (Incentive Gain Condition) or negative (Incentive Loss Condition) affective states. This procedure was validated by an analysis of pre and post manipulation scores on the Multiple Affect Adjective Check List. Following the incentive manipulation subjects participated in an ad-lib drinking situation (Alcohol Taste Rating Rask) believing that the experiment was concerned with the relationship between intelligence, taste sensitivity and alcohol preferences. Contrary to expectation the analysis revealed that more alcohol was consumed by subjects in the incentive gain condition, this effect being most pronounced in heavy social drinkers in the incentive loss and gain conditions. A second analysis demonstrated that significantly less alcohol was consumed by subjects who, prior to ad lib drinking, evinced greater degrees of depressive affect and anxiety. The discussion focuses on the implications of these findings for tension reduction models of alcoholism.