Gender differences in variables relating to alcohol treatment are receiving increased attention. Most reports indicate gender differences in both intrapersonal, interpersonal and environmental trans-situational variables that relate specifically to women's drinking. This study investigates gender differences in the immediate relapse situation itself. These differences are then examined in the context of trans-situational (i.e. across time) affective and cognitive variables and social support. This report is part of a larger study designed to replicate Marlatt's taxonomy of high risk situations for relapse, and to examine alternative methods of describing relapse situations. Three hundred subjects (169 men, 131 women) were recruited from six different inpatient and outpatient programs. Subjects were asked to identify their last relapse before entering the index treatment. Details of the drinking episode itself were also obtained, e.g. number of drinking days in episode, drinking to intoxication, presence of others. Measures of state and trait affect, expectancies, and social support were also collected. There were no gender differences in Marlatt's taxonomy, or in drinking topography during relapse. At baseline, women reported drinking less, and drinking less often than men. However, women drank to intoxication more often than men. Surprisingly, men relapsed alone more often than did women. Men and women report relapsing frequently with same sex friends, and women showed a tendency to relapse in the presence of romantic partners. Men also tend to report more positive mood states during relapse than women, which is congruent with the finding that women report more depression. Severity of drinking pattern was found to be an important confounding variable in these analyses. It is suggested that gender differences found in individuals in treatment for alcohol problems may reflect societal differences between men and women, and that taking these differences into account in designing treatment programs may improve outcomes.