ABSTRACT. Objective: The current longitudinal study was designed to consider the time-varying effects of men's heavy episodic drinking (HED) and drinking setting attendance on college sexual assault perpetration. Method: Freshman men (N = 992) were recruited in their first semester and completed online measures at the end of their first five semesters. Using multilevel models, we examined whether men with higher frequency HED (or party or bar attendance) were more likely to perpetrate sexual assault (between-person, Level 2 effect) and whether sexual assault perpetration was more likely in semesters in which HED (or party or bar attendance) was higher than each individual's average (within-person, Level 1 effect). Results: The between-person effect of HED on sexual assault was not significant after accounting for the between-person effects of antisocial behavior, impersonal sex orientation, and low self-control. The within-person effect of HED on sexual assault perpetration was not significant. However, models substituting frequency of party attendance or bar attendance revealed both between-and withinperson effects. The odds of sexual assault were increased for men with higher bar and party attendance than the sample as a whole, and in semesters in which party or bar attendance was higher than their own average. Supplemental analyses suggested that these drinking setting effects were explained by hookups, with sexual assault perpetration more likely in semesters in which the number of hookups exceeded one's own average. Conclusions: Findings point toward the importance of drinking contexts, rather than drinking per se, as predictors of college men's sexual assault perpetration. (J. Stud. Alcohol Drugs, 78, 5-13, 2017)