Female and male subjects read one of three scenarios depicting a social-sexual interaction between a man and a woman. The scenarios varied in severity of the stimulus male's behavior according to whether sexual harassment had occurred. Subjects then rated the offensiveness of his behavior and whether it constituted sexual harassment. Subjects also rated the stimulus male's behavior on Weiner's (1986) three basic dimensions of causality: locus of causality, or the perceived cause of the behavior as internal or external; stability, or the likelihood of its recurrence; and controllability, or the amount of volitional influence he exercised during the interaction. Additionally, subjects reported their anger at the target male and sympathy for the target female. Findings generally indicated that for both women and men, as the scenarios increased in severity, these ratings were elevated. Gender comparisons also showed elevations in these ratings for female but not for male subjects, and only in the scenarios of lesser severity. Regression analyses revealed attribution variables—control, stability, and anger—as reliable predictors of perceived offensiveness and harassment in the most severe scenario, whereas affects were the most reliable predictors in scenarios of lesser severity. The implications of these findings for perception of sexual harassment are discussed.