We hypothesized that anger repressors would show discrepancies between self-reported anger and cardiovascular and behavioral responses only during harassment. Subjects (N=102) were assigned randomly to condition. In the nonharassment condition, subjects told stories about eight Thematic Apperception Test cards without any harassment. In the harassment condition, subjects told four stories without harassment, and then told four more stories with harassment. Words connoting aggressive behavior and angry/hostile affect were coded from story content. Subjects were classified into low anger expressor, anger repressor, high anger expressor, and defensive anger expressor categories based on median splits of the Anger-Out Subscale and Marlowe-Crowne Social Desirability Scale. Results showed that harassed anger repressors reported anger comparable to that of low anger expressors but less than high expressors, whereas their heart rate (HR) reactivity was comparable to high expressors, but greater than low anger expressors. Increases in anger words did not distinguish repressors from other groups. Repressed anger may represent a distinct anger management style characterized by a discrepancy between acknowledged anger and cardiovascular reactivity--effects that become fully manifest only during interpersonal provocation.