Biased social information processing has been solidly linked to aggressive behavior in children. Additionally, attributional tendencies, the tendency to overattribute deliberately hostile intentions to others, have been shown to distinguish aggressive and nonaggressive male youth. In order to reduce aggressive males' tendency to attribute hostile intentions to peers following ambiguous, negative interactions, an attribution retraining program was implemented in an urban, public elementary school. Subjects, 101 aggressive and nonaggressive African American boys in grades three through five, were randomly assigned to the attributional intervention, to an attention training program, or to a control (no treatment) group. Subjects' reaction to hypothetical peer provocation, teacher ratings of subjects' aggressive behavior, and referrals for formal disciplinary action were assessed and evaluated for statistical and clinical significance. Compared to subjects in the control or attention training groups, aggressive subjects in the attributional intervention group showed significant reduction in bias to attribute hostile intent. These subjects were also rated by their teachers as less reactively aggressive following treatment, and less likely to receive disciplinary action. Nonaggressive subjects experienced no negative effects due to program participation. (ET)