Theory and research using a social-information processing framework indicate that reward-focused (proactive) aggression has different social consequences than defense-focused (reactive) aggression. Students use norms that identify expected and socially approved behaviors as guides to their own actions. Differences in social-cognitive processing characteristics and social status linked to each type of aggression may increase the relevance of some normative sources relative to others. This study fills a gap in the literature by examining the contributions of personal beliefs, classroom beliefs, and classroom rates of aggression to future proactive and reactive aggression. During fall and spring, we observed students' aggression on school playgrounds using a random subsample (n = 254) of consented students from 35 classrooms (Grades 3-6). We calculated classroom rates of proactive and reactive aggression from fall observations. Classroom means for beliefs endorsing retaliation were calculated from surveys of 536 students. Results of multilevel analyses revealed, as hypothesized, that personal beliefs predicted high rates of students' proactive aggression, but not reactive aggression. Classroom beliefs predicted high rates of students' reactive but not proactive aggression. Students in classrooms with high rates of fall proactive aggression showed low spring rates of both types of aggression. In contrast, students in classrooms with high rates of fall reactive aggression displayed high spring rates of proactive and reactive aggression. The latter pattern may represent classrooms in which students continue to struggle against status inequities. The discussion examines how inequities may impact intervention efforts.