We report results from a group-randomized trial of a prevention program aimed at preventing bullying and other aggressive behaviors. Fourth grade classrooms at 28 public elementary schools were assigned to receive selected modules of the Youth Matters prevention curriculum or to a no-treatment control condition. Cross-classified multilevel models were fitted to four waves of data collected over 2 years to test the effect of the intervention on self-reported bullying and bully victimization. No systematic change in bullying other students was observed. In a continuous outcome growth model, bully victim scale scores declined over the course of the study and the rate of decline in victimization was significantly higher in experimental schools relative to control schools. But the results from binary outcome growth models indicate no significant treatment effects on bully status or bully victim status over time. Implications of findings for the implementation of anti-bullying strategies in urban public school settings are discussed.