Aggression, bullying and violence in children and youth are prevalent in Canada (18%) and internationally. The authors evaluated the effectiveness of Roots of Empathy (ROE), a school-based mental health promotion and violence prevention program for children that has been widely implemented but rarely evaluated. Eight school divisions were randomly assigned to either a treatment group that received ROE in 2002-2003 (445 students) or a wait-list control group (315 students). These were compared on three child mental health outcomes (physical aggression, indirect aggression and pro-social behaviour), rated by teachers and students (self-rated). The three wait-list school divisions received ROE in 2003-2004 (new cohort of 265 students) and were compared with the control group from 2002-2003 on the three outcomes, for replication purposes. For both comparisons, the authors report multi-level modelling analyses regarding (1) immediate effects after ROE completion at the end of the school year (pretest to post-test) and (2) long-term ROE effects up to three years after post-test. ROE had replicated, beneficial effects on all teacher-rated outcomes, which were generally maintained or further improved across follow-up. However, ROE had almost no statistically significant or replicated effects on student-rated outcomes. This is the first evaluation to suggest that ROE appears effective when implemented on a large scale under real-world delivery conditions.