Frequently, the media inform us about violent episodes that happen inside and around schools. How is that violence related to students performance? In this paper, I show that, on average, students who attended more violent schools had less proficiency on the SAEB 2003, a test carried out by the Brazilian Ministry of Education, even when I controlled for school, class, teacher and student characteristics. This relation is stronger in the bottom of the proficiency distribution. Furthermore, I find out that school violence increases the probability of teacher turnover among schools, indicating that violence also has an indirect effect on the students proficiency.