This study explored the relationship between levels of aggression between fall of first grade and spring of fifth grade and the likelihood of having a violent court record; and whether this relationship is moderated by lunch status and race, which are used as proxy indicators of social class and markers of increased likelihood for retention in the juvenile justice system. Also, the study examined the utility of aggressive behavior as a screening tool for later violence. It was hypothesized that the screening quality would increase as the ratings became more proximal to the violent events. A community epidemiological sample of 415 public school boys was rated at 6 time points during elementary school regarding their level of aggressive/disruptive behavior. Violence was measured using juvenile police and court records. The results showed that the risk of a later violent arrest increased as levels of aggression during the elementary school years increased. An important finding was that the association peaked in third grade, indicating an important transition point after which the level of variation in the development of aggression decreased. In respect to identifying at-risk boys, three intervention scenarios were compared. When minimizing false positives, the optimal test was found in the fall of first grade, with 83 percent of the high-risk compared to 22 percent of the low-risk boys later having a violent arrest record. When minimizing false negatives, the optimal test was found in the spring of fourth grade (30 percent of the high-risk boys versus 0 percent of the low-risk boys). Focusing on both false positives and false negatives, the optimal test was found in the spring of third grade (52 percent of the high-risk boys versus 14 percent of the low-risk boys). It was concluded that early levels of aggressive behavior are strong and robust predictors of later violence but are of limited utility in the early identification of boys at risk. 1 figure, 4 tables, 42 references