This study used meta-analysis, a form of quantitative synthesis, to extract from the current body of knowledge prospective estimates of the strength of relationship of different precursors to later violent behavior. Observed, predictor-adjusted, and methods-adjusted mean correlations were developed for all predictors with serious violence measured between the ages of 15 and 25. The largest predictive correlations generally fell within the family of antisocial behavior predictors. A rank-order listing of the predictor variables for serious violence based on the methods-adjusted estimates showed that, while substance use was a particularly strong predictor between the ages of 6 and 12, for those 12 to 15 years of age the predictive ability of this construct was negligible. Those aged 6-12 with antisocial friends were slightly less inclined to become violent later while those aged 12-15 with such friendships were much more likely than their peers to become violent later. The corrections for those aged 12-15 are generally higher than those aged 6-12, showing that prediction improved as the time of prediction was closer to the outcome period. Prior delinquency was a strong predictor of serious violent behavior for both age periods while having parental maltreatment was not a particularly strong predictor of serious violence between the ages of 15 and 25. The synthesis of data from nearly 10,000 subjects in 34 studies suggests that serious violent behavior between the ages of 15 and 25 is predictable. No all predictors performed equally well. Performance of a risk factor at one point in the life span was not necessarily indicative of its performance at another point in the life span. When thinking critically about risk factors or using risk factors to select cases for intervention, considering age in addition to a person's status on risk factors is essential. 3 tables, 2 footnotes