Previous research has emphasized the importance of heterogeneity in offense trajectories. Using data from the Oregon Youth Study (OYS), a longitudinal study of at-risk boys interviewed annually from ages 9 to 10 years to ages 23 to 24 years, this study examined childhood and adolescent covariates of observed offending trajectory classes. Six trajectory classes were identified using the latent growth mixture modeling approach: chronic high-level, chronic low-level, decreasing high-level, decreasing low-level, rare, and nonoffenders. Multinomial logistic regressions revealed that very rare offenders and chronic high-level offenders were distinguished by individual, family, and peer factors measured in childhood and adolescence. Only adolescent covariates (deviant peers, various problem behaviors) distinguished among trajectories that engaged in substantive amounts of offending behavior. Overall, there was more specificity than commonality in correlates of distinctive offending trajectories.