Because research shows a close association between offending and victimization, recent work has argued that theories that account for crime should explain victimization as well. The current study uses a new approach to examine the extent of the overlap between offenders who commit violent crime and victims of violence to determine whether it is worthwhile to pursue separate theories to account for these phenomena. Specifically, we take the statistical approach that Osgood and Schreck (2007) developed for analyzing specialization in violent versus property offending and apply it to analyzing tendencies to gravitate toward violent offending versus victimization. In doing so, we treat the differentiation into victim and offender roles as an individual-level latent variable while controlling for confounding between the likelihood that individuals will take either role in violent acts and their overall numbers of encounters with violence (as either offender or victim). Our purpose is to examine 1) whether significant differentiation can be observed between the tendency to be an offender versus the tendency to be a victim, 2) whether any such differential tendency is stable over time, and 3) if it is possible to predict whether individuals will tend toward violent offending versus victimization. Using two waves of data from the National Longitudinal Study of Adolescent Health to explore these objectives, we find significant and stable levels of differentiation between offenders and victims. Moreover, this differentiation is predictable with explanatory variables.