Criminal opportunity theories identify four basic constructs that are central to models of victimization: proximity, exposure, victim attractiveness, and guardianship. Proximity was controlled in this study, allowing for more detailed examination of the other three constructs. The study used discrete-time event history models to estimate the contemporary-time predictors of personal victimization and the time-lagged effect of prior victimization from a five-wave longitudinal study of street youth. The study found strong support for exposure hypotheses with increased victimization risk associated with time on the streets, substance abuse and affiliation with deviant peers. Selected indicators of victim attractiveness and guardianship hypotheses were significant but, overall, the support for these two dimensions was more limited. Prior victimization was associated with approximately a 25o percent increase in the odds of subsequent victimization. The study suggests several interpretations of this effect and it argues for additional situation-specific research to provide further elaboration of victimization theories. Tables, references