The overall multilevel models for physical victimization revealed very few significant predictors at any level of analysis. However, property victimization seemed to be better explained with a more contextual model including prisoner, prison, and interaction variables. In summary, what emerged was distinct dynamics of victimization that needed to be studied separately as specific domains. This might require prison officials to create separate classification tools for identifying an inmate's victimization risk. The majority of the existing literature on inmate victimization considers only one level of analysis, which results in ignoring the interaction effects between inmate- and prison-level variables. To fill the gap in the literature and to explore the possible effects of other prison-level variables, multilevel modeling techniques were used to analyze self-report data from more than 1,000 inmates and 30 prisons in Kentucky, Tennessee, and Ohio. Tables, notes, and references