In the context of a multivariate specification (estimated using the TOBIT procedure), crowding was by far the most influential variable in the predictor stock. Of the four assault types examined, three are positively related to a crowding index, and all crowding-assault relations are nonlinear. When controlling empirically for crowding level, institutional size, staff-inmate ratio, perpentage of staff who are correctional officers, rehabilitative program participation rates and program type, inmate turnover rates, inmate demographics, criminal histories, and unique institutional influences, age was implicated in only one of the four types of assault rates. Measures assumed to be indicators of the deprivation and importation models of prisonization indicate that, at an aggregate level, both models are implicated but are clearly not as important in determining assault rates as is crowding. Applications of these results with regard to prison standards and prison capacities are discussed. (Author abstract)