The research tested the deprivation, management, and "not-so-total" institution models of the causes of prison violence. Results suggested that the deprivation model was least useful in distinguishing ordered from disordered institutions. In addition to the absence of a crowding effect, inmate living conditions did not affect prison order. Moreover, results indicated that prisons are not total institutions, sealed off and unaffected by the circumstances and conditions existing in host communities. Findings suggested that poor prison management is a predictor of rates of assault toward inmates and correctional personnel. However, the likelihood of prison riots is largely independent of structural, managerial, and environmental factors. The data also suggest that inmate educational, vocational, and industrial programs are a useful management tool; such programs should offer not just structure but also opportunities for self-improvement. Tables and 60 references (Author abstract modified)