Abstract A retrospective review was performed on the disciplinary records of 1,659 convicted murderers who had been admitted to the Texas Department of Criminal Justice between February 2001 and November 2003. Institutional violence was disaggregated by type of infraction, inmate or staff victim, extent of injury, and weapon usage to determine baseline prevalence and rates. Characteristics of the perpetrators, including younger age, more serious murder conviction, and longer sentence were associated with a higher incidence of prison assaults. Logistic regression analysis of a restricted sample of 1,440 male, non-death row inmates resulted in modestly predictive models for potentially violent acts (AUC = .668), assaults (AUC = .700), and assaults resulting in serious injury (AUC = .750). The findings suggest that the choice of measure used when operationalizating prison violence is crucial in determining baselines, but may have a more limited effect on identifying correlates and predicting outcomes.