This paper examines links between violent behavior, type and severity of psychopathology, substance abuse comorbidity, and community mental health treatment, using matched data from two surveys: the National Institute of Mental Health Epidemiologic Catchment Area project and the Triangle Mental Health Survey (a North Carolina study of adults with severe and persistent mental illness). Multivariate logistic regression analysis was used to model the risk of violent acts attributable to three domains of independent variables: sociodemographic characteristics, clinical diagnoses and symptomatology, and mental health services utilization. Findings include: (1) Symptom severity was significantly greater in the clinically-selected sample than in the community survey of respondents with comparable diagnoses who self-reported using mental health services; (2) Violence risk was related to psychoticism/agitation in a curvilinear form; (3) In a multivariable model, violence was significantly associated with substance abuse comorbidity, particular psychotic symptoms (perceived threat and loss of internal cognitive controls), and absence of recent contact with a community mental health provider; (4) The relationship between lack of treatment and higher odds of violence was less pronounced among respondents with substance abuse comorbidity; (5) When clinical and services-use variables were taken into account, sociodemographic predictors were not significantly related to violence.