In the early 1980's the National Institute of Mental Health funded a series of parallel surveys in New Haven, Baltimore, Saint Louis, Durham, and Los Angeles for the primary purpose of determining the prevalence of untreated psychiatric disorders in community populations. The data from these surveys also provide the first large-sample estimates of the general prevalence of assaultive behavior among persons with and without diagnosable mental disorders in the community. In the first wave of the Epidemiological Catchment Area (ECA) project, university-based research teams conducted structured diagnostic interviews with approximately 3,000 to 5,000 household-resident adults at each of the five sites, resulting in a combined total of 17,803 community respondents in the pooled database. Additional interviews were conducted with 1,379 residents of nursing homes, prisons, and psychiatric hospitals. The findings presented in this chapter result from recent secondary analysis of the ECA community data for a purpose not intended or anticipated by the original investigators. A limited amount of information on violence was obtained fortuitously in the course of interviews designed to assess the presence of symptom criteria for psychiatric disorders. The diagnostic criteria for two disorders in particular, antisocial personality disorder and alcohol abuse, included violent behavior. In this chapter the initial descriptive analysis and comparisons between alternative indices of violence are based on data from Durham and Los Angeles only, with a pooled sample size of approximately 7,000. The findings show that having a major mental disorder even without substance abuse was associated with a significant increase in the prevalence of violence. Those with both mental disorder and substance abuse were more prone to violence than either group with a single diagnosis. The pattern suggests additive rather than synergistic effects. Implications of the study and its findings for research are discussed. 11 tables, 5 figures, and 16 references