Violence is a direct product of personal behavior conditioned in many ways by environmental factors. Like most modern health problems, it has multiple causes and is not simple to cure. It is susceptible to definition and measurement, however. It can be analyzed, as can a disease, in terms of victim, agent, and circumstances. Based on these assessments, relative risks can be estimated, populations at special hazard defined, and common factors of environment or relationship identified. Once all of this has been done, it is reasonable to expect that intervention strategies can be designed to alter the circumstances most likely to produce violent behavior, to reduce risks, and ultimately to reduce unnecessary and preventable death and disability caused by violent behavior. The process applied to the problem of violence in general and homicide in particular can be divided into two stages. The first is surveillance, which involves collaboration in comparing homicide data held by the police and vital statistics agencies. The second stage is the epidemiologic assessment of precursors to violence. This requires large population data that relate economic status, employment, isolation, crowding, illness in the family, drug or alcohol abuse, or lifestyle to the outcomes of homicide and suicide.