Widely varying estimates of the incidence of child abuse or spouse abuse are a direct result of differences in research definitions and methodology. Standardization of terms should be a top priority for family-violence research. Research should aim to identify and better understand the combination of social and individual factors that lead some persons and not others to behave violently. No research is likely to be any more definitive. Since no single therapeutic technique or service can "cure" family violence, research should use a multifacted approach in assessing treatment efforts. Research on the operational functioning of programs is essential. Special demonstration programs often do more harm than good, since they promise a quick cure when long-term services are required, and they are often a costly duplication of pre-existing services. Because of political and other factors as well as research shortcomings, research often does not impact policy as much as researchers would like. An improved research product could assist in persuading policymakers to incorporate research in their decisionmaking.