For example, within the literature on child abuse through 1978, only 20 percent of studies can be considered empirically sound, and over three-fourths rely on official case records or aggregate statistics. Only 3 percent focused on incidence or prevalence, and only 11 percent examined psychosocial correlates. Further, only a small proportion of research on violence addresses some variation in family violence. Most research attention has been on criminal violence without differentiating the critical features that sets family violence apart from other forms. Within family violence research, underdeveloped theory and imprecise nominal and operational definitions have compromised the comparability, generalizability, and reliability of results. In addition such research has been compromised by methodological problems related to data sources, samples, research design, validity and reliability of measures, response effects, and differences in the levels of analysis. Directions for future research, particularly validations research, are discussed.