Decisions about the occurrence of child abuse are increasingly difficult to make because concepts of what qualifies as reportable child abuse may be broadening. We examined this question by comparing 51 fatal child abuse cases occurring in Georgia between July 1975 and December 1979 to non-fatal cases and to the Georgia population. Overall rates of fatal child abuse were higher for male perpetrators compared with female and black perpetrators compared with white. However, the latter finding varied with economic and geographic status. The highest child abuse fatality rates were found in poor, rural, white families (3.3/100,000 children) and in poor, urban, black families (2.4/100,000 children). Risk factors for fatal abuse included early childhood (RR 6:1), parental teenage childbearing (RR 4:1), and low socioeconomic status. These characteristics were similar to those of the severe child abuse cases noted in the early child abuse literature. Non-fatal cases did not clearly share these risk factors. Severe abuse, here represented by fatal cases, is a distinct subset of reported child abuse, but characteristics associated with it are frequently attributed to all reportable child abuse. Medical personnel should be aware that they cannot rely on the presence or absence of these characteristics in screening for risk of reportable child abuse. Child abuse research should use restricted, stated case definitions. When intervention and prevention programs are being organized, they should not generalize research findings to all forms of child abuse.