The experimental groups consisted of 16 convicted juvenile rapists classified according to those who victims were older than themselves (35 to 90 years old) and those younger than themselves (4 to 9 years old). The control group, selected from juveniles evaluated by the same juvenile court system as the experimental groups or from juveniles currently treated at various mental health settings for nonsexually assaultive behavior, was matched to the experimental groups for age, IQ, and socioeconomic level. Those who sexually assaulted younger victims differed from the other experimental group and the control group. They manifested greater social isolation, physical problems, sexual identification problems, and a lower frequency of aggressive behavior before age 6. Those who raped older victims had social characteristics superficially similar to control subjects. It might be useful to view the rape of older victims as one of various violent, impulsive acts committed by juveniles with a low level of ego integration. Both experimental groups had significantly higher numbers of violent and abusive parents and more sexual identification problems than the control group. Implications are drawn for treatment. 1 table and 9 references.