The 42 women, who were charged in 15 States, were compared with 205 abused women who had not killed or attempted to kill their batterers. A semistructured questionnaire was used to obtain data on both the homicide and the nonhomicide groups. The questionnaire included a section on the woman's family of origin, her relationship with her parents, and any previous physical or sexual abuse she may have experienced. General questions were asked about the background of the abusive mate and about typical interactions in her relationship with him. The women were asked to give narrative accounts of four battering incidents. A discriminant function analysis identified variables that best predicted membership in the homicide group: severity of the woman's injuries, the man's drug use and frequency of intoxication, the frequency with which abusive incidents occurred, forced or threatened sexual acts by the man, the woman's suicide threats, and the man's threat to kill. Most of the women in the homicide group had no previous history of violence. In the abusive relationship, however, the women's attempts to survive with an increasingly violent and unpredictable mate eventually resulted in an act of violence on their part as well. 64 references.