Subjects for this research were recruited from the child protective services caseloads while the control group was selected from clinics and responses to media appeals. All families had children under 4 months to 2 1/2 years old. About half the mothers had not graduated from high school, and only 5 percent were college graduates. Because of recruitment difficulties, only 7 percent of the sample was black. The following variables were tested by home interviews with parents and children, questionnaires, and videotaped behavioral observation and physiological responses: parents' background, personality, child-rearing attitudes, social networks, antecedents to early attachment, ways of handling irritating child behaviors, and parent-child interactions. Subjects were paid $20 when they completed the testing. Both bivariate and multivariate data analysis procedures were used, including zero order correlations, factor analysis, path analysis, and discriminant analysis. The findings revealed sharp differences between the two groups. The abusive and neglectful parents were abused as children, were low in empathy, had few close friends, had been separated from their children, and exhibited sadistic, intolerant, and strict disciplinarian attitudes toward child rearing. They did not communicate with their children and scored low in facilitating behavior in the videotaped parent-child interactions tests. Real differences between parents who inflicted abuse and those who neglected their children were also found. For example, neglect was strongly correlated with problems in pregnancy or early infancy and disagreements over child rearing within the family. Abused children were less ready to learn than were neglected children. Abusive parents maintained higher heart rates and had less heart variability. When both high heart rates and low variability in heart rate were present, subjects' electrodermal skin response took longer to return to normal. Using the same indicators, the ability to accurately predict abusers in the sample was 88 percent. Further research will use a more representative sample and introduce additional variables. Figures, a bibliography of 14 references, and tables are provided. (ERIC author abstract modified)