The thesis underlying this paper is that maltreatment of children is fundamentally a socio-cultural problem, rather than a personal one. The use of an ecological model of human development is advocated to provide insight into the dynamics of child maltreatment, with emphasis on organism-environment interaction, social policy and second-order effects. Hypotheses basic to a program of research on primary and secondary prevention are outlined. Two actual projects within such a program are described. The goal of the first is to develop a multivariate model of the correlates of child maltreatment. The model is to be used to screen neighborhoods for intervention on the basis of the discrepancy between actual and predicted incidence. The second project uses a face-to-face interview procedure to explore the "family biographies" of three contrasting groups: those families reported to Child Protective Services for suspected child maltreatment; those known to be experiencing social stress but known to provide excellent child care; and those who are identified as part of the social network of the former two groups. A preliminary clinical application being considered involves using the interview to locate the "natural reinforcers" in the family's environment. It is hoped that increasing the adequacy of feedback, modeling and reinforcement for good child care through the use of social networks will provide an effective complement to conventional family intervention. (Author/BF)