The study was conducted with a group of 23 mother-child dyads that presented episodes of physical abuse toward the child and a control group of 24 mother-child dyads that did not present any interactional problems. The dyads were observed at home by a group of trained observers using the Standardized Observation Code, Revision III. Three stepwise discriminant function analyses were conducted for variables related to maternal behavior, child behavior, and mother-child interaction. Variables specifically included neutral social approach, neutral instruction, positive response, and aversive response for mothers and interactive prosocial behavior, noninteractive prosocial behavior, interactive deviant behavior, and noninteractive deviant behavior for children. Two variables measured mother-child interactions, appropriate maternal response to child prosocial behavior and inappropriate maternal response to child deviant behavior. Findings indicated that behaviors with a negative valence better classified subjects into groups to which they belonged and that a better classification resulted when variables related to interaction were used in the analysis. More specifically, findings demonstrated that important aspects of maternal behavior were significantly related to abusive or nonabusive status. Along the same lines, child behavior was also associated with group status and the amount of inappropriate maternal behavior directed toward children showed an important association with group status. 47 references and 6 tables