An initial group of 28 mother-child dyads from the general population of Valencia, Spain, participated in the study. The initial sample was used as a participant pool from which mother- child dyads were selected to form high-risk and low-risk groups for child physical abuse. The risk groups were formed by selecting the top 35 percent and the bottom 35 percent of the mothers in terms of their risk for child physical abuse as defined by their scores on a Spanish version of the Child Abuse Potential Inventory. In the final sample, 10 high-risk and 10 demographically similar low-risk mother-child dyads were studied. Trained observers coded maternal-child interaction patterns in the home during five 1-hour periods, using the Standardized Observation Codes system. As expected, high-risk mothers made fewer neutral approaches to their children, displayed more negative behaviors toward their children, and made more indiscriminant responses to their children's prosocial behavior. Expected risk-group differences were not found in the number of neutral instructions or positive responses, albeit the proportion of positive responses out of the total number of positive and negative responses was higher for low-risk mothers. After control for educational differences, risk-group differences remained in the rates of neutral approaches and the number of indiscriminant behaviors made in response to children's prosocial behaviors. The observational data show that high-risk mothers displayed some behaviors similar to those observed in physically abusive mothers. The finding that high-risk mothers made more indiscriminate or noncontingent responses when reacting to their children's prosocial behavior is consistent with a coercive model of child physical abuse. 2 tables and 39 references