Direct observational assessment of parent–child interaction is important in clinical intervention with conduct-problem children, but is costly and resource-intensive. We examined the reliability and validity of a brief measure of parents' relational schemas (RSs) regarding their child. Children (aged 4 to 11 years) and their families receiving treatment at a clinic for externalizing behavior problems ( n = 150) or mood/developmental disorders ( n = 28) were assessed using a multi-method, multi-informant procedure. RSs were coded from Five-Minute Speech Samples (FMSS) using the Family Affective Attitude Rating Scale (FAARS), and were compared with directly observed parent–child interaction and questionnaire measures of family and parental dysfunction and conduct problems. Mothers' and fathers' RS scales were internally consistent and could be reliably coded in under 10 min. Less positive RSs and more negative RSs were associated with higher rates of child conduct problems, and were more characteristic of the speech samples of parents of children with externalizing disorders, compared with clinic control parents. RSs demonstrated some associations with parenting behavior and measures of family functioning and symptoms of parental psychopathology, and predicted conduct problems independently of observed parental criticism. The results demonstrate the reliability and validity of the FAARS assessment of parental RSs in clinic-referred families. This brief measure of parent–child dynamics appears well-suited to ‘real-world’ (i.e., community) clinical settings in which intensive methods of observation are often not feasible.