A sample of 69 families was treated with Parent Training Therapy (PTT). Pre- and post-treatment measures were available for observed child deviancy; ratings of child externalizing behavior by mothers, fathers and teachers; and measures of parenting practices. Baseline, termination, and change scores were calculated for each variable and used to predict 2-year follow-up data for police arrest and out-of-home placement (OHP). When controlling for baseline ratings, none of the termination ratings by teachers, mothers, or fathers made significant contributions. The findings are consistent with the hypothesis that adult participant raters are insensitive to change. On the other hand, termination scores for observed deviancy, parental monitoring, and family problem solving significantly predicted long-term outcome, even after controlling for baseline levels. In addition, the termination measure of parental discipline predicted future OHP after controlling for baseline level. The findings are consistent with the hypothesis that observer-based estimates of child deviancy and careful measures of changes in parenting practices are valid measures of treatment outcome whereas ratings by participant adults are of limited utility.