A meta-analytic review was conducted of 79 controlled studies of children's social skills training. The overall short-term effectiveness of social skills training was seen as moderate. Multiple regression analysis was used to partition the variance in effect size. Diagnostic description of the subjects contributed heavily to the prediction; withdrawn subjects respond better than aggressive youngsters and children not diagnosed as behaviorally atypical. Modeling and coaching techniques are associated with higher effect sizes than social-cognitive procedures or multitreatment packages, but training technique did not emerge as a significant predictor of outcome in the multiple regression analysis. Outcome varied considerably as a function of variable. Effect sizes were higher for indices of social competence (observed social interaction, peer- and self-report, social-cognitive skill) than for more peripheral variables (e.g., global self-concept, academic achievement). Effectiveness at follow-up, though still substantial, was lower than posttreatment indicators. There were few studies with follow-up measurement at posttreatment intervals longer than 3 months.
Read full abstract