Past research has suggested that the aggregation of deviant peers during treatment may cause harmful effects (T. J. Dishion, J. McCord, & F. Poulin, 1999). This study compared the effectiveness of problem-solving skills training groups in which all members had conduct problems ("pure" group condition) with groups that consisted of adolescents with and without behavior problems ("mixed" group condition). Participants were 139 sixth- and seventh-graders (mean age = 12.7; 63% male; 55% White) enrolled at public middle schools. Pre-, post-, and 6-month follow-up data were collected, along with intervention process variables. Results showed that, contrary to the deviancy training hypothesis, adolescents in the pure-group condition engaged in more adaptive in-session behavior and received lower scores on the parent and teacher ratings of externalizing behavior at postintervention than those in the mixed-group condition. Mediation analyses showed that the deviancy training that occurred in the mixed-group condition accounted for their worse postintervention scores, findings consistent with the deviancy training hypothesis.