Seventy-six young, male offenders, attending a Community Home School, were randomly assigned to social skills training (SST) incorporating training in basic and more complex interpersonal skills, attention placebo (APC) and no treatment control (NTC) groups. Within subject changes in specific social skills were monitored in the SST group using a multiple baseline design. The results showed that SST led to definite improvements in some but not all the basic skills, and that these improvements were maintained at 3-month follow-up. In a between groups design SST was found to be significantly superior to APC and NTC groups on the performance of basic skills at posttest. Although the SST group reported significantly less social problems on a social problems questionnaire after training, a similar, though lesser reduction was found for the APC group and the NTC group. On a staff questionnaire of social problems, and independent ratings of social skills, friendliness, anxiety and employability. social workers' ratings of work, school and family relationships, self-reported offending and police convictions, there was no evidence that SST was differentially more effective than the comparison groups. The implications of these findings are discussed in terms both of current research into SST and the treatment of delinquency.