Four 15–17 yr olds attending vocational classes at a residential treatment program for behaviorally disordered youth participated in a social skills training program. Intervention consisting of verbal training and role playing resulted in rapid acquisition of appropriate responses to a supervisor's instructions. However, there was no generalization of the behavior change beyond the intervention site. The addition of a 2nd intervention, self-monitoring, resulted in rapid generalization. The use of a multiple-baseline design across Ss suggested that the changed behavior stemmed from the experimental interventions. Additional data from 1 S suggest that the effects of the interventions were not sequentially dependent. Verbal training and role playing seemed to be a primary contributor to response acquisition, while self-monitoring appeared to facilitate response generalization. It is suggested that an ecological understanding of the natural environment in which behavior change is desired is critical for the development of an effective technology of response generalization.