Three studies were conducted to explore the relationship between social skills instruction and the maintenance and generalization of skills to post-school environments. Three adult women with severe disabilities in their last year of school participated in the studies. The first study addressed the effectiveness of an instructional package to improve social skills and to increase generalization and maintenance of the skills. Skills that had been observed to elicit reinforcement in environments that the students were expected to use following graduation were selected for instruction. Students improved their performance of target skills in the instructional setting. Maintenance and generalization to unstructured social environments were variable. The second study addressed the effectiveness of a "loose" training procedure to facilitate generalization to natural environments. The procedure was effective in improving student performance and in generalization to non-training environments; however, training contingencies were needed to maintain generalization. Interviews were conducted with teachers, employers, co-workers, and acquaintances regarding the social validity of the procedures and the results. The final study addressed variables that may affect social interaction. Interview and observational data were analyzed for content. Results indicated that performance in social interactions may be affected by the perceptions of the interaction partner, the type of the interaction, the settings where the interactions occur, and the relationship between the interactors. Appendices contain observation and interview forms, case examples, coding procedures, and other administrative materials. (Approximately 100 references) (Author/PB)