Abstract
The effect of a series of stimulus changes, including attendant present or absent and agent of token reinforcement in a sorting task, on peer interaction among retarded women was examined. Conversation was not differentially directed toward the agent of reinforcement when that agent was a peer, nor did a peer agent increase the overall rate of peer interaction. Peer interaction did increase when no attendant was present, and during the final phase of the experiment, the high rate of peer interaction was maintained while the attendant was faded back into the situation in a series of discrete steps. The procedure was found to be efficient in terms of staff time and did not adversely affect task performance.