This study was addressed to the problem of applying behavior modification techniques on a group basis to a class of retarded students with "attentional deficits". Seven boys, age 8 to 15 yr, characterized as showing severe "attentional" problems or disruptive behavior in their respective classrooms, participated daily for 30-min sessions in a special class over a 1.5-month period. In each session, verbal instructions were given to the class as a whole. In control sessions, each appropriate instruction-following response by a child produced praise for that child. In experimental sessions, appropriate responses also produced tokens exchangeable for tangible reinforcers after the session. Token reinforcement differentially maintained instruction-following behavior in four children while one responded appropriately to most instructions and a second improved continuously during the study. While the data suggest that the present approach can be successfully applied to the alteration of instruction-following behavior in retarded children, its major contribution may be that of providing objective quantitative information about such behavior.