The effects of a training procedure and two maintenance contingencies on consequencedispensing behavior were investigated. Four peer behavior managers were trained to supervise small groups of subjects (four to six per group) working in programmed math materials and were compared with a teacher skilled in the use of social and point reinforcement and response cost. Manager training was differentially effective in accelerating manager's rates of appropriate social and point dispensing. Having manager reinforcement contingent upon manager consequence-dispensing resulted in moderately higher rates of appropriate social and point dispensing for three of four subjects than did having manager reinforcement contingent upon group study behavior. Two managers exposed to the group performance contingency before the manager performance contingency increased inappropriate social and point-dispensing behaviors to pretraining baseline levels. Subsequent change to the manager performance contingency was effective in reducing the inappropriate dispensing behavior of only one of the two managers.Researchers concerned with the problem of maintaining treatment effects produced in natural settings have suggested the training of social agents within the natural setting to initiate and carry out behavioral intervention (Hartmann, 1970;Patterson and Brodsky, 1966;Patterson, Cobb, and Ray, 1973). Social agents (e.g., parents, teachers, and peers) who are "naturally" in control of reinforcers and punishers, remain in the life setting after treatment is completed and can foster maintenance of gains due to their natural and continuing proximity to the target subject. Wetzel (1969) formalized a triadic intervention model whereby the experimenter trained a "mediator", (e.g., parent, teacher, or peer) to implement a behavioral program designed to modify the deviant behavior of the target or referral subject. With such an intervention model, it was possible for the experimenter to forego interaction with the change target since treatment was implemented and monitored completely by the trained mediator.Early use of the "triadic intervention model" was demonstrated by Ayllon and Michael (1959) and by Williams (1965). In the Williams study, parents were trained to ignore tantrum behavior of their child at bedtime. This -procedure rapidly extinguished that behavior. Ayllon and Michael (1959) trained the nursing staff of a mental hospital in the use of extinction, record keeping, and in administering social reinforcers. These procedures were effective in modifying numerous patient behavior problems; i.e., magazine hoarding, self-feeding, and psychotic talk.More recent utilization of the triadic model has involved the training of teachers to effect behavioral change efficiently in their classrooms.