The study sample consisted of six male youths between 10-and 15-years old who had been committed by the juvenile court to a family-style residential rehabilitation program in Florida. In a token economy system, youths earned points through appropriate behaviors which then entitled them to privileges such as snacks or recreational activities. The baseline evaluation of room-cleaning behavior was determined by an adult observer using a checklist of 15 definitions and was randomly checked by another observer. The following experimental conditions were established: (1) a phase when boys were instructed about room-cleaning requirements, but there were no scheduled consequences for this behavior: (2) an external token system in which a professional staff member evaluated a youth's room-cleaning behavior daily and awarded points; (3) and a self-evaluation token system. The six youths were randomly assigned to two treatment groups which were subjected to different sequences of experimental conditions over a 180-day period. The effects of each condition on room-cleaning behavior are described illustrated in graphs. Under instructions and external token conditions, room-cleaning behavior improved to an appropriate level, but rapidly deteriorated when tokens were not given. The self-evaluation system was effective in increasing and maintaining room-cleaning behavior for all youths. These findings suggest that self-evaluation should be incorporated into token economy systems to maintain self-regulatory behaviors. One note and 22 references are included.