Appropriate classroom behavior was defined in terms of four component categories: on-time in-seat, on-task, social interaction, and assignment completion. A measure of total appropriate classroom behavior was calculated by summing the frequency of target response occurrence for each component behavior. Withdrawal and reinstatement of token reinforcement procedures demonstrated experimental control of total appropriate behavior; however, considerable variability of component behaviors was observed. Token reinforcement control was most clearly demonstrated with on-time and in-seat behavior; assignment behavior was similarly influenced but showed greater variability; on-task behavior was noticeably affected but exhibited an ascending trend toward the reversal phase; and social interaction behavior showed fluctuations which appeared unrelated to reinforcement contingencies. These findings suggest that the use of global, composite measures may mask program effects on important component behaviors. Thus it cannot be concluded that when global appropriate behavior is demonstrated to be controlled by reinforcement contingencies, component or specific behaviors are similarly controlled. Footnotes, one figure, and nine references are provided. (Author abstract modified)