Four pigeons were trained on multiple variable-interval schedules in which components alternated after a fixed number of responses had been emitted. In Part 1, each component change occurred after 20 responses; in Part 2, the number was 40; and in Part 3, the number of responses before change was 10. Component reinforcer rates were varied over five experimental conditions in each of Parts 1 to 3. Component response rates decreased as the specified number of responses per component was increased. However, the relation between component response-rate ratios and component reinforcerrate ratios was independent of the specified number of responses per component, and was similar to that found when components alternate after fixed time periods. In the fourth part of the experiment, the results from Parts 1 to 3 were systematically replicated by keeping the component reinforcer rates constant, but different, while the number of responses that produced component alternation was varied from 5 to 60 responses. The results showed that multiple-schedule performance under component-response-number constraint is similar to that under conventional component-duration constraint. They further suggest that multiple-schedule response rates are controlled by component reinforcer rates and not by principles of maximizing overall reinforcer rates or meliorating component reinforcer rates.Key words: multiple schedules, time allocation, response allocation, response constraints, molar maximizing, melioration, generalized matching, pecking, pigeonsThe generalized matching law (Baum, 1974(Baum, , 1979) provides a convenient description of behavior allocation as a function of reinforcers obtained in both concurrent and multiple variable-interval (VI) schedules. This law suggests that behavior ratios are a power function of obtained reinforcer ratios. For multiple schedules, if the two component responses are subscripted w and r, and if B is the number of responses emitted, T is the time spent in the components, and X is the number of reinforcers obtained, then: log (BwIw) = a log ( 4W) + log c, (1) where the parameter a is called sensitivity to reinforcement