Pigeons obtained food by making four responses on three keys in a specified sequence. Errors produced 5-second timeout periods, during which the keylights were off and responses had no effect. To establish a base line of repeated acquisition, the sequence of correct responses was changed from session to session. Cocaine (3 mg/kg) disrupted the behavior: total errors increased, the relative frequency of perseverative errors increased, the rate of within-session error reduction (learning) decreased and the total trial time (pausing) increased. During repeated drug administration (30-50 sessions), these effects disappeared, i.e., tolerance developed. Tolerance did not develop, however, to cocaine-induced increases or decreases in timeout responding; such effects were nondisruptive in the sense thay they did not reduce the rate of food reinforcement. For comparison, cocaine (3-10 mg/kg) was also studied under a "performance" condition, in which the sequence of correct responses was the same from session to session. Cocaine increased performance errors and produced pausing, but tolerance developed more quickly under the learning condition . The more rapid development of tolerance was presumably due to the stronger stimulus control of behavior under the performance condition.