Under concurrent VI (conc VI) schedules of reinforcement, organisms match the proportion of responses to the proportion of reinforcers obtained from the available options. This result is the basis of the matching law, a major theoretical view of the control of choice between and among available reinforcers. The present experiment examined the extent to which i.v. cocaine self-administration by monkeys under conc VI schedules of reinforcement was predicted by the matching law. One group of rhesus monkeys (n = 4) was prepared with chronic indwelling i.v. catheters and allowed to respond in a two-lever situation under conc VI schedules of reinforcement for cocaine injections. Three doses of cocaine (0.025, 0.05 and 0.1 mg/kg per injection) were made available under various conc VI schedules with an average inter-injection interval of 3 min. The same injection was available for each response, the difference between the options being the schedule of reinforcement. Each dose and schedule condition was in effect for at least ten consecutive sessions and until responding was stable. In a second group of monkeys (n = 3), a comparable experiment was conducted with responding maintained by the delivery of 1-g food pellets. Overall response rate maintained by cocaine was inversely related to dose. In addition, response rate decreased over the course of a session, apparently due to accumulation of cocaine. With regard to choice, more responding was maintained by the schedule that arranged more frequent injections. Choice was well predicted by the matching law, with a consistent tendency toward undermatching but no consistent bias toward one or the other option. Results were similar for behavior maintained by food, though two of three monkeys showed an unexplained bias toward the left lever. With regard to drug self-administration, these results demonstrate that in a choice situation, with all other variables being equal, injections that are available more frequently in time maintain behavior more strongly than less frequently available injections. Undermatching implies that the relative proportion of behavior maintained by the two options is somewhat less than the relative proportion of injections obtained. The finding that choice maintained by cocaine under conc VI schedules was comparable to choice maintained by food extends the generality of the conclusion that drugs and non-drug reinforcers control behavior by similar mechanisms.