Different doses of intravenous cocaine reinforced the lever pressing of rhesus monkeys under two-lever concurrent or concurrent-chain schedules. Under the concurrent procedure, responding produced drug reinforcers arranged according to independent variable-interval 1-min schedules. Under the concurrent-chain procedure, responding in the variable-interval link led to one of two mutually exclusive, equal-valued, fixed-ratio links; completion of the ratio produced a drug reinforcer. Under both procedures, responding on one lever produced a constant dose of 0.05 or 0.1 mg/kg/injection, while on the other lever, dose was systematically varied within a range of 0.013 to 0.8 mg/kg/injection. Preference, indicated by relative response frequency on the variable-dose lever during the variable-interval link, was always for the larger of the doses. Relative response frequencies on the variable-dose lever roughly matched relative drug intake (mg/kg of drug obtained on variable lever divided by mg/kg of drug obtained on both levers). For many dose comparisons, responding occurred and reinforcers were obtained almost exclusively on the preferred lever. Overall variable-interval rates generally were lower than with other reinforcers, and these low rates, under the experimental conditions, may have occasioned the exclusive preferences.