Squirrel monkeys self-administered cocaine under a variable-dose schedule, with the dose varied from injection to injection. As in earlier studies with rats, post-injection pauses varied as a monotonic function of dose, allowing a cocaine dose-effect curve to be obtained during each session. These curves were shifted by pretreatment with dopamine antagonists, demonstrating that this procedure may provide an efficient means of evaluating treatments that affect drug self-administration. However, drug intake eventually became "dysregulated" after extensive training (100-300 sessions), with relatively short pauses following all doses. Dose-sensitivity was restored by adding a 60-s timeout period after each injection, suggesting that dysregulation occurred because the monkeys developed a tendency to self-administer another injection before the previous injection had been adequately distributed. Finally, when the response requirement under the variable-dose schedule was increased from 1 to 10, both the post-injection pause and the rate of responding following the pause ("run rates") were found to vary with dose. The dose-dependency of run rates suggests that post-injection pauses reflect not only motivational factors, such as satiety, but also the direct effects of cocaine on leverpressing.