It has long been hypothesized that human as well as animal cocaine users titrate their intake to maintain a specific level of cocaine reward. This hypothesis predicts that the dose-injection function of each subject individually should be a decreasing function, with no initial, gradual ascending limb. The present study was designed to test this specific prediction. Rats were trained to self-administer cocaine under a continuous reinforcement schedule. After stabilization of cocaine self-administration, all rats were tested with a wide range of i.v. cocaine doses (0.0078-1 mg). To accurately measure the threshold dose of each individual, the pharmacological resolution was set at 0.0078 mg at the four lowest doses. As predicted, individual dose-behavior curves are discontinuous at a threshold dose, with a descending limb but no gradual, ascending limb. Below the threshold, there is no evidence for cocaine self-administration; at and above the threshold, the rate of injections spikes to its maximum and then decreases lawfully with the dose, a decrease that reflects cocaine titration. In all individuals, this critical transition occurred over a dose interval of less than 0.008 mg. This study suggests that the cumulative effects of cocaine maintained during self-administration are all-or-nothing-a conclusion that confirms the regulation hypothesis of cocaine reward. The neurobehavioral consequences of this specific level of cocaine reward remain to be elucidated.