Because most human studies of the neurobiological substrates of the effects of cocaine have been performed with drug-dependent subjects, little information is available about the effects of cocaine in the initial phases of drug use before neuroadaptations to chronic exposure have developed. The purpose of the present study, therefore, was to define the substrates that mediate the initial effects of cocaine in a nonhuman primate model of cocaine self-administration using the 2-[14C]deoxyglucose method. Rhesus monkeys were trained to self-administer 0.03 mg/kg per injection (N = 4) or 0.3 mg/kg per injection (N = 4) cocaine and compared with monkeys trained to respond under an identical schedule of food reinforcement (N = 4). Monkeys received 30 reinforcers per session, and metabolic mapping was conducted at the end of the fifth self-administration session. Cocaine self-administration reduced glucose utilization in the mesolimbic system, including the ventral tegmental area, ventral striatum, and medial prefrontal cortex. In addition, metabolic activity was increased in the dorsolateral and dorsomedial prefrontal cortex, as well as in the mediodorsal nucleus of the thalamus. These latter effects are distinctly different from those seen after the noncontingent administration of cocaine, suggesting that self-administration engages circuits beyond those engaged merely by the pharmacological actions of cocaine. The involvement of cortical areas subserving working memory suggests that strong associations between cocaine and the internal and external environment are formed from the very outset of cocaine self-administration. The assessment of the effects of cocaine at a time not readily evaluated in humans provides a baseline from which the effects of chronic cocaine exposure can be investigated.