Two of the most commonly used procedures to study the abuse-related effects of drugs in laboratory animals are intravenous drug self-administration and reinstatement of extinguished behavior previously maintained by drug delivery. Intravenous self-administration is widely accepted to model ongoing drug-taking behavior, whereas reinstatement procedures are accepted to model relapse to drug taking following abstinence. Previous studies indicate that 5-HT2A receptor antagonists attenuate the reinstatement of cocaine-maintained behavior but not cocaine self-administration in rodents. Although the abuse-related effects of cocaine have been closely linked to brain dopamine systems, no previous study has determined whether this dissociation is related to differential regulation of dopamine neurotransmission. To elucidate the neuropharmacological and neuroanatomical mechanisms underlying this phenomenon, we evaluated the effects of the selective 5-HT2A receptor antagonist M100907 on intravenous cocaine self-administration and drug- and cue-primed reinstatement in rhesus macaques (Macaca mulatta). In separate subjects, we evaluated the role of 5-HT2A receptors in cocaine-induced dopamine overflow in the nucleus accumbens (n = 4) and the caudate nucleus (n = 5) using in vivo microdialysis. Consistent with previous studies, M100907 (0.3 mg/kg, i.m.) significantly attenuated drug- and cue-induced reinstatement but had no significant effects on cocaine self-administration across a range of maintenance doses. Importantly, M100907 (0.3 mg/kg, i.m.) attenuated cocaine-induced (1.0 mg/kg, i.v.) dopamine overflow in the caudate nucleus but not in the nucleus accumbens. These data suggest that important abuse-related effects of cocaine are mediated by distinct striatal dopamine projection pathways.