Converging evidence supports a role for mesocorticolimbic dopaminergic systems in a subject's ability to shift behavior in response to changing stimulus-reward contingencies. To characterize the dopaminergic mechanisms involved in this function, we quantified the effects of subtype-specific dopamine (DA) receptor antagonists on acquisition, retention, and reversal of a visual discrimination task in non-human primates (Chlorocebus aethiops sabaeus). We used a modified Wisconsin General Test Apparatus that was equipped with three food boxes, each fitted with a lid bearing a unique visual cue; one of the cues concealed a food reward, whereas the other two concealed an empty box. The monkeys were trained first to acquire a novel discrimination (eg A(+), B(-), C(-)) in a single session, before experiencing either a reversal of the discrimination (eg A(-), B(+), C(-)) or the acquisition of a completely new discrimination (eg D(+), E(-), F(-)), on the following day. Systemic administration of the D(2)/D(3) receptor antagonist raclopride (0.001-0.03 mg/kg) failed to significantly affect the performance of reversal learning when reversal sessions were run without a retention session. But, raclopride (0.03 mg/kg) significantly impaired performance under the reversal condition when reversal sessions were run right after a retention session; however, it did not affect acquisition of a novel visual discrimination. Specifically, raclopride significantly increased the number of reversal errors made before reaching the performance criterion in the reversal, but not in new learning sessions. In contrast, the D(1)/D(5) receptor antagonist SCH 23390 did not significantly modulate acquisition of a novel discrimination or reversal learning at doses (0.001-0.03 mg/kg, i.m.) that did not suppress behavior generally. In addition, none of the drug treatments affected retention of a previously learned discrimination. The results strongly suggest that D(2)/D(3) receptors, but not D(1)/D(5) receptors, selectively mediate reversal learning, without affecting the capacity to learn a new stimulus-reward association. These data support the hypothesis that phasic DA release, acting through D(2)-like receptors, mediates behavioral flexibility.