The reinstatement of extinguished cocaine self-administration behavior was studied in rats pretreated with N-methyl-D-aspartate receptor antagonists. Rats were trained to self-administer intravenous cocaine (0.32 mg/kg/infusion) during five consecutive daily sessions that were followed by five consecutive daily extinction sessions, during which cocaine was unavailable and cocaine-associated cues (sound and light) were absent. Neither the competitive N-methyl-D-aspartate receptor antagonist D-CPPene (0.3-3 mg/kg) nor the low-affinity N-methyl-D-aspartate receptor channel blocker memantine (1-10 mg/kg) reinstated extinguished responding. Priming injections of intravenous cocaine (Experiment 1), and exposures to cocaine-associated stimuli (buzzer and light; Experiment 2) engendered responding on the reinforced lever in excess of that on the non-reinforced lever. In Experiment 1, administration of D-CPPene or memantine prior to the priming injection of cocaine eliminated the difference between reinforced-lever and non-reinforced-lever response rates. For both D-CPPene and memantine, however, this effect was largely due to increased responding upon the non-reinforced lever rather than to decreased reinforced-lever responding. In Experiment 2, D-CPPene, but not memantine, abolished in a dose-dependent manner the selective increase in reinforced-lever over non-reinforced-lever responding that was induced by exposures to cocaine-related stimuli. This effect of D-CPPene was not due to increased non-reinforced-lever responding. These data help define the boundaries within which N-methyl-D-aspartate receptor antagonists can prevent reinstatement of cocaine-seeking behavior (e.g. type of antagonist used and reinstatement procedure).