The purpose of the present study was to investigate the disruptive effects of cannabinoids on working memory as assessed in the eight-arm radial-maze. Systemic administration of delta 9-THC, WIN-55,212-2, and CP-55,940 increased the number of errors committed in the radial-maze. CP-55,940 was the most potent cannabinoid in impairing memory (ED50 = 0.13 mg/kg). delta 9-THC and WIN-55,212-2 disrupted maze-choice accuracy at equipotent doses (ED50 values = 2.1 and 2.2 mg/kg, respectively). In addition, systemic administration of each of these agents retarded completion time. Whereas the doses of delta 9-THC and CP-55,940 required to retard maze performance were higher than those needed to increase error numbers, WIN-55,212-2 was equipotent in both of these measures. On the other hand, neither anandamide, the putative endogenous cannabinoid ligand, nor cannabidiol, an inactive naturally occurring cannabinoid, had any apparent effects on memory. A second aim of this study was to elucidate the neuroanatomical substrates mediating the disruptive effects of cannabinoids on memory. Intrahippocampal injections of CP-55,940 impaired maze performance in a dose-dependent manner (ED50 = 8 micrograms/rat), but did not retard the amount of time required to complete the maze. The effects of intrahippocampal CP-55,940 were apparently specific to cognition because no other cannabinoid pharmacological effects (e.g., antinociception, hypothermia, and catalepsy) were detected. This dissociation between choice accuracy in the radial-maze and other cannabinoid pharmacological effects suggests that the working memory deficits produced by cannabinoids may be mediated by cannabinoid receptors in the hippocampus.