Rats were trained to run in a spatial, radial arm maze using a procedure to determine two memory functions, working and reference memory. The muscarinic antagonist, not the nicotinic antagonist, impaired both working and reference memory of rats. Scopolamine (0.125, 0.15, and 0.2 mg/kg, IP, 30 min before a session) significantly impaired choice accuracy in the eight-arm maze. In contrast, mecamylamine (5, 10, and 15 mg/kg) did not affect the performance. Huperzine A (0.1, 0.2, and 0.3 mg/kg, IP, 30 min before testing) and physostigmine (0.3 mg/kg, IP, 20 min before testing) could reverse scopolamine-induced deficits in the task. Chronic treatment with huperzine A (0.25 mg/kg, PO, once a day) for 8 consecutive days was as potent as acute treatment on attenuating the scopolamine-induced amnesia.