The effects of ( -)-huperzine A ((5R,9R,11 E)-5-amino-11-ethylidene- 5,6,9,10-tetrahydro-7-methyl-5,9-methanocycloocta[b]pyridin-2(1H)-one), and of the hydrochloride salt of E2020 ((R,S)-1-benzyl-4-[(5,6-dimethoxy-1-indanon)-2-yl]-methyl piperidine) and tacrine (9-amino-1,2,3,4-tetrahydroacridine), on the scopolamine-induced memory deficits in rats were compared in a radial maze, using a 4-out-of-8 baiting procedure. Scopolamine (0.15 mg/kg, i.p.) caused significant impairment in the rats' ability to fulfil the radial maze task. (-)-Huperzine A (0.2-0.4 mg/kg, p.o.; 0.1-0.4 mg/kg, i.p.) had greater efficacy than E2020 (0.6-0.9 mg/kg, p.o.; 0.3-0.6 mg/kg, i.p.) and tacrine (1.5-2.5 mg/kg, p.o.; 0.3-0.6 mg/kg, i.p.) on the improvement of scopolamine-induced working and reference memory errors, respectively. There appeared to be an inverse bell-shape dose-dependent effect for all three compounds tested. The compared data demonstrate that (-)-huperzine A is the most potent and orally active acetylcholinesterase inhibitor of the three, and fits more closely the established criterions for an ideal acetylcholinesterase inhibitor to be used in clinical studies. (C) 1998 Elsevier Science B.V. All rights reserved.