Recent research on genetically modified mice has attributed the amnesic effect of benzodiazepines mainly to the alpha1-containing GABA(A) receptor subtypes. The pharmacological approach, using subtype selective ligands, is needed to complement genetic studies. We tested the effects of the non-selective antagonist flumazenil (0-20.0 mg/kg), the preferential alpha1-subunit selective antagonist beta-carboline-3-carboxylate-t-butyl ester (beta-CCt) (0-30.0 mg/kg), the non-selective agonist midazolam (0-2.0 mg/kg), the preferential alpha1-subunit selective agonist zolpidem (0-3.0 mg/kg), and the non-selective inverse agonist methyl 6,7-dimethoxy-4-ethyl-beta-carboline-3-carboxylate (DMCM) (0-2.0 mg/kg) in the one-trial step-through passive avoidance task in rats. The compounds were administered intraperitoneally, before the acquisition test. Flumazenil and beta-CCt did not affect retention performance. Midazolam and zolpidem induced amnesia in a dose-dependent manner. The complete reversal of amnesia was unattainable. The effects of zolpidem were significantly attenuated by the both, flumazenil (10.0 mg/kg) and beta-CCt (30.0 mg/kg); by contrast, only flumazenil was considerably effective when combined with midazolam. DMCM exerted promnesic effects at 0.2mg/kg, in an inverted U-shape manner. Both antagonists tended to abolish this action. The results indicate that some other alpha-subunit(s), in addition to the alpha1-subunit, contribute to the amnesic actions of non-selective benzodiazepine site agonists in the passive avoidance task. On the other hand, a significant part of the DMCM-induced promnesic effect could involve the alpha1-subunit and/or other putative beta-CCt-sensitive binding site(s).