Mice were tested in a simple automated Y-maze. Total number of arm entries and alternation behaviour were measured. The latter is thought to reflect working memory capacity at a rudimentary level. During an 8 min session, vehicle-treated mice performed 32.4 +/- 7.4 arm entries, 51.0 +/- 12.4% of which were organized in alternations (triplets). The two variables showed a negative correlation. Scopolamine (1.0 mg/kg) significantly enhanced activity, reduced alternation behaviour and diminished the correlation between the two variables. The effects of benzodiazepine receptor inverse agonist, antagonist and agonist beta-carbolines on this spontaneous behaviour and on the effects of scopolamine were examined. The effects of inverse agonists and agonists on locomotor activity were complex in interaction with both vehicle and scopolamine. The scopolamine-induced reduction of alternation behaviour was significantly reversed by the antagonist ZK 93426 but not by inverse agonists; furthermore, partial agonists and agonists showed no effects. It is hypothesized that the interaction of antagonist beta-carbolines with scopolamine is based on a direct GABA-ergic control of cholinergic neurotransmission, and suggests an ability of antagonist beta-carbolines to antagonize amnestic properties of scopolamine.