Pigeons trained to discriminate 0.1 mg/kg flumazenil, proposed as an in-vivo model to study interactions with diazepam-insensitive gamma-aminobutyric acid (GABA)A receptors, were tested with various GABAergic and non-GABAergic compounds. As a result of its pharmacological selectivity, the model was suitable for further examining previously reported flumazenil-like effects of gamma-hydroxybutyrate (GHB). Flumazenil and the GABAA negative modulator Ro 15-4513 produced 82-100% flumazenil-appropriate responding. Diazepam and the direct-acting GABAA agonists muscimol and 4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridine-3-ol (THIP) produced 38-64% flumazenil-appropriate responding. GHB, its precursors 1,4-butanediol (1,4-BD) and gamma-butyrolactone (GBL), and the GABAB agonists baclofen and SKF97541 produced 0-24% flumazenil-appropriate responding. Baclofen shifted the flumazenil dose-response curve to the right and down, possibly involving perceptual masking of the discriminative stimulus effects of flumazenil by agonist activity at GABAB receptors. These masking effects of baclofen were blocked by the GABAB antagonist CGP35348. When CGP35348 was given together with GHB to block its GABAB agonist effects, GHB did not produce flumazenil-appropriate responding. Conceivably, effects of GHB at non-GABAB receptors (e.g. diazepam-sensitive GABAA receptors and GHB receptors) may interfere with the expression of its flumazenil-like discriminative stimulus effects. The asymmetric substitution between GHB and flumazenil is consistent with the hypothesis that the discriminative stimulus effects of GHB consist of several components, not all of which are mimicked by flumazenil.