Non-deprived rats were familiarized with a highly palatable diet until baseline consumption in a 60-min daily access period had stabilised. The benzodiazepine receptor agonist midazolam (1.25-10.0 mg/kg, IP) produced a large, dose-related increase in food consumption during the first 30 min of access. It also produced significant, short-term hyperphagia in animals which had been partially pre-satiated on the diet before drug administration, an effect which was reversible by the benzodiazepine receptor antagonist Ro15-1788. Administered alone, Ro15-1788 (1.25-10.0 mg/kg, IP) had no intrinsic activity in the food consumption test. In contrast, CGS 8216 (2.5-40.0 mg/kg, IP) produced a marked dose-related suppression of food intake. This anorectic effect was shared by two benzodiazepine receptor inverse agonists, FG 7142 and DMCM, which also produced dose-dependent reductions in consumption. The effects on feeding produced by FG 7142 (20 mg/kg, IP) and DMCM (1.25 mg/kg, IP) were reversed by either Ro15-1788 (2.5 and 5.0 mg/kg) or midazolam (5.0 and 10.0 mg/kg). A matched anorectic effect produced by CGS 8216 (40 mg/kg) was not, however, reversed by either Ro15-1788 or midazolam. This suggests that at a high dose CGS 8216 may act by a mechanism different from that of the two inverse agonists. The feeding test described in the report proved sensitive to both hyperphagic and anorectic effects of drugs active at benzodiazepine receptors, pointing to a possible bi-directional control of palatable food consumption.