Baboons and rats were trained under a two-lever, food-reinforced drug discrimination procedure. The training drug was either lorazepam (1.0 mg/kg) or pentobarbital (5.6 mg/kg in baboons, 10.0 mg/kg in rats). Under test conditions, a range of training drug doses occasioned 100% drug lever responding. CGS 8216 (3.2-10.0 mg/kg) combined with lorazepam produced a complete shift to the no-drug lever in both species; this shift was surmountable with higher doses of lorazepam. CGS 8216 (32.0 mg/kg) combined with pentobarbital produced a statistically significant decrease in drug-lever responding in rats, and in baboons CGS 8216 initially, but not subsequently, produced a complete shift to the no-drug lever. Caffeine (0.32-10.0 mg/kg) combined with lorazepam inconsistently decreased drug-lever responding across multiple determinations in baboons and significantly decreased drug lever responding in rats. Caffeine combined with pentobarbital also yielded an inconsistent decrease in drug lever responding in baboons but there was no effect in rats. Thus the most reliable and complete antagonism across species was obtained with the CGS 8216/lorazepam combinations.