Several benzodiazepine receptor ligands were pharmacologically characterized in a battery of functional tests after oral administration in mice, rats, and monkeys. Previous experiments have consistently demonstrated that diazepam exhibits high intrinsic efficacy, bretazenil exhibits intermediate intrinsic efficacy, Ro 42-8773 and Ro 41-7812 both show low intrinsic efficacy, and flumazenil exhibits virtually zero intrinsic efficacy. In the test battery used here it appears that nearly full intrinsic efficacy is required for clear anterograde amnesia or rotarod impairment. In contrast, full protection in the pentetetrazol test was achieved with intermediate-to-high intrinsic efficacy and nearly full protection with lower intrinsic efficacy. In the audiogenic seizure test full anticonvulsant effects were produced with intrinsic efficacy ranging from low to high. Clear inhibition of punished operant responding was observed for all test compounds except for Ro 41-7812 and flumazenil, which exhibit the lowest intrinsic efficacies. All of the test compounds enhanced palatable food consumption, with even those having low intrinsic efficacy producing maximum effects approximating that of diazepam. By additionally taking into consideration the degree of receptor occupancy required to obtain pharmacological activity in each of the tests in this battery it is possible to order the compounds with respect to intrinsic efficacy: diazepam > bretazenil > Ro 42-8773 > Ro 41-7812 > flumazenil. The latter four compounds all exhibited a maximum antagonistic activity in tests involving reversal of meclonazepam- or flunitrazepam-induced central nervous system depression. Thus, using these tests appears to permit the accurate ordinal classification of benzodiazepine receptor ligands for intrinsic efficacy.(ABSTRACT TRUNCATED AT 250 WORDS)