In the clinic, low efficacy positive GABAA modulators might be preferred to high efficacy positive modulators insofar as low efficacy modulators might have comparatively less abuse and dependence liability. Drug discrimination was used to examine the behavioral effects of L-838,417 and bretazenil, two low efficacy positive GABAA modulators that act at benzodiazepine sites, alone and in combination with benzodiazepines and a neuroactive steroid (alfaxolone). In rhesus monkeys (n = 5) discriminating midazolam, alfaxolone substituted for midazolam. In four monkeys, L-838,417 and bretazenil did not substitute for, but rather dose-dependently antagonized, midazolam; L-838-417 and bretazenil, as well as flumazenil, enhanced the midazolam-like effects of alfaxolone. L-838,417 and bretazenil substituted for midazolam in a fifth monkey. In a separate group of rhesus monkeys (n = 3) that received 5.6 mg kg(-1) per day of diazepam and that discriminated flumazenil, L-838,417 and bretazenil substituted for flumazenil. These results demonstrate that L-838,417, bretazenil, and flumazenil can have agonist or antagonist actions in the same animal depending upon whether they are studied in combination with a higher efficacy positive GABAA modulator acting at the same (benzodiazepine) or a different (neuroactive steroid) site. Thus, combinations of low efficacy positive modulators acting at different sites on the GABAA receptor complex could yield drug mixtures with significant therapeutic effects and with reduced abuse and dependence liability, as compared to higher efficacy positive modulators such as currently available benzodiazepines.