This study examined the contribution of endogenous opioids to the antinociception produced by microinjection of the GABAA receptor antagonist, bicuculline, into the rat midbrain ventrolateral periaqueductal gray region. Microinjection of bicuculline (40 ng/0.4 microliter) into the periaqueductal gray produced robust antinociception as measured by the tail-flick latency to noxious heat. This antinociception was partially reversed by intravenous administration of the non-selective opioid antagonist naloxone hydrochloride (1 and 5 mg/kg), indicating that endogenous opioid release is necessary for this effect. To determine whether opioid release in the rostral ventromedial medulla, a major projection target of the periaqueductal gray, contributes to this effect, we microinjected another opioid antagonist, naltrexone, into the rostral ventromedial medulla. Naltrexone in the rostral ventromedial medulla (5 and 10 micrograms/microliter) significantly attenuated bicuculline antinociception elicited from the periaqueductal gray. Cys2, tyr3, orn5, pen7-amide (26.5 nmol), a selective mu-opioid receptor antagonist, also reversed the antinociception when microinjected into the rostral ventromedial medulla. Microinjections of naltrexone (10 micrograms/microliter) or cys2, tyr3, orn5, pen7-amide at sites in the medulla dorsal to the rostral ventromedial medulla were ineffective. None of the antagonists altered baseline tail-flick latencies. These results support the hypothesis that a population of periaqueductal gray neurons produces antinociception through a mu-opioid receptor-mediated action of endogenous opioids in the rostral ventromedial medulla. Thus, two opioid-sensitive pain-modulating brainstem sites are linked by an endogenous opioid synapse in the rostral ventromedial medulla.