Antinociceptive effects elicited from the midbrain may involve both ascending and descending projections from the periaqueductal gray and dorsal raphe nucleus. To investigate the relationship between these different efferent pathways in the rat, we performed a double-labeling study using two retrograde tracers, colloidal gold-coupled wheatgerm agglutinin-apo horseradish peroxidase and a fluorescent dye. One tracer was microinjected in the medullary nucleus raphe magnus; the second was injected into one of several regions rostral to the periaqueductal gray that have been implicated in nociceptive and antinociceptive processes. The results can be grouped into two categories. First, injections into the ventrobasal thalamus, lateral hypothalamus, amygdala, and cerebral cortex labeled neurons in the dorsal raphe nucleus but not in the periaqueductal gray. Up to 90% of these projection neurons were serotonin immunoreactive, and up to 17% were also retrogradely labeled from the nucleus raphe magnus. Second, only injections into the ventrobasal hypothalamus (which included the beta-endorphin-containing arcuate neurons) or into the medial thalamus labeled neurons in the periaqueductal gray itself. Injections into the medial thalamus, but not into the ventrobasal hypothalamus, also labeled neurons in the dorsal raphe nucleus. Up to 20% of the neurons retrogradely labeled from these regions were also retrogradely labeled from nucleus raphe magnus. The presence of large populations of rostrally projecting periaqueductal gray neurons that collateralize to the nucleus raphe magnus implies that activity in ascending projections necessarily accompanies any activation of the periaqueductal gray-nucleus raphe magnus pathway. Possibly, projections from the medial thalamus and medial hypothalamus mediate antinociceptive effects that complement descending inhibition. Finally, possible antidromic activation of these pathways must be considered when interpreting the results of electrical brain stimulation studies.