Pain relief following stimulation of the periaqueductal gray matter (PAG) or periventricular gray matter (PVG) in man has been ascribed to stimulation-induced release of endogenous opioid substances. Forty-five patients were studied and followed for at least 1 year after placement of chronic stimulating electrodes in the PAG or PVG to determine if pain relief due to stimulation could be ascribed to an endogenous opioid mechanism. Three criteria were assessed: the development of tolerance to stimulation; the possibility of cross-tolerance to morphine; and reversibility of stimulation-induced pain relief by the opiate antagonist naloxone. Sixteen patients (35.6%) developed tolerance to stimulation, that is, they obtained progressively less effective pain relief. Twelve (44.4%) of 27 patients undergoing stimulation of the thalamic sensory relay nuclei for treatment of chronic pain (a presumably non-opioid mechanism) also developed tolerance. Morphine sulfate was administered in a blind, placebo-controlled protocol to 10 patients who had become tolerant to PAG-PVG stimulation and none showed evidence of cross-tolerance. Fifteen of 19 patients, already tolerant to morphine at the time of PAG-PVG electrode implantation, experienced excellent pain relief by stimulation, also indicating a lack of cross-tolerance. Twenty-two patients who experienced excellent pain relief from chronic PAG-PVG stimulation received intravenous naloxone in a double-blind, placebo-controlled protocol. Pain intensity as assessed by the visual analog scale was increased to the same degree by both placebo and naloxone. Eight patients showed no increase in pain intensity with either placebo or naloxone. Although tolerance to PAG-PVG stimulation developed in these patients, the frequency of tolerance was similar to that seen in patients undergoing thalamic sensory nuclear stimulation. Since the latter technique presumably relieves pain by a non-opioid mechanism, the development of tolerance to PAG-PVG stimulation does not, in itself, confirm an opioid mechanism. Cross-tolerance between PAG-PVG stimulation and morphine was not seen and cross-tolerance to PAG-PVG stimulation in patients already tolerant to morphine was rare. The pain-relieving effect of PAG-PVG stimulation was reversed to an approximately equal degree by naloxone and placebo. The authors do not believe that, in most patients, pain relief elicited by PAG-PVG stimulation depends on an endogenous opioid mechanism. It appears that other, non-opioid mechanisms are primarily responsible for such pain relief.