Fentanyl, a mu-opioid receptor agonist, produces analgesia while leaving vibrotactile sensation intact. We used positron emission tomography (PET) to study the mechanisms mediating this specific effect in healthy, right-handed human males (ages 18-28 yr). Subjects received either painful cold (n = 11) or painless vibratory (n = 9) stimulation before and after the intravenous injection of fentanyl (1.5 microgram/kg) or placebo (saline). Compared with cool water (29 degrees C), immersion of the hand in ice water (1 degrees C) is painful and produces highly significant increases in regional cerebral blood flow (rCBF) within the contralateral second somatosensory (S2) and insular cortex, bilaterally in the thalamus and cerebellum, and medially in the cerebellar vermis. Responses just below the statistical threshold (3.5 < Z < 4.0) are seen in the contralateral anterior cingulate, ipsilateral insular cortex, and dorsal medial midbrain. The contralateral primary sensory cortex (S1) shows a trend of activation. Except for slight changes in intensity, this pattern is unchanged following a saline placebo injection. Fentanyl reduces the average visual analogue scale ratings of perceived pain intensity (47%) and unpleasantness (50%), reduces pain-related cardioacceleration, and has positive hedonic effects. After fentanyl, but not placebo, all cortical and subcortical responses to noxious cold are greatly reduced. Subtraction analysis [(innocuous water + fentanyl) - (innocuous water + no injection)] shows that fentanyl alone increases rCBF in the anterior cingulate cortex, particularly in the perigenual region. Vibration (compared with mock vibration) evokes highly significant rCBF responses in the contralateral S1 cortex in the baseline (no injection) and placebo conditions; borderline responses (3.5 < Z < 4. 0) are detected also in the contralateral thalamus. Fentanyl has no effect on the perceived intensity or unpleasantness of vibratory stimulation, which continues to activate contralateral S1. Fentanyl alone [(mock vibration + fentanyl) - (mock vibration + no injection)] again produces highly significant activation of the perigenual and mid-anterior cingulate cortex. A specific comparison of volumes of interest, developed from activation peaks in the baseline condition (no injection), shows that fentanyl strongly attenuates both the contralateral thalamic and S1 cortical responses to noxious cold stimulation (P < 0.048 and 0.007, respectively) but fails to affect significantly these responses to vibrotactile stimulation (P > 0.26 and 0.91, respectively). In addition, fentanyl, compared with placebo, produces a unique activation of the mid-anterior cingulate cortex during fentanyl analgesia, suggesting that this region of the cingulate cortex participates actively in mediating opioid analgesia. The results are consistent with a selective, fentanyl-mediated suppression of nociceptive spinothalamic transmission to the forebrain. This effect could be implemented directly at the spinal level, indirectly through cingulate corticofugal pathways, or by a combination of both mechanisms.