Nociception can produce segmental spinal sensitization or descending supraspinal antinociception. We assessed both types of sensory change after surgery during isoflurane-nitrous oxide anaesthesia with or without fentanyl before nociception. Patients undergoing back surgery received fentanyl 3 micrograms kg-1 (n = 15) or placebo (n = 15) before anaesthesia in a prospective, randomized, blinded study. Sensation, pain detection and tolerance thresholds to electrical stimulation were measured before operation at the arm, incision and herniated disc dermatomes (HDD) and 1, 2, 4, 6, 24 h and 5 days after operation, together with pain scores and patient-controlled morphine consumption (duration 24 h). For segmental effects, thresholds were normalized to the thresholds at a distant dermatome (arm). Raw pain thresholds were increased after operation (fentanyl > placebo) and were maximal at 4 h (pain tolerance in HDD: fentanyl +5.2 mA (+62.7%), placebo, +3.8 mA (+44.2%); P < 0.05 vs baseline for both). Normalized sensation thresholds decreased for placebo only (HDD/4 h: placebo, -1.8 (-44.8%), P < 0.05; fentanyl, +0.1 (+5.5%) ns). All changes returned to baseline by 24 h except for the placebo group normalized HDD sensation (d5: placebo, -2.4 (-59.7)%, P < 0.05; fentanyl -0.1 (-5.5%) ns). Pain scores and morphine consumption were similar. The study demonstrated both supraspinal analgesia and spinal sensitization after surgery. Fentanyl administration before operation augmented the former while decreasing the latter, and hence sensitization, especially if neuropathic, may particularly benefit from pre-emptive analgesia.