In a prospective double-blind study, single dose lumbar extradural blockade was instituted in 80 healthy male patients undergoing lower abdominal surgery. Patients were assigned randomly to one of four groups. Each group received treatment with 0.5% bupivacaine 20 ml containing adrenaline 1:200 000. To this solution was added fentanyl 0, 50, 100 or 200 micrograms. The time to regression of analgesic blockade was significantly prolonged (P less than 0.05) only with mixtures containing fentanyl 200 micrograms; no effect was demonstrated with lower doses of the opioid. However, the degree of motor blockade was less intense and incomplete S1-blockade (root-jumping) was more frequent with mixtures containing increasing doses of fentanyl. Patients treated with bupivacaine-fentanyl mixtures shivered less than the control group and showed fewer instances of acute hypotension. The side-effects observed were minor.