This double-blind, placebo-controlled study investigated the efficacy of intermittent doses of intrathecal fentanyl in 30 patients undergoing thoracotomy. They were allocated randomly to three groups, two of which had microspinal catheters inserted into the lumbar subarachnoid space at the end of surgery; the third group acted as a control. Intrathecal fentanyl or 0.9% saline was administered through the catheters and all patients received morphine using a patient-controlled analgesia (PCA) system. Pain scores, morphine consumption and peak expiratory flow rates (PEFR) were recorded on an hourly basis. Intrathecal fentanyl resulted in a faster onset of analgesia (mean visual analogue scale (VAS) score at 1 h = 0.9 compared with 6.3 (95% confidence intervals for the difference -6.8, -4.0) for the other groups; P < 0.001) and significantly lower pain scores at rest, on cough and on movement. PEFR values were consistently higher in the intrathecal fentanyl group. There were no cases of early or delayed respiratory depression.