Sixty elderly patients were given at random either epidural analgesia with bupivacaine 0.75% or general anaesthesia with thiopentone, fentanyl, pancuronium, N2O/O2 for total hip replacement. Preoperatively the patients were of equal physical status with normal and similar laboratory values. All patients were mentally normal for their age. On the 1st postoperative day, the general anaesthesia group had a significantly lower than PaO2 than the epidural group (P less than 0.025). PaO2 in the general anaesthesia group was significantly lower than the preoperative value on the 1st and 3rd days (P less than 0.001 and P less than 0.01, respectively). None of the 29 patients in the epidural group but seven of 31 patients in the general anaesthesia group has significant mental changes postoperatively (P less than 0.01). Five of these patients still had mental changes which reduced the quality of life several months later. In the general anaesthesia group one patient died from acute myocardial infarction. Low postoperative PaO2 might have contributed to this death. Two patients in the epidural group had symptoms of pulmonary embolism postoperatively. Thus elderly patients appear to do better after hip replacement with less deterioration of cerebral and pulmonary functions when given epidural analgesia than when surgery is performed under general anaesthesia. These patients should therefore be offered epidural analgesia whenever possible.