Twenty-seven consecutive patients scheduled for total hip replacement were randomly divided into three groups. The first group had their operations under epidural analgesia, the second whilst breathing halothane spontaneously and patients in the third group were anaesthetised using a modified neuroleptoanaesthetic (NLA) technique. Blood lost during the operation was measured by a colorimetric technique. Closed suction drains were used and all blood lost after the operation collected for measurement. Deliberate hypotensive techniques were not employed. The mean operative blood loss of patients who had their operations under epidural anaesthesia was significantly less than that of patients in the other two groups, being 341-6 +/- 59-1 ml (s.e.m.) for the epidural group, 648-4 +/- 58-3 ml (s.e.m.) for the halothane group and 744 +/- 98-9 ml (s.e.m.) for the NLA group. There was no significant difference in the amounts of blood collected by closed suction drainage in any of the three groups and the mean values for total overall blood loss incurred by procedure were 734-1 +/- 40-7 ml (s.e.m.) for the epidural group, 986-3 +/- 94-9 ml (s.e.m.) for the halothane group and 1168-4 +/- 126 ml (s.e.m.) for the NLA group. In this series only 3 of 10 patients (30%) who had their operation under epidural anaesthesia required transfusion with whole blood, compared with 7 of 9 (78%) in the halothane group, and 7 of 8 (87-5%) in the neuroleptoanalgesia group. From this series it is evident that the need for transfusion of whole blood in total hip replacement can be significantly reduced by the use of epidural analgesia.