This study compared the results of hepatectomy for hepatocellular carcinoma (HCC) using an ultrasonic dissector with those of a combination of the crushing clamp and finger fracture techniques. The crushing clamp and finger fracture method was used from 1989 to 1992 in 96 patients (group 1) and the ultrasonic dissector from 1993 to 1994 in 69 patients (group 2). Data from these two sets of patients were collected prospectively. The groups were comparable in terms of preoperative liver function, tumour size and stage, and the incidence of cirrhosis. Major hepatectomy was performed in 69 patients (72 per cent) of group 1 and in 52 (75 per cent) of those in group 2. Use of the ultrasonic dissector resulted in lower mean (s.e.m.) blood loss (group 1 3.4(0.4) litres versus group 2 2.4(0.2) litres, P = 0.02), lower mean (s.e.m.) blood transfusion requirement (2.2(0.2) versus 1.2(0.2) litres, P = 0.001) and more patients not requiring blood transfusion (8 per cent of group 1 versus 32 per cent of group 2, P = 0.0001). Postoperative complications occurred in 45 patients (47 per cent) of group 1 and 19 (28 per cent) of those in group 2 (P = 0.012). There were no deaths in group 2 whereas the hospital mortality rate in group 1 was 16 of 96 (17 per cent) (P = 0.0004). A wider tumour-free resection margin (mean(s.e.m.) 1.2(0.1) versus 0.9(0.1) cm, P < 0.05) and lower serum bilirubin level throughout the postoperative period were also observed in group 2 patients. The ultrasonic dissector is better than the crushing clamp and finger fracture technique in hepatectomy for HCC.