The diagnostic yield of blind percutaneous liver biopsy is improved when two or three specimens are taken during the procedure by redirecting the needle through a single entry-site, without exposing the patient to a greater risk of complications provided that standard precautions are taken. This study was designed to obtain further data on the safety of this procedure at King Edward VIII Hospital, Durban. During the period 1984-1990 (inclusive) a total of 2,646 biopsies were carried out: a single specimen was obtained in 834 patients, two specimens in 983 patients and three in 829 patients. Complications directly attributable to the procedure occurred in 24 patients who had one specimen, 20 who had two, and 19 who had three specimens taken during the biopsy. A single specimen had been obtained from three of the eight patients who had died, two specimens had been taken from another patient, and three specimens were obtained from the other four patients, i.e. patients in whom two or three specimens were taken did not have a higher incidence of pain, symptomatic hypotension, biliary peritonitis or death than those in whom one specimen was taken. Accordingly, when blind percutaneous needle biopsy of the liver is carried out, two specimens should be obtained by redirecting the needle through a single entry site as this improves the diagnostic yield without increasing complications. The morbidity and mortality associated with liver biopsy in this hospital is, however, high. Good technique, careful monitoring of patients after biopsy and prompt and aggressive resuscitation are essential if the mortality rate is to be reduced.