The clinical course and causes of death in 132 consecutive patients with fulminant hepatic failure and grade III or IV encephalopathy have been reviewed. 105 patients died and in 96 of these an autopsy examination was performed. In 36 patients there was cerebral oedema and the mean age of this group was significantly younger than the other fatal cases. In 28 patients death was attributed to major haemorrhage which originated in the gastrointestinal tract in 25. The prothrombin time ratio was not significantly greater in patients with major bleeding than in those without but they did have a significantly lower platelet count. Sepsis contributed to death in 12 patients. In 25 patients massive hepatic necrosis only was found at autopsy and death was considered to be due solely to hepatic failure. The degree of hepatocyte loss was assessed in 80 fatal cases by a histological morphometric technique on a needle specimen of liver taken immediately post-mortem. The proportion of the liver volume occupied by hepatocytes (hepatocyte volume fraction, HVF) was greatly reduced in all patients (normal 85+/-SD 5 percent) but the mean value was significantly higher in the patients dying with sepsis, cerebral oedema or haemorrhage than in the group in whom death was attributed solely to hepatic failure. There were ten patients in whom liver function was improving at the time of death which was due to cerebral (9) or haemorrhage (1). These observations suggest that many patients presently dying from fulminant hepatic failure may be expected to survive, once more effective therapy is available for the complications of the illness.