Sixty-four patients admitted with acute alcoholic hepatitis, with or without underlying cirrhosis, were randomized regardless of encephalopathy to receive a controlled diet either alone, or supplemented orally, nasogastrically, or intravenously as necessary, with 2000 kCal and 10 g nitrogen daily. Whether this came from a conventional protein source or a branched chain amino acid enriched formulation was also randomly determined. In the absence of renal failure, nitrogen intakes of 10 g or more daily were invariably associated with positive nitrogen balance, but complications of liver dysfunction prevented the attainment of significantly more positive balance in the supplemented groups than in controls. Neither in the series as a whole, nor in any identifiable subgroup of patients, was mortality affected by treatment. Changes in prothrombin time and in measured nutritional parameters during the study did not differ between supplemented and control groups, and the observed changes in midarm muscle circumference appeared to reflect changes in degree of fluid retention. Neither enteral nor parenteral branched chain amino acids showed any consistent effect upon encephalopathy.