SummaryArterial and venous ammonia were studied in different diseases by a modified Conway method.In cirrhosis, the correlation between mental state and arterial ammonia was good; with venous blood, the relationship was poor. In comæof other origin (uremia, cerebral thrombosis, etc.), blood ammonia remained quite normal. After abdominal paracentesis, blood ammonia did not rise. Ammonia concentration in ascitis was different from arterial or peripherical venous blood ammonia.The authors could confirm the existence of an important ammoniacal arterio-venous difference in impending hepatic coma and in hepatic coma. This difference did not completely disappear after warming of the leg. The significance of that arterio-venous difference, the physiopathology of hepatic coma, and the concomitant disturbances of the intermediary metabolism are discussed.The auhors stress the importance of arterial ammonia determination for the adjustment of protein intake to the individual tolerance of each cirrhotic patient.The cases studied here show that protein restriction is not only useful in the treatment of impending hepatic coma; it also improves the more chronic mental troubles which occur in some cases of severe cirrhosis.Nevertheless, food adjustment and antibiotherapy do not prevent the usual complications of cirrhosis, especially upper gastro-intestinal bleedings.