Ammonia is considered the major pathogenetic factor of cerebral dysfunction in hepatic failure. The correlation between total plasma ammonia and the severity of hepatic encephalopathy (HE), however, is variable. Because ammonia that is present in gaseous form readily enters the brain, the correlation with the grade of HE of the pH-dependent partial pressure of gaseous ammonia (pNH3) could be better than that of total arterial ammonia levels. To test this hypothesis, 56 cirrhotic patients with acute episodes of clinical HE (median age, 54 years; range, 21-75) were studied by clinical examination and by long-latency median-nerve sensory-evoked potentials (SEPs) N70 peak, an objective and sensitive electrophysiological measure of HE. pNH3 was calculated from arterial blood according to published methods. The clinical grade of HE correlated (P < .001) with both pNH3 and total ammonia, but correlation was stronger with pNH3 (r = .79 vs. .69, P = .01). A similar correlation was found for N70 peak latency (r = .71 with pNH3 vs. .64 with total ammonia, respectively, P = .08). In summary, arterial pNH3 correlates more closely than total ammonia with the degree of clinical and electrophysiological abnormalities in HE. These findings support the ammonia hypothesis of HE and suggest that pNH3 might be superior to total ammonia in the pathophysiological evaluation of HE.