The fraction of portal venous inflow shunted through portasystemic collaterals was computed by comparing the area of an isotope dilution curve recorded from the hepatic vein after injection into the superior mesenteric or splenic artery with that after injection into the hepatic artery. In seven normal subjects no shunting could be detected, but in forty patients with alcoholic liver disease shunting of mesenteric flow averaged 61.9 per cent and splenic flow 80.1 per cent. In nine of twenty-seven complete studies, there was a discrepancy of greater than 40 per cent between mesenteric and splenic shunting, with splenic shunting usually but not always predominating. No significant relationship between per cent shunting and hepatic blood flow or portal pressure was found. Repeat studies after three to eighteen months in three patients showed progressive increase in the fraction of shunted portal flow. These data indicate that portasystemic collateral circulation may develop independently in the mesenteric and splenic beds in portal hypertension. Furthermore, establishment of an effective collateral circulation in patients with alcoholic liver disease does not necessarily deprive the liver of blood flow or decompress the portal vein.