Eighty-nine patients underwent 95 portasystemic shunts for portal hypertension at our institutions between June 1963 and March 1981. Ninety-three shunts were performed for bleeding varices. Procedures that were performed included 11 Warren shunts, 29 Linton shunts, 28 interposition mesocaval shunts, 26 classic portacaval shunts, and one umbilical to saphenous vein shunt. Thirty-six shunts were performed in Child class A patients (5.5% operative mortality), 37 in Child class B (16.2% operative mortality), and 22 in Child class c patients (36.3% operative mortality). Five-year survival for Child A patients was 74 per cent, 17.4 per cent for Child B patients and 26.3 per cent for Child C patients. Twelve of 14 patients (15.7%) who had recurrent postoperative upper gastrointestinal bleeding were found to have occluded shunts (two Warren, six Linton, one mesocaval, and three portacaval). Of 21 patients who became encephalopathic postoperatively, 17 had alcoholic liver disease. In 15 of these alcoholic patients who survived the immediate postoperative period, encephalopathy correlated directly with continued alcohol consumption. Several conclusions can be drawn from our data: portasystemic shunts can be performed with acceptable morbidity and mortality; rebleeding generally indicates shunt occlusion; encephalopathy can be correlated with continued alcohol consumption after shunting; there appears to be little difference in survival and rebleeding in the various shunt procedures; the poor survival figures for Child B and C patients must make one seriously consider alternative procedures in these poor-risk candidates.