Five hundred four Shunt procedures have been done at Emory University Hospitals between 1971 and 1981 to decompress bleeding esophageal varices. This paper reviews how far the experiences of a prospective randomized study (55 patients) of distal splenorenal shunts against total shunts is supported by the nonrandomized experience (449 patients), and outlines our current methods of management dictated by this experience. The overall operative mortality for 348 selective shunts is 4.1% and for 156 nonselective shunts, 14.1%. The five-year survival following Selective shunt is 59%, and following nonselective shunt is 49%: more than half the selective shunt patients are alive, in contrast to the median survival of 44.5 months for patients having nonselective shunts. Following Selective shunt, the survival in nonalcoholic patients is significantly better than the median survival of alcoholic patients of 57 months. Encephalopathy, reported at three years after surgery in the randomized patients was significantly (p < 0.001) lower after selective shunt (12%) compared to nonselective shunt (52%): in the same population at seven years, all patients with patent nonselective shunts have clinical or subclinical encephalopathy, but only 30% of the selective shunt patients have subclinical encephalopathy. Shunt patency, immediately after surgery, is 93% following selective shunt, with only two documented late thromboses: nine of nine patients, at a mean of seven years, retain patency in the randomized study. Shunt occlusion increases with time after interposition nonselective shunts: seven of 13 are occluded at a mean follow-up of seven years in the randomized study. Portal venous perfusion is retained in 93% of patients seven to ten days after selective shunt, but in no patient with a patent nonselective shunt. Late portal perfusion is maintained in nine of the eleven patients in the randomized group studied at a mean of seven years after selective shunt. Restoration of portal perfusion has led to clearing of encephalopathy and improvement in hepatic function in six patients. The following conclusions are made: (1) selective shunts can be done with low operative mortality, and long-term patency with excellent control of bleeding; (2) hepatic portal venous perfusion has been maintained after selective shunt for ten years, and this is vital for preventing encephalopathy and maintaining hepatic function; (3) long-term survival after selective shunt is better than any reported series for nonselective shunt; and (4) selective shunts are the operative procedure of choice for variceal decompression and nonselective shunts should rarely be performed for elective decompression.