Between October, 1981 and December 1983, 46 consecutive unselected cirrhotic patients who had been admitted for variceal hemorrhage were assigned to propranolol for the prevention of variceal rebleeding and followed for a mean period of 13.3 months (ranges: 1 day, 45 months). At the time of inclusion, 87 p. 100 of patients had alcoholic cirrhosis: ascites was present in 43 p. 100, jaundice in 43 p. 100 and encephalopathy in 33 p. 100 of cases. The goal of this retrospective study was to assess the effectiveness of propranolol in these unselected patients and to determine variables associated with a high rebleeding rate. Twenty-four patients (52 p. 100) rebled from esophageal varices within a mean time of 8.2 months (range: 6 days, 39 months) and 12 of these patients died. The percentage of patients who rebled from varices was 27 p. 100 at 6 months, 43 p. 100 at 1 year, 52 p. 100 at 18 months. Patients with ascites at the time of inclusion were 3.5 times more likely to rebleed than patients without ascites (p = 0.007). Patients with jaundice (total bilirubin ≥ 50 μmol/l) were 3.1 times more likely to rebleed than patients without jaundice (p = 0.01). One year after the beginning of this treatment, the probability of rebleeding was 19 p. 100 in the 17 Pugh-Child class A cirrhotic patients, 60 p. 100 in the 15 class B patients, 55 p. 100 in the 14 class C patients. Neither proper acceptance in taking propranolol nor alcohol withdrawal were associated with significantly modified risks of rebleeding in these unselected alcoholic patients.