To assess the efficacy of beta-blockers in the prevention of rebleeding in selected cirrhotics and to compare the tolerability, safety of, and patient compliance with, a selective and a non-selective beta-blocker, 94 patients were randomly assigned to propranolol (32), atenolol (32), or placebo (30). Randomisation was made at least 15 days after the bleeding episode. Propranolol was given orally in increasing doses until the resting pulse rate was reduced by approximately 25%. Atenolol was given at a fixed dose of 100 mg/day. Patients were followed up for a mean of 357 days. Rebleeding occurred in 14 patients in the placebo group, 10 in the atenolol group and 8 in the propranolol group. The incidence of rebleeding was significantly lower in patients receiving propranolol than in those on placebo (PR vs PL: p less than 0.01, log-rank test). Atenolol was less effective than propranolol (AT vs PL: p = 0.065, log-rank test) but bleeding-free survival was improved for patients on active drugs compared with those patients on placebo (PR vs PL: p = 0.01; AT vs PL: p = 0.05, log-rank test). Retrospective analysis revealed that, whatever the type of treatment, abstinence from alcohol was crucial in preventing rebleeding. It was concluded that beta-blocker treatment is effective in preventing rebleeding from oesophageal varices in carefully selected alcoholic cirrhotics who survive at least 2 weeks after acute variceal haemorrhage and who cease drinking.