We analyzed the clinical course of 85 consecutive, endoscopically verified variceal bleeders. Most patients were alcoholics with advanced stages of hepatic dysfunction. Bleeding was major in all, and medical mortality was 42% at 6 wk. Factors affecting mortality and factors not affecting survival, such as age and comorbid conditions, were identified. Sixty percent of early deaths and 40% of late deaths were attributable to bleeding. Approximately one-third of patients experienced rebleeding within 6 wk, and one-third of survivors experienced subsequent bleeds. In these patients significant improvement with observation was often anticipated, but could not be verified. The majority of deaths associated with variceal bleeding occur soon after the bleeding episode. Those who survive the hospitalization for bleeding may not fare worse than others of similar hepatic functional reserve but who have not experienced bleeding. We demonstrated that statistically significantly different survival curves could be obtained from the same population by changing the zero time for calculating survival. For example, a 1-yr survival rate of 34% for all medically treated patients could be raised to 52% by eliminating from consideration those who failed to survive 2 wk. The long-term survival course of those surviving greater than 2 wk was not statistically significantly different from published series of unselected cirrhotics without bleeding. Variceal bleeding is a common accompaniment of advanced liver disease. We propose that any substantial improvement in long-term survival must improve survival for the early period. If controlled trials are to be done, patients should be matched for hepatic functional reserve and identical zero time for calculating survival must be used.