The risk of developing esophageal cancer in Barrett's esophagus has been estimated at about 10%. This estimate is based primarily on data concerning the prevalence of that association in series of hospitalized patients and autopsies--a practice that tends to exaggerate the risk. We have reviewed retrospectively our experience with 115 patients whom we treated for Barrett's esophagus between September 1962 and March 1983, and have distinguished our data on the prevalence of esophageal cancer from that on incidence. For 8 patients, we found both Barrett's esophagus and esophageal adenocarcinoma during the initial evaluations for a prevalence ratio of 7%. We followed 105 patients not found to have esophageal cancer initially for a total of 350 person-years. Only 2 patients developed adenocarcinoma during that follow-up period for an incidence of 1 case per 175 person-years. This incidence, although some 40-fold greater than that of the general population, is substantially lower than previously estimated. Routine endoscopic and histologic surveillance has been recommended for patients with Barrett's esophagus because of the alleged high incidence of esophageal cancer. Inasmuch as that incidence now appears to be low, we question the value of such surveillance for these patients.