We have previously shown that the endoscopist is able to predict the endoscopic diagnosis in about two thirds of the patients. We report some preliminary findings from two different studies on the ability of symptoms to predict endoscopic findings. Comparison with similar studies in Glasgow and Huddinge suggests that the predictive value of symptoms probably varies between countries and depends on the population dealt with as well as on the methods used for symptom evaluation. Predictive models for peptic ulcer or endoscopic esophagitis based on symptoms showed at best an about 60—70 % sensitivity and specificity. We strongly feel that future studies on the predictive value of symptoms should be more focused on the first step in the decision process in general practice.