Symptomatic patients referred to an open-access upper gastrointestinal endoscopy completed a detailed, self-administered questionnaire aimed at assessing the predictive value of history in dyspepsia. Nine hundred and thirty patients were suitable for analysis. Of these, 29% were found to have organic dyspepsia. A substantial overlap of symptoms and demographic data was found among the various endoscopic diagnoses. Discriminating variables were identified by stepwise logistic regression analysis and included in predictive score models. Pain relieved by antacids, age above 40 years, previous peptic ulcer disease, male sex, symptoms provoked by berries, and night pain relieved by antacids and food were found to predict organic dyspepsia with a sensitivity and specificity of approximately 70%, when applied on the observed material. Similar probabilities were found for score models of peptic ulcer and esophagitis. In general, the low prevalence of organic diseases resulted in low positive and high negative predictive values. Accordingly, the main impact of the predictive models may be to reduce the number of negative endoscopies rather than to predict a precise diagnosis. Independent of disease category and age, 41% of the subjects expressed a fear of malignancy, emphasizing the value of reassurance from a negative endoscopy.