We have compared the sensitivities, specificities, and predictive values of three commercial serological assays for the diagnosis of Helicobacter pylori infection. A qualitative latex method (Pyloriset; Orion Diagnostics), a semiquantitative enzyme-linked immunosorbent assay (ELISA) (GAP test IgG; Bio-Rad), and a quantiative ELISA (Helico-G; Porton Cambridge) were used in 109 untreated dyspeptic patients. The presence of H. pylori was established when the results of culture and/or histology of the gastric biopsies taken were positive. The prevalence of H. pylori infection was 62% (52% in 42 patients younger than 45 years of age and 69% in 67 patients older than 45 years of age). Sensitivities and specificities were 68 and 76% for Pyloriset, 89 and 77% for GAP test IgG, and 82 and 83% for Helico-G. The positive predictive values for all three tests were between 85 and 90%. The predictive values for the absence of disease with a negative result were 62, 82, and 74% for Pyloriset, the GAP test, and Helico-G, respectively. With Helico-G in the younger group (less than 45 years), sensitivity significantly lower (71 versus 87%) and a positive predictive value lower than those for the older group (greater than 45 years) were found. Either the sensitivities and specificities of commercial methods for the measurement of antibodies to H. pylori in serum must be improved or the relationship between the presence of antibodies and the presence of bacteria in the stomach at the time of investigation is too weak to allow the use of serological techniques instead of culture and histological investigation of gastric biopsy material.