To re-evaluate the accuracy of the serological test for Helicobacter pylori, fixation of biopsy specimens with Carnoy's solution (preserving the mucous layer in tissue preparations) followed by immunohistochemical staining (a new histological procedure) was used as the reference histological method instead of 10% formalin fixation followed by hematoxylin-eosin staining (the conventional histological procedure). Biopsy specimens (antrum and body) from 114 patients with gastritis (including non-ulcer dyspepsia) or gastric and/or duodenal ulcers were obtained by endoscopy and used for both bacteriological culture and histological examination. Serum samples were taken from all patients at the time of endoscopy. The serum levels of specific IgG and IgA antibodies for Helicobacter pylori were measured by commercial enzyme immunoassay kits. The reliability of the IgG and IgA measurements was evaluated by analyzing receiver operating characteristic curves obtained using the two histological procedures. With the conventional histological procedure as the reference, the sensitivity and specificity levels of the serological test were 87.2% and 82.1%, respectively. With the new histological procedure as reference, sensitivity and specificity were 94% and 96.7%, respectively. The insufficient accuracy reported for the serological test could be due to false-positive or false-negative results obtained when the conventional histological procedure is used as the reference. The new histological procedure used here revealed that the serological test for Helicobacter pylori is more reliable than previously thought.