A reliable diagnosis of Helicobacter pylori is important in clinical practice and research. To evaluate sensitivity, specificity and inter-observer variation of three histological staining methods for the diagnosis of intragastric H. pylori bacterial flora, four observers assessed the presence of H. pylori in a test set of pairs of gastric biopsies taken from 40 patients during GI-endoscopy. Histological slides of one biopsy of each patient (formalin-fixed and paraffin-embedded) were stained with a Modified Giemsa (MG), the Warthin-Starry (WS), and an immunohistochemical method (IMM) using purified polyclonal H. pylori antiserum (DAKO B471). As a standard, bacterial culture was performed on the adjacent biopsy, using the four quadrants method. Special attention was paid to the presence of non-H. pylori bacterial flora. Twenty out of 40 specimens were H. pylori-positive by culture. Using culture as a standard, sensitivity for H. pylori was 90.0+/-10.0% with MG, 70.0+/-14.1% with WS, and 83.8+/-11.1% with IMM stain. Specificity was 53.8+/-19.3%, 82.5+/-9.6% and 90.0+/-0.0%, respectively. Of 37, 14 and eight false positive scores by all observers for MG, WS and IMM, respectively, non-H. pylori flora was cultured from 17, six, and four of the corresponding adjacent biopsies. Non-H. pylori flora may account for overdiagnosis of H. pylori in gastric biopsies. Kappa values for the variance between the four observers were 0.28 for MG, 0.57 for WS, and 0.83 for IMM. Immunohistochemical staining for H. pylori is highly specific and has a low inter-observer variation. We advise that in situations where gastric histology is the main diagnostic tool, IMM should be used for the detection of H. pylori.