Biopsy and serum specimens were obtained from 95 patients undergoing endoscopy at the University of Zimbabwe Medical School. Common presenting features were epigastric pain, bleeding and dyspepsia. Ulcers were detected in 16 patients (17%), and were more common in men (24%) than in women (7%). Histological examination of biopsies showed that all 95 patients had spiral-shaped organisms that were indistinguishable microscopically from Helicobacter pylori, though the numbers of organisms varied considerably. There was evidence that the degree of inflammation in the mucosa was related to the numbers of H. pylori-like organisms (HPLO) present. Fifty-one biopsy specimens (55%) gave a positive rapid urease test (RUT), with colour change occurring within 4 h. In all but one case, the gastric mucosa from these patients contained moderate to numerous HPLO. We defined the 'gold standard' of H. pylori-associated gastritis as the presence of both moderate to numerous HPLO and moderate to severe inflammation in the gastric mucosa. Using these criteria, RUT had a sensitivity of 67% and a specificity of 68%. Sera from 92 patients were tested for immunoglobulin G antibodies reactive with a glycine-extract antigen of H. pylori, using an enzyme-linked immunosorbent assay (ELISA). Sera giving an indeterminate reaction in the ELISA were also tested by Western blotting. In all, 36 sera (39%) gave a positive ELISA or Western blot reaction. There was poor correlation between serology and RUT results, with only 57% of biopsy specimens from seropositive patients giving a positive RUT, compared with 45% from seronegative patients. Positive serology was found in only 35 patients (61%) with histological evidence of H. pylori-associated gastritis, and the specificity of the test was only 54%. When used in combination with the RUT result, however, 79% of patients with a positive RUT and positive serology had histological evidence of H. pylori-associated gastritis. There was a general trend for increased seroprevalence in patients with mild to moderate atypia. These findings indicate that serology, using an antigen derived from the type strain of H. pylori, is unreliable in detecting H. pylori infection in Zimbabwe. Current studies are aimed at characterizing antigens from organisms isolated from Zimbabwean patients.