The main aims of the Sydney system for the classification of gastritis are to improve uniformity in histopathological reporting and to provide a flexible matrix of rules for grading the histological features. We sought to determine the level of interobserver agreement between pathologists in the application of the Sydney system. Three histopathologists independently examined H & E, alcian blue/PAS and modified Giemsa stained sections of two antral and two corpus gastric biopsies from 69 consecutive dyspeptic patients. After elimination of five unsuitable cases, each observer graded chronic inflammation, polymorph activity, atrophy, intestinal metaplasia and Helicobacter pylori density in the antrum and corpus on a 0-3 scale according to the Sydney system criteria. The pairwise agreement on final diagnosis and the overall and conditional agreement on histological grades were examined by kappa statistics. Agreement on the final diagnosis ranged from 83-94% with kappa values of 0.699 ('good') to 0.887 ('excellent'). Conditional probability of agreement on a diagnosis of H. pylori positive gastritis was 99%, but wider disagreements were apparent in the recognition of H. pylori negative gastritis, reactive gastritis and even normal biopsies. Overall agreement for grade ranged from 70% for antral atrophy to 94% for intestinal metaplasia in the corpus with 'moderate' or 'good' kappa values. We conclude that the diagnostic and grading criteria described in the Sydney system can be applied consistently by histopathologists. The findings underline its potential usefulness in routine practice.