There has been a resurgence of interest in the use of histological grade in predicting prognosis in patients with breast cancer, although several studies have questioned the reproducibility of the techniques used, Interobserver agreement rates of 90% have been reported by the Nottingham group, recognized experts in the field, but a previous study from this institution showed only a 70% agreement rate between two pathologists independently reviewing histological sections. This study reports the results of an attempt to improve internal consistency in reporting histological grade by using established, well defined criteria, and a consensus opinion by pathologists 'trained' in the method. Criteria established by the Nottingham group were used by three pathologists to obtain a consensus histological grade in 98 cases of breast cancer, assessed on two occasions at an interval of 6 months. The intra-observer agreement rate of 86% showed an improvement on the previous interobserver agreement rate, and is comparable to agreement rates of 90% reported by the Nottingham group. Relative disagreement rate (RDR) of 0.10 (95% CI 0.05-0.19) and kappa statistic 0.77 (95% CI 0.66-0.88) also showed improvement over previous results, and compares favourably with other studies, The results suggest 'training' of pathologists in assessing histological grade, strict adherence to guidelines, and a consensus opinion, may all contribute to greater reproducibility.