Sections of neoplasms from 76 female patients with primary operable carcinoma of the breast were independently assessed by 2 pathologists for histological features and assigned a grade score. Relative disagreement rates between pathologists were estimated by use of a log-linear model and found to be similar to those reported by many other groups, but higher than that reported by acknowledged experts. Tumor grade was related to nuclear DNA content as measured by static cytometry, inversely related to oestrogen receptor status and provided some additional prognostic information but, in this small series of patients, did not correlate with short-term survival as closely as other prognostic indicators such as ploidy, tumor size or the extent of lymph node involvement. Patients with Grade III tumors had a particularly poor prognosis, however, there were few patients allotted to Grade III (poorly differentiated tumors), and survival differences between Grades I and II were small; in short-term followup, used alone, grading separated out only a small proportion of patients into useful prognostic groups. This preliminary study emphasizes the need for a careful approach to the use of grading of breast carcinomas in the routine histopathology laboratory. Demonstration of higher levels of interobserver agreement, or concordance with experts in the field, will be necessary before our grading can be incorporated into a prognostic index useful for patient management.