To assess the reliability of the World Health Organization's grading system for breast carcinomas, six observers independently graded 158 tumors. Five of the observers had been simultaneously trained in the same institution. The sixth was one of their tutors. There was agreement in the grading of only 23 cases, most of which were either well or poorly differentiated tumors. If the prognosis breast carcinoma is predicated by the who method, it should be restricted to the two extreme groups. It seems that there is no need to apply the who grading method because these groups could be easily distinguished by the overall impression of the examining pathologists.