Three pathologists, especially interested in breast cancer, reexamined 379 random specimens of invasive breast cancer initially diagnosed at 27 pathological institutes; these were obtained from the computer register of the Danish Breast Cancer Cooperative Group (DBCG). The degree of variation between a) the primary diagnosis (the country as a whole) and each of the 3 pathologists and b) between the 3 pathologists mutually, has been studied with regard to the 2 main groups of carcinoma--infiltrating duct carcinoma (IDC) and infiltrating lobular carcinoma (ILC). The degree of variation was found to be similar between each of the 3 pathologists and the country as a whole, with a kappa value of approx. 0.3 for both types of carcinoma, and a considerably better interobserver variation between the 3 pathologists with a kappa value of approx. 0.7. The most important reason was that the 3 pathologists agreed on the occurrence of more than twice as many cases primarily diagnosed as ILC. A partial explanation is in fact that since the start of DBCG, reports have appeared of variants of ILC not previously known. The diagnosis of malignancy (carcinoma) was almost completely unambiguous.