To investigate whether the degree of differentiation in feline mammary carcinoma (FMC) can indicate the post-surgical survival time (PST), tumours were surgically resected from 55 cats and histologically graded according to a method derived from human breast cancer studies. One year after the resection, 26 cats (47.3%) were alive while 29 (52.7%) had died as a consequence of FMC. Formalin-fixed, paraffin wax-embedded sections stained with haematoxylin and eosin were used to classify the FMCs initially, according to the WHO system. Histological grading was then performed on the basis of three main features: degree of tubule formation, nuclear and cellular pleomorphism, and accurate mitotic count obtained from a defined area. Immunohistochemical examination with an anti-actin antibody was used for the accurate detection of "in situ" carcinomas. Age and histological type were not significantly correlated with the PST. Seven tumours (12.7%) were graded as well-differentiated carcinoma (WDC; grade I), 33 (60%) as moderately differentiated carcinoma (MDC; grade II), and 15 (27.3%) as poorly differentiated carcinoma (PDC; grade III). The tumour-related death rates after the first post-surgical year were 0 in cats with WDC, 14 (42.4%) in those with MDC, and 15 (100%) in those with PDC. Six cats with tumours showing extensive myoepithelial differentiation were all alive after 1 post-surgical year. The grading system seemed to have a good predictive value in respect of grades I and III of FMC but not for grade II. Myoepithelial differentiation may be relevant to clinical prognosis in FMC.