Summary The milk fat globule membrane antibodies HMFGI, HMFG2, NCRC 11 and four of the Mam 6 series, and the lectins peanut agglutinin, wheat germ agglutinin, Concanavalin A, Lotus tetragonolobus and Ulex europaeus I have been applied to 115 stage I and II breast carcinomas (median follow up = 36 months) to assess their value as prognostic markers. Of the milk fat globule membrane antibodies only NCRC 11 staining showed a relationship to development of recurrent disease and overall survival, but this did not act as an independent indicator over and above that provided by histological grade. None of the lectins gave prognostic information, including those whose binding related to node status or grade. It is concluded that for short-term prognosis none of the markers-can given independent prognostic information over and above that provided by histological evaluation.The biological behaviour of breast carcinoma can vary considerably such that there are patients with a very poor prognosis (Blamey et al., 1979) and those who remain free from disease for over 20 years (Brinkley & Haybittle, 1975). Factors such as the extent of spread (both nodal and distant) at the time of presentation are important in predicting behaviour, and this forms the basis of staging systems. However, changes in surgical practice with a reduction in axillary node sampling, plus a need to subdivide node negative early stage patients into poor or good risk categories means that there is a need to maximise the amount of prognostic information which can be obtained from the primary tumour.The immunohistochemical assessment of primary tumours for the expression of various markers of potential prognostic value is one such approach. The use of antibodies to different components of the milk fat globule membrane has resulted in conflicting reports. Wilkinson et al. (1984) concluded that the antibody HMFG1 could be useful as a prognostic indicator, but Berry et al. (1985) found no relationship between the extent of staining with HMFG1 and HMFG2 and survival. Ellis et al. (1985, 1987) used the monoclonal antibody NCRC 11, which although raised against a breast carcinoma metastasis detects high molecular weight glycoproteins similar-to those recognised by antibodies raised against the milk fat globule membrane. They found a clear relationship between immunoreactivity of carcinomas and the clinical course of the disease. Angus et al. (1986), however, using the same antibody, did not find a statistically significant relationship with prognosis. Direct comparison of the staining obtained with HMFG1, HMFG2 and NCRC 11 antibodies for the same group of carcinomas and their relationship to prognosis has not been published.Another approach which has -been evaluated for its prognostic value has been the binding of lectins to primary tumours. Greater reactivity of carcinomas with Helix pomatia was reported as relating to poorer survival (Fenlon et al., 1987;Leathem & Brooks, 1987) as did Ulex europaeus I binding (Fenlon et al., 1987).In the present study a grou...