Immunocytochemical assays (ICA) using monoclonal antiestrogen receptors (ER ICA), antiprogesterone receptors (PR ICA), and monoclonal antibody Ki67 (Ki67 ICA) were performed in 127 breast carcinomas. The immunostaining procedures were applied on frozen tissue sections, tumour imprints, and fine-needle aspirates in order to compare the variations in the distribution of the antigens detected in the three different types of preparations. Positive reactions detected with peroxidase-antiperoxidase and avidinbiotin-peroxidase, and alkaline phosphatase-antialkaline phosphatase complexes were evaluated through a computerized system of image analysis referred to as SAMBA 200 (SAMBA TITN, Grenoble, France). Application programs specifically developed for the analysis of tissue sections and of cytologic preparations were applied. This system allowed a multiparametric, accurate, reliable, reproducible and automatized evaluation of the heterogeneity of the antigenic sites in tumors. For each markers positive cell surface (PS), and integrated and mean optical densities (IOD, MOD) and IOD histograms were compared. It was shown that (1) there was no significant variation in optical densities in cell imprints and aspirates whereas PS significantly (P less than 0.01) differed in both preparations; (2) there were significant differences of the optical densities between tissue sections and cytological preparations, either imprints or aspirates, likely due to randomly cut nuclei in tissue sections; and (3) there was a significant difference between the PS of tissue sections and aspirates but no significant difference between tissue sections and imprints. It is concluded that fine-needle aspiration constitutes a convenient method for cell sampling, reliable for the diagnosis of malignancies. However, it may not reflect the heterogeneity of cell subpopulations in tissue.