The aim of this study was to investigate the biological significance of tumor shape in breast cancer by considering the shape not as a casual event but as an expression of the behavior and natural history of the tumor. The shape was studied by an analytical morphometry procedure and was related to axillary metastases, which up to now are the most meaningful prognostic factors in this disease. Fifty cases of infiltrating breast carcinoma (25 N+ and 25 N-) were investigated. The shape, studied on subgross sections of the tumor, was considered as the result of two components: the subtle contour irregularities and the main distortions of the figure. The procedures used allowed us to distinguish and to parametrize these two components in order to submit them to univariate analysis (Student's t test), a principal components analysis and, finally, a multivariate discriminant analysis (Hotelling test). The utilized analytical procedures by work-station S.A.M. (S.A.M. work station is a product of "Metamorphosis") consisted of three steps: 1) Extraction of tumor function curve obtained by Kth order polynominals which gives a smoothing effect to the original curve; 2) Evaluation of contour irregularities by Fourier harmonic analysis; 3) Evaluation of shape asymmetry by SAE (shape asymmetry evaluator). We considered also the roundness factors of the original and function curves and finally the maximum tumor diameter. Three parameters relating to contour irregularities (sum and mean value of Fourier harmonic amplitude and percentage of superimposed points) and parameters relating to main distortions of the figure (mean value of SAE) were highly significant (p less than 0.001). The roundness factor of the original curve was more significant (p less than 0.001) than that of the function curve (p less than 0.01) whereas maximum tumor diameter was not significant when tested by Student's t test. Multivariate discriminant analysis allowed 20% of error to be reached by using 3 parameters relating to the shape analysis and the two roundness factors. By using 8 parameters, including maximum tumor diameter, the percentage error was 16%. The results obtained, while they stress the usefulness of the employed procedure, reveal that shape of the tumor, together with its dimensions, is an important expression of the biological behavior relating to metastatic spread.