Mycosis fungoides (MF) and Sézary's syndrome (SS) are cutaneous T-cell lymphomas characterized ultrastructurally by the presence of lymphoid cells with deep and narrow nuclear indentations (cerebriform mononuclear cells or CMC). Early diagnosis of MF and SS is difficult because in their early stages they often resemble various forms of chronic, benign skin lesions. By measuring the frequency distribution of a nuclear shape parameter (nuclear contour index or NCI) of lymphoid cells in skin infiltrates using computer assisted planimetry, we tried to classify suspect cases into the malignant and benign groups. From 12 patients with MF or SS (malignant group) and 11 patients with chronic, benign skin lesions (benign group) the frequency distribution of the NCI of the lymphoid cell population was measured. Nonlinear discriminant analysis selected the 70th and 25th percentile of the NCI distribution of the lymphoid cells in the skin infiltrates as parameters by which these patients could be classified correctly into the malignant or benign groups with a probability of over 95%. The predictive value of these parameters was tested on ten patients suspected of having cutaneous T-cell lymphomas. Three cases were classified as benign and 7 as malignant. In a three-year follow up cutaneous T-cell lymphomas did not develop in any of the 3 cases classified as benign, MF developed in 5 of 7 cases classified as malignant, 1 patient has lymphomatoid papulosis and 1 patient is still suspect for MF. These results are compared with those of DNA cytophotometry performed on skin imprint preparations. It is concluded that morphometry of lymphoid cells in skin lesions of patients suspect for MF and SS can make an important contribution to an early diagnosis of MF or SS.