Serum carbohydrate antigen (CA15-3) values were examined in 300 normal subjects in order to determine the standard value of this antigen. The clinical relevance of repeatedly assaying this marker in patients with or without recurrent breast cancer postoperatively was compared with assaying the serum carcinoembryonic antigen (CEA) values. The upper limit of CA15-3 was calculated as being 25.3 U/ml in the normal subjects and the distributions of CA15-3 values were not markedly different among the normal subjects, even if they had been selected according to sex or age. Moreover, no differences were observed among normal women who had been randomly selected according to the age distribution of the breast cancer patients. Thirty samples taken from the breast cancer patients postoperatively revealed values of higher than 25 U/ml and 73 samples showed lower levels. The serum CEA values were positive in 16 samples and negative in 85 samples. Although the accuracy of the CEA assay was about 10 per cent higher than that of the CA15-3 assay, its low positive rate was unsatisfactory for effective use in the breast clinic. The results of this study suggest that serum CA15-3 is not detectable unless there is a relatively large number of tumor cells. The higher false positive rate of the CA15-3 assay should therefore be considered as suggesting recurrence.