We have approached the question of the reliability of the thymidine labeling index (TLI) of breast carcinoma by comparing paired samples. Because the distribution of the TLI is approximately lognormal, comparisons were made after transformation to the natural logarithm. For 38 paired samples from the primary carcinoma (biopsy versus residual tumor in mastectomy specimen), r, 0.89, p less than 0.0001. For 68 primary carcinomas versus their axillary metastases, r, 0.90, p less than 0.0001. Substantial correlation was also observed between the TLIs of 18 primary breast carcinomas and their recurrent or metastatic lesions studied after relapse (r, 0.75, p less than 0.0003), and between 14 pairs of first and second primary carcinomas recovered from the same breast (r, 0.72, p less than 0.003). Good agreement was achieved among three observers in measuring the TLI from a set of breast carcinoma autoradiographs. We conclude that the TLI is sufficiently reliable for semiquantitative use as a means of separating breast carcinomas into groups with slow, moderate, and rapid rates of proliferation. The substantial similarities between the TLIs of breast carcinoma samples from different sites of the primary tumor, between primary tumor and axillary metastases, and between primary tumor and subsequent samples of relapsed tumor suggest that clonal evolution either is infrequent or does not usually affect the cell kinetics of the carcinoma. The similarity of the TLIs of synchronous dual primary carcinomas is an additional interesting finding.