Tumour growth rates, as measured by incorporation of tritiated thymidine, have been reported as being of prognostic importance in breast cancer. We have measured the thymidine labelling index (TLI) of 185 early breast cancers, followed-up for a minimum of 8 years. Above median TLI was associated with higher tumour grade, but not with other prognostic factors. TLI was not predictive of survival in either univariate or multivariate analysis. The inter- and intra-observer reproducibilities of TLI measurements were poor, which may be a factor limiting its usefulness as a prognostic indicator in breast cancer.