A new method is described for analysing the clonal evolutionary dynamics of tumour growth and the lineage relationship of primary tumours to their metastases. It exploits random integrations of transfected plasmid or retroviral infected (proviral) DNA as a means of generating very large numbers of uniquely marked cell clones in a single-step selection whose fates can then be tracked during progressive tumour growth. Using a mouse breast adenocarcinoma we undertook experiments in which syngeneic mice were injected with a mixture of very large numbers of uniquely marked cell clones, only one or a few of which were metastatic, or with reconstituted mixtures containing a genetically tagged metastatic clone with an excess of non-marked non-metastatic tumour cells. Among the results we summarize is the finding that spontaneous metastases developed in a non-random fashion from genotypically distinct cell clones. They were clonal or biclonal at the time of analysis. We also found that the progeny of a single metastatic clone could eventually overgrow the primary tumour. Thus malignant (metastatic) cells may manifest a striking growth advantage within the primary tumour site as well as for dissemination and growth at distant, secondary sites. As a result, late-stage advanced primary tumours, if left intact, may evolve to become biologically similar or equivalent to distant metastases. This 'clonal dominance' phenomenon can reconcile many of the discrepant experimental findings with respect to the putative selective nature of metastatic phenotype. Furthermore, it has important consequences for understanding one source of biological variability in experiments in which different primary tumours are compared to each other or to metastases; it also has implications for theories regarding the clonal origin of neoplasms, and for the physiological and biochemical changes that cause malignant disease.