This review analyzes the concept and evidence in support of a mutator phenotype in human cancer. The large number of mutations reported in tumor cells cannot be accounted for by the low mutation rates observed in normal somatic cells; rather, it must be a manifestation of a mutator phenotype present early during the tumorigenic process. The interaction between increased mutagenesis and clonal selection provides a mechanism for the selection of cells with increased proliferative advantage. The concept of a mutator phenotype in cancer has gained considerable support from the findings of enormous numbers of somatic mutations in repetitive sequences in human tumors. Moreover, cell lines exhibiting microsatellite instability demonstrate an increased mutation frequency in expressed genes. A knowledge of mechanisms that generate multiple mutations in cancer cells has important implications for prevention. For many tumors, a delay in the rate of accumulation of mutations by a factor of two could drastically reduce the death rates from these tumors.