Much recent attention has been paid to the important role of the DNA mismatch repair system in controlling the accumulation of somatic mutations in human tissues and the association of mismatch repair deficiency with carcinogenesis. In the absence of an intact mismatch repair system, cells accumulate mutations at a rate some 1000 times faster than normal cells, and this mutator phenotype is easily measured by the detection of the formation of new variant alleles at microsatellite loci. However, the mismatch repair system is not 100% efficient, even when intact, and the pattern of microsatellite alterations in a wide variety of tumors is consistent with these being due to clonal amplification from tissues that are genetically heterogeneous at microsatellite loci rather than mismatch repair deficiency in the tumor itself. On this basis, it can be estimated that the mutation frequency of microsatellites in normal human tissues is approximately 10(-2) per locus per cell. Similarly, a frequency of mutation at minisatellite loci in normal tissues of around 10(-1) per locus per cell can be estimated. Such elevated levels of mutation are consistent with a recent study of the frequency of HPRT mutation in human kidneys that demonstrated these to be frequent (average 2.5 x 10(-4) in individuals of 70 years or more) and exponentially related to age. Taken as a whole, the data suggest that somatic mutation in human epithelial cells may be some 10-fold higher than in peripheral blood lymphocytes and that the underlying rate of spontaneous mutation is sufficient to account for a large proportion of human carcinogenesis without the need to evoke either stepwise alteration to a mutator phenotype of clonal expansion at all the mutation steps in carcinogenesis. The exponential increase in mutation frequency with age is predictable on the basis that the mutation rate is controlled at the level of repair and that mutation in genes that affect the efficiency of these processes will gradually increase the underlying rate. In addition, the age relatedness of mutation frequency strongly supports the concept that mutation is cell division dependent and that cellular proliferation per se is an important risk factor for cancer. Comparison of somatic mutations with those in the human germline mutation suggests common mechanistic origins and that the high levels of somatic mutation that occur are a direct reflection of the germline mutation rate selected over evolutionary time. Thus, the somatic accumulation of mutations can be seen as a natural process within the human body and cancer a normal part of the human life cycle. This point of view may explain why it has been so difficult to significantly reduce cancer incidence and suggests that, for this to be achieved, the means of altering the natural somatic mutation rate needs to be identified.