The stability of the human genome requires that mutations in the germ line be exceptionally rare events. While most mutations are neutral or have deleterious effects, a limited number of mutations are required for adaptation to environmental changes. Drake has provided evidence that DNA-based microbes have evolved a mechanism to yield a common spontaneous mutation rate of approximately 0.003 mutations per genome per replication (Drake 1991). In contrast, mutation rates of RNA viruses are much larger (Holland et al. 1982) and can approach the maximum tolerable deleterious mutation rate of one per genome (Eigen and Schuster 1977; Eigen 1993). Drake calculates that lytic RNA viruses display spontaneous mutation rates of approximately one per genome while most have mutation rates that are approximately 0.1 per genome (Drake 1993). This constancy of germline mutation rates among microbial species need not necessarily mean constancy of the somatic mutation rates. Furthermore, there need not be a constant rate for somatic mutations during development. In this review, we consider mutations in cancer, a pathology in which there appears to be an increase in the rate of somatic mutations throughout the genome. Moreover, within the eukaryotic genome, as in microbes, there are "hot-spots" that exhibit unusually high mutation frequencies. It seems conceivable to us that many tumors contain thousands of changes in DNA sequence. The major question is: how do these mutations arise, and how many are rate-limiting for tumor progression?