There is increasing evidence that most human cancers contain multiple mutations. By the time a tumor is clinically detectable it may have accumulated tens of thousands of mutations. In normal cells, mutations are rare events occurring at a rate of 10(-10) mutations per nucleotide per cell per generation. We have argued that the mutation rates exhibited by normal human cells are insufficient to account for the large number of mutations found in human cancers, and therefore, that an early event in tumorigenesis is the development of a mutator phenotype. In normal cells, spontaneous and induced DNA damage is balanced by multiple pathways for DNA repair, and most DNA damage is repaired without error. However, in tumor cells this balance may be shifted such that damage overwhelms the repair capacity, resulting in the accumulation of multiple mutations. Our hypothesis is that multiple random mutations occur during carcinogenesis. The sequential mutations that are observed in some human tumors result from selective events required for tumor progression. We consider the possibility that endogenous sources of DNA damage, in particular oxidative DNA damage, may contribute to genomic instability and to a mutator phenotype in some tumors. Endogenous and environmental sources of reactive oxygen species (ROS) are abundant. In tumor cells, antioxidant or DNA repair capacity may be insufficient to compensate for the production of ROS, and these endogenous ROS may be capable of damaging DNA and inducing mutations in critical DNA stability genes. The possibility that oxidative DNA damage could be a significant source of the genomic instability characteristic of human cancers is exciting, because it may be feasible to modulate the extent of oxidative damage through antioxidant therapy. The use of antioxidants to reduce the extent of molecular damage by ROS could delay the progression of cancer.