Investigation of covalent DNA binding in vivo provided evidence for whether a test substance can be activated to metabolites able to reach and react with DNA in an intact organism. For a comparison of DNA binding potencies of various compounds tested under different conditions, a normalization of the DNA lesion with respect to the dose is useful. A covalent binding index, CBI = (mumol chemical bound per mol DNA nucleotide)/(mmol chemical administered per kg body weight) can be determined for each compound. Whether covalent DNA binding results in tumor formation is dependent upon additional factors specific to the cell type. Thus far, all compounds which bind covalently to liver DNA in vivo have also proven to be carcinogenic in a long-term study, although the liver was not necessarily the target organ for tumor growth. With appropriate techniques, DNA binding can be determined in a dose range which may be many orders of magnitude below the dose levels required for significant tumor induction in a long-term bioassay. Rat liver DNA binding was proportional to the dose of aflatoxin B1 after oral administration of a dose between 100 micrograms/kg and 1 ng/kg. The lowest dose was in the range of general human daily exposures. Demonstration of a lack of liver DNA binding (CBI less than 0.1) in vivo for a carcinogenic, nonmutagenic compound is a strong indication for an indirect mechanism of carcinogenic action. Carcinogens of this class do not directly produce a change in gene structure or function but disturb a critical biochemical control mechanism, such as protection from oxygen radicals, control of cell division, etc. Ultimately, genetic changes are produced indirectly or accumulate from endogenous genotoxic agents. The question of why compounds which act via indirect mechanisms are more likely to exhibit a nonlinear range in the dose-response curve as opposed to the directly genotoxic agents or processes is discussed.