The hypothesis is presented that a ground-level DNA damage is unavoidable and must be regarding constitutive to a cell. The genotoxic agents responsible for this type of initiation comprise a number of physiological chemists which are known or suspected to be degraded via chemically reactive species, nitrosamines present in the diet or formed from amines in the stomach, ubiquitous carcinogenic metals, UV- and alpha-irradiation, viruses, and other sources in diet and environment. 2. It is concluded that what is normally called spontaneous tumor incidence is partly due to this constitutive DNA damage. 3. Under this assumption, carcinogens can be divided into two classes, the initiating (genotoxic, DNA-damaging) carcinogens and the non-genotoxic carcinogens which act by modulating any of a number of reactions and side reactions that lead to an increase of the constitutive DNA damage or enhance the chance for the constitutive DNA damage to proceed to a tumor. 4. The carcinogenic potency of a carcinogen is described as a product of persistent DNA damage x mutagenicity x non-genotoxic modulation of the DNA damage, and short-term tests are proposed to determine the contributions of the first two parameters to the potency of an initiating carcinogen. 5. As an approximation to this theoretical approach, a correlation is shown of DNA binding in vivo in the form of a "Covalent Binding Index" to the carcinogenic potency as derived from long-term bioassays. An astonishingly good linear correlation is found with an approximate uncertainty of the estimate of a factor of 10 with a total span of values of 10(6). 6. It is concluded that DNA binding in vivo provides a useful first look at the potency of initiating carcinogens but that additional knowledge is required to assess the organ specificity of initiating carcinogens. 7. The non-genotoxic carcinogens seem to be much less potent than the initiating carcinogens if administered alone. They cannot, however, be spotted on the basis of chemical structure, and there is no short-term test which would allow a good quantitative approach to their carcinogenic potency as was demonstrated for the initiating carcinogens. One main reason for this lack is the wide variety of different mechanisms of action of non-genotoxic carcinogens.