Molecular epidemiologists usually consider the spectrum of p53 mutations found in human tumors to be a signature of the corresponding environmental carcinogen(s). In lung cancer, this signature is the spectrum of G --> T transversions, presumably induced by polycyclic aromatic hydrocarbons (PAH) from cigarette smoke. What complicates the situation, however, is that in the p53 gene the same codons are preferential targets for not only mutagenesis but also tumorigenic selection. In this review, we compare the G --> T spectra induced by PAH o-quinones and diol epoxides with those in lung cancer and show that the main "shaper" of the latter is selection, not mutagenesis. In addition, we propose the approach that allows to distinguish selection and mutagenesis components of the p53 spectra and, therefore, to test the suspect carcinogens for their "in vivo" mutagenic involvement. Collectively, the reviewed basic premises, concepts and data are consistent with the increasing recognition of environmental cancer risk conditions as selecting rather than inducing tumorigenic mutations.