It is an almost consensus opinion that the major carcinogenic risk of tobacco smoke is in its direct mutagenic action on DNA of cancer-related genes. The key data supposedly linking smoke-induced mutations to lung cancer were obtained from the adduct spectrum of the p53 tumor suppressor gene. Results of our analysis of p53 mutations compiled from the International Agency for Research on Cancer p53 database (April 1999 update) and from the literature point to a different causative link. Our new analytical tests focused on complementary base substitutions and showed that it is strand-specific repair of primary lesions and site-specific selection of the resultant mutations that determine the lung cancer-specific hot spots of G:C to T:A transversions along the p53 gene and also their increased abundance in lung tissues as compared with smoke-inaccessible tissues. However, on each of the two strands of p53 DNA, our tests revealed no significant difference between smokers and nonsmokers, either in the frequency of different types of mutations or in the frequency of their occurrence along the p53 gene. Moreover, in both smokers and nonsmokers, there was the same frequency of lung tumors with silent p53 mutations. Accordingly, we offer here a selection-based explanation of why lung cancers with nonsilent p53 mutations are more common in smokers than in nonsmokers. We conclude that physiological stresses (not necessarily genotoxic) aggravated by smoking are the leading risk factor in the p53-associated etiology of lung cancer.