A simple animal model of tobacco smoke carcinogenesis works as follows: Strain A/J mice are exposed for 5 months to tobacco smoke. They are then given a 4-month recovery period in air before being killed. Lung surface tumors are counted and lung tumor multiplicity (average number of tumors per lung, including non-tumor bearing animals) is calculated. Results obtained in four different laboratories during the past 8 years have consistently shown significant increases in lung tumor multiplicities in tobacco smoke exposed animals. While inhaling to tobacco smoke, strain A mice (but not some other strains) fail to gain weight and immediately after smoke exposure only have about 75% of control weight; however, when removed into air, they regain weight rapidly up to control levels. The counting of surface tumors only may occasionally underestimate total number of lung tumors and thus yield false negatives. At the end of the experiment, the mice are 1-year old and about 80% of the tumors are adenomas, the remainder adenomas with carcinomatous foci or adenocarcinomas. Tobacco smoke does not increase the percentage of adenocarcinomas. Studies with filtered tobacco smoke have suggested that benzo(a)pyrene or tobacco smoke-specific nitrosamines cannot account for lung carcinogenesis in mice; the most likely single agent to cause lung tumors is 1,3-butadiene. A major disadvantage of the assay is its low statistical power. While it is easy to detect a 70-100% decrease in lung tumor multiplicity caused by a chemopreventive agent using group sizes of 20-30 animals, the detection of smaller reductions (20-50%) would require group sizes in the hundreds. From all available evidence it must be concluded that the complex mixture of tobacco smoke, a known human carcinogen, is a rather weak rodent carcinogen.