Animal models of carcinogenesis have contributed substan tially to our understanding of neoplastic and preneoplastic events. Typically, laboratory experiments seek specific infor mation on suspected carcinogenic agents and for that reason usually focus on one agent at a time. To test the suspect agent, an adequate yield of neoplastic lesions is desirable, usually several orders of magnitude greater than the frequency observed in humans; that requires the administration of doses many times greater than those prevailing in most human situations. The high yield also requires that animal strains be selected which have a proven susceptibility to the compound, thus concentrating on genetically homogeneous (often inbred) pop ulations of experimental animals. These experimental require ments contrast with most human situations in which mixed, low-dose, carcinogenic influences act on genetically heteroge neous populations. This high-yield high-dose requirement also restricts the study of the precancerous process since it is pur posefully accelerated and, therefore, provides insufficient time and opportunity to observe the precursor stages which may be better expressed in humans. In animal models some of the precancerous changes are observed only when lower doses of weaker carcinogens are used (1). The study of human models of carcinogenesis has begun in recent years. Of necessity, they are based on epidemiological and laboratory observations under circumstances which cannot match the rigorous specifications of experimental designs. However, recent progress in techniques applicable to the human studies now make it possible to test specific links in the chain of events leading to neoplastic transformation of human tissues. To illustrate these points, a proposed model for gastric carci nogenesis in humans will be presented, based on epidemiolog ical, pathological, and clinical observations.