Models of human carcinogenesis, such as the multi-stage model of Armitage and Doll, are designed to explain the age increase in the incidence of cancers in individuals. As a consequence, analyses of population level age-specific death rates via such models are appropriately applied to cohort data where such data are available. In this study a multi-stage model is applied to cohort data for stomach cancer death rates in the U.S. population for nine distinct cohorts observed over a recent 28-year period (1950-1977). The multi-stage model parameters obtained from the analysis of the cohort data show significant differences from the parameters obtained from analyses of cross-sectional mortality data under the assumption of no cohort differences in age-specific stomach cancer death rates.