The overwhelming proportion of colorectal carcinomas are believed to originate as adenomatous polyps (adenomas), and the identi"cation and removal of adenomas is an important component of colorectal cancer prevention e!orts. Mathematical modeling of adenomas can increase our understanding of the natural history and biology of adenomas and colorectal cancer and can help in the e!ort to devise optimal prevention and screening strategies. Here we adapt the multi-stage model of carcinogenesis to the problem of the development and growth of adenomas. We show that, using plausible values for the biological parameters, the model can "t various aspects of adenoma data including adenoma prevalence by age, the size distribution of adenomas, clustering of adenomas within individuals and the correlation between distal and proximal adenomas. Explaining the clustering of adenomas within individuals, as well as other "ndings, requires heterogeneity in risk in the population; we show how such heterogeneity can be related to the distribution of biological parameters in the population. The model can also be adapted to account for adenoma development in two major syndromes related to colorectal cancer, familial adenomatous polyposis and hereditary non-polyposis colorectal cancer.