Wilms' tumor has served as an example of Knudson's two-hit hypothesis of recessive tumor genes, but the genetics has proven to be surprisingly complex. WT1, a tumor suppressor gene on 11p13, is mutated in only a small fraction of Wilms' tumors, and a second chromosomal region, 11p15, harbors a second Wilms' tumor gene also involved in other cancers. In addition, loss of genomic imprinting, or parental origin-specific gene expression of at least two genes, appears to be an early step in Wilms' tumorigenesis and common cancers. Finally, genes on other chromosomes also play a role. I propose a model of Wilms' tumorigenesis in which multiple genetic alterations act within a specific developmental context, accounting for the epidemiological and pathological heterogeneity of Wilms' tumor, as well as the tissue specificity of the tumor types arising from alterations in these genes.