The preferential retention of paternal tumor suppressor alleles in sporadic tumors and the failure to demonstrate genetic linkage between disease predisposition and tumor suppressor loci in familial cases indicates that genome imprinting may be involved in the genesis of some pediatric cancers. A genetic model that invokes the activity of modifier loci (imprinting genes) on alleles to be modified (imprinted genes) is able to account for these data. Genome imprinting may be viewed as a special case of dominance modification, differing from other examples only in that the modification of dominance is dependent on gamete-of-origin. Data from human pediatric tumors, transgenes in the mouse and variegating position-effects in Drosophila, indicate that the net effect of modifier loci is the inactivation of alleles at affected loci. Polymorphism at the level of the modifier loci will result in different degrees of modification between individuals. With respect to tumors, the most important mechanism by which these differences are manifested is cellular mosaicism for the expression of a modified allele. Such characteristics are reminiscent of the behavior of variegating position-effects in Drosophila and the application of this paradigm to human disease phenotypes provides both a mechanism by which differential genome imprinting may be accomplished as well as genetic models that may explain the clinical association of syntenic diseases, the association between tumor progression and specific chromosomal aneuploidy and the unusual inheritance characteristics of many diseases.