In contrast with a wide definition of the 'epigenetic variation', including all changes in gene expression that do not result from the alteration of the gene structure, a more restricted class had been defined, initially in plants, under the name 'paramutation'. It corresponds to epigenetic modifications distinct from the regulatory interactions of the cell differentiation pathways, mitotically stable and sexually transmitted with non-Mendelian patterns. This class of epigenetic changes appeared for some time restricted to the plant world, but examples progressively accumulated of epigenetic inheritance in organisms ranging from mice to humans. Occurrence of paramutation in the mouse and possible mechanisms were then established in the paradigmatic case of a mutant phenotype maintained and hereditarily transmitted by wild-type homozygotes. Together with the recent findings in plants indicative of a necessary step of RNA amplification in the reference maize paramutation, the mouse studies point to a new role of RNA, as an inducer and hereditary determinant of epigenetic variation. Given the known presence of a wide range of RNAs in human spermatozoa, as well as a number of unexplained cases of familial disease predisposition and transgenerational maintenance, speculations can be extended to possible roles of RNA-mediated inheritance in human biology and pathology.