Résumé : Mosaicism in 4-8-cell human embryos analysed by fluorescence in situ hybridisation (FISH) has been widely reported, but few studies have addressed the incidence of mosaicism in more advanced embryonic stages. In the present study we analysed spare human embryos in a case of preimplantation genetic diagnosis (PGD) for increased risk of aneuploidy because of an infertile 47,XYY man. After replacement of two embryos typed as 1818XX at PGD, six spare embryos (not frozen because of their low quality) were re-analysed on Day 5 for PGD confirmation. Out of five embryos typed as 1818XY at PGD, four were diploid mosaic (DM) and one was normal in all cells. The sixth embryo, typed as 18XYY/1818181818X at PGD, was a DM. In spite of the bias of our small series of morphologically low-quality embryos, the surprisingly high proportion of mosaics (which confirms previous findings) questions the validity of PGD, but supports the strategy of transferring only the embryos where two blastomeres gave normal and concordant results at PGD. More data are required to understand the clinical significance of early diploid mosaicism (and its impact on implantation rate) and to determine whether some diploid mosaic embryos might be considered safe for transfer.