Quantitative hemagglutination studies of the MN-hemizygous (M/-) patient and his family reported by German et al. are given together with data on the electrophoretic mobility of their red cells. These results, and those obtained on the cells of a donor of the MU phenotype (MU = M+N-S-s-U+; Mu = M+N-S-s-U-), demonstrate a series of shortcomings in the current ‘precursor transferase’ theory of the biochemical genetics of MN antigens. Another theory is proposed, according to which the effects of the MN genes take place exclusively in the protein part of the glycopeptide. The MN proteins would carry acceptor sites for the antigenic oligosaccharides which are put together by enzymes genetically independent of MN. In M glycoproteins, the acceptor sites are close to each other, in doublets, while in N they are all separate. This model is shown to apply successfully to several difficult problems in MN.