high variability of F 1 Citrus hybrids is, however, a very general phenomenon, not limited to a minority of the individuals. Further, enough is known, in general, of the production of new characters by new combinations of genes in crossing, to warn us against setting any narrow limits to the probable results of recombination in crosses between two highly heterozygous species. On the other hand, a new warning against undue confidence in gene stability is given by Eyster's (1924) recent hypothesis of qualitative division of certain genes. Possibly genes specially affecting other characters are sometimes unstable in the same way as are certain genes relating to variegation. If Citrus is especially notable for the occurrence of such unstable genes, this fact may account for part of the remarkable variability that is observed. POLYEMBRY•ONY Citrus seeds are frequently polyembryonic. Strasburger (1878, 1907) showed that the supernumerary embryos are formed by proliferation of nueellar cells surrounding the embryo sac. These adventitious embryos may be expected, therefore, to reproduce the seed-parent genotype, without variation due to segregation in sporogenesis or to recombination in fertilization. Citrus polyembryony is not entirely due to nucellar budding, however, for in 10 (probably 11) cases in our cultures, among more than 1000 hybrids, two hybrid seedlings have come from one seed. The seeds were planted separately, and all operations on which the reliability of the pedigrees depended were so carefully performed and checked that the single-seed origin of the pairs of seedlings is beyond doubt. The budding and consequent labeling were done with similar care. In eight of the ten cases, both of the original seedlings, as well as trees budded from them, have been positively classified as hybrids; in the ninth case, both of the small seedling trees are almost surely hybrids, as the budded trees certainly are; and in the tenth case one seedling died undescribed, so that its record depends on budded trees .alone. In the eleventh case, one of the two" seedlings" died young, and their separateness below the surface of the soil was not proved. Since in every case the two hybrids seem to be identical in type, in spite of the usual great diversity among hybrids of the same parentage, it is probable that these are all cases of "identical twins," each