The retention of dormant gemmules, and their passage from generation to generation until their development, may seem improbable; but is it more so than the _fact_ which their presence would explain--viz. the transmission of latent structures and their ultimate reappearance? The development of the whole plant from a Begonia leaf implies that these gemmules are very widely distributed through the tissues. The elective affinity of the gemmules for the cells which precede them in growth may be paralleled by the affinity of the male and female germ-cells, as we see in the preference of a plant for the pollen grains of its own over those of closely-allied species, or by the attraction of the minute germs of disease to certain tissues of the body. It is possible that the numerous gemmules thrown off by the cells of a complex structure, such as a feather, “may be aggregated into a compound gemmule.” In the case of a petal, however, where parts as well as the whole are apt to develop, as is seen in the case of “stripes of the calyx assuming the colour and texture of the corolla,” it is more probable that the gemmules are separate and free. The cell itself is a complex structure, and we do not know whether its separate parts are not developed from the separate gemmules of an aggregate. Such an hypothesis explains the fundamental similarity which has already been shown to exist between all modes of reproduction. The gemmules collected in bud or germ-cell are essentially similar; and were it not for the special advantages of sexual reproduction (increased vigour and more marked variation of offspring), we can well believe that it would have been much less general. The formation of graft-hybrids, and the action of the male element on the mother and on future offspring, become intelligible. The antagonism between growth and sexual reproduction in animals, and between increase by buds, etc., and seeds in plants, can be understood by the use of gemmules in one direction preventing their simultaneous use in another.