It must certainly be admitted that the large proportion of abnormal hexamerous flowers depends upon heredity in the instance first quoted; but the hexamerous condition is not an acquired character; it is merely the first appearance of a new innate character. It is not due to the reaction of the vegetable organism under some external stimulus, for it appeared in a plant exposed to conditions similar to those which acted upon the other plant which only produced the normal pentamerous flowers. It must therefore have resulted from the tendencies which were present in the germ from which the plant itself developed, either as a spontaneous change in the germ-plasm or through the combination of two parental germ-plasms—a combination which may lead to the appearance or the reality of a new character. We know that the germ-plasm of each individual is not a simple substance, but possesses a very complex composition, for it consists of a number of ancestral germ-plasms represented in very different proportions. Now, although we cannot learn anything directly about the processes of growth of the germ-plasm, and its resulting ontogenetic stages, yet we do know, chiefly from observations upon man, that the characters of ancestors appear in the offspring in very different combinations and in very different degrees of strength. This may, perhaps, be explained by assuming that in the union of parental germ-plasms which takes place at fertilization, the contained ancestral germ-plasms unite in different ways, and thus come to grow with different strengths. Certain ancestral germ-plasms will meet and together produce a double effect: other opposed germ-plasms will neutralize each other; and between these two extremes all intermediate conditions will occur. And these combinations will not only take place at fertilization, but also at every stage of the whole ontogenetic history, for each stage is represented by its idioplasm, which is itself composed of ancestral idioplasms. We do not yet know enough to be able to prove in detail the manner in which new characters may arise from such a combination of different kinds of germ-plasm. And yet it appears to me that such a view, e.g. in the case of the variation of buds, is by far the most natural. There is indeed a single example in which we can, to some extent, understand how it is that a new character may arise by these means. Certain canary-birds have a tuft of feathers on the head, but if two such birds are paired, their descendants are generally bare-headed, instead of having larger tufts[226]. The formation of a tuft depends upon the fact that the feathers are scanty and in fact absent from part of the skin of the head. Now when the scanty plumage of both parents is combined in the offspring the latter is bare-headed. Hence by the combination of ancestral characters a new character (bare-headedness) is produced, and one which is hardly likely to have ever occurred in the ancestors of existing canaries.