In the case of ordinary plant-hybrids what are designated reversions can only be called so in a wide sense, for the ancestral characters are contained _visibly_ in the parent, although mingled with those of the other parent. In human families, however, there are undoubted cases in which one or more characters of the grandparent reappear in the child which were not in any visible way expressed in the parent, and must therefore have been contained in the parent's germ-plasm in a latent state. And there are both in animals and plants reversions to ancestors lying much further back, to characters and groups of characters which have not been visible for many generations, and the occurrence of these can only be explained on the assumption that certain groups of ancestral determinants have been carried on in the germ-plasm in too small a number to be able to give rise ordinarily to the relevant character. Such isolated determinants may, however, in certain circumstances be strengthened by the amphimixis of two germ-cells both containing small groups of them, and thus augmented they may gain a controlling influence. In this case the chances of the reducing divisions have a part to play, since they bring together the old unvaried ancestral determinants which, as we have seen, may persist in the germ-plasm of any species through a long series of generations. This would, of course, only suffice to bring about a reversion if the determinants of the ancestral species were still contained in the germ-plasm in comparative abundance. If this is no longer the case, something more is necessary, and that is the relative weakness of the more modern determinants.
If two white-flowered species of thorn-apple, _Datura ferox_ and _Datura lævis_, be crossed, there arises a hybrid with bluish violet flowers and brown stalks instead of green. This was interpreted by Darwin as a reversion to violet-flowering ancestral species on both sides, for there are even now a great number of species of _Datura_ with violet flowers and brown stalks. When the two white forms are crossed the reversion takes place every time, not merely in some cases, and we may conclude from this that in both these species there is still such a strong admixture of the same unvaried ancestral ids that they always excel the ids of the two modern species crossed, in strength though certainly not in number. And this superiority must again depend on the fact that similar determinants of the same part are cumulative in their effect, while dissimilars are not.