According to our theory, the ids do not collectively exert a controlling influence on the cells, not even on the germ-cells, whose histological differentiation into spermatic or egg-cells can only depend on control through specific sex-cell determinants. It is the different determinants of the ids that control; transformations of the species will, it is true, depend on transformations of the ids, but this need not necessarily consist in a variation of _all_ the determinants of the id. If, for instance, two species of butterfly, _Lycæna agestis_ in Germany and _Lycæna artaxerxes_ in Scotland, only differ from each other in that the black spot in the middle of the wing in _L. agestis_ is milk-white in _L. artaxerxes_, no other determinants in the id of the germ-plasm can be different except those which control this particular spot. In a majority of the ids in _L. artaxerxes_ the determinants of this spot must have been modified, let us say, to the production of 'milk-white.' This majority will increase very slowly if the white colour has no pronounced advantage for the persistence of the species, but it will increase gradually, as we have already seen, though extremely slowly, through the elimination of those individuals whose germ-plasm at the reducing division has chanced to receive a majority of ids with the old, unmodified determinants, and which have therefore reverted to the ancestral form. This will happen whenever the new character has any use, however small, in maintaining the species. But in most modifications of species quite a number of parts and characters have undergone variation either simultaneously or in rapid succession; in many cases nearly all the details of structure, and therefore almost all the determinants of the germ-plasm, must have varied. We must not, however, assume that all the equivalent determinants, for instance, all the determinants _K_ in all the ids, have varied[9], and above all we must not take for granted that the determinants of different characters or parts of the body, for instance, the determinants _L_, _M_, or _N_, must all undergo variation in an equal number of ids. It will depend on two factors whether a new character is implicit, in the form of varied determinants, in a small or in a very large majority of ids: first, on the relative age of the character, and, secondly, on its value in relation to the persistence of the species. The more important a character is for the species the more frequently is it decisive for the life or death of the individual, and the more sharply will individuals not possessing it be eliminated, and the more rapidly, therefore, will those whose germ-plasm still contains a majority of the unvaried determinants of this character tend to disappear. In this way these determinants will tend to sink down from generation to generation to an ever smaller minority in the germ-plasm of those that survive.