That artificial selection can cause the increase of a part has long been established, but in what way this is possible, and how it can be theoretically explained has hitherto been very obscure, for even if we take the favourable case that both parents possess the desired variation, it cannot be supposed that the characters of the parents are, so to speak, added together in the child; all we can say is that the probability that the children will also exhibit the character in question--for instance, a long or crooked nose--becomes greater. Certainly an increase of the character may result if in both parents the determinants _K_ are present in excess as compared with the heterodynamous determinants _K´_ and _K´´_, for in that case there is an increased probability that, through reducing divisions and amphimixis, there will again be a preponderance of the determinants _K_ composing the germ-plasm of the child, and further, that these determinants _K_ will dominate strongly as compared with the few _K´_'s. It may thus happen that the long nose of the two parents will give rise to a still longer nose in the child, or that parents of considerable bodily size may have still bigger children, but such increase would be confined to one generation, and would not lead to a permanent increase of the character; permanent increase cannot depend merely on the number of the determinants _K_ and on their supremacy over their converse, the determinants _K´_; it must also depend on their own variation, and this again can depend only on germinal selection and not upon personal selection, although the former can be materially assisted by the latter.