[Illustration: FIG. 2. Cross between long and vestigial wings, giving long in F₁. The long-winged hybrid F₁ female is then represented as out-bred to a wild-type male, giving long-winged offspring again—half pure-long and half hybrid-long. The last are represented as again out-bred to wild type, giving long-winged offspring again and of the same two genetic kinds as above.] When the vestigial fly was crossed to normal the mutant character disappeared in the hybrid. If such a hybrid is out-bred to normal all the offspring are again normal, but half of them carry the vestigial gene. If these are out-crossed again still only normal flies appear, Fig. 2. If such out-breeding is continued the vestigial gene will become widely distributed without ever showing itself at the surface, so to speak. If, however, at any time two hybrid flies mate, then a quarter of the offspring will have vestigial wings. It might seem then that the character had appeared for the first time in the race, if one did not know its past. In reality its gene may have been there for some time. Probably many of the recessive defects and malformations that appear in the human race—at least those due to hereditary factors—have had representative genes in the germ-plasm for several generations before they have appeared on the surface. We do not know how widespread recessive genes are in the human germ-plasm. The fact that defective individuals appear in certain communities may be safely interpreted to mean that individuals bearing the same gene have at last come together. On the other hand, the absence of such individuals from the community, at large, may only mean that the chance of suitable combinations is small, and does not mean necessarily that the gene in question is confined to the community within which the defects have been recorded. My illustration may give, however, an entirely erroneous idea as to the chance of a recessive character contaminating the race. If one can control the matings, so that out-breeding takes place each time, the result would undoubtedly be like that in our diagram; but what chance is there for a recessive character, that is neither beneficial nor injurious, if left to itself, to contaminate widely the race with its gene? The answer is that for any one defect there is hardly any chance at all. On the other hand, there is always a possibility that a defect _may_ become widespread despite the chances against each in turn. If a recessive _character_ is selected against each time it appears on the surface, the chance is extraordinarily small that the gene for such a character could ever become widespread in a race. If the recessive character is advantageous, its chance is somewhat better, but still the chance that it may be lost is very great. Let us turn for a moment to the inheritance of a Mendelian dominant character, and to simplify the situation let us first assume that the character itself is neither advantageous nor disadvantageous.