Let it be borne in mind, then (_a_), that inheritance is not only from the immediate parents, but from the whole line of ancestry. The inheritance from the immediate parents is, doubtless, usually greater than from any other _one_ term of the ancestral series--the effect on the offspring of any previous generation becomes, doubtless, less and less as the distance from the offspring increases--yet the _sum_ of the ancestral inheritance is far greater than the immediate parental. Let it also be borne in mind (_b_) that true breeding from one form for many generations creates a fund of heredity in that form, and thus tends to produce fixity, rigidity, or permanence in that form. [Illustration: FIG. 68.] Now, the method of producing artificial breeds, sometimes consciously, sometimes unconsciously, is, briefly, as follows: Suppose it be desired to obtain a variety of an animal, say a dog, having a certain character. We start from a common type, _a_ (Fig. 68). If this type were allowed to breed naturally, the slight divergent variation of offspring represented by the radiating lines would neutralize one another by interbreeding, the individual differences would be “_pooled_” in a common stock, and the species would remain substantially constant. But if among all these slightly divergent varieties we select one, _b_, which seems in the right direction, and ruthlessly destroy all the others (indicated by crossing them out by the circular line), and breed this variety, _b_, only, we shall get again a number of divergent varieties. It may be that the larger number of these will be backward, in the direction of the original type _a_, on account of the ancestral heredity in that direction, but some will again be in the desired direction. Let all the varieties other than the desired one, but especially the backward-going or reverting ones, be again destroyed, and the one kind only selected which seems to be in the right direction, viz., _c_. As we push the form thus from generation to generation in the desired direction, especially if we attempt to hasten too much the process, the resistance to movement--if I may use the expression--in that direction becomes greater and greater (shown by the decreasing distances between the successive points of divergence, _a_, _b_, _c_, _d_, etc.), and the tendency to reversion becomes stronger (shown by the greater number and length of the backward-going lines), until finally it is almost impossible to push any farther. We will suppose that _x_ is such a limit. But if, now, we breed true on the point _x_, destroying the reversions or backward variations for many generations, we will gradually accumulate a fund of ancestral heredity on this point which increases with every added generation, until finally the tendency to reversion becomes small. The variety _breeds true_ without further interference, or with only very general superintendence. Such a permanent variety is called a _race_.