That natural selection differs from most of the other forms of isolation in not being capable of causing _divergent_ or _polytypic_ evolution must at once become evident, if we remember that the only way in which isolation of any form can cause such evolution is by partitioning a given group of intergenerants into two or more groups, each of which is able to survive as thus separated from the other, and so to carry on the evolution in divergent lines. But the distinguishing peculiarity of natural selection, considered as a form of isolation, is that it effects the isolation _by killing off all the individuals which it fails to isolate_: consequently, this form of isolation differs from other forms in prohibiting the possibility of any ramification of a single group of intergenerants into two or more groups, for the purpose of carrying on the evolution in divergent lines. Therefore, under this form of isolation alone, evolution must proceed, palm-like, in a single line of growth. So to speak, the successive generations continuously ascend to higher things on the steps supplied by their own "dead selves"; but in doing so they must climb a single ladder, no rung of which can be allowed to bifurcate in the presence of the uniformity secured _for that generation_ by the free intercrossing of the most fit. Even though beneficial variations may arise in two or more directions simultaneously, and all be simultaneously selected by survival of the fittest, the effect of free intercrossing (in the absence of any other form of isolation) will be to fuse all these beneficial variations into one common type, and so to end in _monotypic_ evolution as before. In order to secure _polytypic_ evolution, intercrossing between the different beneficial variants which may arise must be prevented; and there is nothing to prevent such intercrossing in the process of natural selection _per se_. In order that the original group of intergenerants should be divided and sub-divided into two or more groups of intergenerants, some additional form of isolation must necessarily supervene--when, of course, polytypic evolution will result. And, as Mr. Gulick has shown, the conclusion thus established by deductive reasoning is verified inductively by the facts of geographical distribution.