2. The second difficulty is as follows: Such commencing differences as constitute varieties and species not only would not be preserved and integrated by natural selection unless useful, but would immediately be _swamped by cross-breeding_ with the parental form. But, as the whole divergence commences in varieties, evidently it could not commence at all unless this cross-breeding be in some way prevented. This may, indeed, be done, without the assumption of any new factor of evolution, by _migration_; and, hence, migration must be regarded as an important agent in the creation of new forms, not only by the effect of a new environment, but also by prevention of the swamping of commencing species by cross-breeding with the parental form; but in a crowded locality, without outlet for migration (the very conditions most favorable for severe competitive struggle, and therefore for most potent operation of natural selection; and therefore, also, according to Darwin, for profuse diversification), commencing varieties could not pass into species, because swamped by cross-breeding. Once the divergence reaches the point of cross-sterility--i. e., of species--then, indeed, by true breeding, characters, even though not useful, may be preserved. But how is it to commence? This difficulty has been severely felt by all Darwinists. It seems to us that it is largely met by Dr. Romanes. According to Romanes, no organ is so subject to varietal changes as the _reproductive_, and these in no respect so much as in degrees of fertility. Unfortunately, these changes are not visible, and must be judged of only by the results. It is not uncommon, for example, to find sterility between individuals (sexual incompatibility) who are both of them perfectly fertile with other individuals. Similarly, cross-sterility, partial or complete, is not uncommon between varieties or races, as Mr. Darwin has long ago noticed. It very generally, as we know, occurs between, and, in fact, is constantly used as a test of, species. Now, this cross-sterility with parent stock, which we find so constant a character of species, and which, therefore, must _have commenced as a partial cross-sterility_ in varieties, is it _antecedent or consequent to other variations_? It has been usual to suppose it consequent to a certain amount of divergence, viz., that which constitutes, or at least approaches, species. But, according to Romanes, it is _antecedent_. Among many other variations, this is that one which originates species, because it prevents reversion by cross-breeding with the parent stock, and insures true breeding with its own kind. In a word, it sexually isolates the species. Suppose, then, a species multiplying indefinitely in one locality: trivial variations of many kinds, and in many directions, occur among the offspring. These are merged by cross-breeding into the original type, which, therefore, remains unchanged. But, from time to time, among these variations