may be shown that natural selection
_is_, in some probable cases at all events, able to accumulate
variations in infertility between incipient species.

The simplest case to consider, will be that in which two forms or
varieties of a species, occupying an extensive area, are in process of
adaptation to somewhat different modes of life within the same area. If
these two forms freely intercross with each other, and produce mongrel
offspring which are quite fertile _inter se_, then the further
differentiation of the forms into two distinct species will be retarded,
or perhaps entirely prevented; for the offspring of the crossed unions
will be, perhaps, more vigorous on account of the cross, although less
perfectly adapted to the conditions of existence than either of the pure
breeds; and this would certainly establish a powerful antagonistic
influence to the further differentiation of the two forms.

Now, let us suppose that a partial sterility of the hybrids between the
two forms arises, in correlation with the different modes of life and
the slight external or internal peculiarities that exist between them,
both of which we have seen to be real causes of infertility. The result
will be that, even if the hybrids between the two forms are still freely
produced, these hybrids will not themselves increase so rapidly as the
two pure forms; and as these latter are, by the terms of the problem,
better suited to their conditions of life than are the hybrids between
them, they will not only increase more rapidly, but will also tend to
supplant the hybrids altogether whenever the struggle for existence
becomes exceptionally severe. Thus, the more complete the sterility of
the hybrids the more rapidly will they die out and leave the two parent
forms pure. Hence it will follow that, if there is greater infertility
between the two forms in one part of the area than the other, these
forms will be kept more pure wherever this greater infertility prevails,
will therefore have an advantage at each recurring period of severe
struggle for existence, and will thus ultimately supplant the less
infertile or completely fertile forms that may exist in other portions
of the area. It thus appears that, in such a case as here supposed,
natural selection would preserve those portions of the two breeds which
were most infertile with each other, or whose hybrid offspring were most
infertile; and would, therefore, if variations in fertility continued