The preceding argument, it will be seen, depends entirely upon the assumption that some amount of infertility characterises the distinct varieties which are in process of differentiation into species; and it may be objected that of such infertility there is no proof. This is admitted; but it is urged that facts have been adduced which render such infertility probable, at least in some cases, and this is all that is required. It is by no means necessary that _all_ varieties should exhibit incipient infertility, but only, some varieties; for we know that, of the innumerable varieties that occur but few become developed into distinct species, and it may be that the absence of infertility, to obviate the effects of intercrossing, is one of the usual causes of their failure. All I have attempted to show is, that _when_ incipient infertility does occur in correlation with other varietal differences, that infertility can be, and in fact must be, increased by natural selection; and this, it appears to me, is a decided step in advance in the solution of the problem.[62] _Physiological Selection._ Another form of infertility has been suggested by Professor G.J. Romanes as having aided in bringing about the characteristic infertility or sterility of hybrids. It is founded on the fact, already noticed, that certain individuals of some species possess what may be termed selective sterility--that is, while fertile with some individuals of the species they are sterile with others, and this altogether independently of any differences of form, colour, or structure. The phenomenon, in the only form in which it has been observed, is that of "infertility or absolute sterility between two individuals, each of which is perfectly fertile with all other individuals;" but Mr. Romanes thinks that "it would not be nearly so remarkable, or physiologically improbable, that such incompatibility should run through a whole race or strain."[63] Admitting that this may be so, though we have at present no evidence whatever in support of it, it remains to be considered whether such physiological varieties could maintain themselves, or whether, as in the cases of sporadic infertility already discussed, they would necessarily die out unless correlated with useful characters. Mr. Romanes thinks that they would persist, and urges that "whenever this one kind of variation occurs _it cannot escape the preserving agency_ of physiological selection. Hence, even if it be granted that the variation which affects the reproductive system in this particular way is a variation of comparatively rare occurrence, still, as _it must always be preserved_ whenever it does occur, its influence in the manufacture of specific types _must be cumulative_." The very positive statements which I have italicised would lead most readers to believe that the alleged fact had been demonstrated by a careful working out of the process in some definite supposed cases. This, however, has nowhere been done in Mr.