initial differences are for ever reacting on the environment and on each other, in such a way as to ensure increasing divergence as long as the individuals of the two groups are kept from intergenerating.’” The words of Mr Gulick require close scrutiny. We may admit that “no two portions of a species possess exactly the same average character,” but why should the two, if prevented from interbreeding yet subjected to similar climatic and other conditions, present the phenomenon of “increasing divergence?” The reason assigned by Romanes is the “Law” of Delbœuf, which runs:—“_A constant cause of variation_, however insignificant it may be, changes the uniformity of type little by little, and diversifies it _ad infinitum_.” From this “Law” it follows, says Romanes, on p. 13 of vol. iii. _Darwin and after Darwin_, that “no matter how infinitesimally small the difference may be between the average qualities of an isolated section of a species compared with the average qualities of the rest of that species, if the isolation continues sufficiently long, differentiation of specific type is necessarily bound to ensue.” This deduction involves two important assumptions. The first is, that in each of the separated portions of the given species there is a constant cause of variation operating in one direction in the case of one portion and in another direction in the case of the other. This assumption is, unfortunately, not founded on fact. If we were to take one hundred race-horses and shut them up in one park and one hundred cart-horses and shut them up in another park, and prevent the interbreeding of the two stocks, we should, if Romanes’s tacit assumption be true, see the two types diverge more and more from one another. We know that as a matter of fact they will tend, generation after generation, to become more like one another. Galton’s Law of Regression, of which we have already spoken, and which is supported by ample evidence, clearly negatives this tacit assumption made by Romanes and Gulick. The second assumption upon which their reasoning is based is that there is no limit to the amount of change which can be effected by the accumulation of fluctuating variations; but, as we have already seen (on p. 70), there is a very definite limit and this limit is quickly reached. Thus the arguments of Romanes and Gulick are fundamentally unsound. Mollusca of Sandwich Isles But the fact remains, and has to be accounted for, that, as a general rule, when two portions of a species are separated, so that they are prevented from interbreeding, they begin to diverge in character, and the longer they remain thus separated the greater becomes that divergence. This is an observed fact which cannot be gainsaid.