evolution: it may _change_ the type to any extent during successive generations, and in a single line of change; but it cannot _branch_ the type, unless some other form of homogamy intervenes. Therefore, when Mr. Wallace adduces the well-known fact that all structures vary more or less round a specific mean as proof that natural selection need not be incommoded by free intercrossing, but can of itself produce all the known phenomena of specific evolution, he fails to perceive that his argument refers only to one aspect of such evolution (viz. the transformation of species in time), and does not apply to the aspect with which alone my paper on _Physiological Selection_ was concerned (viz. the multiplication of species in space). The same thing may be shown in this way. It is perfectly obvious that where the improvement of type in a linear series is concerned (monotypic evolution), free intercrossing, far from being a hindrance to the process, _is the very means by which the process is accomplished_. Improvement here ascends by successive steps, in successive generations, simply _because_ of the general intercrossing of the generally most fit with the result that the species, _as a whole_, gradually becomes transformed into another species, _as a whole_. Therefore, it would be mere fatuity in any one to adduce free intercrossing as a "difficulty" against natural selection alone being competent to produce evolution of this kind. But where the kind of evolution is that whereby the species is _differentiated_--where it is required, for instance, to produce different structures in different portions of the species, such as the commencement of a fighting spur on the wing of a duck, or _novel_ characters of any sort in different groups of the species--free intercrossing is no longer a condition to, but an absolute preventive of, the process; and, therefore, unless checked as between each portion of the species by some form of homogamy other than natural selection, it must effectually inhibit any _segregation_ of specific types, or divergence of character. Hence it is that, while no Darwinian has ever questioned the power of unaided selection to cause _improvement of character in successive generations_, in common now with not a few other Darwinians I have emphatically denied so much as the abstract possibility of selection alone causing a _divergence of character in two or more simultaneous lines of change_.