variations will give a new mean nearer the plus or advantageous end of the scale than the old mean. By how much the favourable variations tend to raise the mean standard, by so much will the race tend to advance. But in this process I see no reason why the neutral variations should be eliminated, except in so far as, in the keen struggle for existence, they become relatively unfavourable.

It is clear, however, that the intercrossing and interbreeding which occurs between average individuals on the one hand, and those possessing favourable variations on the other, while it tends gradually to raise the mean standard, tends also at the same time to reduce the advantageous variations towards the mean. It must tend to check advance by leaps and bounds, and to justify the adage, Natura nil facit per saltum. At the same time, it will probably have a greater tendency to reduce to a mean level neutral variations indefinite in direction than advantageous variations definite in direction. Still, it is a most important factor, and one not to be neglected. It tends to uniformity in the species, and checks individualism. It may act as a salutary brake on what we may figuratively term hasty and ill-advised attempts at progress. And at the same time, it favours repeated new experiments in the combination of variations, occasionally, we may suppose, with happy results.

But it does more than this. It tends to check, and, if the offspring always possessed the blended character of both parents, would be absolutely fatal to, divergence of character within the interbreeding members of a species. And yet no fact is more striking than this divergence of character. It is seen in the diversified products of human selection; for example, among pigeons. It is seen in the freedom of nature. Mr. Wallace gives many examples. "Among our native species," he says, "we see it well marked in the different species of titmice, pipits, and chats. The great titmouse, by its larger size and stronger bill, is adapted to feed on larger insects, and is even said sometimes to kill

H

small and weak birds. The smaller and weaker coal-titmouse has adopted a more vegetarian diet, eating seeds as well as insects, and feeding on the ground as well as among trees. The delicate little blue titmouse, with its very small bill, feeds on the minutest insects and grubs, which it extracts from crevices of bark and from the buds of fruit trees. The marsh-titmouse, again, has received its name from the low and marshy localities it frequents; while the crested titmouse is a Northern bird, frequenting especially pine forests, on the seeds of which trees it partially feeds. Then, again, our three common pipits-the tree-pipit, the meadow-pipit, and the rock-pipit, or sea-lark—have each occupied a distinct place in nature, to which they have become specially adapted, as indicated by the different form and size of the hind toe and claw in each species. So the stone-chat, the whin-chat, and the wheat-ear are all slightly divergent forms of one type, with modifications in the shape of the wing, feet, and bill adapting them to slightly different modes of life."* There is scarcely a genus that does not afford examples of divergent species. The question then naturally occurs-How have these divergent forms escaped the swamping effects of intercrossing?

That perfectly free intercrossing, between any or all of the individuals of a given group of animals, is, so long as the characters of the parents are blended in the offspring, fatal to divergence of character, is undeniable. Through the elimination of less favourable variations, the swiftness, strength, and cunning of a race may be gradually improved. But no form of elimination can possibly differentiate the group into swift, strong, and cunning varieties, distinct from each other, so long as all three varieties freely interbreed, and the characters of the parents blend in the offspring. Elimination may and does give rise to progress in any given group as a group; it does not and cannot give rise to differentiation and divergence, so long as interbreeding with consequent interblending of characters be freely

#66 Darwinism," p. 108.

permitted. Whence it inevitably follows, as a matter of simple logic, that where divergence has occurred, intercrossing and interblending must in some way have been lessened or prevented.

Thus a new factor is introduced, that of isolation, or segregation. And there is no questioning the fact that it is of great importance.* Its importance can, indeed, only be denied by denying the swamping effects of intercrossing, and such denial implies the tacit assumption that interbreeding and interblending are held in check by some form of segregation. The isolation explicitly denied is implicitly assumed.

There are several ways in which isolation, or segregation, may be effected. Isolation by geographical barriers is the most obvious. A stretch of water, a mountain ridge, a strip of desert land, may completely, or to a large extent, prevent any intercrossing between members of a species on either side of the barrier. The animals which inhabit the several islands of the Galapagos Archipelago are closely allied, but each island has its particular species or wellmarked varieties. Intercrossing between the several varieties on the different islands is prevented, and divergence is thus rendered possible and proceeds unchecked. It is said that in the Zuyder Zee a new variety of herrings, the fry of which are very small compared with open-sea herrings, is being developed. And the salmon introduced into Tasmania seem to be developing a fresh variety with spots on the dorsal fin and a tinge of yellow on the adipose fin. In the wooded valleys of the Sandwich Islands there are allied but distinct species of land-shells. The valleys that are nearest each other furnish the most nearly related forms, and the degree of divergence is roughly measured by the number of miles by which they are separated. Here there is little or no intercrossing between

Its importance in artificial selection was emphasized by Darwin: "The prevention of free crossing, and the intentional matching of individual animals, are the corner-stones of the breeder's art" ("Animals and Plants under Domestication," ii. 62).

the slow-moving molluses in adjoining valleys; none at all between those at any distance apart.

But even if there are no well-marked physical barriers, the members of a species on a continent or large island tend to fall into local groups, between which, unless the animal be of a widely ranging habit, there will be little intercrossing. Hence local varieties are apt to occur, and varieties show the first beginnings of that divergence which, if carried further and more deeply ingrained, results in the differentiation of species. Geographically, therefore, we may have either complete isolation or local segregation, and in both cases the possibility of divergence.

Another mode of segregation arises also out of geographical conditions. If variations of habits occur (and structure is closely correlated with habit) such that certain individuals take to the mountains, others to the plains or valleys; or that certain individuals take to the forests, others to the open country; the probabilities are that the forest forms will interbreed frequently with each other, but seldom with those in the open, and so with the other varieties. The conditions of forest life or mountain life being thus similar throughout a large area, and life being through elimination slowly but surely adapted to its environment, there might thus arise two distinct varieties scattered throughout the length and breadth of the area, the one inhabiting the mountains, the other the forests. In illustration of this mode of segregation, we may take the case of two species of rats which have recently been found by Mr. C. M. Woodford on one of the Solomon Islands. These two quite distinct species are regarded by Mr. Oldfield Thomas as slightly modified descendants of one parent species, the modifications resulting from the fact that of this original species some individuals have adopted a terrestrial, others an arboreal life, and their respective descendants have been modified accordingly. Thus Mus rex lives in trees, has broad foot-pads, and a long rasp-like, probably semi-prehensile, tail; while Mus imperator lives on the ground, has smaller pads, and a

short, smooth tail. The segregation of these two species has probably been effected by the difference of their mode of life, and each has been adapted to its special environment through the elimination of those individuals which were not in harmony with the condition of their life. It is probable that this mode of segregation has been an important one. And it is clear that in many cases competition would be a co-operating factor in this process, weaker organisms being forced into otherwise uncongenial habitats through the stress of competitive elimination, the weaker forms not perishing, but being eliminated from more favoured areas.

Protective coloration may also be a means of segregation. A species of insects having no protective resemblance might vary in two directions-in the direction of green tints, assimilating their hue to that of vegetation; and in the direction of sandy or dull earthy colours, assimilating them to the colour of the soil. In the one variety elimination would weed out all but the green forms, and these would be left to intercross. In the other variety, green forms would be eliminated, dull-brown forms being left to interbreed. Stragglers from one group into the other would stand a chance of elimination before interbreeding was effected.*

In the case of birds whose freedom of flight gives them a wide range, sometimes almost a world-wide range, it would seem at first sight that their facilities for interbreeding and intercrossing are so great that divergence is well-nigh impossible. And yet the examples of divergence I cited from Mr. Wallace were taken from birds, and it is well known that divergence is particularly well shown in this class. But when the habits of birds are studied attentively, it is found that, wide as is their range, their breeding area is often markedly restricted. The sanderling and knot

* From the absence of interblending in some cases (to be considered shortly), both brown and green forms may be produced; and under certain circumstances, even a power of becoming either brown or green in the presence of appropriate stimuli.