De Vries regards it as a mistake to believe that artificial selection, persevered in for a long time, will succeed in producing a breed which will—as he expresses it—be independent of further selection and will maintain itself in purity. Experience cannot decide this, as we have not command over the unlimited time necessary for selection, but theoretically it is quite intelligible that a variation which had arisen through selection would be more apt to breed true the longer selection was practised, and there is nothing to prevent it becoming ultimately quite as constant as a natural species. For, at the beginning of breeding, we must assume that the variation is contained in only a small number of ids; as the number of generations mounts up, more and more numerous ids with this variation will go to make up the germ-plasm, and the more the breed-ids preponderate the less likelihood will there be that a reversion to the parent-form will be brought about by the chances of reducing-division and amphimixis. That most if not all breeds of pigeon still contain ids of the ancestral form in the germ-plasm, although probably only a small number of them, we see from the occasional reversion to the rock-dove which occurs when species are repeatedly crossed, but that ancestral ids may also be contained in the germ-plasm of long-established natural species is shown by the occurrence of zebra-striping in horse-hybrids. We can understand why these ancestral ids should not have been removed long ago from the germ-plasm by natural selection, since they are not injurious and may remain, so to speak, undetected. It is only when they have an injurious effect by endangering the purity of the new species-type that they can and must be eliminated by natural selection, and this does not cease to operate, as the human breeder does, but continues without pause or break.
I therefore regard it as a mistake on the part of De Vries to exclude fluctuating variation from a share in the transformation of organisms. Indeed, I believe that it plays the largest part, because adaptations cannot arise from mutations, or can only do so exceptionally, and because whole families, orders, and even classes are based on adaptations, especially as regards their chief characters. I need only recall the various families of parasitic Crustaceans, the Cetaceans, the birds, and the bats. None of these groups can have arisen through saltatory, perhaps even retrogressive, 'mutation': they can only have arisen through variation in a definite direction, which we can think of only as due to the selection of the fluctuations of the determinants of the germ-plasm which are continually presenting themselves.
The difference between 'fluctuating' variability and 'mutation' seems to me to lie in this: that the former has always its basis only in a small majority of the ids of the germ-plasm, while the mutation must be present in most of the ids if it is to be stably transmitted from the very first. How that comes about we cannot tell, but we may suppose that similar influences causing variation within the germ-plasm may bring about variation of many ids in the same direction. I need only recall what I have already said as to the origin of saltatory variations, such as the copper-beech and similar cases. The experiments made by De Vries seem to me to give a weighty support to my interpretation of these phenomena. De Vries himself distinguishes a 'pre-mutation period,' just as I have assumed that the variations which spring suddenly into expression have been in course of preparation within the germ-plasm by means of germinal selection for a long time beforehand. At first perhaps only in a few ids, but afterwards in many, a new state of equilibrium of the determinant-system would be established, which would remain invisible until the chance of reducing division and amphimixis gave predominance to a decided majority of the 'mutation-ids.' In the experiments made by De Vries the same seven new 'species' were produced repeatedly and independently of one another in different generations of Œnothera lamarckiana, and we thus see that the same constellations (states of equilibrium) had developed in many specimens of the parent plant, and that it depended on the proportion in which the ids containing these were represented in the seed whether one or another of the new 'species' was produced.
My interpretation, according to which a larger or smaller number of ids were the bearers of the new forms, receives further support from the experiments, for the new species did not always breed true. Thus De Vries found one species, Œnothera scintillans, which only yielded 35-40 per cent. of heirs, or in another group about 70 per cent.; the other descendants belonged to the forms lamarckiana or oblonga, but the number of pure heirs could be increased by selection!
I cannot devote sufficient space to go fully into these very interesting experiments; but one point must still be referred to: the parent form, Œnothera lamarckiana, was very variable from the beginning, that is, it exhibited a high degree of fluctuating variability. This tells in favour, on the one hand, of a deep-rooted connexion between 'variation' and 'mutation,' and, on the other hand, it indicates that saltatory variation may be excited by transference to changed conditions of life—as Darwin in his day supposed, and as I have endeavoured to show in the foregoing discussion. De Vries assumes mutation-periods, I believe rightly; but they are not periods prescribed, so to speak, from within, as those who believe in a 'phyletic force' must suppose; they are caused by the influences of the environment which affect the nutritive stream within the germ-plasm, and which, increasing latently, bring about in part mere variability, in part mutations, just as I have indicated in the section on Isolation (vol. ii. p. [280]), and indeed, in one of my earliest contributions to the theory of descent[30]. In that essay I suggested the conclusion that periods of constancy alternate with periods of variability, basing my opinion on general considerations, and on Hilgendorf's study of the Steinheim snail-shells. According to De Vries's Œnothera experiments we may assume that periods of increased variability may lead to the marked variations sometimes affecting several characters simultaneously, and occurring in many ids, which have hitherto been called 'saltatory' variations, and which we should perhaps do well to call in future, with De Vries, mutations.
[30] Ueber den Einfluss der Isolirung auf die Artbildung, Leipzig, 1872, p. 51.
We cannot yet determine how far the influence of such mutations reaches. I think it is plain that De Vries himself overestimated it, but how many of the species-types which we find to-day depend upon mere mutation can only be decided with any certainty after further investigations. For the present it is well to be clear as to the validity of the general conclusion, that all 'complex,' and especially all 'harmonious,' adaptation, must depend, not upon 'mutation,' but only upon 'variation' guided by selection. As species are essentially complexes of adaptation, originating from a basis of previous complexes of adaptation, there remains, as far as I can see, only the small field of indifferent characters to be determined by mutations, unless indeed we are to include under the term 'mutation' all variations which gain stability, but this would be merely a play upon words. In my opinion there is no definable boundary line between variation and mutation, and the difference between these two phenomena depends solely on the number—larger or smaller—of ids which have varied in the same direction.