Species and Varieties: Their Origin by Mutation 821-826 (1906)
Prof. Hugo de Vries


The comparison of artificial and natural selection has furnished material support for the theory of descent, and in turn been the object of constant criticism since the time of Darwin. The criticisms, in greater part, have arisen chiefly from an imperfect knowledge of both processes. By the aid of distinctions recently made possible, the contrast between elementary species and improved races has become much more vivid, and promises to, yield better results on which to base comparisons of artificial and natural selection.

Elementary species, as we have seen in earlier lectures, occur in wild and in cultivated plants. In older genera and systematic species they are often present in small numbers only, but many of the more recent wild types and also many of the cultivated forms are very rich in this respect. In agriculture the choice of the most adequate elementary forms for any special purpose is acknowledged as the first step in the way of selection, and is designated by the name of variety-testing, applying the term variety to all the subdivisions of systematic species indiscriminately. In natural processes it bears the title of survival of species. The fact that recent types show large numbers, and in some instances even hundreds of minor constant forms, while the older genera are considerably reduced in this respect, is commonly explained by the assumption of extinction of species on a correspondingly large scale. This extinction is considered to affect the unfit in a higher measure than the fit. Consequently the former vanish, often without leaving any trace of their existence, and only those that prove to be sufficiently adapted to the surrounding external conditions, resist and survive.

This selection exhibits far-reaching analogies between the artificial and the natural processes, and is in both cases of the very highest importance. In nature the dying out of unfit mutations is the result of the great struggle for life. In a previous lecture we have compared its agency with that of a sieve. All elements which are too small or too weak fall through, and only those are preserved which resist the sifting process. Reduced in number they thrive and multiply and are thus enabled to strike out new mutative changes. These are again submitted to the sifting tests, and the frequent repetition of this process is considered to give a good explanation of the manifold, highly complicated, and admirable structures which strike the beginner as the only real adaptations in nature.

Exactly in the same way artificial selection isolates and preserves some elementary species, while it destroys others. Of course the time is not sufficient to secure new mutations, or at least these are only rare at present, and their occurrence is doubtful in historic periods. Apart from this unavoidable difference the analogy between natural and artificial selection appears to me to be very striking.

This form of selection may be termed selection between species. Opposed to it stands the selection within the elementary species or variety. It has of late, alone come to be known as selection, though in reality it does not deserve this distinction. I have already detailed the historical evidence which gives preference to selection between species. The process can best be designated by the name of intra-specific selection, if it is understood that the term intra-specific is meant to apply to the conception of small or elementary species.

I do not wish to propose new terms, but I think that the principal differences might better become understood by the introduction of the word election into the discussion of questions of heredity. Election meant formerly the preferential choice of single individuals, while the derivation of the word selection points to a segregation of assemblies into their larger parts. Or to state it in a shorter way, individual selection is exactly what is usually termed election. Choosing one man from among thousands is to elect him, but a select party is a group of chosen persons. There would be no great difficulty in the introduction of the word election, as breeders are already in the habit of calling their choice individuals "élite," at least in the case of beets and of cereals.

This intra-specific selection affords a second point for the comparison between natural and artificial processes. This case is readily granted to be more difficult than the first, but there can be no doubt that the similarity is due to strictly comparable causes. In practice this process is scarcely second in importance to the selection between species, and in numerous cases it rests upon it, and crowns it, bringing the isolated forms up to their highest possible degree of usefulness. In nature it does quite the same, adapting strains of individuals to the local conditions of their environment. Improved races do not generally last very long in practice; sooner or later they are surpassed by new selections. Exactly so we may imagine the agency of natural intra-specific selection. It produces the local races, the marks of which disappear as soon as the special external conditions cease to act:. It is responsible only for the smallest lateral branches of the pedigree, but has nothing in common with the evolution on the main stems. It is of very subordinate importance.

These assertions of course, are directly opposed to the current run of scientific belief, but they are supported by facts. A considerable part of the evidence has already been dealt with and for our closing discussion only an exact comparison remains to be made between the two detailed types of intra-specific selection. In coming to this I will first dwell upon some intermediate types and conclude with a critical discussion of the features of artificial selection, which to my mind prove the invalidity of the conclusions drawn from it in behalf of an explanation of the processes of nature.

Natural selection occurs not only in the wild state, but is also active in cultivated fields. Here it regulates the struggle of the selected varieties and improved races with the older types, and even with the wild species. In a previous lecture I have detailed the rapid increase of the wild-oats in certain years, and described the experiments of Risler and Rimpau in the running out of select varieties. The agency is always the same. The preferred forms, which give a larger harvest, are generally more sensitive to injurious influences, more dependent on rich manure and on adequate treatment. The native varieties have therefore the advantage, when climatic or cultural conditions are unfavorable for the fields at large. They suffer in a minor degree, and are thereby enabled to propagate themselves afterwards more rapidly and to defeat the finer types. This struggle for life is a constant one, and can easily be followed, whenever the composition of a strain is noted in successive years. It is well appreciated by breeders and farmers, because it is always liable to counteract their endeavors and to claim their utmost efforts to keep their races pure. There can be no doubt that exactly the same struggle exempt from man's intrusion is fought out in the wild state.

Local races of wild plants have not been the object for field-observations recently. Some facts however, are known concerning them. On the East Friesian Islands in the North Sea the flowers are strikingly larger and brighter colored than those of the same species on the neighboring continent. This local difference is ascribed by Behrens to a more severe selection by the pollinating insects in consequence of their lesser frequency on these very windy isles. Seeds of the pines from the Himalayas yield cold-resisting young plants if gathered from trees in a high altitude, while the seeds of the same species from lower regions yield more sensitive seedlings. Similar instances are afforded by Rhododendron and other mountain species. According to Cieslar corresponding differences are shown by seeds of firs and larches from alpine and lowland provinces.

Such changes are directly dependent on external influences. This is especially manifest in experiments extending the cultures in higher or in more northern regions. The shorter summer is a natural agent of selection; it excludes all individuals which cannot ripen their seeds during so short a period. Only the short-lived ones survive. Schübeler made very striking experiments with corn and other different cereals, and has succeeded in making their culture possible in regions of Norway where it formerly failed. In the district of Christiania, corn had within some few years reduced its lifetime from 123 to 90 days, yielding smaller stems and fewer kernels, but still sufficient to make its culture profitable under the existing conditions.

This change was not permanent, but was observed to diminish rapidly and to disappear entirely, whenever the Norwegian strain was cultivated in the southern part of Germany. It was a typical improved race, dependent on continual selection by the short summers which had produced it. Similar results have been reached by Von Wettstein in the comparison of kinds of flax from different countries. The analogy between such cultivated local races and the local races of nature is quite striking. The practice of seed-exchange rests for a large part on the experience that the characters, acquired under the definite climatic and cultural conditions of some select regions, hold good for one or two, and sometimes even more generations, before they decrease to practical uselessness. The Probstei, the Hanna and other districts owe their wealth to this temporary superiority of their wheat and other cereals.

Leaving these intermediate forms of selection, we now come to our principal point. It has already been discussed at some length in the previous lecture, but needs further consideration. It is the question whether intra-specific selection may be regarded as a cause of lasting and ever-increasing improvement. This is assumed by biologists who consider fluctuating variability as the main source of progression in the organic world. But the experience of the breeders does not support this view, since the results of practice prove that selection according to a constant standard soon reaches a limit which it is not capable of transgressing. In order to attain further improvements the method of selection itself must be improved. A better and sharper method assures the choice of more valuable representatives of the race, even if these must be sought for in far larger numbers of individuals, as is indicated by the law of Quetelet.

Continuous or even prolonged improvement of a cultivated race is not the result of frequently repeated selection, but of the improvement of the standard of appreciation. Nature, as far as we know, changes her standard from time to time only in consequence of the migrations of the species, or of local changes of climate. Afterwards the new standard remains unchanged for centuries.

Selection, according to a constant standard, reaches its results in few generations. The experience of Van Mous and other breeders of apples shows that the limit of size and lusciousness may be soon attained. Vilmorin's experiments with wild carrots and those of Carrière with radishes lead to the same conclusion as regards roots. Improvements of flowers in size and color are usually easy and rapid in the beginning, but an impassable limit is soon reached. Numerous other instances could be given.

Contrasted with these simple cases is the method of selecting sugar-beets. More than once I have alluded to this splendid example of the influence of man upon domestic races, and tried to point out how little support it affords to the current scientific opinion concerning the power of natural selection. For this reason it is interesting to see how a gradual development of the methods of selection has been, from the very outset, one of the chief aims of the breeders. None of them doubts that an improvement of the method alone is adequate to obtain results. This result, in the main, is the securing of a few per-cent more of sugar, a change hardly comparable with that progress in evolution, which our theories are destined to explain.

Vilmorin's original method was a very simple one. Polarization was still undiscovered in his time. He determined the specific weight of his beets, either by weighing them as a whole, or by using a piece cut from the base of the roots and deprived of its bark, in order to test only the sugar-tissues. The pieces were floated in solutions of salt, which were diluted until the pieces began to sink. Their specific weight at that moment was determined and considered to be a measure of the corresponding value of the beet.. This principle was afterwards improved in two ways. The first was a selection after the salt-solution-method, but performed on a large scale. After some few determinations, a solution was made of such strength as to allow the greater number of the beets to float, and only the best to sink down. In large vessels thousands of beets could be tested in this way, to select a few of the very heaviest. The other improvement was the determination of the specific weight of the sap, pressed out from the tissue. It was more tedious and more expensive, but more direct, as the influence of the air-cavities of the tissue was excluded. It prepared the way for polarization.

This was introduced about the year 1874 in Germany, and soon became generally accepted. It allowed the amount, of sugar to be measured directly, and with but slight trouble. Thousands of beets could be tested yearly by this method, and the best selected for the production of seed. In some factories a standard percentage is determined by previous inquiries, and the mass of the beets is tested only by it. In others the methods of taking samples and clearing the sap have been improved so far as to allow the exact determination of three hundred thousand polarization-values of beets within a few weeks. Such figures give the richest material for statistical studies, and at once indicate the best roots, while they enable the breeder to change his standard in accordance with the results at any time. Furthermore they allow the mass of the beets to be divided into groups of different quality, and to produce, besides the seeds for the continuation of the race, a first-class and second-class product and so on. In the factory of Messrs. Kuhn & Co., at Naarden, Holland, the grinding machine has been markedly improved, so as to tear all cell-walls asunder, open all cells, and secure the whole of the sap within less than a minute, and without heating.

It would take too long to go into further details, or to describe the simultaneous changes that have been applied to the culture of the élite strains. The detailed features suffice to show that the chief care of the breeder in this case is a continuous amelioration of the method of selecting. It is manifest that. the progression of the race is in the main due to great technical improvements, and not solely to the repetition of the selection.

Similar facts may be seen on all the great lines of industrial selection. An increasing appreciation of all the qualities of the selected plants is the common feature. Morphological characters, and the capacity of yielding the desired products, are the first points that strike the breeder. The relation to climate and the dependence on manure soon follow, but the physiological and chemical sides of the problem are usually slow of recognition in the methods of selection. When visiting Mr. de Vilmorin at Paris some years ago, I inspected his laboratory for the selection of potatoes. In the method in use, the tubers were rubbed to pulp and the starch was extracted and measured. A starch-percentage figure was determined for each plant, and the selection of the tubers for planting was founded upon this result. In the same way wheat has been selected by Dippe at Quedlinburg, first by a determination of its nitrogenous contents in general, and secondly by the amount of the substances which determine its value for baking purposes.

The celebrated rye of Schlanstedt was produced by the late Mr. Rimpau in a similar manner and was put on the market between 1880 and 1890 and was received with great favor throughout central Europe, especially in Germany and in France. It is a tail variety, with vigorous stems and very long heads, the kernels of which are nearly double the size of those of the ordinary rye, and are seen protruding, when ripe, from between the scales of the spikelets. It is unfit for poor soils, but is one of the very best varieties for soils of medium fertility in a temperate climate. It is equal in the production of grain to the best French sorts, but far surpassing them in its amount of straw. It was perfected at the farm of Schlanstedt very slowly, according to the current conceptions of the period. The experiment was started in the year 1866, at which time Rimpau collected the most beautiful heads from among his fields, and sowed their kernels in his experiment-garden. From this first culture the whole race was derived. Every year the best ears of the strain were chosen for repeated culture, under experimental care, while the remainder was multiplied in a field to furnish the seeds for large and continually increasing areas of his farms.

Two or three years were required to produce the quantity of seed of each kind required for all the fields of Schlanstedt. The experiment-garden, which through the kindness of Mr. Rimpau I had the good fortune of visiting more than once between 1875 and 1878, was situated in the middle of his farm, at some distance from the dwellings. Of course it was treated with more care, and especially kept in better conditions of fertility than was possible for the fields at large. A continued study of the qualities and exigencies of the élite plants accompanied this selection, and gave the means of gradually increasing the standard. Resistance against disease was observed and other qualities were ameliorated in the same manner. Mr. Rimpau repeatedly told me that he was most anxious not to overlook any single character, because he feared that if any of them might become selected in the wrong way, perchance unconsciously, the whole strain might suffer to such a degree as to make all the other ameliorations quite useless. With this purpose the number of plants per acre was kept nearly the same as those in the fields, and the size of the culture was large enough every year to include the best kernels of quite a number of heads. These were never separated, and exact individual pedigrees were not included in the plan. This mixture seemed to have the advantage of keeping up an average value of the larger number of the characters, which either from their nature or from their apparent unimportance had necessarily to be neglected.

After ten years of continuous labor, the rye of Rimpau caught the attention of his neighbors, being manifestly better than that of ordinary sowings. Originally he had made his cultures for the improvement of his own fields only. Gradually however, he began to sell his product as seed to others, though he found the difference still very slight. After ten years more, about 1886, he was able to sell all his rye as seed, thereby making of course large profits. It is now acknowledged as one of the best sorts, though in his last letter Mr. Rimpau announced to me that the profits began to decline as other selected varieties of rye became known. The limit of productiveness was reached, and to surmount this, selection had to be begun again from some new and better starting point.

This new starting point invokes quite another principle of selection, a principle which threatens to make the contrast between artificial and natural selection still greater. In fact it is nothing new, being in use formerly in the selection of domestic animals, and having been applied by Vilmorin to his sugar-beets more than half a century ago. Why it should ever have been overlooked and neglected in the selection of sugar-beets now is not clear.

The principle in itself is very simple. It agrees that the visible characters of an animal or a plant are only an imperfect measure for its hereditary qualities, instead of being the real criterion to be relied upon, as is the current belief. It further reasons that a direct appreciation of the capacity of inheritance can only be derived from the observation of the inheritance itself. Hence it concludes that the average value of the offspring is the only real standard by which to judge the representatives of a race and to found selection upon.

These statements are so directly opposed to views prevalent among plant-breeders, that it seems necessary to deal with them from the theoretical and experimental, as well as from the practical side.

The theoretical arguments rest on the division of the fluctuating variability into the two large classes of individual or embryonic, and of partial deviations. We have dealt with this division at some length in the previous lecture. It will be apparent at once, if we choose a definite example. Let us ask what is the real significance of the percentage-figure of a single plant in sugar-beets. This value depends in the first place, on the strain or family from which the beet has been derived, but this primary point may be neglected here, because it is the same for all the beets of any lot, and determines the average, around which all are fluctuating.

The deviation of the percentage-figure of a single beet depends on two main groups of external causes. First come those that have influenced the young germs of the plant during its most sensitive period, when still an embryo within the ripening seed. They give a new limitation to the average condition, which once and forever becomes fixed for this special individual. In the second place the young seedling is affected during the development of its crown of leaves, and of its roots, by numerous factors, which cannot change this average, but may induce deviations from it, increasing or decreasing the amount of sugar, which will eventually be laid down in the root. The best young beet may be injured in many ways during periods of its lifetime, and produce less sugar than could reasonably be expected from it. It may be surpassed by beets of inferior constitution, but growing under more favorable circumstances.

Considered from this point of view the result of the polarization-test is not a single value, but consists of at least two different factors. It may be equal to the algebraic sum of these, or to their difference, according to whether the external conditions on the field were locally and individually favorable or unfavorable. A large amount of sugar may be due to high individual value, with slight subsequent deviation from it, or to a less prominent character combined with an extreme subordinate deviation.

Hence it is manifest. that even the results of such a highly improved technical method do not deserve the confidence usually put in them. They are open to doubt, and the highest figures do not really indicate the best representatives of the race. In order to convey this conception to you in a still stronger manner, let us consider the partial variability as it usually shows itself. The various leaves of a plant may noticeably vary in size, the flowers in color, the fruits in flavor. They fluctuate around an average, which is assumed to represent the approximate value of the whole plant. But if we were allowed to measure only one leaf, or to estimate only one flower or fruit, and be compelled to conclude from it the worth of the whole plant, what mistakes we could make! We might indeed hit upon an average case, but we might as easily get an extreme, either in the way of increase or of decrease. In both cases our judgment, would be badly founded. Now who can assure us that the single root of a given beet is an average representative of the partial variability? The fact that there is only one main root does not prove anything. An annual plant has only one stem, but a perennial species has many. The average height of the last is a reliable character, but the casual height of the former is very uncertain.

So it is with the beets. A beet may be divided by its buds and give quite a number of roots, belonging to the same individual. These secondary roots have been tested for the amount of sugar, and found to exhibit a manifest degree of variability. If the first root corresponded to their average, it might be considered as reliable, but if not anyone will grant that an average is more reliable than a single determination. Deviations have as a fact been observed, proving the validity of our assertion.

These considerations at once explain the disappointment so often experienced by breeders. Some facts may be quoted from the Belgian professor of agriculture at Gembloux, the late Mr. Laurent. He selected two beets, from a strain, with the exceptional amount of 23% sugar, but kept their offspring separate and analyzed some 60 of each. In both groups the average was only 11-12%, the extremes not surpassing 14-15%. Evidently the choice was a bad one, notwithstanding the high polarization value of the parent. Analogous cases are often observed, and my countrymen, Messrs. Kuhn & Co., go so far as to doubt all excessive variants, and to prefer beets with high, but less extraordinary percentages. Such are to be had in larger numbers and their average has a good chance of exemption from a considerable portion of the doubts adhering to single excessive cases.

It is curious to note here what Louis de Vilmorin taught concerning this point in the year 1850. I quote his own words: "I have observed that in experiments on heredity it is necessary to individualize as much as possible. So I have taken to the habit of saving and sowing separately the seeds of every individual beet, and I have always found that among the chosen parent-plants some had an offspring with a better average yield than others. At the end I have come to consider this character only, as a standard for amelioration."

The words are clear and their author is the originator of the whole method of plant-breeding selection. Yet the principle has been abandoned, and nearly forgotten under the impression that polarization alone was the supreme guide to be relied upon. However, if I understand the signs rightly, the time is soon coming when Vilmorin 's experience will become once more the foundation for progress in breeding.

Leaving the theoretical and historical aspects of the problem, we will now recall the experimental evidence, given in a former lecture, dealing with the inheritance of monstrosities. I have shown that in many instances monstrosities constitute double races, consisting of monstrous and of normal individuals. At first sight one might be induced to surmise that the monstrous ones are the true representatives of the race, and that their seeds should be exclusively sown, in order to keep the strain up to its normal standard. One might even suppose that the normal individuals, or the so-called atavists, had really reverted to the original type of the species and that their progeny would remain true to this.

My experiments, however, have shown that quite the contrary is the case. No doubt, the seeds of the monstrous specimens are trustworthy, but the seeds of the atavists are not less so. Fasciated hawkweeds and twisted teasels gave the same average constitution of the offspring from highly monstrous, and from apparently wholly normal individuals. In other words the fullest development of the visible characteristic was not in the slightest degree an indication of better hereditary tendencies. In unfavorable years a whole generation of a fasciated race may exhibit exclusively normal plants, without transmitting a trace of this deficiency to the following generation. As soon as the suitable conditions return, the monstrosity reassumes its full development.

The accordance of these facts with the experience of breeders of domestic animais, and of Louis de Vilmorin, and with the result of the theoretical considerations concerning the factors of fluctuation has led me to suggest the method of selecting, which I have made use of in my experiments with tricotyls and syncotyls.

Seedling variations afford a means of counting many hundreds of individuals in a single germinating pan. If seed from one parent-plant is sown only in each pan, a percentage-figure for the amount of deviating seedlings may be obtained. These figures we have called the hereditary percentages. I have been able to select the parent-plants after their death on the sole ground of these values. And the resuit has been that from varieties which, on an average, exhibited 50-55% deviating seedlings, after one or two years of selection this proportion in the offspring was brought up to about 90% in most of the cases. Phacelia and mercury with tricotylous seedlings, and the Russian sunflower with connate seed-leaves, may be cited as instances.

Besides these tests, others were performed, based only on the visible characters of the seedlings. The result was that this characteristic was almost useless as a criterion. The atavists gave, in the main, nearly the same hereditary percentages as the tricotyls and syncotyls, and their extremes were in each case far better constituted than the average of the chosen type. Hence, for selection purposes, the atavists must be considered to be in no way inferior to the typical specimens.

If it had been possible to apply this principle to twisted and fasciated plants, and perhaps even to other monstrosities, I think that it will readily be granted that the chance of bringing even these races up to a percentage of 90% would have been large enough. But the large size of the cultures required for the counting of numerous groups of offspring in the adult state has deterred me from making such trials. Recently however, I have discovered a species, Viscaria oculata which allows of counting twisted specimens in the pans, and I may soon be able to obtain proofs of this assertion. The validity of the hereditary percentage as a standard of selection has, within the last few years, been recognized and defended by two eminent breeders, W. M. Hays in this country and Von Lochow in Germany. Both of them have started from the experience of breeders of domestic animals. Von Lochow applied the principle to rye. He first showed how fallacious the visible characters often are. For instance the size of the kernels is often dependent on their number in the head, and if this number is reduced by the injurious varietal mark of lacunae (Lückigkeit), the whole harvest will rapidly deteriorate by the selection of the largest kernels from varieties which are not quite free from this hereditary deficiency.

In order to estimate the value of his ryeplants, he gathers the seed of each one separately and sows them in rows. Each row corresponds to a parent-plant and receives 200 or 150 seeds, according to the available quantity. In this way from 700 to 800 parent-plants are tested yearly. Each row is harvested separately. The number of plants gives the average measure of resistance to frost, this being the only important cause of loss. Then the yield in grain and straw is determined and calculated, and other qualities are taken into consideration. Finally one or more groups stand prominent above all others and are chosen for the continuation of the race. All other groups are wholly excluded from the "élite," but among them the best groups and the very best individuals from lesser groups are considered adequate for further cultivation, in order to produce the commercial product of the race. As a matter of fact the rye of Von Lochow is now one of the best varieties, and even surpasses the celebrated variety of Schlanstedt. It was only after obtaining proof of the validity of his method that Von Lochow decided to give it to the public.

W. M. Hays has made experiments with wheat at the Minnesota Agricultural Experiment Station. He chose a hundred grains as a proper number for the appreciation of each parent-plant, and hence has adopted the name of "centgener power" for the hereditary percentage.

The average of the hundred offspring is the standard to judge the parent by. Experience shows at once that this average is not at all proportional to the visible qualities of the parent. Hence the conclusion that the yield of the parent-plant is a very uncertain indication of its value as a parent for the succeeding generation. Only the parents with the largest power in the centgener of offspring are chosen, while all others are wholly discarded. Afterwards the seeds of the chosen groups are propagated in the field until the required quantities of seed are obtained.

This centgener power, or breeding-ability, is tested and compared for the various parent-plants as to yield, grade, and percentage of nitrogenous content in the grain, and as to the ability of the plant to stand erect, resist rust, and other important qualities. It is evident that by this test of a hundred specimens a far better and much more reliable determination can be made than on the ground of the minutest examination of one single plant. From this point of view the method of Hays commands attention. But the chief advantage lies in the fact that it is a direct proof of that which it is desired to prove, while the visible marks give only very indirect information.

Thus the results of the men of practice are in full accordance with those of theory and scientific experiment, and there can be little doubt that they open the way for a rapid and important improvement. Once attained, progress however, will be dependent on the selection-principle, and the hereditary percentage, or centgener power or breeding-ability, must be determined in each generation anew. Without this the race would soon regress to its former condition.

To return to our starting point, the comparison of artificial and natural selection. Here we are at once struck by the fact that it is hardly imaginable, how nature can make use of this principle. In some measure the members of the best centgener will manifestly be at an advantage, because they contain more fit specimens than the other groups. But the struggle for existence goes on between individuals, and not between groups of brethren against groups of cousins. In every group the best adapted individuals will survive, and soon the breeding-differences between the parents must vanish altogether. Manifestly they can, as a rule, have no lasting result on the issue of the struggle for existence.

If now we remember that in Darwin's time this principle, breeding-ability, enjoyed a far more general appreciation than at present, and that Darwin must have given it full consideration, it becomes at once clear that this old, but recently revived principle, is not adequate to support the current comparison between artificial and natural selection.

In conclusion, summing up all our arguments, we may state that there is a broad analogy between breeding-selection in the widest sense of the word, including variety-testing, race-improvement and the trial of the breeding-ability on one side, and natural selection on the other. This analogy however, points to the importance of the selection between elementary species, and the very subordinate role of intra-specific selection in nature. It strongly supports our view of the origin of species by mutation instead of continuous selection. Or, to put it in the terms chosen lately by Mr. Arthur Harris in a friendly criticism of my views: "Natural selection may explain the survival of the fittest, but it cannot explain the arrival of the fittest."