Popular Science Monthly/Volume 77/August 1910/The Role of Selection in Plant Breeding


By Professor E. M. EAST


WHEN one attempts to give some idea of the principles and of the methods and scope of plant breeding, the matter falls naturally into two parts, the role of selection and the role of hybridization. This is not because the subject is really thus separable, but because the methods in use fall into these categories. One must, of course, use selection after hybridization, but there are a number of plants of great agricultural value, in which either the flowers are too small for artificial crossing or in which other reasons make it desirable to use simply selection in their improvement. It is of these that this paper will treat.

The particular work discussed has been selected because it will illustrate certain principles, not because it is regarded as more important than other work of like nature. The work of many quiet men who are striving for the good of mankind by their efforts toward the improvement of plants deserves to be mentioned, but unfortunately the limits of a single paper are too narrow to discuss principles and to say much about practise, and knowledge of the former should be made more widespread in order that the latter may be appreciated.

The non-botanical public can not be blamed if it receives comparatively worthless productions with greater acclaim than those of value when the former obtain all the publicity and no voice is raised in protest. Exploitations of new plant introductions of little value have certainly been numerous in the past few years. Perhaps this has been a public benefit, for it has increased the general interest in plant breeding and has stimulated many laymen to study the subject in order to be able to separate the wheat from the tares when dealing with new varieties. It is strange, perhaps, with our reputation for always looking for the dollar sign, that the new agricultural productions of greatest economic value have always received less notoriety than the production of horticultural novelties of limited use and small importance; yet such is the case. It is doubtful whether the production of a new field corn that would increase the yield in the United States by ten per cent, would obtain more than a passing notice from the press; yet such an increase would add $100,000,000 per annum to the wealth of the country, and the individual who was responsible would deserve to be ranked among the greatest benefactors of the commonwealth.

This illustration serves to show something of the extent of the benefits that may be confidently expected from the improvement of cultivated plants; but the full extent of our rightful expectations is at least ten per cent, increase in both quality and quantity of all the great crops of the United States. In fact this is a very conservative forecast based upon what has been accomplished in the past. Men like Haynes with his "Blue stem" wheat and J. S. Learning with his "Learning" corn have perhaps made an even greater percentage increase in the value of the returns from the land upon which their productions have been grown. Their results were obtained largely in the latter half of the last century and even greater advances should be made in the future. This statement is made because, in the last quarter of the nineteenth century, experimental biology was in the same relative position in which chemistry stood in its beginning. During the century chemistry made wonderful advances; during this—the twentieth—century experimental biology will make similar progress. And one of the first and most important applications of the facts discovered will be to guide and direct man in producing new plants and animals by more direct and certain methods.

When one speaks of producing new plants, however, he should not be misunderstood. Man has not yet actually produced new variations (although the time may come when even this is possible); he simply works with the variations which have occurred through natural causes of which little is known. The isolation of a varying plant and from it the production of a variety, or the combination of desirable characters from one strain with other desirable characters from different strains, comprises the total aim and desire of the plant breeder. The idea is simple; to put the idea into practise successfully is often a tedious and difficult task.

As in hybridization the ease with which results can be obtained by selection depends largely upon flower structure. In selection, however, the relative facility with which artificial cross-pollination can be accomplished is of small importance. What one wishes to know is whether cross-pollination or self-pollination takes place naturally. Practically all plants are occasionally cross-fertilized naturally, and many of them have devices whereby they are nearly always crossed; but we are coming to see that cross-fertilization is not as essential to plant life as Darwin endeavored to prove in his "Cross- and Self-fertilization in the Vegetable Kingdom." Wheat, for example, is almost always self-fertilized; yet it has kept its vigor for thousands of years. The importance of this fact to the selectionist is easily seen. If seed from several varieties of wheat is mixed and planted, each variety remains true to its type because of self-pollination, and during the growing season the plants can be compared and any desirable type selected for future propagation. In a cross-pollinated plant like maize this is not the case. The pollen is carried by the wind through long distances and varieties planted close together are continually intercrossed. The isolation of a particular type is not simple as in the case of wheat, but may be prolonged through many generations. Each prize ear selected for future planting will have had at least a few and possibly many of its seeds fertilized by pollen from less desirable strains. When these seeds are grown they of course again fertilize the seeds of the desirable plants with a frequency proportionate to their number.

In the case just cited recourse may be had to artificial self-pollination. Several hundred seeds are thus produced at one operation and the work of isolating the new variety is made materially easier. But suppose we are dealing with red clover where the flowers are small, almost sterile with their own pollen and produce only one seed. In this crop, the long and tedious method of continuous selection just mentioned must be used, for there is no other way. This method is often called the pedigree-culture method. The main idea of the plan is that the seeds of single plants are grown in isolated plots, and the character of the mother plant judged by the characters of the progeny. This method has given much better results than the so-called German method, which consisted in planting a mixed lot of seeds from several of the best plants. For example, the German sugar-beet raisers have for years analyzed large numbers of sugar-beets and have grown their seed from the mother beets showing the highest percentage of sugar. No particular attention was paid to the general average of the progeny of each beet; those were bred from which appeared to be the best as shown by the polariscope sugar test. In this way the amount of sugar produced per acre was gradually increased, but progress was slow and cessation of selection immediately caused the sugar content to decline.

To see the real reason of this we must go back to the time of Darwin. The data from which Darwin proved the doctrine of descent came in large measure from domestic animals and cultivated plants. He saw that plants varied among themselves and that by selection of the variants new types were gradually produced. From these facts he argued that all evolution had taken place by the selection of minute variations and generally through the selective agency of a contest for life taking place among all living organisms. This he called the agency of natural selection. Later, however, Bateson, Korshinsky and de Vries called attention to the fact that many new types of animals and plants are known to have originated suddenly. There was no gradual evolution of the type; it simply appeared fully formed. This hypothesis, called the "mutation theory," found great favor among plant breeders for they knew that many times they had noticed and isolated plants showing new characters from their cultures, and had carefully made selections for further improvement of the new strain, but that generation after generation showed no further progress. LeCouteur, whom de Vries cites as the first known user of the pedigree culture method, had a case in point. From the heterogeneous lot of wheat plants which he was growing, he isolated a uniform type of great merit which he called "Bellevue de Talavera." For years after, this strain was subjected to selection in order to bring about further improvement, but the efforts were made in vain, for no new heritable variations were produced. Yet something was lacking from this theory. Sometimes there did appear to be a gradual improvement by selection. De Vries said that this was merely a temporary improvement made by selection of quantitative variations. He believed that when selection ceased, sooner or later the improved types would return to the original type of the variety from which it had been produced. The real interpretation of the facts and one which fitted all the parts of the puzzle together, came from the work of Johannsen and later investigators. It is an explanation that should have been thought of before, but like many other important discoveries, it was too simple for ordinary minds to grasp. Weismann had shown years before that the inheritance of characters acquired through outside influences during the development of the body was probably mythical. His investigations led him to believe that there is a continuity between the reproductive or germ cells of different generations, and that the body is nothing but a temporary house built to shelter them. Injuries to the house have no effect on the future generations unless the germ cells themselves are affected. Later Boveri and others, through their cytological studies, showed that the future germ cells are laid down at a very early stage in certain animal organisms and that very few cell divisions take place before the maturation of the reproductive organs and the production of active germ cells. The body cells he found to be built up by continuous cell division of a very different part of the original fertilized egg. Since no biologist, however, had found or is likely to find similar cytological phenomena in plants, no one seemed to grasp the idea that here was the key to the question that had been puzzling the plant breeders. Johannsen, however, brought matters straight by his experiments on beans. He found that commercial varieties of beans, though pure in grosser characters, such as color, were actually very mixed types when such characters as length or weight were studied. Several investigations were undertaken on size characters, the characters most rapidly affected by changes in environment. He found that his commercial variety fluctuated around an average size and that when seeds larger or smaller than this type were selected they responded to it in whichever direction the selection was made. The progeny of the selected beans were not so extreme, however, as their parents but regressed toward the average character of the parent race. This was nothing new. Galton had discussed the matter a decade before and had interpreted the regression as due to the "pull toward mediocrity" exerted by former ancestors that must have been on the average mediocre. Johannsen was not satisfied with this interpretation and in order to investigate the subject more thoroughly introduced the individual pedigree culture method, or pure line method as he spoke of it, into his work. All of his plants under experiment were self-fertilized for successive generations, so that all of his future bean progeny were descendants of single individuals from the original commercial variety. Each pure line he found to fluctuate around a typical size just as the commercial variety had done. Some types were exactly the same as the original mixed type, but others fluctuated around averages that would have been considered more or less extreme variations in the original. He then grew extreme variants from each of his pure lines and made the discovery that no progress at all was made by repeated selections of this kind. The progeny of the high extremes and the progeny of the low extremes each were found to fluctuate around the same pure line average. It was quite evident then that in the first place he had been dealing with a mixed race. This mixture consisted of sub-races each with a heritable difference in the character size. These heritable variations, however, were obscured by size fluctuations produced by differences in moisture, sunlight and fertilizer received by the different individual plants. There was even a difference in the size of individual beans on the same plant, due probably to location of some pods in places on the plant more desirable than others for the utilization of the plant's soluble foods waiting to be stored in the seeds. These differences due to immediate environment were not inherited. They behaved exactly as the acquired characters of an animal. This made the rôle of selection clear. The only improvement that selection can achieve is to isolate a substrain if such a substrain or substrains exist in the variety under experiment. When this substrain has been isolated, selection has absolutely no effect, and even if continued for countless generations will have no effect until nature produces one of the heritable changes which are so much rarer than the fluctuations produced by environment. It is also evident that the older idea that improvements made by continued selection;—i. e., gradual isolation of a type—are inconstant, is wrong. The explanation is that since non-inherited fluctuations obscure the heritable variations, only a pure line method can absolutely isolate a pure strain; and in the German method of mass selection with poor control against mediocre pollen, the chances were overwhelmingly in favor of the selected type recrossing with the more commonly cultivated and poorer type from which it came.

To my mind this work should clear up the strife between the critics and the adherents of evolution by mutation. It is evident that there are variations that are inherited and variations that are not inherited. If we call the one a mutation and the other a fluctuation, we have a distinction that will stand analysis. Why should a further distinction be made? De Vries believes mutations to be qualitative, fluctuations quantitative. Nevertheless, quantitative changes that are transmissible occur in much greater numbers than do qualitative changes. Opponents of mutation believe wide jumps appear too seldom to have been a factor in organic evolution, but they can not deny that they do occur. There are too many authentic cases in variation under domestication. Yet no one who has had experience in breeding plants will deny that small variations (not fluctuations) occur with much greater frequency. While it is impossible to prove it, I believe that the mathematical law of error controls the transmissible variations as well as fluctuations. If one could collect a random sample of variations that are inherited he would probably find that a great many forces act as the causes, and therefore as in ordinary probability, the extreme changes—that is, the great variations—occur with less frequency. One should remember, however, that in our present state of physiological knowledge, he can not know with much certainty which of two changes that apparently differ greatly in magnitude is really the greater in the light of the plant's economy.

It might be well before leaving this part of the subject to speak of one other point. In a strain that has been self-fertilized for several generations, gradual progress has sometimes been made by selection. This probably comes about because the parent plant is still hybrid in regard to certain characters, and it is to their recombinations that the intensification or reduction of certain apparently single characters but which are really combinations of separately heritable characters, is due. According to the law of chance with repeated self-fertilizations any strain approaches a constant condition in all of its characters when unselected, but one can not say when this state is reached unless he knows the exact number of hybrid characters in the beginning and can recognize each.

If we were to take up the crops of the United States which owe their present excellence and future prospects in large measure to the isolation of superior strains by selection, we should cover a great majority of the agricultural wealth of the country. Of course natural cross-fertilization and even occasional artificial hybridization have played important parts by causing recombinations of characters, but selection has been the main cause of improvement. Two of the important crops, tobacco and wheat, are very seldom cross-pollinated naturally; nevertheless new types are continually appearing in the fields. To make new varieties

Fig. 1. Types appearing in a Single Field of Maize. A strain like the ear near the center has been isolated.

it simply takes an alert eye for their detection, comparative tests to prove their merit and the time needed to produce a sufficient increase for commercial use. Some of our other important grain crops like oats and rye are more often cross-pollinated, as is also our chief grass crop, timothy. But as maize is probably the most difficult crop to deal with, and is a typical cross-pollinated plant as well as our most important cereal, perhaps it will be of interest to take a short survey of some of the problems with which one has to deal when endeavoring to improve it by selection.

Maize is the only one of our cereals that is monœcious. The tassel contains the pollen or male element while the silks are the stigmas of the female flowers. In order that the pollination of the silks shall be relatively certain, each tassel produces about thirty million pollen grains; and as the ears average less than five hundred seeds apiece, there are about sixty thousand pollen grains produced for each kernel. With such a large amount of superfluous pollen floating around in the air, there is a great deal of inter-crossing between the neighboring plants. This fact has been an obstacle to the improvement of maize, but it has been offset by one advantage it possesses over the other cereals, that of producing large ears. Since each individual ear must be handled and its characters noted at husking time, it is not strange that ears with desirable variations sufficiently striking to catch the eye of the grower have become the parents of numerous distinct varieties. By selecting desirable seed ears and isolating them from other varieties, various strains have been produced that are remarkably uniform in characters such as color that have forcibly attracted the attention of the breeder. Even in these strains, however, there are many natural types growing side by side and continually crossing with each other. There are stalks

Fig. 2. Inherited Abnormalities appearing in Maize. All pollen from these plants and from their normal sister plants must be prevented from maturing.

which bear their ears high and stalks which bear them low, stalks with long and stalks with short ear shanks, stalks with different leaf markings and with notably different tendencies to produce suckers. Differences are everywhere present even in the ears, as is shown in the accompanying photograph (Fig. 1). A large number of these differences are simply fluctuations produced by the environment and are not inherited. The obscuration of heritable variations by the fluctuations and the mixed condition of the natural types makes it a difficult task to isolate the most productive types. Many variations of technique have been proposed for the prosecution of the work, but are all based upon the idea of proving the capacity of a mother ear by the characters of the progeny produced. If a very large number of ears are included in the original stock, it is unquestionable that some of them will transmit more desirable characters than others. It only remains to test them out by growing the seed of each ear in marked plots or rows and gradually eliminating the undesirable types.

The accompanying diagrams, showing the work of the Illinois Agricultural Experiment Station in their experiments in selecting for high and low protein content, and high and low oil content, admirably illustrate the rapidity with which progress can be made by selecting only from the maternal side, even in the face of constant intercrossing. This work the writer believes has given a complete

Fig. 3. Diagrammatic representation of the Results of the Illinois Agricultural Experiment Station in selecting for high and for low protein content. Y, per cent, protein in crop; X, generations; h, high protein strain; l, low protein strain.

corroboration of Johannsen's conclusions on pure lines. This interpretation has been made, however, from their published data, and the Illinois station should not be held responsible. This work of breeding to change the composition of maize was started in 1896 with a hazy Darwinian idea that as corn was known to vary in composition, continuous selection of extreme variations would produce a continuous change in type. A very old type—Burr's White—furnished the foundation stock. A chemical analysis was made of parts of the individual ears each year, and the extreme ears planted. From the first, the four lines above mentioned were planted in isolated plots and were continually selected in the same direction. After ten generations the average crop of the high protein line had reached 14.26 per cent., while the low protein line was only 8.64 per cent.; the high oil strain had reached 7.37 per cent., while the low oil strain was reduced to 2.66 per cent. These facts clearly show the rapidity with which results can be obtained by this method of selection even with a crop that is often cross-fertilized. But the diagrams show other facts. The published records show that the variability of the race was but little, if any, reduced by continuous selection. With extreme variants comparatively as far removed from each year's type, available for planting in each successive generation, the gain each year should have been at the same rate, if the Darwinian interpretation of the role of selection were correct. On the contrary, we notice that the regular curve fitted to the crop averages for ten generations, is first concave showing great progress made by selection, is later convex as progress becomes slower, and last becomes horizontal

Fig. 4. Diagrammatic Representation of the Results of the Illinois Agricultural Experiment Station in selecting for high and for low oil content. Y, per cent, oil in crop; X, generations; h, high oil strain; l, low oil strain.

as no more progress results. It is very evident that the original stock was a mixed race containing sub-races of various composition intermingled by hybridization. Selection rapidly isolated these sub-races. The isolation was practically complete at the eighth generation in the case of the protein strains and the ninth generation in the oil strains. After this selection accomplished nothing. That the effect of selection was simply the isolation of a sub-race and not a continuous response, is further demonstrated by the fact that in 1903 another plot was started with seed from the isolated high oil strain. After four years' cessation of selection, the average composition of the crop remained the same, showing that after complete isolation of a homogeneous type no retrogression of the selected character occurs unless intercrossing with mediocre strains takes place. Fluctuation in composition still appears, but this is the non-inherited kind produced by external conditions.

It is sometimes somewhat difficult to see why selection of this kind should yield results slowly. There are indeed many points concerning which little is known. One may picture to himself, however, that where crossing is always likely to occur and where the apparent character is in reality a combination of a number of separately inherited characters, many thousands or even millions of individuals would have to be grown to run a fair chance of obtaining the most desirable combination. By growing a few individuals in which the desired character is intensified in successive generations, the combination wanted may be obtained with the use of smaller numbers.

I have stated that nothing can be accomplished by selection after a pure line or genotype as Johannsen calls them is isolated, unless a new transmissible variation is produced by nature. The questions then arise: how often may such changes be expected? and, what is their nature? Such changes are of two kinds,[2] progressive where a new character appears, or retrogressive where a character is lost. But little can be said as to their relative frequency. Undoubtedly some species are in a more unstable condition than others and give more of such variations, as de Vries has already suggested. On the other hand, certain unknown combinations of external conditions may favor germcell changes. They are both rare, the progressive changes being relatively much less frequent than the retrogressive changes, but they are sufficiently common for several to have come within the knowledge of every experienced breeder.

There is another type of variation much more closely related to changes occurring in "pure lines" than is generally supposed. I refer to what is commonly known as bud variation or vegetative sports. Retrogressive variations of this kind are probably no rarer than the same kind of changes occurring in pure lines. No authentic progressive variations (as distinguished from digressive) are known. In my own experience in growing eight hundred species and varieties of tuberous solanums (largely potato varieties), fifteen retrogressive variations have been noticed, and the changes that occurred were exactly like those occurring in seed-propagated strains.

The relative value of progressive and retrogressive variations is difficult to estimate. In organic evolution the former must have been far more valuable; commercially the latter are often of great worth. We may cite, for example, the great value of the bush or dwarf varieties of beans, peas and tomatoes that have originated as retrogressions.

In closing I should like to call attention to a fact both of evolutionary and of commercial importance. The first generation of crosses between nearly related types generally grows more vigorously than the pure types themselves. If the fertility is not impaired, they even fruit more freely. This is undoubtedly the explanation of Burbank's quick growing hybrid walnuts, but if they were self-pollinated and grown for another generation a large percentage of the progeny would lose this character. In naturally self-pollinated types like tobacco, one sees the phenomenon expressed as greater vigor in a cross; in a continually intercrossed species like maize the same thing is shown by a loss of vigor when the plants are self-pollinated. It is clear then that if pure strains of maize are gradually isolated by selection, by the same token they lose in vigor and productiveness. The original mixed strain may contain

Fig 5. Effects of inbreeding in Maize. Outer ears inbred four generations. Middle ear the result of their crossing, first generation.

sub-strains some of which are much more productive than others. The less productive types may be discarded, but at the same time there is a loss of vigor from the fact that they are withdrawn from hybrid combinations. The logical procedure, then, is to isolate two high-yielding types, combine them by hybridization, and grow only the first generation of the cross. This is not mere theory, for by using such methods I have obtained from 100 to 200 bushels of shelled corn per acre on small plots. Unfortunately, this method can not be used to advantage on many crops, but in the case of maize the procedure is simple. There are many breeders using the isolation method of improvement. From

Fig. 6. Results of crossing two Inbred Strains of Maize. At the right average of the parents after three generations inbreeding, 61 bushels per acre. At the left crop of first generation cross of the inbred strains, 101 bushels per acre.

them the grower obtains two strains and plants them in alternate rows. At flowering time all of the male flowers or tassels are removed from one of the plants of the varieties before they shed their pollen. All the ears that these plants produce are crossed with the other variety. It is this seed that produces the vigorous plants.

This method might be made the basis for some very valuable work in forestry. It is quite conceivable that many important timber trees might be found where nearly related species or varieties would cross readily. Experiment would show how great an increase in rapidity of growth could be expected., and whether such an increase would pay for the increased expense of hand hybridized seed.

Fig. 7. Silver Hill. Normal at left. Bud variation isolated by selection at the right.

One may summarize by saying that two important points cover the whole rôle of selection. The first point is that nature continually causes variations to appear in plants. The majority of these variations are simply accelerations or retardations of development of the whole or of certain parts of the plant due to good or bad environment at critical stages of the plant's growth. These variations are not inherited because the reproductive or germ cells are not affected. Other variations, however, are being constantly produced by nature—though much more rarely—which do affect the reproductive cells and are transmitted to the plant's progeny. These variations are the basis of selection. They are constant from the beginning and remain so unless changed by a second variation affecting the same constituent in the reproductive cells that is due to develop the character in question.

The second point to be remembered is that the whole aim and action of selection is to detect the desired heritable variants among the useful commercial plants and through them to isolate a race with the desired characters. When this is accomplished, selection can then do nothing until nature steps in and produces another desirable variation.

In other words, the results of selection are not continuous. Selection does not gradually perfect a character. The production of heritable variations is intermittent and the intermissions may be long. If the practical results seem to be parts of a continuous process, it is because of the imperfect methods at hand to isolate the desirable variations from their combinations with undesirable characters formed by natural hybridization.

  1. This paper is based on a series of poular lectures delivered at the Bussey Institution of Harvard University April and May, 1910. A second paper will follow.
  2. De Vries also gives a third kind, digressive variations, such as occur when a character previously possessed by but latent in the plant appears. This class is unnecessary. Digressive characters appear either through the loss of a complementary inhibiting factor or the gain of a complementary factor necessary for it to become active.