Notwithstanding these considerations, I will now give some examples in order to show that fluctuating variability plays a prominent part in these improvements. Of course it is the third in importance in the series. First comes the choice of the material from the a.s.semblage of species, elementary species and varieties. Hybridization comes next in importance. But even the hybrids of the best parents may be improved, because they are no less subject to Quetelet's law than any other strain. Any large number of hybrids of the same ancestry will prove this, and often the excellency of a hybrid variety depends chiefly, or at least definitely, on the selection of the best individuals. Being propagated only in a vegetative way, they retain their original good qualities through all further culture and multiplication.

As an ill.u.s.trative example I will take the genus _Canna_. Originally cultivated for its large and bright foliage only, it has since become a flowering plant of value. Our garden strains have originated by the crossing of [760] a number of introduced wild species, among which the _Canna indica_ is the oldest, now giving its name to the whole group. It has tall stems and spikes with rather inconspicuous flowers with narrow petals. It has been crossed with _C. nepalensis_ and _C. warczewiczii_, and the available historic evidence points to the year 1846 as that of the first cross. This was made by Annee between the _indica_ and the _nepalensis_; it took ten years to multiply them to the required degree for introduction into commerce. These first hybrids had bright foliage and were tall plants, but their flowers were by no means remarkable.

Once begun, hybridization was widely practiced. About the year 1889 Crozy exhibited at Paris the first beautifully flowering form, which he named for his wife, "Madame Crozy." Since that time he and many others, have improved the flowers in the shape and size, as well as in color and its patterns. In the main, these ameliorations have been due to the discovery and introduction of new wild species possessing the required characters. This is ill.u.s.trated by the following incident. In the year 1892 I visited Mr. Crozy at Lyons. He showed me his nursery and numerous acquisitions, those of former years as well as those that were quite new, and which were in the process of rapid [761] multiplication, previous to being given to the trade. I wondered, and asked, why no pure white variety was present. His answer was "Because no white species had been found up to the present time, and there is no other means of producing white varieties than by crossing the existing forms with a new white type."

Comparing the varieties produced in successive periods, it is very easy to appreciate their gradual improvement. On most points this is not readily put into words, but the size of the petals can be measured, and the figures may convey at least some idea of the real state of things.

Leaving aside the types with small flowers and cultivated exclusively for their foliage, the oldest flowers of _Canna_ had petals of 45 mm.

length and 13 mm. breadth. The ordinary types at the time of my visit had reached 61 by 21 mm., and the "Madame Crozy" showed 66 by 30 mm. It had however, already been surpa.s.sed by a few commercial varieties, which had the same length but a breadth of 35 mm. And the latest production, which required some years of propagation before being put on the market, measured 83 by 43 mm. Thus in the lapse of some thirty years the length had been doubled and the breadth tripled, giving flowers with broad corollas and with petals [762] joined all around, resembling the best types of lilies and _Amaryllis_.

Striking as this result unquestionably is, it remains doubtful as to what part of it is due to the discovery and introduction of new large flowered species, and what to the selection of the extremes of fluctuating variability. As far as I have been able to ascertain however, and according to the evidence given to me by Mr. Crozy, selection has had the largest part in regard to the size, while the color-patterns are introduced qualities.

The scientific a.n.a.lysis of other intricate examples is still more difficult. To the practical breeder they often seem very simple, but the student of heredity, who wishes to discern the different factors, is often quite puzzled by this apparent simplicity. So it is in the case of the double lilacs, a large number of varieties of which have recently been originated and introduced into commerce by Lemoine of Nancy. In the main they owe their origin to the crossing and recrossing of a single plant of the old double variety with the numerous existing single-flowered sorts.

This double variety seems to be as old as the culture of the lilacs. It was already known to Munting, who described it in the year 1671. Two centuries afterwards, in 1870, a new description [763] was given by Morren, and though more than one varietal name is recorded in his paper, it appears from the facts given that even at that time only one variety existed. It was commonly called _Syringa vulgaris azurea plena_, and seems to have been very rare and without real ornamental value.

Lemoine, however, conceived the desirability of a combination of the doubling with the bright colors and large flower-racemes of other lilacs, and performed a series of crosses. The "_azurea plena_" has no stamens, and therefore must be used in all crosses as the pistil-parent; its ovary is narrowly inclosed in the tube of the flower, and difficult to fertilize. On the other hand, new crosses could be made every year, and the total number of hybrids with different pollen-parents was rapidly increased. After five years the hybrids began to flower and could be used for new crosses, yielding a series of compound hybrids, which however, were not kept separate from the products of the first crosses.

Gradually the number of the flowering specimens increased, and the character of doubling was observed to be variable to a high degree.

Sometimes only one supernumerary petal was produced, sometimes a whole new typical corolla was extruded from within the first. In the same [764] way the color and the number of the flowers on each raceme were seen to vary. Thousands of hybrids were produced, and only those which exhibited real advantages were selected for trade. These were multiplied by grafting, and each variety at present consists only of the buds of one original individual and their products. No constancy from seed is a.s.sumed, many varieties are even quite sterile.

Of course, no description was given of the rejected forms. It is only stated that many of them bore either single or poorly filled flowers, or were objectionable in some other way. The range of variability, from which the choices were made, is obscure and only the fact of the selection is prominent. What part is due to the combination of the parental features and what to the individual fluctuation of the hybrid itself cannot be ascertained.

So it is in numerous other instances. The dahlias have been derived from three or more original species, and been subjected to cultivation and hybridization in an ever-increasing scale for a century. The best varieties are only propagated in the vegetative way, by the roots and buds, or by grafting and cutting. Each of them is, with regard to its hereditary qualities, only one individual, and the individual characters were selected at the same time with the [765] varietal and hybrid characters. Most of them are very inconstant from seed and as a rule, only mixtures are offered for sale in seed-lists. Which of their ornamental features are due to fluctuating deviation from an average is of course unknown. _Amaryllis_ and _Gladiolus_ are surrounded with the same scientific uncertainties. Eight or ten, or even more, species have been combined into one large and multiform strain, each bringing its peculiar qualities into the mixed ma.s.s. Every hybrid variety is one individual, being propagated by bulbs only. Colors and color-patterns, shape of petals and other marks, have been derived from the wild ancestors, but the large size of many of the best varieties is probably due to the selection of the extremes of fluctuating variability. So it is with the begonias of our gardens, which are also composite hybrids, but are usually sown on a very large scale. Flowers of 15 cm. diameter are very showy, but there can be no doubt about the manner in which they are produced, as the wild species fall far short of this size.

Among vegetables the potatoes afford another instance. Originally quite a number of good species were in culture, most of them having small tubers. Our present varieties are due to hybridization and selection, each of them being propagated only in the vegetative way.

[766] Selection is founded upon different qualities, according to the use to be made of the new sort. Potatoes for the factory have even been selected for their amount of starch, and in this case at least, fluctuating variability has played a very important part in the improvement of the race.

Vegetative propagation has the great advantage of exempting the varieties from regression to mediocrity, which always follows multiplication by seeds. It affords the possibility of keeping the extremes constant, and this is not its only advantage. Another, likewise highly interesting, side of the question is the uniformity of the whole strain. This is especially important in the case of fruits, though ordinarily it is regarded as a matter of course, but there are some exceptions which give proof of the real importance of the usual condition. For example, the walnut-tree. Thousands of acres of walnut-orchards consist of seedling trees grown from nuts of unknown parentage. The result is a great diversity in the types of trees and in the size and shape of the nuts, and this diversity is an obvious disadvantage to the industry. The cause lies in the enormous difficulties attached to grafting or budding of these trees, which make this method very expensive and to a high degree uncertain and unsatisfactory.

[767] After this hasty survey of the more reliable facts of the practice of an as.e.xual multiplication of the extremes of fluctuating variability, we may now return to the previously mentioned theoretical considerations. These are concerned with an estimation of the chances of the occurrence of deviations, large enough to exhibit commercial value.

This chance may be calculated on the basis of Quetelet's law, whenever the agreement of the fluctuation of the quality under consideration has been empirically determined. In the discussion of the methods of comparing two curves, we have pointed to the quartiles as the decisive points, and to the necessity of drawing the curves so that these points are made to overlie one another, on each side of the average. If now we calculate the binomium of Newton for different values of the exponent, the sum of the coefficients is doubled for each higher unit of the exponent, and at the same time the extreme limit of the curve is extended one step farther. Hence it is possible to calculate a relation between the value of the extreme and the number of cases required. It would take us too long to give this calculation in detail, but it is easily seen that for each succeeding step the number of individuals must be doubled, though the length of the steps, or the amount of increase of the quality [768] remains the same. The result is that many thousands of seedlings are required to go beyond the ordinary range of variations, and that every further improvement requires the doubling of the whole culture. If ten thousand do not give a profitable deviation, the next step requires twenty thousand, the following forty thousand, and so on.

And all this work would be necessary for the improvement of a single quality, while practice requires the examination and amelioration of nearly all the variable characters of the strain.

Hence the rule that great results can only be obtained by the use of large numbers, but it is of no avail to state this conclusion from a scientific point of view. Scientific experimenters will rarely be able to sacrifice fifty thousand plants to a single selection. The problem is to introduce the principle into practice and to prove its direct usefulness and reliability. It is to Luther Burbank that we owe this great achievement. His principles are in full harmony with the teachings of science. His methods are hybridization and selection in the broadest sense and on the largest scale. One very ill.u.s.trative example of his methods must suffice to convey an idea of the work necessary to produce a new race of superlative excellency. Forty thousand blackberry and raspberry [769] hybrids were produced and grown until the fruit matured.

Then from the whole lot a single variety was chosen as the best. It is now known under the name of "Paradox." All others were uprooted with their crop of ripening berries, heaped up into a pile twelve feet wide, fourteen feet high and twenty-two feet long, and burned. Nothing remained of that expensive and lengthy experiment, except the one parent-plant of the new variety. Similar selections and similar amount of work have produced the famous plums, the brambles and the blackberries, the Shasta daisy, the peach almond, the improved blueberries, the hybrid lilies, and the many other valuable fruits and garden-flowers that have made the fame of Burbank and the glory of horticultural California.

[770]

LECTURE XXVII

INCONSTANCY OF IMPROVED RACES

The greater advantages of the as.e.xual multiplication of extremes are of course restricted to perennial and woody plants. Annual and biennial species cannot as a rule, be propagated in this way, and even with some perennials horticulturists prefer the sale of seeds to that of roots and bulbs. In all these cases it is clear that the exclusion of the individual variability, which was shown to be an important point in the last lecture, must be sacrificed.

Seed-propagation is subject to individual as well as to fluctuating variability. The first could perhaps be designated by another term, embryonic variability, since it indicates the fluctuations occurring during the period of development of the germ. This period begins with the fusion of the male and female elements and is largely dependent upon the vigor of these cells at the moment, and on the varying qualities they may have acquired. It comprises in the main the time of the ripening of the seed, and [771] might perhaps best be considered to end with the beginning of the resting stage of the ripe seed. Hence it is clear that the variability of seed-propagated annual races has a wider range than that of perennials, shrubs and trees. At present it is difficult to discern exactly the part each of these two main factors plays in the process. Many indications are found however, that make it probable that embryonic variability is wider, and perhaps of far greater importance than the subsequent partial fluctuations. The high degree of similarity between the single specimens of a vegetative variety, and the large amount of variability in seed-races strongly supports this view.

The propagation and multiplication of the extremes of fluctuating variability by means of seeds requires a close consideration of the relation between seedling and parent. The easiest way to get a clear conception of this relation is to make use of the ideas concerning the dependency of variability upon nourishment. a.s.suming these to be correct in the main, and leaving aside all minor questions, we may conclude that the chosen extreme individual is one of the best nourished and intrinsically most vigorous of the whole culture. On account of these very qualities it is capable of nourishing all of its organs better and also its seeds. In other words, the seeds [772] of the extreme individuals have exceptional chances of becoming better nourished than the average of the seeds of the race. Applying the same rule to them, it is easily understood that they will vary, by reason of this better nourishment, in a direction corresponding to that of their parent.

This discussion gives a very simple explanation of the acknowledged fact that the seeds of the extremes are in the main the best for the propagation of the race. It does not include however, all the causes for this preferment. Some are of older date and due to previous influences.

A second point in our discussion is the appreciation of the fact that a single individual may be chosen to gather the seed from, and that these seeds, and the young plants they yield, are as a rule, numerous. Hence it follows that we are to compare their average and their extremes with the qualities of the parents. Both are of practical as well as of theoretical interest. The average of the progeny is to be considered as the chief result of the selection in the previous generation, while the extremes, at least those which depart in the same direction, are obviously the means of further improvement of the race.

Thus our discussion should be divided into [773] two heads. One of these comprises the relation of the average of the progeny to the exceptional qualities of the chosen parent, and the other the relation of exceptional offspring to the exceptional parents.

Let us consider the averages first. Are they to be expected to be equal to the unique quality of the parent, or perhaps to be the same as the average of the whole unselected race? Neither of these cases occur.

Experience is clear and definite on this important point. Vilmorin, when making the first selections to improve the amount of sugar in beets, was struck with the fact that the average of the progeny lies between that of the original strain and the quality of the chosen parent. He expressed his observation by stating that the progeny are grouped around and diverge in all directions from some point, placed on the line which unites their parent with the type from which it sprang. All breeders agree on this point, and in scientific experiments it has often been confirmed. We shall take up some ill.u.s.trative examples presently, but in order to make them clear, it is necessary to give a closer consideration to the results of Vilmorin.

From his experience it follows that the average of the progeny is higher than that of the race at large, but lower than the chosen parent. [774]

In other words, there is a progression and a regression. A progression in relation to the whole race, and a regression in comparison with the parent. The significance of this becomes clear at once, if we recall the constancy of the variety which could be obtained from the selected extreme in the case of vegetative multiplication. The progression is what the breeder wants, the regression what he detests. Regression is the permanency of part of the mediocrity which the selection was invoked to overcome. Manifestly it is of the highest interest that the progression should be as large, and the regression as small as possible.

In order to attain this goal the first question is to know the exact measure of progression and regression as they are exhibiting themselves in the given cases, and the second is to inquire into the influences, on which this proportion may be inc.u.mbent.

At present our notions concerning the first point are still very limited and those concerning the second extremely vague. Statistical inquiries have led to some definite ideas about the importance of regression, and these furnish a basis for experimental researches concerning the causes of the phenomenon. Very advantageous material for the study of progression and regression in the realm of fluctuating variability is afforded by the [775] ears of corn or maize. The kernels are arranged in longitudinal rows, and these rows are observed to occur in varying, but always even, numbers. This latter circ.u.mstance is due to the fact that each two neighboring rows contain the lateral branches of a single row of spikelets, the ages of which however, are included in the fleshy body of the ear. The variation of the number of the rows is easily seen to comply with Quetelet's law, and often 30 or 40 ears suffice to give a trustworthy curve. Fritz Muller made some experiments upon the inheritance of the number of the rows, in Brazil. He chose a race which averaged 12 rows, selected ears with 14, 16 and 18 rows, etc., and sowed their kernels separately. In each of-these cultures he counted the rows of the seeds on the ears of all the plants when ripe, and calculated their average. This average, of course, does not necessarily correspond to a whole number, and fractions should not be neglected.

According to Vilmorin's rule he always found some progression of the average and some regression. Both were the larger, the more the parent-ear differed from the general average, but the proportion between both remained the same, and seems independent of the amount of the deviation. Putting the deviation at 5, the progression calculated from his figures is [776] 2 and the regression 3. In other words the average of the progeny has gained over the average of the original variety slightly more than one-third, and slightly less than one-half of the parental deviation. I have repeated this experiment of Fritz Miller's and obtained nearly the same regression of three-fifths, though working with another variety, and under widely different climatic conditions.

The figures of Fritz Muller were, as given below, in one experiment. In the last column I put the improvement calculated for a proportion of two-fifths above the initial average of 12.

Rows on Average of rows 12 + 2/5 of parent ears of progeny Difference 14 12.6 12.8 16 14.1 13.6 18 15.2 14.4 20 15.8 15.2 22 16.1 16.0

Galton, in his work on natural inheritance, describes an experiment with the seeds of the sweet pea or _Lathyrus odoratus_. He determined the average size in a lot of purchased seeds, and selected groups of seeds of different, but for each group constant, sizes. These were sown, and the average of the seeds was determined anew in the subsequent harvest they yielded. These figures agreed with the rule of Vilmorin and were calculated in the manner [777] given for the test of the corn. The progression and regression were found to be proportionate to the amount of the deviation. The progression of the average was one-third, and the regression in consequence two-thirds of the total deviation. The amelioration is thus seen to be nearly, though not exactly, the same as in the previous case.

From the evidence of the other corresponding experiments, and from various statistical inquiries it seems that the value of the progression is nearly the same in most cases, irrespective of the species used and the quality considered. It may be said to be from one-third to one-half of the parental deviation, and in this form the statement is obviously of wide and easy applicability.

Our figures also demonstrate the great preeminence of vegetative varieties above the improved strains multiplied by seeds. They have a definite relation. As.e.xually multiplied strains may be said to be generally two times or even three times superior to the common offspring. This is a difference of great practical importance, and should never be lost sight of in theoretical considerations of the productive capacity of selection. Multiplication by seed however, has one great advantage over the as.e.xual method; it may be repeated. The [778] selection is not limited to a single choice, but may be applied in two or more succeeding generations. Obviously such a repet.i.tion affords a better chance of increasing the progression of the average and of ameliorating the race to a greater degree than would be possible by a single choice. This principle of repeated selection is at present the prominent feature of race improvement. Next to variety-testing and hybridizing it is the great source of the steady progression of agricultural crops. From a practical standpoint the method is clear and as perfect as might be expected, but this is not the side of the problem with which we are concerned here. The theoretical a.n.a.lysis and explanation of the results obtained, however, is subject to much doubt, and to a great divergence of conceptions. So it is also with the application of the practical processes to those occurring in nature.

Some a.s.sume that here repeated selection is only of subordinate importance, while others declare that the whole process of evolution is due to this agency. This very important point however, will be reserved for the next lecture, and only the facts available at present will be considered here.

As a first example we may take the ray-florets of the composites. On a former occasion we have dealt with their fluctuation in number and [779]

found that it is highly variable and complies in the main with Quetelet's law. _Madia elegans_, a garden species, has on the average 21 rays on each head, fluctuating between 16 and 25 or more. I saved the seeds of a plant with only 17 rays on the terminal head, and got from them a culture which averaged 19 rays, which is the mean between 21 and 17. In this second generation I observed the extremes to be 22 and 12, and selected a plant with 13 rays as the parent for a continuation of the experiment. The plants, which I got from its seeds, averaged 18 and showed 22 and 13 as extremes. The total progression of the average was thus, in two generations, from 21 to 18, and the total regression from 13 to 18, and the proportion is thus seen to diminish by the repet.i.tion rather than to increase.

This experiment, however, is of course too imperfect upon which to found general conclusions. It only proves the important fact that the improved average of the second generation is not the starting-point for the further improvement. But the second generation allows a choice of an extreme, which diverges noticeably more from the mean than any individual of the first culture, and thereby gives a larger amount of absolute progression, even if the proportion between progression and regression remains [780] the same. The repet.i.tion is only an easy method of getting more widely deviating extremes; whether it has, besides this, another effect, remains doubtful. In order to be able to decide this question, it is necessary to repeat the selection during a series of generations. In this way the individual faults may be removed as far as possible. I chose an experiment of Fritz Muller, relating to the number of rows of grains on the ears exactly as in the case above referred to, and which I have repeated in my experimental garden at Amsterdam.

I started from a variety known to fructify fairly regularly in our climate, and exhibiting in the mean 12-14 rows, but varying between 8 and 20 as exceptional cases. I chose an ear with 16 rows and sowed its seeds in 1887. A number of plants were obtained, from each of which, one ear was chosen in order to count its rows. An average of 15 rows was found with variations complying with Quetelet's law. One ear reached 22 rows, but had not been fertilized, some others had 20 rows, and the best of these was chosen for the continuation of the experiment. I repeated the sowing during 6 subsequent generations in the same way, choosing each time the most beautiful ear from among those with the greatest number of rows. Unfortunately with the increase of the number the [781]

size of the grains decreases, the total amount of nourishment available for all of them remaining about the same. Thus the kernels and consequently the new plants became smaller and weaker, and the chance of fertilization was diminished in the ears with the highest number of rows. Consequently the choice was limited, and after having twice chosen a spike with 20 and once one with 24 rows, I finally preferred those with the intermediate number of 22.

This repeated choice has brought the average of my race up from 13 to 20, and thus to the extreme limit of the original variety. Seven years were required to attain this result, or on an average the progression was one row in a year. This augmentation was accompanied by an accompanying movement of the whole group in the same direction. The extreme on the side of the small numbers came up from 8 to 12 rows, and cobs with 8 or 10 rows did not appear in my race later than the third generation. On the other side the extreme reached 28, a figure never reached by the original variety as cultivated with us, and ears with 24 and 26 rows have been seen during the four last generations in increasing numbers.

This slow and gradual amelioration was partly due to the mode of pollination of the corn. [782] The pollen falls from the male spikes on the ears of the same plant, but also is easily blown on surrounding spikes. In order to get the required amount of seed it is necessary in our climate to encroach as little as possible upon free pollination, aiding the self-pollination, but taking no precautions against intercrossing. It is a.s.sumed that the choice of the best ears indicates the plants which have had the best pollen-parents as well as the best pistil parents, and that selection here, as in other cases, corrects the faults of free intercrossing. But it is granted that this correction is only a slow one, and accounts in a great degree for the slowness of the progression. Under better climatic conditions and with a more entire isolation of the individuals, it seems very probable that the same result could have been reached in fewer generations.

However this may be, the fact is that by repeated selection the strain can be ameliorated to a greater extent than by a single choice. This result completely agrees with the general experience of breeders and the example given is only an instance of a universal rule. It has the advantage of being capable of being recorded in a numerical way, and of allowing a detailed and definite description of all the succeeding generations. The entire harvest of all [783] of them has been counted and the figures combined into curves, which at once show the whole course of the pedigree-experiment. These curves have in the main taken the same shape, and have only gradually been moved in the chosen direction.

Three points are now to be considered in connection with this experiment. The first is the size of the cultures required for the resulting amelioration. In other words, would it have been possible to attain an average of 20 rows in a single experiment? This is a matter of calculation, and the calculation must be based upon the experience related above, that the progression in the case of maize is equal to two-fifths of the parental deviation. A cob with 20 rows means a deviation of 7 from the average of 13, the incipient value of my race.

To reach such an average at once, an ear would be required with 7 x 5/2 = 17-1/2 rows above the average, or an ear with 30-32 rows. These never occur, but the rule given in a previous lecture gives a method of calculating the probability of their occurrence, or in other words, the number of ears required to give a chance of finding such an ear. It would take too long to give this calculation here, but I find that approximately 12,000 ears would be required to give one with 28 rows, which was the highest number attained in [784] my experiment, while 100,000 ears would afford a chance of one with 32 rows*. Had I been able to secure and inspect this number of ears, perhaps I would have needed only a year to get an average of 20 rows. This however, not being the case, I have worked for seven years, but on the other hand have cultivated all in all only about one thousand individuals for the entire experiment.

Obviously this reduction of the size of the experiment is of importance.

One hundred thousand ears of corn could of course, be secured directly from trade or from some industrial culture, but corn is cultivated only to a small extent in Holland, and in most cases the requisite number of individuals would be larger than that afforded by any single plantation.

Repeated selection is thereby seen to be the means of reducing the size of the required cultures to possible measures, not only in the experimental-garden, but also for industrial purposes. A selection from among 60,000-100,000 individuals may be within reach of Burbank, but of few others. As a rule they prefer a longer time with a smaller lot of plants. This