Species and Varieties: Their Origin by Mutation (1906)
Lectures delivered at the University of California
Hugo DeVries
Professor of Botany in the University of Amsterdam




Terminology is an awkward thing. It is as disagreeable to be compelled to make new names, as to be constrained to use the old faulty ones. Different readers may associate different ideas with the same terms, and unfortunately this is the case with much of the terminology of the science of heredity and variability. What are species and what are varieties? How many different conceptions are conveyed by the terms constancy and variability? We are compelled to use them, but we are not at all sure that we are rightly understood when we do so.

Gradually new terms arise and make their way. They have a more limited applicability than the old ones, and are more narrowly circumscribed. They are not to supplant the older terms, but permit their use in a more general way.

[310] One of these doubtful terms is the word sport. It often means bud-variation, while in other cases it conveys the same idea as the old botanical term of mutation. But then all sorts of seemingly sudden variations are occasionally designated by the same term by one writer or another, and even accidental anomalies, such as teratological ascidia, are often said to arise by sports.

If we compare all these different conceptions, we will find that their most general feature is the suddenness and the rarity of the phenomenon. They convey the idea of something unexpected, something not always or not regularly occurring. But even this demarcation is not universal, and there are processes that are regularly repeated and nevertheless are called sports. These at least should be designated by another name.

In order to avoid confusion as far as possible, with the least change in existing terminology, I shall use the term "ever-sporting varieties" for such forms as are regularly propagated by seed, and of pure and not hybrid origin, but which sport in nearly every generation. The term is a new one, but the facts are for the most part new, and require to be considered in a new light. Its meaning will become clearer at once when the illustrations afforded by [311] striped flowers are introduced. In the following discussion it will be found most convenient to give a summary of what is known concerning them, and follow this by a consideration of the detailed evidence obtained experimentally, which supports the usage cited.

The striped variety of the larkspur of our gardens is known to produce monochromatic flowers, in addition to striped ones. They may be borne by the same racemes, or on different branches, or some seedlings from the same parent-plant may bear monochromatic flowers while others may be striped. Such deviations are usually called sports. But they occur yearly and regularly and may be observed invariably when the cultures are large enough. Such a variety I shall call "ever-sporting." The striped larkspur is one of the oldest garden varieties. It has kept its capacity of sporting through centuries, and therefore may in some sense be said to be quite stable. Its changes are limited to a rather narrow circle, and this circle is as constant as the peculiarities of any other constant species or variety. But within this circle it is always changing from small stripes to broad streaks, and from them to pure colors. Here the variability is a thing of absolute constancy, while the constancy consists in eternal changes. Such apparent [312] contradictions are unavoidable, when we apply the old term to such unusual though not at all new cases. Combining the stability and the qualities of sports in one word, we may evidently best express it by the new term of eversporting variety.

We will now discuss the exact nature of such varieties, and of the laws of heredity which govern them. But before doing so, I might point out, that this new type is a very common one. It embraces most of the so-called variable types in horticulture, and besides these a wide range of anomalies.

Every ever-sporting variety has at least two different types, around and between which it varies in numerous grades, but to which it is absolutely limited. Variegated leaves fluctuate between green and white, or green and yellow, and display these colors in nearly all possible patterns. But there variability ends, and even the patterns are ordinarily narrowly prescribed in the single varieties. Double flowers afford a similar instance. On one side the single type, on the other the nearly wholly double model are the extreme limits, between which the variability is confined. So it is also with monstrosities. The race consists of anomalous and normal individuals, and displays between them all possible combinations of normal and monstrous [313] parts. But its variability is restricted to this group. And large as the group may seem on first inspection, it is in reality very narrow. Many monstrosities, such as fasciated branches, pitchers, split leaves, peloric flowers, and others constitute such ever-sporting varieties, repeating their anomalies year by year and generation after generation, changing as much as possible, but remaining absolutely true within their limits as long as the variety exists.

It must be a very curious combination of the unit-characters which causes such a state of continuous variability. The pure quality of the species must be combined with the peculiarity of the variety in such a way, that the one excludes the other, or modifies it to some extent, although both never fully display themselves in the same part of the same plant. A corolla cannot be at once monochromatic and striped, nor can the same part of a stem be twisted and straight. But neighboring organs may show the opposite attributes side by side.

In order to look closer into the real mechanism of this form of variability, and of this constant tendency to occasional reversions, it will be best to limit ourselves first to a single case, and to try to gather all the evidence, which can be obtained by an examination of the hereditary relations of its sundry constituents.

[314] This may best be done by determining the degree of inheritance for the various constituents of the race during a series of years. It is only necessary to apply the two precautions of excluding all cross-fertilization, and of gathering the seeds of each individual separately. We do not need to ascertain whether the variety as such is permanent; this is already clear from the simple fact of its antiquity in so many cases. We wish to learn what part each individual, or each group of individuals with similar characters, play in the common line of inheritance. In other words, we must build up a genealogical tree, embracing several generations and a complete set of the single cases occurring within the variety, in order to allow of its being considered as a part of the genealogy of the whole. It should convey to us an idea of the hereditary relations during the life-time of the variety.

It is manifest that the construction of such a genealogical tree requires a number of separate experiments. These should be extended over a series of years. Each should include a number of individuals large enough to allow the determination of the proportion of the different types among the offspring of a single plant. A species which is easily fertilized by its own pollen, and which bears capsules with [315] large quantities of seeds, obviously affords the best opportunities. As such, I have chosen the common snapdragon of the gardens, Antirrhinum majus. It has many striped varieties, some tall, others of middle height, or of dwarfed stature. In some the ground-color of the flowers is yellow, in others it is white, the yellow disappearing, with the exception of a large mark in the throat. On these ground-colors the red pigment is seen lying in streaks of pure carmine, with white intervals where the yellow fails, but combined with yellow to make a fiery red, and with yellow intervals when that color is present. This yellow color is quite constant and does not vary in any marked degree, notwithstanding the fact that it seems to make narrower and broader stripes, according to the parts of the corolla left free by the red pigment. But it is easily seen that this appearance is only a fallacious one.

CybeRose note: The striped snapdragons were originaged by Vilmorin.

The variety of snapdragon chosen was of medium height and with the yellow ground-color, and is known by horticulturists as A. majus luteum rubro-striatum. As the yellow tinge showed itself to be invariable; I may limit my description to the red stripes.

The strain I received from Mr Fothergill's tended more towards spots.

Some flowers of this race are striped, others are not. On a hasty survey there seem to be three types, pure yellow, pure red, and stripes [316] with all their intermediate links of narrower and broader, fewer and more numerous streaks. But on a close inspection one does not succeed in finding pure yellow racemes. Little lines of red may be found on nearly every flower. They are the extreme type on this side of the range of variability. From them an almost endless range of patterns passes over to the broadest stripes and even to whole sections of a pure red. But then, between these and the wholly red flowers we observe a gap, which may be narrower by the choice of numerous broad striped individuals, but which is never wholly filled up. Hence we see that the red flowers are a separate type within the striped variety.

This red type springs yearly from the striped form, and yearly reverts to it. This is what in the usual descriptions of this snapdragon, is called its sporting. The breadth of the streaks is considered to be an ordinary case of variability, but the red flowers appear suddenly, without the expected links. Therefore they are to be considered as sports. Similarly the red forms may suddenly produce striped ones, and this too is to be taken as a sport, according to the usual conception of the word.

Such sports may occur in different ways. Either by seeds, or by buds, or even within the single spikes. Both opposite reversions, [317] from striped to red and from red to stripes, occur by seed, even by the strictest exclusion of cross-fertilization. As far as my experiments go, they are the rule, and parent-plants that do not give such reversions, at least in some of their offspring, are very rare, if not wholly wanting. Bud-variations and variations within the spike I have as yet only observed on the striped individuals, and never on the red ones, though I am confident that they might appear in larger series of experiments. Both cases are more common on individuals with broad stripes than on plants bearing only the narrower red lines, as might be expected, but even on the almost purely yellow individuals they may be seen from time to time. Bud-variations produce branches with spikes of uniform red flowers. Every bud of the plant seems to have equal chances to be transformed in this way. Some striped racemes bear a few red flowers, which ordinarily are inserted on one side of the spike only. As they often cover a sharply defined section of the raceme, this circumstance has given rise to the term of sectional variability to cover such cases. Sometimes the section is demarcated on the axis of the flower-spike by a brownish or reddish color, sharply contrasting with the green hue of the remaining parts. Sectional variation may be looked at as a [318] special type of bud-variation, and from this point of view we may simplify our inquiry and limit ourselves to the inheritance of three types, the striped plants, the red plants and the red asexual variants of the striped individuals. In each case the heredity should be observed not only for one, but at least for two successive generations.

Leaving these introductory remarks I now come at once to the genealogical tree, as it may be deduced from my experiments:

This experiment was begun in the year 1892 with one individual out of a large lot of striped plants grown from seeds which I had purchased from a firm in Erfurt. The capsules were gathered separately from this individual and about 40 flowering plants were obtained from the seeds in the following year. Most of them had neatly striped flowers, some displayed broader stripes and spare flowers were seen with one [319] half wholly red. Four individuals were found with only uniform red flowers. These were isolated and artificially pollinated, and the same was done with some of the best striped individuals. The seeds from every parent were sown separately, so as to allow the determination of the proportion of uniform red individuals in the progeny.

Neither group was constant in its offspring. But as might be expected, the type of the parent plant prevailed in both groups, and more strongly so in the instances with the striped, than with the red ones. Or, in other words seed-reversions were more numerous among the already reverted reds than among the striped type itself. I counted 2% reversion in the latter case, but 24% from the red parents.

Among the striped plants from the striped parents, I found some that produced bud variations. I succeeded in isolating these red flowering branches in paper bags and in pollinating them with their own pollen, and subjected the striped spikes of the same individuals to a similar treatment. Three individuals gave a sufficient harvest from both types, and these six lots of seeds were sown separately. The striped flowers repeated their character in 98% of their offspring, the red twigs in only 71%, the [320] remaining individuals sporting into the opposite group.

In the following year I continued the experiment with the seeds of the offspring of the red bud-variations. The striped individuals gave 95%, but in the red ones only 84% of the progeny remained true to the parent type.

From these figures it is manifest that the red and striped types differ from one another not only in their visible attributes, but also in the degree of their heredity. The striped individuals repeat their peculiarity in 90-98% of their progeny, 2-10% sporting into the uniform red color. On the other hand the red individuals are constant in 71-84% of their offspring, while 16-29% go over to the striped type. Or, briefly, both types are inherited to a high degree, but the striped type is more strictly inherited than the red one.

Moreover the figures show that the degree of inheritance is not contingent upon the question as to how the sport may have arisen. Bud-sports show the same degree of inheritance as seed-sports. Sexual and asexual variability therefore seem to be one and the same process in this instance. But the deeper meaning of this and other special features of our genealogical tree are still awaiting further investigation. It seems that much important evidence might [321] come from an extension of this line of work. Perhaps it might even throw some light on the intimate nature of the bud-variations of ever-sporting varieties in general. Sectional variations remain to be tested as to the degree of inheritance exhibited, and the different occurrences as to the breadth of the streaks require similar treatment.

In ordinary horticultural practice it is desirable to give some guarantee as to what may be expected to come from the seeds of brightly striped flowers. Neither the pure red type, nor the nearly yellow racemes are the object of the culture, as both of them may be had pure from their, own separate varieties. In order to insure proper striping, both extremes are usually rejected and should be rooted out as soon as the flowering period begins. Similarly the broad-striped ones should be rejected, as they give a too large amount of uniform red flowers. Clearly, but not broadly striped individuals always yield the most reliable seed.

Summing up once more the results of our pedigree-experiment, we may assert that the striped variety of the snapdragon is wholly permanent, including the two opposite types of uniform color and of stripes. It must have been so since it first originated from the invariable uniform [322] varieties, about the middle of the last century, in the nursery of Messrs. Vilmorin, and probably it will remain so as long as popular taste supports its cultivation. It has never been observed to transgress its limits or to sport into varieties without reversions or sports. It fluctuates from one extreme to the other yearly, always recurring in the following year, or even in the same summer by single buds. Highly variable within its limits, it is absolutely constant or permanent, when considered as a definite group.

Similar cases occur not rarely among cultivated plants. In the wild state they seem to be wholly wanting. Neither are they met with as occasional anomalies nor as distinct varieties. On the contrary, many garden-flowers that are colored in the species, and besides this have a white or yellow variety, have also striped sorts. The oldest instance is probably the marvel of Peru, Mirabilis Jalappa, which already had more than one striped variety at the time of its introduction from Peru into the European gardens, about the beginning of the seventeenth century. Stocks, liver-leaf (Hepatica), dame's violet (Hesperis), Sweet William (Dianthus barbatus), and periwinkles (Vinca minor) seem to be in the same condition, as their striped varieties were already quoted [323] by the writers of the same century. Tulips, hyacinths, Cyclamen, Azalea, Camellia, and even such types of garden-plants as the meadow crane's-bill (Geranium pratense) have striped varieties. It is always the red or blue color which occurs in stripes, the underlying ground being white or yellow, according to the presence or absence of the yellow in the original color mixture.

All these varieties are known to be permanent, coming true during long series of successive generations. But very little is known concerning the more minute details of their hereditary qualities. They come from seed, when this is taken from striped individuals, and thence revert from time to time to the corresponding monochromatic type. But whether they would do so when self-fertilized, and whether the reversionary individuals are always bound to return towards the center of the group or towards the opposite limit, remains to be investigated. Presumably there is nowhere a real transgression of the limits, and never or only very rarely and at long intervals of time a true production of another race with other hereditary qualities.

In order to satisfy myself on these points, I made some pedigree-cultures with the striped forms of dame's violet (Hesperis matronalis) [324] and of Clarkia pulchella. Both of them are ever-sporting varieties. The experiments were conducted during five generations with the violet, and during four with the striped Clarkia, including the progeny of the striped and of the monochromatic red offspring of a primitive striped plant. I need not give the figures here for the numerical relations between the different types of each group, and shall limit myself to the statement that they behaved in exactly the same manner as the snapdragon.

It is worth while to dwell a moment on the capacity of the individuals with red flowers to reproduce the striped type among their offspring. For it is manifest that this latter quality must have lain dormant in them during their whole life. Darwin has already pointed out that when a character of a grandparent, which is wanting in the progeny, reappears in the second generation, this quality must always be assumed to have been present though latent in the intermediate generation. To the many instances given by him of such alternative inheritance, the monochromatic reversionists of the striped varieties are to be added as a new type. It is moreover, a very suggestive type, since the latency is manifestly of quite another character than for instance in the case of Mendelian hybrids, and probably more allied to those instances, [325] where secondary sexual marks, which are as a rule only evolved by one sex, are transferred to the offspring through the other.

Stripes are by no means limited to flowers. They may affect the whole foliage, or the fruits and the seeds, and even the roots. But all such cases occur much more rarely than the striped flowers. An interesting instance of striped roots is afforded by radishes. White and red varieties of different shapes are cultivated. Besides them sometimes a curious motley sort may be seen in the markets, which is white with red spots, which are few and narrow in some samples, and more numerous and broader in others. But what is very peculiar and striking is the circumstance, that these stripes do not extend in a longitudinal, but in a transverse direction. Obviously this must be the effect of the very notable growth in thickness. Assuming that the colored regions were small in the beginning, they must have been drawn out during the process of thickening of the root, and changed into transverse lines. Rarely a streak may have had its greatest extension in a transverse direction from the beginning, in which case it would only be broadened and not definitely changed in its direction.

This variety being a very fine one, and more agreeable to the eye than the uniform colors, is [326] being more largely cultivated in some countries. It has one great drawback: it never comes wholly true from seed. It may be grown in full isolation, and carefully selected, all red or nearly monochromatic samples being rooted out long before blooming, but nevertheless the seed will always produce some red roots. The most careful selection, pursued through a number of years, has not been sufficient to get rid of this regular occurrence of reversionary individuals. Seed-growers receive many complaints from their clients on this account, but they are not able to remove the difficulty. This experience is in full agreement with the experimental evidence given by the snapdragon, and it would certainly be very interesting to make a complete pedigree-culture with the radishes to test definitely their compliance with the rules observed for striped flowers.

Horticulturists in such cases are in the habit of limiting themselves to the sale of so-called mixed seeds. From these no client expects purity, and the normal and hereditary diversity of types is here in some sense concealed under the impurities included in the mixture from lack of selection. Such cases invite scrutiny, and would, no doubt, with the methods of isolation, artificial pollination, and the sowing of the seeds separately from each parent, yield [327] results of great scientific value. Any one who has a garden, and sufficient perseverance to make pure cultures during a series of years might make important contributions to scientific knowledge in this way.

Choice might be made from among a wide range of different types. A variety of corn called "Harlequin" shows stripes on its kernels, and one ear may offer nearly white and nearly red seeds and all the possible intermediate steps between them. From these seeds the next generation will repeat the motley ears, but some specimens will produce ears of uniform kernels of a dark purple, showing thus the ordinary way of reversion. Some varieties of beans have spotted seeds, and among a lot of them one may be sure to find some purely red ones. It remains to be investigated what will be their offspring, and whether they are due to partial or to individual variation.

The cockscomb (Celosia cristata) has varieties of nearly all colors from white and yellow to red and orange, and besides them some striped varieties occur in our gardens, with the stripes going from the lower parts of the stem up to the very crest of the comb. They are on sale as constant varieties, but nothing has as yet been recorded concerning their peculiar behavior in the inheritance of the stripes. [328] Striped grapes, apples and other fruits might be mentioned in this connection.

Before leaving the striped varieties, attention is called to an interesting deduction, which probably gives an explanation of one of the most widely known instances of ever-sporting garden plants. Striped races always include two types. Both of them are fertile, and each of them reproduces in its offspring both its own and the alternate type. It is like a game of ball, in which the opposing parties always return the ball. But now suppose that only one of the types were fertile and the other for some reason wholly sterile, and assume the reversionary, or primitive monochromatic individuals to be fertile, and the derivative striped specimens to bloom without seed. If this were the case, knowledge concerning the hereditary qualities would be greatly limited. In fact the whole pedigree would be reduced to a monochromatic strain, which would in each generation sport in some individuals into the striped variety. But, being sterile, they would not be able to propagate themselves.

Such seems to be the case with the double flowered stocks. Their double flowers produce neither stamens nor pistils, and as each individual is either double or single in all its flowers, the doubles are wholly destitute of seed. [329] Nevertheless, they are only reproduced by seed from single flowers, being an annual or biennial species.

Stocks are a large family, and include a wonderful variety of colors, ranging from white and yellow to purple and red, and with some variations toward blue. They exhibit also diversity in the habit of growth. Some are annuals, including the ten-week and pyramidal forms; others are intermediates and are suitable for pot-culture; and the biennial sorts include the well-known "Brompton" and "Queen" varieties. Some are large and others are small or dwarf. For their brightness, durability and fragrance, they are deservedly popular. There are even some striped varieties. Horticulturists and amateurs generally know that seed can be obtained from single stocks only, and that the double flowers never produce any. It is not difficult to choose single plants that will produce a large percentage of double blossoms in the following generation. But only a percentage, for the experiments of the most skilled growers have never enabled them to save seed, which would result entirely in double flowering plants. Each generation in its turn is a motley assembly of singles and doubles.

Before looking closer into the hereditary peculiarities of this old and interesting ever-sporting [330] variety, it may be as well to give a short description of the plants with double flowers. Generally speaking there are two principal types of doubles. One is by the conversion of stamens into petals, and the other is an anomaly, known under the name of petalomany.

The change of stamens into petals is a gradual modification. All intermediate steps are easily to be found. In some flowers all stamens may be enlarged, in others only part of them. Often the broadened filaments bear one or two fertile anthers. The fertility is no doubt diminished, but not wholly destroyed. Individual specimens may occur, which cannot produce any seed, but then others of the same lot may be as fertile as can be desired. As a whole, such double varieties are regularly propagated by seed.

Petalomany is the tendency of the axis of some flowers never to make any stamens or pistils, not even in altered or rudimentary form. Instead of these, they simply continue producing petals, going on with this production without any other limit than the supply of available food. Numerous petals fill the entire space within the outer rays, and in the heart of the flower innumerable young ones are developed half-way, not obtaining food enough to attain [331] full size. Absolute sterility is the natural consequence of this state of things.

Hence it is impossible to have races of petalomanous types. If the abnormality happens to show itself in a species, which normally propagates itself in an asexual way, the type may become a vegetative variety, and be multiplied by bulbs, buds or cuttings, etc. Some cultivated anemones and crowfoots (Ranunculus) are of this character, and even the marsh-marigold (Caltha palustris) has a petalomanous variety. I once found in a meadow such a form of the meadow-buttercup (Ranunculus acris), and succeeded in keeping it in my garden for several years, but it did not make seeds and finally died. Camellias are known to have both types of double flowers. The petalomanous type is highly regular in structure, so much so as to be too uniform in all its parts to be pleasing, while the conversion of stamens into petals in the alternative varieties gives to these flowers a more lively diversity of structure. Lilies have a variety called Lilium candidum flore pleno, in which the flowers seem to be converted into a long spike of bright, white narrow bracts, crowded on an axis which never seems to cease their production.

It is manifestly impossible to decide how all such sterile double flowers have originated. [332] Perhaps each of them originally had a congruent single-flowered form, from which it was produced by seed in the same way as the double stocks now are yearly. If this assumption is right, the corresponding fertile line is now lost; it has perhaps died out, or been masked. But it is not absolutely impossible that such strains might one day be discovered for one or another of these now sterile varieties.

Returning to the stocks we are led to the conception that some varieties are absolutely single, while others consist of both single-flowered and double-flowered individuals. The single varieties are in respect to this character true to the original wild type. They never give seed which results in doubles, providing all intercrossing is excluded. The other varieties are ever-sporting, in the sense of this term previously assumed, but with the restriction that the sports are exclusively one-sided, and never return, owing to their absolute sterility.

The oldest double varieties of stocks have attained an age of a century and more. During all this time they have had a continuous pedigree of fertile and single-flowered individuals, throwing off in each generation a definite number of doubles. This ratio is not at all dependent on chance or accident, nor is it even variable to a remarkable degree. Quite on the contrary [333] it is always the same, or nearly the same, and it is to be considered as an inherent quality of the race. If left to themselves, the single individuals always produce singles and doubles in the same quantity; if cultivated after some special method, the proportion may be slightly changed, bringing the proportion of doubles up to 60% or even more.

Ordinarily the single and double members of such a race are quite equal in the remainder of their attributes, especially in the color of their flowers. But this is not always the case. The colors of such a race may repeat for themselves the peculiarities of the ever-sporting characters. It often happens that one color is more or less strictly allied to the doubles, and another to the singles. This sometimes makes it difficult to keep the various colors true. There are certain sorts, which invariably exhibit a difference in color between the single and the double flowers. The sulphur-yellow varieties may be adduced as illustrative examples, because in them the single flowers always come white. Hence in saving seed, it is impossible so to select the plant, that an occasional white does not also appear among the double flowers, agreeing in this deviation with the general rule of the eversporting varieties.

I commend all the above instances to those [334] who wish to make pedigree-cultures. The cooperation of many is needed to bring about any notable advancement, since the best way to secure isolation is to restrict one's self to the culture of one strain, so as to avoid the intermixture of others. So many facts remain doubtful and open to investigation, that almost any lot of purchased seed may become the starting point for interesting researches. Among these the sulphur-yellow varieties should be considered in the first place.

In respect to the great questions of heredity, the stocks offer many points of interest. Some of these features I will now try to describe, in order to show what still remains to be done, and in what manner the stocks may clear the way for the study of the ever-sporting varieties.

The first point, is the question, which seeds become double-flowered and which single-flowered plants? Beyond all doubt, the determination has taken place before the ripening of the seed. But though the color of the seed is often indicative of the color of the flowers, as in some red or purple varieties, and though in balsams and some other instances the most "highly doubled" flowers are to be obtained from the biggest and plumpest seeds, no such rule seems to exist respecting the double stocks. Now if one half of the seeds gives doubles, and [335] the other half singles, the question arises, where are the singles and the doubles to be found on the parent-plant? The answer is partly given by the following experiment. Starting from the general rule of the great influence of nutrition on variability, it may be assumed that those seeds will give most doubles, that are best fed. Now it is manifest that the stem and larger branches are, in a better condition than the smaller twigs, and that likewise the first fruits have better chances than the ones formed later. Even in the same pod the uppermost seeds will be in a comparatively disadvantageous position. This conception leads to an experiment which is the basis of a practical method much used in France in order to get a higher percentage of seeds of double-flowering plants.

This method consists in cutting off, in the first place the upper parts of all the larger spikes, in the second place, the upper third part of each pod, and lastly all the small and weak twigs. In doing so the percentage is claimed to go up to 67-70%, and in some instances even higher. This operation is to be performed as soon as the required number of flowers have ceased blossoming. All the nutrient materials, destined for the seeds, are now forced to flow into these relatively few embryos, and it is clear that [336] they will be far better nourished than if no operation were made.

In order to control this experiment some breeders have made the operation on the fruits when ripe, instead of on the young pods, and have saved the seeds from the upper parts separately. This seed, produced in abundance, was found to be very poor in double flowers, containing only some 20-30%. On the contrary the percentage of doubles in the seed of the lower parts was somewhat augmented, and the average of both would have given the normal proportion of 50%.

Opposed to the French method is the German practice of cultivating stocks, as I have seen it used on a very large scale at Erfurt and at other places. The stocks are grown in pots on small scaffolds, and not put on or into the earth. The obvious aim of this practice is to keep the earth in the pots dry, and accordingly they are only scantily watered. In consequence they cannot develop as fully as they would have done when planted directly in the beds, and they produce only small racemes and no weak twigs, eliminating thereby without further operation the weaker seeds as by the French method. The effect is increased by planting from 6-10 separate plants in each pot.

It would be very interesting to make comparative [337] trials of both methods, in order to discover the true relation between the practice and the results reached. Bath should also be compared with cultures on open plots, which are said to give only 50% of doubles. This last method of culture is practiced wherever it is desired to produce great quantities of seeds at a low cost. Such trials would no doubt give an insight into the relations of hereditary characters to the distribution of the food within the plant.

A second point is the proportional increase of the double-flowering seeds with age. If seed is kept for two or three years, the greater part of the grains will gradually die, and among the remainder there is found on sowing, a higher percentage of double ones. Hence we may infer that the single-flowered seeds are shorter lived than the doubles, and this obviously points to a greater weakness of the first. It is quite evident that there is some common cause for these facts and for the above cited experience, that the first and best pods give more doubles. Much, however, remains to be investigated before a satisfactory answer can be made to these questions.

A third point is the curious practice, called by the French "esimpler," and which consists in pulling out the singles when very young. It seems to be done at an age when the flower-buds [338] are not yet visible, or at least are not far enough developed to show the real distinctive marks. Children may be employed to choose and destroy the singles. There are some slight differences in the fullness and roundness of the buds and the pubescence of the young leaves. Moreover the buds of the doubles are said to be sweeter to the taste than those of the singles. But as yet I have not been able to ascertain, whether any scientific investigation of this process has ever been made, though according to some communications made to me by the late Mr. Cornu, the practice seems to be very general in the environs of Paris. In summer large fields may be seen, bearing exclusively double flowers, owing to the weeding out of the singles long before flowering.

Bud-variation is the last point to be taken up. It seems to be very rare with stocks, but some instances have been recorded in literature. Darwin mentions a double stock with a branch bearing single flowers, and other cases are known to have occurred. But in no instance does the seed of such a bud-variant seem to have been saved. Occasionally other reversions also occur. From time to time specimens appear with more luxurious growth and with divergent instead of erect pods. They are called, in Erfurt, "generals" on account [339] of their stiff and erect appearance, and they are marked by more divergent horns crowning the pods. They are said to produce only a relatively small number of doubles from their seeds, and even this small number might be due to fertilization with pollen of their neighbors. I saw some of these reversionary types; when inspecting the nurseries of Erfurt, but as they are, as a rule, thrown out before ripening their seed, nothing is exactly known about their real hereditary qualities.

Much remains to be cleared up, but it seems that one of the best means to find a way through the bewildering maze of the phenomena of inheritance, is to make groups of related forms and to draw conclusions from a comparison of the members of such groups. Such comparisons must obviously give rise to questions, which in their turn will directly lead to experimental investigation.