ol li ii -D rr r-=1 i-R D CD r-R a m a a VOL. n. PI. i. ANTIRRHINUM MAJUS LUTEUM RUBRO-STRIATUM, With Bud Variation. VOL. II. PL II. B.C. A-C. -CHRYSANTHEMUM SEGETUM PLENUM, D. CHRYSANTHEMUM SEGETUM. VOL. If. PI. III. O LATA. O. LAMARCKIANA. O. NANELLA ptm.Hood&Larkir-L-.LhLcn.ior: '' VOL. II. PL IV. B A. HYOSCYAMUS PALLIDUS X NIGER. A. H. niger. B. H. pallidus. .n.Hood &Lark:n,Litk,London,W.C , in UJ z CQ D o o cc 1/3 Z H Z O O O o UJ H cc X 6 t, o c . 0) u o: d o o ' THE MUTATION THEORY EXPERIMENTS AND OBSERVATIONS ON THE ORIGIN OF SPECIES IN THE VEGETABLE KINGDOM BY HUGO DE VRIES PROFESSOR OF BOTANY AT AMSTERDAM TRANSLATED BY PROF. J. B. FARMER AND A. D. DARBISHIRE VOLUME II THE ORIGIN OF VARIETIES BY MUTATION WITH NUMEROUS ILLUSTRATIONS AND SIX COLORED PLATES CHICAGO THE OPEN COURT PUBLISHING COMPANY LONDON AGENTS KEGAN PAUL, TRENCH, TRUBNER & CO., LTD. 1910 . . . COPYRIGHT BY THE OPEN COURT PUBLISHING Co. 1910 3 CONTENTS. PART I. THE ORIGIN OF HORTICULTURAL VARIETIES. PAGE I. THE SIGNIFICANCE OF HORTICULTURAL VARIETIES IN THE THEORY OF SELECTION 3 1. Variability in Garden Plants 3 2. The Doctrine of the Increase in Variability in One Direction Brought About by Selection 9 II. LATENT AND SEMI-LATENT CHARACTERS 18 3. Eversporting Varieties 18 4. Half Races and Half Curves 26 5. Trifolinm Pratense Quinquefolium, An Eversporting Race 36 III. THE DIFFERENT MODES OF ORIGIN OF NEW SPECIES 56 6. Horticultural and Systematic Varieties and Elementary Species 56 7. Progressive, Retrogressive and Degressive Formation of Species 65 IV. THE SUDDEN APPEARANCE AND THE CONSTANCY OF NEW VARIETIES 76 8. Examples of Constant Races 76 9. Sterile Varieties 88 10. Instances of Races which Have Arisen Suddenly in Nature 95 11. Horticultural Varieties which Have Arisen Suddenly. 99 V. ATAVISM 104 12. Atavism by Seeds and Buds * 104 13. Vilmorin's Suggestion as to the Origin of Striped Flowers 113 14. Antirrhinum Majus Striatnm 120 iv Contents, PAGE 15. Hesperis Matronalis 136 16. Clarkia Pulchella ; . . . 144 17. Plantago Lanceolata Ramosa 148 VI. EXPERIMENTAL OBSERVATION OF THE ORIGIN OF VARIETIES. . 161 18. The Origin of Chrysanthemum Segetum Plenum .... 161 19. Double Flowers and Flowerheads 194 20. The Origin of Linaria Vulgaris Peloria 201 21. Heritable Pelorias 220 VII. NON-ISOLABLE RACES 227 22. Trifolitim Incarnatum Quadrifolium 227 23. Ranunculus Bulbosus Semi-Plenus 243 24. Variegated Leaves 265 25. Alternating Annual and Biennial Habit 291 VIII. NUTRITION AND SELECTION OF SEMI-LATENT CHARACTERS. 307 26. Increased Nutrition Favors the Development of the Anomaly 307 27. The Influence of External Conditions and of Manur- ing 315 28. The Periodicity of Semi-Latent Characters 323 29. The Choice of Seeds in Selection 332 PART II. THE ORIGIN OF EVERSPORTING VARIETIES. I. TRICOTYLOUS RACES 343 1. The Occurrence of Tricotyls as Half Races and Inter- mediate Races 343 2. Tricotyls, Hemi-Tricotyls and Tetracotyls 356 3. The Influence of Tricotyly on the Arrangement of Leaves 365 4. Tricotylous Half Races 3/9 5. Tricotylous Intermediate Races Do Not Arise by Se- lection 393 6. The Isolation of Tricotylous Intermediate Races .... 417 7. Partial Variability of Tricotyly 444 8. The Influence of External Conditions on Tricotyly . . 450 II. SYNCOTYLOUS RACES 457 9. Hemi-Syncotyly, Syncotyly, Amphi-Syncotyly 457 10. Helianthus Annuus Syncotyleus 466 11. Improvement of a Hemi-Syncotylous Race 476 12. Atavistic Races 481 Contents. v PAGE 13. The Influence of External Conditions on Hereditary Values 485 III. THE INCONSTANCY OF FASCIATED RACES 488 14. The Inheritance of Fasciations 488 15. Half Races with Heritable Fasciation 502 16. Eversporting Varieties with Heritable Fasciation . . . 508 17. The Significance of the Atavists 514 IV. HERITABLE SPIRAL TORSIONS 527 18. Spiral Disposition of the Leaves 527 19. Rare Spiral Torsions 537 20. Spirally Twisted Races 543 21. The Significance of the Atavists 554 PART III. THE RELATIONS OF THE MUTATION THEORY TO OTHER BRANCHES OF INQUIRY. I. THE CONCEPTION OF SPECIES ACCORDING TO THE THEORY OF MUTATION 567 1. Systematic Biology and the Theory of Mutation 567 2. Progressive, Retrogressive and Degressive Mutations. 569 3. The Theoretical Distinction Between Species and Va- rieties 578 4. The Practical Conception of Species 589 5. The Parallel Between Systematic and Sexual Relation- ship 592 II. THE RANGE OF VALIDITY OF THE DOCTRINE OF MUTATION. . 599 6. The Significance of the Available Evidence 599 7. The Explanation of Adaptations 606 8. Vegetative Mutations 614 III. THE MATERIAL VEHICLES OF THE HEREDITARY CHARACTERS 631 9. Darwin's Pangenesis 631 10. Intracellular Pangenesis 639 11. The Pangenes as Bearers of the Hereditary Char- acters 643 IV. GEOLOGICAL PERIODS OF MUTATION 651 12. The Periodicity of Progressive Mutations 651 13. Iterative Formation of Species 661 14. The Biochronic Equation , 663 INDEX 675 LITERATURE. LIST OF THE AUTHOR'S PAPERS BEARING ON THE THEORY OF MUTATION. (See the List at the Beginning of the First Volume.) a. Tntracellular Pangenesis. Translated from the German by Prof. C. Stuart Gager. Chicago : The Open Court Pub- lishing Co., 1910. b. Fluctuating Variability and Mutability. Eine zweigipfelige Yariationscurve. Roux' Archiv fur Entwicklungsmechanik der Organismen, 1895, II, Heft I. Archiv. Neerl., 1895. Sur les courbes Galtoniennes des monstruosites. Bull, scient. France et Bel- gique, 1898, T. XXVII, p. 395. Over het omkeeren van halve Galton-curven. Botanisch Jaarboek, Gent, 1898, X, p. 27. Ueber Curvenselection bei Chrysanthemum segetum. Ber. d. d. bot. Ges., 1899, Bd. XVII, Heft 3. De zaadkweekeryen te Erfurt. Het Nederlandsch Tuinbouwblad, 1891, p. 327. Gladiolus nanceianus, ibid., VIII, Jan. 1892. Tulipa Greigi, ibid., May 1892. Caladium, ibid., July 1892. Caladium's van ALFRED BLEU, ibid., July 1892. Dubbele Seringen, ibid., Sept. 1892. Grootbloemige Canna's I and II, ibid., Dec. 1892. Amaryllis, ibid., IX, Sept. 1893. c. Spiral Torsions. Ueber die Erblichkeit der Zwangsdrehung. Ber. d. d. bot. Ges., 1889, VIII, p. 7. Eenige gevallen van Klemdraai by de Meekrap. Bot. Jaarboek, Gent, 1891, III, p. 74- Monographic der Zwangsdrehungen. Jahrb. f. wiss. Bot., 1891, XXIII, pp. 13- 206, Plates II-XI. Bydragen tot de leer van den Klemdraai. Bot. Jaarboek, 1892, IV, p. 145. Eine Methode, Zwangsdrehungen aufzusuchen. Ber. d. d. bot. Ges., 1894, Bd. XII, Heft 2. On Biastrepsis in Its Relation to Cultivation. Annals of Botany, 1899, XIII, P- 395- d. Fasciations and Other Anomalies. Sur un spadice tubuleux du Peperomia maculosa. Archiv. Neerl., 1891, T, XXIV, p. 258. viii Literature. Over de erfelykheid der fasciatien. Bot. Jaarboek, Gent, 1894, VI, p. 72. Over de erfelykheid van synfisen. Bot. Jaarboek, 1895, VII, p. 129. Krfelyke monstrositeiten in den ruilhandel der Bot. Tuinen. Bot. Jaarboek, 1897, IX, p. 66. Een epidemic van vergroeningen. Bot. Jaarboek. 1896, VIII, p. 66. Sur la culture des monstruosites. Cps. rs. de 1'Acad. des Sc.. Paris, 1899. Sur la culture des fasciations des especes annuelles et bisannuelles. Revue gene- rale de botanique, 1899, T. XI, p. 136. Ueber die Abhangigkeit der Fasciation vom Alter bei zvveijahrigen Pflanzen. Bot. Centralblatt, 1899, LXXVII. Ueber die Periodicitat partieller Variationen. Ber. d. d. bot. Ges., 1899, XVII, Heft 2, p. 45. Over het periodisch ontreden van anomalien op monstreuze planten. Bot. Jaar- boek, 1899, XI, p. 46. Sur !a periodicite des anomalies dans les plantes monstrueuses. Archiv Xeerl., Serie II, T. III. Over verdubbeling van Phyllopodien. Bot. Jaarboek, 1893, V, p. 108. Ueber tricotyle Rassen. Ber. d. d. bot. Ges., 1902, Bd. XX, Heft 2. e. Unit-Characters. ADAM'S Gotiden regen (Cytisus Adami). Album der Natuur, 1894. Hybridizing of Monstrosities. Journ. Roy. Hortic. Soc., 1899. Sur la fecondation hybride de 1'albumen. Cps. rs. de 1'Acad. de Paris, 1899 and Ref. Biol. Centralbl., 1900. Sur la fecondation hybride de rendosperme chez le Mais. Revue generale de botanique, 1900, T. XII, p. 129. Sur la loi de disjonction des hybrides. Cps. rs. de 1'Acad. de Paris, 1900. Das Spaltungsgesetz der Bastarde. Ber. d. d. bot. Ges., 1900, Bd. XVIII, Heft" 3- Ueber erbungleiche Kreuzungen. Ber. d. d. bot. Ges., 1900. Bd. XVIII. Heft 9. Sur les unites des caracteres specifiques. Revue generale de botanique, 1900, T. XII, p. 257. The Law of Separation of Characters in Crosses. Journ. Roy. Hortic. Soc., 1901, XXV. Part 3. On Artificial Atavism. Proceed. Americ. Hortic. Soc., 1902. La loi de MENDEL et les caracteres constants des hybrides. Cps. rs. de 1'Acad. de Paris, 1903. Anwendung der Mutationslehre auf die Bastardirungsgesetze. Ber. d. d. bot. Ges., 1903, Bd. XXI. p. 45. Befruchtung und Bastardirung; ein Vortrag, 1903. Leipsic, Veit & Co. PART I. THE ORIGIN OF HORTICULTURAL VARIETIES. I. THE SIGNIFICANCE OF HORTICULTURAL VARIETIES IN THE THEORY OF SELECTION. i. VARIABILITY IN GARDEN PLANTS. DARWIN based his theory of selection, in great part, on the well-known horticultural principle that new varie- ties are obtained by seeking for small deviations with subsequent isolation and selection. Variations which at their first appearance almost escape observation can be worked up by the skill of the gardener ; in doing so varia- bility is seen to increase, and in favorable cases, very rapidly. In this way a new form arises, which answers the purposes and rewards the labors of the breeder. We have all heard how beautiful double varieties have resulted from the appearance of single flowers in which only one stamen and this often only partially was transformed into a petal. In the first volume we dealt with this practice more than once, and pointed out how liable it is to give rise to misunderstanding when applied to the elucidation of the problem of specific differentiation (Vol. I, 23, pp. 176-185). The object of the present Part is to collate the relevant data and to show what light they throw on this all-important problem. Of course we can only go .^o far as the incomplete and scanty character of the material will allow. 4 The Significance of Horticultural Varieties. The development of the statistical treatment of varia- tion which took place after DARWIN'S time, allows of an altogether different conception of the phenomena than was possible some fifty years ago. It was shown that the fluctuation of characters is due to their development to a greater or less degree. But the character in ques- tion does not vary in any other than these two directions. The variation is linear (Vol. I, p. 118). It increases or diminishes but creates nothing new. New characters can arise, so to speak, alongside of it, but they arise independently of the fluctuation of the old ones. This applies to the case before us. The variations which the horticulturist looks for and then works up are not variations of the old characters', such may indeed give rise, by selection, to improved races, but not to new types (Vol. I, p. 82). The required deviations are anomalies, as in the example of the origin of double flowers, just cited. When such an anomaly arises we may be sure that the new character already existed in the internal organization of the plant. Where it springs from and how it arose is a matter of indifference to the breeder : he has got it and can work it up. In other words : ' 'The first condition necessary for raising a nov- elty is to possess it" (Vol. I. p. 185). , In this connection two cases are distinguished in prac- tice according as one is dealing with apparently invariable forms, or with forms exhibiting a high degree of fluctu- ating variability. In the former case all that has to be done is to isolate the novelty and to free it of possible impurities introduced by crossing. If this can be done without much difficulty the variety is perfect and con- stant from the beginning and needs only a few years of multiplication before it can be put on the market (Vol. Variability in Garden Plants. 5 I, pp. 77-80). Many white flowered varieties afford good examples of this kind of novelty. But it is very different with the second case. A nov- elty which exhibits fluctuating variability in a high de- gree seldom makes its first appearance in a full state of development. As a rule it is very slightly developed at first. The novelty is betrayed, as the expression is, by a quite small trace or indication. From the scientific standpoint we have to regard this as a minus-variant, i. e., as an extreme variant in the minus direction of the new character (Vol. I, p. 51). And it is plain that the seeds of such a variant of the new variety will, when sown in the garden, soon give the mean value of the character in question. This process is, as we can easily see, fundamentally a phenomenon of regression (Vol. I, Figs. 18 and 19, pp. 73 and 84) ; but to the breeder it is a progressive change, and by no means an inconsiderable one, since on it the success of his operations largely depends. This apparent paradox, however, has been a great obstacle to the under- standing of these phenomena. But, to us, it explains in a very clear way the initial and rapid increase in varia- bility; for it is obvious that an approximation to the mean value will take place much more easily and rapidly than a departure from it. The breeder can now either rest content with this "regressive advance"; or he can endeavor to raise the new form above its mean value by choosing plus-variants as seed-parents. But in the latter case the value of the new form remains dependent on the continuance of se- lection (Vol. I, p. 80). Note's dealing with this process of breeding are not rare in horticultural literature, but they are generally 6 The Significance of Horticultural Varieties. short and lack precision being much inferior in this re- spect to the accurate accounts that are given of artificial crossings. I shall bring together the most important facts that I have been able to find, in the following sec- tion (2). In order to penetrate more deeply into these phenom- ena I have endeavored to apply this method to a series of cases. With the help of control experiments, and by keeping detailed records, I succeeded in finding out how such novelties usually develop themselves. Just as hap- pens in practice, I was successful with some cases but not with others. And the correspondence between my results and the experience of breeders seems to me to be so com- plete that my experiments may simply be taken as in- stances of the method under discussion. I propose to distinguish, therefore, between highly variable and only slightly variable novelties. The lat- *ter are generally assumed to be instances of single var- iations which arise suddenly. In the case of these I shall, therefore, only have to discuss their origin and the question of their constancy. (Chapter IV of this Part.) Much more important from the critical standpoint are the varieties with a high degree of fluctuating variability, i. e., those very cases which passed for instances of the origin of new characters by artificial selection (Chapters II and VIII). As examples of this I refer to variegated leaves and to double and striped flowers. If we now compare, from a theoretical standpoint, this high variability with the normal examples which we dealt with before (Vol. I, pp. 47-52 etc.) we shall see that the two are not exactly the same. In variegated leaves the yellow alternates with the green, in semi-double flowers the petaloid stamens alternate with normal ones Variability in Garden Plants. 7 and so forth. Therefore we are not here dealing with the variable development of a single quality, but with the simultaneous operation, or rather with the conflict, of two qualities. For in proportion as the one or the other of them prevails the plant will be more or less variegated, double and so forth. One of the characters is the old, normal one, pertaining to the original species. The other is the new, abnormal one pertaining to the variety in pro- cess of formation, in fact the anomaly. And the conflict of these two antagonistic types affords at least a partial explanation of this extraordinary variability. The green color itself is only very slightly variable, and the pure yellow or golden varieties, in which the green is entirely absent, are equally uniform (varietates anreae, for example Pyrcthnim Partlicniuin aureum). Of "double" forms there are two types ; the ordinary highly variable more or less double sorts, and on the other hand the sterile varieties which exhibit this peculiar- ity to its full extent in all their flowers (see Ranunculus acris petalomana, Vol. I, Fig. 40, p. 194). In this case the types with a high degree of fluctuating variability might be considered as a connecting link between two almost invariable forms, the normal single and the pet- alomanous types. If we regard this principle as an explanation of the case in point we arrive at the conception of intermediate forms with two antagonistic characters strking for the mastery, and possessing a remarkably high degree of vari- ability as a result of this struggle. The extent of this variability differs from case to case : in the most extreme examples whole organs or even whole individuals can come exactly to resemble one of the types between which they oscillate. Pure green or, on the other hand, pure 8 The Significance of Horticultural Varieties. yellow or even white leaves or seedlings are by no means rare in variegated varieties. But the resemblance is only superficial. The green minus variant of the variegated type does not belong to the original species, nor the yel- low plus variant to the golden variety; as may often be seen by sowing the seeds of such extreme types. I propose to call such varieties intermediate races, and if neither of the two antagonistic characters preponder- ates too much over the other, balanced races or ever- sporting varieties 1 (see 3). If we attempt to make a statistical study and graph- ical description of the variability in such intermediate forms we must obviously not expect such simple and straightforward curves as those which describe the var- iability of normal characters (Vol. I, p. 48). In prin- ciple we may expect to obtain figures which simultane- ously exhibit the two magnitudes that is to say com- pound curves such as have been studied by LUDWIG, BATESON, PEARSON, DAVENPORT and others. It is evi- dent that they will present very different forms according to the mutual proportion of the two characters (see be- low 3-5). At the same time it is clear that in such cases selection may lead to special results which will often be due to the impossibility of transgressing the characters of the two limiting types (see 5 and Fig. 3). The two following generalizations may be derived from the facts we have been discussing. 1. Some horticultural varieties owe their existence to a single nezv character. These are usually not more vari- able than the original species and as a rule just as con- stant from seed. Very frequently the novelty consists in the loss or latency of a character of the parent spe- 1 See Species and Varieties, Their Origin by Mutation, p. 309. Increase in Variability in One Direction. 9 cies. In cases where the origin of such a novelty is satis- factorily known it always happened suddenly. For the combination of several characters in the same variety see Vol. I, p. 197. 2. On the other hand some horticultural varieties are compound types which owe their existence to the associa- tion of two (or more) antagonistic characters. The two characters tend to exclude one another more or less com- pletely and struggle for the upper hand; from this there results a very high degree of variability in their mani- festation (as in variegation, stripes, doubleness and so forth). These forms usually first appear as minus var- iants, i. e., with a slight degree of development of the abnormality in which condition they are sought for and isolated and subsequently improved by selection. The artificial production, therefore, of such a form is not a sudden one but a process of gradual improvement Their first origin however remains unknown. 2. THE DOCTRINE OF THE INCREASE IN VARIABILITY IN ONE DIRECTION BROUGHT ABOUT BY SELECTION. One of the most attractive parts of the doctrine of selection is that according to which variability may be increased by selection. Many observations, especially in horticulture, seem to support this view ; which, if it were true would afford an almost irrefutable argument in favor of the prevailing belief in the omnipotence of natural selection ("Vol. I, p. 119). Varieties are said to be incipient species. By selecting the individuals which deviate most from the type of the species it is believed to be possible to attain first to varia- tions and then to varieties. To these is ascribed the 10 The Significance of Horticultural Varieties. tendency to become fixed and so to become races : in the same way these races would later be transformed into species. This is the generally accepted view. This view is based, as I attempted to show in the first part of the first volume, on an unwarranted extension of DARWIN'S theory of selection. DARWIN argued from the results obtained in horticulture ; but these, at least as described in the works of the best authorities, do not seem to me to justify such an extension. According to the prevailing view, man has the power to produce any desired amelioration in any species ready to hand. All characters vary and all that need be done would be to isolate the extreme variants and to breed further from them. The process takes some time of course but in many species the experiment is already lasting about half a century. But the advances which have been made, and which are of the very greatest prac- tical importance, do not tally with this assumption. On the contrary we learn from them that for much that has been attained much has proved unattainable. The comparative studies of systematists show us that almost everywhere there exist imperceptible transitional stages between the smallest differences and perfectly dis- tinct species. This forms a weighty argument for, but no proof of, the prevalent view. For we have to reckon here with transgressive variability (Vol. I, Part II, 25, p. 430), which tends to blur the boundaries of related groups. I have indicated in the foregoing section (1) the principles on which in my opinion an elucidation of the process in question must be based. If a small anomaly is found in a wild or cultivated species, and a new and constant form is raised from this by selection, the whole Increase in Variability in One Direction. 11 sequence of events may have the appearance of having been gradually brought about by the free will of the ex- perimenter ; whereas as a matter of fact the result was attained mainly by good luck. If we look through the literature of horticulture we shall soon see that this illusion has not taken in the really efficient breeder. He knows perfectly well that neither the beginning nor the end of such an experiment is under his control. It is only between these two limits that everything depends on his skill. The first indication of an anomaly in a pure species appears by chance ; and it is a well-known rule in horti- culture that the breeder should always be on the lookout for such chance occurrences. It does not matter how small the deviation is so long as it is an anomaly (p. 4). When such a deviation has once been found it lies with the breeder to perfect it and bring it to its full develop- ment. But the ever present, more or less considerable, fluctuating variations of normal characters are of no use for this purpose; by their means many varieties may be made better and prettier, but they can give rise to nothing really new. The best horticultural authorities are in agreement on this point. CARRIERE for example says : "Uhorticul- tenr ne pent faire naitre les varietes" and in greater de- tail in reference to double flowers: "Le point de depart des fleurs donbles est en dehors de notre puissance coin me de nos calcnls; nons ne ponvons rien, on a pen pres rien, sur le fait initiatif; nons ne ponrons qnc le saisir lorsqn'il se presente; nons ne pouvons pas le provoqner; c'est nn effct, dont la canse nous est inconnnc." 1 A well-known 1 E. A. CARRIERE, Production et fixation dcs rarictcs dans les vegctaux, 1865, p. 64 and p. 15. 12 The Significance of Horticultural Varieties. English breeder, WILLIAM PAUL, says : l 'He who is seeking to improve any class of plants, should watch narrowly and seize with alacrity any deviation from the fixed character However unpromising in appearance at the outset, he knows not what issues may lie concealed in a variation." SALTER also said that the greatest diffi- culty lies in finding a small initial deviation ; but when this has once been found all the rest lies within our power, however small the variation may be. And DARWIN, who cites this, 2 always emphasized its great importance when- ever he had occasion to refer to it. In other words, which we have already often quoted : The main condition necessary to produce a novelty is to be in possession of its first step. And yet as is well known the attempt is not by any means always successful. Sometimes the variation dis- appears without leaving a trace behind ; in which case of course all further efforts to deal with it are in vain. Unfixable deviations of this kind are, according to my experience, the occasional manifestation of latent characters. What the breeder wants to find are those cases in which the chance anomaly has already become a heritable although hidden race. If this has happened the anomaly will, in the first place, easily manifest itself, if the conditions of life are not quite unfavorable and in the second can rapidly be developed to the level of a good horticultural variety. So far as the available data enable -us to judge, breed- ing experiments of this kind always follow the same course. Hosts of examples can be found. Extensive 1 Contributions to Horticultural Literature, 1892. Nature, Vol. 46, P- 583- 2 Variations of Animals and Plants, II, p. 249. See also Part I, p. 267 et seq. Increase in Variability in One Direction. 13 sowings repeated year after year avail nothing if chance does not play its part. Anemone coronaria plena arose in the nurseries of WILLIAMSON in England as a single plant, which exhibited a slight petaloid broadening of one of the stamens. 1 From the seed of this specimen a race has been started, the flowers of which became fully double in the course of a few generations. The double varieties of roses, Campanulas, and many other garden plants have arisen in the same way. I saw a bed of mignonette (Reseda odorata) some of which had double spikes, in a nursery at Erfurt. The spikes were fasciated, the flowers were broader and the whole plant fuller, more compact and handsomer than the species. The plants of this bed had been produced from the seed of two fasciated specimens which had accidentally appeared the year be- fore. The normal were weeded out and the abnormal saved and allowed to set seed with a view to putting a new r variety on the market. In cases such as this, selection has a twofold object. In the first place the variety must be isolated, that is purged of the impurities resulting from free crossing. It must also be actually improved by selection. The first indications of doubling are, as we have seen, single super- numerary petals, or in composites single supernumerary ray florets on the disc ; the first indication of a new color is often very pale : slit leaves and petals are indicated by quite small imaginations, combs (Vol. I, p. 191) appear as small outgrowths. All these qualities had to be im- proved by selection up to the level of the mean of the character and then even perhaps beyond. An improvement of this kind, when once started, is effected not only rapidly but with increasing swiftness. DARWIN, he. cit., TT, p. 269. 14 The Significance of Horticultural 1 \irietics. Hence the illusion of an increase in variability. The ex- planation is simply this that, as shown in the preceding section ( 1), we first find a minus variant of the new character, which, in accordance with the law of regres- sion, approaches not the character of the old species but the mean value of the new variety, as soon as it is iso- lated. And this takes place easily and swiftly since the new variety in this case behaves like an improved race on the cessation of selection or under reversed selection (Vol. 1, 14, p. 122). The progress made by this improvement and through the purification from the results of crossing is often so rapid that it can be expressed in terms of a geometrical series. This generalization does not attain to the rank of a law, but my meaning will become clearer by citing an example. HOFMEISTER sowed the seeds of plants of Papaver somniferum polycepJialiim, 1 which he had found growing between normal examples of the species. By selecting the fruits which were richest in supernumerary carpels, but without isolation, he effected the following increase in the number of abnormal examples in the suc- ceeding generations : Year: 1863 1864 1865 1866 1867 Percentage of abnormal plants: 6% 17% 27% 69% 97% Geometric series: 8 16 32 64 (100) These figures, as we see, do not differ considerably from a geometric series. I do not lay much stress on the fact, but I have myself more than once obtained similar series of figures in breeding experiments. The limits that can be reached are as little under the control of the breeder as the starting-points that had to 1 Allgememe Morphologic, p. 565. See our Fig. 27 on p. 138 of the first volume; also HOFFMANN, Bot. Zcitg., 1881, p. 397, and VER- LOT, Production et fixation des varietes, p. 88. Increase in Variability in One Direction. 15 be waited for. This is most forcibly brought out by the fact that numerous horticultural varieties are still at exactly the same level as they were at the time of their introduction. The most vigorous selection continued over long periods of time has only rarely succeeded in effecting a further improvement in the same direction. We are familiar with hosts of variegated plants, but Aurea vari- eties are very rare. Flowers with petalomany are sterile, and the plants can only be multiplied by vegetative meth- ods. But it is quite clear that this difficulty is by no means the cause of their rarity. Amongst composites we occa- sionally find isolated heads without tongue florets, but how small is the number of discoid varieties. I once found an example of Coreopsis tinctoria in my cultures, which exhibited only some spare ray florets, but although I isolated the plant, the abnormality did not reappear from its seed. Catacorolla (an outward doubling of the corolla so as to form lappets) occurs almost only as a commercial race in Gloxinia superba. Fistulous compo- sites are rare ; there is room on the market for monoph- yllous and laciniate varieties of many species, if only we could make them. But so long as chance does not put them into our hands, all our labor is in vain. Nevertheless, all plants no doubt possess numerous latent characters. Any culture carried out on a sufficiently large scale, or continued for several years, will give con- vincing proof. In fact it is often very difficult to keep races free from anomalies. Agrostemnia Githago, Raph- anus Rhaphanistnun and many other species contain an almost inconceivable number. Amongst garden plants desirable novelties must obviously be rare now because thev must have been already found and put on the market ; 16 The Significance of Horticultural Varieties. useless and indifferent anomalies are common enough, especially in extensive cultures. When a new horticultural variety has been isolated and "fixed," that is to say improved and rid of impuri- ties by a few years' cultivation no considerable further improvement in the same direction is to be expected. Only two ways of progress are still at hand. These are to wait for the chance appearance of a new abnormality in the same strain, or to combine the new character with others by crossing. The former method is dependent on chance and therefore often unsatisfactory. The sec- ond is almost sure to succeed, and thus it is always chosen. Each new character is immediately transferred to nu- merous other varieties of the species and a corresponding number of novelties obtained in this way. Thus LE- MOINE transferred the double flowers of a single double lilac to several dozen varieties, and the Cactus Dahlia was, very soon after its introduction, obtainable in almost every shade of color and doubleness. Ordinarily this process is described in the opposite way that is to say, it is claimed that the properties of the old varieties are transferred to the new type. In this way there appears a vast series of varieties forming a new group co-exten- sive with the older forms of the original species. Thus a single new character can double the number of varieties. Petunias, Zinnias and Fuchsias are familiar examples of the application of this method in former times, Gladiolus, Begonia and many others of its recent application. The ostrich-feather Chrysanthemum (with ciliated petals) arose about thirty years ago in a single variety (Alph. Hardy), but can now be obtained in large numbers of forms. The doctrine of the onesided increase of variability Increase in Variability in One Direction. 17 selection is based, therefore, as far as the available data enable us to decide, on the existence of strains with heritable but hitherto latent characters. Such races are highly variable, and their existence is betrayed when they first are met with, by trifling anomalies which however can easily be worked up by selection. As a result they rapidly depart from tlie type of the species but only be- cause they approach their neiv type: and as soon as this has been reached bv isolation or exceeded bv selection j *f it is just as difficult to effect any further improvement as in ordinary improved races. These varieties cannot be evoked at will; we have to wait till they chance to ap- pear. Nor when once fully developed can we improve them further. Nothing but chance that is to say some unknown factor can as yet overstep these two limits; selection can effect no more than the most transparent illusion of any thing approaching complete control. II. LATENT AND SEMI-LATENT CHARACTERS. 3. EVERSPORTING VARIETIES. Before I proceed to deal with the results which have been obtained, in horticulture, with these highly variable varieties it is desirable, in order to clear up the concep- tions involved, to fix our attention on the various stages which may be interpolated between a species and a simple and constant variety derived from it. We will start from the fact that the chance appear- ance of an anomaly by no means always opens up the way to the acquisition of a novelty. One example out of many will suffice. Pitchers (Figs. 16, 106, and 109, Vol. I, pp. 61, 470, 484) are usually found as quite rare and isolated variations, 1 but in some species of plants, such as Magnolia and Tilia, tolerably frequently. But a vari- ety as rich in these structures as, for example, Tri folium pratensc quinquc folium is in 4- and 5-merous leaves does not exist, although it would obviously attract attention and pay the trouble of breeding experiments. 2 This shows that an anomaly discovered by chance may be the expression of a latent character which cannot be brought to its full state of development. Besides this 1 Over de erfelykheid van synfisen, Knildkundig Jaarbock, Gent, 1895, P- 129- 2 A variety of Ficus religiosa, with all its leaves changed into pitchers, has since been introduced into Europe by Mr. PRAIN, the Director of Kew-gardens. (Note of 1910.) Ever sporting Varieties. 19 extreme but very common mode of appearance two other cases are possible, according to my experience : 1. When the seeds of an abnormal individual are sown the anomaly is repeated from time to time in a few or more individuals, remaining rare or only appearing in a feeble state of development. Selection may improve it, but only to a very inconsiderable extent. 2. Under favorable circumstances the anomaly may increase rapidly both in the degree of its development and in the number of individuals which present it. A so-called constant race is formed in the course of a few genera- tions. It is subject to a high degree of fluctuating vari- ability in respect to the character in question and is largely dependent on cultivation. I propose to term the first type of characters scini- latcnt and to distinguish amongst latent characters be- tween the genuine completely latent ones and those which occasionally come to light or the semi-latent ones. This term refers to the behavior of the character in the race as a whole; a semi-latent character may remain latent in many individuals and organs and be active in others. A true latent character on the other hand only very rarely becomes active. If we study these three cases statistically, trying to plot the variation of the anomaly in the form of a curve (p. 8) we generally obtain the following results: First cose. The genuine latent characters appear so rarely that they do not afford sufficient material for a curve. Second case. Semi-latent characters must be studied in combination with their antagonistic active characters, and are expressed by half curves (Fig. 1, p. 28), from which a two-sided curve mav be derived bv selection 20 Latent and Semi-Latent Characters. (Fig. 2, p. 34), the apex of which however does not become very distant from that of the half-curve. Third case. The characters are first expressible by half curves because they are minus variants; but after isolation the curve very soon becomes a two-sided one with a new apex. The new variety reaches its full de- velopment and is maintained without further selection. A schematic presentation of the conflict between two antagonistic characters is given below : The normal character is: The anomaly is: I. active latent. II. active semi-latent. III. An equilibrium is maintained. IV. semi-latent active. V. latent active. I do not of course suppose that no further cases are possible, that there may not for example be various stages of semilatency. The facts at our disposal do not admit of any such definite statement. On the other hand it must be stated that the scheme I have given covers the cases which have been so far collected ; we shall soon see large numbers of examples of the main cases, whereas of others I have not yet found any. In the above table I obviously represents the normal, original species, and V the slightly variable and constant variety derived from it. The three other numbers repre- sent the three intermediate forms of which the two first (II and III) correspond to actuality whilst the fourth merely follows from the scheme. I am rather doubtful as to its occurrence in nature. It is necessary to introduce special names for the first two intermediate forms. I shall therefore call them both intermediate races, one of which- No. II I shall Eversporting Varieties. 21 call a half race, and No. Ill a middle race. The word race is obviously not used here in the sense of an im- proved rare (Vol. I, p. 80) but simply means a heritable form. Instead of middle race I shall usually employ the more convenient term of eversporting variety. 1 Two examples to which reference has already been made will serve to illuminate the foregoing discussion. EXAMPLES. VARIEGATED LEAVES. DOUBLE FLOWERS. I. Original species. Green. Simple. II. Half -race. Rarely variegated. Occasional petaloid stamens. III. Eversporting variety. Var. variegata. Var. plena. V. Constant variety. Var. aurea. Var. petalomana. The parallelism of these two groups rests on the assumption that the same character appearing in a state of full development would give rise to the constant golden and to the fully double varieties; 2 and that it is by their mixture with the antagonistic character that variegated and half-double varieties arise. The object of this as- sumption is solely to present the matter more clearly ; for in cases of segregation the characters behave slightly differently (see p. 124). There are many examples of half races and ever- sporting varieties; the former constitute a very con- siderable part of the material of teratology and afford suitable material for the experimental study of monstros- ities. The same holds good for many eversporting vari- eties, and I shall have to recur to this point in the second part of this volume with especial reference to twisted stems and fasciations. Half races as a rule exhibit their 1 See Species and Varieties, Their Origin by Mutation, Chapters XI-XV, pp. 309-459- 2 See 19 and especially 24 (on variegation). 22 Latent and Semi-Latent Characters. abnormality too seldom to be of any use, or at any rate to be of more than of secondary value, in horticulture. On the other hand the eversporting varieties highly con- tribute to the diversity among horticultural plants. Nu- merous varieties with variegated leaves, with striped or double flowers, with double heads amongst the compo- sites, belong to this group. The Formae cristatae of many ferns, the combs in the flowers of Primula sinensis, Cyc- lamen persicum, Begonia etc., the polycephaly of Rapaver, the catacorolla of Gloxinia superba, and a series of other more or less rare instances may also be adduced. It is, obviously, not necessary that all the forms named should exist for every pair of antagonistic characters. In many cases the intermediate races are absent and in J others one or two of them. It is, likewise, not necessary that the pure type corresponding to a certain intermediate race should exist. We can, in such cases, very often reconstruct it by the help of analogy. The following are instances which will be described more fully later on in this part, in which the corresponding constant vari- ety is still failing. SPKC.BS. HA LF - RACE . Tri folium pratense wild four-leaved T. p. guinque folium. clover Trifolium incarnatum T. i. quadrifolium unknown. Ranunculus bulbosus R. b. semiplenus unknown. Chrysanthemum inodcrum unknown C. i. plenissimum. Chrysanthemum segetum C. s . grandiflorum C. s. plenum. Caltha palustris furnishes another instance; it ex- hibits in nature a half race with supernumerary petals and is represented on the market by a uniformly double sterile variety exhibiting petalomany. Camellia japonica presents the two types of doubling in different varieties. Eversporting I 'arietics. 23 The remarkable fistulous and monophyllous varieties, so well known as examples of partial atavism, are further instances of eversporting types (Vol. I, Fig. 38, p. 193, and Fig. 15 of this volume), together with the viviparous grasses (Poa alpina liuipara, Poa bnlbosa livipara, etc.) and a number of other viviparous forms such as Agave vk'ipara and so forth. 1 If the constant variety corre- sponding to a certain intermediate race does not exist, this latter is usually classed as a variety in the case of middle races, but as a heritable anomaly in the case of half races. It is, further, very probable that many natural spe- cies which attract attention by the high degree of vari- ability of some particular character are really in a way intermediate races, i. e., that they owe their multiformity to the co-existence of two antagonistic characters. In- stead of entering further into this very attractive subject I shall content myself with citing the case of Acacia dk'ersi folia which owes its name and its nature to the conflict between the two characters of bipinnate leaves and flattened petioles without leaflets (phyllody of the stalks). The question of the constancy of these intermediate races is a very important one. I propose to deal with it when referring to individual cases in detail; and the only general statement I shall make now is that both con- stant and inconstant intermediate races exist. On the one hand there are those cases in which an overstepping of the limits between these two races is apparently as rare as the mutations by which new species arise, and 1 See GOEBEL, Organografihie, I, pp. 153-159; E. H. HUNGER, Ueber einige vivipare Pflamen. Diss. Rostock, 1887. Bot. Jahresber., 1888, T. XVI, I, p. 421, and also, especially, CLOS, in Actcs du congrcs international dc botaniquc, Paris, Sept., 1900, p. 7. 24 Latent and Semi-Latent Characters. in which at least, in spite of every precaution and care, I have not yet succeeded in obtaining the one race from the other. (Trifolium incarnatum qnadrij oliiim , T. pra- tensc qiiinqucfolium, Ranunculus bulbosus scmiplcnus.) On the other hand are those races which when cultivated on a sufficiently large scale give rise every year to indi- viduals which seems to overstep the otherwise fixed lim- its of the race. These are therefore inconstant inter- mediate races. I regard this phenomenon as one of atavism, at any rate in those cases where, as in my own observations, they revert from an eversporting variety to the type of the parent species without however ac- quiring the constancy of the latter. Atavistic phenom- ena of this kind are well known in striped flowers and variegated leaves ; and I have also found very striking examples of it in Linaria vulgaris pcloria and Plantago lance olata ramosa (20 and 17). Besides the cases which fall into the two categories just discussed, I succeeded in finding a third in which one intermediate race arose from the other very rarely and only in isolated cases. I have seen two cases of this so far. One was the origin of Linaria vulgaris pcloria from L. v. hemi'peloria ( 20); the other was the formation of the double Chrysanthemum segetum plenum (Plate II), from C. s. grandiflorum with 21 instead of 13 tongue-florets ( 18). Linaria vulg. peloria is probably an intermediate race, on account of its inconstancy; whereas L. vulg. hcniipeloria (with stray peloric flowers) is obviously a half race. The origin of the former from the latter presumably occurs in nature from time to time. My Chrysanthemum segetnm plenum is a novelty in the horticultural sense of the term, being just as double as the double varieties of other composites; so far as I Eversporting Varieties. 25 know it has not as yet arisen anywhere else. It consti- tutes an eversporting variety like a number of other double composites which are analogous to it; and arose in my experimental garden, not from the original species, but from a variety known in the trade as C. s. grandi- flonun, which forms a first step towards it in respect of the number of its tongue florets, and is therefore to be regarded as a half race. 1 Let us now briefly summarize the foregoing dis- cussion : 1. There exist both in the cultivated state and in nature a series of forms which are either not constant or highly variable, a state of affairs which is probably due to the interaction of two antagonistic characters. 2. Of these two characters one is to be regarded as normal, that is to say, as belonging to the parent species ; the other as the abnormal. 3. Where the former preponderates, teratological half races with their half curves are the result. 4. If the two maintain an equilibrium, there are formed what I have called middle races, intermediate races, or eversporting varieties, of which many examples are to be found amongst garden varieties and "heritable" teratological races. 5. The high degree of fluctuating variability of the eversporting variety, its occasional discovery in nature and in cultivation, and the possibility which it affords of the working up of striking novelties by means of iso- lation and selection, afford an explanation of the major- 1 The numerous apices of the curves describing variation in the number of rays in composites, which have received no explanation so far, tend however to make the application of this conception diffi- cult See also the origin of Dahlia variabilis fistulosa in my cultures ( ii, p. 100. 26 Latent and Semi-Latent Characters. ity of the phenomena which led DARWIN to his theory of the slow transformation of species. For at that time it was believed that the inception of this process was to be sought in the variation of a character already exist- ing, whereas as a matter of fact the variation in question is independent of the fluctuation of the existing char- acters. 6. The origin of a constant variety or a new species could be easily imagined to occur in this way: First a half race would arise from a pure race, then from this half race a middle race and lastly, from this latter, a new constant form. But this would be pure fancy, since it is without any basis of fact. Besides in many cases the intermediate stages are entirely wanting. 4. HALF RACES AND HALF CURVES. The study of anomalies must be based on the theory that external factors can only be efficient in altering the form of the plant if the power to react to them (or the potentiality for the change) is already present. 1 'The induction of malformations by external causes is no more than the manifestation of latent potentialities," says GOE- BEL. 2 Every plant possesses a whole host of such latenV potentialities. A single plant of Plantago lanceolata may be ramosa, stipitata, and bracteata\ it may have splil leaves and pitchers composed of one or two leaves; and it may exhibit abnormal twisting and forked ears, or present a whole series of other anomalies. The seeds of a single self-fertilized plant will very often give rise to 1 See Intracellulare Pangcnesis, p. 194. 2 GoEBEL, Organo graphic, p. 158. Half Races and Half Curves. 27 a whole series of malformations. Many cultivated plants, such as Cyclamen, Pelargonium and Fuchsia, are particu- larly productive of such abnormalities. The internal factors may either be latent or semi- latent. In the former case the characters are either not manifested, or only exceptionally, as in the pinnate leaves of the red clover (Fig. 46) and as in the numerous cases of pitchers which have been found once, or only at long intervals, in the same species. In the second case they appear more or less regularly, often yearly, and in many specimens. For example I have observed the for- mation of pitchers on Magnolia obovata in the various botanical gardens which I have visited ; and this species as well as its near allies bears pitchers with us every year. 1 In both cases these potentialities are heritable. This is proved by their frequence in the case of the semi-latent characters and rendered extremely probable in that of the latent ones by their occasional reappearance. Latent arid semi-latent characters constitute what we may call the outer range of the forms of a species. The inner range of forms consists of the normal characters of a species which are exhibited during its normal life or are only induced by such common stimuli as w r ounds, mutilations, darkness, or the uncovering of subterranean organs and so forth. They are part of the innermost essence of the species. But the countless latent charac- ters belong just as much to the essence of the species, especially when they have formed part of the inner range in some remote ancestor and are therefore atavistic. And it is just this outer range which presents the best indica- 1 Over de erfelykheid van Synfisen, Bot. Jaarb. d. Gesellsch. Do- donaea, Gent, 1895, P- I2 9- In the course of ten years I have observed about 100 pitchers on Magnolia. 28 Latent and Semi-Latent Characters. tions of descent and therefore of systematic relationship. It fully deserves and repays the attention given to it, as CELAKOWSKY'S admirable papers show. It is to be hoped that others will, following the lines laid down by HEIN- RICHER, undertake the task of rendering these characters more amenable to study by cultivation, and so bring an increased number of them to light. The manifestations of latent characters are so rare that they scarcely ever lend themselves to statistical study (p. 19). When they recur from time to time they are seen to be extremely vari- able, since as a rule even the B I rarest anomalies are not quite the same each time they appear. It is easily seen in such cases that the varia- bility is a unilateral one; but the construction of curves usually fails owing to the sparsity of the material. The half races are much more favorable in this re- spect. Here the deviations are by no means so very rare. As a rule the normal character still preponderates, but material sufficient for statistical study can often be found without difficulty. It shows clearly that the variation is a unilateral one. The apex of the curve is the mean of the normal character, and the deviations all lie on the same side. And in ordinary cases they are less numerous 1 Ber. d. d. hot. Ges., Vol. XTT, 1894, P- IQ7. Plate X. 5678345 Fig. i. Half Curves. A, Caltha palustris. Curve of the num- ber of petals in 416 flowers. B, Weigelia amabilis, Curve of the slips of the corolla in 1145 flowers. 1 Half Races and Half Curves. 29 the further they deviate from the type of the species. Fig. I gives a couple of examples at A and B. A gives the number of petals of Caltha palustris in a locality not far from Hilversum; the flowers, where the species is pure, are pentamerous. But in this place there occurred flowers with 5-8 petals in the following proportions : Flowers with 5678 Petals. Relative number 72% 21% 6% 1% Weigelia amabilis, also, has normally pentamerous flowers ; but it often varies in a minus direction. I found in 1145 flowers on three bushes in our garden (Fig. IB): Number of slips in the corolla 345 Number of flowers 61 196 888 Half curves differ from the half of a normal curve because the height of the mean, i. e., the number of nor- mal cases, is too great. Such curves do not display the variability of the character given by the highest ordinate, but that of another character which is concealed in the normal flowers. 1 Half- or unilateral curves are widely distributed in nature. Where they occur they point to the existence of half races. Nevertheless middle races can, under cer- tain circumstances, as we have already pointed out (p. 20) exhibit half curves; just as, on the other hand, the half 1 Half curves are therefore compound curves. Their apex cor- responds to the mean value of the normal character ; their flank is the expression of the semi-latent character. If the normal character, in the material at our disposal, does not vary it has no curve of its own, which accounts for the absence of a flank on the other side. This for example is the case for curves based on numbers, when the nor- mal number is constant or practically constant as in the case of the three-leaved clover or pentamerous flowers. If the normal character is distinctly, though slightly, variable, as in the case of data based on measurements, the half curve has a flank on the other side, but it is very steep. I do not propose to pursue this point any further here, since it is merely my object to show that half curves are only a special case of asymmetrical curves. 30 Latent and Semi-Latent Characters. curve of a half race can be tran formed into a bilateral curve by selection and high nutrition. I shall recur to this point shortly. The well-known researches of FRITZ MULLER with Abutilon give instances of half curves. 1 MULLER ob- tained the following figures from a culture with seeds from flowers with six petals: 145 with 5 petals, 103 with 6, and 13 with 7. Of more recent investigations we may mention those of BATESON and PERTZ with Veronica BiiA'bannin according to which the normal cases always composed 70-90% of the culture in spite of the selection of the extreme variants in petal-number as seed-parents, the remaining 30-10% being composed of abnormal cases in a rapidly diminishing series. 2 The fruits of Aqmlegia are pentamerous, but 6-, and still more rarely 7-merous, ones occur. The fruit of the cotton is also pentamerous, but I have found several tetramerous and occasional trimerous ones. Papavcr Argenwne has tetramerous flowers, but specimens with 5, and less often with 6 petals, also occur ; by sowing seeds from the latter I was not able to obtain any increase in the number. Duplications of leaves, concrescence of umbel-rays in UmbclUferae, of the fruit stalks of Cruciferae, of the fruits themselves in the Compositac and so forth, the adnation of an axial bud with its axillary branch and a number of other anomalies behave as half races. The abnormal cases, which are of course infinitely rarer than the normal ones, become rarer in proportion as they de- part from the normal. It is unnecessary to give a longer list here, I may just mention the catacorolla on the outer 1 HERMANN MULLER, Die Befruchlung dcr Bhimen, p. 450. 2 W. BATESON and Miss D. F. M. PERTZ, Notes on the Inheritance of Variation in the Corolla of Veronica Bu.vbaumii. Proceed. Cam- bridge Phil. Soc., Vol. X, Pt. II, p. 78. Half Races and Half Curves. 31 side of the corolla in a half race of Linaria vulgaris which I have studied for a few generations, and for which the half curves have recently been plotted and in- vestigated by GARJEANNE. 1 It is w r ell known that every species has a tendency, as the expression is, to vary in certain definite directions ; in these the deviations occur fairly frequently, in others either not at all or very seldom. The number of anom- alies is by no means an unlimited one for a given species, hut strictly limited. One expression of this phenomenon is the fact that one species tends to produce and repeat one particular abnormality, and another species, another. This general fact, with which we are familiar in vague expressions of this kind, can be made the starting point of valuable experimental investigations. For what are we to understand by "tendency" in these cases? In my opinion simply the existence of a half race or sometimes even of an eversporting variety. These two types of races are, so far as my experience reaches, perfectly dis- tinct, and in numerous cases amenable to experimental study; they are things with nothing intrinsically vague about them although they are sometimes blurred in their manifestation, under a superficial examination, on ac- count of the high degree of fluctuating variability which they exhibit. > If we take a plant which presents this tendency to a particular anomaly and cultivate its progeny, isolating it with an eye to this tendency, we shall usually find that we are dealing with an intermediate race of the kind of which we have spoken. I shall refer to an instance in the fol- lowing section ( 5) ; but this will be only one out of 1 A. J. M. GARJEANNE, Beobachtungen und Culturversuche ilber clue Bluthenanomalie von Linaria vulgaris. Flora, 1901, Vol. 88, p. 78; with Plates IX and X. 32 Latent and Semi-Latent Characters, many. It is frequently uncertain, at first at any rate, whether besides the half race, the "species" itself exists in pure condition, that is to say, a race in which the char- acter in question is not semi-latent but latent. But when, as is so often the case, the species is widely distributed but the half race is only observed locally, we are evi- dently fairly safe in assuming the separate existence of both. Anomalies which are very common in nature point to the existence of eversporting varieties ; those which are rare, to half races. In the former case they are often reckoned among the characters proper to the species, as for instance the remarkable lateral fruitlets on the fruits of Tetragonia expansa, which were included by DE CAN- DOLLE in his diagnosis of the species, in his Prodromus. 1 Other well-known instances are the incomplete apetaly of Ranunculus auriconius, 2 as well as the branched ears of Lolium perenne ramosum which seem to be relatively common everywhere in my own country. LENECEK S records lime-trees with 20-30% of their leaves trans- formed into pitchers; and with us trees with single pitchers, and others which produce large numbers of them every year are met with from time to time (Vol. I, Fig. 106, p. 470). In many cases we know both the half race and the middle race of the same, or of closely related, species. For example, there grows very commonly here a form of Plantago major (/. bractcata] which bears more or A. DE CANDOLLE, Prodromus Regni Vegetabilis. See also EICH- LER, Bli'tthendiagramme, II, p. 120. z WINTER, Journ. of Bot., Vol. 35, 1897, P- 406. This form also grows in our garden and in our country in the wild condition. 3 O. LENECEK, Mitth. d. naturw. Vercins, Vienna, 1893. Found not far from Leitmeritz. Half Races and Half Curves. 33 less numerous green bracts on the lower parts of the spikes. The well-known Plantago major rosea of our gardens, all of the bracts of which are green and fairly large, constitutes the complementary, and constant, ever- sporting variety. Besides Papavcr somniferum polyceph- alum (Vol. I, Figs. 27-28, pp. 138-139) which is to be regarded as an eversporting variety, there are polyceph- alous half races of P. commntatnm and several other spe- cies which in my cultures behave in quite a different man- ner from the former, in response to selection. Besides the favorite Varietates cristatae of our cultivated ferns we occasionally find, in nature, wild species with a split leaf. Cclosia cristata, the cockscomb, is an exceedingly interesting eversporting variety, 1 besides which fasciated half races in numerous other genera are known. 2 But I must refrain from the citation of further instances. Just as a species can as a rule be distinguished from its nearest allies by two or several characters, so a half race can manifest as semi-latent anomalies two or more characters which are latent in the species in question. Nor is this by any means rare. In the case of characters which deviate in the opposite direction from the type of the species, "double half-curves" may be formed which have two unequal flanks. The number of petals of Hypcricum perforatum varies in this way, in this neigh- borhood, round a mean of 5 ; on the one side going fre- quently to 4 and rarely to 3, and on the other side rarely to 6. The corolla of Campanula rotundifolia often varies from 5 to 6 and 7, and rarely from 5 to 4 and 3. 3 1 See the second part of this volume. " Botanisch Jaarbock, Gent, 1894, p. 72. 3 See also Ber. d. d. hot. Ges., Vol. XII, 1894, P- 2 2 , where further examples will be found. 34 Latent and Semi-Latent Characters. Selection and nutrition have as usual a great effect on half races. I shall not deal exhaustively with this point until the end of this part, but will give here a brief discussion of the general principles underlying it in order to prepare a proper understanding of the question. Our discussion of the phenomena of fluctuating vari- ability in the third part of the first volume led us to the conclusion that selection and nutrition usually operate in the same manner on the individual characters of plants. Fig. 2. Influence of selection and nutrition on the half race Ranunculus bulbosus scmiplcnus. A, Half Curve after several years of culture. B, Curve of the 12 best individuals (i. e., those richest in petals). C, Curve of the best plant. 1 Positive selection and plenty of food enhance the devel- opment of a character, whilst selection in a minus direc- tion or defective nutriment operate in the opposite direc- tion. Polycephaly in Papaver soinniferum behaves in this way, 2 and, as we shall see later on, all the other anomalies which have been tested do so, as well. Half curves can thereby be transformed into unilateral ones (Fig. 2) , either 1 Ber. d. d. bot. Ges., Vol. XII, 1894, Plate X, Fig. 4. 8 Vol. I, Part I, pp. I35-M3. Half Races and Half Curve. 35 by making r. special curve from plants which exhibit the largest number of abnormalities, or by making a curve from a race bred from such plants. But an improved race of this kind remains dependent on selection and high nutrition, and is soon lost if these are withheld. 1 One instance will suffice. Achillea Millcfoliinn has white flowers, but occasional specimens have red ones. From this I have bred a race with red flowers, which some- times even attain the deep red of dark wine. After four years of stringent selection, all the plants in successful cultures were more or less red. But if the plants were grown too close or were on poor land, more than half were white, and when I made further sowings without selection the proportion of reds rapidly reverted to its original small amount. On the other hand there is on the market the well-known Begonia semperfiorens atro- pnrpurea Vcrnon as a constant dark brownish red vari- ety. Eversporting varieties revert rapidly under minus- selection, but it is seldom possible to eradicate their char- acter altogether as I experienced when working with the adnation of the lateral branches to their main stems in Aster Tripolium and Bidcns grandiflora, and as I shall describe later in greater detail in the case of Celosia cris- tata. (See Part II of this volume.) In conclusion, we see that in nature as well as in cultivation (especially in the case of horticultural vari- eties and other anomalies) intermediate forms between the original species and its constant variety are often met with. The two commonest are the half race and the middle race or eversporting variety. The former has a half curve, the latter a bilateral one. Both occur in 1 Vol. I, Part I, 14, p. 122. 36 Latent and Semi-Latent Characters. numerous species and genera, either together, or sepa- rately. Both are easily influenced by high nutrition and selection, but are usually quite distinct and only appar- ently connected bv transitional forms. 5. TRIFOLIUM PRATENSE QUINQUEFOLIUM, AN EVER- SPORTING RACE. Four-leaved clovers are notoriously rare in nature, but it is perfectly easy to have many hundreds of them, provided a hereditary race can be obtained. Isolated ex- Fig. 3. Tri folium pratense quinquc folium, five-leaved and seven-leaved leaves of clover. The left leaf, A, shows a transition towards the 6-merotis leaf in the splitting of one of its leaflets. amples of this race seem to occur sporadically in nature ; all that has to be done is to find, to isolate, and to multiply them. (Fig. 3.) In the following section I shall describe the history of a particular race. I shall do so largely with a view to emphasizing the contrast between a middle race and a Trifolium Pratcnsc Quinque folium. 37 half race. In a half race the latent or semi-latent character is very seldom visible, perhaps in one leaf or on one plant amongst many thousands, and after several years of se- lection it is only on isolated individuals that two or three specimens of the anomaly may be found. In the middle race, or eversporting variety as I call it in contradistinction to the half race, the anomaly is by no means so rare. Most of the leaves consist of from 4-7 leaflets, and plants without such scarcely occur at all even in the absence of any selection. Trifoliate leaves are not wanting; indeed no plant is without them, par- ticularly in its early stages and on weak branches. On the other hand pure five-leaved or pure seven- leaved races do not as yet exist : I mean forms which do not revert. 1 There is no ground for supposing that we may not succeed, some time, in obtaining at least a constant seven-leaved variety. But for this to happen the right combination of unknown causes must chance to offer itself (see 2) ; and this has not yet occurred in the case before us. When a variable race has been found in nature the next step is to isolate it. And if, as is the case with red clover, isolated individuals of the species are sterile, two or three of them must be cultivated together, or if this is not possible one or several generations must be grown as a rule in order to purify the race of the effects of cross- ing. But this is easily effected. Further, the character can be improved by selection within the limits of varia- bility in the new race, just as in the case of pure specific characters. When once the furthest point in this direc- tion has been reached, and this usually occurs after a 1 Or at least revert as rarely as four-leaved individuals occur in the ordinary clover, which are in reality also partially atavistic. 38 Latent and Semi-Latent Characters. few generations, further improvements are only to be expected from a corresponding amelioration of the con- ditions of cultivation. In this way I succeeded in the beginning in improving my four-leaved clover, but after 1895, in spite of continuous and stringent selection, no further improvement has been observed. I shall there- fore confine myself to a description of the first seven generations. These were : 1st Generation. 1886-89. Two plants from Loosdrecht , 2nd 1890. Four plants with some four- and five-leaved leaves. 3rd " 1891. 36 % abnormal leaves per plant. ( S With isolated abnormal seedlings. 4th 1 " 1892. ] C With 60% seedlings of which the first. second or third leaf was tetramerous. 5th " 1893. C With 55% seedlings with compound pri- mary leaf. 6th " 1894. C With 96-98% seedlings with compound primary leaf. 7th " 1895. C With 95-97% seedlings with compound primary leaf. To proceed to a more detailed account I begin with the examples collected in the field. 2 I found them near Loosdrecht on the edge of a road which was covered with grass. They bore several tetramerous and one pentam- erous leaf and seemed therefore to afford better oppor- tunities than the usual find which often is only a single four-leaved clover leaf in a meadow. I transplanted them to my garden, where they lived for another three years. Here the anomaly not only reappeared but in- 'The result for this year is a double one. vS" (spring) refers to the crop of 1892 itself. C (crop) to the record of the seedparents in terms of the seedlings raised from their seeds (see p. 40) : Similarly with the subsequent years. 2 Over het omkcercn van halve Galton-curven, Kruidkundig Jaar- boek, Gent, Vol. X, 1898, pp. 27-54 with Plate I. Trifoliiun Pratcnsc Quinque folium. 39 creased, on account, doubtless, of the improved conditions of life. In July and September, 1889, I counted 46 tetramerous and 19 pentamerous leaves amongst a large number of normal ones. But there was no sign of a 6- or 7-foliate leaf on these two parent plants of my race. I saved seed from them in the autumn of 1889 and sowed it next spring on a bed in my experimental garden. I obtained something over one hundred plants of which about one-half showed at least one four-leaved leaf. The rest were removed either in July before they flowered, or whilst they were in flower. On September 1, I selected the four plants which bore the largest number of ab- normal leaves, and destroyed the rest. These four bore 64 tetramerous and 44 pentamerous leaves. Of the de- stroyed plants the best had only an average of about 5 abnormal leaves per plant. This year again there were no instances of 6- or 7-foliate leaves. In 1891 I obtained the third generation from the seeds of these four selected plants, sowing in the garden as before. It consisted of 300 plants on which I exam- ined 8366 leaves when they were beginning to flower. Of these 1117 or 14% were tetra- or pentamerous. Leaves with 6 or 7 leaflets were not observed ; they were first seen in August and September of that year. The number of plants with at least one quadri foliate leaf also ex- hibited an advance. There were about 50% of them in 1890, but now there were nearly 80%. These plants had on an average about four tetramerous and as many pentamerous leaves. At the beginning of August I chose the twenty best individuals and destroyed all the rest. I only harvested seed from the nine best plants among them and in the following spring only sowed the seeds of a single seed-parent which seemed to me to be the 40 Latent and Semi-Latent Characters. very best of all, 36% of its leaves being composed of more than 3 leaflets. In the spring of 1892 I sowed the seed in pans in the greenhouse attached to my laboratory instead of in the beds as before. The advantages of this were (1) that more seeds germinated and (2) that the examination of the seedlings was greatly facilitated. They stayed in the pans until the unfolding of the third leaf, were then looked through, and the best ones transplanted into pots with manured garden soil. Amongst the several hundred seedlings there were 18 in which the quadruplicity was already manifest among the first leaves. Only these specimens were planted out; during the summer they bore a large number of tetra- and pentamerous leaves ; and some 6- and 7-foliate ones, which appeared now for the first time in considerable numbers. With this, the isolation of the five-leaved race of clo- ver was brought to an end. The elaboration of the ordi- narily latent or semi-latent character had been fully ac- complished. The race could, like any other, be improved by selection but it could not be expected to change its character any further in the process. Of course I did not omit to effect this further im- provement. But there was no point in paying further attention to the characters of the adult plants, since dif- ferences could now only be found in them by a statistical examination of all their leaves. And it was found to be practically impossible to carry out this scrutiny with the necessary detail, for the plants soon become too big to be grown in pots. Therefore in order to make curves it is necessary to defoliate the plants, and this can not be done until after the choice of the seed-parents, whose Trifolium Pratense Quinquefolium. 41 leaves must obviously neither be removed nor even dam- aged. For these reasons it is desirable to effect the selection in the seedling stage, or at any rate before transplanting. This process had already been begun in the spring of 1892 and needed therefore only to be perfected by con- tinued selection. And the result justified my expecta- tions. In the spring of 1893 I sowed the seed of the 18 plants of the year before, already referred to, separately for each seedparent. I recorded the seedlings when the third leaf had unfolded. If all the leaves were normal, I straightway weeded out the plant; but if one or more of its leaves had a supernumerary leaflet I preserved it. Of the 3409 seedlings which I examined 2471 were normal and 938 were not, i. e., about 30 %. 1 Of course the re- maining 70% must also be abnormal, but the anomaly was not yet recognizable in the seedlings. Some of them which I transplanted produced, as adult plants, leaves with from 4 to 7 leaflets in large numbers. I determined the percentage production of abnormal seedlings in this manner for 16 of the 18 seed-parents; the values were distributed over them as follows : 10-20% 21-30% 31-40% 41-50% 51-60% 61-70% Seed-parents^ 173221 I further chose from this series a seed-parent pro- ducing 60% abnormal seedlings. It had itself had in its early stages a compound primordial leaf, which fact also marked it out for the continuation of the race. It will be found in the table on p. 38 under 1892 C. Amongst the seedlings from the seeds of this parent 1 Botan. Jaarbock, Gent, Vol. X, p. 37, where the two figures have been transposed by an oversight. 42 Latent and S end-Latent Characters. several occurred with trifoliate (instead of single) pri- mordial leaves (Fig. 4). I only selected these as seed-bear- ers, for transplanting, and I effected a considerable simpli- fication in my cultures by adopting this mark as a cri- terion for all further selection of stock plants. For the definitive selection could now be made 2-3 weeks after sowing, and it was not necessary to pay any further attention to the development of the character; this was fully insured. Nevertheless I took care by means of further experiments to sat- isfy myself that there exists a fairly close relation be- tween a large number of 4-7- merous leaves on a plant and a high percentage of abnor- mal seedings produced by it. In July 1893 I only saved the 12 best plants raised from the seed of the plant of 1892 with 60% abnormal offspring. With the excep- tion of two plants they all bore not only 4-6-foliate leaves, but even some 7- merous ones. The four best had 27, 30, 33 and 34 of this latter type. There were no leaves with more than seven leaflets. The plant with 34 7-merous leaves also produced the highest percentage of abnormal seedlings, as shown by the result of the sowing in the following spring. Of 209 seedlings produced, 51 had a bimerous, and 61 a trimer- ous, primordial leaf, i. e., 55% of abnormalities. It was therefore chosen as seed-parent (see p. 38). It should Fig. 4. Tri folium pratense quin- quefoliuni. A, Seedling with a trifoliate primordial leaf. B, C, Seedlings with single and bimerous primordial leaves ; these two latter types were regarded in my race as atavistic. Trifoliimi Pralcnsc Omnque folium. 43 be remarked that in previous years seedlings with a com- pound primordial leaf had either been entirely absent or at any rate very rare. 1 In the summer of 1894 I only bred offspring from the plant with 55% of abnormalities in its seedlings, and of these only the twenty best, with compound primordial leaf and the next leaf tetra-pentamerous. These only did I allow to flower and to bear seed. The result was recorded by means of the same characters in the following spring. For eleven plants it was 70-90%, for five others 91-96%, and for the two best 98-99% seedlings with compound primary leaf. And the higher the number the greater was the percentage of trifoliate, as opposed to bimerous, primordial leaves. The same high percentage was obtained in the culture of the next year, 1895, in the seventh generation of my experiment. Since then the race has remained constant under the same conditions of selection. I have employed this constant and highly abnormal race for a series of observations and experiments, to the more important of which I shall now refer, 2 for they are well qualified to afford us some insight into the nature of such a race. This race exhibits a high degree of var- iability, which is due to the possession of a semi-latent character besides that which it has obviously inherited from the parent species. The extent to which this paren- tal heritage, the normal trifoliate leaf, is developed de- pends on the conditions of life of the plant. And, speak- 1 See the remarks in 22 relating to the size of the seed in Tri- foliuin incarnatum. In the five-leaved clover, especially in later years, practically all the seedlings had compound primordial leaves, so that this character had nothing to do with the size of the seed. 2 For a detailed account see the oft-cited paper in Kniidkundig Jaarboek, Vol. X. 44 Latent and Semi-Latent Characters. ing generally, favorable conditions favor the characters O O J ' of the race, and unfavorable ones those of the species (see below, 26). This is only a special case of the well-known prin- ciple : Every injury increases the tendency to atavism . l In the first place let us consider the periodicity. The number of multipartite leaves increases with the indi- vidual strength both on the whole plant and on the sep- arate branches. And if, at the end of growth, weakness supervenes this number again decreases. Let us examine Fig. 5. It is a photograph of a strong young branch which was removed on August 1, 1900. The lowest leaf was nearly withered ; it was small and had the inversely egg-shaped form of the leaflets which is characteristic of the leaves of the young red clover, It consisted of only 3 leaflets. The two following leaves were markedly larger and stronger, of a more elliptical form and tetramerous. Then follows a 6- and then a 7-merous leaf, after which the leaves again return to the simpler types. The branch photographed was chosen for its regu- larity; and yet a pentamerous leaf is absent from the ascending series. Most of the branches, even on the best plants, were less regular : indeed it often happened that tetramerous leaves were succeeded by some trimerous ones, and so forth. 2 What has been stated concerning the lateral branches is also true of the rosette of radical leaves whose axis 1 That is, reversion of the race to its parent species, for the char- acter of the race is itself, morphologically speaking, a reversion to a more remote ancestor. 2 For exact figures the reader is referred to : Ueber die Periodici- tat der partiellen Variationen, in Ber. d. d. bot. Ges., 1899, Vol. XVII, p. 48. Trifolinin Pratcnsc Quinquefolium. 45 is, of course, the main stem of the whole plant. Here also the number of leaflets per leaf first increases, on the average, and then decreases, with many fluctuations how- ever. The branches themselves exhibit a certain periodic- ity since the lower ones contain a smaller quantity of Fig. 5. Tri folium pratcnse quinquefolium, 1900, showing the periodicity of the anomaly on a branch. Beginning from below the leaves have 3 4 4 6 7 5 leaflets. abnormal leaves than those next above them, whilst the highest of all are poorer again. If therefore the conditions are favorable to a branch in its earliest stages it will develop more 4-7-merous 46 Latent and S end-Latent Characters. leaves. And it is obvious that such leaves will extend both above and below the maximum of the period in direct proportion to their number. \Yhence it again fol- lows that the better nourished the plant is, the earlier will the abnormality appear. And this is true both of the individual L ranches and of the rosette of radical leaves, and therefore of the whole plant. From these conclusions the converse rule may be de- duced that the earlier a seedling produces its first tetram- erous leaf, the greater will probably be the number of ab- normal leaves on the adult plant. The most abnormal plants will probably be those which in the seedling stage had a compound primary leaf. Experience has proved the truth of this rule throughout my experiments. If we now take another glance at the table on page 38 we see that the character recorded has gradually shifted in the course of generations and as a result of selection. The more the improvement advanced the earlier could selection be effected. In the third generation I kept 300 plants in the beds to be selected from; since the fourth generation I have carried out the selection in the seed- pans and only planted out the few best (e. g., 10-20) to act as seed-parents. It is possible, therefore, within the limits of such a race, on the one hand to effect an increase in the number of multipartite leaves, and on the other to reduce it In- reversed selection. In both cases we go as far as pos- sible from the mean of the race, without, however, suc- ceeding in overstepping its definite boundaries. Let us see what selection is able to effect in the two cases, and let us begin with the former. It is the question of in- tensifying the anomaly to its extreme limit. / o J A striking peculiarity of my race is the fact that leaves Trifolium Prafcusc Quinque folium. 47 with more than seven leaflets have never, or only ex- tremely rarely, been produced. As a matter of fact a duplication of the leaves by splitting, which is so common among other plants, 1 occurs in my race also, and if it affects a pentamerous leaf, makes a 10-merous one of it. But that is the expression of another latent character which we are not concerned with here. Apart from these I have not yet found in my cultures, in spite of the most careful search, a single instance of a leaf with more than 7 leaflets. The character of my race is the quinquefoliate leaf which is usually in the majority ; the remaining types are grouped round it in accordance with OUETELET'S law, so far as the tendency to symmetry permits this. For it is clear that this tendency does not favor the regularity of the curve of variation. The increase in the number of leaflets from 3 to 4 takes place by the lateral splitting of one of the lateral leaflets (see Fig. 3 A), one of the lateral veins becoming the primary vein of the new leaflet. Transitions such as that figured are certainly fairly rare, but all degrees of them, down to a splitting of the small partial stalk of the leaflet, occur from time to time. If only one leaflet is split, the leaf becomes asymmetrical ; but if the two lateral leaflets split, the whole may remain symmetrical. The duplication can extend to the terminal leaflet and turn a vein of this, either on one side or on both sides, into the primary vein of a new leaflet. In this way the 6- and 7-merous leaves arise ; the former are asymmetrical, the latter symmetrical. The statistical examination of large numbers proves that the symmetrical leaves predominate over the asym- metrical ones. The plant seems to prefer to retain its 1 DELPTNO, Teoria general? dclla FiJlotassi, 1883, p. 197. 48 Latent and Semi-Latent Characters. symmetry even in the anomalies. This is brought out in the curves by the relative shortness of the ordinates corresponding to 4 and 6 (Fig. 6). Let us return to the processes of selection. The mean of the race is a pentamerous leaf, which varies within fairly narrow limits, never (or hardly ever) less than Fig. 6. Trifolium pratcnse quinque folium. A, Normal curve of the number of leaflets in the leaf. B, Curve of an atavistic individual. C, Curve of the maximum de- gree of abnormality, 1894. three or more than seven leaflets being produced. Selec- tion can therefore be either in the direction of the 7 or the 3. In both cases the original symmetrical curve be- comes unilateral. But in the former case the improve- ment of the race is pushed on as strongly as possible, in the latter the reverse happens until it can hardly be dis- Trifolinm Pratcnsc Quinque folium. 49 tingiiished from the ordinary instances of the rare four- leaved clover. A glance at the table on page 38 will show that my race was only very slightly developed at first, and had to be brought to its normal type by isolation and selection. But in spite of this selection it is not so constant that it does not occasionally give rise to atavistic individuals. On the other hand individuals with a maximum develop- ment of the character of the race are from time to time produced. And these extremes are sometimes both found within the limits of a single culture. I observed this in 1894 with plants which had been raised from the seeds of a single individual in the third generation (1891, p. 38). The seed-parent in question had survived the winter and did not ripen its seed until the second year. In July, 1894, there was a large num- ber of strong plants of the same age, of which I chose the seven best for a detailed examination of their leaves. Some of the oldest leaves were already withered, the youngest not yet unfolded ; these were not recorded. Each of these seven plants was plotted in the form of a curve, one (Fig. 6 A) gave the normal curve of the race, an- other (B) was atavistic, whilst all the rest had their highest ordinate at 7. I have only given the mean value for these five (C). These three groups gave the following percentage of leaves with the number of leaflets written above them : Number of leaflets: 34567 Number of leaves counted A. Normal example: 17 16 37 14 16 172 B. Atavistic example: 75 19 5 1 216 C. Extreme variants: 12 9 22 17 40 97 1 1 Mean number per plant. 50 Latent and Scnii-Latcnt Characters. These figures are presented graphically in Fig. 6. It Will be seen that the normal curve is a symmetrical one slightly depressed, however, over the ordinates of the even numbers as a result of that symmetry which we discussed above. The two other lines form half curves; in both of them the apex coincides with one extreme. The curve B, of the atavistic individual, is almost the same as the curve which was the dominant one in the first years of my experiment when there were, as yet, no 6-7- foliate leaves (p. 38). It is an ordinary half curve of variation, which is characteristic of the half races of semi-latent anomalies. The curve C is, however, re- versed; it displays the predominance of the racial char- acter over the antagonistic one which is that of the original species. It also shows the preference for sym- metrical leaves. If atavistic individuals are used as seed-parents the character of the race can be observed to vanish more or less completely in a short time. I carried out an experi- ment of this kind in the years 1896-1898, after the race had reached its maximum development in 1894-1895 as described on page 38. Within the space of three gen- erations this race has retrogressed so far that the plants could no longer be recognized as belonging to it. For the purposes of this reversed selection I chose, from the plants which had borne a large number of 5-7-merous leaves in 1895, those seedlings of which the primary leaves were single and the first leaves trifoliate. With a few exceptions they had all developed occasional tetra- pentamerous leaves by the middle of June. Three of the exceptional ones were isolated before flowering, they sub- sequently developed a few multipartite leaves. But when their seeds o- e rrnmated it was seen that they were not Trifolinm Pratcnsc Quinque folium. 51 only not poorer in seedlings with compound leaves but even slightly richer; they were therefore not chosen for the continuation of the experiment. I chose the seeds of three plants of 1896 which had given rise to no more than 2-3% seedlings with compound primary leaves. Atavistic seedlings only were transplanted, but in the following summer (1897) even those bore some tetra- pentamerous leaves, almost without exception. On the other hand 6-7-merous leaves were almost entirely ab- sent, and the race had thus returned to the condition described by the unilateral curve of the first year of the experiment (1891-1892). Some plants produced noth- ing but trifoliate leaves during the whole of the summer and the following spring. In 1898 I made another culture of atavists from the seeds harvested in 1897. This was therefore the third atavistic generation. But two thirds of the generation raised still consisted of plants with some tetra-pentam- erons leaves, and therefore possessed this character in a far higher degree of development than ordinary red clover. This stringent, thrice occurring reversed selec- tion had therefore considerably reduced the development of the anomaly but had not succeeded in destroying or even in concealing the fact that the culture belonged to the pentamerous race. I also made an experiment on the influence of ex- ternal conditions on the development of multipartite leaves. There are two ways of dealing with experiments of this kind ; we may either subject the different parts of the same plant to diverse conditions of life or similar samples of seed to diverse treatments from germination onwards. In the former case we determine the effect on the grown plant. This is however seldom great, inas- 52 Latent and Semi-Latent Characters. much as the plant is most sensitive in its early stages. In this form of the experiment we can, so to speak, only investigate the last vestiges of its former susceptibility. Far more striking results are to be expected from experi- ments with seedlings; but here a great uniformity in the samples of the seeds is necessary for the results to be reliable. It is not sufficient to mix the seed, but it is advisable to harvest seed from two or three or still better from a single seed-parent of known and pure ancestry. It is even better to allow the influences that are to be investigated to operate during the development of the seed on the parent plant. In accordance with these considerations, therefore, I cut one of my plants into two parts, one of which I transplanted into poor sandy soil but the other into good garden soil, and allowed them to set seed. I was thus able to study both the direct effect on the plant and also the indirect effects on the succeeding generation. (See Vol. I, Part III, pp. 521-522.) The experiment, which was carried out during the years 1892-1894, was made with a single individual which arose from the stock plant for 1891, mentioned on page 38 and marked S. This plant had, when it germinated in 1892, a bimerous primordial leaf, and in the same year bore seeds which, when sown in the spring of 1893, gave rise to about 40% seedlings with a tetramerous leaf. As soon as this was visible in the seeclpan the choice was made and the parent plant, which I had kept through the winter in a bed, was cut in two and trans- planted into the above mentioned kinds of soil. Both halves grew well, although not with equal luxuriance; they flowered in July, were pollinated from the various plants around them composing the main culture of that Trifolium Pratcnsc Quinquc folium. 53 year, and set seed in August. At this time I examined an equal number of leaves on the two halves and ob- tained the following result : Number of leaflets: 34567 On garden soil: 12 25 34 20 18 On sandy soil: 18 19 35 19 17 The behavior of the two halves was identical ; the difference in the soil exerted no visible effect. Moreover the seeds on the two halves were of about the same size and produced in roughly equal numbers. The two sets were harvested separately and sown in the following spring (1894) in pans. When the young plants had about 3 leaves they were examined. Calling a plant with a tetra- or a pentamerous leaf "abnormal" the result was : Seeds from garden soil 30% abnormal " sandy soil 24% The experiment involved 150 and 200 seedlings. The abnormal ones were further sorted according to the composition of their primary leaves. Leaflets 123 Totals Seeds from sandy soil 24 10 13 47 41 garden soil 16 12 13 41 Both counts therefore gave a difference in favor of the better nourished seeds. For further investigation I selected those which appeared most abnormal from both series, i. e., the seedlings with a trimerous primordial leaf, and planted them out under similar conditions. In July when each plant had twenty or more stems, I pulled them up, selecting for examination the ten best plants from each group; i. e., those ten, the leaves of which numbered about 100 per plant. The leaves were recorded separately for each individual, and as there happened to 54 Latent and Semi-Latent Characters, be practically no difference between the several individuals in each group, I calculated the mean for the two sets in percentages. Number of leaflets per leaf 34567 From seeds from garden soil 14 13 25 16 32 " sandv soil 39 13 23 10 15 Difference 25 +2 +6 +17 The effect of the treatment in the previous year is now perfectly plain. The curves for both groups have become unilateral but in the case of the better nourished ones the apex is at seven, and for the others at three leaflets per leaf. Conversely we may conclude that, in the experiment described on page 47 and graphically exhibited in Fig. 6, the atavists were produced by poorly nourished and the maximal variants by highly nourished seeds. And the following generalization about anomalies seems to be justified: that the nutrition of the seed on the parent plant is the most important factor influencing the devel- opment of the anomaly (Vol. I, pp. 521-522). Let us now briefly summarize the results of this experiment. I began by finding in the field two plants belonging to a five-leaved race, which however as the result of indifferent nutrition for several generations only developed tetra-pentamerous and no 6-7-foliate leaves. By better cultivation and by the continued selec- tion of the most abnormal individuals, no doubt those which happened to have been best fed, a race was evolved in the course of a few generations with a number of leaf- lets per leaf varying between 4 and 7 round a mean of five. After this selection had been repeated four or five times maximal variants were produced the majority of Trifolinm Pratcnsc Quinque folium. 55 the leaves of which were 7-foliate. At the same time there were still "atavists" in the seventh generation the apex of whose curve was over 3 leaflets. The atavists however really belong to their race as is shown by the fact that even after repeated selection in an atavistic direction they produce far more quadri foliate leaves than the normal red clover (or more exactly, the corresponding- wild half race of the red clover). The better the seeds are fed on the parent plant the greater is the development of the anomaly on the indi- viduals produced by them. Poor seeds give rise to ata- vists, good ones to extreme variants. My experiment extends over ten generations. It gives no support to the view that the five-leaved race was, so to speak, caught in the act of developing its character, or that it could give rise to a higher type with- out further mutation. It is a highly variable, but constant variety. III. THE DIFFERENT MODES OF ORIGIN OF NEW SPECIES, 6. HORTICULTURAL AND SYSTEMATIC VARIETIES AND ELEMENTARY SPECIES. The opinion has of late been often expressed, by VON WETTSTEIN in particular, that there is no ground for the assumption that all species have arisen in the same way. 1 There is no difficulty in applying this view to the theory of mutation, although one of the chief objects of this book is to show that ordinary or fluctuating variability does not provide material for the origin of new species. But this does not exclude the possibility of different modes of origin of new species. The simultaneous origin of species in groups, in definite periods, such as I have described in the case of Ocnothcra Lainarckiana, must constitute for me the main type of this process, until the origin of species has been experimentally studied in other cases. Such experiments would have to study the phe- nomenon before and during the first appearance of the new type. Inferences drawn from data obtained after its appearance can hardly be considered as decisive. This essential type explains in my opinion in the first 1 R. v. WETTSTEIN, Der Saison-Dimorpliismus ah Ausgangspunkt fiir die Bildung neuer Artcn im Pflanzenrcich, Ber. d. cl. hot. Ges., Vol. XIII, 1895, P- 3O3 ; and particularly the same author's Desccn- denztheoretische Untersuc/ningcii; I. Untersuchungen iibcr den Sai- son-Dimorphismus im Pflanzenrcich; Denkschr. d. Mat. Naturw. Classe d. k. Akad. d. Wiss., Vienna, 1900. Horticultural and Systematic Varieties. 57 place the progressive origin of species, that formation of new characters to which in the main the evolution of the plant kingdom is due. On the other hand there is a whole series of other types which are now, so far as it is possible to judge, mainly confined to the lateral branches of the phyletic tree. With regard to these however we must content ourselves at present with indirect methods of investigation. DARWIN'S statement that varieties are incipient spe- cies is well known. So also are the words of one of the most famous authorities 1 on horticulture, VERLOT : Toute varicte a d'abord existe a I'ctat de variation. These two generalizations are evidently based on phenomena en- tirely different from those with which we have become familiar in Ocnothera. They constitute, so to speak, the other extreme of the series. I propose therefore now to investigate the manner in which "variations" in the sense of so-called structural abnormalities or anomalies (and not the individuals which exhibit variation in accordance with QUETELET'S law) arise, and how they result in the origin of "species." But here we come across an obstacle on the very threshold of the inquiry in the manifold meanings of the word variety. 2 It will soon become clear that horticultural and systematic varieties are to be considered as categories of entirely different values. But both DARWIN'S and VERLOT'S sentences just quoted are based on data ob- tained from horticultural varieties ; and we must now 1 B. VERLOT, Production ct fixation dcs varictcs, 1865, p. 100. 5 The general conception of this term is that formulated by CAR- RIERE in the following words: "On nomine varictc tout individu qui, par quclque caractcrc quc ce soit, se distingue d'un on de plusicurs autres avec Icsqucls on le compare ct quon considere comme apparte- nant a un meme type specifique (Production et fixation dcs varietes, 1865, p. 6). 58 The Different Modes of Origin of new Species. inquire how far their transference to systematic varie- ties is justified. 1 The origin of horticultural varieties will therefore be submitted to a critical and experimental examination. But before we do this I think it advisable to consider first the meaning which is attached to the term variety in systematic works, and secondly the various ways in which species can arise. And we shall find that whilst there is no question that the mode of origin of horti- cultural varieties is often analogous to that of so-called "good" species, this parallel is by no means so common as the present form of the doctrine of descent would lead one to believe. To begin with systematic varieties : Here we find we can draw a pretty natural line between what we called ele- mentary species on the one hand, and real systematic varieties on the other. In connection with this antithesis I think it desirable, after what has already been said on this topic in the first volume, 2 to lay especial stress on the fundamental dif- ference between these two conceptions. LINNAEUS and his pupils describe the elementary species as varieties ; JORDAN, DE BARY, and others who argue from experi- mental data, refer to all forms as species. The terms "species" and "variety" have become so familiar that it is no longer possible to effect any radical change in their definition. For their exact meaning we have to refer to the works of LINNAEUS himself. His 1 For some interesting observations relating to the origin of new forms, see the papers by F. KRASAN in ENGLER'S Botanischc Jalir- biichcr. Vol. XIII, Pts. 3-4; Vol. XXVIII, Pts. i, 2. and 5, and also his Mittheilungen ilbcr Cultiirvcrsuchc mit Potcntilla arcnaria, Graz, 1900. 2 See Vol. I, 7, "Species, Subspecies and Varieties," especially pp. 169-172. Horticultural and Systematic Varieties. 59 conception of them is now common property, and in my opinion our best course is to interfere with that con- ception as little as possible. There can be little question that the difference between variants and variations is becoming more and more widely recognized. Variants are what we call individual devia- tions ; they are instances of fluctuating variability. The characters which distinguish them disappear under suit- able cultivation and are therefore to be regarded as in- constant. In systematic works they are not as a rule gi ven a place, or merely briefly mentioned, or, lastly, treated as a Forma, which is the lowest subdivision of the system; e. g., Forma alpestris, Forma aquatica. But this can only be done when the relationship of the form is sufficiently known ; lack of material in the case of exotic plants, or incomplete investigation of indigenous species of course would make this impossible, and such forms have therefore often first been described as vari- eties or even as species. 1 In many cases of course the true relationship is still unknown and the systematic grouping, therefore, to be considered as provisional: as for instance in the case of Anthyllis Vnlneraria alpcstris, Limosclla aquatica caulesccns, Carlina acaulis caulcsccns, and so forth. BONNIER'S researches on Alpine plants, discussed in detail above (Vol. I, p. 146), have demonstrated that some of these differences are not even instances of indi- vidual but of partial variability. From the two halves of a single individual can be grown the form character- istics of the plains and the Forma montana. 1 For example Ranunculus aconitifolius L. in alpibus minor, caule 3-5 floro ; R. aconitifolius altior KOCH, caule multifloro, fol. laciniis longius acuminatis, in montibus humilioribus = R. platanifolius L. mant. 79 (Keen, Synopsis, p. 12). 60 The Different Modes of Origin of new Species. But the large number of cases of forms preliminarily described as varieties but which possibly may be only variants, is one of the most considerable obstacles in this inquiry. LINNAEUS himself followed two distinct rules in sub- dividing his species. According to the one the species was regarded as the type from which the varieties were derived ; according to the other, however, the species was regarded as a collective group which embraced a certain number of units of equal value. The separation is sharp and definite and LINNAEUS was obviously perfectly con- scious of its reality. In the derived varieties the series begins with /? followed by y, 3, e etc. ; it is taken for granted that the type or Forma geniiina represents the a. In a homonomous series there is no such Forma gcnuina, and the series of varieties therefore begins with a. Let us consider the two cases separately and let us begin with the second. LINNAEUS'S homonomous varieties, a, /?, y etc., are sometimes arranged in groups, and sometimes not (as in Teucrium Poliiiin, Lavandula Spica, etc.). In the former case the species falls into two or several subspecies, each of which again may include one or several varieties. For instance Euphorbia e.rigua has two subspecies acuta and retusa, the first of which consists of one and the second of two varieties. Beta mdgaris has the well-known sub- species mbra and Cicla; the first of these embraces five, the second two varieties. In these species there is no Forma typica or Forma gcnuina. The variety which is named first has no other priority over the others. In such cases the species is a group of similar com- position to that of a genus and of a family ; since in these no particular species or genus is regarded as the proto- Horticultural and Systematic Varieties. 61 type from which the rest would be merely derived forms. Species of this kind are therefore obviously and avowedly collective species. LlNDLEY, A. P. DE CANDOLLE, ALPHONSE DE CAN- DOLLE and other eminent systematists consider the col- lective species without Forma typica to be the only really existing type. Species must be subdivided in exactly the same way as genera, says the last named of these authors in his Phy to graphic. 1 LINDLEY splits up his species of roses on the same principle; Rosa rubiginosa into 8, R. spinosissima into 9 varieties, etc. DE CANDOLLE deals with the difficult and numerous subgenera and elementary forms of Brassica in the same way in the second volume of his Systema Vegctabiliwn. DE CANDOLLE calls the units, which in such cases are treated as varieties, "Ics elements de I'espece"; 2 they are related to the species as these are to the genera and as the genera to the families. But the majority of botanists regard varieties as forms which have been derived from the species. For them the species is the type, the real entity, from which the varieties have arisen by small changes. They follow the course taken by LINNAEUS who based his diagnoses, in the vast majority of cases, on one of the forms of a species and arranged the rest in a lower grade under this. The origin of the varieties from the species was simply inferred from a priori premises as I have already shown in the first volume, this origin having only been directly observed in isolated cases of horticultural prod- ucts; for the majority and certainly the most important 1 ALPH. DE CANDOLLE, La Phytographie on I'art de decrire les vegetaux, 1880, pp. 74-82. Much of the argument set forth in the text is due to this excellent work. z Loc. cit., p. 80. 62 The Different Modes of Origin of new Species. cultivated varieties are as old or even older than cultiva- tion itself. If we examine a number of such derived forms in any systematic work or flora, it immediately becomes evident that the same kind of differences recur in the most widely j separated families, genera and species. Everywhere vari- eties present series of parallel forms. The recurrence of white flowered varieties in numerous species with blue or red flowers is so familiar a phenomenon, that often all reference to them is omitted. LINNAEUS himself knew that nearly all such species had a white variety. If the color of a flower is compound, and if one of the compo- nents is lacking, a white flower with a dark center often results and is known as a ]\ir. bicolor (for example Cyno- (llossmn offieinale bicolor, Agrostenuua coronaria bicolor} ; or the dark patches are absent as in Geutiana piinctata concolor, which case is exactly analogous to that of Arum niacitlatuni iiiimacnlatinu. Often too, the clothing of hair is lacking either on the whole plant or, when only certain parts are densely hirsute in the "species," on these. The nomenclature of the series of parallel forms, under this heading, is par- ticularly rich in terms which all indicate the same prop- erty, as for example : Papaver dubium ylabrum, Biscn- tella laevigata glabra, Arabis ciliata ylabrata, Arabis hir- suta ylabcrrima, Veronica spicata nitens, Amyydalus Per- sica laciis, Eritrichium namun leiospermum, Paeonia corallina (pcrcyrina) leiocarpa, etc. Thornless forms are usually termed incnnis; they oc- cur in Ranunculus arvensis, Genista ycrmanica, Robinia Pseud-Acacia and many others. The ] T arietas ciliata occurs in Cytisus prostratus and in C. spinescens, also in Lotus corniculatus, etc. A dense clothing of hair is the Horticultural and Systematic Varieties. 63 distinguishing feature of Solatium Dulcamara tomcnto- sum, Veronica scntcllata pubescens, Melissa officiualis vil- losa, Galeopsis Ladanum canescens, Vicia lutca hirta, Lotus corniculatus hirsutus, etc. The patches of color at the base of the petals are often absent in Papavcr orientate, in Erodiitm cicutarium and many other plants. Such names as ochrolcuca, purpuras- cens, intcgrifolia, serratifolia, angustifolia, latifolia de- note varieties each one of which may recur in several unrelated species. Finally I may mention the red berries which occur as a varietal character in Empetum nigrum and characterize the red variety of the gooseberry; and the yellow berries of Atropa Belladonna lutca and Daphne Mezereum album which are only selected examples from a long series of such varieties. All these forms differ from their species in the fact that a particular one of their characters is either devel- oped to a greater extent (hirsuta, ciliata, purpurascens,) or on the other hand very slightly developed or entirely absent. The absence of a character may also be a case of ex- treme rarity in the vegetable kingdom such as the straw- berries without runners, and the peculiar Finns Abies aclada, with its tall absolutely unbranched stem, which has been figured by ScHROTER. 1 Fragaria t'csca mono- phylla (Vol. I, Fig. 38, p. 193), Robinia Pseud-Acacia monophylla, Fra.rinus Ornus monophylla,- and a mo- nophyllous form of Melilotus cocrulea (Fig. 12 on page 87) belong to the same category. The varietal names enumerated above almost always occur, in systematic works, in series which begin not with 1 C. SCHROTER, Die Vielgestaltigkeit dcr Fichte, 1898, pp. 52-53. 2 A. BRAUN, Verjiingung, 332. Here also the earlier literature will be found; and some facts concerning Rubus Idaeus monophyllus. 64 The Different Modes of Origin of new Species. a. but with ft and are therefore considered as having been derived from a Forma typica or gcnuina and not as being of equal value with this. Evidently the principle on which they are classified is borrowed from a consideration of horticultural varieties. This proceeding, however, is only justified in the relatively rare cases in which horti- cultural varieties can be demonstrated to be younger than the species. Besides this the geographical distribution of the forms in question is often employed to decide which is the species, and which are the varieties. If one particular form is wide-spread and another only local or sporadic in its appearance it is obvious that the former will be regarded as the older and therefore as the species. Often this fits in conveniently with the fact that the species was discovered earlier than the variety, so that instead of disturbing the classification in the system all that had to be done was to range the variety under the species. The exigencies of space prevent me from going into further detail here. What I have already said may suffice to show that the systematic term "variety" means two fundamentally different things to LINNAEUS and the later systematists : 1. Homonomous Forms, amongst which even LIN- NAEUS could not select one as a type for the others ; "Elements de 1'espece" (DE CANDOLLE) or ele- mentary species. 2. Derived Forms, which are distinguished from the type of the species only by the decreased or in- creased development of a particular quality; or by its complete absence: True varieties. 1 Amongst these, again, the simple invariable types are to be distinguished from the intermediate or eversporting races. (See 3-4-) Progression, Retrogression and Degression. 65 I think it undesirable that these two types of sub- divisions of the species should continue to be denoted by the same term. The simplest plan would be to refer to the former as elementary species and only to the latter as varieties, and I hope that this limitation of the terms will come into general use. The question, however, is a purely systematic one and belongs to the department of descriptive science. For as soon as it is treated from the experimental standpoint the whole difference disappears. Many of the best vari- eties prove, when tested by sowing, to be as constant as elementary species, so that a separation on the basis of constancy is out of the question. # * # On the basis of the foregoing discussion I treat the homonomous subdivisions of the LINNEAN species as elementary species and eventually denote them with bi- nary names. In the case of derivative varieties, however, I prefer to make no definite choice ; I regard for example Chelidonium laciniatmn Miller and Chelidonlum inajus laciniatnni as equally justifiable. And when for instance several species in the same genus have white flowered or glabrous varieties, a binary nomenclature would obviously be much too cumbrous. 1 7. PROGRESSIVE, RETROGRESSIVE AND DEGRESSIVE FORMATION OF SPECIES. A glance at the phylogeny of the vegetable kingdom reveals the fact that all species cannot have arisen in the same way. Progressive development is due to the con- dor instance if specific names like that of Agrostemma nicae- cnsis for Agrostemma Githago pallida were generally used for white flowered varieties. 66 The Different Modes of Origin of new Species. tinual formation of new characters, to increasing differ- entiation. Nevertheless the great multiformity of spe- cies within the orders and families is only in part due to this progressive process, but to a large extent to an in- finite variety of combinations of characters already exist- ing. This is combined in innumerable cases with in- o stances of regression ; that is, with the absence of characters which are otherwise proper to the group to which the species belongs. Sinm and Bcrula have, for example, simple pinnate leaves within the group of the Umbclli ferae with doubly pinnate leaves ; and the assump- tion is that they have arisen from the latter by a simple loss. Similarly Primula aeanlis stands in the middle of a group containing the Primulas, Androsace etc. with umbellate inflorescences, and the same inference is drawn as to its origin. The same is true of a host of other cases, and even for whole groups. For instance DELPINO holds, as is well known, that the Monocotyledons have arisen from the lower Dicotyledons by the loss of a whole series of characters. Cases such as these are spoken of as instances of retrogressive metamorphosis. And it is probably not too much to say that there are possibly more species on the face of the earth at present that have arisen on retro- gressive than on progressive lines. The question is often debated whether, in retrogres- sion, the characters absolutely disappear or only become invisible, or latent. There is much evidence for the latter view, derived largely from the great variety of atavistic structures (youth forms, subvariations on the lower internodes of lateral branches, the form of the leaf in suckers, the effects of parasites, anomalies, reversions to the ancestral form by bud-variations, etc.). Latency Progression, Retrogression and Degression. 67 is without doubt the general rule. That an actual internal loss may also occur is probable on general grounds, but very difficult to demonstrate in a given case. For every positive result points to latency, and nothing but a nega- tive result after exhaustive investigation could warrant the conclusion that a character had absolutely disap- peared. The multiformity of species within the larger groups is also due to a phenomenon which DARWIN calls parallel variation. I refer to the repeated appearance of the same new character in related or remote groups. 1 Climb- ing and tendril-bearing plants, parasites, saprophytes and insectivorous plants, decussate phyllotaxy, are a few names from a vast number of instances. One of the greatest difficulties for the systematist, the question as to the mono- or polyphyletic origin of many characters is a problem of a similar nature. For example, are the siliqua and silicula in the Cruci ferae, or is the position of their embryo to be regarded as an indication of mono- or polyphyletic origin? Do the Sympetalae with an in- ferior ovary originate from other Sympetalae or from epigynous Choripetalae ? Have the Gymnosperms arisen once or oftener from the vascular Cryptogams? We do not know, because, on such points, the highest authorities are not in agreement. And so long as these differences of opinion exist it will be difficult to approach the question as to the cause of the parallel formation of specific char- acters whether they arise from a common latent source, or afresh each time with any hope of success. The origin of svstematic and horticultural varieties 1 On this point see also my Intracellulare Pangenesis, English translation by Prof. C. Stuart Gager (Chicago, The Open Court Publishing Co., 1910). 68 The Different Modes of Origin of new Species. is evidently clue, in the vast majority of cases, to retro- gressive development or latency, as I have already at- tempted to show. There is a close analogy between the formation of these varieties and that of certain species. The origin of varieties (such as Var. hirsutissinia, spinosissima and ciliata) as the result of the intensification of characters is a much rarer phenomenon. This form of variety, which seems to be of very little importance in the evolu- tion of the vegetable kingdom, may be called subprogres- sive, and the phenomenon of its origin subprogressive formation of species. The parallel, retrogressive, and subprogressive modes of origin have this in common that they only provide new combinations and do not contribute new units or any es- sentially new elements to the progressive evolution of the vegetable kingdom. In this respect they stand in sharp contrast to progressive formation of species. There is another series of phenomena to be mentioned here, of still less significance in the phylogeny of plants. The first of these is the manifestation of old, latent char- acters. A whole series of anomalies are so widely dis- tributed in the vegetable kingdom, or at least among flowering plants, that it is almost impossible not to as- sume a common cause for them. This cause must be a latent character that has arisen in some common ances- tor and therefore must be of great antiquity. The com- monest and best known example of a widely distributed anomaly of this kind is that of fasciation, instances of which in almost any desirable number of species can be collected in the course of a few years. It appears that almost every species amongst Coni ferae and Monocotyl- edons, but especially among Dicotyledons, can exhibit Progression, Retrogression and Degression. 69 fasciations. 1 As a constant horticultural variety it occurs in Celosia cristata; but as a specific character, so far as I know, it does not occur. This is, however, true of the almost equally widely distributed split leaves (as in Boeh- incria biloba), of adherences (Solatium), of flowers on leaves (Hclwinyia rusciflora and others) and of numer- ous other anomalies of which CASIMIR DE CANDOLLE has given a valuable general account. 2 He calls them "Varia- tions taxinomiques" ', whereas anomalies which do not occur as specific characters such as fasciation, twisting, virescence and sterile varieties, are designated by him as ataxinomous. I take the following further instances from his list : Connation of opposites leaves, which oc- curs normally in Dipsacus, Lonicera and others, or of the cotyledons (amphisyncotyly, normal in Sicyos) ; Pitch- ers, normal in Sarracenia etc., and in the peltate leaves, as for instance Eucalyptus citriodora; foliacious excres- cences on the leaves, normal in Senecio sagittifolins from Uruguay, and on petals, normal in Petaquia saniculae- folia and as a sport in Clarkia clcgans\ Synanthy in Lonicera, and so on. For our purposes, however, the question is not which anomalies can also occur as specific characters, but con- versely which specific characters can also occur as anom- alies in other species. For our task is to deal with the problem of species and especially to provide an answer to the question how far their characters can be derived from more or less widely distributed latent qualities which have existed for a long time in the vegetable king- dom or in particular groups of it. With this end in view 5-6- 1 See Vol. I, Figs. 34 and 35 on pp. 182-183. 2 C. DE CANDOLLE, Rcmarqucs sur la tcralologie vegctale, 1896, pp. 70 The Different Modes of Origin of new Species. I shall supplement the examples named with a few more ; they serve to show how general this parallelism between anomalies and specific characters is. Thus, for example, Polygonum viviparum and Agave vivipara bear adven- titious buds or bulbils normally in the inflorescences ; but I found them also as an anomaly in Aloe vcrrncosa and Sa.vifraga nmbrosa. A spiral involution is normally exhibited by the flowerstalks of Vallisneria and Cyclamen, and it occurs as a variety in the stalks of Juncus spiralis, and as an anomaly in Scirpns lacnstris of which latter a beautiful instance came under my notice. Hypocotyl- ous buds are, for example, normally present in Linaria and Limiin ; they occur as an anomaly in Siegesbeckia 1 according to BRAUN, and I have also observed them in Phaseolns multiflorus. The numerous flowerbuds on the leaf stalk of Cucumis sath'iis as described by CASPARY 2 are analogous to the buds scattered on the internodes of Begonia phyllonianiaca. The bulbs of Gladiolus carry their lateral conns on stalks; I observed the same mode of connection as an anomaly in Hyacinthus oriental is. MASTERS has collected a series of teratological cases 3 of buds on leaves, which may be regarded as parallel to the normal instance of the same phenomenon furnished by Bryophyllum. We see therefore that a large number of specific characters are analogous to taxinomous anomalies. The latter recur in related forms, but much more frequently in more or less remote groups. In so far as they are due to a common cause, they point to the widespread 1 A. RRAUN, Verh. d. hot. Vcreins Brandcnb., XII, 1870, p. 151. 2 CASPARY, Ueber Bliithensprosse auf Slattern, Schriften d. phys Gesellsch., Konigsberg, 1874, p. 99 and Table II. 3 MASTERS, Vegetable Teratology, p. 170. Progression, Retrogression and Degression. 71 existence of latent characters. I shall refer to this mode of formation of species as degressive. In it, that which arises is always something new, and often something strikingly new, but usually without any clear relation to the progressive development on the main lines of evolu- tion. They form, rather, lateral improvements of types already existing. Degressive formation of species is therefore due to the activation of long established latent characters. Of these, as GOEBEL has shown in his Organographie, there are two types to be distinguished. 1 Either the character in question was active in the earlier ancestors, or it was not. In the former case we have an instance of reversion or atavism, and moreover a genuine systematic reversion, at least inasmuch as the ancestral relation can be demon- strated. In the other case we have only the development of a specific character from a taxinomous anomaly. It is perhaps hardly necessary to state that the appli- cation of this criterion of grouping can only be effected at the present moment in a relatively small number of cases. The information at our disposal does not as yet meet the demands of such a system. On the physiolog- ical side, however, the question of prime importance is only the distinction between the chief groups; so that we will only lay stress on that point here . Bearing this limitation in mind we can summarize what we have already said, as follows : THE ORIGIN OF NEW SPECIES. A. By the formation of new characters: Progressive specific differentiation. B. Without the formation of new characters. 1 K. GOEBEL, Organographie, Vol. I, p. 170. 72 The Different Modes of Origin of new Species. B\, By characters already existing becoming la- tent: Retrogressive specific differentiation, cases of atavism. 82- By the activation of latent characters : De- gressive specific differentiation. a. From taxinomous (latent) anomalies. b. As genuine atavism. #3. From hybrids. This list does not of course claim to be complete. There is no doubt a whole series of further types which can be more or less easily ranged under or parallel to these subdivisions. On the other hand it is at once clear that the distinction between A and B is, in the present state of our knowledge, the important thing, and more- over that it will suffice as a basis for experimental in- quiry. But before I proceed to illustrate this antithesis I will offer some remarks on the last section (3). New species can arise from hybrids but specific char- acters cannot arise by means of hybridization ; or, we may say that with regard to the production of mutations, hy- brids behave just like ordinary species, except that ac- cording to the prevalent view they are slightly more prone to it. The existence of a vast number of species, however, is due to the various combinations of characters which also exist in closely allied or in remote species. And it is evident that, by crossing, characters can be com- bined which have not appeared in the same genealogical line but in distinct though allied species. Thus for ex- ample by crossing Oenothera nibrinervis with O. nanella I obtained an O. rubrinervis-nanella which has remained constant for many generations without segregation and Progression, Retrogression and Degression. 73 without reversion. And a host of new species have doubtless arisen on similar lines. Coming now to the discussion of the difference be- tween our two groups A and B, we draw a distinction between progressive specific differentiations or the origin of new specific characters on the one hand, and rctro- and degressive specific differentiation, which consists in the activation or latency of potentialities already in ex- istence. It is obvious that a premutation is necessary for pro- gressive but not for retro- and degressive differentiation. For in the case of the former the new potentialities must first arise before they can become visible externally, whilst in the case of the two latter we are only dealing with potentialities already existing. I propose therefore to apply the results obtained with Oenothera Laniarckiana and the conclusions regarding the premutation period to which we arrived, to the further elucidation of this ques- tion. 1 It is of course a purely speculative discussion that we are embarking on, but one which will, in my opinion, materially help in clearing the ground. And I may therefore say, in anticipation, that this theory is supported by the experiments to be recorded in this sec- tion and most strikingly by the history of my Linaria z'lilc/aris peloria (see 20). I have already stated, in Vol. I, Part II, that I regard the mutational period in Oenothera Laniarckiana as a type of the mode of origin of species in general; that is to say, of the essential form of that process, the pro- gressive type. 2 We often find in the vegetable kingdom analogous groups of closely related species which are 1 See Vol. I, Part II and especially 31, p. 490. 2 Vol. I, p. 259- 74 The Different Modes of Origin of nciv Species. usually ranged as elementary species of larger species, but sometimes regarded by the best authorities as "good" species. The group most closely related to our experi- mental series is that of Oenothera bicnnis or the subgenus Onagra; 1 more remotely connected are the groups of Hicracium, Rosa, etc., or of Draba vcrnci, Viola tricolor etc. Such groups appear to us as the relics of past periods of mutation. The new forms which arise from such periodical mutations are to be regarded as homonomous subdivisions of the older species or as elementary species. It is natural in such periods not only that new specific characters should arise, but that old established latent ones should reappear more easily than at other times ; and among the mutations of Oenothera Lamarckiana our O. nanclla is undoubtedly analogous to typical horticul- tural dwarf varieties, and 0. laei'i folia to those systematic varieties which arise by the loss of a character. These latter, however, and similar retro- and degres- sive changes are an entirely different matter. The essen- tial condition for their production is always present, and all that is needed is the external stimulus to induce the mutation. This, it appears, need not occur periodically, nor affect several characters at the same time. New horti- cultural varieties appear at irregular intervals of time, and here and there in the area of cultivation of the spe- cies. But it is equally certain that we have to do in such such cases exclusively, or almost exclusively, with retro- gressive and degressive changes. 2 Analogy and paral- lelism are universal, and their effects sometimes go so far that the characters of the species fall into the back- ground. Double flowers look so much alike that one 'See Vol. I, p. 439; and 31, p. 490. 2 1 am excluding from this consideraton the effects of crossing. Progression, Retrogression and Degression. 75 often cannot tell, even from the best illustrations, to which genus or family a given instance belongs. I shall therefore throughout this Part attempt to describe the origin of horticultural varieties as exhaus- tively as possible. In the present state of our knowledge they form in my opinion the pattern of retrogressive and degressive formation of species ; just as the mutations of Oenothera were the pattern of progressive changes. Together they give us some idea of the main lines along which specific differentiation takes place in nature, at the present time as well as in the past. In conclusion : Progress on the main lines of descent results from the production of new characters ; but the extraordinary variety of forms results from the occa- sional disappearance of characters already existing, or from the activation of latent ones (retrogression, de- gression, atavism). IV. THE SUDDEN APPEARANCE AND THE CONSTANCY OF NEW VARIETIES. 8. EXAMPLES OF CONSTANT RACES. Horticultural varieties are generally constant ; excep- tions to this rule are usually noted expressly in the text- books. Most varieties are not only constant from seed but also pure. By constant is meant that in ordinary cultivation they produce no more impurities than are un- avoidable (that is to say, at most 3%). Absolute purity means that when isolated under experimental conditions the seeds reproduce their own variety without exception. Constancy in this case is complete, but it is seldom of practical interest to bring either the old established sorts or the novelties to this pitch of purity, or even to find out how closely they approach it. This has, however, been repeatedly done by scientific investigators and especially by DAR\VIN and HOFFMANN. 1 Insufficient familiarity with the danger of chance cross- ings robbed the results of the older investigators of much of their value as evidence, except of course in those cases where the race proved constant. The large number of observations of instances of complete constancy were ob- 1 See the Riickblick auf meinc Culturvcrsuchc of the latter author in the Botanischc Zeitung, 1881, and the literature cited there. IHNE and SCHROTER have given a complete list of HOFFMANN'S papers in the obituary of him in Bcrichte d. d. bot. Gescllsch., Vol. X, 1892, p. 18 of the last part. Examples of Constant Races. 77 viously made under conditions which excluded the effects of crossing. In spite of the existence of these experimental data, it is still the general view that varieties are inconstant forms. That which distinguishes them from true spe- cies is supposed to be their faculty of giving rise to occasional and not even rare reversions to the type of the species. This, moreover, is supposed to be a proof of their assumed relation to their species. Every one of course is free to choose his own defi- nition of a variety. But he who makes inconstancy an essential part of the definition will have to exclude a very large number and perhaps the most important of our horticultural varieties, and regard them as elementary species. I have spent much time in the endeavor to test the constancy of horticultural and also of wild varieties with a view partly of directly satisfying myself as to their purity and partly of finding inconstant forms for sub- sequent experiments. I have usually started from seed but sometimes, in the case of perennial varieties, from bought plants. Whenever possible the visits of insects were excluded and the plants artificially pollinated. But in the great majority of cases pollination has to be left to bumblebees and moths, and we must be content in providing complete isolation. The most important point is the extent of the experi- ment. Absolute constancy can obviously never be di- rectly demonstrated. The space and time needed for other experiments seldom permit the bringing to flower of more than a few thousands of plants of one sort. And even if this is continued for several years the possi- bility of the occurrence of rare cases of atavism (e. g., 78 Sudden Appearance and Constancy. once in a million) can not be excluded. The object of experimentation cannot therefore be to demonstrate ab- solute constancy. The best plan usually is to be content with a few hundred individuals; it is even often impos- sible to get sufficient seed for more. Experiments on a smaller scale should only serve to confirm the results ob- tained in other cases; but even if they only do this they are, in my opinion, by no means without value. Fig. /. Bidcns tripartita. Type without ray florets. The nearest that we can get to demonstration of ab- solute constancy is to make observations on races which grow in vast quantities in certain districts and are never- theless true to their type. In these cases the constancy is so striking as to induce many systematists to regard the form as a species. Amongst the better known ex- Examples of Constant Races. 79 amples of this are the Discoidca forms of many com- posites. MOOUIN TANDON regarded the Discoid 'ea, i. e., the *** fj form without tongue-florets, as the Peloria of the Com- posites. 1 They are generally regarded as having arisen from the Radiata forms. Sometimes the discoid form is Fig. 8. Senecio Jacobaea L. (/. radiata}. commoner than that with rays, and then the Discoidea form may be found described in systematic works as a species and the Radiata as the variety ; as for instance in Bldcns tripartita (Fig. 7), and B. cernua, 2 although B. 1 Teratohgie vegctale, p. 179. 2 KOCH, Synopsis Florae Gcrmanicae, p. 309. 80 Sudden Appearance and Constancy. grandiflora, B. bipinnata, and B. atropurpurea are well- known species with ray florets. B. trlpartlta and B. cermta are very common in Holland and I have often tried to find or to obtain examples with ligulate florets, Fig. g. Senecio Jacobaea discoidcus, KOCH. but as yet in vain. By this fact both forms are proved to be constant as thoroughly as a proof can be. In other countries, however, the varietates radiatae are known to Examples of Constant Races. 81 occur. Similarly Scnccio Jacobaca has a Forma radiata and a F. discoidea 1 (Figs. 8 and 9) both of which grow in this country and are absolutely constant. The Discoi- deus grows in thousands of specimens in the dunes in the province of North Holland; but the Forma radiata grows equally abundantly in South Holland; both are amongst the commonest and most widely distributed spe- cies of our flora. For twenty and more years I have had them under observation, and never saw any trace of ad- mixture or reversion ; the two varieties were always ab- solutely pure in the respective localities. Of late, how- ever, there have been some cases of intermingling near the limits of their areas, probably as the result of seed transportation. The two sorts can therefore be regarded as absolutely constant. 2 Matricaria Chamomilla discoi- dea 3 has proved equally constant in my experimental garden, but MURR mentions the occasional occurrence of heads with ravs. 4 In 1897 I raised from the seeds of a j single plant of M. discoidea 575 plants, all of which were without ligulate florets. On these I only harvested the seeds of the weakest branches of the higher orders and raised 460 plants in 1898, all of which again were with- 1 See Vol. I, p. 196. 2 A valuable summary dealing with this point is given by. J . MURR, Strahllosc Bliltlien bei heimischen Kompositen, Deutsche Bot. Monatsschr., Vol. 14, 1896, pp. 161-164. See also Botan. Jahrcsber., T. 24, 2, p. n, where rare instances of forms with rays belonging to normally discoid species and rayless flowers on normally radiata forms, are given. I cite Senecio Jacobaea as an instance of the latter, in opposition to the observations given in the text. An attempt to discriminate half races amongst these forms (See 3, p. 18) would probably lead to valuable results. 3 For an account of the rapid spread of this form in Norway see JENS HOLMBOE, Nogle Ugraesplanters Invandring i Norge, 1900. Nyt Magaz. f. Natitrv., Vol. XXXVIII, p. 187 (with map). The variety is there also fully constant. 4 J. MURR, loc. at., pp. 161-164. 82 Sudden Appearance and Constancy. out ligulate florets. From these plants I only harvested the poorest possible seed on the latest branches after cutting away the main stem and the stronger branches; but from this seed, as before, I obtained nothing but Discoidea (750 plants in 1899). Flowerheads without, or almost without, rays also occasionally occur in races usually normal in this respect. Examples of this have occurred in my experimental gar- den in Chrysanthemum coronarimn, Coreopsis tinctoria, Dahlia striata nana and others. 1 In the first volume I cited numerous examples of constant varieties 2 and showed 3 that many of them were certainly one or two centuries old ; in fact as old, or probably even older than, the cultivation of their species itself. The varieties are generally as constant as the wild elementary species, of which Draba vcrna and Viola tricolor* w^ere cited as examples. Belonging to the same group are the two remarkable types, which HERMANN MULLER has distinguished in Iris Pseudacorus, of which the one with narrow openings to the flower is adapted for pollination by Rhingia, whilst the other is adapted for pollination by bumble bees ; 5 IRWIN LYNCH has re- cently compiled a very complete and valuable list of 1 Further examples are given by MURR, loc. cit. ~ See p. 196. Examples are afforded by GMLLON-strawberries (Vol. I, Fig. 7, p. 34) and by Chelidonium laciniotnin (Vol. I, Fig. 36, p. 190). 3 On page 183 of the first volume will be found a list of the vari- eties known to MUNTING (1671) and still cultivated. 4 See Vol. I, Figs. 3 and 4, pp. 22 and 23. For the constancy of the elementary species of Viola tricolor see also V. B. WITTROCK, Viola Studicr in Acta Horti Bergiani, Vol. II, No. I, 1897 (Cultures extending over three years). H. MULLER, Die Bcfruchtung der Blumen, p. 68. Examples of Constant Races. 83 constant varieties, based on data given by gardeners and botanists as well as on his own observations. 1 It is a common opinion amongst gardeners that white flowered varieties are among the most constant. They are very plentiful and easy to control. From the cases as yet investigated it may be assumed that hybrids be- tween them and the colored species will be colored also, and therefore soon and easily discovered ; so that in the purification or fixation of these varieties the hybrids are usually removed soon and altogether, which is a very important thing for approaching constancy. Sev- eral investigators have tested the purity of white vari- eties. For instance HiLDEBRAND 2 worked with white Hyacinths, Delphinium Consolida, Matthiola incana and Lathyrus odoratus', HOFFMANN with Linnm usitatissi- 11111111 album ; 3 HOFMEISTER for thirty years with Digitalis parviflora alba ; 4 PREHN with Scabiosa alba? etc. I myself have made similar observations. I started by buying a few plants of each of some varieties of perennial species, allowing them to flower on isolated spots and then saving and sowing their seed. Wherever the isolation was complete all the offspring, with a single exception (Aquilegia chrysantha), were white flowered. The following were the species tested in this way (I give in parentheses after each one the number of plants which were raised from their seed and observed in flower) : Cam- pan ula pyram idalis alba (26), C. persicifolia alba ( 1 044 ) , 1 IRWIN LYNCH, The Evolution of Plants, Journ. Roy. Hort. Soc., Vol. XXV, Pt. i, pp. 34-37, Nov. 1900. 2 HILDEBRAND, Die Farbcn der Bliitlicn, p. 7p. 3 HOFFMANN, Botan. Zeitung, 1876, p. 566. See also the very complete list of constant white varieties given by CARRIERE, pp. 12-13, and the literature cited there. 4 HOFMEISTER, Allgemeine Morphologic, p. 556. 5 J. PREHN, Schr. Naturw. Vereins Hoist ein, Vol. X, 1895, p. 259. 84 Sudden Appearance and Constancy. Catananche cocntlca alba (5), Hyssopus officinalis albus (198), Lobelia syphilitica alba (537), Lychnis chalcedo- nica alba (401), Polemonium dissectum album (126), Sak'ia sylvestris alba (296). The following white vari- eties of annual species I also found to be perfectly con- stant : Chrysanthemum coronariiim album (400), Godetia amocna, white Pearl (15), Linum usitatissimum album (779), Phlox Drummondi alba (50), Silene Armeriaalba (617). Among wild species I subjected especially Ero- dium ciciitarium album, which is common in Holland, to a severe test. In this form the pigment characteristic of the species is lacking both in the leaves and in the flowers. I found the variety constant through five gen- erations in my experimental garden, not a single red plant appearing although the sowings were conducted on a very large scale. Later I collected seeds of the "car. alba from another locality and found it also to be con- stant (43 specimens). Other color varieties usually prove equally pure if the seeds of plants that have been isolated are sown. In some cases this fact is so generally known that they have, on this ground, been raised to the rank of species, as, for instance, Anagallis (arvensis} coendca. In 1897 I had 25 examples of this variety flowering on an isolated spot, and from these in 1898 I had 866 plants which were without exception blue. Tetragonla e.rpansa, whose leaves and flowers are normally reddish brown, has given rise to a pure green form which has been raised to the rank of a species under the name of T. crystalline,. This I also found perfectly constant. In 1898 I sowed about 600 fruits obtained from a culture of 1897. Each fruit contains from 6 to 10 and often more seeds, which ger- minate sooner or later, some of them not until after a Examples of Constant Races. 85 few years have elapsed. In the course of the first sum- mer 3975 seeds germinated, during the second 1082, during the third 88, and during the fourth 90. All the 5235 seedlings thus raised were green without a trace of the red pigment, and belonged therefore to the T. crys- tal Una. In this case, therefore, the seeds which germinate late are just as constant as those which germinate early. 1 In other cases where the constancy is just as complete but happens to be less well known, the sorts in question are "only" regarded as varieties. Some of these forms even seem to be wholly unknown in botanical circles, 2 as for instance, Silcnc Armcria rosca whose color is inter- mediate between that of the species and that of the white variety and which is not a hybrid but an old established perfectly constant sort and just as "good" as the other two. In 1898 I had about 4000 plants raised from the seed of isolated plants of 1897 of the Var. fiorc roseo in flower; they were all of the same color as the parent plants. The same result was obtained on a smaller scale in subsequent years. Clarkia pulchella carnca behaved in the same way (50 specimens). I also found the pale flowered Agrostemma Githago nicaccnsis constant (for 10 years), and Hyoscyamus (niger) pallid us (40 spec.) and Agrostemma coronaria bicolor (349 spec.). Further examples of the same phenomenon are afforded by the yellow Chrysanthemum coronarium, the varieties of the flax with white and with yellow seeds, many varieties without the dark patches at the base of the petals, which are characteristic of the species, as in Papaver soinni- fermn Danebrog, Papaver conimutatuni, Madia elegans (Fig. 10) and others. 1 This is not the case with Trifolium incarnatum quadrifolium (See 22). 2 See Bot. Zeitung, 1900, p. 234. 86 Sudden Appearance and Constancy. Another interesting constant variety is Chelidonium majus latipetalum. (Fig. 11), for the possession of which I am indebted to Prof. J. W. MOLL in Groningen. It differs from C. majus in its petals which are so broad that their edges overlap so that they form an unbroken Fig. 10. Madia elegans. Fig. ii. A, B, Chelidonium majus latipetalum. C, D, Chelidonium ma jus. crown instead of an open cross. I found it to be con- stant through several generations. The constancy of the fasciated variety of Myosotis alpestris; Victoria with its broad, many-petalled central flowers, is likewise well known, as is also that of Linaria Examples of Constant Races. 87 vulgaris tricalcarea, 1 of many glabrous forms like Lych- nis vespertina glabra, of thornless types like Datura Stra- monium incnnis, 2 etc., all of which I have tested per- sonally. Space does not permit of the compilation here of anything like a complete list of constant varieties. mm Fig. 12. Melilotus coendca monophylla. Each leaf consists of a single blade but is more or less deeply incised. One of the lateral leaflets in the middle leaf on the right of the figure appears to be quite free. A, a bract from the inflorescence; here the leaves are least incised. 1 J. H. WAKKER, Linaria rulgaris, Nederl. Kruidk. Archief, 1889, with plate X. 2 See Fig. 5 on page 31 of Vol. I. 88 Sudden Appearance and Constancy. I shall conclude by referring to a race of Mclilotus coerulca, the possession of which I owe to the kindness of Prof. M. W. BEYERINCK (p. 63). Its leaflets 1 are fused to a single blade in which the three main nerves still diverge from the base. The blade moreover has three distinct tips, the depth of the indentations between these being subject to considerable fluctuating variability. Not rarely the three parts are only united from the base to halfway up or less, and sometimes they are separated almost to the base and in rare cases even entirely so. All these forms may occur on the same plant. But there was no reversion in my experimental sowings ; every plant exhibited this monophylly to a greater or less ex- tent. 9. STERILE VARIETIES. One of the greatest difficulties presented by the cur- rent doctrine of selection lies, as I have pointed out more than once in the first volume of this work, in the fact that the gradual origin of species, which is presupposed by it, has never been observed. In every case in which observations have been made sufficiently close to the origin of a new form, they indicate a sudden change. We do not find those gradual transitions which the doc- trine of selection would lead us to expect. The new form may be highly variable, and in that way the limits be- tween it and the parent species may sometimes overlap; but, as I have already shown (Vol. I, 25, p. 430) trans- gressive variability of this kind only provides a morpho- logical transitional series and not a genetic one. My object in the present chapter is to bring together 1 This form has been described by WYDLER, Flora, 1860, p. 56, and occasionally since. Sterile Varieties. 89 a list of further instances partly from the literature and partly from my own observations in order to place my conclusions on a broader basis of facts. The difficulty of this task is increased by the fact that it often seems impossible to show how those cases, in which other in- vestigators believed that they had detected transitional series, are to be explained on the theory of mutation. This is especially so where the authors have simply relied on comparative investiga- tions. The results of these can usually be explained, no doubt, by the supposition of transgres- sive variability, but a proof can only be given if the phenomena in question are investigated by statistical methods. In strong contrast to these doubtful cases, however, there is a long series of observations in which the absence of transitions is practically certain. Perhaps the most striking of these are the sterile varieties which consti- tute one of the most serious ob- stacles to the current doctrine of selection, at any rate as regards its exclusive application. DARWIN himself repeatedly cited them as objections and examined them minutely. Fig. 13. A flower of Li- Hum candidum plenum. The thalamus is changed into a long stalk on which the narrow per- fectly white petals are spirally arranged. 90 Sudden Appearance and Constancy. In the case of the vast majority of sterile varieties we know neither how, when nor where they arose. They are propagated by vegetative methods and have been from time immemorial. But they differ from their sup- posed parent species so markedly that they take rank with the best varieties. Nobody supposes that they have arisen gradually. The first instance that I give is Lilium candidum plenum, a form which is on the market. It is a well- known variety, the bulbs of which are offered every year by dealers in bulbs, in their catalogues. Instead of flowers it has long stalks clothed with petals (Fig. 13). The stalk is the elongated thalamus ; the petals are nar- row and dead white, and of the color and structure of the petals of other white lilies. In each flower they continue to appear for several weeks ; the lowest may be brown and withered before the uppermost have unfolded. Figure 13 shows a fairly short flower; they are often twice as long. Stamens and carpels are never formed ; the apex consists of a compact bud of the youngest petals. We do not know how the variety first arose. It was first described by G. VROLIK in 1827, after he had al- ready seen it flowering for 20 years in the Botanical Garden in Amsterdam. 1 It is therefore nearly a century old. In horticultural literature it is not referred to until much later, about 1840. 2 Another well-known sterile garden plant is the green Dahlia (Dahlia varidbilis viridiflora). The flowerheads 1 G. VROLIK, Over ccn rankroruiige ontwikkeling van wittc Iclic- blocnicn. Verhandelingen der eerste klasse v. h. k. Nederl. Instituut van Wet. te Amsterdam, Part I, 1827, pp. 295-301, with one table. The spike with five flowers figured there, is still preserved in our collection. 2 See MERAT, Ann. Soc. d'hortic. dc Paris, 1841-1845, and VERLOT, loc. cit.j 1865, p. 91. Sterile Varieties. 91 are destitute of flowers; and the thin, transparent bracts are transformed into small green leaves. The variety is much cultivated in gardens, partly as a curiosity and partly because their green "flowers" do not wither but remain fresh on the plant ; which renders it of a decora- tive effect until far into the autumn. 1 The variety arose in a crop of seedlings about the middle of the last cen- tury in Boskoop in Holland, and since then has been grown from tubers. It occasionally bears isolated red ray florets but, so far as I know, never sets seed. Some years ago I obtained what seems to be a new and hitherto undescribed form of green Dahlia through the kindness of Messrs. ZOCHER & Co. in Haarlem. It is not known whence this form came because it was at first taken for the type of green Dahlia we have just been considering. It differs from this however in the fact that the green heads are not of the normal form and size but transformed into long green leaf -bearing spikes like that figured in Fig. 14 with the exception of the clump at the top. This form produced elongated flowers of this kind in great numbers in the nursery garden ; but it could never, so to speak, bring its growth to a conclusion. They grow until the autumn and often longer, and frequently attain a length of 30 centimeters and more. They behaved in exactly the same way in my garden until last year when I manured them heavily. Then there appeared from a few of the green "flowers" in late autumn a little head at the uppermost end (Fig. 14). This unfolded, but consisted of green bracts only; it contained neither flow- ers nor seeds. The plant is therefore perfectly sterile. Another variety closely analogous with this is the 1 See the literature in PENZIG'S Teratologic, IT, p. 71. 92 Sudden Appearance and Constancy. wheat ear carnation (Dianthns Ca- ryophyllns spicatus). In this case we find instead of the flowers small green ears which are formed of green bracts arranged crosswise. This sterile form does not appear to be in general cultivation, although it is perennial ; nevertheless it turns up here and there in crops of seedlings especially of mixed varieties. I cul- tivated a plant arisen in this way ; most of the ears were sterile as usual, but some of them produced a flower at their top. From these I obtained several germinative seeds. 1 The green rose has been known from time immemorial, but the green Pelargonium Donalds a modern prod- uct. In both cases the petals and stamens are transformed into green leaves. They are said to be perfectly sterile, and are only propagated by cuttings. Many double flowers never pro- duce seed, and this is especially true of those forms which do not develop structures intermediate between sta- mens and petals, but are described as instances of Petalomania. 2 Ra- 1 After having been sown these seeds have repeated the wheat-ear variety (Note of 1909). 2 K. GOEBEL, Beit rage zur Kcnntniss gc- Fig. 14. Elongated Green fi'illtcr Blilthen, Jahrb. f. iciss. Bot., Vol. Dahlia, a new variety. XVII, pp. 217-219, and elsewhere. Sterile Varieties. 93 nuncidus acris (Vol. I, Fig. 40, p. 194), Caltha palustris, Anemone nemorosa, Hepatica triloba, Tropaeolum ma jus florc plena, Clematis recta, Barbarea vulgar is floribus plenis and many others are alleged in horticultural litera- ture to be perfectly sterile. Double varieties of compo- sites also afford instances such as Achillea P tar mica, Ageratum mexicanum (some varieties), Pyrethriim ro- seum, etc. Others, such as Anthemis nobilis, are known to bear seed from time to time and so do not belong here. Viburnum Opulus, Hydrangea hortensea, Muscari co- mosum plumosum and others have become sterile by the transformation of their reproductive flowers into sterile ornamental ones. Bananas and other seedless fruits have already been dealt with in the first volume (p. 195). Many varieties of the sugar-cane never set seed, such as the Cheribon cane which is the richest in sugar. This variety, which extends over vast regions, consists of a single individual ; that is to say, it is derived from one single unknown stock plant and has always been propa- gated by cuttings or so-called Bibits only. Robinia Pseud-Acacia inermis 1 is also said never to flower; and is only propagated by grafting. If this sterility affects annual species or such as cannot be permanently reproduced by vegetative means, the sterile form must vanish sooner or later. Such forms hardly deserve the name of variety, and are usually spoken of as monstrosities. But, in regard to their ori- gin, they are perfectly analogous with the sterile forms of which we have already treated. In the first volume (p. 195, Fig. 41), I gave the sterile maize as such an instance. More remarkable still is the unbranched Fir (Pinus excelsa aclada or monocaulis) which SCHROTER 1 DE CANDOLLE, Physiologic, II, p. 735. 94 Sudden Appearance and Constancy, has described in his excellent monograph. The whole plant consists of a single branchless stem, which is merely slightly swollen at the limits of every year's growth ; the needles remain adherent for a long time. 1 This form has appeared in diverse localities. SCHROTER records 4 examples from Italy, one from Baden, several from Westphalia, Mittelfranken and Bohemia, and some from Mariabrunn near Vienna. The majority of these plants reached a height of 1-2 meters, a few of them as much as 5-6 meters; some of them are still alive RIMPAU has described an instance of sterile rye. 2 Ears of this rye appeared almost every year during a period of more than ten years; they were often much and sometimes excessively branched, especially in years and localities where the rye was very thin. But as ears of this kind occurred on plants which also bore normal ones, the repeated appearance of the anomaly may per- haps be due to inheritance. And lastly, instead of giving a further record of the numerous existing sterile varieties, let me refer to Nitella syncarpa, which has recently been described by A. ERNST, and which bears, instead of oogonia, incompletely devel- oped antheridia which never produce spermatozoids. 3 The examples in question were observed near Zurich, and were completely sterile. 1 See p. 63 and C. SCHROTER, Ucbcr die Fichtc (Picca cxcclsa Link) Vierteljahrsschr. d. nat. Ges. in Zurich, Jahrg. XLIII, 1898, Parts 2 and 3, pp. 50-53, Fig. 18. This valuable work contains a very full review of the varieties, forms, and monstrosities of this highly "variable" tree. 2 Deutsche landivirthschaftliche Presse, Berlin, October 4, 1899, where photographs of monstrous ears of rye are given. 3 ALFRED ERNST, Ueber Pscudo-Hcrmaphroditismus bci Nitella syncarpa. Flora, 1901, Vol. 88, Part I, with Plates I-III. Races which Have Arisen Suddenly in Nature. 95 10. INSTANCES OF RACES WHICH HAVE ARISEN SUDDENLY IN NATURE. In nature, elementary species are, as is well known, not connected with their closest allies by transitional forms. Nevertheless fluctuating and transgressive varia- bility frequently bring about the appearance of contin- uous series, which however on closer examination espe- cially by statistical methods dissolve into perfectly dis- tinct component units. 1 In very many cases these tran- sitional forms are absolutely lacking, and the separation of a particular form as variety, subspecies, elementary species, or even species, is mainly founded on their ab- sence. The absence of transitional stages in the case of forms which have been long familiar and are widely distributed obviously tells little concerning their mode of origin. Results are more likely to follow from the investigation of those cases in which the types in question are local in occurrence and in which, therefore, if transitional stages should occur, one would expect to find them in the locality inhabited by the plant. In some instances an exhaustive and minute study of the geographical distribu- tion of certain varieties has led to the discovery of the center from which distribution took place. ASCHERSON and MAGNUS obtained a result of this kind with the pale fruited varieties of the European species of Vac- ci nil tin and some related Ericaceae. 2 In inquiries of this kind, the absence of transitions at the present time points to the conclusion that they may never have existed. 'See Vol. I, Part II, 25, p. 430. 2 P. ASCHERSON and P. MAGNUS, I'crhandl. d. k. k. zool.-botau. Gesellschaft in iricn, 1891, p. 677. 96 Sudden Appearance and Constancy. In contradistinction to these more or less distributed varieties, there is a whole series of records scattered through the literature of cases in which a new form has been found on a particular spot under circumstances which warrant the conclusion that it has arisen exactly there and fairly recently. In such cases transitional forms are always lacking, a fact which proves pretty conclu- sively that such have not been produced in the origination of the form. In Part II of the first volume two cases afforded by Oenothcra Lamarckiana were described in detail : I refer to the appearance of O. brevistylis and O. lacvi folia on the original locality at Hilversum. Both species proved, when tested, to be perfectly constant from seed, without any atavism; and transitional forms were not seen in the field. If these species had arisen where I found them, their origin must have taken place between the year of the introduction of the species and the first year in which I discovered them; that is, between 1870 and 1886 (See Vol. I, p. 266). The most important and accurate observation of such an occurrence is that which has recently been recorded by SOLMS-LAUBACH, and deals with a species newly arisen from Capsella Bursa Pasioris. 1 This was found by Professor HEEGER in the market place near Landau amongst the ordinary Shepherd's purse and called C. Hccgcri after him, by SOLMS. It occurred in 1897 and 1898 in very small numbers and only on this one spot. In its vegetative parts it is exactly like C. Bursa pastoris, from which it only differs in the form of its fruits. But the differentiating characters are of the rank of some of 1 H. GRAF zu SOLMS-LAUBACH, Crucifcrcn-Stiidicn, Botanische Zeitung, 1900, Heft X, Oct., i, 1900, pp. 167-190, Plate VIII. Races which Have Arisen Suddenly in Nature. 97 those which serve to separate genera amongst the Cru- ci ferae. The fruits of Capsella Heegeri are oval, and about as thick as they are broad. The seeds are notorrhizous. The valves lack the firm anatomical structure, character- istic of the normal valve, but are soft and full of sap, a condition which may be considered as due to arrested development. On the weaker branches in the autumn, deviations from this type occur which revert more or less to that of C. Bursa ; moreover the flowers and young fruits may develop into malformations, as the result of the attacks of Cystopus candidus, which closely resemble those of C. Bursa pastoris. The seeds of isolated plants of C. Heegeri gave rise solely to the parent type (382 examples) without rever- sion to C. Bursa. There can therefore be scarcely any doubt that C. Heegeri is a good elementary species which arose from C. Bursa in 1897, or a few years previously, somewhere near Landau. It is moreover a species which is dis- tinguished from its nearest allies by characters of far greater systematic importance than those which separate many species of known origin. I myself found a Stcllaria Holostea apetala not far from Wageningen in Holland under similar circum- stances (1889), and also in the same year the well-known Capsella Bursa Pastoris a petal a 1 near Horn in Lippe. But I did not succeed in obtaining seed from either of them. In 1888 I collected some seed of Lychnis I'csper- tiua not far from Hilversum and obtained some per- fectly glabrous plants by sowing it. The new variety L. v. glabra proved fully constant as soon as I was able 1 See PENZIG, Tcratologic, I, p. 267. 98 Sudden Appearance and Constancy. to isolate it, and has maintained itself up to the present day without ever reverting. So far as published data go, forms which have sud- denly appeared in nature, or have not previously been noticed, prove constant, provided that cross-pollination is guarded against. In the opposite case they will prove themselves pure as soon as they can be isolated. One of the oldest cases in point is the constancy of Ranun- culus art'cnsis incnnis which was established by HOFF- MANN. 1 The majority of records refer to trees of which the larger number of varieties, if not all, according to DARWIN himself, have arisen suddenly, 2 such as the weeping oak, the weeping white hawthorn, etc. 3 A single specimen 4 of Fagns sylvatica aspleniifolia was found in a wood in Lippe-Detmolcl and could be multiplied from seed. According to LOUDON, Ta.rus baccata fastigiata was found in 1780 growing wild in Ireland; 5 but no pure seedlings of it have been obtained since only one speci- men was observed (a female one). The above list of cases is not a rich one ; but it makes no claim to completeness. The observations in point are, with few exceptions, relatively incomplete inasmuch as there is always the possibility that the first discovery of the new species or variety may have been preceded by a long period of evolution. If we assume this to be true, the absence of transitional forms and the constancy of the 1 HOFFMANN, Bot. Zeitung, 1878, p. 273, where several other examples will be found. 2 DARWIN, Variations, I, pp. 461-463. 3 Further examples are given by BRAUN, Vcrjungung, p. 333 (the sudden origin of red-leaved varieties of Quercus, Corylus, etc.). 4 RATZEBURG, cited by BRAUN in Abh. d. k. Akad. Berlin, 1859, p. 217. 5 L. BEISSNER, Handbuch der Nadclhohkundc, 1891, p. 169. A great number of further examples is given in this work. Horticultural \\irieties Arisen Suddenly. 99 new form are the only arguments for its sudden appear- ance. 11. HORTICULTURAL VARIETIES WHICH HAVE ARISEN SUDDENLY. It is a matter of common knowledge that horticultural varieties have very often arisen by sports. But opinions differ on two points. One is an empirical one and relates to the question of constancy; the other relates to the meaning of the word variety. The two points are nar- rowly bound up with one another. If the new form is not constant and pure from seed but frequently reverts to the parent species it is usually supposed to be derived from that species and is treated as a subdivision of it. But if the new form is as constant as the parent species, the empirical means of demonstrating its relationship are lacking, and the conclusions are drawn from his- torical data and based on analogy; a proceeding which, as we all know, often leads to differences of opinion. Besides the historical records the main point in such cases is always the proof of the constancy from seed. But inasmuch as the interest of the practical man only extends to the question whether the variety can be con- veniently multiplied by seed and is not concerned with the possibility of occasional reversions, such information, especially in older cases, can only be accepted with cau- tion. \Yith this reservation, I propose to give a brief review of some of the better known instances. But before I do this I will call attention to a very beautiful variety which I have not yet found described nor seen in trade- catalogues, but which has appeared in my own cultures. 100 Sudden Appearance and Constancy. Fig. 15 represents a single Dahlia, whose ray florets are all transformed into long and broad tubes which are open above. The same thing occurs in many other com- posites, for instance in Chrysanthemum segetum fistu- losum, Coreopsis tinctoria fistulosa, etc. On the analogy of these cases I propose to call this new Dahlia, DaJilia variabilis fistulosa. This variety arose from a crop raised Fig. 15. Dahlia variabilis fistulosa , a new variety which has appeared in my cultures. from the seeds of D. far. Jul. Chretien, a dwarf single Dahlia with red flowers of the color of red lead, the tubers of which I had bought in 1892 in Lyons. From the seeds which I saved in that year from this variety, I raised in 1893 several plants of which one had a white flower. I only sowed seeds of this in 1894. 1 It was I have unfortunately not yet succeeded in fertilizing Dahlias artificially by their own pollen. Horticultural Varieties Arisen Suddenly. 101 in the crop thus raised that the plant which bore the flower head shown in Fig. 15 appeared. The color was dark carmine red, not that of red lead. The flower heads were all fistulous from the beginning of June until well into October; but the later flowers manifested the ab- normality in varying degrees. Either the base only of the tube was closed; or only some of the ray florets had the form of a tube. The plant had to be left to free crossing with its neighbors so that no observations of real value as to its constancy could be made. Neverthe- less this was pretty considerable, for, from the seeds of my fistulosa I raised 43 plants in 1895 of which 25, that is to say more than half, had the characters of the new variety. The origin of Chclidoniuni laciniatum from C. ma/us was described in detail in the first volume (p. 189, Figs. 36 and 37) ; where a series of other cases will also be found. VERLOT (loc. tit., p. 34) describes Ageratmn cocruleuin nanuin as a novelty which is sometimes sterile, but sometimes occurs as a fertile and constant variety. Verbena hybrida, "a fleur couronnee" arose about 1889 from the variety "a fleur d'auricule," it immediately proved constant and after only two years was put on the market by E. FOURGEOT of Paris. 1 Robinia Pseud-Acacia rosea was found by DECAISNE in a crop of ordinary Acacias; and Glcditschia sinensis incruiis arose in the same way, as also did Sophora japonica pendula which appeared in M. JOLY'S nursery garden in Paris about 1S00. 2 In 1860 a new strawberry "Rcus van Zuidwyk" appeared in Boskoop. Its leaves and fruits were larger 1 See his Catalogue for 1891. 2 VERLOT, loc. cit., pp. 59, 92, 93. 102 Sudden Appearance and Constancy. and altogether better than any varieties then known; it was constant from the first and spread rapidly. I shall conclude this summary with a reference to the new species of Tomatoes which BAILEY has recently described. 1 He describes the origin of two new forms which he has called Upright and Mikado and which arose in his cultures. They differ from one another and from the parent species by more definite and more numerous characters than many among the older forms which are recognized as good species in the genus Lycopersicum. They arose suddenly as usual and were propagated by seed. The observations recorded in this and the two preced- ing sections, which are far from constituting complete lists, show that the origin of varieties and of elementary species both in the garden and in the field is amenable to experimental investigation, for the phenomenon is by no means so rare as is generally believed. The botanist will investigate the indifferent and useless forms with just the same result as the profitable ones, to which alone, of course, the practical man pays attention. The cultures need not be very extensive to afford novelties from time to time, though these must not be expected the first or every year. Once obtained, all that there is to be done is to isolate them as soon as they appear and pollinate them artificially. But it is far more important to go back to their ancestors, partly not to lose the historical evidence, but mainly in order to sow the seeds of these ancestors again and to find out if the novelty will be again produced, and if possible to discover the conditions which determine its appearance. Unfortunately there are many plants which do not lend themselves to such experiments, *L. H. BAILEY, Survival of the Unlike. Horticultural yaricties Arisen Suddenly. 103 either because they produce no seed or yield too small a harvest when self -fertilized or because they cannot be artificially fertilized on a sufficiently large scale or be- cause the number of seeds produced, even under normal conditions, is too small. Moreover one is almost abso- lutely confined to annual or biennial species or to such perennial ones as flower freely in the first year. But in spite of these difficulties and of the incom- pleteness of the observations made hitherto, we may safely conclude from them the possibility of an experi- mental study of the origin of horticultural varieties. 1 1 1 shall describe an experiment of this kind with Linaria vul- gar is pcloria in 20. V. ATAVISM. 12. ATAVISM BY SEEDS AND BUDS. HOFMEISTER in his Allgemeine Morphologic defines atavism in these words : "The occurrence of reversions, the offspring of a variety of known origin resembling the parent type" (p. 559). According to the meaning of the word "known" in the above definition the term atavism may embrace quite a series of phenomena of the most diverse importance. It may mean either that this origin must have been actu- ally observed, or that it can be inferred with sufficient certainty from comparative and systematic studies. If we are merely dealing with morphological questions this distinction may appear unessential, but as soon as our object is to test by experiment the results obtained, it becomes of the highest importance. For to obtain true experimental proof of atavism it is obvious that the origin of the forms should be known directly by observa- tion. The origin of a whole series of varieties and ele- mentary species from their parent forms, however, is sufficiently established by the historical evidence relating to their first appearance. It seems therefore feasible to confine our attention to such cases and to draw a distinc- tion between physiological and phylogenetic atavism. The former is reversion to actually known ancestors, the latter to systematic ancestors. Atavism by Seeds and Buds. 105 But before I proceed to examine these two forms of atavism more closely, I think it desirable to state that I here use the word "atavism" in its narrower sense, for in its wider sense it embraces so large a group of phe- nomena that it would not be possible to deal with them all within the limits at my disposal. It seems worth while to indicate the more important of these types because they are often confused with one another and because results obtained with one form are often taken to apply to an- other, simply because they both go by the same name. We must first of all draw a sharp distinction between atavism as applied to variability and as applied to muta- bility. In the first case we are dealing with the phenom- ena presented by a single heritable character ; in the latter, with the conflict of two or more. In the improvement of races the offspring do not resemble the selected pa- rents, they always revert partly towards the mean of their ancestors. We are of course dealing in such cases with the phenomenon of regression which was fully dis- cussed in the first volume (pp. 82 and 120) ; and it would be better to refer to all those individuals of less value which are eliminated in selection as regressive and those which exceed the level attained by their parents as progressive. But it is customary to call the former atav- ists ; and, as a matter of fact, they exhibit the degree of development of the characters in question as it was mani- fested by their grandparents and more remote ancestors, and not as in their parents. They could perhaps be called "curve atavists," since this term does not suggest a reversion beyond the curves in question. 1 The most fascinating section of the subject of atav- ism is that which deals with so-called "youth" forms 1 See the pedigree of the many-rowed maize, Vol. I, p. 73, Fig. 18. 106 Atavism. and with related phenomena. GOEBEL'S admirable in- vestigations have demonstrated the wide distribution of these phenomena and their great importance to the theory of descent. 1 It is now a matter of common knowledge that many plants, and indeed whole groups of species, exhibit characters when young which they either lack in the adult state, or which in later life appear only under definite circumstances. BEISSNER'S discovery 2 that whole genera of cultivated Coniferae, such as Rctinospora, are only youth-forms of other known types such as Thuya; and REINKE'S investigations 3 into the earlier stages of Leguminosae, as well as the work of many others, have resulted in the accumulation of a mass of information relating to this subject. Sium and Bcrnla in their early stages have the doubly pinnate and finely slit leaves of their close allies ; the thorns of Berberis on the so-called suck- ers revert to the foliate form. These phenomena, how- ever, fall mostly within the sphere of systematic botany, and only concern the study of variability in so far as they are dependent on external influences. We must further exclude from our considerations the effects of crossing. The so-called reversions of the horticulturists which are brought about either by acci- dental crosses with the parent or by unconsciously using hybrid seed, certainly occupy a very prominent place in the practice of horticultural selection, but they should be rigidly excluded from scientific speculations. And 1 K. GOEBEL, Ucbcr Jugcndformen von Pflanzen und dercn kilnst- Viche Wiederhervorrufung. Sitzungsber. d. k. bayr. Akad. d. Wiss., Vol. 26, 1896, Part III. For further references see GOEBEI/S Organo- graphie dcr Pftanzcn, Part I, 1898. 2 L. BEISSNER, Handbuch der Nadelhohkunde, 1891. 3 J. REINKE, Untcrsiichungen fiber die Assimilationsorgane dcr Leguminoscn, I-III and IV-VII. Jahrbiicher fiir wissensch. Botan., Vol. XXX, Parts i and 4, pp. I and 71, 1897. Atavism by Seeds and Buds. 107 this is true not only of those cases in which the cause of the reversion is perfectly plain, but still more of those in which the facts observed may lead us to suspect a cross either in the previous generation or in more remote years. By excluding such cases, however, the apparent abun- dance of data relating to experimental atavism is very much reduced ; but it is obviously better to build on a few reliable facts than on the highly insecure basis formed by the numerous data which have hitherto been collected. With these reservations I shall now turn to the dis- tinction between physiological and phylogenetic atavism. Each has its own sphere. The object of the study of the former is to discover the laws to which this form of variation conforms. That of the latter is to discover the ancestors of the species in question either by the ob- servation of chance deviations, or by cultures and selec- tion. HEINRICHER'S extensive studies in the genus Iris show how fruitful may be the application of selection in the study of phylogenetic atavism. 1 The cultivated plants of this group are well known to be highly variable, and the favorite Iris Kaempferi with its large flowers affords numerous opportunities for the study of tetram- erous and pentamerous flowers and of other variations. HEINRICHFR, starting from occasional anomalies pre- sented by Iris pallida, and working on a methodical sys- tem of selection, has raised an atavistic race which he calls Iris pallida abavia. 2 The individual anomalies could not, it is true, be fixed although they were selected for three generations, but a series of new types gradually 1 CARRIERE, Production et fixation des varictcs, 1865, p. 65. 2 E. HEINRICHER, Versuche ilber die Vererbung von Riickschlags- erscheinungen. Jahrb. f. wiss. Bot, Vol. 24, Part I, 1892, and Iris pallida abavia in Biolog. Centralbl., Vol. XVI, No. i, p. 13, 1896. 108 Atavism. appeared and threw a definite light on the probable na- ture of their common ancestor. This is regarded as being an extinct form, with an hexamerous perigon of equal petals, and six stamens. A still living form, Iris falci- folia, possesses such a perigon but has only three stamens. The reader who is interested in this branch of in- quiry and in the highly important results which it has afforded, is referred to the works of this author for further information. I now return to the main question, viz., that of phys- iological atavism. Here we are concerned not with the production of new forms but with an inquiry into the processes which underlie the reappearance of preexisting characters. The character in question is, therefore, one that is still retained in that species from which the one under investigation is descended. Atavism is in this case to be regarded as an oscillation between two empir- ically known extremes. The field of oscillation can ob- viously not be very considerable, for only in cases of very close relationship is the common origin of two forms historically known to us. In this restricted province also, atavism may be brought about by fluctuating variation as well as by muta- tion. In the case of the former it is merely a transitory phenomenon and dependent on external conditions; but in the second case it leads to the origin of a race which externally resembles the ancestors of its parent form. Variational atavism seems to be a phenomenon which plays a large part in the sphere of semi-latent characters. As an example of this I cite the case, described above, of the five leaved clover ( 5, p. 36) which always bears a certain number of trifoliate leaves especially under unfavorable conditions. These trifoliate leaves obviously Atavism by Seeds and Buds. 109 constitute a reversion to the normal clover leaf but, on the other hand, they are merely the extreme variants in the curve of the five-leaved race (Fig. 6, p. 48). A similar state of affairs pre- vails in numerous cases of semilatency where the range of variation of a character is occasioned by the antag- onism of two characters. Mutational atavism must obviously be as rare as mu- tation itself. The reversion of striped flowers to self- colored ones, the heritable atavism of Plantago lanceo- lata ramosa, and the incon- stancy of the peloric Lina- ria, are facts which we shall have to consider below. Physiological atavism can be manifested by plants propagated by seeds or by buds. In the case of the former definite proof is only possible under excep- tionally favorable circum- stances; in the case of the Fi f } 6 - .Cephalotaxus pedunculata fastigiata. The mam stem bears latter it is at once evident the upright branches with leaves /T-" 1^- A \ '-ni 1 inserted on all sides, character- ing. 16 at A). ihe pub- istic O f the var iety; but has pro- lished records of atavism uced at A, where a branch has been cut off close, several in crops of seedlings are branches with flat spreading bi- 1 i 1 serial leaves such as are char- always subject to the StlS- acte ristic of the parent species. 110 Atarisni. picions indicated above. I mean that they occur so rarely and in so few individuals that the possibility of a previous cross, by means of insects, with the pollen of allied forms, even if growing a long way off, can never be quite ex- cluded. It is only in cases in which, as in that of Ociw- tlicra sciutiUaiis (Vol. I, pp. 245 and 377), a species pro- duces a large number of atavistic individuals every year, that the phenomenon easily lends itself to experimental study. On account of the circumstances indicated, it is not possible to say whether atavism in plants propagated by seed is a common or a rare phenomenon. It is certainly much rarer than the practical gardener usually imagines. I have observed in my cultures a number of cases which might have been called atavistic with more or less cer- tainty, but only the cases of regularly inconstant races, such as those of Plantago and Linaria, and the phenom- ena presented by striped flowers, to be described shortly, seem to me to be sufficiently well established to be ad- duced as instances of atavism. Atavism by bud-variation, on the other hand, is a well-known phenomenon. One of the best instances is shown in Fig. 16. It represents a vertical branch of a bush of Ccphalota.rus pedunculata fastigiata (Podocarpus Koraiana Hort}. Below the middle of the figure can be seen the place where a branch has been cut off, and from the side of its base some lateral branches have arisen with flat spreading leaves (Fig. 16 A). 1 The variety Fastigiata has erect branches only and their leaves are inserted on all sides ; but the branches at A have the structure of the parent species, C. pedunculata ; their 1 For a series of interesting experiments relating to this subject see Mutations ct traumatismcs by L. BLARINGHEM (Note of 1909). Atavism by Seeds and Buds. Ill leaves project to right and left, and their side branches are horizontal, making the whole shoot flat with definite dorsal and ventral surfaces. The bush which grows in our garden and bears several branches with similar bud- variations, I owe to the kindness of MESSRS. ZOCHER & Co., nurserymen in Haarlem. The variety can only be propagated by cuttings, as it never flowers, 1 and these produce reversions of this kind pretty regularly, both in the nursery of Messrs. ZOCHER & Co. and elsewhere. It appears to have been first observed in 1863 by CARRIERS in Paris, 2 and since that time by many others. This re- markable case is well worthy of a closer study. The perfectly analogous Ta.rus baccata fastigiata never ex- hibits atavism by bud-variations, so far as I know. 3 The phenomena of bud-variation have hitherto not received from botanists the attention they deserve. In a few cases we know that the phenomenon is preceded by a s-ectorial segregation, as for instance in striped flowers (13) and variegated leaves ( 24) ; but as a rule there is no available information even on this point. Another point which awaits investigation is the nature of the offspring of self -pollinated bud-variants. 4 It seems cer- tain that new types sometimes arise in this way, but much of the proof in favor of this will not bear scrutiny. Under these circumstances it seems desirable to direct more gen- eral attention to this phenomenon 5 by means of some 'BEISSNER, Handbuch, loc. cit., p. 181. 2 CARRIERE, loc. cit., p. 44, with Figs, i and 2; see also CARRIERS, Traitc general dcs Conifcres, p. 717; and JAMES VEITCH & SONS, A Manual of the Conifcrae, iSSi, p. 308. 3 See CARRIERE, loc. cit., and BEISSNER, Hand-buck, loc. cit., p. 169. 4 In the older records attention is seldom paid to pollination ; see the literature in CARRIERE, loc. cit., p. 59, and DARWIN, Animals and Plants, I, 525; II, 442, etc. 5 CARRIERE gives a very complete list ; he. cit., pp. 42-56 ; see also 112 Atavism. further examples. They are taken mainly from woody plants because herbaceous and especially annual plants, with the exception of the instances named and of hybrids, very seldom exhibit bud-variations. Green branches on red-leaved bushes and trees are not rare and are for instance often seen in the variety atropurpiirca of Corylus Avellana, C. tiibulosa, Betula alba, and in the copper beech. The red bananas with their red fruits have given rise to a green variety with yellow fruit in spite of the fact that they are sterile. 1 BRAUN mentions an example of Kerria japonica plena which pro- duced some branches with single flowers. 2 On a garden Hortensia producing only large sterile flowers, FOCKE observed a branch bearing inflorescences with little fertile flowers in the middle of a circle of large ornamental ones as in the wild form. 3 Trees with laciniate leaves habitually give rise to re- versions on solitary branches, as for instance Fagus syl- vatica aspleniifolia, Carpinus Betulns hetcrophylla, Sam- bucus nigra laciniata, Cytisus Laburnum quercifolia, Vitis and others. (BRAUN, loc. cit.) The same is true of Sali.r babylonica crispa, of the parsley grape, of nec- tarines, and especially of roses and bulbs (Hyacinthus, Gladiolus, etc.) although the possibility of previous crosses makes the latter cases still doubtful. In conclusion, this list shows that the series of cases which are amenable to experimental study is by no means small. On the other hand the number of examples is sufficient to demonstrate the pretty general occurrence HOFFMANN, Bot. Zeitung, 1881, p. 395; DARWIN, loc. cit., I, pp. 476- 530; HOFMEISTER, Allgemcine Morphologic, p. 560, etc. 1 FR. MULLER, Flora, Vol. 84, 1897, pp. 96-99. *Abh. d. k. Akad. Berlin, 1859, p. 219. 3 Abh. d. Naturf. Vere'ms Bremen, Vol. 14, 1897, p. 276. The Oriyin of Striped Flozvcrs. 113 of reversion of varieties to their parent species, and there- fore to suggest that the characters of the latter were not lost when the variety originated, but only became latent. 13. VILMORIN'S SUGGESTION AS TO THE ORIGIN OF STRIPED FLOWERS. One of the oldest and best-known instances both of bud - variations and of sectorial splitting is af- forded by certain so- called variegated garden flowers and particularly by the annual Larkspurs, Delphinium Ajacis and D. Consolida. All phases of the phenomenon can be followed in this case with great ease, for from time immemorial these varieties have borne flowers which show the most varied striping on a background of a dif- ferent color; and they also produce flowers a half or a third or some other fraction of which uniformly bears the color which commonly only appears in stripes (Fig. 19). Flowers of this Fig. 17. Delphinium Consolida stria- i i , turn plenum. A plant in flower. kind may be scattered over the whole plant, but are oftener distributed in such 114 Atavism. a way that those on one side of a spike are uniform and those on the other striped. 1 Flowers which are inserted at the boundaries of the two regions exhibit on one side the color of one sector and on the other half, the stripes of the other. A diagram of such a branch is shown in Fig. 18 in which the flowers Nos. 1, 4, 6, 9, and 11 are dark blue, Nos. 2, 5, 7, 10, 12, and 13 pale red with scat- Fig. 18. Delphinium Consolida stria- turn plenum. Diagram of a branch of which the left half was blue, and of which the right bore flowers with fine blue stripes on a pale red background. 1899. Fig. 19. A sectorial flower of the same variety. The whole right half was dark blue ; the left, pale red with scattered blue stripes. tered blue stripes, and Nos. 3 and 8 half blue and half striped. I obtained this branch in my culture of 1899; similar cases are not at all rare. Branches with nothing but blue flowers also occur, but the seeds obtained from the self-fertilization of such flowers gave rise in my gar- den to the striped variety and not to a pure blue progeny. 1 Exactly the same phenomenon is seen in the seedcoats of Pi sum. The minute purple spotting characteristic of some green-skinned varieties sometimes takes the form of a deep uniform purple. These uniformly purple seeds produce the ordinary form with small purple spots and no more full purples than are usually produced. (Trans- lator's Note.") The Oriyin of Striped Flowers. 115 On the other hand a certain percentage (often 6% and more) of the plants raised from the seeds of striped flowers and especially of sectorial branches are usually uniform blues. 1 The phenomena of segregation which we have been describing are quite common in striped flowers, and any one can observe them in Dahlia varidbilis striata (Vol. I, Fig. 14, p. 54), Mirabilis Jala pa, Verbena and many other favorite garden flowers. Sectorially colored flow- ers appear to manifest a tendency towards a simple pro- portion between the two parts. Frequently exactly half of the flower is atavistic, sometimes a quarter or three quarters. I observed the proportion % in white and red striped tulips and in partially dark blue and partially pale blue flowers of Iris xiphioides, etc. In these cases the various types frequently occur on the same plant, or in the case of plants grown from bulbs, on examples raised through vegetative propagation from a single original bulb; for instance on the tulips and Iris just mentioned there were also flowers of which one-half of each was atavistic. Sectorial variability often occurred in my cultures, as for instance in the flowerheads of Helichr\suin brac- teatiun and the flowers of Papaver nudicaide (Fig. 20), in both of which cases stripes or sectors of the color belonging to the parent species were superimposed on the paler background of the variety. A common balsam (Impatiens Balsam ina') whose flowers were usually white with fine red stripes bore a branch with red flowers only in my garden. The whole breadth of the fasciated stem 1 A point of great interest to investigate would be the relation between sectorial variability and cell division in the vegetation cone ; clues which might lead to the solution of many important questions would probably be afforded by such an inquiry. 116 Atavism. of the striped cockscomb or Celosia variegata cristata, is traversed by longitudinal stripes of different colors, yellow and red according to the variety. Dahlias how- ever exhibit the most prodigal wealth of color of all variegated flowers, especially those varieties known as Fancy-flowers. 1 In this case the color is in some way connected with the amount of doubling, which often ex- hibits sectorial variations and bud-variations at the same time. 2 Striped Dahlias give rise to these partial variants sometimes very rarely, but sometimes in such abundance that a good variety is often ex- posed to the danger of being lost thereby. In most of the cases we have to do with two types which are manifested in various degrees of association and separation. Cases in which more than two forms are combined and which there- fore may produce by bud- Fig. 20. Papaver nudicaule. Yel- low variety with dark orange stripes. variation two or more types on the same plant, besides the normal one, have been described; but they were probably hybrids. Central dissociation seems to be a very rare phenomenon, but of Mad. H. Vonrchy, a variety which usually has 1 See GROOMBRIDGE'S Treatises on Florist's Flowers; The Dahlia, 1853, and the extensive literature which has appeared since. 2 VILMORIN-ANDRIEUX, Les fteurs de pleine terre, first edition, p. 340. The Origin of Striped Flowers. 117 white flowers with red stripes, I have seen a head whose outer ray florets were dark red whilst the inner ones formed a disc of pure white with only very occasional red stripes. In the center the unmodified fertile yellow disc florets were seen. I have observed the same phe- nomenon in a few other 'cases. The striped varieties of Cyclamen persicum are said to bear in some instances only variegated flowers one year and from the same bulb uniformly colored atavistic flowers the next year. Centaurea Cyanns, the blue corn flower or blue bottle, has a brown variety with double flowerheads which is / highly variable in color; it is far from being fixed yet, as a plantbreeder in Erfurt expressed it to me. I culti- vated it for five years, always selecting the purest and darkest brown specimens in small numbers as seed- parents. The race produced reversions to the blue form every year. Some plants bore blue flowers exclusively, in others the blue color appeared in segments or in stripes on some of the heads. No advance was brought about by this selection. The examples given must suffice to show the impor- tance of the striped flowers of horticulture. A Var. striata of a number of species is advertised in the cata- logues ; it is open to any one, therefore, to cultivate them. The Var. alba of many other species often reveals on closer inspection scattered stripes of the color of the parent species; these stripes can easily be intensified by isolation and selection as I shall show in one of the following sections ( 16). Str-iped flowers 1 are also of great importance in the 1 Spotted flowers may possibly behave differently ; but up to the present time I have not grown them. 118 Atai'isjn. science of variability and mutability, and especially in that of atavism of which they perhaps afford the most beauti- ful examples. As such they have been dealt with espe- cially by Louis VILMORIN whose theory we will now pro- ceed to examine. 1 VILMORIN starts from the observation that striped flowers only occur on those species which are themselves colored, but which also possess a white variety; or if the color of the flower is composed of red and yellow the uniform yellow variety may behave like the white (Mira- bil'iSf Antirrhinum). The first variety to arise is the white (or yellow) from which later on the striped form originates and VILMORIN explains this as a partial re- version to the parent species. White varieties of a large number of decorative plants have arisen in cultivation, and in fact many favorite ones in M. VILMORIN'S own nurseries. They can usually be easily "fixed" in the course of a few years ; that is to say, they are generally constant from the very beginning but have to be purged of the consequences of unavoidable crosses, and this takes a few years, as a rule. The striped sorts do not appear in this period, the hybrids resulting from the crosses are like the parent species and segregate into this and the pure white variety. The striping is not the result of crossing therefore ; moreover in such cases deliberate crossing has only resulted in the production of self-colored and not of variegated flowers. Also, when such hybrids exhibit sectorial variation, the color is in large patches and not in fine stripes. It is not until the white varieties have attained com- plete purity and have proved constant for a considerable 1 Societe Pliilomatiqiic dc Paris, Seance du 17 Janvier, 1852, Pro- ccs-verbaux, p. 9; Notices sur I' amelioration des plantes par le semis, 1886, p. 39; and B. VERLOT, Sur la fixation des varietes, 1865, pp. 62-66. The Origin of Striped Flowers. 119 number of generations that the striping appears. It af- fects almost necessarily, so it seems, every cultivated white or yellow variety. Some are worth putting on the market ; others are not. Amongst the latter VILMORIN ( 1852) has mentioned as an example Clarkia pnlchella, from bought seeds of the white variety of which the striped form has also appeared in my cultures (see 16). The same thing happened with Browallia erecta and Coiu- melina tuberosa. Geranium pratense is only to be bought in two forms, white and blue. I ob- tained seeds from two plants which were bought as Var. alba and raised from them, besides pure whites, plants with all grades of color ar- rangement from striping and secto- rial variations to complete blue (Fig. 21). If it is thought desirable to put the striped variety on the market it must be purified by selection. The striping first appears as single fine i -in TC Fig- 21. Geranium fira- streaks on occasional flowers. It tc ,i sc cl!bum with pie _ these plants are isolated and their seeds sown separately the majority of the plants raised are pure white, but occasional ones are produced with broader and more numerous stripes. The seeds of these are saved, and so on. The object is to isolate the striped race from the white, and this can be attained in the course of a few years. On the other hand the breeder has to fight against the tendency of the striped form to return to the full blue either by buds or through seeds. It is to guard against this that VILMORIN recommends bald bine and white flowers. The dark parts of the petals were blue ; the others white. 120 Atavism. the selection of seeds from the palest examples of the striped forms. Convolvulus tricolor was the first species in which this mode of origin of the striped form was observed (1840). It was followed by Gomphrena globosa, Antir- rhinum ma jus album and luteum, Ncmophila insignis, Por- tulacca grandiflora, and others. Of recent years a large number of blotched varieties have been obtained in vari- ous nurseries; and always, so far as is known, in the same way, by so-called partial reversion of a white or yellow variety to the red or blue color of the parent species. In the following sections we will therefore examine in detail some cases of striped flowers as instances of physiological atavism. 14. ANTIRRHINUM MAJUS STRIATUM. (With Plate I.) Amongst the numerous cultivated varieties of the Snapdragon one group is distinguished by the possession of striped flowers. A bed of these produces a fine and varied show of color. On the other hand the horticul- turist's handbooks state that, whilst the remaining sorts are practically constant, the striped ones leave much to be desired in this respect. 1 Such a statement naturally in- vites the investigator to inquire into the mode of inheri- tance of this character. The striped varieties owe their character to the fact that the normal red color of the wild snapdragon is con- fined to broader or narrower longitudinal stripes. Where the red is absent the pure color of the background be- comes visible. This may be either white, rose, yellow ^ILMORIN'S Blumengartnerei, 3d Ger. ed., Vol. I, 1896, p. 756. Antirrhinum Ma jus Striatum. 121 or sulphur as in the corresponding self-colored varieties. 1 It must further be mentioned that each of these types may exist in a tall, medium or dwarf form. In the ex- periment to be described the form I have used was Antirrhinum majns lutcum rubro-striatum of medium height. The richness of types of marking in these striped varieties is very great. The stripes may be sparse and very fine so that the flowers appear at first glance to be pure yellow or white; or the stripes may be bold and broad and very numerous in such a way that the yellow (or the white) appears in about equal parts with the red. Often half of a flower is entirely red whilst the other half is striped, and so on. 2 If we buy seeds of the striped sorts and sow them, the crop raised is considerably less true than is usual in sowings of bought seeds. In 1899 I sowed samples of different varieties of Antirrhinum ma jus and obtained 26% unstriped individuals from A. in. album rubro- striatnm, and \9% from A. m. tut cum rubro-striatum. In other cases a far higher degree of purity is usually obtained, e. g., in A. m. hit cum I found only 2% im- purities. The admixtures in the striped varieties were in the vast majority of cases uniform reds and therefore closely allied to them. Other deviations were not more numer- ous in the striped forms than in any other variety. The reason for the abundance of the red flowered individuals has been disclosed by subsequent culture ; it is to be sought in the incomplete inheritance of the striped char- 1 A. m. album rubro-striatum, A. m. sulphureum rubro-venosum, A. m. pumilum roscum rubro-striatum, etc. 2 ViLMORiN, Flcurs dc plclnc tcrrc, p. 723. 122 Atavism. acter. For if the seeds of striped individuals which have been artificially self-fertilized are harvested and sown, we usually obtain some plants with uniformly red flowers. The striped varieties therefore give rise to red plants from time to time, and in my cultures, which extend over about eight years, A. m. Intciim rubro-striatum has done so almost every year in spite of being self-fertilized. As the original wild form is uniformly colored (that is, not striped, for the color itself is composed of white, red and yellow) the loss of the striping may be regarded as a case of atavism. Moreover this phenomenon of atavism was exhibited by my cultures in two other forms (Plate I) : on the one hand as a bud-variation in which whole branches of a plant with striped flowers revert to the red type; on the other hand as a lateral or sectorial variation, to adopt HEINSIUS'S term, 1 in which one side of the spike bears uniform flowers, whilst the other bears striped ones. Let us examine these two cases more closely. In the case of bud-variation a striped plant bears a branch all of whose flowers are red, without striping. If, as is usually the case, the plant flowers on 6-8 or more lateral branches the abnormality is very striking. A single plant very seldom bears two branches with red flowers, and it scarcely ever happens, if indeed it ever does, that the terminal portion of the main stem has red, and the branches striped flowers. As a rule it is one of the lower stronger branches which is atavistic and seldom one of the higher weaker ones. I occasionally found a tertiary branch with red flowers, i. e., a lateral twig of a striped branch. As might be expected, the coarsely 1 H, W. HEINSIUS, Over bonte bladercn, Genootschap v. Natmir-, Genees- en Heelkunde, Biologische Sectie, May, 7, 1898, p. 2. Antirrhinum Mo jus Striatum. 123 striped plants exhibit a stronger tendency to produce bud-variations than the finely striped ones. Sectorial variation is very diverse in the manner of its manifestation. I found it as a rule on the main stem, but also on the branches. If the inflorescence is looked at from above, i. e., in projection, one sector is red whilst the rest is white. This red sector often consists of a narrow red stripe only, or of one-half or three-quarters of the whole. As a rule the abnormality extends from the base to the top of the spike ; but it may also be con- fined to part of it, especially when it consists of a narrow line only. A single red flower on an otherwise striped spike is by no means a rare occurrence. On the borders of the two sectors the flowers are often striped on one side and red on the other. As in the case of bud-varia- tions it is the coarsely striped individuals which are most prone to the sectorial dissociation of color. The red color occurs not only on the corolla but also on the stamens. In finely striped flowers the stamens are, as a rule, yellow ; in flowers with broad stripes they are striped or red. The individual stamens in the same flower are usually dissimilar in respect to this character; yet it is difficult to find a strong contrast within a single flower, e. g., a single stamen which is almost red, and another nearly yellow. I have spent much trouble in the attempt to find such flowers, especially in those that had one longitudinal half almost or entirely without stripes. But I did not discover any definite relation between the striping on the stamens and that on the corresponding parts of the corolla. As a matter of fact pure yellow flowers never occur in this race. To a superficial observer it may seem as if they were not rare and even that the red stripes may 124 Atavism. be lacking on whole spikes and sometimes on entire plants. But such absence is only apparent; closer in- spection will reveal the existence of very fine red stripes. I never found a branch on which they were quite lacking, nor a plant, nor even a twig which had reverted to the variety, A. m. hit cum. On inflorescences on which the striping is very meager it may sometimes occur that on a single flower no stripes can be found ; but this is merely an extreme case of that partial variability which all organisms exhibit. This negative result based on eight years' experience is important because it shows us that we are not dealing here with a segregation into two components, e. g., A. ma jus rubrum and A. ma jus hit cum. If we want to speak of a segregation the two units would be the red striped and the uniformly red form. A glance at a bed of these plants is sufficient to re- veal the fact that the breadth of the red stripes exhibits individual variability; moreover that, as might be ex- pected, plants with very fine and those with very coarse red stripes are the rarest. In 1897 I tried to find out if it were possible to express this variability in the form of a curve. At first it seemed impossible to obtain an ac- curate measure of the striping, for it seemed practically unfeasible to determine the sum of the breadths of all the stripes in a flower and to express this sum in proportion to the circumference of the corolla. I succeeded, how- ever, in attaining my object in the following way : I had the average flower on the main stem of every plant in a bed picked by an assistant, and then I endeavored to ar- range these in a series according to their color, ascending from the almost yellow to the completely red. With a group of between one and two hundred flowers this sue- Antirrhinum Ma jus Striatnm. 125 ceeded better than I had anticipated ; for at the end there turned up a certain number of groups which corresponded sufficiently closely to equal subdivisions of a scale to warrant their selection as ordinates. I admit of course that this method is not free from the personal factor; but for the case under consideration it sufficed, since, when the same group of flowers was sorted again, the result agreed sufficiently well with the first trial. I plotted three curves in this way in 1897; each was based on one typical flower of the terminal spikes of all the plants flowering on a bed. The three beds contained the offspring of three individual striped plants of the 1894 harvest, seeds of which had been saved and sown separately; but whose flowers had been left to be polli- nated by insects in the midst of a larger culture. More- over the seed-parents were selected without reference to the degree of their striping, and so the curves give an idea of the average composition of the commercial race, I thus obtained the following table : STRIPES COLOR-EFFECT A B C Almost absent Lemon yellow (g) 6 4 Very fine Yellow 9 18 Narrow Dark yellow 2 12 30 1-2 mm broad Reddish yellow 5 15 53 1-3 mm broad Narrowly striped (s) 18 22 84 1-5 mm broad Coarsely striped 28 22 31 1-6 mm broad Broadly striped (b) 42 21 16 Broad fields Half yellow, half red 26 12 10 Uniform red Red (R) 37 9 15 Number of individuals 158 128 261 These figures are exhibited in the form of a curve in Fig. 22; in the case of the figures under C the scale or unit of the ordinates is half of that selected for A and B. 126 Atavism. The result of this inquiry shows that the first eight groups merge continuously into one another ; but that between the striped and red flowers a broad gulf is fixed. The red are not connected with the striped by a series of transitional forms as the lemon yellow are with the broad striped ; red flowers with small yellow patches may occur, but they are at most very rare. The shape of the curves is far more regular than I had anticipated; but the reds obviously have no place in it; I mean, they are far too numerous in proportion. They are therefore obviously not the extreme variants of the series but constitute a group which is perfectly distinct from the striped although the size of this group varies directly with the amount of striping in the other. After the composition of the commercial race had been determined in this way, my next task was to dis- cover the nature of the offspring resulting from the self- fertilization of the individual components of this diverse assemblage. I have confined the solution of this problem to the three chief types : finely striped, coarsely striped, and uniformly red. Let us begin with the two former groups. The offspring of the parent plant A (Fig. 22 and table on page 125) contained many coarsely striped individ- uals (Fig. 22b} ; when they were in flower I transplanted some very coarsely striped ones to a special bed, picked off all their flowers and young fruits and enclosed all the buds which subsequently opened to insure self-fertili- zation. In the same way I treated some plants from the bed B (Fig. 22B) with almost yellow flowers. I har- vested and sowed the seeds of each plant separately. In August, 1898, when the beds were in full flower, T determined the amount of striping by the method al- Antirrhinum Majus Striatum. 127 ready employed, taking care that the boundaries between the individual groups corresponded as closely as pos- sible with those of the previous year. I succeeded in c I .'/. ft. b. Fig. 22. Antirrhinum inajus hiteum rubro-striatum. A, B, C, curves showing the degree of striping amongst the offspring of three insect-fertilized plants, 1897. g< lemon yellow, almost without red stripes; s, narrowly striped; b, broadly striped; R, uniform red. See table, page 125. recording the rather scanty offspring of four coarsely striped parents. The result is given below. (The indi- vidual seed-parents of 1897 are denoted as Ai--A4.) OFFSPRING OF THE COARSELY STRIPED SEED-PARENTS. Stripes Less than 4 mm broad 1-5 mm broad 1-6 mm broad Broad fields Uniform red AI A z A 3 A* Totals 3 2 6 8 19 5 4 6 9 24 7 8 5 6 26 2 2 5 9 Totals 17 14 19 28 78 These figures are exhibited graphically in Fig. 23B. As the extent of this experiment was relatively small and especially as the proportion of self-colored plants appeared to me very small, I repeated it in the following year. I chose from the broad striped bed of this culture 128 Atavism. a beautiful typical plant with broad stripes but without any broad patches on the corolla, and fertilized it with its own pollen in a bag. In 1899 I raised from its seeds about 250 plants, which covered a bed of about four square meters, and nearly all of which flowered on the main stem and on several lateral branches. There were only a few finely striped individuals amongst them, whereas the majority were very coarsely marked. But the proportion of uniformly red plants was considerable: Striped individuals 160 64% Red individuals 91 36% Total 251 That is to say, about one-third of the olants had re- verted to a uniform red color. The offspring of the almost yellow parents showed the following distribution of the various types of colora- tion (Bi--B4 refer to the individual seed-parents and to the groups of offspring arising from them) : OFFSPRING OF THE YELLOW PARENTS. Stripes BI B 2 B* 4 Totals Nearly absent 6 5 12 1 24 Very fine 3 7 18 2 30 Narrow 3 6 12 2 23 1-2 mm broad 9 7 18 3 37 1-3 mm broad 7 4 22 2 35 1-5 mm broad 3 1 4 1-6 mm broad Broad fields Uniform red Totals 28 29 85 11 153 See Fig. 23A. These tables, and Fig. 23 which has been constructed from them, show that two races have been produced by the selection and self-fertilization of the extreme variants. Antirrhinum Majns Striatitin. 129 One of them, A, consists almost solely of finely striped individuals and contains no red ones. The other, B, consists almost entirely of broadly striped ones together with 11-36% of uniformly red ones. But the separation is not nearly so sharp as between the striped on the one hand and the red on the other, inasmuch as the two curves overlap. Fig. 23. Antirrhinum majus hitcum rubro-striatum. Curves to illustrate the distribution of color amongst the off- spring of self-fertilized individuals from the culture on which Fig. 22 is based. Experiment in selection with broadly and narrowly striped flowers. Curves represent- ing the offspring: A, of the finely striped seed-parents Bi_B 4 ; B, of the broadly striped seed-parents Ai_A 4 . See tables on pp. 127 and 128. For the signification of g, Sj b, R, see previous figure. We now come to the most important part of the ex- periment, the question of the inheritance of the red character. On account of this greater importance I had already given it previously much attention. Here we are concerned not merely with the inheri- tance of the red flowers in general, but with the study of the special cases already distinguished. First we have to consider the red seed variants, then the bud-variants and lastly the single red flowers on striped racemes. Finally it should be possible to test the red stamens of 130 Atavism. striped flowers but I have not yet come across suitable material for the investigation of this point. In 1892 I had raised from bought seed of A. ma jus liiteum rubro-striatiun a large bed of plants the flowers of which were all striped. I gathered the seed of one individual for the next year's crop (1893). I obtained about 40 flowering plants in this way; the majority bore flowers with fine stripes, and here and there flowers occurred of which one-half was a uniform red. There were four plants which only bore pure red flowers. Of these I selected the strongest, enclosed their spikes in bags and fertilized their flowers with their own pollen. Besides these I dealt in the same way with two striped plants, with few and fine stripes. As soon as the seeds germinated in the following spring a difference became visible : the seedlings from the seed of striped plants had green foliage, those from the red, however, were reddish brown. This difference was particularly striking on the under surface of the later leaves of the young plants. On the former bed 152 plants flowered, on the latter 71. Both groups consisted of plants with striped flowers and plants with red ones, but as I had expected, in very different proportions. The proportions in the offspring from the two types of parents were as follows : STRIPED RED Finely striped parents 98% 2% Red flowered parents 24% 76% Most of the striped flowerr were finely striped coarsely striped plants only occurred in the proportions of 6 and 7%. The characters of both races are therefore heritable but, so to speak, incompletely so. We may describe the Antirrhinum Majus Striatum. 131 production of individuals of the opposite race in both cases as atavism. The striped offspring of the red parents resemble their grandparents. The red offspring of the striped parents resemble the wild species, that is, their very remote ancestors. Thus the difference in the in- tensity of inheritance could be expressed in the state- ment that the influence of the nearer ancestors is greater than that of the remoter ones. But this is merely a re- statement of the facts in conventional terminology. It affords no clue to the solution of the problem. Amongst the finely striped individuals in the culture under consideration there were thirteen plants which had, besides the striped terminal portion of the main stem and the several striped lateral branches, one or two twigs with red flowers exclusively. A good opportunity was thus offered of studying inheritance in bud-variants. I owed it to the fact that the seeds had been sown early, the plants had been grown far apart and the ground well manured ; circumstances which together brought about a profuse branching in all the plants. I trans- planted these individuals to a separate spot, picked off all the open flowers and young fruits and superfluous twigs, and enclosed 1-2 striped and 1-2 red spikes in bags to insure pure self-fertilization. PERCENTAGES Plant No. Red Striped Totals Red Striped From the red } spikes I 1 73 27 100 73 27 2 21 12 33 63 37 3 25 5 30 77 23 1 3 93 96 4 96 2 75 75 100 3 1 36 37 3 97 From the striped spikes I I obtained a sufficient harvest of the striped and red spikes of the same plant from three individuals only. 132 Atavism. These produced in the summer of 1895 the result shown in the table on page 131. In other words, the average intensity of inheritance for striped spikes was 98% and for red ones 71%. If we compare these figures with those derived from the previous generation we do not observe any appre- ciable difference between them. In other words, the in- tensity of inheritance exhibited by the red bud-variants is essentially the same as that of the red seed-variants. In the following year I continued this experiment through one more generation by self-fertilizing some striped and some red individuals amongst the offspring of the bud-variants. The seeds of three striped parents gave rise to 67 offspring that flowered, only 5% of which were red ; the seeds of the five red seed-parents, however, gave rise to 127 offspring of which 84% were red. (The percentages in the five individual groups were 71-78-84- 88 and 100.) Thus the proportions were similar to those of the previous year. ANTIRRHINUM MAJUS LUTEUM RUBRO-STRIATUM. SEED- AND BUD-VARIATION (ANNUAL CULTURES). Year 1896 (1895) 95% Striped Striped plant 84% Red. I Red plants 1895 98% Striped (1894) Striped twigs 71% Red. I Red twigs. 1894 1893 1892 98% Striped 90 % Striped plants 76% Red. I 10% Red plants f Striped plant Antirrhinum Ma jus Striatiun. 133 I have exhibited on the opposite page the whole ex- periment in the form of a pedigree. The result of our experiment can be given in yet another form. The intensity of inheritance in the finely striped spikes in successive generations produced by self- fertilization was always about 95-98%. The intensity of the inheritance of the red character in the various subdivisions of the experiment was as follows : 1. For seed variants 76% 2. For bud variants 71% 3. For the offspring of bud variants . .84% Average 77 % Finally I have endeavored to investigate the mode of inheritance in the case of sectorial variation ; that is, of spikes which on one lateral part bear striped flowers and on the others red ones. It is obvious that this phe- nomenon may be due to two entirely different causes. First the red flowers may be genuine bud- variants and, in such cases, they will presumably exhibit an intensity of inheritance which corresponds with that found for the bud-variants dealt with above. But it may also hap- pen that on a very coarsely striped spike some of the flowers may possess this striping in so extreme a degree that they appear uniformly red. In this case their mode of inheritance will presumably not differ from that of the remaining flowers on the same spike. The latter was the case in the only experiment which I have so far had the opportunity of making. In the summer of 1898 I employed for this purpose a broadly striped plant from the crop referred to on page 128. One side of its terminal spike bore red and the other striped flowers. There were 8 of the former and 7 of the latter. I enclosed the whole branch, before it flowered, in 134 Atavism. a bag, fertilized each flower with its own pollen, and gathered the seeds separately. Five fruits of each color ripened, though some of them contained little seed. I sowed the seed in 1899 on ten separate beds; they flow- ered in July. On each bed one saw at a glance that about half the plants bore exclusively red whereas the other, slightly larger half, bore striped flowers. I recorded the numbers separately for the ten groups ; but do not con- sider it necessary to give the separate numbers. There flowered : From the seeds of: Plants Reds Average 1. Red flowers 67 33 %} 2. Striped flowers 137 46% f * * * The result of all the experiments described above may be summarized in the following theses : 1. Antirrhinum inajiis Intcnm rubro-striatum (Plate I) is an inconstant race consisting of striped and of red flowered plants. 2. The striping of the commercial race varies con- tinuously, but the continuity does not include the red ones; these are separated by a gulf from the striped (Fig. 22). _ 3. The intensity of inheritance of the finely striped plants is about 95-98%. They pass into the red type either when propagated by seeds or by buds. 4. In the same way the broadly striped individuals produce many more reds ; the mean of three experiments (11-36-42) was about 30%. 5. The red plants resemble the wild ancestral form externally but are not constant as this is. The intensity of inheritance of their character is only about 70-85% : and the remainder of their offspring revert to the striped Antirrhinum Ma jus Striatum. 135 type. I have not yet observed this to happen by means of bud-variation. 6. Antirrhinum ma/its lutciim does not arise from these striped and red races. 7. If we compare the forms which we have been considering, 1 with the half races and middle races which we distinguished in 3, p. 18, we find that between the two constant elementary species (the systematic species, A. iiiajus and the systematic variety, A. majns luteum) there exist two intermediate forms which are perfectly distinct from these two, but not from one another. We can distinguish, o. The eversporting variety, A. ma jus luteum striatum, with striped flowers and a high degree of fluctuating variability, from which a faintly striped and a broadly striped race can be raised by selection. These three races however merge continuously into one another. b. The atavistic type in this race is uniformly red, but with incomplete inheritance and gives rise, when self- fertilized, in each generation to about 25% striped indi- viduals besides the red ones. In contrast with the previously described cases, the transition from the atavistic type to the eversporting variety and the reverse process here occur every year but always with a slight gap. The red type arises from the striped race by seeds and by buds, but the striped race has, hitherto, arisen from the atavistic type only by 1 The mode of inheritance in the coarsely striped individuals will have to be more closely investigated ; so also must sectorial varia- tion. Moreover the experiment should be repeated with other striped varieties, and the spotted forms investigated to see if they behave in the same way. But it is most important that pure cultures of the various types should be made by breeding for several generations. For this purpose the tall varieties should be chosen preferably, since they promise a much better harvest than the half-dwarf ones which I employed in my experiment. 136 Atavism. seeds. The transition from the red to the striped oscil- lates round 25%, the transition from the striped to the red is largely dependent on the degree of striping, which points to the existence of factors as yet incompletely understood. It may perhaps be mentioned here in anticipation that the varieties of Hesperis and Clarkia (15 and 16) with striped flowers behave in the same way , whilst both in Plantago ( 17) and Linaria vulgar is peloria (20) the eversporting variety is inconstant and reverts more or less easily to the atavistic type. 15. HESPERIS MATRONALIS. The flowers of the dame's violet are violet as the name indicates. There are three varieties on the market : a white flowered, a double, and a dwarf variety, all of which are con- stant so far as I know. A Forma lUacina and a "mixed" sort are of- fered in the catalogues. The plants are perennial; if the seed is sown in the spring, the majority of the plants will not flower until the following year; but if the seed is sown as soon as it ripens, or is allowed to fall on the ground instead of being harvested, the plants generally flower the next year. I have employed both of these methods at different times. I obtained my seeds in 1890 from a mixed group of white and violet flowered plants which were growing in our Botanical Garden. I grew them for two genera- Fig. 24. Hesperis matronalis. A flow- er of the pale finely striped form, with half of one of its petals dark violet. Hcsperis matronalis. 137 tions and found that the "white" were not pure white but pale lilac. Then I kept only plants of this variety through the winter, and first examined them in 1894, when they were in full flower. They flowered in isola- tion and partly pollinated themselves with their own pollen, partly were fertilized by insects. In later years also I have not enclosed this species in bags but have either grown them in an isolated position and left them to be pollinated by insects, or have had them flowering in a little greenhouse entirely built of fine metal gauze, where they fertilized themselves. My object was to test the degree of inheritance of the pale, the lilac, and the violet types separately. I shall first give a summary of my experiment. In this table, W denotes whitish, L lilac, and V violet (that is, the color of the wild species). The numbers in each case are per- centages of the particular culture ; where the culture was too small I have omitted the numbers. HESPERIS MATRONALIS. (WHITISH, LILAC, AND VIOLET IN PERCENTAGES.) 1900, 1899 annual and biennial 1898 annual 1897 annual 38 W. 30 L. 32 V. 50 W. 28 L. 22 V. ^ i i i 92 U W. 6 L. 2 V. j V. 7 ' . 1895 annual and biennial 1894 29 v W. 57 L. 14 V I ^ Before I come to the description of this experiment it is necessary to give some more details as to the varia- bility of the color of the flower. 138 Atavism. Plants with pure white flowers such as those belong- ing to the variety Alba did not occur in my cultures. I have compared the Alba and also the Alba plena directly with my plants. Certainly the difference is sometimes very slight, especially as the petals of Alba acquire a pale lilac color when they fade. They are all gradations between the whitest examples and those with the full lilac color; the variability in this case is perfectly con- tinuous. But between the lilacs and the violets there is always a gap; the darkest lilacs seem to be about half as dark as the violets ; intermediate stages do not occur. The vast majority of the plants have all their petals of the same color, but mixed conditions also occur. As in other cases there are striped flowers, sectorial and bud- variations. Examples of these three groups appeared in various years in my cultures but only sparingly. On the striped petals the stripes were fine and rare, but they ex- hibited the dark violet hue of the original species. The instances of sectorial variation have so far been occa- sional dark flowers on pale clusters, and on the other hand flowers of which one-half of a petal was whitish and the other violet (Fig. 24). Bud-/variations occurred on plants with very pale flowers, especially when they were richly branched and flowered on into the autumn. They were always stray twigs on the lower part of the main stems ; their flowers were all of the normal violet color. But so far I have not been able to obtain seed from them. A glance at a large bed reveals the general distribu- tion of color. At once the pale flowers are seen to be in the majority, whereas the whitish on the one hand and the lilac on the other are obviously rarer. The violet stand out conspicuously because they are not connected with the rest by any gradations. Except for this the varia- is Matronalis. 139 tion is so continuous that it is almost impossible to ex- press it in numbers. I have tried to arrange the plants in groups and to count the numbers of each group. And I give the numbers obtained in this way, only with the object of conveying to the reader the general impression which a bed makes on the observer, for it is inevitable that the limits between the groups should be somewhat arbitrary. Nevertheless I trust that I have succeeded in keeping fairly well the same limits between the groups during the successive years of my experiment, and this is the most important point. For the purposes of these color valuations I picked a flowering cluster, if possible the terminal one, from each of the plants on a bed, brought them to my house and sorted them there. I made out the following more or less clearly defined groups : W. Whitish, always without stripes. W\. Almost white; buds and withering petals almost white. 77 7 2. White suffused with lilac, not darker when withering. JI 7 s- Very pale lilac; buds lilac; only slightly darker when withered. L. Lilac, sometimes striped or spotted. LI. Definitely lilac, although pale; darker than Ws. Lo. Lilac; half as dark as V. V. Violet, the color of the typical species. I shall now give the composition of the culture of 1898 which was raised from the seeds of plants with whitish flowers. On July 14, I sorted 250 individuals by the method described and found : 140 Atavism. HESPERIS FROM THE SEED OF WHITISH FLOWERED PLANTS. W 5% IV 2 57% \ 92% W. W^ 30% } LI 4% 2% I 6% L. V 2% 2% V. I determined the composition of the cultures of the next year, 1 899, in the same way ; they were both raised from the seeds of lilac plants. One of them (5th gen- eration) flowered partly in 1899 and partly in 1900; but the other only in 1899. The result was as follows: HESPERIS FROM THE SEEDS OF LILAC FLOWERED PLANTS. Color 1st Experiment (5th Gen.) 2d Experiment (3d Gen.) W^ 3% \ 4% \ W 2 15% 38% W 22% - 50% W. W 3 20% > 24% J L L \ II \ . L 2 ll \ - V 32% = 32% V 22% = 22% V. The first list is based on 155 flowering plants, and the second on 219. The seeds of the whitish Hespcris, therefore, in this experiment, produce their like with a small percentage of lilacs and violets. The seeds of lilacs, on the other hand, give rise to the three types in about equal numbers, though it must be remembered that the limit between /T.3 and LI is to a certain extent arbitrary. I have not yet made a sufficient number of observa- tions on the inheritance of the violet color in this race. In the only experiment which I have carried out, only five plants flowered and they had the same color as their parents. Let us now pass on to a detailed description of the Hcspcris Matronalis. 141 experiment. It began in 1894 with seven plants which had already flowered in 1893 and had been noted as lilac flowered. Many of their flowers were more or less striped, some of them produced in August the violet bud- variations mentioned above, when the rest of the flowers had been through blooming for a long time. Seed was only saved from the lilac flowered branches; a part of it was sowed in August, the rest as soon as it was ripe. Most of it germinated in the following February and March; more than half of these plants produced stems and flowered in August. I obtained altogether 234 plants in flower of which 29% were pale, 57% were lilac and 14% normal violet. I selected the strongest plants from among the most typical of each group and transplanted them in the autumn to three as isolated spots as I could find in my garden. Here they grew freely, branched abundantly and flowered in the following year (1895) for a second time. There were three violet plants which however set very little seed. This was sown and the offspring flow- ered in the summer of 1897 in a conservatory. I took precautions to prevent their being visited by insects in order to render impossible the transference of their pollen to the other plants. As soon as the color of the flowers could be determined with certainty for any plant, this was pulled up. There were, as I have already stated, only five plants and their flo\vers were violet. I did not allow the lilac flowered plants to flower in this year but kept them for the next. Of the plants with pale flowers w r hich had been planted out separately in the autumn of 1895, only one plant flowered in 1896. Its seeds \vere sown immediately and gave rise to 12 plants which flowered in the summer of 1897; they were all 142 Atavism. pale with no more than the faintest indication of the lilac color. The seeds were sown in pans in the autumn, the seedlings were pricked out in November and planted out in April 1898 on a large bed. In June 250 individ- uals flowered, and the percentage composition of the color, as given above, was determined. Then the four lilac individuals falling into the group Lo were taken up and transplanted with all possible care to the metal gauze greenhouse. Before doing so all open flowers and young fruits were of course removed. It may be noted that in this experiment the lilac flowered individ- uals began to flower conspicuously later than the pale and violet ones. The seeds of these four plants were sown partly in October and partly in November, separately for each parent. Only one of the four resultant groups flowered in the following year (1899) ; the rest remained in the rosette stage and flowered in 1900. The proportions of the various colors were very much the same in the four groups. I recorded them separately but did not find any significant differences. The numbers in the first column (1st Experiment, 5th Gen.) on the table on page 140 give the composition of the whole culture. I transplanted some lilac plants of the first crop (1895), but only kept one of them which caught my eye with its beautifully striped flowers. It grew up into a sturdy plant, flowered in 1898 in an isolated spot and set an abundance of seed. From this 219 flowering o plants were raised in 1899, and their colors are recorded in the last column of the table on page 140 (2d Experi- ment, 3d Gen.). If we consider the results of these experiments, ex- Hcspcris Matronalis. 143 tending over seven years, in their relation to other known facts we find that we can distinguish the following races : 1. Hcspcris matronalis alba, the constant commercial variety. 2a. A whitish, pale lilac, seldom or never striped sort (Wi-Ws), which can reproduce the violet color by sectorial, bud- and seed-variation ; violet seed-variation about 2 r /c ; lilac offspring about 6 C ' ( . 2b. A lilac, often striped or spotted, race which gives rise to an inconstant but mostly considerable number of / whitish and violet offspring. Its color merges contin- uously into that of No. 2a. but is sharply separated from No. 3. 3. A violet variety which has arisen from 2a and 2b j and is presumably inconstant, on the analogy of Antirrhi- num majus. 4. Hcspcris matronalis, the original, constant, violet species. The analogy with the corresponding races of An- tirrhinum majus seems to me to be obvious and can be expressed as follows : 1. The systematic variety which is perfectly con- f -1 v stant (H. in. alba, A. maj. Intcnm). 2. The eversporting variety with lilac or striped flowers (H. m. lilacina, A. maj. lutcnm striatum). It can be split by selection in a plus and in a minus direction ; into a pale lilac, or finely striped race on the one hand, and on the other into the dark lilac and frequently striped dame's violet and the broadly striped snapdragon. 3. The self-colored but inconstant atavistic type which has the color but not the constancy of the original species. 4. The original violet, or red, perfectly constant spe- cies (Hcspcris matronalis. Antirrhinum majus). 144 16. CLARKIA PULCHELLA. A white variety of this pretty red species is offered by seedsmen. 1 Besides this a striped race sometimes oc- curs which has more or less numerous red bands of vary- ing breadth on the petals. 2 The red in these cases has the same intensity as that of the species. Moreover the white flowers are not pure white ; a very delicate but distinctly visible red flush can be seen on any bed of them in full flower. Sometimes occasional plants or individual flowers are somewhat richer in pigment, so that it is at once obvious that they are not pure white. I have only made an in- complete series of experiments with this plant because it does not lend itself easily to artificial fertilization and, as a rule, does not stand transplanting while in flower. But the results ob- tained suffice to demonstrate their essential correspondence with those obtained with Antirrhinum and Hes peris. We can distinguish in this as in the other two cases between a pale race poor in stripes and a richly striped one ; moreover these two races possess the characters of the corresponding ones in the two species named. But in Clarkia the broad stripes appear chiefly as sectors, as There is also a variety, Carnea, which is constant so far as my experience goes. 2 See p. 119. It was referred to by VILMORIN and by B. VERLOT, Production et fixation dcs varictes, 1865, p. 64. Fig. 25. Clarkia pulchella. A white flower of which one petal and a half are dark red, while there are dark red stripes here and there on the other two petals. Clarkia I'ulchclla. 145 for instance, whole or half petals; I shall therefore call such flowers and plants sectorial. In 1896 I had a bed of about 50 plants all of the flowers of which were whitish. The majority bore no red stripes, or only such fine ones and so rarely that they were oyerlooked, which is always possible since the plants produce very many and rapidly fading flowers. Only one plant stood out amongst the rest ; at the end of July it bore a flower with two red petals and at the beginning of August a petal the middle third of which was also colored red. Otherwise, the bed was practically white throughout the summer. Some of the seed of the whites was saved. From the seeds of a white flowered specimen I ob- tained in 1897 a culture of about 100 plants. Amongst these again there was only one sectorial example ; I saved its seeds separately although it had been fertilized by insects in the midst of the others. In the majority of these others I had not seen red stripes, but on a few of them there had been some insignificant ones. The seeds of the pale flowered plants gave rise to a generation equally poor in stripes ; in 1898 I only saw one striped one amongst 30. This race therefore remained poor in red sectors as a result of a continued selection of almost white plants. From the seeds of the sectorial plant I at once ob- tained a race which was rich in red petals and red sec- tions of petals, and often produced whole red flowers and twigs with red flowers only (Bud-variation). I grew it for two generations (1898 and 1899). The seeds for the first were gathered in 1897 from a seed-parent which had not been isolated; in 1898, however, I pulled up all of the non-sectorial plants whilst they were in 146 flower and on the remaining seed-parents only harvested the seeds from those flowers which opened after tHat operation. The single sectorial plant of 1897 bore one flower with one, and another with two red petals. Their seeds were harvested separately and sown. The other flowers were pale; I also harvested their seed separately. The first named seeds, naturally few in number, gave rise in 1898 to about 40 plants which flowered; the latter to 200. In both groups the red stripes and sectors were remarkably numerous in comparison with the previous year. At the end of July I found amongst the former about 25%, and amongst the latter 23^ sectorial plants. Besides these, a plant bearing red flowers exclusively, occurred in the former group. If I had repeated these observations from time to time the two percentages would of course have been considerably increased. But in order to isolate the sectorial plants I pulled up all those which up to that time had exhibited only few and narrow stripes. As already mentioned, I harvested seed only from the fruits of those flowers which had opened after this operation. I saved two kinds of seed : one was from a number of sectorial flowers which I had marked on a large group of individuals ; the other was from a par- ticularly striking plant which I had also marked, and which had a fair number of sectorial and occasional per- fectly red flowers, exhibiting also red bud-variations on its lower branches. I harvested seeds only from the narrow striped flowers of this plant. I have one more case of sectorial variation to men- tion before I proceed to give the results obtained from this harvest. A green lateral branch in an inflorescence on an otherwise white or finely striped plant had a Clarkia Pnlchclki. 147 narrow red longitudinal line on it which was not much broader than a flower stalk and extended over four inter- nodes. The upper, lower, and middle flowers of the tract stood on this line ; the two former were completely red, the middle one only partly so. The two flowers oc- cupying intermediate positions but on the green side of the branch were almost white. The culture of 1899 was richer in sectorial plants than that of 1898, as the isolation of the seed-parents would have led us to expect. From the mixed seeds referred to above, I had about 300 plants of which five were wholly red whilst the proportion of sectorial ones was 40%. The single selected seed-parent, however, gave rise to only 50 offspring which flowered, of which one was red, whilst the proportion of sector ials mounted to 70% . The average number of reds in the two cultures was 1-2% ; and that of sectorial plants 45%. These experiments show that the pale flowered plants, selected as seed-parents, give rise to a fairly constant progeny amongst which the proportion of sectorial plants is quite small. The progeny of sectorial plants, on the other hand, consists of about 45% broadly striped and 1-2% red plants, the remainder being pale tinged with red, or at any rate very poor in stripes. The cultures of the pale flowered plants are ordinarily in flower some weeks before the first stripes appear; but in the beds of sectorial plants the red may be seen among the very first flowers. Here also the white flowered ones are always in a large majority ; among a thousand flowers of this race I counted 34 striped and 8 sectorial ones, that is to say only 4% altogether. 148 Atavism. 17. PLANTAGO LANCEOLATA RAMOSA. Plantago lanccolata is one of of those plants which are remarkably rich in anomalies. PENZIG mentions a considerable number of them such as leafy stalks, ears the tops of which bear tufts of foliage leaves, 1 forked spikes with two or more tips, torsions etc. These and many other malformations such as split leaves, pitchers consisting of one or more leaves, occur commonly in this o - neighborhood and also in my cultures. It is worth men- tioning that all or nearly all of these abnormalities can occur in the same race, and sometimes indeed in a single stout individual. Evidently every plant must contain a number of latent or semi-latent characters which lie out- side its proper range of form ; these characters consti- tute, as I have already said, the outer range of the forms of the species (p. 27). A form also frequently mentioned 2 in the literature of the subject is one with branched ears (Plantago lan- ccolata rainosa). 3 In this variety sessile secondary spikes are produced in the axils of the bracts at the base of the main ear. They are often small, but sometimes nearly attain the size of the central ear. Their number is highly variable. Under good conditions of cultivation each head may have from 2-7 lateral ears, but on single ears the number may rise to 20 and more (Figs. 26, 27). I have been carrying out experiments on the inheri- tance of this ramosa-character since 1887. It proved to I 1 have often picked these tufts and made cuttings of them ; they take most quickly and grow to strong rosettes of radical leaves, the ears arising from which may repeat the phenomenon of the tufting to a certain extent (Plantago lanceolata coronata). ~ PENZIG, Teratologie, II, p. 252. "Kruidkundig Jaarboek, Gent, 1897, pp. 76 and 91. Plantayo Lanceolata Ranwsa. 149 be only partial. In spite of the most careful selection and isolation during the time of flowering this race every year produces plants not one of whose spikes, even when there are a hundred to the plant, exhibits the smallest trace of branching. They are obviously to be regarded as atavists. The proportion in which these atavists occur seems to be fairly constant, fluctuating however from year to year. It can be slightly increased or diminished by the choice of favorable or unfavorable seed-parents; but it does not seem possible to effect an essential and per- manent improvement by continued selection, at least not to a degree that would open a chance of altogether elim- inating the atavism. In the first years of my cultures I did not pay partic- ular attention to this phenomenon ;" moreover my experi- ments were on too small a scale to afford numerical data of any value. But I found atavists as well as ramosa- plants every year, although I always collected my seeds from the former. I did not determine the proportion until the fifth generation (1892) was reached. I should state that I have isolated my seed-parents every year, cutting off as many as possible of their unbranched ears before they flowered. Pollination which had to be left to the wind was therefore confined to the group of se- lected seed-parents, whose number scarcely ever ex- ceeded 10. It was as pure as it was possible to have it. I obtained the following figures : GENERATION PERCENTAGE OF ATAVISTS 5. 1892 46% 6. 1894 50 %, 58%, 59% 7. 1897 47% 8. 1898 45%, 56%, 59% 8. 1900 52% 150 Atavism. Plantago lanccolata rainosa, therefore, produces a pro- portion of about one-half atavistic individuals every year. The variability in the fig- ures given is at least in part dependent on external in- fluences (nutrition and selec- tion). Closer examination of the individual years proves the truth of this. In 1892 I had 48 plants in flower ; nine of these plants produced split leaves and pitchers at the time when they were being transplanted, about three weeks after the seed had been sown, and seemed espe- cially desirable on this ac- count. In the summer they turned out to be all rainosa- plants with richly branched spikes. They were cultivated the following year also ; and the sixth generation was raised partly from their seed and partly from the seed har- vested in 1892 from two other seed-parents. From the latter were raised 103 plants which flowered, of Fig. 26. Plantago lanceolata which 50 c/ c , atavists, ramosa. A whole plant. this proportion being nearly the same for the two seed-parents. In order to investi- gate this, the seeds of the individual seed-parents were, Plantayo Lanccolata Ramosa. 151 as usual, sown separately. The higher figures 58 % and 59% were derived from the offspring of a plant which had been divided into two in 1893, after which one-half of it was grown on sand and the other on ordinary garden soil. I shall have to revert to the effect of this treatment on the plant itself; but it will be observed that the differential treatment had no visible effect on the offspring of the two halves. (The numbers of indi- viduals dealt with in the two cultures were 57 and 60 respectively. ) The seeds of the typical individuals of my race of 1894 I sowed in 1897 under normal conditions, as usual (seed sown in the greenhouse; seedlings pricked out into pots, and afterwards transplanted to the beds). The seed had been saved from two plants with richly branched ears. It produced a culture of 70 plants which flowered and contained 47% atavists. Whilst flowering was pro- ceeding I transplanted all the ramosa-plants whose ears were only slightly branched, and marked among the re- mainder a specimen which seemed to be the most pro- fusely branched. I harvested seed from those flowers only which protruded their stigma after this separation had taken place and after the atavists had been weeded out. Seed was harvested separately from each plant. In the following summer (1898, 8th generation) it was seen that the seed of the best seed-parent had only pro- duced 45% atavists (among 100 plants that flowered). The seeds of the average seed-parents gave 56%, and those of the worst, 59%. Selection had therefore a dis- tinct, although not a very great effect. It should be re- marked that the number of average seed-parents was 8, and that of the worst ones 10. The composition of the progeny was determined separately for each seed-parent, 152 Atavism. but the differences were not greater than the extent of the experiment would lead us to expect. There were 1033 offspring from the average seed-parents and 732 from the ten worst plants. The two separate cultures which deviated most from the mean contained 37 % and 65% atavists respectively. The value of 52% given Fig. 27. Plantago lanceolata ramosa. A, B, C, three branched ears. above for the same generation but grown in 1900, will be dealt with below. Bud-variations occur in this as in the inconstant races of other species, although very rarely. In such cases it is one or several lateral rosettes which varv. The struc- ff ture of our plant is a very simple one. The stem of the Plantayo Lanceolata Ramosa. 153 seedling grows out into a short, somewhat oblique, rhi- zom which produces a rosette of radical leaves. Ears are formed in the axils of the higher leaves but rosettes of the second order groxv out from the axils of the lower ones. In the second summer the primary and secondary rosettes behave in the same way, again producing ears above and secondary rosettes below. If the plant grows very robustly it may consist of as many as 10-20 single rosettes ; if it is a raiuosa every rosette produces branched ears, at least on some stalks. Sometimes all the ears of the whole plant are branched, in which case it is per- fectly easy to see that there is no bud-variation. In its second year a single plant may often produce more than 50 branched ears. The culture of 1897 contained a plant which exhibited a bud-variation. The seeds of its branched ears, har- vested in the first year, had produced 89 individuals that flo\vered, of which 36 (40%) were atavists. The plant in question consisted, in the autumn of its second year, of more than 25 single rosettes which were carefully isolated, and planted separately. Only the seven strong- est ones survived this operation. I kept them all in their pots until a sufficient number of ears were visible and then planted them out on two distant beds. On the one I planted four rosettes with unbranched ears, on the other, three with branched ears. The four former pro- duced, together, over 200 strong ears, all unbranched with the exception of a single one which bore a small lateral branch at its base. The three latter formed both unbranched and more or less richly branched inflores- cences, but during the whole summer the unbranched ears were all cut off before they flowered. The harvest from the two beds, gathered and sown separately, gave 154 Atai'isjn. rise to two cultures in 1900. They had the following composition : Ears of seed-parent Extent of culture Atavists With branched ears Branched 44 individuals 52 % 48 % Unbranched 206 individuals 92 % 8 % The rosettes with branched ears gave rise to rather more atavists than the seed of the branched inflorescences of the same plant in the first year (52% as against 40% ) , which was probably due to the fact that it had a less sunny position in 1899 than in 1897. But the rosettes with unbranched ears, although they were in a good po- sition in 1899 and grew very healthily, gave a progeny dissimilar to that hitherto produced by any of the branched plants of this race (see Table on page 149 which gives the results of more than 25 individual sowings from separate seed-parents). The four lateral rosettes with unbranched ears, there- fore, formed in this case a clear instance of bud-variation, producing a race poor in branched ears. The question of the constancy of the atavists in my race is a point of considerable interest. Hitherto I have found them completely constant. With a view to test- ing this I did not weed out the atavists in the fifth gen- eration in 1894, but simply cut off all their ears before the branched plants flowered, and repeated this opera- tion from time to time when new ears appeared before they could protrude their stamens. After the harvest I weeded out all the branched individuals ; most of the atavists survived the winter and flowered luxuriantly J in 1895 in isolation. The majority of them produced over one hundred ears per individual. I harvested the seeds separately for each seed-parent. The sowings took place in 1896 and in 1897. They Plantago Lanccolata Ramosa. 155 gave rise respectively to three and six cultures derived from the nine seed-parents. Each culture consisted of from 35 to 100 plants, making together 600 flowering individuals bearing 4000 inflorescences. These were un- branched without exception. The question suggests itself, whether the seed-atavists and the bud-atavists belong to the same type. On the one hand it is possible that the constancy of the former is not always so absolute as it appeared in my experiment. On the other hand, branched bud-variants might occa- sionally appear in the race derived from the atavistic bud-variants, and such might have been the cause of the occurrence of branched individuals (8%) in my culture of 1900. But further investigations are necessary to provide a satisfactory answer to this question. Plantago lanccolata ramosa, therefore, gives rise to atavistic individuals, cither by seed (about 50%) or by buds (very seldom) which are either absolutely, or at least in a high degree, constant from seed. It still remains to describe briefly the fluctuating vari- ability of our race of plantains. This is considerable, and conforms to the common laws; especially is it de- pendent to a large extent on external conditions and, within certain limits, capable of being altered by selec- tion. The observations, which I now shall give, refer to true ramosa-plants, and not to atavists and bud-varia- tions. The variability of this race corresponds with that of other monstrous races inasmuch as the curve describ- ing it is dimorphic. 1 During July and August 1893 I 1 Sur Ics courbes galtoniennes dc.: monstruosites, Bull. Scientif. de la France et de la Belgiqne, public par A. GIARD, XXVII, 1896, P- 397- 156 Atavism. picked all the ears of a small group of plants, and ob- tained the following figures: Ears without any branching . . . 191 Ears with one lateral ear 80 Ears with two lateral ears .... 136 Ears with three lateral ears .... 93 Ears with four lateral ears .... 33 Ears with five lateral ears .... 12 Sum of ears 545 The degree of branching in this group was fairly low; nevertheless the apex of the curve of the atavistic ears is distinct from that of the branched ones. This phenomenon could indeed be easily observed, even with- out any counting, on account of the relative scarcity of heads bearing a single lateral ear, a fact which I have also observed repeatedly since. This is a character of the eversporting variety and suggests the possibility that the one-branched ears which are so common in nature (where the rauwsa-form, as is well known, is not at all rare) presumably constitute the half race; but I have not investigated this point. The number of compound ears per plant, and the degree of branching in each, are to a great extent de- pendent on the conditions of life. The stronger the growth of the whole plant, and the richer the foliage, the more pronounced will the anomaly be. Therefore, a more profuse branching of the individual ears usually goes hand in hand with a richness of branched inflores- cences. The branching also manifests a certain periodic- ity. The young plants almost always begin with un- branched ears; it is not until later that the monstrosity appears, gradually increasing in strength. Then towards the end of the summer I often observed a diminution in the amount of branching and often the formation of Plantago Lanccolata Rcunosa. 157 more numerous unbranched ears. In the second summer often almost all the ears on healthy individuals are branched even when their number reaches 50-60 per plant. In the first year I found that as a rule there were 10-20 branched ears, and sometimes as many as 30 or even more occurred on each plant. In fact we may assume that, on the average, and with ordinary methods of cultivation, about one-third of the ears will be branched during the first summer; for instance, in 1898 I found amongst 439 ears on 30 individuals 136 or 31% which were branched. It goes without saying that the atavistic individuals were excluded from these countings. I have also made some direct experiments to deter- mine the influence of individual vigor on the develop- ment of the anomaly. In the first place I have grown very weak plants and have then got them to grow stronger gradually. For this purpose I made use of the plantain's well-known property of producing buds from its roots. As the roots are all very thin, the plantlets obtained in this way are very weak at first, nor do they grow up as quickly as seedlings. For the purpose of this experiment I selected (March 1893) ten plants which had had 10-25 branched ears each in the previous year. I pulled them out of the ground, cut off the mass of their roots and planted these, throwing away the rosettes and any leaf-buds that might be present. I put the roots of each individual straight into the ground without separating them. Radical buds were produced in hundreds, often so many from one bundle of roots that there was not room for all of them to develop. In the middle of June, that is, after about three months, they began to flower. At first there were only 40% branched ears, with only one or two lateral 158 Atavism. ears (on the 46 first flowerstalks). In the next 100 the proportion mounted to 60%, and 3-4 partite inflores- cences also occurred. Later on, about the middle of July, the first stalks with five lateral ears appeared, and the number of branched ears gradually increased to 70%, and in August the strongest rosette had 67 ears of which 52 were branched, i. e., about 78%. A question at one time much discussed was whether adventitious buds had the power to reproduce the varia- tions and anomalies of the parent plant. At that time malformations were not regarded as heritable, but since the inheritance of monstrosities has become generally recognized, 1 it must be considered evident that adven- o titious buds will behave like normal ones; and the only question that can arise is whether they are more liable to produce bud-variations or not. If they are weak the abnormal character will be less pronounced; but if their strength is equal to that of ordinary buds the abnormal character must be developed to the same extent. It is therefore almost superfluous to lay much stress on the reproduction of the branched ears from the radical shoots of our plantain. The rest of my experiments deal with divided plants. In the spring of 1893 I selected for this purpose two fine rosettes that had survived the winter and which had proved to be particularly rich in branched ears in the previous year. Both plants were divided as equally as possible into halves. Of the first plants one-half was planted in sand and of the other one-half was put in the shadow of a tree, the control halves of both plants being cultivated under ordinary conditions for the purpose of l Erfelyke Monstrositeiten, Kruidkundig Jaarboek, Gent, 1897, p. 62. Plantago Lanccolata Raniosa. 159 comparison. At the beginning of the period of flowering no difference was discernible in either experiment be- tween the two halves, but it gradually became visible during the course of the summer. I picked off all the ears from the culture in sand at the end of July and at the end of August; here is a record of them: Number of lateral ears per primary ear Totals 012345 July 28th -! Sand 3 3 4 6 J i 20 < Control 9 7 9 6 31 A , ( Sand 14 10 12 8 3 1 48 '( Control 12 2 10 7 6 2 39 The difference though slight is distinct. It is more clearly brought out if the mean number of lateral ears j o per primary ear is calculated. In August in the plants on sand this was 1.5, in the control half 2. A similar effect was produced by shade which exerted a most deleterious effect on the whole growth of my experimental plants as will be seen from the small number of ears produced. I obtained the following figures in the same way as in the previous experiment. Number of lateral ears per primary ear Totals 0123456 Tnlv 28th * Shade 7627520 29 [uly M Control 1 1 2 8 19 20 1 52 . 1 \ Shade 15 1 1 2 19 ( Control 21 9 20 16 10 3 79 The mean number of subsidiary ears per primary ear in August in the shadow half was 0.5 and in the control half 2.0. In conclusion, the results of the whole series of ex- periments which has lasted over more than ten years may be summarized as follows : The Plantago lanccolata rainosa of ;//v experiment constitutes an "inconstant" 160 Atavism. middle race or eversporting variety ; that is to say, a race which produces in every generation a fairly constant proportion of atavists. This proportion is about 50 P- 1080. This tree stood in the garden of M. SALADIN DE BUDE near Geneva. Many cuttings made from the double-flowered branch have been distrib- uted. 202 Observation of the Origin of Varieties. MEISTER wrote the following words at the end of his account of pelorias. 1 "One of the most remarkable features of the varia- tions of plants is, without question, the sharpness and suddenness of the origin of profound deviations from the normal form of structures such as we see it in the phe- nomena just considered, in many analogous cases, and especially in the formation of monstrosities. The new form does not come into existence bv the gradual sum- *> O mation of small deviations in one direction, during suc- ceeding generations; it appears all at once, perfectly distinct from the original form." This highly important and perfectly correct state- ment rests even now simply on the absence of transi- tional forms, and does not rest on direct observation. If the peloria had originated by a gradual process it would be reasonable to suppose that at least in some of the relatively numerous instances the intermediate steps would have been found ; but as this was not the case it was concluded that they did not exist and therefore that the origin of the variety had been immediate. 2 But it is hardly necessary to point out that nothing short of direct observation can furnish the final proof. Direct observation will moreover inaugurate a new stage in the study of this remarkable phenomenon, by making 1 W. HOFMEISTER, Allgcmc'me Morphologic dcr Gcivachsc, 1868, p. 564. 2 On the pelorias of Linaria, especially of L. spuria, see H. VOCHTING, Ueber Bluthcnanomalien, Jahrb. fur wiss. Botan., Vol. XXXI, No. 3, 1893, and L. Josx, Bliithenanomalien bei Linaria spuria, Biolog. Centralblatt, Vol. XIX, 1899, p. 145. Also J. H. WAKKER, Over pelorien, Ned. Kruidk. Archief, Vol. V, p. i, July 1889, with Plate X. P. VUILLEMIN, Monstruosites chez le Linaria vulgar is, Bull. Soc. Sc., Nancy, Dec. 1893, with one plate (Vol. XIII, 1894, P- 33)- W. and A BATESON, On Variations in the Floral Symmetry, Journ. Linn. Soc. Bot., Vol. 28, 1871, p. 381. The Origin of Linaria J^ul (juris Pcloria. 203 accessible to investigation the mode of its appearance and the external causes to which it is due. For these reasons I have endeavored to induce the occurrence of the Pcloria from the ordinary form in my experimental garden. It is obvious that the success of such an experiment, at least at first, is dependent on chance. This chance however can be favored by making the cultures as extensive as possible, and by widely vari- able conditions of life. Fortune has favored me, and after seven years' work my object has been attained. ThePeloria appeared quite suddenly in the fifth and sixth generation of my culture. The signification of my observations will be more properly understood if I premise my account of them with a short gen- eral and historical account of the subject, referring Fi s- 38. A B Linaria vulgaris. J . C, D, Pelonc flowers. the reader for the litera- ture to the following section (21) and to PENZIG'S TeratologieS Peloric flowers in Linaria I'ulgaris' 2 were first dis- covered, as is well known, in 1742 by ZIOBERG on an island near Upsala and described by LINN^US in the 1 O. PENZIG, PHanzciv-Teratologic, Vol. II, p. 195. ! The Pelorias have five spurs: Pcloria ncctaria. But there is also a Peloria ancctaria in which the flowers are regular but without spurs. See PENZIG, loc. cit., and VERLOT, Production des varictcs, p. 90. This variety is nearly sterile, setting very little seed, but it breeds true. 204 Observation of the Origin of Varieties. Amoenitates academicae. 1 The plant grew there together with the ordinary Linaria and formed a "constant" race Fig. 39. Linaria vulgaris pelorla. A richly branched stem of a plant of the second generation. Raised in 1898 from seed of the first generation of 1897 and photographed in August 1900. All flowers are peloric. through propagation by the buds on its roots. All the flowers of this plant were peloric (as in Fig. 39). LIN- 1 Amoen. acad., I, p. 55, p. 280 (1744). See MOQUIN-TANDON, Pfiansen-Teratologie, 1842, p. 170, and HOFMEISTER, loc. cit., p. 563. The Origin of Linaria Vulgaris Pcloria. 205 described this form, which was new then, under the name of Pcloria, derived from the Greek WA 6 h + 1 p 1895 1897 57 h -f 1 p h -f 1% p h + 1% / 1894 1896 1899 V H h h h We will begin the further account of the experiment with the parent plants (H) of the peloric race (1893). As I had not of course observed anything extraordinary up to that time I only sowed a little of its seed. This was done in pans in the greenhouse; the young plants were transferred into pots with manured soil until they were planted out in June. As a result of this treatment 212 Observation of the Origin of Varieties. they all flowered in the first year, 58 plants in all, of which 45 were clicotylous and 13 tricotylous. Amongst the dicotyls there were eleven plants, each of which bore one, two, or three peloric flowers, while in one case a peloric flower replaced a whole raceme. Amongst the tricotyls I did not find any such flowers, partly because the majority of these \vere removed by the middle of August; but there appeared amongst them one plant which bore peloric flowers exclusively on all of its stems and their branches. It bore no seed in spite of repeated careful pollination, partly with pollen from the neigh- boring plants ; it survived the winter and flowered freely in the following year, again producing exclusively peloric flowers. This experiment seemed to suggest that the Peloria arose from the hemipeloric parent in a proportion of about 1-2%. So in order to obtain closer knowledge of this proportion, I made a larger sowing in 1896 from the same lot of seed, and was able to plant out about 1850 seedlings in pots. By the middle of July some wholly peloric individuals had appeared, which were promptly taken up and transferred to a remote part of the garden. The further examples of Peloria which appeared from time to time, were planted beside them. By the middle of August all healthy plants were in flower and were recorded. There were altogether 16 totally peloric plants and 1759 with ordinary flowers, and here and there occasional peloric structures. This gives a total of 1775 plants which flowered, of which 1% (strictly speaking 0.9%) belonged to the new peloric variety. For the harvest the flowers of the best peloric plants were enclosed in parchment bags and each fertilized with The Origin of Linaria Vulgaris Pcloriu. 213 the pollen of another peloric plant. I also selected a beautiful hemipeloric plant which bore a profusion of flowers, one of the earlier of which was peloric. It set a quantity of seed after self-pollination. 1 I repeated the experiment in 1899 with the rest of the seed of the parent plant H, and obtained the same result, as was to be expected. I raised slightly over 300 flower- ing plants, of which 3 were wholly peloric; that is to say, a proportion of \% again. I observed on the rest a certain number of stray peloric flowers during the course of about two months. These three cultures constituted the fifth generation of my experiment. The sixth generation therefore could be raised from the seeds of the hemipeloric plants in it. I did this partly in 1895 from the plants of 1894, and partly in 1897 from those of 1896. The plants which bore the seed had flowered in bags and had been fertilized partly by their own pollen and partly by pollen which I had transferred from one seed-parent to the other. In both cases the mutation was repeated. Wholly peloric individuals again arose from hemipeloric ances- tors, in spite of the smallness of the crops occasioned by the poorness of the harvest. In 1895 I raised 17 flowering individuals from seeds of the dicotylous plants mentioned on page 559; two of them were wholly peloric, all their flowers being of this type. In 1897 I sowed the seed of the fine hemipeloric plant of 1896 referred to above, but obtained only 7 flowering individuals, one of which again, ho\vever, was wholly peloric. I come now to the consideration of the question as [ This frequently fails in Linaria vulgaris, but sometimes succeeds more or less completely on very vigorous plants. 214 Observation of tlic Origin of Varieties. to whether the mutants are immediately constant from seed. An almost insurmountable obstacle in the way of providing an answer to this question is the low fertility, or rather the almost complete sterility, of the peloric flowers. Practically no results can be obtained with self-pollination, and when artificially fertilized with one another's pollen the majority of the flowers set no seed. I have pollinated thousands of flowers in the course of several years, only to obtain a little over one hundred fertile seeds. Under these circumstances it is obviously difficult to avoid mistakes ; stray pollen grains may happen to reach the stigma from distant groups of normal plants, by the agency of insects, or in the operation of artificial pollination. 1 These circumstances evidently tend to in- validate the conclusion in cases in which the abnormality would seem to be incompletely inherited. Only three of the wholly peloric plants of 1896 set seed in that year. From this seed only 8 plants were raised ; five of them had one-spurred flowers and 3 were wholly peloric. I kept the peloric plants of 1896 through the winter, and took much trouble in 1897 in the attempt to fertilize their flowers. Every other day I pollinated all the open flowers with pollen from two other seed- parents. I obtained a very small quantity of seed most of which was empty (0.2 cc). About 100 seeds ger- minated, but some of the young plants were so weak that they soon died. 79 plants flowered most of which were very vigorous and branched freely ; 75 were wholly peloric, and 4 normal, the latter being removed as soon as possible. The former exhibited great variability in the structure of their flowers, but did not produce a single one-spurred corolla. During July and August they 1 Such crosses give normal one-spurred individuals. The Origin of Linaria Vulgaris Pcloria, 215 filled an entire bed of over 3 square meters in extent, with hundreds of vigorous spikes which bore exclusively peloric flowers. I again obtained only a very small harvest from this bed (0.3 cc) ; it was the result partly of artificial and partly of insect pollination, the plants flowering in suffi- cient isolation. Very few of the seeds germinated (1899) and only 32 plants flowered ; 28 of them were peloric but 4 were normal. The progeny of the peloric race was therefore a mixed one, in the three experiments which were continued over two generations. It consisted altogether of 3 -\- 75 -p- 28 = 106 peloric and 5 + 4-(-4 13 normal (including hemipeloric) individuals, a total of 119 with about atavists. WILLDENOW (see p. 206) also found the peloria character inherited, though incompletely. As already stated, however, insufficient isolation may have played some part in bringing about this result, but hardly to such a degree that we might infer from our experiments that the peloria comes true. If we now look back over this experiment, which occupied 13 years, its result may be summarized as fol- lows : 1. Linaria vulgaris hemipelona is a race with an in- herited semi-latent character, which manifests it- self from time to time among thousands of flow- ers, but seldom in more than one instance on a plant. It is widely distributed in the wild state. 2. From it the Linaria vulgaris peloria may arise but the conditions under which this happens are not yet understood. 3. This origin is a mutation ; it takes place suddenly, and without any visible preparation. Especially 216 Observation of the Origin of Varieties. in those individuals from the seeds of which the mutation arises the latent character is not more highly or more often developed than in the rest of the race. 4. The mutation is repeated in successive genera- tions. I observed it for two years, but did not follow it further. 5. The mutation occurred in about \ c /c of the indi- viduals. 6. The new character was exhibited by the mutants, in a full state of development, in all their flowers ; although it was subject to considerable fluctuating variability. 7. The mutants are to a large extent, perhaps even perfectly, constant from seed. The intensity of inheritance observed was about 90%, but it is probably more. * * * Let us next see how these results can be applied to the explanation of the occurrence of the Peloria in the free state. Wholly peloric plants have been found wild by numerous botanists and in the most diverse localities ; but, so far as the published information extends, always as rarities. They maintained themselves during a larger or shorter period of years by means of their radical buds, perhaps produced some scanty seed but could not spread nor reach new localities by this means. They must there- fore have originated in each case in the spot where they were found. I imagine that this origin is determined everywhere by the same general laws, and thence conclude that it occurs in the wild state in the same manner as in the particular case observed by me, i. e., from Linaria i'itl- The Origin of Linaria Vulgaris Pcloria. 217 garis hemipeloria, and always suddenly. The very gen- eral occurrence of this race and the fact that intermediate forms between it and the fully developed Pcloria have never been mentioned by botanists, give support to this hypothesis If this view is correct we have here a mutation which is not limited to a period but continues to appear from time to time during the course of the ages. Its appear- ance in every single case is independent of the others, at least so far as external conditions are concerned. In this sense it is polyphyletic. A point which favors this view is the fact that it is not a member of a definite group of mutations as are the subspecies of Draba vcrna, Viola tricolor and others. Linaria vulgaris, it is true, frequently gives rise to other kinds of variations such as the Pcloria ancctaria and the Catacorolla, both of which have occasionally appeared in my own cultures, but nothing is on record concerning the relation between these and the Pcloria ncctaria which I have studied. If we compare these results with those which we have described above for Antirrhinum inajits striatuin ( 14, p. 134), we see that Linaria I'nlg. hemipeloria is obviously a half race; and that L. vulg. peloria, whose partial con- stancy seems analogous to that of the striped snapdragon, may perhaps be regarded as parallel to this. These two races fluctuate so as to approach one another, so to speak, occasionally overstepping the common boundary either in single flowers (L. vulg. hemipeloria) or in whole plants (L. vulg. pel o Ha). ^ 5JJ JJC We now come to the most important point to which our results and conclusions lead us namely the com- 218 Observation of the Origin of Varieties. parison of this mutation with those of Ocnothera La- marckiana. The two processes have several features in common, but possess others which are more or less strongly opposed. The points of similarity are : the sudden and imme- diate origin, the repeated appearance, the mutation- coefficient of about 1% (see Vol I, Part II, 14, p. 337), the completeness of the new type, and its high degree of heritability. These common characters justify the description of the origin of Linaria mdgaris peloria as a mutation. 1 But it is a mutation of a special kind. The structural change does not extend to all parts of the plant, but is confined to the flowers; in their youth the two types cannot be distinguished. In the mutations of Oenothera Lamarckiana the new characters are analogous to the specific characters of related species already existing; in the case of Linaria no such analogy exists. On the contrary the new character in Linaria occurs as a variety in numerous other species, and even in distantly related 1 LINNAEUS, as is well known, expressed the view that the Peloria is a hybrid between the common Linaria vulgaris and some other un- known plant. Its comparative sterility favored this view, but as the second^of the two parents could not be found this view has since been given up. Here, however, I might discuss the possibility that L. vulg. hemipcloria might be a cross between L. vulgaris (apeloria) and L. vulg. peloria. If this were so the appearance of the latter from the former would perhaps have to be regarded not as a muta- tion, but as a segregative process in a hybrid race. If this view were true the Peloria should first have arisen from the Apeloria, without the mediation of the Hemipcloria, a process which has still to be observed. It is, however, no more than a pure assumption that the hybrid Apeloria 'X Peloria would be a Hemipcloria; in fact our knowledge of other cases would lead us to suppose that it would be like one of the parents, in this case the Apeloria, and so long as there is no direct information on any of these points a further dis- cussion of this view seems barren. Moreover it is by no means cer- tain that Linaria vulgaris apeloria exists at all, or ever has existed ; the variety, in this genus particularly, may well be older than the species. The Origin of Linaria Vulgaris Pcloria. 219 plants. Lastly the mutation in Linaria does not appeal- along with others in space and time, but occasionally, and scattered perhaps over the whole area of the parent form and probably over the whole period of the life of this race. The mutations of Ocnotlicra Lamarckiana necessi- tated the assumption of a definite premutation, but the origin of the Pcloria is obviously a phenomenon of a different kind. Pcloria is often regarded as an instance of atavism. 1 The correctness of this interpretation obviously depends primarily on whether this term is used in a narrow or a broad sense. Atavism is a reversion to ancestral char- acters ; in the narrow sense to the complete type of par- ticular ancestors, in the wider it refers only to single characters. But it is clear that the spurs which form a distinctive character of the genus Linaria must be older than the species L. vulgaris, which cannot therefore have had ancestors without the spur but with the other char- acters of the species ; so that L. vulgaris ancctaria can occupy no place in the series of ancestors. The sym- metry is ever so much older and L. vulgaris with regular flowers has certainly never existed amongst the ancestors of the common toadflax. Moreover the sterilitv of the peloric plants does not favor such a view. If the Pcloria must be regarded as atavistic, this view can mean no more than the assertion that it has arisen by the loss or latency of a character of the common Linaria. Therefore we are concerned here with a retro- gressive mutation, and the question arises, how far the differences between this case and the progressive muta- tions with which we have become familiar in Oenothera 1 See L. JOST, Biolog. CentralbL, 1899. p. 149. 220 Observation of the Origin of Varieties. are thus to be explained. The explanation is so simple that it follows directly from the preceding discussion. It is merely necessary to point out that the most impor- tant condition for a character to become latent is its pres- ence ; and this explains how it is possible that the Peloria so often appears over the whole area of distribution of the species. Neither a premutation nor a period of muta- tion is necessary for such an occurrence. If the loss or latency (for the inner potentiality is ob- viously not lost but only becomes inactive) affects single flowers we have a partial atavism, but if it affects the whole plant we have the complete and heritable Peloria. It is on this basis that the atavistic phenomena of the striped flowers, of the many-spiked Plantago (17, p. 148) and of the peloric Linaria fall in line. They are retrogressive phenomena, reversions to ancient charac- ters which have externally become lost but are still pres- ent in a latent state. Their agreement with one another on the one hand, and their contrast with the progressive mutations of Oenothera Lamarckiana, on the other, thus receive a satisfactory explanation. 21. HERITABLE PELORIAS. Pelorias are very rarely met with in nature as a specific character. As an instance I may quote Mentha aquatica, the apical flowers of which according to SCHIM- PER'S discovery are always regular and consequently peloric, 1 and the orchid Uropedium Lindenii, which is regarded as the peloric form of Cypripedmm caudatnin.- *A. BRAUN, Abh. d. Berliner Akad., 1859, p. 112; and DELPINO, Mem. R. Instit. di Sci., Bologna, 5 Ser., Vol. I, 1890, p. 269. 2 A. BROGNIART, Ann. Sc. nat., 3 Ser.l, Vol. XIII, p. 113 (Plate 2) ; and J. M. JANSE, Maandblad voor Natuurwetenschappen, Vol. XIV, No. 3, 1887, p. 29. Uropedium Lindenii appears to be by no Heritable Pelorias. 221 In a state of cultivation peloric races are also very rare, and the common Gloxinia superba erecta with its numer- ous color varieties and hybrids is the best generally known cultivated example. 1 Our present knowledge of the origin of Linaria vul- yaris pcloria as described in the foregoing pages, justifies us in attempting to form some idea concerning the origin of such forms in these perfectly analogous cases and also to sketch the details of this idea on a basis, or back- ground of facts. But there are still difficulties in the way. The low fertility and the incomplete constancy of the Peloria distinguish it from true species. 2 Most systematists would evidently not consider L. vulgaris pcloria to be a true species unless the common L. vulgaris were ex- tinct. Besides the examples named, there is a whole series of heritable cases of peloria, which either appear as rare anomalies, or are familiar cultivated races, and repeat the abnormality regularly and in a fairly large number of individuals every year. 3 In both cases, however, the development of the anomaly is, as usual, in a high degree dependent on external conditions. There are, as we have stated in the foregoing section means rare in Colombia (South America) ; it was discovered there by LINDEN in 1843 (LINDEN, Pescatorea, Iconographie des Orchi- dees, 1860, Plate II. l The spurless varieties of certain species of Viola and Tropae- olum may also be regarded as pelorias : see the following page. 2 From this point of view it would be very important to know whether the Mentha and Uropedium cited are perfectly constant, that is, never produce atavists without pelorias. 3 It is extremely doubtful whether, besides these, there are pelorias, the origin of which is solely due to external influences and does not need the existence of a corresponding internal poten- tiality. 222 Observation of the Origin of Variety. (20), various kinds of pelorias according as one or another form of the petals of the parent species has be- come the one which prevails in the subspecies. In spur- bearing species they are distinguished as Peloria nee- tana and anectaria. Both possess a very low degree of fertility but are, so far as is known, heritable. Peloric flowers without spurs are well known in Linaria, 1 An- tirrhinum, 2 Viola? Tropaeolum, 4 etc/' There are few heritable peloric races beyond those which have been named. The best known are Cor yd alls solida peloria which in GODRON'S experiments 6 was found to transmit the abnormality through a series of genera- tions, and Digitalis pnrpurca monstrosa (Fig. 42). This latter, the peloric foxglove, has been a favorite garden plant for a long time, and has often been the subject of morphological investigations. The oldest descriptions and figures are due to my predecessor G. VROLIK, whose preparations are still to be seen in the collection at Amster- dam. 7 Since his time the variety has been cultivated in our botanical garden more or less regularly, and is still growing there. 8 It is very constant; its peloric flowers *C. BILLOT, Annotations a la Flore de France et d'AUcmagnc, quoted in Bot. Zeitung, 1872, p. 278. 2 J. T. C. RATZEBURG, Animadversiones ad peloriarum indolcm, 1825, Plate I, Figs. 64-76. 3 J. C. COSTERUS, P clones du Viola tricolor, Archiv. Neerl., Vol. XXIV, p. 142, Table II ; DE CANDOLLE, Organo graphic, PI. 45. 4 E. VON FREYHOLD, Ueber Pelorienbildung bei Tropaeohim adun- cum, Botan. Zeitung, 1872, p. 725 and Plate IX. 5 D. A. GODRON, Mem. Acad. Stanislas, 1865 and 1868 (Delphi- nium chinense, etc). 8 GODRON, loc. cit., 1868, pp. 3-8, Cultures from 1862-68, with more than fifty peloric plants. 7 G. VROLIK, Ueber cine sonderbare Wucherung der Bhnnen bci Digitalis purpurea, Flora, 1844,?. i, Plates I and II; also Fortge- setzte Beobachtungcn iiber die Prolification von Digitalis purpurea, Flore, 1846, p. 97, Plates I and II. 8 The following selection of references may be of use : W. F. R. Heritable Pclorias. 223 are, however, highly variable and only too frequently accompanied by other malformations. The commonest of these are an increase in the number of organs, the formation of catacorollas and the production of a secon- dary raceme from the axis of the flower. These are the cases which are most commonly described and figured in literature. In order to find more regu- lar and even perfectly pentamerous flowers we must look to the tops of the weak lateral branches of vigorous plants (Fig. 42) ; these hardly ever proliferate, are often still pleiomerous, but there will also occur amongst them flowers with a per- fectly regular corolla with five lips and five erect stamens. The peloric flowers of Digitalis purpurca are always terminal, whether they occur on the main stem or on branches. The same is true of most other Scrophu- SURINGAR, Plantaardige Monstruositcitcn, K. Akad. v. Wetensch., Amsterdam, 1873, 2d. R., Vol. VII, Plates I-II P. MAGNUS, Digitalis pur pur ea, Sitzungsber. Prov. Brandenb., Vol. XXII, 1880, p. J. C. COSTERUS, Teratologische Verschynsclcn by Digitalis pur- purca, Ned. Kruidk. Archief, 1885, Plate VII. ANGEL GALLARDO, Fasciacion, Proliferation y Sinantia, Ann. Miis. Nacion., Buenos Aires, Vol. VI, p. 37, PI. 3; also Sobre algunas onomalias de Digitalis purpurca (with complete bibliography), same journal, Vol. VII, pp. 37-72. Fig. 42. Digitalis purpurca mon- strosa. A lateral branch with a terminal pentamerous peloria. 224 Observation of the Origin of Varieties. lariaccac, 1 and of many other families, especially or- chids. 2 The relation between this position and the reg- ular form of the flower is still without a proper explana- tion; and the question whether the anomaly is due to high nutrition or to the absence of the factor which determines the bilateral symmetry or both, still awaits a definitive answer. Laterally situated peloric flowers are very rare but sometimes occur as we have seen in Linaria Fig. 43. Antirrhinum ma jus. A, Peloric flower from the middle of an otherwise normal raceme, August 1899. Two slips of the corolla stand erect ; the other three are bent downward. B, Normal flower of the same spike. vulgaris hemipcloria (Fig. 41, p. 207), and as is shown by Antirrhinum majus (Fig. 43), etc. Of great impor- tance, also, is the hitherto little noticed fact that in Digi- talis and one or two other cases, the peloric terminal flower opens first of all, whilst the order of opening of all the other flowers on the stem is normal, i. e., aero- petal. 1 EICHLER, Bliithendiagramme , I, p. 208. 2 PFITZER, in ENGLER and PRANTL'S Naturl. Pfiansen-Familien : Orchid., p. 61. For further information on pelorias of Orchids see PENZIG, Mem. Soc. nat. Sc. Cherbourg, Vol| XXIX,, 1894, pp. 79-104. Heritable Pelorias. 225 Peloric flowers occur as chance anomalies in a large number of plants. A specimen! of Scroplmlaria nodosa which I have had growing for the last ten years pro- duced them abundantly. On the other hand my cul- tures of Antirrhinum majiis although of twelve years duration and carefully guarded gave rise to no more than two peloric flowers, one of which is shown in Fig. 43A. Both sprang from the middle of the racemes, that is, they were lateral. I have also observed occasional cases, of peloria on Acsculus Hippocastamim, Melam- pyrnm pratensc, Orobanche Galii, 1 Cytisus Laburnum , etc. In my cultures of 1892 a peloric flower occurred on a plant of Lupinus luteus. The tube-shaped peloric flowers of the cultivated Calceolarias are also well known. In these and similar cases the mode of inheritance has still to be investigated. In this respect the observations of PEYRITSCH are of great importance He has shown that in the case of peloria in Lconiinis Cardiaca, an an- nual Labiate, the anomaly can be reproduced from seed whether this originates from the peloric or the normal flowers of the same plant. PEYRITSCH'S memoir is one of the most valuable of those which deal with peloria, and is indeed an almost complete monograph so far as the Labiates are con- cerned. 2 He has also investigated the influence of the environment on the anomaly as occurring in a series of Labiates. 3 I select the following observations for notice here : 1 See also W. F. R. SURINGAR, Orobanche Galii, Ned. Kruidk. Archief, 1874, Vol. I, p. 330, Plate 18. 2 T. PEYRITSCH, Ueber Pelorien bci Lobiotcn, Sitzber. d. k. Akad. d. Wiss., Vienna, Vol. XL, Part I, 1869, p. 343, Plates I- VI; and Vol. XLII, ist section, 1870, p. 497, Plates I-VIII. s J. PEYRITSCH, Untcrsuch. fiber die Aetiologie pelorischer Bliiten- 226 Observation of the Origin of Varieties. Lamiwn inacnlatuin and Galeobdolon luteiun commonly produce peloric flowers in the neighborhood of Vienna. They often bear them every year on the same plant, but one or more years are sometimes skipped. A sunny position increases the number of anomalous flowers whilst dense shade diminishes it ; consequently one locality often furnishes instances of peloria in several species of Lab- iates (e. g., Calainintha and others), whilst the same species growing together in another locality will not pro- duce a single symmetrical flower or only very few. When- ever the conditions affecting a plant were improved by cutting down timber, peloria occurred in profusion, and the transference of a plant to a sunny spot in a garden often resulted in its appearance. Other authors, and particularly VUILLEMIN/ also assert that the conditions of life play an important part in inducing the anomaly, provided that the inherited potentiality for it is present. bildungen, Denkschr. d. k. Akad., Vienna, Vol. XXXVIII, Part II, 1877, with Plates I- VIII. See also GOEBEL, Organo graphic, I, p. 163. 1 Loc. cit., 1894, P- 33- VII. NON-ISOLABLE RACES. 22. TRIFOLIUM INCARNATUM QUADRIFOLIUM. Few experiences are so well fitted for enabling us to obtain an insight into the nature of specific characters as the failure of an experiment in selection. I am not speak- ing of practical experiments because in such cases the breeder is often disappointed by the fact that the result is not superior to what he has already, or is not suitable for cultivation on a large scale from other causes. This kind of failure only concerns the practical breeder and does not affect the scientific investigator. The object of the latter is simply to find out whether a race specified beforehand can be obtained or not. According to the theory of selection almost anything ought to be obtainable. Almost all characters manifest fluctuating variability to the extent requisite for selection. If the range of variation is considerable, selection should proceed rapidly; if it is within narrower limits it should merely require longer series of generations; and if, more- over, the familiar but undemonstrated opinion is assumed that fluctuating variability increases as the result of the selective process, there is no reason why in any given case the attempt to breed a desired race should not suc- ceed. But this discussion, in my opinion, only applies to ordinary fluctuating variability, and if thus limited, I 228 Non-Isolable Races. willingly agree with the prevailing view. In the sphere of mutability, on the other hand, matters are entirely different. Here species, subspecies, varieties, races, etc. arise by mutations which are induced by rendering active a hitherto latent or semi-latent character. The first con- dition for a desired mutation therefore is the existence of the character in question in a latent or semi-latent state. Without this nothing can be achieved, at least in the present state of science, and it is only in the case of semi-latency that we can have any sort of evidence that the desired character is present. Horticultural breeders are well known to be continually on the lookout for any such indication. 1 But the presence of a latent character is not of itself sufficient, according to my experience, to insure the suc- cess of an experiment in selection. For many an experi- ment has failed in spite of years of labor. This proves nothing in itself, because it is often due to lack of sufficient experience, and this experience can only be acquired by carrying out a successful experiment in an analogous case ; in other words, by making exactly the same experiment with a related plant, preferably with another species of the same genus. For this reason I have more than once endeavored to breed a race analogous to one already existing in a closely related species of the same group, which is either on the market, or has appeared in my own cultures. Ex- perience has taught me that the end may often be attained with greater or less ease according to circumstances ; but that in many other cases, so it appears, insurmountable obstacles bar the way. A very definite and simple case is afforded by the 1 See Vol. I, Part I, 25, p. 188; and this volume, Part I, 2, p. o. Tri folium Incarnatum Quadrifoliiun. 229 attempt to breed a five-leafed race of the crimson clover (Trifolium incarnatum) analog- ous to the five-leafed race of the red clover (Trifolium pratense) which has already been described ( 5, p. 36). I started the ex- periment in 1894; since then I have devoted a great amount of trouble to the task without any result, until in 1900 I gave it up. The attempt simply does not suc- ceed, with my material at any rate. The object was worth a great effort. At first I believed that I had artificially made the five- leaved red clover, or as it is often expressed, that I had cre- ated it. The gradual develop- ment of my theory, however, led me to doubt the correctness of this opinion. It seemed possible that I had merely found the race already existing in nature, but in a condition in which it was not recognizable as such. Eight years however had gone by since the beginning of that culture, and it was practically out of the question to go back to it. I re- Tig-. 44. Trifolium incar- 1 i ,1 f . natum. A flowering solved therefore to endeavor to branch w ith a single 4 - raise a new five-leafed clover and fo r liate leaf; . the . result . of an experiment in se- selected the crimson clover. This lection lasting six years. 230 Non-Isolable Races. choice was largely determined by the fact that there were no published records of 4- or 5-foliate leaves of this clover, 1 which means that the character, if present in a latent state, is much rarer than in the red clover. I take this opportunity of calling attention to the inestimable value of PENZIG'S "Teratology," This lies per- haps rather on the negative than on the positive side, for it is of course possible to collect the main literature relating to a given question oneself, although not with- out the expenditure of a great amount of time; but if one is not a teratologist by profession, it seems hardly possible without some such help, to satisfy oneself that Fig. 45. Tri folium incarnatum, 4-foliate leaves, the middle one with incomplete segregation of a lateral leaflet. absolutely no records relating to a particular phenomenon exist. The first step in a purely scientific breeding experi- ment evidently is to find out whether the deviation in question has occurred before, and if so, whether it is rare or common. My belief is that the commoner anomalies are heritable characters with a high index of inheritance (often about 30-40% or more), but that the rarer ones are the occasional expressions of latent or semi-latent characters. These are also inherited in their latent state, and if they turn up here and there this latent condition must probably be widely distributed. 1 0. PENZIG, Pflanzcnteratologie, Vol. I, 1890, p. 385, where incarnatum is not even mentioned. T. Trifolium Iiicanialum Quadrifolium. 231 If Trifolium incarnatum with 4-foliate leaves had often been mentioned it would therefore seem probable that a five-leaved race of it occurs in nature, although just as little separated from the ordinary crimson clover as the five-leaved race of the ordinary clover is from this. Latent characters, in my opinion, are often older than the species which bear them. I regard the division of the leaf into four blades in this case as an atavistic phenom- enon, and I believe that this latent potentiality is as old as the whole group of clovers with trifoliate leaves (Trifolium, Mcdicago, Mclilotns etc.), that is, older than the individual gen- era of this group. In many spe- cies this power of reproducing quadri foliate leaves may have been completely lost, for it is mentioned in PENZIG'S book only for a relatively small number of them. In others, however, it has Fl > ^. Trifolium An atavistic pinnate leat. persisted to the present day. If the trifoliate leaves of the clovers are derived from Papilionaceae with pinnate ones, the multi foliate leaves which they occasionally produce must evidently be re- garded as atavistic phenomena. The correctness of this view is proved by those very rare cases in which, in the races in question, pinnate leaves appear instead of the ordinary multi foliate ones. I have observed this from / time to time in my Trifolium pratcnse quinqne folium (Fig. 46) and the same thing has been found by other authors in Trifolium minus and Trifolium re pens. I have myself found 4- and 5-foliate leaves in Medi- 232 Non-Isolable Races. cago lupnlina, whilst BRAUN has observed them in M. saliva. They are well known in T. pratense and T. rcpcnSj and WYDLER has recorded 4-foliate leaves in Lotus major and Tetragonolobus biflorns. In some suc- cessive sowings which I made with Medicago lupnlina I found the character to be inherited although in a mod- erate degree only, but I have not continued the experi- ment. But let us return to the crimson clover. The question is, what prospects were present at the beginning of the experiment, and what may be expected from such ex- periments in general ? There are three main possibilities to be considered. \Ye may find at the beginning of the experiment (See 3 p. 20) : 1. A race which often exhibits the anomaly in ques- tion, and bears it as a heritable character, i. e., an ever- sporting variety ; 2. A half-race with a semi-latent anomaly which is only occasionally manifested; 3. An ordinary plant of the species with the character in question in a latent condition. In the first case the race already exists and all that is necessary is to isolate it ; in the second it may possibly be obtained ; in the last there is little prospect of doing so. In order to present a clearer idea of the mutual rela- tions of these three cases let us examine Tri folium repens and T. pratense. That the anomaly is by no means very rare is testified in both cases by the popular belief in the so-called lucky four-leaved clover as well as by common experience. If looked for in a field of clover, or in a meadow, or along the roadside, a four-leaved clover will be found from time to time. If repeated attempts are made to find them they will certainly prove to be rare Tnfolium Incarnatum Quadrifoliuin. 233 but not so rare as we might have imagined. I have found them almost every year, and often quite soon after I had been asked for one. On the other hand there is on the market the 5-foliate T. repcns air o pur pur cum which is often cultivated in gardens for its dark brown leaves, and for T. pratense I have described the five-leaved form in detail in 5. Plants of T. pratense are sometimes found in the field with two or more 4- or multi foliate leaves. I found one in 1866 in the Cronesteyn estate near Leyden, and an- other in 1886 near Loosdrecht. The first had several 4-foliate leaves, and also some 5-6-foliate ones. I se- cured the former but did not cultivate it; the latter formed the starting point of my race. In view of my present knowledge I must assume that in both cases the plants already belonged to the race when I found them ; and I also consider it as probable that this race had arisen on these very spots, or at least not far from them. 1 Whether the same race can also be produced from the occasional stray four-leaved clovers I do not, of course, know ; but I anticipate that the attempt would sometimes succeed and at other times fail. If this view is confirmed by future experiments we shall have proof of the exist- ence of the two races, the eversporting variety and the half-race, existing simultaneously within the limits of a single species. For the present we must be satisfied with the knowledge that there exists a race rich in anomalous leaves in the red and in the white clover, and one in the crimson clover which bears the character only in the semi-latent state. I shall now proceed to the description of the latter. In the winter of 1894-95 I bought a kilo of the seed 1 A polyphyletic origin, therefore, as in Linaria vulgaris peloria. 234 Non-I salable Races. of the ordinary crimson clover and sowed part of it on a bed of about five square meters. Two of the seedlings were tricotylous and one was tetracotylous, and these were transplanted to a special bed as soon as possible in the hope that tkey would exhibit the desired abnor- mality. This hope was based on the principle of the correlation between different kinds of anomalies. 1 If a plant exhibits an anomaly in its early stages it will, ac- cording to this principle, be more likely than any other individual in the same culture to give rise to other devia- tions later on. In this particular case my expectation was fulfilled, for the tetracotylous plant produced one 4-foliate and one 5-foliate leaf in the course of the sum- mer. Such were not found on any other plant, either during the course of the experiment or at the end of July when the plants were in full bloom and were pulled up and minutely examined. There were about a thousand plants. I left the three selected specimens to flower together and sowed their seeds in April 1896. Over 600 seedlings came up, all of them with only two cotyledons. In all of them the first leaf was single, which is the general rule in clovers (Fig. 47 A). The second and third leaves developed in May; they were quite normally trifoliate, with the exception of one, of which one of the three leaflets was" split laterally, although not completely di- vided. The form of this blade was similar to that figured in Fig. 45 B. About 250 individuals of the whole group were planted out. The seed had been sown in pans; the young plants were transplanted into pots and were planted in the beds ir the middle of May. At the end of June, 1 Eine Methode, Zwangsdrehungen aufzusuchcn, Ber. cl. d. hot. Ges., Vol. XII, 1894, p. 25. Trifoliiun Incarnatum Qiiadrifolinm. 235 at the beginning of the flowering period, several indi- viduals exhibited one or more 4-foliate leaves ; the anom- aly was therefore a heritable one. Moreover the multiplication of the blades had also increased considerably as the result of selection, as the following figures prove. These refer to the offspring of that seed-parent which had already exhibited the anomaly in the previous year. There were 90 of them; among the offspring of the tricotylous parents "4-leaves" were not entirely absent, but they were relatively scarce, and Fig. 47. Trifolium Incarnatum. A, a seedling with normal primary leaf. B D, seedlings with 2- and 3-foliate leaves. The former arise from the larger; the latter from the smallest seeds. the whole group was consequently pulled up at the be- ginning of the following period. About % (58 out of 90) of the rest were perfectly normal without any in- crease of the number of leaflets. On the average they had about 10 stems and 100 leaves per plant. The re- maining plants formed a half-curve 1 of the following composition. The first row gives the number of 4- or 5-foliate leaves per plant, and the second the number of 1 4, p. 26 ; and Ueber halbe Galton-Curven, Ber. d. d. hot. Ges., 1894, Vol. XII, p. 197- 236 Non-Isolable Races. individuals on which these numbers were observed (cul- ture of 1896) : Abnormal leaves 1 23456789 Individuals 58 10 12 4 2 2 1 1 The 58 normal plants were pulled up. Of the rest four were weak and died; there remained 28 which all flowered together. Their seed was harvested separately after the number of 4- and 5-foliate leaves on each parent had been recorded. In March 1897 I sowed a part of this seed in pans, separately for each seed-parent. The object of this was to find out whether there was any difference between the individual seed-parents with regard to the number of anomalous offspring which they produced. From an examination of the pans it was easily seen that the ab- normality had already appeared in the primary leaves of some of the seedlings. In the great majority of cases these were perfectly normal, consisting of one leaflet as in the whole of the previous generation. In some cases however this primary leaf consisted of two or three leaflets (Fig. 47 B-C). Such occurred in the crops raised from 6 of the 21 plants whose seeds had been sown. Each seed-parent had given a crop of about 300 seed- lings. Five of the crops contained not more than 2 ab- normal seedlings, but the remaining one had a very large number, namely 14 amongst 335 seedlings or about 4%. It is worthy of notice that the parent of this crop had only had two 4-foliate leaves itself and thus had not given the least sign that it would produce offspring with so much higher a degree of the abnormality. Moreover I could not find any relation between the number of ab- normal leaves on the other seed-parents and the pro- portion of abnormal offspring raised from their seeds. Trifolium Incarnatum Quadrifolium. 237 The plant with nine 4- or 5-foliate leaves did not give rise to a single anomaly amongst 300 seedlings. Amongst breeders of animals it is generally recog- nized that the visible characters of an animal are of very little use as an indication of its value for breeding. The offspring which the animal has already produced afford a much more reliable indication. On the basis of the choice of the seedlings, the 14 abnormal offspring of the seed-parent with 4% were planted out in the beds in June 1897, together with the seven next best plants. The latter produced very few 4- to 5-foliate leaves per plant, the first and 1 in eleven cases, but 9, 9 and 4 in three cases. There was there- fore no marked advance on the previous year in this respect. The progress was just as inconsiderable in the harvest of that year. The percentage of abnormal individuals amongst the seedlings ranged in 1898 between 1 and 4% and in one case reached 6%. On the other hand all the (19) seed-parents investigated had at least one and usu- ally two or more seedlings with a divided primary leaf. But here again no relation was manifested between the number of abnormal seedlings and the number of 4- or 5-foliate leaves on the seed-parents which produced them. 227 seedlings were planted out, most of which were perfectly normal at the time of flowering. I obtained the following half -curve (1898) : Number of multifoliate leaves per plant: 12345 Individuals: 188 29 7 1 1 1 That is to say, about 20% of individuals with the inherited anomaly in from 1 to 5 of the whole number 238 Non-Isolable Races. of leaves counted on the plant (about 100). The numbers were therefore smaller in this than in the previous year. For this culture I had planted out the normal and ab- normal seedlings of the most abnormal seed-parents and some abnormal seedlings of the remaining seed-parents. No essential difference between these three groups could be detected when they were recorded at the time of flow- / ering. Pitcher formation was observed both amongst the seedlings and during the later stages ; this is another indication of correlation amongst the various characters. In the summer of 1898, 41 of the selected plants fur- nished a sufficient quantity of seed. In the following spring I determined the proportions of seedlings with compound primary leaves in the crops from each of these parents and reckoned them in percentages. The compo- sition of the 1898 harvest with respect to this character was : Percentage of abnormal offspring 1 2 3 4 5 8 11 15 16 20 24 27 Parents 3 12 75421211111 That is to say, a considerable advance which at once becomes evident if this series of figures is compared with that given above for the 1897 harvest (1-4 and 6%). This advance has moreover taken place in spite of the falling off in the number of 4-foliate leaves in the seed- parents. In the spring of 1899 I only selected seedlings with trifoliate primary leaves for transplanting (see Fig. 47C), and only from amongst the offspring of the four seed- parents with from 15-24% abnormal offspring. At the time of flowering, however, my hopes were disappointed. In the middle of July there were amongst 120 richly branched flowering plants 45% without the anomaly, Trifolium Incarnatum Quadrifolium. 239 27% with a single abnormal leaf each, and 28% with two to four 4- to 5-foliate leaves each. That is to say, 55% abnormals as against 20% in the previous year which indicated a marked advance. But my hope of obtaining a leaf with more than five leaflets was not fulfilled. In spite of repeated search I never found one. Nor did I obtain plants rich in four- bladecl leaves ; for there were none with more than four of them. Therefore I have since abandoned the hope of breed- ing a race of four-leaved clover, corresponding to my Trifolium pratcnse quinque folium, from this material. A striking feature of this experiment is the apparent absence of a relation between the degree of abnormality of the adult plants and that of the seedlings. For the paucity of four-bladed leaves in the grown plants seems incompatible with the abundance of multi foliate primary leaves in the seedlings from which they grew. The failing of this relation has led me to the dis- covery of a most remarkable connection between this variability and the size of the seeds, for the smallest seeds are those which give rise in the largest number to compound primary leaves. Small seeds germinate somewhat later than larger ones and also give rise to weaker plants. It had often struck me that the selection of the most abnormal of the seedlings was frustrated by the fact that many of the individuals with compound primary leaves were too weak to be planted out, or died soon after the process. It also struck me that all the seedlings in a pan could not be recorded at the same time. At first view the plants ap- pear to germinate very regularly, and hundreds in the same pan seem to unfold their leaves at the same moment. 240 Non-I salable Races. At this point they were recorded and, if the first leaf was single, were usually pulled up. Those which were saved were usually weaker, more stunted and backward in growth. Several had not yet unfolded their first leaves, and amongst them a great number of the anom- alies were found when the examination was repeated a few days afterwards. I then convinced myself by a very simple experiment of the correctness of these conclusions. All that was necessary was to isolate the large and the small seeds in a sample and to sow them separately. But as there is no absolute limit between the two it was necessary to know how many seeds should be separated out, as the smallest. And this can only be done by the number of anomalies, i. e., compound primary leaves, they produce. I therefore selected a sample of seed whose capacity for producing anomalies I already knew. This was 15%; the sample was derived from a single seed-parent. I separated the seeds into three categories, small, inter- mediate and large. All in all there were 217 seeds of which 17 did not germinate. The characters of those which did are as follows: Number of divisions in the primary leaf. 123 2-3 Small seeds 31 9 16 12.5 % Intermediate seeds 50 2 1 1.5 % Large seeds 88 2 1 1.5% 169 13 18 15.5 % It will be seen that almost all the abnormal seedlings are derived from the smallest seeds. The seedlings from the large seeds had, with a single exception which was an abnormal one, unfolded their primary leaves in May, and fourteen days after the seed had been sown ; the same Trifolium Incarnatum Quadrifoliiun. 241 is true of the intermediate seeds with the exception of four, two of which were abnormal. 22 normal plantlets developed from the smallest seeds in the same time ; the 9 other normal and the 25 abnormal ones did not unfold their first leaf until the third week. 1 These facts show further that the number of seed- lings with abnormal primary leaves does not depend simply on the degree of fixation of the variety. It de- pends mainly on the proportion of small seeds. This, however, in its turn, depends on the size of the harvest. Fig. 48. Trifolium incarnatum. Monstrous seedlings from the smaller seeds. A, B, D, with two to four primary leaves ; C, with a double leaf with broad flat peduncle. In the 41 samples which composed the harvest of 1898 there were 8 with 8-27% abnormals; these samples con- sisted of from 0.3 to 1.5 cc. of seed. The remaining samples consisted of from 2 to 5 cc. of seed and the number of abnormals produced ranged between and 5%. From these facts we see that the weaker individ- 1 In stocks also the seeds which produce plants with double flowers and those which give rise to "singles" have different rates of germination, as is well known. An investigation of the seeds of inconstant varieties, or, as they are usually described, varieties which have not "yet" been fixed, would be certain to reward the inquirer with many interesting discoveries. 242 Non-Isolablc Races. uals, which gave a poorer harvest, gave rise to the larg- est number of abnormals amongst their offspring. I have repeated the same experiment with the harvest of 1899, with the seeds of four separate seed-parents, but as I did not know their capacity for producing abnormals in advance, the difference was not so striking. The large seeds gave rise to 2-4%, the small ones to 3-13% abnor- mals. Altogether seedlings from 2758 large and from 617 small seeds were examined. Two questions present themselves in connection with the interpretation of these experiments: (1) Can the position in which the small seeds are chiefly produced on the plant, be determined? 1 (2) Are the germs of the small seeds perhaps the better nourished ones ; is there, for instance, just as much nutriment brought to them as to the large seeds, but must they, for want of room or for other reasons, utilize it in some other way? I recommend these problems for further study, and may perhaps in the mean time record a few facts bear- ing on them which I have observed. In the crimson clover, monstrosities occur much more frequently among the seedlings from small than among those from large seeds. The latter are almost all perfectly normal. The small seeds often produce plants with supernumerary cotyledons, or with two or more primary leaves (instead of one) or with divided peduncles, symphyses in the leaves and other malformations (Fig. 48). Unfortunately it is often difficult to keep these individuals alive and to bring them to flower. Let us now cast a final glance over the whole course of the experiment. 1 In stocks, according to CHATE, Culture des Giro-flees , the seeds which produce double-flowered plants arise chiefly from the lower half of the pods of the strongest racemes of the plant. Ranunculus Bnlbosus Semiplenus. 243 A crimson clover plant with some quadri foliate leaves was obtained by the selection of tricotylous and tetra- cotylous seedlings in conformity with previously studied laws of correlation. The anomaly proved to be heritable and has maintained itself until now, during six genera- tions (1895-1900). It was improved by selection but only within very narrow limits. Plants with more than five leaflets per leaf have not as yet arisen, nor have plants bearing ten or more 4- to 5-foliate leaves, and it is nearly always the "small" seeds which give rise to seedlings with compound primary leaves. But the chief result is that the desired race, rich in 4- foliate leaves, T. incarnatum quinqucfoUuni' anal- ogous with Trifolium pratcnse quinquefolium, did not arise. 1 23. RANUNCULUS BULBOSUS SEMIPLENUS. Double flowers are common phenomena amongst the buttercups. 2 They occur not only in the cultivated Ra- nunculi (R. asiaticus} but also in several wild species. The doubling may be either complete and brought about by petalomania as in the Ranunculus acris shown in Fig. 40, Vol. I, p. 194; or it may be more or less incomplete when caused by the transformation of a varying number of stamens into petals (R. acris, R. auricoinus, R. Phih- notis, R. r opens etc.). In Ranunculus bulbosus, the bulbous buttercup, the 1 The same thing no doubt occurs also in other cases. The at- tempt to breed from occasional anomalies a constant race endowed with the particular variation, in some cases succeeds, but in others does not. For instance I have for many years endeavored to raise from the occasional polycephaly in Papaver cotnmutatuni a race with as beautiful crowns as those which characterize the familiar Papai'er somnifcrum polycephalum (see Vol. I, p. 138, Fig. 27), but in vain. - See PENZIG, Pfianzenteratologie, Vol. I, pp. 181-189. 244 Non-holable Races. stamens are often (either all, or only some of them) transformed into petals with the result that dense double flowers are produced. 1 These have been described by Fig. 49. Ranunculus bulbosus semiplenus. A , the bulb ; A' and A", its leaves from the axils of which the flower- ing stems S arise ; E, terminal flower of the main stem ; S, secondary flowers partly broken off; T, tertiary flow- ers. (See p. 256). various authors. 2 In the neighborhood of Amsterdam this variety does not occur, so far as I know. On the 1 Loc. cit., p. 185. Fasciated stems with broadened terminal flow- ers are also met with occasionally in the Ranunculus bulbosus in Holland. 2 Compare the Ranunculus bulbosus Alcae of Naples, described by TERRACCIANO, Nov. Atti d. R. Instit. Napoli, 1895, Vol. VIII, No. 7. Ranunculus Bulbosus Semiplenus. 245 other hand on plants growing as they often do in sandy localities, the flowers often possess a slightly increased number of petals. In these abnormal flowers there are usually six or, rarely, seven petals, very seldom more than 10-12. They are as a rule ordinary petals, but there sometimes occur some that are much smaller and narrower and are ob- viously metamorphosed stamens. This metamorphosis is often only partial, and the famil- iar intermediate stages are exhib- ited. The abortive stamens are usually to be found among the most peripheral ones ; but they are not necessarily the outermost ones, di- rectly adjoining the corolla. 1 The potentiality of this doubling is therefore present in a semi-latent condition in the wild plants of this species growing in this neighbor- hood. I regard this race, therefore, as a half race in contradistinction to the normal double race which is only known to me from the pub- lished records. Obviouslv the two J possess the same character; which Fig. 50. Ranunculus bul- bosus semiplenus. A flower with 31 petals (partly petalodic sta- mens) ; the only one amongst 4425 flowers. It occurred on a qua- ternary branch in my culture of October, 1892. See the series of figures on page 252. is, however, active in the one case but latent or semi-latent in the other. It seemed to me important to discover whether it was possible to obtain the double from the half race by selec- tion. According to the views advanced in this work this should be possible, but not every attempt need neces- sarily be successful. But if it does succeed the change 1 See GOEBEL, Jahrb. f. wiss. Bot., Vol. XVII, pp. 217-219. 246 Non-Isolablc Races. must be brought about suddenly, and, under ordinary conditions of culture, be effected in the course of a few years. In this way the double variety may have arisen from time to time in the wild state; and in the same manner the present half race may perhaps, in the course of time, undergo this change. This transformation, however, cannot be simply the result of careful selection. A mutation is needed; and we know as little about the causes of mutations as about the method of inducing them artificially. Mutations are known to occur with moderate frequency both in breed- ing experiments and in nature, but, up to the present, their occurrence has been a matter of chance (10 and 11, pp. 95-103). In my experiment such a mutation did not occur, although it extended over five generations. 1 The half race was distinctly improved by repeated and very strin- gent selection. It became at the end very rich in extreme or almost extreme variants, but it was just in these that it proved to be so remarkably constant. In its five gen- erations it reached a point which did not seem to me likely to be exceeded by further selection. It produced occasional flowers with more than 1 5 petals, and a single one with 31, but the mean number of the petals in its selected individuals did not exceed 9-10. The double variety did not arise from it, in spite of every effort. 1 The fluctuating variability of the semi-latent character in Ranun- culus bulbosus scmiplctnis seems to cover a much wider range of forms than in Trifolium. There the extremes are 3 and 7 leaflets; in the buttercup they are 5 and 31 and perhaps more petals. From this it does not, however, follow that the variation is greater in the one case than in the other, but only that the variation is expressed by a larger number of divisions in the latter case, i. e., that there are more scale characters in the curve. Ranunculus Bulbosus Semiplenus. 247 I conclude, therefore, that in this case the half race cannot be transformed into the double race by simple selection, but only by an internal change a mutation the external causes of which are still unknown to us. Proceeding now to the detailed description of my experiment, I begin with the half race in the conditions in which I originally found it in nature. I found the half race growing in 1886 and 1887 in a sunny and sandy spot not far from Hilversum, where I have often seen it since. The bulbous buttercup grew there in abundance ; most of the flowers were normal, but a considerable number had more than five petals. I shall refer to these latter for convenience of expression as pleiopetalous. For several years I have recorded the flowers in that locality. I give the records of 1886 and 1887, each of which relate to 300-400 flowers. The data are given as percentages. Number of petals 5 6 7 8 9 10 11 12 13 14 Flowers in 1886 91.5 5.5 1.2 0.6 0.6 0.3 0.3 Flowers in 1887 90 7 2 0.5 0.5 The two series 1 agree as closely as could be expected and the records for the other years also fall in line. The maximum of the curve (see Fig. 51 H) is over the normal number of petals ; and from it the curve falls rapidly. It is a so-called half GALTON-curve. Flowers with less than five petals do not occur in this locality. The great steepness of this curve is due to the fact that on many of the plants no pleiopetalous flowers were found on the days when the observations were made. But this does not mean that the half race is mixed with 1 Ueber halbe Galton-Curven ah Zeichen discontinuirlicher Varia- tion. Ber. d. d. bot. Gesellschaft, Vol. XII, 1894, p. 197, where some of the series of figures given below can also be found. 248 Non-holable Races. a pure race with five petals only. For the plants in ques- tion were either weaklings, or exhibited pleiopetalous flowers on other days. I was often able to observe that on many plants six-petalled flowers occur on one day but not on another. The 6-7-petalled flowers are found from the beginning of the flowering period, but the higher figures do not occur till later, as is also known to be the case in other instances of double flowers. In 1887 I moved some plants in which the abnormal- ity was well developed, to my garden, where they flowered again in the following summer and set eed. These plants constitute the first generation of my experiment. Since then I have sown seed every year, but only part of the plants, sometimes one-half, sometimes two-thirds, pro- duced flowering stems in the first year, and I have always confined my attention to these, throwing away those which did not bloom during the summer. I have some- times kept some of the best examples of the half race through the winter for secondary experiments, but I shall return to these later on. During the period 1889-1892 the second to the fifth generation of the half race were grown in this manner, the extent of the cultures being gradually increased. I always harvested my seed from the most abnormal in- dividuals, which I selected by simply cutting off the flowers with five petals from all the plants. The numbers of these on the individual plants were recorded in some years but not in others. Pollination was left to the bees, but no definite effects of cross-fertilization have been traceable in the results of the experiments. The first two years of the experiment (1889 and 1890) need only a brief reference. Plants without pleio- petalous flowers or with only very few, were removed Ranunculus Bulbosus Semiplenus. 249 as soon as possible, or were deprived of their flowers: of the rest, only the seeds of flowers with six and more f> petals were saved. But this process is not one of selec- tion, as will be shown by means of some special experi- ments which were instituted later. The result of selection could be seen in 1891 in the best examples of the half race, but in 1892 (the fifth generation of the culture) in nearly all the plants. The number of petals increased in every respect, The apex >H7887. Fig. 51. Ranunculus bulbosus semiplenus. Experiments in selection during the period 1887-1892. H 1887, curve of the wild form; E 1891, curve of the abnormal plants in 1891 ; A 1891, curve of the selected seed-parents in 1891. 1892, curve of the whole crop in August 1892. The num- bers at the base refer to the number of petals per flower. of the curve shifted to 9 and 10 petals and even further; that is to say, the mean of the half race (9-10 petals) was separated by selection from that of the pure species (5 petals), a point which is rather striking because this was not effected in my experiment with Trifoliuin in- carnatmn. The course of the whole experiment is ex- hibited graphically in Fig. 51 which is composed of four curves. The first (H 1887) exhibits the countings given 250 Non-Isolable Races. above, which were made in the original locality. Then there are two curves for 1891. In this year I had a culture of about four square meters from which I re- moved, about the beginning of August all plants which had not produced any pleiopetalous flowers as well as those which had not yet bloomed. For two weeks I counted every flower which appeared on the remaining specimens. They amounted to 128 and the various de- grees of the anomaly were distributed over them as follows : Number of petals 5 6 7 8 9 10 11 12 13 Number of flowers 45 24 28 17 8 4 1 1 The curve E 1891 in Fig. 51 is based on these figures. It is a half curve like the previous one, but without the steep apex. The disappearance of this is due partly to cultivation and to the repeated selection, but partly also to the fact that the individuals with the smallest number of pleiopetalous flowers had been removed before the counting took place. After these data had been determined I carried out a still further selection. Several plants had not produced a single flower with more than seven petals. These were removed in the middle of August and observations on the rest were continued. There were 18 plants, all of which were selected for seed-parents as being the best representatives of the race. I counted all the flowers which bloomed from August 15 to 31, and obtained the following numbers: 1 Number of petals 5 6 7 8 9 10 11 12 13 14 Number of flowers 9 17 39 64 45 37 15 9 6 2 1 In the preliminary account of this experiment, mentioned above, this series is given separately in two curves, one of which refers to the most abnormal plant, the other to the rest. Ranunculus Bulbosus Semiplenus. 251 Total 243. The curve (Fig. 51, A 1891) has be- come two-sided. It has no maximum at 5 but a very definitely pronounced one at 8. It is composed of ob- servations made on 18 plants which differ little from one another. Individuals with these characters occurred neither in the original locality nor at the beginning of my experiment. The sowing, in 1892, of the seeds of these selected individuals gave rise to above 300 plants which were coming into flower from July 21 to August 31. The curve for 1892 in Fig. 51 refers to these. Those which flowered later were examined separately and will be de- scribed afterwards. On all the flowers which opened between the dates named the petals were counted, and the numbers entered in my notebook separately for each plant. I give the totals, which relate to 4425 flowers. The numbers of petals were distributed over these as follows : Petals 5 6 7 8 9 10 11 12 13 14 15 16-31 Flowers 409 532 638 690 764 599 414 212 80 29 18 20 The curve which is now an index of the degree of de- velopment of the whole race, agrees fairly closely with that of the selected seed-parents of the previous year ( 1891), as can be seen from a comparison of the curves A 1891 and 1892 in Fig. 51. The apex of the curve, however, has advanced a whole petal. There has been no regression as is the case in the selection of active char- acters, but a progression such as is usually characteristic of the selection of semi-latent characters. The change in the right half of the curve is also im- portant although not given in Fig. 51. It consists in the occurrence of more extreme variants. In the previous 252 Non-Isolable Races. generation there were no flowers with more than 14 o petals. Now there are 38, distributed as follows : Petals 15 16 17 18 19 20 21 22 23 31 Number of flowers 18 852111011 It should be noted, however, that they were found amongst a group of 4425, and therefore only amount to about \% (0.86%). But as not a single one was found in 1891 amongst 243 flowers, a genuine, although only a slight, advance has taken place. The great majority of the 295 plants which were flowering in August and formed the 1892 culture, had F. Fg. 52. Ranunculus bulbosus semiplenus. Composition of the fifth generation in 1892. A, the curve of some "ata- vists" ; M, the curve of intermediate individuals ; V, the group of extreme variants. The figures refer to the number of petals per flower. individual curves whose maximum was at 9. But amongst their number were variants and extreme variants also. On the one hand there were "atavists" with a maximum on the ordinate of five petals, i. e,, with a one-sided curve, as in those from the original locality ; on the other hand there were variants on the plus side which bore on the average eleven petals per flower. In one case even a mean of 13 petals was reached. These curves were two-sided, and not, as in the five-leaved race of the red clover, in- Ranunculus Bulbosus Semiplenus. 253 versely one-sided. But we are dealing here merely with a cumulative effect within a half race and not with an isolated, fully developed race. I have chosen a number of variants from the two categories, have added up the number of their petals, and obtained the following data : Petals: 5 6 7 8 9 10 11 12 13 14 15 16 17 18 23 Number of flowers /$. 66 34 21 18 15 11 720010000 " M: 13 14 22 28 51 26 16 12 6 4 2 1 1 V: 9 11 26 39 62 79 148 84 30 8 4 3 2 1 1 These data are exhibited graphically in Fig. 52. They relate to three small groups of individuals, chosen in such a way that the curves of the individual plants did not exhibit any considerable deviations from the mean of the group. A is the curve of the twelve atavists extracted from the whole series of observations ; the maxima of all their curves were at 5 petals. M is a curve represent- ing ten plants grown from the seeds of a single seed- parent. V is the curve composed of all the plants the apex of whose individual curves lay above 10. There were 22 of them altogether; the apices of their curves were at 11, with three exceptions which were at 12 and 13, but these curves did not exercise any marked effect on the shape of the average curve of the whole group. If Fig. 51 and Fig. 52 are compared a most remark- able similarity will be observed. The latter figure gives the composition of my race at the end of a process of selection extending over five generations, the former re- lates to the separate stages in this process. This mutual resemblance lies in the fact that the original half curve (Fig. 51, H 1887) continues to appear throughout the process, although it is a little flattened; it occurs in 1891 (Fig. 51, E 1891) and also in 1892 (Fig. 52 A). Ata- vistic fluctuation therefore is still exhibited by my race 254 Xon-Isolable Races. in spite of the repeated selection. 1 The curve M has a more normal shape than the corresponding curve "1892" in Fig. 51; which is obviously due to the fact that the former represents a homogeneous group whilst the latter is a composite curve embracing all the groups of this culture. Curve V is related to curve M as it would be if we were dealing with ordinary fluctuating variability ; it is simply shifted to one side. 2 It seems obvious that the race could still be improved by sowing the seeds of those seed-parents the apices of whose curves are at 11, i. e., that these curves could be shifted still further to the right. I have made some such sowings since 1892, but only on a small scale and not without interruption. They were not intended as a con- tinuation of the experiment. The number of petals per flower increased slightly, but the type itself was not es- sentially altered. I especially never saw a trace of any- thing like the origin of a double flower. In order to find out whether there was any likelihood that the type of my race would in the near future mani- fest an improvement I made the following calculation. The 295 plants of which the culture of 1892 consisted, arose from the seeds of 21 seed-parents. I selected the ten best of these parental groups and plotted the curves for all the offspring of each seed-parent. The curves proved to differ very little from one another. Their apices all fell over nine petals, with one exception, which was over ten. I should say that in making the calcula- tion I have left the groups which contained less than 300 1 Whereas as a result of the selection of active characters the whole curve is shifted; see Vol. T, p. 73, Fig. 18, and the third part of the first volume. ' Sec Vol. T. Fig. 116. on page 536. Ranunculus Bulbosus Semiplenus. 255 flowers out of consideration. But even these did not manifest any notable differences. I then compared these ten curves with the part-curves determined from the parents themselves (i e., with the number of flowers counted on the seed-parents) and found no correlation. As a matter of fact the seed-parent with the smallest number of pleiopetalous flowers had the offspring with the largest number. The following four curves of the offspring of four seed-parents are extracted from my records. Under M are given the numbers of petals in the seed-parents of 1891. M NUMBER OF PETALS PER FLOWER 5678 9 10 11 12 13 14 IS 16 17 18-23 Totals C 510 37 47 81 81 85 102 47 31 63410 525 C 610 25 67 80 75 117 77 75 45 30 10 6 1 2 3 613 C 611 54 53 62 78 87 60 59 37 10 4 4 1 1 1 511 C 711 52 57 76 77 95 64 26 13 460 Another fact which points in the same direction is that the plant which seemed to be far the best in the summer of 1891, inasmuch as the apex of its curve was over 11-12 petals, had offspring whose character cor- r^sponded exactly with that of the whole culture of 1892. The improvement on the seed-parent therefore did not justify the expectation of a real advance. For these reasons I then discontinued the experiment. It seemed to me that the impossibility of raising the double race from my half race by simple selection was placed beyond all doubt. This result could only be ex- pected from a further mutation. The extensive material afforded by these cultures has been utilized to find out how far the number of petals per flower in the half race is determined, apart from selection, bv internal causes, and how far bv external. 256 Non-Isolable Races. I found it to be dependent only to a very slight degree on the former but in a high degree on the latter. The first question that presented itself was : To what principles of distribution does the number of petals on the individual plants correspond. Is this number de- termined by the situation of the flower or by external factors or by both? With regard to the position of the flower BRAUN in his Verjilngung distinguishes between a strengthening and a weakening system of branching. In the former the branches increase in strength, though often but slightly, with each new degree of division; in the latter each secondary branch is weaker than the branch on which it is borne. Ranunculus bulbosus be- longs to the former category (see Fig. 49 on page 244). The main stem (E) is surpassed by the vigorous lateral shoots (S), which arise directly from the tuber, and these in their turn are excelled by their own (tertiary) branches (T in the figure). The same thing continues with further growth until ultimately the process is re- versed and weaker branchlets are produced. The more vigorous a branch is, the larger and stronger, as a rule, will be its flower. If we now compare the number of petals on the flowers of this half race, with their position on the branches of the various orders, we are often struck by an apparently definite correlation. But this is only due to the fact that such cases produce a more vivid impression than the opposite ones. When a detailed record is made the latter are found to be just as numerous as the former. In September 1892 I determined the number of petals, and the position on the plant, of 1197 flowers on 82 plants ; and plotted a curve for each position. Here how- ever I only give the means of the curves. Ranunculus Bulbosus Semiplemts. 257 NUMBER OF FLOWERS PETALS PER FLOWER A. On the ma:'n stem: 1. Terminal flower 75 697 9.3 2. Secondary flowers 221 2005 9.1 3. Tertiary " 134 1237 9.3 B. From the tuber: 4. Secondary flowers 259 2419 9.3 5. Tertiary 397 3716 9.4 6. Quaternary 111 1014 9.1 1197 11088 9.3 It is evident that the number of petals in the various groups is practically the same. 1 Even the seeds of pleiopetalous flowers are by no means better. In gathering the harvest of 1891 I col- lected the seed on each plant in a separate bag with ref- erence to the number of petals of the flowers. The flow- ers were labelled for this purpose at the time of flowering. In the culture of 1892, therefore, the plants were arranged in groups, first according to their seed-parents, and sec- ondly according to the petal-number of the flowers from which the seed had been gathered. I then grouped all of the figures by the latter character and obtained the fol- lowing result : NUMBER OF PETALS IN THE MEAN NUMBER NUMBER FLOWERS WHICH OF PETALS OF FLOWERS FURNISHED THE SEED OF THE OFFSPRING 1 COUNTED C5-7 C 8 C 9 C 10-11 C 12-14 Average 8.6 Total 5560 1 In the weakening system of branching on the other hand the contrary seems to be the rule ; so for instance in my cultures of Sapo- naria officinalis with 5-10 petals, in Chrysanthemum scgetiini ( 18) etc. ! The mean is slightly lower here than in the previous table be- cause that only refers to countings made in September. (See later.) 8.3 932 8.7 1072 8.5 1217 8.6 1420 8.7 919 258 Non-Isolable Races. Here again there is no discernible correlation. I have obtained the same result in other years. From this we see that in this case at least selection must not be founded on the different flowers of a plant but on the individual plants. However, the possible influence of the various grades of branching independently of the number of petals remains to be investigated. But whereas no internal causes were found which determined the pleiopetaly in the individual flowers, the external causes could be discovered the more readily. This character follows the general rule ; for the higher the nutrition and the more favorable the environment the more petals are produced per flower. The following experiments and observations will prove this. I shall first refer to an observation for which un- fortunately I can give no numerical corroboration, but which may throw some light on the independence of the character of the flowers, of the order of branching. In the summer of 1892 when I examined all the flowers of my culture, and recorded the number of their petals twice a week, I was struck by the fact that the high numbers fell on particular days whilst on other days only low or intermediate numbers were observed. This would seem to indicate that during the development of the flowers in May and June pleiopetaly is influenced by weather conditions, in such a way that flowers which are in the susceptible period of their development during fine weather will produce more petals, quite independently of the order of the branch which bears them. This conclusion is supported by another set of obser- vations. In September 1892 the flowers, on the whole, produced more petals than they did in August of the same year. Or, to be more accurate, the number was greater Ranunculus Biilbosus Seniiplenns. 259 on those plants which opened their first flower in Sep- tember, than on those which had already begun to flower in July and August. The number of individuals of the former group was 77; they produced 1134 flowers du- ring the period ending with the beginning of November, when I stopped recording. In the other group there were 295 plants which flowered, and they produced 4425 flow- ers. The distribution was as follows: 1 Petals: 5 6 7 8 9 10 11 12 13 14 15 16-31 Oldest plants. 409 532 638 690 764 599 414 212 80 29 18 20 September plants: 40 52 126 165 204 215 177 104 35 8 4 Fig. 53. Ranunculus bulbosus sctniplcinis. A, curve of the plants flowering in August ; S, curve of those flowering in Septemher. The figures at the base refer to the num- ber of petals per flower. These figures are exhibited graphically in Fig. 53 ; they have been reduced for convenience of comparison so that the numbers in the two groups are about the same. The apex of the curve of the early flowering plants is over the 9 ; it is the same curve which has already been given in Fig. 51 on page 249 for the year 1892. The other curve has its apex over the 10, and also remains above the other curve in the right half of its course. The cause of this difference can only lie in the re- tarded germination. Either the seeds which germinate later are intrinsically more productive of pleiopetalous 1 See above, p. 250 ywd Fig. 5: v 1892). 260 Non-1 'salable Races. flowers 1 (like the small, late-germinating seeds of the crimson clover), or germination in the height of the summer in better and particularly in warmer weather favors development in such a way that the flowers are richer in petals; for the plants which flowered in July and August, germinated for the largest part during the cold and unfavorable weather experienced in May shortly after they had been sown. I first made an experiment to determine the influence of nutrition on pleiopetaly in 1890. I had wintered the selected plants of 1889, and in March transplanted half of them on a bed of pure sand, and the other half on a bed of ordinary garden soil. Only two-thirds (i. e., 12) of the plants of the former lot flowered, whilst all of the latter did. On the sandy bed I counted the petals of all the flowers and about twice their number on the control bed by simply picking off all the open flowers on alternate days. I examined in all 75 and 147 flowers respectively. The following is the result reckoned in percentages for convenience of comparison : Number of petals: 5 6 7 8 9 10 On the bed of sand- 73 23 400 On garden soil: 53 26 14 5 1 1 The plants on the better soil produce distinctly fewer five-petalled and more 7-10 petalled flowers. It is per- haps permissible to conclude from this that the steep drop of the curve from the wild locality, where the soil was sandy is, to a large extent at any rate, due to low nutri- tion. For presumably the same plants would exhibit a higher degree of pleiopetaly if grown on better soil and 1 With regard to this, it would be of great interest to find out in this and other plants the degree of development of the anomaly in such individuals which do not germinate until two or three years after the sowing of the seed Ranunculus Bulbosus Semiplenus. 261 so give rise to a less steep curve, just as in the experiment under consideration. I made a corresponding experiment in the summer of 1891, on the effect of manured and unmanured garden soil, with the race which was by that time considerably improved (Fig. 51 and page 250). The manuring was done with guano ; the two beds lay next to one another and were of the same size. On each was sown half of the harvest of several plants which had been very pro- ductive of pleiopetalous flowers in 1890. In the course of the summer 1 59 flowers on the unmanured bed opened and were recorded and 376 on the manured. The rela- tion between these two numbers is the best measure of the effect of the manure. The results, reckoned in percent- ages, are as follows: Petals: 5 6 7 8 9 10 11 12 13 14 Without manure: 12 15 25 21 12 10 3 1 1 With guano: 14 15 17 21 14 9 4 3 2 1 Without manure the apex of the curve was over the 7 and there were very few flowers with more than eleven petals : with manure the apex was over the 8, and there were distinctly more pleiopetalous flowers. In both the above experiments the control material consisted of other individuals than those used for the experiment itself. It is possible, however, to subject the same plant alternately to favorable and unfavorable in- fluences, and when this is done the same result is ob- tained as in the previous cases. With this object I trans- planted a series of the best plants of 1892 to a very dry bed in the spring of 1893. I left them there, and did not water them although the weather was continually dry. They suffered visibly under this treatment and some of them even produced fewer flowers than in the 1892 1893 1892 1893 No. 1 25 14 11 9 No. 2 43 19 9 5 No. 3 9 14 10 6 No. 4 44 5 8 5 No. 5 12 18 10 8 No. 6 16 21 9 8 262 Non-Isolablc Races. previous summer. I have a record, which has been al- ready alluded to, of the number of petals of all the flowers of each of the plants of 1892 ; these were recorded in the same way in 1893. But I only give here the mean numbers of petals per flower. PLANTS NUMBER OF MEAN NUMBER OF nwwKVKCV* FLOWERS PETALS PER FLOWER 2 4 4 3 2 1 The anomaly was thus diminished on every single plant as the result of transplanting to dry earth. The results of all these experiments prove that the production of more than five petals in a flower is inde- pendent of the position of this flower on the plant, but on the other hand is dependent in a high degree on the external conditions under which the particular flower passes its early stages, i. e., the most susceptible period of its existence. The number of petals varies directly with the vigor of the plant, the moisture and richness of the soil, the warmness of the weather and even the amount of sunshine during this susceptible period. Cultivation in the garden is therefore bound to con- vert the steep half curve of the wild locality (Fig. 51 for 1887) into a flatter one which will gradually extend to higher numbers of petals and will ultimately develop a new apex. This process, however, takes place more conveniently and more certainly, if the cultivation is combined with selection (see the same figure). The latter process picks Ranunculus Bnlbosus Serniplenus. 263 out the plants which manifest the anomaly most abun- dantly and most strongly ; these must, however, according to the facts given, as a rule, be the best nourished ones, i. e., the most favored by their environment. l ; '>r on the same bed, even if it has been uniformly prepared with the greatest care, the conditions under which neighboring plants grow are often very different. One seed may germinate in a place in which moisture is better retained ; another may germinate in almost dry soil. Some germi- nate on warm and fine days and are in consequence ahead of their less favored brothers for their whole lives; and so on. 1 And so it is that the several plants from seeds of the same seed-parent sown on the same day and on the same bed, are necessarily exposed to diverse condi- tions of life. Amongst them selection picks out the best and therefore, at least as a general rule, the most highly nourished ones. Selection, so to speak, only precipitates the operation of these external factors ; as w r e have pointed out before in connection with Papa e'er somniferum poly- cephalum. 2 Selection and cultivation have, therefore, worked in the same direction in my experiment for four genera- tions. They have about doubled the mean number of petals per flower, having brought it, in fact, to 9-10; they have produced, amongst several hundred plants and several thousand flowers, no more than three flowers with more than twenty petals (C 21, C 23 and C 31), i. e., not essentially more than would be expected according to OUETELET'S law from the actual mean and the amplitude of variation. These flowers occurred perfectly fortui- tously on plants which were not particularly favored oth- erwise, the means of the curves being only 10 for each 1 See Vol. I, p. 138. 2 See Vol. I, p. 140. 264 Non-I salable Races. of the three plants. We are thus justified in concluding that by the selection of these plants as seed-parents the mean of the race might further be slightly improved during the course of some years, but that these extreme variants afforded no more hope than did the others, of the attainment of the double race. Cultivation and selection cooperate in the direction of the desired end; they lead the half race measurably fur- ther on this line, but it is not through them that the object can be attained. The half race remains a half race, in spite of every effort and care, the semi-latent character expresses itself oftener and oftener, but it does not succeed in becoming the equal of the normal active characters, i. e., in constituting the mean character of a new race, independent of the continuance of selection and favorable cultural conditions. To arrive at this result a process of an entirely differ- ent nature is evidently required. According to the cur- rent theory of selection the goal would be reached if the experiment could be continued for tens or hundreds of years. But the course of the experiment we have de- scribed does not support this view ; it shows, on the con- trary, that all that can ever be gained by nutrition and selection has already been secured in these five genera- tions. The actual result is the production of an elite race which has a mean number of 9 petals in the flowers, under the favorable conditions of culture which obtained ; and gives rise, according to environmental conditions, on the one hand, to better variants (with a mean of about 11-13, or perhaps a few more, petals) while, on the other, it throws off atavists with a half GALTON curve (see Fig. 52 on page 252). It is my opinion, howeve r , that if the culture of the Varie gated Leaves. 265 half race were still continued, the double race would some day appear quite suddenly, and that it would then, after a short but sufficient isolation, persist as a constant, though highly variable, race. 1 24. VARIEGATED LEAVES. Variegated plants have long been great favorites in the garden, and their great instability has contributed largely to the development of the horticultural concep- tion of a variety, for the variations in their color pattern are practically unlimited. Hardly any two leaves are alike, and many species have a whole series of dappled and flecked varieties. They also possess the striking property of continually and conspicuously reverting to the species to which they belong. Such reversions occur either amongst seedlings or as bud-variations, and since on shrubs and trees these latter often remain for many vears and not rarely in more than one instance on the * / same plant, they can be seen by every one. In this way these bud-variations have come to be regarded as a suf- ficient proof of the idea that varieties are derivative and unstable structures, which always tend to revert to their parent species. Especially in the first half of the eighteenth century were plants with speckled and striped leaves very much sought after. 2 About that time the well-known English gardener THOMAS FAIRCHILD possessed more than one hundred varieties of them in his garden, and afterwards SCHLECHTENDAHL published a list from which it can be seen that variegation is distributed over the whole *!. e., as an eversporting variety with a wide amplitude of varia- tion which however would not alter in the course of the generations. "MEYEN, Pfiansen-Pathologie, 1841, p 282. 266 Non-Isolable Races. vegetable kingdom and occurs in all the larger groups and especially in most families of flowering plants. 1 At that time some of the most widely cultivated forms were the ribbon grass, Phraguiitcs arundinacca I'aricgata, and the variegated holly, Ilex Aquifoliiiin. Both are still much grown in gardens, the ribbon grass being rela- tively uniform, the holly highly variable. Of the latter there is a variety with white-edged leaves, besides the ordinary one with flecked leaves. Phragmites is differ- ent in many respects from genuine variegated plants and is much less variable in its character. The Ilc.r, however, is highly variable and often bears green shoots which may soon supersede the others on account of the greater facility with which they can obtain nourishment. A fine variegated bush of this species, or of any other, may be- come entirely green, whenever the green branches are not cut away every year. Thus it is probable that many specimens of the holly, which are now quite green, were originally variegated and were bought and planted as such. On closer examination we often find on them an occasional variegated twig which proves the correctness of this supposition. This is also the case with the horse chestnut, of which many older trees still living were planted at a time when the variegated variety was in special favor. Since then their foliage has become green and their original character is no longer seen. But an occasional checkered branch, or even the numerous small twigs with white leaves along the main stem, betrays the original variegated condition of the specimen. In the same way many cases of single variegated twigs on green bushes and trees are not to be regarded as the indi- 1 SCHLECHTENDAHL, Liiuiaea, 1830, V, p. 494. Very little seems to be known about variegated mosses and thallophytes. Variegated Leaves. 267 cation of something new but as a reminiscence of times long past when these varieties were in general favor. Variegation is classified under several headings. In the first place there are the yellow and the white varieties. In the former the chlorophyll is only insufficiently pro- duced, but in the latter even the xanthophyll or carotin is lacking; 1 and a more or less abortive development of the chloroplasts is usually correlated with the absence of these pigments. 2 Further we distinguish marginate, flecked and striped sorts. The former seem to constitute a variety for them- selves and are much rarer than the latter; they appear to be good races, that is, to be as constant as any ordinary garden variety, but I shall have little to say about them in this part. The most characteristic and best known example of them is the white bordered holly to which we have alreadv referred. 3 / Whether a plant is flecked or striped depends as a rule on the mode of venation of the leaves. Many varie- gated monocotyledons have striped leaves (Agave, Con- vaUaria majalis, Phormium tenax, Tradescantia repens, etc.) whereas the dicotyledons are usually flecked or streaked. The incomplete development of the chlorophyll ob- 1 See T. TAMMES, Ueber Carotin, Flora, 1900. 2 For further information on this point see the elaborate ana- tomical studies of A. ZIMMERMANN, Ucber die Chromatophoren in panachirten Slattern, in Beitrage zur Morphologic und Physiologic der Pflanzenzelle, Heft II, 1891, pp. 81-111, and Ber. d. d. hot. Ges. VIII, 1890, p. 95. Also H. TIMPE, Beitrage zur Kenntniss der Panachirung, Inaug.-Diss., Gottingen, 1900. 5 Marginate forms are commonly supposed by gardeners to be more stable than flecked ones. This fact was noted by MORREN in 1865. (Heredite de la panachure, Bull. Acad. roy. Belg., T. XIX. 2d series, p. 225). VERLOT however maintains the opposite opinion (Des Varie- tes, 1865, p. 74). For information relating to variegated varieties of Ilex see FOCKE, Abh. d. Naturw. Ver. zu Bremen, Vol. V, pp. 401-404. 268 Non-Isolable Races. viously results in an insufficient assimilation of carbonic acid gas. Thus the variegated parts grow less vigor- ously and are less resistant than the corresponding green ones. The Cy perns alter nifolius of our greenhouses, the Aspidistra elatior and a number of other favorite varie- ties show this clearly. Arundo do mix often attains a height of three or more meters whereas its striped vari- ety is scarcely half that height. Leaves of the variegated Aspidistra very often have one of their longitudinal halves green, but the other colorless. In such cases the leaf is distorted owing to the insufficient growth of the colorless half. The same thing happens in many other cases. The yellow leaves and parts of leaves, however, are not entirely without the green coloring matter, nor wholly without the power of assimilation. Most of them give a green extract when put into alcohol, and if examined under the microscope patches of green tissue can be found here and there, especially near the veins. The power to sustain life, however, is often lacking and the leaves die shortly after their growth is completed. Therefore, a high degree of the anomaly is not in favor, because the plants which possess it often become disfigured by the edges of their leaves turning brown. Many plants in which the variegation has gone too far die in their very early stages, while others have not sufficient strength to flower and bear seed. This latter circumstance is of special interest because it follows that plants with a high degree of variegation as a rule can have no part in the propagation of the variety. 1 In the opinion of some 1 It is perhaps scarcely necessary to state that these remarks do not apply to brown and purple leaves or those with red spots. For information on this point see STAHL'S excellent article Ueber bunte Laubbldttcr, Ann. Jard. Bot. Buitenzorg., Vol. XIII, Pt. 2, 1896, p. 137. Variegated Leaves. 269 authors another fact is connected with this, viz., that varieties which have both variegated leaves and double flowers are much rarer than would have been expected from the prevalence of these two anomalies in horti- culture. 1 In variegated plants, as is well known, not only the leaves are flecked. Their stems and calices are also often variegated, and the same is true of the fruits (pears, grapes, the siliquae of cabbage, Barbarea vulgaris, Chei- ranthus Chciri, Alyssum maritinium, Acer, Ilex, Aego- podiuin, Ligusticum, etc.). 2 I have also sometimes found galls on variegated oaks to be variegated, especially in the case of the beautiful orbicular galls of Cynips Kollari. I shall now proceed to the important question of the inheritance of this abnormality or the degree of fixing as it is usually called. As already stated I shall exclude from consideration the white-flecked 3 and the marginate forms of variegation, and shall confine myself to the ordinary cases of yellow variegated leaves. I shall give the numerical proofs of my conclusions later, and shall now proceed to deal with the question whether variegated sorts are half races or intermediate races (see Chapter II of this part). In my opinion the great majority of the variegated garden varieties are intermediate races, as for instance Barbarea vulgaris: whereas wild plants which occasion- ally present this character represent half races. Their 1 B. VERLOT, Sur la production et la fixation des varictes dans Ics Mantes d'ornement, 1865, p. 75. Also MORREN, Hercditc dc la pana- chure, loc. cit., p. 226. 2 MORREN, loc. cit., p. 233. 3 1 have not myself made any observations on this phenomenon (Albicatio, Albinismus} and the published records of it are very scanty. The fine white-variegated Humulus japonicus variegatus would be well worth experimenting with. 2/0 Non-Isolable Races. multiformity and instability corroborate this view. It is only the commonness of variegated sorts and the great interest which attaches to them which brings them to be regarded as analogous to the best constant varieties. o o Moreover this view is supported by the general opinion that a complete development of the yellow color would characterise the supposed constant variety, but that it would at the same time of necessity lead to the destruc- tion of the plants. In this conception variegation is re- garded as an incomplete anomaly whose complete con- dition would involve its own destruction ; but this view is incorrect. 1 Complete yellow varieties are not only pos- sible and capable of existence but actually well known in horticulture, although the number of such forms is small. Instances can be found in seedsmen's catalogues; e. g., Sainbucus uiyra aurca and Fra.vinus c.rcclsior aurea, also the aurca varieties of Chrysanthemum carinatum, Mira- bilis Jala pa, Scabiosa atropurpurca, Hiunulus japomcus (liitcsccns) etc. These plants, so far as I know, are all either yellowish-green or golden-yellow. 2 They also ap- pear to be very constant and never or very seldom to revert to the green type. I have made a number of ex- perimental sowings on a large scale of the seeds of the ordinary golden-yellow variety of Chrysanthemum Par- /