but he now admits ('Improvement of the Cereals' 1873) that this was an error.)
The following case is much more remarkable, and evidently perplexed Gartner, whose strong wish it was to draw a broad line of distinction between species and varieties. In the genus Verbascum, he made, during eighteen years, a vast number of experiments, and crossed no less than 1085 flowers and counted their seeds. Many of these experiments consisted in crossing white and yellow varieties of both V. lychnitis and V. blattaria with nine other species and their hybrids. That the white and yellow flowered plants of these two species are really varieties, no one has doubted; and Gartner actually raised in the case of both species one variety from the seed of the other. Now in two of his works (16/15. 'Kenntniss der Befruchtung' s. 137; 'Bastarderzeugung' s. 92, 181. On raising the two varieties from seed see s. 307.) he distinctly asserts that crosses between similarly-coloured flowers yield more seed than between dissimilarly-coloured; so that the yellow-flowered variety of either species (and conversely with the white-flowered variety), when crossed with pollen of its own kind, yields more seed than when crossed with that of the white variety; and so it is when differently coloured species are crossed. The general results may be seen in the Table at the end of his volume. In one instance he gives (16/16. 'Bastarderzeugung' s. 216.) the following details; but I must premise that Gartner, to avoid exaggerating the degree of sterility in his crosses, always compares the MAXIMUM number obtained from a cross with the AVERAGE number naturally given by the pure mother-plant. The white variety of V. lychnitis, naturally fertilised by its own pollen, gave from an AVERAGE of twelve capsules ninety-six good seeds in each; whilst twenty flowers fertilised with pollen from the yellow variety of this same species, gave as the MAXIMUM only eighty-nine good seeds; so that we have the proportion of 1000 to 908, according to Gartner's usual scale. I should have thought it possible that so small a difference in fertility might have been accounted for by the evil effects of the necessary castration; but Gartner shows that the white variety of V. lychnitis, when fertilised first by the white variety of V. blattaria, and then by the yellow variety of this species, yielded seed in the proportion of 622 to 438; and in both these cases castration was performed. Now the sterility which results from the crossing of the differently coloured varieties of the same species, is fully as great as that which occurs in many cases when distinct species are crossed. Unfortunately Gartner compared the results of the first unions alone, and not the sterility of the two sets of hybrids produced from the white variety of V. lychnitis when fertilised by the white and yellow varieties of V. blattaria, for it is probable that they would have differed in this respect.
Mr. J. Scott has given me the results of a series of experiments on Verbascum, made by him in the Botanic Gardens of Edinburgh. (16/17. The results have since been published in 'Journ. Asiatic Soc. of Bengal' 1867 page 145.) He repeated some of Gartner's experiments on distinct species, but obtained only fluctuating results, some confirmatory, the greater number contradictory; nevertheless these seem hardly sufficient to overthrow the conclusion arrived at by Gartner from experiments tried on a larger scale. Mr. Scott also experimented on the relative fertility of unions between similarly and dissimilarly-coloured varieties of the same species. Thus he fertilised six flowers of the yellow variety of V. lychnitis by its own pollen, and obtained six capsules; and calling, for the sake of comparison, the average number of good seed in each of their capsules one hundred, he found that this same yellow variety, when fertilised by the white variety, yielded from seven capsules an average of ninety-four seed. On the same principle, the white variety of V. lychnitis by its own pollen (from six capsules), and by the pollen of the yellow variety (eight capsules), yielded seed in the proportion of 100 to 82. The yellow variety of V. thapsus by its own pollen (eight capsules), and by that of the white variety (only two capsules), yielded seed in the proportion of 100 to 94. Lastly, the white variety of V. blattaria by its own pollen (eight capsules), and by that of the yellow variety (five capsules), yielded seed in the proportion of 100 to 79. So that in every case the unions of similarly-coloured varieties of the same species were more fertile than the unions of dissimilarly-coloured varieties; when all the cases are grouped together, the difference of fertility is as 100 to 86. Some additional trials were made, and altogether thirty-six similarly-coloured unions yielded thirty-five good capsules; whilst thirty-five dissimilarly- coloured unions yielded only twenty-six good capsules. Besides the foregoing experiments, the purple V. phoeniceum was crossed by a rose-coloured and a white variety of the same species; these two varieties were also crossed together, and these several unions yielded less seed than V. phoeniceum by its own pollen. Hence it follows from Mr. Scott's experiments, that in the genus Verbascum the similarly and dissimilarly-coloured varieties of the same species behave, when crossed, like closely allied but distinct species. (16/18. The following facts, given by Kolreuter in his 'Dritte Fortsetzung' ss. 34, 39, appear at first sight strongly to confirm Mr. Scott's and Gartner's statements; and to a certain limited extent they do so. Kolreuter asserts, from innumerable observations, that insects incessantly carry pollen from one species and variety of Verbascum to another; and I can confirm this assertion; yet he found that the white and yellow varieties of Verbascum lychnitis often grew wild mingled together: moreover, he cultivated these two varieties in considerable numbers during four years in his garden, and they kept true by seed; but when he crossed them, they produced flowers of an intermediate tint. Hence it might have been thought that both varieties must have a stronger elective affinity for the pollen of their own variety than for that of the other; this elective affinity, I may add of each species for its own pollen (Kolreuter 'Dritte Forts.' s. 39 and Gartner 'Bastarderz.' passim) being a perfectly well-ascertained power. But the force of the foregoing facts is much lessened by Gartner's numerous experiments, for, differently from Kolreuter, he never once got ('Bastarderz.' s. 307) an intermediate tint when he crossed the yellow and white flowered varieties of Verbascum. So that the fact of the white and yellow varieties keeping true to their colour by seed does not prove that they were not mutually fertilised by the pollen carried by insects from one to the other.)
This remarkable fact of the sexual affinity of similarly-coloured varieties, as observed by Gartner and Mr. Scott, may not be of very rare occurrence; for the subject has not been attended to by others. The following case is worth giving, partly to show how difficult it is to avoid error. Dr. Herbert (16/19. 'Amaryllidaceae' 1837 page 366. Gartner has made a similar observation.) has remarked that variously-coloured double varieties of the Hollyhock (Althea rosea) may be raised with certainty by seed from plants growing close together. I have been informed that nurserymen who raise seed for sale do not separate their plants; accordingly I procured seed of eighteen named varieties; of these, eleven varieties produced sixty-two plants all perfectly true to their kind; and seven produced forty-nine plants, half of which were true and half false. Mr. Masters of Canterbury has given me a more striking case; he saved seed from a great bed of twenty-four named varieties planted in closely adjoining rows, and each variety reproduced itself truly with only sometimes a shade of difference in tint. Now in the hollyhock the pollen, which is abundant, is matured and nearly all shed before the stigma of the same flower is ready to receive it (16/20. Kolreuter first observed this fact, 'Mem. de l'Acad. de St. Petersburg' volume 3 page 127. See also C.K. Sprengel 'Das Entdeckte Geheimniss' s. 345.); and as bees covered with pollen incessantly fly from plant to plant, it would appear that adjoining varieties could not escape being crossed. As, however, this does not occur, it appeared to me probable that the pollen of each variety was prepotent on its own stigma over that of all other varieties, but I have no evidence on this point. Mr. C. Turner of Slough, well known for his success in the cultivation of this plant, informs me that it is the doubleness of the flowers which prevents the bees gaining access to the pollen and stigma; and he finds that it is difficult even to cross them artificially. Whether this explanation will fully account for varieties in close proximity propagating themselves so truly by seed, I do not know.
The following cases are worth giving, as they relate to monoecious forms, which do not require, and consequently cannot have been injured by, castration. Girou de Buzareingues crossed what he designates three varieties of gourd (16/21. Namely Barbarines, Pastissons, Giraumous: 'Annal. des Sc. Nat.' tome 30 1833 pages 398 and 405.), and asserts that their mutual fertilisation is less easy in proportion to the difference which they present. I am aware how imperfectly the forms in this group were until recently known; but Sageret (16/22. 'Memoire sur les Cucurbitaceae' 1826 pages 46, 55.), who ranked them according to their mutual fertility, considers the three forms above alluded to as varieties, as does a far higher authority,