breeders, that this segregation of characters took place during the reduction division. At that time breeders, just as definitely as now, planned experiments in hybridizing different varieties or species to secure certain recombinations of desired characters in the hybrids. The experiments in citrus hybridization conducted by Mr. W. T. Swingle and the speaker were planned in 1893 entirely on this basis, yet the principle was in no sense of the word original with us, but was at that time well understood by all practical breeders. This understanding, the speaker thinks, was largely derived from the investigations of Naudin, though various investigators contributed to it.
With a full understanding of the knowledge and practises of the breeders of two decades ago, it must be admitted that the conception of unit characters and Mendelian segregation was necessary to clarify this knowledge and bring out the latent possibilities of the material presented by nature for the use of the breeder, and it is doubtful whether we even yet adequately comprehend the almost infinite possibilities open to us.
To understand breeding to-day we must clearly understand the conception of unit characters. We no longer conceive the species, race or variety, as a fixed ensemble of characters. Following De Vries, we now commonly conceive the species or variety to be made up of a certain number of unit characters, that are in large measure associated together by the accident of evolution or breeding and which are separable entities in inheritance. We may liken these unit characters to bricks used in the construction of a building, each separate and yet dependent on the others for the maintenance of the structure; as each unit character is dependent on the other unit characters for the maintenance of the plant body. We may think of these unit characters as organic elements similar to chemical elements, that by their recombination through hybridization, form new compounds—new plants—of distinctly different appearance, but which in turn do not affect the unit characters, which may again be separated and led to form other compounds, again resulting in distinct organisms. Related species may possess many distinct unit characters, but ordinarily would be expected to possess many similar unit characters. Cultivated races or varieties ordinarily would differ only in a few unit characters, and difference in a single unit character would be sufficient to give a distinct and recognizable race or variety. Indeed, the difference between two varieties of a single unit character might mean that one variety would be exceedingly valuable and the other practically worthless. De Vries asserts that unit characters are discontinuous in inheritance and do not exhibit transitional forms. A plant can not be hairy and at the same time smooth, or a fruit yellow and at the same time red. While there is yet much difference of opinion on these questions the preponderance of evidence certainly favors the unit character conception.
If, then, we recognize that species are made up of unit characters and that different species differ in the possession of different unit characters, the great problem in the evolution of species becomes the question of how the new unit character is acquired. Have all unit characters existed from the beginning, or are new unit characters being continuously acquired? A few years ago we supposed that new characters, if acquired in any form, must be seized upon, as it were, by natural selection and preserved, or otherwise that they would be swamped by intercrossing and lost. We now know from Mendelian analysis that
