This seedling in the various species of clover usually begins with a first leaf above the cotyledons of a different structure from those that follow. It has only one blade instead of three. But in my variety the increase of the number of the leaflets may extend to these primary organs, and make them binate or even ternate. Now it is obvious that an individual, which begins with a divided primary leaf, will have a greater tendency to produce a large number of supernumerary leaflets than a plant which commences in the ordinary way. Or in other words, the primary leaves afford a sure criterion for the selection, and this selection may be made in the seed-pans. In consequence, no young individual with an undivided primary leaf was planted out. Choosing the 20 or 30 best specimens in the seed-pan, no further selection was required, and the whole lot could be left to cross-fertilization by insects. The observation of this distinguishing mark in the young seedlings has led to the discovery of another quality as a starting-paint for further [348] selection. According to the general rule of pedigree-culture, the seeds of each individual plant are always saved and sowed separately. This is done even with such species as the clover, which are infertile when self-pollinated, and which are incapable of artificial pollination on the required scale, since each flower produces only one seed. My clover was always left free to be pollinated by insects. Obviously this must have led to a diminution of the differentiating characters of the individual plants. But this does not go far enough to obliterate the differences, and the selection made among the seedlings will always throw out at least a large part of those that have suffered from the cross. Leaving this discussion, we may inquire closer into the nature of the new criterion afforded by the seedlings. Two methods present themselves. First, the choice of the best seedlings. In the second place it becomes possible to compare the parent-plants by counting the number of deviating seedlings. This leads to the establishment of a percentage for every single parent, and gives data for comparisons. Two or three hundreds of seeds from a parent may easily be grown in one pan, and in this way a sufficiently high degree of accuracy may be reached. Only those parents that give [349] the highest percentage are chosen, and among their progeny only the seedlings with trifoliolate primary leaves are planted out. The whole procedure of the selection is by this means confined to the glasshouse during the spring, and the beds need not be large, nor do they require any special care during the summer.