One of them produced during the ensuing summer one four-bladed and one five-bladed leaf. The seeds were saved separately and sown the following spring and the expected result could soon be seen. Among some 250 individual plants I counted 22 with one or two deviations, and 10 with from three to nine four- or five-bladed leaves. Proportions nearly similar have been observed repeatedly. Better nourished individuals have produced more deviating leaves on one plant, partly owing to the larger number of stems and branches, and poor or average specimens have mostly been without any aberration or with only one or two abnormal leaves. No further improvement could be attained. Quadrifoliolate leaves were always rare, never [360] attaining a number that would put its stamp on a whole bed. I have endeavored to get some six- and seven-bladed crimson clover leaves, but in vain; selection, culture of many hundreds of individuals, manure, and the best possible treatment has not been adequate to produce them. Of course I am quite convinced that a repetition of my experiment on a far larger scale would yield the desired types, but then only in such rare instances that they would have no influence whatever on the average, or on the improvement of the race. The eighth generation in the year 1903 has not been noticeably better than the second and third generations after the first selection. In comparing this statement with the results gained in the experiment with the red clover, the difference is at once striking. In one case a rich variety was isolated, and, by better treatment and sharp methods of selection, was brought up in a few years to its highest pitch of development. In the other case a very weak race was shown to exist, and no amount of work and perseverance was adequate to improve it to any noticeable degree.