five generations, often in large numbers, selecting always those which had the highest number of petals, throwing out the remainder and saving the seed only from the very best plants. I got a strain of selected plants with an average number of nine petals in every flower, and found among 4,000 flowers four having 20 petals or more, coming up even to 31 in one instance. But such rare instances had no influence whatever on the selection, since they were not indicative of individual qualities, but occurred quite accidentally on flowers of plants having only the average number of petals. Now double flowers are widely known to occur in other species of the buttercups, both in the cultivated varieties and in some wild forms. For this reason it might be expected that through a continuous selection of [358] the individuals with the largest numbers a tendency to become double would be evolved. Such, however, was not the case. No propensity to vary in any definite direction could be observed. Quite on the contrary, an average condition was quickly reached, and then remained constant, strongly counteracting all selection. Such experiences clearly show that the same anomaly may occur in different species, and no doubt in strains of the same species from different localities, according to at least two different standards. The one is to be called the poor, and the other the rich variety. The first always produces relatively few instances of the deviation, the last is apt to give as many of them as desired. The first is only half-way a variety, and therefore would deserve the name of a half-race; the second is not yet a full constant variety, but always fluctuates to and fro between the varietal and the specific mark, ever-sporting in both directions. It holds a middle position between a half-race and a variety, and therefore might be called a "middle-race." But the term ever-sporting variety seems more adequate to convey a right idea of the nature of this curious type of inheritance. From this discussion it will be seen that the behavior of the crimson clover is not to be considered [359] as an exception, but as a widely occurring type of phenomenon, occurring perhaps in all sorts of teratologic deviations, and in wide ranges of species and genera. Hence it may be considered worth while to give some more details of this extended experiment. Ten years ago (1894-5) I bought and sowed about a pound of seed of the crimson clover. Among many thousands of normal seedlings I found two with three and one with four cotyledons. Trusting to the empirical rules of correlation, I transplanted these three individuals in order to isolate them in the flowering period.