246 Non-Isolahlc Races. 



must be brought about suddenly, and, under ordinary 

 conditions of culture, be effected in the course of a few 

 years. In this way the double variety may have arisen 

 from time to time in the wild state; and in the same 

 manner the present half race may perhaps, in the course 

 of time, undergo this change. 



This transformation, how^ever, cannot be simply the 

 result of careful selection. A mutation is needed; and 

 we know as little about the causes of mutations as about 

 the method of inducing them artiiicially. Mutations are 

 known to occur wn'th moderate frequency both in breed- 

 ing experiments and in nature, but, up to the present, 

 their occurrence has been a matter of chance (§§10 and 

 11, pp. 95-103). 



In my experiment such a mutation did not occur, 

 although it extended over five generations.^ The half 

 race was distinctly improved by repeated and very strin- 

 gent selection. It became at the end very rich in extreme 

 or almost extreme variants, but it was just in these that 

 it proved to be so remarkably constant. In its five gen- 

 erations it reached a point which did not seem to me 

 likely to be exceeded by further selection. It produced 

 occasional flowers with more than 15 petals, and a single 

 one with 31, but the mean number of the petals in its 

 selected individuals did not exceed 9-10. 



The double variety did not arise from it, in spite of 

 everv effort. 



* The fluctuating variability of the semi-latent character m Ranun- 

 culus hulbosus sonif^lenus seems to cover a much wider range of 

 forms than in Trifolium. Tlicre the extremes are 3 and 7 leaflets ; 

 in the buttercup they are 5 and 31 and perhaps more petals. From 

 this it does not. however, follow that the variation is greater in the 

 one case than in the other, but only that the variation is expressed 

 by a larger number of divisions in the latter case, i. e., that there are 

 more scale characters in the curve. 



