No. 613] EXPERIMENTS CONCERNING EVOLUTION 41 



of this kind will lead to physiological impossibilities, 

 monsters, etc. Here, then, is again the limit for selection 

 of factorial quantities. It need hardly be added that such 

 selection is necessarily orthogenetic. 



Our own experiments in this line are, as far as they go 

 at present, in perfect accord with Castle's work. We have. 



namely, selection in F^ If a' given pair of multiple a! 

 lelomorphs differs in regard to the quantity of the fac- 

 torial substance and this quantity is subject to fluctuation 

 around a mean, the variability of the character in F l is 

 caused by the usual agencies producing fluctuations as 

 well as by the different combinations of the parental 

 quantitative values. Selection in F x ought, therefore, to 

 influence the curve in F 2 in a certain number of cases, 

 namely, when the plus or minus individuals are genetically 

 plus or minus. Within the normal segregation of light 

 and dark individuals in the 3:1 ratio a shifting of the 

 mean for lightness and darkness must take place. In a 

 series of such experiments we had a number of positive 

 results. The following Table VIII may serve as an ex- 



TABLE VIII 



We believe that these facts and interpretations have a 

 definite bearing on the problem of evolution. The first 

 step in the differentiation of species which occurs in na- 

 ture seems to be the formation of geographic races. The 

 entire bulk of modern evidence in ecology tends to show 

 the existence of clearly defined local forms for very re- 

 stricted areas. For example, the ichthyologists differ- 

 entiate forms of Salmonids and Coregonids for prac- 

 tically every river and lake; in the same way in the lower 



