No. 519] THE EFFECTIVENESS OF SELECTION 



I II 



with races of complicated descent; they plunge us at once 

 into all the difficulties due to interweaving, blending and 

 transfer of characters from one genotype to another. 

 But if we stick closely to the general propositions already 

 stated, we shall have a guide. MacCurdy and Castle 

 got by selection all sorts of conditions lying between the 

 extremes with which they started. But did they get any- 

 thing lying outside these extremes, as would be required 

 in order to show that we can by selection make evolu- 

 tionary progress ? As I read their results, they did not. 

 Their experiments are most important for many prob- 



us evidence that methodical selection can produce any- 

 thing beyond combinations of what already exists; hence 

 they do not help us in getting from Amceba to man. 



The work of the German breeders who have for years 

 practised methodical selection for improvement of agri- 

 cultural races clears up at once under the genotype idea, 

 as the analyses of Fruwirth, v. Kiimker and others show 



but what it does is to purify a contaminated race-a 

 process which, owing to the laws of inheritance, may re- 

 quire several generations. 



I have spoken only of those experiments which seem 

 at first view to show the efficacy of selection; brevity 

 requires me to pass without mention over investigations 

 which, while not carried on with pure lines, support and 

 reinforce the conclusions drawn from such work. Such 

 for example are the fundamental experiments of Tower, 

 the recent work of Pearl, of Shull, and many of the experi- 

 ments of de Vries. 



Thus far the dissecting knife of the pure line idea 



wSgsof seTectlom Am? one who uses it with pre- 

 cision will find what an important advance its exact 

 formulation by Johannsen marks over even such an 

 analysis as that given by de Vries. In the work of de 

 Vries, as in that of many recent authors, the selection 



