462 



TEE AMERICAN NATURALIST [Vol.L 



capacity for forming such pigment. If we select parents 

 because they have medium-brown hair we get offspring 

 with blond or golden or red hair even, and it would take 

 a long time to get a pure brown-haired race by this 

 method of selecting. If the trait that we are trying to 

 improve is very sensitive to environment, then the somatic 

 conditions will be a very bad index of the germinal ; but 

 if not sensitive, the somatic may be a good index of the 

 germinal. Thus, since in the fruit fly the number of 

 bristles varies closely with food conditions, selection of 

 breeders merely on the somatic state will not lead to 

 much, if any, genetic progress. 



The question of the potency of selection in nature 

 comes back to the matter of value of the soma as an index 

 of the germ plasm. If blue eye color affords insufficient 

 protection from tropical conditions elimination of the 

 blue-eyed individual kills off blue-eyed plasm and with 

 each death the race is rapidly purified. But there is a 

 limit to the process because brown eyes are preserved 

 equally whether the germ plasm does or does not carry 

 the blue-eyed condition. As there are twice as many 

 simplex as duplex brown eyes, the complete elimination 

 of the blues will not be brought about quickly. 



When, therefore, we hear a breeder say: "I made this 

 character by selection" what he really means is : Somatic 

 variations in the desired direction were afforded; there 

 was a large correlation between somatic and germinal 

 conditions, so that I was able, merely by choosing as 

 breeders individuals showing the desired trait somati- 

 cally, to get a race with the determiners of the character 

 pure, or practically pure, in the germ plasm. 



One other difference of opinion there seems to be be- 

 tween selectionists and the others. Castle evidently 

 doubts if factors for characters are always discrete and 

 do not change. He is inclined to hold that there may be 

 genes which vary pari passu with a variation of a "unit 

 character" and in somewhat the same degree. This 

 view seems to be quite in accord with an expectation that 



