No.SSo] MODIFICATION OF CHARACTERS. 



563 



Anticipating the conclusions which will be reached in 

 this paper, it may be pointed out that the swamping effect 

 is not so serious a check upon progressive evolution as 

 might be supposed, (1) because blending or modification 

 of a new character only takes place in certain crosses and 

 may be accompanied by segregation even in some of 

 those, and (2) because Mendelian characters usually come 

 out "pure" when crossed with the form from which they 

 were derived. Hence when Mendelian characters arise 

 through mutations in nature it may be expected that they 

 will be able to perpetuate themselves and spread, espe- 

 cially when dominant, unless they place the organism at a 

 disadvantage in the struggle for existence. The modifi- 

 cation of a Mendelian character will come, not from cross- 

 ing with its parent form but with a more distantly related 

 species. 



Some writers appear to believe that it is practically 

 impossible to modify a unit character because it is repre- 

 sented in the germ plasm by a "gene" whose essential 

 characteristic is its unmodifiability. But if we consider 

 that each unit character is a difference which has arisen 

 through a change in one element of the germ plasm, prob- 

 ably in a chromosome, then it would seem possible that if 

 introduced into a foreign cytoplasm the chromosome may 

 become subject to permanent modification. 



Castle and Phillips (1914) have produced evidence 

 from hooded rats tending to show that selection may 

 modify a unit character in certain cases, although the 

 nature of this result is not yet fully analyzed. They 

 moreover show that the hooded character is modified by 

 a cross. Davenport (1906) in his experiments with poul- 

 try, concluded that unit characters are frequently modi- 

 fied by crossing. He says (p. 80) : 



