272 



THE AMERICAN NATURALIST [Vol. LI 



through the accumulation of lesser variations by natural 

 selection. 



But Castle^^ still contends vigorously that a single ge- 

 netic character may undergo quantitative change under 

 selection, and if this be true the difference finally attained 

 by two forms differing only in one factor might follow 

 from the accumulation of an indefinite number of slight 

 variations. Still the complete abandonment of Castle's 

 position would not save Punnett's argument. For if re- 

 ported observations and inferences concerning Droso- 

 phila^^ are correct, a red-eyed strain has given rise to 

 white by mutation and this in turn to eosin, which being 

 crossed with the original red gave in the generation 

 offspring in the proportion of three red to one eosin. 



It is ciuite immaterial whether one explains these rela- 

 tions upon the hypothesis of multiple allelomorphs, or, 

 as Punnett^^ prefers, upon the assumption of complete 

 coupling of factors. In the former case one must admit 

 that by breeding experiments the end product of a series 

 of mutations can not be indubitably distinguished from 

 one that results from a single modification of the affected 

 factor. In the latter, one *ust grant that the occurrence 

 of the grandparental types in the offspring of hybrid par- 

 ents in the ratio of three to one is no proof that the grand- 

 parents themselves differed in respect to one character 

 alone, or that the difference between the two resulted from 

 changes occurring at one time in the germinal constitu- 

 tion of an ancestor of one of them. Hence, in so far as 

 tliis argument is concernAJ, there is in the case of P. 

 ludytes, for example, no assured reason for supposing 

 that the aristolocliia-Y\\.Q form did not attain its present 

 appearance by a series of steps, of which a number of the 

 later at least were preserved by virtue of the advantage 

 they conferred upon the individuals in which they ap- 



