564 



THE AMERICAN NATURALIST [Vol. LV 



to too small size of the random sample could be eliminated by the 

 rearinf? of larger numbers ; the effects of the interplay of numer- 

 ous environmental factors could not be thus eliminated. 



Kioni rh.' i'oiN^ooing considerations it must be clear that Castle 

 IS iihoirri lirr too sanofuine as to the value of his method when 



Ir is p.'i hiips lint to be expected that results more than approxi- 

 mately correct, would be cfiven by this mctliod, unless fairly large 

 numbers of both L\ ami imlividuah have been studiedS' 



I believe the conclusion is justified that even a fairly large" 

 number of individuals of the and P, could not be expected 

 to give correct estimates of the number of factors interacting 

 in any case. As between tlie method of Castle in estimating the 

 number of hypothetical duplicate factors operating in any case 

 on the basis of the change they produce in standard devia- 

 tions, and the method of Punnett. of formulating a genotypic 

 situation on the basis of a small number of definitely weighted 

 factors, I am convinced that the latter method is much to be 

 preferred, even though it does not lend itself readily to a 

 "general treatment of blending inheritance." 



George II. Shull 



Dr. Siiri.i. lias kniilly snu iiic his luanuscript in advance of 



The difficult ics u liidi he thinks might be encountered in apply- 

 ing the method, whii h I have suggested, for estimating the num- 

 ber of genetic tactots involved in cases of blending inheritance, 



1. The possible uiHMiual iiiHiience of the several factors which 

 are responsible J'of a case of blending inheritance, making it 

 difficult to estimate their timuber from the total effect observed. 



2. The possibility that M.me of the factors may bo dominant 

 in character and others not. 



:i. The possibility that some may be positive in action and 



