THE AMERICAN NATURALIST [Vol. XL V 



it does not, the discrepancy is in favor of the pure linist. 

 But comparing these values with 1.000 ± .044 for "Vik- 

 toria" and 1.000 ± .046 for "Kapital," I think we must 

 admit that the evidence is strongly in favor of a differ- 

 entiation in these lines of peas. 



III. The Beabing of These Data upon the Genotype 

 Theory of PIeeedity 



To the conclusion that these pure lines of peas are 

 differentiated among themselves and that speaking 

 roughly their means "form a Quetelet's Curve," there 

 can be no objection, although this conclusion is by no 

 means justified by Roemer's own analysis of his data. 

 The assumption that these facts lend any support what- 

 ever to the genotype theory 1 " seems to me to rest not only 

 upon the most slipshod reasoning, but upon a complete 

 disregard of simple biological precautions. 



The chief of the pertinent reasons follow. 



First. There i.s no Evidence of Line or (idiotypic Con- 

 stancy or Heredity. 



By definition the genotype is a rigid organic entity, 

 distinguished by breeding true from generation to gen- 

 eration, with the exception of mutations which are com- 

 pletely inherited and fluctuations which are not inherited 

 at all. The mean of the line remains the same from 

 generation to generation (except for rises and falls due 

 to periodic environmental changes) ; the variations 

 around these means are absolutely incapable of inhent- 



Roemer and his supporters of course assume this to be 

 true for the Pisum series dealt with, but no fragment of 

 evidence is adduced to show either (a) that these means 

 remain the same from generation to generation, or (b) 

 that selection within the pure line is ineffective. The 

 condition is even worse than this. Roemer measured the 



