650 



THE AMERICAN NATURALIST 



[Vol. XLVI 



tremes in one generation, extremes that were never seen 

 when but few individuals were produced. This sort of a 

 thing is not hypothetical. It is mathematically demon- 

 strable that with the same variability (a + b) n expanded 

 gives an increase in the number of classes as the total 

 number of individuals increases. It is, moreover, sup- 

 ported by the experimental evidence of De Vries on se- 

 lecting for higher number of rows in maize. I, myself, 

 by using greater numbers obtained an increase in protein 

 in maize in one generation comparable to that obtained 

 by the Illinois Agricultural Station in three generations. 



Castle further argues that decrease in regression 

 toward the original mean supports his view. On the 

 other hand, this is exactly what should take place on as- 

 suming the truth of the fixed factor conception, as has 

 been shown by Jennings. 



Again, the selected races when crossed with wild races 

 both act as simple recessives, but the extracted plus in- 

 dividuals are more pigmented. This is what I should 

 expect. The extracted plus individuals would be more 

 pigmented when existing in small numbers, because the 

 modifying factors are several. If several thousand 

 progeny were grown, however, recombinations would 

 show a more varied result. And as a matter of fact, ex- 

 tracted recessives from the plus race are not precisely 

 comparable in their fluctuation to the selected race with 

 which the wild was crossed. They are more variable than 

 the progeny of an inbred hooded individual of the same 

 grade as the parent used in the cross with the wild. I 

 do not think that one has a right to say, therefore, that 

 there were no modifying genes present in various com- 

 binations in the extracted recessives. 



When the selected lines were crossed together, more- 

 over, the resulting progeny were somewhat intermediate 

 and variable. The grandchildren were more variable. 

 This is what should result from our assumptions. The 

 animals are homozygous as far as having a pattern is 

 concerned, but they differ in several genes that affect 

 the development of the pattern. 



