D4 



THE AMERICAN NATURALIST 



[Vol. L 



Castle's that he regards that evidence as totally worth- 

 less. It has not so appealed to other workers. 4 Further- 

 more I think it can be shown that methodologically my 

 treatment of the problem of inheritance of fecundity 

 stands on precisely the same plane as Mendelian work in 

 general, and Professor Castle's Mendelian work in par- 

 ticular. This I shall now try to do. 



The essence of a test of a Mendelian hypothesis lies in 

 this : the genetic constitution of the parents of an array 

 of offspring necessitates that the individual offspring 

 bearing different segregating characters, or different 

 segregating categories of the same character, shall occur 

 in definite numerical proportions. If the observed nu- 

 merical proportions of the offspring agree, within the 

 limits of error due to random sampling, with the propor- 

 tions expected from the Mendelian hypothesis, then this 

 fact constitutes valid evidence in support of the hypoth- 

 esis. If no exceptions to this rule appear and a sufficient 

 number of agreeing cases are adduced the hypothesis is 

 regarded as demonstrated. The number of cases neces- 

 sary to constitute a proof is a purely individual matter. 

 What one person will consider sufficient to establish proof 

 another will not. 



.Vow in the case of fecundity in fowls, Pearl and Sur- 

 face 5 first established that the Barred Plymouth Kock 

 stock at the Maine Experiment Station was not homo- 

 zygous in respect of winter egg production, but that it 

 contained, with frequent occurrence, individuals of high 

 fecundity, and also individuals of low fecundity. The 

 race not being homozygous with respect to fecundity, it 

 was possible to test the Mendelian inheritance of this 



*Cf., for example, Morgan, T. H., "Heredity and Sex," New York, 1913, 



The Determination of 



ion of Sex," Cambridge, 1914, and Jo- 

 .kten Erblichkeitslehre," Zweite Ausgabe. 



Data on the Inher 



