VARIATION AND MENDELIAN INHERITANCE 381 



pigmentation. Considering then, the fourteen remaining fig- 

 ures, representing characters in respect to which the parent 

 races differ unmistakably, we have— 



F2 greater than Fi in 8 cases 

 F2 less than Fi in 2 cases 

 F2 equal to Fi in 4 cases. 



Let us add that in two of the instances in the first of these 

 groups, the differences are scarcely larger than their probable 

 errors, while in only a few of the entire fourteen do they attain 

 anything approaching statistical certainty. 



I think it is plain, therefore, that the exponents of the 'mul- 

 tiple factor' hypothesis will derive rather cold comfort from the 

 figures which I have to offer, even though the table as a whole 

 may show a slightly preponderant increase of variability in the 

 second hybrid generation. Of course, one answer is obvious. 

 For studies such as these I have not chosen 'favorable' material. 

 In these wild races, it may be contended, the number of factor 

 mutations for each character has been so great that segregation 

 cannot be expected to manifest itself appreciably in these small 

 series. Such arguments are as unanswerable as they are 

 unconvincing. 



Experience warns me that another objection is likely to be 

 made to the validity of these results, although it is an objection 

 which I believe to be utterly irrelevant when brought in this 

 connection. It will be pointed out that each of the parent races 

 with which I am dealing is in reality not pure, but is a mixture 

 of genetically distinct strains. As evidence of this will be cited 

 the wide range of variability within each race, and the fact that, 

 for two characters at least, I have shown these variations to be 

 hereditary. Even if all this is granted, however, and the con- 

 tentions of the pure-line school be admitted in full, it still seems 

 to me inevitable that we should, on Mendelian principles, have 

 an increased variability in the F2 generation of hybrids between 

 two such mixed races. For there would in general be more fac- 

 torial differences between representatives of two geographical 

 races than between two individuals of the same race. The Fi 

 generation from such a cross would present more heterozygosis 



THE JOURNAL OF EXPERIMENTAL ZOOLOGY, VOL. 30, NO. 3 



