No. 543] FACTORS AND METHODS OF EVOLUTION 137 



sperm, such a sport must, in nature, be eliminated. 

 Under domestication it is continued by trimming away 

 the feathers that cover the vent. Similarly, winglessness 

 in male fowl renders copulation difficult because the 

 wings serve the cock as balancers while treading the hen. 

 These then are examples of characteristics that must be 

 eliminated in nature. In the case of certain striking 

 colors in poultry there is evidence that they are selected 

 against; their possession gives their owner a handicap. 



On the other hand certain new characteristics of fowl 

 may be preserved because apparently they offer no handi- 

 cap. Thus in the rumpless fowl the oil gland is absent 

 and the birds seem to be none the worse on that account ; 

 their plumage is bright and quite as resistant to a wet- 

 ting as that of birds with an oil gland at the base of the 

 tail. The striking fact that our experimental work yields 

 is the great number of new characters that seem to bear ■ 

 no relation to fitness or unfitness, but are truly neutral. 

 Thus I can not find that polydactylism, shank-feathering 

 or its absence, and the lower grades of single, pea and 

 rose comb have any adaptive significance for poultry. 

 One can invent adaptive explanations for them or then- 

 absence in birds, but there is no reason for thinking that 

 the explanations are significant. On the other hand, 

 there is accumulating considerable experimental support 

 for Darwin's theory of sexual selection; but of this it is 

 early to speak. On the whole, I think it may be fairly 

 said that experimental work supports the principle of 

 selective elimination but finds many characters that are 

 wholly neutral. 



To sum up, modern experimental study of heredity has 

 given-a new formulation to the problem of evolution and 

 has given definite data on the method of evolution. It 

 formulates the problem of evolution as the problem of the 

 nature and origin of the germinal determiners ol char- 

 acters. It has shown that, for the most part, the new 

 determiners arise one at a time and are independent r. 

 one another, may occur in any combination and may be 



