EVOLUTION 441 



selective death-rate, can never give a stable new form, 

 which would continue without this selection unless the 

 type c has been made identical with the character a of 

 maximum fertility. Both are probably functions of the 

 environment, and if they do not coincide a balance must 

 somehow be maintained between them. 



It is impossible to proceed far with the theory of 

 evolution without being drawn up by still unsolved 

 problems in genetic selection. For example, we have 

 seen how with some form of sexual selection differentia- 

 tion of a local race is a possibility. But our account 

 of this process was given on the assumption that all 

 individuals, or all " mid -parents " (p. 421) are equally 

 fertile. But suppose this assumption not to be true ? 

 Then unless both the differentiated types become centres 

 of maximum fertility, there will, if change of environment 

 suspends periodic selection, at once set in a progressive 

 change of one or both types towards the character a of 

 maximum fertility of the undifferentiated race. It seems, 

 therefore, of utmost importance for biologists to deter- 

 mine : — 



(i.) Whether fertility is really correlated with any other 

 characters. Is genetic selection a reality? 



(ii.) If genetic selection is real, how does a differentia- 

 tion of the character a of maximum fertility take place, 

 when there is differentiation of type ? 



S 5. — On the Reality of Genetic Selection 



One of the most direct methods of ascertaining whether 

 genetic selection exists is to take, say, 1000 mated 

 individuals of one sex and find their type. Then take 

 the same thousand individuals and reckon them once for 

 each individual offspring of the same sex that they have. 

 If fertility be no function of the character under investiga- 

 tion, then it follows that the type will not be sensibly 

 changed ; if, on the other hand, it be, then there will be 

 a sensible change of type. Unfortunately, I have again 

 no sufficiently wide data, but the following statistics are 



