EVOLUTION 439 



as the case may be, with regard to any organ or character. 

 Now if fertility be correlated with this character, we 

 should expect some values of the character to be more 

 fertile than others, and accordingly the given distribution 

 could not possibly reproduce itself in the next generation. 

 If fertility varies from individual to individual or from 

 pair to pair, being any function of their characters, then 

 no form of life can be stable ; without a selective 

 death-rate periodically reducing the new generation to a 

 distribution absolutely like to that of the old, the type 

 must progressively change. This selection by relative 

 reproductivity, I have ventured to term reproductive 

 selection, or acting on a suggestion from Mr. Francis 

 Galton, genetic selection. Those who deny natural 

 selection must, if we can show the existence of genetic 

 selection, still admit that evolution is a reality, for genetic 

 selection produces a continuous change of type which 

 can only be held in check by more or less stringent 

 selection. 



Examining the point a little more carefully, let us 

 suppose the organ or character described by the numerical 

 value a to be most favourable to fertility ; organs deviating 

 from a either in excess or defect will belong to less fertile 

 individuals, while extreme deviations will be sensibly 

 sterile. Now three cases are possible : — 



(i.) The organ or character a falls actually outside the 

 range of organs found in the given race. 



(ii.) The organ or character a falls within the range 

 observed for individuals of the given race, but does not 

 coincide with the modal value of the organ or character, 

 say the value c. 



(iii.) The organ or character a coincides with the 

 modal value c. 



Supposing in the first case the more character the 

 more fertility, we mark a positive correlation between the 

 character and fertility. The individuals with more of the 

 character will be relatively more fertile than those with 

 less, and accordingly whatever be the original distribution 

 of frequency the mode c must progress in the direction 



