Inheritance of Variations 73 



marked. Measuring by the coefficient of correlation, we find 

 that when the parent differs from the average, the progeny 

 tend to inherit somewhere from one-tenth to three-tenths of 

 this peculiarity. That is, the race shows a slight tendency 

 to break up into several races, hereditarily distinct. (For 

 details see Jennings, 1916.) 



For the investigator who has searched in vain for years 

 to find in uniparental reproduction any tendency for a single 

 race to evolve into several, such faint indications are excit- 

 ing; here we begin to get hold of the beginnings of evolution. 

 Most of the grounds on which we believe that evolution 

 occurs are inferential ; we believe that it must have occurred, 

 in order to account for the diversities that we find now 

 existing. But can we actually see it occur? 



To carry our work farther, we begin to exercise selection 

 within the single family. On the one hand we select all the 

 long-spined individuals and place them together; on the 

 other hand we select all the short-spined ones and place them 

 together. In the long-spined group we continue to save for 

 generation after generation only the individuals that are 

 long-spined; in the short-spined group only the offspring 

 with short spines. In the same way we select other sets for 

 numerous spines and for few spines ; for large shells and 

 for small shells ; for many teeth and for fewer teeth. 



And now as we keep this up for generation after genera- 

 tion we find that the correspondence between parent and 

 progeny becomes more and more marked. We find that 

 our single family is breaking up into many different groups, 

 which differ from one another hereditarily. We get finally 

 what appear to be two diverse races, one with long spines, 

 the other with short spines (Figures 24, M and F), 

 the difference continuing for generation after generation. 

 A third set (L) has constantly large shells, while others 



