ILLUSTRATION OF STATISTICAL RESULTS 333 



is evident. Thus Messrs. Rommel & Philipps (1906) have shown 

 in regard to Poland China hogs : (i) that there has been an 

 increase of '48 in the size of litter in the twenty years between 

 1882 and 1902, and (2) that the size of litter is a character trans- 

 mitted from mother to daughter. " It would appear proved that, 

 by judicious selection for breeding purposes of sows from large 

 litters, the average for the breed may be increased." 



Prof. Karl Pearson has been led by rigorous statistical methods 

 to statements like the following : 



" If selection were to act upon our 5' 9" Englishmen, and 

 the 6' among them were the type best fitted to survive, then 

 with fairly stringent selection it would not take more than six 

 generations to produce a type sensibly 6' high, and this type 

 would be permanently established even if selection ceased. . . . 

 Our determination of the quantitative strength of heredity is 

 thus seen to give values quite intense enough to produce 

 rapid and permanent changes of type, when selection is 

 stringent." 



Prof. Pearson has worked out the following case. Suppose 

 the mean height of a population be 5 ft. 8 in., that a start is 

 made with individuals 6 ft. 2 in., and that for successive genera- 

 tions individuals of this height are selected as parents. It is 

 calculated that in the first generation the offspring would show 

 0*62 of the particular quality selected (h), viz. 6 in. of deviation 

 above the general mean height. It is calculated that after two 

 generations the offspring will show o'Szh, after three generations 

 Q'SgJi, and so on up to O'g2h. Thus by persistent selection 

 an array of individuals would result, almost all of whom were 

 over six feet in height. 



But if at a given generation the artificial selection of tall 

 parents stops, and the tall array is left to inbreed, there will be 

 a gradual sinking back towards the mean height of the population. 



The importance of definite conclusions of this kind can hardly 

 be overestimated. 



