DESCENT WITH CHANGE 



379 



would be obtained by the statistical treatment of nearly all fluctuat- 

 ing characters among the members of any large group of organisms, 

 or of the size of the grains in a handful of sand, or the deviations 

 of shots from the bull's-eye in a shooting match. Therefore the 

 variations with respect to a given character very closely approxi- 

 mate the expectation from the mathematical theory of probability, 

 or chance, and the reasonable conclusion is that such finely-graded 

 fluctuating variations are a resultant of a large number of factors, 

 each of which contributes its slight and variable quota to the ex- 

 pression in a given individual. (Figs. 240, 241.) 



The question is, what results are obtained by breeding from 

 individuals which exhibit such a fluctuating variation to, let us 



Inches 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 72 

 Per, one 2 3 7 18 34 80 135 163 183 163 115 73 41 16 8 5 2 



Fig. 241. — Normal variability curve plotted from measurements of the 

 height of 1052 women (population). The height of each rectangle is propor- 

 tional to the number of individuals of each given height. (From Kellicott, 

 after Pearson.) 



say, a greater degree than that of the mean of a mixed popula- 

 tion? One will perhaps expect, and rightly, that the offspring 

 usually will exhibit the character to a less degree than the parents 

 but to a greater degree than the population. The top (mode) of the 

 curve will have moved, so to speak, slightly in the direction of selec- 

 tion. Now, by continuing generation after generation to select as 

 parents the extreme individuals, is it possible, with due allowance 

 for some regression, to take one step after another indefinitely, or 

 until the character in question is expressed to a degree which 

 did not exist previously? The experience of practical breeders 



