320 STATISTICAL STUDY OF INHERITANCE 



an average height differing but little from that indicated by the 

 formula. In regard to all these statistical conclusions, it must 

 be carefully borne in mind that they cannot be applied to indi- 

 vidual cases. " Of the individual we can assert nothing as certain, 

 only state the probable. The individual varies owing to the vari- 

 ability of the gametes, and we know nothing of the particular 

 gametes which fused to give the stirp, of which he is the product. 

 All we know in heredity is what degree of resemblance there is 

 on the average. . . . The statistician dealing with heredity is 

 like the physicist dealing with the atom — he can say little or 

 nothing of the individual, his knowledge is of the group containing 

 great numbers." (Pearson, op. cit., p. 457.) 



Regression and Correlation. — As the term regression, used by 

 Galton to describe the extent to which an average son is more like 

 the mean of the stock than his father is, has been often misunder- 

 stood to imply something in the nature of a " throwback," it is 

 probably desirable to get rid of it and to substitute for it the tech- 

 nical term correlation, which expresses the extent to which a son 

 approximates nearer to his father than to the average of the stock. 



The term " regression " which Mr. Galton introduced into 

 biometry is not really a biological term. As the late Prof. 

 Weldon pointed out in an interesting lecture, there may be 

 regression between two different sets of results of dice-throwing 

 if the second set of results is in some way, but not entirely, 

 dependent upon the first. He protested against regarding 

 regression "as a peculiar property of living things, by virtue 

 of which variations are diminished in intensity during their 

 transmission from parent to child, and the species is kept true 

 to its type " (1906, p. 107). 



If a set of fathers deviate, in respect to some character, a certain 

 amount from the general mode of the whole population, their 

 sons will, in respect to the same character, vary about a mode 

 which is between the paternal deviation and the mode of the 

 whole population. This is filial regression. 



