CORRELATION 107 



affected, as regards their value for representing quanti- 

 tatively a hereditary relationship between two indi- 

 viduals, by the fact that the statistics from which they 

 are derived show the existence of a marked correlation 

 between husbands and wives in point of stature, 

 amounting, indeed, to as much as 0-28 the result 

 of what is technically described as selective mating. 

 In the absence of such a relation between the statures of 

 the parents, the correlation between parent and child 

 might be expected to be distinctly less than that 

 between pairs of brothers or sisters. 



The term correlation replaces to some extent the 

 older term regression employed by Galton. When 

 speaking of regression the facts already described are 

 regarded from a slightly different point of view. It 

 is sometimes found convenient to speak of the regres- 

 sion of the mean stature of an array of sons toward 

 the mean of the general population, instead of speaking 

 of the correlation between the filial mean and the 

 value of the parental class. 



Regression represents the extent to which the 

 average son is more like the mean of the general 

 population than his father is. Correlation, on the 

 other hand, indicates the amount by which the son is 

 more like his parent than he is to the average of the 

 general population. Thus, instead of being exactly 

 like their parents, children are said to show regression 

 towards the mean of the general population to which 

 both parents and children belong. 



In a case where the mean height of the fathers is 

 identical with the mean height of the sons examined, 



