HEREDITY 487 



The coefficient of heredity is therefore nothing more nor less than 

 the correlation coefficient (r) obtained from a regression table 

 in which two sets of individuals related by descent are tabulated 

 with respect to the same character. The methods of rinding 

 the coefficient of heredity are precisely the same as those already 

 described for finding the coefficient of correlation ; indeed, the 

 correlation coefficient of a regression table is the coefficient of 

 heredity. 



It is manifest that this correlation table may be constructed 

 not only between mid-parents and offspring, but between fathers 

 and sons, between grandfathers and grandsons, between mothers 

 and sons (or daughters), between uncles and nephews, between 

 brothers and sisters, between brothers and brothers, and, indeed, 

 between persons connected by any ties of consanguinity what- 

 ever, direct or indirect. In each case the correlation coefficient 

 becomes a good measure of hereditary resemblance. 



If a regression table be constructed between fathers and 

 mothers, a correlation would still be found, though the two are 

 united by no blood lines except those common to the race in 

 general. Such correlation comes entirely through selection, and 

 its coefficient (r) is commonly called the coefficient of cross 

 heredity or " assortative mating." It is a good measure of the 

 degree of selection involved in mating. 



The table on the next page gives some of the coefficients of 

 heredity that have been determined for different relatives. 



Pearson remarks : " We see that on the average the intensity 

 of parental correlation is about 0.3 to 0.5 ; of grand parental, 

 about 0.15 to 0.3; and of fraternal, about 0.4 to 0.6, the 

 latter correlation being somewhat reduced when the fraternity 

 consists of members of opposite sexes." 



Regression coefficient. The regression coefficient here is 

 computed exactly the same as the regression coefficient from 

 any other correlation table, and it has the same uses, namely for 

 prediction ; that is to say, for example, knowing the deviation of 

 a group of mid-parents from their mean, what deviation shall 

 we expect on the part of their offspring ? 



The use of the word " regression" in the term "regression 

 coefficient " is likely to lead to confusion. We must not assume 



