46 SOME RECENT MISINTERPRETATIONS OF THE 



namrly -50. But there is an essential difference between the two 

 cases; pairs of brothers or sisters have a relatively high correlation 

 with each other, namely -50, but parents have a low correlation with 

 each other, on the average in the middle classes, perhaps, -15, due to 

 assortative mating; in the lower classes this sinks almost to zero. 

 The following table gives the results: 



TABLE III. 



Showing the Effect of Intercorrelations between Relatives, when a Judg- 

 ment has to be formed of the Character in an Individual (A) from 

 a knowledge of it in his Relatives. 



Multiple Correlation Coefficients 



Thus we see that unless there be assortative mating, we do not 

 better matters by introducing for the purposes of prediction the four 

 grandparents. And even with assortative mating two parents are 

 worth more than six brothers or sisters. Two parents who have not 

 selected their likes would form as good a basis for prediction as an 

 infinity of brethren, were such indeed available. Now the correlation 

 between brethren is only -50, and that between environmental factors 

 is often higher, thus this illustration will indicate how very little 

 additional intensity of correlation is gained by combining inter- 

 correlated factors. We do not add the correlations due to separate 

 factors as Mr Carr-Saunders' paper appears to suggest. If the factors 

 are absolutely independent, as when persons marry without assortative 

 mating, then the multiple correlation coefficient is the square root of 

 the sum of the squares of the correlations combined. 



Thus if we took 100 independent environmental factors with a 

 strength of about -05, the resulting combined or multiple correlation 



