602 



THE AMERICAN NATURALIST [Vol. XLII 



difficult to assert that this resemblance is due to other causes than 

 those assortative processes which have just been shown to produce quite 

 a sensible degree of resemblance in husband and wife." 



Pearson is "prepared to accept with some reservation a 

 sensible but probably not very large infective action from the 

 available statistics of pulmonary tuberculosis." The question 

 of assortative mating is an important one, and a knowledge of 

 the amount to be allowed under various circumstances seems to 

 me to be a necessary adjunct for recorrecting all the correla- 

 tion coefficients of heredity which have so far been obtained by 

 the London workers. Their coefficients agree fairly well, but 

 they are all distinctly higher than the theoretical — fraternal 

 are about .50 to .60 instead of the theoretical .40; paternal .40 

 to .50, the theoretical being .30; and so with the most remote 

 relationships, especially the first cousin resemblances. 



It is evident that if assortative mating be in general the 

 strong force that Pearson has shown it to be, then it must in 

 most investigations raise the correlation coefficients for heredity. 

 To make this clear — tall fathers have on the average tall sons, 

 though their average height is less than that of the fathers, 

 due to the principle of regression, but now if it happens that all 

 the tall fathers have tall wives, then the sons will get an added 

 height from the influence of the tall mothers and w T ill seem to 

 resemble their fathers more than they do from the real paternal 



Among royal families assortative mating is a disturbing factor 

 is at a minimum, for here the marriages are so often arranged 

 by others than the parties most concerned, or are the result of 

 some important state policy, that the question of individual 

 selection is nearly, though I believe not quite, eliminated. This 

 may be the reason why the coefficients for heredity found in the 

 study of royalty are so much nearer the theoretical. 



It may be well, in closing, to say a word about the general 

 question of correlation coefficients as affording a proof of the 

 influence of pure heredity. It may be asked— do the coefficients 

 really prove anything more than a general resemblance between 

 relatives? May this not be due to heredity in some cases and 

 to environment in others, or a combination of both, in most cases? 

 Personally I do not feel that the coefficients alone afford all the 

 desired proof. Analysis of the material, separating the cases 

 into classes in which environment has had greater or less time 



