K. Pearson and A. Lke 375 



of local races, and thai iiicii and wonien inatf witliin tlieir locality. Now it 

 appears to mo that this argiiiiient would be far inore valid, it' iny mateiial was 

 drawu in bulk fruni local lower midtUo and artizan classus. But it is vory 

 doiibtful how far it is true of the iniddlo classes, such as provide the stiidents at 

 the London Colleges. The niiddle classes undoubtedlymarry in their ovvn " sets," 

 but these aie hardly looal sets. Further, a wide .series of assortative mating 

 observations have been niade on another, wholly dififerent class of characters, in 

 which local race is regarded, and the coefficieiits conic out as high as in the 

 present data. Hence, I think, we are forced to the conclusioii that the bulk 

 of the observed resemblance in physica! characters between parents is (lue U> a 

 direct, if quite unconscious, selection of liko bv liko, aiid possibly in a contributory 

 degree to a likeness in parents for the characters under considei'ation emphasising 

 their fertility. 



The ainount of " consciousness " in the selection niay possibly be measured by 

 the difference between the statu re-stature correlation and those for span-span, 

 and foreann-forearm. 



(vi) Cross-Assortative Mating in Man. 



The second part of Table IL gives the cross-coefficients, for e.xample, the 

 correlation between husband's stature and wife's forearm. We might d priori, 

 perhaps, anticipate that the correlation lietween a hrst organ in the husband 

 and a secoud in the wife, would be equal to the correlation between the second 

 in the husband and the tirst in the wife. This is actually the case for .span and 

 forearm, and, perhaps, we niay consider for stature and span ; the results for 

 stature and forearm are less close than we might have anticipated, but the work 

 has been revised without the discovery of any error. Relations such as : 



do not, however, appear to be theoretically necessary. 



The problem now arises : are cross correlations between characters in husband 

 and wife, solel}' due to selection of direct characters ? 



1 think this may be roughly tested ni the tolluwiug manner. Suppose unly 

 these Organs to be selected and the direct selection coetficients to be p^, p^, pe«, 

 as before. They may be foumi frijiii equations (i)* and we have the values: 



Pi„ = -2:374., p^ = 0053, p,, = -1043. 



Thus there is most immediate selection of stature, a sensible selection of 

 forearm, and practically none of span. 



' These give numerically : 



■2804 = p,o + -ö'JlDpsj + -SSlS/jj« , 



•1989 = -Sgiep,« + P3i + -öOSTpjj , 



■ 1977 = -3818^,., + ■y087pj, + pjs . 



48—2 



