382 THE AMERICAN NATURALIST [Vol. LVI 



length, head breadth and cephalic index in a series of Dutch 

 families. He has himself calculated a coefficient of correlation 

 of .039 ± .034 for the cephalic index of husband and wife in 

 389 families.^ We have felt it desirable to determine the cor- 

 relation for length and width of head, as well as that for index. 



Because of a suggestion by Pearson {loc. cit.) that the cor- 

 relation apparently indicating assortative mating may be really 

 due in some cases to an association of fertility with homogamy, 

 we have thought it desirable to calculate all the coefficients of 

 correlation in two ways: (1) by using the actual number of 

 parents, and (2) by weighting the parents with the number of 

 offspring indicated in Frets' tables.® 



The correlation coefficients are as follows: 

 Length of husband's head and length of wife's head: 

 Parents only, r = + .0487 ± .0377. r/Er = 1.29. 



r=+ .0616 ± .0376. r/Er = 1.63.. 

 Breadth of husband's head and wife's head: 



Parents only, r=+ .1197 ± .0372. r/Er = 3.22. 



r= + .1184 ± .0372. r/Er^ 3.18. 



Parents only, r = + .0231 ± .0377. r/Er = 0.61. 



r = — .0546 ± .0376. r/Er = 1.44. 



The constants are with one exception positive in sign. That 

 for the breadth of husband and breadth of wife may perhaps 

 be considered statistically significant in comparison with its 

 probable error. The others, particularly that for the cephalic 

 index, can not be so considered. 



The coefficients may, therefore, indicate a slight assortative 

 mating for the dimensions of the head. The coefficients, in 

 common with those for physical characters other than stature, 

 are relatively low. That the correlation for the cephalic index 

 is so low is a point of particular interest. If cephalic index be 

 a character of great importance in distinguishing races, and if 

 correlations which have been demonstrated between the physi- 



5 Frets, G. P., " Erfelijkheid, eorrelatie en regressie," Genetica, 3: 1-27. 

 1921. 



«Wo are able to abstract from Frets' tables 319 pairs of parents an 

 which there were no indications of typographical errors when different 

 tables were checked against each other. These had a total of 1328 recorded 



