126 



IRENE BARNES TAEUBER 



imately equal Negro and White ancestry would tend to increase the vari- 

 ability of the population as a whole, while a type of crossing which tended 

 to increase the proportion of unmixed Negro matings would decrease the 

 variability of the total population. 



Herskovits has previously shown that there is a definite curvilinear rela- 

 tionship between the color of husbands and wives. Table 2 gives the 

 average color of wives according to the color of their husbands, and vice 

 versa, based on the same families Herskovits used in his study, plus addi- 

 tional families measured in rural West Virginia. The coefficient of regres- 

 sion of wives on husbands is .44; of husbands on wives, .41. 6 "The darkest 

 women .... seem to marry men of about their own color or lighter, but 



TABLE 2 



Average skin color {per cent N) of wives for increasingly dark husbands, and of husbands 



for increasingly dark wives 



the lighter women, and those not pronouncedly dark, marry men darker 

 than themselves." 7 



Both husbands and wives were divided into those with less than 45 per 

 cent N (light), 45 to 67 per cent N (medium), and 67 per cent N or more 

 (dark). The coefficient of contingency for the color of husbands and wives, 



6 The skewness of the distributions makes the validity of the use of correlation co- 

 efficients extremely questionable, but within the limits of this objection, the r between 

 the percentage of N pigmentation of husbands (y) and wives (x) is +.42, while rjyx is .51 

 and rj yx is .52. The difference between r and -q yx is 3.22 times the probable error of f . 



7 Herskovits, M. J., "Social selection in a mixed population," Proceedings of the National 

 Academy of Sciences, vol. xii, 1926, pp. 587-593. 



