Assortative marriage for specific cognitive abilities in two ethnic groups.

Abstract

Age-adjusted spouse correlations for 15 tests of specific cognitive abilities were relatively small in samples of Americans of Japanese Ancestry (AJA) and Americans of European Ancestry (AEA). Median correlations based upon data from 148 AJA couples and 555 AEA couples were + .07 and + . 13, respectively. When effects of educational attainment were statistically removed, resulting partial correlations became even smaller. These results suggest that there is relatively little assortative mating for specific cognitive abilities, especially for that part of ability which is independent of educational level. Assortative mating, as evidenced by nonrandom marriage, has long been of interest to human biologists. The following advice as to assortative mating may be found in Plato s Laws : Let there be one word concerning all marriages: Every man shall follow, not after the marriage which is most pleasing to himself, but after that which is most beneficial to the state. For somehow every one is by nature prone to that which is likest to himself, and in this way the whole city becomes unequal in property and in disposition (Hutchins, 1952, p. 707). Since assortative mating influences the correlation between relatives and the genetic variance in a population (Crow and Kimura, 1970), it is important to assess its extent for a wide range of characters. The literature concerning assortative marriage for both physical and behavioral characters has been reviewed by Spuhler (1967, 1968) and by Vandenberg (1972). The highest spouse correlations reported are those for age, in the range of -I.50 to +.99. Spouse correlations for physical characteristics and personality ratings are usually positive, but low, with median values of about + .15. In contrast, moderately high husband-wife correlations for performance on tests of cognitive ability have been reported. In a recent review by Jensen Supported by NSF grant GB-34720 and grant HD-06669 from the National Institute of Child Health and Human Development. 2Behavioral Biology Laboratory, University of Hawaii, Honolulu, Hawaii 96822. institute for Behavioral Genetics, University of Colorado, Boulder, Colorado 80302. Human Biology May 1976, Vol. 48, No. 2, pp. 343-352. ® Wayne State University Press, 1976 This content downloaded from 157.55.39.78 on Thu, 23 Jun 2016 05:10:00 UTC All use subject to http://about.jstor.org/terms 344 R.C. Johnson et al. (1975), the unweighted mean of 43 spouse correlations for various measures of mental ability was +.45. Although spouse similarity for various measures of cognitive ability is well documented, the cause of this correlation remains unknown. As Warren (1966, p. 289) stated in an earlier discussion of assortative mating, Whether it occurs because people actually seek spouses with characteristics similar to their own, or whether people are sorted spatially by some characteristic and then marry someone at hand, still has not been determined. Since educational attainment is a function of mental ability, the observed spouse similarity for this ability could be due to either or both of the following: (1) propinquity mediated by educational attainment, and/or (2) the practice of seeking a mate similar to oneself in intellect. Warren (1966), for example, found that observed spouse correlations with respect to socio-economic origin were reduced by one-half when effects of educational level were statistically controlled. To the extent that cognitive ability is correlated with educational level, statistically adjusting mental ability for educational level will overcorrect the spouse correlation. Nevertheless, as long as the results are interpreted conservatively, a partial correlation analysis is appropriate for such data. For the purpose of such analysis, cognitive ability may be considered to consist of two parts: one part which is correlated with educational level, and one part which is not. A partial correlation of the test performances of husband and wife, holding educational level constant, will thus assess spouse similarity for that part of cognitive ability which is independent of educational attainment. There is another factor, in addition to propinquity and mate seeking on the basis of intellectual similarity, which can yield a spouse correlation for cognitive ability. When cross-sectional data are examined, some change in mental ability occurs as a function of age (Wilson et al. 1975). It was previously noted that the correlation between mates for age is very high. Thus, when the age range of subjects is wide, at least some of the observed spouse similarity for cognitive ability may be due to their close similarity in age. Because of this age factor, it is essential that appropriate age adjustment be employed in studies of assortative marriage for cognitive ability. It should be noted that a simple linear adjustment may not always be appropriate. For some tests, the age function is highly curvilinear (Wilson et al. 1975). The primary objective of the present study is to report spouse correlations for specific cognitive abilities in two different ethnic groups. A secondary objective is to assess changes in spouse correlations as a function of This content downloaded from 157.55.39.78 on Thu, 23 Jun 2016 05:10:00 UTC All use subject to http://about.jstor.org/terms Assortative Marriage o Cognitive Abilities 345 statistically controlling age and educational level. Only one relatively small study (N=55 pairs) of assortative marriage for specific cognitive abilities has been previously reported (Williams, 1975), and, according to Jensen (1973), there are no previous reports of assortative marriage for mental ability in non-Caucasian groups. Materials and Methods As described in a previous report (DeFries et al. 1974), a large scale family study is currently being conducted in Hawaii with the objective of assessing genetic and environmental influences on performance on various tests of cognitive ability. During the initial two years of testing, data have been obtained from members of 996 families (3,268) individuals). A family consists of both biological parents, 60 years of age or younger, and one or more children, 14 years of age or older. Data are obtained on 15 cognitive variables, various environmental measures, blood groups and enzyme systems, and dermatoglyphics. Details concerning the testing procedure are provided in Wilson et al. (1975). The names of the 15 tests, in order of administration, are presented in Table 1. Also shown are test times and estimated reliabilities. Phenotypic correlations among the 15 cognitive variables were obtained for the two major ethnic groups represented in the sample, 648 Americans of Japanese Ancestry (AJA) and 1, 971 Americans of European Ancestry (AEA). Resulting phenotypic correlation matrices were subjected to principal component analysis with varimax rotations. Communalities of one were used, and the factors retained for rotation were those for which the eigenvalues were greater than one (Kaiser, 1960). Four readily interpretable factors emerged for both racial groups: spatial visualization, verbal, perceptual speed and accuracy, and visual memory. Coefficients of congruence (Tucker, 1951) were +.99 for each factor. A similar analysis was undertaken for three different age groups (14-20, 21-40, 41-60 years). Coefficients of congruence were H-. 98 or greater for each of the 12 comparisons (three age comparisons for each of four factors). These results indicate that cognitive structure (as indexed by correlational structure) is nearly identical in the two ethnic groups (DeFries et al. 1974) and is invariant across the age range sampled (Wilson et al. 1975). There are 555 AEA couples and 148 AJA couples for whom there are no missing data (listwise deletions). Spouse correlations were obtained only from data on these couples. This content downloaded from 157.55.39.78 on Thu, 23 Jun 2016 05:10:00 UTC All use subject to http://about.jstor.org/terms 346 R.C. Johnson et al.