1 Assortative Mating.

Any non-random pairing of mates on the basis of factors other than biological relatedness is subsumed under the general category of assortative mating. Mating based on relatedness is termed inbreeding, and will not be examined in this book. We discuss assortative mating under the general heading of genotype-environmental effects for two main reasons. First, when assortment is based on some aspect of the phenotype, it may be influenced by both genetic and environmental factors. Second, assortative mating may affect the transmission, magnitude, and correlation of both genetic and environmental effects. In human populations, the first indication of assortative mating is often a correlation between the phenotypes of mates. Usually, such correlations are positive. Positive assortment is most marked for traits in the domains of education, religion, attitudes, and socioeconomic status. Somewhat smaller correlations are found in the physical and cognitive domains. Mating is effectively random, or only very slightly assortative, in the personality domain. We are not aware of any replicated finding of a significant negative husband-wife correlation, with the exception of gender! Assortative mating may not be the sole source of similarity between husband and wife -- social interaction is another plausible cause. A priori, we might expect social interaction to play a particularly important role in spousal resemblance for habits such as cigarette smoking and alcohol consumption. Two approaches are available for resolving spousal interaction from strict assortative mating. The first depends on tracing the change in spousal resemblance over time, and the second requires analyzing the resemblance between the spouses of biologically related individuals (see Heath, 1987). Although the usual treatment of assortative mating assumes that spouses choose one another on the basis of the trait being studied (primary phenotypic assortment), we should understand that this is only one model of a process that might be more complicated in reality. For example, mate selection is unlikely to be based on an actual psychological test score. Instead it is probably based on some related variable, which may or may not be measured directly. If the variable on which selection is based is something that we have also measured, we call it correlated variable assortment. If the correlated trait is not measured directly we have latent variable assortment. In the simplest case, the latent variable may simply be the true value of trait of which the actual measure is just a more or less unreliable index. We then speak of phenotypic assortment with error. Once we begin to consider latent variable assortment, we recognize that the latent variable may be more or less genetic. If the latent variable is due entirely to the social environment we have one form of social homogamy (e.g., Rao et al., 1974). We can conceive of a number of intriguing mechanisms of latent variable assortment according to the presumed causes of the latent variable on which mate selection is based. For example, mating may be based on one or more aspects of the phenotypes of relatives, such as parents' incomes, culinary skills, or siblings' reproductive history. In all these cases of correlated or latent variable assortment, mate selection may be based on variables that are more reliable indices of the genotype than the measured phenotype. This possibility was considered by Fisher (1918) in what is still the classical treatment of assortative mating. Clearly, the resolution of these various mechanisms of assortment is beyond the scope of the conventional twin study, although multivariate studies that include the spouses of twins, or the parents and parents-in-law of twins may be capable of resolving some of these complex issues (see, e.g., Heath et al., 1985). Even though the classical twin study cannot resolve the complexities of mate selection, we have to keep the issue in mind all the time because of the effects of assortment on the correlations between relatives, including twins. When mates select partners like themselves phenotypically, they are also (indirectly) choosing people who resemble themselves genetically and culturally. As a result, positive phenotypic assortative mating increases the genetic and environmental correlations between relatives. Translating this principle into the context of the twin study, we will find that assortative mating tends to increase the similarity of DZ twins relative to MZ twins. As we shall see, in twins reared together, the genetic effects of assortative mating will artificially inflate estimates of the family environmental component. This means, in turn, that estimates of the genetic component based primarily on the difference between MZ correlations and DZ correlations will tend to be biased downwards in the presence of assortative mating.