What is the tendency of people with similar characteristics to marry one another?

12.3.1.2 Assortative Mating

Assortative mating is the tendency for people to choose mates who are more similar (positive) or dissimilar (negative) to themselves in phenotype characteristics than would be expected by chance. If these characteristics are genetically determined, positive assortative mating may increase homozygosity in the population. An important difference between inbreeding and positive assortative mating is that inbreeding affects all loci, while assortative mating affects only those that play a role in the phenotype characteristics that are similar. Clinical examples of positive assortative mating are those between individuals who are profoundly hearing impaired or blind, which in some cases may be attributable to the same genotypes.

Assortative Mating

Assortative mating is nonrandom mating based on phenotypes rather than between relatives. Positive-assortative mating or negative-assortative mating occurs if the mated pairs in a population are composed of individuals with the same phenotype more often, or less often, than expected by random mating, respectively. Positive-assortative mating is in some ways analogous to inbreeding in that similar phenotypes, which might have similar genotypes, are more likely to mate than random individuals from the population. Some types of assortative mating are also similar to inbreeding in that they do not change allele frequencies but do affect genotype frequencies. On the other hand, negative-assortative mating may result in balancing selection and the maintenance of genetic variation. Many assortative mating models do change allele frequencies because the proportion of individuals in the matings differs from the proportion in the population. An important point to remember is that assortative mating affects the genotype frequencies of only those loci involved in determining the phenotypes for mate selection (and genotypes at loci nonrandomly associated with those loci), whereas inbreeding affects all loci in the genome.

In a survey of assortative mating studies, Jiang et al. (2013) found that most of the examples of assortative mating were for positive-assortative mating. There appears to be positive-assortative mating for a number of traits in humans, such as height, skin color, and intelligence, although the consequent phenotypic correlation is often not very large. In addition, there also appears to be positive correlations among mates in humans that have particular phenotypes, such as deafness, blindness, or small stature. Of course, there are many different genetic (and nongenetic) causes for deafness, blindness, or small stature so that such a phenotypic correlation may not result in a genetic correlation (for deafness, see Nance and Kearsey, 2004). Rather strong positive-assortative mating may occur in plants when a pollinator forages at a given height or is attracted to a given flower color and, as a result, tends to pollinate plants similar to the ones where the pollen was collected. Similar effects may also occur when flowering time is variable, and only plants that flower simultaneously pollinate each other.

Jiang et al. (2013) found few examples of negative-assortative mating in their review. Some examples are, however, in some plants where successful fertilization occurs only between individuals with different flower types. Although less generally accepted, another example is in populations where rare males (or females) have a mating advantage over more common types. Some reports suggest that negative-assortative mating in mammals and other vertebrates may be based on major histocompatibility complex (MHC) differences.

The overall support for MHC-based, negative-assortative mate choice in humans is mixed and contentious. The most widely known example is the ‘t-shirt study’ in which female Swiss university students ranked the smell of t-shirts worn by male students on characteristics such as pleasantness (Wedekind et al., 1995). The findings of this study suggested that females preferred the odor of males that differed at MHC genes, except when they were on birth control pills, in which case they preferred males that were similar at MHC genes! Further, a follow-up study examining some of the same pairs found no correlation between the rankings for the two different studies.

Several recent studies have examined the correlation of mates for MHC in humans compared to the correlation of genes in the rest of the genome. The first such study found a small but significant (partly due to high statistical power) negative correlation in 30 couples at the MHC region of −0.043 compared to the average in the rest of genome of −0.00 016 (Chaix et al., 2008). However, this level of assortment appears small in a biological sense and nine other genomic regions had higher levels. Subsequently, Derti et al. (2010) concluded that the findings of Chaix et al. (2008) were not statistically robust and they found nonsignificant results in another small-sized, independent sample. These sample sizes appear much too small to draw inference about genomic assortative mating but a study underway of assortative mating in about 10 000 pairs using genome-wide SNP data (P. Visscher, personal communication) should give definitive estimates.

A striking example of negative-assortative mating is in wolves from Yellowstone National Park for gray and black coat color (Hedrick et al., 2016b). Black coat color in wolves is caused by a dominant allele at a beta-defensin gene. In the surveys of mating pairs at Yellowstone from 1995 to 2014, 166 out of 261 (64%) of the matings were between different color wolves, either gray males × black females or black males × gray females (Table 3) with a significant negative correlation of −0.27.

Table 3. The number of matings observed between gray and black wolves in Yellowstone National Park from 1995 to 2014 (Hedrick et al., 2016b)

3 Random or assortative mating?

Size-assortative mating has been reported in the intertidal, gonochoric prosobranchs Olivella biplicata (Edwards 1968) and Littorina littorea (in one of three populations; Erlandsson and Johannesson 1994), in the freshwater prosobranch Viviparus ater (Staub and Ribi 1995), and in the simultaneously hermaphroditic nudibranch Chromodoris zebra (Crozier 1918). In O. biplicata, size-assortative mating can be explained by a zonation of individuals of different size: larger animals live further up the shoreline and smaller ones lower (Edwards 1969). In the slug C. zebra. size-assortative mating can be explained by a physical constraint. Sexually mature slugs range in body length from 4 to 18 cm and two individuals that differ greatly in size are unable to bring their reproductive organs together. Chromodoris zebra can copulate at all hours of the day throughout the year; individuals have plenty of time to match the size of a potential mate and if they are unequal in size they move apart to search for another mate (Crozier 1918).

The terrestrial pulmonate Achatina fulica is a protandrous hermaphrodite. ‘Young adults’, which produce only sperm, continue to grow for a further 3–6 months to became true hermaphrodites (‘old adults’), which produce both sperm and eggs (Tomiyama 1993). In a natural population in Japan, 72% protandric and 28% hermaphroditic snails were recorded; copulations between hermaphroditic individuals occurred more frequently than would be expected under random mating (Tomiyama 1996).

Random mating with respect to size has been observed in the intertidal prosobranchs Littorina rudis and Littorina nigrolineata (Raffaelli 1977), and in a natural population of the hermaphroditic sea hare Aplysia califarnica (Pennings 1991). In terrestrial pulmonates, mating has been reported to be random with respect to shell size (Cepaea nemoralis: Wolda 1963; Helix pomatia and Arianta arbustorum; Baur 1992a), shell colour and banding pattern (C. nemoralis: Schilder 1950; Schnetter 1950; Lamotte 1951; Wolda 1963), and degree of relatedness (A. arbustorum: Baur and Baur 1997). In contrast to most benthic marine and freshwater gastropods, courtship and copulation in intertidal and terrestrial gastropods is restricted to periods of favourable environmental conditions. It has been suggested that because of the time-constrained activity and high costs for locomotion, the best strategy for a snail is to mate with the first mating partner available to minimize the risk of either ending up without any mating at all or drying up during mating (Baur 1992a). The resulting random mating pattern does not imply random fertilization of eggs, because multiple mating and sperm storage offer opportunities for sperm competition (see below). Furthermore, the structure and morphology of the sperm storage site (spermatheca), fertilization chamber and sperm-digesting organ offer opportunities for sperm selection by the female function of the hermaphrodite (cryptic female choice; Eberhard 1996).

15.2 THE STRUCTURE OF SPOUSAL RESEMBLANCE IN ATTITUDES

In our discussion of assortative mating for social attitudes we have taken it as given that spouses select one another on the basis of the same combinations of characteristics that generate the factors derived from the correlations within individuals. Put another way, the pattern of correlations between mates is assumed to be like that within individuals. There is no particular reason why this should be the case. Indeed, for cognitive and socio-economic variables there is a persistent indication that mates may take a fairly “cavalier” attitude toward the descriptive refinements of sociologists and cognitive psychologists in choosing their partners (Eaves et al., 1984). Thus, even though psychologists recognize specific cognitive abilities in designing tests, it is possible that mate selection and cultural transmission are based on linear combinations of these primary factors. A similar process may operate with social attitudes. Mates may select one another on the basis of higher-order combinations of primary factors, for example, so that there could be cross-correlations between primary factors of spouses that are not reflected strongly in the phenotypic correlations within individuals. Even though it may be possible to identify a large number of factors from the analysis of the phenotypic correlations between the items, it may be that the individual dimensions of attitudes are not recognized in the process of mate selection.

Table 15.2 gives the correlations between spouses for the raw responses to the 42 individual items of the British quota sample. These are all significantly greater than zero, and range from 0.13 (item 22) to 0.42 (item 21). Factor analysis of the items revealed five primary factors that could be matched to those extracted from the London twin sample: “religion”, ‘authoritarianism”, “socialism”, “prejudice” and “permissiveness”.

Table 15.2. Marital correlations of items and primary factors.

Do these factors play a part in mate selection, or are different combinations of items involved? The canonical correlations between spouses define those linear combinations of the items that maximize the correlations between mates. If mate selection were truly based on a single dimension then we should expect only one significant canonical correlation. If the five primary factors operate independently in assortative mating then we expect at least five independent dimensions of resemblance between the attitudes of husbands and wives reflected in at least five significant canonical correlations. Table 15.3 summarizes the results of this analysis. Indeed, if we can accept the statistical tests as appropriate to these data then there are nine or ten independent dimensions on which spouses select one another with respect to their attitudes. It turns out that the canonical variates are difficult to interpret, so the analysis was repeated using only the factor scores on the five primary factors that showed cross-cultural stability in Feingold's study. Table 15.4 gives the correlations within and between spouses for these factors.

Table 15.3. Canonical correlation analysis of husband and wife responses to social-attitude items.

Table 15.4. Marital correlations and cross correlations of primary factors.a

The pattern of spousal correlations for the five primary factors mirrors the structure of phenotypic correlations very closely. The correlations are very high indeed and comparable to those we reported for the R and T scales in the British sample. The diagonal of the matrix of correlations between mates contains the highest values. These are the correlations for the same variables rather than between different variables. The off-diagonal correlations between spouses, however, also reflect the structure inherent in the attitudes of the individuals in the sample. The correlation between “religion” and “prejudice”, for example, is − 0.40 for husbands and − 0.44 for wives. The correlation between husband's religion and wife's prejudice is − 0.33, and the reciprocal correlation is − 0.30.

Therefore once again there is much to suggest that mates select one another on the basis of traits that resemble those that factor analysis derives from the item correlations in the population. There is no obvious structure to the spousal correlations beyond that found in the correlations between the primary factors. The canonical correlations derived from this matrix (Table 15.5) confirm the results of our inspection.

Table 15.5. Canonical-correlation analysis of husband and wife social-attitude factor scores.

All five canonical correlations are highly significant, and the first three are quite large. The pattern of loadings of the factors on the canonical variates is very similar for husbands and wives (Table 15.6). The important feature of the data, however, is that mate selection or the interactions between mates occurs on a “trait-by-trait” basis. In contrast to what has been claimed for cognitive and educational variables (Eaves et al., 1984), mate selection for attitudes does not occur on some one-dimensional combination of the attitudes of potential spouses. Furthermore, since the main correlations arise between identical factors rather than across different factors, there would seem to be little justification of any principle of “complementary needs” as far as mate selection for social attitudes is concerned.

Table 15.6. Husband and wife factor loadings on canonical variates.

2 Preference for a Mate Differing at the Major Histocompatibility Complex

In (positive) assortative mating, psychologically, behaviorally, or physically similar individuals pair up. Negative assortative, or disassortative, mating occurs as a result of attraction between dissimilar individuals. An example of positive marital assortment is the tendency of deaf persons to marry one another. Another example is the high correlation between the intelligence quotients of spouses. In the latter example, however, it is not known whether someone with an IQ comparable to one's own is actually preferred as a marriage partner. Rather, people who marry one another have often experienced similar environments as children, and since the environment is a major determinant of IQ, spouses may incidentally have similar IQs.

Recently, Wedekind et al. (1995) suggested that humans prefer a mate differing at the major histocompatibility complex (MHC). MHC is an essential part of the vertebrate immune system. It comprises many genes, each of which is highly variable and exists in many alternative forms called alleles. In the human, the genes are called HLA-A, HLA-B, HLA-C, etc. (HLA=human leucocyte antigen.) For illustrative purposes, let us focus on the HLA-A gene and assume that four alleles are present, which will be called A1, A2, A3, and A4. The pair of alleles in an individual defines the genotype of that individual. The genotypes could be A1A1, A2A2, A3A3, or A4A4, which are called homozygotes, or A1A2, A1A3, A1A4, A2A3, A2A4, or A3A4, which are called heterozygotes.

Mates can share two alleles (as when the genotypes of husband and wife are A1A2 and A1A2), one allele (A1A2 and A2A3, for example), or no alleles (A1A3 and A2A4, say). The hypothesis of disassortative mating asserts that sexual attraction is negatively correlated with the number of alleles shared. Hence, an individual of genotype A1A1, say, given a choice of three different partners whose genotypes are A2A2, A1A2, and A1A1, say, is predicted—all other things being equal—to prefer the first (A1A1 and A2A2 share no alleles) over the second (A1A1 and A1A2 share one allele) over the third (A1A1 and A1A1 share two alleles).

Theoretically speaking, disassortative mating for MHC may make evolutionary sense. Yamazaki et al. (1976) argue as follows. Heterozygotes for MHC genes have a higher fitness than homozygotes, because the presence of two alleles rather than just one would permit an immunological response to a wider range of antigens. Although the genotype that an individual is born with cannot be altered, the genotype of offspring is under some personal control. Namely, by choosing an MHC-dissimilar partner, the chances are improved that offspring will be heterozygotes. A second reason why disassortative mating for MHC might be favored by natural selection is that it leads to avoidance of inbreeding.

A preference for MHC-dissimilar mates is fairly well established in house mice. There is good evidence, at least in mice, that this preference is mediated by body odor, which is influenced by the MHC genes. More precisely, mice imprint on the MHC identities of the individuals by, or with, whom they are raised, who under normal circumstances would be extended family. They subsequently choose a mate differing in MHC from these individuals. Since, close relatives are on average more similar for their MHC than unrelated individuals, the resulting preference would be for MHC-dissimilar mates.

The human evidence is of two kinds. First, in a provocative experiment, male subjects were asked to each wear a T-shirt, and female subjects to rate the odors of these T-shirts for pleasantness (sexiness). The subjects, all of whom were students at a Swiss university, were also typed for their HLA-A, HLA-B, and HLA-DR genes. It turned out that the females preferred the odors of T-shirts worn by MHC-dissimilar males to those worn by MHC-similar males (Wedekind et al. 1995).

Second, a direct test of whether humans mate disassortatively for MHC should be possible by typing married couples. To date, such an analysis has been done on the Hutterites (a North American reproductive isolate of European ancestry), South American Indians from the lower Amazon basin, and the Japanese. Only among the Hutterites has dissortative mating been demonstrated (Ober et al. 1997). In the other two populations, mating is apparently random with regard to MHC.

It is not clear why such contradictory results have emerged. One possibility is that, since the preference for MHC-dissimilar mates is weak, it may sometimes be overwhelmed by other biological or cultural factors. In particular, a preference for cousin marriages will mask any tendency towards disassortative mating. However, there is no evidence of cousin marriages among the couples sampled in the above mentioned studies.

Disassortative Mating

Disassortative mating (sometimes called negative assortative mating) occurs when mates are chosen to be more phenotypically dissimilar than would arise by chance alone. Disassortative mating is not only the opposite of assortative mating in terms of the phenotypes displayed by mating pairs but also in its evolutionary and genetic consequences. This can be shown by the simple one-locus, two allele model of 100% disassortative mating given in Table 3.7. In this model, every genotype has a distinct phenotype and mates at random only with those individuals with a different phenotype, with no gender effects.

Table 3.7. A Model of 100% Disassortative Mating at a Single Locus With Two Alleles, A and a, With Each Genotype Having a Distinct Phenotype

SUM = GAA × GAa + GAA × Gaa + GAa × Gaa is used to standardize the mating frequencies so that they sum to 1.

As can be seen from Table 3.7, this system of mating produces many heterozygotes and few homozygotes—just the opposite of assortative mating. For example, suppose we started out with Hardy–Weinberg genotype frequencies with p = 0.25, with an initial heterozygote frequency of 0.375. Then in a single generation of disassortative mating as given by Table 3.7, the frequency of heterozygotes would increase to 0.565. Unlike the assortative mating model, in this case the allele frequency also changes from 0.25 to 0.326, so disassortative mating is a strong evolutionary force at the single locus level. However, with p = 0.326, the expected heterozygosity under random mating is 0.439, so there is still a heterozygous excess under disassortative mating with f = −0.286. Hence, disassortative mating resembles avoidance of inbreeding, but unlike avoidance of inbreeding, it only affects the loci contributing to the phenotype for which disassortative mating is occurring and loci in linkage disequilibrium with them. In addition, unlike avoidance of inbreeding, disassortative mating alters allele frequencies and tends to stabilize them at intermediate levels.

At the multi-locus level, disassortative mating can bring together into the same family alleles that have opposite effects on phenotypes. This could potentially generate some linkage disequilibrium, but by also causing excesses of heterozygosity, disassortative mating dissipates linkage disequilibrium much more rapidly than random mating (recall, recombination only changes gamete frequencies in double heterozygotes). Hence, disassortative mating is not as effective as assortative mating in generating or maintaining linkage disequilibrium.

A potential example of disassortative mating in humans is the major histocompatibility complex (MHC) (Laurent and Chaix, 2012a,b). MHC is a genomic region containing multiple genes coding for molecules whose role is to present self- and nonself-derived peptide antigens to T cells, thereby playing a critical role in immune response and in organ transplant success. MHC is a 3.6 megabase-pair long region located on the short arm of chromosome 6 in the human genome. Many of these same MHC genes influence body odor, and studies in other species and possibly humans indicate disassortative mating at MHC mediated by olfactory cues (Havlicek and Roberts, 2009). As expected for a region under disassortative mating, the MHC region shows a significantly higher level of heterozygosity than other regions of the human genome (Laurent and Chaix, 2012b). However, many studies do not indicate disassortative mating at MHC, and a metaanalysis of MHC effects on human mating revealed both MHC-dissimilar and MHC-similar matings in various studies (Winternitz et al., 2017). This seemingly contradictory pattern appears to be an artifact of population ethnic heterogeneity in observational studies that tend to indicate assortative mating versus experimental studies with more control over sociocultural biases that tend to indicate disassortative mating or mating for diverse MHC mates (Winternitz et al., 2017). In many areas of the world, human populations from diverse geographical areas and with different cultures have been brought together, as will be discussed in detail in Chapter 6. North America is one such area, and many of the studies on MHC have been performed on North American populations. Assortative mating by “ethnicity” has been historically quite strong and reduces genetic admixture among the descendants of these historic populations, although assortative mating by “ethnicity” has been diminishing with each successive generation (Sebro et al., 2017). Although “ethnicity” is not a genetic trait per se, it is often associated with some degree of genetic differentiation that reflects the historical origins of the parental populations that have been brought together into a single geographic region (Chapter 6). Hence, assortative mating in North America by “ethnicity” has resulted in deviations from Hardy–Weinberg and linkage disequilibrium for those loci that were differentiated between the parental population gene pools (Sebro et al., 2017), which includes the MHC cluster. When “ethnicity” and other sociocultural biases that influence mate choice are not controlled, it appears as if there is assortative mating for MHC, but when these factors are eliminated or controlled, it appears as if there is disassortative mating for MHC (Winternitz et al., 2017).

Hybridization and Mate-Choice Learning

Although mate-choice imprinting often results in positive assortative mating, typically with conspecifics, the potential exists for misimprinting, or imprinting on the wrong species. Hybridization between species of Darwin’s finches, for example, is known to occur and is thought to result from misimprinting. Additionally, crossfostering experiments conducted in the wild have demonstrated that some bird species will imprint on a foster parent of another species, resulting in heterospecific pairings.

Heterospecific matings could result in hybrid offspring and hybrid zones are not uncommon in nature. What role then, if any, does mate-choice imprinting play in hybrid zones? Using an artificial neural network, Brodin and Haas demonstrated that phenotypes of pure species are learned faster and better than those of hybrids, potentially leading to selection against hybrids. Further spatial simulations combined with empirical data on dispersal demonstrate that mate-choice imprinting can maintain a hybrid zone under natural conditions.

Hybridization and Mate-Choice Learning

Although mate-choice imprinting often results in positive assortative mating, typically with conspecifics, the potential exists for misimprinting, or imprinting on the wrong species. Hybridization between species of Darwin’s finches, for example, is known to occur and is thought to result from misimprinting (Grant and Grant, 2008, 2009). Mate-choice copying can also lead to heterospecific matings if a heterospecific model is perceived to be of high quality (Hill and Ryan, 2006) or if models mate with heterospecific males (Schlupp et al., 1994; Heubel et al., 2008).

Heterospecific matings caused by mate-choice learning could lead to the production of viable hybrid offspring. Thus, mate-choice copying may play a role in promoting hybridization and the establishment of hybrid zones; an area of research argued to be heavily overlooked (Varela et al., 2018). The role of mate-choice learning in an established hybrid zone was, however, previously explored using an artificial neural network approach. Brodin and Haas showed that phenotypes of pure species are learned faster and better than those of hybrids, potentially leading to selection against hybrids (Brodin and Haas, 2006). Further spatial simulations combined with empirical data on dispersal demonstrated, however, that mate-choice learning can maintain a hybrid zone under natural conditions (Brodin and Haas, 2009).

A recent framework exploring the mechanisms and conditions under which speciation or hybridization is expected with mate-choice copying was put forth by Varela et al. (2018).The authors conclude that hybridization resulting from mate-choice copying increases a species’ opportunity for speciation events caused by adaptive introgression and radiation as well as by hybrid speciation and reinforcement (Varela et al., 2018).

Narrow Genetic Base

The rubber breeding material used in Asia has a very narrow genetic base, having originated from seeds collected by Sir Henry Wickham in a very small area of Brazil. The initial breeding base came from about 2000 seedlings from the Wickham collection imported into Asia in 1877. Most of these seedlings went to Sri Lanka, with much smaller numbers going to Malaya and Indonesia. There were some other introductions to Southeast Asia in later years but due to their poor yield they have not been used in subsequent breeding programs. Hence, rubber breeding in Asia (and in Africa) is based on limited germplasm.

The intensive use of directional selection for high yield, phenotypic assortative mating, and the extensive use of clonal propagation in the early breeding programs have aggravated the problem of the narrow genetic base. As a result, some undesirable features have become apparent which can be illustrated by the following:

Most of the clones bred to date in Malaysia (and other Southeast Asian countries) can be traced to seven “primary clones.” Some inbreeding has therefore featured in the early breeding programs. Some of these primary clones are themselves related, so inbreeding may be even greater.

There is a diminishing return obtained from breeding effort. There was a substantial yield increase in the early phases of breeding, but this has diminished in later phases (see below).

Genetic erosion seems to have occurred as materials selected in one region may be found to be less adaptable in another. For example, Southeast Asian selections are susceptible to South American leaf blight (SALB) which could be due to lack of resistance in the original Wickham collection or subsequent loss through genetic erosion following selection in the absence of the disease.

16.2.2 Reliable environmental transmission

The previous model assumes that all the environmental factors are involved in assortative mating but that they are not transmitted with complete reliability between generations. This model predicts a high correlation between the environmental determinants of spouses. It is difficult to conceive of a variable that is so completely correlated between spouses and yet that is not transmitted easily between parents and children. Religious affiliation, for example, is very highly correlated between spouses, but is also transmitted readily to children. We thus consider an alternative model for the cultural component of conservatism that assumes, in effect, that not all the environmental effects on conservatism are transmissible, but those that are transmissible are perfectly correlated between spouses and transmitted with complete reliability to their children. This model approximates the kinds of prediction that would follow if the cultural component of conservatism were a secondary consequence of the non-genetic inheritance of a variable such as religious affiliation.

In Figure 16.2 we give such a modified model for the effects of assortative mating for the environment. We postulate a latent environmental variable R that is perfectly correlated between spouses and between parents and children. The “culturar” path coefficient b is set equal to ½. When the marital correlation for R is unity, this value for b ensures that all members of the nuclear family are completely correlated for R. There are residual environmental effects on the measured phenotype, which are assumed to be uncorrelated between family members. The model predicts the following correlations:

rMZ=h2+r2,rDZ=12h2+r2,rspouse=r2.

Note that the predicted correlation for DZ twins must be at least as high as that between spouses. This constraint is clearly violated by the small sample of Australian spouses and nearly so by the London sample.

Fitting this model by weighted least squares yields the results in Table 16.2. The model fails in both samples, suggesting that the non-genetic inheritance of attitudes cannot be explained simply by postulating a latent variable that is perfectly correlated between the members of nuclear families.

Table 16.2. Results of fitting model assuming complete family resemblance for transmissible environmental effect.