In the human species, the two uniparental genetic systems (mitochondrial DNA and Y chromosome) exhibit contrasting diversity patterns. It has been proposed that sex-specific behaviours, and in particular differences in migration rate between men and women, may explain these differences. The availability of high-density genomic data and the comparison of genetic patterns on autosomal and sex chromosomes at global and local scales allow a reassessment of the extent to which sex-specific behaviours shape our genome. In this article, we first review studies comparing the genetic patterns at uniparental and biparental genetic systems and assess the extent to which sex-specific migration processes explain the differences between these genetic systems. We show that differences between male and female migration rates matter, but that they are certainly not the only contributing factor. In particular, differences in effective population size between men and women are also likely to account for these differences. Then, we present and discuss three anthropological processes that may explain sex-specific differences in effective population size and thus human genomic variation: (i) variance in reproductive success arising from, for example, polygyny; (ii) descent rules; and (iii) transmission of reproductive success.
Since the 1990s, sex-specific genetic variation within and among human populations has been characterized using markers located on the maternally inherited mitochondrial DNA (mtDNA) and on the paternally inherited Y chromosome. Most studies converged towards the view that the paternal and the maternal components of human genetic variation tell contrasting stories. At a global scale, Y chromosome markers exhibit a higher degree of differentiation among populations than mtDNA markers, while the mtDNA-based coalescent time estimated for the human species is significantly longer than the Y chromosome-based coalescent time (Wilder et al. 2004b). Patterns at local scales also underscore the lack of congruence between the two sexes. Several explanations have been proposed to explain these patterns, including differences in mtDNA and Y chromosome mutation rates, as well as the possibility of different selective pressures shaping the diversity of these nonrecombinant genetic systems. The discrepancy observed between the two uniparental genetic systems may also result from sex-specific behaviours, and several studies point to differences in migration rate of men and women as being the main explanatory variable. The recent advent of high-density genomic data, and more precisely the comparison of genetic patterns for autosomal and sex chromosomes at global scales, allows a revisitation of these questions and confirm that sex-specific behaviours may indeed shape genetic diversity in the human genome. We feel it is timely to establish a state of the art of what we know on this topic.
Insights from Y chromosome and mitochondrial DNA differentiation patterns
The first studies investigating sex-specific genetic diversity reported a higher level of genetic differentiation for the Y chromosome than for mitochondrial DNA (Seielstad et al. 1998; Stoneking 1998). In these studies, genetic differentiation levels between populations were estimated by FST, which measures the proportion of the global genetic diversity that is explained by differences between populations. The authors found that Y-Chr FST were several times higher than mtDNA FST. It has been later argued that these smaller FST found for mtDNA did not reflect a difference in behaviour between men and women but were rather because of a bias in the choice of genetic markers (Wilder et al. 2004a). When unbiased markers were used, no difference was found between mtDNA FST and Y-Chr FST. More generally, comparing mitochondrial DNA and Y chromosome is extremely tricky because differences between them could also reflect selective pressures acting differently on these two genetic systems, as well as the great disparity in their molecular characteristics (such as size, copy number, cellular localization, coding proportion and mutation rate) (Cummins 2001). To circumvent this molecular bias, Balaresque et al. (2006) chose markers that are comparable in terms of molecular characteristics and that are located in similar genomic contexts: they chose markers on the X and Y chromosomes that are homologous because of a common autosomal origin. Their results, based on a data set comprising populations from different geographical areas worldwide, revealed that between-population differentiation is always greater for the Y chromosome than for the X chromosome.
From sex-specific behaviours to sex-specific genetic patterns
Which sex-specific behaviour(s) may have given rise to these lower levels of genetic differentiation observed for female-specific makers compared with male-specific markers? The first explanation given by Seielstad et al. (1998) and by Stoneking (1998) was that women disperse more than men; in other words, women are the moving sex. Indeed, FST values are lower when two populations are genetically similar. Populations become more genetically similar when they exchange migrants. If women migrate more than men, female-specific genetic differentiation between populations would be therefore lower than male-specific genetic differentiation. However, an alternative explanation is possible: men-specific markers could be more driven by genetic drift than women-specific markers. Genetic drift tends to decrease the genetic diversity within populations and to increase the differentiation between populations. Larger FST values observed for Y-Chr could therefore result from a larger magnitude of genetic drift acting on men-specific markers. It is well known that the magnitude of genetic drift is inversely proportional to the effective population size Ne. Consequently, FST depends inversely to the product of effective population size and migration: when more migration occurs, this product is larger and FST is lower. When the effective population size is smaller, FST is higher. The mathematical relationship between FST, effective population size and migration is well known under the island model (Wright 1931). While this island model may not be the most realistic for human populations, as it assumes equal migration between all populations whatever the distance between them, it is the most commonly used in human population genetics studies. In any case, while the exact relation between FST, effective population size and migration rate may differ between models, the inverse relation between FST on the one hand and effective population size and migration rate on the other hand should remain valid whatever the model.
We review below the literature regarding these two alternative explanations. In particular, we show that differences between men and women migration rates are surely not the only factor that matters, but that differences in effective population size between men and women are also likely to shape diversity patterns of the human genome. Then, we discuss three processes that may explain sex-specific differences in effective population size: (i) variance in reproductive success that can arise from polygyny; (ii) descent rules; and (iii) transmission of reproductive success. We hope to shed light on the influence of sex-specific behaviours on the evolution of human genomic diversity and will encourage and guide future interdisciplinary research on this topic. We also hope that this review will contribute to the development of more realistic models of human genetic evolution that will depart from the oversimplistic assumption of random mating and integrate human socio-demographic specificities.
Genetic evidence for sex-specific migration rates
The first hypothesis proposed for explaining higher Y-chr than mtDNA differentiation between populations is that women would have migrated more than men during human history (Seielstad et al. 1998; Stoneking 1998). This would be due to patrilocality: a residence rule that forces women to settle in the same residential area as their husband’s family. According to this rule, men reside continuously in the same location, while women migrate. Distribution of residential rules among populations shows that residence is patrilocal in 70% of human populations (after getting married, the woman settles down with her family-in-law), matrilocal in 20% of the cases (the man settles down with his family-in-law), and neolocal (the spouses settle down together in a new place) or multilocal (they change residences throughout their life) in the remaining cases (Burton et al. 1996; Godelier 2004). Although patrilocality is common nowadays, this is not an exclusive behaviour in humans and the extent to which this behaviour was dominant in the past remains unclear.
At local scales, sex-specific differentiation patterns often correlate with residence rules
To test this hypothesis, Oota et al. (2001) compared matrilocal vs. patrilocal populations in the hill tribes from Thailand. Their study was based on the comparison between Y-Chr and mtDNA diversity. They show that genetic differentiation was higher for mtDNA than for Y-Chr in matrilocal populations where women are the phylopatric sex, yet the opposite result was found in patrilocal populations where men are the phylopatric sex. This remarkable study showed the influence of social organization on genetic diversity at a local scale. This study was further confirmed in several geographical areas (see Bustamante & Ramachandran 2009; Segurel et al. 2008; Wilkins & Marlowe 2006 for reviews), although exceptions may occur. In particular, Besaggio et al. (2007) showed that differences in diversity between mtDNA and Y-Chr among populations in Northern Thailand result from difference in residential rules, with the exception of the Hmong Mien. According to the authors, this exception is explained by a particular cultural trait of the Hmong Mien, who adopt children from neighbouring communities to enlarge their labour force, with the consequence of maintaining a constant inflow of non sex-specific genetic diversity. Kumar et al. (2006) compared the genetic diversity of matrilocal populations from India with that of the surrounding patrilocal populations. No difference in haplotype diversity was found with surrounding patrilocal populations because none of them was exogamous enough: migration was too low to have an impact on genetic differentiation. Gunnarsdottir et al. (2011) did a similar study in Sumatra and found differences in mitochondrial genetic diversity between matrilocal and patrilocal populations when they used complete mtDNA sequence. Therefore, they claimed that their study ‘highlights the importance of using complete mtDNA sequences for such analyses, as using only partial sequences (as done in previous studies) can give misleading results’. Nevertheless, their study also showed that when using mean pairwise differences rather than haplotype diversity with HVS1 sequences, significant differences were also found between the matrilocal and patrilocal populations. This shows that the HVS1 sequence is informative enough only when the appropriate parameter is used. One must also note that Kumar et al. (2006) computed an analogous measure of the mean pairwise diversity (Da distance) and showed no differences between matrilocal and patrilocal with this parameter. Thus, even if these studies confirm that sex-specific migration rates associated with resident rules are partly shaping sex-specific genetic diversity, other cultural traits may also be important at the local scale.
Genetic differentiation patterns depend on the scale of the study and on past residential rules
Interestingly, Wilkins & Marlowe (2006) showed that the geographical scale at which genetic diversity is estimated also matters. At a local scale, Y-Chr FST values are larger than mtDNA FST in patrilocal societies, but this result does not hold when comparing more distant populations: the excess of Y-Chr differentiation in comparison with mtDNA decreases with geographical distance. The proposed explanation is that when comparing geographically distant populations, genetic data trace back to an older common ancestry than when comparing nearby populations. In other words, when comparing geographically distant populations, the genetic variability under study is more influenced by migrational processes that have occurred in a more remote past. Assuming matrilocal populations were more frequent in the past, this could explain why the excess of Y-chr FST over mtDNA FST is not found when comparing geographically distant populations. In this case, matrilocality may have been the main process shaping sex-specific diversity and Y-Chr differentiation can even become smaller than mtDNA differentiation.
What do we know about past residence rules?
This hypothesis is consistent with the work of Holden & Mace (2003) who addressed the question of matrilocal ancestry of the Bantu- and Bantoid-speaking populations from sub-Saharan using a phylogenetic approach. The authors showed that the ancestral state of most of these currently patrilocal and matrilocal populations was matrilocality, with a shift to patrilocality concomitant with the arrival of animal husbandry. Jordan et al. (2009) reached the same conclusion for Austronesian societies with a similar method. Other studies showed that most forager societies are multilocal, with no tendency for patrilocality (Marlowe 2004). If we assume that these populations are a good proxy for our Palaeolithic past, then patrilocality would be recent in our history, perhaps occurring during the Neolithic. In Europe, Fortunato (2011b) used cross-cultural data into a phylogenetic framework and inferred an ancestral state of Indo-European populations in Neolithic to be patrilocal or neolocal but not matrilocal. Ancient DNA analysis of Neolithic remains, located in France and dated at 5000 years old, confirmed this result by showing a smaller Y-Chr diversity than mtDNA diversity for these samples (Lacan et al. 2011).
Looking back further in the past of our most closely related species (the Neanderthals), Lalueza-Fox et al. (2010) provide a remarkable analysis of the mitochondrial diversity from 12 Neanderthal remains found in a 49 000-year-old grave in Spain. Among the six adults of the sample, the three males carried the same mtDNA lineages, whereas the three females carried each a different mtDNA lineage. They concluded that these men were kin having mated with immigrant women, thus suggesting that Neandertal was patrilocal. However, an alternative explanation could be that Neanderthal was matrilocal, with male-skewed kin migration processes. Indeed, this study also showed that mtDNA diversity is significantly lower in Neanderthal than in any random subsample of sequences from unrelated modern Europeans. This suggests that mitochondrial diversity was low within Neanderthal groups, a result consistent with matrilocality, rather than patrilocality. So, their conclusion that Neandertal was patrilocal requires further studies for a firm demonstration.
Finally, going into a much deeper past, using ratios of strontium isotopes to identify the geological substrate on which lived Australopithecus africanus and Paranthropusrobustus in South Africa, Copeland et al. (2011) showed that individuals who have spent their lifetime in the place where they were buried had larger teeth than individuals who moved during their lifetime. Given the strong sexual dimorphism of these species, they associated small teeth to female and large teeth to males, and thus proposed that Australopithecus females dispersed more than Australopithecus males, supporting the view that patrilocality is very ancient.
In conclusion, it is still not clear whether our ancestors were matrilocal, patrilocal or multilocal, notwithstanding the fact that perhaps changes occurred several times during our history. As a consequence, the hypothesis that matrilocality explains differences in genetic structure according to the geographical scale remains unanswered. However, at local scales, there is strong evidence across several studies that residence rules shape sex-specific genetic differentiation.
Genetic evidence for sex-specific effective population size
FST also depends on effective size
The fact that women are genetically less structured than men has been typically explained by a higher migration rate of women, in relation to the dominant tradition of patrilocality in present-day populations. Yet, differences in effective population size between men and women could be an alternative explanation, as the population differentiation parameter FST depends inversely on the product of the effective number of individuals within each population and the migration rate among populations. Indeed, effective population size influences the amount of genetic drift: populations with smaller effective size tend to have lower level of intrapopulation diversity and higher level of interpopulation differentiation. Consequently, a smaller effective number of men in comparison with women could explain the discrepancy between Y chromosome and mtDNA level of differentiation. It could explain in the same time the younger Y chromosome-based coalescence time (c. 100 000 years old and younger) in comparison with mtDNA-based coalescence time (c. 200 000 years old), as the coalescence time is proportional to the effective population size (see Wilder et al. 2004b and references therein).
Sex-biased dispersal and differences in male and female effective population size are therefore intermingled in giving rise to genetic structure measured for uniparentally inherited markers. To untangle these two components, Balaresque et al. (2006) combined Y-Chr and X-Chr FST in an island model and then estimated that a wide range of migration and demographic scenarios is compatible with the observed level of sex-specific genetic differentiation. Possible scenarios range from a sex-specific migration process with the female migration rate being three times larger than the male migration rate, to sex-specific demographic process where the female effective population size has to be 11 times higher than male effective population size.
Autosomal and X chromosome genetic variation patterns at a regional scale suggest differences between male and female effective population size
Another way to untangle these two components is to jointly analyse autosomal and X-linked markers, and to investigate differentiation patterns for these markers. In the absence of differences in male and female effective population size, X chromosome differentiation is expected to be always larger than autosomal genetic differentiation. These methods also have the advantage of limiting the problems associated with sex-specific selective pressures that can occur on the nonrecombining uniparental markers because of hitch-hiking effects. However, it may not completely circumvent this problem of selective pressures. Recent positive selection is indeed expected to be more effective for recessive mutations on the X chromosome than on the autosomes, because males will express any X-linked recessive mutation. Nevertheless a recent study by Gunnarsdottir et al. (2011) using whole genomes did not find such a selective effect but rather concluded that demographic factors explain the differences between X and autosomal genetic diversity. Using autosomal and X chromosome markers, the ratio of male/female effective population size and male/female migration rate that are compatible with the data can be estimated. Using this approach, Segurel et al. (2008) showed that the X and autosomal differentiation patterns among Turkic populations from Central Asia result not only from a higher migration rate of women, but also from a sex-specific difference in effective population size that is higher for females than for males. On the other hand, comparing autosomal and X-Chr genetic patterns in a worldwide population sample, and assuming a pure divergence model, Ramachandran et al. (2004) showed no differences in effective population size between men and women at this much broader geographical scale. The difference between the two studies cannot be explained by the method used: when applying Ramachandran et al. (2004)’s method to Central Asia, Segurel et al. (2008) confirmed the higher migration rate and effective size for women in these Turkic populations. One explanation for no sex-specific differences at a worldwide scale but a strong one at the regional scale could be the geographical scale factor expressed by Wilkins & Marlowe (2006). Performing similar studies in other parts of the world would certainly be interesting.
At a global scale, autosomal/X chromosome genetic variation studies yield conflicting results
A more global approach has been undertaken to compare male and female effective population size of our species. Hammer et al. (2008) and Keinan et al. (2009) estimated the ratio of genetic diversity between the X chromosome and autosomes. X chromosome is present in two copies in females and in one copy in males, whereas autosomes are present in two copies in both sexes. Consequently, in the absence of any sex-specific process, the expected ratio of diversity between X chromosome and autosomes markers is 0.75. The two studies yielded contradictory results. Hammer et al. (2008)’s study of six worldwide population samples yielded a ratio higher than 0.75 and therefore concluded that female effective population size is larger than male effective population size. By contrast, Keinan et al. (2009) found a ratio below 0.75 in populations outside Africa and concluded to an accelerated genetic drift on the X chromosome during human migrations out of Africa because of a female bottleneck. As pointed out by Bustamante & Ramachandran (2009), these two studies differ in the data used: Hammer et al. (2008) had a limited number of genetic data on a large number of individuals, while Keinan et al. (2009) studied a limited number of individuals but with a large genetic coverage. To test for a selective hypothesis, Hammer et al. (2010) contrasted, for a given genomic region, the ratio of X chromosome to autosome diversity with the genetic distance of this region to the closest gene on the X chromosome. They showed that this ratio is lower when the region under study is closer to a gene. Thus, selection could in part explain this X/autosome ratio. However, Gunnarsdottir et al. (2011) refute this explanation by analysing 68 whole genomes and concluded that demographic factors, rather than selective factors, explain the X/autosomal genetic diversity ratio.
Methodological issues and demographic scenarios that may bring together autosomal/X chromosome variation studies
Emery et al. (2010) recently proposed that the lack of congruence between these two studies comes from differences in the summary statistics used by the two teams. In particular, they showed that the one used by Hammer (based on ratio of diversity estimates) detects the differences in sex-specific effective population size that occurred on a longer timescale than the statistics used by Keinan et al. (2009) (based on ratio of FST values). They concluded that extant patterns of human genomic variation are consistent with both an earlier time when female effective population size was larger than male effective population size and a ‘recent time’ when male effective population size was larger than that of females. This larger male effective population size is only found when comparing African vs. non-African populations but not when comparing Eurasian populations. Indeed, a study by Keinan & Reich (2010) has shown that a model of continuous male-favoured migration from Africa into non-African populations before the split of Asians and Europeans can account for the magnitude of the larger effective population size observed for males. This process can further be amplified by the fact that bottlenecks do not have the same impact on X and autosomes and can lead to X vs. autosome bias in the ratio of genetic diversity (Pool & Nielsen 2007). It has to be noted that Emery et al.’s (2010) model is a population split model without further migration. Interestingly, this lower female effective population size following the colonization of a new territory is also found at a more recent time scale during the colonization of Quebec in the XVIIth century (Tremblay & Vezina 2010). Using genealogical data, they demonstrated that female effective population size is half the effective size of males. Indeed, one-third of the founders were women. Moreover, in the subsequent generations, there were more fluctuations of female lineages than of male lineages, which further decreased female effective population size.
Labuda et al. (2010a) addressed these questions by measuring the ratio of population recombination rates between the X-Chr and the autosomes, rX/rA. The rationale of this approach is that the X-Chr recombines only in female meiosis, whereas autosomes undergo crossovers in both sexes. The ratio rX/rA thus reflects the female-to-male breeding ratio. After correction, they found a breeding ratio larger than two in Africa and Europe and between one and two in East Asia, thus providing another evidence for a sex differences in effective population size (Labuda et al. 2010b).
In summary, almost all studies converge towards the view that both sex-specific migration and sex-specific differences in effective population size influence human genomic variation, although future studies will help further clarify the relative importance of these two sex-specific factors. In the second part of this article, we present three socio-demographic processes that can explain sex-specific differences in effective population size. The first one is the larger variance in male reproductive success resulting from polygyny. The second is the existence of particular ‘descent rules’ in some human populations. The third is the cultural transmission of reproductive success.
One of the critical parameters that can explain a lower effective population size for males as compared to females is a difference in variance in reproductive success between males and females. Indeed, a larger variance in reproductive success decreases the effective population size (Kimura & Crow 1963). Data on human lifetime reproductive success are scarce, but when they exist, they all show a higher or equal variance in reproductive success among males compared with females.
The first proposed explanation is polygyny. Polygyny creates the demographic conditions for a larger variance in reproductive success in males compared with females, because females are always the limiting sex for reproductive success because of their limited period of fertility and their strong and time-consuming investment in each offspring (pregnancy, breastfeeding, etc.) Because our species exhibits sexual dimorphism, even at a moderate level, mild polygyny is thought to characterize our species’ evolutionary past (Alexander et al. 1979). Moreover, ethnological studies showed that polygyny is a frequent practice in human societies. Revising the classification of Murdock (1967), Marlowe (2000) proposed the following distribution among populations: 17% of populations are monogamous, 51% are slightly polygynous, 31% are generally polygynous and 1% are polyandrous. It should be noted that there are several measures used to evaluate the intensity of polygyny: Hartung (1982)’s measured the percentage of polygynously married women, Betzig (1986) measured the maximum harem size and Low (1988) proposed a new measure combining the percentage of polygynously married men with the percentage of polygynously married women. He also compared different measures of polygyny and showed reassuringly strong correlations between these measures.
Interestingly, White (1988) distinguished two types of polygyny. First, ‘wealth-increasing’ polygyny described societies where women’s labour generates wealth, and most men are able to become polygynists with age conditional on their survival. The second is called ‘exceptional men’ polygyny and is characterized by the fact that most of the family resources are generated by the husband. In this case, polygyny usually depends on the exceptional productivity of particular men (such as hunters or shamans). As we will show, the distinction between these two types of polygyny, that is, ‘wealth-increasing’ and ‘exceptional men’, is important as it does not have the same impact on the variance in reproductive success.
Polygyny influences the variance in male reproductive success, but less often than thought
A key phenomenon in polygyny is that when few males reproduce more, some males cannot find a mate. Frequent polygyny is expected to create a shortage of wives, and thus to increase the variance in male reproductive success, but it does not always do so. First, demographic regulation of polygyny occurs indeed when the age at marriage of males is delayed. Pison (1986) performed a detailed demographic study of the Peul in Senegal. He showed that delayed age at marriage can compensate for the sex ratio bias created by polygynous behaviours in an expanding population. Indeed, in such populations, young age classes are always found at larger proportions than older age classes. Men can therefore marry several younger women without creating a shortage of females for the other males. In this case, we do not expect a strong variance in male reproductive success.
Pison’s study further showed that polygyny correlates with age so that a high proportion of males eventually become polygynous. A transversal analysis of these data shows that 26% of adult males do not have a wife when 31% have more than one, and thus, the estimated variance in mating success is large. Longitudinal studies show, however, that almost all men will eventually marry, at first with one wife, then with more wives as they get older. While polygyny is common in these populations (85% of the males surviving to age 50 years have at least two wives), celibacy is extremely low in these populations: the percentage of male without any wife after 40 years is <2%. From White’s (1988) analyses, this type of polygyny is the most frequent among the societies described as ‘full polygyny’ and account for more than ¾ of these societies.
Secondly, social regulation can exist: among the Pawnee, who have high rates of polygyny, polyandry was also common and allowed mature reproductive women who had married young as second wives to take young men as second husbands, while providing a first wife to young men who would not otherwise have one (White 1988).
Polygyny may also, in certain cases, influence the variance in female reproductive success
We may also wonder whether polygyny also affects the variance in female reproductive success, decreasing thus the female effective population size. The reproductive costs or benefits of polygyny are difficult to assess because polygynously married women may show decreased fertility. For example, in Africa, it has been documented that monogamously married women are 1.5 times more fertile than polygynously married women (Brass et al. 1968). Several factors have been cited as possible explanations for the higher fertility of monogamous wives. One explanation is that infertility in woman precipitates polygyny in men (Pison 1986). A second explanation is based on a detailed study on longitudinal data showing that the level of fertility of a woman decreases with the age of her husband (Garenne & Walle 1989) and that age-related male reduced fertility plays a much larger role than a lower frequency of intercourse. Therefore, in the case of ‘age-related’ polygyny, second or third wives tend to have a lower reproductive success. A third explanation is that infant mortality of polygynous wives is higher in some cases (Strassmann 2003). All these factors can potentially increase the variance in female reproductive success while not substantially increasing that of males (Fix 1976).
Nevertheless, a decrease in reproductive success of females in polygynous marriages does not always occur. Borgerhoff Mulder (1988) showed no effect of marital status on female reproductive success in the Kipsigis. Her data show that completed family sizes are smaller among women married to older husbands than among those married to men closer to their own age. This effect was primarily because of higher rates of offspring mortality in the former group. However, the co-wife cooperation in polygynous marriages can confer benefits that offset the disadvantage of being married to an older man (Borgerhoff Mulder 1990).
Furthermore, as for the men, the female celibacy rate is extremely low in these societies and widows and divorced women remarry quickly. For example, in Pison’s (1986) study, the celibacy rate is <1% for females older than 25 years, even if 12% of women become widows in the 10 years following their first marriage and if 26% of women divorce in the first 3 years of their first marriage. All women spend their full lifetime reproductive period married, thus decreasing the variance among women that would have occurred because of difference in the length of the period during which they are married.
We can conclude that, for men, polygyny may not increase variance in reproductive success as often as commonly thought. For women, it is far from clear whether polygyny tends to increase or decrease variance in reproductive success. It is likely to vary strongly between populations according to the type of practiced polygyny and the demographic regime.
Does polygyny increase the ratio of male/female variance in reproductive success?
In an attempt to evaluate the potential intensity of sexual selection, Brown et al. (2009) present data regarding male and female variance in reproductive success in societies where they found demographic data. The critical variable to estimate is the excess of variance in male reproductive success over variance in female reproductive success. As pointed out by Clutton-Brock (1988), most studies comparing male and female reproductive success have relied on short-term data, as complete data on lifetime reproductive success are scarce in humans. In their study, Brown et al. (2009) present data on 18 human populations. They show that the variance in reproductive success is on average higher for male than for female in polygynous populations, but that this ratio varies greatly between human populations: from 0.8 in Finland to 4.75 in an extremely polygynous population of the Dogon in Mali (Brown et al. 2009). They also show that the social norm of monogamy vs. polygamy is not a good predictor of the variance in reproductive success. For example, monogamous societies with extensive serial monogamy have ratios similar to those of polygynous societies. By contrast, some polygamous societies exhibit a ratio close to one. In her recent study of a serial monogamous society, Borgerhoff Mulder (2009) showed that males and females have the same variance in reproductive success because serial monogamy leads not only to serial polygyny but also to serial polyandry. Conversely, Jokela et al. (2010) showed that in monogamous societies, variance in offspring number was 10% higher in men, because of serial monogamy. It has to be noted that most of these studies did not calculate lifetime reproductive success but the number of surviving offspring of married adults and that no data are available on the proportion of these offspring that will eventually get married. Based on Brown et al.’s (2009) data, Fortunato & Archetti (2010) compared polygynous vs. monogamous and show no significant differences in the male variance in reproductive success.
To what extent can polygyny have an impact on sex-specific effective population size?
The extent to which polygyny might explain sex-specific genetic differentiation is difficult to assess. For example, in a phylogenetic study that aimed to reconstruct the history of marriage strategies in Indo-European-Speaking societies, Fortunato (2011a) showed that the ancestral state for Indo-European societies is monogamy, a pattern that can be traced back to the Neolithic. Nevertheless, Europe also shows an increase in Y chromosome differentiation compared with mtDNA differentiation (Seielstad et al. 1998). Fortunato et al.’s inference excludes polygyny as being the explicative factor for this sex-specific difference in genetic diversity. Nevertheless, one explanation could be that the monogamous ancestral behaviour was only social, with serial monogamy being the norm, which would have increased the male variance in reproductive success and hence decreased male effective population size. Conversely, ancient polygyny perhaps left traces on European genetic diversity as suggested by Dupanloup et al. (2003). However, it is difficult to infer how strong polygyny was in the past. For example, what we see now in Africa may have been changed by European colonization, which reduced wealth in some colonized countries, and thus may have decreased the level of polygyny achieved through wealth accumulation. Conversely, in some cases, colonization has led to the injection of earned wealth and has increased the level of polygyny as documented by Borgerhoff Mulder (1988) in the Kipsigis. If present-day forager societies are a good proxy for our past social organization, we were a mildly polygynous species (Marlowe 2003).
Using Wade’s formula (Wade 1979), we can evaluate the expected impact of polygyny on effective population size. Table 1 presents these values for different levels of polygynous behaviour.
Table 1. Estimated ratio of male/female effective population size for different levels of polygyny
|Mildly polygynous societies—forager societies||5% with 2 wives |
90% with 1 wife
5% with 0 wife
|0.1||2.0 (2.4) |
|Mildly polygynous societies—forager societies||10% with 2 wives |
80% with 1 wife
10% with 0 wife
|0.2||2.0 (2.8) |
|Strong polygyny—Dogon|| || ||3.47 (7.3) |
|Super male||1% with 10 wives |
90% with 1 wife
9% with 0 wife
|0.9||2.0 (5.6) |
This table shows that polygyny alone cannot explain the difference in male and female effective population sizes. Indeed, the worldwide differences in mtDNA or X-chr differentiation vs. Y chromosome are compatible with ratios of male/female effective population sizes (assuming sex-specific migration rates being equal) ranging from 8.5 (Seielstad et al. 1998) to 11 (Balaresque et al. 2006). These ratios are much higher than the ratios in male/female effective population sizes that even the most extreme scenario of polygyny presented in Table 1 could explain. This shows that polygyny alone cannot explain the sex bias in levels of genetic differentiation. However, it can contribute to it, through interactions with patrilocality, descent rules and transmission of reproductive success, as discussed in the following sections. Polyandry could have a symmetrical effect by increasing the variance in female reproductive success, but we expect this effect to be weaker.
Human societies do not differ only in the level of polygyny and in their residence rules. Other parameters of social organization may also influence the sex-specific genetic diversity such as descent rules. Indeed, descent is cognatic in 39% of the populations, which means that the children belong to the kin groups of both parents, the kind of family organization most represented in industrialized societies. Alternatively, descent can be unilineal: in 45% of the populations, the children are affiliated to the kin group of the father (patrilineal) and in 12% of the populations to the kin group of the mother (matrilineal) (Godelier 2004). In some rare cases, descent can be bilineal.
Descent rules may affect sex-specific genetic patterns at a regional scale
To understand the influence of descent rules on the evolution of genetic diversity, one study compared two kinds of societies in Central Asia that have different descent rules (Chaix et al. 2004): Turkic populations have a patrilineal descent, while Indo-Iranian populations have a cognatic descent. More precisely, Turkic societies are organized into so-called patrilineal descent groups, that is, lineages, clans and tribes. Several lineages form a clan, several clans form a tribe. Individuals belonging to the same descent group (lineage, clan or tribe) claim to share a recent common ancestor on the paternal line. The individuals are usually able to trace the genealogies back to their lineage’s ancestor (7–10 generations depending on the populations), but not to their clan or tribe’s ancestors. In addition, lineages and clans are exogamous entities, whereas tribes are endogamous entities. On the other hand, Indo-Iranian agriculturist societies are organized in cognatic families, and marriage rules are based on kinship and geographical proximity with a strong preference for first-cousin marriages. Both societies are patrilocal and mildly polygynous.
In the Turkic patrilineal groups, the comparison between genealogical data and Y chromosome data showed that this social organization is mirrored by the genetic structure. Indeed, the Y chromosome-based kinship coefficient is higher among individuals belonging to the same descent group than among individuals chosen at random in the population, demonstrating that the individuals from the same descent group share a common ancestor on the paternal line that is more recent than the ancestor of the population. This is true at the lineage level and in most cases at the clan level, in agreement with the story told by oral tradition in these populations. This is however not true at the tribe level, showing that links between individuals belonging to the same tribe are cultural rather than biological (Chaix et al. 2004).
This patrilineal social organization has a strong impact on the intrapopulation genetic diversity and interpopulation genetic differentiation. Turkic populations, but not Indo-Iranian populations, exhibited a significant loss of intrapopulation genetic diversity for their Y chromosome. Chaix et al. (2007) showed that the average number of individuals carrying the same Y chromosome haplotype was much higher in patrilineal Turkic populations than in cognatic Indo-Iranian populations, but no difference was observed at the mitochondrial level. Regarding population differentiation, the overall genetic differentiation was much higher for the Y chromosome, as compared to mitochondrial DNA, among the patrilineal Turkic populations vs. . In cognatic Indo-Iranian populations, the difference between Y chromosome differentiation and mitochondrial differentiation is less strong vs. (Segurel et al. 2008). Because the main difference between Turkic and Indo-Iranian social organization is the rule of descent, with the Turkic populations having patrilineal descent and the Indo-Iranian populations having cognatic descent (both groups are patrilocal and mildly polygynous), it is likely that the larger discrepancy between Y-chr and mtDNA FST observed in Turkic populations, as compared to Indo-Iranian populations, results from the patrilinearity of Turkic populations. This shows that patrilocality and polygyny are not the only parameters of social organization that may shape sex-specific genetic differentiation patterns, but that descent rules also matter.
How do descent rules influence male and female effective population sizes?
To understand the processes through which descent rules influence the evolution of genetic diversity, one has to take into account the dynamics of patrilineal social organization, involving descent group fissions, fusions and extinctions. Whenever a fission event occurs, the descent groups are not formed randomly and related men tend to cluster together. This particular dynamics of lineal fissions increases the relatedness among men belonging to the same descent group and decreases the relatedness between descent groups. In the short term, this dynamics of lineal fission may increase the genetic diversity of the Y chromosome, with different alleles being fixed in different descent groups, but because of the inherent demographic stochasticity of such organization, some of the descent groups will go extinct, thus reducing the genetic diversity of the Y chromosome. The mitochondrial diversity is not (or less) affected by such local extinctions because marriages are exogamous, and migrations of women between descent groups tend to homogenize the mitochondrial diversity among descent groups. An analogy can be made between the Y chromosome diversity of patrilineal populations and the so-called ‘several small populations’ case of the SLOSS model (Frankham et al. 2003), whereas the mtDNA diversity can be compared with the ‘single large population’ case. Such dynamics of lineal fissions, associated with local extinctions, can explain the reduction in the effective number of men, as compared to women. The effect of fission–fusion on effective population size has first been documented in the Yanomama from South America (Smouse et al. 1981).
These observations support the hypothesis that descent rules may influence the genetic diversity of human populations at a local scale. As patrilinearity is more common than matrilinearity in present-day human populations, it is possible that descent rules generally reduce the effective number of men relative to women, thus contributing to the higher Y chromosome differentiation observed at large geographical scales. However, unilineal descent is not widespread among present-day hunter–gatherers (39%), while is found at higher frequency in horticultural and pastoral societies (68% and 75% respectively) (van den Berghe 1980). Consequently, it is likely to have emerged recently relative to the timescale of human evolution, along with plant and animal domestication. Assuming that patrilineal descent is a recent invention, similarly to patrilocality, the question of its influence of human genomic diversity at large geographical scale remains open.
Transmission of reproductive success
A key parameter that is still too seldom studied is the transmission of reproductive success. As we will see, this transmission can strongly reduce the effective population size (Nei & Murata 1966; Sibert et al. 2002).
Is the intergenerational transmission of reproductive success genetic or cultural?
Although it can occur through genetic transmission, we expect this process to be widespread in humans because of widespread cultural transmission. The debate around biological vs. social transmission of fertility is a long-standing debate that started in the 19th century. Murphy (1999) gives an excellent review of this debate. More recently, it has been documented that fertility could be genetically transmitted: in the Huterrite, Pluzhnikov et al. (2007) give strong arguments for a genetic heritability of fertility. Nonzero additive genetic variance has also been estimated in some pre-industrial populations based on historical records (Pettay et al. 2005).
There are several reasons why we should also expect the transmission of reproductive success to be cultural. First, the fact that this reproductive success could be culturally transmitted is expected as in humans, founding a large family can be seen as a cultural behaviour related to family structure, and it has been shown that cultural traits related to the family are mainly transmitted by the parents (Borgerhoff Mulder et al. 2006).
Secondly, it has been shown in many human societies that reproductive success correlates with wealth, either material, embodied and relational wealth. Wealthier individuals, particularly men, have higher reproductive success than less wealthy ones (see also Nettle & Pollet 2008 for a review), so that resources inherited by children from their parents can be an important determinant of their future reproductive success (e.g. Low 1991; Mace 1996). A recent study also showed that even in premodern societies representing four production systems (Hunter–gatherers, horticulturalists, pastoralists and agriculturalists), wealth is passed on from one generation to the next (Borgerhoff Mulder et al. 2009; Smith et al. 2010). Three broad classes of wealth were defined for this study: embodied (body weight, grip strength, practical skills and reproductive success), material (land, livestock and household goods) and relational (social ties in food-sharing networks and other forms of assistance). The fact that reproductive success depends on wealth and that wealth can be transmitted leads to cultural inheritance of reproductive success through wealth transmission. Borgerhoff Mulder et al. (2009) computed the transmission of reproductive success in 11 populations. Among them, eight showed nonzero intergenerational correlation ranging from 0.066 to 0.213 with an average of 0.135. These values are similar to the low correlations between parental and offspring fertility found in many pre-demographic transition populations (Murphy 1999) and to those found in studies based on long-term historical records (Austerlitz & Heyer 1998; Pluzhnikov et al. 2007; Helgason et al. 2003; Pettay et al. 2005). Conversely, no correlation of reproductive success was detected in the Dogon from Mali (Cazes 2009).
Thirdly, heritability in reproductive success can also be seen in term of kin cooperation. Individuals belonging to large kin networks may benefit from stronger social support, resulting in more offspring. This form of cooperative breeding, also known as allocare (i.e. the fact that not only the mother but other individuals in the society are important for child survival), is important in humans (Hrdy 2009; Kokko et al. 2002). According to evolutionary theory, it is expected that these helpers are close kin (Clutton-Brock 2002), and this will create a correlation in reproductive success from one generation to the next.
One of the major criticisms about the existence of transmission of reproductive success comes from the work of Fisher (1958),who demonstrated that any traits highly correlated with fitness should have a small heritability (zero heritability) as selection will erode the additive variance. Reproductive success is highly correlated with fitness; therefore, no heritability of this trait is expected according to Fisher’s model. However, this model does not allow for a high mutation rate that could restore genetic variance at each generation. The potential importance of mutation for restoring additive variance has been acknowledged in long-term selection experiments where additive genetic variance seems never to erode (Lynch & Walsh 1998). Transmission of reproductive success can be described in its simplest way with a selection-mutation model, where ‘mutation’ stands for the error in the transmission of the cultural trait. This simple population genetic model shows that with a high ‘mutation’ rate, equilibrium with polymorphism can be achieved. This ‘high mutation rate’ is particularly relevant for fertility transmission, which is probably much less rigid than genetic transmission.
Intergenerational transmission of reproductive success can be detected from genealogical or genetic data
In practice, the transmission of reproductive success can be measured with either demographic or genetic data. The demographic data needed for measuring this process should span two generations, which is difficult to get in the field. To circumvent this problem, we have designed a method based on genetic data that infers the existence of such transmission (Blum et al. 2006; Sibert et al. 2002). This method is based on the imbalance shape of the coalescent tree under fertility transmission (Sibert et al. 2002). It has been applied to uniparental mitochondrial data in a worldwide set of populations, showing a higher fertility transmission in forager societies as compared to food producers (Blum et al. 2006). One explanation of the higher fertility transmission observed in foragers is that female reproductive success is more dependent on female kin cooperation in foragers than in food producers. This female kin cooperation cannot be explained by a bias in matrilocality for the forager societies as all of them were either multilocal or patrilocal. Alternatively, social status can be transmitted through the female line, as for instance in the Maoris (Murray-McIntosh et al. 1998). Using another approach based also on genetic data, Lansing et al. (2008) show no evidence of male dominance transmission, but there are some concerns over their methods (Heyer 2011). Finally, it should be noted that this process is not specific to human populations. Cultural and/or genetic transmission of reproductive success has also been documented in some animal species (Heyer et al. 2005), such as whales (Whitehead 1998), dolphins (Frere et al. 2010) and cheetahs (Kelly 2001).
Why does the transmission of reproductive success influence sex-specific genetic patterns?
It is well documented that transmission of traits linked to reproductive success are often sex-biased. Several societies transmit wealth in an asymmetrical way. Hartung (1982) used Murdock’s (1967) data to show that high male bias for inheritance of real property or movable property is common. Even when no material wealth is transmitted, relational or embodied wealth transmission can be sex-biased. These phenomena can lead to a sex-specific transmission of reproductive success. This phenomenon was first documented by Neel (1970) who reported that in Yanomami polygyny is transmitted from father to son and reduces the level of Y chromosome diversity. Conversely Murray-McIntosh et al. (1998) suggest that the low diversity of mtDNA in the Maori could be explained by the female transmission of social rank. In this society, social rank is a good indicator of reproductive success as children of high-ranking female have a higher survival rate. In several historical populations, sex-specific transmission of reproductive success has been measured (Pluzhnikov et al. 2007). Indeed, if a cultural trait is highly linked to fitness and if variance in reproductive success is stronger in one sex than another, we expect asymmetrical transmission of this trait to the sex with higher variance. This is analogous to Trivers & Willard’s (1973) rule, which states that under good conditions, parents should invest more in the offspring sex that has the highest variance in reproductive success through surviving and/or mating.
In the cases where the transmission of reproductive success is sex specific, it can induce differences in male vs. female effective population sizes, which will be amplified by a higher variance in reproductive success in the sex with the higher level of transmission. We therefore believe that when it exists, asymmetrical transmission of reproductive success is a major evolutionary force that can explain strong sex-specific differences in genetic diversity. This sex-biased transmission of reproductive success is likely to be common in human societies as sex-biased transmissions of cultural traits like wealth or prestige are common and are correlated with reproductive success. Furthermore, kin cooperation can be sex-biased.
Factors influencing sex-specific migration rates and effective sizes interact
All factors that we discussed here (residence, alliance and descent rules and transmission of reproductive success) interact in fact to various degrees in human populations and can lead to many different scenarios. For example, in Yanomami, the transmission of reproductive success is concomitant with the transmission of polygyny (Neel 1970). Moreover, the reduction in effective population size through the transmission of reproductive success is amplified by a higher variance of reproductive success: when this variance is higher, the reduction in Ne is higher for a given level of fertility transmission. For instance, in the Saguenay-Lac-Saint-Jean population, both factors were necessary to explain the high frequency of severe recessive disorders (Austerlitz & Heyer 1998). The interaction of both processes decreases the effective population size 20-fold, and the effective population size decreases through time despite the demographic growth of the population (Mourali-Chebil & Heyer 2006). In the Saguenay-Lac-Saint-Jean population, the estimated effective population size based on demographic data is 17 000. When variance in reproductive success is incorporated in the model, it decreases to 12 000, and when transmission of reproductive success is also included, it drops further to 900 individuals. In her study on cheetahs, Kelly (2001) showed the importance of the interaction of both processes: variance in reproductive success creates a ratio between effective population size and census size of 0.45 and with the transmission of reproductive success it further decreases to 0.15. Similarly, the variance in reproductive success because of polygyny may interact with descent rules and residence rules to further impact human genomic variation. Some societies are at once polygynous, patrilocal and have patrilineal clans, and the combination of these factors may lead to dramatic decreases in Y chromosome diversity, in comparison with mtDNA diversity, as observed in some ethnic groups of New Guinea (Kayser et al. 2003).
Local processes may create global effects
Island South-East Asia is also a good geographic example for how local processes may impact human genomic variation at large geographical scales. Indeed, large disparities between Y chromosome and mtDNA genetic markers were observed across societies of Island South East Asia and the Pacific: mtDNA markers show a higher proportion of Asian markers, while Y chromosome markers show a higher proportion of Melanesian ancestry (seeKayser 2010 for a review). This pattern was recently confirmed by X vs. autosome comparisons (Cox et al. 2010; Wollstein et al. 2010). It has been proposed that this rather dramatic asymmetrical pattern may be the result of a colonization and expansion process by Austronesians from the West, in which a small proportion of Austronesian women married with non-Austronesian men. The children of such matings have the Y chromosome of their non-Austronesian father, the mtDNA of the Austronesian mother, and as the result of the dominant matrilocality practised by most Austronesian societies in the past (Jordan et al. 2009), the children of such matings would have been incorporated into the Austronesian communities. Thus, it was shown that a simple model of just 2% marriage of Austronesian women with non-Austronesian men over a period of 50 generations, in the context of matrilocality, could explain the pattern of sex-specific diversity (Lansing et al. 2011). This example demonstrates how a local process of low intensity, here in relation to the residence rule, but maintained over a large number of generations, may impact human genomic variation at large geographic scales.
Directions for future studies
Other factors that we have not addressed in this review can also have an impact on sex-specific drift like kin migration (Fix 1999) or sex differences in generation time. Regarding the latter, the mean generation time that have been measured on genealogies are 28 years for females and 31 years for the males in Iceland (Helgason et al. 2003), 29.5 years for females and 33.9 years for the males in Quebec (Tremblay & Vezina 2000); in other societies, the difference in generation time can be higher, for example, the generation time is 27.2 and 40.2 for females and males, respectively, in the Peul from Senegal (Pison 1986). Fenner (2005) estimated a mean generation time of 25.6 for females and of 31.5 for males in hunter–gatherer societies. Thus, during the same time period, more drift occurs in the female lines than in the male lines, as the female generation time is smaller. As a consequence, mtDNA genetic diversity may decrease faster than its Y chromosome counterpart. It is also expected that generation time is positively correlated to variance in reproductive success because reproduction is conditional on survival: the later men reproduce, for instance because of later age at first or last reproduction, the largest the proportion of men that do not survive to achieve these reproductive events. This relationship between generation time and variance in reproductive success has not yet been explored and the way this relationship impacts effective population size remains unknown.
Moreover, many of the factors that we have tackled in this review need more attention to understand better their influence on human genomic variation. Further analyses of genealogical databases would be very useful. Such analyses would be especially valuable if genetic data are also available, allowing a comparison of both types of data. Conversely, when no demographic data are available, genetic polymorphisms could be investigated by themselves to detect these cultural processes (e.g. Blum et al. 2006; Lansing et al. 2008). In this context, more statistical and theoretical work is needed. On the other hand, a very large amount of genetic data is becoming available through genome wide analyses (high-density genotype data, full genome sequencing) and will increase our power to decipher complex cultural processes. A remaining question would be to assess to which extent these cultural processes may be somehow responsible for the low genetic diversity characterizing our species, in comparison with its closest relatives (Gagneux et al. 1999).
We are thankful to the two anonymous reviewers, Nicolas Perrin and Tim Vines for their helpful comments on this manuscript. This work is funded by ANR SoGen and by ANR Altérité Culturelle.
E.H., R.C. and F.A. are broadly interested in the inference of cultural, demographic and selective factors shaping the genomic diversity of human populations. E.H. and R.C. are mainly focused on field work and data analysis. F.A. is more interested in theoretical aspects and works also on other species. S.P. is a biodemographer interested in the coevolution of life-history traits with sociality in human. Currently, E.H. and S.P. are professor and lecturer (respectively) at the Museum National d’Histoire Naturelle (Paris, France), while F.A. and R.C. are researchers at the Centre National de la Recherche Scientifique (Paris, France).