The adaptive function of female multiple mating in socially monogamous species has been variously debated over the past few decades (Westneat et al., 1990; Birkhead and Møller, 1992; Westneat and Stewart, 2003), however so far no obvious conclusion has been reached (Wan et al., 2013). Proposed hypotheses include non-adaptive mechanisms, for example pleiotropic gene effects such as the outcome of selection on males (Forstmeier et al., 2011), or convenience polyandry where females submit to extra-pair copulations to avoid harassment (Forstmeier et al., 2014). Of the adaptive hypotheses proposed to explain female extra-pair mating behavior, the “good genes” and “compatible genes” hypotheses in which females gain indirect fitness benefits have been the most extensively examined (reviewed in Griffith et al., 2002; Neff and Pitcher, 2005; Akçay and Roughgarden, 2007; Griffith and Immler, 2009; Prokop et al., 2012).

At the genic level, the effects of good genes and compatible genes differ in their functional mechanism. Under selection for good genes, a male's genetic quality is additive and conveyed through genetic covariance with phenotypic indicator traits (Zahavi, 1975; Hamilton and Zuk, 1982; Kirkpatrick and Ryan, 1991; Andersson, 1994; Kirkpatrick, 1996; Kokko et al., 2002, 2003). Therefore, a trait indicating good genes benefits can be used as proxy for genetic quality, and hence for the heritable fitness benefits (in the form of either increased viability or attractiveness) that are conferred to offspring by the mate choice decisions of all individuals regardless of their own genotype. Under the good genes hypothesis, females are therefore predicted to allocate paternity toward males with extreme forms of the trait. In contrast, under the compatible genes hypothesis, not all females stand to derive the same benefits from mating with a particular male (Neff and Pitcher, 2005; Puurtinen et al., 2009). Instead, it is the compatibility of male and female genotypes that act, through either heterozygote overdominance or avoidance of the deleterious effects of inbreeding, to produce genetically superior offspring (Zeh and Zeh, 1996, 1997; Jennions, 1997; Tregenza and Wedell, 2000). Inbreeding depression is a consequence of intragenomic conflict caused by increased shared genetic similarity by common descent and lowered functional genome-wide heterozygosity (Coulson et al., 1998). These genetic consequences have the potential to influence fitness through their detrimental effects on offspring birth weights, growth, and survival (Keller and Waller, 2002). Thus, the compatible genes hypothesis predicts that paternity should be biased toward genetically dissimilar males.

Theory predicts that good genes and compatible genes benefits are not mutually exclusive and therefore that females should be selected to exploit both types of benefits in mate choice (Colegrave et al., 2002; Neff and Pitcher, 2005; Puurtinen et al., 2005, 2009). Yet few empirical studies have demonstrated simultaneous optimization of both types of benefit (but see Foerster et al., 2003). As a possible explanation, Colegrave et al. (2002) hypothesized that when selection for both good genes and compatible genes operates in a population, unless there are large numbers of males of equally high additive genetic quality available to all females, selection for compatibility will weaken directional sexual selection for good genes. It has therefore been argued that females cannot simultaneously select the most ornamented and most genetically compatible male, and that female optimization of both good genes and compatible genes benefits can only operate via plasticity in mate choice in relation to the relative fitness advantages of each strategy (e.g., Roberts and Gosling, 2003; Oh and Badyaev, 2006; Mays et al., 2008). This antagonistic mechanism of selection could explain the so far equivocal support for the good genes and compatible genes hypotheses (reviewed in Griffith et al., 2002; Mays and Hill, 2004; Akçay and Roughgarden, 2007; Mays et al., 2008; Forstmeier et al., 2014).

Because extra-pair paternity (EPP) is purported to be a mechanism for socially monogamous females to circumvent the negative impacts of social inbreeding (Birkhead and Møller, 1998), compatible genes theory predicts that females should allocate paternity to genetically dissimilar extra-pair males (EPMs; Pusey and Wolf, 1996; Keller and Waller, 2002; Wheelwright et al., 2006; Kempenaers, 2007). Although the compatible genes hypothesis is emerging as an important component of EPP through its influence on offspring viability (Blomqvist et al., 2002; Masters et al., 2003; Richardson et al., 2005; Dowling and Mulder, 2006; Freeman-Gallant et al., 2006; Cohas et al., 2007; Rubenstein, 2007; Brouwer et al., 2011; Arct et al., 2015) the exact selective mechanism through which it acts remains unclear (reviewed in Griffith and Immler, 2009). Potential mechanisms can arise through pre- or post-copulatory (i.e., cryptic female choice; Eberhard, 1991; Colegrave et al., 2002; Birkhead and Pizzari, 2002) mate choice, or through sperm selection in which the sperm of more compatible males are more likely to fertilize an egg (Wilson et al., 1997; Zeh and Zeh, 1997; Birkhead and Møller, 1998; Tregenza and Wedell, 2000). These processes may operate disparately in different animal species. For example, while sophisticated olfactory mechanisms present in mammals, as well as visual and acoustic learning in birds with cooperative breeding systems, commonly facilitate the kin recognition mechanism necessary for female choice based on genetic compatibility (e.g., Pusey and Wolf, 1996; Mateo, 2003; Boulet et al., 2009; Leclaire et al., 2013), the role of pre- or post-copulatory female choice for genetic compatibility in socially monogamous birds remains unconvincing because there is scarce empirical support for the recognition of unfamiliar kin mechanism required for such precise copulatory control (Jamieson et al., 2009; but see Arct et al., 2010).

However, patterns in EPP can also be indirectly affected by demographic parameters such as population social structure; although to date this has been rarely considered in the EPP literature (Reid et al., 2015). Arguably the most common example of population structuring occurs when individuals mate non-randomly with respect to relatedness. In several socially monogamous species, studies have shown that individuals pair with relatives more frequently than expected through random mating (e.g., Foerster et al., 2006; Szulkin et al., 2009; Brekke et al., 2012). If related social-pairs then engage in extra-pair mating, compatible genes benefits could be gained automatically without specific mate choice for more genetically compatible partners, simply because of the high likelihood that the EPM would be less-related to the female than the within-pair male (WPM).

Recently, Reid et al. (2015) demonstrated that female song sparrows, Melospiza melodia, paired to related social males can lower offspring coefficients of inbreeding simply by engaging in random extra-pair mating. This finding suggests that social structuring by itself has the potential to influence the fitness outcomes of EPP without the need for pre- or post-copulatory kin recognition mechanisms. However, Reid et al. (2015) did not include any sexually selected traits in their analyses. This information is important to consider because social structuring could also have potential consequences for the predicted opposing interaction of selection for good genes and compatible genes. For example, in a non-randomly inbreeding population, females engaging in EPP for good genes benefits also stand to gain compatibility benefits as an automatic outcome of extra-pair mating. Under these circumstances, antagonistic selection between good genes and compatible genes effects would be minimized. Therefore, social structuring could provide a pathway through which the benefits of compatible genes and good genes could operate simultaneously. We test this hypothesis by investigating population social structure and its effect on the fitness outcomes of EPP in a naturally occurring population of the New Zealand tui, Prosthemadera novaeseelandiae.

Tui are socially monogamous but have a very high rate of EPP (57% of all offspring in 72% of all broods), and therefore the potential exists for females to gain significant indirect benefits from extra-pair mating (Wells et al., 2015). A previous study on a population of this honeyeater (Meliphagidae), revealed that both within-pair and EPP success are positively related to the size of a male's white throat feather ornament, suggesting mating preference for good genes (Wells et al., 2015). However, mate choice for genetic compatibility could also be an important adaptation to avoid inbreeding in this species because natal philopatry is common in both sexes and is exacerbated by competition for breeding territories near nectar resources (Stewart, 1980; Bergquist, 1985; Stewart and Craig, 1985).

In this study, we use offspring growth rates to quantify potential fitness benefits for genetic compatibility gained from extra-pair mating in this same population of tui, and examine how social structure can influence these potential benefits by increasing the occurrence of non-random social mating. To investigate population social structure, we first determine whether social-pairs are more genetically related than expected through random mating. We then test whether within-pair and extra-pair genetic relatedness influences male paternity success in addition to male ornamentation. If EPP functions as an inbreeding avoidance mechanism, females paired to related social partners are predicted to engage in extra-pair copulations and EPP should be biased toward genetically dissimilar males. Finally we test for increased offspring performance as a predicted fitness outcome of mating with genetically dissimilar EPMs (Zeh and Zeh, 1996; Puurtinen et al., 2009). Because growth rates are an important component of offspring performance (Lindstrom, 1999) that are associated with non-additive genetic quality (Keller and Waller, 2002; Kempenaers, 2007; Pitcher and Neff, 2007; Dziminski et al., 2008; Rosivall et al., 2009), we specifically examine the effect of genetic relatedness of biological parents on offspring growth rates, and compare within-pair and extra-pair maternal half-siblings to determine whether genetically dissimilar males provide compatibility benefits to females.

Theory predicts that compatible genes selection should be greatest in populations where the cost of incompatibility is high (Pusey and Wolf, 1996; Colegrave et al., 2002; Michalczyk et al., 2011). This would occur, for example, in inbred populations where inbreeding depression can have detrimental effects on offspring performance (Keller and Waller, 2002). In our study, this prediction is borne out by several lines of evidence. First, social pairings between kin in tui are relatively common and occur more frequently than expected by chance, increasing the risk of inbreeding depression. Second, the offspring of genetically dissimilar biological parents had faster growth rates than the offspring of highly-related pairs, suggesting that inbreeding can reduce fitness under some circumstances. Third, EPMs were significantly less-related to the female than were the WPMs they cuckolded, EPY generally possessed faster growth rates than WPY, and EPY tended to have faster growth rates than their WPY half-siblings, which suggest that extra-pair mating helps females to circumvent the costs of social inbreeding. Together these findings provide empirical evidence that extra-pair mating in tui can provide compatible genes benefits.

The demonstrated adverse effect of inbreeding on offspring fitness in this study suggests that the high rate of EPP in tui could, at least in part, be an inbreeding avoidance mechanism that has evolved in response to significant natal philopatry (Birkhead and Møller, 1998; Foerster et al., 2006). Yet, we found no evidence that social inbreeding was compensated for by a facultative increase in frequency of EPP. Our findings are thus congruent with studies in other passerines in which females mated with genetically dissimilar EPMs and yet the genetic similarity of the social-pair did not predict the occurrence of EPP (Foerster et al., 2006; Suter et al., 2007; Brekke et al., 2012). Comparably, two of these studies also found some support for non-random social inbreeding (Foerster et al., 2006; Brekke et al., 2012). In our study, this was primarily driven by the social pairings of some highly-related individuals. We therefore argue that although female tui gain compatible genes benefits from EPP, this mating pattern is not a consequence of direct selection on pre- or post-copulatory female extra-pair mate choice (Fossøy et al., 2008) but is rather the automatic consequence of within-pair mating decisions. In support of this hypothesis, if highly-related pairs are removed from our analysis, extra-pair mating appears to be random with respect to genetic relatedness. Nevertheless, this mating pattern does not preclude that EPP in tui evolved as an inbreeding avoidance mechanism, because the social structuring of pairs with high genetic relatedness negates the requirement for female extra-pair mate choice based on genetic compatibility in order for EPP to be adaptive.

Indeed, mate choice based on relatedness would require the existence of sophisticated recognition of unfamiliar kin mechanisms. Although kin recognition by phenotype matching is common among mammals with refined olfactory mechanisms (e.g., Pusey and Wolf, 1996; Mateo, 2003; Boulet et al., 2009; Leclaire et al., 2013) and in cooperative breeders where visual or acoustic cues may be important in kin selection (Komdeur and Hatchwell, 1999; Hatchwell et al., 2001; Cornwallis et al., 2009; McDonald and Wright, 2011), few studies have definitively identified the methods through which socially monogamous birds can assess relatedness (Jamieson et al., 2009). Consequently, significant doubt still surrounds the ability of females to select mates based on genetic compatibility through either pre- or post-copulatory mechanisms (see Griffith, 2007 for a review). Instead, the EPP success of genetically dissimilar tui males likely derives from their superior post-copulatory fertilization success (Tregenza and Wedell, 2000; Griffith and Immler, 2009). Critically, this system does not require an active female preference or facultative manipulation of paternity. For example, under the “genetically loaded raffle” model (Ball and Parker, 2003) females may mate with many different males but only sperm from the most compatible males achieve fertilization (Wilson et al., 1997; Birkhead, 1998; Birkhead and Pizzari, 2002; Bernasconi et al., 2004; Griffith and Immler, 2009; Pryke et al., 2010). Foerster et al. (2006) similarly argued that sperm selection was responsible for the EPP success of genetically dissimilar blue tit, Cyanistes caeruleus, males. The tui system could be considered a specific form of the genetically loaded raffle driven by population social structure; if females are paired to highly-related social males, any chosen EPM should achieve significant paternity success because of his high likelihood of having more compatible sperm.

Previous studies have shown that population social structuring can influence rates of EPP in some species (e.g., Cohas and Allainé, 2009). However, the extent to which social structure influences the fitness outcomes of extra-pair mating has been little researched. To our knowledge, our study is the first to demonstrate that females increase offspring fitness by gaining compatible genes benefits from EPP simply as a result of population social structuring. Our findings are similar to those of a recent study (Reid et al., 2015) which found that female song sparrows, M. melodia, alter offspring inbreeding coefficients through random extra-pair mating due to their propensity to pair with related social partners more frequently than expected by random mating alone. These two studies demonstrate that population social structure can influence the fitness consequences of extra-pair mating without the overt need for an evolutionary mechanism based on kin recognition. However, Reid et al. (2015) did not include any aspect of sexually selected traits in their study, therefore it was unclear how selection for good genes might be affected. Although we do not include a measure of good genes selection here, this was the focus of another study on this tui population during the same time period (Wells et al., 2015). In contrast to the findings of Reid et al. (2015), the study found that extra-pair mating in this population of tui is not random: EPMs are more ornamented than their WPM counterparts, and males with smaller throat ornaments are more likely to be cuckolded (Wells et al., 2015). Theory predicts that females cannot gain good genes and compatible genes benefits from the same male (Mays et al., 2008), and that when these two mating strategies coexist in a population, selection for one benefit will weaken selection for the other (Colegrave et al., 2002). However, this may not be the case when compatibility benefits are the outcome of social structure, as we argue here. Most tui broods containing EPP are sired by only one EPM (Wells et al., 2015). Therefore, it is unlikely that female tui select different EPMs for good genes and compatible genes benefits, as has been demonstrated in blue tits (Foerster et al., 2003). Instead, any females paired to relatives that seek extra-pair copulations with more ornamented EPMs also stand to gain compatible genes benefits because of the higher likelihood that the chosen EPM will be more genetically dissimilar to her than the WPM. However, we stress that our finding that EPY have faster growth rates than WPY does not alone distinguish whether this is due to good genes or compatible genes benefits, or indeed whether it is a result of both processes, and examination of the fitness outcomes of good genes matings is needed to confirm this. Nevertheless, our findings highlight a potential pathway through which EPP in tui can provide the opportunity to simultaneously gain good genes and compatible genes benefits. Thus, population social structuring can facilitate a means by which selection for good genes and compatible genes can operate through EPP with little opposing selection.

The relatively stable pair-bonds and non-random social mating with respect to relatedness in tui then raises the question exactly what benefits, if any, being paired to a relative might provide (reviewed in Kokko and Ots, 2006)? One possibility is that inclusive fitness benefits gained from inbreeding and/or the costs of avoiding inbreeding may outweigh any cost of inbreeding depression (Lehmann and Perrin, 2003; Parker, 2006; Szulkin et al., 2013; Lehtonen and Kokko, 2015). Alternatively, individuals may not actively choose related social partners, yet inbreeding may provide females with some correlated benefit. For example, females may pair with relatives if local males are more likely than immigrants to procure high quality territories. This hypothesis is particularly relevant to tui, in which natal philopatry (Stewart, 1980; Bergquist, 1985; Stewart and Craig, 1985) exacerbated by social aggregations and competition for territories occurring at nectar resources at the beginning of the breeding season (Stewart, 1980), could render such a strategy adaptive.

Like most studies on a large, open population, our study has some limitations. First, our sample sizes, and hence the power of our tests, in this study are small, and should be replicated with a larger dataset. Second, due to the correlative nature of this study, we are unable to differentiate between the effects of relatedness itself on offspring fitness and between potential confounding factors such as age or condition. For example, individuals of low genetic quality or condition may experience limited breeding opportunities and may therefore be more likely to mate with relatives. Third, we cannot rule out the possibility that the greater fitness observed in the young of less related parents is the result of maternal effects. For example, females may preferentially invest more in extra-pair offspring that are deemed to be of greater reproductive value (Magrath et al., 2009; Tschirren et al., 2012). This is a possibility in our study because females choose EPMs based on phenotypic quality in this population. Finally, in including only fledged offspring in our tests we may have inadvertently created a bias, or reduced the power of our tests by reducing our sample size. However, the outcome of this was deemed to be preferable to potentially reporting results based on offspring that had extremely low growth rates that were unrelated to parental relatedness.

In conclusion, our study provides valuable insight into the effect that population social structure can have on the fitness outcomes of extra-pair mating in a large and naturally occurring population. Whether cryptic female choice or sperm selection is responsible for the demonstrated EPP success of genetically dissimilar males is outside the scope of the present study. Irrespective of the exact selective mechanism, our findings demonstrate that females gain compatible genes benefits from extra-pair mating, but that social inbreeding rather than active extra-pair mate choice is responsible. Moreover, in our study, females from highly-related pairs engaging in extra-pair mating for good genes benefits gained genetic compatibility benefits as an additional aside. Thus contrary to theoretical predictions (Colegrave et al., 2002; Mays et al., 2008) we argue that selection for compatible genes need not always weaken selection for good genes when genetic compatibility is an outcome of population social structuring. Such multiple genetic benefits could compensate for any costs of female extra-pair behavior and extra-pair mating could therefore be adaptive for females under such circumstances.

SW and WJ conceived of and designed the study. SW and WJ collected field data. DG provided funding for molecular work. SW analyzed the data with input from DG and BJ. SW, JD, and WJ interpreted the data. SW wrote the manuscript with contributions from JD and WJ.

This study was funded by Landcare Research, New Zealand, and Auckland Council, New Zealand. SW was supported by a Commonwealth Doctoral Scholarship.