Female differential allocation in response to extrapair offspring and social mate attractiveness

Abstract Renewed debate over what benefits females might gain from producing extra‐pair offspring emphasizes the possibility that apparent differences in quality between within‐pair and extra‐pair offspring are confounded by greater maternal investment in extra‐pair offspring. Moreover, the attractiveness of a female's social mate can also influence contributions of both partners to a reproductive attempt. Here, we explore the complexities involved in parental investment decisions in response to extra‐pair offspring and mate attractiveness with a focus on the female point of view. Adult zebra finches paired and reproduced in a colony setting. A male's early‐life diet quality and his extra‐pair reproductive success were used as metrics of his mating attractiveness. Females paired with males that achieved extra‐pair success laid heavier eggs than other females and spent less time attending their nests than their mates or other females. Extra‐pair nestlings were fed more protein‐rich hen's egg than within‐pair nestlings. Females producing extra‐pair offspring had more surviving sons than females producing only within‐pair offspring. Collectively, results show that females differentially allocate resources in response to offspring extra‐pair status and their social mate's attractiveness. Females may also obtain fitness benefits through the production of extra‐pair offspring.


| INTRODUC TI ON
While the benefits of extra-pair paternity to males of pair-bonding species are broadly acknowledged, there is less agreement on possible benefits to females from extra-pair mating (Forstmeier et al., 2014;Griffith et al., 2002;Kempenaers & Schlicht, 2010).
A likely source of confounding maternal effects is the differential allocation of resources to extra-pair offspring (EPOs) or clutches containing EPOs versus within-pair offspring (WPOs) or clutches containing only WPOs (Ferree et al., 2010;Magrath et al., 2009;Schmoll, 2011;Tschirren et al., 2012). Specifically, in species with biparental care, the differential allocation hypothesis predicts that a female will provide higher parental investment to offspring of an attractive or high-quality male-that is, one that she judges to possess "good genes"-than to an unattractive/low-quality male (Burley, 1988(Burley, , 2019Horvathova et al., 2012;Sheldon, 2000). This hypothesis, typically framed in the context of within-pair matings, is applicable to investment in EPOs if females seek extra-pair matings with males that are more attractive than their social mates, as some studies have found (Forstmeier et al., 2002;Kempenaers & Schlicht, 2010;Wilson et al., 2019). The differential allocation hypothesis can be extended to EPO investment because offspring of extra-pair partners are expected to possess superior heritable traits, including sons that display greater secondary sexual trait expression. An alternative possibility to the "good genes" hypothesis is that greater expression of secondary sexual traits of adult extra-pair sons results from differential allocation of resources toward these young by mothers (Tschirren et al., 2012). Of course, these possibilities are not mutually exclusive, particularly since a pattern of greater maternal investment in extra-pair offspring would be paradoxical unless such offspring tend to have higher reproductive value than withinpair offspring.
Several variables likely interact to influence the extent to which females benefit from extra-pair matings. To date, studies of maternal investment in response to extra-pair offspring have most often focused on primary reproductive allocation, notably egg size (Bolund et al., 2009;Krist et al., 2005;Tschirren et al., 2012), which may be positively associated with hatchling survival and development of superior adult trait expression (Krist, 2011;Wagner & Williams, 2007).
However, less is known about parental expenditure in extra-pair offspring during the incubation and provisioning phases. Since social parents often contribute to incubation and offspring provisioningwhich can be costly to caregivers (Alonso-Alvarez et al., 2004;Monaghan & Nager, 1997;Nord & Williams, 2015;Owens & Bennett, 1994;Williams, 2018)-these reproductive phases likely contribute additional sources of variation in allocation of resources to WPOs versus EPOs. A bird's own extra-pair mating tendencies may also affect its parental investment patterns. Males that sire more EPOs may invest less in their social mates' clutches due to the time and resources required to seek extra-pair mates (Ball et al., 2017;Crouch & Mason-Gamer, 2018) while, as noted above, females may increase investment in broods that contain EPOs (Schmoll, 2011).
In addition to the considerations enumerated above, the extent to which preferred male traits are heritable is usually unknown, and environmental conditions typically influence expression of such traits (Cornwallis & Uller, 2010;Griffith et al., 1999).
In this paper, we report results from a breeding experiment in which male breeders had been raised on high-or low-quality diets, with female breeders raised on an intermediate diet. Results from this breeding population focusing on male reproductive performance have been reported elsewhere (Wilson et al., 2019) and show that developmental stress negatively impacts secondary sexual traits and son production. To examine the female perspective here, we include two measures of the quality/attractiveness of females' social mates.
The first is male early diet quality, which positively impacts adult male expression of secondary sexual traits (Naguib & Nemitz, 2007;Wilson et al., 2019); females from this population (Burley et al., 2018) and another (Spencer et al., 2005) prefer males raised under higherdiet conditions. The second metric is male extra-pair mating success, which is presumed to be greater for higher-quality males and has been previously used as an index of attractiveness in studies on this species (Bolund et al., 2009;Houtman, 1992). Male extra-pair success is thought to reflect genetic as well as environmental determinants of variation (Jennions & Petrie, 2000;Neff & Pitcher, 2005).
Since a female's mate choice and reproductive allocation decisions are likely to depend on her mate's developmental history as well as his genetic quality-and she may well lack information to assess the relative contribution of environmental versus heritable effects to male phenotype-use of these two measures may provide insight into a broad range of male phenotypic attributes that influence female extra-pair mating decisions.
By manipulating early-life diet quality of male zebra finches and measuring subsequent reproductive investment as well as withinpair and extra-pair reproductive success, we address the following questions: Do females invest more in EPOs or EPO-containing clutches? Do they invest more in WPOs sired by attractive males? Do males invest differentially in offspring care based on their own attractiveness and/or the presence of EPOs in the clutches they raise? In line with the differential allocation hypothesis, we predicted that females with attractive social mates would invest more in offspring at all phases of reproduction and that attractive males would show lower parental care. In light of previous findings for this species (Tschirren et al., 2012), we also expected that females would invest more in EPOs regardless of the quality of their social mates. Lacking evidence that male zebra finches contribute to the EPOs they sire or that they can detect their mates' EPOs in the clutches they raise, we did not make specific predictions about male caregiving behavior in relation to EPOs. Finally, we ask whether females gained direct fitness benefits through production of EPOs, a possibility that has seldom been addressed empirically, by assessing female overall reproductive success in relation to her extra-pair success.

| Founder rearing conditions and flight initiation
Zebra finches (Figure 1) selected as founders of this breeding experiment were raised in one of four large outdoor aviary flights (3 × 12 × 2.3 m), each of which contained 64 breeders (32 of each sex). To reduce relatedness between male and female founders and to control founder rearing diet, two of the rearing flights were used to generate female founders, while the remaining flights were used to generate male founders. Access to boiled hen's egg differed among rearing flights, but in other respects (size, resources, microclimate), all rearing flights were virtually identical. Female founders were reared in flights in which hen's egg was made available three times a week (LAB diet). In order to manipulate male early diet quality, male founders were reared in either a flight in which egg was provided daily (HI diet) or one in which egg was never provided (LO diet). All diet regimes were maintained until birds were selected for this study. Hen's egg has an amino acid profile similar to that of halfripe grass seed (Allen & Hume, 1997), which is seasonally available to zebra finches in the wild and has a higher protein content than ripe grass seed. All rearing flights provided breeders and fledged offspring with ad libitum access to a commercial mix of ripe grass seed for estrildines, cuttlefish bone, ground oyster shell and water, and green vegetables three times a week. All resources were provided to this ground-feeding species on the aviary floor.
Offspring produced in rearing flights were banded with numbered, closed metal leg bands when they were 7-14 days old. Once offspring reached 45 +/-3 days of age, they were housed in single-sex cages (68 × 50 × 54 cm) at standard densities (10 birds) inside their rearing flight until they reached adulthood (100 days of age) in order to provide developing birds visual and acoustic contact with adults; such exposure is important for imprinting on visual and acoustic traits (Bischof et al., 2002;Bolhuis, 1991;Immelmann, 1975). Birds remained in single-sex cages until they were selected for use in the present study. Additional details on founder rearing conditions and the history of this experimental colony can be found elsewhere (Wilson et al., 2019).
In total, 32 females and 32 (16 HI and 16 LO) males were selected to found the breeding colony for the current experiment. Once selected, birds were held in single-sex flights and maintained on the LAB diet for 6 weeks before they were released into the breeding flight, a protocol that allowed LO-diet males the opportunity to become familiar with the availability of hen's egg as a food resource.
LO-and HI-diet males did not differ in their consumption of egg during this period (two-sample t test: t = 1.24, df = 30 p = .22).
During this time, male phenotype data were collected and a 25μl blood sample was collected from the brachial vein of each bird for genetic parentage analyses. HI-diet males had a greater expression of two secondary sexual traits compared to those raised on the LO-diet (redder beaks, larger cheek patches), supporting the use of natal diet as a metric of attractiveness (see Wilson et al., 2019 for details). Founders varied from 6 to 13 months of age (X̄ +/-SD: females-279 +/-53 days old; males-354 +/-50 days old) at the start of breeding and had no prior breeding experience. All birds were judged to be in excellent overall condition and no more than 2 siblings from each family were employed (males were derived from 20 families and females from 24 families). Lastly, all birds were banded for identification during observations using colors for which zebra finches show no band preference (Burley, 1985).

| Breeding conditions and reproductive measures
Founders of both sexes were released simultaneously into a single (3 × 12 × 2.3 m) outdoor aviary flight and allowed to pair and breed for 5 months. The flight was maintained on the LAB diet. Ample nest sites (~2.5/ breeding pair) and nest material (dried Bermuda grass and feathers) were provided. The time it took each pair to F I G U R E 1 Female and male zebra finches were not found on the day they were laid, as evidenced by the discovery of two or more unmarked eggs in a nest on the same day; such eggs were not included in analyses of lay order due to ambiguity. Since relatively few eggs in a single clutch hatched on the same day (12.2%), we were able to reliably track the egg from which each offspring hatched by noting which egg was "missing" when a new, unmarked nestling was found. The down feathers of new hatchlings were colored with nontoxic markers to track hatch order. When two nestlings appeared on the same day, the older nestling (assessed by having drier down feathers or, secondarily, weighing more) was assigned to the egg of earlier lay order.
Nestlings were banded with seamless numbered bands between 7 and 14 days of age. When independent offspring reached 45 +/-3 days of age, they were caught and housed within the natal flight in cages containing other same-sex offspring of similar age.
At this time, a 25μl blood sample was collected from a brachial vein for genetic parentage assignment.

| Social and genetic parentage assignment
Both genetic parentage and social parentage were tracked in order to identify which birds were successful in producing surviving EPOs. Social parentage was assigned to each clutch through regular observations of active nests beginning when eggs first appeared in a nest and ending once the last nestling fledged. Genetic analyses involved isolation and amplification of DNA through PCR with fluorescently labeled primers corresponding to 8 highly polymorphic microsatellite loci previously identified for zebra finches (Forstmeier et al., 2007). CERVUS 3.0 (Kalinowski et al., 2007) was used to assign parentage based on correspondence at these 8 loci.
Genetic analyses of blood samples collected from all founders and 186 offspring were used to assign genetic parentage based on 6 or more unambiguous loci.
Because only offspring that survived to independence were gen- pairs produced at least one surviving offspring). Of the male breeders included, 12 were raised on the HI diet and 13 on the LO diet. Nine females and 9 males (3 LO and 6 HI) raised one or more EPOs to independence during the course of the experiment (range: 1-6 EPOs).

| Parental nest attendance
The amount of time parents spent attending their nests was re-

| Nestling provisioning
Nestling provisioning was recorded in order to examine whether it differed based on nestling status (EP or WP) and/or male attractiveness (male natal diet and extra-pair success). Since seed and egg differ in nutrient quality, we measured provisioning of these food types separately on days when egg was provided, so that relative amounts of egg versus seed feeding could be assessed.
Immediately prior to provisioning sampling sessions, founders were provided with a bowl of egg placed on the aviary floor and allowed to feed for 3 min. If birds flushed in response to a disturbance, timing was suspended until bird(s) returned to the egg bowl. The bowl was then removed and, after an additional 5 min, experimenters entered the aviary and scored crop contents of all nestlings. This protocol was repeated to produce 2 samples of nestling provisioning on each sampling day. At the end of the testing period, the remaining egg was left in the flight for the rest of the day. This design was selected after preliminary trials confirmed that (1) increasing the foraging interval from 3 to 5 min did not significantly increase the number of adults that consumed egg (although the same individuals often revisited food bowls more times) and (2) interference competition at the food bowl was uncommon (see Discussion).
Nestlings ranged in age from 0 (hatching) to 15 days (X̄ +/-SD:  The RM LMM predicting egg mass included the three main effects, with female extra-pair success included as the fixed effect "Egg Status (EP or WP)"; egg status was assigned to each individual egg based on offspring genetic analysis. Lay order was included as a covariate, and mother's identity was included as a random variable. The RM LMM predicting clutch size included the three main effects, with female extra-pair success included as the fixed effect "Clutch Status (EP or WP only)"; here "EP" denotes that one or more EPOs were present in the nest, and "WP only" denotes that all surviving nestlings were WPOs. Mother's identity was included as a random variable.

| Analyses
The RM LMM predicting nest attendance time included founder sex as a fixed effect in addition to the three main effects, where the fixed effect for female extra-pair success was "Clutch Status (EP or WP only)"; as for the model predicting clutch size, "EP" denotes that one or more EPOs were present in the nest, and "WP only" denotes that all surviving nestlings were WPOs. Interactions between founder sex and other main effects were considered based on a priori expectations of sex differences in nest attendance time (Wilson et al., 2017). Observation day (defined above) was included as a covariate and, since first clutches can be less successful, clutch number was also included as a covariate. Mother's identity was included as a random variable.
The RM LMMs predicting nestling provisioning included nestling sex as a fixed effect in addition to the three main effects, with female extra-pair success scored as the fixed effect "Nestling Status (EP or WP)," which was assigned to each individual nestling. Nestling age on the day of sampling and clutch number were included as covariates. Nestling identity was included as a random variable.
Three-way ANOVAs were performed to assess how female extra-pair success and the two measures of mate attractiveness influenced the total number of offspring that females produced.
Numbers of sons and daughters were also analyzed separately since several studies have shown for this species that several environmental and parental conditions influence the relative production of the two sexes (Bradbury & Blakey, 1998;DeKogel, 1997;Foster & Burley, 2007;Kilner, 1998;Martins, 2004). Since the sample size was relatively small (N = 25 pairs), interactions were not included in these models. In these analyses, females were dichotomously categorized as having produced 1 or more surviving EPOs (having obtained "extra-pair success") or having produced no surviving EPO (not having obtained "extra-pair success"). Similar two-way ANOVAs were run to assess variation in production of EPOs by females as a function of social mate attractiveness.
Finally, Pearson's tests were used to ask whether the total num-

| RE SULTS
Whether or not a female produced one or more extra-pair offspring was independent of her social mate's attractiveness (Fisher's exact: male natal diet-p = .12; male extra-pair success-p = .39).

| Incubation-phase parental effort
Nest attendance time was affected by male extra-pair success and the interaction between male extra-pair success and breeder sex.
Males that achieved extra-pair success spent the most time attending their nests (p < .002), while their mates spent the least (p < .001).

Males without extra-pair success spent less time attending their
nests than males that achieved extra-pair success (z = 3.11; p = .002) and a similar amount of time as their mates (z = 1.30; p = .19) (Table 3; Figure 3).

| Nestling-phase parental effort
1.00 +/-0.12; WP: 1.24 +/-0.06) than WP nestlings. This pattern was driven by the interaction between nestling extra-pair status and male natal diet (Table 4): EP nestlings with HI-diet social fathers were provisioned with the highest amounts of egg and the lowest amounts of seed ( Figure 4).

| Reproductive success
Total reproductive success between social mates was highly correlated (Pearson's correlation: R 2 = 0.811; p < .0001) but EP reproductive success was not (Pearson's correlation: R 2 = 0.207; p = .32) ( Figure 5). Analyses revealed predictors of son production, but not daughter production or total offspring production. Females that produced one or more EPOs during the study produced more surviving sons than females that produced only WPOs (Table 5; Figure 6).

Males' natal diet and extra-pair success did not impact their social
mates' extra-pair offspring production (p > .18; Table 6). Lastly, females did not produce more EPOs of one sex (Fisher's exact p = .39).

| D ISCUSS I ON
Differential allocation has been demonstrated in both free-living and captive populations (Burley, 1986;Horvathova et al., 2012). As shown here and elsewhere (e.g., Rutstein et al., 2004;Gilbert et al., 2006;Johnsen et al., 2005), food availability need not be limited to observe patterns of differential allocation, because parental expenditure can be limited by numerous environmental and physiological trade-offs (e.g., Svensson & Nilsen, 1997;Williams, 2012). Breeders may facultatively increase the number of eggs laid and/or offspring reared per brood when food availability increases, for example, and then become time/energy limited as a result. In short, the availability of ample food in a laboratory environment does not imply that breeders fail to experience costs of parental care (Burley, 1988). In this light, and as discussed in detail below, some patterns described here number of surviving sons, which leaves open the possibility that females benefit by producing offspring who will themselves experience greater reproductive success.
It is important for researchers to acknowledge the range of plausible interpretations of their results. Here, we need to keep in mind that variation in mate quality may generate positive assortative mating patterns (Burley, 1986;Holveck & Riebel, 2010) even though various constraints-ecological and informational-on such assortment may lead to differential allocation (Burley, 1986(Burley, , 2019; given this, it is possible that patterns described here may also relate to assortative mating patterns. Since female mating quality can be difficult to assess, assortative pairing and differential allocation can often not be parsimoniously resolved.

| Egg phase
Egg mass increased with laying order but EP eggs were not heavier than WP eggs (Table 2) and, unlike reports on other species (Cordero et al., 1999;Krist et al., 2005;Magrath et al., 2009), in the present study EP eggs were not more likely to be laid near the beginning of a clutch, where mortality is typically lower (Ferree et al., 2010;Magrath et al., 2009). Nonetheless, there was evidence of female differential allocation based on social mate quality during the egg phase: Females produced heavier eggs when mated to males with extra-pair success (Figure 2a), and females mated to HI-diet males tended to produce larger clutches (Table 2).
Differential allocation toward eggs in response to male quality/ attractiveness has been demonstrated previously, both in this species (Gilbert et al., 2006;Arnold et al., 2016;but see Bolund et al., 2009) and other avian species (Cunningham & Russell, 2000;Horvathova et al., 2012). The effects of greater allocation toward offspring sired by attractive males often last into adulthood (Arnold et al., 2016;Cunningham & Russell, 2000;Gilbert et al., 2006), confounding demonstration of possible genetic benefits from sires to offspring.
An alternative, and nonmutually exclusive, interpretation is that pairs formed assortatively based on male attractiveness and female fecundity. Since egg mass is heritable (Christians, 2002;Potti, 1999) and can be influenced by female early-life environment-with better environments associated with heavier eggs and larger clutches (Griffith & Buchanan, 2010;Monaghan et al., 1996;Potti, 1999)-high-quality males may have secured mates that were more fecund. Assortative pairing in zebra finches has been reported in both free-pairing colony experiments (Burley, 1986;Burley & Foster, 2006) and mate choice experiments (Holveck & Riebel, 2010), F I G U R E 6 Number of offspring produced by females based on EP success (X̄ +/-SE). Circles indicate females that produced one or more surviving EPOs during the study (N = 9). Squares indicate females that produced only WPOs (N = 16). *p < .05. Data correspond to Table 5 Variable

TA B L E 6
Number of extra-pair offspring produced by each female that survived to independence but it is unclear whether such assortment extends to fecundity, since mixed results are reported for males' tendency to choose based on female fecundity (Martin & Burley, in press;Wang et al., 2017). Thus, the relative contributions of differential allocation and assortative mating to the egg mass patterns found here remain unclear.
Females paired to LO-diet males invested less in EP eggs relative to WP eggs (Figure 2b), which is contrary to expectations, since males raised on the LO diet have lower expression of secondary sexual traits (Burley et al., 2018;Wilson et al., 2019), and males successful in obtaining extra-pair copulations are expected to be of above-average quality. This finding is, however, potentially consistent with a previous study on this species that reported compensatory investment in eggs in response to social mate quality (Bolund et al., 2009). Specifically, since egg production is costly (Stearns, 1992;Visser & Lessells, 2001;Williams & Miller, 2003), we envision that females with low-quality social mates may benefit by reducing investment toward individual EP eggs when brood-mates compete to be fed. Since maternal fitness would likely be reduced if EP offspring-presumed to be of higher genetic quality-were to out-compete WP offspring for food, a tactic of reduced investment in EP eggs may result in equilibration of broodmate competitive ability, while permitting females to produce EP offspring of high genetic quality. The variation in egg mass investment found in this study highlights the potential for genetic and maternal effects to interact in how they drive observed allocation toward EPOs and WPOs. Greater theoretical attention to which aspects of environmental variation are likely to favor production of EPOs (e.g., Eliassen & Kokko, 2008), as well as to the context-dependent roles of differential allocation and reproductive compensation (Burley, 2019) in production of EPOs and WPOs, would likely advance our understanding of these complex but ecologically relevant patterns.

| Incubation phase
Patterns of nest attendance showed differences in how parents divide this shared duty based on male attractiveness but did not suggest occurrence of female differential allocation. Zebra finch pairs exhibit plasticity in incubation behavior (Gilby et al., 2013;Wilson et al., 2017;Zann & Rossetto, 1991), and this plasticity likely reflects differences in individual condition or quality as well as differences in sexual conflict/cooperation (e.g., Gorman & Nager, 2003;Wilson et al., 2017). While avian incubation is typically considered to be female-led (Burley & Johnson, 2002;Moore & Varricchio, 2016;Tullberg et al., 2002), if a male increases his nest attendance time, a comparable decrease in his social mate's attendance time is expected, since pairs are likely aiming for some optimal total amount of incubation (Jones, 1989;Wilson et al., 2017). Our finding that total nest attendance time did not vary between diet treatments (Table 3) is consistent with this expectation. Although males that sire EPOs are typically predicted to invest less in their social mates' clutches (due in part to the time and resources required to seek extra-pair mates-Crouch & Mason-Gamer, 2018), we found the opposite pattern here (Figure 3). This result may reflect that zebra finches are a gregarious species: They nest in colonies and typically feed in aggregations (McCowan et al., 2015;Zann, 1996). Under these circumstances, male search costs for extra-pair partners may be low.
An important additional consideration is that male quality likely influences the cost of seeking extra-pair partners: Attractive males should obtain extra-pair mates with lower effort and may therefore experience little trade-off between nest attendance and extra-pair activities (Burley et al., 1994). If so, females mated to such males may then experience greater time to feed early in the clutch cycle, likely contributing to the egg mass pattern observed here (Figure 2a).

| Nestling phase
EP nestlings raised by pairs containing attractive males-as measured by early diet quality-received greater provisioning of nutrient-rich egg. We interpret this pattern as consistent with our prediction of differential allocation of egg resources to offspring of attractive males. To explain this conclusion, we develop several salient points and consider alternative hypotheses. First, we note that the EP nestlings that were fed a greater amount of egg experienced a corresponding decrease in the amount of seed they were fed ( Figure 4). This inverse relationship is to be expected given  (Criscuolo et al., 2011;Labocha et al., 2015).
However, it is also likely that parents with offspring judged to have higher reproductive value were more willing to take such risks.
Another possible explanation for egg-provisioning patterns that is important to consider is that LO-diet males may have had lower tendency to feed egg to offspring as a result of their prior unfamiliarity with egg as a food item. This concern, however, is offset by the finding that males of the two rearing treatments did not differ in their tendency to consume egg in samples collected during the 6-week interval they spent on the LAB diet prior to the start of the experiment. The palatability of hen's egg to naïve birds has also been demonstrated for wild zebra finches (Burley et al., 1992).
A final possibility important to consider is that egg-provisioning patterns may have been caused by interference and/or scramble competition, which could influence results, especially if HI-diet males and their mates were superior competitors. The potential significance of interference competition is offset, however, by the finding of no differences in egg-eating patterns between HI-and LO-diet males during preliminary trials (see above). More broadly, interference competition at food resources is minimal in our colonies; this pattern likely reflects the natural history of the species, which feeds in flocks on the ground, where birds are vulnerable to predation and where any disruption leads the flock to scatter. In the feeding trials performed here, up to 8 adults at a time crowded onto the rim of the bowl of egg, collected egg in their beaks and quickly flew away; other birds, waiting nearby, then took their turn. The amount of egg was sufficient to ensure that some usually remained after the end of sampling, and trials conducted during protocol development indicated that birds were not excluded from feeding during trials (see Methods); these methods likely minimized effects of scramble competition on results. While we do not contend that competition was entirely absent during feeding trials, the considerations addressed here add weight to our interpretation that the greater relative provisioning to EPOs of HI-diet males reflects a pattern of differential allocation of high-quality nutrients to these offspring. While further research to explore questions about foraging risk assessment and the relative roles of the two parents in egg feeding would facilitate interpretation of these results, they are generally consistent with the idea that EPOs have higher reproductive value.

| Reproductive success
Contemporary frameworks for the evolution of extra-pair behavior have highlighted the potential for extra-pair behavior to be maladaptive for females of pair-bonding species (Arnqvist & Kirkpatrick, 2005;Forstmeier et al., 2011Forstmeier et al., , 2014Kempenaers & Schlicht, 2010) but persist in populations due to the benefits of extra-pair tendencies to males combined with weak or absent sexlimited expression of such tendencies (Forstmeier et al., 2011).
By contrast, our results suggest that females may obtain a fitness benefit by producing more sons, a result that complements a previous report of female zebra finches investing more reproductive resources in sons sired by EP males (Tschirren et al., 2012).
In theory, such benefits could arise through greater caregiving toward EPOs and their clutch mates (as seen here in selective egg feeding patterns) which may enhance survival as well as mating quality of offspring in EPO clutches. Collectively, results point to the importance of remaining open-minded about the possibility that females may obtain direct fitness benefits from extra-pair activities through changes in parental investment as well as the more frequently invoked possibility of indirect fitness benefits (Akçay & Roughgarden, 2007;Birkhead & Pizarri, 2002;Iwasa & Pomiankowski, 1991;Jennions & Petrie, 2000;Neff & Pitcher, 2005;Tschirren et al., 2012). Another possibility that merits investigation is that those females predisposed to producing more sons on the basis of genetics and/or condition are more likely to accept extra-pair copulations. Such females might stand to gain more from EPOs, since sons produced with an EP male are expected to have high reproductive value.

| CON CLUS ION
Results of this paper indicate that greater attention to the effects of extra-pair mating on differential allocation of resources to offspring

ACK N OWLED G M ENTS
We thank Michelle Hernandez for collecting blood samples; Michelle Manalo and Cole Symanski for assistance with collecting reproductive investment data and phenotype information, respectively; John Avise for access to supplies and laboratory space for genotyping; Andrey Tatarenkov for advice on genotyping; and students in Bio 199 for help with colony maintenance. This research was supported by the University of California, Irvine.

CO N FLI C T O F I NTE R E S T
KMW and NTB declare no conflicts of interests.