Partner retention as a mechanism to reduce sexual conflict over care in a seabird

The costs of caring for offspring are predicted to lead to an evolutionary con ﬂ ict between parents, where each parent bene ﬁ ts if the other provides most of the care. However, in many biparental species, breeding partners remain together for multiple breeding attempts and their respective future reproductive ﬁ tness prospects are therefore intertwined. Since an increase in current care by a long-term partner reduces that partner's future investment and longevity, individuals do not automatically only bene ﬁ t when their partner provides care. We tested whether selection will favour individuals that reduce the burden of care falling on long-term partners, thus decreasing evolutionary con ﬂ ict over parental care. Using a seminatural, captive colony of black-headed gulls, Chroicocephalus ridibundus , we show that bene ﬁ ts of long-term partner retention can indeed reduce sexual con ﬂ ict. Long-term partners had less intense courtships and were more behaviourally compatible, and individuals in long-term pairs spared each other's resources by increasing their own parental investment compared to those with new partners. Lastly, we demonstrate that high partner compatibility in newly formed pairs can select for commitment to the pair bond. Our results highlight that compatibility bene ﬁ ts of long-term partnerships can increase selection for mate retention and increase parental investment. © 2023 The Authors. Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour. This is an open access article under the CC BY license (http://creativecommons.org/licenses/ by/4.0/).

Across the animal kingdom, there are many species in which breeding partners work together to raise their joint offspring (Kleiman & Malcolm, 1981;Gross & Sargent, 1985;Trumbo, 1996;Cockburn, 2006).While such biparental care is relatively rare compared to other reproductive systems (Reynolds et al., 2002), its repeated evolution and maintenance in numerous different taxa has attracted decades of research attention because it is theoretically unstable.Providing parental care is costly (Smith, 1995;Santos & Nakagawa, 2012): investing resources in offspring increases the likelihood that a parent succumbs to disease (Nordling et al., 1998;Knowles et al., 2009) or predation (Veasey et al., 2001;Li & Jackson, 2003) or depletes its reserves so much that it cannot afford to fully invest in subsequent reproduction (Landwer, 1994;Golet et al., 2004).Selection is therefore expected to act on both male and female parents to limit their own investment in offspring and force their breeding partner to provide the care necessary to rear offspring to independence (Trivers, 1972;Houston et al., 2005;Lessells, 1999).Our current views of reproductive systems (Thomas & Sz ekely, 2005), parental behaviour (Royle et al., 2002) and even population level dispersal patterns (Kuijper & Johnstone, 2018) are all rooted in the idea that parents are in an evolutionary conflict.In the quest to reveal how biparental care can none the less evolve, evolutionary biologists have identified several ecological mechanisms and behavioural adaptations that can help to overcome parents' differing fitness interests (e.g.Bebbington & Hatchwell, 2016;Johnstone et al., 2014;Johnstone & Hinde, 2006;Harrison et al., 2009;Lessells & McNamara, 2012).
While this body of work has contributed substantially to our understanding of biparental care, it largely relies on the assumption that parents' fitness interests will always be maximized if the partner takes on the costly investment in parental care (Griffith, 2019).Apparent mitigation of this evolutionary conflict, or evidence for cooperation between caring parents, has been shown to occur when the benefits of providing care are exceptionally high (Trillmich, 2010;Reme s et al., 2015;Vincze et al., 2017).However, the potential costs of withholding care and forcing the partner to increase its investment are less well understood (Griffith, 2019).One important factor that determines whether parents might pay a cost to forcing a current breeding partner to provide care is the future value of that partner.If the current breeding partner is sufficiently important for a focal parent's future survival and/or reproductive success, the costs of that partner providing care are also paid by the focal parent because losing the partner reduces its future survival and reproductive prospects.This argument is already well known in its most extreme conceptualization: under 'true monogamy' where individuals mate for life and do not remate if the current partner dies, parental fitness interests are completely aligned (Lessells, 2006;Parker, 1985).However, the same concept applies to a lesser extent even when partners only expect to remain together for a short time after the current breeding attempt, whether due to reproductive benefits, social support or resource defence (see Griffith, 2019).Although the idea that parents might be under selection to cooperate if their future reproductive fitness relies on the future investment of their partner has been articulated elsewhere (McNamara et al., 1999;Mock et al., 1996;Patrick et al., 2020), empirical tests of its application in real-world systems are rare.
We argue that the existence of costs of care provided by a breeding partner can further reduce parental conflict via selection on traits that maximize the chance of partner retention, which in turn indirectly alter parents' optimal level of investment in current reproduction in two ways.First, individuals that benefit from breeding again with a current partner (for example because doing so increases reproductive success or because alternative partners are hard to find) should be under selection to reduce the costs of care for that partner to maximize the chance that the partner survives, stays in the partnership and has the necessary resources to breed with them again (Coulson, 1966;Mills, 1973).Once the costs of care for the partner are reduced, the partner's own optimal level of investment increases and becomes more aligned with that of the focal parent, reducing the extent of conflict between them.Second, individuals that benefit from retaining their current partner may also be selected to signal their commitment to the breeding partnership to reduce the risk of being divorced (Mills, 1994).Assuming such signals are honest (i.e. the costs of signalling increased commitment are only worthwhile if the focal individual truly intends to commit), the breeding partner gains assurance that the partnership will continue.As a result, the net fitness benefit per unit of their care increases as it includes the additive benefit of sparing the resources of the focal individual, which has signalled its intention to breed with that partner again.Once the future fitness benefits of partner care increase, the optimal level of investment becomes closer to that of the focal individual, again reducing conflict.
In this study, we make use of a well-documented example of how retained partners can contribute towards future fitness, known as the 'mate familiarity' effect (Black, 1996), to test whether breeding partners that are important for each other's future fitness are more likely to invest in each other and have lower levels of conflict over care.Across a variety of taxa, individuals repeatedly breed with the same partner over successive seasons (Kleiman, 1977;Black, 1996;Whiteman & Côt e, 2004), which suggests that for many species, such a strategy can be beneficial over serially divorcing and remating for each new breeding attempt.Indeed, there is abundant evidence for benefits of long-term pair bonds that cannot be explained through breeding experience or breeder age, including improved reproductive success (e.g.Mills, 1973;S anchez-Macouzet et al., 2014;van de Pol et al., 2006) and behavioural coordination (Prior et al. 2020), and increased overwinter survival (Culina et al., 2015).Although empirical tests of this are still largely missing, familiar partners may also benefit from reduced costs of negotiation over parental care (Lessells & McNamara, 2012) if they are already familiar with each other's quality and investment capacity.In all these cases, a more established breeding partner can be inherently more valuable than a newer one; individuals that force such established partners to provide super-optimal levels of parental care may pay the cost of losing that partner's contribution to future reproduction.However, two things remain unclear.First, we must test whether individuals that benefit from retaining a current partner actually invest more in that partner.Second, we must determine whether individuals invest in highly compatible new partners (leading to the establishment of valuable long-term pair bonds), or instead only start investing in partners once a long-term bond has been established.
We aimed to determine whether the fitness cost of care provided by valuable future breeding partners can indeed reduce parental conflict over care through its influence on behaviours affecting partner investment or social commitment.To do so, we studied a captive breeding colony of black-headed gulls, Chroicocephalus ridibundus.Black-headed gulls are long-lived, socially monogamous seabirds found throughout Eurasia and northern America (Burger et al., 2020).Both males and females are involved in all aspects of parental care including building and defending the nest, incubating eggs and provisioning nestlings (Van Rhijn & Groothuis, 1985).Pair bonds can last many years, although divorce and remating after partner death both occur relatively frequently (ca.20% of captive pairs divorce or remate in a given year (Bebbington and Groothuis, n.d.) and our captive colony always contains a number of unpaired individuals that are available to enter the breeding pool, should they find a partner).Thanks to a relatively high divorce rate and annual release of some birds from the colony, older individuals are regularly paired with younger ones in our populations, which allows us to distinguish effects of pair bond duration from those of age.While a mate familiarity effect on parental care and fitness has yet to be demonstrated in blackheaded gulls, two aspects of the species' ecology would suggest that longer partnerships may provide fitness benefits.First, pair bonds are entirely severed during the nonbreeding season (with no obvious spatial or behavioural link between known partners) and must be re-established each spring during a period of intense and elaborate courtship rituals, of which performance synchrony is a prominent component (Van Rhijn & Groothuis, 1987).This complex and time-consuming re-establishment of the pair bond might be shorter and more efficient when performed with a long-term mate than with a new one.Second, offspring require a prolonged and intense period of parental care (Sk orka et al., 2012), which we predict may also be easier to coordinate when breeding partners are familiar with each other (Prior et al., 2020).The current study is based on data collected from a large captive colony kept in seminatural conditions, where the reproductive behaviours of individuals have been monitored for over 10 years.Together, this suggests that the black-headed gull is a highly suitable system for testing the hypothesis that the value of a long-term partner can reduce the extent of parental conflict over care.
We first tested whether the mate familiarity hypothesis holds in this species, that is, there is variation in the value of a current partner.(H1) Courtship costs hypothesis: long-term pair bonds are beneficial because they reduce the time and energy cost of mate searching and courtship.We predicted that long-term pairs display less often and that their courtship period is shorter than that of more recently formed pairs, leading to earlier onset of breeding.(H2) Behavioural coordination hypothesis: long-term pair bonds are beneficial because they allow pairs to develop a high degree of behavioural coordination, thereby increasing the efficiency of joint behaviours like courtship and parental care.We predicted that long-term pair displays are more synchronous, and that the division of incubation duties is more coordinated (but see Patrick et al., 2020 for an alternative prediction).
Having established that long-term partners are more valuable than new partners, we then tested whether birds in long-term pair bonds invest more in maintaining their current partner.(H3) Partner investment hypothesis: individuals in long-term pair bonds aim to reduce the costs of parental care for their partner.We predicted that long-term pairs have a more equal incubation share, as both members of the breeding pair are selected to maximize their own investment and reduce that of their partner; this results in an overall more equal division of labour.We also predicted that males provide more direct resources to females through increased frequency of courtship regurgitation (Mills, 1994), and that females reduce male costs of care by increasing prenatal resource allocation (larger egg volume relative to clutch size; Salas et al., 2020).(H4) Partner commitment hypothesis: individuals in long-term pair bonds provide behavioural cues of their commitment to the bond.We predicted that birds in long-term pairs spend more time on two visually conspicuous nesting behaviours: nest building and territory defence displays.Nest building is usually performed when both partners are at the nest (Beer, 1963) and territorial displays are visual indicators of individual motivations (e.g.Tinbergen, 1956;Bukaci nska and Bukaci nski, 1994), suggesting that both can act as valuable signals to the breeding partner.We also predicted that individuals in long-term pairs signal their commitment to the partnership by actively joining in a greater proportion of courtship displays initiated by their partner, and by participating less in extrapair courtship displays (Mills, 1994).
Lastly, we tested whether investment in valuable partners already occurs in highly compatible new pairs, thus driving the mate familiarity effect and subsequent selection on mate retention in this species.(H5) Partner compatibility hypothesis: the benefits of remaining with a highly compatible partner select for mate retention and result in high compatibility and low conflict in longterm pair bonds.If selection on retention of valuable partners already begins early in the partnership, we expected that newly formed pairs with a high level of compatibility (as demonstrated by a high degree of behavioural synchrony during initial courtship) would invest more in their partner and in their commitment to the pair bond than pairs with low compatibility.If partner compatibility only arises after multiple joint breeding attempts, we expected that all individuals in new partnerships would show low levels of investment in their partner and the pair bond.

Study Species and Housing
We studied a captive population of black-headed gulls that has been continuously housed and monitored at the Groningen Institute for Evolutionary Life Sciences since 2010.The colony of ca.100e120 birds is kept in a large (350 m 2 ) aviary with sufficient space for flying, ample potential nesting sites and two pools for swimming and drinking.The roof is covered with transparent material and the floor with sand.Birds are subject to natural light and temperature regimes and are fed ad libitum with pelleted trout food (E-3P Stella, Trouw Nutrition, Putten, Netherlands) and supplemented three times a week with cat food (Huismerk 3-mix kattenbrok, Arie Blok, Woerden, Netherlands) soaked in water.In the nest-building period, birds are provided with straw to use as nest material, along with numerous natural and seminatural shelters to support nest building and chick rearing.
The ancestors of the current population were either reared from eggs or nestlings or captured as adults from wild black-headed gull colonies in the north of the Netherlands.The captive population was first established in the 1980s; since that time, birds have been periodically released and the colony supplemented with new wild eggs and birds (most recently in 2010) to maintain a healthy, outbred and, importantly, wild-type population exhibiting as close to natural behaviours as is possible in captivity.As in the wild, pairs in the aviary construct and defend nests on the ground from mid-April onwards, with the first eggs being produced at the beginning of May (median 4 May).Modal clutch size in both wild populations and the captive population is three (range 1e4; Lundberg & V€ ais€ anen, 1979; K. Bebbington & T.G.G.Groothuis, personal observation).All birds in the colony have been molecularly sexed (following Griffiths et al., 1998) and are fitted with a unique combination of four or five coloured leg rings for individual identification.Of the 115 individuals currently in the colony, 91were hatched in captivity after 2009 and are hence of known age; the remaining 24 were at least 1 year old in 2010.Each breeding season since 2012, the members of each breeding pair have been identified and all nests were monitored.
This study was based on 42 breeding pairs that nested in the breeding season of 2020.We excluded from the analysis two homosexual pairs and one heterosexual pair that formed very late in the season after the majority of the observations were performed.We focused on the reproductive behaviours of pairs from the onset of courtship up until the end of the incubation period, when all eggs were removed prior to hatching for use in another experiment.For each pair, we used the historical data to determine pair bond duration and the age of both birds.Effects of pair bond duration are notoriously difficult to separate from those of age (Fowler, 1995;van der Pol et al., 2006;S anchez-Macouzet et al., 2014); in our population, remating after divorce or partner death is not uncommon (on average, there are nine new pairs each year in our population and only nine of the pairs used in this study had been together since 2010), meaning that these two variables are at least partially decoupled.To maximize our ability to partition variation between age and pair bond duration, we classified birds into either 'new' or 'established' pairs, depending on whether they had spent more than one breeding season together before the start of 2020.This classification allowed us to compare 22 new pairs with 20 established pairs in 2020 and simultaneously account for age effects in our analyses (see Statistical Analyses for details).

Courtship data
We conducted courtship observations across the whole colony every 3e4 days from when courtship displays were first observed (mid-April) until the start of incubation (mid-May).All observations were conducted by a single observer who was blind with regard to the breeding history and current pairings of birds in the colony.Observations took place between 0830 and 1500 hours; we recorded over 1100 courtship displays over a cumulative total of 18.2 h.During each observation session, we monitored the entire colony and categorized courtship displays based on previous descriptions for this species (Tinbergen, 1959;Van Rhijn & Groothuis, 1987) with the addition of two displays ('circling' and 'walking together') that we frequently observed in our population.We also recorded courtship feeding, where the male regurgitates food for the partner, and attempted or successful copulation.Some displays can be performed in a sequence, especially early in the season (see Van Rhijn & Groothuis, 1987 for a full description); where this was the case, we recorded the event as one multistage display (e.g.forward-to-upright).Not all courtship displays involved two individuals: in some cases, a focal individual either displayed alone (no discernible intended recipient) or directed its display towards a specific individual that did not respond by joining in the display.For each courtship display, we therefore recorded the identities of the individual(s) involved and, where applicable, whether the intended recipient also participated in the display.Lastly, we recorded whether displays were performed synchronously.Synchronous displays (birds displaying together in perfect unison) were always easy to distinguish from nonsynchronous ones, where the movement of one pair member lagged behind the other.

Nesting data
As soon as nesting began (late April), we recorded observations of territory defence (aggressive displays or physical fights) in addition to courtship displays.During each observation session, we also checked for nest building every 3 min and recorded the identity of any birds that were actively building (carrying nest material, adding material to the nest or using the body to make a nest cup) at that time point.We obtained complete clutch sizes for 38 of the 42 nests, excluding two nests where eggs went missing after laying (making it impossible to determine final clutch size) and two nests where we had evidence of conspecific brood parasitism (a persistent but infrequent occurrence in both our population and wild colonies; Van Rhijn & Groothuis, 1985;Duda & Che ˛tnicki, 2012).For a subset of nests (N ¼ 31) we also measured the length and width of eggs and calculated egg volume following Hoyt's (1979) formula for the closely related Larus occidentalis: 0.497 Â length Â width 2 .

Parental care data
After incubation of complete clutches had begun (early May), we collected incubation data over 3 nonconsecutive days using small cameras (GoPro Hero 5), which were placed at various locations above the nesting area and programmed to take a photo every 2 min from 0800 until 2200 hours.We obtained incubation data for 23 pairs.We determined which parent was on the nest in each photo by identifying the colour rings of the incubating bird or its nearby partner; where these could occasionally not be seen, we distinguished members of the pair based on sexual size dimorphism (males are 15% larger than females; Glutz von Blotzheim and Bauer, 1982) or on their unique tail and head markings (see Appendix Fig. A1).We were able to confirm the identity of nestattending birds in the vast majority of photos (mean ¼ 98% of photos identified per nest, range 89e100%), and only used these photos in our calculations of incubation behaviour.

H1: courtship costs hypothesis
To test whether courtship is shorter and less intense in established pairs, leading to an earlier onset of breeding, we calculated courtship display frequency (the number of times the pair were seen displaying to each other/total observation hours since their first courtship display was seen), the Julian date of courtship onset (when the pair was first seen displaying together), courtship duration (days between courtship onset and egg laying) and Julian date of first egg laying.Courtship display data were collected for all 42 pairs and courtship duration was available for 41 pairs (one pair performed courtship and built a nest but did not produce eggs).

H2: behavioural coordination hypothesis
To quantify each pair's behavioural coordination, we calculated courtship display synchrony as (1) number of displays the pair performed synchronously together as a proportion of all the displays where one or both partners displayed to each other, and (2) number of synchronous displays as a proportion of only those displays where both individuals actively joined in the display.We quantified a pair's level of incubation coordination according to nest attendance: in each photo: the nest could be attended by a single parent, both parents or neither parent.For each nest, we therefore calculated coordination in incubation as the proportion of photos where only one parent was in attendance.With respect to behavioural coordination, both double attendance and no attendance constitute poor coordination between parents (Spoon et al., 2006).Cases of double attendance only included those where both birds were physically on the nest, usually both standing over the eggs or one attempting to sit on/displace the other during an incubation bout.While a parent being present on the nest when it is not its turn to incubate may be displaying a higher motivation to care, in doing so it reduces the efficiency of parental labour division; such double attendance indicates that partners are not well aware of when each other's incubation bouts are due to end.We also quantified a second measure of incubation coordination by measuring the proportion of mid-bout incubation switches, where one parent would replace the other, currently incubating parent very briefly, before the initially incubating parent returned and continued its incubation bout.We considered a mid-bout incubation switch to have occurred if the identity of the incubating parent was maleefemaleemale or femaleemaleefemale, in three consecutive nest photos.Since the displaced parent rarely left the nest area and resumed incubating before the next photo was taken 2 min later, it seems unlikely that such mid-bout switches have a function (for example, allowing the initially incubating bird to temporarily leave to get food).Instead, we consider them to be an example of poor communication and coordination between parents over their respective incubation contributions which not only increase the level of activity around the nest but also expose the eggs (albeit briefly) to lower temperatures and threats from conspecifics and predators.We calculated the number of mid-bout switches as a proportion of the total number of photos taken.

H3: partner investment hypothesis
To quantify partner investment (the investment in the breeding partner from the viewpoint of a focal parent), we determined how equally breeding partners divided incubation effort between them ('incubation inequality') as the absolute deviation of the male's incubation contribution from 0.5.To quantify male investment in the breeding partner, we calculated regurgitation frequency (the number of times a male was seen regurgitating for his partner/total observation hours since the pair's first courtship display was seen).To quantify female investment in the breeding partner, we determined egg volume relative to clutch size by weighting the volume of each egg by the final clutch size and then calculating a mean volume across the clutch (N ¼ 28 nests where both final clutch size and egg volumes were known).

H4: partner commitment hypothesis
To quantify the extent to which each individual signalled its commitment to its partner, we calculated nest-building effort (the number of nest-building observations as a proportion of all the nest-building checks performed since nesting began) and territory defence rate (number of nest defence displays/total observation hours since nesting began) for all 84 breeding individuals.We also determined pair display participation as the proportion of all within-pair displays that the focal bird actively participated in, and display infidelity as the proportion of all an individual's displays (including regurgitation and copulation attempts) that were either directed towards or actively involved a bird other than the breeding partner.Courtship interactions can be difficult to distinguish from competitive interactions; however, since neither males nor females defend physical resources or spatial territories in the prebreeding period, there was no reason to expect interactions between members of the opposite sex to be competitive in nature.To limit these extrapair displays to those of a sexual nature rather than those signalling competitive status, we therefore restricted the observations to only include those between members of the opposite sex.

H5: partner compatibility hypothesis
To test whether highly compatible new pairs were also more likely to invest in their partner and pair bond, we restricted our data set to new pairs and tested whether those with high courtship display synchrony, a potential early signal of behavioural compatibility, went on to invest more in their partner and the pair bond (see H3 and H4 above).

Statistical Analyses
All analyses were carried out in R version 3.6.0(R Core Development Team 2019).We inspected our data set for outliers and checked for collinearity by correlating variables against each other and examining variance inflation factors (VIFs).Of particular importance was our ability to distinguish between effects of pair bond duration and age of the parents, which are usually correlated (Sanchez-Macouzet et al., 2014).This is also the case in our population (R 2 ¼ 0.46), but the relatively high divorce rate and the fact that birds are regularly released into the wild (to maintain population health) means that older individuals are regularly paired with younger ones; as a result, VIFs were <2 for all our models (Zuur et al., 2010).
Since we collected numerous and often correlated variables to test each of our hypotheses, we performed principal component analyses (PCA) to avoid multiple testing and reduce the variables relating to each hypothesis to one or two dimensions.For some pairs we were not able to obtain data on incubation behaviour (N ¼ 19) and clutch size-controlled egg volume (N ¼ 14).Prior to the two PCAs that involved these variables (behavioural coordination and partner investment, see below), we therefore imputed missing values using the 'imputePCA' function from the missMDA package (Josse & Husson, , 2016) in R. Tests of this method have demonstrated that PCA results are robust to imputation of missing values of 15e25% (Josse & Husson, , 2016; Severson et al., 2017) and these are therefore unlikely to affect the descriptive quality of our principal components.However, we also ran the two abovementioned PCAs using only nests for which we had no missing values; the results of these analyses were qualitatively the same in spite of the reduced sample size and are shown in Appendix Tables A1 and A2.All variables were scaled and centred before PCA.We retained principal components (PC) with an eigenvalue >1.Variables were considered of strong influence if they loaded on a PC with values greater than 0.4 or smaller than À0.4.
We performed a total of four PCAs to produce dependent variables for each of our first four hypotheses and used the results of these PCAs to test H5: (H1) courtship costs: pair display frequency, date of courtship onset, courtship duration and date of first egg laying; (H2) behavioural coordination: number of synchronous displays (both as a proportion of all pair displays and as a proportion of only those displays where both pair members actively participated), and the proportion of mid-bout incubation switches and single-parent attendance during incubation; (H3) partner investment: incubation inequality, mean egg volume (controlled for clutch size) and male regurgitation frequency; (H4) partner commitment: nest-building intensity, nest defence intensity, display fidelity and display participation.
For each of these four PCAs, we extracted PC scores for each breeding pair or individual and used these as dependent variables in separate models for each PC.For the first three analyses (courtship cost, behavioural coordination and partner investment, H1eH3), each data point corresponded to a breeding pair, and we constructed linear models to test for an effect of pair bond duration on each PC, including chronological age of the pair to control for age effects.Since male and female partner ages are correlated in our population (R 2 ¼ 0.42), we used the mean of male and female age.In the analysis of partner commitment (H4), data points corresponded to individuals, and we constructed linear mixed models with pair ID as a random effect to control for nonindependence of birds from the same breeding pair, and included individual age and sex as covariates.We also tested for an interaction between sex and pair bond duration to determine whether the effect of pair bond duration varies between males and females.To test whether highly compatible new pairs were also more likely to invest in their partner and pair bond (H5, Partner compatibility), we ran the PCAs for partner investment and partner commitment on a data set restricted to new pairs (outputs from these PCAs were qualitatively similar to those on the full data set, see Appendix Table A3) and built linear models to test whether the four PCs for investment and commitment were predicted by a pair's degree of courtship display synchrony.For all analyses, we obtained final models by comparing Akaike information criterion values (corrected for small sample size, AICc) of models with and without additional fixed effects and interactions, considering a difference of AICc >2 to indicate a significantly more parsimonious model.Parameters for terms not present in the final models were obtained by reintroducing them individually into the final model.

Ethical Note
This study was performed on a permanently captive population of wild-type black-headed gulls that live in a large aviary with all the necessary resources and stimulation required to support natural behaviours (see above).Behavioural data used in this study were collected from inside an observation tunnel next to the aviary, which greatly reduces disturbance to the colony.The research is performed with permission of the Animal Welfare Body of the University of Groningen and under a permit issued by the Dutch Competency Authority (permit number AVD1050020209349).

H1: Courtship Costs Hypothesis
The PCA of courtship costs resulted in one PC with eigenvalue >1, which explained 56.4% of the original variation.PC1 reflected low courtship intensity: it was characterized by low display rates, late courtship onset, short courtship duration and early egg laying (Table 1).
The model with both pair bond duration and pair age was not a better fit than the model with only pair age (DAICc ¼ 1.65) but did perform better than the model including only pair bond duration (DAICc ¼ 2.05).Established pairs scored marginally, but significantly, higher on the PC for low courtship intensity (Table 2), meaning that they performed courtship displays less often, started courting later and had a shorter courtship, leading to earlier breeding (Fig. 1).Independent of their pair bond duration, older pairs had lower scores on the PC for low courtship intensity, suggesting that they displayed more often, started courting earlier and took longer to end the courtship and begin egg laying (Table 2).

H2: Behavioural Coordination Hypothesis
Behavioural coordination variables were reduced to two PCs with eigenvalues >1, explaining 86.7% of the original variation.PC1 was linked to low behavioural coordination: low proportions of synchronous displays and low proportions of single parent attendance at the nest (Table 1).PC2 was also associated with low behavioural coordination, particularly with respect to incubation: this component was characterized by a high proportion of midbout incubation switches and low proportions of single parent attendance (Table 1).
The model with both pair bond duration and pair bond age as predictors of PC1 outperformed a model with only pair bond duration (DAICc ¼ 3.06) but was only marginally better than the model with only pair age (DAICc ¼ 1.82).There was a weak positive effect of pair bond duration on behavioural coordination: established pairs had significantly lower PC1 scores (Table 2), suggesting that they displayed more synchronously (Fig. 2) and were less likely to either overlap or both be absent from the nest during incubation.Pairs consisting of two older birds had lower behavioural coordination than those with younger birds, independent of their pair bond status (Table 2).The best two models for PC2 contained a single predictor of either pair bond duration or mean age (DAICc ¼ 1.45); the model containing both terms was worse than the one containing only age (DAICc ¼ 2.26) and equal to the model containing only pair bond duration (DAICc ¼ 0.82).There was no effect of pair bond duration on PC2 (Table 2), but PC2 increased with the mean age of the pair, again suggesting that pairs of older birds had lower behavioural coordination (Table 2).

H3: Partner Investment Hypothesis
The partner investment PCA produced two PCs with eigenvalues >1 that collectively explained 80.7% of the original variation.PC1 described low partner investment and was characterized by infrequent male regurgitation and greater incubation inequality (Table 1).PC2 was also linked to low partner investment, being characterized by small egg volumes and greater incubation inequality (Table 1).
All three models (with either pair bond duration, age or both as predictors) of PC1 were indistinguishable (DAICc < 2).There was no effect of pair bond type or pair age on PC1 (Table 3).The best two models for PC2 contained a single predictor of either pair bond duration or mean age (DAICc ¼ 1.80); the model containing both terms was worse than the one containing only pair bond duration  (DAICc ¼ 2.42) and equal to that containing only age (DAICc ¼ 0.62).Established pairs had significantly lower PC2 scores (Table 3), suggesting that they shared incubation more equally and that females laid larger eggs than those in new pairs (Fig. 3).There was no effect of pair age on PC2 (Table 3).

H4: Partner Commitment Hypothesis
Our partner commitment variables were reduced to two PCs with eigenvalues >1, which explained 64.4% of the original variation.Both PCs described high partner commitment: PC1 was linked to high nest-building and territory defence rates, and to high levels of within-pair display participation (Table 1).PC2 described high fidelity with respect to low levels of extrapair displays and high levels of within-pair display participation (Table 1).
The top model for PC1 contained only sex (DAICc > 2).Males were significantly more committed to nesting-related behaviours, having higher PC1 scores (Table 3), but models containing other predictors performed significantly worse and no other predictors had an effect on PC1 (Table 3).Sex was also the only predictor present in the top model for PC2 (DAICc > 2).There was a nonsignificant tendency for males to have lower PC2 scores than females, and no other predictors had an effect on PC2 (Table 3).

H5: Partner Compatibility Hypothesis
There was no relationship between a new pair's score on either of the investment PCs and their degree of display synchrony during courtship (Table 4), suggesting that new pairs of higher compatibility did not invest more in their partner.There was also no relationship between display synchrony and a new pair's score on the first partner commitment PC (Table 4).However, pairs with high display synchrony had a higher score on the second partner commitment PC (Table 4).This means that new pairs with high display synchrony also had high commitment to the pair bond (high levels of within-pair display participation and low levels of extrapair displays; Fig. 4).

DISCUSSION
In this study, we aimed to test for benefits of remaining with long-term partners and determine whether this added partner value results in parents investing more in their partner and the pair bond.We found evidence for benefits of mate familiarity in blackheaded gulls: pairs that were already established in previous years had less intense courtships than new pairs and were also more behaviourally coordinated during courtship and incubation.Parents in established relationships also invested more in their partner: females in established pairs laid larger eggs and incubation share was more equal in established pairs than in new pairs.In addition, we found that individuals in new partnerships with high display synchrony signalled their commitment to the partnership more than those that were less synchronous during courtship, suggesting that selection can favour partner investment even before a long-term pair bond has developed and may explain the existence of mate familiarity effects and the subsequent selection on long-term partner investment in this species.Taken together, our results suggest that mutual benefits of valuable long-term breeding partnerships might act to increase levels of investment in current breeding partners, which may in turn decrease the level of parental conflict over care in many species.

Future Fitness Value of Long-term Partners
We predicted that whether a focal parent pays a cost to forcing its current partner to provide super-optimal levels of parental care depends on the extent to which it gains fitness benefits from breeding with that partner again.Lending support to evidence of the mate familiarity effect from other species, we found that established black-headed gull pairs had a shorter and less intense courtship (Fig. 1).Established pairs displayed less often, and individuals involved in frequent courtship displays must sacrifice time and energy that can otherwise be invested in finding food, establishing a nest site and (especially in the case of egg-laying females) conserving energy for imminent reproduction.Engaging in frequent courtship displays may also put individuals at greater  1).Squares and bars denote mean and standard errors per group; dots represent raw data.1).Squares and bars denote mean and standard error per group; dots represent raw data.risk of predation, both because the act of displaying itself is visually conspicuous (Magnhagen, 1991) and because predator vigilance might be compromised during reproductive behaviours (Blanchard et al., 2017;Hess et al., 2016;Rosa & Murphy, 1994).In addition, the courtship period of established pairs started later and was shorter, leading to earlier onset of egg laying (Fig. 1).The main benefit of shorter and more efficient courtship in birds is that it might enable females to start laying earlier, which is linked to reproductive success in many species (reviewed in Verhulst & Nilsson, 2008).Indeed, female black-legged kittiwakes, Rissa tridactyla, and redbilled gulls, Larus novaehollandiae, are known to lay earlier when mated to a familiar partner (Coulson, 1966;Mills, 1973).Taken together, it therefore seems likely that black-headed gulls that retain established partners can expect future reproductive benefits of a relatively cheap courtship.Indeed, reduced costs of courtship have been invoked to explain mate familiarity effects elsewhere (Chardine, 1987;Griggio & Hoi, 2011); this is likely to be an important selective force on mate retention in any species where reproduction is preceded by a long or intense courtship period and pairs generally mate for more than one reproductive season (reviewed in Irwin, 1996).

New
In addition to reduced courtship intensity, we also found evidence for increased behavioural coordination among established pairs.These pairs performed a greater proportion of their displays synchronously, suggesting that birds in existing partnerships were better able to predict and coordinate with their partner's intentions and movements.We are not aware of any studies that have quantified the degree of movement synchrony within the displays of species with mutual courtship.In alliances of male bottlenose dolphins, Tursiops aduncus, synchrony of surfacing behaviour is associated with the strength of the alliance (Connor et al., 2006) and monogamous pairs of zebra finches, Taeniopygia guttata, have more coordinated movements than nonpartners (Prior et al., 2020), suggesting that synchrony is at least partly linked to familiarity or affinity.While it is unlikely that such movement synchrony could have a direct effect on reproductive success, it can serve as an indication of the level of behavioural compatibility of a pair with respect to parenting behaviours, which itself is an important predictor of reproductive success in many species (e.g.Burtka & Grindstaff, 2015;Ouyang et al., 2014;Spoon et al., 2006).In line with this, we also found that established pairs were more coordinated in their incubation behaviour.Nests of established pairs were more often attended by a single parent, suggesting that established pairs are better able to coordinate their respective incubation bouts to minimize both overlap of nest attendance and miscommunication that results in the nest being left unattended (Spoon et al.,  .The relationship between courtship display synchrony and commitment to the pair bond in 22 newly formed black-headed gull pairs.Partner commitment was calculated as a principal component (PC2) describing a high proportion of courtship displays to the breeding partner and a low proportion of displays to extrapair individuals (see Table 1).Line and shading represent regression slope and standard error; dots represent raw data.2006).Coordination of parental care has been linked to decreased parental conflict over care (Johnstone et al., 2014;Patrick et al., 2020) and increased reproductive success (e.g.Bebbington & Hatchwell, 2016;Mariette & Griffith, 2015), suggesting a potential fitness benefit to the higher coordination seen in established pairs.In black-headed gulls, where egg dumping and egg stealing are relatively common both in captive (Van Rhijn & Groothuis, 1985; K. Bebbington & T.G.G.Groothuis, personal observation) and in wild colonies (Duda & Che ˛tnicki, 2012;Van Rhijn & Groothuis, 1985), parents that successfully coordinate and alternate their nest attendance such that one partner is always present have a higher ability to guard the nest against conspecifics.Minimizing overlap in nest attendance to ensure that both parents do not needlessly attend the nest together is also potentially beneficial in that it optimizes the ability of nonincubating parents to replenish their reserves (e.g. by foraging), potentially increasing the effectiveness of their care when they return to the nest (Kavelaars et al., 2021).

Selection for Increased Partner and Pair Bond Investment
We predicted that parents that benefit from continued association with their current partner should be under selection to retain that partner.One way in which parents can increase the chance of retaining a valuable partner is to reduce the partner's costs of care, thereby reducing its risk of mortality or excessive resource depletion.We found support for the partner investment hypothesis: established pairs shared incubation more equally than new pairs, suggesting that partners in established pairs aim to reduce costs of care for each other.In addition, females in established pairs laid larger eggs (relative to clutch size) than those in new pairs, which we suggest is a mechanism to spare the resources of the male parent.In another gull species, females that have skipped a year of breeding have been shown to lay larger eggs when mated to familiar males (Salas et al., 2020), suggesting that, potentially like black-headed gulls, females with high levels of resources to invest in reproduction do so more when their partner's resources are considered more valuable.While future studies are needed to confirm such an effect in black-headed gulls, the idea that hatchlings from smaller eggs require more posthatching care is supported by studies showing that females lay smaller eggs when they expect to receive more help with nestlings (e.g.Russell et al., 2007;Taborsky et al., 2007), and evidence from burrowing beetles that parental care removes negative effects of small egg size on larval survival (Monteith et al., 2012).Interestingly, the frequency of male regurgitation, which might provide males with a means of investing in valuable female partners, was not related to pair bond duration.This is surprising, since regurgitation has been shown to be nutritionally important for female gulls: the frequency of regurgitation tends to peak shortly before egg laying (black-backed gulls, Brown, 1967;common terns, Sterna hirundo, Nisbet, 1973;kittiwakes, Neuman et al., 1998;whiskered terns, Chlidonias hybrida, Ledwo n & Neubauer, 2018), and regurgitation is also related to reduced female foraging requirements in red-billed gulls (Tasker & Mills, 1981) and increased clutch and third egg sizes in glaucous-winged gulls, Larus glaucescens (Salzer & Larkin, 1990).In red-billed gulls, males that regurgitated more frequently for their partner were more likely to remain mated to that individual in the following year (Mills, 1994), which suggests that partner investment as a mechanism to retain valuable breeding bonds does occur in at least some systems.The lack of effect in our population may be an artefact of the ad libitum feeding regime, which frees the females from energetic constraints on egg production.
We also predicted that individuals that benefit from retaining their current partner might be under selection to signal their commitment to the current pair bond as a means of demonstrating their attractiveness or quality as a partner and avoiding costly divorce.Evidence in another gull species suggests that a female's participation in extrapair copulation attempts was negatively linked to the amount of male regurgitation she received from her partner (Mills, 1994), suggesting that social signalling can indeed be an important indication of partner value.However, we found no support for the partner commitment hypothesis in black-headed gulls: individuals in established pair bonds were not more likely to engage in conspicuous nesting behaviours, and were also not less faithful to their partner in terms of their engagement in either within-pair or extrapair displays.It is worth noting that certain factors that we tested under the behavioural coordination hypothesis, namely the prevalence of double nest attendance and the proportion of mid-bout switches during incubation, could also be interpreted as representing signals of commitment to nesting.However, we found no evidence that mid-bout incubation switches were more common in established pairs, as we would expect if this behaviour is indicative of partner commitment.Established pairs were also less likely to attend the nest together, which again is the opposite of what we would expect if individuals show more commitment to valuable partners.Displays between birds outside a pair bond were not uncommon, representing ca.25% of all courtship and occurring throughout the courtship period.While commitment to within-pair displays does not seem to be greater in long-term pairs, it would none the less be extremely interesting to test whether birds that engaged in more extrapair displays are more likely to divorce in the following year.
In contrast to established pairs, the principal strategy for retaining valuable new partners appeared to be signalling commitment to the pair bond.We found that new pairs with a high level of display synchrony (a potential indicator of general behavioural compatibility) were more likely to commit to the pair bond, performing fewer extrapair displays and participating in a higher proportion of within-pair displays.This result suggests that selection on retention of valuable partners already begins in the first years of a partnership, but that the mechanism of retention is different.While established pairs retained their partners by sparing the partner's reproductive costs (e.g. by contributing more to incubation or prenatal resources, see above), new pairs seemed to signal their intention to retain that partner by remaining socially faithful.Signalling commitment to the pair bond may be less costly than actively reducing the costs of parental care for a valuable partner; for individuals in new partnerships where the future fitness benefit of retaining their partner is still uncertain, adopting the former strategy may be less risky and thus more strongly selected for.We therefore suggest that the mechanism driving the mate familiarity effect in this species, and the subsequent selection on investment in valuable partners, is selection on the retention of compatible partners in early pair formation.

Partner Value as a Mitigator of Sexual Conflict
We present evidence for mate familiarity effects in black-headed gulls that may alter the optimum parental investment of breeding partners and ultimately affect the level of parental conflict over care.Although we tested our hypotheses in a captive population, the colony is only very recently removed from its wild ancestry; the behaviours and underlying mechanisms we report on are likely to be very similar to those observed in wild colonies.Indeed, the fact that we found effects of pair bond duration on courtship and reproductive behaviours, despite the presence of ad libitum resources, suggests that mate familiarity may have even more severe effects on parental investment decisions in free-living populations and species.It is also important to mention that we only measured parental investment until the end of the incubation period, rather than also incorporating nestling feeding.In general, there is little evidence that increased investment in incubation affects investment in offspring provisioning (Nord & Nilsson, 2012;Sakaluk et al., 2018) but incorporating parental investment across the whole spectrum of reproductive behaviours will certainly give a better picture of how partner value alters investment in different aspects of the pair bond.While the results we present here therefore warrant repeating in wild populations and over complete reproductive cycles, the collective evidence strongly points to a role for partner value in mitigating parental conflict over care, which can have profound consequences for our understanding of reproductive behaviour and the evolution of mating systems.
In this study, we were only able to demonstrate the value of retained partners and reproductive investment decisions that correlate with that value, but not that fitness differs between individuals that do and not invest in valuable partners.In other words, it remains to be seen whether individuals that force valuable partners to provide costly care indeed suffer punishment in terms of increased risk of divorce, or partner death or abstinence from future breeding and, crucially, that this reduces their future reproductive success.Interestingly, punishment or threat of punishment was recently shown to have considerable influence on the theoretical outcomes of parental negotiation of parental care (Barbasch et al., 2020).This opens up the possibility that the future fitness consequences of current reproductive investment reach far beyond the traditional consideration of life history trade-offs and can include social selection on partner-mediated costs, cooperation and punishments.

Conclusions
In this study, we have shown that future benefits of retaining a current partner are linked to increased investment in that breeding partner and in the pair bond.We argue that selection for mate retention can be one important mechanism for reducing the level of conflict over parental investment in species with biparental care and suggest that social factors such as partner value might be a key determinant of reproductive decisions.More broadly, selection to retain valuable social partners might be a general mechanism to facilitate alignment of individual fitness optima and solve the evolutionary puzzle of how cooperation can prevail among fundamentally selfish organisms.(a) (b) . Incubation photos demonstrating that differences in size and tail and head markings can be used to distinguish between breeding partners in black-headed gulls.In (a), the female (left) is visibly smaller and has more white tail feathers than the male (right).The black 'hood' on the head of the male is also much fuller and extends further back on the head.In (b), the distinct patterning on the tail feathers of the female (top) and male (bottom) can be seen.

Figure 3 .
Figure3.Degree of investment in the breeding partner by black-headed gulls.Differences between new (N ¼ 22) and established (N ¼ 20) gull pairs are displayed with respect to the principal component of low partner investment (PC2), which describes small egg size and unequal share in incubation (Table1).Squares and bars denote mean and standard errors per group; dots represent raw data.

Figure 2 .
Figure 2. Behavioural synchrony in black-headed gulls.Differences between new (N ¼ 22) and established (N ¼ 20) gull pairs are displayed with respect to the principal component of low behavioural synchrony (PC1), which describes low display synchrony and low proportions of time where only one partner was present at the nest during incubation (see Table1).Squares and bars denote mean and standard error per group; dots represent raw data.

Figure 4
Figure4.The relationship between courtship display synchrony and commitment to the pair bond in 22 newly formed black-headed gull pairs.Partner commitment was calculated as a principal component (PC2) describing a high proportion of courtship displays to the breeding partner and a low proportion of displays to extrapair individuals (see Table1).Line and shading represent regression slope and standard error; dots represent raw data.

Table 1
Principal component analysis on variables relating to four hypotheses about parental conflict in black-headed gulls, showing loadings and correlation with the original data (plus significance) for each factor Factors that load with !0.4 or À0.4 are highlighted in bold.Total variance refers to the percentage of variance in the original data set that is explained by a given PC.

Table 2
Parameter estimates for models of courtship and coordination variables in relation to pair bond duration in black-headed gull pairs Significant predictors are highlighted in bold.Values for established pairs are presented relative to new pairs.Intensity of courtship displays in black-headed gulls.Differences between new (N ¼ 22) and established (N ¼ 20) gull pairs are displayed with respect to the principal component of low courtship intensity (PC), which describes low frequency of courtship, late onset of courtship and short courtship duration (see Table1).Squares and bars denote mean and standard errors per group; dots represent raw data.

Table 4
Model estimates for the effect of courtship display synchrony (explanatory variable) on principal components of partner or pair bond investment in 22 new black-headed gull pairs

Table 3
Parameter estimates for models of partner investment and commitment in relation to pair bond duration in black-headed gull pairs Significant predictors are highlighted in bold.Values for established pairs are presented relative to new pairs, and values for males are presented relative to females.

Table A2
Parameter estimates for models relating to behavioural coordination and partner investment in black-headed gulls, using only pairs for which no data were missing Significant predictors are highlighted in bold.Values for established pairs are presented relative to new pairs.Principal component analysis of variables relating to partner investment and social commitment in newly formed black-headed gull pairs, using only pairs for which no data were missing Factors that load with !0.4 or À0.4 are considered important contributors to a principal component and are highlighted in bold.

Table A1
Principal component analysis of variables relating to behavioural coordination and partner investment in black-headed gulls, using only pairs for which no data were missing Factors that load with !0.4 or !À0.4 are considered important contributors to a principal component and are highlighted in bold.