Mating reduces responsiveness to sexual stimuli in females but not in males

In

Mating behaviour is complex, and deciding on the best way to design mate choice experiments is often difficult (Dougherty, 2020).One key experimental design decision relates to the mating history ('mating status') of the animals used in the experiment: in other words, whether to use mated (nonvirgin) or unmated (virgin) individuals during mating trials (Tanner et al., 2019;Dougherty, 2020;Richardson & Zuk, 2023).The term 'mating history' is shorthand for 'has participated in a successful mating interaction since sexual maturity'.In species that are not strictly monogamous, both mated and unmated individuals can be used for mating experiments.However, there are several reasons to think that mated and unmated individuals may behave differently from each other.One important expected difference between mated and unmated animals is their responsiveness to sexual signals, also often referred to as the 'receptivity to mating.'In this context, responsiveness reflects the motivation to mate, and refers to the likelihood that an individual will respond positively to the presentation of any potential mating opportunity, or to any valid signals produced by a member of the opposite sex (Bailey, 2008;Brooks & Endler, 2001;Edward, 2015).Responsive individuals are expected to engage with any sexual stimuli quickly and more often (Bailey, 2008).This contrasts with an individual with a strong mating preference, who will only respond to the subset of stimuli that they find attractive (Edward, 2015;Rosenthal, 2017).A standard approach to measuring responsiveness is therefore to record the average response of a subject to the presentation of multiple mate options (Brooks & Endler, 2001); doing this enables us to distinguish a heightened response to all valid sexual stimuli from a mating preference for a specific stimulus.
In species with internal fertilization, mated females are often reported to be less responsive to sexual stimuli compared to unmated females (e.g.Houde, 1987;Judge et al., 2010;Modak et al., 2021).Ultimately, this is thought to arise due to the different benefits associated with the first mating compared to any subsequent matings.Female reproductive success is often more strongly determined by the number of eggs produced than the number of matings (Clutton-Brock & Parker, 1992;Trivers, 1972).Therefore, the reproductive benefit of the first mating is always high, as it allows females to produce fertilized eggs.Failing to respond to a potential mating partner before the first mating also comes with the extra risk of a failure to mate before death (Barry & Kokko, 2010;Greenway et al., 2015;Kokko & Mappes, 2005).For females, the benefit of each subsequent mating is typically lower than the first, and the cost of failing to respond to a viable mating partner is greatly reduced.This leads to the prediction that females should be more responsive (and show weaker mating preferences) when they are unmated.Indeed, in some species, mating induces changes in female reproductive physiology which greatly reduce receptivity to mating (known as a 'refractory period'), often for several days or weeks (Ringo, 1996;Wedell, 2005).For example, in the fruit fly Drosophila melanogaster, mating triggers a cascade of neuronal and physiological changes which cause females to drastically increase rejection behaviour towards males for several days (Chapman et al., 2003;Feng et al., 2014).In some internally fertilizing species, the length of the refractory period can also partly be under male control, because the seminal fluid contains accessory proteins which interact with female physiology to suppress her remating (Smith et al., 2017;Wigby et al., 2020).Lengthening the refractory period benefits the male by reducing the chance he will have to share paternity with a rival.
Mating history is not typically considered to be an important determinant of male responsiveness.This is because males tend to invest less in reproduction than females, either in terms of the size of gametes or the allocation of parental care, and so have a greater potential reproductive rate (Clutton-Brock & Parker, 1992;Trivers, 1972).This means that male reproductive success tends to increase more steeply with each subsequent mating (the Bateman gradient) than for females (Bateman, 1948;Janicke et al., 2016;Kokko et al., 2012).Males therefore benefit more from achieving as many matings as they can.Importantly, the relative investment in reproduction is not fixed across species, and sexual selection theory therefore predicts that mating history should have a strong effect on responsiveness in whichever sex invests the most.It would therefore be especially useful to examine males of 'sex-rolereversed' species (Hare & Simmons, 2019).However, it is worth noting that the cost of failing to respond to a viable mating partner does still differ between mated and unmated males.This could in theory lead to an increased responsiveness for unmated males, although it is not clear how strong this effect is.
A loss of responsiveness after mating is commonly claimed as a justification for avoiding the use of mated females in mating experiments (e.g.Houde, 1987).This decision may be justified if mating commonly leads to a large reduction in responsiveness.However, this is not always the case (e.g.Kodric-Brown & Nicoletto, 2001;McNamara et al., 2004), and there has been no attempt to formally quantify this effect using data from multiple animal species.Richardson and Zuk (2023) recently used meta-analysis to test the prediction that mated females should show stronger mating preferences than unmated females.They found no overall difference in the strength of mate choice between mated and unmated females; instead, the strength of mate choice appears highly variable across species and contexts (Richardson & Zuk, 2023).This high variation in mating behaviour may make it difficult to find a clear signal attributable to one causal factor.Further, while mating history is considered frequently in relation to female behaviour, it is less often considered in relation to male behaviour.Nevertheless, several studies have found that males become less responsive after mating (e.g.Fischer & King, 2012;Wang et al., 2016).Meta-analysis is useful for identifying patterns when empirical results are highly variable (Koricheva et al., 2013).I here used meta-analysis of 44 studies to quantify how mating influences the responsiveness of female and male animals to sexual stimuli, in species with internal fertilization.Because females typically benefit less from additional matings than males, I predicted that mating should reduce responsiveness in females, but not for males.

METHODS
Throughout I follow the reporting guidelines suggested by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses in Ecology and Evolutionary Biology (PRISMA Eco-Evo) framework by O'Dea et al. (2021).

Literature Search and Study Screening
I relied primarily on the 7158 papers examining statedependent mating behaviour collected by Dougherty (2021aDougherty ( , 2023a)).This database was produced by obtaining all papers cited by two reviews of state-dependent mate choice (Ah- King & Gowaty, 2016;Cotton et al., 2006), and keyword searches using the online databases Web of Science and Scopus on 13 August 2019.See Dougherty (2021a) for full details of the search strings used.I exported all 7158 studies into the systematic review software Rayyan (Ouzzani et al., 2016), and manually screened all titles and abstracts using my inclusion criteria (see below).All articles were screened by myself: while Rayyan can use machine learning to make inclusion decisions, I did not make use of this function.From these searches I identified 60 relevant studies.I also searched Google Scholar on 26 August 2022 using the following phrases: 'virgin[mated] females[males] were[are] more[less] responsive ' and 'virgin[mated] females[males] respond[ed] more[less]'.These searches resulted in two additional relevant studies, for a total of 62 studies which I then read in full.The literature search and study screening process is summarized in Fig. A1.
To be included in the meta-analysis, a study needed to: (1) focus on internally fertilizing animals (not humans); (2) record some behaviour that relates to sexual receptivity; (3) present behavioural data for mated and unmated individuals (of either sex); and (4) provide appropriate data to calculate an effect size.
I considered mating status in relation to an individual's entire reproductive life.I only included studies where virginity could be confirmed with a high confidence.This meant I only considered laboratory-reared individuals that had been separated into singlesex groups prior to sexual maturity, or wild-caught individuals if they were guaranteed to be sexually immature (Iglesias-Carrasco et al., 2019;Maklakov et al., 2003;Meyer & Uetz, 2019;Sivalinghem et al., 2010;Tak acs et al., 2003;Uetz & Norton, 2007;Zeh & Zeh, 2007), or to have not yet entered the mating pool (Hoeck & Garner, 2007).I excluded eight studies that used wild-caught individuals for which virginity could not be confirmed.I included studies that presented individuals with members of the opposite sex or with their sexual signals.
I included two categories of sexual behaviour.First, the 'chooser behaviour' category included any behaviour reflecting sexual interest in members of the opposite sex or their signals (orientation and approach, association or interaction with, mating attempts), or behaviour expressed in response to sexual interest by members of the opposite sex (receptive displays, acceptance or rejection of a mating attempt).Second, the 'mating' category included mating frequency and mating latency.Behaviours in the first category primarily reflect the decisions of the choosing individual, whereas behaviours in the second category may be the result of decisions made by both mating partners.This is important, because males are often less attracted to mated females (Bonduriansky, 2001;Richardson & Zuk, 2024).Therefore, changes in female responsiveness in terms of mating or mating latency could result entirely from changes in male courtship behaviour.I considered behaviours measured in terms of occurrence, duration, rate or latency.I did not include courtship or sexual signalling behaviour, as this has already been analysed (Dougherty, 2021a).I classed individuals as more responsive when they: (1) exhibited more receptive behaviours or fewer rejection behaviours, (2) spent more time associating with, interacting with or attempting to mate with potential mating partners, or less time rejecting sexual advances, (3) oriented towards, approached or attempted to mount potential mating partners sooner, or (4) were more likely to mate or mated sooner.

Effect Size Calculations
I used the standardized mean difference (Hedges' g) in responsiveness between mated and unmated individuals as the response variable in the analysis (Hedges & Olkin, 1985).I coded effect sizes as positive when unmated individuals were more responsive, and negative when unmated individuals were less responsive.Hedges' g was either calculated using reported means and standard deviations or converted from common statistical tests (e.g.t test, chi-square test, ManneWhitney U test) using the equations in Koricheva et al. (2013, p. 200).I used the online tool WebPlotDigitizer v4 (https://apps.automeris.io/wpd/) to extract means and standard deviations from bar plots.One study compared responsiveness of the same individuals before and after mating using a paired t test (Gemeno Marín et al., 2007).Here, I assumed a correlation of 0.5 between the two measures of responsiveness.In this case, the effect size calculation is the same as for an independent-measures t test.I extracted more than one effect size from 15 studies, either because the study reported multiple valid responsiveness behaviours, or because both male and female responsiveness was reported.I found two cases in which an effect size could be calculated but the sign of the effect could not be determined: one in males (Cook, 1994) and one in females (Kodric-Brown & Nicoletto, 2001).These data points are traditionally excluded from meta-analyses.However, this systematically biases the data set against the inclusion of nonsignificant results (Harts et al., 2016).I therefore included these two data points but assigned them a value of zero.

Statistical Analysis
All statistical analyses were performed using R v4.1.2(R Development Core Team, 2021).To account for the potential nonindependence of correlations from closely related species, I first constructed a supertree containing all 38 species using the Open Tree of Life (OTL) database (Hinchliff et al., 2015), and the Rotl v3.1 (Michonneau et al., 2016) and Ape v5.7 (Paradis et al., 2004) R packages.Given the absence of accurate branch length data for these trees, branch lengths were first set to one and then made ultrametric using Grafen's method (Grafen, 1989).
Meta-analyses were performed using the R package Metafor v3.4 (Viechtbauer, 2010).I performed a multilevel meta-regression using the rma.mv function in Metafor, with Hedges' g as the response variable, weighted by study sampling variance, and two categorical fixed factors: the sex of the focal individual (male or female) and the behavioural category ('chooser behaviour' or 'mating').Because females typically benefit less from additional matings than males, I predicted that mating would reduce responsiveness for females, but not for males.I included the behavioural category as a fixed factor to test whether changes in responsiveness are due to chooser decisions or to changes in attractiveness.If the latter is most important, then the change in responsiveness following mating will be larger for studies recording mating behaviour compared to those recording chooser behaviours.The model also included phylogeny, study ID and observation ID as random factors.Phylogeny was incorporated into the model using a correlation matrix, calculated assuming that traits evolve via Brownian motion.An observation level random factor (observation ID) is required to correctly estimate residual heterogeneity.I used the QM statistic to test whether the mean effect size differs for males and females.I also used a minusintercept model to determine the mean effect size, plus 95% confidence intervals, for males and females separately.Effect size estimates are significantly different from zero if their 95% confidence intervals do not overlap zero.
I used I 2 as a measure of total heterogeneity, which is the proportion of variance in effect sizes that is not attributable to sampling (error) variance (Higgins et al., 2003).I used the method of Nakagawa and Santos (2012) to partition heterogeneity among the random factors in the model.I 2 values of 25, 50 and 75% are considered low, moderate and high, respectively (Higgins et al., 2003).Heterogeneity above 75% is common to ecological metaanalyses incorporating data from many species (Senior et al., 2016).
I searched for two signs of publication bias.First, I tested for a change in the average effect size over time, which could arise if studies with nonsignificant results are more likely to be published as a research field ages (Jennions & Møller, 2002).Second, I tested whether there was a significant relationship between effect size and study variance driven by asymmetry caused by 'missing' effect sizes.This may provide evidence for publication bias if 'missing' effect sizes are small in magnitude and come from studies with small sample sizes (a 'small study effect': Koricheva et al., 2013).I tested for this using the same multilevel meta-regression as described above, but with either publication year or study variance as a fixed effect.

RESULTS
I obtained data from 41 papers and 38 species overall: 38 estimates from 26 species and 28 studies for females, and 22 estimates from 15 species and 16 studies for males (Table 1).For females, only two vertebrate species were represented; the rest were arachnids (seven species) or insects (17 species).For males, three vertebrates, two arachnids and 10 insect species were represented.Most studies used laboratory-reared individuals (31 of 38 estimates for females; 20 of 22 estimates for males).
Female responsiveness was more strongly affected by mating history than male responsiveness (Q M ¼ 10.64, P ¼ 0.001, N ¼ 60; Fig. 1).Females were more responsive to male signals when they were unmated (Hedges' g ¼ 0.56, 95% CI ¼ 0.24 to 0.88; Fig. 1).In contrast, there was no difference in responsiveness to female signals between mated and unmated males (Hedges' g ¼ 0.08, 95% CI ¼ À0.27 to 0.42; Fig. 1).The effect of mating history on responsiveness did not depend on whether responsiveness was measured in terms of chooser behaviour or mating behaviour (Q M ¼ 0.92, P ¼ 0.34, N ¼ 60).Overall heterogeneity was high (total I 2 ¼ 84.5%), with 20% being attributable to phylogenetic history, and the remaining 64.5% attributable to observation level differences.Study explained a negligible amount of heterogeneity.We found no evidence for publication bias: the mean effect size was not related to study year (Q M ¼ 0.31, P ¼ 0.58, N ¼ 60) or study precision (Q M ¼ 0.12, P ¼ 0.73, N ¼ 60).

How Many Studies Use Only Unmated Females?
I performed an informal survey of the sexual selection literature, to identify studies that collate information on the mating status of subjects in mating experiments.I identified two previous studies that reported the mating history of females used in mate choice experiments.Tanner et al. (2019) surveyed 24 studies examining female mate choice in field crickets (Gryllidae) reared in the laboratory and found that 79% of studies used only unmated females.Richardson and Zuk (2023) recently performed a meta-analysis testing whether studies using mated females recorded stronger mating preferences than studies using unmated females, including data from both vertebrates and arthropods.Just over half (53%) of 303 studies examined used only unmated females, and for some types of studies this percentage was much higher, reaching 74% for studies examining choice of related versus unrelated males (Richardson & Zuk, 2023).To supplement these data, I surveyed studies used in two recent meta-analyses examining mate choice in relation to the environment (Dougherty, 2021b) and individual state (Dougherty, 2023a), focusing on studies using laboratoryreared animals.This data set included data from arthropods and vertebrates.Of 82 studies examining female mate choice, 76% used  only unmated females (Fig. 2).Of 27 studies examining male mate choice, 44% used only unmated males (Fig. 2).

DISCUSSION
This meta-analysis provides clear quantitative evidence that females of internally fertilizing species become less responsive to sexual stimuli after their first mating.This seems to be a general trend, rather than being limited to a few well-studied taxa, and is not primarily driven by male mating preferences.Notably, my nonsystematic survey of the mate choice literature shows that most studies examining female mate choice use exclusively unmated females as subjects.In contrast, I found no evidence that mating alters male responsiveness, and this likely explains why studies of male mate choice do not tend to favour using unmated over mated individuals.
The reduction in female responsiveness following mating can be explained by the fact that female reproductive fitness often does not relate strongly to the number of matings she attains (Clutton-Brock & Parker, 1992;Janicke et al., 2016;Trivers, 1972).This means that females typically benefit more from investing in offspring produced from a single mating than by pursuing further matings.A postmating reduction in responsiveness will benefit females if it reduces her chance of obtaining further matings that are not beneficial (Arnqvist & Nilsson, 2000).Alternatively, it may be harmful, if it results in increased harassment or aggression by males (Arnqvist & Rowe, 2005).This also means that males could waste more time and energy trying to court recently mated females that are unresponsive, although in many species males appear to avoid this cost by preferentially courting and mating with unmated females (Bonduriansky, 2001;Richardson & Zuk, 2024), and females can advertise their mating status through changes in chemical cues (Thomas, 2011).
I found no consistent evidence that mating alters male responsiveness to sexual stimuli.The difference between females and males is likely due to differences in overall reproductive investment.Females generally invest more in each reproductive episode than males, so that the relationship between the number of matings and reproductive success is typically steeper for males.This means that males benefit more from pursuing matings than females (Clutton-Brock & Parker, 1992;Janicke et al., 2016;Trivers, 1972).Notably, sexual selection theory predicts that we should see a postmating reduction in male responsiveness in species where males invest more in reproduction than females (often referred to as 'sex-role-reversed' species; Fritzsche et al., 2021).However, there are no such species in the current data set.Nevertheless, it also follows that we should see a reduction in male responsiveness whenever mating starts to become costly for males, even without full reversal of sex roles.It would be informative to examine the responsiveness of males that invest heavily in mating (for example in species with large nuptial gifts: Lewis & South, 2012), or to compare subsequent matings after which sperm or other resources may become depleted (e.g.Macartney et al., 2021).Another reason for sexual dimorphism in response to mating may relate to the nature of internal fertilization, in which males almost always donate gametes to the female.This provides males with the opportunity to influence female receptivity via compounds transferred in the seminal fluid (e.g.Chapman et al., 2003;Feng et al., 2014).Here, changes in female sexual receptivity after mating could be an adaptive male strategy to reduce the degree of sperm competition he faces.However, so far evidence for such an effect has been found in only a few well-studied insect species (Avila et al., 2011).Even in sex-role-reversed species, females cannot typically do this to males (although pipefish females may transfer eggs to males along with ovarian fluid: Watanabe et al., 2000;Paczolt & Jones, 2010).
This meta-analysis has several limitations.First, the sample size, only 41 studies, was relatively small.Partly this is due to my inclusion criteria: seven studies were excluded because statistical issues prevented me from calculating an effect size, and eight were excluded because the authors could not confirm mating status with high confidence.My literature searches were also not exhaustive: I used a sample of collected studies from a previous meta-analysis (Dougherty, 2021a(Dougherty, , 2023a) ) to make the searches more manageable.I therefore expect there are more published papers that I did not include.Nevertheless, my hope is that the sample I have collected is unbiased and representative.Second, studies typically only record responsiveness once following mating.This means it is not clear how long the reduction in responsiveness following mating lasts, for females or males.If responsiveness is only reduced for a short-time following mating, then its ecological relevance may be limited.Even extreme postmating reductions in female responsiveness may relax as stored sperm are depleted.However, studies typically retest individuals within a few days of mating.This means that the effect sizes collated here most appropriately reflect changes in responsiveness following a recent mating.Third, the best test of this question would measure the responsiveness of the same individuals before and after their first mating (Dougherty, 2023b), as this would control for any initial differences in responsiveness between individuals.However, only eight studies in the meta-analysis (seven studies for females, two for males) did this.This method also results in an additional experimental issue: mating behaviour can be dependent on prior mating experience (e.g.Jordan & Brooks, 2012).Experiments using a repeated-testing approach may therefore need creative ways to control prior mating experience.

Experimental Design Considerations
The present meta-analysis shows a clear trend for mating to reduce female responsiveness in internally fertilizing species.This result has important practical implications for the design of mating experiments because individuals that do not respond during mating trials (either from one or multiple tests) are often excluded from any analyses (Kokko & Jennions, 2015;Rosenthal, 2017).Therefore, one clear advantage of using only unmated females is that a smaller initial sample size will result in the same number of responsive individuals.Using unmated females also has other experimental design benefits, such as removing the need to account for differences in prior experience (Dougherty, 2020).
Therefore, there will often be clear practical benefits of using only unmated females in mating experiments.However, I suggest that we should be wary of doing this, for three reasons.First, mating history may alter other important aspects of mating behaviour.For example, individuals may change their mating preferences based on their experience of previous mates (e.g.Jordan & Brooks, 2012).However, note that recent meta-analyses failed to find any effect of mating on the production of male sexual signals (Dougherty, 2021a) or the strength of mate choice in either sex (Richardson & Zuk, 2023;Dougherty, 2023a).Additionally, males are known to behave differently when encountering mated and unmated females, courting unmated females more (Bonduriansky, 2001) and investing more in ejaculate and seminal fluid when mating with previously mated females (e.g.Lüpold et al., 2010;Sirot et al., 2011)).Second, responsiveness is an important behaviour in and of itself, which may tell us something about female preferences.This is because failing to respond can be a form of choice, especially when mates are encountered singly (Dougherty & Shuker, 2015).Removing nonresponders thus risks skewing the results in favour of responsive individuals.This may be especially problematic if individuals are nonresponsive for a reason; for example, they may be more attractive or in better condition (Dougherty, 2023a).
Additionally, having a large proportion of nonresponsive individuals in an experiment might suggest that the sexual stimuli you present them with are not very stimulating (Dougherty, 2020).
The third reason to consider including mated individuals in mate choice experiments is that in natural populations unmated individuals may often be in the minority.If this is the case, experimental practices could lead to systematic differences between laboratory and field studies.It is currently difficult to know how much of a problem this is, as mating frequency in the wild is unknown for most populations.Studies quantifying female mating history in insects typically find that the proportion of unmated females in a population is less than 50%, and is often below 10% (Rhainds, 2010).However, there is significant variability in mating frequency estimates, and data from individual species may record the proportion of females that are unmated at over 50% (Rhainds, 2010(Rhainds, , 2019)).Extrapolation of insect data to other animal groups is difficult, however, as mating probability is likely to be affected by a range of factors, including the population density, the adult sex ratio, when individuals are sampled, and the mating and breeding system of each species (Rhainds, 2010(Rhainds, , 2019)).Importantly, a much higher proportion of unmated individuals might occur in species where mating is monopolized by a few dominant individuals or where breeding sites are limited.For example, in many bird species, newly mature males are not sexually competitive and do not breed in their first year (Both et al., 2017).This means that, without good data from natural populations, researchers should be cautious about claiming ecological realism.

Figure 2 .
Figure 2. Histogram showing the mating history of males and females used in the mate choice experiments collected by Dougherty (2021b, 2023a).

Table 1
Studies included in the meta-analysis