Olfactory camouflage and communication in birds

Smell is a sensory modality that is rarely considered in birds, but evidence is mounting that olfaction is an important aspect of avian behaviour and ecology. The uropygial gland produces an odoriferous secretion (preen oil) that can differ seasonally and between the sexes. These differences are hypothesized to function in olfactory camouflage, i.e. minimizing detection by nest predators (olfactory crypsis hypothesis), and/or intraspecific olfactory communication, particularly during breeding (sex semiochemical hypothesis). However, evidence for seasonal and sex differences in preen oil is mixed, with some studies finding differences and others not, and direct evidence for the putative function(s) of seasonal variation and sex differences in preen oil remains limited. We conducted a systematic review of the evidence for such changes in preen oil chemical composition, finding seasonal differences in 95% of species (57/60 species in 35 studies) and sex differences in 47% of species (28/59 species in 46 studies). We then conducted phylogenetic comparative analyses using data from 59 bird species to evaluate evidence for both the olfactory crypsis and sex semiochemical hypotheses. Seasonal differences were more likely in the incubating than non‐incubating sex in ground‐nesting species, but were equally likely regardless of incubation strategy in non‐ground‐nesting species. This result supports the olfactory crypsis hypothesis, if ground nesters are more vulnerable to olfactorily searching predators than non‐ground nesters. Sex differences were more likely in species with uniparental than biparental incubation and during breeding than non‐breeding, consistent with both the olfactory crypsis and sex semiochemical hypotheses. At present, the data do not allow us to disentangle these two hypotheses, but we provide recommendations that will enable researchers to do so.


I INTRODUCTION
All animals produce odours, either as metabolic by-products or as chemicals secreted by specialised glands. These odours can provide information about the producer that can be used during interspecific interactions (e.g. to detect the presence of potential predators or prey) or during intraspecific interactions (e.g. to assess the age, sex, relatedness, or genetic compatibility of a potential mate). In birds, body odours can derive from various sources, including faeces, blood, stomach oils, powder down, plumage, and from secretions of the anal gland, salt gland, salivary gland, ear glands, sebokeratocytes, or skin (Hagelin & Jones, 2007). Recently, much attention has focused on the odour-producing role of the uropygial or preen gland (Moreno-Rueda, 2017;Whittaker & Hagelin, 2020). The preen gland, located near the base of the tail, is present in almost all bird species (Johnston, 1988;Moreno-Rueda, 2017). The gland secretes preen oil, a complex mixture of wax esters (monoesters, diesters, and triesters) and other compounds (e.g. alcohols, alkanes, aldehydes, carboxylic acids, ketones; reviewed in Campagna et al., 2012). Early work on preen oil was primarily descriptive, but there has been a remarkable growth in preen oil research, particularly with respect to its putative functions (reviewed in Moreno-Rueda, 2017; summarized in Fig. 1).
Over the past 20 years, researchers have begun to explore preen oil from the perspectives of ecotoxicology [effects of environmental pollutants on preen oil composition, a role for preen oil in pollutant depuration (L opez-Perea & Mateo, 2019;Grieves et al., 2020)]; chemical defence [antimicrobial/antiparasitic activity, predator repellence, olfactory crypsis (Burger et al., 2004;Reneerkens et al., 2007a;Martín-Vivaldi et al., 2010)]; vector attraction (preen oil as an attractant to parasite vectors such as mosquitoes; reviewed in Martínez-de la Puente et al., 2020); species recognition and speciation [testing for chemical signatures of preen oil useful for taxonomic classification (Zhang, Du & Zhang, 2013;Gabirot et al., 2016)]; and intraspecific communication [reproductive and social signalling (reviewed in Caro, Balthazart & Bonadonna, 2015;Whittaker & Hagelin, 2020)]. Researchers have also continued to study the mechanisms underlying preen oil production and chemistry [e.g. diet, endocrine regulation, symbiotic microbes (Thomas et al., 2010;Whelan et al., 2010;Whittaker et al., 2019b)]. Despite this growth in research, the mechanisms of preen oil production and variationas well as the  putative functions of preen oilare still poorly understood across all research areas. Thus, there is ample opportunity for researchers to make novel and valuable contributions to our understanding of preen oil production and its function in birds.
Some of the functions of preen oil, including waterproofing, feather maintenance, and pollutant depuration, depend on its physical (i.e. oily, waxy) structure. In addition to these structural functions, preen oil is also odoriferous and considered to be a major source of avian body odour (Hagelin & Jones, 2007;Caro et al., 2015). Accordingly, preen oil has been hypothesized also to act as an infochemical (Müller et al., 2020) during intraspecific interactions (reviewed in Moreno-Rueda, 2017), or as a deleterious cue that reduces detection by predators, such that downregulation of its production, or volatility, would be indicative of olfactory crypsis.
Crypsis is the avoidance of detection through camouflage (Stevens & Merilaita, 2009). While most studies of crypsis involve vision, crypsis can also involve olfactory concealment (Ruxton, 2009). Birds in a nest can emit odours at all life stages (as eggs, chicks, and adults) and may be vulnerable to olfactorily searching nest predators such as mammals as a result. Birds should therefore benefit from olfactory crypsis at the nest (Shutler, 2019), especially since nest predation is a primary cause of reproductive failure (Martin, 1993). As such, birds might alter their odours to become less detectable to predators, especially during the critical period of nesting. By contrast, the use of sex semiochemicals for intraspecific chemical communication during breeding suggests that individuals might alter their odours to convey information to and/or modulate their detectability by conspecifics.
We systematically reviewed the literature on seasonal and sex differences in preen oil composition to investigate two non-mutually exclusive hypotheses. First, the 'olfactory crypsis hypothesis' posits that incubating birds switch from more odorous to less odorous preen oil during incubation as a means of reducing odour cues at the nest, thereby protecting eggs and young from olfactorily searching predators (Reneerkens, Piersma & Damsté, 2002;Reneerkens et al., 2007a). Because less-odorous (higher molecular weight) preen oil is presumably more costly to produce, and perhaps also to apply (Reneerkens et al., 2007b), it is predicted to be secreted only during incubation, when the benefits of crypsis outweigh the costs of production (Reneerkens, Piersma & Damsté, 2006). This hypothesis predicts an effect of both breeding stage and incubation type on the chemical composition of preen oil. Preen oil changes should occur specifically during incubation and only in the incubating sex, leading to sex differences in uniparentally incubating, but not biparentally incubating, species. Changes in preen oil composition specifically associated with incubation should have evolved primarily in species under strong selective pressure from olfactorily searching nest predators (Reneerkens et al., 2006). Notably, this hypothesis assumes that nest predators should be better at detecting low molecular weight than high molecular weight preen oil (Reneerkens, Piersma & Damsté, 2005).
Next, we introduce the 'sex semiochemical hypothesis', which posits that sex differences in preen oil are associated with mate recognition (identifying the appropriate sex to mate with) and/or mate choice (identifying a suitable, e.g. genetically compatible, mate). The sex semiochemical hypothesis predicts that sex differences in the chemical composition of preen oil should be found only during breeding (particularly during mate pairing and egg laying), and that birds should use preen oil odour cues to discriminate between the sexes and/or among individuals. We expand on these two hypotheses further in Fig. 2.
The olfactory crypsis and sex semiochemical hypotheses are based on the odoriferous nature of preen oil. However, preen oil may also serve as a chemical defence against a range of parasites, including eggshell bacteria, feather-degrading bacteria, chewing lice, and mosquitoes (reviewed in Moreno-Rueda, 2017), and such antiparasitic defence does not require preen oil to be odoriferous (though chemical defences can indeed be odorous). Thus, the antiparasitic defence hypothesis is also non-mutually exclusive with the olfactory crypsis and sex semiochemical hypotheses. Due to a paucity of data, we were not able to conduct a comparative analysis to test for general support of this hypothesis, and therefore focused our analyses on the odour-based hypotheses.
Under the olfactory crypsis hypothesis, we predicted that, in uniparentally incubating species, only the incubating sex would show a shift in preen oil composition while in biparentally incubating species, both sexes would show shifts; thus, we expected that seasonal differences in preen oil chemical composition would be more common in the incubating sex. We also expected to find seasonal differences more commonly in species with nests more vulnerable to olfactorily searching predators (i.e. nests that are located on or near the ground compared to nests placed at height or in remote, inaccessible locations such as on cliffs). Similarly, we also predicted that sex differences in preen oil would be more likely in species with uniparental than biparental incubation. Under the sex semiochemical hypothesis, we predicted that sex differences in the chemical composition of preen oil would be more likely during breeding than non-breeding. To test these predictions, we conducted a comparative analysis of the available literature that tested for seasonal and sex differences in the preen oil of all bird species for which data were available.

II METHODS
(1) Literature review We systematically reviewed studies that tested for an effect of season and/or sex on the chemical composition of preen oil. We screened the abstracts of 187 publications and the full text of 66 publications, retaining 55 publications (35 on seasonal differences and 46 on sex differences, including 26 publications addressing both seasonal and sex differences) that corresponded to our inclusion criteria. Details of the systematic review and the data used for analysis are available as online Supporting Information (Appendix S1, S2, Fig. S1, Tables S1 and S2).
(2) Preen oil chemical differences Various analytical and statistical methods have been used to evaluate chemical differences in preen oil composition (Table S1). Given the diversity of methodologies used, if a significant chemical difference was observed at α = 0.05, we recorded it as such. Thus, we created binary response variables of 'sex difference' and 'seasonal difference' (yes/no). (3) Seasonal differences We tested whether sex-specific seasonal changes are related to incubation and nest ecology. We obtained data on seasonal differences for 91 occurrences, defined as data on a given sex for a given species. For each occurrence, we recorded whether the sex exhibited a significant (α = 0.05) seasonal change in preen oil composition (yes/no), whether the sex incubates (yes/no), whether the species nests on the ground (ground/non-ground; details below), and the timescale of the study (within breeding season/across breeding and non-breeding seasons; details below; Table S4). Thus, a species could be included multiple times in our analysis if it was included in multiple studies. Information about incubation and nest ecology was obtained from the Handbook of the Birds of the World (del Hoyo, Elliott & Christie, 2009). In some species, only one parent incubates, but the incubating parent can be of either sex (e.g. western sandpiper, Calidris mauri). Because a mix of both sexes would be incubating in any given study population for such species, we categorized these species as biparentally incubating. For studies on captive birds, we inspected the methods to confirm that seasonality was established using appropriate methods (e.g. by using artificial light cycles for birds kept indoors).
Seasonal changes can occur at different timescales (within the breeding season, within the non-breeding season, and across the breeding and non-breeding seasons). To interpret any biological functions of preen oil changes, it is necessary to consider the timescale of the changes. We categorized timescale as 'within breeding season' (spanning nest building, egg laying, incubation, and brood care), and 'across breeding and non-breeding seasons' (where nonbreeding encompasses fledging through winter, up to the start of nest building the following year). Studies conducted within the breeding season either compared samples from different periods within the breeding season (e.g. across mating, incubation, and brood care) or measured the effect of date on preen oil composition. Studies conducted across the breeding and non-breeding seasons either compared samples from the breeding and non-breeding season, or compared samples collected regularly throughout the year (e.g. monthly).
In total, our data set on seasonal differences comprised 91 occurrences (where one occurrence corresponds to one sex) from 43 species and 25 studies (Table S4). Effect sizes (Cohen's d) could be calculated for only three studies [using an online calculator (Lenhard & Lenhard, 2016), Table S4] and were therefore not used for analysis.
(4) Sex differences We tested whether sex differences in the chemical composition of preen oil are related to season and incubation type. For each species, we recorded whether a significant (α = 0.05) sex difference was detected in the composition of preen oil (yes/no), the season in which preen oil was sampled (breeding/non-breeding; where breeding includes nest building, egg laying, incubation, and brood care, and nonbreeding encompasses fledging through winter, up to the start of nest building the following year), and the incubation type (uniparental/biparental; Table S5). Analysing sex differences during specific breeding periods (e.g. mate choice, incubation, chick rearing) would be more informative than distinguishing only breeding and non-breeding, but most studies sampled birds across multiple breeding stages, and we therefore could not conduct such an analysis. Also, in most cases, the nature and direction of sex differences were not explicitly recorded, so we could not include this information in our analyses. For studies on free-living birds, breeding stage dates and incubation type were verified using the Handbook of the Birds of the World (del Hoyo et al., 2009). For studies on captive birds, we inspected the methods to confirm that birds were brought into breeding condition using appropriate methods (e.g. using natural light cycles for birds in outdoor aviaries or by using artificial light to photostimulate birds kept indoors).
In total, our data set on sex differences comprised 75 occurrences (where one occurrence corresponds to one season) from 49 species and 39 studies (Table S5). As with seasonal differences, because effect sizes could be calculated for only a limited number of studies (21 , Table S5), we did not use effect sizes in our analysis.

(5) Statistical analyses
Our full data set included 59 species and 45 studies. We conducted comparative analyses for each model (seasonal differences, sex differences) using generalized linear mixed models (GLMMs) with Markov chain Monte Carlo techniques under a Bayesian statistical framework, using the package MCMCglmm (Hadfield, 2010) in R (R Development Core Team, 2017) that allowed us to control for phylogenetic dependency. The first model (seasonal differences) was run for 13 × 10 6 iterations, with a burn-in phase of 10000 and a thinning interval of 3500, resulting in a sample size of 3712. The second model (sex differences) was run for 10 × 10 6 iterations, with a burn-in phase of 5000 and a thinning interval of 2000, resulting in a sample size of 4998. These parameters were chosen to ensure model convergence  (Hadfield, 2010). Because we had no a priori predictions about the values of these parameters, both models were fit using a weakly informative inverse-gamma prior (Hadfield, 2010). We verified the absence of autocorrelation, verified convergence with the Gelman-Rubin diagnostic (Gelman & Rubin, 1992), and assessed the significance of fixed effects (at α = 0.05) by checking whether their 95% credible interval spanned 0.
Our first model included seasonal difference as a binary response variable (yes/no) and the fixed effects incubation (sex incubates/sex does not incubate), nest ecology (ground/ non-ground nesting), timescale (within breeding season/across breeding and non-breeding seasons), and the interaction term incubation × nest ecology. Our second model included sex difference as a binary response variable (yes/no) and the fixed effects season (breeding/non-breeding) and incubation type (uniparental/biparental).
For both models, we included species as a random effect because some species were used in multiple studies, and because some species were tested at two times of year (sex differences) or in both sexes (seasonal differences). We included phylogeny as a random effect to control for potential effects of phylogenetic relatedness. We calculated the phylogenetic relatedness between species using the consensus tree of 1000 phylogenetic trees (Stage2 MayrAll Hackett backbone) generated on birdtree.org (Jetz et al., 2012). Finally, we verified that the inclusion of random effects improved the fit of the models, indicated by a lower deviance information Fig. 3. Distribution of species studied with respect to seasonal (blue) and sex (orange) differences in preen oil chemical composition in birds. Orders highlighted in purple were studied with respect to both seasonal and sex differences. No order was studied with respect to seasonal differences only. Phylogeny is based on Hackett et al. (2008); gulls (family Laridae) and sandpipers (family Scolopacidae) belong to the order Charadriiformes. Illustrations created by M.G. using Microsoft PowerPoint. criterion (DIC) score. These analyses are detailed in the Supporting information (Appendices S3 and S4). Detailed sample sizes used in each analysis are available in Table S6.

III RESULTS
(1) Literature review Of the 55 studies included in our systematic review, 35 investigated seasonal differences (60 species) and 46 investigated sex differences (59 species) in preen oil composition, with 26 of these papers investigating both seasonal and sex differences. While 76 species have been investigated, most studies (61) involved just two phylogenetic orders, Passeriformes (songbirds, 32 species) and Charadriiformes (gulls and shorebirds, 29 species; Fig. 3). Seasonal differences were found in 95% (57/60) of species studied and sex differences were detected in 47% (28/59) of species studied.
(2) Seasonal differences The probability of detecting a seasonal change in preen oil composition was related to the interaction between incubation and nest ecology (posterior mean = 287.64, 95% CI = [66.39, 543.24], P MCMC = 0.01; Table 1, Fig. 4). To elucidate the direction of the interaction, we performed separate analyses for ground-nesting species (45 occurrences) and non-groundnesting species (46 occurrences). For ground-nesting species, seasonal differences were more likely in the incubating than the non-incubating sex (posterior mean = 286.66; 95% CI = [98.27, 494.14], P MCMC < 0.001), whereas for non-ground-nesting species, seasonal differences were apparent regardless of which sex incubated (posterior mean = 51.73, 95% CI = [−66.78, 182.74], P MCMC = 0.29). Timescale had no effect on the probability of detecting seasonal changes (Table 1). Accounting for phylogeny and species increased the fit of the models slightly but had little effect overall (Table S7). Phylogeny and species explained 7 and 5% of the total variance, respectively (Table S8).
(3) Sex differences The probability of detecting sex differences in preen oil composition was related to both breeding stage and incubation type (  Fig. 5). Accounting for phylogeny and species increased the fit of the models slightly but had little effect on the overall model results (Table S7). Phylogeny and species explained 9 and 5% of the total variance respectively (Table S8).

IV DISCUSSION
This study reviewed and analysed the literature on olfactory crypsis and sex semiochemicals and found support for both hypotheses. Seasonal changes in the chemical composition of preen oil were nearly ubiquitous. Consistent with predictions derived from the olfactory crypsis hypothesis, the Table 1. Summary of phylogenetically controlled Markov chain Monte Carlo generalized linear mixed effects models to investigate factors affecting seasonal and sex differences in preen oil chemical composition. The model on seasonal differences (91 occurrences) tests whether the occurrence of seasonal differences (no = 0; yes = 1) depends on incubation (sex does not incubate = 0; sex incubates = 1), nest ecology (non-ground-nesting = 0; ground-nesting = 1), the timescale of the study (within breeding season = 0; across breeding and non-breeding season = 1), and the interaction between incubation and nest ecology. The model on sex differences (75 occurrences) tests whether the occurrence of sex differences (no = 0; yes = 1) depends on the species' incubation type (uniparental = 0; biparental = 1) and the season (non-breeding = 0; breeding = 1) likelihood of detecting a seasonal change in preen oil composition was related to the interaction between incubation and nest ecology such that seasonal differences were more likely in the incubating sex, but only in ground-nesting species. For non-ground-nesting species, seasonal changes were equally likely, regardless of which sex incubated. By contrast, sex differences were less ubiquitous than seasonal differences, occurring in less than half of the species studied. Consistent with predictions of both the sex semiochemical and olfactory crypsis hypotheses, the likelihood of detecting sex differences in preen oil composition was related to both breeding stage and incubation type. Specifically, sex differences were more likely during breeding than non-breeding, and in species with uniparental than biparental incubation. It should be noted that our results on the probabilities of seasonal and sex differences may be overestimates if there is publication bias in favour of significant results. On the other hand, the probabilities of seasonal and sex differences may also be underestimated, since some studies were not designed specifically to test for such differences (e.g. in cases where studies sampled across breeding and/or non-breeding stages, and/or had small sample sizes), and as a result could not or did not detect any differences in preen oil composition. With a more appropriate design, such studies may have detected seasonal and/or sex differences in preen oil composition.
At present, there are insufficient data to disentangle these non-mutually exclusive hypotheses. Thus, our work is not the definitive test of these two hypotheses, but it is the best we can achieve to date. Below, we review current support for the olfactory crypsis and sex semiochemical hypotheses and offer recommendations for more direct hypothesis testing.
(1) Olfactory crypsis Evidence for a role of preen oil in olfactory crypsis is currently limited. Studies on the preen oil composition of 27 ground-nesting shorebird species (order Charadriiformes) revealed a seasonal shift from monoesters to diesters at the onset of breeding (Piersma, Dekker & Sinninghe Damsté, 1999;Reneerkens et al., 2002Reneerkens et al., , 2006Reneerkens et al., , 2007a, with diester secretion being maintained during incubation and chick-rearing (Reneerkens et al., 2002(Reneerkens et al., , 2006. Remarkably, diesters were secreted equally in both sexes in species where both sexes incubate, only in males in species where only males incubate, and mainly in females in species where only females  incubate (Reneerkens et al., 2007a). Because diesters are less volatile than monoesters, these authors hypothesized that seasonal changes in the preen oil of incubating birds enhance olfactory crypsis by reducing olfactory cues at the nest, thereby limiting detection by olfactorily searching nest predators (Reneerkens et al., 2002(Reneerkens et al., , 2007a. In support of this hypothesis, a trained dog was better at detecting preen oil composed of monoesters than diesters (Reneerkens et al., 2005).
Other studies have also interpreted seasonal and sex differences in preen oil composition using the olfactory crypsis hypothesis. For example, the preen oil of Eurasian moorhens (Gallinula chloropus) is less volatile during breeding than nonbreeding, and olfactory crypsis was proposed as an explanation (L opez-Perea & Mateo, 2019). In mallards (Anas platyrhynchos), a shift from monoesters to diesters at the onset of breeding was observed in females but not in males (Jacob et al., 1979;Bohnet et al., 1991). The shift was first thought to be involved in mate choice, by providing an olfactory cue that males might use to identify females in breeding condition (see Section IV.2). But given that diesters are less volatile than monoesters, that only females incubate in this species, and that mallard nests are exposed to mammalian predators (Johnson, Sargeant & Greenwood, 1989), this shift may be more relevant for olfactory crypsis than intraspecific communication and mate choice.
Finally, a New Zealand study on 13 non-ground-nesting passerine species compared the preen oil of introduced species that co-evolved with mammalian predators to that of native species that have a long evolutionary history without mammalian predators. Consistent with the olfactory crypsis hypothesis, preen oil was less volatile during breeding than non-breeding in introduced but not in native species (Fluen, 2008). However, sample sizes were low and preen oil volatiles were not lower in females for species with female-only incubation (Fluen, 2008), as would be predicted by the olfactory crypsis hypothesis. Overall, most studies speculate on the role of seasonal and sex variation in preen oil in maintaining olfactory crypsis without providing evidence. Additional studies on the ability of predators to detect preen oil secreted during various breeding and life-cycle stages are thus warranted.
Based on the olfactory crypsis hypothesis, we predicted that seasonal differences in preen oil composition would be more common in the incubating than non-incubating sex(es), especially in species with nests that are more vulnerable to olfactorily searching predators, such as ground nesters. We also predicted that sex differences in preen oil composition would be more common during breeding than non-breeding, and in uniparentally incubating than biparentally incubating species. We found support for all three predictions. Consistent with our first prediction, seasonal differences were more frequently detected in sexes that incubate than in sexes that do not incubate, but only in ground-nesting species (i.e. species more likely exposed to olfactorily searching predators). This suggests that preen oil changes are indeed associated with incubation in species that are under stronger selection pressure from olfactorily searching predators, supporting the olfactory crypsis hypothesis. However, our findings also highlight that olfactory crypsis cannot explain our findings for non-groundnesting species. Birds exhibited seasonal changes in preen oil regardless of which sex(es) incubated, suggesting that there are also other explanations for seasonal changessuch as intraspecific chemical communication, as predicted under the sex semiochemical hypothesis.
Consistent with our second and third predictions, sex differences in preen oil composition were more common during breeding and in species with uniparental incubation. This is consistent with both the olfactory crypsis and sex semiochemical hypotheses. Additional information about the nature of sex differences could allow us to disentangle these two hypotheses. For example, less-volatile preen oil during breeding compared to non-breeding could corroborate a cryptic function, while more volatile preen oil during breeding compared to non-breeding could corroborate a signalling function.
We used a comparative analysis to re-evaluate Reneerkens' olfactory crypsis hypothesis, which predicts that seasonal changes in the preen oil of incubating birds is primarily due to mammalian predation (Reneerkens et al., 2002(Reneerkens et al., , 2007a. Based on our results, we propose expanding the definition of olfactory crypsis to consider other biologically relevant factors. First, nest predation can be as high during the nestling stage as during the incubation stage (Pietz & Granfors, 2000; although we note this is not always the case), so olfactory crypsis could be important during both breeding stages. This is consistent with diesters being secreted until the end of the chick-rearing period in shorebirds (Reneerkens et al., 2006). Second, preen oil could reduce the detectability of nests in two main ways: preen oil could enhance the crypsis of brooding (i.e. incubating and chick rearing) parents, thereby masking nest odours while adults are on the nest, but preen oil could also enhance crypsis of the eggs and chicks directly, if preen oil is transferred from parents to the offspring. Evidence that preen oil is transferred directly from parents to eggs and chicks is limited (but see Soler et al., 2014), so the mechanisms and efficacy of preen oil transfer at nests are worth exploring further. Third, olfactory crypsis may be applicable not only to non-volatile (e.g. diesters) but also to volatile compounds. For example, certain volatile compounds could enhance crypsis if they blend in with the olfactory background of the nest (i.e. background matching; Soini et al., 2007). Comparing preen oil compounds secreted during incubation and brooding with background odours of the nest and surrounding environment could help determine whether olfactory background matching is occurring. Finally, the olfactory crypsis hypothesis could apply not just to mammals, but to any olfactorily searching nest predators, regardless of taxon (e.g. insects, snakes, and even birds; Shutler, 2019).
In this study, we estimated the vulnerability to olfactorily searching nest predators by describing the nest ecology of the species and distinguishing ground-nesting and nonground-nesting species. We assumed that ground nests are more vulnerable to olfactorily searching predators, because ground nests are more commonly depredated by mammals  (Söderström et al., 1998;Zuria et al., 2007; but see Angelstam, 1986;Mallord et al., 2012) and because mammalian nest predation usually occurs at night and is mostly olfactorily based (Whelan et al., 2010;Cox, Thompson & Reidy, 2013). However, this assumption is simplistic. Both ground nests and non-ground nests are susceptible to predation by three taxa in particular: mammals, snakes, and birds (and, to a lesser extent, insects; Thompson, 2007). The prevalence of nest predation by each taxon may depend more on habitat characteristics than nest ecology (Martin, 1995;Thompson, 2007;Reidy & Thompson, 2012). Moreover, each predator taxon can use multiple cues to detect nests. Mammals and snakes depredate nests mostly at night (e.g. Cox et al., 2013;DeGregorio et al., 2014) using olfactory cues (Ford & Burghardt, 1993;Whelan et al., 2010) but may also use other cues [e.g. visual cues (Mullin & Cooper, 1998;Stake et al., 2005;Dawson et al., 2014)]. Birds commonly depredate nests during the day (Reidy & Thompson, 2012) and rely mainly on visual cues, but may also use olfactory cues (e.g. Buitron & Nuechterlein, 1985;Molina-Morales et al., 2020). A more accurate proxy of vulnerability to olfactorily searching nest predators would be, for example, the incidence of such predators weighted by the likelihood of nest detection by olfaction, but such a measure was impossible to obtain for the species and populations included in our analyses. Although simplistic, we consider that nest ecology is a reasonable proxy of vulnerability to olfactorily searching predators in the absence of site-specific information on predation dynamics. Furthermore, if olfactory crypsis prevents some (although not all) predator detections, it could still be sufficiently beneficial to have evolved. Overall, although evidence from experimental studies with natural predators and from studies on taxa other than shorebirds are still lacking, our results and literature review provide compelling support for a role of preen oil in olfactory crypsis.
(2) Sex semiochemicals Preen oil chemical cues are increasingly thought to play a role in avian mate choice and reproduction (Balthazart & Taziaux, 2009;Caro & Balthazart, 2010;Caro et al., 2015;Whittaker & Hagelin, 2020). Reproductive signals or cues should differ between the sexes and reflect aspects of quality or condition (Johansson & Jones, 2007), and there is growing evidence that preen oil provides odour cues of sex that at least some bird species respond to. Thus, we proposed the sex semiochemical hypothesis, positing that sex differences in preen oil are associated with reproduction and preen oil odour cues are involved in mate recognition and/or mate choice. The sex semiochemical hypothesis predicts that there should be an effect of breeding stage (breeding versus nonbreeding) on preen oil. Indeed, sex differences were more common in breeding birds, suggesting a role for preen oil in reproductive chemical signalling.
The preen oil of several passerine species becomes more volatile during the breeding season [e.g. white-throated sparrows, Zonotrichia albicollis (Tuttle et al., 2014), gray catbirds, Dumetella carolinensis (Shaw et al., 2011), dark-eyed juncos, Junco hyemalis (Soini et al., 2007)], and birds may use these preen oil odour cues to attract mates and compete with same-sex conspecifics (Whittaker & Hagelin, 2020). Such findings argue against the chemical crypsis hypothesis, at least for some species. An increased volatility of preen oil chemical cues could serve to advertise for mates and/or to compete with samesex conspecifics (e.g. via territorial scent marking), and such signals might reinforce or enhance other indicators of sex, breeding status, or dominance, such as song characteristics, plumage traits, and other sexually selected ornaments. However, sex differences in preen oil are often, but not always, associated with a greater abundance and/or diversity of chemical compounds in the preen oil of females (Whittaker & Hagelin, 2020), who often display fewer sexually selected ornaments than males. This apparent female emphasis on chemical differences in preen oil may be driven by three main factors: intersexual advertisement (e.g. of female receptivity and/or quality) and physiological priming effects on males; intrasexual competition (e.g. territorial scent marking, dominance, and reproductive suppression); and maternal behaviours (e.g. maternal care, mother-offspring recognition, chemical protection of eggs and nestlings) (Whittaker & Hagelin, 2020). Additional experiments testing for evidence of a role for preen oil in intersexual advertisement, intrasexual competition, and parental behaviours are warranted.
Shifts in the preen oil composition of breeding birds may also act as indicators of quality (Whittaker & Hagelin, 2020). In dark-eyed juncos, females with more 'female-like' odour and males with more 'male-like' odour both produce more offspring (Whittaker et al., 2013). Further, males with more 'male-like' odour have more surviving nestlings (regardless of nestling paternity) while males with more 'female-like' odour have more extrapair young in their home nest (Whittaker et al., 2013). In the lance-tailed manakin (Chiroxiphia lanceolata), the likelihood that offspring survive to fledging increases with male microsatellite heterozygosity (a proxy for genome-wide heterozygosity), and this almost certainly reflects genetic quality, because male manakins do not provide parental care (Sardell, Kempenaers & Duval, 2014). Furthermore, some preen oil components are correlated with increased heterozygosity in males, suggesting that females could use preen oil odour cues to evaluate male heterozygosity (Whittaker et al., 2019a).
In species where it has been investigated, the chemical composition of preen oil is associated with MHC genotype, part of the adaptive immune system, such that individuals with more similar preen oil composition are more similar at MHC [e.g. in black-legged kittiwake, Rissa tridactyla  and song sparrows, Melospiza melodia (Slade et al., 2016;Grieves et al., 2019c)]. This suggests that preen oil may provide cues of relatedness and/or genetic compatibility. Notably, such cues are detectable to at least some bird species [blue petrels, Hydrobates caerulea (Leclaire et al., 2017) and song sparrows (Grieves et al., 2019c)]. To understand better the role of preen oil chemical cues in avian reproduction, more information is needed on which sexes Biological Reviews 97 (2022)  Most of the studies included in our comparative analysis did not test birds' ability to discriminate between the sexes, but evidence for sex discrimination was found in all six of the studies that did (Zhang et al., 2010;Whittaker et al., 2011a;Amo et al., 2012;Mihailova, 2014;Grieves et al., 2019a;Van Huynh & Rice, 2019). In breedingcondition Passeriformes, both sexes spent more time with male odour in dark-eyed juncos (Whittaker et al., 2011a) and spotless starlings (Sturnus unicolor; Amo et al., 2012). By contrast, both sexes spent more time with opposite sex odour in black-capped chickadees (Poecile atricapillus), Carolina chickadees (Poecile carolinensis; Van Huynh & Rice, 2019), and song sparrows (Grieves et al., 2019a). In Psittaciformes, female budgerigars (Melopsittacus undulatus) spent more time with male odour (Zhang et al., 2010), and female crimson rosellas (Platycercus elegans) spent more time on nest boxes treated with male odour than female odour (Mihailova, 2014), suggesting a preference for these odour types.
Evidence for sex discrimination was also found in studies that were not included in our analysis (because sex differences in preen oil composition were not measured). In Galliformes, male domestic chickens (Gallus gallus) more frequently mount and copulate with females that have an intact preen gland, but this preference is abolished in anosmic males (Hirao et al., 2009). In Charadriiformes, crested auklets (Aethia cristatella) of both sexes approached model birds treated with male odour more closely than they approached models treated with female odour; this study used a synthetic odour mimicking two major components of auklet odour (Jones et al., 2004). Importantly, these sex-discrimination tests were all performed on birds in breeding condition. Taken together, these results suggest that the ability to use odour cues to discriminate conspecific sex is widespread in birds.
(3) Mechanisms of seasonal and sex differences Seasonal and sex differences in preen oil composition may be related to changes in diet, preen gland microbes, and circulating hormone levels. Such shifts may provide both protection from predators (via olfactory crypsis) and indirect cues of reproductive status (i.e. readiness to breed) that play a role in both intersexual signalling and intrasexual competition (via sex semiochemicals; Whittaker & Hagelin, 2020).

(a) Diet
Many avian species change their diet at the onset of the breeding season (Bairlein & Gwinner, 1994). As such, seasonal differences in preen oil composition may also be affected by changes in diet. To our knowledge, no studies have tested whether natural seasonal dietary changes affect preen oil composition, but laboratory studies have shown that diet affects preen oil composition in captive birds (Apandi & Edwards, 1964;Thomas et al., 2010, p. 201;Kanakri et al., 2016). However, captive birds fed a constant diet still exhibit seasonal changes in preen oil (Reneerkens et al., 2007b;Whelan et al., 2010;Tuttle et al., 2014;Potier et al., 2018;Grieves et al., 2020), demonstrating that differences in diet can only partly explain seasonal changes in preen oil composition. Sex differences in the chemical composition of preen oil may be driven partly by sex differences in diet, which is common in species with size dimorphism [e.g. seabirds (Phillips et al., 2011), raptors (Catry et al., 2016)] or with spatial segregation during foraging (e.g. shorebirds; Catry et al., 2012).

(b) Symbiotic microbes
Preen gland microbes can also influence the chemical composition of preen oil , 2010Whittaker et al., 2019b). Gland microbiota can differ seasonally, which may be associated with seasonal changes in bacterial loads (e.g. an increase during the breeding season; Rodríguez-Ruano et al., 2018) that can then affect preen oil composition. Preen gland microbiota can also differ between the sexes (Pearce et al., 2017;Rodríguez-Ruano et al., 2018, but see Whittaker et al., 2019bGrieves et al., 2021). Given that nests can harbour unique microbial communities (Jacob et al., 2014;van Veelen, Salles & Tieleman, 2017), sex differences in symbiotic microbes may be driven by sex differences in time spent at the nest (Saag et al., 2011;Goodenough et al., 2017). Seasonal changes in diet could also contribute to changes in preen gland microbes, but to our knowledge this has only been explored in avian gut microbiota (Grond et al., 2018).

(c) Hormones
Seasonal and sex differences in the chemical composition of preen oil may be driven at least partly by endogenous changes in circulating levels of sex steroid hormones such as oestradiol and testosterone. Oestradiol injections trigger a shift from monoesters to diesters in mallard preen oil (Bohnet et al., 1991). Testosterone implants have variable effects on preen oil composition across species, triggering increases in some compounds (Abalain et al., 1984;Whittaker et al., 2011b) and decreases in others (Whelan et al., 2010). Thus, seasonal and sex differences in preen oil are likely at least partly related to physiological changes associated with reproduction.

V RECOMMENDATIONS FOR FUTURE RESEARCH
(1) Sampling and study design Based on our comparative analysis, we found support for both the olfactory crypsis and sex semiochemical hypotheses. In most cases, the studies we reviewed do not consider the nature of seasonal and sex differences; that is, information on which sex(es) exhibited changes, and details on which chemicals changed (and how they changed) are rarely reported. Seasonal changes in preen oil composition have been detected at fine timescales [e.g. less than a week in red knots, Calidris canutus (Reneerkens et al., 2007b), less than 2 weeks in dark-eyed juncos and song sparrows (Whittaker et al., 2011b;Grieves et al., 2018)], and preen oil appears to be more subject to these finer scale seasonal changes during the breeding than the non-breeding season (e.g. Reneerkens et al., 2002). Preen oil composition may thus be more stable (i.e. less variable) during the non-breeding season. That said, if there is selection on a specific mix of preen oil compounds during breeding, one might predict that preen oil should be less, not more, variable during the breeding season, or at least during specific stages of breeding; these contrasting possibilities are worth further study. Interestingly, the speed of seasonal changes in preen oil can be altered in captivity, as shown in red knots, where the shift to diesters was two times slower in captive (4 weeks) than wild individuals (2 weeks; Reneerkens et al., 2007b). Our comparative analysis revealed that seasonal changes in preen oil were detected independently of the timescale of the study (within breeding or across the breeding and non-breeding seasons). Seasonal changes may have occurred before incubation, supporting olfactory crypsis, but may also have occurred at other times. Sampling preen oil at regular intervals across breeding and nonbreeding stages would clarify the timescale over which preen oil changes. In addition, knowing at which specific period(s) (e.g. mate choice, incubation, chick rearing) seasonal changes occur will allow more specific, testable predictions about the function of changes in preen oil composition to be made.
Based on our findings that sex differences depend on season, and seasonal differences were nearly ubiquitous, these factors should be carefully considered in sampling design and analysis. If sex differences are of interest, sampling should be conducted during the breeding season, and the breeding stages (e.g. pair formation, egg laying, incubation, brood care) during which sampling occurs should be recorded. If sex differences are not of interest, it may be ideal to sample during the non-breeding season when sex differences may be less likely to be observed, as this may reduce potential confounds. The conditions, date(s), and duration of sampling should always be taken into consideration.
(2) Hypothesis testing The olfactory crypsis and sex semiochemical hypotheses are not mutually exclusive. Studies that examine changes in preen oil over a finer timescale (e.g. comparing different periods within the breeding season, such as pre-breeding, incubation, and chick guarding; Reneerkens et al., 2002), quantify hatching and fledging success (Whittaker et al., 2013), and measure additional variables of interest (e.g. measures of quality; Whittaker et al., 2019a) should provide important insights into the functions of avian preen oil in crypsis and social signalling.
The existing literature concerns mainly two bird orders (shorebirds and passerines). Most studies supporting the olfactory crypsis hypothesis have been conducted on shorebirds (but see Fluen, 2008), while studies supporting the sex semiochemical hypothesis have been conducted on predominantly passerines and shorebirds (Table S4). This could be due to a taxonomic bias, as evidence of olfactory crypsis was first collected in shorebirds (Reneerkens et al., 2002), and evidence for sex semiochemicals is rapidly accumulating in passerines (Whittaker & Hagelin, 2020). Alternatively, this could be because seasonal and/or sexual variation in preen oil chemical composition has different functions in each taxon, or that such differences depend on the environment or ecology of a given species. These hypotheses could be tested by analysing variation in preen oil composition in ground-breeding passerines and shorebirds that co-occur in the same habitat, taking care to collect and analyse preen oil separately from the pre-breeding stage (mate choice, pairing, nest building) and during incubation and brood care.

(a) Olfactory crypsis hypothesis
Here, we outline specific predictions that should be tested to evaluate support for the olfactory crypsis hypothesis.
(1) Predators should be less able to detect nests treated with the preen oil of incubating birds compared to preen oil of non-incubating birds. To test this, field experiments should be conducted, although we recognize that experiments on nest predation are difficult to implement in the field. For example, one could measure the predator detection rate of artificial nests where eggs are smeared with preen oil secreted during incubation versus outside of incubation, compared to nests with no preen oil treatment. Such an experiment should ideally be combined with chemical analyses to verify and quantify chemical differences among treatments.
(2) During incubation, preen oil should become less volatile in the incubating sex compared to the nonincubating sex. This can be tested by taking repeated measurements of preen oil collected from incubating and non-incubating birds and performing chemical analyses to measure the volatility of preen oil at different time points.
(3) Preen oil may be transferred from parents to eggs and/or chicks. To test for evidence of preen oil transfer, one could search for traces of preen oil on the eggs and/or chicks, or determine (e.g. using video recording; Martín-Vivaldi et al., 2014) whether adults actively deposit preen oil onto the eggs and/or chicks. Currently, preen oil transfer has only been documented in a single species (Eurasian hoopoe, Upupa epops). In this species, preen oil becomes malodorous during breeding , suggesting it does not provide olfactory crypsis but may instead repel predators and/or parasites. (4) Preen oil could increase olfactory crypsis at the nest via background matching (Soini et al., 2007). To test this, one could analyse the chemical composition of preen oil secreted by the incubating parent(s) compared to the chemical composition of the nest and surrounding environment (e.g. using headspace sampling; Díez-Fern andez et al., 2021) and assess whether preen oil is more chemically similar to the environment than would be expected by chance.
(b) Sex semiochemical hypothesis Here, we outline specific predictions that should be tested to evaluate support for the sex semiochemical hypothesis.
(1) Birds should use preen oil odour cues to discriminate between the sexes and among individuals of varying quality. There is growing evidence for avian olfactory sex discrimination in the literature (e.g. Zhang et al., 2010;Whittaker et al., 2011a;Amo et al., 2012;Grieves et al., 2019a;Van Huynh & Rice, 2019). Additional experiments should be designed to test for evidence of olfactory discrimination based on indicators of genetic quality and compatibility such as genomewide heterozygosity (Whittaker et al., 2019a) and MHC genotype (Grieves et al., 2019c).
(2) Preen oil should become more volatile and/or more abundant (e.g. to increase detectability) during breeding than non-breeding. This can be tested by taking repeated measurements of preen oil collected from breeding and non-breeding birds and performing chemical analyses to measure the volatility of preen oil and the abundance of compounds of interest (i.e. those that have been associated with sex or individual differences) at different time points and between the sexes. (3) The preen oil preferences of birds tested in a laboratory should translate to mate choice in the wild. Such experiments are difficult to perform, not least because numerous factors affect mate choice in the field, but one could start by looking for evidence of nonrandom mating based on features birds have been shown to discriminate using odour cues, such as MHC genotype (Grieves et al., 2019c) or the relative 'maleness' of preen oil composition (Whittaker et al., 2011a). (4) Mate choice based on preen oil odour cues should also be linked to measures of mate quality and fitness. For example, in lance-tailed manakins, male reproductive success is correlated with microsatellite heterozygosity chicks of more heterozygous males are more likely to fledge, and heterozygosity is correlated with lower proportions of certain preen oil chemicals, but whether female manakins use preen oil odour during male mate choice is still unknown (Whittaker et al., 2019a). Odour preferences can be tested using a two-choice maze, and preferences can subsequently be linked back to field data on metrics such as heterozygosity, quality, and fitness.

VI CONCLUSIONS
(1) Determining the functions of preen oil chemical differences in birds has the potential to shift our understanding of avian behaviour. (2) We conducted an extensive literature review to evaluate the evidence for seasonal and sex differences in the chemical composition of avian preen oil (a proxy of avian body odour). Seasonal differences were nearly ubiquitous, while sex differences were found in almost half of the species studied. (3) We conducted a comparative analysis to test two hypotheses that may explain seasonal and sex differences in preen oil: the olfactory crypsis and the sex semiochemical hypotheses. (4) Our comparative analyses on both seasonal and sex differences supports the olfactory crypsis hypothesis. However, direct evidence of a role for preen oil differences in olfactory crypsis is still lacking, notably from experimental studies with natural predators and from studies on species other than shorebirds. (5) Our comparative analysis supports the sex semiochemical hypothesis. Evidence for the sex semiochemical hypothesis is growing, but more research is needed to connect preen oil differences to odour preferences, measures of quality, and mating success in the wild. (6) We suggest numerous predictions that can be tested to allow researchers to disentangle the olfactory crypsis and sex semiochemical hypotheses. Doing so will enable us to gain deeper insights into the role of chemical masking and chemical signalling in birds.

VII ACKNOWLEDGEMENTS
We thank the Wissenschaftskolleg Berlin for the time to formulate ideas (B.A.C., I.C.C., T.S.), two anonymous reviewers for comments that improved this manuscript, J. Winternitz for advice on statistical analysis, D. Whittaker for helpful discussion, and F. Bonadonna

X. Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Studies included in the literature review of seasonal and sex differences in preen oil chemical composition. Table S2. Occurrences (i.e. sex in each species or season in each species) included in the analyses of seasonal and sex differences in preen oil chemical composition.  Table S3. Definitions used to categorize studies on preen oil according to their main focus. Table S4. Species and studies included in comparative analysis testing whether sex-specific seasonal changes in preen oil chemical composition are related to incubation and nest ecology. Table S5. Species and studies included in a comparative analysis testing whether sex-specific changes in preen oil chemical composition are related to season (breeding/nonbreeding) and incubation type (uniparental/biparental). Table S6. Sample sizes for the analyses testing for factors affecting the likelihood of detecting seasonal and sex differences in avian preen oil chemical composition. Table S7. Effect of the inclusion of the random effects (species, phylogeny) on the fit of the Markov chain Monte Carlo generalized linear mixed effects (MCMCglmm) models on seasonal differences and sex differences in preen oil composition. Table S8. Mean proportion of the total variance explained by the random effects (phylogeny, species, and residual variance) in the two models fitted to explain seasonal and sex differences in preen oil chemical composition across species. Appendix S1. Supplementary methods. Appendix S2. References for studies included in Tables S4  and S5. Fig. S1. PRISMA flowchart for the systematic review and comparative analysis on seasonal and sex differences in preen oil chemical composition. Appendix S3. R code for the comparative analysis on seasonal differences in preen oil composition. Appendix S4. R code for the comparative analysis on sex differences in preen oil composition.