Decision rules for egg-color-based rejection by two cavity-nesting hosts of the brown-headed cowbird

By placing their young in other species' nests, obligate avian brood parasites offload the costs of parental care onto their hosts. In turn, hosts must either expend energy raising the foreign young as their own (Davies and Brooke, 1989; DuRant et al., 2013; Hauber, 2006; Hepp et al., 1990; Hoover and Reetz, 2006; Ruiz-Raya et al., 2018) or recognize parasitic young and remove them. The host's failure to detect parasitism can negatively affect both current (Hauber, 2003; Hoover, 2003) and future reproductive success (Antonson et al., 2020a; Hoover and Reetz, 2006); in response, many host lineages have evolved a wide range of defenses against parasitism to combat these substantial costs to fitness (Stevens, 2013). The resulting arms races between hosts and brood parasitic birds represent some of nature's best studied and most widely understood coevolutionary interactions (Stoddard and Hauber, 2017).

One frequently evolved host defense is the rejection of parasite eggs, which, despite some costs and errors (Lorenzana and Sealy, 2001; Sealy and Neudorf, 1995; Stokke et al., 2002), is generally a valuable defense mechanism (Stevens, 2013). For this behavior, the host's ability to accurately detect a parasite's egg based on its color depends on the host's color perception capabilities (Cassey et al., 2008), its visual acuity (Luro et al., 2020) and its cognitive abilities (Dixit et al., 2022). Therefore, eggshell coloration is the most common and effective features used by hosts for discriminating parasitic eggs (Samaš et al., 2021). We have long assumed that hosts use an acceptance threshold, whereby hosts should reject an egg when the difference in color between it and the host’s egg exceeds this threshold (Reeve, 1989). Although the exact cognitive processes underpinning egg recognition are not fully understood (Hanley et al., 2017, 2021), most underlying theory assumes that hosts respond to perceivable differences between their own eggs (or an internal template of their egg) and the parasite's egg (Moskát et al., 2014; Shizuka and Lyon, 2020; Stevens et al., 2013). Thus, because egg colors vary exclusively between blue-green and brown (Hanley et al., 2015a), hosts should reject eggs that are either bluer or browner than their own if the perceived color differences exceed the acceptance threshold (multiple thresholds hypothesis).

Contrary to these expectations, five hosts facing various forms of parasitism (conspecific or interspecific by cowbird, Molothrus spp., or cuckoo, Cuculus spp.) from different lineages and geographic regions did not reject experimentally introduced foreign eggs based solely on the perceived color differences between the model eggs and their own eggs (Abolins-Abols et al., 2019; Hanley et al., 2017, 2019b). Instead, these hosts preferentially rejected foreign eggs with colors on only one side of their eggs' natural phenotypic ranges (single threshold hypothesis). Specifically, they rejected eggs that they perceived as browner, but not bluer, than their own (Fig. 1). These findings raise important questions regarding the sensory ecology of hosts (Hanley et al., 2021), which are complicated because the host species in these prior studies all laid blue-green eggs. Brown-egg-laying species may respond by preferentially rejecting eggs bluer than their own, or they may lack a color-biased response entirely and simply reject foreign eggs based on their overall perceived differences (either bluer or browner), or (less likely) preferentially reject browner eggs (as the other hosts did). These questions can only be addressed by experiments assessing how brown-egg-laying hosts perform on similar egg color discrimination tasks. Ideally, such experiments should be conducted on both a blue-egg-laying host and a brown-egg-laying host that have similar nest types so that the eggshell color discrimination tasks are as comparable as possible.

Most studies on egg recognition have been conducted in open nesting species, but many parasitic species (Antonson et al., 2020b; Jamie et al., 2020; Spottiswoode, 2013) and individual parasitic females (Reboreda et al., 2013; Rutila et al., 2002; Tuero et al., 2007) exploit dome- or cavity-nesting hosts. Unlike open nests, cavity nests may not always provide enough light to stimulate the four single cone photoreceptors that are necessary to mediate color vision in birds (Olsson et al., 2015; Toomey et al., 2016), which could affect a host's ability to distinguish colors in a cavity nest. Nonetheless, there are some examples of cavity-nesting birds that respond to dramatic differences in eggshell colors in the context of foreign egg rejection (Grim et al., 2011; Hauber et al., 2014; Zhang et al., 2021) and birds can adapt their visual systems from bright to dim illumination within 1 s (Chaib et al., 2023). However, there are fewer examples of how hosts of avian brood parasites discriminate between fine-grained variation in eggshell coloration within dark cavity nests, where poorer light conditions could make egg color discrimination more challenging (Endler, 1993; Honza et al., 2014).

One such recent study on the cavity-nesting and immaculate blue-egg-laying common redstart (Phoenicurus phoenicurus) demonstrated that this host of the common cuckoo (Cuculus canorus) responded to model eggs based on their colors (Manna et al., 2020). Like several open-nesting hosts (Abolins-Abols et al., 2019; Hanley et al., 2017, 2019b), the cavity-nesting common redstarts discriminated eggs based on perceived color differences on one end of their phenotypic range (i.e. those they would perceive as browner than their own), supporting the single threshold model. These rejection patterns are surprising, because their parasite lays highly mimetic immaculate blue eggs, which could be either bluer or browner than their own (Igic et al., 2012). Together, these results suggest that light levels in common redstart cavity nests are sufficient to make relatively fine-grained decisions based on eggshell colors and that common redstarts use the same color-biased decision rule as other hosts (e.g. Abolins-Abols et al., 2019; Hanley et al., 2017, 2019b). Although these findings are illuminating (pun intended), we do not know whether other cavity nesters respond similarly. Furthermore, because past research focused predominantly on blue-egg-laying species, it remains unclear how brown-egg-laying species would respond to the same color discrimination task. For example, would they instead preferentially reject eggs bluer than their own?

To address these questions, we tested eggshell color discrimination abilities of two brown-headed cowbird (Molothrus ater; hereafter, ‘cowbird’) hosts that nest in cavities: eastern bluebirds (Sialia sialis) and house wrens (Troglodytes aedon). Brown-headed cowbirds are generalist brood parasites, impacting over 200 species of passerine host species in North America, and they lay eggs varying from white to light blue with variable dark brown maculation (Peer and Sealy, 2004). By contrast, eastern bluebirds typically produce light-blue immaculate eggshells, although their eggs can occasionally be white (Siefferman et al., 2006), and house wrens generally produce eggs so finely speckled that they appear reddish-brown over their entire surface, but the maculation of these eggs can be quite variable (Hodges et al., 2020; Thompson et al., 2022). The eastern bluebird rejects cowbird eggs at a rate of 65%, which is the highest of any cavity-nesting cowbird host (Peer et al., 2006; Woodward and Woodward, 1979). However, North American house wrens reject cowbird eggs at extremely low rates (Peer and Sealy, 2004) as they are only infrequently parasitized by cowbirds (Johnson, 2020). These low parasitism rates are likely due to house wrens' preference for nest cavities with entrance holes that are too small for cowbirds to enter or for their propensity to build protective nest architecture in front of the nest cup (Pribil and Picman, 1997; Stanback et al., 2013). However, we can still learn a great deal from this species about avian color perception and how they use color to inform their decisions. This is especially so because both hosts respond to rejection experiments using model eggs (our personal observations) and they readily nest in artificial cavities (i.e. nest boxes), which facilitates field experimentation.

In this study, we tested the hypothesis that both host species use a single threshold model for egg rejection and, therefore, only reject model eggs with colors that are at the opposite end of the egg color gradient (Fig. 1). To do so, we presented model eggs painted one of six distinct colors that spanned the natural brown to blue avian eggshell color gradient (Canniff et al., 2018; Hanley et al., 2015a). Because eastern bluebird and house wren natural egg phenotypes differ markedly, this design allowed us to determine not only whether both hosts respond to perceived variation in color, but also whether either host responds differently to eggs that are bluer or browner than their own. Specifically, we predicted that eastern bluebirds should preferentially reject eggs browner than their own, whereas house wrens should preferentially reject eggs bluer than their own. Alternatively, if these two species use a multiple thresholds decision-rule for egg rejection, we expected that they would reject eggs based solely on perceived color differences between model eggs and their own (either browner or bluer; Fig. 1). Simultaneously, our study design allowed us to test whether birds preferentially set a single threshold at any particular side of the phenotypic range or whether they preferentially reject eggs browner than their own (e.g. Hanley et al., 2017). By testing two host species with egg colors on opposing sides of the color spectrum and with similar nest types, we provide a more reliable test of these hypotheses because both species should have different thresholds and therefore elicit opposing responses.

We monitored eastern bluebird nests from mid-March through late July 2021 at several field sites within 20 km of Davidson, North Carolina, USA (35°29′N, 80°50′W), and house wren nests from April through late July 2020 and 2021 at the Mackinaw study area (40°39′N, 88°53′W) in north-central Illinois, USA. Although we include two years of house wren data, we only included females without prior experience with these model eggs to rule out the possibility that females learned from their previous year's exposure to egg models (Moskát et al., 2014). We also included year as a covariate in all resulting models on house wrens because only second clutches were manipulated in 2020 whereas only first broods were studied in 2021.

Light levels at the nest cup were dim for both eastern bluebirds (median=106.5 lux, interquartile range=177.75 lux) and house wrens (median=37.5 lux, interquartile range=176.5 lux, see Supplementary Materials and Methods). Although there is interspecific variation in the minimum amount of light necessary to adequately perform color discrimination (Kelber and Lind, 2010; Kelber et al., 2017; Olsson et al., 2015), these nest light levels suggest that 92.5% and 82.5% of eastern bluebird and house wren nests, respectively, exceeded the threshold beyond which color discrimination is not limited by light levels (∼10 cd m⁻²). Another 5% and 10% of eastern bluebird and house wren nests, respectively, were dark enough to expect light levels to limit color discrimination (between 0.1 and 10 cd m⁻²). Finally, 2.5% and 7.5% of eastern bluebird and house wren nests, respectively, were likely too dark for any color discrimination. Based on our understanding of how light limits color discrimination (Kelber and Lind, 2010; Kelber et al., 2017; Olsson et al., 2015), we would expect color discrimination to be possible within the majority of these boxes. However, the color discrimination abilities of hosts at a subset of nests would likely be reduced owing to limited light availability and, more rarely, not possible at all.

For both host species, we placed a single model egg in their nests that was similar to the dimensions and mass of their own eggs. Eastern bluebirds received wooden eggs (mean±s.e.m.: length=21.94±0.02 mm, width=16.26±0.02 mm, mass=1.9±0.04 g) that were commercially available (7/8″ model, American Woodcrafter Supply) and which approximated the reported ranges of their own eggs (length=18.1–24.3 mm, width=14.7–19.2 mm, mass=2.15–3.85 g; Gowaty and Plissner, 2020). Eastern bluebirds are grasp ejectors, and use their beaks to remove eggs from their nests (authors’ personal observations). Eastern bluebirds rejected these wooden model eggs at a similar rate (66%; M.T.S. and M.E.H., unpublished data) to the previously reported rejection rate (65%) of natural cowbird eggs of comparable size and mass (Peer et al., 2006) when the wooden models were painted to appear white with speckling, despite the fact that these artificial eggs had similar dimensions but a slightly lower mass than natural cowbird eggs (length=18–25 mm, width=15–18 mm, mass=2.5–3.6 g; Strausberger, 1998). House wrens received a single 3D-printed egg model (Igic et al., 2015) with dimensions (model egg means±s.e.m.: length=16.94±0.01 mm, width=11.93±0.02 mm, mass=1.25±0.002 g) that also approximated those of their own eggs (length=14.80–18.99 mm, width=11.29–13.84 mm, mass=1.02–1.88 g; Johnson, 2020). House wrens frequently puncture eggs to remove them from other birds' nests, but our wren-sized model eggs could not be pierced and therefore were removed by grasping.

Each model egg was painted one of six colors along a blue-green to brown gradient (Canniff et al., 2018) that spanned the natural variation across all avian eggshell coloration (Hanley et al., 2015a) (Fig. 1). Host nests were assessed for rejection 4 days (house wren) or 7 days (eastern bluebird) after the insertion of the model eggs. These periods were selected owing to logistical constraints at each study site, but are justified because rejection of avian eggs generally occurs within 2 days of deployment in most rejecter species (e.g. Hanley et al., 2015b). Thus, these rejection periods would allow for ample time for birds to reject, while also limiting the time researchers spent at their nests. The experimental egg was considered ‘accepted’ if the model egg remained in the nest cup at the end of these respective periods, whereas the egg was considered ‘rejected’ if the egg disappeared entirely from the nest cavity or was no longer within the incubated clutch (e.g. on the nest rim). We excluded nests when predation occurred during the experimental period, or whenever a response could not be reliably assessed over the rejection period. In two known cases, house wrens rejected a model egg on the sixth day, but we coded this as an acceptance because it had remained in the nest for at least 4 days. The final sample size resulted in 143 house wren nests and 97 eastern bluebird nests, after excluding abandoned nests, depredated nests and nests where pairs had prior exposure to these experimental eggs.

We measured the reflectance spectra (from 300 to 700 nm) for a subset of model eggs and abandoned natural host eggs of each species (7 eastern bluebird eggs, 8 house wren eggs) using a reflectance spectrometer (Jaz, Ocean Optics, Dunedin, FL, USA). We determined the photoreceptor quantum catch by modeling the visual system for each species using the pavo package in R (Maia et al., 2019). For these calculations, we assumed that the eastern bluebirds and house wrens both have single cone photoreceptor sensitivities corresponding to the average ultraviolet-sensitive avian viewer and double cone photoreceptor sensitivities of the blue tit (Cyanistes caeruleus) (Ödeen et al., 2011). The light available in nest boxes can vary with both time (time of day and year) and the direction of the entrance hole; therefore, to avoid making assumptions about the quality of light, we calculated quantum catches assuming an idealized light environment but scaled these for relatively dim conditions to account for the low light availability in nest boxes by multiplying the values by 500 (the ‘scale’ option in the vismodel function in pavo). Importantly, although light availability within these boxes was low (see above), data on light intensity thresholds suggest that it was sufficiently bright for color discrimination in most cases (Kelber and Lind, 2010; Kelber et al., 2017). We then used the receptor noise-limited model (Vorobyev and Osorio, 1998) to calculate the chromatic and achromatic contrasts between the average host egg and each egg-model type (the six colors used for each host species), using a Weber fraction of 0.1 for the long-wavelength-sensitive photoreceptor, the typical assumption for these models (Olsson et al., 2018). For these calculations, we used the average densities for the four photoreceptors of 22 species for which this information is known (Hart, 2001), because these data are not available for the eastern bluebird or the house wren, specifically. Finally, because we were uncertain of the threshold between photopic (visual signals from only cones) and mesopic vision (visual signals from both rods and cones) for either species, we ran visual models accounting for neural noise assuming that noise is independent of light availability (Vorobyev and Osorio, 1998) as well as models that assumed that noise was proportional to the sum of both neural and receptor noise, which accounts for the effect of dim-light conditions on color discrimination (Antonov et al., 2011; Holveck et al., 2010; Osorio et al., 2004). Furthermore, we performed a sensitivity analysis (Delhey et al., 2013) to ensure that our analyses were robust to variation in cone densities and the individual reflectance spectra of host eggs (see Supplementary Materials and Methods).

The visual models produce estimates of perceivable differences in units of just-noticeable differences (JND), where a JND greater than 1 represents a stimulus that is above the discriminable threshold and a JND less than 1 represents a stimulus that is unable to be discerned (Igic et al., 2012). We then plotted each egg-model type within the avian tetrahedral color space and recorded the relative position to the hosts' eggs to determine whether the egg model was bluer or browner than the host's egg (Hanley et al., 2019b). When the model was bluer or brighter than the host's eggs, we multiplied the chromatic contrast or achromatic contrast (respectively) by 1, but when it was browner or darker, we multiplied these values by −1. These data also allowed us to determine the direction of difference (e.g. browner or bluer and brighter or darker) of each model egg, which we refer to as either directional chromatic contrast or directional achromatic contrast. These visual models predicted the degree to which our model eggs would be perceived as bluer or browner and lighter or darker than each host egg (Table 1). Directional chromatic and achromatic contrasts should predict host rejection if they use a single threshold (Fig. 1), while chromatic and achromatic (non-directional) contrasts should predict host rejection if they use multiple thresholds (Fig. 1). Because we do not know the light conditions within the nest boxes when each host made its decision, we present both models that assume light was not limiting and models that assume light was limiting, understanding that in most cases the true degree of discriminability would be somewhere in between.

We predicted a dichotomous host response (reject or accept) using generalized linear models assuming a binomial distribution (logit link function). For each species, we ran separate models for chromatic and directional chromatic contrasts, addressing the multiple and single threshold hypotheses, respectively. Each model controlled for date of the first laid egg in the nest (ordinal date) and clutch size (continuous), as well as year (categorical) for the house wren dataset. Whole model significance was assessed via a likelihood ratio test comparing a null model (intercept only) with each model for each host species. Unfortunately, despite our best efforts, we were unable to make each egg model color equally bright; the model eggs used in our study were uniformly darker than eastern bluebird eggs and lighter than house wren eggs. Therefore, we also analyzed whether directional achromatic contrasts (i.e. perceived differences in brightness) impacted host response using similar generalized linear models. In this case, we opted to run and compare separate models because we detected a large variance inflation factor (>3), suggesting that significant multicollinearity between the perceived chromatic and achromatic contrasts of house wren eggs would impact the interpretation of our models (Freckleton, 2010; Zuur et al., 2010). Therefore, we opted to compare models within each species via AICc weights (Burnham and Anderson, 2004; Burnham et al., 2011). If these models produced similar results to the chromatic or directional chromatic models, then attributing host response to color alone would be challenging. These models also provide additional information about how hosts respond to parasitic eggs with noticeably different brightness to their own. Analyses were performed in R v. 4.0.3 (https://www.r-project.org/).

Animal care and experimental methods were approved by the University of Illinois Urbana-Champaign and Davidson College Institutional Animal Care and Use Committees (protocol numbers 17259 and 2-21-124, respectively) and under the Illinois Department of Natural Resources (NH21.6220, NH20.6220, NH20.0004, and NH21.0004) and North Carolina Wildlife Resources Commission (22-SC00428), the US Fish & Wildlife Service (MB692148-0 and MB08861A-10), and the US Geological Survey (09211 and 22742).

We found low rejection rates for both species, with eastern bluebirds rejecting only 13 of 97 eggs (13.4%) and house wrens rejecting only 5 of 49 eggs (10.2%) in 2020 and 18 of 103 eggs (17.4%) in 2021.

The probability of egg rejection in eastern bluebirds was significantly and positively related to chromatic contrast between the model eggs and the average color of an eastern bluebird egg (Table 2). To a lesser, but statistically similar extent, their responses were also predicted by directional chromatic contrasts to their own eggs (Table 2, Fig. 2). Both models were within two AICc scores of one another; therefore, neither model provided a notably better fit to these data. Conversely, achromatic contrasts between eastern bluebird eggs and the model eggs did not predict their responses (Table 2). These findings were robust to variation in photoreceptor densities and individual eggshell reflectance spectra (Fig. S1, see Supplementary Materials and Methods).

Unlike the eastern bluebirds, house wren responses were not predicted by chromatic contrast between the model eggs and their own (Table 2). House wren responses were, however, predicted by directional chromatic contrasts, representing the color of the model eggs relative to their own eggs. In this case, house wrens were more likely to reject eggs that were bluer than their own, rather than browner than their own (Table 2, Fig. 2). As with eastern bluebird responses, house wren responses were not predicted by achromatic contrasts between their own eggs and model eggs.

Analyses using perceptual models for both species that assumed that receptor noise was related purely to neural noise, where light is not limiting, produced virtually identical results (Table S1).

Assessing the salient cues for egg recognition is vital for our understanding of coevolutionary dynamics between hosts and their obligate brood parasites (Hauber et al., 2015a). Previous research found that blue-egg-laying hosts rejected model eggs that were browner than their own, but accepted model eggs bluer than their own, even if the colors of both model eggs were equally dissimilar to those of their own eggs (sensuHanley et al., 2017). This could suggest strong selection for bluer eggs in parasite populations. Unfortunately, those prior studies never investigated the response of brown-egg-laying host species, which may use a single threshold to preferentially reject eggs either bluer or browner than their own, or may use a multiple threshold (i.e. rejections with no color bias). Here, we discovered that the brown-egg-laying house wren preferentially rejected eggs bluer than their own eggs (Fig. 2), and thus exhibited color-biased rejection behavior consistent with the single threshold model as found in several other species (Abolins-Abols et al., 2019; Hanley et al., 2017, 2019a; Manna et al., 2020). Importantly, this suggests that a host's single threshold can be based on relevant self-referential cues (e.g. species laying browner eggs reject bluer eggs, whereas species laying bluer eggs reject browner eggs) rather than always being biased toward a particular color. Unlike previous studies, we found that eastern bluebirds rejected eggs that were either bluer or browner than their own eggs (Fig. 2). Thus, our evidence suggests that bluebirds may set multiple thresholds, whereas house wrens set a single threshold.

We found that the rejection behavior of the blue-egg-laying eastern bluebird was more likely predicted by the absolute perceived chromatic contrast to their own eggs; therefore, eastern bluebirds most likely use multiple thresholds for egg color discrimination tasks as found in many species (Cassey et al., 2008; Hauber et al., 2015b; Stoddard and Stevens, 2011). However, further tests are necessary to rule out whether they reject based on directional chromatic contrasts (single thresholds) as more recently found in other hosts (Abolins-Abols et al., 2019; Hanley et al., 2017, 2019a; Manna et al., 2020). In fact, bluebirds were more likely to reject brown eggs, overall (Table 2, Fig. 2), and our statistical models predicting eastern bluebird responses were unable to differentiate between models including chromatic and directional chromatic contrasts, as both models were significant and within two AICc scores of one another, making it challenging to distinguish between these alternative models (Symonds and Moussalli, 2011). Future studies should present a wider range of experimental eggs, including eggs that are far bluer than the eastern bluebird's eggs to distinguish between these two competing hypotheses.

By contrast, house wren rejection behaviors were best explained by the single threshold model (solid lines, Fig. 1), where they reject dissimilar bluer eggs but accept equally dissimilar browner eggs. Importantly, because house wrens frequently puncture eggs they are rejecting, we may have underestimated their rejection rates by using (impenetrable) plastic model eggs; however, this should not have resulted in different rejection rates for egg models of different colors. House wrens also exhibited a significant year effect on rejection rates. This statistical effect of year may be because the 2020 experiment was conducted during the second brood only (first-egg dates 19 June–21 July), whereas the 2021 experiment was conducted during the first two broods (7 May–13 July). Thus, the differences may actually reflect differences in brood order rather than year per se. Although house wren rejection behaviors appear adaptive, it is noteworthy that they contrast with the pattern found so far in other hosts, all of which have rejected browner eggs over bluer eggs (Abolins-Abols et al., 2019; Hanley et al., 2017, 2019b; Manna et al., 2020). Overall, these new findings suggest that hosts likely evolve a self-referential decision-rule, accepting eggs with similar colors to their own eggs (i.e. similar positions within the color space) rather than rejecting egg colors based on absolute perceivable differences per se. Such behavioral responses can have important implications on the arms races between hosts and parasites, including the coevolution of host defenses and, in turn, parasitic deceptions (Stevens et al., 2013; Stokke et al., 2007).

Despite low light levels, these cavity-nesting hosts still rely on eggshell color when making decisions within the nest (albeit in different ways). Thus, even though darker nest environments should reduce these hosts' color discrimination ability within their cavity nests (Honza et al., 2014; Kelber et al., 2017), our data suggest that light levels were sufficiently high for color discrimination. These results corroborate previous findings on a cavity-nesting host of the common cuckoo, the common redstart, in which rejection rates are also predicted by chromatic rather than achromatic differences (Manna et al., 2020). More broadly, research has shown that cavity-nesting species can respond to color variation despite being in extremely dim environments (Antonov et al., 2011; Dugas, 2015; Holveck et al., 2010; Honza et al., 2014). Although the timing of egg-rejection decisions is not well studied in wild birds (but see Hauber et al., 2015b, 2019), such color discrimination could occur if birds make decisions while their eyes are bright-adapted (Maziarz and Wesołowski, 2014; Węgrzyn et al., 2011; Wesolowski and Maziarz, 2012; Zele and Cao, 2015) shortly after returning to the dimly lit nest or while their eyes are in a mesopic state (Węgrzyn et al., 2011; Wyszceki and Stiles, 1982; Zele and Cao, 2015). Although the results from light-limited models in our study did not differ from the perceptual models without light limitations (see Supplementary Materials and Methods), these light-limited models do suggest that color differences between the blue-green coloration of bluebird eggs and bluer egg models may be particularly difficult when performing color discrimination in dim light conditions (Table 1). Thus, if some bluebirds experience lower light than others, we do expect that bluer eggs would be harder to detect than browner eggs. Overall, our results are consistent with prior evidence that has demonstrated that birds rely on color variation for discrimination even in quite dim environments (Honza et al., 2014); therefore, future tests of color discrimination within low-light environments could be particularly fruitful.

Here, we demonstrate that two cavity-nesting host species with very different egg phenotypes (Fig. 1C) use color discrimination in response to experimental parasitism (Fig. 2). Each species was biased toward colors unlike those of their own eggs; however, the single threshold model was only supported in the house wren. Further tests are necessary to definitively rule out eastern bluebirds using single thresholds, but our evidence suggests that they most likely use multiple thresholds when making egg-rejection decisions. These responses provide further support of the hypothesis that hosts do not simply respond to perceivable differences (Hanley et al., 2017), but treat certain colors (i.e. ‘own’ versus ‘other’) as different despite having the capacity to differentiate those colors (Harnad, 1987). Furthermore, neither species relied on differences in eggshell brightness for egg recognition, suggesting that light was sufficiently bright for color discrimination within their cavity nests. We encourage future research to directly measure and experimentally manipulate light within cavities to better understand how low light availability limits host-egg recognition, as well as the underlying co-evolutionary implications of such a finding. Although we have much to learn about how light conditions within cavity nests affect egg recognition and the detection of other colorful signals (Dugas, 2015), in this study we have demonstrated the (limited) color discrimination abilities of two cavity nesters.

We are grateful for the help and support in collecting data from Julia Barnfield, Rachael DiSciullo, Paige Duncan, Finlay Holston, Faith Kipnis, Tifani Panek, Christine Poppe, Max Rollfinke, Thomas Seabourn, and the 2020 and 2021 Wren Crew undergraduate students at Illinois State University. In addition, we are grateful for very helpful comments from Almut Kelber, Kaspar Delhey and two anonymous reviewers.