Hosts of avian brood parasites can avoid the reproductive costs of raising genetically unrelated offspring by rejecting parasitic eggs. The perceptual cues and controls mediating parasitic egg discrimination and ejection are well studied: hosts are thought to use differences in egg color, brightness, maculation, size and shape to discriminate between their own and foreign eggs. Most theories of brood parasitism implicitly assume that the primary criteria to which hosts attend when discriminating eggs are differences between the eggs themselves. However, this assumption is confounded by the degree to which chromatic and achromatic characteristics of the nest lining co-vary with egg coloration, so that egg–nest contrast per se might be the recognition cue driving parasitic egg detection. Here, we systematically tested whether and how egg–nest contrast itself contributes to foreign egg discrimination. In an artificial parasitism experiment, we independently manipulated egg color and nest lining color of the egg-ejector American robin (Turdus migratorius), a host of the obligate brood parasitic brown-headed cowbird (Molothrus ater). We hypothesized that the degree of contrast between foreign eggs and the nest background would affect host egg rejection behavior. We predicted that experimentally decreasing egg–nest chromatic and achromatic contrast (i.e. rendering parasitic eggs more cryptic against the nest lining) would decrease rejection rates, while increasing egg–nest contrast would increase rejection rates. In contrast to our predictions, egg–nest contrast was not a significant predictor of egg ejection patterns. Instead, egg color significantly predicted responses to parasitism. We conclude that egg–egg differences are the primary drivers of egg rejection in this system. Future studies should test for the effects of egg–nest contrast per se in predicting parasitic egg recognition in other host–parasite systems, including those hosts building enclosed nests and those parasites laying cryptic eggs, as an alternative to hypothesized effects of egg–egg contrast.
Obligate brood parasites circumvent the costs of parental care and lay their eggs in the nests of other species (Davies, 2000). By accepting the burden of raising genetically unrelated offspring, brood parasite hosts suffer major fitness costs (Øien et al., 1998; Lorenzana and Sealy, 2001; Hauber, 2003a,,b; Hoover, 2003). The rejection of foreign eggs in the nest is an effective defense against brood parasitism (Rothstein, 1975; Grim et al., 2011; Kilner and Langmore, 2011), which places reciprocal selective pressure on parasites to evolve egg coloration and/or maculation to match that of its host. This then selects for increasingly fine-tuned discrimination by hosts (Davies and Brooke, 1989; Stoddard and Stevens, 2010,, 2011; Davies, 2011). Such an arms-race is a canonical example of co-evolutionary processes driving both perceptual and signaling mechanisms (Davies and Brooke, 1989; Davies, 2011; Igic et al., 2012; Stoddard et al., 2014).
The proximate, perceptual controls underlying egg rejection behavior have been intensively studied in various brood parasite–host systems (Kilner and Langmore, 2011). Generally, an egg should be perceived as foreign if it differs beyond a given threshold from the variation present within a host female's natural clutch (Reeve, 1989; Rodríguez-Gironés and Lotem, 1999). Such recognition is dependent on a number of factors, including the population parasitism rate (Davies et al., 1996), the number of host eggs present, and the timing of egg parasitism (e.g. Moskát and Hauber, 2007). Hosts' acceptance thresholds also vary according to experience, even within a single clutch (Hauber et al., 2006). Hosts can respond to differences in eggshell background color (Avilés et al., 2005,, 2010; Honza et al., 2007; Honza and Polacˇiková, 2008; Moskát et al., 2008; Bán et al., 2013; Croston and Hauber, 2014a), maculation pattern (Lawes and Kirkman, 1996; Lahti and Lahti, 2002; López-de-Hierro and Moreno-Rueda, 2010; Spottiswoode and Stevens, 2010), egg brightness (Lahti, 2006; Gloag et al., 2014), egg size (Rothstein, 1982; Marchetti, 2000) and egg shape (Guigueno and Sealy, 2012) when discriminating their own from foreign eggs.
While above-threshold visual contrast is increasingly known to induce egg rejection among brood parasite hosts, it is not firmly established whether comparing own versus foreign eggs is a more reliable cue than other visual comparisons available in the host's nest environment (Endler and Mielke, 2005; Thorogood and Davies, 2013). For example, relatively few studies have examined whether and how nest lining color influences parental behavior (but see Bailey et al., 2015). Regarding parasitic egg rejection by host parents, the role of egg–nest contrast has similarly not been well established (Siefferman, 2006), and only a handful of studies have experimentally tested the hypothesis that visual contrasts between eggs and their background (i.e. the nest lining) affect egg rejection decisions (Gloag et al., 2014; Honza et al., 2014). Growing evidence suggests that there is selective pressure for brood parasites to evolve dark, cryptic eggs among Australasian cuckoo–host systems, making egg detection by hosts or, rather, by competing parasites difficult because the eggs blend in with the nest background (Langmore et al., 2005,, 2009; Gloag et al., 2014). While similar arguments have also been made for other host–parasite systems (Mason and Rothstein, 1987; Honza et al., 2011,, 2014), experimental tests of whether egg–nest contrast affects parasitic egg discrimination in the context of both natural and experimental egg color variation are lacking.
We focused on the North American brown-headed cowbird [Molothrus ater (Boddaert 1783); hereafter, cowbird]–American robin (Turdus migratorius Linnaeus 1766; hereafter, robin) parasite–host system. Robins are a suitable study species in that they are one of fewer than 30 documented cowbird host species to eject cowbird eggs at rates above 75% (Briskie et al., 1992; Peer and Sealy, 2004), allowing for the testing of specific sensory hypotheses mediating egg rejection in this system. Previous work on this species pair showed that natural and model cowbird eggs are perceptually distinct from natural (conspecific) robin eggs: natural parasite eggs are rejected from 100% of experimental nests, whereas conspecific natural and model robin eggs are not rejected (Briskie et al., 1992; Croston and Hauber, 2014a; Rothstein, 1982; Fig. 1A). Because egg color variability within robin clutches is significantly lower than egg color variability between clutches, robins may compare foreign eggs against the relatively low color variability present within the entire clutch in their egg rejection decisions (Abernathy and Peer, 2014, Croston and Hauber, 2015; see also Fig. 1B).
In this host–parasite system, artificially colored and natural eggs also exhibit strongly and positively correlated chromatic contrast against both natural robin eggs and natural robin nest linings, as measured by avian visual modeling (Fig. 1B). Similar to robins' intra-clutch color variability, natural robin nest linings show low spectral variability across the avian visible range (supplementary material Fig. S1) as well as low avian-perceived chromatic and achromatic contrasts when compared against each other (supplementary material Fig. S2). Further, avian-perceived visual chromatic and achromatic contrasts between robin eggs and natural nest linings are generally lower than that between cowbird eggs and robin nests (Fig. 1B). Thus, egg–nest contrast potentially confounds the degree to which we understand egg–egg contrasts to serve as the necessary and/or sufficient cues for parasitic egg discrimination in this and other host–parasite systems.
Here, we hypothesized that artificial eggs that more closely resemble the nest background (i.e. are cryptic) are more likely to be accepted. We experimentally tested the degree to which egg–nest contrast affects egg rejection, independent of egg–egg contrast, predicting that increasing/decreasing egg–nest contrast (thereby rendering eggs less/more cryptic), would increase/decrease parasitic egg ejection rates. Alternatively, egg–nest contrast may not itself affect hosts' rejection decisions, which would support the role of foreign versus own egg differences themselves as the primary cue for parasitic egg discrimination. To establish the degree to which egg–nest contrast per se influences parasitic egg discrimination, we manipulated the nest-lining color of robin nests (Fig. 2) in an artificial brood parasitism experiment.
We parasitized robin nests with plaster-of-Paris eggs painted the same colors as our nest lining manipulations (cowbird ground color-mimetic – hereafter, beige; blue–green – hereafter, robin-mimetic; and red), and whose rejection rates in non-manipulated nests are known from our published work (Table 1, Fig. 2; these egg colors and their rejection rates in natural nests were sourced from Croston and Hauber, 2014a). To determine the extent to which we successfully manipulated artificial egg–nest lining contrast, we conducted avian visual modeling analyses on egg and nest-lining reflectance spectra (Fig. 3), and analyzed raw reflectance spectra themselves (see Materials and methods) as a methodological check. We specifically predicted that artificially increasing the visual contrasts (measured as just-noticeable differences, or JNDs, from visual modeling analyses) between experimental parasitic eggs and the nest background would result in increased rejection rates, while artificially decreasing contrast would decrease rejection rates (Table 1). We then tested our predictions by assessing the extent to which artificial egg–nest lining achromatic and chromatic contrasts predicted egg rejection rates.
Covariation of egg–egg versus egg–nest contrasts with published egg rejection rates
Natural robin eggs (which elicit no ejection; Briskie et al., 1992) possessed significantly lower chromatic contrasts than natural cowbird eggs (which are always ejected) when compared with natural robin eggs sourced from different conspecific nests (U1=21.77, P<0.0001; Fig. 1A). In parallel, natural robin eggs possessed significantly lower chromatic contrasts against natural robin nest linings relative to natural cowbird eggs (U1=6.00, P=0.01, Fig. 1A). Further supporting our claim that there is a quantitative confound between egg–egg chromatic contrast and egg–nest chromatic contrast, we found a strong and significant positive relationship between artificial egg–natural robin egg and artificial egg–natural nest chromatic JNDs (F1,4=30.24, P=0.0053; Fig. 1B) by including natural and artificial egg stimuli analyzed in Croston and Hauber (2014a). We also found that color variation among natural robin nest linings is low (supplementary material Figs S1 and S2), suggesting that the nest lining itself presents a reliable cue to be used by robins to perceptually discriminate own from foreign eggs.
Perceptual outcomes of egg/nest lining color manipulations
We found that natural robin egg–egg chromatic contrasts (mean±s.e.m. 1.91±0.30) were significantly lower than natural egg–natural nest lining contrasts (3.98±0.19; U1=9.60, P<0.001; Fig. 1). In contrast, achromatic natural egg–egg contrasts (2.84±0.65) were not significantly different from egg–natural nest lining contrasts (3.70±1.53; U1=0.154, P=0.69) when compared with natural nest lining. Together, these results suggest that it is chromatic contrast against the natural nest lining that may provide a strong cue against which to compare foreign eggs.
We compared avian-perceived chromatic differences between all eggs and nest-lining colors to test our predictions outlined in Table 1. We found a significant effect of nest-lining color (Nbeige nests=15, Nred nests=15, Nrobin-mimetic nests=15, Nnatural nest=5) on chromatic contrast among beige eggs (H3=44.63, P<0.0001). All pairwise comparisons were significant (P<0.05; Fig. 4A): beige eggs had the highest chromatic contrast in red nests, followed by robin-mimetic nests, natural nests, then beige nests. We also found a significant effect of nest-lining color on chromatic contrast among robin-mimetic eggs (H3=42.67, P<0.0001). Red nests had the highest chromatic contrast with robin-mimetic eggs, followed by beige nests, natural nests and robin-mimetic nests. The amount of chromatic contrast between robin-mimetic eggs and nests differed significantly among all pairs (P<0.05), except between beige and natural nests (Fig. 4C). Lastly, we found a significant effect of nest-lining color on chromatic contrast among red eggs (H3=45.00, P<0.0001). All pairwise comparisons were significant (P<0.05; Fig. 4E); red eggs in robin-mimetic nests had the highest chromatic contrast, followed by beige nests, natural nests and red nests.
The analyses above were conducted using an ultraviolet-sensitive (UVS) perceptual model for robin vision (based on Aidala et al., 2012), and we carried out a separate set of visual model analyses using violet-sensitive (VS) visual model parameters (see Materials and methods). The results followed the same chromatic contrast patterns as above, although JND values were generally much larger using this model than in our UVS visual model (Fig. 4; supplementary material Fig. S3A,C,E). Similarly, chromatic distance analyses of chromatic principal component (PC)2 and PC3 scores (supplementary material Table S1, Fig. S4) of raw reflectance spectra (as a measure of chromatic distance) between eggs and nest linings corroborated the patterns seen in both of our visual modeling analyses (Fig. 4A,C,E; supplementary material Fig. S5) and followed the same pattern when compared against rejection rates as chromatic JNDs (Fig. 5A; supplementary material Fig. S6A and Fig. S7A).
We also compared avian-perceived achromatic differences between all eggs and nest-lining colors (Fig. 4B,D,F). We found a significant effect of nest-lining color on achromatic contrast among beige eggs (H3=43.54, P<0.0001; Fig. 4B). There was also a significant effect of nest-lining color on achromatic contrast among robin-mimetic eggs (H3=38.03, P<0.0001; Fig. 4D) and red eggs (H3=41.38, P<0.0001; Fig. 4F). Neither our VS visual modeling analysis (supplementary material Fig. S3) nor our PC1 distances (as a measure of achromatic distance – see Materials and methods; supplementary material Fig. S5B,D,F) paralleled the visual contrasts in our achromatic UVS visual model. However, because neither PC1 distances nor achromatic JNDs (from either visual modeling analysis) between artificial eggs and nest linings were significantly related to rejection rate (see below; Fig. 5B, supplementary material Figs S6B and S7B), we only included achromatic JNDs in further behavioral analyses so as to be consistent with our analysis of chromatic JNDs. Because of the similarities in chromatic contrasts for both the VS visual model and analysis of chromatic principal components, we focused on our primary UVS visual modeling data (JNDs) for our behavioral analyses (see below).
We conducted a total of 94 artificial parasitism experiments with model eggs (Nbeige eggs=34, Nrobin-mimetic eggs=29, Nred eggs=31) in nests with beige (N=17), red (N=19) and robin-mimetic (N=12) linings. When combined with egg rejection rate data of artificial egg colors in natural nests from Croston and Hauber (2014a), the mean chromatic contrasts in egg–nest treatments were not significantly related to the rejection rate in our egg–nest manipulations, and the regression slope was slightly negative and thus in the direction opposite to our predictions (F1,10=0.49, P=0.50, R2=0.05; Fig. 5A). Similarly, when natural nest data were removed from this analysis, the relationship trended in the same direction but remained non-significant (F1,7=0.25, P=0.63, R2=0.03). Achromatic contrasts were also not significantly related to rejection rate in our nest manipulations both with natural nest rejection data from Croston and Hauber (2014a) included (F1,10=0.43, P=0.53, R2=0.04; Fig. 5B) and without these data (F1,7=0.0004, P=0.98, R2<0.01).
A Friedman ANOVA revealed a consistent effect of egg color on robin egg rejection behaviors: relative egg rejection rates were consistently ordered as beige>red>robin-mimetic eggs across, and irrespective of, the three colors of experimental and one natural nest lining types (χ22=8.00, P=0.018; Fig. 5C). To confirm these results, we fitted binomial generalized linear mixed models (GLMMs) to further describe predictors of egg rejection. In order to be conservative in the analysis and interpretation of our data, we first controlled for individual females' known propensity to consistently reject or accept foreign eggs irrespective of egg coloration (Croston and Hauber, 2014b; supplementary material Table S2). Although only one nest site was significant (red nest, all egg colors accepted; P=0.04) in the model, we also removed a second site that approached significance (robin-mimetic nest, all egg colors rejected; P=0.06) to be conservative, thereby excluding two sites at which female robins responded to neither egg color nor nest lining treatment. We therefore excluded a total of six experiments at these two nests from further behavioral analysis, leaving N=88 experiments analyzed in subsequent models (supplementary material Table S2B; Table 2).
We then combined our dataset with published egg rejection data of artificial eggs in natural, non-manipulated robin nests (Croston and Hauber, 2014a). The full model significantly predicted artificial egg rejection/acceptance outcome (χ28=41.19, P<0.0001). The only significant predictor of egg rejection in this model was egg color (χ22=39.77, P<0.0001; Table 2A). Last, we fitted a GLMM including all above predictors, as well as chromatic and achromatic contrast between egg and nest-lining colors. Again, the whole model significantly predicted egg acceptance/rejection behavior (χ210=42.82, P<0.0001), but neither chromatic nor achromatic egg–nest JNDs were a significant predictor of egg rejection. As in our above models, the only significant predictor of egg rejection was egg color (χ22=34.05, P<0.0001; Table 2B).
To further confirm these results, we ran post hoc tests on the single significant predictor (egg color) in the final GLMM model (Table 2B). The post hoc χ2 test of egg color against a reject/accept outcome variable showed a significant difference in egg rejection behavior by egg color (χ22=40.39, P<0.0001; Fig. 5C). Irrespective of nest type, beige eggs were rejected in 33 out of 39 trials, robin-mimetic eggs were rejected in five out of 35 trials, and red eggs were rejected in 23 out of 43 trials. When split by nest type (Fig. 5C, Table 1), there was a significant difference in rejection rate of each egg type in beige nests (χ22=24.69, P<0.0001). In beige nests, beige eggs were rejected in 12 out of 13 trials, robin-mimetic eggs were rejected in 0 out of nine trials and red eggs were rejected in four out of 12 trials. Further analysis showed that beige eggs were rejected significantly more often than both robin-mimetic eggs (χ21=23.27, P<0.001) and red eggs (χ21=10.34, P=0.0013; Table 1). Red eggs were similarly rejected more often than robin-mimetic eggs (χ21=5.17, P=0.02; Table 1). There was no significant difference in egg rejection by egg color in robin-mimetic nests (χ22=4.68, P=0.10; Table 1). In robin-mimetic nests, beige eggs were rejected in six out of eight trials, robin-mimetic eggs were rejected in two out of eight trials and red eggs were rejected in four out of six trials. There was also no significant difference in egg rejection by egg color in red nests (χ22=3.97, P=0.14; Table 1). In red nests, beige eggs were rejected in eight out of 11 trials, robin-mimetic eggs were rejected in three out of 10 trials and red eggs were rejected in six out of 11 trials (Table 1).
Natural nest linings represent a reliable cue against which robins could compare own versus foreign eggs; natural robin nests have low variation in raw reflectance spectra (supplementary material Fig. S1) and avian-perceived chromatic and achromatic visual contrasts across different nests (supplementary material Fig. S2). Furthermore, egg–egg contrasts between natural and artificial egg colors are positively related to egg–nest contrasts in robin nests, thus potentially confounding the interpretation of host–parasite egg rejection studies focusing on egg–egg contrasts only. Yet, our experimental manipulations of nest lining did not reliably alter egg rejection rates. Although we successfully altered the degree of egg–nest visual contrast both above and below natural levels (Table 1, Fig. 4; supplementary material Figs S3 and S5), we show here that the degree of perceivable color difference between foreign eggs and the nest background does not induce a predictable change in rejection rates of foreign eggs in the American robin. We minimally predicted egg–nest contrast would affect rejection rate of red eggs, which are rejected at intermediate rates in natural nests (Fig. 5C; Croston and Hauber, 2014a). Here, red eggs were rejected at intermediate rates irrespective of nest-lining color. Similarly, ejection rates of beige eggs and robin-mimetic eggs remained high and low, respectively, in all experimental nest-lining color conditions (Fig. 5C).
All same-color egg–nest combinations produced the lowest chromatic contrast (i.e. were the most cryptic) when compared with other nest types (e.g. beige egg–beige nest), while different egg–nest combinations consistently yielded high chromatic contrasts (Fig. 4A,C,E). However, the degree of egg–nest chromatic contrast did not have a significant effect on rejection rates in our linear regression analysis (Fig. 5A), and remained non-significant in our GLMM analysis (Table 2B). There was similarly no discernible pattern, nor significant predictive effect, of achromatic contrast on egg rejection (Fig. 4B,D,F, Fig. 5B,C, Table 2). Because only four natural robin eggs went missing throughout the course of this study, excluding the predation of the entire nest (see Materials and methods), we conclude that rejection responses by robins were specifically directed at experimental egg colors, and that manipulation of the nest lining did not induce rejection of the robins' own eggs.
Based on the consistent patterns of relative egg rejection rates between different artificial colors, irrespective of nest type (Fig. 5C, Table 2), we are therefore confident in rejecting the hypothesis that altering egg–nest contrast affects egg rejection in American robins. Unfortunately, robin identity, breeding age, prior experience with natural cowbird parasitism and/or prior experience with our own experimentation were unknown in this study. We also did not collect data on whether robins were flushed during experimental parasitism events, a factor which is known to affect egg rejection behavior in the congeneric European blackbird (T. merula; Hanley et al., 2015). Though age and experience may also influence egg rejection decisions in other brood–parasite host systems, with more experienced individuals typically more likely to correctly identify and reject parasitic eggs (e.g. Moskát et al., 2014a), it is not clear to what extent experience influences rejection decisions in American robins in our study population.
Evidence for parasitic egg crypsis via egg–nest color matching in other brood parasite systems is increasingly well documented in enclosed-nesting species. For example, some bronze-cuckoos (Chalcites spp.) have evolved dark egg pigmentation, which is cryptic in the domed nests of their hosts (Langmore et al., 2009). Manipulations could next establish whether host species and/or competing parasites respond differentially to parasitic eggs (Gloag et al., 2014) when experimentally illuminating the nest interior (Cassey, 2009; Honza et al., 2014) or when altering egg–nest contrasts independent of egg–egg contrast (this study). Whether cowbird eggs have a cryptic function in host nests has also not been studied in detail across different Molothrus cowbird–host systems (but see Mason and Rothstein, 1987; Siefferman, 2006). For example, cowbird eggs may be cryptic or difficult to see in the open cup nests of the eastern phoebe (Sayornis phoebe), which are often built under eaves/bridges or in caves and may be less illuminated than the open cups of robin nests; in turn, phoebes always accept cowbird parasitism (Hauber, 2003a; Peer and Sealy, 2004). Conversely, cowbird eggs have a greater avian-perceivable chromatic contrast against natural robin nest linings than do robin eggs themselves (Fig. 1A), making it unlikely that cowbird eggs are at all cryptic in robin nests.
That the rejection of foreign eggs does not depend on the degree of contrast between eggs and the nest lining (this study) provides support for earlier findings in hosts of egg–mimetic brood parasites that egg rejection is driven mechanistically by differences between foreign and own eggs (Cassey et al., 2008; Stevens et al., 2013; Moskat et al., 2014b). In contrast, cowbird eggs in robin nests are exceptional to this pattern: Croston and Hauber (2014a) showed that while robins' responses to artificial egg colors are generally predicted by chromatic JNDs differentiating foreign versus host eggs, artificial cowbird ground color-mimetic (beige) eggs are rejected in 100% of trials, despite their relatively low avian-perceivable chromatic difference from robin eggs (Fig. 1B). Our experimental manipulations showed that neither chromatic nor achromatic contrasts differentiating foreign eggs from nest linings were significant predictors of egg rejection – thus, cowbird egg rejection is likely the result of comparison between host and foreign eggs in robins (Croston and Hauber, 2014a). Future work should investigate the role of egg–nest contrast in egg rejection using ordinarily non-ejecting hosts.
We should note that higher chromatic contrasts do not necessarily correspond to more robust behavioral responses (Ham and Osorio, 2007). For example, chromatic JNDs differentiating artificial parasitic eggs and natural robin eggs do seem to drive rejection in robins. Specifically, cowbird-mimetic model eggs are rejected at the highest rates despite having relatively low chromatic contrast from robin eggs (Croston and Hauber, 2014a). In the present study, the two visual models showed similar patterns of chromatic contrasts between eggs and nest linings, and our supplementary analyses of physical distance using chromatic PCs largely confirm the outputs of both of our visual models. Despite the corroboration of our visual contrast analyses, we cannot assume that higher JND values in the supra-threshold range necessarily correlate with stronger behavioral responses.
Another caveat in this, and other studies based on the analysis of avian visual modeling data is that the magnitude of the chromatic difference (whether between eggs or between eggs and nests) is not always a linear means of predicting egg rejection (or any vision-dependent) behavior. Chromatic distance is but one component of broader sensory/perceptual (de la Colina et al., 2012) and cognition-dependent (Hauber and Sherman, 2001; Moskát and Hauber, 2007; Croston and Hauber, 2015) processes that ultimately result in the complex behavioral decision to accept or reject a parasitic egg. For example, there are a growing number of studies showing that perceptual difference alone does not fully explain patterns of egg rejection behavior (Moskát and Hauber, 2007; Moskát et al., 2010; Cassey et al., 2008; Stoddard and Stevens, 2011; Bán et al., 2013; Stevens et al., 2013; Croston and Hauber, 2014a).
Aside from specific perceptual/cognitive processes mediating egg rejection behavior, variation in the predictive power of avian visual models may be partly due to the physiological assumptions made within visual sensory models themselves. Specifically, visual models are based on a limited subset of bird species, including a handful of UVS oscines, none of which are common hosts of brood parasites (Grim et al., 2011; Aidala et al., 2012). Specifically, for this study we used parameters for the robin's visual system from the congeneric European blackbird (Turdus merula). This potentially confounds the degree to which we can model and understand host–parasite co-evolution to shape hosts' perceptual sensitivities. It is possible, then, that the visual models used in this and in previous studies do not accurately represent the sensory physiology of the American robin. Likewise, inter-individual differences in sensory physiology could confound our results, such that egg rejection reflects unaccounted-for differences in individual sensory physiology rather than at the level of decision making. Accordingly, within-species differences in sensory physiology have recently been described in the brown-headed cowbird (Fernández-Juricic et al., 2013). Future studies should endeavor not only to obtain and incorporate species-specific models of avian sensory physiology but also to describe the degree of inter-individual variation at both the behavioral and physiological levels.
We have shown here that egg–nest contrast is not a significant predictor of egg rejection by the American robin. Instead, egg rejection in robins is statistically explained, and likely perceptually driven, by differences between the hosts' own eggs and foreign egg colors. Future work should focus on improving visual models by incorporating physiologically appropriate, individual-specific cone densities/absorbance spectra, as well as nest site-specific egg, nest lining and ambient light availability data.
MATERIALS AND METHODS
All behavioral experiments were conducted in the vicinity of Ithaca, Tompkins County, NY, USA, from May to July of the 2013 breeding season. We located active robin nests (N=48), as defined by dry nest content, warm eggs and/or defense or attendance by adult robins, through focusing on suitable nest sites near human-built structures, as this species is highly commensal (Sallabanks and James, 1999). Nest sites were also located with the help of local citizens via advertising in community Listserv and businesses, and returning to locations with known robin nests from previous years (Croston and Hauber, 2014a,b).
After an active nest containing eggs was located, it was assigned in a balanced random procedure to an experimental nest type (one treatment per nest) and sequential egg treatments (one to three artificial eggs per nest). Robin nests were assigned one of three artificially colored nest linings, and paired with an artificial egg of one of the same three colors (see below for artificial egg and nest details; Fig. 2). Painted felt nest linings (see below for details) were inserted and affixed to the inner bottom lining of robin nests using fast-drying, non-toxic glue (Liquid Fusion®). An experimental egg was then added to the clutch without replacement (removal of one host egg), following methods used by Briskie et al. (1992) for American robins. Although egg replacement by cowbirds has been documented in one-third of parasitized yellow warbler (Setophaga petechia) nests (Sealy, 1992) and in most parasitized eastern phoebe (S. phoebe) nests (Hauber, 2003a), the addition of an experimental egg does not affect rejection rates in related European Turdus thrushes (Davies and Brooke, 1989; Grim et al., 2011) and allowed us to compare our new data with previous studies on robins (Rothstein, 1982; Briskie et al., 1992; Croston and Hauber, 2014a). Following the initiation of an experiment, we remained within sight of the nest to ensure that the new nest lining was not removed by adults upon their return to the nest. Nest lining removal occurred in only 3% of trials, and we returned and replaced the lining. If the experimental nest lining was removed by an adult robin three consecutive times, the experiment was abandoned at that nest. This occurred at only one nest site throughout the entire study.
All nests were checked daily after each experiment was initiated. Eggs were considered rejected if they were missing from a nest upon the return visit, unless the entire clutch was missing (presumed predation) or nestlings had begun to hatch (to avoid conflating egg rejection with eggshell removal, as in nest sanitation: Hauber, 2003c). If an artificial egg remained in the nest on the 5th day after addition, it was considered accepted (Rothstein, 1975; Briskie et al., 1992). In a previous study using the same focal robin population, all ejected model eggs were rejected within 1–4 days of being parasitized (mean 1.69 days; Croston and Hauber, 2014a), justifying a 5 day acceptance threshold. If a model egg remained in the nest through to hatching, we continued monitoring for up to 3 days post-hatching because of well-documented asynchronous hatching in robin broods (Sallabanks and James, 1999; Z.A. personal observations). Following the acceptance or rejection of a first experimental egg, a second egg of a different color was introduced. Up to three different eggs were introduced into robin nests in this way during the laying and incubation periods. The same egg color was not introduced repeatedly into the same nest. The experimental protocols followed in this study were approved by the Hunter College Institutional Animal Care and Use Committee, and all experiments conducted on private properties were done so with the express permission and mostly enthusiastic support from the landowners (Hauber, 2003a; Wagner et al., 2013).
Experimental eggs and nest linings
We constructed model cowbird eggs within the natural variation of natural brown-headed cowbird egg shape, size (21×16 mm) and mass (2.6–3.4 g) as documented near our field site in upstate New York, USA (Lowther, 1993; Croston and Hauber, 2014a,b; Z.A. personal observation). Model eggs were made from plaster of Paris, using the same silicone molds that were used by Croston and Hauber (2014a,b). Experimental nest-lining inserts were circular discs cut from white felt to fit the bottom of the robin's nest cup dimensions (mean disc diameter 94 mm) at our study site (Fig. 2). Eggs and felt were then painted red, natural cowbird ground color-mimetic (beige) or blue–green (robin-mimetic), using the same latex or acrylic paint as used in Croston and Hauber (2014a). We utilized the three egg and nest lining colors by considering the general shape and peak of their reflectance curves and by the relative photon catches of each avian cone photoreceptor (Endler, 1990; Endler and Mielke, 2005; Fig. 3), predicted to induce sharply different sensory responses of the UV-sensitive visual range of American robins (Aidala et al., 2012). We also chose these three egg/nest colors because they represented known behavioral variation in egg ejection responses in natural nests within the same population of robins: beige (100% rejected), red (64% rejected) and robin-mimetic (0% rejected; Croston and Hauber, 2014a). These extreme and intermediate egg color rejection rates allowed us to design a two-tailed experiment, whereby both increased, decreased and unchanged rejection rates would be predicted as a result of our experimental manipulations (Table 1).
As an internal experimental control for our invasive manipulations, we monitored the fate of naturally laid robin eggs in each clutch: a total of four robin eggs (at N=48 nests monitored, mean natural clutch size per nest=3.3 eggs) went missing during our study in 2013 (outside of complete nest predation events), implying that egg rejection responses were limited to experimental model eggs, and that own-egg rejection was not related to experimental manipulation of the nest lining. It was unclear in these instances whether these eggs were missing as a result of partial depredation events or failed rejections. In turn, as experimental controls for the nest lining manipulation, we contrasted the data from all of our experiments (single- and multiple-presentation nests in 2013 with the published behavioral egg rejection data in natural, non-manipulated robin nests from Croston and Hauber (2014a), barring those from the two excluded sites in the GLMM model described below and in supplementary material Table S2. We acknowledge the limitation that using the published egg rejection data from natural nests is at best a partial methodological control for our nest lining manipulations, and a full experimental control should conceivably include adding a see-through felt, or felt dyed with a natural nest reflectance matching color. Furthermore, those data were derived mostly during the 3 years prior to our experiments; however, egg rejection rates did not vary between years in our study population (Croston and Hauber, 2014a,b).
Spectral measurements and visual modeling
We obtained spectral measurements of natural robin (N=76) and cowbird (N=15) eggs by combining our dataset from 2013 with that of Croston and Hauber (2014a). In 2013, we also collected reflectance spectra from natural robin nest linings (N=19), as well as from our artificial eggs and nest backgrounds. Spectral measurements were taken with an Ocean Optics USB2000 Miniature Fiber Optic Spectrometer, connected to a laptop computer running OOIBase32 software, and using a UV-Vis DT mini-lamp light source (Ocean Optics, Inc. Dunedin, FL, USA) or an Ocean Optics Jaz spectrometer with UV-VIS light source (Ocean Optics, Inc.). All measurements were taken at a 90 deg angle to the egg or nest-lining surface. We took nine measurements each from individual nests, linings and eggs: three measurements each from the nests' upper inner cup, lower inner cup and bottom; and three measurements each from the blunt pole, middle portion and narrow pole of natural and artificial eggs (Croston and Hauber, 2015). The spectrometer was re-calibrated frequently, using the Ocean Optics WS-1 white reflectance standard and a dark reference made from a cardboard box, lined with black felt, and pierced to create a small hole for the probe (blocking any incident light; Igic et al., 2009; Igic et al., 2010; Croston and Hauber, 2014a). We averaged the nine spectra per egg/nest to generate a composite spectra profile for each egg and nest included in our visual modeling analyses. As a methodological check, we compared the mean achromatic and chromatic spectra of each nest lining area prior to compiling composite natural nest lining spectra.
Visual modeling analyses were conducted using AVICOL v.6 (Gomez, 2006). We applied a 15 nm triangular correction to raw spectra, available as a function within AVICOL, to attenuate and minimize the effect of spectrometer noise on the visual model. We ran a tetrachromatic receptor noise-limited color opponency model (Vorobyev and Osorio, 1998), assuming noise independent of the neural signal, and set the Weber fraction to 0.1 (Vorobyev et al., 1998; Igic et al., 2010; Croston and Hauber, 2014a,b, 2015). This type of opponency contrast model is preferable over avian visual models only accounting for properties of the photoreceptors themselves because such models do not agree with behavioral psychophysics data (see Vorobyev and Osorio, 1998). The model incorporates maximal absorbance and relative densities of each cone type as well as other physiological variables such as oil droplet and ocular media transmittance, allowing for analysis of both chromatic and achromatic contrasts (Vorobyev and Osorio, 1998; Vorobyev et al., 1998).
Because no photoreceptor absorbance or relative cone density data are currently available for robins, we approximated photoreceptor abundance and relative cone density based on published data of the closely related UVS European blackbird (Hart et al., 2000). The use of a congener Turdus may be suitable as the American robin is predicted to also possess a UVS SWS1 photopigment, based on the results of our molecular genetic analyses of the SWS1 opsin gene of the robin (Aidala et al., 2012). In this model, we set the relative cone densities (UVS: 1, SWS: 1.78, MWS: 2.21, LWS: 1.96) based on cone density data measured by Hart et al. (2000). Ambient light level irradiance data of a generic ‘open-cup’ nesting species were extracted from Avilés et al. (2008) and were kindly provided by Igic et al. (2012), as ambient light levels can affect both the risk of parasitism and parasitic egg detection (Langmore et al., 2005; Muñoz et al., 2007; Avilés, 2008; Honza et al., 2011).
Achromatic contrasts were calculated by summing MWS and LWS cone spectra (Osorio and Vorobyev, 2005,, 2008; Gomez, 2006), as their combined sensitivities are thought to be comparable to those of the non-color-sensitive rod and double cone (Osorio et al., 1999) photoreceptors across avian taxa (Hart et al., 1998,, 2000; Igic et al., 2009). Using the model parameters described above, AVICOL generated separate chromatic and achromatic perceptual distances between two objects as JNDs; a calculated JND value greater than 1.0 suggests that two stimuli are discriminable from one another, while a JND less than 1.0 suggests that they are not (Gomez, 2006).
Although our visual modeling is based on the known retinal physiology of a closely related UVS Turdus species, the European blackbird, we augmented our visual modeling analyses by also computing a VS visual model as differences in retinal physiology between the European blackbird and the American robin are unknown. In this second model, we used the cone absorbance spectra from the VS rock pigeon (Columba livia; Bowmaker et al., 1997; Vorobyev and Osorio, 1998) and the relative cone densities (UVS: 1, SWS: 1.9, MWS: 2.2, LWS: 2.1) of the peafowl (Pavo cristatus) as measured by Hart (2002). All other visual modeling parameters remained the same as in our UVS visual model.
Using these distance scores, we examined the relationship between (a)chromatic distance and rejection rates between experimental eggs and nest linings using linear regression. These additional analyses of our spectral data allowed for increased explanatory power of our behavioral results as they relate to visual contrast, and complemented the statistical and qualitative conclusions drawn from our JND analyses.
In order to confirm that our natural nest lining composite spectra were representative of all three nest areas measured, and not biased towards one nest area over the others, we compared avian-perceived (a)chromatic differences between natural nest lining areas (upper inner cup, lower inner cup and bottom). No achromatic or chromatic within-nest area comparison was higher than 1.65 JNDs. Because the visual contrasts between nest areas were so low, we used the composite natural nest lining spectra including the nine measurements from the three nest areas in all analyses. In order to show that there exists a methodological confound between egg–egg chromatic contrast and egg–nest chromatic contrast, we conducted non-parametric Mann–Whitney U-tests between natural robin and cowbird eggs against conspecific natural robin eggs and natural robin nest linings, respectively. We also conducted a linear regression analysis to test the relationship between egg–natural robin egg and egg–natural robin nest lining chromatic contrasts using artificial egg stimuli sourced from Croston and Hauber (2014a). We next confirmed that our nest lining manipulations resulted in experimental alteration of chromatic and achromatic contrasts between eggs and nests, using non-parametric Kruskal–Wallis rank sums tests and post hoc pairwise comparisons following the Wilcoxon method. Prior to analysis, we randomized our comparisons such that only one egg–nest combination was used in each type of egg–nest contrast comparison.
We examined the statistical relationship between (a)chromatic egg–nest and egg–egg contrasts and rejection rates for both natural and artificial eggs using linear regression analyses. A non-parametric 2-way Friedman ANOVA was run to test whether nest color affected egg rejection behavior across the different egg color stimuli. To further examine the role of egg–egg and egg–nest contrasts in parasitic egg rejections, we fitted binomial GLMMs (with accept/reject as the outcome variables) using Firth-adjusted bias estimates to determine the degree to which nest color influenced egg rejection behavior. In these models, we included egg color, nest color, nest site, experimental date, presentation order and natural clutch size as predictor variables (Table 2; supplementary material Table S2). We first examined the known effects of the same individual female robins' tendencies to consistently accept or reject differently colored eggs (Croston and Hauber, 2014b) across nest treatments by nesting site within nest color (supplementary material Table S2). In these models, we only included nest sites where more than one egg had been presented. After controlling for individual females' tendencies to accept or reject experimental eggs irrespective of egg/nest treatments (supplementary material Table S2), we included rejection rates in natural nests (Croston and Hauber, 2014a) for the three egg colors used in this study. We included the same predictors listed above except nest site. Lastly, we included chromatic and achromatic JNDs between eggs and nests in the GLMM model to explicitly test the role of avian-perceived contrasts in egg rejection frequencies. Post hoc analyses of significant predictors in this final GLMM were run using chi-square tests. All analyses were run using JMP v. 10 (SAS Institute, Inc., Cary, NC, USA), Statview 5.1 (SAS Institute, Inc.), and GraphPad Prism v. 6 (GraphPad Software, Inc., La Jolla, CA, USA). Figures were compiled and edited using Adobe Creative Suite 5 (Adobe Systems, Inc., San Jose, CA, USA).
We would like to thank the kind residents of Tompkins County, whose generosity made this study possible. We thank Michael Webster, and his lab, at the Cornell University Laboratory of Ornithology for their hospitality and assistance with field work. We would also like to thank Esteban Fernandez-Juricic, James Gordon, Tomáš Grim, Cheryl Harding, Brani Igic, David Lahti, Csaba Moskát, Arnon Lotem, M. Cassie Stoddard and Marlene Zuk for discussions.
Z.A. and M.E.H. designed this study. Z.A., R.C., J.S. and L.T. conducted the experiments and collected data. Z.A. and M.E.H. analyzed the data and Z.A. wrote the first draft of the manuscript, with all authors contributing to critical interpretation of data and results, and the writing and revision of subsequent drafts of the manuscript.
Funding for this study was provided by the Human Frontiers Science Program [M.E.H.], the PSC-CUNY grant [M.E.H.], the Provost's Office of Hunter College [M.E.H.], the Vice-Chancellor's Office for Research of the City University of New York [M.E.H.], the Raab Presidential Fellowship Program at Hunter College [J.S.], the McNair Scholars Program at Hunter College [L.T.], the Animal Behavior Society [Student Research Grants to Z.A. and R.C.], and the National Science Foundation [GK-12 STEM Fellowship to Z.A.].
The authors declare no competing or financial interests.