Predator perception across space and time: relative camouflage in a colour polymorphic salamander

Abstract

The ecological and evolutionary dynamics maintaining colour polymorphisms are important because of the relationship between polymorphism and niche evolution, species formation and the maintenance of genetic variation. Colour polymorphisms are often hypothesized to be maintained by interactions with visual predators, either via apostatic selection or by being differentially camouflaged in different habitats, seasons or populations. The Eastern Red-backed Salamander (Plethodon cinereus) contains two common colour morphs, a striped morph and a lead (unstriped) morph. Presently, it is not known which morph is better camouflaged, or the degree to which camouflage is context dependent. To address this question, we studied three populations that vary in morph frequency over two seasons. In each population, in the fall and the spring, we collected reflectance measurements from salamanders, and the distribution of background colours. Given avian, snake and mammalian visual models, we found that both salamander and background colours differed by population and season. In general, we found striped morphs were better camouflaged than lead morphs against most background types; however, the level of camouflage was dependent on population, season and body region. These results highlight the value of considering spatial and temporal dimensions when testing hypotheses regarding the origin and maintenance of polymorphisms.

contrast, Eastern Red-backed Salamander, morph, Plethodon cinereus, reflectance, spectrometer, visual model

INTRODUCTION

Within populations, individuals that effectively blend into the background of their natural environment have a lower probability of being discovered by visual predators (Cott, 1940; Ruxton, Sherratt & Speed, 2004). However, camouflage is context dependent, and distinct colours can vary in their relative conspicuousness as a function of where and when predation occurs, as well as between the divergent visual systems of diverse predators (Cain & Sheppard, 1954; Endler, 1980, 1990; Endler & Greenwood, 1988). Thus, co-evolutionary dynamics between predators and prey often vary over space and time, promoting the evolution of phenotypic polymorphisms within and between populations (Thompson, 2005; Bond, 2007; Klomp et al., 2014).

Colour polymorphism is the presence of two or more distinct, genetically determined colour morphs within a single interbreeding population (Ford, 1945; Huxley, 1955). How and why distinct morphs evolve is an active field of study, but visually mediated predation is often postulated to maintain many colour polymorphisms (Punzalan, Rodd & Hughes, 2005; Fitzpatrick, Shook & Izally, 2009; Farallo & Forstner, 2012). One commonly hypothesized mechanism for the maintenance of colour morphs within populations is a form of negative frequency-dependent selection known as apostatic selection, in which common morphs are preyed upon more frequently and rarer morphs are preyed upon less frequently as a consequence of the specific search image of a predator (Clarke, 1962, 1969; Greenwood, 1984; Allen, Raison & Weale, 1998; Bond, 2007). Similarly, neophobia, the fear of unfamiliar stimuli, can result in negative frequency-dependent selection (Sherratt, 2011; Aubier & Sherratt, 2015; Crane & Ferrari, 2017). Variation in selection in space and time, combined with gene flow among populations, can also maintain a polymorphism. For example, alternative colour morphs may be adapted to match different backgrounds (Endler, 1980; Bond & Kamil, 2006), and may differ in their relative camouflage as a function of habitat, season and light conditions (Endler, 1990). Consequently, the processes that maintain colour morphs are likely to occur in a geographic mosaic (Thompson, 2005; Calsbeek et al., 2012).

The Eastern Red-backed Salamander, Plethodon cinereus, has a widespread distribution in northeastern North American forests and is the most abundant terrestrial vertebrate in the northeastern United States (Burton & Likens, 1975). There are two common colour morphs of P. cinereus, a striped morph and a lead (or ‘unstriped’) morph (Highton, 1959). The striped morph has a red stripe running down the centre of a black dorsum, whereas the lead morph lacks the red dorsal stripe. The morphs are distinct in dorsal coloration and were originally described as separate species (Highton, 1959). Many populations of P. cinereus are polymorphic, with varying morph frequencies, or are monomorphic for the striped morph; however, monomorphic lead populations are rare (Moore & Ouellet, 2014, 2015; Hantak et al., 2015). The striped/lead dimorphism is common within the genus Plethodon, with at least ten species exhibiting this polymorphism (Highton, 2004).

Despite the number of studies on the striped and lead colour morphs of P. cinereus (reviewed in Anthony & Pfingsten, 2013), we know little with regard to the function of the distinct dorsal colours (but see Fitzpatrick et al. 2009; Kraemer, Serb & Adams, 2016). Many elements of the behaviour and ecology of the morphs appear to differ, including metabolic rate, temperature associations, diet, stress levels, tail breakage rates, territoriality and disease prevalence (Lotter & Scott, 1977; Moreno, 1989; Venesky & Anthony, 2007; Reiter, Anthony & Hickerson, 2014; Paluh et al., 2015; Venesky et al., 2015). In addition, the morphs have been shown to mate assortatively by colour in at least one population (Anthony, Venesky & Hickerson, 2008; Acord, Anthony & Hickerson, 2013). These studies demonstrate that the striped morph is usually the competitive dominant. However, no studies on P. cinereus have examined multiple populations over multiple seasons and years, which is essential for quantifying whether differing selection across space and time contributes to the maintenance of the polymorphism.

A number of evolutionary processes could operate to maintain the polymorphism, such as social selection, sexual selection and predator–prey interactions. In this study, we tested hypotheses about how the colour polymorphism in P. cinereus influences relative camouflage to potential predators. To investigate these questions, we used a spectrometer to quantify reflectance values of salamander body regions, as well as the distribution of colours on the forest floor. The degrees of relative camouflage of the two morphs were tested under divergent visual systems (avian, snake, mammalian), seasonal variation (fall vs. spring) and lighting conditions (direct sunlight vs. forest shade). First, we evaluated whether dorsal salamander colours and background (leaf litter and soil substrate) colours vary across seasons and populations. We predicted morph and soil colours would remain consistent, but leaf litter colours would vary by season and population. Second, we tested whether different body regions (dorsum, venter and side) of the morphs were differentially camouflaged to diverse vertebrate predators. Salamander sides may be viewed by snakes and small mammals, and examining the venter may be of ecological relevance because individuals within Plethodon occasionally flip onto their backs when threatened (Brodie, 1977, pers. observ.). Finally, we quantified whether seasonal and spatial variation in background colours alters detectability of the morphs. We predicted striped morphs would be more camouflaged than lead morphs in the fall due to a greater abundance of newly fallen leaves with a high proportion of red colours, and that lead morphs would be more camouflaged against relatively decomposed spring leaf litter and soil substrates compared to striped morphs.

METHODS

Salamander sampling and reflectance measurements

Individuals of P. cinereus were collected from three populations in northern Ohio that vary in morph frequency: Squire Valleevue Farm (Squire; 100% striped), Manatoc Scout Reservation, directly adjacent to Cuyahoga Valley National Park (CV; 80% striped) and the Heineman property on South Bass Island (SBI; >99% lead; Hantak et al., 2015). In fall 2014 and spring 2015, 15 individuals/morph (120 total) were collected from each population. All salamanders were euthanized with tricaine methanesulfonate (MS-222). Immediately following, ten reflectance measurements were recorded along the mid-sagittal section of the dorsum of each salamander, encompassing the area where a dorsal stripe could be present (Fig. 1). In addition, five reflectance measurements were recorded along the mid-sagittal region of the venter, as well as three measurements from the side of each salamander (spring field season only). Spectral reflectance measurements were recorded with an Ocean Optics Jaz UV/Vis spectrometer (Model EL 200) with a Jaz-PX Xenon light source and a QR400-ANGLE-UV reflectance probe. The light probe was fitted with a Mikopark CSH-45° holder to reduce specular reflectance, standardize reflectance measurements and exclude ambient light (Endler, 1990). Each reflectance spectrum was measured in 1 nanometre (nm) intervals from 300 to 700 nm. Dorsal, ventral and side reflectance measurements from each salamander were averaged and smoothed by body region for each individual.

Figure 1.

(A), representative lead and striped morphs of Plethodon cinereus. (B–D) spectral reflectance of P. cinereus dorsal colours by season, population and morph. (B) striped and lead morphs from CV, (C) striped morphs from Squire and (D) lead morphs from SBI. Lead morphs nearly overlap completely. (E–G) spectral reflectance of substrate colours by season and population. (E) CV, (F) Squire, (G) SBI. For all graphs, the centre line represents the mean spectral reflectance, and the shading around the line represents the standard error.

At each of the three sites, 100 reflectance measurements were collected from spring leaf litter and fall leaf litter (600 measurements total) within 1 m of where a salamander was located, and from soil under cover objects in the spring where salamanders were found (5–10 soil measurements/salamander, 300 measurements total; Fig. 1). These substrate types represent backgrounds on top of which P. cinereus could be viewed by a predator. All spring and fall leaf litter measurements were obtained 2 days following rainfall. Individuals of P. cinereus forage in leaf litter during moist conditions (Jaeger, 1980); therefore, this represents an optimal time for visual predators to discover salamanders (Kuchta, 2005; Venesky & Anthony, 2007).

To describe colour, we calculated mean brightness, hue and chroma (hereafter termed ‘colour’) from reflectance measurements (Endler, 1990; Andersson & Prager, 2006; Kemp et al., 2015). Brightness is the total intensity of light that is reflected; hue roughly corresponds to the verbal definition of colour and was calculated as the wavelength at peak reflectance; and chroma is a measure of colour saturation and was calculated as the relative difference between the maximum and minimum reflectance while taking into account mean brightness (Maia et al., 2013). These measures of colour are predator independent and were used to quantify colour variability. Colour variables were obtained using the R (version 3.2.2, R Core Team, 2015) package pavo, version 0.5.2 (Maia et al., 2013).

Visual models

Visual models include reflectance values from the body regions of salamanders, background reflectance, predator spectral sensitivities and irradiance. To quantify how well potential predators discriminate between the morphs of P. cinereus given a set of background colours and ambient light conditions, we applied the visual model developed by Vorobyev et al. (1998). Colour distances were obtained by calculating chromatic (ΔS) and achromatic (ΔL) contrasts, which correspond to colour (hue and saturation) and brightness (luminance), respectively. These calculations estimate the contrast between an object and a background in units of just-noticeable differences (JNDs), where a value of 1 approximates the minimum difference between an object and the background that is detectable to a given predator (Vorobyev et al., 1998). Due to variation in ambient light conditions, the distance between the viewer and target, or the length of time an object is viewed by a predator, a JND of 1 is not an absolute threshold, but is useful as an approximate criterion (Kemp et al., 2015).

To quantify the level of camouflage among colour morphs and populations against each background type, the spectral sensitivities of the tetrachromatic blue tit (Parus caeruleus; Hart et al., 2000), trichromatic common garter snake (Thamnophis sirtalis; Sillman et al., 1997) and the dichromatic thirteen-lined ground squirrel (Ictidomys tridecemlineatus; Jacobs, Neitz & Crognale, 1985) were used. The common garter snake is a natural predator of P. cinereus, but the blue tit and thirteen-lined ground squirrel are not. However, the blue tit visual system is well characterized and is similar to corvids, which are common predators of North American salamanders (Murray, Pearl & Bury, 2005; Kuchta, Krakauer & Sinervo, 2008). Similarly, the thirteen-lined ground squirrel has a visual system that resembles several mammalian mesopredators (Jacobs et al., 1985), which also prey on salamanders (Brodie, Nowak & Harvey, 1979; Kuchta, 2005; Anthony & Pfingsten, 2013; Kraemer et al., 2016).

Finally, two irradiance measures were used in the visual models, forest shade and standard daylight (Endler, 1993). Contrast values did not differ with irradiance type; thus, we only report values calculated using forest shade irradiance. Full visual model and contrast value calculations can be found in Maia et al. (2013).

Statistical analyses

Salamander colour variables as well as seasonal and population differences in substrate colours were compared using analyses of variance (ANOVAs; Supporting Information, Table S1) and multivariate analyses of variance (MANOVAs). We used a full factorial ANOVA to test for colour morph, seasonal and population differences in chromatic and achromatic contrast, which were calculated using avian, snake and mammalian visual models against the three substrate types. We ran separate models for chromatic and achromatic contrasts, and for avian, snake and mammalian visual models. Significance was assessed with Tukey’s HSD post hoc tests. Tetrahedral (four cones), trichromatic (three cones) and dichromatic (two cones) colour space models, which are measures of colour (hue and saturation) overlap, were created to illustrate the visibility of morphs against background types given avian, snake and mammalian predators (Goldsmith, 1990; Stevens, Stoddard & Higham, 2009). All statistical analyses were conducted in R version 3.2.2.

RESULTS

Variation in salamander and background colours

Combined colorimetric variables of salamander dorsal measurements differed between seasons and populations (F_3,112 = 4.9, P < 0.001; Fig. 1). In addition, the colour of salamander sides differed among populations (F_3,56 = 4.7, P < 0.001), and the colour of ventral measurements differed by season and population (F_3,112 = 2.6, P = 0.008). Accordingly, salamanders from fall and spring and among populations were kept separate for subsequent analyses. Spring and fall leaf litter differed by season and population (F_2,114 = 20.2, P < 0.001; Fig. 1), and soil substrate differed among populations (F_2,57 = 7.7, P < 0.001; Fig. 1). Thus, spring leaf litter, fall leaf litter and soil substrate from each population and season were also kept separate for subsequent analyses.

Salamander camouflage

From the perspective of avian and snake predators, the dorsum, side and venter of all salamanders were discriminable in chromatic and achromatic colour space against all background types (JND > 1; Fig. 2; Supporting Information, Tables S1–S3). From the perspective of a mammalian predator, the chromatic contrast of the dorsum of the lead morphs was discriminable against all background types (JND > 1), whereas striped morphs did not stand out against any background type (JND < 1; Fig. 2C). Dorsal achromatic contrast was discriminable against all background types for both colour morphs (Supporting Information, Fig. S1C). Chromatic and achromatic contrasts of all salamander sides and venters were discernable against all background types (Supporting Information, Figs S2, S3).

Figure 2.

Dorsal chromatic contrast values for the avian (A), snake (B) and mammalian (C) visual systems. Bars show means (±SE). Contrast values that lie below the grey horizontal line represent groups that are indistinguishable to the predator. Y-axes values vary by plot.

Seasonal and spatial variation in camouflage

Dorsal contrasts: avian visual model

Dorsal contrasts using the avian visual model differed between background type and population for both chromatic contrast (F_6,168 = 24.57, P < 0.001, adj. R² = 0.64) and achromatic contrast (F_6,168 = 25.24, P < 0.001, adj. R² = 0.67). Pairwise comparisons revealed that striped individuals exhibited lower chromatic contrast compared to lead morphs against spring and fall leaf litter, but there was no difference in morph conspicuousness against soil substrate (Fig. 2A; Table 1A). Achromatic contrast varied by population (Supporting Information, Fig. S1A; Table 1A). Tetrahedral colour space plots, which demonstrate the visual overlap between morphs and background types at each population, illustrate the similarity between striped individuals and spring and fall leaf litter, whereas both morphs show similar overlap with the soil substrate (Fig. 3A).

Table 1.

Tukey’s HSD post hoc test P-values for dorsal Plethodon cinereus (A) avian, (B) snake and (C) mammalian contrast values

	Fall leaves				Spring leaves				Soil substrate
	CV striped	CV lead	Squire striped	SBI lead	CV striped	CV lead	Squire striped	SBI lead	CV striped	CV lead	Squire striped	SBI lead
A
CV striped	–	1.000	<0.001	0.713	–	<0.001	0.655	<0.001	–	<0.001	0.980	0.134
CV lead	<0.001	–	<0.001	0.727	0.002	–	<0.001	<0.001	1.000	–	<0.001	<0.001
Squire striped	0.106	<0.001	–	<0.001	1.000	0.032	–	0.086	0.538	0.276	–	0.883
SBI lead	<0.001	<0.001	<0.001	–	<0.001	0.014	<0.001	–	0.877	0.639	1.000	–
B
CV striped	–	0.527	<0.001	1.000	–	<0.001	0.726	<0.001	–	<0.001	0.675	0.998
CV lead	1.000	–	0.279	0.842	0.590	–	<0.001	<0.001	1.000	–	<0.001	<0.001
Squire striped	0.006	<0.001	–	0.001	1.000	0.553	–	<0.001	1.000	1.000	–	0.996
SBI lead	0.590	0.135	0.816	–	0.001	0.512	0.001	–	0.820	0.384	0.870	–
C
CV striped	–	0.003	<0.001	0.031	–	<0.001	0.684	<0.001	–	<0.001	0.237	0.971
CV lead	<0.001	–	1.000	1.000	<0.001	–	<0.001	<0.001	<0.001	–	0.454	<0.001
Squire striped	1.000	<0.001	–	0.944	1.000	<0.001	–	<0.001	1.000	<0.001	–	0.004
SBI lead	<0.001	<0.001	<0.001	–	<0.001	0.039	<0.001	–	<0.001	0.972	<0.001	–

Comparisons below the diagonal of each substrate type are chromatic contrasts, whereas values above the diagonal are achromatic contrasts. Significant comparisons are in bold.

Figure 3.

(A) tetrahedral colour space plots representing the visual perspective of an avian predator. Hue is estimated from the angle of each point to the blue (short; s), green (medium; m), red (long; l) and UV (u) cone colour channels (x-y-z axes), and saturation is measured as the distance from the achromatic origin (blue centre star) to each individual point. (B) trichromatic colour space plots representing the visual perspective of a snake predator. Hue is estimated by short (s) and medium (m) and long (l) cone colour channels, and saturation is measured as the distance from the achromatic origin (blue centre star). (C) dichromatic colour space plots representing the visual perspective of a mammalian predator. Hue is estimated by short (s) and medium/long (m/l) cone colour channels, and saturation is measured as the distance from the achromatic origin (blue centre star). For (A–C), each plot represents a population, and each point represents the colour of an individual salamander or an average of 5–10 substrate measurements to a visual predator.

Dorsal contrasts: snake visual model

Dorsal contrasts using the snake visual model differed between background type and population for both chromatic contrast (F_6,168 = 6.24, P < 0.001, adj. R² = 0.32) and achromatic contrast (F_6,168 = 26.18, P < 0.001, adj. R² = 0.63). Pairwise comparisons revealed that striped morphs exhibited lower chromatic contrast compared to lead morphs from SBI against spring leaf litter (Fig. 2B; Table 1B). Striped morphs from Squire were more camouflaged than striped and lead morphs from CV against fall leaf litter (Fig. 2B; Table 1B). There was no difference in conspicuousness between morphs or populations against soil substrates (Fig. 2B; Table 1B). Dorsal achromatic contrast varied by background type and population (Supporting Information, Fig. S1B; Table 1B). Trichromatic colour space plots illustrate similar colour overlap between striped and lead morphs and soil substrate (Fig. 3B). However, striped morph colour overlaps more with spring leaf litter, and striped morphs from Squire overlap more with fall leaf litter, whereas lead morph colours overlap less with these background colours (Fig. 3B).

Dorsal contrasts: mammalian visual model

Dorsal contrasts using the mammalian visual model differed by background type and population for both chromatic contrast (F_6,168 = 14.70, P < 0.001, adj. R² = 0.81) and achromatic contrast (F_6,168 = 21.03, P < 0.001, adj. R² = 0.61). Pairwise comparisons revealed that striped morph dorsal chromatic contrast was lower than lead morph contrast (Fig. 2C; Table 1C). Striped morphs were indistinguishable from the backgrounds they appeared against, but lead morphs were not (Fig. 2C). The achromatic contrast of striped and lead morphs varied by population (Supporting Information, Fig. S1C; Table 1C). Dichromatic colour plots demonstrate the high degree of colour overlap between striped morphs and all background types, whereas lead morphs did not overlap with any background type (Fig. 3C).

Side and ventral contrasts: avian visual model

Chromatic contrast of salamander sides differed by background type and population (F_3,112 = 2.97, P = 0.035, adj. R² = 0.69); however, achromatic contrast did not differ (F_3,112 = 1.64, P = 0.183, adj. R² = 0.34). Pairwise comparisons show that against soil substrate and spring leaf litter chromatic contrasts of lead morph sides from SBI were significantly less camouflaged than lead morph sides from CV and all striped morphs (Supporting Information, Fig. S2A, Table S2A). No other comparisons of salamander sides differed.

The chromatic contrast of salamander venters differed by background type and population (F_6,168 = 83.64, P < 0.001, adj. R² = 0.94), as did achromatic contrast (F_6,168 = 50.19, P < 0.001, adj. R² = 0.71). Overall, pairwise comparisons of the ventral chromatic and achromatic contrasts of lead morphs from SBI were less camouflaged against soil substrate and spring leaf litter (Supporting Information, Fig. S3A, Table S3A). Conversely, against fall leaf litter, the venters of lead morphs from SBI were more camouflaged than lead morphs from CV and striped morphs (Supporting Information, Fig. S3A, B, Table S3A).

Side and ventral contrasts: snake visual model

Chromatic contrasts of salamander sides differed by background type and population (F_3,112 = 3.57, P = 0.016, adj. R² = 0.64); however, achromatic contrast did not differ (F_3,112 = 2.48, P = 0.065, adj. R² = 0.46). Pairwise comparisons showed that against soil substrate the chromatic contrasts of lead morph sides from SBI were significantly less camouflaged than lead morphs from CV and all striped morphs (Supporting Information, Fig. S2C, Table S2B). Lead morph sides from SBI were less camouflaged against spring leaf litter compared to lead and striped morphs from CV (Supporting Information, Fig. S2C, Table S2B). Striped morphs from Squire were less camouflaged than striped morphs from CV against spring leaf litter (Supporting Information, Fig. S2C, Table S2B).

Chromatic contrasts of salamander venters differed by background type and population (F_6,168 = 47.54, P < 0.001, adj. R² = 0.84), as did achromatic contrasts (F_6,168 = 69.42, P < 0.001, adj. R² = 0.81). In general, pairwise comparisons revealed that the ventral chromatic and achromatic contrasts of lead morphs from SBI were less camouflaged against soil substrate and spring leaf litter (Supporting Information, Fig. S3C, D, Table S3B). However, the venter of lead morphs from SBI was more camouflaged than lead morphs from CV and striped morphs against fall leaf litter (Supporting Information, Fig. S3C, D, Table S3B).

Side and ventral contrasts: mammalian visual model

The chromatic contrasts of salamander sides differed by background type and population (F_3,112 = 4.51, P = 0.005, adj. R² = 0.66), as did achromatic contrasts (F_3,112 = 6.40, P < 0.001, adj. R² = 0.63). In general, pairwise comparisons of salamander side chromatic and achromatic contrasts revealed that lead morphs from SBI were conspicuous against soil substrate and spring leaf litter, but there was no difference in chromatic contrast between lead morphs from CV and SBI against soil substrate (Supporting Information, Fig. S2E, F, Table S2C).

Both chromatic (F_6,168 = 75.22, P < 0.001, adj. R² = 0.93) and achromatic (F_6,168 = 92.00, P < 0.001, adj. R² = 0.86) contrasts of salamander venters differed by background type and population. Pairwise comparisons of ventral chromatic and achromatic contrasts showed that lead morphs from SBI were less camouflaged against soil substrate and spring leaf litter (Supporting Information, Fig. S3E, F, Table S3C). The chromatic and achromatic contrasts of the venter of lead morphs from SBI against fall leaf litter were more camouflaged than lead morphs from CV and striped morphs (Supporting Information, Fig. S3E, F, Table S3C).

DISCUSSION

Geographic variation is an intrinsic part of the evolutionary process, driven by differences in abiotic and biotic conditions, including co-evolving interactions among morphs (Thompson, 2005). However, despite steady interest in the biology of colour polymorphisms, we still have a poor understanding of how spatial and temporal processes impact the evolutionary dynamics of colour morphs within and among populations (Gray & McKinnon, 2007; Corl et al., 2010; Hugall & Stuart-Fox, 2012), including how commonly morphs are exposed to divergent selection pressures (Gosden & Svensson, 2008; Calsbeek et al., 2012). With reflectance measurements of the polymorphic salamander, P. cinereus, and three background types, we examined whether spatial and temporal differences in coloration influenced visual perception of the morphs by avian, snake and mammalian predators. We found that salamander and background colours varied across populations and seasons. An unexpected finding in our study was that, in general, lead morphs were more conspicuous to visual predators. However, we also found there was spatial and temporal variation in the relative degree of morph camouflage.

In P. cinereus, the striped/lead polymorphism is genetic with a simple genetic architecture (Highton, 1959, 1975). However, individual colours differ within and among populations, which can be due to genetic differences, environmental differences or an interaction between genes and the environment. Using reflectance measurements of P. cinereus, we found that the average coloration of populations varied between spring and fall. The striped morph expressed a brighter red coloration in the fall, whereas the lead morph was brighter in the spring. We did not anticipate these seasonal alterations in colour; however, a study by Kraemer, Kissner & Adams (2012) demonstrated gradual colour change in the striped and erythristic (entirely orange-red) morphs of P. cinereus in captivity. The authors suggested colour change may have occurred due to diet, stress or natural seasonal changes; our study suggests that natural seasonal changes may have played a role. Seasonal changes in colour are a common biological phenomenon, and have been documented in many organisms, including insects (Tauber, Tauber & Masaki, 1986), birds (Delhey et al., 2006), mammals (Aldous, 1937; Caro, 2009), reptiles (Johnston, 1994; Boback & Siefferman, 2010) and frogs (Wente & Phillips, 2003).

In our study of camouflage, we found that the striped morphs were more camouflaged in dorsal coloration compared to lead morphs to the avian, snake and mammalian predators. However, there was spatial variation in the degree of morph camouflage. For example, the avian and snake visual models were unable to discriminate between striped and lead morphs against soil substrate. In addition, striped morphs from Squire were more camouflaged to snakes than striped morphs from CV or lead morphs against fall leaf litter.

Across seasons and populations, our results indicate that when visual predators are the agents of selection, striped morphs benefit from better camouflage relative to lead morphs, suggesting that relative camouflage may not play a role in the maintenance of this polymorphism. This raises the question of why the striped morph does not go to fixation. It may be that predator-mediated selection does not play an important role in the maintenance of the colour polymorphism, but rather the polymorphism is under, for example, strong social or sexual selection. Another possible explanation is that apostatic selection or neophobia plays a role in maintaining the polymorphism. For example, using clay model replicas of striped and lead morphs of Plethodon in an experimental setup, Fitzpatrick et al. (2009) found that rare morphs benefitted from lower rates of avian predation. By contrast, another clay model study of predation on P. cinereus did not find evidence for apostatic selection by mammalian predators (Kraemer et al., 2016). Such incongruence may be a by-product of different experimental methods. For instance, the salamander replicas used by Fitzpatrick et al. (2009) contained a food reward on the underside, and were conducted in an open field, whereas Kraemer et al. (2016) deployed clay models without a reward in forested habitats. Additional studies within a geographic framework may aid our understanding of the role of predator-mediated selection in the maintenance of the striped/lead polymorphism.

In our study, much of the variation in contrast values among populations was due to differing substrate colours in different populations (Figs. 1, 3). Soil composition in Ohio consists of 12 distinct series, which are defined by combinations of soil attributes. Our CV and Squire sites are in soil series 8 (Mahoning-Canfield-Rittman-Chili), whereas SBI is in soil series 1 (Hoytville-Nappanee-Paulding-Toledo; ODNR 2017). Common tree species at the three study locations vary as well. Sugar maple (Acer saccharum) is common at all three sites, but American beech (Fagus grandifolia) and tulip poplar (Liriodendron tulipefera) are common at CV and Squire. Red oak (Quercus rubra) is also common at CV, whereas shagbark hickory (Carya ovata) and red maple (Acer rubrum) are abundant at Squire. Conversely, common hackberry (Celtis occidentalis), American basswood (Tilia americana) and the invasive Amur honeysuckle (Lonicera maackii) dominate SBI. These different assemblages of tree species create spatial variation in leaf litter colours.

Our study is not the first to find that morph-specific predation risk is highly dependent on background colours (Forsman et al., 2011; Karpestam, Merilaita & Forsman, 2014; Kraemer & Adams, 2014). For instance, Karpestam, Merilaita & Forsman, (2013) found predator perception of the divergent colour morphs of the Pygmy Grasshopper (Tetrix subulata) was strongly dependent on whether the habitat was burnt, unburnt or intermediate. They also found a correlation between predator detection rates and colour morph frequencies. In P. cinereus, striped morphs are typically at a higher frequency in polymorphic populations, and are more commonly fixed within populations (Moore & Ouellet, 2015), suggesting specific settings are required for the maintenance of the lead morph. For example, the fitness of lead morphs may be more dependent on behavioural or ecological attributes. A study by Fisher-Reid et al. (2013) found that lead morphs from monomorphic populations on Long Island, New York, have one more costal groove than striped morphs, and elongation in salamanders is associated with increased fossoriality (Wake, 1966). In addition, other factors, such as temperature, parasitism and competition between the morphs, have been shown to be correlated with morph frequencies in several taxa (Gibbs & Karraker, 2006; McLean & Stuart-Fox, 2014). With the wide range of documented differences in ecology, behaviour and coloration between the morphs of P. cinereus, it may be that multiple evolutionary processes interact to contribute to the maintenance of this polymorphism (Merilaita, 2001; McLean & Stuart-Fox, 2014).

CONCLUSIONS

How colour morphs are maintained within and among populations is a long-standing question in evolutionary biology (Ford, 1945; Huxley, 1955). Recent studies of geographic variation in colour polymorphic species have aimed to elucidate the role of geography in divergence, including species formation (Corl et al., 2010; Ozgo, 2011; Hugall & Stuart-Fox, 2012; McLean & Stuart-Fox, 2014). Our study demonstrates the importance of studying polymorphism in a geographic framework. A single population at a single point in time provides a snapshot of the ecological and evolutionary dynamics impacting a trait; however, many evolutionary interactions vary over space and time, and a consideration of larger spatial scales is often required to fully understand the evolutionary dynamics involved in trait evolution (Brodie, Ridenhour & Brodie, 2002; Thompson, 2005; Gosden & Svensson, 2008; Siepielski, Dibattista & Carlson, 2009; Kuchta & Wake, 2016). Future studies of predator-mediated selection, relative camouflage, social interactions and gene flow, preferably in a geographic framework and over multiple years, would be beneficial in deciphering the role of selection in the maintenance of the striped/lead colour polymorphism in P. cinereus.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article at the publisher’s website:

Table S1. Univariate analyses of variance (ANOVAs) examining seasonal and population effects on the coloration of Plethodon cinereus. Significant effects are in bold.

Table S2. Tukey’s HSD post hoc test P-values for side Plethodon cinereus (A) avian, (B) snake and (C) mammalian contrast values. Comparisons below the diagonal of each substrate type are chromatic contrasts, whereas values above the diagonal are achromatic contrasts. Significant comparisons are in bold.

Table S3. Tukey’s HSD post hoc test P-values for ventral Plethodon cinereus (A) avian, (B) snake and (C) mammalian contrast values. Comparisons below the diagonal of each substrate type are chromatic contrasts, whereas values above the diagonal are achromatic contrasts. Significant comparisons are in bold.

Figure S1. Dorsal achromatic contrast values for the avian (A), snake (B) and mammalian (C) visual systems. Bars show means (±SE). Contrast values that lie below the grey horizontal line represent groups that are indistinguishable to the predator. Y-axes values vary by plot.

Figure S2. Side chromatic (A) and achromatic (B) contrast values for the avian visual system. Side chromatic (C) and achromatic (D) contrast values for the snake visual system. Side chromatic (E) and achromatic (F) contrast values for the mammalian visual system. Bars show means (±SE). Contrast values that lie below the grey horizontal line represent groups that are indistinguishable to the predator. Y-axes values vary by plot.

Figure S3. Ventral chromatic (A) and achromatic (B) contrast values for the avian visual system. Ventral chromatic (C) and achromatic (D) contrast values for the snake visual system. Ventral chromatic (E) and achromatic (F) contrast values for the mammalian visual system. Bars show means (±SE). Contrast values that lie below the grey horizontal line represent groups that are indistinguishable to the predator. Y-axes values vary by plot.

ACKNOWLEDGEMENTS

Financial support was provided to M.M.H. and S.R.K. by Ohio University. We thank Kyle Brooks, Olivia Brooks and Eric Leach for their assistance with data collection, and Rafael Maia for his expert assistance with the R package pavo. We also thank three anonymous reviewers for their helpful comments on this manuscript. Kyle Brooks provided the photograph of the striped and lead morphs of P. cinereus (Fig. 1A). All salamanders were collected with Ohio Department of Natural Resources Wild Animal Permit (17-19) to M.M.H., and all research was conducted under approval of the Ohio University Institutional Animal Care and Use Committee (12-L-050).