TY - JOUR
AU - Partridge, Julian, C
AB - Abstract An animal’s 3D form, combined with the directional lighting that is typical of many natural light environments, often results in the production of self-shadows, which may increase prey detectability to visual predators or vice versa. In terrestrial animals, countershading patterning, a luminance gradient from dark dorsal to pale ventral pigmentation, acts to counterbalance this effect by essentially reversing the distribution of light incident across the body surface. Although widespread among aquatic predators and prey, it is unclear whether countershading facilitates camouflage through elimination of self-shadows (self-shadow concealment [SSC]), enhances the match between an animal’s radiance and that of the background for multiple viewing angles (background matching [BM]), or a combination of both. We used clay models of a color-changing freshwater fish to determine the optimal patterning for SSC in different light environments, and we compared this to the skin reflectance profile of fish held under the same conditions. Fish adjusted their countershading pattern in response to changes in the light environment, but the observed reflectance profiles did not match the modeling predictions for optimal SSC. Thus fish adjusted their body pigmentation to match the viewing background rather than to conceal their ventral shadows. We suggest that different selection pressures resulting from the dissimilar characteristics of light in air and water have resulted in convergence of similar countershading phenotypes in terrestrial and aquatic prey. INTRODUCTION In both terrestrial and aquatic environments, predators and prey interact in a 3D world that rarely has a uniform light field (Penacchio et al., 2015a). Animal coloration that serves for concealment from predators (or prey) must therefore not only function to obscure salient morphological features of prey, such as their eyes (Kelley et al., 2013; Kjernsmo et al., 2016) and body edges (Cuthill et al., 2005; Stevens and Cuthill, 2006), but must also eliminate the shadows that are an inherent component of 3D form and nonuniform illumination. Objects with 3D form are readily detectable by vision because, in most natural environments, the upper surface of the object is mainly lit from overhead whereas the lower surface remains more or less in shadow, generating a strong radiance gradient across the object’s surface. It has been suggested that animals that have dark dorsal pigmentation and light ventral pigmentation (referred to as “countershading coloration”) can counteract the effect of overhead light falling on the body surface thereby largely eliminating the production of self-shadows (Poulton, 1888; Thayer, 1896), a mechanism referred to as “Self-Shadow Concealment” or “SSC” (Kiltie, 1988). SSC is thought to enhance crypsis by reducing intra-body visual contrasts and/or by rendering the prey animal perceptually “flat” and thereby less detectable or recognizable by potential predators (Poulton, 1890; Thayer, 1896; Cott, 1940; Hailman, 1977; Ruxton et al., 2004b; Rowland, 2009; Allen et al., 2012). Countershading may also facilitate concealment in a 3D world because predators view prey against diverse visual backgrounds (and vice versa) (Poulton, 1890; Thayer, 1896; Penacchio et al., 2015a) that differ depending on visualization direction. Thus a dark dorsal surface allows aerial and aquatic animals to remain concealed when seen (e.g., by an overhead predator) against the dark background provided by the ground or deeper water (Wallace, 1889; Beddard, 1895; Cott, 1940). It was previously predicted that a light ventral side will make prey cryptic when viewed from below and seen against the background of open sky or downwelling underwater light (Wallace, 1889). However, because the ventral surface of an animal is illuminated from below and, as light is scattered as it is reflected from the ground, light incident on the ventral surface of an animal will always be orders of magnitude less than that of downwelling light (Johnsen, 2002; Penacchio et al., 2015a). Thus, when viewed from below, an animal’s ventral surface can never match the downwelling irradiance unless the animal produces its own ventral illumination. The phenomenon whereby an animal’s radiance is adjusted to match that of the background against which it is seen is known as Background Matching (BM) (Cott, 1940) and is considered to be one of the most common strategies for camouflage (Merilaita and Stevens, 2011). Historically, the distinction between SSC and BM has not been clearly made (Poulton, 1890; Thayer, 1896; Cott, 1940; Penacchio et al., 2015a). However, SSC and BM can be considered as separate processes because an animal with optimal countershading for SSC will still be detectable unless its radiance also matches that of the visual background, that is, BM (Penacchio et al., 2015a). Likewise prey that achieve overall BM (i.e., for average body radiance) will still be detectable if their self-shadows produce internal edge contrasts, and they will remain recognizable if predators use a search image based on 3D form (Penacchio et al., 2015a). Although a number of studies with artificial prey and natural predators have provided strong support for the protective benefits of countershading (De Ruiter, 1956; Turner, 1961; Edmunds and Dewhirst, 1994; Rowland et al., 2007; Rowland et al., 2008), until recently, it has remained unclear whether countershaded animals achieve crypsis by concealing their 3D form (i.e., SSC), by matching the background (i.e., BM) at different viewing angles, or a combination of both. Recent theoretical and empirical evidence suggests that SSC is the predominant mechanism of concealment of 3D form in terrestrial animals (Allen et al., 2012; Penacchio et al., 2015a; Penacchio et al., 2015b; Cuthill et al., 2016). However, the predictions of optimal SSC in aquatic environments have never been tested experimentally, despite the fact that some of the most compelling examples of countershading coloration are found in aquatic environments (e.g., Körner, 1982; Caro, 2009; Claes et al., 2010). A number of studies have suggested that BM is a more likely determinant of countershading coloration in aquatic environments than SSC because of the optical characteristics of light in water (Johnsen, 2002; Ruxton et al., 2004a; Ruxton et al., 2004b; Kelley and Merilaita, 2015). During its passage through water, light is scattered and attenuated by dissolved and suspended particulate matter, causing light to be diffused and less directional than light in terrestrial systems (Ruxton et al., 2004b). The physical properties of light in water, and particularly the extreme asymmetry of the underwater light field, have led to the suggestion that selection for a dark dorsal surface in aquatic environments is likely to be particularly strong (Johnsen, 2014a). Surprisingly, however, almost all research to date has focused on countershading in terrestrial animals, (reviewed by Rowland, 2009). This study adopts a novel approach to establish the relative importance of the visual background, viewing angle and overhead illumination in predicting “countershading coloration,” in a freshwater fish that can change color. The use of a color-changing animal provides unique insights into the link between body coloration and the environment because, within the physiological constraints of the pigments involved, the adopted coloration should reflect the animal’s detected change in social/environmental conditions. Many fishes change color, including their overall broad-spectrum reflectance, with color change occurring over time periods ranging from seconds to weeks (Sköld et al., 2016). The western rainbowfish (Melanotaenia australis) was selected for this study because it is a small freshwater fish that exhibits countershading, and can adjust its coloration over a period of days/weeks in response to the visual background in which they are maintained (Rodgers et al., 2010; Kelley and Merilaita, 2015; Kelley et al., 2016). The predominant “color” change is due to changes in the dispersion of (black/brown) melanin pigment in the skin changing dermal lightness or darkness rather than hue. In this study, we experimentally evaluated the putative roles and efficacy of SSC and BM in explaining changes in rainbowfish countershading patterning in response to different visual backgrounds and for different predator viewing angles. In 2 separate experiments, fish were allocated to identical aquaria but with varying illumination (high or low) and varying substrate reflectance (Experiment 1; black or white) or with varying illumination and varying horizontal background reflectance (Experiment 2; black or white), allowing us to differentiate whether SSC or BM models best explained the phenotypic changes in countershading that were observed in each case. We determined the optimal countershading gradient for SSC by making clay models of western rainbowfish and photographing them underwater under the same experimental lighting conditions used in Experiments 1 and 2. We generated and tested a number of predictions about the effects of the light environment on changes in fish skin reflectance under the experimental conditions for concealment by both SSC and BM (Table 1). Table 1 Predicted changes in reflectance for concealment by SSC and BM for different viewing angles following manipulation of illumination level (high or low), substrate reflectance (BS: black substrate, WS: white substrate) and horizontal background reflectance (BW: black walls, WW: white walls) Mechanism . Body reflectance . Illumination level . Substrate reflectance . Horizontal background reflectance . Concealment by SSC Dorsal half No change Paler on WS Paler on WW Ventral half No change Darker on WS Darker on WW D–V gradient No change Steeper on BS Steeper on BW Concealment by BM when viewed from the side Dorsal half No change No change Paler on WW Ventral half No change Paler on WS Paler on WW D–V gradient No change No change No change Concealment by BM when viewed from above Dorsal half Paler for high Paler on WS — Ventral half — — — D–V gradient — — — Concealment by BM when view from below Dorsal half — — — Ventral half Paler for high No change — D–V gradient — — — Mechanism . Body reflectance . Illumination level . Substrate reflectance . Horizontal background reflectance . Concealment by SSC Dorsal half No change Paler on WS Paler on WW Ventral half No change Darker on WS Darker on WW D–V gradient No change Steeper on BS Steeper on BW Concealment by BM when viewed from the side Dorsal half No change No change Paler on WW Ventral half No change Paler on WS Paler on WW D–V gradient No change No change No change Concealment by BM when viewed from above Dorsal half Paler for high Paler on WS — Ventral half — — — D–V gradient — — — Concealment by BM when view from below Dorsal half — — — Ventral half Paler for high No change — D–V gradient — — — For concealment by SSC, it is assumed that prey are predominately viewed from the side (Kiltie, 1988). Dorsal half and ventral half refer to the mean reflectance of the upper and lower 50% of the body (when viewed from the side), whereas the D–V gradient is the change in percentage body reflectance across a D–V transect along the mid line of the fish and is always positive for countershaded prey. Open in new tab Table 1 Predicted changes in reflectance for concealment by SSC and BM for different viewing angles following manipulation of illumination level (high or low), substrate reflectance (BS: black substrate, WS: white substrate) and horizontal background reflectance (BW: black walls, WW: white walls) Mechanism . Body reflectance . Illumination level . Substrate reflectance . Horizontal background reflectance . Concealment by SSC Dorsal half No change Paler on WS Paler on WW Ventral half No change Darker on WS Darker on WW D–V gradient No change Steeper on BS Steeper on BW Concealment by BM when viewed from the side Dorsal half No change No change Paler on WW Ventral half No change Paler on WS Paler on WW D–V gradient No change No change No change Concealment by BM when viewed from above Dorsal half Paler for high Paler on WS — Ventral half — — — D–V gradient — — — Concealment by BM when view from below Dorsal half — — — Ventral half Paler for high No change — D–V gradient — — — Mechanism . Body reflectance . Illumination level . Substrate reflectance . Horizontal background reflectance . Concealment by SSC Dorsal half No change Paler on WS Paler on WW Ventral half No change Darker on WS Darker on WW D–V gradient No change Steeper on BS Steeper on BW Concealment by BM when viewed from the side Dorsal half No change No change Paler on WW Ventral half No change Paler on WS Paler on WW D–V gradient No change No change No change Concealment by BM when viewed from above Dorsal half Paler for high Paler on WS — Ventral half — — — D–V gradient — — — Concealment by BM when view from below Dorsal half — — — Ventral half Paler for high No change — D–V gradient — — — For concealment by SSC, it is assumed that prey are predominately viewed from the side (Kiltie, 1988). Dorsal half and ventral half refer to the mean reflectance of the upper and lower 50% of the body (when viewed from the side), whereas the D–V gradient is the change in percentage body reflectance across a D–V transect along the mid line of the fish and is always positive for countershaded prey. Open in new tab Under the predictions of SSC, we expected that fish should exhibit a weak dorsal–ventral (D–V) gradient when the substrate reflectance is high because the relatively high upwelling illumination requires a relatively dark ventral skin color to minimize ventral shadowing. In contrast, black substrates, which increase the difference between the radiance incidence on the dorsal and ventral parts of the fish, should result in a stronger D–V gradient (due to a lighter ventral surface) for concealment by SSC. As horizontal radiance increases, there is relatively less light incident on the upper surfaces of the body to cause ventral shadowing. The predictions of SSC relate to the relative amount of downwelling to upwelling light, thus we expected no effect of illumination level on changes in the dermal reflectance gradient. Under a model of BM, we expected no change in the D–V reflectance gradient according to the experimental lighting conditions. When viewed from the side, we expected fish to become paler on white horizontal backgrounds and darker on black horizontal backgrounds, resulting in a uniform reflectance distribution across the body surface and an even skin radiance similar to the background radiance (Kelley and Merilaita, 2015). When viewed from above, we expected dorsal reflectance to be increased when fish are held on white substrates, and to be reduced when on dark substrates, to reduce conspicuousness to predators that attack from overhead. When viewed from below, ventral reflectance should be higher under high light (HL) intensity conditions and reduced under low light (LL) intensity conditions to reduce conspicuous to visual predators. MATERIALS AND METHODS Fish origin and maintenance Adult rainbowfish (M. australis [Castelnau, 1875]) used in this experiment were captured from Coondiner Creek (latitude: −23.0043, longitude: 119.6213), a tributary of the Fortescue River in the Pilbara region of Western Australia in October 2013. Predators at this site include those that attack from above such as herons, as well as mid-water piscivorous predators such as cormorants and the spangled perch (Leiopotherapon unicolor [Günther, 1859]), that are likely to target rainbowfish by attacking from the side or from below. Prior to the experiments, all fish were maintained in mixed sex tanks (80 × 50 × 31 cm; approx. equal sex ratio) and fed a diet of commercially prepared flake food, frozen blood worm, and live Artemia nauplii. The temperature of the aquarium water was 26 ± 1 °C and the lighting was set to a 12:12 h light:dark cycle. All animal experiments were conducted following the Australian code for the care and use of animals for scientific purposes and were approved by UWA Animal Ethics Committee (approval no. RA/3/100/1176). Experimental design A total of 24 replicate aquaria (30 × 30 × 30 cm) were used in this study to individually isolate fish so that color change occurred independently of social cues that can influence color change (Kelley et al., 2016). Particular light environments can be stressful for some fishes, thus we monitored fish twice daily for signs of stress (which we did not observe), and we minimized the length of the experimental procedure. Six tanks were allocated to each treatment and arranged in a block design, with multiple blocks conducted for each experiment (Experiment 1: 4 blocks; Experiment 2: 2 blocks). Different individuals were used for each block of the experiment such that no individual was used more than once. Lighting was provided by overhead dual fluorescent tubes (one each of Sylvania Grolux™ 30W and GE Tri-Tech plus™ 6500k/f30 daylight) on a 12 h light:dark cycle. Light intensity in the treatment tanks was manipulated using neutral density (ND) filters, which absorb and transmit light equally across the visible spectrum, placed below the lights. The HL intensity treatment involved the use of a LEE 0.15ND filter, which has a nominal transmittance of 69.3% (400–700 nm), placed below the light source, whereas the LL intensity light treatment comprised a double layer of LEE 0.9ND filter (nominal transmittance 13.7%) to achieve a transmittance of 1.6% (400–700 nm). Absolute spectral irradiance (300–700 nm) was measured for each experimental light condition in Experiments 1 and 2 (see below) using a calibrated USB4000 spectrometer (Ocean Optics Inc., Dunedin, FL) fitted with a 600 μm diameter fiber and CC-3-UV cosine corrector. The probe was placed 10 cm below the water’s surface and adjusted to measure upwelling, downwelling, and sidewelling irradiance in each tank and for each experiment. No adjustment was made for the effect of immersion in water on the nominally cosine angular response function of the CC-3-UV corrector, which is designed for use in air, as we required only relative measurements of spectral irradiance. (i) Experiment 1—manipulation of illumination level and substrate reflectance In the first experiment, substrate reflectance was altered using either black or white Corflute™ extruded polypropylene sheet placed in the base of each tank. The black Corflute had a measured (by methods described above, but using a bare collection fiber) reflectance value of 6.1% over the spectrum from 400 to 700 nm, and the white a value of 81% relative to a white standard (WS-1, Ocean Optics Inc.). To generate a homogenous visual background when the fish was viewed from the side, standard copy paper was printed grey to achieve a reflectance midway between that of the black and white substrates (~45%). The grey paper was laminated and used to line the inside walls of the experimental tanks, to avoid internal reflection from the glass, and to provide approximately uniform side-welling light in the aquaria. Thus, the treatment groups for the first experiment were HLBS (high light, black substrate), HLWS (high light, white substrate), LLBS (low light black substrate) and LLWS (low light white substrate). A total of 76 fish were used for Experiment 1 (HLBS: n = 20; HLWS: n = 21; LLBS: n = 18; LLWS: n = 17). (ii) Experiment 2—manipulation of illumination level and horizontal background reflectance In the second experiment, light intensity was manipulated as described above (HL or LL) but we manipulated the reflectance of the visual background when viewed from the side (hereafter “horizontal background”) by covering the sides of the tank with either black or white Corflute™ (reflectance values given above). The substrate of the tanks was laminated grey paper, as described above, in all 4 experimental treatments of Experiment 2. Thus the treatment groups for the second experiment were HLBW (high light, black walls), HLWW (high light, white walls), LLBW (low light, black walls), and LLWW (low light, white walls). For both Experiments 1 and 2, one side of each tank was left open to allow assessment of fish health throughout the experiment. A total of 46 fish were used in Experiment 2 (HLBW: n = 12; HLWW: n = 12; LLBW: n = 11; LLWW: n = 11). (iii) Calculating changes in fish body reflectance Fish in both experiments were photographed at the start of the experimental blocks, before they were allocated to different treatments, and 2 weeks after the start of this time, when they had changed “color.” Note that we use the term “color change” when referring to changes in the dermal reflectance of the fish that primarily occur due to changes in the dispersion of black/brown melanin pigment. Each fish was photographed using a Nikon D7100 digital SLR fitted with a 60mm AFS macro lens. Photographs were taken out of the water using a light diffusion tent lit by 2 halogen floodlights (500W). The camera was positioned on a tripod set at a fixed height (30 cm) above the fish. All photographs were recorded in Nikon proprietary raw file format (.NEF) using the aperture priority setting (set at f/7.1), and shutter speeds of 1/640 s or 1/800 s (automatically selected) with all in-camera processing turned off. The fish were photographed right side down on a photography slate with a color standard (based on a mini GretagMacbeth ColourChecker®) and calibration scale in frame. We did not photograph the fish in water, or photograph the fish from dorsal/ventral viewing angles, as this would have required additional handling and anesthesia. Anesthesia was not used, as it is known to darken fish coloration, even over short time scales (Gray et al., 2011) and fish remained out of the water for <5 s. (iv) Predicting optimal SSC for different light environments We calculated optimal fish body reflectance for SSC by photographing clay models of rainbowfish under the same lighting conditions used in Experiments 1 and 2. The models were made using casts of 4 adult rainbowfish that were euthanized with an overdose of MS222 and immediately frozen at −20 °C. The right side of each frozen fish’s body was gently pressed into silicon rubber (Elastosil M4503; Fibreglass & Resin Sales Pty Ltd, Perth, Australia), and the mould was left to cure overnight. Four models were formed using plaster of Paris and left to dry before being painted with a white, enamel-based craft paint (Micador Group, Melbourne, Australia). The models were not intended to recreate the reflectance properties of the fish’s skin, but to provide a reconstruction of the 3D natural body shape of the fish (because we expect that changes in reflectance over the body will depend on body shape). Images of the models were captured in each of the experimental lighting and visual background conditions by suspending the fish model and a mini GretagMacbeth ColourChecker® on a 10 cm strip of transparent plastic. Three photographs of each model fish were taken using the aperture priority mode (f/7.1) and variable shutter speeds as before. We produced the predicted optimal pattern of SSC by inverting the mean reflectance profile of the models for each lighting condition and visual background in Experiments 1 and 2. Image Analysis The RAW (.NEF) images were firstly converted to 16 bit TIFF file format using DCRAW software (available at https://www.cybercom.net/~dcoffin/dcraw/, last accessed 27 June 2017), which allows the user to specify the processes involved in the conversion and ensures that no data are lost or altered by file compression (see Appendix A1 for DCRAW settings). The resulting TIFF images were then split into 4 RGB channels using the Image J plugin “DeBayer” (http://www.umanitoba.ca/faculties/science/astronomy/jwest/plugins.html, last accessed 27 June 2017). The green channel (G2) was selected for all subsequent analyses, providing image information in the mid-wavelength range, which preliminary investigations revealed, comprised the most information about patterning. Imaging software ImageJ 1.47 (https://imagej.nih.gov/ij) was used to quantify the mean body reflectance and the D–V reflectance gradient of each fish. To quantify mean dorsal and ventral body reflectance, we split the image of each fish into dorsal and ventral halves by aligning the midpoints of 6 lines drawn dorso-ventrally at predetermined positions on the body, recording the mean pixel value for each dorsal and ventral half of the body. We calculated the D–V reflectance profile for each fish and for each fish clay model by drawing a line dorso-ventrally at the widest part of the body on the digital photographic image and using the ImageJ built-in “plot profile” function to obtain a pixel value for each D–V position on the body. The body size of the fish/model was standardized between 0 (most dorsal position) and 100 (most ventral position) and reflectance profiles were linearly interpolated to give pixel values for 100 positions along the D–V transect for each individual fish/model. All pixel values were subsequently converted to percentage reflectance using the methodology described in the Supplementary Material (ESM1). Statistical analyses We calculated the changes in mean fish dorsal and ventral body reflectance (R, expressed as percentage reflection) over the 2-week experimental period (i.e., R after treatment − R before treatment; thus positive values indicate increased body reflectance and negative values indicate decreased body reflectance) to use as dependent variables in subsequent analyses. Preliminary analyses revealed that using the experimental block (number of replicates of the experiment performed) as a random effect had no effect in our models. Thus we used linear models to test for an effect of light intensity and substrate reflectance (Experiment 1) or light intensity and horizontal background reflectance (Experiment 2), and their interactions, on changes in mean fish dorsal and ventral percentage reflectance. Previous work has revealed that the extent of color change on different backgrounds is dependent on previous skin coloration (Kelley et al., 2016), thus we used previous skin reflectance (i.e., before allocation to treatment) as a covariate in the models. Fish body length (SL) was also included as a covariate in our initial models to consider the effect of fish size on color change. Diagnostic plots were used to assess for overdispersion and to examine the distribution of the residual errors. Non-significant terms were removed from the model in a stepwise manner and the final models are presented in each case. All statistical modeling was performed using the software program R, version 3.2.2. To calculate the optimal SSC in each of the experimental lighting conditions we tested the fit of a number of linear and non-linear models to the relationship between pixel position on the D–V profile and percentage reflectance, using the software CurveExpertPro 2.20™. The best fitting model was evaluated by inspection of the AICc values and was determined to be a segmented linear regression with a break point estimated at approximately 49 (i.e., midway down the body). A segmented regression model (y = ax + b; x < 49; cx + d; x ≥ 50) was therefore fit to the clay model profiles, as well as the reflectance profiles of each individual fish before and after their allocation to a treatment. Models were fit using the package “Segmented” in R (https://cran.r-project.org/package=segmented). The effect of the experimental lighting conditions on the D–V reflectance gradients was tested using linear models (as described above), using the 2 slopes as the dependent variables and using light intensity and substrate reflectance (Experiment 1) or light intensity and horizontal background reflectance (Experiment 2), and their interactions, as predictors. RESULTS Experiment 1: Manipulation of light intensity and substrate reflectance The linear models revealed a significant interaction between previous skin reflectance and substrate reflectance on the change in dorsal percentage reflectance (Table 2). On a white substrate there was a positive relationship between previous skin reflectance and the change in dorsal percentage reflectance (i.e., fish that were paler before the experiment showed a greater increase in dorsal reflectance (i.e., becoming paler still) than those that were darker; r2 = 0.16, t36 = 2.83, P = 0.008; Figure 1a). There was no relationship between previous skin reflectance and the change in dorsal reflectance for fish held on a black substrate (r2 = 0.02, t36 = −0.51, P = 0.61; Figure 1a). The results for changes in mean ventral reflectance were similar (Table 2); on a white substrate, fish with relatively pale undersides showed a greater increased in reflectance than those with darker ventral surfaces (r2 = 0.20, t36 = 3.23, P = 0.003; Figure 1b), but the relationship was not significant for fish on a black ventral surface (r2 = 0.02, t36 = 0.22, P = 0.83; Figure 1b). Illumination level had an effect on the change in ventral reflectance, but not on dorsal reflectance (Table 2). Specifically, fish held in tanks with high illumination showed a greater increase in ventral reflectance (mean change in ventral R%; high = 9.84 ± 1.75, low = 4.30 ± 1.36) than those held in tanks under low illumination (Table 2) and there was no effect of previous skin reflectance on change in ventral coloration (Table 2). Table 2 Final set of linear models testing for an effect of illumination level (high or low) and substrate reflectance (Experiment 1) or illumination level and horizontal background reflectance (b; Experiment 2) on changes in dorsal and ventral skin reflectance . . Experiment 1 . Experiment 2 . Variable . Effect . df . Estimate . SE . t . P . df . Estimate . SE . t . P . Dorsal R Intercept −1.58 2.29 0.69 0.49 19.9 5.79 3.44 0.001 Background (W) 1, 71 −2.20 3.72 −0.59 0.56 1, 42 6.92 1.14 6.09 <0.001 Light intensity (L) 1, 71 −1.53 1.06 −1.45 0.15 1, 42 2.85 1.15 2.48 0.017 Previous R 1, 71 −0.04 0.16 −0.25 0.81 1, 42 −0.63 0.22 −2.85 0.007 Background (W)* Previous R 1, 71 0.91 0.28 3.30 0.002 — — — — — Ventral R Intercept 4.56 4.28 1.06 0.29 110.13 17.32 6.36 <0.001 Background (W) 1, 71 −10.5 6.63 −1.58 0.12 1, 41 4.58 2.10 2.18 0.035 Light intensity (L) 1, 71 −6.04 1.89 −3.20 0.002 1, 41 0.18 2.11 0.08 0.934 Previous R 1, 71 0.04 0.13 0.30 0.76 1, 41 −1.38 0.17 −8.29 <0.001 Background (W)* Previous R 1, 71 0.62 0.20 3.05 0.003 — — — — — SL — — — — — 1, 41 −5.44 2.93 −1.86 0.071 . . Experiment 1 . Experiment 2 . Variable . Effect . df . Estimate . SE . t . P . df . Estimate . SE . t . P . Dorsal R Intercept −1.58 2.29 0.69 0.49 19.9 5.79 3.44 0.001 Background (W) 1, 71 −2.20 3.72 −0.59 0.56 1, 42 6.92 1.14 6.09 <0.001 Light intensity (L) 1, 71 −1.53 1.06 −1.45 0.15 1, 42 2.85 1.15 2.48 0.017 Previous R 1, 71 −0.04 0.16 −0.25 0.81 1, 42 −0.63 0.22 −2.85 0.007 Background (W)* Previous R 1, 71 0.91 0.28 3.30 0.002 — — — — — Ventral R Intercept 4.56 4.28 1.06 0.29 110.13 17.32 6.36 <0.001 Background (W) 1, 71 −10.5 6.63 −1.58 0.12 1, 41 4.58 2.10 2.18 0.035 Light intensity (L) 1, 71 −6.04 1.89 −3.20 0.002 1, 41 0.18 2.11 0.08 0.934 Previous R 1, 71 0.04 0.13 0.30 0.76 1, 41 −1.38 0.17 −8.29 <0.001 Background (W)* Previous R 1, 71 0.62 0.20 3.05 0.003 — — — — — SL — — — — — 1, 41 −5.44 2.93 −1.86 0.071 Previous skin reflectance was included as a covariate in all models. Background refers to substrate reflectance in Experiment 1 and horizontal background reflectance in Experiment 2. Missing information is where a term is not included in the final model. Significant effects are shown in bold. Open in new tab Table 2 Final set of linear models testing for an effect of illumination level (high or low) and substrate reflectance (Experiment 1) or illumination level and horizontal background reflectance (b; Experiment 2) on changes in dorsal and ventral skin reflectance . . Experiment 1 . Experiment 2 . Variable . Effect . df . Estimate . SE . t . P . df . Estimate . SE . t . P . Dorsal R Intercept −1.58 2.29 0.69 0.49 19.9 5.79 3.44 0.001 Background (W) 1, 71 −2.20 3.72 −0.59 0.56 1, 42 6.92 1.14 6.09 <0.001 Light intensity (L) 1, 71 −1.53 1.06 −1.45 0.15 1, 42 2.85 1.15 2.48 0.017 Previous R 1, 71 −0.04 0.16 −0.25 0.81 1, 42 −0.63 0.22 −2.85 0.007 Background (W)* Previous R 1, 71 0.91 0.28 3.30 0.002 — — — — — Ventral R Intercept 4.56 4.28 1.06 0.29 110.13 17.32 6.36 <0.001 Background (W) 1, 71 −10.5 6.63 −1.58 0.12 1, 41 4.58 2.10 2.18 0.035 Light intensity (L) 1, 71 −6.04 1.89 −3.20 0.002 1, 41 0.18 2.11 0.08 0.934 Previous R 1, 71 0.04 0.13 0.30 0.76 1, 41 −1.38 0.17 −8.29 <0.001 Background (W)* Previous R 1, 71 0.62 0.20 3.05 0.003 — — — — — SL — — — — — 1, 41 −5.44 2.93 −1.86 0.071 . . Experiment 1 . Experiment 2 . Variable . Effect . df . Estimate . SE . t . P . df . Estimate . SE . t . P . Dorsal R Intercept −1.58 2.29 0.69 0.49 19.9 5.79 3.44 0.001 Background (W) 1, 71 −2.20 3.72 −0.59 0.56 1, 42 6.92 1.14 6.09 <0.001 Light intensity (L) 1, 71 −1.53 1.06 −1.45 0.15 1, 42 2.85 1.15 2.48 0.017 Previous R 1, 71 −0.04 0.16 −0.25 0.81 1, 42 −0.63 0.22 −2.85 0.007 Background (W)* Previous R 1, 71 0.91 0.28 3.30 0.002 — — — — — Ventral R Intercept 4.56 4.28 1.06 0.29 110.13 17.32 6.36 <0.001 Background (W) 1, 71 −10.5 6.63 −1.58 0.12 1, 41 4.58 2.10 2.18 0.035 Light intensity (L) 1, 71 −6.04 1.89 −3.20 0.002 1, 41 0.18 2.11 0.08 0.934 Previous R 1, 71 0.04 0.13 0.30 0.76 1, 41 −1.38 0.17 −8.29 <0.001 Background (W)* Previous R 1, 71 0.62 0.20 3.05 0.003 — — — — — SL — — — — — 1, 41 −5.44 2.93 −1.86 0.071 Previous skin reflectance was included as a covariate in all models. Background refers to substrate reflectance in Experiment 1 and horizontal background reflectance in Experiment 2. Missing information is where a term is not included in the final model. Significant effects are shown in bold. Open in new tab Figure 1 Open in new tabDownload slide Effect of previous % skin reflectance on color change for fish exposed to black (filled circles) or white (open circles) visual backgrounds. Percentage change in dorsal and ventral reflectance is shown for fish in Experiment 1 (a, b; manipulation of substrate reflectance and light intensity) and for fish in Experiment 2 (c, d; manipulation of horizontal background reflectance and light intensity). Lines indicate least squares regression fits for black (solid line) and white backgrounds (dashed line) (panel a; solid line, y = −0.05x + 0.98, r2 = 0.02; dashed line, y = 0.82x – 0.57, r2 = 0.16; panel b; solid line: y = 0.03x + 2.10, r2 = 0.03, dashed line: y = 0.61x – 7.17, r2 = 0.20; panel c; solid line, y = −0.31 + 13.05, r2 = 0.02, dashed line: = −0.97x + 37.2, r2 = 0.18; panel d; solid line, y = −1.24x + 74.0, r2 = 0.48, dashed line, y = −1.56x + 95.8, r2 = 0.72). N = 38 (black) and N = 38 (white) for Experiment 1 and N = 23 (black) and 23 (white) for Experiment 2. Note that high reflectance values indicate pale skin whereas low reflectance values indicate dark skin. Figure 1 Open in new tabDownload slide Effect of previous % skin reflectance on color change for fish exposed to black (filled circles) or white (open circles) visual backgrounds. Percentage change in dorsal and ventral reflectance is shown for fish in Experiment 1 (a, b; manipulation of substrate reflectance and light intensity) and for fish in Experiment 2 (c, d; manipulation of horizontal background reflectance and light intensity). Lines indicate least squares regression fits for black (solid line) and white backgrounds (dashed line) (panel a; solid line, y = −0.05x + 0.98, r2 = 0.02; dashed line, y = 0.82x – 0.57, r2 = 0.16; panel b; solid line: y = 0.03x + 2.10, r2 = 0.03, dashed line: y = 0.61x – 7.17, r2 = 0.20; panel c; solid line, y = −0.31 + 13.05, r2 = 0.02, dashed line: = −0.97x + 37.2, r2 = 0.18; panel d; solid line, y = −1.24x + 74.0, r2 = 0.48, dashed line, y = −1.56x + 95.8, r2 = 0.72). N = 38 (black) and N = 38 (white) for Experiment 1 and N = 23 (black) and 23 (white) for Experiment 2. Note that high reflectance values indicate pale skin whereas low reflectance values indicate dark skin. Experiment 2: manipulation of light intensity and horizontal background reflectance There was a significant effect of horizontal background reflectance on changes in fish dorsal and ventral percentage reflectance, and the extent of dermal color change depended on previous skin reflectance (Table 2). On a white horizontal background, there was a negative relationship between previous dorsal skin reflectance and the observed change in reflectance (r2 = 0.18, t21 = −2.39, P = 0.026), whereas on a black horizontal background this relationship was not significant (r2 = 0.02, t21 = −1.19, P = 0.25). In contrast to Experiment 1, on a white horizontal background fish with darker dorsal surfaces showed a greater change in reflectance, becoming paler than those with less dark dorsal halves (Figure 1c). Changes in ventral reflectance were affected by previous skin reflectance for fish on both white (r2 = 0.72, t21 = −7.59, P < 0.001) and black (r2 = 0.48, t21 = −4.59, P < 0.001) horizontal backgrounds, being greatest in fish with previously low ventral skin reflectance (Figure 1d). In contrast to Experiment 1, illumination level had an effect on the change in dorsal, but not ventral reflectance (Table 2). Following the lighting treatment, fish under low-level illumination had higher dorsal reflectance than those under high illumination (high = 33.0 ± 1.01, low = 36.1 ± 1.17), and there was no effect of previous skin reflectance on ventral color change (Table 2). Comparing predicted SSC with observed fish D–V reflectance gradients (a) Modeling optimal patterning for SSC The optimal pattern of pigmentation for SSC differed according to the visual background (Figure 2). The optimal pigmentation for concealment by SSC is a steep linear reflectance gradient on the upper body surface (slope 1), which is higher for black substrates or black horizontal backgrounds than for white substrates or white horizontal backgrounds (Figure 2; Table 3). The reflectance gradient for the ventral surface of the fish models (slope 2) is flat in all treatments (Figure 2). Figure 2 Open in new tabDownload slide Predicted skin reflectance profiles (mean ± SD) under an optimal model of SSC for lighting conditions that vary in substrate reflectance (a; Experiment 1) and horizontal background reflectance (b; Experiment 2). Black lines = black substrate/horizontal background; grey lines = white substrate/horizontal background. A smoothing function has been applied to the reflectance profiles (span = 0.15). Images of the models photographed in the treatments are also shown. HLBS: high light, black substrate; HLWS: high light, white substrate; HLBW: high light, black walls; HLWW: high light, white walls. The reflectance profiles of the models in the low illumination treatments (not shown) were negligible for both experiments due to the very low light levels. Figure 2 Open in new tabDownload slide Predicted skin reflectance profiles (mean ± SD) under an optimal model of SSC for lighting conditions that vary in substrate reflectance (a; Experiment 1) and horizontal background reflectance (b; Experiment 2). Black lines = black substrate/horizontal background; grey lines = white substrate/horizontal background. A smoothing function has been applied to the reflectance profiles (span = 0.15). Images of the models photographed in the treatments are also shown. HLBS: high light, black substrate; HLWS: high light, white substrate; HLBW: high light, black walls; HLWW: high light, white walls. The reflectance profiles of the models in the low illumination treatments (not shown) were negligible for both experiments due to the very low light levels. Table 3 Predicted D–V gradients (a) under a model of concealment by SSC generated from fish model reflectance profiles . . (a) . (b) . . . Predicted parameters (models) . Observed parameters (live fish) . Experiment . Treat . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . 1 HLWS 0.65 ± 0.01 0.62–0.67 0.00 ± 0.01 −0.02 to 0.02 0.98 0.49 ± 0.00 0.41–0.58 0.31 ± 0.04 0.23–0.40 0.94 HLBS 1.71 ± 0.05 1.61–1.81 0.27 ± 0.05 0.17–0.36 0.97 0.44 ± 0.05 0.34–0.55 0.66 ± 0.01 0.56–0.77 0.94 LLWS 0.04 ± 0.00 0.03–0.04 0.00 ± 0.00 −0.01 to 0.00 0.84 0.64 ± 0.05 0.54–0.75 0.42 ± 0.05 0.32–0.53 0.95 LLBS 0.06 ± 0.01 0.05–0.07 0.04 ± 0.01 0.02–0.04 0.87 0.54 ± 0.06 0.43–0.66 0.52 ± 0.06 0.40–0.63 0.93 2 HLWW 0.74 ± 0.02 0.70–0.78 −0.01 ± 0.02 −0.05 to 0.03 0.97 0.45 ± 0.04 0.37–0.52 0.38 ± 0.04 0.31–0.46 0.96 HLBW 1.34 ± 0.03 1.27–1.40 0.08 ± 0.03 0.02–0.14 0.98 0.43 ± 0.04 0.35–0.50 0.56 ± 0.04 0.47–0.64 0.96 LLWW 0.02 ± 0.00 0.01–0.02 0.00 ± 0.00 0.00–0.01 0.67 0.47 ± 0.03 0.40–0.54 0.35 ± 0.04 0.27–0.42 0.96 LLBW −0.13 ± 0.03 −0.19 to −0.08 0.05 ± 0.01 0.03–0.07 0.74 0.43 ± 0.04 0.36–0.51 0.46 ± 0.04 0.38–0.54 0.96 . . (a) . (b) . . . Predicted parameters (models) . Observed parameters (live fish) . Experiment . Treat . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . 1 HLWS 0.65 ± 0.01 0.62–0.67 0.00 ± 0.01 −0.02 to 0.02 0.98 0.49 ± 0.00 0.41–0.58 0.31 ± 0.04 0.23–0.40 0.94 HLBS 1.71 ± 0.05 1.61–1.81 0.27 ± 0.05 0.17–0.36 0.97 0.44 ± 0.05 0.34–0.55 0.66 ± 0.01 0.56–0.77 0.94 LLWS 0.04 ± 0.00 0.03–0.04 0.00 ± 0.00 −0.01 to 0.00 0.84 0.64 ± 0.05 0.54–0.75 0.42 ± 0.05 0.32–0.53 0.95 LLBS 0.06 ± 0.01 0.05–0.07 0.04 ± 0.01 0.02–0.04 0.87 0.54 ± 0.06 0.43–0.66 0.52 ± 0.06 0.40–0.63 0.93 2 HLWW 0.74 ± 0.02 0.70–0.78 −0.01 ± 0.02 −0.05 to 0.03 0.97 0.45 ± 0.04 0.37–0.52 0.38 ± 0.04 0.31–0.46 0.96 HLBW 1.34 ± 0.03 1.27–1.40 0.08 ± 0.03 0.02–0.14 0.98 0.43 ± 0.04 0.35–0.50 0.56 ± 0.04 0.47–0.64 0.96 LLWW 0.02 ± 0.00 0.01–0.02 0.00 ± 0.00 0.00–0.01 0.67 0.47 ± 0.03 0.40–0.54 0.35 ± 0.04 0.27–0.42 0.96 LLBW −0.13 ± 0.03 −0.19 to −0.08 0.05 ± 0.01 0.03–0.07 0.74 0.43 ± 0.04 0.36–0.51 0.46 ± 0.04 0.38–0.54 0.96 Estimates of slopes 1 (dorsal profile) and 2 (ventral profile) are shown, along with the standard error (SE), 95% confidence intervals (CI) and adjusted r2 values The D–V gradients observed in the real fish, photographed after 2 weeks spent in the same experimental treatments, are shown in (b). Open in new tab Table 3 Predicted D–V gradients (a) under a model of concealment by SSC generated from fish model reflectance profiles . . (a) . (b) . . . Predicted parameters (models) . Observed parameters (live fish) . Experiment . Treat . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . 1 HLWS 0.65 ± 0.01 0.62–0.67 0.00 ± 0.01 −0.02 to 0.02 0.98 0.49 ± 0.00 0.41–0.58 0.31 ± 0.04 0.23–0.40 0.94 HLBS 1.71 ± 0.05 1.61–1.81 0.27 ± 0.05 0.17–0.36 0.97 0.44 ± 0.05 0.34–0.55 0.66 ± 0.01 0.56–0.77 0.94 LLWS 0.04 ± 0.00 0.03–0.04 0.00 ± 0.00 −0.01 to 0.00 0.84 0.64 ± 0.05 0.54–0.75 0.42 ± 0.05 0.32–0.53 0.95 LLBS 0.06 ± 0.01 0.05–0.07 0.04 ± 0.01 0.02–0.04 0.87 0.54 ± 0.06 0.43–0.66 0.52 ± 0.06 0.40–0.63 0.93 2 HLWW 0.74 ± 0.02 0.70–0.78 −0.01 ± 0.02 −0.05 to 0.03 0.97 0.45 ± 0.04 0.37–0.52 0.38 ± 0.04 0.31–0.46 0.96 HLBW 1.34 ± 0.03 1.27–1.40 0.08 ± 0.03 0.02–0.14 0.98 0.43 ± 0.04 0.35–0.50 0.56 ± 0.04 0.47–0.64 0.96 LLWW 0.02 ± 0.00 0.01–0.02 0.00 ± 0.00 0.00–0.01 0.67 0.47 ± 0.03 0.40–0.54 0.35 ± 0.04 0.27–0.42 0.96 LLBW −0.13 ± 0.03 −0.19 to −0.08 0.05 ± 0.01 0.03–0.07 0.74 0.43 ± 0.04 0.36–0.51 0.46 ± 0.04 0.38–0.54 0.96 . . (a) . (b) . . . Predicted parameters (models) . Observed parameters (live fish) . Experiment . Treat . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . Slope 1 ± SE (a) . 95% CI . Slope 2 ± SE (c) . 95% CI . r2 . 1 HLWS 0.65 ± 0.01 0.62–0.67 0.00 ± 0.01 −0.02 to 0.02 0.98 0.49 ± 0.00 0.41–0.58 0.31 ± 0.04 0.23–0.40 0.94 HLBS 1.71 ± 0.05 1.61–1.81 0.27 ± 0.05 0.17–0.36 0.97 0.44 ± 0.05 0.34–0.55 0.66 ± 0.01 0.56–0.77 0.94 LLWS 0.04 ± 0.00 0.03–0.04 0.00 ± 0.00 −0.01 to 0.00 0.84 0.64 ± 0.05 0.54–0.75 0.42 ± 0.05 0.32–0.53 0.95 LLBS 0.06 ± 0.01 0.05–0.07 0.04 ± 0.01 0.02–0.04 0.87 0.54 ± 0.06 0.43–0.66 0.52 ± 0.06 0.40–0.63 0.93 2 HLWW 0.74 ± 0.02 0.70–0.78 −0.01 ± 0.02 −0.05 to 0.03 0.97 0.45 ± 0.04 0.37–0.52 0.38 ± 0.04 0.31–0.46 0.96 HLBW 1.34 ± 0.03 1.27–1.40 0.08 ± 0.03 0.02–0.14 0.98 0.43 ± 0.04 0.35–0.50 0.56 ± 0.04 0.47–0.64 0.96 LLWW 0.02 ± 0.00 0.01–0.02 0.00 ± 0.00 0.00–0.01 0.67 0.47 ± 0.03 0.40–0.54 0.35 ± 0.04 0.27–0.42 0.96 LLBW −0.13 ± 0.03 −0.19 to −0.08 0.05 ± 0.01 0.03–0.07 0.74 0.43 ± 0.04 0.36–0.51 0.46 ± 0.04 0.38–0.54 0.96 Estimates of slopes 1 (dorsal profile) and 2 (ventral profile) are shown, along with the standard error (SE), 95% confidence intervals (CI) and adjusted r2 values The D–V gradients observed in the real fish, photographed after 2 weeks spent in the same experimental treatments, are shown in (b). Open in new tab (b) Fish D–V reflectance gradients The reflectance profiles measured from images of the real fish were approximately linear and different to those under the predictions of SSC (Table 3; Figure 3). In both experiments, there was no effect of light intensity, horizontal background, or their interaction on the dermal reflectance profiles of fish (all P > 0.05; see Supplementary Material, ESM3). Thus, the observed reflectance profiles of the fish did not meet the predictions of optimal patterning under a model of SSC, but displayed a linear radiance profile that is consistent with BM. Figure 3 Open in new tabDownload slide Observed reflectance profiles of all fish evaluated before (a) and after (b) exposure to treatments varying in light intensity and substrate reflectance (Experiment 1) and light intensity and horizontal background reflectance (Experiment 2). Position on the D–V axis refers to the standardized position on the fish’s body profile (when viewed laterally), where 0 is the most dorsal point and 100 is the most ventral point. Lines represent mean and shaded region represents ±1 SD; refer to Figure 2 for legend. Experiment 1 samples sizes: HLWS = 21, LLWS = 17, HLBS = 20, LLBS = 18; Experiment 2 sample sizes: HLWW = 12, LLWW = 11, HLBW = 12, LLBW = 11). Figure 3 Open in new tabDownload slide Observed reflectance profiles of all fish evaluated before (a) and after (b) exposure to treatments varying in light intensity and substrate reflectance (Experiment 1) and light intensity and horizontal background reflectance (Experiment 2). Position on the D–V axis refers to the standardized position on the fish’s body profile (when viewed laterally), where 0 is the most dorsal point and 100 is the most ventral point. Lines represent mean and shaded region represents ±1 SD; refer to Figure 2 for legend. Experiment 1 samples sizes: HLWS = 21, LLWS = 17, HLBS = 20, LLBS = 18; Experiment 2 sample sizes: HLWW = 12, LLWW = 11, HLBW = 12, LLBW = 11). DISCUSSION Increasing empirical and theoretical evidence suggests that concealment of shadows is the primary explanation for the evolution of countershading coloration in terrestrial animals (Allen et al., 2012; Penacchio et al., 2015a; Penacchio et al., 2015b; Cuthill et al., 2016). However, in an aquatic system, our modeling of optimal SSC under different visual conditions, coupled with the actual changes in fish dermal pigmentation that we observed, is consistent with a function of concealment by BM rather than concealment of self-shadows. Our findings support the notion that differences in the optical properties of air and water, as well as the viewing angles between predators and prey, may have resulted in the same countershading pattern, but one that has arisen under different evolutionary patterns of selection in terrestrial and aquatic environments. Our strongest evidence that color change does not conform to concealment under SSC is the finding that the reflectance gradients of our model clay fish under a model of optimum SSC were very different to those observed in the experimental fish under the same viewing conditions. Under the model of optimum SSC we expected a steeper reflectance gradient to counterbalance the higher ratio of downwelling: upwelling illuminance when fish were placed on a dark substrate and a shallower reflectance gradient due to the relative increase in upwelling light in environments with highly reflective substrates. We also expected that black horizontal backgrounds would result in steeper D–V reflectance gradients than white horizontal backgrounds because the ratio of downwelling to upwelling illumination is high (and results in increased ventral shadowing) when sidewelling radiance is low. However, our measurements of the reflectance gradient of fish skin pigmentation in our experimental conditions revealed no effect of substrate reflectance or horizontal background reflectance on fish D–V reflectance gradients, suggesting that changes in skin pigmentation do not function to counteract the production of self-shadows. Indeed, the D–V gradients of live fish were approximately linear, generating an even skin tone across the body, which minimizes strong contrast boundaries between the darker dorsal surface and pale underside that would otherwise increase conspicuousness to visual predators. Thus, it appears that both in the wild and in the laboratory (Kelley and Merilaita, 2015), fish adjust the relative amounts of melanin pigmentation on their dorsal and ventral surfaces, while maintaining an even transition from dark to light body coloration. Our findings suggest that fish appear to incorporate radiance information from multiple angles to inform changes in skin pigmentation and maximize BM from a variety of predator viewing angles. Substrate reflectance affected changes in fish skin pigmentation as predicted if fish need to conceal themselves from overhead predators, such as birds. Thus, the dorsal surface of fish became paler when fish were placed on a highly reflective (white) substrate than when on a less reflective (dark) substrate, suggesting that this change in skin reflectance serves for BM when viewed from above. However, the ventral reflectance of fish (which would not be visible to an aerial predator) also increased on a white background, suggesting that upwelling radiance influences BM from other viewing angles. Thus, in contrast to our predictions for concealment by SSC, fish are increasing the reflectance of their ventral surface to match the high upwelling radiance, which would facilitate BM when viewed from the side. It is unlikely that selection for BM when viewed from below could result in the evolution of a pale surface, because the amount of upwelling radiance (which would be required to illuminate the ventral surface) is orders of magnitude lower than the level of downwelling radiance, although it might be partially beneficial for animals living in shallow waters and over high reflectance substrates, for example, coral sand. Some animals produce their own light and bioluminescent counter-illumination is an effective method of camouflage (Young and Roper, 1976; Widder, 1999), especially for deep-sea fishes (Jerlov, 1976; Johnsen, 2014b). In freshwater fishes, changes in ventral reflectance for BM when viewed from below will always be suboptimal, yet a pale underside will still partially reduce contrast and hence may still confer a selective advantage. In both experiments, the level of observed color change was largely affected by previous skin reflectance. This could be indicative of the physiological constraints associated with melanin dispersion and aggregation. One possibility is that melanin aggregation may occur more rapidly than melanin dispersal, leading to the observed “carryover effects” associated with previous skin reflectance on pale backgrounds but not dark backgrounds. This situation might arise if fish are more likely to encounter dark backgrounds (e.g., dark refuges) in the wild than pale ones, making them predisposed to display “rapid” skin darkening for concealment in darkened habitats. We observed that changes in ventral reflectance on both black and white horizontal backgrounds were strongly dependent on previous skin (ventral) reflectance. There could be more constraints acting on changes in reflectance on the ventral surface than the dorsal one, for example, if ventral patterning plays a role in social signaling as well as camouflage (Kelley et al., 2016). It is important to note that the spectral range of artificial light is extremely restricted compared to that of natural sunlight (see Supplementary Material, ESM2, for comparison of irradiance in the wild and in the laboratory) and in aquaria, light is reflected off the surrounding glass walls rather than scattered and selectively absorbed by organic matter. Thus it is likely that the predictions of concealment using SSC differ in natural and artificial lighting environments, and some evidence for this is provided in the Supplementary Material (ESM4; Figure 1). Nonetheless, we predict that the reflectance profiles of fish measured in their natural habitats will not conform to the expectations of SSC, but will be similar (i.e., approximately linear profiles) to those observed in the experiments presented here (Supplementary Material, ESM4; Figure 2). Our experiments were not designed to replicate natural visual environments, but to disentangle the mechanisms of concealment by BM and SSC. Further field-based studies, incorporating variable lighting conditions and multiple predator viewing angles, are required to confirm that countershading serves for 3D BM in aquatic prey. Although SSC is the primary explanation for the evolution of countershading in terrestrial systems, it remains unclear how SSC influences the visual perception of the viewer (Penacchio et al., 2015a). Computer vision models have demonstrated that a camouflage breaking mathematical operator that detects the visual cues associated with prey convexity (i.e., 3D form) is very effective at detecting prey that are lit from overhead that generate self-shadows, but fails to detect simulated prey that exhibit countershading patterning (Tankus and Yeshurun, 2009). Detection might be impeded in countershaded prey due to the elimination of the strong transition gradient generated by light falling unevenly on the body surface, making the 2D difference in light intensity less detectable to the viewer by reducing intra-body contrast. Alternatively, prey may be less detectable because the visual cues that are used by predators to detect 3D objects are absent. These scenarios can only be addressed using models of predator 3D perception and by linking prey contrast with predator detection to determine whether predators actively search for 3D prey. This warrants further attention, as most studies of prey patterning have not considered the 3D component of predator-prey dynamics. In summary, our observations of the optimal “color” change responses of fish, coupled with our modeling of SSC for different visual conditions and viewing angles suggests that SSC is unlikely to account for the evolution of countershading patterning in aquatic animals. This supports the view that the physical properties of the light environment in terrestrial and aquatic environments have led to divergent mechanisms of concealment but a similar pattern phenotype. Aquatic environments are likely to have favored BM as a strategy, allowing for overall concealment in the open water, where there are few opportunities to hide. SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online. FUNDING J.L.K. was supported from an Australian Research Council Linkage Project (LP120200002) with industry partners Rio Tinto and BHP Billiton and The School of Animal Biology at The University of Western Australia provided the research funds for this project. J.C.P. was supported for this work from The Gorgon Barrow Island Net Conservation Benefits Fund, and Australian Research Council Discovery Project Grants DP160102658 and DP14010211. We are extremely grateful to Sami Merilaita, Callum Donohue, and our anonymous reviewers for providing valuable comments that improved this manuscript. We would also like to thank John Endler, Jan Hemmi, Laura Kelley, and Hannah Rowland for insightful discussions on countershading patterning. Data accessibility: Analyses reported in this article can be reproduced using the data provided by (Kelley et al., 2017). REFERENCES Allen WL , Baddeley R , Cuthill IC , Scott-Samuel NE . 2012 . A quantitative test of the predicted relationship between countershading and lighting environment . Am Nat . 180 : 762 – 776 . Google Scholar Crossref Search ADS PubMed WorldCat Beddard FE . 1895 . Animal coloration. An account of the principal facts and theories relatiing to the colours and marking of animals . London : Swan Sonnenschein & Co . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Caro T . 2009 . Contrasting coloration in terrestrial mammals . Philos Trans R Soc Lond B Biol Sci . 364 : 537 – 548 . Google Scholar Crossref Search ADS PubMed WorldCat Claes JM , Aksnes DL , Mallefet J . 2010 . Phantom hunter of the fjords: camouflage by counterillumination in a shark (Etmopterus spinax) . J Exp Mar Biol Ecol . 388 : 28 – 32 . Google Scholar Crossref Search ADS WorldCat Cott HB . 1940 . Adaptive coloration in animals . London : Methuen & Co. Ltd . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Cuthill IC , Sanghera NS , Penacchio O , Lovell PG , Ruxton GD , Harris JM . 2016 . Optimizing countershading camouflage . Proc Natl Acad Sci . 113 : 13093 – 13097 . doi: 10.1073/pnas.1611589113 . Google Scholar Crossref Search ADS PubMed WorldCat Cuthill IC , Stevens M , Sheppard J , Maddocks T , Párraga CA , Troscianko TS . 2005 . Disruptive coloration and background pattern matching . Nature . 434 : 72 – 74 . Google Scholar Crossref Search ADS PubMed WorldCat De Ruiter L . 1956 . Countershading in caterpillars . Archives Neerlandaises de Zoologie . 11 : 285 – 341 . Google Scholar Crossref Search ADS WorldCat Edmunds M , Dewhirst RA . 1994 . The survival value of countershading with wild birds as predators . Biol J Lin Soc . 51 : 447 – 552 . Google Scholar Crossref Search ADS WorldCat Gray SM , Hart FL , Tremblay MEM , Lisney TJ , Hawryshyn CW . 2011 . The effects of handling time, ambient light, and anaesthetic method, on the standardized measurement of fish colouration . Canad J Fish Aquat Sci . 68 : 330 – 342 . Google Scholar Crossref Search ADS WorldCat Hailman JP . 1977 . Optical signals: animal communication and light . London : Indiana University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jerlov NG . 1976 . Marine optics . Amsterdam, The Netherlands: Elsevier . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Johnsen S . 2002 . Cryptic and conspicuous coloration in the pelagic environment . Proc R Soc Lond B Biol Sci . 269 : 243 – 256 . Google Scholar Crossref Search ADS WorldCat Johnsen S . 2014a . Hide and seek in the open sea: pelagic camouflage and visual countermeasures . Ann Rev Mar Sci . 6 : 369 – 392 . Google Scholar Crossref Search ADS WorldCat Johnsen S . 2014b . Hide and seek in the open sea: pelagic camouflage and visual countermeasures . In: Carlson CA , Giovannoni SJ , editors. Annual review of marine science . Vol 6 . Palo Alto (CA) : Annual Reviews . p. 369 – 392 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Kelley JL , Fitzpatrick JL , Merilaita S . 2013 . Spots and stripes: ecology and colour pattern evolution in butterflyfishes . Proc Biol Sci . 280 : 20122730 . Google Scholar Crossref Search ADS PubMed WorldCat Kelley JL , Merilaita S . 2015 . Testing the role of background matching and self‐shadow concealment in explaining countershading coloration in wild‐caught rainbowfish . Biol J Lin Soc . 114 : 915 – 928 . Google Scholar Crossref Search ADS WorldCat Kelley JL , Rodgers GM , Morrell LJ . 2016 . Conflict between background matching and social signalling in a colour-changing freshwater fish . R Soc Open Sci . 3 : 160040 . Google Scholar Crossref Search ADS PubMed WorldCat Kelley JL , Taylor I , Hart NS , Partridge JC . 2017 . Data from: aquatic prey use countershading camouflage to match the visual background . Dryad Digital Repository . http://dx.doi.org/10.5061/dryad.g669j . OpenURL Placeholder Text WorldCat Kiltie RA . 1988 . Countershading: universally deceptive or deceptively universal? Trends Ecol Evol . 3 : 21 – 23 . Google Scholar Crossref Search ADS PubMed WorldCat Kjernsmo K , Grönholm M , Merilaita S . 2016 . Adaptive constellations of protective marks: eyespots, eye stripes and diversion of attacks by fish . Animal Behaviour . 111 : 189 – 195 . Google Scholar Crossref Search ADS WorldCat Körner HK . 1982 . Countershading by physiological colour change in the fish louse Anilocra physodes L. (Crustacea: Isopoda) . Oecologia . 55 : 248 – 250 . Google Scholar Crossref Search ADS PubMed WorldCat Merilaita S , Stevens M . 2011 . Crypsis through background matching . In: Stevens M , Merilaita S , editors. Animal Camouflage . Cambridge, UK : Cambridge University Press . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Penacchio O , Lovell PG , Cuthill IC , Ruxton GD , Harris JM . 2015a . Three-dimensional camouflage: exploiting photons to conceal form . Am Nat . 186 : 553 – 563 . Google Scholar Crossref Search ADS WorldCat Penacchio O , Cuthill IC , Lovell PG , Ruxton GD , Harris JM . 2015b . Orientation to the sun by animals and its interaction with crypsis . Funct Ecol . 29 : 1165 – 1177 . Google Scholar Crossref Search ADS WorldCat Poulton EB . 1888 . Notes in 1887 upon lepidopterus larvae & c . Trans Entomol Soc London . 1888 : 595 – 596 . OpenURL Placeholder Text WorldCat Poulton EB . 1890 . The colours of animals, their meaning and use, especially considered in the case of insects . New York : D. Appleton and Company . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Rodgers GM , Kelley JL , Morrell LJ . 2010 . Colour change and assortment in the western rainbowfish . Animal Behav . 79 : 1025 – 1030 . Google Scholar Crossref Search ADS WorldCat Rowland HM . 2009 . From Abbott Thayer to the present day: what have we learned about the function of countershading? Philos Trans R Soc Lond B Biol Sci . 364 : 519 – 527 . Google Scholar Crossref Search ADS PubMed WorldCat Rowland HM , Cuthill IC , Harvey IF , Speed MP , Ruxton GD . 2008 . Can’t tell the caterpillars from the trees: countershading enhances survival in a woodland . Proc Royal Society B . 275 : 2539 – 2545 . Google Scholar Crossref Search ADS WorldCat Rowland HM , Speed MP , Ruxton GD , Edmunds M , Stevens M , Harvey IF . 2007 . Countershading enhances cryptic protection: an experiment with wild birds and artificial prey . Animal Behav . 74 : 1249 – 1258 . Google Scholar Crossref Search ADS WorldCat Ruxton GD , Sherratt T , Speed MP . 2004a . Avoiding attack: the evolutionary ecology of crypsis, warning signals and mimicry . Oxford, UK : Oxford University Press . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Ruxton GD , Speed MP , Kelly D . 2004b . What, if anything, is the adaptive function of countershading? Animal Behav . 68 : 445 – 451 . Google Scholar Crossref Search ADS WorldCat Sköld HN , Aspengren S , Cheney KL , Wallin M . 2016 . Chapter four-Fish chromatophores—from molecular motors to animal behavior . Int Rev Cell Mol Biol . 321 : 171 – 219 . Google Scholar Crossref Search ADS PubMed WorldCat Stevens M , Cuthill IC . 2006 . Disruptive coloration, crypsis and edge detection in early visual processing . Proc Biol Sci . 273 : 2141 – 2147 . Google Scholar Crossref Search ADS PubMed WorldCat Tankus A , Yeshurun Y . 2009 . Computer vision, camouflage breaking and countershading . Philos Trans R Soc Lond B Biol Sci . 364 : 529 – 536 . Google Scholar Crossref Search ADS PubMed WorldCat Thayer AH . 1896 . The law which underlies protective coloration . Auk . 13 : 124 – 129 . Google Scholar Crossref Search ADS WorldCat Turner ERA . 1961 . Survival values of different methods of camouflage as shown in a model population . Proc Zoo Soc London. Wiley Online Library . 136 : 273 – 284 . Google Scholar Crossref Search ADS WorldCat Wallace AR . 1889 . Darwinism: an exposition of the theory of natural selection - with some of its applications . London and New York : Macmillan and Co . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Widder EA . 1999 . Bioluminescence . In: Archer SN , Djamgoz MBA , Loew ER , Partridge JC , Vallerga S , editors. Adaptive mechanisms in the ecology of vision . Boston : Kluwer . p. 555 – 581 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Young RE , Roper CF . 1976 . Bioluminescent countershading in midwater animals: evidence from living squid . Science . 191 : 1046 – 1048 . Google Scholar Crossref Search ADS PubMed WorldCat Author notes " Handling editor: Bob Wong © The Author 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
TI - Aquatic prey use countershading camouflage to match the visual background
JF - Behavioral Ecology
DO - 10.1093/beheco/arx093
DA - 2017-09-01
UR - https://www.deepdyve.com/lp/oxford-university-press/aquatic-prey-use-countershading-camouflage-to-match-the-visual-shSOQzO5vU
SP - 1314
VL - 28
IS - 5
DP - DeepDyve
ER -