Contrasting stripes are a widespread feature of group living in birds, mammals and fishes
Grouping is a widespread form of predator defence, with individuals in groups often performing evasive collective movements in response to attack by predators. Individuals in these groups use behavioural rules to coordinate their movements, with visual cues about neighbours’ positions and orientations often informing movement decisions. Although the exact visual cues individuals use to coordinate their movements with neighbours have not yet been decoded, some studies have suggested that stripes, lines, or other body patterns may act as conspicuous conveyors of movement information that could promote coordinated group movement, or promote dazzle camouflage, thereby confusing predators. We used phylogenetic logistic regressions to test whether the contrasting achromatic stripes present in four different taxa vulnerable to predation, including species within two orders of birds (Anseriformes and Charadriiformes), a suborder of Artiodactyla (the ruminants), and several orders of marine fishes (predominantly Perciformes) were associated with group living. Contrasting patterns were significantly more prevalent in social species, and tended to be absent in solitary species or species less vulnerable to predation. We suggest that stripes taking the form of light-coloured lines on dark backgrounds, or vice versa, provide a widespread mechanism across taxa that either serves to inform conspecifics of neighbours' movements, or to confuse predators, when moving in groups. Because detection and processing of patterns and of motion in the visual channel is essentially colour-blind, diverse animal taxa with widely different vision systems (including mono-, di-, tri-, and tetrachromats) appear to have converged on a similar use of achromatic patterns, as would be expected given signal-detection theory. This hypothesis would explain the convergent evolution of conspicuous achromatic patterns as an antipredator mechanism in numerous vertebrate species.
Group living characterizes the lives of many animals including flocking birds, schooling fishes, and herding ungulates. Grouping is often a defensive strategy to reduce the likelihood of individuals being attacked through detection, dilution, confusion, or self-herd effects [1–4]. Because isolated individuals are more likely to be targeted by predators [5–8], individuals in groups often move to maintain cohesion by escaping from predators in the same direction as other group members [9,10]. Many models aimed at explaining such coordinated movements assume animals rely primarily on visual information to infer neighbours' movements and positions , although the mechanosensory lateral line system may also play a role in communication among social fishes [12,13]. These assumptions seem appropriate considering that birds are highly visually oriented animals [14,15], and many fishes and mammals similarly rely on vision when interacting with conspecifics .
Because vision appears essential for many species relying on coordinated and cohesive escape from predators, particular visual cues may be used by social animals to promote coordination. Numerous grouping species display conspicuous stripes on their bodies (figure 1), with differences in design related to their particular body plans. In mammals and fishes, the majority of which have elongated bodies parallel to the substrate, stripes often run parallel to the longitudinal axis of the body. In birds, contrasting lines are often displayed in open wings. In all groups, stripes are typically achromatic. Light-coloured lines, often pure white, appear on an otherwise darker (i.e. melanized) plumage. Brooke  found that flocking shorebirds tended to have white wing stripes, while solitary species did not, and suggested that these white marks could provide a conspicuous flash to other group members promoting cohesive take-off, or could aid in coordinated movement. In such species, a role of sexual selection as a driver of such body patterns may be dismissed, as these marks are shown year-round by both males and females in otherwise sexually dichromatic species that display both breeding and non-breeding plumages . Whether the stripes in other social taxa, in particular the contrasting bands on the bodies of some social mammals, such as gazelles , or the stripes of different species of fish [19–22] play a role in coordination remains unclear.
Others have suggested that such body stripes, when combined in a moving group, may produce ‘dazzle’ patterns that could reduce either the likelihood of predators attacking or their attack success rates [6,7,23]. Contrasting stripes or lines have been hypothesized to provide camouflage or function to startle predators when they are suddenly displayed by fleeing prey [7,24,25,26]. Dazzle patterns may also reduce the ability of a predator to judge the speed and direction of their prey . Although there are no empirical data from wild animals, computer simulations using humans as model predators suggest that striped targets are among the most difficult to capture. Nonetheless, simpler patterns, such as all-grey colourations, are equally difficult to catch, and thus evidence that stripes play an adaptive role in dazzle patterning remains inconclusive . Earlier studies comparing dorsal patterns of snakes concluded that longitudinal stripes tended to occur in species with antipredator strategies favouring fast escape over active defence (e.g. ). On the other hand, garter snakes (Genus Tamnophis) display highly contrasting longitudinal stripes and include the most social snakes, with communal hibernation and mating . These investigations in reptiles already point to potential links between striped patterns and both antipredator and social behaviour.
Expanding on Brooke's  original idea for shorebirds, here, we separately analyse colouration data on shorebirds, waterfowl, ruminants, and fishes, while controlling for phylogeny, to assess whether social species are more likely to display stripes on their bodies and investigate their possible function. We posit that social species may use contrasting stripes to aid in coordinated movement or dazzle predators during coordinated movement. According to these hypotheses, we predicted that for any given taxa, contrasting stripes should be more prevalent in group-living species than in solitary ones. Here, we test this prediction in species that rely on group cohesion when fleeing from attacking predators, including herding ungulates, flocking birds, and schooling fishes.
2. Material and methods
(a) Study species
We selected three different vertebrate taxa: two orders of birds (shorebirds within Charadriiformes, and waterfowl (Anseriformes)), one suborder of mammals (Ruminantia), and fishes in the community of the Eastern Tropical Pacific Ocean with species in the Orders: Anguilliformes, Clupeiformes, Beloniformes, Perciformes, Scombriformes, and Istiophoriformes, all of which include species which have contrasting achromatic stripes (i.e. dark and light lines) (figure 1), as well as other species with no such patterns. We investigated two large bird groups where the presence/absence of wing marks had caught the attention of previous researchers (i.e.  for shorebirds and  for Anseriformes). In these bird groups, stripes are only exposed in flight, and thus potentially when the individuals attempt to escape from a predator. Both shorebirds and ducks are more vulnerable and exposed to a large set of predators when on land or on the water. In those situations, camouflage may be used to remain undetected (with stripes being hidden), but when detected by predators, stripes become exposed during flight. Mammals and fishes are different in that they do not have foldable limbs, with stripes being permanently visible. For these groups, we studied the most prominent stripes; longitudinal stripes in mammals, and both longitudinal and vertical stripes in fishes. The latter were analysed separately as previous studies reported that stripe orientation in fishes is linked to body shape , and to different ecological and behavioural traits (e.g. ). The taxa under investigation included both solitary and social species that often forage or rest in groups. In these taxa, individuals that detect the predator or are attacked often increase their speed during escape movements, alerting others of the predator's attack [31,32].
We used the same dataset and species' ecological attributes as Brooke  who assessed the occurrence of stripes in shorebirds (excluding gulls, auks, and allies). Our study included 205 species in this order; although Brooke's previous study included 210 (we excluded five species unavailable in the taxonomy offered in BirdTree, see below). Brooke investigated several ecological variables, such as migration, habitat, feeding technique, and flocking behaviour and found that only flocking behaviour was related to the presence of stripes on the wings. For our analysis, we included (a) whether the species was mainly migratory or sedentary, as the marks might help in group formation during migration trips, (b) whether the species tended to form flocks or was mainly solitary, and (c) wing-chord length as a proxy for overall size. For flocking behaviour, Brooke  followed Hayman et al.  and classified species as non-flocking if terms such as ‘singly’ or ‘in small groups’ were used in their respective descriptions. Species were classified as flocking if the term ‘flock’ was used, or if groups exceeding 20 birds were mentioned. Our response variable was the presence/absence of white lines on the wings (1 present, 0: absent). We did not analyse patches on the tails and rumps that were not significantly linked to sociality in Brooke's  previous study.
To capture the difference in stripe expression that we knew a priori distinguished the smaller ducks from the larger geese and swans  we included 148 representative members of all families of waterfowl (this order comprises 180 species in total). We used a binomial response variable indicating the presence or the absence of contrasting stripes on the wings following Hegyi et al. . In particular, we unified their scores of 1, 2, and 3 into a single score, i.e. 1 = presence of white patch in the wings, 0 = absence of white patch. Hegyi et al.  did not include geese and swans (Anserinae) in their analyses, as these species rarely have wing patches. For the few species that do present white wing patches, we scored flying birds as displayed in the identification guide of Madge and Burn , and followed a similar procedure to Hegyi et al.  as described above. For species that are predominantly white, like some swans, elongated dark patches on the wings were scored as presence of stripes.
In contrast to the other taxa included in this study, a majority of the species of Anseriformes are highly social during at least part of their annual cycle. Therefore, for this order, we could not distinguish between solitary or social species. As an explanatory variable, therefore, we used body mass as a proxy for body size, and a binomial variable indicating if the species is migratory or resident (again to assess whether marks might assist group formation during migrations). We predicted that wing marks would be associated with smaller ducks that are often hunted in the air by falcons, whereas these lines should be largely absent in the larger geese and swans that are rarely hunted by aerial predators. The different susceptibility to predation according to size affects their flying behaviour: ducks rank among the faster flyers among birds and tend to fly in dense flocks, whereas geese and swans fly with a lower flapping rate and usually adopt v-flight formations .
Given the sexual dimorphism in many Anseriformes and the absence of sexual differences in wing patch expression , we selected male body mass to increase variation (female body mass produced similar results—not shown). Body mass data were taken from Figuerola & Green . The migratory status of the species was taken from the distribution maps of the identification guide by Madge & Burn . For partially migratory species, we classified the species as migratory if the area of breeding range abandoned in winter was larger than the resident breeding area (present all year around).
Many species within the Artiodactyla, to which Ruminantia belong, have linear contrasting stripes on their flanks and dorsum (i.e. narrow bands or stripes differing in colour from the surface on either side of them). Some species may have other markings of different shapes (i.e. linear or not) on their legs, rump, and faces. Caro & Stankowich  previously characterized the presence and absence of contrasting marking in different body parts in 198 species of artiodactyls and investigated their function. Here, we built our own dataset to assess whether the longitudinal stripes on the flanks of Ruminantia are related to their degree of sociality. Similar to above, and to remain consistent, we recorded the presence or absence of longitudinal stripes on the bodies of animals as a binary response variable, given the definition of a longitudinal marking by Caro & Stankowich  as a stripe in both flanks contrasting in colouration with the belly below and the dorsum above. Mammals may also have vertical stripes, as with the patterning of zebras, but in the Ruminantia that we studied (e.g. Tragelaphus species) these are very narrow and faint lines that possibly serve to provide crypsis , and therefore were not considered in our analysis. Social species were defined as the ones forming herds including more than one parent–offspring unit or more than three unrelated individuals. Conversely, solitary species are defined as the ones in which individuals tend to live alone or in parent–offspring units. Smaller markings on other body parts (i.e. legs, neck, or head) were not considered in the present study, although they may serve a similar function. Body size data were obtained from Smith et al. .
We selected representative species of the main Class of fishes, the Actinopterygii. Among them, we chose species from the fish community of the Eastern Tropical Pacific Ocean, following the guide by Allen & Robertson  and the online-actualized version https://biogeodb.stri.si.edu/sftep/en/pages. Because many tropical species of fish can be colourful, we chose species with several patterns, including species that were both striped and unstriped on the sides of their body. We analysed vertical and horizontal lines separately, as they have been suggested to convey different visual information, or relate to different life-history traits and habitat features [19,21,22]. Our searching criteria were as follows: first, and following the phylogenetic order of the book, we identified species with some form of stripe, either parallel to the horizontal axis—i.e. horizontal stripes or down the vertical axis—i.e. vertical stripes (or bars as in Seehausen et al. ). After this initial selection, we went through our database, and for each species with vertical or horizontal stripes, we randomly identified another species within the same genus, or genus within the same family, that did not have either of those stripes (plain species as in ). In our final database of fishes, most species (90%) belonged to the Perciformes, but with some representatives of additional Orders (Clupeiformes, Beloniformes, Scombriformes, and Istiophoriformes, and just two species of Anguilliformes). Finally, and to align coherently with the other taxa in this study, we only retained primary prey species (defined as species that remain small enough to be eaten by larger fish throughout their lives), and recorded as binary response variables the presence/absence of one or more stripes (of any colour), and their sociality level (i.e. solitary versus schooling). For assignment of sociality, we followed the natural history notes of Allen & Robertson . Some species are always in schools (sardines, mackerels), others aggregate for migrations (and we included them as schooling) and for others, especially the reef species, we considered them as social if the minimum size of groups reported exceeded 50 individuals in any part of their life cycle.
3. Phylogenetic analysis
We tested whether the presence of stripes on the body correlated with the evolution of group living, or other attributes such as body mass or ecological characteristics, using phylogenetic logistic regression [39–44]. In comparative analyses of trait evolution, both ‘complete’ trees (i.e. those that include unsampled species using polytomy solvers) and ‘sequenced species only’ trees have their advantages and disadvantages . First, birth–death polytomy resolvers used to infer ‘complete trees’ place species onto trees randomly and independently from the trait's values, thus potentially breaking down natural patterns of trait phylogenetic structure . Similarly, ‘sequenced species only’ trees, while not suffering from this problem, contain fewer species that are potentially non-randomly sampled, risking excluding their trait values from analyses . Accordingly, here we followed Upham et al.'s  recommendations and compared analyses run on samples from both complete and sequenced-only trees (see electronic supplementary material).
For each taxonomic group, we computed a 50% majority-rule consensus tree using SumTrees v. 4.4.0 in DendroPy v. 4.4.0 [47,48] following the guidelines of Rubolini et al.  to summarize individual gene trees for comparative studies. By using this method, we obtained a single (optimal) consensus tree for each group. According to Holder et al. , a 50% majority-rule consensus tree can be seen as an optimal summary of the posterior distribution of Markov Chain Monte Carlo (MCMC) trees. We downloaded phylogenetic subsets from broad phylogenetic syntheses. For the two bird taxa, we used BirdTree (, http://birdtree.org). In particular, we downloaded phylogenetic trees from the Hackett tree distributions (1000 trees from ‘All species’ + 1000 trees from ‘Sequenced species’ distributions). For the Ruminantia taxa, we used the species-level trees of extant Mammalia of Upham et al.  (1000 trees from ‘birth–death node-dated completed trees’ + 1000 trees from ‘birth–death node-dated DNA-only’ distributions). Lastly, for fish taxa, we downloaded data from the fish tree of life (100 trees from ‘all-taxon assembled’ distributions plus the ‘DNA-only’ time-calibrated tree; ). The final consensuses ‘complete trees’ were composed of 205 Charadriiformes, 148 Anseriformes, 197 Ruminantia, and 159 fish species, and the final consensuses ‘sequenced species only’ trees were composed of 144 Charadriiformes, 137 Anseriformes, 182 Ruminantia, and 120 fish species.
To assess whether stripes were associated with group-living behaviour, we used phylogenetic logistic regression analyses with the function phyloglm (logistic_MPLE method, 1000 independent bootstrap replicates, and p-values computed using Wald tests) in the phylolm R package v. 2.6 [39–44]. We evaluated model fit with binned residual plots—specifically, whether standard error bands from these plots contain 95% of the binned residuals—(binnedplot function from the arm v. 1.10-1 R package; ) and computing an R2 statistic suitable for statistical models with correlated errors, such as the phylogenetic logistic regression models used here (R2 function from the rr2 R package; ). Then, we used the fitted coefficients from the phyloglm models to plot the phylogenetic logistic regression curves using the plogis function of the R package ‘stats’.
We found significant associations between presence of stripes and the level of sociality in all taxa that we investigated, and these results were consistent across complete and sequenced-only trees. Flocking Charadriiformes (shorebirds) were more likely to have white wing stripes than non-flocking species (phyloglm: , p = 0.001; table 1, figures 2a and 3a). Migratory status or body size (wing-chord length), however, were not significantly related to the presence or absence of wing markings (phyloglm: migratory status, p = 0.599; body size, p = 0.121; table 1). In Anseriformes (waterfowl), most species of ducks had white wing stripes on dark backgrounds, whereas most of the geese did not. In these species, body mass was significantly correlated to the presence or the absence of stripes, with smaller species more likely to have stripes than larger ones (phyloglm: , p = 0.001; table 1, figures 2b and 3b). Similar to the Charadriiformes, migratory status was not related to the presence or absence of wing stripes (phyloglm: , p = 0.346; table 1). A model excluding the three seemingly outlying data points (i.e. species with body mass greater than 10 000 g) yielded essentially similar results (phyloglm: ; body mass, p = 0.003; migratory status, p = 0.337).
Social species of ruminants were also more likely to have stripes on the flanks than non-social species (phyloglm: ; p = 0.003; table 1, figures 2c and 3c). Body mass, however, was not related to the presence or the absence of stripes (phyloglm: body mass, p = 0.133; table 1).
In fishes, sociality was associated with the presence or the absence of horizontal stripes on the body, with social species more likely to have horizontal stripes (phyloglm: ; p < 0.001; table 1, figures 2d and 3d). Vertical lines (bars), however, were associated in the opposite direction to social species (i.e. social species were less likely to have vertical stripes; phyloglm: vertical stripes, ; p = 0.009; table 1).
Results for the subset of species for which actual sequence data are available are given as electronic supplementary material, and include phyloglm models (electronic supplementary material table S1), and phylogenetic trees (electronic supplementary material, figure S1).
Stripes on the plumage, fur, and scales of prey species were strongly associated with group living across taxa. While different roles for the contrasting marks on animals' bodies had previously been suggested in independent studies of shorebirds, waterfowl, ruminants, and fishes [17,18,21,22,30], evidence for their importance in social behaviour remained inconclusive. In the all-social waterfowl, wing stripes were related to the species' body mass; a strong predictor of predation risk [55,56]. Given this, and given predation is a strong driver of group living, we suggest these contrasting lines play an important role in social antipredator defences of at least three different vertebrate classes; mammals, birds, and fishes.
The finding that stripes are associated with social behaviour is even more striking considering the visual systems of these taxa are known to vary widely. However, the combination of adjacent achromatic lines or bands – mostly black/dark and white (except in some fishes where structural colours also play a role) – suggests that colour perception is not necessary for their role. Indeed, detection of patterns and of motion is achieved by achromatic mechanisms, and thus is essentially colour-blind . These different taxa, therefore, appear to have converged on a similar use of achromatic patterns, as would be expected given signal-detection theory. Indeed, white bands with no pigment combined with melanized areas may be generated by virtually any vertebrate , and the stripes in the groups we studied were indeed overwhelmingly monochromatic. Hence, these markings may be just as efficiently detected in either mono-, di-, tri-, or tetra-chromatic species [59–61], and visible in scotopic conditions where most species' vision becomes effectively monochromatic owing to only rods remaining functional . It is noteworthy, however, that the species in our study tend to live in open habitats. Anseriformes and Charadriiformes often live on open water, marshlands, and coastlines, and the ruminants in our study tend to graze on open grasslands . Stoner et al.  also noted that artiodactyls with contrasting lines on their flanks were diurnal and lived in open habitats. The fishes in our sample similarly live in illuminated shallow waters. Hence, while nothing precludes that contrasting stripes remain functional both in full daylight and in dim light conditions, detection of these stripes is likely to be easier and more applicable to diurnal species that live in relatively open habitats where contrasting stripes are more conspicuous.
While the association between stripes and social behaviour appears widespread, these markings’ functional role remains unresolved. Two primary (non-mutually exclusive) functions have been proposed for these stripes. First, they could be used by conspecifics to inform neighbours of rapid movement and aid in coordinated movement , or second, to reduce predation success through flash, dazzle, or confusion effects [7,23]. In many cases of birds (including the Charadriiformes and Anseriformes studied here) wing markings are often only exposed when a bird takes flight, but remain hidden when the bird folds its wings, with cryptic dorsal plumages intended for background matching . This sudden display of contrasting lines has led authors to consider them primarily as flash or dazzle marks destined to confuse an approaching predator that might misjudge the speed or direction of the prey's escape . While such an effect could also be achieved by a single individual exposing its wings during its escape flight, the flashes from multiple individuals would likely produce a stronger stimulus, and could hence be used by social animals to reduce per capita predation risk. Indeed, because predators prioritize distance estimation, their perception of horizontal contours may be disrupted by such movements, viewing such collectively moving prey as blurred images . There is also the suggestion that predators and prey may have different systems for distance estimation and object recognition, evidenced by their often different pupil shape—vertically elongated in ambush predators such as cats and vipers, and horizontally elongated in prey species such as the Ruminantia . Therefore, while stripes may confuse predators from long-range, at short-range, these markings may inform individuals of neighbours' motion, indirectly indicating to others the presence of a predator, and perhaps aiding in more coordinated take-offs .
In contrast to the wing markings of birds that are only exposed when birds fly, longitudinal lines of mammals and fish are permanently visible. What roles do these markings play in social behaviour? Caro & Stankowich  suggested several functions for the stripes in Ruminantia, including pursuit deterrence, intraspecific signalling, and crypsis. In fishes, some authors have suggested that stripes may disrupt the body shape making it unrecognizable for predators  or that stripes may make it difficult for a predator to focus on a specific individual target . Indeed, because visual acuity scales with eye or body size in ray-finned fishes , larger predatory species are more likely to be able to resolve stripes than smaller shoaling species. This may suggest that stripes are important in the confusion effect, horizontal lines serving to reduce the ability of predators to judge the speed and direction of prey . Denton & Rowe  further suggest that stripes in the mackerel Scomber scombrus help to coordinate shoaling behaviour, because stripes are perceived differently with changes in body orientation (although they referred to the vertical stripes of the dorsal side of a fish's body). Once again, therefore, the association of grouping with these stripes suggests that both confusion and coordination effects may play an important role for these species. Further studies on the functional role of horizontal and vertical stripes are clearly warranted.
In fishes, however, we found opposite trends in relation to whether horizontal (longitudinal) or vertical (bars) stripes were associated with social behaviour. Shoaling fishes were less likely to have vertical stripes, but more likely to have horizontal stripes. These findings are consistent with other studies, where open-water cichlids that engage in shoaling behaviour tend to have horizontal stripes , whereas cichlids that engage in aggressive territory defence are likely to have vertical stripes [68–71]. Moreover, the presence of a horizontal stripe is thought to reduce aggressiveness in neotropical and West African riverine cichlids [69–71]. And while neither vertical nor horizontal stripes were associated with social behaviour in butterflyfishes, diagonal stripes were negatively correlated with butterflyfishes' shoaling tendencies . Our wider taxonomic analysis, therefore, appears largely consistent with previous research documenting both positive and negative effects of differently orientated stripes on social behaviour. Indeed, Barlow  previously concluded that stripes and bars act as both social signals and antipredator adaptations in fishes.
While our result confirms many social species have stripes and contrasting marks, there are species which form large moving groups in open environments without any conspicuous contrasting markings. Indeed, the European starling Sturnus vulgaris, that does not display visually contrasting patterns in the plumage , yet displays some of the most coordinated and collective behaviour in the animal kingdom [73–77]. Clearly, stripes are not a prerequisite for coordinated and cohesive movement. Similarly, the presence of stripes is not entirely restricted to social species, with some stripe patterns functioning as aposematic signals, as in striped skunks Mephitis mephitis , or deimatic signals that startle predators . Stripes may also have other functions, for instance, melanin-based plumage patches have been suggested to signal dominance in social passerines, with black patches located around the head and breast being assessed during close encounters with conspecifics . Also in passerines, a study on leaf warblers (Phylloscopidae) with varying widths of wing bars revealed that species with the broader wing bars lived in darker habitats. Displaying males, therefore, attempt to increase their conspicuousness to competing conspecifics in darker environments . Indeed, these species are solitary, and their contrasting wing marks have evolved in contexts different to predation avoidance or grouping. In addition, smaller head stripes in numerous taxa may have evolved to conceal the eye, a prominent structure in most species that may attract the attention of predators or alert prey . These eye stripes appear to be a case of coincident disruptive camouflage [24,81] and would therefore not be linked to increased sociality. Nevertheless, while there are cases where social species have no stripes, and cases where stripes serve multiple functions in animals, our results suggest that all else being equal, social species are more likely to have stripes than non-social species.
Using a comparative method where we accounted for phylogenetic signal, we found that contrasting and achromatic stripes are more likely to be present across a broad taxonomic range of social species that are often predated upon. It appears that these stripes may aid in group coordination, or act to increase the confusion effect, both of which may reduce per capita predation risk. These achromatic patterns even suggest that communication signals for alerting and recruiting others may even work within mixed-species groups that are so prevalent among social birds (including ducks and waders), mammals (including gazelles), and fish . Our work represents a rare example of aligned phenotypic characteristics across a wide range of taxonomically diverse animals.
All data needed to evaluate the conclusions in the paper are present in the paper and/or the electronic supplementary materials. Supplementary data (raw datasets) are available at Figshare (doi:10.6084/m9.figshare.12423623). An earlier version of the manuscript was uploaded as a preprint in BioRxiv .
J.J.N. conceived the hypotheses, designed the study, contributed to the dataset, and drafted the manuscript; J.D. carried out the statistical analyses, helped in drafting the manuscript, and prepared figures and supplementary material; M.C.B. contributed to the dataset and critically revised the manuscript; A.R. contributed to the dataset and helped draft the manuscript. J.H.R. refined hypotheses and helped draft the manuscript. M.B. contributed to the dataset and helped draft the manuscript. All authors gave final approval for publication and agree to be held accountable for the work performed therein.
We declare we have no competing interests.
A.R. was supported by Juan de la Cierva programme, Spanish Ministry of Economy, Industry and Competitiveness (grant no. IJCI-2015-23913). J.D. was supported by a Marie Curie Global fellowship (grant no. 886532). J.E.H.-R. was supported by the Whitten Lectureship in Marine Biology, and a Swedish Research Council grant no. 2018-04076.
We thank Enrique Figueroa-Luque for his help in building the dataset for mammals, and Christopher Swann for the fish picture. Three anonymous reviewers greatly helped to improve the first version of the manuscript.
Published by the Royal Society. All rights reserved.
Cott HB. 1940Adaptive coloration in animals. London, UK: Methuen & Co. Google Scholar
Hayman P, Marchant J, Prater T. 1986Shorebirds: an identification guide to the waders of the world. London, UK: Christopher Helm. Google Scholar
Madge S, Burn H. 1987Wildfowl: an identification guide to the ducks, geese and swans of the world. London, UK: Helm. Google Scholar
Allen GA, Robertson DD. 1998Peces del PacÍfico oriental tropical. México: Smithsonian Tropical Research Institute. Conabio-Sierra Madre-CEMEX. Google Scholar
McGraw KJ. 2006Mechanics of melanin-based coloration. In Bird coloration, vol. 1. Mechanisms and measurements (eds Hill GE, McGraw K), pp. 243-294. Cambridge, MA: Harvard University Press. Google Scholar
Nelissen M. 1975Contribution to the ethology of Simochromis diagramma (Günther) (Pisces, Cichlidae). Acta Zool. Pathol. Antverp. 61, 31-46. Google Scholar
Voss J. 1977Les livrees ou patrons de coloration chez les poissons Cichlides africains. Rev. Franc. Aquariol. 4, 34-80. Google Scholar
Svensson L, Mullarney K, Zetterström D, Grant PJ. 1999Collins bird guide: the most complete guide to the birds of Britain and Europe, 2nd ed. London, UK: Harper Collins Publishers. Google Scholar
Senar JC. 2006Color displays as intrasexual signals of aggression and dominance. In Bird coloration vol 2 function and evolution (eds Hill GE, McGraw KJ), pp. 87-136. Cambridge, MA: Harvard University Press. Google Scholar
Goodale E, Beaucham G, Ruxton G. 2017Mixed-species groups of animals. London, UK: Academic Press. Google Scholar