Different colour morphs of the poison frog Andinobates bombetes (Dendrobatidae) are similarly effective visual predator deterrents

Aposematism is the use of warning signals to advertise unpleasant or dangerous defences to potential predators. As the effectiveness of this strategy depends on predator learning, little variation is expected in aposematic warning signals, as similar signals facilitate predator learning. However, warning signals are frequently variable in aposematic species. Such variability could arise as a result of geographic variation in the interpretation that local predators give warning signals. We tested this divergent learning hypothesis in the polytypic poison frog Andinobates bombetes (Anura: Dendrobatidae), focusing on visual predators. Our study was conducted in two populations of this species located in the Western Andes of Colombia, where individuals at some localities exhibit red dorsolateral stripes, while those in others exhibit yellow dorsolateral stripes. We deployed paraffin models imitating both forms of A. bombetes in size and colouration, as well as dull‐coloured controls, at sites inhabited by either red‐striped or yellow‐striped frogs. Red and yellow models were attacked at similar rates at both sites, and brown models were attacked more frequently at one of the sites. These results suggest that red and yellow colourations function as similarly effective aposematic signals for primarily visual predators, regardless of the form previously experienced by these predators. Therefore, our results do not support the hypothesis of divergent predator learning as a driver of the polytypism present in this species. Finally, we discuss other mechanisms that may be involved in the evolution and maintenance of this polytypism.     

Conditions for the spread of conspicuous warning signals: a numerical model with novel insights

The initial evolution of conspicuous warning signals presents an evolutionary problem because selection against rare conspicuous signals is presumed to be strong, and new signals are rare when they first arise. Several possible solutions have been offered to solve this apparent evolutionary paradox, but disagreement persists over the plausibility of some of the proposed mechanisms. In this paper, we construct a deterministic numerical simulation model that allows us to derive the strength of selection on novel warning signals in a wide range of biologically relevant situations. We study the effects of predator psychology (learning, rate of mistaken attacks, and neophobia) on selection. We also study the how prey escape, predation intensity, number of predators, and abundance of different prey types affects selection. The model provides several important results. Selection on novel warning signals is number rather than frequency dependent. In most cases, there exists a threshold number of aposematic individuals below which aposematism is selected against and above which aposematism is selected for. Signal conspicuousness (which increases detection rate) and distinctiveness (which allows predator to distinguish defended from nondefended prey) have opposing effects on evolution of warning signals. A more conspicuous warning signal cannot evolve unless it makes the prey more distinctive from palatable prey, reducing mistaken attacks by predators. A novel warning signal that is learned quickly can spread from lower abundance more easily than a signal that is learned more slowly. However, the relative rate at which the resident signal and the novel signal are learned is irrelevant for the spread of the novel signal. Long-lasting neophobia can facilitate the spread of novel warning signals. Individual selection via the ability of defended prey to escape from predator is not likely to facilitate evolution of conspicuous warning signals if both the resident (cryptic) morph and the novel morph have the same escape probability. Predation intensity (defined as the proportion of palatable prey eaten by the predator) has a strong effect on selection. More intense predation results in strong selection against rare signals, but also strong selective advantage to common signals. The threshold number of aposematic individuals is lower when predation is intense. Thus, the evolution of warning signals may be more likely in environments where predation is intense. The effect of numbers of predators depends on whether predation intensity also changes. When predation intensity is constant, increasing numbers of predators raises the threshold number of aposematic individuals, and thus makes evolution of aposematism more difficult. If predation intensity increases in parallel with number of predators, the threshold number of aposematic individuals does not change much, but selection becomes more intense on both sides of the threshold.     

Investigating Müllerian mimicry: predator learning and variation in prey defences

Inexperienced predators are assumed to select for similarity of warning signals in aposematic species (Müllerian mimicry) when learning to avoid them. Recent theoretical work predicts that if co-mimic species have unequal defences, predators attack them according to their average unpalatability and mimicry may not be beneficial for the better defended co-mimic. In this study, we tested in a laboratory environment whether a uniform warning signal is superior to a variable one in promoting predator learning, and simultaneously whether co-mimics are preyed upon according to their average unpalatability. There was an interaction of signal variation and unpalatability but inexperienced birds did not select for signal similarity in artificial prey; when the prey was moderately defended a variable signal was even learnt faster than a uniform one. Due to slow avoidance learning, moderately defended prey had higher mortality than highly defended prey (although this was not straightforward), but mixing high and moderate unpalatability did not increase predation compared with high unpalatability. This does not support the view that predators are sensitive to varying unpalatability. The results suggest that inexperienced predators may neither strongly select for accurate Müllerian mimicry nor affect the benefits of mimicry when the co-mimics are unequally defended.     

Aposematic coloration, luminance contrast, and the benefits of conspicuousness

Many organisms use warning, or aposematic, coloration to signaltheir unprofitability to potential predators. Aposematicallycolored prey are highly visually conspicuous. There is considerableempirical support that conspicuousness promotes the effectivenessof the aposematic signal. From these experiments, it is welldocumented that conspicuous, unprofitable prey are detectedsooner and aversion learned faster by the predator as comparedwith cryptic, unprofitable prey. Predators also retain memoryof the aversion longer when prey is conspicuous. The presentstudy focused on the elements of conspicuousness that conferthese benefits of aposematic coloration. Drawing on currentunderstanding of animal vision, we distinguish 2 features ofwarning coloration: high chromatic contrast and high brightness,or luminance, contrast. Previous investigations on aposematicsignal efficacy have focused mainly on the role of high chromaticcontrast between prey and background, whereas little researchhas investigated the role of high luminance contrast. Usingthe Chinese mantid as a model predator and gray-painted milkweedbugs as model prey, we found that increased prey luminance contrastincreased detection of prey, facilitated predator aversion learning,and increased predator memory retention of the aversive response.Our results suggest that the luminance contrast component ofaposematic coloration can be an effective warning signal betweenthe prey and predator. Thus, warning coloration can even evolveas an effective signal to color blind predators.     

Predator mixes and the conspicuousness of aposematic signals

Conspicuous warning signals of unprofitable prey are a defense against visually hunting predators. They work because predators learn to associate unprofitability with bright coloration and because strong signals are detectable and memorable. However, many species that can be considered defended are not very conspicuous; they have weak warning signals. This phenomenon has previously been ignored in models and experiments. In addition, there is significant within- and among-species variation among predators in their search behavior, in their visual, cognitive, and learning abilities, and in their resistance to defenses. In this article we explore the effects of variable predators on models that combine positive frequency-dependent, frequency-independent, and negative frequency-dependent predation and show that weak signaling of aposematic species can evolve if predators vary in their tendency to attack defended prey.     

Why are wasps so intimidating: field experiments on hunting dragonflies (Odonata: Aeshna grandis)

The mechanisms of aposematism (unprofitability of prey combined with a conspicuous signal) have mainly been studied with reference to vertebrate predators, especially birds. We investigated whether dragonflies, Aeshna grandis, avoid attacking wasps, Vespula norwegica, which are an unprofitable group of prey for most predators. As a control we used flies that were painted either black or with yellow and black stripes. The dragonflies showed greater aversion to wasps than to flies. Black-and-yellow-striped flies were avoided more than black ones, suggesting that aposematic coloration on a harmless fly provides a selective advantage against invertebrate predators. There was no significant difference in reactions to black-painted and black-and-yellow wasps, indicating that, in addition to coloration, some other feature in wasps might deter predators. In further experiments we offered dragonflies artificial prey items in which the candidate warning signals (coloration, odour and shape) were tested separately while other confounding factors were kept constant. The dragonflies avoided more black-and-yellow prey items than solid black or solid yellow ones. However, we found no influence of wasp odour on dragonfly hunting. Dragonflies were slightly, but not significantly, more reluctant to attack wasp-shaped prey items than fly-shaped ones. Our results suggest that the typical black-and-yellow stripes of wasps, possibly combined with their unique shape, make dragonflies avoid wasps. Since black-and-yellow stripes alone significantly decreased attack rate, we conclude that even profitable prey species (i.e. Batesian mimics) are able to exploit the dragonflies' avoidance of wasps. Copyright 2003 Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour.      

Predator experience on cryptic prey affects the survival of conspicuous aposematic prey

Initially, aposematism, which is an unprofitable trait, e.g. noxiousness conspicuously advertised to predators, appears to be a paradox since conspicuousness should increase predation by naive predators. However, reluctance of predators for eating novel prey (e.g. neophobia) might balance the initial predation caused by inexperienced predators. We tested the novelty effects on initial predation and avoidance learning in two separate conspicuousness levels of aposematic prey by using a 'novel world' method. Half of the wild great tits (Parus major) were trained to eat cryptic prey prior to the introduction of an aposematic prey, which potentially creates a bias against the aposematic morph. Both prey types were equally novel for control birds and they should not have shown any biased reluctance for eating an aposematic prey. Knowledge of cryptic prey reduced the expected initial mortality of the conspicuous morph to a random level whereas control birds initially ate the conspicuous morph according to the visibility risk. Birds learned to avoid conspicuous prey in both treatments but knowledge of cryptic prey did not increase the rate of avoidance learning. Predators' knowledge of cryptic prey did not reduce the predation of the less conspicuous aposematic prey and additionally predators did not learn to avoid the less conspicuous prey. These results indicate that predator psychology, which was shown as reluctance for attacking novel conspicuous prey, might have been important in the evolution of aposematism.     

Pyrazine odour makes visually conspicuous prey aversive

Unpalatable insects frequently adopt multimodal signals to ward off predators, incorporating sounds and odours into their colourful displays. Pyrazine is an odour commonly used in insect warning displays, and has previously been shown to elicit unlearned biases against common warning colours, e.g. yellow and red in naive predators. We designed two experiments to test for similar effects of pyrazine on the conspicuousness of prey, perhaps the most ubiquitous aspect of aposematic coloration. In the first experiment, we offered predators (Gallus gallus domesticus) a choice between conspicuous crumbs and cryptic crumbs in the presence or absence of pyrazine. In the second experiment, we manipulated the birds' experience of conspicuous prey during an initial training phase. Only in the presence of pyrazine did birds show a bias against conspicuously coloured food, and this occurred whether or not they had previously experienced food that contrasted with the background. This emergent behaviour relied upon the visual and odorous signal components being presented together. These unlearned, yet hidden, responses against conspicuousness demonstrate that there are initial benefits to prey being conspicuous when the multimodal nature of warning signals is accounted for.     

Aposematism and gregariousness: the combined effect of group size and coloration on signal repellence

Many aposematic species have evolved an aggregated lifestyle, and one possible advantage of grouping in warningly coloured prey is that it makes the aposematic signal more effective by generating a greater aversion in predators. Here we investigate the effect of prey group size on predator behaviour, both when prey are aposematic and when they are not aposematic, to separate the effects of warning coloration and prey novelty. Naive domestic chicks (Gallus gallus domesticus) were presented with either solitary or groups of 3, 9 or 27 live larvae of the aposematic bug Tropidothorax leucopterus. Other naive chicks were presented with larvae of the non-aposematic bug Graptostethus servus either solitary or in groups of 27. Attack probability decreased with increasing group size of aposematic prey, both when birds were naive and when they had prior experience, whereas prey gregariousness did not affect the initial attack probability on the G. servus larvae. In a separate experiment, groups of mealworms were shown to be even more attractive than solitary mealworms to naive chicks. We conclude that the aversiveness of prey grouping in this study can be explained as increased signal repellence of specific prey coloration, in this case a classical warning coloration. These experiments thus support the idea of gregariousness increasing the signalling effect of warning coloration.     

Prey community structure affects how predators select for Mullerian mimicry

Müllerian mimicry describes the close resemblance between aposematic prey species; it is thought to be beneficial because sharing a warning signal decreases the mortality caused by sampling by inexperienced predators learning to avoid the signal. It has been hypothesized that selection for mimicry is strongest in multi-species prey communities where predators are more prone to misidentify the prey than in simple communities. In this study, wild great tits (Parus major) foraged from either simple (few prey appearances) or complex (several prey appearances) artificial prey communities where a specific model prey was always present. Owing to slower learning, the model did suffer higher mortality in complex communities when the birds were inexperienced. However, in a subsequent generalization test to potential mimics of the model prey (a continuum of signal accuracy), only birds that had foraged from simple communities selected against inaccurate mimics. Therefore, accurate mimicry is more likely to evolve in simple communities even though predator avoidance learning is slower in complex communities. For mimicry to evolve, prey species must have a common predator; the effective community consists of the predator's diet. In diverse environments, the limited diets of specialist predators could create 'simple community pockets' where accurate mimicry is selected for.     

Did aggregation favour the initial evolution of warning coloration? A novel world revisited

From experiments using novel prey signals to avoid innate reactions to traditional signals, Alatalo & Mappes (1996, Nature, 382, 708-710) concluded that gregariousness would have selected for warning coloration as it originated for the first time, whereas a solitary prey distribution would not. We have investigated this suggestion in experiments using the same novel prey and background symbols and wild-caught great tit, Parus major, predators. We compared the attack rate on cryptic unpalatable and aposematic unpalatable prey in either a solitary or an aggregated treatment. In the aggregated treatment we found no difference in attack rate on cryptic and aposematic prey. In the solitary treatment the attack rate on aposematic prey was significantly lower after one attack and at the end of the experiment. Thus, we conclude that, in so far as these experiments mimic an original predator-prey relationship, they do not give support to the idea that aggregation would have favoured the evolution of warning coloration in unpalatable prey. Copyright 2000 The Association for the Study of Animal Behaviour.     

Five species of passerine bird differ in their ability to detect Batesian mimics

Multiple predators affect the evolution of aposematic signals in nature and these predators may substantially differ in terms of ecological and cognitive parameters. However, most experimental studies testing the evolution of Batesian mimics use only a single species of predator (usually the great tit or a domestic chick). Therefore, in the present study, we experimentally tested the responses of five passerine predators to an artificially made Batesian mimic (a cockroach equipped with the warning pattern of the red firebug) with respect to their dietary ecology. Half of the individuals of each species were fed on unmodified roaches before the experiment, whereas the other half were fed with mealworms and thus had no previous experience with roaches. We found that Batesian mimics were better protected than inconspicuous prey against inexperienced great tits and robins alone. The other three bird species showed high level of neophobia; therefore, the effect of warning coloration could not be assessed. We also found that experienced birds attacked a greater number of Batesian mimics compared to inexperienced individuals of all tested species, with the exception of blackcaps. In the great tits, robins, and blue tits, a significant number of experienced birds attacked the Batesian mimic, which was possibly the result of a learned search image for a roach. Our results suggest that using a limited array of predators to describe evolutionary processes forming the diversity of antipredatory strategies of the prey may be biased and need not describe the situation occurring in nature.     

Warning signals, receiver psychology and predator memory

This review identifies four receiver psychology perspectives that are likely to be important in the design and evolution of warning signals. Three of these perspectives (phobia, learning and prey recognition) have been studied in detail, and I include a brief review of recent work. The fourth, a memory perspective, has received little attention and is developed here. A memory perspective asks, 'how might warning signals function to reduce forgetting of avoidances between encounters?'. To answer this question I review data from psychology literature that describe important features of animal long-term memory. These data suggest that components of warning signals may function to reduce forgetting (and therefore increase memorability) by (1) preventing forgetting of learnt prey discriminations; (2) jogging the memories of forgetful predators; and (3) biasing forgetting in favour of prey avoidance when the warning signal of a defended aposematic species is copied by an edible Batesian mimic. A combination of a learning and a memory perspective suggests that the features of aposematic prey that accelerate avoidance learning may also be the features that decelerate forgetting processes. If correct, this would have important implications for the comprehension of signal design. Finally, I suggest that the cryptic appearance of an edible prey may decelerate predator learning and accelerate predator forgetting, to the benefit of the prey. In terms of learning and memory, crypsis may be an antisignal. Copyright 2000 The Association for the Study of Animal Behaviour.     

No evidence for differential survival or predation between sympatric color morphs of an aposematic poison frog

Because variation in warning signals slows down the predator education process, aposematic theory predicts that animal warning signals should be monomorphic. Yet, warning color polytypisms are not uncommon in aposematic species. In cases where warning signal variants are separated geographically, adaptation to local predators could explain this variation. However, this cannot explain the persistence of sympatric polymorphisms in aposematic taxa. The strawberry poison frog (Oophaga pumilio) exhibits both allopatric and sympatric warning color variation in and around the Bocas del Toro archipelago of Panama. One explanation that has been proposed for the rapid diversification of O. pumilio coloration in this archipelago is low predation; if island populations have few predators, stabilizing selection would be relaxed opening the door for diversification via selection or genetic drift. Using a combination of mark-recapture and clay model studies, we tested for differences in survival and predation among sympatric red and yellow color morphs of O. pumilio from Bastimentos Island. We found no evidence for differential survival or predation in this population, despite the fact that one morph (red) is more common and widely distributed than the other (yellow). Even in an area of the island where the yellow morph is not found, predator attack rates were similar among morphs. Visual modeling suggests that yellow and red morphs are distinguishable and conspicuous against a variety of backgrounds and by viewers with different visual systems. Our results suggest that general avoidance by predators of red and yellow, both of which are typical warning colors used throughout the animal kingdom, may be contributing to the apparent stability of this polymorphism.     

Is the hibiscus harlequin bug aposematic? The importance of testing multiple predators

Aposematism involves predators learning conspicuous signals of defended prey. However, prey species utilize a wide range of chemical (or physical) defenses, which are not likely to be equally aversive to all predators. Aposematism may therefore only be effective against a physiologically sensitive subset of potential predators, and this can only be identified through behavioral testing. We studied the emerging model organism Tectocoris diophthalmus (Heteroptera: Scutelleridae), an aposematically colored but weakly defended shieldback stinkbug, to test the efficacy of its defenses against a suite of predator types. We predicted the bugs' defenses would be ineffectual against both experienced and naïve birds but aversive to predaceous insects. Surprisingly, the opposite pattern was found. Both habituated wild passerines and naïve chickens avoided the bugs, the chickens after only one or two encounters. To avian predators, T. diophthalmus is aposematic. However, praying mantids showed no repellency, aversion, or toxicity associated with adult or juvenile bugs after multiple trials. Comparison with prior studies on mantids using bugs with chemically similar but more concentrated defenses underscores the importance of dose in addition to chemical identity in the efficacy of chemical defenses. Our results also emphasize the importance of behavioral testing with multiple ecologically relevant predators to understand selective pressures shaping aposematic signals and chemical defenses.     

Diversity in warning coloration: selective paradox or the norm?

Aposematic theory has historically predicted that predators should select for warning signals to converge on a single form, as a result of frequency‐dependent learning. However, widespread variation in warning signals is observed across closely related species, populations and, most problematically for evolutionary biologists, among individuals in the same population. Recent research has yielded an increased awareness of this diversity, challenging the paradigm of signal monomorphy in aposematic animals. Here we provide a comprehensive synthesis of these disparate lines of investigation, identifying within them three broad classes of explanation for variation in aposematic warning signals: genetic mechanisms, differences among predators and predator behaviour, and alternative selection pressures upon the signal. The mechanisms producing warning coloration are also important. Detailed studies of the genetic basis of warning signals in some species, most notably Heliconius butterflies, are beginning to shed light on the genetic architecture facilitating or limiting key processes such as the evolution and maintenance of polymorphisms, hybridisation, and speciation. Work on predator behaviour is changing our perception of the predator community as a single homogenous selective agent, emphasising the dynamic nature of predator–prey interactions. Predator variability in a range of factors (e.g. perceptual abilities, tolerance to chemical defences, and individual motivation), suggests that the role of predators is more complicated than previously appreciated. With complex selection regimes at work, polytypisms and polymorphisms may even occur in Müllerian mimicry systems. Meanwhile, phenotypes are often multifunctional, and thus subject to additional biotic and abiotic selection pressures. Some of these selective pressures, primarily sexual selection and thermoregulation, have received considerable attention, while others, such as disease risk and parental effects, offer promising avenues to explore. As well as reviewing the existing evidence from both empirical studies and theoretical modelling, we highlight hypotheses that could benefit from further investigation in aposematic species. Finally by collating known instances of variation in warning signals, we provide a valuable resource for understanding the taxonomic spread of diversity in aposematic signalling and with which to direct future research. A greater appreciation of the extent of variation in aposematic species, and of the selective pressures and constraints which contribute to this once‐paradoxical phenomenon, yields a new perspective for the field of aposematic signalling.     

Honest signaling and the uses of prey coloration

Abstract Although signal reliability is of fundamental importance to the understanding of animal communication, the extent of signal honesty in relation to antipredator warning signals has received relatively little attention. A recent theoretical model that assumed a physiological linkage between pigmentation and toxicity suggested that (aposematic) warning signals may often be reliable, in the sense that brightness and toxicity are positively correlated within prey populations. Two shortcomings of the model were (1) the requirement among predators for an innate aversion to brightly colored prey and (2) the assumption that prey can generate only bright coloration and not cryptic coloration. We evaluated the generality of predictions of reliable signaling when these shortcomings were removed. Without innate avoidance of bright prey, we found a positive brightness-toxin correlation when conspicuous prey coloration provided an additional fitness benefit unrelated to predation. Initially, this correlation could evolve for reasons unrelated to prey signaling; hence, aposematism might represent a striking example of exaptation. Given a choice between using pigmentation for bright or for cryptic coloration, crypsis was favored only in conditions of very low or very high resource levels. In the latter case, toxicity correlated positively with degree of cryptic coloration. Predictions of toxin-signal correlation appear robust, but we can identify interesting conditions in which signal reliability is not predicted.     

The role of predators in maintaining the geographic organization of aposematic signals

Selective predation of aposematic signals is expected to promote phenotypic uniformity. But while these signals may be uniform within a population, numerous species display impressive variations in warning signals among adjacent populations. Predators from different localities who learn to avoid distinct signals while performing intense selection on others are thus expected to maintain such a geographic organization. We tested this assumption by placing clay frog models, representing distinct color morphs of the Peruvian poison dart frog Ranitomeya imitator and a nonconspicuous frog, reciprocally between adjacent localities. In each locality, avian predators were able to discriminate between warning signals; the adjacent exotic morph experienced up to four times more attacks than the local one and two times more than the nonconspicuous phenotype. Moreover, predation attempts on the exotic morph quickly decreased to almost nil, suggesting rapid learning. This experiment offers direct evidence for the existence of different predator communities performing localized homogenizing selection on distinct aposematic signals.     

Effect of sex and bright coloration on survival and predator‐induced wing damage in an aposematic lantern fly with startle display

1. Aposematic coloration in prey promotes its survival by conspicuously advertising unpalatability to predators. Although classical examples of aposematic signals involve constant presentation of a signal at a distance, some animals suddenly display warning colours only when they are attacked. 2. Characteristics of body parts suddenly displayed, such as conspicuous coloration or eyespot pattern, may increase the survival of the prey by startling the predator, and/or by signalling unpalatability to the predators at the moment of attack. 3. The adaptive value of such colour patterns suddenly displayed by unpalatable prey has not been studied. We experimentally blackened the red patch in the conspicuous red–white–black hindwing pattern displayed by an unpalatable insect Lycorma delicatula White (Hemiptera: Fulgoridae) in response to predator's attack. 4. There was no evidence that the presence of the red patch increased prey survival over several weeks. We hypothesise that predators generalised from the red–white–black patches on the hindwings of unpalatable L. delicatula to any similar wing display as a signal of unpalatability. Because a higher proportion of males than females stay put at their resting sites, displaying their wings in response to repeated attacks by predators, wing damage was more frequent in males than in females. 5. To our knowledge, this is the first experimental test of an adaptive role of aposematic signals presented by unpalatable prey during sudden displays triggered by direct predatory attack.     

Why aren't warning signals everywhere? On the prevalence of aposematism and mimicry in communities

Warning signals are a striking example of natural selection present in almost every ecological community – from Nordic meadows to tropical rainforests, defended prey species and their mimics ward off potential predators before they attack. Yet despite the wide distribution of warning signals, they are relatively scarce as a proportion of the total prey available, and more so in some biomes than others. Classically, warning signals are thought to be governed by positive density-dependent selection, i.e. they succeed better when they are more common. Therefore, after surmounting this initial barrier to their evolution, it is puzzling that they remain uncommon on the scale of the community. Here, we explore factors likely to determine the prevalence of warning signals in prey assemblages. These factors include the nature of prey defences and any constraints upon them, the behavioural interactions of predators with different prey defences, the numerical responses of predators governed by movement and reproduction, the diversity and abundance of undefended alternative prey and Batesian mimics in the community, and variability in other ecological circumstances. We also discuss the macroevolution of warning signals. Our review finds that we have a basic understanding of how many species in some taxonomic groups have warning signals, but very little information on the interrelationships among population abundances across prey communities, the diversity of signal phenotypes, and prey defences. We also have detailed knowledge of how a few generalist predator species forage in artificial laboratory environments, but we know much less about how predators forage in complex natural communities with variable prey defences. We describe how empirical work to address each of these knowledge gaps can test specific hypotheses for why warning signals exhibit their particular patterns of distribution. This will help us to understand how behavioural interactions shape ecological communities.