Behavioural, ecological, and evolutionary aspects of diversity in frog colour patterns

22 Pages·2016·4.64 MB·English

Biol.Rev.(2016),pp.000–000. 1
doi:10.1111/brv.12269
Behavioural, ecological, and evolutionary
aspects of diversity in frog colour patterns
∗
Bibiana Rojas
CentreofExcellenceinBiologicalInteractions,DepartmentofBiologyandEnvironmentalSciences,UniversityofJyvaskyla,POBox35,Jyva¨skyla¨
FI40001,Finland
ABSTRACT
The role of colours and colour patterns in behavioural ecology has been extensively studied in a variety of contexts
and taxa, while almost overlooked in many others. For decades anurans have been the focus of research on acoustic
signalling due to the prominence of vocalisations in their communication. Much less attention has been paid to the
enormous diversity of colours, colour patterns, and other types of putative visual signals exhibited by frogs. With
the exception of some anecdotal observations and studies, the link between colour patterns and the behavioural and
evolutionary ecology of anurans had not been addressed until approximately two decades ago. Since then, there has
been ever-increasing interest in studying how colouration is tied to different aspects of frog behaviour, ecology and
evolution.HereIreviewtheliteratureonthreedifferentcontextsinwhichfrogcolourationhasbeenrecentlystudied:
predator–preyinteractions,intraspecificcommunication,andhabitatuse;andIhighlightthoseaspectsthatmakefrogs
an excellent, yet understudied, group to examine the role of colour in the evolution of anti-predation strategies and
animalcommunicationsystems.Further,Iarguethatinadditiontonatural-historyobservations,moreexperimentsare
neededinordertoelucidatethefunctionsofanurancolourationandtheselectivepressuresinvolvedinitsdiversity.To
conclude, I encourage researchers to strengthen current experimental approaches, and suggest future directions that
maybroadenourcurrentunderstandingoftheadaptivevalueofanurancolourpatterndiversity.
Key words: colouration, predator–prey interactions, visual communication, sexual selection, conflict resolution,
spaceuse.
CONTENTS
I. Introduction .............................................................................................. 2
II. Predator–preyinteractions ............................................................................... 2
(1) Camouflage .......................................................................................... 2
(2) Aposematismandmimicry ........................................................................... 4
(a) Thepuzzleofpolymorphicwarningsignals ....................................................... 8
(b) Honestyinwarningsignals ....................................................................... 10
(3) Conspicuouscolourationrevealedthroughmovementorbehaviour ................................. 11
III. Intraspecificcommunication ............................................................................. 11
(1) Matepreferencesandassortativemating ............................................................. 11
(a) Colourscanattractbothmatesandpredators .................................................... 13
(b) Sexualdichromatism ............................................................................. 13
(2) Intrasexualcompetitionandconflictresolution ....................................................... 13
IV. Habitatselectionandspaceuse .......................................................................... 15
V. Futuredirections ......................................................................................... 16
VI. Conclusions .............................................................................................. 16
VII. Acknowledgements ....................................................................................... 17
VIII. References ................................................................................................ 17
* Addressforcorrespondence(Tel:[phone];E-mail:bibiana.rojas@jyu.fi).
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
2 BibianaRojas
I. INTRODUCTION behavioural and evolutionary ecology of anurans had not
been properly addressed until approximately two decades
‘Anextensivesurveyoftheorganicworldthusleadsustotheconclusion ago. Previous studies focused mostly on the inheritance of
that colour is by no means so unimportant or inconstant a character colour patterns (Davison, 1963; Resnick & Jameson, 1963;
as at first sight it appears to be; and the more we examine it the more Fogleman,Corn&Pettus,1980;Blouin,1989);butseeNevo,
convinced we shall become that it must serve some purpose in nature, 1973, for an early study on the selective pressures involved
andthatbesidescharmingusbyitsdiversityandbeautyitmustbewell inthemaintenanceofcolourpolymorphismincricketfrogs).
worthyofourattentivestudy,andhavemanysecretstounfoldtous.’ – Recently, more attention has been paid to the enormous
A.R.Wallace(1877,p.643) diversityofcolours,colourpatterns,andothertypesofvisual
Signals are supposed to evolve so that the signal-to-noise signals displayed by frogs (e.g. Ho¨dl & Ame´zquita, 2001),
and how those signals are tied to different aspects of frog
ratio (the contrast between the signal and the background
ecology and behaviour. Here I aim to review the diversity
noise) is maximised (Endler, 1992, 1993a; Bradbury &
of frog colouration in relation to behaviour and ecology.
Vehrencamp, 2011), while signal degradation is minimised
I do so whilst focusing on those cases in which colour
(Endler, 1992). Selection also tends to favour signals with
patternsarevisualsignalsontheirown,describinghowthese
a high efficacy not only in terms of their transmission and
signals are currently thought to function in the context of
detection, but also in their ability to elicit a response in
anti-predation strategies, intraspecific communication and
the receiver that increases the sender’s fitness (Guilford
habitat use. Lastly, I suggest future directions within each
& Dawkins, 1991) while maintaining the receiver’s fitness
context that might fill some of the current gaps in frog
unaffected. An exception to this is deceptive signals, which
colouration research; and highlight the need to strengthen
deliver incorrect information about the signaller and thus
and broaden the current experimental approaches in order
benefit the sender at the expense of the receiver (Wiley,
towidenourunderstandingoftheadaptivevalueofdiversity
1983,1994;Mokkonen&Lindstedt,2015).
in frog colouration, and to identify the candidate selective
Different modalities of communication entail diverse
pressuresthatmightbeshapingsuchdiversity.
advantagesandconstraintsonthesignalsinvolved(Bradbury
&Vehrencamp,2011).Acousticsignalscanbeadvantageous
overlongdistancesbecausesoundwavescantravelforlonger
througheitherairorwaterwithoutdegradingcomparedto, II. PREDATOR–PREYINTERACTIONS
for example, light. This is particularly useful for nocturnal
animals given the low or non-existent light levels at night
Colourationmayhaveanenormouseffectonanimals’fitness
(Bradbury & Vehrencamp, 2011). Chemical or olfactory
because of its adaptive function as an interspecific signal in
signals are also good in low-light scenarios, but they rely
thecontextofpredation,amongothers.Whilesomeanimals
heavilyonthecharacteristicsofthemediuminwhichtheyare
gain protection from predators by blending with their sur-
transmitted(Bradbury&Vehrencamp,2011).Visualsignals,
roundings(camouflage;Edmunds,1974),manyspeciesalso
ontheotherhand,workwellovershorttomediumdistances
haveconspicuouscolourpatternsthatwarnpredatorsabout
provided there is a lack of physical obstacles between
theirunprofitability(Ruxton,Sherratt&Speed,2004).The
the sender and the receiver, suitable contrast with the
latterstrategyisreferredtoasaposematism(Poulton,1890).
background,andaminimumofenvironmentallight(Endler,
1992, 1993a); they can have different shapes and sizes, and
(1) Camouflage
may convey information either on their own (static signals;
e.g.diverseandconspicuouscolourpatternsinbirdplumage; Camouflageinvolves a seriesof strategies that preventprey
Andersson, 1994) or when accompanied or enhanced by from being detected or recognised by predators (Edmunds,
repeated movements or displays (dynamic signals; e.g. the 1974;Stevens&Merilaita,2009).Suchstrategiesinclude,for
extension of a coloured dewlap in combination with head instance,crypsisandmasquerade.Commontypesofcrypsis
bobbinginAnolislizards;Losos&Chu,1998). arebackgroundmatching,wherebyananimalresemblesits
The role of colour patterns as visual signals in animal background colouration and thus avoids detection (Endler,
behaviour and ecology has been studied extensively in a 1988), and disruptive colouration, which involves markings
variety of contexts and taxa, but has been neglected in that make it difficult for the predator to distinguish the
others. For example, for decades anurans have been the outline or shape of the prey (Thayer, 1909; Cott, 1940).
focusofstudiesonacousticsignallingduetotheprominence Masquerade, whereby animals resemble an uninteresting
of vocalisations in their communication system. Not only objectintheirsurroundings(i.e.arock,aleaf,astick,etc.),
hasitbeendemonstratedthatfrogsemitcallswithdifferent preventspreyrecognitioninstead(Skelhornetal.,2010).Both
functions (Gerhardt & Huber, 2002), but they have also crypsis and masquerade ultimately deprive the predator of
beenshowntohaveoutstandingsensoryabilitiesthatallow key information about the prey and, therefore, constitute
them to be both physiologically (Capranica & Moffat, a form of deception (Caro, 2014; Mokkonen & Lindstedt,
1983) and behaviourally (Ame´zquita etal., 2011) tuned to 2015). Camouflage is a widespread anti-predator strategy
the characteristics of their own species-specific acoustic among anurans, which is reflected in the prevalence of
signals. By contrast, the link between colouration and the earthycolourssuchasdifferentshadesofgreen,brownand
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
Functionandevolutionoffrogcolourdiversity 3
greyinmanyspecies(Wells,2007).Itcanbeeffectiveonits phenotypes (apostatic selection), because the fitness of each
own,asinyoungPristimantiszeuctotyluswhosecolourpatterns morph will be inversely related to its frequency in the
resemble those of the mossy substrate (Fig. 1A), or Hyla population(Allen&Greenwood,1988;Endler,1991a;Bond
japonica,whichchangesitsdorsalcolourtomatchthatofthe & Kamil, 1998; Bond, 2007). Thus, by being polymorphic
background(Choi&Jang,2014);butcanalsobeenhanced the per capita predation risk of a given species might be
withparticularbehaviours.IndividualsofCraugastorfitzingeri, reduced,astheabundanceofanyofitsmorphswouldbelow
forexample,becomecompletelyimmobileaftereveryjump comparedtoamonomorphicspecies(Endler,1991a).Despite
and can also hide their head in the leaf litter; experiments the prevalence of colour pattern polymorphisms among
withhumanobserverswhoknewexactlywheretosearchbut cryptically coloured anurans, the mechanisms that allow
still could not easily locate individuals have demonstrated for their maintenance remain rather obscure. According to
thatthedetectionofthesefrogscanbeextremelydifficultfor Hoffman&Blouin(2000),thereisevidencethatvariationin
predators(Cooper,Caldwell&Vitt,2008). colourpatternscanbeinheritedandstrongindicationsthat
Colour pattern polymorphisms [the simultaneous predationcouldindeedbetheselectivepressurebehindsuch
occurrence of two or more forms within a population diversity, but with a few exceptions (Tordoff, 1980; Morey,
with the rarest form occuring at frequencies higher than 1990;Wente&Phillips,2005)mostevidencetodateremains
thoseexpectedbymutationpressure(Ford,1945)]arevery merelycorrelational.
common among cryptically coloured species, and frogs are Although there are many frog examples illustrating the
no exception (Hoffman & Blouin, 2000; Wells, 2007). For benefits of background matching, evidence describing the
example,thefrequenciesofcolourpatterns(grey,green,and function of disruptive colouration or masquerade is very
red) in two species of Acris (A. gryllus and A. crepitans) are limited and mostly anecdotal. Indeed, despite the fact
correlated locally with variations in substrate colour (Nevo, that it has been suggested to be common among anurans
1973). Remarkably, these species show seasonal variation (Wells,2007),disruptivecolourationhasnotbeenpurposely
in the frequencies of each colour pattern, suggesting that examined experimentally (Rudh & Qvarnstro¨m, 2013).
a morph that is, for example, favoured in spring when Examples of what might be disruptive colouration are
everything is green, will not be favoured in the autumn, markings such as light dorsal stripes, and blotches or spots
whenthebackgroundvegetationwillbemorered.Likewise, thathinderdefinitionofthebodyshapeorbreakupthelimbs
in Dendropsopus (formerly Hyla) labialis, a frog species with (Wells,2007),aswellaslaterallinesthatcrosstheeyeswhile
an extensive latitudinal and altitudinal distribution, there confounding their shape (Toledo & Haddad, 2009; Amat,
are at least five distinct morphs whose frequency seems Wollenberg&Vences,2013).Thebest-knownillustrationof
to be correlated with the predominant background at masquerade,ontheotherhand,isprobablyfoundinspecies
each location. For instance, individuals with spotted colour living in forests, which can presumably trick predators by
patternsaremorecommoninpopulationsathighelevations, lookinglikedeadleaves(Duellman&Trueb,1994),suchas
where the background vegetation is dominated by mosses somespeciesofthegenusRhinella(Fig.1B–D).
(Ame´zquita,1999).Thissuggeststhatcolourpolymorphism Attempts to demonstrate the adaptive value of any of
in this species may have evolved as a form of crypsis theseformsofcamouflage,ortotestspecificallyfortherole
(Ame´zquita, 1999). Interestingly, colour patterns also seem ofpredatorsastheselectivepressurebehindthisdiversityin
toberelatedtobodysize,sothatgreen-dominatedmorphs protective colour patterns (i.e. through apostatic selection),
are smaller than brown ones (Ame´zquita, 1999), pointing could benefit from the use of well-established protocols
at a potential link between colouration and life-history employed in other systems. For example, experiments with
traits that has been surprisingly understudied in anurans. bothhumanandavianpredatorsforagingoneitherartificial
A long-term study on Eleutherodactylus coqui in Puerto Rico prey items, or virtual prey on computer screens have
revealed the existence of 21 distinct pattern morphs whose demonstrated that certain colour patterns increase prey
frequenciesalsodifferamongpopulationsandarecorrelated survival,viabackgroundmatchingordisruptivecolouration,
with the background colouration (Woolbright & Stewart, implicating the role of visual predators as a selection agent
2008).Individualswithlongitudinaldorsalstripesweremost on the evolution of diverse protective colouration (Allen
common in grassland areas, whereas individuals with spots & Clarke, 1968; Allen & Greenwood, 1988; Bond &
and bars were more common in the forest (Woolbright Kamil, 2006; Fraser etal., 2007; Karpestam, Merilaita &
& Stewart, 2008). Likewise, females of Rhinoderma darwinii, Forsman, 2013, 2014). An alternative approach to tackle
which are mostly brown, are found on brown substrates, similar kinds of questions, which has been widely used for
whereas males, which can be either green or brown, research on aposematism (see Section 2.2), is to deploy
are distributed across brown, green and brown–green artificial prey in the field and document and compare
backgrounds(Bourke,Busse&Bakker,2011). attackratesondifferentmorphsbynaturalvisualpredators
Ingeneral,theassociationbetweenbodyandbackground (Cuthilletal.,2005;Valkonenetal.,2011;Farallo&Forstner,
colouration suggests that frogs try to reduce predation risk 2012). Studies involving actual frogs would also be highly
through crypsis. Cryptic polymorphic species may have an informative, as they can account for decisions made by
advantage over monomorphic ones if they are exposed to the individual frogs themselves. For instance, in an elegant
predatorsthatpreydisproportionatelyonthemostcommon experimentinvestigatingwhetherindividualsoftwodifferent
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
4 BibianaRojas
(A) (B)
(C) (D)
Fig.1. AnexampleofbackgroundmatchinginayoungPristimantiszeuctotylus(A);andexamplesofmasquerading,whereafemale
Rhinellamargaritiferaisshowntoresembleadeadleaf(B–D).Images(C)and(D)areforthesameindividualphotographedatdifferent
distances.Photocredits:BibianaRojas.
morphs of Hyla regilla expressed colour pattern-mediated (2) Aposematismandmimicry
microhabitat selection, chemical cues of a snake predator
Aposematismisananti-predatorstrategythroughwhichprey
were used to evaluate how predator presence affected
warnpredatorsabouttheirunprofitability(presenceoftoxins
the frogs’ choice (Wente & Phillips, 2005). The authors
orphysicaldefencessuchasspinesorirritanthairs)bymeans
found that, in the presence of predator cues, both green
ofspecificcolourpatternsthatactaswarningsignals(Poulton,
and brown frogs preferred a substrate that matched their
1890; Cott, 1940; Ruxton etal., 2004; Rojas, Valkonen &
own colour. In the absence of predator cues, however,
Nokelainen, 2015b). This strategy works in such a way
only green frogs exhibited a significant preference for a
that predators learn the association between the warning
matchingbackground,whichsuggestsapossiblegenetically
signalsandtheunprofitabilityoftheprey,andsubsequently
linkedassociationbetweenphenotype(i.e.dorsalcolouration)
avoid them (Endler, 1991a; Endler & Mappes, 2004;
and behaviour (i.e. preference for a matching background)
Ruxtonetal.,2004).
(Wente&Phillips,2005).
Several species of frogs exhibit conspicuous colouration
Finally,thecombinationofthesekindsofexperimentswith
and have a wide array of skin toxins (Wells, 2007), such as
methods that allow for detailed analyses of colour patterns
variousspeciesofMantella(Fig.2A,B)andthe‘Tomatofrogs’
while accounting for predator perception (Endler, 1978,
1984, 1990, 2012; Osorio & Srinivasan, 1991; Vorobyev (genus Dyscophus, Microhylidae; Fig. 2C) from Madagascar
& Osorio, 1998; Vorobyev etal., 1998; Endler & Mielke, (Garraffo etal., 1993a); the Corroboree frogs (Pseudophryne
2005;Kempetal.,2015;Renoult,Kelber&Schaefer,2015) corroboree; Fig. 2D) and other myobatrachids from Australia
opens a broad range of possibilities to study the evolution (Dalyetal.,1990);BrachycephalusephippiumfromBrazil(Fig.2E)
of camouflage strategies in anurans. For instance, a field (Sebben etal., 1986); and numerous species of Bufonids in
experiment where dummies with different colour patterns the genera Melanophryniscus (Fig. 2F) (Garraffo etal., 1993b;
aredeployedondifferentbackgroundstoevaluatehowattack Grant etal., 2012) and Atelopus (Fig. 3A, B; Kim, Kim &
rates differ among groups could be complemented with a Yotsu-Yamashita, 2003) from South and Central America.
surveyoftheactualfrogsintheirnaturalhabitat,recording However, probably the best-known example of aposematic
informationontheexactspotwhereeachindividualisfound. frogs are the dart poison frogs (Dendrobatidae; Fig. 4)
Ifnotonlythisinformationiscollected,butalsostandardised (Stynoski, Schulte & Rojas, 2015). The varied toxins found
photographs are taken of both the individual frog and its inthisNeotropicalfrogfamily(Daly&Myers,1967;Myers
microhabitat, similarities between the frog colour patterns &Daly,1976,1983;Dalyetal.,1994,2002)aresequestered
anditsbackgroundcouldbemeasured,andtheaccuracyof fromtheirspecialiseddiet(Saporitoetal.,2004,2007a),which
camouflagequantified. consistsmainlyofants,termites,mitesandotherarthropods
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
Functionandevolutionoffrogcolourdiversity 5
(A) (B)
(C) (D))
(E) (F)
Fig.2. Aposematicfrogs(A)Mantellabaroni;(B)Mantellaaurantiaca;(C)Dyscophusguineti,knownasthe‘Tomatofrog’;(D)Pseudophryne
corroboree;(E)Brachycephalusephippium;and(F)Melanophryniscusrubriventris.Photocredits:A–C,GerardoGarcía;D,J.P.Lawrence;E,
TaranGrant;F,MarcosVaira.
foundintheleaflitter(Toft,1995;Darstetal.,2005).Toxins possiblythreetimes(Santosetal.,2003;Vencesetal.,2003).
vary noticeably within the family in composition, amount, Thecombinationofbrightcoloursandhightoxicityinthese
andpower,butmostarelipophilicalkaloids(Saporitoetal., frogs has traditionally been put forward as an example of
2012).Onedendrobatidspecies,Phyllobatesterribilis(Fig.4D), aposematism(Myers&Daly,1983;Poughetal.,2001),even
has the most potent non-proteolytic (alkaloid) toxin among thoughthefirstexperimentalattemptstoshowpredatoraver-
vertebrates, batrachotoxin (Myers, Daly & Malkin, 1978). sionofcolourfuldendrobatidsonlytookplaceafewyearsago
Each of these golden yellow or metallic orange frogs can (Saporitoetal.,2007b).Whileithasbeensuggestedthatsome
have up to 1.2mg of toxin which, if it comes into contact crabs,snakes,beetlelarvaeandspidersfeedondendrobatid
withanopenwound,couldpotentiallybelethaltohumans tadpoles (Gray & Christy, 2000; Stynoski etal., 2014), and
inadoseaslowas200μg(Myersetal.,1978). someseeminglytoxin-resistantsnakesmightfeedonjuveniles
During the last 15years, various studies have demon- (Myersetal.,1978),themajorpredatorsofadultpoisonfrogs
stratedanevolutionarylinkbetweencolourationandtoxicity are still not known with certainty, presumably due to the
in dendrobatid frogs (Summers & Clough, 2001; Santos, frogs’successatdeterringpredators.Asindicatedbyanecdo-
Coloma & Cannatella, 2003; Summers, 2003; Darst etal., talobservationsandexperimentsinthefield,ants,Paraponera
2005), suggesting that bright colouration has evolved inde- clavata(Fritz,Rand&Depamphilis,1981),andspiders,Cupi-
pendentlyatleastthreetimes.Dietspecialisation,whichisin ennius coccineus (Szelistowski, 1985) and Sericopelma rubronitens
turn linked with higher levels of toxicity (Darst etal., 2005), (Gray,Kaiser&Green,2010),rejectthemasprey;butthere
might have itself evolved independently at least two, but are also some accounts of fish (Santos & Cannatella, 2011),
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
6 BibianaRojas
(A) (B)
(C)
Fig.3. Atelopusaff.franciscus(A)dorsaland(B)ventralcolouration.(C)Geographicvariationinthedorsalandventralcolouration
ofMelanophryniscusrubriventris.Photocredits:A,B,BibianaRojas;C,MarcosVaira.
snake (Fig. 5A) (Ringler, Ursprung & Ho¨dl, 2010; Lenger, models representing local frogs are usually less frequently
Berkey&Dugas,2014),andspider(T.Larsen,personalcom- attacked than dull models, models representing novel
munication; Fig. 5B) predators. Experiments with frog clay morphs, or familiar models placed on novel backgrounds,
models(seebelow),suggestthatpoisonfrogscouldbesubject atleastforcoloursresemblingthemorphsofOophagapumilio
toattackbybirdsandcrabs.Theseresultsareinagreement in Costa Rica (Saporito etal., 2007b; Hegna etal., 2011;
withatleastoneobservationofacrabfeedingonanindivid- Stuart, Dappen & Losin, 2012) and Dendrobates tinctorius in
ualofOophagahistrionicaintheChoco´ regionofColombia(A. FrenchGuiana(Noonan&Comeault,2009;Rojas,Rautiala
Ve´lez&S.Ko¨rting,personalcommunication;Fig.5C),and &Mappes,2014b).Interestingly,anexperimentcarriedout
twoobservationsofadultrufousmotmots(Baryphthengusmar- inIslaColo´n(Panama´)showedthatthelocal,greenmorphof
tii)consumingoneindividualofD.auratuswithnoapparent O.pumiliowasattackedatsignificantlyhigherfrequencythan
negative effects (Master, 1999) or feeding individuals of O. theforeign,redmorphfromthemainland(Hegna,Saporito
pumiliototheiroffspring(Alvarado,Alvarez&Saporito,2013) & Donnelly, 2013). According to the authors, this result
(Fig. 5D). Additional evidence obtained in studies incorpo- suggeststhatredmightbeamoreefficientpredator-deterrent
rating taxon-specificvisionmodelling(Maan &Cummings, warningsignalregardlessofwhatthelocalsignalis.Finally,
2012; Crothers & Cummings, 2013; Dreher, Cummings & movement has been shown to affect attack rates on clay
Pro¨hl,2015)indicatethatthecolourpatternsofpoisonfrogs models of different colours (Paluh, Hantak & Saporito,
are indeed likely to be designed to signal primarily to birds 2014). In a study comparing attack rates on stationary
and crabs. However, it is important to note that both these brown and red clay models to those on moving models
vision models and the experiments with clay models are of the same colours, Paluh etal. (2014) demonstrated that
unabletoassesstheimportanceofpredatorssuchassnakes, not only was bird predation significantly higher on moving
which do not rely predominantly on visual cues for prey brown frog models, but also significantly lower on moving
detection, but use mostly olfactory, thermal or movement red frog models. These findings provide evidence of the
cuesinstead(Saviola,McKenzie&Chiszar,2012). significance of prey movement for visual predators, and
Furthersupportfortheroleofpoisonfrogcolourpatterns highlighttheimportanceofincorporatingelementsthatoffer
as an anti-predator strategy has been obtained in recent more representative measures of predation in the wild into
studies.Variousfieldexperimentshaveshownthatcolourful claymodelexperiments.
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
Functionandevolutionoffrogcolourdiversity 7
(A) (B) (C)
(D) (E) (F)
(G ) (H) (I)
Fig.4. Dartpoisonfrogsrepresentthebest-knownexampleofaposematismamonganurans.(A)Adelphobatesgalactonotus;(B)Ameerega
bilinguis; (C) Dendrobates truncatus; (D) Dendrobates tinctorius; (E) Phyllobates terribilis; (F) Dendrobates auratus; (G) Oophaga granulifera; (H)
Phyllobates lugubris and (I) Ranitomeya imitator. Photo credits: A,C, Taran Grant; B,F,H,I, J.P. Lawrence; D,G, Bibiana Rojas; E,
RobertoMa´rquez.
In addition to aposematism per se, studies demonstrate fantastica(Fig.6F)(Yeageretal.,2012;butseeChouteauetal.,
that both Batesian (a palatable species mimicking the 2011).Recentexperimentswithchickenshavedemonstrated
colouration of a defended one; Bates, 1862) and Mu¨llerian that models and mimics in this complex might indeed
(twodefendedspeciessharingsimilarcolourpatterns;Mu¨ller, share the costs of predator learning (Stuckert, Venegas
1878) mimicry exist among frog species. In Ecuador, the & Summers, 2014b). Furthermore, a comparison between
aposematic frogs Ameerega (Epipedobates) bilinguis (Fig. 4B) the alkaloid profiles of mimics and models has confirmed
and A. parvula occur parapatrically and serve as models thatallco-mimicspossesschemicaldefences(Stuckertetal.,
to their non-defended mimic, Allobates zaparo, which adopts 2014a). Reciprocal learned avoidance by predators and
the corresponding colouration of its model at each locality possession of secondary defences by all the species in
(Darst&Cummings,2006;Darst,Cummings&Cannatella, the mimetic complex are two fundamental assumptions of
2006). Likewise, Allobates femoralis (Fig. 6A, right), an Mu¨llerian mimicry (Mu¨ller, 1878). Among frogs, however,
Amazonian species with a wide geographic distribution these assumptions have been tested and confirmed only for
and great interpopulational variation in the colouration of the R. imitator complex thus far. Another case of Mu¨llerian
both their inguinal and axillary patches, has been thought mimicrywasrecentlyproposed,withoutfurtherexperimental
to be a Batesian mimic of Ameerega hahneli (Fig. 6A, left) support,wheretheleptodactylidLeptodactyluslineatus(Fig.6C)
(Ame´zquita etal., 2009). Ranitomeya imitator from Peru, on previouslythoughtnottobechemicallydefended,resembles
the other hand, is by far the best-known example of the colouration of the dendrobatid Ameerega picta (Fig. 6B)
a Mu¨llerian mimetic radiation in amphibians (Symula, (Pratesetal.,2012).Also,phylogeneticanalysesusedtostudy
Schulte & Summers, 2001; Twomey etal., 2013; Twomey, the evolution of colour patterns in Malagassy poison frogs
Vestergaard & Summers, 2014). This species (Fig. 6H–K) (genusMantella)suggestthattheconvergenceincolouration
has diverged in both colour pattern and brightness among between M. madagascariensis and M. baroni (Fig. 2A), which
populations to resemble the colour patterns of its putative occursympatrically,mayrepresentanothercaseofMu¨llerian
defended models: R. variabilis (Fig. 6D, highland morph; mimicry(Schaefer,Vences&Veith,2002).Thishypothesis,
Fig. 6G, lowland morph), R. summersi (Fig. 6E), and R. however, has not been tested either. Therefore, given the
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
8 BibianaRojas
(A) (B)
(C) (D)
Fig.5. Knownpredatorsofdendrobatidfrogs:(A)thesnakeRhadinaeadecorata,feedingonOophagapumilio;(B)wolfspider(Lycosidae)
preying on Ameerega trivittata; (C) crab holding an individual Oophaga histrionica; (D) rufous motmot (Baryphthengus martii) taking an
O.pumiliotoitsnest.Photocredits:A,MatthewDugas;B,TrondLarsen;C,AlejandroVe´lez;D,RalphSaporito.
great variability in toxicity and colour patterns within (Fig. 7A–D) (Maan & Cummings, 2012), O. histrionica
anurans, and the co-occurrence of toxic with nontoxic (Ame´zquitaetal.,2013),Dendrobatestinctorius(Rojas&Endler,
species in wide geographical ranges, it seems likely that 2013) (Fig. 7E–H) and Melanophryniscus rubriventris (Fig. 3C)
evenmoreexamplesofbothtypesofmimicryarewaitingto (Bonansea&Vaira,2012).Nevertheless,giventhatlevelsof
beuncovered. variation have not been investigated in many species, it is
highlylikelythatmoreexamplesareyettobediscovered.
Although there are several instances of polymorphic
(a) Thepuzzleofpolymorphicwarningsignals
aposematic populations, the mechanisms allowing warning
Warningsignalvariabilitymayreducetheabilityofpredators signal polymorphisms to persist are not yet fully
to learn and retain the association between colour patterns understood.Recentapproachessuggestanumberofpossible
and distastefulness (Greenwood, Wood & Batchelor, 1981; explanations: (i) an interaction between natural and sexual
Mallet & Joron, 1999; Exnerova´ etal., 2006). As a result of selection (Nokelainen etal., 2012; Crothers & Cummings,
stabilisingselection,it is expectedthat aposematicpreywill 2013; Cummings & Crothers, 2013); for example, in the
have little to no variation in their warning signals (Endler, aposematic wood tiger moth (Parasemia plantaginis), where
1988;Joron&Mallet,1998;Endler&Mappes,2004;Darst males can be either yellow or white, yellow individuals
& Cummings, 2006). Hence, polymorphisms should be have been found to be better protected from predators,
selected against in aposematic species (Endler & Mappes, whereas under certain circumstances whites have a higher
2004). However, there are several cases of aposematic matingsuccess(Nokelainenetal.,2012;Gordonetal.,2015).
species exhibiting colour polymorphisms in nature, for In O. pumilio from Solarte Island (in the Bocas del Toro
exampleladybirds(O’Donald&Majerus,1984;Ueno,Sato archipelago),thestabilisingselectionexertedbypredatorsto
& Tsuchida, 1998), beetles (Borer etal., 2010) and moths keepcolourpatternsuniformismostlikelycounteractedby
(Hegna, Galarza & Mappes, 2015). In anurans, despite directionalsexualselectionfavouringbrightermales,which
many aposematic species showing geographic variation in are preferred by females (Maan & Cummings, 2009). (ii)
colouration(Myers&Daly,1983;Lo¨ttersetal.,1997;Noonan Spatio-temporalvariationinselection(Endler&Rojas,2009;
& Gaucher, 2006; Wollenberg etal., 2008; Chouteau etal., Galarzaetal.,2014; Nokelainenetal.,2014), forinstancein
2011; Hoogmoed & Avila-Pires, 2012; Ame´zquita etal., the composition of local predator communities, which may
2013; Brusa etal., 2013), within-population variability in generate a geographic mosaic of selection throughout the
warning signals has been reported only for Oophaga pumilio distribution range of a species. (iii) A link between colour
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
Functionandevolutionoffrogcolourdiversity 9
(A) (B) (C)
(D) (E) (F) (G)
(H) (I) (J) (K)
Fig. 6. Examples of mimicry among poison frog species. (A) Ameerega hahneli (left) is thought to be a model for Allobates femoralis
(right);Leptodactyluslineatus(C)hasbeensuggestedtobeaMu¨llerianmimicofAmeeregapicta(B).Thebest-knowncaseofaMu¨llerian
mimicry system is that of Ranitomeya imitator. (D)–(G) are the model species: Ranitomeya variabilis, highland morph; R. summersi, R.
fantastica,R.variabilislowlandmorph.ThedifferentmorphsofR.imitatorareshownbelow:(H)spotted;(I)banded;(J)varadero;and
(K)striped.Photocredits:A,JasonL.Brown;B,MauricioPacheco;C,QuentinMartínez;D–KEvanTwomey.
(A) (B) (C) (D)
(E) (F) (G) (H)
Fig.7. ExamplesofgeographicvariationinthecolourpatternsofOophagapumilio(A–D);andintrapopulationvariationinDendrobates
tinctorius(E–H).Photocredits:A–D,J.P.Lawrence;E–H,BibianaRojas.
patternsandafitness-relatedtrait(orgroupoftraits),either Rojas, Gordon & Mappes, 2015a). (iv) Relaxed selection
behaviouralorphysiological.Examplesofthisaredifferential towards warning signals or predator generalisation, which
activity patterns and investment in immune defences involvespredatoravoidanceofonesignalandtheexpansion
accordingtocolour,asinwoodtigermoths,ortheassociation of this aversion towards other signals similar enough to
between movement type and colour patterns in the dyeing the one learnt (Darst etal., 2006; Ame´zquita etal., 2013;
poison frog (Dendrobates tinctorius) (Nokelainen, Lindstedt & Richards-Zawacki, Yeager & Bart, 2013). (v) Non-adaptive
Mappes, 2013; Rojas, Devillechabrolle & Endler, 2014a; forces such as hybridisation among geographic variants or
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety
10 BibianaRojas
drift (Thompson, 1984; Gray & McKinnon, 2007; Medina 2012). This is because, if a trade-off between the resources
etal., 2013). Whatever the mechanism, for polymorphisms investedinwarningsignalsandchemicaldefencesisassumed,
to be maintained there should be no differences in fitness thebenefitofmoreconspicuoussignalsisalowriskofprey
amongthemorphs. beingattackedafterdetection,whereasthecostisareduced
defence level, which will bring decreased probabilities of
(b) Honestyinwarningsignals survival upon attack (Holen & Svennungsen, 2012). Toxins
such as those found in dendrobatid, mantellid or bufonid
The ‘handicap principle’ (Zahavi, 1975) suggests that
frogs are believed to be mainly sequestered from their
selection should favour signals that provide reliable
diet (Daly etal., 1997; Clark etal., 2005; Hantak etal.,
informationaboutanindividual’squality.Thesesignalsmust
2013), which consists primarily of leaf-litter arthropods.
representcostsforthesignaller,sothattheyareunaffordable
This means that, as suggested by previous studies (Wang,
for individuals whose quality is lower. Although suggested
2011; Maan & Cummings, 2012), variation in toxicity
originallyforsexualsignals,honestsignalshavethepotential
levels among populations of the same species might be
toevolveinpredator–preyrelationshipsaswell(Guilford&
duetovariationinpreyavailability.However,thisdoesnot
Dawkins,1993).
necessarily explain why some species (or individuals within
Foraposematicindividuals,itislikelythatwarningsignals
the same population) are more toxic, or more conspicuous,
arereliableindicatorsofprey’sunprofitability(e.g.toxicity),
thanothers.
as only well-defended prey can afford the costs of being
Findingcorrelations,eitherpositiveornegative,between
easily detectable by predators (Sherratt, 2002). However,
warning signals and the quality or quantity of chemical
detectability is not necessarily the only cost aposematic
defences in the species mentioned above has been very
species may incur. Theoretical studies have shown that
informativeonthedynamicsofpredator–preyrelationships
simultaneousinvestmentinpigmentsandchemicaldefences
in anurans. However, empirical studies tend to miss one
may trade off within the same individual (Blount etal.,
(or more) of the key elements about honesty, such as
2009). Thus, warning signals are likely to be honest signals
evidence that predators really pay attention to variation
of the quality (or quantity) of chemical defences, as only
in toxicity and/or see (and care about) the difference
well-defendedindividualscouldprofitablytoleratethecosts
amongsignals.Oftenonlyacorrelationbetweentoxicityand
ofstrongwarningsignals.
colouration is found, and the rest is assumed. Therefore,
Apositivecorrelationbetweenwarningsignalsandtoxicity
as Summers etal. (2015) state in a recent review, one
hasbeenfoundacrossspeciesofdartpoisonfrogsinastudy
key question is how to differentiate between quantitative
where phylogenetic constraints were taken into account,
honesty and other ways in which conspicuousness and
and where both colouration and toxicity were considered
toxicity may be correlated without involving honest
‘either/or’ traits (Summers & Clough, 2001). However,
signalling.
quantitative approaches to whether poison frogs’ warning
The next steps thus should also include observations
signals are honest or not have yielded mixed evidence.
and experiments leading to a better understanding of the
Forinstance,amongdifferentpopulationsofOophagapumilio
costs associated both with toxin sequestration and storage,
colouration has been suggested to be an honest indicator
and with the production of warning signals (i.e. pigments).
of toxicity, such that the brightest populations are also
For example, in species with variable colour patterns, it
the most toxic (Maan & Cummings, 2012). However, a
negative correlation between these two traits has also been could be tested whether in cohorts of individuals raised
found, both between and within species. The conspicuous fromlarvaeinthesameconditions,juvenilecolourpatterns
speciesAmeerega(Epipedobates)bilinguishasalowertoxicitylevel are correlated with life-history traits such as size and time
than its less-conspicuous counterpart, Ameerega (Epipedobates) to metamorphosis. Additionally, juveniles could be raised
parvulus (Darst etal., 2006). Notably, experiments with in semi-natural enclosures differing in diet restrictions to
chicks revealed that both increased conspicuousness and see whether resource availability affects colouration, or
increased toxicity are equally effective when it comes to toxicity, or both. A primer on this is a recent study on
predatordeterrence(Darstetal.,2006).Likewise,indifferent Dendrobates auratus, where the effects of rearing tadpoles on
populations of Oophaga granulifera conspicuous colouration eitherahigh-foodorlow-fooddietweretestedonbodysize
and toxicity are inversely related, such that the most and luminance at metamorphosis (Flores etal., 2013). The
conspicuous populations show the lowest levels of toxicity. authors found that in metamorphs raised on a high-food
In fact, these populations, where individuals are bright diet body size and luminance were negatively correlated,
red, lack four of the alkaloids present in populations with whereas this correlation was positive in froglets reared on
the less conspicuous yellow or green colouration (Wang, a low-food diet. According to Flores etal. (2013), these
2011). According to Blount etal. (2009), this scenario is findings suggest either a trade-off in resource allocation, or
plausiblewhenresourceavailabilityishigh,assuggestedfor developmentalplasticityaimedatminimisingpredationrisk
theseven-spotladybird,Coccinellaseptempunctata(Blountetal., atthemostvulnerable(early)stages.Life-historytrade-offsin
2012).However,arecenttheoreticalapproachfoundthat,at relationtofrogcolourationmaythereforebemorecommon
equilibrium,anegativecorrelationbetweenconspicuousness than usually thought, especially if the resource-allocation
anddefencestrengthdoesnothold(Holen&Svennungsen, hypothesisholds.
BiologicalReviews(2016)000–000©2016CambridgePhilosophicalSociety