225

HERPETOLOGICAL JOURNAL 17 225ndash236 2007HERPETOLOGICAL JOURNAL 17 225ndash236 2007HERPETOLOGICAL JOURNAL 17 225ndash236 2007HERPETOLOGICAL JOURNAL 17 225ndash236 2007HERPETOLOGICAL JOURNAL 17 225ndash236 2007

Specialist or generalist Feeding ecology of the Malagasypoison frog Mantella aurantiaca

Cindy WoodheadCindy WoodheadCindy WoodheadCindy WoodheadCindy Woodhead11111 Miguel Vences Miguel Vences Miguel Vences Miguel Vences Miguel Vences22222 David R Vieites David R Vieites David R Vieites David R Vieites David R Vieites33333 Ilona Gamboni Ilona Gamboni Ilona Gamboni Ilona Gamboni Ilona Gamboni11111 Brian L Brian L Brian L Brian L Brian LFisherFisherFisherFisherFisher44444 amp Richard A Griffiths amp Richard A Griffiths amp Richard A Griffiths amp Richard A Griffiths amp Richard A Griffiths11111

1Durrell Institute of Conservation and Ecology University of Kent Canterbury UK2Zoological Institute Technical University of Braunschweig Germany

3Museum of Vertebrate Zoology and Department of Integrative Biology University of California Berkeley USA4Department of Entomology California Academy of Sciences San Francisco USA

We studied the diet of a population of free-ranging Mantella aurantiaca an alkaloid-containing poison frog fromMadagascar As in other poison frogs this species is thought to sequester alkaloids from arthropod prey Among preymites and ants are known to regularly contain alkaloids and mites appear to be a major source of dietary alkaloids in poisonfrogs We predicted that mites and ants would constitute the most important prey item for these frogs Prey inventorieswere obtained during the rainy season by stomach flushing 23 adult male and 42 adult female frogs from one populationMales had smaller body sizes than females and ate smaller prey items but males and females displayed no differencesin the number of prey items consumed The numerical proportion of ants in most specimens was surprisingly low (11in males and 15 in females) while mites were slightly more frequent (34 in males and 18 in females) Other preyitems consumed in large proportions were flies and collembolans Comparing the total of 5492 arthropod prey itemswith 1867 arthropods sampled from the frogsrsquo leaf litter habitat the proportion of prey classes did not significantlydiffer among the samples indicating a low degree of prey electivity in this population Our data suggest that not all poisonfrogs exhibit a continuous and active preference for feeding on ants and mites but instead some may consume highproportions of ants due to a high abundance of ants in their environment

Key words Amphibia ant feeding Mantellidae Madagascar prey choice

Correspondence Miguel Vences Zoological Institute Technical University of Braunschweig Spielmannstr 838106 Braunschweig Germany E-mail mvencestu-bsde

INTRODUCTIONINTRODUCTIONINTRODUCTIONINTRODUCTIONINTRODUCTION

Specialization in foraging and feeding is known to be amajor trigger for evolutionary novelty and adaptive

radiation (Streelman amp Danley 2003) However in am-phibians habitat rather than food choice tends to causeresource partitioning (Toft 1985) In addition the strong-est factor influencing the radiations of anuran amphibiansmay be the striking diversification of reproductive modesand larval development (eg Wake 1982 Duellman ampTrueb 1986 Dubois 2005) Nevertheless numerousfrogs have evolved adaptations related to feeding mode(Nishikawa 1999 2000 Meyers et al 2004) Among themost fascinating of these are the alkaloid-containingmicrophagous and myrmecophagous taxa Alkaloidswhich supposedly play a role in defence from predatorsare found in the skins of poison frogs from four differentfamilies the neotropical Dendrobatidae (various genera)and Bufonidae (Melanophryniscus) the AustralianMyobatrachidae (Pseudophryne) and the MadagascanMantellidae (Mantella) (Daly et al 1987)

Among these alkaloid-containing taxa thedendrobatids and Mantella especially are relativelysmall diurnal and brightly coloured frogs Their preymainly consists of small arthropods with ants and mitesforming the majority of the diet (Simon amp Toft 1991 Toft1995 Caldwell 1996 Vences amp Kniel 1998 Summers ampClough 2001 Clark et al 2005 Darst et al 2005) Conse-

quently these frogs possess skull and tongue modifica-tions such as the reduction of maxillary and vomerineteeth and tongue width that may be adaptations for in-gesting small prey (Vences et al 1998)

Recent research suggests these frogs take up their al-kaloids from arthropod prey (eg Daly et al 1994 Daly1998 Daly et al 2002) with mites and ants contributingmost of their alkaloids (Saporito et al 2004 Clark et al2005 Takada et al 2005 Saporito et al 2007)Microphagousmyrmecophagous feeding and relatedspecializations of skull and tongue skin alkaloidsaposematic coloration and diurnal behaviour may consti-tute a closely linked suite of adaptations (Caldwell 1996Vences et al 1998) for which the successive chain of evo-lutionary novelty remains largely undetermined

The genus Mantella comprising the Malagasy poisonfrogs belongs to a radiation endemic to Madagascar andthe Comoro island of Mayotte (Glaw amp Vences 2003Vences et al 2003) All are considered members of thefamily Mantellidae (Frost et al 2006) Mantella containsabout 17 species of brightly-coloured diurnal frogs inhab-iting most of the bioclimatic and vegetation zones ofMadagascar (Daly et al 1996 Vences et al 1999a) Thecolour patterns of several species such as the black-yel-low-orange Mantella baroni and M madagascariensisthe black-orange M cowani and the uniformly orange Mmilotympanum and M aurantiaca are probablyaposematic This attractiveness has made Mantella

226

popular in the pet trade (Rabemananjara et al in press)and has led to their use as flagship species for habitat pro-tection (eg Zimmermann 1996) Indeed according toIUCN Red List categories three species of Mantella arecurrently considered Vulnerable two species Endan-gered and five species (M aurantiaca M cowani Mexpectata M milotympanum M viridis) Critically En-dangered (Andreone et al 2005) Habitat destruction isbelieved to constitute the primary threat to these specieswith the exception of M cowani which has also beenovercollected for the pet trade (Andreone ampRandrianirina 2003 Vences et al 2004)

Recent molecular data on Mantella has improved ourunderstanding of their phylogeny and aided in the evalu-ation of their genetics for conservation purposes(Schaefer et al 2002 Vences et al 2004 Chiari et al 20042005 Vieites et al 2006) Ecological studies on Mantellaare needed for conservation purposes (Andreone et al2005) to advance our understanding of the convergentevolution of coloration (Chiari et al 2004) and to identifyhow these frogs take up alkaloids from arthropods (Clarket al 2005) Yet such field studies remain remarkablyscarce Besides anecdotal information on habitat and col-lection localities (eg Daly et al 1996) only a few studieson distribution range population density predators andthe reproduction of single species have been published(eg Heying 2001ab Rabemananjara et al 2005 Vieiteset al 2005) Preliminary data on diet of Mantella werecollected by Vences amp Kniel (1998) for M betsileo Mharaldmeieri M laevigata and M nigricans RecentlyClark et al (2005) examined the stomach contents ofMantella baroni M bernhardi and Mmadagascariensis focusing on both the taxonomic com-position and alkaloid content of prey They found severalalkaloid-containing ants and millipedes to be major com-ponents of Mantella food indicating that preyspecialization may have been responsible for the evolu-tion of this frogrsquos alkaloid uptake system

In the present paper we provide data on the prey com-position of Mantella aurantiaca an aposematicuniformly orange species known to contain alkaloids(Daly et al 1996) By comparing the stomach contents ofthese frogs with the food available in leaf litter sampleswe hoped to determine whether the arthropods consumedby Mantella aurantiaca were due to active prey selectionor to a background abundance of arthropods in the envi-ronment

MATERIALS AND METHODSMATERIALS AND METHODSMATERIALS AND METHODSMATERIALS AND METHODSMATERIALS AND METHODS

Study siteStudy siteStudy siteStudy siteStudy site

The study was conducted in a forest bordering the natu-ral flood plain of the Torotorofotsy swamp (18ordm5229S48ordm2221E 960 m asl) near Andasibe Madagascar Veg-etation in the area consisted of a sparse forest(approximately 70ndash90 canopy cover) with many vinesand occasional shrubs We defined one transect 100 mlong and 10 m wide in an area of high Mantella abun-dance Some 50 m of this transect bordered a streamFieldwork was carried out from 20 to 22 February 2004

corresponding to the end of the period of peak activityand reproduction for these frogs

Frog processingFrog processingFrog processingFrog processingFrog processing

A total of 65 adult frogs was collected between 0600 and1700 and processed immediately after capture at a nearbycampsite Specimens were sexed based on the presence orabsence of the whitish femoral glands present in malesonly For each frog we measured snoutndashvent length(SVL) to the nearest 005 mm with callipers and mass (M)to the nearest 005 g with a Pesola scale

Stomach flushing was performed by inserting a smallflexible bevel-ended human plastic catheter (Cookrsquosprecutaneous entry TFE catheter 22 gauge) while thefrog was inverted (Legler amp Sullivan 1979 Opatrnyacute1980) During the insertion water was pushed gentlythrough the catheter with a large syringe (20 cm3) to pre-vent injury to the frog Once the catheter was insertedcompletely gentle water pressure was applied until thestomach contents were expelled into a receptacle Thiswas done until no more prey items were expelled andtested by touching the ventral section of the frogs exter-nally Stomach contents were preserved in 70 ethanolAfter stomach flushing frogs were marked by toe-clip-ping and released along their transect of origin

Leaf litter collectionLeaf litter collectionLeaf litter collectionLeaf litter collectionLeaf litter collection

Forty leaf-litter samples were taken from the same transectimmediately after the final frogs were processed to avoidaltering food availability over time All of the leaf litterwithin a 1 m times 1 m quadrat was removed from the forestfloor and placed in cloth mesh bags Samples wereweighed and divided in fractions of 005 kg and wereprocessed within a week of collection Each fraction wasplaced for several days in a Berlese funnel trap All leaveswere subsequently checked by hand to collect any re-maining arthropods All arthropod specimens werepreserved in 95 ethanol

Identification of arthropodsIdentification of arthropodsIdentification of arthropodsIdentification of arthropodsIdentification of arthropods

Stomach content samples were examined in a Petri dishwith a Harris micrometergraticule scale (1 cm long subdi-vided into 01 mm) (Griffiths 1986 Griffiths amp Mylotte1987) The lengths and widths of all organic matter in-cluding organisms particles and vegetation fragmentswere measured to the nearest 005 mm Lengths were con-sidered the longest anterior to posterior apices excludingappendages such as antennae mandibles and oviposi-tors Widths were measured at the widest girthperpendicular to length Erected wings were not countedas part of the width All organisms were identified to thelowest taxonomic level possible The size of fragmentedspecimens was estimated based on other specimens withintact bodies Arthropods from leaf litter samples wereprocessed in the same manner

All invertebrate length and width measurements weresorted into identification and size categories Roughmorphospecies were used in Diptera and consecutivelynumbered Volumes of arthropod specimens were calcu-lated using formulae (see Appendix) that bestapproximated the volumes of individual invertebrate cat-

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

227

egories (Griffiths 1986 Griffiths amp Mylotte 1987) Antswere identified to genus and morphospecies The pres-ence of each ant species per stomach content wasassessed

Statistical analysisStatistical analysisStatistical analysisStatistical analysisStatistical analysis

Data were entered into spreadsheets tracking frognumber date time field site location SVL M and numberof prey items in the stomach (NPI) Individual frog preydimensions were averaged to calculate average preylength (APL) average prey width (APW) average preyvolume (APV) and total stomach content volume (SCV)for each individual frog Separate calculations were madefor unmarked and recaptured frogs Values were com-pared between males and females and between unmarkedand recaptured frogs using t-tests to identify possiblechanges in food intake in frogs after stomach flushingConsumed prey and dietary availability were comparedusing the Strauss Food Selection Index (Strauss 1979) forprey occurrence numerical abundance in percent and vol-ume in percent (Hyslop 1980) Correlation and linearregressions were performed with SPSS 100

RESULTSRESULTSRESULTSRESULTSRESULTS

Prey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and size

We obtained stomach contents from 65 frogs of which 23were males and 42 were females Thirteen of these frogswere recaptured and stomach-flushed a second time Thenumber of prey items and average stomach content vol-ume were found to be significantly lower in recapturedfrogs of both sexes (t-tests Plt0005) although the frogsdid not differ significantly in size measurements Recap-tured frogs also appeared to have consumed smaller preybut the differences were not significant (P=0056) MeanNPIplusmnSD was 244plusmn405 in recaptured frogs vs 845plusmn596 inunmarked frogs Mean SCV was 32plusmn41 mm3 vs 131plusmn95mm3 This is probably due to the short time span betweencaptures which may not have been enough for the frogsto reach ldquoaveragerdquo levels of food items Consequently allfurther calculations were based on first-time capture data

SVL was 174ndash203 mm (mean = 190 SD = 08 mm) inmales 184ndash306 mm (mean = 222 SD = 19 mm) in femalesSixty-three of the unmarked frogs contained prey in theirstomachs two males had empty stomachs In total 5492prey items were identified 3880 items in females 1612items in males Male frogs were smaller than females (t-tests for SVL MV MH Plt0001 for all comparisons) andate smaller less voluminous prey items than females (t-tests for APL APV Plt005 for all comparisons)However females and males displayed no differences inthe number of prey items consumed (t-test for NPIP=0151) Comparing all prey items found in males withthose in females significant differences were found inprey length (t-test on IPL Plt0001) and prey volume (t-test for IPV Plt005) Average length of all prey itemsconsumed plusmnSD was 119plusmn082 mm (02ndash86 mm) in malesand 141plusmn112 mm (02ndash47 mm) in females Average volumeplusmnSD was 012plusmn064 mm3 in males and 017plusmn073 mm3 in fe-males The smallest prey items in males and females weremites and collembolans The largest observed prey items

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 1Fig 1Fig 1Fig 1Fig 1 The relationships between average prey length(APL) number of prey items (NPI) and snoutndashvent length(SVL) in male and female Mantella aurantiaca Blacksymbols and continuous regression lines are femaleswhite symbols and dashed lines are males a)Relationship between APL and SVL Slopes of regressionlines y=00006x+15131 P=0988 for females y=ndash00221x+16568 P=0792 for males Therelationship is not significant if male and female dataare analysed together (P=0079) b) Relationship of NPIand SVL Slopes of regression lines y=87567xndash10181 P=0092 for females y=84975xndash83943P=0531 for males The relationship is significant ifmale and female data are analysed together (Plt005)c) Relationship of APL and NPI Slopes of regressionlines y=ndash5354x+17405 Plt005 for females y=ndash57318x+14775 P=0117 for males Therelationship is significant if male and female data areanalysed together (Plt005)

228

were insect larvae isopods beetles centipedes andamphipods of 6ndash9 mm as well as one 47 mm centipedefound in a female frog Of all prey items 958 hadlengths of lt3 mm 829 were lt2 mm and 365 lt1 mm

Our data show no obvious trends for larger frogs to eatlarger prey (Fig 1a) Although females consumed largerprey this was apparently not due to their larger sizesHowever both sexes showed an increase in NPI con-sumed as SVL increased (Fig 1b) APL and NPI werenegatively correlated both when considering males andfemales together and for female data alone (Plt005) indi-cating that individuals that consumed larger prey had onaverage consumed fewer prey items (Fig 1c)

All prey categories were found in the stomachs of bothmales and females with the exception of wingedFormicidae and Lepidoptera larvae which were found ex-clusively in females The percent occurrence of preycategories as well as percentage and volume of preyitems per category (Spearman rank correlation Plt001Fig 2) was highly correlated between males and femalesFemales ate a larger number of soft-bodied Diptera 1while males consumed larger quantities of mites than fe-males Among males mites were the most frequent preyitems (341) followed by flies (244) collembolans(130) and ants (113) In females flies were the mostfrequent prey item (361) followed by mites (181)ants (153) and collembolans (147) However in bothmales and females ants made up the highest proportion

of stomach content volume (263 and 288 respec-tively) followed by flies (224 and 246 respectively)

Prey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availability

Arthropod availability and prey selection was based ondata from 11 leaf-litter samples with 1867 individual poten-tial prey items and compared against the same 65 frogsamples (42 females and 23 males) containing a total of5492 prey items in their stomachs All leaf-litter and stom-ach samples contained mites collembolans andnon-winged ants Flies (Diptera 1) were rare in leaf-littersamples but common in stomach samples This discrep-ancy may have been due to a methodological bias as theflies may have evaded capture within the leaf-litter sam-ples Other flies (Diptera 2) winged ants and thysanuranswere likewise absent from leaf-litter samples suggestingeither low numbers or evasion of capture Stomachslacked the prey classes Diptera 3 Blattodea Dermapteraand Nematoda although some large unidentified arthro-pod parts present in some frog stomachs may havebelonged to Blattodea Leaf-litter and stomach contentsamples were correlated in terms of occurrence numbersand volumes of prey categories (Spearman rank correla-tion Plt005) In all three categories frog diet and preyavailability in leaf litter mirrored each other (Figs 3ndash4 Ta-ble 1) The Strauss Food Selection Index (Strauss 1979)indicated low electivity regarding all prey classes for bothnumbers (Fig 5) and volumes of arthropods There may

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 2 Fig 2 Fig 2 Fig 2 Fig 2 Comparison of numerical abundance of different prey items in the diet of male and female Mantellaaurantiaca Values are averages of percentages for individual stomach contents and are correlated between sexes(Spearman rank correlation rs=0842 Plt001) Abbreviations Homoptera (Hom) Hemiptera (Hem) non-Formicidae (NF) winged (W) non-winged (NW)

229

also have been a negative election of mites and a positiveelection of flies (Diptera 1) but the latter may be due tomethodological issues (see above)

Stomachs contained ants from nine generaCrematogaster (1 morphospecies) Cataulacus (1)Anochetus (1) Paratrechina (1) Hypoponera (2)Strumigenys (3) Tetramorium (6) Monomorium (4) andPheidole (7) Frequency of occurrence was highest in onemorphospecies of Pheidole (present in 48 stomach con-tents) and two morphospecies of Monomorium (presentin 45 and 24 stomach contents) All other antmorphospecies were present in ten or fewer stomach con-tents only

DISCUSSIONDISCUSSIONDISCUSSIONDISCUSSIONDISCUSSION

Prey electivity inPrey electivity inPrey electivity inPrey electivity inPrey electivity in Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca

Our data indicate that the prey of Mantella aurantiacaconsists mainly of arthropods of small size different frommost other mantellid species where numerous large preyitems are found (Vences et al 1999b) Although ants werenot the most numerically common prey items (14) theycomprised the bulk of the prey consumed (28) in termsof volume Mites which have recently been shown to be amajor source of dietary alkaloids in poison frogs (Takadaet al 2005 Saporito et al 2007) were present in highnumbers (23) but only made up 5 of the volume of thestomach contents However it is insufficiently under-

stood whether volume or individual numbers of a particu-lar prey is more important to determine alkaloid contentand composition of poison frogs If all individuals of oneprey type contain similar alkaloids in similar concentra-tion and alkaloids are stored in the skins of the frogs forlong periods then volume is probably most important Ifonly some individuals of a certain prey category containalkaloids prey individuals differ in the concentration ofcontained alkaloids andor alkaloids are stored only for alimited time in the skins of the frogs then it is probablymore important that the frogs eat large numbers of thisprey category in order to ensure that at regular intervalssome prey individuals with high alkaloid concentrationare consumed

Ant species of Tetramorium were present in twelvestomach contents (summarized for the fourmorphospecies of this genus) One species of this genushad previously been found to contain two differentpumiliotoxin alkaloids (Clark et al 2005) A species ofParatrechina was present in a single stomach these antsare known to contain a variety of alkaloids and togetherwith other species of formicine ants could potentiallycontribute to the dietary alkaloids of Mantellaaurantiaca Several other arthropods found in the stom-ach contents such as millipedes and beetles couldpotentially provide dietary alkaloids as well (eg Ayerand Browne 1977 Daly et al 2002 Saporito et al 2003Clark et al 2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 3 Fig 3 Fig 3 Fig 3 Fig 3 Comparison of prey consumed with prey availability Columns show the percentage of stomach and littersamples that contained different prey items Data are correlated (Spearman rank correlation rs=0564 Plt001)Abbreviations as in Fig 2

230

Prey size did not appear to be a function of frog bodysize Though larger frogs eat more prey it is of the sameaverage size as prey eaten by smaller frogs This con-trasts with data from other studies which found frogs ofgreater snoutndashvent length consuming fewer but largerprey (eg Berry 1966 Brooks 1982 Hirai 2002Hodgkinson amp Hero 2003 Ramiacuterez-Bautista amp Lemos-Espinal 2004) The fact that large and small individuals ofMantella aurantiaca consumed arthropods of the sameaverage size indicates little preference for different-sizedprey

This is confirmed by the comparisons of prey in theleaf litter versus the stomach contents The electivityanalysis provided little evidence for prey selection In-deed cases where preference was apparent (eg for flies)may in fact be artefacts of our collection method ThusMantella aurantiaca appeared to consume prey in pro-portion to its availability The one distinct exception weremites ndash these were consumed at lower proportions eitherbecause 1) they are difficult to find due to their small sizeand cryptic habits or 2) there is active discriminationagainst such small prey Few large prey items were con-sumed but such large arthropods were apparently alsorare in the leaf litter Hence M aurantiaca followed amicrophagous ndash but not a myrmecophagous ndash foragingmode but this pattern is a reflection of the abundance ofsmall prey in the environment at the study site and in thestudy period MacNally (1983) studied a similar case inwhich two species of Ranidella mainly consumed small

prey below 15 mm in length but with a very low propor-tion of ants reflecting the patterns of availability ofdifferent prey types and prey sizes

Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other MantellaMantellaMantellaMantellaMantella

Previously published results have observed a high per-centage (74) of ants in stomachs of Mantellaharaldmeieri M nigricans M laevigata and M betsileo(Vences amp Kniel 1998) and 67 ants in M baroni Mbernhardi and M madagascariensis (Clark et al 2005)The percentages of ants consumed by M aurantiaca aremuch greater than the proportion of ants consumed bymost other mantellid frogs analysed (generaAglyptodactylus Boophis Gephyromantis Laliostomaand Mantidactylus) usually because many of these spe-cies consume much larger prey (Vences et al 1999b) Atfirst glance a comparison of our results with these pub-lished data seems to indicate an important differencebetween M aurantiaca and other Mantella specieswhereas congeneric species appear to be ant specialistsM aurantiaca does not display such a preference

However previous studies did not include data onfood availability and therefore it is not yet possible todetermine whether the dietary pattern of Mantellaaurantiaca is common among other species in this ge-nus Our data do not support the idea of Mantellaaurantiaca as a specialist although this species may dis-play different feeding patterns with clear preference forants in other populations or in other seasons However it

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 4 Fig 4 Fig 4 Fig 4 Fig 4 Comparison of prey consumed with prey availability Columns show the percentage numerical abundance ofprey items in stomachs and leaf litter Data are correlated (Spearman rank correlation rs=0613 Plt001)Abbreviations as in Fig 2

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

226

popular in the pet trade (Rabemananjara et al in press)and has led to their use as flagship species for habitat pro-tection (eg Zimmermann 1996) Indeed according toIUCN Red List categories three species of Mantella arecurrently considered Vulnerable two species Endan-gered and five species (M aurantiaca M cowani Mexpectata M milotympanum M viridis) Critically En-dangered (Andreone et al 2005) Habitat destruction isbelieved to constitute the primary threat to these specieswith the exception of M cowani which has also beenovercollected for the pet trade (Andreone ampRandrianirina 2003 Vences et al 2004)

Recent molecular data on Mantella has improved ourunderstanding of their phylogeny and aided in the evalu-ation of their genetics for conservation purposes(Schaefer et al 2002 Vences et al 2004 Chiari et al 20042005 Vieites et al 2006) Ecological studies on Mantellaare needed for conservation purposes (Andreone et al2005) to advance our understanding of the convergentevolution of coloration (Chiari et al 2004) and to identifyhow these frogs take up alkaloids from arthropods (Clarket al 2005) Yet such field studies remain remarkablyscarce Besides anecdotal information on habitat and col-lection localities (eg Daly et al 1996) only a few studieson distribution range population density predators andthe reproduction of single species have been published(eg Heying 2001ab Rabemananjara et al 2005 Vieiteset al 2005) Preliminary data on diet of Mantella werecollected by Vences amp Kniel (1998) for M betsileo Mharaldmeieri M laevigata and M nigricans RecentlyClark et al (2005) examined the stomach contents ofMantella baroni M bernhardi and Mmadagascariensis focusing on both the taxonomic com-position and alkaloid content of prey They found severalalkaloid-containing ants and millipedes to be major com-ponents of Mantella food indicating that preyspecialization may have been responsible for the evolu-tion of this frogrsquos alkaloid uptake system

In the present paper we provide data on the prey com-position of Mantella aurantiaca an aposematicuniformly orange species known to contain alkaloids(Daly et al 1996) By comparing the stomach contents ofthese frogs with the food available in leaf litter sampleswe hoped to determine whether the arthropods consumedby Mantella aurantiaca were due to active prey selectionor to a background abundance of arthropods in the envi-ronment

MATERIALS AND METHODSMATERIALS AND METHODSMATERIALS AND METHODSMATERIALS AND METHODSMATERIALS AND METHODS

Study siteStudy siteStudy siteStudy siteStudy site

The study was conducted in a forest bordering the natu-ral flood plain of the Torotorofotsy swamp (18ordm5229S48ordm2221E 960 m asl) near Andasibe Madagascar Veg-etation in the area consisted of a sparse forest(approximately 70ndash90 canopy cover) with many vinesand occasional shrubs We defined one transect 100 mlong and 10 m wide in an area of high Mantella abun-dance Some 50 m of this transect bordered a streamFieldwork was carried out from 20 to 22 February 2004

corresponding to the end of the period of peak activityand reproduction for these frogs

Frog processingFrog processingFrog processingFrog processingFrog processing

A total of 65 adult frogs was collected between 0600 and1700 and processed immediately after capture at a nearbycampsite Specimens were sexed based on the presence orabsence of the whitish femoral glands present in malesonly For each frog we measured snoutndashvent length(SVL) to the nearest 005 mm with callipers and mass (M)to the nearest 005 g with a Pesola scale

Stomach flushing was performed by inserting a smallflexible bevel-ended human plastic catheter (Cookrsquosprecutaneous entry TFE catheter 22 gauge) while thefrog was inverted (Legler amp Sullivan 1979 Opatrnyacute1980) During the insertion water was pushed gentlythrough the catheter with a large syringe (20 cm3) to pre-vent injury to the frog Once the catheter was insertedcompletely gentle water pressure was applied until thestomach contents were expelled into a receptacle Thiswas done until no more prey items were expelled andtested by touching the ventral section of the frogs exter-nally Stomach contents were preserved in 70 ethanolAfter stomach flushing frogs were marked by toe-clip-ping and released along their transect of origin

Leaf litter collectionLeaf litter collectionLeaf litter collectionLeaf litter collectionLeaf litter collection

Forty leaf-litter samples were taken from the same transectimmediately after the final frogs were processed to avoidaltering food availability over time All of the leaf litterwithin a 1 m times 1 m quadrat was removed from the forestfloor and placed in cloth mesh bags Samples wereweighed and divided in fractions of 005 kg and wereprocessed within a week of collection Each fraction wasplaced for several days in a Berlese funnel trap All leaveswere subsequently checked by hand to collect any re-maining arthropods All arthropod specimens werepreserved in 95 ethanol

Identification of arthropodsIdentification of arthropodsIdentification of arthropodsIdentification of arthropodsIdentification of arthropods

Stomach content samples were examined in a Petri dishwith a Harris micrometergraticule scale (1 cm long subdi-vided into 01 mm) (Griffiths 1986 Griffiths amp Mylotte1987) The lengths and widths of all organic matter in-cluding organisms particles and vegetation fragmentswere measured to the nearest 005 mm Lengths were con-sidered the longest anterior to posterior apices excludingappendages such as antennae mandibles and oviposi-tors Widths were measured at the widest girthperpendicular to length Erected wings were not countedas part of the width All organisms were identified to thelowest taxonomic level possible The size of fragmentedspecimens was estimated based on other specimens withintact bodies Arthropods from leaf litter samples wereprocessed in the same manner

All invertebrate length and width measurements weresorted into identification and size categories Roughmorphospecies were used in Diptera and consecutivelynumbered Volumes of arthropod specimens were calcu-lated using formulae (see Appendix) that bestapproximated the volumes of individual invertebrate cat-

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

227

egories (Griffiths 1986 Griffiths amp Mylotte 1987) Antswere identified to genus and morphospecies The pres-ence of each ant species per stomach content wasassessed

Statistical analysisStatistical analysisStatistical analysisStatistical analysisStatistical analysis

Data were entered into spreadsheets tracking frognumber date time field site location SVL M and numberof prey items in the stomach (NPI) Individual frog preydimensions were averaged to calculate average preylength (APL) average prey width (APW) average preyvolume (APV) and total stomach content volume (SCV)for each individual frog Separate calculations were madefor unmarked and recaptured frogs Values were com-pared between males and females and between unmarkedand recaptured frogs using t-tests to identify possiblechanges in food intake in frogs after stomach flushingConsumed prey and dietary availability were comparedusing the Strauss Food Selection Index (Strauss 1979) forprey occurrence numerical abundance in percent and vol-ume in percent (Hyslop 1980) Correlation and linearregressions were performed with SPSS 100

RESULTSRESULTSRESULTSRESULTSRESULTS

Prey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and size

We obtained stomach contents from 65 frogs of which 23were males and 42 were females Thirteen of these frogswere recaptured and stomach-flushed a second time Thenumber of prey items and average stomach content vol-ume were found to be significantly lower in recapturedfrogs of both sexes (t-tests Plt0005) although the frogsdid not differ significantly in size measurements Recap-tured frogs also appeared to have consumed smaller preybut the differences were not significant (P=0056) MeanNPIplusmnSD was 244plusmn405 in recaptured frogs vs 845plusmn596 inunmarked frogs Mean SCV was 32plusmn41 mm3 vs 131plusmn95mm3 This is probably due to the short time span betweencaptures which may not have been enough for the frogsto reach ldquoaveragerdquo levels of food items Consequently allfurther calculations were based on first-time capture data

SVL was 174ndash203 mm (mean = 190 SD = 08 mm) inmales 184ndash306 mm (mean = 222 SD = 19 mm) in femalesSixty-three of the unmarked frogs contained prey in theirstomachs two males had empty stomachs In total 5492prey items were identified 3880 items in females 1612items in males Male frogs were smaller than females (t-tests for SVL MV MH Plt0001 for all comparisons) andate smaller less voluminous prey items than females (t-tests for APL APV Plt005 for all comparisons)However females and males displayed no differences inthe number of prey items consumed (t-test for NPIP=0151) Comparing all prey items found in males withthose in females significant differences were found inprey length (t-test on IPL Plt0001) and prey volume (t-test for IPV Plt005) Average length of all prey itemsconsumed plusmnSD was 119plusmn082 mm (02ndash86 mm) in malesand 141plusmn112 mm (02ndash47 mm) in females Average volumeplusmnSD was 012plusmn064 mm3 in males and 017plusmn073 mm3 in fe-males The smallest prey items in males and females weremites and collembolans The largest observed prey items

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 1Fig 1Fig 1Fig 1Fig 1 The relationships between average prey length(APL) number of prey items (NPI) and snoutndashvent length(SVL) in male and female Mantella aurantiaca Blacksymbols and continuous regression lines are femaleswhite symbols and dashed lines are males a)Relationship between APL and SVL Slopes of regressionlines y=00006x+15131 P=0988 for females y=ndash00221x+16568 P=0792 for males Therelationship is not significant if male and female dataare analysed together (P=0079) b) Relationship of NPIand SVL Slopes of regression lines y=87567xndash10181 P=0092 for females y=84975xndash83943P=0531 for males The relationship is significant ifmale and female data are analysed together (Plt005)c) Relationship of APL and NPI Slopes of regressionlines y=ndash5354x+17405 Plt005 for females y=ndash57318x+14775 P=0117 for males Therelationship is significant if male and female data areanalysed together (Plt005)

228

were insect larvae isopods beetles centipedes andamphipods of 6ndash9 mm as well as one 47 mm centipedefound in a female frog Of all prey items 958 hadlengths of lt3 mm 829 were lt2 mm and 365 lt1 mm

Our data show no obvious trends for larger frogs to eatlarger prey (Fig 1a) Although females consumed largerprey this was apparently not due to their larger sizesHowever both sexes showed an increase in NPI con-sumed as SVL increased (Fig 1b) APL and NPI werenegatively correlated both when considering males andfemales together and for female data alone (Plt005) indi-cating that individuals that consumed larger prey had onaverage consumed fewer prey items (Fig 1c)

All prey categories were found in the stomachs of bothmales and females with the exception of wingedFormicidae and Lepidoptera larvae which were found ex-clusively in females The percent occurrence of preycategories as well as percentage and volume of preyitems per category (Spearman rank correlation Plt001Fig 2) was highly correlated between males and femalesFemales ate a larger number of soft-bodied Diptera 1while males consumed larger quantities of mites than fe-males Among males mites were the most frequent preyitems (341) followed by flies (244) collembolans(130) and ants (113) In females flies were the mostfrequent prey item (361) followed by mites (181)ants (153) and collembolans (147) However in bothmales and females ants made up the highest proportion

of stomach content volume (263 and 288 respec-tively) followed by flies (224 and 246 respectively)

Prey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availability

Arthropod availability and prey selection was based ondata from 11 leaf-litter samples with 1867 individual poten-tial prey items and compared against the same 65 frogsamples (42 females and 23 males) containing a total of5492 prey items in their stomachs All leaf-litter and stom-ach samples contained mites collembolans andnon-winged ants Flies (Diptera 1) were rare in leaf-littersamples but common in stomach samples This discrep-ancy may have been due to a methodological bias as theflies may have evaded capture within the leaf-litter sam-ples Other flies (Diptera 2) winged ants and thysanuranswere likewise absent from leaf-litter samples suggestingeither low numbers or evasion of capture Stomachslacked the prey classes Diptera 3 Blattodea Dermapteraand Nematoda although some large unidentified arthro-pod parts present in some frog stomachs may havebelonged to Blattodea Leaf-litter and stomach contentsamples were correlated in terms of occurrence numbersand volumes of prey categories (Spearman rank correla-tion Plt005) In all three categories frog diet and preyavailability in leaf litter mirrored each other (Figs 3ndash4 Ta-ble 1) The Strauss Food Selection Index (Strauss 1979)indicated low electivity regarding all prey classes for bothnumbers (Fig 5) and volumes of arthropods There may

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 2 Fig 2 Fig 2 Fig 2 Fig 2 Comparison of numerical abundance of different prey items in the diet of male and female Mantellaaurantiaca Values are averages of percentages for individual stomach contents and are correlated between sexes(Spearman rank correlation rs=0842 Plt001) Abbreviations Homoptera (Hom) Hemiptera (Hem) non-Formicidae (NF) winged (W) non-winged (NW)

229

also have been a negative election of mites and a positiveelection of flies (Diptera 1) but the latter may be due tomethodological issues (see above)

Stomachs contained ants from nine generaCrematogaster (1 morphospecies) Cataulacus (1)Anochetus (1) Paratrechina (1) Hypoponera (2)Strumigenys (3) Tetramorium (6) Monomorium (4) andPheidole (7) Frequency of occurrence was highest in onemorphospecies of Pheidole (present in 48 stomach con-tents) and two morphospecies of Monomorium (presentin 45 and 24 stomach contents) All other antmorphospecies were present in ten or fewer stomach con-tents only

DISCUSSIONDISCUSSIONDISCUSSIONDISCUSSIONDISCUSSION

Prey electivity inPrey electivity inPrey electivity inPrey electivity inPrey electivity in Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca

Our data indicate that the prey of Mantella aurantiacaconsists mainly of arthropods of small size different frommost other mantellid species where numerous large preyitems are found (Vences et al 1999b) Although ants werenot the most numerically common prey items (14) theycomprised the bulk of the prey consumed (28) in termsof volume Mites which have recently been shown to be amajor source of dietary alkaloids in poison frogs (Takadaet al 2005 Saporito et al 2007) were present in highnumbers (23) but only made up 5 of the volume of thestomach contents However it is insufficiently under-

stood whether volume or individual numbers of a particu-lar prey is more important to determine alkaloid contentand composition of poison frogs If all individuals of oneprey type contain similar alkaloids in similar concentra-tion and alkaloids are stored in the skins of the frogs forlong periods then volume is probably most important Ifonly some individuals of a certain prey category containalkaloids prey individuals differ in the concentration ofcontained alkaloids andor alkaloids are stored only for alimited time in the skins of the frogs then it is probablymore important that the frogs eat large numbers of thisprey category in order to ensure that at regular intervalssome prey individuals with high alkaloid concentrationare consumed

Ant species of Tetramorium were present in twelvestomach contents (summarized for the fourmorphospecies of this genus) One species of this genushad previously been found to contain two differentpumiliotoxin alkaloids (Clark et al 2005) A species ofParatrechina was present in a single stomach these antsare known to contain a variety of alkaloids and togetherwith other species of formicine ants could potentiallycontribute to the dietary alkaloids of Mantellaaurantiaca Several other arthropods found in the stom-ach contents such as millipedes and beetles couldpotentially provide dietary alkaloids as well (eg Ayerand Browne 1977 Daly et al 2002 Saporito et al 2003Clark et al 2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 3 Fig 3 Fig 3 Fig 3 Fig 3 Comparison of prey consumed with prey availability Columns show the percentage of stomach and littersamples that contained different prey items Data are correlated (Spearman rank correlation rs=0564 Plt001)Abbreviations as in Fig 2

230

Prey size did not appear to be a function of frog bodysize Though larger frogs eat more prey it is of the sameaverage size as prey eaten by smaller frogs This con-trasts with data from other studies which found frogs ofgreater snoutndashvent length consuming fewer but largerprey (eg Berry 1966 Brooks 1982 Hirai 2002Hodgkinson amp Hero 2003 Ramiacuterez-Bautista amp Lemos-Espinal 2004) The fact that large and small individuals ofMantella aurantiaca consumed arthropods of the sameaverage size indicates little preference for different-sizedprey

This is confirmed by the comparisons of prey in theleaf litter versus the stomach contents The electivityanalysis provided little evidence for prey selection In-deed cases where preference was apparent (eg for flies)may in fact be artefacts of our collection method ThusMantella aurantiaca appeared to consume prey in pro-portion to its availability The one distinct exception weremites ndash these were consumed at lower proportions eitherbecause 1) they are difficult to find due to their small sizeand cryptic habits or 2) there is active discriminationagainst such small prey Few large prey items were con-sumed but such large arthropods were apparently alsorare in the leaf litter Hence M aurantiaca followed amicrophagous ndash but not a myrmecophagous ndash foragingmode but this pattern is a reflection of the abundance ofsmall prey in the environment at the study site and in thestudy period MacNally (1983) studied a similar case inwhich two species of Ranidella mainly consumed small

prey below 15 mm in length but with a very low propor-tion of ants reflecting the patterns of availability ofdifferent prey types and prey sizes

Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other MantellaMantellaMantellaMantellaMantella

Previously published results have observed a high per-centage (74) of ants in stomachs of Mantellaharaldmeieri M nigricans M laevigata and M betsileo(Vences amp Kniel 1998) and 67 ants in M baroni Mbernhardi and M madagascariensis (Clark et al 2005)The percentages of ants consumed by M aurantiaca aremuch greater than the proportion of ants consumed bymost other mantellid frogs analysed (generaAglyptodactylus Boophis Gephyromantis Laliostomaand Mantidactylus) usually because many of these spe-cies consume much larger prey (Vences et al 1999b) Atfirst glance a comparison of our results with these pub-lished data seems to indicate an important differencebetween M aurantiaca and other Mantella specieswhereas congeneric species appear to be ant specialistsM aurantiaca does not display such a preference

However previous studies did not include data onfood availability and therefore it is not yet possible todetermine whether the dietary pattern of Mantellaaurantiaca is common among other species in this ge-nus Our data do not support the idea of Mantellaaurantiaca as a specialist although this species may dis-play different feeding patterns with clear preference forants in other populations or in other seasons However it

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 4 Fig 4 Fig 4 Fig 4 Fig 4 Comparison of prey consumed with prey availability Columns show the percentage numerical abundance ofprey items in stomachs and leaf litter Data are correlated (Spearman rank correlation rs=0613 Plt001)Abbreviations as in Fig 2

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

227

egories (Griffiths 1986 Griffiths amp Mylotte 1987) Antswere identified to genus and morphospecies The pres-ence of each ant species per stomach content wasassessed

Statistical analysisStatistical analysisStatistical analysisStatistical analysisStatistical analysis

Data were entered into spreadsheets tracking frognumber date time field site location SVL M and numberof prey items in the stomach (NPI) Individual frog preydimensions were averaged to calculate average preylength (APL) average prey width (APW) average preyvolume (APV) and total stomach content volume (SCV)for each individual frog Separate calculations were madefor unmarked and recaptured frogs Values were com-pared between males and females and between unmarkedand recaptured frogs using t-tests to identify possiblechanges in food intake in frogs after stomach flushingConsumed prey and dietary availability were comparedusing the Strauss Food Selection Index (Strauss 1979) forprey occurrence numerical abundance in percent and vol-ume in percent (Hyslop 1980) Correlation and linearregressions were performed with SPSS 100

RESULTSRESULTSRESULTSRESULTSRESULTS

Prey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and sizePrey use in relation to frog sex and size

We obtained stomach contents from 65 frogs of which 23were males and 42 were females Thirteen of these frogswere recaptured and stomach-flushed a second time Thenumber of prey items and average stomach content vol-ume were found to be significantly lower in recapturedfrogs of both sexes (t-tests Plt0005) although the frogsdid not differ significantly in size measurements Recap-tured frogs also appeared to have consumed smaller preybut the differences were not significant (P=0056) MeanNPIplusmnSD was 244plusmn405 in recaptured frogs vs 845plusmn596 inunmarked frogs Mean SCV was 32plusmn41 mm3 vs 131plusmn95mm3 This is probably due to the short time span betweencaptures which may not have been enough for the frogsto reach ldquoaveragerdquo levels of food items Consequently allfurther calculations were based on first-time capture data

SVL was 174ndash203 mm (mean = 190 SD = 08 mm) inmales 184ndash306 mm (mean = 222 SD = 19 mm) in femalesSixty-three of the unmarked frogs contained prey in theirstomachs two males had empty stomachs In total 5492prey items were identified 3880 items in females 1612items in males Male frogs were smaller than females (t-tests for SVL MV MH Plt0001 for all comparisons) andate smaller less voluminous prey items than females (t-tests for APL APV Plt005 for all comparisons)However females and males displayed no differences inthe number of prey items consumed (t-test for NPIP=0151) Comparing all prey items found in males withthose in females significant differences were found inprey length (t-test on IPL Plt0001) and prey volume (t-test for IPV Plt005) Average length of all prey itemsconsumed plusmnSD was 119plusmn082 mm (02ndash86 mm) in malesand 141plusmn112 mm (02ndash47 mm) in females Average volumeplusmnSD was 012plusmn064 mm3 in males and 017plusmn073 mm3 in fe-males The smallest prey items in males and females weremites and collembolans The largest observed prey items

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 1Fig 1Fig 1Fig 1Fig 1 The relationships between average prey length(APL) number of prey items (NPI) and snoutndashvent length(SVL) in male and female Mantella aurantiaca Blacksymbols and continuous regression lines are femaleswhite symbols and dashed lines are males a)Relationship between APL and SVL Slopes of regressionlines y=00006x+15131 P=0988 for females y=ndash00221x+16568 P=0792 for males Therelationship is not significant if male and female dataare analysed together (P=0079) b) Relationship of NPIand SVL Slopes of regression lines y=87567xndash10181 P=0092 for females y=84975xndash83943P=0531 for males The relationship is significant ifmale and female data are analysed together (Plt005)c) Relationship of APL and NPI Slopes of regressionlines y=ndash5354x+17405 Plt005 for females y=ndash57318x+14775 P=0117 for males Therelationship is significant if male and female data areanalysed together (Plt005)

228

were insect larvae isopods beetles centipedes andamphipods of 6ndash9 mm as well as one 47 mm centipedefound in a female frog Of all prey items 958 hadlengths of lt3 mm 829 were lt2 mm and 365 lt1 mm

Our data show no obvious trends for larger frogs to eatlarger prey (Fig 1a) Although females consumed largerprey this was apparently not due to their larger sizesHowever both sexes showed an increase in NPI con-sumed as SVL increased (Fig 1b) APL and NPI werenegatively correlated both when considering males andfemales together and for female data alone (Plt005) indi-cating that individuals that consumed larger prey had onaverage consumed fewer prey items (Fig 1c)

All prey categories were found in the stomachs of bothmales and females with the exception of wingedFormicidae and Lepidoptera larvae which were found ex-clusively in females The percent occurrence of preycategories as well as percentage and volume of preyitems per category (Spearman rank correlation Plt001Fig 2) was highly correlated between males and femalesFemales ate a larger number of soft-bodied Diptera 1while males consumed larger quantities of mites than fe-males Among males mites were the most frequent preyitems (341) followed by flies (244) collembolans(130) and ants (113) In females flies were the mostfrequent prey item (361) followed by mites (181)ants (153) and collembolans (147) However in bothmales and females ants made up the highest proportion

of stomach content volume (263 and 288 respec-tively) followed by flies (224 and 246 respectively)

Prey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availability

Arthropod availability and prey selection was based ondata from 11 leaf-litter samples with 1867 individual poten-tial prey items and compared against the same 65 frogsamples (42 females and 23 males) containing a total of5492 prey items in their stomachs All leaf-litter and stom-ach samples contained mites collembolans andnon-winged ants Flies (Diptera 1) were rare in leaf-littersamples but common in stomach samples This discrep-ancy may have been due to a methodological bias as theflies may have evaded capture within the leaf-litter sam-ples Other flies (Diptera 2) winged ants and thysanuranswere likewise absent from leaf-litter samples suggestingeither low numbers or evasion of capture Stomachslacked the prey classes Diptera 3 Blattodea Dermapteraand Nematoda although some large unidentified arthro-pod parts present in some frog stomachs may havebelonged to Blattodea Leaf-litter and stomach contentsamples were correlated in terms of occurrence numbersand volumes of prey categories (Spearman rank correla-tion Plt005) In all three categories frog diet and preyavailability in leaf litter mirrored each other (Figs 3ndash4 Ta-ble 1) The Strauss Food Selection Index (Strauss 1979)indicated low electivity regarding all prey classes for bothnumbers (Fig 5) and volumes of arthropods There may

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 2 Fig 2 Fig 2 Fig 2 Fig 2 Comparison of numerical abundance of different prey items in the diet of male and female Mantellaaurantiaca Values are averages of percentages for individual stomach contents and are correlated between sexes(Spearman rank correlation rs=0842 Plt001) Abbreviations Homoptera (Hom) Hemiptera (Hem) non-Formicidae (NF) winged (W) non-winged (NW)

229

also have been a negative election of mites and a positiveelection of flies (Diptera 1) but the latter may be due tomethodological issues (see above)

Stomachs contained ants from nine generaCrematogaster (1 morphospecies) Cataulacus (1)Anochetus (1) Paratrechina (1) Hypoponera (2)Strumigenys (3) Tetramorium (6) Monomorium (4) andPheidole (7) Frequency of occurrence was highest in onemorphospecies of Pheidole (present in 48 stomach con-tents) and two morphospecies of Monomorium (presentin 45 and 24 stomach contents) All other antmorphospecies were present in ten or fewer stomach con-tents only

DISCUSSIONDISCUSSIONDISCUSSIONDISCUSSIONDISCUSSION

Prey electivity inPrey electivity inPrey electivity inPrey electivity inPrey electivity in Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca

Our data indicate that the prey of Mantella aurantiacaconsists mainly of arthropods of small size different frommost other mantellid species where numerous large preyitems are found (Vences et al 1999b) Although ants werenot the most numerically common prey items (14) theycomprised the bulk of the prey consumed (28) in termsof volume Mites which have recently been shown to be amajor source of dietary alkaloids in poison frogs (Takadaet al 2005 Saporito et al 2007) were present in highnumbers (23) but only made up 5 of the volume of thestomach contents However it is insufficiently under-

stood whether volume or individual numbers of a particu-lar prey is more important to determine alkaloid contentand composition of poison frogs If all individuals of oneprey type contain similar alkaloids in similar concentra-tion and alkaloids are stored in the skins of the frogs forlong periods then volume is probably most important Ifonly some individuals of a certain prey category containalkaloids prey individuals differ in the concentration ofcontained alkaloids andor alkaloids are stored only for alimited time in the skins of the frogs then it is probablymore important that the frogs eat large numbers of thisprey category in order to ensure that at regular intervalssome prey individuals with high alkaloid concentrationare consumed

Ant species of Tetramorium were present in twelvestomach contents (summarized for the fourmorphospecies of this genus) One species of this genushad previously been found to contain two differentpumiliotoxin alkaloids (Clark et al 2005) A species ofParatrechina was present in a single stomach these antsare known to contain a variety of alkaloids and togetherwith other species of formicine ants could potentiallycontribute to the dietary alkaloids of Mantellaaurantiaca Several other arthropods found in the stom-ach contents such as millipedes and beetles couldpotentially provide dietary alkaloids as well (eg Ayerand Browne 1977 Daly et al 2002 Saporito et al 2003Clark et al 2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 3 Fig 3 Fig 3 Fig 3 Fig 3 Comparison of prey consumed with prey availability Columns show the percentage of stomach and littersamples that contained different prey items Data are correlated (Spearman rank correlation rs=0564 Plt001)Abbreviations as in Fig 2

230

Prey size did not appear to be a function of frog bodysize Though larger frogs eat more prey it is of the sameaverage size as prey eaten by smaller frogs This con-trasts with data from other studies which found frogs ofgreater snoutndashvent length consuming fewer but largerprey (eg Berry 1966 Brooks 1982 Hirai 2002Hodgkinson amp Hero 2003 Ramiacuterez-Bautista amp Lemos-Espinal 2004) The fact that large and small individuals ofMantella aurantiaca consumed arthropods of the sameaverage size indicates little preference for different-sizedprey

This is confirmed by the comparisons of prey in theleaf litter versus the stomach contents The electivityanalysis provided little evidence for prey selection In-deed cases where preference was apparent (eg for flies)may in fact be artefacts of our collection method ThusMantella aurantiaca appeared to consume prey in pro-portion to its availability The one distinct exception weremites ndash these were consumed at lower proportions eitherbecause 1) they are difficult to find due to their small sizeand cryptic habits or 2) there is active discriminationagainst such small prey Few large prey items were con-sumed but such large arthropods were apparently alsorare in the leaf litter Hence M aurantiaca followed amicrophagous ndash but not a myrmecophagous ndash foragingmode but this pattern is a reflection of the abundance ofsmall prey in the environment at the study site and in thestudy period MacNally (1983) studied a similar case inwhich two species of Ranidella mainly consumed small

prey below 15 mm in length but with a very low propor-tion of ants reflecting the patterns of availability ofdifferent prey types and prey sizes

Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other MantellaMantellaMantellaMantellaMantella

Previously published results have observed a high per-centage (74) of ants in stomachs of Mantellaharaldmeieri M nigricans M laevigata and M betsileo(Vences amp Kniel 1998) and 67 ants in M baroni Mbernhardi and M madagascariensis (Clark et al 2005)The percentages of ants consumed by M aurantiaca aremuch greater than the proportion of ants consumed bymost other mantellid frogs analysed (generaAglyptodactylus Boophis Gephyromantis Laliostomaand Mantidactylus) usually because many of these spe-cies consume much larger prey (Vences et al 1999b) Atfirst glance a comparison of our results with these pub-lished data seems to indicate an important differencebetween M aurantiaca and other Mantella specieswhereas congeneric species appear to be ant specialistsM aurantiaca does not display such a preference

However previous studies did not include data onfood availability and therefore it is not yet possible todetermine whether the dietary pattern of Mantellaaurantiaca is common among other species in this ge-nus Our data do not support the idea of Mantellaaurantiaca as a specialist although this species may dis-play different feeding patterns with clear preference forants in other populations or in other seasons However it

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 4 Fig 4 Fig 4 Fig 4 Fig 4 Comparison of prey consumed with prey availability Columns show the percentage numerical abundance ofprey items in stomachs and leaf litter Data are correlated (Spearman rank correlation rs=0613 Plt001)Abbreviations as in Fig 2

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

228

were insect larvae isopods beetles centipedes andamphipods of 6ndash9 mm as well as one 47 mm centipedefound in a female frog Of all prey items 958 hadlengths of lt3 mm 829 were lt2 mm and 365 lt1 mm

Our data show no obvious trends for larger frogs to eatlarger prey (Fig 1a) Although females consumed largerprey this was apparently not due to their larger sizesHowever both sexes showed an increase in NPI con-sumed as SVL increased (Fig 1b) APL and NPI werenegatively correlated both when considering males andfemales together and for female data alone (Plt005) indi-cating that individuals that consumed larger prey had onaverage consumed fewer prey items (Fig 1c)

All prey categories were found in the stomachs of bothmales and females with the exception of wingedFormicidae and Lepidoptera larvae which were found ex-clusively in females The percent occurrence of preycategories as well as percentage and volume of preyitems per category (Spearman rank correlation Plt001Fig 2) was highly correlated between males and femalesFemales ate a larger number of soft-bodied Diptera 1while males consumed larger quantities of mites than fe-males Among males mites were the most frequent preyitems (341) followed by flies (244) collembolans(130) and ants (113) In females flies were the mostfrequent prey item (361) followed by mites (181)ants (153) and collembolans (147) However in bothmales and females ants made up the highest proportion

of stomach content volume (263 and 288 respec-tively) followed by flies (224 and 246 respectively)

Prey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availabilityPrey use in relation to prey availability

Arthropod availability and prey selection was based ondata from 11 leaf-litter samples with 1867 individual poten-tial prey items and compared against the same 65 frogsamples (42 females and 23 males) containing a total of5492 prey items in their stomachs All leaf-litter and stom-ach samples contained mites collembolans andnon-winged ants Flies (Diptera 1) were rare in leaf-littersamples but common in stomach samples This discrep-ancy may have been due to a methodological bias as theflies may have evaded capture within the leaf-litter sam-ples Other flies (Diptera 2) winged ants and thysanuranswere likewise absent from leaf-litter samples suggestingeither low numbers or evasion of capture Stomachslacked the prey classes Diptera 3 Blattodea Dermapteraand Nematoda although some large unidentified arthro-pod parts present in some frog stomachs may havebelonged to Blattodea Leaf-litter and stomach contentsamples were correlated in terms of occurrence numbersand volumes of prey categories (Spearman rank correla-tion Plt005) In all three categories frog diet and preyavailability in leaf litter mirrored each other (Figs 3ndash4 Ta-ble 1) The Strauss Food Selection Index (Strauss 1979)indicated low electivity regarding all prey classes for bothnumbers (Fig 5) and volumes of arthropods There may

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 2 Fig 2 Fig 2 Fig 2 Fig 2 Comparison of numerical abundance of different prey items in the diet of male and female Mantellaaurantiaca Values are averages of percentages for individual stomach contents and are correlated between sexes(Spearman rank correlation rs=0842 Plt001) Abbreviations Homoptera (Hom) Hemiptera (Hem) non-Formicidae (NF) winged (W) non-winged (NW)

229

also have been a negative election of mites and a positiveelection of flies (Diptera 1) but the latter may be due tomethodological issues (see above)

Stomachs contained ants from nine generaCrematogaster (1 morphospecies) Cataulacus (1)Anochetus (1) Paratrechina (1) Hypoponera (2)Strumigenys (3) Tetramorium (6) Monomorium (4) andPheidole (7) Frequency of occurrence was highest in onemorphospecies of Pheidole (present in 48 stomach con-tents) and two morphospecies of Monomorium (presentin 45 and 24 stomach contents) All other antmorphospecies were present in ten or fewer stomach con-tents only

DISCUSSIONDISCUSSIONDISCUSSIONDISCUSSIONDISCUSSION

Prey electivity inPrey electivity inPrey electivity inPrey electivity inPrey electivity in Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca

Our data indicate that the prey of Mantella aurantiacaconsists mainly of arthropods of small size different frommost other mantellid species where numerous large preyitems are found (Vences et al 1999b) Although ants werenot the most numerically common prey items (14) theycomprised the bulk of the prey consumed (28) in termsof volume Mites which have recently been shown to be amajor source of dietary alkaloids in poison frogs (Takadaet al 2005 Saporito et al 2007) were present in highnumbers (23) but only made up 5 of the volume of thestomach contents However it is insufficiently under-

stood whether volume or individual numbers of a particu-lar prey is more important to determine alkaloid contentand composition of poison frogs If all individuals of oneprey type contain similar alkaloids in similar concentra-tion and alkaloids are stored in the skins of the frogs forlong periods then volume is probably most important Ifonly some individuals of a certain prey category containalkaloids prey individuals differ in the concentration ofcontained alkaloids andor alkaloids are stored only for alimited time in the skins of the frogs then it is probablymore important that the frogs eat large numbers of thisprey category in order to ensure that at regular intervalssome prey individuals with high alkaloid concentrationare consumed

Ant species of Tetramorium were present in twelvestomach contents (summarized for the fourmorphospecies of this genus) One species of this genushad previously been found to contain two differentpumiliotoxin alkaloids (Clark et al 2005) A species ofParatrechina was present in a single stomach these antsare known to contain a variety of alkaloids and togetherwith other species of formicine ants could potentiallycontribute to the dietary alkaloids of Mantellaaurantiaca Several other arthropods found in the stom-ach contents such as millipedes and beetles couldpotentially provide dietary alkaloids as well (eg Ayerand Browne 1977 Daly et al 2002 Saporito et al 2003Clark et al 2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 3 Fig 3 Fig 3 Fig 3 Fig 3 Comparison of prey consumed with prey availability Columns show the percentage of stomach and littersamples that contained different prey items Data are correlated (Spearman rank correlation rs=0564 Plt001)Abbreviations as in Fig 2

230

Prey size did not appear to be a function of frog bodysize Though larger frogs eat more prey it is of the sameaverage size as prey eaten by smaller frogs This con-trasts with data from other studies which found frogs ofgreater snoutndashvent length consuming fewer but largerprey (eg Berry 1966 Brooks 1982 Hirai 2002Hodgkinson amp Hero 2003 Ramiacuterez-Bautista amp Lemos-Espinal 2004) The fact that large and small individuals ofMantella aurantiaca consumed arthropods of the sameaverage size indicates little preference for different-sizedprey

This is confirmed by the comparisons of prey in theleaf litter versus the stomach contents The electivityanalysis provided little evidence for prey selection In-deed cases where preference was apparent (eg for flies)may in fact be artefacts of our collection method ThusMantella aurantiaca appeared to consume prey in pro-portion to its availability The one distinct exception weremites ndash these were consumed at lower proportions eitherbecause 1) they are difficult to find due to their small sizeand cryptic habits or 2) there is active discriminationagainst such small prey Few large prey items were con-sumed but such large arthropods were apparently alsorare in the leaf litter Hence M aurantiaca followed amicrophagous ndash but not a myrmecophagous ndash foragingmode but this pattern is a reflection of the abundance ofsmall prey in the environment at the study site and in thestudy period MacNally (1983) studied a similar case inwhich two species of Ranidella mainly consumed small

prey below 15 mm in length but with a very low propor-tion of ants reflecting the patterns of availability ofdifferent prey types and prey sizes

Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other MantellaMantellaMantellaMantellaMantella

Previously published results have observed a high per-centage (74) of ants in stomachs of Mantellaharaldmeieri M nigricans M laevigata and M betsileo(Vences amp Kniel 1998) and 67 ants in M baroni Mbernhardi and M madagascariensis (Clark et al 2005)The percentages of ants consumed by M aurantiaca aremuch greater than the proportion of ants consumed bymost other mantellid frogs analysed (generaAglyptodactylus Boophis Gephyromantis Laliostomaand Mantidactylus) usually because many of these spe-cies consume much larger prey (Vences et al 1999b) Atfirst glance a comparison of our results with these pub-lished data seems to indicate an important differencebetween M aurantiaca and other Mantella specieswhereas congeneric species appear to be ant specialistsM aurantiaca does not display such a preference

However previous studies did not include data onfood availability and therefore it is not yet possible todetermine whether the dietary pattern of Mantellaaurantiaca is common among other species in this ge-nus Our data do not support the idea of Mantellaaurantiaca as a specialist although this species may dis-play different feeding patterns with clear preference forants in other populations or in other seasons However it

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 4 Fig 4 Fig 4 Fig 4 Fig 4 Comparison of prey consumed with prey availability Columns show the percentage numerical abundance ofprey items in stomachs and leaf litter Data are correlated (Spearman rank correlation rs=0613 Plt001)Abbreviations as in Fig 2

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

229

also have been a negative election of mites and a positiveelection of flies (Diptera 1) but the latter may be due tomethodological issues (see above)

Stomachs contained ants from nine generaCrematogaster (1 morphospecies) Cataulacus (1)Anochetus (1) Paratrechina (1) Hypoponera (2)Strumigenys (3) Tetramorium (6) Monomorium (4) andPheidole (7) Frequency of occurrence was highest in onemorphospecies of Pheidole (present in 48 stomach con-tents) and two morphospecies of Monomorium (presentin 45 and 24 stomach contents) All other antmorphospecies were present in ten or fewer stomach con-tents only

DISCUSSIONDISCUSSIONDISCUSSIONDISCUSSIONDISCUSSION

Prey electivity inPrey electivity inPrey electivity inPrey electivity inPrey electivity in Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca Mantella aurantiaca

Our data indicate that the prey of Mantella aurantiacaconsists mainly of arthropods of small size different frommost other mantellid species where numerous large preyitems are found (Vences et al 1999b) Although ants werenot the most numerically common prey items (14) theycomprised the bulk of the prey consumed (28) in termsof volume Mites which have recently been shown to be amajor source of dietary alkaloids in poison frogs (Takadaet al 2005 Saporito et al 2007) were present in highnumbers (23) but only made up 5 of the volume of thestomach contents However it is insufficiently under-

stood whether volume or individual numbers of a particu-lar prey is more important to determine alkaloid contentand composition of poison frogs If all individuals of oneprey type contain similar alkaloids in similar concentra-tion and alkaloids are stored in the skins of the frogs forlong periods then volume is probably most important Ifonly some individuals of a certain prey category containalkaloids prey individuals differ in the concentration ofcontained alkaloids andor alkaloids are stored only for alimited time in the skins of the frogs then it is probablymore important that the frogs eat large numbers of thisprey category in order to ensure that at regular intervalssome prey individuals with high alkaloid concentrationare consumed

Ant species of Tetramorium were present in twelvestomach contents (summarized for the fourmorphospecies of this genus) One species of this genushad previously been found to contain two differentpumiliotoxin alkaloids (Clark et al 2005) A species ofParatrechina was present in a single stomach these antsare known to contain a variety of alkaloids and togetherwith other species of formicine ants could potentiallycontribute to the dietary alkaloids of Mantellaaurantiaca Several other arthropods found in the stom-ach contents such as millipedes and beetles couldpotentially provide dietary alkaloids as well (eg Ayerand Browne 1977 Daly et al 2002 Saporito et al 2003Clark et al 2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 3 Fig 3 Fig 3 Fig 3 Fig 3 Comparison of prey consumed with prey availability Columns show the percentage of stomach and littersamples that contained different prey items Data are correlated (Spearman rank correlation rs=0564 Plt001)Abbreviations as in Fig 2

230

Prey size did not appear to be a function of frog bodysize Though larger frogs eat more prey it is of the sameaverage size as prey eaten by smaller frogs This con-trasts with data from other studies which found frogs ofgreater snoutndashvent length consuming fewer but largerprey (eg Berry 1966 Brooks 1982 Hirai 2002Hodgkinson amp Hero 2003 Ramiacuterez-Bautista amp Lemos-Espinal 2004) The fact that large and small individuals ofMantella aurantiaca consumed arthropods of the sameaverage size indicates little preference for different-sizedprey

This is confirmed by the comparisons of prey in theleaf litter versus the stomach contents The electivityanalysis provided little evidence for prey selection In-deed cases where preference was apparent (eg for flies)may in fact be artefacts of our collection method ThusMantella aurantiaca appeared to consume prey in pro-portion to its availability The one distinct exception weremites ndash these were consumed at lower proportions eitherbecause 1) they are difficult to find due to their small sizeand cryptic habits or 2) there is active discriminationagainst such small prey Few large prey items were con-sumed but such large arthropods were apparently alsorare in the leaf litter Hence M aurantiaca followed amicrophagous ndash but not a myrmecophagous ndash foragingmode but this pattern is a reflection of the abundance ofsmall prey in the environment at the study site and in thestudy period MacNally (1983) studied a similar case inwhich two species of Ranidella mainly consumed small

prey below 15 mm in length but with a very low propor-tion of ants reflecting the patterns of availability ofdifferent prey types and prey sizes

Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other MantellaMantellaMantellaMantellaMantella

Previously published results have observed a high per-centage (74) of ants in stomachs of Mantellaharaldmeieri M nigricans M laevigata and M betsileo(Vences amp Kniel 1998) and 67 ants in M baroni Mbernhardi and M madagascariensis (Clark et al 2005)The percentages of ants consumed by M aurantiaca aremuch greater than the proportion of ants consumed bymost other mantellid frogs analysed (generaAglyptodactylus Boophis Gephyromantis Laliostomaand Mantidactylus) usually because many of these spe-cies consume much larger prey (Vences et al 1999b) Atfirst glance a comparison of our results with these pub-lished data seems to indicate an important differencebetween M aurantiaca and other Mantella specieswhereas congeneric species appear to be ant specialistsM aurantiaca does not display such a preference

However previous studies did not include data onfood availability and therefore it is not yet possible todetermine whether the dietary pattern of Mantellaaurantiaca is common among other species in this ge-nus Our data do not support the idea of Mantellaaurantiaca as a specialist although this species may dis-play different feeding patterns with clear preference forants in other populations or in other seasons However it

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 4 Fig 4 Fig 4 Fig 4 Fig 4 Comparison of prey consumed with prey availability Columns show the percentage numerical abundance ofprey items in stomachs and leaf litter Data are correlated (Spearman rank correlation rs=0613 Plt001)Abbreviations as in Fig 2

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

230

Prey size did not appear to be a function of frog bodysize Though larger frogs eat more prey it is of the sameaverage size as prey eaten by smaller frogs This con-trasts with data from other studies which found frogs ofgreater snoutndashvent length consuming fewer but largerprey (eg Berry 1966 Brooks 1982 Hirai 2002Hodgkinson amp Hero 2003 Ramiacuterez-Bautista amp Lemos-Espinal 2004) The fact that large and small individuals ofMantella aurantiaca consumed arthropods of the sameaverage size indicates little preference for different-sizedprey

This is confirmed by the comparisons of prey in theleaf litter versus the stomach contents The electivityanalysis provided little evidence for prey selection In-deed cases where preference was apparent (eg for flies)may in fact be artefacts of our collection method ThusMantella aurantiaca appeared to consume prey in pro-portion to its availability The one distinct exception weremites ndash these were consumed at lower proportions eitherbecause 1) they are difficult to find due to their small sizeand cryptic habits or 2) there is active discriminationagainst such small prey Few large prey items were con-sumed but such large arthropods were apparently alsorare in the leaf litter Hence M aurantiaca followed amicrophagous ndash but not a myrmecophagous ndash foragingmode but this pattern is a reflection of the abundance ofsmall prey in the environment at the study site and in thestudy period MacNally (1983) studied a similar case inwhich two species of Ranidella mainly consumed small

prey below 15 mm in length but with a very low propor-tion of ants reflecting the patterns of availability ofdifferent prey types and prey sizes

Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other Prey electivity of other MantellaMantellaMantellaMantellaMantella

Previously published results have observed a high per-centage (74) of ants in stomachs of Mantellaharaldmeieri M nigricans M laevigata and M betsileo(Vences amp Kniel 1998) and 67 ants in M baroni Mbernhardi and M madagascariensis (Clark et al 2005)The percentages of ants consumed by M aurantiaca aremuch greater than the proportion of ants consumed bymost other mantellid frogs analysed (generaAglyptodactylus Boophis Gephyromantis Laliostomaand Mantidactylus) usually because many of these spe-cies consume much larger prey (Vences et al 1999b) Atfirst glance a comparison of our results with these pub-lished data seems to indicate an important differencebetween M aurantiaca and other Mantella specieswhereas congeneric species appear to be ant specialistsM aurantiaca does not display such a preference

However previous studies did not include data onfood availability and therefore it is not yet possible todetermine whether the dietary pattern of Mantellaaurantiaca is common among other species in this ge-nus Our data do not support the idea of Mantellaaurantiaca as a specialist although this species may dis-play different feeding patterns with clear preference forants in other populations or in other seasons However it

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Fig 4 Fig 4 Fig 4 Fig 4 Fig 4 Comparison of prey consumed with prey availability Columns show the percentage numerical abundance ofprey items in stomachs and leaf litter Data are correlated (Spearman rank correlation rs=0613 Plt001)Abbreviations as in Fig 2

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

231

is also possible that the high proportion of ants found insome Mantella species simply reflects a high abundanceof ants in their environment Indeed a study currentlyunderway by T Razafindrabe and co-workers indicatesthat in many Mantella populations regardless of breed-ing season the numerical proportion of ants consumedwas lower than that previously encountered (about 70Vences amp Kniel 1998 Clark et al 2005) although theseresults are based on far fewer individuals per populationthan our study In the study by Vences amp Kniel (1998) nomarked preference for small-sized prey (fruit flies) versuslarger prey (crickets) was found in various species ofMantella while such a preference was more obvious forthree species of dendrobatid frogs of the generaDendrobates and Phyllobates (especially for Dleucomelas) Although these experimental results on cap-tive animals may be biased by the habituation of thesespecimens to particular kinds of prey the results wouldbe in agreement with less feeding specialization inMantella as compared to dendrobatids

Temporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in preyTemporal variation in prey

Most Mantella habitats in Madagascar are characterizedby marked seasonality and the frogs undergo at leastpartial hibernation (during the cool dry austral winter)This may involve a trade-off between the need to accumu-late fat reserves rapidly and the selective uptake ofalkaloid-containing prey for chemical defence This hy-pothesis would predict different strengths of preyelectivity during different periods of frog activity Frogsmay change their diet and possibly also their diet prefer-ences in different time periods if for example preferredprey become seasonally less abundant This scenario hasalready been suggested for other tropical communitiesstudied during whole-year periods (Whitfield ampDonnelly 2006) During our study the population of Maurantiaca was still reproducing but was getting close tothe end of the reproductive season and this may be oneexplanation for the low prey electivity encountered asadults may spend less time searching for prey and rely

more on the food they can encounter close to their breed-ing grounds Long-term variations in the alkaloid profilesof populations of dendrobatids have already been ob-served by Daly et al (1987) and recent studies bySaporito et al (2006) demonstrate large seasonal fluctua-tions in alkaloid profiles of populations of Oophaga(previously Dendrobates) pumilio from Panama that mayreflect different food composition or different feedingstrategies at different times of the year Long-term tempo-ral variation in alkaloid profiles has also been observed inMantella baroni (Clark et al 2006) and is hypothesizedto reflect turnover in arthropod communities

Ant specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoidersAnt specialists versus ant avoiders

Based on observations mainly of lizard data Huey ampPianka (1981) remarked that species are either widely for-aging or sit-and-wait predators (but see Perry amp Pianka1997 Perry 1999) Moreover these behaviours correlatewith different types of prey consumed widely foragingspecies concentrate on prey that is sedentary unpredict-ably distributed and clumped such as ants and termitesSimilarly based on large surveys of consumed and avail-able prey for diurnal tropical litter frogs Toft (1980ab1981) distinguished between active foragers whichelected ants and sit-and-wait foragers which avoidedants A weak avoidance of ants was also reported by Hiraiamp Matsui (2000) for Glandirana rugosa which had a highnumerical abundance but a low volumetric proportion ofants in its diet and consumed a lower proportion of antsthan was found in the environment Among nocturnalanurans species of the family Microhylidae may be antand mite specialists (Toft 1981) which is consistent withother studies (eg Berry 1965) and with their specializedtongue morphology and feeding behaviour (Meyers etal 2004)

In dendrobatids Toft (1980a) found that the positionof these frogs along a specialistndashgeneralist continuumwas similar irrespective of the type of calculation appliedelectivity or simple niche breadth calculated from propor-tions of prey categories Consequently Darst et al (2005)

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

Fig 5Fig 5Fig 5Fig 5Fig 5 Strauss Food Selection Index for numerical abundance of prey found in leaf litter and frog stomach samplesAbbreviations as in Fig 2

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

232

used values of niche breadth as proxies for dietary spe-cialization in dendrobatid poison frogs In dendrobatidsmyrmecophagy appears to have evolved multiple timesand in most cases was associated with the recurrent evo-lution of noxiousness and aposematism (Caldwell 1996Summers amp Clough 2001 Darst et al 2005) HoweverDarst et al (2005) also discussed the case of Allobateszaparo in which stomach content analyses from differentpopulations gave conflicting results regarding the pro-portion of ants consumed (74 vs 11 26 and 34)This species is a phylogenetically basal species of

dendrobatid (eg Vences et al 2003 Santos et al 2003)and its variability in feeding strategies may therefore begreater than in the more specialized dendrobatids (eg thegenus Dendrobates sensu lato)

Our results suggest caution if a generalistndashspecialistcontinuum of myrmecophagy or microphagy is to bebased on niche breadth data alone niche breadth may insome poison frog species or populations reflect a true di-etary specialization in terms of positive electivity ofcertain prey categories and in other cases may just reflectfrequencies of prey in the environment Comparing data

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Prey Prey Prey in prey leaf Prey Prey Leaf prey leafnumbers numbers leaf numbers litter volumes volumes litter volumes litter(males) (females) litter (males numbers (males) (females) volumes (total) volume

numbers andfemales)

HEXAPODA

Thysanura 1 4 0 01 00 00 03 00 00 00 Collembola 209 572 309 142 166 154 518 36 79 12 Thysanoptera 6 24 20 05 11 02 10 08 01 03 HomopteraHemiptera adults 10 22 1 06 01 35 84 01 14 00 HomopteraHemiptera immatures 1 7 8 01 04 01 38 16 05 06 Orthoptera 1 3 1 01 01 02 23 04 03 01 Hymenoptera Formicidae 182 593 348 141 187 526 1875 904 282 313 Hymenoptera Formicidae winged 0 13 0 02 00 00 35 00 04 00 Hymenoptera others 19 23 1 08 01 57 41 05 12 02 Diptera 1 393 1402 2 327 01 447 1607 01 241 00 Diptera 2 4 7 0 02 00 06 21 00 03 00 Coleoptera 14 26 12 07 06 12 47 194 07 67 Larvae Lepidoptera 0 6 3 01 02 00 194 151 23 52 Larvae Diptera 2 3 4 01 02 02 03 62 01 21 Larvae Coleoptera 9 21 14 05 08 267 234 608 59 211 Larvae undetermined 1 1 1 00 01 02 124 00 15 00 Other Hexapoda 0 0 5 00 03 00 00 71 00 25

CRUSTACEA

Amphipoda 51 124 25 32 13 160 460 299 73 104 Isopoda 52 69 8 22 04 57 183 68 28 24 Myriapoda Diplopoda 1 2 2 01 01 03 07 06 01 02 Chilopoda 2 10 4 02 02 06 375 47 45 16Arachnida Acari 550 703 1051 228 564 153 308 223 54 77 Araneae 9 31 21 07 11 12 40 161 06 56 Pseudoscorpiones 1 2 12 01 06 00 01 02 00 01

OTHERS

Arthropoda parts undetermined 4 16 0 04 00 27 113 04 16 01 Organic matter 90 196 12 52 06 69 164 15 27 05

Total 1612 3880 1864 100 100 2000 6508 2886 100 100

Table 1 Table 1 Table 1 Table 1 Table 1 Total numbers and percentages and volume (in mm3) and percentage of volumes of the most commonprey categories found in the stomach contents of male and female Mantella aurantiaca and in leaf litter samplesfrom frog habitat Organic matter refers to undetermined remains probably of accidentally consumed leaf litter

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

233

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

from dendrobatids to those of other alkaloid-containingfrogs may also reveal that the generalist versus specialistcontinuum and the active-search-foraging versus sit-and-wait foraging continuum are not always in fullcorrespondence with each other Mantella are clearly ac-tively foraging and diurnal species and like dendrobatidsshow continuous small movements Food capture ismostly by tongue only (versus fully leaping forward) atechnique probably associated with lower metaboliccosts (Toft 1981) Nevertheless Mantella at least onsome occasions do not seem to actively elect small preyor ants and how strongly this trend varies among spe-cies populations and seasons remains to be studied

ACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTSACKNOWLEDGEMENTS

We are grateful to Olivier Sam Mamy and Jean-NoelNdriamiary of the Mitsinjo team in Andasibe for their helpin the field We further thank Edouard Randriamitso andNandi Fatroandriantjafinontjasolokitova for field com-panionship Rainer Dolch for logistical assistance DICEfor funding David Sewell for support and Laura Maryand David Eklund and David and Marvalee Wake forlogistical support in California Ralph A Saporito andseveral anonymous reviewers helped to significantly im-prove previous manuscript drafts The Malagasyauthorities provided authorizations for research andspecimen exports under a cooperation accord betweenthe universities of Amsterdam and Antananarivo and theAssociation National pour la Gestion des AiresProtegeacuteesndashANGAP Financial assistance to M Venceswas provided by the Volkswagen and BIOPAT founda-tions DR Vieites was supported by a University of Vigogrant for research in foreign countries BL Fisher wassupported in part from NSF DEB-0344731

REFERENCESREFERENCESREFERENCESREFERENCESREFERENCESAndreone F Cadle JE Cox N Glaw F Nussbaum

RA Raxworthy CJ Stuart SN Vallan D ampVences M (2005) Species review of amphibianextinction risks in Madagascar conclusions from theGlobal Amphibian Assessment Conservation Biology19 1790ndash1802

Andreone F amp Randrianirina JE (2003) Itrsquos not carnivalfor the harlequin mantella Urgent actions needed toconserve Mantella cowani an endangered frog from thehigh plateau of Madagascar Froglog 59 1ndash2

Ayer WA amp Browne LM (1977) The ladybug alkaloidsincluding synthesis and biosynthesis Heterocycles 7685ndash707

Berry PY (1965) The diet of some Singapore Anura(Amphibia) Proceedings of the Zoological Society ofLondon 144 163ndash174

Berry PY (1966) The food and feeding habits of thetorrent frog Amolops larutensis Journal of Zoology149 204ndash214

Brooks GR Jr (1982) An analysis of prey consumed bythe anuran Leptodactylus fallax from Dominica WestIndies Biotropica 14 301ndash309

Caldwell JP (1996) The evolution of myrmecophagy andits correlates in poison frogs (family Dendrobatidae)

Journal of Zoology 240 75ndash101Chiari Y Andreone F Vences M amp Meyer A (2005)

Genetic variation of an endangered Malagasy frogMantella cowani and its phylogeographic relationship tothe widespread M baroni Conservation Genetics 61041ndash1047

Chiari Y Vences M Vieites DR Rabemananjara FBora P Ramilijaona Ravoahangimalala O amp Meyer A(2004) New evidence for parallel evolution of colourpatterns in Malagasy poison frogs (Mantella)Molecular Ecology 13 3763ndash3774

Clark VC Rakotomalala V Ramilijaona O Abrell L ampFisher BL (2006) Individual variation in alkaloidcontent of poison frogs of Madagascar (MantellaMantellidae) Journal of Chemical Ecology 32 2219ndash2233

Clark VC Raxworthy CJ Rakotomalala V Sierwald Pamp Fisher BL (2005) Convergent evolution of chemicaldefense in poison frogs and arthropod prey betweenMadagascar and the neotropics Proceedings of theNational Academy of Sciences of the USA 102 11617ndash11622

Daly JW (1998) Thirty years of discovering arthropodalkaloids in amphibian skin Journal of Natural Products61 162ndash172

Daly JW Andriamaharavo NR Andriantsiferana M ampMyers CW (1996) Madagascan poison frogs(Mantella) and their skin alkaloids American MuseumNovitates 3177 1ndash34

Daly JW Kaneko T Wilham J Garraffo HM SpandeTF Espinosa A amp Donnelly MA (2002) Bioactivealkaloids of frog skin combinatorial bioprospectingreveals that pumiliotoxins have an arthropod sourceProceedings of the National Academy of Sciences of theUSA 99 13996ndash14001

Daly JW Myers CW amp Whittaker N (1987) Furtherclassification of skin alkaloids from neotropical poisonfrogs (Dendrobatidae) with a general survey of toxicnoxious substances in the Amphibia Toxicon 25 1023ndash1095

Daly JW Secunda SI Garraffo HM Spande TFWisnieski A amp Cover JF Jr (1994) An uptakesystem for dietary alkaloids in poison frogs(Dendrobatidae) Toxicon 32 657ndash663

Darst CR Meneacutendez-Guerrero PA Coloma LA ampCannatella DC (2005) Evolution of dietaryspecialization and chemical defense in poison frogs(Dendrobatidae) a comparative analysis AmericanNaturalist 165 56ndash69

Dubois A (2005) Developmental pathway speciation andsupraspecific taxonomy in amphibians 1 Why are thereso many frog species in Sri Lanka Alytes 22 19ndash37

Duellman WE amp Trueb L (1986) Biology of AmphibiansNew York McGraw-Hill

Frost DR Grant T Faivovich J Bain RH Haas AHaddad CFB de Saacute RO Channing A WilkinsonM Donnellan SC Raxworthy CJ Campbell JABlotto BL Moler P Drewes RC Nussbaum RALynch JD Green DM amp Wheeler WC (2006) Theamphibian tree of life Bulletin of the American Museumof Natural History 297 1ndash370

234

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l

Glaw F amp Vences M (2003) Introduction to amphibiansIn The Natural History of Madagascar 883ndash898Goodman SM amp Benstead JP (eds) Chicago andLondon University of Chicago Press

Griffiths RA (1986) Feeding niche overlap and foodselection in smooth and palmate newts Triturusvulgaris and T helveticus at a pond in mid-WalesJournal of Animal Ecology 55 201ndash214

Griffiths RA amp Mylotte VJ (1987) Microhabitatselection and feeding relations of smooth and wartynewts Triturus vulgaris and T cristatus at an uplandpond in mid-Wales Holarctic Ecology 10 1ndash7

Heying HE (2001a) Mantella laevigata (climbingmantella) aborted predation Herpetological Review 3234

Heying H (2001b) Social and reproductive behaviour in theMadagascan poison frog Mantella laevigata withcomparisons to the dendrobatids Animal Behaviour 61567ndash577

Hirai T (2002) Ontogenetic change in the diet of the pondfrog Rana nigromaculata Ecological Research 17 639ndash644

Hirai T amp Matsui M (2000) Myrmecophagy in ranidfrog Rana rugosa specialization or weak avoidance toant eating Zoological Science 17 459ndash466

Hodgkison S amp Hero JM (2003) Seasonal sexual andontogenetic variations in the diet of lsquodecliningrsquo frogsLitoria nannotis Litoria rheocola and Nyctimistes dayiWildlife Research 30 345ndash354

Huey RB amp Pianka ER (1981) Ecological consequencesof foraging mode Ecology 62 991ndash999

Hyslop EJ (1980) Stomach contents analysis ndash a reviewof methods and their application Journal of Fish Biology17 411ndash429

Legler JM amp Sullivan LJ (1979) The application ofstomach-flushing to lizards and anurans Herpetologica35 107ndash110

MacNally RC (1983) Trophic relationships of twosympatric species of Ranidella (Anura) Herpetologica39 130ndash140

Meyers JJ OrsquoReilly JCO Monroy JA amp NishikawaKC (2004) Mechanisms of tongue protraction inmicrohylid frogs Journal of Experimental Biology 20721ndash31

Nishikawa KC (1999) Neuromuscular control of preycapture in frogs Philosophical Transactions of the RoyalSociety of London 354 941ndash954

Nishikawa KC (2000) Feeding in frogs In Feeding inTetrapod Vertebrates Form Function Phylogeny 117ndash141 Schwenk K (ed) London Academic Press

Opatrnyacute E (1980) Food sampling in live amphibiansVěstniacutek Československeacute Společnosti Zoologickeacute 44 268ndash271

Perry G (1999) The evolution of search modes ecologicalversus phylogenetic perspectives American Naturalist153 98ndash109

Perry G amp Pianka ER (1997) Animal foraging pastpresent and future Trends in Ecology and Evolution 12359ndash364

Rabemananjara F Bora P Cadle JE Andreone FRajeriarison E Talata P Glaw F Vences M amp

Vieites DR (2005) New records distribution andconservation of Mantella bernhardi an endangered frogspecies from south-eastern Madagascar Oryx 39 339ndash342

Rabemananjara FCE Rasoamampionona Raminosoa NRavoahangimalala Ramilijaona O Andreone F BoraP Carpenter AI Glaw F Razafindrabe T VallanD Vieites DR amp Vences M (in press) Malagasypoison frogs in the pet trade a survey of levels ofexploitation of species in the genus Mantella BollettinoRegionale di Scienze Naturali di Torino

Ramiacuterez-Bautista A amp Lemos-Epinal JA (2004) Diets oftwo syntopic populations of frogs Rana vaillanti andRana brownorum from a tropical rain forest in southernVeracruz Meacutexico Southwestern Naturalist 49 316ndash320

Santos JC Coloma LA amp Cannatella DC (2003)Multiple recurring origins of aposematism and dietspecialization in poison frogs Proceedings of theNational Academy of Sciences of the USA 100 12792ndash12797

Saporito RA Donnelly MA Garraffo HM SpandeTF amp Daly JW (2006) Geographic and seasonalvariation in alkaloid-based chemical defenses ofDendrobates pumilio from Bocas del Toro PanamaJournal of Chemical Ecology 32 795ndash814

Saporito RA Donnelly MA Hoffman RL GarraffoHM amp Daly JW (2003) A siphonotid millipede(Rhinotus) as the source of spiropyrrolizidine oximes ofdendrobatid frogs Journal of Chemical Ecology 292781ndash2786

Saporito RA Donnelly MA Norton RA GarraffoHM Spande TF amp Daly JW (2007) Oribatid mitesas a major dietary source for alkaloids in poison frogsProceedings of the National Academy of Sciences of theUSA 104 8885ndash8890

Saporito RA Garraffo HM Donnelly MA EdwardsAL Longino JT amp Daly JW (2004) Formicine antsan arthropod source for the pumiliotoxin alkaloids ofdendrobatid poison frogs Proceedings of the NationalAcademy of Sciences of the USA 101 8045ndash8050

Schaefer H-C Vences M amp Veith M (2002) Molecularphylogeny of Malagasy poison frogs genus Mantella(Anura Mantellidae) homoplastic evolution of colourpattern in aposematic amphibians Organism Diversityand Evolution 2 97ndash105

Simon MP amp Toft CA (1991) Diet specialization insmall vertebrates mite-eating in frogs Oikos 61 263ndash278

Streelman JT amp Danley PD (2003) The stages ofvertebrate evolutionary radiation Trends in Ecology andEvolution 18 126ndash131

Strauss RE (1979) Reliability estimates for Ivlevrsquoselectivity index the forage ratio and a proposed linearindex of food selection Transactions of the AmericanFisheries Society 108 344ndash352

Summers K amp Clough ME (2001) The evolution ofcoloration and toxicity in the poison frog family(Dendrobatidae) Proceedings of the National Academyof Sciences of the USA 98 6227ndash6232

Takada W Sakata T Shimano S Enami Y Mori NNishida R amp Kuwahara Y (2005) Scheloribatid mites

235

Poison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecologyPoison frog feeding ecology

as the source of pumiliotoxins in dendrobatid frogsJournal of Chemical Ecology 31 2403ndash2415

Toft CA (1980a) Feeding ecology of thirteen syntopicspecies of anurans in a seasonal tropical environmentOecologia 45 131ndash141

Toft CA (1980b) Seasonal variation in populations ofPanamanian litter frogs and their prey a comparison ofwetter and drier sites Oecologia 47 34ndash38

Toft CA (1981) Feeding ecology of Panamanian litteranurans patterns in diet and foraging mode Journal ofHerpetology 15 139ndash144

Toft CA (1985) Resource partitioning in amphibians andreptiles Copeia 1985 1ndash21

Toft CA (1995) Evolution of diet specialization inpoison-dart frogs (Dendrobatidae) Herpetologica 51202ndash216

Vences M Chiari Y Raharivololoniaina L amp Meyer A(2004) High mitochondrial diversity within and amongpopulations of Malagasy poison frogs MolecularPhylogenetics and Evolution 30 295ndash307

Vences M Glaw F amp Boumlhme W (1998) Evolutionarycorrelates of microphagy in alkaloid-containing frogs(Amphibia Anura) Zoologischer Anzeiger 236 217ndash230

Vences M Glaw F amp Boumlhme W (1999a) A review of thegenus Mantella (Anura Ranidae Mantellinae)taxonomy distribution and conservation of Malagasypoison frogs Alytes 17 3ndash72

Vences M Glaw F amp Zapp C (1999b) Stomach contentanalyses in Malagasy frogs of the genera TomopternaAglyptodactylus Boophis and Mantidactylus (AnuraRanidae) Herpetozoa 11 109ndash116

Vences M amp Kniel C (1998) Mikrophage undmyrmecophage Ernaumlhrungsspezialisierung beimadagassischen Giftfroumlschen der Gattung MantellaSalamandra 34 245ndash254

Vences M Kosuch J Boistel R Haddad CFB LaMarca E Loumltters S amp Veith M (2003) Convergentevolution of aposematic coloration in neotropical poisonfrogs a molecular phylogenetic perspective OrganismsDiversity and Evolution 3 215ndash226

Vieites DR Chiari Y Vences M Andreone FRabemananjara F Bora P Nieto-Romaacuten S amp MeyerA (2006) Mitochondrial evidence for distinctphylogeographic units in the endangered Malagasypoison frog Mantella bernhardi Molecular Ecology 151617ndash1625

Vieites DR Rabemananjara FEC Bora PRazafimahatratra B Ramilijaona Ravoahangimalala Oamp Vences M (2005) Distribution and populationdensity of the black-eared Malagasy poison frogMantella milotympanum Staniszewski 1996 (AmphibiaMantellidae) In African Biodiversity MoleculesOrganisms Ecosystems 197ndash204 Huber BA andLampe KH (eds) Proceedings of the 5th InternationalSymposium on Tropical Biology Museum Koenig BonnSpringer Verlag

Wake MH (1982) Diversity within a framework ofconstraints amphibian reproductive modes InEnvironmental Adaptation and Evolution 87ndash106Mossakowski D amp Roth G (eds) Stuttgart amp NewYork Gustav Fischer Verlag

Whitfield SM amp Donnelly MA (2006) Ontogenetic andseasonal variation in the diets of a Costa Rican leaf-litterherpetofauna Journal of Tropical Ecology 22 409ndash417

Zimmermann H (1996) Der Schutz des Regenwaldes undein kleines Froumlschchen in Ost-Madagaskar Stapfia 47189ndash218

Accepted 3 October 2007

236

APPENDIXAPPENDIXAPPENDIXAPPENDIXAPPENDIX

Formulae used to calculate volumes of prey items Abbreviations as follows height (H) widthFormulae used to calculate volumes of prey items Abbreviations as follows height (H) widthFormulae used to calculate volumes of prey items Abbreviations as follows height (H) widthFormulae used to calculate volumes of prey items Abbreviations as follows height (H) widthFormulae used to calculate volumes of prey items Abbreviations as follows height (H) width(W) length (L)(W) length (L)(W) length (L)(W) length (L)(W) length (L)

Formula Prey categories

V=pr2H CollembolaColeoptera larvaeDiplopodaDiptera 1 2 and 3Diptera larvaeFormicidae (winged)Hymenoptera (non-winged)Insect larvaeLepidoptera larvae

V=01LW Organic matter (vegetation)

V=(23)pr2H Formicidae (non-winged)

V=(13) pr2H Acari

V=pr2H(12) ChilopodaThysanoptera

V=(12)(13) pr2H Thysanura

V=p(43)times(05L)times(05H)times(025H)times(15) AmphipodaBlattodeaColeopteraDermapteraIsopodaPseudoscorpionesUnidentified insects

V=p(43)times(05L)times(05W)times(025W) HomopteraHemiptera adults and immatures

V=p(43)(05L)2times(05W) Organic matter (seed-like 2)

V=p(43)(05L)times(05W)2 AraneaeArthropod partsOrganic matter (dirt)Organic matter (seed-like 1)Orthoptera

C Woodhead C Woodhead C Woodhead C Woodhead C Woodhead et a l et a l et a l et a l et a l