Received 14March 2003 Accepted 9 July 2003 Publishedonline 2October 2003

Multipleoverseas dispersal in amphibians Miguel Vences 1* ,David R.Vieites 1,2, Frank Glaw3,HennerBrinkmann 1, Joachim Kosuch 4,Michael Veith 4 and AxelMeyer 1 1Departmentof Biology, University ofKonstanz, 78457 Konstanz, Germany 2Laboratoriode Anatom ´õ aAnimal,Departamento de Ecolox ´õ a e Biolox´õ aAnimal,Facultade de Ciencias Biolo´xicas, Universidade de Vigo,Buzo ´n137,36201 Vigo, Spain 3ZoologischeStaatssammlung, Mu ¨nchhausenstrasse21, 81247 Mu ¨nchen, Germany 4Institut fu¨rZoologie,Universita ¨tMainz, Saarstrasse21, 55099 Mainz, Germany Amphibians are thought tobe unable to disperse over oceanbarriers becausethey donot tolerate the osmoticstress of salt water.Their distribution patternshave thereforegenerally beenexplained byvicari- ancebiogeography. Here,we present compelling evidencefor overseasdispersal offrogs in theIndian Oceanregion basedon the discovery oftwo endemic species on Mayotte. This island belongsto the Comoroarchipelago, which is entirely volcanic andsurrounded by seadepths of more than 3500 m.This constitutesthe first observation ofendemic amphibians onoceanic islands that didnot have any past physical contactto other land masses.The twospecies of frogs had previously beenthought tobe non- endemicand introduced from Madagascar, butclearly representnew species based on their morphological andgenetic differentiation. They belong tothe genera Mantidactylus and Boophis in thefamily Mantellidae that is otherwiserestricted to Madagascar, andare distinguishedby morphology andmitochondrial and nuclearDNA sequencesfrom mantellid speciesoccurring in Madagascar. This discovery permits usto updateand test molecular clocksfor frogs distributedin this region. The newcalibrations are in agreement with previous rate estimatesand indicate two further Cenozoictransmarine dispersal eventsthat had previously beeninterpreted as vicariance: hyperoliid frogs from Africa toMadagascar ( Heterixalus ) and from Madagascar tothe Seychelles islands ( Tachycnemis ).Our resultsprovide thestrongest evidence so far that overseasdispersal ofamphibians existsand is norare exception,although vicariance certainly retains muchof its importance in explaining amphibian biogeography. Keywords: Amphibia; Mantellidae;Madagascar; Comoros;phylogeny; biogeography

1. INTRODUCTION andPacific regions.Some of theselandmasses were prob- ably belowsea-level in thepast, and their overseascoloniz- Amphibians are akey group in historical biogeography ation following emergencehas been hypothesized (Hedges becausethey are oftenthought tobe unable to disperse et al. 1992; Hedges1999). However,all ofthese islands over saltwater barriers (Duellman &Trueb 1986; Meirte are madeup at least partly by continentalfragments for 1999; Bossuyt& Milinkovitch 2001; Inger &Voris 2001; which past land connectionscannot be excluded (Heaney Brown& Guttman2002). They are well knownto be 1985; Whitmore 1987; Crother &Guyer1996). Geologi- extremely sensitiveto osmotic stress and do not survive in cal data are seldomsufficiently definitive toascertain the salt water,although somespecies of frogs tolerate or par- full submersionof a landmass,and small emerging tially inhabit brackish water (Balinsky 1981). Therefore, remains wouldbe sufficent to harbour relict amphibian amphibians are consideredto be excellent models for vica- populations. riance scenariosas explanation for general biogeographic Adifferentsituation is that ofoceanic islands that never patterns,and major biogeographic hypotheseshave been had physical contactto other landmasses.These are of influencedby theoccurrence of endemic amphibians on completevolcanic origin or werebuilt by coral reefs.They islandsor continents(Duellman &Trueb 1986; Rich- are surroundedby deepwaters that make land connec- ards& Moore1996; Worthy et al. 1999; Bossuyt& tionsthrough fluctuating sealevels impossible, andthey Milinkovitch 2001; Brown& Guttman2002). One are mostly tooyoung toassume vanished connections to important argument for suchinterpretations, ever since drifting continents.No endemicamphibian speciesare Darwin (1859), has beenthat heretoforth noendemic knownfrom truely oceanicislands. Some are populatedby amphibians wereknown from oceanicislands. By contrast, non-endemicfrogs orsalamanders,but these are knownor reptiles are presenton many islandsand some are known assumedto have beenintroduced. Such has beenthought tobe excellent over-water dispersers(Censky et al. 1998; tobe the case for Mayotte,an island belonging tothevol- Arnold2000; Schoener et al. 2001). canicComoro archipelago in theIndian Ocean,located Amphibians are widespreadon many archipelagos, for betweenAfrica andMadagascar. Mayotteis separated instanceon the Philippines andin theSunda, Caribbean from Madagascar by ageographical distanceof 300 km andby seadepths of more than 3600 m,and its origin datesback nofurther than 10–15 Myr ago (Emerick & *Authorand present address for correspondence: Institutefor Biodiversity and Ecosystem Dynamics, Mauritskade 61,1092 AD Amsterdam, The Duncan1982; Nougier et al. 1986). The twofrog species Netherlands ([email protected]). knownfrom Mayotteare seenas conspecific with taxa

Proc.R. Soc.Lond. B (2003) 270, 2435–2442 2435 Ó 2003 TheRoyal Society DOI10.1098/ rspb.2003.2516 2436M. Vencesand others Multiple overseasdispersal in amphibians

from Madagascar andof allochthonousorigin (Blommers- threemitochondrial (12S and 16SrRNA, tRNA Val) gene Schlo¨sser& Blanc 1991; Meirte1999). fragments fromrepresentatives of allmantellid genera, The evolutionand biogeography ofthe highly diverse, subgenera and speciesgroups to resolvethe relationships butstrongly endangeredanimal diversity ofMadagascar within the family;representatives of two other ranoidfam- andother islandsin thewestern Indian Oceanhave been ilies(Rhacophoridae: Polypedates ; Ranidae: Rana) were subjectto intense debates in recentyears (Krause et al. usedas the outgroup. 1997; Murphy &Collier 1997; Jansa et al. 1999; (iii)We chose members of majorclades of ranoidfrogs Bossuyt& Milinkovitch 2001; Farias et al. 2001; (Bossuyt &Milinkovitch2000), including previously Meegaskumbura et al. 2002; Raxworthy et al. 2002). unstudiedAfrican taxa, and of other familiesthat could Hypotheseson thetime ofthe origin ofthesefaunas must beinformative regarding biogeographic relationshipsin the largely rely onphylogenies ofextant taxa becauseno ter- IndianOcean region: to resolvethe relationshipsamong restrial or freshwaterfossils are knownfrom theTertiary these deepclades, we analyseda morecomprehensive period (65–2Myr ago) ofMadagascar (Krause et al. dataset includingtwo nuclear(rhodopsin, tyrosinase) and

1997). Deepvicariance has oftenbeen invoked to explain fourmitochondrial (12S and 16SrRNA, tRNA Val, cyto- theorigin ofMadagascar ’sendemicvertebrates chrome b)genefragments. Asalamanderand representa- (Duellman &Trueb 1986; Richards &Moore1996; tivesof archaicfrogs (familiesDiscoglossidae and Pipidae) Murphy &Collier 1997; Bossuyt& Milinkovitch 2001; wereused as hierarchicaloutgroups. Farias et al. 2001): their ancestorssupposedly evolved in isolation after thebreakup ofthe southern supercontinent Gondwana.During this geological process,Madagascar (b) DNAextraction, amplification andsequencing had beenseparated from other landmassesin theJurassic DNA was extracted fromtissue samples preserved in ethanol andCretaceous (Briggs 2003). Recentphylogenies ofcha- and sequencedon ABI 3100and ABI377automated meleonsand rodents ( Jansa et al. 1999; Raxworthy et al. sequencersafter directamplification using primersfrom pre- 2002), however,proposed area cladograms that are notin viousstudies (Palumbi et al. 1991;Bossuyt &Milinkovitch accordancewith thesuccession of plate tectonicalevents. 2000;Vences et al. 2003)or that weredeveloped for this work Dispersal scenariostherefore seem plausible for these (sequencesin a 5 9–39 directiongiven only for new primers;F, groups buthave notbeen considered for amphibians, forward primers;R, reverseprimers). Cytochrome b (up to which in thewestern Indian Oceanregion are mostly rep- 1016bp): CBJ10933(F); Cytb-a (F);MVZ15L-mod(F) —AAC resentedby frogs.Caecilians occuron the Seychelles and TWATGGCCCMCACMATMCGWAA;Cytb-c (R); continentalAfrica andAsia, whereas salamanders are CytbAR-H-mod (R) —TAW ARGGRTCYTCKACTGGTT completely absent.Except for theenigmatic Seychellean G.Tyrosinase (exon1; 632 bp): Tyr-1b (F);Tyr-1d (F);Tyr- sooglossids,all anuransfrom theSeychelles and Madagas- 1a(F); Tyr-F40 (F) —AARGARTGYTGYCCIGTITGG; car are includedin thesuperfamily Ranoidea,a highly Tyr-Fx3 (F)—ACTGGCCCAYTGTHT TYTACAAC; diversegroup oflargely unsolvedsystematics (Duellman & Tyr-Fx4 (F)—YTGGCCYWYTGTNTTYTAYAAC;Tyr- Trueb 1986; Feller &Hedges1998; Vences& Glaw 1g (R);Tyr-1e (R);Tyr-SPA (R) —GAIGAGAARAARGAI 2001). GCTGGGCT.Rhodopsin (exon1; 334 bp): Rhod-ma (F) — Here,we report onour recent discovery that theCom- AACGGAACAGAAGGYCC;Rhod-1a (F);Rhod-md (R) — oro frogs representpreviously undescribedspecies GTAGCGAAGAARCCTTC;Rhod-1d (R);Rhod-1c(R). endemicto Mayotte. We use mitochondrial andnuclear 12SrRNA and tRNA Val (ca.700bp): 12SA-L (F);12SB-H(R); DNA sequencesto demonstrate the close phylogenetic 16SR3(R). 16SrRNA (5 9 fragment; ca.650bp): 16S-L3(F); relationships ofthese species to the endemic Malagasy 16SA-H(R); 16SrRNA (3 9 fragment; ca.550bp): 16SA-L(F); radiation ofmantellid frogs,thereby providing evidence 16SB-H(R). The moleculardataset was complementedby for their overseasdispersal from Madagascar. Their origin sequencesavailable from GenBank (see http:/ /www4.ncbi. is further usedas a newcalibration point ofa molecular nlm.nih.gov/).The 238new sequencesobtained inthis study clock.We thereby contributeto the elucidation of pro- (12S,55 sequences;16S, 93; Cyt b,19;Rhod, 52;Tyr, 19)have cessesof amphibian overseasdispersal in theIndian Ocean beendeposited in the GenBankdatabase underthe accession region andthe origin ofMadagascar ’senigmatic fauna. numbersAY341580 –AY341817.

(c) Phylogenetic analysis 2. MATERIALAND METHODS Geneswere submitted to separate and combinedanalyses

(a) Taxonand gene sampling with Paup¤ (Swofford 2002)after exclusionof allgapped and To understand(i) the distinctnessof the Comorofrogs; (ii) hypervariable regionsof the rRNA and tRNA genesand the theirphylogenetic relationships among mantellids;and (iii)their third positions of cytochrome b,which areknown to befully relationshipsto other frog groups fromthe IndianOcean region, saturated at the levelof anuran families(Graybeal 1993). we compiledthree datasets that differedin the compositionof Exploratory analyses includingthese hypervariable sitesdid not taxa and ofDNAfragments. resultin relevantdifferences of cladogramtopologies. Maximum likelihood(ML) heuristicsearches with 10random addition (i)We sequenceda fragment ofthe mitochondrial16S rRNA sequencereplicates were carried out underthe treebisection – genefrom all available mantellid species, usually from sev- reconnectionbranch-swapping option, after determiningthe eralindividuals and populations, and assembledsequences substitution modelfor eachdata subset by hierarchicallikeli- fromover 250 individuals and 120species. These data will hood ratio tests as implementedin M odelTest, v. 3.06 bepresented elsewhere. (Posada& Crandall1998). For the among-Mantellidaedataset (ii)We sequenced fragments of onenuclear (rhodopsin) and of the concatenatedrhodopsin, 12SrRNA, 16S rRNA and

Proc.R. Soc.Lond. B (2003) Multiple overseasdispersal in amphibians M.Vencesand others 2437

tRNAVal sequences(total numberof includedbase pairs: (ii)Age estimates and 95%prediction confidence intervals of 1875bp), ageneraltime-reversible (GTR 1 I 1 G)substitution majornodes were calculated by regressionanalysis. For modelwas selected( 2lnL = 26025.7305),with empiricalbase this secondapproach we usedrhodopsin sequencesonly, frequencies(freqA = 0.3551;freqC = 0.2199;freqG = 0.1644; becausefor this gene,data werealso available from three freqT = 0.2607)and substitution rates ([A –C] = 3.7878; frog speciespairs that probably aroseby vicarianceat the [A–G] = 10.5790; [A–T] = 5.7768; [C–G] = 1.1797; [C–T] = endof the Mediterraneansalinity crisis in the Messinian 29.2325; [G–T] = 1),a proportion of invariablesites of 0.4042 (5.3Myr ago): Ranacretensis– R. cerigensis , Alytes dickhil- and agamma distributionshape parameterof 0.6582.For the lenii–A. maurus , Pelobatescultripes– P. varaldii . higher-levelrelationship dataset ofthe concatenatedrhodopsin, tyrosinase,cytochrome b,12SrRNA,16S rRNA, and tRNA Val Weconsistently assumed ancient ages for the calibration sequences(total numberof includedbase pairs:2625 bp), a points (oldage of the root; oldestage of Comoroislands; Tamura-Nei (TrN 1 I 1 G)substitution modelwas selected assumption ofvicariancefor the Mediterraneanfrogs), although (2lnL = 30463.0156),with empiricalbase frequencies morerecent dispersals to Mayotte orover the Mediterranean, (freqA = 0.3371;freqC = 0.2527;freqG = 0.1497;freqT = ora younger age of the root cannot beexcluded. Any possible 0.2606)and substitution rates ([A –G] = 3.3874; [C–T] = bias of the calibrationswill therefore be directed towards an 5.2170;all other rates = 1),a proportion of invariablesites of overestimateof the ages of divergence.This isconservative if 0.2851and agamma distributionshape parameterof 0.6848. young ages of divergenceare to bedemonstrated herein. Non-parametric MLbootstrapping with 100full heuristic searcheswas carriedout for the two sets of taxa. Analysesusing 3. RESULTS 2000replicates under maximum parsimony and neighbour- joiningmethods resultedin identical topologies and similar (a) Discovery andrelationships of Comoro frogs bootstrap values.Bayesian posterior probabilities were calcu- Our intensivesurveys on three Comoro islands latedusing M rBayes,v. 2.01(Huelsenbeck & Ronquist 2001). (Mayotte,Moheli andGrande Comoro) resulted in Inaccordance with the selectionof complexsubstitution models amphibian findingson Mayotteonly. We encountered two by ModelTest (Posada& Crandall1998), we set the MLpara- distinctfrog speciesthat correspondedto those previously metersto correspondto the the GTRmodelwith site-specific recorded as Mantidactylusgranulatus (Blommers- rate variation(some sites invariant and others followinga Schlo¨sser& Blanc 1991; Meirte1999) and Boophistephra- gamma distribution),and proportion ofdifferentbase categories eomystax (Meirte1999). However,the former differed estimatedfrom the data. Weran aminimumof 300000 gener- distinctly in morphology (e.g.smaller size,larger femoral ations, sampling treesevery 10 generations. The initialset of glands,white single versusblackish paired vocal sac)and generationsneeded before convergence on stable likelihood advertisementcalls from M.granulatus, while thelatter values(burnin) was set at 6 –7%based onour own empirical waslarger, had adifferentiris coloration (reddishversus evaluation.Competing phylogenetic hypotheses weretested with golden-brownish)and a more granular dorsal skinthan B. non-parametriclikelihood tests (Shimodaira& Hasegawa 1999). tephraeomystax .Despitethese differences, the species were Fordetails of primers,voucher specimens and phylogenetic clearly assignable tothegenera Mantidactylus and Boophis methods,see electronic Appendices A –E,availableon The ofthe family Mantellidae (Vences& Glaw2001), based RoyalSociety ’sPublicationsWeb site. onfemoral glands in the Mantidactylus males andoverall similarity with B.tephraeomystax from Madagascar. (d) Molecular clock calibrations Both Comorofrogs, which will beformally described The oldestavailable age ofMayotte (8.7Myr ago; Nougier et in aforthcoming paper, werecommon on Mayotte and al. 1986)provides a calibrationof the age of the splitbetween inhabited secondaryhabitats, oftenclose to human settle- the Comoroanand Malagasy sisterspecies. In addition to the ments.A comparison with morphological data from over two Comoro–Madagascar splits,we usedthe divergence 5000 voucherspecimens of all nominal mantellid species betweenAfrican and South Americanrepresentatives of the examined by us,and with mitochondrial haplotypes (16S strictlyfreshwater aquatic familyPipidae at 101Myr ago rRNA) ofmore than 120 describedand undescribed man- (Pitman et al. 1993)as afurther calibrationpoint. The origin tellids,confirmed that thesespecies have neverbeen found of lissamphibianshas beenestimated (Kumar & Hedges1998) onMadagascar. Minimum pairwise sequencedivergence at 360 ± 14Myrago; an ancientordinal divergence of the Lis- ofthe Comoro haplotypes ascompared with Malagasy samphibia istherefore possible. specieswas 5%. Ageestimates were based entirelyon nuclear genes to exclude Phylogenetic relationships ofthe Comoro frogs were possibleinfluences of saturation inthe mitochondrialdataset. revealed by amolecular phylogeny basedon 1875 bp of Weused two separate approaches. onenuclear and three mitochondrial genesfrom 47 rep- resentativesof all subgeneraand species groups ofman- (i)The MLtreeobtained using the completedataset was sub- tellids (figure 1). The Comorofrogs weredeeply nested mittedto non-parametricrate smoothing, which minim- within theMantellidae. The cladogram placedthem assis- izesancestor-descendent local rate changes inthe absence ter speciesof the Malagasy M. wittei and of the B. of rate constancy (Sanderson1997). Branch lengths had doulioti/B.tephraeomystax clades,supported by maximum previouslybeen recalculated based ontyrosinase and rho- bootstrap values andBayesian posterior probabilities dopsinsequences only. The age of the root (salamander – (100%) andby non-parametric likelihood ratio testsof frog split)was fixedat 370Myr ago (ageof first fossil alternative topologies (seeelectronic Appendix F). Acom- tetrapods), but alternativesearches with the root at parison oftheir 16S rRNA haplotypes with thosefrom five 250Myr ago (ageof first frog ancestor: Triadobatrachus ) populations of M. wittei,fourpopulations of B.tephraeo- werealso performed. mystax andfive populations of B.doulioti, sampled over

Proc.R. Soc.Lond. B (2003) 2438M. Vencesand others Multiple overseasdispersal in amphibians

92 Md. albolineatus Vences& Glaw2001), endemicto Madagascar andthe 100 100 Md. liber Comoros,with five genera that previously (Blommers- 100 Md. depressiceps Schlo¨sser& Blanc 1991) had beenassigned to three differ- 87 93 Md. domerguei 100 100100 Md. blommersae entfamilies. The twoSeychellean taxa occupiedvery dif- 100 100 Md. n. sp. (Comoros) ferentpositions on the cladogram. Nesomantis was the 100 100 Md. wittei sistergroup ofthetwo major Neobatrachian lineages,the P 99 100 100 Md. kely 100 87 100 Md. sarotra Hyloidea andRanoidea (B andC in figure 2; Feller & 100 Md. grandisonae Hedges1998), butan alternative positionas basal hyloid Md. madinika 99 couldnot be significantly excludedby likelihood ratio 100 100 Mt. madagascariensis 99 100 Mt. laevigata tests. Tachycnemis wasvery closely related totheMalagasy

100 Md. massorum e 100 a Heterixalus in thefamily Hyperoliidae (Richards &Moore n

100 Md. peraccae i l

l 1996; Vences et al. 2003).

100 Md. redimitus e t

94 100 Md. granulatus n The application ofnon-parametric rate smoothing to 100 Md. sculpturatus a 84 100 M branch lengthsbased on nuclear genes only (figure 2) 100 99 Md. striatus 100 91 100 Md. horridus placedthe divergence between the endemic Malagasy taxa 100 Md. asper (mantellids) andother ranoidsinto thelate Cenozoic.Sev- 61 84 Md. aff. ulcerosus eral relevant splits had muchyounger ages in themid- or 86 100 100 99 Md. biporus LateCenozoic, namely thosebetween Seychellean, Mala- 100 100 100 Md. charlotteae 100 Md. opiparis gasy andAfrican hyperoliids ( Tachycnemis , Heterixalus , 95 100 Md. lugubris Hyperolius),andbetween African andAsian species of the 100 100 Md. sp. aff. lugubris genera Rana and Hoplobatrachus ,andof thefamily Rhaco- 82 100 Md. femoralis

99 100 Md. ambreensis e phoridae (Chiromantis and Polypedates ).Theseresults were a n

Md. grandidieri i corroborated andreceived statistical significance by the

98 A. madagascariensis m o 95% confidenceintervals calculatedusing rhodopsin P t

100 L. labrosum s o 62 B. doulioti i divergencesin aregressionanalysis (seeelectronic Appen- l 100 63 B. n. sp. (Comoros) a

L dix G).The resulting rate estimatesof 0.03 –0.1% rhodop- 96 100 B. tephraeomystax 2 1 21 100 100 sindivergence lineage Myr correspondedwell toa P 100 B. xerophilus

B. idae e further independentcalibration usingthe synapsid/ diapsid a n

B. goudoti i split 310 Myr ago ( Homo/Gallus divergence0.03% 100 h 2 1 2 1 B. boehmei p lineage Myr )(Kumar &Hedges1998). 99 62 100 o

B. occidentalis o 100 100 B. microtympanum B 91 B. luteus 100 B. sibilans 4. DISCUSSION 100 B. rappiodes 99 (a) First evidence for frogoverseas dispersal to 100 100 B. viridis 100 B. vittatus oceanic islands 100 B. marojezensis The Comorosare entirely volcanic andhave neverhad directcontact to any continentallandmass (Nougier et al. Figure 1. Phylogeny of 47 speciesof theendemic Malagasy – 1986). Notcounting the presently submerged Geyser sea- Comoroan frog familyMantellidae. Thetree wasobtained mount,Mayotte is theoldest of these islands. Its origin byML analysisof adatasetof 1875 bpof nuclear and mitochondrial genes. Numbers atnodes are results of ML hasbeen estimated at 7.7 ± 1Myr ago (Nougier et al. bootstrapping (above, 100 replicates) and Bayesian posterior 1986), while analysesbased on the K –Ar methodyielded probabilities (below, 280 000 generations; every tenth evenyounger estimatesfor theoldest shield building vol- generation sampled) in percentages (values below 50% not canism(5.41 ± 0.26 Myr ago; Emerick &Duncan1982). shown). Genera are abbreviated asfollows: Md, The lowesthistorical sea-level in theIndian Ocean(Haq Mantidactylus ; Mt, Mantella, A, Aglyptodactylus ; L, et al. 1987) was 2145 ± 5mat 18 400 yr ago (Colonna et Laliostoma; B, Boophis. Rana temporaria , R.temporalis and al. 1996). By analysing British Admiralty nautical charts Polypedates cruciger were used asoutgroups. Thenewly 2110 and758 after partially digitalizing andgeoreferenc- discovered speciesfrom Mayotte(Comoros) are printed in ing them in ageographic information system,we estimate bold. Thethree lineages specialized to reproduction in that underthese lower sea-levels the Malagasy northwest- stagnant water (pond breeders) are marked witha ‘P’. erncoast and Mayotte were still separatedby alinear minimum distanceof more than 250 km andby ocean their completedistribution areas (Vences& Glaw2002), depthsof more than 3400 mbelowsea level. Although the confirmedthese relationships. The Mantidactylus from Comoroslie onan oceanic ridge, aCenozicland connec- Mayotte,furthermore, differed very distinctly from M. tion toMadagascar seemsout of the question (Krause et wittei by advertisementcalls, andby awidearray ofmor- al. 1997; Raxworthy et al. 2002). phological characters (broader head,longer handsand Manynew species of frogs have beendescribed from feet,more strongly enlarged terminal finger disks). Madagascar in recentyears (Vences& Glaw2001), but mostare morphologically closeto known species and (b) Higher-level relationshipsand age of Indian many werealready presentin historical collections.Fur- Ocean anurans thermore, thesenew taxa are usually discoveredin primary Aphylogeny reconstructedusing 2625 bp oftwo mid-altitude rainforests.The Comorofrogs inhabit sec- nuclearand four mitochondrial genes(figure 2) corrobor- ondary lowlandhabitats that have beenexhaustively sur- atedthe mantellid clade(Bossuyt & Milinkovitch 2000; veyedin Madagascar, andthe Mantidactylus from Mayotte

Proc.R. Soc.Lond. B (2003) Multiple overseasdispersal in amphibians M.Vencesand others 2439

100/100 Boophis tephraeomystax e 63/98 Boophis n. sp. 1 a d i

96/100 l Aglyptodactylus madagascariensis l e

99/100 t

Laliostoma labrosum n a M 100/100 Mantidactylus n. sp. e 2 a d

Mantidactylus wittei i r

100/100 Chiromantis xerampelina o h

5 p

Polypedates cruciger o c a

Nyctibatrachus major h R 100/100 Rana galamensis 6

Rana temporalis e a d 100/100 Ptychadena mascareniensis i n

E Petropedetes parkeri a Indirana cf. leptodactyla R Hoplobatrachus crassus 100/100 7 100/100 C Hoplobatrachus occipitalis 93/100 Astylosternus diadematus Astylosternidae e

97/100 Leptopelis natalensis a d i i

Arthroleptis variabilis Arthroleptidae l

64/98 100/100 o A D 91/100 r Tachycnemis seychellensis e p

100/100 Heterixalus tricolor 4 3 y H e

100/100 Hyperolius viridiflavus a d e i a Bufo melanostictus e n

100/100 d a o i

B f d l i u

Leptodactylus fuscus y s 73/84 t s B c o

Nesomantis thomasseti a l d g e o

Pipa parva o 67/99 a t o d p i 100/100 Hymenochirus boettgeri S e p i L P Xenopus laevis e a d

Alytes muletensis i s s o l g o c s 250 200 150 100 50 0 Myr ago i D

Figure 2. Phylogeny and divergence times of major clades of frogs in thewestern Indian Ocean region, basedon MLanalyses of 2625 bpof nuclear and mitochondrial genes. Salamander sequences were used asthe outgroup. Aphylogram withbranch lengths basedon nuclear gene divergences only wassubmitted to non-parametric rate smoothing. Thenumbers atnodes are results of MLbootstrapping (left, 100 replicates) and Bayesian posterior probabilities (right, 470 000 generations, every tenth generation sampled) in percentages (values below 50% not shown). Themauve horizontal barsshow mean and 95% confidence intervals of ageestimates of young dispersal events byregression analysisof pairwise divergences of rhodopsin sequences; mauvecircles indicate calibrations. Thecoloured vertical barsencompass theperiod of separation of Madagascar – Greater India from Africa (Rabinowitz et al. 1983), of South America from Africa (Pitman et al. 1993) and of Greater India from Madagascar(Storey et al. 1995; but see Briggs 2003). Major clades are coded bycapital letters: A,Neobatrachia; B, Hyloidea; C,Ranoidea; D,Arthroleptoidei; E,Ranidei. Distribution of taxaon continents isindicated bysymbols (see inset map). Cenozoic dispersal events are coded bynumbers (blackfill, oceanic dispersal): 1, 2, mantellines from Madagascarto Mayotte(Comoros); 3,hyperoliids from Africa toMadagascar;4, hyperoliids from Madagascarto theSeychelles; 5, rhacophorids from Asiato Africa; 6, 7, ranids from Asiato Africa. is morphologically distinctive.It istherefore very unlikely distribution patternsof extant biotas (Whitmore 1987; that thesespecies also occuron Madagascar buthave been Moss& Wilson1998). The Philippines wereprobably sofar overlooked.Overseas dispersal remains theonly connectedto the Asian mainland through land bridges in conceivable explanation for thepresence of mantellid frogs thePleistocene (Heaney 1985; Voris 2000). Compelling onMayotte, and our phylogenetic resultsdemonstrate evidencethat Sundaand Philippine amphibians diddis- that twoindependent dispersal eventsfrom Madagascar perseamong islandshas beenpublished, but amphibian took place. biogeographers working in theseregions usually avoided Several other examples ofputative overseasdispersal in explicit statementson overseas dispersal (e.g.Inger & amphibians have beenmentioned in theliterature. The Voris 2001; Brown& Guttman2002). In fact,the com- Philippines harbour alargely endemicfrog faunabut were plex geological history ofthese regions makesit difficult probably entirely submergedduring theOligocene, leaving toidentify unequivocaltransmarine dispersal events.Vica- subsequentdispersal asmost probable origin ofthese riance explanations are also still being putforward to amphibians (Heaney1985; Brown& Lomolino1998). explain theorigin ofCaribbean (Crother &Guyer1996) The sameapplies toparts ofSulawesiand the Moluccans andPacific amphibians (Worthy et al. 1999). (Heaney1985), andendemic frogs are also knownfrom All examples ofamphibians onoceanic islands of fully Fiji (Worthy et al. 1999). Jamaica, mostof Cuba and volcanic origin sofar refer tonon-endemicspecies of pre- PuertoRico were covered by theocean as well (Buskirk sumedor demonstratedorigin by human translocation: 1985), yet have endemicfrog faunas(Hedges 1999). Ptychadenamascareniensis and Bufogutturalis onMauritius Molecular clockdata indicatethat mostCaribbean islands andReunion; Hylameridionalis and Rana perezi on the werecolonized by frogs subsequentto their final isolation Canary archipelago; Trituruscarnifex on Madeira; Eleuther- by thesea (Hedges et al. 1992). However,all ofthese land- odactylus,Dendrobates and Bufo specieson Hawaii; Scinax massesare fully orpartly ofcontinentalorigin. The Sunda onGalapagos (Staub1993; Pleguezuelos1997; Kraus et region has beeninfluenced by drifting continentaland al. 1999; Snell &Rea 1999). Consequently,our discovery oceanicmaterial, especially micro-continentalblocks, ofendemicComoro frogs is thefirst instanceof endemic whichbear thepotential ofhaving shapedmuch of the amphibian speciesoccurring onfully volcanic andoceanic

Proc.R. Soc.Lond. B (2003) 2440M. Vencesand others Multiple overseasdispersal in amphibians

Seychelleanhyperoliids(Blommers -Schlo¨sser& Blanc 1991). The toleranceof xeric environmentsmay bea key adap- IA Rana (4–26 Myr ago) S tation enabling anuransto cross ocean barriers, possibly A E S m L resting onleafs of rafting trees.Ocean currents at present o Hoplobatrachus (4–12 Myr ago) L r E f H favour arafting from northwesternMadagascar toMay- Chiromantis (33–51 Myr ago) C otte,whereas in theEarly Tertiary they may have tempor-

Y

E arily favouredrafting from Africa toMadagascar (Krause S et al. 1997). O S The early radiation ofthe Ranidei clade(Vences & O R Mantidactylus M Glaw2001), which containsthe families Mantellidae O (<8.7 Myr ago) C (Madagascar –Comoros),Rhacophoridae (mainly Asia) A andthe paraphyletic Ranidae (figure 2), has beenrelated I C F R totheir arrival in Asia onthe drifting Indian continent A Boophis Tachycnemis (Bossuyt& Milinkovitch 2001), andthis fitsour molecular (<8.7 Myr ago) (11–21 Myr ago) age estimates.An alternative scenariolinked themajor split betweenranoid andhyloid Neobatrachians tothe

R

A separation ofAfrica andSouth America (Feller &Hedges C 1998), consideringthe largely Old World –NewWorld dis- Heterixalus S (19–30 Myr ago) A junctionin thediversity centresof both groups.This G A hypothesisimplies that theRanoidea had notyet radiated D A at thetime ofseparation ofthe Madagascar –India conti- M nentfrom Africa. The indication (figure 2) that theSeych- ellean sooglossids(genus Nesomantis)are deeplydivergent from other Neobatrachians provides somesupport for this hypothesis.However, because our sampling ofthe Hylo- idea is limited (twofamilies, Bufonidaeand Leptodactyli- Figure 3. Schematicrepresentation of amphibiandispersal in daeonly), the divergence estimates associated with this thewestern Indian Ocean region, and between Asiaand Africa. Age estimateranges of colonization events are based enormousradiation, which containsover 2000 speciesin on combined evidence of regression confidence intervals and 10 families, are tentative only. non-parametric rate smoothing results, except for thetwo Our resultsdo not rule outthe possibility that vicariance Comoro colonizations thatwere used ascalibrations. The hasplayed animportant role in shaping currentamphibian three Asia–Africa dispersal events (grey arrows) mayhave distributions.In fact,there is little doubtthat amphibians benefitted from land connections, while transmarine belong tothe lower endof the relative dispersal ability dispersals must beassumed between Africa, Madagascar, spectrum(Inger &Voris 2001; Brown& Guttman2002), Comoros and theSeychelles. asalso indicatedby their absencefrom mostoceanic islands.However, strict assumptionsthat their distribution islands,and thereby themost reliable evidencefor over- hasexclusively oralmost exclusively beenshaped by vicar- seasdispersal in this vertebrate class. ianceand terrestrial dispersal (e.g.Duellman &Trueb 1986; Bossuyt& Milinkovitch 2001) are notwarranted (b) Gondwananversus post-Gondwananorigin of according toour results. ranoidfrogs Our molecular clockdatings suggestat least twofurther (c) Thevertebrate colonization ofMadagascar instancesof transmarine frog dispersal in thewestern Accordingto our results, hyperoliids have colonized Indian Ocean,namely ofhyperoliid frogs from Africa to Madagascar by overseasdispersal subsequentto its separ- Madagascar ( Heterixalus ),andfrom Madagascar tothe ation from theAfrican mainland. In theabsence of a con- Seychelles( Tachycnemis ).Geneticdivergences between vincing phylogenetic resolutionamong ranids (cladeE in thesegenera are onthe same order ofmagnitude asthose figure 2), similar origins cannot yet beexcluded for other betweenComoroan and Malagasy mantellids, andnot Malagasy frog lineages,such as the Mantellidae. This is reconciliable with aGondwananvicariance. Thesedisper- in agreement with thefact that Madagascar almost exclus- sal eventsare estimatedto have takenplace in theOligo- ively harbours relatively modernlineages ofnon-marine ceneand Miocene. In addition,at least threedispersals vertebrates (Vences et al. 2001). Only fourto five ofthese from Asia toAfrica, discussedbut not dated by Kosuch have afossilrecord dating back intothe Mesozoic (see et al. (2001), also tookplace in theTertiary according electronicAppendix H),andthe available paleontological toour results (figure 3): ofrhacophorid treefrogs data from theMalagasy LatestCretaceous suggest a biotic (Chiromantis )in theEocene, and of Rana and Hoploba- change in deeptime (Krause et al. 1997, 1999). Wesur- trachus in theMiocene. However, the latter twotaxa prob- veyedand re-assessed literature data of20 vertebrate ably crossedthe Arabian peninsulaland connection, cladeswith representativesin Madagascar andof reliably whereasthe ancestors of Chiromantis may have usedland knownphylogenetic relationships (seeelectronic Appen- bridges in theTethys sea(Kosuch et al. 2001). dix H).Groupsoriginating by Gondwananvicariance Within theMantellidae, the Comoro frogs are deeply wouldbe expected to show biogeographic affinities to nestedin lineages ofarboreal or semi-arboreal pond India, whichwas last connectedto Madagascar breeders(figure 1), which are adaptedto living in unfor- (Rabinowitz et al. 1983; Pitman et al. 1993; Storey et al. estedareas. The sameis truefor theMalagasy and 1995; Briggs 2003). However,the closest relatives ofthe

Proc.R. Soc.Lond. B (2003) Multiple overseasdispersal in amphibians M.Vencesand others 2441

Malagasy taxa occurin Africa in 12 casesand in India – Emerick, C.M.&Duncan, R.A.1982 Age progressive volcan- Asia in only threecases. The area cladograms ofonly three ismin theComores Archipelago, western Indian Ocean and outof 20 groups coincidewith thesuccession of plate tec- implications for Somali platetectonics. Earth Planetary Sci. tonicevents. The Malagasy faunatherefore contains an Lett. 60, 415–428. important componentthat originated by dispersal from Farias, I.P., Orti, G., Sampaio, I., Schneider, H.&Meyer, A. 2001 Thecytochrome b gene asa phylogenetic marker: the Africa, possibly usingcurrently submergedislands as step- limits of resolution for analyzing relationships among cichlid ping stones(Krause 1997; McCall 1997). Multiple et al. fishes. J.Mol.Evol. 53, 89–103. additional dispersal eventsfrom Madagascar totheCom- Feller, A.E.&Hedges, S.B.1998 Molecular evidence for the orosand Seychelles, and partly back toAfrica, sub- early history of living amphibians. Mol.Phylogenet. Evol. 9, sequentlyobscured this original biogeographic signal 509–516. (Jansa et al.1999; Raxworthy et al.2002). This contrib- Graybeal, A.1993 Thephylogenetic utility of cytochrome b: utedto the current diversity in origins among terrestrial lessons from bufonid frogs. Mol.Phylogenet. Evol. 2, 256– andfreshwater vertebrates in theIndian Oceanregion. 269. Haq, B.U., Hardenbol, J.&Vail, P.R.1987 Chronology of For contributing crucial samples,primer information or help fluctuating sealevels since theTriassic. Science 235, 1156– during fieldwork, weare indebted to MagalieDelorme, 1167. Annemarie Ohler, Liliane Raharivololoniaina, Kathrin Heaney, L.R.1985 Zoogeographic evidence for Middle and Schmidt and Henk Wallays. WethankErnst Mayrfor valuable LatePleistocene land bridges tothePhilippine Islands. Mod. discussions. Thecomments of two anonymous referees con- Quat. Res.Southeast Asia 9, 127–143. siderably helped in improving an earlier draft of thismanu- Hedges, S.B.1999 Distribution patterns of amphibiansin the script. Thiswork wassupported bygrants of theDeutsche WestIndies. In Regional patterns ofamphibian distribution: a Forschungsgemeinschaft (GL314/1-1, VE247/1-2) and the global perspective (ed. W.E.Duellman), pp. 211 –254. Balti- Deutscher Akademischer Austauschdienst. We are grateful to more, MD:John Hopkins University Press. theMalagasy and Comoroan authorities for research and Hedges, S.B., Hass, C.A.&Maxson, L.R.1992 Caribbean export permits. biogeography: molecular evidence for dispersal in West Indian terrestrial vertebrates. Proc.Natl Acad.Sci. USA 89, REFERENCES 1909–1913. Huelsenbeck, J.P.&Ronquist, F.2001 M rBayes: Bayesian Arnold, E.N.2000 Using fossils and phylogenies to under- inference of phylogenetic trees. Bioinformatics 17, 754–755. stand evolution of reptile communities on islands. In Isolated Inger, R.F.&Voris, H.K.2001 Thebiogeographical relations vertebrate communities in the tropics (ed. G.Rheinwald), pp. of thefrogs and snakesof Sundaland. J.Biogeogr. 28, 309–324. Bonn: Museum A.Koenig. ( Bonn.Zool. Monogr. 863–891. 46.) Jansa, S.A., Goodman, S.M.&Tucker, P.K.1999 Molecular Balinsky, J.B.1981 Adaptation of nitrogen metabolism tohyp- phylogeny and biogeography of thenative rodents of Mada- erosmotic environment in amphibia. J.Exp. Zool. 215, gascar(Muridae: Nesomyinae): atestof thesingle-origin 335–350. hypothesis. Cladistics 15, 253–270. Blommers-Schlo¨sser, R.M.A.&Blanc, C.P.1991 Amphibi- Kosuch, J., Vences, M., Dubois, A., Ohler, A.&Bo¨hme, W. ens (premie`re partie). 75, 1–379. Faune de Madagascar 2001 Out of Asia: mitochondrial DNAevidence for an Bossuyt, F.&Milinkovitch, M.C.2000 Convergent adaptive Oriental origin of tiger frogs, genus Hoplobatrachus . Mol. radiations in Madagascanand Asian ranid frogs reveal Phylogenet. Evol. 21, 398–407. covariation between larval and adult traits. Proc.Natl Acad. Kraus, F., Campbell, E.W.&Allison, A.1999 Eleutherodac- Sci. USA 97, 6585–6590. tylus frog introductions to Hawaii. Herpetol. Rev. 30, 21–25. Bossuyt, F.&Milinkovitch, M.C.2001 Amphibiansas indi- Krause, D.W., Hartman, J.H.&Wells, N.A.1997 LateCre- cators of EarlyTertiary ‘out-of-India ’ dispersal of ver- taceous vertebrates from Madagascar.Implications for biotic tebrates. Science 292, 93–95. Briggs, J.C.2003 Thebiogeographic and tectonic history of changesin deep time. In Natural change and human impact (ed. S.M.Goodman &B.D.Patterson), pp. India. J.Biogeogr. 30, 381–388. in Madagascar 3–43. Washington, DC:Smithsonian Institution Press. Brown, J.H.&Lomolino, M.V.1998 Biogeography . Sun- derland, MA:Sinauer. Krause, D.W., Rogers, R.R., Forster, C.A., Hartman, J.H., Brown, R.F.&Guttman, S.I.2002 Phylogenetic systematics Buckley, G. A.&Sampson, S.D.1999 TheLate Cretaceous of the Rana signata complex of Philippine and Bornean vertebrate fauna of Madagascar:implications for Gondwanan stream frogs: reconsideration of Huxley ’smodification of paleobiogeography. GSA Today (Publ. Geol. Soc. Am. ) 9, 1–7. Wallace’sline atthe Oriental –Australian faunal zone inter- Kumar, S.&Hedges, S.B.1998 Amolecular time-scale for face. Biol. J.Linn.Soc. 76, 393–461. vertebrate evolution. Nature 392, 917–920. Buskirk, R.E.1985 Zoogeographic patterns and tectonic his- McCall, R.A.1997 Implications of recent geological investi- tory of Jamaicaand thenorthern Caribbean. J.Biogeogr. 12, gations of theMozambique Channel for themammalian 445–461. colonization of Madagascar. Proc.R. Soc. Lond. B 264, 663– Censky, E.J., Hodge, K.&Dudley, J.1998 Over-water disper- 665. (DOI 10.1098/rspb.1997.0094.) salof lizards due to hurricanes. Nature 395, 556. Meegaskumbura,M., Bossuyt, F., Pethiyagoda, R., Manamendra- Colonna, M.,Casanova, J.&Camoin, G.1996 Sea-level Arachchi, K., Bahir, M., Milinkovitch, M.C.&Schneider, C.J. changes and d 18Orecord for thepast 34 000 yr from May- 2002 Sri Lanka:an amphibianhot spot. Science 298, 379. otte reef, Indian Ocean. Quat. Res. 46, 335–339. Meirte, D.1999 Batraciens. In Lafaune terrestre de Mayotte Crother, B.I.&Guyer, C.1996 Caribbean historical biogeog- (ed. M.Louette), pp. 114 –135. Tervuren, Belgium: Royal raphy: wasthe dispersal-vicariance debate eliminated byan Museum for Central Africa. ( Annls.Mus. r. Afr. cent. (Sc. extraterrestrial bolide? Herpetologica 52, 440–465. Zool.) 284.) Darwin, C.1859 The originof species .London: John Murray. Moss, S.J.&Wilson, M.E.J.1998 Biogeographic impli- Duellman, W.E.&Trueb, L.1986 Biologyof amphibians . New cations from theTertiary palaeogeographic evolution of York: McGraw-Hill. Sulawesiand Borneo. In Biogeography and geological evolution

Proc.R. Soc.Lond. B (2003) 2442M. Vencesand others Multiple overseasdispersal in amphibians

of SE Asia (ed. R.Hall &D.Holloway), pp. 133 –155. estadisticasde 34? In Informe Galapagos 1998–1999 (ed. P. Leiden, TheNetherlands: Backhuys. Ospina& E.Munoz), pp. 65 –71. Quito, Ecuador: Funda- Murphy, W.J.&Collier, G.E.1997 Amolecular phylogeny cio´n Natura. for aplocheiloid fishes(Atherinomorpha, Cypridontiformes): Staub, F.1993 Fauna ofMauritius and associated flora . Maurit- therole of vicariance and theorigins of annualism. Mol. Biol. ius: Pre´cigraph. Evol. 14, 790–799. Storey, M., Mahoney, J.J., Saunders, A.D.,Duncan, R.A., Nougier, J., Cantagrel, J.M.&Karche, J.P.1986 TheCom- Kelley, S.P.&Coffin, M.F.1995 Timing of hot spot- ores archipelago in thewestern Indian Ocean: volcanology, related volcanism and thebreakup of Madagascarand India. geochronology and geodynamic setting. J.Afr.Earth Sci. 5, Science 167, 852–855. 135–145. Swofford, D.L.2002 Paup¤ .Phylogenetic analysis usingparsi- Palumbi, S.R., Martin, A., Romano, S., McMillan, W.O., mony (¤ and other methods), v. 4b10 .Sunderland, MA: Sinauer. Stice, L.&Grabowski, G.1991 The simple fool’s guide to Vences, M.&Glaw, F.2001 When molecules claim for taxo- PCR, v. 2.0.Honolulu, Hawaii: University of Hawaii. nomic change: new proposals on theclassification of Old Pitman III, W.C., Cande, S., LaBrecque, J.&Pindell, J.1993 World treefrogs. Spixiana 24, 85–92. Fragmentation of Gondwana: theseparation of Africa from Vences, M.&Glaw, F.2002 Molecular phylogeography of South America. In Biological relationships between Africaand Boophis tephraeomystax :atestcase for east-westvicariance in South America (ed. P.Goldblatt), pp. 15 –34. NewHaven, Malagasyanurans (Amphibia, Anura, Mantellidae). CT:Yale University Press. Spixiana 25, 79–84. Pleguezuelos, J.M.(ed.) 1997 Distribucio´nybiogeograf ´õ a de los Vences, M., Freyhof, J., Sonnenberg, R., Kosuch, J. &Veith, M. Anfibios yreptiles deEspan˜a y Portugal.Granada, Spain: Uni- 2001 Reconciling fossils and molecules: Cenozoic diver- versity of Granada. gence of cichlid fishesand thebiogeography of Madagascar. Posada, D.&Crandall, K.A.1998 M odelTest:testing the J.Biogeogr. 28, 1091–1099. model of DNAsubstitution. Bioinfomatics 14, 817–818. Vences, M., Kosuch, J., Glaw, F., Bo¨hme, W.&Veith, M. Rabinowitz, P.D., Coffin, M.F.&Falvey, D.1983 Thesepar- 2003 Molecular phylogeny of hyperoliid treefrogs: biogeo- ation of Madagascarand Africa. Science 220, 67–69. graphic origin of Malagasyand Seychellean taxaand re- Raxworthy, C.J., Forstner, M.R.J.&Nussbaum,R. A.2002 analysisof familial paraphyly. J.Zool.Syst. Evol. Res. 41, Chameleon radiation byoceanic dispersal. Nature 415, 205–215. 784–786. Voris, H.K.2000 Mapof pleistocene sealevels in southeast Richards, C.M.&Moore, W.S.1996 Aphylogeny for the Asia: shorelines, river systemsand time durations. J. African treefrog familyHyperoliidae basedon mitochondrial Biogeogr. 27, 1153–1167. rDNA. Mol.Phylogen. Evol. 5, 522–532. Whitmore, T.C.1987 Biogeographical evolution ofthe Malay Sanderson, M.J.1997 Anon-parametric approach to estimat- Archipelago .Oxford: Clarendon. ing divergence times in theabsence of rate constancy. Mol. Worthy, T.H., Anderson, A.J.&Molnar, R.E.1999 Biol. Evol. 14, 1218–1231. Megafaunalexpression in aland without mammals:the first Schoener, T.W., Spiller, D.A.&Losos, J.B.2001 Natural fossil faunasfrom terrestrial deposits in Fiji (Vertebrata: 79 restoration of thespecies –area relation for alizard after a Amphibia, Reptilia, Aves). Senck. Biol. , 237–242. hurricane. Science 294, 1525–1528. Shimodaira, H.&Hasegawa,M. 1999 Multiple comparison As this paper exceedsthe maximum lengthnormally permitted, the of log-likelihoods withapplications to phylogenetic infer- authors have agreedto contributeto production costs. ence. Mol.Biol. Evol. 16, 1114–1116. Snell, H.L.&Rea, S.1999 ElNin˜o 1997–1998 en Gala´pagos: Visithttp:/ /www.pubs.royalsoc.ac.uk to seeelectronic appendices to sepuede estimar 120 an˜os de variaciones clima´ticas con this paper.

Proc.R. Soc.Lond. B (2003)

These are an electronic appendices to the paper by Vences et al. 2003 Multiple overseas dispersal in amphibians. Proc. R. Soc. Lond. B 270, 2435–2442. (doi:10.1098/rspb.2003.2516)

Electronic appendices are refereed with the text. However, no attempt has been made to impose a uniform editorial style on the electronic appendices.

Table of Contents

Appendix A. Primers used for DNA amplification and sequencing

Appendix B. Voucher specimens and Genbank accession numbers

Appendix C. Substitution models used in Maximum Likelihood Phylogenetic analysis

Appendix D. Details of Bayesian phylogenetic analysis

Appendix E. Separate analysis of gene fragments

Appendix F. Testing alternative topologies with Shimodaira-Hasegawa tests

Appendix G. Estimates of divergence times

Appendix H. A survey of the terrestrial and freshwater vertebrate families of Madagascar, their biogeographic relationships and fossil ages

Appendix I. References used in electronic appendices

Appendix A. Primer sequences used for DNA amplification and sequencing

All primers are given in 5'-3' direction. Primers marked with an asterisk were newly developed for this paper.

Cytochrome b

F-Primer 1 (1): CBJ10933 - TAT GTT CTA CCA TGA GGA CAA ATA TC

F-Primer 2 (1): Cytb-a - CCA TGA GGA CAA ATA TCA TTY TGR GG

*F-Primer 3: MVZ15L-mod - AAC TWA TGG CCC MCA CMA TMC GWA A

R-Primer 1 (1): Cytb-c - CTA CTG GTT GTC CTC CGA TTC ATG T

*R-Primer 2: CytbAR-H-mod - TAW ARG GRT CYT CKA CTG GTT G

Tyrosinase (exon 1)

F-Primer 1 (1): Tyr-1b - AGG TCC TCY TRA GGA AGG AAT G

F-Primer 2 (1): Tyr-1d - TCC TCC GTG GGC ACC CAR TTC CC

F-Primer 3 (1): Tyr-1a - AGG TCC TCT TRA GCA AGG AAT G

*F-Primer 4: Tyr-F40 - AAR GAR TGY TGY CCI GTI TGG

*F-Primer 5: Tyr - Fx3 - ACT GGC CCA YTG THT TYT ACA AC

*F-Primer 6: Tyr - Fx4 - YTG GCC YWY TGT NTT YTA YAA C

R-Primer 1 (1): Tyr-1g - TGC TGG CRT CTC TCC ART CCC A

R-Primer 2 (1): Tyr-1e - GAG AAG AAA GAW GCT GGG CTG AG *R-Primer 3: Tyr-SPA - GAI GAG AAR AAR GAI GCT GGG CT

Rhodopsin (exon 1)

*F-Primer 1: Rhod-ma - AAC GGA ACA GAA GGY CC

F-Primer 2 (1): Rhod-1a - ACC ATG AAC GGA ACA GAA GGY CC

*R-Primer 1: Rhod-md - GTA GCG AAG AAR CCT TC

R-Primer 2 (1): Rhod-1d - GTA GCG AAG AAR CCT TCA AMG TA

R-Primer 3: Rhod-1c - CCA AGG GTA GCG AAG AAR CCT TC

12S rRNA & tRNAVal

F-Primer 1 (2): 12SAL - AAA CTG GGA TTA GAT ACC CCA CTA T

R-Primer 1 (2): 12SBH - GAG GGT GAC GGG CGG TGT GT

R-Primer 2: 16SR3 - TTT CAT CTT TCC CTT GCG GTA C

16S rRNA (5' fragment)

F-Primer (3): 16SL3 - AGC AAA GAH YWW ACC TCG TAC CTT TTG CAT

R-Primer (3): 16SAH - ATG TTT TTG ATA AAC AGG CG

16S rRNA (3' fragment)

F-Primer (2): 16SAL - CGC CTG TTT ATC AAA AAC AT

R-Primer (2): 16SBH - CCG GTC TGA ACT CAG ATC ACG T

Appendix B. Voucher specimens and Genbank accession numbers

Voucher specimens are deposited in the herpetological collections of the Museo Regionale di Scienze Naturali, Torino (MRSN), Université d'Antananarivo, Département de Biologie Animale (UADBA), Laboratorio de Biogeografía, Universidad de los Andes, Mérida, Venezuela (ULABG), Zoologisches Forschungsinstitut und Museum A. Koenig, Bonn (ZFMK), Zoologische Staatssammlung München (ZSM). Some UADBA numbers are preliminary fieldnumbers of F. Glaw and M. Vences (UADBA-FG/MV and UADBA-MV), some MRSN numbers are preliminary fieldnumbers of F. Andreone and J. E. Randrianirina (MRSN-FAZC and MRSN-RJS).

Asterisks mark sequences newly obtained in this study.

Mantellidae

Aglyptodactylus madagascariensis: AF249068 (cytochrome b), AF249103 (rhodopsin), AF249166 (tyrosinase), AF249007 (12S and tRNAVal), AY341678* (ZSM 183/2002, Tolagnaro, Madagascar) (16S [5' fragment]), AF249036 (16S [3' fragment]).

Boophis new species (Comoros): AY341733* (cytochrome b), AY341796* (rhodopsin), AY341752* (tyrosinase), AY341610* (12S and tRNAVal), AY341667* (16S [5' fragment]), AY341716* (16S [3' fragment]); all sequences from specimen ZSM 658/2000 (Tsingoni, Mayotte).

Boophis boehmei: AY341798* (rhodopsin), AY341612* (12S and tRNAVal), AY341669* (16S [5' fragment]), AY341717* (16S [3' fragment]); all sequences from specimen UADBA-MV 2001.1205 (Andasibe, Madagascar).

Boophis doulioti: AY341792* (rhodopsin), AY341608* (12S and tRNAVal), AY341663* (16S [5' fragment]), AF215334 (ZFMK 66690, Kirindy, Madagascar) (16S [3' fragment]); all other sequences from specimen ZSM 185/2002 (Nahampoana, Madagascar).

Boophis goudoti: AY341797* (rhodopsin), AY341611* (12S and tRNAVal), AY341668* (16S [5' fragment]), AJ315917 (not preserved; Col des Tapias, Madagascar) (16S [3' fragment]); all other sequences from specimen UADBA-MV 2001.557 (Andringitra, Madagascar).

Boophis idae: AY341795* (rhodopsin), AY341609* (12S and tRNAVal), AY341666* (16S [5' fragment]), AY341715* (16S [3' fragment]); all sequences from specimen ZSM 45/2002 (Andasibe, Madagascar).

Boophis luteus: AY341800* (rhodopsin), AY341614* (12S and tRNAVal), AY341671* (16S [5' fragment]), AJ315916 (UADBA-FG/MV 2000.063, Andasibe, Madagascar) (16S [3' fragment]); all sequences from specimen UADBA-FG/MV 2000.63 (Andasibe, Madagascar).

Boophis marojezensis: AY341803* (rhodopsin), AY341617* (12S and tRNAVal), AY341674* (16S [5' fragment]), AJ315923 (ZSM 326/2000, Vohidrazana, Madagascar) (16S [3' fragment]); all other sequences from specimen ZSM 189/2002 (Vohidrazana, Madagascar).

Boophis microtympanum: AY341799* (rhodopsin), AY341613* (12S and tRNAVal), AY341670* (16S [5' fragment]), AJ315918 (16S [3' fragment]); all sequences from specimen ZSM 393/2000 (Col des Tapias, Madagascar).

Boophis occidentalis: AY341806* (rhodopsin), AY341620* (12S and tRNAVal), AY341677* (16S [5' fragment]), AY341720* (16S [3' fragment]); all sequences from specimen ZSM 44/2002 (Antoetra, Madagascar).

Boophis rappiodes: AY341804* (rhodopsin), AY341618* (12S and tRNAVal), AY341675* (16S [5' fragment]), AJ314815 (16S [3' fragment]); all sequences from specimen ZSM 347/2000 (Andasibe, Madagascar).

Boophis sibilans: AY341801* (rhodopsin), AY341615* (12S and tRNAVal), AY341672* (16S [5' fragment]), AY341718* (16S [3' fragment]); all sequences from specimen ZSM 39/2002 (Andasibe, Madagascar).

Boophis tephraeomystax: AF249070 (cytochrome b), AY341793* (rhodopsin), AF249168 (tyrosinase), AF249009 (12S and tRNAVal), AY341664* (16S [5' fragment]), AJ312116 (16S [3' fragment]); all sequences from specimen UADBA-FG/MV 2000.379 (Sambava, Madagascar).

Boophis vittatus: AY341802* (rhodopsin), AY341616* (12S and tRNAVal), AY341673* (16S [5' fragment]), AY341719* (16S [3' fragment]); all sequences from specimen UADBA-FG/MV 2000.82 (Tsaratanana, Madagascar).

Boophis viridis: AY341805* (rhodopsin), AY341619* (12S and tRNAVal), AY341676* (16S [5' fragment]), AJ314818 (16S [3' fragment]); all sequences from specimen ZSM 338/2000 (Andasibe, Madagascar).

Boophis xerophilus: AY341794* (rhodopsin), AF249008 (12S and tRNAVal), AY341665* (16S [5' fragment]), AF215335 (16S [3' fragment]); all sequences from specimen ZFMK 66705 (Kirindy, Madagascar).

Laliostoma labrosum: AF249096 (cytochrome b), AF249106 (rhodopsin), AF249169 (tyrosinase), AF249010 (12S and tRNAVal), AY341679* (UADBA-MV 2001.289, Ankarafantsika, Madagascar) (16S [5' fragment]), AF249037 (16S [3' fragment]).

Mantella laevigata: AY263277 (rhodopsin), AY341607* (12S and tRNAVal), AJ438538 (16S [5' fragment]), AF215279 (16S [3' fragment]).

Mantella madagascariensis: AF249076 (cytochrome b), AF249101 (rhodopsin), AF249164 (tyrosinase), AF249005 (12S and tRNAVal), AJ438892 (16S [5' fragment]), AF249049 (16S [3' fragment]).

Mantidactylus new species (Comoros): AY341731* (cytochrome b), AY323742* (rhodopsin), AY341750* (tyrosinase), AY341585* (12S and tRNAVal), AY341639* (16S [5' fragment]), AY330888* (16S [3' fragment]); all sequences from specimen ZSM 652/2000 (Mont Combani, Mayotte).

Mantidactylus albolineatus: AY341766* (rhodopsin), AY341580* (12S and tRNAVal), AY341635* (16S [5' fragment]), AY341701* (16S [3' fragment]); all sequences from specimen ZSM 250/2002 (Andasibe, Madagascar).

Mantidactylus ambreensis: AY341788* (rhodopsin), AY341603* (12S and tRNAVal), AY341659* (16S [5' fragment]), AY324822* (ZSM 492/2000, Montagne d'Ambre) (16S [3' fragment]); all other sequences from specimen ZSM 634/2001 (Tsaratanana, Madagascar).

Mantidactylus asper: AY341783* (rhodopsin), AY341598* (12S and tRNAVal), AY341653* (16S [5' fragment]), AJ314802 (16S [3' fragment]); all sequences from specimen UADBA- FG/MV 2000.17 (Mandraka, Madagascar).

Mantidactylus biporus: AY341784* (rhodopsin), AY341599* (12S and tRNAVal), AY341655* (16S [5' fragment]), AF215322 (ZFMK 70481, Masoala) (16S [3' fragment]); all other sequences from specimen ZSM 122/2002 (Andranofotsy, Madagascar).

Mantidactylus blommersae: AY341770* (rhodopsin), AY341584* (12S and tRNAVal), AY341638* (16S [5' fragment]), AF317688 (16S [3' fragment]); all sequences from specimen UADBA-FG/MV 2000.65 (Andasibe, Madagascar).

Mantidactylus charlotteae: AY341790* (rhodopsin), AY341605* (12S and tRNAVal), AY341661* (16S [5' fragment]), AY341713* (16S [3' fragment]); all sequences from specimen ZSM 127/2002 (Andranofotsy, Madagascar).

Mantidactylus depressiceps: AY341775* (rhodopsin), AY341590* (12S and tRNAVal), AY341645* (16S [5' fragment]), AF215326 (ZFMK 60131, Andasibe) (16S [3' fragment]); all other sequences from specimen ZSM 688/2001 (Andasibe, Madagascar).

Mantidactylus domerguei: AY341768* (rhodopsin), AY341582* (12S and tRNAVal), AY341636* (16S [5' fragment]), AF317689 (16S [3' fragment]); all sequences from specimen ZSM 353/2000 (Mantasoa, Madagascar).

Mantidactylus massorum: AY341776* (rhodopsin), AY341591* (12S and tRNAVal), AY341646* (16S [5' fragment]), AY341705* (16S [3' fragment]); all sequences from specimen MRSN-FAZC 6737 (Ambolokopatrika, Madagascar).

Mantidactylus grandidieri: AY341789* (rhodopsin), AY341604* (12S and tRNAVal), AY341660* (16S [5' fragment]), AY341712* (16S [3' fragment]); all sequences from specimen UADBA-MV 2001.1201 (Andasibe, Madagascar).

Mantidactylus grandisonae: AY341771* (rhodopsin), AY341640* (16S [5' fragment]), AF215315 (16S [3' fragment]); all sequences from specimen ZFMK 66669 (Ambato, Madagascar).

Mantidactylus granulatus: AY341779* (rhodopsin), AY341594* (12S and tRNAVal), AY341649* (16S [5' fragment]), AJ314794 (ZSM 645/2001, Tsaratanana) (16S [3' fragment]); all other sequences from specimen MRSN-FAZC 8011 (Nosy Be, Madagascar).

Mantidactylus horridus: AY341781* (rhodopsin), AY341596* (12S and tRNAVal), AY341651* (16S [5' fragment]), AY341708* (16S [3' fragment]); all sequences from specimen UADBA 10002 (Tsaratanana, Madagascar).

Mantidactylus kely: AY341769* (rhodopsin), AY341583* (12S and tRNAVal), AY341637* (16S [5' fragment]), AF317690 (16S [3' fragment]); all sequences from specimen ZSM 363/2000 (Ambatolampy, Madagascar).

Mantidactylus liber: AY341774* (rhodopsin), AY341589* (12S and tRNAVal), AY341644* (16S [5' fragment]), AJ314801 (16S [3' fragment]); all sequences from specimen ZSM 491/2000 (Montagne d'Ambre, Madagascar).

Mantidactylus lugubris: AY341785* (rhodopsin), AY341600* (12S and tRNAVal), AY341656* (16S [5' fragment]), AY341710* (16S [3' fragment]); all sequences from specimen ZSM 166/2002 (Mantady, Madagascar).

Mantidactylus new species (aff. lugubris): AY341786* (rhodopsin), AY341601* (12S and tRNAVal), AY341657* (16S [5' fragment]), AY341711* (16S [3' fragment]); all sequences from specimen ZSM 171/2002 (Mantady, Madagascar).

Mantidactylus madinika: AY341772* (rhodopsin), AY341587* (12S and tRNAVal), AY341642* (16S [5' fragment]), AY341703* (16S [3' fragment]); all sequences from one paratype specimen from Antsirasira, Madagascar.

Mantidactylus femoralis: AY341787* (rhodopsin), AY341602* (12S and tRNAVal), AY341658* (16S [5' fragment]), AY324812* (16S [3' fragment]); all sequences from specimen UADBA-MV 2001.1277 (Andasibe, Madagascar).

Mantidactylus opiparis: AY341791* (rhodopsin), AY341606* (12S and tRNAVal), AY341662* (16S [5' fragment]), AY341714* (16S [3' fragment]); all sequences from specimen UADBA-MV 2001.1069 (Marolambo region, Madagascar).

Mantidactylus peraccae: AY341777* (rhodopsin), AY341592* (12S and tRNAVal), AY341647* (16S [5' fragment]), AY341706* (16S [3' fragment]); all sequences from specimen MRSN-RJS 109 (Tsaratanana, Madagascar).

Mantidactylus redimitus: AY341778* (rhodopsin), AY341593* (12S and tRNAVal), AY341648* (16S [5' fragment]), AY341707* (16S [3' fragment]); all sequences from specimen ZSM 152/2002 (Vohidrazana, Madagascar).

Mantidactylus sculpturatus: AY341782* (rhodopsin), AY341597* (12S and tRNAVal), AY341652* (16S [5' fragment]), AY341709* (16S [3' fragment]); all sequences from specimen ZSM 95/2002 (Mantady, Madagascar).

Mantidactylus striatus: AY341780* (rhodopsin), AY341595* (12S and tRNAVal), AY341650* (16S [5' fragment]), AJ314796* (16S [3' fragment]).

Mantidactylus aff. ulcerosus: AF249102* (rhodopsin), AF249006* (12S and tRNAVal), AY341654* (16S [5' fragment]), AF215319 (16S [3' fragment]); all sequences from specimen ZFMK 66659 (Ambato, Madagascar).

Mantidactylus wittei: AY341732* (cytochrome b), AY323743* (rhodopsin), AY341751* (tyrosinase), AY341586* (12S and tRNAVal), AY341641* (16S [5' fragment]), AF317691 (UADBA-FG/MV 2000.123, Ambanja) (16S [3' fragment]); all other sequences from specimen ZSM 405/2000 (Benavony, Madagascar).

Ranidae

Hoplobatrachus crassus: AF249090 (cytochrome b), AF249109 (rhodopsin), AF249172 (tyrosinase), AF249013 (12S and tRNAVal), AY341688* (16S [5' fragment]), AY014375 (16S [3' fragment]).

Hoplobatrachus occipitalis: AJ564733* (cytochrome b), AJ564730* (rhodopsin), AJ564729* (tyrosinase), AJ564734* (12S and tRNAVal), AY341689* (16S [5' fragment]), AY014373 (16S [3' fragment]); all sequences from specimen ZFMK-WB 02 (Mauritania).

Indirana cf. leptodactyla: AF215392 (16S [5' fragment]), AY341686* (16S [3' fragment]); sequences from specimen ZFMK uncatalogued (Ooty, India).

Indirana sp. (aff. leptodactyla): AF249080 (cytochrome b), AF249123 (rhodopsin), AF249186 (tyrosinase), AF249027 (12S and tRNAVal).

Nyctibatrachus major: AF249084 (cytochrome b), AF249113 (rhodopsin), AF249176 (tyrosinase), AF249017 (12S and tRNAVal), AY341687* (16S [5' fragment]), AF215397 (16S [3' fragment]).

Petropedetes parkeri: AY341813* (rhodopsin), AY341757* (tyrosinase), AY341628* (12S and tRNAVal), AY341694* (16S [5' fragment]), AY341724* (16S [3' fragment]); all sequences from a specimen from Nlonako (Cameroon).

Petropedetes sp. (aff. parkeri): AY341738* (cytochrome b); sequence from a specimen from Cameroon.

Ptychadena mascareniensis: AY341734* (cytochrome b), AY341809* (rhodopsin), AY341753* (tyrosinase), AY341624* (12S and tRNAVal), AY341690* (16S [5' fragment]); all sequences from specimen ZSM 258/2002 (Nahampoana, Madagascar ).

Rana temporalis: AF249083 (cytochrome b), AF249118 (rhodopsin), AF249181 (tyrosinase), AF249022 (12S and tRNAVal), AY341683* (16S [5' fragment]), AF215390 (16S [3' fragment]).

Rana temporaria: AF249078 (cytochrome b), AF249119 (rhodopsin), AF249182 (tyrosinase), AF249023 (12S and tRNAVal), AY341684* (16S [5' fragment]), AF249048 (16S [3' fragment]).

Rana cretensis: AY148010 (rhodopsin).

Rana cerigensis: AY148009 (rhodopsin).

Rhacophoridae

Polypedates cruciger: AF249089 (cytochrome b), AF249124 (rhodopsin), AF249187 (tyrosinase), AF249028 (12S and tRNAVal), AY341685* (16S [5' fragment]), AF215357 (16S [3' fragment]).

Astylosternidae

Astylosternus diadematus: AY341735* (cytochrome b), AY341810* (rhodopsin), AY341754* (tyrosinase), AY341625* (12S and tRNAVal), AY341691* (16S [5' fragment]), AY341723* (16S [3' fragment]); all sequences from a specimen from Nlonako (Cameroon).

Arthroleptidae

Arthroleptis variabilis: AY341737* (cytochrome b), AY341812* (rhodopsin), AY341756* (tyrosinase), AY341627* (12S and tRNAVal), AY341693* (16S [5' fragment]), AF124107 (ZFMK 68794, Cameroon) (16S [3' fragment]); all other sequences from a specimen from Nlonako (Cameroon).

Hyperoliidae

Leptopelis natalensis: AY341736* (cytochrome b), AY341811* (rhodopsin), AY341755* (tyrosinase), AY341626* (12S and tRNAVal), AY341692* (16S [5' fragment]), AF215448 (16S [3' fragment]); all sequences from specimen ZFMK 68785 (Mtunzini, South Africa).

Hyperolius cf. viridiflavus: AF249066 (cytochrome b), AF249098 (rhodopsin), AF249161 (tyrosinase), AF249002 (12S and tRNAVal).

Hyperolius viridiflavus: AY341695* (16S [5' fragment]), AF215440 (16S [3' fragment]); sequences from specimen ZFMK 66726 (Barberton, South Africa).

Tachycnemis seychellensis: AY341739* (cytochrome b), AY341814* (rhodopsin), AY341758* (tyrosinase), AY341629* (12S and tRNAVal), AY341696* (16S [5' fragment]), AF215451 (16S [3' fragment]).

Heterixalus tricolor: AY341740* (cytochrome b), AY323741* (rhodopsin), AY341759* (tyrosinase), AY341630* (12S and tRNAVal), AY341697* (16S [5' fragment]), AY341725* (16S [3' fragment]); all sequences from specimen ZSM 700/2001 (Ankarafantsika, Madagascar).

Bufonidae

Bufo melanostictus: AF249082 (cytochrome b), AF249097 (rhodopsin), AF249001 (12S and tRNAVal), AF249061 (16S [3' fragment]).

Bufo asper: AY263257 (16S [5' fragment]).

Leptodactylidae

Leptodactylus fuscus: AY341741* (cytochrome b), AY323746* (rhodopsin), AY341760* (tyrosinase), AY341631* (12S and tRNAVal), AY263262* (16S [5' fragment]), AY263226* (16S [3' fragment]); all sequences from specimen ULABG 4591 (Canaima, Venezuela).

Sooglossidae

Nesomantis thomasseti: AY341742* (cytochrome b), AY323744* (rhodopsin), AY341761* (tyrosinase), AY341632* (12S and tRNAVal), AY341698* (16S [5' fragment]), AY330889* (16S [3' fragment]).

Pipidae

Pipa parva: AY341743* (cytochrome b), AY323734* (rhodopsin), AY341762* (tyrosinase), AY341633* (12S and tRNAVal), AY341699* (16S [5' fragment]), AY333690* (16S [3' fragment]).

Xenopus laevis: M10217 (cytochrome b), S62229 (rhodopsin), AY341764* (tyrosinase), M10217 (12S and tRNAVal), M10217 (16S [5' fragment]), AY341727* (16S [3' fragment]).

Hymenochirus boettgeri: AY341744* (cytochrome b), AY323735* (rhodopsin), AY341763* (tyrosinase), AY341634* (12S and tRNAVal), AY341700* (16S [5' fragment]), AY341726* (16S [3' fragment]).

Pelobatidae

Pelobates cultripes: AY323736* (rhodopsin).

Pelobates varaldi: AY341815* (rhodopsin).

Discoglossidae

Alytes muletensis: AY341728* (cytochrome b), AY323731* (rhodopsin), AY341747* (tyrosinase), AY341621* (12S and tRNAVal), AY341680* (16S [5' fragment]), AF224729 (16S [3' fragment]); all sequences from a specimen preserved in the ZFMK (Mallorca, Spain).

Alytes maurus: AY341816* (rhodopsin).

Alytes dickhillenii: AY341817* (rhodopsin).

Urodela

Ambystoma mexicanum: AY341745* (cytochrome b), U36574 (rhodopsin), Y10947 (12S and tRNAVal), Y10947 (16S [5' fragment]), Y10947 (16S [3' fragment]).

Hynobius kimurae: AY341746 (cytochrome b), AY341765* (tyrosinase).

Appendix C. Substitution models used in Maximum Likelihood Phylogenetic analysis

We tested the goodness-of-fit of nested substitution models for homogeneous data partitions of ingroup taxa by a hierarchical likelihood ratio test (HLRT). Modeltest (4) version 3.06 was used to calculate the test statistic δ = 2 log Λ with Λ being the ratio of the likelihood of the null model divided by the likelihood of the alternative model (5). The best fitting model was used for Maximum Likelihood phylogenetic analyses using PAUP* (6).

Before applying the HLRT and any phylogenetic analysis, we excluded all gapped and hypervariable regions of the rRNA and tRNA genes. Rather than trying to fit these regions (which mainly correspond to loops in the tertiary structure of the rRNA molecules) into secondary structure models (7), we preferred to exclude all sections in which homology was not immediately obvious, as well as all gapped positions. Additionally, we excluded the third positions of cytochrome b which are known to be fully saturated at the level of anuran families (8).

The following substitution models were selected by the hierarchical likelihood ratio test (HLRT) as implemented in Modeltest:

For the among-Mantellidae dataset of the concatenated rhodopsin, 12S rRNA, 16SrRNA and tRNAVal sequences, a GTR+I+G substitution model (-lnL = 26025.7305) with empirical base frequencies (freqA = 0.3551; freqC = 0.2199; freqG = 0.1644; freqT = 0.2607) and substitution rates ([A-C] = 3.7878; [A-G] = 10.5790; [A-T] = 5.7768; [C-G] = 1.1797; [C-T] = 29.2325; [G- T] = 1), a proportion of invariable sites of 0.4042 and a gamma distribution shape parameter of 0.6582.

For the higher-level relationship data set of the concatenated rhodopsin, tyrosinase, cytochrome b, 12SrRNA, 16S rRNA, and tRNAVal sequences, a TrN+I+G substitution model (-lnL = 30463.0156), with empirical base frequencies (freqA = 0.3371; freqC = 0.2527; freqG = 0.1497; freqT = 0.2606) and substitution rates ([A-G] = 3.3874; [C-T] = 5.2170; all other rates = 1), a proportion of invariable sites of 0.2851, and a gamma distribution shape parameter of 0.6848.

For the higher-level relationship data set of nuclear genes only (rhodopsin, tyrosinase), a HKY+I+G substitution model (-lnL = 8573.2109) with empirical base frequencies (freqA = 0.2261; freqC = 0.3242; freqG = 0.2048; freqT = 0.2449), a transition/transversion ratio of 2.1689, a proportion of invariable sites of 0.3788, and a gamma distribution shape parameter of 1.25.

For the higher-level relationship data set of rRNA and tRNA genes only (12S rRNA, 16S rRNA, tRNAVal), a GTR+I+G substitution model (-lnL = 17708.6992) with empirical base frequencies (freqA = 0.3763; freqC = 0.2075; freqG = 0.1800; freqT = 0.2362) and substitution rates ([A-C] = 3.8266; [A-G] = 7.5824; [A-T] = 5.2028; [C-G] = 0.6641; [C-T] = 24.8576; [G-T] = 1), a proportion of invariable sites of 0.1657, and a gamma distribution shape parameter of 0.5400.

Appendix D. Details of Bayesian phylogenetic analysis

Bayesian phylogenetic analyses were performed using the program MrBayes, version 2.01 (9). The analysis consisted of Maximum Likelihood (ML) comparisons of trees in which the tree topology and ML parameters were permuted using a Markov chain Monte Carlo method with Metropolis-coupling (MCMCMC) (9), and sampled periodically according to the Metropolis- Hastings algorithm. Because Modeltest (4) suggested complex substitution models (GTR+I+G; see previous section), we set the ML parameters in MrBayes as follows: "lset nst = 6" (the GTR model), "rates = invgamma" (site-specific rate variation, with some invariant sites and other sites with a gamma distribution), and "basefreq = estimate" (proportion of different base categories estimated from the data).

All our analyses employed one cold chain and three incrementally heated chains. We run 500,000 generations in the higher-level phylogeny (Fig. 5) and 300,000 generations in the mantellid phylogeny (fig. 4). Trees were sampled every ten generations. The initial set of generations needed before convergence on stable likelihood values (burnin) was set at 6-7% based on own empirical evaluation.

Appendix E. Separate analysis of gene fragments

The following figures show the results of additional analyses of the two data sets. For the Mantellidae data set, we present the maximum likelihood (ML) phylogram with maximum parsimony (MP) bootstrap values. For the higher-level relationship data set, we show (a) a ML analysis of nuclear genes (rhodopsin and tyrosinase) at the amino acid level, (b) a ML analysis of nuclear genes only at the amino acid level, but using Alytes muletensis (Discoglossidae) as outgroup instead of a salamander, ( c) a ML analysis of nuclear genes only at the nucleotide level, and (d) a ML analysis based on ribosomal genes only (12SrRNA, 16SrRNA, tRNAVal). Bootstrap values are shown for ML, MP and Neighbor-joining (NJ).

Fig. 4. ML phylogram showing relationships among the Mantellidae, obtained by heuristic searches in PAUP* using settings suggested by Modeltest. The numbers are bootstrap values in percent under MP (2000 replicates).

Fig. 5. ML phylogram showing higher-level relationships among Gondwanan frogs. Based on amino acid sequences of nuclear genes (rhodopsin and tyrosinase; VT substitution model (10)). The numbers are quartet puzzling values in percent under ML (10,000 puzzling steps, obtained using Tree-Puzzle (11)), bootstrap values under MP (2000 replicates; obtained using PAUP*, with 10 random addition sequence replicates) and NJ (obtained using Mega (12), under minimum evolution criterion using an empirically determined gamma distribution shape parameter with eight rate categories).

77 Mantidactylus new species Comoros ML 99 NJ 95 Mantidactylus wittei MP Aglyptodactylus madagascariensis 67 Boophis tephraeomystax 98 95 Boophis new species Comoros

0.1 substitutions/site Laliostoma labrosum 82 80 Rana galamensis 60 76 Rana temporalis 69 87 41 Hoplobatrachus crassus 100 100 Hoplobatrachus occipitalis 96 90 Chiromantis rufescens 93 Polypedates cruciger ---- 61 Indirana cf leptodactyla 49 Nyctibatrachus major 54 Ptychadena mascareniensis 39 76 48 Petropedetes parkeri 96 93 54 Tachycnemis seychellensis 83 87 41 100 Heterixalus tricolor 100 ---- Hyperolius viridiflavus 55 69 Leptopelis natalensis 30 79 72 ---- Astylosternus diadematus 46 Arthroleptis variabilis 40 Leptodactylus fuscus 87 59 Nesomantis thomasseti 65 Pipa parva 71 94 Hymenochirus boettgeri 89 Xenopus laevis

Alytes muletensis

Ambyostoma mexicanum

Fig. 6. ML phylogram showing higher-level relationships among Gondwanan frogs. The numbers are bootstrap values in percent under ML, MP, and NJ. Based on amino acid sequences of nuclear genes (rhodopsin and tyrosinase). Because of the high divergence of the salamander sequences, this separate tree was calculated using the basal discoglossid frog Alytes muletensis as outgroup. See legend of Fig. 5 for further details of analysis.

76 52 Mantidactylus new species Comoros ML 99 NJ ------94 Mantidactylus wittei MP Aglyptodactylus madagascariensis 62 Boophis tephraeomystax 41 32 ---- 45 Boophis new species Comoros ---- Laliostoma labrosum

0.1 substitutions/site Nyctibatrachus major 50 Ptychadena mascareniensis 62 40 32 56 Petropedetes parkeri 45 Indirana cf leptodactyla ---- 95 49 Chiromantis rufescens 90 43 92 Polypedates cruciger 87 Hoplobatrachus crassus 100 76 83 100 Hoplobatrachus occipitalis 67 79 74 47 65 Rana galamensis 96 92 Rana temporalis 94 56 Tachycnemis seychellensis 88 45 100 Heterixalus tricolor 55 100 ---- Hyperolius viridiflavus 67 29 32 Leptopelis natalensis ------81 ---- 32 Astylosternus diadematus ---- 66 Arthroleptis variabilis

Leptodactylus fuscus

Nesomantis thomasseti ---- Pipa parva 70 56 94 70 Hymenochirus boettgeri 89 Xenopus laevis

Alytes muletensis

Fig. 7. ML phylogram showing higher-level relationships among Gondwanan frogs. The numbers are bootstrap values in percent under ML, MP and NJ. Based on nucleotide sequences of nuclear genes (rhodopsin and tyrosinase). See legend of Fig. 5 for further details of analysis.

84 Chiromantis rufescens 100 ML 98 Polypedates cruciger NJ 84 MP Mantidactylus new species Comoros 100 100 Mantidactylus wittei

Aglyptodactylus madagascariensis 88 Boophis tephraeomystax 100 0.05 substitutions/site 100 Boophis new species Comoros

Laliostoma labrosum

83 Rana galamensis 100 100 Rana temporalis 97 Hoplobatrachus crassus 100 40 100 Hoplobatrachus occipitalis 99 95 Nyctibatrachus major 59 Indirana cf leptodactyla 93 81 Petropedetes parkeri 63 100 Ptychadena mascareniensis 100 83 Astylosternus diadematus 94 59 Leptopelis natalensis 62 ---- Arthroleptis variabilis 72 100 51 94 92 Hyperolius viridiflavus 100 63 100 Tachycnemis seychellensis 35 86 79 100 Heterixalus tricolor 98 Leptodactylus fuscus

Nesomantis thomasseti 78 Pipa parva 72 22 100 61 Hymenochirus boettgeri 100 Xenopus laevis

Alytes muletensis

Fig. 8. ML phylogram showing higher-level relationships among Gondwanan frogs. Obtained by heuristic searches in PAUP* with TBR branch-swapping and 10 random sequence addition replicates, under the substitution model proposed by Modeltest. The numbers are bootstrap values in percent under ML (100 replicates), Maximum Parsimony (2000 replicates), and Neighbor-joining (2000 replicates). Based on nucleotide sequences of ribosomal genes

(12SrRNA, 16SrRNA, tRNAVal) only. Although the ML tree did not place Nesomantis (Sooglossidae) as sister group of all Neobatrachia, such a topology was suggested by the MP and NJ bootstrap consensus trees (66% and 63% support), in agreement with the trees based on nuclear genes (see Figs. 5-7).

Appendix F. Testing alternative topologies with Shimodaira-Hasegawa tests

Non-parametric likelihood ratio tests (Shimodaira-Hasegawa tests - SH tests) (13) as implemented in PAUP* (6), were used to test alternative topologies (chosen a priori) in the phylogenetic trees obtained by Maximum Likelihood searches. This is the only statistically valid method currently available for multiple topology testing (14). We run SH tests with 2000 bootstrap replicates and full optimization settings. The following tables show the topologies tested, their likelihood scores, the difference in likelihood score as compared to the best tree, and the significance of this difference.

In the mantellid phylogeny, we tested alternative positions of the two Comoroan species, keeping the overall topology unchanged. The results, summarized in Table 1, show that the general position of these taxa is statistically significant. The only topologies that could not be excluded were (a) permutations of the position of Mantidactylus sp. (Comoros) relative to M. wittei, M. domerguei and M. blommersae, and (b) of Boophis sp. (Comoros) relative to B. tephraeomystax, B. doulioti and B. xerophilus. Considering any of these species other than M. wittei and B. tephraeomystax as sister taxa of the Comoroan species would result in higher genetic divergences of the Comoroan taxa, thus reinforcing the molecular clock calculations performed by suggesting even younger origins of the lineages studied. In contrast, hypotheses placing the Comoroan taxa as sister groups to each other, or as basal to their respective lineages (subgenera or species groups) could be significantly rejected. This provides evidence that the Comoroan taxa are deeply nested within their clades, and strongly confirms their double independent origin. Although the Comoroan Mantidactylus had previously been included in Mantidactylus granulatus, a relationship to this species or its relatives (M. redimitus, M. sculpturatus) could also be excluded with high significance. In the taxa set used to assess higher-level relationships, the SH-tests significantly excluded relationships between the two Seychellean taxa Tachycnemis and Nesomantis, confirming the independent origin of the two Seychellean frog lineages. In contrast, the arrangement of the different mantellid subfamilies was not sufficiently clarified, and also several alternative positions of Nesomantis could not be significantly excluded, indicating that our data set is insufficient for resolving satisfyingly the deep splits among the main frog lineages.

Table 1. Results of SH-tests in the Mantellidae data set. M. sp. and B. sp. refer to the Mantidactylus and Boophis species from Mayotte (Comoros). Significances of P<0.05 are marked with an asterisk.

Tree Description of ree permutation Likelihood (-lnL) Diff (-lnL) Significance A0 Best ML tree 26013.35215 ------A1 M. sp. sister to M. domerguei / M. blommersae 26042.78229 29.43014 0.5185 A2 M. sp. sister to M. wittei / M. domerguei / M. blommersae 26042.78229 29.43014 0.5185 A3 M. sp. sister to M. domerguei 26074.46334 61.11119 0.0730 A4 M. sp. sister to M. blommersae 26074.84296 61.49082 0.0695 A5 M. sp. sister to M. grandisonae / M. kely / M. sarotra 26090.12128 76.76914 0.0160* A6 M. sp. sister to M. grandisonae 26108.24787 94.89572 0.0045* A7 M. sp. sister to M. kely / M. sarotra 26108.24787 94.89572 0.0045* A8 M. sp. sister to M. kely 26187.82919 174.47704 0.0000* A9 M. sp. sister to M. sarotra 26187.82919 174.47704 0.0000* A10 M. sp. sister to all Blommersia 26090.12128 76.76914 0.0160* A11 M. sp. sister to M. madinika / Mantella spp. 26164.94468 151.59254 0.0000* A12 M. sp. sister to M. madinika 26205.35170 191.99955 0.0000* A13 M. sp. sister to species of Guibemantis / Pandanusicola 26133.00009 119.64794 0.0005* A14 M. sp. sister to species of Guibemantis / Pandanusicola 26188.96063 175.60848 0.0000* A15 M. sp. sister to species of Guibemantis / Pandanusicola 26188.96063 175.60848 0.0000* A16 M. sp. sister to M. granulatus 26381.78107 368.42892 0.0000* A17 M. sp. sister to M. redimitus 26381.78107 368.42892 0.0000* A18 M. sp. sister to M. granulatus / M. redimitus 26346.10407 332.75192 0.0000* A19 M. sp. sister to M. sculpturatus / M. granulatus / M.redimitus 26338.20727 324.85512 0.0000* A20 B. sp. sister to M. sp. 26465.21993 451.86778 0.0000* A21 M. sp. sister to B. sp. 26443.05343 429.70129 0.0000* A22 B. sp. sister to B. tephraeomystax / B.doulioti 26014.00790 0.65576 0.9875 A23 B. sp. sister to B. tephraeomystax 26016.30986 2.95772 0.9785 A24 B. sp. sister to B. xerophilus 26068.06433 54.71218 0.1380 A25 B. sp. sister to B. doulioti / B. tephraeomystax / B. xerophilus 26069.01388 55.66174 0.1245 A26 B. sp. sister to B. idae 26117.88218 104.53003 0.0035* A27 B. sp. sister to all pond breeding Boophis 26117.88218 104.53003 0.0035* A28 B. sp. sister to all Boophis 26173.48306 160.13091 0.0000* A29 B. sp. sister to all brook breeding Boophis 26173.48306 160.13091 0.0000*

Table 2. Results of SH-tests in the higher-level relationship data set. Significances of P<0.05 are marked with an asterisk.

Tree Description of ree permutation Likelihood (-lnL) Diff (-lnL) Significance B0 Best ML tree 30444.99194 ------B1 Different position of Nesomantis 30447.52743 2.53548 0.940 B2 Different position of Nesomantis 30508.07683 63.08489 0.085 B3 Different position of Nesomantis 30510.36233 65.37039 0.077 B4 Different position of Nesomantis 30516.91704 71.92510 0.060 B5 Nesomantis sister to Ranoidea 30449.36063 4.36869 0.935 B6 Nesomantis sister to Ranidei 30509.67657 64.68463 0.044* B7 Nesomantis sister to Arthroleptoidei 30508.79564 63.80370 0.050 B8 Nesomantis sister to Tachycnemis 30644.78662 199.79468 0.000* B9 Tachycnemis sister to Nesomantis 31004.34740 559.35546 0.000* B10 Rhacophoridae sister of Mantellidae 30451.63068 6.63874 0.847 B11 Rhacophoridae sister of Mantellidae 30450.20593 5.21398 0.859 B12 Different arrangement of mantellid subfamilies 30448.63226 3.64031 0.942 B13 Different arrangement of mantellid subfamilies 30452.15834 7.16640 0.891 B14 Different arrangement of mantellid subfamilies 30489.00288 44.01094 0.217

Table 3. (next page). Topologies of trees included in Shimodaira-Hasegawa non-parametric likelihood ratio tests (dataset: Mantellidae). Taxa: 1, Polypedates cruciger; 2, Mantidactylus depressiceps; 3, Mantidactylus albolineatus; 4, Mantidactylus liber; 5, Mantidactylus domerguei; 6, Mantidactylus blommersae; 7, Mantidactylus new species Comoros; 8, Mantidactylus wittei; 9, Mantidactylus grandisonae; 10, Mantidactylus kely; 11, Mantidactylus sarotra; 12, Mantidactylus madinika; 13, Mantella madagascariensis; 14, Mantella laevigata; 15, Mantidactylus fimbriatus; 16, Mantidactylus peraccae; 17, Mantidactylus sculpturatus; 18, Mantidactylus redimitus; 19, Mantidactylus granulatus; 20, Mantidactylus asper; 21, Mantidactylus striatus; 22, Mantidactylus horridus; 23, Mantidactylus grandidieri; 24, Mantidactylus aff ulcerosus; 25, Mantidactylus biporus; 26, Mantidactylus charlotteae; 27, Mantidactylus opiparis; 28, Mantidactylus lugubris; 29, Mantidactylus sp aff lugubris; 30, Mantidactylus mocquardi; 31, Mantidactylus ambreensis; 32, Aglyptodactylus madagascariensis; 33, Laliostoma labrosum; 34, Boophis idae; 35, Boophis xerophilus; 36, Boophis tephraeomystax; 37, Boophis doulioti; 38, Boophis new species Comoros; 39, Boophis microtympanum; 40, Boophis occidentalis; 41, Boophis goudoti; 42, Boophis boehmei; 43, Boophis luteus; 44, Boophis sibilans; 45, Boophis rappiodes; 46, Boophis viridis; 47, Boophis vittatus; 48, Boophis marojezensis; 49, Rana temporaria; 50, Rana temporalis.

Tree Topology

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A0 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (7, (5,6))), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A1 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (7, (8, (5,6))), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A2 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (6, (5,7))), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A3 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5, (6,7))), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A4 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (7, (9, (10,11))))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A5 (47,48)))), (49,50)) (1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), ( (7,9), (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A6 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (7, (10,11))))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A7 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (11, (7,10))))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A8 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10, (7,11))))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A9 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), (7, ( (8, (5,6)), (9, (10,11))))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A10 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), (7, (12, (13,14)))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A11 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), ( (7,12), (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A12 (47,48)))), (49,50))

(1, ( ( ( ( ( (7, (2, (3,4))), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A13 (47,48)))), (49,50))

(1, ( ( ( ( ( ( (2,7), (3,4)), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A14 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (7, (3,4))), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A15 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( ( ( (17, (18, (7,19))), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))), (15,16))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A16 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( ( ( (17, (19, (7,18))), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))), (15,16))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A17 (47,48)))), (49,50)) (1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( ( ( (17, (7, (18,19))), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))), (15,16))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A18 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( ( ( (7, (17, (18,19))), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))), (15,16))), (32,33)), ( (34, (35, (36, (37,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A19 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (8, (7,38))), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36,37))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A20 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( (8, (5,6)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36, (37, (7,38))))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A21 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (38, (36,37)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A22 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (37, (36,38)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A23 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, ( (35,38), (36,37))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A24 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (38, (35, (36,37)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A25 (47,48)))), (49,50)) (1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( ( (34,38), (35, (36,37))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A26 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (38, (34, (35, (36,37)))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A27 (47,48)))), (49,50))

(1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), (38, ( (34, (35, (36,37))), ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A28 (47,48))))), (49,50)) (1, ( ( ( ( ( (2, (3,4)), ( ( (5,6), (7,8)), (9, (10,11)))), (12, (13,14))), ( (15,16), ( ( (17, (18,19)), (20, (21,22))), (23, ( ( ( (24,25), (26,27)), (28,29)), (30,31)))))), (32,33)), ( (34, (35, (36,37))), (38, ( ( ( (39, (40, (41,42))), (43,44)), (45,46)), A29 (47,48))))), (49,50))

Table 4. Topologies of trees included in Shimodaira-Hasegawa non-parametric likelihood ratio tests (dataset: higher-level rfelationships). Taxa: 1, Alytes muletensis; 2, Chiromantis rufescens; 3, Rana galamensis; 4, Mantidactylus new species Comoros; 5, Mantidactylus wittei; 6, Boophis tephraeomystax; 7, Boophis new species Comoros; 8, Aglyptodactylus madagascariensis; 9, Laliostoma labrosum; 10, Rana temporalis; 11, Polypedates cruciger; 12, Indirana cf leptodactyla; 13, Nyctibatrachus major; 14, Hoplobatrachus crassus; 15, Hoplobatrachus occipitalis; 16, Ptychadena mascareniensis; 17, Astylosternus diadematus; 18, Leptopelis natalensis; 19, Arthroleptis variabilis; 20, Petropedetes parkeri; 21, Hyperolius viridiflavus; 22, Tachycnemis seychellensis; 23, Heterixalus tricolor; 24, Bufo melanostictus; 25, Leptodactylus fuscus; 26, Nesomantis thomasseti; 27, Pipa parva; 28, Hymenochirus boettgeri; 29, Xenopus laevis; 30, Ambystoma mexicanum

Tree Topology B0 ( (1, ( ( ( ( ( ( ( ( ( (2,11),13), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( ( (17,18),19), (21, (22,23)))), (24,25)),26), ( (27,28),29))),30) B1 (30, (1, ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (26, (24,25))), (29, (27,28))))) B2 (30, (1, (26, ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25)), (29, (27,28)))))) B3 (30, (1, ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25)), (26, (29, (27,28)))))) B4 (30, ( (1,26), ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25)), (29, (27,28))))) B5 (30, (1, ( ( (26, ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23))))), (24,25)), (29, (27,28))))) B6 (30, (1, ( ( ( (26, ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15))), ( (19, (17,18)), (21, (22,23)))), (24,25)), (29, (27,28))))) B7 (30, (1, ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), (26, ( (19, (17,18)), (21, (22,23))))), (24,25)), (29, (27,28))))) B8 (30, (1, ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (23, (22,26))))), (24,25)), (29, (27,28))))) B9 (30, (1, ( ( ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21,23))), (24,25)), (22,26)), (29, (27,28))))) B10 (30, (1, ( (26, ( ( ( ( ( ( (2,11), ( (4,5), ( (6,7), (8,9)))), (13, (3,10))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25))), (29, (27,28))))) B11 (30, (1, ( (26, ( ( ( ( ( (13, ( (2,11), ( (4,5), ( (6,7), (8,9))))), (3,10)), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25))), (29, (27,28))))) B12 (30, (1, ( (26, ( ( ( ( ( ( (13, (2,11)), (3,10)), ( ( (4,5), (8,9)), (6,7))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25))), (29, (27,28))))) B13 (30, (1, ( (26, ( ( ( ( ( ( (13, (2,11)), (3,10)), ( ( (4,5), (6,7)), (8,9))), (12, (16,20))), (14,15)), ( (19, (17,18)), (21, (22,23)))), (24,25))), (29, (27,28))))) B14 (30, (1, ( (26, ( ( ( ( ( ( (13, (2,11)), (3,10)), ( (4,5), ( (6,7), (8,9)))), (12, (16,20))), (14,15)), ( (17,19), (18, (21, (22,23))))), (24,25))), (29, (27,28)))))

Appendix G. Estimates of divergence times

Since our results indicated younger ages of most anuran cladogenetic events within the Anura as compared to previous publications, we followed a multiple conservative approach, to avoid any methodological error that could mask our results. This implied (a) using the oldest available geological dating of the origin of Mayotte for calibration of the split between the Comoroan Boophis and Mantidactylus from their Malagasy sister groups (8.7 MYr ago), (b) usage of an extremely old age estimate of separation between salamanders and frogs (370 million years, the age of the first tetrapod fossils) (15), (c) usage of the oldest age estimate of the breakup of Africa and South America (101 million years) (16) to calibrate the vicariance between the fully freshwater aquatic Hymenochirus (Africa) and Pipa (South America) (17).

We followed two separate approaches to estimate the ages of the relevant splits in our phylogeny: (i) non-parametric rate smoothing using the program r8s (written by M. J. Sanderson) including the calculation of confidence intervals; as input we used a phylogram with the topology obtained by ML searches of the complete (mitochondrial and nuclear) dataset but with branch lengths based on the nuclear gene sequences only. (ii) Regression of pairwise divergences between taxa in the rhodopsin gene, using additional vicariant species pairs from the Mediterranean area for calibration, and calculation of prediction confidence intervals (18).

7.1 Non Parametric Rate Smoothing

Maximum Likelihood (ML) heuristic searches were carried out with PAUP* (6) under the substitution model proposed by Modeltest (4), with a random sequence addition sequence (10 replicates). The obtained tree was saved without branch lengths, and loaded again after excluding of all mitochondrial sequences and constraining to ML settings proposed by MODELTEST for the nuclear dataset alone. The obtained phylogram was saved to a treefile including branch lengths, and submitted to non parametric rate smoothing (19) using the program r8s, written by M. J. Sanderson, on a Pentium III computer with Linux 2.2.18 partition. We preferred NPRS over penalized likelihood (PL) (20) following again a conservative approach, since it usually produces more extended confidence intervals. NPRS calculations were carried out with the Powell algorithm, fixing the age of the root at 370 or 250 MYA,

the ages of the splits between the two Boophis and the two Mantidactylus at 8.7 MYA, and the age of the split between Pipa and Hymenochirus at 101 MYA. Confidence intervals were calculated using a cutoff value of 4.0. However, due to the lack of crossover points, confidence intervals could not be calculated for the young dispersal events but only for the older splits in the cladogram.

Root fixed at 250 MYr ago Root fixed at 370 MYr ago Fixed ages used for calibration Root 250 MYr ago 370 MYr ago Pipidae 101 MYr ago 101 MYr ago Boophis (Mayotte-Madagascar) 8.7 MYr ago 8.7 MYr ago Mantidactylus (Mayotte-Madagascar) 8.7 MYr ago 8.7 MYr ago

Age estimate Hyperolius-Heterixalus 22.05 MYr ago 26.55 Heterixalus-Tachycnemis 11.28 MYr ago 13.57 Hoplobatrachus occipitalis - H. crassus 8.94 MYr ago 10.29 Rana galamensis - R. temporalis 23.03 MYr ago 26.44 Polypedates - Chiromantis 28.37 MYr ago 32.70

Age estimate and confidence intervals Nesomantis - Neobatrachia 150.50 (118.07-180.56) MYr ago 194.40 (127.61-273.17) MYr ago Hyloidea-Ranoidea 137.57 (105.63-168.39) MYr ago 176.16 (111.13-255.81) MYr ago Ranidei-Arthroleptoidei 92.71 (66.73-126.75) MYr ago 113.63 (70.88-113.68) MYr ago Basal splits within Ranidei 62.79 (41.78-96.99) MYr ago 73.34 (44.96-149.52) MYr ago

7.2 Regression analysis of pairwise Rhodopsin divergences

The calibration of the phylogram submitted to NPRS (see above) was based on three different splits that could be geologically dated (two colonizations of the Comoros, and the vicariance split between the African and South American pipids). To confirm these calibrations with independent data, we sequenced homologous rhodopsin fragments in three species pairs of the Mediterranean region which are likely to have been separated by vicariance at the end of the Messininian salinity crisis of the Mediterranean Sea, 5.2-5.3 MYr ago (21, 22): Rana cretensis (Crete) vs. Rana cerigensis (Rhodos), Pelobates cultripes (Spain) vs. Pelobates varaldii (Morocco), Alytes dickhillenii (Spain) vs. Alytes maurus (Morocco). The two Rana species belong to the European green frogs (or water frogs), which have been successfully used to calibrate a protein clock based on geologically dated Mediterranean sea barriers (21), making it highly probable that the assumed age reflects their biogeographic history. The two Pelobates and Alytes species, respectively, are sister species within their genera (22), and their separation is likelily to have occurred at the opening of the strait of Gibraltar.

The 1-2 substitutions which were consistently found between the species of these pairs correspond well to the 5 substitutions between the Comoroan and Malagasy Mantidactylus species, indicating a rhodopsin substitution rate of 0.1-0.3 substitutions (0.03-0.1%) per lineage per MYr. This agrees with the rate calculated using the ancient split between Xenopus and Pipa (0.2 substitutions per lineage per MYr, corresponding to 0.06%). It also agrees with the placement of the diapsid/synapsid split at 310 MYr ago (23) as exemplified by the Homo/Gallus divergence of 62 substitutions (0.1 substitutions or 0.03% per lineage per MYr).

Since the amount of substitutions between the two Rana (three substitutions) and Hoplobatrachus (three substitutions), as well as between the hyperoliids Heterixalus, Tachycnemis and Hyperolius (6-9 substitutions) were in a similar order of magnitude as in the Mediterranean and Comoran taxa used for calibration, it is highly improbable (and could clearly be excluded by 95% confidence intervals) that these divergences are due to ancient (Mesozoic) vicariance. Such vicariance, for the three hyperoliids, would imply a dramatic three to ten-fold decrease of substitution rate (down to 0.01% per lineage per MYr) for which there is no indication in the branch lengths of the phylograms (see section 5).

Calibration Pairwise divergences Age

(# substitutions) (MYr ago)

Pelobates varaldi - P. cultripes 2 5.3 Alytes maurus - A. dickhillenii 1 5.3 Rana cretensis - R. cerigensis 1 5.3 Mantidactylus wittei - M. sp. (Mayotte) 5 8.7 Xenopus - Pipa 37 101

Age estimate Pairwise divergences Age_mean Age_min Age_max (# substitutions) (MYr ago) (MYr ago) (MYr ago) Chiromantis - Polypedates 16 43.1471 36.3026 50.7792 Rana temporalis - R. galamensis 3 7.9301 3.9599 12.6879 Hoplobatrachus occidentalis - H. crassus 3 7.9301 3.9599 12.6879 Tachycnemis - Heterixalus 6 16.0571 11.4236 21.4782 Heterixalus - Hyperolius 9 24.1841 18.8873 30.2685 Nesomantis - Neobatrachia Min 41 110.8721 98.5001 124.0317 Nesomantis - Neobatrachia Max 49 132.5441 118.4033 147.4725 Neobatrachia Min 33 89.2001 78.5969 100.5909 Neobatrachia Max 51 137.9621 123.3791 153.3327 Ranoidea Min 29 78.3641 68.6453 88.8705 Ranoidea Max 42 113.5811 100.988 126.9618 Ranidei Max 33 89.2001 78.5969 100.5909 Mantellidae-Rhacophoridae Min 10 26.8931 21.3752 33.1986 Mantellidae-Rhacophoridae Max 16 43.1471 36.3026 50.7792

Appendix H. A survey of the terrestrial and freshwater vertebrate families of Madagascar, their biogeographic relationships and fossil ages

Vertebrate group Name of Malagasy taxon non-Malagasy sister taxon Distribution of non- Area cladogram First fossil record of family *** Malagasy sister taxon agreement? (or higher inclusive taxon) Osteichthyes: Cichlidae Paretroplus Etroplus (24) India-Sri Lanka (24) Yes (24, 54) Eocene (25) Osteichthyes: Aplocheiloidea Pachypanchax Oryzias (26) India-Sri Lanka (26) Yes (26) Cyprinodontiformes: Eocene (27, 28)

Amphibia: Hyperoliidae Heterixalus Hyperolius - Afrixalus (3) Africa (3) No (3) Ranoidea: Cretaceous (Cenomanian, 90-97 MYr ago) (29)** Amphibia: Ranoidei Mantellidae Rhacophoridae? (1) India-Asia? (1) Uncertain Ranoidea: Cretaceous (possibly yes) (Cenomanian, 90-97 MYr ago) (29)** Amphibia: Microhylidae Microhylidae: Cophylinae, Microhylidae gen. Unknown Unknown Eocene (29) Dyscophinae, Scaphiophryninae Squamata: Typhlopidae Typhlops spp. Unknown Unknown Unknown Miocene (15) Squamata: Boidae Sanzinia, Acrantophis Boinae gen. (30) South America (30) No (30) Paleocene (31) Squamata: Colubridae Mimophis spp. Psammophiinae* Africa* No* Oligocene (15) Squamata: Chamaeleonidae Calumma, Furcifer Chamaeleo (32, 33) Africa (32, 33) No (32, 33) Paleocene (15) Squamata: Scincidae Mabuya spp. Mabuya spp. (34) Africa (34) No (34) possibly Cretaceous/Jurassic? (15) Squamata: Scincidae Scincinae gen. Scincinae gen.* Africa* No* possibly Cretaceous/Jurassic? (15) Squamata: Gekkonidae Lygodactylus spp. Lygodactylus spp. (35)* Africa (35)* No (35)* Eocene (36) Squamata: Gekkonidae Phelsuma spp. Rhoptropella ocellata (35)* Africa (35),* No (35)* Eocene (36) Squamata: Gekkonidae Blaesodactylus spp. Homopholis spp. (35) Africa (35) No (35) Eocene (36) Squamata: Gerrhosauridae Zonosaurus spp. Gerrhosauridae gen. (37, 38) Africa (37, 38) No (37, 38) Cordylidae s. l. : Eocene (15), maybe Latest Cretaceous? (39) Squamata: Opluridae Opluridae spp. Tropiduridae? South America (40) No (40) Iguanidae s. l.: upper Cretaceous (15, 53) (Iguanidae s.l.) (40) Crocodylia: Crocodylidae Crocodylus niloticus Crocodylus niloticus* Africa* No* Upper Cretaceous (15) Testudines: Podocnemidae Erymnochelys Podocnemis spp. (41, 42) South America (41, 42) No (41, 42) Eocene (41) madagascariensis Testudines: Testudinidae Geochelone spp., Kinixys Geochelone spp. (43) Africa (43) No (43) Cretaceous (Neocomian, 127-144 MYr ago) (44) Mammalia: Rodentia Nesomyinae gen. Muridae gen. (45, 46) Africa (45, 46) No (45, 46) Paleocene (47) Mammalia: Lemuriformes Lemuroidea gen. Lorisidae gen. (48, 49) Africa and Asia (48, 49) No (48, 49) Oligocene (50) Mammalia: Carnivora Herpestidae gen. Herpestidae gen. (47) largely Africa (47) Probably no Paleocene (47) Mammalia: Afrosoricida Tenrecidae gen. Potamogale spp. (51) Africa (51, 52) No (51, 52) Miocene (47)

* According to own observations and unpublished data of A. Schmitz (Bonn)

** Information of ranid fossils of the African Cenomanian have been published (29) but not adequately described so far; in the absence of confirmation, and in the light o recently proposed new classificatons (52), we doubt on the reliability of this familial assignation and refer the findings to the Ranoidea in a preliminary way. *** Seychellean taxa not considered

Appendix I. References used in electronic appendices

1. Bossuyt, F., & Milinkovitch, M. C. (2000) Proc. Natl. Acad. Sci. USA 97, 6585-6590.

2. Palumbi, S. R. et al., (1991) The Simple Fool's Guide to PCR, Version 2.0. (Privately published, Univ. Hawaii).

3. Vences, M., Kosuch, J., Glaw, F., Böhme, W. & Veith, M. (in press) J. Zool. Syst. Evol. Res..

4. Posada, D. & Crandall, K. A. (1998) Bioinfomatics 14, 817-818.

5. Huelsenbeck, J. P. & Crandall, K. A. (1997) Ann. Rev. Ecol. Syst. 28, 437-466.

6. Swofford, D. L. (2002) PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods) (Sinauer, Sunderland, MA), Version 4b10.

7. Kjer, K. M. (1995) Mol. Phylogenet. Evol. 4, 314-330.

8. Graybeal, A. (1993) Mol. Phylogenet. Evol. 2, 256-269.

9. Huelsenbeck, J. P. & Ronquist, F. (2001) Bioinformatics 17, 754-755.

10. Müller, T. & Vingron, M. (2000) J. Comp. Biol. 7, 761-776.

11. Schmidt, H. A., Strimmer, K., Vingron, M. & von Haeseler, A. (2000) Tree-Puzzle, version 5.0. (Heidelberg).

12. Kumar, S., Tamura, K., Jakobsen, I. B. & Nei, M. (2001) Bioinformatics 17, 1244-1245.

13. Shimodaira, H. & Hasegawa, M. (1999) Mol. Biol. Evol. 16, 1114-1116.

14. Whelan, S., Liò, P. & Goldman, N. (2001) Trends in Genetics 17, 262-272.

15. Carroll, R. L. (1988) Vertebrate Paleontology and Evolution (Freeman, New York).

16. Pitman III, W. C., Cande, S., LaBrecque, J. & Pindell, J. (1993) in Biological relationships between Africa and South America, ed. Goldblatt, P. (Yale Univ. Press, New Haven), pp. 15-34.

17. Duellman, W. E. & Trueb, L. (1986) Biology of Amphibians (McGraw-Hill, New York).

18. Hillis, D. M., Mable, B. K. & Moritz, C. (1996) in Molecular Systematics, 2nd Edition, eds. Hillis, D. M., Moritz, C. & Mable, B. K. (Sinauer, Sunderland, MA), pp. 515-543.

19. Sanderson, M. J. (1997) Mol. Biol. Evol. 14, 1218-1231.

20. Sanderson, M. J. (2002) Mol. Biol. Evol. 19, 101-109.

21. Beerli, P., Hotz, H. & Uzzell, T. (1996) Evolution 50, 1676-1687.

22. Fromhage, L., Veith, M. & Vences M. (submitted).

23. Kumar, S. & Hedges, S. B. (1998) Nature 392, 917-920.

24. Farias, I. P., Orti, G., Sampaio, I., Schneider, H. & Meyer, A. (2001) J. Mol. Evol. 53, 89-103.

25. Murray, A. M. (2001) Biol. J. Linn. Soc. 74, 517-532.

26. Murphy, W. J. & Collier, G. E. (1997) Mol. Biol. Evol. 14, 790-799.

27. Lundberg, J. G. (1993) in Biological relationships between Africa and South America, ed. Goldblatt, P. (Yale Univ. Press, New York), pp. 156-199.

28. Berg, L. S. (1958) System der rezenten und fossilen Fischartigen und Fische (Verlag der Wissenschaften, Berlin).

29. Sanchiz, B. (1998) Encyclopedia of Palaeoherpetology, Part 4. Salientia (Pfeil, München).

30. Vences, M., Glaw, F., Kosuch, J., Böhme, W. & Veith, M. (2001) Copeia 2001: 1151-1154.

31. Rage, J.-C. (2001) Palaeovertebrata 30: 111-150.

32. Raxworthy, C. J., Forstner, M. R. J. & Nussbaum, R. A. (2002) Nature 415, 784-787.

33. Townsend, T. & Larson, A. (2002) Mol. Phylogenet. Evol. 23, 22-36.

34. Mausfeld, P., Vences, M., Schmitz, A. & Veith, M. (2000) Mol. Phylogenet. Evol. 17, 11-14.

35. Kluge, A. G. & Nussbaum, R. A. (1995) Misc. Publ. Mus. Zool. Univ. Michigan 183, 1-20.

36. Rösler, H. (2000) Gekkota 2, 28-153.

37. Odierna, G., Canapa, A., Andreone, F., Aprea, G., Barucca, M., Capriglione, T. & Olmo, E. (2002) Mol. Phylogenet. Evol. 23, 37-42.

38. Lang, M. (1991) Bull. Inst. Roy. Sci. Nat. Belg. 61, 121-188.

39. Krause, D. W., Hartman, J. H. & Wells, N. A. (1997) in Natural Change and Human Impact in Madagascar, eds. Goodman, S. M. & Patterson, B. D. (Smithsonian Inst. Press, Washington), pp. 3-43.

40. Schulte II, J. A., Macey, J. R., Larson, A. & Papenfuss, T. J. (1998) Mol. Phylogenet. Evol. 10, 367-376.

41. Noonan, B. P. (2001) J. Biogeogr. 27, 1245-1249.

42. Georges, A., Birrelli, J., Saint, K. M., McCord, W. & Donnellan, S. C. (1998) Biol. J. Linn. Soc. 67, 213-246.

43. Caccone, A., Amato, G., Gratry, O. C., Behler, J. & Powell, J. R. (1999) Mol. Phylogenet. Evol. 12, 1-9.

44. Hirayama, R., Brinkman, D. B. & Danilov, I. G. (2000) Russ. J. Herpetol. 7, 181-198.

45. Jansa, S. A., Goodman, S. M. & Tucker, P. K. (1999) Cladistics 15, 253-270.

46. J.-Y. Dubois, D. Rakotondravony, C. Hänni, P. Sourrouille, F. Catzeflis (1996) J. Mammal. Evol. 3, 239-260.

47. McKenna, M. C. & Bell S. K. (1997) Classification of Mammals above the Species Level (Columbia Univ. Press, New York).

48. Yoder, A. D. (1996) in Biogeography of Madagascar, ed. Lourenço W. R. (Orstom, Paris), pp. 245-258.

49. Yoder A. D., Irwin, J. A. & Payseur, B. A. (2001) Syst. Biol. 50, 408-424.

50. Marivaux, L., Welcomme, J. L., Antoine, P.-O., Metais, G., Baloch, I. M., Benammi, M., Chaimanee, Y., Ducrocq, S. & Jaeger, J.-J. (2001) Science 294: 587-591.

51. Douady, C. J., Catzeflis, F., Kao, D. J., Springer, M. S. & Stanhope, M. J. (2002) Mol. Phylogenet. Evol. 22, 357-363.

52. Murphy, W. J. et al. (2001) Science 294, 2348-2351.

53. Rage, J. C. (1996) in Biogeography of Madagascar, ed. Lourenço, W. R. (Orstom, Paris), pp. 27-35.

54. Vences, M., Freyhof, J., Sonnenberg, R., Kosuch, J. & Veith, M. (2001) J. Biogeogr. 28, 1091-1099.