Modeling body size evolution in Felidae under alternative phylogenetic hypotheses.

0.59 MB·English

by  Diniz-Filho, Jose Alexandre Felizola

GeneticsandMolecularBiology,32,1,170-176(2009)
Copyright©2009,SociedadeBrasileiradeGenética.PrintedinBrazil
www.sbg.org.br
ResearchArticle
Modeling body size evolution in Felidae under alternative phylogenetic
hypotheses
JoséAlexandreFelizolaDiniz-Filho1andJoãoCarlosNabout2
1DepartamentodeBiologiaGeral,InstitutodeCiênciasBiológicas,UniversidadeFederaldeGoiás,
Goiânia,GO,Brazil.
2ProgramadeDoutoradoemCiênciasAmbientais,UniversidadeFederaldeGoiás,Goiânia,GO,Brazil.
Abstract
Theuseofphylogeneticcomparativemethodsinecologicalresearchhasadvancedduringthelasttwentyyears,
mainlyduetoaccuratephylogeneticreconstructionsbasedonmoleculardataandcomputationalandstatisticalad-
vances. We used phylogenetic correlograms and phylogenetic eigenvector regression (PVR) to model body size
evolutionin35worldwideFelidae(Mammalia,Carnivora)speciesusingtwoalternativephylogeniesandpublished
bodysizedata.Thepurposewasnottocontrastthephylogenetichypothesesbuttoevaluatehowanalysesofbody
sizeevolutionpatternscanbeaffectedbythephylogenyusedforcomparativeanalyses(CA).Bothphylogeniespro-
duced a strong phylogenetic pattern, with closely related species having similar body sizes and the similarity de-
creasingwithincreasingdistancesintime.ThePVRexplained65%to67%ofbodysizevariationandallMoran’sI
valuesforthePVRresidualswerenon-significant,indicatingthatboththesemodelsexplainedphylogeneticstruc-
turesintraitvariation.EventhoughourresultsdidnotsuggestthatanyphylogenycanbeusedforCAwiththesame
power,orthat“good”phylogeniesareunnecessaryforthecorrectinterpretationoftheevolutionarydynamicsofeco-
logical,biogeographical,physiologicalorbehavioralpatterns,itdoessuggestthatdevelopmentsinCAcan,andin-
deed should, proceed without waiting for perfect and fully resolved phylogenies.
Keywords:autocorrelation,bodysize,Felidae,phylogeneticeigenvectorregression,phylogenies.
Received:December12,2007;Accepted:May5,2008.
Introduction al. 2005). Furthermore, some studies have suggested that
incorporatingphylogeneticstructureintodataanalysesal-
Phylogenetic comparative methods developed since
lowed a better understanding of the processes underlying
the middle 1980’s are now commonly applied in areas of
ecological,behavioralandphysiologicaldata(Hansenand
biological research, such as ecology, physiology and be-
Martins,1996;Diniz-Filho,2001).
havior,toexplainhowphylogeneticpatternsinthetraitsof
Mainlybecauseoftheimprovementsandpopulariza-
species can be associated to adaptive evolution (Martins,
tion of DNA sequence techniques and other molecular
2000; Freckleton et al., 2002). It is also well-known that
markers,therehasbeenamarkedincreaseinthenumberof
species,orothertaxonomicunits,maynotrepresentinde-
availablephylogeniesthatcanbeusedasabasisincompar-
pendentobservationsforstatisticalanalysessuchasregres-
ative analysis (Pagel, 1999; see Felsenstein, 2004). More
sion and correlation (Felsenstein, 1985; Martins and
importantlyforcomparativeanalyses,therearenowtech-
Garland,1991;Martinsetal.,2002).Hence,manydifferent
niques that combine phylogenies from different sources
formsofcomparativeanalyseshavebeendevelopedtotake
andbasedondifferentdatatypes,suchasmorphology,mo-
intoaccountthelackofindependenceamongspeciesdueto
lecular data or behavior, to generate complete, or nearly
phylogenetic relationships (i.e., phylogenetic autocorre-
complete, trees for very large taxonomic groups (super-
lation)andtocorrectlyapproximatetheTypeIerrorsofsta-
trees)(Bininda-Emonds,2004).
tistical analyses of correlation between traits or between
The first supertree was built for Primates (Purvis,
traits and other components of environmental variation
1995)andanearlycompletesupertreeformorethan95%
(Martins and Garland, 1991; Gittleman and Luh, 1992;
ofcurrentmammalspecieshasrecentlybeenpublishedby
MartinsandHansen,1996;Martinsetal.,2002;Garlandet
Bininda-Emondsetal.(2007).However,thefirstcomplete
supertreeforalllivingspeciesinalargetaxawasgenerated
SendcorrespondencetoJoséAlexandreFelizolaDinizFilho.De-
forworldwideCarnivora(Bininda-Emondsetal.1999)and
partamento de Biologia Geral, Instituto de Ciências Biológicas,
notonlyencompassedall271livingCarnivoraspeciesbut
Universidade Federal de Goiás, Caixa Postal 131, 74001-970
Goiânia, GO, Brazil. E-mail: [email protected]. was comparatively well resolved with a relatively small
Diniz-FilhoandNabout 171
number of polytomies for most subclades and has been have some important differences (Figure 1), with, for in-
widelyusedincomparativeanalyses(Sechrestetal.,2002.; stance, the first divergence of felids occurring at 10.78
Diniz-Filho and Tôrres, 2002; Tôrres and Diniz-Filho, millionyears(My)intheJNphylogenyandat16.2Myin
2004; Diniz-Filho, 2004; Diniz-Filho et al., 2007). More theSTphylogeny.Morerelevant,however,werethediffer-
recently, however, Johnson et al. (2006) proposed a fully encesintreetopology,withtheSTphylogenyhavingmany
resolved felid phylogeny derived from 22,789 base pairs polytomies,andthetwophylogeniesalsohavingsomedif-
from autosomal, X-linked, Y-linked and mitochondrial ferencesinthecladesstructure(Figure1),with,forexam-
genes, with many important differences in relation to the ple,PatherabeingbasaltoothercladesintheJNphylog-
widelyusedsupertree. eny,whichhelpstoexplainsomeofthedifferencesfound
Itisimportanttoconsiderthat,despiteofthefactthat laterinthecomparativeanalyses.
well-resolved phylogenies are the core of comparative Moran’sIisthemostcommonlyusedcoefficientin
analyses, there are still many uncertainties regarding the spatial and phylogenetic autocorrelation analyses and al-
methods and data needed to reconstruct phylogenies and lows the description of the phylogenetic structure of the
supertrees(Webbetal.,2002).Thus,despitetheincreasing data (or model residuals) and the comparison of the ob-
useofcomparativemethods(Carvalhoetal.,2005),itisal- servedvalueswiththeexpectedvaluesbasedondifferent
waysimportanttounderstandhowmodelingtraitevolution evolutionary models (Gittleman and Kot, 1990; Diniz-Fi-
from comparative methods are sensitive to errors and un- lho,2001;Diniz-FilhoandTôrres,2002).MoranMoran’sI
certainties regarding tree topology (Martins and Garland, isgivenby
1991; Martins, 1996; Martins and Housworth, 2002) and (cid:10)(cid:16)(cid:16)(y (cid:9)y)(y (cid:9)y)w (cid:13)
other related problems, such as taxon sampling (Ackerly, (cid:3)n(cid:6)(cid:12) i j ij(cid:15)
2000).Itis,therefore,ofparamountimportancetoevaluate I (cid:2)(cid:4)(cid:5)S (cid:7)(cid:8)(cid:12) i j (cid:16)(y (cid:9)y)2 (cid:15)
howerrorsinphylogenyreconstructionmayaffectthere- (cid:11)(cid:12) i (cid:14)(cid:15)
i
sultsofcomparativeanalyses,aswhenmodelingtraitevo-
lution.Inthispaperweusedphylogeneticcorrelogramsand wherenisthenumberofobservation(species),y andy are
i j
phylogeneticeigenvectorregression(PVR)tomodelbody thebodysizevaluesinthespeciesiandj, yistheaverageof
size evolution in worldwide Felidae (Mammalia, Carni- yandw isanelementofthematrixW.Actually,several
ij
vora) under alternative phylogenies. Our purpose was not Moran’sIarecalculatedforthesamevariable,bycreating
toestablishthevalidityofanyofthephylogenies,butrather successiveWmatricesinwhichw =1ifthepairi,jofspe-
ij
toevaluatehowanalysesofpatternsinbodysizeevolution ciesiswithinagivenphylogeneticdistanceclassinterval,
can be affected by choosing one of them as the basis for establishedbasedonD(indicatingthatspeciesare“linked”
comparativeanalyses. inthisclass),andw =0otherwise.Inthispaper,4classes
ij
withequalintervals(timeslices)wereused,andSindicates
Material and Methods thenumberofentries(connections)ineachWmatrix.The
valueexpectedunderthenullhypothesisoftheabsenceof
Body masses (in g) for 35 Felidae species (Table 1)
spatialautocorrelationis-1/(n-1).Moran’sIusuallyvary
distributedworldwidewereobtainedmainlyfromSmithet
between -1.0 for maximum negative autocorrelation and
al.(2003).Valuesweretransformedtodecimallogarithms
1.0 for maximum positive autocorrelation. Computation
priortotheanalysis.Becauseweweremainlyinterestedin
details of the standard error of this coefficient (se(I)) are
broad-scalecomparativepatterns,bodysizesofmalesand
giveninLegendreandLegendre(1998),sothatMoran’sI
females were averaged for each species. Also, although
canbetestedbyassuminganormaldistributionofse(I).In
sexual dimorphisms exists for the group, the relatively
thiscaseastandardnormaldeviate(SND={I-E(I)}/se(I))
magnitude of male-female differences when compared to
largerthan1.96indicatesthatse(I)issignificantatthe5%
interspecificdifferencesacrosstheentirecladeissosmall
probabilitylevel(p=0.05).TheBonferronicorrectioncan
that this averaging is not likely to affect the conclusions
be used for conservative statistical decisions, so that a
reachedinthispaper.
correlogramwillbeconsideredsignificantatp<0.05only
We used the phylogenies proposed by Bininda-
if one of Moran’s I is significant at p = 0.05/4 distance
Emonds et al. (1999), designated the supertree (ST) phy-
classes.
logeny,andJohnsonetal.(2006),designatedtheJohnson
Itisalsopossibletouseadirectmodelingstrategyto
(JN) phylogeny, to establish pairwise phylogenetic dis-
describepatternsofbodysizeevolution.Diniz-Filhoetal.
tancematricesD,whichwerethebasisformodelingbody
(1998) developed a new technique called phylogenetic
sizeevolutionusingMoran’sIphylogeneticcorrelograms
eigenvectorregression(PVR)topartitionthetotalvariance
(Gittleman and Kot, 1990; Diniz-Filho, 2001) and phylo-
genetic eigenvector regression (Diniz-Filho et al., 1998). ((cid:17)2T)ofatraity(e.g.,bodysize)measuredinasetofspe-
Forallanalysis,thespeciesCatopumabadia,Feliscatus, cies into phylogenetic ((cid:17)2 ) and unique variances or eco-
P
FelisilbycaandPardofelisbadiawereeliminatedbecause logical((cid:17)2 )variances,suchthat(cid:17)2 =(cid:17)2 +(cid:17)2 .Theidea
S T P S
theywereabsentinbothphylogenies.Thetwophylogenies beingthataphylogenycanbeexpressedasasetoforthogo-
172 Felinephylogeniesandphylogeneticanalysis
Table1-FelidspeciesfortheSupertree(ST)*andJohnson(JN)†phylogenies.Thescientificnamesaremostlythesameinboththephylogenies,theex-
ceptionsbeingthatthegenusnamesinparenthesesareasgivenintheJNphylogeny.Commonnamesandroundedmeanbodymassvaluesarealsogiven.
Code Scientificname Commonname Approximatemeanbodymass(kg)-
1 Pantheraleo Lion 159
2 Pantherapardus Leopard 52
3 Pantheraonca Jaguar 85
4 Pantheratigris Tiger 163
5 Unciauncia Snowleopard 33
6 Neofelisnebulosa Cloudedleopard 15
7 Pardofelismarmorata Marbledcat 3
8 Lynxcanadensis Canadianlynx 10
9 Lynxlynx Eurasianlynx 19
10 Lynxpardinus Iberian,orSpanish,lynx 11
11 Lynxrufus Bobcat 27
12 Catopuma(Pardofelis)temminckii Asiangoldencat 8
13 Profelis(Caracal)aurata Africangoldencat 11
14 Leopardustigrinus Oncillaorlittlespottedcat 2
15 Oncifelisgeoffroyi Geoffroy’scat 3
16 Oncifelisguigna Kodkod 3
17 Oncifeliscolocolo Pampascat 4
18 Oreailurusjacobita Andeanmountaincat 8
19 Leoparduspardalis Ocelotorpaintedleopard 12
20 Leoparduswiedii Margay 3
21 Felismargarita Sandcat 3
22 Felisnigripes Black-footedcat 1
23 Felissilvestris Wildcat 5
24 Felisbieti Chinesemountaincat 6
25 Felischaus Junglecat 7
26 Otocolobusmanul Pallas,orsteppe,cat 3
27 Caracalcaracal African,orPersian,lynx 12
28 Leptailurus(Caracal)serval Serval 12
29 Prionailurusbengalensis Leopardcat 3
30 Prionailurusviverrinus Fishingcat 9
31 Prionailurusrubiginosus Rustyspottedcat 1
32 Prionailurusplaniceps Flat-headedcat 4
33 Herpailurus(Puma)yagouaroundi Jaguarundicat 7
34 Pumaconcolor Cougar 54
35 Acinonyxjubatus Cheetah 51
*Bininda-Emondsetal.(1999).†Johnsonetal.(2006).-MainlyfromSmithetal.(2003).
nalvectorsobtainedbyaneigenanalysisofaphylogenetic take into account a different number of predictors, of the
distancematrixD.Thesevectorscanthenbeusedaspre- multiple regression model of the trait y against the eigen-
dictors of y in any form of linear or non-linear modeling. vectorsinVprovidesanestimateofthephylogeneticsignal
For analogous applications in a spatial context see Diniz- inthedata((cid:17)2 /(cid:17)2 ).
P T
FilhoandBini(2005)andGriffithandPeres-Neto(2006). There are different ways to establish how many
Thus,PVRfollowsthestandardframeworkofgenerallin- eigenvectorsofDshouldbeusedformodeling.Inthework
earmodels,suchthat describedinthepresentpaperweusedanexhaustivesearch
strategybasedonAkaikeInformationCriterion(AIC)de-
y=V(cid:18)+(cid:19)
scribedbyBurnhamandAnderson(2002).Whenperform-
ing the PVR, the AIC value of a particular model (i.e.,
whereVistheorthogonaleigenvectorsextractedfromthe
basedonsomecombinationofeigenvectors)wasgivenby
double-centered phylogenetic distance matrix D, and (cid:18) is
thepartialregressioncoefficients.TheR2value,adjustedto AIC=nln((cid:17)2 )+2K
S
Diniz-FilhoandNabout 173
alternative models (i.e., 2m model minus the model with
interceptonly),allowingtheestablishmentofthemostpar-
simoniousmodelforexplainingbodysizevariationbased
on phylogenetic eigenvectors. Moran’s I correlograms of
thesePVRresidualswerealsoappliedtotestifthemodel
was effective in taking phylogenetic autocorrelation into
account.
Results
The correlograms based on the two phylogenies
showed slightly different patterns, although there was a
generaltrendofpositiveautocorrelationinthefirsttwodis-
tanceclasses,withalargervalueinthesecondclassdueto
themorerecenttimeslices,coupledwithanegativeauto-
correlation in the last distance class (Figure 2; Table 2)
(Diniz-Filho and Tôrres, 2002). In both phylogenies,
closelyrelatedspecieshavesimilarbodysizesandthissim-
ilaritydecreaseswhenincreasingdistancesintimearecon-
sidered, stabilizing in the last distance class (Table 2).
However,despitetheexistenceofamonotonicrelationship
betweenthetwocorrelogramswhenusingtheJNtreethere
was a much lower Moran’s I in the first distance classes,
suggesting that there is some relatively high dissimilarity
betweenmorecloselyrelatedspecies.
Theproportionofvariationexplainedbythephylog-
enyobtainedfromPVRissimilarwhenbasedonthetwo
phylogenies, despite the fact that the correlation between
thefirsteigenvectorsextractedfromthesematricesisrela-
tively low (0.415). When based on the ST phylogeny, a
model with 6 eigenvectors (summing 81.6% of the varia-
tion in original distance matrix) was selected by the AIC
criterionastheminimummodel,withR2=65.1%ascom-
pared with R2 = 69.1% for the full model with 10 eigen-
vectors. On the other hand, only three eigenvectors
(explaining 45% of the total distances) were selected by
Figure1-Phylogeniesusedinthisstudyshowingtherelationshipsbe- AICcriteriontoexplainbodysizevariationbasedontheJN
tween the 35 felid species included in the analysis. In (A) the ST phylogeny, and they explained 67.5% of the variation in
(Bininda-Emonds et al. (1999). Biol Rev 74:143-175) and (B) the JN bodysizeasagainstR2=70.6%forthefullmodelwith10
(†Johnsonetal.(2006).Science311:73-77)phylogeniesthenumbersre-
eigenvectors. In both cases, all Moran’s I values for the
latetothespeciesshowninTable1.
PVR residuals were non-significant, indicating that these
models were sufficient to explain phylogenetic structures
where(cid:17)2Sisthevarianceofthepreviouslydefinedspecific intraitvariation.
component(PVRresiduals)andKisthenumberofparame-
ters(numberofeigenvectors,plustheinterceptandresidual Discussion
variance(cid:17)2 ).Thevalueof(cid:17)2 canbeusedasaproxyfor
S S
Itisalwaysimportanttocheckhowresultsfromcom-
thelikelihoodofthemodelgiventhedata,butthisapproxi-
parativeanalysesoftraitevolutionareaffectedbyerrorsor
mationisvalidonlyifmodelerrorsareindependent,nor-
uncertaintiesinthephylogeny(MartinsandGarland,1991;
mallydistributedandwithconstantvariance(Burnhamand
Martins, 1996), when conflicting or alternative phylogen-
Anderson,2002).Thebestmodelsarethosewiththelowest
etichypothesesareavailableforthegroupoforganismsun-
AICvaluesbutmultiplePVRmodelscanbegenerated,in
der scrutiny. Since the number of available phylogenies
whichcaseanewspatialanalysisinmacroecology(SAM) (includingsupertrees)israpidlyincreasingintheliterature,
module software (Rangel et al., 2006; Diniz-Filho et al. the simple solution is frequently to model trait evolution
2008)wasusedtogenerateallpossiblecombinationofthe underthesemultiplephylogeniesandcomparetheoutcome
first10eigenvectorsfromDaspredictors,producing1023 ofcomparativeanalyses,asperformedhere.
174 Felinephylogeniesandphylogeneticanalysis
Table2-Moran’sI,standarderror(SE)andstandardnormaldeviate(SND)obtainedforbodysizeandthephylogeneticeigenvectorregression(PVR)re-
sidualforeachphylogenyusedinthisstudy.
Supertree(ST)phylogeny Johnson(JN)phylogeny
Moran’sI SE SND Moran’sI SE SND
Bodysize
1 0.840 0.142 6.140*** 0.550 0.209 2.770**
2 0.882 0.253 3.605** 0.979 0.153 6.597***
3 0.046 0.092 0.820 0.093 0.036 3.436**
4 -0.144 0.020 -5.882*** -0.292 0.027 -9.753
PVRresiduals
1 -0.196 0.143 -1.170 -0.013 0.211 0.078
2 -0.251 0.255 -0.868 0.221 0.155 1.618
3 -0.066 0.093 -0.391 -0.094 0.036 -1.817
4 -0.003 0.020 1.361 -0.011 0.027 0.682
**p<0.01;***p<0.001.
Our analyses show that, despite the differences be- Indeed, many papers have reported relatively strong
tweenthetwophylogeniestested,phylogeneticautocorre- phylogenetic patterns of body size variation between spe-
lation patterns in body size of world felids revealed by cies,sothisisprobablyduetorealprocessesthatgenerate
correlogramsandPVRwerequalitativelysimilar.Thissug- strong phylogenetic inertia in this trait, almost independ-
gests that previous general inferences about macroecolo- entlyoftaxonomicscaleandresolution.Hence,changesin
gical and macroevolutionary patterns using the Bininda- phylogenyareunlikelytoaffecttheestimateofthispattern
Emondsetal.(1999)supertreemayberobusttochangesin toomuch,atleastusingstatistically-basedtechniquessuch
phylogeny,atleastintermsofrelativemagnitudeofphylo- asPVRandcorrelograms.Somemoregeneralstudieshave
genetic patterns. For example, Diniz-Filho and Tôrres shownconsistentpatternsregardingthephylogeneticstruc-
(2002)showedstrongpatternsinbodysizeforNewWorld turingofdifferenttraitswith,forexample,morphological
Carnivora,usingphylogeneticeigenvectorregressionanda traits usually being more phylogenetically structured than
significant correlation between body size and geographic behavioral or ecological traits (Freckleton et al., 2002;
rangesize,whereasDiniz-Filhoetal.(2007)recentlyused Morales,2000;seealsoCarvalhoetal.,2005).Sincethese
PVR to decouple adaptive and historical components of observationswereundertakeninameta-analyticalcontext,
Bergmann`srule(i.e.,increaseinbodysizetowardshigher involving many different clades from different regions of
latitudes) in European Carnivora. The correlogram shape theworld,theyareunliketohavebeenstronglyaffectedby
andPVRresultsobtainedinthesestudiesfordifferentsub- changesintheparticularphylogeniesusedintheanalyses.
setsofCarnivoraarequalitativelysimilartothethoseob-
Itisdifficulttogeneralizetheresultsdescribedinthis
tainedbyusforfelidsalone.
paper, although it is possible to evaluate why statistical
methodsusedherewouldworkfineandprovidesimilarre-
sultsunderalternativephylogenies.Becauseeigenanalysis
generates a hierarchical structure of representation of
phylogenetic patterns of relationship from the root to the
tipsofthephylogeny,PVRisprobablylesssensitivetoer-
rorsinphylogenythatpresumablytendtooccurinmorere-
cent branches (Diniz-Filho et al., 1998). If discussions
aboutthevalidityoftherelationshiparemainlyfocusedon
recentnodeswithlessclearsignals,therewillbereasonable
estimates using different trees if trait evolution is station-
ary.Evenso,inouranalysesthereweresomeimportantdif-
ferencesindeepbranches(suchastherelationshipbetween
Panthera and other subclades), although the correlogram
suggeststhat,forbothphylogenies,covarianceinbodysize
has been stronger in recent times, probably due to non-
linearcomponents(stabilizingselection)involvedinbody
Figure2-Phylogeneticcorrelogramsforfelidbodysizevariationbased
onthephylogeniesshowninFigure1. sizeevolution(Diniz-Filho,2004).Correlogramsmayalso
Diniz-FilhoandNabout 175
benotverysensitivewhenusinglargedistanceclassesbe- Diniz-FilhoJAF,Sant’AnaCERandBiniLM(1998)Aneigen-
tweentimeslices,asusedhere,becauseofthesmallnum- vector method for estimating phylogenetic inertia. Evolu-
berofspecies.Eveniftherearedifferentphylogenies,itis tion52:1247-1262.
Diniz-FilhoJAF(2001)Phylogeneticautocorrelationunderdis-
likelythatmostofthepairwisecomparisonsfallwithinthe
tinctevolutionaryprocesses.Evolution55:1104-1109.
same distance classes, creating similarity between corre-
Diniz-FilhoJAFandTôrresNM(2002)Phylogeneticcompara-
lograms.
tivemethodsandthegeographicrangesize-Bodysizerela-
However, other patterns detected from phylogenies
tionship in new world terrestrial carnivora. Evol Ecol
may be more sensitive to errors and uncertainties and 16:351-367.
shouldbetestedinfuturestudies.Forinstance,estimation Diniz-FilhoJAF(2004)Phylogeneticdiversityandconservation
of phylogenetic diversity, based on summing branch priorities under distinct models of phenotypic evolution.
lengths or the ages of the most recent common ancestors ConservBiol18:698-704.
(Sechrest et al., 2002; Diniz-Filho, 2004; Tôrres and Diniz-FilhoJAFandBiniLM(2005)Modellinggeographicalpat-
Diniz-Filho, 2004) may be more sensitive to changes in ternsinspeciesrichnessusingeigenvector-basedspatialfil-
ters.GlobEcolBiogeogr14:177-185.
phylogenysinceassemblagecompositionswillbeconsid-
Diniz-FilhoJAF,BiniLM,RodriguezMA,RangelTFLVBand
ered independently (reducing sample sizes) so that errors
Hawkins BA (2007) Seeing the forest for the trees: Parti-
wouldbemagnifiedatthesesmallergeographicalscales.
tioning ecological and phylogenetic components of berg-
It is worth noting that our main conclusions do not
mannsruleinEuropeancarnivora.Ecography30:598-608.
meanthatanyphylogeny,independentlyoftheresolution,
Diniz-FilhoJAF,RangelTFLVBandBiniLM(2008)Modelse-
canbeusedincomparativeanalysiswiththesamepower, lection and information theory in geographical ecology.
or that “good” phylogenies are not necessary to give the GlobEcolBiogeogr17:479-488.
correctinterpretationstotheevolutionarydynamicsofeco- Felsenstein J (1985) Phylogenies and the comparative method.
logical, biogeographical, physiological or behavioral pat- AmNat125:1-15.
terns.However,resultssuchasthosepresentedinthispaper Felsenstein J (2004) Inferring Phylogenies. Sinauer Associates,
supporttheviewthatdevelopmentsincomparativeanaly- Sunderland,663pp.
sis in many fields of biology can, and indeed must, occur Freckleton RP, Harvey PH and Pagel M (2002) Phylogenetic
analysisandcomparativedata:atestandreviewofevidence.
withinthecontextofdealingwithless-than-perfectornot
AmNat160:712-726.
fullyresolvedphylogenies.
GarlandT,BennettAFandRezendeEL(2005)Phylogeneticap-
proachesincomparativephysiology.JExpBiol208:3015-
Acknowledgments
3035.
GittlemanJLandKotM(1990)Adaptation:Statisticsandanull
WorkbyJAFDFhasbeencontinuouslysupportedby
model for estimating phylogenetic effects. Syst Zool
productivityfellowshipsfromtheBrazilianAgencyCNPq
39:227-241.
andtheFundaçãodeAmparoàPesquisadaUniversidade
Gittleman JL and Luh HK (1992) On comparing comparative
Federal de Goiás (FUNAPE/UFG). Work by JCN is sup-
methods.AnnuRevEcolSyst23:383-404.
portedbyaCAPESdoctoralfellowship.
GriffithDAandPeres-NetoP(2006)Spatialmodelinginecology:
The flexibility of eigenfunction spatial analyses. Ecology
References 87:2603-2613.
Hansen TF and Martins EP (1996) Translating between micro-
AckerlyDD(2000)Taxonsampling,correlatedevolutionandin-
evolutionary process and macroevolutionary patterns: The
dependentcontrasts.Evolution54:1480-1492.
correlation structure of interspecific data. Evolution
Bininda-EmondsORP,GittlemanJLandPurvisA(1999)Build- 50:1404-1417.
ing large trees by combining phylogenetic information: A JohnsonWE,EizirikE,Pecon-SlatteryJ,MurphyWJ,AntunesA,
complete phylogeny of the extant Carnivora (Mammalia). TeelingEandO’BrienSJ(2006)ThelateMioceneradiation
BiolRev74:143-175. of modern felidae: A genetic assessment. Science 311:73-
Bininda-Emonds ORP (2004) Phylogenetic supertrees: Com- 77.
bining information to reveal the Tree of Life. Springer, LegendrePandLegendreL(1998)NumericalEcology.2ndedi-
Dordrecht,568pp. tion.Elsevier,Amsterdam,853pp.
Bininda-Emonds ORP, Cardillo M, Jones KE, MacPhee RDE, MartinsEPandGarlandJrT(1991)Phylogeneticanalysesofthe
Beck RMD, Grenyer R, Price SA, Vos RA, Gittleman JL correlatedevolutionofcontinuouscharacters:Asimulation
andPurvisA(2007)Thedelayedriseofpresent-daymam- study.Evolution45:534-557.
mals.Nature446:507-512. MartinsEP(1996)Conductingphylogeneticcomparativeanaly-
Burnham KP and Anderson DR (2002) Model Selection and seswhenthephylogenyisnotknown.Evolution50:12-22.
Multimodel Inference: A Practical Information-Theoretic Martins EP and Hansen TF (1996) The statistical analysis of
Approach.2ndedition.Springer-Verlag,NewYork,488pp. interspecificdata:Areviewandevaluationofphylogenetic
CarvalhoP,Diniz-FilhoJAFandBiniLM(2005)Theimpactof comparativemethods.In:MartinsEP(ed).Phylogeniesand
Felsenstein’s “phylogenies and comparative method” on theComparativeMethodinAnimalBehavior.OxfordUniv.
evolutionarybiology.Scientometrics62:53-66. Press,England,pp22-27.
176 Felinephylogeniesandphylogeneticanalysis
Martins EP (2000) Adaptation and the comparative method. Sechrest W, Brooks TM, Fonseca GAB, Konstant WR, Mitter-
TrendsEcolEvol15:295-299. meierRA,PurvisA,RylandsABandGittlemanJL(2002)
Hotspotsandtheconservationofevolutionaryhistory.Proc
MartinsEP,Diniz-FilhoJAFandHousworthEA(2002)Adaptive
NatlAcadSciUSA99:2067-2071.
constraints and the phylogenetic comparative method: a
Smith FA, Lyons SK, Morgan-Ernest SK, Jones KE, Kaufman
computersimulationtest.Evolution56:1-13.
DM, Dayan T, Marquet PA, Brown JH and Haskell JP
MartinsEPandHousworthEA(2002)Phylogenyshapeandthe
(2003) Body mass of late Quaternary mammals. Ecology
phylogeneticcomparativemethod.SystBiol51:873-880.
84:3403.
Morales E (2000) Estimating phylogenetic inertia in Tithonia TôrresNMandDiniz-FilhoJAF(2004)Phylogeneticautocorre-
(Asteraceae): A comparative approach. Evolution 54:475- lationandevolutionarydiversityofcarnivore(mammalia)in
484. conservation units of the new world. Genet Mol Biol
27:511-516.
PagelM(1999)Inferringthehistoricalpatternsofbiologicalevo-
WebbCO,AckerlyDD,McPeekMAandDonoghueMJ(2002)
lution.Nature401:877-884.
Phylogeniesandcommunityecology.AnnuRevEcolSyst
Purvis A (1995) A composite estimate of primate phylogeny.
33:475-505.
PhilosTransRSocLondBiolSciSeriesB348:405-421.
RangelTFLVB,Diniz-FilhoJAFandBiniLM(2006)Towardsan AssociateEditor:JoãoS.Morgante
integratedcomputationaltoolforspatialanalysisinmacro-
ecology and biogeography. Glob Ecol Biogeogr 15:321- Licenseinformation:Thisisanopen-accessarticledistributedunderthetermsofthe
CreativeCommonsAttributionLicense,whichpermitsunrestricteduse,distribution,and
431. reproduction in any medium, provided the original work is properly cited.