hsi h uhrmnsrp cetdfrpbiainadhsudroefl errve u a o entruhtecopyediting, the Please through Record. been of Version not the has and but version review this peer between full differences undergone to 10.1111/evo.13516 lead has doi: may and as which article publication process, this for cite proofreading accepted and manuscript pagination author typesetting, the is This Author Manuscript epritgaino e-adplotlgcldt omove era. to genomic data the paleontological into d and approach neo- the of that integration hopeful deeper am I da disparity. molecular morphological for in priors calibration node sugges disparity, Bayesian morphological construct of units mot in Brownian lengths phylogenetic fixe branch of on formulations taxa most fossil Unlike of positions taxa. me the Bayesian reconstruct new to a data describe trait I datasets. facil morphological also large may of They mo structures. generating morphological by in issues variation these improve can inferen data phylogenetic morphometric to depend applied their rarely by Though hindered morphology. been have th they While synthesis, of organisms. stages extinct and extant challe from major evidence merging One biology. evolutionary living in from goal life fundamental of tree comprehensive a developing Jointly 2 1 opoercdata morphometric A -al [email protected] E-mail: eateto clg n vltoayBooy Universit Biology, Evolutionary and Ecology of Department BSTRACT aoieParins-Fukuchi Caroline aeinpaeeto osl npyoeisuigquantita using phylogenies on fossils of placement Bayesian hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 1 , 2 1 eeec re opsdo extant of composed trees reference d iganwfaeoktruhwihto which through framework new a ting tt h ai n betv aggregation objective and rapid the itate igadepoecmaaiepatterns comparative explore and ting n osltx a ogbe a been long has taxa fossil and e rdtoa n geometric and traditional ce, g a tme rmdfclisin difficulties from stemmed has nge opooia hlgntc further phylogenetics morphological srbdhr ilhl ofcltt a facilitate to help will here escribed neo ulttv ecitosof descriptions qualitative on ence fMcia,AnAbr I USA MI, Arbor, Ann Michigan, of y o oes hsmto expresses method this models, ion ergru ast quantify to ways rigorous re s fot aersle nvarying in resulted have efforts ese hdta eeae quantitative leverages that thod tive Author Manuscript rza 01 Simões 2011; Brazeau ag lds n rvd opligeiec nfvro in Ronquist of 2011; favor (Pyron in evidence analyses compelling provide and clades, large unders increased have evolution, prob character upon morphological based approaches, These matrices. betwee morphological times divergence and ’total-evidence’ relationships of the proliferation reconstruct rapid evolutionary the in in topic decade contentious, past yet central, a been long eeaiaino h ue-atrmdlo uloiesubs nucleotide of model Jukes-Cantor mod the ‘Mk’ of so-called generalization the A use model. generally adequately methods to probabilistic difficult be often can phylogenentics uniaiefaeokt vlaeterlaiiyo morp of reliability the m evaluate it to subjectively, framework made quantitative typically are judgments researchers these that since characters exclude to filtered frequently tde yipoigteacrc n ramn funcertai (Slater of dynamics treatment comparative and and accuracy the improving by studies hc hsi h aei miia aaesi nla (Golo unclear is datasets empirical in case the t is 2014), this Hillis which and (Wright simula parsimony upon outperform based approaches work previous Although states. character osl ersne ymrhlg ln Wes20;Wiens 2009; (Wiens alone morphology by the improve represented can fossils species extant to representing sought data has molecular work re Previous to times. species divergence extant and mo and phylogeny with fossil molecular of of capability compatibility the the understand explore to nearly need in a data been molecular of unavailability the is organisms ikno 95 Hawkins (Haus 1995; Wilkinson researchers different cause from can achieved process results this between of charac nature subjective discrete The assigning of assessment. practice Syst general taxa. the fossil from of stem preservation poor freque and also error is systematic there hemiplasy, and evolution someti convergent may as datasets morphological in to discordance susceptible Although frequently are data morphological example, leigtruhniei opooia aaesb levera by topology. datasets reference morphological in off noise by through reconstructions filtering fossil improve st further Their can results. approach accurate yield m can that phylogenies shown molecular have (2010) Stamatakis da and well-behaved Berger phylogen in observations, of placements accuracy fossil the of decrease estimation improve to tendency a charac have morphological not of does inclusion the that suggest studies Introduction: saohrcalne h iceecaatrmtie most matrices character discrete the challenge, another As osatsuc fdfclywe onl siaigphyl estimating jointly when difficulty of source constant A opooia aapeetohruiu hlegsimport challenges unique other present data Morphological hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This h oeo osldt nrcntutn hlgn mn li among phylogeny reconstructing in data fossil of role The tal. et tal. et 07.A naddsuc fptnilba,teemtie a matrices these bias, potential of source added an As 2017). 97 ctadadPnigo 00 Scotland 2000; Pennington and Scotland 1997; tal. et tal. et 02.Ti a eetbt ae-adneontological and paleo- both benefit can This 2012). 02 uno 2018). Guindon 2012; 2 ehd htse osimultaneously to seek that methods upc ob oolsos However, homoplasious. be to suspect ooia traits. hological adn feouinr ep across tempo evolutionary of tanding iigadfsi aauigcladistic using taxa fossil and living n d lososta scaffolding a that shows also udy mtcsucso icrac often discordance of sources ematic irclyifr eosrcinof reconstruction inform ciprocally escmrsn iigadfsi species fossil and living comprising ters ilg.Ti a aietdoe the over manifested has This biology. radPec 91 lie 1995; Pleijel 1991; Presch and er rn tagtowr en of means straightforward a ering igifrainfo h molecular the from information ging ipaigbae rmsedn signal. misleading or biased displaying l osltx.A eut hr has there result, a As taxa. fossil all yb fbnfi oitouea introduce to benefit of be ay e ttst aatruhqualitative through taxa to states ter tysbtnilniesemn from stemming noise substantial ntly ofe l 07.Eprcldatasets Empirical 2017). al. et boff aes xadn pnthese upon Expanding tasets. eosrcino eainhp among relationships of reconstruction l(ei 01.Ti sa is This 2001). (Lewis el oprtn osl nphylogenetic in fossils corporating ao reoclbedisagreement irreconcilable major t netmto fdvrec times divergence of estimation in nty iuinta accommodates that titution e aahssgetdta Mk-based that suggested has data ted e eetbooia rcse such processes biological reflect mes rsn,rsacesemploying researchers present, t blsi oeso oeua and molecular of models abilistic eetn n odtosunder conditions and extent he tal. et pooia aat better to data rphological omnyepoe in employed commonly eemn hte h nlso of inclusion the whether determine tcrcntutos n can and reconstructions, etic n opyoeei nlss For analysis. phylogenetic to ant tospaigfsiso fixed on fossils placing ethods gn ewe iigadextinct and living between ogeny 00.Terslso these of results The 2010). igognsshas organisms ving tal. et 2003; k ealso re possible Author Manuscript nosseto ilaigrsls(a 98 Goloboff 1998; (Rae results misleading or 2001) Wiens inconsistent 1993; (Thiele data propos morphological been quantitative have approaches Several easil more data. to present-absence discretized typically are but phylogenetics, nomtvns vrsm iceie hrce codings. character t discretized continuous some that over possibility informativeness the exploring while equal, chara be discrete to p similarly in perform traits characters discrete continuous and ex continuous I of 2017), performance (Parins-Fukuchi relative study previous a the In Nevertheless, unexplored. 2002). MacLeod 1988; 1973, (Felsenstein s,adtesniiiyo oooia eosrcint t to reconstruction topological present. of sensitivity the and use, eea rvospyoeei tde sn parsimony-ba using studies phylogenetic is previous this several which to extent the quantify to needed previo be 2 will point studies explored I Although represe variation. by interspecific codings character discretized potentia over 2) information and subjecti matrices, reducing character 1) discrete by constructing available when levels two Conti on data. benefits fossil morphologi poorly-sampled from analyzing extracted when be beneficial can that morphometr information of of space amount state continuous The holistical phylogeny. to inform d potential with the occur offer that morphometrics subjectivity geometric and bias to relating concerns iceectgre.A eut tmyotnb preferable be often may it result, a binning As by datasets categories. continuous discrete in information of amount the ohsiae oes(Wright models sophisticated resul (Ronquist erratic times to divergence lead of can estimates data assumptions and the assumptions from model significantly between depart to likely also are Catalano rcutstasoiin(aLo 01 02 Catalano 2002; 2001, regions (MacLeod homologous transposition truly Procrustes reco from derived in workers are other that of coordinates success the by 2002 alleviated (Felsenstein partially morphospace in homologou rotations out from parsing resulting in difficulties potent and upon characters primarily across based con analysis, analyze conce phylogenetic to raised in have tools data available authors of previous lack addition, the In to framework. due part in be may eeoe sn otnoscaatr oee ne Browni under modeled characters continuous using developed 2016). rdtoa iermrhmti esrmnshv ogbe long have measurements morphometric linear Traditional o l fteeraos otnostat aebe sugges been have traits continuous reasons, these of all For saohrsuc fcniuu ris emti morphome geometric traits, continuous of source another As ca the have shape and size morphological quantify that Tools h aletsuisivsiaigpoaiitcmethods probabilistic investigating studies earliest The tal. et hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 00 mt n edik 03,bthv o andsubstan gained not have but 2013), Hendricks and Smith 2010; tal. et 06,tesadr ymti kmdlrmisi frequent in remains model Mk symmetric standard the 2016), tal. et 06.Atog eetsuishv rpsdmore proposed have studies recent Although 2016). 3 tal. et l rsrighr-o phylogenetic hard-won preserving lly i rqetmsac sfil nla at unclear fairly is mismatch frequent his nlz hmalongside them analyze y yicroaealdmnin fsaeto shape of dimensions all incorporate ly tr hnpyoeei aflf sstto set is half-life phylogenetic when cters at a xedphylogenetic extend may raits tn h ulrneo observed of range full the nting e ehd (González-José methods sed yoeei neec.Ifudthat found I inference. hylogenetic tal. et sy(aisFkci21) future 2017), (Parins-Fukuchi usly srcigpyoeyfo landmark from phylogeny nstructing oaayecniuu risdirectly. traits continuous analyze to a ro tmigfo covariance from stemming error ial srt hrces prahssc as such Approaches characters. iscrete .Hwvr hs ocrshv been have concerns these However, ). neseicvrainfo variation from variation interspecific s fteM oe.Ti ormatch poor This model. Mk the of fpyoeei neec were inference phylogenetic of oee,teecnyield can these However, . n bu h s fmorphometric of use the about rns dfrtedsrtzto of discretization the for ed lrdtruhsmltosthe simulations through plored ruehsrmie relatively remained has use ir hthv enpoel lge using aligned properly been have that uu risi eea a engender may general in traits nuous h aei ies miia datasets. empirical diverse in case the eba fe nonee when encountered often bias ve n-cldvrainit shared into variation fine-scaled 06,admyi rnil reduce principle in may and 2006), sadhg netit nposterior in uncertainty high and ts a aaes hc a be may which datasets, cal cdt ih loices the increase also might data ic nmto B)(Cavalli-Sforza (BM) motion an iuu ri aai probabilistic a in data trait tinuous e saptnilalternative potential a as ted 00 ooofadCatalano and Goloboff 2010; nepoe nmorphological in employed en rcdt aesonuiiyin utility shown have data tric aiyt leit ayo the of many alleviate to pacity iltato.This traction. tial tal. et 2008; Author Manuscript eut e mlmnain xs,wt nyCnM ntePH the (Höhna in functionality ContML such only providing with exist, use implementations few rarely result, are still but gai methods, since comparative have phylogenetic models trait Continuous ba models. approaches substitution trait discrete by thes popularity sequencing, DNA in abund overshadowed of the emergence to the part with in available Due became 1973). Felsenstein 1967; Edwards and n siaigtepstoso osltx ln hsscaff this along taxa fossil of positions the estimating and relationship extant ar fixing taxa involves fossil procedure and alternative extant flexibl where very analyses, is total-evidence RevBayes to procedure. searching tree simple very nesod uniaietateouinmgtas proce c also (Landis directional might as evolution such trait Quantitative patterns by understood. violated be still draw may repeated model many with timespans timescal longer short over over distribution whe distribution cases non-Gaussian in a even to that, conform suggests nor This of assumption theorem. BM limit the central datasets, While empirical lengths. most branch in and violated topology the by Gaus multivariate defined a covariances to according wher distributed in equal, are For are phylogeny Mk. states to character comparison of in frequencies data stationary empirical to fit its’ improves xlrdol h lcmn fasnl ao tatime. a at taxon single a exti of recently placement and the extant only only explored of implementat placement the the but to well, restricted performed phylogeni approach on the taxa that individual found places that method a described iaeefml fflwrn lns(hn20)adcarnivora and 2009) (Chen e plants analyze flowering also of I family implementation. Vitaceae the of behavior Us the inference. explore associat phylogenetic obstacles in pressing data most morphometric the geometric of m some to tackle similarly to tree, seeks resulting the morpho across allows visualized morpholog and easily of procedure, terms estimation in the lengths simplifies branch This treats here Rev used from model Departing BM BM. under modeled characters n quantitative are variance unit in expressed are lengths branch when tree rte nteG agae h orecd spbil avail publicly is code packa source This The https://github.com/carolinetomo/cophymaru. language. Go the in written Software: Materials: and p Methods evolutionary the infer to lengths. (MCMC) branch Carlo Monte chain measu Markov morphometric uses geometric and traditional of comprised lhuh ieteM oe,B sfil ipitc tmyof may it simplistic, fairly is BM model, Mk the like Although, nti ae,Idsrb e prahta lcsmultiple places that approach new a describe I paper, this In l oslpaeetaaye eepromduigtenws new the using performed were analyses placement fossil All tal. et hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 03,btteietfibiiybtenB n oecomplicat more and BM between identifiablility the but 2013), tal. et 06.Hwvr h HLPipeetto ssa uses implementation PHYLIP the However, 2016). 4 sn h eut famlclranalysis, molecular a of results the using s lclrpyorm.Teapoc here approach The phylograms. olecular eetmtstepstoso osltaxa fossil of positions the estimates ge a Masmsta risa h i fa of tip the at traits that assumes BM eas dacrigt tssadsde jumps sudden and stasis to according ed lig rvosy Revell Previously, olding. siae iutnosy An simultaneously. estimated e o hlgntcifrne sa As inference. phylogenetic for d c aa nadto,teauthors the addition, In taxa. nct indsrbto,wt e of set a with distribution, sian e pnMro nucleotide Markov upon sed s hs olpeit Gaussian-like a into collapse these es, ecaatrsaecagsmybetter may changes state character re e infiatpplrt in popularity significant ned dwt h s ftaiinland traditional of use the with ed n iuae aa aiaeand validate I data, simulated ing ag,btteefc sntwell not is effect the but hange, suigqatttv aaTeauthors data.The quantitative using es .Tesadr hlgntcBM phylogenetic standard The s. prcldtst ersnigthe representing datasets mpirical oia iprt costx obe to taxa across disparity logical eet,rsetvl.Temethod The respectively. rements, o eeoe o h td was study the for developed ion tclear. ot bea resfwr at software free as able tne h kmdlasmsthat assumes model Mk the stance, aiycnotnb utfidb the by justified be often can mality aml (Jones mammals n ,hwvr ti ehp etsuited best perhaps is it however, e, prahswr quickly were approaches e LPpcaeadRevBayes and package YLIP n iceecaatrdt that data character discrete ant h keulbimasmto is assumption equilibrium Mk the ell osl nmlclrtesusing trees molecular on fossils cldvrec ahrta time. than rather divergence ical fwr package oftware e ereo eiiiythat flexibility of degree a fer tal. et aeet ffsisand fossils of lacements 21) h phylogenetic the (2015), dmdl cosa across models ed tal. et cophymaru tal. et 2015) (2015) Author Manuscript aiu ieiod(EL aclto ecie yFelsens by described calculation (REML) likelihood maximum absolute (in likelihood: divergence the of of period Computation long a b or long evolution case, of discrete len rate the branch to the Similarly to analogue models. continuous substitution a as per of deviations roughly standard thought morphological of disp units morphological in the measured reflect which lengths, Branch models. o iutnosyetmtn iegnetmsadfossil mol product and in times calibrations divergence as estimating used simultaneously fossils for place importantly to more method Perhaps the using article. pr this a of such scope of the feasibility outside and behavior The perf effectively traits. would continuous approach accommoda this can addition, that In prior tree fossils. a of specification the whi A requires topology estimated. and is times parameter divergence rate estimate a simultaneously and time, absolute to scaled Bte n ig20;O’Meara 2004; King and (Butler h iuae n miia aagnrtdfo hsstudy, this model from motion generated Brownian t data by empirical separated and is simulated trait the co each using where species file extant data of PHYLIP-formatted tree reference user-specified a on esrdi bouetm n h ntnaeu aeparame rate instantaneous the and time prod absolute the in by measured defined taxa between variances with distributed, n rnhn ie skon h oli yial oobtain ( to parameter typically rate is the goal of the (MLE) known, as times branching and time itiue admvralswt en0advariance and 0 mean with variables random distributed itiuin(ulradKn 04 O’Meara 2004; da King the and of (Butler likelihood distribution the calculate to variance-covar phylogeny a ultrametric construct generally These (PCMs). i methods used implementations conventional most than calculation ahrta s h opttoal xesv C likelihood VCV expensive computationally the use than Rather ntpcluae eerhr mlypyoeei Mmodels BM phylogenetic employ researchers usage, typical In h prahsta eciei hsppralrl pnthe upon rely all paper this in describe I that approaches The where h ehd htIdsrb s lgtydfeetparamet different slightly a use describe I that methods The with where t , σ 2 dX(t) X t oehr sarsl,rt n bouetm r confounded are time absolute and rate result, a As together. 0 hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This ie h ri au at value trait the gives stetm eiaieo h hnei trait in change the of derivative time the is σ oeaieeouinr aears clades. across rate evolutionary examine to ) tal. et t 0 . 06 ne M risaeasmdt emultivariate be to assumed are traits BM, Under . 2006) dX ar V E ( ( t X ( tal. et = ) X t = ) 5 t = ) σdB 06.Sneteemtostettetopology the treat methods these Since 2006). X σ 0 ( 2 dt t t r opooia aigusing dating morphological orm ) hslast h xetto htover that expectation the to leads This . ac egh ol eetete rapid a either reflect could lengths ranch a suigamliait normal multivariate a assuming ta, hswudas raecruaiywhen circularity create also would this , lcmns h ehdetmtsthe estimates method the placements, cdr sntudrto,adfalls and understood, not is ocedure oenpyoeei comparative phylogenetic modern n X tnostat otie ihna within contained traits ntinuous vial online. available eaayigcniuu ris this traits, continuous analyzing le ie hsitrrtto ol be could interpretation This site. clrdtn.T vroeteneed the overcome To dating. ecular ie ln htlineage. that along time) c ftereouinr distance evolutionary their of uct tog ti osbeto possible is it lthough ennutaerctesta include that trees non-ultrametric te e ( ter ac VV arxfo dated, a from matrix (VCV) iance and b.Eape a egendfrom gleaned be can Examples abs. h aiu ieiodestimate likelihood maximum the rt ewe aa r thus are taxa, between arity tsotie rmdiscrete from obtained gths rzto n likelihood and erization en(93.Fl eiain fthe of derivations Full (1973). tein aiirB oe fevolution of model BM familiar σ dB(t) ): hr rnhlntsare lengths branch where aclto,Iuetereduced the use I calculation, orsodn onormally to corresponding ntetatadtree and trait the in (1) (2) (3) Author Manuscript cln’mv ecie yYn 21,p 225). p. (2014, Yang t by for described given derivations move the scaling’ follow subcl updates to length multiplier branch l a for Branch applying 287). randomly by p. detail and (2014, in Yang individually described in pruned is ratio move is proposal The tip MH fossil two. appropriate the into when split merged is t are inserted the tree is procedure, the this of sta In rest the 1). the of (Fig. to case tees specific unrooted a for is move This (SPR) proposal. rando a by generate performed to are taxon taxa fossil bran of and positions points topological insertion fossil of distribution posterior bra of estimate Carlo rough Monte initial Markov-chain the using length tree expected ainefo h ist rpgt rmtetp oteroot: the to tips the from propagate to tips the from variance h nenlnd uigtelklho optto ttep the at computation node, likelihood internal the during node internal the eneult h otatbtentecaatrstates, character the between contrast the to equal mean ainecluae stesmo ahcidedge, child each of sum the as calculated variance h PIC, The tec fthe of each at otre rvra,adtelg likelihood, log- proced the this and In traversal, algorithm. postorder ‘pruning’ fr a computed using is (PICs) likelihood contrasts tree (198 The below. Felsenstein briefly by given summarized also are algorithm and likelihood h oa o-ieiodo h tree, the of log-likelihood total The oeua ao:1 a uiom,2 xoeta,ad3 c a 3) and exponential, 2) (Rannala (uniform), flat 1) different three canon: tested molecular and implemented I phylogenetics, in Priors: hsmto ssaMtooi-atns(H loih (Has algorithm (MH) Metropolis-Hastings a uses method This ic h siaino rnhlntsfo otnostrai continuous from lengths branch of estimation the Since tal. et x internal hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This n v 01.Tecmon iihe ro loofr h pint option the offers also prior Dirichlet compound The 2011). internal nenlnodes. internal scluae tec nenlnd n sda h characte the as used and node internal each at calculated is , sas xeddb vrgn h egh ftecidndst nodes child the of lengths the averaging by extended also is L node = 1 2 x v internal internal ∗ L o (2 log tree L tree scluae ysmigtelglklhoscalculated log-likelihoods the summing by calculated is = = π o ( log + ) L v ( = x internal node 1 node ∗ 6 X n scluae smliait oml iha with normal, multivariate as calculated is v v =1 v 2 v 1 1 ( + ) v + 1 and 1 L + + + node hlnts eragmnso the of Rearrangements lengths. ch ( ( x v v v v x 2 v 1 1 v 1 2 c lengths. nch rnhlnt rosdrvdfo the from derived priors length branch 2 2 l rnn n enetn fossil a reinserting and pruning mly 2 ete’utpir r’proportional or ’multiplier’ the he ln ihafl eiaino the of derivation full a with along , + and ∗ b n rcltn(02,and (2012), Freckleton and 1b) r,ec nenlnd svstdi a in visited is node internal each ure, rn oe h delnt fthe of length edge The node. arent ( + ) dso h re Hpooa ratios proposal MH tree. the of ades : ∗ mon iihe ro after prior Dirichlet ompound dr ute rnn n regrafting and pruning subtree ndard v hl h deuo hc h tip the which upon edge the while , mtepyoeei independent phylogenetic the om v v oeg egh htln h fossil the link that lengths edge wo 2 2 si eaieyucatdterritory uncharted relatively is ts 1 ) x ) ntsaeudtdboth updated are engths ) x 2 1 ig 90 osmlt the simulate to 1970) tings tec utnigeg and edge subtending each at − x 2 ) 2 e h cl fthe of scale the set o tt representing state r lo the allow o (4) (6) (7) (5) Author Manuscript Randomly select anew insertion branch Randomly select and prune single fossil Regraft fossil branch along new edge A B A A B B Fossil with unknownFossil with placement Refererence topology hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This F E E E F 7 C D C C D D F Author Manuscript expressions:: acltdo h eeec re(xldn osltx) N taxa). fossil (excluding tree reference the on calculated aheg egho h rndtretxnte ( tree three-taxon ti pruned the the at ’traits’ of as length treated edge calc are each are and PICs root, node. new each the along subtending rerooted is introdu tree procedure the iterative procedure, the using retaine computed likelihood are highest lengths the e yields all that along point inserted insertion individually is tree, fossil guide Each molecular the estimated. to fossils using adds achieved successively are step placements initial These tree. starting tree: starting ML rough a Generating nrsvl nmrhmti aaesta ndsrt data discrete excluded in be e than be datasets might morphometric problem in This intrusively signal. discordant and noisy display analysis. under dataset the the starti reflect b non-random of of specification construction the straightforward inform allows lengths branch proce the this while from trees, achieved topologies furt the without trapp here, used becoming applications not to and propensity cautiously, strong interpreted a be has should it t that interpreting is with own problem their b One can simulation. lengths MCMC starting during These used tree. complete the on optimized complet a with along calculations. algorithm, conve the likelihoods of the explanation and detailed values their until lengths branch fec iewt h eeec re nti rcdr,tel the procedure, this In tree. we reference of the set with a site computes each This of purpose. develop this I for sites. (2010) reliable Stamatakis more favor to signal noisy through eipeetdteapoc sdi h otLpormt ge to program ContML the in used approach the re-implemented I iue1. Figure n ao udeivle nteueo opooia aais data morphological of use the in involved hurdle major One sites: concordant for Filtering fos the of all for assigned been has placement initial an Once n each on rerooting successively by iterated is process This priori a hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This admfsi rn n erf procedure. regraft and prune fossil Random ntebsso ecie neiblt.A eut hr i there result, a As unreliability. perceived of basis the on v v v ˆ ˆ ˆ 1 2 3 j j j = = = j j j P P P =1 =1 =1 n n n ( ( ( x x x 1 2 3 j j j − − − 8 v x x x i scmue nltclyuigthe using analytically computed is ) 2 1 1 n n n j j j )( )( )( x x x .A ahse,ME ftebranch the of MLEs step, each At d. 1 2 3 itn rnhso h re ihthe with tree, the of branches xisting gtesfrteMM n rosthat priors and MCMC the for trees ng j j j es ic risaemc eslkl to likely less much are traits since sets, tpieadto.Ulk otL this ContML, Unlike addition. stepwise x,telklho ( likelihood the ext, ueaeue nyt osrc starting construct to only used are dure n ool h oslpstosare positions fossil the only so and dapoeueaatdfo egrand Berger from adapted procedure a ed so he-ao re h L of MLE The tree. three-taxon a of ps g.Flesen(91 ie more a gives (1981) Felsenstein rge. e CCsacig nthe In searching. MCMC her e y(esnti 91) nthis In 1981a). (Felsenstein by ced − − − klho ( ikelihood eiaino h L rnhlength branch MLE the of derivation e ac eghpir.Ti procedure This priors. length ranch sdt nombac eghpriors length branch inform to used e di oa pia sarsl,it result, a As optima. local in ed x x x pce omnfs vnmore even manifest to xpected ltdt aho h he edges three the of each to ulated gt ae pnteconcordance the upon based ights 3 3 2 d ftete n acltn the calculating and tree the of ode j j j ) ) ) erslso h Lapoc on approach ML the of results he is h rnhlntsare lengths branch the sils, hi rqettnec to tendency frequent their L eaea prxmt ML approximate an nerate ref fec ieis site each of ) L edt filter to need a s )o ahst is site each of n) (10) (8) (9) Author Manuscript acltdaogeach along calculated ieiodi h eeec reta 5o oeo h random the of more equal or is 95 weight than the tree where reference value the binary in generate a likelihood value 3) the and where 100, 1, by ob and divided value 0 the between equals value weig weight point then the floating are where sites values, The integer 100. whole and 1) 0 between weight a possessing admte,tewih ftest siceetdb n.Thu one. by incremented is th site along the higher of is weight site the a the tree, lengths of random Branch likelihood the formed. If is tree draws. full random a until manner stepwise a hrces hydfe nterproe asmn iewe site Parsimony purpose. their p in during differ calculated they weights characters, site addition the an to As terminology datasets. similar integ noisy and particularly floating in the use than greatest o penalty filters harsher it a Since enforces scheme accepted. are M likelihood hill-climbing the inefficient increase very that a becomes algorithm reducin scal substantially MCMC the rejected, the in be increase to large proposals the all to nearly cases due most likely in is practical scheme most integer probably are schemes binary and simulation MCMC during computations likelihood subsequent where and hsyed egtvco hti h aelnt stechar the as length same the is that vector weight a yields This if if and napiain on htitgrwihigcue orM poor caused weighting integer that found I application, In fe h egt r optduigteiptgiete,th tree, guide input the using computed are weights the After δnj δnj if: 0 = hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This if: 1 = n f10pyoeisgnrtdrnol yscesvl graf successively by randomly generated phylogenies 100 of W − → W − → W − → W − → W − → L L j int ref ref j binary j binary j int j int = 9 L > L < > < n X 100 =1 1 = 0 = 95 95 n n δ nj h fcec fteagrtm neffect, In algorithm. the of efficiency the g lnt,atog hs rcdrsshare procedures these although note, al tdsodn ie opeey h binary the completely, sites discordant ut gt r neddt omlz the normalize to intended are ights rioyaayi fmulti-state of analysis arsimony ftelglklhos hscauses This log-likelihoods. the of e eeec reta h current the than tree reference e rshms n omgtb of be might so and schemes, er andfo qain1,2 a 2) 11, equation from tained h ormxn civdb the by achieved mixing poor The . ete sinduiguniform using assigned then re ,site s, rmterno oprsnis comparison random the from d re,ad0i esta 95: than less if 0 and trees, s. tduigoeo he schemes: three of one using hted o1i h iedslydahigher a displayed site the if 1 to M iig n otefloating the so and mixing, CMC yaesoe,adue nall in used and stored, are ey ce arx ihec site each with matrix, acter erh ic nyproposals only since search, L j eevsteitgrweight: integer the receives igndsin nodes ting (11) (12) (13) (14) (15) (16) (17) Author Manuscript ilgclpoes iia fetmyi rnil occur principle in may effect similar misleading A shared exhibit process. sites biological many when occur sourc can common studies fairly a Nevertheless, morphology. traits of discrete aspects than traits continuous between the covariance 2002), 1988, (Felsenstein phylogenetics has covariance morphometric Although inducin covariance. of of effect effect the confounding examining in interested primarily contin was from site reconstructions that handicap 2017) significantly (Parins-Fukuchi itself study previous a covaria in no showed displaying I from (ie. expected independently evolved sampling was of level trait the examine to sensible d most morphometric that quantitative as comprehensive such more methods of future, assembly the In one only "leaf"). from "pelvis", is compiled "cranium", sampling often of are level which This datasets, datasets. continuous empirical two the as size are them generate to used scripts https://github.com/carolinetomo/fossil_placement_te all m and with trees, distribution Newick exponential sets, an Br from function. drawing t "multi2di" by star the randomly a using into resolved tree randomly true was the which collapsing us by simulated generated were were traits trees All perform 2012). were (Revell simulations package dirty) 25, phytools and 10, (clean of trait intervals All in traits trees. “dirty” “clean sets 50 with of combined consisting t and alignments explore To generated pruned. also were I conained fossils signal, tree five resulting the The when 0.5. tree of reference birth-deat parameter a death under a simulated and was 1.0 us tree tree The the BM. along under positions simulated sin their a estimated From and simulations. taxa of "fossil" set a performed I implementation, pr the when results Simulations: to compared and thoughtfully, applied be Nevertheless, estimates. credibility interp posterior to standard means straightforward poss a datasets offers in framework support Bayesian posterior increasing ou of filtered effect are the signal, discordant sh yielding to processes tre appear other guide that molecular characters the relationships, that species assuming extant By prior. a as functions oerlal.A eut h ud rei sdt identify to used is tree placements. guide fossil the inform result, a likelihood, As tree overall reliable. the more to th topology decrease reference to the designed with is t here deployed to approach spaces weighting state differing with characters of contribution oepoetebhvo fteeapoce ne different under approaches these of behavior the explore To etitdtesmltost arysalnme ftrai of number small fairly a to simulations the restricted I ML an in implemented originally was procedure this Although hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 10 sts. h leigapoc,a n ro,should prior, any as approach, filtering the wsgicn ro,hmpay rreflect or homoplasy, error, significant ow hc,iely huddsrb similar describe should ideally, which, , ostat,ad2 eas nti td I study this in because 2) and traits, uous eoealte egh ncnrs,tesite the contrast, In length. tree overall he sigmn icratcharacters.The discordant many essing rd-mhszd hspoeuehas procedure This de-emphasized. or t duigte"atM ucini the in function "fastBM" the using ed nta ihihigsga ae obe to taken signal highlighting instead c ihohrsts.Ti sbcue1) because is This sites). other with nce rissmltdaogtetu tree, true the along simulated traits ” ei orao oepc greater expect to reason no is re admniewtottepotentially the without noise random g n 0 aaeso 0characters 50 of datasets 100 ing plg sn h d2ut"function, "di2multi" the using opology nsuisuigcontinuous using studies in ie htaems ieyt reliably to likely most are that sites n 0tat eeae ln random along generated traits 50 and enepesda ao ocr in concern major a as expressed been arycmo mn existing among common fairly vial at available iecvrac osnti n of and in not does covariance wise ferri oeua phylogenetic molecular in error of e n aeprmtro ..Random 1.0. of parameter rate a ing otiuino ie htdisagree that sites of contribution e o sntused. not is ior e h eutn upr ausas values support resulting the ret l iuae re rndfive pruned I tree, simulated gle eefc fcnitn n noisy and conflicting of effect he aaes u tpeet tseemed it present, at but datasets, inldet oelegitimate some to due signal 1tx,laiga36-taxon a leaving taxa, 41 sbcueti eetdasimilar a reflected this because ts rvdsa cuaeve of view accurate an provides e rtoaaoia ein (eg., regions anatomical two or a e o1 h iuae data simulated The 1. to set ean xsigdtst.Ec simulated Each datasets. existing oe,wt it aaee of parameter birth a with model, h nhlntswr hnassigned then were lengths anch etnsadvldt the validate and settings srbdhr a norg the encourage may here escribed otx,teapiainhere application the context, Author Manuscript lopromdaaye nadtsto 33 adakcoordi landmark 3D 33 of dataset a on analyses performed also reta was likelihood mean highest the e with under analysis performed the were each, t runs ensure To redundant several analysis. optima, unweighted local an to compared perfor and were sites Analyses the convergence. analys for All checked models. visually across and run stability Individual assess taxa. to fossil priors the length of scaffol positions this the Using estimate to supplement. data online the in available hlgntcapoce htepiil s emti mor geometric use explicitly c that a approaches represents phylogenetic coordinate landmark of each rows data, stor to morphometric to corresponding used columns, those in to arranged similar measurements matrices character as input are r w omnsucso otnostatdt,ads eec the were In so and types. data, data trait possible continuous different of sources 3-dimension common of two composed are data li morphometric 1) geometric of 2) comprised and datasets empirical on analyses performed analyses: Empirical calcu were distances placement and scripts. summary Tree represent run. trees MCMC MAP the distributions. tree of summaries (MAP) miia hrce arcs rc ls n eeec tr reference and https://github.com/carolinetomo/fossil_placement_te files, trace matrices, character Empirical o h xattx rm8nceradclrpatgnsgath genes chloroplast and (Soltis nuclear system Chen 8 PHLAWD from from acquired taxa were extant data the These for taxa. sp Vitaceae seed fossil and 8 pollen and from gathered measurements continuous 51 aln ewe n .Paeetacrc a vlae us evaluated was accuracy Placement y to 1. tree and simulated 0 the between in falling tips two dist between These distance branch. possible no insertion of reconstructed number the the from counting branch by calculated betwe was distances o This the mean calculated placements. a then with I data method, simulated placement the the for exce employed (ESS) was sizes prior sample effective length the for that run ensure were to simulations checked MCMC weights. supplemented site were applying These without approaches. floating pe and I binary schemes, the weighting of performance relative the explore ore fcnitaderror. pro and to conflict matrices of character sources across signal of distribution careful the characters, morphological ac covariance discrete by and mislead molecular being toward risk tra inherent continuous greater although so, And characters. morphological oepiil etteifraieeso emti morpho geometric of informativeness the test explicitly To oass h tlt ftenwapoc naayigcontin analyzing in approach new the of utility the assess To siae h hlgntcpstoso osl sn mo a using fossils of positions phylogenetic the estimated I hs iuae aaeswr hnue orcntuttep the reconstruct to used then were datasets simulated These hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This tal. et 01.Tesqec lgmn sdt osrc h ud tr guide the construct to used alignment sequence The 2011). cophymaru 11 mlmnaino h ehd hs characters these method, the of implementation sts. e r l vial nieat online available all are ees el dniybooia n systematic and biological identify perly swr u o 00000generations 30,000,000 for run were es eepromdudraltrebranch three all under performed were s nlssi motn opoel dissect properly to important is analysis ig nlzdtemrhlgcldata morphological the analyzed I ding, hmti aa(Catalano data phometric ne eedvddb h largest the by divided were ances iceetat,wt homologous with traits, discrete e tmtie a o eesrl carry necessarily not may matrices it osststa tde ae on based studies than sites ross ermaueet n proportions, and measurements near ined. fre eosrcin sn both using reconstructions rformed ntetu n eosrce fossil reconstructed and true the en edamaueo lcmn error placement of measure a ield dd20 h xoeta branch exponential The 200. eded lm,smlryt previous to similarly olumn, e ihbnr egt ple to applied weights binary with med rdfo ebn sn the using Genbank from ered c obnto fstig.For settings. of combination ach h igems ape reduring tree sampled most single the a CCrn eenttapdin trapped not were runs MCMC hat n the ing aa ntecs ftegeometric the of case the In taxa. ercdt nfsi lcmn,I placement, fossil in data metric yaaye fteniydatasets noisy the of analyses by e eaaigtetu insertion true the separating des cmn ape cos17extant 147 across sampled ecimens pooia arxcmrsdof comprised matrix rphological llnmr oriae.These coordinates. landmark al ae ersnig4 xatad5 and extant 46 representing nates oe ots h ehdacross method the test to hosen ..T vlaeteacrc of accuracy the evaluate To 1.0. f osmrhlgcldt,I data, morphological uous aeet ftefiefsis To fossils. five the of lacements 20) osrce ud tree guide a constructed I (2009). ,0,0 eeain and generations 1,000,000 ae sn utmPython custom using lated aiu posteriori a maximum tal. et 2010). eis ee Author Manuscript xic oslcrioa rna(Jones crania carnivoran fossil extinct ifr hsueo opoercdt ssmlrt htused that to similar al. is et data morphometric detail of the use Although this taxon. differ, each the by contains occupied column landmark each single posit that Z such or Y, ’aligned’ X, are aligned characters the represented character (Klin each st MorphoJ where original using the transposition of Procrustes supplement to data subjected the from obtained was taxa lcmn ro yhl,ee huhtesga-oniera signal-to-noise the though even half, o by application error scheme), placement weighting the no in with distant prior si quite exponential dirty are and that clean positions of fossil number reconstructs equal an contain error alignments placement when prior) Diri exponential compound the the under under (unweighted, t weighting within binary fall (with to 7.8% tend between char positions the fossil of estimated half the where signal, case extreme the exc in accurate, even fairly Nevertheless, are reconstructions e the 0.3% noise, I displaying random correct, 2). always and nearly 1 (Tables is taxa reconstruction fossil placing in accurate generally Simulations sampled all across Discussion mean and the Results as packag summarized DendroPy were the lengths with bundled Branch is that script esti SumTrees MAP the the material to (supplementary compared placements were concordant These generally clade. each summar of (MCC) support credibility the clade maximum as distributions tree Exp tree. starting 1.0. ML of the values of mean length assigned tree total para the alpha as assigned specified were priors Dirichlet als were supplement. sites data float-weighted binar and the unweighted using using weighted analyses were Sites above. described method settings. all under (>1000) convergence. values assess ESS large to 20 v1.6 for Tracer run using were examined simulations visually MCMC perform and likelihood. priors, mean length highest branch the t three with all As under measurements. data of canid set transformed remaine this taxa on the between performed preserved distances procedure relative This the since 2016). Team dataset, Core relativ (R of R absolu pattern in the original function increase the to maintaining traits while the site of each achiev all in scaled difficulties I causes this, which remedy surface, MCMC scale. the absolute in an peaks on (variance) dispersion low very showed eosrcin ffsi lcmnsfo h iuae dat simulated the from placements fossil of Reconstructions simu the than complex more were datasets empirical the Since g lengths branch and trees starting used I datasets, both For 21) h eutn risdslydpyoeei signa phylogenetic displayed traits resulting The (2010). hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This tal. et 05.Arfrnete opsdo h 6extant 46 the of composed tree reference A 2015). 12 os ae(8 ro ne the under error (18% case worst ecretrgo ftete,exhibiting tree, the of region correct he aine costx sn h scale() the using taxa across variances e p hnniebcmssevere. becomes noise when ept h ae ia nlsswere analyses Final same. the d rro vrg.I h rsneof presence the In average. on rror .MCsmaiswr bandusing obtained were summaries MCC ). evrac vnyars aaeel at evenly taxa across evenly variance te ebr 01.Ti ile matrix a yielded This 2011). genberg i sqiehigh. quite is tio h bec fniytraits, noisy of absence the n naeaeudrteepnnilprior exponential the under average on eeso . n eaparameters beta and 1.0 of meters h egtn rcdrsreduces procedures weighting the f ntemto ecie yCatalano by described method the in dsvrlrn,rtiigteoewith one the retaining runs, several ed oriae noedmnino a of dimension one in coordinates efre,adaepeetdi the in presented are and performed, o oheprcldtst achieved datasets empirical Both nnilbac eghpir were priors length branch onential o ainecnrsl nnarrower in result can variance Low urudteaayia approaches analytical the surround s o foelnmr.These landmark. one of ion cesdslycmltl random completely display acters d.Teecodntswere coordinates These udy. o h nlaaye.Intermediate analyses. final the for y eVtca aae,Iaaye the analyzed I dataset, Vitaceae he ,bttetasoe coordinates transposed the but l, SkmrnadHle 2010). Holder and (Sukumaran e e.Adatog h procedure the although And tes. nrtdfo h og ML rough the from enerated he ro)ad17.6% and prior) chlet n CCcnegne To convergence. MCMC ing st hwdta h ehdis method the that showed asets e.Teesmaismaximize summaries These ies. ae,hwvr n yielded and however, mates, 0000gnrtos and generations, ,000,000 trees. inlpeeti h original the in present signal ae aa umrzdthe summarized I data, lated Author Manuscript repstosars l 0 elctso ahdtstdiv dataset nodes. r each between of separating replicates the nodes 100 of all under across number fossils positions average true simulated the as placing measured when is Error error Mean 1: Table hsmto o h nlsso miia datasets. empirical of analysis the dr for improves method performance this correct the the applied, within are average schemes on weighting fossils place to manages still t accuracy, reconstruction exp decreases be analyses might unweighted noise this significant in developed where method datasets, the morphometric applying analy of unweighted prospect to the t compared toward 20 noise to and 2 signal by of accuracy proportion improving with noise, particu random appear misleading calibrations weight binary thr The filter Stama effectively estimation. and to Berger appears of also adaptation approach The calibration clear. estim is well, signal performs when method error the computa that the shows of datasets validation clean a of provide and fossils of placement reference the of regions th particular under to calculated localized weights is which conflict to extent othe the among determine signal informative to dis correctly may but traits taxa, continuous of cases, subset these s In be taxa. may t certain noise dirty only reality, the in However, since simulations, noise. of random set completely en this than in rather examined likelihood not the was t to because sites is noisy This of contribution trees. cas of in regions beneficial different remain between could varies scheme weighting float the advant that apparent the despite However, analyses. unweighted datasets. noisy particularly sche analyze weighting to binary which with combined Thi prior accuracy. branch placement Dirichlet in with increase dataset, largest simulated the the displaying nois of the noisiness on the prior with exponential increases to weight compared binary error the in imp using reduction The when comparisons. greatest the is of Dirichlet one compound Pla but all 2). (Table in datasets Dirichlet simulated compound the on prior length branch 0cen+5 it .9 .8 0.528 0.009 0.489 0.396 0.485 0.012 0.420 0.195 0.291 0.009 unweighted 0.153 float_weights 0.024 dirty binary_weights 50 + clean 50 dirty 25 + clean 50 dirty 10 + clean 50 clean 50 dataset vrl,tesmltosdmntaeteefiayo h met the of efficacy the demonstrate simulations the Overall, im shows weighting binary priors, length branch both Across the In cophymaru hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This mlmnain h opudDrcltpiroutperforms prior Dirichlet compound the implementation, 13 uhniydtst oimprove to datasets noisy ough ddb h aiu osbept length path possible maximum the by ided tree. emto efrspeitby and predictably, performs method he al fetv tdaigwt rampant with dealing at effective larly ieyecuigte.Ti possibility This them. excluding tirely eutsget httecompound the that suggests result s efla egtn ceelmt the limits scheme weighting float he lymsedn inlaogsome among signal misleading play h 0cen5 it dataset dirty clean+50 50 the ust.Frhrwr ilb needed be will work Further subsets. r egbrod oee,when However, neighborhood. eeterrlwrudrthe under lower error cement g fbnr egtn,i spossible is it weighting, binary of age at eegnrtdt reflect to generated were raits etdtst.Teipoeetalso improvement The datasets. iest emyb h da oethrough mode ideal the be may me otwihigshm aywhen vary scheme weighting float e inlipeetto.Teanalysis The implementation. tional tn oslpaeet ihvr low very with placements fossil ating ce.Atog nrdcn os to noise introducing Although ected. sial,hglgtn h rms of promise the highlighting astically, n cee eutn na2% a in resulting scheme, ing e.Teerslsso promise show results These ses. rcue odslydsodnein discordance display to tructured sweetedsrbto fnoise of distribution the where es okt h nlsso large-scale of analysis the to work cntutdpaeet rmtheir from placements econstructed ai’(00 ieweight site (2010) takis’ oeetehbtdudrthe under exhibited rovement o o h phylogenetic the for hod msdpnigo h relative the on depending imes rvdacrc vrfla and float over accuracy proved xoeta rnhlnt prior. length branch exponential h exponential the Author Manuscript meoisswildei Ampelocissus biologic t generate noise, to this noise Despite this runs. through all filter across adequately time molecula to the the favor of to 95% failing least sites at 51 procedure of calibration 10-12 weight with The dataset, S1). Fig. 2, (Fig. results dataset: Vitaceae datas each nodes. of between replicates length 100 separ all nodes across of positions number true the average their the under as fossils measured simulated is placing Error prior. when error Mean 2: Table paradoxa cosrn,adwe ohteepnniladeprclcom empirical and exponential the both when and runs, across hi epciegenera. respective their Parthenocissus_clarnensis ihn xatseis ohgopwt eaae non-monop separate, with group both species, extant no with 0cen+5 it .3 .6 0.522 0.012 0.501 0.390 0.468 0.009 0.402 0.189 0.234 0.009 unweighted 0.120 float_weights 0.021 dirty binary_weights 50 + clean 50 dirty 25 + clean 50 dirty 10 + clean 50 clean 50 dataset plcto ftefsi lcmn ehdt h iaeed Vitaceae the to method placement fossil the of Application l i fteetx r tbei hi lcmns groupi placements, their in stable are taxa these of six All . hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This lcdwti crown within placed Palaeovitis_paradoxa and , meossrooseae Ampelopsis Cissus and , 14 eaae yol oefrom node a only by separated , iscru jackesiae Cissocarpus l hr ldswt h xatmmesof members extant the with clades share all tdvddb h aiu osbepath possible maximum the by divided et lyraoal results. reasonable ally eeldsbtnilniei the in noise substantial revealed eeec reoe h admtrees random the over tree reference r ebnr egtn ceeappeared scheme weighting binary he yei ruso crown of groups hyletic tn eosrce lcmnsfrom placements reconstructed ating on iihe rosaeapplied. are priors Dirichlet pound opudDrcltbac length branch Dirichlet compound tstsoe eeal positive generally showed ataset gwti h aeclades same the within ng hc ersn genera represent which , Palaeovitis ii tiffneyi Vitis Cissus . , Author Manuscript hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This Cissocarpus_jackesiae_0.55 Cissus_hypoglauca Cissus_simsiana Parthenocissus_vitacea Parthenocissus_henryana Parthenocissus_dalzielii Cissus_vinosa Cissus_adnata Cissus_verticillata Cissus_diffusiflora Cissus_discolor Parthenocissus_quinquefolia Cissus_erosa Cissus_repens Rhoicissus_digitata Cissus_tweediana Cissus_striata Cissus_rubiginosa Yua_thomsonii Cissus_duarteana Ampelocissus_acetosa Cissus_rotundifolia Cissus_assamica Ampelopsis_cordata Ampelopsis_aconitifolia Parthenocissus_heptaphylla Cissus_hastata Cissus_obovata Cissus_trianae Rhoicissus_revoilii Cissus_campestris Cissus_cornifolia Vitis_rotundifolia Parthenocissus_feddei Parthenocissus_himalayana Parthenocissus_laetevirens Ampelopsis_humulifolia Rhoicissus_tridentata Vitis_vinifera Cissus_alata Nekemias_megalophylla Clematicissus_angustissima Cissus_antarctica Ampelopsis_japonica Vitis_magnisperma_1.0 Cissus_fuliginea Parthenocissus_clarnensis_0.96 Cissus_tiliacea Ampelocissus_erdwendbergiana Cissus_oblonga Ampelopsis_rooseae_0.42 Vitis_wilsoniae Cissus_hexangularis Cissus_oliveri Cissus_decidua Leea_compactiflora Cissus_penninervis Nekemias_grossedentata Cissus_producta Nothocissus_spicifera Cissus_microcarpa Cissus_reniformis Cissus_tuberosa Rhoicissus_rhomboidea Nekemias_cantoniensis Vitis_vulpina Cissus_granulosa Leea_angulata Leea_guineensis Clematicissus_opaca Cissus_quadrangularis Vitis_chunganensis Vitis_piasezkii Ampelopsis_delavayana Cissus_descoingsii Ampelocissus_abyssinica Leea_asiatica Vitis_aestivalis Vitis_amurensis Leea_aequata Leea_indica Ampelopsis_glandulosa Vitis_betulifolia Ampelocissus_acapulcensis Vitis_flexuosa Ampelocissus_polystachya Ampelocissus_thyrsiflora Leea_aculeata Ampelocissus_tomentosa Vitis_tiffney_0.62 Ampelocissus_gracilis Vitis_latifolia Leea_congesta Ampelocissus_barbata Ampelocissus_martinii Ampelocissus_parvisemina_0.62 Ampelocissus_africana 15 Pterisanthes_eriopoda Pterisanthes_polita Cissus_paullinifolia Cissus_heyneana Cissus_gongylodes Palaeovitis_paradoxa_1.0 Cissus_sterculiifolia Cissus_biformifolia Ampelocissus_javalensis Ampelocissus_divaricata Ampelocissus_wildei_1.0 Cissus_stipulata Author Manuscript t lcmn,atraigbtenpaeeta h center the at with placement between alternating placement, its ge ihpeiu eut Asnadd uzn21;Jones 2014; Muizon de and (Amson results previous with agree lcmnsof placements anvrndataset: Carnivoran sup prior. posterior length higher branch and Dirichlet Di stability compound greater the the under by taxa supported empiric two the these to for extended inferred be placements can observation this binar If the datasets. with combined when especially prior, exponential parvisemina Ampelocissus supplement. data the in cla available the is only tree displays Newick Figure (MCC). fr clades summarized credible is maximally Valu Topology red. placement. in for highlighted support are posterior branches and taxa Fossil prior. xoeta n iihe ros ihpseirspotv support posterior with priors, Dirichlet and exponential dnia otetplg sdb oe n olaus(2015 colleagues and Jones by used topology the to identical pacifica Piscophoca oe ta. lcn h w aai oense oiin H position. Muison. nested and more Amson a of in results taxa the two with the placing al., et Jones ne h xoeta ro,the prior, exponential u the support Under posterior higher achieves t and prior, This respectively. exponential runs, prior Dirichlet and exponential rbblt Fg 1,adicesst .2udrteDiric the under 0.62 to magnisperma increases and S1), (Fig. probability aiyOoeia.Nvrhls,lk h rbe aai t in taxa of problem placement the like Nevertheless, Odobenidae. family magnus n nistnec ogopa h etro h re Interes tree. the of center the at emlongii group Enaliarctos to tendency its in and nlsso h anvrndtstas ile eeal r generally yielded also dataset carnivoran the of Analysis a placements their in unstable are fossils two remaining The h iuain hwta h opudDrcltpirachi prior Dirichlet compound the that show simulations The iue2. Figure Vulpes ihnteCndefml,wiei aoial nw steo the as known canonically is while family, Canidae the within and hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This Pontolis iaeefsi lcmnsifre ne h opudDir compound the under inferred placements fossil Vitaceae lentl eovsit ldssae ycrown by shared clades into resolves alternately icpoapacifica Piscophoca Otocyon and and ipasrasrnl ekspot ohi em fispos its of terms in both support, weak reassuringly displays h atrpaeeti nieyt ecret eas tp it because correct, be to unlikely is placement latter The . cohc longirostris Acrophoca Allodesmus lentl cuiscae hrdb crown by shared clades occupies alternately meoissparvisemina Ampelocissus , cohc longirostris Acrophoca eansal cosrn,bt ipa eksupport. weak display both runs, across stable remain nlaco emlongii Enaliarctos 16 ifr lgtyfo h oooygnrtdby generated topology the from slightly differs xnsospo upr ne the under support poor shows axon le f02 n .4 respectively. 0.54, and 0.23 of alues drtecmon iihe prior. Dirichlet compound the nder otosre ne h compound the under observed port fteuroe oooy rgrouping or topology, unrooted the of ltpir(i.2.Similarly, 2). (Fig. prior hlet lrsls ti esnbet rfrthe prefer to reasonable is it results, al sfloigfsi i aesindicate labels tip fossil following es , ceeadapidt noisy to applied and scheme y igy lhuhtepaeet of placements the although tingly, eVtca xml bv,the above, example Vitaceae he lcmn hw . posterior 0.2 a shows placement mtepseiruigtestof set the using posterior the om nlaco emlongii Enaliarctos ). and wvr hspaeeti consistent is placement this owever, aoal eut Fg ) The 3). (Fig. results easonable ihe ro.Ti nepeainis interpretation This prior. richlet Cissus ecnann l osl.Tefull The fossils. 6 all containing de otlsmagnus Pontolis rs rnhlnt priors. length branch cross tal. et vshge cuayta the than accuracy higher eves Allodesmus and 05.Tepaeetof placement The 2015). l xatmme of member extant nly Vitis Ampelocissus cltbac length branch ichlet or eov npositions in resolve Nekemias smr rai in erratic more is and , eirdensity terior laces Allodesmus Vitis ne the under Pontolis nthe in Author Manuscript Nandinia_binotata Cryptoprocta_ferox hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This Genetta_genetta Ichneumia_albicauda Paradoxurus_hermaphroditus Civettictis_civetta Galidia_elegans Ursus_americanus Tremarctos_ornatus Ailuropoda_melanoleuca Pontolis_magnus_0.31 Fossa_fossana Melursus_ursinus Mephitis_mephitis Lynx_rufus Meles_meles Otocyon_megalotis Eupleres_goudotii Potos_flavus Gulo_gulo Vulpes_vulpes Crocuta_crocuta Ailurus_fulgens Acinonyx_jubatus Procyon_lotor Enaliarctos_emlongii_0.23 Martes_pennanti Nasua_nasua Enhydra_lutris Eumetopias_jubatus Arctocephalus_gazella Callorhinus_ursinus Zalophus_californianus Otaria_flavescens Allodesmus_sp._0.18 Neophoca_cinerea Monachus_monachus 17 Arctocephalus_australis Phoca_largha Phoca_vitulina Pusa_sibirica Erignathus_barbatus Mirounga_leonina Odobenus_rosmarus Histriophoca_fasciata Pusa_caspica Cystophora_cristata Piscophoca_pacifica_0.78 Halichoerus_grypus Hydrurga_leptonyx Acrophoca_longirostris_0.79 Author Manuscript xetwe lcmnsaeecpinlysal uha wit as such stable exceptionally are display placements placements, when fossil except the in uncertainty of estimates support. posterior display m names represent tip lengths fossil Branch cl trees. maximum of the distribution as posterior displayed are Placements prior. Dirichlet otnostat a eetoe iceetat ne cer under traits discrete over benefit can traits continuous characters: morphological discrete vs priors. Continuous across b similar same placements the fossil exhibit and not does dataset supp carnivoran posterior The of Vitaceae. sensitivity apparent examina this Closer of chang prior. significance exponential not the do above) to (listed compared behavior Dirichlet stable i with fossils taxa several six across the higher is prior Dirichlet compound seilyiprati rge aasc as such taxa ‘rogue’ in important especially ipae yteqatttv risi hpdb uhdisco such by shaped is traits geneal quantitative to the due by estimated displayed c be more cannot become tree may scaffolding also reliable procedure weighting t The steps high. additional becomes of possibility wi the work and Further discrepancy, lower. this is signal noisy/misleading significant of simulati performs amount the procedure by weighting shown the is high, as becomes addition, In variation. the of remove source to transformation cautious Procrustes be using should coordinates authors data, morphometric consider geometric to caveats to several are there promising, generally approach: the to datasets. Caveats such gather and times different model, important at combine, be data will morphological it of develops, sources area different this in progress As into information here. discrete struct such of combine gain to and incorporati straightforward loss the the used, as are such that qualitatively, data described continuous of (Parins-Fu type datasets the morphological on from gleaned potent be and can 1998), that (Rae schemes discretization from stemming hsmyb h aewe n ftecnen ttdb Felsens by stated concerns the of any when case the be may This extra be cannot characters phylogenetically-informative oooisacpe ssafligtes rb xedn t extending by or discordance. t trees, to scaffolding extensions as future accepted in topologies accommodated be principle in could otnosvrainit iceectgre (Goloboff categories discrete direc into traits variation continuous continuous analyze that methods principle, In nbt aaes lcmn ne h xoeta rnhle branch exponential the under placement datasets, both In iue3. Figure rvoswr netgtn h erdto fphylogenet of degradation the investigating work Previous lhuhtepromneo hsnwapoc nsmltda simulated on approach new this of performance the Although ept h oeta tlt npyoeeis hr a b may there phylogenetics, in utility potential the Despite hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This oslpaeet nerdfo h anvrndtstusi dataset carnivoran the from inferred placements Fossil ii magnisperma Vitis 18 tal. et tdfo emti opoercdata. morphometric geometric from cted a otfisteapoc hnnoise when approach the fortifies hat n eeal o otro support, posterior low generally ing r esrst ro hieosre in observed choice prior to measures ort l r rfrbeoe hs htbin that those over preferable are tly ae,adtems prpit asto ways appropriate most the and cales, ancrusacs(Revell circumstances tain rs a ehlfl twudbe would It helpful. be may ures, h iaeedtst h oiin of positions The dataset. Vitaceae the n nisue hnapyn hsmethod this applying When use. its in rhlgcldsaiy austrailing Values disparity. orphological emdlt comdt such accommodate to model he gcldsodne rweesignal where or discordance, ogical h opoercfaeokdescribed framework morphometric the dne(Mendes rdance fet frtto n3 pc sa as space 3D in rotation of effects n,we h inlt-os ratio signal-to-noise the when ons, no etrsta a euniquely be can that features of on uh 07.Nvrhls,depending Nevertheless, 2017). kuchi obte nesadtebhvo of behavior the understand better to infiatyudrtecompound the under significantly e yls cuaeyta hnthe when than accurately less ly ini eddt etrdtriethe determine better to needed is tion h d rdblt umr fthe of summary credibility ade hvo,wt ohbac support branch both with ehavior, lb eddt sestesuc of source the assess to needed be ll emto yrlxn h ubrof number the relaxing by method he opoel lg landmark align properly to 06,det hi viac ferror of avoidance their to due 2006), ae hr useful where cases e csga vrtm a mle that implied has time over signal ic a obte rsreinformation preserve better to ial meoisswildei Ampelocissus rnhspotudrthe under support Branch . ghpiryed conservative yields prior ngth en(02 antb overcome, be cannot (2002) tein mlctdi ae hr a where cases in omplicated deprcldt appears data empirical nd gtecompound the ng tal. et 08.This 2018). hsis This . tal. et 2008). Author Manuscript hlgntcpaeeto osltx.Atog ti ots most is Revell it by Although developed taxa. fossil of placement phylogenetic approaches: other data. to adding a Comparison by that except unlikely achieved is be it can occurs, placement is sho this to When due information sampling. weak taxonomic present simply may quali fossils or where proportions, and measurements linear traditional ySihadHnrcs(03 hnuigasm-adakgeo semi-landmark a in using variation when morphological (2013) Hendricks and Smith inconsis by exhibit characters geometrically-defined when or eeoeet coscaatr.Ti a endn ncont in done been exte has model This explore characters. to across important heterogeneity be will a it However, traits the equal. rescaled be I above, an performed under analyses traits i empirical continuous gap all major that One assumption collection. the data concerns of stage r the analysis at data subjectively quantitative through encou signal would in This concordance judgement. by subjective datasets through morphological than in rather bias of sources approach common data-centric an reduces more taxa a across of development densely sampling the when facilitate p occur by may datasets that morphometric noise large the very data to empirical method and the simulated applying the on a calibrations approaches weight of variety the a under collected data morphometric total-e into (Zhang data T continuous questions. incorporating biological of certain feasibility addressing in useful be adapting may approach, implem data-filtering native the a including feature describe, not does RevBayes RevBay Although as such framework. packages, phylogenetics existing in implement analysis. fo dating of the support in posterior probabilities using pr by in approach, could, placement which fossil 2018), (Guindon dating phylogenet molecular the Bayesian in in uncertainty calibrations accommodates fossil that of developed placement insta the For for questions. priors certain generate to suited proces better estimation and the datasets, splitting canon, t phylogenetic Although the relationships. in extant and extinct m ’total-evidence’ both recent estimate from differs f also achieved accuracy here the described St upon and improve Berger can of matrices adaptation character the Finally, taxa. her fossil implementation extinct the b addition, lengths In branch that procedure. require estimation not does approach my instance, h ehdta eciehr ifr usatal rme from substantially differs here describe I that method The oigfrad twl eipratt xlr h behavior the explore to important be will it forward, Moving wo here describe I that method the that noting worth also is It tal. et hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 05 . 2015) tal. et 21) tetnsterapoc nsvrlipratways important several in approach their extends it (2015), Conus nis nteecss tmyb eesr orsr ousing to resort to necessary be may it cases, these In snails. 19 iec opooia lc analyses clock morphological vidence c,teapoc eemyb sdto used be may here approach the nce, h rsn rcdr oti framework this to procedure present the aiecaatr.Fnly hr r cases are there Finally, characters. tative nosae a ebnfiili certain in beneficial be may stages into s oedtn.Anwmto a been has method new A dating. node tleiec ehd r sfltools useful are methods otal-evidence tosta ekt simultaneously to seek that ethods te hnb tepigt filter to attempting by than ather losfrteetmto flong of estimation the for allows e eto eksga,sc swsfound was as such signal, weak or tent ahst ota h ainei e to set is variance the that so site each t mtks praht filtering to approach amatakis’ lsseov ne hrdrt.I the In rate. shared a under evolve alysis omrhlgclpyoeeisthat phylogenetics morphological to slclbain steprior the as calibrations ssil i a nld nepoaino the of exploration an include may his essget h osblt of possibility the suggests sets noscaatr opstv fetby effect positive to characters inuous h prahpeetdhere presented approach the n mlrt h oslpaeetmethod placement fossil the to imilar cpaeeto oeclbain in calibrations node of placement ic aea xlrto fcnitand conflict of exploration an rage dsmln cee.Tescesof success The schemes. sampling nd cldt ie ipiyn the simplifying time, to scaled e o xsigmtos h method The methods. existing rom ygetrcranyi their in certainty greater ny sosta comdt rate accommodate that nsions oiigamast le through filter to means a roviding leigdt arcsstatistically matrices data filtering toig ngooi and geologic in rtcomings nil,b obndwt my with combined be inciple, itn prahst the to approaches xisting ras uhafaeokwould framework a Such organs. d s n dpe oatotal-evidence a to adapted and es, fti ehdwe ple to applied when method this of l esrihfradto straightforward be uld ercmto ocapture to method metric naino h oe htI that model the of entation For . Author Manuscript Schraiber usadn usin costete flife. of e tree better the will across here questions collecti developed outstanding the approach in the advances to New refinements era. with genomic the in better collection, scale character approach in the subjectivity using surrounding data morphologic issues morphometric for of important Analysis be will pace. it keep life, data of genomic tree of the use across its in advances ex phylogenetics the molecular in meth displayed these uncertainty of of refinement assessment and further conservative simulated for the need Although a life. and of imperfections tree the in fossils place Conclusions: method. the extensions an future (Huelsenbeck in heterogeneity rate modeling to approaches h ehddsrbdhr rvdsanwmasfrbiologist for means new a provides here described method The tal. et hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 21)uigaGmast-aemdl n dpigti ral or this adapting and model, site-rate Gamma a using (2013) 20 n ilhl opooia aaesto datasets morphological help will and mlsapa norgn.As encouraging. appear amples lpyoeeisadplotlg to paleontology and phylogenetics al miia nlssso several show analyses empirical uhr 07 ilb e priority key a be will 2007) Suchard d no opoercdt,combined data, morphometric of on upmrhlg osekt major to speak to morphology quip oase udmna questions fundamental answer to d,teoealacrc and accuracy overall the ods, hw eewl ept improve to help will here shown orlal n confidently and reliably to s ternative Author Manuscript ooof .A n aaao .A 06 n eso .,incl 1.5, version Tnt 2016. A. S. Catalano, and A. P. Goloboff, ooof .A,Mtoi .I,adQitrs .S 06 Cont 2006. S. A. Quinteros, and I., C. Mattoni, A., P. Goloboff, esnti,J 98 hlgne n uniaiecharac quantitative and Phylogenies 1988. J. Felsenstein, akn,J . uhs .E,adSoln,R .19.Primary 1997. W. R. Scotland, and E., C. Hughes, A., chara J. ordered Hawkins, of effect The 1991. W. Presch, and L. D. Hauser, com for calculations likelihood Fast 2012. P. R. Freckleton, phylogeni characters, Quantitative 2002. J. Felsenstein, atns .K 90 ot al apigmtosuigmar using methods sampling carlo Monte 1970. K. W. Hastings, uncertainty: calibration for Accounting 2018. S. Guindon, Pucc and R., Cúneo, A., W. Neves, I., Escapa, R., González-José, ön,S,Lni,M . et,T . osa,B,Lartillot, B., Boussau, A., T. Heath, J., M. Landis, S., Höhna, aaao .A,Glbf,P . n inii .P 00 Phy 2010. P. N. Giannini, and A., P. Goloboff, A., S. Catalano, egr .A n tmtks .21.Acrc fmorphology of Accuracy 2010. A. Stamatakis, and A. S. Berger, ulr .A n ig .A 04 hlgntccmaaiea comparative Phylogenetic 2004. A. A. King, and A. M. Butler, in methods coding character Problematic 2011. D. M. Brazeau, aal-fra .L n dad,A .16.Pyoeei a Phylogenetic 1967. W. A. Edwards, and L. L. Cavalli-Sforza, mo,E n eMio,C 04 e uohgu hcd(mam phocid durophagous new A 2014. C. Muizon, de and E. Amson, References hn .2009. I. Chen, esnti,J 91.Eouinr re rmgn freque gene from trees Evolutionary 1981b. J. Felsenstein, sequenc dna from trees Evolutionary 1981a. J. Felsenstein, esnti,J 93 aiu-ieiodetmto fev of estimation Maximum-likelihood 1973. J. Felsenstein, hlgntcmorphometrics. phylogenetic Evolution and such. hrcesadcaatrstates. character and characters reconstruction. Biometrika phylogeny and shape Systematics dul nrcal”problem. intractable” “doubly phylog provides Nature traits modularized continuous of analysis s flnmr aai hlgntcframework. phylogenetic a in data landmark of use Palaeontology procedures. evolution. adaptive Society Linnean the of Journal Biological 1–9. pages likelihood. maximum under placement aenoeeo euadcnieain nmnciesasp seals monachine on considerations and peru of neogene late nigmxmmlklho estimates. likelihood maximum finding Evol. Mol. J. characters. eodo seeds of record Cladistics 5(16:775. 453(7196): , hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This 71:97–109. 57(1): , m .Hm Genet. Hum. J. Am. 91:445–471. 19(1): , m .Hm Genet. Hum. J. Am. itr fVtca nerdfo opooybsdphylog morphology-based from inferred Vitaceae of History 76:368–376. 17(6): , () 940–947. 3(5): , 25:523–548. 12(5): , nvriyo Florida. of University . Cladistics 26:589–601. 22(6): , m Nat. Am. ae 27–44. pages , () 243–265. 7(3): , 6() 683–695. 164(6): , Cladistics yt Biol. Syst. 55:471. 25(5): , Cladistics 93:233. 19(3): , Evolution 23:221–238. 32(3): , 0() 489–498. 104(3): , 33:275–283. 13(3): , 21 56:1229–1242. 35(6): , Cladistics s n morphometrics. and es, ters. ce n uniaiecharacters: quantitative and ncies aeinmlclrdtn sa as dating molecular Bayesian s aiu ieiodapproach. likelihood maximum a es: ltoaytesfo continuous from trees olutionary aaieanalyses. parative ntcsgasi ooevolution. homo in signals enetic . or,B . Huelsenbeck, R., B. Moore, N., nulRve fEooyand Ecology of Review Annual dn ulipeetto of implementation full a uding oeei opoerc i:the (i): morphometrics logenetic ayi.mdl n estimation and models nalysis. tr nphylogenetic on cters noscaatr nlzdas analyzed characters inuous opooyadtereffects. their and morphology hylogeny. 65:539–549. 26(5): , o hisadterapplications. their and chains kov ayi:amdln prahfor approach modeling a nalysis: bsdpyoeei fossil phylogenetic -based ael,H .20.Cladistic 2008. M. H. iarelli, oooyassessment, homology ai:Crioa rmthe from Carnivora) malia: ora fSystematic of Journal n n h fossil the and eny ehd nEcology in Methods Morphology, Author Manuscript ulebc,J .adScad .A 07 oprmti m nonparametric A 2007. A. M. Suchard, and P. J. Huelsenbeck, oqit . atlo,N,adPilp,M .21.Closing 2016. J. M. Phillips, and N., Lartillot, F., Ronquist, eel .J,Mhe,D . enls .G,adSae,G .20 J. G. Slater, and G., R. Reynolds, L., D. Mahler, J., Phylogeneti L. 2008. Revell, C. D. Collar, and J., L. Harmon, J., L. co Revell, phylogenetic for package r an phytools: 2012. J. L. i Revell, partial paradox, Tail 2011. Z. Yang, and T., char Zhu, continuous B., of Rannala, use the for basis logical The 1998. C. T. Rae, oqit . lpsen . ihlsn . Schulmeister L., Vilhelmsen, S., Klopfstein, F., Ronquist, eds .K,FetsGnae,J . crie,J . an G., J. Schraiber, A., J. Fuentes-Gonzalez, K., F. Mendes, dat morphometric in signals Phylogenetic 2002. N. MacLeod, pale quantitative in phylogeny of role The 2001. N. MacLeod, phylo estimating to approach likelihood A 2001. O. P. Lewis, pac software integrated an Morphoj: 2011. P. C. Klingenberg, ’er,B . n,C,Snesn .J,adWiwih,P .20 C. P. Wainwright, and J., M. Sanderson, C., Ané, C., B. O’Meara, Phylogen 2013. M. Liang, and G., J. Schraiber, J., M. Landis, the of Impact 2015. A. Goswami, and B., J. Smaers, E., K. Jones, oeTa 2016. Team Core R a fossils using estimation time Divergence reconst 2011. phylogeny A. for R. coding Pyron, character On 1995. F. Pleijel, improv can traits continuous of Use 2017. C. Parins-Fukuchi, crie,J . otvy . s,T . n rm .B 03 I 2013. B. R. Brem, and Y., T. Hsu, Y., Mostovoy, G., J. Schraiber, etn cosst aevariation. rate across-site testing language. i model-specification interactive phylogenetic an Bayesian and Revbayes: models 2016. F. Ronquist, and P., J. sn oa-vdnedating. total-evidence using aesuywt h iadaoi roosevelti. anolis lizard the phylogen with ultrametric study an case into taxa hypothesized or extinct, rate. and things). inference. length branch bayesian in systematics. Austria. Vienna, Computing, Statistical 02 oa-vdneapoc odtn ihfsis app fossils, with dating hymenoptera. to approach total-evidence A 2012. fcniuu ri vlto sn likelihood. using evolution trait continuous of under by affected bioRxiv are traits quantitative about inferences phylogeny Paleobiology data. character traits. continuous of evolution the in jumps finding resources ecology Molecular skull. carnivoran the in evolution of rates and disparity lissamphibia. Biol. Syst. ae276642. page , ehd cl Evol. Ecol. Methods yt Biol. Syst. 0:138. 100: , hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This Cladistics 72:226–240. 27(2): , yt Biol. Syst. yt Biol. Syst. yt Biol. Syst. :ALnug n niomn o ttsia Computing Statistical for Environment and Language A R: 74:591–601. 57(4): , 43:221–228. 14(3): , 04:466–481. 60(4): , 16:973–999. 61(6): , 06:913–925. 50(6): , hl rn.R o.B Soc. R. Trans. Phil. () 217–223. 3(2): , 12:353–357. 11(2): , ytmtcBiology Systematic oeua ilg n evolution and biology Molecular Evolution 22 Evolution 7(69:20150136. 371(1699): , yt Biol. Syst. 66:975–987. 56(6): , 94:1027–1035. 69(4): , . ury .L,adRsisn .P. A. Rasnitsyn, and L., D. Murray, S., , M vltoaybiology evolutionary BMC yn eelgcldiscordance. genealogical lying 05:922–933. 60(5): , ruction. an .W 08 Evolutionary 2018. W. M. Hahn, d yt Biol. Syst. sn otnoscaatrdt:A data: character continuous using y h a ewe ok n clocks and rocks between gap the eyfo iceemorphological discrete from geny opooia phylogenetics. morphological e idt h al aito fthe of radiation early the to lied tcaayi sn éyprocesses: lévy using analysis etic prtv ilg adother (and biology mparative aefrgoercmorphometrics. geometric for kage emnltx n h rgn of origins the and taxa terminal s a. booia aaanalysis. data obiological etfiblt,adifleta priors influential and dentifiability, 22:193–204. 62(2): , 5 lcn rpi,recently cryptic, Placing 15. inl vltoayprocess, evolutionary signal, c cesi phylogenetic in acters opooy hp and shape Morphology, to o comdtn and accommodating for ethod ersra-qai rniinon transition terrestrial-aquatic frneuiggraphical using nference frigeouinr histories evolutionary nferring Cladistics 6 etn o ifrn rates different for Testing 06. 54:726–736. 65(4): , 91:325–335. 29(1): , 13:309–315. 11(3): , onainfor Foundation R . 51:8. 15(1): , Author Manuscript in,J .20.Plotlg,gnmc,adcombined-da and genomics, Paleontology, 2009. J. J. Wiens, phylo morphological in analysis Character 2001. J. the J. Wiens, character: perfect the of grail holy The 1993. K. libr Thiele, python a Dendropy: 2010. T. M. Holder, and J. Sukumaran, ikno,M 95 oprsno w ehd fcharacter of methods two of comparison A 1995. M. Wilkinson, in,J . uznk,C . oned . edr .W,Mul W., T. Reeder, T., Townsend, A., C. Kuczynski, J., J. Wiens, ag .2014. Z. Yang, hn,C,Salr . lpsen . et,T . n Ronqu and A., T. Heath, S., Klopfstein, T., Stadler, C., Zhang, rgt .M,Lod .T,adHli,D .21.Mdln ch Modeling 2016. M. D. Hillis, and T., G. Lloyd, M., A. Wright, rgt .M n ils .M 04 aeinaayi sn s a using analysis Bayesian 2014. M. D. Hillis, and M. A. Wright, ctad .adPnigo,R .2000. T. R. Pennington, and R. Scotland, ctad .W,Omta,R . n ent,J .20.Phyloge 2003. R. J. Bennett, and G., R. Olmstead, W., R. Scotland, iõs .R,Clwl,M . ac,A,adNdm .L 07 G 2017. L. R. Nydam, and A., Palci, W., M. Caldwell, R., T. Simões, ltr .J,Hro,L . n lao .E 02 Integrat 2012. E. M. Alfaro, and J., L. Harmon, J., G. Slater, ots .E,Sih .A,Cliee . udc,K . Tan J., K. Wurdack, N., Cellinese, A., S. Smith, E., D. Soltis, mt,U .adHnrcs .R 03 emti morphometri Geometric 2013. R. J. Hendricks, and E. U. Smith, mrv hlgn siainfrfsi taxa? fossil for estimation phylogeny improve Biol. Syst. data. Bioinformatics oeua aacag h lcmn ffsi taxa. fossil of placement the s change higher-level data in molecular fossils and phylogenomics Combining 2010. ne h oslzdbrhdahprocess. birth–death fossilized the under hlgntcaaye fmrhlg hog h s fprio of use the through morphology of analyses phylogenetic dis from phylogeny One of estimation for parsimony outperforms 297–308. hlgntcanalysis phylogenetic data. profiling e1003255. transcriptional from regulation pathway of morphology. arcs ult fcaatrcntutosrmiscrit remains constructions character of quality matrices: nisempyoey 7gns 4 taxa. 640 genes, 17 phylogeny: Angiosperm hlgne mrvsifrneo ri evolution. trait of inference improves phylogenies 198–219. 33(2): eui-orge,N . akr .B,Moe .J,Carlsward J., M. Moore, B., J. Walker, F., N. Refulio-Rodriguez, aa xrcigpyoeei inlfo atoo shell gastropod from signal phylogenetic extracting data: (0:e109210. 9(10): , Cladistics 05:689–699. 50(5): , hsatcei rtce ycprgt l ihsreserved. rights All copyright. by protected is article This oeua vlto:asaitclapproach statistical a evolution: Molecular yt Biol. Syst. 61) 1569–1571. 26(12): , () 275–304. 9(3): , R Press. CRC . 24:539–548. 52(4): , oooyadsseais oigcaatr for characters coding systematics: and Homology ytmtcbiology Systematic mrcnJunlo Botany of Journal American 23 ytmtcBiology Systematic Evolution yt Biol. Syst. clrgrls fsize. of regardless ical xodUiest Press. University Oxford . s. ldsi ramn fmorphometric of treatment cladistic LScmuainlbiology computational PLoS n osl ihmolecular with fossils ing apyoeeis a oeua data molecular can phylogenetics: ta eeis rbesadsolutions. and problems genetics: ,D . rcigo,S F., S. Brockington, C., D. k, yt Biol. Syst. s,F 05 oa-vdnedating Total-evidence 2015. F. ist, 52:228–249. 65(2): , 61) 3931–3944. 66(12): , rt opooia data. morphological crete rs. r o hlgntccomputing. phylogenetic for ary umt etl phylogeny: reptile quamate hrce utsa phylogenetic as suites character c 96:674–688. 59(6): , ay .G,adSts .W. J. Sites, and G., D. cahy, construction. rce hnehtrgniyin heterogeneity change aracter ml ieiodmodel likelihood imple yt Biol. Syst. .S., B. , yrcntuto:terl of role the reconstruction: ny 81:87–99. 58(1): , attaxon-character iant 2 366–385. 62: , tal. et 84:704–730. 98(4): , 54:602–611. 65(4): , Cladistics 2011. Cladistics 11(3): , 9(10): , PLoS ,