The Phylogeny of the Ornithischian

Journal of Systematic Palaeontologyhttp://journals.cambridge.org/SYP

Additional services for Journal of Systematic Palaeontology:

Email alerts: Click hereSubscriptions: Click hereCommercial reprints: Click hereTerms of use : Click here

The phylogeny of the ornithischian dinosaurs

Richard J. Butler, Paul Upchurch and David B. Norman

Journal of Systematic Palaeontology / Volume 6 / Issue 01 / March 2008, pp 1 - 40DOI: 10.1017/S1477201907002271, Published online: 25 September 2007

Link to this article: http://journals.cambridge.org/abstract_S1477201907002271

How to cite this article:Richard J. Butler, Paul Upchurch and David B. Norman (2008). The phylogeny of the ornithischian dinosaurs. Journal ofSystematic Palaeontology, 6, pp 1-40 doi:10.1017/S1477201907002271

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/SYP, IP address: 132.248.181.49 on 11 Aug 2014

Journal of Systematic Palaeontology 6 (1): 140 Issued 22 February 2008doi:10.1017/S1477201907002271 Printed in the United Kingdom C The Natural History Museum

The phylogeny of the ornithischiandinosaursRichard J. ButlerDepartment of Earth Sciences, University of Cambridge, Downing Street, Cambridge, CB2 3EQ, UK &Department of Palaeontology, The Natural History Museum, Cromwell Road, London, SW7 5BD, UK

Paul UpchurchDepartment of Earth Sciences, University College London, Gower Street, London, WC1E 6BT, UK

David B. NormanDepartment of Earth Sciences, University of Cambridge, Downing Street, Cambridge, CB2 3EQ, UK

SYNOPSIS Ornithischia is a familiar and diverse clade of dinosaurs whose global phylogeny hasremained largely unaltered since early cladistic analyses in the mid 1980s. Current understanding ofornithischian evolution is hampered by a paucity of explicitly numerical phylogenetic analyses thatconsider the entire clade. As a result, it is difcult to assess the robustness of current phylogenetichypotheses for Ornithischia and the effect that the addition of new taxa or characters is likely to haveon the overall topology of the clade.

The new phylogenetic analysis presented here incorporates a range of new basal taxa and charac-ters in an attempt to rigorously test global ornithischian phylogeny. Parsimony analysis is carried outwith 46 taxa and221 characters. Although the strict component consensus tree showspoor resolutionin a number of areas, application of reduced consensus methods provides a well-resolved pictureof ornithischian interrelationships. Surprisingly, Heterodontosauridae is placed as the most basalgroup of all well-known ornithischians, phylogenetically distant from a stem-dened Ornithopoda,creating a topology that is more congruent with the known ornithischian stratigraphical record. Thereis no evidence for a monophyletic Fabrosauridae, and Lesothosaurus (the best-known fabrosaur)occupies an unusual position as the most basal member of Thyreophora. Other relationships withinThyreophora remain largely stable. The primitive thyreophoran Scelidosaurus is the sister taxon ofEurypoda (stegosaurs and ankylosaurs), rather than a basal ankylosaur as implied by some previousstudies.

The taxonomic content of Ornithopoda differs signicantly from previous analyses and basalrelationships within the clade are weakly supported, requiring further investigation. Hypsilopho-dontidae is paraphyletic, with some taxa (Agilisaurus, Hexinlusaurus, Othnielia) placed outsideof Ornithopoda as non-cerapodans. Ceratopsia and Pachycephalosauria are monophyletic and areunited as Marginocephalia; however, the stability of these clades is reduced by a number of poorlypreserved basal taxa.

This analysis reafrms much of the currently accepted ornithischian topology. Nevertheless, in-stability in thepositionandcontent of several clades (notablyHeterodontosauridaeandOrnithopoda)indicates that considerable future work on ornithischian phylogeny is required and causes problemsfor several current phylogenetic denitions.

KEY WORDS vertebrate palaeontology, Ornithischia, systematics, cladistics, Dinosauria

E-mail: [email protected]; Email: [email protected]; Email: [email protected]

Contents

Introduction 1Institutional abbreviations 2

Previous analyses of ornithischian phylogeny 2Traditional classications 2Cladistic studies 3

Material and Methods 5The aim of the analysis 5Phylogenetic framework 5Selection of ingroup taxa 5

2 R. J. Butler et al.

Supraspecic taxa 5Included species level taxa 7Excluded species level taxa 11

Selection of outgroup taxa 11

Analyses 12Search methods 12Testing the support for relationships 12

Randomisation tests 12Bremer support 12Bootstrapping 14

Results 16Ornithischian monophyly 16Pisanosaurus mertii 17Heterodontosauridae 17Fabrosaurids 18Thyreophora 19Basal neornithischians 19Ornithopoda 20Hypsilophodontidae 20Marginocephalia 21Ceratopsia 21Pachycephalosauria 22

Conclusions 22Stability, instability, and future directions in ornithischian phylogeny 22Implications for phylogenetic taxonomy of Ornithischia 22

Acknowledgements 23

References 23

Appendix 1 Specimens and references used for coding operational taxonomic units 29

Appendix 2 Character list 30

Appendix 3 Data matrix 35

Appendix 4 Tree descriptions 40

Introduction

The clade Ornithischia represents a major grouping withinDinosauria, the most familiar and widely popularised ofall extinct organisms. An ornithischian, Iguanodon Mantell,1825, was the second dinosaur genus to be named, whileof the three taxa explicitly included within Dinosauria byOwen (1842), two (Iguanodon and Hylaeosaurus) were laterrecognised as ornithischians. Since these early discoveries, alarge number of genera and species of ornithischians havebeen named: a recent review (Weishampel et al. 2004a)recognised over 180 valid genera. The earliest ornithischi-ans are known from the Carnian stage of the Late Trias-sic (Casamiquela 1967) and the clade disappeared in themass extinctions at the end of the Cretaceous. During thistime ornithischians achieved a global distribution and arenow known from every continent, including Antarctica (e.g.Hooker et al. 1991; Weishampel et al. 2004b).

Ornithischians appear to have been extremely scarceduring the Late Triassic (Sereno 1997) and remained un-common (although apparently more diverse) during the Earlyand Middle Jurassic; during this time interval terrestrial ver-

tebrate faunas are dominated by saurischians (Weishampelet al. 2004b). Ornithischians became much more abundantduring the Late Jurassic and Early Cretaceous and ornith-ischian diversity peaked during the Campanian stage of theLate Cretaceous. Late TriassicEarly Jurassic ornithischianswere generally small-bodied (12 m in length) bipedal curs-ors (e.g. Lesothosaurus diagnosticus: Thulborn 1972; Sereno1991a), but during the rest of the Mesozoic they diversifiedinto a considerable range of morphologies and sizes, withmany groups reverting to quadrupedality. The vast majorityof ornithischians are believed to have been herbivorous, al-though some basal forms have been interpreted as potentiallyomnivorous (Barrett 2000).

One major problem in understanding ornithischianevolution is that, to date, there are few published numer-ical phylogenetic analyses dedicated solely to Ornithischia(Weishampel 2004). It is difficult, therefore, to assess the re-lative phylogenetic support for ornithischian clades, to lookat the effects upon phylogenetic results of adding or deletingtaxa or characters, or to test alternative hypotheses of taxonor character evolution. Here, we present a new phylogen-etic analysis of Ornithischia, as part of an on-going study ofornithischian phylogeny.

Phylogeny of ornithischian dinosaurs 3

Institutional abbreviations

BMNH = Natural History Museum, London, UKBRSMG = Bristol City Museum and Art Gallery, Bris-

tol, UKBP = Bernard Price Institute for Palaeontological

Research, Johannesburg, South AfricaBYU = Earth Science Museum, Brigham Young

University, Provo, Utah, USACAMSM = Sedgwick Museum, University of Cam-

bridge, Cambridge, UKCEUM = Prehistoric Museum, College of Eastern

Utah, Price, Utah, USACV = Chongqing Natural History Museum,

Chongqing, Peoples Republic of ChinaGCC = Museum of the Chendgu University of

Technology (formerly Chendgu College ofGeology), Chengdu, Peoples Republic ofChina

GI = Geological Institute, Ulaanbaatar, Mongo-lia

GSM = Geological Survey Museum, Keyworth,UK

GZG = Geowissenschaftliches Zentrum der Uni-versitat Gottingen, Gottingen, Germany

IGCAGS = Institute of Geology, Chinese Academyof Sciences, Beijing, Peoples Republic ofChina

IVPP = Institute of Vertebrate Paleontology andPaleoanthropology, Bejing

MCZ = Museum of Comparative Zoology, Harvard,USA

MB = Museum fur Naturkunde, Berlin, GermanyMCF-PVPH = Museo Carmen Funes, Paleontologa de

Vertebrados Plaza Huincul, ArgentinaMNA = Museum of Northern Arizona, Flagstaff,

USAMOR = Museum of the Rockies, Bozeman,

Montana, USAMPM = Museo Padre Molina, Rio Gallegos, Santa

Cruz, ArgentinaMWC = Museum of Western Colorado, Grand Junc-

tion, Colorado, USAOUM = University Museum of Natural History, Ox-

ford, UKPVL = Fundacion Miguel Lillo, Universidad

Nacional de Tucuman, San Miguel de Tu-cuman, Argentina

ROM = Royal Ontario Museum, Toronto, Ontario,Canada

SAM-PK = Iziko South African Museum, Cape Town,South Africa

SDSM = South Dakota School of Mines, Rapid City,South Dakota, USA

SGWG = Sektion Geologische Wissenschaften Gre-ifswald, Ernst-Moritz Universitat, Greif-swald, Germany

UC = University of Chicago, Chicago, USAUCMP = University of California Museum of Pale-

ontology, Berkeley, USAYPM = Peabody Museum of Natural History,

Yale University, New Haven, Connecticut,USA

ZDM = Zigong Dinosaur Museum, Dashanpu,Peoples Republic of China

ZPAL = Institute of Paleobiology of the PolishAcademy of Sciences, Warsaw, Poland.

Previous analyses ofornithischian phylogeny

Traditional classications

Although material from the clade began to be described andnamed in the early 19th century (Mantell 1822, 1825, 1833),the recognition that ornithischians formed a grouping dis-tinct from other reptiles (including other dinosaurs) was notreached until the work of Seeley (1887). Seeley was the firstto identify and articulate a fundamental morphological dicho-tomy within the then described dinosaurian taxa. He recog-nised two orders, distinguished mainly on the basis of pelvicanatomy, that he named Ornithischia and Saurischia. Majorsubdivisions of Seeleys Ornithischia (Stegosauria, Ornitho-poda, Ceratopsia) were identified by Marsh (1877a, 1881,1890) and Marsh (1894) later grouped these subdivisionstogether as the order Predentata, generally considered syn-onymous with Ornithischia.

Nopcsa (1915) proposed a subdivision within Ornithis-chia, between the bipedal, unarmoured forms (ornithopods)and a new suborder that he named Thyreophora (compris-ing ankylosaurs, stegosaurs and ceratopsians). The group-ing Thyreophora was often ignored by later workers, butthe name was revived by Norman (1984) and Sereno (1984,1986), although its current usage (for a clade consisting ofall armoured ornithischians) differs somewhat from that pro-posed by Nopcsa (i.e. in current usage ceratopsians are notmembers of Thyreophora).

Romer (1956) divided Ornithischia into four suborders:Ornithopoda (including all bipedal forms), Stegosauria, An-kylosauria and Ceratopsia. Most bipedal and relatively un-specialised taxa were included within Ornithopoda; othersuborders were believed to be derived from within the or-nithopods, although only the ceratopsians were explicitlylinked with a particular group of ornithopods (the psit-tacosaurids).

Thulborn (1971) suggested that most of the majorgroups of ornithischians (e.g. iguanodontids, pachycephalo-saurs, ceratopsians) were descended from a Late TriassicLate Cretaceous hypsilophodont plexus, an implicitly para-phyletic grade of small, primitive, bipedal ornithopods. An-kylosaurs and stegosaurs were considered to occupy a prim-itive position outside of this plexus, sharing ancestors withthe earliest hypsilophodontids. The details of this classifica-tion were questioned by Galton (1972), who removed Ech-inodon becklesii and Fabrosaurus australis from the familyHypsilophodontidae to form the family Fabrosauridae (seealso Galton 1978), which included those taxa he consideredto represent the most basal known ornithischians. Maryanska& Osmolska (1974) emphasised the morphological distinct-ness of the pachycephalosaurs, previously included as thefamily Pachycephalosauridae within Ornithopoda, and cre-ated a new suborder of Ornithischia, Pachycephalosauria.

In summary, pre-cladistic ornithischian classificationstended to recognise either four (Romer 1956) or five


(Maryanska & Osmolska 1974) suborders within Ornithis-chia. One of these suborders (Ornithopoda) was an implicitlyparaphyletic grouping of taxa, defined on the basis of ple-siomorphic characters (e.g. bipedality). Only a few workers(e.g. Thulborn 1971; Galton 1972) attempted to identify thepattern of interrelationships between or within these clades.

Cladistic studies

The first cladistic studies of ornithischian phylogeny werepublished simultaneously by Norman (1984), Milner &Norman (1984) and Sereno (1984); the results of Norman(1984) and Sereno (1984) are shown in Fig. 1A, B. Norman(1984) proposed that Ornithischia could be divided into twomajor groupings: Thyreophora, comprising the ankylosaursand stegosaurs, and Ornithopoda, which Norman expanded toinclude ceratopsians. Norman positioned fabrosaurs as basalornithopods, suggested that ceratopsians and iguanodontians(referred to as dryosauroideans in his cladogram) shareda common ancestor to the exclusion of hypsilophodontids,but considered the position of heterodontosaurids and pachy-cephalosaurs to be problematic and unresolved. The phylo-geny of Milner & Norman (1984) concentrated on relation-ships within Ornithopoda and effectively represented a subsetof the analysis of Norman (1984).

The phylogeny presented by Sereno (1984) differedsignificantly from that of Norman (1984). Sereno proposedthat Fabrosauridae was polyphyletic and positioned Lesotho-saurus diagnosticus (previously included within Fabrosaur-idae) as the most basal known ornithischian. His conceptionof Ornithopoda was much more restrictive than that of previ-ous workers and excluded ceratopsians, pachycephalosaursand fabrosaurs, while including heterodontosaurids. Hisconception of Thyreophora also differed substantially fromthat of Norman (1984), comprising a clade consisting of an-kylosaurs, stegosaurs, pachycephalosaurs and ceratopsians.

The phylogeny of Cooper (1985) was similar in manyaspects (Fig. 1C) to that of Sereno (1984). However, Cooperpositioned Heterodontosauridae as the sister group to thePachycephalosauriaCeratopsia clade and considered Fab-rosauridae to represent the most basal clade within Ornitho-poda.

Maryanska & Osmolska (1985) outlined a phylogeny(Fig. 1D) that differed in several key points from that ofNorman (1984) and Sereno (1984). Maryanska & Osmolskaproposed that ankylosaurs and stegosaurs did not share acommon ancestor but, instead, represented serial outgroupsto more derived ornithischians, and that heterodontosauridsformed the sister group to a clade consisting of Ornitho-poda (including Lesothosaurus diagnosticus), Pachycephalo-sauria and Ceratopsia. Following Sereno (1984), Maryanska& Osmolska united Pachycephalosauria and Ceratopsia toform a clade to the exclusion of ornithopods.

The most influential published work on ornithischianphylogeny was produced by Sereno (1986) and his res-ults (Fig. 1E) have dominated subsequent understandingof ornithischian phylogeny (see, e.g. Weishampel et al.1990, 1992, 2004a; Fastovsky & Weishampel 1996, 2005;Currie & Padian 1997; Sereno 1997, 1999a). Sereno (1986)modified his earlier (Sereno 1984) hypothesis by uniting or-nithopods, pachycephalosaurs and ceratopsians (followingNorman (1984) and Maryanska & Osmolska (1985)) in aclade that he termed Cerapoda. Within Cerapoda, hetero-

dontosaurids were positioned as basal ornithopods, whilethe clade containing Pachycephalosauria and Ceratopsia wasnamed Marginocephalia. Sereno followed Norman (1984) inrestricting Thyreophora to ankylosaurs, stegosaurs and twobasal armoured forms (Scutellosaurus lawleri, Scelidosaurusharrisonii), while thyreophorans and cerapodans were unitedas the clade Genasauria. Sereno continued to consider Fab-rosauridae polyphyletic and positioned Lesothosaurus dia-gnosticus as the sister taxon of Genasauria.

Following the work of Sereno (1986), ornithischianworkers tended to focus on relationships within the majorornithischian clades; e.g. the phylogeny of basal Ornitho-poda has been analysed by Weishampel & Heinrich (1992),Winkler et al. (1997), Scheetz (1998, 1999), Weishampelet al. (2003) and Norman et al. (2004c), amongst others.However, there have been only a few attempts to test theglobal phylogeny of Ornithischia.

Sereno (1997: figs 1, 2) presented an ornithischiancladogram, but this was not supported by a data matrix or in-formation about analyses. In a review paper, Sereno (1999a)considered ornithischians within a larger scale analysis of di-nosaurian phylogeny. This included nine separate data sets,each of which analysed separate portions of the dinosaur-ian tree. Four of these data sets (data sets 14) dealt withornithischians. Data set 1 analysed basal dinosaurian phylo-geny and, within this framework, tested both ornithischianmonophyly and interrelationships. Data sets 24 analysed thewithin-clade phylogeny of Thyreophora, Ornithopoda andMarginocephalia. Results generally supported the findings ofSereno (1986) and differed only in the inclusion of additionaltaxa (e.g. the basal thyreophoran Emausaurus ernsti) and theidentification of the Late Triassic taxon Pisanosaurus mertiias the most basal ornithischian. Monophyly of Thyreophora,Ornithopoda and Marginocephalia (and a taxonomic contentfor these clades consistent with the phylogenetic results ofSereno (1986)), was assumed prior to analysis.

Buchholz (2002) carried out an ornithischian analysiswith 27 taxa and 97 characters. Taxonomic sampling was re-stricted mostly to taxa generally considered as ornithopods,with Marginocephalia coded as a composite taxon and basalthyreophorans excluded from the analysis. Buchholz foundsupport for a sister-grouping of heterodontosaurids and mar-ginocephalians and for paraphyly of hypsilophodontids. Un-fortunately, although a character-list was published, a matrixwas not provided and these results cannot be reassessed.

Liu (2004) tested the global phylogeny of Ornithischiawith a large-scale analysis including 44 taxa and 326 char-acters. Interesting results were reported: Fabrosauridae andHypsilophodontidae were found to be paraphyletic, Leso-thosaurus was positioned as a basal member of Thyreo-phora, Agilisaurus louderbacki (often considered to repres-ent a basal ornithopod, e.g. Norman et al. 2004c) groupedoutside of Cerapoda and Marginocephalia and Iguanodontiawere united as sister groups. Unfortunately, this analysis has,to date, been published in abstract form only. It is, therefore,impossible, at this stage, to reanalyse the data. However, thereported results of Liu (2004) clearly highlight the import-ance of new phylogenetic analyses.

One of us (Butler 2005) included a cladistic analysis(23 taxa, 73 characters) of Ornithischia within a review of thefabrosaurid ornithischians of the upper Elliot Formation ofsouthern Africa. That analysis recovered interesting results,including positioning heterodontosaurids and the Middle


Figure 1 Previous numerical ornithischian phylogenies, simplied and redrawn from the original publications. A, Norman (1984); B, Sereno(1984); C, Cooper (1985); D, Maryanska & Osmolska (1985); E, Sereno (1986 ). H1, H2, alternative positions proposed by Norman (1984) for theclade Heterodontosauridae; P1, P2, alternative positions proposed by Norman (1984) for the clade Pachycephalosauria.


Jurassic taxa Agilisaurus louderbacki and Hexinlusaurusmultidens outside of Cerapoda; that analysis represented anearly iteration of the analysis presented here and is super-seded by the present study.

Although a number of major phylogenetic studies of Or-nithischia have been published (Norman 1984; Sereno 1984,1986, 1999a; Cooper 1985; Maryanska & Osmolska 1985),the majority have failed to include crucial information, in-cluding: (1) a charactertaxon data matrix; (2) details of thespecimens and references used in coding operational taxo-nomic units; (3) tree searching methods; (4) the number andscores of the most parsimonious trees recovered by searchmethods; (5) tests of data robustness and support for particu-lar clades (e.g. bootstrapping, decay analysis). Most previousstudies have simply presented a fully-resolved tree and listsof apomorphies for particular clades. It is neither possible toindependently rerun these analyses and recover their results,nor is it possible to assess data robustness, the relative sup-port for clades, or the support for alternative phylogenetichypotheses.

With the exception of the preliminary study of Butler(2005), Sereno (1999a) is the only published ornithischiananalysis that includes a data matrix that can be rerun andreanalysed. However, there are problems with this analysis.Only a limited number of ornithischian taxa were includedand, in addition, monophyly of major clades (such as Mar-ginocephalia) was assumed prior to analysis; unfortunately,monophyly of clades such as Marginocephalia are questionsthat ornithischian analyses still need to resolve.

Material and Methods

The aim of the analysis

The aim of this analysis is to test the global phylogeny of Or-nithischia, concentrating on the phylogenetic relationshipsof basal forms. Questions concerning the monophyly of theDinosauria, the phylogenetic relationships of basal dino-sauriformes and basal saurischians, and the relationships ofderived taxa within the major ornithischian clades, are bey-ond the scope of this analysis. The intention is to test thephylogenetic framework upon which the current understand-ing of ornithischian evolution is based.

Phylogenetic framework

All published cladistic analyses have shown Dinosauria toform a monophyletic clade (e.g. Gauthier 1986; Benton& Clark 1988; Novas 1996; Sereno 1999a; Benton 2004;Langer 2004; Langer & Benton 2006). Dinosauria includestwo subclades: Saurischia and Ornithischia (Gauthier 1986;Sereno 1986). A number of outgroups to Dinosauria havebeen identified; the most proximate of which appear to be sev-eral dinosauriform taxa, exemplified by Lagerpeton (Ser-eno & Arcucci 1993), Marasuchus (Sereno & Arcucci 1994)and Silesaurus (Dzik 2003; Langer & Benton 2006). Suc-cessively more distant outgroups to Dinosauria within Archo-sauria include Pterosauria (although see Bennett 1996; Peters2000), Scleromochlus, Crurotarsi, Proterochampsidae, Eu-parkeria, Erythrosuchus and Proterosuchus (Sereno 1991b;Benton 2004).

Table 1 provides phylogenetic definitions for ornithis-chian clade names discussed in the text (modified from Ser-eno 1998, 1999b; Wagner 2004). There is a conflict in the lit-erature between the names Neornithischia (Cooper 1985) andCerapoda (Sereno 1986), which have both been applied tothe clade consisting of ornithopods, pachycephalosaurs andceratopsians (e.g. Sereno 1999a; Weishampel 2004). Here,we follow Buchholz (2002) and Barrett et al. (2005) by us-ing both names: Neornithischia is applied to a stem-basedclade, while Cerapoda is used for a node-based clade. In ad-dition, our use of Ornithopoda differs from that of those au-thors who have defined this taxon as a node-based clade util-ising Heterodontosaurus as an internal specifier (e.g. Sereno1998); we instead use Ornithopoda for a stem-based clade(Buchholz 2002; Wagner 2004; Norman et al. 2004c).

Selection of ingroup taxa

A taxonomic review of Ornithischia was carried out and anumber of supraspecific, species level and outgroup oper-ational taxonomic units (OTUs) were selected. Weishampelet al. (1990, 1992, 2004a) served as the source for this review.Coding of taxa for cladistic analysis was based, where pos-sible, on first-hand examination of specimens, supplementedwith information from the literature. Appendix 1 provides de-tails of the references and specimens used for coding OTUs.The choice of ingroup taxa is discussed in greater detail be-low and outgroup taxa are discussed in the following section.

Supraspecic taxaA number of authors (e.g. Wiens 1998; Prendini 2001) havesuggested that supraspecific taxa should be avoided in phylo-genetic analysis when possible, as the coding of such taxais problematic, and simulations tend to suggest that splittingsuch taxa into species level terminals provides better results.Ideally, therefore, any analysis of Ornithischia should util-ise only species level terminal taxa. Nevertheless, the use ofspecies level taxa as exemplars for major clades was con-sidered impractical for this analysis for a number of reasons.First, the choice of exemplar taxa is not always obvious. Forinstance, the clade Ankylosauria is nearly universally accep-ted as monophyletic and is well-supported by anatomicalevidence. However, there is little consensus as to phylogen-etic relationships within Ankylosauria (e.g. Kirkland 1998;Carpenter 2001; Vickaryous et al. 2001, 2004; Parish 2003)and justifying the use of one taxon, or several taxa, as exem-plars is difficult. In addition, many of the apparently basal an-kylosaur taxa that might be used as exemplars are fragment-ary (e.g. Cedarpelta bilbeyhallorum, Mymoorapelta maysi),incompletely described in the literature (Gastonia burgei), orbased upon juvenile material (Liaoningosaurus paradoxus).In such a situation it can be advantageous to code a supra-specific taxon to represent the clade. The use of supraspecifictaxa has the additional advantage of reducing the number ofOTUs that must be included in the analysis. This is important,because it allows heuristic searches to be carried out in anacceptable timeframe and allows much more detailed exam-ination of the data. For these reasons, selected supraspecifictaxa were included in this analysis.

As outlined by Bininda-Emonds et al. (1998), the cor-rect use of supraspecific taxa in phylogenetic analyses hastwo requirements. Firstly, the taxa must be monophyletic.Only supraspecific taxa that are generally accepted as


Table 1 Phylogenetic denitions for the major ornithischian clades discussed in this analysis (modied from: Sereno 1998, 1999b; Buchholz2002; Wagner 2004).

Clade name Phylogenetic denition

Dinosauria Owen, 1842 Triceratops horridusMarsh, 1889, Passer domesticus (Linnaeus, 1758), their mostrecent common ancestor and all descendents.

Saurischia Seeley, 1887 All dinosaurs more closely related to Passer domesticus (Linnaeus, 1758) than toTriceratops horridusMarsh, 1889.

Ornithischia Seeley, 1887 All dinosaurs more closely related to Triceratops horridusMarsh, 1889 than to eitherPasser domesticus (Linnaeus, 1758), or Saltasaurus loricatus Bonaparte & Powell,1980.

Genasauria Sereno, 1986 Ankylosaurus magniventris Brown 1908, Stegosaurus stenopsMarsh, 1877a,Parasaurolophus walkeri Parks, 1922, Triceratops horridusMarsh, 1889,Pachycephalosaurus wyomingensis (Gilmore, 1931), their most recent commonancestor and all descendents.

Thyreophora Nopcsa, 1915 All genasaurians more closely related to Ankylosaurus magniventris Brown, 1908than to Parasaurolophus walkeri Parks, 1922, Triceratops horridusMarsh, 1889, orPachycephalosaurus wyomingensis (Gilmore, 1931).

Eurypoda Sereno, 1986 Ankylosaurus magniventris Brown, 1908, Stegosaurus stenopsMarsh, 1877a, theirmost recent common ancestor and all descendents.

Ankylosauria Osborn, 1923 All ornithischians more closely related to Ankylosaurus magniventris Brown, 1908than to Stegosaurus stenopsMarsh, 1877a.

Stegosauria Marsh, 1877a All ornithischians more closely related to Stegosaurus stenopsMarsh, 1877a than toAnkylosaurus magniventris Brown, 1908.

Neornithischia Cooper, 1985 All genasaurians more closely related to Parasaurolophus walkeri Parks, 1922, thanto Ankylosaurus magniventris Brown, 1908 or Stegosaurus stenopsMarsh, 1877a.

Cerapoda Sereno, 1986 Parasaurolophus walkeri Parks, 1922, Triceratops horridusMarsh, 1889, their mostrecent common ancestor and all descendents.

Ornithopoda Marsh, 1881 All genasaurians more closely related to Parasaurolophus walkeri Parks, 1922, thanto Triceratops horridusMarsh, 1889

Marginocephalia Sereno, 1986 Triceratops horridusMarsh, 1889, Pachycephalosaurus wyomingensis (Gilmore,1931), their most recent common ancestor and all descendents.

Ceratopsia Marsh, 1890 All marginocephalians more closely related to Triceratops horridusMarsh, 1889 thanto Pachycephalosaurus wyomingensis (Gilmore, 1931).

Pachycephalosauria Maryanska & Osmolska, 1974 All marginocephalians more closely related to Pachycephalosaurus wyomingensis(Gilmore, 1931) than to Triceratops horridusMarsh, 1889.

monophyletic were utilised here. Secondly, it must be pos-sible to code them as a single OTU in a manner that maintainstheir position on a cladogram with respect to a solution in-cluding all species. Several authors (e.g. Bininda-Emondset al. 1998; Wiens 1998) suggest that the ancestral method,whereby the character states of a hypothetical ancestor (thegroundplan) are reconstructed on the basis of prior phylo-genetic analyses, is the most successful method of codingsupraspecific taxa. The ancestral method (the methodologyis outlined by Langer & Benton 2006) was used in this ana-lysis. The eight supraspecific OTUs used are discussed ingreater detail below:

1. Ankylosauria. Ankylosauria is defined as all taxa moreclosely related to Ankylosaurus magniventris than toStegosaurus stenops (Sereno 1998) and includes thesubclades Ankylosauridae and Nodosauridae. Includedtaxa and diagnostic features are listed by Vickaryouset al. (2004). The known temporal range of the clade isCallovian to Maastrichtian (Middle JurassicLate Creta-ceous).

The most recent review of Ankylosauria (Vickaryouset al. 2004) recognised over 40 valid species. The phylo-geny assumed here for character coding represents a con-sensus of the following published phylogenies: Kirkland

(1998), Vickaryous et al. (2001), Hill et al. (2003) andVickaryous et al. (2004). The phylogenetic analysis ofCarpenter (2001) is not used, because it utilised a com-partmentalisation technique and does not, therefore, rep-resent a global phylogeny.

2. Stegosauria. Stegosauria is defined as all taxa moreclosely related to Stegosaurus stenops than to Ankylo-saurus magniventris (Sereno 1998). A full listing of in-cluded taxa and synapomorphies supporting monophylyof the clade is given in Galton & Upchurch (2004b). Ste-gosaurs form a relatively small but well-known and well-supported clade of ornithischians, known mostly from theMiddle to Late Jurassic, with fragmentary forms knownfrom the Early Cretaceous.

Sereno & Dong (1992) provided the first phylo-genetic analysis of stegosaurs, but considered only afew taxa. They proposed that Huayangosaurus taibaiirepresents the most basal member of the clade, withDacentrurus armatus positioned as the sister-groupto all more derived stegosaurs. Galton & Upchurch(2004b) have provided the most complete analysis todate; the basal positions of Huayangosaurus and Da-centrurus were confirmed by their analysis, but littleresolution was found amongst more derived stego-saurs.


3. Rhabdodontidae. Rhabdodontidae is defined as Zalmoxesrobustus, Rhabdodon priscus, their common ancestor andall of its descendents (Weishampel et al. 2003) and thetemporal range of the clade extends from the Late San-tonian to the Maastrichtian (Late Cretaceous). Synapo-morphies supporting the clade are given by Weishampelet al. (2003).

4. Dryosauridae. Dryosauridae includes the taxa Dry-osaurus altus, Dryosaurus lettowvorbecki, Valdosauruscanaliculatus and Valdosaurus nigeriensis (Norman2004) and is defined as Dryosaurus altus and all taxamore closely related to it than to Parasaurolophus walkeri(Sereno 1998). The clade is known from the Late Jurassicand Early Cretaceous. Potential synapomorphies of thisclade include: lacrimal inserts into notch in the maxilla;very wide brevis shelf on the ilium; large, deep pit on thefemoral shaft, at the base of the fourth trochanter; digit Iof the pes is absent or vestigal.

5. Ankylopollexia. Ankylopollexia is defined as Campto-saurus dispar, Parasaurolophus walkeri, their commonancestor and all descendents (Sereno 1998). A listing ofincluded taxa and synapomorphies supporting monophylyof the clade is given by Norman (2004). The clade ex-tends from the Kimmeridgian to the Maastrichtian (LateJurassicLate Cretaceous).

Ankylopollexia was erected by Sereno (1986) for or-nithopods exhibiting derived features of the teeth andmanus, in particular the modification of manual digit I toaccommodate a spine-like pollex. Within this clade areCamptosauridae and Styracosterna, both of which havebeen given stem-based phylogenetic definitions by Sereno(1998). The monophyly of Ankylopollexia is universallysupported by phylogenetic analysis and its interrelation-ships are relatively well understood (e.g. Norman 2002,2004).

6. Pachycephalosauridae. Pachycephalosauridae is definedas all taxa more closely related to Pachycephalosauruswyomingensis than to either Homalocephale calathocer-cos or Goyocephale lattimorei. Known taxa are restrictedto the Late Cretaceous (CampanianMaastrichtian) andsynapomorphies are given in Sereno (2000).

A number of explicit, numerical phylogenetic analysesof Pachycephalosauridae have been carried out in recentyears (Sereno 2000; Williamson & Carr 2002; Sullivan2003; Maryanska et al. 2004).These studies indicate abasal position for Stegoceras and the existence of a de-rived clade containing Tylocephale, Prenocephale andPachycephalosaurus, amongst others.

7. Psittacosauridae. Psittacosauridae is defined as all taxamore closely related to Psittacosaurus mongoliensis thanto Triceratops horridus and contains the genera Hong-shanosaurus and Psittacosaurus. Synapomorphies of theclade are given by Sereno (2000). The genus Psit-tacosaurus is probably the most speciose and diverse ofall dinosaur genera, although the exact number of spe-cies recognised is controversial (Sereno 1990b; You &Dodson 2004). Relationships within the clade are poorlyunderstood. Members of Psittacosauridae are known fromthe Early Cretaceous.

8. Unnamed taxon (Coronosauria + Leptoceratopsidae).This unnamed clade is defined as Leptoceratops gracilis,Protoceratops andrewsi and Triceratops horridus, theircommon ancestor and all of its descendents and includes

taxa ranging from the Turonian to the Maastrichtian (LateCretaceous). Synapomorphies of this node are given bySereno (2000), Makovicky (2001) and Makovicky &Norell (2006).

Neoceratopsian phylogeny has undergone a number ofrigorous studies in recent years (Sereno 2000; Makovicky2001; Xu et al. 2002; You & Dodson 2003, 2004;Chinnery 2004; Makovicky & Norell 2006) and a broad-scale consensus has been reached that Liaoceratops andArchaeoceratops represent basal taxa and that other neo-ceratopsian taxa form a distinct clade. This clade ofmore derived neoceratopsians remains unnamed and,generally, comprises a clade known as Leptoceratop-sidae and a clade known as Coronosauria, which in turncomprises Protoceratopsidae and Ceratopsoidea (Sereno2000; Makovicky 2001; Xu et al. 2002; Chinnery 2004).

Included species level taxaFollowing the selection and definition of supraspecific taxa,the status of all taxa not included in one of these derivedclades was assessed, using first-hand observations and theliterature. For each taxon a decision was made as to whetherit should be included in phylogenetic analysis or not. Includedspecies level taxa are discussed in this section; excluded spe-cies level taxa (and the reasons for exclusion) are discussedin the following section.

Abrictosaurus consors (Thulborn, 1974) is known froma single partial skull and postcranial skeleton (BMNHRUB54, holotype; formerly UCL B54) from the upperElliot Formation of Lesotho (Early Jurassic: HettangianSinemurian). Autapomorphies have not previously beendefined for Abrictosaurus and this taxon is provisionally dia-gnosed by the following combination of characters: archeddiastema between premaxilla and maxilla present; enlargedcaniniform teeth absent from the premaxilla and dentary. Thesingle specimen of Abrictosaurus was initially described as anew species of the heterodontosaurid Lycorhinus (Thulborn1974); however, it can clearly be distinguished from Ly-corhinus on the basis of dental characters (Hopson 1975).Hopson (1975) tentatively referred the specimen BMNHA100 (discussed below) to Abrictosaurus; however, this re-ferral has not been supported by subsequent work. Abricto-saurus and BMNH A100 are included here as separate OTUs.

Agilisaurus louderbacki Peng, 1990 is known from acomplete, articulated skull and postcranial skeleton (ZDMT6011, holotype) from the Lower Shaximiao Formation ofSichuan Province, Peoples Republic of China (Middle Jur-assic: ?Bajocian, Chen et al. 1982; ?BathonianCallovian,Dong & Tang 1984). Agilisaurus can be distinguishedby the following autapomorphies: presence of a palpeb-ral/supraorbital bar that traverses the width of the orbit;anteriormost dentary teeth conical, resembling premaxillaryteeth; presence of a series of low, anterolaterally directedridges on the orbital portion of the frontal; presence of anexcavated area immediately anterior to the antorbital fossa(modified from Barrett et al. 2005).

Anasibetia saldiviai Coria & Calvo, 2002 is knownfrom a partial skeleton with skull fragments (MCF-PVPH-74, holotype) as well as a number of referred speci-mens (see Coria & Calvo 2002) from the Lisandro Form-ation of Neuquen Province, Argentina (Late Cretaceous:Cenomanian). This taxon can be diagnosed on the basis of its


anteroventrally orientated occipital condyle and the presenceof an ilium with preacetabular process longer than 50% ofthe total ilium length (modified from Coria & Calvo 2002).

Archaeoceratops oshimai Dong & Azuma, 1997 isknown from a well-preserved skull and postcranial material(IVPP V11114, holotype; IVPP V11115, paratype) from theXinminbao Group of Gansu Province, China (Early Creta-ceous: AptianAlbian) and is characterised by the presence ofan excavation on the lateral surface of the ischiadic peduncleof the ilium, as well as by a unique character combination(modified from You & Dodson 2003).

The specimen BMNH A100 (formerly UCL A100)comprises a partial, disarticulated skull from the upper El-liot Formation of South Africa. Assignment of this speci-men to Lycorhinus (Thulborn 1970b, 1974; Gow 1990) wasnot based on unique characters, but on general similarity. Anumber of subsequent authors criticised the referral of thisspecimen to Lycorhinus: Galton (1973a: caption to fig. 2)referred BMNH A100 to Heterodontosaurus sp.; Charig &Crompton (1974) considered BMNH A100 to be generic-ally distinct from both Heterodontosaurus and Lycorhinus;while Hopson (1975) provisionally referred BMNH A100 toAbrictosaurus.

The taxonomy of the Elliot Formation heterodontosaur-ids is poorly resolved and requires further work. At presentthere is no consensus as to the taxonomic status of BMNHA100; however, this specimen is known from relatively com-plete and informative cranial remains and has received a de-tailed description (Thulborn 1970b) and is thus included inthe phylogenetic analysis here as a separate OTU.

Bugenasaura infernalis Galton, 1995 was erected for apartial skull and postcranial fragments (SDSM 7210, holo-type) from the Hell Creek Formation of South Dakota, USA(Late Cretaceous: Maastrichtian). This taxon is diagnosed bythe following features: no edentulous region at the anteriorend of the premaxilla; very deeply recessed cheek tooth row,with a massive and deep dentary and a very prominent over-hanging ridge (with a braided appearance) on the ventralpart of the maxilla; distal end of palpebral obliquely trun-cated with ridges medially (modified from Galton 1999). Anew skeleton of Bugenasaura is known (MOR 979; R.J.B.pers. obs. 2004); however, this specimen has not been fullyprepared and is currently undescribed.

Chaoyangsaurus youngi Zhao et al. 1999 is based upona partial skull and fragmentary postcranial elements (IG-CAGS V371, holotype) from the Tuchengzi Formation ofLiaoning Province, China (Middle or Late Jurassic: MiddleJurassic, Zhao et al. 1999; Tithonian, Weishampel et al.2004b). Chaoyangsaurus is distinguished by the follow-ing autapomorphic features: quadratojugal overlaps posteriorside of the quadrate shaft; quadrate slopes strongly anteriorly;ridge present between the planar lateral and ventral surfacesof the angular (modified from Zhao et al. 1999).

Echinodon becklesii Owen, 1861b is based upon frag-mentary cranial material (see Norman & Barrett 2002) fromthe Purbeck Formation of England (Early Cretaceous: Berri-asian). Echinodon can be diagnosed by the presence of one,or possibly two, caniniform teeth situated at the anterior endof the maxilla (Norman & Barrett 2002).

Emausaurus ernsti Haubold, 1990 is known from a par-tial skull and postcranial fragments (SGWG 85, holotype),from an unnamed unit in Germany (Early Jurassic: Toarcian).Emausaurus can be diagnosed by the possession of a large,

triangular plate-like palpebral, the robust lateral margin ofwhich bears a number of low ridges.

Gasparinisaura cincosaltensis Coria & Salgado, 1996is known from numerous specimens (see Coria & Salgado1996; Salgado et al. 1997) from the Rio Colorado Forma-tion of Patagonia, Argentina (Late Cretaceous: ConiacianSantonian) and is diagnosed by the following characters:anteroposteriorly wide ascending process of lacrimal con-tacts ventral process of postorbital posteriorly; infratemporalfenestra bordered entirely ventrally by quadratojugal; apexof arched dorsal margin of infratemporal fenestra positionedposterior to mandibular articulation; fully fused greater andlesser trochanters; condylid of femur laterally positioned(modified from Coria & Salgado 1996).

Goyocephale lattimorei Perle et al. 1982 is known froma relatively complete skeleton with partial skull (GI SPS100/1501, holotype), from an unnamed unit, Mongolia (LateCretaceous: ?late Santonian or early Campanian). One auta-pomorphy has been identified: the lateral margin of the skullis sinuous in dorsal view (Sereno 2000).

Heterodontosaurus tucki Crompton & Charig, 1962 isknown from a nearly complete skull (SAM-PK-K337, holo-type) from the Clarens Formation (= Cave Sandstone) ofSouth Africa (Early Jurassic: Sinemurian) and a referredskull and postcranial skeleton (SAM-PK-K1332, Santa Lucaet al. 1976; Santa Luca 1980) from the upper Elliot Form-ation of South Africa. A number of features may be auta-pomorphic for this taxon, although it is possible that somemay prove to be present in other, poorly known, heterodon-tosaurids, or may eventually prove to be ornithischian ple-siomorphies. These possible autapomorphies include: dorsalprocess of premaxilla does not form contact with nasals; an-terior, accessory opening present within the antorbital fossa;squamosalquadratojugal contact is anteroposteriorly broad;paroccipital processes are very deep dorsoventrally; paired,deep recesses on the ventral surface of the basisphenoid;basisphenoid processes are extremely elongated; cingulumis completely absent on cheek-teeth; ischium with elongateflange on lateral margin.

Hexinlusaurus multidens (He & Cai, 1983) is knownfrom two partial skulls and postcranial skeletons (ZDMT6001, holotype; ZDM T6002, paratype) from the LowerShaximiao Formation of China. Hexinlusaurus can be distin-guished by the presence of a marked concavity that extendsover the lateral surface of the postorbital (Barrett et al. 2005).

Homalocephale calathocercos Maryanska &Osmolska, 1974 is known only from a skull and par-tial postcranial skeleton (GI SPS 100/1201, holotype) fromthe Nemegt Formation of Mongolia (Late Cretaceous: ?lateCampanian or early Maastrichtian). Homalocephale isdiagnosed by the presence of a postacetabular process of theilium that is crescent-shaped and ventrally deflected (Sereno2000).

Hypsilophodon foxii Huxley, 1869 is known from nu-merous specimens (see Galton 1974a) from the WessexFormation of the Isle of Wight, UK (Early Cretaceous: Bar-remian). Autapomorphies have not been previously definedfor Hypsilophodon, but include the presence of a large fora-men in the ascending process of the maxilla that communic-ates medially with the antorbital fossa (Galton 1974a: fig. 3).Although material of Hypsilophodon has been reported fromcontinental Europe (Sanz et al. 1983) and North America(Galton & Jensen 1979), none of this material can be


confidently referred to this taxon and, at present, Hypsilo-phodon is only known from the UK.

Jeholosaurus shangyuanensis Xu et al. 2000 is knownfrom two specimens (IVPP V12529, holotype; IVPPV12530, referred) from the Yixian Formation of LiaoningProvince, China (Early Cretaceous) and is characterised bythe following combination of characters: six premaxillaryteeth; foramina present on dorsal surface of nasal; large fo-ramen present in quadratojugal; predentary about 1.5 timesas long as the premaxilla; pedal phalanx 34 times longerthan other phalanges of pedal digit 3 (modified from Xuet al. 2000).

Lesothosaurus diagnosticus Galton, 1978 is knownfrom a number of nearly complete skulls and disarticulatedpostcranial skeletons, while Stormbergia dangershoeki But-ler, 2005 is known from three partial skeletons. Both taxa arefrom the upper Elliot Formation of South Africa and Leso-tho. A full discussion of the hypodigm and diagnosis of eachtaxon can be found in Butler (2005).

Liaoceratops yanzigouensis Xu et al. 2002 is knownfrom two complete skulls (IVPP V12738, holotype; IVPPV12633, referred specimen) from the Lower Yixian Forma-tion of Liaoning Province, China (Early Cretaceous). Liao-ceratops is characterised by the following features: suturesbetween premaxilla, maxilla, nasal and prefrontal intersect-ing at a common point high on the side of the snout; posses-sion of several tubercles on the ventral margin of the angular;a foramen on the posterior face of the quadrate near the artic-ulation with the quadratojugal; small tubercle on the dorsalborder of the foramen magnum; thick posterior border of theparietal frill (Xu et al. 2002).

Lycorhinus angustidens Haughton, 1924 is known froma left dentary (SAM-PK-K3606, holotype) and two provi-sionally referred specimens (BP/1/4244, left maxilla, holo-type of Lanasaurus scalpridens Gow, 1975, referred to Ly-corhinus angustidens by Gow 1990; BP/1/5253, left max-illa, referred to Lycorhinus angustidens by Gow 1990) fromthe upper Elliot Formation of South Africa. There has beenconsiderable controversy over the validity of Lycorhinus an-gustidens. Haughton (1924) named Lycorhinus for a partialleft dentary that he believed represented a cynodontid syn-apsid. Crompton & Charig (1962) reidentified Lycorhinus asa heterodontosaurid and, later Charig & Crompton (1974)considered it a nomen dubium. Thulborn (1970b) assignedthe specimen BMNH A100 to Lycorhinus; however, this as-signment was not supported by most subsequent authors(Galton 1973a; Charig & Crompton 1974; Hopson 1975,1980). Hopson (1975) demonstrated that Lycorhinus could bedistinguished from other heterodontosaurids (Abrictosaurusand Heterodontosaurus). Finally, Gow (1990) referred themaxillae BP/1/4244 (holotype of Lanasaurus Gow, 1975)and BP/1/5253 to Lycorhinus.

The validity of Lycorhinus requires reassessment andonly a preliminary diagnosis is suggested here, based uponthe marked medial curvature of the dentary and maxillarytooth rows (Gow 1990). As discussed by Hopson (1975), aunique combination of plesiomorphic and derived charactersis probably also diagnostic for Lycorhinus.

Micropachycephalosaurus hongtuyanensis Dong, 1978is known from a partial skull and skeleton (IVPP V5542,holotype) from the Wangshi Formation of ShandongProvince, Peoples Republic of China (Late Cretaceous:Campanian). The holotype of Micropachycephalosaurus is

extremely fragmentary and many elements were erroneouslyidentified in the original description. A full review and rede-scription is being prepared (R. J. B. & Q. Zhao, unpublishedresults). Although Sereno (2000) has suggested that autapo-morphic features are absent, the presence of prominent vent-ral grooves on the most posterior dorsal vertebrae appearsto be autapomorphic for Micropachycephalosaurus and thistaxon is here included in the phylogenetic analysis.

Orodromeus makelai Horner & Weishampel, 1988 isknown from abundant and well-preserved material (seeScheetz 1999) from the Upper Two Medicine and JudithRiver formations of Montana, USA (Late Cretaceous: ?LateCampanian). Orodromeus is characterised by the followingcombination of characters: prominent boss on anterolateralmaxilla; subnarial depression on premaxillamaxilla bound-ary; midline depression on nasals; boss on jugal; postorbitalwith distinct projection into orbit; dentition plesiomorphicwith ridges absent lingually and labially.

Othnielia rex (Marsh, 1877b) is known from a num-ber of specimens (YPM 1915, holotype, left femur; referredspecimens (see Galton 1983) include: BYU ESM-163R, ar-ticulated, near-complete postcranial skeleton described byGalton & Jensen 1973) from the Morrison Formation ofthe USA (Late Jurasssic: KimmeridgianTithonian). YPM1915 is the holotype of Nanosaurus rex Marsh, 1877b, whichwas made the type species of the genus Othnielia by Galton(1977). This specimen (an isolated femur) lacks obvious auta-pomorphies, although it may be diagnosable on the basis of aunique character combination. Referral of specimens to Oth-nielia follows Galton (1983), pending a review of the validityof this taxon.

Parksosaurus warreni (Parks, 1926) is known from asingle, relatively complete, skull and skeleton (ROM 804,holotype) from the Horseshoe Canyon Formation of Alberta,Canada (Late Cretaceous: Maastrichtian). Parksosaurus isdistinguished by a dorsoventrally broad posterolateral pro-cess of the premaxilla and a postorbital process of the jugalthat expands posterodorsally.

Pisanosaurus mertii Casamiquela, 1967 is known froma single partial skeleton (PVL 2577, holotype) from the Is-chigualasto Formation (Late Triassic: Carnian) of Argentina.As discussed by Sereno (1991a), this taxon can be diagnosedby the following characters: anteroposterior depth of distalend of the tibia is greater than maximum transverse width;calcaneum is transversely narrow.

Scelidosaurus harrisonii Owen, 1861a is known fromseveral partial skeletons from the Lower Lias of Dorset, Eng-land (Early Jurassic: late Sinemurian). Owen (1861a) de-scribed material which he referred to Scelidosaurus, includ-ing a femur (GSM 109560), articulated knee-joint (BMNH39496), ungual phalanx (GSM 10956), a partial juvenileskeleton (Philpott Museum, Lyme Regis, unnumbered, castsare catalogued as BMNH R5909) and a near-complete skull(BMNH R1111). Owen (1863) described the near-completepostcranial skeleton associated with the skull BMNH R1111,while Lydekker (1888) later designated the articulated knee-joint as the lectotype. Newman (1968) recognised that thematerial described by Owen (1861a) represents a compositeof theropod (GSM 109560; BMNH 39496; GSM 10956) andornithischian (Philpott Museum, juvenile skeleton; BMNHR1111) material, and Charig & Newman (1992) formallydesignated the skull and postcranial skeleton (BMNH R1111)as a replacement lectotype. The juvenile described by Owen


(1861a) probably represents a second individual of Scelido-saurus (Galton 1975). Further material has come to light inrecent years, including BMNH R6704 (Rixon 1968; Charig1972: fig. 6A; considered as a possible new taxon of basalthyreophoran by Coombs et al. 1990), BRSMG Ce12785(Barrett 2001) and CAMSM X 39256. Reports of the genusScelidosaurus in the Kayenta Formation of Arizona (Padian1989) and the Lower Lufeng of China (Lucas 1996: see Tat-isaurus, below) cannot be substantiated at present. No dia-gnosis based upon synapomorphies has ever been publishedfor Scelidosaurus and a full diagnosis must await redescrip-tion of this taxon (D. B. Norman, unpublished results).

Scutellosaurus lawleri Colbert, 1981 is known fromseveral partial skeletons (MNA P1.175, holotype; MNAP1.1752, paratype; for referred material see Rosenbaum &Padian 2000) described by Colbert (1981) and Rosenbaum &Padian (2000) from the Kayenta Formation of Arizona, USA(Early Jurassic: SinemurianPliensbachian). Unique featuresinclude: dorsal and ventral margins of the preacetabular pro-cess of the ilium are drawn out medially into distinct flangeswhich converge upon one another anteriorly; elongate tail ofat least 58 caudal vertebrae (R. J. B. & S. C. R. Maidment,unpublished results).

Stenopelix valdensis Meyer, 1857 is known from asingle partially articulated postcranial skeleton, preservedas impressions in sandstone blocks, from the ObernkirchenSandstein of Germany (Early Cretaceous: Berriasian); latexcasts prepared by Sues & Galton (1982) make detailed exam-ination of the material possible. The ischium of Stenopelixhas the following autapomorphies: distinct bend at mid-shaft;broadest at mid-shaft and tapers anteriorly and posteriorly;blade is transversely arched distally, being ventrally convexand dorsally concave (modified from Sereno 1987).

Talenkauen santacrucensis Novas et al. 2004 is basedupon a partial skull and postcranial skeleton (MPM-10001,holotype) from the Pari Aike Formation, Santa CruzProvince, Argentina (Late Cretaceous: Maastrichtian). Thistaxon can be diagnosed by the presence of well-developedepipophyses on cervical 3 and plate-like uncinate processeson the rib-cage (Novas et al. 2004).

The genus Tenontosaurus Ostrom, 1970 is character-ised by the following features: dorsoventrally tall maxilla,nearly full height of the rostrum; orbit square; 12 cervicalvertebrae; elongate tail (59+ caudals). Two species are re-cognised and included here: Tenontosaurus tilletti Ostrom,1970 is known from abundant material (see Forster 1990)from the Cloverly Formation of Montana, USA (Early Creta-ceous: AptianAlbian) and Tenontosaurus dossi Winkleret al. 1997 is known from two specimens from the TwinMountains Formation of Texas, USA (Early Cretaceous).The species are distinguished by the retention in Tenonto-saurus dossi of premaxillary teeth and a postpubic processequal in length to the ischium (Winkler et al. 1997).

Thescelosaurus neglectus Gilmore, 1913 is knownfrom numerous specimens (see Galton 1997) from theLance Formation of Wyoming, the Hell Creek Formation ofMontana, the Scollard Formation of Alberta and the French-man Formation of Saskatchewan (Late Cretaceous: ?lateCampanianMaastrichtian). The taxonomic history of Thes-celosaurus has been summarised by Galton (1995, 1997)and his taxonomic assignments are followed provisionallyhere, with Thescelosaurus edmontonensis being considereda junior synonym of T. neglectus. It should be noted that

synonymy is based on general similarity, rather than on auta-pomorphic features (see Galton 1995); future revision of theThescelosaurus material may indicate the presence of two,or more, distinct taxa. One possible autapomorphy is recog-nised here: a large notch or foramen within the supraoccipital,dorsal to the foramen magnum (see Galton 1997).

Wannanosaurus yansiensis Hou, 1977 is known froma partial skull and postcranial skeleton (IVPP V4447, holo-type) and some referred postcranial elements (IVPP V4447.1,paratype), from the Xiaoyan Formation of Anhui Province,China (Late Cretaceous: Campanian). Wannanosaurus is dis-tinguished by the extreme flexure of the humerus, with prox-imal and distal ends set at approximately 30 to one another(modified from Sereno 2000).

Yandusaurus hongheensis He, 1979 is known from apartial skull and postcranial material (GCC V20501, holo-type) from the Upper Shaximiao Formation of Sichuan,Peoples Republic of China (Late Jurassic: ?Oxfordian,Weishampel et al. 2004b). All of the anatomical features pre-viously used to diagnosis Yandusaurus (He & Cai 1984) havewider distributions amongst basal ornithopods. However, oneautapomorphic feature is apparent in the holotype (R. J. B.,pers. obs. 2004; Barrett et al. 2005); the midposterior cer-vicals have circular, pit-like depressions developed at thebase of their postzygapophyses.

Zephyrosaurus schaffi Sues, 1980 is known from an in-complete skull and postcranial fragments (MCZ 4392, holo-type) from the Cloverly Formation of Montana, USA (EarlyCretaceous: AptianAlbian) and is characterised by the fol-lowing unique combination of characters: prominent bosson anterolateral maxilla; short, massive, triangular palpeb-ral; boss on jugal; postorbital with distinct projection into or-bit; dentary/maxillary teeth with numerous subparallel ridgesconnecting to marginal denticles.

Excluded species level taxaA number of taxa have been considered non-diagnosablenomina dubia by most recent reviews and are here excludedfrom analysis. These taxa include Camptosaurus leedsi,Fabrosaurus australis, Geranosaurus atavus, Hypsilopho-don wielandi, Laosaurus celer, Laosaurus minimus, Lus-itanosaurus liasicus, Nanosaurus agilis and Sanpasaurusyaoi.

Many ornithischian or putative ornithischian taxa havebeen erected solely or largely on the basis of dental re-mains, including: Alocodon kuehnei, Crosbysaurus harri-sae, Drinker nisti, Ferganocephale adenticulatum, Galto-nia gibbidens, Gongbusaurus shiyii, Krzyzanowskisaurushunti, Lucianosaurus wildi, Pekinosaurus olseni, Phyllodonhenkeli, Protecovasaurus lucasi, Revueltosaurus callenderi,Siluosaurus zhangqiani, Stegosaurus madagascariensis,Taveirosaurus costai, Tecovasaurus murryi and Trimucrodoncuneatus. Recent work has demonstrated that at least someof these taxa pertain to non-ornithischian clades (Parker et al.2005a; Irmis et al. 2007) and the taxonomic validity of manyof these tooth taxa is additionally questionable (Weishampelet al. 2004a). Furthermore, tooth taxa add little new anatom-ical information to the analysis and suffer from extremelyhigh (more than 95%) levels of missing data. For these reas-ons these taxa have been excluded.

Technosaurus smalli, from the Cooper Canyon Form-ation (Late Triassic: Norian) of Texas, was described as afabrosaurid ornithischian by Chatterjee (1984) on the basis


of a single fragmentary skeleton. Sereno (1991a) sugges-ted that the holotype of Technosaurus contains elements ofboth an ornithischian and a sauropomorph. Irmis et al. (2005,2007) agreed that the specimen is a composite of at least twotaxa, but suggested that the posterior portion of the lower jawis referable to the pseudosuchian Shuvosaurus, whereas otherparts of the holotype may represent a Silesaurus-like taxon.They do not consider any of the material to be ornithischian.In view of the considerable confusion as to the associationof the holotype of Technosaurus, it is here excluded fromphylogenetic analysis.

Norman et al. (2004b) considered Tatisaurus oehleriand Bienosaurus lufengensis from the Lower Lufeng Form-ation (Early Jurassic: Sinemurian) of China as valid taxaand possible basal thyreophorans. However, reassessment ofTatisaurus (Norman et al. 2007) suggests that it is a no-men dubium. Autapomorphic characters are not evident inthe original description of Bienosaurus (Dong 2001) and thevalidity of this taxon is uncertain. As a result, both taxa areexcluded from the current phylogenetic analysis.

Xiaosaurus dashanpensis from the Lower ShaximiaoFormation of Sichuan Province, China (Middle Jurassic:?Bajocian, Chen et al. 1982; ?BathonianCallovian, Dong& Tang 1984) is based upon very fragmentary remains, butappears to be diagnosable on the basis of the possession of aproximally straight humerus that lacks the medial curvatureseen in all other basal ornithischians (Barrett et al. 2005).However, the whereabouts of the holotype and referred ma-terial are currently unknown (Barrett et al. 2005) and un-available for further study and the original description (Dong& Tang 1983) provides few anatomical details. As a result,we do not include Xiaosaurus in the present study.

The location of the holotype of Gongbusaurus wu-caiwanensis is currently unknown (X. Xu, pers. comm.,2004) and the original description (Dong 1989) is brief andpoorly figured and does not allow the recognition of auta-pomorphies, or a unique character combination. Therefore,despite the fact that recent reviews (e.g. Norman et al. 2004c)retain Gongbusaurus wucaiwanensis as a valid taxon, wehere exclude it from the phylogenetic analysis.

Notohypsilophodon comodorensis was described byMartnez (1998) on the basis of a partial skeleton from theBajo Barreal Formation, Chubut Province, Argentina (LateCretaceous: ?Cenomanian). Martnez (1998) listed a numberof potential autapomorphies of Notohysilophodon; however,most of these appear to have wider distributions within Or-nithischia (e.g. the reduction of the deltopectoral crest of thehumerus is seen in other South American ornithopods, Novaset al. 2004) or represent plesiomorphies (distal end of fibulareduced, astragalus with stepped proximal surface). Noto-hypsilophodon may represent a valid taxon, but this cannotbe ascertained from the published description and this taxonis provisionally excluded from the phylogenetic analysis.

Four taxa (Atlascopcosaurus loadsi, Fulgurotheriumaustrale, Leaellynasaura amicagraphica, Quantassaurus in-trepidus) have been named on the basis of cranial and post-cranial material from the Early Cretaceous of Australia (Rich& Vickers-Rich 1989, 1999). Although all of these taxa havebeen considered valid by a recent review (e.g. Norman et al.2004c), the type specimens of all four taxa are fragmentaryand unambiguous autapomorphies have not yet been defined.Referral of additional material to any one of these taxa isproblematic. Although we accept that some or all of these

taxa may prove to be diagnostic with further study, we hereprovisionally exclude them from phylogenetic analysis aswe have been unable to examine the majority of the materialfirst-hand.

A number of taxa (Auroraceratops rugosus,Changchunsaurus parvus, Xuanhuaceratops niei, Yamacer-atops dorngobiensis, Yinlong downsi) were described afterthe current analysis was carried out and thus have not beenincluded. We plan to include these taxa in future iterationsof this analysis.

Selection of outgroup taxa

Three outgroup taxa were chosen, based upon the phylo-genetic framework outlined above. Recent, comprehens-ive, phylogenetic analyses of basal dinosaurs have demon-strated that Herrerasaurus ischigualastensis Reig, 1963 isthe most basal known member of Saurischia (Langer 2004;Langer & Benton 2006). In addition, this taxon is knownfrom well-preserved complete material and has been extens-ively described (Novas 1993; Sereno 1993; Sereno & Novas1993). Marasuchus talampayensis (Romer, 1972) representsa well-known dinosaurian outgroup and Euparkeria capensisBroom, 1913 represents a basal archosaur, phylogeneticallydistant from Dinosauria and lacking the numerous derivedspecialisations seen in many other more proximate dinosaur-ian outgroups, such as pterosaurs and crurotarsans.

Analyses

Search methods

The full matrix (Appendix 3) consists of 46 taxa (43 in-group taxa and 3 outgroup taxa), coded for 221 characters(Appendix 2). The data matrix was constructed using theNEXUS Data Editor (http://taxonomy.zoology.gla.ac.uk/rod/NDE/nde.html). Prior to analysis, safe taxonomic reduc-tion (Wilkinson 1995c) was carried out using the pro-gram TAXEQ3 (Wilkinson 2001a). Safe taxonomic reduc-tion identifies taxa that can be excluded without affectingthe inferred relationships of the remaining taxa. The matrixdoes not contain any taxonomic equivalents and all taxa wereincluded in subsequent analyses.

Analyses were carried out in PAUP 4.0b10 (Swofford2002); all characters are treated as unordered and equallyweighted and polymorphisms are treated as uncertainty.Branches with a minimum length of zero were collapsed dur-ing searches (the -amb option); this setting recovers onlystrictly supported trees (Nixon & Carpenter 1996; Kearney& Clark 2003), but can result in trees that are not of min-imum length and cannot be considered as most parsimonioustrees (MPTs); (Wilkinson 1995a). As a result we filteredthe resultant set of trees to ensure that only minimum lengthtrees were retained. Analysis was conducted using a heuristicsearch with 10,000 replicates and TBR branch-swapping,each starting tree being produced by random stepwise addi-tion.

The analysis recovered 3787 trees; filtering these treesso that only minimum length trees were retained resultedin 756 MPTs of 477 steps (Consistency Index (CI) = 0.505,Retention Index (RI) = 0.732, Rescaled Consistency Index(RC) = 0.370). Strict and 50% majority-rule component


Figure 2 Strict component consensus (left) and 50% majority-rule consensus (right) of 756 most parsimonious trees (MPTs) produced byanalysing a data matrix of 46 taxa and 221 characters. Values above nodes on the Strict component consensus represent bootstrap proportions.Values beneath nodes on the Strict component consensus indicate Bremer support. Bremer support values of +1 or less are not shown. Numbersbeneath nodes on the 50% majority-rule consensus indicate the percentage of MPTs in which that node appears (nodes with no values beneaththem appear in all MPTs).

consensus trees (Fig. 2) and an Adams consensus tree, werecalculated using PAUP. The strict component consensus(SCC) tree contains two major polytomies and contains amuch lower degree of resolution than the majority-rule orAdams consensus trees; the latter observation suggests thatthe low degree of resolution in the SCC tree results froma number of taxa acting as wildcards, as a result of high

amounts of missing data, or character conflict, or both. Amaximum agreement subtree was also calculated that ex-cludes 8 taxa (Echinodon, Lycorhinus, Bugenasaura, Jeholo-saurus, Talekauen, Thescelosaurus, Yandusaurus, Zephyro-saurus; see Fig. 3).

An additional search was carried out using a demonstra-tion version of TNT (Tree Analysis Using New Technology)


Figure 3 Maximum agreement subtree of 756 most parsimonious trees (MPTs) produced by analysing a data matrix of 46 taxa and 221characters. Eight (Echinodon, Lycorhinus, Bugenasaura, Jeholosaurus, Talekauen, Thescelosaurus, Yandusaurus and Zephyrosaurus) of theoriginal taxa are excluded.

v1.0, downloaded from www.zmuc.dk/public/phylogeny. ANew Technology search was carried out, using a randomaddition-sequence, 1000 replicates and default settings forthe Sect. Search, Ratchet, Drift and Tree Fusing op-tions. The search recovered 119 trees of 477 steps; the con-sensus of these trees matched the consensus of the 756 MPTsrecovered by PAUP. That TNT failed to find trees shorterthan 477 steps suggests that this is the minimum tree length.

The set of 756 MPTs recovered by PAUP forms the basisfor subsequent discussion.

In an attempt to resolve further relationships com-mon to all 756 MPTs and to identify the most unstabletaxa, reduced consensus techniques (Wilkinson 1994, 1995b,2003) were applied to the data. The most commonly usedconsensus methods are strict component consensus (SCC)trees, which include all terminal taxa and all the clades


(components) common to all MPTs. However, the strict com-ponent method has problems of insensitivity and may fail torepresent relationships that are common to the set of MPTs,but cannot be expressed as shared clades (Wilkinson 2003).Reduced consensus methods identify n-taxon statements;n-taxon statements express cladistic relationships (e.g. A andC are more closely related to each other than either is to E),but need not include all terminal taxa. Reduced consensusmethods delete unstable taxa to produce more informativeconsensus trees, which represent n-taxon statements.

Reduced consensus was applied using the strict pro-gram of REDCON 3.0 (Wilkinson, 2001b) and the resultscorroborated using RADCON (Thorley & Page 2000), identi-fying a profile of eight strict reduced consensus (SRC) trees,the first of which includes all taxa and corresponds in to-pology to the SCC tree of the 756 MPTs. The remainingSRC trees exclude one or more unstable wildcard taxa, res-ulting in an increase in resolution. Seven taxa (Echinodon,Lycorhinus, Zephyrosaurus, Talenkauen, Yandusaurus, Gas-parinisaura, Parksosaurus) are identified as unstable by theseanalyses. We combined six of the SRC trees (those exclud-ing Echinodon, Lycorhinus, Zephyrosaurus, Talenkauen andYandusaurus) to produce an informative derivative SRC tree(Fig. 4). We use this derivative SRC tree as the basis foroptimisation of synapomorphies (Appendix 4) and for muchof the subsequent discussion, and it represents our preferredhypothesis of interrelationships.

Testing the support for relationships

Randomisation testsPAUP was used to run a Permutation Tail-Probability (PTP)test using 1000 randomised replicates of the reduced dataset (Faith & Cranston 1991; Kitching et al. 1998). The ran-domised replicates are created by randomly permuting thecharacter states assigned to taxa, decreasing character con-gruence to a level that would be expected by chance alone.The MPT length is then calculated for each replicate and thedistribution of MPT lengths for the replicates is comparedto the length of the original MPT. The PTP test has beencriticised (Bryant 1992; Carpenter 1992) and Kitching et al.(1998) suggested that it could best be used as an independentevaluation of the explanatory power of the data set, ratherthan as a criterion for acceptance or rejection of any partic-ular cladogram. In this case, the results of this test indicatethat the most parsimonious tree length (477 steps) lies out-side the range of minimum tree lengths obtained from therandomised data (P = 0.001). This indicates that a signific-ant phylogenetic signal is present in the data set and is notcompletely obscured by character conflict and missing data.

Bremer supportTraditional decay analysis, or Bremer support, measuresthe number of additional steps required before the clade islost from the strict consensus of near-minimum length clado-grams (Bremer 1988; Kitching et al. 1998). Bremer supportwas calculated for nodes present in the SCC tree by search-ing in PAUP for the shortest trees not compatible with aparticular node, using the CONVERSE option.

Bremer support values are shown in Fig. 2. Most nodeshave a decay index of +1, i.e. they are absent from the strictconsensus of all trees of 478 steps or less. Stronger support is

found only within the clades Thyreophora and Iguanodontia,but even here support is relatively low. However, it is possiblethat a few unstable wildcard taxa, such as those identifiedby reduced consensus techniques (see above), can obscuresupport for relationships, resulting in lower decay indicesthan might be otherwise expected.

Wilkinson et al. (2000) proposed a new technique,double decay analysis (DDA), which provides Bremer sup-port for all strictly supported n-taxa relationships. An attemptwas made to apply DDA using RADCON (Thorley & Page2000); however, the large size of the data set meant thatthis approach was not feasible due to time and memory con-straints. For particular areas of interest PAUP was used towrite backbone constraints that could then be used to testthe decay indices of n-taxa statements (see individual ex-amples below). This allowed an assessment of the effect thatwildcard taxa have upon the Bremer support for clades.

BootstrappingBootstrap analysis generates pseudoreplicate data sets byrandomly sampling with replacement a proportion of thecharacters, deleting some characters randomly and reweight-ing other characters randomly. The MPTs are generated foreach pseudoreplicate and the degree of conflict between res-ulting MPTs is assessed using a 50% majority rule consensustree. Clades that are supported by a large number of charac-ters, with low levels of homoplasy, would be expected to havehigh bootstrap values, whereas bootstrap values should belower for clades supported by only a few, or by homoplastic,characters. The bootstrap should, therefore, be considered asa one-sided test of a cladogram (Kitching et al. 1998): groupsthat are recovered are supported by the data, but groups thatare not recovered (or that have low bootstrap values) cannotbe rejected.

A bootstrap analysis was carried out using PAUP with1000 replications. To allow the analysis to be carried outwithin a reasonable length of time the MAXTREES optionof PAUP was set to 1000. This means that for each pseu-doreplicate data set the search for MPTs was truncated once1000 trees had been found. Figure 2 shows the results ofthe bootstrap analysis. Bootstrap support is weak throughoutmuch of the tree.

Unstable wildcard taxa can obscure levels of bootstrapsupport for relationships (Wilkinson 2003). One potentialsolution is to use the majority-rule bootstrap reduced con-sensus (MBRC) technique developed by Wilkinson (1996).The MBRC method provides bootstrap proportions for alln-taxon statements and is implemented in the REDBOOTprogram of REDCON 3.0. Unfortunately, REDBOOT couldnot be used for this analysis due to the large size of thedata matrix. An alternative method for assessing the im-pact of wildcards on bootstrap support is used here. Alltrees recovered by the bootstrap analysis were saved to aNEXUS treefile. The five unstable taxa removed in the de-rivative SRC tree (Echinodon, Lycorhinus, Zephyrosaurus,Talenkauen and Yandusaurus) were pruned a posteriori fromthis set of bootstrap trees, with duplicate topologies cre-ated by this deletion being collapsed. Following pruning ofthese five taxa a new majority-rule bootstrap reduced con-sensus tree was generated containing recalculated bootstrapvalues (Fig. 4). For further areas of particular interest addi-tional potentially unstable taxa were identified and pruned


Figure 4 Derivative strict reduced consensus tree derived by a posteriori pruning of ve unstable taxa (Echinodon, Lycorhinus,Zephyrosaurus, Talenkauen and Yandusaurus) from the set of 756 most parsimonious trees (MPTs) generated by the full analysis. The numberabove each node is a unique identier used in the tree description (see Appendix 4). The number beneath a node represents the bootstrapproportion for that node (taken from the reduced bootstrap analysis). Note the increased levels of bootstrap support for a number of nodeswhen compared to Fig. 2. Abbreviations: ORN., Ornithischia; HETERODONT., Heterodontosauridae; GENA., Genasauria; NEORN., Neornithischia;ORNITH., Ornithopoda; MARG., Marginocephalia; CERAT., Ceratopsia; PACHY., Pachycephalosauria.


and bootstrap values recalculated. The effect of wildcardson bootstrap values is discussed further below.

Results

Ornithischian monophyly

Most of the characters identified as synapomorphic for Or-nithischia (Appendix 4) have been identified by previousauthors (Norman 1984; Sereno 1984, 1986, 1991a, 1999a;Cooper 1985; Maryanska & Osmolska 1985; Weishampel2004; Norman et al. 2004a; Butler 2005). Some new potentialornithischian synapomorphies are proposed by this analysis.The presence of a buccal emargination has previously beenconsidered (Sereno 1986, 1999a) to be synapomorphic for aless inclusive clade of ornithischians, Genasauria, and absentin the basal ornithischian Lesothosaurus diagnosticus. Thisanalysis alternatively suggests that the presence of a weak orincipient buccal emargination (generally correlated with thepresence of cheeks, see Galton 1973a) is a synapomorphyof Ornithischia. As noted by Butler (2005: 204), in Lesotho-saurus there is a weak anteroposteriorly extending ridge (re-ferred to, below, as the maxillary ridge) above the row of ex-ternal maxillary foramina, forming the ventral margin of theexternal antorbital fenestra. Below this eminence the lateralsurface of the maxilla is gently bevelled such that the maxil-lary tooth row is slightly inset along at least the posterior two-thirds of its length. A similar weak medial offset of the toothrow is seen in other basal ornithischians such as Abricto-saurus consors (BMNH RUB54) and Scutellosaurus lawl-eri (Colbert 1981) and is proposed here to be homologouswith the well-developed buccal emargination seen in manyother ornithischians. Irmis et al. (2007) criticised Butler(2005) for using the same coding for the weak buccalemargination of Lesothosaurus and the well-defined buccalemargination of taxa such as Heterodontosaurus tucki. How-ever, the coding of Butler (2005) referred only to the presenceof an emargination (not how well-developed it was), whichwe propose is homologous in Lesothosaurus and Hetero-dontosaurus. Irmis et al. (2007) additionally stated that inornithischians the maxillary ridge was separated from theventral margin of the external antorbital fenestra. This is truefor some ornithischians (e.g. Hypsilophodon foxii, Galton1974a) but in many basal ornithischians such as Lesotho-saurus (BMNH R8501) and Heterodontosaurus (SAM-PK-K1332; Norman et al. 2004c: fig. 18.1) the maxillary ridgedoes form the ventral margin of the external antorbital fen-estra.

Another potential ornithischian synapomorphy is thesize and position of the posttemporal foramen. In ornith-ischian outgroups the posttemporal foramen is relativelylarge and positioned on the boundary between the parietaland the paroccipital process (e.g. Euparkeria capensis, Ewer1965: fig. 2B; basal sauropodomorphs, Galton & Upchurch2004a; Herrerasaurus ischigualastensis, Sereno & Novas1993: fig. 8C), whereas in basal ornithischians (e.g. Lesotho-saurus diagnosticus, Sereno 1991a, Sereno 1991a: fig. 11C;Heterodontosaurus tucki, Weishampel & Witmer 1990b: fig.23.1), the posttemporal foramen is reduced in size and en-tirely enclosed by the paroccipital process. This character wasindependently identified as an ornithischian synapomorphyby Langer & Benton (2006).

Some previously suggested ornithischian synapo-morphies are not supported by this analysis. For exampleSereno (1999a) suggested that an elongate posterolateralprocess of the premaxilla diagnoses the clade; however,an elongate posterolateral process is present in the basalarchosaur Euparkeria (Ewer 1965), the basal saurischianHerrerasaurus (Sereno & Novas 1993) and the problem-atic non-ornithischian dinosauriform Silesaurus opolen-sis (Dzik 2003), and may be plesiomorphic for Dinosauria(Maryanska & Osmolska 1985).

Many of the characters identified by this and previ-ous analyses as synapomorphic for Ornithischia describe theanatomy of the dentary and maxillary teeth. These synapo-morphies have been used to refer taxa named on the basisof isolated teeth to Ornithischia (e.g. Hunt & Lucas 1994).However, recent discoveries suggest that ornithischian-liketeeth have evolved a number of times within Archosauria.For instance, recently discovered cranial and postcranial ma-terial of the putative ornithischian Revueltosaurus callenderisuggests that this taxon is actually a non-ornithischian archo-saur that appears to be phylogenetically closer to crocodylo-morphs than to dinosaurs and that its ornithischian-like den-tition evolved independently (Parker et al. 2005a, b; Irmiset al. 2007). In addition, ornithischian-like teeth have beendescribed in Silesaurus opolensis (Dzik 2003), while manyof the characteristic features of the ornithischian dentitionalso occur in taxa as diverse as therizinosaurs, basal sauro-podomorphs and aetosaurs, suggesting that dental characters(perhaps not surprisingly) may be subject to particularly highlevels of homoplasy.

The identification of ornithischian-like dental characterstates in non-ornithischian taxa highlights the problems in-herent in referring fragmentary material to specific clades.As noted by Butler et al. (2006) and Irmis et al. (2007), mostTriassic taxa named upon the basis of isolated teeth cannot,in general, be referred to Ornithischia with certainty unlessthey demonstrate, or are associated with material demon-strating, unique synapomorphies of an ornithischian clade(i.e. features that are not independently synapomorphic forother clades).

Pisanosaurus mertii

Pisanosaurus mertii is generally believed to be the oldestknown ornithischian (Casamiquela 1967), but both the as-sociation of the material included within the holotype speci-men and its phylogenetic position, have proved controversial.Sereno (1991a) proposed that the holotype specimen con-tained material from more than one individual; he suggestedthat the skull fragments, partial impression of the pelvis anddistal hind limb might belong together, but that the fragment-ary scapula and other postcranial bones were too small to bereferred to the same individual. However, Bonaparte (1976)noted that the vertebrae were recovered in articulation withthe skull and that the skeleton may have been complete priorto weathering, and a recent review supported the idea thatthe type specimen represents a single individual (Irmis et al.2007).

Bonaparte (1976) referred Pisanosaurus to Heterodon-tosauridae on the basis that both share subcylindrical, closelypacked cheek teeth, with wear facets forming a more orless continuous surface extending along the tooth row. How-ever, Weishampel & Witmer (1990a) and Sereno (1991a)


considered Pisanosaurus to be the most basal known or-nithischian. Weishampel & Witmer (1990a) suggested thatthe similarities between Pisanosaurus and heterodontosaur-ids are plesiomorphic, while Sereno (1991a: 174) noted that:The [wear] facets . . . do not form a continuous occlusal sur-face as occurs in Heterodontosaurus [tucki]. More recently,Norman et al. (2004a) have once again emphasised the mor-phological similarities between the cranial material of Pis-anosaurus and that of heterodontosaurids.

The SCC tree (Fig. 2) recovered by this analysis posi-tions Pisanosaurus in an unresolved polytomy at the base ofOrnithischia. However, the 50% majority-rule consensus tree(Fig. 2), maximum agreement subtree (Fig. 3) and the derivat-ive SRC tree (Fig. 4) all support the position of Pisanosaurusas the most basal known ornithischian. These consensus treesadditionally position heterodontosaurids as a monophyleticclade of non-genasaurians, close to the base of Ornithischia(discussed below). Pisanosaurus and heterodontosaurids arenot separated in the SCC (this is the result of the instability ofthe fragmentary wildcard taxon Echinodon becklesii); onlyone node separates Pisanosaurus from heterodontosaurids inthe derivative SRC tree. This node is weakly supported bybootstrap proportions: it does not appear in the total-evidencebootstrap analysis (support of only 37%) and has support ofonly 54% in the reduced bootstrap analysis (Fig. 4). Only onecharacter (Character 206, Appendix 2) unambiguously sup-ports this node. Constraining Pisanosaurus and heterodon-tosaurids to form a monophyletic clade requires only one ad-ditional step. This suggests that the two opposing phylogen-etic positions for Pisanosaurus, as either a non-genasaurianbasal ornithischian (Sereno 1991a) or a heterodontosaurid(Bonaparte 1976), are not necessarily mutually exclusive.Pisanosaurus may indeed represent a heterodontosaurid andthe evidence for this should be reconsidered by future work.

Heterodontosauridae

The position of heterodontosaurids within Ornithischia is oneof the most problematic areas in ornithischian phylogeny andthere is no current consensus on this topic. Four alternativephylogenetic positions have been proposed: (1) as basal or-nithopods (e.g. Crompton & Charig 1962; Thulborn 1971;Galton 1972; Santa Luca et al. 1976; Sereno 1984, 1986,1999a; Gauthier 1986; Weishampel 1990; Weishampel &Witmer 1990b; Smith 1997; Norman et al. 2004c); (2) asthe sister group to Marginocephalia (Maryanska & Osmolska1984; Cooper 1985; Olshevsky 1991; Zhao et al. 1999; Buch-holz 2002; You et al. 2003; Norman et al. 2004c; Xu et al.2006); (3) as the sister group to Ornithopoda + Margino-cephalia (Cerapoda) (Norman 1984; Maryanska & Osmolska1985; Butler 2005); (4) as the most basal well-known ornith-ischians (Bakker & Galton 1974; Olsen & Baird 1986). Thisanalysis supports the fourth of these positions and the phylo-genetic support for this is discussed below, although a fullreview of the anatomical evidence will be presented else-where (R. J. B., unpublished results).

Four taxa (Heterodontosaurus tucki, Abrictosaurus con-sors, Echinodon becklesii, Lycorhinus angustidens) and onespecimen (BMNH A100) previously referred to Heterodon-tosauridae were included in this analysis. These taxa do notform a clade in the SCC tree, but are included in a largepolytomy at the base of Ornithischia (Fig. 2). However, re-duced consensus trees indicate that this basal polytomy is

the result of the unstable and problematic taxon Echinodon(which is highly fragmentary, with 83% missing data); ex-clusion of Echinodon results in the remaining four OTUsforming a heterodontosaurid clade, while further exclusionof Lycorhinus (86% missing data) in the derivative SRC treeresolves relationships within Heterodontosauridae (Fig. 4).

The heterodontosaurid node is weakly supported bybootstrap proportions in the total-evidence bootstrap ana-lysis (support of only 19%); however, bootstrap support isconsiderably higher (68%) in the reduced bootstrap analysis(Fig. 4), suggesting that low bootstrap support for Heterodon-tosauridae is probably the result of fragmentary and unstablewildcard taxa. To test whether similar hidden Bremer sup-port exists for Heterodontosauridae we wrote a backboneconstraint that specified that Heterodontosaurus, Abricto-saurus and BMNH A100 form a clade to the exclusion ofother ornithischians, but did not specify the position of thefive unstable taxa (Echinodon, Lycorhinus, Zephyrosaurus,Talenkauen and Yandusaurus) discussed above and removedin the derivative SRC tree. The Bremer support for this back-bone constraint was +2, again suggesting that low supportfor Heterodontosauridae is the result of unstable wildcardtaxa.

The MPTs recovered in this analysis do not supporta link between heterodontosaurids and ornithopods (phylo-genetic hypothesis 1, above). In order to test this further weran the following constrained analyses: firstly, Ornithopoda(sensu Weishampel 1990; Sereno 1999a), containing hetero-dontosaurids, hypsilophodontids and iguanodontids, wasconstrained as monophyletic, although Hypsilophodontidae(sensu Weishampel & Heinrich 1992) was not constrained asa monophyletic clade; secondly, a backbone constraint wasspecified that required that Heterodontosaurus and Ankylo-pollexia be more closely related to each other than eitheris to marginocephalians, thyreophorans or ornithischian out-groups. This constraint does not specify the complete contentof Ornithopoda.

Templeton non-parametric tests were carried out usingPAUP that compared trees within the profile of 756 MPTsgenerated by the unconstrained analysis with all trees re-covered by the constrained analyses. Ideally all MPTs shouldbe compared with all trees recovered by the constrained ana-lyses; however, as this is time intensive (this would involve756 separate Templeton tests) a subset of the MPTs (every25th MPT) was used.

Un

Date post:	11-Oct-2015
Category:	Documents
Upload:	fernanda-ramirez-corchado
View:	43 times
Download:	0 times

The Phylogeny of the Ornithischian

Documents