Phylogenetic and Biogeographic Assessment of Ornithischian Diversity Throughout the Mesozoic: A Species-Level Analysis from Origin to Extinction

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations Summer 2013 Phylogenetic and Biogeographic Assessment of Ornithischian Diversity Throughout the Mesozoic: A Species-Level Analysis from Origin to Extinction Marc Richard Spencer University of Iowa Copyright 2013 Marc Richard Spencer This dissertation is available at Iowa Research Online: Recommended Citation Spencer, Marc Richard. "Phylogenetic and Biogeographic Assessment of Ornithischian Diversity Throughout the Mesozoic: A Species-Level Analysis from Origin to Extinction." PhD (Doctor of Philosophy) thesis, University of Iowa, Follow this and additional works at: Part of the Geology Commons

2 PHYLOGENETIC AND BIOGEOGRAPHIC ASSESSMENT OF ORNITHISCHIAN DIVERSITY THROUGHOUT THE MESOZOIC: A SPECIES-LEVEL ANALYSIS FROM ORIGIN TO EXTINCTION by Marc Richard Spencer A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Geoscience in the Graduate College of The University of Iowa August 2013 Thesis Supervisor: Associate Professor Christopher A. Brochu

3 Copyright by MARC RICHARD SPENCER 2013 All Rights Reserved

4 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL PH.D. THESIS This is to certify that the Ph.D. thesis of Marc Richard Spencer has been approved by the Examining Committee for the thesis requirement for the Doctor of Philosophy degree in Geoscience at the August 2013 graduation. Thesis Committee: Christopher A. Brochu, Thesis Supervisor Jonathan M. Adrain Ann F. Budd John M. Logsdon, Jr. Peter J. Makovicky

5 To my family. All that I have become and ever hope to become is because of your love and support. ii

6 We must not believe the many, who say that only free people ought to be educated, but we should rather believe the philosophers who say that only the educated are free. Epictetus Discourses iii

7 ACKNOWLEDGEMENTS First and foremost, I would like to sincerely thank my advisor, Chris Brochu. Your insight, patience, and guidance were and continue to be the most invaluable asset I have in my academic life. To my committee members Jonathan Adrain, Nancy Budd, John Logsdon, and Pete Makovicky whether it was through course instruction, your scholarly papers, or just general discussions, each of you has left a tremendously beneficial indelible mark on my nascent career and for that I cannot thank you enough. I am indebted to Eric Wilberg and members, both past and present, of the University of Iowa Systematics and Vertebrate Paleontology Discussion Group for helpful comments and support throughout the course of my graduate career. I would not be, could not be, where I am today without the moral support of my loving family and friends. Chris Brochu generously provided the skeletal reconstructions used in Figures 1.10, , , Julia McHugh and Pete Makovicky provided some very helpful photographs of several South African specimens. Peter Galton (Univ. Bridgeport) provided many helpful photographs of Othnielosaurus consors. In the early stages of my biogeographic analysis, Paul Upchurch (UCL) provided extremely helpful insight and guidance, particularly with dealing with the temperament of TreeFitter, and I am truly grateful for his assistance. Funding for my dissertation was generously provided by the Max and Lorraine Littelfield Fund, the Swade Fund, the University of Iowa Department of Geoscience Strategic Initiative Fellowship, the University of Iowa Graduate College, the University of Iowa Graduate Student Senate, and T. Anne Cleary International Dissertation Research Fellowship. Logistical support has been and continues to be provided by the staff and patrons of Joe s Place, IC. Additional support throughout the course of writing this dissertation (in the form of background noise and the occasional necessary distraction) iv

8 was provided by the creators of Friends, Scrubs, and Psych. The dulcet tones of Boyz II Men also provided much-needed soothing vibes throughout the writing process. Museum collections are central to paleontology and other fields of evolutionary biology and natural history. It follows, then, that the curators and collections managers of those collections are invaluable to science. Therefore, I would like to thank the following individuals for access to collections: Sandra Chapman and Lorna Steel (NHMUK); Amy Henrici (CM); Jessica Cundiff and Mara Lyons (MCZ); Dan Brinkman (YPM); Michael Brett-Surman (USNM); Mark Norell and Carl Mehling (AMNH); Roger Smith and Sheena Kaal (SAM); Bruce Rubidge and Bernard Zipfel (BP); Pete Makovicky, Olivier Rieppel, and Bill Simpson (FMNH); and Rodrigo Pellegrini and Dave Parris (NJSM). v

9 TABLE OF CONTENTS LIST OF TABLES... viii LIST OF FIGURES... ix LIST OF ANATOMICAL ABBREVIATIONS USED IN THE TEXT... xi CHAPTER I. ORNITHISCHIAN PHYLOGENY AND THE LABILITY OF PROBLEMATIC BASAL TAXA...1 Abstract...1 Introduction...2 Background...3 Taxonomic and Phylogenetic History of Problematic Ornithischians...9 Materials and Methods...23 Taxon and Character Sampling Strategies...23 Phylogenetic Analysis...31 Results...32 Discussion...37 Clade Support...38 Ornithischian Phylogeny...45 Conclusion...47 II. THE MORPHOLOGY AND ANCESTRAL BODY PLAN OF ORNITHISCHIA...48 Abstract...48 Introduction...49 Morphological History of Basal Ornithischia...51 Methods...56 Phylogenetic Analysis...56 Ancestral State Reconstruction...57 Results...58 Discussion and Description...61 Cranial Osteology...62 Mandibular Osteology...66 Dentition...68 Postcranial Osteology Axial...69 Postcranial Osteology Pectoral Girdle and Forelimb...71 Postcranial Osteology Pelvic Girdle and Hindlimb...78 Postcranial Osteology Accessory Ossifications...82 Comparison with Ornithischian Subclades...82 Conclusion...83 III. DISTRIBUTIONAL PATTERNS IN ORNITHISCHIA: A COMPREHENSIVE SPECIES-LEVEL BIOGEOGRAPHIC ANALYSIS...84 vi

10 Abstract...84 Introduction...85 Biogeography...86 Ornithischian Biogeography...87 Materials and Methods...91 Ornithischian Phylogeny...91 Event-Based Parsimony Tree Fitting...94 Input Hypotheses...97 Global Biogeographic Analysis of Ornithischia...99 Discussion Time-Sliced Analyses Conclusion IV. ORNITHISCHIAN PHYLOGENY AND ITS BROADER IMPACT ON PHYLOGENY ESTIMATION, CHARACTER EVOLUTION, AND BIOGEOGRAPHY Introduction Phylogenetic Analysis Ancestral Body Plan and Character Evolution Biogeographic Analysis Conclusion APPENDIX A. INSTITUTIONAL ABBREVIATIONS APPENDIX B. SPECIES INCLUDED IN THE PHYLOGENETIC ANALYSIS APPENDIX C. CHARACTER LIST AND CHARACTER STATE DESCRIPTIONS APPENDIX D. TAXON-CHARACTER MATRIX FOR PHYLOGENETIC ANALYSIS APPENDIX E. APOMORPHY LIST FOR THE PHYLOGENETIC ANALYSIS PRESENTED IN CHAPTER I APPENDIX F. CHARACTER DESCRIPTIONS AND ASSOCIATED CHARACTER STATES FOR THE RECONSTRUCTION OF THE ANCESTRAL ORNITHISCHIAN DINOSAUR APPENDIX G. SPECIES, GEOLOGIC AGE, AND THE PHYLOGENETIC REFERENCE LIST APPENDIX H. BIOGEOGRAPHIC ANALYSIS NEXUS FILE FORMAT AND TAXA REFERENCES vii

11 LIST OF TABLES Table 2.1. Traditional ornithischian synapomorphies defined by Sereno (1986) Area cladogram hypotheses Number of significant events on each of the eight alternative area cladograms C1 Characters that include relative measurements F1 Reconstructed ancestral states for the hypothetical ancestral ornithischian (A.O.), Genasauria (Gen.), Neornithischia (Neo.), Thyreophora (Thy.), Cerapoda (Cer.), and Marginocephalia (Mar.) viii

12 LIST OF FIGURES Figure 1.1. Phylogeny of Dinosauria presented in Bakker and Galton (1974) Traditional ornithischian phylogeny following Sereno (1986) Right maxilla of SAM-PK-K426 (Lesothosaurus diagnosticus) Ornithischian phylogeny of Butler (2005) Consensus ornithischian phylogeny from Butler et al. (2007, 2008, 2010, 2011) Preliminary phylogenetic analysis of Ornithischia by Spencer (2007) Problematic South African heterodontosaurids Strict consensus tree of all 990 MPTs resulting from the phylogenetic analysis of 54 taxa and 247 characters Maximum agreement subtree of all 990 MPTs found during the phylogenetic analysis Competing phylogenetic hypotheses for the placement of A) Heterodontosauridae, B) Lesothosaurus, and C) Ornithopoda Summary phylogeny of Ornithischia from the phylogenetic analysis presented in Chapter I Reconstructed cranium of the hypothetical ancestral ornithischian Reconstructed mandible of the ancestral ornithischian Reconstructed axial elements of the ancestral ornithischian Left forelimb of Heterodontosaurus tucki (SAM-PK-K1332) in lateral view Forelimb of Abrictosaurus consors (NHMUK RU B54) in ventral view Reconstructed pectoral girdle elements of the ancestral ornithischian Right forelimb of Heterodontosaurus tucki (SAM-PK-K1332) in dorsal view Distal (A) left tibia of Stormbergia dangershoeki (NHMUK R11000) and (B) right tibia of Lesothosaurus diagnosticus (SAM-PK-K1106) in posterior view Reconstructed pelvic girdle elements of the ancestral ornithischian Mesozoic time scale and the break-up of Pangaea...88 ix

13 3.2. Ornithischian global diversity throughout the Mesozoic Time-calibrated composite phylogeny for Ornithischia Generalized break-up sequence of Pangaea throughout the Mesozoic Optimal area cladogram for the comprehensive species-level phylogeny Time-calibrated phylogeny of Ornithopoda Time-calibrated phylogeny of basal ornithischians and Thyreophora Time-calibrated phylogeny of basal neornithischians and Marginocephalia Eight alternative area cladograms Early Jurassic distribution of ornithischians and the single OAC that best explains the biogeographic pattern Middle Jurassic distribution of ornithischians and the single OAC that best explains the biogeographic pattern Late Jurassic distribution of ornithischians and the two OACs that best explain the biogeographic pattern Early Cretaceous distribution of ornithischians and the single OAC that best explains the biogeographic pattern Middle Cretaceous (Aptian-Albian) distribution of ornithischians and the two OACs that best explain the biogeographic pattern Late Cretaceous distribution of ornithischians and the single OAC that best explains the biogeographic pattern End-Cretaceous (Maastrichtian) distribution of ornithischians and the three OACs that best explain the biogeographic pattern Complete reconstruction of the ancestral hypothetical ornithischian dinosaur E1. Strict consensus tree from the phylogenetic analysis presented in Chapter I x

14 LIST OF ANATOMICAL ABBREVIATIONS USED IN THE TEXT aofo antorbital fossa as - astragalus d dentary dia diastema ep extensor pits h - humerus mc (1-4) metacarpal olp olecranon process of the ulna ph (I-II)-(1-2) phalanx plpt posterolateral process of the tibia prdc posterior ridge on the distal crown of the cheek tooth ra radius rae - radiale sp sternal plate spl splenial ul ulna ule - ulnare un 1-2 manual ungual xi

15 1 CHAPTER I ORNITHISCHIAN PHYLOGENY AND THE LABILITY OF PROBLEMATIC BASAL TAXA Abstract Monophyly of derived ornithischian clades such as Stegosauria, Iguanodontia, Ceratopsia, and Pachycephalosauria has been corroborated in numerous analyses. The relationships of these clades within Ornithischia, however, are less certain. Moreover, the placement of taxa such as Lesothosaurus diagnosticus, Heterodontosauridae, and traditional basal ornithopods (basal cerapodans and neornithischians) are unresolved. Lesothosaurus was generally considered one of the basalmost ornithischians; however, until recently, it was never included as an ingroup taxon in a quantitative species-level phylogenetic analysis of all representative ornithischians. Heterodontosaurids have been recovered as basal ornithopods, basal cerapodans, basal ornithischians, and as the sister group to Marginocephalia. Ornithopoda traditionally considered a well-supported clade containing hypsilophodontids and Iguanodontia collapsed into a polytomy at the base of Cerapoda in several recent analyses and the relationships of these taxa to Marginocephalia are unresolved. A phylogenetic analysis was performed on 52 ornithischian species representing both basal and derived taxa and a matrix of 247 discrete characters. The resultant phylogeny is fairly well resolved with several interesting results. Lesothosaurus was recovered as the basalmost neornithischian; a result noted in only a few analyses. Additionally, it requires nine extra steps to move

16 2 Lesothosaurus outside of Genasauria. Heterodontosauridae was robustly recovered at the base of Ornithischia. A monophyletic Ornithopoda was not recovered, and, if traditional clade membership is considered (including heterodontosaurids and basal neornithischians such as Othnielosaurus and Agilisaurus; Ornithopoda sensu lato), then the taxon is polyphyletic (i.e., heterodontosaurids as basal ornithischians). The results presented here lend support to several hypotheses, both traditional and recent, of the placement of Lesothosaurus, heterodontosaurids, ornithopods, basal cerapodans, and basal neornithischians. With a fairly well-resolved phylogeny at least among the basalmost taxa and the major clades of ornithischians (except for Ornithopoda), it is possible to address questions of character evolution and biogeography within a strict phylogenetic framework. Furthermore, the uncertainty around what comprises stem and basal cerapodans underscores the need for an even larger taxonomic sampling to illuminate ornithischian relationships. Introduction Ornithischia is a well-known and diverse extinct clade of herbivorous dinosaurs including such icons as the spike-tailed stegosaurs and the horned ceratopsians. Ornithischians achieved a global distribution during a time of fundamental changes to global continental configuration and climate (Weishampel et al., 2004b). They were the dominant large herbivores in many Mesozoic faunas, and their diversification may have been coupled with the radiation of flowering plants (e.g., Butler et al., 2009a, b, 2010a). They are the ideal group for large-scale analyses of biodiversity, biogeography, and

17 3 coevolution. Ornithischians are central to studies of the evolution of herbivory (e.g., Barrett, 2000), life history (e.g., Sampson et al., 1997), physiology (e.g., Goodwin and Horner, 2004), and sexual dimorphism in the fossil record (e.g., Padian and Horner, 2011). Ornithischia is also one of the few extinct clades to which model-based phylogenetic methods (e.g., Bayesian inference) have been applied (e.g., Prieto-Márquez, 2010), making it a model clade for empirically exploring the influence of missing data, taxon sampling, input model parameters, and character coding on quantitative phylogeny estimation methods with morphological data. Many of these questions, though, rely directly or indirectly upon a thoroughly sampled and well-resolved phylogenetic hypothesis for the entire clade, one that is currently lacking in the literature. Background Dinosauria, first coined by Owen (1842), is subdivided into two major clades, Saurischia and Ornithischia (Seeley, 1887). Cope (1866) was the first to consider the group of dinosaurs now known as ornithischians to have been descended from a single common ancestor. Since its first formal designation by Seeley (1887), Ornithischia has been generally accepted as a monophyletic group (e.g., Thulborn, 1971a; Galton, 1972; Bakker and Galton, 1974; Bonaparte, 1976; Gauthier, 1986; Weishampel and Witmer, 1990b). Marsh recognized several groups that are still recognized today, though the taxonomic content has changed, including Stegosauria (1877), Ornithopoda (1881), and Ceratopsia (1890), which he later grouped together into Predentata (Marsh, 1894; Predentata is the junior synonym of Ornithischia). Like Marsh, various workers after Seeley (1887) designated subgroups within Ornithischia based on overly simplistic traits

18 4 such as bipedal forms or armored forms. Nopcsa (1915) coined the name Thyreophora to include all armored forms, though he included ceratopsians as well as stegosaurs and ankylosaurs. Romer (1956) suggested that Ornithopoda was the ancestral stock for Stegosauria, Ankylosauria, and Ceratopsia, based on their bipedal nature and lack of specialized armor. As such, Ornithopoda became a wastebasket taxon to which all ornithischians were assigned that did not fit into Ceratopsia, Stegosauria, and Ankylosauria. Bakker and Galton (1974) suggested the monophyly of Dinosauria and in doing so, illustrated the relationships of the major groups of ornithischians (including Romer s [1956] four orders) and saurischians, including birds. The notion of dinosaur monophyly was not new, nor was the hypothesis of dinosaurian origins for modern birds; however, Bakker and Galton s (1974) study was among the first to depict the relationships in the form of an evolutionary branching diagram (Fig. 1.1; though, Thulborn [1971a:fig. 4] suggested ornithischian monophyly as well). Gauthier (1986) cladistically tested Bakker and Galton s hypothesis and corroborated the monophyly of not only Dinosauria, but also its subclades Saurischia (including birds) and Ornithischia. His analysis, which focused on saurischian dinosaurs, was complemented by independent analyses of Ornithischia (Norman, 1984; Sereno, 1984, 1986; Cooper, 1985; Maryańska and Osmólska, 1985); though, Gauthier s (1986) was the first to truly quantify these relationships. The general pattern of ornithischian relationships that has prevailed since Gauthier (1986) is based predominantly on Sereno (1986). Ornithischia (Fig. 1.2) has been parsed out into its basal forms consisting of Pisanosaurus mertii, Lesothosaurus

19 Figure 1.1 Phylogeny of Dinosauria presented in Bakker and Galton (1974). Saurischia has been collapsed here for clarity. Modified from Bakker and Galton (1974). 5

20 Figure 1.2 Traditional ornithischian phylogeny following Sereno (1986). Pisanosaurus is added here following Sereno (1997, 1999). Sereno s phylogeny was the most influential and dominant view of ornithischian phylogenetics for more than two decades. 6

21 7 diagnosticus, and Genasauria (Sereno, 1997, 1999; Weishampel, 2004; Norman et al., 2004a). Thyreophora and Cerapoda are the two subdivisions that have typically comprised Genasauria (Sereno, 1986, 1997, 1999; Weishampel, 2004). Thyreophorans include the armored dinosaur groups Stegosauria (Huayangosaurus taibaii + Stegosauridae) and Ankylosauria (Nodosauridae + Ankylosauridae) as well as basal sister taxa Scutellosaurus lawleri, Scelidosaurus harrisonii, and Emausaurus ernsti (Norman et al., 2004b). Cerapoda is a more taxonomically-diverse clade than its sister taxon Thyreophora and has been divided into Ornithopoda and Marginocephalia (Sereno, 1997, 1999; Weishampel, 2004). Sereno (1997, 1999) suggested that Cerapoda, a name he coined (Sereno, 1986), was a junior synonym of Neornithischia, a name that Cooper (1985) erected. However, these two names are taxonomically different as Cooper s Neornithischia excludes Ornithopoda, though it does include Neornithopoda consisting of ceratopsians, pachycephalosaurs, and heterodontosaurids. Neornithischia has been resurrected in recent studies, beginning with Butler (2005), as a more inclusive clade that contains a more exclusive Cerapoda plus several stem taxa. Ornithopoda has been further divided into Heterodontosauridae and Euornithopoda (Weishampel, 1990; Norman et al., 2004c). This differs from Sereno s (1986) original classification in which he termed Euornithopoda as the more inclusive clade that included Heterodontosauridae and Ornithopoda. Euornithopoda (sensu Sereno, 1986) consists of Hypsilophodontidae and Iguanodontia. Hypsilophodontidae has recently been demonstrated as a paraphyletic grade of ornithopods (Weishampel et al.,

22 8 2003; Norman et al., 2004c; Butler et al., 2007, 2008). Recent analyses suggest that Ornithopoda might not be monophyletic (Spencer, 2007; Butler et al., 2008, 2010, 2011). In strict consensus trees of these analyses, the clade collapses into a polytomy at the base of Cerapoda. In some most parsimonious trees, Iguanodontia has been recovered as more closely related to Marginocephalia than to other ornithopods (e.g., hypsilophodontids ). Monophyly of Iguanodontia has been supported in nearly all phylogenetic analyses that have tested the group. The sister taxon to Ornithopoda, Marginocephalia, is divided into Pachycephalosauria and Ceratopsia (Sereno, 1986, 1997, 1999; Weishampel, 2004). Following Norman (1984) and Sereno (1984, 1986), most analyses have focused on the interrelationships of ceratopsians (e.g., Sereno, 1989, 1990; Dodson, 1990, 1997a, 1997b; Xu et al., 2002; You and Dodson, 2003, 2004; Zhou et al., 2006; Xu et al., 2006) or pachycephalosaurs (e.g., Sereno, 1989, 2000; Maryańska, 1990; Maryańska et al., 2004), but rarely on Marginocephalia as a whole. Since Sereno (1986), the overwhelming majority of phylogenetic analyses focused solely on derived clades within Ornithischia (e.g., Stegosauria, Galton and Upchurch, 2004; Ankylosauria, Vickaryous et al., 2004; Ornithopoda, Norman et al., 2004c; Iguanodontia, Norman, 2004; Hadrosauridae, Horner et al., 2004; Pachycephalosauria, Maryańska et al., 2004; Ceratopsia, You and Dodson, 2004). Indeed, prior to Butler (2005, and various iterations of his initial data set that followed), a largescale, quantitative phylogenetic analysis of basal ornithischian dinosaurs had never been published (Weishampel, 2004).

23 9 Taxonomic and Phylogenetic History of Problematic Ornithischians The type and only known specimen of Pisanosaurus is from the Ischigualasto Formation (Carnian) of Argentina (Casamiquela, 1967). The material consists of partially articulated skeletal elements, including a fragmentary dentary and maxilla and central portion of the right pelvis including the proximal ischium and pubis as well as the femoral head, the right tibia and fibula, and fragments of the pes and several dorsal (and possibly cervical) vertebrae. Bonaparte has produced the only published reconstruction of Pisanosaurus (1976:fig. 8). Casamiquela (1967) noted that Pisanosaurus is the earliest known ornithischian (and, to date, still is the earliest known valid ornithischian) and placed it in Ornithopoda as a basal member within a new clade termed Pisanosauridae. Galton (1972) indicated that, although it exhibits several plesiomorphic ornithischian features, Pisanosaurus was a hypsilophodontid, whereas, basal ornithischians Fabrosaurus australis and Echinodon becklesii were members of a new family, the Fabrosauridae. Galton (1974b) placed Pisanosaurus as the sister taxon to hypsilophodontids, then reverted back to his original hypothesis (1972) and placed Pisanosaurus as a basal member of Hypsilophodontidae (1974a), concluding that hypsilophodontids were derived from cursorial fabrosaurids and that Pisanosaurus was among that lineage. Bonaparte (1976) concluded that Pisanosaurus represented a basal member of Heterodontosauridae, despite the lack of heterodonty as exhibited in Heterodontosaurus and Lycorhinus. Weishampel and Witmer (1990b) contended that although Pisanosaurus shares several features with heterodontosaurids such as occlusal tooth wear, until further discoveries are made, these characters appear to be the product of convergent evolution.

24 10 The presumed presence of the predentary and buccal emargination are characters that make Pisanosaurus an ornithischian dinosaur. Casamiquela (1967), Bonaparte (1976), and Irmis et al. (2007) indicate a derived ornithischian feature, the absence of an external mandibular fenestra (contra Sereno [1991] and Butler [2005]). The presence of an opening in the posterior mandibular ramus, as indicated by Irmis et al. s (2007:fig. 6) personal observations of the irregular broken margins, appears to be an unnatural break (also indicated by Sereno s [1991:p.174] observation that the medial mandibular fossa maintains openings at the anterior and ventral margins it seems evident that this putative external mandibular fenestra has fragmented further between Sereno s observations and Irmis et al. s [2007] observations). Another derived ornithischian feature of Pisanosaurus is the extensive wear facets on the maxillary and dentary teeth (Bonaparte, 1976; Sereno, 1991). Sereno (1991) and Irmis et al. (2007) suggest that Bonaparte s (1976) presumed posteroventral position of the pubis is unwarranted due to the actual preservation of the pelvic impression. Sereno (1991) and Irmis et al. (2007) agree that the pubis was directed anteroventrally and lacked a prepubic process, thus making these ancestral traits of Ornithischia. Collectively, these characters have led some (Sereno, 1991; Norman et al., 2004a) to suggest that the holotype may be a chimera. According to Irmis et al. (2007) and Bonaparte s in situ map of the preserved semi-articulated skeletal elements (1976:fig. 1), the material assigned to the holotype is likely from one individual. Therefore, it is likely that, given its age and plesiomorphic features, Pisanosaurus represents one of the most basal ornithischians known.

25 11 Lesothosaurus has been considered one of the basalmost ornithischian dinosaurs and historically has been considered the sister taxon to all other ornithischians (Genasauria; Sereno, 1986, 1997, 1999; Weishampel and Witmer, 1990a; Norman et al., 2004a). One character that is considered to be the ancestral state for all ornithischians that Lesothosaurus reportedly did not possess is buccal emargination (the maxillary and dentary tooth rows are marginal). However, there is at least incipient emargination present in the maxillae of Lesothosaurus (e.g., SAM-PK-K426; Fig. 1.3). While the lack of emargination is plesiomorphic for Dinosauria (saurischians and related dinosaur outgroups do not demonstrate buccal emargination ancestrally), it appears that all taxa with preserved maxillae and/or dentaries possess at least some rudimentary degree of buccal emargination (including the heterodontosaurid Abrictosaurus, NHMUK RUB54). Originally, Ginsburg (1964) described a partial dentary with three preserved teeth as the holotype for a new species Fabrosaurus australis. He considered it most closely related to Scelidosaurus harrisonii based on the similarity in the morphology of the dental characters, but regarded Fabrosaurus as a primitive ornithischian due to the smaller size of the specimen and the vertical nature of the teeth as opposed to the recumbent appearance of Scelidosaurus teeth. In a series of papers, Thulborn (1970a, 1971b, 1972) described far more complete specimens that he referred as the cranial and postcranial remains of Fabrosaurus. Based on his observations, he noted that the affinity to Scelidosaurus was inaccurate based on, among other things, the lack of dermal armor and the uncertain relationship of Scelidosaurus to other ornithischians (Thulborn, 1970a)

26 Figure 1.3 Right maxilla of SAM-PK-K426 (Lesothosaurus diagnosticus). SAM-PK- K426 is similar to other maxillae in that it demonstrates a weakly developed maxillary ridge providing evidence of at least incipient buccal emargination. Scale bar = 1 cm. 12

27 13 and placed Fabrosaurus in Hypsilophodontidae (Thulborn, 1972). Galton (1972) noted the primitive nature of Fabrosaurus as well as several other taxa including Echinodon becklesii and Nanosaurus rex (= Othnielia rex, which has since been redescribed and reassigned Othnielosaurus consors [Galton, 2007]) in that the lateral surface of the maxilla is flat or only incipiently indented (little or no buccal emargination), thus separating them from the rest of ornithischian dinosaurs. He erected Fabrosauridae as the basalmost group of ornithischians to reflect this distinction. Charig and Crompton (1974) noted that the only feature of the holotype of Fabrosaurus that was characteristic of Ornithischia was the plesiomorphic appearance of the teeth. They suggested that the genus Fabrosaurus and the species F. australis should be considered nomina dubia and that no other specimens be referred to either. Upon further review of Fabrosaurus material, Galton (1978) reassigned Thulborn s (1970a, 1971b, 1972) specimens to a new genus and species Lesothosaurus diagnosticus and kept the holotype of Ginsburg s (1964) specimen within the taxon Fabrosaurus australis. The major difference between Fabrosaurus and Lesothosaurus, according to Galton (1978), was the presence of special foramina (for the access of replacement teeth) in the material from Fabrosaurus, but not present in Lesothosaurus. However, Gow (1981) and Sereno (1991) indicated that these special foramina were present in Lesothosaurus (e.g., SAM-PK-K426; Fig. 1.3) and so this feature could not be used as an autapomorphy for Fabrosaurus. Weishampel and Witmer (1990a) and Sereno (1991) considered Fabrosaurus and F. australis nomina dubia. Sereno (1986, 1991) and Weishampel and Witmer (1990a) consider the family Fabrosauridae a nomen dubium based on its polyphyletic grouping (taxa such as Echinodon and Scutellosaurus are more closely

28 14 related to other ornithischians than to Lesothosaurus). In response to Galton (1978), Weishampel and Witmer (1990a), and Sereno (1991), Thulborn (1992) reaffirmed his conclusions about Fabrosaurus australis and considered Galton s (1978) Lesothosaurus diagnosticus as an invalid junior synonym. Other than Peng (1992, 1997), no known analysis or description has considered Fabrosaurus australis a valid taxon. Crompton and Charig (1962) described the primitive ornithischian Heterodontosaurus tucki, though the family Heterodontosauridae was not erected until after its description (Kuhn, 1966). Hopson (1975) considered three taxa, Heterodontosaurus, Abrictosaurus, and Lycorhinus, based primarily on tooth morphology to be most closely related within Heterodontosauridae. In this assignment, he considered Lycorhinus to be the basalmost heterodontosaurid. Thulborn (1970b) described a specimen that he concluded was so similar to Lycorhinus angustidens that it was conspecific and further suggested that these were similar to Heterodontosaurus tucki and that the genus Heterodontosaurus should be considered invalid and assigned the new material to Lycorhinus tucki. This was later shown to be an incorrect assignment and Heterodontosaurus tucki was resurrected (Galton, 1973). In an analysis of basal ornithopods, including heterodontosaurids, Norman et al. (2004c) agreed with Hopson (1975), and included Abrictosaurus consors, Echinodon becklesii, and Lycorhinus angustidens within the clade. There are, however, two taxa that have at least been provisionally assigned to Heterodontosauridae but have been either considered a synonym of another species (i.e., Lanasaurus scalpridens Gow, 1975 = Lycorhinus angustidens Haughton, 1924) or, due to the fragmentary nature of the specimen and lack of autapomorphies, has been considered a nomen dubium (i.e.,

29 15 Geranosaurus atavus Broom, 1911). Neither of these taxa has been included in a phylogenetic analysis as distinct species. Lanasaurus, however, has been included in several analyses coded as Lycorhinus (Norman et al., 2004c; Butler et al., 2008). The taxonomic validity of Lanasaurus is discussed below. Heterodontosauridae has historically been considered the basalmost group within Ornithopoda (Weishampel and Witmer, 1990b; Norman et al., 2004c); however, it has been suggested that it might be outside Cerapoda (e.g., Butler, 2005; Spencer, 2007) and has been recovered as the basalmost clade of Ornithischia, Pisanosaurus notwithstanding (Butler et al., 2008, 2010, 2011). Xu et al. (2006), however, described a new basal ceratopsian from China, Yinlong downsi, and in the process recovered Heterodontosauridae as the sister taxon to Marginocephalia. They concluded that several of the dental characters used by Norman et al. (2004c) to define the Heterodontosauridae enlarged premaxillary teeth, chisel-shaped crowns of the maxillary teeth with denticles restricted to the apical third, prominent mesial and distal ridges on the maxillary teeth can also be used to diagnose the clade comprised of Heterodontosauridae and Marginocephalia (Heterodontosauriformes). Support for this hypothesis was weak and the few analyses since Xu et al. (2006) have failed to recover this relationship (e.g., Butler et al., 2008, 2010, 2011). The members of Heterodontosauridae are largely poorly preserved and it is likely one of the reasons that the within-clade relationships are unresolved. Norman et al. (2011) and Sereno (2012) evaluated the interrelationships of Heterodontosauridae. Norman et al. (2011) provided a detailed description of the cranial anatomy of Heterodontosaurus and a taxonomic discussion of putative South African

30 16 heterodontosaurids (Abrictosaurus, Lanasaurus, Geranosaurus, Lycorhinus, and NHMUK RU A100 [formerly BMNH A100]). Sereno (2012), in describing a new species, Pegomastax africanus, also provided a taxonomic review of heterodontosaurids, but included also Manidens, Fruitadens, Echinodon, and Tianyulong in his discussion and phylogenetic analysis. Sereno recovered two distinct clades, a Laurasian clade of Echinodon, Fruitadens, and Tianyulong and a Gondwanan clade of the South African taxa and the South American taxon Manidens. There is another putative heterodontosaurid from the Upper Triassic Laguna Colorada Formation, Argentina (Báez and Marsicano, 2001). This heterodontosaurid also extends the clade s temporal range into the Triassic and is only the third known Triassic ornithischian, the others being Pisanosaurus and Eocursor. The issues that plague heterodontosaurids are similar to those that plague basal neornithischians, cerapodans, and ornithopods. Poorly preserved or poorly described taxa have confounded the resolution at the base of Neornithischia and Cerapoda (e.g., Talenkauen, Yandusaurus, Othnielosaurus, Micropachycephalosaurus). These taxa are essential for determining the overall relationships of ornithischians. The major clades (i.e., Ankylosauria, Stegosauria, Iguanodontia, Ceratopsia, Pachycephalosauria) are well established but their relationships within Ornithischia are less well supported and the phylogenetic positions of the basal taxa of these clades are not as clear-cut. Neoceratopsians have received considerable attention (e.g., Sampson et al., 2010), but basal ceratopsians less so (e.g., Xu et al., 2006). As other derived clades have received more attention, the phylogenetic resolution has diminished (e.g., the dissolution of Hypsilophodontidae; Weishampel et al. 2003).

31 17 Ornithopoda, according to Sereno (1998), is a node-based clade defined as the most recent common ancestor of Heterodontosaurus tucki Crompton and Charig, 1962 and Parasaurolophus walkeri Parks, 1922 and all of its descendants, whereas Butler et al. (2008) indicate that it is a stem-based clade defined as all genasaurians more closely related to Parasaurolophus walkeri Parks, 1922 than to Triceratops horridus Marsh The lability of Heterodontosauridae renders Sereno s (1998) definition problematic and suggests that Butler et al. s (2008) definition is more stable. As mentioned above, Norman (1984) and Sereno (1984, 1986), independently, were the first to compile an exhaustive list of synapomorphies that diagnose Ornithischia. Since that time, myriad phylogenetic analyses have been performed on more exclusive clades of Ornithischia (e.g., Forster, 1990; Weishampel and Heinrich, 1992; Coria and Salgado, 1996; Weishampel et al., 2003; Dodson et al., 2004; Galton and Upchurch, 2004; Horner et al., 2004; Maryańska et al., 2004; Norman, 2004; Norman et al., 2004b, c; Vickaryous et al., 2004; You and Dodson, 2004; Maidment et al., 2006, 2008; Xu et al., 2006). There are, however, a limited number of published analyses that consider basal ornithischians (Butler, 2005; Langer and Benton, 2006; Butler et al., 2007, 2008, 2010, 2011). Langer and Benton (2006) conducted a phylogenetic analysis on basal dinosaurs, including basal saurischians and dinosauriforms, to evaluate dinosaur origins. The analysis did not compare the interrelationships among the basal ornithischians as it just considered Pisanosaurus mertii and a composite-coded Ornithischia versus all other dinosaurs and Silesaurus opolensis, a dinosauriform from southern Poland (Dzik, 2003). Butler s (2005) analysis evaluated the basal ornithischian affinities with the oldest and

32 18 most primitive known taxa, which included a newly described ornithischian, Stormbergia dangershoeki. His analysis challenged long-standing phylogenies of basal ornithischians and recovered Heterodontosauridae at the base of Cerapoda, an unresolved Ornithopoda, and Lesothosaurus as the basalmost member of Neornithischia (Fig. 1.4). Traditionally, Lesothosaurus had been considered the a priori outgroup to all other ornithischians (Genasauria), likely based on the supposed lack of buccal emargination and the presence of an external mandibular fenestra. Prior to Butler (2005), there are no known analyses that a priori place Lesothosaurus within the ingroup to be tested with other basal ornithischians. Butler et al. (2007) considered basal ornithischian affinities and included a complete description of a new taxon, Eocursor parvus, which the previous study of Butler (2005) also included, but referred to it as the then-undescribed specimen SAM- PK-K8025. This analysis suggested that Pisanosaurus and heterodontosaurids are basal to all other ornithischians. Contrary to Butler (2005), Lesothosaurus was recovered in a polytomy at the base of Genasauria with Thyreophora and Neornithischia. Eocursor was recovered as the sister taxon to Genasauria. Butler et al. (2008), building off of the data matrix used in Butler et al. (2007), recovered a monophyletic Neornithischia, but the basal relationships were unresolved (Fig 1.5). Heterodontosauridae, in the strict consensus tree, collapsed into a polytomy at the base of Ornithischia with Pisanosaurus and Lesothosaurus recovered as the basalmost thyreophoran. Butler et al. (2010) recovered a monophyletic Heterodontosauridae; however, the within-clade relationships were unresolved. They included two new heterodontosaurid taxa, Fruitadens haagarorum and Tianyulong confuciusi into the matrix. Lesothosaurus was recovered as the basalmost neornithischian and Eocursor was the sister taxon to Genasauria. The basal

33 19 relationships of Cerapoda, however, were unresolved with the basal members of Ornithopoda collapsed into a polytomy. Iguanodontia, though, was recovered as a monophyletic group. The poor resolution is due, in part, to the sparse and fragmentary nature of the early ornithischian record. However, it also underscores the need for a thorough evaluation of the taxa, characters, character descriptions, and character codings used for evaluating ornithischian phylogeny. In the first iteration of the current matrix (see Materials and Methods), Spencer (2007) analyzed a data set of 17 ingroup taxa and 97 characters. The resulting phylogeny was well resolved (Fig. 1.6) and recovered several curious groupings. Heterodontosauridae was recovered as the sister taxon to Cerapoda and Eocursor was the basalmost neornithischian. A monophyletic Ornithopoda was not recovered; however, the taxon sampling was limited to a composite-coded Iguanodontia along with Hypsilophodon and Thescelosaurus. More curious still, a fabrosaurid clade (Lesothosaurus outside of Stormbergia + Agilisaurus) was recovered as the sister taxon to Thyreophora. As discussed below, the taxon and character sampling strategies employed by Spencer (2007) and Butler et al. (2008, and all other versions of that matrix that followed) may not be optimal given the available taxa and the methods employed for those analyses.

34 Figure 1.4 Ornithischian phylogeny of Butler (2005). Lesothosaurus and Eocursor (referred to in Butler [2005] as SAM-PK-8025) are recovered as the basalmost neornithischians and Heterodontosauridae is the sister taxon to Cerapoda. A monophyletic Ornithopoda was not recovered. 20

35 Figure 1.5 Consensus ornithischian phylogeny from Butler et al. (2007, 2008, 2010, 2011). Relatively few clades have been established among the analyses of various iterations of Butler et al. s data set. A taxonomically-limited Ornithopoda, Ceratopsia, and Thyreophora are the only major clades to be recovered. An Asian clade consisting of Haya + (Jeholosaurus + Changchunsaurus) was also recovered in Makovicky et al. (2011), a clade that does not contradict Butler et al. s (2011) inclusion of Changchunsaurus. NHMUK A100 was formerly referred to as BMNH A

36 Figure 1.6 Preliminary phylogenetic analysis of Ornithischia by Spencer (2007). Lesothosaurus (as the basal member of a fabrosaurid clade) was recovered as a basal thyreophoran and Heterodontosauridae was the sister taxon to Cerapoda. A limited sampling of ornithopods (Iguanodontia, Hypsilophodon, and Thescelosaurus) did not form a clade. The data set used in the current analysis is an extension of Spencer (2007). 22

37 23 Materials and Methods Taxon and Character Sampling Strategies There are more than 170 named genera of ornithischian dinosaurs, many of them monospecific (approximately 200 total species; Weishampel et al., 2004a). Many of these genera belong to well-supported derived clades such as Hadrosauroidea and Ceratopsia. However, approximately genera have been considered basal taxa based on either their assumed phylogenetic position given their original descriptions (e.g., Lesothosaurus diagnosticus) or their basal phylogenetic position within larger, yet still well supported, clades (e.g., Scutellosaurus lawleri as a basal thyreophoran, Yinlong downsi as a basal marginocephalian). Recently, taxonomic revisions have shrunk the number of basal taxa due in large part to either ambiguity in diagnosable characters (e.g., Bugenasaura infernalis; Boyd et al., 2009) or, more commonly for historically basal ornithischians, the establishment of so-called tooth taxa, those taxa that have been erected based solely on dental morphology that has been assumed to be an explicitly ornithischian character but was later demonstrated to be present in non-ornithischian archosaurs (e.g., Technosaurus smalli, Revueltosaurus callenderi [Parker et el., 2005; Irmis et al., 2007; Nesbitt et al., 2007]). Recent analyses by Butler and colleagues (Butler, 2005; Butler et al., 2008; 2010; 2011; Norman et al., 2011) as well as other studies that added to and complemented the taxa and characters of those studies (Zheng et al., 2009; Makovicky et al., 2011; Pol et al., 2011) have illuminated new hypotheses of relationships among these basal taxa.

38 24 To expand on this premise, taxa were selected for this analysis based on three primary criteria: 1) validity of the species (e.g., Lanasaurus, Stormbergia; see below); 2) availability of specimens for firsthand examination; and 3) a well-described and wellillustrated description of the species. The first two criteria were paramount in taxon selection. Whenever possible, a species was included if it was available for firsthand examination. With respect to the last criterion, for example, Makovicky et al. (2011) presented a recent, detailed description of Haya griva that included not only a wellexplained morphological description, but also a fairly well illustrated description with numerous photographs. For some older descriptions, (e.g., Agilisaurus louderbacki, Peng, 1992), newer redescriptions (e.g., Barrett et al., 2005) were used as supplements and/or surrogates if they provided more detailed information due to, e.g., further preparation since the time of original description, new material, more detailed photographic information. The taxa chosen for this analysis represent the overwhelming majority of species that currently are or historically have been considered basal ornithischians. The validity of one specimen, in particular, deserves some explanation because of its potential phylogenetic importance. The heterodontosaurid specimen NHMUK RU A100 has been the subject of much debate since its description (Thulborn, 1970b) and the debate has recently been reinvigorated by Norman et al (2011) and Sereno (2012). Thulborn (1970b) was the first to mention NHMUK RU A100 in the literature and referred it to the genus Lycorhinus. NHMUK RU A100 has been attributed to a distinct genus (Charig and Crompton, 1974; Butler et al., 2008, 2010, 2011), Lycorhinus (Thulborn, 1970b; Sereno, 2012), Abrictosaurus (Hopson, 1975, 1980; Weishampel and Witmer, 1990; Norman et al., 2004c), and Lanasaurus (Norman et al., 2011). Lanasaurus

39 25 has been synonymized with Lycorhinus by several authors (Hopson, 1980; Gow, 1990; Weishampel and Witmer, 1990; Norman et al., 2004c; Butler et al., 2008, 2010, 2011; Sereno, 2012); however, this is a dubious assignment because the type and referred specimens of Lanasaurus (BP/1/4244 and BP/1/5253, respectively) are both left maxillae and the type and only known specimen of Lycorhinus (SAM-PK-K3606) is a left dentary. Lycorhinus (SAM-PK-K3606; Fig. 1.7D, 1.7E) is considered a valid species based on a unique suite of characters: the diastema between the dentary caniniform and the first postcaniniform tooth is short, subequal to the width of the first postcaniniform tooth crown; tightly packed postcaniniform crowns that slightly overlap; possesses a mediolateral swelling at the base of the crown (a cingulum ); and crowns expand mesiodistally above the root forming a neck. Lanasaurus (BP/1/4244; Fig. 1.7B) is diagnosed by a strongly medially bowed maxillary tooth row and a prominent posterior ridge along the distal crown edge. The taxonomic status of both Lycorhinus and Lanasaurus likely depends on the status of NHMUK RU A100. NHMUK RU A100 consists of a partial disarticulated skull in a sandstone block. The cranial fragments include the lateral aspect of the right maxilla (Fig. 1.7A), and the medial aspect of the partial right premaxilla (not figured), the medial aspect of the left dentary (Fig. 1.7C), and a highly fragmented right dentary (not figured). As in Lanasaurus, the maxillary tooth crowns in NHMUK RU A100 possess a prominent distal ridge (Fig. 1.7A, 1.7B); however, it is unclear how much the maxilla in NHMUK RU A100 bows medially, as it does in Lanasaurus. Both Norman et al. (2011) and Sereno (2012) agree that the type and referred specimens of Lanasaurus (BP/1/4244 and BP/1/5253) can be tentatively linked to NHMUK RU A100 based on the prominent

40 Figure 1.7 Problematic South African heterodontosaurids. A) Right maxilla in lateral view of NHMUK A100 (formerly BMNH A100). B) Left maxilla in lateral view of Lanasaurus scalpridens (BP/1/4244). C) Left dentary in medial view of NHMUK A100. D) Silicon mold of left dentary in lateral view of Lycorhinus angustidens (SAM-PK- K3606). E) External mold and caniniform of left dentary (SAM-PK-K3606). The maxillary tooth crowns of NHMUK A100 (A) and BP/1/4244 (B) are enlarged to illustrate the prominent posterior ridge on the distal crown surfaces. The proportional size of the diastema between the dentary caniniform and the first postcaniniform tooth is highlighted for NHMUK A100 (C) and SAM-PK-K3606 (D, E). Scale bar = 1 cm. 26

41 27 posterior ridge on the distal surface of the crowns and possibly by the medial bowing of the maxilla. I agree with both assessments of the distal tooth crown ridges; however, it is less certain that the slight medial bowing in the right maxilla of NHMUK RU A100 is natural and not an artefact of preservation. Unfortunately, the type and only known specimen of Lycorhinus has been damaged since its initial description by Haughton (1924). The actual specimen (SAM- PK-K3606) is an external mold preserving only a partial dentary caniniform and an associated silicon mold that preserves very good detail of the labial aspect of the postcaniniform teeth. NHMUK RU A100, however, preserves only the medial aspect of the left dentary. Norman et al. (2011) suggest that, based on the abbreviated diastema in Lycorhinus and the lack of overlap in the dentary tooth crowns in NHMUK RU A100, the two are distinct from each other. Sereno (2012) suggests that NHMUK RU A100 is referable to Lycorhinus based on the medial bowing noted in the silicon mold. The silicon mold of the type of Lycorhinus is slightly bowed, but that appears to be a feature of the mold itself as the actual external mold is not noticeably bowed. Based on firsthand observations of the characters on each specimen (Fig. 1.7), I agree with Norman et al. (2011) and Sereno (2012) that SAM-PK-K3606 is a valid taxon (Lycorhinus). I also agree with Norman et al. (2011) and Sereno (2012) that BP/1/4244, BP/1/5253, and NHMUK RU A100 likely represent the same taxon. For this analysis, I follow Norman et al. (2011; contra Sereno, 2012) and tentatively assign NHMUK RU A100 to Lanasaurus scalpridens. Future preparation or CT work done on NHMUK RU A100 may help shed light on this controversial, yet phylogenetically important, South African heterodontosaurid.

42 28 Although Butler and colleagues (2005; Butler et al., 2008; 2010; 2011) have sampled nearly all recognized representative basal taxa including those taxa that are firmly nested within well-established clades (e.g., Scelidosaurus harrisonii in Thyreophora) there are some issues with their taxon assignments and character codings. For several characters, Butler et al. (2008, and in subsequent iterations of the same matrix) coded all three outgroups (Euparkeria, Marasuchus lilloensis, and Herrerasaurus) with derived characters states. For example, character 174 in Butler et al. (2008) the length of the postacetabular process as a percentage of the total length of the ilium was scored as 2 for Euparkeria, 2 for Marasuchus, and 1 for Herrerasaurus and is only scored as the plesiomorphic state (0) for three taxa: Scelidosaurus and in the supraspecific taxa Stegosauria and Ankylosauria. In this analysis, putative thyreophorans (i.e., Scelidosaurus harrisonii, Huayangosaurus taibaii, Dacentrurus armatus, Stegosaurus stenops, Euoplocephalus tutus) all maintain a distinct restriction of the postacetabular process relative to all other included taxa. However, the preacetabular process is greatly elongated relative to other ornithischians. Conversely, Euparkeria possesses an incipient postacetabular process but no preacetabular process that is nearly half the total length of the ilium, similar to derived cerapodans such as ceratopsians. Furthermore, Butler et al. (2008, 2010, 2011) use several supraspecific taxa in their phylogenetic analyses: Stegosauria, Ankylosauria, Rhabdodontidae, Dryosauridae, Ankylopollexia, Pachycephalosauridae, Psittacosauridae, and an unnamed taxon consisting of two supraspecific taxa, Coronosauria and Leptoceratopsidae (the previous version of the current data set [Spencer, 2007] also utilized several supraspecific taxa).

43 29 The issue of assumed monophyly is typically the major objection to using supraspecific taxa (e.g., Yeates, 1995); however, as noted above, there have been myriad phylogenetic analyses that have supported monophyly among these taxa (relationships of taxa within those clades notwithstanding). More germane to the current study is the issue of phylogenetic accuracy due to interspecific variation among the characters (Wiens, 1998). Supraspecific taxa are more frequently associated with polymorphic character states (Rice et al., 1997), and thus, can decrease phylogenetic accuracy (Wiens, 1998). Indeed, there are 17 polymorphisms among the supraspecific taxa in Butler et al. (2008), 36 in Butler et al. (2010), and 38 in Butler et al. (2011). To avoid the issues of assumed monophyly or decreased phylogenetic accuracy, the obvious alternative to supraspecific taxa is to sample all known species and analyze them simultaneously. This, of course, can be logistically problematic due to numerous reasons including the availability of a species, either because of a lack of firsthand access to specimens, or because the species lacks a formal description or is inadequately described in the literature. Prendini (2001) lists several criteria for selecting exemplar taxa in lieu of using supraspecific taxa (see also Brusatte, 2010), including maximizing the amount of interspecific variation among the species within that taxon and using multiple species, particularly those that are basal within that supraspecific taxon. I have followed Prendini s (2001) criteria for selecting species that represent these supraspecific taxa. With respect to the derived thyreophoran clade, Stegosauria, I chose two wellknown basal taxa, Huayangosaurus and Dacentrurus (e.g., Maidment et al., 2008) and the derived member, Stegosaurus, representing the widest variability among stegosaurs. Unfortunately, I was unable to confidently gather enough information on multiple

44 30 ankylosaurs to include more than just the derived taxon, Euoplocephalus. For derived cerapodans Pachycephalosauria and Ceratopsia, I chose taxa that have been traditionally and independently recovered as basal members. Additionally, these taxa also represent a wide variability among their individual morphologies and, thus, at least partially fulfill Prendini s (2001) criteria for exemplar taxon sampling. A data matrix consisting of 54 species (two outgroup and 52 ingroup taxa) and 247 discrete morphological characters was compiled in Mesquite (Maddison and Maddison, 2011). The data set is largely modified from previous unpublished work (Spencer, 2007) and from Butler et al. (2008). New characters, character descriptions, and codings have been added, updated, and/or modified based on personal observations from firsthand examination of specimens and new species descriptions (e.g., Fruitadens haagarorum, Butler et al., 2010; Haya griva, Makovicky et al., 2011; Manidens condorensis, Pol et al., 2011; Pegomastax africanus, Sereno, 2012). The basal archosauriform Euparkeria capensis was chosen as an outgroup taxon for purposes of character polarity. The basal saurischian dinosaur Herrerasaurus ischigualastensis was also chosen as an outgroup taxon for character polarity (for some characters that were not present in Euparkeria) and to test ingroup monophyly of Ornithischia. Taxa included in this analysis are listed in Appendix B along with the sources of morphological information (institutional abbreviations used in Appendix B are denoted in Appendix A). Character state descriptions are included in Appendix C and the taxon-character matrix for the phylogenetic analysis (see below) is included in Appendix D

45 31 Phylogenetic Analysis Although there are other methods available for phylogenetic analyses namely, model-based approaches such as maximum likelihood and Bayesian inference maximum parsimony is used here because it is most appropriate for discrete morphological characters (e.g., Siddall and Kluge, 1997; Rindal and Brower, 2011). Model-based approaches rely explicitly on a priori assumptions of character evolution, assumptions that may be unrealistic for morphological characters (Spencer and Wilberg, 2013). Therefore, a maximum parsimony analysis on the abovementioned data matrix was performed in TNT v. 1.1 (Goloboff et al., 2008). All characters were treated as equally weighted and unordered. As noted above, Euparkeria capensis was designated the outgroup taxon. Because TNT only allows one outgroup designation, Herrerasaurus ischigualastensis was technically considered part of the ingroup during the analysis. Most parsimonious trees (MPTs) were found with a heuristic search ( traditional search in TNT) using 1,000 random addition sequences with tree bisection and reconnection (TBR) branch swapping, holding ten trees per replicate. All zero-length branches were collapsed if they were not supported in any of the MPTs ( rule 1 according to Coddington and Scharff [1994]). More than ten trees were found in some replicates, so to make sure that the initial analysis found as many MPTs as possible, another round of searches on that set of trees using random addition sequences with TBR was carried out. To assess node stability within the tree, nonparametric bootstrap proportions (Felsenstein, 1985) and decay indices (Bremer 1988, 1994) were calculated. Bootstrap proportions (5,000 pseudoreplicates) were assessed in TNT using GC frequencies (the default setting in TNT) rather than absolute frequencies. GC frequencies take into

46 32 account not only the amount of support in favor of a group, but also the evidence against a group by measuring the difference between the amount of times a given group is recovered and the amount of times it is contradicted (Goloboff et al., 2003). Decay indices were determined in TNT on all suboptimal trees five steps longer than the MPTs (a total of 9,076 trees found with a heuristic search using 1,000 random addition sequences with TBR, holding ten trees per replicate). Results The phylogenetic analysis resulted in 990 MPTs. A strict consensus tree of all MPTs was constructed to summarize the component relationships within Ornithischia (Fig. 1.8). Each MPT had a length (TL) of 606 steps. The average consistency index (CI), average retention index (RI), and average rescaled consistency index for all characters in the data set were calculated to determine the amount and extent of homoplasy within the matrix. The CI and RI were and 0.721, respectively, with a RC of The resulting phylogeny based on the strict consensus tree is largely well resolved with the exception at the base of Cerapoda. Ornithischia is a robustly supported monophyletic group, with Pisanosaurus as the basalmost ornithischian dinosaur. Heterodontosauridae is here recovered as a clade at the base of Ornithischia, though the interrelationships of heterodontosaurids are less well resolved. A clade comprised of strictly South African heterodontosaurids is also present, with a grouping of Heterodontosaurus + Pegomastax, along with a polytomy including Abrictosaurus,

47 Figure 1.8 Strict consensus tree of all 990 MPTs resulting from the phylogenetic analysis of 54 taxa and 247 characters. The tree statistics include: TL = 606, CI = 0.452; RI = 0.721; RC = The numbers at each node represent Bremer Decay Indices/Bootstrap Proportion Values. 33

48 34 Lycorhinus, and Lanasaurus (including NHMUK A100). Manidens, Tianyulong, Fruitadens, and Echinodon form a polytomy at the base of Heterodontosauridae. Eocursor is the sister taxon of Genasauria, which is further divided into Thyreophora and Neornithischia. Thyreophora is a well-resolved and well-supported clade consisting of Ankylosauria (here represented only by Euoplocephalus) and Stegosauria (represented by Huayangosaurus + (Dacentrurus + Stegosaurus)) as well as successively more distantly related taxa Scelidosaurus, Emausaurus, and Scutellosaurus at the base. Lesothosaurus is the basalmost neornithischian, followed by Stormbergia, Agilisaurus, Hexinlusaurus, and Cerapoda. The relationships within Cerapoda are poorly resolved. Other than a monophyletic Marginocephalia and Iguanodontia as well as a clade comprised of Gasparinisaura + Parksosaurus, the rest of the taxa collapsed into a polytomy at the base of Cerapoda. The current analysis did not recover a monophyletic Ornithopoda. To determine if there was any topological structure in the resulting phylogeny that may be obscured by taxa that have conflicting phylogenetic positions among the MPTs, I used TNT to calculate the combinable components, or semi-strict, consensus tree (Nelson, 1979; Bremer, 1990) and the maximum agreement subtree (MAS). The resulting topology of the combinable components tree (not shown) was similar to the strict consensus tree (Fig. 1.8) with two exceptions: Lanasaurus forms a clade with Heterodontosaurus + Pegomastax and Chaoyangsaurus is the sister taxon to all other ceratopsians. The rest of the topology was identical to the strict consensus tree. Unlike most other consensus methods that contain the same taxa as the original trees, the MAS details the greatest amount of terminal taxa and components (clades)

49 35 common among two or more input trees by pruning labile taxa ( common pruned trees according to Finden and Gordon [1985]). In the present study, the subtree analysis resulted in the pruning of 11 taxa: three heterodontosaurids Fruitadens, Echinodon, Lycorhinus; one thyreophoran Stegosaurus; five basal cerapodans Yandusaurus, Thescelosaurus, Orodromeus, Anabisetia, Othnielosaurus; and two marginocephalians Stenopelix, Micropachycephalosaurus. The MAS (Fig. 1.9) is fully resolved and illustrates a pectinate Heterodontosauridae, with Manidens and Tianyulong as successively more distantly related sister taxa to a clade of South African species. Cerapoda is divided into Marginocephalia and a monophyletic Ornithopoda. The Asian taxon Jeholosaurus is the basalmost ornithopod, followed by successively more derived taxa including the Asian taxon Haya, the European Hypsilophodon, and the North American Zephyrosaurus. Recent attempts at reconstructing the evolutionary history of Ornithischia have resulted in phylogenies that are largely unresolved and poorly supported (Butler et al., 2008, 2010, 2011; Makovicky et al., 2011; though, the strict consensus of Pol et al., 2011 is also fairly well resolved). With the exception of the basal polytomy for Cerapoda, the phylogeny here is fairly well resolved with respect to most basal taxa. A detailed list of unambiguous apomorphies for each node is located in Appendix E; however, there are some salient characters for well-established clades that are detailed below.

50 Figure 1.9 Maximum agreement subtree of all 990 MPTs found during the phylogenetic analysis. Eleven taxa were pruned to produce a fully resolved topology: Fruitadens, Echinodon, Lycorhinus, Stegosaurus, Yandusaurus, Thescelosaurus, Orodromeus, Anabisetia, Othnielosaurus, Stenopelix, and Micropachycephalosaurus. 36

51 37 Discussion When compared to the competing phylogenetic hypotheses for the most problematic taxa (Fig. 1.10), the phylogeny presented here illuminates two of them. Heterodontosaurids are basal ornithischians, sister taxon to Genasauria, possessing several plesiomorphic features including long forelimbs and large hands. The heterodontosaurid skull, though, is highly derived with a unique dentition and specialized masticatory apparatus (e.g., Sereno, 2012). Despite its generalized ornithischian body plan, Lesothosaurus is firmly nested within Genasauria, here recovered as the basalmost neornithischian. It shares with other basal neornithischians a reduced forelimb, as seen in Agilisaurus (Peng, 1992, 1997). Despite the poor resolution at the base of Cerapoda, a monophyletic Ornithopoda cannot be ruled out as a possibility. Much of the issue regarding ornithopod monophyly is subjective based on the phylogenetic definition. Here, I adopt the definition proposed by Butler et al. (2008) that Ornithopoda is a stem-based clade defined as all genasaurians more closely related to Parasaurolophus walkeri Parks, 1922 than to Triceratops horridus Marsh A stem-based phylogenetic definition provides more stability given that it is nested within a node-based clade (node-stem triplets, sensu Sereno, 1998) Cerapoda is a node-based taxon defined as Parasaurolophus walkeri Parks, 1922, Triceratops horridus Marsh 1889, their most recent common ancestor and all its descendants (Butler et al., 2008). However, the definition as applied here would render Ornithopoda equivalent to Iguanodontia. More work is needed to parse out the

52 38 relationships among basal neornithischians and cerapodans to address the monophyly of Ornithopoda. Clade Support Ornithischia Monophyly of Ornithischia has rarely been in question since it was formally erected by Seeley (1887). The inclusion of Pisanosaurus in phylogenetic analyses of ornithischians has been limited; however, when it has been included, it has always been recovered as a basal species (Langer, 2004; Butler, 2005; Langer and Benton, 2006; Butler et al., 2008; 2010; 2011; Makovicky et al., 2011; Pol et al., 2011). Though, beginning with Bonaparte s (1976) redescription of Pisanosaurus, it has been suggested by several authors that it may belong to Heterodontosauridae (e.g., Weishampel and Witmer 1990b; Sereno, 2012). And, although Pisanosaurus is here recovered as the basalmost known ornithischian, it only takes two additional steps to make it sister to all other heterodontosaurids. It does, however, take an additional eight steps to align Pisanosaurus with Gondwanan heterodontosaurids. Ornithischia is a well-supported monophyletic group and is here diagnosed by eight unambiguous cranial and dental synapomorphies: buccal emargination along the maxilla (27.1); anterodorsal margin of the coronoid process is formed by the posterodorsal margin of the dentary (113.1); subtriangular maxillary and dentary teeth (130.1); maxillary and dentary teeth lack recurvature (144.1); maxillary and dentary teeth are closely packed or even overlapping (145.1); maxillary and dentary cheek tooth crowns are mesiodistally expanded above the root (146.1); maxillary and dentary teeth achieve maximum apicobasal height at the middle of the tooth row (148.1); and the

53 Figure 1.10 Competing phylogenetic hypotheses for the placement of A) Heterodontosauridae, B) Lesothosaurus, and C) Ornithopoda. A.BO, Heterodontosaurids as basal ornithischians; A.Ne, Heterodontosaurids as basal neornithischians; A.Orn, as basal ornithopods, A.Mar, Heterodontosaurids as sister to Marginocephalia. B.BO, Lesothosaurus as a basal ornithischian; B.T, Lesothosaurus as a basal thyreophoran; B.Ne, Lesothosaurus as a basal neornithischian. C.M, Monophyletic Ornithopoda; C.P, Paraphyletic Ornithopoda. 39

54 40 apicobasal height of the mid-dentary and maxillary tooth crowns is less than 1.5 times the mesiodistal width. Heterodontosauridae Heterodontosauridae has been at the center of recent debate regarding its phylogenetic position (e.g., Norman et al., 2004c; Butler, 2005; Xu et al., 2006; Irmis et al., 2007; Butler et al., 2008). Here, its position as a basal clade of ornithischians is consistent with the most recent analyses (e.g., Butler et al., 2008; Norman et al., 2011; Sereno, 2012). To make Heterodontosauridae a basal ornithopod group (similar to, e.g., Sereno, 1986, 1997, 1999; Norman et al., 2004c), it takes an additional 14 steps. To align heterodontosaurids with marginocephalians (e.g., Xu et al., 2006), it would require an additional ten steps. And, to shift heterodontosaurids to the base of Cerapoda (e.g., Butler, 2005), it requires an additional nine steps. In Sereno s (2012) recent description of Pegomastax and analysis of heterodontosaurid relationships, he reported two distinct geographic clades an unresolved Laurasian clade consisting of Echinodon, Fruitadens, and Tianyulong, and a Gondwanan clade consisting of Lycorhinus, Pegomastax, Manidens, Abrictosaurus, and Heterodontosaurus. A Gondwanan clade was recovered here (what Sereno [2012] termed Heterodontosaurinae); however, the Laurasian taxa are unresolved in a polytomy at the base of Heterodontosauridae. In the MAS (Fig. 1.9), the taxonomically-limited Laurasian taxa form a paraphyletic grade to the Gondwanan clade. It only takes one additional step, though, to place the Laurasian heterodontosaurids in an unresolved sister clade to the Gondwanan species. Based on the data set used here, it is evident that Heterodontosauridae is a wellsupported basal ornithischian clade, outside of Genasauria. Heterodontosauridae is here

55 41 diagnosed by two unambiguous synapomorphies: maximum dorsoventral depth of the dentary is at least 25% of its length (109.1); and maxillary and dentary teeth have a ridge (ectoloph) separating the cingulum from the rest of the crown surface. Genasauria When Sereno (1986) first coined Genasauria, one of the most salient features of the group was the presence of buccal emargination along the maxilla ( cheeks ); however, as demonstrated above, that appears to be a character present, at least incipiently, in all ornithischians. Sereno s (1986) argument was advanced by the perception that Lesothosaurus was the basalmost ornithischian (with uncertainty about Pisanosaurus). It takes an additional nine steps to place Lesothosaurus as a basal ornithischian (in a slightly more derived position than Pisanosaurus). Genasauria is here diagnosed by two unambiguous synapomorphies: the penultimate phalanx of manual digits 2 and 3 is shorter than the first phalanx (179.0); and the lack of well-developed extensor pits on the dorsal surfaces of the metacarpals and manual phalanges (182.0). Thyreophora Thyreophora has also been a well-supported clade since the first cladistic analyses began evaluating ornithischian relationships (e.g., Norman, 1984; Sereno, 1984, 1986). In several recent analyses (e.g., Spencer, 2007; Butler et al., 2008), Lesothosaurus has been recovered as the sister to Thyreophora. In the current analysis, it only takes one additional step to shift the phylogenetic position of Lesothosaurus to be more closely related to Thyreophora. Thyreophora is here diagnosed by four unambiguous synapomorphies: triangular skull shape in dorsal view (37:1); cortical remodeling of skull dermal bone (96:1); parasagittal row of dermal osteoderms along the body (245:1); and lateral row of keeled osteoderms along the body (246:1).

56 42 Neornithischia Neornithischia was erected by Cooper (1985), but was rarely used in ornithischian systematics, with authors preferring Cerapoda Sereno, 1986 because the taxonomic makeup of the group (Ornithopoda + Marginocephalia) was more consistent with prevailing phylogenetic hypotheses (e.g., Weishampel, 1990; Norman et al., 2004c). In recent analyses (e.g., Butler et al., 2008), Neornithischia has been reinstated as a more inclusive clade than Cerapoda. I follow the phylogenetic definition of Butler et al. (2008) in that Neornithischia Cooper, 1985 is a stem-based clade defined as all genasaurians more closely related to Parasaurolophus walkeri Parks, 1922 than to Ankylosaurus magniventris Brown, 1908 or Stegosaurus stenops Marsh, Neornithischia is diagnosed by one unambiguous synapomorphy: the posttemporal foramen is located within the paraoccipital process, very near to its dorsal margin (84.1). As noted above, in previous iterations of this matrix, Lesothosaurus has been recovered as the sister taxon to Thyreophora. Here, Lesothosaurus is recovered as the basalmost neornithischian, sister to a clade comprised of Stormbergia + all other neornithischians. Knoll et al. (2010) suggested that, based on histological analyses, Butler (2005), in his description of Stormbergia, mistakenly assigned subadult or adult specimens of Lesothosaurus to the new taxon, Stormbergia. Butler (2005) indicated that the absence of a tab-shaped obturator process on the ischia of Lesothosaurus as well as a long, medial symphysis of the ischia without a shallow, dorsal groove are what separate it from Stormbergia. However, Some of the features noted by Butler (2005) and eschewed by Knoll et al. (2010) are present elsewhere in basal ornithischians: tab-shaped obturator process in, e.g., Tianyulong, Eocursor (Zheng et al., 2009; Butler, 2010; respectively); shallow, dorsal groove in, e.g., Tianyulong (Zheng et al., 2009); and the presence of an

57 43 elongate medial ischial symphysis in primitive and non-dinosaurian archosaurs (Hutchinson, 2001). Given the phylogenetic placement here of Lesothosaurus with respect to Stormbergia, the taxonomic validity of Stormbergia should be reassessed; however, in light of the presence of these features in other basal taxa, Stormbergia is tentatively considered a valid taxon. Cerapoda Cerapoda, consisting of Ornithopoda and Marginocephalia, has been a well-supported group since Sereno (1986) first recognized the group. Though, some of the clade membership has changed since that time (e.g., Heterodontosauridae is no longer accepted as a cerapodan). Cerapodans are diagnosed by seven unambiguous synapomorphies: six or more sacral vertebrae (157.3); and accessory articulation between the pubis and the sacrum (158.1); a medioventral acetabular flange is no longer present (196.1); the presence of a strongly developed supraacetabular flange (197.1); the femoral head is separated from the greater trochanter by a marked constriction (219.1); a reduced anterior ( lesser ) trochanter (220.3); and anterior trochanter is located proximally, near the femoral head (221.1). Ornithopoda Ornithopoda was not recovered here as a monophyletic group, at least in the strict consensus of all MPTs (Fig. 1.8). The majority of taxa that have been considered ornithopods (e.g., Hypsilophodon, Othnielosaurus, Iguanodontia) collapsed into a polytomy at the base of Cerapoda. However, after pruning Yandusaurus, Thescelosaurus, Orodromeus, Anabisetia, and Othnielosaurus (the MAS, Fig. 1.9), Ornithopoda is recovered as a clade. The MAS does not parse out a basal clade of Asian ornithopods (i.e., Jeholosaurus and Haya) as reported by Butler et al. (2011) and Makovicky et al. (2011). However, one taxon that was included in Butler et al. (2011)

58 44 and Makovicky et al. (2011), Changchunsaurus parvus, was not included in this analysis. In Butler et al. (2011), Changchunsaurus and Jeholosaurus formed a clade at the base of Ornithopoda (Haya had not yet been described) and in Makovicky et al. (2011) Haya, Changchunsaurus, and Jeholosaurus formed a clade at the base of Ornithopoda. Makovicky et al. (2011) coded Haya into the matrix of Butler et al. (2011) and adjusted the taxonomic status of two taxa (following Boyd et al. [2009], Bugenasaura was merged with Thescelosaurus). It should be noted, though, that in Butler et al. (2011) the Asian clade was recovered only in the MAS, not in the strict consensus. The strict consensus trees in both Butler et al. (2011) and Makovicky et al. (2011) were very poorly resolved (as in Fig. 1.5); however, Makovicky et al. (2011) did recover an Asian clade in both the MAS and the strict consensus tree. With the current topology, it only requires one additional step to unite Jeholosaurus and Haya. Inclusion of Changchunsaurus in the data set here may help resolve some of the basal relationships of Cerapoda. Iguanodontia There has been an increase recently in the number of analyses focusing on basal and derived relationships among iguanodontians (e.g., McDonald et al., 2010; Prieto-Márquez, 2010). The iguanodontian taxa included in this analysis focused solely on the basalmost known species from numerous previous analyses (e.g., Weishampel and Heinrich, 1992; Weishampel et al., 2003; Norman, 2004; corroborated most recently in McDonald et al., 2010). Although Iguanodontia was recovered as a clade here, there were no unambiguous synapomorphies that supported its monophyly. Marginocephalia Of the recent published phylogenetic analyses that have focused on basal ornithischian relationships (Butler et al., 2008, 2010, 2011), none has recovered a monophyletic Marginocephalia. Like Iguanodontia, Marginocephalia was

59 45 here recovered as a monophyletic group but there were no unambiguous synapomorphies that supported it. However, one of the features that has historically diagnosed the clade, a parietosquamosal shelf that obscures the occiput in dorsal view (75.1, as per Sereno, 1986), is ambiguously optimized for the clade because the basal taxa Micropachycephalosaurus, Chaoyangsaurus, Stenopelix, and Wannanosaurus do not preserve the character. Ornithischian Phylogeny Previous iterations of the current data set (Spencer, 2010, 2012) have suggested similar relationships as those demonstrated here. The taxon sampling varied somewhat across all three analyses (e.g., the 2010 analysis was more limited in scope); however, there has been an increase in character sampling in each successive analysis. Spencer (2010) included 42 ingroup taxa and 222 characters. Spencer (2012) was largely similar to the current analysis, including 54 ingroup taxa and 227 characters. The 42 ingroup taxa from the original analysis are all included in the current analysis. The resulting phylogeny was poorly resolved and weakly supported. All heterodontosaurids collapsed into a polytomy at the base of Ornithischia along with Pisanosaurus and Eocursor. Lesothosaurus was recovered as the basalmost thyreophoran and a taxonomically-limited Ornithopoda was recovered (Anabisteia + Iguanodontia; all other putative ornithopods e.g., Othnielosaurus, Thescelosaurus, Hypsilophodon collapsed into a polytomy at the base of Cerapoda). Moreover, there were no unambiguous synapomorphies that united Ornithopoda and none that supported the placement of Lesothosaurus as a basal thyreophoran.

60 46 Spencer (2012) had a slightly different taxonomic makeup than the current analysis. Several putative basal ornithopods (e.g., Talenkauen, Anabisetia) and basal marginocephalians (e.g., Yinlong, Wannanosaurus) were excluded from the analysis and there was a substantial increase in the number of derived iguanodontians (e.g., Iguanodon, Barilium, Corythosaurus, Bactrosaurus). Unfortunately, there was no substantial increase in other derived taxa such as stegosaurs, ankylosaurs, and marginocephalians. As in the previous analysis, the phylogeny was poorly resolved and weakly supported. Heterodontosaurids, again along with Pisanosaurus and Eocursor, collapsed into a polytomy at the base of Ornithischia. A monophyletic Ornithopoda was not recovered, though all included iguanodontians formed a clade. Most of the character support for the recovered clades was ambiguous, though there was one unambiguous synapomorphy that supported Lesothosaurus as a basal thyreophoran the dorsal groove on the ischium (character in this analysis, shared with Scutellosaurus). There are several consistencies across the two previous preliminary analyses and the current phylogeny. In each of the two previous phylogenetic analyses, all heterodontosaurids fell out as basal ornithischians. Lesothosaurus is robustly nested within Genasauria, although its within-clade relationships are less certain. The base of Cerapoda is very poorly resolved. Though, MASs in the previous two analyses pulled Othnielosaurus out of Cerapoda and it was one of the pruned taxa in this analysis. An undescribed, nearly complete articulated skeleton of Othielosaurus from the Morrison Formation (Kimmeridgian-Tithonian) of Wyoming, USA, is currently on display in the Sauriermuseum Aathal, Switzerland (specimen number unknown; P. Galton pers. comm.). Given the lack of phylogenetic resolution in the position of

61 47 Othnielosaurus (and other basal neornithischians and cerapodans), this specimen could provide a firmer understanding of the morphology at the base of Neornithischia. Conclusion The overall topology recovered here reveals the lability of certain problematic taxa and represents one of the most comprehensive species-level phylogenies of basal ornithischians to date. It helps build on and complement the few other analyses of basal ornithischian relationships that have been published (Butler et al., 2008, 2010, 2011). This analysis supports the hypothesis that heterodontosaurids are basal ornithischians with a mixture of plesiomorphic and derived traits. With the increasing amount of quantitative analyses sampling basal ornithischians, it is becoming more evident that Lesothosaurus is a genasaurian; however, its position within Genasauria remains unclear. The monophyly of Ornithopoda is more problematic. Without consensus of a formal phylogenetic definition for the clade based on stable taxa, its monophyly will likely remain tenuous. The phylogeny presented here provides a framework upon which numerous questions can be asked and addressed regarding character evolution and biogeographic scenarios. More broadly, the coincident rise in ornithischian diversity in the middle Cretaceous with flowering plants offers an interesting test case in coevolutionary hypotheses. But most importantly, ornithischians achieved a global distribution during a time of fundamental changes to global continental configuration and climate.

62 48 CHAPTER II THE MORPHOLOGY AND ANCESTRAL BODY PLAN OF ORNITHISCHIA Abstract Ornithischians first appear in the Late Triassic, but their early fossil record is sparse. The earliest known forms are relatively small, bipedal, and possess morphological features that are associated with herbivory. Furthermore, the phylogenetic position of many of these early ornithischians remains unresolved. Previous work has focused primarily on cranial anatomy because ornithischians demonstrate elaborate solutions for herbivory (e.g., tooth batteries) and possess extensive cranial ornamentation (e.g., crests). Little has been done, however, to evaluate the general body plan (cranial and postcranial) of ornithischians. Several of the earliest known ornithischians (e.g., Lesothosaurus, Stormbergia, Heterodontosaurus) were examined and evaluated with respect to more derived forms (e.g., thyreophorans, ornithopods, marginocephalians). To obtain a generalized cranial and postcranial body plan for the ancestral ornithischian, though, these taxa must be placed in a phylogenetic context to polarize the characters at the base of the tree. A maximum parsimony analysis was carried out to reconstruct ornithischian phylogeny. Taxon sampling focused on basal members, though members of derived clades were represented where applicable. The resultant phylogeny is fairly well resolved among basal members, with the exception of basal cerapodans and putative ornithopods, but several important patterns at the base of Ornithischia are evident. These phylogenetic

63 49 patterns help polarize character optimizations that allow inferences regarding the ancestral body plan of Ornithischia. These results suggest that the postcranial morphology of these taxa likely reflects the ancestral ornithischian body plan. The ancestral ornithischian had a generalized Lesothosaurus-like skull and was bipedal with ossified tendons along the sacral vertebrae, likely extending anteriorly to at least the posteriormost dorsals. The forelimb was long relative to the hindlimb (more than 50% of the hindlimb length) and possessed a large manus with deep, well-developed extensor pits for a grasping hand, more similar to that of early saurischians. Other, more derived bipedal ornithischians possess a relatively short forelimb, typically far less than half the length of the hindlimb. Basal saurischians (e.g., Herrerasaurus, Eoraptor) from the Late Triassic approximate this condition and these results suggest that early dinosaurs were more similar than previously thought. The analysis presented here provides a window into the evolution of dinosaurs in the Late Triassic and Early Jurassic. Introduction Ornithischian dinosaurs, first appearing in the Middle Triassic (based on ghost lineages; e.g., Nesbitt et al., 2010), were the dominant large herbivores in many Mesozoic faunas, particularly of Laurasia during the Cretaceous (Weishampel et al., 2004b). They are an ideal group for large-scale analyses of biodiversity, biogeography, and coevolution. Ornithischians are central to studies of the evolution of herbivory, life history, physiology, and sexual dimorphism in the fossil record.

64 50 Ornithischians possess several unique skeletal specializations, most of which are related to modifications for herbivory. Previous studies that analyzed the putative diet of ornithischians have focused primarily on the skull and dentition (e.g., Barrett, 2000). Many cranial features of ornithischians lend themselves to rigorous analyses of function such as potential hinge points in the skull (e.g., between the maxilla and lacrimal in ornithopods; e.g., Norman and Weishampel, 1985) and occlusal patterns on the teeth (e.g., wear facets on the teeth of heterodontosaurids; e.g., Sereno, 2012). These characters all suggest an evolutionary trend toward more complex mechanisms for processing plant material. In fact, it had been suggested that the coincident rise in angiosperm plants and complex dental batteries among iguanodontians and ceratopsians in the Early Cretaceous was correlated (e.g., Wing and Tiffney, 1987). However, several recent studies suggest that this correlation may not be statistically corroborated (e.g., Weishampel and Jianu, 2000; Barrett and Willis, 2001), though sampling biases and inconsistencies inherent in the fossil record may be responsible for the lack of correlation (Butler et al., 2011). Characters in the postcranial skeleton can also shed light on the diet, as well as the habit, of ornithischians. For example, the presence of an opisthopubic pelvis has been long-thought to accommodate a larger gut tract to facilitate the breakdown of plant material (e.g., Rasskin-Gutman, 1997). The fore- and hindlimbs, though, have received little attention with respect to inferences about the ancestral condition for ornithischian dinosaurs, particularly in light of new phylogenetic hypotheses of Ornithischia. Here, I present an updated ornithischian phylogeny and use it to estimate the ancestral bauplan for Ornithischia.

65 51 Morphological History of Basal Ornithischia Ornithischian dinosaurs have been known to paleontology since Mantell (1825) first described the jaw and tooth of a large iguana-like reptile (Iguanodon). Approximately 200 currently valid ornithischian species are now known (Weishampel et al., 2004a) and many have been assigned to large subclades within Ornithischia such as Thyreophora, Ornithopoda, and Ceratopsia. These are groups that have largely been defined by generalized or ambiguous features such as armored (e.g., Thyreophora; Nopcsa, 1915) or bipedal (e.g., Ornithopoda; Romer, 1956) and represent the precladistic consensus of ornithischian taxonomy (Thulborn, 1971a). Thulborn (1971a) first suggested ornithischian monophyly based on a derivation from a pseudosuchian ancestral stock (including Euparkeria as one of the closest relatives of ornithischians). The origins of Ornithischia stemmed from what he termed a hypsilophodont plexus of early lycorhinids (Lycorhinus, Geranosaurus, and Heterodontosaurus; though, he considered Heterodontosaurus to be a junior synonym of Lycorhinus) and fabrosaurids (Fabrosaurus, Pisanosaurus). The large caniniform of the lycorhinids as well as the recessed maxilla (=buccal emargination) set them apart from the less conspicuous fabrosaurids because fabrosaurids lacked caniniforms, premaxillary teeth that extend to the tip of the snout, and lack buccal emargination. These fabrosaurid features have since been reevaluated and shown to be ambiguous given the preservation of Pisanosaurus and the taxonomic status of Fabrosaurus (e.g., Sereno, 1986, 1991; Norman et al., 2004a). Thulborn (1971a) suggested that a reticulate radiation of ornithischians crossed the evolutionary hypsilophodont/iguanodont boundary based on peculiar morphologies

66 52 exhibited by, e.g., Thescelosaurus, in that it possesses premaxillary teeth (as hypsilophodonts do), but its femur is longer than its tibia (as in iguanodonts ). His classification of ornithischians was ambivalent toward the origin of both stegosaurs and ankylosaurs and suggested that the lycorhinids were an ephemeral group that lasted only for a brief period of time in the early history of Ornithischia. Thulborn (1971a) characterized several other groups (e.g., ceratopsians, pachycephalosaurs, hadrosaurs), but made it a point that all these other ornithischians were derived from a hypsilophodont stock, likely from taxa similar to Pisanosaurus and Fabrosaurus. Sereno (1984) and Norman (1984) independently noted ornithischian synapomorphies in their attempts to reconstruct ornithischian phylogeny. Sereno s (1984) apomorphy list featured four dental characters, six cranial characters, and five postcranial characters, whereas Norman s (1984) apomorphy list included three cranial characters, eight postcranial characters, and no dental characters. Both lists, though, did agree on three characters: a predentary, opisthopubic pelvis, and pedal digit V reduced to only the metatarsal. Sereno (1984) also suggested that prosauropods were closely related to ornithischians than they were to saurischians because they share with Ornithischia a transversely-flattened and denticulate tooth crown, reduction in the recurvature of the premaxillary and maxillary teeth, a medially-offset anterior tooth crown, and lengthened vomer. Many dental features, while potentially useful in diagnosing clades accompanied by other cranial/postcranial characters, tend to be problematic when used as the primary evidence for phylogenetic relatedness, primarily due to convergent ecological habit (e.g., Parker et al., 2005; Irmis et al., 2007).

67 53 Table 2.1 Traditional ornithischian synapomorphies defined by Sereno (1986). Dental a Low, triangular-shaped tooth crowns in lateral view Recurvature absent in maxillary and dentary teeth a Well-developed neck separating crown from root a Overlap of adjacent crowns in maxillary and dentary teeth Maximum tooth size attained near the central or posterocentral portion of the tooth row Cranial b Ossified palpebral a Anterior tip of the premaxilla roughened and edentulous Horizontal or broadly arched premaxillary palate a Strong premaxillary posterodorsal process on the lateral aspect of the face, excluding the maxilla from the margin of the external nares Anterior process of the quadratojugal relatively short; straplike shape Anterior ramus of the quadratojugal broader than dorsal ramus Ventral margin of the antorbital fossa parallels the maxillary tooth row Antorbital fossa relatively smaller; separated from the maxilla-nasal suture Antorbital fenestra relatively smaller Prefrontal with long posterior ramus overlapping the dorsal surface of the frontal abc Predentary a Anterior coronoid margin formed by the posterodorsal process of the dentary Postcranial Five or more sacral vertebrae Preacetabular process of the ilium elongate and dorsoventrally narrow abc Opisthopubic pelvis a Pubic shaft elongate and rodlike a Pubic symphysis restricted to the distal end Distal puboischial symphysis Proximal puboischial plate relatively narrower dorsoventrally Obturator notch between the pubis and ischium, rather than a foramen entirely encased within the pubis Transversely-flattened prepubic process a Ischial symphysis restricted to the distal end

68 54 Table 2.1 Continued Distally expanding, ventromedially angling, bladelike ischial shaft Flat lateral surface of the greater trochanter broader anteroposteriorly bc Pendent fourth trochanter ab Pedal digit V reduced to just metatarsal b Ossified epaxial tendons Source: Sereno, P. C Phylogeny of the bird-hipped dinosaurs (Order Ornithischia). Natl. Geogr. Res. 2: a Sereno, P. C The phylogeny of the Ornithischia: a reappraisal, p In Reif, W. E., and Westphal, F. (eds.), Third Symposium on Mesozoic Terrestrial Ecosystems, Short Papers. Attempto Verlag, Tübingen. b Norman, D. B A systematic reappraisal of the reptile order Ornithischia, p In Reif, W. E., and Westphal, F. (eds.), Third Symposium on Mesozoic Terrestrial Ecosystems, Short Papers. Attempto Verlag, Tübingen. c Cooper, M. R A revision of the ornithischian dinosaur Kangnasaurus coetzeei Haughton, with a classification of the Ornithischia. Ann. S. Afr. Mus. 95:

69 55 Cooper (1985) similarly suggested that Ornithischia was diagnosed by an opisthopubic pelvis and a predentary, but also agreed with Norman (1984) that the group was further diagnosed by a supraorbital element (=palpebral) and a pendent fourth trochanter of the femur. However, Cooper (1985) also included wear facets on the tooth crowns as a diagnosable feature. Undoubtedly, the most influential analysis of global ornithischian relationships based on a cladistic assessment of morphology was Sereno (1986). In it, he built upon his previous analysis (Sereno, 1984) and expanded the character sampling. He listed 32 dental, cranial, and postcranial characters to diagnose Ornithischia (Table 2.1). Many of these characters are based off the condition observed in Lesothosaurus (Sereno, 1986), as he considered it to be the basalmost ornithischian. One major difference observed in Lesothosaurus that is not noted in basal saurischians such as Herrerasaurus and Eodromaeus, is a shortened forelimb (Sereno and Novas, 1993; Martinez et al., 2011). In Lesothosaurus, the forelimb is less than half the length of the hindlimb (~40%), whereas in those basal saurischians and even some basal ornithischians (e.g., Heterodontosaurus; Santa Luca, 1980), the forelimb is rather long (~60%) relative to the hindlimb. Additionally, the manus of basal saurischians and some basal ornithischians (e.g., Heterodontosaurus) is large with deep, well-developed extensor pits on the distal ends of the metacarpals and non-ungual phalanges of digits I- III (Butler, 2010; Martinez et al., 2011). As the hypotheses of ornithischian relationships have changed (e.g., Butler et al., 2008), so have the interpretations of character evolution within the clade. For example, the weak buccal emargination present in Lesothosaurus and Abrictosaurus may not be the

70 56 ancestral condition for Ornithischia if these taxa particularly Lesothosaurus are nested higher in the tree than previously supposed. Recently, O Leary et al. (2013) were able to reconstruct the hypothetical ancestor of placental mammals and confirm what had been long assumed about the ancestral species; namely, that it is a small and rodentlike taxon. They also determined that the hypothetical ancestor was insectivorous, scansorial, and that the young were born hairless and with their eyes closed (their data set consisted of phenomic characters, including soft tissue, and molecular data). Additionally, they were able to push the divergence of crown Placentalia more recently into the Paleocene (contra Meredith et al., 2011). Given the geologic age, phylogenetic position, and geographic distribution of basal ornithischian taxa, as well as the geologic age and geographic distribution of basal saurischians (e.g., Eodromaeus, Herrerasaurus) and non-dinosaurian archosauriforms (e.g., Euparkeria, Marasuchus), it is likely that the location and time of divergence for ornithischian dinosaurs is well-constrained to southern Gondwana (southern South America or southern Africa) in the Middle to Late Triassic. The results that stem from this analysis, though, can inform our hypotheses of how ornithischians evolved from this location in time and space into the diversity noted throughout the Mesozoic. Methods Phylogenetic Analysis To properly evaluate the ancestral body plan for Ornithischia, the group first needs to be placed in a phylogenetic context. Therefore, I compiled a data set consisting

71 57 of 52 ingroup taxa focusing on basal members (e.g., Eocursor, Pisanosaurus, Lesothosaurus, heterodontosaurids), though several members of derived clades (e.g., Huayangosaurus, Tenontosaurus, Psittacosaurus) were included. The archosauriform Euparkeria capensis and the basal saurischian Herrerasaurus ischigualastensis were chosen as outgroups. The taxa were scored for 247 discrete morphological characters, all treated as equally weighted and unordered. A maximum parsimony analysis was performed in TNT v. 1.1 (Goloboff et al., 2008). Specifics for the phylogenetic analysis are presented in Chapter I above and in Appendices B-D. Ancestral State Reconstruction The data set and the resulting trees from the phylogenetic analysis were imported into Mesquite v (Maddison and Maddison, 2011). Mesquite allows for numerous ways to evaluate character evolution across individual trees and groups of trees alike. To assess the most parsimonious reconstructions of all ancestral character states supported by all trees, I used the Trace All Characters option. Mesquite allows the user to select a clade of interest in this case, Ornithischia and with that clade selected, the ancestral states were calculated across all MPTs from the phylogenetic analysis via the assigned parsimony model. Under parsimony reconstruction of ancestral states, the data were treated with the unordered character assumption for categorical data by which any state change is counted as one step (e.g., changing from state 0 to state 1 is just as costly as changing from state 1 to state 3) so as to minimize the amount of ad hoc assumptions of character evolution during the analysis.

72 58 Fossil data sets tend to be associated with missing data due to incompleteness inherent in the rock record and because of incomplete or fragmentary taxa (e.g., Wiens and Morrill, 2011). Additionally, some characters are necessarily dependent on other characters (mother characters and daughter characters) e.g., character 4: presence or absence of a ventrally-directed and anteriorly-keeled rostral bone is dependent upon whether or not the taxon possesses a rostral bone (character 3). Mesquite treats these characters (i.e., inapplicable characters) in the same way it treats missing data. Because of this, the ancestral state reconstructions can include estimates that are not biologically meaningful. The results below detail only those characters that, unlike characters 4 and 5 (both related to the presence of a rostral bone; Appendix C), have the potential to legitimately be present in the ancestral taxon. Inapplicable characters were ignored when reconstructing the hypothetical ancestral ornithischian. Additionally, when the mother character was ambiguous, a daughter character was not reconstructed. Exceptions to these rules are noted where appropriate below. The same method as above was repeated for several other prominent clades of ornithischian dinosaur, including: Genasauria, Thyreophora, Neornithischia, Cerapoda, and Marginocephalia. All characters were reconstructed at each of the nodes that include these taxa. However, only the ancestral ornithischian is illustrated below. Results The results of the phylogenetic analysis are as they were in Chapter I (Fig. 2.1, a simplified phylogeny from the results in the preceding chapter). The strict consensus tree

73 59 was fairly well resolved among basal taxa and robustly recovered Pisanosaurus as the most basal ornithischian. Heterodontosauridae was recovered as a monophyletic group outside of Genasauria. Genasauria is divided among two major clades, Thyreophora and Neornithischia. Thyreophora, a generally uncontroversial clade that consists of Eurypoda and successively more distantly related taxa, Scelidosaurus, Emausaurus, and Scutellosaurus. Unlike previous recent analyses of basal ornithischian relationships (Chapter III; Spencer, 2007; Butler et al., 2008), Lesothosaurus was not recovered here as a basal thyreophoran, but rather as the basalmost neornithischian (as in Butler et al., 2010, 2011). The base of Cerapoda, primarily due to the instability of putative basal ornithopods, is unresolved. The ancestral state reconstruction analysis permitted scoring of 85% of the characters in the data matrix (210 of 247 characters). The remaining 37 characters were either ambiguous (26 characters) or inapplicable (11 characters). The overwhelming majority of the reconstructed character states for the ancestral ornithischian were scored for the plesiomorphic state. Only 21 of the 210 characters scored (10%) were reconstructed as the derived condition. Appendix F lists all the characters and indicates those that were reconstructed, those that were ambiguous, and those that were inapplicable. As with the ancestral ornithischian, the ancestral reconstructions of Genasauria, Thyreophora, Neornithischia, Cerapoda, and Marginocephalia also possessed several characters that were ambiguous or inapplicable (Table F1). For Genasauria, 88% of the characters were scored (25 ambiguous, 10 inapplicable). Thyreophora also had 88% of scored characters (25 ambiguous, 10 inapplicable). Neornithischia had 85% of the

74 Figure 2.1 Summary phylogeny of Ornithischia from the phylogenetic analysis presented in Chapter I. Large, well-supported clades have been collapsed for clarity. 60

75 61 characters scored (26 ambiguous, 11 inapplicable), whereas, Cerapoda had 92% (21 ambiguous, 8 inapplicable) and Marginocephalia had 90% (22 ambiguous, 8 inapplicable) scored. Discussion and Description The cranial osteology of ornithischian dinosaurs is tightly interconnected with specializations for herbivory, even among the basal members (Weishampel, 2004). As such, aspects of jaw mechanical function are of particular interest (e.g., Rybczynski et al., 2008), as are studies of sexual dimorphism (e.g., Thulborn, 1974; Hopson, 1975), ontogeny (e.g., Sampson et al., 1997), and physiology (e.g., Goodwin and Horner, 2004). Because of the highly modified nature of many ornithischian skulls of both derived (e.g., ceratopsians) and basal (e.g., heterodontosaurids) taxa, it is difficult to estimate the ancestral condition without analyzing the characters quantitatively. The postcranial skeleton of ornithischians is less problematic than the skull, but there still are modifications that have been made to the body plan for accommodating herbivory in addition to remaining bipedal, at least ancestrally (Weishampel, 2004). Postcranial material for many of the taxa in this analysis is generally more fragmentary and preserves less phylogenetic information than the cranial material. In fact, 25 of the 37 ambiguous or inapplicable character reconstructions are from postcranial characters. However, much about the general morphology of the ancestral ornithischian can be inferred from the characters that were able to be reconstructed.

76 62 The general morphology for the body plan of the ancestral ornithischian hereafter, AO is discussed below, with more specific information and particular attention paid to those characters that have been optimized on the phylogeny presented in Figure 2.1 and based on the data matrix evaluated here. Cranial Osteology (Fig. 2.2) The premaxilla of AO maintains an edentulous anterior portion (the width of about one premaxillary tooth crown) that is rugose, which suggests that it possessed a (likely) keratinous rhamphotheca (horny beak), similar to that observed in Lesothosaurus (Sereno, 1991) but lacks an anterior premaxillary foramen. The oral margin of the premaxilla is level with the maxillary tooth row yet it is unclear if the premaxillary palate is arched as in most ornithischians. The posterolateral process is wide and tall but does not contact the lacrimal, as in Agilisaurus (Peng, 1992) or Lesothosaurus (NHMUK R8501). The narial aspect of the body of the premaxilla slopes steeply toward the oral margin. The dorsal process overlaps the nasals to a larger extent than observed in Agilisaurus (Peng, 1992, 1997) but similar to the condition observed in Lesothosaurus (Sereno, 1991). The external nares are relatively small and are bounded anteriorly, posteriorly, and ventrally by the premaxilla. A distinct subtriangular narial fossa extends ventrally to near the ventral margin of the premaxilla. The nasals are thin and possess a deep elliptical fossa along the medial suture, as in Heterodontosaurus (SAM-PK-K337, SAM-PK-K1332). The external nares are encased dorsally by the nasals, which contact the premaxilla to form an internarial bar as in Lesothosaurus (NHMUK 8501).

77 Figure 2.2 Reconstructed cranium of the hypothetical ancestral ornithischian. The numbers refer to the characters in Appendix F that were used to reconstruct the ancestral states illustrated here. 63

78 64 Relatively large (with respect to all other ornithischians), external antorbital fenestrae are present in AO, similar to the condition seen in basal saurischians (e.g., Eoraptor, Sereno et al., 1993; Eodromaeus, Martinez et al., 2011; Herrerasaurus, Sereno and Novas 1993), though more triangular than ovoid. The fenestrae are bounded predominantly by the maxilla anterodorsally and ventrally, with only a small contribution of the lacrimal to the posterior border. The maxilla is robust and triangular in lateral view. It also lacks any anterior processes that laterally overlap a corresponding medial process of the premaxilla. A dorsally located slot for the lacrimal to insert into the maxilla is not present as it is in Lesothosaurus (NHMUK R8501). Buccal emargination is present in AO, but the extent to which the maxillary tooth row is inset is unknown. Abrictosaurus (NHMUK RUB54) and Lesothosaurus (SAM-PK-K426) demonstrate incipient buccal emargination; however, nearly all other basal ornithischians (e.g., Heterodontosaurus, SAM-PK-K337; Pisanosaurus, Sereno, 1991; Irmis et al., 2007) show marked emargination. The presence or absence of an accessory ossification that extends freely into the orbit (palpebral) was noted as ambiguous in the analysis; however, all ornithischians that preserve the orbit possess a palpebral (supraorbital) or evidence of the attachment of a palpebral. Therefore, one rodlike palpebral was most likely present in AO that did not traverse the entire width of the orbit. Like other basal ornithischians, this free palpebral likely attached to the prefrontal and lacrimal in the anterodorsal region of the orbit Unlike Agilisaurus (Peng, 1992; Barrett et al., 2005), the jugal of AO is not excluded from the antorbital fossa as in Lesothosaurus (NHMUK R8501) and Hexinlusaurus (Barrett et al., 2005). The orbital ramus is deeper than it is wide, but not

79 65 deeper than the posterior ramus. The jugal is fairly nondescript, lacking any lateral flanges or bosses as in heterodontosaurids or basal marginocephalians. The jugalpostorbital bar is fairly thin with an elongate overlapping scarf joint that, along with the posterior process, forms the anteroventral margin of the infratemporal fenestra. The posterior process overlaps the quadratojugal laterally. The postorbital is relatively smooth and T-shaped as in most basal ornithischians. Its contact with the parietal is unclear, but if it does, it likely only narrowly contacts it in dorsal view within the supratemporal fossa as in Lesothosaurus (Sereno, 1991). A modest depression behind the orbital bulge of the postorbital was also likely present. The quadratojugal is L-shaped and faces laterally. Its posteroventral margin approaches the mandibular condyle of the quadrate. The quadrate is posteriorly concave in lateral view. Unlike Agilisaurus (Peng, 1992, 1997; Barrett et al., 2005), the head of the quadrate is excluded from the dorsolateral corner of the infratemporal fenestra. The paraquadratic foramen is situated on the quadrate-quadratojugal boundary, visible in posterolateral view. The medial condyle is larger than the lateral condyle on the mandibular articular surface. The paired frontals are subrectangular and form the anteromedial boundary of the supratemporal fenestrae. The supratemporal fenestrae are oval to circular, similar to those observed in Lesothosaurus (NHMUK R8501). As in Lesothosaurus, the parietal septum is fairly narrow and smooth and forms the medial border of the supratemporal fenestrae. The lateral borders are formed by a smooth, rodlike postorbital-squamosal bar, as in most other basal ornithischians.

80 66 In posterior view, the paroccipital processes are short and deep and extend out laterally and expand distally. Unlike other basal ornithischians, the posttemporal foramen is situated between the paroccipital process and the parietal. The location of the posttemporal foramen in AO is similar to the condition of Euparkeria (SAM-PK-K5867; Ewer, 1965) and Herrerasaurus (Sereno and Novas, 1993). The supraoccipital forms the entire margin of the foramen magnum, whereas, the basioccipital contributes to the ventral margin and is not excluded by the exoccipitals. The basisphenoid is roughly equal in size to the basioccipital. There is no prooticbasisphenoid plate as in Lesothosaurus (NHMUK RU17, R8501; Sereno, 1991); however, as in Lesothosaurus, the basisphenoid forms the lateral portion of the knobshaped basal tubera. Unlike Heterodontosaurus (SAM-PK-K337; Norman et al., 2011), the basipterygoid process projects anteroventrally, though likely not to the extent observed in Lesothosaurus. Limited information is available for the remainder of the skull elements of AO; however, it is clear that there was no contact between the pterygoids and maxillae at the posterior end of the tooth row and the maxillae did not contact at the midline, thereby precluding a premaxilla-vomer communication. Mandibular Osteology (Fig. 2.3) In the current analysis, the presence of a predentary in AO is ambiguous; however, in all other ornithischians for which there is evidence, a predentary is present or assumed to be present as in Pisanosaurus (Bonaparte, 1976; Sereno, 1991; Irmis et al.,

81 Figure 2.3 Reconstructed mandible of the ancestral ornithischian. The numbers refer to the characters in Appendix F that were used to reconstruct the ancestral states illustrated here. 67

82 ). Therefore, it is likely that AO possessed a predentary that is relatively short such that the premaxillary teeth oppose the anteriormost dentary teeth. The articular surface of the dentary for the predentary is relatively flat, though the dorsal and ventral margins of the dentary converge anteriorly. The dentary is generally thin, unlike NHMUK A100 (here referred to Lanasaurus; Norman et al., 2011), and possesses an anterior shallow groove that serves as the vascular tract from the anterior dentary foramen (as noted by Sereno, 2012). The tooth row is generally straight in lateral view. The posterodorsal process of the dentary forms the anterodorsal margin of the rudimentary coronoid process. The coronoid process is entirely located posterior to the tooth row. The angular forms less than half the maximum height of the mandibular ramus, as in Lesothosaurus (NHMUK R8501). In basal saurischians like Herrerasaurus and some basal ornithischians like Heterodontosaurus, there is an external mandibular fenestra situated on the dentary-angular-surangular boundary and AO possessed a similar fenestra. A small, shallow groove (and possibly a foramen, though it is unclear) is present on the anterior surface of the surangular, like in Heterodontosaurus (SAM-PK-K1332). The retroarticular process is longer than the glenoid facet and the level of the jaw joint is, at most, slightly depressed relative to the tooth row. Dentition (Figs. 2.2, 2.3) The number of premaxillary teeth for AO is ambiguous. In Tianyulong, there are only two premaxillary teeth (Zheng et al, 2009). In most other heterodontosaurids, there are generally three premaxillary teeth (though, it is debatable as to the number of

83 69 premaxillary teeth are in Fruitadens; Galton, 2007; Butler et al., 2010; Sereno, 2012). In Herrerasaurus (Sereno and Novas, 1993), there are four premaxillary teeth and in other basal ornithischians such as Haya (Makovicky et al., 2011), there are five (Lesothosaurus has six premaxillary teeth). Here, I have reconstructed AO with four premaxillary teeth (Fig. 2.2), but the results of the analysis suggest four possibilities: four (the plesiomorphic condition, 0), five (character state 1), six (character state 2), or three (character state 3). The teeth are conical, of similar size, and the crowns do not expand above the root. The maxillary and dentary tooth crowns (Figs. 2.2, 2.3) are not mesially or distally curved and are all relatively short and subtriangular. There are likely no wear facets among the teeth as noted in heterodontosaurids and Pisanosaurus (Sereno, 1991) and enamel is symmetrically distributed on both sides of the crowns, unlike ornithopods. Apicobasally-extending ridges are present; however, they are not confluent with the marginal denticles. Though the tooth crowns are not labiolingually swelled above the root, they are swelled mesiodistally, creating an intermediate step between lacking cingula and possessing fully developed crown expansions above the root. There are no alveolar foramina (or special foramina as in Galton [1978]) associated with the maxillary and dentary tooth rows. Postcranial Osteology Axial (Fig. 2.4) The axial skeleton of AO is not unlike most other basal ornithischians. There are likely between seven and nine cervical vertebrae (Heterodontosaurus has nine; SAM-PK- K1332, as does Manidens; Pol et al., 2011). There is no fusion between the intercentrum

84 Figure Reconstructed axial elements of the ancestral ornithischian. The numbers refer to the characters in Appendix F that were used to reconstruct the ancestral states illustrated here. 70

85 71 of the atlas and the neural arches, unlike in heterodontosaurids (Sereno, 2012) and there are no prominent epipophyses (in Heterodontosaurus, the epipophyses are hypertrophied; SAM-PK-K1332). There are 15 dorsal vertebrae and the zygopophyseal articulations of the dorsal vertebrae are flat, with the prezygapophyses facing dorsally and the postzygapophyses facing ventrally, as in most basal ornithischians such as Stormbergia (Butler, 2005) and Scutellosaurus (Colbert, 1981). The sacral count, like the cervical vertebrae, is uncertain. There are likely four or more sacral vertebrae, though it is possible that there are as few as two. Additionally, the sacral ribs are all roughly equal length. Unlike Heterodontosaurus (SAM-PK-K1332) there is no accessory articulation between the sacrum and the pubis. The caudal series is modest, possessing around 50 vertebrae, similar to Lesothosaurus (Sereno, 1991) and Heterodontosaurus (SAM-PK-K1332; Santa Luca, 1980). The proximal neural spines are at least as tall as the associated centra. Chevrons are rodlike, occasionally with a distal expansion. There are no ossified sternal ribs or gastralia. It is unclear if AO would have had ossified clavicles or sternal plates. Subrectangular sternal plates are present in Heterodontosaurus (SAM-PK-K1332). Postcranial Osteology Pectoral Girdle & Forelimb (Fig. 2.7) The scapula is at least as long as the humerus as in Abrictosaurus (NHMUK RUB54) and Heterodontosaurus (SAM-PK-K1332), though, unlike other basal ornithischians such as Agilisaurus (Peng, 1992, 1997), Hexinlusaurus (Barrett et al.,

86 ), Hypsilophodon (Galton, 1974), and Psittacosaurus (Sereno, 1990, 2000). There is a well-developed, spinelike acromion as in Tianyulong (Zheng et al., 2009), Heterodontosaurus (SAM-PK-K1332; Santa Luca, 1980), and Stormbergia (SAM-PK- 1105; Butler, 2005). The humerus is straight, large (more than 60% of the length of the femur), and possesses a well-developed and anteriorly-directed deltopectoral crest as in Heterodontosaurus (SAM-PK-K1332; Fig. 2.5) and Abrictosaurus (NHMUK RUB54; Fig. 2.6). Tianyulong possesses a well-developed deltopectoral crest as well, but the humerus itself is not nearly as long (Zheng et al., 2009). The manus (Fig. 2.7) is extremely large for most basal ornithischians, though a large manus is present in most heterodontosaurids (except for Tianyulong; Zheng et al., 2009) and in Herrerasaurus (Sereno, 1993). Additionally, the carpus is much like that of heterodontosaurids as well as basal saurischians Eodromaeus and Herrerasaurus (Martinez et al., 2011) in that there are numerous bones typically a radiale, ulnare, and pisiform in the proximal row; an os centrale (or os intermedium; Santa Luca, 1980); and a carpal for each of the five metacarpals in the distal row. This is very unlike the situation in other basal ornithischians with only two or three carpals (e.g., Lesothosaurus, Sereno, 1991; Hypsilophodon; Galton, 1974). Metacarpals 1 and 5 are substantially shorter than metacarpal 3. The penultimate phalanges of digits 2 and 3 are longer than the first phalanx. As in heterodontosaurids and Eocursor (Fig. 2.7) there are deep, well-developed extensor pits on the distal ends of the metacarpals and non-ungual phalanges (Butler, 2010). The unguals are strongly recurved and possess a prominent flexor tubercle. This suggests a powerful grasping manus, as is the case with basal saurischians such as

87 Figure 2.5 Left forelimb of Heterodontosaurus tucki (SAM PK-1332) in lateral view. Note the size of the manus relative to the rest of the forelimb and the strongly recurved unguals. Scale bar = 2 cm. 73

88 Figure 2.6 Forelimb of Abrictosaurus consors (NHMUK RU B54) in ventral view. Given the poor preservation of the available material, determining the presence/absence of the characters presented here is difficult. Scale bar = 2 cm. 74

89 Figure 2.7 Reconstructed pectoral girdle elements of the ancestral ornithischian. The numbers refer to the characters in Appendix F that were used to reconstruct the ancestral states illustrated here. 75

90 76 Herrerasaurus (Sereno, 1993) and in at least some heterodontosaurids (e.g., Heterodontosaurus, Figs. 2.5, 2.7). There has been some debate surrounding the pollex in Heterodontosaurus with respect to the natural position and degree of extension. Bakker and Galton (1974) suggested an attainable state of hyperextension in the pollex as it tracked medially, capable of being held off the ground in the event that Heterodontosaurus was walking quadrupedally as seen in basal sauropodomorphs (e.g., Eoraptor; Sereno et al., 1993). This was summarily dismissed by Santa Luca (1980) and others (e.g., Weishampel and Witmer, 1990; Sereno, 2012), suggesting that the digit was much straighter than Bakker and Galton (1974) had suggested. The specimen in question (SAM-PK-K1332; Fig. 2.8) has been damaged over time and the natural position of the digit is unclear. Regardless, the highly asymmetrical distal condyles of metacarpal 1 indicate that, during flexion, the pollex deflected toward the palm. Unfortunately, little can be said of the forearm of AO. Based on the aforementioned forelimb characters, though, it seems likely that the ulna had a fairly prominent olecranon process and that, as in Heterodontosaurus (SAM-PK-K1332), Abrictosaurus (NHMUK RUB54), Herrerasaurus (Sereno, 1993), Eoraptor (Sereno et al., 1993), and Eodromaeus (Martinez et al., 2011), the manus was nearly half the length of the combined humeral-ulnar length.

91 Figure 2.8 Right forelimb of Heterodontosaurus tucki (SAM PK-1332) in dorsal view. Note the labeled extensor pits for ligament attachment to facilitate digit hyperextension. Scale bar = 2 cm. 77

92 78 Postcranial Osteology Pelvic Girdle & Hindlimb (Fig. 2.10) The preacetabular process of the ilium is relatively abbreviated anteriorly and laterally compressed. The dorsal margin of the ilium above the partially perforate acetabulum is relatively straight. Much ambiguity surrounds the postacetabular process of the ilium. The process is not as long as in Orodromeus (Horner and Weishampel, 1988) but not as abbreviated as in Heterodontosaurus (SAM-PK-K1332). The brevis shelf/fossa faces likely ventrolaterally. The presence of a supraacetabular crest or flange is uncertain. It is present in Heterodontosaurus, Lesothosaurus, Stormbergia, and Eocursor, but not in most other basal ornithischians such as Abrictosaurus. The pubic and ischial peduncles are robust and project ventrally. The ischium is relatively straight and bladelike and lacks a tab-shaped obturator process as in Eocursor, Tianyulong, and Lesothosaurus (unlike Stormbergia; contra Knoll et al., 2010). The ischia form a median symphysis along at least half the length of the shaft. It is ambiguous as to whether or not the ischium for AO possessed a dorsal groove as in Tianyulong and Stormbergia (unlike Lesothosaurus; contra Knoll et al., 2010). Because of the uncertainty of Pisanosaurus, it is unclear if AO would have had a fully-developed opisthopubic pelvis (Sereno, 1991; Irmis et al., 2007; contra Bonaparte, 1976). But, given that all other ornithischians maintain an opisthopubic pelvis and that the material for Pisanosaurus is equivocal, I am reconstructing AO with an opisthopubic pelvis. The femur is bowed anteriorly in lateral view. As in Eocursor, Lesothosaurus, Fruitadens, Stormbergia, and Scutellosaurus, the femoral head is confluent with the

93 79 greater trochanter, forming a shallow groove (fossa trochanteris). In other basal ornithischians (e.g., basal cerapodans), the head of the femur is separated by a distinct constriction. The fourth trochanter is located entirely on the proximal half of the femur, though its general shape (e.g., pendent, prominent flange) is uncertain. The anterior distal end of the femur lacks an extensor intercondylar groove, but the posterior intercondylar groove is fully open. Both lateral and medial condyles are subequal in size as in Valdosaurus (NHMUK R184, R185). The distal tibia does not possess a posterolateral process to support the fibula as in Stormbergia and Lesothosaurus (Fig. 2.9) and is not as expanded as the proximal tibia. As in most other basal ornithischians, the fibula is well-formed as in basal saurischians such as Herrerasaurus (Rauhut, 2003). The astragalus and calcaneum are separate elements, not fused together as in Heterodontosaurus, Tianyulong, and Fruitadens. The calcaneum also lacks a facet for the tibia. The medial distal tarsal articulates with metatarsal 3 only and the metatarsals are closely packed with each other along more than half of the their length and then spread distally, as in Abrictosaurus and Tianyulong (and other basal ornithischians). The pedal unguals, like the manual unguals, are narrow and pointed and appear very clawlike, as in many basal ornithischians.

94 Figure 2.9 Distal (A) left tibia of Stormbergia dangershoeki (NHMUK R11000) and (B) right tibia of Lesothosaurus diagnosticus (SAM PK-1106) in posterior view. Note the prominent posterolateral process that supports the fibula (not shown) anteriorly to strengthen the ankle joint. Scale bar = (A) 3 cm; (B) 1 cm. 80

95 Figure 2.10 Reconstructed pelvic girdle and limb elements of the ancestral ornithischian. The numbers refer to the characters in Appendix F that were used to reconstruct the ancestral states illustrated here. 81

96 82 Postcranial Osteology Accessory Ossifications (Fig. 2.4) As in all other ornithischians (except for all stegosaurs other than Huayangosaurus), AO possessed ossified epaxial tendons along at least the sacral and posteriormost dorsal vertebrae. Ossified hypaxial tendons (present in Tianyulong and Hypsilophodon) are not present in AO. Comparison with Ornithischian Subclades (Table F1) The reconstructed ancestor of Genasauria differs, not including those characters that are ambiguous or inapplicable, from AO in 11 characters (Table F1). Four of those characters, though, are related to the manus (176, 179, 182, 183). The ancestral genasaurian, which is mirrored nearly entirely in the ancestral thyreophoran and neornithischian, has a reduced hand relative to AO (Fig. 2.7). The ancestral neornithischian differs from the ancestral genasaurian only in the position of the posttemporal foramen; however, the ancestral thyreophoran differs from the ancestral genasaurian in two characters in the posterior portion of the skull (36 wide orbital ramus of the jugal; 96 cortical remodeling of dermal bone) and with the presence of osteoderms. The ancestral cerapodan and marginocephalian are more disparate from AO and the ancestral genasaurian, which is expected given that these are more derived clades nested within Neornithischia.

97 83 Conclusion The characteristics presented here indicate that many of the features postulated for herbivory in ornithischians were present in the last common ancestor of all ornithischians. There is still some ambiguity as to the presence or form of a few of these features (e.g., opisthopubic pelvis; Irmis et al., 2007). Each of these characters, present in some form or another, in basal ornithischians illuminates the ancestral bauplan for Ornithischia. The skull of the ancestral ornithischian was not unlike that of Lesothosaurus; not overly elongate relative to the rest of its body, but slender and gracile. Its maxillary and dentary teeth were loosely packed in modestly-formed inset jaw margins. These cheek teeth were leaf-shaped with denticles present that provided occlusal surfaces for mechanical processing of vegetation. The premaxillary teeth, set behind a keratinous beak as indicated by the edentulous premaxilla, were peglike and unornamented. The overall similarities between the basal ornithischians and basal saurischians from the Late Triassic suggest that early dinosaur skeletal morphology, particularly the postcrania, was more similar than previously thought. The view presented here provides a glimpse into the evolution of dinosaurs in the Late Triassic and Early Jurassic.

98 84 CHAPTER III DISTRIBUTIONAL PATTERNS IN ORNITHISCHIA: A COMPREHENSIVE SPECIES-LEVEL BIOGEOEGRAPHICAL ANALYSIS Abstract Ornithischians, obligately terrestrial animals and therefore ideal taxa for studying global biogeographic patterns, first appear in the Late Triassic of southern Gondwana and achieve a nearly worldwide distribution by the late Early Jurassic. Because of their terrestrial habit, analyzing ornithischian distribution patterns can potentially illuminate unknown or underlying geographic signals such as dispersal events that might otherwise be obscured by continental fragmentation patterns. The early fossil record for ornithischians is sparse and predominantly Gondwanan, but the latter half of ornithischian history in the Mesozoic is characterized by a Laurasian distribution. Ephemeral connections between North America and Asia and the island conglomerate that was Europe presents ample opportunities to test patterns of dispersal throughout the late Mesozoic. Here, I present a global species-level biogeographic analysis of all Ornithischia throughout the Mesozoic. In total, 176 basal and derived species were included in a phylogenetic framework, built from the concatenation of multiple phylogenies of subclades (e.g., Hadrosauroidea, Stegosauria), to illustrate the overall phylogeny of Ornithischia. A parsimony optimization of areas from which each species was recovered delineates the putative region of origin for each major clade. Statistical tests suggest that several routes of dispersal between North America and Asia as well as

99 85 between Europe and North America and Europe and Asia were significantly more abundant than expected by chance. For larger well-established clades such as Thyreophora and Marginocephalia, ad hoc hypotheses of dispersal across Laurasia are required, either via island hopping from North America to Europe or across ephemeral land bridges between North America and Asia throughout the Cretaceous. The basal radiation of Ornithopoda is problematic in that there are multiple independent dispersals into present-day Asia, Australia, North America, and Europe that are inconsistent with the sequence of landmass fragmentation. These results suggest the presence of short-lived connections within and among these landmasses. Introduction As obligately terrestrial animals, ornithischian dinosaurs are ideal for analyzing global geographical vicariance and dispersal patterns. These patterns can potentially illuminate previously unknown geological features such as archipelagos and land bridges that have been otherwise unnoticed or uncorroborated in the geological record. The early ornithischian record is sparse and predominantly Gondwanan, but the group achieved a global distribution shortly after the Triassic-Jurassic mass extinction (Weishampel, 1990; Upchurch et al., 2002; Weishampel et al., 2004a, b; Holtz et al., 2004). Ornithischian global phylogeny has undergone substantial systematic revision recently (e.g., Butler et al., 2010; Chapter I, above). The positions of major subclades have shifted within the ornithischian tree; thereby, potentially altering previously proposed areas of origin for major groups.

100 86 Biogeography The field of biogeography has been around as long as evolutionary biology. The notion that evolutionary relationships can inform hypotheses of geographic distributions was even part of Darwin s (1859) seminal work. Ever since Darwin, though, biogeography has undergone substantial revision, particularly after Hennig (1966) published his methods on how best to estimate phylogenies. The thought process is that if two taxa share a similar feature, it is hypothesized that they inherited that feature from a common ancestor. From there, we can build a pattern of relationships based on these features. Similarly, in biogeography, if two closely related taxa share similar distribution patterns, it is hypothesized that they share a common origin (Ronquist and Sanmartín, 2011). Several methods have been developed to analyze biogeographic distributions. These include, e.g., component analysis (Nelson and Platnick, 1981), Brooks parsimony analysis (Brooks, 1985), tree reconciliation analysis (Page, 1995), and paralogy-free subtree analysis (Nelson and Ladiges, 1996). Of these, only tree reconciliation analysis considers the evolutionary processes that can potentially create distribution patterns (Ronquist, 1998, 2002). More recently, improvements to tree reconciliation analysis (e.g., DIVA; Ronquist, 1997) have developed cost assignments to different evolutionary processes (e.g., dispersal and vicariance events; see below). As statistical inferences of distribution patterns have improved, they have been more frequently incorporated into larger-scale analyses of global organismal phylogenies (e.g., Upchurch et al., 2002).

101 87 Ornithischian Biogeography The earliest known ornithischians are from the Late Triassic of southern Pangaea, the presumed location of their evolutionary origin (Figs. 3.1, 3.2; Chapters I and II, above). By the Early and Middle Jurassic, ornithischians had achieved a worldwide distribution (Figs. 3.1, 3.2), ranging from southern Gondwana (present-day southern South America and South Africa) to northern and eastern Laurasia (present-day east Asia). By the Late Cretaceous, though, nearly all ornithischians known were confined to Laurasia, with only scant evidence of ornithischians located in the Gondwanan continents (Figs. 3.1, 3.2). Historical biogeography among dinosaurs has been evaluated in two primary ways: qualitative assessments of both large and small clades according to literal interpretations of the group s fossil record (e.g., Galton, 1977, 1982; Le Loeuff et al., 1992; Fastovsky and Weishampel, 1996), or quantitative assessments using statistical inferences to infer processes of distribution (e.g., Upchurch et al., 2002; Holtz et al., 2004). A qualitative assessment based on a literal reading of the fossil record is by far the most common practice, typically because small clades of endemic species lend themselves to uncontroversial assertions regarding areas of endemism or patterns of vicariance (Holtz et al., 2004). Because of their predominantly terrestrial habit, the biogeography of dinosaurs should best be explained by vicariance (Humphries and Parenti, 1999). However, because of the relatively incomplete nature of the fossil record, this prediction of vicariance is not the preferred overall explanation for dinosaur distributions (Sereno, 1997, 1999; Upchurch et al., 2001). Unfortunately, as noted in Ronquist (2002; see also Sanmartín et

102 Figure 3.1 Mesozoic time scale and the break-up of Pangaea. This figure is used as a scale in Figure 3.2. Ages given on the time scale were obtained from Gradstein et al. (2004). Size of time bins is not to scale. Paleogeographic reconstructions are Ron Blakey. 88

103 Figure 3.2 Ornithischian global diversity throughout the Mesozoic. Red stars indicate approximate locations of taxa and skeletal reconstructions represent the different clades that appear in those locations. Yellow arrows indicate where on the temporal scale the particular time slice is located. Approximate locations of taxa were obtained from Weishampel et al. (2004b). Paleogeographic reconstructions are Ron Blakey. 89

104 90 al., 2001; Sanmartín and Ronquist, 2004; Ronquist and Sanmartín, 2011), most methods used for testing patterns are not sensitive enough to detect dispersals, extirpations (local extinctions), and duplications (in the form of vicariance via reticulate evolution). Upchurch et al. (2002) presented the most comprehensive quantitative large-scale analysis of dinosaur biogeography to date. They used time-sliced tree reconciliation analysis (see below) to analyze the global geographic distribution of Dinosauria. Several studies since then have also analyzed global patterns of distribution, but most of these were either focused on different questions such as latitudinal gradient (e.g., Mannion et al., 2011) or used different methods of statistical inference such as unweighted pair-group method with arithmetic averages (UPGMA; e.g., Holtz et al., 2004). In their phenetic analysis of Dinosauria, Holtz et al. (2004) used UPGMA to parse out patterns of taxa that explained their distributions within the localities in which they are found (e.g., Stegosaurus, Allosaurus, and Camarasaurus from the Morrison Formation). Their analysis of global dinosaur biogeography sampled localities, rather than time bins, and so the temporal constraint was actually finer than that of Upchurch et al. (2002). There are some non-trivial issues with the analyses presented in both Upchurch et al. (2002) and Holtz et al. (2004). Upchurch et al. (2002), due to technical limitations with the software available, were only able to sample 100 or fewer taxa. They did attempt to maintain the widest selection of species across all clades and geographic locations. Holtz et al. (2004), unlike Upchurch et al. (2002), used supraspecific taxa (77 total) in their analysis, which, in some instances, covered disparate geographic areas (e.g., Pachycephalosauria, present in Europe [Stenopelix], Asia [e.g., Homalocephale], and

105 91 North America [e.g., Pachycephalosaurus]). Both of these analyses utilize a temporal component, either directly ( time slicing in Upchurch et al., 2002) or indirectly (geologic units/localities in Holtz et al., 2004). However, this can be problematic because time bins can be subjective divisions based on, e.g., worker bias or preservation of the rock record. A more suitable approach to avoid some of these issues relating to temporal biases or composite taxa may be to analyze events (e.g., speciation, extirpation; Ronquist, 2002, Sanmartín and Ronquist, 2004). And, in the context of temporal segregation (e.g., time slicing; Hunn and Upchurch, 2001; Upchurch and Hunn, 2002), using these events may aid in discriminating between dispersals and vicariant events throughout the Mesozoic. These analyses, like many of the qualitative studies (e.g., Sereno, 1997, 1999), suggest that dispersal may be responsible for a large portion of dinosaur distributions. However, until recently, testing the significance of dispersals, duplications, and local extinctions was problematic (Ronquist, 2002). Here, I present a comprehensive specieslevel biogeographic analysis of Ornithischia throughout the Mesozoic, testing for patterns of vicariance, dispersal, duplication, and extinction. This represents the largest specieslevel analysis of ornithischians known. Materials and Methods Ornithischian Phylogeny A near-exhaustive species list of ornithischians was compiled and evaluated based on three criteria: species validity, known paleogeographic range and geologic age, and

106 92 phylogenetic information. Only currently valid species were included in this analysis. If a taxon has been described but little is known about its geologic age or paleogeographic range, it was excluded from the analysis. Only species that have been included in a recent (within the last ten years) quantitative phylogenetic analysis were selected (Appendix G). Thirteen phylogenetic analyses were combined to create a composite phylogeny of 176 species of ornithischians (Fig. 3.3; Appendix G). Twelve of the phylogenies were published previously. This biogeographic analysis began prior to the most recent version of the data set used in Chapters I and II above. As a consequence, several of the basal relationships of a few taxa are slightly different (according to the semi-strict consensus tree, not the strict consensus). For example, Lesothosaurus, prior to the phylogenetic analysis presented in Chapter I, was recovered as the basalmost thyreophoran. Additionally, Othnielosaurus and Yandusaurus were recovered outside of Cerapoda along with a Late Cretaceous clade comprised of Thescelosaurus, Parksosaurus, and Gasparinisaura. Weak support for a hypsilophodontid clade pulled out Hypsilophodon, Jeholosaurus, Orodromeus, and Zephyrosaurus as the basalmost ornithopods. Because this phylogeny had a wide sampling of basal species, it was used as the phylogenetic framework upon which to build the rest of the composite tree for more derived clades (as in Upchurch et al., 2002). As noted in Chapter I, logistical issues prevented a single, comprehensive phylogenetic analysis of all, or nearly all, ornithischian dinosaurs. Following the method of Upchurch et al. (2002), I was able to create a comprehensive phylogeny by combining trees from other published analyses that had at least one overlapping taxon. None of the twelve phylogenies had competing phylogenetic positions for any of the taxa in the

107 Figure 3.3 Time-calibrated composite phylogeny for Ornithischia. Taxon stratigraphic and geographic ranges are mapped onto the terminals. Numerical ages for each Epoch were obtained from Gradstein et al. (2004). 93

108 94 original phylogenetic analysis (or amongst each other) and, therefore, there are no contradicting phylogenies that were concatenated. Typically, a supertree or supermatrix analysis is conducted because of conflicting phylogenetic hypotheses; however, supertree analyses generally lose character information as the matrices are discarded in favor of only analyzing the competing topologies (e.g., Goloboff and Pol, 2002). A supermatrix approach is much more desirable for determining the optimal topology (or topologies) among competing hypotheses (though, missing data can be a confounding issue with supermatrices; de Queiroz and Gatesy, 2007). In the current analysis, however, these issues are not problematic as the topologies used are non-contradictory. Therefore, as in Upchurch et al. (2002), the combined phylogenies listed in Appendix G were used to create the composite phylogeny for the biogeographic analysis. Event-Based Parsimony Tree Fitting The composite phylogeny of 176 species was imported into TreeFitter v. 1.3 b1 (Ronquist, 2003). TreeFitter is a software program that, unlike most other parsimonybased historical biogeography, coevolution, and gene tree programs, employs an eventbased technique to tree fitting. Traditionally, biogeographic methods have relied mainly on tenuous relationships between distributional patterns and biogeographic processes. As such, hypotheses that include both vicariant and dispersal events often get muddled (Sanmartín and Ronquist, 2004). Page s (1995) TreeMap was the first program to implement events (i.e., vicariance, dispersal, extinction, and duplication) that are associated with a cost. In these

109 95 event-based tree-fitting methods, costs are inversely related to their likelihoods such that the more likely the event, the less costly it is for the hypothesis (Ronquist, 1998). Thus, the most parsimonious reconstruction is that which minimizes the cost across all implied events. Methods used for historical biogeography have a long history of use in analyses of coevolution and host-parasite systems, but they rarely have been used to parse out patterns of vicariance and/or dispersal (Page, 1995). Tree-fitting methods, such as dispersal-vicariance analysis (DIVA; Ronquist, 1997), do not force vicariance patterns to fit hierarchical patterns (Sanmartín and Ronquist, 2004). But, as Sanmartín and Ronquist (2004) point out, the lack of statistical power in DIVA prevents it from recognizing truly strong (significant) vicariance patterns when they do, in fact, exist. DIVA maximizes codivergence events (vicariant events), but it becomes sensitive to dispersals and duplications (reticulate geography) and can wash out any vicariant signal (Ronquist, 2002). As such, it is not statistically powerful enough to assess phylogenies that are mixed vicariance-and-dispersal events or those that lean heavily toward dispersal patterns (Ronquist, 2002). Here, I used TreeFitter s four-event model of event-based parsimony inference (Ronquist, 2003). The four-event model, also known as tree reconciliation (Page, 1995), assesses the frequency of: vicariance allopatric speciation due to a restrictive barrier; duplication allopatric or sympatric speciation due to a temporary barrier; sorting extinction (local) of a lineage in an area; and switching dispersal between isolated areas. Sanmartín and Ronquist (2004:fig. 5) addressed the issue of how to assess a cost to these events prior to determining the fit of an organism phylogeny to an area cladogram.

110 96 They indicate that each reconstruction dictates sets of correlated ancestral distributions and biogeographic events that could have potentially created an area cladogram. The cost assessed to the reconstruction is the sum of the costs of the events. The reconstruction with the minimum total cost is thus the most parsimonious reconstruction because it explains the distribution of the taxa in the fewest number of costly events. For example, an organism phylogeny that can best be explained by a fully vicariant sequence of events from the area cladogram is less costly, and therefore more parsimonious, than an organism phylogeny that is wrought with dispersal events to explain its area cladogram. Sanmartín and Ronquist (2004) liken this to a typical maximum parsimony analysis using explicit enumeration or heuristic search of stepwise addition sequences followed by branch swapping. Ronquist (2002) tested cost assignments for each of the four events. He determined positive cost assignments for each event was more appropriate with respect to the principle of parsimony, but that setting those costs was more problematic. However, using hypothetical data sets, he determined that for vicariance and duplication events, equal cost assignments were most appropriate because both produce phylogeneticallyconstrained patterns (in TreeFitter, the default cost assignment for both events is 2.0). Extinction events were slightly more costly, so he scored them as 1.0 in TreeFitter. Switching, or dispersal, events was far more costly and so are assigned the default cost of 0.2 in TreeFitter (Ronquist, 2002, 2003). Tree reconciliation analyses (TRA), such as DIVA (Page, 1995), have been updated to include time slices (Upchurch et al., 2002) because theoretical analyses have suggested that temporal ranges influence distribution patterns (Grande, 1985; Hunn and Upchurch, 2001, 2002; Upchurch and Hunn, 2002). These patterns seem to dissolve over

111 97 time as more and more patterns are superimposed on them (Grande, 1985; Upchurch et al., 2002). By incorporating these time slices into TRA, it is possible to parse out a potentially strong vicariant signal that would have been otherwise obscured because it failed to recover the optimal area cladogram or it failed the randomization tests (Upchurch et al., 2002). However, time slicing can also be potentially problematic because these time slices create organism cladograms that are less robust with respect to areas of origination for lineages. Input Hypotheses As noted above, Upchurch et al. (2002) provided the most comprehensive quantitative global biogeographic analysis of Dinosauria known to date. However, the dinosaur phylogeny used has dramatically changed since its publication. In particular, the base of Ornithischia has been drastically revised (e.g., Lesothosaurus is now well supported within Genasauria and Heterodontosauridae is deeply rooted at the base, rather than being more highly nested; Chapter I). The phylogeny presented here (Fig. 3.3) had to be adjusted, though, because at present TreeFitter cannot accommodate polytomies. Therefore, any clade that maintained a polytomy was assessed to determine how to prune species from it without losing geographic information. Fortunately, nearly all the polytomies consisted of taxa that were from the same geographic area (e.g., numerous ceratopsians from North America, heterodontosaurids from Africa). Because of this, I was able to apply Rosen s Rule (Rosen, 1979), whereby any polytomous taxa in a monophyletic group that occur in the same geographic area can be reduced to just one representative taxon without decreasing the amount of biogeographic information. There

112 98 was one polytomy that could not be resolved by pruning taxa via Rosen s Rule. In Stegosauria, there is a nested polytomy consisting of the African taxon Paranthodon, the European Loricatosaurus, the Asian Tuojiangosaurus, and the North American Stegosaurus. I chose to first prune Loricatosaurus and Paranthodon because two taxa just outside the polytomous clade, Kentrosaurus and Dacentrurus, are from Africa and Europe, respectively. I then reran the analysis, but pruning Paranthodon and Tuojiangosaurus, and then again by pruning Paranthodon and Stegosaurus (Paranthodon was pruned in each analysis because the sister taxon to the polytomous clade is the African Kentrosaurus, thus making Paranthodon redundant; Rosen, 1979). In each instance, the results were equivocal. Forty total taxa were pruned the majority of derived taxa in Ceratopsia were pruned because they all are North American species. Appendix H details the list of the remaining 136 species as well as the NEXUS file created for TreeFitter. Aside from just determining the optimal area cladogram for an organismal phylogeny, TreeFitter can also be used to test the fit of an organism phylogeny on any area cladogram. To test the significance of the optimal area cladogram that Upchurch et al. (2002) found for Ornithischia in light of the new phylogenetic hypotheses, I input their tree into the NEXUS file. Other geographic hypotheses were tested that have been suggested, including the currently-accepted generalized geological hypothesis of continental landmass break-up throughout the Mesozoic (Fig. 3.4; e.g., Smith et al., 1994; Smith and Rush, 1997; Olsen, 1997) and the hypothesis that a fragmented Europe and Africa were in communication for a substantial portion of the latter half of the Mesozoic (Ezcurra and Agnolin, 2012), as well as a few that appear possible by observing the

113 99 phylogeny in (Table 3.1, Appendix H). In particular, I wanted to test the potential association of Australian taxa with the South American species to see if there is any significant pattern that could be ascribed to a trans-antarctic route, provided that it holds that southern South America and the Antarctic Peninsula were connected across the present-day Drake Passage (e.g., Smith et al., 1994; Smith and Rush, 1997). Global Biogeographic Analysis of Ornithischia To determine the optimal area cladogram for Ornithischia, I performed a heuristic search in TreeFitter holding 100 trees for each stepwise addition, followed by tree bisection and reconnection (TBR) branch swapping with a reconnection limit of 20 nodes (nearly identical to a full standard TBR search as per Sanmartín and Ronquist [2004]). All multistate (widespread) taxa were treated using the recent option default setting in TreeFitter whereby the widespread distribution of taxa is assumed to be due to recent dispersal and only one area within the distribution is considered ancestral (Ronquist, 2003; Sanmartín and Ronquist, 2004). A total of 6,210 rearrangements were tried to find the most parsimonious reconstruction. To test the alternative area relationship hypotheses, I permuted each of the terminals for the geographical hypotheses 1,000 times and fit them to the organismal phylogeny. Because TreeFitter does not discard duplicate pseudoreplicates, there is the potential for artificially inflating the significance for results. With seven areas (continents) and the four events outlined above, that gives a total of 840 possible permutations (7*6*5*4 = 840) that can lead to a 19.05% increase in Type I error rates.

114 Figure 3.4 Generalized break-up sequence of Pangaea throughout the Mesozoic. Numerical dates obtained from Gradstein et al. (2004). The colors associated with the continental landmasses are constant throughout the chapter. 100

115 101 Table Area cladogram hypotheses. Optimal area cladogram from Upchurch et al. (2002) Modified from Upchurch et al. (2002) to demonstrate a relationship between South America and Australia and show a communication between Africa and Laurasia Modified from Upchurch et al. (2002) as above, but with a sister group relationship between Europe and Africa Currently accepted consensus hypothesis of continental fragmentation throughout the Mesozoic South America-Australia relationship H1 (SA,(AU,(AF,(EU,(AS,NA))))) H2 ((SA,AU),(AF,(EU,(AS,NA)))) H3 (((AS,NA)(EU,AF)),(SA,AU)) H4 ((AU,(AF,SA)),(NA,(EU,AS))) H5 (((SA,AU),NA),((EU,AF),AS)) Europe-Africa relationship, sister to the Americas Europe-Africa relationship, sister to Asia-Australia H6 (((NA,SA),(EU,AF)),(AS,AU)) H7 ((NA,SA),((EU,AF),(AS,AU))) Europe-Africa relationship, nested H8 ((((AF,EU),NA),AS),(SA,AU)) within Laurasia Note: H(1-8) refers to the assigned hypotheses noted in the text.

116 102 Therefore, significance values for these geographical hypotheses were considered significant at the p= or lower (0.05 [0.1905*0.05]). As for the optimal area cladogram analysis, all multistate taxa were treated using the default recent option (Sanmartín and Ronquist, 2004). I then calculated the frequency of the four biogeographic events for each hypothesis. The significance of each of these events was tested by permuting the organismal terminals. The first half of the Mesozoic is dominated by a single or loosely aggregated landmass (Pangaea); however, the latter half, the one in which ornithischians flourished, is marked by continuous fragmentation (e.g., Smith et al., 1994). More problematic is that during the Cretaceous, ornithischians were present mostly in Laurasia, a fragmented landmass that was likely ideal for dispersal events. By permuting the terminals, TreeFitter maximizes its sensitivity to dispersal and duplication events (Ronquist, 2003; Sanmartín and Ronquist, 2004), which are those predicted to occur in Cretaceous Laurasia and those preliminarily reported in the first part of this analysis above. Analysis of the output above suggested that dispersal was a driving force in many of the area hypotheses. Therefore, I tested the significance of the suggested hypotheses (Appendix H) and tested whether or not there was a signal detected for directional asymmetry in each of the area cladograms (e.g., a significantly higher frequency of dispersals from North America to Asia). To do so, I executed the dispersal hypotheses in TreeFitter and then randomized the terminals via 1,000 permutations to calculate how many of these events were significantly more abundant then expected by chance. Some of the hypotheses listed below were excluded from an individual analysis if the events described in the event set did not occur on that particular area cladogram.

117 103 In addition to testing the overall phylogeny of Ornithischia against the various hypotheses of landmass break-up, I also performed an analysis on individual time slices of the Mesozoic (Hunn and Upchurch, 2001; Upchurch and Hunn, 2002). Although ornithischians first appear in the Triassic Period, there are only two taxa (Pisanosaurus and Eocursor) that exist in the contiguous landmass of Pangaea and, thus, hypotheses of the processes of their biogeographic distribution is not necessary. However, seven time slices that coincide with Upchurch et al. s (2002) analysis contain a sufficient number of ornithischian taxa: Early Jurassic, Middle Jurassic, Late Jurassic, Early Cretaceous, middle Cretaceous (Aptian-Albian), Late Cretaceous, and the end-cretaceous boundary (Maastrichtian). As above, I calculated the optimal area cladogram for each time slice and tested the fit of each of the four events. Discussion The optimal area cladogram (OAC) for the biogeographic analysis of Ornithischia reveals several interesting relationships. An evaluation of 6,904 trees resulted in one most parsimonious reconstruction with a total cost of (Fig. 3.5). The OAC illustrates close relationships between Asia and Africa and between North America and Europe. The connection between North America and Europe (which, in turn, is more closely associated with Australia than to South America, Africa, or Asia) is driven predominantly by the distribution of ornithopods (Fig. 3.6). Basal ornithopods include the European Hypsilophodon and Callovosaurus, the North American taxa Orodromeus, Zephyrosaurus, Drinker, Tenontosaurus, and Dryosaurus, and include all Australian taxa

118 Figure 3.5 Optimal area cladogram for the comprehensive species-level phylogeny. The fit cost of this OAC is

119 Figure 3.6 Time-calibrated phylogeny of Ornithopoda. Stratigraphic ranges are noted for each terminal. Numerical dates given for the geological ages were obtained from Gradstein et al. (2004). 105

120 106 (with the exception of the ankylosaur Minmi) such as Atlascopcosaurus, Fulgurotherium, Leaellynasaura, Qantassaurus, and Muttaburrasaurus. The association between Asia and Africa (and collectively with South America), is likely driven by the early diversification of ornithischians such as Pisanosaurus from South America, followed by the basal split of African taxa: first, Eocursor, then the split of Genasauria into Thyreophora (Fig. 3.7) and Neornithischia (Fig. 3.8). The basalmost taxa in Neornithischia and Thyreophora are Stormbergia and Lesothosaurus, respectively. The strong connection with Asia likely stems from the associations of Agilisaurus, Hexinlusaurus, and Yandusaurus, all Asian taxa that are more closely related to Cerapoda than is Stormbergia. As noted in Chapter I, Lesothosaurus, despite being well-nested in Genasauria, is still labile with respect to the bases of Neornithischia and Thyreophora. The phylogeny presented here differs slightly from the topology presented in Chapter I. There, Lesothosaurus is recovered as the basalmost neornithischian (though, it takes only one additional step to move it to the base of Thyreophora) and here, as a result of a previous iteration of the data set presented in Chapter I, it is positioned at the base of Thyreophora. It may appear to be problematic, geographically, if Lesothosaurus is a basal thyreophoran given that Scutellosaurus is North American and Emausaurus and Scelidosaurus are European. However, a putative basal thyreophoran from the Tiourarén Formation of Niger (Middle Jurassic) supports the hypothesis that thyreophorans may have an early widespread distribution into North America, Europe, and Africa and Lesothosaurus could provide evidence with respect to a starting point for the ancestral range of the clade (Ridgwell and Sereno, 2011).

121 Figure 3.7 Time-calibrated phylogeny of basal ornithischians and Thyreophora. Stratigraphic ranges are noted for each terminal. Numerical dates given for the geological ages were obtained from Gradstein et al. (2004). 107

122 Figure 3.8 Time-calibrated phylogeny of basal neornithischians and Marginocephalia. Stratigraphic ranges are noted for each terminal. Numerical dates given for the geological ages were obtained from Gradstein et al. (2004). 108

123 109 The results shown in the OAC do not seem to follow closely the distribution of some derived clades like Ankylosauria, Ceratopsia, and Pachycephalosauria. Each clade is heavily dominated by North American and Asian taxa. This suggests that the nearly fully pectinate ornithopod phylogeny, containing numerous taxa that are distributed throughout Europe and North America (as well as some African taxa) at the base of Iguanodontia, up the tree through Asian and African taxa near the middle of the spine of the tree (base of Hadrosauroidea), to a mixture of North American, European, and Asian taxa (Fig. 3.6). Derived hadrosauroids (Saurolophidae) contain a mix of predominantly North American and Asian taxa, with a few European taxa and one lone Gondwanan species, South American Secernosaurus. Each of the eight alternative area cladograms (Table 3.1) was significantly more costly than the OAC (p<0.0405; Fig. 3.9). Vicariance signals among the nodes of the eight alternative cladograms were extremely low and dominated by dispersals. Duplication events and extirpations also played a role, though they were more heavily concentrated at the terminals for the analysis. Not all of the alternative area cladograms include relationships implied by the input dispersal event hypotheses (Appendix H). As such, some of the events could not be tested for significance. Indeed, all of the events that could be tested for were related to Laurasian communication. That is, dispersal routes between North America and Europe, North America and Asia, and Asia and Europe (Table 3.2). These dispersal routes are in contrast to Ezcurra and Agnolin (2012), who found that substantial communication between Africa and Laurasia existed, particularly via the island conglomerate that encompassed Europe throughout the majority of the Mesozoic.

124 Figure 3.9 Eight alternative area cladograms. H(1-8) refers to the cladograms detailed in Table 3.1. The cost assigned to each cladogram is given next to H(1-8). 110

125 111 Table Number of significant events on each of the eight alternative area cladograms. Alternative Hypothesis H1 H2 H3 H4 H5 H6 H7 H8 Dispersal Route p-value Asymmetrical dispersal from North America to Asia 0.78 Asymmetrical dispersal from Europe to Asia 0.59 Asymmetrical dispersal from Europe to North America 0.28 Asymmetrical dispersal from North America to Asia 0.74 Asymmetrical dispersal from Europe to Asia 0.79 Asymmetrical dispersal from Europe to North America 0.24 Asymmetrical dispersal from North America to Asia 0.77 Asymmetrical dispersal from Europe to Asia 0.91 Asymmetrical dispersal from Europe to North America 0.04 Asymmetrical dispersal from North America to Asia <0.01 Asymmetrical dispersal from Europe to Asia <0.01 Asymmetrical dispersal from Europe to North America <0.01 Asymmetrical dispersal from North America to Asia 0.62 Asymmetrical dispersal from Europe to Asia 0.97 Asymmetrical dispersal from Europe to North America 0.45 Asymmetrical dispersal from North America to Asia <0.01 Asymmetrical dispersal from Europe to Asia <0.01 Asymmetrical dispersal from Europe to North America <0.01 Asymmetrical dispersal from North America to Asia 0.21 Asymmetrical dispersal from Europe to Asia 0.38 Asymmetrical dispersal from Europe to North America 0.59 Asymmetrical dispersal from North America to Asia 0.74 Asymmetrical dispersal from Europe to Asia 0.48 Asymmetrical dispersal from Europe to North America 0.18 Note: Significance was determined based on p < H(1-8) refers to the area cladogram in Figure 3.9. Values in bold are significant.

126 112 This discrepancy can be explained by several possibilities. First, the data sets used between the current analysis and Ezcurra and Agnolin are entirely different. Ezcurra and Agnolin performed their analysis on a large sampling of Mesozoic archosaurs that spanned Laurasia and Gondwana nearly equally. They also incorporated Upchurch et al. s (2002) time-sliced TRA. Ezcurra and Agnolin s analysis represents a complete picture of faunal interchange among Cretaceous and early Paleogene archosaurs. The analysis here is focused solely on explanations for the distribution of ornithischian taxa throughout the Mesozoic. Of the events present in the original OAC of Upchurch et al. (2002; H1), none yielded any significant asymmetry. The same held true for alternative area cladograms H2, H5, H7, and H8. However, significant asymmetrical dispersal from North America to Europe was detected in H3 and all three signals were detected in alternative area cladogram H6. Furthermore, the current consensus hypothesis for Mesozoic continental fragmentation (H4) also had significant dispersal events for all three hypotheses, supporting the hypotheses put forward by both qualitative and quantitative assessments (e.g., Sereno, 1997, 1999; Upchurch et al., 2002; Holtz et al., 2004, Ezcurra and Agnolin, 2012) that dispersals play a much larger role in shaping Mesozoic biogeographic patterns. Time-Sliced Analyses Reticulate evolution, Ronquist s (2002) duplication events, can be a confounding problem, even when assigning costs to such an event. An alternative to testing the presence of duplication events across the entire phylogeny of Ornithischia is to parse out any signal among temporally-segregated bins that preserve ornithischian

127 113 diversity (Hunn and Upchurch, 2001; Upchurch and Hunn, 2002). Upchurch et al. (2002) selected several time bins based on geologically-significant temporal events. In addition to the Late, Middle, and Early Jurassic time slices, the Early and Late Cretaceous slices, I also incorporated a middle Cretaceous (Aptian-Albian) time slice because of the proposed separation of Africa from other Gondwanan landmasses and communication with Laurasia (e.g., Ezcurra and Agnolin, 2012). Also added to the analysis was a time slice of the latest Cretaceous (Maastrichtian) prior to the end-cretaceous mass extinction based on analyses of regional biogeographic patterns being influenced primarily by the global marine regression (e.g., Sampson et al., 2010). The OACs for the seven major time slices, like the OAC for the entire clade throughout the Mesozoic, yielded no statistically significant events other than a dispersal event between Africa and North America + Europe in the Early Jurassic (p<0.03; Fig. 3.10). In the Early Jurassic, there are eight taxa (heterodontosaurids, basal thyreophorans, and basal genasaurians) across three modern-day continents (Africa, Europe, and North America). North American and European taxa form a derived clade with respect to the African taxa and correlates well when compared to the favored geological hypothesis (i.e., North America and Europe are more closely associated with each than with Africa; Smith et al., 1994). The Middle Jurassic consists of eight species, including heterodontosaurids, stegosaurs, and basal neornithischians, and three continents, including South America, Europe, and Asia (Fig. 3.11). The lack of significant events for this time slice is likely due to South American Manidens, in this phylogeny, the sister taxon to Asian

128 Figure 3.10 Early Jurassic distribution of ornithischians and the single OAC that best explains the biogeographic pattern. The cost of the OAC is Only one significant event was determined in this time slice a dispersal from Africa to North America + Europe (p<0.03). 114

129 Figure 3.11 Middle Jurassic distribution of ornithischians and the single OAC that best explains the biogeographic pattern. The cost of the OAC is There were no significant events found. Manidens is the only species from the Middle Jurassic that is found in the southern continents. 115

130 116 Tianyulong. It is the only southern continent species in the time slice and is not positioned outside a monophyletic group consisting of all Laurasian taxa. The Late Jurassic consists of eighteen species mainly stegosaurs and basal cerapodans across four continents, including all Laurasian continents plus Africa (Fig. 3.12). The lack of statistical significance in this time slice, like the Middle Jurassic, is due in part to the several African taxa and their relationships to taxa in all three Laurasian continents. The two OACs illustrate that communication between Africa and Laurasia may have begun earlier than previously thought (Ezcurra and Agnolin, 2012). In the Early Cretaceous, 25 species across five continents also yielded no statistically significant events (Fig. 3.13). The intercalation of Gondwanan taxa (African and Australian) suggest that the undersampled diversity of ornithischians is driving the lack of statistical significance for this time bin. The Aptian and Albian in the middle of the Cretaceous detail the broadest distribution of ornithischians, occurring on six continents (three Laurasian and three Gondwanan). The only differences between the two competing OACs, though, are the positions of South America and Asia (3.14). With Africa nested with Europe, it further corroborates Ezcurra and Agnolin s (2012) conclusion that Africa and Laurasia maintained a fairly well established connection through the Cretaceous. More than 70 taxa across four continents (three Laurasian continents, South America) constitute ornithischian diversity in the Late Cretaceous making it the most species-rich time bin for Ornithischia (Fig. 3.15). Most of the taxa belong to either Ceratopsia or Hadrosauroidea, but all other major clades have representatives. The South American taxa are intercalated throughout the basal positions of the tree, confusing the

131 Figure 3.12 Late Jurassic distribution of ornithischians and the two OACs that best explain the biogeographic pattern. The cost of the OACs is There were no significant events found. Kentrosaurus, Paranthodon, and Dysalotosaurus, all African taxa, confound the biogeographic pattern. 117

132 Figure 3.13 Early Cretaceous distribution of ornithischians and the single OAC that best explains the biogeographic pattern. The cost of the OAC is There were no significant events found. 118

133 Figure 3.14 Middle Cretaceous (Aptian-Albian) distribution of ornithischians and the two OACs that best explain the biogeographic pattern. The cost of the OACs is There were no significant events found; however, this time slice shows the greatest amount of geographic distribution among ornithischians than at any other time in the Mesozoic. 119

134 Figure 3.15 Late Cretaceous distribution of ornithischians and the single OAC that best explains the biogeographic pattern. The cost of the OAC is There were no significant events found. The Late Cretaceous exhibits the greatest amount of species richness; however, the overwhelming amount of taxa are known from North America and Asia. 120

135 121 statistical signal for this time bin. Interestingly, the South American taxa are all associated with North American species. Nearly 40 species, with representatives from most major clades of Ornithischia that appear in the Cretaceous, across four continents all three Laurasian continents and South America makeup the total ornithischian biodiversity at the end of the Cretaceous (Fig. 3.16). Only one taxon, Secernosaurus, is located in South America, whereas, all others are from Laurasia. There appears to be two competing hypotheses for this time bin between differing clades, which is why there is a lack of statistical support. Ankylosaurs and marginocephalians all appear to be dispersing between Asia and North America and the hadrosauroids all appear to be dispersing between Europe, Asia, and North America. Conclusion The time-calibrated phylogeny of Ornithischia, OAC, alternative area hypotheses, and time-sliced OACs yield several striking trends. Vicariance, except on very local scales and at very fine geographical resolution, is nearly absent from the large-scale biogeographic analysis. This is expected as the less speciose derived ornithischian clades offer less complicated patterns of biogeographic distributions. However, there are exceptions within these clades. For example, the presence of the ankylosaur Minmi in the Early Cretaceous of Australia when the rest of the clade is found in Laurasia requires an ad hoc hypothesis of dispersal to Australia from Laurasia. Alternatively, this could suggest that a relict population of basal ankylosaurs (or basal thyreophorans) has yet to be described from Gondwanan localities (e.g., Ridgwell and Sereno, 2011).

136 Figure 3.16 End-Cretaceous (Maastrichtian) distribution of ornithischians and the three OACs that best explain the biogeographic pattern. The cost of the OACs is There were no significant events found. Most major clades are represented in the Maastrichtian and all but one species, Secernosaurus, are known from Laurasian continents. 122

137 123 A similar scenario also presents itself with stem- and basal iguanodontians. The Australian taxa, whose geographical distribution in association with their phylogenetic positions, imply multiple ad hoc hypotheses of dispersal from Laurasia to Australia or, as above, suggest the presence of relict Gondwanan populations of basal cerapodans yet to be described. Given that ornithischians are obligately terrestrial animals, crossing large expanses of water to establish founder populations on distant islands and/or continents is unlikely, though by no means improbable. Another possibility is that the phylogeny is incorrect. While most phylogenies, particularly those that are composed of extinct taxa, are rarely static, the likelihood that a radically different phylogeny would illuminate vicariant scenarios of distribution is expected to be low. More likely, however, is the presence of undescribed relict populations of ornithischians.

138 124 CHAPTER IV ORNITHISCHIAN PHYLOGENY AND ITS BROADER IMPACT ON PHYLOGENY ESTIMATION, CHARACTER EVOLUTION, AND BIOGEOGRAPHY Introduction The utility and application of phylogenetic techniques in evolutionary biology is wide-ranging and diverse. Despite advances in our approaches to phylogeny estimation and biogeographical reconstruction, issues regarding character evolution and dispersal and vicariance patterns remain. Fossil groups provide an insight into these issues due to varying degrees of missing data, conflicting character data from different published analyses, and unknown or uncorroborated biogeographical patterns. Ornithischia, one of the two major groups of dinosaurs, is itself an exemplar taxon with which to evaluate approaches to phylogeny estimation, character evolution, and biogeographical reconstruction because of labile taxa within the clade, differing character codings among published data matrices, and problematic or complicated biogeographical patterns. Large-scale, comprehensive phylogenies for all of Ornithischia including both basal and derived taxa were lacking, preventing the testability of these issues. Because of the lack of a well-resolved species-level phylogeny including basal and derived members, our understanding of the initial ornithischian radiation in the Late Triassic and Early Jurassic was unclear, further complicating our ability to test biogeographic scenarios.

139 125 Phylogenetic Analysis Poor resolution and the general lack of statistical support across the ornithischian phylogeny underscored the need for greater taxon and character sampling to illuminate the phylogenetic positions of problematic, labile taxa. The data set assembled here constitutes one of the largest species-level data sets assembled for Ornithischia and is based on comprehensive empirical assessments of a wide sample of ornithischians with an emphasis on basal taxa, helping to resolve the relationships of historically labile taxa. Heterodontosauridae is a monophyletic group outside of Genasauria, corroborating recent studies, though the current analysis did not recover distinct Laurasian and Gondwanan clades as has been previously reported. Lesothosaurus, traditionally considered as one of the most basal ornithischians, is here recovered as the basalmost neornithischian more closely related to cerapodans than to thyreophorans. The resultant phylogeny shows good resolution among basal taxa; however, most members of Cerapoda collapse into a polytomy recovering only a monophyletic Iguanodontia, Marginocephalia, and a few other derived clades. The phylogenetic placement of these taxa is critical for assessments of character evolution for more derived clades and ancestral state reconstructions for Ornithischia in general, as well as determining areas of origination for clades across the tree. More broadly, though, the phylogenetic analysis based on comprehensive taxonomic sampling, including both basal and derived species, provides the potential backbone for numerous studies on, e.g., the evolution of herbivory at a time when terrestrial ecosystems were changing from a gymnosperm-dominated environment to one

140 126 that saw the rise and global expansion of angiosperms (e.g., Butler et al., 2009a, b). The coincident rise in ornithischian diversity in the middle Cretaceous with flowering plants offers an interesting test case in coevolutionary hypotheses. But most importantly, ornithischians achieved a global distribution during a time of fundamental changes to global continental configuration and climate. Ancestral Body Plan and Character Evolution The hypothetical ancestral ornithischian was a bipedal cursor that had a large, grasping manus that may have been useful in facilitating a potentially omnivorous diet, perhaps for fossoriality (Fig. 4.1). Given that most phylogenetic analyses of large subclades within Ornithischia use Lesothosaurus as an outgroup, the reconstructed ancestral ornithischian here indicates that some characters, particularly for the forelimb, may not be appropriate when determining character polarity. More broadly, whereas the standard for assessing morphological character evolution is to evaluate the accelerated and delayed transformations for all characters on a single most parsimonious tree, I have demonstrated here that reconstructing ancestral character states may be a more suitable alternative (see also O Leary et al., 2013). With this approach, a more accurate estimation of character evolution can be assessed by determining the characters more likely to change and at which nodes, allowing evaluation among all most parsimonious trees. Furthermore, comparisons of the hypothetical ancestral taxon of a given clade could provide information on the utility and accuracy of

141 Figure 4.1 Complete reconstruction of the ancestral hypothetical ornithischian dinosaur. 127