THE UNIVERSITY OF CALGARY

Skull Morphology of the Ankylosauria

by

Matthew K. Vickaryous

A THESIS

SUBMITED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

MASTER OF SCIENCE

DEPARTMENT OF BIOLOGICAL SCIENCES

CALGARY, ALBERTA

JANUARY, 2001

O Matthew K. Vickaryous 2001 National Library Bibliothéque nationale I*l of Canada du Canada Acquisitions and Acquisitions et Bibliographie Services services bibliographiques 395 Wellington Street 395, rue Wellington Ottawa ON KIA ON4 Ottawa ON KIA ON4 Canada Canada Your Rie Votre rrlftimce

Our fite Notre dfbrance

The author has granted a non- L'auteur a accordé une licence non exclusive licence allowing the exclusive permettant à la National Library of Canada to Bibliothèque nationale du Canada de reproduce, loan, distribute or sell reproduire, prêter, distribuer ou copies of this thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la forme de microfiche/film, de reproduction sur papier ou sur format électronique.

The author retains ownership of the L'auteur conserve la propriété du copyright in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substantial extracts fiom it Ni la thèse ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation. Abstract

The vertebrate head skeleton is a fundamental source of biological information for the study of both modern and extinct taxa. Detailed analysis of structural modifications in one taxon frequently identifies developmental and / or functional features widespread amongst a more inclusive clade of organisms. Members of the Ankylosauria, a group of armoured, herbivorous dinosaurs, provide an opportunity to review innovative cranial architecture of a fossil taxon. The generalized skull condition of the Ankylosauria is surveyed, providing the first description of the basicranium, and a reappraisal of the palate and mandible. The previous identification of neomorphic cranial elements, such as "tabulars" and "prevomers", is not supported. Osteological terminology used to describe the Ankylosauria is reviewed in connection with the rnorphological overview. This descriptive evaluation establishes the foundation for subsequent detailed investigations of selected taxa. Euoplocephalus tutus Lambe (1 902),frequently regarded as the archetypal ankylosaur, is characterized by an akinetic craniurn with a distinctive pattern of cranial sculpturing, the presence of a modified "ciliary" supraorbital, and relatively small, fluted teeth. A new taon of ankylosaur, "Gobisaurus" gen. et sp. nov., is characterized by a narrow premaxillary rostrum, a prominent orbit, an enlarged nasal vestibule and an elongated rostral process of the vomer. The developmental processes giving rise to cranial ornarnentation are reviewed using the comparative approach, identifying two discrete rnechanisms that operate independently. A cladistic analysis of 22 ankylosaur taxa is conducted, thereby establishing a phylogenetic position for "Gobisaurus". The results of the cladistic analysis are compared with a consensus estimate of previously published hypotheses using a supertree analysis. The cladogram - supertree cornparison illustrates the importance of detailed cranial information in effectively clustering members of the Ankylosauria. ACKNOWLEDGEMENTS

An enormous number of people played important roles throughout the course of this research, and many of them deserve special recognition. 1 would like to thank the members of my examination committee, Drs. A.P. Russell, P.J. Currie, M.A. Katzenberg, G. Pritchard and H. 1. Rosenberg, for carefully scrutinizing the penultimate product of my Master's research. My supervisor Dr. A.P. Russell has been particularly patient and longsuffering. Many thanks. Drs. A.P. Russell, P.J. Currie, H.I. Rosenberg, M.A. Katzenberg, B. Hallgrimsson, D.B. Brinkman (TMP) and LM. Witmer (Ohio University, Athens) al1 helped to mould, shape and pervert my knowledge of skeletal anatomy. Mr. W.D. Fitch kindly permitted me to play with, cut apart and yammer on about many of the fine specimens in his care at the University of Calgary, Museum of Zoology. 1 have no doubt benefited from the many frank and frequently perturbing conversations with my cohorts J.M. Lavigne (Talisman Energy, Calgary), M. Getty (Utah Museum of Natural History, Salt Lake City), K. Kucher (TMP), and L. Claessens (Harvard University, Cambridge). J.T. Lumb (Metafore, Calgary) and N. Rybczynski (Duke University, Durham) are to be congratulated not only for their years of polite discourse and encouragement, but also for their burliness in excavating ankylosaurs in the dark. My colleagues in the Vertebrate Morphology and Palaeontology Research Group have al1 contributed significantly to my current outlook. Thanks to G.L. Powell, L. McGregor, E. Snively, M.J. Ryan, M. Thompson, P.J. Bergmann, J. Peng, A.M. Gutierrez and T. Higham. Specimens studied during the course of this research came from a variety of institutions, and special thanks is owed to the following museum curators and collection managers: A. Newman, J.D. Gardner, J. Wilke, M. Laframbois and L. Cook (TMP); R.C. Fox (UAVLP); M.A. Norrell (AMNH); K. Carpenter (DMNH); K. Shepard (CMN); and Y. Kobayashi (FPDM). I also owe thanks to the staff of Prehistoric Animal Structures (East Coulee, Alberta) and Canada Fossils (Calgary, Alberta) for keeping me posted on ankylosaurs moving through their workspaces. Dr. R. Etheridge (San Diego State University) provided some crucial information iv about the development of ornamentation in squamates and supplementary information about ankylosaur material was gleaned through conversations with Drs. K. Carpenter (DMNH), J.I. Kirkland (Utah Geological Survey, Salt Lake City) and Mr. P. Penkalski (North Carolina State, Raleigh). The gurus of systematics, Drs. O.R.P. Bininda-Emonds (Leiden University, Leiden, The Netherlands) and S. Modesto (ROM),deserve special credit for permitting me to bother them so frequently. Thanks-eh! Ankylosaur cranial rnaterial was thin-sectioned with unparalleled skill by Mr. J. Resultay (University of Calgary, Department of Geology and Geophysics). Preparation of ankylosaur material at the TMP during the course of this research was conducted with prudence and finesse by Ms. W. Sloboda and Messrs. K. Kucher and M. Mitchell. Ms. D. Sloan (TMP), Dr. P. Johnston (TMP), Mr. G. Campbell and Mr. A.L. Vickaryous assisted with much of the photography and R. Humphries, L. Curtis and L. Morris performed the CT scanning at the Foothills Hospital, Calgary. This research was supported by the Duerksen Mernorial Scholarship, a Jurassic Foundation Grant, two Heaton Student Support Grants, a University of Calgary Graduate Scholarship, a University of Calgary Thesis I Dissertation Research Grant, an NSERC Operating Grant awarded to Dr. A.P. Russell and the infinite patience of J.T. Lumb. TABLE OF CONTENTS ... Abstract ...... III Acknowledgements ...... iv Table of Contents ...... vi List of Tables ...... xii... List of Figures...... xiii List Of ~ymbols.Abbreviations. And Nomenclature ...... Chapter 1. Introduction ...... 1.1 Background Rationale...... 1.2 The Ankylosauria ...... 1.2.1 Euoplocephalus tutus Lambe (1902) .. 1.2.2 " Gobisaurus" gen. et sp . nov...... 4

Chapter 2. Fossil Materials and Methods ...... 7 2.1 Specimens ...... 7 2.1 .1 Post-mortem Deformation ...... 9 2.1 -2 Euoplocephalus;A Question of Synonymy...... IO 2.2 Methodology ...... II

Chapter 3. Ankylosaur Head Skeleton Morphology ...... 15 3.1 Overall Morphology ...... -15 3.2 Ossa Cranii ...... 17 3.2.1 Rostral Region ...... 17 3.2.1 .1 Os premaxillare ...... -18 3.2.1.2 Os maxillare (rostrodorsal secondary palate) .. -20 3.2.1.3 Osnasale ...... 23 3.2.1.4 Os prefmntale ...... -24 3.2.1 .5 Os lacrimale ...... -25 3.2.2 Temporal Region...... -26 3.2.2.1 Ossa supraorbitalia (processus cornuum supraonbitalia) ...... 26 3.2.2.2 Os postorbitale (lamina postocularis) ...... 28 3.2.2.3 Os jugale (arcusjugale) ...... 30 3.2.2.4 Os frontale...... 31 3.2.2.5 Os parietale (margo nuchae)...... 32 3.2.2.6 Os squamosum (processus cornuum squamosi) ...... 33 3.2.2.7 Os quadratojugale (processus cornuum quadratojugalis) ...... 35 3.2.2.8 "Os tabulare" ...... 36 3.2.3 Palatal Region ...... 37 3.2.3.1 Ossa vomera (sepfum ca vitas internasalis)...... 37 vi

4.2.2.7 Os quadrafojugale (processus cornuum quadratojugalis)...... 91 4.2.3 Palatal Region ...... 91 4.2.3.1 Ossa vornera (septum cavitas internasalis)...... 91 4.2.3.2 Os palatinum (caudoventral secondary palate) 92 4.2.3.3 Os pterygoideum ...... 92 4.2.3.4 Os ectopterygoideum...... -93 4.2.4 Occipital / 8asicranial Region...... 93 4.2.4.1 Os supraoccipitale...... 93 4.2.4.2 Os exoccipitale ...... -94 4.2.4.3 Os basioccipitale ...... 94 4.2.4.4 Ossa ofica ...... 94 4.2.4.5 Os basisphenoidale ...... 95 4.2.4.6 Os laferosphenoidale ...... 95 4.2.4.7 Os parasphenoidale ...... 95 4.2.4.8 Ossa interorbifale ...... 95 4.2.4.9 Os quadrafum ...... 96 4.3 Ossa Mandibulae ...... 96 4.3.1 Mandibular Region ...... 96 4.3.1 .1 Os predentale...... 97 4.3.1.2 Os dentale ...... 97 4.3.1 .3 Os spleniale ...... 97 4.3.1.4 Osangulare...... 98 4.3.1 .5 Os supra-angulare ...... 98 4.3.1.6 Os coronoideurn...... 98 4.3.1 .7 Os prearficulare ...... 98 4.3.1 .8 Os atticulare...... 98 4.4 Dentes ...... ~...... 99 4.5 Surnmary ...... 99

Chapter 5. Morphological Description of the Cranium of "Gobisaurus" gen. et sp . nov...... 137 5.1 Overall Cranial.. Morphology ...... 137 5.2 Ossa Cranil ...... 139 5.2.1 Rostral region ...... 1 39 5.2.1 .1 Os premaxillare ...... 139 5.2.1.2 Os maxillare (rostrodorsal secondary palate). 140 5.2.2 Temporal Region ...... 141 5.2.2.1 Ossa supraorbitalia. . (processus cornuum supraoh~talra)...... 141 5.2.2.2 Os postorbitale (lamina postocularis) ...... 142 5.2.2.3 Os jugale (arcusjugale) ...... 142 5.2.2.4 Os panefale (margo nuchae)...... 142 5.2.2.5 Os squamosum @rocessus cornuum squamosi) ...... 142 viii 5.2.2.6 Os quadratojugale (processus cornuum quadrafojugalis)...... -143 5.2.3 Palatal Region ...... 143 5.2.3.1 Ossa vomera (septum ca vitas internasalis).... 143 5.2.3.2 Os palatinum (caudoventral secondary palate) ...... 144 5.2.3.3 Os pterygoideom ...... 144 5.2.3.4 Os ectopferygoideum ...... 145 5.2.4 Occipital 1 Basicranial Region ...... 145 5.2.4.1 Os supraoccipitale ...... 145 5.2.4.2 Os exoccipitale ...... 145 5.2.4.3 Os basioccipitale ...... 146 5.2.4.4 Os basisphenoidale ...... 146 5.2.4.5 Os quadratum ...... 146 Denfes ...... 147 Surnrnary ...... 147 Chapter 6. Cranial Omamentation ...... 163 6.1 The Comparative Method...... 163 6.1. 1 Phylogenetic lnference ...... 165 6.1.2 Ahistorical Extrapolatory Modeling ...... 166 Methods and Materials...... 167 Hypothesis 1 : Co.ossification ...... 169 6.3.1 Development of Cranial Ornamentation in Extant . Scleroglossans ...... 169 6.3.2 Osteological Correlates ...... 170 6.3.3 The Hypothesis ...... 171 6.3.4 Testing the Hypothesis...... 171 6.3.5 Conclusions ...... 175 Hypothesis 2: Elaboration ...... 175 6.4.1 Development of Cranial Ornamentation in Extant lguanians ...... 176 6.4.2 Osteological Correlates ...... 177 6.4.3 The Hypothesis ...... 177 6.4.4 Testing the Hypothesis ...... 177 6.4.5 Conclusions ...... 178 A Synthetic Approach ...... 178

Chapter 7. Phylogenetic Analysis of Ankylosaur Crania .....m....m..mmmmmm .... 190 7.1 Cladistic Analysis ...... 191 7.1.1 Character Analysis ...... 192 7 11 Characters of the dorsum of the craniurn and overall cranial morphology...... 193 7.1 .1.2 Characters of the palate ...... 195 7.1.1.3 Characters of the respiratory passageways ... 196 7.1.1.4 Characters of the pterygoid cornplex. occiput and basicranium ...... 197 7.1 .1.5 Characters of the quadrate ...... 198 7.1 .1.6 Other characters ...... 198 7.1.1.7 Characters not used here but that have been used or discussed elsewhere ...... 199 7.1.2 Cladistic Results and Discussion...... 199 7.1.2.1 Cladistic analysis of al1 taxa under consideration...... 199 7.1.2.2 Cladistic analysis of al1 taxa coded for greater than 55% of the characters ...... 199 7.1.2.3 Cladistic analysis of all taxa coded for greater than 80% of the characters ...... 201 7.1.2.4 Cladogram cornparison ...... 201 7.2 Systematics...... -201 7.3 Reference Hypothesis - The Ankylosaur Supertree ...... 206 7.3.1 Supertree Methodology ...... -207 7.3.2 Supertree Estimate...... -208 7.3.3 Supertree - Cladistic Comparison ...... 208 7.3.3.1 Supertree - figure 7.1 comparison ...... 209 7.3.3.2 Supertree - figure 7.2 cornparison ...... 209 7.3.3.3 Supertree - figure 7.3 comparison ...... 209 7.3.3.4 Supertree cornparison ...... 210 7.3.3.5 Supertree conclusion ...... 210 7.4 Concluding Comments ...... 210

Chapter 8. Summary and Prospectus...... 233 8.1 The Skull ...... 233 8.2 The Ankylosauria ...... 233 8.3 EuplocephalustutusLarnbe(1902) ...... 235 8.4 "Gobisaurus" gen. et sp . nov...... 236 8.5 Ornamentation Development ...... 237 8.6 Systematics...... -238 8.7 Prospectus ...... 239 Literature Cited ...... -1 Appendix 2.1. Nomenclatural abbreviations...... 257

Appendix 4.1. Skull measurements for seleded specimens of Euoplocephalus and "Gobisaurus" referred to in the text ...... 262

Appendix 6.1. Extant material examined. including information on specimen type...... 264 Appendix 7.1. Character state distribution among the 24 taxa considered in this study (including two outgroups)...... 266

Appendix 7.2. Character state distribution for the supertree analysis of 14 previously published estimates and three cladograms generated from the evaluation of strictly cranial data (figures 7.1 - 7.3)...... 267 List of Tables

Table 3.1. Element composition of each topographic region of the skull and a survey of element distribution amongst rnembers of the Archosauna...... 78

Table 3.2. Individual tooth counts for the various dentigerous elements of the taxa

discussed in the text...... , , ...... 80

Table 4.1. Skull material referred to the taxon EuopIocephalus tutus presently accessioned in museurn collections...... 134

Table 4.2. Additional fossil material examined...... 136

Table 7.1. Skuil material examined during the course of this study...... 230

Table 7.2. Source of original data for the supertree analysis (including the cladistic analyses), and the number of characters therein provided...... 232

xii LIST of FIGURES

Figure 2.1. Lateral profile of three specimens of Euoplocephalus tutus Lambe

(1902)...... , ...... 14

Figure 3.1. Schematic of the Euoplocephalus cranium illustrating the rostral region, and al1 the purported constituent elements thereof...... 59

Figure 3.2. Schematic of the Euoplocephalus craniurn illustrating the temporal region, and al1 the purported constituent elements thereof...... 61

Figure 3.3. Schematic of the Euoplocephalus cranium illustrating the palatal region, and al1 the purported constituent elements thereof...... 63

Figure 3.4. Schematic of the Euoplocephalus cranium illustrating the occipital 1 basicranial region, and al1 the purported constituent elements thereof...... 65

Figure 3.5. Schernatic of the Euoplocephalus basicranium, right lateral view. 67

Figure 3.6. Schematic Euoplocephalus mandible illustrating the mandibular region...... 69

Figure 3.7. Schematic illustrations of the right lateral profile of ankylosaur crania discussed in the text...... 71

Figure 3.8. Schematic illustrations of the dorsal profile of the ankylosaur wania discussed in the text...... 73

Figure 3.9. Schematic illustrations of the ventral view of Pinacosaurus grangen...... 75 xiii Figure 3.10. Schematic illustrations of the occipital view of ankylosaur crania discussed in the text...... 77

Figure 4.1. Euoplocephalus tutus Lambe (1902) holotype cranium (NMC021 0) ...... ~...... 103

Figure 4.2. Euoplocephalus tutus Lambe (1902) craniurn (TMP 91.1 27.1 ). ... 105

Figure 4.3. Cranial sculpturing of Euoplocephalus tutus Lambe (1 902)...... 107

Figure 4.4. Schematic rostrodorsal profile of Euoplocephalus tutus Lambe (1902)...... 109

Figure 4.5. Schematic of the right lateral profile of the cranium of Euoplocephalus tutus Lambe (1902)...... 11 1

Figure 4.6. Euoplocephalus tutus Lambe (1902) cranium with truncated palate, ventral view (TMP 97.59.1)...... 11 3

Figure 4.7. Euoplocephalus tutus Lambe (1902) cranium, taphonomically broken (TMP 96.75.1)...... 115

Figure 4.8. Euoploœphalus tutus Lambe (1902) cranium, ventral view of palate, original and interpretive illustration (TMP 97.132.1 )...... 117

Figure 4.9. Transverse sections through the rostral region of Euoplocephalus tutus Lambe (1902). interpretive illustrations (AMNH 5405)...... 119

Figure 4.1 0. Euoplocephalus tutus Lambe (1902) dentition, in situ and isolated. 121 xiv Figure 4.1 1. Euoplocephalus tutus Lambe (1902) craniun, ventral view of occipital 1 basicranial region, original and interpretive illustration (TMP 97.1 32.1 )...... 123

Figure 4.1 2. Euoplocephalus tutus Lambe (1 902) cranium, occipital view of occipital I basicranial region, original and interpretive illustration (TMP 91.127.1 )...... 125

Figure 4.1 3. lsolated cranial elements referred to Euoplocephalus...... 127

Figure 4.14. Euoplocephalus tutus Lambe (1 902) basicranium, original and interpretive illustration (FDMJ-V-31)...... 129

Figure 4.1 5. Euoplocephalus tutus Lambe (1902) mandible, missing coronoid but including predentary (AMNH 5405)...... 131

Figure 4.16. Euoplocephalus tutus Lambe (1902) mandible including coronoid, but missing predentary (UALVP 31)...... 133

Figure 5.1. "Gobisaurusngen. et sp. nov. holotype cranium (IVPP V 12563). 150

Figure 5.2. Schematic of the lateral profile of the cranium of "Gobisaurus" gen. et sp. nov...... 152

Figure 5.3. Transverse sections through the rostral region of "Gobisaurus" gen. et sp. nov., interpretive illustrations (IVPP VI2563) ...... 154

Figure 5.4. "Gobisaurus* gen. et sp. nov. cranium, temporal region and dentition (IVPP VI2563)...... 156 Figure 5.5. "Gobisaurus" gen. et sp. nov. craniurn, ventral view of palate, original and interpretive illustration (IVPP V 12563)...... 158

Figure 5.6 "Gobisaurus" gen. et sp. nov. cranium, ventral view of the occipital 1 basicranial region, original and interpretive illustration (IVPP VI2563)...... 160

Figure 5.7. "Gobisaurusngen. et sp. nov. cranium, occipital view of the occiput, interpretive illustration (IVPP VI2563)...... 162

Figure 6.1. Phylogenetic inference and the "extant bracketn...... 181

Figure 6.2. Modified phylogeny of the Amniota, providing a phylogenetic framework for the taxa discussed in the text...... 183

Figure 6.3. Heloderma suspectum subadult specimen, radiographic image illustrating the pervasive development of osteoderms...... 185

Figure 6.4. cf. Euoplocephalus thin section from an isolated fragment of cranium representing the rostral region...... 187

Figure 6.5. Phrynosoma modestum neonate, cleared and double-stained, dorsal view...... 189

Figure 7.1. A strict consensus tree of the 20,065 most parsimonious cladograms (126 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence...... 213

Figure 7.2. A strict consensus tree of the 54 rnost parsimonious cladograms (120 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence produced by eliminating Animantam. Nodocephalosaurus, Sauropelfa. Shanxia and "Sfruthiosaurus" ...... 215

Figure 7.3. A strict consensus tree of the 6 most parsimonious cladograms (114 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence produced by eliminating Minmi and Talarurus in addition to those removed from figure 7.2...... 217

Figure 7.4. Schematic map of China and Mongolia illustrating the approximate locality of "Gobisaurus"gen . et sp . nov ...... 219

Figure 7.5. Schematic map illustrating the approximate locality of the type specimen of Euoplocephalus tutus Lambe (1 902)...... 221

Figure 7.6. The supertree methodology...... 223

Figure 7.7. Supertree - figure 7.1 comparison...... 225

Figure 7.8. Supertree - figure 7.2 cornparison...... 227

Figure 7.9. Supertree - figure 7.3 comparison...... 229 INSTITUTIONAL ABBREVIATIONS

AMNH, American Museum of Natural History, New York, New York, United States of America; ARPC, Anthony P. Russell (personal collection); BM(NH), British Museum (Natural History), London, England; CEUM, College of Eastern Utah, Price, Utah, United States of America; DMNH, Denver Museum of Natural History, Denver, Colorado, United States of America; FPDM-V, Fukui Prefectura' Dinosaur Museum (Vertebrate collection), Katsuyama, Japan; GI, Geological Institute, Section of Paleontology, Ulaanbaatar, Mongolia; IVPP, lnstitute of Vertebrate Paleontology and Paleoanthropology, Beijing, People's Republic of China; LACM, Natural History Museum of Los Angeles County, Los Angeles, California, United States of America; MOR, Museum of the Rockies, Bozeman, Montana, United States of America; NMC, Canadian Museum of Nature, Paleobiology Division (National Museum of Canada), Ottawa, Ontario; PBPC, Philip J. Bergmann (personal collection); ROM, Royal Ontario Museum, Toronto, Ontario; SMU, Shuler Museum of Paleontology, Dallas, Texas, United States of America; TMP, Royal Tyrrell Museum of Palaeontology, Drumheller, Alberta; UALVP, University of Alberta Laboratory of Vertebrate Palaeontology, Department of Geology, Edmonton, Alberta; UCMZ, University of Calgary Museum of Zoology, Calgary, Alberta; USNM, National Museum of Natural History (formerly the United States National Museum) Washington, D.C., United States of America; ZPal, Paleobiological lnstitute of the Polish Academy of Sciences, Warsaw, Poland.

OTHER ABBREVIATIONS SVL, snout - vent length, in millimetres Chapter 1 Introduction

11 Background Rationale

In the preface to his treatise The Development of the Verfebrate Skull, de Beer (1937) wrote: ".. . I have been dismayed at the lack of CO-ordination displayed by the numerous workers in this field, leading as it has to a most confused and redundant nomenclature, and to neglect of a number of morphological principles and features of interest which escaped recognition largely because the objects of cornparison bore different names." Although taken slightly out of context, de Beer's astute criticism highlights an ever present problem facing comparative biologists -- the recognition and identification of patterns of similarity, herein summarized by the somewhat nebulous concept of homology (sensu Hall, 1994). The vertebrate head skeleton appears to be particularly susceptible to such nomenclatural conundrums. Primitively assisting with the physical acquisition and processing of food, the maintenance and control of respiratory flow and protection and support of the central nervous system and special sense organs (Russell and Thornason, 1993))the skull is both functionally and structurally an integrated by-product of evolutionary consequence. Consider the broad cranial diversity exhibited within even finite clades such as extant eutherians - from the minimalistic (e.g. mysticete whales) to the inflated (e.g. elephants); from the foreshortened (e.g. many microchiropteran bats) to the protracted (e-g. anteaters). The loss and 1 or morphological alteration 1 coalescence of individual elements within an ever changing developmental timeframe combine to obfuscate the determination of homologues. The problem is particularly acute for vertebrate palaeontologists, where the material is generally scant and fragmentary, and may reflect an aberrant morphology with no modern analogues. The curent effort attempts to address this issue, on a limited scale, by morphologically reviewing, in detailed, consistent fashion using multiple specimens, the head skeleton of an unusual fossil taxon. Following Witmer's (1995) notion of an "inverted pyramid of inference", a comprehensive understanding of skeletal anatomy is a precept to al1 future endeavours dealing with fossil vertebrates (e.g. soft-tissue inferences, functional morphology, behaviour, community structure and ecology). A central tenet to such a review is the notion that detailed analyses of innovative structural modifications in one group of organisms can frequently indicate developmental and 1 or functional features widespread, albeit less noticeably, amongst a more inclusive group of organisms. In other words: "1 hope, therefore, that this work may be consulted with profit by vertebrate morphologists, palaeontologists, and human anatomists for the purposes of both study and research" (de Beer, 1937, p. vii). The purpose of this contribution is two-fold: (1) to establish a fundamental osteological framework to form the foundation of future palaeobiological inferences (sensu Witmer, 1995; 1997); and (2) to derive a practical, systematic methodology for approaching morphological descriptions within an unusual clade of fossil vertebrates, including the adoption of a pragmatic, uniform nomenclature. This contribution is arranged sequentially, building the framework, beginning with a descriptive review of the generalized skull condition for the clade under consideration (essentially a synthesis of information pertinent to all rnembers; Chapter 3). Subsequent ckiapters document a comparison of the generalized pattern with that of a genus previously described from several specimens (Chapter 4), and then with that of a previously undescribed taxon, known from a single specimen (Chapter 5). Following the osteological appraisal, problems of development (Chapter 6) and genealogy (Chapter 7) are investigated.

1.2 The Ankylosauria

The Ankylosauria is a monophyletic assemblage of quadrupedal herbivorous dinosaurs, frequently characterized, in part, by the pervasive development of parasagittal rows of osteoderms across the dorsolateral surfaces of the body, and unusual skull architecture. Detaits of ankylosaur cranial morphology form the main body of the thesis (Chapters 3, 4, 5, 6 and 7). The novel head skeleton morphology demonstrated by ankylosaurs, coupled with a relative abundance of material (relative in a palaeontological sense; i.e. more than two) provides a unique opportunity to qualitatively review this fossil taxon. The present research focuses on establishing a descriptive protocol for application to the clade

Ankylosauria (Chapter 3) through the examination of 42 specimens and a review of the literature (see Table 7.1), and then applies this methodology to specific taxa.

1.2.3 Euoplocephalus tutus Lambe (1902)

Euoplocephalus tutus Lambe (e1SC12)is the subject of a comprehensive reappraisal (Chapter 4), incorporating the physical examination of nineteen of the 25 known partial to complete skulls (see Table 4.1 ). Perhaps the best- represented ankylosaur taxon in North America, Euoplocephaluswas originally collected in 1897 and described in 1902 on the basis of a partial, fragmentary skull (the right mandible of this specimen has not yet been prepared) and incomplete postcranium (NMC 0210, holotype) by Canada's first dinosaur palaeontologist, Lawrence M. Lambe. Lambe referred to this specirnen as Stereocephalus tutus ( Stereos [Latin], sol id; kephale [ancient Greek], head; tutus [Latin], protected from danger, safe, secure), a name later found to be occupied. He renamed the specimen Euoplocephalus (Euoplos [ancient Greek], well armed) in 1910. Subsequent explorations into the badlands of southern Alberta and parts of Montana resulted in the collection of numerous other specimens referred to this taxon (Coombs, 1978). However, fragmentary, taphonomically distorted material created problems for early workers. During the 19201s,two "new" taxa of ankylosaurs, resembling (to some degree) Euoplocephalus, were described:

ûyoplosaurus acutosquameus Parks i924 (the cranial material was restricted to the caudalmost area of the occiput, a mandible and some broken teeth) (Dyo [ancient Greek], two, double; oplos [ancient Greek], armoured; acutus [Latin], sharp, pointed; squameus [Latin], covered with scales); and Anodontosaurus lambei Sternberg 1929 (a taphonomically truncated and depressed cranium and left mandible) (An [ancient Greek], without; odontos [ancient Greek], teeth; lambei, named after L.M. Lambe). Beginning in the late 1960's Walter Coombs Jr., perhaps the most significant figure in the study of ankylosaur biology, reviewed virtually al1 the material then referred to the Ankylosauria. He surrnised that while Euoplocephalus, Dyoplosaurus and Anodontosaurus did demonstrate subtle differences in morphology, this appeared to be largely the result of taphonomic deformation. Unable to assign the specimens to two or more discrete taxa, Coombs (1978) asserted that only a single genus-species combination was valid, the senior available synonyrn being Euoplocephalus tutus. Coombs also established the taxon Scolosaurus cutleri Nopcsa 1928 (Skolos [ancient Greek], thorn; sauros [ancient Greek], lizard; cufleri, named after W.E. Cutler) as a subjective junior synonym of Euoplocephalus on the basis of postcranial characters, and found Palaeoscincus asper Lambe 1902 (a single tooth) (Palaios [ancient Greek], ancient; skincus [ancient Greek], skink; asper [Latin], pointy, bristling, disagreeable to the touch) to be a nomen dubium. Although some questions as to the appropriateness of this synonymy have been fomarded (Carpenter, 1982;

1997a; b; Penkalski, 1998),no formal reanalysis has been published.

1.2.2 "Gobisaurus" gen. et sp. nov.

During the course of this work, an opportunity arose to evaluate a heretofore undescribed ankylosaur cranium, originally collected by a joint Sino-Soviet expedition in 1959-1 960. Largely forgotten for decades, the specimen resurfaced during a joint exploration / expedition venture between the lnstitute of Vertebrate Paleontology and Paleoanthropology (IVPP; Beijing), the Canadian Museum of Nature (formerly the National Museum of Canada; NMC; Ottawa) and the Royal Tyrrell Museum of Palaeontology (TMP; Drumheller), and became part of a major traveling exhibit (the Dinosaur World Tour; Currie, 1997b; see also Vickaryous et al., submitted) during the period 1990-1 997. Following the established approach utilized to describe the general ankylosaur condition

(Chapter 3) and a well represented taxon (Le. Euoplocephalus; Chapter 4), this cranium, herein referred to as "Gobisaurus" gen. et sp. nov. (to be offkially designated elsewhere; see Vickaryous et al., submitted), was similarty reviewed (see Chapter 5).

Further commentary on these issues is provided in this thesis, subsequent to the detailed descriptions of Euoplocephalus tutus and "Gobisaurus" gen. et sp. nov. being furnished (Chapters 4 and 5). This opens avenues to the reconsideration of cranial ornamentation in ankylosaurs and of the significance of cranial morphology to the systematics of ankylosaurs. Chapter 2 Fossil Materials and Methods

Osteological nomenclature for the head skeleton of ankylosaurs, as for many archosaurian clades, may best be described as anarchic (Witmer, 1994), lacking interna1 consistency and any degree of consensus. Witmer (1994) initiated a movement to codify and stabilize the anatomical nomenclature of archosaurs, following the established terminology of the Nomina Anatomica Veterinaria (NA V; 1983) and Nomina Anatomica Avium (NAA; Baumel and Witmer, 1993; see also Witmer, 1997). This nomenclatural methodology is herein adopted and applied to the Ankylosauria, with the intention of introducing a measure of uniformity to the descriptive process. To facilitate communication and consistency, vernacular terminology is used throughout the text, with a complete listing of Latin and English equivalents provided in Appendix 2.1. In accordance with the NAV and NAA, the directional terms of ventral (Ventralis) and dorsal (Dorsalis) have been substituted for the expressions anterior and posterior. Furthermore, the terms rostral (Rostralis) and caudal (Caudalis) replace use of the terrns superior and inferior. The terms "left" (Sinister) and "right" (Dexter) are used with reference to the specimen being exarnined in dorsal view. Articulation (Arficulatio) is employed in the broad sense as the general term for al1 joints - fibrous (A. fibrosae), cartilaginous (A. cartilagineae) and synovial (A. synoviales). An asterisk ("") following an accession number indicates a type specimen.

2.1 Specimens

During the course of this study, a large number of ankylosaur specimens was examined, including original and cast fossil material referred to Euoplocephalus tutus Larnbe (1902) (sensu Coombs, 1978; n = 19; see Table 4.1) and the type and only known specimen of "Gobisaurus" gen. et sp. nov. (IVPP VI2563'). Additionally, material assigned to other ankylosaur taxa (n = 23; Table 7.1) was also reviewed (see also Table 4.2). Supplementary information was gleaned from the Iiterature, principally from Coombs, 1971; Maryanska, 1977; Tumanova, 1987; Coombs and Maryanska, 1990; and Molnar, 1996. Use of the taxonomic clade Euoplocephalus tutus follows the application of Coombs, 1978, and includes the subjective synonyms (different specimen names based on different type specimens representing the same taxon; Grande and Bernis, 1998) of Palaeoscincus asper Lambe 1902, Dyoplosaurus acufosquameus Parks 1924, Scolosaurus cuflen Nopcsa 1928 and Anodontosaurus lambei Sternberg 1929. lncluded arnongst the materials examined during the course of this study were specimens referred to Euoplocephalus and specimens formerly referred to as Palaeoscincus (NMC 1349*, a single tooth) and Anodonfosaurus (NMC 8530*; Fig. 2.1A) (Scolosaurus is not known from skull material and will not be considered further). Unfortunately, specimens formerly referred to as Dyoplosaurus (ROM 784* and USNM 11892) were unable to be personally observed. However, photographs of USNM 11892 provided by Mr. Penkalski (North Carolina State University, Raleigh, NC) and Mr. Tanke (TMP) did permit most of the salient features to be assessed. A recently collected specimen from Montana, MOR 433 (purportedly a new genus of ankylosaurid, albeit one closely resembling Euoplocephalus; Penkalski, pers. comm.), was also not personally examined. However, based on the overall morphology (as gleaned from photographs provided by Mr. Penkalski) it too was placed within the clade Euoplocephalus tutus. The descriptive morphology of the ankylosaur skull presented in Chapter 3, relies upon a synthesis of al1 of the taxa under consideration (i.e. Euoplocephalus, "Gobisaurus"and other ankylosaurs), with an emphasis upon identifying the "general" morphological pattern. Subsequently, Chapters 4 and 5 detai 1 specific taxa (Euoploœphaius and " Gobisaurus" respectively), and apomorphic features and / or deviant morphologies from the "general" morpkological pattern. Chapter 6 draws upon the overall skull architecture of ankylosaurs in order to elucidate the developmental pattern of cranial ornamentation. Details of the morphological pattern of individual taxa are further considered in the phylogenetic analysis presented in Chapter 7, again involving the synthesis of al1 taxa under consideration. In the concluding chapter (Chapter 8),promising areas for future consideration are reviewed in light of this revised level of comprehension of ankylosaur skull architecture.

2.1.1 Post-mortem Deformation

In most instances, each specirnen was compared and contrasted with conspecifics (where possible) or other related taxa in order to gauge any extrinsically induced anatomical distortion. Unlike most disciplines in biology, palaeontological specimens frequently demonstrate severe post-mortem morphological alteration, herein summarized as taphonomic deforrnation (Fig. 2.1). Taphonomic deformation refers to any changes and / or alterations that occur between the spatial relationships of hoor more anatomical points on a specimen during the transition of that specimen from the biosphere into the lithosphere or geological record, and includes al1 the processes of preservation and the pre- and post-burial histories of that specirnen's organic remains (Lyman, 1994; see also Efremov, 1940). One possible agent of taphonomic deformation is the physical burial of organic material within layers of water-saturated sediment. As a result of an increase in pressure on the surrounding sediments (caused by the weight of overlying accumulation), it is hypothesized that buried specimens are subjected to forces that may induce plastic deformation. Upon subsequent re-exposure to the atmosphere, such specimens are subjected to water and 1 or wind erosion, thereby truncating surfaces and further remodeling structures and 1 or features. Other factors, such as the quality of the preservation and the care and skill associated with its excavation, collection, preparation and curation also add degrees of uncertainty with respect to the relationship of the form of the preserved material and its appearance in tife. Owing to the unusual architecture of the ankylosaur skull, viz. the near complete obliteration of cranial sutures in adult-sized specimens, the closure of temporal and antorbital fenestrations and the elaboration development of cranial ornamentation (see Chapters 3 and 6), specimens frequently undergo some degree of plastic deformation without total dissociation. While this type of taphonomic deformation has led previous workers to suggest that ankylosaur crania were naturally "depressed", or that the orbits were elliptical in shape (e.g. Gilmore, 1923; Sternberg, 1929; Fig. 2.1A), fracture patterns, asymmetrical morphology, bending I twisting of features and rare, virtually unaltered specimens indicate these morphologies are artificial.

2.1.2 Euoplocephalus; A Question of Synonymy

The taxon Euoplocephalus has been the subject of numerous taxonornic debates (see Chapter 1). Although it has not been fornially challenged in the literature since Coombs (1978) synonyrnized Palaeoscincus, Dyoplosaurus, Scolosaurus and Anodontosaurus with Euoplocephalus, several authors have suggested that this clustering rnay be overly ambitious (e-g. Carpenter, 1997a; Penkalski, 1998). Critical to the evaluation of both arguments is a detailed understanding of morphology. Grande and Bemis (1998) reviewed the issue of nomenclature and morphological diagnosis of taxa (specifically dealing with amiid fishes) and sunised that "There seems to be a strong compulsion to name a new species in paleontology even when the material or available morphological information does not warrant itl'. They cited four causes that may lead to such superfiuous taxonomy: (1) failure to examine enough known specimens; (2) failure to consider the effects of development on morphology; (3) inadequately preserved and I or prepared material; and (4) the assumption that a taxon is distinctive because it cornes from a different locality or geologic horizon than previously reported (Grande and Bemis, 1998). All of these concerns, and in particular causes 1 and 2, apply to the issue of validity for the synonymy of Euoplocephalus. This contribution is primarily fowsed on the morphological evaluation of the skull of the taxon Euoplocephalus, and al1 the junior synonyms ascribed to it (specifically Oyoplosaurus and Anodontosaurus), with complete impartiality as to the outcorne. As will be noted later (and astutely predicted by Coornbs, pers. comm., 1995), there are no morphological features of the skull that warrant segregation of the taxon Euoplocephalus into other genera or species. Not unexpectedly, each specimen may be discretely identified on the basis of subtle features (see Fig. 2.1 ), although frequently these appear to be more the result of taphonomic deformation than the result of naturally occurring individual variation. Thus, there does not exist, at this tirne, enough morphological evidence to support any subdivision of the clade Euoplocephalus, or enough evidence to support the resurrection of either Dyoplosaurus or Anodontosaurus.

2.2 Methodology

Several approaches were utilized in order to assess the rnorphology of al1 the taxa under consideration. Perhaps the most obvious is traditional qualitative observation, coupled with photography and illustration. Careful osteological analysis was typically approached through the examination of multiple specirnens made available synchronously, in order to minimize any confusion concerning taphonomic deformation (see above). Selected specimens of Euoplocephalus (viz. AMNH 5405, TMP 91.127.1, TMP 97.59.1 and TMP 97.132.1) and the holotype of *Gobisaurus" (IVPP 12563') were subjected to non-invasive computer tornography (CT) scanning at the Foothills Hospital, Calgary, using a GE Medical Systems CT HiSpeed Advantage scanner. Two isolated Euoplocephalus teeth (TMP 87.36.99and TMP 92.36.1226) were viewed using scanning electron microscopy. The teeth were sputter coated with gold (to a thickness of approximately 20 nm) using a Semprep2 sputter coater, viewed with a Hitachi 5-450 scanning electron microscope in the Microscopy and lmaging Facility at the Faculty of Medicine, University of Calgary, and photomicrographed using Polaroid Type 55 PIN film. Other teeth were observed using a Wild M5 bifocal dissecting microscope (see Table 4.2). Histological (fossil bone) thin-sections were generated according to standard geological procedures for rnicroscopic sections (Resultay, pers. corn., 1997). Two specimens were physically sectioned either coronally (TMP 67.20.20) or transversely (TMP 98.1 15.2) to produce thin wafers. The wafers were then glued to glass slides and ground down to a thickness of approximately 0.03mm.The resulting preparations were then viewed and photomicrographed using a 35mm camera mounted on a Leitz Ortholux 2 Polarizing Microscope, with the polarizer engaged. Minerals in the fossil material have a crystalline structure that alters the transmittance of polarized light, thus highlighting microscopic structures present (e.g. Haversian systems). These thin-sections were compared to those previously created from AMNH 5403. All measurements, to the nearest 0.5mmI were taken three times, using vernier calipers, and averaged. Figure 2.1. Left lateral profile of three specimens of Euoplocephalus futus Lambe (1 902). A. NMC 8530*,the type specimen of Anodontosaurus lambei Sternberg 1929. B. AMNH 5405. C.TMP 96.75.1. Note the variability in morphology, especially evident in the shape of the orbit and the rostrum. NMC 8530* is an excellent example of a taphonomically depressed craniurn. TMP 96.75.1 is both foreshortened and broken, again the result of taphonomic deformation (in particular, the effects of subarid exposure to the elements). AMNH 5405 is extrernely well preserved, and represents a closer approximation to the actual cranial architecture of Euoplocephalus than either NMC 8530* or TMP 96.75.1. AMNH 5405 collected from the Judith River Group of what is now Dinosaur Provincial Park. NMC 8530" and TMP 96.75.1 both collected frorn the Horseshoe Canyon Formation. Scale bar, 100 mm.

Chapter 3 Ankylosaur Skull Morphology

The ankylosaur skull (cranium plus mandibles; sensu Hildebrand, 1988), arguably represents one of the Ieast understood areas of archosaur osteology. The extremely divergent, apomorphic architecture and (putative) plethora of neomorphic (ectopic) elements, have frustrated most previous descriptive efforts, often resulting in the mis- or over-interpretation of the represented material. Owing to the profound importance of the skull in interpreting the biology and relationships of any vertebrate fossil taxon (Hanken and Hall, 1993),a Jack of resolution of this fundamental source of anatomical information presents a serious obstacle that has hampered al1 past research, and threatens to continue to do so. This chapter provides an overview of ankylosaur skull morphology, and identifies a number of problems that have hindered interpretation and assessment of ankylosaur systematics and functional morphology.

3.1 Overall Morphology

The characteristic morphology of the ankylosaur skull consists of a robust, stout cranium (Fig. 3.1), with both the orbits and the externat nares oriented rostrolaterally (although the latter are subject to a degree of variation), and a comparatively gracile mandible with an abbreviated retroarticular process and reduced wronoid bone. Furthermore, three major architectural novelties uniquely diagnose members of this clade: (1 ) the secondary closure of the dorsotemporal (= supratemporal), antorbital, and mandibular fenestrae (although there is no inherent assumption that these erasures are necessarily related); (2) the almost complete obliteration of cranial sutures across the dorsum; and (3) the pervasive development of osseous ornamentation, including boss-like protuberances dorsal to the orbits and at the caudodorsal and caudoventral borders of the cranium. Conjointly, al1 three architectural modifications impede element identification and the establishment of homologies (especially across the dorsolateral surfaces of the cranium) through the obkiscation of traditional morphological landmarks. Additional characteristic features of the ankylosaurian skull include a partially edentulous rostrum, rnedially deflected tooth rows on both the cranium and the mandibles, a slender, dorsoventrally depressed predentary, and a small mandibular (= mental) symphysis at which the contralateral elements never fuse. Ankylosaur teeth are relatively small, laterally compressed, and denticulated along the carina (Coombs and Maryanska, 1990). The cranium is akinetic (contra Coombs, 1971), afthough evidence suggests that in at least one taxon (Euoplocephalus tutus) the predentary-dentary contact was mobile, and the quadrate-articular jaw joint may not have been restricted to a simple orthal movement (Rybczynski and Vickaryous, in press). As a result of this unusual skull morphology, previous workers have found it difficult to discuss the ankylosaur skull in terms of its constituent parts, with disarticulated elements being rare and synarthroses not being readily demonstrable. Recently however, a number of unusual ankylosaur specimens have been collecteci that display neither obliterated sutura1 contacts along the dorsum, nor a dorsum completely embossed with extensive osseous ornamentation. Although representative specimens are only known from three taxa (subadult specimens of Pinacosaurus grangen from China, Minmi sp. from Australia, and an undescribed taxon from the United States herein referred to as the Cedar Mountain ankylosaur), they provide critical insight into the morphology of ankylosaur skull elements in general. Such fortuitous specimens, however, do not represent the majority of taxa and thus an alternative methodology for the discussion of skull morphology is required. That implemented here is based upon the arbitrary division of the skull into several mutually exclusive topographical, but not necessarily developrnental, regions (Figs. 3.1 - 3.5, Appendix 3.1). It is acknowledged that owing to the arbitrary nature of the topographical division, several elernents cross the boundaries of these topographical regions. For the purposes of this contribution, each element is included within the region in which it predominantly resides (i.e. where the majority of it is contained), and each element is reported for one region only (Appendix 3.1 ). The rostral region (Fig. 3.1) represents al1 the elements composing the cranium rostral to a hypothetical transverse section placed immediately in advance of the orbits that contribute, at least in part, to the dorsal and lateral surfaces. The temporal region (Fig. 3.2) represents the caudal complement to the rostral region, composed of the dorsolateral surfaces (excluding the elements defining the interorbital opening) of the cranium caudal to a hypothetical transverse section placed immediately in advance of the orbit. The palatal region (Fig. 3.3) is composed of al1 the elements situated along the ventral surface of the cranium covering the roof of the oral cavity, caudal to the rostral region contributions to the palate (viz. the premaxilla and maxilla). The occipital 1 basicranial region (Figs. 3.4, 3.5) is composed of al1 the elements present in caudal view of the cranium, including the basicranium proper and associated structures, and the elements composing the interorbital wall, but excluding those elements visible prirnarily in lateral view (viz. the parietal, squamosal and quadratojugal elements of the temporal region). The lower jaw is presented as the mandibular region (Fig. 3.6). Since the cranial anatomy of ankylosaurs is so poorly known, the placement of particular elements into specific topographical regions is largely based upon their corresponding position in successive outgroups (viz. stegosaurs and Lesothosaurus; see Coombs and Maryanska, 1990; Sereno, 1991; Sereno and Zhimin, 1992).

3.2 Ossa Cranii

Unless otherwise stated, al1 cranial elements are paired.

3.2.1 Rostral Region (Fig. 3.1 )

Delineation of the topographie margins of constituent elements is virtually impossible in the majority of ankylosaur taxa, as the sutura1 contacts are obscured by the presence of cranial ornamentation and 1 or the obliteration of sutures. Therefore, commentary concerning element morphology, and arthrology is frequently limited to a few disparate taxa.

3.2.1.1 Ospremaxillare (Figs. 3.1A, 3.2A13.3A1 3.4A, 3.7, 3.8, 3.9) The premaxilla is situated at the rostral terminus of the cranium proper, bounding (in part) the external naris (and associated nasal vestibule) and (when present) the paranasal aperture, and partial1y roofing the oral cavity. Although variable between taxa, the premaxilla frequently occupies a large proportion (up to 50%; Maryanska, 1977; Molnar, 1996) of the overall length of the rostral region. In palatal view, the premaxilla forms the majority of the tomium. The tomial crest (crista tomialis premaxillare) may alternatively be restricted to an extreme rostral position, a condition directly correlated with the presence of premaxillary teeth (e.g. Gargoyleosaurus; C hapter 7, section 7.2, character 17), extend caudally across the premaxillary-maxillary contact in confluent alignment with the maxillary tooth row (e.g. Edmontonia), or extend caudally across the premaxillary-maxillary contact before terrninating in a position lateral to the maxillary tooth row (e.g. Saichania) (see also Chapter 7, section 7.2, character 16). Amongst the Ankylosauria, the presence of premaxillary teeth is relatively rare (see Appendix 3.2). Contralateral premaxillary elements contact one another along the sagittal plane and, when visible, the interpremaxillary suture appears to have been a butt joint (i.e. not serrated; Carpenter et al., in press). In palatal view the interpremaxillary suture is frequently obliterated, although its presence rnay be inferred by a shallow longitudinal furrow, and / or an oblong depression. A premaxillary notch (incisera premaxillare) (= interpremaxillary notch; Sereno, 1999) of variable size and morphology (ranging from an inverted IV'to an inverted "U") may incise the interpremaxillary suture at its rostroventralrnost point. The presence of this notch appears to be a phylogenetically significant characteristic of many ankylosaurs, the implications of which are discussed in Chapter 7, section 7.4 (see character 15). Directly caudal to the premaxillary notch, along the palatal surface of each premaxilla, is a caudodorsally inclined parasagittal incisive foramen (foramen incisivium). Some ankylosaurs (e. g. Pinacosaurus, Saichania; Maryanska, 1977, also Ankylosaurus) are reported to have a second, more caudally positioned pair of parasagittal formina (Fig. 3.9). These openings have been referred to as conduits for Jacobsen's organ, but such homology is questionable. Carpenter et al. (in press) report that the Cedar Mountain ankylosaur lacks incisive foramina. Externally, the premaxilla may bear a number of foveae (foveae corpusculorum nervosorum) and shallow, vertically oriented furrows. Dorsolaterally, the morphology of the premaxilla, and the position of sutura1 contacts with adjacent elements (e.g. the nasals) is obscured. Characteristically, however, the premaxillary-maxillary contact may be traced along the palate as a narrow furrow that undulates rostrally from a position ventral to the external nares before circumventing the alveolar ridge of the maxilla, and continuing caudomedially. Contact between the prernaxilla and the vomer along the palatal surface is frequently indistinct, although when present the vomer appears to invade the premaxilla dong the interpremaxillary suture. In Minmi, the Cedar Mountain ankylosaur and subadult Pinacosaurus, the dorsolateral morphology of the premaxilla is at least partially visible (see Maryanska, 1977; Molnar, 1996; Carpenter et al., in press). Subadult Pinacosaurus exhibit a short, relatively linear premaxillary-maxillary contact in lateral view, perpendiwlar to the long axis of the cranium (Fig. 3.7A). An unnamed triangular process of the premaxilla incises the rostral border of the maxilla-nasal contact. Although present, the premaxillary-nasal contact is not weil exemplified in any Pinacosaurus specimen. In Minmi the contact is interpreted as being obliquely oriented, directed caudodorsally at a 45' angle and extending to a position 113 of the distance along the length of the maxilla, wherein it becomes reoriented as a vertical abuttment (Molnar, 1996; Fig. 3.78). Contact between the premaxilla and maxilla ends at the caudal edge of the external naris, thereby excluding any direct connection with the nasals. Molnar (1996) states that the premaxilla overlaps the lateral surface of the maxilla. The premaxilla (CEUM 10405) of the Cedar Mountain ankylosaur partly resembles that of Minmi, in that the premaxillary-maxillary contact is obliquely oriented caudodorsally (Fig. 3.7C). At its caudal terminus, however, the premaxilla tapers to a deltaic projection, before turning rostrally to contact the external naris. As preserved, the sutura1 contacts of the premaxilla with both the maxilla and the nasal appear to be butt joints. Rostrodorsomedially, a process from each premaxilla, the internasal bar (pila internasalis), extends caudodorsally, divorcing the two external nares. Reportedly in some taxa (e-g. Tsagantegia, Tumanova, 1993; also see Shamosaurus, Tumanova, 1983; Talarurus, Tumanova, 1987) the internasal bar conjoins and fuses with the vomers sagittally, forming the rostralmost extension of the internasat cavity septum (see section 3.2.3.1). Many ankylosaurs further demonstrate a subdivision of the external nares proper (within the nasal vestibule) by an additional premaxillary process or septum (Chapter 7, section 7.2, character 24). Orientation of this bony nasal septum (sepfum nasi osseum) is highly variable between taxa, ranging from vertical (e.g. Euoplocephalus) to horizontal (e.g. Saichania). Partitioning of the external naris by the bony septum is frequently disproportionate, with the medialmost aperture (the nasal aperture) being larger. Although preliminary details of the respiratory system and cranial pneumatics are discussed in Chapter 7, it is pertinent to mention here that the iateralmost aperture (the paranasal aperture) in some taxa leads to a paranasal sinus cavity that may be variably enclosed within the premaxilla (Chapter 7, section 7.2,character 26). Pinacosaurus appears to demonstrate an additional large opening within the confines of the nasal vestibule. Maryanska (1977) suggested that this is a supplementary paranasai aperture, leading to a sinus cavity within the premaxilla.

3.2.1.2 Os maxillare (+ rostrodorsal secondary palate) (Figs. 3.f A, 3.2A1 3.3A1 3.4A, 3.7A1 3.7B, 3.9) The maxilla, the principal element bearing cranial dentition (Appendix 3.2), is situated along the lateral aspects of the rostral region, caudally adjacent to the premaxilla. In palatal view, the maxilla rnay contribute to the caudal terminus of the tornial crest (crista tomialis maxillare; see also section 3.2.1.1 ). Apomorphic for the Ankylosauria, the maxillary tooth row is inset medially from the lateralmost border of the maxilla, thereby bisecting the palatal surface into buccal and lingual portions (Russell, 1940). The buccal portion or emargination (= cheek pouch: Lee, 1996) varies in cross-sectional morphology, from a relatively flat, planar shelf (e.g. Gargoyleosaurus), to one which is deeply vaulted (e.g. Ankylosaurus) (Chapter 7, section 7.2, character 22). In some taxa, direct contact between the premaxillary tomial crest and the maxillary tooth row decouples the premaxillary palate from the buccal emargination (e.g. Edmontonia). Alternatively, lateral displacement of the tomial crest perrnits confluence between these two regions (e.g. Ankylosaurus; also see section 3.2.1.1 and Chapter 7, section 7.2, character 17; Coornbs, Igïl). The lingual portion of the maxilla (medial to the tooth rows) undulates dorsally as a smooth, vaulted surface. Proximal to the alveoli proper, along the alveolar border, lies a sub-parallel row of foramina (the "special foramina" of Edmund, 1957), within which the crowns of replacement teeth or the roots of the occluding dentition are often identifiable. In ankylosaurs these foramina are frequently irregular in distribution and rnorphology, although the general diameter corresponds to the width of the cusp. Dorsomedially beyond the "special formina", the lingual portion of the maxilla alternatively contributes tcl the rostralmost border of the internal naris (choana) (e.g. Gargoyleosaurus) or a palatal shelf, the purported rostrodorsal secondary palate (Coombs and Maryanska, 1990; anterior maxillary shelf, Tumanova, 1983; also see Chapter 7, section 7.2, character 21) forming the floor of the respiratory passage (cavum naso and the rostral margin of the internal naris. In addition to this lingual maxillary shelf (lamina mediana maxillare), the rostrodorsal secondary palate also receives contributions from the vomer, although the sutura1 contacts between these elements are invariably damaged and / or obliterated. Caudolaterally, the maxilla may also contribute to a longitudinal ridge leading to the caudoventrally oriented secondary palate (see section 3.2.3.2). The rnaxillary tooth row resides along a ventrally projecting ridge, the alveolar border (processus alveolaris) (= dentigerous ridge: Russell, 1940), noted earlier as the line of subdivision between buccal and lingual portions of the maxilla. Characteristically, the alveolar border of ankylosaurs is medially deflected, with the minimum span between juxtaposed tooth rows occurring within the rostralmost third of the element (Chapter 7, section 7.2, charactei- 18). Typically, maximum separation occurs at the caudalmost tooth positions. Caudal to the last tooth position, the alveolar border becomes rugose, forming an alveolar tuberosity (tuber alveolare). Laterally, the maxillary tornium frequently obscures the rostralmost tooth positions of the alveolar border. Although linear, the alveolar border tends to be slightly caudoventrally oriented with respect to the long axis of the cranium. In addition to the aforementioned contacts along the palate with the premaxilla (section 3.2.1.1 ) and the vomer, the maxilla also contacts the jugal, ectopterygoid, and palatine, although in many taxa these junctures are obliterated and 1 or not exposed. Subadult specimens of Pinacosaurus illustrate this area particularly well, although contact between the maxilla and palatine remains indistinct. The maxillary-jugal contact is serrate, while the maxillary- ectopterygoid contact appears to be a scarf joint. Delineation of the topographic margins of the maxilla along the dorsolateral surface of the cranium is virtually impossible in most taxa. Specimens of Pinacosaurus and Minmi, however, confirm that the maxilla contacts the premaxilla rostrally, the nasal rostrodorsally, the lacrimal caudodorsally, and the jugal caudally (Figs. 3.7A, B). These taxa further verify the complete absence of the antorbital fenestra. In the Cedar Mountain ankylosaur, the maxilla (CEUM unnumbered) is fragmentary, and does not display any of its sutura1 contacts with adjacent elements. The maxilla may (e-g. Minmi, Molnar, 1996; Fig. 3.7B) or may not (e.g. Pinacosaurus; Figs. 3.7A) contribute to the caudal margin of the externai naris, although the situation is equivocai for the majority of taxa. Both the maxillary-nasal and the rnaxillary- lacrimal synarthroses of Pinacosaurus appear to be serrate. In Minmi these contacts are less distinct.

3.2.1.3 Osnasa/e(Figs. 3.1A, 3.2A, 3.7A13.78, 3.8A, 3.8B) Despite being considered the largest element in terms of surface area of the ankylosaur head skeleton (Maryanska, 1977), the nasal bone proper has rarely received consideration in the literature. Frequent circuitous references to the element in the context of regionalized topographic positions (Le. the mid- dorsal antorbital area; see Russell, 1940; Lee, 1996; Sullivan, 1999) belies Our current state of knowledge. However, the presence of this element in specimens of both Minmi and subadult Pinacosaurus permits a limited morphological assessment. Specimens of both Pinacosaurus and Minmi demonstrate a similar nasal morphology with regard to the dorsal contribution (Figs. 3.8A, B). The nasal is parasagittal, directly caudal to the premaxilla, abutting its contralateral counterpart along a lengthy continuation of the interpremaxillary suture (the internasal suture) (see section 3.2.1.1). At its caudalmost extremity, the nasal contacts the frontal in line with the midpoint of the orbit, along a lengthy transversely oriented serrate suture. Rostrally, the nasal expands laterally, giving the element a subrectangular (Minm~]or pentagonal (Pinacosaurus) appearance in dorsal view. Along this oblique border, the nasal contacts the prefrontal at an interdigitating synarthrosis. Rostrally, the nasal forms a junction with the premaxilla that, as noted earlier, is not preserved well enough for description. The rnorphology of the lateral contribution (Figs. 3.7A1 B) differs between Pinacosaurus and Minmi. In the former, the nasal arcs perpendicularly around the prefrontal-lacrimal complex (located in advance of the orbital cavity), before terminating at the premaxillary-maxillary contact. This lateral process of the nasal is situated as a vertically oriented rectangular strip covering the rostralmost half of the antorbital region. In Minmi, the maxilla and prefrontal are interpreted as precluding the nasal from contributing to the lateral aspect of the antorbital region (Molnar, 1996). A specimen of Talarurus (Tumanova, 1987) illustrates the complete truncation of both secondary palates (rostrodorsal and caudoventral), thus permitting the investigation of the medial (nasal) surface of the nasal proper. Extrapolating their topographic position, this specimen suggests that the nasal contributes to the roof of the nasal cavity (cavum nas& including the sagittally subdividing internasal cavity septum (see section 3.2.3.1), although none of the sutural contacts can be clearly visualized.

3.2.1.4 Os prefronfale (Figs. 3.1A, 3.2A1 3.7A, 3.8A, 3-88) With the exception of three taxa (Pinacosaurus, Minmi, and the Cedar Mountain ankylosaur), poor preservation of the sutural contacts delineating the ankylosaurian prefrontal precludes description. Within the these taxa, the prefrontal is caudolateral to the nasal, rostrolateral to the frontal, rostromedial to the supraorbital(s), and dorsal to the lacrimal. In dorsal view the prefrontal is typically subrectangular, oriented obliquely rostrolaterally with respect to the sagittal plane, and forms the continuation of the lateral margin of the rostrum (following the nasals). Laterally, the prefrontal may either be al1 but absent (Minmi; although Molnar [A 9961 speculates that a slender rod-like element in advance of the lacrimal may be a process of the prefrontal), or contribute to a minor rectangular portion of the antorbital area (Pinacosaurus, Cedar Mountain ankylosaur; Figs. 3.7A1 B). Carpenter et al. (in press) report that the prefrontal contributes to the rostromedial orbital wall in the Ceclar Mountain form. Medial to the prefrontal- lacrimal contact Carpenter et al. (in press) identified a novel depression, the so- called prefrontal fossa. Contact between the prefrontal and the nasal is a scarf joint, with either the nasal overlapping the prefrontal (Cedar Mountain ankylosaur; Carpenter et al., in press) or the prefrontal ovetlapping the nasal (Tumanova, 1987, presumably with reference to subadult Pinacosaurus). The condition in Minmi is unreported. All other contacts are serrate, with the exception of the planar abuttment of the prefrontal to the lacrimal.

3.2.1.5 Os lacrimale (Figs. 3.1A, 3.2A13.7) Although rarely identifiable as a distinct element, the lacrimal (alternatively spelled lachryrnal; Carpenter et al., in press) represents the rostralmost of the circumorbital bones (as evidenced by specimens of Pinacosaurus, Minmi and the Cedar Mountain ankylosaur -- Sullivan [A9991 reported the "prominent" appearance of a similarly placed lacrimal in Nodocephalosaurus, but, the description is sketchy and published photographs and illustrations are not conclusive). Laterally, the lacrimals of both Pinacosaurus and the Cedar Mountain ankylosaur are subrectangular, with the caudalmost edge in each rnarking the rostral rim of the orbit (Figs. 3.7A, C). Orientation of the element varies between the two taxa, with the lacrimal of subadult Pinacosaurus being rostrodorsally inclined, while the long axis of the Cedar Mountain form is parallel to the main axis of the cranium. In comparison to these two taxa, the reported morphology of the lacrimal of Minmi differs considerably, consisting of a slender wedge that follows the rostral margin of the orbit, tapering towards the ventral contact with the maxilla, (Fig. 3.78). The long axis is vertically oriented, with the lacrimal terminating at the rostrodorsal border of the orbit. Along the rostromedial border of the orbit, the lacrimal is perforated by the caudal nasolacrimal aperture (apertura eaudalis [orbitalis] canalis nasolacnmalis; = lacrimal foramen, Maryanska, 1977; Sullivan, 1999; = lacrimal duct, Carpenter et al., in press), communicating via a canal (canalis nasolacnmalis) with the maxillary sinus. Tumanova (1987) reported a second lacrimal foramen in (an) undisclosed ankylosaur(s), although the homologies of this purported canal are unclear. The lacrimal contacts the maxilla rostroventrally, the jugal caudoventrally, and the supraorbital(s) caudodorsally. Contact between the laaimal and supraorbital(s) is serrated. In subadult Pinacosaurus and the Cedar Mountain ankylosaur, the lacrimal contacts the prefrontal rostrodorsally (Figs. 3.7A, 3.7C, 3.8A, 3.8C). The condition in Minmi is difkult to determine, although it appears that the lacrimal does not abutt the prefrontal (see section 3.2.1.4; Fig. 3.78). Medially, a process of the lacrimal of Pinacosaurus contributes to the rostromedial wall of the orbital cavity (Maryanska, 1977; Tumanova, 1987). This orbital process reportedly contacts the " sphenethmoid " (= accessory ossification; Maryanska, 1977) within the orbit, although neither the process nor the accessory ossification is preserved well enough for description (see section 3.2.2.8).

3.2.2 Temporal Region (Fig. 3.2)

With the exception of a few taxa, elements of the temporal region, like their rostral counterparts, are poorly delineated. As a result, evaluation of element morphology and arthrology is limited.

3.2.2.1 Ossa supraorbitalia (= ossa palpebrae, os postfrontale) (Figs. 3.1 A, 3.2A, 3.7, 3.8) The supraorbital bones (= palpebrals: Coombs, 1971), positioned as the name suggests, constitute a cornplex, or series, of irregularly shaped elements spanning the dorsal margin of each orbital cavity. In addition to their circumferential location with respect to the orbit, the supraorbitals may be characterised by the frequent appearance of ornamental elaboration(s) (processus cornuum supraorbitalium), resernbling wedge-like bosses (e.g. Ankylosaurus, Saichania) or knob-shaped prominences (e.g. Edmontonia, Silvisaurus) (Chapter 7, section 7.2, character 5). For the majority of ankylosaur taxa, identification of the supraorbitals is restricted to their ornamental embellishment (Le. boss or protrusion), and consequently their number, exact morphology, and arthrology are rarely demonstrable. The supraorbital series is most clearly illustrated in subadult specimens of Pinacosaufus (Figs. 3.7A, 3.8A). Above each orbit are three individual supraorbital elernents. The rectangular rostral supraorbital (= presupraorbital: Maryanska, 1971 ; 1977) and rectangular caudal supraorbital (= postsupraorbital: Maryanaska, 1971; 1977) contribute equally to the lateral margin of the orbit (Fig. 3.8A). The pentagonal medial supraorbital (= postfrontal, Maryanska: 1971; 1977), interposed between the prefrontal, frontal, postorbital and rostral and caudal supraorbitals, forms the bulk of the orbital roof (Fig. 3.8A). Laterally, only the rostral and caudal supraorbitals are visible (Fig. 3.7A). Rostrally, the complex maintains a lengthy undulating contact with the prefrontal obliquely along the lateral face of the cranium, before abruptly terminating at the lacrimal, near the dorsal third of the orbit. The caudalmost supraorbital, forming the majority of the prominent boss, mirrors the contribution of the rostral supraorbital to the dorsal orbital margin before articulating with the postorbital. Similar to the supraorbital complex - prefrontal contact, the postorbital synarthrosis is elongate (Fig. 3.7A). All sutural contacts appear to be serrated. A supraorbital element was also identified in Minmi (Molnar, 1996), although the morphology of this element diverges considerably from Pinacosaurus. Dorsally, only a single supraorbital element is visible, consisting of a mediolaterally compressed pentagon interposed between the prefrontal and "squamosal" (Fig. 3.78). Laterally, the supraorbital is planar, and makes contact with the "squamosal" caudally, the lacrimal rostrally, and an unidentified element, possibly an osteoderm or a second supraorbital (Molnar, 1996), caudoventrally. None of the contacts are preserved well enough for description. Carpenter et al. (in press) note the presence of a supraorbital complex in the Cedar Mountain ankylosaur (CEUM l236O*, holotype), although no sutural contacts were identified (Fig. 3.7C). An extremely unusual element, herein referred to as a modified "ciliary" supraorbital, is infrequently found associated (even articulated) with the roof of the orbital cavity of Euoplocephalus. Shaped like a quarter-sphere, this supraorbital is not integrated into the cranium proper, but remains unconstrained, and is interpreted as having been associated with the eyelid (see discussion in Chapter 4, section 4.4.1). Lee (1 996) described a pair of unremarkable sub- ovoid bony elements affiliated with the skeleton of Pawpawsaums as "palpebrals", analogous to the modified supraorbitals of Euoplocephalus. Although presumed to articulate with the dorsal rim of the orbit (thereby acting as a laterally oriented osseous awning; Lee, 1996), the "palpebral" of Pawpawsaurus was not found in association with the orbital cavity. While the homology of this "palpebral" with either the rnodified supraorbital of Euoplocephalus or any other ankylosaurian element remains equivocal, it should be noted that several taxa of modern crowdylians (e.g. Alligator, Osteolaemus, Paleosuchus; see lordansky, 1973; Brochu, 1999) demonstrate a similarly positioned, loosely articulating supraorbital.

3.2.2.2 Os postorbitale (+ lamina postocularis) (Figs. 3AA, 3.2A, 3.7A, 3.7C, 3.8A, 3.8C) The postorbital is the caudalmost contribution to the circumorbital series, although its exact morphology for rnost taxa cannot be determined. When evident, each postorbital may be further subdivided into three distinctive regions; a dorsal surface, a Iateral surface, and a medial (postocular) septum (Maryanska, 1977). This tripartite morphological arrangement is well preserved in subadult

specimens of Pinacosaurus (Figs. 3.7A, 3.8A). Across the dorsum, the . postorbital is situated as a narrow, intrusive rectangle between the caudally positioned squamosal and the rostrally placed medial supraorbital. The medialmost terminus of the postorbital contacts the frontal, and perhaps the rostrolateral corner of the parietal. Moving laterally, the dorsal surface of the postorbital (Fig. 3.8A, C) expands longitudinally, forming a deltaic wedge between the supraorbital and squamosal bosses. In profile, the lateral aspect of the postorbital (Fig. 3.8) is irregularly shaped, arching around the caudal supraorbital before contacting the ventralmost half of the orbital margin. Rostroventrally, the postorbital makes an abbreviated contact with the jugal. The vertically oriented caudal articulation with the quadratojugal (= secondary dermal plate: Maryanska, 1977) is elongate, terminating dorsally at the squamosal contact. Apomorphic for the Ankylosauria is the postocular shelf (lamina postocularis; see Chapter 7, section 7.2, character 41), a rostrally concave transverse wall of bone occluding the caudal border of the orbital cavity. The postocular shelf is a composite structure, receiving major contributions from the postorbital, jugal and (reportedly) the laterosphenoid (Maryanska, 1977). The integrated nature of this shelf prevents an atomistic segregation of the constituent elements, and requires a synthesised description. In subadult Pinacosaurus, the postocular shelf is erected at the caudalmost margin of the orbit, corresponding to the juncture between the postorbital and the jugal. Maryanska (1977) reports that the medial (ocular) septum of the postorbital contributes to the bulk of the vertically oriented wall, while the jugular contribution is largely restricted to the ventralmost margin. The shelf extends rnedially, contacting the basicranium near the basioccipital-basisphenoid articulation. The postorbital element of Minmi is not readily identifiable as a distinct entity. Molnar (1996) integrated this element into the "squamosal complex" (section 3.2.2.6). Presence of a postocular shelf could not be determined. Morphology of the postorbital (CEUM 10352) in the Cedar Mountain ankylosaur differs somewhat from that of the Pinacosaurus and Minmi. The dorsal surface forms a subtriangular wedge between the rostrally Iocated supraorbital complex and the caudally located squamosal (Fig. 3.8C). The postorbital also makes abbreviated contacts with both the parietal (medially) and the quadratojugal (caudolaterally). In profile, the postorbital of CEUM 10352 appears to contact the supraorbital complex near the apex of the orbit before circumventing the entire caudal margin of this opening (Fig. 3.7C). Contact between the postorbital and jugal is not preserved. Virtually the entire caudal margin of the postorbital articulates with the vertically oriented quadratojugal, before terminating at the squamosal contact. In striking contrast to the condition noted in Pinacosaurus, the postocular shelf of the Cedar Mountain form is poorly developed. In addition, Carpenter et al. (in press) note a contribution to the shelf by an interna1 process of the frontal. 3.2.2.3 Os jugale(Figs. 3.1A, 3.2A, 3.3Al 3.4A, 3.7A, 3.78, 3.9) The jugal forms the ventralmost rnargin of the orbital cavity, spanning a characteristically shallow suborbital arch (arcus jugalis) (Sereno, 1999). Ventrally, the jugal provides a short, tapering continuance of the buccal emargination, frequently including an obtuse extension of the tomial crest (tubera tomialis jugalis). Subadult specimens of Pinacosaurus confirm earlier interpretations (e.g. Coombs, 1971) of the jugal articulating with the maxilla rostrally, the lacrimal rostrodorsomedially, the postorbital caudodorsomedially (giving rise to the postocular shelf; section 3.2.2.2), and the quadratojugal caudally (Maryanska, 1971; 1977; Figs. 3.7Al 3.8A, 3.9). Additionally, the jugal may form a narrow contact with the ectopterygoid along the palate, caudomedial to the tooth rows. The jugal and the pterygoid do not appear to come into direct apposition with one another (contra Tumanova, 1987). The majority of articulating surfaces along the jugal in subadult Pinacosaurus specimens appear to be scarf joints, although the postorbital contact is interdigitated. Molnar (1996) described the jugal of Minmi as a large, triangular element, forming the ventral margin and floor of the orbital cavity, and the entire caudoventral portion of the cranium (Fig.3.7B). It remains unclear whether or not the jugal contributes to a postocular shelf. Caudally, the jugal overlaps the quadratojugal, thus completely obscuring it from lateral view. Neither the caudodorsal contact with the "squamosal" nor the rostral contacts with the lacrimal and 1 or maxilla are preserved well enough for description. The morphology of the jugal of the Cedar Mountain ankylosaur is poorly delineated, with the suture between the jugal and postorbital being obscured (Carpenter et al., in press). Carpenter et al. suggest, however, that the jugal is precluded from contacting the quadratojugal by the intervening postorbital. Recently a number of authors have described triangular or cornuate ornamentation associated with the jugals of a variety of ankylosaurs (e.g. the jugal dermal plate of Pawpawsaurus, Lee, 1996; the jugal derrnal plate of Minmi, Molnar, 1996; the jugal " horn" of Gargoyleosaurus, Carpenter et al., 1998; the jugal scutes of Sauropelta and Silvisaurus, Carpenter and Kirkland, 1998; the jugal spine of Gastonia, Kirkland, 1998). However, personal examination of specimens of Pawpawsaurus (cast of SMU 73203*), Sauropelta (AMN H 3035), Gargoyleosaurus (DMNH 27726*) and Gastonia (CEUM 1307*) suggests that this purported association is in error, and that the caudoventral ornamentation of ankylosaurs is generally restricted to the quadratojugal (section 3.2.2.7). The condition for Silvisaurus and Minmi could not be verified.

3.2.2.4 Osfrontaie(Figs. 3.1A, 3.2A, 3.5A, 3.7A13.7CI 3.8) Obscured by cranial ornamentation and 1 or obliteration of the delineating sutures, the ankylosaur frontal is poorly known. In both subadult Pinacosaurus (Maryanska, 1971; 1977) and the Cedar Mountain ankylosaur (Carpenter et al., in press), the rectangular frontal is positioned as the caudal continuance of the parasagittal nasal (Figs. 3.8A13.8C). Along with the nasal, the frontal adjoins the prefrontal (rostrally), and also contacts the medialmost supraorbital (laterally), and the parietal (caudally). AI1 the sutura1 contacts, including that between the juxtaposed frontal elements, are serrated in Pinacosaurus, while the condition in the Cedar Mountain arikylosaur cannot be determined, with the exception of the interdigitating frontal-nasal abuttment. ln Minmi the frontals have coalesced into a single, unpaired element, with a hexagonal outline (Fig. 3-88). This unified frontal is bounded rostrally by the nasals, laterally by the prefrontals (rostrolaterally) and the "squamosals" (caudolaterally), and caudally by the parietal. In addition to forming the medialmost segment of the orbital cavity roof, the ventral surface of the ankylosaur frontal contributes to a number of composite partitioning structures, as inferred by unusually preserved specimens of Talarurus (Tumanova, 1987) exhibiting cornplete palatal truncation (section 3.2.1.3). Extrapolating a general topographic position for the frontal based on specimens of Pinacosaurus and Minmi, this element plays an undetermined role in the formation of the sagittally oriented septum subdividing the paired respiratory pathways (the internasal cavity septum) and the rostrolaterally oriented wall (lamina rostralis orbifalis) that encloses the rostral margin of the orbital cavity. Additionally, a ventral process of the frontal rnay contribute a variable component to the postocular shelf (Carpenter et al., in press). Furthermore, Maryanska (1977) suggests that the interna1 surface of the subadult Pinacosaurus frontal also bore transverse ridges corresponding to the rostralmost limits of the basicranium, and shallow furrows near the frontal-parietal contact demarcating the articulation point for the laterosphenoids. Grooves within the sagittal septum subdividing the paired respiratory paths of Talarurus have been interpreted as the conduits for transrnitting the olfactory nerves (Tumanova, 1987).

3.2.2.5 Os parietak (Figs. 3. IA, 3.2A, 3.5A13.7A1 3.7C1 3.8, 3.10) The parietal is situated as the caudalmost basicranial roofing element, additionally forming the medial portion of the occipital margin. Although rarely delineated, individual parietal elements have been described for subadult Pinacosaurus, Minmi, and the Cedar Mountain ankylosaur. Unlike most ornithischians (Romer, 1956; Sereno, 1991) including Minmi and the Cedar Mountain ankylosaur, the rectangular parietal of Pinacosaurus is paired, illustrating a distinct caudal continuance of the sagittal suture rostrally, flanked by the premaxillae, nasals, and frontals (Fig. 3.8A). In dorsal view, each Pinacosaurus parietal articulates with a laterally positioned squamosal and a caudolaterally located "tabular" (Maryanska, 1971; 1977; see below, section 3.2.2.10). Rostrolaterally, the parietal makes a short contact with the postorbital. In occipital view, the parietal is situated dorsal to the supraoccipital, with each element articulating via a single, ventrally projecting knob-like process. This process is received within a shallow juxtaposed parasagittal depression along the dorsal surface of the supraoccipital. Contact between the parietal and the supraoccipital is generally obscured from dorsal view by a caudally projecting nuchal shelf (rnargo nuchae) (Chapter 7, section 7.2, character 4 2; see also Fig. 3.10). The nuchal shelf is frequently characterised by transversely oriented, raised craniai sculpturing (Chapter 7, section 7.2, character 11 ). When viewed dorsally, the unpaired parietals of both Minmi and the Cedar Mountain ankylosaur are trapezoidal, with the lateral margin of each following a rostromedial course from caudal to rostral (Figs. 3.8B1C). The parietal of Minmi contacts the frontal rostrally, the "squarnosal" rostrolaterally, and a pair of unidentified elements ("dermal ossicles", Molnar, 1996) laterally. The nuchal shelf does not project beyond the occiput, and consquently the occipital condyle and the paroccipital processes are visible dorsally. In occipital view, the parietal contacts the supraoccipital and the exoccipital (viz. the paroccipital process), although the nature of this union (Le. fused or sutured) is unclear (Fig. 3.10). The Cedar Mountain ankylosaur parietal (CEUM 10332) contacts the frontal rostrally, the supraorbital(s) rostrolaterally, and the squarnosal laterally (Fig. 3.8C). The parietal receives a minor contribution to the lateral margin via a narrow process of the postorbital, interposed between the supraorbital and the squamosal. It is unclear as to whether the nuchal shelf obscures the occiput dorsally. Poor preservation precludes description of contact between the parietal and the basicranium. Specimens of both Euoplocephalus (AMNH 5405; see Coombs, 1971 ) and subadult Pinacosaurus (Maryanska, 1977) are purported to have had a mobile contact between the parietal(s) and the supraoccipital. Coombs (1971 ) suggested that this joint might permit a limited degree of cranial kinesis (metakinesis) in this Euoploephalus specimen, although he was unable to find additional examples within the genus. Personal observations have similarly been unable to identify this mobile contact in any examined specimen, including AMNH 5405 (Chapter 4, section 4.2.2.5). The condition in Pinacosaurus remains unexplored.

3.2.2.6 Os squamosum (Figs. 3.1A, 3.2A, 3.3A1 3.4A13.7A1 3.78, 3.8, 3.9, 3.10) Together with the ventrally positioned quadratojugal (section 3.2.2.7),the dorsally located squamosal forms the caudolateral margin of the cranial ornamentation. Variable expression of squamosal morphoiogy, and elaborations thereof (section 3.5), may (e.g. Ankylosaurus) or may not (e-g.Edmontonia) obscure the occiput in dorsal view. The squarnosal is flanked (as evidenced in subadult Pinacosaurus, Minmi and the Cedar Mountain ankylosaur) by the parietal medially and, reportedly, one or more unusual ossifications and I or neomorphic elements caudally (Fig. 3.8). Additionally, the squamosal of most taxa contributes to a distinctive cranial adornment (processus cornuum squamosorum), typically a pyramidal-like horn core (e.g. Ankylosaurus) or knob- like boss (e.g. Shamosaurus) Chapter 7, section 7.2, character 6). The irregularly-shaped squamosal of Pinacosaurus is rostrolaterally enveloped by the postorbital (Figs. 3.7A, 3.8A). In dorsal view, the squamosal contacts a purported "tabular" (caudornedially) and a second element of uncertain homology (the so-called "secondary dermal-squamosal" [Maryanska, 19771, positioned caudolaterally). However, persona1 examination of several subadult Pinacosaurus specimens revealed neither the "tabular" or the alleged "secondary dermal-squamosal" as distinct frorn the squamosal proper. Consequently, for the purposes of this treatment, the "tabuiar" and "secondary dermai-squamosal" are subsumed within the squamosal proper. In occipital view, the squarnosal contributes to the dorsolateral aspect of the occipital margin, articulating synarthrotically with the quadratojugal dorsal to the latter's contact with the quadrate (Fig. 3.1 OA). The squamosal boss is pyramidal. Ventrally, a large cup-shaped depression, the quadrate articulation surface (coty/usquadrati squamosalis), receives the squamosal condyle of the quadrate and the dorsal margin of the paroccipital process. Maryanska (1977) suggests that this junction was mobile, providing subadult Pinacosaurus with some degree of streptostyly, although this interpretation awaits further investigation. Sutura1 contacts between the squarnosal and the postorbital, quadratojugal and parietal are al1 serrated. The "squarnosal" of Minmi appears to represent a coalescence of the squamosal proper, the postorbital, and the medialmost supraorbital (= postfrontal; Molnar, 1996). In dorsal view this complex is hexagonal, contacting a supraorbital rostrolaterally, the prefrontal rostrally, the frontal rostromedially, and the parietal caudomedially (Fig. 3-88). The caudal margin of the squamosal conjoins with a small, oblong ossification Molnar (1996) refers to as a "dermal ossicle". Homology of this neomorphic element remains uncertain. Laterally, the "squamosal" is triangular, with the pyramidal apical embellishment directed caudodorsatly. Together, the "squamosal" and the jugal form the entire postorbital profile (Fig. 3.78). In occipital view, the "squamosal" adjoins with the parietal medially and the paroccipital process of the exoccipital ventrally (Fig. 3.1OB). With the exception of the jugal - "squamosal" scarf joint, none of the sutura! contacts delineating the "squamosal" are clear. The squamosal (CEUM 10345) of the Cedar Mountain ankylosaur is not well preserved (Fig. 3.8C). However, it is apparent that the squamosal protuberance is rounded, not pyramidal, and that the ventral surface receives the squamosal condyle of the quadrate.

3.2.2.7 Os quadratojugale(Figs. 3.1A, 3.2A, 3.3A, 3.4A13.7A, 3.7C13.8A1 3.9, 3.1 O) Situated as the ventral counterpart of the squamosal protuberance, the quadratojugal is the location of the caudoventralmost contribution of the cranial ornamentation. In some taxa (e.g. Saichania), the quadratojugal is frequentty misinterpreted as overgrowing the laterotemporal fenestra (e.g. Tumanova, 1987; Carpenter et al., in press). However, white the quadratojugal rnay obscure the laterotemporal fenestra laterally (Coombs, 1971; Coombs, 1978; Coombs and Maryanska, 1990), this opening is never completely obstructed. Virtually al1 ankylosaur taxa (with the possible exception of "Struthiosaurus") develop quadratojugal protuberances (processus cornuum quadratojugaiis), ranging in morphology from rounded bosses (e.g. Edmontonia) to deltaic flanges (e.g. Pinacosaurus) (Chapter 7, section 7.2, character 7). The quadratojugal protuberance may (e.g. Pinacosaurus) or rnay not (e.g. Edmontonia) obscure the articular coridyle of the quadrate in lateral view. In occipital view, the quadratojugal lies lateral to, and in approximately the same transverse plane as, the quadrate (Coombs and Maryanska, 1990). Contact with the quadrate occurs near the articular condyle, suggesting to at least one previous author (Russell, 1940) that the quadratojugal contributes to the jaw joint. Although specimens of subadult Pinacosaurus illustrate the morphology (including sutura1 delineation) of the quadratojugal, previous descriptive interpretations appear to be in error. Maryanska (1979; 1977) originally described the quadratojugal of Pinacosaurus as a small elernent, contacting the quadrate medially, the jugal rostrally, the postorbital rostrolaterally, and a triangular "secondary dermal plate" laterally (also see Coombs, 1978; Coombs and Maryanska, 1990, for similar interpretations concerning other taxa; Fig. 3.7A). This "secondary dermal plateJ1in turn contacts the squamosal dorsally (specifically, the element identified as the "secondary dermal-squamosal"; Maryanska, 1977). However, the purported "dermal plateJ'fused to the quadratojugal is, in fact, an outgrowth of the quadratojugal proper (see section 3.5). The quadratojugal of Minmi, as interpreted by Molnar (1996), does not contribute significantly to the lateral surface of the cranium. However, given the presence of a large, deltaic flange at the caudoventral corners of the cranium, this conclusion seems highly suspect. In occipital view the quadratojugal contacts the quadrate medially, and the "jugal complexJ1laterally (Fig. 3.1 OB). In the Cedar Mountain ankylosaur, the quadratojugal (CEUM 10417, CEUM 10561) contacts the quadrate medially. Poor preservation prevents further cornmentary.

3.2.2.8 "Os tabulare" (see Fig. 3.8A) The "tabular" is a purported temporal region elernent, located along the occipital margin of a subadult Pinacosaurus specimen (2.Pal. No. MgD-III1: Maryanska, l97l; 1977). Dorsally, this element is situated caudolateral to the parietal and caudomedial to the "squamosal complex" (= squamosal + "secondary dermal-squamosal": Maryanska, 1971; 1977; Fig. 3.8A). However, the presence of a "tabular" element cannot be confirmed in other subadult Pinacosaurus specimens (IVPP uncatalogued). While this doeç not preclude the existence of such an unusual ossification, it does render its occurrence as questionable. 'Tabulars" remain unreported in other ankylosaur taxa. Given the undemonstrative nature of the "tabular" elements, the homologies clearly have yet to be established. The tabular element proper is not present in any living tetrapod taxon. In addition, the genealogically closest taxa reported to possess a tabular (basal diapsids such as araeoscelidids [AraeoscelM and younginiforms [Youngina]that lie well outside the Archosauria [sensu Romer, 1956; Carroll, 1988; Evans, 1988; Benton, 19981) remain phytogenetically distant from any dinosaur. Therefore, the term "tabular" is used in the nominal sense only, with no implicit assumption of homology.

3.2.3 Palatal Region (Fig. 3.3)

Frequently the exact configuration of the elements contributing to the palatal region is rendered indistinct through the obliteration of cranial sutures, poor preservation, and incomplete preparation. Details provided herein are generally drawn from numerous specimens that demonstrate particular structures and / or contacts to varying degrees.

3.2.3.1 Ossa vomera (+ septum cavitas intemasalis) (Figs. 3.3A, 3.4A, 3.9) Bisecting the palatal region in the sagittal plane, the vomer contributes to a vertically oriented septum (sepfum cavitas internasalis) (= internasal bony septum, Tumanova, 1993; = vomer plate (sic), Carpenter et al., 1998) that partitions the palatal vaults, and variably subdivides the entire rostmm (see below). Along the ventralmost (palatalmost) margin (the vornerine keel, see below), a faint longitudinal intervomeral suture may (e-g. Pinacosaurus) or may not (e.g. Gasfonia), be evident. Several authors (e.g. Carpenter el al., 1998; Kirkland, 1998) have implied that in some taxa the vorner is unpaired. However, this conclusion appears to be erroneously based upon the obfuscation of the intervomeral suture. Rostrally, the vomer contacts the premaxilla at the caudal margin of the interpremaxillary suture (section 3.2.A.1) and spans the rostromedial margin of the interna1 naris. A reported rostral contact with the prevomer element (e.g. Lee, 1996) is herein considered erroneous (section 3.2.3.5). The premaxilla may preclude rostrolateral contact between the vomer and the maxilla (e.g. Pinacosaurus; Fig. 3.9),although when conjoined (thereby uniting the vomer with the lingual maxillary shelf; section 3.2.1 -2)the vomer contributes to the rostrodorsal secondary palate (e.g. Ankylosaurus; see Coombs, 1971; Coombs and Maryanska, 1990). The vomer continues caudally, contacting the rostral face of the pterygoid. Any relationship between the vomer and the basicranium (viz. the interorbital ossifications) remains uncertain. Caudolateralty, the vomer frequently contributes to the caudoventral secondary palate (section 3.2.3.3),along with the palatine, pterygoid and maxilla, although none of these contacts are preserved well enough for description. As noted above, the vomer forms the ventral extension of the internasal respiratory septum. Dorsally, this partition may (e.g. Edmontonia [Coombs, 1971; 1978; Coombs and Maryanska, 1990], Tsagantegia [Tumanova, 9931) or rnay not (e.g. Pawpawsaurus [Lee, 19961, Gargoyleosaurus [Carpenter et al., 19983) contact the cranial roof, thereby subdividing the paired respiratory passages (cavum nasi) (Chapter 7, section 7.2, character 20). Dorsal to the palatal vault, the internasal respiratory septum receives variable contributions from the nasal (section 3.2.1 -3) and frontal (section 3.2.2.4), although the morphology and distribution of the sutura1 contacts remains uncertain. In many ankylosaurs the vomerine keel (carina vomeris) (the ventrally projecting palatal margin of the conjoined vomers), extends beyond the level of the rnaxillary tooth rows (e.g. Edmontonia [TMP 98.71 .Il,Saichania [Maryanska, 19771). However, the condition in most taxa is unclear, owing to variable damage along this margin. In cross-section, the morphology of the ventral margin of the vomerine keel is variable, ranging from wedge-shaped (e.g. Edmontonia; Carpenter, 1990) to incised (with an inverted 'VI-shaped notch; e.g. Nodocephalosaurus; Sullivan, 1999) to spherical (e.g. Panoplosaurus; Carpenter, 1990). 3.2.3.2 Os palatinum (+ caudoventral secondary palate) (Figs. 3.3A, 3.4A, 3.9) The palatine resides within the coronal plane, situated caudal to the intemal naris, medial to the maxilla and jugal, and rostromedial to the pterygoid and ectopterygoid. Medially, the palatine contacts the vomer, and (reportedly) the rostralmost aspect of the basicranium (Maryanska, 1977). Along the ventral surface, the palatine demonstrates a variety of shallow crests, troughs and depressions (Coombs, 1971 ; Maryanska, 1977), although poor preservation and inconsistent preparation frequently obscure any detail. Maryanska (1977) exarnined the fractured palates of a variety of Asian ankylosaur taxa (including Saichania, Tarchia and subadult Pinacosaurus), and suggested that pneumatic cavitations ramify the palatines proper. Each palatine may (e.g. Saichania, Tianzhenosaurus) or may not (e.g. Edmontonia, Pawpawsaurus) form a smal 1, circular embayrnent along its junction with the rostral face of the pterygoid (the "posterior palatal foramen (vacuity]": Maryanska, 1971; 1977; Sullivan, 1999; Fig. 3.9). In addition to the margin of the internal naris and palatal vacuity, the palatine conjoins with the maxilla and vomer to form the caudoventral secondary palate Chapter 7, section 7.2, character 21). Extending caudally, this ridge begins to flare medially near the midpoint of the maxillary tooth row, forming a concave margin. The internal naris and the termination of the nasal cavity proper is located deep within the vestibule, defined laterally by the longitudinal ridge and caudally by the concave margin (Coombs, 1971; 1978; Coombs and Maryanska, 1990). The exact configuration and subsequent contribution of each individual osteological component forming the caudoventral secondary palate cannot be ascertained.

3.2.3.3 Os pterygoideum (Figs. 3.3A, 3.4A, 3.9) Positioned as the transverse partition demarcating the caudal termination of the palate, the pterygoid demonstrates a characteristically complex morphology that appears to be correlated, in part, with the presence (or absence) of a caudoventral secondary palate. Coombs (1971 ) divided the pterygoid into three major components; a central body contacting the vomers (rostrodorsally) and the basicranium (caudodorsally), a rostrolateral mandibular rarnus, and a caudolateral articulation with the quadrate. The central body (corpus pterygoideus) may be further subdivided into a rostral face and a caudal margin. Orientation of the rostral face corresponds to the development of a caudoventral secondary palate (Chapter 7, section 7.2, character 27). In taxa without the caudoventral secondary palate (e.g. Edmontonia, Pawpawsaurus) (Chapter 7, section 7.2, character 21 ), the pterygoid forms a caudoventrally oblique concave surface that issues from the interna1 naris. A sagittal ridge subdivides the paired pterygoid vaults, continuing caudally from the vomerine keel. At the caudal extremity of this midline ridge, the pterygoids converge to a ventrally projecting subtriangular zenith. Morphology of the interpterygoid suture is not discernible. The oblique caudoventral orientation of the rostral face of the pterygoid conceals the rostral portion of the basicranium and the rnorphology of the caudal margin of the pterygoid body in ventral view. Alternatively, taxa in which a caudoventral secondary paiate is present (e.g. Ankylosaurus, Saichania) show a vertical orientation of the rostral face of the pterygoid. However, this lamina is not oriented in a strict transverse manner, but rather forms a 'W'-shaped wall when viewed ventrally. From the sagittal point of contact with the vomer, the pterygoid juts caudolaterally, followed by an immediate realignment rostrolaterally. The rostrolateral apex subsequently melds with the mandibular ramus. Caudally adjacent to the vomeral - pterygoid contact, a deltaic cleft, the interpterygoid vacuity (incisura interptevgoideus) (= interpterygoid depression, Tumanova, 1993),demarcates the junction between the individual pterygoid elements. Lateral to the interpterygoid vacuity, a parasagittal foramen of uncertain homology pierces the central body of each pterygoid. The caudal margin of the pterygoid body may (e.g. Saichania) or rnay not (e-g. Ankylosaurus) be fused with a basipterygoid process (section 3.2.4.6; Chapter 7, section 7.2, character 30). Similar to the rostral face of the pterygoid body, morphology cif the mandibular ramus (processus mandibularis pterygoideum) (= mandibular brace; Coombs, 1971 ) is wrrelated with the development of the caudoventral secondary palate (Chapter 7, section 7.2, character 29). In taxa without this secondary palate (e.g. Edmonfonia), the mandibular ramus is directed parasagittally from the rostrolateral surface of the pterygoid (Coombs, 1971; i978). The mandibular ramus of taxa with a caudoventral secondary palate (e.g. Ankylosaurus) arises from a more medial position along the pterygoid body, with a distinct rostrolateral orientation. In both situations, the mandibular ramus contacts the ectopterygoid near the terminal limits of the alveolar border of the maxilla. The quadrate process of the pterygoid (processus quadraticus pferygoideum) is a transversely deep, rostrocaudally narrow brace joining the (dermato)cranium to the suspensory apparatus. Corretated with the presence of a caudoventral secondary palate, both the quadrate process and the mandibular process arise at a common position along the lateral margin of the pterygoid body. In taxa without a caudoventral secondary palate, the quadrate process originates from the caudolateral margin of the pterygoid body (see also section 3.2.4.9). The pterygoid of the Cedar Mountain ankylosaur (present on the holotype, CEUM 12360*; Carpenter et al., in press) is described as being longitudinally elongate, with a hook-like process at the caudolateral margin. Carpenter et al. (in press) refer to this recurved process as an analog of the trochlear process of pleurodire testudinates, a structure otherwise unknown amongst members of the Dinosauria. Unfortunately, the rnorphology of this purported process remains unclear owing to the poor presenratian of the specimen. Furthermore, the orientation and morphology of both the mandibular ramus and the quadrate process cannot be ascertained.

3.2.3.4 Os ectopterygoideum (Fig. 3.3A, 3.4A, 3.9) The ectopterygoid is a small, wedge-shaped element, interjected between the mandibular ramus of the pterygoid ventrornedially, the palatine dorsally, and the maxilla laterally (although Carpenter et al. [in press] report that the ectopterygoid of the Cedar Mountain ankylosaur contacts the lacrimal, not the maxilla laterally; section 3.2.3.3).While the precise limits of the ectopterygoid are difficult to discem on any specimen, shallow furrows permit a tentative identification of the aforementioned contacts. The ectopterygoid variably contributes to the rostral border of the suborblital fenestra (fenestra suborbitalis) and either the caudoventral secondary palate (e.g. Ankylosaurus) or (in those taxa without this additional "palate") the caudal margin of the interna1 naris (e.g. Edmontonia).

3.2.3.5 "Os prevomera" ("osparadoxum") (not figured) The prevomer is an infrequently referenced element, purported to contribute to the palatal surface. Although virtually indistinguishablefrom the rest of the premaxillary palate, Lee (1996) identified prevomers, including a prevomeral formen, on the cranium of Pawpawsaurus. Subsequent investigation (including examination of a cast of the holotype, SMU 73203; see afso Carpenter and Kirkland, 1998) does not corroborate Lee's findings, however, and the presence of a "prevomer" in Pawpawsaurus is herein considered doubtful. The prevomer identified by earlier authors (Gilmore, 1930; Russell, 1940) is actually a vomer.

3.2.4 Occipital IBasicranial Region (Figs. 3.4, 3.5)

Although devoid of the characteristic osseous ornamentation previously noted for both the rostral and temporal regions, demarcation of individual elements in the Occipital 1 Basicranial Region is similarly ambiguous. Extensive fusion, particularly between elements of the basicranium proper, and incomplete preparation obscure most sutura1 contacts, and many identifications remain tentative. 3.2.4.1 Os supraoccipitale (Figs. 3.3A, 3.4A, 3.8A, 3.1 0) The supraoccipital is an unpaired trapezoidal element forming the dorsal rim of the foramen magnum. In occipital view, the supraoccipital contacts the exoccipital and opisthotic rostrolaterally and the parietal dorsally. The presence of epiotics, discrete centres of ossification developing independent of, but subsequently coalescing with, the supraocciptals (Currie, 1997a), is unreported in the Ankylosauria. A narrow, variably developed inverted 'Yshaped crest is oriented vertically between the foramen magnum and the nuchal shelf of the parietal. In most taxa the sagittal component (crista nuchalis sagittalis) bifurcates immediately dorsal to the foramen magnum, forming a pair of caudolaterally inclined awning-like crests (crisfa nuchalis transversa). The rostrocaudal pitch of these transverse crests correlates with the orientation of the foramen magnum (i.e. a caudoventralfy directed foramen magnum has caudoventrally oriented transverse crests; see also Chapter 7,section 7.2, character 37). Frequently, paired fossae with subtending rugosities flank the sagittal crest.

3.2.4.2 Os exoccipitale (+ processus paroccipita/is) (Figs. 3.3A, 3.4A, 3.5A 3.9,3.1 0) Constituting the caudalmost parasagittal component of the occipital / basicranium region, the exoccipital flanks the foramen magnum and, along with the opisthotic, contributes to the paroccipital process (processus paroccipitalis). This prorninent, dorsoventrally deep transverse buttress spans the occiput between the foramen magnum and the quadrate. Long axis orientation of the paroccipital process is variable, ranging from strictly transverse (e.g. Ankylosaurus) to caudolaterally oblique (e.g. Edmontonia) (see also Chapter 7, section 7.2, character 33). Additionally, the exoccipital may provide a caudolateral contribution to the occipital condyle (section 3.2.4.3; Chapter 7, section 7.2, character 34). This condition is rnost pronounced in Minmi (Molnar, 1996), with each exoccipital forming a third of the condyle. Along the occiput, the exoccipital articulates with the supraoccipital dorsomedially, the basioccipital ventromedially, and the quadrate laterally (Fig. 3.1 O). The exoccipital (viz. the paroccipital processes) may (e.g. Ankylosaurus) or may not (e.g. Tarchia) fuse with the quadrate near the squamosal condyle (Chapter 7, section 7.2, character 39). Rostrally, the exoccipitat coalesces with the opisthotic, obliterating al1 sutura1 contacts. Maryanska (1977) reports that the exoccipital of subadult Pinacosaurus contributes to both the caudoventral margin of the fenestra ovalis and the caudal wall of the inner ear. However, this conclusion is Iikely in error, as these features are more commonly associated with the opisthotic (Romer, 1956; Starck, 1979; see also 3.2.4.4). In Pawpawsaurus, three distinct foramina reportedly (Lee, 1996) perforate the exoccipital - basioccipital synarthrosis. Lee (1996) interpreted the rostralmost opening as the comrnon foramen for the glossopharyngeal (IX), vagus (X) and accessory (XI) nerves, the median opening for the exit of the interna1 jugular vein, and the caudalmost opening for the hypoglossal (XII) nerve. In contrast, Tumanova (1987) identified the foramina for cranial nerves IX, X and XI in several Asian ankylosaurs (e.g. Amtosaurus, Pinacosaurus, Saichania, Talarurus and Tarchia) as occurring along the exoccipital - opisthotic contact (Fig. 3.5). In most dinosaurs, however, the exoccipital - opisthotic complex reportedly provides the rostrodorsal border of the foramina for cranial nerves IX, X, and potentially XI, while the foramen for cranial nerve XII perforates the exoccipital proper (Currie, 1997a). In dorsal view, the nuchal shelf of the parietal(s) rnay (e.g. Ankylosaurus) or may not (e.g. Edmontonia) conceal the caudolateral extremities of the paroccipital processes (Chapter 7, section 7.2, character 12).

3.2.4.3 Os basioccipifale (+ condylus occipitalis) (Figs. 3.3A, 3.4A, 3.5A13.9, 3.1 0) The basioccipital is an unpaired element contributing to the ventralmost margin of the foramen magnum, the caudoventral portion of the basicranium proper and the majority of the occipital condyle (condylus occipitalis). In occipital view, the stout occipital condyle varies in morphology, ranging from reniform (e.g. Ankylosaurus) to ovoid I round (e.g. Edmontonia) (Chapter 7, section 7.2, character 35), and may be oriented either directly caudally (e-g. Tarchia) or caudoventrally (e-g. Saichania) (Chapter 7, section 7.2, character 36). A shallow, sagittally oriented furrow, the median condylar notch (incisura mediana condyli) is situated along the mid-dorsal margin of the occipital condyle (corresponding to the mid-ventral rim of the foramen magnum). Contacts between the basioccipital and the exocciptal - opisthotic complex along the occiput are frequently obscured, although shallow furrows occasionally delineate each element. The basioccipital frequently provides the caudoventral border for the foramina of cranial nerves IX, X and XI (Tumanova, 1987; Fig. 3.5; however, see Lee, 1996 for a different interpretation). Ventrally, the basisphenoid articulates with the rostral margin of the basioccipital, and contributes to the basal tubera (section 3.2.4.5). Caudal to the basal tubera the basioccipital becomes sellar, and may give rise to a variably short "neck" leading to the occipital condyle (see Chapter 7, section 7.2.7).

3.2.4.4 Ossa otica (os prooficum, os opisthoticum) (Fig. 3.5A) The otic eiements (the prootic and opisthotic) form the lateral walls of the basicranium, harbouring the vestibular apparatus and perilymphatic spaces (Le. the inner ear). Despite providing variable contributions to the osseous otic complex of other dinosaurs (Sereno, 1991; Baumel and Witmer, 1993; Currie, 1997a), neither the epiotic nor the metotic has been reported for members of the Ankylosauria, and their presence within this clade remains equivocal. Furtherrnore, the columella (stapes) has yet to be identified. With few exceptions the sutural contacts between the prootic and opisthotic are obliterated, although the relative configuration of each element may be inferred by the position of the intervening fenestra ovalis (fenestra vesifibuli). Maryanska (1977) described the prootic from a specimen of Pinacosaurus (2.Pal. No. MgD-II) as a massive elernent, contacting the laterosphenoid rostrally, the basisphenoid rostroventrally, the supraoccipital caudodorsomedially and the opisthotic and exoccipital caudally. All of the sutural contacts, Save that of the prootic - supraoccipital union, are reportedly unfused. The foramen for the trigeminal nerve (V) is situated along the prootic - laterosphenoid contact, while the prootic - basisphenoid juncture hosts the foramen for the facial nerve (VII) (Fig. 3.5). Unfortunately, persona1 examination of several uncatalogued specimens of subadult Pinacosaurus was unable to confirm any of these interpretations, owing to poor preservation. The prootic of Pawpawsaurus (Lee, 1996) is described as a crescentic element, articulating with the laterosphenoid rostrally, along the concave border, and the ventral basisphenoid, caudal opisthotic and dorsal parietal, along the convex border. Similar to Pinacosaurus, the foramen for cranial nerve V is situated along the prootic - laterosphenoid contact. In contrast, however, the foramen for cranial nerve VI1 penetrates the wall of the prootic proper, just caudal to the foramen for cranial nerve V. Additionally, Lee (1996) reports that foramina for both the oculomotor (III) and abducens (VI) reside at the juncture between the prootic, laterosphenoid and basisphenoid. The opisthotic, positioned as the caudalmost component of the otic cornplex, articulates with the prootic rostrally, the exoccipital caudally, the basioccipital ventrally, and the supraoccipital caudodorsomedially (Fig. 3.5). lnvariably the opisthotic fuses with the exoccipital (Norman and Faiers, 1996; Currie, 1997a), and rnost descriptions of the latter element inadvertently include opisthotic contributions (e-g. the inner ear contribution Maryanska [1977] describes for the exoccipital appears to refer to the wnjoined opisthotic). The opisthotic, coupled with the aforementioned exoccipital, provides the rostral / rostrodorsal border of the forarnina for cranial nerves IX, X and XI (Tumanova, 1987; Norman and Faiers, 1996; Currie, 1997a; section 3.2.4.2; Fig. 3.5).

3.2.4.5 Os basisphenoidale (tuberculum basilare; processus basipterygoideus) ( Figs. 3.3A, 3.4A, 3.5A, 3.9, 3.1 0A) Forming the floor for the midbrain (i.e. the mesencephalon), the basisphenoid adjoins the parasphenoid rostrally, the laterosphenoid rostrodorsally, the prootic caudodorsally and the basioccipital caudally. Contact between the basioccipital and basisphenoid is readily demarcated by the basal tubera (tubercu/um basilare; = basitubera: Carpenter et al., in press) (section 3.2.4.3). Basal tubera morphotogy is variable, ranging from bulbous convexities (e.g. Pawpawsaurus) to rugose crests (e-g. Edmontonia) (Chapter 7, section 7.2, character 32). The basisphenoid may (e.g. Tsagantegia) or may not (e.g. Edmontonia) be longer than the basioccipital (Chapter 7, section 7.2, character 31 ). In ventral view, prior to contacting the parasphenoid, a parasagittal basipterygoid process (processus basiptevgoideus) diverges obliquely from the rostrolateral limits of the basisphenoid. This short, blunt process may (e.g. Saichania) or may not (e.g. Ankylosaurus) fuse with the rostrally adjacent caudal margin of the pterygoid body (Chapter 7, section 7.2, character 30). Caudoventral inclination of the pterygoid body frequently conceals the basipterygoid process from ventral view. Along the dorsally positioned laterosphenoid - prootic contact, the basisphenoid provides variable contributions to the ventral rnargins of three apertures; the foramina for cranial nerves III, VI and VI1 (sections 3.2.4.4 and 3.2.4.6; Lee, 1996; Norman and Faiers, 1996; Currie, 1997a; Fig. 3.5). The dorsal (cerebral) surface of the basisphenoid is distinguished by a prominent, sagittally positioned pituitary fossa (sella turcica). The caudal margin of the pituitary fossa is marked by the slender, transversely oriented dorsum sellae. Paired canals for the interna1 carotid (canalis caroticus cranialis; = vidian canals: Lee, 7996; Carpenter et al., in press) and palatine (canalis orbifalis) arteries emanate from the base of the pituitary fossa (see also Maryanska, 1977; Norman and Faiers, 1996). In some taxa, the foramina for cranial nerves VI (Pinacosaurus; Maryanska, 'l977), IV and VI 1 (cf. Polacanthus; Norman and Faiers, 1996) reportedly pierced the base of the pituitary fossa.

3.2.4.6 Os laterosphenoidale (Fig. 3.5A) The laterosphenoid contacts the parasphenoid and interorbital ossifications rostrally, the skull roof (i.e. the frontal - parietal complex) dorsally, the basisphenoid ventrally, and the prootic caudally (Fig. 3.5). Laterally, in addition to enclosing the forebrain, the laterosphenoid contributes transversely oriented braces to the postocular wall (section 3.2.2.2).A number of apertures for cranial nerves either pierce or receive variable marginal contributions from the laterosphenoid, including the foramina for the optic (II) and trochlear (IV) nerves (situated within the region of the interorbital - laterosphenoid confluence), the foramen for cranial nerve V (residing within the laterosphenoid - prootic synarthrosis) and the foramina for cranial nerves III and VI (piercing the laterosphenoid articulation with the prootic - basisphenoid complex) (Fig. 3.5).

3.2.4.7 Os parasphenoidale (Fig. 3.5A) The unpaired, sagittally positioned parasphenoid forms the rostroventral border of the basicranium proper, behveen the rostrolaterally oriented basipterygoid processes (section 3.2.4.5). Sutural contacts between the parasphenoid and adjacent elements (viz. the rostrodorsal frontal, caudodorsal interorbital ossifications and laterosphenoid, and caudal basisphenoid) are not visible in any taxon, and consequently contributions to the margins of cranial nerve foramina remain unclear. A prominent parasphenoid rostrum (rostrum parasphenoidale; = cultriform process, Romer, 1956; = presphenoid rod, Coombs, 1971; = basisphenoid rostrum, Tumanova, 1987) extends rostrodorsally from the parasphenoid - basisphenoid contact to underlie the caudal margin of the pterygoids (Fig. 3.5A). Ventrally, the parasphenoid rostrum may be viewed deep within the interpterygoid vacuity. The rostral extremity of the parasphenoid rostrum is contiguous with the internasal cavity septum.

3.2.4.8 Ossa interorbifale (= "os presphenoidale", "os orbitosphenoidale", "os sphenethmoidaley') (Fig. 3.5A) Within the ankylosaur orbit, as many as three basicranial elements (herein referred to as the interorbital ossifications) have been purported to contribute to the medialmost cavity wall. None of these three elements, the "orbitosphenoidn, "presphenoid", and "sphenethmoidn(= accessory ossification; Maryanska, 1977), can be clearly delineated in any specimen, and consequently al1 reported homologies are herein considered equivocal. In addition to defining the medial wall of the orbital cavity, the interorbital ossifications also transmit cranial nerve I and vascular channels, although there is little consensus on the reported identifications. The rostralmost of the putative interorbital ossifications is the "sphenethmoid" (Tumanova, 1987). That of subadult Pinacosaurus contacts the orbital process of the lacrimal (section 3.2.1.5) rostrolaterally within the orbit. Although indistinct, Maryanska (1977) reports that the "sphenethmoid" of (adult) Saichania articulates with the conjoined presphenoids and orbitosphenoids caudomedially. Caudomedial to the "sphenethmoid" is the "presphenoid". Contact between the "presphenoid" and the "orbitosphenaid" is typically indistinct, such that these two elements becorne integrated to form the "orbitosphenoid-presphenoid cornplex (= interorbital septum; Maryanska, 1977). In Saichania, each bilateral complex forms a parasagittally oriented wall that concurrently subdivides the contralateral orbital cavities. An "orbitosphenoid-presphenoid" complex has also been identified in specimens of Tarchia (Maryanska, 1977) and Pawpawsaurus (Lee, 1996). In both Saichania and Tarchia, foramina pierce the "orbitosphenoid- presphenoid" complex for passage of cranial nerve II and an ophthalmic vesse1 (artery and / or vein). The "orbitosphenoid-presphenoid" complex of Pawpawsaurus reportedly transmits the olfactory (1), optic (II) and trochlear (IV) nerves (Lee, 1996). Carpenter et al. (in press) identified an orbitosphenoid in the Cedar Mountain ankylosaur, pierced by cranial nerve II.

3.2.4.9 Os guadratum (Figs. 3.3A, 3.4A, 3.9, 3.10) The quadrate forms the principal element of metautostylic jaw suspension, positioned as a vertical brace between the cranium proper and the mandible (Romer, 1956; Kardong, 1998). In lateral profile, the rostrocaudally narrow body of the quadrate (corpus ossis quadrafi) may (e-g. Edmonfonia) or may not (e-g. Ankylosaurus) be bowed, with the rostral margin convex and the caudal margin concave (Chapter 7, section 7.2, character 38). Furthermore, the long axis of the quadrate body in many ankylosaurs appears to demonstrate a variable degree of rostrocaudal inctination, but it remains unclear as to whether this inclination represents the actual morphology of the quadrate or a taphonomic artefact (Coombs, 1971; Rybczynski and Vickaryous, in press). In occipital view, the quadrate is transversely broad. The dorsal extremity of the central body of the quadrate, the squamosal condyle (condylus squamos~],articulates with the quadrate articulation surface of the squamosal (section 3.2.2.6) and the distal ends of the paroccipital processes. The caudal surface of the squamosal condyle may (e-g. Ankylosaurus) or may not (e.g. Tarchia) fuse with the rostral surface of the paroccipital process (Chapter 7, section 7.2, character 39). Contact between the squamosal condyle of the quadrate and the quadrate articulation surface of the squamosal is presently unclear in the majority of specimens, owing to various degrees of incomplete preparation and poor preservation. Antipodeal to the squamosal condyle of the quadrate is the mandibular condyle (condylus mandibularis). Morphology of the mandibular condyle is highly variable, ranging from ovoid (e.g. Panoplosaurus) to bicondylar (e-g. Ankylosaurus), although taphonornic deformation of most specirnens prevents a phylogenetically detailed evaluation. Lateral to the mandibular condyle, the quadrate is embraced by a narrow buttress from the quadratojugal. Quadratojugal protuberances frequently conceal the mandibular condyle of the quadrate from lateral view (Chapter 7, section 7.2, character 40). Ventral to the midpoint of the quadrate body, a slender, rostromedially directed pterygoid process (processus pterygoideus) adjoins wi th the pterygoid. Although this junction is generally fused, several taxa preserve the contact as a scarf joint (e.g. subadult Pinacosaurus, Siivisaunis; Coombs, 1971). The pterygoid processes may (e.g. Ankylosaurus) or rnay not (e.g. Saichania) be in transverse alignment with the caudal margin of the pterygoids (Chapter 7, section 7.2, character 28). 3.2.4.1 0 "Os epipterygoideum" (not figured) The putative epipterygoid is a narrow, cylindrical ossification located adjacent to the lateral wall of the basicranium (Coombs and Maryanska, 1990). When present, the epipterygoid is situated rostroventrally oblique alongside the lateral surface of the basicranium. Distally, each end of the epipterygoid is dilated. Epipterygoids are poorly known amongst ornithischians, and are currently reported only from specimens of Saichania, Tarchia, Euoplocephalus (Chapter 4, section 4.5.9), subadult Pinacosaurus (Maryanska, 1977; Tumanova, 1987; Coombs and Maryanska, 1990) and some Asian pachycephalosaurs. The unusual disposition (in most dinosaurs the epipterygoids are vertically oriented) and somewhat arbitrary distribution of the epipterygoids in ankylosaurs calls into question the homology of this element.

3.3 Ossa Mandibulae

3.3.1 Mandibular Region (Fig. 3.6)

Unless otherwise stated, al1 mandibular elements are paired. The morphology of the ankylosaur mandible, similar to that of the cranium, is characteristically peculiar to the clade, with unusual, meandering tooth rows and a complete absence of external rnandibular fenestrae. Although frequently obscured by poor preservation, the majority of the sutures are visible (i.e. not obliterated). However, specimens are uncornmon (in comparison with crania) and detailed descriptions in the literature are rare. Consequently, the following cornmentary is limited to generalizations concerning a few disparate taxa. Unless otherwise noted, ail contacts are scari joints.

3.3.1.1 Ospredentaie (Fig.3.6A) The unpaired predentary is positioned as the mandibular counterpart to the premaxillary rostrum, forming a rostrocaudally broad, dorsoventrally depressed edentulous tomium (tomialis predentaiis). Crescentic in dorsal (and ventral) view, the predentary conjoins the paired dentaries rostral to the mandibular symphysis within a transverse depression (section 3.3.1-2). Evidence from at least one taxon (Euoplocephalus; Rybczynski and Vickaryous, in press) suggests that the predentary - dentary articulation permitted some degree of limited movernent (Le. amphiarthritic). A short, ventrally projecting sagittal protuberance of the predentary (tuberculum predentale) (Fig. 3.6A) contacts the mandibular symphysis. Both the dorsal and the ventral surfaces rnay be perforated by a variety of foramina and shallow foveae. The rostralmost edge of the predentary is irregularly scalloped.

3.3.1.2 Os dentale (Figs. 3.6B,C) The dentary is the only dentigerous element of the mandible (Appendix 3.2). The mandibular tooth row resides within the alveolar border (processus alveolaris [mandibulael), a dorsal crest running the length of the dentary. In dorsal view, the course of the alveolar border (and consequently the mandibular tooth row) ranges from virtually linear (e.g. Gargoyleosaurus) to a medially deflected parabola (e.g. Tarchia; Tumanova, 1987). Furthermore, in lateral profile the alveolar border invariably undulates dorsoventrally. Consequently, the overall morphology of the tooth row resembles that of a medially folded arch, with displacement in three-dimensions along its length (rostrocaudally, dorsoventrally and laterornedially). Rostrally, contralateral dentary elements articulate with one-another at the dorsoventrally abbreviated rami of the mandibular symphysis (pars symphysialis) (Fig. 3.68). The mandibular symphysis emanates from the main body of the dentary (corpus dentale) immediately rostral to the tooth row. Symphysis morphology relates to the degree of tooth row displacement: in taxa with a minimal1y fiexed alveolar border (e. g. Panoplosaurus), the symphysis is rostrocaudally long; in taxa with a more arched alveolar border (e.g. Tarchia), the symphysis is constricted and subcircular. Various ridges and rugosities along the symphyseal margin permit articulation of the juxtaposed contributions, and suggest a limited degree of mobility was possible (Rybczynski and Vickaryous, in press). Along the rostralmost border of the dentary, the locus for predentary contact is a transverse pitted depression (sulcus predentale), contiguous across the symphysis (section 3.3.1 -1; Rybczynski and Vickaryous, in press; Fig. 3.6C). In cross-section, this predentary depression is concave, with numerous foveae coalescing to excavate the caudal margin. ln lateral view, the dentary articulates with the surangular (caudally), the angular (caudoventrally), and the aforernentioned predentary (Fig. 3.6A). Nurnerous forarnina of uncertain homology pierce the dentary ventral to the alveolar border. Along the media1 surface, subparallel to the alveolar border, reside numerous "special forarnina" (Edmund, 1957))wherein the cusps of replacement teeth may be seen. Medially, the ventral portion of the dentary is concealed via contact with the splenial (Fig. 3.6B). Beyond the rostralmost extremity of the splenial, a longitudinal groove leading away from an aperture at the dentary- splenial contact continues towards the mandibular symphysis (Fig. 3.68). The homology of this "Meckelian groove" (= Meckelian canal, Coombs, 1971; = sulcus Meckelii, Tumanova, 1987) has yet to be established.

3.3.1.3 Os spleniale (Fig. 3.68) The splenial is positioned as a thin overlay of bone along the ventromedial surface of the mandible. Near the rostral terminus, at the splenial - dentary contact, the splenial contributes to the rostral aperture of the "Meckelian groove" (Fig. 3.6B). Ventral to the caudalmost tooth, the splenial is pierced by a large intermandibular foramen (foramen intermandibularis caudalis). The majority of the horizontally oriented dorsal margin of the splenial overlaps the medial surface of the dentary along the length of the alveolar border. Adjacent to the adductor fossa, the spleniat underlies and braces the rostroventral rnargin of the prearticular. Ventrally, the splenial has a lengthy contact with the angular, although osseous ornamentation obscures the morphology of this articulation (Fig. 3.68). 3.3.1.4 Os angulare (Fig. 3.6C) Diametrically opposite to the splenial, the angular resides along the ventrolateral margin of the mandible (Fig. 3.6C). In most taxa, the angular is completely embossed by osseous ornamentation (Chapter 6) and consequently the morphology of the sutura1 contacts with adjacent elements is unclear. However, in specimens of subadult Pinacosaurus that lack the extensive development of ornamentation, the lateral profile of the angular is triangular, with an elongated ventral base. Rostrodorsally, the angular articulates with the dentary while the caudodorsal margin adjoins with the surangular (Fig. 3.6C). In some taxa (e.g. Edmontonia), the angular wraps medially around the ventral margin of the mandible, and contacts the prearticular.

3.3.1.5 Os supra-angulare (Figs. 3.68, C) The surangular is situateci in the caudodorsolateral quarter of the mandible, contributing to both the lateral margin of the adductor fossa (fossa adductoris mandibulae) and the retroarticular process (section 3.3.1 -7; Fig. 3.6C). Furthermore, the surangular conjoins with the coronoid to form the coronoid process (contra Coombs, 1978; section 3.3.1.6). Laterally, the surangular articulates with the dentary (rostrodorsally), angular (rostroventrally) and prearticular (caudally) (Fig. 3.6C). Along its medial surface, the surangular adjoins the coronoid (rostrornedially) and articular (medially), and may make an abbreviated contact with the splenial (rostroventromedially). Contact between the angular and prearticular within the adductor fossa is unclear. Frequently the coronoid process of the angular is pierced by a large foramen of uncertain affinity.

3.3.1.6 Os coronoideum (Fig. 3.68) The coronoid is a small, rarely identified, chevron-shaped element contacting the medial surface of the alveolar border proximal to the adductor fossa, and the rostromedial margin of the coronoid process (processus coronoideus) (Fig. 3.6B). Along the margin of the coronoid process the coronoid is rugose.

3.3.1.7 Os prearticulare (Figs. 3.68, C) In conjunction with the surangular, the prearticular contributes to the abbreviated retroarticular process (processus retroarficularis) (Fig. 3.6B, C). In general, the retroarticular process is deeper than long, and bears rugosities along its entire caudal margin. Ventrally, the prearticular (including the retroarticular process) is underlain by the splenial. In medial profile, the prearticular contacts the splenial, and sometimes a rnedial process of the angular (e.g. Edmontonia) rostroventrally, the coronoid rostrodorsally and the articular dorsally. The prearticular forms the entire caudomedial margin of the adductor fossa.

3.3.9.8 Os articulare (Fig. 3.6B) Bounded by the surangular (laterally) and the prearticular (caudomedially), the articular accommodates the quadrate articular fossa (fossa articularis quadratica). Medially, a process of the articular (and a continuation of the quadrate articular fossa), the medial mandibular process (processus mandibulae medialis) extends perpendicular to the long axis of the mandible. A weakly developed ridge may subdivide the articular surface either mediolaterally (e.g. Edmontonia; Coombs, 1971 ) or rostrodorsally (e.g. Euoplocephalus; Rybczynski and Vickaryous, in press). The rostral terminus of the articular extends within the adductor fossa, and cannot be determined at this time.

3.4 Dentes

The dentition of ankylosaurs has received extensive treatment elsewhere (e-g. Coombs, 1971; 1990; Coombs and Maryanska, 1990; Galton, 1986; Weishampel and Norman, 1989; Carpenter, 1990; and Rybczynski and Vickaryous, in press), and hence the following contribution is a brief summary of pertinent morphology. Ankylosaur teeth are relatively small, in contrast to body size (e.g. Euoplocephalus is reportedly 6 - 7 m in body length with a cranial length of 460 mm [Coombs and Maryanska, 1990], yet the teeth attributed to this taxon are invariably less than 7.5 mm both mesiodistally and from the base of the cusp to the apex) with straight, unbranched roots. Premaxillary teeth are not common within the Ankylosauria (Appendix 3.2; Chapter 7,section 7.2, character 17), although when present they are usually conical and slightly recurved, with a smooth crown. Maxillary and mandibular teeth are laterally compressed, with apical denticulate carinae (8 - 17 denticles; Coornbs and Maryanska, 1990). Crown morphology may either be smooth or fluted. These vertical flutes or grooves may (e.g. Edmonfonia) or may not (e.g. Euoplocephalus) correspond to the disposition of denticles. Frequently, the base of the crown becomes swollen, forming a cingulum (Chapter 7, section 7.2, character 19). Unfortunately, in situ dentition in ankylosaurs is rare (Rybczynski and Vickaryous, in press), and morphological variation with reference to tooth position (e.g. maxillary versus dentary) remains ambiguous.

3.5 Summary

This treatment establishes the generalized pattern of skull morphology for the Ankylosauria. The cranium and rnandibular elements are readily diagnosable osteological constructs. The peculiar morphology of the skull is characterized by the secondary loss of several fenestrations otherwise considered ubiquitous within the Dinosauria, the near complete obliteration of sutura1 contacts between adjacent elements and the development of an all-encompassing cranial ornamentation. Arbitrary subdivision of the skull into discrete regions permits an effective review of al1 pertinent osteological features, in a logical, progressionary sequence. Although most individual elements mnnot be readily delineated, structural features (such as bosses, shelves or foramina) provide a means by which their rough position and morphology may be inferred. A synthesis of available information permits a reappraisal of the palatal and mandibular regions and the first ever detailed description of the basicranium. Furtherrnore, the architecture and taxonornic distribution of such composite structures as the secondary palates, the internasal respiratory septum and the basal tubera are addressed. A review of curent evidence does not support the identification of neomorphic elements such as "secondary dermal-squamosals", "tabulars", "prevomers" and "epipterygoids" as regular constituents of the ankylosaur head skeleton.

The descriptive procedure employed here provides the foundation for subsequent, more specific morphological analyses. The following two chapters adopt the aforementioned descriptive procedure and apply it to the morphological reappraisal of a taxon previously described from a nurnber of specimens (Chapter 4) and the primary description of the type and only known specimen of a heretofore unknown clade (Chapter 5). Figure 3.1 Schematic of the Euoplocephalus cranium illustrating the rostral region, and al1 the purported constituent elements thereof. A. Oblique rostrodorsolateral view. B. Lateral view. Rostral region defined by the area in advance of a transverse plane located at the rostral extremity of the orbital cavities (represented by the dashed-square in A. and the white vertical line in B.). See text (Chapter 3, sections 3.1 and 3.2.1)for details. Rostral region elernents expressed in bold font and bold outlines; non-rostral region elernents expressed in italic font and faint outlines. All abbreviations listed in Appendix 1.A.

Figure 3.2 Schematic of the Euoplocephalus craniurn illustrating the temporal region, and al1 the purported constituent elements thereof. A. Oblique caudodorsolateral view. B. Lateral view. Temporal region defined by the area caudal to a transverse plane located at the rostral extremity of the orbital cavities (represented by the dashed-square in A. and the white vertical line in B.). See text (Chapter 3, sections 3.1 and 3.2.2)for details. Temporal region elements expressed in bold font and bold outlines; non-temporal region elements expressed in italic font and faint outlines. All abbreviations listed in Appendix 1.1.

Figure 3.3 Schematic of the Euoplocephalus cranium illustrating the palatal region, and al1 the purported constituent elements thereof. A. Oblique caudoventral view. B. Ventral view. Palatal region defined by the elements situated along the ventral surface of the cranium, in advance of the basicranium proper, covering the majority of the roof of the oral cavity (represented by the dashed-square in A. and the white box in B.). See text (Chapter 3, sections 3.1 and 3.2.3)for details. Palatal region elements expressed in bold font and bold outlines; non-palatal region elements expressed in itaiic font and faint outlines. Al1 abbreviations listed in Appendix 1. 1 .

Figure 3.4 Schematic of the Euoplocephalus cranium illustrating the occipital / basicranial region, and al1 the purported constituent elernents thereof. A. Oblique caudoventral view. B. Ventral view. Occipital / braincase region defined by al1 the elements present in caudal view, including the basicranium proper and associated structures 1 elements (represented by the dashed-square in A. and the white box in B.). See text (Chapter 3, sections 3.1 and 3.2.4) for details. Occipital 1 braincase region elements expressed in bold font and bold outlines; non-occipital / braincase region elements expressed in italic font and faint outlines. All abbreviations listed in Appendix 1.A.

Figure 3.5 Schernatic of the Euoplocephalus basicraniurn, right lateral view. A. Basicranial components. B. Basicranial foramina. See text (Chapter 3, section 3.2.4) for details. All abbreviations listed in Appendix 1.1. inter orb latero fr /

exo -

for n II fen oval \ for n IV

for n XII

for n IX, X, XI Figure 3.6 Schematic Euoplocephalus mandible illustrating the mandibular region. A. Predentary, dorsal view. B. Medial view, right mandible. C. Lateral view, (mirror imaged) right mandible. Teeth and alveoli omitted. See text (Chapter 3, sections 3.1 and 3.3) for details. All abbreviations listed in Appendix 1.1. pro cor sur

art

"Meckelian"groove for interman caud 'Ir retro / / sulcus ~reart predent Figure 3.7 Schematic illustrations of the right lateral profile of ankylosaur crania discussed in the text. A. Pinacosaurus grangen (after Maryanska, 1971; 1977 and specimens). B. Minmi sp (after Molnar, 1996). C. Cedar Mountain ankylosaur (after Carpenter et al., in press). All abbreviations listed in Appendix 7.1. Not to scale. SOT ,lac Figure 3.8 Schematic illustrations of the dorsal profile of ankylosaur crania discussed in the text. A. Pinacosaurus grangen (after from Maryanska, 1971; 1977 and specimens). B. Minmi sp (after Molnar, 1996). C. Cedar Mountain ankylosaur (after Carpenter et al., in press). Al1 abbreviations listed in Appendix 1.1. Not to scale. ost? Figure 3.9 Schematic illustrations of the ventral view of Pinacosaurus grangen (after from Maryanska, 1971; 1977 and specirnens). All abbreviations listed in Appendix 4.1. Not to scale. fen posttemp exo Figure 3.10 Schematic illustrations of the occipital view of ankylosaur crania discussed in the text. A. Pinacosaurus grangen (after from Maryanska, 1971 ; 1977 and specimens). B. Minmi sp (after Molnar, 1996). All abbreviations listed in Appendix 1.1. Not to scale. "fen laterotemp" SOCC #

matrix Table 3.1. Element composition of each topographic region of the skull and a survey of element distribution amongst members of the Archosaufla. The clades Crocodylia and Aves include extant members; al1 others are strictly fossil taxa. Checkmarks indicate the presence of the element in question; question marks denote questionable presence. Redlined areas suggest the elernent is mis-identified and 1 or of dubious affinity. See text for details.

Table 3.2. Individual tooth counts for the various dentigerous elernents of the taxa discussed in the text. Question marks ("?") denote that these elernents have yet to be collected and 1 or are not preserved well enough to determine an approximate count; not applicable abbreviation ("da") denotes the confirrned absence of teeth. Taxa not presently under consideration shown in redline. Chapter 4 Morphological Re-description of the Skull of Euoplocephalus tutus Lambe (1902)

First collected over 100 years ago, Euoplocephalus tutus Lambe (1902) remains one of the best represented and most frequently discussed ankylosaur taxa (see Chapter 1). Characterized by a unique pattern of cranial sculpturing (see below; Chapter 6) and relatively small, fluted teeth, Euoplocephalus has corne to epitomize much of what is currently understood with regard to the Ankylosauria. Despite an abundant fossil record (more than 40 partial skeletons currently residing in museum collections around the world [Coombs and Maryanska, 1990]), a review of the literature suggests that a fundamental understanding of skull anatomy and architecture is not presently available. Euoplocephalus is herein morphologically reviewed and re-described on the basis of personal observations of 19 original crania, including the holotype NMC 021O*, and six mandibles (Appendix 4.1). This database was supplemented with photographs and previously published information pertaining to four specimens not personally examined. In addition to osteological details gleaned from traditional observalional methods, several specimens were subjected to non- invasive cornputer tomographic (CT) scanning, permitting the examination of numerous deeply situated structures. This chapter provides a synthesis of information obtained from the osteological examination of the aforementioned Euoplocephalus material, following the descriptive procedure outlined in Chapter 3, with the goal of providing a detailed qualitative morphological description of a representative member of the Ankylosauria. For the purposes of this treatment, except where noted, the skull morphology of Euoplocephalus closely resembles that of other ankylosaurs (see Chapter 3), and consequently, the details provided hereafter concentrate on novel developments and previously discussed character states (see also Chapter 7). 4.1 Overall Morphology

Sirnilar to other mernbers of the Ankylosauria, specimens of Euoplocephalus frequently, perhaps inevitably, undergo some degree of post- mortem plastic deformation (Chapter 2). Consequently, the rnorphology expressed by most specimens does not necessarily reflect the original (unmodified) skull condition. In an effort to circumvent the pitfalls associated with descriptions influenced by taphonomic deformation, the following effort relies heavily on the integration of data gleaned through the investigation of multiple specimens, with an acute awareness of comparative vertebrate osteology and the recognition of post-mortem events as a source of potential error. The skull of Euoplocephalus is known from at least 25 partial to virtually complete specirnens (Figs. 4.1, 4.2; see also Appendix 4.1 ), ranging in rostrocaudal length (for cornplete specimens only) from 305 to 41 1 mm (Appendix 4.1 ). In dorsal view (Fig. 4.2), the cranium roughly resembles an apically truncated equilateral triangle and consequently, the craniurn is wider than long (see Table 4.1 ; Chapter 7, section 7.2, character ?). Length of the mandible, from a line connecting the rostral-most extent of the premaxillary process of the dentary to the caudal-most tip of the retroarticular process, is approximately 70 - 80% of cranial length (Appendix 4.1 ). All the material examined is generally considered to represent adult or near-adult sized individuals, based on the ubiquitous developrnent of cranial ornamentation and the virtual obliteration of sutura1 contacts. The pattern of cranial sculpturing is highly distinctive (see below), arising caudodorsal to the external naris (and accommodating nasal vestibule) and extending caudad across the dorsolateral surface of the cranium (Figs. 4.1, 4.2, 4.3, 4.4A, 4.5; see also below and Chapter 6), and along the lateral surface of the mandible (viz. the angular). Rostroventral to the nasal vestibule however, the premaxilla is characteristically unornamented (Figs. 4.3, 4.5; see also section 4.2.1.1). Crowning each nasal vestibule, caudodorsal to the premaxilla, is a rugose, transversely elongate supranarial (vestibwlar) arch of ornamentation (Figs. 4.3A, 4.4A14.5), marking the abrupt initiation of cranial sculpturing. lmmediately caudodorsal to the supranarial (vestibular) arch, prominent furrows partition the ornamentation of the antorbital area into a unique mosaic consisting of one of two successive unpaired polygons (if both are present, the caudalmost is larger) in the sagittal plane, flanked by a pair of adjacent subrectangles (Figs. 4.3AI 4.4A). Caudolateral to the supranarial (vestibular) arch, furrows along the lateral aspect of the antorbit subdivide the area into two large polygons (Figs. 4.3A1 4.5A), and demarcate the wedge-like supraorbital boss (section 4.3.1 ; Chapter 7, section 7.2, character 9). Proximal to the orbits, furrows across the dorsum becorne increasingly reticulate, shallow and less symmetrically arranged, resufting in a random distribution of srnaller, ill-defined polygons (Fig. 4.2A). Along the temporal region, this escalating pattern of sculptured vagueness culminates in a completely amorphous texture, with the exception of conspicuous furrows segregating the variably aciculate squamosal bosses and quadratojugal projections (Chapter 7, section 7.2, characters 6 and 7; see also Fig. 4.2 and sections 4.3.5 and 4.3.6; Figs. 4.2, 4.3, 4.5A) and the nuchal shelf (Chapter 7, section 7.2, character 11 ; Fig. 4.3C).

Of the four pairs of cranial openings that typically characterize archosaurs, none of the antorbital, dorsotemporal or mandibular fenestrae are externally visible along the skull of Euoploeephalus, similar to the condition noted in other ankylosaurs. The laterotemporal fenestra is present (contra Sternberg, l928), although it remains partially obscured in lateral view by the squamosal- quadratojugal complex (Chapter 7, section 7.2, character 4; see section 4.2.2). The external naris is situated within a shallow, poorly defined nasal vestibule along the rostral surface of the prernaxilla (see section 4.3.1; Figs. 4.3A14.58). Although frequently characterked as being transversely elongate (Gilmore, 1923; Sternberg, 1929), the long axis of the external naris is actually vertically oriented (Fig. 4.3A14.58), with the former condition being the result of taphonomic deformation. The orbit is relatively small (diameter is generally less than 20% of cranial length; see Appendix 4.1 and Figs. 4.3A14.5A), and directed Izterally, and slightly rostrally. The antorbital area is linear in dorsal view (Fig. 4.2)) tapering medially towards the contact with the prernaxilla. The maximum distance between the supraorbital bosses (ranging from 314 to 364 mm) is subequal to the maximum distance across the squamosal bosses (ranging from 275 to 408 mm) (Chapter7, section 7.2, character 10) (Appendix 4.1). In lateral profile, the roof of the rostral region is dorne-like, while the roof of the temporal region is flat (Chapter 7, section 7.2, characters 2 and 3; see also Fig. 4.3At 4.5A).

4.2 Ossa Cranii

Unless otherwise noted, al1 cranial elements are paired. A number of unusually preserved specimens demonstrating either variable truncation (via taphonomic events) along the palate (e.g. TMP 97.59.1; Fig. 4.6) or fortuitous natural breaks (TMP 96.75.1; Fig. 4.7), permit examination of the internai surfaces of various elements of both the rostral and temporal regions. In addition, a specimen from the American Museum of Natural History, New York (AMNH 5337))that was previously physically sectioned (in a transverse plane across the antorbit) and various CT scanned specimens supplemented the descriptive effort.

4.2.1 Rostral Region(Figs.4.1B,4.3A14.5,4.6,4.7,4.8,4.8,4.9,4.10)

With the exception of a segment of the interpremaxillary suture, none of the synarthroses between adjacent rostral elements along the dorsum are visible (see Chapter 3; Fig. 4.5B). Therefore, topographic position and morphological expression of many of the rostral elements (viz. the nasal, prefrontal and lacrimal) cannot be directly inferred. In addition, presence of the nasolacrimal aperture has yet to be demonstrated in any specimen. However, the premaxilla and maxilla are demonstrated prominently along the palate, permitting a more detai led description.

4.2.1.1 Ospremaxillare(Fig. 4.1B1 4.4A1 4.5, 4.6, 4.7, 4.8, 4.9) The premaxilla is broad, edentulous (Chapter 7, section 7.2, character 17) and contributes to the medial, ventral and part of the lateral margins of the nasal vestibule (Fig. 4.5B). Although devoid of cranial ornamentation, the external surface of the premaxilla frequently bears a srnall number of shallow foveae and a network of variably oriented furrows (Fig. 4.4A, 4.5B). Dorsolateral to the nasal vestibule, cranial sculpturing obscures al! contacts with adjacent elernents (see above). The ventral margin of the premaxillary tomial crest is variably scalloped. In lateral profile the premaxillary tomial crest conjoins with the maxillary tomial crest along the lateral margin of the rostrum (Chapter 7, section 7.2, character 16) although the juncture is well defined by the abrupt development of cranial sculpturing caudolateral to the nasal vestibule (Figs. 4.3A, 4.5A). In ventral view, the premaxillary rostrum (= palatine, Lambe, 1902; Fig. 4.1B), is rectangular (Chapter 7, section 7.2, character 13) (Figs. 4.2B14.8). A premaxillary notch is present (Chapter 7, section 7.2, character 15; Fig. 4.58), resembling a small, narrow isosceles triangle, with the apex incising the interpremaxillary suture (Figs. 4.4, 4.5). Caudal to the premaxillary notch is a small (generally less than 3 mm in diameter) parasagittal incisive foramen. Although sornewhat variable, the incisive foremen of Euop/ocephalus is oriented dorsolaterally. Unfortunately, the course of each incisive foramen could not be traced in either truncated specimens (Fig. 4.6) or transverse CT scans (Fig. 4.9). Contact between the contralateral premaxillae frequently occurs within a shallow, sagittally oriented longitudinal depression (Fig. 4.8). The interpremaxillary suture is a butt joint. The prernaxilla is not perforated by the internat naris or any other large fenestration (confraNopcsa, 1928; Sternberg, 1929; see Fig. 4.8). Caudally, the premaxilla joins the vomer and the maxilla along the rostrodorsal secondary palate. Nopcsa (1928, Plate V, Figure 4) labelled this point of contact, between the premaxilla, maxilla and vomer, as the vomer proper. The premaxillary - maxillary contact (= palatine - pterygoid contact, Lambe, 1902; Fig. 4.1 B) along the palate is generally preserved as a shallow furrow, variably pierced by foramina, undulating from a lateral position rostromedially then caudornedially around the dentigerous embayment of the maxilla (i.e. the alveolar border of the maxilla). Contact between the vomers and the premaxillary- maxillary articulation is variably fused, and usually indistinct, although the vomers do not appear to invade the premaxillary palate. The premaxilla does not directly contact the palatine (contra Sternberg, 1929). Specimens with a truncated palate (Fig. 4.6) and CT scans conducted in the transverse plane of the rostrum (Fig. 4.9) dernonstrate an uninterrupted connection between the internasal bar (separating the nasal vestibule frorn the sagittal plane) and the rostralmost extension of the internasal cavity septum of the vorner (see also Figs. 4.7B, C). This uninterrupted subdivision gives rise to a linear, bilaminar sagittal partition between adjacent nasal cavities (see section 4.2.3.1 ). Within the nasal vestibule resides a rostrornedially positioned, vertically oriented suboval nasal aperture (the external naris) and a caudolaterally positioned, horizontally oriented elliptical paranasal aperture, segregated by a vertical process of the premaxilla (Chapter 7, section 7.2, character 24) (Fig. 4.58, 4.6, 4.9C).

4.2.1.2 Os maxillare (+ rostrodorsal secondary palate) (Figs. 4.4 B, 4.6, 4.8, 4.9, 4.1 O) The maxilla is the only dentigerous element of the cranium (contra Sternberg, 1929), with an average of 22 alveoli arranged in a single occlusal row along the alveolar border (Fig. 4.1 0;see Appendix 3.2). The flat "bony plates" identified by Sternberg (1929) in his description of NMC 8530' (the holotype of Anodontosaurus lambei; see Chapter 1; Fig. 1.1 ) actually represent the lateral surfaces of the maxillae, displaced ventraily by taphonomic distortion. Across the dorsurn, the morphological boundaries of the maxil!a are not visible, except at the aforementioned premaxillary tomial crest contact (see above). In lateral profile, immediately caudal to the premaxillary - maxillary contact, the maxillary tomial crest arches dorsally, thereby reducing the depth of the tomial margin and exposing the alveolar border of the maxilla (Figs. 4.3A, 4.5A). The shallow maxillary tomial crest continues caudad to adjoin with the jugal tuberosity (Chapter 7, section 7.2, character 16). Inset between, and running parallel with, the maxillary tomial crest and the adjacent alveolar border, the buccal emargination is a deepiy concave furrow (Chapter 7, section 7.2, character 22; see also Fig. 4.8). Along the ventromedial surface of the maxilla, a row of "special foramina" (Edmund, 1957) (diameter of each "special foramen" 80 - 90% that of the mesial - distal crown breath) parallels the course of the alveolar border (Fig. 4.1 OA). Rostrodorsomedially the lingual maxillary shelf (= pterygoid, Lambe, 1902), along with a rninor contribution from the vomer (see section 4.2.3.1), forms the rostral margin of the internal naris (= pterygoid vacuity, Lambe, 1902; Fig. 4. AB) and the rostrodorsai secondary palate (Fig. 4.8). (Chapter 7, section 7.2, character 21). The maxilla continues to circumvent the internal naris laterally, forrning a prominent medial longitudinal ridge. Along the palatal surface, the tooth rows (and alveolar borders) are medially inflected (Fig. 4.8). Maximum separation occurs at the caudalmost positions, with the rows converging towards the midpoint before diverging at the rostral extreme (Chapter 7, section 7.2, character 18). Tooth row length is less than 40% of cranial length (Appendix 4.1 ). In many specimens (e.g. AMNH 5405, TMP 97.132.1 ), an alveolar tuberosity is present at the caudal terminus of the tooth row (Fig. 4.5).

4.2.2 Temporal Region (Figs. 4.3, 4.5A, 4.7A, 4.11,4.12,4.13)

Similar to the rostral region, none of the sutura1 contacts between adjacent elernents of the temporal region are delineated on any of the crania examined. Therefore, discussion of individual elements is generally limited to elaborations thereof and / or structures they are presumed to contribute to. Information from intact specimens is supplemented by a number of isolated fragments of ankylosaur cranial material (viz. a supraorbital, squamosai and quadratojugal; Fig. 4.13), recently collected from the late Campanian sediments of Dinosaur Provincial Park, southern Alberta. Given the collection locality and morphology of these elements, they are considered herein to represent disarticulated remnants of Euoplocephalus. Furthermore, specimens truncated along the palate (Fig. 4.6) or fortuitously broken (Fig. 4.7) also permit examination of the deep structures of the temporal region (see section 4.2). The temporal region also provides a minor caudodorsal contribution to the nasal cavity proper, as it exits the rostral region and the bilaminar internarial septum.

4.2.2.1 Ossa supraorbitalia (processus cornus supraorbitalia) (Figs. 4.3 4.5A14.7A) The supraorbital(s) of Euoplocephalus is (are) characterized by the pronounced development of a dorsolaterally oriented longitudinal wedge-like boss, with a variably acute apex (Chapter 7, section 7.2, character 5). At present it is unclear if, similar to Pinacosaurus, Minmi and (maybe) the Cedar Mountain ankylosaur, the area dorsal to the orbital cavity includes more than one ossification. However, an isolated supraorbital referred to Euoplocephalus (TMP 88.1 06.5) suggests that the wedge-like boss is restricted to a single element (in contrast to the condition noted for Pinacosaurus; Chapter 3, section 3.2.2.1). This small (35 mm maximum length) subpyramidal element demonstrates a rostrocaudally elongate, laterally compressed keel leading to an aciculate apex and a continuous interdigitating sutura1 surface encircling the majority of the base. In both size and shape, TMP 88.1 06.5 closely resembles the combined rostral and lateral supraorbitals of subadult Pinacosaurus (see Figs. 3.7A. 3.8A). Unique to Euoplocephalus is the unusual quarter-sphere shaped modified "ciliary" supraorbital (= palpebral, Coombs, 1972) (Fig. 4.1 1; see also Chapter 3, section 3.2.2.1 ). This ossification, invariably located within the orbital cavity and frequently coarticulating with the roof of the orbit, is currently known from three specimens, AMNH 5238, AMNH 5404 and TMP 97.132.1 (Fig. 4.11 ; see Appendix 4.1). When viewed externally, the convex surface is slightly rugose, with a parabolic dorsal margin and a weakly bowed, laterally deflected ventral margin. The media1 surface is smooth and deeply concave. Given its position, Coombs (1972) hypothesized that the "ciliary" supraorbital of Euoplocephalus functioned as a bony shutter, occluding the orbit when adducted.

4.2.2.2 Os postorbitale (lamina postocularis) (Figs. 4.1 1,4.14A) A well-developed postocular shelf transversely occludes the caudal border of the orbital cavity (Chapter 7, section 7.2, character 41) (Fig. 4.1 1). The concave rostral face adjoins the basicraniurn in the same transverse plane as the basal tubera, and extends laterally to a position just medial to the rostral border of the quadratojugal boss. Contributions to the postocular shelf from any of the jugal, frontal or laterosphenoid cannot be discretely ascertained (Fig. 4.14A).

4.2.2.3 Os jugale (arcusjugale) (Fig. 4.5A) Providing a caudal continuance to both the maxillary tomial crest and the buccal emargination, the jugal grades insensibly with the surrounding cranial elements. In lateral profile, the suborbital jugal arch is rnorphologically unremarkable and not distinctly ornamented.

4.2.2.4 Os frontale (not explicitly figured) Although the frontal cannot be discretely identified in any given specirnen, a cornparison of frontal position in subadult Pinacosaurus and Minmi, with structures present in various taphonomically damaged Euoplocephalus crania suggests that the frontal contributes to a number of composite structures, including the rostral wall of the orbit, the internarial cavity septum, and the nasal cavity proper.

4.2.2.5 Os parietale (margo nuchae) (Fig. 4.1 2) Along the caudalmost border of the cranium, the nuchal shelf is embossed with raised, transverse ornamentation (C hapter 7, section 7.2, character 1 1). Fusion between the nuchal shelf and the supraoccipital is complete, and transverse CT scans demonstrate the complete absence of a metakinetic joint (contra Coombs, 1971). In dorsal view, the nuchal shelf completely conceals the basicranium and occipital elements from dorsal view (Chapter 7, section 7.2, character 12) (Fig. 4.2A).

4.2.2.6 Os squamosum (processus cornuum squamosi, cotylus quadrati sguamosaljs) (Figs. 4.3A, 4.3C, 4.5AI 4.7A, 4.1 2, 4.1 3A14.13C) Positioned at the caudodorsolateral extremity of the cranium, the pyramidal squarnosal boss (Chapter 7, section 7.2, character 6) demonstrates a considerable degree of non-taphonomic morphological variability, especially at the apex, ranging from sornewhat obtuse (e.g. AMNH 5403) to sharply tapering (Fig. 4.12). Furthermore, the surface texture of the boss may either be smooth (Fig. 4.13A) or rugose (e.g. AMNH 5405), although it is invariably embellished by numerous small foramina. An unusual condition is noted in AMNH 5337, where virtually the entire apex of the squamosal boss has become excavated by a large caudolateratly oriented pathological aperture of unknown etiology. ln both lateral profile and dorsal view, the boss is characteristically segregated from the surrounding ornamentation by a network of prominent furrows (Figs. 4.3A14.3C, 4.5A, 4.1 3A). Diametrically opposite the apex of the boss, the ventromedial margin of the squamosal forrns the shallow concavity of the quadrate articulation surface (4.13C). Although typically occluded from view by the presence of the quadrate, examination of an isolated right squamosal attributed to Euoplocephalus (TMP 93.36.79; Fig. 4.1 3 A, C) perrnits investigation of the quadrate articulation surface. While there is no evidence of fusion between the squamosal and the quadrate condyle, streptostylic cranial kinesis appears to be unlikely, given the integrated nature of the quadrate with the elements of the temporal region (see section 4.2.4.9). 4.2.2.7 Os quadratojugale (processus cornuum quadratojugalis) (Figs. 4.3A, 4.5AI 4.9, 4.11, 4.12, 4.13B, 4.13D) The quadratojugal protuberance is perhaps the most salient feature of the cranium, situated as a pronounced caudolaterally directed triangular flange, obscuring the adjacent quadrate articular condyle from lateral view (Fig. 4.5A). In ventral view, the ventromedial aspect of the quadratojugal embraces the quadrate proximal to the articular condyle (Fig. 4.1 1). Opposite to the quadratojugal - quadrate contact, the caudomedial surface of the quadratojugal flange is frequently excavated by large (10 - 15 mm diameter) depressions. Although these excavations are present, to variable degrees, in al1 crania examined, their significance is not currently understood.

4.2.3 Palatal Region (Figs. 4.1 BI 4.6, 4.8,4.9)

Historically, the palatal region has played an important rote in early descriptive efforts, owing to the lack of cranial ornamentation and the frequent retention of sutura1 contacts. Recently prepared material, coupled with new data generated through CT scans and fortuitously broken specimens, permits a more exhaustive commentary.

4.2.3.1 Ossa vomera (septum cavitas internasalis) (Figs. 4.1 B, 4.6, 4.7, 4.8, 4.98) Along the sagittal plane, the juxtaposed vorner elements (identified by Lambe [1902] as the conjoined presphenoid and basisphenoid [Fig. 4.1 BI, and by Nopcsa [1928] as "longitudinal septa") becorne laminated and extend dorsally to form the internasal cavity septum. Contra Sternberg (1929), the vomer is paired. Transverse CT scans (Fig. 4.9), physically sectioned specimens (AMNH 5403), and crania having undergone taphonomic truncation (Fig. 4.6) demonstrate the complete longitudinal bisection of the rostral and palatal regions by this septum (Chapter 7, section 7.2, character 20). Rostrally, the vomers narrowly incise the interpremaxillary suture, and provide a small contribution to both the margin of the intemal naris and the rostrodorsal secondary palate (Fig. 4.8). At the caudal terminus the vomer conjoins with the body of the pterygoid. Despite available evidence, contact between the basicraniurn and the vorner is equivocal. Along the palatal margin, the vornerine keel lies variably within or just ventral to the coronal plane of the tooth rows. In cross-section, the vomerine keel is incised (resernbling an inverted "V-shaped trough).

4.2.3.2 Os palathum (caudoventral secondary palate) (Figs. 4.1 B, 4.8) Situated at the caudal rnargin of the interna1 naris, the palatine (= vomer, Lambe 1902; Fig. 4.1 B) adjoins with the vomer medially and the maxilla laterally to form the caudoventral secondary palate (Chapter 7, section 7.2, character 21 ) (Fig. 4.8). Transverse CT scans, a physically sectioned specimen and fortuitously broken crania confirm that the caudoventral secondary palate underlies the rostrodorsal secondary palate. A small "posterior palatal foramen" (sensu Maryanska, 1971; 1977) is not present along the caudal border of the palatine. However, a large aperture is present near the midpoint of the palatine (Fig. 4.8), leading to a caudodorsally extending recess. The homology of this aperture and cavity remains unclear.

4.2.3.3 Os pterygoideum (Figs. 4.8, 4.1 1) Correlated with the developrnent of a caudodorsal secondary palate, the central pterygoid body is vertically oriented (Chapter 7, section 7.2, character 27). Rostrally, the pterygoid body adjoins with the vomer, contributing to the caudalmost extremity of the vomerine keel, and the aforementioned secondary palate (Fig. 4.8). Frorn the caudal terminus of the vomerine keel, the rostral face af the central pterygoid body is directed in an oblique caudolateral orientation, and thence into the transverse plane (Figs. 4.8, 4.1 1). After a short distance, the transverse plane of the central pterygoid body gives rise to the rostroventrolaterally oriented mandibular ramus (Figs. 4.8, 4.1 1; Chapter 7, section 7.2, character 29) (= part of the ectopterygoid, Sternberg, 1929, labeled as the "transversumJ'Nopcsa, 1928, Plate VI Figure 4). Distally, the mandibular ramus is slightly expansive and rugose. The quadrate process of the pterygoid bifurcates from the central pterygoid caudal to the rnandibular ramus. Oriented caudolaterally, the quadrate process is a deep, narrow brace typically fusing with the pterygoid process of the quadrate (Chapter 7, section 7.2, character 28) (Figs. 4.8, 4.1 1). Aside from the scarf joint between the quadrate and the pterygoid, none of the sutures between the pterygoid and the adjacent elements can be discerned.

4.2.3.4 Os ecfopterygoideum (not figured) The ectopterygoid is a small, robust caudomedially oriented wedge, situated dorsal to the mandibular ramus. None of the contacts between the ectopterygoid and the maxilla, pterygoid or palatine are evident.

4.2.4 Occipital 1 Basicranial Region (Figs. 4-78, 4.7C, 4.1 1, 4.12, 4.14)

Despite nurnerous specimens, identification of individual elements composing the basicranium remains problematic, owing to a virtual elimination of intervening sutura1 contacts. The elernents discussed herein were identified based on structural features (viz. cranial nerve foramina) that are presumed to rernain consistent within the clade Ankylosauria.

4.2.4.1 Os supraoccipifale (Fig. 4.12) Identifiable only by inference, the supraoccipital is fused completely with adjacent elements of the occiput. Roofing the forarnen magnum, the supraoccipital forms a prorninent pair of caudoventromedially oriented transverse nuchal crests (= oval tuberosities, Nopcsa, 1928), segregated from one another by a small, incisive notch. Within the sagittal plane, inset deep to the caudal margin of the nuchal shelf and dorsal to the foramen magnum, is a narrow sagittal nuchal crest. Flanking the sagittal nuchal crest is a pair of shallow fossae with subtending rugosities. 4.2.4.2 Os exoccipitale (processus paroccipitalis) (Figs. 4.1 1, 4.1 2, 4.14) With the exception of the prominent paroccipital process, the exoccipital cannot be discretely identified. Projecting laterally from either side of the sagittally positioned foramen magnum (Chapter 7, section 7.2, character 33), the paroccipital processes deepens dorsoventrally towards the distal terminus (Fig. 4.1 2). The paroccipital process does not fuse with the squamosal head of the quadrate (Chapter 7, section 7.2, character 39), although the two elements come into close apposition with one-another. Ventral to the paroccipital process is a subovoid posttemporal fenestra (Fig. 4.12). Contact between the paroccipital process and the opisthotic is not preserved (Fig. 4.14A). The contribution, if any, of the exoccipital to the rnargin of the foramen for cranial nerve XI rernains unclear (Fig. 4.148). Dorsally, the exoccipital is completely obscured from view by the nuchal shelf (Chapter 7, section 7.2, character 12) (see Fig. 4.2A).

4.2.4.3 Os basioccipitale (condylus occipitalis) (Fig. 4.1 1, 4.12, 4.14) The caudoventrally protrusive occipital condyle (Chapter 7, section 7.2, character 36) is the most diagnostic feature of the basioccipital. In occipital view, the occipital condyle is reniform (Chapter 7, section 7.2, character 35), with a shallow median condylar notch incising its dorsum (Fig. 4.1 2). The articular surface is smooth and featureless, and does not appear to receive any contributions from the exoccipitals (Chapter 7, section 7.2, character 34). In lateral profile, the occipital condyle is not offset from the rest of the basicranium by a neck (Fig. 4.14B). Although al1 the sutura1 contacts have becorne obliterated, the basioccipital is presumed to contribute to the ventral margin of the foramina for cranial nerves IX, X and XI (Fig. 4.14B) and to contact the basisphenoid near the reduced basal tubera.

4.2.4.4 Ossa otica (os prooticum, os opisthoticum) (Fig. 4.1 4A) Although not discretely demonstrable morphologically, the otic elements circumscribe the fenestra ovalis and consequently, their general topographie position may be inferred. 4.2.4.5 Os basisphenoidale (tuberculum basilare; processus basipterygoideus) ( Fig. 4.1 1, 4.1 4) Positioned as the rostral continuation of the basioccipital, the basisphenoid gives rise to a pair of short, stout, rostrally (and slightly laterally) oriented basipterygoid processes (Fig. 4.1 4B). Although the basipterygoid processes corne into close proximity with the caudal face of the central pterygoid body, they do not directly contact this surface (Chapter 7, section 7.2, character 30). Caudal to the basipterygoid process, along the ventral surface of the basisphenoid near the contact with the basioccipital, emerge a pair of rugose, crest-ljke basal tubera (Chapter 7, section 7.2, character 32) (Fig. 4.1 1). Contributions, if any, to the basal tubera by the basioccipital is currently unclear. The basisphenoid is invariably shorter than the basioccipital (Chapter 7, section 7.2, character 31). Specific contributions to the ventral margin of the foramina for cranial nerves III, VI and VI1 are assumed, although not explicitly evident (Fig. 4.1 4B). Along the rostroventral margin of the basisphenoid is a large parasagittal opening for transmitting the palatine artery.

4.2.4.6 Os laterosphenoidale (Fig. 4.14) Medially adjacent to the postocular shelf, the laterosphenoid occupies a small obscure position along the dorsolateral wall of the basicraniurn. It remains unclear as to whether the laterosphenoid contributes to the postocular shelf. Topographically the element is bounded by the margins of the foramina for cranial nerves 11, It Il IV and V (Fig. 4.1 48).

4.2.4.7 Os parasphenoidale (Fig. 4.14A) Generally observed coursing rostrodorsally deep within the interpterygoid vacuity, the rod-like parasphenoid rostrum is the only satient feature of the parasphenoid.

4.2.4.8 Ossa interorbitalis (Fig. 4.14A) The interorbital elements are not readily demonstrable in any specimen.

4.2.4.9 Os quadrafum (Figs. 4.7B14.7C. 4.1 1, 4.12) The rectilinear quadrate (Chapter 7, section 7.2, character 28) is situated as a tripartite connector, linking elements of the occipital / basicranial (viz. the exoccipital), temporal (squamosal and quadratojugal), palatal (pterygoid) and mandibular (articular) regions. Among these points of contact, only the quadrate- articular jaw joint is considered to have had extensive mobility (see below). The squamosal condyle does not fuse with either the ventral surface of the squamosal (the quadrate articulation surface; Fig. 4.13C) or the paroccipital process (Chapter 7, section 7.2, character 39). However, the fused union of the quadrate to both the quadratojugal (laterally) and the pterygoid (rostromedially) (Chapter 7, section 7.2, character 28) appears to prohibit any possibility of streptostyly (contra Coombs, 1971) (Fig. 4.1 1). Although frequently misaligned as a result of taphonomic deformation, the long axis orientation of the quadrate (as properly reconstructed) is near vertical (contra Haas, 1969; Coombs, i971 ; see Rybczynski and Vickaryous, in press). In lateral view, the pronounced quadratojugal flange completely obscures the articular condyle of the quadrate (Chapter 7, section 7.2, character 40) (Fig. 4.12; see also Figs. 4.2C, 4.5A14.1 1).

4.3 Ossa Mandibulae

Unless otherwise stated, al1 mandibular elements are paired.

4.3.1 Mandibular Region (Fig. 4.15, 4.1 6)

Despite the relative abundance of crania, the lower jaw of Euoplocephalus is infrequently collected, and typically, when recovered, is represented by only one-half the pair (Le. a right or a left rnandible). 4.3.1 .l Os predentale (Figs. 4.15A, B) Concomitant with the broad premaxillary rostrum, the predentary is transversely wide. Externally, the surface of the predentary is rugose and variably foveated, with a small predentary tuber on the dorsal surface in the sagittal plane (Fig. 4.1 5B). The rostral margin is crenellated, although no distinct predentary tomium develops, and the distal rami are directed caudolaterally. In cross-section, the predentary is dorsoventrally depressed. Contact between the predentary and the dentary has been interpreted as amphiarthitic (Rybczynski and Vickaryous, in press; see Chapter 3, section 3.3.1. A).

4.3.1.2 Os dentale (Figs. 4.1 5C,4.1 SD, 4.16B) Although similar to the maxilla in arranging the dentition into a single occlusal row, the alveolar border of the dentary is longer than its cranial counterpart. Conversely however, the dentary has an average of 21 teeth. The course of the alveolar border is both medially deflected and dorsoventrally convex, giving the dentigerous aspect of the mandible the appearance of a medially folded arch. Rostrally the mandibular ramus is greatly reduced, both dorsoventrally and rostrocaudally, and is reoriented perpendicular to the long axis of the body of the dentary. In addition, the dorsal aspect of the rostral margin is deeply incised by the predentary sulcus. The articulation surface between the contralateral mandibular rami (i.e. the mandibular symphysis) is variably rugose. In Iateral view the alveolar border appears to arch dorsal to the main body of the dentary (Fig. 4.15~).Near this parabolic elevation, numerous large foramina pierce the dentary surface. Medially, a subparallel row of "special foramina" (Edmund, 1957) flanks the alveolar border. Approaching the rami of the mandibular symphysis, a large "Meckelian groove" incises the ventromedial border of the dentary. Caudally, this groove can be traced back to an aperture between the dentary and the splenial.

4.3.1.3 Os spleniale (Figs. 4.1 5C,4.16B) Overlapping the ventral margin of the dentary in medial view, the splenial circurnscribes, in part, the aperture of the "Meckelian grooven and completely delimits the large caudal interrnandibular foramen.

4.3.1.4 Os angulare (Fig. 4.1 5D) Invariably, the angular is embellished by rugose cranial ornementation resembling a single elongate osteoderm with a conspicuous ventrolaterally oriented keel. The angular is visible along the ventralmost margin of the splenial in medial view, although it does not specifically contribute to this surface, and it does not make contact with the prearticular.

4.3.1.5 Os supra-angulare (Fig. 4.1 5D) Close to its caudal border, the surangular is pierced by a large rostromedially oriented foramen of uncertain homology. Along the medial surface, this foramen opens at the caudalmost boundary of the adductor fossa (Fig. 4.16).

4.3.1.6 Os coronoicfeum (Fig. 4.16) With the possible exception of the predentary, the coronoid is the least represented element of the mandible. Presently identifiable in only two specimens (AMNH 5337 and UALVP 31 ; Fig. 4.16), the coronoid gives rise to a small, thin coronoid process (Fig. 4.16A), the medial surface of which is embossed with caudodorsally oriented striae.

4.3.1.7 Os prearticuiare (Fig. 4.1 5B, 4.15C, 4.16B) The retroarticular process is mediolaterally compressed, with a sharp keel- like ventral margin and rugosities along the caudoventral border.

4.3.1.8 Os articulare (not figured) The articular is transversely broad, extending medially from a position just caudal to the adductor fossa (Fig. 4.1 BA). The dorsal surface of the articular (the quadrate-articular fossa) bears an obtuse, mediolaterally directed fold that partitions the area into a rostral and a caudal component (Rybczynski and Vickaryous, in press).

4.4 Dentes (Fig. 4.10)

Euoplocephalus lacks a premaxillary dentition (Chapter 7, section 7.2, character 17; see also Appendix 3.2). lndividual teeth of both the maxilla and the dentary are generally small (cusp width up to 7.5 mm, maximum cusp height 7.5 mm; compare with cranial length of 41 1 mm) and lack a distinct cingulum (Chapter 7, section 7.2, character 19) (Figs. 4.1OC, D). Although frequently truncated (through both natural Wear and taphonomic processes; see Rybczynski and Vickaryous, in press) the apical carina may bear as many as 17 denticles. Crown morphology is variably fluted, with numerous vertically oriented, shallow furrows adorning both the labial and lingual surfaces (Figs. 4.10C1 D). Fluting distribution does not correspond with position of the denticles. The taxon Palaeoscincus asper Lambe (1902), known only from a single tooth, was synonymized with Euoplocephalus, on the basis of general morphology (Coombs, IWO).

4.5 Summary

Euoplocephalus is a well represented Late Cretaceous ankylosaur from the Western lnterior of North America. Despite frequent anecdotal references, the most detailed descriptive efforts of this taxon date back to the 1920's. An exhaustive review of the majority of skuli specimens referabte to Euoplocephalus has identified a number of morphological features heretofore unreported, underreported and I or erroneously reported. The cranium is akinetic, with neither metakinetic nor streptostylic joints. Although present, the laterotemporal fenestra is not visible in lateral view. The pattern of cranial sculpturing across the rostrum is unique, with bilateral supranarial (vestibular) arches and flanking subrectangular ornaments, and one or hounpaired median polygons. The long axis of the external naris is vertically oriented. A paranasal aperture is located caudolatera! to each external naris, within a shallow nasal vestibule. The broad, edentulous, unornamented premaxillary rostrum has a narrow premaxillary notch. The premaxillary tomial crest contacts the maxillary tomial crest lateral to the tooth row. The maxillary tooth rows are medially inflected, giving rise to an "hour glass" configuration when viewed ventrally. The buccal emargination is deeply concave. The rostrodorsal secondary palate is mainly derived from the maxillae. The presence of modified supraorbitals resembling ossified "eyelids" is described from a third specimen. The supraorbital boss tapers dorso!aterally to form a wedge-like protuberance. The postocular shelf is prominent and occludes the entire caudal margin of the orbital cavity. The pyramidal squamosal boss is variable in morphology, ranging from blunt ta sharp. The nuchal shelf has cranial sculpturing. The quadratojugal forms a conspicuous flange that completely obscures the articular condyle of the quadrate from lateral view. The ventral margin of the vomerine keel projects beyond the plane of the tooth rows. The vomer contributes to the internasal cavity septum, cornpletely bisecting the rostral region of the cranium. The palatine has a large aperture of undetermined homology. The pterygoid contributes to the caudoventral secondary palate (along with the palatine), with a vertical central body, a rostroventrolaterally oriented mandibular ramus and a caudolaterally oriented quadrate process. The pronounced paroccipital process is directed laterally and is obscured from dorsal view by the nuchal shelf. The occipital condyle is caudoventrally oriented, kidney-shaped in occipital view and derived from the basioccipital alone. The basipterygoid process does not contact the main body of the pterygoid and the basal tubera is a rugose crest. The quadrate does not fuse with either the quadrate articulation surface of the squamosal or the paroccipital process. Maximum mandibular length is four-fifths the length of the cranium. The predentary is broad and edentulous, with an abbreviated predentary tuber on the dorsal surface. The alveolar border of the dentary is longer than that of the maxilla, although it contains fewer teeth, and resernbles a medially folded arch. The angular is completely embossed with sculpturing. The coronoid is small and gives rise to a reduced coronoid process. The articular has an obtuse, transversely oriented crest subdividing the glenoid into rostral and caudal cornponents. All teeth are relatively small with a variable degree of fluting and no cingulum.

This chapter summarizes and reappraises the cranial morphology of Euoplocephalus, providing the fundamental osteological framework for further palaeobiological investigations, including the phylogenetic placement of this taxon within the clade Ankylosauria. A complete discussion detailing the taxonomie position of Euoplocephalus and a specific summary of diagnostic characters is presented in Chapter 7, section 7.2. Figure 4.1. Cranium of Euoplocephalus tutus Lambe (1902) in dorsal (A.) and ventral (B.) views, with the rostrum directed towards the top of the page. NMC 021 O*, holotype. Scale bar, 100 mm. Abbreviations in parentheses reflect the original element interpretation by Lambe (1902). Al1 abbreviations listed in Appendix 2.1 . s mx-

VO - (para I bs) Figure 4.2. Cranium of Euoplocephalus tutus Lambe (1902) in dorsal (A.) ventral (B.), right lateral (C.)and occipital (D.) views. TMP 91 .IZ.1. Scale bar, 100 mm.

Figure 4.3. Cranial sculpturing of EuopIocephalus tutus Lambe (1902). A. Oblique rostrodorsalateral view of AMNH 5405. Scale bar, 100 mm. B. Lateral view of right supraorbital boss of TMP 91.127.1. Scale bar, 50 mm. C. Dorsal view of the left temporal region of TMP 91.127.1. Scale bar, 50 mm. All abbreviations listed in Appendix 2.1. pro corn sorb cranial sculpturing \ pro corn sq

/ pro corn ql

pro corn sq margo nuc \

pro corn sorb

b. / pro corn sorb Figure 4.4. A. Rostrodorsal profile of Euoplocephalus tutus Lambe (1902) demonstrating of the characteristic pattern of cranial scul pturing. AMN H 5405. Scale bar, 50 mm. B. Dorsal view of cranial osteoderrn referred to Euoplocephalus. TMP 89.36.183. Scale bar, 10 mm. Note the similarity in morphology between the embossed cranial sculpturing and the cranial osteoderm.

Figure 4.5. A. Schematic of right lateral profile of the cranium of Euoplocephalus tutus Lambe (1902), adapted from AMNH 5405 (Coombs, 1978). Scale bar, 100 mm. B. Oblique rostrodorsal view of the rostrum of Euoplocephalus tufusLambe (1 902), AMNH 5405. Scale bar, 50 mm. All abbreviations listed in Appendix 2.1. pro corn sorb

pro corn l / 91 pro man pt cr tm prnx A.

apt nas \ PpOsm pila internas \ apt paranas

interpremaxillary suture

/ \ in pmx cr trn pmx

B. Figure 4.6. Ventral view of the truncated palatal region of Euoplocephalus tutus Lambe (1 902). TMP 97.59.1,original and interpretive illustration. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1 sp n osm sp I cav internas

pro alv- B. Figure 4.7. Cranium (broken) of Euoplocephalus tutus Lambe (1902) in left lateral (A.) and right medial (B., C.) views, original (A., B.) and interpretive illustration (C.). TMP 96.75.1.Scale bar, 100 mm. All abbreviations listed in Appendix 2.1 pro corn sq pro corn sorb \ I

V sp cav internas Figure 4.8. Ventral view of the palate of Euoplocephalus tutus Lambe (1902), original and interpretive illustration. TMP 98.83.1.Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. in pmx rostrodorsal secondary palate

emargination

/ secondary palate ,t ,t pro man qu Figure 4.9. Transverse sections (B. - D.) through the rostrai region of Euoplocephalus tutus Lambe (1902), based on computer tomography of AMNH 5405 (A.), interpretive illustrations. Stippled areas indicate regions that could not be properly evaluated owing to the presence of large amounts of radio-opaque minerals. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. cav nas

C ?s mx

L-- sp osm nas

~vesnas =cr tm pmx Figure 4.1O. Dentition of Euoplocephalus tutus Lambe (1 902),in situ and isolated. A. Medial view of left rnaxilla with in situ dentition, AMNH 5405, original photograph. B. Media1 view of left maxilla with in situ dentition, AMNH 5405, interpretive illustration. Scale bar, t 0 mm. Scanning electron micrographs of isolated teeth, TMP 92.36.1226(C.) and 87.36.99(D.). Scale ber, 10 mm.

Figure 4.11. Ventral view of the occipital 1 basicranial region of Euoplocephalus tutus Larnbe (1902), original and interpretive illustration. TMP 98.83.1. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. pro man pt pro man qu cor pt vacinpt /

lam postoc Figure 4.12. Occipital view of occipital / basicranial region of Euoplocephalus tutus Lambe (19O2), original and interpretive illustration. TMP 91.127.1. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. damaged margo nuch 1

cor

I pro'man pt ros pmx fen posttemp Figure 4.13. lsolated cranial elernents referred to Euoplocephalus. Left squamosal, in occipital (A.) and mediai (C.)views. TMP 93.36.79.Left quadratojugal, in lateral (B.) and mediai (D.) views. Scale bar, 20 mm. All abbreviations listed in Appendix 2.1. suture contact for sq cor-

pro corn qj

cot qu sq \

suture for contact with qj Figure 4.14. Right lateral profile of the basicranium of Euoplocephalus tutus Lambe (1 902), original and interpretive illustration. FPDM-V-35. Scale bar, 50 mm. AH abbreviations listed in Appendix 2.1. lam postoc

opis rb / ros para Pro exo \ exo - 8 S

for n VI1 for n III, IV Figure 4.1 5. Mandibular region of Euoplocephalus tutus Lambe (1902), AMNH 5405. Predentary in dorsal (A.) and oblique rostrodorsal (B.) views. Mandible in medial (C.),and lateral (D.) views (mirror imaged). Scale bar, 100 mm. AH abbreviations listed in Appendix 2.1.

Figure 4.16. Mandibular region of Euoplocephalus tutus Lambe (1902), UALVP 31. Mandible in rriedial view (B.) (scale bar, 100 mm), with a magnification of the coronoid element (A.). All abbreviations listed in Appendix 2.1 . fos add Table 4.1. Skull material referred to the taxon Euoplocephalus tutus presently accessioned in museurn collections. Institutional abbreviations listed in chapter 1. Occurrence refers to the stratigraphic position material was collected from. Occurrence abbreviations: DP Fm., Dinosaur Park Formation (late Campanian, Alberta); HSC Fm., Horseshoe Canyon Formation (early Maastrictian, Alberta); JR Gp.,Judith River Group (late Campanian, Alberta); TM Fm., Two Medicine Formation (early Maastrictian, Montana); ?: unknown Iocality. Use of the more inclusive Judith River Group stratigraphic package was applied to specimens whose exact providence within Dinosaur Provincial Park, Alberta remains unclear. See Literature Cited for a complete reference citation. Asterisk indicates type specimen. Specimens not personally examined shown in redline.

NMC 021 0 * cranium JR Gp. Table 4.1. Continued.

. .. .

TMP 97.59.1 cranium HSC Fm. none cranium + right rnandible + "ciliary" TMP 97.1 32.1 DP Fm. none supraorbitals TMP 98.83.1 cranium DP Fm. none UALVP 31 cranium + lefi mandible JR Gp. Gilmore, 1923

NMC 8530 * cranium + right mandible HSC Fm. Sternberg, 1929 Table 4.2. Additional fossil material examined. Unless otherwise noted, al1 specimens are referrable to the taxon Euopiocephalus tutus. Comment abbreviations: disart, disarticuIated; taph worn, taphonomicaliy wom (e.g. water abraided); SEM, specimen examined using scanning electron microscopy. "Broken" designation suggests that the specimen was physically damaged, with clean fracture danes.

Palaeoscincus asper Lambe 1902 holotype Stereocephalus tutus Lambe 7 902 holotype skull roof fragment broken right squamosal disart + broken right quadratojugal thin-sectioned right quadratojugal fragment broken + taph worn left quadratojugal disart + broken right mandible fragment broken + taph wom

SEM right supraorbital disart

cranial osteoderrn disart

right quadratojugal fragment broken + taph worn skull roof fragement broken SEM

right squamosal disart + broken right quadratojugal fragment broken + taph wom broken rostral region fragment thin-sectioned Chapter 5 Morphological Description of the Cranium of "Gobisaurus" gen. et sp. nov.

Amongst the fossil material collected by the Sino-Soviet expeditions (1959-1 960) to the Alshan Desert of China was a virtually complete, heretofore unknown ankylosaur. Characterized by an enlarged orbit and elongated rostral process of the vomer, this taxon, herein referred to as "Gobisaurus" gen. et sp. nov., is described on the basis of a single, well-ossified cranium (sensu stricto), IVPP V12563* (holotype; Fig. 5.1). Forrnal description of this taxon, including nomenclatural designation, is forthcoming (Vickaryous et al., submitted). Until such time, the designate "Gobisaurus" remains unofficial. In addition to details gleaned from traditional observational methods, the holotype was subjected to non-invasive cornputer tomographic (CT) scanning, permitting the examination of several deeply situated structures (see below). Although postcranial material attributed to this specimen is reported to exist (Zhao, pers. comm., 2000),it is presently unavailable for study. A cast of the holotype cranium (TMP 90.300.74) is available for examination at the TMP. For the purposes of this treatment, except where noted, the cranial morphology of "Gobisaurus" closely resembles that of other ankylosaurs (Chapter 3), and consequently, the details provided hereafter will concentrate on novel features and previously discussed character states (see also Chapter 7).

5.1 Overall Morphology

The cranium of "Gobisaurus" is virtually complete (Fig. 5. l),measuring 457 mm from a line connecting the rostral-most tips of the premaxillary tomia to the caudal edge of the occipital condyle (see Appendix 4.1). In dorsal view (Fig. 5.1A), the cranium is triangular and slightly longer than wide (445 mm between the flaring apices of the quadratojugal projections; see section 5.4.8; also see Chapter 7, section 7.2, character 1; Appendix 4.1). The entire dorsum is covered with a rugose, pockmarked texture that lacks any organized pattern of ornamentation (i.e. cranial sculpturing sensu Coornbs, 1971; see Chapter 3, section 3.5) (Chapter 7, section 7.2, characters 8 and 9). Except for a short segment of the sagittal interpremaxillary suture (Fig. 5.28), none of the cranial sutures is visible in either dorsal (Fig. 5.1A) or lateral view (Figs. 5.1C15.2). Neither the antorbital nor the dorsotemporal fenestrae are externally visible. Furthermore, transverse computer tomographic (CT) scans taken at 3 mm increments parallel to the sagittal plane (Figs. 5.3) failed to identify any discrepancy in cranial thickness along the entire length of the cranium (including the bony antorbital wall and the skull table). The laterotemporal fenestra is present, however, but remains obscured by the squamosal-quadratojugal complex in lateral view (Chapter 7, section 7.2, character 4). The remaining openings, the rostrolaterally directed external naris and paranasal aperture (Fig. 5.2) and orbit (Figs. 5.2A15.4A) are prominent. The roughly elliptical nasal vestibule accounts for over 23% of cranial length (long axis is 106.6 mm), and is subdivided medially by an osseous nasal septum (see section 5.3.1 ; Fig. 5.2). The orbit of Gobisaurus is ovoid, with the long axis oriented such that the narrowest end is directed rostroventrally (Figs. 5.2A15.4A). Its long axis (91.5 mm) encompasses 20% of the cranial length. A narrow rim of bone encircles the orbit, gaining prominence along the rostral and ventral borders. The antorbital area is swollen (Figs. 5.1A15.1 B, 5.26, 5.3, 5.4A, 5.5), extending from the caudal edge of the external naris to a position just rostrodorsal to the orbit. A morphological reciprocity exists between the elliptical nasal vestibule, the ovoid orbit, and the bulbous antorbital area, with the antorbital area situated as a trapezoid-like wedge between the prominent openings. The maximum distance between the supraorbital bosses (Fig. 5.1A) (350 mm; also see section 5.4.1) is nearly equivalent to the maximum distance across the squamosal bosses (352 mm; also see section 5.4.4; also see Chapter 7, section 7.2, character 10; Appendix 4.1). In lateral profile, both the rostral and the temporal regions are relatively flat (see Chapter 7, section 7.2, characters 2 and 3) (Figs. 5.1C, 5.2A, 5.3).

5.2 Ossa Cranii

Unless othennrise stated, al1 cranial elements are paired.

5.2.1 Rostral region (Fig. 5.2, 5.3, 5.5)

Except where previously noted, sutural contacts between individual elements of the rostral region are indistinguishable in rostrodorsolateral views. Thus, the topographical position and morphological expression of many rostral elements (viz. the nasals, prefrontals, and lacrimals) and features (e.g. the nasolacrimal aperture) cannot be ascertained discretely. Fortunately, several of these elements have maintained distinct sutural junctions that are visible along the palate (e.g. the premaxillae and maxillae), such that they merit separate commentary.

5.2.1.1 Os premaxillare (Figs. 5.2, 5.3C, 5.3D15.5) The premaxilla is edentulous (Chapter 7, section 7.2, character 17), encloses the rostral half of the nasal vestibule (Fig. 5.2))and bears externally a number of shallow foveae and vertically oriented furrows. As previously noted, only a short segment of the sagittal inter-premaxillary suture is visible, with al1 other contacts along the dorsal and lateral surfaces completely obscured. In ventral view, the narrow premaxillary palate is deeply incised by a deltaic premaxillary notch (Fig. 5.5) (Chapter 7, section 7.2, character 15), from which a short segment of the interprernaxillary suture is derived. The premaxillary tomial crest continues caudolateraily from the notch along the ventrolateral border of the premaxillary palate, adjoining the maxillary tomial crest at the premaxillary- maxillary contact (Chapter 7, section 7.2, character 16) (Fig. 5.5). A caudodorsally inclined parasagittal incisive forarnen (6 mm in diameter) is situated directly caudal to the premaxillary notch. Transverse CT scans enables tracing of the incisive foramen caudad to a position just rostral to the external naris, where it then becomes indistinct (not figured). Contact between the premaxillae along the palate is restricted to a narrow bridge media1 to the incisive forarnina (Fig. 5.5). Caudal to this bridge, the premaxillae are separated by elongate rostral extensions of the vorners (processus vomera; see section 5.5.1 ). The premaxillary-rnaxillary contact along the palate is preserved as a shallow furrow, undulating rostromedially then caudomedially around the dentigerous embayrnent of the maxilla (i.e. the alveolar border of the maxilla). Near the contact with the vomers, the premaxillary-maxillary articulation is not preserved. Rostral to the external naris, along the dorsornedial surface of the cranium, the internasal bar slopes obliquely at a 45' angle, giving the rostrum a protracted appearance (Figs. 5.1 Cl 5.2). Within each expansive nasal vestibule, a caudodorsally oblique bony nasal septum partitions each external naris into rostral and caudal apertures (Chapter 7, section 7.2, character 24) (Fig. 5.2). CT scans indicate that the nasal cavity proper opens externally at the rostral narial aperture (the nasal aperture; apertora nasalis), whereas the caudal aperture (the paranasal aperture; apertura paranasalis) leads to a paranasal maxillary sinus cavity (see Chapter 3, section 3.7) (Chapter 7, section 7.2, character 26) (Fig. 5.3).

5.2.1.2 Os maxillare (rostrodorsal secondary palate) (Figs. 5.38, 53C,5.4, 5.5) The topographical borders of the maxilla are indistinguishable from the rest of the cranial elements, except at the aforementioned premaxillary contact along the palate (see above). The tomial crest continues caudolaterally from the premaxillary-maxillary synarthrosis to a position near the rnidpoint of the alveolar border. lnset between, and running parallel with, the maxillary tomial crest and the alveolar border is a shallow buccal emargination (Fig. 5.5). In cross section this depression begins rostrally as concave furrow, partially obscuring the maxillary teeth from lateral view (see Fig. 5.2A). Immediately, however, the caudal continuance of the emargination becomes increasingly superficial and flattened (see Fig. 5.4, 5.5; also see Chapter 7, section 7.2, character 22). Medially, the lingual portion of the maxilla is poorly preserved, and the presence of "special foramina" (Edmund,. 1957) is equivocal. Furthermore, the sutural contacts between the maxilla and the adjacent elements of the palatal region (Le. the vomer, ectopterygoid and palatine) are not evident. Hotvever, the lingual maxillary shelf may be traced dorsomedially, contributing, along with the vomer, to the rostrodorsal secondary palate (Chapter 7, section 7.2, character 21 ) (Fig. 5.5). Unfortunately, the exact configuration of these elements remains unc!ear. Bilaterally, each half of the rostrodorsal secondary palate is inclined caudornedially, and forms the rostral margin of the interna1 naris. The tooth rows and alveolar borders are maximally separated at their caudalmost positions (127 mm) and converge towards the rostnim (95 mm). Close to the contact with the premaxillae, the tooth rows are parallel (see Chapter 7, section 7.2, character 18) (Fig. 5.5). Tooth row length is approximately 26.7% (122 mm) of cranial length. Most of the alveolar border is damaged, and an accurate tooth count based upon alveoli is not possible. However, nine in situ teeth are present in the left maxilla, with four more in the right (Fig. 5.4). An alveolar tuberosity cannot be discretely identified.

5.2.2 Temporal Region (Fig. 52A, 5.4, 5.6,5.7)

Similar to the rostral region, none of the sutural contacts between adjacent elements of the temporal region are delineated. Therefore, discussion of individual elements is limited to elaborations thereof and / or structures they are presumed to contribute to.

5.2.2.1 Ossa supraorbitalia (processus cornuum supraorbltalia) (Fig. 5.2A, 5.4A) The supraorbital adornment is a blunt, ill-defined convex protrusion, with its apex oriented dorsolaterally (Chapter 7, section 7.2, character 5). When viewed laterally, the protrusion conjoins with a narrow, variabty developed rim of bone encircling the orbital margin (Fig. 5.4A).

5.2.2.2 Os postorbitale (lamina postocularis) (Fig. 5.6) A well-developed postocular shelf transversely occludes the caudal border of the orbital cavity (Chapter 7, section 7.2, character 41). The rostral face is weakly concave before becoming contiguous with the jugal arch. Medially, the postocular shelf contacts the basicraniurn near the caudal terminus of the basal tubera. Contributions to the postocular shelf from any of the jugal, frontal or laterosphenoid cannot be discretely identified.

5.2.2.3 Os jugale (arcusjugale) (Fig. 5.2A) Concomitant with the enorrnous size of the orbit, the jugal arch is a slight, longitudinal brace. Ventrally, a short continuation of the shallow buccal emargination encroaches upon the rostraimost rnargin. A perceptible, blunt continuance of the maxillary tomial crest projects along the ventrolateral margin of the jugal arch, before coalescing with the quadratojugal protuberance.

5.2.2.4 Os parietale (margo nuchae) (Fig. 5.7) The nuchal shelf lies in the coronal plane, roofing the occiput and concealing al1 but the caudodorsal extremity of the paroccipital process from dorsal view (Chapter 7, section 7.2, character 12). Contact between the nuchal shelf and the underlying supraoccipital is not preserved well enough for description. Raised sculpturing is not present along the caudal margin of the nuchal shelf (Chapter 7, section 7.2, character 11 ) (see Fig. 5.1 A).

5.2.2.5 Os squamosum (processes cornuum squamosi + cotylus guadrati squamosalis) (Figs. 52A, 5.4A, 5.7) Sirnilar to the supraorbital protrusion, the squarnosal process is a blunt convexity (Chapter 7, section 7.2, character 6). Located at the caudodorsolateral angle of the cranium, the squamosal process has low relief and is largely indistinguishable from the rest of the temporal region. In occipital view, the apex of the squamosal process is oriented slightly more laterally than the aforementioned supraorbital protrusion (see section 5.3.1). Ventrally, the dorsolateral coalescence of the paroccipital process of the exoccipital and the squamosal condyle of the quadrate is deeply situated within the quadrate articulation surface of the squamosal (also see sections 5.5.2 and 5.5.9). A ventrally oriented ridge emanating from the caudal margin of the squamosal process partially obscures the paroccipital - squamosal condyle contact.

5.2.2.6 Os quadrafojugale (processes cornuum quadrafojugalis)(Figs. 5.2A, 5.4A15.6, 5.7) The prominent quadratojugal protuberance is situated as a lobular flange lateral to the articular condyte of the quadrate (Chapter 7, section 7.2, character 7) (Figs. 5.6, 5.7). ln occipital view, the ventrolateral orientation of the flange is redirected laterally towards the apex, permitting the exposure of the articular condyle of the quadrates in lateral profile. A medial process of the quadratojugal contacts the quadrate proximal to the articular condyle.

5.2.3 Palatal Region (Fig. 5.5)

The palatal region is largely complete, although variably preserved. Obliteration of synarthroses in this region prevents a comprehensive morphological discussion of each constituent element.

5.2.3.1 Ossa vomera (+ septum cavitas intemasalis) (Figs. 5.3C,5.3D, 5.5) The paired vomers are completely fused along the vomerine keel, which while variably damaged, extends ventrally beyond the plane of the prernaxillary palate. Rostrally, elongate processes of the vomers intrude deeply into the premaxillary portion of the rostrum, wherein they unite with the roof of the premaxillary palate (Fig. 5.5). Unlike the vomerine keel, the ventral rnargin of the rostral process does not project beyond the transverse plane of the premaxillary palate. CT scans (Figs. 5.3C1D) indicate that the rostral processes of the vomers are triangular in cross-section, and form a wedge between the juxtaposed premaxillae. In addition, the vomer contributes to the internasal cavity septum, completely segregating the paired respiratory paths (Chapter 7, section 7.2, character 20). Sutural contacts between the individual vomer elements, and between the vomer and the pterygoid, palatine and maxilla, are not visible.

5.2.3.2 Os palatinum (caudoventral secondary palate) (Fig. 5.5) Extensive fracturing of the palatines has rendered complete preparation inappropriate, and, consequently, contact between the palatine and adjacent rnaxilla, pterygoid and ectopterygoid is not discernible. Furthermore, the ventral (palatal) surface of the palatine is devoid of ornate structures (e.g. crests and depressions). However, the palatine does contribute to the caudoventral secondary palate (Chapter 7, section 7.2, character 21) and, along with the vomer, defines the caudal limit of the internal naris. Although the concave caudal margin of the internal naris is present, the narrow, longitudinal crest directed rostrally along the medial surface of the maxilla is variably damaged.

5.2.3.3 Os pterygoideum (Figs. 5.5, 5.6, 5.7) The central body of the pterygoid is vertically oriented (Chapter 7, section 7.2, character 27),and abuts the vomer and the caudodorsal secondary palate rostrally (Fig. 5.5),although the sutura1 contacts between these elements are obliterated. A rostrolaterally-directed mandibular ramus fuses with the ectopterygoids near the caudal terminus of the maxillary tooth row. Distally, the mandibular ramus is expansive, slightly rugose and laterally oriented (Chapter 7, section 7.2, character 29). The laminar quadrate process of the pterygoid is directed caudolaterally, towards the quadrate (Chapter 7, section 7.2, character 28). The suture between the pterygoid and the quadrate is not preserved. The pterygoid forms the rostromedial border of a large, elliptical suborbital fenestra. 5.2.3.4 Os ecfopferygoideum (not figured) The ectopterygoid is short, deep, and rostrolaterally oriented, although none of the sutures demarcating the contacts with the maxilla, palatine and pterygoid are preserved.

5.2.4 Occipital / Basicranial Region (Figs. 5.6, 5.7)

Poor preservation and sutura1 obliteration impedes the identification of many individual elements (viz. the prootic, opisthotic, laterosphenoid, parasphenoid and interorbital ossifications) and most structural features of the basicranium (e-g. foramen ovale, foramina for cranial nerves and vascular channels, the parasphenoid rostrum). Except where noted, none of the occipital 1 basicranial elements are discretely preserved. Where possible, morphological structures have been assigned to elements with which they are most commonly associated. The ventral surface of the basicraniurn is square in outline, and perceptibly inclined rostrodorsally.

5.2.4. i Os supraoccipitale (Fig. 5.7) A narrow sagittal nuchal crest is present between the nuchat shelf of the parietal and the foramen magnum. Bifurcating ventrally from the sagittal crest are caudolaterally directed transverse nuchal crests. The transverse nuchal crests, and associated foramen magnum, are directed cauclally and slightly ventrally (Chapter 7, section 7.2, character 37). The area of the occiput flanking the sagittal nuchal crest is rugose and foveated.

5.2.4.2 Os exoccipifale (processus paroccipitalis) (Figs. 5.6, 5.7) The caudolaterally directed paroccipital process (maximal width for the pair 282 mm) (Chapter 7, section 7.2, character 33) expands distally, and fuses with the quadrate (Chapter 7, section 7.2, character 39). A rugose muscle scar caps the lateralmost edge of the paroccipital process, and continues ventrally along the lateral border of the quadrate. The subovoid posttemporal fenestra is visible ventral to the paroccipital process (Fig. 5.7). The opisthotic contribution to the paroccipital process is not preserved.

5.2.4.3 Os basioccipitaie (condylus occipitalis) (Figs. 5.6, 5.7) The massive, caudally directed occipital condyle (80 mm wide by 44 mm high) (Chapter 7, section 7.2, character 36) is reniform in occipital view (Chapter 7, section 7.2, character 35) (Fig. 5.7). Dorsally, the occipital condyle bears a shallow median condylar notch. In lateral profile, the occipital condyle is offset from the rest of the basicranium by a short neck, and the ventral margin is sellar. Along the occiput, the condylar processes of the exoccipitals adjoin the basioccipital at a pair of shallow, dorsornedially oblique furrows (Chapter 7, section 7.2, character 34).

5.2.4.4 Os basisphenoidale (processus basipterygoideus, tuberculum basilare) ( Fig. 5.6) Each basipterygoid process is thick, robust and rostrolaterally oriented. The basipterygoid process does not, however, contact the caudal margin of the central pterygoid body, and the narrow gap rostral to each process is filled by rnatrix (Chapter 7, section 7.2, character 30). Each basal tubera is well defined as a rugose, longitudinally oriented crest (Chapter 7, section 7.2, character 32). The relative length of the basisphenoid compared with the basioccipital is presently unclear (Chapter 7, section 7.2, character 31 ).

5.2.4.5 Os quadratum (Figs. 5.2, 5.4A, 5.6, 5.7) The prominent quadrate dominates the occiput as a transversely broad buttress, fusing the quadratojugal laterally, paroccipital process dorsally (Chapter 7, section 7.2, character 39) (Fig. 5.7) and pterygoid (medially). In lateral profile, the quadrate is linear, without any rostrocaudal undulations (Chapter 7, section 7.2, character 38), and the quadrate condyle is readily visible ventral to the quadratojugal flange (Chapter 7, section 7.2, character 40) (Figs. 5.2, 5.4A). Contact between the quadrate and pterygoid occurs along an oblique rostromedial union (Chapter 7, section 7.2, character 28). The nature of the contact between the quadrate and the squamosal remains uncleat.

5.3 Dentes (Fig. 5.48)

The premaxillary rostrum of "Gobisaurus" is edentulous (Chapter 7, section 7.2, character 17; see also Appendix 3.2). Each maxillary tooth is characterized by its relatively large size (cusp width up to 9.5 mm, maximum cusp height 11 mm; compare with cranial length of 457 mm), swollen base and development of an incipient cingulum along the lingual surface (Chapter 7, section 7.2, character 19). The apical series of denticles is asymmetrically disposed across the medial - distal length of the tooth, but an exact count is not possible. Surface embellishments (e.g. fluting) are not present on any of the teeth preserved.

5.4 Surnmary

A new genus and species of large ankylosaur from the Early Cretaceous of China is documented, based on a single cranium (a formal taxonornic description is provided elsewhere: Vickaryous et al., submitted). The taxon "Gobisaurus" gen. et sp. nov. may be readily distinguished from most other ankylosaurs by the narrow premaxillary rostrurn, enlarged orbit and narial vestibule, and rugose cranial ornamentation not organised into cranial sculpturing. The laterotemporal fenestra is visible in occipital view. The antorbital area is laterally distended. The paired prernaxillae are edentulous, cleaved by a deltaic premaxillary notch and separated along their length by elongated rostral processes of the vomers. The premaxillary tomial crest adjoins the maxillary tomial crest lateral to the tooth row. The paranasal aperture is located caudal to the external naris. The tooth rows are parallel to one another rostrad and diverge from one another caudad. The buccal emargination is relative shallow and fiat. The rostrodorsal secondary palate receives contributions from the maxilla and the vomer. The supraorbital and squamosal bosses are blunt and ill-defined. The pronounced postocular shelf encloses the caudal rnargin of the orbital cavity. The jugal arch is narrow. The nuchal shelf lacks cranial ornamentation. Despite the presence of a large, conspicuous lobate Range, the quadratojugal does not obscure the quadrate condyle from lateral view. The ventral margin of the vomerine keel projects beyond the plane of the tooth rows. Dorsally, the vomer contributes to the internasal cavity septum, completely subdividing the rostral region in the sagittal plane. The caudoventral secondary palate receives contributions from the palatines and pterygoidç. The main body of the pterygoid is vertically oriented, with a rostrolaterâl'iy directed mandibular ramus and a caudolaterally directed quadrate process. The paroccipital process is caudolaterally directed, fused with the quadrate and not obscured by the nuchal shelf in dorsal view. The occipital condyle is large, reniform in occipital view, caudally directed and receives contributions from the exoccipitals. The basipterygoid process is robust, rostrolaterally oriented and does not fuse with the main body of the pterygoid. The basal tubera is a parasagittal rugosity. The teeth are relatively large, with a reduced cingulum and no surface fluting.

This chapter summarises the cranial morphology of "Gobisaurus", providing the necessary osteological underpinning for al1 subsequent palaeobiological investigations, including the phylogenetic placement of this taxon within the clade Ankylosauria. A complete discussion detailing the taxonomic validity of "Gobisaurus" and a specific summary of diagnostic characteristics is presented in Chapter 7, section 7.2. Figure 5.1. Cranium of "Gobisaurus"gen. et sp. nov. in dorsal (A.), ventral (B.), right lateral (C.)and occipital (D.) views. lVPP V l2563*, holotype. Scale bar, 100 mm.

Figure 5.2. A. Schematic of the lateral profile of the cranium of "Gobisa~rus~~gen. et sp. nov. IVPP V12563*, holotype. Scale bar, 100 mm. B. Left rostrolateral oblique view of "Gobisaurus"gen. et sp. nov. lVPP VI2563*, holotype. Scale bar, 50 mm. All abbreviations listed in Appendix 2.1. antorbital region pro corn sorb apt nas \ \ I

- fov cor ner arc

tm pmx pro corn qj 1 \ pro man pt \ pro man qu sp n osm

antorbital area apt nas /

in pmx sp n osm \ pro man pt B. I apt paranas cr trn pmx Figure 5.3. Transverse sections (B. - D.) through the rostral region of "Gobisaurus" gen. et sp. nov. based on CT scans of IVPP Vl2563*, holotype (A.), interpretive illustrations. Stippled areas indicate regions that could not be properly evaluated owing to the presence of large arnounts of radio-opaque minerals. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. ?s rnx

.-0- . .

B cav nas

Ipro vo

provo Figure 5.4. A. Left rostrolateral oblique view of "Gobisaurus" gen. et sp. nov. IVPP V12563*, holotype. Scale bar, 100 mm. B. Right lateral view of right maxilla of "Gobisaurus" gen. et sp. nov. IVPP VI 2563*,holotype. Scale bar, 10 mm. All abbreviations listed in Appendix 2.1 . pro corn sorb pro corn sq antorbital area \ \ 1

orb

pro corn qj 1 I cd qu 1 pro man pt

/ pro alv Figure 5.5. Ventral view of the palate of "Gobisaurus"gen. et sp. nov. IVPP V12563*, holotype, original and interpretive illustration. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. in pmx

reconstructed secondary palate

cor pt palate Figure 5.6. Ventral view of the occipital I basicranial region of "Gobisaurus"gen. et sp. nov. IVPP VI2563*, holotype, original and interpretive illustration. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. pro bpt promanpt Vacint~t/

tub b Figure 5.7. Occipital view of occiput of "Gobisaurus" gen. et sp. nov. IVPP V'i 2563*,holotype, interpretive illustration. Scale bar, 100 mm. All abbreviations listed in Appendix 2.1. cr nuc sag in med .cd / pro corn

for mag

fen laterotemp

'fen posttemp cd I qu pro man pt Chapter 6 Cranial Ornamentation

The most conspicuous synapomorphy of the ankylosaur head skeleton is the pervasive developrnent of osseous cranial ornamentation (Coombs, 1971; Sereno, 1986; 1999; Coombs and Maryanska, 1990; Carpenter, 1997a; b), defined here as elaborate osteological constructs and 1 or modifications of pre- existing skull element(s) not appearing to have a primary role in food acquisition or locomotion (sensu Vickaryous and Rybczynski, in press). Elements of the dorsal and lateral surfaces of the cranium (viz. the rostral and temporal regions) and the lateral surface of the mandible frequently, perhaps inevitably, become embossed with a continuous, amorphous, foveated texture (see Figs. 5.1A, 5.1 Cl 5.2, 5.4). Furthermore, these surfaces may also exhibit a series of shallow, interconnecting furrows that subdivide the continuous ornamentation into a mosaic of polygons and cornuate bosses, resulting in the manifestation of cranial sculpturing (Coombs, 1971; see Figs. 4.1A, 4.2A, 4.2C, 4.3, 4.4A, 4.5A). Variants of cranial sculpturing have often been cited as a means by which rnembers of the Ankylosauria may be differentially clustered (e.g. the Nodosauridae vs. the Ankylosauridae; see Coombs, 1971; Coornbs and Maryanska, 1990; Sereno, 1986; Carpenter, 1990). However, despite its ubiquitous, defining presence, ankylosaur cranial ornamentation (and cranial sculpturing) has received scant developmentat, and subsequent biotogical, consideration. This chapter evaluates cranial ornamentation in the context of modern developmental mechanisrns, and outlines a research application whereby biological processes may be effectively interpreted within the limited confines of the fossil record.

6.1 The Comparative Method

Previous efforts at biologically evaluating ankylosaurian cranial ornamentation (including the cranial sculpturing variant) have focused on two competing developmental hypotheses: (1) overlying bony plates (osteoderms) develop within the integument before contacting, and subsequently coossifying with, the underlying elements of the skull proper (the coossification hypothesis); or (2) individual skull elements undergo some degree of osteological elaboration (the elaboration hypothesis). Neither of these hypotheses have, however, been rigorously examined, as developmental mechanisms are not subject to discrete preservation. The issue is reinvestigated here with the application of the comparative method (sensu Bock, 1989). The comparative method, by definition, is an a priori evaluation of inference based upon biological comparisons, permitting a tenable extrapolation of theorized (unpreserved) attributes using formalized, falsifiable procedures (Bock, 1989; Bryant and Russell, 1992; Witmer, 1995). Inference of function, behaviour, physiology, development and / or soft tissue morphology using the comparative method generally follows one of two principal (although not necessarily exclusive) techniques (Bock, 1989; Witmer, 1995); phylogenetic inference and (or) ahistorical extrapolatory modeling. Phylogenetic inference (the historical method of Bock, 1989) is based upon the integration of established phylogenetic information with morphological corollaries, in order to evaluate the structure (and the potential for its development) in the context of an organism's genealogy. Recently, phylogenetic inference has been embodied as crown-group phylogenetic bracketing (Bryant and Russell, 1992) and the extant phylogenetic bracket (Witmer, 1995; 1997; see also Weishampel, 1995). Alternatively, ahistorical extrapolatory modeling permits the illation of unpreserved attributes based strictly upon the preserved morphology of a fossil taxon (Bryant and Russell, 1992; Weishampel, 1995; Witmer, ? 995; 1997; the non-historical method of Bock, 1989; the paradigm method of Lauder, 1995). The methodotogy is reliant upon the establishment of "biological generalizations", conservative, invariable relationships between structures, tissues, or patterns of development and organisms. Consequently, the identification of a particular osteological correlate perrnits the extrapolation of a particular unpreserved feature (Bryant and Russell, 1992; Weishampel, 1995; Lauder, 1995). In contrast to the phylogenètic approach, the extrapolatory approach enables the inference of attributes that may not be present in phylogenetically proximate taxa (Bryant and Russell, 1992).

6.1 .l Phylogenetic lnference (Fig. 6.1 )

The following information is presented as a brief summarization of the methodology outlined by Bryant and Russell (1992) and Witmer (1995; 1997). The comparative approach of phylogenetic inference for fossil vertebrate taxa relies upon two important precepts; (1) the determination of phylogenetic polarity and (2) the identification of osteological correlates. Phylogenetic inferences pertaining to a fossil taxon are based upon the attributes of the two most closely related extant outgroups (the sister group and the next most closely related outgroup). These taxa form the "extant bracket" (Fig. 6.1). Polarity determination requires that character state change be mapped out upon a phylogenetic hypothesis (i.e. a cladogram). As the character state of the fossil taxon is unknown, it may be treated as both present and absent (equivocal) (see Fig. 6.1 D). Consequently, a fossil taxon with a particular osteological feature may, or may not, have had the unpreserved attribute in question. The extant bracket is then examined to determine if the same osteological feature, and accompanying unpreserved attribute (the so-called "osteological correlate"), is present arnongst the two most closely related outgroups. If both the extant bracketing taxa possess the suspected attribute, then the fossil taxon assessment iç considered decisively positive (Fig. MC); Le. the fossil taxon probably did possess the unpreserved attribute. If on1y one extant bracketing taxon possessed the suspected attribute, then the fossil taxon assessment is considered equivocal (Fig. 6.1 D). If neither extant bracketing taxon possess the suspected attribute, then the fossil taxon assessment is considered decisively negative (Fig. 6.1 8);i.e. the fossil taxon probably did not posses the fossil attribute. For the Dinosauria (ankylosaurs included), the extant phylogenetic bracket is composed of crocodylians and avians (Fig. 6.1A). A review of modern skull material from a variety of avian lineages indicates that osseous cranial ornamentation is rare, and when present does not resemble that of ankylosaurs (e.g. the culmen crests of some Coraciiformes and the pneumatic cranial casques of Casuariiformes). Conversely, modern crocodylians invariably express some degree of osseous cranial ornamentation, generally characterized by networks of shallow pits and grooves that superficially resemble the cranial ornamentation of some ankylosaurs. However, close inspection reveals that the morphology and distribution of the crocodylian pitting differs considerably from that of ankylosaurs. Furthermore, modern crocodylians Jack any semblance of cranial sculpturing (including cornuate bosses) and rarely obliterate any suturat junctions. Consequently, while the developmental mechanism giving rise to the pitted morphology of crocodylians (reportedly a localized pattern of differential resorption of the periosteum; de Buffrenil, 1982) is not precluded from occurring within the Ankylosauria, it is not herein considered to play a major role. In this instance, both extant bracketing taxa fail to demonstrate the osteological correlates required for a decisively positive assessment; i.e. neither modern avians nor modern crocodylians possess osteological correlates of the cranial ornamentation of ankylosaurs. Consequently, application of phylogenetic inference as a means of evaluating the developmental mechanisni of ankylosaur cranial ornamentation is of limited utility.

6.1.2 Ahistorical Extrapolatory Modeling

In contrast to phylogenetic inference, ahistorical extrapolatory modeling reconstructs unpreserved attributes based upon recognized form-function correlations or biomechanical interrelationships (Hopson and Radinsky, 1980; Bryant and Russell, 1992), in the absence of proximate genealogy. Examination of remote outgroups, such as squamates and mammals, identifies patterns of osseous cranial ornamentation sirnilar to that noted in ankylosaurs. This common pattern of development is herein considered to represent a "biological generalization" (Bryant and Russell, 1992), and, as such, this study relies heavily upon the ahistorical extrapolatory approach (Bryant and Russell, 1992; also see Weishampel, 1995; Witmer, 1995).

6.2 Methods and Materials

The development of osseous cranial ornamentation is nearly ubiquitous within the Amniota; thus, a comprehensive systematic examination is beyond the scope of this analysis. For reasons of practicality (relative size and abundance, ease of transport), the comparative (homoplasious) portion of this study focuses on an examination of the development of cranial ornamentation in extant squamates. The Squamata (Fig. 6.2) is a monophyletic clade encompassing the majority of taxa commonly considered as structural grade "reptiles". Squamates may be subdivided into two major lineages: the Iguania and the Scleroglossa (Fig. 6.2). Members of both lineages exhibit a wide array of cranial ornamentation, including osseous and soft-tissue structures. A survey of alcohol- preserved and skeletal specimens representing most of the major squamate clades (Appendix 6.1 ), supplemented with data from the Iiterature (see reviews by Gadow, 1901; Camp, 1923; and Estes et al., 1988), suggests that particular patterns of osteology are strongly correlated with phylogeny. The presence of both cranial and postcranial osteoderms (ectopic dermal ossifications supporting [an] overlying epidermal scale[s]; Romer, 1956; Hildebrand, 1988; Vickaryous et al, in press) in squamates appears to be restricted to members of the Scleroglossa (in particular scincoids and anguimorphs [exclusive of the Ophidia], as well as sorne members of the Gekkota and Lacertoidea). In contrast, members of the lguania (with the exception of Amblyrhynchus cristafus) rarely, if ever, develop osteoderrns (Gadow, 1901; Camp, 1923). However, numerous iguanian taxa exhibit the development of osseous horn cores, bosses (e.g. phrynosomatids and some acrodonts) and exostoses (polychrotids). Extant iguanians served as the basis for modeling the elaboration hypothesis. Of primary import for this study were several embryo and adult comparative series of Phrynosoma (including P. cornutum, P. hernandezi, and P. modestum), embryonic Chamaeleo pumilus, neonate C. calyptratus, and adult C. jacksonii, C. montium, and C. parsoni (Appendix 6.1 ). Extant scleroglossans were similarly employed as models for testing the coossification hypothesis. The principal specimens examined for this study were subadult and adult Heloderma suspectum, adult Gerrhosaurus major, and adul t Tarentola maurifanica (see Appendix 6.4 ). The developmental models were generated through observations made from a combination of cleared and double-stained, akohol-preserved, fresh- frozen, and dried skeletal specimens. Alcohol-preserved and fresh-frozen specimens were subjected to radiographic imaging. A superficial consideration of osseous cranial ornamentation in mammals suggests that the developmental pathways noted in squamates are representative of a more inclusive clade. Amongst extant mammals, only dasypodids (armadillos) commonly develop osteoderms (Martin et al., 2001). Morphologically the condition appears to parallel that of scleroglossans. The elaboration of the frontals in the Bovidae is likewise similar to the developrnental process noted in iguanians. A review of original ankylosaur material was based largely upon two taxa; Euoplocephalus tutus (Fig. 4.1) and Pinacosaurus grangen (Fig. 3.7A13.8A, 3.9, 3.1OA). Both taxa were medium-sized ankylosaurid ankylosaurs from the Late Cretaceous of North America and Asia respectively. Selection of these taxa was based on the relative abundance of available cranial material. While the majority of Euoplocephalus specimens appear to represent adult-sized individuals, a number of undescribed elements referable to subadult individuals have recently been identified (see Chapter 4, sections 4.3.1, 4.3.5, 4.3.6;see Fig. 4.13). Nearly al1 the material referable to Pinacosaurus represents subadult individuals. Wherever possible, additional material assigned to other ankylosaur taxa was examined. 6.3 Hypothesis 1: Coossification

Overwhelrningly, the majority of dinosaur researchers have advocated (usually incidentally) the hypothesis of osteoderm coossification with the skull as the developmental mechanisrn responsible for ankylosaur cranial ornamentation (e.g. Lambe, 1902; 1919; Brown, 1908; Sternberg, 1929; Russell, 1940; Sereno, 1997). Often this interpretation has been invoked circumstantially (and inappropriately) during discussion of morphology, but it has yet to be critically tested or substantiated.

6.3.1 Developrnent of Cranial Ornamentation in Extant Scleroglossans

A review of the literature suggests that the development of osteoderms (both cranial and postcranial) in scleroglossans is a characteristic that undergoes repetitive reversal within the clade (Gadow, 1901; Camp, 1923; Arnold, 1973; Zylberberg and Castanet, 1985; Estes et al., 1988). Morphologically there is a high degree of variability arnongst members dernonstrating the condition, ranging from the imbricated "shingle" morphotype of Gerrhosaurus (Scincoidea) and Anguis (Anguimorpha), to the polygonal "pavement" morphotype of Heloderma (Anguimorpha; see Fig. 6.3) and Tarenfola (Gekkota). Developmental mechanisms investigated thus far have been IargeIy restricted to the polygonal "pavement" morphotype, although al1 osteoderms are presumed to be homologous (Camp, 1923; Moss, 1969). Moss (1969) investigated the ontogenetic development of osteoderms in the anguimorphan Heloderma horridium. Presumptive osteoderms develop as domed regions of thick collagen within the dermis, first appearing over the ossified head skeleton, then spreading caudally (Fig. 6.3). Topographic distribution of these collagen "domes" over the cranium bears no direct relationship to the underlying, and well-established, cranial elements. As the individual matures, the collagen "domes" increase in size before finally becoming ossified. With subsequent development the osteoderms may fuse directly to the cranium (e.g. Heloderma), or remain suspended within the overlying dermis (e.g. sorne gekkotans such as Tarentola). An ontogenetic study of the anguimorphan Anguis fragilis by Zylberberg and Castanet (1985) indicates that the "shingle" morphotype undergoes a similar process.

6.3.2 Osteological Correlates

Due to the nature of osteoderm development, a number of biological generalizations (sensu Bryant and Russell, 1992) can be made with regard to the resulting osteology. Osteoderms develop within the dermis, superficial to the head and postcranial skeletons, and, as such, are not confined to the topographic limitations of individual elements. Thus, osteoderms may originate in positions that overlap several elements and / or sutural boundaries. Additionally, osteoderms may form in regions where underlying skeletal elements are absent (e.g. between adjacent ribs, superficial to the temporal fenestrae and orbits; Fig. 6.3). Consequently, the position and morphology of the cranial openings and the sutural arrangement of cranial elements may become obscured in mature individuals. Prior to fusion with the cranium, however, it is possible to remove the osteoderms by removing the integument. While the gross morphology of postcranial osteoderms may be highly variable, those that develop over the skull generally form polygons (Fig. 6.3). These polygons appear to develop as the incipient osteoderms (centres of ossification within individual collagen "domes") expand radially and begin to encroach upon one-another. Continued growth is constrained by adjacent (incipient) osteoderms, thereby giving rise to the polygonal morphology. Abutment of adjacent polygonal osteoderms may cover the cranial openings with ectopic (extra-cranial) derrnal bone (e.g. the supraorbital ossifications of the gekkotan Tarentola; Bauer and Russell, 1989). The polygonal configuration of the cranial osteoderms closely resembles the morphological condition noted in the carapace of placodonts (Westphal, 1976) and sorne mammals (e-g. dasypodids). A brief analysis of dasypodid material (Appendix 6.1 ) suggests that the presence of osteoderms in mammals is derived from developmental processes similar to those of scleroglossans. Dried adult Dasypus novemcinctus skeletal material and alcohol-preserved adult Chaetophractus sp. demonstrates the presence of both the polygonal "pavement" morphotype, and the imbricated "shingle" morphotype osteoderms, covering the entire dorsal surface of the head and body. However, no osteoderms were identified through the radiographic imaging of alcohol preserved fetal Chaetophractus vellerosus. Owing to the presence of the osteological correlates identified in scleroglossans, dasypodids are presumed to undergo a similar developmental process.

6.3.3 The Hypothesis

Despite their phylogenetically distant relationship to ankylosaurs, many scleroglossans appear to present a pattern of cranial ornamentation that is morphologically congruent with them. On the basis of this gross similarity it may be hypothesized that ankylosaur cranial ornamentation developed in an analogous manner - i.e. osteoderms overlying the craniurn fuse with the skull, resulting in a polygonal pattern of ornamentation that obscures sutural contacts and cranial openings. Prior to maturation (and subsequent fusion), the cranium of ankylosaurs should not exhibit cranial ornamentation in any areas associated with this mode of formation.

6.3.4 Testing the Hypothesis

In order to test the coossification hypothesis, osteological correlates within the Ankylosauria must be identified. Requirements for satisfying the hypothesis of osteoderrn coossification include: (A ) complete absence of ornamentation in immature specimens; (2) the presence of ornamentation that obscures (overlaps) the sutural arrangement in mature specimens; and (3) the presence of osseous ornamentation in regions of the skull without underlying cranial elements. The ubiquitous presence of cranial ornamentation has been purported to obscure the sutural arrangement of the head skeleton of ankylosaurs (Coombs and Maryanska, 1990). Evidence to support this assertion is derived from the examination of material referred to Pinacosaurus grangeri. Several specimens of Pinacosaurus (IVPP uncatalogued; Maryanska, 1971; 1977) illustrate the morphology of individual skull elements (see Chapter 3, Figs. 3.7A, 3.8A, 3.9, 3.10A). Based, in part, on the lack of cranial ornamentation, and the relatively small and unfused nature of the skeletal elements, these specimens are considered to represent subadult or immature individuals. Examination of the holotype (AMNH 65233,a larger specimen with ornamentation, suggests that the state of development of cranial sculpturing is ontogeneticaily regulated. Assuming that skull morphology is relatively conservative (although see Chapter 3 for details), comparison of subadult Pinacosaurus crania with those of other related ankylosaurids suggests there is no relationship between the individual cranial elements and any overlying polygonal ornamentation. Subadult cranial material is not yet known for al1 taxa (see Chapter 3), thus thwarting a more corn prehensive systematic review. The majority of ankylosaur taxa are known, at least in part, from cranial material considered to represent adult or mature individuals. In most cases, the specimens demonstrate extensive development of cranial ornamentation (most often cranial sculpturing). With the exception of some specific topographical regions (e.g. the premaxillary rostrum), the sutural arrangement across the skull roof in these specimens is unknown, and presumed to be obscured by the presence of cranial ornamentation (see Chapter 3, sections 3.2.1 and 3.2.2). In addition to the widespread development of postcranial osteoderrns, a number of ankylosaur taxa have been found preserved with in situ osseous "ciliary" supraorbitals (Euoplocephalus tutus, AMNH 5238,5404; Coombs, 1972; Fig. 4.1 1; see also Chapter 4, section 4.3.1) and buccal ossifications (osseous cheek plates; Edmontonia rugosidens, AMN H 5381 ; Panoplosaurus mirus, NMC 2759). Al1 of these dermal elements develop in the apparent absence of underlying skeletal tissue. The hypothesis of osteodermal coossification giving rise to cranial ornamentation was originally dismissed by Coombs (1971) in his comprehensive review of the Ankylosauria. Coombs cited two major flaws with the argument, one based on gross morphology, the other on bone histology. Prior to Coombs' research, no subadult ankylosaur skull had been described and thus al1 the cranial material he examined demonstrated extensive development of cranial ornamentation. He reasoned that if the cranial ornamentation of ankylosaurs was the result of osteoderms coossifying with the skull surface, then a specimen should exist where the cranium proper was visible. Despite referring to an unseen subadult specimen of Pinacosaurus (via communication with Teresa Maryanska), he submitted that "No such specimen exists.. ." (Coombs, 1971:157). Coombs found further evidence to refute the coossification hypothesis by examining the bone histology of an ankylosaur skull. He suggested that if overlying osteoderms did contact and fuse with the cranium, then an appreciable thickening of the skull roof should be noted (when compared to other non- pachycephalosaurian ornithischians). A single cranium of Euoplocephalus tutus (AMNH 5403) was transversely sectioned across the antorbital region. Examination of the cross-sections indicated that the thickness of the Euoplocephalus skull roof was "modest" (Coombs, 1971 : 156), and generally comparable to that of most other ornithischians. He also noted that there was no indication of a juncture between the cranium and the ornamentation. This led Coombs to surmise that Euoplocephalus, as a representative of the Ankylosauria, did not have osteoderrns fused to the skull roof proper. A review of the present evidence, supplemented by new cranial thin- sections, accounts for the misgivings of Coombs (1971 ). Detailed examination of subadult Pinacosaurus specimens confirms the absence of cranial ornamentation in immature ankylosaurs. Additionally, an undescribed bony plate (TMP 89.36.1 83; Fig. 4.48) may represent unfused cranial ornament. The element has a rugose and pitted texture, similar to that of postcranial osteoderms (to which it was originally referred). However, the specimen is thin (40mm thick) and flat, with a slightly concave upper surface, and is hexagonal in dorsal view. A comparison of this specimen with other known ankylosaur postcranial osteoderms suggests that the morphology is unique. It most closely resembles the polygons associated with the rostral cranial sculpturing of Euoplocephalus, and is herein referred to as an unfused cranial osteoderm. The original thin-sections of Coombs (from AMNH 5403) have been augmented with new material (TMP 67.20.20 and 98.1 15.2) in order to review ankylosaur cranial bone histology. Two main histological layers may be differentiated (Fig. 6.4); a thin (~2.5mm) layer of isolated, unorganized primary osteons (Fig. 6.4, labelled "a") superficial to deeper, extensively remodefled Haversian bone (Fig. 6.4, labelled "b)(Coombs, 1971 ). Thorough examination of al1 thin-sections failed to identify any sutural junctions, either between overlying osteoderms and the cranium, or between individual cranial elernents. The highly reorganized nature of the bone histology suggests that the cranium undenivent continuai remodelling throughout ontogeny, and that any sutural contacts have long since been obliterated. The functional implications of remodelled bone in the ankylosaur cranium are not fully understood, although previous work has suggested that sutural fusion adds increased rigidity and strength to the head skeleton (Herring, 1974). Furthermore, a superficial layer of bone lacking a strict organizational pattern (Le. woven bone), coupled with the presence ornarnentation, has been postulated to be a structural rnechanism for stress diffusion (Coldiron, 1974). Experimental work has suggested that rnicroscopic fractures spread as stress is focused at the end of the crack (Coldiron, 1974). If this point stress may be diffused and distributed across a Iarger surface area - e.g. at lacunae within the layer of woven bone or along a foveated surface - movement of the crack may be arrested. Thus, in a teleological sense, the peripheral layer of an ankylosaur skull may have arisen in response to dispersing and dissipating any stresses incurred (e.g. through predation or intraspecific combat). 6.3.5 Conclusions

All the fundamental requirements for corroborating the hypothesis of osteoderm coossification are fulfilled by ankylosaur cranial osteology. The expression of cranial sculpturing in ankylosaurs is governed by ontogeny, and the entire clade demonstrates a propensity for the production of osseous (ectopic) dermal tissue. The supetficial furrows that subdivide the cranium (giving rise to the cranial sculpturing) are herein considered to represent the areas of coossification between adjacent cranial osteoderms. The unusual distribution of cranial sculpturing (i.e. concentrated around the rostrum) may reflect the degree of interaction bekveen the epidermis and the dermis. However, the rote of the epidermis in mediating osteodermal growth is not presently understood. In the anuran Hyla septentrionalis, dense connective tissue of the dermis is effectively replaced by bone over areas of the cranium (Trueb, 1966). This integurnentary derivative then coossifies with the cranium, creating a "casque". The lack of epidermal scutes may partially explain the absence of polygonal organisation in this secondary dermal bone. Whether the production and subsequent coossification of dermal bone to the skull is responsible for al1 the cranial ornamentation noted in ankylosaurs has yet to be addressed. The subadult crania of Pinacosaurus grangeri, white devoid of cranial sculpturing, do demonstrate small horn-like bosses over the orbits and at the caudal corners of the skull. In addition, Jacobs et al. (1994) noted the presence of "excrescences" on some disarticulated elements of a small (presumably) subadult nodosaurid cranium. The development of these structures cannot be accounted for by osteoderm coossification.

6.4 Hypothesis 2: Elaboration

Coombs (1971) rejected the notion of osteoderrn coossification as the predominant developmental mechanism of cranial omamentation, due to a number of perceived morphological and histological inconsistencies. Alternatively, he proposed that the osseous ornamentation of the cranium was the result of individual cranial elements becoming elaborated under the influence of epidermal structures. Coombs noted that elaborate modifications of the cranium were not without precedent amongst the Ornithischia (e.g. the premaxillary-nasal crests of lambeosaurines, the squamosal-parietal frills of neoceratopsians). Therefore, while the possibility of some extra-cranial contributions to the ankylosaur skull were not entirely precluded, they were thought to play a minor role. A more recent review of the Ankylosauria by Coombs and Maryanska (1990) put forward a less polarised view by stating that the cranial omamentation might be the result of either elaboration of the cranium or the coossification of osteoderms.

6.4.1 Development of Cranial Ornamentaiion in Exiant lguanians

Analysis of an ontogenetic sequence of cleared and double-stained Phrynosorna specirnens provided the basis for this discussion of the development of cranial ornamentation in iguanians. Members of the taxon Phrynosoma are characterised, in part, by the presence of laterally and caudally directed cranial horns (Montanucci, 1987). These horns can initially be distinguished as protuberances of the squamosals and parietal of neonates (e-g. Phrynosoma modesfum; Fig. 6.5). Throughout ontogeny the protuberances may increase in size and become modified in appearance to form a variety of horns and bosses that are always confined to individual cranial elements. Cranial sutures are generally not obscured by the development of osseous ornamentation, although the degree of elaboration is taxonomically variable. Examination of cleared and double-stained, alcohol-preserved, and dried skeletal material of various specimens of Chamaeleo corroborates these findings. The role of epidermal structures in the developrnent of cranial protuberances is not presently understood, although the horns of phrynosornatids and Chamaeleo jacksonii are sheathed by a single, much enlarged conical scale in life. 6.4.2 Osteological Correlates

Early ontogenetic development is characterised by osseous outgrowths from cranial elements, resulting in the development of horn cores, bosses and protuberances in iguanians. Cranial ornamentation derived from elaborations of the head skeleton proper cannot be removed at any time during ontogeny without direct physical damage to the skeletal elements of origin. In addition, the cranial ornamentation of iguanians is restricted to individual elements and thus does not generally transgress sutura1 boundaries, except sometimes during late growth stages (e.g. Chamaeleo jacksonii). This condition appears to be paralleled by bovids in the development of frontal horn cores.

6.4.3 The Hypothesis

The morphology of the posterolaterally directed horn cores and superficial protuberances on iguanian crania appears to parallel the condition of cranial ornamentation demonstrated by ankylosaurs. Based on this osteological resemblance it is hypothesised that the cranial ornamentation of ankytosaurs developed in an analogous manner - Le. by elaboration of individual cranial elements.

6.4.4 Testing the Hypothesis

Requirements for satiçfying the hypothesis of cranial elaboration include: (1) the presence of ornamentation in immature specimens; (2) the restriction of ornamentation to individual elements. The developmental nature of ankylosaur skulls is such that the majority of cranial sutures are rarely visible. Examination of material referred to subadult Pinacosaurus grangeri (IVPP uncatalogued; Maryanska, 1971; 1977) (see Chapter 3 and Figs. 3.7A, 3.8A, 3.9, 3.10) and Euoplocephalus (see Chapter 4, sections 4.3.1, 4.3.5and 4.3.6; Fig. 4.13) suggests that this feature is related to ontogeny. While the cranial sutures are readily apparent (the result of an apparent lack of osteodermal ossification with the cranium; see section 6.3),they are not without osseous ornamentation. The squamosals, quadratojugals and supraorbitals al1 demonstrate the incipient development of cornuate I pyramidal bosses (see Chapter 3, sections 3.2.2.1,3.2.2.6 and 3.2.2.7). These bosses are direct outgrowths of individual cranial elements. Currently little is known about the osteology of subadult nodosaurids. Several disarticulated cranial elements from the Albian of Texas suggest that at least some subadult nodosaurids developed rugose, encrusting elaborations ("excrescences"; Jacobs et al., 1994). A number of nodosaurid taxa are also known to have developed incipient supraorbital, squamosal and quadratojugal bosses (e-g. Pawpawsaurus; see also Chapter 7, section 7.2, characters 5, 6 and 7)-

6.4.5 Conclusions

Osteological evidence suggests that the cranial ornamentation of ankylosaurs is partly the result of elaboration of outgrowths from cranial elements. Such expression is consistent throughout known ontogenetic stages, at least among members of the Ankylosauridae. There is no evidence, however, to support the hypothesis that al1 of the cranial ornamentation of ankylosaurs is derived in this way.

6.5 A Synthetic Approach

A cornparison of osteological correlates associated with extant developmental processes with those demonstrated by the fossil record, permits the re-appraisal of ontogenetic rnechanisrns in extinct taxa. Previous work had suggested two alternative and competing hypotheses in order to explain the development of cranial ornamentation in ankylosaurs. A review of modern squamates demonstrates the independent occurrence of each process within selected taxa, and presents the opportunity to determine the resulting osteological expression of each mechanism. Cornparisons of these skeletal correlates with ankylosaur cranial material suggest that ankylosaurs ernploy bath developmental processes. The degree of expression of each mechanism is highly variable among, and even within, taxa. Thus, the expressed pattern may be utilized to diagnose a taxon (e.g. ankylosaurids vs. nodosaurids), or a particular ontogenetic stage. Carpenter (1990) noted the variability of cranial ornamentation expressed within the taxon Edmontonia rugosidens, and suggested that it rnay reflect a state of maturation. Contention over the taxonomie validity of the rnany synonyms of Euoplocephalus is based in part on subtle characters associated with cranial ornamentation morphology (Penkalski, 1998; in press). Resolution of these issues is likely embedded within a greater overall understanding of developmental mechanics. Figure 6.1. Phylogenetic inference and the "extant bracket". A. Phylogenetic relationships of members of the Archosauria. The extant taxa are positioned as the two successive outgroups to the extinct clade. Taxa under specific consideration in bold. Explanation of nodes: node (1) Archosauria; node (2) Dinosauria; node (3) Ornithischia; node (4) Eurypoda. B. A decisively negative assessment for an unpreserved attribute, Le. neither bracketing taxon has the attribute. C. A decisively positive assessment for an unpreserved attribute, i.e. both bracketing taxa have the attribute. D. Equivocal assessrnent for an unpreserved attribute, Le. one taxon has the attribute, the other does not. See text for details. Modified from Witmer, 1997. Crocodylia

.m... & . Lesothosaurus

. m.. rnlmmil @ Ankylosauri . 3 :m.: Stegosauria

Aves

phy logenetic r- Crocody lia: (-) absent bracket -4 Ankylosau ria: (-) absent Aves: (-) absent B. r-Crocodylia: (+) present >An kylosau ria: (+) present - Aves: (+) present r- Crocody lia: (+) present Ankylosau ria: (+/-) equivocal

D. Aves: (-) absent Figure 6.2. Modified phylogeny of the Amniota, providing a phylogenetic framework for the taxa discussed in the text (after Estes et al., 1988; Witmer, 1997). Explanation of nodes: node (1) Squamata; node (2) Iguania; node (3) Scleroglossa. 1MAMMALIA r-CORYTOPHANIDAE *& *& PHRYNOSOMATIDAE POLYCHROTIDAE ACRODONTA GEKKOTA 3 ANGUIMORPHA LACERTOIDEA SCINCOIDEA CROCODY LIA ANKYLOSAURIA -AVES Figure 6.3. Heloderma suspectum subadult specimen (UCMZ[R] 2000.00 1; SVL = 192 mm), radiographic image illustrating the pervasive development of osteoderms. Note the pronounced development of osteoderms in the head region as compared to areas further caudal. Scale bar, 10 mm.

Figure 6.4. cf. Euoplocephalus thin-section (TMP 98.1l5.2), from an isolated fragment of cranium representing the rostral region. Section was taken in the transverse plane, with the periosteum towards the top of the page. Position of arrows indicate the junction between the two main histological layers. "a"the superficial layer of woven bone. "bthe deeper layer of highly remodelled Haversian bone. Scale bar, 1 mm.

Figure 6.5. Phrynosoma modestum neonate, cleared and double-stained (LACM 123344; SVL = 30 mm), dorsal view (right eye removed). Arrow indicates the position of the occipital (parietal) horn cores. Scale bar, 1 mm.

Chapter 7 Phylogenetic Analysis of Ankylosaur Crania

In order to estimate a phylogenetic position for "Gobisaurus"gen. et sp. nov. within the Ankylosauria, and, as a corollary, effectively review the hypothesized interrelationships of the ankylosaurian lineage, a cladistic analysis involving 22 ankylosaurs (including "Gobisaurus")was conducted. Given the limited material currently described and attributed to "Gobisaurus" (i.e. a single cranium), this evaluation was limited to characteristics of the cranium proper. A summary of taxa and specimens examined / sources consulted is presented in Table 7.1. Due to their fragmentary nature, several ankylosaurs known from partial cranial material (viz. Amtosaurus rnagnus [scored for five of the 41 characters], Hierosaurus sternberii [scored for one of the 41 characters], Maieevus disparoserratus [scored for five of the 41 characters] and Niobrarasaurus coleii [scored for four of the 41 characters]) were omitted from the analysis. Two outgroups were used to polarize the character states; Lesothosaurus diagnosticus, a basal ornithi schian , and Huayangosaurus taibaii, a basal stegosaur. Given the lirnited scope of this analysis, any genealogical inferences rnust be placed into the wntext of those advanced for ankylosaurs in general based upon more extensive representation of material. A cornparison of the results generated by this cladistic analysis is further considered in light of previously published patterns of the lineage, thereby testing the significance of cranial data in the establishment of ankylosaurian phylogeny. Rather than contrasting the cladistic hypotheses presented here with individual systematic appraisals, a reference hypothesis, based upon a consensus of previously circulated systematic estimates was compiled into a so-called supertree (sensu Sanderson et al., 1998). 7.1 Cladistic Analysis

A data set of 41 cranial characters (see section 7.2; Appendix 7.1) was cladistically analyzed using the program Phylogenetic Analysis Using Parsimony (PAUP), version 3.1.1 (Swofford, 1993) on a Macintosh II si cornputer. AH characters utilized were assumed to be of equal weight and al1 multistate characters were left unordered. The heuristic search algorithm was employed, providing a relatively rapid solution, although one which does not examine every possibility and therefore may not be optimal (Swofford, 1993). Character state optimization was conducted using the accelerated transformation algorithm (ACCTRAN; favoring repeated reversals over repeated origins) (Wiley et al., 4991). Multiple equally most parsimonious solutions were summarized using the strict consensus option, expressing regions of conflict (i.e. inconsistent topographies; Wiley et al., 1991) as polychotomies. By convention, phylogenetic hypotheses (i.e. cladograms) generated from cladistic analyses are statistically summarized with a number of basic tree measures, al1 of which PAUP 3.1.1 provides upon completion of the analysis. These measures inciude tree length (TL), consistency index (CI) and homoplasy index (HI). TL refers to the sum of al1 the character state changes along each branch of the cladogram. In general, the shortest tree represents the most parsimonious hypothesis (Wiley, et al., 1991). CI is a measurement of how well the data matrix "fits" the cladogram, with a low level of homoplasy resulting in a high Cl (the highest being 1.O) and a high level of homoplasy resulting in a low CI (Wiley, et al., 1991). Specificaliy, CI is equivalent to the minimum number of possible character state changes for a given cladogram divided by the actual number of character state changes. HI measures the amount of homoplasy, i.e. the actual CI subtracted from an optimal (theoretical) CI of 1.O (Wiley et al., 1991).

In accordance with current phylogenetic philosophy (see Ghiselin, 1984; Rowe, 1987; de Queiroz and Gauthier, 1990; 1992; Sereno, 1998) al1 phylogenetic taxa are unranked (in the Linnaean hierarchical sense). Furthermore, taxonomic definitions (clades delineated on the basis of common ancestry) are considered distinct from taxonomic diagnoses (clades characterized by synapomorphies). All suprageneric taxonomic nomenclature follows the definitions of Sereno (1998).

7.1.l Character Analysis

The characters selected for analysis are adapted (except for characters 10 and 32) and modified from previously published sources. Modifications to terminology have been made, where necessary, to improve clarity and precision. This has enabled a more extensive array of taxa to be effectively and accurately coded. A brief (but not exhaustive) summary of original sources is provided below for each character. Continuously variable (mensural) characters pose problems for character state designation, as the demarcation points are essentially arbitrary. Previously, however, such characters have been extensively empioyed in ankylosaur systernatics and thus have therefore been sirnilarly applied herein. In the case of such characters, precise andlor relative statements of proportion have been given whenever possible, and coding has been applied accordingly. While a number of original and cast specimens were consulted, published descriptions, illustrations and photographs were also relied upon in order to identify character states in many taxa. Additionally, systematic evaluation of ankylosaur cranial material is burdened with identifying the effects of taphonomic deformation of a highly unusual head skeleton. Many features, such as an extremely depressed lateral profile, suggest that the material recovered from the fossil record may not accurately portray the skeletal morphology as it existed in life. In an effort to circumvent erroneous interpretations of taphonomically induced cranial features, multiple specimens were examined (where possible), and a conscious effort was undertaken to recognize and identify al1 areas of non-biological morphology. 7.1.1.1 Characters of the dorsum of the cranium and overall cranial morphology 1. Maximum cranial dimension: cranium length greater than cranial width (O); cranial width equal to or greater than cranial length (1) (Coombs 1978; Coombs and Maryanska 1990; Tumanova 1987; Barrett et al. 1998; Kirkland 1998; Hill 1999; Sereno 1999). Maximum cranial length was determined by measuring the overall distance from the rostral-most tip of the premaxillary rostrum sagittally along the palatal surface to the caudal-most edge of the occipital condyle. Maximum cranial width was determined by measuring the distance (across the palatal surface) between the lateral-rnost edges of the quadratojugals. In taxa where the quadratojugals project far ventrally, careful attention was afforded to the estimation of cranial width, attempting to identify any taphonornic distortion.

2. Cranial roof in lateral profile rostral to orbits: flat (0);dome-1 ike (1) (Lee 1996; Carpenter et al. 1998; Sereno 1999).

3. Cranial roof in la feral profile caudal to orbits: f lat (0);dome-li ke (1) (Lee 1996; Hi11 1999).

4. Laferotemporal fenesfra in lateral vjew. open (O); closed (1) (Coombs 1978; Tumanova 1987; Carpenter et al. 1998; Kirkland 1998; Hill 1999).

5. Supraorbitai boss: absent (0);present, rounded protuberance, laterally oriented (1); present, longitudinal ridge, dorsolaterally oriented (2)(Sereno 1986; Lee 1996; Hill 1999).

6. Squamosal boss: absent (0);present, rounded protuberance (1); present, pyramidal protuberance (2) (Coombs 1978; Tumanova 1987; Sereno 1986; 1999; Barrett et al. 1998; Carpenter et al. 1998; Kirkland 1998; Hill 1999). 7. Quadratojugal projection: absent (0); present, rounded protuberance (1); present, deltaic protuberance (2) (Coombs 1978; Tumanova 1987; Sereno 1986; 1999; Barrett et al. 1998; Kirkland 1998; Hill 1999).

8. Cranial ornamenfation in a transverse plane across the anforbit absent (0); present, arnorphousfill-defined (1); present, distinct pattern of sculpturing consists of three or more flat polygons (2); present, distinct pattern of sculpturing consists of three or more bulbous polygons (3); present, distinct pattern of sculpturing consists of one or two flat polygons (4) (Coombs 1978; Kirkland 1998; Hill 1999; Sereno 1999).

9. Shallow furrows demarcate a single area of cranial ornamenfation between the external nares: absent (O); present (1) (Coombs 1978; Kirkland 1998).

IO. Skull table morphology: width between squarnosals greater than or equal to width between supraorbitals (O); width between squamosals less than width between supraorbitals (1) (new). The skull table refers to the area bounded by a transverse plane between the lateral-most points of the squarnosals and a transverse plane between the lateral-most points of the supraorbitals.

11. Raised nuchal sculpturing: absent (O); present (1) (Kirkland 1998; Hill 1999). A localized pattern of cranial sculpturing extending transversely across the dorsal surface of the nuchal shelf.

12. Nuchal shelf does not obscure occiput in dorsal view (0); obscures occiput in dorsal view (1) (Coombs 1978; Barrett et al. 1998; Carpenter et al. 1998; Kirkland 1998; Hill 1999). 7.1.1.2 Characters of the palate 13. Maximum premaxillary rostrum dimension: prernaxillary palate length equal to or greater than premaxillary palate width (O); premaxillary palate length less than premaxillary palate width (1) (Kirkland 1998; Hill 9 999; Sereno 1999). Maximum premaxillary palate length was determined by measuring the distance (sagittal plane along the palatal surface) between the midpoint of a line extended between the medial edges of the rostral-rnost tomial crests, and the midpoint of a line extended between the rostral-most maxillary teeth. Premaxillary palate width was determined by measuring the maximum transverse distance between the medial surface of the tomia and / or premaxillary teeth.

14. Maximum premaxillary rostrum width: less than the distance between the caudal-most maxillary teeth (O); equal to or greater than the distance between the caudal-most maxillary teeth (1) (Tumanova 1987; Kirkland 1998). Premaxillary rostrum width was determined by measuring the transverse distance between the lateral surfaces of the premaxillae at their maximum. The distance between the caudalmost maxillary teeth was determined by measuring the transverse distance between the lateral surfaces of the tooth-rows.

15. Premaxillary notch: absent (O); present (1) (Kirkland 1998; Hill 1999; Sereno ? 999).

16. Premaxillary tomia: restricted to an extreme rostral position (0); extend caudally, continuous with maxillary tooth row (1); extends caudaily, lateral to maxillary tooth row (2) (Carpenter et al. 1998; Hill 1999; Sereno 1999).

17. Premaxillary feefh: present (0);absent (1) (Coombs and Maryanska 1990; Carpenter et al. 1998; Kirkland 1998; Hill 1999; Sereno 1999). See also Appendix 3.2. 18. Maxillary tooth rows: straight, converge caudomedially (O); curved into an hourglass-shape, converge and then diverge (1) (Coombs 1978; Coombs and Maryanska 1990; Sereno 1986; 1999; Lee 1996; Carpenter et al. 1998; Kirkland 1998; Hill 1999).

19. Maxillary tooth cingulurn:absent (0); present (1) (Coombs 1978; Kirkland 1998; Hill 1999; Sereno 1999).

20. Longjtudinal bisection of the rostrum by the internasal cavity septum: incornplete (0); complete, (1) (Sereno 1986; 1999; Coombs and Maryanska 1990; Lee 1996; Carpenter et al. 1998; Hill 1999).

21. Secondary palate: absent (O); rostrodorsal palatal arch only (1); rostrodorsal and caudoventral palatal arches (2) (Coombs 1978; Sereno 1986; Tumanova 1987; Coombs and Maryanska 1990; Lee 1996; Carpenter et al. 7998; Kirkland 1998).

22. Buccal emargination: flat (O); strongly concave (1) (Coombs 1978; Sereno 3 986; Lee 1996).

7.1.1.3 Characters of the respiratory passageways 23. External naris proper: not visible in rostral view (0); visible in rostral view (1) (Tumanova 1987; Carpenter et al. 1998; Kirkland 1998; Hill 1999).

24. Nasal septum: absent (0); present, horizontal (1); present, vertical (2) (Hill 1999; Sereno 1999).

25. Nasal cavity proper: relatively linear orientation {O); convoluted (1) (Coombs 1978; Tumanova 1987; Kirkland 1998; Hill 1999; Sereno 1999). 26. Paranasal sinus cavities: absent (O); present (1) (Coombs 1978; Tumanova 1987; Sereno 1986; 1999; Carpenter et al. 1998; Hill 1999). Fortuitous broken specimens, physical sectioning, and non-invasive cornputer tomography have revealed blind-ended sinus cavities in some ankylosaur taxa (Coombs, 1978; Witmer, 1997). White these sinuses have generally been associated with the presence of nasal septa, the nature of this relationship remains to be assessed. A recently prepared specimen of Edmontonia rugosidens (TMP 98.71.1 ) indicates that members of this taxon do possess premaxillary sinus cavities, a feature previously unrecorded in the literature.

7.1 A.4 Characters of the pterygoid cornplex, occiput and basicranium 27. Rostral face of pterygoid body: directed caudally (0); directed vertically or rostrally (1) (Tumanova 1987; Sereno 1999).

28. Caudal margin of the pterygoid: rostral to (O); or in transverse alignment with (1) the ventral margin of the pterygoid process of the quadrate (Lee 1996; Sereno 1999).

29. Mandibular ramus of the pterygoid: directed parasagittally (O); rostrolaterally (1) (Coombs 1978).

30. Basipterygoid process - pterygoid contact: f used (O); unfused (1) (Coombs 1978; Tumanova, 1987; Sereno 1999).

31. Basisphenoid length: greater than basioccipital length (O); less than basioccipital length (1) (Sereno 1986; 1999).

32. Basal tubera morphology bulbous (O); rugose crest (1) (new) 33. Paroccipital process directed: caudolaterally (O); laterally (1) (Tumanova 1987; Carpenter et al. 1998; Hill 1999).

34. Occipital condyle composition: multiple elements (O); basioccipital only (1) (Sereno 1986; 1999; Kirkland 1998; Hill 1999).

35. Occipital condyle morphology in occipital view reniform (O); ovoid / round (1) (Sereno 1986; 1999; Tumanova 1987; Coombs and Maryanska 1990; Lee 1996; Carpenter et al. 1998; Hill, 1999).

36. Occipital condyle orientation: directly caudally (O); caudoventrally (1) (Sereno 1986; Kirkland 1998; Hill 1999).

37. Foramen magnum orientation: directly caudally (O); caudoventrally (1) (Coombs 1978).

7.1 .1.5 Characters of the quadrate 38. Quadrate lateral profile: bowed, rostrally convex, caudally concave (O); flat (1) (Hill 1999).

39. Quadrafe - paroccipital process contact: unfused (0); fused (1) (Tumanova 1987; Coombs and Maryanska 1990; Lee 1996; Carpenter et al. 1998; Hill 1999).

40. Quadrate condyle, lateral view visible (0); obscured by the quadratojugal(1) (Tumanova 1987).

7.1 .1.6 Other characters 41. Postocular shelf. absent (O); present (1) (Coombs 1978; Coombs and Maryanska 1990; Sereno 1986; 1999; Barrett et al. 1998). 7.1.1.7 Characters not used here but that have been used or discussed elsewhere Occipital condyle "ne&? Many ankylosaurs have previously been characterized by the presence or absence of a short, stout process connecting the occipital condyle to the main body of the basioccipital (see Coombs, 1978; Coornbs and Maryanska, 1990; Kirkland, 1998). However, personal examination of numerous cranial specimens representing several ankylosaur taxa revealed this purported feature to be highly variable, even within a species, and difficult to discretely diagnose and effectively score. Furthermore, many publications fail to make any reference to this feature and do not illustrate the basicranial area. Consequently, the utility of this feature as a clustering characteristic awaits further investigation.

7.1.2 Cladistic Results and Discussion

7.1.2.1 Cladistic analysis of al1 taxa under consideration (Fig. 7.1) Analysis of the data matrix resulted in 20,065 equally most parsimonious cladograms (TL =126; CI = 0.405; HI = 0.595). A strict consensus tree is presented in figure 7.1. Although largely unresolved, this analysis resuits in a novel genealogical position for both Animantanc ramaljonesi (i.e. within the An kylosauridae; contra Carpenter et al. 1999) and "Strufhiosaurus" transilvanicus (as a member of the basal polychotomy defining the Ankylosauria; contra Coombs and Maryanska, 1990). In addition, this hypothesis supports the notion of Minmi sp. as an ankylosaurid (Sereno 1998; 1999; contra Molnar 1980; Coombs and Maryanska 1990). However, as the scope of this analysis was restricted to aspects of the cranium, these statements await subsequent testing.

7.1.2.2 Cladistic analysis of al1 taxa coded for greater than 55% of the characters (Fig. 7.2) lncomplete preservation prevented a large number of taxa from being effectively coded for many of the characters. To examine the effect that these taxa had on the phylogenetic hypothesis, two additional analyses were conducted: the first eliminated al1 taxa that were able to be coded for 55% or fewer of the total number of characters; the second eliminated al1 taxa coding for fewer than 80% of the characters. For the first analysis, Shanxia tianzhenesis (24.4%), Nodocephalosaurus kirtlandensis (36.6%),"Struthiosaurus" (36.6%), Animantarx (41.6%), and Sauropelta edwardsi (51.2%) were eliminated. This elimination resulted in a reduction from 20,065 (for the complete data set) to 54 equally most parsimonious cladograrns being identified (TL = 120; Cl = 0.425; Hl = 0.575). Missing data frequently obscure synapomorphy assignment to any particular node. In situations where, under different optimizations, the synapomorphy defined a more inclusive clade, the character(s) was (were) considered ambiguous (denoted by placement within brackets). A strict consensus tree (Fig. 7.2) illustrates the following points of agreement: (1 ) The Ankylosauria is monophyletic, as defined by three unambiguous characters (quadratojugal projection present [characten]; amorphous antorbital cranial ornamentation [character 81; quadrate-paroccipital process fused [character 391) and four ambiguous characters (squamosal boss present [character 61; raised nuchal sculpturing present [character II]; rnaxillary teeth with cingulum [character 191; paranasal sinus cavities present [character 261). (2) The Ankylosauridae is rnonophyletic, as defined by two unambiguous characters (premaxillary palate wider than long [character 131; premaxillary notch present [character 151) and three ambiguous characters (rostral face of pterygoid vertically or rostrally inclined [character 271; mandibular process of the pterygoids directed rostrolaterally [character 291; postocular shelf present [character 411). (3) Shamosaurus and "Gobisaurus"are sister taxa (4) Tsagantegia is the sister taxon to the remaining ankylosaurines (5) Ankylosaurus and Euoplocephalus are sister taxa (6) Edmonfonia and Panoplosaurus are sister taxa 7.1.2.3 Cladistic analysis of al1 taxa coded for greater than 80% of the characters (Fig. 7.3) The third analysis further eliminated Minmi sp. (Molnar 1996) (56.1 %) and Talarurus plicatospineus (58.5%). This resulted in a further reduction to six equally most parsimonious trees of 1 14 steps being generated (CI = 0.447; HI = 0.553). The strict consensus tree is given in figure 7.3. While this hypothesis is considerably better resolved than the previous cladograms (i.e. there are fewer polychotomies as compared to Figs. 7.1 and 7.2), the results are consistent with the earlier proposed genealogies (see above).

7.1.2.4 Cladogram cornparison All three cladograms (Figs. 7.1 - 7.3) illustrate the following congruent statements: (1) the Ankylosauria is monophyletic; (2) Sharnosaurus and "Gobisaurus" are sister taxa; and (3) Edmontonia and Panoplosaurus are sister taxa. Furthermore, sequential examination of these hypotheses supports the notion that the elimination of poorly coded taxa may increase the potential resolution of a phylogenetic hypothesis (although see Wu et al. 1996 for contradictory results).

7.2 Systematics

Archosauna Cope 1869

Dinosauria Owen 1842 Ornithischia Seeley 1888

Thyreophora Nopcsa 1915 Eurypoda Sereno 1986

Ankylosauria Osborn 1923 Emended cranial diagnosis (based on figure 7.3) - Members of the Ankylosauria differ from al1 other Ornithischia by the following unequivocal adult characters: presence of pronounced ventrolateral elaboration of quadratojugal; developrnent of cranial ornamentation in transverse plane across rostrum; and fused contact between paroccipital process and quadrate. In addition, ankylosaurs also demonstrate the following ambiguous characters: presence of discrete squamosal boss; raised nuchai sculpturing; paranasal sinus cavity(ies); and occipital condyle made of basioccipital only.

Ankylosauridae Brown 1908

Emended cranial diagnosis (based on figure 7.3) - Members of the Ankylosauridae differ from al1 other Ankylosauria by the following unequivocal adult characters: prernaxillary palate wider than long; and presence of premaxillary notch. In addition, ankylosaurids also demonstrate the following ambiguous characters: rostral surface of pterygoid either vertical or slightly rostraily inclined; mandibular ramus of pterygoid rostrolaterally directed; occipital condyle oriented caudoventrally; foramen magnum oriented caudoventrally; and postocular shelf.

"Gobisaurus"gen. nov.

Type and only species - "Gobisaurus" gen. et sp. nov.

Distribution - Ulanhushao (Suhongtu) Formation, Early Cretaceous (Aptian - ?Albian) of the Alshan Desert, Nei Mongol Zizhiqu (Inner Mongolia), China (Fig. 7.4). Etymology - Gobi (English) refers to the geographic locale; sauros (ancient Greek), lizard.

Cranial diagnosis - Large ankylosaurid with orbit 20% of cranial length and nasal vestibule 23% of cranial length; robust basipterygoid process not fused to main pterygoid body; elongate premaxillary process of the vomer visible in palatal view. Similar to Shamosaurus in having deltaic dorsal profile, with premaxiilae forming the narrow apex. Differs from Shamosaurus in that cranium longer than wide; lacks discernible cranial sculpturing over antorbital region; maxillary tooth rows relatively shorter; maximum rostral width less than distance between caudal-most maxillary teeth; and reduced supraorbital boss (see Chapter 7, Figs.

"Gobisaurus" gen. et sp. nov. Figures 5.1 - 5.7

Holotype - IVPP V 12563*

Type locality - IVPP V 12563* is believed to have been collected from the same general locality as the large theropod Chilantaisaurus maortuensis, approximately 60 km north of Chilantai (Jilantai), on the east side of Chilantai Salt Lake (Chilantaiyen Chih), Maortu, Alashan Desert, Nei Mongol Zizhiqu (Inner Mongolia), China.

Craniai diagnosis - As for the genus, as it is the only known species. Cranial description - See Chapter 5.

Ankylosaurinae (Brown 1908)

Emended cranial diagnosis (based on figure 7.3) - Members of the Ankylosaurinae differ from al1 other Ankylosauridae by the following unequivocal adult characters: nuchal shelf obscures the occiput in dorsal view; and quadrate condyle obscured in lateral view by quadratojugal. In addition, ankylosaurines also demonstrate the following ambiguous characters: sagittally positioned vomers completely bisecting the rostrum; and convoluted nasal cavity proper.

Euoplocephalus Lambe 1910 (= the following junior objective synonym: Stereocephalus Lambe 1902 1 Arribalzaga 1884; = the following junior subjective synonyms: Dyoplosaurus Parks 1924; Scolosaurus Nopcsa 1928; Anodontosaurus Sternberg i929).

Type and only species - Euoplocephalus tutus Lambe ( 1902)

Distribution - The genus is currently known from Late Cretaceous (mid-Campanian to Late Maastrichtian) deposits of the Western lnterior of North America. Specifically, material referable to Euoplocephalus has been collected from the Judith River Group (particularly the Dinosaur Park Formation) and Horseshoe Canyon Formation of southern Alberta, and the Two Medicine Formation of Montana (see Fig. 1.2; Table 4.1 ).

Etymology - Euoplos (ancient Greek), well arrned; kephale (ancient Greek), head. Emended cranial diagnosis - Medium to large sized ankylosaurid with unique pattern of cranial sculpturing across rostral region; small teeth with fluting not corresponding to position of denticles along apical carina; no cingulurn; "ciliary" supraorbital located within orbital cavity; and shallow nasal vestibule. Similar to Ankylosaurus in having small teeth; distinct pattern of three or more flat polygons forming the cranial sculpturing across the antorbital area; and caudal margin of the pterygoid in transverse alignment with the ventral margin of the pterygoid process of the quadrate. Differs from Ankylosaurus in that premaxilla is virtually devoid of cranial ornamentation; intranarial cranial sculpturing forms more than one polygon in transverse plane; externat naris proper is visible in rostral view; and quadrate - paroccipital process contact unfused. Differs from other ankylosaurines (except Tsagantegia) in having a vertical narial septum. Differs from al1 other ankylosaurines in that the maxillary tooth rows converge rnedially and diverge rostrally and caudally.

Euoplocephalus tutus Lambe (1902) Figures 2.1,4.1 - 4.16 (= the following junior objective synonym: Stereocephalus tutus Lambe 1902; = the following junior subjective synonyms: Dyoplosaurus acutosquameus Parks i924; Scolosaurus cutleri Nopcsa 1 928; Anodontosaurus Iambei Sternberg 1929. The name Palaeoscincus asper Lambe 1902 is considered a nomen dubium [Coombs, 19781).

Holotype - NMC 021 O* Type Locality - Canadian Museum of Nature (formerly the National Museum of Canada) locality number P-189704, Upper Cretaceous, Dinosaur Park Formation (formerly the Oldman Formation), East side of mouth of Berry Creek at the Red Deer River, Alberta, Canada. Unfortunately, the actual quarry site was never formally placed ont0 a map, nor were any specific CO-ordinatesgiven, and consequently the exact locality is presently unclear (Fig. 1.2).

Etymology - tutus (Latin), protected from danger or han; safe, secure.

Emended cranial diagnosis - As for the genus, as it is the only known species.

Cranial description - See Chapter 4.

7.3 Reference Hypothesis - The Ankylosaur Supertree

The significance of cranial data for the elucidation of ankylosaur phylogeny was reviewed through a set of direct comparisons between the cladistic hypotheses proposed above (Figs. 7.1 - 7.3) and a series of reference hypotheses. Each cladistic I reference hypothesis couplet examined the same set of taxa. The reference hypotheses were derived from a supertree analysis, whereby numerous previously published data sets are synthesized, providing a concise consensus of earlier systematic efforts. In effect, a supertree is a "majority rule summary" (Bininda-Emonds et al., 1999), integrating al1 manner of data (including those traditionally considered incompatible such as morphology and rnolecular evidence) by coding for tree topography and not character states. Previous supertree analyses have been complied for extant primates (Purvis, 1995a) and carnivorans (Bininda-Emonds et al., 1999). The curent instance is the first attempt at conducting a similar approximation for a fossil clade. The advantages of such a technique include the possibility of a greater variety of sources analyzed (including cladistically and non-cladistically generated trees); al1 data being combined equally (small sets of primary data are not necessarily overshadowed by larger sets); and the inclusion of sources that may possess different sets of terminal taxa (Purvis 1995a; b; Bininda-Emonds et al., 1999). However, the supertree method is not without drawbacks; it is heavily biased against new and 1 or unorthodox systematic statements and taxa for which iittle information is available collapse branches and result in polychotomies. In the case of fossil vertebrates, many taxa are known from meager material, much of which is not diagnostic and / or extremely fragmentary. To circumvent this issue, taxa that have not yet received systematic representation in the literature (beyond assumptions of suprageneric assignment) were excluded. While it is acknowledged that this somewhat arbitrary elimination of taxa partially undermines the basis for conducting a supertree estimate, there appears to be littte point in estimating large, unresolved genealogical patterns. Ultirnately, taxa were selected based on the relative amount of systematic information available. In addition, no distinction was made between data sources with regard to differential weighting.

7.3.1 Supertree Methodology

A brief summary of the methodology (as outlined by Bininda-Emonds et al., 1999) is as follows (see also Purvis, 1995a; 6). Data for the supertree analysis were drawn from two major sources: phylogenetic analyses (such as those of Kirkland, 1998 and Sullivan, 1999) and non-phylogenetic based branching (tree) diagrams (such as those of Coombs, 1978 and Tumanova, 1987). Each data set was individually coded (see below) and the results entered into a consensus matrix (Appendix 7.2; see also Fig. 7.6A, B). Nodes on the branching diagrams (and cladograms) were coded into a series of binary characters: taxa defined by the node under examination were scored a "1"; sister taxa to the defined node were scored a "O" (Baum, 1992; Regan, 1992). Following the methodology of Purvis (1995b). missing taxa were scored a "?". A hypothetical all-zero outgroup was used to polarize the statements. The resulting matrix (Appendix 7.2) was constructed using the data editor of MacClade 3.05 (Maddison and Maddison, 1992) and analyzed using the heuristic search algorithm of PAUP 3.1.1 (Swofford, 1993).

7.3.2 Supertree Estirnate (Fig. 7.4)

An example supertree estimate was conducted by coding 107 characters from 14 previously published source trees (Table 7.2, Appendix 7.2). The resulting hypothesis is virtually resolved (see Fig. 7.6C). Within the Ankylosauria, three distinct, monophyletic lineages are apparent; (1) the Ankylosauridae (al1 ankylosaurs closer to Ankylosaurus than to Panoplosaurus), (2) Ankylosaurinae (al1 ankylosaurids closer to Ankylosaurus than either Shamosaurus or Minmi) and (3) the Nodosauridae (al1 ankylosaurs closer to Panoplosaurus than to Ankylosaurus) (Sereno, 1998). The clade Nodosaurinae (al1 nodosaurids closer to Panoplosaurus than to either Sarcolestes or Hylaeosaurus; Sereno, 1998) could not be identified based on the current data set. It should be noted that in this preliminary analysis, only taxa having received previous systematic representation in the literature (beyond arbitrary suprageneric assignments) were considered. Consequently, Animantam, "Gobisaurus", Shanxia and Tianzhenosaurus could not be considered at this tirne.

7.3.3 Supertree - Cladistic Cornparison

The following supertree analyses incorporate topographic data generated from the previously described cladistic estimates based on cranial characters (Figs. 7.1 - 7.3). 7.3.3.1 Supertree - figure 7.1 cornparison (Fig. 7.7) Analysis of the supertree data matrix including topographic data from Figure 7.1 (consisting of al1 taxa under consideration) resulted in 1608 equally most parsimonious trees of 144 steps (CI = 0.784; HI = 0.21 5). A strict consensus tree is presented in figure 7.7. The supertree differs from the cranially generated cladogram on two points; the position of "Strufhiosaurus" and the position of Pawpawsaurus (see Fig. 7.5). Although both estimates agree that (1) Edmontonia and Panoplosaurus are sister taxa and (2) Shamosaurus and "Gobisaurus" are sister taxa, neither hypothesis is well resolved.

7.3.3.2 Supertree - figure 7.2 comparison (Fig. 7.8) Analysis of the supertree data matrix including topographic data from figure 7.2 (eliminating al1 taxa coding for fewer than 55% of the cranial characters; Appendix 7.2) resulted in 126 equally most parsimonious trees of 143 steps (CI = 0.818; Hl = 0.182). A comparison of the strict consensus tree with the cladograrn figure 7.2 (Fig. 7.8) indicates that the two estimates are virtually identical, differing only in the relative position of Gargoyleosaurus within the Ankylosauridae and the degree of resolution of the clade Nodosauridae (see Fig. 7.8). Both hypotheses share the following points of agreement: (1) The Ankylosauridae is monophyletic (2) The Nodosauridae is monophyletic (3) Ankylosaurus and Euoplocephalus are sister taxa (4) Shamosaurus and "Gobisaurus" are sister taxa (5) Edmontonia and Panoplosaurus are sister taxa

7.3.3.3 Supertree - figure 7.3 cornparison (Fig. 7.9) A final supertree analysis, including topographic data from figure 7.3 (eliminating al1 taxa coding for fewer than 80% of the cranial characters; Appendix 7.2), resulted in seven equally most parsimonious trees of 141 steps (CI = 0.844; HI = 0.156). A comparison of the supertree estimate with the cladogram presented in figure 7.3 (Fig. 7.9) suggests that while the supertree hypothesis better resolves the interrelationships between members of the Nodosauridae, the genealogy of ankylosaurine ankylosaurids remains largely indeterminate in both versions.

7.3.3.4 Supertree comparison AH three supertrees (Figs. 7.7A, 7.8A, 7.9A) illustrate the following congruent statements: (1 ) the Nodosauridae is monophyletic; (2) Shamosaurus and "Gobisaurus" are sister taxa; and (3) Edmontonia and Panoplosaurus are sister taxa. Unresolved genealogical patterns within the Ankylosauria (specifically within the Ankylosauridae) suggest an incornplete fossil record and 1 or poor representation within the literature.

7.3.3.5 Supertree concl usion Supertree analysis is a useful tool for integrating numerous, potentially non-compatible data sets in order to provide a comparative foi1 against which recently generated hypotheses may be contrasted. Among ankylosaurs, the conclusions generated by both cladistic and supertree analyses do appear to closely resemble one-another (see Figs. 7.7 - 7.9), testifying to the importance of atomizing the cranium for use in characterizing (and clustering) taxa. Although somewhat limited in scope, evaluations of the cranium have traditionally provided the majority of characteristics by which ankylosaurs have been genealogically appraised.

7.4 Concluding Comments

As a source of phylogenetic information, the cranium is, unquestionably, one of the most important resources available. The extent to which this osteological region has shaped (or perhaps prejudiced) Our understanding of genealogy may be qualifted by a direct comparison between a reference hypothesis (summarizing al1 systematic efforts) and a phylogenetic evaluation relying solely upon cranial data. The obvious reliance upon the cranium as a rich source of systematic data (see Hanken and Hall, 1993) is further exemplified by a number of preservational features of ankylosaurs, including the preference for institutions to collect skulls over (less dramatic) postcrania and the difficulty in taxonomically assessing incomplete postcrania. In addition, there appears to be a heavy bias against the investigation and evaluation of Asian taxa, even for well-represented clades such as Pinacosaurus (illustrated by the collapse of nodes into polychotornies). Although the cranium clearly provides a profusion of systematic information, a more cornplete review of ankylosaur phylogenetics and evolution is situated within a more detailed evaluation of a less skewed data set. Figure 7.1. A strict consensus tree of the 20,065 most parsimonious cladograms (126 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence. For each taxon, unambiguous ("unequivocal") characters are followed by ambiguous ("equivocal") characters in brackets, arbitrarily selected from a single most parsimonious cladogram. Presence of a dash ("-Il) symbol pnor to a character indicates the reversal of the character; an asterisk (""') indicates an unresolved homology. Thus, the statement "Gobisaurus? (-5), suggests that "Gobisaurus"is characterized by a single character (character 5) that undergoes a reversal, and which may be a synapomorphy for a more inclusive clade under a different set of opimization parameters (Le. not ACCTRAN), given the large amount of missing data present. "Gobisaurus. (-5); Shamosaurus: 1, 8, -30, (-14, -27); Edmontonia: 27, 36; Panoplosaurus: -5, 11. The open star symbol indicates these taxa have previously been considered members of the Nodosauridae, contra the findings presented here (see text for details). SHAMOSAURUS O O

O SAICHANIA O O

PINACOSA URUS O 1- 1- TIANZHENOSAURUS 1- 1- TARCHIA TALARURUS O mO EUOPLOCEPHALUS "traditionat" ANKYLOSAURUS ankylosauriâs TSAGANTEGIA ANIMANTARX* O MlNMl O O NODOCEPHALOSAURUS i

O SHA NXIA O

k PA WPAWSAURUS "traditional" nodosaurids

SAUROPELTA O -k .O O O SlL VISAURUS O O "STRUTHIOSAURUS" """"" HUAYANGOSAURUS Figure 7.2. A strict consensus tree of the 54 most parsimonious cladograms (120 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence produced by eliminating Animantarx, Nodocephalosaurus, Sauropelta, Shanxia, and "Struthiosaurus" (taxa coded for fewer than 55% of the available characters; see Appendix 7.1 for data matrix). "Gobisaurus": (-5); Shamosaurus: 1, 8, -14, -30, (-27); Euoplocephalus: 18; Ankylosaurus: 9, -23, -27, 39; Tsagantegia: 20, 31, 33, (5, -27); Gargoyleosaurus: -1 9, 22; Edmonfonia: 27, 36; Panoplosaurus: -5, (1 1); Huanyangosaurus: 38. %OBISAURUS" 4 SHAMOSAURUS

TIANZHENOSAURUS TARCHIA TALARURUS EUOPLOCEPHALUS 1 ANKYLOSAURUS -TSAGANTEGlA MINMI GASTONIA GARGOYLEOSAURUS PANOPLOSAURUS EDMONTONIA PA WPALVSAURUS SlL VISAURUS HUA YANGOSAURUS LESOTHOSAURUS Figure 7.3. A strict consensus tree of 6 most parsimonious cladograms (114 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence produced by eliminating Minmi and Talarurus (taxa that coded fewer than 80% of the available characters) in addition to those removed from figure 7.2 (Anirnantarx, Nodocephalosaurus, Sauropelta, Shanxia, and "StruthiosaurusJ';taxa coded for fewer than 55% of the available characters; see Appendix 7.1 for data matrix). "Gobisaurus": (-5); Shamosaurus: 1,8, -1 4, -30, (27); Tarchia: -36, (20); Euoplocephalus: 18; Ankylosaurus: 9, -23, (27, 39); Tsaganfegia: -31, -33, -36, (5, 27); Gastonia: 2, 28, -30, (34);Gargoyleosaurus: 22; Edrnontor!ia: 27, 36; Panoplosaurus: -5, (1 1); Nuayangosaurus: 38. "GOBISAURUS" SHAMOSAURUS SAICHANIA PINA COSAURUS TIANZHENOSAURUS TARCHlA EUOPLOCEPHALUS ANKYLOSAURUS TSAGANTEGIA GASTONIA GARGOYLEOSAURUS - -- PANOPLOSAURUS 7,8939, (6, - EDMONTONlA 1l,26,34)

I PA WPAWSA URUS I SlL VlSAURUS I HUAYANGOSAURUS ' LESOTHOSAURUS Figure 7.4. Schematic map of China and Mongolia illustrating the approximate locality of "Gobisaurus"gen. et sp. nov. (black star symbol). Circles denote urban centres (capital cities underlined). Scale bar, 1000 km. Nei Mongol Zizhique

People's Republic of China Figure 7.5. A. Schematic map of North America indicating the position of Alberta and Montana. B. Schematic map of Alberta. Star symbol indicates the position of Dinosaur Provincial Park; circles denote urban centres (provincial capital underlined). The type specimen of Anodontosaurus lambei Sternberg 1929 (NMC 8530') was collected frorn near Drumheller. C. Schematic map of Dinosaur Provincial Park, indicating the location of Berry Creek, near which Lambe collected both the type specimen of Euoplocephalus tutus Lambe (1902) (NMC 021 O*) and the type specimen of Palaeoscincus asper Lambe 1902 (NMC 1349*). Berry Creek J Figure 7.6. The supertree methodology. A. Previously published an kylosaur systematic evaluation (modified from Coombs, 1978), with nodes labeled (1 - 5) and a hypothetical outgroup added. B. Consensus data matrix generated from the aforementioned systematic evaluation, with nodes coded into a series of binary characters ("O1'for sister taxa to the defined node; "1" for the node under consideration) and the hypothetical outgroup scored as al1 zeros (Purvis, 1995b; see section 7.2.1 ). C. An example reference hypothesis generated from the supertree analysis of 14 previously published sources, coding for a total of 107 characters (see Appendix 7.2). Only taxa receiving previous systematic representation in the literature (beyond arbitrary suprageneric assignments) considered. Consequently, the positions of Animantarx, "Gobisaur~s'~,Shanxia, and Tianzhenosaurus are not figured. ANKYLOSAURUS EUOPLOCEPHALUS ANKYLOSAURUS 1 1 1 O O EUOPLOCEPHALUS 1 1 1 00 PANOPLOSAURUS PlNACOSAURUS 1 1 O O O PAlVOPLOSAURüS 1 O O 1 1 A SIL WSAURUS SIL VISAURUS 10011 SAUROPEL TA 10010 OUTGROUP O0000 OUTGROUP

ANKYLOSAURUS EUOPLOCEPHALUS SAICHANIA TARCHIA PlNACOSA URUS TALARURUS

GARGOYLEOSAURUS GASTOMA -NODOCEPHAL OSAURUS F- r EOMONTONlA I PANOPLOSAURUS SIL WSAîJRUS SAUROPELTA œSTRUTH/OSAURUS" I -PA WPA WSAURUS -OUTGROUP Figure 7.7. Supertree - figure 7.1 comparison. A. A strict consensus of 1608 most parsimonious reference hypotheses (144 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on the supertree analysis of 14 previously published sources and the tree topography scored from figure 7.1. B. Figure 7.1, a strict consensus of the 20,065 most parsimonious cladograrns (126 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence. Italicized letters (in circles) indicate regions where the two estimates differ: a, the position of "Sfruthiosaurus"; b, the position of Pawpa wsaurus.

Figure 7.8. Supertree -figure 7.2 cornparison. A. A strict consensus of 126 most parsimonious reference hypotheses (143 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on the supertree analysis of 14 previously published sources and the tree topography scored from figure 7.2. B. Figure 7.2,a strict consensus of the 54 most parsimonious cladograms (120 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence. ltalicized letters (in circles) indicate regions where the two estimates differ: a, the relative position of Gargoyleosaurus; b,the position of Pa wpa wsaurus.

Figure 7.9.Supertree - figure 7.3 comparison. A. A strict consensus of seven most parsimonious reference hypotheses (141 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on the supertree analysis of 14 previously published sources and the tree topography scored from figure 7.3. B. Figure 7.3, a strict consensus of the six most parsimonious cladograms (1 14 steps each) illustrating the hypothesized genealogy of the Ankylosauria based on cranial evidence. Italicized letters (in circles) indicate regions where the two estimates differ: a, the degree of resolution of the Ankylosauridae; b, the position of Pawpa wsaurus.

Table 7.1. Skull material examined during the course of this study. Institutional abbreviations listed in Chapter 1. Asterisk C) indicates type specimen.

Brown, 1908; Coombs, 1978; Coombs and Maryanska, AMNH 5895*; AMNH 5214; NMC 8880 1990 NMC 8531"; AMNH 3076'; AMNH 5381'; Sternberg, 1928; Gilmore, 1930; Carpenter, 1990; Coombs AMNH 5665'; USNM 118682 and Maryanska, 1990 see Appendix 4.1 Chapter 4 DMNH 27726" Carpenter et al., 1998 CEUM 1307* Kirkland, 1998 IVPP V12563* Chapter 5 none Molnar, 1996; Coombs and Maryanska, 1990 none Sullivan, 1999 Russell, 1940; Carpenter, 1990; Coombs and Maryanska, NMC 2759*;TMP 83.25.2 l99O cast of SMU 73203 tee, 1996; Carpenter and Kirkland, 1998 AMNH 6523*; four uncatalogued subadult Gilmore, 1933; Maryanska, 1971; 1977; Tumanova, 1987; specimens from the IVPP Coombs and Maryanska, 1990 Maryanska, 1977; Tumanova, 1987; Coombs and cast of GI SPS 1001151 Maranska, 1990 Coombs and Maryanska, 1990; Carpenter and Kirkland, AMN W 3035 rd 1998 W O Table 7.1. Continued.

none Barrett et al., 1998 Eaton, 1960; Coombs and Maryanska, 1990; Carpenter none and Kirkland, 1998 cast of BM(NH) R4966 Coombs and Maryanska, 1990; Weishampel et al., 1991 Maryanska, 1977; Tumanova, 1987; Coombs and none Maranska, IWO none Tumanova, 1987; Coombs and Maryanska, 1990 none Pang and Cheng, 1998 none Tumanova, 1993

Weisharnpel and Witmer, 1990; Sereno, 1991 Galton, -lWO; Sereno and Zhimin, 1992

Footnotes: Specimens referred to the taxon Edmontonia rugosidens Specimen referred to the taxon Edmontonia longiceps Table 7.2. Source of original data for the supertree analysis (including the cladistic analyses), and the number of characters therein provided.

Bakker, 1988 Carpenter, 1998 Coombs and Maryanaska, 1990 Coombs, 1978 Fastovsky and Weishampel, 1996 Hill, 1999 Kirkland, 1998 Lee, 1996 Maryanska, 1977 Sereno, 1986 Sereno, 1999 Sullivan, 1999 Tumanova, 1983 Tumanova, 1987

Figure 7.1 Figure 7.2 Figure 7.3 Chapter 8 Summary and Prospectus

"... I rnust make it clear that this work in no way pretends to be a study of the complete rnorphology of the skull, . . . " (de Beer, 1937).

8.1 The Skull

Notwithstanding the paramount role the skull plays in deterrnining much of the biology of vertebrate taxa, detailed osteological treatments for many clades do not exist. Owing to problems with the establishment of homologies, lineages with atypical morphologies appear to be frequently omitted frorn consideration. Vertebrate palaeontologists are further encumbered with the disfiguring effects of taphonomic deformation. This effort seeks to approach the issue of evaluating unusual skull morphologies using a comparative evaluation. In order to enhance the utility of such descriptive work (both within and beyond the taxa under consideration), this review has adopted the nornenclatural methodology previously ascribed to birds and (to a lesser degree) mammals (Le. the Nomina Anatomica Avium [Baurnel and Witmer, 19931; Nomina Anafomica Veferinana [1983]; see also Witmer, 1997). The systematic rnorphological approach employed permitted each element to be explicitly reviewed in a rational progression, thereby promoting discrete analysis of the individual components of the integrated skull architecture.

8.2 The Ankylosauria

The Ankylosauria is a monophyletic assemblage of herbivorous dinosaurs (Archosauria: Ornithischia) with a distinctive skull morphology. Members differ frorn al1 other ornithischians in demonstrating a unique combination of adult characters, including the absence of the antorbital, dorsoternporal and mandibular fenestrae, the nearly complete obliteration of cranial sutures and the pervasive developrnent of cranial ornamentation across the skull dorsum (Coombs and Maryanska, 1990; Carpenter, 1997a; b; Sereno, 1998; see also Chapter 3, section 3.1). Although both the dorsotemporal and mandibular fenestrae remain open in al1 non-ankylosaurian ornithischians, Witmer (1997) noted that there is an independent trend (Le. homoplasy) within the various lineages towards the reduction of the antorbital cavity (the area bounded by the antorbital fenestra) (see also Sereno, 7 986). However, the basic pattern of ornithischian cranial architecture does not lend itself to casual comparison with that of ankylosaurs. This study focused on examining possible congruent patterns of morphological similarity within the Ankylosauria, thus permitting the effective review of these evolutionary trends in the context of successive outgroups. The unique morphology of the ankylosaur cranium rnay explain certain aspects of the preservational quality observed in specirnens, particularly those from North America. In contrast to most other dinosaurs, ankylosaur crania are rarely found disarticulated (pers. obs.). Rather, the crania frequently (perhaps inevitably) undergo sorne degree of post-rnortem plastic deformation, whether it be depressional, compressional or a cornbination of the two (see Chapter 2, section 2.1 -1). This deformation appears to be the result of sorne forrn of taphonomic process. Complete fusion of the sutures and the developrnent of craniai omamentation may preclude the cranium from readily becoming disarticulated. The resulting plastic deformation has deceived rnany previous workers, and has hindered an accurate analysis of the actual morphological pattern of the skull. Furtherrnore, the virtual lack of elernent delineation across the dorsum has obliged most authors to focus upon aspects of the palate and occiput. A review of the literature suggests that a fundamental understanding of ankylosaur skull morphology is presently not available. Despite previous work, much of the head skeleton remains undescribed or inadequately described. Biological information derived from inaccurate descriptions rnay lead (or may have already led) to erroneous interpretations and impede subsequent research. Moreover, the fundarnental role of the skeleton in providing the foundation for the inference of unpreserved attributes (see Bryant and Russell, 1992; Witmer, 1995; 1997) predicts that baseline errors are amplified within higher order suppositions (e.g. soft tissue inference and evolutionary patterns). ln contrast to most previous work (with the notable exception of Coombs, 1978), this effort is comparative, with a rigorous reappraisal and subsequent synthesis of multiple specimens of multiple taxa. To facilitate ease of comparison and description, the ankylosaur skull was subdivided into five mutually exclusive topographic regions (rostral, temporal, palatal, occipital 1 basicranial and mandibuiar). Although the assignment of particular elements to such topographic regions is to some extent arbitrary, it permits a more definitive and extensive evaluation to be conducted. In terms of the overall architectural patterns, the resulting morphological review does not differ significantly from those forwarded by most authors. However, this reappraisal does provide more detail on a greater variety of features 1 elements afid permits a more substantiated assessment of the pattern within the clade as a whole. In addition, the incorporation of a standardized terminology, previously established for extant taxa, will enable future efforts to exploit a more inclusive morphological assessment (i.e. outside the Ankylosauria).

8.3 Euoplocephalus tutus Larnbe (1902)

Following the establishment of the basic morphological pattern and descriptive procedure of the generalized ankylosaur skull (Chapter 3))two taxa were specifically reviewed; one previously described from several specimens (Euoplocephalus tutus Lambe [1902]), the other a single specimen of a "new" taxon ("Gobisaurus"gen. et sp. nov.). With the aforementioned review of the clade Ankylosauria providing the morphological framework, these descriptive efforts focused on comparing the general "synthesized" pattern with individual, discrete lineages. Chapter 4 focused on a review of Euoplocephalus, a well- represented taxon from the Late Cretaceous of Alberta and Montana. Euoplocephalus is something of an "archetype" for the Ankylosauria, forming the basis for much of what is currently known about armoured dinosaur biology (see Haas, 1969; Coombs, 1972; 1978; 1990; 1995; Coombs and Maryanska, 1990; Carpenter, 1997a; b). Specific morphological considerations (following the aforementioned descriptive approach), coupled with a cornparison of the generalized skull architecture provided in Chapter 3, enabled a more comprehensive appraisal than was formerly possible. Although it has not been officially contested since Coombs' 1978 review, some workers believe the taxon Euoplocephalus rnay represent more than one genus - species combination (e.g. Carpenter, 1997b). Grande and Bernis (1998) argued that within the fossil record, there is little justification in distinguishing new taxa without sufficient osteological evidence. Detailed morphological evaluation of specimens attributed to Euoplocephalus, including type material of junior synonyms, failed to identify any discrete morphotypic clusters. Certainly each specirnen demonstrates a unique morphological configuration. However, as noted earlier, taphonomic deformation is common, and is no doubt a major contributing factor (see Fig. 2.1 ). Furthermore, al1 of the differences are subtle, and intraspecific and ontogenetic morphological variation cannot be ruled out. Consequently, there exists no unique combination of primitive and derived characters and no compelling autapomorphies (unique - derived characters) that preferentially group any specimens together. Thus, this review concludes definitively that, at the present time, cranial evidence supports the recognition of a monotypic Euoplocephalus.

8.4 "Gobisaurus" gen. et sp. nov.

Description of the taxon "Gobisaurus" permitted the morphological assessrnent of an undescribed clade (Chapter 5)based on limited skeletal evidence, following the methodological procedures previously outlined. Although highly distinctive when compared to such North American forms as Euoplocephalus, the morphology of "Gobisaurus" is not entirely unique. Shamosaurus scutatus Tumanova 1983, from the Early Cretaceous (Aptian - Albian) of Mongolia demonstrates a number of shared morphological features. lncluded amongst these are: a deltaic dorsal profile; a narrow premaxillary rostrum; rounded squamosal bosses and quadratojugal protuberances; and large, prominent rostrolaterally oriented orbits and external nares (in Shamosaurus, the orbits and external nares represent approximately 18% and 19% respectively of the cranial length). However, both taxa may be diagnosed by a suite of unique apomorphic (derived characters) and plesiomorphic (primitive characters) character state ccmbinations (see Chapter 7, Appendix 7.1 ). Furthermore, each taxon is definable by one or more autapomorphies (Shamosaurus reportedly has a maxillary tooth row approximately 50% of the cranium length ~umanova,1983; Coombs and Maryanska, 19901; "Gobisaurus" has elongated premaxillary processes of the vomer visible in palatal view). Consequently, the validity of "Gobisa~rus~~as a distinct taxon is supported by its morphology. As a corollary, the utility of the descriptive procedure established in Chapter 3 as a mechanism for the morphological assessment of the skull for the purposes of phylogenetic evaluation is favorable.

8.5 Ornamentation Development

The establishment of a descriptive overview of skull morphology also provided the necessary anatomical foundation for subsequent investigation into less discretely presenred aspects of vertebrate biology. In this instance, the development of cranial ornamentation in members of the Ankylosauria was assessed. Cranial ornamentation is cited as being the most distinct and readily diagnostic aspect of ankylosaur skull anatomy (see Chapter 6). Previous efforts at elucidating the developmental mechanism responsible for cranial ornamentation have enlisted a circularly-based approach, hypothesizing ontogenetic patterns based strictly upon the end result (i.e. the adult condition). This study inwrporated a more widely encompassing comparative approach. Cranial ornamentation was reviewed in a broad variety of extant taxa, both in terms of developmental mechanisms and morphological outcomes. Using the procedures for the inference of unpreserved attributes (summarized by Bryant and Russell, 1992; Witmer, 1995; 1997), earlier work was reappraised in the context of correlating similar patterns of morphology with particular patterns of development. The end result serves as an example of the utility of the comparative approach in the establishment of unpreserved attributes of organisms.

8.6 Systematics

The skull is a major contributor of systematic information. Careful review of cranial anatomy permits the disclosure of potentially unreported sources of phylogenetic data and a reassessrnent of previously derived cranial characters. Within the clade Ankylosauria, the reevaluation of both cranial rnorphology and osteological nomenclature permits a more extensive character list to be generated, and a more inclusive assemblage of taxa to be effectively coded (see Chapter 7). The results obtained represent the most exhaustive phylogenetic anaiysis of the Ankylosauria ever undertaken. As noted in previous work on the skull (Trueb, 1993; Zusi, 1993),the absence of a well-supported and generally accepted phylogeny constrains Our capacity to effectively evaluate evolutionary statements with regard to specific aspects of the biology of any taxon. The significance of the cranium as a source of systematic data was reviewed by comparing the results of the cranial phylogenetic hypotheses with a series of reference hypotheses. A reference hypothesis is a consensus estimate of tree topography for al1 previously published systematic statements. In this instance, a supertree analysis was employed to generate the reference hypothesis. Supertree analysis sanctions the integration of a variety of data sources (e.g. evolutionary diagrams and cladograms) with potentially variable terminal taxa, and approaches a "majority rule summary" (sensu Bininda-Emonds et al., 1999) for ail the data under consideration. The results indicate that the cranial cladograms closely approximate the reference hypotheses. The implications are two-fold, although not necessarily exclusive of one-another; (1) many previous efforts have relied heavily upon information gleaned from the skull, and (2) even in the absence of other osteological material (e-g. the postcranium), the cranium may be used to effectively cluster members of this lineage.

8.7 Prospectus

"My answer is that in morphology general principles are founded on matters of quite intricate detail and require detail for their illustration.. ." (de Beer, 1937).

Detailed, qualitative considerations of morphology form the foundation for the effective evaluation of many higher order biological inferences, including functional morphology, community structure and genealogy. Within the fossil record, osteological components frequently offer the only material through which palaeobiological inferences may be approached. However, just as comprehensive treatments offer the opportunity to enlighten, limited and cursory morphological reviews have the potential to misdirect and deceive. Furtherrnore, the effectiveness of any study is limited by its capacity to influence future efforts. Thus, the value of this study is related to its potential to compel additional comprehensive evaluations. The descriptive procedure, while effective, stands to be irnproved by the use of diagnostic imaging. This study incorporated, to a limited extent, the use of cornputer tornographic (CT) scanning as a mechanism for evaluating deeply situated structures non-invasively. Unfortunately, the effectiveness of CT scanning is related, in part, to the fossilization process. During the course of this research, four specimens of Euoplocephalusand the holotype of "Gobisaurusf' (see Chapter 2, section 2.2) were CT scanned. Despite the fact that these specimens were coliected from a variety of different palaeoenvironments (e.g. the Dinosaur Park and Horseshoe Canyon Formations, Alberta, the Ulanhushao Formation, China), al1 of the mania demonstrated a large amount of radio-opaque minerals. Consequently, in this investigation, the CT imaging was of lirnited utility. However, concurrent with these investigations, a similar CT review of a specimen from another ankylosaur taxon contemporaneous with Euoplocephalus, Edmontonia rugosidens (Gilmore 1930), was exttemely effective (pers. obs.). Supplementary surveys of specimens may reveal other candidates for CT analysis. Another important avenue for investigation is the reappraisal of material attri buted to Pinacosaurus grangeri Gilmore 1933. Pinacosaurus is the only ankylosaur known frorn multiple subadult and adult specimens. Although morphology of subadult material has received treatment (Maryanska, 1971; Maryanska, 1977), adult material is not well known. In addition, no reviews integrating a cornparison of ontogenetic stages exist. In lieu of fortuitous discoveries of subadult - adult combinations elsewhere, these specimens represent the rnost çignificant untapped resource for future ankylosaur research. Literature Cited

Arnold, E.N. 1973. Relationships of the Palearctic lizards assigned to the genus Lacerta, Algyroides and Psammodromous (Reptilia, Lacertidae). Bulletin of the British Museum (Natural History) 25: 291-366.

Bakker, R.T. 1988. Review of the Late Cretaceous nodosaurid Dinosauria: Denversaurus schlessmani, a new armot-plated dinosaur from the latest Cretaceous of South Dakota, the last survivor of the nodosaurians, with comments on stegosaur-nodosaur relationships. Hunteria l(3): 3-23.

Barrett, P.M., Hailu, Y., Upchurch, P. and Burton A.C. 1998. A new ankylosaurian dinosaur (Ornithischia: Ankylosauria) from the Upper Cretaceous of Shanxi Province, People's Republic of China. Journal of Vertebrate Paleontology l8(2): 376-384.

Baszio, S. 1997. Systematic palaeontology of isolated dinosaur teeth frorn the Latest Cretaceous of South Alberta, Canada. Courier Forschungsinstitut Senckenberg 196: 33-77.

Bauer, A.M. and A.P. Russell. 1989. Supraorbital ossifications in geckos (Reptilia: Gekkonidae). Canadian Journal of Zoology 67: 678-684.

Baurn, B.R. 1992. Combining trees as a way of combining data sets for phylogenetic inference, and the desirability of combining gene trees. Taxon 41:

Baumel, J. J. and Witmer, L.M. 1993. Osteologia. In Handbook of Avian Anatomy: Nomina Anatomica Avium, 2ndEdition. Edited by J. J. Baumel, A.S. King, J.E. Breazile, H.E. Evans, and J.C. Vanden Berge. Nuttall Ornithological Society, Cambridge, pp. 45-1 32. de Beer, G.R. 1937. The Development of the Vertebrate Skull. Oxford University Press, Oxford.

Benton, M.J. 1998. Vertebrate Palaeontology. Chapman and Hall, New York.

Bininda-Emonds, O.R.P., Gittleman, J.L. and A. Purvis. 1999. Building large trees by combining phylogenetic information: a complete phylogeny of the extant Carnivora (Mammalia). Biological Reviews 74: 143-175.

Bock, W. J. 1989. Principles of biological comparison. Acta Morphologica Neerlando-Scandinavica 27: 17-32.

Brochu, C.A. 1999. Phylogenetics, taxonomy, and historical biogeography of Alligatoroidea. Journal of Vertebrate Paleontology 19 (Supplement to number 2): 9-100.

Brown, B. 1908. The Ankylosauridae, a new family of armored dinosaurs frorn the Upper Cretaceous. Bulletin of the American Museum of Natural History 24: 187-201.

Bryant, H.N. and A.P. Russell. 1992. The role of phylogenetic analysis in the inference of unpreserved attributes of exti nct taxa. Philosophical Transactions of the Royal Society of London, B 337: 405-41 8. de Buffrenil, V. 1982. Morphogenesis of bone ornamentation in extant and extinct crocodilians. Zoomorphology 99: 155-1 66. Camp, C.L. 1923. Classification of the lizards. Bulletin of the American Museum of Natural History 48: 289-481.

Carpenter, K. 1982. Skeletal and dermal armor reconstruction of Euoplocephalus tutus (Ornithischia: Ankylosauridae) from the late Cretaceous Oldman Formation of Alberta. Canadian Journal of Earth Sciences 19: 689-697.

Carpenter, K. 1990. Ankylosaur systematics: example using Panoplosaurus and Edmontonia (Ankylosauria: Nodosauridae). In Dinosaur Systematics: Approaches and Perspectives. Edited by K. Carpenter and P.J. Currie. Cambridge University Press, Cambridge, pp. 281-298.

Carpenter, K. 1997a. Ankylosauria. In Encyclopedia of Dinosaurs. Edited by P.J. Currie, and K. ~adian.Academic Press, New York, pp. 16-20.

Carpenter, K. 1997b. Ankylosaurs. ln The Complete Dinosaur. Edited by J. Farlow and M. Brett-Surman. Indiana University Press, Indianapolis, Indiana, pp. 307-316.

Carpenter, K. and Kirkland, J.I. 1998. Review of Lower and Middle Cretaceous ankylosaurs from North America. Lower and Middle Cretaceous Terrestrial Ecosystems, New Mexico Museum of Natural History and Science Bulletin 14: 249-270.

Carpenter, K., Miles, C., and Cloward, K. 1998. Skull of a Jurassic ankylosaur (Dinosauria). Nature 393: 782-783.

Carpenier, K., Kirkland, J.I., Burge, D., and Bird J.. 1999. Ankylosaurs (Dinosauria: Ornithischia) of the Cedar Mountain Formation, Utah, and their stratigraphic distribution. Vertebrate Paleontology in Utah, Miscellaneous Publication, Utah Geological Survey 99-1: 243-251. Carpenter, K., Kirkland, J.I., Burge, D. and J. Bird. In press. Disarticulated skull of a primitive ankylosaund from the Lower Cretaceous of eastern Utah. In. The Armored Dinosaurs. Edited by K. Carpenter and J.I. Kirkland. Indiana University Press, Indianapolis.

Carroll, R.L. 1988. Vertebrate Paleontology and Evolution. W. H. Freeman and Company, New York.

Coldiron, R.W. 1974. Possible functions of ornament in labyrinthodont amphibians. Occasional Papers of the Museum of Natural History, The University of Kansas, Lawrence, Kansas 35: 1-19.

Coombs, W.P., Jr. 1971. The Ankylosauria. Ph.D. dissertation, Columbia University, New York, 487p.

Coombs, W. P., Jr. 1972. The bony eyelid of Euoplocephalus (Reptilia, Ornithischia). Journal of Paleontology 46(5): 637-650.

Coombs, W.P., Jr. 1978. The families of the ornithischian dinosaur order Ankylosauria. Palaeontology 21: 143-170.

Coombs, W.P. 1990. Teeth and taxonomy in ankylosaurs. ln Dinosaur Systematics: Approaches and Perspectives. Edited by K. Carpenter and P.J. Currie. Cambridge: Cambridge University Press, pp. 269-279.

Coombs, W.P. 1995. Ankylosaurian tail clubs of middle Campanian to early Maastrichtian age from western North America, with description of a tiny club from Alberta and discussion of tail orientation and tail club function. Canadian Journal of Earth Sciences 32: 902-912. Coombs, W.P., Jr. and Maryanska T. 1990. Ankylosauria. In The Dinosauria. Edited by D.B. Weishampel, P. Dodson, and H. Osrnolska. University of California Press, Berkeley, pp. 456-483.

Currie, P.J. 1997a. Braincase anatomy. In Encyclopedia of Dinosaurs. Edited by P.J. Currie and K. Padian. Academic Press, New York, pp. 81-85.

Currie, P.J. 1997b. Sino-Soviet Expeditions. In Encyclopedia of Dinosaurs. Edited by P.J. Currie and K. Padian. Academic Press, New York, pp. 661 -662.

Eaton, T.H., Jr. 1960. A new armored dinosaur form the Cretaceous of Kansas. University of Kansas Paleontological Contribution 25: 1-24.

Edmund, G.A. 1957. On the special foramina in the jaws of many ornithischian dinosaurs. Contributions to the Royal Ontario Museum Division of Zoology and Palaeontology 48: 3-1 4.

Estes, R., de Queiroz, K. and J. Gauthier. 1988. Phylogenetic relationships within the Squamata. In Phylogenetic Relationships of the Lizard Families: Essays Commemorating Charles L. Camp. Edited by R. Estes and G. Pregill. Stanford University Press, Stanford, California, pp. 11 9-270.

Etheridge, R. and K. de Queiroz. 1988. A phylogeny of the Iguanidae. In Phylogenetic Relationships of the Lizard Families: Essays Commemorating Charles L. Camp. Edited by R. Estes and G. Pregill. Stanford University Press, Stanford, California, pp. 283-337.

Evans, S.E. 1988. The early history and relationships of the Diapsida. In The Phylogeny and Classification of the Tetrapods, Volume 1: Amphibians, Reptiles, Birds. Systematics Association Special Volume 35A. Editied by M. J. Benton. Clarendon Press, Oxford, pp. 221-260. Fastovsky, D.E. and D.B. Weishampel. 1996. The Evolution and Extinction of the Dinosaurs. Cambridge University Press, Cambridge.

Frost, D.R. and R. Etheridge. 1989. A phylogenetic analysis and taxonomy of iguanian lizards (Reptilia: Squamata). The University of Kansas Museum of Natural History Miscellaneous Publications 81: 1-61?.

Gadow, H. 1901. Amphibia and Reptiles. The Cambridge Natural History, volume VIII. Macmillan and Co., Ltd., London.

Galton, P.M. 1986. Herbivorous adaptations of tate Triassic and Early Jurassic dinosaurs. In The Beginning of the Age of Dinosaurs. Edited by K. Padian. Cambridge: Cambridge University Press, pp. 203-221.

Galton, P.M. IWO. Stegosauria. In The Dinosauria. Edited by D. B. Weishampel, P. Dodson and H.Osmolska. Berkeley: University of California Press, pp. 435-455.

Ghiselin, M.T. 1984. "Definition", "character", and other equivocal terms. Systematic Zoology 33: 104-11 0.

Gilmore, C.W. 1923. A new species of Coryfhosaurus with notes on other Belly River Dinosauria. Canadian Field-Naturalist 37: 46-52.

Gilmore, C.W. 1930. On dinosaurian reptiles frorn the Two Medicine Formation of Montana. Proceedings of the United States National Museum 77(16): 1-39.

Gilmore, C.W. 1933. On the dinosaurian fauna of the ken Dabasu Formation. Bulletin of the Arnerican Museum of Natural History 67: 23-78. Grande, L. and W.E. Bernis. 1998. A comprehensive phylogenetic study of ami id fishes (Amiidae) based on comparative skeletal anatomy. An empi rical search for interconnected patterns of natural history. Journal of Vertebrate Paleontology 18 (Supplement to number 1): 1-690.

Haas, G. 1969. On the jaw muscles of ankylosaurs. American Museum Novitates. 2399: 1-1 1.

Hall, B.K. 1994. Introduction. In Homology : the hierarchical basis of comparative biology. Academic Press, New York, pp. 2-1 7.

Hanken, J. and B.K. Hatl. 1993. General Introduction. ln The Skull: Volume 1 Development. The University of Chicago Press, Chicago, pp. ix-xii.

Herring, S.W. 1974. A biometric study of suture fusion and skull growth in peccaries. Anatomy and Embryology 146: 167-180.

Hildebrand, M. 1988. Analysis of Vertebrate Structure, Third Edition. John Wiley and Sons, Inc., New York.

Hill, R.V. 1999. Phylogenetic relationships arnong Ankylosauria: an analysis of cranial characters. Journal of Vertebrate Paleontology 19 (supplement to No. 3): 51A.

Hopson, J.A. and L.B. Radinsky. 1980. Vertebrate paleontology: new approaches and new insights. Paleobiology 6: 250-270.

Hu, S.-Y. 1964. Carnosaurian remains from Alshan, lnner Mongolia. Vertebrata Palasiatica 8:42-63, lordansky, N.N. 1973. The skull of the Crowdilia. ln Biology of the Reptilia, Volume 4, Moprhology (D). Edited by C.Gans. Academic Press, New York, pp. 256-284.

Jacobs, L.L., Winkler, D.A., Murry, P.A. and J.M. Maurice. 1994. A nodosaurid scuteling from the Texas shore of the Western lnterior Seaway. ln Dinosaur Eggs and Babies. Edited by K. Carpenter, K.F. Hirsch, and J.R. Horner. Cambridge Unversity Press, Cambridge, pp. 337-346.

Kardong, K.V. 1998. Vertebrates: Comparative Anatomy, Function, Evolution. Wm. C. Brown Publishers, Oxford.

Kirkland, J. 1. 1998. A polacanthine ankylosaur (Ornithischia: Dinosauria) from the Early Cretaceous (Barremian) of eastern Utah. Lower and Middle Cretaceous Terrestrial Ecosystems, New Mexico Museum of Natural History and Science Bulletin I4:271-28l.

Lambe, L.M. 1902. New genera and species from the Belly River Series (mid- Cretaceous). Contributions to Canadian Palaeontoiogy; Geological Survey of Canada 3: 25-81.

Lambe, L.M. 1910. Note on the parietal crest of Centrosaurus apertus and a proposed new generic name for Stereocephalus tutus. Ottawa Naturalist 14: 149-151.

Lambe, L.M. 1919. Description of a new genus and species (Panoplosaunrs mirus) of an armoured dinosaur from the Belly River Beds of Alberta. Proceedings and Transactions of the Royal Society of Canada 13: 39-51. Lauder, G.V. 1995. On the inference of function from structure. In Functional Morphology in Vertebrate Paleontology. Edited by J. J. Thornason. Cambridge: Cambridge University Press, pp. 1- 18.

Lee, Y.-N. 1996. A new nodosaurid ankylosaur (Dinosauria: Ornithischia) from the Paw Paw Formation (Late Albian) of Texas. Journal of Vertebrate Paleontology 16(2): 232-245.

Martin, R.E., Pine, R.H. and A.F. DeBlase. 2001 . A Manual of Mammalogy: With Keys to Families of the World, Third Edition. McGraw-Hill Companies, Inc., New York.

Maryanska, T. 1971. New data on the skull of Pinacosaurus grangeri (Ankylosauria). Palaeontologica Polonica 25: 45-53.

Maryanska, T. 1977. Ankylosauridae (Dinosauria) from Mongolia. Palaeontologia Polonica 37: 85-151.

Meszoely, C. 1970. North American anguid lizards. Bulletin of the Museum of Comparative Zoology, Harvard University 139: 87-149.

Molnar, R. E. 1996. Prelirninary report of a new ankylosaur from the Early Cretaceous of Queensland, Australia. Memoirs of the Queensland Museum 39(3): 653-668.

Montanucci, R.R. 1987. A phylogenetic study of the horned lizards, genus Phrynosoma, based on skeletal and external morphology. Contributions in Science 390: 1-36.

Moss, M.L. 1969. Comparative histology of dermal sclerifications in reptiles. Acta Anatomica 73: 510-533. Moss, M.L. 1972. The vertebrate dermis and the integumentary skeleton. American Zoologist 12: 27-34.

Nomina Anatomica Veterinana, third edition. 1983. Edited by R.E. Habel, J. Frewein and W.O. Sack. International Cornmittee on Veterinary Anatomical Nomenclature, Vienna.

Nopcsa, F. 1928. Palaeontological notes on reptiles. Geologica Hungarica Series Palaeontologica 4 : 1-84.

Norman, D.B. and T. Faiers. 1996. On the first partial skull of an ankylosaurian dinosaur from the Lower Cretaceous of the Isle of Wight, southern England. Geological Magazine 133(3): 299-31 0.

Pang, Q. and Cheng, 2. 1998. A new ankylosaur of Late Cretaceous from Tianzhen, Shanxi. Progress in Natural Science 8(3): 326-334.

Parks, W.A. 1924. Dyoplosaurus acutosquameus, a new genus and species of armoured dinosaur; with notes on a skeleton of Prosaurolophus maximus. University of Toronto Studies (Geological Series) 18: 5-25.

Penkalski, P. 1998. A preliminary systematic analysis of the Ankylosauridae (Ornithischia: Thyreophora) from the Late Campanian of North America. Journal of Vertebrate Paleontology 18 (Supplement to number 3): 69A-70A.

Purvis, A. 1995a. A composite estimate of primate phylogeny. Philosophical Transactions of the Royal Society of London, B 348: 405-421.

Purvis, A. 1995b. A modification to Baum and Ragan's method for combining phylogenetic trees. Systematic Biology 44(2): 251 -255. de Queiroz, K. 1987. Phylogenetic systematics of iguanine lizards. A comparative osteological study. University of California Publications, Zoology 118: 1-203. de Queiroz, K. and J.A. Gauthier. 1990. Phylogeny as a central principle in taxonomy: Phylogenetic definitions of taxon names. Systernatic Zoology 39: 307-322. de Queiroz, K. and J.A. Gauthier. 1992. Phylogenetic taxonomy. Annual Review of Ecology and Systematics 23: 449-480.

Ragan, M.A. 1992. Phylogenetic inference based on matrix representation of trees. Molecular Phylogenetics and Evolution 1: 53-58.

Romer, A.S. 1956. Osteology of Reptiles. The University of Chicago Press, Chicago.

... Rowe, T. Definition and diagnosis in the phylogenetic system. Systematic ZOOIO~Y36: 208-21 1.

Russell, L.S., 7940. Edmontonia rugosidens (Gilmore) an armored dinosaur from the Belly River Series of Alberta. University of Toronto Studies, Geology Series . . 43: 1-28.

Russell, A.P. and J. J. Thornason. 1993. Mechanical analysis of the mammalian head skeleton. In. The Skull: Volume 3 Functional and Evolutionary Mechanisms. The University of Chicago Press, Chicago, pp. 345-383.

Rybczynski, N. and M.K. Vickaryous. In press. Evidence of complex jaw movement in the Late Cretaceous ankylosaurid Euoplocephiaus tutus (Dinosauria: Thyreophora). In. The Armored Dinosaurs. Edited by K. Carpenter and J. 1. Kirkland. Indiana University Press, Indianapolis.

Sanderson, M.J., Purvis, A. and C. Henze. 1998. Phylogenetic supertrees: assembling the trees of life. Trends in Ecology and Evolution 13: 105-109.

Sampson, S.D. and Witmer, L.M. 1999. Novel narial anatomy in ceratopsid dinosaurs. Journal of Vertebrate Paleontology 19(supplement to Number 3): 72A-73A.

Sereno, P.C. 1986. Phylogeny of the bird-hipped dinosaurs (Order Ornithischia). National Geographic Research 2(2):234-256.

Sereno, P.C. 1991. Lesothosaurus, "fabrosaurids", and the early evolution of Ornithischia. Journal of Vertebrate Paleontology II: 168-197.

Sereno, P.C. 1997. The origin and evolution of dinosaurs. Annual Review of Earth and Planetary Sciences 25: 435-489.

Sereno, P.C. 1998. A rationale for phylogenetic definitions, with application to the higher-level taxonomy of Dinosauria. Neues Jahrbuch fur Geologie und Palaeontologie 21 0:41-83.

Sereno, P.C. 1999. The evolution of dinosaurs. Science 284: 21 37-2147.

Sereno, P.C.and D. Zhimin. 1992. The skull of the basal stegosaur Huayangosaurus taibaii and a cladistic analysis of the Stegosauria. Journal of Vertebrate Paleontology 12(3): 31 8-343. Starck, D. 1979. Cranio-cerebral relations in recent reptiles. In Biology of the Reptilia, Volume 9, Neurology A. Edited by C.Gans. Academic Press, New York, pp. 1-36.

Sternberg, C.M. 1929. A toothless armoured dinosaur from the Upper Cretaceous of Alberta. National Museum of Canada Bulletin 54: 28-33.

Sullivan, R. M. 1999. Nodocephalosaurus kirflandensis, gen. et sp. nov., a new ankylosaurid dinosaur (Ornithischia: Ankylosauria) from the Upper Cretaceous Kirtland Formation (Upper Campanian), San Juan Basin, New Mexico. Journal of Vertebrate Paleontology lg(1): 126-1 39.

Swofford, D. L. 1993. PAUP: Phylogenetic Analysis Using Parsimony, version 3.1.1. Computer software and documentation distributed by Illinois Natural History Survey, Champaign, Illinois.

Trueb, L. 1966. Morphology and development of the skull in the frog Hyla septentrionalis. Copeia 1966: 563-573.

Trueb, L. 1993. Patterns of cranial diversity among Iissamphibians. In The Skull: Volume 2 Patterns of Structural and Systematic Diversity. The University of Chicago Press, Chicago, pp. 255-343.

Tumanova, T.A. 1977. (New data on the ankylosaur Tarchia gigantean). Paleontological Zhurnal1977: 92-100. (In Russian.)

Tumanova, T.A. 1983. (The first ankylosaur from the Lower Cretaceous of Mongo li a). So vmestnaya So ve fsko-Mongol'skaya Paleon fogichekaya Ekpeditsiya, Trudy 24: 11 0-120. (In Russian with English summary.) Tumanova, T.A. 1985. Skull morphology of the ankylosaur Shamosaurus scutatus from the Lower Cretaceous of Mongolia. In Les Dinosaures de la Chine a la France. Edited by P. Taquet and C. Sudre. Museum d'Histoire Naturelle de Toulouse, France and Museum d'Histoire Naturelle, Chongqing, People's Republic of China, pp. 73-79.

Tumanova, T.A. 1987. (The armored dinosaurs of Mongolia). Sovmestnaya Sovetsko-Mongol'skaya Paleontogichekaya Ekpeditsiya, Trudy 32: 1-80. (In Russian with English summary.)

Tumanova, T.A. 1993. A new armored dinosaur from south-eastern Gobi. Paleontological Journal 27(2):1 19-1 25.

Vickaryous, M. K., Russell, A.P., and P.J. Currie. In press. The cranial ornamentation of ankylosaurs (Ornithischia: Thyreophora): reappraisal of developmental hypotheses. ln. The Armored Dinosaurs. Edited by K. Carpenter and J. 1. Kirkland. Indiana University Press, Indianapolis.

Vickaryous, M. K., Russell, A. P. and P.J. Currie. Submitted (December, 2000). A new ankylosaurid (Dinosauria: Ankylosauria) from the Late Cretaceous of China, with comments on ankylosaurian relationships and osteological nomenclature. Canadian Journal of Earth Sciences.

Vickaryous, M. K. and N. Rybczynski. In press. Omamentation. ln Dinosaur Biology. Edited by P.J. Currie and K. Padian. Academic Press, New York.

Weisharnpel, D.B. 1995. Fossils, function, and phylogeny. ln Functional Morphology in Vertebrate Paleontology. Edited by J. J. Thomason. Cambridge: Cambridge University Press, pp. 34-54. Weishampel, D.B. and D.B. Norman. 1989. Vertebrate herbivory in the Mesozoic; Jaws,plants, and evolutionary metrics. Geological society of Arnerica. Special Paper 238: 87-1 00.

Weishampel, D. B. and L.M. Witmer. 1990. Lesothosaurus, Pisanosaurus, and Technosaurus. In The Dinosauria. Edited by D.B. Weishampel, P. Dodson and H. Osmolska. Berkeley: University of California Press, pp. 416-425.

Weishampel, D. B., Grigorescu and D. B. Norman. 1991. The dinosaurs of Transylvania. National Geographic Research and Exploration 7(2):1 96-21 5.

Westphal, F. 1976. The dermal armour of some Triassic placodont reptiles. In Morphology and Biology of Reptiles, Linnean Symposium Series No. 3. Edited by A. d'A. Bellairs and C.B. Cox. Academic Press, New York, pp. 31 -41.

Witmer, L.M. 1994. A revised terminology of the facial osteology of archosauriforms. Journal of Vertebrate Paleontology ?4(supplement to number 3): 53A.

Witmer, L.M. 1995. The extant phylogenetic bracket and the importance of reconstructing soft tissues in fossils. In Functional Morphology in Vertebrate Paleontology. Edited by J. J. Thornason. Cambridge: Cambridge University Press, pp. 19-33.

Witmer, L.M. 1997. The evolution of the antorbital cavity in archosaurs: a study in soft-tissue reconstruction in the fossil record with an analysis of the function of pneumaticity. Journal of Vertebrate Paleontology 17(supplementto Number 1), Memoir 3. Witmer, L.M. and Sampson, S.D. 1999. Nasal conchae and blood supply in some dinosaurs: physiological implications. Journal of Vertebrate Paleontology 19(supplement to Number 3): 85A.

Wu, X.-C., Brinkman, D.B., and Russell, A.P. 1996. A new alligator from the Upper Cretaceous of Canada and the realtionships of early eusuchians. Palaeontology 39(2): 351-375.

Zusi, R.L. 1993. Patterns of diversity on the avian skull. In The Skull: Volume 2 Patterns of Structural and Systematic Diversity. The University of Chicago Press, Chicago, pp. 391 -437.

Zylberberg, L. and J. Castanet. 1985. New data on the structure and the growth of osteoderms in the reptile Anguis fragilis L. (Anguidae, Squamata). Journal of Morphology 186: 327-342. Appendix 2.1. Nomenclatural abbreviations

ang os angulare angular bone apt nas *apertum nasalis nasal aperture apt paranas *apettura paranasalis paranasal aperture apt pal "apertura palatinus palatal aperture arc jug amsjugalis jugal arch art os articulare articular basi basis cranii basicranium bo os basioccipifale basioccipital bone bs os basisp henoidale basisphenoid bone cav nas cavum nasi nasal cavity proper cd man qu condylus mandibularis quadrati mandibular condyle of the quadrate cd occ condylus occipitalis occipital condyle ch choana intemal naris "cil"sorb supraorbitaliâ accessona "cilary" supraorbital cor qu corpus quadmti body of the quadrate cor pt corpus ptetygoideum body of the pterygoid cor0 os comnoideum coronoid bone cot qu sq soty/a quadratica squamosi quadrate articulation surface of the squamosal cr nuc sag srista nuchalis sagittalis sagittal nuchal ridge cr nuc trans zrista nuchalis transversa transverse nuchal crest cr tm mx ?rista tomialis maxillare maxillary torniat crest cr tm pmx xista fomialis pmmaxillare premaxillary tomial crest d mx 3'entes maxillare maxillary teeth Appendix 2.1. Continued

dent os dentale dentary bone ecpt os ectopterygoideum ectopterygoid bone exo os exoccipitale exoccipital bone fen lateroternp fenestra laterotemporalis laterotemporal fenestra fen oval fenestra vestibuli oval window fen posttemp fenestra posttempomlis postternporal fenestra fen suborb fenestra suborbitalis suborbital fenestra for inci foramen incisivium incisive foramen for inteman caud foramen intermandibularis caudalis caudal intermandibular foramen for mag fommen magnum forarnen magnum for n il foramen n opticus optic nerve forarnen for n III fommen n ocuiornotons oculomotor nerve foramen for n IV foramen n tmchlearis trochlear nerve forarnen for n V fomrnen n tdgeminus trigeminal nerve foramen for n VI1 foramen n facialis facial nerve foramen for n IX foramen n glossophatyngeus glossopharyngeal nerve foramen for n X foramen n vagus vagus nerve forarnen for XI foramen n accessorius accessort nerve forarnen for pal a fommen palatinus a foramen for the palatine artery fos add fossa adductoris mandibulae mandibular adductor fossa fov cor ner foveae corpusculorum nervosomm sensory pits fr os frontale frontal bone Appendix 2.1. Continued.

median condylar notch in prnx *incisura premaxillare premaxillary notch interorb ossa interorbitalis interorbital bones os jugale jugal bone lac os lacdmale lacrimal bone lam postoc *lamina postocularis postocular shelf latero os latemsphenoidale laterosphenoid bone man ossa mandibulae mandibular bones margo nuc *marge nuchae nuchal shelf mx os rnaxillam maxillary bone nas os nasale nasal bone opis os opisfho ficum opisthotic bone orb orbifa orbit otic ossa ofica otic elements pal os palatinum palatine bone Par os parietale parietal bone pars symph pars symphysialis dentale mandibular symphysis para os parasphenoidale parasphenoid bone pila internas *pila intemasalis internasal bar Pmx os premaxillarir prernaxillary bone PO os postohitale postorbital bone prf os prefmntale prefrontat bone Appendix 2.1. Continued.

- - pro alv processus alveola ris alveolar border pro bpt processus basipteryg oideus basipterygoid process pra corn qj *processus cornuum quadratojugalis quadratojugal projection pro corn sorb *processus comuum supraorbitalis supraorbital boss pro corn sq *processus comuum squamosi squamosal boss pro cor0 processus coronoideus coronoid process pro ex0 processus paroccipitalis paroccipital process pro man pt pmcessus mandibularis pterygoideum mandibular rarnus of the pterygoid pro man qu processus mandibulafis quadratum mandibular process of the quadrate Pro qu pt processus quadmticus pterygoideum quadrate process of the pterygoid Pro qu processus quadraticus quadratojugalis quadrate process of the quadratojuga! pro retro processus retroarticula ris retroarticular process pro vo *pmcessus vomera rostral process of the vomers pro0 os pmoticum prootic bone pt os pterygoideus pterygoid bone os quadmtojugale quadratojugal bone os quadratum quadrate ros man rostrum mandibulae mandibular rostrurn ros pmx mstium premaxillare premaxillary rostrum s mx sinus maxillans rnaxillary sinus cavity SOCC supraocci pitalis supraoccipital bone sorb ossa supraorbitalis supraorbital bones sp cav internas *septum ca vitas internasalis internasal cavity septum sp n osrn septum nasi osseum nasal septum sple os spleniale splenial bone

Appendix 4.1. Skull material referred to the taxon Euoplocephalus fufus presently accessioned in rnuseum collections. Institutional abbreviations listed in Chapter 1. Occurrence refers to the stratigraphic horizon material was collected from. Occurrence abbreviations: DP Fm., Dinosaur Park Formation (late Campanian, Alberta); HSC Fm., Horseshoe Canyon Formation (early Maastrichtian, Alberta); JR Gp., Judith River Group (late Campanian, Alberta); TM Fm., Two Medicine Formation (early Maastrichtian, Montana). Use of the more inclusive Judith River Group stratigraphic package was applied to specimens whose exact providence within Dinosaur Provincial Park, Alberta remains unclear. See Literature Cited for a complete reference citation. Asterisk indicates type specimen.

NMC 0210 * cranium JR Gp. Lambe, 1902; 1910

cranium HSC Fm. Coombs, 1978 cranium, "ciliary" supraorbitals ? Coombs, 1972 cranium, complete mandible, DP Fm. Haas, 1969 endocast cranium, thin sections DP Fm. Coombs, 1978 cranium, "ciliary" supraorbitals DP Fm. Coombs, 1972; 5978

cranium TM Fm. none

none TMP 85.36.330 right mandible JR Gp. none TMP 91.127.1 cranium DP Fm. none cranium DP Fm. none cranium, right mandi ble HSC Fm. none Appendix 4.1. Continued.

TMP 97.59.1 cranium HSC Fm. none cranium, right mandible, "ciiliary" TMP 97.132.1 DP Fm. none supraorbitals TMP 98.83.1 cranium DP Fm. none UALVP 31 cranium, left mandible JR Gp. Gilrnore, 1923

NMC 8530 * cranium, right rnandible HSC Fm. Sternberg, 1929 Appendix 6.1. Extant material examined, including information on specimen type. Abbreviations: C&S, cleared and doubled-stained for bone and cartilage; SD, dried skeletal material; FF, fresh-frozen; and PR, alcohol- preserved whole specimens. All specimens assigned to one of three different ontogenetic stages: neo, nemate 1 embryo; subad, subadult; ad, adult. Asterisk (') denotes unnumbered specimen. See Prefatory pages for collection acronyms.

Cotytophanes cristatus 1; ad PBPC*

ARPC* UCMZ(W1978-11 UCMZ/R/I 975-67

UCMZIWI986-21 PBPC*

Phrynosoma comutum 1; neo LACM l9897(lZl85) LACM 4307 UCMZ/R/1979-6; 7 ARPC* ARPC* LACM 1Q6692(5536); l9694(5538) TMP 90.7.162 2; neo LACM 123343; 123344

Chamaeieo calyptratus 2; neo ARPC* TMP 90.7.350 PBPC* ARPC* Amendix 6.1. Continued

ARPC* ARPC* UCMZIW1975-1O6 UCMZIRII975-98

ARPC* ARPC IO ARPC* ARPC*

UCMZlWl975-115; 116 UCMZWl975-119; 120 PBPC" PBPC*

UCMZ/Wl975-137 ARPC* UCMZIW2000.001 TMP 90.7.26 UCMZIWI 993-4 UCMZ/R/1976-32 Appendix 7.1. Character state distribution among the 24 taxa considered in this study (including two outgroups). Characters are listed numencally to correspond with those presented in the text. Primitive states scored with a "O", missing data with a "?",and al1 others indicate the denved condition. The symbol "%" refers to the percentage of character states scored for each taxon. Taxa not under consideration demarcated with redlining. Appendix 7.2. Character state distribution for the supertree analysis of 14 previously published systematic estimtaes and three cladograms generated from the evaluation of strictly cranial data (figures 7.1 - 7.3). Appendix 7.2. Continued