sign in sign up
<<
Home , Gastonia (dinosaur), Hesperosaurus, Hoplitosaurus, Scutellosaurus, Tail club, Talarurus

University of Alberta

Evolution, biomechanics, and function of the of ankylosaurid (: )

by

Victoria Megan Arbour ©

A thesis submitted to the Faculty of Graduate Studies and Research in partial fulfillment of the requirements for the degree of

Master of Science in Systematics and Evolution

Department of Biological Sciences

Edmonton, Alberta Spring 2009 Library and Archives Bibliotheque et 1*1 Canada Archives Canada Published Heritage Direction du Branch Patrimoine de Pedition 395 Wellington Street 395, rue Wellington OttawaONK1A0N4 Ottawa ON K1A 0N4 Canada Canada

Your We Votne r&temnce ISBN: 978-0-494-54648-2 Our file Notre r6f6rence ISBN: 978-0-494-54648-2

NOTICE: AVIS:

The author has granted a non­ L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduce, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lnternet, preter, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distribute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non­ support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantiels de celle-ci substantial extracts from it may be ne doivent §tre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis.

1+1 Canada This thesis is dedicated to all of my mentors, past and present, who have helped me get to where I am today. Abstract

Modified distal caudal vertebrae (the handle) and large terminal (the knob) form the tail club of ankylosaurid dinosaurs. It has been assumed that the club was used as a weapon, but the biological feasibility of this behaviour has not been tested. Computed tomographic scans of tail clubs referred to , and measurements of free caudal vertebrae, were used to estimate the impact force of clubs of various sizes. Tails are modeled as segments for which mass, muscle cross-sectional area, torque, and angular acceleration are calculated. Large knobs could generate sufficient force to break bone during impacts, but average and small knobs could not. Finite element analyses showed that stress is distributed evenly along the handle, preventing fracture of the club during impacts. Tail clubbing behaviour is feasible in ankylosaurids, but it remains unknown whether the tail was used for interspecific defense, intraspecific combat, or both. Acknowledgements

I extend many thanks to my supervisor, Philip Currie, for the opportunity to conduct this research and for his guidance throughout the course of this project. I am also grateful to Eva Koppelhus and my committee members Alison Murray and Michael Caldwell for their advice and assistance. The students of the

Laboratory for Vertebrate have provided endless academic support and friendship and I thank them all.

Thanks to the many people who provided access and assistance at their institutions: C. Mehling (AMNH), K. Shepherd and M. Feuerstack (CMN), L. Ivy and K. Carpenter (DMNH), B. Demchig (MPC), D. Evans and B. Iwama (ROM),

J. Gardner and B. Strilisky (TMP), and S. Harpham and P. Correia (Univ. of

Alberta Anthropology). I thank B. Demchig, P. Currie, E. Koppelhus, M. Ryan and the staff of Nomadic Expeditions for providing me with the opportunity to visit collections and conduct fieldwork in .

G. Pinto and P. Bell helped prepare UALVP 47273 and I thank M. James for assistance with preparation of UALVP 31. The University of Oslo graciously provided a cast of ZPAL MgD-l/113. P. Currie, E. Snively, and M. Burns provided specimen photographs.

Many thanks to R. Lambert for allowing access to the ABACUS CT scanning facility at the University of Alberta Hospital, and to G. Schaffler and the

ABACUS technicians for help during CT scanning of TMP 1983.36.120, UALVP

47273, and UALVP 16247. ROM 788 was scanned at CML Healthcare in

Mississauga, Ontario, and I thank T. Ladd for his assistance organizing the scan. Many, many thanks to D. Evans and B. Iwama for preparing ROM 788 for CT scanning in the midst of a new gallery opening! M. James, M. Burns, and E.

Snively assisted with transporting specimens for CT scanning. J. Li and M.

Lawrenchuck (Materialise) provided technical assistance with Mimics, and Anne

Delvaux (Beaufort Analysis, Inc.) provided technical assistance with Strand7. H.

Mallison (Museum fur Naturkunde, Berlin) provided advice on digital imaging of .

I have been fortunate to have had funding from both the Natural Sciences and Engineering Research Council and Alberta Ingenuity. The University of

Alberta Graduate Students Association, the Department of Biological Sciences, and the Research Institute have also provided funds for travel and CT scanning, and their assistance is gratefully acknowledged.

Finally, I cannot ever express enough thanks to my parents, Edith and

Joseph Arbour, to my sister Jessica Arbour, and to Peter Maguire, for their kindness, support, and faith in me. Thank you all. Table of Contents

Abstract Acknowledgements List of Tables List of Figures Institutional Abbreviations Chapter 1: Introduction and background 1 1.1 Introduction 1 1.2 Phylogeny of the 3 Chapter 2: Morphology of ankylosaurid pelves and tails and the evolution of ankylosaurid tail clubs 7 2.1 Introduction 7 2.2 Materials and methods 8 2.3 Pelvic morphology of Euoplocephalus 9 2.4 Caudal morphology 13 2.4.1 Nodosaurids 13 2.4.2 Basal ankylosaurids 14 2.4.3 Derived ankylosaurids 17 2.5 Discussion 35 2.5.1 Number of represented by material referred to Euoplocephalus 35 2.5.2 Causes of tail club variation 43 2.5.3 Evolution of the ankylosaurid tail club 44 2.6 Conclusions 48 Chapter 3: Ankylosaurid tail and pathologies 50 3.1 Introduction 50 3.2 Description 51 3.2.1 AMNH 5409, pelvis and .... 51 3.2.2 AMNH 5337, pelvis and sacrum 52 3.2.3 AMNH 5245, two anterior free caudals 53 3.2.4 TMP2005.09.75, sacrocaudal 55 3.2.5 TMP 1985.36.70, proximal free caudal 55 3.2.6 AMNH 2062, anterior free caudal vertebra 56 3.2.7 ROM 1930, anterior free caudal vertebra and last free caudal vertebra 56 3.2.8 TMP 1992.36.344, posterior free caudal vertebra 58 3.2.9 AMNH 5404, two free caudal vertebrae 58 3.2.10 ROM 1930, haemal arch 60 3.2.11 TMP 1983.36.120, tail club 60 3.2.12 ROM 788, tail club 62 3.3 DISCUSSION 63 3.3.1 Diagnoses 64 3.3.1.1 AMNH 5245, AMNH 2062, ROM 1930 haemal arch and spine, TMP 1985.36.70, and TMP 2005.09.75: osteomyelitis and periostitis 65 3.3.1.2 ROM 1930, proximal free caudal vertebra and last free caudal vertebra 70 3.3.1.3 AMNH 5404, AMNH 5337, and AMNH 5409: metastatic cancer...70 3.3.1.4 ROM 788, TMP 1983.36.120, and the challenges of identifying pathologies in osteoderms ...73 3.3.2 Frequency and patterns of pathologies in ankylosaurid taxa 76 3.3.3 Behavioural implications 79 3.4 Conclusions 82 Chapter 4: Ankylosaurid caudal musculature 84 4.1 Introduction 84 4.2 Review of epaxial and intrinsic trunk and tail muscles of extant reptiles 85 4.3 Ossified tendons of ornithischian dinosaurs 88 4.4 Inferred tail musculature of ankylosaurids 89 Chapter 5: Analysis of tail swinging ability in ankylosaurid dinosaurs 97 5.1 Introduction 97 5.2 Materials and methods 97 5.3 Description of club internal morphology from CT scans 98 5.4 Analysis of tail club motion and impact force 104 5.4.1 Analysis of a small knob and tail, ROM 784/UALVP 47273 108 5.4.1.1 Determining the volume of the tail bones and muscles 110 5.4.1.2 Determining knob volume 117 5.4.1.3 Angle of articulation between free caudal vertebrae 119 5.4.1.4 Calculating T, I, and co 122 5.4.1.5 Sensitivity analyses 129 5.4.2 Analysis of a large tail and knob, ROM 788/AMNH 5245 136 5.4.2.1 Estimates of bone and muscle mass and volume 136 5.4.2.2 Calculating T, I, and co 141 5.4.3 Analysis of a mid-sized tail and knob, UALVP 16247 141 5.4.3.1 Estimates of bone and muscle mass and volume 142 5.4.3.2 Calculating T, I, and w 143 5.5 Discussion 144 5.6 Conclusions 146 Chapter 6: Finite element analyses of ankylosaurid tail clubs 148 6.1 Introduction 148 6.2 Materials and Methods 149 6.2.1 Analysis 1 154 6.2.2 Analysis 2 154 6.2.3 Analysis 3 155 6.2.4 Analysis 4 155 6.2.5 Analysis 5 157 6.2.6 Analysis 6 158 6.3 Results 159 6.3.1 Analysis 1: Effect of knob size and impact force 159 6.3.2 Analysis 2: Impact site analysis 165 6.3.3 Analysis 3: Stress distributions in the handle vertebrae 165 6.3.4 Analysis 4: Postural role of the haemal arches 168 6.3.5 Analysis 5: Material properties 169 6.3.6 Analysis 6: Neural spine shape in different ankylosaurid taxa 169 6.4 Discussion 172 6.5 Conclusions 176 7.0 Conclusions 177 8.0 Literature Cited 182 Appendix 1 211 List of Tables

Chapter 1: Table 1.1: Stratigraphic information for ankylosaurs discussed in this thesis.

Chapter 2: Table 2.1: Neural spine angles. Table 2.2: Knob widths arranged by size. Table 2.3: Tail club distribution within Ankylosauridae.

Chapter 3: Table 3.1: Possible causes of blastic and lytic lesions. Table 3.2: Differentiating lesions caused by blastomycosis, tuberculosis, and cancer. Table 3.3: Summary of diagnoses.

Chapter 5: Table 5.1: Actual and ideal values for dimensions of the centra in ROM 784, in mm. Table 5.2: Actual and ideal values for dimensions of the tail in ROM 784, in mm. Table 5.3: Comparison of the handle vertebra neural spine length. Table 5.4: Area of zygapophyseal overlap. Table 5.5: Summary of volumes, areas, and masses for ROM 784/UALVP 47273. Table 5.6: Rotational inertias for ROM 784/UALVP 47273. Table 5.7: Torques for ROM 784/UALVP 47273. Table 5.8: Segment angular rates of movement. Table 5.9: Summary of results of sensitivity analyses for ROM 784/UALVP 47273. Table 5.10: Actual and ideal values of the centra in AMNH 5245. Table 5.11: Actual and ideal values for dimensions of the tail in AMNH 5245, in mm. Table 5.12: Comparison of the handle vertebra neural spine length in AMNH 5245 and ROM 788. Table 5.13: Velocities and forces for ROM 788/AMNH 5245. Table 5.14: Cross-sectional area of UALVP 16247 handle and comparisons with ROM 788 and UALVP 47273. Table 5.15: Velocities and forces for UALVP 16247.

Chapter 6: Table 6.1: Material Properties used in finite element analyses. Table 6.2: Summary of forces used in finite element analyses. Table 6.3: Weight and torque. Table 6.4: Maximum and minimum stress values and locations. Table 6.5: Peak stresses in Analysis 6. List of Figures

Chapter 1: Figure 1.1: Phylogeny of the Ankylosauria.

Chapter 2: Figure 2.1: Diagram of tail terminology. Figure 2.2: Ankylosaurid pelves. Figure 2.3: CMN 8521, longiceps (), anterior caudal vertebra. Figure 2.4: Basal ankylosaurid caudal vertebrae. Figure 2.5: Skeletal and dermal reconstruction of preserved elements of ROM 784, Dyoplosaurus (holotype). Figure 2.6: Ankylosaurid free caudal and handle vertebrae. Figure 2.7: ROM 1930, Euoplocephalus, articulated free caudals. Figure 2.8: Variations in ankylosaurid free caudal vertebrae. Figure 2.9: Unfused centra in the handle. Figure 2.10: Variation in ankylosaurid handle vertebrae. Figure 2.11: Ankylosaurid knobs in dorsal view. Figure 2.12: Ankylosaurid knobs in anterior and posterior view. Figure 2.13: Graph showing the relationships between various vertebral dimensions and knob size in Euoplocephalus. Figure 2.14: Specimens that can be confidently referred to Euoplocephalus tutus, based on first cervical half ring morphology. Figure 2.15: Dermal armour arrangement in the tails of basal and derived ankylosaurids.

Chapter 3: Figure 3.1: Lytic lesions on the pelvis. Figure 3.2: Rugose blastic lesions of the neural spine and transverse processes. Figure 3.3: Exostoses. Figure 3.4: Metastatic cancer in AMNH 5404. Figure 3.5: ROM 1930 haemal arches. Figure 3.6: Pathologies of the tail club knobs. Figure 3.7: Examples of blastic and lytic lesions in extant vertebrates. Figure 3.8: Distal portion of the tail of AMNH 5403, .

Chapter 4: Figure 4.1: Cross-sectional reconstructions of ankylosaurid caudal musculature. Figure 4.2: Origins of tail muscles on the pelvis. Figure 4.3: Ossified tendons in ROM 784, Dyoplosaurus. Figure 4.4: ROM 784 tail club in dorsal view, showing maximum width of the handle muscles Chapter 5: Figure 5.1: CT scanning UALVP 47273 at the University of Alberta ABACUS facility. Figure 5.2: UALVP 47273 transverse views of handle Figure 5.3: UALVP 47273 sagittal views. Figure 5.4: UALVP 16247, coronal plane. Figure 5.5: Transverse sections through knobs. Figure 5.6: Diagram showing the approximate right lateroflexion of the tail in Euoplocephalus, and showing the definition of the half angle of articulation 0. Figure 5.7: Diagrammatic representation of composite tails used in this study. Figure 5.8: Graph comparing the length of the neural spine of the handle vertebrae in ROM 784 and UALVP 47273. Figure 5.9: ROM 784 knob in dorsal view, showing the maximum distal width of handle muscles. Figure 5.10: Partial of a Siberian ibex (Capra sibirica). Figure 5.11: Determining the maximum angle of rotation in ankylosaurid free caudal vertebrae.

Chapter 6: Figure 6.1: Three-dimensional models of specimens used in this study. Figure 6.2: Impact stresses in UALVP 47273. Figure 6.3: Impact stresses in TMP 1983.36.120 and UALVP 16247 Figure 6.4: Impact stresses in ROM 788 Figure 6.5: Results from analyses 4 and 5. Figure 6.6: Effects of differing material properties. Figure 6.7: Two dimensional FEA of neural arches in dorsal view. Institutional Abbreviations

AMNH - American Museum of Natural History, New York, New York

BMNH - Department of Palaeontology, The Natural History Museum, London

CCM - Carter County Museum, Ekalaka, Montana

CMN - Canadian Museum of Nature, Ottawa, Ontario

DMNH - Denver Museum of Nature and Science, Denver, Colorado

MACN PV - Vertebrate Palaeontological Collection of the Museo Argentino de Ciencias Naturales "Bernardino Rivadavia", Buenos Aires Province, Argentina

MPC - Geological Institute, Section of Palaeontology and Stratigraphy, Academy of Sciences of the Mongolian People's Republic, Ulaanbataar, Mongolia (KID: Korea Mongolia International Dinosaur Project Collection; PJC: Nomadic Expeditions Collection)

PIN - Palaeontological Institute, Russian Academy of Sciences, Moscow,

ROM - , Toronto, Ontario

TMP - Royal Tyrrell Museum of Palaeontology, Drumheller, Alberta

UALVP - University of Alberta Laboratory for Vertebrate Paleontology, Edmonton, Alberta

ZPAL - Zoological Institute of Paleobiology, Polish Academy of Sciences, Warsaw, Poland

Other Abbreviations:

CT - Computed Tomography

FEA - Finite Element Analysis Chapter 1: Introduction and background

1.1 Introduction

The interpretation of behaviours in extinct is made difficult by the lack of anatomical information and modern analogues. As such, behaviours may be inferred without substantial supporting evidence. Advances in comparative biomechanics, such as computer simulations of biomechanical models, coupled with traditional studies of morphology and anatomy, provide additional evidence to test the biological feasibility of possible behaviours in extinct animals.

Ankylosaurid dinosaurs were massive, quadrupedal, herbivorous ornithischian dinosaurs with extensive dermal ossifications on the head, neck, back, and tail (Vickaryous et al. 2004). Parks (1926) described the first ankylosaurid tail club (ROM 784, Dyoplosaurus acutosquameus), but did not comment on its potential function. Maleev (1952) interpreted the tail club of

Talarurus as the 'striking end' of the tail, and referred to it as a mace. He later described a tail club of Syrmosaurus (=Pinacosaurus granger!) as a double- edged axe, and suggested that the robust neural and haemal arches and the presence of large strands of ossified tendons indicated that strong muscles would have been employed in tail-swinging (Maleev 1954). Coombs (1971, 1979,

1995) discussed possible muscles associated with swinging the tail and tail club, and the range of motion possible in the tail.

Ankylosaurids were certainly capable of swinging the tail laterally, and the large knob and stiffened handle vertebrae suggest reinforcement against strong

1 impacts. However, little work has been done to determine the feasibility of such a behaviour. The aim of this study is to determine whether the tail clubs of ankylosaurid dinosaurs could have been used as defensive weapons. The hypothesis that ankylosaurids used their tails as weapons can be tested by asking several questions. First, what would be the maximum force and stress that an ankylosaurid could deliver with its tail club knob? Would the force be sufficient to damage muscles or bones of predators or intraspecific opponents?

How were stress and strain resulting from strong impacts dissipated throughout the club? If the vertebrae or knob osteoderms fractured under normal impact forces, this would suggest that the primary purpose of the knob was not for delivering forceful blows.

Comparisons of the morphologies of the tails and pelves of different taxa

(Chapter 2) provides insight on the evolution of tail clubs within the

Ankylosauridae, and indicates that many genera can be differentiated based on

pelvic and caudal morphology. Derived ankylosaurids appear to have been prone to injuries and diseases of the caudal vertebrae and tail clubs, and these are discussed in Chapter 3. Muscle reconstructions (Chapter 4) using extant alligators as a guide suggest that M. longissimus caudae was large, and that M.

caudofemoralis longus and M. iliocaudalis were important for moving the tail

laterally. Computed tomography (CT) scans of several tail clubs (Chapter 5)

provide information on the internal structure of tail clubs, and provide data used for estimates of tail club mass, rotational inertia, and impact force. Finally, finite

element analyses (FEA) of 3-dimensional models derived from CT scans

2 (Chapter 6) are used to understand the distribution of stress and strain in swinging tail clubs.

1.2 Phylogeny of the Ankylosauria

Vickaryous et al. (2004) conducted the most recent and inclusive phylogenetic analysis of the Ankylosauria (Fig. 1.1). Ankylosaurs include all eurypods closer to magniventris Brown 1908 than to armatus Marsh 1877 (Sereno 1998). Ankylosaurs have highly modified with prominent ornamentation and lack fenestrae except the orbit and nares, and have an imperforate acetabulum, a sacrum composed of coossified dorsal, sacral, and caudal vertebrae, and multiple rows of osteoderms on the body. The analysis by Vickaryous et al. (2004) found that all ankylosaurs fall into one of two : the Ankylosauridae, closer to Ankylosaurus than mirus

Lambe, 1919 (Sereno 1998), or , closer to Panoplosaurus than

Ankylosaurus (Sereno 1998). Ankylosaurids are all found in or North

America, except for Minmiparavertebrata Molnar, 1980, which is found in

Australia. Ankylosaurids have broad, triangular skulls, and derived members each have a tail club. Gargoyleosaurus parkpinorum Carpenter et al., 1998 is the sister taxon to all other ankylosaurids, and burgei Kirkland, 1998 and

Minmi are basal members of this group. Gobisaurus domoculus Vickaryous,

Russell, Currie et Zhao, 2001 and Shamosaurus scutatus Tumanova, 1983 from

Asia are the outgroup to the Ankylosaurinae, which includes Tarchia gigantea

Maryanska, 1977 and Tsagantegia longicranialis Tumanova, 1993 before

3 splitting into a North American (Ankylosaurus + Euoplocephalus tutus

Lambe, 1910) and an Asian clade (Pinacosaurus grangeri Gilmore, 1933,

Pinacosaurus mephistocephalus Godefroit, Pereda-Suberbiola, Li et Dong, 1999,

Saichania chulsanensis Maryanska, 1977, plicatospineus Maleev,

1952, and Tianzhenosaurus youngi Pang et Cheng, 1998). Nodocephalosaurus kirtlandensis Sullivan, 1999 is tentatively considered an ankylosaurine; it is found in North America but has cranial sculpturing similar to and Tarchia.

Carpenter (2001) placed Cedarpelta bilbeyhallorum Carpenter, Kirkland, Burge et

Bird, 2001 within the Ankylosauridae, but Vickaryous et al. (2004) recovered

Cedarpelta as a nodosaurid.

The analysis by Vickaryous et al. (2004) relies heavily on cranial characters, and so excludes several taxa with poorly preserved or missing crania. Carpenter (2001) included more taxa than Vickaryous et al. (2004), and assumed a priori that the Polacanthidae is a valid taxon. Carpenter (2001) assigned Gastonia burgei, Gargoyleosaurus parkpinorum, marshi

Lucas, 1902, armatus Mantell, 1833, Mymoorapelta maysi Kirkland et Carpenter, 1994, and foxii Hulke, 1881 to the Polacanthidae.

Aletopelta coombsi Ford et Kirkland, 2001 was assigned to the Ankylosauridae by Ford and Kirkland (2001), but is regarded as a nomen dubium by Vickaryous et al. (2004). An analysis by Vickaryous et al. (2001) recovered Animantarx ramaljonesi Carpenter, Kirkland, Burge et Bird, 1999 as an ankylosaurid.

Antarctopelta oliveroi Salgado et Gasparini, 2006 may belong to the

Ankylosauridae, although it also shares characteristics with the nodosaurids. Lu

4 et al. (2007) described Crichtonsaurus benxiensis Lu, Ji, Gao et Li, 2007, and assigned it to the Ankylosauridae; Crichtonsaurus bohlini Dong, 2002, was regarded as a nomen dubium by Vickaryous et al. (2004). Stratigraphic information for ankylosaur taxa discussed in this thesis is provided in Table 1.1.

Gargoyleosaurus Minmi * Gastonia

Gobisaurus Shamosaurus

Tsagantegia

Tarchia

« Saichania

Talarurus

Pinacosaurus grangeri

Pinacosaurus mephistocephalus

Tianzhenosa urus

Ankylosaurus

Euoplocephalus

Nodosauridae

Figure 1.1: Phylogeny of the Ankylosauria, modified from Vickaryous et al. (2004). Node 1, Ankylosauria; node 2, Ankylosauridae; node 3, Ankylosaurinae. Nodosaurids lack tail clubs. Ankylosaurus, Euoplocephalus, Gargoyleosaurus, and Gastonia are from North America, Minmi is from Australia, and the remainder are from Asia. Table 1.1 Stratigraphic information for ankylosaurs discussed in this thesis, arranged in increasing stratigraphic order. Data summarized from Vickaryous et al. (2004), unless otherwise noted. Taxon Formation and Location Range Gargoyleosaurus Morrison Fm. (Wyoming), United States -Tithonian (Late ) Mymoorapelta Morrison Fm. (Colorado), United States Kimmeridgian-Tithonian () Gastonia Lower Cedar Mountain Fm. (Utah), United Berriasian-Hauterivian (Early States ) Hylaeosaurus Hastings Beds (East Sussex, West Sussex), Late Berriasian- (Early England Cretaceous) Hoplitosaurus Lakota Fm. (South Dakota), United States () Polacanthus Lower Greensand (Dorset), Wessex Fm., Barremian (Early Cretaceous) Vectis Fm., Lower Greensand (Isle of Wight), England Gobisaurus Ulansuhai Fm. (Nei Mongol Zizhiqu), People's Aptian/?Albian (Early Cretaceous) Republic of China Minmi Bungil Fm. (Queensland), Australia Aptian/Albian (Early Cretaceous) Shamosaurus Hilhteeg Svita (Dornogov', Ovorkhangai), Aptian-Albian (Early Cretaceous) Mongolia Animantarx Upper Cedar Mountain Fm. (Utah), United early (Late States Cretaceous)

Crichtonsaurus benxiensis (Lu Sunjiawan Fm. (Liaoning), People's Republic of Cenomanian- (Late et al. 2007) China Cretaceous) Crichtonsaurus bohlini Sunjiawan Fm. (Liaoning), People's Republic of Cenomanian-Turonian (Late China Cretaceous) Talarurus Bayanshiree Svita (Dornogov', Omnogov'), Cenomanian-Campanian (Late Mongolia Cretaceous) Tsagantegia Bayanshiree Svita (Dornogov'), Mongolia Cenomanian- () Pinacosaurus granger] Djadokhta Fm., Beds of Alag Teeg (Omnogov'), ?late Santonian-?middle Campanian Mongolia; Djadokhta Fm. (Nei Mongol Zizhiqu), (Late Cretaceous) Wangshi Group (Shandong), People's Republic of China Pinacosaurus Djadokhta Fm. (Nei Mongol Zizhiqu), People's ?late Santonian or early Campanian mephistocephalus Republic of China (Late Cretaceous) Tianzhenosaurus Huiquanpu Fm. (Shanxi, Hebei), People's Late Cretaceous Republic of China Saichania Baruungoyot Fm., Red Beds of Hermiin Tsav ?middle Campanian (Late (Omnogov'), Mongolia Cretaceous) Aletopelta (Ford and Kirkland Point Loma Fm. (California), United States upper Campanian (Late Cretaceous) 2001) Antarctopelta (Salgado and Santa Marta Fm. (North James Ross Island), upper Campanian (Late Cretaceous) Gasparini 2006) Antarctica Euoplocephalus Upper Two Medicine Fm. (Montana), United late Campanian-early Maastrichtian States; Dinosaur Park Fm., Horseshoe Canyon (Late Cretaceous) Fm. (Alberta), Canada Nodocephalosaurus Lower Kirtland Fm. (New Mexico), United late Campanian or early States Maastrichtian (Late Cretaceous) Tarchia Baruungoyot Svita, Nemegt Fm., White Beds of ?late Campanian-early Maastrichtian Hermiin Tsav (Omnogov'), Mongolia (Late Cretaceous) Ankylosaurus Hell Creek Fm. (Montana), Lance Fm. late Maastrichtian (Late Cretaceous) (Wyoming), United States,; Scollard Fm. (Alberta), Canada Chapter 2: Morphology of ankylosaurid pelves and tails and the evolution of ankylosaurid tail clubs1

2.1 Introduction

In order to better understand the evolution and function of ankylosaurid tail clubs, a detailed survey of the morphology of tail clubs is necessary. This will highlight inter- or intraspecific variation in club morphology. In this study, ankylosaur tail clubs, free caudal vertebrae, chevrons, and pelves were examined (Appendix 1).

Ankylosaurid tail clubs are composed of a handle and knob (terminology sensu Coombs 1995). The handle is formed of interlocking and variably fused distal caudal vertebrae, representing approximately the distal half to third of the tail (Fig. 2.1). Large osteoderms at the terminus of the tail enclose approximately three handle vertebrae, forming the knob. The knob always contains two large lateral osteoderms (major plates), and at least three small, paired and unpaired osteoderms at the terminus (minor plates). The anterior portion of the tail consists of unfused, non-interlocking caudal vertebrae (free caudals), similar to those found in other dinosaurs. The last free caudal vertebrae is intermediate in form between the free caudal vertebrae and handle vertebrae The anteriormost caudal vertebra usually fuses via the transverse processes to the ilia, and is here called a sacrocaudal (after Parks 1926).

Portions of this chapter have been submitted for publication. Arbour, V., Sissons, R.L., and Burns, M.E. A redescription of the ankylosaurid dinosaur Dyoplosaurus acutosquameus Parks, 1924 (Ornithischia: Ankylosauria) and a revision of the . Journal of Vertebrate Paleontology, submitted 22 August 2008. All text and figures in this chapter are by V. Arbour. Permission from coauthors can be found in Appendix 2. Figure 2.1: Diagram of tail terminology used in this thesis. Ankylosaurid tail reconstructed from ROM 784; ROM 784 lacks the transitional caudal vertebra and the anterior portion of the pelvis. Scale equals 1 m.

2.2 Materials and methods

Where possible, material was examined firsthand and measured using digital calipers and a measuring tape. Each measurement was made three times and averaged. Selected measurements of an Ankylosaurus tail club (AMNH

5214) on display in the galleries of the American Museum of Natural History were measured using a laser squared against the glass and moved along a tape measure. Some measurements were obtained using photographs and ImageJ. If a scale is present in the image, ImageJ can be calibrated to measure distances, angles, and areas. Direct observation is supplemented by figures and descriptions from the literature.

8 2.3 Pelvic morphology of Euoplocephalus

Two complete (AMNH 5337 and AMNH 5409) and several partial pelves

(AMNH 5245, AMNH 5470, BMNH R5161, ROM 784, ROM 1930, and TMP

1982.9.3) are referred to Euoplocephalus. The pelves of BMNH R5161 and ROM

784 are parts of incomplete articulated specimens, ROM 1930 includes a skull and partial postcranium, AMNH 5337 includes a skull, both cervical half rings, and partial postcranium, AMNH 5245 includes partial postcranium, and AMNH

5409, AMNH 5470 and RTMP 1982.9.3 are isolated elements. A partial pelvis and sacrum were collected with UALVP 31, but have not been prepared yet. All of these specimens were collected from Alberta, Canada, with the exception of

AMNH 5470, which was collected from Montana, U.S.A.

AMNH 5337 (Fig. 2.2) and 5409 are similar in overall morphology and size, although AMNH 5337 lacks ischia. The ilia are broad and diverge from the midline anteriorly; the preacetabular process is long, while the postacetabular process is short. Each tapers distally, is directed ventromedially, and has a rounded, curved attachment with the .

No pubes can be discerned on the specimen. The centra of four dorsosacral, two sacral, and two sacrocaudal vertebrae fuse to form a sacral rod.

The neural spines of the sacrals and dorsosacrals (with the exception of the last sacral) are fused to form a continuous blade. Sacral and dorsosacral ribs project at right angles to the centra, and have flat, horizontal dorsal surfaces. This condition is also observed in the incomplete specimens AMNH 5245, AMNH

9 s sd sc2 sc3 fd

Figure 2.2: Ankylosaurid pelves. A) AMNH 5409, Euoplocephalus (modified from Coombs 1971), dorsal view, anterior is up. The sacral fenestrae are oval, and the sacral and dorsosacral transverse processes project from the centra laterally or slightly posterolaterally; B) ROM 784, Dyoplosaurus acutosquameus (holotype) dorsal view, anterior is up. The anteriormost sacral fenestrae are trapezoidal, and the more posterior sacral fenestrae are oval. The sacral transverse processes project anterolaterally from the centra, and the sacrocaudal transverse processes project laterally or slightly posterolaterally; C) AMNH 5409, posterior view. The ischia are medioventrally directed; D) ROM 784, oblique right anterior view. The ischia are ventrally directed. E) Pelvic region of ROM 784, Dyoplosaurus, in oblique left dorsolateral view, anterior is to the left. The sacral transverse processes are anteroposteriorly tilted ventrally, compared to the sacrocaudal transverse processes. Scale bars equal 10 cm. Abbreviations: fc, free caudal; il, ilium; is, ischium; os, ; s, sacral; sc, sacrocaudal.

10 5470, BMNH R5161, and TMP 1982.9.3. Distally, the sacral ribs expand anteroposteriorly before contacting the ilia. The dorsosacral ribs do not expand distally, and do not fuse as strongly with the ilia. AMNH 5337 and AMNH 5409 both have unusual excavations at the junction of the ilium and fourth dorsosacral rib. In AMNH 5409, the excavation is on the right side, and the fourth dorsosacral rib does not contact the right ilium, whereas in AMNH 5337, the excavation is on the left side and the fourth dorsosacral rib still contacts the ilium. These excavations will be discussed in greater detail in Chapter 3.

ROM 784 is strikingly different in pelvic morphology compared to all other referred Euoplocephalus specimens, with anteriorly directed, anteroventrally tilted second sacral ribs (Fig. 2.2). The sacral ribs on the sacrocaudals are broad with flat dorsal surfaces. In contrast, the ribs on the third sacral are tilted along the transverse axis so that the flat dorsal surfaces are at angles of approximately

25° from the horizontal. The sacrocaudals have sacral ribs directed posteriorly at approximately the same angle, while the third sacral has sacral ribs directed anteriorly. This results in dorsal sacral foramina with different shapes. The posteriormost foramen is roughly oval, while the outline of the anteriormost foramen is an isosceles trapezoid. The sacral foramina decrease in size posteriorly. The ischia attach at right angles to the ilium, forming a sharp 90 degree angle when viewed anteriorly. The ischia are broken approximately halfway down, and the proximal halves are not directed medially. The right ilium is badly damaged, and the anteriormost portion of the left preacetabular blade has been reconstructed.

11 Penkalski (2001) noted differences in the number of vertebrae incorporated into the pelvis, the shape of the acetabulum, and the angle of divergence between the ilia. It is difficult to confirm any differences in the number of vertebrae. Eight are preserved in AMNH 5337: four dorsosacrals, three sacrals, and one sacrocaudal. AMNH 5409 also has four dorsosacrals, three sacrals, and one sacrocaudal; although the last vertebra is mostly sculpted, the right sacrocaudal transverse process is real. TMP 1982.9.3 has four dorsosacral vertebrae, two sacrals with transverse processes, and one sacral without preserved ribs. BMNH R5161 (holotype of cutleri) has four dorsosacrals, three sacrals, and two sacrocaudals (Nopsca 1928, plates VI and

VII). The last sacrocaudal in BMNH R5161 and ROM 784 are not fused to the other sacrocaudals, and so are less likely to be preserved in isolated or partial pelves, which could explain the difference in sacrocaudal number between these specimens and AMNH 5337 and AMNH 5409.

The shape of the acetabulum could not be confirmed in most specimens:

AMNH 5337, AMNH 5409, ROM 784, and TMP 1982.9.3 are mounted and difficult to examine ventrally, and the other specimens do not preserve the acetabulum. Most ilia vary somewhat in the amount of divergence from the midline, but the effects of taphonomy and ontogeny are unclear.

12 2.4 Caudal morphology

2.4.1 Nodosaurids

Nodosaurid caudal vertebrae are, perhaps surprisingly, more robust than ankylosaurid caudals. CMN 8531, the holotype of Edmontonia longiceps,

Sternberg, 1928, has heart-shaped caudal centra (Fig. 2.3). The transverse processes are stout, with oval cross sections, both on anterior and more posterior caudals. The transverse processes have broad bases at the centra, and do not usually form an acute angle with the centrum in anterior or posterior view.

The neural arch and neural spine are robust, and the neural spine is oval in cross section. The prezygapophyses and postzygapophyses have broad articular faces, and are long and well separated from the neural spine. The haemal arch, where preserved, is robust.

Figure 2.3: CMN 8521, Edmontonia longiceps (holotype), anterior caudal vertebra in anterior view. Scale equals 5 cm.

13 Numerous vertebrae are known for Sauropelta edwardsi Ostrom, 1970, including anterior and posterior caudal vertebrae. Anterior caudals are robust, with massive, rodlike, anteriorly-directed transverse processes and round centra.

The neural spine expands distally and the pre- and postzygapophyses are large and well separated from the neural spine. Distal caudal centra are hexagonal in anterior and posterior view, and are anteroposteriorly longer than anterior caudals. The neural spine is more slender but the pre- and postzygapophyses are still robust.

2.4.2 Basal ankylosaurids

Caudal vertebral morphology of several "polacanthid" ankylosaurs is consistent with the morphology seen in the free caudal vertebrae of ankylosaurids. Many of the "polacanthid" ankylosaurs are probably basal members of the Ankylosauridae. In this study, polacanthid ankylosaurs are considered basal ankylosaurids (as in Vickaryous et al. 2004, contra Carpenter

2001), and Mymoorapelta and Polacanthus are included in this discussion.

Six caudal vertebrae were preserved in the holotype specimen of

Gargoyleosaurus parkpinorum DMNH 27726 (Fig. 2.4). These have amphicoelus centra with round anterior and posterior faces. The prezygapophyses are long, especially in the more distal vertebrae, while the postzygapophyses are not well separated from the neural spines. Neural spines are bladelike, laterally slender, distally tapering, and directed dorsoposteriorly. Transverse processes taper

14 Figure 2.4: Basal ankylosaurid caudal vertebrae. A) DMNH 27726, Gargoyleosaurus parkpinorum (holotype), anterior caudal vertebra, anterior view. B) MWC 5819, Mymoorapelta maysi (holotype), two distal caudal vertebrae, right lateral view and C) dorsal view. Scale bars equal 5 cm. Photographs of MWC 5819 provided by M. Burns and used with permission. distally, and project laterally from the centra. Distally, the centra become anteroposteriorly longer.

More than 200 caudal vertebrae referred to Gastonia have been collected from the Ruby Ranch Member of the . Sixty-seven were examined during the course of this project, with the aim of better understanding the amount of variation in caudal vertebral morphology in a single taxon. Anterior, middle and distal caudals in varying states of preservation were measured and classified into one of three shape categories, based on the anterior face of the centrum: circular, heart-shaped, and intermediate. Most of the vertebrae had circular anterior faces (52, or 78%), while 9 (13%) had heart- shaped anterior faces and 6 (9%) had intermediate morphologies. Heart-shaped morphologies seem to correspond to the more anterior caudal vertebrae, as

15 these vertebrae were generally larger and had more robust transverse processes. Transverse processes project at right angles to the centrum and are usually directed lateroventrally. The prezygapophyses are well separated, long and sticklike, while the postzygapophyses are smaller and not well separated from the neural spine. The neural spine itself is rarely preserved.

Postzygapophyses decrease in size distally in the tail, more posterior neural spines become progressively longer and lower relative to the more anterior neural spines, and the centra become more elongate anteroposteriorly compared to the more anterior centra. The neural canal is usually circular to oval, and rarely triangular.

Mymoorapelta, from the of western Colorado, was originally considered a nodosaurid (Kirkland and Carpenter 1994), then a polacanthid (Carpenter 2001). The anterior caudal has a heart-shaped centrum.

The neural spine is relatively short, and laterally expanded distally. The prezygapophyses are broad and widely separated from each other, while the postzygapophyses are poorly separated from the neural spine. The transverse processes are long, lateroventrally directed, and slightly expanded distally. In more distal caudals (Fig. 2.4), the transverse processes are much shorter and project laterally, the neural spine is more tapered and bladelike, and the prezygapophyses are sticklike. The neural spine of one vertebra is enclosed laterally by the prezygapophyses of the next successive vertebra. Two distal caudals interlock: the neural spine of one is bordered laterally by the prezygapophyses of the next successive vertebra. The prezygapophyses are

16 long and taper to points, and the centra are anteroposteriorly elongate.

Transverse processes are present only as small ridges on the lateral surfaces of the centra. The haemal arch is fused to the posterior end of the centrum, and the haemal spine is elongate and hatchet-like.

Nopsca (1905) figures several caudal vertebrae belonging to Polacanthus.

The largest, most anterior caudal has a heart-shaped centrum in posterior view, and a stout based, robust neural spine. A more distal caudal has a circular centrum, less robust transverse processes, and a neural spine that is laterally expanded distally. Two distal caudals have more anteroposteriorly elongate centra, with long, sticklike prezygapophyses. The neural spines are broken, but they do not appear to have formed an interlocking structure with the prezygapophyses of the following vertebrae. The transverse processes are ridges on the lateral sides of the centra. Blows (1987, 2001) figured a "caudal end mass" that includes several vertebrae, ossified tendons, two oval osteoderms, and a rod-like bone. This caudal end mass does not appear to represent a true tail club, as the vertebrae do not seem to form the interlocking handle structure.

2.4.3 Derived ankylosaurids

Caudal vertebrae of ankylosaurid and nodosaurid ankylosaurs are distinct, and ankylosaurid free caudal vertebrae may also be taxonomically useful at the level of genus and species. An ankylosaurid free caudal vertebra differs in the shape of the centrum, the direction the transverse process projects from the

17 centrum, the nature of the contact between the transverse process and centrum, and the shape of the prezygapophysis and postzygapophysis. The neural spines, transverse processes and postzygapophyses progressively decrease in size posteriorly, and the centra become shorter and narrower. In the handle, the centra are anteroposteriorly elongate, with dorsoventrally depressed anterior and posterior faces that may be rectangular to figure-eight shaped. The neural spines are long and low, and embraced by the elongate prezygapophyses of the subsequent vertebrae. Postzygapophyses are absent in the handle, and the transverse processes are generally absent. Ossified tendons are only found associated with the handle vertebrae.

Free caudal vertebrae referred to Euoplocephalus show great variation, and it is sometimes difficult to determine what variation is due to taphonomic, ontogenetic, individual, and intrageneric variation, and which variation represents taxonomic differences at a suprageneric level. All of the centra of free caudal vertebrae referred to Euoplocephalus have circular or subcircular articular faces.

Notochordal prominences may be present to varying degrees on the anterior face, posterior face, both, or neither. Neural spines are typically short, laterally compressed, and tapering; neural spines with expanded distal tips are generally rugose and probably pathological. Neural canals are triangular. Pre- and postzygapophyses may be broad or narrow and cylindrical. The transverse processes project anterolaterally from the centrum, and generally form sharp, acute angles ventrally with the centrum, in anterior or posterior view. The transverse processes of all but the most proximal free caudals taper distally.

18 Haemal spines, when preserved, are laterally compressed and tapering, and haemal canals are usually oval.

ROM 784, the holotype of Dyoplosaurus acutosquameus Parks, 1924, includes the best example of a North American ankylosaurid caudal series (Fig.

2.5). ROM 784 preserves three sacral vertebrae, one sacrocaudal vertebra, eleven free caudal vertebrae, and eleven handle vertebrae (sensu Coombs

1995). This appears to represent the entire caudal series, although the distalmost free caudal may be missing. The centrum of the sacrocaudal (after Parks 1924; first free caudal in Penkalski 2001) is not fused to the last sacral. The centrum is not as long as each of those of the sacral series, but is approximately the same shape and size as the first caudal. The neural arch is complete and similar to those of the caudals. The caudosacral ribs, intermediate in shape between the sacral and caudal ribs, are broad, like the sacral ribs, but taper like the caudal ribs. While the caudal ribs taper to points, the caudosacral ribs taper to rounded edges, which are fused to, but distinct from, the postacetabular blades of the ilia.

The ribs are also curved anterolateral^, like the caudal ribs, although not as strongly.

ROM 784 has eleven free caudals (Fig. 2.6) preserved in articulation. The articular faces of the centra are mostly obscured. The posterior articular face is visible on the second free caudal, and the anterior face is visible on the third free caudal. The centra have been crushed and are skewed clockwise, but the centra are round in articular view, and anteroposteriorly constricted in lateral view. They are slightly amphicoelous, and no notochordal prominence is noted where

19 Figure 2.5: Skeletal and dermal armour reconstruction of preserved elements of ROM 784, Dyoplosaurus acutosquameus (holotype), using Euoplocephalus tutus (Carpenter 1982) as a guide for missing elements. Anterior is to the top. Scale equals 1 m.

20 Figure 2.6: Ankylosaurid free caudal and handle vertebrae. A) MPC KID2007.167, Talarurus, isolated vertebra in right lateral view, anterior is to the right. ROM 784, Dyoplosaurus (holotype), B) section of the handle in dorsal view, showing the tightly interlocking nature of the neural spines and prezygapophyses, anterior is to the right, C) tail club in oblique right posterolateral view, showing the two sets of ossified tendons, D) second free caudal vertebra in posterior view, and E) first and second free caudal vertebrae in right lateral view. F) ROM 1930, Euoplocephalus, transitional free caudal vertebra in left lateral view. Note the presence of the prezygapophyses from the first handle vertebra. Scale bars in A, B, D, E, and F equal 5 cm, scale bar in C equals 10 cm. Abbreviations: ns, neural spine; prz, prezygapophysis; prz2, prezygapophysis of first handle vertebra. Photograph of MPC KID2007.167 provided by P. Currie and used with permission.

21 articular faces are visible. There is a 2-3 cm gap between each of the free caudals; the presence of similar-sized gaps between articulated caudal vertebrae in ROM 1930 suggests that these gaps represent intervertebral cartilage, although the large size of the gaps may be a result of taphonomy. Centrum width decreases posteriorly (from 93.12 mm to 74.91 mm), and all free caudals are generally wider than tall. The neural arch is massive and the neural canal is teardrop-shaped to oval. The neural spines are slender and bladelike, with laterally expanded distal ends. The distal expansions progressively decrease posteriorly along the caudal series. The neural spines are directed dorsoposteriorly. Neural spine height and width decrease posteriorly, but length does not change measurably (neural spine height of first free caudal = 39.35 mm, last free caudal = 38.86 mm). Prezygapophyses and postzygapophyses have distinct articular surfaces. The lengths of the prezygapophyses vary somewhat but do not regularly decrease or increase posteriorly. The postzygapophyses decrease in length posteriorly, until only stubs remain on the ninth and tenth free caudals, and are absent on the eleventh. Chevrons are present on the fifth, sixth, seventh and eighth free caudals, but are obscured distally and only visible on the right side of the specimen. The chevrons are bladelike and are angled posteroventrally.

The tail club is composed of ten caudal vertebrae forming the handle, and several osteoderms (overlapping the most distal vertebrae) forming the terminal knob. The centra are visible in dorsal view in handle caudals one through six

(caudals 13 through 19). The anterior articular face is visible on the first handle

22 centrum; it is concave and lacks a notochordal prominence. The centrum is elongate with a slight hourglass shape in dorsal view. Centra decrease posteriorly in width, and appear strongly fused to each other. The boundaries between centra can be difficult to distinguish. Each neural spine and its associated prezygapophyses are highly modified to create a V-shaped structure that interlocks with the preceeding neural spine (Fig. 2.6). The neural spine is triangular, and fits into the slot between the next, more posterior pair of prezygapophyses. Neural spines typically end in sharp points, but some are more tongue-like. In lateral view, the dorsal surfaces of the neural spines and prezygapophyses are almost perfectly horizontal, although the neural spines are slightly concave on the dorsal surfaces. There is often a small pit on the dorsal surface of the neural spine just beyond the terminus of the adjacent, more anterior neural spine. The prezygapophyses of the first handle vertebra are more lightly built than the rest, and the neural spine has a small, sharp dorsal ridge, which is absent on each of the remaining neural spines. Transverse processes are short stubs on the first handle vertebra, small swellings on the second handle vertebra, and absent on the remaining handle vertebrae. The chevron is visible only on the first handle vertebra, and is broad and distally blunt.

ROM 1930 has twelve free caudals, one additional caudal than the number observed in ROM 784. Seven of these (caudals four through ten) are preserved in articulation in two blocks of matrix (Fig. 2.7), along with dermal

("skin") impressions and armor; the remaining caudals are free of matrix and have varying degrees of completeness. Numerous chevrons are also preserved

23 Figure 2.7: ROM 1930, Euoplocephalus, articulated free caudals, right lateral view. Anterior is to the right.

in ROM 1930. The free caudals of ROM 1930 are similar in all respects to ROM

784, except for being proportionately larger. The twelfth and last free vertebra differs markedly from the other free caudals (Fig. 2.6). The centrum has a greater length and lower height in proportion to the other free caudals, and appears intermediate in morphology to the free caudals and handle caudals. The neural spine is dorsoventrally shorter, more strongly directed posteriorly, and anteroposteriorly longer. The prezygapophyses are much longer than those on the tenth and eleventh vertebrae. The postzygapophyses are absent. The neural spine is overlapped posteriorly by two large bladelike prezygapophyses, which closely match the morphology of prezygapophyses of the handle vertebrae in other tail club specimens (e.g. ROM 784, UALVP 47273). This morphology is present in Tarchia (Maryanska 1977), but is not observed in the free caudals of

ROM 784, which may lack its last free caudal. There is a large (8 cm) gap between the last free caudal and first handle vertebra in ROM 784. The lengths of the free caudal centra are 5-6 cm, and are usually spaced 1-2 cm apart. An additional free caudal would consequently fit in this gap. It is not known why the

24 last free caudal, which would have been partly fused to the handle, is missing in

ROM 784, even though all other free caudals appear to be present.

CMN 8530 (holotype of Anodontosaurus lambei Sternberg, 1929), differs significantly from ROM 784, ROM 1930 (Fig. 2.8), and isolated TMP caudals.

The centrum of CMN 8530 is relatively elongate compared to other specimens of

Euoplocephalus. The centrum is also strongly octagonal, with pronounced vertices in anterior and posterior views, while the centra in all other

Euoplocephalus are circular to subcircular and lack distinct vertices in anterior or posterior views.

Figure 2.8: Variations in ankylosaurid free caudal vertebrae: A) ROM 1930, Euoplocephalus tutus, proximal free caudal, anterior view. The centrum is circular, the neural spine and transverse processes taper distally, and there is sharp angle ventrally where the transverse process meets the centrum; B) CMN 8530, holotype of Anodontosaurus lambei {=Euoplocephalus), proximal free caudal, anterior view. The centrum is octagonal with distinct vertices, and the neural and haemal spines taper distally; C) AMNH 5404, Euoplocephalus tutus, mid- series free caudal, posterior view. The centrum is subcircular, the neural spine and transverse processes are comparatively more robust, and the transverse process does not make a sharp angle ventrally where it meets the centrum; D) AMNH 5895, Ankylosaurus magniventris (holotype), proximal free caudal, anterior view. The centrum is square, the neural spine is more robust compared to Euoplocephalus, and the transverse processes taper distally. Scale bars equal 5 cm.

25 AMNH 5404 includes four free caudals that differ slightly from those of other Euoplocephalus (Fig. 2.8). The transverse processes are broken, but where they attach to the centrum they form an obtuse angle in anterior or posterior views. The transverse processes are also much rounder in cross section, compared to the dorsoventrally flattened condition found in most other

Euoplocephalus specimens.

Nodocephalosaurus from the Campanian of New Mexico has two referred caudal vertebrae figured by Sullivan and Fowler (2006). Nodocephalosaurus differs from Euoplocephalus in having rounded, robust transverse processes directed laterally and somewhat anteroventrally from the centrum, which form an obtuse angle in anterior or posterior view where they meet the centrum.

The holotype of Ankylosaurus (AMNH 5895) includes one complete and several partial free caudal vertebrae (Fig. 2.8). Ankylosaurus differs from

Euoplocephalus in having a more square centrum and transverse processes that project more laterally from the centrum.

Comparisons with Asian ankylosaurids are made primarily from previously published figures and descriptions. A complete caudal vertebra belonging to

Talarurus is figured by Maleev (1956, fig. 8). The centrum has a short anteroposterior length and is approximately square in anterior view. The transverse processes are slender and tapering, and are directed anterolateral^.

Caudal vertebrae of "Dyoplosaurus cf. giganteus" (=Tarchia gigantea

Maryanska 1977) were described and figured by Maryahska (1971, plate VI, figs.

1 and 2). This specimen included the tenth or eleventh caudal vertebra, and the

26 last free caudal fused to the first handle vertebra. The tenth or eleventh vertebra has a round centrum with a large, distinct notochordal prominence. The transverse processes are tubercles located about halfway down the centrum.

This differs from the condition in ROM 784, where the transverse processes are distinct on all of the free caudals, and tubercles are present on the first handle vertebra. The neural arch of the last free caudal is fused to the neural arch of the first handle vertebra.

Caudal vertebrae are known in both Pinacosaurus grangeri and

Pinacosaurus mephistocephalus. In P. mephistocephalus, the transverse processes disappear quickly posteriorly in the sequence. The anterolateral^ directed transverse processes also nearly disappear well before the first handle vertebra in P. grangeri (Syrmosaurus viminicaudus in Maleev 1954, fig. 4).

Pinacosaurus ninghsiensis Young 1935 (=P. grangeri, Maryanska 1971) shows a similar morphology, but the transverse processes project straight laterally from the centrum, as do those of Heishansaurus pachycephalus Bohlin, 1953 (=P. grangeri, Maryanska 1971). Free caudal vertebrae are also known for

Crichtonsaurus benxiensis (Lu et al. 2007), and appear typically ankylosaurid

(contra Lu et al. 2007).

Ankylosaurid tail clubs are composed of numerous interlocking vertebrae

(the handle), ossified tendons, and large terminal osteoderms forming the knob

(Fig. 2.1). All knobs include two major plates, one on each side of the handle vertebrae, and a varying number of minor plates that comprise the distal end of the knob. Numerous tail clubs have been referred to Euoplocephalus, whereas

27 most other ankylosaurid taxa each have only one or a few known tail clubs.

Euoplocephalus knobs are highly variable; Coombs (1995) reviewed this variation and suggested that ontogeny, intraspecific variation, taxonomic variation, or a combination thereof may be responsible. Four specimens in this study (AMNH 5245, ROM 784, ROM 788, and UALVP 47273) have substantial portions of the handle preserved. The length of the prezygapophyses appears to increase posteriorly in AMNH 5245, ROM 788, and UALVP 47273, while the length decreases in ROM 784. The length of the neural spine increases posteriorly in ROM 788, decreases in ROM 784 and UALVP 47273, and remains the same in AMNH 5245. Centra are fused (Fig. 2.9) along the handle in ROM

784, UALVP 47273, and possibly in ROM 788. Centra do not appear to be fused

(Fig. 2.9) in Ankylosaurus (AMNH 5419), AMNH 5245 {Euoplocephalus),

Figure 2.9: A) Unfused centra in the handle of AMNH 5245, Euoplocephalus, marked by arrowhead. B) Unfused centra in the handle of MPC PJC2007.123, Pinacosaurus granger!. Right lateral views, anterior is to the right in both photos. Abbreviations: ex, most anterior centrum in specimen; cy, second most anterior centrum in specimen; ot, ossified tendon; px, prezygapophysis of most anterior vertebra in specimen; py, ot ex cy prezygapophysis of second most anterior vertebra; pz, prezygapophysis of third most anterior vertebra. Scale in A equals 10 cm, scale in B equals 3 cm. B juvenile Pinacosaurus granger/ (MPC PJC 2007-123), and Tarchia (ZPAL MgD-

I/43).

Broadly, tail clubs can be divided into two categories (Fig. 2.10): those with handle vertebrae that have sharply pointed neural spines (Euoplocephalus,

Pinacosaurus, Saichania, and Talarurus), and those with handle vertebrae that

have broad, tonguelike neural spines (Ankylosaurus, Tarchia). All

Euoplocephalus have sharply pointed neural spines in the handle. The angle of

divergence of the neural spine and prezygapophyses (Fig. 2.10), measured from

the posterior point of the neural spine in dorsal view, was measured in

Ankylosaurus, Euoplocephalus, Saichania, Talarurus, and Tarchia, (Table 2.1).

Figure 2.10: Variation in ankylosaurid handle vertebrae: A) MPC KID2007.167, Talarurus, isolated vertebra in dorsal view. Also shown are the dimensions for neural spine and prezygapophysis length, maximum neural arch width, and neural spine angle; B) AMNH 5245, Euoplocephalus, section of handle in oblique dorsal view; C) UALVP 47948 (cast of ZPAL MgD-l/113), Tarchia, section of handle in dorsal view; D) AMNH 5214, Ankylosaurus, section of handle in oblique dorsal view. Anterior is to the right. Scale bars equal 5 cm. Photograph of MPC KID2007.167 provided by P. Currie and used with permission.

29 Euoplocephalus and Saichania have similar angles (-22°, average).

Ankylosaurus AMNH 5214 has an average of 63°, whereas a larger but weathered Ankylosaurus, CCM V03, has an average of 53°. Two specimens referred to Tarchia have different morphologies and angles: UALVP 47948 has an average of 35° and a more tonguelike morphology, while ZPAL MgD-l/43 has pointed neural spines with an average of 23°. It is possible that these two specimens do not both represent the same taxon. Talarurus MPC KID2007.167 is an isolated vertebra that falls within the range of variation of Euoplocephalus and Saichania. Compared to the differences between genera, there is no difference in neural spine angle among specimens referred to Euoplocephalus.

Table 2.1 Neural spine angle. See Fig. 2.10 for explanation of measurement. Taxon Specimen # Vertebra # Notes 1 2 3 4 5 6 7 8 9 Mean Ankylosaurus AMNH 5214 69 49 75 64 57 63 ImageJ, from Carpenter (2004) Ankylosaurus CCM V03 54 53 53 ImageJ, from Carpenter (2004) Dyoplosaurus ROM 784 20 17 26 19 20 22 20 28 30 22 Protractor Euoplocephalus AMNH 5245 35 20 15 20 22.5 Protractor Euoplocephalus ROM 788 30 25 26 20 22 21 24 Protractor Euoplocephalus UALVP 26 14 21 20 28 22 Protractor 47273 Pinacosaurus MPC PJC 26 26 26 ImageJ, from granger! 2007-123 B photo by V. Arbour Saichania PIN 21 16 19 21 22 25 25 30 22 ImageJ, from 3142/251 Tu ma nova (1987)

Tarchia UALVP 34 35 33 39 34 41 42 37 Protractor 47948 Tarchia ZPAL MgDI/ 29 19 ImageJ, from 43 photo by P. Currie Talarurus MPC 16 16 ImageJ, from KID2007.167 photo by P. Currie

30 The greatest variation in tail clubs lies in the shape of the knob (Figs. 2.11,

2.12). While some differences in size and shape may be correlated with ontogeny, it is also possible that some of these differences represent real taxonomic differences. Tail club knobs (Table 2.2) range in width from slender

(<200 mm), to average (200 - 500 mm), to large (>500 mm). The Ankylosaurus knob (AMNH 5214) and Tarchia knob (ZPAL MgD I/43) are both large, which would at first appear to be correlated with the more 'robust' tonguelike interlocking vertebrae. However, the largest knobs, AMNH 5245 (593 mm) and

ROM 788 (572 mm) are referred to Euoplocephalus, and have the sharply- pointed, less robust morphology. To determine whether there was a relationship between the size of the handle vertebrae and the size of the knob, measurements of the handle vertebrae (neural spine length, prezygapophysis length, and neural arch width) of AMNH 5245, ROM 784, ROM 788, and UALVP

47273 were averaged and plotted against knob width. Larger vertebral dimensions should correspond to larger individuals. There was no increase in vertebral elements associated with larger knob widths, indicating that larger individuals did not necessarily bear larger knobs (Fig. 2.13). Knob size may vary by taxon, age, or individual. If knob size is linked to ontogeny, these results suggest that the knob may have developed relatively late in life when the individual ankylosaur was near adult size, rather than the knob increasing in size in proportion to increasing body size.

31 Figure 2.11: Ankylosaurid knobs in dorsal view, showing variation in shape and number of osteoderms. AH are referred to Euoplocephalus, except for B (Dyoplosaurus) and J (Tarchia). Slender knobs: A) UALVP 47273, B) ROM 784, C) ROM 7761. Average-sized knobs: D) CMN 40605, E) UALVP 16247, F) TMP 83.36.120, G) TMP 94.168.1. Large knobs: H) AMNH 5245, I) ROM 788, J) ZPAL MgD-l/43. Scale equals 10 cm. Photo of ZPAL MgD-l/43 provided by P. Currie and used with permission.

32 Figure 2.12: Ankylosaurid knobs, showing variation in osteoderm shape and position of the keel. A) AMNH 5245, Euoplocephalus, oblique posterior view. B) CMN 40605, Euoplocephalus, posterior view. C) ROM 7761, Euoplocephalus, posterior view. D) ROM 784, Dyoplosaurus, posterior view. E) TMP 83.36.120, Euoplocephalus, posterior view. Scale equals 10 cm.

Figure 2.13: Graph showing the relationships between various AMNH 5245 600 B vertebral dimensions and knob size in Euoplocephalus. A) Knob ROM 788 i maximum width versus mean E i maximum handle centrum width. 500 E^ i x B) Knob maximum width versus n i •• H mean handle neural spine length. JZ TMP 83.36.120 30° izip 30° C) Knob maximum width versus T3 40M M O w o M O " S I II dk o o> mean handle prezygapophysis °X '§; length. to + IS*. °1 N> GO I 30W -* N) ^8 o + Si 00 _» E 00 X bi CO 20CH ROM 784 .Q UALVP 47273 O 10CH

0 20 40 60 80 100120140160180200 Vertebral Dimension (mm)

33 Table 2.2 Knob widths arranged by size. Taxon Specimen # Width Notes (mm) Euoplocephalus ROM 7761 133 Euoplocephalus UALVP 47273 155 Dyoplosaurus ROM 784 166 Euoplocephalus CMN 40605 299 Euoplocephalus UALVP 16247 355 Euoplocephalus TMP 1983.36.120 418 Euoplocephalus AMNH 5405 -420 Knob in two halves; estimate only Euoplocephalus TMP 94.168.1 467 Euoplocephalus MACN Pv 12554 470 Measurement by P. Currie and E. Snively Ankylosaurus AMNH 5214 470 ImageJ, from Carpenter (2004) Euoplocephalus ROM 788 572 Euoplocephalus AMNH 5245 593

In dorsal view (Fig. 2.11), major plates are generally straight to sigmoidal along the medial edge, and oval to circular on the lateral edge. An unusual knob,

AMNH 5245 (Fig. 2.11), is triangular and pointed laterally. Many knobs have distinct longitudinal keels at the midheight or higher on the major plates (Fig.

2.12). The axes of most keels are laterally-directed, but in a few specimens

(AMNH 5245, UALVP 16247) the axis is directed dorsolateral^. The major plates extend closer to the midline on the dorsal side than on the ventral side; there is usually a gap between the major plates, with the exception of CMN 40605 (Fig.

2.11). ROM 784 has a slender tail club knob with a relatively smooth surface texture (Fig. 2.11), and pronounced, sharp keels on the major plates. Other slender knobs include ROM 7761 and UALVP 47273 (Fig. 2.11), as well as

Saichania PIN 3142/251.

34 2.5 Discussion

2.5.1 Number of species represented by material referred to Euoplocephalus

Euoplocephalus tutus (originally Stereocephalus tutus Lambe, 1902) is currently one of three valid taxa of ankylosaurids from North America, along with the more southern Nodocephalosaurus, and the younger Ankylosaurus. It is the most common North American ankylosaurid, and specimens referred to this genus make up the core of this study. However, there is substantial variation among specimens referred to Euoplocephalus. It is important to understand how many taxa of ankylosaurids are represented by Euoplocephalus, in order to put differences in tail club morphology and mechanics into evolutionary perspective.

It is possible that different vertebral and knob morphologies may correspond to different mechanical responses to tail club swinging, and this may have implications for use of the tail club in different genera.

Coombs (1978) synonymized three taxa with Euoplocephalus tutus:

Dyoplosaurus acutosquameus, Scolosaurus cutleri, and Anodontosaurus lambei.

Dyoplosaurus (ROM 784) includes the only complete caudal series of a specimen referred to Euoplocephalus, as well as a pelvis, limbs, skull fragment and armour. Scolosaurus lacks the skull and distal half of the tail, but preserves the remainder of the skeleton, including armour. Anodontosaurus includes a complete but crushed skull, as well as a partial postcranium. Many specimens have been referred to Euoplocephalus, and variation within this taxon has been reviewed by Penkalski (2001). Carpenter (1982) reconstructed the skeleton and

35 armour of Euoplocephalus using primarily Dyoplosaurus and Scolosaurus, and suggested that Scolosaurus and Anodontosaurus may be distinct from

Euoplocephalus based on cervical half ring morphology. Penkalski (2001) suggested that Scolosaurus and Dyoplosaurus may be distinct from

Euoplocephalus and each other, but did not formally resurrect either name.

Penkalski (2001) also noted that ROM 813 differs significantly from Scolosaurus, but did not determine whether ROM 813 was referable to Euoplocephalus.

In Alberta, Euoplocephalus is found throughout the Dinosaur Park

Formation, as well as the Horseshoe Canyon Formation (Ryan and Evans 2005).

Specimens referred to Euoplocephalus have also been found in Montana, in the

Judith River formation (Penkalski 2001). In contrast, nodosaurid species in

Alberta have more restricted stratigraphic ranges: Edmontonia rugosidens

Gilmore, 1930 is found in the lower part of the ,

Panoplosaurus mirus in the upper part, and Edmontonia longiceps is found in the

Horseshoe Canyon Formation (Ryan and Evans 2005). In fact, many ornithischian species in the Dinosaur Park Formation of Alberta have quite restricted stratigraphic ranges at different levels throughout the formation.

Furthermore, the Oldman, Dinosaur Park, and Horseshoe Canyon Formations each host their own unique faunas. The Oldman Formation includes

Albertaceratops nesmoi Ryan, 2007, Centrosaurus brinkmani Ryan and Russell,

2005, and Brachylophosaurus canadensis Sternberg, 1953 (Ryan and Evans

2005). Ryan and Evans (2005) recognized three biozones within Dinosaur

Provincial Park. The Centrosaurus-Corythosaurus biozone is found in the lowest

36 20 to 25 metres, and includes Centrosaurus apertus Lambe, 1904,

Chasmosaurus russelli Sternberg, 1940 (Holmes et al. 2001), and Gryposaurus notabilis Lambe, 1914 (Ryan and Evans 2005). The Styracosaurus-

Lambeosaurus lambei biozone in found in the next 25 m, and includes

Chasmosaurus belli Lambe, 1914 (Holmes et al. 2001), Styracosaurus albertensis Lambe, 1913 (Ryan et al. 2007), Lambeosaurus lambei Parks, 1923, and Prosaurolophus maximus Brown, 1916 (Evans and Reisz 2007). An upper fauna near the Lethbridge Coal Zone (at the top of the Dinosaur Park Formation) contains rare but distinct species, including an undescribed pachyrhinosaur,

Chasmosaurus irvinensis Holmes et al., 2001, and Lambeosaurus magnicristatus

Sternberg, 1935; Prosaurolophus is also present (Ryan and Evans 2005).

Lambeosaurus lambei and Lambeosaurus magnicristatus are stratigraphically separated within the Dinosaur Park Formation (Evans and Reisz 2007), as are

Centrosaurus apertus and Styracosaurus albertensis (Ryan et al. 2007).

Tyrannosaurids show a similar pattern: Daspletosaurus torosus Russell, 1970 is known from the Oldman Formation, and a second species of Daspletosaurus from the Dinosaur Park Formation is being described by Currie and Bakker.

Gorgosaurus libratus is the most common tyrannosaurid in the Dinosaur Park

Formation, whereas Albertosaurus sarcophagus Osborn, 1905 is only known with certainty from the Horseshoe Canyon Formation (Currie 2003). If only one species of ankylosaurid is found throughout the Dinosaur Park Formation and the

Horseshoe Canyon Formation, this raises an interesting question: why is this single taxon present in both formations, while other dinosaur taxa are being

37 replaced? It is possible that several taxa are represented by specimens currently referred to Euoplocephalus tutus. Anatomical/phylogenetic comparisons and a detailed stratigraphic analysis of all specimens referred to Euoplocephalus tutus is beyond the scope of this project, but could provide much information on the number of taxa represented.

Few dinosaur species are found in the same stratigraphic interval in both

Montana and Alberta. For example, Styracosaurus ovatus Gilmore, 1930 and

Prosaurolophus blackfeetensis Homer, 1992 are found in the Two Medicine

Formation in Montana, whereas Styracosaurus albertensis and Prosaurolophus maximus are found in the Dinosaur Park Formation in Alberta (Horner 1992;

Ryan et al. 2007; Ryan and Evans 2005). An exception is Albertaceratops nesmoi Ryan, 2007, which is found in both the lower part of the Oldman

Formation and the lower part of the Judith River Formation (Ryan 2007). Lehman

(2001) has suggested that omithischians had narrow geographical and ecological ranges during the Late Cretaceous of North America. Again, if Euoplocephalus tutus is the only ankylosaurid in the Campanian of Montana and Alberta, this raises interesting palaeobiogeographical questions. Penkalski (2001) noted differences in Euoplocephalus specimens from Montana and Alberta and suggested that they may represent different taxa. This has been neither confirmed nor rejected during the course of this study.

A significant difficulty with referring specimens to Euoplocephalus tutus is the poor quality of the holotype, CMN 0210 (Fig. 2.14). CMN 0210 is a fragmentary skull and a nearly complete first cervical half ring. In addition, the

38 small size and morphology of the skull and half rings suggest this individual was a subadult at the time of death, further complicating the interpretation of ankylosaurid fossils referred to Euoplocephalus. The skull fragment represents the portion of the snout between the orbits and nares. The lateral edges are preserved but the anterior and posterior edges are broken. On the dorsal surface, a large (8.5 cm wide) nasal plate is preserved centred over the midline, surrounded by smaller osteoderms ranging from 2 to 5 cm wide and separated by 3 to 6 mm. Preserved on the ventral surface are the nasal cavities and a partial vomer.

The first cervical half ring in CMN 0210 may provide diagnostic information. Unfortunately, the cervical half rings were not recovered for ROM

780. Ankylosaurid half rings are composed of an underlying band of fused, paired, smooth-textured dermal plates, with varying numbers of keeled osteoderms fused to the dorsal surface. Ankylosaurids have two cervical half rings, with the first half ring smaller than the second half ring. The first half ring of

CMN 0210 has five preserved osteoderms (Fig. 2.14); all half rings have an even number of large osteoderms so it can be assumed that a sixth osteoderm was present in life. These osteoderms form symmetrical pairs, with the medial pair lying closest to the midline, the distal pair covering the distal tips of the band, and the lateral pair in between (terminology after Penkalski 2001). Both distal and lateral osteoderms are preserved, and the right medial osteoderm is missing in

CMN 0210. Each pair of osteoderms has a unique shape (contra Penkalski

2001). Each osteoderm of the medial pair has a wide oval base, and the keel

39 Figure 2.14: Specimens that can be confidently referred to Euoplocephalus tutus, based on first cervical half ring morphology: A) CMN 0210, holotype, skull fragment in dorsal view. Anterior is up; B) CMN 0210, holotype, partial first cervical half ring, posterior view; C) UALVP 31, skull, dorsal view, anterior is up; D) UALVP 31, partial first cervical half ring, anterior view; E) AMNH 5406, first cervical half ring, anterior view. Scale bars equal 5 cm.

40 is aligned anteroposteriorly. Each lateral osteoderm has a sigmoidal keel that cuts diagonally across the band, and has a narrower base than the more medial plate. The distal osteoderms are compressed and highly excavated ventrally.

Penkalski (2001) noted that there was variation in first half ring morphology among specimens referred to Euoplocephalus, and focused on the variable number (four versus six) of osteoderms on the underlying band. Osteoderm shape is more diagnostic, and only a few specimens (AMNH 5406 and UALVP

31) can at present be confidently referred to Euoplocephalus tutus.

AMNH 5406 includes a complete, although distorted, first cervical half ring

(Fig. 2.14), and a fragment of the second half ring, as well as a scapula, , , , pes, and several vertebrae. Six osteoderms are present on the first half ring, and match the morphology observed in CMN 0210 (Fig.

2.14). UALVP 31 matches the morphology observed in CMN 0210. UALVP 31

(Fig. 2.14) consists of a relatively complete skull (Gilmore 1923), both cervical half rings, vertebrae, ribs, limb elements, part of the sacrum, the right ilium, and osteoderms. Currently, only the skull and cervical rings have been prepared; the sacrum, ilium, and a scapula are undergoing preparation. When the remainder of the skeleton has been prepared, UALVP 31 will provide additional information on the morphology of Euoplocephalus.

ROM 784 appears to be distinct from all other referred specimens of

Euoplocephalus tutus (Arbour et al., in review), with significant differences in the pelvis compared to all other referred specimens. ROM 784 has caudal vertebrae with the most common morphology (circular to subcircular centra, transverse

41 processes projecting anterolateral^ from the centrum and forming acute angles ventrally where they meet the centrum) among specimens referred to

Euoplocephalus, but differs from AMNH 5404 and CMN 8530. Like all cf.

Euoplocephalus, ROM 784 has handle vertebrae with sharply pointed neural spines. However, ROM 784 has a small, elongate tail club knob, with sharp keels on the major plates. Most significantly, ROM 784 differs from Euoplocephalus in having anterolateral^ projecting, and somewhat ventrally directed sacral ribs on the ?second sacral vertebra, which forms a butterfly-like arrangement of the sacral fenestrae, and in having ischia that meet the ilia at right angles. Arbour et al. (in review) have removed this specimen from Euoplocephalus tutus. Currently, there are no sacral vertebrae or pelves that can be definitely referred to

Euoplocephalus, although a sacrum from UALVP 31 was collected and is undergoing preparation. There is no overlap among AMNH 5406, CMN 0210, and ROM 784. It is possible that ROM 784 and UALVP 31 may share similar pelvic morphology, which would mean that ROM 784 could be confidently referred to Euoplocephalus. In this case, AMNH 5245, AMNH 5337, AMNH 5409,

AMNH 5470, and TMP 1982.9.3 would represent distinct species or genera from

Euoplocephalus tutus, and Dyoplosaurus acutosquameus would once again be a junior subjective synonym of Euoplocephalus. Regardless, ROM 784 represents a distinct taxon from all other North American ankylosaurids as they are presently known, and should not be synonymized with Euoplocephalus until that synonymy can be demonstrated. Dyoplosaurus acutosquameus should therefore be considered as a valid taxon until the genus Euoplocephalus can be revised with

42 additional information from AMNH 5406 and UALVP 31. At least two species,

Dyoplosaurus acutosqumaeus and Euoplocephalus tutus, are represented by material currently referred to Euoplocephalus tutus.

2.5.2 Causes of tail club variation

Coombs (1995) suggested that the difference between ROM 784 and other clubs referred to Euoplocephalus could be related to taxonomic variation, sexual dimorphism, ontogeny, or individual variation. It is difficult to test for evidence of sexual dimorphism in extinct vertebrates, and the presence of slender knobs across several taxa does not currently support the hypothesis that slender knobs represent a sexual dimorph of Euoplocephalus or any ankylosaurid taxon. Growth in ankylosaurids is poorly understood at present, making it difficult to test for evidence of ontogeny in tail clubs. Juvenile

Pinacosaurus lack tail club osteoderms completely (Currie 1993). Tail club knob size does not appear to be associated with smaller vertebrae in specimens referred to Euoplocephalus, although the sample size is small. It is possible that slender tail clubs represent immature individuals. ROM 784 represents a relatively small individual, compared to many Euoplocephalus specimens, and has a slender tail club, but it is unknown whether Dyoplosaurus was a smaller- bodied taxon, or if ROM 784 was simply not a large, mature adult. Interestingly, although AMNH 5214 has large, robust handle vertebrae, the centra are unfused, while in smaller ankylosaurids such as Dyoplosaurus, the centra are fused in the handle. Some amount of variation is undoubtedly due to individual variation, but

43 this is difficult to quantify or describe, because every tail club knob is unique in some way, be it shape, texture, number of minor plates, or a combination thereof.

At present, it is not possible to determine whether or not tail club knob shape is of taxonomic value, although there is some evidence that knob size may be linked to ontogeny. It is possible that knob shape is too plastic to be used as a taxonomic indicator. If knob shape is a useful taxonomic character, then UALVP

47273 could also be referred to Dyoplosaurus.

2.5.3 Evolution of the ankylosaurid tail club

No tail clubs have been recovered for several basal members of the

Ankylosaurinae and it is unclear at what time the tail club evolved. Table 2.3 summarizes data from Vickaryous et al. (2004), and includes the age ranges of taxa included in their phylogenetic analysis. Gastonia and Gargoyleosaurus, the oldest ankylosaurids, lack tail clubs, and a tail club is unknown in Minmi.

Shamosaurus and Tsagantegia are known from cranial material only, and the postcranium of Gobisaurus has not been described. Gobisaurus and

Shamosaurus form the outgroup to the Ankylosaurinae, all of which are known to have had tail clubs, except for Nodocephalosaurus and Tsagantegia for which only cranial and fragmentary postcranial material is known. Based on the current phylogenetic and stratigraphic information for the Ankylosauridae, tail clubs must have evolved sometime between the Hauterivian and late Santonian (Table 2.3).

Tianchiasaurus nedegoapeferima Dong, 1993 is a Middle Jurassic ankylosaur from the Junggar Basin, China. Dong (1993) interpreted small, flat plate as a tail

44 club knob, which would make this the specimen with the earliest tail club. Based on the photograph provided (Dong 1993, Plate IV-C), this specimen bears little resemblance to any other known tail club knobs or partial knobs, and, lacking remnants of a handle, cannot confidently be identified as a knob. Therefore, there are no known ankylosaurid tail clubs from the Jurassic. Additional postcranial material from Gobisaurus or Shamosaurus could provide interesting new information on the timing of tail club evolution. Did tail clubs appear suddenly in ankylosaurids, or was there a gradual shift towards fused, interlocking vertebrae and larger terminal osteoderms? Among ankylosaurids with tail clubs, there do not seem to be any trends in tail club evolution. Despite the range of morphologies exhibited by tail club osteoderms, overall tail club morphology is relatively conservative, with most taxa exhibiting similar features in the handle vertebrae, as well as two large lateral knob osteoderms, and several small terminal knob osteoderms. Tail clubs do not become noticeably larger throughout the history of ankylosaurids, and most tail club size variation is probably a result of ontogeny.

Blows (2001) suggested that ossified tendons in Polacanthus may have indicated modification of the tail for swinging a tail-end weapon. The presence of ossified tendons is basal for omithischians (Sereno 1986), and they are present in basal thyreophorans such as and (Norman et al.

2004). Ossified tendons are present in the basal stegosaur taibaii Dong, Tang, and Zhou, 1982 (Maidment et al. 2006), but have been lost in

45 Table 2.3 Tail club distribution within Ankylosauridae, in order of increasing stratigraphic level. Taxon Tail club Range, from Vickaryous et al. (2004) present/absent Gargoyleosaurus Absent Kimmeridgian-Tithonian Gastonia Absent Berriasian-Hauterivian Minmi Unknown Aptian/Albian Gobisaurus Unknown Aptian/?Albian Shamosaurus Unknown Aptian-Albian Tsagantegia Unknown Cenomanian-Santonian P. grangeri Present ?late Santonian-?middle Campanian P. mephistocephalus Present ?late Santonian or early Campanian Tianzhenosaurus Present Late Cretaceous Talarurus Present Cenomanian-Campanian Tarchia Present ?late Campanian-early Maastrichtian Saichania Present ?middle Campanian Euoplocephalus Present late Campanian-early Maastrichtian Nodocephalosaurus Unknown late Campanian or early Maastrichtian Ankylosaurus Present late Maastrichtian stegosaurids (Sereno 1986, Maidment et al. 2008), except for Stegosaurus mjosi

Maidment et al. 2008 (= mjosi, Carpenter, Miles, et Cloward,

2001). This was considered a reversal by Maidment et al. (2008), whereas

Carpenter et al. (2001) argued that the absence of ossified tendons could not be used to define the . Whereas ossified tendons most likely aided in tail club swinging, they cannot be used as evidence for this behaviour, because ossified tendons are found throughout the Ornithischia. Stegosaurs are also hypothesized to have swung their tails defensively (Carpenter et al. 2005), but they generally lack ossified tendons.

Basal ankylosaurids ("polacanthids") such as Gastonia are noted for having plate-like spines, or "splates" (terminology sensu Blows 2001), which were present along the lateral edges of the tail (Fig. 2.15). Several ankylosaurids have similar arrangements, with wedge-shaped, keeled lateral caudal osteoderms present in Pinacosaurus, Saichania, Tarchia, and possibly 46 Dyoplosaurus. The major plates of the tail club knob are almost always keeled, and resemble enlarged, dorsoventrally inflated splates. Of particular note is a nearly complete specimen of Saichania (MPC 100/1305) that has triangular, wedge-shaped, keeled osteoderms around the pelvis and anterior portion of the tail (Fig. 2.15). The osteoderms near the terminus (the knob is not preserved) are rounded in dorsal view, similar to those of the knob. The keel on the most distal caudal splates in Gastonia is probably homologous to the keel on the knob osteoderms of more derived ankylosaurids.

Figure 2.15: Dermal armour arrangement in the tails of basal and derived ankylosaurids. A) MPC 100/1305, mounted skeleton of Saichania, oblique lateral view. Splate osteoderms were preserved as mounted associated with the pelvis and caudal vertebrae. Anterior is to the left. B) DMNH 45577, Gastonia, dorsal view of block with four caudal vertebrae and associated splate osteoderm and ossified tendons. Abbreviations: c, centrum; f, ; fc, free caudal vertebra; il, ilium; os, osteoderm; ot, ossified tendon.

47 2.6 Conclusions

Ankylosaurid tail clubs appear suddenly in the record, between the

Hauterivian and late Santonian, and at present are known only in the

Ankylosaurinae. Tiachiasaurus, from the Middle Jurassic, did not bear a tail club.

Polacanthus, a basal ankylosaurid or polacanthid, also lacked a tail club.

Mymoorapelta is possibly a basal ankylosaurid and has interlocking distal caudal vertebrae that may represent the early of handle evolution. Knob osteoderms are homologous with the most distal wedge-shaped lateral caudal osteoderms found in basal ankylosaurids such as Gastonia, based on the presence of a laterally-oriented keel in both types of osteoderms.

Knob osteoderms are highly variable in shape and size, but no trends are apparent through time and most tail club size variation is probably a result of ontogeny. Large knobs are not correlated with larger individuals, which suggests that the knob may have developed when the was near adult size.

Overall tail club morphology is conservative through time, with most taxa exhibiting similar features in the knob and handle. However, there are distinct differences in the free caudal vertebrae of different taxa, such as the shape of the centra, the direction the transverse processes project from the centra, the nature of the contact between the transverse processes and the centra, and the number of vertebrae that bear transverse processes. Some ankylosaurid taxa can be differentiated by the shape of the neural spine in dorsal view in the handle vertebrae. Differences in the pelvis and caudal vertebrae of ROM 784 indicate that Dyoplosaurus is a distinct taxon from Euoplocephalus. Inclusion of these

48 postcranial characters would allow the inclusion of many problematic taxa (i.e. most 'polacanthids') in phylogenetic analyses.

49 Chapter 3: Ankylosaurid tail and pelvis pathologies2

3.1 Introduction

Ankylosaurid dinosaurs are notable for their tail clubs, consisting of modified vertebrae that form a rigid handle, and large terminal osteoderms forming a knob (Coombs 1995). It has been assumed that the tail club was used as a weapon in intraspecific antagonistic behaviour, as an interspecific defense against predators, or both (Coombs 1995). Pathologies in ankylosaurid tail elements, including vertebrae of the handle, free caudal vertebrae, chevrons, and the knob osteoderms, may provide evidence of active use of the tail as a weapon in these dinosaurs. Fracture calluses and evidence of infection may point to traumatic injuries sustained during combat. This chapter describes pathological tail elements from several individual ankylosaurids, including pathological free caudal vertebrae, chevrons, and knob osteoderms. Active use of the tail as a weapon may have resulted in more injuries to the tail compared to ankylosaurs

(basal ankylosaurids and nodosaurids) that lacked tail clubs. If this is true, then at least some pathologies associated with the pelvis, caudal vertebrae, and tail club in derived ankylosaurids should be preserved in the fossil record. Fracture calluses and evidence of infection may point to traumatic injuries sustained during combat. Pathologies should be more common in ankylosaurids with tail clubs than in ankylosaurs without tail clubs. Patterns in the types and distributions of pathologies in ankylosaurid tail elements may shed light on the

2 This chapter has been submitted for publication. Arbour, V. Ankylosaur tail and pelvis pathologies. Cretaceous Research, submitted 14 October 2008. 50 use of the tail. A survey of ankylosaurid caudal vertebrae, tail clubs, and pelves resulted in the identification of numerous abnormal elements (Appendix 1).

3.2 Description

3.2.1 AMNH 5409, pelvis and sacrum

AMNH 5409 (Fig. 3.1) is an isolated pelvis and sacrum of Euoplocephalus tutus, and includes a complete right ilium, mostly complete left ilium, complete left ischium, and eight vertebrae forming the sacral rod. The ilia are broad and diverge from the midline anteriorly. The ischia taper distally and are directed ventromedially. No pubes are present. The neural spines of the vertebrae are fused to form a continuous blade. Sacral and dorsosacral ribs project at right angles to the centra, and have flat, horizontal dorsal surfaces. Distally, the sacral • ribs expand anteroposteriorly before contacting the ilia. The first through fourth dorsosacral ribs do not expand distally, and do not fuse as strongly with the ilia as the sacral ribs do. The right fourth dorsosacral does not contact the right ilium.

Where the rib should meet the ilium, there is instead an elongate, oval excavation, 99 mm long by 44 mm wide, which undercuts the dorsal surface of the ilium by a minimum of 67 mm (Fig. 1). The surface bone around the lesion has a slightly ridged texture, with the ridges generally oriented anteroposteriorly.

The distal edge of the dorsosacral rib is rounded and smooth. Anterior to the large lesion is a smaller (approximately 10 mm diameter) lesion that only shallowly excavates the ilium. The internal surfaces of both lesions are smooth.

51 Figure 3.1: Lytic lesions on the pelvis. A) AMNH 5409 in anterodorsal view. B) Detail of erosive lesion on right ilium of AMNH 5409, anterior is down. Second, smaller lesion marked by arrowhead. Note ridged texture surrounding large lesion. C) Detail of erosive lesion on left ilium of AMNH 5337, anterior is up. Scales = 10 cm.

3.2.2 AMNH 5337, pelvis and sacrum

AMNH 5337 (Euoplocephalus) includes a skull, , limb elements, and a well-preserved pelvis and sacrum, including both ilia, both ischia, and the eight vertebrae that form the sacral rod. Overall the pelvis is similar to that of

AMNH 5409, with broad, divergent ilia, and ventromedially directed, tapering ischia. The neural spines are fused, forming a blade. The neural spine of the last vertebra is free. There is a semicircular excavation on the left ilium at

52 approximately the position where the left fourth dorsosacral should meet the ilium

(Fig. 3.1). This excavation deeply undercuts the ilium at its posterior and lateral boundaries. This specimen is on display in a glass cabinet, so no measurements could be taken, but the excavation is at least 100 mm in diameter. The bone surrounding the excavation is smooth. The left and right fourth dorsosacral ribs do not meet the left and right ilia, and the distal tips have the same striated texture as those of the other ribs.

3.2.3 AMNH 5245, two anterior free caudals

AMNH 5245 (Euoplocephalus) includes several dorsal vertebrae, the sacrum, several free caudal vertebrae, part of the handle of the tail club and a large (width = 590 mm) complete knob. AMNH 5245 also includes two fused vertebrae, and the first and second free caudal vertebrae (Fig. 3.2). The first caudal includes a complete centrum, neural arch and spine, prezygapophyses, postzygapophyses, and the left transverse process. The centrum is 102 mm wide by 89 mm tall. The second caudal is represented only by a partial neural spine and prezygapophyses; the centrum is not preserved. The neural spines of the two caudals are fused. Distally, the neural spine of the first caudal is rugose and thickened. The second caudal appears to be offset ventrally from its normal position, based on the position of the prezygapophyses of the first free caudal and the postzygapophyses of the sacrocaudal.

53 Figure 3.2: Rugose blastic lesions of the neural spine and transverse processes. A) TMP 1985.36.70, anterior caudal vertebra, posterior view. The neural spine is swollen distally. B) Detail of TMP 1985.36.70 neural spine, left lateral view. A draining sinus is indicated by the arrowhead. C) Detail of neural spine of TMP 2005.09.75, possibly a first free caudal vertebra. Note the similarity in texture to TMP 1985.36.70. D) AMNH 2062, ?mid free caudal vertebra, anterior view. The left transverse process is dorsoventrally thickened. E) AMNH 5245, two anterior free caudal vertebrae, left lateral view. The neural spines are fused together and there is rugose new bone growth on both. Scales equal 5 cm.

54 3.2.4 TMP 2005.09.75, sacrocaudal

TMP 2005.09.75 (Euoplocephalus) is probably a first free caudal vertebra, based on the flared shape of the distal ends of the transverse processes. It includes a complete centrum, neural arch and spine, prezygapophyses, broken postzygapophyses, and haemal arch and spine. The centrum is 119 mm high and 121 mm wide, and represents a large individual of Euoplocephalus. The neural spines in Euoplocephalus taper distally, but in TMP 2005.09.75, the neural spine is expanded and swollen distally (Fig. 3.2). It has a rugose texture, with numerous small perforations. The rugose area is situated largely on the posterior edge of the neural spine.

3.2.5 TMP 1985.36.70, proximal free caudal vertebra

TMP 1985.36.70 (Euoplocephalus) was collected from a macrovertebrate bonebed in Dinosaur Provincial Park. It is a nearly complete free caudal vertebra, including a complete centrum, neural arch, and haemal arch. The transverse processes are present but broken at the distal termini. The centrum is 121 mm high and 129 mm wide, and likely came from a large individual. Overall, this vertebra bears a striking resemblance to TMP 2005.09.75, with its swollen and rugose neural spine (Fig. 3.2). This rugose texture extends onto the bases of the postzygapophyses, but is absent on the articular faces. Rugosity is also present at the intersection of the neural arch and transverse processes, and the base of the right transverse process is marked by several cloacae (sinuses).

55 3.2.6 AMNH 2062, anterior free caudal vertebra

AMNH 2062 is a cast of the holotype specimen of Heishansaurus pachycephalus Bohlin, 1953, synonymized with Pinacosaurus granger! by

Maryanska (1971). The specimen includes the centrum and partial left transverse process. The left transverse process is expanded and has an irregular surface texture (Fig. 3.2). A large sinus is present on the dorsal surface of the transverse process.

3.2.7 ROM 1930, anterior free caudal vertebra and last free caudal vertebra

ROM 1930 (Euoplocephalus) includes a skull and partial postcranium, including almost all of the free caudal vertebrae. Several of these vertebrae are articulated with osteoderms and ossicles on the left lateral side, while others are disarticulated and free of the matrix. The third largest free caudal vertebra includes the centrum, neural arch, and postzygapophyses. The neural spine, haemal arch, and both transverse processes are missing (the right transverse process has been reconstructed). Centrally located at the base of the posterior face of the centrum is an abnormal swelling of bone (Fig. 3.3). This small projection is approximately 10 mm wide, projects ventrally by approximately 10 mm, and extends 20 mm anteriorly along the ventral surface of the centrum. The swelling has a smooth texture and there are no cloacae.

The last free caudal vertebra includes the centrum, neural arch, prezygapophyses, and part of the neural spine. This vertebra is probably the last

56 Figure 3.3: Exostoses. A) ROM 1930, anterior free caudal vertebra, posterior view. There is an unusual swelling ventrally. B) ROM 1930, last free caudal vertebra, right lateral view. Prezygapophyses from the first handle vertebra are attached to the neural spine of this vertebra. Posteriorly, there is a spicule of bone on the ventral side of the centrum. C) TMP 1992.36.344, posterior free caudal vertebra, oblique right lateral view. A smooth, round bump is visible on the centrum near the base of the neural arch. Scales equal 5 cm.

free caudal in the series, because it has fragments of the broad, elongate,

flattened prezygapophyses of the first handle vertebra surrounding the neural

arch (Fig. 3.3). The haemal arch and transverse processes are not preserved. A

spicule of bone is located at the base of the caudal face of the centrum, on the

right lateral edge. This spicule, unlike that observed in the proximal caudal

vertebra, is distinct from the centrum. It is 20 mm long, 15 mm tall, and

approximately 10 mm wide. The spicule projects beyond the posterior border of

the centrum. It has a smooth texture.

57 3.2.8 TMP 1992.36.344, posterior free caudal vertebra

TMP 1992.36.344 (Euoplocephalus) is an isolated posterior free caudal vertebra, based on its overall size and the relatively small neural and haemal spines. The vertebra includes the centrum, complete neural arch, and complete haemal arch, and is missing the transverse processes. Lateral to the right side of the neural arch, and on the dorsal aspect of the centrum, is a 10 mm long prominent oval bump with sharp borders (Fig. 3.3).

3.2.9 AMNH 5404, two free caudal vertebrae

AMNH 5404 (Euoplocephalus) includes a skull, first cervical half ring, limb bones, ribs, and vertebrae. Five caudal vertebrae are preserved, two of which show evidence of pathologies. The largest preserved free caudal is 131 mm wide by 102 mm tall, and includes the centrum, and neural arch. There is a laterally- oriented excavation on the ventral edge of the posterior face of the centrum (Fig.

3.4).

The second-largest preserved vertebra is 123 mm wide by 87 mm high, and includes a centrum, neural arch, and the bases of both transverse processes. There are large excavations between the transverse processes and the centrum, and on the ventral side of the centrum - overall, the centrum has the appearance of 'caving in' on itself (Fig. 3.4). The excavation between the centrum and left transverse process undercuts the centrum and distorts its outline in anterior view. The excavation on the ventral side is more pronounced anteriorly. Exposed trabeculae show no evidence of remodeling. The notochordal

58 Figure 3.4: Metastatic cancer in AMNH 5404. Largest preserved caudal in A) anterior and B) ventral views. Note cavities under transverse processes, along lateral sides of centrum, and on the ventral aspect. C) Detail of the lytic lesion under the left transverse process and 'puckered' appearance nearby on the centrum. D) Second largest preserved caudal in posterior view; note the lytic lesion near the ventral side of the centrum. Scales in A, B and D equal 5 cm, scale in C equals 3 cm. prominence is an unusual, elongate, dorsoventrally aligned oval bump (unlike the more common approximately circular bump found in many ankylosaurid caudal vertebrae). The external bone surface of the unaffected areas has a typical, smooth texture.

59 3.2.10 ROM 1930, haemal arch

Loose chevrons not associated with vertebrae are preserved in ROM

1930. The largest of these is deformed and clearly pathological in nature (Fig.

3.5). All of the other preserved chevrons are Y-shaped in cranial and caudal views, with tapering, blade-like haemal spines. The pathological chevron has a thickened, distally expanding haemal spine, and is bulbous and rounded in proximal view. The surface texture is relatively smooth, and there are no obvious large sinuses, although proximally there is a series of deep furrows.

Figure 3.5: ROM 1930 haemal arches. A) Pathological haemal arch, anterior view. B) Normal haemal arch, anterior view. C) Same pathological haemal arch as in A, left lateral view. D) Same normal haemal arch as in B, left lateral view. Scale equals 5 cm.

3.2.11 TMP 1983.36.120, tail club

TMP 1983.36.120 is a partial Euoplocephalus tail club collected from Dinosaur

Provincial Park (Fig. 3.6). TMP 1983.36.120 consists of a wide and massive

60 A , _ . _ __ B C

Figure 3.6: Pathologies of the tail club knobs. A) ROM 788 in oblique left lateral view. Note the subparallel furrows dipping to the right. Examples of more typical Euoplocephalus knob textures: B) CMN 2252, and C) CMN 2253. D) TMP 1983.36.120 in posterior view. The right major plate is taller than the left major plate; the lesion is at the apex of the swelling. E) TMP 1983.36.120 in dorsal view, showing the discrete lytic lesion on the right major plate. Scale in A, D, and E equals 10 cm, in B and C equals 2 cm.

tail club knob, measuring 42 cm in width, with at least eleven osteoderms (two major plates and at least nine minor plates). The major plates are the largest osteoderms in the knob and are positioned laterally. These are approximately symmetrical in dorsal view. The minor plates make up the distal, posterior portion of the tail club and have a variety of shapes and sizes. Some of the tail club vertebrae are preserved in this specimen, although they are mostly obscured by numerous thick ossified tendons.

61 The right major plate of TMP 1983.36.120 has an erosional pit on the dorsal surface near the lateral edge. An area of bone 20 mm wide by 25 mm longhas been excavated at least 20 mm into the osteoderm. The erosion undercuts the dorsal surface of the osteoderm. The interior surface of the erosion is smooth, remodeled bone. The surface texture on the left major plate and much of the right major plate is bumpy. Fine pitting is present on the outer surface near the large excavation, although it is unclear if this is due to pathology or taphonomic processes. Although the left and right major plates appear symmetrical in dorsal view, the right major plate is much taller in posterior view.

In posterior view, the right major plate has a conical appearance with the erosion at the tip of the cone on the dorsal surface, which contrasts with the essentially flat dorsal surface of the left major plate.

3.2.12 ROM 788, tail club

ROM 788 is a nearly complete ankylosaurid tail club, consisting of a complete knob and most of the handle vertebrae (Fig. 3.6). The knob is massive and measures nearly 590 mm at the widest point; the left major plate is 225 mm wide, while the right major plate is 163 mm wide. Individual minor plates are difficult to distinguish at the distal end. The specimen has undergone some taphonomic deformation; the handle vertebrae are crushed laterally and some of the asymmetry in the major plates may be due to crushing.

The left major plate has a highly unusual surface texture. The right major plate and the medial part of the left major plate have a bumpy texture. Towards

62 the lateral edge of the dorsal side of the osteoderm, and continuing onto the lateral side, are large, deep furrows and ridges. Several of the furrows are close to 1 cm deep. The furrows are approximately parallel, although they occasionally branch and merge, and some extend for more than 10 cm. The furrows do not seem to correspond to any internal cavities within the osteoderm. A computed tomography (CT) scan of this specimen did not show any obvious internal abnormalities, and unfortunately the furrows on the left major plate were outside the field of view of the scanner.

3.3 DISCUSSION

Paleopathology, the study of ancient disease, is frequently limited to the examination of pathologies of the skeleton. Pathological changes in the skeleton can result from mechanical stress (trauma), changes in blood supply, inflammation of bony or surrounding soft tissues, infection, hormonal/nutritional/metabolic upsets, and tumours (White 2000). Few diseases produce changes in the skeleton, and lesions (areas of damage) caused by different processes can look similar. Although disease of the human skeleton has been relatively well studied, both in clinical and anthropological settings, there are still several diseases (for example, brucellosis and sarcoidosis; Rothschild and Martin 2006) known to affect the human skeleton that have not been observed in defleshed skeletons. Furthermore, pathologies in most nonhuman vertebrates have received comparatively little attention, with few examples of skeletal pathology outside of veterinary and agricultural settings. Because of

63 these challenges, the study of paleopathology in dinosaurs is necessarily limited in the precision with which diagnoses can be made, and in the palaeobiological interpretations of these diagnoses.

Excellent discussions of skeletal pathologies in humans include Steinbock

(1976), Aufderheide and Rodriguez-Martin (1998), and Resnick (2002), and pathologies encountered in veterinary practice are discussed by McGavin and

Zachary (2007). Rothschild and Martin (1993; 2006) discuss paleopathologies in the fossil record of both humans and other vertebrates. White (2000) divides pathologies into six categories: trauma, infectious disease, circulatory disorders, metabolic disorders, tumours, and arthritis. Lytic lesions are those in which bone has been eaten away, while blastic lesions are those in which extra bone has been deposited. The specimens described in this study can be separated into those with lytic lesions (AMNH 5337, AMNH 5404, and AMNH 5409), those with blastic lesions (AMNH 5245, ROM 788, ROM 1930, TMP 1985.36.70, TMP

1992.36.344, and TMP 2005.09.75), and those with a combination of the two

(TMP 1983.36.120). A summary of the possible diseases responsible for the abnormalities in this study is provided in Table 1.

3.3.1 Diagnoses

No specimen in this study shows evidence of fluoride poisoning, or anything resembling an osteoma, or the 'sunburst' appearance of osteogenic sarcoma. Diffuse osteopenia was not observed in any specimens, which excludes iron deficiency, sickle-cell anemia, thalassemia, leukemia, and

64 osteoporosis. Degenerative joint disease and rheumatoid arthritis only affect the diarthrodial joints. Gout is not commonly found in the vertebrae, and no specimens in this study have an ivory-like texture such as that found in gout.

Vertebrae of the handle in some ankylosaurids appear to be fused at the centra.

However, this does not appear to be the result of pathological processes such as

DISH, ankylosing spondylitis, or vertebral collapse, as these conditions tend to result in exuberant new bone growth or vertebral collapse. Tumours, infections, and trauma remain as possible explanations for the various abnormalities observed in ankylosaurid tails.

3.3.1.1 AMNH 5245. AMNH 2062. ROM 1930 haemal arch and spine, TMP

1985.36.70. and TMP 2005.09.75: osteomyelitis and periostitis

Five specimens (AMNH 2062, AMNH 5245, ROM 1930 haemal arch, TMP

1985.36.70, and TMP 2005.09.75) have reactive new bone growth with a rugose, perforated appearance. None of the new bone resembles neoplasias, such as osteomas, osteogenic sarcomas, or osteochondromas. The rugose bone textures on these specimens most likely represent either osteitis or periostitis. Osteitis is an inflammation of the cortex and periostitis is inflammation of the periosteum and do not need to be caused by infectious agents (Resnick 2002). The rugose textures are consistent with osteomyelitis and treponemal infections.

Treponemal infections are caused by spirochaete bacteria, and produce widespread periostitis throughout the skeleton (Steinbock 1976). However, treponemal infections rarely affect the vertebral column, and the modifications

65 Table 3.1: Causes of blastic and lytic lesions. Symptoms summarized from Steinbock (1976), Rothschild and Martin (1993), Aufderheide and Rodriguez-Martin (1998), Resnick (2002), Rothschild and Martin (2006), and McGavin and Zachary (2007). Disease Blastic or Disease Skeletal Characteristics category Lytic Arthritis Blastic Diffuse idiopathic Dripping candle wax appearance, ossification of spinal skeletal ligaments, intervertebral disk space spared. hyperostosis (DISH) Blastic Ankylosing Calcification of anulus fibrosus, bamboo-like appearance, spondylitis apophyseal joints fuse. Lytic Degenerative joint Affects only diarthrodial joints. disease

Lytic Rheumatoid Affects only diarthrodial joints. arthritis

Lytic Schmorl's nodes Found only on vertebral endplates.

Lytic Gout Dense reactive new bone with ivory-like texture. Discrete "scooped out" lesions with overhanging edges. Most commonly found in metatarsals and metacarpals Tumours Blastic Osteomas Smooth, small, dense, rounded blastic lesions, most commonly found on skull. Blastic Osteogenic Spiculated bone, "sunburst" effect. sarcoma Blastic Osteochondroma Projection from bone surface with core of trabecular bone. Lytic Multiple myeloma Sharply localized dissolution of bone, lesions are 5-20 mm and occasionally coalesce. No new reactive bone. Lytic Chondroblastomas Only found at sites of endochondral ossification.

Lytic Giant cell tumours Usually only affects large tubular bones. Lytic Chondrosarcomas Lobulated endosteal appearance. Lytic Metastatic cancer Lesions have "space-occupied' appearance. Circulatory Lytic Iron deficiency Diffuse osteopenia, expansion of marrow spaces. Disorders Lytic Sickle-cell anemia Diffuse osteopenia, expansion of marrow spaces.

Lytic Thallasemia Diffuse osteopenia, expansion of marrow spaces.

Lytic Leukemia Diffuse osteopenia, lytic lesions, onion skin appearance of diaphyseal bone. Metabolic Lytic Osteoporosis Porous cancellous bones, vertebral collapse, long bone Disorders fractures. Lytic Fluoride poisoning Broad, flat sheets of periosteal bone, localized nodular new (fluorosis) bone, narrowing of the medullary cavity and spinal canal. Infectious Blastic and Pyogenic Irregular, pitted surface. Draining abscesses and sinuses. Disease Lytic osteomyelitis Involucrum of reactive bone around original bone. Blastic Periostitis Thickening of cortical bone, irregular surface texture. Lytic Tuberculosis Little to no reactive bone, lesions on vertebrae confined to centrum, vertebral collapse, no joint fusion. Lytic Actinomycosis Predominantly affects the jaws. Lytic Treponemal Difficult to tell apart from other bacterial infections. Widespread disease (syphilis, periostitis, "saber shin" deformity, and draining sinuses. pinta, yaws, bejel) Brucellosis Nodular, reniform reactive bone. [Note: The skeletal impact of brucellosis is not known in humans, but has been observed in moose (Alces alces) (Honour and Hickling 1993)] Lyuu Fungal infections Lesions have sharp boundaries, "punched out" appearance. Occasional periosteal reactive bone. Trauma Blastic Fractures Fracture callus, offset bones.

66 are usually slight (Steinbock 1976). Osteomyelitis is the infection of bone and marrow (Resnick 2002). Pyogenic or suppurative osteomyelitis is an inflammation of bone and marrow caused by pus-producing bacteria, and can be acute, subacute, or chronic in nature (Aufderheide and Rodriguez-Martin 1998).

Bone may become infected directly through trauma, via the spread of a nearby infection in the soft tissues, or haematogenously (Resnick 2002). Rothschild and

Martin (2006) refer to all bony infections as osteomyelitis, and subdivide it into pyogenic osteomyelitis and non-pyogenic osteomyelitis. Non-pyogenic osteomyelitis includes tuberculosis, other mycobacterial diseases, brucellosis, and fungal diseases. It should be noted that turtles, squamates, crocodilians, and birds do not produce liquid pus such as that found in mammals; instead, a hard, lumpy, caseous ("cheeselike") substance called a fibricess, rather than an abscess, is formed (Hutchzermeyer and Cooper 2000). Birds and crocodiles are the nearest extant relatives of dinosaurs, so it is reasonable to assume that dinosaurs may have employed a similar inflammatory response.

Aufderheide and Rodriguez-Martin (1998) describe the main features of osteomyelitis as well as the forms it can take. Osteomyelitis involves the destruction of bone as well as reactive new bone formation, forming bone with irregular, pitted surfaces. Identifying features include abscesses and draining sinuses, and an involucrum of reactive bone around the original bone (see infected Ursus americanus bone, Fig. 3.7). Osteomyelitis most frequently affects the long bones. Bacteria may enter bones directly through an open wound, or may originate from elsewhere in the body and travel via the bloodstream to

67 Figure 3.7: Examples of blastic and lytic lesions in extant vertebrates. A) Osteomyelitis in an Ursus americanus (black bear) right and , anterior view. The shaft of the tibia is greatly expanded and fused to the fibula. The arrow marks a large draining sinus, and the arrowhead marks an enthesophyte. Note the porous, irregular surface texture. B) A human rib, with a tumour. Cancellous bone is preferentially destroyed, undercutting the bone surface. Little remodeling of the trabeculae is evident. C) A human innominate with a lytic lesion caused by tuberculosis. The lesion is large, and cortical and cancellous bone are equally affected. Trabeculae are smoothed. Scales equal 5 cm. cause hematogenous osteomyelitis. A Brodie's abscess is a form of osteomyelitis that can only be diagnosed radiologically, because the abscess is covered by normal bone. Sclerosing osteomyelitis of Garre is a fusiform thickening of the cortex of a long bone. Periostitis is similar to osteomyelitis, but is not itself a disease; it is inflammation of the periosteum and involves only the cortical bone

(White 2000). Periostitis may result from an infection, but can also be caused by trauma, and produces an irregular bone surface (Aufderheide and Rodriguez-

Martin 1998). Injury of the periosteum can lead to nodular bony osteophytes, or

68 bony enthesophytes (Fig. 3.7) representing new bone at the insertions of tendons and ligaments (McGavin and Zachary 2007).

Aufderheide and Rodriguez-Martin (1998) note that pyogenic osteomyelitis usually results in bone with an irregular, pitted surface, with abscesses and cloacae, and an involucrum of reactive bone around the original bone. Normal, smooth-textured bone is visible near the affected areas of all four specimens

(Fig. 3.2). Periostitis can also produce irregular bone surfaces, but only involves the cortex (Aufderheide and Rodriguez-Martin 1998). It is difficult to determine on each specimen whether only the cortex is affected. The haemal spine of ROM

1930 is greatly expanded and furrowed, but bears no cloacae (Fig. 3.5). It is possible that this specimen represents a case of periostitis, rather than pyogenic osteomyelitis. Rothschild et al. (2005) described expanded limb bones in a duck that appear similar to the condition in ROM 1930. This widening was interpreted as either a neoplasm or infection.

Bone fractures heal by way of a callus, a fibrous mass that surrounds the broken ends of bone and is calcified later in the healing process (Aufderheide and Rodriguez-Martin 1998). Fracture calluses are resorbed overtime and may eventually disappear altogether (Aufderheide and Rodriguez-Martin 1998). If the affected bones had been fractured and the rugose new bone represents a fracture callus, some offset in the affected areas would be expected, but this was not observed.

69 3.3.1.2 ROM 1930, proximal free caudal vertebra and last free caudal vertebra

Two vertebrae of ROM 1930 have spicules of bone on the ventral sides of the centra, and TMP 1992.36.344 has a prominent bump on the right dorsolateral side of the centrum. These are similar to spicules reported on a distal caudal vertebra of Majungasaurus (Farke and O'Connor 2007). Farke and O'Connor

(2007) suggested that it could be developmental in origin, or an enthesopathy.

The spicules in ROM 1930 and bump in TMP 1992.36.344 are similar, and also probably represent developmental abnormalities or enthesopathies.

3.3.1.3 AMNH 5404. AMNH 5337, and AMNH 5409: metastatic cancer

Hershkovitz et al. (1998) provide a differential diagnosis for differentiating between blastomycosis (a fungal disease), tuberculosis (a bacterial disease), and cancer. The pathological vertebrae of AMNH 5404 (Fig. 3.4) share several characteristics with cancer and tuberculosis, and there is little evidence to support the interpretation that the lesions were caused by a fungal infection. In blastomycosis, lesions do not coalesce, as they do in AMNH 5404. AMNH 5404 shares almost all characteristics with cancer, including expansile, coalescent lesions, no trabecular remodeling, no reactive surrounding bone, and unequal dissolution of cortical and trabecular bone (see example in a human rib, Fig. 3.7).

Intriguingly, only the centrum appears to be affected, which is a characteristic of tuberculosis. Unlike tuberculosis, the subjacent trabecular bone shows no evidence of 'smoothing' (see example in a human innominate, Fig. 3.7). The

70 most likely diagnosis for the unusual condition in AMNH 5404 appears to be cancer.

Rothschild et al. (1998) examined characteristics to differentiate metastatic carcinoma, multiple myeloma, and leukaemia (cancers that can affect the skeleton). Multiple myeloma lesions are characterized by spheroid geometry

(Rothschild et al. 1998). Lesions of metastatic cancer frequently have an elliptical or irregular shape, although circular lesions are possible. Unlike multiple myeloma, in metastatic cancer the lesions frequently retain a shell of cortical bone. Lesions are generally much smaller in leukemia than in multiple myeloma or metastatic carcinoma. Using the criteria in Rothschild et al. (1998), AMNH

5404 appears to represent a case of metastatic cancer: the lesions retain a shell of cortical bone, and have irregular shapes (Table 3.2).

Table 3.2: Differentiating lesions caused by blastomycosis, tuberculosis, and cancer; modified from Hershkovitz et al. (1998). Lesion character Blastomycosis Tuberculosis Cancer AMNH 5404 Expansile Present Absent Present Present Fronts of resorption Present Absent Present ? Zones of resorption Absent Present Absent ? Trabecular remodeling Minimal Minimal Absent Absent Smoothing/remodeling Absent Occurs Absent Absent of subjacent trabecular Equal dissolution of Yes No No No cortical and trabecular bone Reactive surrounding Slight Slight Absent Absent bone Coalescence of lesions Absent Frequent Occurs Present Affects posterior Yes No Yes No vertebral elements

Using this differential diagnosis, the lesions on the ilia of AMNH 5337 and

AMNH 5409 (Fig. 3.1) may also represent metastatic cancer. These lesions are expansile, do not appear to involve trabecular remodeling or smoothing (although

71 this is difficult to confirm in AMNH 5337), and cancellous bone is more greatly affected than cortical bone. In AMNH 5409, there appears to be some reactive bone surrounding the lesion, which is not usually associated with cancer. Small ridges around the lesion may represent periostitis. These ridges do not appear to be present in AMNH 5337.

Centrosaurines and chasmosaurines occasionally have large extra fenestrae and shallow, 'punched out lesions' on the frill (Tanke and Farke 2007).

These bear some resemblance to the smooth lesions on the Euoplocephalus pelves, although the pelvic lesions are more irregular in shape. Tanke and Farke

(2007) did not find any support for the interpretation that these represent traumatic injuries. Punched out lesions also differ from most lesions resulting from infection, because the punched out lesion does not penetrate the cortex and always has a smooth texture. Squamosal fenestrae appear to be related to bone resorption at thin areas, and not to infection or trauma; punched out lesions also do not seem to be due to trauma, but may represent an unknown disease process with no modern analogues (Tanke and Farke 2007). It is also possible that the lesions on AMNH 5337 and AMNH 5409 result from resorption of unnecessary bone - ankylosaurid pelves are massive elements compared to those of other groups of dinosaurs.

72 3.3.1.4 ROM 788, TMP 1983.36.120, and the challenges of identifying pathologies in osteoderms

Osteoderms are found in many unrelated extinct and extant tetrapods, including turtles, aetosaurs, phytosaurs, crocodilians, thyreophoran dinosaurs, titanosaurid dinosaurs, some theropod dinosaurs, some squamates, and cingulatan mammals. Despite their wide distribution among extant tetrapods, the effects of disease on osteoderms have not been well documented. Typical ankylosaur osteoderms can have a variety of textures (Ford 2000; Burns in press); many are pitted and rugose and superficially resemble bones with pathological conditions in other taxa. Ankylosaur osteoderms also have a wide variety of shapes, both on a single individual animal and between different species (Coombs 1995; Ford 2000; Burns, in press). No ankylosaur skeletons are preserved with the complete set of osteoderms, so the typical diversity of morphologies is unknown in most ankylosaur taxa. Osteoderms with unusual or unique morphologies may be interpreted merely as normal variation. As such, diseased ankylosaur osteoderms, such as those of the tail club knob, can be difficult to distinguish from normal osteoderms.

The impact of disease on osteoderms has not been well documented in extant organisms, which makes interpretation of the features seen in ROM 788 and TMP 1983.36.120 difficult. Hutchzermeyer (2003) lists several diseases that affect the skin and scales of farmed crocodilians (alligator pox, caiman pox, various fungal infections), but no mention is made of the impact of these diseases on the underlying osteoderms. Wild and domestic turtles can suffer

73 from a variety of diseases that affect the dermal bones of the shell. Shell rot affects the carapace or plastron, or both, and is characterized by erosive lesions.

It is caused by injury to the shell and poor environmental conditions leading to bacterial or fungal infections, whereas poor nutrition and infection can predispose turtles to shell rot (Rosskopf and Shindo 2003). Garner and colleagues (1997) reported severe necrosis of the shell (both keratin and bone) in two species of wild turtles, but the cause was unknown. No reports were found of squamates with diseased osteoderms.

There are a few reports of dermal armour pathologies in the fossil record.

Lucas (2000) described pathological osteoderms from the aetosaur

Paratyphothorax. Three osteoderms were fused into a single mass by a rough, irregular bony overgrowth. Several large sinuses were present in the bony overgrowth. Lucas (2000) interpreted this pathology as an infection resulting in new bone growth. Sawyer and Erickson (1998) surveyed a suite of bones belonging to numerous individuals of the crocodilian Leidyosuchus. Numerous osteoderms had unusual pitting and canals, along with new bone growth and thickening. These were interpreted largely as periosteopathy or osteopathy, which Sawyer and Erickson (1998) defined as reactive hypertrophy of the periosteum and subperiosteal bone, respectively. Necrotizing dermatitis was postulated as a possible cause for erosion of an osteoderm referred to

Diplocynodon (Wolff et al. 2007). The osteoderm showed loss of surface texture, and Wolff et al. (2007) suggested that a fungal infection in the overlying skin spread downwards onto the osteoderm, eroding the surface ornamentation.

74 McWhinney et al. (2001) described Stegosaurus tail spikes with reactive new

bone growth, ridges and furrows, sinus tracts, and surfaces with a pitted or filigree appearance. McWhinney et al. (2001) concluded that infection began

after the tail spikes were fractured, which resulted in chronic osteomyelitis.

The unusual morphology of TMP 1983.36.120 (Fig. 3.6) may be a result of

individual variation or pathology. Other tail club osteoderms referred to

Euoplocephalus have a variety of textures and morphologies, and there is always

some amount of asymmetry between the two major plates. Surface texture of tail

club osteoderms can be smooth to dendritic (ROM 784, ROM 7761), although

typically they are rugose and bumpy (CMN 2252, CMN 2253, UALVP 16247; Fig.

3.6). Although surface textures vary, none of these tail clubs have such large, well-defined pits. The presence of a single large lesion, and the distinctly

asymmetrical morphology of TMP 1983.36.120 differs from that observed in other

Euoplocephalus tail clubs, and does not appear to be related to individual

variation. The lesion on TMP 1983.36.120 includes reactive new bone growth

and an approximately circular, well-defined lytic lesion. The new bone growth and

cloaca are indicative of pyogenic osteomyelitis. The remodeled bone on the

interior of the lytic lesion excludes the possibility that the erosion is due to

postmortem damage. TMP 1983.36.120 differs from other specimens with

pyogenic osteomyelitis in that the reactive new bone does not differ significantly

in texture from the rest of the osteoderm. This may be related to the unusual and

irregular surface texture already present in Euoplocephalus osteoderms. This is

75 the first report of an ankylosaur osteoderm with evidence of pyogenic osteomyelitis.

The unusual furrowed texture of ROM 788 (Fig. 3.6) may or may not be pathological in nature. The absence of cloacae suggests that the furrows are not a result of pyogenic osteomyelitis, and instead this may represent a case of periostitis. In crocodilians, osteoderms lack a periosteum (Vickaryous and Hall

2008); mammalian osteoderms eventually develop a periosteum (Vickaryous and

Hall 2006). If ankylosaur osteoderms also lacked a periosteum, inflammation of the periosteum would be impossible. The cause of the furrows in ROM 788 is unknown, but an infectious or inflammatory process may have been responsible.

3.3.2 Frequency and patterns of pathologies in ankylosaurid taxa

A total of 193 caudal vertebrae from a variety of ankylosaurid taxa, representing all parts of the caudal series, were examined over the course of this study (Appendix 1). These represent at least seven taxa: Ankylosaurus,

Euoplocephalus, Gargoyleosaurus, Gastonia, Pinacosaurus grangeri, Talarurus, and Tarchia. These represent at least 26 individuals, excluding specimens from bonebeds (MPC PJC 2007 Pinacosaurus and DMNH Gastonia). Ten (13%) pathological vertebrae were observed in 77 vertebrae referred to

Euoplocephalus. Of these, all pathologies were found in the free caudal vertebrae, and none were recognized in the handle vertebrae. Of eight observed partial free caudal vertebrae of Pinacosaurus, one has a pathological transverse process. No pathological vertebrae were observed in any other ankylosaurid

76 Table 3.3: Summary of diagnoses- Specimen Abnormality Diagnosis AMNH 2062 Swollen transverse process with Pyogenic osteomyelitis cloaca AMNH 5245 Two free caudal vertebrae fused at Pyogenic osteomyelitis (?following zygapophyses, rugose neural trauma) spines AMNH 5337 Lytic lesion on ilium and two Metastatic cancer or idiopathic abnormal sacral ribs anomaly AMNH 5404 Lytic lesions on centra Metastatic cancer AMNH 5409 Lytic lesion on ilium and abnormal Metastatic cancer or idiopathic sacral rib anomaly ROM 788 Unusual knob osteoderm texture Periostitis or idiopathic anomaly ROM 1930 Bone spicules on two centra Enthesophyte or developmental abnormality ROM 1930 Inflated, rugose haemal spine Osteitis, periostitis, or neoplasm TMP Lytic lesion on knob osteoderm; Pyogenic osteomyelitis 1983.36.120 asymmetrical knob osteoderm heights TMP 1985.36.70 Rugose, swollen neural spine and Pyogenic osteomyelitis transverse processes on caudal vertebra TMP Sharply bordered bump on Enthesophyte or developmental 1992.36.344 centrum, next to neural arch abnormality TMP 2005.09.75 Rugose, swollen neural spine and Pyogenic osteomyelitis transverse processes on caudal vertebra

taxa, although the number of vertebrae examined was much smaller compared to

Euoplocephalus. Of three Euoplocephalus pelves examined (AMNH 5337,

AMNH 5409, and TMP 1982.9.3), two had abnormalities.

No pathologies were found in a survey of 67 isolated caudal vertebrae

belonging to Gastonia sp., collected from a single quarry and representing

multiple individuals. Gastonia is a basal ankylosaurid (Vickaryous et al. 2004)

that lacked a tail club. Many of the Gastonia vertebrae are taphonomically

distorted, with centra obviously crushed or distorted laterally, but none had

excavations, rugosities, or bone spicules. No pathological chevrons were

present, either. No pathologies were found in the basal ankylosaurid

77 Gargoyleosaurus, which also lacks a tail club, although only a few vertebrae are preserved in the holotype (and only) specimen.

A survey of nodosaurid caudal vertebrae was beyond the scope of this study, but at least one pathological nodosaurid specimen was encountered.

AMNH 5303 (Sauropelta) includes a series of 17-18 articulated caudal vertebrae from the posterior portion of the tail, including the terminal vertebra. The most posterior nine to ten vertebrae are fused together, the tail is kinked ventrally at the ?11th vertebra in the series, and the posterior portion of the fused vertebrae are offset to the left at this point (Fig. 3.8). This vertebra appears to bear a fracture callus, suggesting that the distal portion of the tail suffered a trauma, that the vertebra healed in a misaligned position, and that some of the posterior vertebrae may have been dislocated from their normal position.

Pathologies in ankylosaurid tail vertebrae have only been observed in free caudals, and not in vertebrae of the tail club handle. The distal portions of the vertebrae were more affected by osteomyelitis, osteitis, or periostitis, while the

Figure 3.8: Distal portion of the tail of AMNH 5403, Sauropelta, left lateral view. The distal caudals are kinked and fused together. Scale equals 5 cm.

78 centra were affected by cancer or idiopathic anomalies. The most frequently affected parts of the vertebra are the neural spine and transverse processes.

One haemal spine was affected, and two centra were affected. Fusion of two free caudal vertebrae was observed in AMNH 5245. It appears that anterior free caudals are more frequently pathological than are free caudals that are more distal in series.

3.3.3 Behavioural implications

One of the main objectives of paleopathology is to provide insight on the behaviour and lifestyles of these animals. Evidence of disease can provide support for proposed behaviours. Infected Clidastes caudal vertebrae with embedded shark teeth, and shark-bitten Platecarpus vertebrae, are indicative of shark attacks on mosasaurs (Rothschild and Martin 1993). Rothschild and Martin

(1987) suggested that mosasaurs suffered from avascular necrosis (resulting from caisson disease, "the bends"), and that this was evidence of their deep- diving habits. Tyrannosaurids suffered from gout, which may be linked to a diet rich in red meat (Rothschild et al. 1997). Stegosaurus specimens with fractured and infected tail spikes suggest active use of the tail as an offensive or defensive weapon (McWhinney et al. 2001).

Ankylosaurid dinosaurs are thought to have used the tail clubs as a weapon (Coombs 1995). Fractured and healing caudal vertebrae and tail club knobs could provide evidence of active use of the tail, as could tendon avulsions.

The handle of the tail club is a long, rigid structure, and fractures might be

79 expected to be most common in this region of the tail. No fracture calluses were apparent in any of the handle vertebrae examined in this study. It is possible that the vertebrae of the handle were well-equipped to absorb the stress and strain of forceful impacts, and did not fracture easily, or, that tail clubs were not used for forceful impacts. It is also possible that the absence of pathological handle specimens reflects the small sample size in this study.

Infected vertebrae may also indicate active use of the tail, although bone

infections can arise both from direct injury and infection, or haematogenously from elsewhere in the body (Rothschild and Martin 2006). Instances of osteomyelitis in extant wild and captive animals are not usually the result of trauma: osteomyelitis in a captive bottlenose dolphin (Tursiops truncatus), wild

leatherback sea turtle (Dermochelys coriacea), and wild crow (Corvus

brachyrhynchos) were attributed to haematogenous infection (Daoust 1978;

Ogden et al. 1981; Alexander et al. 1989).

Several vertebral elements show evidence of infection, especially the

neural spines of more anterior free caudal vertebrae. The infection is usually

restricted to the area of the neural spine dorsal to the zygapophyses. The

absence of infected caudal vertebrae in the basal ankylosaurid Gastonia, despite

a similar sample size to that of Euoplocephalus, may suggest that: 1) basal

ankylosaurids did not actively use the tail as a weapon, 2) basal ankylosaurids

may have used the tail as a weapon, but that the forces involved were less likely to result in trauma and infection, or 3) basal ankylosaurids were less prone to

bony infections in the tail, regardless of cause. Pathological caudals in the

80 nodosaurid Sauropelta indicate that nodosaurids were also susceptible to trauma in the tail, even though frequent use of the tail as a weapon has not been proposed for this group.

TMP 1983.36.120 shows that the tail club had become infected in this individual, although how the infection originated is unknown. Unlike the

Stegosaurus tail spikes described by McWhinney et al. (2001), the

Euoplocephalus osteoderms show no obvious fracture calluses. The tail club may have become infected following a trauma that did not fracture the massive osteoderms, or the infection may have originated elsewhere in the body and spread to the tail club. Evidence for intraspecific combat resulting in bone infection in extant animals is mixed. For example, Hoets and Bunch (1992) described a wild male Dall sheep (Ovis dalli dalli) with asymmetrical horns.

Examination of the skull after death showed chronic osteomyelitis contained between the horn core and horn sheath; the sheath was thickened and rough in the areas overlying the infected bone. Hoets and Bunch (1992) suggest that trauma to the horn, from butting other rams, may have caused the infection. Erb et al. (1996) observed that adult male fur seals (Arctocephalus gazella) bite rival males during territorial fights, and that only adult male fur seals had dental and mandibular infections. However, these authors also note that most of this intraspecific combat is ritualistic, and that most fights were minor. Instead, they suggested that osteomyelitis in the mandible of a male fur seal was likely the result of trauma incurred when heavy surf smashed the seal into rocks or boulders on the beach. It is possible that the infection in TMP 1982.36.120 is a

81 result of trauma to the tail club. However, it is equally likely that the infection in the osteoderm was haematogenous in nature. The overlying keratinous sheath that is hypothesized to have covered the osteoderms may have become enlarged and rugose over the bony lesion, as was the case with the Dall sheep described by Hoets and Bunch (1992). The unusual furrows in ROM 788 could be related to an infection in the overlying keratinous sheath, but no reports of a similar morphology in extant animals could be found.

The identification of metastatic cancer in Euoplocephalus does not provide any information on the behaviour of ankylosaurids. However, this finding is of interest because it is the first time cancer has been identified in an ankylosaurid.

Metastatic cancer was diagnosed in an unidentifiable dinosaur bone fragment from the Jurassic by Rothschild et al. (1999). In a radiological survey of more than 10,000 vertebrae from all of the major groups of dinosaurs, Rothschild et al.

(2003) only found evidence for cancer in the Hadrosauridae, and only found metastatic cancer in one, out of 528 examined, Edmontosaurus vertebra.

3.4 Conclusions

Pathologies in ankylosaurid tail elements were found in 13% of

Euoplocephalus vertebrae, and can be attributed to infections, cancer, and idiopathic processes. No healing fractures were found, and no pathologies were found in the handle vertebrae. The relatively common infection of neural spines and transverse processes of more anterior caudals is an interesting pattern, although the cause remains unknown. This study also describes the first known

82 case of metastatic cancer in a dinosaur outside of the Hadrosauridae. There are many possible explanations for the pathologies of the tails and pelves described in this paper. Infections could be unrelated to traumatic injuries and could indicate other systemic diseases. Injuries can be related to inter- or intraspecific combat, mating, predation, or random accidents. At present, the pathologies of the tails and pelves cannot be used as evidence of tail-clubbing in ankylosaurids.

However, infections of two tail club knobs could be related to trauma induced by forceful impacts. Currently, there are few details regarding what disease processes affect osteoderms, how to recognize the effects of disease in osteoderms, and how commonly osteoderms in various taxa are affected by disease. A greater knowledge of the pathology of disease in osteoderms may help elucidate the causes of lesions in osteoderms in the fossil record, which in turn may be useful in interpreting ancient behaviours in extinct animals.

83 Chapter 4: Ankylosaurid caudal musculature3

4.1 Introduction

In order to understand the mechanics of tail swinging, the muscles of the tail and pelvis in ankylosaurids must be reconstructed. Of particular interest are the caudal epaxial and hypaxial muscles, and some muscles of the hindlimb, in particular the Musculus caudofemoralis longus and M. caudofemoralis brevis.

Crocodilians are used as the main comparative analogue, and are suitable for two reasons: 1) crocodilians represent one pole of the extant phylogenetic bracket for nonavian dinosaurs (Witmer 1995), and 2) crocodilians have long, muscular tails, which are capable of generating large forces; crocodiles use their tails actively during swimming (Manter, 1940), and to propel themselves into a

'death roll' for rotational feeding (Fish et al. 2007). Although birds are more closely related to ankylosaurs, crocodilian tails more closely resemble those of ankylosaurs in relative length, number of vertebrae, and size of processes for muscle attachment. Muscle reconstructions in this paper use previously published studies of crocodilian anatomy and dinosaurian muscle reconstructions, and comparisons with lizards and birds that complete the phylogenetic bracket for ankylosaurs.

3 This chapter has been submitted for publication. Arbour, V., and Snively, E. Biomechanics and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora). The Anatomical Record A, submitted 10 October 2008. All text and figures in this chapter are by V. Arbour. E. Snively provided comments on an early draft of this chapter. Permission from the coauthor can be found in Appendix 2. 84 4.2 Review of epaxial and intrinsic trunk and tail muscles of extant reptiles

Muscles of the tail are serially homologous with the dorsal spinal musculature. Organ (2006b) reviewed the thoracic epaxial muscles in extant archosaurs. In the dorsal region, they are divided into M. iliocostalis, M. longissimus dorsi, and M. transversospinalis. In the caudal region, M. iliocostalis is absent, M. longissimus dorsi is replaced by M. longissimus caudae, and M. transversospinalis is present.

M. transversospinalis includes several subunits: the intrinsic vertebral muscles Mm. interarcuales, Mm. interarticulares superiores, and Mm. interspinals, and the outer M. multifidus, M. semispinalis (further divided into M. articulospinalis and M. tendinoarticularis), and M. spinalis. Mm. interspinales and

Mm. interarcuales connect the anterior and posterior edges of successive neural spines. Mm. interarticulares superiores connect the zygapophyses of successive vertebrae (Organ 2006b).

M. multifidus, M. semispinalis, and M. spinalis act synergistically to bend the tail laterally (Chiasson, 1962). Lateral to Mm. interspinales and Mm. interarcuales lies M. multifidus, which originates from a tendinous sheet, passes over two vertebrae, and inserts on the posterodorsal surface of the third neural spine (Organ 2006b). M. spinalis is lateral to M. multifidus, and interweaves with

M. longissimus (Cong 1998). Organ (2006b) states that M. spinalis originates near the prezygapophysis, whereas Chiasson (1962) gives the origin as the transverse process and segmenting fascia. M. spinalis inserts on the anterodorsal surface of the neural spines of the fifth to seventh vertebrae from its

85 origin, as well as onto the lower surfaces of the overlying osteoderms (Organ

2006b). Lateral to M. spinalis is M. semispinalis, which is subdivided into M. articulospinalis and M. tendinoarticularis (Organ 2006b). M. articulospinalis originates on the base of the neural spine, passes four vertebrae, and divides into three tendinous heads that insert on 1) the overlying osteoderms, 2) the posterodorsal surface of the neural spine, and 3) the fascia of M. tendinoarticularis (Organ 2006b). M. tendinoarticularis is lateral to M. articulospinalis and is formed of subunits of anteriorly-pointing cones of myosepta (Organ 2006b).

M. longissimus caudae is found lateral to M. transversospinalis, is composed of posteriorly-pointing cones of myosepta (Organ 2006b), and acts to move the spinal column laterally (Chiasson 1962). M. longissimus caudae interweaves with M. spinalis, and is the longest and strongest component of the epaxial musculature (Cong 1998). It originates on the anteromedial side of the ilium (Organ 2006b), and posteriorly on the caudal transverse processes in a direction opposite that of the M. longissimus dorsi (Cong 1998). It passes three vertebrae and inserts on the dorsal distal parts of the transverse processes

(Organ 2006b). In most crocodylians, the cones are enlarged in diameter and length adjacent to the first three to four caudal vertebrae, compared to those in the lumbar and sacral region, which decrease in size proportionally to the vertebrae (Frey et al. 1989). M. tendinoarticularis is subsequently reduced in size

(Frey et al. 1989). However, in Gavialis, M. longissimus caudae has an evenly decreasing diameter from the sacral region posteriorly (Frey et al. 1989). Mm.

86 intertransversarii connect the ventral sides of the transverse processes (Cong

1998), and are considered part of M. longissimus in Crocodylia (Tsuihiji 2007).

M. ilioischiocaudalis is a massive muscle that extends along the entire length of the tail of crocodylians (Cong 1998). It originates tendinously from the posterior of the blade of the ilium and from the posterolateral corner of the ischium, and inserts on each caudal (Cong 1998).

M. caudofemoralis is subdivided into M. caudofemoralis longus and M. caudofemoralis brevis. In crocodylians M. caudofemoralis longus originates on the transverse processes of the anterior caudal vertebrae, and inserts tendinously on the fourth trochanter of the femur and to the shank (Chiasson

1962). M. caudofemoralis longus of lizards originates from fascia and septa associated with the anterior edges of the transverse processes, the haemal spines, and ventral portion of the centrum (Russell et al. 2001). M. caudofemoralis is present in birds; M. caudofemoralis pars caudalis is slender and acts to twitch the tail (Vanden Berge and Zweers 1993). In crocodylians, by contrast, M. caudofemoralis is a massive, strong muscle that acts to draw the femur caudally (Cong 1998). In Alligator mississippiensis, Caiman crocodylus,

Crocodylus acutus, Osteolaemus tetraspis, Palaeosuchus trigonatus, and

Tomistoma schlegeli, this muscle extends to the thirteenth or fourteenth caudal vertebra, whereas in Gavialis it extends to the tenth caudal vertebra (Frey et al.

1989). M. caudofemoralis brevis inserts on the lateral surface of the ischium and on the ventral surface of the transverse processes of the last sacral and first caudal vertebrae (Chiasson 1962). It inserts on the posterior proximal half of the

87 femur and draws the hindlimb posteriorly (Chiasson 1962). M. caudofemoralis brevis is not present in Gavialis (Frey et al. 1989).

4.3 Ossified tendons of ornithischian dinosaurs

Organ (2006b) reviewed the arrangement of ossified tendons of the thoracic region in crocodiles and birds, and provided a detailed interpretation of the ossified tendon lattice of iguanodontoid dinosaurs. Crocodiles and birds both possess ossified tendon lattices associated with M. transversospinalis; crocodiles have two sets of tendons that run posteroventrally and one set that runs posterodorsally, whereas birds have the opposite arrangement. These tendon sets are hypothesized to correspond with the M. articulospinalis, M. spinalis, and

M. tendinoarticularis subunits of M. transversospinalis. M. interarticularis, M. interspinales, and M. longissimus are found in both birds and crocodilians, leading Organ (2006b) to infer their presence in dinosaurs. M. multifidus is not found in birds, and no tendons corresponding to this muscle are found in iguanodontoids. Organ (2006b) examined thoracic epaxial musculature only, and determined that parallel bundles of ossified tendons along the transverse processes belonged to M. iliocostalis or M. longissimus, based on their position.

An ossified tendon trellis in Chasmosaurus irvinensis is similarly arranged to that in iguanodontoids, representing subunits of M. transversospinalis. As in iguanodontoids, M. multifidus appears to be absent in C. irvinensis (Holmes and

Organ 2007).

88 Coombs (1995) briefly discussed ossified tendons in ankylosaurids, and suggested that caudal ossified tendons represented M. caudofemoralis, M. iliocaudalis, and various intrinsic axial muscles. Coombs (1995) suggested that these long, distal ossified tendons were used to transmit large forces from the more proximally located caudal muscles.

The ossified tendons of the handle are best preserved in ROM 784,

Dyoplosaurus acutosquameus. Parks (1924) recognized three series of tendons on the dorsolateral sides of the handle, and four on the distal, ventral side of the tail. Here two sets of tendons are recognized on the dorsolateral sides. The ventral side of the specimen is not exposed. The inner set of tendons has an imbricated appearance, whereas the tendons of the outer layer are parallel with a braided appearance (Fig. 4.1). The inner tendons are shorter in length compared to the long outer tendons, and have smaller diameters. The inner set of tendons is slightly dorsal to the outer set. Posteriorly, the inner and outer set converge towards the knob, whereas anteriorly the two sets are distinctly separated. The tendons are posterodorsally oriented, and the inner set more strongly so. The anteriormost outer tendons are parallel and vertically stacked. The inner set of tendons inserts at either the midpoint of the centrum or the neural arch.

4.4 Inferred tail musculature of ankylosaurids

Ankylosaurids probably had epaxial musculature in the caudal region similar to that of extant crocodilians, and this was likely also similar to the thoracic epaxial musculature inferred in other omithischian dinosaurs. The

89 M. spinalis

M. longissimus caudae »

Figure 4.1: Ossified tendons in ROM 784, Dyoplosaurus, oblique right lateral view. This specimen is the best example of the arrangement of ossified tendons in ankylosaurid tail clubs. M. spinalis is represented by the inner set of imbricated tendons, and M. longissimus caudae is represented by the outer set of parallel to braided tendons. The ossified tendons continue underneath the knob osteoderms (arrowhead). Scale equals 10 cm. intrinsic vertebral muscles Mm. interspinales, Mm. interarcuales, and Mm. interarticulares superiores were probably present. M. multifidus may or may not have been present in ankylosaurids, as its presence in other ornithischians is neither supported nor refuted (Organ 2006b).

Based on comparisons with the work of Organ (2006) and Holmes and

Organ (2007), the posterodorsally-oriented, inner set of ossified tendons alongside the handle probably represents M. spinalis. Organ (2006) considered parallel bundles of tendons along the transverse processes as representing M. iliocostalis or M. longissimus dorsi. Because M. iliocostalis is not present along the caudal vertebrae, it is likely that the parallel, outer set of tendons in ROM 784 represents M. longissimus caudae. M. transversospinalis was present, and is

90 represented in the distal portion of the tail by ossified tendons from the M. spinalis subunit. It is unknown whether M. semispinalis was present, and if so, how large it was in the caudal region.

Symmetrical ridges located approximately halfway along the lateral edge of the ilium of AMNH 5409 (Fig. 4.2) likely correspond to the origin of M. longissimus caudae. These ridges are more than 5 cm long, and suggest that M. longissimus caudae was large, at least proximally. Coombs (1979) suggested that the rugose lateral edges of ankylosaurid ilia corresponded to the origin of M. longissimus dorsi, although this would have resulted in an unusually long M. longissimus dorsi. The transverse processes are not large or robust in ankylosaurids, and these would have limited the size of M. longissimus caudae posteriorly along the tail. Parallel ossified tendons along the handle may represent the distal portion of M. longissimus caudae; if so, this would suggest that M. longissmus caudae was an important tail-swinging muscle. The handle vertebrae lack transverse processes, although there are occasionally bumps or ridges along the lateral sides of the centra (e.g. ROM 784), which may represent the insertion of M. longissimus caudae.

Coombs (1979) reconstructed ankylosaur pelvic muscles with separate M. iliocaudalis and M. ischiocaudalis. According to Coombs (1979), M. iliocaudalis originated from a massive blunt knob at the caudal end of the ilium and inserted only along the proximal caudals, and this interpretation is accepted here (Fig.

4.2). M. ischiocaudalis originated from the distal terminus of the ischium, and

Coombs (1979) suggested that this muscle was probably not involved in tail

91 Figure 4.2: Origins of tail muscles on the pelvis. A) AMNH 5409 (Euoplocephalus) pelvis, posterior right dorsolateral view. M. ischiocaudalis originates at the distal terminus of the ischium. The origin of M. longissimus caudae is marked by a long, pronounced ridge and rugose area on the lateral aspect of the ilium. The posterior terminus of the ilium is partially reconstructed. B) AMNH 5337 (Euoplocephalus) pelvis, dorsal view, showing an unreconstructed posterior terminus of the left ilium. M. iliocaudalis originates from a large knob. C) AMNH 5409, same view as A, with reconstructed musculature. The muscles are cut posteriorly to show their relationships in cross-section. M. caudofemoralis longus originates on the transverse processes of the free caudal vertebrae, and inserts on the fourth trochanter of the femur (not shown). M. transversospinalis originates and inserts on the neural spines. Scales equal 10 cm. Abbreviations: ca = M. caudofemoralis longus, il = M. iliocaudalis, is = M. ischiocaudalis, lo = M. longissimus caudae, tr = M. transversospinalis.

92 swinging, due to the vertical orientation of the ischium (Fig. 4.2). M. caudofemoralis longus (Fig. 4.2) inserted onto the distally located fourth trochanter of ankylosaurids (Coombs 1979).

Ankylosaurid tail musculature is reconstructed in cross section in Figure

4.3. Because the morphology of various subunits of M. transversospinalis is uncertain, and because there is little osteological evidence for the size of these divisions, the entire M. transversospinalis system is depicted rather than its components. In the free caudal vertebrae, M. transversospinalis would have occupied the area closest to the neural spine. M. longissimus caudae is here reconstructed as a large muscle occupying the area lateral to M. transversospinalis to the distal terminus of the transverse process.

Ventrally, M. caudofemoralis longus is the largest muscle, and is reconstructed occupying an area between the transverse process and the stout portion of the haemal arch. Cong (1998) shows M. ilioischiocaudalis of Alligator sinensis forming the outer boundary of the ventral tail musculature, between the transverse process and the distal portion of the haemal spine. Ankylosaurids likely had a small M. ischiocaudalis (Coombs 1979), which is here reconstructed occupying the area near the ventral terminus of the haemal spine. Cong (1998) also shows that there is a varying amount of fat between the M. caudofemoralis longus and M. ilioischiocaudalis in the anterior portion of the tail, which reduces in size posteriorly. These fat deposits leave no correlates for reconstruction in ankylosaurids, and so are excluded here. In crocodilians, the musculature of the tail is to a certain extent limited by the vertebrae, but the cross-sectional profile

93 Figure 4.3: Cross-sectional reconstructions of ankylosaurid caudal musculature. A) Anterior free caudal vertebra, modified from TMP 85.26.70. M. transversospinalis is not divided into its subunits. The relative sizes of all muscles are speculative, especially M. iliocaudalis and M. ischiocaudalis. B) More muscular reconstruction of A, with muscles bulging past the transverse processes and the neural and haemal spines. C) Posterior free caudal vertebra, reconstructed from TMP 2007.20.100. M. iliocaudalis may not have extended very far posteriorly along the tail, in which case M. ischiocaudalis may have occupied the area reconstructed as M. ischiocaudalis here. D) Musculature of the handle, reconstructed from a CT scan image of UALVP 47273 at approximately the midlength of a tail club. M. caudofemoralis originates on the transverse processes, and because the handle caudals lack transverse processes, it is not likely that this muscle extended onto the handle. M. transversospinalis and M. longissimus caudae are represented by ossified tendons in many tail club specimens. The size of M. iliocaudalis is speculative. The width of M. longissimus caudae is equivalent to the maximum space between the major osteoderms of the tail club knob. Scales equal 5 cm.

of the tail changes greatly from anterior to posterior (Cong 1998). A conservative

reconstruction of the muscles of the tail of ankylosaurids would have an elliptical

cross-sectional outline, with none of the muscles bulging past the transverse

processes, or neural and haemal spines.

94 Figure 4.4: ROM 784 tail club in dorsal view, showing maximum width of the maximum width of handle muscles. Scale equals 5 cm. handle muscles

There are fewer osteological correlates for muscle attachments in the handle vertebrae, and it is even more difficult to estimate the cross-sectional outline of the muscles than in the free caudal vertebrae. However, one clue that may indicate muscle area is the amount of space between the knob osteoderms.

Their bumpy or dendritic texture suggests that they were covered by a keratinous sheath, and not muscle. In crocodilians, the epaxial musculature is firmly connected to the dermis (Gasc 1981), and tendons of M. spinalis insert on the basal sides of osteoderms (Organ 2006b). The width between the two major knob osteoderms in dorsal view must have been the maximum width of the handle muscles (Fig. 4.4). M. transversospinalis is represented by ossified tendons in the handle, and probably occupied the area dorsal and lateral to the neural arch. The outer set of ossified tendons in ROM 784 may represent M. longissimus caudae, which would have occupied the space lateral to the centrum. M. caudofemoralis longus was likely absent along the handle vertebrae, and it is unknown whether or not M. ischiocaudalis was present in this region. In these reconstructions, M. iliocaudalis occupies the space ventral and lateral to the haemal arch.

95 All of the epaxial musculature would function to bend the tail laterally, and an anteriorly large M. longissimus caudae might imply that the tail could be swung quite forcefully. A problem with trying to understand which muscles may have contributed the most to tail-swinging actions is the lack of understanding of tail muscle function in extant analogues. Further research on the function of large muscles in alligator tails would help clarify the reconstructed musculature of ankylosaurid tails.

96 Chapter 5: Analysis of tail swinging ability in ankylosaurid dinosaurs.4

5.1 Introduction

The large knob and stiffened handle vertebrae of ankylosaurid tail clubs suggest reinforcement against impacts. However, the biomechanics of the tail and tail-swinging in ankylosaurids have not been studied in detail. This chapter examines the internal morphology of tail clubs using data from computed tomography (CT) scans. Data from CT scans and observation of other fossil specimens are used to calculate functional dynamics of ankylosaurid tails, to determine potential values for impact forces. If the function of ankylosaurid tail clubs is to deliver forceful blows, then tail club swings should be able to generate large impact forces and stresses.

5.2 Materials and methods

Four ankylosaurid tail clubs were CT scanned, to provide information on their internal structure, and to derive three dimensional models for use in volume estimates and FEA. ROM 788 and UALVP 47273 have substantial portions of the handle preserved, and represent examples of small and large knobs, respectively. UALVP 16247 and a cast of TMP 1983.36.120 do not preserve much of the handle and have average-sized knobs. TMP 1983.36.120, UALVP

4 This chapter has been submitted for publication. Arbour, V., and Snively, E. Biomechanics and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora). The Anatomical Record A, submitted 10 October 2008. All calculations, text, and figures in this chapter are by V. Arbour. E. Snively contributed advice on calculating muscle forces and provided comments on an early draft of the chapter. Permission from the coauthor can be found in Appendix 2. 97 16247, and UALVP 47273 were scanned at the University of Alberta Hospital

Alberta Cardiovascular and Stroke Research Centre (ABACUS), on a Siemens

Somatom Sensation 64 CT scanner, at 1 mm increments (Fig. 5.1). ROM 788 was scanned at CML Healthcare Imaging in Mississauga, Ontario, at 2 mm increments. All CT scans were viewed using the software program OsiriX, and interpreted using a grayscale colour palette for density values.

IS Figure 5.1: CT scanning (j§l UALVP 47273 at the University of Alberta ABACUS facility.

5.3 Description of club internal morphology from CT scans

CT scans provide information about the internal structure of the handle vertebrae, the knob osteoderms, and the relationships between the vertebrae, ossified tendons, and knob. The scans also provide information about the differences between small and large clubs. UALVP 47273 provided the best data, because of the quality of the scan and because it is relatively complete. ROM

788 was scanned in two pieces (knob and handle). The knob width was only slightly smaller than the aperture of the scanner, and was slightly larger than the

98 field of view. As a result, the lateral edges of both major osteoderms were partially excluded from the scan. Most of the knob was obscured by artifacts resulting from beam hardening and the partial volume effect (Zollikofer and

Ponce de Leon 2005), possibly caused by ferrous minerals infilling the pore spaces in the knob, or because the knob was too large for the X-rays to penetrate uniformly. Even with the artifacts, the borders of the specimen can usually be determined, except for the dorsal border of the vertebra in the centre of the knob. Some artifacts are present in the scan of UALVP 16247, but these are not prominent and are easily distinguished from the bone.

In ROM 788 and UALVP 47273, the centra are comprised of low density cancellous bone, whereas the neural and haemal arches are dense compact bone (Fig. 5.2). The ossified tendons are similarly dense. The neural and haemal canals are radiolucent in the scans, indicating that they have been infilled with minerals. Transversely, the neural canal is circular to oval. The haemal canal is always completely enclosed by bone. The centra are at times difficult to discern in the knob, but the neural and haemal canals are visible until near the terminus of the knob. In UALVP 47273, the neural canal seems to end at approximately the anterior border of the minor plates that comprise the distal end of the knob

(Fig. 5.3).

In UALVP 16247, the shape and number of the vertebrae in the knob is best viewed in coronal view (Fig. 5.4). Three vertebrae are preserved in the knob, and the last vertebra extends almost to the posterior terminus of the knob.

The posterior two vertebrae are completely enclosed laterally by the major

99 Figure 5.2: UALVP 47273 transverse views of handle, anterior view. A) At approximately the midlength of a handle vertebra. The neural canal is completely enclosed by the neural arch, and parts of three neural arches are visible. The haemal arch is poorly preserved in this section. B) Posterior to the midlength of the vertebra. The neural canal is not completely enclosed by the neural arch, and parts of only two neural arches are visible. The haemal canal is robust and the centrum is cancellous. Scale equals 5 cm. Abbreviations: C, centrum, HA, haemal arch, HC, haemal canal, NA1, neural arch of the centrum in the slice, NA2, neural spine of the anterior vertebra, NA3, prezygapophyses of the posterior vertebra, NC, neural canal, OT, ossified tendon.

osteoderms. The anterior two vertebrae are partially exposed dorsally, but the terminal vertebra is dorsally covered by the minor plates. In dorsal view, the two anterior vertebrae have the characteristic elongate hourglass shape found in handle vertebrae. The terminal vertebra is abbreviated in length, with a length of less than one third that of the penultimate vertebra. The terminal vertebra is roughly triangular in dorsal view, rather than hourglass-shaped. Figure 5.3: UALVP 47273 sagittal views, left lateral view. A) Section at approximately the mid- width of the club. Most of the centra appear to be fused at the anterior and proximal faces (arrow with open head), although one joint does not appear fused (arrow with closed head). B) Section at approximately the mid-width of the left half of the club. The neural canal extends to approximately the anterior terminus of the minor plates at the distal end of the knob (arrowhead). The three narrow, vertically stacked structures at the anterior of the handle are ossified tendons. Scale equals 10 cm.

Figure 5.4: A) UALVP 16247, coronal plane, dorsal view, at approximately mid-height of the knob. B) Interpretive illustration of A, showing the shapes of the vertebrae, highest density areas (white), medium density areas (light grey), and lowest density areas (dark grey). The neural canal and vascular canals in the osteoderms are indicated by black. Scale equals 5 cm.

101 In some tail clubs (e.g. AMNH 5245), successive tail club centra are not fused at the anterior and posterior ends. However, in sagittal view of UALVP

47273, bright zones at the articular ends of the centra, and a lack of distinct spaces between the centra, seem to indicate fusion of successive handle vertebrae (Fig. 5.3). However, this may result from mineralization of the space between vertebrae. Vertebrae appear fused in ROM 788, although the specimen has been partially reconstructed and painted. The CT scan of the ROM 788 handle does not clarify whether or not the vertebrae are fused at the centra.

Ossified tendons are preserved alongside the handles in ROM 788,

UALVP 16247, and UALVP 47273. In UALVP 47273, the ossified tendons are visible between the osteoderms and vertebra (Fig. 5.5). Tendons are comprised predominantly of compact bone.

In UALVP 47273, the osteoderms each have a relatively thin compact cortex, and are predominantly cancellous (Fig. 5.5). The cortex is slightly thicker on the right major plate than on the left, especially at the keel. This compact bone is absent on the dorsal and ventral medial edges of the major plates. The minor plates at the distal tip of the knob are somewhat denser than the major plates.

Pores are approximately radially oriented near the outer edges of the osteoderms, and have a more random distribution medially. Some large pores can be traced several centimeters dorsally from the ventral border of the knob osteoderms. In transverse sections through the major plates, there are patches of low density (Fig. 5.5). These change shape anteroposteriorly, but remain symmetrical between the osteoderms.

102 Figure 5.5: Transverse sections through knobs, anterior view. A) UALVP 47273, at approximately the midlength of the knob. Note the symmetrical patches of low density in the middle of the osteoderms. The vertebra is cancellous and has a distinct boundary with the osteoderms. The arrowhead marks three vertically stacked ossified tendons between the left major osteoderm and the vertebra. B) UALVP 47273, approximately 2 cm posterior from A. The outer layer of compact bone is thicker in the right osteoderm than the left. The neural canal is distinct. Symmetrical patches of low density are also present in this section. C) UALVP 16247, at approximately 1/3 from terminus of knob. The outer layer of denser bone is relatively thicker compared to UALVP 47273. The boundary between the vertebra and the osteoderms is less distinct compared to UALVP 47273, although the neural and haemal canals are distinct. Note the difference in the position of the keel in UALVP 47273 and UALVP 16247. D) ROM 788, approximately 1/3 from terminus of knob. Artifacts obscure most of the fine details. The density in the osteoderms appears relatively more uniform compared to UALVP 47273 and UALVP 26347, although this may be in part due to the artifacts. The boundary between the vertebra and the osteoderms is indistinct. The arrow marks the support jacket in which the knob was scanned, and the arrowhead marks the CT scanning tray. Scales in A, B, and C equal 5 cm, scale in D equals 10 cm.

103 5.4 Analysis of tail club motion and impact force

Alexander et al. (1999), Carpenter et al. (2005), and Snively and Russell

(2007) have investigated the dynamics of vertebral flexion in fossil vertebrates;

Alexander et al. (1999) estimated tail blow energy in glyptodonts, Carpenter et al.

(2005) calculated impact force in Stegosaurus spikes, and Snively and Russell

(2007) investigated tyrannosaurid necks. A method similar to that employed by

Carpenter et al. (2005) is used here, as this method is the most detailed, and using this method allows mechanics of stegosaur and ankylosaur tail impacts to be precisely compared. Carpenter et al. (2005) measured a large mounted

Stegosaurus and modeled the tail as a series of five rigid links, with the anterior and posterior boundaries of the links defined by the large plates that occur above the vertebrae of Stegosaurus. Ankylosaurids were not limited by such large plates in the tail region, although many ankylosaurids (Dyoplosaurus,

Pinacosaurus, Saichania, and Tarchia) have laterally-oriented, wedge-shaped osteoderms along the lateral sides of the tail. The complete caudal armour is not known in Euoplocephalus. Therefore, osteoderms other than the knob osteoderms of the tail club are ignored, for both mass estimates and possible limits on the range of motion of the tail. ROM 784 (Euoplocephalus) has eleven free caudal vertebrae, eleven visible handle vertebrae, and a transitional free caudal vertebra was not preserved. No movement would have been possible between the transitional free caudal vertebra and the first handle vertebra.

Therefore, there would have been twelve free caudal segments and one tail club segment, for a total of thirteen segments to model in the tail. The ossified

104 tendons are not included in this analysis, because their role in tail club swinging and impacts is unresolved.

The impact force of the knob is related to the acceleration and mass of the tail club segment. Actual acceleration and mass cannot be directly measured, and so must be calculated and inferred using other properties. Properties that can be directly measured on fossil specimens include length, width, and height of each vertebral segment. From these, the volume of each bone segment can be calculated, and the mass of the bone can be calculated using estimates of modern bone density. Muscle height, length, and width can be estimated for each segment, which provides approximations of muscle cross-sectional area, volume, and mass. Another property that can be directly estimated from the fossils is the angle of articulation of each segment, 20 (where 0 represents the half angle of articulation of each segment). The half angle of articulation is the maximum amount of lateral deflection from the neutral position (Fig. 5.6).

maximum right lateroflexion

neutral axis Figure 5.6: Diagram showing the approximate right lateroflexion of the tail in Euoplocephalus, and the half angle of articulation 9.

105 From these properties, the following quantities can be calculated for ankylosaur tail dynamics:

2 Rotational inertia: I = J,mjr

Angular velocity: w = hmAr^0— = — At dt

A i . ,• Aeo dco Angular acceleration: a = hmA,^0 — = —

A? dt

Torque: T = la\ T = r(F sin(j)); T = rFt\ T = r±F

2 2 2 Kinetic energy for a rotating body: KE = ^mjr co , or KE =—Ico

Translational velocity: V = cor

Carpenter et al. (2005) use the following equation for rotational inertia, I:

I = \ x2pdx = x3 L> 3 3

Where l_i is the distance from the proximal end of the segment to the base of the tail, L2 is the distance from the distal end of the segment to the base of the tail, p is the average mass density per unit length, and x is the variable of integration between U and L2.

Muscles pull on one half of the width of each link at the proximal end, which generates torque in each segment.

T = rxF

** ~ V*xs A* muscle ) 2 In these equations Axs (in cm ) is the cross-sectional area of muscle at the proximal end of the segment, and Pmuscie is the specific tension of the muscle, the

2 force the muscle can exert per unit of cross-sectional area (N/cm ). Axs is determined by calculating the total cross-sectional area of the tail (represented by an ellipse), and subtracting the ellipse representing the cross-sectional area of the centrum. If the rotation axis is in the centre of the segment, then r± is the distance from the centre of the segment to the line of force, or the outside of the segment; this is equivalent to half of the width of the segment (w/2). Therefore:

j* _ V*xs A* muscle AW )

2

The impact velocity and impulse are related to u>, rotational velocity, and a, rotational acceleration. OJ is additive along the tail, so the velocity increases from segment to segment (summation of velocities, Hildebrand and Goslow

2001). u) and a can be related to I, T, and the angle through which each segment moves, 8.

w = dt J\dt y \dt)

2(doi\

Rearranging for w gives a> = —.

/ do) \ ldco\ 26 CD = t\ , SO t Kdt J t

dco 26 — = a , so ta = dt t 107 2d Rearranging for t gives t 2 = — a T T = la , a = —

_ 2 261 . 261 So t = —, and t = J— . T \ T

Then t can be substituted into u> = 20/t to express GO in terms of 0,1, and T.

26 co = 261

co2 =26<->/j 2 T 261

I26T co = .

Because u> is additive along the tail segments, u)C|Ub is:

26ATA , 26BTB _ 20^ wciub = -'—^^ + -1—— + - -'to// V *tail-FCl V "'to!7-FCl-FC2-fC3-FC4-FC5-FC6-FC7-FC8-FC9-FC10-FCH

5.4.7 Analysis of a small knob and tail, ROM 784/UALVP 47273

ROM 784 includes all free caudals, except for the final, transitional free caudal, and the entire tail club. UALVP 47273 is a partial tail club with similar proportions to ROM 784. Not all measurements are possible on each specimen, but a composite of the two specimens, with parts scaled to an equivalent size, A

•H**-p: ... ! * ' ' • 111 •»?•

•*%

Figure 5.7: Diagrammatic representation of composite tails used in this study. A) ROM 784/UALVP 47273 composite tail. ROM 784 elements are indicated by light grey. UALVP 47273 elements are indicated by dark grey. The black vertebra represents the transitional vertebra not found in either specimen, but found in ROM 1930. Its presence is inferred by the gap at this location in ROM 784. The light purple area represents the free caudal tail frustum, and the dark purple area represents a single free caudal tail segment. The orange area represents the transitional tail frustum, and the pink area represents the handle volume. B) UALVP 16247 reconstructed tail. Only the knob is preserved (dark grey); the rest of the tail is reconstructed from measurements of ROM 784 (black). C) AMNH 5245/ROM 788 composite tail. AMNH 5245 elements are light grey, ROM 788 elements are dark grey, and elements reconstructed from ROM 784 are black. can be used to model movement in the tail (Fig. 5.7).

5.4.1.1 Determining the volume of the tail bones and muscles

Calculating the volume of bone and muscle in the tail requires three steps:

1) calculating the volume of the moveable, free caudal portion of the tail, 2) calculating the volume of the handle, and 3) calculating the volume of the knob.

For this study, each vertebra and the subsequent disk space represent a segment (Fig. 5.7). Ankylosaurid vertebrae each have an approximately circular centrum in anterior view, with width exceeding height slightly in ROM 784, and height exceeding width in ROM 1930. This variation may be related to taphonomy, ontogeny, individual variation, or sexual dimorphism. The neural spine and haemal spine are approximately equal in height. Centrum height and width, neural spine and haemal spine height, and transverse process length decrease posteriorly, whereas centrum length increases posteriorly.

Based on the above reconstruction of ankylosaur caudal muscles, the volume of these muscles was much greater that those of their associated neural and haemal arches. The volumes of these osseous structures are difficult to estimate and will be ignored, and only the volume of the centra will be used for calculating segmental and total muscle volume. The shape of each segment

(centrum + subsequent disk space) can be represented by a truncated cone with an elliptical base (an elliptical frustum). The equation to determine the volume of any pyramidal frustum is:

110 Where A^ is the area of the base of the pyramid, A2 is the area of the plane truncating the pyramid, and h is the height from Ai to A2. The area of an ellipse is:

A = ixDxD2

Where Di and D2 are the major and minor axes of the ellipse. To calculate the volume of the vertebral segment, segment length I (centrum length + length of disk space), width D-i, and height D2must be known. Volume is then calculated as:

Unfortunately, not all of these parameters are known for every segment in ROM

784. Some of the vertebrae are crushed and distorted, which yields measurements that do not necessarily decrease from one vertebra to the next.

To compensate for these problems, an 'ideal' ROM 784 is constructed.

Measurements of centrum height, width, and length were plotted as a scatterplot and slope of the lines of best fit calculated. The slope of each line is then used to calculate new heights, lengths, and widths, which are then used to calculate volume (Table 5.1).

Ill Table 5.1: Actual and ideal values for dimensions of the centra in ROM 784, in mm. Ideal segment length calculated using y = -0.29x + 76.17; ideal segment height calculated using y = -1.35x + 73.94; ideal segment width calculated using y = -2.00x +91.36. Vertebral Measured Inter­ Total Ideal Centrum Ideal Centrum Ideal Volume Segment Centrum vertebral Length Segment Height Segment Width Segment of # Length Cartilage Length Height Width elliptical Space frustrum Length (mm3) (approx­ imate) 1 60.49 20 80.49 75.88 - 72.59 63.05 89.36 1514000 2 59.66 20 79.66 75.59 93.12 71.25 82.6 87.36 1447000 3 48.33 20 68.33 75.30 - 69.90 - 85.36 1381000 4 48.64 20 68.64 75.02 87.86 68.55 - 83.36 1317000 5 54.44 20 74.44 74.73 68.41 67.21 - 81.36 1255000 6 59.06 20 79.06 74.44 85.19 65.86 - 79.36 1194000 7 49.49 20 69.49 74.15 68.66 64.52 - 77.36 1135000 8 58.86 20 78.86 73.86 - 63.17 67.03 75.36 1078000 9 - - - 73.57 - 61.82 52.64 73.36 1022000 10 - - - 73.28 - 60.48 - 71.36 9682000 11 - - - 72.99 74.91 59.13 62.64 69.36 9159000 72.70 57.79 67.36

Calculating the volume of muscle in the flexible portion of the tail is more complicated. In crocodilians, the musculature of the tail is to a certain extent limited by the vertebrae, but the cross-sectional profile of the tail changes greatly from anterior to posterior (Cong 1998). A conservative reconstruction of the muscles of the tail of ankylosaurids would have an elliptical cross-sectional outline, with none of the muscles bulging past the transverse processes or neural and haemal spines. If this is the case, the shape of the tail as a whole would mimic the shape of the centra, and the tail can be modeled as a series of truncated elliptical cones just like the centra. This reconstruction ignores the muscles of the pelvis that continue caudally.

The heights of the neural and haemal spines, and lengths of the transverse processes, were measured in ROM 784 and ROM 1930. As before, measurements for all elements could not be obtained as some vertebrae were missing some or all of these elements. The 'ideal' neural spine, haemal spine, and transverse process values were calculated as above. The height of a tail

112 segment is the sum of the heights of the haemal spine, centrum, and neural spine. The width of a tail segment is the sum of the width of the centrum and the length of both transverse processes. The volumes were calculated as for the centra (Table 5.2). To obtain the volume of the muscles, the volume of the centra is subtracted from the total volume of the tail.

Table 5.2: Actual and ideal values for dimensions of the tail in ROM 784, in mm. Ideal neural spine and haemal spine heights calculated using y = -4.22x + 115.19, transverse process length calculated using y = -10.46x + 113.36. Ideal total height equals ideal neural spine plus centrum plus haemal spine heights. Ideal total width equals ideal centrum width plus two ideal transverse process lengths. Vertebral Measured Ideal Ideal Measured Ideal Ideal Ideal Volume of Segment neural neural haemal transverse transverse total total elliptical # spine spine spine process process height width frustrum height height height length length (mm3) 1 104.91 110.97 110.97 130.77 102.90 294.53 295.16 19580000 2 99.01 106.75 106.75 111.36 92.44 284.75 272.24 17330000 3 94.62 102.53 102.53 57.54 81.98 274.96 249.32 15190000 4 106.65 98.31 98.31 43.11 71.52 265.18 226.40 13180000 5 108.36 94.09 94.09 54.91 61.06 255.39 203.48 11290000 6 101.57 89.87 89.87 41.05 50.6 245.61 180.56 9515000 7 88.5 85.65 85.65 31.98 40.14 235.82 157.64 7859000 8 84.29 81.43 81.43 32.41 29.68 226.04 134.72 6321000 9 66.81 77.21 77.21 46.38 19.22 216.25 111.80 4897000 10 64.91 72.99 72.99 0 8.76 206.47 88.88 3667000 11 68.96 68.77 68.77 0 0 196.68 69.35 3004000 64.55 64.55 0 186.89 67.35

The handle vertebrae are not as easily represented geometrically, and measurements of the heights, widths, and lengths of the centra were not possible in ROM 784 because the specimen is partially embedded in matrix. However, a partial tail club with similar vertebra and knob proportions (UALVP 47273) has been CT scanned. This specimen can be scaled to the size of ROM 784, and measurements of the volume of this specimen can be substituted for ROM 784.

CT scan data was imported in OsiriX. CT scan slices were exported as

TIFF files at 10 mm intervals (plus an additional slice representing 5 mm), totaling a length of 475 mm. These images were analyzed using imageJ.

Regions of interest (ROIs) were traced manually based on density contrasts in the image. ROIs for the handle vertebrae included the total cross-sectional areas, 113 and the areas of the neural arch plus neural canal, neural canal, centrum, haemal arch, and haemal canal. The total cross-sectional area is multiplied by slice thickness to find the volume of each slice, and these results are then summed to find the volume of the club. Volumes of the compact neural and haemal arches, cancellous centra, and 'empty' neural and haemal canals can be calculated in the same manner. Using this method, the total volume of the handle vertebrae in UALVP 47273 is 1025 cm3.

ROM 784 is slightly larger in than UALVP 47273. The proportions of the knob cannot be used to scale UALVP 47273 to ROM 784, because knob size does not seem to be correlated with vertebra size (see Chapter 2).

Measurements of the length of the neural spine on each handle vertebra were plotted on a scatterplot, and the slope was calculated. The slope was similar for both ROM 784 (-4.01) and UALVP 47273 (-4.56), and so the length of the neural spine was chosen as an appropriate scaling measure (Fig. 5.8). ROM 784 is

109% the length of UALVP 47273 using this measure (Table 5.3). The width of the knob of ROM 784 is 107% that of UALVP 47273.

120 Figure 5.8: Graph comparing the length of the neural spine of the handle vertebrae in ROM 784 100- ROM 784 and UALVP 47273. ROM 784 is y = -4.01x + 104 represented by the solid line and squares. UALVP 47273 is represented by the | ad dashed line and diamonds.

6M c 0) 4Q\ UALVP 47273 y = -4.56x + 102

201

2 4 6 8 10 12 Vertebra Number

114 Table 5.3: Comparison of the handle vertebra neural spine length. Handle vertebra neural spine length (mm) Mean Knob (mm) width (mm) 1 2 3 4 5 6 7 8 9 ROM 784 101.14 97.27 86.07 89.61 82.65 76.26 77.08 88.31 55.52 83.77 166 UALVP 47273 - - - - 84.16 70.98 59.4 79.11 57.3 70.19 155 ROM 784 as a % of UALVP 47273 101.83 93.08 77.06 89.58 103.21 92.95 93.37 UALVP 47273 as a % of ROM 784 98.21 107.44 129.76 111.63 96.89 108.79 107.10

The length of the club of ROM 784, from the anterior of the first handle vertebra to the posterior terminus of the knob, is 127.2 cm. The measured length of the handle, plus the length of the knob, in UALVP 47273 is 70.8 cm. Scaling by 1.09 gives a length of 77.2 cm. Scaling the measured volume by 1.093 gives a volume of 1330 cm3. UALVP 47273 is an incomplete club; subtracting 77.2 cm from a total length of 127.2 cm gives a missing length of 50.0 cm. The average cross-sectional area of each slice is 21.15 cm2, which scaled to ROM 784 is

25.13 cm2. Multiplying this average by 50.0 cm provides an estimate of 1260 cm3 for the volume of the missing area in UALVP 47273. This actually underestimates the likely missing volume, because in ROM 784 the centra of the first two handle vertebrae are slightly larger than the rest of the centra. Summing the scaled up volume of the measured portion of UALVP 47273 (1220 cm2), and the estimated volume of the missing portion (1260), yields a bone volume of 2470 cm3.

ROM 784 is probably missing a vertebra in the middle of the series. ROM

1930 has a vertebra with a transitional morphology between the free caudal vertebrae and the handle vertebrae, and zygapophyses from a handle vertebra

115 are fused to the neural arch of this transitional vertebra. A large gap between the last free caudal vertebra and the first handle vertebra, and the fact that the prezygapophyses of the first handle vertebra are broken, suggests that a transitional free caudal vertebra was present in ROM 784 as well. To model this vertebra, an additional, twelfth 'ideal' vertebra was constructed using the 'ideal' free caudal equations, and the volume was calculated as 865 cm3.

Calculating the total volume of the handle, with muscles reconstructed, is more difficult than with the free caudals. There are fewer osteological correlates for muscle attachments in the handle vertebrae. As discussed in Chapter 4, the width between the medial sides of the two major knob osteoderms in dorsal view must have been the maximum width of the handle muscles (Fig. 4.4). A CT scan cross-sectional slice of the handle of UALVP 47273 provided the basis for reconstructing the musculature. This reconstruction was then measured using

ImageJ, giving a cross-sectional area of 60.0 cm2 (71.3 cm2 scaled to ROM 784).

The first two handle vertebrae in ROM 784 are larger than the more posterior handle vertebrae, and have small bumps where the transverse processes are located in the more anterior caudals. To approximate the musculature of the free caudals tapering onto the handle, a frustum from the anterior of the transitional free caudal vertebra to the posterior of the second handle vertebra was calculated. The length of the first two handle vertebrae in

ROM 784 is 18.99 cm. Using this length, the 'ideal' dimensions for the transitional free caudal musculature, and an area of 71.3 cm2 as the top of the frustum, a volume of 7640 cm2 was calculated. The remaining length of the club is 108.20

116 cm. Subtracting the length of the knob (23.3 cm in UALVP 47273, scaled to 25.4 cm) gives the remaining length of handle for which total volume must be calculated. The handle vertebrae do not taper much posteriorly, and for the purposes of this study it is assumed that the total volume of the tail in the handle did not taper posteriorly either. Therefore, the cross-sectional area of 71.3 cm2 can be multiplied by the length to obtain a volume of 4970 cm2.

With the total volumes of the various tail segments, and the volumes of the vertebrae, the volume of muscle can be calculated. The total volume of the tail

(excluding the knob) is 12610 cm3, and the total volume of the vertebrae is 3250 cm3. Subtracting the volume of the vertebrae from the total volume gives a muscle volume of 9357 cm3.

5.4.1.2 Determining knob volume

The knob of ROM 784 is not easily modeled using simple geometry, and measurements of all dimensions could not be obtained because the knob is partially embedded in matrix. Instead, the volume of UALVP 47273 was calculated by tracing the area of CT scan slices in ImageJ and multiplying by slice thickness (1 mm). ROIs included the entire club, each osteoderm, the cancellous area in each osteoderm, and areas of exceptionally low density. The volume of the knob of UALVP 47273 is 1551 cm3, and the scaled volume is 2008 cm3.

It is difficult to reconstruct with certainty the size and shape of the probable keratinous sheath that would have covered each of the knob

117 Figure 5.9: Partial skull of a Siberian ibex (Capra sibirica), a bovid, oblique left lateral view; horn sheath has been removed from the left horn core. Horn sheath morphology does not necessarily closely match the underlying horn core.

osteoderms. In many horned ungulates, the morphology of the horny sheath does not closely match the size and shape of the inner bony horn core (Fig. 5.9)

(Picard et al. 2006). Keratinous coverings in Alligator mississippiensis osteoderms appear to conform more closely to the shape of the underlying osteoderm, and particularly augment the shape of the keel, if present (Vickaryous and Hall 2008). A specimen of the basal thyreophoran Scelidosaurus with preserved integument indicates that thyreophoran osteoderms were covered in a thin layer of skin or horny keratin (Martill et al. 2000). The size of the keratinous sheath probably does not greatly affect the rotational inertia of the tail, and is not included in the following calculations. However, the size and shape of the sheath would play a role in absorbing stress and strain upon impact; Snively and Cox

(2008) found that the thickness of keratin covering a pachycephalosaur dome reduced the strain in the bone during impacts. This will be considered in the finite element analyses in Chapter 6.

118 5.4.1.3 Angle of articulation between free caudal vertebrae

An important variable for determining forces, velocities, and impulses is the amount of rotation possible between each free caudal vertebra. Ankylosaurid free caudal vertebrae had limited vertical motion, but were capable of lateral motion. The maximum angle of rotation is the maximum left and right divergence from midline. The maximum half angle of rotation is the maximum divergence in one direction from the midline. Ideally, a complete specimen with all or most vertebrae preserved and prepared out of the matrix could be manipulated to manually measure the maximum half angle of rotation between each vertebra.

Whereas ROM 784 preserves almost all of the caudal vertebrae, it is embedded partially in matrix and the vertebrae cannot be moved to measure angles.

Several other specimens have two or three vertebrae in sequence and prepared out of the matrix (AMNH 5404, ROM 1930), but in these specimens the zygapophyses are not complete between vertebrae, and so the maximum half angle of rotation could not be determined. An alternative method for determining the half angle of rotation is presented here.

Stevens and Parrish (1999) found that the synovial capsules of the pre- and postzygapophyses or extant birds constrained the amount of movement between each vertebral joint. In birds, zygapophyseal facets must overlap by approximately 50%. Dzemski and Christian (2007) examined flexibility in ostrich

(Struthio camelus) necks and skeletonized necks of camels (Camelus bactrianus) and giraffes (Giraffus camelopardalis), and found that maximum lateral flexion is limited by the overlap between the zygapophyseal joint facets. In

119 lateral flexion of ostrich necks, the overlap of the joint facets was equivalent to the rim of one facet covering between one eighth and one quarter the long diameter of the corresponding facet. Muscles along the neck reduced the lateral flexion if a long segment of the neck was flexed. Extreme lateral flexions of the necks of living ostriches were close to the values obtained from neck skeletons.

Dzemski and Christian (2007) found that the maximum intervertebral lateral flexions in the necks of the ostrich and camel, which both have very flexible necks, are below 25°.

The studies by Stevens and Parrish (1999) and Dzemski and Christian

(2007) provide a guideline (the amount of contact between the zygapophyseal joints) by which maximum angles of rotation can be determined in ankylosaur tails. TMP 2007.20.80, an isolated free caudal vertebra, has complete prezygapophyses and postzygapophyses. In ROM 784 and ROM 1930, each successive free caudal vertebra is approximately 3% smaller in width than the preceding vertebra. A dorsal photograph of TMP 2007.20.80 was rotated by 0°,

5°, 10°, 15°, 20°, and 25°. The axis of rotation follows that of Snively and Russell

(2007), at approximately the midpoint between the prezygapophyses. The photograph was scaled by 103% to create a preceding vertebra. The rotated original photograph and enlarged photograph were overlain so that the zygapophyses articulated (Fig. 5.10). The prezygapophyses of the rotated image and the area covered by the postzygapophyses of the enlarged image were meaured in ImageJ (Table 5.4).

120 Figure 5.10: Determining the maximum angle of rotation in ankylosaurid free caudal vertebrae. In the left column, a dorsal view of TMP 2007.20.80 is on the left (anterior is to the right), and a 3% larger copy is on the right. The vertebrae are separated by a 2 cm gap representing the intervertebral disk space. The left vertebra is rotated from 0 to 25 degrees, in 5 degree increments, from A to F. The right column shows the articular faces of the prezygapophyses in light grey, and the area covered by the postzygapophyses in darker grey, for each rotation. Scale equals 5 cm. Table 5.4: Area of zygapophyseal overlap (mm). Angle Area of Area of Area of Area of Left post- Right post- left pre- right left right zygapophysis zygapophysis zygapo- pre- post- post- area/left area/right physis zygapo- zygapo- zygapo- prezygapophysis prezygapophysis physis physis physis area, % area, % 0 7.116 7.844 2.616 1.389 36.76 17.71 5 7.116 7.844 2.650 1.630 37.24 20.78 10 7.116 7.844 1.845 2.240 25.93 28.56 15 7.116 7.844 1.337 2.336 18.79 29.78 20 7.116 7.844 0.419 3.152 5.888 40.18 25 7.116 7.844 0 4.670 0 59.54

The original photograph is not perfectly aligned, so that the non-rotated original photo and enlarged photo do not articulate perfectly. This explains why the area in contact in the right zygapophyses is less than 25% when the angle of rotation is zero. However, this method provides an effective way to estimate the maximum angle, even if the photograph is not perfectly aligned, or if the specimen is slightly taphonomically distorted. The maximum lateral flexion of the caudal vertebrae are estimated to have been between 5° to 10° from the neutral position, and would have almost certainly been less than 20°.

5.4.1.4 Calculating T, I, and a).

Table 5.5 summarizes the volume of bone and muscle, proximal cross- sectional area of muscle, mass of bone and muscle, total mass, length, and total mass per unit length for each segment of the tail.

122 Table 5.5: Summary of volumes, areas, and masses for ROM 784/UALVP 47273. Segment Muscle Muscle Muscle Bone Bone Total Length Mass Cross- Volume Mass Volume Mass Mass (g) (cm) per unit 3 3 Sectional (cm ) (g)(P = (cm ) (g)(P = length Area (cm2) 1.0 g/ 1.98 g/ (g/cm) cm3) cm3) FC1 2526 18070 18070 1514 2998 21070 7.59 2776 FC2 2239 15880 15880 1447 2864 18740 7.56 2479 FC3 1965 13810 13810 1381 2734 16550 7.53 2197 FC4 1706 11860 11860 1317 2607 14470 7.50 1929 FC5 1460 10030 10030 1255 2484 12520 7.47 1675 FC6 1228 8321 8321 1194 2364 10680 7.44 1435 FC7 1011 6724 6724 1135 2247 8972 7.42 1210 FC8 806.7 5243 5243 1078 2134 7377 7.39 998.7 FC9 616.7 3874 3874 1022 2024 5898 7.36 801.7 FC10 440.7 2699 2699 968.2 1917 4616 7.33 629.9 FC11 299.5 2089 2089 915.9 1813 3902 7.30 534.6 Club 273.1 9357 9357 5346 10580 19940 134.48 148.3

Rotational inertia (Table 5.6) for each segment was calculated using

I = f—& L_/> where p is the mass per unit length, calculated in Table 6. L2 and l_i change for each segment and each I. cy A 4 O u U o LL. u. li. O •* £ CM o o o Li. u. - o H H- CM" "S u o — u. su. u.

§1 o

CO CD t-- CO l-EO d CD f~ co •* ™o o°.o P"o CD Itail- J FC6- I FC 9 CO T- t *- •*- *- *~ FC5 - FC2 - •FC 8 CO o CO Li. •FC4 - •FC7 - 00 o ( CO in •

T— m en •*,„ C(-~O CO o> «b To - «?b j *~ T- l~*."b- T~ T— T CM T-

•* t^ en CM 00 ^r CO O5,, O IT) "* „ ^t- Pb Pb •*"o P<°„b 56 4 - °*o S2 T *- cri CO T- "tf t- CM T- CM •<- CM

r^ CM O O) co CO CO LO^ r^ CO CM ^ -* „ l«- C-

T o Pb Pb T O ^b 97 3 m 1- ^b ^-.b J11 *- *- CM T- CO T- CD r- CM T- CM i- CM

CO CM CO T— o o en w co CO CO CM o „ CM CMm m ± r* -v CO"^ •* m WOO Pb Pb °°P mo °o ^ o m r- CO T- N- T- CO T- CO T- CO

•* O) ^J- t CM CD CO O) 0•*0 co (^ CO „ CDm CM 05 r- o CO ^"b Pb ^o •^ o «Tb ""b Po ~"b ^:b P°o o CM i- T- ^~ CO T- CD T- CO T- CJ> T- T~ T— *- ^~ CO r- •* T" •*

CO CD o> o IT) CD CO CO" co'; CO" -5 §1 *:b

•* Tf m CO h- CO O) in lO CO CM CM lO Tf t-*»; o> °™»o ^b Oi"o Pb "o Pb <>i"b ^"(^ o 47 5 IT) T- CO T- *~~ *~~ T~~ T~ *~ ^~ ^~ T~ T— •* T- in i- in

t O CO CM •*^ o> ^ o CO Pb Pb CM T- CM CM T- CM i-

CD E CM CO D) o J3 (D O o O O O o w o o o o o U- LL o o o h-

124 Torque (Table 5.7) is calculated for each segment using

2 2 T = \ *sA muscle A ) carpenter et al. (2005) used 39 N/cm and 78 N/cm as the upper and lower bounds for the range of forces that muscles can exert. Snively and Russell (2007) note that the amount of force a muscle can exert is related to its cross-sectional area and length, the geometry of muscle fibers, and the composition of muscle fibers. The of fibers associated with ankylosaurid tail muscles is difficult to assess, and the body temperature of ankylosaurids is unknown. Seebacher (2003) suggested that ankylosaurs did not evolve endothermy, whereas Gillooly et al. (2006) provided evidence that most large dinosaurs were inertial homeotherms. For the purposes of this study, it is assumed that ankylosaurids had muscle physiology comparable to those of extant homeotherms.

Muscle volume, fiber length, and pennation angle can be used to estimate the physiological cross-sectional area (PCSA) of a muscle, which relates to muscle force (Fukunaga et al. 2001). The force a muscle produces per unit area is its specific tension (ST), and specific tension multiplied by the PCSA yields the contraction force of the muscle (Fukunaga et al. 2001). It is impossible to estimate most of the factors involved in calculating PCSA for fossil taxa, but

PCSA is probably close to anatomical cross-sectional area (ACSA) in fusiform muscles (Snively and Russell 2007). Muscle force can thus be represented by the following equation:

F = ACSA* ST Specific tension can be estimated from studies of extant vertebrates, as it is relatively uniform in vertebrate muscle that is shortening by concentric contraction (Snively and Russell 2007). Specific tension has been found to range between 15 to 24 N/cm2 in a variety of extant vertebrates (Close 1972; Keshner et al. 1997; Bamman et al. 2000; Maganaris et al. 2001). Ankylosaurid muscle forces (Table 5.7) are calculated using 20 N/cm2 as a typical specific tension value, and are also calculated using Carpenter et al.'s (2005) values of 39 N/cm2 and 78 N/cm2 for the purposes of comparison with Stegosaurus.

Table 5.7: Torques for ROM 784/UALVP 47273. Segment Muscle Force Force Force Link half Torque Torque Torque cross- (half of (half of (half of width, at base at base at base sectional muscle muscle muscle w/2 (m) of link of link of link area at cross- cross- cross- F(w/2), F(w/2), F(w/2), proximal section sectional section 20 39 78N end of al area area al area N/cm2 N/cm2 (Nm) segment multiplie multiplied multiplie (Nm) (Nm) (cm2) dby20 by 39 N/ dby78 N/cm2) cm2) N/cm2 FC1 2526 25260 49260 98520 0.1476 3728 7269 14540 FC2 2239 22390 43650 87310 0.1361 3047 5942 11880 FC3 1965 19650 38320 76640 0.1247 2450 4777 9555 FC4 1706 17060 33260 66520 0.1132 1931 3765 7530 FC5 1460 14600 28470 56940 0.1017 1485 2897 5793 FC6 1228 12280 23950 47910 0.0903 1109 2162 4325 FC7 1011 10110 19710 39410 0.0788 796.5 1553 3106 FC8 806.7 8067 15730 31460 0.0674 543.4 1060 2119 FC9 616.7 6167 12030 24050 0.0559 344.7 672.2 1344 FC10 440.7 4407 8593 17190 0.0444 195.8 381.9 763.7 FC11 299.5 2995 5841 11680 0.0347 103.9 202.6 405.1 Club 273.1 2731 5325 10650 0.0337 91.95 179.3 358.6

26T The u) term is calculated using w •• . The sum of the u> terms (Table 5.8) is

26ATA 2dBTB 2eLrL

Ltail ltail-FC\ ltail-FCl-FC2-FC3-FC4-FC5-FC6-FCT-FC8-FC9-FCl0-FCU

Ankyiosaurids may have initiated a taii swing from the neutral position of the tail extended straight from the hips and without any lateroflexion between the caudal vertebrae. However, a more forceful impact would be achieved if the tail was swung from the maximum deflection of one side to the maximum deflection on the other side. Using 7.5° as an average half angle of articulation, the angle of articulation between each free caudal vertebra was 15°. An ankylosaurid tail could probably swing through at least 100°.

Table 5.8: Segment angular rates of movement. The angle io f articulation is 15 degrees, or 0.2618 radians. Torque at Torque at Torque at Cumulative base of base of base of moment of link link link inertia (sum sequence sequence sequence of columns w term u) term w term (Table (Table 5.7) (Table 5.7) in Table (rad/s) (rad/s) (rad/s) Segment 5.7) (20N) (39N) (78N) 5.6) (20N) (39N (78N) FC1+2+3+4+5+6 +7+8+9+10+11+ club 3728 7269 14540 6411 0.5518 0.7730 1.093 FC2+3+4+5+6+7 +8+9+10+11+ club 3047 5942 11880 5475 0.5398 0.7564 1.070 FC3+4+5+6+7+8 +9+10+11+club 2450 4777 9555 4685 0.5232 0.7333 1.037 FC4+5+6+7+8+9 +10+11+club 1931 3765 7530 4017 0.5016 0.7031 0.9944 FC5+6+7+8+9+ 10+11+club 1485 2897 5793 3453 0.4746 0.6654 0.9410 FC6+7+8+9+10+ 11+club 1109 2162 4325 2973 0.4419 0.6196 0.8762 FC7+8+9+10+11 +club 796.5 1553 3106 2564 0.4033 0.5655 0.7998 FC8+9+10+11+ club 543.4 1060 2119 2212 0.3586 0.5029 0.7112 FC9+10+11+club 344.7 672.2 1344 1908 0.3075 0.4313 0.6100 FC10+11+club 195.8 381.9 763.7 1643 0.2498 0.3504 0.4955 FC11+club 103.9 202.6 405.1 1409 0.1964 0.2755 0.3896 club 91.95 179.3 358.6 1202 0.2001 0.2807 0.3970 Total 4.7487 6.6312 9.3780

The angular rate of movement of the club (u)ciub) is between 4.75 rad/s and

9.38 rad/s. The length of the tail from the anterior end of the first free caudal vertebra, to the posterior end of the knob, is 216.4 cm. If a tail club was swung laterally, the impact site would not be at the posterior end of the knob, but somewhere along the lateral edge of one of the major plates of the knob. These osteoderms are sharply keeled laterally, and the maximum width is at approximately 15.69 cm from the posterior tip of the knob. Using this as an impact site, the impact site is 200.7 cm from the anterior face of the first free caudal. With this, the impact velocity of the club can be calculated:

Yimpact ~ ^club'"

= (4.75 rad/s)(2.01 m) = (6.63 rad/s)(2.01 m) = (9.38 rad/s)(2.01 m)

= 9.6m/s =13.3m/s = 18.9 m/s

This is much larger than the values obtained for Stegosaurus (8.7 - 12 m/s) by

Carpenter et al. (2005).

Using the mass of the club segment (19.94 kg, Table 1), the impulse delivered by the club can be calculated:

/ =m V u club '"club' impact = (19.94 kg)(9.56 m/s) = (19.94 kg)(13.3 m/s) = (19.94 kg)(18.9 m/s)

= 190 kg m/s = 266 kg m/s = 376 kg m/s

Carpenter et al. (2005) assumed a stopping time of 1/3 s. Using this, the maximum force exerted on the target can be calculated, as the impulse/interval.

17 _ ^club max stop

= (190 kgm/s)/(0.333s) = (266 kgm/s)/(0.333s) = (376 kgm/s)(0.333 s)

= 571N =797N =1127N

The stress exerted by the impacting club is Fmax over the area of impact.

The site of impact is the lateral keel of one of the major knob osteoderms. The amount of area involved in the impact can vary. If the sharpest part of the keel is the site of impact (height = -0.20 cm), and 1 cm of length is involved, then the area of impact is 0.20 cm2.

F impact A

impact

= 571 N/0.20 cm2 = 797 N/0.20 cm2 = 1127 N/0.20 cm2

= 2900N/cm2 = 4000 N/cm2 = 5600 N/cm2

5.4.1.5 Sensitivity analyses

There are several factors that could affect the results in ROM 784/UALVP

47273 that should be examined:

1. Bone mass. Differences in the density of cancellous and compact bone

will affect the mass estimates for each segment, and in particular the

mass of the tail club segment.

2. Muscle reconstructions. Differences in the amount of muscle

reconstructed affects the mass of each segment and the cross-sectional

area used to calculate torque.

3. Angle of articulation between free caudal vertebrae. The angle of

articulation is difficult to determine precisely, and may be too low or too

high. The maximum angle of articulation may also change posteriorly

along the tail.

4. Site of impact on club. The site of impact could be more posterior or

anterior on the knob. 5. Area of impact. The area of impact could be greater or smaller, depending

on the shape of the keratinous sheath, and whether the impact is along a

sharp or blunt keel, or on the rounded surfaces of the knob osteoderms.

6. Stopping time.

Each of these variables will be changed systematically with the composite

ROM 784/UALVP 47273 tail, and the results are summarized in Table 5.9.

Carpenter et al. (2005) used 1.98 g/cm3 when estimating segment mass.

Ankylosaurid handle vertebrae have cancellous centra and compact neural and haemal arches, and the knob is predominantly cancellous. To understand the role that bone mass plays in tail impact forces, a more accurate estimate of mass is needed. In the baseline analysis, the neural arch, haemal arch, and transverse processes were not modeled, and they are again excluded here. Additionally, changes in mass will affect the calculations for rotational inertia and impulse.

Because the tail club segment is so much larger than the rest of the tail segments, and because only the tail club segment is used to calculate impulse, it is reasonable to exclude the free caudal vertebrae from this sensitivity analysis.

The relative proportions of compact vs. cancellous bone in the handle vertebrae was determined by using ImageJ to calculate the cross-sectional area of the centrum and the neural and haemal arches in several transverse sections of the handle. The centrum was approximately 38% the total cross-sectional area of a handle vertebra. Extrapolating this to the handle as a whole (including the transitional vertebra), the volume of cancellous centra was 1251.10 cm3, and the volume of compact neural and haemal arches was 2086.27 cm3. Using average

130 co O 1*- o o o O O sz o co CD CD •<- CN IO N- T— T—

a> o I-- o o o O o co co CD o CD T- CO CD N- CO o co CM CO O CD 106 0 1330 0 371 0 39 9 3990 0 •*r co m CN CO rf CO TT 409 0 798 0

5 Si 165 0 339 4 1 337 0 362 9 233 0 330 0 220 0 35 6 266 0 293 0 572 0 76 2 28 6 2860 0 952 0

o si co 134 0 103 0 143 0 130 0 140 0 105 0 116 0 113 0 113 0 113 0 U. IE 65 2 92 2 87 1 x- O 376 0

T3 co co oo" co« •* T- CD O 1.0 1 94 8 72 5 65 2 92 2 61 6 99 1 79 8 UL E 46 1 t~- co 79 8 79 8 CM T"

CM CO CM" o» co co CD O 67 9 67 5 72 6 33 0 66 0 71 0 57 2 57 2 57 2 46 7 44 1 in io in T

06 2 o m m oo 34 1 30 7 37 6 37 6 37 6 37 6 37 6 44 6 47 7 21 7 43 4 46 7 co co

CD f- CM 15 3 31 6 24 1 33 7 30 7 33 0 -> E 21 7 20 5 CM CM 26 6 26 6 26 6 26 6 26 6

5 922 r~- in o o o o o CD CO en CD 11 0 15 5 14 7 22 5 24 2 -> -Q 22 0 23 6 ^ T~ T— T_ T_ T_ T—

co CD •* oq in CD m co oq co oo co oo cb o iri • z 20. 3 20. 2 CM CM ^ T~ T_ T_ T_ T— T_

T3 co •<* co "* ^ CO CO CO CO CO CD co •* •^ c\i co CO CO CO co CO 10. 9 15. 4 10. 3 16. 7 > E 7.7 0

5 co oo •5T co in in in in m 13. 4 10. 2 11. 0 11. 9 > -2 9.5 4 5.5 1 7.7 9 7.3 6 co C3> CD CD CD a> o>

CM co p co co co co co co oo hig h co r-i co co co co co co co 10. 8 11.7 5 5.4 1 7.6 6 7.2 3 05 CD CD CD ci CD' CD

•a co LO CD co i— CO co co co co co co co oo •<3- CD T— CD CD T- O co CD CD CD CD CD CD CD I E CD co iril~- I in co CD CD CD CD CD CD CD

CO CM CO in E in in in to in •<- O CD 3.8 8 I-2 rj: 2.7 4 5.4 8 3.6 6 5.9 5 LO If) •* •*' CO •<* •* •*'_ E •* •* .S co O) CO c I- (1) t- ts CD 7; CD 5 o a.c o S • CD CO m a.C O 8 o a. £ 2 P> o oq o o o CO o o in N- Fre e caudat < io CM N o handl e T^ CN E o o CM o o m S.5? = cp 35

131 density values for cancellous (1 g/cm3) and compact bone (2 g/cm3), also used by Snively and Cox (2008), yields a mass of 1251.10 g and 4172.54 g, respectively. The knob is varying densities of cancellous bone with a relatively thin layer of compact bone, and is here modeled as cancellous bone (1 g/cm3), giving a mass of 2008.32 g. The bone mass of the tail club segment is therefore

7432.36 g, which is less than the estimate using 1.98 g/cm3 as an average. The total mass of the tail club segment (including muscles) is 16.79 kg.

Reducing the mass of the tail club segment reduces the rotational inertia of this segment, and thereby increases the u)C|Ub to 5.12, 7.15, and 10.12 rad/s using 20, 39, and 78 N/cm2 of muscle force, respectively. This increases the impact velocity, which would have allowed ankylosaurids to strike more quickly.

Decreasing the tail club mass also decreases the rotational inertia, making the club easier to wield. However, the reduction in tail club segment mass also reduces the impact force and stress. There is a trade-off between reducing the mass of the tail club and increasing impact velocity, and increasing the mass of the tail club and increasing impact force and stress.

The amount of muscle that would have powered the tail is subjective. For the baseline analysis, it was assumed that muscles did not bulge outwards past the neural and haemal spines and the transverse processes. Reconstructing the muscles in this way allows the tail to be modeled as a larger frustum containing the centra frustum. However, the muscles may have been much larger than depicted in this reconstruction. The areas of two reconstructions (Fig. 4.3) were compared in ImageJ. TMP 1985.26.70 was reconstructed with conservative

132 musculature, and with bulging muscles. The cross-sectional area of the segment was 1310.25 cm2 for the conservative estimate, and 1872.26 cm2 for the larger estimate. The larger reconstruction is 143% the size of the conservative estimate. Using this value, the values of the cross-sectional areas of the tail segments in the baseline analysis can be scaled upwards, and the maximum force recalculated. The half width of each segment is left unchanged, because the reconstructed muscles do not necessarily bulge laterally past the transverse processes.

Increasing the amount of muscle powering the tail increases the cross- sectional area of muscle in each segment, which in turn increases the torque of each segment. The impact velocity, impulse, force, and impact stress are increased. Increasing the muscle also adds to the mass of the tail club segment

(23.6 kg), which increases the impulse. Even with larger muscles, the impact stress is still less than would be required to break bone.

In the baseline analysis, 15° was selected as a probable maximum angle of articulation between each pair of the free caudal vertebrae. Maximum angles of articulation of 5°, 10°, and 20°, and the effects of decreasing the amount of rotation posteriorly along the tail were examined, starting at 15° and moving to 0° at the articulation between the intermediate caudal and first handle vertebra. The degree of rotation between the free caudal vertebrae was calculated by graphing the rotation between the pelvis and first free caudal as 15° and the rotation between the intermediate caudal and first handle vertebra as zero, then taking the slope of the line (y=1.25x +16.25) and calculating the amount of rotation for

133 the vertebrae in between. The variables r, mciub, tstoP, A\mpacu and lciub are the same as those used in the baseline analysis.

Reducing the angle of articulation reduces u)ciub, which results in a lower impact velocity, impulse, impact force, and impact stress, whereas increasing the angle of articulation increases these factors. Decreasing the angle of articulation posteriorly along the tail also reduces these factors. If the vertebrae of the handle were not fused and rigid, and instead were able to rotate freely like the free caudal vertebrae, then impact velocity in the knob would increase, as would impulse, force, and stress. A flexible distal tail would be able to deliver more forceful blows than a rigid tail. It may be that the handle is necessary for postural reasons, to keep the knob elevated above the ground; Coombs (1995) suggested that the tail club did not drag. Or, the handle may be necessary for absorbing the shock of impact. Handles may represent a structural trade-off between maximum velocity and strength.

To examine the role of the handle in tail swinging, a hypothetical tail composed entirely of free caudal vertebrae is constructed. There are at least eleven, and probably twelve vertebrae in the handle. With eleven free caudal vertebrae and one missing transitional vertebra, the total number of vertebrae in

ROM 784 would have been 24. Even assuming that all of the vertebrae were free caudal vertebrae, the knob would still enclose the last two caudal vertebrae. This means there would be 23 segments (22 vertebrae and the knob). The length of the knob is 23.10 cm, which scaled to ROM 784 is 25.18 cm. Following the same procedure as for the baseline analysis, with the main changes being the

134 calculation of torque and rotational inertia for the extra segments, the value of

Wciub is determined to be 8.31 rad/s to 11.76 rad/s. The absence of a handle increases u)ciub, impact velocity, impulse, maximum force, and impact stress, and these values would likely be even greater if a more accurate tail with lengthening distal caudals could be reconstructed. However, the maximum stress of 7010

N/cm2 is still far less than the 104 N/cm2 necessary to fracture compact bone. A knob of ROM 784 size could be swung at the end of a tail without modified handle vertebrae. There may be an upper size limit for larger knobs, such as those in ROM 788 and AMNH 5245.

Changing the site of impact on the club changes the value of r, the distance from the base of the tail to the site of impact. For this analysis, the club impact points are assumed to be near the distal tip of the club (10 cm from the distal terminus, r = 2.06 m) and near the anterior margin of the knob (29 cm from the distal terminus, r = 1.87). Moving the impact site anteriorly decreases the impact velocity, impulse, maximum force, and impact stress. Moving the impact site posteriorly increases these factors. In both cases, the change is small. The location of the impact site may have affected the distribution of stress within the tail club, a possibility addressed in the finite element analyses below.

The area of impact is determined by the shape of the keratinous sheath that would have covered the knob. Because the shape of the sheath is unknown, the area of impact is speculative. The area of impact may have varied greatly depending on where the site of impact was, and what the knob was impacting. In the baseline study, an area of 0.20 cm2 was chosen as a reasonable

135 approximation. The bluntness or sharpness of the keel of the sheath would also affect the impact area.

The area of impact only affects the impact pressure, because impact pressure is Fmax/Aimpact- The larger the area of impact, the greater the area the force is distributed over, and the lower the impact stress. Reducing the area of impact from 0.20 cm2 to 0.10 cm2 doubles the impact pressure. An impact area of

2 cm2 yields a relatively low impact pressure. A blow from the sharp keel of an ankylosaurid knob would be more destructive than a blow from the more rounded distal end of the knob, or from the rounded faces of the knob osteoderms.

Altering the stopping time does not affect the impact velocity or impulse, but does affect maximum force (because this represents impulse over the stopping time) and pressure. Decreasing the stopping time from a third to a tenth of a second nearly triples the maximum force of the impact.

5.4.2 Analysis of a large tail and knob, ROM 788/AMNH 5245

5.4.2.1 Estimates of bone and muscle mass and volume

Whereas ROM 784 and UALVP 47273 are two of the smallest tail clubs,

AMNH 5245 and ROM 788 have the widest knobs encountered during the course of this research. A complete caudal series is not available in either of these specimens. ROM 788 includes the knob and most of the handle. AMNH 5245 includes the knob, some of the handle, and two anterior free caudals that may represent the first and second free caudals. An ideal free caudal series can be

136 constructed in the same manner as for ROM 784. Measurements of the vertebra were only possible in the first free caudal, so for the purposes of these estimates the proportions of the vertebrae in AMNH 5245 are assumed to decrease in the same manner as ROM 784. This is a reasonable assumption because the proportions of ROM 1930 (which has overall larger vertebrae than ROM 784) decrease at the same rate as in ROM 784. The width of the centrum in the first vertebra could not be calculated because it is broken. To estimate the width, the width:height ratio for each ROM 784 free caudal segment (vertebra and disk space) was determined, then used to calculate the width in AMNH 5245. To calculate the proportions of ideal AMNH 5245, the same m value is used as calculated for ideal ROM 784, and the b value is changed to the measurement of the first free caudal vertebra in AMNH 5245 (Table 5.10). It is also assumed that

AMNH 5245 has 11 free caudal vertebrae and one transitional vertebra, as in

ROM 784.

Table 5.10: Actual and ideal values of the centra in AMNH 5245, in mm. Ideal segment length is calculated using y = - 0.2889x + 109.5); ideal centrum width is calculated using y = -2.001x + 102.38; ideal centrum height is calculated using y = -1.346x + 88.91. Vertebral Width: Ideal Measured Ideal centrum Measured Ideal centrum Volume of segment Height in segment centrum width centrum height elliptical number ROM 784 length width height frustum (mm ) 1 1.231 109.5 102.38 102.4 88.91 88.91 3074000 2 1.226 109.2 - 100.4 - 87.56 2960000 3 1.221 108.9 - 98.38 - 86.22 2848000 4 1.216 108.6 - 96.38 - 84.87 2738000 5 1.211 108.3 - 94.38 - 83.53 2631000 6 1.205 108.0 - 92.38 - 82.18 2526000 7 1.199 107.7 - 90.38 - 80.83 2423000 8 1.193 107.4 - 88.38 - 79.49 2323000 9 1.187 107.1 - 86.38 - 78.14 2225000 10 1.180 106.9 - 84.38 - 76.80 2129000 11 1.173 106.6 - 82.37 - 75.45 2036000 12 1.166 106.3 - 80.37 - 74.10 1945000 106.0 - 78.37 - 72.76

137 The musculature of the tail in AMNH 5245 (Table 5.11) is modeled in the same manner as for ROM 784.

Table 5.11: Actual and ideal values for dimensions of the tail in AMNH 5245, in mm. Ideal neural spine height and haemal spine height are calculated using y = -4.220x + 192.85); ideal transverse processes length is calculated using y = -10.46x + 156.18. Ideal total height equals neural spine height plus haemal spine height plus centrum height. Ideal total width equals centrum width plus two transverse process lengths. Vertebral Measured Ideal Ideal Measured Ideal Ideal Ideal Volume of Segment neural neural haemal transverse transverse total total elliptical # spine spine spine process process height width frustrum height height height length length 2 (mm ) 1 192.85 192.9 192.9 156.18 156.2 474.6 414.7 65140000 2 188.6 188.6 - 145.9 464.8 392.2 50950000 3 184.4 184.4 - 135.5 455.0 369.23 46720000 4 180.2 180.2 - 125.0 445.3 346.4 42640000 5 176.0 176.0 - 114.5 435.5 323.5 38720000 6 171.8 171.8 - 104.1 425.7 300.5 34950000 7 167.5 167.5 - 93.62 415.90 277.61 31330000 8 163.3 163.3 - 83.16 406.11 254.69 27860000 9 159.1 159.1 - 72.70 396.33 231.77 24540000 10 154.9 154.9 - 62.24 386.54 208.85 21370000 11 150.7 150.7 - 51.78 376.76 185.93 18340000 12 146.4 146.4 - 41.32 366.97 163.01 5620000 142.2 142.2 - 30.86 357.19

The handle vertebrae of AMNH 5245 are partially embedded in matrix, but once again a tail club with similar vertebra and knob proportions (ROM 788) is available. In this case, AMNH 5245 is the less complete specimen, so it is scaled to the size of ROM 788. ROM 788 was CT scanned and the data was analyzed in ImageJ as for UALVP 47273.

AMNH 5245 is scaled to the size of ROM 788 using the length of the neural spine. Measurements of the length were plotted on a scatterplot and the slope was found to be similar in AMNH 5245 (y = 3.43x + 80.16) and ROM 788 (y

= 4.28x + 119.5). ROM 788 is 158% the length of AMNH 5245. The width of the knob of AMNH 5245 is 103% the width of ROM 788. As such, all of the free caudal vertebrae segments are scaled 158%.

138 Table 5.12: Comparison of the handle vertebra neural spine length (mm) in AMNH 5245 and ROM 788. Handle vertebra neural spine length (mm), see Figure 2.2 for explanation. Mean Knob (mm) width (mm) 1 2 3 4 5 6 7 8 9 10 AMNH - - - 100.22 - - 87.66 108.72 119.95 - 104.14 593 5245 ROM - - 100.10 142.57 158.75 137.69 158.04 222.47 124.69 133.95 147.28 572 788 ROM - - - 142.26 - - 180.29 204.63 103.95 - 157.78 96.5 788 as a % of AMNH 5245 AMNH - - - 70.295 - - 55.467 48.870 96.199 - 67.708 104 5245 as a % of ROM 788

The length of the preserved part of the club in ROM 788 is 126.8 cm, and eight vertebrae are visible. Ten vertebrae are visible in ROM 784. If ROM 788 had the same number of vertebra in the tail, and the average length of the vertebra is 9.32 cm (the length of the club minus the length of the knob, 74.58 cm, divided by 8 vertebrae), then the length of the tail club (handle + knob + missing vertebrae) would have been 146.8 cm. Included in the tail club segment for modeling purposes is the transitional vertebra, with an estimated length of

8.98 cm (scaled to 14.19 cm), giving a total tail club segment length of 160.9 cm.

The average cross-sectional area of each handle CT slice is 45.10 cm2, which multiplied by the length of the handle (94.58 cm) gives a bone volume of

4265 cm2. The transitional vertebra has a volume of 12120 cm2 when scaled to

ROM 788. The volume of the knob was partially measured using ImageJ as for

UALVP 47273; however, the knob was wider than the field of view of the scanner and the lateral edges of the knob osteoderms were not scanned. The missing portion of the knob osteoderms can be represented by an ellipsoid, where the volume is: V = —jxabc 3

Where a, b, and c are the three axes of the ellipsoid. The axes a and b (length and width) were measured by overlying a semi-transparent coronal CT section of

ROM 788 over a dorsal photograph of the specimen, and measuring the length and width of the missing part of each osteoderm in ImageJ. The height (c) of the missing portion of each osteoderm was measured from a transverse CT section in ImageJ. The measured volume of the knob was 20809.96 cm3, and the missing volume of the left and right osteoderms was 14120 cm3 and 18080 cm3, respectively. This gives a total volume of the knob of 53 000 cm3. The bone volume of the tail club segment (transitional free caudal vertebra + handle vertebrae + knob) totals 69390 cm3.

The muscle volume is calculated by determining the total volume of the tail club and subtracting the volume of the transitional free caudal and handle vertebrae. The maximum width of the handle muscles is the width between the major osteoderms of the knob, which for ROM 788 was measured as 19.48 cm in

ImageJ. In UALVP 47273, the average cross-sectional area of the reconstructed tail (muscles + vertebrae) was 60.00 cm2, and the width between the osteoderms was 10 cm. Assuming that the muscles in ROM 788 are proportionately larger, the average cross-sectional area of the tail is 116.9 cm2. Multiplying this by the length of the handle (94.58 cm) gives a tail volume of 11100 cm3, and subtracting the volume of the handle vertebrae gives a muscle volume of 6787 cm3. The transitional tail segment is 14500 cm3 (scaled to ROM 788), and subtracting the volume of the vertebra gives a muscle volume of 10780 cm3.

140 5.4.2.2 Calculating T, I, and co.

Torque, rotational inertia, and angular acceleration are calculated in the same manner as for ROM 784/UALVP 47273. The angular rate of movement of the club (u)ciub) is between 4.7569 rad/s and 9.3942 rad/s. The length of the tail from the anterior end of the first free caudal vertebra, to the posterior end of the knob, is 348.66 cm. If the impact site is located at approximately the maximum width of the tail (located roughly 20 cm anterior to the posterior terminus of the knob), then the impact site is 328.66 cm from the anterior face of the first free caudal vertebra. This is used to calculate the impact velocity. The mass of the club segment (154.97 kg, Table 5) is used to calculate the impulse delivered by the club. The maximum force is calculated using a stopping time of 1/3 s, and the impact stress is calculated assuming an impact area of 0.20 cm2 as for UALVP

47273/ROM 784. The results are summarized in Table 5.13.

Table 5.13: Velocities and forces for ROM 788/AMNH 5245. = l*>dub - 4.76 rad/s Wclub ~ '6.6 4 rad/s wciub 9.39 rad/s Velocity (m/s) 15.66 21.85 30.89 Impulse (kgm/s) 2427 3387 4788 Force (N) 7281 10160 14360 Stress (N/cm2) 36400 50800 71810 Stress (MPa) 364.0 508.0 718.1

5.4.3 Analysis of a mid-sized tail and knob, UALVP 16247

UALVP 47273/ROM 784 and ROM 788/AMNH 5245 represent the extreme ends of the range of widths in knobs measured in this study. Most tail clubs are around 40 cm wide, and two examples of average-sized tail clubs were

CT scanned (UALVP 16247 and TMP 1983.36.120). These are both fragmentary clubs with only the knob preserved in UALVP 16247 and a small fragment of the

141 handle in TMP 1983.36.120. As such, any estimates of lengths, volumes and masses will be more tentative for these clubs compared to UALVP 47273/ROM

784 and ROM 788/AMNH 5245. Nevertheless, an estimate can be made by

'extruding' the handle from the knob, and then reconstructing the free caudal vertebrae posteriorly to anteriorly using information from the 'ideal' ROM 784 vertebrae.

5.4.3.1 Estimates of bone and muscle mass and volume

UALVP 16247 represents an average-sized tail club knob. It is the most fragmentary specimen in this study, as it is an isolated knob lacking a handle.

However, a rough estimate of tail dimensions can be made for UALVP 16247, to provide estimates of impact forces for the most common knob size.

A CT scan of UALVP 16247 was measured as for ROM 788 and UALVP

47273. The volume of the knob was determined to be 6486 cm3. Some of the terminal handle vertebrae are visible in transverse section within the knob.

Several cross-sectional areas were traced and averaged to provide an estimate for the average cross-sectional area of the handle, which can be compared to the values obtained for UALVP 47273 and ROM 788 (Table 5.14).

Table 5.14: Cross-sectional area (cm2) of UALVP 16247 handle and comparisons with ROM 788 and UALVP 47273. Specimen Average cross- UALVP 16247 as a % of sectional area of other specimens the handle ROM 788 45.10 80.90% UALVP 47273 21.15 172.5% UALVP 47273 scaled to ROM 784 25.25 144.5% UALVP 16247 36.48 -

142 UALVP 47273/ROM 784 represents a more complete composite specimen than ROM 788/AMNH 5245. The value of UALVP 47273 scaled to

ROM 784 (144.5%) is thus used as the scaling factor for UALVP 16247. This can be used to scale the proportions of the free caudal vertebrae, handle vertebrae, and muscles in UALVP 47273/ROM 784 to reconstruct the missing elements in

UALVP 16247. The square root of 144.5% is used to determine the linear proportions of the vertebrae, using the ideal values of ROM 784, as well as the proportions of the tail segments.

The volume of the tail club can be calculated using proportions from

UALVP 47273 scaled to ROM 784. The length of the ROM 784 tail club is 127.21 cm, and the length of the handle (total club length minus the length of the knob) is 100.96 cm. The length of the handle in UALVP 16247 is therefore 121.35 cm, and multiplying by the average handle cross-sectional area gives a handle vertebrae volume of 4427 cm3. The estimated cross-sectional area of the handle tail segment in UALVP 47273/ROM 784 was 71.29 cm2, which scaled to UALVP

16247 is 103.01 cm2. The volume of the handle tail segment is therefore 12500 cm3.

5.4.3.2 Calculating T, I, and to.

The same method for calculating torque and rotational inertia is employed here as for ROM 784/UALVP 47273 and AMNH 5245/ROM 788. Tables including the mass, torque, and rotational inertia are not included, and the calculations of velocity, impulse, force and stress (Table 5.15) use the same equations as for

143 ROM 784/UALVP 47273, and AMNH 5245/ROM 788. The angular rate of movement of the club (wciub) is between 3.75 rad/s and 7.41 rad/s. The length of the tail from the anterior end of the first free caudal vertebra, to the posterior end of the tail club knob, is 269.96 cm. Choosing the maximum width of the knob as the impact site (18 cm anterior to the posterior terminus of the knob), the impact site is 251.96 cm from the anterior face of the first free caudal vertebra. The mass of the club segment is 35.94 kg. As for UALVP 47273/ROM 784, and

AMNH 5245/ROM 788, a stopping time of 1/3s is used to calculate force, and an impact area of 0.2 cm2 is used to calculate impact stress.

Table 5.15: Velocities and forces for UALVP 16247.

Wciub = 4.76 rad/s u)C|Ub = 6.64 rad/s u)ciub = 9.39 rad/s Velocity (m/s) 9^45 13.20 18.67 Impulse (kgm/s) 320.7 474.6 671.1 Force (N) 962.2 1424 2014 Stress (N/cm2) 4811 7119 10070 Stress (MPa) 48.11 71.19 100.7

5.5 Discussion

The inferred angular accelerations of ROM 788/AMNH 5245, UALVP

16247, and UALVP 47273/ROM 784 were similar because the proportions of the tail were all modeled from UALVP 47273/ROM 784. However, there was a great difference in impact velocities and forces because of the differences in mass and length of the tail club segment in each tail. Sensitivity analyses for the functional calculations show that the bone and muscle mass, the location of the impact site, the area of impact, and the stopping time influence the impact force and stress for each tail club. However, changing these parameters within biologically reasonable bounds does not appear to influence the calculated range of impact forces.

Ankylosaurids with large knobs could deliver more forceful blows than ankylosaurids with small knobs, despite lower tail rotational inertias in the latter.

Impact stress results for small clubs are similar to those found for Stegosaurus spikes. Carpenter et al. (2005) determined that a Stegosaurus spike could exert

360-510 N of force when swung, which they argue was more than enough to damage tissue and bone. They estimated a spike-tip impact area of 0.28 cm2, which would create an impact stress of 1300-1800 N/ cm2. In contrast, ROM 784 could exert a force of 797-1127 N, using the specific tensions used by Carpenter et al. (2005), and 571 N using a more realistic specific tension, creating an impact stress of 2900-5600 N/cm2. Carpenter et al. (2005) use -100 MPa (104

N/cm2) as the maximum shear strength of living cortical bone; Currey (2002) summarizes several papers which give values between 64 and 84 MPa for shear strength. It does not appear that a Stegosaurus spike could puncture bone, nor could the tail club in ROM 784. This seems reasonable, as the knob in ROM 784 is small in comparison to others. UALVP 16247 represents average knob width, and could impact with a force of 962-2014 N, and exert an impact stress of 4811

-10 070 N/cm2. Average-sized knobs may have been able to break bone during impacts. An ankylosaurid with the proportions of ROM 788/AMNH 5245 could create an impact force of 7281-14 360 N, an impact stress of 36 400 - 71 810

N/cm2, and could easily break bone during a tail club impact.

145 The handle reduces the maximum angular acceleration of the terminal tail segment, in comparison to a tail composed of free caudal vertebrae. This would have reduced the impact force, preventing damage to the distal caudal vertebrae.

Many ankylosaurid tail club specimens are broken at the handle just anterior to the knob (AMNH 5245, ROM 788, TMP 1983.36.120, TMP 1994.168.1), suggesting that a weak point exists at this location. This will be examined using finite element analysis in Chapter 6.

Decreasing the density of the knob osteoderms in the sensitivity analyses increases the impact force by decreasing the club mass, and therefore increasing the impact velocity. This also decreases the rotational inertia of the knob, which may have made it easier to wield. However, cancellous bone is more susceptible to strain during impacts. The role of different material properties in the club will be examined using finite element analyses in the following chapter.

5.6 Conclusions

Data from CT scans provided new information on the internal morphology of tail clubs. Up to three vertebrae may have been enclosed within the knob, and handle vertebrae may have fused together in some individuals. Centra are composed of cancellous bone, whereas the neural and haemal arches are compact. Knob osteoderms are cancellous with areas of low density, and ossified tendons are present underneath the knob osteoderms. Low density osteoderms would have reduced the rotational inertia of the tail club and allowed for greater impact velocities, which increases impact force and stress.

146 The range of motion between vertebrae was probably no more than 15°, which would allow the tail to swing through at least 100°. Angular acceleration for

all tails in this study was between 4.75 and 9.39 rad/s; this may underestimate

the acceleration of larger tails such as ROM 788/AMNH 5245. The composite tail

ROM 784/UALVP 47273 could impact with 571 to 1127 N of force, exerting an

impact stress of 29 to 56 MPa, well below the 100 MPa required to break bone in

shear according to Currey (1988). The reconstructed UALVP 16247 tail could

exert 962 to 2014 N of force and 48 to 100 MPa of stress; tail clubs of average

size may have been able to fracture bone during forceful blows. The composite

tail ROM 788/AMNH 5245 impacted with 7281 to 14360 N of force and 364 to

717 MPa of stress and would have almost certainly broken bone on impact. The

hypothesis that ankylosaurids could use their tail clubs as weapons was not

rejected in this study; ankylosaurids with average to large-sized knobs were

capable of fracturing bone on impact and therefore these clubs could have been

useful as weapons. Additionally, small and average knobs could likely cause pain

and damage to soft tissues, without breaking bone. If large tail club knobs were

able to impact with enough force to break bone in shear, how did the tail club

itself resist fracture? This will be explored using finite element modeling in

Chapter 6.

147 Chapter 6: Finite element analyses of ankylosaurid tail clubs5

6.1 Introduction

Finite element analysis is a powerful tool for understanding the biomechanics of extant and extinct organisms. Finite element analysis is used to reconstruct stress, strain, and deformation in structures. Rayfield (2007) provides an overview of the finite element method and its uses in palaeontology. Stress occurs in a structure when a force (load) is applied, and represents force per unit area. Tensile stresses are, by convention, represented by positive values, and compressive stresses by negative values. Strain is the change in length divided by the original length of a structure after a load is applied.

Finite element analysis of dinosaur fossils has predominantly dealt with theropod skulls (Rayfield et al. 2001; Mazzetta et al. 2004; Rayfield et al. 2004;

Rayfield 2005a; Rayfield 2005b; Rayfield et al. 2007; Shychoski et al. 2007), with fewer studies on ornithischian skull mechanics (Farke et al. 2007; Maidment and

Porro 2007; Porro 2007; Snively and Cox 2008). Analyses of the postcranial skeleton are rarer, and have included the metatarsus of a tyrannosaurid (Snively and Russell 2002), dromaeosaurid claws (Manning et al. 2007), and ossified tendons (Organ 2006a) and pedal morphology (Moreno et al. 2007) of ornithopods.

5 Portions of this chapter have been submitted for publication. Arbour, V., and Snively, E. Biomechanics and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora). The Anatomical Record A, submitted 10 October 2008. All text, calculations, and figures in this chapter are by V. Arbour. E. Snively provided advice on finite element modeling and comments on an early draft of this chapter. Permission from the coauthor can be found in Appendix 2. This is the first study to use FEA to investigate biomechanics in ankylosaurs. Four ankylosaurid tail clubs are examined to understand the distribution and magnitude of stresses within the club under simulated impact conditions. We assume that if stress magnitudes within the club are greater than is necessary to fracture bone, then tail clubs were not likely used as weapons.

Understanding the distributions of stresses will provide information on the function of the handle.

6.2 Materials and Methods

The first five analyses described in the results section examine tail clubs referred to Euoplocephalus. These analyses use 3 dimensional (3D) meshes constructed from computed tomography scans of four tail clubs specimens, using the computer software Mimics (mesh generation and editing) and Strand7 (finite element analysis). The sixth analysis described in the results section compares handle vertebrae of several ankylosaurid taxa, and uses 2 dimensional (2D) models constructed from photographs of various specimens, using the computer software Adobe Illustrator (model creation) and COMSOL Multiphysics (finite element analysis).

CT scans of four partial tail clubs (ROM 788, UALVP 16247, and UALVP

47273) were used to create 3D models for use in FEA (Fig. 6.1). The computer program Mimics was used to create a 3D model and mesh for each specimen, and to apply material properties to each mesh. A mask over the desired portion of the scan is created using the thresholding function. Each slice is manually

149 Figure 6.1: Models used in this study. UALVP 47273 in A) dorsal, B) ventral, C) right lateral and D) posterior view. Note the well preserved interlocking neural arches in A, haemal arches in B and ossified tendons in C. UALVP 16247 in E) dorsal, F) posterior, and G) left lateral view. TMP 83.36.120 in H) ventral, I) left lateral, J) posterior, and K) oblique dorsal view. ROM 788 in L) oblique dorsal, Nl) ventral, N) posterior, and O) left lateral view. The lateral edges of the knob were excluded from the scan; photos of the specimen are overlain in M and N to show the missing portions. Ridges on the knob in L and M are artefacts resulting from poor scan quality and manual editing in this region. Note the well-preserved haemal arches in M. All images created in Mimics from computed tomography scans. Photograph in N by R. Sissons and used with permission. Scale equals 10 cm.

150 edited using the 'multiple slice edit' function to both add and remove mask, to fill in cracks in the specimen and remove artifacts and unwanted parts of the scan

(the scanning bed, support jackets, etc.). A 3D model was then calculated and inspected for artifacts. A 3D mesh was created and exported as a NASTRAN file.

The default settings in Mimics produce a mesh with too many elements, which will not work properly in the FEA software Strand7 (Strand7 can handle a mesh of approximately 1 million elements or less). The mesh size is reduced by grouping voxels in the xy and z planes; this results in a loss of fine surface features, such as the knob osteoderm texture, but the model is still an accurate representation of specimen geometry. Once a mesh has been created, material properties can be assigned. Mimics calculates grey values of the CT images and displays these as a histogram. The number of materials can be specified, and material properties can be manually entered. The mesh is then exported as a

NASTRAN file for use in Strand7.

ROM 788 was scanned in two pieces (the knob and the majority of the handle), and data from the two CT scans was combined to make a single model for FEA. Both CT scans were cleaned in Mimics as for the other models. Each model was exported as a surface STL file and imported into a Mimics project file.

The STL models were aligned appropriately and then joined using the Boolean

Unite function in the Segmentation module. The united model was then decimated using the reduce triangles, smooth, and remesh functions in the

Mimics Remesher. This remeshed, united model was then imported into Strand7.

The missing, lateral edges of each major osteoderm, which were outside of the

151 field of view of the CT scanner, could not be reconstructed. No additional meshing is needed for models in NASTRAN format, but the model of ROM 788 required additional automatic and manual cleaning in Strand7 to remove triangles with free edges. The surface mesh was then converted to a solid mesh.

Strand7 is a finite element analysis software program. Material properties, a constraint, and a load are applied to each mesh and then analyzed for both stress and strain results using the linear solver. Table 6.1 lists the material properties used in the different analyses, and Table 6.2 lists the forces, constraints, and other variables used for each mesh of each analysis. Impact forces were calculated as described below, in the functional dynamics section.

Stress and strain results were displayed both as 3D surface plots, and as 2D cross-sections at various locations within the specimen. Strand7 can produce coloured contour and vector plots; tensile stresses are positive values, and compressive stresses are negative values.

Each specimen provides different benefits and limitations for analysis.

UALVP 47273 is a relatively complete specimen, and allows for analysis of the knob and handle together. However, a mesh of less than 5 million elements does not show the details of the individual neural and haemal arches. In order to better understand the stress distribution in the handle vertebrae, a smaller model was created by removing all but the last two of the visible handle vertebrae and the knob. The original model was edited slice by slice in Mimics to remove the extra vertebrae. In this manner, an impact force could be applied to the knob and the stress distribution in the handle vertebrae could be observed. Appropriate forces

152 could then be applied to a single vertebra isolated from the handle in the same

manner. Additionally, UALVP 47273 represents a small knob morphology that is

not representative of most ankylosaurid knobs. ROM 788 is the largest specimen

in this study, but the handle and knob are separate elements, and the lateral sides of the knob osteoderms were not included in the CT scan. UALVP 16247 is an isolated knob, but represents the average knob size in Euoplocephalus, and the CT scan of this specimen had few artifacts. As such, the effects of differing

bone densities and material properties are best analyzed in this specimen. The

cast of TMP 1983.36.120 cannot be used to examine material properties, but can

be compared with the similarly-sized UALVP 16247.

Table 6.1: Material properties used in finite element analyses

Density Young's Poisson's Comments (kg/m3) modulus ratio (Pa) Compact 2000 20e9 0.4 Density: Human 1.5-2.0 (Wirtz et al. 2000) bone Young's modulus: Alligator mississippiensis cortical 12 020, Crocodylus sp. cortical 5630, Geochelone niger 13780 (Currey 1988); Varanus exanthematicus cortical 22 800 (Erickson and Keaveny 2002) Poisson's ratio: Human cortical 0.22 to 0.47 (Peterson and Dechow 2003) Cancellous 1000 8e9 0.4 Density: Human 0.1-0.7 kg/m3 (Wirtz et al. bone 2000) Young's modulus: Human 774 (Peterson and Dechow 2003) Keratin 1300 2.5e9 0.4 Young's modulus: Ramphastos toco 6.7 GPa (Seki et al. 2006); Struthio camelus claw 1.84, 1.33 GPa (Bonser 2000); avian feather 2.5 GPa (Bonser and Purslow 1995), bovine hoof 261-418 MPa (Frank et al. 2006); Gekko gekko setae 1.6 GPa, Ptyodactylus hasselquistii setae 1.4 GPa (Peattie et al. 2007) Poisson's ratio: bovine hoof 0.38 (Frank et al. 2006)

153 Table 6.2: Summary of forces (N) used in finite element analyses Analysis ROM TMP UALVP UALVP UALVP 47273 UALVP 47273 788 1983.36.120 16247 47273 knob+vertebrae single vertebra 1 1016 9 960 570 0 1127 - 2 1016 960 570 - 0 3 - 570 570 200 4 - 39 39 1029 - 126 - 1029 - 960 570 - 200

6.2.1 Analysis 1

Analysis 1: Three specimens with different knob sizes were used to examine the effect of knob size and impact force on tail clubs. For each model, the anterior face of the centrum of the most anteriorly located part of the handle was constrained. A force was applied to both a small and large area at approximately the mid-height and mid-length of the left major osteoderm of each knob. This force was oriented at right angles into the osteoderm. The impact force for each knob was applied to each node in both the small and large impact area analyses. This is reasonable because any given point on the knob would impact with approximately the same amount of force. For this analysis, the knobs were assigned uniform material properties of cancellous bone.

6.2.2 Analysis 2

Analysis 2: Impacts did not necessarily always occur at the same location on the tail club. Impacts were simulated on the handle just anterior to the knob, and on the distal end of the knob, to understand how stress distribution changes as impact site changes. The most realistic forces were used for both ROM 788 154 and UALVP 47273, and the meshes were given the material properties of cancellous bone.

6.2.3 Analysis 3

Analysis 3: The vertebrae of the handle are highly modified compared to the free caudal vertebrae in ankylosaurids. To understand the distribution of stress within the handle vertebrae, two models were constructed from the CT scan of UALVP 47273. First, only the knob and two preceding handle vertebrae were manually isolated and meshed in Mimics ("knob+vertebrae" model). In

Strand 7, a force was applied at the midlength and midheight of the left lateral osteoderm, as for Analysis 1. The model was constrained at the anteriormost vertebra, on the medial faces of the prezygapophyses, the anterior face of the centrum, and the medial sides of the anterior projection of the haemal spine.

Results of the stress distribution in these models were then applied to a second model of a single handle vertebra ("single vertebra" model), which was also manually isolated and meshed in Mimics. Properties of cancellous bone were applied to the model. To simulate a tail club with unfused centra, an additional analysis, where the centrum was not constrained, was conducted for both the knob+vertebrae and isolated vertebra models.

6.2.4 Analysis 4

The unusually robust haemal arches of ankylosaurid tail clubs may play a role in postural support of the large knob. Impact forces are assumed to be

155 directed in the horizontal plane, but gravity would also act to pull downward on the tail club. Coombs (1995) noted that ankylosaurids probably did not drag their tails on the ground, although the tail may not have been held high off of the ground. The weight of each tail club knob is calculated using the volumes and masses determined in chapter 5, multiplied by gravitational acceleration (9.81 m/s2). Torque was calculated using both the length of the entire tail and the length of the tail club only. UALVP 47273 is the only specimen in this study that preserves the knob and handle together. Handle vertebrae become moderately larger as knob size increases, but the two are not linearly correlated (see chapter

5). As such, it is reasonable to apply the forces and torques derived for each knob (ROM 788, UALVP 16247, and UALVP 47273) to the model of UALVP

47273, for the purposes of comparing large and small knob weights. UALVP

47273 was constrained at the anterior face of the anteriormost vertebra, and the force was applied to a single node at the estimated centre of mass of the knob.

To understand the distribution of stress within a single vertebra, this force was also applied to the knob+vertebrae model and the single vertebra model (Table

6.3).

Table 6.3: Weight and torque. Specimen Knob Knob r r taj| Force Torque, Torque, volume mass .'. ... . '. , . (N) entire tail club v / 3X „ \ entire tail club (m) ' ...... , (cm) (kg) tail(Nm) (Nm) UALVP 2008 3.98 (m2.1) 6 1.34 39.01 84.24 52.26 47273 UALVP 6486 12.84 2.70 1.72 126.00 340.09 216.65 16247 ROM 788 53 007 104.95 3.49 1.60 1029.56 3593.16 1647.30 6.2.5 Analysis 5

Analysis 5: Knob osteoderms have regions of high, medium, and low density, which may affect the distribution of stress and strain throughout the club.

Strait et al. (2005) found that elastic properties affect quantitative strain data in finite element analyses, although overall strain patterns are similar using different elastic properties. Elastic properties for ankylosaur bone cannot be known.

However, a range of different properties from various taxa were used to estimate material properties in tail clubs (Table 6.2).

Regions of differing density were calculated using Mimics for the knob of

UALVP 16247 and an isolated handle vertebra of UALVP 47273. UALVP 16247 was loaded over a small area on the left lateral osteoderm, as for Analysis 1, and

UALVP 47273 was loaded on the neural spine as for Analysis 3.

Knob osteoderms were likely covered by a keratinous sheath in life.

Snively and Cox (2008) showed that the relative thickness of a horny covering on pachycephalosaur domes would have greatly influenced the distribution and magnitude of stresses within the osseous dome. To simulate the effects of a keratinous sheath, a new mask was created for UALVP 47273 in Mimics. In

Mimics, the outline of a thin keratinous sheath was traced for each slice of the knob osteoderms and added to the overall mask, and the grayscale values in the resulting model were assigned material properties for cancellous bone and keratin.

Additional analyses were conducted using two dimensional models in

COMSOL. The outline of a transverse section through the knob of both UALVP

157 16247 and UALVP 47273 was traced, as well as areas of low density in each osteoderm, and hypothetical keratinous coverings on each osteoderm. These coordinate outlines were exported as CAD DXF files, imported into COMSOL, coerced to solid, and assigned material properties as per the 3D models. The section models were constrained at the dorsal and ventral borders of the centrum

(equivalent to the midline of the knob) and loaded as for the 3D models.

6.2.6 Analysis 6

Many ankylosaurid taxa {Dyoplosaurus, Euoplocephalus, Pinacosaurus,

Saichania, and Talarurus) have sharply pointed neural spines in dorsal view.

However, Tarchia has less sharply pointed neural spines, and Ankylosaurus has

U-shaped, tongue-like neural spines. The response of these neural spine shapes to stress was modeled in COMSOL Multiphysics. The shape of the neural arch in dorsal view was traced in Adobe Illustrator from photographs of specimens directly observed, and from specimens in the literature, and these vector lines were exported as CAD .dxf files. These files were then imported into COMSOL, coerced to solid, and extruded to approximate the 3-dimensional shape of the neural arch. The material properties of compact bone were used for the neural arches, based on information from CT scans and naturally broken specimens.

The neural arch was constrained along the medial faces of the prezygapophyses, which would have abutted the next anterior neural spine. Each model was loaded along the left lateral side of the neural spine, from the posterior tip to the junction with the prezygapophyses, with the force directed mediolateraily. A separate set

158 of analyses loaded each model at the posterior tip of the prezygapophyses, with the force directed anteriorly. All models were loaded with 200 N.

6.3 Results

6.3.1 Analysis 1: Effect of knob size and impact force

The impact forces applied to UALVP 47273 (Fig. 6.2) had little effect on the club. Forces applied over a larger area had a greater effect, but even relatively large impact forces did not approach stress values indicating that bone would break. Stress is greatest at the impact site, and where the constraint has been applied (Table 6.4). In reality, the tail club would not be rigidly constrained, but would be free to flex laterally at the joint between the penultimate and transitional free caudal vertebrae. Stress also becomes concentrated in some locations along the handle that correspond to breaks in the specimen, and is not biologically meaningful. Tensile stress is found from the impact site to the distal terminus of the left half of the knob. Tensile stress is also particularly high at between the anterior terminus of the left major knob osteoderms and the handle, whereas compressive stress is found in the same location on the right side of the tail club. Maximum stress is found within the constrained area of the handle, and minimum stress is found distal to the impact site on the knob. The magnitude of the force did not change the distribution of stress within the club, but did change the absolute values of the maximum stress. Varying the impact area also changed the absolute values of the maximum stress. In lateral view, stress

159 Figure 6.2: Impact stresses in UALVP 47273. Impact at midlength of knob, A) stress contour plot (-100-100 Pa), B) stress vector plot (-600-600 Pa), left lateral view. Impact on handle anterior to knob, C) stress contour plot (-75-75 Pa), D) stress vector plot (-600-600 Pa). Impact on distal tip of knob, E) stress contour plot (-75-75 Pa), F) stress vector plot (-600-600 Pa). G) Stress vectors on anterior face of handle, impact at midlength of knob (-978-1639 Pa). H) Stress vectors on knob, dorsal view, impact at midlength of knob (-75-75 Pa).

vectors were oriented radially from the impact site and lengthwise along the

handle. In dorsal view, stress vectors were oriented transversely across the

handle and formed a complex swirling pattern on the knob around the impact site.

160 Table 6.4: Maximum and minimum stress values and locations. Compress;iv e stress is negative, tensile stress is positive. Analysis Model Impact Variable Maximum Stress (Pa) Minimum Force Stress (N) Impact Area X Y Z location Location 1 ROM 788 10 160 small 9150 8264 16351 Right Anterior to anterior face impact site of handle 10160 large 103426 93388 184760 Right Anterior of anterior face left of handle osteoderm TMP 1000 Small -587 -383 -695 Constraint Midline, 1983.36. midlength of 120 knob UALVP 960 Small -1073 -837 -1308 Midline, Midline, 16247 constraint anterior 960 Large -11547 -9126 -14238 Midline, Midline, constraint anterior 1420 Small -1668 -1316 -2055 Midline, Midline, constraint anterior 1420 Large -16841 -13310 -20767 Midline, Midline, constraint anterior UALVP 570 Small -1368 -1215 -2758 Left side, Knob 47273 anterior face posterior, of centrum right of midline 570 Large -21295 -18893 -42961 Left side, Knob anterior face posterior, of centrum right of midline 1127 Small -2127 -1874 -4307 Left side, Knob anterior face posterior, of centrum right of midline 1127 Large -40750 -36151 -82216 Left side, Knob anterior face posterior, of centrum right of midline Impact location 2 ROM 788 10160 Handle 77776 57925 121921 Right Posterior anterior face lateral face of vertebra of left osteoderm 10160 Knob distal 39730 35732 70743 Right Posterior end tip anterior face of knob, of vertebra midline UALVP 570 handle -3569 -2914 -7463 Left side, 2/3 length of 47273 anterior face knob, left of of centrum midline 570 Knob distal -2546 -2291 -5102 Left side, Midline, tip anterior face midlength of of centrum knob Materials 3 UALVP 960 Compact -206 -175 -233 Posterior end Dorsal to 16247 and of right impact cancellous osteoderm UALVP 570 Compact, 809 833 1220 Anterior face Distal knob, 47273 cancellous, of centrum, left of midline keratinous right side sheath UALVP 200 Compact -688 -110 -292 Medial face Lateral side, 47273 and of left left haemal single cancellous prezygapo- spine vertebra physis, posterior Table 6.4, continued: Maximum and minimum stress values and locations. Compressive stress is negative, tensile stress is positive. Constraint 4 UALVP 570 Centrum, -175 -64 -39 Impact site Posterior 47273 prezyga- knob, knob+ pophyses, midline vertebrae haemal spine UALVP 570 Prezyga- 216 103 50 Medial face Anterior tip 47273 pophyses, of right of haemal knob+ haemal prezygapo- spine, left vertebrae spine physis, side posterior UALVP 200 Prezyga- -2389 -1297 -1443 Medial face Lateral side, 47273 pophyses, of left left haemal single haemal prezygapo- spine vertebra spine physis, posterior Constraint 5 ROM 788 1029 Centrum, -269 -249 -505 Right anterior Dorsal to Prezyga- face of impact site pophyses centrum UALVP 39 Anterior 22 25 35 Constraint Distal tip of 47273 handle knob UALVP 39 Prezyga- 5 12 4 Load Distal tip of 47273 pophyses, knob knob+ haemal vertebrae spine

Impact forces produced low absolute stresses on TMP 1983.36.120 and

UALVP 16247 (Fig. 6.3). Compression is found on the left osteoderm and is greatest at the site of impact, whereas tension is found on the right osteoderm and near the constraint. Tensile stress is also concentrated at the boundary between the major and minor plates. Stress vectors were oriented radially from the impact site on the lateral face of the osteoderm, anteroposteriorly on the left major osteoderm in dorsal view, and mediolaterally on the right major osteoderm in dorsal view. In anterior view, the stress vectors converged towards the constraint, forming a clockwise swirl. Figure 6.3: Impact stresses in TMP 83.36.120, A) stress vector plot (-75-75 Pa), dorsal view, and B) stress contour plot (-50-50 Pa), oblique posterodorsal view. Impact stresses in UALVP 16247: C) stress vector plot (-75-75 Pa), dorsal view, and D) stress contour plot (-30-30 Pa), oblique posterodorsal view.

In ROM 788 (Fig. 6.4), compressive stress was found at the impact site, with tensile stress immediately adjacent to the impact site rapidly changing to approximately neutral stress throughout the rest of the osteoderm. Tensile stress is found at the boundary of the knob osteoderms and handle, with compressive stress concentrated along the midline of the knob dorsally and tensile stress ventrally. Stress vectors radiate from the impact site and form a complex, swirling pattern in dorsal view at the knob and anterior view at the constraint. They are oriented anteroposteriorly along the handle in lateral view, and mediolaterally in dorsal view. The anterior face of the handle centrum of ROM 788 experienced tensile stress on the right half and compressive stress on the left half, similar to

163 Figure 6.4: ROM 788 impact stresses. A) Stress contour plot (-150-150 Pa), right lateral view. Note the area of concentrated compression anterior to the knob. B) Stress contour plot (-60-60 Pa), left lateral view. High tensile stress is found at the anterior borders of the prezygapophyses. C) Stress contour plot (-5784-5596 Pa), anterior view. D) Stress vector plot (-1500-1500), anterior view. E) Stress vector plot (-1000-1000 Pa), closeup of knob and handle junction, oblique dorsal view. that observed in UALVP 47273. The medial face of the right prezygapophysis experienced tension, and the lateral face experienced compression; the reverse was found in the left prezygapophysis. Tensile stress was also found within bone surrounding the neural canal. Along the handle, tensile stress was found at the anterior edges of the prezygapophyses on the right side. There is an area of concentrated compressive stress on the right side of the handle approximately 5 cm anterior to the knob. The haemal arch experienced neutral stress for much of its length, with increasing tensile stress near the constraint.

164 6.3.2 Analysis 2: Impact site analysis

Altering the impact site did not change the distribution of stresses at the constraint (Fig. 6.2). Compression was found at the impact site. Impacts to the handle resulted in almost zero stress within the knob. Peak stress did not greatly increase or decrease based on impact location, and was always found within the constraint. The location of minimum stress changed in each model (Table 6.4).

Stress vectors radiate from the impact site on the handle. In dorsal view, stress vectors on the knob are oriented mediolaterally, and in lateral view they are oriented dorsoventrally.

An impact near the distal tip of the knob results in stress vectors oriented anteroposteriorly in lateral view of the knob and handle, and mediolaterally in dorsal view. The distribution of stress along the handle did not change, and shifted distally in the knob. Tensile stress radiated anteriorly through the left half of the minor plates, and compressive stress did the same on the right half.

6.3.3 Analysis 3: Stress distributions in the handle vertebrae

Compressive stress was found at the impact site on the left major osteoderm, dorsally between the left major osteoderm and handle, and on the right half of the anterior face of the centrum, where the model was constrained

(Fig. 6.5). The midline of the centrum had stress near zero, approximating a neutral axis. Tensile stress was found dorsally and anteriorly between the right major osteoderm and the handle, and on the left half of the anterior face of the centrum. Peak stress when 570 N was applied to a small area on the knob was

165 Figure 6.5: Results from analyses 4 and 5. UALVP 47273 knob+vertebrae, impact force, centrum constrained, stress contour plots in oblique left anterolateral view A) -100 to 100 Pa, B) - 25 to 25 Pa; C) anterior view, -100 to 100 Pa; and oblique left dorsolateral view E) -100 to 100 Pa, F) -25 to 25 Pa. UALVP 47273 knob+vertebrae, impact force, centrum unconstrained, stress contour plots in D) anterior view, -125 to 125 Pa; and oblique left dorsolateral view G) -50 to 50 Pa, H) -25 to 25 Pa; stress vector plot in oblique left dorsolateral view, -125 to 125 Pa. UALVP 47273 knob+vertebrae, knob weight, stress contour plot in I) dorsal view, -15 to 15 Pa, L) anterior view, -15 to 15 Pa; O) stress vector plot in left lateral view, -125 to 125 Pa. UALVP single vertebra, impact force, centrum unconstrained, -250 to 250 Pa, in J) dorsal view, K) oblique left dorsolateral view, and M) -250 to 250 Pa, anterior view.

166 well below that required to break bone in shear (175 Pa). Within the prezygapophyses, stresses were greater posteriorly and decreased to nearly zero at the anterior termini. Changing the constrained area of the model changed the distribution of stresses within the vertebrae. When only the prezygapophyses were constrained, peak stress was found on the posterior part of the right prezygapophysis, within the constrained area. Tensile stress was concentrated below the right prezygapophysis on the anterior face of the centrum, but dissipated abruptly away from the prezygapophysis.

Stress vectors in the unconstrained centrum model were complex. In dorsal view of the knob, stress vectors are oriented mediolaterally in the right osteoderm, and in the left osteoderm collectively form a swirling pattern, inclined anteroposteriorly. In left lateral view, vectors were oriented lateroposteriorly along the neural spine, but became undulate along the prezygapophyses. Along the centrum, vectors were oriented approximately anteroposteriorly, looping ventrally onto the haemal spine. The anterior projection of the haemal spine had approximately dorsoventrally directed stress vectors. In right lateral view, stress vectors are oriented dorsoventrally on the neural spine, right prezygapophysis, centrum, and posterior portion of the haemal spine. The anterior projection of the haemal spine has approximately laterally oriented vectors. Dorsally, anteroposteriorly directed vectors from the left side of the neural spine and haemal spine arced across the neural arch and haemal arches, becoming mediolaterally oriented on the right side of each spine. Stress vectors loop mediolaterally around the right prezygapophysis.

167 The location and value of the peak stresses were used to estimate a force for an analysis of a single vertebra (Fig. 6.5). A 200 N force was applied to several nodes on the left lateral side of the neural spine, with the force directed medially at approximately right angles to the neural spine. This is consistent with the orientation of the stress vectors in the knob+vertebrae model, where the anteroposteriorly-oriented stress vectors in the right prezygapophyses arc mediolaterally at the location where the preceding neural spine would have interlocked with the prezygapophyses. Stress vector orientation in the isolated vertebra model was consistent with that seen in the knob+vertebra model, confirming an appropriate force direction. Compressive stress was concentrated where the model was loaded, but became tensile abruptly, anterior to the load.

6.3.4 Analysis 4: Postural role of the haemal arches

Tensile stress was found at the junction of the prezygapophyses, but not along their medial faces (Fig. 6.5). Low tensile stresses were observed on the anterior face of the centrum dorsal to the haemal canal. Ventrally, tensile stress is found irregularly along the haemal arches. In lateral view, the knob experienced low tensile stress ventrally, and low compressive stress dorsally. In lateral view, the pattern of vectors within the handle was similar to that in

Analysis 4. In dorsal view, the vectors are oriented anteroposteriorly along the knob osteoderms, the neural spines, and both right and left prezygapophyses.

168 6.3.5 Analysis 5: Material properties

In the keratinous sheath UALVP 47273 model (Fig. 6.6), the distribution of stresses within the handle and knob did not change noticeably compared to the normal UALVP 47273 model. Compressive stress at the impact site was surrounded by a halo of tensile stress, which was not observed in the bone model. The keratinous sheath slightly reduced the peak stress at the constraint.

The overall distribution of stresses in the UALVP 47273 isolated vertebra model

(Fig. 20) did not change when the material properties were changed, although the stresses appeared more diffuse compared to the single material property model. Material properties affected the external distribution of stress in UALVP

16247 slightly; there was an increase in tensile stress at the anterior of the right major osteoderm. Two dimensional models of UALVP 16247 (Fig. 20) had higher strain values in the inner low density areas of the osteoderms, compared to the outer cortex, in models lacking a keratinous sheath. When a keratinous sheath was modeled, strain was localized to the keratinous layer at the site of impact and strain values were reduced in the bone.

6.3.6 Analysis 6: Neural spine shape in different ankylosaurid taxa

The basal ankylosaurid ("polacanthid") Mymoorapelta lacks a tail club, however, the distal caudal vertebrae are modified to interlock in a manner similar to the handle vertebrae of more derived ankylosaurids. In Mymoorapelta, the prezygapophyses are shorter than the neural spine, and the neural spine is broadest at its midlength and constricted near the junction with the

169 Figure 6.6: Effects of differing material properties. UALVP 47273 with simulated keratinous covering, oblique left lateral view: A) mesh resulting from material property assignment in Mimics, where dark blue is assigned the material properties of keratin and all other colours are assigned the properties of cancellous bones, and B) stress contour plot of results (-150-150 Pa). UALVP 47273 isolated vertebra with two material properties, oblique left anterolateral view: C) stress contour plot (-600-600 Pa) of results of mesh D) with neural and haemal arches assigned properties of compact bone and the centrum assigned properties of cancellous bone. UALVP 16247 with two material properties, oblique left anterolateral view: E) mesh where greens and blues are assigned the properties of compact bone and reds, yellows and oranges are assigned the properties of cancellous bone, and F) stress contour plot of results (-50-50 Pa). G) UALVP 16247, transverse section at approximately the midlength of the knob, first principle strain results with an outer compact zone, an inner cancellous zone, and a simulated keratinous covering over the left osteoderm. prezygapophyses. In contrast, in ankylosaurid handle vertebrae, the prezygapophyses are as long as or longer than the neural spine, and the neural spine is broadest at the junction with the prezygapophyses.

Under lateral loading, stress was concentrated at the junction between the two prezygapophyses, and on the lateral sides of the neural spine at this location

(Fig. 6.7). A neutral axis of low stress occurs along the midline of the neural

170 |Max: 8.193 * 10 Figure 6.7: Two dimensional FEA of neural arches in dorsal view, von Mises stress results. A) Mymoorapelta MWC 5819. B) Euoplocephalus AMNH 5245. C) Talarurus MPC KID2007.167 D) Dyoplosaurus ROM 784. E) Tarchia UALVP 47948. F) Ankylosaurus AMNH Min: 0.332 5214. Outline in A traced from photograph provided by M. Burns and Max: 2.852 * 106 used with permission. Outline in F traced from Carpenter (2005).

Min: 8.602 *10"3

Max: 2.457 * 10b

Min: 6.497 * Iff3

Max: 1.774 *106

!Min: 2.993* 10"8

Max: 1.385 MO6

Min: 6.693 * 10"*

Max: 4.787 MO5

iMin: 6.487 * Iff5 171 spine. The boundaries of this axis are sharpest in the most narrow, sharply pointed neural spines, and the axis becomes more diffuse in Tarchia and

Ankylosaurus, and almost unrecognizable in Mymoorapelta. The prezygapophyses experienced neutral stress.

Peak stress was greatest in Euoplocephalus, and lowest in Ankylosaurus when all models were scaled relative to each other (Table 6.5). Peak stress was greatest in Talarurus when all models were scaled to the same length.

Mymoorapelta also had a low peak stress similar to that observed in

Ankylosaurus.

Table 6.5: Peak stresses in neural arch analyses Taxon and specimen number Models scaled relative Models scaled to same to each other length as ROM 784 Ankylosaurus, AMNH 5214 4.787* 10s 4.865*10° Dyoplosaurus, ROM 784 1.774 * 106 1.774 *106 Euoplocephalus, UALVP 47273 2.852 * 106 2.852 *10b Mymoorapelta, MWC 5819 8.193 * 105 9.81 *10 b Talarurus, MPC KID2007.167 2.457 * 106 2.914 *10b Tarchia, UALVP 47948 1.385 * 106 1.387 *10b

6.4 Discussion

Ankylosaurid tail clubs appear well equipped to handle the stresses incurred by lateral impact forces. Stress is distributed evenly throughout the knob and handle, with stress concentrating at the constraint. Small areas of impact result in peak stresses well below the amount required to break bone in shear.

Larger areas of impact result in larger peak although these are still well below that required to break bone in shear. However, the analyses in this study ignore the role of soft tissues in controlling and reducing stress within the tail club. Ligaments, tendons, and muscles connecting successive vertebrae, as well as intervertebral disks, may all have acted to absorb forces along the handle; no part of the handle would have been completely constrained, and even a small amount of flexibility between successive vertebrae may have sufficed to prevent tail clubs from breaking during impacts. Additionally, the analyses in this study do not model the free caudal vertebrae, and the effects of tail club impacts in this region of the tail are unknown.

The interlocking neural spines and prezygapophyses of the handle stiffened the distal portion of the tail, providing a support for the large terminal osteoderms. The handle reduces the maximum angular acceleration of the terminal tail segment, in comparison to a tail composed of free caudal vertebrae.

This would have reduced the impact force, preventing damage to the distal caudal vertebrae. Many ankylosaurid tail club specimens are broken at the handle just anterior to the knob (AMNH 5245, ROM 788, TMP 1983.36.120, TMP

1994.168.1), suggesting that a weak point exists at this location. Concentrated compressive stress (-650 Pa) was noted in approximately this location in ROM

788, but not in UALVP 47273. This may be a result of the size difference between the two knobs.

The components of the neural arch are arranged to resist lateral bending.

The prezygapophyses are long and tall, and do not dorsally overlap the neural spine of the preceding vertebra. Detailed FEA of the handle vertebrae indicate that the elongate prezyagpophyses act to dissipate stress anteriorly. They also reduce stress within the centrum in handles with unfused vertebrae. In ROM 788,

173 tensile stress was concentrated at the anterior edges of the prezygapophyses on the impact side. In the model, these edges are fused to the handle. In reality, there is some space between the prezygapophyses and neural spine of successive vertebrae, which would have allowed for a small amount of flexibility, and tensile stress may not have concentrated in this location. However, stress at this location in the model suggests that soft tissues in this area (possibly associated with Mm. interarticulares superiores), may have experienced greater tensile stress than elsewhere between the prezygapophyses and neural spines.

Two dimensional analyses of the neural arches showed that the rounded neural spine of Ankylosaurus experienced lower peak stresses compared to the more pointed neural spines in other taxa, independent of size. Mymoorapelta also experienced low stresses independent of size. This suggests that neural spines with broad, rounded dorsal outlines are better at absorbing impact stresses. If Mymoorapelta represents a transitional form between basal ankylosaurids and derived ankylosaurids (in terms of handle evolution), then why do most ankylosaurids have the slender, tapering neural spine morphology?

Additional study is needed to investigate the possible benefits, if any, that the slender morphology provides.

The haemal arches of the handle are unique among dinosaurs as a robust, nearly continuous tube of bone on the ventral side of the centra. The anterior projection has a ventral groove that receives the posterior projection of the preceding arch. This groove becomes ventrally enclosed posteriorly, and the posterior projection of the preceding arch becomes completely surrounded by the

174 subsequent haemal arch. The haemal arch appears to be adapted to resist vertical bending of the club, and may play a role in maintaining the neutral posture of the tail. Although in some specimens the centra are fused in the handle, in many specimens the centra are unfused, and the tube-like, robust haemal arches may act as a strut that would have kept the knob held off of the ground without requiring additional muscular effort. This tube of bone would act to keep the handle from sagging, and would therefore keep the neural arches

properly aligned to resist lateral bending. Finite element analyses simulating the weight of the club had low peak stresses, and evenly distributed stress throughout the handle.

Porro (2008) found that material properties and force did not change the distribution of stress within the skull of Heterodontosaurus, and only changed the

magnitude of the maximum stress. However, the direction of force changed the distribution of stress within the skull. This is also true for the ankylosaurid tail

clubs: changes to the material properties, magnitude of force, and area of impact size in the 3D analyses only changed the peak stress magnitude. Changes in the

location of impact altered the distribution of stress, and loading the models for

impact force versus weight altered the distribution of stress as well.

Decreasing the density of the knob osteoderms in the sensitivity analyses

increases the impact force by decreasing the club mass, and therefore increasing the impact velocity. This also decreases the rotational inertia of the knob, which

may have made it easier to wield. However, cancellous bone is more susceptible

to strain during impacts. Two-dimensional models showed that a hypothetical

175 keratinous covering reduced strain within the cancellous region of the knob. A keratinous sheath may have been important in preventing damage to the underlying bone during impacts.

6.5 Conclusions

The hypothesis that tail clubs could be used as weapons was not rejected, although it is impossible to determine whether or not ankylosaurids actually engaged in tail swinging behaviour. Tail clubs of all sizes were capable of withstanding impact forces estimated in Chapter 5. Soft tissues may also have played a role in preventing bone fracture during forceful impacts. The derived distal caudal vertebrae of ankylosaurids appear well adapted for distributing impact stress along the handle and for supporting the weight of the knob. A keratinous covering may have reduced strain within the low density knob osteoderms.

176 7.0 Conclusions

The results of this study indicate that ankylosaurids were capable of using their tail clubs as weapons for delivering forceful blows. The most important pieces of evidence that support this conclusion are:

1. Functional dynamics: Ankylosaurid tail clubs with knobs of average to

large size were capable of generating impact stresses greater than that

required to break bone in shear.

2. Finite element analysis: Ankylosaurid tail clubs were capable of

withstanding these estimated impact forces without risking fracture.

Additional evidence that suggests the function of the tail club was for delivering forceful impacts includes:

1. Knob osteoderms are composed of low density bone, which would have

lightened the knob. This would have increased impact forces while

decreasing rotational inertia and making the club easier to wield as a

weapon.

2. Muscle scars on the pelvis indicate the presence of a large M. longissimus

caudae, which may have been involved in tail swinging.

3. Two infected tail club knobs may indicate traumatic injury to the knob,

suggesting active use of the tail as a weapon.

This study modeled tail club impact forces with the assumption that the lateral movement of the tail began only at the anterior free caudal vertebra, and

177 does not incorporate movement of the body using the hips and hindlimbs. This simplified model almost certainly underestimates the impact force of a tail club, and if ankylosaurids engaged in this behaviour then the hips and hindlimbs would probably have played an important role in tail swinging.

Finite element analyses of tail club impacts assume that the tail club was rigidly constrained at the anterior face of the handle, even though this is not the case in reality. Improvements to finite element analysis techniques may allow for modeling of 'soft' constraints with some flexibility at the anterior face of the handle. Additionally, this study did not examine the stress distribution within the free caudal vertebrae. An interesting avenue for future research would be to examine impact stress and stress from the weight of the tail club in the free caudal vertebrae using the finite element method.

Ankylosaurid tail clubs could have been used for delivering forceful blows, but was their main function for defense against predators, for intraspecific combat, or both? ROM 784 and UALVP 47273 have small knobs and represent smaller individuals than ROM 788 and AMNH 5245. However, overall body size and knob width are not proportionately correlated (see Chapter 2). ROM 784 and

UALVP 47273 probably represent almost fully mature individuals. This suggests that ankylosaurid knobs were not primarily used as defensive weapons: a weapon that is not functional until very late in life would probably not have a selective advantage over a weapon that is of use earlier in life. Small juvenile

Pinacosaurus did not have knobs at all (Currie 1993). Life history curves similar to those created for tyrannosaurids (Erickson et al. 2006) would be useful in

178 plotting growth of the knob in relation to growth of the individual, although these results may not be possible to obtain in dermal ossifications.

An alternative hypothesis is that tail clubs evolved for use in intraspecific combat, although this is difficult to test directly. Knobs may have grown only at reproductive maturity, and may have been used during courtship battles. The two competitors may have swung tail clubs at the flanks of the opponent, which can be compared with flank-butting in bovids such as Bison bison (Reynolds et al.

2003), and head-clubbing (necking) in Giraffus camelopardalis (Simmons and

Scheepers 1996). Flank butting in bison often results in goring and rib fractures

(Reynolds et al. 2003), and giraffe necking can result in leg fractures, opponents being knocked unconscious, and death (Simmons and Scheepers 1996). If ankylosaurids engaged in a similar behaviour using tail clubs, we might expect to see a larger number of rib injuries in ankylosaurids compared to other groups of dinosaurs. Many ankylosaurid rib specimens show evidence of breakage and subsequent healing (pers. obs.), but it is unknown what percentage of specimens are pathological, and how this compares to other dinosaur groups. Tail clubs with large knobs were undoubtedly effective deterrents against bipedal predators.

However, the exclusive use of tail clubs as a defensive weapon is not supported by the results of this study.

The evolutionary origin of the tail club remains unresolved. Handle vertebrae represent modified distal caudal vertebrae, but there are few intermediate forms. Mymoorapelta has interlocking distal caudals that are similar to those of more derived ankylosaurids, and which may represent the precursor

179 to the tail club handle. Unfortunately, the phylogenetic position of Mymoorapelta is unresolved. This study has found that characters of the pelvis, caudal vertebrae and tail club can be used to diagnose ankylosaurid taxa, and found support for the separation of Dyoplosaurus from Euoplocephalus; incorporation of additional postcranial characters into existing ankylosaur character matrices could help resolve the phylogenetic position of 'polacanthids', many of which may be basal ankylosaurids. Future research could investigate the early evolution of the ankylosaurid tail club by including postcranial characters in phylogenetic analyses of the Ankylosauria.

Basal and derived ankylosaurids frequently have wedge-like, keeled plates (splates) along the lateral sides of the tail, and knob osteoderms are probably enlarged versions of the most distal of these splates. A scenario for the evolution of the knob and tail club may have involved basal ankylosaurids engaging in tail swinging behaviour, using the splates along the sides of the tail for slashing. A similar behaviour can be seen in the extant lizard Uromastyx aegypticus, which thrashes its spiny tail at predators while wedged into crevices

(Cooper et al. 2000). Large terminal splates may have been advantageous, perhaps for delivering heavier blows. Eventually, these became enlarged to form the tail club knob. Early modifications to the distal vertebrae may have facilitated tail swinging, and later may have become necessary to support the weight of the tail club knob. These larger osteoderms may have evolved in response to predators, or, more likely, were used in intraspecific combat.

180 Handle morphology appears to be relatively conservative through time.

Most ankylosaurids have slender, tapering neural spines in the handle.

Ankylosaurus has neural spines that are broad and rounded in dorsal view. Two dimensional FEA showed that this morphology results in lower peak stresses within the neural arch. This may have been an adaptation for withstanding greater impact forces than those calculated for Euoplocephalus. The one known

Ankylosaurus tail club (AMNH 5214) has a knob within the size range of the largest Euoplocephalus knobs, but with substantially larger and more robust handle vertebrae. Ankylosaurus is a much larger animal than Euoplocephalus, and perhaps impact stresses were greater in Ankylosaurus tail clubs. It is also possible that AMNH 5214 does not represent the maximum knob size in

Ankylosaurus. However, Mymoorapelta also has broad, rounded neural spines in dorsal view. If this rounded morphology is better suited to dissipate stresses during impacts, why do most ankylosaurids have slender, tapering neural spines in the handle? Again, the phylogenetic placement of Mymoorapelta is important for interpreting the evolution of the ankylosaurid tail club.

Ankylosaurid tail clubs appear well adapted for delivering and sustaining forceful impacts. However, this does not confirm that ankylosaurids actually engaged in tail swinging behaviour, and it does not determine whether or not they were exclusively used for defense from predators or for intraspecific combat, or both.

181 8.0 Literature Cited

Alexander, J.W., Solangi, M.A., and Riegel, L.S. 1989. Vertebral osteomyelitis

and suspected diskospondylitis in an Atlantic bottlenose dolphin (Tursiops

truncatus). Journal of Wildlife Diseases 25:118-121.

Alexander, R.M., Farina, R.A., and Vizcaino, S.F. 1999. Tail blow energy and

carapace fractures in a large glyptodont (Mammalia, Xenarthra).

Zoological Journal of the Linnean Society 126:41-49.

Arbour, V. In review. Ankylosaur tail and pelvis pathologies. Cretaceous

Research. Submitted 14 October 2008.

Arbour, V., Sissons, R.L., and Burns, M.E. In review. A redescription of the

ankylosaurid dinosaur Dyoplosaurus acutosquameus Parks, 1924

(Ornithischia: Ankylosauria) and a revision of the genus. Journal of

Vertebrate Paleontology. Submitted 22 August 2008.

Arbour, V., and Snively, E. In review. Biomechanics and function of the tail club

of ankylosaurid dinosaurs (Ornithischia: Thyreophora). The Anatomical

Record. Submitted 10 October 2008.

182 Aufderheide, A.C., and Rodriguez-Martin, C. 1998. The Cambridge Encyclopedia

of Human Paleopathology. Cambridge University Press, 496 pp.

Bamman, M.W., Newcomer, B.R., Larson-Meyer, D., Weisner, R.L., and Hunter,

G.R. 2000. Evaluation of the strength-size relation in vivo using various

muscle size indices. Medicine and Science in Sports and Exercise

32:1307-1313.

Blows, W.T. 1987. The armoured dinosaur Polacanthus fox/from the Lower

Cretaceous of the Isle of Wight. Palaeontology 30:557-580.

Blows, W.T. 2001. Dermal armor of the polacanthine dinosaurs. In: Carpenter, K.

(ed.) The Armored Dinosaurs. Indiana University Press, pp. 363-385.

Bonser, R.H.C. 2000. The Young's modulus of ostrich claw keratin. Journal of

Materials Science Letters 19:1039-1040.

Bonser, R.H.C, and Purslow, P.P. 1995. The Young's modulus of feather

keratin. Journal of Experimental Biology 198:1029-1033.

Bohlin, B. 1953. Fossil reptiles from Mongolia and Kansu. Sino-Swedish

Expedition Publication 37:1-113.

183 Brown, B. 1908. The Ankylosauridae, a new family of armored dinosaurs from

the Upper Cretaceous. American Museum of Natural History Bulletin

24:187-201.

Burns, M.E. In press. Taxonomic utility of ankylosaur (Dinosauria, Omithischia)

osteoderms: Glyptodontopelta mimus Ford, 2000-a test case. Journal of

Vertebrate Paleontology.

Carpenter, K., Miles, C, and Cloward, K. 1998. Skull of a Jurassic ankylosaur

(Dinosauria). Nature 393:782-783.

Carpenter, K. 2001. Phylogenetic analysis of the Ankylosauria. In: Carpenter, K.

(ed.) The Armored Dinosaurs. Indiana University Press, pp. 455-483.

Carpenter, K., Kirkland, J.I., Burge, D.L., and Bird, J. 1999. Ankylosaurs

(Dinosauria: Omithischia) of the Cedar Mountain Formation, Utah and

their stratigraphic distribution. In: Gillette, D.D. (ed.) Vertebrate

Paleontology in Utah. Utah Geological Survey Miscellaneous Publications

99-1:243-251.

184 Carpenter, K., Kirkland, J.I., Burge, D., and Bird, J. 2001. Disarticulated skull of a

new primitive ankylosaurid from the Lower Cretaceous of Eastern Utah. In:

Carpenter, K. (ed.) The Armored Dinosaurs. Indianapolis: Indiana

University Press, pp. 211-238.

Carpenter, K., Miles, C.A., and Cloward, K. 2001. New primitive stegosaur from

the Morrison Formation, Wyoming. In: Carpenter, K. (ed.) The Armored

Dinosaurs. Indianapolis: Indiana University Press, pp. 55-75.

Carpenter, K., Sanders, F., McWhinney, L.A., and Wood, L. 2005. Evidence for

predator-prey relationships: examples for and Stegosaurus. In:

Carpenter, K. (ed.) The Carnivorous Dinosaurs. Indianapolis: Indiana

University Press, pp. 325-350.

Chiasson, R.B. 1962. Laboratory anatomy of the alligator. Dubuque, Iowa: WM.

C. Brown Company Publishers, 56 pp.

Close, R.I. 1972. Dynamic properties of mammalian skeletal muscle.

Physiological Reviews 52:129-197.

COMSOL, Inc. 2005. COMSOL Multiphysics. Los Angeles, California.

185 Cong, L. 1998. The gross anatomy of Alligator sinensis Fauvel. Beijing: Forestry

Publishing House, 388 pp.

Coombs, W.P., Jr. 1971. The Ankylosauria. Ph.D. dissertation, Columbia

University, New York. 487 pp.

Coombs WP, Jr. 1979. Osteology and myology of the hindlimb in the

Ankylosauria (Reptilia, Omithischia). Journal of Paleontology 53:666-684.

Coombs, W.P., Jr. 1995. Ankylosaurian tail clubs of middle Campanian to early

Maastrichtian age from western North America, with description of a tiny

tail club from Alberta and discussion of tail orientation and tail club

function. Canadian Journal of Earth Sciences 32:902-912.

Cooper, W., Jr., van Wyk, J.H., Mouton, P. Le F.N., Al-Johany, A.M., Lemos-

Espinal, J.A., Paulissen, M.A., and Flowers, M. 2000. Lizard antipredatory

behaviors preventing extraction from crevices. Herpetologica 56:394-401.

Currie, P. J. 1991. The Sino-Canadian dinosaur expeditions, 1986 - 1990.

Geotimes 36:18-21.

186 Currie, P.J. 2003. Cranial anatomy of tyrannosaurid dinosaurs from the Late

Cretaceous of Alberta, Canada. Acta Palaeontologica Polonica 48:191-

226.

Currey, J.D. 1988. The effect of porosity and mineral content on the Young's

modulus of elasticity of compact bone. Journal of Biomechanics 21:131-

139.

Currey, J.D. 2002. Bones: Structure and Mechanics. New Jersey: Princeton

University Press, 456 pp.

Daoust, P.Y. Osteomyelitis and arthritis caused by Salmonella typhimurium in a

crow. Journal of Wildlife Diseases 14:483-485.

Dong, Z. 1993. An ankylosaur (ornithischian dinosaur) from the Middle Jurassic

of the Junggar Basin, China. Vertebrata PalAsiatica 31:258-266.

Dong, Z. 2002. A new armored dinosaur (Ankylosauria) from Beipiao Basin,

Liaoning Province, Northeastern China. Vertebrata PalAsiatica 40:276-

285.

Dong, Z.-M., Tang, Z., and Zhou, S.-W. 1982. Note on the new Mid-Jurassic

stegosaur from Sichuan Basin, China. Vertebrata PalAsiatica 20:83-88.

[Chinese 83-86; English 871. Dzemski, G., and Christian, A. 2007. Flexibility along the neck of the ostrich

(Struthio camelus) and consequences for the reconstruction of dinosaurs

with extreme neck lengths. Journal of Morphology 268:701-714.

Erb, E., Shaughnessy, P.D., and Norman, R.J. de B. 1996. Dental and

mandibular injury in an Antarctic fur seal, Arctocephalus gazelle, at Heard

Island, Southern Ocean. Journal of Wildlife Diseases 32:376-380.

Erickson, G.M., Catanese, J., and Keaveny, T.M. 2002. Evolution of the

biomechanical material properties of the femur. The Anatomical Record

268:115-124.

Erickson, G.M., Currie, P.J., Inouye, B.D., and Winn, A.A. 2006. Tyrannosaur life

tables: an example of nonavian dinosaur population biology. Science

313:213-217.

Evans, D.C., and Reisz, R.R. 2007. Anatomy and relationships of Lambeosaurus

magnicristatus, a crested hadrosaurid dinosaur (Ornithischia) from the

Dinosaur Park Formation, Alberta. Journal of Vertebrate Paleontology

27:373-393.

188 Farke, A.A., and O'Connor, P.M. 2007. Pathology in Majungasaurus

crenatissimus (: Abelisauridae) from the Late Cretaceous of

Madagascar. In: Sampson, S.D., and Krause, D.W. (eds.) Majungasaurus

crenatissimus (Theropoda: Abelisauridae) from the Late Cretaceous of

Madagascar. Society of Vertebrate Paleontology Memoir 8, pp. 180-183.

Farke, A.A., Chapman, R.E., and Andersen, A. 2007. Structural properties of the

frill of Triceratops. Royal Tyrrell Museum Ceratopsian Symposium, 22-23

September 2007; Short Papers, Abstracts, and Programs, 44-47.

Fish, F.E., Bostic, S.A., Nicastro, A.J., and Beneski, J.T. 2007. Death roll of the

alligator: mechanics of twist feeding in water. The Journal of Experimental

Biology 210:2811-2818.

Ford, T.L. 2000. A review of ankylosaur osteoderms from New Mexico and a

preliminary review of ankylosaur armor. In: Lucas, S.G. and Heckert, A.B.

(eds.), New Mexico Museum of Natural History Bulletin No. 17, pp. 157-

176.

Franck, A., Cocquyt, G., Simoens, P., and De Belie, N. 2006. Biomechanical

properties of bovine claw horn. Biosystems Engineering 93:459-467.

189 Frey, E., Riess, J., and Tarsitano, S.F. 1989. The axial tail musculature of recent

crocodiles and its phyletic implications. American Zoologist 29:857-862.

Fukunaga, T., Miyatani, M., Tachi, M., Kouzaki, M., Kawakami, Y., Kanehisa, H.

2001. Muscle volume is a major determinant of joint torque in humans.

Acta Physiologica Scandinavica 172:249-255.

Garner, M. M., Herrington, R., Howerth, E. W., Homer, B. L, Nettles, V. F., Isazo,

R., Shotts, E. B., Jr., Jacobson, E. R. (1997). Shell disease in river cooters

(Pseudemys concinna) and yellow-bellied turtles (Trachemys scripta) in a

Georgia (USA) lake. Journal of Wildlife Diseases, 33, 78-86.

Gasc, J.-P. 1981. Axial musculature. In: Gans C, and Parsons TS (eds). Biology

of the Reptilia, vol. II. London: Academic Press, pp. 355-435.

Gilmore, C.W. 1923. A new species of Corythosaurus with notes on other Belly

River Dinosauria. The Canadian Field-Naturalist 37:46-52.

Gilmore, C.W. 1930. On dinosaurian reptiles from the Two Medicine Formation of

Montana. Proceedings of the United States National Museum 77:1-39.

Gilmore, C.W. 1933. Two new dinosaurian reptiles from Mongolia with notes on

some fragmentary specimens. American Museum Novitates 679:1-20.

190 Gillooly, J.F., Allen, A.P., and Charnov, E.L. 2006. Dinosaur fossils predict body

temperatures. PLoS Biology 4(8): e248. DOI:

10.1371/joumal.pbio.0040248.

Godefroit, P., Pereda Suberbiola, X., Li, H., and Dong, Z.-M. 1999. A new

species of the ankylosaurid dinosaur Pinacosaurus from the Late

Cretaceous of Inner Mongolia (P.R. China). Bulletin de L'institut Royal des

Sciences Naturelles de Belgique, Sciences de la Terre 69(Supp. B):17-36.

Hershkovitz, I., Rothschild, B.M., Dutour, O., and Greenwald, C. 1998. Clues to

recognition of fungal origin of lytic skeletal lesions. American Journal of

Physical Anthropology 106:47-60.

Hildebrand, M., and Goslow, G. 2001. Analysis of Vertebrate Structure, 5th

Edition. Toronto: John Wiley & Sons, Inc., 635 pp.

Hoets, M., and Bunch, D. 1992. Cranial asymmetry in a Dall sheep ram (Ovis

dalli dalli). Journal of Wildlife Diseases 28:330-332.

Holmes, R., and Organ, C. 2007. An ossified tendon trellis in Chasmosaurus

(Ornithischia: Ceratopsidae). Journal of Palaeontology 81:411-414.

191 Holmes, R.B., Forster, C, Ryan, M., and Shepherd, K.M. 2001. A new species of

Chasmosaurus (Dinosauria: Ceratopsia) from the Dinosaur Park

Formation of southern Alberta. Canadian Journal of Earth Sciences

38:1423-1438.

Honour, S., and Hickiing, K.M.H. 1993. Naturally occurring Brucella suis Biovar4

infection in a moose (Alces alces). Journal of Wildlife Diseases 29:596-

598.

Horner, J. 1992. Cranial morphology of Prosaurolophus (Omithischia:

Hadrosauridae) with description of two new hadrosaurid species and an

evaluation of hadrosaurid phylogenetic relationships. Museum of the

Rockies Occasional Paper 2. Bozeman, Montana: Museum of the

Rockies. 119 pp.

Hulke, J.W. 1881. Polacanthus foxii, a large undescribed dinosaur from the

Wealden Formation in the Isle of Wight. Philosophical Transactions of the

Royal Society 172:653-662.

Huchzermeyer, F. W. 2003. Crocodiles: biology, husbandry and diseases.

Cambridge, Massachusetts: CABI Publishing, 337 pp.

192 Huchzermeyer, F.W., and Cooper, J.E. 2000. Fibriscess, not abscess, resulting

from a localised inflammatory response to infection in reptiles and birds.

Veterinary Record 147:515-517.

Hulke, J.W. 1887. Supplemental note on Polacanthus foxii, describing the dorsal

shield and some parts of the endoskeleton, imperfectly known in 1881.

Philosophical Transactions of the Royal Society 178:169-172.

Keshner, E.A., Statler, K.D., and Delp, S.L. 1997. Kinematics of the freely moving

head and neck of the alert cat. Experimental Brain Research 115:257-266.

Kilbourne, B., and Carpenter, K. 2005. Redescription of Gargoyleosaurus

parkpinorum, a polacanthid ankylosaur from the Upper Jurassic of Albany

County, Wyoming. Neues Jahrbuch fur Geologie und Palaontologie-

Abhandlungen 237: 111-160.

Kirkland, J.I. 1998. A polacanthine ankylosaur (Ornithischia: Dinosauria) from the

Early Cretaceous (Barremian) of eastern Utah. In: Lucas, S.G., Kirkland,

J.I., and Estep, J.W. (eds.) Lower and Middle Cretaceous terrestrial

ecosystems. New Mexico Museum of Natural History and Science Bulletin

14, pp. 271-281.

193 Kirkland, J.I., and Carpenter, K. 1994. North America's first pre-Cretaceous

ankylosaur (Dinosauria) from the Upper Jurassic Morrison Formation of

western Colorado. BYU Geology Studies 40:25-42.

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.

Leahey, L, Molnar, R., and Salisbury, S. 2008. The cranial osteology of Minmi

sp., a basal ankylosauromorph (Omithischia: Dinosauria) from the Early

Cretaceous (Albian) Allaru Formation of Richmond, North-Western

Queensland, Australia. Journal of Vertebrate Paleontology 28:104A.

Lehman, T.M. 2001. Late Cretaceous dinosaur provinciality. In: Tanke, D.H., and

Carpenter, K. (eds.) Mesozoic Vertebrate Life: New Research Inspired by

the Paleontology of Philip J. Currie, pp. 310-328. Bloomington: Indiana

University Press, pp. 310-328.

Lu, J., Ji, Q., Gao, Y., and Li, Z. 2007. A new species of the ankylosaurid

dinosaur Crichtonsaurus (Ankylosauridae: Ankylosauria) from the

Cretaceous of Liaoning Province, China. Acta Geologica Sinica (English

Edition) 81:883-897.

194 Lucas, F.A. 1902. A new generic name for Stegosaurus marshi. Science 16:435.

Lucas, S.G. 2000. Pathological aetosaur armour from the Upper of

Germany. Stuttgarter Beitrage zur Naturkunde Serie B (Geologie und

Palaontologie) 281:1-6.

Maganaris, C.N., Baltzopoulos, V., Ball, D., and Sargeant, A.J. 2001. In vivo

specific tension of human skeletal muscle. Journal of Applied Physiology

90:865-872.

Maidment, S.C.R., Wei, G., and Norman, D.B. 2006. Re-description of the

postcranial skeleton of the Middle Jurassic stegosaur Huayangosaurus

taibaii. Journal of Vertebrate Paleontology 26:944-956.

Maidment, S., and Porro, L. 2007. Homology and function of the ornithischian

dinosaur palpebral. Journal of Vertebrate Paleontology 27:111 A.

Maidment, S.C.R., Norman, D.B., Barrett, P.M., and Upchurch, P. 2008.

Systematics and phylogeny of Stegosauria (Dinosauria: Omithischia).

Journal of Systematic Palaeontology 6:367-407.

Maleev, E.A. 1952. [A new ankylosaur from the Upper Cretaceous of Asia].

Doklady Akademii Nauk S.S.S.R. 87:273-276. [In Russian]

195 Maleev, E.A. 1954. [The armored dinosaurs of the Cretaceous Period in

Mongolia (Family Syrmosauridae)]. [Works of the Paleontological Institute

of the Academy of Sciences of the USSR] 48:142-170. [In Russian;

translation by Robert Welch]

Maleev, E.A. 1956. [Armoured dinosaurs from the Upper Cretaceous of

Mongolia]. [Works of the Paleontological Institute of the Academy of

Sciences of the USSR] 62:51-91. [In Russian]

Manning, P., Ali, J., McDonald, S., Mummery, P., and Sellers, W. 2007.

Biomechanics of dromaeosaurid claws: application of x-ray

microtomography, nanoindentation and finite element analysis. Journal of

Vertebrate Paleontology 27:111 A.

Mantell, G.A. 1833. Geology of the South East of England. London: Longman,

Rees, Orme, Brown, Green and Longman, 415 pp.

Manter, J.T.1940. The mechanics of swimming in the alligator. Journal of

Experimental Zoology 83:345-358.

Marsh, O.C. 1877. Notice of a new and gigantic dinosaur. American Journal of

Science Series 3, 14:87-88.

196 Martill, D.M., Batten, D.J., and Loydell, D.K. 2000. A new specimen of the

thyreophoran dinosaur cf. Scelidosaurus with soft tissue preservation.

Palaeontology 43:549-559.

Maryanska, T. 1971. Remains of armoured dinosaurs from the uppermost

Cretaceous in Nemegt Basin, Gobi Desert. Palaeontologia Polonica

21:23-32.

Maryanska, T. 1977. Ankylosauridae (Dinosauria) from Mongolia. Palaeontologia

Polonica 37:85-151.

Materialise. 2008. Mimics Version 12. Leuven, Belgium.

Mazzetta, G.V., Cisilino A.P., and Blanco, R.E. 2004. Mandible stress distribution

during the bite in Carnotaurus sastrei Bonaparte, 1985 (Theropoda:

Abelisauridae). Ameghiniana 41:605-617.

McGavin, M.D., and Zachary, J.F. 2007. Pathologic Basis of Veterinary Disease,

4th Edition. Missouri: Mosby, Inc., 1476 pp.

197 McWhinney, L. A., Rothschild, B. M., and Carpenter, K. 2001. Posttraumatic

chronic osteomyelitis in Stegosaurus dermal spikes. In: Carpenter, K. (ed.)

The Armoured Dinosaurs. Indianapolis: Indiana University Press, pp. 141—

156.

Molnar, R.E. 1980. An ankylosaur (Omithischia: Reptilia) from the Lower

Cretaceous of southern Queensland. Memoirs of the Queensland

Museum 20: 77-87.

Moreno, K., Carrano, M.T., and Snyder, R. 2007. Morphological changes in pedal

phalanges through omithopod dinosaur evolution: A biomechanical

approach. Journal of Morphology 268:50-63.

Nopcsa, F.B. 1905. Notes on British Dinosaurs. Part II: Polacanthus. Geological

Magazine 2:241-250.

Nopsca, F. 1928. Palaeontological notes on reptiles. Geologica Hungarica, Seria

Palaeontologica 1:1-84.

Norman, D.B., Witmer, L.M., and Weishampel, D.B. 2004. Basal Thyreophora. In:

Weishampel, D.B., Dodson, P., and Osmolska, H. (eds.) The Dinosauria,

2nd Edition. Berkeley, California: University of California Press, pp. 335-

342.

198 Ogden, J.A., Rhodin, A.G.J., Conlogue, G.J., Light, T.R. 1981. Pathobiology of

septic arthritis and contiguous osteomyelitis in a leatherback turtle. Journal

of Wildlife Diseases 17: 277-287.

Organ, C.L. 2006a. Biomechanics of ossified tendons in omithopod dinosaurs.

Paleobiology 32:652-665.

Organ, C.L. 2006b. Thoracic epaxial muscles in living archosaurs and omithopod

dinosaurs. The Anatomical Record Part A 228A:782-293.

The OsiriX Foundation. 2007. Osirix. Available at http://www.osirix-viewer.com/.

Geneva, Switzerland.

Ostrom, J.H. 1970. Stratigraphy and paleontology of the Cloverly Formation

(Lower Cretaceous) of the Bighorn Basin area, Wyoming and Montana.

Peabody Museum of Natural History Bulletin 35: 1-205.

Pang, Q., and Cheng, Z. 1998. A new ankylosaur of Late Cretaceous from

Tianzhen, Shanxi. Progress in Natural Science 8:326-334.

Parks, W.A. 1924. Dyoplosaurus acutosquameus, a new genus and species of

armored dinosaur; with notes on a skeleton of Prosaurolophus maximus.

University of Toronto Studies, Geological Series 18:1-35.

199 Peattie, A.M., Majidi, C, Corder, A., Jull, and R.J. 2007. Ancestrally high elastic

modulus of gecko setal B-keratin. Journal of the Royal Society Interface

4:1071-1076.

Penkalski, P. 2001. Variation in specimens referred to Euoplocephalus tutus. In:

Carpenter, K. (ed.) The Armored Dinosaurs. Indianapolis: Indiana

University Press, 261-298.

Picard, K., Festa-Bianchet, M., and Thomas, D. 1996. The cost of horniness:

heat loss may counter sexual selection for large horns in temperate

bovids. Ecoscience 3:280-284.

Porro, L. 2007. Feeding and jaw mechanism in Heterodontosaurus tucki using

finite element analysis. Journal of Vertebrate Paleontology 27: 131A.

Porro, L. 2008. Accuracy in finite element modeling of extinct taxa: sensitivity

analyses in Heterodontosaurus tucki. Journal of Vertebrate Paleontology

28:128A.

Rasband, W. S. 2008. ImageJ. Available at http://rsb.info.nih.gov/ij/. U. S.

National Institutes of Health, Bethesda, Maryland, USA.

200 Rayfield, E.J., Norman, D.B., Horner, C.C., Horner, J.R., Smith, P.M., Thomason,

J.J., and Upchurch, P. 2001. Cranial design and function in a large

theropod dinosaur. Nature 409:1033-1037.

Rayfield, E.J. 2004. Cranial mechanics and feeding in rex.

Proceedings of the Royal Society of London Series B 271:1451-1459.

Rayfield, E.J. 2005a. Aspects of comparative cranial mechanics in the theropod

dinosaurs Coelophysis, Allosaurus, and Tyrannosaurus. Zoological

Journal of the Linnean Society 144:309-316.

Rayfield, E.J. 2005b. Using finite-element analysis to investigate suture

morphology: a case study using large carnivorous dinosaurs. Anatomical

Record Part 283A:349-365.

Rayfield, E.J. 2007. Finite element analysis and understanding the biomechanics

and evolution of living and fossil organisms. Annual Review of Earth and

Planetary Sciences 35:541-576.

Resnick, D. 2002. Diagnosis of Bone and Joint Disorders, 4th Edition. New York:

W.B. Saunders Company, 5760 pp.

201 Reynolds, H.W., Gates, C.C., and Glaholt, R.D. 2003. Bison (Bison bison). In:

Feldhamer, G.A., Thompson, B.C., and Chapman, J.A. (eds.) Wild

Mammals of North America: Biology, Management, and Conservation.

Baltimore, Maryland: JHU Press, pp. 1009-1060.

Rosskopf, W.J., Jr., and Shindo, M.K. 2003. Syndromes and conditions of

commonly kept tortoise and turtle species. Seminars in Avian and Exotic

Pet Medicine 12:149-161.

Rothschild, B.M., and Martin, L.D. 1987. Avascular necrosis: occurrence in diving

Cretaceous mosasaurs. Science 236:75-77.

Rothschild, B.M., and Martin, L.D. 1993. Paleopathology: Disease in the Fossil

Record. Boca Raton: CRC Press, 386 pp.

Rothschild, B. M., and Martin, L. D. 2006. Skeletal impact of disease.

Albuquerque: New Mexico Museum of Natural History & Science Bulletin

33, 226 pp.

Rothschild, B.M., and Panza, R.K. 2005. Epidemiologic assessment of trauma-

independent skeletal pathology in non-passerine birds from museum

collections. Avian Pathology 34:212-219.

202 Rothschild, B.M., Tanke, D., and Carpenter, K. 1997. Tyrannosaurs suffered from

gout. Nature 387:357.

Rothschild, B.M., Hershkovitz, I., and Dutour, O. 1998. Clues potentially

distinguishing lytic lesions of multiple myeloma from those of metastatic

carcinoma. American Journal of Physical Anthropology 105:241-250.

Rothschild, B.M., Witzke, B.J., Hershkovitz, I. 1999. Metastatic cancer in the

Jurassic. The Lancet 354: 398.

Rothschild, B.M., Tanke, D.H., Helbling, M., and Martin, L.D. 2003. Epidemiologic

study of tumors in dinosaurs. Naturwissenschaften 90:495-500.

Ryan, M.J. 2007. A new basal centrosaurine ceratopsid from the Oldman

Formation, southeastern Alberta. Journal of Paleontology 81:376-396.

Ryan, M. J., and Evans, D. C. 2005. Ornithischian dinosaurs. In: Currie, P.J., and

Koppelhus, E.B. (eds.) Dinosaur Provincial Park: A spectacular ancient

ecosystem revealed. Indianapolis: Indiana University Press, pp. 312-348.

203 Ryan, M.J., and Russell, A.P. 2005. A new centrosaurine ceratopsid from the

Oldman Formation of Alberta and its implications for centrosaurine

and systematics. Canadian Journal of Earth Sciences 42:1369-

1387.

Ryan, M.J., Holmes, R., and Russell, A.P. 2007. A revision of the late

Campanian centrosaurine ceratopsid genus Styracosaurus from the

Western Interior of North America. Journal of Vertebrate Paleontology

27:944-962.

Salgado, L, and Gasparini, Z. 2006. Reappraisal of an ankylosaurian dinosaur

from the Upper Cretaceous of James Ross Island (Antarctica).

Geodiversitas 28:119-135.

Sawyer, G.T., and Erickson, B.R. 1998. Paleopathology of the Paleocene

crocodile Leidyosuchus (=Borealosuchus) formidabilis. St. Paul: The

Science Museum of Minnesota Monograph Vol. 4: Paleontology, 38 pp.

Seebacher, F. 2003. Dinosaur body temperatures: the occurrence of endothermy

and ectothermy. Paleobiology 29:105-122.

204 Seki, Y., Kad, B., Benson, D., and Meyers, MA 2006. The toucan beak:

structure and mechanical response. Materials Science and Engineering C

26:1412-1420.

Sereno, P.C. 1986. Phylogeny of the bird-hipped dinosaurs (Order Ornithischia).

National Geographic Research 2:234-256.

Sereno, P.C. 1998. A rationale for phylogenetic definitions, with application to the

higher-level taxonomy of Dinosauria. Neues Jahrbuch fur Geologie und

Palaontologie-Abhandlungen 210:41-83.

Shychoski, L, Rayfield, E., and Sakamoto, M. 2007. The biomechanics of scale:

a phylogenetic and ontogenetic investigation of tyrannosaurid dinosaurs

using geometric morphometries and the finite element method. Journal of

Vertebrate Paleontology 27:146A.

Simmons, R.E., and Scheepers, L. 1996. Winning by a neck: sexual selection in

the evolution of giraffe. The American Naturalist 148:771-786.

Snively, E., and Cox, A. 2008. Structural mechanics of pachycephalosaur crania

permitted head-butting behavior. Palaeontologia Electronica 11: 17p.

205 Snively, E., and Russell, A. 2002. The tyrannosaurid metatarsus: bone strain and

inferred ligament function. Senckenbergiana Lethaea 82:35-42.

Snively, E., and Russell, A.P. 2007. Craniocervical feeding dynamics of

Tyrannosaurus rex. Paleobiology 33:610-638.

Steinbock, R.T. 1976. Paleopathology Diagnosis and Interpretation: Bone

Diseases in Ancient Human Populations. Illinois: Charles C Thomas

Publisher, 423 pp.

Sternberg, CM. 1929. A toothless armoured dinosaur from the Upper

Cretaceous of Alberta. Bulletin of the National Museum of Canada 54:34-

37.

Stevens, K.A., Parrish, J.M. 1999. Neck posture and feeding habits of two

Jurassic sauropod dinosaurs. Science 284:798-799.

Strait, D.S., Wang, Q., Dechow, P.C., Ross, C.F., Richmond, B.G., Spencer,

M.A., and Patel, B.A. 2005. Modeling elastic properties in finite-element

analysis: how much precision is needed to produce an accurate model?

The Anatomical Record Part A 283A:275-287.

Strand7 Pty Ltd. 2008. Strand7. Sydney, Australia.

206 Sullivan, R.M. 1999. Nodocephalosaurus kirtlandensis gen. et sp. nov., a new

ankylosaurid dinosaur (Omithischia: Ankylosauria) from the Upper

Cretaceous (Late Campanian) Kirtland Formation of New Mexico. Journal

of Vertebrate Paleontology 19:126-139.

Sullivan, R.M., and Fowler, D.W. 2006. New specimens of the rare ankylosaurid

dinosaur Nodocephalosaurus kirtlandensis (Omithischia: Ankylosauridae)

from the Upper Cretaceous Kirtland Formation (De-na-zin Member), San

Juan Basin, New Mexico. In: Lucas, S.G. and Sullivan, R.M. (eds.) Late

Cretaceous vertebrates from the Western Interior. New Mexico Museum of

Natural History and Science Bulletin 35, p. 259-261.

Tanke, D.H., and Farke, A.A. 2007. Bone resorption, bone lesions, and

extracranial fenestrae in ceratopsid dinosaurs: a preliminary assessment.

In: Carpenter, K. (ed.) Horns and . Indianapolis: Indiana University

Press, pp. 319-348.

Tsuihiji, T. 2007. Homologies of the Longissimus, lliocostalis, and hypaxial

muscles in the anterior presacral region of extant Diapsida. Journal of

Morphology 268:986-1020.

207 Tumanova, T.A. 1983. [The first ankylosaur from the Lower Cretaceous of

Mongolia]. Trudy Sovmestnaya Sovetsko-Mongolskaya

Paleontologicheskaya Ekspeditsiya 24:110-120. [In Russian with English

summary]

Tumanova, T.A. 1993. [Concerning a new armored dinosaur from the

southeastern Gobi]. Paleontologicheskii Zhurnal 27(2):92-98. [In Russian

with English summary]

Vickaryous, M.K., and Hall, B.K. 2006. Osteoderm morphology and development

in the nine-banded armadillo, Dasypus novemcinctus (Mammalia,

Xenarthra, Cingulata). Journal of Morphology 267:1273-1283.

Vanden Berge, J.C., and Zweers, G.A. 1993. Myologia. In: Baumel, J.J., King,

A.S., Breazile, J.E., Evans, H.E., and Vanden Berge, J.C. (eds.)

Handbook of Avian Anatomy: Nomina Anatomica Avium, Second Edition.

Cambridge, Massachusetts: Nuttall Ornithological Club, pp. 189-250.

Vickaryous, M.K., and Hall, B.K. 2008. Development of the dermal skeleton in

Alligator mississippiensis (Archosauria, Crocodylia) with comments on the

homology of osteoderms. Journal of Morphology 269:398-422.

208 Vickaryous, M.K., Russell, A.P., Currie, P.J., and Zhao, X.-J. 2001. A new

ankylosaurid (Dinosauria: Ankylosauria) from the Lower Cretaceous of

China, with comments on ankylosaurian relationships. Canadian Journal

of Earth Sciences 38:1767-1780.

Vickaryous, M.K., Maryanska, T., and Weishampel, D.B. 2004. Ankylosauria. In:

Weishampel, D.B., Dodson, P., and Osmolska, H. (eds.) The Dinosauria,

2nd Edition. University of California Press, pp. 363-392.

White, T.D. 2000. Human Osteology, 2nd Edition. San Diego: Academic Press,

563 pp.

Wirtz, D.E., Schiffers, N., Pandorf, T., Radermacher, K., Weichert, D., and Forst,

R. 2000. Critical evaluation of known bone material properties to realize

anisotropic FE-simulation of the proximal femur. Journal of Biomechanics

33:1325-1330.

Witmer, L.M. 1995. The Extant Phylogenetic Bracket and the importance of

reconstructing soft tissues in fossils. In: Thomason, J.J. (ed.) Functional

Morphology in Vertebrate Paleontology. New York: Cambridge University

Press, pp. 19-33.

209 Wolff, E.D.S., Fowler, D.W., and Bonde, J.W. 2007. A possible case of

necrotizing dermatitis in the crocodylian Diplocynodon, from the Oligocene

of the Isle of Wight, United Kingdom. Historical Biology 19(2):203-207.

Young, C.C. 1935. On a new nodosaurid from Ninghsia. Palaeontologia Sinica

Series C 11:1-35.

Zollikofer, C.P.E., Ponce de Leon, M.S. 2005. Virtual reconstruction: a primer in

computer-assisted paleontology and biomedicine. New Jersey: John Wiley

& Sons, Inc., 333 pp.

210 Appendix 1

Material examined. Taxonomic assignment of specimens is based on museum catalogue information and previously published identifications.

Omithischia indet.

TMP 1987.48.90, TMP 1998.93.65, TMP 2003.12.166

Stegosauria

Stegosaurus sp.—AMNH 548

Ankylosauria indet.

CMN 842, DMNH 2451, TMP 1966.17.28, TMP 1983.36.120, TMP

1993.36.76, TMP 1993.36.421, TMP 1994.168.1, TMP 1998.93.55, TMP

2005.12.43

Ankylosauridae

Ankylosaurus magniventris— AMNH 5214, AMNH 5895 (holotype)

Euoplocephalus tutus— AMNH 5211, AMNH 5245, AMNH 5337, AMNH

5403, AMNH 5404, AMNH 5405, AMNH 5406, AMNH 5409, AMNH 5470,

CMN 0210 (holotype), CMN 268, CMN 349, CMN 2234, CMN 2251, CMN

2252, CMN 2253, CMN 8530, CMN 40605, ROM 784, ROM 788, ROM

211 1930, ROM 7761, TMP 1982.9.3, TMP 1985.36.70, TMP 1991.36.321,

TMP 1992.36.334, TMP 1997.132.01, TMP 2000.57.3,

Gargoyleosaurus parkpinorum—DMNH 27726

Gastonia sp.—DMNH 45412, DMNH 45416, DMNH 45430, DMNH 45577,

DMNH 45686, DMNH 46984, DMNH 46986, DMNH 47007, DMNH 47026,

DMNH 47093, DMNH 47098, DMNH 47259, DMNH 47172, DMNH 47194,

DMNH 48065, DMNH 48078, DMNH 48087, DMNH 48103, DMNH 48426,

DMNH 48478, DMNH 48770, DMNH 48776, DMNH 49542, DMNH 49543,

DMNH 49604, DMNH 49601, DMNH 49608, DMNH 49616, DMNH 49703,

DMNH 49704, DMNH 49708, DMNH 49715, DMNH 50598, DMNH 51736,

DMNH 51743, DMNH 51756, DMNH 51950, DMNH 51959, DMNH 51962,

DMNH 51964, DMNH 52602, DMNH 52604, DMNH 52608, DMNH 52621,

DMNH 52629, DMNH 52631, DMNH 52636, DMNH 52790, DMNH 53067,

DMNH 53071, DMNH 53074, DMNH 53075, DMNH 53077, DMNH 53113,

DMNH 53117, DMNH DMNH 53120, 53267, DMNH 53340, DMNH 53350,

DMNH 53452, DMNH 53451, DMNH 53487, DMNH 53732, DMNH 53733,

DMNH 53736, DMNH 53883, DMNH 53899

Pinacosaurus grangeri— AMNH 2062 (cast of holotype, Heishansaurus pachycephalus), MPC PJC 2006-158, MPC PJC 2007-123, TMP

1990.301.2, TMP 1990.301.4.

212 Saichania chulsanensis—MPC 100/1305.

Talarurus plicatospineus—MPC KID2007.167.

Tarchia gigantea—UALVP 47948 (cast of ZPAL MgD-l/113), ZPAL MgD-

I/43.

Ankylosauridae indet—TMP 1984.121.33, TMP 1993.36.76, TMP

2005.09.75, TMP 2007.020.0080, TMP 2007.020.0100

Nodosauridae

Edmontonia longiceps—CMN 8251 (holotype)

Edmontonia sp.— TMP 2000.12.083

Panoplosaurus mirus—CMN 2769 (holotype)

Sauropelta edwardsi—AMNH 3032 Hadrosauridae indet.

TMP 1980.16.18, TMP 1981.23.8, TMP 1987.36.221, TMP 1990.36.399,

TMP 1993.36.115, TMP 1991.36.147, TMP 1991.36.513, TMP

1993.36.601, TMP 1996.666.11

Ceratopsia indet.

TMP 1981.18.83, TMP 1993.36.412

Centrosaurus sp.—TMP 1996.12.219

Ornithomimidae indet.

TMP 1979.14.715, TMP 1993.36.155

Tyrannosauridae indet.

TMP 1992.36.352, TMP 1997.12.204

Specimens from the University of Alberta Department of Anthropology:

Human Osteology Collection

9794 V8S2, 9794 V2067, 9794 340, 992.11.2h PA981Q, 9794066 25

Zooarchaeology Collection

Ursus americanus—981.55

Ursus arctos—982.12.1

Gulo luscus—982.23.2

Alcesalces—981.13 Michael Bums Department of Biological Sciences CW 405 Biological Sciences Centre University of Alberta Edmonton, AB, Canada T6G 2E9

10 November 2008

Dear Mr. Burns,

I am completing a Masters thesis at the University of Alberta entitled "Evolution, biomechanics, and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora)". I would like permission to allow inclusion of the following material in the thesis and permission for the National Library to make use of the thesis (i.e. to reproduce, loan, distribute, or sell copies of the thesis by any means and in any form or format).

These rights will in no way restrict republication of the material in any other form by you or by others authorized by you.

The co-authored chapter to be reprinted is:

Arbour, V., Sissons, R.L., and Burns, M.E. A redescription of the ankylosaurid dinosaur Dyoplosaurus acutosquameus Parks, 1924 (Ornithischia: Ankylosauria) and a revision of the genus. Submitted 22 August 2008 to the Journal of Vertebrate Paleontology.

If these arrangements meet with your approval, please sign this letter where indicated below and return it to me in the enclosed return envelope. Thank you for your assistance in this matter.

Sincerely,

Victoria Arbour

Permission Granted for the Use Requested Above:

Signature Print Name Date Robin Sissons Department of Biological Sciences CW 405 Biological Sciences Centre University of Alberta Edmonton, AB, Canada T6G 2E9

10 November 2008

Dear Ms. Sissons,

I am completing a Masters thesis at the University of Alberta entitled "Evolution, biomechanics, and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora)". I would like permission to allow inclusion of the following material in the thesis and permission for the National Library to make use of the thesis (i.e. to reproduce, loan, distribute, or sell copies of the thesis by any means and in any form or format).

These rights will in no way restrict republication of the material in any other form by you or by others authorized by you.

The co-authored chapter to be reprinted is:

Arbour, V., Sissons, R.L., and Burns, M.E. A redescription of the ankylosaurid dinosaur Dyoplosaurus acutosquameus Parks, 1924 (Ornithischia: Ankylosauria) and a revision of the genus. Submitted 22 August 2008 to the Journal of Vertebrate Paleontology.

If these arrangements meet with your approval, please sign this letter where indicated below and return it to me in the enclosed return envelope. Thank you for your assistance in this matter.

Sincerely,

Victoria Arbour

Permission Granted for the Use Requested Above:

Signature Print Name Date Dr. Eric Snively Department of Biological Sciences CW 405 Biological Sciences Centre University of Alberta Edmonton, AB, Canada T6G 2E9

10 November 2008

Dear Mr. Snively,

I am completing a Masters thesis at the University of Alberta entitled "Evolution, biomechanics, and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora)". I would like permission to allow inclusion of the following material in the thesis and permission for the National Library to make use of the thesis (i.e. to reproduce, loan, distribute, or sell copies of the thesis by any means and in any form or format).

These rights will in no way restrict republication of the material in any other form by you or by others authorized by you.

The co-authored chapter to be reprinted is:

Arbour, V., and Snively, E. Biomechanics and function of the tail club in ankylosaurid dinosaurs (Ornithischia: Thyreophora). Submitted 10 October 2008 to the Anatomical Record, Part A.

If these arrangements meet with your approval, please sign this letter where indicated below and return it to me in the enclosed return envelope. Thank you for your assistance in this matter.

Sincerely,

Victoria Arbour

Permission Granted for the Use Requested Above:

Signature Print Name Date