DIVERSITY AND SYSTEMATICS OF MARSUPIAL LIONS FROM THE RIVERSLEIGH WORLD HERITAGE AREA AND THE EVOLUTION OF THE THYLACOLEONIDAE

ANNA K GILLESPIE

Submitted in fulfilment of the requirements for the degree of Doctor of Philosophy in the School of Biological, Earth and Environmental Sciences, University of New South Wales

2007 i

ABSTRACT

The fossil record of marsupial lions (family Thylacoleonidae) from Australian Oligo- Miocene deposits is generally poor. Study of new material of this family collected from Oligo-Miocene limestone sediments of the Riversleigh World Heritage Area, northwestern Queensland adds significant new information about previously described species and also indicates a greater diversity of thylacoleonids during this period of geological time. Two new genera and five new species are described.

Reassessment of the holotype of the type species of Priscileo, P. pitikantensis, indicates it shows stronger affinities to species of the genus Wakaleo than it does to Priscileo roskellyae. Priscileo is regarded here to be a junior synonym of Wakaleo. The cranium and lower dentition of Priscileo roskellyae show significant morphological differences from species of Wakaleo, and this species is referred to a new genus, Lekaneleo. Distinctive morphological differences are identified in the M3s of Wakaleo oldfieldi and W. vanderleueri, species previously distinguished only by relative size differences in their dentitions.

Functional morphological assessment of postcranial remains of species of Wakaleo suggests that they were probably scansorial or arboreal, but does not support a previous hypothesis of a fossorial habit.

Cladistic analyses of the interrelationships of marsupial lions support the referral of Priscileo pitikantensis to the genus Wakaleo. The monotypic genus Microleo is the sister-group to all remaining thylacoleonid taxa. Species of Lekaneleo are the sister- group to a Wakaleo/Thylacoleo clade. Intraordinal relationships of thylacoleonids were also investigated. Phylogenetic analyses of the interrelationships of Diprotodontia that included representatives of all extinct vombatiform families as well as extant taxa were conducted employing cranial and dental morphological characters. These analyses provide support for the hypothesis that Thylacoleonidae are members of the suborder Vombatiformes. ii

Two species of Wakaleo (W. oldfieldi and W. vanderleueri) present in Riversleigh deposits are also found at other localities - respectively, the Leaf Locality of central Australia (Kutjamarpu LF) and the Small Hills Locality of northern Australia (Bullock Creek LF) - and suggest age estimations of the relevant Riversleigh sites of early Miocene and late Miocene. The phyletic evolution of Wakaleo suggests that some Riversleigh deposits are probably late Oligocene in age. iii

TABLE OF CONTENTS

Abstract i Table of contents iii List of figures vii List of tables ix Acknowledgements x Publications xii

Chapter 1. Introduction 1

Chapter 2. Materials and Methods 3

Chapter 3. Thylacoleonid Diversity 10

Chapter 4. Re-assessment of Priscileo pitikantensis Rauscher 1987 and the identification of a plesiomorphic species of Wakaleo from Riversleigh 18 4.1 Introduction 18 4.2 Re-assessment of Priscileo pitikantensis Rauscher, 1987 19 4.3 Systematics 22 4.4 Specific attribution of a plesiomorphic Wakaleo species from Riversleigh 24 4.5 Wakaleo hilmeri sp. nov. 28

Chapter 5. Postcranial bones of Wakaleo hilmeri 77 5.1 Introduction 77 5.2.0 Description and Comparisons 81 5.2.1 Vertebral column 81 5.2.2 Cervical vertebra 81 5.2.3 Lumbar vertebrae 82 5.2.4 Sacrum 86 5.3.0 Pelvic girdle 90 5.3.1 Ilia 90 5.4.0 Forelimb 91 5.4.1 Humerus 91 5.4.2 Radius 97 5.4.3 Ulna 102 5.5.0 Manus 103 5.5.1 Hamatum 103 5.5.2 Trapezium 107 5.5.3 Metacarpal I 109 5.5.4 Pollex 111 5.5.5 Digits 111 5.6.0 Pes 115 iv

5.6.1 Calcaneum 115 5.7 Amended description of the scapula of Wakaleo pitikantensis 116

Chapter 6. Wakaleo sp. cf. W. hilmeri 119 6.1 Introduction 119 6.2 Wakaleo sp. cf. W. hilmeri 119

Chapter 7. New material of Wakaleo oldfieldi and W. vanderleueri 128 7.1 Introduction 128 7.2 New material of W. oldfieldi 128 7.3 New material of W. vanderleueri 143

Chapter 8. Lekaneleo roskellyae new combination 152 8.1 Introduction 152 8.2 Systematics 153 8.3 Description of skull, maxillae and dentary 157

Chapter 9. Lekaneleo deminutivus sp. nov. 200 9.1 Introduction 200 9.2 Systematics 200

Chapter 10. Microleo attenboroughi gen. et sp. nov. 207 10.1 Introduction 207 10.2 Systematics 207

Chapter 11. Thylacoleonid intrafamilial relationships 219 11.1 Historical background 219 11.2 Aims 222 11.3 Methods 223 11.3.1 Ingroup and outgroup taxa 223 11.3.2 Character selection and description 224 11.3.3 Parsimony analyses 243 11.4 Results 245 11.5 Discussion 251

Chapter 12. Thylacoleonid interfamilial relationships 258 12.1 Use of names 258 12.2 Historical background 258 12.3 Methods 264 12.3.1 Ingroup and outgroup taxa 264 12.3.2 Character selection and description 265 12.3.3 Parsimony analyses 284 12.4 Results 284 12.5 Discussion 286

Chapter 13. Functional assessment of the postcrania of W. hilmeri sp. nov. and W. pitikantensis 296 13.1 Introduction 296 v

13.2 Methods and materials 299 13.3 Results and discussion 305

Chapter 14. Biostratigraphy, Biochronology and Palaeoecology of Riversleigh thylacoleonids 334 14.1 Introduction 334 14.2 Riversleigh geology 334 14.3 Riversleigh Biostratigraphy and Biocorrelation 338 14.4 Intracontinental correlations 342 14.5 Diversity and palaeoenvironmental indications 348

Chapter 15. Conclusions 352

References 356

Appendices 369 Appendix 1: Dentition measurements of all thylacoleonid specimens from Riversleigh and of comparative specimens from other localities 369 Table A: Measurements of upper and lower cheekteeth for all species of Wakaleo 369 Table B: Upper cheek dentition measurements for species of Lekaneleo 371 Table C: Lower cheek dentition measurements for specimens of Lekaneleo roskellyae 371 Table D: Dentition measurements for specimens of Microleo attenboroughi 371 Table E: Comparative measurements of cranial dimensions of species of thylacoleonids 372 Table F: Comparative measurements of upper tooth rows and palate dimensions of Wakaleo hilmeri, W. vanderleueri, W. pitikantensis and Lekaneleo roskellyae 373 Table G: Dentary measurements of Wakaleo hilmeri, W. oldfieldi, W. vanderleueri and Lekaneleo roskellyae (mm) 373 Appendix 2: Comparative specimens used in postcranial and functional morphological studies 374 Appendix 3: Specimens used in intrafamilial and interfamilial phylogenetic analyses 377 Appendix 4: Nodal apomorphies for most- parsimonious trees from intrafamilial analysis: A. unordered-character states; B. terminal taxa; C. ordered character- states 379 Appendix 5: Nodal apomorphies for most-parsimonious trees from interfamilial analysis: A. unordered character states; B. ordered character-states 388 Appendix 6: Publications……………………………………..in sleeve on back page vi

Gillespie, A. 1997. Priscileo roskellyae sp.nov. (Thylacoleonidae, Marsupialia) from the Oligocene-Miocene of Riversleigh, northwestern Queensland). Memoirs of the Queensland Museum. 41(2): 321-327. Wroe, S., Myers, T.J., Wells, R.T. and Gillespie, A. 1999. Estimating the weight of the Pleistocene marsupial lion, Thylacoleo carnifex (Thylacoleonidae: Marsupialia): implications for the ecomorphology of a marsupial super-predator and hypotheses of impoverishment of Australian marsupial carnivore faunas. Australian Journal of Zoology. 47: 489-498. Archer, M., Arena, R., Bassarova, M., Black, K., Brammall, J., Cooke, B., Creaser, P., Crosby, K., Gillespie, A., Godthelp, H., Gott, M., Hand, S.J., Kear, B., Krikmann, A., Mackness, B., Muirhead, J., Musser, A., Myers, T., Pledge, N., Wang, Y., and Wroe S. 1999. The evolutionary history and diversity of Australian mammals. Australian Mammalogy. 21: 1-45. vii

LIST OF FIGURES

Fig. 3.1 Distribution of fossil sites yielding thylacoleonid species 11 Fig. 4.1 Priscileo pitikantensis, SAM P37719 holotype 20 Fig. 4.2 Comparative measurements of maxilla alveoli and M2s of SAM P37719 and Riversleigh specimens 20 Fig. 4.3 Cranium of Wakaleo hilmeri sp. nov., QM F45200 holotype, dorsal view 30 Fig. 4.4 Cranium of Wakaleo hilmeri, QM F45200, profile view 32 Fig. 4.5 Cranium of Wakaleo hilmeri, QM F45200, ventral view 37 Fig. 4.6 Cranial measurements of Wakaleo hilmeri, QM F45200 42 Fig. 4.7 Occiput of Wakaleo hilmeri, QM F45200 46 Fig. 4.8 Left basicranial region of Wakaleo hilmeri, QM F45200 49 Fig. 4.9 Wakaleo hilmeri, QM F23443, palate; Rackham’s Low Lion Site 49 Fig. 4.10 Wakaleo hilmeri, QM F45200, left cheek dentition 54 Fig. 4.11 Wakaleo hilmeri, AR 19249, right dentary; a. lingual; b. buccal; c. measurements 59 Fig. 4.12 Wakaleo hilmeri, AR 19249; a. paired dentaries; b. left dentary, stereo occlusal view 62 Fig. 5.1 Wakaleo hilmeri cervical vertebra 80 Fig. 5.2 Wakaleo hilmeri lumbar vertebra 80 Fig. 5.3 Wakaleo hilmeri lumbar vertebrae 80 Fig. 5.4 Wakaleo hilmeri sacrum 80 Fig. 5.5 Wakaleo hilmeri ilia 89 Fig. 5.6 Wakaleo hilmeri humerus 92 Fig. 5.7 Humeri of Wakaleo hilmeri and Wakaleo pitikantensis 92 Fig. 5.8 Thylacoleonid humeri, anterior view 94 Fig. 5.9 Wakaleo hilmeri radius 98 Fig. 5.10 Radii of species of Wakaleo and Thylacoleo carnifex 98 Fig. 5.11 Wakaleo hilmeri ulna 101 Fig. 5.12 Wakaleo hilmeri unciform 104 Fig. 5.13 Wakaleo hilmeri trapezium 104 Fig. 5.14 Wakaleo hilmeri metacarpal I 110 Fig. 5.15 Wakaleo hilmeri proximal phalanges: A. right proximal phalanx digit; B. left pollex, proximal phalanx 110 Fig. 5.16 Wakaleo hilmeri left pollex and right digit II 110 Fig. 5.17 Wakaleo hilmeri calcaneum 114 Fig. 5.18 Wakaleo pitikantensis (SAM P37720) scapula 114 Fig. 6.1 Wakaleo sp. cf. W. hilmeri, QM F23446, occlusal view 120 Fig. 6.2 Wakaleo sp. cf. W. hilmeri, QM F23446, profile view 120 Fig. 6.3 Wakaleo sp. cf. W. hilmeri, QM F23446, measurements of maxilla 120 Fig. 7.1 Wakaleo oldfieldi, QM F20895, left dentary 130 Fig. 7.2 Measurements of QM F20895 133 Fig. 7.3 Wakaleo oldfieldi, QM F36437, right dentary, stereo occlusal view 133 viii

Fig. 7.4 Wakaleo oldfieldi, QM F23428, left maxilla fragment, Jim’s Jaw Site 135 Fig. 7.5 Wakaleo oldfieldi, AR 19248, M1, Kutjamarpu LF 135 Fig. 7.6 Wakaleo oldfieldi M2; a.QM F31398 and b. AR 3272. 135 Fig. 7.7 Wakaleo oldfieldi, QM F31294, M3 135 Fig. 7.8 Wakaleo vanderleueri, left dentary, QM F36453 144 Fig. 7.9 Dimensions of Wakaleo vanderleueri, QM F36453, left dentary 146 Fig. 7.10 Wakaleo vanderleueri, QM F41204, left P3 146 Fig. 7.11 Wakaleo vanderleueri, QM F52182, M3 146 Fig. 7.12 Wakaleo vanderleueri, QM F31352, left canine 146 Fig. 8.1 Cranium of Lekaneleo roskellyae, QM F23453 holotype, dorsal view 156 Fig. 8.2 Cranium of Lekaneleo roskellyae, QM F23453 holotype, profile view 158 Fig. 8.3 Cranium of Lekaneleo roskellyae, QM F23453 holotype, ventral view 161 Fig. 8.4 Cranial measurements of Lekaneleo roskellyae, QM F23453 164 Fig. 8.5 Basicranium of Lekaneleo roskellyae, QM F23453 171 Fig. 8.6 Lekaneleo roskellyae, QM F23442, right maxilla, stereo pair 176 Fig. 8.7 Lekaneleo roskellyae, QM F40116, right maxilla, stereo pair 176 Fig. 8.8 Lekaneleo roskellyae, QM F20191, right dentary 180 Fig. 8.9 Dimensions of Lekaneleo roskellyae, QM F20191, right dentary 180 Fig. 8.10 Lekaneleo roskellyae, QM F29623, right dentary 182 Fig. 9.1 Lekaneleo deminutivus sp. nov., QM F24549 holotype, P3 201 Fig. 10.1 Microleo attenboroughi gen. et sp. nov., QM F41143 holotype 209 Fig. 10.2 Microleo attenboroughi, QM F42676, M2 209 Fig. 11.1a-c Hypotheses of thylacoleonid relationships 220 Fig. 11.2a Consensus of 9 most-parsimonious trees obtained from unordered parsimony analysis 244 Fig. 11.2b One of 9 equally-parsimonious trees, unordered characters 244 Fig. 11.3 Consensus of 21 equally-parsimonious trees from ordered analysis 246 Fig. 11.4 One of 21 equally-parsimonious trees, ordered characters 247 Fig. 11.5a-b Consensuses of 4 equally-parsimonious trees obtained from unordered analysis with Lekaneleo deminutivus excluded 249 Fig. 11.6 Consensuses of 3 equally-parsimonious trees obtained from ordered analysis with Lekaneleo deminutivus excluded 249 Fig. 11.7 Possible interrelationships of thylacoleonids incorporating taxonomic changes 256 Fig. 12.1a - c Hypotheses of thylacoleonid relationships 260 Fig. 12.2 The most-parsimonious tree resulting from unordered analysis 283 Fig. 12.3 The most-parsimonious tree resulting from ordered analysis 285 Fig. 13.1 Humerus measurements 301 Fig. 13.2 Humeri of thylacoleonids and comparative marsupial taxa 308 Fig. 13.3 Medial (A) and anterior (B) views of the ulna of Wakaleo hilmeri and selected marsupials 317 Fig. 13.4 Anterior view of the radii of thylacoleonids and selected marsupials 320 Fig. 14.1 Geographic map of the Riversleigh area showing the location of the main tertiary fossil-bearing areas 335 ix

Fig. 14.2 Correlation of Tertiary deposits based on stage of evolution of Wakaleo species, biocorrelation of other marsupial species and superposition, following Archer et al. (1997) 343 Fig. 14.3 Diversity and hypothetical interrelationships of thylacoleonids through time 347

LIST OF TABLES

Table 3.1 Marsupial lion diversity: species, ages and localities 12 Table 4.1 Summary of measurements of humeri and radii of Thylacoleo carnifex from South Australian localities (mainly Naracoorte Caves) 26 Table 4.2 Dimensions of cheekteeth of Wakaleo hilmeri 52 Table 5.1 Comparative measurements of humeral features of Wakaleo hilmeri, Lekaneleo pitikantensis and Thylacoleo carnifex 95 Table 7.1 Cheek dentition measurements of Wakaleo oldfieldi and W. vanderleueri 137 Table 8.1 Upper and lower cheek dentition measurements for Lekaneleo roskellyae 178 Table 11.1 Taxa character state matrix for intrafamilial parsimony analysis 242 Table 12.1 Taxa character state matrix for interfamilial parsimony analysis 281 Table 13.1 Humeral measurements of thylacoleonids and selected comparative marsupial taxa 303 Table 14.1 Distribution of marsupial lion species in Riversleigh sites 339 x

ACKNOWLEDGEMENTS

Many people have assisted me throughout this undertaking. I would like to thank the following people for discussing, reading and constructively criticising various sections of this work: Professor M. Archer, Dr. Suzanne Hand, Karen Black, Vera Weisbecker, Robin Beck, Dr. Dirk Megirian and anonymous reviewers. Dr. Ken Aplin provided assistance with the intricacies of basicranial anatomy.

Many thanks go to Henk Godthelp for sharing his knowledge of and infectious enthusiasm for palaeontology, for his patient guidance and teaching of preparatory and field work skills over many years, and for providing numerous references.

I would like to thank the postgraduate students of the Vertebrate Palaeontology Laboratory, University of New South Wales, for their helpful discussions and assistance with references. In particular, I would like to thank Dr. Rick Arena, Dr. Mina Bassarova, Robin Beck, Dr. Kirsten Crosby, Julien Louys, Jacqueline Nguyen, Karen Roberts, Kenny Travouillon and Vera Weisbecker. I would also like to thank Robin Beck and Peter Gillespie for assistance with PAUP and matters pertaining to cladistics. I thank past postgraduate students of the Riversleigh Research Project, in particular Dr. Bernard Cooke, Dr. Jeanette Muirhead, Dr. Troy Myers, Dr. Paul Willis, Dr. John Scanlon, Dr Steve Wroe and Dr. Walter Boles, for their assistance and good company during many long, happy and memorable hours spent in the field at Riversleigh, and whose work has been a source of inspiration during the course of this research. xi

I am indebted to Dr. Jeanette Muirhead for her illustrations and photographs of the earliest thylacoleonid specimens recovered from Riversleigh, in particular those of the “P. roskellyae” cranium. I thank Jenny Brammall and Karen Black for their advice and assistance regarding the photographing of specimens. I thank Dr. Natalie Warburton for taking measurements of Notoryctes material.

I am extremely grateful to the following people for providing access to comparative material: Jim MacNamara and Neville Pledge from the South Australian Museum; Sandy Ingleby and Tish Ennis from the Mammal Section, the Australian Museum; Robert Jones from the Palaeontology Section, the Australian Museum; Gavin Dally, Dirk Megirian and Peter Murray from the Museum and Art Gallery of the Northern Territory; Joanne Wilkinson from the Queensland Museum; Dermot Henry from the National Museum of Victoria; Craig Reid from the Queen Victoria Museum and Art Gallery, Launceston. I would also like to thank Sandy Ingleby, Tish Ennis, Robert Jones and Jim McNamara, for providing space in their respective departments to allow me to undertake comparative work.

As the senior preparator of the Riversleigh project I have had the privilege (and fun) of preparing many of the specimens I have studied, however, I would also like to thank my co-preparators, Henk Godthelp, Karen Black and Sally Cowan at the University of New South Wales and Benita Johnston at the Riversleigh Fossil Centre, Mt Isa, for their care and skill in preparing the Riversleigh fossils.

The Riversleigh project has been supported by the Australian Research Grants Council, the Department of the Environment, Sport and Territories, National Estate Grants Scheme (Queensland), Queensland National Parks and Wildlife Service, The Australian and Queensland Museums, The Australian Geographic Society, The Linnean Society of New South Wales, the University of New South Wales and Pasminco Pty Ltd. Special thanks must go to the many volunteers, staff and postgraduate students from UNSW who have assisted in the collecting and processing of the Riversleigh material. xii

Very special thanks go to Karen Black for providing encouragement, comradery and constructive comments regarding all things thylacoleonid.

Finally, special thanks go to my family, Michael, Nicola and my mother for their extreme tolerance and patience throughout the completion of this work. This work is dedicated to them and to the memory of my father. PUBLICATIONS

Publications resulting from or contributed to by this research are listed below:

Gillespie, A., 1997. Priscileo roskellyae sp. nov. (Thylacoleonidae, Marsupialia) from the Oligocene-Miocene of Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum. 41 (2): 321-327.

Wroe, S., Myers, T.J., Wells, R.T. and Gillespie, A. 1999. Estimating the weight of the Pleistocene marsupial lion, Thylacoleo carnifex (Thylacoleonidae: Marsupialia): implications for the ecomorphology of a marsupial super-predator and hypotheses of impoverishment of Australian marsupial carnivore faunas. Australian Journal of Zoology. 47: 489-498.

Archer, M., Arena, R., Bassarova, M., Black, K., Brammall, J., Cooke, B., Creaser, P., Crosby, K., Gillespie, A., Godthelp, H., Gott, M., Hand, S.J., Kear, B., Krikmann, A., Mackness, B., Muirhead, J., Musser, A., Myers, T., Pledge, N., Wang, Y., and Wroe S. 1999. The evolutionary history and diversity of Australian mammals. Australian Mammalogy. 21: 1-45. 1

CHAPTER 1

INTRODUCTION

Thylacoleonids, or marsupial lions, are an extinct group of marsupials that inhabited Australia from at least the late Oligocene to the Pleistocene. This group is characterised by a blade-like third premolar and posterior molars that are reduced in size and sometimes in number. The evolution of marsupial lions culminated in Thylacoleo carnifex, a species whose premolar was enlarged into an extraordinary secateur-like cutting blade and whose dentition is arguably the most unique within Mammalia and one that has been described as the most extreme development of carnivory. Very little is known about the ancestors of this species, the pre- Pleistocene fossil record of marsupial lions having been disappointingly depauperate. An overview of the diversity of marsupial lions known at the commencement of this study is presented in Chapter 3.

The Oligocene-Miocene freshwater limestone deposits of the Riversleigh World Heritage Fossil Locality in north-western Queensland have yielded numerous fossils of marsupial lions. To date, thylacoleonid material has been recovered from 18 sites at Riversleigh. This material includes complete skulls, mandibles and postcranial elements and vastly improves the fossil record of thylacoleonids for this period of geological time. These fossils indicate a greater diversity for this marsupial family at this time than was previously thought and also allow new insights into their origins.

The first aim of this research was to document the diversity of marsupial lions within these deposits by identifying and describing the species present. Descriptions of new genera and species, and of significant new material of currently known species, are undertaken in Chapters 4 to 10. This work has also required re-evaluations of previously described species and has resulted in amendments to the systematics of some taxa. 2

An area of controversy that surrounds thylacoleonids is that of their ancestry. Currently, there are conflicting hypotheses regarding their affinities within Diprotodontia. The historic view is that they descended from a phalangeroid ancestor, although, more recent studies suggest they are a plesiomorphic member of the vombatiform marsupials. The second aim of this study was to investigate the interrelationships of marsupial lions within the family (Chapter 11) and, more significantly, to obtain a clearer understanding of their relationships within Diprotodontia (Chapter12) by incorporating the new taxa and morphological data provided by the Riversleigh fossils into phylogenetic analyses. These analyses were carried out using the principles of cladistics which are briefly discussed in the methods section (Chapter 2).

Previously, studies of the functional morphology of marsupial lions have been restricted to two species, Thylacoleo carnifex and Priscileo pitikantensis. The former is interpreted to have been a cursorial species that may have had the ability to leap (Finch and Freedman 1986, 1988) and assessment of the limited remains of the latter species suggest that it may have been scansorial or fossorial (Rauscher 1983). The third aim of this study was to assess the functional morphology of the postcranial elements recovered from Riversleigh by comparison with the skeletal morphologies of marsupials of known locomotor categories (Chapter 13). Postcranial elements of P. pitikantensis were included in this analysis because of their similarities to the Riversleigh material and to test previous locomotory hypotheses that have been proposed for that species.

The fourth aim of this study was to use the species identified at Riversleigh as biocorrelative tools to help correlate sites within the Riversleigh WHA and on a broader intracontinental level (Chapter 14). Using these correlations in conjunction with an improved knowledge of the phyletic evolution of marsupial lions, previous hypotheses of the ages of Riversleigh sites and other Australian Oligo-Miocene vertebrate localities are assessed. A further aim of this study was to assess the diversity of thylacoleonids in relation to the palaeoenvironments proposed to have occurred during this period of geological time. 3

CHAPTER 2

METHODS and MATERIALS

2.1 Riversleigh Fossils

The fossils that are the focus of this study have been recovered from sites within the Riversleigh World Heritage Area which is located approximately 5 km west of the Riversleigh homestead (19o02’ S, 138o45’ E) in the southern part of the Lawn Hill National Park, northwestern Queensland. Ongoing collection of fossils from this locality has been carried out by Professor Michael Archer and colleagues from the University of New South Wales since 1981. A history of the discoveries and description of the field work undertaken at Riversleigh is provided in Archer et al. (1991). The vast majority of the fossiliferous limestone blocks that have been collected have been sent to the Vertebrate Palaeontology Laboratory at the University of New South Wales for processing. Supplementary processing of this material has also been carried out in Mt Isa by Alan Rackham Snr in the 1980s and, since 2000, by the Riversleigh Fossil Centre. Fossils in the limestone are recovered by acid-etching, a process that involves dissolution of the limestone in a weak concentration of acetic acid. Fragile fossil specimens are strengthened by consolidation with a liquid resin (Bedacryl) that is thinned with acetone. Once free of the limestone, the fossils are sorted and registered (signified by the prefix QM F or a temporary AR prefix, see below) prior to being studied. Following completion of this research, the Riversleigh fossil material is housed in the palaeontological collection of the Queensland Museum, Brisbane.

Much of this research has required the comparative study of other fossil specimens as well as osteological material of extant marsupial species. Comparative specimens of thylacoleonids and other extinct marsupial families were made available by the Australian Museum, South Australian Museum, Queensland Museum and the 4

Museum of Arts and Sciences of the Northern Territory. Comparative osteological specimens of extant species was made available by the Mammal Section of Australian Museum, the School of Biological, Earth and Environmental Sciences at the University of New South Wales and The Queen Victoria Museum and Art Gallery of Launceston.

2.2 Terminology

Abbreviations As indicated above, the Riversleigh fossils are registered as part of the palaeontological collection of the Queensland Museum and are indicated by the prefix QM F. Initially, specimens processed at the University of New South Wales were recorded with an AR prefix, indicating the temporary collection of Professor Michael Archer, University of New South Wales. AR numbers were later replaced by a QM F number. Other institutional abbreviations used within this work are as follows: AM MF or AM F = palaeontological collection of the Australian Museum, Sydney; AM M or AM S = mammal collection of the Australian Museum, Sydney; CPC = Commonwealth palaeontological collection, Canberra; NTM P = palaeontological collection of the Museum and Art Gallery of the Northern Territory, Darwin; NMV P = palaeontological collection of the National Museum of Victoria, Melbourne; SAM P = palaeontological collection of the South Australian Museum, Adelaide; UCMP = palaeontological collection of the University of California, Berkeley; UNSWZ = teaching collection of the School of Biological, Earth and Environmental Sciences, University of New South Wales. QVM = The Queen Victoria Museum and Art Gallery, Launceston.

Riversleigh Site Names Approximately 300 fossil sites have been identified on the limestone plateaux at Riversleigh. These sites are usually isolated areas of bone-bearing limestone and are designated with a particular name by their discoverer. In some cases, if the bone- bearing area extends over a relatively large area, smaller bony areas within that site are numbered (for example, COA 1, COA 2, COA 3). This numbering has been 5 employed to safe-guard against the possibility that the area may actually represent a number of different fossil sites that differ in age or faunal content.

Nomenclature Numbering of the teeth is topological and does not imply serial homology. Thus P1-3 is used for the premolars and the remaining molars are M1-4. Within the text, the position of a tooth within the tooth-row is indicated by the letters I (incisor), C (canine), P (premolar) or M (molar). Teeth of the upper dentition are indicated by a superscript number (e.g. P3) that also indicates the position of the tooth in the tooth row, or a forward stroke following the numeral (eg. P3/). Lower teeth are indicated by a subscript number (e.g. P3), or by a forward stroke separating the letter from the number (e.g. P/3). Reference to both upper and lower teeth is indicated if the number is neither superscript nor subscript (e.g. P3).

Anatomical terms used, in particular those for the postcranial bones and musculature, is the anglicised equivalent of the Latin term from the Nomina Anatomica Veterinaria (1992). Anatomical nomenclature used for the basicranium is based on Archer (1976a) and Aplin (1987). Terminology for the periotic is based on Norris (1994) and Crosby and Norris (2003).

Measurements Dental and osteological measurements were made in millimetres using PAV digital callipers accurate to 0.01mm and rounded to the nearest 0.1mm. Because of their large size, measurements of some limb bone lengths were made to the nearest millimetre with a millimetre ruler. All measurements were made at least three times.

2.3 Systematics

The classification proposed by Aplin and Archer (1987) is followed in this study. In this classification, the Order Diprotodontia is divided into two suborders, Vombatiformes and Phalangerida. Vombatiformes is divided into two Infraorders, Phascolarctomorphia and Vombatomorphia. Phascolarctomorphia contains the 6 family Phascolarctidae and Vombatomorphia contains the living family, Vombatidae, as well as a diverse array of extinct families that includes the Diprotodontidae, Palorchestidae, Wynyardiidae, Ilariidae, Maradidae and Thylacoleonidae. The Suborder Phalangerida contains the “possum” families (Phalangeridae, Miralinidae, Ektopodontidae, Burramyidae, Pseudocheiridae, Petauridae, Tarsipedidae, Acrobatidae and Pilkipildridae) and the macropodoid families (Macropodidae and Potoroidae). This classification system is very similar to that proposed by Woodburne (1984) but differs in its introduction of an additional hierarchical rank in the Vombatiformes and replaces Woodburne’s “Phalangeriformes” with the new taxonomic name, Phalangerida, because of taxonomic synonymy of the former name (Aplin and Archer 1987).

2.4 Cladistic Methodology

Over the last thirty years the use of cladistic analysis has overtaken evolutionary systematics and phenetics as the preferred method for investigating the evolutionary relationships of species. This method, originally proposed by Hennig (1966), is based upon the identification of shared, derived character states amongst the groups of taxa under study. This methodology aims at identifying monophyletic groups, i.e. a group of species that includes an ancestral species and all of its descendants. Evolutionary systematics has fallen into disfavour because it lacks a consistent and definable methodology (Smith 1994). Its methodology generally relies on the knowledge and intuition of the investigator to determine the ‘importance’ of features. Such features may be unique to the taxon and thus provide no information regarding that species’ relationship to other taxa. Consequently, the taxonomic groups produced by evolutionary systematics may be artificial and not reflect phylogeny. Phenetic analysis focuses upon the overall similarity of an organism, however no attempt is made to determine whether some similarities are the result of convergence or parallelism. Hence, this approach will not necessarily reflect the phylogenetic relationships of the taxa under investigation. 7

Cladistics has been broadly embraced by researchers investigating systematics. There are now a number of journals and an increasing literature devoted to exploring and refining aspects of this methodology (e.g. Cladistics and Systematic Zoology).

The first step in a cladistic analysis involves the identification of independent, intrinsic characters in the taxa that are the focus of the study (ingroup) and the determination of the different states (or transformations) of these characters. The shared derived character states, or synapomorphies, critical to cladistic analysis are usually determined by comparing the character states of the ingroup with those of an ‘outgroup’, ideally the closest relative or sister group of the ingroup. “For a given character with 2 or more states within a group, the state occurring in related groups (i.e. sister group) is assumed to be the plesiomorphic (primitive) state” (Watrous and Wheeler 1981). The determination of whether a character state is plesiomorphic or apomorphic (derived) is also termed character polarisation. The ontogenetic history of a character may also be used to polarise character states, however, the paucity of embryological information regarding most species severely limits the use of this criterion at present.

The major challenge of cladistic methodology (as with both phenetics and evolutionary systematics) is the difficulty in determining whether character state similarities are the result of common ancestry or the result of homoplasy (convergence or reversal). The main approach to solving the problem of homoplasy is through the use of the principle of parsimony which reasons “that given several solutions to a problem, the most economical should be accepted” (Scotland 1992). Using this approach results in a search for the phylogenetic tree with the least number of steps or character state changes and thus, theoretically, the least amount of homoplasy. This tree is referred to as the optimal tree (Wiley et al. 1991).

A number of different parsimony criteria can be used in constructing phylogenetic trees. The two main types of criteria are Wagner parsimony and Fitch parsimony. In Wagner parsimony changes in character states occur in an ordered sequence in a transformation series. Thus, for a character to change from state 0 to state 2, it must 8 do so by going through state 1. Character states are also allowed to reverse freely. In Fitch parsimony character states in a transformation series are treated as unordered so that a character state is free to transform into any other state without requiring to pass through any intermediate states. For example, a character state may transform directly from state 0 to state 2 without going through state 1.

When a number of parsimonious trees are produced by a parsimony analysis, a single consensus tree is generated from non-contradictory components of these trees. However, because they are constructed from sets of cladograms, the consensus tree may not be “supported as most parsimonious by the original data and in some cases may even contain components not found in any of the fundamental cladograms” (Kitching et al. 1998, p.140). A number of different consensus methods may be employed to analyse sets of cladograms. Three of the methods used in this study are strict, majority-rule and Adam’s consensus methods.

Strict and majority-rule consensus methods use the frequency of informative groups within a set of most-parsimonious cladograms (Kitching et al. 1998). Strict consensus trees contain only those components, or monophyletic groups, that are common to all competing cladograms. Majority-rule consensus trees contain all groupings that are found in more than 50% of the competing most-parsimonious cladograms. This consensus method is frequently used when many trees are being compared (Kitching et al. 1998). Adam’s consensus trees aim to improve the “resolution” between conflicting trees. This is done by relocating those taxa that are found in varying positions on competing cladograms to the nearest node they have in common. A potential problem with Adam’s consensus trees is that they may produce groupings that do not occur in any of the original parsimony trees.

A number of different indices have been formulated to measure the amount of homoplasy in a data matrix. The indices used in this study include the ensemble consistency index (CI), the rescaled consistency index (RC) and the retention index (RI). 9

The ensemble consistency index provides an indication of how well the entire data set fits a particular cladogram (Wiley et al. 1991). A high CI value indicates that the distribution of character states fits the tree well and that there is little homoplasy. A low CI value indicates a high level of homoplasy. A number of problems have been noted regarding the use of the ensemble consistency index. Firstly, if autapomorphies or uninformative characters are included, this results in an inflated CI value. Secondly, increasing the number of taxa results in a decrease in the CI value. The retention index (RI) has been introduced to combat some of the problems associated with the CI. The retention index measures the proportion of terminal taxa that retain the character state identified as a synapomorphy for that group (Smith 1994). The rescaled consistency index (RC) is the product of the consistency index and the retention index. This index excludes characters that do not contribute to the “fit” of the tree. It is considered a superior index because it excludes autapomorphies and totally homoplastic characters, preventing them from artificially inflating the measure of fit (Wiley et al. 1991).

A number of phylogenetic computer programmes and packages are available for parsimony analysis. In this study the computer package PAUP, Phylogenetic Analysis Using Parsimony, version 4.0b (Swofford 1998) is used for both intrafamilial and interfamilial analyses (see Chapter 11 and 12).

Further details of the methods used in the phylogenetic analyses and functional morphological studies are described in the relevant chapters. 10

CHAPTER 3

THYLACOLEONID DIVERSITY

3.1 Introduction

Prior to the recovery of thylacoleonid fossils from the Riversleigh World Heritage Fossil Area, this family was represented by three genera and a total of seven species. A list of these species, their ages, and the localities at which they have been found is shown in Table 1. Although thylacoleonids have a fossil history that stretches from the late Oligocene through to the Pleistocene, a period spanning approximately 25 million years, their diversity during this time has never been high. Throughout this interval, no more than two species of marsupial lion have been known to co-exist at any one time. The late Oligocene taxon, Priscileo pitikantensis from central Australia, may have overlapped with Wakaleo oldfieldi. The middle and later stages of the Miocene epoch are characterised by two apparently temporally separate species of Wakaleo, W. vanderleueri and W. alcootaensis. Two species of the genus Thylacoleo are evident in the Pliocene; the medium sized T. crassidentatus and the smaller T. hilli. These species were possibly sympatric (Archer and Dawson 1982). A single, large species, Thylacoleo carnifex, was present in the Pleistocene.

3.2 Priscileo pitikantensis Rauscher, 1987

This taxon represents the oldest known thylacoleonid and was recovered from the late Oligocene Ngapakaldi Local Fauna, Etadunna Formation, at the southwestern end of Lake Pitikanta, South Australia (Rauscher 1987). The species is poorly known because of the fragmented and limited nature of the fossil material. It is represented by right and left maxillary fragments that contain only a single tooth - a portion of the crown of the left M2 (Rauscher 1987). Postcranial material of this species was also recovered with dental material and a partial right jugal, and includes a partial scapula, a humerus, a radius, a magnum and an ectocuneiform bone. Studies of elements of 11

Figure 3.1 Distribution of fossil sites yielding thylacoleonid species

3

5

4 9 13

27 2 6 12 1 10/ 14 11 17 36 16 29 35 7 28 15 32 20 18 34 8 33 19 31 26 24 22 25 23 30 21

37

38 Priscileo pitikantensis Wakaleo oldfieldi Wakaleo vanderleueri Wakaleo alcootaensis Thylacoleo hilli Thylacoleo crassidentatus Thylacoleo carnifex

Riversleigh World Heritage Fossil Area

Numbers refer to localities listed in Table 3.1 12 the forelimb suggest that this small, cat-sized species may have been arboreal (Rauscher 1983; Murray and Megirian 1990).

This species is also regarded as the most plesiomorphic thylacoleonid taxon, possessing a primitive upper dentition which includes all three premolars as well as four molars. In addition, the dentition is absolutely smaller in dimensions than those of all other marsupial lions (Rauscher 1987).

Table 3.1: Marsupial Lion Diversity: species, ages and localities. ______SPECIES AGE LOCATION

Number indicates locality on map in Fig. 3.1 ______

Priscileo pitikantensis late Oligocene (26-24 Ma)1 1Lake Pitikanta, S.A.

Wakaleo oldfieldi late Oligocene (24 Ma)1 2Leaf Locality, Lake Ngapakaldi, S.A.

Wakaleo vanderleueri mid-Miocene (12 Ma) 3Small Hills Locality, Bullock Creek, N.T.

Wakaleo alcootaensis late Miocene (8 Ma) 4Paine Quarry, Alcoota Stn., N.T.

Thylacoleo crassidentatus Pliocene (5.3-3.4 Ma) 5Bluff Downs, NE Qld; 6Chinchilla, SE Qld; 7Bow, N.S.W.

Thylacoleo hilli Pliocene (5.3 Ma) 8Curramulka, S.A.; 7Bow, N.S.W.; (including T. sp. cf. T. hilli) 9Mt. Etna, QLD.

Thylacoleo carnifex* Pleistocene (2 Ma- Holocene) 10Darling Downs,11 King River, 12Dalby,13 Spring Creek, 14Gore, Qld. 15Wellington Caves, 16Moree,17Bingara, 18Geurie, 19Wombeyan Caves, 20Lake Victoria, N.S.W. 21Lake Colongulac, 22Lake Colac, 23Buchan Caves, 24Strathdownie, 25DuckPonds, Vic. 26NaracoorteCaves, 8Curramulka, 27Kalamurina, 28Port Pirie, 29Buckalowie Caves, 30Mt.Gambier, 31Normanville, 32Bundey, 33Lake Fowler, S.A. 34Mammoth Caves, 35Balladonia Soak, 36Madura Cave, W.A. 37Scotchtown Cave, 38Montague, Tas. ______Ma = millions of years ago 1. Woodburne et al. (1993) * locality data from Pledge (1977) and Finch (1981) 13

3.3 Wakaleo oldfieldi Clemens and Plane, 1974

This species is the oldest of the Wakaleo lineage. It was collected from the late Oligocene, Kutjamarpu Local Fauna in the Wipajiri Formation at Lake Ngapakaldi in South Australia (Clemens and Plane 1974). This dog-sized marsupial lion is known from few specimens. The holotype is a single dentary containing an incisor, P3, M1, and alveoli for an M2, M3, and a single-rooted tooth between the incisor and P3 (Clemens and Plane 1974). The only other described specimens are two isolated lower teeth. No upper teeth or postcranial bones are known for this species.

3.4 Wakaleo vanderleueri Clemens and Plane, 1974

This species was first recovered in 1967 from the Camfield Beds, Bullock Creek Local Fauna, northcentral Northern Territory and was estimated to be mid Miocene in age (Clemens and Plane 1974). The holotype specimen is a right dentary fragment containing M1, alveoli for M2-3, and partial alveolus for I1 (Clemens and Plane 1974). It was distinguished from that species on the basis of its slightly larger size and the greater prominence of ridges descending from the apical cusp on M1.

More material of this species has since been recovered from the Bullock Creek LF and includes four dentaries, a maxillary fragment, a complete cranium, and a radius fragment (Megirian 1986; Murray, Wells and Plane 1987; Murray and Megirian 1990).Comparative study of the dentaries with that of W. oldfieldi indicates that the two species show little distinct morphological variation and it has been hypothesised that they represent a morphocline, with W. vanderleueri representing the slightly larger and younger extreme (Murray and Megirian 1990).

The proportions of the skull indicate that W. vanderleueri was a moderate to large dog-sized predator and/or scavenger that inhabited the open woodland or tree savannah of mid-Miocene Australia (Murray and Megirian 1992). Studies of the radius fragment suggest that this species may have been more terrestrial than Priscileo pitikantensis (Murray and Megirian 1990). 3.5 Wakaleo alcootaensis Archer and Rich, 1982 14

Wakaleo alcootaensis is the largest and youngest species within Wakaleo. It is known from a single left maxillary fragment that was recovered from the Waite Formation, Alcoota Local Fauna, Northern Territory and is estimated to be 8 million years old (Archer and Rich 1982). The specimen contains a large, blade-like P3, a large, broken, triangular M1, and alveoli (with roots) for M2. Although the region anterior to P3 is preserved, there is no evidence of any alveoli for the anterior premolars, however, the species may have retained M3 (Archer and Rich 1982). This species was much larger than other Wakaleo species, its P3 being twice the length of that of W. oldfieldi. On the basis of this specimen and because of the lack of any known upper teeth for W. oldfieldi and W. vanderleueri, Archer and Rich (1982) proposed the absence of both upper and lower anterior premolars (P1-2) as synapomorphic for the genus.

3.6 Thylacoleo crassidentatus Bartholomai, 1962

In 1956 Jack Woods reviewed the specimens referred to Thylacoleo Owen, 1859 in the collections of the Queensland Museum and noted that specimens from the Chinchilla district were possibly distinct from those of T. carnifex. The recovery of a further right dentary from this Pliocene locality supported Woods' earlier speculation and Alan Bartholomai named the new species T. crassidentatus (Bartholomai 1962). The dentary of the new species differed from T. carnifex in having a much broader posterior end on P3, a much stouter M1, and more robust M2. Further material referable to this species has also been collected from the Bluff Downs Local Fauna from the Allingham Formation of north Queensland and from the Bow Local Fauna in northern New South Wales, both Pliocene in age (Archer and Wade 1976; Archer and Dawson 1982). These additional specimens have allowed a more complete description of the species and brought to light further differences from T. carnifex (Archer and Dawson 1982). The most significant specimen was a left skull fragment from the Bluff Downs Local Fauna which preserved C1-M1 and alveoli for I1-3 and M2 which indicated that T. crassidentatus possessed a molar (M2) that is lost in T. carnifex, and that the first upper molar was proportionately longer compared with P3. 15

3.7 Thylacoleo hilli Pledge, 1977

This species is the most poorly known within Thylacoleo and was named on the basis of a single upper P3 collected from Town Cave at Curramulka on the Yorke Peninsula in South Australia (Pledge 1977). The major feature of the tooth which distinguishes it from other species of Thylacoleo is its small size, being half the length of the equivalent tooth of T. carnifex and three-fifths the length of this tooth in T. crassidentatus. Unfortunately, the type locality could not be dated, although Pledge (1977) suggested a late Miocene or early Pliocene age. Support for a Pliocene age has been provided by the discovery of a right dentary specimen containing I1 and P3 referable to this species (Thylacoleo sp. cf. T. hilli) from the Pliocene Bow Local Fauna near Bow in New South Wales (Archer and Dawson 1982). The presence of this species in the Bow Local Fauna is significant because it is the only known fauna to contain apparently sympatric species of marsupial lion, i.e. T. crassidentatus and T. sp. cf. T. hilli. Further material of T. hilli has recently been recovered from the limestone caves of Mt. Etna in Queensland (Hocknull and Zhao 2005).

3.8 Thylacoleo carnifex Owen, 1858

Thylacoleo carnifex is the largest species of marsupial lion, approaching the size of a lion, and is believed to have been a medium-paced cursorial predator capable of grasping prey with its forelimbs (Finch and Freedman 1988). It is the only thylacoleonid found in Pleistocene deposits. As well as being the first known, it has the most complete fossil record and has been the most studied species of marsupial lion. Fossil specimens of the species were first collected from the Wellington Valley in New South Wales in the early 1830s by Major Thomas Mitchell. Sir Richard Owen, the renowned English palaeontologist, on receiving fragmentary remains that included the trenchant premolar, was able to determine the carnivorous function of the teeth before he was able to determine the identity or affinities of their possessor. In 1846 Owen was sent the remains of a skull from Lake Colongulac in Victoria and on the basis of this specimen and a mandible from King Creek, Darling Downs, Queensland, the species was first named in the Encyclopaedia Britannica (1858) and 16 described in detail in 1859 in the Philosophical Transactions of the Royal Society, London. Comparative studies with other carnivorous marsupials and placentals enabled Owen to determine the marsupial affinities of the species (Owen 1866). During the late 1800s a number of other thylacoleonid specimens were designated as either new species of Thylacoleo or as new genera by a number of researchers including Stutchbury (Schizodon 1853), Krefft (Plectodon; Mylodon (?) australis 1870, Thylacoleo robustus 1872), McCoy (Thylacoleo oweni 1876), De Vis (Prochaerus celer 1886) and Owen (Thylacopardus australis 1888). Archer and Dawson (1982) reviewed the holotypes of these species and their names and concluded that Thylacoleo is the senior synonym of Plectodon, Schizodon and Thylacopardus and Thylacoleo carnifex is a senior synonym of Mylodon australis, Thylacopardus australis, Thylacoleo oweni and Thylacoleo robustus.

The dental formula for T. carnifex is I1-3/1, C1/0, P1-3/1-3, M1/1-2. The unique feature of this dentition is the extreme elongation of the third premolar to form an extraordinary cutting blade. Owen (1859) stated that the tooth signified the most extreme development of carnivory and that the animal was "one of the fellest and most destructive of predatory beasts". Owen's assessment of the unique of dentition of the species sparked controversy. Flower (1868) vigorously argued for an herbivorous habit and the polemic was pursued for almost one hundred years. Relatively recent studies of the wear patterns on the teeth, and on masticatory mechanisms, support Owen's original carnivorous conclusion (Finch 1982; Wells, Horton and Rogers 1982).

Specimens of Thylacoleo carnifex have been recovered from Pleistocene deposits all around Australia (Table 1). The most prolific sites have been Wellington Caves in New South Wales and Victoria Cave at Naracoorte in South Australia. The most complete specimen of the species was recovered from Marshall’s Ponds Creek near Moree in New South Wales in 1966. This locality yielded an almost complete skeleton of an adult, a skull of a juvenile and the crushed skull of a pouch-young. Eileen Finch (1982) described these specimens and studied the skeleton to assess the functional morphology, diet, and locomotion of the species (Finch 1982; Finch and Freedman 1982, 1986, 1988). An articulated left and right manus and partial left pes 17 have also been described from Victoria Cave in Naracoorte, South Australia (Wells and Nichol 1977). 18

CHAPTER 4

REASSESSMENT OF Priscileo pitikantensis Rauscher, 1987 AND THE IDENTIFICATION OF A PLESIOMORPHIC SPECIES OF Wakaleo FROM RIVERSLEIGH

4.1 Introduction

The genus Wakaleo Clemens and Plane, 1974 is comprised of species of medium- sized marsupial lions that exhibit loss of the anterior premolar/s and posterior molar/s. Fossils of this genus are not common. Wakaleo oldfieldi Clemens and Plane, 1974 is known from a single dentary and an isolated tooth collected from the Kutjamarpu Local Fauna in South Australia and W. alcootaensis Archer and Rich, 1982 from a single maxillary fragment recovered from the Alcoota Local Fauna in the Northern Territory. Wakaleo vanderleueri Clemens and Plane, 1974 has a better record, being known from a skull, a number of dentaries and maxillary fragments recovered from the Bullock Creek Local Fauna in the Northern Territory (Clemens and Plane 1974; Murray, Wells and Plane 1987; Megirian 1986; Murray and Megirian 1990).

Fossil material of a new, smaller species of Wakaleo has been recovered from the Riversleigh World Heritage Area, Lawn Hill National Park northwestern Queensland. This material includes a beautifully preserved, near-complete cranium from Hiatus Site, a palate and associated postcranial bones from the Rackham's Low Lion Site and a number of dentaries from White Hunter Site. These fossils provide significant new information regarding the dental morphology of the genus especially in relation to the genus Priscileo. Previously, the major features used to distinguish species of Wakaleo from those of Priscileo have been size and the loss of anterior premolars and posterior molars (Rauscher 1987, Murray et al. 1987). The Riversleigh taxon shares many dental and cranial features with W. vanderleueri that strongly support its referral to the genus Wakaleo but it also exhibits a number of 19 similarities with Priscileo pitikantensis, the most significant being the primary diagnostic feature of the genus Priscileo, i.e. presence of three premolars and four molars. These similarities suggest the possibility that they may represent the same taxon and has prompted a re-evaluation of the holotype of P. pitikantensis (SAM P37719). A review of this specimen involving comparisons with species of Wakaleo and Priscileo roskellyae is presented in this chapter and is followed by a description of the plesiomorphic Wakaleo taxon from Riversleigh. The diagnosis of the genus Wakaleo is amended.

4.2 Re-assessment of Priscileo pitikantensis Rauscher, 1987

In 1961, the remains of a small marsupial lion were collected from the Etadunna Formation at Lake Pitikanta, South Australia by R.A. Stirton and R.H. Tedford of the University of California (Berkeley). These fossils included a fragmented palate containing a single molar (M2) and various postcranial elements. Unlike the species of thylacoleonids (species of Wakaleo and Thylacoleo) known at that time, this specimen was small and alveoli indicated it possessed a fourth molar. In contrast to species of Wakaleo, this specimen also possessed two anterior premolars. Based on these differences, Rauscher (1987) placed the species in a new genus and named it Priscileo pitikantensis. Specimens of a plesiomorphic Wakaleo taxon (QM F45200, QM F23443 and QM F24680) recovered from sites at Riversleigh also have this dental formula.

Comparisons between the holotype of P. pitikantensis (SAM P37719) and the Riversleigh material are limited because of the incompleteness of the former specimen. The only tooth that is present in SAM P37719 is M2, which preserves only the protocone and a crenulated trigon basin. The buccal flank of the crown is missing and slivers of enamel are also lost along the lingual margin (see Fig. 4.1). The remains of this crown show more similarities to the M2 of species of Wakaleo than to that of P. roskellyae (QM F23453, QM F40116 and QM F23442). These similarities include: (1) the triangular outline of the trigon basin. This shape results primarily from the postprotocrista being directed buccally rather than posteriorly. In contrast, in 20

Fig. 4.1: Priscileo pitikantensis, SAM P37719 holotype, left maxilla with damaged M2; Lake Pitikanta.

M M2 P3 M3 J 4 P2 B M

C D E F 1 H C A G I

Specimen A B C FED G H JI No. SAM P37719 9.5 12.4 5.3 4.8 3.6 3.0 38.818.821.4 5.2

QM F45200 13.112.1 6.0 5.5 4.9 4.8 43.522.523.0 5.9

QM F23443 13.712.7 6.8 5.2 5.0 3.8 46.121.925.7 5.6

QM F24680 13.410.7 6.9 6.2 4.3 4.9 42.722.023.0 6.2

Fig. 4.2 : Comparative measurements of maxillae alveoli and M2 s of SAM P37719 and Riversleigh specimens (QM F).

Drawing modified from Rauscher (1987; holotype of P. pitikantensis SAM P37719) 21

P. roskellyae the trigon basin extends posteriorly toward the metaconule (postprotocrista more posteriorly-directed) resulting in a squarer basin shape; (2) the lingual margin of the crown is similar in shape and robustness. The crown supporting the protocone dominates the lingual margin being relatively large and bulbous, relatively tall, thick and broadly-rounded lingually. The crown below the metaconule is relatively reduced. The combination of these features gives the tooth a more triangular outline. In contrast, in P. roskellyae the crown supporting the protocone is less bulbous and does not dominate the lingual margin and its outline is less triangular. In this species the lingual edge is straighter (or squared off) and shorter, and the protocone has a relatively more anterior position. The proportion of the lingual side of the crown below the metaconule is much greater than in species of Wakaleo; (3) the width of the trigon basin relative to the width of the tooth. The buccal flank of the trigon basin in P. pitikantensis appears to be relatively steep and suggests that the fractured edge of its basin is very close to the longitudinal crest that would have marked its buccal margin. Although the buccal flank of the crown is missing, the distance between the broken edge of the trigon basin and the buccal edge of roots indicates the buccal flank (stylar shelf) was relatively broad. Species of Wakaleo have a moderately broad stylar shelf unlike P. roskellyae in which the trigon basin extends nearly all the way to the buccal margin of the crown; (4) the postprotocrista is positioned more buccally due to the bulbous lingual margin of the crown whereas in P. roskellyae the postprotocrista and premetaconulecrista lie closer to the lingual margin; (5) the absence of a distinct cusp in the metaconule position is a feature characteristic of species of Wakaleo in contrast to the cuspate condition in P. roskellyae.

Comparative dental measurements of P. pitikantensis and the plesiomorphic species of Wakaleo are limited to M2 length and alveolar dimensions of the upper cheek teeth and indicate that P. pitikantensis is slightly smaller than Riversleigh specimens in all measurements (see Fig. 4.2). Comparison of the average length of P3 of P. roskellyae (QM F23453, QM F23442, QM F40116, see Appendix 1: Table B) with that of SAM P37719 indicate that the latter is much closer in size to species of Wakaleo. 22

On the basis of the above mentioned similarities in M2 morphology, dental formula (Wakaleo sp. nov.) and overall size, the holotype of P. pitikantensis is hereby referred to the genus Wakaleo. Further support for its referral to Wakaleo is provided by cladistic analyses of the interrelationships of marsupial lions carried out in this research (see Chapter 11). In both unordered and ordered parsimony analyses SAM P37719 consistently paired with the plesiomorphic Wakaleo taxon from Riversleigh and failed to form a monophyletic clade with other species of Priscileo. The generic name Priscileo Rauscher, 1987 is therefore regarded as a junior synonym of Wakaleo Clemens and Plane, 1974. The only feature SAM P37719 shares with P. roskellyae is its dental formula which is a symplesiomorphy for thylacoleonids. Priscileo roskellyae demonstrates generic level differences in its cranial and dental morphology to species of Wakaleo and Thylacoleo and renaming and further description of this species is undertaken in Chapter 8. Further reference to roskellyae in this chapter is made using its new generic name, Lekaneleo.

4.3 Systematics

SYSTEMATICS

Superorder MARSUPIALIA Illiger, 1811 Order DIPROTODONTIA Owen, 1866 Family THYLACOLEONIDAE Gill, 1872

Wakaleo Clemens and Plane, 1974

Type species: Wakaleo oldfieldi Clemens and Plane, 1974

Other species: Wakaleo vanderleueri Clemens and Plane, 1974 Wakaleo alcootaensis Archer and Rich, 1982 Wakaleo hilmeri sp. nov.

Revised Generic Diagnosis: Species of Wakaleo are distinguished from those of

Thylacoleo in: retaining M3-4/3-4; P3/3 being proportionately shorter relative to 23

M1/1; P3 having a prominent posterior cusp that lies distinctly anterior to the end of the crown in; P3 having a short and strongly-developed anterolingual crest or cusp (rather than a long vertical crest); having an M1 that is triangular/subtriangular and that lacks a neoanterobuccal cusp (sensu Archer and Rich 1982); having an upper molar row that is not strongly medially-inflected; possessing a diastema between C1 and P1; having a relatively well-developed posterobuccal crest on P3; P3 having a moderately developed anterobuccal crest that is connected to the posterobuccal crest by a distinct valley; having the frontal fail to contact the squamosal posteriorly; having a tympanic wing consisting of the alisphenoid and squamosal; having a relatively narrow postglenoid process that does not curl posteriorly; having a proportionally shorter frontal and longer parietal and in lacking a postorbital bar.

Species of Wakaleo differ from species of Lekaneleo (Chapter 8) in being larger and more robust, in the presence of P1 and P2 being variable and in having: narrower and deeper talonid basins on M1 and M2; no more than two single-rooted teeth between 2 I1 and P3; a triangular or sub-triangular trigon basin on M ; a relatively tall buccal margin relative to the protocone on M2; a cranium with tall sagittal and nuchal crests; the dorsal surface of the neurocranium predominantly concave; the alisphenoid in the region of the attachment of the internal pterygoid muscle predominantly concave; a postglenoid cavity that does not extend into the postglenoid process; a prominent rostral tympanic process (sensu Crosby and Norris 2003) on the periotic; the entire anterior wall of the mastoid process formed by the squamosal and through which runs the stylomastoid canal; a relatively narrower outer ear canal; supraorbital processes that extend beyond the lateral edge of the infraorbital margin of the maxilla; more robust molars; teeth with greatly exposed roots; canine alveoli that reach or extend beyond the posterior margin of the anterior palatal fenestra. An additional possible feature distinguishing Wakaleo species from Lekaneleo species is the presence of a transverse canal. Species of Wakaleo differ from the monotypic genus Microleo in being significantly larger and in having: posteriorly broader P3 with lingually bowed longitudinal blade; P3 (in profile) with a straight or gently-curving, un-notched blade; the posterior cusp 24 of P3 located at the posterior margin of the crown; a distinct anterolingual crest on

P3; and the paracone on M2-3 taller than the protocone.

Wakaleo pitikantensis new combination

Holotype: SAM P37719, a left maxillary fragment with partial alveoli for C-P3, broken M1-2, alveoli for M 3-4, and two right maxilla fragments with partial alveoli for C-P3, roots of M1-2, alveoli for M3-4.

Referred material: SAM P37720, fragments of right nasal and jugal, proximal right scapula lacking spine, left humerus, right radius, distal left radius, four rib fragments, left magnum, left ectocuneiform, phalanx. UCMP 88457, crushed right humerus.

Type Locality: Lake Pitikanta (UCMP locality V-5857), Etadunna Formation, Tirari Desert, South Australia.

Diagnosis: Differs from all other species of Wakaleo in being smaller. Differs from 1-3 4 all other species of Wakaleo, except W. hilmeri sp. nov., in retaining P and M /4.

Description: The holotype and referred material were originally described as Priscileo pitikantensis by Rauscher (1987; pp. 423-32, figs. 1-9). Amendment to the description of the scapula of SAM P37720 is given in Chapter 5 (5.7).

4.4 Specific attribution of a plesiomorphic Wakaleo species from Riversleigh

In view of the similarities in M2 morphology, dental formula and size between the Lake Pitikanta Wakaleo and the plesiomorphic Wakaleo specimens from Riversleigh, consideration was given to the possibility that they may represent one species. Comparisons were made of both the dental and postcranial material of Wakaleo pitikantensis and the Riversleigh taxon. Comparative measurements of the dentitions of these specimens were limited to M2 length (midcrown) and alveolar dimensions 25 of the upper cheek teeth (see Fig. 4.2) and indicate that SAM P37719 is smaller (approximately 10%) than Riversleigh specimens. The greatest difference was recorded for the length of the maxilla between the canine and P3 alveoli (measurement A), W. pitikantensis being 20% shorter for this dimension. Comparison to the total sample of M2 specimens from Riversleigh (n = 7) indicates that all are larger than W. pitikantensis (see Appendix 1: Table A).

A much greater size difference is apparent between the postcranial elements of the Riversleigh specimen from Rackham’s Low Lion Site (see Figs. 5.7-5.9) and those of W. pitikantensis (SAM P37720). The humerus from Riversleigh (QM F23443) shows a similar morphology to that from Lake Pitikanta, however it is larger by a factor of 21% (humeral width; see Chapter 13). Similarly, comparison of the radius fragment from Riversleigh (QM F52239) with the radius of W. pitikantensis indicates that although the morphology is generally similar, the former is wider and deeper by a factor of 26%. Morphologically, the Riversleigh radius differs from the Lake Pitikanta radius in having a larger lateral styloid process and in this feature more closely resembles the radius of W. vanderleueri (see Chapter 5). Whether these size differences reflect specific differences or encompass the range found in a population is difficult to determine with so few specimens available. Archer and Dawson (1982) examined tooth size in samples of Thylacoleo carnifex from Wellington Caves and Naracoorte Caves and found that the range from the latter sample fell within that of the former and that both samples had low variability. Within the Wellington Caves sample the maximum size difference in the P3 sample was 18% and the maximum size difference between P3s was 24%. Measurements of humeri and radii of T. carnifex from Naracoorte Caves were made to gain an indication of the size range of these elements in a thylacoleonid species (see Table 4.1). Measurements of the humeri indicate a maximum difference of 16% in total length and of 19% for distal width. Measurements of the radii show a maximum difference of 14% in distal width 26

Table 4.1: Summary of measurements (mm) of humeri and radii of Thylacoleo carnifex from South Australian localities (mainly Naracoorte Caves).

Thylacoleo N mean observed range S.D. carnifex limb measures humerus length 9 215.5 200 - 232 10.55 humerus distal 10 75.9 67.8 - 82.1 4.93 width radius distal 8 42.8 39.7 - 45.5 2.02 width radius distal 8 19.7 18.4 - 21.7 1.24 depth

and 18% for distal depth. The value for humeral width is slightly less than the difference between the W. pitikantensis and the Riversleigh humerus, but the differences in radial measurements are much less than those observed for Wakaleo specimens.

Wakaleo vanderleueri was originally distinguished from W. oldfieldi primarily on the basis of differences in dental dimensions (Clemens and Plane 1974). This size difference was not substantial, being 1.3mm for M1 length and ranging from 1.1- 2.9mm for P3 length. Although the morphological differences between the respective M1s were relatively minor, the geographic separation of the sites involved and their possible age difference supported their recognition as separate species. Murray and Megirian (1990) have suggested that the successive increase in tooth size through time that is observed in Wakaleo species from W. oldfieldi, through W. vanderleueri and continuing to W. alcootaensis are indicative of a morphocline. On the basis of this trend, one could predict that earlier stages of this morphocline would be smaller than W. oldfieldi and are also likely to demonstrate gradual changes in tooth size. Both W. pitikantensis and the plesiomorphic Wakaleo taxon from Riversleigh support this model because they are smaller than W. oldfieldi and are from sites that are regarded to be older than those yielding W. oldfieldi (see Chapter 14). Wakaleo pitikantensis was collected from the Lake Pitikanta locality, Etadunna Formation, 27 which is estimated to be late Oligocene (24-26my) in age (Woodburne et al. 1993). The Riversleigh specimens have been collected from sites regarded as part of Riversleigh's System A sediments which are also estimated to be late Oligocene in age (Archer et al. 1989, 1997; see Chapter 14). Although accurate dating of the Riversleigh sites has yet to be achieved, it is possible that the size difference between W. pitikantensis and the Riversleigh taxon may also reflect different stages in the Wakaleo morphocline with the slightly smaller W. pitikantensis representing an earlier stage and the larger Riversleigh taxon representing a slightly later stage.

The possibility that the Riversleigh and Lake Pitikanta specimens may represent one species cannot be dismissed because size differences may reflect either regional differences (only partially reflected in dental measurements), age differences of individuals or sexual dimorphism within populations. It is unlikely that age is a key factor in accounting for the size differences between the Lake Pitikanta and Riversleigh specimens because all appear to be adult individuals. Fully formed alveoli for the posterior molars are present in all the maxillae indicating that all are adults. This age estimation is also supported by the fused epiphyses on the humeri and radii. Within populations of the Pleistocene thylacoleonid Thylacoleo carnifex, although dental measurements show no evidence of sexual dimorphism (Finch and Freedman 1982; Archer and Dawson 1982), Archer and Dawson (ibid.) suggest that the differences observed in skulls collected from Wellington Caves probably reflect sexual dimorphism. It is not known whether this difference is also reflected in the limbs of T. carnifex because postcranial remains of this species from this locality have not been identified.

On the basis of the overall larger size of the Riversleigh material, differences in M2 morphology, probable age difference, as well as the large geographic separation of the sites (more than 1000 km), it is considered to be a separate species to W. pitikantensis. Additional and more complete specimens of W. pitikantensis from the late Oligocene localities of central Australia are needed to resolve questions of conspecifity between these taxa. 28

4.5 Systematics

SYSTEMATICS

Superorder MARSUPIALIA Illiger, 1811 Order DIPROTODONTIA Owen, 1866 Family THYLACOLEONIDAE Gill, 1872

Wakaleo Clemens and Plane, 1974

Type species: Wakaleo oldfieldi Clemens and Plane, 1974

Other species: Wakaleo vanderleueri Clemens and Plane, 1974 Wakaleo alcootaensis Archer and Rich, 1982 Wakaleo pitikantensis new combination

Wakaleo hilmeri sp. nov.

(Figs. 4.3 – 4.9, Table 4.2)

Holotype: QM F45200, a near-complete skull with transverse fracture in the region of postorbital constriction, containing C1 P3 M1-3 and alveoli for I1-3 P2 and M4.

Type Locality: Hiatus Site, Riversleigh World Heritage Area, Lawn Hill National Park, northwestern Queensland.

Paratypes: From Rackham's Low Lion Site, QM F23443, a palate preserving C1 P3 M1-3 and alveoli for P1-2 and M4, and an edentulous left dentary containing only the roots of P3 M1-2 and alveoli for I1 M3-4. From Dirk's Towers Site, QM F24680, a left maxillary fragment containing M2 and alveoli for P3 M1, 3-4; QM

F30250, an upper left M1; and QM F30073, a lower left M1. From White Hunter Site, AR 19249, left and right dentaries, both containing P3 M1-3 and alveoli for I1, 29

P1-2 M4; QM F30465, a left dentary containing I1 P3 M1-2 and alveoli for P2, M3- 4, and QM F31376, a left dentary containing I1, P3, M1-2.

Referred material: NTM P 91167-3, a right dentary containing P3 M1-4 from Burnt Offering Site, Riversleigh World Heritage Area, northwestern Queensland.

Age: Hiatus, Rackham's Low Lion, White Hunter and Burnt Offering Sites are assigned to Riversleigh’s System A sequence of sediments (Creaser 1997). System A deposits have been interpreted as late Oligocene or possibly older (Archer et al. 1997) in age (see Chapter 14). Dirk’s Towers is regarded as part of Riversleigh’s System B sediments and is estimated to be early Miocene in age (Archer et al. 1997).

Etymology: named for Fred Hilmer, Vice Chancellor of the University of New South Wales.

Species Diagnosis: Wakaleo hilmeri differs from all other species of Wakaleo except W. pitikantensis in: retaining M4/4; being smaller, its P3/3 being two-thirds the length of P3/3 of W. vanderleueri and W. oldfieldi, and half the length of P3 of W. alcootaensis; having a more gentle decreasing molar gradient; P3 being relatively similar in length to M1; M1 having with a squarer lingual margin.

Wakaleo hilmeri differs from W. pitikantensis in being larger and in having a relatively longer diastema between C1 and P1.

It also differs from W. oldfieldi and W. vanderleueri in having: two teeth located between I1 and P3; M1 and M2 with relatively broader talonid basins; the entoconid and hypoconid on M1 more widely separated; P3 with a more sharply defined anterolingual crest and more concave surface between that crest and the anterior crest (rather than a broadly convex one). 30

nasal premaxilla

lachrymal foramen maxilla lachrymal

jugal

supraorbital frontal 47.1 process

frontal squamosal depression fossa alisphenoid

squamosal buttress sagittal crest subsquamosal parietal foramen

supraoccipital nuchal crest

0 10 20 30 mm Fig. 4.3: Cranium of Wakaleo hilmeri, holotype QM F45200, dorsal view. 31

Wakaleo hilmeri also differs from W. vanderleueri in having a cranium that: is relatively smaller, especially in length; is flatter dorsally; has broad dorsally- depressed frontals which have supraorbital margins that are raised and thickened; has relatively smaller and more anteriorly located supraorbital processes; has an occiput which is shorter, more rounded, deeper bilateral depressions and lacks a central septum; has a nuchal crest with central posterior concavity; has a cavernous postglenoid cavity not divided horizontally by a septum; has a narrower glenoid fossa but a broader postglenoid process; has a more prominent medial glenoid process; has a flatter tympanic wing; has a more obtuse rostral tympanic process on the periotic; has buttressing on the medial surface of the zygomatic arch; lacks a masseteric process. It also differs in having a smaller canine such that the canine alveoli does not extend beyond the posterior margin of anterior palatal foramen and in having M2 with a more strongly developed metaconule that results in a squarer lingual margin.

It also differs from W. alcootaensis in retaining P1 and P2, and in having a less arcuate molariform tooth row.

Description of the cranium of Wakaleo hilmeri

General The description is that of a skull recovered from Hiatus Site (QM F45200) and supplemented by QM F23443 and QM F24680. The skull, although slightly smaller than that of W. vanderleueri, is generally similar in much of its morphology and exhibits a robustness seen in that species. It is similar in width across the occiput but it is relatively shorter, being 88% of the length of W. vanderleueri. The cranium is twice the length of Lekaneleo roskellyae. The face is very broad and has a relatively short rostrum with forward facing orbits. Posteriorly, there is a prominent sagittal crest and a prominent, overhanging, nuchal crest. A strong postorbital constriction occurs slightly posterior to the midpoint of the skull.

The skull is preserved in two halves, the break between these halves runs across the region of the postorbital constriction such that the anterior half consists of the 32

sphenorbita parietal l

foramen supraorbita sagittal rotundum fronta l lachrymal lachrymal squamosa foramina nasa

subsquamosal premaxill foramen orbitosphenoi canine postglenoi palatin zygomatic maxill d alisphenoi infraorbital arch jugal foramen pterygoid

Fig. 4.4: Cranium of Wakaleo hilmeri, holotype QM F45200; Hiatus Site. Profile view. Scale bar = 25mm 33 rostrum and the palate, and the posterior half of the skull consists of the neurocranium and posterior halves of the zygomatic arches. Originally, the right molar row and right zygomatic arch were attached to the posterior half of the skull (see Fig. 4.5), however, the tenuous link by which it was attached could not support its weight. For ease of study, as well as symmetry, this piece was reattached to the remainder of the palate.

Most of the palate is intact although most of the incisors, right and left P2, and the right P3, are missing. The left tooth row preserves the crowns of C1, P3, M1-3. Only the base of I1 is present, this tooth having been broken off at the level of the palate. Alveoli are also present for I2-3, P2 and M4. The right tooth row preserves M1-4 (the anterolingual edge of M1 is damaged), the base of C1, the broken anterior root of P3, and alveoli for I1-3 and P2.

Rostrum

The rostrum is broad and narrows slightly anteriorly. In frontal view, the rostrum is gently rounded laterally and dorsally. In profile, the rostrum is slightly convex, the curvature being greatest at the anterior margin of the nasals while posteriorly the outline is quite flattened. Directly posterior to the canines, the rostrum has a dorsoventral depth of 39mm and width of 38.5mm. The dorsal surface of the rostrum forms a distinct depression between the supraorbital processes, just posterior to the nasals.

Nasal The nasals are long and narrow and measure 71mm in length. In dorsal view they are narrow at their anterior extremity and widen at the level of the posterior edge of the narial opening, forming a triangular, convex roof over this opening. The nasals widen again at the level of the superior lachrymal foramen and then narrow posteriorly forming a deep M-shaped contact with the frontals. Laterally, the nasals form a gently-arched contact with the premaxilla and maxilla. A moderately-sized 34 foramen lies near the lateral edge of the nasal at approximately halfway along its length and directly dorsal to the premaxillary-maxillary suture.

Premaxilla In frontal view, the premaxillae are gently convex. Together with the convex nasals they form a heart-shaped nasal aperture which has a maximum width of 26mm and an anterior depth of 18.7mm (between the tip of the nasals and the nasal cavity floor). The premaxilla laminae are broadly separated ventrally and enclose the roots of the incisors. Dorsally, the laminae converge near the nasopremaxillary contact. In lateral view, the premaxillae are taller than they are long and are slightly sinusoidal in outline. On the dorsal margin the nasopremaxillary suture is slightly arched, rising posteriorly. The premaxillo-maxillary suture runs from the midpoint of the canine alveolus in a dorsal direction, curving posteriorly over the dorsal third before contacting the nasal. The premaxillo-maxillary suture is 35.3mm long and forms an angle of approximately 80o with the horizontal axis. Ventrally, the premaxilla bears alveoli for three incisors and is incised by the anterior palatal fenestra. The alveolus for I1 is the largest, I2 is the smallest, and that of I3 is larger than I2 but smaller than I1. The sockets for the roots of all the incisors curve posteriorly. The alveoli for both I1s lack their anterolateral walls. These alveoli are large and longitudinally elliptical. The left I1 alveolus retains the root of the I1 which in cross-section measures 10.4 mm lengthwise and 7.6mm in width. The socket for the root is large and extends posteriorly above the I2 and I3 alveoli. The alveolus of I2 lies immediately posterior to I1, is slightly anteroposteriorly compressed, and measures 3.2mm in length and 4.1mm in width. The socket for its root is much shorter than that for I1, being 6-7mm in length. The alveolus for I3 lies immediately posterior to I2 and is relatively round. It measures 5.1mm in length and 4.2mm in breadth, and is relatively shallow in depth. The premaxilla extends for 2.5mm posterior to the I3 alveolus where it forms the anterior wall of the canine alveolus. 35

Maxilla In lateral view, the maxilla is roughly rectangular in outline, and is taller than it is broad. Its dorsal margin forms a short (14.6mm) convex contact with the nasal, and extends in a posteroventral direction to make contact with the anterolateral edge of the frontal. The maxilla is laterally convex. A large infraorbital foramen is located in the lower third of the maxilla about midway between the premaxilla and the suture with the jugal. This foramen is anteriorly-directed, subtriangular in shape, and measures 3.1mm in width and 4.7mm in height. On the left side a small foramen is located 1mm below the infraorbital foramen. The zygomatic process of the maxilla extends laterally forming an angle with the median plane of 70o, and forms a broad, anteriorly-directed, anterior margin of the orbit. Its articular contact with the jugal runs mainly in a ventral direction down the anterior surface of the zygomatic arch and then turns in a more posterolateral direction along the ventral surface. The maxilla forms a deep suborbital shelf at the anteroventral margin of the orbit. A number of large pits and furrows lie at the posterior edge of the infraorbital shelf just lateral to the posterior infraorbital foramen. Medially, the maxilla contacts the lachrymal. The maxillolachrymal suture courses posteriorly from the anterior edge of the orbit to the posterior infraorbital foramen. At this foramen the maxilla contacts the palatine, the maxillopalatine suture running posteriorly along the medial wall of the infraorbital canal and then descending posteroventrally to the palatal surface approximately 10mm behind M4. On the palatal surface this suture then runs anteriorly to the level of M4 and then turns medially toward the posterior palatal fenestra. In ventral view, the posterior palatal fenestrae extend anteriorly into the maxilla, forming its posterior margin in the central area of the palate. Although missing in QM F45200, in QM F23443 a small spinous process extends posteriorly into the posterior palatal fenestrae at the region of intermaxillary suture.

The maxilla bears the posterior half of the canine alveolus, one or two single-rooted anterior premolars, a doubled-rooted third premolar, and four molars which progressively decrease in size. A number of specimens indicate that the anterior premolar number is variable. The palate specimen QM F23443 bears alveoli for two single-rooted anterior premolars with the P1 alveolus being located 3.5mm posterior 36 to the canine and approximately 2.5mm anterior to the P2 alveolus. The second premolar alveolus is smaller (1.7mm diameter) and is located at the base of P3. In contrast, the holotype QM F45200 and maxilla fragment QM F24680 possess an alveolus for only a single anterior premolar located at the base of P3. This location suggests this alveolus is that of P2.

Jugal Neither jugal is completely intact, the zygomatic arches missing portions from their central regions; more of the left jugal is preserved than the right. The jugal is thick and extends in a posterolateral direction producing a broad, robust face. Its lateral surface is convex longitudinally and its medial or orbital surface is concave. Anteriorly, the jugal contacts the maxilla and also makes a brief contact with the lachrymal at its anterodorsal extremity. From its anterior border with the maxilla, the jugal descends slightly and then curves dorsally, producing a concave dorsal edge on the anterior half of the zygomatic arch. The jugal contacts the squamosal on the dorsal edge of the zygomatic arch 22.1mm posterior to its contact with the lachrymal. The jugal then runs below the squamosal becoming dorsoventrally thinner and wider at its posterior extremity. A distinct ridge, the masseteric border, runs along the ventral third of its lateral surface. The jugal surface ventral to this border is angled medially. The jugal terminates 8.3mm anterolateral to the glenoid process. On its ventral surface, just anterior to the jugosquamosal suture, the jugal forms a weak ridge that runs transversely across the zygomatic arch. A slight circular depression lies immediately anterior to this ridge.

Palate The palate is broad but narrows anteriorly and to a lesser extent posteriorly. From the front of the palatal arch to the anterior margin of the posterior palatal fenestra the palate measures 60.0mm. The width between the canines is 19.9mm, between the anterior roots of P3 is 39.7mm, and between the alveoli of M4 is 33.4mm. The palate is widest between P3/M1 (41.6mm). A narrow palatal arch runs between the rows of incisors. At the level of I3, elongate anterior palatal fenestrae incise the palate. These 37

anterior palatal 1 fenestra I premaxilla I2 I3 maxilla C palatal ridges P2 fracture 3 jugal P M1 M2 posterior palatal 3 fenestra M M4

transverse palatine palatine presphenoid process lateral transverse canal alisphenosquamosal pterygoid pterygoid suture process foramen ovale alisphenoid entocarotid squamosal bs foramen postglenoid process medial glenoid process postglenoid cavity tympanic wing periotic basioccipital mastoid paroccipital epoxy repair of fracture (---) occipital condyle condylar foramen 0 1 2 3 cm Fig. 4.5: Ventral view of the cranium of Wakaleo hilmeri, holotype QM F45200. (bs = basisphenoid) 38 fenestrae are 11-12mm long and approximately 2.5mm wide and continue posteriorly to just beyond the canines. The lateral margins of the fenestrae are on a more ventral level than their medial margins. The premaxillary bone that lies between the fenestrae is 6.5mm wide anteriorly, and broadens posteriorly (narrowing the fenestrae). The premaxillo-maxillary suture runs transversely between the medial edge of the canine alveolus and the lateral margin of the fenestrae. Posterior to the canines the palate begins to widen markedly, reaching its maximum width just posterior to P3. The palatal arch is gently concave across this area, becoming flatter in the centre of the palate. Two, short (4-5mm) palatal ridges lie on the lateral margins of the palate. An anterior ridge lies approximately 6.4mm medial to the P2 alveolus and extends anteriorly parallel to the diastema. A second ridge lies approximately 6.5mm medial to the posterior root of P3. Both ridges are angled anteromedially. A small foramen lies 10mm medial to the anterior root of the left P3. A pair of larger foramina pierces the palate at the level of M1/M2. Numerous pinhole-sized foramina lie along the lateral margins of the palate, just medial to the molars. The posterior part of the palate is dominated by large posterior palatal fenestrae (21.0mm wide) that extend from the level of M2 to the transverse palatine process (28.5mm long). On the right side of the palate the postalveolar process extends posteriorly for approximately 20mm. A remnant of the transverse palatine process measuring 4mm deep (anteroposteriorly) is evident near the posterior end of the post-alveolar process. The highly intercalated maxillopalatine suture courses from the lateral margin of the posterior palatal fenestra (at approximately the level of M2) in a posterolateral direction, and reaches the lateral margin of the postalveolar process approximately 9mm posterior to the M4 alveolus.

Lachrymal The lachrymals on QM F45200 are nearly complete; the right is lacking its ventral margin and the left is damaged such that it is pierced by a large hole in the centre of its surface. The lachrymal forms the anteromedial wall of the orbit and anteriorly extends for a short distance onto the rostrum. Two foramina are located in the anterior portion of the lachrymal, one above the other. The superior foramen is located on the posterior edge of the anterior orbital margin. This foramen is small 39

(diameter 2.3mm) and directed dorsally. The inferior lachrymal foramen is larger, vertically elliptical, and located in a small, laterally-directed fossa. The anteromedial edge of the orbit lying directly anterior to the superior foramen is raised and extremely rugose. The maxillolachrymal suture which lies anterior to the foramina is highly convoluted. Dorsally, the suture arches posteriorly, contacting the frontals, and then descends in a posteroventral direction across the flat medial wall of the orbit. Posteroventrally the lachrymal makes contact with the palatine just dorsal to the posterior opening of the infraorbital canal. Laterally, the lachrymal extends a short distance forming the anterodorsal margin of the orbit.

Palatine Both palatine bones on QM F45200 are damaged. The transverse fracture of the skull passes through these bones and they are preserved on both the anterior and posterior halves of the skull. The right palatine is lacking its dorsal margin and part of its posterior margin and the left is lacking a central portion. In lateral view, the palatine articulates dorsally with the frontal bone and extends anteriorly as a narrow strip along the medial wall of the infraorbital canal where it is bordered dorsally by the lachrymal and ventrally by the maxilla. The centre of the palatine is pierced by the sphenopalatine foramen which is large and round, measuring 3.0mm in diameter. A much smaller foramen lies 2mm posterior to the sphenopalatine foramen. The palatine is bordered posteriorly by the alisphenoid, by the orbital wing of the presphenoid posterodorsally, and by the pterygoid posterolaterally. At its posterolateral border with the alisphenoid, just lateral to the foramen rotundum, the palatine forms a small posterolaterally-directed process. Ventrally, it forms the lateral margins of the posterior palatal fenestrae and extends anteriorly to the level of

M2-3. Posteriorly, the ventral surface of the palatine narrows slightly and forms a transverse palatine process approximately 14.5mm posterior to M4. Only the lateral margins of this process are preserved in QM F45200. Small posterolateral palatine foramina are located 3mm anterior to the transverse palatine processes.

Vomer 40

Both nasopharyngeal canals are filled or partly filled with matrix. At the anterior margin of the posterior palatal fenestrae a small segment of the sagittal part of the vomer can be seen. This part of the vomer forms a rounded septum and forms a relative steep medial wall of the nasopharyngeal meatus. Posteriorly, its articulation with the presphenoid and palatine is obscured.

Neurocranium

The neurocranium is narrow anteriorly and broad posteriorly. Relatively tall sagittal and nuchal crests are distinctive features on the dorsal surface. Ventrally, the tympanic region is rather open; the tympanic wing is moderate in size and the outer ear canal is rather broad, affording a relatively unimpeded view of this region.

Frontal The frontals are broken posteriorly, the fracture in the skull passing through this region of the skull. In dorsal view, the frontals are broad anteriorly and strongly constricted posteriorly. The anterior region of the frontals is marked by the presence of distinct triangular supraorbital processes that project laterally into the orbit. The processes are rugose and are approximately 5.2mm thick. The width of the frontals between the supraorbital processes is 47.1mm. Between these two processes the frontals form a broad depression that narrows posteriorly. The anterior surface of the depression bears many longitudinal striations and the posterior surface is covered by numerous small pits. The posteriorly-converging lateral margins of the depression are raised and thickened (5.1mm) and commence anteriorly as slight ridges on the supraorbital processes and become more prominent as they converge.

Ethmoid The transverse fracture through the skull of QM F45200 exposes the cribriform plate of the ethmoid. The posterior surface of the plate can be observed on the anterior half of the skull. The plate is 14.7mm wide and is divided in two by a vertical septum. The surface of the plate is pierced by numerous foramina. Dorsally, the cribriform 41 plate contacts the inner table of the frontal bone, a dense area of frontal bone 13.5mm deep that is also perforated with numerous small foramina.

Parietal The parietals are long and relatively narrow. Anteriorly, their articulation with the frontal is obscured because of damage associated with the transverse cranial fracture. Dorsally they form a tall, thickened ridge that together forms a strong sagittal crest. Below the sagittal crest the parietal laminae curve concavely and extend laterally to meet the squamosal. The dorsal edges of the squamosal laminae are missing (approximately 5mm) and the underlying parietal surface bears many vertical grooves. At the posterior end of the sagittal crest the parietals do not continue to ascend onto the nuchal crest but form a posteriorly-concave margin with the squamous part of the supraoccipital bone, approximately 10mm anterior to the nuchal crest. The parietals do extend to the nuchal crest at their posterolateral extremity.

Presphenoid The presphenoid is not complete, the transverse fracture of the skull running through the anterior region of this bone. In lateral view, the presphenoid forms the anterior floor and roof of the sphenorbital fissure and is bordered anterolaterally by the palatine, dorsally by the frontal, posteriorly by the basisphenoid and posterolaterally by the alisphenoid. The sphenorbital fissure is approximately 9mm long, is elongate and relatively deep. Dorsal to the sphenorbital fissure, the presphenoid forms the anterior half of a broad and deep fossa which is delineated superiorly by the prominent infratemporal crest and inferiorly by a thin, sharp, horizontal crest (the alisphenoid forms the posterior part of this fossa). Ventrally, the presphenoid is a narrow tongue of bone which lies anterior to the basisphenoid and is flanked on either side by the palatine. Centrally, it forms a longitudinal septum which becomes narrower and more prominent anteriorly. Laterally, it forms a highly convoluted suture with the palatine. 42

10.4 x 7.6 (length x breadth)

11.3 x 2.5 3.2 x 4.1 5.1 x 4.2

7.5 x 5.3

3.1 x 1.5

39.0

41.6

164

52.5

16.8

35.4

73.1

Fig. 4.6: Cranial measurements of Wakaleo williamsi, QM F45200. Measurements in mm. 43

Basisphenoid The basisphenoid is roughly triangular in shape, being narrow anteriorly where it contacts the presphenoid and broadening posteriorly where it is strongly fused to the basioccipital. It is flat and inclined anterodorsally. It is flanked by the pterygoids.

Pterygoid Most of the pterygoids have been lost and only their basal attachments to the basisphenoid and alisphenoid are preserved. More of the left bone is preserved than the right, the left measuring 34.4 mm in length. In ventral view, they extend from the posterior margin of the palatine to the opening for the internal carotid artery, curving laterally where they contact the basisphenoid.

Alisphenoid The alisphenoid forms much of the anteroventral floor of the braincase. In lateral view, it makes a brief contact with the frontal and parietal dorsoanteriorly and forms the posterior part of the sharp and prominent infratemporal crest. This crest becomes rounder posteriorly and laterally becomes contiguous with the anterior edge of the zygomatic process. In this region the alisphenoid makes a broad contact with the squamosal and forms a small dorsal strip along the anterior edge of the zygomatic process before descending to the ventral surface at the anterior end of the medial glenoid process. Ventral to the infratemporal crest and immediately dorsal to the sphenorbital fissure, the alisphenoid forms the posterior part of a broad fossa. At the posterior margin of the fissure this fossa has a dorsoventral depth of 10.0mm. The presphenoid forms the anterior part of this fossa and forms a vertical sinuous suture with the alisphenoid. Ventral to the sphenorbital fissure the alisphenoid forms the roof of the large, round foramen rotundum (3.0mm diameter) and then extends anteriorly as a small wing that contacts the palatine. Ventral and slightly posterior to this wing, it forms a flattened, rounded, laterally-projecting process (lateral pterygoid process). Ventral to this process the alisphenoid articulates with the pterygoid bone. Posterior to this process the alisphenoid forms a very shallow, oval-shaped depression, the pterygoid fossa (for the attachment of the medial pterygoid muscle). 44

In ventral view, the alisphenoid extends laterally from the pterygoid fossa to the medial base of the medial glenoid process as a broad, ventrally-concave surface. At the medial base of the medial glenoid process it makes a distinct contact with the squamosal. At the posterior margin of the pterygoid fossa lies the large foramen of the transverse canal. This opening is directed posterolaterally. Approximately 10mm posterior to the transverse canal lies the foramen of the entocarotid. A broad sulcus runs posterolaterally from the entocarotid foramen and is overlain posteriorly by the periotic. Approximately 8mm posterolateral to the foramen of the transverse canal lies the foramen ovale which has a diameter of 3.5mm and is anteriorly-directed. Directly posterolateral to the foramen ovale the alisphenoid forms the tympanic wing. The tympanic wing is rounded posteriorly, is ventrally convex, and has a transverse width of approximately 8-9mm. The tympanic wing contacts the squamosal at its anterior, medial and lateral margins, and a distinct suture with this element can be seen on the anterior wall of the squamosal epitympanic sinus. Significantly, the contribution of the alisphenoid to the tympanic floor varies between the left and right sides. On the right side the floor is formed completely by this element, however, on the left the squamosal forms a small contribution on the anterolateral edge of the floor.

Squamosal The squamosal covers most of the lateral wall of the brain case. It is bordered dorsally by the parietal. On the left side much of the dorsal edge (5mm) is missing although evidence of its contact is marked by the presence of numerous fine vertical grooves on the surface of the parietal. In lateral view, the squamosal forms a convex wall over the lower two-thirds of the brain case and extends laterally to form a broad zygomatic process. Anteriorly, it contacts the alisphenoid, but does not make contact with the frontal. Just dorsal to the anterior margin of the zygomatic process lies a horizontally-elongate depression approximately 5mm in diameter and which is anteriorly-directed. As the zygomatic process curves anteriorly its lateral edge rises to form a tall, thin margin that tilts medially to slightly overhang the orbit. The medial surface of the zygomatic process bares two buttress-like vertical swellings. The lateral surface is convex and steep, and pierced by a foramen approximately 10mm anterior to the glenoid notch. 45

In ventral view, the squamosal is bordered by the alisphenoid medially and the mastoid posteriorly. The ventral surface of the zygomatic wing bears a transverse, elongate, rounded fossa for the condyle of the mandible (14.5mm in width). The glenoid fossa opens laterally through a small notch. On the ventral surface of the right zygomatic wing, anterolateral to the glenoid notch, the surface is rugose and depressed marking the articular surface for the jugal. Posterior to the glenoid fossa the squamosal develops a broad, deep, rounded postglenoid process that measures 14mm across at the base and approximately 11.5mm in depth. At the medial extent of the glenoid fossa the squamosal develops a moderate ridge, the medial glenoid process. This process runs in a posteromedial direction from the orbital margin to the tympanic cavity where it projects as a small process just lateral to the tympanic wing and 4.5mm medial to the postglenoid process. The apex of the ridge is rugose. On the medial side of the medial glenoid process the squamosal contacts the alisphenoid.

The squamosal is the main contributor to the roof of the auditory region. Medial to the posterior end of the medial glenoid process the squamosal inflates anteriorly forming the squamosal epitympanic sinus which measures approximately 6mm in diameter. In ventral view the sinus is roofed by an alisphenoid tympanic wing and at the anterior base of the tympanic wing this element lies against the squamosal such that if the wing were broken at its base it would appear as a bilaminar structure. Medially, the epitympanic sinus is bordered by a small process from the alisphenoid tympanic wing and the periotic.

In posterolateral view, the posterior surface of the postglenoid process is broad and gently convex. At the dorsal margin of the process the squamosal thickens, forming a bridge between the posterior margin of the zygomatic process and the brain case. As the zygomatic process connects with the brain case it is pierced by a large, oval, subsquamosal foramen which is posteriorly-directed and measures 3.4mm in diameter. A small, triangular depression lies ventral to this foramen at the posterodorsal edge of the external auditory meatus. Posterior to the subsquamosal foramen a narrow strip of the squamosal forms the lateral margin of the nuchal crest and abuts the mastoid posteriorly. Medial to the postglenoid process the squamosal inflates anterodorsally to form a deeply recessed postglenoid cavity. 46

50mm

sagittal supraoccipital

zygomatic arch nuchal crest epoxy 27.7

squamosa postglenoid 19.1 mastoid 35.3 condyloid foramen mastoid process paroccipital process 73.2 foramen magnum

Fig. 4.7: Morphology and dimensions of the occiput of Wakaleo hilmeri, holotype QM F45200. Measurements in mm. 47

Basioccipital A transverse fracture, marked by a thin line of epoxy resin, runs across the posterior region of the basioccipital. The basioccipital is hexagonal in outline and bears a central longitudinal crest which is flattened anteriorly and becomes sharper posteriorly. The central crest is flanked by broad, deep fossae. The lateral margins of the basioccipital, medial to the entocarotid foramen (alisphenoid) and the periotic, are raised, thickened and rugose. A large foramen pierces the basioccipital posteriorly near the central crest and a smaller foramen opens anteriorly in the anterior part of the left fossa. A pair of moderately-large foramina (opening for rami of the hypoglossal nerve) lies at the posterolateral margins of the basisphenoid. Both foramina lie in a small depression and are directed anteroventrally. One foramen lies anteromedial to the other.

Occipital Region

The occipital region is broad, measuring 73.2mm in width. Its dorsal margin is strongly arched and is formed by a prominent overhanging nuchal crest. The depth from the top of the nuchal crest to the ventral surface of the occipital condyles is 45.0mm. As the lateral edges of the nuchal crest ascend toward the sagittal crest they course anteriorly and form a posteriorly-concave recess in the centre of the nuchal crest. This results in the crest having a broad W-outline when viewed dorsally.

In posterior view the occiput has an undulating surface composed of three fossae, one central and two lateral. The supraoccipital is lens-shaped and bears the deep central fossa. Its height between the nuchal crest and the dorsal rim of the foramen magnum is 27.7mm. In dorsal view, the squamous part of the supraoccipital is evident on the anterior surface of the nuchal crest where it forms a small depression on each side of the sagittal crest. At its lateral margins the supraoccipital is bordered by the mastoid dorsally and exoccipital ventrally. The supraoccipito-exoccipital and supraoccipito-mastoid sutures course in a dorsolateral direction from just dorsal of the occipital condyles to the nuchal crest. The central fossa bears a very weak median external occipital crest and is pierced by four large foramina and numerous 48 pinhole foramina. The dorsolateral margins of the fossa are rugose. The central fossa is separated from 48 the lateral fossa by broad, gently-rounded crests that run in a ventromedial direction from the nuchal crest toward the foramen magnum. The large lateral fossae are formed at the region where the supraoccipito-exoccipital suture meets the supraoccipito-mastoid suture.

The occipital condyles are large and prominent, and project posteriorly beyond the level of the nuchal crest. The condyles are rounded posteriorly and ventrally, narrowing toward their ventral margins where they are separated by an intercondyloid notch that is 11.7mm wide. The foramen magnum is large, measuring 12.0mm in height and 19.1 mm in width. It is oval in shape and opens posteriorly. A pair of foramina lies on the internal margins of the foramen magnum at the base of each occipital condyle. Both foramina are directed posteromedially. The dorsal foramen is very large (dorsoventral diameter approximately 3.5mm) and its ventral margin is developed into a short, posteriorly-directed sulcus. This foramen is the external opening of the condyloid canal. The second foramen lies anteroventrally to the condyloid foramen, is round and much smaller (diameter 1mm), and is the opening for the hypoglossal nerve.

Between the occipital condyle and the paroccipital process is the ventral condylar fossa which is concave and has a width of approximately 4mm. The posterior edge of the right paroccipital process is damaged. The paroccipital processes are rounded and project posteroventrally to a level similar to that of the mastoid process but well short of the ventral margin of the occipital condyles. A small condyloid foramen lies on the posterior surface of the paroccipital just dorsal to the condyloid fossa. The ventral surface of the paroccipital process forms a narrow longitudinal process that has a sharp lateral edge which abuts against the mastoid process. At its anterolateral margin, the paroccipital process forms a slight projection over the mastoid process and anteriorly is separated from the posterior end of the periotic by a sinus approximately 25mm in diameter. Medially, the exoccipital fuses strongly with the basioccipital. Near the lateral margins of the occiput the mastoid forms a relatively broad crescentic band (approx. 10.3mm wide) between the paroccipital and the narrow lateral squamosal margin. The dorsomedial half of the mastoid surface is depressed and forms part of the lateral fossa of the occiput. 49

medial glenoid process jugal alisphenosquamosal suture

transverse canal glenoid fossa foramen ovale postglenoid process entocarotid foramen postglenoid foramen tympanic wing

petrosal postglenoid cavity

rostral tympanic process post-tympanic process of squamosal

mastoid process of petrosal

posterior lacerate paroccipital foramen process

Fig. 4.8: Left basicranial region of Wakaleo hilmeri, QM F45200 holotype; Hiatus Site.

Fig. 4.9: Wakaleo hilmeri, QM F23443, palate; Rackham’s Low Lion Site. Scale bar = 50 mm 50

Auditory Region

Outer ear canal The auditory ossicles and ectotympanics in QM F45200 are missing, hence there is an unobstructed view of the outer ear canal. In ventral and lateral aspect the outer ear canal is very broad, measuring approximately 8mm in anteroposterior diameter. The canal walls are formed by the squamosal. The anterior wall of the canal is formed by the gently curving posterior surface of the postglenoid process which bears a broad (4mm wide) but shallow sulcus at its medial margin that continues into the postglenoid cavity. The postglenoid cavity is large and inflates anteriorly and dorsally. At the posteromedial margin of the postglenoid cavity is a small, deep, oval fossa into which opens the canal from the subsquamosal fossa. A narrow transverse sulcus courses across the anterior wall of the postglenoid cavity and connects (piercing the fossa's anterior wall on the left side) with this fossa. At the ventral margin of this fossa is a small, posteriorly-opening foramen, the postglenoid foramen. The postglenoid cavity is separated from the epitympanic sinus by a thick septum from which projects a small spur at approximately the septum’s midpoint.

The posterior wall of the outer ear canal is dominated by a large anteriorly-concave, laterally-directed sulcus, the stylomastoid notch. Immediately dorsal to the notch, the wall of the canal forms a small depression that is bounded dorsally by a slight transverse crest. This depression would have been a recess for the ectotympanic.

Periotic Both periotics are preserved and are in their original, in-vivo positions. Only their ventral and lateral surfaces can be observed and described. The posterior region of the right periotic has been repaired and the morphological details of this region are slightly obscured by a thin line of epoxy resin.

The periotic lies in a large sinus that is bordered medially by the basioccipital, laterally by the squamosal, anteriorly by the alisphenoid, and posteriorly by the paroccipital process. In lateral view, the periotic is subrectangular in shape with its long axis oriented in an anteroventral-posterodorsal direction. Along this axis the 51 periotic has a length of 14mm. In ventral view the periotic is thin and flattened anteriorly and is broad and rounded posteriorly.

Tympanic Face: The anterior region of the promontorium is rather flattened while the posterior part of the tympanic face is convex. Three fenestrae are located across the dorsal margin of the tympanic face: the central opening, the fenestra vestibuli, is horizontally oval and opens laterally; posteroventral to the fenestra vestibuli lies the large foramen of the fenestra cochleae, which is semicircular in shape and is posteriorly-directed; anterodorsal to the fenestra vestibuli lies the small foramen for the facial nerve (secondary facial foramen) which is dorsoventrally oval and opens posteriorly. A broad sulcus runs posteriorly from the secondary facial foramen, coursing along the dorsal margin of the fenestra vestibuli for approximately 2.5mm. Directly ventral to the fenestra vestibuli and anterior to the fenestra cochleae, a large prominent process, the rostral tympanic process, projects from the ventrolateral portion of the periotic. In lateral view, this process can be seen lying between tympanic wing anteriorly and the mastoid process posteriorly. The process is broader at its base, rounded ventrally, and narrows laterally to a blunt point.

Anterolateral to the secondary facial foramen the periotic forms a narrow epitympanic wing which is bordered anterodorsally by the roof of the squamosal epitympanic sinus, appearing to pass medially beneath the edge of the squamosal. At the posterior end of the epitympanic wing, lateral to the secondary facial foramen, the periotic forms a deep fossa, the incudal fossa. Posterior to the incudal fossa and lateral to the fenestra cochlea lies the fossa for the stapedius muscle. A low, rounded septum separates these two fossae. The stapedial fossa is deep and has two depressions within its anterior half.

Ventral Surface: In ventral view the periotic appears triangular, being narrow anteriorly and becoming rounder and broader posteriorly. Medially, it lies against the basioccipital. Its lateral surface is flattish and inclined anteroventrally toward the epitympanic sinus. Its posterior margin is gently rounded and is marked laterally by the prominent rostral tympanic process. A large sinus, the tympanic sinus (= the mastoid epitympanic sinus of Archer 1976) separates the posterior end of the periotic 52

Table 4.2: Dimensions of cheekteeth of Wakaleo hilmeri (mm)

** = average of left and right tooth rows; a = alveolus measurement

Upper cheekteeth

Specimen # Riversleigh Site/ P3 P3 M1 M1 M2 M2 M3 M3 M4 M4 P3 -M4 P3:M1 P3 / Local Fauna length width length width length width length width length width P3 – M4 QM F23443** Rack. Low Lion 12.2 7.1 9.4 9.2 6.0 6.3 4.5 4.9 3.5a 35.8a 1.30 0.34 QM F45200** Hiatus 11.8 7.0 8.8 8.6 6.3 6.9 4.7 5.1 34.0 1.34 0.34 QM F24680 Dirk’s Towers 12.9a 7.8a 8.9a 9.3a 6.3 7.6 4.9a 4.3a 4.4a 4.0a 35.8a 1.44a 0.36 QM F30250 Dirk’s Towers 9.7 9.9 QM F23441 Upper 9.0 9.0 QM F20573 White Hunter 5.7 7.3 QM F30378 White Hunter 6.2 7.5 QM F23444 White Hunter 4.7 4.4 QM F23801 Neville’s Garden 6.5 7.8 AR 16653 White Hunter 5.6 6.3 QM F24660 Creaser’s Ramp 4.7 4.7

Lower cheekteeth

Riversleigh site/ P3 P3 P3 M1 M1 M1 M2 M2 M2 M3 M3 P3:M1 Specimen # Local Fauna length anterior posterior length anterior posterior length anterior posterior length width width width width width width width AR 19249** White Hunter 10.1 4.9 5.9 8.7 7.1 6.5 7.4 6.3 6.1 5.8 4.9 1.16 QM F30465 White Hunter 9.5 5.4 6.0 10.0 6.6 7.1 7.7 6.0 6.1 0.95 QM F31376 White Hunter 8.6 4.4 8.7 6.1 6.1 7.1 5.7 5.4 0.98 NTM P91167-3 Burnt Offering 12.2 9.3 8.5 7.3 7.3 6.2 5.7 1.31 NTM P91171-4 Burnt Offering 11.5 6.5 7.4 QM F30073 Dirk’s Towers 9.4 6.6 6.4 QM F23449 Upper 9.6 6.2 5.6 53 from the paroccipital. At the medial end of this sinus lies a large, deep, transverse sulcus that forms the posterior lacerate foramen.

Mastoid The mastoid part of the periotic is well developed posteriorly into an elongate process. Lateral to the fenestra cochleae lies an elongate stapedial fossa. This fossa is relatively broad. The lateral wall of the fossa forms the medial edge of the mastoid process. The posterior margin of the fossa forms a small ridge that runs posterolaterally to the mastoid process (the caudal tympanic process). In ventral view, the mastoid is narrow anteromedially and widens laterally to form a rounded, obtuse process. The posterior edge of the left mastoid process is quite rugose, bearing a number of small tubercles. On the medial edge of the ventral surface a small transverse sulcus runs into the tympanic sinus. Anteriorly, the mastoid makes a broad contact with squamosal.

Upper Dentition (Fig. 4.9, 4.10 & Table 4.2)

The description of the upper dental morphology of the species is based mainly upon the holotype from Hiatus Site (QM F45200). Variations observed in the paratypes are referred to where appropriate.

Incisors: Only the alveoli for the incisors have been preserved, indicating the presence of three. Descriptions of the alveoli are found above in the description of the palate. Canines: the canines are the first known for any species of Wakaleo. The canines in QM F45200 are broken and hence the following description is based on QM F23443. The canines are short, recurved slightly, with a broad, oval base. The major cusp is anteriorly situated. A small longitudinal ridge ascends anteriorly and posteriorly from the occlusal tip separating the lingual and buccal surfaces. Enamel is present on lingual and buccal surfaces from the tip to about half way up the tooth. At the level of the alveolus, the root of the canine runs almost horizontally into the tooth socket.

Premolars: no anterior premolar crowns are preserved, only their alveoli are present. 54

stylar basin 10 mm p longitudinal pa me blade pa me pa ac me pa

2 P me

prpaclc mcl prtrb pr pr alc pr 3 1 P M M2 M3 M4

ac = anterior cusp; alc = anterolingual cusp; me = metacone; mcl = metaconule; pa = paracone; pr = protocone; prpaclc = preparaconulecrista; trb = trigone basin (M4 is mirror–image from right side)

Fig. 4.10: Wakaleo hilmeri, QM F45200, left cheek dentition; Hiatus Site. 55

P1: the presence of P1 is variable. An alveolus for P1 is present in QM F23443 but is absent in the holotype (QM F45200). The alveolus in QM F23443 lies 3.5 mm posterior to the canine alveolus, is larger than the P2 alveolus, is round, and measures 2.4mm in width and 3.3mm in length. P2: the P2 alveolus is small and lies at the base of the anterior root of P3. In QM F45200 the alveolus is 1.6 mm wide and 3.1mm long. P3: the left P3 is intact but worn on its lingual side. Only the anterior root of the right P3 is preserved. P3 is sectorial, having an anterior and posterior cusp that are joined by a longitudinal blade. This blade is notched so that in profile the complete cutting edge of the tooth is W-shaped. The lingual occlusal edge of the longitudinal blade is worn exposing a thin arc of dentine between the anterior and posterior cusp. This wear would have been produced by the buccal edge of the longitudinal blade of P3. On the posterior margin of the lingual surface an area of enamel above the posterior cusp has spalled off. Wear facets do not normally occur in this region of a thylacoleonid P3 and it is uncertain whether this damage has occurred pre or post- mortem. The anterior and posterior cusps are subequal in height. An anterior crest ascends from the anterior cusp, curving lingually, to form a small vertical crest before merging with the base of the crown. The lingual surface bears a distinct anterolingual crest that ascends from the anterior cusp, terminating about half way up the crown, where it forms a prominent rounded cusp on the swollen lingual surface below the anterior root. A narrow, fluted valley lies between the anterior crest and the anterolingual crest. The posterior portion of the lingual surface forms a broad convex surface below the posterior root.

The anterior buccal surface is gently convex. An anterior buccal blade curves posterodorsally from the anterior cusp and terminates near the base of the crown just below the point where the anterior and posterior roots meet. This crest becomes more prominent half way up the crown. The central region of the buccal surface bears a broad valley which connects the anterior buccal blade and the convex buccal surface above the posterior cusp. A posterior crest runs from the posterior cusp to the posterior margin of the tooth (the posterior portion of the longitudinal blade). At the 56 posterior end of the blade a small posterior buccal crest curves anterodorsally. A small basin lies between this crest and the posterior cusp.

Molars: the molar series comprises four molars. The molars decrease in size posteriorly. In QM F45200 the left M4 is missing and both M1's are well worn, the right M1 also exhibiting damage along its anterior edge. The isolated M1 from Dirk's Towers (QM F30250) is unworn and the following description is based on the two specimens. M1: is subtriangular in shape. The protocone, metacone and paracone are all well developed. The paracone is slightly taller than the metacone and is connected to the latter by a longitudinal crest that is notched midway. A short preparacrista runs anterolingually from the paracone and joins stylar cusp B at the anterior edge of the tooth. This blade is contiguous with the posterior crest that ascends from P3. From stylar cusp B an anterobuccal crest ascends the anterior buccal edge of the crown. A small anterior crest runs lingually from stylar cusp B across the front edge of the tooth. From the metacone a short postmetacrista runs posteriorly to the posterior margin of the tooth where buccally it connects with the posterior end of the stylar basin and lingually it connects with a short lingual crest that merges with the posterior surface of the crown. The buccal surface of M1 bears a stylar basin that runs from the posterobuccal surface of the paracone to the posterobuccal corner of the tooth. The basin narrows on the buccal surface of the metacone. On the unworn molar QM F30250, it is evident that the anterior and buccal margins of the stylar basin are formed by a small ridge which ascends half way up the buccal flank of the paracone in a postero-oblique direction before turning posteriorly and running horizontally to the posterior end of the tooth. Numerous fine ridgelettes line each side of this ridge. A small, vertical, kinked ridge also runs up the midsection of the buccal face from the stylar ridge to the base of the crown between the anterior and posterior roots. No similar crest is obvious on QM F45200 or QM F23443 but it may possibly have been obliterated by wear on these specimens. On the lingual face of the paracone a crescent-shaped preparaconulecrista runs transversely, terminating at a notch about two-thirds of the way between the paracone and protocone. At the lingual end of this crest is a weak paraconule that bears a shearing facet caused by 57 wear against the posterior face of the metaconid of M1. The protocone is well developed. A short preprotocrista runs anterolaterally from the protocone to a notch at approximately the midpoint of the anterior edge of the trigon basin. A postprotocrista runs posteriorly to a small notch about two-thirds of the way between the protocone and the metaconule. This crest is slightly worn and forms a narrow, shearing surface. A premetaconulecrista runs posteriorly from the notch to the metaconule. A short postmetaconulecrista curves posterolaterally, merging with the posterior margin of the tooth. This crista is also slightly worn flat just posterior to the metaconule. A short, crescentic blade ascends the lingual face of the metacone. The crests connecting the major cusps form the margins of a subtriangular trigon basin. The trigon basin is lightly crenulated. M1 has three roots, the posterior being the largest and the lingual and anterior buccal being subequal in size.

In QM F45200 and QM F23443, wear has resulted in the loss of enamel from the three primary cusps and much of the curvature of the crests is lost leaving relatively distinct, flat shearing surfaces. The anterior crescentic crest on the lingual flank of the paracone becomes much straighter and the area of the tooth anterior to this crest becomes a relatively flat facet. Similarly, the area anterior to the protocone, anterior and posterior to the posterior crescentic crest, the lingual edge of the longitudinal crest, and the medial portion of the trigon basin are also worn to flat, shearing surfaces.

M2: this molar in the holotype is heavily worn and description is supplemented by QM F24680 and QM F23443. M2 is triangular or nearly heart-shaped in occlusal outline, is dominated by a large trigon basin, and has an anteriorly broad but posteriorly narrow stylar shelf. The paracone is the tallest cusp and is relatively lingually located because of the broad stylar shelf. From the paracone a preparacrista runs anteriorly to the anterior margin of the tooth (stylar cusp B). From stylar cusp B a crest runs lingually along the anterior margin of the tooth and is eventually obscured by the extensive oval wear facet on the anterolingual margin of the protocone. A postprotocrista curves posterolaterally and connects with a small metaconule at the posterolingual corner of the tooth in QM F24680 (indiscernible in 58

QM F23443). From the metaconule, a postmetaconulecrista runs laterally to the posterior margin of the tooth where it joins the postmetacrista ascending from the metacone. The metacone is more posteriorly positioned than in M1. The lingual face of the metacone is worn and forms a near-vertical shearing facet. A lingual crest runs from the metacone to the metaconule. In QM F23443, this crest is just discernible on the left tooth and is barely evident on the more heavily worn right tooth. A narrow basin containing crenulations is formed between this lingual crest and the posterior edge of the tooth. From the paracone, a crescent-shaped crest runs lingually towards the protocone, ending at a notch about midway between the two cusps. Crenulations run down the anterior and posterior faces of this crest. A narrow anterior basin is formed between this crest and the anterior margin of the tooth. A longitudinal crest between the paracone and metacone is formed by the contiguous postparacrista and premetacrista. The meeting of these two crests is usually marked by a number of small grooves that fan out from this point on both the medial and lateral sides of the crest. On QM F24680 these grooves are particularly deep and well formed and practically cause a separation of the two crests. A stylar shelf, broader anteriorly than posteriorly, is present on the buccal flank. Both the anterior and posterior surfaces are convex and are connected by a narrow vertical valley that lies at the level of the junction of the postparacrista and premetacrista. The buccal surface adjacent to the paracone shows some variability: in the holotype and QM F24680 an anterobuccal ridge runs uninterrupted in an arc from the paracone to the end of the postparacrista, while on QM F23443, QM F20573 and AR 16653 this ridge is interrupted by a notch at its apex. A small vertical ridge is also present on the buccal surface adjacent to the metacone. This ridge runs anterodorsally from the metacone and meets the small ridges and grooves associated with the valley that lies on the buccal flank. The trigon basin is triangular in shape and well crenulated. M2 has three roots, the medial and posterior roots are subequal in size and the anterior root is the smallest.

M3: In QM F45200, moderately heavy wear obscures much of the detail on the crowns and greater detail can be seen on the left M3 of QM F23443. M3 is triangular in occlusal outline. The paracone and protocone are subequal in height and are taller than the metacone. A small crest runs lingually from the paracone, ascending and 59

A

B

C 35.3

3.1 5.7 7.5 8.5 9.9

11.7

22.5

98.2

Fig. 4.11 : Right dentary of Wakaleo hilmeri, paratype AR 19249; White Hunter Site. A. lingual view B. buccal view C. measurements in mm. 60 then descending toward the protocone. A very narrow anterior basin is formed between this crest and the anterior margin of the tooth. Wear facets are evident on the posterolingual surface of the paracone, on the protocone, and to a lesser extent on the metacone. The trigon basin is triangular and heavily crenulated. M3 is three- rooted, the medial root being the largest, and the posterior root is slightly larger than the anterior root.

M4: only the right M4 is preserved in QM F45200. The crown is triangular in occlusal outline and the occlusal surface is worn flat, forming a small circular basin. In lateral view, the protocone is the highest cusp and the metacone and paracone are subequal in height. M4 is three-rooted.

Dentary (Fig. 4.11 & Fig. 4.12)

Description of the dentary and lower dentition of W. hilmeri is primarily based on a left and right dentary (AR 19249) that were recovered from the same small block of limestone from White Hunter Site. Although it is difficult to achieve a close fit at their symphyseal joints due to the presence of epoxy glue, it is highly probable that the two jaws are from the same individual because of the high degree of similarity in size and morphology of the mandibles and their teeth. These two dentaries are the most complete of the five known dentary specimens and preserve P3, three of the four molars, and the posterior structures of the dentary. Both dentaries have suffered longitudinal fractures that pass through the horizontal and ascending rami. Fractures are also present across the posterobuccal corner of M3 of the left dentary, and in the lingual side of M1-2 and diagonally across M3 of the right dentary. Where appropriate, the descriptions will be supplemented by reference to the other dentaries from White Hunter site (QM F31376 and QM F30465). These dentaries both preserve I1, P3, M1-2 but exhibit extensive damage to their horizontal rami and are missing their postalveolar structures. An edentulous left dentary (QM F23443) was recovered along with the palate from the Rackham's Low Lion site and is attributed to the latter on the basis of its size, shape and length of the tooth row. This specimen preserves the dentary from the I1 alveolus to the condyle. Unfortunately its buccal 61 surface has been worn away removing all trace of the tooth crowns. The coronoid and angular processes are both badly damaged.

The right dentary of AR 19249 is the more complete of the pair. Its coronoid process, masseteric fossa and angular process are preserved intact. In dorsal view the dentary is bowed buccally and the tooth row forms a gentle arc from the P3, which is more laterally located, to the M4 which is more medially located. The dentary measures 98.2mm in length from the I1 alveolus to the articular condyle. The horizontal ramus has a depth of 22.5mm between M2 and M3. The apical region of the coronoid process bears a number of fractures and is displaced laterally. The depth of the ascending ramus, from the top of this process to the ventral margin, is 53.3mm. The anterior edge of the coronoid process forms an angle of 120o to the horizontal ramus. Its anterior margin is thickened, becoming thinner as it rises posteriorly. The last molar of the tooth row, M4, is located at the base of the ascending ramus.

The symphyseal area of articulation is large and roughly rectangular in shape. The anterior edge of the symphysis is missing. The symphyseal scar extends posteriorly to the level of the posterior root of P3. A small, deep sulcus lies at the lower 3-4mm of the posterior margin of the symphysis. The lingual surface of the horizontal ramus adjacent to the symphysis and below M1 is rounded and becomes flatter posteriorly. A shallow depression runs along the centre of the ramus from the level of M2 towards the medial pterygoid fossa. A thin crest (6mm long) lies at the base of the ascending ramus, 4mm posterior to the M4 alveolus, and is tilted dorso-posteriorly. The mandibular foramen lies 23mm posterior to M4 at the anterior edge of the deep depression of the pterygoid fossa. This foramen opens posteriorly, is vertically elongate and measures 3.6mm deep. The pterygoid fossa is very deep and pushes the postero-inferior part of the ramus, and therefore the masseteric fossa, laterally.

Posteriorly, the inflected angle expands lingually to form a well developed triangular shelf that extends posteriorly to approximately the posterior margin of the condyle. The medial border of the shelf is thickened, measuring 6mm at its thickest point. The 62

A

B

Fig. 4.12: Wakaleo hilmeri, AR 19249 paratype; White Hunter Site.

A. Paired right & left dentaries. Scale bar = 50mm. B. Stereo occlusal view of left dentary. 63 buccal surface of the horizontal ramus is gently rounded anteriorly and becomes more convex below M3/4, immediately anterior to the masseteric fossa. A large mental foramen lies ventral to the anterior root of P3. In QM F30465 this foramen is placed slightly more posteriorly, and is ventral to the posterior root of P3. A second, smaller, mental foramen is located approximately 8mm ventral to M1 and opens posteriorly. The masseteric fossa extends anteriorly to just behind the level of M4, although in QM F30465 it extends to just behind the level of M3. Anteriorly, the fossa is moderately deep while posteriorly, it forms a small depression anterior to the condyle. The margin of the fossa is strongly defined anteriorly by the prominent masseteric crest. Dorsally, the fossa is marked by the posterior curve of this crest towards the mandibular notch, and ventrally by its posterior extension towards the angular process. A small masseteric foramen lies at the inferior third of the fossa, 12mm posterior to the M4 alveolus. The posterobuccal portion of the angular process expands laterally to form a rounded process. The angular process is broad and has a width of 29.8mm.

The condyle is broad (18.1mm) and rounded. It lies at approximately the same height as the molar row and 41mm posterior to the M4 alveolus. In dorsal view the condyle has a sinusoidal outline and is slightly thicker on the lingual side (7.4mm) than the buccal side (6.0mm).

Lower Dentition (Figs. 4.11 & 4.12, Table 4.2)

Description of the lower dentition is primarily based on the paired dentaries AR 19249. These dentaries preserve P3 and M1-3, and alveoli for I1, P1-2, and M4. Description of I1 is based on QM F31376. Variations in morphology in other specimens are noted where appropriate.

I1: is caniniform and procumbent. The apical portion of the crown is completely covered in enamel. The medial tip of the incisor bears a small ridge that descends for a short distance before splitting into distinct superior and inferior ridges. Between these ridges the interincisive surface forms a trough. The dorsal surface of the incisor 64 is divided into two areas by a low crest that runs along the medial third of the tooth from the tip to the alveolus. The narrower dorsomedial surface is concave and the broader dorsolateral surface is convex. The proximal portions of the incisors in both QM F31376 and QM F30465 are badly damaged. The left dentary of AR 19249 and QM F30465 indicate that root of I1 runs deeply to below P3.

Premolars: the region anterior to P3 bears two small alveoli. P1 alveolus: The anterior-most premolariform alveolus is elongate and lies approximately 5mm posterior to the I1 alveolus. It has a width of 1.4mm and opens anteriorly into a longitudinal depression that is contiguous with the alveolus suggesting that the tooth projected in a horizontal direction rather than directly vertically.

P2 alvelous: In AR 19249 a second premolar alveolus lies approximately 1mm anterior to the lingual base of P3. It is round and measures 1.8mm long and 1.4mm wide. It is relatively shallow and probably held a small, single-rooted, button-like P2. In QM F30465 this alveolus lies directly against the base of P3.

P3: P3 is blade-like and similar in length to M1. As in other species of Wakaleo, the crown slopes backwards and consists of a tall anterior cusp and a shorter posterior cusp joined by a buccally-curving longitudinal blade. An anterolingual crest descends from the anterior cusp, curving slightly posteriorly as it nears the base of the crown. A narrow V-shaped valley descends from the anterior cusp to the base of the crown between the anterior and anterolingual crests. A short posterolingual crest descends in a lingual direction from the posterior cusp, curving slightly anteriorly as it nears the base of the crown. Between the antero- and posterolingual crests the surface of the tooth is concave with its area of greatest curvature at the posterior third. The buccally surface is gently convex. A weak crest descends for a short distance from the anterior cusp and merges with the crown about half way down the tooth. The anterior root supports the anterior third of the crown and the larger posterior root supports the remaining two-thirds. 65

M1: is similar to those of other species of Wakaleo. Anteriorly, a tall trigonid bears an apical cusp. An anterior crest descends from this cusp to the anterior margin of the tooth and is contiguous with the longitudinal blade of P3. M1 abuts and interlocks P3, the anterolingual margin forming a slight lip that follows the curve of the posterior end of P3. On the buccal side of the trigonid an obtuse crest descends from the apex and merges with the crown about two-thirds of the way down the tooth. At the base of this crest, above the anterior root, and also above the posterior root, the buccal surface of the crown is slightly inflated. The lingual surface of the trigonid is rounded and bears a short, weak crest that descends from the apex. The talonid basin is broad, oval in outline and lined with weak ridges. The posterior margin of the talonid basin is formed by two crests; a short steep crest runs medially from the entoconid and connects with a longer and more gently sloping crest that runs laterally from the hypoconid. In the unworn specimen QM F31376, ridges similar to those lining the talonid basin descend towards each other from the hypoconid and entoconid leaving a notch at the back of the talonid. A posthypocristid runs medially across the back of the tooth terminating ventrally and posterior to the entoconid. In QM F30465 wear and/or damage at the back of the tooth especially at the hypoconid produces a lower posterobuccal margin and the posthypocristid is very truncated. This lower posterior margin results in a less prominent notch in the posterior edge of the talonid. The posterior face of the crown is relatively flat. The molar has two roots; the posterior root is slightly larger than the anterior.

M2: is rectangular in outline and consists of two basins; anteriorly, the crown is tall and bears a small subsquare trigonid basin, while posteriorly there is a large rectangular talonid basin. The metaconid is slightly taller than the protoconid and the entoconid and hypoconid are weakly developed, less tall and subequal in height. The buccal side of the tooth is rounder and broader than the lingual side. On both lingual and buccal sides there is a slight constriction midcrown. The lingual edge of the talonid basin is formed by two crests, the postmetacristid and the preentocristid, that meet approximately midway along the tooth. The buccal edge of the talonid basin is formed by a postprotocristid that descends posteriorly from the protoconid and connects with a relatively horizontal prehypocristid that curves medially toward the 66 posterior margin of the tooth. The talonid basin is lined with weak ridges. M2 has two roots of approximately similar size.

M3: is subrectangular in outline. The sides of the crown are gently rounded and the occlusal surface bears a shallow, lightly-crenulated basin. The anterior margin of the tooth is slightly taller than the posterior margin. There are two roots of approximately equal size.

M4: the crown is not preserved. The alveolus is small and slightly oval in shape. There is no evidence of a septum indicating the presence of two roots on either the left or right dentary. This suggests either that the two roots were so compacted that there was no room for a septum to develop, or possibly the tooth had a single root. In either case, the tooth was only weakly anchored.

Comparisons

Comparisons were made with the following cranial specimens: Wakaleo pitikantensis (SAM P37719, holotype), L. roskellyae (QM F23453, holotype; QM F23442), W. vanderleueri (CPC 26604), W. alcootaensis (maxillary fragment, NTM P1, holotype), Thylacoleo crassidentatus (QM F10622, Archer and Dawson 1982, Fig. 4) and T. carnifex (AM F4659). Comparative measurements of the crania of thylacoleonid taxa are shown in Table E (Appendix 1). Dentary and dental comparisons were also made with the following additional specimens: L. roskellyae (QM F29623, QM F20191), W. oldfieldi (SAM P17925, QM F20895, maxillary fragment QM F23428), and W. vanderleueri (NTM P85553, QM F36453, QM F41204) Thylacoleo carnifex (AM F62525, AM F734), T. crassidentatus (QM F7762, QM F10623, AR1400), T. hilli (SAM P17621, cast) and Thylacoleo sp. cf. T. hilli (AM F63584). Comparative palatal and dentary measures of thylacoleonids are shown in Table F and Table G, in Appendix 1.

The cranium of W. hilmeri shares features characteristic of W. vanderleueri but also exhibits others only seen in Lekaneleo roskellyae. In general appearance, the skull of W. hilmeri is most similar to that of W. vanderleueri. Like the latter, the skull is 67 robust and broad, has a slight anteroposterior flexure, prominent supraorbital processes, tall sagittal and nuchal crests, and has a similar basicranial morphology. The supraorbital processes differ from those of W. vanderleueri in being relatively shorter and shallower and, as in L. roskellyae, are more anteriorly located. The rostrum differs slightly from W. vanderleueri in being more laterally rounded and lacks the anteorbital depression seen in that species. The region of the dorsum of the skull shows the greatest contrast with W. vanderleueri; it lacks the prominent dorsal inflation of the frontals which contributes to its relatively shallower skull depth. This region of the skull shows more similarities to L. roskellyae, exhibiting a distinct dorsal depression of the frontals that is laterally bordered by strong ridges. The cranium exhibits a postorbital constriction which is also seen in W. vanderleueri and L. roskellyae. The region of the neurocranium shares most similarities with that of W. vanderleueri. These two taxa have tall sagittal and nuchal crests that descend steeply onto the temporal fossae and the parietals are predominantly concave, becoming convex only at their lateral extremity. In contrast, the neurocranium of L. roskellyae lacks these tall crests and its parietals are predominantly convex.

The morphology of the occipital region is similar to those of W. vanderleueri and T. carnifex. It is similar in absolute width to W. vanderleueri but is not as tall (Table E; Appendix 1). The deep posterior concavity of the central region of the nuchal crest is also seen in T. carnifex but is extremely weak in W. vanderleueri and practically absent in L. roskellyae. The lateral occipital fossae of W. hilmeri are much deeper than those of other species of marsupial lions.

The zygomatic arch of W. hilmeri projects laterally dorsal to the posterior root of P3, M1 and the anterior root of M2. This is most similar to L. roskellyae. The larger species, W. vanderleueri and T. carnifex, are somewhat similar but their larger teeth result in a posterior shift of the M2 (or loss in the latter species) relative to the zygomatic arch. Rauscher (1987) indicated that in W. pitikantensis the anterior root of the zygomatic arch projects anterolaterally dorsal to M1-2, the position also seen in L. roskellyae. However, most of the anterior portion of the zygomatic arch 68

(including the area dorsal to P3) of W. pitikantensis is missing and thus this specimen does not provide conclusive evidence of its orientation in this species.

Unlike W. vanderleueri, both W. hilmeri and L. roskellyae lack a masseteric process on the zygomatic arch. In comparison to W. vanderleueri, the maxillary contribution to the arch is shallower and thinner in W. hilmeri. In addition, the maxillojugal suture is more vertically oriented than in W. vanderleueri. In the latter, the malar tuberosity forms an angle of 45o to the vertical (Murray et al. 1987) and the anterior portion of the jugal extends anterodorsally towards the anterior margin of the lachrymal foramen, forming a more oblique suture. The anterior infraorbital region also lacks the strong depression seen in this region in W. vanderleueri.

The palate of W. hilmeri is proportionately shorter than that of W. vanderleueri but otherwise shows a strong resemblance to the latter, especially in shape and depth (Table F; Appendix 1). Both species exhibit a narrow but deep anterior arch which widens posteriorly, achieving a maximum width and relatively deep palate at the level of P3/M1. Additionally, both taxa exhibit palatal ridges.

Basicranium The basicranial region of W. hilmeri is extremely similar in overall appearance and size to that of W. vanderleueri. Both taxa have a relatively open tympanic region which is partially floored by a moderately sized tympanic wing that does not reach the mastoid process. Both taxa have a tympanic wing that consists mainly of the alisphenoid overlying a smaller squamosal contribution. In W. hilmeri the squamosal contribution is anterior and anterolateral. In W. vanderleueri the squamosal contribution is also anterior, however any lateral contribution from this element is difficult to discern in the holotype. The tympanic wings of these two taxa do differ slightly in shape and size, that of W. hilmeri being flatter and shorter, and with a much greater distance separating it from the medial edge of the mastoid process. Both taxa have similarly shaped mastoid and paroccipital processes and differ from T. carnifex which has a better developed paroccipital. 69

The region of the pterygoid fossa, in particular the region of attachment of the inferior head of the lateral pterygoid, is similar to that seen in W. vanderleueri and T. carnifex but differs from that of L. roskellyae. This area in Wakaleo is a shallow, triangular fossa that has a slightly deeper circular fossa at its anteromedial margin, directly anterior to the foramen for the transverse canal. In contrast, in L. roskellyae this region, although also triangular, is flattened anteriorly and is weakly rounded or slightly inflated posteriorly. In addition, L. roskellyae differs from W. hilmeri and W. vanderleueri in appearing to lack a transverse canal.

The region of the postglenoid process in W. hilmeri exhibits a number of differences when compared to other thylacoleonid taxa. The morphology of the postglenoid process is most similar to that of W. vanderleueri but in that species the process is transversely narrower and its posterior surface is more rounded. In contrast, the posterior surface in W. hilmeri bears a shallow sulcus that extends onto its surface from the lateral margin of the postglenoid cavity. In addition, W. hilmeri lacks the distinct ventral process at the posterior end of the jugal that is present in W. vanderleueri. The medial glenoid process of W. hilmeri is more prominent and more laterally positioned than that of W. vanderleueri. The postglenoid cavity in W. hilmeri also differs from that of W. vanderleueri in lacking the horizontal septum seen in the latter. In distinct contrast to both species of Wakaleo, the postglenoid cavity in L. roskellyae extends laterally into the postglenoid process.

Detailed comparison of the periotic of W. hilmeri and W. vanderleueri was limited because of its displacement and the presence of matrix around the bone in the latter comparative specimen (CPC 26604). However, the periotics in both taxa appear to be quite similar in morphology with the most striking shared feature being the presence of a prominent rostral tympanic process. In contrast, the rostral tympanic process on the periotic of L. roskellyae is extremely weak and is only a gently rounded structure. The promontorium of the L. roskellyae periotic also differs in bearing a strong angular ridge at about the midpoint of the surface whereas W. hilmeri has a slight ridge 2mm from its anterior edge. 70

Upper Dentition In contrast to cranial features, W. hilmeri shares many dental features with W. pitikantensis and L. roskellyae. These include: (1) presence of three premolars (unlike W. vanderleueri which has 1 or 2); (2) presence of four molars; (3) similar position of the anterior premolars relative to P3; (4) similar size of P1 and P2 alveoli relative to each other (P1 is larger than P2); (5) similar proportional length of P3 relative to rest of the molar row (based on molar and alveolar measurements); and (6) a crenulated trigon basin on M2.

The positions and relative sizes of the incisor and canine alveoli in W. hilmeri are similar to those in both L. roskellyae and W. vanderleueri, although the canine alveoli are proportionally smaller than those of the latter. In W. hilmeri and L. roskellyae the canine alveoli are approximately level with the posterior end of the anterior palatal fenestrae, whereas in W. vanderleueri the alveoli extend much further posteriorly. The canines of W. hilmeri are relatively large and are similar to those of W. vanderleueri (see Chapter 7, 7.3) but differ to those of Thylacoleo carnifex which are more peg-like. A corollary of the smaller size of the canines in W. hilmeri is the relatively longer diastema between the canine and the anterior premolar in comparison to W. vanderleueri.

The anterior premolar number appears to be variable in W. hilmeri. The palatal specimen QM F23443 is similar to Lekaneleo roskellyae in having two anterior premolars with the first premolar being larger than the second, and in having a similar premolar location, i.e. P1 approximately halfway between the canine and P3, and P2 at the base of P3. However, the holotype QM F45200 lacks P1 and only bears an alveolus for P2 at the base of P3. The maxillary fragment QM F24680 further complicates this situation by appearing to lack any anterior premolars, possessing two pits along its diastema which more closely resemble foramina than true alveoli. These two latter specimens show a pattern similar to that seen in Wakaleo vanderleueri which has either a single anterior premolar or none at all, although the location of this premolar when present is more anterior (two-thirds of the way between the canine and P3) than in W. hilmeri. 71

The P3 of W. hilmeri is approximately 1.5 times the length of P3 in L. roskellyae, is two-thirds the length of W. vanderleueri, and half the length of W. alcootaensis. Despite similar palatal dimensions in W. hilmeri and W. vanderleueri, the relative size of P3 to M1 in W. hilmeri is more similar to the proportions seen in L. roskellyae (Table F; Appendix 1). Wakaleo vanderleueri has a longer P3 relative to M1. The general morphology of P3 is similar to those of L. roskellyae, Wakaleo oldfieldi and W. vanderleueri. On the buccal surface, the anterior buccal crest in W. vanderleueri is straighter and shorter and the posterior crest is much more prominent than in W. hilmeri. The roots of P3 are relatively similar in size to those of W. oldfieldi but relatively larger than those of L. roskellyae and not as massive as those of W. vanderleueri.

The molars of W. hilmeri are similar in size to those of W. pitikantensis, although the dimensions of the latter are based mainly on alveolar measurements. In general, the molar morphology of W. hilmeri is similar to that of L. roskellyae and W. vanderleueri. The most striking dental feature in W. hilmeri is its possession of four molars, a trait shared with Microleo attenboroughi (Chapter 10), W. pitikantensis and L. roskellyae. In contrast, the other species of Wakaleo exhibit a loss of either one or two posterior molars. Loss of posterior molars is also a characteristic of species of Thylacoleo.

The occlusal outlines of M1and M2 are similar to those of L. roskellyae and these molars have a relatively squarer lingual margin compared to the triangular forms of W. oldfieldi and W. vanderleueri. The squarer shape is the result of greater development of the posterolingual corner of the tooth and is probably related to the slightly stronger development of the metaconule. Although the metaconule is quite worn in many specimens of W. hilmeri, it is more strongly developed than in W. oldfieldi and W. vanderleueri, but is not as prominent as in L. roskellyae. The metaconule of M1 in W. oldfieldi and W. vanderleueri is extremely reduced and in combination with the relatively large protocone contributes to the triangular shape of their molars. Wakaleo hilmeri is also similar to W. vanderleueri in having the molar 72 crowns supported on tall, exposed roots. The triangular outline of M3-4 is the shape common to other species of Wakaleo. The decreasing molar gradient of W. hilmeri is more moderate than that of W. vanderleueri and this is reflected in the comparatively greater size of its M3 relative to M2.

Lower Dentition In general, the morphology of the dentary and lower dentition of W. hilmeri is similar to that of W. oldfieldi and W. vanderleueri, and differs mainly in its smaller size and presence of an extra premolar and a fourth molar. Its smaller dental dimensions are reflected in P3 length which is generally three-quarters the length of W. oldfieldi and only two-thirds the length of W. vanderleueri. Its smaller size is reflected also in the length of the cheektooth row (P3-M4) which, although containing an extra molar, is less than those of the other two Wakaleo species (Table G; Appendix 1).

Wakaleo hilmeri possesses a procumbent incisor similar to those seen in W. oldfieldi and W. vanderleueri but is slightly more gracile. The medial surface of the incisor also differs slightly from those of the latter species in bearing a longitudinal groove which is enclosed by superior and inferior ridges. In W. oldfieldi and W. vanderleueri, the medial surface is flatter, probably the result of wear between the incisors. In addition, the buccal flange of the W. hilmeri incisor is slightly less prominent than that of W. oldfieldi.

Wakaleo hilmeri differs from W. oldfieldi and W. vanderleueri in bearing two alveoli between I1 and P3. The anterior-most of these alveoli is similar in its shape and location to the single alveolus that is present in those two species. The location of the second alveolus (i.e. at the lingual base of P3) is also seen in other thylacoleonids that possess more than one alveolus anterior to P3 (i.e. species of Thylacoleo and L. roskellyae).

The general morphology of P3 is essentially similar to those of W. oldfieldi, W. vanderleueri and L. roskellyae. The morphology of the anterolingual region is more similar to that of L. roskellyae and differs from the two larger Wakaleo species in 73 having a relatively more sharply defined anterolingual crest and in having a relatively concave surface between this crest and the anterior crest, rather than a broadly convex one. The anterolingual crest is also slightly more vertical and anteriorly located than in W. oldfieldi and W. vanderleueri, descending towards the posterior third of the anterior root rather than curving toward the posterior margin of the root as in those species.

The morphology of M1 is similar to those of other Wakaleo species, the major differences being its smaller size and the structure of the posterior edge of the talonid basin. It approaches the length of W. oldfieldi but is distinctly shorter than that of W. vanderleueri. Although the molar is proportionally narrower, its talonid basin is similar in absolute width to those of the larger species. Unlike W. oldfieldi and W. vanderleueri, which bear a very narrow notch at the posterior edge of the talonid basin, the notch in W. hilmeri is relatively broad. M2 is similar in shape to that of W. oldfieldi and W. vanderleueri, however, it is slightly smaller in size and has a proportionally broader trigonid basin at the front of the tooth and a talonid basin which is proportionally and absolutely broader than those of the latter species.

Although sharing some dental similarities with L. roskellyae, W. hilmeri also exhibits some distinct differences. Firstly, it is much larger, P3 being longer by a factor of 40- 50%. Secondly, the talonid basins of the molars of W. hilmeri are relatively narrower and much deeper. Thirdly, M1 lacks the transverse crest that runs from the entoconid towards the hypoconid in L. roskellyae. Consequently, it also lacks the small basin that lies at the posterolingual corner of the talonid basin in L. roskellyae. Fourthly, the metaconid and protoconid on M2 of W. hilmeri are relatively taller, the protoconid is in a more lingual position, and the trigonid basin is more elevated and narrower.

Discussion 74

The diagnostic dental characters for species of Priscileo were initially considered to be the presence of M4, P3/M1 length ratio less than 1.70, and a crenulated trigon basin on M2 (Rauscher 1987). Following the recognition of a plesiomorphic species of Wakaleo from Riversleigh that shared the same dental formula as species of Priscileo, Gillespie (1997) amended the diagnosis to P3 less than 12mm long, M1 and M2 square in basal outline with a posterolingual metaconule and the zygomatic arch projecting anterodorsally to M 2-3. Wakaleo hilmeri exhibits all these dental characteristics except the last, its M2 being more subtriangular in outline. An additional feature shared with species of Priscileo is the presence of three upper premolars. In contrast to these dental similarities, W. hilmeri exhibits a neurocranium and basicranial morphology that is significantly different from that of Lekaneleo roskellyae and one that is unequivocally characteristic of species of Wakaleo. The cranial features shared with Wakaleo vanderleueri include large sagittal and nuchal crests, well developed supraorbital processes, lateral palato-pterygoid processes, shallow and concave pterygoid fossa, a medially-confined postglenoid cavity that does not invade the postglenoid process, deep fossa below the infratemporal crest, relatively deep central fossa on the occiput, a dorsal and oblique subsquamosal foramen, and a large rostral tympanic process on the periotic. All these features are synapomorphies for species of Wakaleo and hence support the referral of the Riversleigh material to this genus.

Previously, loss of the anterior upper and lower premolars has been identified as a diagnostic feature of the genus Wakaleo (Archer and Rich 1982, Murray et al. 1987), however the presence of these teeth in W. hilmeri and W. pitikantensis suggests that this character is symplesiomorphic within thylacoleonids and cannot be used to separate the genera Wakaleo and Lekaneleo. The diagnostic utility of this condition is still retained for assessing relationships within Wakaleo and suggests that W. hilmeri and W. pitikantensis are plesiomorphic to the remaining species of Wakaleo. The two other diagnostic features of Lekaneleo that are shared with W. hilmeri (four molars and a square lingual base of M1) also appear to be symplesiomorphic characters of marsupial lions and similarly lose their diagnostic utility (see Chapter 11). 75

Premolar identification in Wakaleo The Riversleigh specimens indicate that the upper premolar number is variable intraspecifically, some specimens exhibiting the full complement of P1-3 while others fail to develop P1 and/or P2. The reasons for such variability are unclear. In QM F24680 the alveoli for the anterior premolars appear malformed. In this specimen the region between the canine alveolus and the P3 alveolus is comparatively short. In addition, there is a small accessory canal which lies between the infraorbital foramen and the base of anterior root of P3 that opens internally into the infraorbital canal. Studies of tooth abnormalities within marsupials indicate that missing teeth mostly occur in the region of the premolars (Archer 1975) and thus the appearance of QM F24680 may be an example of abnormal development.

Wakaleo vanderleueri also shows similar variation in the presence of its anterior premolar which is usually identified as P2. It is interesting to note that in species of Lekaneleo and Thylacoleo, and in W. hilmeri, P1 is located approximately midway between the canine and P3 and P2 abuts the base of P3. In the cranial specimen of W. vanderleueri (CPC 26604), Murray et al. (1987) identified the sole premolar alveolus, which lies anterior to P3, as that of P2. This alveolus lies nearly midway along the diastema between the canine and P3. Considering the pattern shown by other thylacoleonids, it seems more likely that this alveolus corresponds to P1 and that P2, normally positioned at the base of P3, has been lost as a result of the hypertrophy of P3.

The complete dental formula for W. hilmeri is: I1-3/1 C1/0 P1-3/1-3 M1-4/1-4 , however, as indicated by Archer and Dawson (1982), the homology of the single- rooted teeth that lie between the procumbent I1 and P3 in thylacoleonids is open to speculation; they may represent posterior incisors and a canine, or a canine and anterior premolars. In species of Wakaleo, the anterior-most alveolus with its anteriorly-directed opening is similar to that seen in the mandibles of many diprotodontian taxa and is probably related to the presence of the root of the enlarged 76

I1 that characterises the order. In W. hilmeri, the close proximity of the second alveolus to P3 suggests that it probably corresponds to P2.

Within the Thylacoleonidae (i.e. Wakaleo and Thylacoleo), the general trend through time has been one of increasing P3 length while simultaneously reducing the number of molars. This trend reflects the increasing emphasis on the shearing function of P3, indicating a trend to a more carnivorous habit. The dentitions exhibited by W. hilmeri and W. pitikantensis (i.e. smaller P3 and presence of M4) suggest that these species placed less emphasis on the cutting of food items and the possibility of a more omnivorous diet. 77

CHAPTER 5

POSTCRANIAL BONES OF Wakaleo hilmeri sp. nov.

5.1 Introduction

The postcranial skeleton of most species of thylacoleonids is poorly known. Only the bones of the Pleistocene species Thylacoleo carnifex have been found in any number or articulation to enable further detailed study: the almost complete skeletons of an adult and juvenile were recovered from Moree in New South Wales. From this material studies have been made of the functional morphology of both the vertebral column and the limbs of this species (Finch 1982, Finch and Freedman 1988). In addition to these specimens, an articulated right manus and left pes of T. carnifex from Victoria Cave, Naracoorte World Heritage Area in South Australia has been described (Wells and Nichol 1977).

Very few postcranial bones have been recovered or described for other marsupial lion taxa. Postcranial elements of Wakaleo pitikantensis from Lake Pitikanta in South Australia have been described by Rauscher (1987), including a proximal right scapula, left humerus, right radius, distal left radius, four rib fragments, a left capitatum (magnum), left ectocuneiform, a phalanx and a crushed right humerus. The only element described for Wakaleo vanderleueri is a distal radius recovered from the Bullock Creek Local Fauna in the Northern Territory (Murray and Megirian 1990). Foot elements probably belonging to Wakaleo alcootaensis have been recovered from the Alcoota Local Fauna but are yet to be described (Murray and Megirian 1990).

After the initial discovery of the marsupial lion palate (Fig. 4.9, Chapter 4) from the Rackham's Low Lion Site (QM F23443) further processing of a small amount of limestone from this site yielded a number of postcranial elements which on comparison to similar elements of other thylacoleonids and other marsupial groups 78 indicated they were indeed those of a marsupial lion. Another factor supporting the association of this material is the size of the elements; all the elements appear to be from a medium sized mammal and are of a size appropriate for an animal with a palate of the size originally found at the site. Additionally, besides some crocodilian elements, no other taxa have been recovered from the site. The Wakaleo postcranial material recovered from this site includes a cervical vertebra, two lumbar vertebrae, sacrum, fragments of both ilia, distal right humerus, distal left radius, left hamatum, left trapezium, left first metacarpal, proximal phalanx of the left pollex, right proximal phalanx and medial phalanx of digit II, and the left calcaneum. That material is described in this chapter.

Rauscher (1983) assessed the functional morphology of the humerus and radius of W. pitikantensis and concluded that the species may have been fossorial. The recovery from Riversleigh of a humerus of W. hilmeri provides an opportunity to compare the form and functional morphology of this element with that of W. pitikantensis. Considering the possible conspecific associations of these specimens, study of the osteology and functional morphology of this element may aid in determining the relationships of these specimens.

A numerical parsimony analysis of the postcranials was not attempted due to the limited nature of the skeletal material available. However, the elements were compared against those of other thylacoleonid taxa where available, and against other extinct and extant marsupial taxa from a range of families within the Order Diprotodontia Owen, 1866 with the aim of identifying morphological similarities that may assist in providing insights into the phylogenetic relationships of the family. The comparative taxa and specimens are listed in Appendix 2.

For ease of reading, within the text comparative taxa are referred to by their generic names (for example, Thylacoleo for T. carnifex, Nimbadon for N. lavarackorum and so on) except for species of Trichosurus where T. vulpecula fuliginosus and T. caninus are referred to by their specific titles and T. vulpecula carries the generic nomen. It should be noted that at no stage does reference to the generic name imply referral to all members of the genus, only those members listed in Appendix 2. 79

An attempt was made to look at more than one individual of a species, however, in some cases skeletons were missing the relevant bones and only elements from a single specimen were available. In the case of fossil taxa, often only a single element or a fragmentary element was available, for example, the only relevant postcranial material available for Wynyardia bassiana was a cast of the ilia. Original fossil material of Ngapakaldia tedfordi was unobtainable and again only casts were available, and where casts of certain elements were unavailable, comparison was made from descriptions in Munson (1992).

While undertaking comparative studies of thylacoleonid postcranials I discovered a previously unidentified part of the scapula of the holotype of Wakaleo pitikantensis Rauscher, 1987 in a collection of bone fragments collected with the holoytpe. An amended description of this element is included here. Functional assessment of the postcrania of W. hilmeri and W. pitikantensis is given in Chapter 13.

Prefixes correspond to the following: AM F, AM M, AM P, or AM S = Australian Museum, Sydney; AR = specimens temporarily housed in the collection of Michael Archer, University of New South Wales; CSIRO CM = CSIRO mammal collection, Canberra; NTM = Northern Territory Museum, Darwin; QM F = Queensland Museum, Brisbane; QVM = Queen Victoria Museum, Launceston; SAM or SAM P = South Australian Museum, Adelaide; UNSWZ = University of New South Wales Zoology Department, Sydney; WAM = Western Australian Museum, Perth.

Terminology The bones are described on the basis that the animal is postured in a plantigrade, four-footed position. In this position the back of the hand and foot is dorsal and the palmar surfaces are ventral. The anatomical directions of anterior and posterior equate to cranial and caudal, respectively. In descriptions of the limb elements anterior also refers to extensor muscle surfaces and posterior refers to flexor muscle surfaces. The terms distal and proximal are used relative to the distance from the vertebral column. Anatomical terms are based on those from the Nomina Anatomica Veterinaria (1992), however, I have replaced the Latin with the English equivalent as used in Miller, Christensen and Evans (1964). 80

A B

Fig. 5.1: Wakaleo hilmeri QM F52183, cervical vertebra from Rackham’s Low Lion Site. Fig. 5.2: Wakaleo hilmeri QM F52184, A. anterior view B. lateral view lumbar vertebra (L6), anterior view.

A A

B B

L6 L5

Fig. 5.3: Wakaleo hilmeri lumbar vertebrae, < posterior anterior >

QM F52185 (L5) and QM F52184 (L6). Fig. 5.4: Wakaleo hilmeri QM F52186, sacrum. A. lateral view B. dorsal view A. lateral view B. dorsal view 81

5.2 DESCRIPTIONS AND COMPARISONS

5.2.1 VERTEBRAL COLUMN

5.2.2 Cervical vertebra (Fig. 5.1)

One cervical vertebra is preserved. The vertebra is lacking its lateral margins including most of the transverse processes and the edges of the anterior and posterior articular surfaces. The tip of the spinous process is also missing. Based on comparisons with the cervical sequence in other marsupial taxa this vertebra possibly represents the fifth vertebra (C5).

The vertebral body is dorsoventrally compressed and is rectangular in outline. Anteriorly, the body is 18.8mm wide and 10.0mm high. The anterior surface of the body is missing some of its epiphyseal plate and has an undulating surface, being convex centrally and concave bilaterally. Posteriorly, the body is 18.1mm wide and 10.1mm high and has a rounded dorsal margin. The ventral surface of the body is flattened and measures 10mm anteroposteriorly. The vertebral arch is gently rounded and the vertebral foramen has a height of 6.3mm and a width of 14.0mm. The pedicles are short and the right pedicle is pierced by a large nutrient foramen. In dorsal view, the arch is positioned predominantly over the posterior half of the vertebral body. The lamina of the arch is moderately deep measuring 8.7mm anteroposteriorly. The cranial margin of the lamina is gently concave and the caudal margin is straight. A large intervertebral foramen is present in the lamina on the left side and a number of small foramina are present on the right. The remains of the cranial articular processes indicate their facets face craniodorsally. The caudal articular processes are more complete. The facet of the more intact left process has a width of 6.8mm. The processes face caudoventrally and are angled 35o to the horizontal. The transverse processes are poorly preserved. The more complete left process has a basal length of 12.5mm. A pair of transverse foramina lies at the base of the pedicles. These foramina are round (3mm diameter) and would have carried 82 the vertebral artery. The basal 5mm of the spinous process is preserved. This process is directed caudally and forms an angle of 45o to the vertebral body.

Comparisons The W. hilmeri cervical vertebra is most similar (see Appendix 3 for list of comparative taxa) to those of Thylacoleo and the posterior cervical vertebrae of Trichosurus although it is relatively smaller than the vertebrae of the former and much larger than those of the latter. Although the Thylacoleo vertebrae exhibit taller and more dome-shaped vertebral foramina, they display similarities in the shape of their centra, the posterior positioning of the arch relative to the centrum, the posteriorly-inclined spines, the anterior protrusion of the edges of the anterior surface of the centrum, and the angle of orientation of the articular facets.

The similarities shown by the Trichosurus vertebrae include dorsoventral compression of the vertebral foramen resulting in a relatively low, oval canal, an arch with a moderately anteroposteriorly-deep lamina and a posteriorly-inclined spinous process. Munson (1992) indicates an overall similarity between a vertebra of Ngapakaldia and those of Trichosurus which therefore suggests a possible similarity between the vertebra of W. hilmeri and those of Ngapakaldia.

In contrast, the posterior cervical vertebrae of Vombatus, Nimbadon, Phascolarctos, and Sarcophilus have vertebral foramina that are relatively much deeper and are triangular (Vombatus and Nimbadon), or round (Phascolarctos and Sarcophilus) in shape. The cervical vertebrae of Vombatus and Phascolarctos also differ greatly from the W. hilmeri vertebra in having extremely short (anteroposteriorly) vertebral bodies and laminae.

5.2.3 Lumbar vertebrae (Fig. 5.2 & 5.3)

Two near-complete articulating lumbar vertebrae are preserved. The posterior articular processes of a third vertebra are attached to the most anterior of these vertebrae. A number of features strongly suggest that these vertebrae represent the last two in the lumbar sequence. Firstly, the centra of the two vertebrae are large and 83 are similar in size to that of the first sacral vertebra. Secondly, the caudal facets of the most posterior vertebra articulate well with the cranial facets of the first sacral vertebra. Thirdly, both vertebrae have large transverse processes and those of the posterior vertebra are more anteriorly-curved than those of its anterior neighbour, a characteristic also seen in the last lumbar vertebrae of Vombatus, Dasyurus and Sarcophilus. Fourthly, the posterior accessory processes on the posterior vertebra are smaller than those of the anterior vertebra, a characteristic also seen in the last two lumbar vertebrae of most marsupial species.

The skeleton of Thylacoleo carnifex from Moree in New South Wales is reported to possess seven lumbar vertebrae (Finch 1981, Finch and Freedman 1986), however, it is not unusual to find individual animals with numbers of vertebrae that vary from the characteristic species' number. Usually there is a loss or gain of a vertebra (Davis 1964). Dublin (1903) indicated that the number of thoracolumbar vertebrae in marsupials is usually nineteen, i.e. thirteen thoracic and six lumbar. Observation of the marsupial skeletons used in this comparative study supports this finding and therefore I propose that the lumbar vertebrae recovered represent the fifth (L5) and sixth (L6). (Alternatively, assuming that W. hilmeri possesses the same number of lumbar vertebrae as Thylacoleo carnifex (L1 - L7), these vertebrae possibly represent L6 and L7.)

Both vertebrae exhibit minor damage. The anterior and posterior surfaces of both centra are eroded and the apices of the neural spines are missing. On L5 the left transverse process is missing and the left anterior and posterior articular processes are damaged. The posterior articulating facets and base of the neural spine of L4 are attached to the anterior articular facets of L5. The anterior articular facets of L6 are damaged and the right transverse process has a fracture at its base and is ventrally displaced.

L5 The centrum of L5 is large and deep. Its anterior articular surface forms a rounded square in outline and is slightly wider (22.4mm) than it is deep (19.9mm). The posterior articular surface is more oval in shape, has a maximum width of 26.1mm 84 and height of 20.7mm. Between the two articular surfaces the body of the vertebra narrows creating a concave surface on its lateral and ventral surfaces. Two large foramina are located on the ventral surface, one each side of the slight midline ridge that runs along the length of the vertebra. The anterior articular processes are tall and extend anterodorsally from the neural arch forming an angle of 50o to the horizontal axis of the centrum. The anterior articular facets are directed medially and articulate with the posterior articular processes of L4. A large foramen is located at the posterior margin of each anterior articular process. The right posterior articular process is intact, the left is damaged. The posterior articular facets are small, round, convex and directed ventrolaterally, articulating with the anterior articular facets of the second vertebra. The intact right transverse process is dorsoventrally compressed and curves anteriorly in the horizontal plane. The base of a small spine is located on the posterior margin of the transverse process 10mm from the vertebral arch. Only about 15mm of the base of the neural spine is preserved. The neural spine is laterally compressed. In lateral view its base is 16.0mm long and tilts anteriorly at an angle of approximately 70o to the horizontal axis. The posterior margins of the pedicles bear moderate accessory processes. The right process is approximately 3mm wide and 3mm long (its tip is lost) and only the base of the left process is preserved. The accessory processes project posteriorly and lie laterally to the anterior articular processes of L6.

L6 L6 is similar in morphology to L5. The centrum is deep and its anterior surface is relatively square in outline, becoming narrower ventrally. The anterior surface has a maximum width of 25.1mm and is 20.5mm in height. The posterior surface of the vertebral body is oval, its width (27.3mm) being greater than its height (20.3mm). The vertebral body has a constricted waist similar to that of L5. A large foramen is located on the ventral surface of the body, on the left side of a slight central ridge. In profile, the anterior articular processes of this vertebra also form an angle of 50o with the horizontal axis of the centrum. The articular facets of the anterior articular processes are slightly damaged; the dorsal margin of the right facet and the dorsal half of the left are missing. The facets are tall, concave and directed medially at an angle of approximately 60o. The posterior articular facets are smaller, slightly 85 convex and directed ventrolaterally. The transverse processes curve anteriorly, are dorsoventrally compressed, but are thicker and rounder than those of L5. Their dorsal and ventral surfaces are convex and their lateral extremities indicate the processes were also ventrally directed. A small rugose ridge runs anteroposteriorly across each transverse process approximately 13mm from the vertebral body. There are very small accessory processes on the posterolateral margins of the pedicles. These processes are approximately 2mm wide at their base and project posteriorly for 1mm. The remnant of the neural spine is 24mm high indicating that the spine was relatively tall. It is laterally compressed and anteroposteriorly deep, measuring 16.5mm long at its base. In lateral view, the spine tilts anteriorly at an angle of 70o to the horizontal axis of the centrum.

Comparisons It was hoped that the Thylacoleo carnifex skeleton from Moree may provide a useful sequence of lumbar vertebrae for comparison, however, the vertebral arches and transverse processes of these vertebrae are crushed or broken near their bases (Finch 1981). Numerous isolated lumbar vertebrae of Thylacoleo were examined for comparison. Unfortunately, the majority of these were lacking neural spines and transverse processes, but, in general these vertebrae had relatively deeper and rounder centra, wider vertebral foramen and more laterally placed articular facets. The transverse processes of Thylacoleo, although anteriorly-directed, also have flatter dorsal and ventral surfaces compared to those of W. hilmeri. The neural spines are similarly tall (32mm: Finch 1981) but are slightly posteriorly-inclined in contrast to those of W. hilmeri.

In overall morphology the posterior lumbar vertebrae of W. hilmeri most closely resemble those of Trichosurus vulpecula. Like W. hilmeri, the Trichosurus vertebrae have round, relatively deep centra and possess moderately long transverse processes that curve anteriorly and ventrally (although they are relatively thinner dorsoventrally). Their neural spines are also relatively long but differ slightly in their inclination; L5 leans slightly posteriorly while L6 is more upright. The posterior lumbar vertebrae of the dasyurids Dasyurus maculatus, D. viverrinus and Sarcophilus, also possess large anteroventrally curved transverse processes, 86 however, they differ in having the articular processes more laterally positioned and in having short, posteriorly-inclined neural spines (Sarcophilus only). The posterior lumbar vertebrae of the wombats examined in this study are less robust than those of W. hilmeri: the centra are dorsoventrally compressed and the pedicles are short. The transverse processes of the wombats are thinner and longer, and like those of Ngapakaldia (Munson 1992), lie in the horizontal plane. Phascolarctos differs greatly in having relatively smaller centra, short and anteroposteriorly deep neural spines, very short transverse processes, and like the wombats, have articular processes that are relatively long with facets that are more horizontally inclined. The lumbar vertebrae of Nimbadon also differ by having squarer articular surfaces of the centra, relatively shorter and dorsally tilted transverse processes, anteroposteriorly lengthened and shorter neural spines (like Phascolarctos), larger vertebral foramina and larger anterior articular processes.

5.2.4 Sacrum (Fig. 5.4)

The sacrum consists of the fused bodies of two sacral vertebrae. The sacrum exhibits a small degree of damage: the top of the spinous process and the cranial edge of the right articular process of the first sacral vertebra are missing and there are numerous pits and holes over much of the sacral surface. The sacrum is roughly wedge-shaped with the first sacral vertebra (S1) being much wider and deeper than the second vertebra (S2). In ventral aspect the total length of the vertebral bodies is 48.7mm. The body of S1 is broad and deep, measuring 22.6mm in length and 19.0mm in anterior height. The anterior articular surface of this body is damaged and rough. The body of S2 is broadly oval, measuring 21.9mm in length and 13.9mm in posterior height. Its posterior articular surface is tilted slightly dorsally.

The dorsal surface is dominated by the neural spines and articular processes of both vertebrae. The neural spine of S1 is tall, measuring 25mm in height, laterally compressed and is inclined posteriorly. About 11mm below its apex the spine straightens a little and at this point the spine is 10.8mm deep and 4.7mm wide. At its base the neural spine measures approximately 13.8mm in length and 6.5mm in 87 breadth. The neural spine of S2 is approximately 7.0mm long and 1.9mm wide at its base, and is approximately 8mm high.

S1 bears a pair of large anterior articular processes and a pair of much smaller posterior processes. The two articular facets of the anterior processes are separated by a maximum width of approximately 19.0mm. The intact left process bears a mamillary process on its dorsal margin below which is a large concave facet that is directed medially. This facet is 10.7mm deep and 15.3mm long. The smaller posterior articular processes are fused with the anterior processes of S2. The total width across the fused processes is 15.6mm. The fused articular processes create an intervertebral foramen between the spines. A thin sliver of bone runs longitudinally across this foramen. The posterior processes of S2 are much larger and more dorsally positioned than its anterior processes. The articular facets on the posterior processes are directed ventrolaterally at an angle of 45o to the horizontal and have a maximum dorsoventral width of 11.1mm.

The transverse process of S1 forms a large lateral process, the wing of the sacrum. Posteriorly, this wing fuses with the smaller transverse process of S2. The fused transverse processes form a dorsoventrally- deep (21mm) and long (34.2mm) articular facet for the ilium. The ilial facet is a rugose surface, rounded anteriorly and tapering posteriorly. The facet is nearly vertical and is tilted slightly dorsally. The articular surface is uneven and can be divided into an anterior half and posterior half. The anterior half is roughly circular in outline and has a raised, convex, anterior margin which develops posteriorly into a large near-vertical channel that descends in an anteroventral direction. The posterior half of the articular facet is nearly fusiform in shape, its dorsal margin forming a distinct ridge. The anterior portion of this area is slightly convex and the larger posterior portion is concave.

The posterior edges of the transverse processes of S2 form small posteriorly- projecting processes. A pair of pelvic sacral foramina is formed by the fusion of the lateral processes and lies lateral to the fused bodies and medial to the posterior third of the ilial articular facet. These foramina are round and have a diameter of approximately 5.1mm. 88

On the ventral surface of the sacrum the two vertebral bodies form a smooth longitudinally concave surface. The ventral surface of the transverse process of S1 curves convexly longitudinally and concavely laterally. The lateral edge of the transverse process of S1 projects ventrally and thickens forming a small, round, ventrally-directed facet. This process suggests that the medial surface of the ilium had a strong concave or nearly right-angled surface. The ventral surface of the transverse process of S2 is gently concave.

Comparisons The sacrum of W. hilmeri is most like that of Trichosurus vulpecula and T. caninus, except it is twice the size. The similarities shared by these sacra include a sacrum consisting of two fused sacral vertebrae, a very tall neural spine on S1 and a shorter but relatively tall spine on S2, a relatively deep sacral wing with a well-developed anterior region and a deep, sigmoidal ilial articular facet.

The sacrum of W. hilmeri is similar to that of Thylacoleo in being deep and robust with deep articular facets for the ilia. The neural spines are relatively similar in shape and inclination although those of Thylacoleo have a longer base. Not withstanding these similarities the two marsupial lion taxa show an interesting difference in the number of fused vertebrae that contribute to the sacrum; unlike W. hilmeri, Thylacoleo has three sacral vertebrae. In this respect W. hilmeri also differs from Vombatus and Phascolarctos (4 vertebral bodies), and Nimbadon, Ngapakaldia, Dasyurus and Sarcophilus (3 vertebral bodies). Many of the differences between the two thylacoleonid sacra stem from this difference in the number of vertebrae; e.g. Thylacoleo has three neural spines and has 2 pairs of sacral foramina. However, the sacral wing of Thylacoleo is also relatively longer and is significantly broader anteriorly, creating a much flatter and more dorsally oriented ilial facet. The posterior articular facets on its posterior vertebra are also relatively much smaller than those of W. hilmeri. 89

A B

right

left

Fig. 5.5: Wakaleo hilmeri QM F52187, left and right ilia.

A. lateral surfaces B. medial surfaces 90

Although the sacra of the wombats are nearly twice the width of W. hilmeri they have much shallower vertebral bodies, sacral wings and ilial articular surfaces. Unlike W. hilmeri, Phascolarctos and the wombats (except in S1) also have low and anteroposteriorly-lengthened neural spines (the posterior 3 spines are nearly fused together in Phascolarctos). Although the centra of Nimbadon and Sarcophilus are relatively deep, the sacra of these species also have low neural spines.

5.3 PELVIC GIRDLE

5.3.1 Ilium (Fig. 5.5)

The only parts of the pelvis that are preserved are the alae of the ilia. Both ilia are broken near their sacral articular facets. Enough of the posterior part of the left ilium is preserved to allow it to form an articulation with the sacrum. Both ilia show various amounts of damage, more of the right bone is missing than the left. The left ilial fragment has a maximum length of 70.5mm. It has a large triangular portion of bone missing from the centre of its medial surface and another area of bone missing from the posterior region on its lateral surface. A number of smaller fractures also run across its surface. The right ilial fragment is slightly shorter, having a maximum length of 62.3mm, and bears a relatively large hole (about 12 mm wide) in its ventral margin about midway along the bone. The ventral part of its anterior edge is also missing.

The ilia are thick, deep and robust. The anterior end of the left ilium has a depth of 39.1mm. At their anterior ends the ilia widen, extending slightly in a ventral direction, and also flare slightly laterally. Most of the medial surface (iliac fossa) is flat and smooth. Posteriorly, approaching the sacral articular area, the surface becomes concave, the ventral edge curling medially forming a cradle (approximately 7.5mm wide at the posterior break of the left ilium) in which sits the ventral portion of the sacral wings. The more complete lateral surface of the right ilium indicates that this surface is relatively flat anteriorly and becomes strongly convex posteriorly. The broken posterior edge of the right ilium shows that the bone in this region was approximately 12.4mm thick. 91

Comparisons The ilia of W. hilmeri are most similar to those of Trichosurus vulpecula, being mediolaterally compressed, having a relatively flat medial surface and a convex lateral surface. No comparative specimens observed possessed the distinctive medioventral channel for the sacrum although a small medioventral lip was present in some specimens of Trichosurus. Munson (1992) indicates that the os coxae of Ngapakaldia are extremely similar to those of Trichosurus. The ilia of both taxa are mediolaterally compressed although the iliac crest is slightly broader in the former species. These observations suggest a degree of similarity between the ilia of Ngapakaldia and W. hilmeri. The ilium of Thylacoleo has a straighter iliac crest and shows a greater degree of medial inflection resulting in a stronger lateral flaring of the ventral iliac spine. The ilia of W. hilmeri lack the distinct crest which lies near the flattened ventrolateral margin of the Thylacoleo ilium. The ilia of Phascolarctos and Nimbadon show a greater degree of anteroventral and lateral curvature of the alae, and the ventral margin of the former is relatively flat, forming a distinct medial ridge. The ilia of Vombatus and Phascolarctos are least similar being dorsoventrally flattened and having large, laterally flaring alae.

When articulated with the sacrum and lumbar vertebrae, the ilia of Wakaleo are relatively short and are much like those of Trichosurus and Dasyurus, extending anteriorly to the level of the L5/L6 articulation. The ilia of Phascolarctos and the wombats are comparatively longer extending to L5 and towards L4.

5.4 FORELIMB

5.4.1 Humerus (Fig. 5.6, 5.7 and 5.8)

The distal half of a right humerus is preserved and includes that part from the distal centimetre of the deltopectoral crest to the articulating condyles of the trochlea and capitulum. The surface of the shaft contains many small fractures and is missing areas of surface lamellae. A large fracture runs through the base of the shaft from near the superior margin of the lateral epicondylar crest to the entepicondylar 92 foramen. This large fracture displaces the anterior base of the shaft from the distal expansion of the 92

A B

Fig. 5.6: Wakaleo hilmeri humerus, QM F23442. A. anterior view B. posterior view Scale bar = 50 mm

AB

Fig. 5.7 : Humeri of Wakaleo hilmeri and Wakaleo pitikantensis (SAM P37720; Lake Pitikanta). Scale bar = 50 mm A. anterior view B. posterior view 93 humerus by two millimetres. A fracture also runs through the base of the strut which encloses the entepicondylar foramen. A large hole penetrates the surface bone on the posterior surface just medial to the lateral epicondylar crest. A shallow hole of similar dimensions is also present on the lateral surface of the shaft midway between the base of the deltopectoral crest and the epicondylar crest. The lateral edge of the lateral epicondylar crest and the lateral epicondyle is slightly abraded.

The remains of the humerus indicate it was stout bone with a moderately thickened shaft and a broad distal extremity. The distal part of the shaft is triangular in cross- section and bears the distal portion of the deltopectoral crest on its anterior surface. The anterior edge of the crest curves medially. Approximately 23mm distal to the deltopectoral crest a flange of bone extends laterally, forming the lateral epicondylar crest. The anterior surface of the flange is gently concave. The crest is thicker and more prominent on the proximal margin of the flange. The lateral epicondylar crest extends distally to the lateral epicondyle, just dorsal to the capitulum. Medial to the lateral epicondyle and proximal to the capitulum the humeral surface forms a rugose circular depression. Ventral to the deltopectoral crest the anterior edge of the shaft curves medially connecting with a strut of bone that encloses a large, oval-shaped entepicondylar foramen. This strut is bowed anteriorly. Distal to the entepicondylar foramen the humerus projects medially forming a gently-rounded entepicondyle. The medial surface of the condyle is strongly scarred from muscle and ligament attachments. A smoothly rounded arched surface 4.2mm wide separates the entepicondyle from the trochlea. The articular surface of the trochlea is gently convex and roughly circular in outline, its superior border forming an arch that continues laterally to connect with the articular margin of the capitulum. The capitulum is larger and rounder than the trochlea and its lateral surface is roughened. On the posterior surface of the humerus a small and very shallow olecranon fossa lies dorsal to the articular margins of the trochlea and capitulum.

Comparisons On the basis of distal width, the Wakaleo hilmeri humerus from Riversleigh is 21% larger than the humerus of W. pitikantensis from Lake Pitikanta, South Australia 94

Wakaleo Wakaleo Thylacoleo pitikantensis hilmeri carnifex

Fig. 5.8: Thylacoleonid humeri, anterior view. 95

(SAM P37720) and is approximately 30% smaller than that of Thylacoleo carnifex (see Table 5.1).

Table 5.1: Comparative measurements of humeral features of Wakaleo hilmeri, Wakaleo pitikantensis and Thylacoleo carnifex. Measurements are in mm.

SPECIES distal lat. shaft width at shaft depth at width epicondylar dorsal lat. dorsal lat. Specimen # crest height epicond. crest epicond. crest Wakaleo hilmeri QM F23443 52.9 46.0 16.6 14.7 Wakaleo pitikantensis 43.5 36.5 11.7 11.6 SAM P37720 Thylacoleo carnifex AM F106768 72.3 51.0 25.7 21.1 SAM P40007 76.4 59.9 29.0 23.1 SAM P38793 76.0 54.4 30.4 22.3

In general morphology the Riversleigh humerus (QM F23443) is more similar to the humerus from Lake Pitikanta (SAM P37720) than to Thylacoleo. In both QM F23443 and SAM P37720 the entepicondylar foramina are more distally positioned and the strut of bone that encloses this foramen curves anteriorly, unlike that of Thylacoleo which has a more proximal foramen and a flatter strut. In both the Riversleigh and Lake Pitikanta specimens the lateral epicondylar ridge is relatively taller, extending proximally well beyond the entepicondylar foramen while in Thylacoleo the lateral epicondylar crest extends only slightly proximal to the entepicondylar foramen. The entepicondyles of the Riversleigh and Lake Pitikanta humeri are more distomedially located relative to the trochlea in comparison to Thylacoleo where the entepicondyle is more dorsally (or proximally) positioned. The anterior surface of the shaft that lies lateral to the entepicondylar foramen and just proximal to the trochlea is relatively more convex in the Wakaleo humeri than in Thylacoleo. Additionally, the Wakaleo humeri also have shallower olecranon fossae than Thylacoleo. In anterior view, the articular surfaces of the capitulum and trochlea are also similar, and the lateral margins of their capitula angles toward the lateral epicondylar crest while that of Thylacoleo has a more vertical lateral margin. In both 96

Wakaleo humeri the trochlea surface is relatively round, located in the central third of the distal end of the humerus and account for approximately 50% of the total articular width. The trochlea of Thylacoleo, in contrast, is medially elongated and accounts for nearly two-thirds of the total width of the radio-ulnar articular surface. The W. hilmeri humerus differs slightly from that of W. pitikantensis in having a relatively thicker distal shaft ventral to the deltopectoral crest and in having the distal end of this crest show a greater degree of medial overhang.

The W. hilmeri and W. pitikantensis humeri show a high degree of morphological similarity to Trichosurus vulpecula and T.v. fuliginosus. Both subspecies of Trichosurus have a medially extended entepicondyle, a relatively straight shaft with the end of the deltoid crest proximally located, a capitulum that is taller than the trochlea, and a very shallow olecranon fossa. The W. pitikantensis humerus is also strongly similar to that of T.v. fuliginosus in the gross morphology of the bone but also in the relative proportions of many of the structures; for example, the total length of the humerus and width of the shaft are 70% that of W. pitikantensis. Similarly, its distal width is 60% that for W. pitikantensis.

The humeri of Nimbadon and Ngapakaldia (Munson 1992) are similar to W. hilmeri in having a broad distal end and broad lateral epicondylar crest, a thick strut enclosing the entepicondylar foramen and shallow olecranon fossa. However, they differ in having a proximally positioned entepicondyle, a transversely elongate trochlea, a medially-bowed shaft and a relatively tall and in Nimbadon, a thin deltopectoral crest. Although the humerus of W. hilmeri is similar in absolute distal width to Vombatus, the latter differs by having a relatively shorter and more triangular (in cross-section) shaft, a broader deltopectoral crest, a taller lateral epicondylar crest, a capitulum and trochlea that are centrally located and that are similar in height, a thinner strut enclosing the entepicondylar foramen, and a deeper olecranon fossa. The humerus of Phascolarctos is less robust being relatively smaller and much narrower, the entepicondyle is much more proximally positioned, the trochlea is comparatively elongate, and the deltopectoral crest and lateral epicondylar crest are narrower. The humeri of the dasyurids Sarcophilus and Dasyurus, being relatively long and narrow, show least similarities. In addition, the entepicondylar 97 foramen of Sarcophilus is relatively small while in Dasyurus it is completely absent. The dasyurids also differ in having trochleae and capitulae that are similar in size to each other, although the trochlea is distinctly deeper. These articular structures fuse to form a relatively narrow distal articular surface which has straight, ridge-like lateral and medial edges. These species also differ in having a narrow lateral epicondylar crest, a laterally-bowed shaft (in Dasyurus) and a very deep and broad olecranon fossa.

5.4.2 Radius (Fig. 5.9 & 5.10)

The distal portion of a left radius, measuring 57.5mm in length, is preserved. Most of the medial margin of the shaft is missing. The shaft is roughly oval in cross-section and widens medio-laterally near the epiphyseal suture line where it measures 21.9mm in breadth. The shaft is deepest approximately 10mm above the articular surface and becomes slightly thinner proximally. The anterior surface is convex and bears a large and prominent muscle scar along much of the lateral half of its surface, the scar extending from the proximal break to within 17mm of the epiphyseal joint. This scar possibly represents the attachment for part of the extensor pollicis muscle group, i.e. the extensor pollicis brevis muscle (Barbour 1963; Murray and Megirian 1990). The lateral edge of the anterior surface is rounded. Distally, the lateral surface flattens and forms a broad (approx. 5mm) rugose depression in the region of the distal ulnar articulation (just proximal to the epiphyseal suture line). The lateral border forms a relatively angular edge with the posterior surface of the shaft. The posterior surface of the shaft is flattened and gently concave. The medial edge of the distal shaft is rounded.

The distal carpal articulating surface is roughly rectangular in outline and bears a large medial styloid process and a smaller, posteriorly-located, lateral styloid process. In dorsal view the medial styloid process is triangular, its medial surface being relatively flat while its articular surface is convex. In distal aspect, a convex ridge extends from the medial styloid process across the central portion of the articular surface, broadening laterally, to the lateral styloid process. This surface represents the 98

A B C

Fig. 5.9 Wakaleo hilmeri QM F52239, radius. A. anterior surface B. posterior surface C. distal facet

Wakaleo Wakaleo Wakaleo Thylacoleo pitikantensis hilmeri vanderleueri carnifex

Fig. 5.10: Radii of species of Wakaleo and Thylacoleo carnifex (all distal ends except W. pitikantensis) 99 articular surface with the scapholunar bone (assuming the scaphoid and lunar bones have fused in this species, as in Thylacoleo). This central articular ridge is delineated from a small posterior shelf by a narrow transverse groove. A small rugose depression lies at the anterior margin of the articular surface near the anterolateral base of the medial styloid process. The posterior articular surface forms a smooth elongate ledge that is broader medially and narrows laterally. The posterior edge of this ledge protrudes to form a slight crest along the epiphyseal joint line.

Comparisons Although the distal radius fragment of W. hilmeri is similar in size to that of W. vanderleueri, in most morphological features it is more similar to the smaller radius of W. pitikantensis. The W. vanderleueri radius is only 0.8mm wider at the distal epiphyseal joint line (22.7mm), but above the joint line its shaft is narrower (15.5mm) and much more anteroposteriorly compressed (Fig. 5.9).

The Riversleigh radius is similar to those of W. pitikantensis and T. carnifex in having a shaft with an anterior surface that is convex and bears strong scarring indicating an aponeurotic or tendinous muscle attachment (Bryant and Seymour 1990). As indicated above, this scarring possibly represents the origin for the extensor pollicis brevis muscle. In comparison to Trichosurus vulpecula (proximal half), the attachment for this muscle within thylacoleonids seems to be more distally located (Barbour 1963). In contrast, the anterior surface of the W. vanderleueri radius is quite flattened and scarring in this region is practically non-existent. It is possible that the attachment for the extensor pollicis muscle may have been more proximally located in that animal. The rugose distal surface on the lateral edge of the Riversleigh radius is also seen in other thylacoleonid specimens, the only difference being that its surface is relatively flatter and broader than that of W. pitikantensis and W. vanderleueri. The medial surfaces of the distal shaft and medial styloid process are also similar to those of W. pitikantensis being gently convex with a weak swelling at the epiphyseal joint line. This contrasts markedly with the medial surfaces in W. vanderleueri and Thylacoleo which both bear a distinct medial tuberosity on the epiphyseal joint line. 100

In general, the carpal articular surface is most similar to those of W. pitikantensis and W. vanderleueri radii. The Riversleigh radius shares with these specimens an articular surface that has an anteroposteriorly central medial styloid process (more posterior in Thylacoleo), a transverse central convex ridge and a small anterior depression at the anterolateral base of the medial styloid process. In T. carnifex, its larger medial styloid process results in this depression being more centrally and less medially situated.

The lateral styloid process of the Riversleigh radius is more prominent than that of W. pitikantensis and in this respect it is more similar to W. vanderleueri. The smooth convex posterior shelf of the articulating surface is most similar to that of the W. pitikantensis radius although it narrows more abruptly at its lateral end. In contrast, the posterior articular shelf in W. vanderleueri bears a distinct depression at its medial end near the proximal base of the medial styloid process.

The distal shaft of the W. hilmeri radius is similar to the distal shaft of the wombats, Nimbadon and Ngapakaldia in being slightly anteroposteriorly compressed. The remnant of the W. hilmeri radius is relatively straight and suggests that the whole shaft was relatively straight like that of W. pitikantensis, Ngapakaldia and the phalangerids Trichosurus and Spilocuscus. In contrast, the shafts of Phascolarctos and Nimbadon are moderately curved and those of the wombats are strongly curved. The distal articular surface of the W. hilmeri radius, like those of W. pitikantensis and W. vanderleueri, differs from most of the comparative taxa (wombats, Nimbadon, Ngapakaldia, phalangerids, dasyurids) in being rectangular in outline rather than triangular. The triangular outline in these groups is caused by the presence of a tubercle on the anterior surface near the epiphyseal suture line. This tubercle is most strongly developed in Vombatus and Trichosurus. The distal articular surface of Phascolarctos is most similar to that of W. hilmeri in having a roughly rectangular outline, a weakly developed anterior tubercle, and in having a slight ledge on its posterior articular surface. Some taxa, for example Ngapakaldia, Sarcophilus and Trichosurus, also bear small tubercles on the medial margin of the epiphysis. Interestingly, although these are lacking on the W. hilmeri and W. pitikantensis radii, they are present on the radius of W. vanderleueri. 101

A B C

Fig. 5.11: Wakaleo hilmeri QM F52238, ulna.

A. anterior view. B. medial view. C. lateral view. 102

5.4.2 Ulna (Fig. 5.11)

Only the proximal fragment of the left ulna is preserved. This fragment is 47.3mm in length and includes the olecranon process and most of the articular facet for the humerus. The olecranon is robust and laterally compressed, and in profile is rectangular in outline. The medial surface bears two fossae, a small, triangular proximal fossa and a larger elongate fossa that lies adjacent to the trochlea notch. A moderate tuberosity (medial tuberosity) lies at the proximo-posterior corner of the medial surface and is probably the site of attachment for the long head of the triceps brachii muscle. The lateral surface has strong muscle scarring over most of the posterior portion of its surface. The proximal end of the olecranon bears a raised and roughened edge around its anterior and medial margins. The well-developed medial tuberosity and strong rugosities on both the proximal and lateral surfaces of the olecranon suggests well-developed triceps brachii. The posterior surface is proximally broad and flat, has rugose margins, and narrows distally. The humeral facet (trochlea notch) is saddle-shaped being convex laterally and concave proximo- distally. The dorsal margin of the facet, the anconeal process, forms a small, sharp, anteriorly-projecting lip. The medio-distal margin of the facet, the coronoid process, is broken and measures 14.0mm in width at the break. The process projects medially and is concave. A small proximal remnant of the facet for the radius, the radial notch, lies directly distal to the lateral half of the trochlea notch. This remnant is semicircular, flat and directed anterolaterally.

Comparisons Comparisons of the Wakaleo ulna with those of other marsupial taxa indicate that it is most similar to that of Trichosurus, Thylacoleo and Spilocuscus: these taxa have a relatively broad trochlea notch, an anterolaterally placed radial notch, a relatively small anconeal process and a robust olecranon process although that of Thylacoleo tapers proximally and is consequently less rectangular in profile. Another feature these taxa share with Wakaleo is the presence of a tuberosity on the postero-proximal corner on the medial side of the olecranon. The trochlea notch of Spilocuscus differs slightly in being taller on its medial edge. Unfortunately, the loss of the coronoid 103 process from the Wakaleo specimen prevents comparison of this structure which is quite long and prominent on the ulna of Thylacoleo.

The olecranon of Phascolarctos also shows some similarities to Wakaleo hilmeri in having a broad trochlea notch and small anconeal process although the olecranon lacks the posterior width and tapers slightly proximally. In addition, Phascolarctos lacks the distinct medial tuberosity, has a shallower medial fossa, and the radial notch is slightly more laterally located. The proximal ulna of Nimbadon also shows a number of differences, having a tapering olecranon, a broader trochlea notch and extremely wide radial notch. One feature it does share with Wakaleo is the presence of the medial tuberosity.

The olecranon of the wombats shows the greatest contrast to that of Wakaleo. The olecranon of wombats is relatively longer, broader anteriorly, more strongly medially inflected, and has an extremely prominent anconeal process. Additionally, the lateral half of the trochlea notch is well-developed and highly concave while the medial half is relatively reduced, the coronoid process projects more strongly medially and has a more horizontally-inclined surface, and the radial notch is significantly broader and more concave. Wombats also exhibit the medial tuberosity on the olecranon but it is more anteriorly positioned being located about half way between the anterior and posterior margins of the medial surface. The proximal ulna of Ngapakaldia is quite wombat-like in having an anteriorly-broad and medially-curved olecranon process and large coronoid process. However, unlike wombats, the radial notch is flat and the anconeal process is small (Munson1992) and as such Ngapakaldia exhibits few similarities with that of Wakaleo.

5.5 MANUS

5.5.1 Hamatum (Fig. 5.12)

The left hamatum is a many-faceted, asymmetrical bone. The dorsal half of its medial surface is roughened as a result of damage. Its maximum anterior width is 104

16.8mm. The dorsal surface is triangular and rugose. The anterior and medial edges of the 104

AB C

Fig. 5.12: Wakaleo hilmeri QM F52240, hamatum (unciform). A. anterior view (MCIV facet) B. dorsal view C. medial view Scale bar = 10mm

A BC D

E Fig. 5.13: Wakaleo hilmeri QM F52241, trapezium. A. lateral view (trapezoid facet) B. anterior view (MCI facet) C. posterior view (scapholunar facet) D. anterolateral view (MCI facet left, trapezoid facet right) E. Trapezium articulating with pollex Scale bar = 10mm 105 dorsal surface are perpendicular to each other and the posterolateral border forms the hypotenuse of this triangle. The dorsal surface is slightly concave and rises anterolaterally to form a small ridge at the anterior margin. The posterolateral margin bears a small concavity anteriorly and develops a small ridge posteriorly. The anterior surface of the hamatum consists of a large triangular, concave facet and a lateral rectangular process, the hook, that projects anterolaterally and ventrally. On the basis of studies of the manus of Thylacoleo (Wells and Nichol 1977; Finch and Freedman 1988) and from observations of the manus of Vombatus, this triangular facet articulates mainly with metacarpal IV (MC IV) and the medial part of metacarpal V (MC V). Lateral to the MC IV facet the anterodorsal surface of the hook forms a strongly concave facet for MC V. The anterior portion of the ventral (palmar) surface bears the ventrally projecting rectangular hook. The hook has a maximum anteroposterior width of 9.5mm. The ventromedial surface of the hook is slightly concave. The palmar surface is bordered on its posterior margin by the smooth ventral edge of the articular facet for the cuneiform. The anteromedial corner of the palmar surface is incised by a 2.3mm-broad channel that continues onto the medial surface.

The medial face of the hamatum is vertical and has a roughened (damaged) surface. This surface would have articulated with the capitatum. The outline of the medial surface is C-shaped with a distinct convex posterior margin, and a concave, ventrally-truncated anterior margin. At its anteroventral corner the medial face is incised by a narrow sulcus that widens onto the medial surface. Ventral to the sulcus the medial surface protrudes anteriorly as a rounded process. The posterior surface of the hamatum bears a deep, squarish articular facet for the cuneiform. This facet measures 11.4mm dorsoventrally and its maximum mediolateral width is 11.0mm. The surface of the facet is nearly saddle-shaped being mainly convex but with a slight concavity on its lateral margin and just below its dorsal margin. Lateral to the cuneiform facet, on the posterior surface of the MC V facet and the hook, lies a broad, ovoid depression.

Comparisons 106

The hamatum of W. hilmeri differs greatly from that of Thylacoleo and is most similar in total morphology to Trichosurus although the latter is much smaller. Strong similarities are also shared with Nimbadon and Ngapakaldia. Like W. hilmeri, in those species the hamatum is roughly triangular in dorsal view and has relatively similar proportions, the anterior and medial surfaces being of similar length, and the dorsoventral depth similar to the width of the flattened dorsal surface. All the main articular facets also show strong morphological similarities with those of Wakaleo. In contrast, the hamatum of Vombatus, Lasiorhinus and most noticeably Thylacoleo show less similarity, being extremely broad but anteroposteriorly compressed. This change in shape has also resulted in differences in their articular surfaces.

The palmar surface of the Wakaleo hamatum is similar to that of Trichosurus, Nimbadon, Ngapakaldia and Phascolarctos. These species have a relatively rectangular hook that projects ventrally and laterally. Ngapakaldia differs slightly by having a hook that is relatively longer anteroposteriorly and extends further posteriorly. In contrast, the hook of Thylacoleo is rounded and thickened and its posteroventral surface is convex rather than concave. In Vombatus the hook is also rounded but projects ventrally and less anteriorly.

The articular facets for the capitatum in Trichosurus, Nimbadon, Ngapakaldia and Phascolarctos show a basic morphology similar to that of Wakaleo: a roughened sulcus lies at the anteroventral corner and a roughly round and shallow concave facet lies on the posterior half of the surface. The articular facet for the capitatum in Thylacoleo differs in being relatively long (anteroposteriorly) but dorsoventrally shallow and its dorsoanterior margin forms a distinct overhang above the concave facet that curves onto the anterior surface. The facet for the capitatum in Vombatus is also different being quite short and posteriorly convex.

Thylacoleo lacks the distinct triangular-shaped MC IV facet. Instead, this facet is comparatively shallow and bears a large tuberosity at its lateral margin and another at its ventromedial margin. This facet articulates mainly with MC IV while the lateral tuberosity together with the anterior margin of the hook forms the broad articular surface for MC V. 107

The facet for MC IV on the hamatum of Trichosurus, Nimbadon and Ngapakaldia have roughly the same triangular shape as seen in Wakaleo. Phascolarctos differs slightly by having a slightly broader, more rectangular facet. The facet for MC V in these taxa is also relatively similar to that seen in Wakaleo being relatively broad and strongly concave anteriorly. Ngapakaldia differs in having a taller facet and in being broader dorsally. The facet in Phascolarctos additionally differs in being relatively shallow and weakly concave.

Although the facets for the metacarpals of Vombatus are similar in size to those of W. hilmeri, the MC IV facet is flatter and its lateral margin forms a gentle ridge with the MC V facet and not the sharp ridge seen in W. hilmeri. Further, the MC V facet in Vombatus lacks the depth and strong concavity seen in Wakaleo.

5.5.2 Trapezium (Fig. 5.13)

The trapezium is from the left manus and is a small triangular bone measuring 13.0mm in height and 11.5mm in width. The trapezium bears three major articular facets. Its anterior surface bears two articular facets, one medial and the other lateral. The medial facet is tall, triangular and directed anteromedially and articulates with the first metacarpal (MCI). The dorsal half of this facet is concave and the ventral half is convex. The anterolateral surface of the trapezium is tall, narrow and rectangular in outline measuring 11.7mm in height and has a maximum anteroposterior depth of 6.8mm. This surface bears the articular facet for the trapezoid. This facet is tall and tapers at its dorsal and ventral margins. The surface of the facet is convex dorsally and concave ventrally. At the base of the anterior surface a triangular rugose area separates the two articular facets. The posterior surface of the trapezium is triangular in outline and is dominated by a large ventrally-directed channel on the medial two-thirds of its surface. The channel measures 4.3mm wide and 8.4mm long. Lateral to this channel and on the ventrolateral corner of the posterior surface is a small triangular facet. This facet would have articulated with the distal surface of the scapholunar. The ventral surface 108 of the trapezium is U-shaped, narrow and roughened and forms part of the palmar surface of the hand.

Comparisons The trapezium of Thylacoleo is quite unlike that of Wakaleo being much more anteroposteriorly elongate, anteriorly-narrow and flattened, and taller posteriorly. The articular facets for the trapezoid and MCI are longer and more concave and the deep channel on the proximal surface is lacking.

The trapezium of Wakaleo is most similar in shape to that of Trichosurus except it is approximately three times the size. In both taxa the trapezium is triangular in outline, the points of the triangle being superior and inferior lateral and medial. A broad subtriangular facet for MCI dominates the medial half of the anterior surface. The MCI facet in both species is concave dorsally and convex ventrally. In both species, the proximal surface bears a broad concave channel medially and a small triangular facet for the scapholunar laterally. The trapezium of Ngapakaldia, although slightly larger, is also generally morphologically similar to that of Wakaleo. The former also exhibits a triangular shape and differs in having a relatively longer and narrower facet for MCI and a slightly longer and more convex facet for the trapezoid. The articular region for the scapholunar appears to be shallower and shorter although this maybe an artefact of the casting process.

The trapezium of Phascolarctos is also similar to that of Wakaleo, although it is half the size. The trapezium of Phascolarctos has a flatter MCI facet and lacks the convexity seen in Wakaleo, and has a much shallower channel on its proximal surface. The trapezium of Vombatus also differs slightly, being less triangular in outline because it has an arched ventral margin, has flatter articular surfaces for MCI and the trapezoid, and lacks the deep concave channel on its proximal surface. The trapezium of Nimbadon shows few similarities to that of Wakaleo, differing in general shape, curvature and shape of nearly all articular facets and in the size and shallowness of the scapholunar articular region. 109

5.5.3 Metacarpal I(MCI) (Fig. 5.14)

The first metacarpal from the left manus is preserved. It is a broad, stout, dorsoventrally compressed bone measuring 17.9mm long. It has a relatively short shaft length of 8mm, a mid-shaft width of 8.9mm and depth of 4.9mm. Its proximal and distal ends bear large articular facets. Its distal articular facet is a prominent convex surface that extends onto the dorsal and ventral surfaces. This facet is medially offset and medially directed. The proximal facet for articulation with the trapezium is broad and extends slightly onto the dorsal surface and thickens at its ventrolateral margin. The lateral half of the facet is convex and the medial half of the facet concave.

Comparisons The MCI of W. hilmeri is very similar to that of Thylacoleo although the latter is comparatively larger and relatively shorter. The MCI of Thylacoleo curves medially, the medial side being much shorter than the lateral side, and its distal facet is broad and medially-directed. The proximal facet is also similar to that of W. hilmeri in being dorsolaterally convex and medially-concave. Although the MCI of Phascolarctos is a straighter bone and is relatively narrower, it is similar in having a distal facet that is also slightly medially-directed, however it does differ in having a proximal facet that is predominantly convex and lacks the broad medial concavity seen in Wakaleo. The MCI of Trichosurus is similar to that of Phascolarctos but lacks the medial orientation of the distal facet and has a slightly flatter proximal facet. The MCI of Nimbadon and Ngapakaldia are short, stout bones but lack the strong medial curvature of the shaft and medial direction of the distal facet although the enlargement of the lateral half of this facet results in the proximal phalanx being angled medially. The greatest contrast is seen in Vombatus in which the MCI curves in the opposite direction, i.e. laterally, towards the second digit. In addition, the proximal facet differs in being predominantly convex but this convexity does not extend onto the dorsal margin of the facet and the facet lacks the broad medial concavity present in Wakaleo. 110

A B Fig. 5.14: Wakaleo hilmeri QM F52242, left first metacarpal (MCI).

A. dorsal view B. ventral view

A Fig. 5.15: Wakaleo hilmeri QM F52243 (A) and QM F52244 (B), proximal phalanges.

A. right proximal phalanx, digit II, lateral view. B. left proximal phalanx, pollex, medial view. B

AB Fig. 5.16: Wakaleo hilmeri digits.

A. left pollex (MCI and prox. phalanx: QM F52242 and QM F52244)

B. right digit II (prox. & medial phalanx: QM F52243 and QM F52245). 111

5.5.4 Pollex: left proximal phalanx (Fig. 5.15 & 5.16)

The proximal phalange of the left pollex is preserved. This bone is relatively long and narrow and measures 25.1mm in length. Proximally, the phalange is broad (12.2mm) and forms a wide concave facet that encloses the distal facet of MCI. The medial margin of the proximal fact extends further posteriorly than the lateral margin. In lateral view, the phalanx is slightly arched. In dorsal view, the shaft narrows just before the large, round, distal articular facet. On the ventral surface of the shaft the posterior edges are slightly raised and rugose indicating sites of tendon attachment.

5.5.5 Digit II: right proximal phalanx (Fig. 5.15 & 5.16)

This bone is very similar to the proximal phalanx of the pollex being long and narrow, but it does show some features lacking in that bone. The medial edge of the bone is longer (27.9mm) than the lateral edge (25.6mm), the medial edge extending further posteriorly. Consequently, the proximal facet of the phalanx (for articulation with MCII) forms an angle of 25o to the longitudinal axis of the bone. In profile, the element is gently arched and has a mid-shaft depth of 4.6mm. The dorsal surface of the shaft is convex while the ventral surface is relatively flat with a small rounded tuberosity on each edge approximately 5mm behind the distal articular condyle. The posterior edges on the ventral surface of the shaft are slightly raised and roughened. The proximal articular facet consists of two concave facets separated by a weak dorsoventral ridge that runs slightly medial to the midline of the bone. The lateral concavity is directed posteriorly and the medial concavity extends further posteriorly and is directed in a more posterolateral direction.

5.5.6 Digit II: right medial phalanx (Fig. 5.16B)

The medial phalanx is a short, squat, dorsoventrally compressed bone. It measures 13.0mm in length, its posterior width is 10.4mm, anterior width 8.2mm and mid- shaft depth 3.9mm. The proximal articular facet is kidney shaped, its surface bearing 112 twinned, shallow, concave facets that are directed posteriorly and slightly dorsally. Its distal facet is a broad, biconvex surface that bears a central groove.

Comparisons Comparison of the Wakaleo phalanges with those of Thylacoleo shows that there is a high degree of morphological similarity between the corresponding elements. Both species have a medially curving MCI and have relatively long proximal phalanges relative to the medial phalanges. The greatest difference is seen in the morphology of the MCI of the pollex, the MCI of Thylacoleo being much broader with a more strongly medially-directed distal articular facet. The proximal phalanx that articulates with this MCI also shows greater posterior development of the lateral half of its MCI articular facet. The asymmetry of these surfaces results in the two articulated bones forming a much more medially-directed digit.

Although relatively longer and narrower, the proximal phalanx of the second digit of Wakaleo is very similar to that of Thylacoleo, both bones exhibiting the posterior extension of their medial sides. The Wakaleo medial phalanx is also morphologically similar to the medial phalanx in Thylacoleo being short and dorsoventrally flattened.

Although approximately half the length and about one third the width and depth, the proximal phalanges of Trichosurus are extremely similar to those of Wakaleo being relatively long and slender, arched and showing an increased width posteriorly. As in Wakaleo, the proximal phalanx of the pollex is much longer than its corresponding metacarpal. The articular facets are also very similar, the proximal facets being gently concave although that of digit II does not show the same degree of posterior extension of the medial side as seen in Wakaleo. The medial phalanges of Trichosurus, like that of Wakaleo, are short, being approximately half the length of proximal phalanges, however, they are much more slender than those of Wakaleo. The condyles on the distal articular facet are relatively well-developed and separated by a deeper groove.

The proximal phalanges of the pollex of Phascolarctos and Nimbadon like that of Wakaleo have a slightly longer posterior medial margin but differ in a number of 113 aspects; they are proportionally shorter (being similar in length to their MCI), show less arching of the shaft and less posterior widening, and have a deeper groove on the distal articular facet. Phascolarctos also differs in having a shallower proximal facet and the posterior extension of the medial side is only weakly developed. The proximal phalanx of Nimbadon also differs in having a relatively broader and deeper shaft, in having the medial part of the proximal facet being smaller than the lateral part, and in having a more prominent posteriorly-projecting process at the dorsal margin of the proximal facet.

The proximal phalanx of digit II of Wakaleo is similar to those of Phascolarctos and Nimbadon being relatively long (approximately twice the length of the medial phalanx) but the latter taxa lack the arched shaft and posterior widening seen in the former. This bone in Phascolarctos also has a deeper groove between the condyles on the distal articular facet that extends strongly onto the ventral surface and also shows weaker posterior extension of the medial half of the proximal facet. This phalanx in Nimbadon is generally similar to that of Wakaleo in the shape of the facet for MCII although the whole facet is relatively more concave. Nimbadon also differs from Wakaleo in having a thicker shaft, and in having a distinctly deeper and broader concave groove on the distal articular facet.

The medial phalanges of Phascolarctos and Nimbadon are generally similar to that of Wakaleo in being short and stout but are relatively cylindrical and lack the dorsoventral compression seen in the latter. Phascolarctos also differs from Wakaleo in having a deeper groove on the distal articular facet and a more developed ridge separating the condylar surfaces on the proximal facet. The medial phalanx of Nimbadon also differs by having a more strongly biconcave proximal facet which has a prominent dorsal process that severely limits extensor movement of the medial phalanx on the proximal phalanx.

The proximal phalanges of Vombatus differ from those of Wakaleo in showing a considerable reduction in length; that of the pollex is the same length as its MCI (not longer than the MCI) while that of digit II is only half the length of the MCII and 114 only 25% longer than the medial phalanx. The proximal phalanx of the Vombatus pollex 114

Fig. 5.17: Wakaleo hilmeri QM F52246, calcaneum, dorsal view.

ABcoracoid

scapular tuberosity

glenoid fossa

subscapular

Fig. 5.18 Wakaleo pitikantensis scapula, SAM P37720 ; Lake Pitikanta.

Scale bar = 10 mm A. anterior view B. medial view C. lateral view C 115 also lacks the dorsal arch seen in Wakaleo but is similar in showing an increased width posteriorly and slightly greater development of medial side of the proximal facet. The medial phalanges of Vombatus, being short and squat, are similar to the medial phalanx of Wakaleo.

5.6.0 PES

5.6.1 Calcaneum (Fig.5.17)

A left calcaneum is preserved that displays damage to the medial and lateral edges of the astragalus articular facet as well as along most of its lateral surface. It is a relatively long, mediolaterally-compressed, rectangular bone in which the ventral surface bows concavely. Its anteroposterior length measures 41.2mm. The anterior half bears three articular facets while the posterior half consists of the calcaneal tuber. The calcaneal tuber measures 15.5mm dorsoventrally and is 17.9mm long. The medial surface of the tuber is relatively flat, tilts dorsomedially and its caudal margin is thickened in a slight ridge. The posterior surface of the tuber is relatively flat. The anterior surface of the calcaneum forms a large triangular facet that would have articulated with the proximal surface of the cuboid. The points of this triangle are positioned laterally, medioventrally and mediodorsally. The lateral point extends furthest anteriorly while the dorsomedial point is the most posterior. The medioventral part of the facet projects ventrally and helps to form the concave ventral margin of the bone. The facet for the cuboid is broad and in dorsal view is strongly concave, the concavity becoming angular near the centre of the facet such that the lateral half of the facet is directed medially and the medial half of the facet is directed anteriorly. On the dorsal surface of the calcaneum, proximal to the cuboid facet, are the articular facets for the astragalus. Most of the medial portion of the facet is intact (only the medial edge is missing) but the lateral facet is missing. The medial facet is rectangular in outline, has a gently concave surface, and is directed dorsally. Almost the entire dorsal half of the lateral surface of the calcaneum is lost. A long tubercle is present at the anterior end of the lateral surface.

Comparisons 116

The calcaneum of W. hilmeri shows a similar basic plan to that of Thylacoleo but is proportionally much narrower and its calcaneal tuber much longer. Both species have a broad, slightly concave medial astragalar facet, and a distinctly angular cuboid facet, although in Thylacoleo the medial half of the facet is proportionally larger, more dorsally directed and protrudes further anteriorly. The greatest difference is seen in the calcaneal tuber which in Thylacoleo is relatively shorter and its dorsomedial and posterior surfaces are much more rounded. The position of the lateral tuberosity in Thylacoleo also differs in being more centrally located.

In general morphology, although it is nearly twice the size, the calcaneum of W. hilmeri is most similar to that of Trichosurus. The calcanea of both these species are relatively long and narrow, have a cuboid facet that has a distinct dorsovental groove creating a sharply angular surface, have a broad bifaceted astragalus facet and a straight and relatively narrow calcaneal tuber. The broad articular facet for the astragalus, with a convex lateral facet and concave medial facet, is common to most of the Vombatiformes (sensu Aplin and Archer 1987), however, the calcanea within this group show other differences in morphology. For example, although the cuboid facet on the calcaneum of Ngapakaldia is similar to that of W. hilmeri, the former differs by having a proportionately broader, shallower and curved calcaneal tuber. The calcaneum of Phascolarctos, although relatively slim, differs by having a steeper and less angular cuboid facet, lacks the anterior protrusion of the lateral edge of this facet, and has an anteroposteriorly longer medial facet for the astragalus. The calcanea of the wombats differ greatly in being extremely curved and in having a cuboid facet that is steep, gently concave and ventromedially reduced.

5.7 AMENDED DESCRIPTION OF THE SCAPULA OF Wakaleo pitikantensis (SAM P37720) FROM LAKE PITIKANTA, SOUTH AUSTRALIA.

A broken right scapula of the holotype of W. pitikantensis (SAM P37720) collected from the Ngapakaldi Local Fauna was described by Rauscher (1987; p. 425). After examination of this bone and a container of unidentified bony fragments of the holotype, I discovered that one of the fragments was the anteromedial edge of the 117 glenoid fossa. This fragment bore the scapular tuberosity (supraglenoid tuberosity) and coracoid process of the scapula. Because these structures are important scapular features an amended description of this bone follows (Fig 5.18).

Most of the broad, thin, dorsal part of the scapula (i.e. the areas of attachment for the supraspinous, infraspinous and subscapularis muscles, and the spine) are missing. The ventral end of the spine that bears the remnants of the base of the acromion process is preserved.

The glenoid fossa is elliptical in outline, the anterior end being narrower than the posterior. The glenoid fossa measures 21.5mm anteroposteriorly (from the scapular tuberosity to the posterior rim of the fossa) and 13.9mm at its broadest diameter. The fossa is shallow. The anterior end of the fossa narrows to a relatively sharp, anteroventrally projecting point on the scapular tuberosity. The scapular tuberosity lies at the anterior margin of the fossa and is relatively flat and broad, measuring 4.9mm across its base and 6.1mm dorsoventrally. The scapular tuberosity is the point of attachment for the biceps brachii muscle and the tuberosity’s strong development suggests this muscle was also well-developed. Medial to the scapular tuberosity a short coracoid process curves medially. Its anterior surface forms a broad (5.7mm) convex cap, its lateral edge projecting slightly. Dorsal to the coracoid process the cranial border of the scapula is concave, forming a relatively deep scapular notch. The posterior border of the scapula is thick and roughened. On the medial side of the posterior border, just above the glenoid fossa, is a deep and narrow furrow. This furrow is 12.1mm long and would have been the site of attachment for the caput longum of the triceps brachii and teres minor muscles.

Comparisons The remains of the W. pitikantensis scapula indicate it was a relatively robust bone, similar in stoutness to those of Vombatus and Phascolarctos. In outline, the glenoid fossa is most similar to that of Vombatus although that species exhibits a straighter lateral margin of the fossa. In other marsupial species the anterior end of the glenoid fossa is distinctly narrowed giving the fossa a piriform outline. However, in contrast to the relatively cupped surface of this fossa seen in Vombatus, the fossa of W. 118 pitikantensis is more similar to that of Trichosurus v. fuliginosus and Phascolarctos, being open and shallow. Wakaleo pitikantensis is also similar to the latter in having a relatively distinct scapular tuberosity unlike Vombatus in which the tuberosity is barely noticeable. The scapula of Thylacoleo also shows a number of morphological differences; the lateral edge of the glenoid fossa is straighter, the scapula notch is much shallower, the scapular tuberosity is broader and the coronoid process extends further ventrally. Additionally, its subscapular fossa is deeper near the margin of the glenoid fossa unlike the W. pitikantensis scapula which has greater depth dorsally.

A functional assessment of the postcrania of W. hilmeri and W. pitikantensis is given in Chapter 13. 119

CHAPTER 6

Wakaleo sp. cf. W. hilmeri

6.1 Introduction

The first thylacoleonid fossil recovered from Riversleigh Station was a left maxillary fragment collected from Tedford’s original D Site on the D Site plateau. This specimen differs from those of Wakaleo hilmeri (QM F45200 and QM F23443) in its robustness (reflected in the greater width of its molars and lack of M4. Although having a similar molar number to W. oldfieldi and W. vanderleueri, the D site specimen is smaller than these marsupial lions.

6.2 Wakaleo sp. cf. W. hilmeri

Wakaleo sp. cf. W. hilmeri

(Figs 6.1, 6.2 & 6.3)

Holotype of Wakaleo hilmeri: QM F45200, a cranium from Hiatus Site, Riversleigh World Heritage Area, Lawn Hill National Park, northwestern Queensland.

Wakaleo sp. cf. W. hilmeri specimen: QM F23446, a left maxillary fragment containing broken P1, P3, M1-3 and alveoli for C1 and P2.

Locality: D Site, Riversleigh World Heritage Area, northwestern Queensland, Australia. 120

Fig. 6.1: Wakaleo sp. cf. W. hilmeri, QM F23446,left maxilla from D Site. Stereo occlusal photo. Scale bar = 10 mm

Fig. 6.2: Wakaleo sp. cf. W. hilmeri, QM F23446 left maxilla from D Site. Profile view.

31.4

12.6 9.6 6.0 5.2

3 P M1 M2 P2 M3 P1 P2 P1

50.3

Fig. 6.3: Measurements of QM F23446, Wakaleo sp. cf. W. hilmeri, left maxilla (in millimetres). 121

Age: D Site is assigned to Riversleigh’s System A group of sediments which is interpreted to be approximately late Oligocene in age (Tedford 1967, Creaser 1997, Archer et al. 1997).

Description The maxilla is robust and the portion preserved extends from the premaxilla- maxillary suture and posterior border of the canine alveolus to slightly posterior to M3. The anterior dorsal, and posterior margins of the maxillary bone are missing. The specimen retains P1, P3, and M1-3, however, the crown of P1 and buccal portions of P3 and M1 are missing. There is an alveolus for P2. The enamel is roughened and obscures much of the fine detail on the tooth surfaces. Enamel is missing from the anterobuccal surface of P3, posterobuccal corner of M1, anterolingual edge of M2, and the buccal and posterior occlusal surface of M3.

Lachrymal Only a small, triangular, anterior portion of the lachrymal is preserved. Two lachrymal foramina are present. The superior foramen is smaller than the inferior foramen. Dorsal to the superior lachrymal foramen, the lachrymomaxillary suture is highly convoluted and the lachrymal surface is strongly rugose. Anterior and ventral to the foramen, the lachrymomaxillary suture appears smooth and round, however, this may be the result of damage. The lachrymal foramen (2.0mm x 2.5mm) lies in a small rounded fossa, is slightly oval, directed laterally and has gently rounded edges.

Maxilla The maxilla extends from the posterior margins of the canine alveolus and the premaxilla/maxillary suture to 4 mm posterior to M3, measuring 50.3mm in total length. The region anterior to the infraorbital fossa is missing, exposing the ventral, finely perforated, surface of the canine alveolus and the premaxilla/maxillary boundary at the posterior edge of the I3 alveolus. The infraorbital foramen is obliquely oval (3.6 mm x 5.8 mm) and directed anteriorly. The maxilla surface between the infraorbital foramen and the jugal suture is dorsoventrally convex and 122 continues dorsally to form a smooth surface with a very slight anteorbital depression anterior to the lachrymal foramen. The malar tuberosity measures 26.0 mm dorsoventrally and is 14.5mm thick. The malar tuberosity is oval in section and forms an angle of approx. 60o to the vertical. The ventral surface of the maxillary wing of the zygomatic arch bears a small, rough, slightly elongated masseteric process. The infraorbital surface of the maxillary wing is gently concave, its lateral margin forming a rounded crest which ascends laterally along the zygomatic arch. The large infraorbital canal is enclosed by the maxilla and opens posteriorly just below the maxillopalatine suture.

Palate The maxillary portion of the palate is robust. The width of the palate from the medial base of the posterior root of P3 to the interpalatal suture is 16.7 mm. The depth of the bone at this medial suture measures 5.3 mm. There is a short diastema of 3.9mm between the canine and P1 alveolus. Small foramina are present along this diastema. The alveolus for P1 is oval in shape and is 1.8mm wide and 2.8mm long. The alveolus for P2 is 2.1mm wide and 2.3mm long and lies 3.7mm posteriorly to P1, opening onto the anteromedial edge of the anterior root of P3, producing a triangular alveolar outline.

Small foramina are located between the roots of P3 and between M1 and M2. Large, anteriorly-directed, palatal foramina are located 6mm medial to P1 and 13mm medial to the lingual root of M1. Two shallow fossae are located medial to the base of P3 and M1. Each fossa is bordered anteriorly by a palatal ridge that runs in an anteromedial direction. The anterior-most ridge runs parallel to the line of the diastema from the base of P3 to 5mm posterior to the canine alveolus. A second ridge runs anteromedially from the base of the posterior root of P3 towards the medial border of the maxilla. The posterolateral margin of the anterior palatal fenestra is preserved but the medial edge and anterior medial septum is missing. 123

Dentition Canine: the anterodorsal portion of the canine alveolus is missing, exposing the floor of the tooth socket. The anterior margin of the incomplete alveolus is 4.9mm wide and suggests that the complete alveolus would have had a slightly wider diameter. Posteriorly, the alveolus narrows and continues for approximately 13mm to just anterior to the infraorbital foramen. The shape of the alveolus indicates that the root of the canine curved horizontally as it ran backwards.

P1 and P2: the tip of P1 is broken off at the level of the palate exposing the dentine and the root canal. The basal remains of the tooth narrows towards its broken surface suggesting the missing crown probably tapered to a point. P2 is missing. Its alveolus is located at the anteromedial base of P3 and is separated from P1 by a short diastema.

P3: is blade-like and dominated by a longitudinal crest. The posterior buccal corner is missing and the enamel is damaged on the anterior buccal surface and around the base of the crown. The anterior half of the tooth is much narrower (6.0mm) than the posterior half (8.4mm), the basal region of the latter being inflated buccally and lingually, producing a subtriangular occlusal outline. The anterior and posterior cusps are subequal in height. The longitudinal crest joining these two cusps is worn, exposing a broad strip of dentine. From the anterior cusp an anterior crest runs towards the front edge of the tooth turning slightly lingually and thickening near the base of the crown. A prominent anterolingual crest ascends from just above the exposed dentine at the anterior cusp and broadens near the base of the crown. On the buccal surface a small anterobuccal crest ascends from the anterior cusp and a short crest runs anteriorly from the posterior cusp. The P3 is anchored by an anterior and posterior root, the latter being larger than the former and supporting slightly more than 50% of the tooth.

M1: has a bulbous appearance and is triangular in occlusal outline. The anterior buccal edge is missing and enamel on the posterior buccal surface of the crown has 124 been lost. The paracone is taller than the metacone. These two cusps are linked by a longitudinal crest that has a shearing facet on its lingual face and on which a thin strip of dentine is exposed. From the paracone an anterior crest runs to stylar cusp B at the anterior edge of the tooth. This crest is contiguous with the longitudinal crest of P3. From stylar cusp B a transverse anterior crest ascends lingually merging with the anterior margin of the tooth about half way between the paracone and protocone. A vertical crest ascends the lingual face of the paracone terminating at a notch at the level of the trigon basin. The protocone is prominent and elevated and the crown above this cusp is deep and bulbous. A postprotocrista runs posteriorly to a metaconule. A short postmetaconulecrista runs posteriorly to merge with the posterior edge of the tooth. The metacone is worn, exposing a small area of dentine. A posterolingual crest runs from the metacone toward the metaconule terminating at a small notch. On the buccal surface, a stylar basin runs between the paracone and metacone. The lateral edge of the stylar basin and the surface of the trigon basin are crenulated. M1 has three roots, a large lingual root and two smaller anterior and posterior buccal roots.

M2: is triangular in occlusal outline. Enamel is missing from the lingual anterior margin. The paracone is taller than the metacone. An anterior crest runs from the paracone to stylar cusp B at the anterior edge of the tooth. A short, narrow, anterior cingulum runs across the anterior margin to connect with stylar cusp B. On the lingual face of the paracone, a robust anterolingual crest ascends, terminating at a notch at the level of the trigon basin. From this notch a preprotocrista runs lingually to an elevated protocone. The base of the crown below the protocone is expanded giving the tooth a bulbous appearance. A postprotocrista runs posteriorly to a reduced metaconule. A posterolingual crest ascends the lingual face of the metacone. A small, narrow basin lies between this crest and the posterior margin of the tooth. The stylar shelf is broad anteriorly and narrows posteriorly. A buccal crest traverses the stylar shelf from stylar cusp B to the base of the metacone and forms a narrow anterior stylar basin. The trigon basin is concave and crenulated. Three roots anchor M2; the lingual root is largest while the anterior buccal root is larger than the posterior buccal root. 125

M3: is tritubercular and triangular in occlusal outline. The crown is relatively tall and bulbous. Enamel is missing from the buccal and posterior occlusal surfaces. The posterior half of the paracone is missing as is the metacone. The paracone was probably taller than the metacone. The protocone is rounded and prominent. A short preparacrista runs from the paracone to the anterior edge of the molar. A very narrow anterior cingulum (crest) runs along the anterior margin between stylar cusp B and the protocone. An anterolingual crest ascends the lingual face of the paracone, running lingually to join the preprotocrista. From the protocone a postprotocrista runs posteriorly toward the damaged metacone. The trigon basin is subtriangular and lined with crenulations. There are three roots, the lingual being the largest followed by the anterior and then the posterior. The crown is very bulbous.

Discussion

The maxilla, QM F23446, resembles Wakaleo hilmeri in a number of respects: a) its dentition is only slightly larger than that of W. hilmeri; for example, its P3 length of 12.6mm compares closely to a mean length of 12.0 mm for the P3 of W. hilmeri;b) although the maxilla seems relatively robust, occlusion against mandibles of W. hilmeri (AR 19249, QM F31376 and QM F30465) results in a remarkably close fit; c) QM F23446 and some specimens of W. hilmeri exhibit a full complement of premolars (P1-3), the only ones to do so within the genus; d) the position of the anterior palatal fenestra is similar to that seen in W. hilmeri, i.e. the fenestrae extend beyond the canine alveolus; e) there is a strong possibility of age equivalence between D Site and the Riversleigh System A deposits from which W. hilmeri has been recovered.

The most significant difference between QM F23446 and W. hilmeri is its lack of a fourth molar. Because molar number is an important specific diagnostic feature within Thylacoleonidae, the significance of the loss of M4 in QM F23446 requires careful consideration. It is possible that the absence of this molar may be an example of a dental abnormality. In marsupials, missing teeth mostly occur in the region of the premolars and rarely in the molar region (Archer 1975). Additionally, where non- 126 traumatic molar loss does occur, other teeth are usually also missing (ibid). In QM F23446 there is no evidence of an alveolus or malformed alveolus or tooth crypt posterior to M3 and all three premolars are present, which suggests that M4 absence is unlikely to be an abnormality. Within Wakaleo, loss of the fourth molar is characteristic of W. vanderleueri, W. oldfieldi and probably also W. alcootaensis. Therefore, employing this character as the sole criterion for species determination, QM F23446 could possibly be referred to any of those three species. Of these species, the most likely is W. oldfieldi because of its smaller size and older age. Comparison of QM F23446 with a partial maxilla and upper cheekteeth referred to W. oldfieldi (QM F11852 and QM F23427) indicates that its P3 and M1 are slightly smaller in length and breadth. In contrast, occlusion of QM F23446 against mandibles of W. oldfieldi (SAM P17925 and QM F20895) produces a very poor fit, the mandibles being significantly oversized to provide adequate occlusion. The mismatch between the maxilla and the mandibles is sufficiently large that it is difficult to consider them as conspecific. Apart from molar number, QM F23446 shares a number of other features with W. vanderleueri; these include an anteorbital depression, deep malar tuberosity and an incipient zygomatic masseteric process. However, three factors argue strongly against the referral of this specimen to W. vanderleueri. The first factor is size - the P3 of QM F23446 is significantly smaller in absolute size than the P3 of W. vanderleueri, being approximately two-thirds the length of the latter. The second factor is the size difference between the anterior palatal fenestra which in QM F23446 extends posteriorly to beyond the posterior margin of the canine and contrasts with that of W. vanderleueri in which the fenestrae end well short of the canine alveolus. The third factor is age; the postulated age of D Site sediments is late Oligocene (Tedford 1967, Archer et al. 1997) while W. vanderleueri is known from deposits estimated to be mid Miocene (Plane and Gatehouse 1968, Stirton et al. 1968) and early late Miocene (Myers et al. 2001) in age (see Chapter 14).

Despite the similarities of QM F23446 to W. hilmeri as noted above, the significance of the absence of M4 is difficult to dismiss. Consequently, a referral of this specimen to W. hilmeri would be conservative if not tentative. An alternative hypothesis that 127 should be considered is that QM F23446 represents an intermediate or transitional form between W. hilmeri and W. oldfieldi. Its slightly greater size and robustness (in comparison to W. hilmeri) and the presence of plesiomorphic (P1-3) and apomorphic (loss of M4) features support this transitional scenario. This specimen may represent a further, earlier, stage in the oldfieldi/vanderleueri morphocline. 128

CHAPTER 7

NEW MATERIAL OF Wakaleo oldfieldi AND Wakaleo vanderleueri

7.1 Introduction

The only previously known specimens of Wakaleo oldfieldi Clemens and Plane, 1974 were a left dentary (the holotype SAM P17925) and some isolated lower teeth (UCMP102677 and UCMP102678), all collected from the Leaf Locality, Wipajiri Formation, South Australia. Isolated upper molars of this species were collected from this locality by M. Archer in 1981 and are described here. Wakaleo vanderleueri Clemens and Plane, 1974, in contrast, is the best known species of Wakaleo with a complete skull, and maxillary and dentary specimens having been collected from the Small Hills Locality near Bullock Creek, Camfield Beds, Northern Territory (Murray, Wells and Plane 1987; Megirian 1986; Murray and Megirian 1990). Although relatively well known compared to other species of Wakaleo, its fossil record is by no means abundant. Additional material of both W. oldfieldi and W. vanderleueri has been recovered from sites in the Riversleigh World Heritage Area, northwestern Queensland, increasing the geographical ranges of both taxa.

This new material includes previously unknown or poorly preserved teeth which indicate significant morphological interspecific differences between these otherwise only weakly differentiated taxa. Measurements of the teeth of these specimens as well as those of comparative specimens are shown in Table 7.1.

7.2 New material of Wakaleo oldfieldi

The most complete specimen from Riversleigh referred to Wakaleo oldfieldi is a left dentary (QM F20895) from the Cleft of Ages Site (COA) which retains P3 and the complete molar row (M1-3). This dentary together with an isolated tooth (QM 129

F36437) provide the first information about the morphology of M3. The upper dentition of W. oldfieldi has not previously been described. A maxillary fragment recovered from Jim’s Jaw Site that contains P3 and M1 has been assigned to this species as has two isolated upper teeth (M2 and M3) from COA Site. Additional isolated upper molars (M1 and M2) collected from the type locality in South Australia (Leaf Locality, Lake Ngapakaldi; Stirton et al. 1967) are also referred to this species.

SYSTEMATICS

Family THYLACOLEONIDAE Gill, 1872 Genus Wakaleo Clemens and Plane, 1974

Wakaleo oldfieldi Clemens and Plane, 1974

Holotype: SAM P17925, a left dentary with I1, P3, M1-2 from the Kutjamarpu Local Fauna, Leaf Locality, UCMP loc. V6213, Lake Ngapakaldi, Wipajiri Formation, South Australia (Stirton et al. 1967).

Referred material: The following specimens were collected from the Leaf Locality, Kutjamarpu Local Fauna: AR 19248, a left M1. AR 3272, a right M2.

The following specimens were collected from COA Site: QM F20895, a left dentary retaining P3, M1-3 and alveoli for I1 and a single-rooted tooth anterior to P3. QM F20304, a left dentary fragment containing M1. QM F36437, a right dentary fragment containing M3. QM F31295, a left M1. QM F31398, a right M2. 130

A

B

C

Figure 7.1: Wakaleo oldfieldi, QM F20895, left dentary with P3 M1-3, COA Site. A. buccal view. B. lingual view. C. stereo occlusal view. Scale bars = 10 mm 131

QM F31294, a left M3. QM F23428, a left premaxillary and maxillary fragment containing P3 and M1 and alveoli for I1-3 from Jim’s Jaw Site. QM F23427, juvenile left and right P3 and M1 from Helicopter Site. QM F24745, a left M2, from Keith’s Chocky Block Site (KCB). QM F36466, a left M3, from Golden Steph Site. QM F23452, a right M1 from Henk’s Hollow Site.

Localities and ages of referred material: Kutjamarpu Local Fauna, Leaf Locality, eastern strand of Lake Ngapakaldi, Tirari Desert, South Australia. This site is estimated to be late Oligocene in age (Woodburne et al. 1993). The following sites are found within the Riversleigh World Heritage Area, northwestern Queensland: Helicopter Site is regarded to be a Riversleigh System B deposit and is estimated to be early Miocene in age; Jim’s Jaw Site, Golden Steph Site, Keith’s Chocky Block Site, Henk’s Hollow Site (all System C deposits) and COA Site are estimated to be middle Miocene in age (Archer et al. 1997).

Revised specific diagnosis: Differs from W. pitikantensis and W. hilmeri in being 4 larger and lacking M /4. Differs from W. vanderleueri in having P3 absolutely smaller in length and M3 with a subrectangular outline and a crenulated talonid basin. Additional possible characters: P3:M1 length, less than 1.2.

Description and comparisons

The dentary QM F20895 (Fig. 7.1 and 7.2) retains P3, M1-3, and alveoli for I1 and a single-rooted tooth anterior to P3. It lacks most of the posterior structures including the coronoid process, the masseteric fossa and the angular process. The dentary was fractured longitudinally and bears a thin line of epoxy through the tooth row and through the base of the horizontal ramus. In terms of size and morphology, this dentary and its dentition is nearly identical to that of the holotype (SAM P17925 in Appendix 1: Table A). Additionally, its undamaged M2 provides a more complete 132 picture of the morphology of this molar. The only previously known M2 of W. oldfieldi was an isolated tooth (UCMP 102677) described with the holotype (Clemens and Plane 1974). That tooth was worn and missing fragments of enamel. The M2 of QM F20895 also exhibits wear as well as repair to its lingual surface.

Comparison with UCMP 102677 indicates that although the Riversleigh tooth is slightly larger, they are essentially similar. The trigonid basin of UCMP 102677 lacks a portion of its posterobuccal margin. In QM F20895, the posterior margin of this basin forms a transverse crest that at its buccal and lingual extremities forms the high points of the basin's rim. The talonid basin of UCMP 102677 also lacks part of its posterobuccal and lingual margins. In QM F20895 the rim of the talonid basin is unbroken. The lingual margin of the basin is slightly taller than the buccal margin and rises to a high point at the posterolingual corner of the rim. The buccal margin of the talonid basin rises posteriorly to a lower and more obtuse point at the posterobuccal corner of the basin. The buccal surface of the crown below this point bears a slight vertical ridge. The posterior edge of the basin is formed by two short crests which descend towards each other from these lingual and buccal high points. Morphologically, the M2 of W. oldfieldi is basically similar to that of W. vanderleueri although specimens of the latter show a wide range of variation in the development of the crests on the trigonid and in the width and shape of the talonid basin.

The description of M3 is based on QM F20895 and an isolated tooth, QM F36437 (Fig. 7.3). Both teeth exhibit partial damage and show moderate to heavy wear. QM F20895 bears a longitudinal fracture through its crown and has a line of epoxy running across its anterior edge and along its buccal occlusal edge. The buccal side of the crown is lacking enamel from its posterior half and also bears a fracture across the anterior two-thirds of its surface. Fractures also run through the base of the crown and through both roots of QM F36437. Small pieces of enamel have been lost from the lingual base and the anterobuccal base of its crown. 133

36.5

12.4 10.7 8.3 5.1

23.2

84.9

Fig. 7.2: Measurements (mm) of QM F20895, Wakaleo oldfieldi left dentary; COA Site.

Fig. 7.3: Wakaleo oldfieldi, QM F36437, right dentary containing M3; 134

The M3 of QM F20895 is much smaller than M2 and measures 5.1mm in length and 4.7mm in width. It appears to have a subrectangular outline, however, this may be a result of displacement of the crown after fracturing. QM F36437 measures 4.9mm in length and 4.5mm in width and differs in having a slightly squarer outline. The anterior edge of QM F20895 is relatively flat and conforms to the shape of the posterior edge of the M2. In profile, the posterior end of the tooth is distinctly taller than the anterior end. In occlusal view, the buccal and posterior sides of the molar are rounded while the lingual edge is relatively straight. The occlusal surface of the tooth bears a shallow, oval-shaped talonid basin. The lingual edge of the basin is relatively straight and well worn. The buccal edge of the basin bows buccally and shows moderate wear. The surface of the basin appears to have been rugose. The tooth has two roots of approximately equal size that are both directed in a slightly posterior direction. The crown morphology of QM F36437 is essentially the same but differs in having a slightly broader and rounder talonid basin. In addition, unlike QM F20895, the highest point on the tooth appears to be at its anterobuccal corner, however, the anterolingual corner of the tooth exhibits a small triangular area of dentine and appears to have been affected by both breakage and wear. The lingual margin of the basin is slightly lower than the buccal margin, and is comparatively more heavily worn.

The M3 of W. oldfieldi is absolutely shorter in length but proportionately wider than that of W. vanderleueri. This is the only molar that exhibits a distinct morphological difference between the two taxa. While the crown of W. oldfieldi is dominated by a crenulated talonid basin, that of W. vanderleueri completely lacks this basin. In addition, in occlusal view, the outline of the crowns also differ with that of W. oldfieldi being subrectangular, having relatively broader anterior and posterior margins, while the crown of W. vanderleueri is narrower and is more fusiform in shape.

Upper dentition QM F23428, from Jim’s Jaw Site, consists of two fragments of the rostrum, one of the left premaxilla, the other of the left maxilla (Fig. 7.4). This material preserves a 135

Fig. 7.4: Wakaleo oldfieldi, QM F23428, left maxilla fragment; Jim’s Jaw Site. Stereo pair. Scale bar = 5mm

Fig. 7.5: Wakaleo oldfieldi, AR19248, M1; Kutjamarpu LF. Stereo pair. Scale bar = 5mm

A B

Fig.7.6: Wakaleo oldfieldi M2: A. QM F31398, COA Site; B. AR 3272; Kutjamarpu LF. Stereo pair. Scale bar = 2mm

Fig.7.7: Wakaleo oldfieldi QM F31294, M3; COA Site. Stereopair. Scale bar = 2mm 136 badly fragmented I1 and alveoli for I1-3, and P3, M1 and the anterobuccal corner of M2. The description of M1 is supplemented by AR19248 from the Leaf Locality. A small, posterior portion of the premaxilla that extends from the root of I1 to the premaxillomaxillary suture is preserved. This fragment is approximately 35mm long and contains the fractured remains of I1, alveoli for I 2-3 and the anterior half of the alveolus for the canine. The remnant of I1 indicates that it was much larger than the other two incisors and was relatively long. Its alveolus had an anteroposterior width of approximately 10mm. The ventral margins of both the I2 and I3 alveoli are damaged. The I2 alveolus is smaller than I3, is oval in shape and anteroposteriorly compressed. The I2 alveolus measures 2.4mm long and 4mm wide at its ventral margin. The I3 alveolus lies directly behind I2, is relatively large and round, and measures 4.1mm long and 5mm wide.

The canine alvelous lies approximately 4mm posterior to I3. The anterior half of this alveolus is preserved and measures approximately 5mm in width. The anterior wall of the alveolus is formed by the heavily pitted posterior surface of the premaxilla and indicates that the canine was positioned at the junction of the premaxilla and maxilla.

P3: This tooth is badly fractured and much of the anterobuccal part of the crown is missing. The basal part of the crown below the posterior cusp is also missing. It measures 15.1mm in length and has an anterior width of 6.7mm and posterior width of 7.2mm. It resembles the P3 of W. hilmeri and W. vanderleueri, possessing a longitudinal blade between an anterior and posterior cusp. This blade is worn and a strip of dentine approximately 1mm wide is exposed along most of its length. The anterior crest thickens at its basal extremity much more so than in W. vanderleueri. On the lingual face of the crown a well developed anterolingual cuspule lies below the anterior cusp. On the buccal surface, a short posterobuccal crest connects with the posterior end of the longitudinal crest.

M1: This molar in QM F23428 is missing enamel from its posterolingual edge and much of the roots below the buccal half of the crown. AR 19248 (Fig.7.5) is 137

Table 7.1: Cheek dentition measurements of Wakaleo oldfieldi and W. vanderleueri (mm). ^ = data from Clemens and Plane (1974); * = data from Murray and Megirian (1990); [ ] = estimated measure; a=alveolus measure.

Lower Cheek Teeth Riversleigh P3 P3 P3 M1 M1 M1 M2 M2 M2 M3 M3 P3:M1 Species site/ length anterior posterior length anterior posterior length anterior posterior length width Specimen no. Local Fauna width width width width width width W. oldfieldi QM F20895 COA 1 12.4 5.6 6.7 10.7 8.6 7.7 8.3 6.6 5.3 5.1 4.7 1.15 QM F20304 COA 2A 10.1 8.8 6.9 QM F31295 COA4 11.6 8.2 7.4 QM F36437 COA2A 4.9 4.5 QM F23452 Henk’s Hollow 11.9 7.3 6.5 SAM P17925^ Kutjamarpu 12.4 6.2 7.8 10.4 8.7 7.2 7.5a 6.3a 6.1a 5.0a 5.5a 1.19 UCMP 102677^ Kutjamarpu 7.5 6.3 W. vanderleueri QM F36453 Encore 14.3 6.8 7.7 12.1 8.6 7.7 8.0 6.1 5.4 5.6 3.8 1.18 QM F41204 Encore 15.5 6.4 7.7 1.30 QM F36454 Encore 8.4 5.1 QM F29712 Encore 12.5 8.5 7.6 QM F36281 Encore 7.5 QM F51409 Encore 9.2 6.4 6.1 QM F52182 Encore 6.3 4.2 NTM P87108-5* Bullock Creek [14.5] 5.7 7.2 12.0 8.2 7.4 9.8 6.3 6.1 [1.20] NTM P87108-6* Bullock Creek [14.2] 6.3 7.6 11.3 [8.5] 7.7 9.4 6.3 6.8 [1.25] NTM P85553-4* Bullock Creek 13.9 [7.1] 9.6 [6.5] 7.7 5.5 5.2 [5.9] 1.44 NTM P8695-97* Bullock Creek 8.2 6.6 6.2 6.0 4.7 NTM P8695-96* Bullock Creek 5.8 4.1 CPC 13527* Bullock Creek 11.7 7.9 7.7 138

Table 7.1 cont.: Cheek dentition measurements of Wakaleo oldfieldi and W. vanderleueri (mm).

Upper Cheek Teeth Species Riversleigh site/ P3 P3 M1 M1 M2 M2 M3 M3 Specimen No. Local Fauna length width length width length width length width W. oldfieldi QM F23427 Helicopter 13.2L 6.4 10.1 10.0 QM F11852 Jim’s Jaw 14.1 7.5 10.6 10.3 QM F31398 COA 4 7.3 8.4 QM F31294 COA 2A 5.1 5.6 QM F36466 Golden steph 5.5 5.2 QM F24745 KCB 7.5 8.1 AR 19248 Kutjamarpu 11.6 10.6 AR 3272 Kutjamarpu 6.5 7.1 W. vanderleueri QM F30384 Encore 8.5 CPC 26604* Bullock Creek 17.9 9.2 11.2 11.0 7.0 9.5 5.0 5.5 NTM P87103-9* Bullock Creek 7.5 8.4 5.6 5.8 139 undamaged and exhibits less wear than QM F23428. Both molars are roughly triangular in outline. The paracone is the tallest of the three major cusps. It connects with the preparacone by a short anteriorly-running crest. From the preparacone a short crest runs down the anterobuccal corner to the base of the crown. A second crest runs lingually from the preparacone across the front margin of the crown, becoming weaker and merging with the crown approximately half way to the protocone. A postparacrista runs posteriorly, descending to a notch where it meets the short, ascending, premetacrista. These crests together with the preparacrista form a longitudinal shearing crest that would be contiguous with the shearing crest of P3. The metacone is slightly shorter than the paracone. A short postmetacrista descends posteriorly to the rear margin of the tooth. A distinct stylar shelf lies along the buccal edge of the crown. This shelf is delineated by a crest which descends from the paracone and then runs posteriorly, passing approximately 2mm below the metacone and connects with a small vertical crest at the posterobuccal corner of the crown. A number of small notches incise the edge of the shelf about half way along its length. The protocone is distinctly lower than the other cusps. Its base is broadly rounded. Lingual to the paracone and running along the anterior margin of the trigon basin is the remnant of a short crest that terminated near the notch at the anterior border of the crown. A weakly developed metaconule is located at the posterolingual margin of the trigon basin and is linked to the protocone by a gently rounded crest. The metaconule connects with the metacone by a posteriorly-bowed crest that also forms the posterior wall of the trigon basin. A small crest runs from the metaconule around the posterior edge of the tooth and connects with the postmetacrista. The trigon basin is well crenulated. The tooth has three roots. The posterobuccal root is the largest and is antero-posteriorly compressed. The anterobuccal root is slightly smaller than the posterobuccal, and the lingual root is the smallest.

M1 of W. oldfieldi shows the triangular shape generally characteristic of species of Wakaleo. It is essentially similar in size and morphology to the M1 of W. vanderleueri (CPC 26604 in Appendix 1: Table A) although slight differences are evident in the relative height of the paracone and the region of the tooth posterior to the metaconule. The paracone of both W. oldfieldi specimens is relatively lower and closer to the protocone and metacone than in W. vanderleueri. In the latter species 140 the paracone is very tall in relation to its protocone. The posterior region of W. oldfieldi (AR19248) shows a slightly greater development, producing a more angular posterior margin than is evident in W. vanderleueri. In the latter, the posterolingual edge of the crown curves in a gently curving arc. The M1 of W. oldfieldi is distinctly larger than that of Wakaleo hilmeri.

M2: This molar is represented by two isolated teeth, AR 3272 from Leaf Locality and QM F31398 from COA Site (Fig. 7.6). The former tooth is heavily worn and exhibits the remnants of most of the topological features of the crown. Because of the extent of wear in AR 3272 the description is based mainly on QM F31398. Although technically quadrilateral, this tooth is roughly subtriangular in outline, and is much broader anteriorly than posteriorly. It measures 7.3mm in length and has a maximum width of 8.4mm. The paracone and protocone are subequal in height. The crown below the paracone is inflated buccally. A small valley on the buccal surface of the crown marks where this part of the crown meets the crown supporting the metacone. A small preparaconule lies anterobuccally of the paracone to which it is linked by a very short crest. From the preparaconule a crest runs along the front edge of the tooth connecting with the preprotocrista about half way along the anterior margin of the trigon basin. A crest runs lingually from the paracone, descending and then slightly ascending for a short distance, before meeting with the preprotocrista and anterior crest. These crests form the anterior margin of the trigon basin. The crest from the paracone runs almost parallel to the anterior crest and these form the walls of a narrow trough at the anterior edge of the crown. The crown below the protocone is quite bulbous. A postparacrista runs posteriorly from the paracone and connects with a short premetacrista about two-thirds of the way along the buccal edge of the trigon basin. Two small crests, probably remnants of the stylar shelf, lie on the buccal surface of the crown below the paracone. The largest of these crests runs posteriorly from the preparaconule, slightly ascending the buccal surface. The shorter crest ascends the buccal surface approximately two-thirds of the way along the postparacrista. The metacone lies near the posterobuccal corner of the tooth and is slightly shorter than the other cusps. A short crest runs from the metacone towards the posterobuccal corner of the crown. The posterior edge of the crown is rounded. The posterior margin of the trigon basin is defined by a crest that runs lingually from 141 the metacone and then, at about one-third of its length, angles anterolingually towards the protocone. The trigon basin is relatively deep and heavily crenulated. Three roots anchor the tooth.

The morphology of QM F31398 is essentially similar to the M2 of other species of Wakaleo. It is very similar to but slightly larger than that of Wakaleo hilmeri. It also differs from that taxon in having a more prominent protocone that results in the molar having a more triangular outline. QM F31398 is very similar in dimensions and general shape to the M2 of W. vanderleueri, although it has a much flatter and broader posterior margin than that species. Although the M2 in the comparative specimen of W. vanderleueri (CPC 26604) is heavily worn, the remaining morphology indicates the presence of the same main features that are present in QM F3139, i.e. large trigon basin, narrow anterior trough, buccal longitudinal crest and prominent protocone.

M3: this molar is represented by an isolated tooth, QM F31294 which shows only slight wear (Fig. 7.7). This tooth is triangular in outline and measures 5.1mm in length and 5.6mm in width. The crown consists of subtriangular crenulated basin. The paracone and metacone lie at the anterobuccal and posterobuccal corners, respectively. The tooth has three roots that are subequal in size. The roots splay out slightly beyond the margin of the crown.

QM F31294 is most similar in shape and absolute size to that of W. vanderleueri. It is also very similar in its shape and morphology to the M3 of Wakaleo hilmeri although it is slightly larger.

Comments Clemens and Plane (1974) described the rim of the trigonid basin of M2 as being contiguous and lacking cusps. The Riversleigh M2 (QM F20895) also shows this structure and although prominent individual cusps are not discernible, it is possible that the high points on the basin's posterior buccal and lingual corners represent respectively, the protoconid and metaconid. Similarly, the high points on the 142 posterolingual and posterobuccal corners of the talonid basin possibly represent respectively, the entoconid and the hypoconid. Further support for this hypothesis comes from the buccal surface of the crown which bears a slight vertical ridge below the posterobuccal corner and suggests buttressing of a previously more prominent cusp in this region of the crown.

Although no upper dentition was previously known for W. oldfieldi, the similarities in size and morphology of its lower dentition to that of W. vanderleueri (Murray and Megirian 1990), allowed the reasonable assumption that its upper dentition would likewise be similar to, but probably slightly smaller than that species. This assumption is supported by the Riversleigh material. The specimens from Jim’s Jaw Site and Helicopter Site although similar in morphology to W. vanderleueri, differ in the size of P3, which is significantly shorter than that of W. vanderleueri. The isolated upper molars from COA are very similar in dimensions and morphology to those of W. vanderleueri (Appendix 1: Table A) and consideration had to be made to the possibility that these molars were of that species. A combination of factors however, supports the referral of these isolated teeth to W. oldfieldi. Firstly, all the Wakaleo lower teeth recovered from this site have been assigned to this taxon (on the basis of morphology and the size of P3). The second factor is the probable older age of this site. COA Site has not yielded the range of derived mammalian taxa typical of those recovered from the early late Miocene Riversleigh locality (Encore Site) that has produced undisputed W. vanderleueri material (Myers et al. 2001; see below). Of the few species to be described from COA, the vombatid Rhizophascolonus crowcrofti has also been recovered from South Australia’s late Oligocene Kutjamarpu LF. Further support for an older age for COA Site comes from the relatively plesiomorphic nature of its macropodoids (Cooke 1977) compared to those from Encore Site (Myers et al. 2001, Kirkham 2004).

7.3 New material of Wakaleo vanderleueri

Previously, all specimens of Wakaleo vanderleueri have been collected from the Bullock Creek Local Fauna, Camfield Beds, Northern Territory. All the dentaries of 143

W. vanderleueri recovered from that fauna exhibit various amounts of damage to the cheek tooth row (P3-M3) (Megirian 1986, Murray and Megirian 1990). The new Riversleigh dentary, QM F36453, is important because it preserves a completely undamaged cheek tooth row and is the first known to contain a complete P3. In addition, the relatively unworn nature of its M3, together with another isolated specimen, permits the description of this molar for the first time. Other noteworthy specimens in the Riversleigh sample include a canine, a tooth also previously unrecorded for this species, and a molar from a juvenile individual. All Riversleigh specimens that have been referred to this species have been collected from Encore Site.

Wakaleo vanderleueri Clemens and Plane, 1974

Holotype: CPC 13527, a right dentary fragment containing M1, from the Small Hills Locality, Bullock Creek Local Fauna, Camfield Beds, Northern Territory.

Referred material: the following specimens are from Encore Site: QM F36453, a left dentary containing P3, M1-3 QM F29712, a left dentary containing M1 QM F36454, a left dentary containing M2 and partial I1 and M1 QM F52182, an isolated left M3 QM F41204, an isolated right P3 QM F30384, a P3 (anterior half only) QM F36281, a left M1(juvenile) QM F31352, a left C1 144

A

B

C

c

Figure 7.8: Wakaleo vanderleueri QM F36453, left dentary; Encore Site.

A. buccal view B. lingual view C. stereo occlusal view Scale bars = 20mm 145

Locality and age of referred material: Encore Site from the Riversleigh World Heritage Area, northwestern Queensland is estimated to be early late Miocene (9myr) in age (Myers et al. 2001).

Revised specific diagnosis: This revision relates only to features of the dentary: W. vanderleueri differs from W. oldfieldi in having P3 larger absolutely and relative to M1; and in having M3 oval in outline, bearing a small longitudinal ridge and lacking a crenulated talonid basin.

Description and comparisons

QM F36453 is a left dentary containing P3, M1-3 (Fig.7.8 and 7.9, Table 7.1). This specimen has been referred to W. vanderleueri on the basis of its size and the morphology of M3. The molars and especially the P3 in this dentary are much larger than those in W. oldfieldi and are very similar in size to those recovered from the Camfield Beds (Bullock Creek LF; Table 7). The morphology of the dentary and teeth of W. vanderleueri has previously been described in Megirian (1986) and Murray and Megirian (1990). Consequently, the following descriptions will focus on those teeth that were previously incomplete or undescribed, i.e. P3 and M3.

Description of the P3 is based on QM F36453 as well as an isolated tooth, QM F41204 (Fig. 7.10). The Riversleigh P3 specimens support Megirian's (1986) observation of a P3 from the Bullock Creek LF that lacked its lingual side, i.e. it appears to show little difference to the P3 of W. oldfieldi. The P3 of QM F36453 has a longitudinal blade that descends posteriorly from an apex above the anterior root, curving lingually approximately two-thirds of the way along its length, and then rises slightly before reaching the posterior apex of the tooth. The anterior and posterior lingual crests that descend from their respective apices are similar to those seen in W. oldfieldi. The anterior lingual crest is variably developed in the Riversleigh specimens. In QM F36453 there is evidence of a very weak, angled ridge near the base of the crest while in QM F41204 the upper and more-vertical part of the crest is relatively 146

39.9 14.3 12.1 8.0 5.6

23.3

72.4

Figure 7.9: Dimensions of Wakaleo vanderleueri, QM F36453, left dentary; Encore Site. Buccal view. Measurements in mm.

Figure 7.10: Wakaleo vanderleueri, Fig. 7.11: Wakaleo vanderleueri, QM F52182, M3; QM F41204, left P3; Encore Site. Encore Site. Stereo pair. Scale bar = 2mm Stereo pair. Scale bar = 5mm

A B

Figure 7.12 Wakaleo vanderleueri QM F31352, left canine; Encore Site.

A. buccal view. B. lingual view. Scale bar = 5 mm 147 prominent. The relative sizes of the two roots are also similar to those of W. oldfieldi with the posterior root being slightly larger than that of the anterior.

The M3 in QM F36453 is well preserved and its crown shows little wear (Fig. 7.8c). In occlusal view the tooth is oval and measures 5.6mm long and 3.8mm wide. QM F52182 (Fig. 7.11) is larger, lacks roots and is also relatively unworn. In profile, M3 has an anterior apex and becomes lower posteriorly. A poorly developed longitudinal crest descends from the anterior apex and runs along the centre of the crown. A short anterior crest runs from the anterior apex in an anterolingual direction towards the base of the crown. In QM F52182 a short, weak crest appears to link the anterior crest to the longitudinal crest. The development of the longitudinal crest is slightly different between these teeth. In QM F52182 the posterior end of the crest ascends slightly to a relatively short posterior apex. From this point a short lingual crest curves around the posterolingual margin of the crown. In QM F36453 the longitudinal crest becomes very weak in the posterior third, appearing to merge with the crown at the rear of the tooth. A tiny cuspule lies at the posterior end of the tooth. The enamel in this region of the tooth appears rough and it is possible that the weakness of the longitudinal crest in this region may be the result of wear. Both the lingual and buccal surfaces of the crown are broadly rounded. The occlusal surface forms a small depression on the lingual side of the longitudinal crest. The lingual side of the longitudinal crest is slightly steeper than the buccal side and bears a number of small corrugations that are most prominent adjacent to the depression. There are two roots of subequal size. The posterior root extends posteriorly beyond the posterior margin of the crown.

Although the M3 of W. vanderleueri is present in two Bullock Creek specimens (NTM P8695-96 and NTM P8695-97), the crown morphologies of these teeth have been obscured by heavy wear. The least worn of these teeth, NTM P8695-96, shows a similar oval occlusal outline to the Riversleigh molars and an even stronger resemblance in profile. The M3 of W. oldfieldi differs markedly from that of W. vanderleueri in bearing a crenulated talonid basin, having a generally squarer occlusal outline, and having a relatively horizontal occlusal profile. 148

An isolated M1, QM F36281, is probably from a juvenile animal. This assessment is based on the appearance of the tooth which differs markedly from all the other specimens in a number of ways. Firstly, only the upper half of the trigonid carries the characteristic shiny brown enamel. The remainder of the crown is a lighter bone- colour and has a matt finish and is finely rugose. Secondly, all the major crests on the trigonid (i.e. the anterior, anterobuccal and anterolingual) are very prominent and are devoid of wear. Additionally, no wear is apparent on the remnant of the talonid basin. This appearance suggests that the tooth was only partially erupted.

A left canine (QM F31352, Fig. 7.12) is attributed to this species on the basis of its size and morphology. This tooth is bilaterally compressed, tapers gently to a single lingually-located point, and is slightly recurved. At the base of the crown it has an anteroposterior length of 7.8mm. Only the posterobuccal portion of the root (approximately 10.3mm long) is preserved. The lingual and buccal surfaces are separated by a lingually displaced crest that ascends both the anterior and posterior margins from the apical point. This results in the canine having a broad, rounded, buccal surface and a shallower and less rounded lingual surface. The crest on the anterior edge of the tooth forms a slight lip at its base and overall is much more developed than the posterior crest which is much flatter. Although now missing from a part of its base, it is evident that enamel completely covered the buccal surface of the crown. On the lingual surface enamel is evident only towards the point of the tooth. The morphology of this tooth is very similar to that of the canines found in Wakaleo hilmeri and differs markedly from the peg-like canines of Thylacoleo carnifex and T. crassidentatus.

Discussion

The new material from Riversleigh documents for the first time the morphologies of M3 of both Wakaleo oldfieldi and W. vanderleueri. The M3 of W. oldfieldi is relatively more plesiomorphic and although it has a reduced trigonid, it retains a crenulated talonid basin. In contrast, the M3 of W. vanderleueri is much more derived having lost any semblance of either the trigonid or talonid basins and is also relatively narrower. 149

Murray and Megirian (1990) compared dental material of Wakaleo vanderleueri from the Bullock Creek LF with W. oldfieldi from the Kutjamarpu LF and found minimal size differences between the two species and in some cases there was an overlap of tooth size. These authors were unable to find any discrete morphological characters by which to distinguish the two species and therefore concluded that the two species were parts of a chronocline in which the older and smaller species, W. oldfieldi, was considered the “systematically valid species” and the younger and larger species, W. vanderleueri, was designated a chronospecies. The gradual change of characters or groups of characters that is demonstrated by chronospecies often results in arbitrary specific designations of the transitional forms within a chronocline.

Identification made here of distinct morphological differences between the M3s of W. oldfieldi and W. vanderleueri allows for a clearer, less arbitrary distinction between the two species and supports the recognition of W. vanderleueri as a distinct “systematically-valid” species. The slight changes in morphology exhibited by both taxa over time, notably the increase in length of P3 and reduction in size of M2 and M3, are probably indicative of a Wakaleo morphocline which, as indicated in Chapter 4, probably began with W. pitikantensis and extended to W. alcootaensis. Similar morphoclines are also exhibited within the Thylacoleonidae by the Pliocene taxon, Thylacoleo crassidentatus and the Pleistocene taxon, T. carnifex (Archer and Dawson 1982).

Previous investigations of the lower cheek teeth of the two Wakaleo species found that the P3 of W. vanderleueri was consistently longer than that of W. oldfieldi (Murray and Megirian 1990). The new specimens from Riversleigh support this finding. The P3 of W. oldfieldi (QM F20895) has the same length (12.4mm) as that of the only other known specimen and is distinctly smaller than that of W. vanderleueri. The specimens attributed to W. vanderleueri have much larger premolars and contribute to an average P3 length of 14.5mm for this species. Although interspecific differences are identifiable in the M3, as 150 indicated by Murray and Megirian (1990), specific attribution of other isolated teeth (especially anterior cheek teeth) will remain difficult and arbitrary because of the similarities in morphology of these teeth and because of the smaller size differences between equivalent molars. This is exemplified by two isolated teeth, QM F23452 and QM F24745 from Henk’s Hollow Site and Keith’s Chocky Block Site, respectively. Although large enough to be referred to W. vanderleueri, these teeth have been referred to W. oldfieldi on the basis of these sites’ less derived mammalian faunas and earlier age estimations. Murray and Megirian (1990) indicate that such differences could be caused by a number of factors including individual variation, allometry, or possibly sexual dimorphism.

The different morphologies of the M3s of W. oldfieldi and W. vanderleueri possibly reflect functional and therefore also dietary differences between these taxa. In W. oldfieldi, the broad, crenulated basin-like structure of the M3 and the relatively broader talonid basin of M2 suggests a more crushing and grinding function for these teeth while the rather simple structure of the M3 of W. vanderleueri suggests a reduction of the grinding function and a posterior extension of the cutting or shearing function, a function that is more characteristic of the anterior region of the cheek tooth row. Approximation of the Encore dentary (QM F36453) with the skull of W. 3 vanderleueri (CPC 26604) indicates that M3 occludes against the basin-shaped M in such a manner that its longitudinal crest lies between the paracone and protocone and extends towards the metacone, the buccal side of the longitudinal crest working against the lateral edge of the trigon basin of M3. The protocone of M3 occludes with the anterolingual edge of the longitudinal crest and not against the anterior wall of the talonid basin as it does in M1 and M2. The protocone on both M1 and M2 in CPC 26604 exhibit heavy wear facets whereas the protocone on M3 is relatively unworn suggesting either a loss or reduction of function for this structure. CPC 26604 displays a distinct longitudinal pattern of wear that extends across all three molars from the protocone of M1 to the metacone of M3. This linear wear pattern is suggestive of a heightened importance of a cutting function for the molars in W. vanderleueri. Whether the functional morphology of W. oldfieldi reflects omnivory 151 or insectivory is open to speculation, but the changes in dental morphology exhibited by W. vanderleueri suggest a change to a more carnivorous diet. 152

CHAPTER 8

Lekaneleo roskellyae

8.1 Introduction

Rauscher (1987) described the thylacoleonid genus Priscileo on the basis of a fragmentary palate and some associated postcranial remains. That material was collected from the Ngapakaldi Local Fauna in the Etadunna Formation at Lake Pitikanta, South Australia. The taxon described, P. pitikantensis, was much smaller than any other previously known thylacoleonid and also exhibited a plesiomorphic dentition possessing four molars.

A well-preserved skull of an even smaller species of marsupial lion was recovered from Upper Site at Riversleigh. This new species was referred to the genus Priscileo on the basis of its size and because, like P. pitikantensis, it possessed four molars and was named P. roskellyae (Gillespie 1997, see Appendix 6). The discovery of a new plesiomorphic species of Wakaleo that also retains M4 has necessitated a review of the holotype of the type species of Priscileo, Priscileo pitikantensis (SAM P37719). As discussed in Chapter 4, the morphology of M2 of that specimen closely resembles that of species of Wakaleo and supports its referral to the genus Wakaleo. The generic name Priscileo Rauscher is therefore regarded as a junior synonym of Wakaleo Clemens and Plane.

Although SAM P37719 shows close affinities with species of Wakaleo, the holotype of P. roskellyae (QM F23453) does not. The morphology of the dentition and cranium show significant differences from those of species of Wakaleo and support its recognition as a separate genus. In this chapter a new generic name is provided and described for this taxon. Gillespie (1997) described the upper dentition from the cranium (QM F23453). Additional maxillary and mandibular material of this species 153 has since been recovered from other Riversleigh sites. This chapter describes the cranium of the holotype as well as the additional dentary and maxillary specimens of this species. Amendments are made to the diagnosis of both the genus and this species based on this material.

The great majority of the skull description is based on observation and measurement of the actual skull, however, prior to the writing of the description the specimen was loaned for casting and unfortunately during this process some areas of the skull, in particular the palate, vomers, ethmoidal region, and right auditory region, were badly damaged. Consequently, some of the observations of these areas have come from photographs and detailed drawings made of the skull prior to it sustaining damage. These photographs and drawings have been included to provide the maximum information regarding the original features of the skull.

8.2 Systematics

SYSTEMATICS

Order DIPROTODONTIA Owen, 1866 Suborder VOMBATIFORMES Woodburne, 1984 Family THYLACOLEONIDAE Gill, 1872

Lekaneleo gen. nov.

Type species: Lekaneleo roskellyae new combination

Generic diagnosis: differs from all species of Wakaleo and Thylacoleo in being smaller and in having: a dentary that possesses three single rooted teeth between I1 and P3; lower molars with talonid basins that are nearly as broad as the crowns; an M1 that is relatively square in basal outline and is not elongated posterobuccally and does not abut and overlap M2; smaller and thinner supraorbital processes; the 154 posterior region of the nasal dorsally inflated; parietals that are predominantly convex and relatively large in comparison to the squamosal and that roof most of the neurocranium; a broadly convex neurocranium; a relatively small squamosal contribution to the neurocranial wall; small sagittal and nuchal crests; a relatively flat or slightly convex pterygoid fossa; the postglenoid cavity laterally expanded into the postglenoid process; the squamosal only partially contribute to the posterior wall of the outer ear canal; the stylomastoid sulcus course through the mastoid process and not the squamosal, and; a relatively broader outer ear canal.

Also differs from species of Wakaleo in having: M2 with a trigonid basin positioned buccally not centrally; M2 with hypoconid taller than the protoconid and metaconid; a rounder and weaker infratemporal crest and; the supraorbital processes positioned relatively anteriorly. Also differs from species of Wakaleo in lacking: a prominent rostral tympanic process on the periotic; a deep concave fossa between the infratemporal crest and the sphenorbital fissure, and; lateral alisphenopalatine processes below the foramen rotundum.

Also differs from species of Thylacoleo in having four molars, in having both alisphenoid and squamosal elements contribute to the tympanic wing and in lacking a squamosal-frontal contact.

Differs from Microleo attenboroughi in being larger and in having P3: constricted where the anterior and posterior roots meet; broader posteriorly and having a relatively flat posterior margin; with a broader and deeper valley between the anterobuccal and posterobuccal crests; with a lingually-bowed longitudinal blade; a relatively horizontal posterior longitudinal blade and the posterior end of the crown similar in height to the posterior cusp; that has a prominent anterolingual crest; that has a sharper anterobuccal crest that is directed more posteriorly; lack a small buccal crest anterior to the anterobuccal crest; with a shallow posterobuccal basin and a relatively shorter and weaker posterobuccal crest. Also differs from Microleo attenboroughi in M2 having: greater lingual development posteriorly; paracone taller than the protocone; with a relatively weaker metaconule that is more lingually 155 located relative to the protocone; a more strongly developed preprotoconule crista, and; the postprotoconule crista running more posteriorly rather than in a posterobuccal direction. Also differs in M2 lacking a small notch on the lingual margin of the trigon basin between the protocone and metaconule.

Etymology: “lekane” is derived from Greek meaning basin and refers to the broad talonid basins present on the lower molars; and “leo” is Latin for lion.

Lekaneleo roskellyae new combination

(Figs 8.1 - 8.10, Table 8.1)

Holotype: QM F23453, a nearly complete skull missing the right basicranial and auditory regions, and both zygomatic arches. There is partial damage to the anterior region of the premaxilla, palate, and left auditory region. Both left and right cheek tooth rows contain P3 and M1-2. Alveoli for left and right I1-3, C1, P1-2 and M3-4 are preserved.

Referred specimens: QM F23442, a right maxilla containing P3, M1-2, and alveoli for C1, P1-2, and partial alveolus for M3; QM F40116, a right maxilla fragment containing a fragmentary P3, and M1-2 and damaged alveoli for P1-2 and M3-4 ;

QM F31605 an M3 ; QM F20191 a right dentary containing P3, M1-2, and alveoli for P1-2 and M3-4 ; QM F29623 a right dentary fragment containing the base of I1, P3, M1, and alveoli for P1-2 as well as small alveolus between I1 and P1; QM F51418 a right M2 ; QM F23455 an M2. 156

premaxilla nasal

maxilla

lachrymal foramen lachrymal

maxillojugal lachrymal suture

frontal 25.2 posterior opening of the infraorbital canal supraorbital crest supraorbital process zygomatic arch

palatine

sagittal crest squamosal parietal

occiput 10mm

Fig. 8.1: Cranium of Lekaneleo roskellyae, holotype, QM F23453, Upper Site. Dorsal view. (Drawing by Jeanette Muirhead)

157

Type Locality: The holotype is from Upper Site, Godthelp Hill; QM F23442 is from White Hunter Site (WH); QM F29623 is from Dirk's Towers Site (DT); QM F31605 is from Camel Sputum Site; QM F40116 is from Jim's Carousel Site (JC); QM F23455 is from Henk’s Hollow Site; QM F20191 and QM F51418 are from Cleft of Ages Site (COA). All these sites are located within the Riversleigh World Heritage Area.

Age: Upper Site, Dirk's Towers Site, and Camel Sputum Site are regarded as part of Riversleigh's System B sequence which is estimated to be early Miocene in age (Archer et al. 1997). White Hunter Site is a basal locality within Riversleigh System A sediments and is interpreted to be late Oligocene in age (Archer et al. 1997). Henk’s Hollow Site, JC Site and COA Site are part of Riversleigh’s Gag Plateau Sequence of sediments (Creaser 1997) and these sediments are regarded as part of System C (Archer et al. 1997) and estimated to be middle Miocene in age (Archer et al. 1997).

Etymology: This species is named in honour of the former Australian Minister of Arts, Sport, the Environment, Tourism and Territories, the Hon. Ros Kelly, who provided significant support for the Riversleigh Project.

Revised specific diagnosis: differs from L. deminutivus sp. nov. in: being larger; P3 not tapering posteriorly; lacking a notch on the longitudinal blade; possessing a shallower valley anterior to the posterobuccal crest and; the anterior crest appearing to consist of a single crest.

8.3 Description of skull, maxillae and dentary of L. roskellyae

General Description

The skull of Lekaneleo roskellyae is small, approximately 90mm long, and is similar in size and shape to the Brushtail Possum (Trichosurus vulpecula). In dorsal view the rostrum tapers anteriorly and the orbits are relatively large and directed anteriorly. 158

ethmoidal zygomatic foramen arch nasal maxilla lachrymal frontal parietal squamosal occipital

21.1 10.2

mastoid premaxilla alisphenoid foramen subsquamosal palatine rotundum foramen infraorbital glenoid fossa foramen sphenopalatine foramen sphenorbital fissure suture of jugal orbitosphenoid 10m

Figure 8.2: Profile of the cranium of Lekaneleo roskellyae, holotype, QM F23453.

(Drawing and photo by Jeanette Muirhead)

159

The frontal region is broad, slightly depressed, and bounded laterally by distinctive triangular supraorbital processes that continue posteriorly as converging supraorbital crests. These paired crests continue posteriorly to the occiput as a small sagittal crest. The cranium displays a marked constriction at the posterior margin of the frontals and then expands posteriorly to form a broad, low braincase. The neurocranial wall is dominated by a large rectangular parietal. The squamosal occupies the inferior third of the neurocranial wall and the zygomatic arch forms a relatively low point of attachment on this wall. In profile, the skull is moderate in depth and is deepest at the level of M1. The dorsal surface is gently rounded and the depression of the frontal region results in a flattening of its outline. Dorsal inflation of the posterior region of the nasals forms a rostral high point immediately anterior to the frontals. The occiput is broad and gently arched. Its surface is relatively vertical and bears a small overhanging nuchal crest. A large foramen magnum opens posteriorly from the inferior half of the occiput.

Ventrally, the palate is generally broad and rhomboid-like in shape although the region forward of the sectorial premolar narrows. Large confluent posterior palatal fenestrae dominate the posterior half of the palate. The cheektooth rows converge slightly posteriorly. The basicranium is broad. The tympanic cavity is large and is posteroventrally open. A tear-shaped periotic lies at the medial margin of this cavity. A broad remnant of a tympanic wing lies at the anterior margin of the cavity. The external auditory meatus is broad. The postglenoid process is gently rounded and its posterior surface is invaded by a laterally-expanded postglenoid cavity.

Rostrum The rostrum is broad and posteriorly it is dorsally inflated. In frontal aspect the rostrum is arched, being more rounded dorsally than laterally. In profile, the dorsal outline of the rostrum is slightly convex, rising from the anterior tip of the nasals to a high point at the posterior region of the nasals.

The nasals are long (35.7mm), being one-third the total length of the skull. They are gently rounded and in their posterior half are dorsally inflated along the internasal suture. In dorsal aspect they are wide anteriorly, narrowing midway along their 160 length, and then widen again posteriorly. They form a broad but shallow W-shaped contact with the frontals. The anterior margins of the nasals are laterally concave; their anterior tips together form a pointed nasal process which projects over the premaxillae and forms the dorsal boundary of the nasal aperture. In frontal view the medial margins of the nasals are thin and ventrally concave and together form the roof of a deep dorsal meatus within the nasal cavity. The ventral margins of the nasals are thickened and laterally are overlain by the dorsal margins of the premaxillae. Situated on the medial edge of the thickening is a longitudinal sulcus which develops posteriorly into a slender medial septum that is contiguous with a septum on the ethmoid. Posteriorly, the ventral surface widens and forms a fossa in the posterior third. The internal structures of the nasal cavity are missing.

In profile, the premaxillae are short and deep. In frontal aspect they are gently arched and, with the nasals, form a rounded nasal aperture which measures 12.1mm in width and 10.0mm in depth. The premaxillae are thin dorsally but inflate ventrally to accommodate the roots of the incisors and canine. In profile, the nasopremaxillary suture curves posteriorly in a gentle arc, reaching an apex at the posterodorsal margin of the premaxilla. The premaxillomaxillary suture commences near the posterior border of the canine alveolus and rises almost vertically to the nasal bone. The premaxillae contain alveoli for three incisors; the anterior is the largest and the second is smaller than the third. Although the anterior margins of the I1 alveoli are missing, the remains of the sockets indicate the alveoli were anteroposteriorly oval and approximately three times larger than the two posterior incisors. The more intact right alveolus measures 5.5mm in length and 3.5mm transversely at its broken margin. The alveolus for I2 is slightly triangular and small, the right measures 1.7mm in length and 2.1mm in breadth. The socket for I3 is basically round and the right measures 2.5mm lengthwise and 2.5mm transversely. The premaxilla continues posteriorly from the I3 alveolus for a short distance (1.5mm) and forms the anterior wall of the canine alveolus. The canine alveolus is oval in outline and large, measuring 3.8mm in length and 2.4mm in breadth. The anterior two-thirds of the anterior palatal fenestra are 161

premaxilla A I1 B I2 anterior palatal fenestra I3 C maxilla P1 P2

posterior palatal 3 fenestra P M1 vomer M2 M3 palatine M4

presphenoid pterygoid fossa basisphenoid foramen ovale alisphenoid medial glenoid squamosal process tympanic wing alisphenosquamosal suture postglenoi d process posterior lacerate postglenoid foramen mastoid process basioccipital paroccipital condylar foramen (Photo: Jeanette Muirhead) periotic 10 mm

Fig. 8.3: Ventral surface of the cranium of Lekaneleo roskellyae, holotype, (QM F23453). A. original undamaged skull B. Anatomy of skull.

162 situated within the premaxilla. A thin bar of the premaxilla forms most of the medial border of this fenestra.

The maxillae are short and deep. The infraorbital foramen is large, laterally- compressed, anteriorly-directed, and located low in the maxilla directly anterior to the anterior root of P3. Dorsal to the infraorbital foramen and anterior to the lachrymal, the maxilla is slightly depressed. The zygomatic process of the maxilla is narrow and protrudes posterolaterally at an angle of approximately 50o to the median plane. The anterolateral edge of the zygomatic wing bears an oblique, recessed, and roughened surface for the articulation with the jugal. This articular surface is a shallow triangular recess and is bordered anteriorly by a long, narrow sulcus and posteriorly by shorter sulcus. The shallowness of the recess suggests that the anterior portion of the jugal was thin and tapered anteriorly. The ventral surface of the maxillary process of the zygomatic arch is convex and smooth. Ventrally, the maxilla measures 35.6mm from the premaxillomaxillary suture to the posterior border of the postalveolar process. The maxilla contains alveoli for the posterior half of the canine, two small anterior premolars, a large P3, and four molars that sequentially decline in size.

Palate The palate is generally broad but tapers markedly anteriorly and less so posteriorly. Total length of the palate is 49.2 mm. Elongate anterior palatal fenestrae are located adjacent to I3 and the canine; the fenestrae are 7.7mm long and approximately 1.7mm wide. The premaxillomaxillary suture intersects the lateral edge of these fenestrae at their midpoint and continues on the posterior third of the medial edge. Anterior to these fenestrae the palatal arch is broad and flattened. The arch measures 6.1mm between the I2 alveoli. Posterior to the fenestrae the palate widens, reaching its broadest point between the posterior roots of P3 (25.2mm). The extensive damage to the palate makes it difficult to determine whether palatal ridges were present. A small palatal foramen is located medial to, and midway between the canine and P1 alveoli. Smaller foramina are also located near the anterior and posterior roots of P3. Numerous minute pits/foramina are scattered in a band running just medial to the 163 tooth row from the canine to M1. Large posterior palatal fenestrae arise at the level of M1 and measure 21.8 mm in length, slightly less than half the total length of the palate. At their posterior limit the combined fenestrae measure 13.7mm in breadth. Medially, small incomplete palatal septa converge to project posteriorly into the fenestrae. Posterior to M1 the lateral margins of the palate continue as tapering processes that are longitudinally bisected by the maxillopalatine suture. The surface of these processes tilts dorsomedially, the medial margin forming a smooth ridge as it arches toward the presphenoid.

Lachrymals The lachrymal forms the anteromedial wall of the orbit. Anteriorly, a narrow strip of the lachrymal projects onto the rostrum. Within this anterior strip are two lachrymal foramina, one positioned above the other; the superior foramen is small, round and dorsally-directed and the inferior foramen is large, oval, and anterolaterally-directed. The maxillolachrymal suture courses anteriorly in an arc around the foramina descending laterally onto the dorsal surface of the zygomatic arch where it makes contact with the jugal. Here, the lachrymal forms a small wing which lies on the posterodorsal surface of the zygomatic process of the maxilla. Ventrally, the lachrymal contacts the maxilla. Posteriorly it contacts the palatine inferiorly and the frontal superiorly.

Palatine The palatine forms a broad wing at the posterior region of the palate. Within the orbit the anterior limit of the palatine lies just posterior to the infraorbital canal where it contacts the lachrymal. Laterally it is bordered by the maxilla. The anterior region completely encloses the sphenopalatine foramen which is round and measures 1.7mm in diameter (right side). Anterodorsally the palatine is bounded by the frontal and posterodorsally by the presphenoid. Posteriorly, it contacts the alisphenoid. A small, narrow posterolateral palatine foramen is located at its lateral extremity near the maxillopalatine suture. Ventrally, the palatines form the medial margins of the postalveolar processes. The maxillopalatine suture runs longitudinally along each 164

5.2 1.3 x 2.0 (length x breadth) 1.7 x 1.8 3.1 x 2.3

13.6 1.6 x 1.1 1.2 x 1.1 26.1

26.6 23.1

15.7 92.8

29.8 19.0

8.0

36.0

Figure 8.4: Cranial measurements of Lekaneleo roskellyae, QM F23453. Measurements in mm. 165 process and turns medially at the level of M3. The width of the posterior region of the palate between the maxillopalatine sutures is 18.9mm. The posterolateral palatine foramen is located approximately 3.5mm anterior to the posterior margin of the postalveolar process. The posterior margins of these processes are slightly damaged and it is difficult to determine whether a transverse post-palatine bar was present. The medial margin of the postalveolar process is gently rounded, the palatine ascending medially to form a broad arch over the posterior part of the narial passage. In ventral view, the palatine is bounded medially along most of its length by the vomer. Posterior to the vomer there is short contact (5mm) with the presphenoid and then as the palatine tapers posteriorly, there is a broad contact with the basisphenoid. At its posterolateral margin the palatine contacts the alisphenoid.

Vomer Although present when originally recovered, the anterior part of the vomer has since been lost. In ventral view the vomers are broad anteriorly, narrowing posteriorly to form thin, medial, bilateral processes which are separated by a narrow ridge of the presphenoid. Laterally, the vomer articulates with the palatines, the vomeropalatine suture curving medially as it runs posteriorly. The vomers extend posteriorly to within 4.4mm of the presphenoid-basisphenoid suture.

Neurocranium and Basicranium Frontal The frontal region of the skull is broad anteriorly and shows marked constriction posteriorly. Dorsally, the frontals combine to form a broad, shallow, tear-shaped fossa. Laterally, the margins of this fossa are formed by triangular, thin, supraorbital processes that project laterally to the level of the maxillopalatine suture. The width of the skull between the supraorbital processes is 25.2mm. Posteriorly, the converging margins of the fossa are formed by rounded supraorbital crests which, after meeting, continue posteriorly as the sagittal crest. The dorsal anterior margin of the frontals is slightly raised and overlain by the nasals. The nasofrontal contact is marked by a shallow W-shape suture. Anterior to the supraorbital processes the frontal projects anteriorly toward the premaxillomaxillary suture. Along its anterolateral margin the frontal articulates with the lachrymal. Laterally, the frontal 166 forms the steep ventromedially-inclined wall of the orbit. As the frontals constrict posteriorly this wall becomes laterally concave and dorsoventrally convex. Ventrally, the main contact is with the palatine and with presphenoid ventroposteriorly. A small, posteriorly-directed ethmoidal foramen opens at the posterodorsal margin of the frontopalatine suture, approximately 7mm anterior to the sphenorbital fissure. In dorsal view, posterior to the supraorbital processes, the frontals narrow forming a distinct biconcave postorbital constriction that measures 14.0mm in width. At their posterolateral margins, the frontals begin to expand laterally as they contribute to the anterior wall of the neurocranium and are here overlain by the parietals. Immediately anterior to the frontoparietal suture and approximately 10 mm below the supraorbital crest, an infratemporal crest descends along the line of the suture to the alisphenoid. A small fossa lies anterior to the infratemporal crest and dorsal to the sphenorbital fissure, and accentuates the prominence of the former. Ventral to the parietal the frontal contacts the alisphenoid. The frontoalisphenoid suture measures approximately 8mm and runs in a dorsoposterior direction.

Ethmoid The ethmoid is incomplete. Although most of the turbinates are missing, the cribriform plate is mostly intact as is the posterior edge of the mesethmoid. The cribriform plate is bounded dorsally and laterally by the frontal bones and ventrally by the vomer and palatine bones. Most of the left portion of the cribriform plate is preserved, however much of the central portion of the right side has been lost. The inferior half of the plate curves anteriorly. The dorsal edge of the plate is bilaterally concave and forms the ventral margin of the frontal sinus. The plate is perforated by foramina which are concentrated over the inferior half and along the lateral margins.

Parietal The parietals roof most of the neurocranium. The left parietal is intact and the lamina of the right parietal is badly fractured and missing from its posterolateral and anterior margins, and the contact with the squamosal has been obscured on this side of the skull. The parietals are long and broad. The left parietal measures 37.5mm from its anterior margin to the nuchal crest posteriorly. Transversely it measures approximately 14.0mm between the interparietal line and the squamosoparietal 167 suture. The parietals are convex for most of their length turning concave in the posterior third as the posterior margin rises to form a broad nuchal crest. At their medial margin they form a low but moderately thick sagittal crest. The sagittal crest begins at the posterior convergence of the supraorbital crests. A very narrow sulcus runs along the interparietal articulation. The anterior border of the parietal is anteriorly convex and overlies the frontal bone. The frontoparietal suture curves anteriorly from the sagittal crest (following the temporal line) and descends posteroventrally to meet the alisphenoid. The alisphenoparietal suture is short (5.3mm) and dorsally convex. Posterior to the alisphenoid the parietal has a long contact with the squamosal.

Presphenoid Ventrally, the presphenoid can be seen as a rounded median bar separating the horizontal laminae of the vomer. Beyond the rear margins of the vomer the presphenoid widens slightly and articulates laterally with the palatines and posteriorly with the basisphenoid. Externally, the orbital wing of the presphenoid lies just anterior to the sphenorbital fissure and extends anteriorly, articulating with the frontal dorsally, palatine ventrally, and basisphenoid posteriorly. A small, anteriorly- directed foramen is located in the dorsal part of the orbital wing between the ethmoidal foramen and the sphenorbital fissure.

Basisphenoid The right posterolateral edge the basisphenoid is missing. Minor damage is also evident on the left posteroventral surface. The basisphenoid is roughly rectangular, widening slightly in its posterior third. It measures 17.4mm in length between the presphenobasisphenoid suture and the basisphenobasioccipital suture. The basisphenoid is inclined anteriorly at an angle of 10o to the horizontal plane. The anterior region of its ventral surface continues the contour of the presphenoid and is gently rounded, however, over the posterior two-thirds its surface becomes flatter. The lateral edges of the basisphenoid are ventrally-tilted, slightly depressed and markedly roughened for articulation with the basal margin of the pterygoid. Laterally, the basisphenoid is strongly fused to the alisphenoid such that it is difficult to discern the basisphenoalisphenoid suture. On the lateral edge of the basisphenoid a 168 sulcus for the pterygopalatine nerve extends from the anterior palatine contact to the posterior third where it swings laterally on to the alisphenoid and terminates. This groove would have been roofed by the ala of the pterygoid (both pterygoids are missing). Posteriorly, the basisphenoid widens slightly through bilateral processes which project posterolaterally. A large entocarotid canal runs through each process and opens posteriorly near the anterior point of the periotic and the medial edge of the foramen ovale. Only the left canal has been preserved, the posterior ventral margin of which has been broken away exposing the posterior roof of the canal and obliterating the entocarotid foramen. The entocarotid canal measures 1.8mm in diameter approximately midway along its length. Although a suture is not evident, it is possible that the alisphenoid forms the lateral wall of the entocarotid canal.

Alisphenoid Both alisphenoid bones are damaged. Most of the right alisphenoid is missing, however, the left is mostly intact although there is some minor damage to its posteroventral area. The alisphenoids are broad and floor most of the anterolateral part of the neurocranium. In lateral view, the alisphenoid extends anteriorly to roof the posterior part of the sphenorbital fissure, contacting the frontal anterodorsally and the presphenoid anteroventrally. Anterodorsally, it makes a short contact with the parietal and posterodorsally it makes a longer contact with the squamosal. The infratemporal process that lies at the posterior margin of the frontal continues posteroventrally onto the alisphenoid and forms a small, robust, triangular prominence at the ventral limit of the process just ventral to the parietoalisphenoid contact. Ventral to the sphenorbital fissure the alisphenoid articulates with the presphenoid medially and the palatine laterally. A large, anteriorly-directed foramen rotundum lies just posterior to the alisphenopalatine suture. Lateral to the foramen rotundum, the lateral edge of the alisphenoid bears a small ventrolaterally-projecting ridge. This ridge runs posteriorly from the alisphenopalatine suture to approximately the midpoint of the alisphenoid (measuring 7.07mm in length).

In ventral aspect the alisphenoid articulates with the basisphenoid medially, pterygoid ventromedially, and squamosal posteriorly. A small ridge (7.6mm in length) runs along the alisphenoid-basisphenoid boundary, slightly flooring and 169 forms the lateral wall of the pterygopalatine groove. Immediately lateral to this ridge the alisphenoid forms a triangular pterygoid fossa that is bounded laterally by the “lateral pterygoid ridge”. The narrower and flatter anterior region of the pterygoid fossa is separated from the broader, slightly convex, posterior part by a low anterolaterally-directed ridge. No foramina for a transverse canal are present in the skull.

Squamosal Both squamosal bones are damaged. On the left side most of the bone is preserved although the posterolateral portion of the tympanic wing is missing allowing a view of the internal surfaces of the tympanic cavity. On the right side, much of the lamina and posterior portion of the element is damaged or missing. Both bones are lacking the anterior zygomatic processes. In lateral view, the squamosal lamina is relatively long and narrow and covers approximately the ventral third of the brain case above the level of the zygomatic limb. In lateral view, the squamosparietal suture forms a gentle arch and measures 24.5mm in length. The anterior two-thirds of the lamina is gently convex and the posterior third is concave. Anteriorly, the squamosal contacts the alisphenoid. The alisphenosquamosal suture descends posteroventrally toward the anterior edge of the zygomatic limb and continues on the ventral surface in a posteromedial direction toward the tympanic wing of the alisphenoid. The posterior portion of the zygomatic limb is anteroposteriorly broad and dorsally concave. At its dorsolateral margin the limb develops a thin squamosal crest. Posterior to the zygomatic limb, a broad, transversely-elliptical subsquamosal foramen extends across the lateral margin of the squamosal to within 2mm of the lateral edge of the nuchal crest. This foramen measures 5.2mm in width and 1.6mm in height and internally develops into a shallow, horizontally-triangular recess. Two small foramina are clearly visible within the subsquamosal foramen. The anterior-most foramen is the larger of the two, is directed anteriorly, and connects with the postglenoid foramen. The posterior foramen is smaller and is directed medially. 170

Ventrally, the loss of the ectotympanic and the lateral half of the tympanic wing allow an unobstructed view of the posterior ventral region of the squamosal and the tympanic cavity. At the tympanic wing the squamosal extends medially to overlie the alisphenoid and forms the lateral half of the anterior wall of the tympanic cavity. The ventral surface of the zygomatic limb although damaged, appears to have provided a flat articular surface for the posterior process of the jugal. The glenoid fossa forms a shallow triangular depression and measures 6.8mm anteroposteriorly. The anterior region of the fossa (articular eminence) is flattened while posteriorly a moderate groove (the mandibular fossa) lies directly anterior to the postglenoid process. The mandibular fossa measures 9.0mm wide (medial glenoid process to the glenoid notch). The medial glenoid process is a low, broad crest that extends from the anterior edge of the zygomatic limb to the base of the tympanic wing, running just laterally to the alisphenosquamosal suture. The postglenoid process is broad, measuring 7.7mm across its base. It is thin anteroposteriorly and thickens laterally, and is rounded ventrally. Its posterior surface is broad and concave and forms the postglenoid cavity. Posterior to the postglenoid process the squamosal forms the roof and posterior wall of the outer ear canal.

Occipital Region On the right side of the skull the lateral and inferior margins of the occiput are missing. Ventrally, most of the basioccipital is missing and only a fragment on the right where it contacts the exoccipital and periotic has been preserved. This fragment bears a rounded ridge along its lateral edge. At its widest point the basioccipital was approximately 18.6mm wide. Viewed posteriorly, the occipital region is broad and measures approximately 35.9mm wide between paroccipital processes and 21.5mm deep from the dorsal nuchal crest to the broken ventral margin of occipital condyles. The supraoccipital forms a broad arch across the back of the cranium. Dorsally, the supraoccipital forms a small nuchal crest where it contacts the parietals. A medial, rounded, occipital crest extends from the nuchal crest to the dorsal margin of the foramen magnum and separates a pair of supraoccipital fossae. Large supraoccipital foramina are located lateral to the medial crest and smaller foramina are distributed across the supraoccipital fossae. The ventral margin of the supraoccipital forms a rounded and slightly thickened arch that is the dorsal labium of the foramen magnum. Internally, the anterior margin of the supraoccipital extends forward to underlie the parietals, forming an interparietal process. The internal surface of the 171

alispheno-squamsosal suture

alisphenoid

squamosal

sq broken tympanic wing showing alispheno- squamosal suture within the tympanic cavity

sqw = epitympanic wing of squamosal

alispheno-squamosal suture

alisphenoid medial glenoid process foramen ovale

internal carotid squamosal canal

broken glenoid tympanic wing fossa postglenoid epitympanic wing process of petrosal postglenoid promontorium cavity of periotic postglenoid foramen posterior lacerate squamosal epitympanic sinus foramen incudal fossa stylomastoid mastosulcusid proc ess paroccipital process 10 mm

Fig. 8.5 : Basicranium of Lekaneleo roskellyae, holotype QM F23453, Upper Site. (Drawing by Jeanette Muirhead) 172 supraoccipital bears three fossae, one medially and two laterally. The medial fossa is bounded by paired internal occipital crests. Dorsolateral to each crest is a short sulcus.

Both exoccipitals are damaged, the right more so than the left. The occipital condyles are incomplete, missing their ventral parts. The superior borders of the condyles gradually emerge from the dorsal condyloid fossae. A small sulcus is located on the medial surface of each condyle and within this sulcus lies a hyperglossal foramen. Only the dorsal margin of the foramen magnum is preserved. This foramen is approximately 12mm at its widest point. The right paroccipital is missing and the left exhibits minor damage to the posterolateral surface. The paroccipital has a broad base and ventrally is similar in depth to the mastoid process. The paroccipital- mastoid suture traverses the outer third of the occiput, descending in a slightly lateral direction.

Auditory Region

Periotic Neither ectotympanic bone nor any of the auditory ossicles are preserved and the loss and damage to the tympanic wings allows an unobstructed ventral and lateral view of the petrosals on both sides of the skull. The extensive damage to the right tympanic wing has exposed most of the tympanic face of the periotic on this side. The more complete left basicranial area provides information about the relationships of the bones and structures in the auditory region. In ventral and lateral view, the periotic is teardrop-shaped, its tapered anterior end lying between the basioccipital, basisphenoid, and the alisphenoid. Its broad, rounded, posterior portion is bounded by the exoccipital posteromedially and the squamosal posterolaterally. The length of the right periotic is approximately 9.3mm.

Tympanic Face The predominant feature on the tympanic face is the pars cochlearis which is broad and has an almost flat surface that carries a slight central depression. An angular, anteriorly-directed ridge delineates the promontorium from the pars cochlearis. The 173 promontorium is triangular and its tip slightly overlaps the ventral posterolateral margin of the basisphenoid (on the sulcus for the internal carotid). The ventromedial margin of the promontorium appears to have been broken and forms a sharp ridge that becomes obtuse as it merges with posterior part of the pars cochlearis. Slightly anterior to the fenestra cochleae, the posteroventral margin of the pars cochlearis is developed into a rounded but angular, vertical ridge that weakens as it approaches the fenestra vestibuli. In lateral view, the anterior and posterior ridges on the tympanic face gives the pars cochlearis a diamond-shaped outline. The fenestra cochleae is vertically-oval and opens posterolaterally. It lies at the posterior margin of a broad, flat posterolaterally-directed basin (fossula cochleae). The fenestra vestibuli lies anterodorsally to the fenestra cochleae. It is anteroposteriorly oval, inclined anteroventrally and is slightly larger than the fenestra cochleae. The dorsal margin of the fenestra vestibuli forms a sharp lip. This labium also forms the ventral margin of the sulcus for the facial nerve. The foramen for the facial nerve lies anterodorsally to the fenestra vestibuli, is vertically narrow and opens posteriorly. From this foramen the facial nerve sulcus runs posteriorly, directly above the fenestra vestibuli, and connects posteriorly with a large fossa, the stapedial fossa. This fossa is deeper posteriorly and bears a notch on its ventrolateral margin for the stylomastoid nerve. On the left side of the skull the stapedial fossa is partially hidden by the mastoid process of the petrosal which lies ventral to this fossa.

The more extensively exposed right petrosal allows an unobstructed view of the morphology of its anterior surface. The hiatus fallopi lies 2.8mm anterior to the secondary facial nerve foramen. Dorsal and anterolateral to the hiatus fallopi the petrosal expands to form an epitympanic wing. This small, tear-shaped, relatively deep, concave fossa extends anteriorly to the alisphenosquamosal suture. Dorsolaterally it is bounded by the squamosal hypotympanic sinus. Posterior to the epitympanic wing and separated from it by a steep septum is the incudal fossa. It is roughly triangular in shape and is bordered laterally by the squamosal and medially by the sulcus of the facial nerve. At the lateral edge of the incudal fossa the squamosal is developed into a septum which separates the fossa from the squamosal epitympanic sinus.

174

Cerebellar Face Damage to the right parietal allows a clear but restricted view of the cerebellar face of the left periotic. The cerebellar face has a roughly pentagonal outline. The main features on this face are two large openings, the internal auditory meatus and the subarcuate fossa. The internal auditory meatus is small, round, and deep and lies ventromedially to the subarcuate fossa. No internal septum can be distinguished in its opening. Its anteroventral rim opens gently onto the cerebellar face. A narrow sulcus (probably for the facial nerve) runs from the anterior edge of the meatus to the anterior edge of the cerebellar face. A second, shorter sulcus also exits the meatus immediately ventral to the larger sulcus but merges with the cerebellar face about half way to the anterior edge of the periotic. The subarcuate fossa is large and separated from the internal auditory meatus by a moderately broad septum. The septum widens and projects anteriorly, developing a slight process on the prefacial commissure. The subarcuate fossa opens gently in an anterodorsal direction. Its ventral rim is rounded while its anterodorsal margin is flat and expands to form a sharp dorsal crest (the crista petrosa). Anterior to this crest the lamella expands to form a narrow concave flange that articulates with the squamosal and contributes to the rear portion of the middle cranial fossa.

Ventromedial face The ventral edge of the promontorium forms a sharp longitudinal crest that becomes obtuse at the posterior margin of the pars cochlearis. This crest forms the lateral wall of a deep, broad, trough that extends anteroposteriorly along the medial half of the ventral surface of the periotic. The posterolateral half of the trough is pitted. The posterolateral extremity of the basisphenoid lies in the anterior edge of this trough and the remaining part of the trough is the articular surface for the basioccipital. A shorter longitudinal ridge, approximately half the length of the ventral crest, lies in the anterior half of the trough and creates a smaller medial sulcus, the sulcus of the inferior petrosal sinus. Posterior to the short ridge, the medial surface forms a shelf as it curves toward the cerebellar face. Posterior to the articular trough and the ventral edge of the promontorium is a deep transverse sulcus (partially obscured by the exoccipital) which possibly represents the posterior lacerate sulcus or the sulcus of the sigmoid sinus. The sulcus is deeper at its lateral end and bears three shallow 175 furrows in its lateral half. A broad septum separates the lateral end of the sulcus from the stapedial fossa. The lateral wall of this septum is pierced by a small foramen from which courses a small sulcus in a dorsolateral direction.

The medial end of the mastoid process of the petrosal is larger than the lateral end. Two relatively large sulci lie on each side of the process and results in the process having a rather hour-glass shape when viewed ventrolaterally. The enlarged medial end of the mastoid partly encloses the sulcus that courses along its posteromedial border with the paroccipital. The sulcus on the anterior margin of the mastoid carries the stylomastoid division of the facial nerve and lies mainly within the mastoid rather than the squamosal.

Outer Ear Canal In ventral and lateral view the outer ear canal is very broad. Posterior to the medial glenoid process it measures 5mm wide and widens to 8mm at the external margin of the canal. It is enclosed anteriorly, dorsally and posteriorly by the squamosal bone. The anterior wall of the outer ear canal is formed by the posterior surface of the postglenoid process. This surface is transversely broad and kidney-shaped, posteriorly-concave, and forms the large postglenoid cavity. The cavity expands anteriorly and dorsally under the medial glenoid process. The lamina in the dorsolateral part of the cavity is fractured and it is not possible to discern whether any foramina were present in this region. The medial part of the cavity contains a small shallow fossa that is separated from the larger lateral fossa by a small septum. This medial fossa is triangular in shape and has a large posterior opening, the postglenoid foramen. A second, smaller foramen is located on the anterior margin of the fossa just medial to the lateral septum and is directed anteriorly. A small spur of bone lies directly ventral to this foramen. The posterior wall of the canal is formed mainly by the squamosal which lies against the anterior surface of the mastoid process and terminates about half way down this process.

Upper Dentition The upper dentition of the holotype QM F23453 is described in Appendix 6, (Gillespie 1997). 176

Fig. 8.6: Lekaneleo roskellyae right maxilla, QM F23442, White Hunter Site Stereo pair. Scale = 5 mm

Fig. 8.7: Lekaneleo roskellyae, right maxilla, QM F40116, Jim’s Carousel Site Stereo pair. Scale bar = 5 mm 177

Additional maxilla specimens of Lekaneleo roskellyae (Figs 8.6 and 8.7) Two additional maxillary specimens have been recovered from the Riversleigh World Heritage Area. Both specimens are right maxillae, one from White Hunter Site (QM F23442) and the other from Jim’s Carousel Site (QM F40116). QM F23442 is smaller than QM F40116 and retains P3 M1-2 and alveoli for P1-2 and C1 (damaged). QM F40116 also retains P3 M1-2 but its P3 is badly damaged and only the posterolingual part of the crown is preserved. Alveoli for P1-2 and M3 are also present in this specimen but most are damaged. The trigon basins of the molars of QM F23442 display prominent crenulations, however, those of QM F40116 exhibit a much greater degree of wear. These specimens are compared to the holotype (QM F23453). Measurements of all specimens are shown in Table 8.1.

Canine: the position of the canine alveolus in QM F23442 is similar to that in the holotype, lying approximately 3mm anterior to the P1 alveolus.

Anterior premolars: the alveolus for P1 in both QM F23442 and QM F40116 lies slightly anterior (approx. 1mm) to the alveolus for P2. In QM F23442 the P1 alveolus is slightly larger than the P2 alveolus and is similar to the condition seen in the holotype. P1 alveolus is damaged in QM F40116 and its size cannot be determined. In both specimens, the location of P2 is similar to that in the holotype, lying immediately anterior and slightly lingual to the base of P3.

P3: the intact P3 that is present in QM F23442 is essentially similar to that of the holotype except it is slightly smaller.

M1: M1 of both specimens, and especially that of QM F40116, shows essentially the same morphology as that of the holotype. QM F23442 shows a slight difference in the region of the stylar basin. Rather than having a long uninterrupted stylar basin that is bordered buccally by a continuous ridge, this basin in QM F23442 is divided approximately half way along its length by the postparacrista which swings across 178 the basin and connects with the buccal margin. Additionally, rather than having a single 178

Table 8.1: Measurements of upper and lower dentitions of Lekaneleo roskellyae (mm)

Upper Dentition Riversleigh P3 P3 P3 M1 M1 M2 M2 M3 M3 M4 M4 P3:M1 P3 / 3 4 Specimen # Site length anterior posterior length width length width length width length width P – M width width (max.) (max.) QM F23453 (L) Upper 8.3 4.2 5.1 5.8 6.2 4.3 4.8 3.3a 3.5a 3.0a 2.5a 1.41 0.35 (R) 8.1 4.2 5.1 5.7 6.0 4.2 4.9 3.3a 3.5a 2.8a 2.5a QM F23442 WH 7.6 3.4 4.3 5.8 6.1 4.1 4.7 1.31 QM F40116 JC [9.6] [5.1] 6.5 6.8 4.5 5.4 1.47

Lower Dentition

Riversleigh P3 P3 P3 M1 M1 M2 M2 M3 M3 M4 M4 P3-M4 P3 / Specimen # Site length anterior posterior length width length width length width length width P3 – M4 width width QM F20191 COA 6.4 3.6 4.4 6.8 4.8 5.2 4.4 3.7a 2.6a 2.7a 2.2a [25] 0.25 QM F29623 DT 7.3 3.5 4.3 6.3 4.5 QM F51418 COA 4.9 3.9

a = alveolus; [ ] = estimate

179 crest marking the buccal boundary of the stylar basin, QM F23442 exhibits a number of separate cuspules that are probably the original stylar cusps. The most prominent of these cuspules lies midway along the stylar basin slightly anterior to the buccally- directed postparacrista.

M2: this molar in both QM F23443 and QM F40116 is similar to that of the holotype.

Dentary

No dentary was found with the skull (QM F23453) from Upper Site. The description of the dentary of L. roskellyae is based on two right dentaries, QM F20191 from COA Site and QM F29623 from Dirk's Towers Site (Figures 8.8 - 8.10). Neither of these dentaries is complete but each retains some features not present in the other and these variations will be referred to where appropriate. QM F20191 is the more complete specimen, preserving most of the horizontal ramus including P3 M1-2, alveoli for P1-2 and M3-4, and the anterior half of the masseteric fossa. QM F29623 preserves only the anterior half of the horizontal ramus and bears I1(proximal portion) P3 M1 and three small alveoli between I1 and P3.

QM F20191 shows that the vertical axis of the horizontal ramus is twisted such that the anterior region is tilted buccally while the posterior region flexes lingually, consequently the posterior portion of the molar row overhangs the lingual flank of the ramus. The dorsal surface of the ramus is slightly concave. The alveoli of M3-4 are medially rotated and are located on the anterior portion of the ascending ramus. The anterior margins of these alveoli are lower than the posterior margins. The dentary fragment is deepest below P3. On the lingual surface of the ramus a deep symphyseal scar (8.0mm depth) extends from the I1 alveolus to below the posterior root of P3 (12.5 mm). The dorsal edge of the posterior portion of the scar forms a slight medial overhang. Two small foramina pierce the articulating surface, one centrally and the other 4mm posterior. Another small, anteriorly-directed foramen 180 lies dorsal to the symphyseal scar 1mm behind the I1 alveolus. Below the M3-4 alveoli the lingual 180

A

B C

Fig. 8.8: Lekaneleo roskellyae, right dentary, QM F20191. Scale bar = 10mm A. buccal view B. lingual view C. Stereo occlusal view

25.1

5.2 6.8 6.4

11.5

36.4

Fig. 8.9: Dimensions of Lekaneleo roskellyae, right dentary, QM F20191. Buccal view. Measurements in mm. 181 surface of the ramus is concave. The angular process commences slightly posterior to M4. An oval-shaped mandibular foramen is exposed on the broken posterior edge of the dentary and penetrates to the level of M1. The buccal surface of the horizontal ramus is gently convex. In QM F29623, a large mental foramen is located at the base of the anterior root of P3 and a smaller foramen is located below the anterior root of M1. The ramus narrows above the root of I1 and anterior to P3. The bone in this area and around the I1 alveolus is perforated by numerous small foramina. The masseteric fossa extends anteriorly to the level of M3/M4. A small posteriorly-directed masseteric foramen is located in the anterior region of the fossa.

Lower Dentition

Measurements of the lower dentition are shown in Table 8.1. The molar row is buccally bowed and is twisted so that the anterior cheekteeth (P3-M2) are oriented vertically and the posterior molars (M3-4) are tilted medially.

QM F29623 bears three small alveoli between the I1 alvelous and P3. The anterior- most of these is the largest, is subtriangular in outline and has a broad (1.3 mm), anteriorly-directed opening. Directly behind and slightly medial to this alveolus lies a small, round, shallow alveolus. The third alveolus lies slightly laterally to the second and abuts the anterolingual base of the anterior root of P3. The third alveolus is larger than the second but smaller than the first.

I1: in QM F29623 the basal 6.5mm of the incisor is preserved and protrudes from the horizontal ramus at an angle of 40o. In cross-section I1 is mediolaterally compressed. The medial surface is flattened and the lateral surface is rounded. The incisor measures 5.1 mm in depth and 2.9 mm in breadth at the level of the alveolus. In the centre of the broken surface of the tooth there is a large neural canal. The proximal end of a small crest is present on the dorsobuccal margin of the incisor. This crest is most obvious in cross-sectional view. Enamel is unevenly distributed over the tooth; 182

A B

C

Figure 8.10: Lekaneleo roskellyae right dentary, QM F29623. Scale bar = 10mm

A. lingual view B. buccal view C. occlusal view stereo pair 183 it covers most of the buccal surface and continues onto the dorsal and ventral margins of the medial surface but is absent from most of the medial surface. The root of the incisor continues posteriorly to below P3.

P3: is blade-like. In occlusal view the crown is subtriangular in outline, the posterior half of the tooth being broader than the anterior half. A longitudinal blade descends from a tall anterior cusp to a shorter posterior cusp, curving lingually at the posterior third of the tooth. A narrow shearing facet is evident along the buccal edge of the longitudinal blade. In profile, the anterior edge of P3 is gently rounded. An anterior crest runs down the anterior edge of P3 and curves lingually at the base of the crown. In QM F20191 the anterior crest bifurcates near the base of the crown, forming a triangular facet on the anteromedial surface. The buccal surface is gently convex. A weak ridge descends from the anterior cusp down the anterobuccal flank towards the base of the crown. In QM F20191, a weak, notched, posterobuccal crest descends from near the posterior cusp. In QM F29623, extension of the longitudinal wear facet obscures any evidence of a similar crest. A lingual crest runs from the anterior cusp to the base of the crown above the anterior root. Between the anterior crest and the anterolingual crest the crown forms a narrow valley. Posterior to the anterolingual crest the lingual surface is broadly concave with the curvature increasing markedly at the point where the longitudinal blade curves lingually. A short posterolingual crest descends from the posterior cusp in a slightly anterior direction. In QM F20191 the anterior edge of this crest bears small buttress-like ridges. The ridges in QM F29623 may have been obscured by wear. In QM F20191 a small ridge of enamel runs across the base of the crown from the base of the anterolingual crest towards the base of the posterolingual crest. On the lingual surface of the crown the inferior margin of the enamel forms a relatively straight line from the anterior edge of the tooth to midway across the posterior root. From the latter point the enamel border rises, following the inferior margin of the posterolingual crest. The posterior root is larger than the anterior root.

M1: the M1 of QM F20191 is broader than that of QM F29623 and shows slight differences in morphology. In occlusal view M1 is rectangular with a slight 184 narrowing at the anterior margin. The front edge of M1 is concave and conforms to the convex curvature of the posterior end of P3. Anteriorly there is a tall trigonid and posteriorly, a large talonid basin. The apical cusp of the trigonid, the protoconid, is lingually displaced. An anterior crest descends from this cusp in a slightly buccal direction to the anterior edge of the tooth where it is contiguous with the longitudinal blade of P3. On QM F20191 a shearing facet is present on the buccal edge of this crest. The concave anterior edge of the tooth is developed into a small transverse crest. The anterolingual surface of the trigonid is mainly convex and bears a small tubercule just medial to the protoconid. A small fossa lies directly lingual to the anterior crest near the anterior margin of the tooth. An anterobuccal crest descends from the protoconid to the swollen base of the crown above the anterior root. The trigonid of QM F20191 is broader than that of QM F29623 and its anterobuccal crest is more strongly developed. Small horizontal ridges are present along the anterobuccal edge of this crest in the former. The buccal flank of the talonid is rounded and forms a slight valley where it meets the anterobuccal crest. In QM F29623 the anterobuccal crest connects with the buccal edge of the talonid basin. The posterior flank of the protoconid descends steeply and forms the anterior wall of the talonid basin. A lingual crest descends posteromedially from the protoconid and joins the lingual edge of the talonid basin. The talonid basin is highly crenulated. In QM F29623, this basin is deeper and narrower, especially posteriorly. From the base of the lingual crest a short preentocristid runs posteriorly to an entoconid. In QM F20191 a small transverse crest cuts across the posterolingual corner of the talonid basin, running from the entoconid to about three-quarters of the way across the posterior margin. This transverse crest is not present in QM F29623. Posteriorly, a postentocristid descends around the posterolingual corner of the tooth, meeting the posthypocristid at the nadir of the rear end of the molar. The posterior root is larger than the anterior root.

M2: is rectangular in outline and rotated medially relative to M1. The hypoconid is the highest cusp. The protoconid and metaconid are subequal in height and the entoconid is the lowest cusp. Anteriorly, a small triangular trigonid basin lies between the confluent protocristid and metacristid and the anterior edge of the tooth. This basin in QM F20191 is much broader than the basin in QM F51418. In the 185 former specimen the basin stretches across most of the anterior part of the tooth while in the latter specimen the basin is restricted to the anterobuccal corner. The talonid basin is broad and shallow. Both the anterior basin and the talonid basin are crenulated. The anterior and posterior roots appear to be subequal in size.

Neither M3 nor M4 is preserved in either specimen.

Comparisons

Comparisons were made against the crania of Wakaleo vanderleueri (CPC 26604), W. hilmeri (QM F45200), Thylacoleo crassidentatus (QM F10622), T. carnifex (AM F4659) and against the palate of W. pitikantensis (SAM P37719). The dentaries and lower dentitions were compared against those of W. oldfieldi (SAM P17925 and QM F20895), W. hilmeri (AR 19249, QM F30465 and QM F31376), W. vanderleueri (NTM P85553-4 and QM F36453), Thylacoleo carnifex (AM MF 734), T. crassidentatus (QM F7762) and T. sp. cf. T. hilli (AM F63584).

The cranium of Lekaneleo roskellyae is absolutely smaller and more gracile than that of any other thylacoleonid for which crania are known. It is approximately half the length and width of W. vanderleueri and W. hilmeri and approximately one third the length and width of T. carnifex (see Appendix 1; Table E, F and G). Although the skull of L. roskellyae exhibits a broadness generally characteristic of marsupial lions, it is proportionately narrower.

In general appearance, the cranium of L. roskellyae differs most from other thylacoleonids in the region of the neurocranium, which is proportionately longer and broader, and which lacks large sagittal and nuchal crests. The region of the basicranium, although showing similarities to Wakaleo, also exhibits a number of significant differences.

In profile, the skull of L. roskellyae shows some similarity to W. hilmeri, both taxa having relatively flattish or gently curving dorsal profiles that are mainly the result of the depressed frontal bones that lie between the supraorbital processes. The 186 depression is much more marked in L. roskellyae than in W. hilmeri. In contrast, the frontals of W. vanderleueri are highly inflated giving the skull a distinctly arched profile, the highest point of the skull occurring in this region of inflation. The profile of T. carnifex is also slightly arched and the apex of the skull in that species is more posteriorly situated, near the frontoparietal suture. The supraorbital processes are much thinner and less procumbent than in species of Wakaleo.

In lateral view, the depression of the frontals in L. roskellyae accentuates differences in the rostrum which, because of expansion of the posterior region of the nasals, is more dorsally inflated than in T. carnifex and species of Wakaleo. In profile, the region of the rostrum above the narial opening is less curved than in species of Wakaleo and has a gentle slope that is more similar to that of T. carnifex. The sides of the rostrum are less rounded than in these larger thylacoleonids.

The nasal bones of L. roskellyae are similar in shape to those of other thylacoleonids, however, they are most similar to those of T. carnifex in their broad, rounded, dorsal curvature, their less-rounded anterior ends and in their relative length (i.e. approximately one-third the total skull length). In contrast, the nasals of species of Wakaleo are slightly pinched, are posteriorly depressed and relatively longer. In profile, the nasal bones of L. roskellyae are proportionately much deeper than those of Wakaleo species or T. carnifex.

The premaxilla of L. roskellyae differs from those of Wakaleo species and T. carnifex in being less convex and in being proportionately shorter in height and length. In general shape it is most similar to species of Wakaleo although the premaxillomaxillar suture is more vertically inclined than in W. vanderleueri. In Lekaneleo roskellyae this suture, as in W. hilmeri, intersects the posterior part of the canine alveolus whereas in both W. vanderleueri and T. carnifex this suture intersects the anterior part of the canine alveolus.

The maxilla of L. roskellyae is generally similar in shape to those of species of Wakaleo although it is less robust. The outline of the maxilla is most similar to that of W. vanderleueri in forming a relatively vertical suture with the premaxilla, 187 however, it is proportionately longer than the maxilla of that species. The premaxillomaxillar suture intersects the nasals at approximately their midpoint as it does in W. vanderleueri. In contrast, the maxillae of T. carnifex are dorsally narrow and their articulation with the premaxillae is inclined posteriorly and intersects with the posterior third of the nasals. The contours of the maxillae anterior to the orbit are more similar to T. carnifex being gently rounded dorsoventrally. This region lacks the anteorbital depression seen in species of Wakaleo and which is most strongly developed in W. vanderleueri. The anterior region of the lachrymal in L. roskellyae resembles that of T. carnifex being relatively narrow and also lacks the relatively large lachrymal fossa present in Wakaleo species. However, unlike T. carnifex and similar to species of Wakaleo, the anterior margin of the orbit is gently rounded (in T. carnifex it is developed into a sharp ridge) and the lachrymal foramina are situated relatively low down in the anterior margin of the orbit.

The maxillary wing of the zygomatic arch is more gracile than those of Wakaleo species and T. carnifex. Its long, oblique maxillojugal suture is most similar to that of W. vanderleueri, whereas that of W. hilmeri and T. carnifex is more vertically oriented. The relative narrowness of the maxillojugal articular surface suggests the jugal was less robust than those of the larger thylacoleonids. In contrast to W. vanderleueri, the zygomatic wing of the maxilla lacks a masseteric process as it does in W. hilmeri and T. carnifex.

A characteristic feature of thylacoleonids that is exhibited by L. roskellyae is its supraorbital processes although these are relatively small and thin in comparison to those of species of Wakaleo and T. carnifex. In dorsal view, these processes in L. roskellyae do not project beyond the suborbital margin of the maxilla, however, in species of Wakaleo they project well into the orbit and in Thylacoleo carnifex they develop into a postorbital bar.

As in species of Wakaleo, the palatine bone of L. roskellyae completely encloses the sphenopalatine foramen. In contrast, this foramen in T. carnifex is greatly enlarged and the palatine only forms its posterior border. Lekaneleo roskellyae, however, does 188 differ from species of Wakaleo in lacking the robust lateral processes that lie just laterally of the foramen rotundum where the palatine meets the alisphenoid.

The palate of L. roskellyae exhibits a shape characteristic of marsupial lions being broad and tapering markedly anteriorly. The palate more closely resembles those of species of Wakaleo rather than species of Thylacoleo, being flatter than and not as arched as the latter and in having a relatively long (anteroposteriorly) hard palate and moderately-sized posterior palatal fenestra. The posterior palatal fenestra are relatively larger than those in Wakaleo species accounting for slightly less than half the total length of the palate; in Wakaleo species they account for approximately one third of the palatal length while in T. carnifex and T. crassidentatus they are greatly enlarged. The anterior palatal fenestrae in L. roskellyae are proportionately longer than those in Wakaleo species and Thylacoleo species, extending posteriorly well beyond the canine alveoli.

Posterior to the hard palate the roof of the posterior nasopharyngeal passage is proportionately broader and more gently arched than in W. vanderleueri and T. carnifex and is more similar to that of W. hilmeri. The septum on the presphenoid is similar to the rounded structure in W. hilmeri and unlike that of W. vanderleueri which is sharper and thinner. The medial edges of the postalveolar processes in L. roskellyae are rounded like those in T. carnifex. They are unlike those of W. vanderleueri and W. hilmeri which are flat and have a relatively sharp edge. The crania of Wakaleo species and T. carnifex possess transverse palatine processes, however, the presence of these structures in L. roskellyae is indeterminate because of damage at the posterior margins of the palate.

The postorbital constriction in L. roskellyae occurs at approximately the midpoint of the skull. This is similar to the area of constriction in T. carnifex and is slightly more anterior to the area of constriction seen in W. hilmeri. In W. vanderleueri the constriction occurs more posteriorly, at about the posterior two-thirds of the skull. In dorsal view, the neurocranium of L. roskellyae differs significantly from those of other thylacoleonids in being proportionately longer and wider. The majority of the neurocranium is roofed by the parietal bones which are distinctly convex and are 189 proportionately broader and longer than those of W. hilmeri, W. vanderleueri and T. carnifex. In contrast, the parietals of those species are broadly concave and form taller and more robust sagittal crests. Associated with these differences in the parietals are the differences seen in the squamosal contribution to the neurocranial wall. In L. roskellyae the squamosal covers the inferior third of the lateral neurocranium. In both species of Wakaleo, and even more so in T. carnifex, the squamosal covers approximately 50 to 70% of the lateral neurocranium. The zygomatic limb of the squamosal is positioned low down in the neurocranium of L. roskellyae, a position similar to those in the comparative crania. The posterior dorsal margin of the arch is low and similar to that seen in both species of Wakaleo and lacks the rounded margin and steeply descending crest of T. carnifex. At the posterior margin of the neurocranium the broad, laterally-positioned, subsquamosal fossa is more similar to those seen in species of Wakaleo. In T. carnifex the subsquamosal fossa is round and positioned more dorsomedially.

Viewed laterally, the relationships of the bones of the anterior neurocranium of L. roskellyae are similar to those of species of Wakaleo; i.e. the alisphenoid projects dorsally to lie between the squamosal and the frontal and contacts the parietal along its dorsal margin. In T. carnifex, the alisphenoid does not extend dorsally to contact the parietal but is bordered dorsally by the frontal which extends posteriorly to contact the squamosal. The infratemporal crest of L. roskellyae is not as sharp or as strongly developed as those in species of Wakaleo. It also differs by being inclined posterodorsally and does not extend as far anteriorly, descending to about the midpoint of the sphenorbital fissure. In Wakaleo species the anterior part of this crest is more horizontal and extends to the anterior margin of the sphenorbital fissure.

The morphology of the alisphenoid below the infratemporal crest also differs from that of Wakaleo species. In L. roskellyae most of this region is gently convex and is generally more similar to that of T. carnifex. In contrast, species of Wakaleo exhibit a distinct concave fossa between the infratemporal crest and the sphenorbital fissure, the margin with the latter also taking the form of a sharp crest. The region of the pterygoid fossae in Lekaneleo roskellyae also differs from those of species of Wakaleo and T. carnifex; whereas the medial pterygoid fossa in the latter taxa are 190 predominantly concave, in L. roskellyae the fossa has a small flat anterior region that is separated by a small ridge from a broader slightly convex posterior surface. The area of attachment for the lateral pterygoid also differs, again being more convex in L. roskellyae, and is relatively flatter or concave in Wakaleo spp. and T. carnifex.

The occiput of L. roskellyae has a gently-arched outline that is most similar to that of W. hilmeri and is not steeply arched like those of W. vanderleueri and T. carnifex. The occipital surface lacks the deep medial fossa seen in those species. However, like those species, L. roskellyae exhibits a medial septum but differs in it being relative short and sharp. The nuchal crest of L. roskellyae is relatively smaller than those in species of Wakaleo and T. carnifex and lacks the prominent anterior curvature seen in W. hilmeri and T. carnifex. The short and relatively broad paroccipital processes are similar to those in Wakaleo species and T. carnifex.

The basicranium of L. roskellyae is more similar to those of species of Wakaleo than to Thylacoleo carnifex. The damage to the basicranial regions of both L. roskellyae, W. hilmeri and W. vanderleueri provide evidence that these species have a tympanic wing consisting of two components, the alisphenoid and the squamosal, unlike T. carnifex in which the tympanic wing is constructed completely by the squamosal. Although the posterior extent of the tympanic wing is not known in L. roskellyae, the bilaminar structure of the portion that is preserved suggests that in its entirety it was probably similar to the tympanic wing seen in species of Wakaleo. As in other marsupial lions the basioccipital is broad.

The region of the glenoid fossa of L. roskellyae is similar to those of W. hilmeri and W. vanderleueri being relatively broad with a shallow mandibular fossa and bordered medially by a low medial glenoid process. In contrast, T. carnifex has a ventrally convex articular eminence, a large mandibular fossa, and a deep, steeply-sided medial glenoid process. The position of the postglenoid process is similar to that in species of Wakaleo, lying relatively laterally to the mastoid process and is unlike T. carnifex, in which the process has a more medial position, lying directly anterior to the mastoid. As in Thylacoleo carnifex, the postglenoid process of L. roskellyae is relatively short and broad, however, it is relatively much thinner. The posterior 191 surface of this process in both these taxa is concave although the causes of the concavity are different. In L. roskellyae it is caused by a lateral extension of the postglenoid cavity and in T. carnifex it is caused by a posterior curling of the postglenoid process such that the process nearly reaches the mastoid process and forms a partial floor to the postglenoid cavity. In species of Wakaleo the postglenoid process is deep and narrow and has a slightly convex posterior surface.

The postglenoid cavity varies considerably within thylacoleonids. In L. roskellyae it is relatively broad, expanding laterally to invade the posterior surface of the postglenoid process. This fossa in species of Wakaleo is relatively smaller and does not extend laterally into the postglenoid process. The fossa in W. vanderleueri differs from that of W. hilmeri in being divided horizontally by a septum. Thylacoleo carnifex possesses a very large and broad postglenoid process which encloses the postglenoid fossa and prevents it from having such a direct posterior opening as in Lekaneleo and species of Wakaleo.

Compared to species of Wakaleo and to T .carnifex, the width of the outer ear canal in L. roskellyae is proportionately much larger. At its most lateral margin (between the squamosal portion of the mastoid and the lateral edge of the postglenoid process) it is 8.8mm wide and is only slightly narrower than that of W. vanderleueri (approx. 10 mm) and half the diameter of T. carnifex (approx. 20mm). As in species of Wakaleo, the medial part of the canal has a broad ventral opening. However, in T. carnifex the ventral opening of the canal is greatly reduced by the posterior curling of the postglenoid process.

The mastoid process of the petrosal is generally similar to those in species of Wakaleo, but L. roskellyae differs in having the medial portion of the process enlarged. The enlarged medial end of the mastoid partly encloses the sulcus that courses along its posteromedial border with the paroccipital. In W. hilmeri this sulcus is shallower and lies more centrally on the ventral surface of the mastoid process. In L. roskellyae the sulcus for the stylomastoid division of the facial nerve lies mainly within the mastoid process unlike in species of Wakaleo and in T. carnifex in which 192 it courses through the squamosal which is medially expanded across the anterior surface of the mastoid.

The periotic of L. roskellyae is more similar in shape and position to those of Wakaleo species than to that of T. carnifex. As in species of Wakaleo there is a crescentic epitympanic wing at the anterolateral margin with the alisphenoid, although this structure is relatively larger and deeper in L. roskellyae. Lekaneleo roskellyae differs from species of Wakaleo in having a periotic surface that is more convex, in lacking a prominent rostral tympanic process, and has a relatively taller septum at the posterior end of the stapedial fossa. The posteroventral margin of the L. roskellyae periotic also differs from that of species of Wakaleo in bearing a broader and deeper fossa between the fenestra cochleae and the posterior lacerate foramen.

Upper dentition The dental formula of Lekaneleo roskellyae (I1-3 C1 P1-3 M1-4) is similar to that of Wakaleo hilmeri and W. pitikantensis. The size pattern of the incisor alveoli (i.e. I1 > I3 > I2) is similar to those of species of Wakaleo, T. crassidentatus and T. carnifex. The position of the canine in L. roskellyae (i.e. closer to I3 than to P1) is also similar to that seen in specimens of W. hilmeri that possess P1. The molar gradient is similar to that seen in species of Wakaleo.

The P3 of L. roskellyae is 25 - 66% smaller than P3 in species of Wakaleo; it is 25% smaller than W. hilmeri, 44% smaller than W. vanderleueri, and 66% smaller than W. alcootaensis (see Appendix 1; Table F and G). The morphology of P3 is similar to those taxa, especially W. hilmeri, but a number of differences are also evident. Unlike most species of Wakaleo, in occlusal view, the P3 exhibits a narrowing of the crown between the anterior and posterior roots. The P3 of W. hilmeri also shows this feature although the constriction on the lingual side is less pronounced. In contrast, in W. vanderleueri the lingual constriction is completely lost and the lingual surface is strongly convex. In comparison to species of Wakaleo, its longitudinal blade is relatively straight and the whole tooth has a less bowed appearance, especially compared to those of W. vanderleueri and W. alcootaensis. The anterior crest is 193 slightly rounder than that of W. hilmeri and differs greatly from the relatively steep crests of other species of Wakaleo. The anterior cusp is marginally taller than the posterior cup, a condition also seen in W. alcootaensis, although the heavy wear exhibited on the sole existing premolar of the latter may obscure the original cusp heights. In W. vanderleueri and W. hilmeri, specimens of which also demonstrate moderate to heavy wear, the anterior and posterior cusps are approximately equal in height.

The anterior crest is most similar to that of W. hilmeri having a rounder curvature and curving lingually at its base. In W. vanderleueri, the lingual curve of the anterior crest is short and most of the crest runs vertically and in W. alcootaensis the crest terminates in a cuspule. The lingual crest in L. roskellyae is most similar to that of species of Thylacoleo being long and relatively narrow, although it is not as sharp. It lacks the robust basal cuspule that is present in species of Wakaleo. The anterior buccal crest, although moderately developed, does not exhibit the basal thickening seen in W. hilmeri and is relatively longer and less robust than those in W. vanderleueri and W. alcootaensis. The horizontally-bowed posterior longitudinal blade that runs from the posterior cusp to the posterior edge of the tooth is also present in all species of Wakaleo. Unlike those taxa, the posterior longitudinal blade of Thylacoleo hilli descends from the posterior cusp in a gently downward curve. As in W. hilmeri and W. vanderleueri the P3 of L. roskellyae exhibits a small buccal crest at the posterior end of the crown. This crest in Wakaleo vanderleueri differs by commencing a short distance before the end of the posterior blade and is more vertically oriented. This buccal crest is not present in species of Thylacoleo or in W. alcootaensis.

The molar morphology of L. roskellyae is generally similar to that of species of Wakaleo and most closely resembles that of W. hilmeri, however, a number of differences are present in the morphology of M1. Firstly, unlike the predominantly triangular-shaped M1and M2 seen in species of Wakaleo (i.e. W. vanderleueri, W. oldfieldi and W. alcootaensis), those of L. roskellyae are relatively square in outline. In L. roskellyae the posterior edge of M1 is relatively perpendicular to the buccal 194 edge and its posterobuccal corner is rounded. Consequently, a distinct gap lies between the posterobuccal corner of M1 and the anterior edge of M2. In contrast, in species of Wakaleo the buccal margin of M1 is lengthened relative to its lingual margin. This results in the posterobuccal corner of M1 abutting against M2 and consequently the posterior-most point of M1 (i.e. posterobuccal corner) lies at a level that is posterior to most of the anterior margin of M2. Secondly, the M1 of L. roskellyae has a relatively broader stylar shelf and thus a relatively more lingual location of the longitudinal blade complex that runs through the paracone and metacone. In species of Wakaleo the stylar shelf is narrower and hence the longitudinal blade complex is positioned more laterally. A third difference is seen in the position of the protocone which is less-lingually located in comparison to Wakaleo species. The more medially located centrocrista and protocone results in M1 having a relatively narrower trigon basin. As in W. hilmeri (and probably W. vanderleueri), M1 possesses a metaconule, although in L. roskellyae its position is comparatively more lingual.

The M2 of L. roskellyae is most similar to that of W. hilmeri, having a relatively square outline and a metaconule. As in species of Wakaleo, the trigon basin is crenulated. It differs from W. hilmeri and W. pitikantensis in its smaller size and possession of a broader, more fusiform posterior basin. It also differs from W. pitikantensis in its possession of a metaconule. No metaconule is apparent on M2 of W. pitikantensis (Rauscher 1987); in L. roskellyae the edge of the postprotocrista rises posteriorly towards the metaconule unlike the remnant edge in W. pitikantensis which remains relatively horizontal. In addition, the lingual margin of W. pitikantensis is not as square as that of L. roskellyae.

Dentary and Lower Dentition The dentary and lower dentition is significantly smaller than those of any species of Wakaleo or Thylacoleo (see Appendix 1; Table G). Comparison against other thylacoleonid dentaries was limited because of the fragmentary nature of both L. roskellyae dentaries. Comparison of the depth of the dentary relative to the length of the tooth row indicates that L. roskellyae has a shallower dentary compared to 195 species of Wakaleo and Thylacoleo. The tooth row of L. roskellyae exhibits the gentle buccal curvature also seen in species of Wakaleo and Thylacoleo.

Although the general dental morphology is similar to that of species of Wakaleo (i.e. there is a procumbent I1, a blade-like P3 and basin-like, crenulated molars) L. roskellyae exhibits a number of significant differences. Firstly, L. roskellyae differs from other thylacoleonids in having three teeth present between I1 and P3. All thylacoleonids have teeth in this region, the number varying with the genus; species of Thylacoleo bear two peg-like teeth (P1 and P2) that sit in small, vertically- oriented alveoli, Wakaleo oldfieldi and W. vanderleueri has one alveolus and W. hilmeri has a maximum of two.

P3 of L. roskellyae differs from those of other species of thylacoleonids in having a comparatively deep concave surface between the anterior and anterolingual crests. In W. hilmeri and T. carnifex this surface is shallower and broader. In contrast, in W. vanderleueri and W. oldfieldi this surface is broadly convex. The anterolingual crest of L. roskellyae is sharper and more prominent than those of W. oldfieldi and W. vanderleueri and in this condition it is more similar to W. hilmeri and species of Thylacoleo. In comparison to other thylacoleonids this crest is also relatively long and is most similar to that of Thylacoleo sp., cf. T. hilli. P3 of QM F20191 exhibits a bifurcation of the anterior crest at its base. The resulting flat, triangular surface is probably an articular facet for P2, the alveolus of which lies directly below. Bifurcation of this crest is also present in Thylacoleo sp., cf. T. hilli but in that species the facet lies laterally to the P1 alveolus and is well anterior of the P2 alveolus.

Generally, the lower molars of L. roskellyae are most similar to those of species of Wakaleo, but they differ in having much broader talonid basins relative to the width of the tooth. M1 and M2 of L. roskellyae have basins that are nearly as wide as their crowns unlike those of Wakaleo species which have basins that are much narrower relative to the crown width. M2 also differs in having a talonid basin with a much squarer outline and in having a trigonid basin that is relatively broader, having a 196 width nearly three-quarters that of the anterior width of the tooth. The trigonid basin of M2 in species of Wakaleo is less than half the width of the tooth and is square in outline rather than triangular.

Discussion

The diagnosis for Lekaneleo is amended to include features identified on the cranial and dentary material of L. roskellyae. Initially, QM F23453 was referred to the genus Priscileo Rauscher because it shared a number of significant characters with P. pitikantensis, i.e. presence of M4, relatively small P3 and relatively square M2 (Gillespie 1997). Re-evaluation of the P. pitikantensis holotype with more complete material of the new plesiomorphic Wakaleo taxon (W. hilmeri) indicates that most of the features that were formerly diagnostic for Priscileo are also present in hilmeri, a taxon that in all other regards sits comfortably in the genus Wakaleo. Accordingly, the species pitikantensis is herein referred to the latter genus (see Chapter 4).

Although L. roskellyae shares certain features of its upper dentition with W. hilmeri, the morphological differences in its cranium and lower dentition to those of species of Wakaleo strongly support the allocation of this species to a separate genus. These differences include; its small size, steeply-sided rostrum, thin supraorbital processes, small sagittal and nuchal crests, relatively flat or convex pterygoid fossa, relatively broad parietals, horizontal subsquamosal foramen, large posterior palatal fenestra that have gently sloping lateral margins, a postglenoid cavity that is laterally expanded into the posterior surface of a thin postglenoid process, its lack of a prominent rostral tympanic process on the periotic, relatively narrow mastoid process, reduced squamosal contribution to the posterior wall of the external auditory meatus, a stylomastoid sulcus that courses through the mastoid, lack of lateral alisphenopalatine lateral processes, presence of three teeth between I1 and P3 and very broad talonid basins on the lower molars.

It is probable that many of the distinctive features of the cranium of Lekaneleo roskellyae, i.e. its small sagittal and nuchal crests, thin supraorbital processes, depressed frontals, relatively convex pterygoid fossa, weak rostral tympanic process, 197 frontoalisphenoid contact, bilaminar tympanic wing and large parietals are also plesiomorphic features within thylacoleonids (see Chapters 11 and 12). Similarly, the distinguishing features of the lower dentition, i.e. the presence of an extra tooth in the lower antemolar dentition and broad, shallow talonid basins also suggest a further plesiomorphic condition.

In their review of Thylacoleo, Archer and Dawson (1982) identified the two teeth that lie between I1 and P3 as probably being premolars (i.e. P1 and P2) although they point out that it is difficult to determine the homology of these teeth. Similarly, the identification of the three teeth that lie anterior to P3 of L. roskellyae is also speculative. In both L. roskellyae and W. hilmeri the posterior-most alveolus that lies between I1 and P3 is vertically-oriented and lies directly against the base of P3 suggesting that this tooth was probably P2. In L. roskellyae, the vertical orientation of the alveolus that lies directly in front of that alveolus (i.e. P2) strongly suggests that it was P1. The anterior-most alveolus in L. roskellyae opens almost antero- horizontally, an orientation similar to that of the single alveolus present in W. oldfieldi. Whether this tooth represents a canine or posterior incisor is difficult to determine. The adult formula for the lower dentition is thus; I1, ?I2 I3 or C1, P1-3, M1-4.

Murray et al. (1987) separated thylacoleonids into two subfamilies, the Thylacoleoninae and the Wakaleoninae, tentatively placing Priscileo pitikantensis Rauscher 1987 along with species of Thylacoleo in the Thylacoleoninae. The characters used to separate the two subfamilies are listed below (Table 8.2). Considering the limited nature of the pitikantensis material available at that time, referral to the Thylacoleoninae was based primarily on the presence of P1.

Table 8.2: Characters distinguishing the subfamilies Wakaleoninae and Thylacoleoninae (Murray et al. 1987).

Wakaleoninae Thylacoleoninae 198

1 P1/1 absent P1/1 present 2 emphasis of tooth size & deep incurvation anterior to P3 development 3 distribution of shearing function shearing function mainly P3/3 throughout molar row 4 squamosal/alisphenoid bilaminar squamosal tympanic wing tympanic wing 5 position of ventrolateral course of alisphenosquamosal suture alisphenosquamosal suture medial to tympanic wing 6 relatively larger alisphenoid wing reduction of upper canine 7 lack of frontal/squamosal contact frontal/squamosal contact on lateral on neurocranium cranial wall 8 total absence of postorbital bar presence of complete postorbital bar 9 lack of fusion of the basisphenoid fusion of basisphenoid and and basioccipital basioccipital 10 position of premax/max. suture weak/absent transverse palatal ridges relative to canines and anterior palatal fenestrae 11 tribosphenic, conule-bearing, 3 possible that M4 in P. pitikantensis rooted upper molars is rotated clockwise about 20 with respect to long axis of tooth row

The new cranial material of L. roskellyae and Wakaleo hilmeri and the referral of pitikantensis to Wakaleo, has altered the utility of some these features, for example, 1 the presence or absence of P /1 is lost to both subfamilies because of its presence in the taxon pitikantensis and W. hilmeri. Of the ten remaining wakaleonine characters, L. roskellyae exhibits nine and differs only by degree in the posterior location of the premaxillomaxillary suture relative to the canine. Of the ten remaining thylacoleonine characters, only the condition of the transverse palatal ridges is evident in L. roskellyae. With the referral of pitikantensis to Wakaleo, the thylacoleonine diagnostic trait of possible clockwise rotation of M4 is lost (or transferred to wakaleonines). On the weight of the number of wakaleonine characters displayed by L. roskellyae it is evident that Lekaneleo could be placed within this subfamily. Cladistic analysis of the intrafamilial relationships of thylacoleonids is presented in Chapter 11. The similarity of many of the features of the dentition and basicranium of L. roskellyae and W. hilmeri supports a hypothesis of shared ancestry for these genera. The presence of both L. roskellyae and W. hilmeri in sediments 199 estimated to be late Oligocene in age suggests that this ancestor may have existed in the early Oligocene.

The dental features of Lekaneleo roskellyae, in particular the small P3 and molar size, square molar shape, presence of a metaconule, and full premolar and molar complement are plesiomorphic features within Thylacoleonidae (see Chapters 11 and 12). However, these same features clarify some questions about the relationships of the family with the Order Diprotodontia. It was once commonly believed that thylacoleonids evolved from a diprotodontian ancestor, probably a phalangerid, which had quadritubercular upper molars including a hypertrophied metaconule (Krefft 1872, Broom 1898, Bensley 1903, Ride 1964, Archer 1976). Archer and Rich (1982) hypothesised that the tritubercular shape of the molars of Wakaleo alcootaensis were secondarily derived from an ancestral quadritubercular shape through suppression of the metaconule. More recently it has been suggested that the triangular molars of species of Wakaleo are plesiomorphic for the family (Murray et al. 1987). The primitive dental features exhibited by L. roskellyae, especially the presence of a metaconule and square molar shape, however, tend to support the original quadritubercular diprotodontian hypothesis. 200

CHAPTER 9

Lekaneleo deminutivus sp. nov.

9.1 Introduction

In 1956 an isolated P3 of a thylacoleonid was recovered from Town Cave, Curramulka, South Australia (Pledge 1977). Morphological similarity with the premolar of species of Thylacoleo indicated affinity with that genus, however, the premolar differed in being only half the length of the equivalent tooth in T. carnifex. On the basis of the tooth’s relatively small size and the interpretation that it was a permanent tooth from an adult individual, it was determined to represent a new species, Thylacoleo hilli (Pledge 1977).

At Riversleigh, a similar situation is evident with the discovery of a very small thylacoleonid P3 from Dirk’s Towers Site. Although similar in overall morphology to the P3 of Lekaneleo roskellyae, it differs in a number of features and is only two- thirds the length.

9.2 Systematics

SYSTEMATICS

Superorder MARSUPIALIA Illiger, 1811 Order DIPROTODONTIA Owen, 1866 Family THYLACOLEONIDAE Gill, 1872 Genus Lekaneleo, Gillespie 2006

Lekaneleo deminutivus sp. nov.

(Fig. 9.1) 201

A BBB

C

longitudinal D anterior cusp blade posterior cusp posterior longitudinal blade

anterolingual crest

5.7mm

Fig. 9.1: Lekaneleo deminutivus sp. nov., QM F24549 holotype,

left P3; Dirk’s Towers Site. Scale bar = 5 mm

A. lingual view B. buccal view C. stereo occlusal view D. major cusps and crests 202

Holotype: QM F24549, a left P3.

Type Locality: Dirk’s Towers Site, Riversleigh World Heritage Area, Lawn Hill National Park, northwestern Queensland.

Age: Dirk’s Towers Site is regarded as being part of Riversleigh’s System B sequence which is interpreted to be early Miocene in age (Archer et al. 1997).

Etymology: the species name is from the Latin deminutivus, meaning diminutive. The name refers to the small size of the tooth (and probably the animal as a whole).

Species diagnosis: Differs from Lekaneleo roskellyae in P3: being smaller; lacking a constriction of the crown where the anterior and posterior roots meet; narrowing posteriorly and lacking lingual and buccal inflation of the posterior crown; having a weak anterolingual crest; having a deep notch on the longitudinal blade between the anterior cusp and posterior cusp; having a deeper curvature on the posterior longitudinal blade and in the anterior crest consisting of two parts.

Differs from Microleo attenboroughi in having: a greater separation between the anterior and posterior cusps; a broader buccal valley between these cusps; a weaker posterobuccal crest and shallower posterobuccal basin; the posterior longitudinal blade ascend to the posterior edge of the tooth at a more gentle angle; the wear facets directed lingually rather than posteriorly and in lacking an anterobuccal crest anterior to the anterior cusp.

Description

The tooth, a left P3 (Fig. 9.1), measures 5.7mm in length and 2.6mm at its widest point (over the posterior root). It is roughly fusiform in occlusal outline, with the anterior end narrowing more abruptly than the posterior end. In occlusal view the buccal edge of the tooth is relatively straight while most of the lingual margin bulges lingually. The crown bears an anterior and posterior cusp linked by a V-shaped or notched longitudinal blade. This notch occurs two-thirds of the way between the two 203 cusps. The anterior cusp is slightly taller than the posterior cusp. In profile the occlusal edge resembles an "M" with a slight upstroke at the end. Between the anterior and posterior cusps the lingual edge of the longitudinal blade bears a narrow shearing facet. The anterior crest consists of two parts (inferior and superior); the inferior part of the crest runs along the anterior margin from the anterior cusp to about half way up the crown where it meets a vertical but lingually located crest that continues to the base of the tooth. A small, near-vertical, anterolingual crest descends from the anterior cusp, merging with a slight swelling at the crown base. A steep, broad valley lies between the anterior crest and the anterolingual crest. Posterior to the anterolingual crest the base of the crown is swollen, forming a rounded shelf that diminishes over the posterior root. A small piece of enamel is missing from the lingual base of the crown above the posterior root. From the posterior cusp the longitudinal blade descends and then rises toward the posterior margin of the tooth. The lingual edge of this part of the longitudinal blade also bears a narrow shearing facet. At the posterior margin of the tooth the longitudinal blade curves buccally forming a small posterobuccal crest. The buccal surface of the crown below the anterior and posterior cusps is broadly convex. A broad but sharply angled valley links the two cusps. The posterobuccal surface of the crown forms a small basin between the posterior cusp and the posterobuccal crest. The posterior root is slightly larger than the anterior and the roots are slightly bowed toward each other.

Comparisons

The P3 of L. deminutivus shows most similarity to that of L. roskellyae. Although it is two-thirds the size (Appendix 1; Table B), it shows a similar development of the anterior and posterior cusps and the longitudinal blade. It differs from L. roskellyae in lacking a constriction of the crown between the anterior and posterior roots. Additionally, it differs in having the crown taper posteriorly (in L. roskellyae the posterior part of the crown widens, especially lingually), has a deep notch in the longitudinal blade and a relatively deep curvature of the posterior longitudinal blade, and exhibits relatively weak development of the anterolingual crest. Although it is evident that the anterior crest is made up of two crests, the general curvature (i.e. 204 having a basal lingual curve) of the crest is similar to that in L. roskellyae and species of Wakaleo.

The P3 of L. deminutivus also shows some similarities in size and general morphology to that of Microleo attenboroughi, however, the latter differs in having the anterior and posterior cusps more closely positioned, and thus it has a much narrower valley between these cusps on the buccal surface. Microleo attenboroughi also differs in having a more strongly developed posterobuccal crest and deeper posterobuccal basin, in possessing a small anterobuccal crest anterior to the anterior cusp, and in having the posterior longitudinal blade ascend to the posterior edge of the tooth at a much more acute angle so that the posterior edge of the crown is rather low relative to the posterior cusp. Like L. deminutivus, M. attenboroughi exhibits a lingual swelling of the crown between the anterior and posterior roots, but this swelling is much greater in the latter. Lekaneleo deminutivus also lacks the narrowing of the crown directly anterior to this swelling as seen in Microleo. Another difference is seen in the direction of the wear facet on the longitudinal blade, which in Microleo is directed more posteriorly, rather than lingually as in L. deminutivus.

Discussion

Compared to all other species of thylacoleonids, QM F24549 exhibits a morphology that is most similar to the P3 of Lekaneleo roskellyae (cusp placement, development of the longitudinal blade and posterobuccal crest and broad buccal valley) hence it is allocated to that genus. Although similar in size to the P3 of Microleo attenboroughi, its morphological differences preclude it from being assigned to that genus. Designation as a new species of Lekaneleo is based on its smaller size with respect to L. roskellyae and its morphological differences, especially its fusiform shape, lack of constriction of the crown between the anterior and posterior roots, weak anterolingual blade and deeply notched longitudinal blade. 205

The 30 % size difference between QM F24549 and the P3s of L. roskellyae, although not as great as the difference between Thylacoleo hilli and T. carnifex (50%), is sufficiently large to support specific differentiation. This degree of difference in P3 size is greater than that observed between species of Wakaleo (14 to 26%, see Appendix 1; Table A) and is also greater than the difference between samples of T. crassidentatus and T. carnifex (20%) from Queensland. This size difference is also greater than the interspecific size difference seen in morphologically similar populations of Quaternary kangaroos (e.g. 27% between Macropus titan and M. giganteus; Bartholomai, 1975). Intraspecific variation in P3 size in large samples of T. carnifex from New South Wales and South Australia does not exceed 20% (Finch and Freedman 1982).

Consideration was also given to the possibility that QM F24549 may represent a deciduous premolar of L. roskellyae, however, weight of evidence tends to support the conclusion that this tooth is a permanent premolar. To date, no deciduous premolars have yet been identified for any species of thylacoleonid. Although T. carnifex boasts a relatively abundant fossil record, the dentaries of its pouch young have been shown to possess a fully functional P3 and a deciduous premolar is lacking (Finch 1971, Pledge 1975). Although absence in the fossil record does not preclude the possibility that this tooth may be a deciduous premolar, at least two observations lend support to the belief that the tooth is from an adult individual. Firstly, the roots of the tooth are relatively long and well-formed and show a slight tendency to curve toward each other, providing no indication of being impinged upon by a descending permanent tooth. Additionally, this tooth only has two roots while the deciduous premolars of other diprotodontians (e.g. Nimbadon lavarackorum, Propalorchestes novaculacephalus, Trichosurus vulpecula) display three splayed roots. Secondly, the degree of wear on the tooth suggests it is from an adult animal; the presence of distinct shearing facets and the appearance of pale underlying regions of dentine along the lingual edge of the longitudinal blade, especially near the anterior and posterior cusps, indicate fully functional usage of the tooth. Enamel has also been spalled off the posterolingual surface, although it is not clear whether this is the product of normal wear or has occurred post-mortem. 206

Predicting the likely size or form of a thylacoleonid deciduous premolar is also made difficult by the variable nature (and lack of information) of the deciduous premolar in other vombatiform taxa. Observation of juvenile skulls of Phascolarctos cinereus (the apparent closest living relative of thylacoleonids) in the Australian Museum collection (M35651, M2068, M32254, M7364, M7486) indicate that the adult P3 is present at a relatively early age: all skulls lacked any sign of a deciduous premolar and exhibited a fully formed P3 and M1 although the canine and most of the posterior molars had yet to erupt. The more distantly related vombatiform Nimbadon lavarackorum (AR 18062) possesses a deciduous premolar that is relatively small and cuspate in comparison to the permanent tooth. The deciduous premolar of Propalorchestes novaculacephalus (QM F50605) is relatively large and very similar in size and morphology to the adult P3 and is present in the tooth row with the adult molars M 1-3. Unlike Phascolarctos, juveniles of Trichosurus vulpecula in the Australian Museum collection (M11625, M5311, M5074, M633, M7706, and M9692) all displayed a relatively small bicuspid deciduous premolar and M1 although most of the posterior molars had yet to erupt.

The recovery from Dirk’s Towers Site of three species of thylacoleonid, L. deminutivus, L. roskellyae and W. hilmeri, indicates a level of thylacoleonid sympatry previously unseen at any fossil locality.

The phylogenetic relationships of L. deminutivus are examined in Chapter 11. 207

CHAPTER 10

Microleo attenboroughi gen. et sp. nov.

10.1 Introduction

Paired maxillary fragments of a very small thylacoleonid were recovered from Neville’s Garden Site at Riversleigh. The teeth in these specimens are the smallest known for any thylacoleonid. In addition, although the general morphology of the premolar and molars are characteristically those of a marsupial lion, some features of the dentition are unique within Thylacoleonidae.

10.2 Systematics

SYSTEMATICS

Superorder MARSUPIALIA Illiger, 1811 Order DIPROTODONTIA Owen, 1866 Family THYLACOLEONIDAE Gill, 1872

Microleo gen. nov.

Type Species: Microleo attenboroughi sp. nov.

Diagnosis: distinguished from all other thylacoleonids by its very small size and in having P3 with the following features: a tapering posterior end that lacks lingual and buccal inflation; the anterior cusp linked to posterior cusp by a straight, not lingually-bowed, longitudinal blade; the posterior longitudinal blade straight rather than buccally deflected and ascending steeply to the posterior margin of tooth; the posterior margin approximately half the height of the posterior cusp; a deep rather than shallow posterobuccal basin; a very weak anterolingual blade; a rounded 208 posterolingual blade that connects to a prominent lingual swelling at the junction of the anterior and posterior roots; a deep valley that lies immediately anterior to posterolingual blade and; a small vertical crest on the buccal surface between anterior blade and anterobuccal blade. It also differs from other thylacoleonids in having: M2 with distinct metaconule; the lingual margin of its trigon basin exhibiting a small notch slightly anterior to the metaconule; and the protocone taller than the paracone.

Microleo also differs from species of Lekaneleo in having the P2 alveolus large relative to P3 and positioned anteriorly rather than anterolingually to P3; M2 with the protocone more lingually positioned relative to the metaconule; a weaker preprotoconule crista; the postprotoconule crista running in a more posterobuccal direction rather than posteriorly; and a more prominent metacone and metaconule.

Microleo is additionally distinguished from all species of Wakaleo, except W. hilmeri and W. pitikantensis, in possessing P1 and M4, and from species of Thylacoleo by having M3-4.

Etymology: micro from the Greek meaning small, and leo from the Latin meaning lion.

Microleo attenboroughi sp. nov.

(Figs. 10.1, 10.2)

Holotype: QM F41143, palatal fragment consisting of partial right and left maxillae. The right maxilla contains P3 and M2 and alveoli for P1-2 and M1; the anterior margin of the P1 alveolus is incomplete. The left maxilla contains M2-3, however, the crowns of P3 and M1 are absent having been broken off at the level of the palate. An alveolus for M4 and the maxillary root of the zygomatic arch are also preserved. 209 10mm

A B

P1

2.7 P2

P3 P3 A 4.6 B

1 M1 3.5 M ~16

M2 3.3 M2

3.3 3.6 2.8 M3

Fig. 10.1: Microleo attenboroughi, gen. et sp. nov., QM F41143 holotype, right (A) and left (B) maxilla, stereo occlusal views and measurements (mm); Neville’s Garden Site.

a

b l

p 2.5 mm

Fig. 10.2: Microleo attenboroughi, gen. et sp. nov., QM F42676 paratype, a left M2. Stereo occlusal view. a = anterior, p = posterior, l = lingual 210

Referred material: QM F42676, a left M2 from Neville's Garden Site.

Type Locality: Neville's Garden Site, Godthelp Hill, Riversleigh World Heritage Area, Lawn Hill National Park, northwestern Queensland.

Age: Neville’s Garden Site is regarded as being part of Riversleigh “System B” sediments (Creaser 1997) which are estimated to be early Miocene in age (Archer et al. 1997).

Specific Diagnosis: as for genus.

Etymology: attenboroughi, named in honour of Sir David Attenborough for his enthusiasm and dedication to the promotion of natural history.

Description

The holotype of Microleo attenboroughi is a fragmentary palate consisting of portions of both maxillae. More of the left maxilla is preserved than the right. The right maxillary fragment measures 15.2 mm in length and preserves P3 and M2. Alveoli are also present for M1. Two small premolar alveoli lie anterior to P3, the anterior margin of the anterior-most alveolus is missing. The left maxilla fragment measures 22.1 mm anteroposteriorly and contains M2-3. Unfortunately, the crowns of P3 and M1 have been broken away at the level of the palate. Alveoli for M4 are present; only the anteromedial one is complete, the lateral and posterior sockets are broken. The region anterior to P3 is damaged exposing the internal walls of the P2 alveolus and infraorbital canal. Only narrow portions (3-5mm) of the palate medial to both tooth rows are preserved. Other fragments of bone from the posterior cranium are preserved including a partial atlas.

Dentition 211

The alveoli of P1 and P2 are preserved. These alveoli are small and round. The P1 alveolus is broken but appears to be similar in size to the P2. The P2 alveolus measures 1.0mm long and 0.9mm wide. The P2 alveolus sits at the anterior base of P3, directly behind that of P1.

P3 is fusiform in occlusal outline. A serrated longitudinal blade runs the length of the tooth. The serrations are created by two principal cusps, one anterior, the other posterior. The anterior cusp is lower than the posterior although it is evident that the height of the anterior cusp has been reduced by wear. In profile, the longitudinal blade forms a weak "W" outline. The anterior cusp bears a posterolingually directed wear facet, the wear exposing a rounded area of dentine. Slight wear facets are also discernible on the lingual edge of the longitudinal blade anterior to both the anterior cusp and the posterior cusp and posterior to the posterior cusp. A crest extends anteriorly from the anterior cusp and ascends with a medial curve, merging with the anterior base of the crown. A small cuspule lies anterior to the anterior cusp at the anterobuccal edge of the longitudinal blade. On the buccal surface of the tooth a rounded, vertical crest or blade ascends from each principle cusp and merges with the base of the crown. Posterior to the posterior cusp, the longitudinal blade curves dorsolaterally forming a discrete posterobuccal crest. In so doing, a small, deep basin is formed on the posterobuccal surface between the posterior cusp and the posterior margin of the tooth. On the lingual surface, a near-vertical crest runs from the anterior cusp to a small basal swelling over the anterior root. Immediately posterior to the anterolingual crest a deep valley curves anteriorly from the posterior cusp to the base of the anterolingual crest. A rounded posterolingual crest ascends from the posterior cusp and runs slightly anteriorly, merging with a larger basal swelling approximately midway between the two roots. P3 has two roots, the posterior being slightly larger than the anterior.

No crown of M1 is preserved. In the left maxillary fragment the crown is broken away at the level of the palate indicating a relatively square basal outline and a slight lingual expansion of the base of the protocone. M1 has three roots, the medial and posterior root being subequal in size and larger than the anterior root. 212

M2 is subtriangular in outline and dominated by a large trigon basin. A paracone is located at the anterobuccal margin. From the paracone, a small preparacrista runs forward to the anterior edge of the tooth and a postparacrista runs posteriorly, swinging buccally to connect with the buccal margin of the tooth midway along the crown. These connecting cristae form the lingual margin of a small basin that lies between the paracone and the anterobuccal edge of the tooth. A weak anteriorly- curving crest runs lingually from the paracone towards the protocone. A narrow basin runs between this crest and the anterior margin. The protocone is similar in height to the paracone but shows slightly more wear. The base of the crown bearing the protocone bulges lingually. A rounded postprotocrista runs posteriorly to a metaconule at the posterolingual margin of the tooth. The postprotocrista is notched just anterior to the metaconule. The metacone is shorter than the paracone and protocone but taller than the metaconule. A transverse crista runs from the metacone to the metaconule and is notched in one or two places between the two cusps. A narrow basin lies between this incised crista and the posterior edge of the tooth. The trigon basin lacks crenulations although weak buttresses flank the notches on the posterior flank. M2 has three roots.

M3 is subtriangular in shape and dominated by a large trigon basin. The paracone and protocone are subequal in height. From the paracone a short preparacrista runs anteriorly to the edge of the crown and a postparacrista curves posterobuccally connecting with the buccal margin half way between the paracone and metacone. These two cristae and the anterobuccal edge enclose a small, shallow, anterobuccal basin. A third crest runs from the paracone in a lingual direction, becoming very weak as it approaches the protocone. A very small valley lies between this crest and the anterior edge of the crown. The base of the protocone is gently rounded and the underlying crown is deep but not lingually-inflated. A postprotocrista runs posteriorly and curves laterally to connect with the metacone. The metacone is weak. The trigon basin lacks crenulations. M3 is anchored by three roots. Facial skeleton 213

Description of the facial skeleton is based on the left maxillary fragment which is the larger of the two fragments. Only that part of the maxilla just dorsal and posterior to the infraorbital foramen and the wing of the zygomatic arch is preserved. The base of the infraorbital foramen is broken away showing a triangular-shaped internal canal which opens anteriorly. Dorsal and posterior to the foramen the maxilla gently curves convexly towards the wing of the zygomatic arch. At the base of the wing the curvature becomes slightly concave. The base of the zygomatic arch projects posterolaterally dorsal to M1 - M2. The lateral extremity of the maxillary root tapers posteriorly and bears a narrow groove near its inferior border marking the articulation area of the jugal. A remnant of the jugal lies on the anterodorsal surface of the zygomatic arch.

The internal dorsal surface of the maxilla consists of three fossae. The anterior fossa lies anterior to the level of the infraorbital foramen and the middle and posterior fossae are separated by a remnant of the lachrymal. Most of the anterior fossa is missing, only the lateral wall above the foramen is intact. Along this wall a shallow, tapering channel runs posteriorly, and is pierced on its superior margin by a small foramen which opens posteromedially. The middle fossa is oval and narrows anteriorly. A sulcus runs along its lateral wall, the anterior course of which is obscured by a break in the wall, although it seems to continue anteriorly, descending into the anterior part of the fossa and/or connects with the shallow channel on the lateral wall of the anterior fossa. Posteriorly, the sulcus terminates just anterior to the lachrymal. The fragment of lachrymal that remains is approximately 2.5mm wide, 2 mm deep and 1mm in height. Its anterior surface is vertical while the posterior surface is slightly concave. The posterior fossa of the maxilla forms the anterior floor of the orbit and here is concave and broad. A small foramen is present in its lateral margin where the zygomatic wing commences. Medially, the maxilla is pierced by the posterior opening of the infraorbital canal. Posteriorly, the maxilla extends to the level of the damaged alveoli of M4, forming a flat medial shelf at the inferior margin of the orbit. Two transverse fractures run across this shelf above M3.

A small, triangular remnant of the jugal is evident on the superior margin of the zygomatic arch. The medial surface of this bone is smooth and concave, its lateral 214 surface, convex. The dorsal margin of the jugal is narrow and forms a tiny projection at its anterior end, below which lies the opening for a shallow sulcus that traverses its anteromedial surface. This sulcus runs parallel to the maxillojugal suture and terminates at the level of the lachrymal. This sulcus is probably a canal for the lachrymal gland. Medially, the jugal makes brief contact with the remains of the lachrymal. Posterior to this contact, the maxillojugal suture curves superiorly and laterally along the anteromedial surface of the zygomatic arch.

Description of referred specimen: A left M2 (Fig. 10.2), QM F42676, was recovered from the concentrate associated with the holotype material and is interpreted to be from the same individual. The tooth is rectangular in occlusal outline and measures 3.0mm long and 2.5mm wide. A large, shallow talonid basin dominates the tooth. The paraconid, protoconid, metaconid and hypoconid are subequal in height. A weak crest runs across the anterobuccal corner of the talonid basin creating a shallow anterobuccal basin which separates the paraconid from the rest of the talonid basin. The anterobuccal, anterolingual and posterobuccal corners of the tooth are rounded. The talonid basin rises more steeply to the hypoconid and the posterolingual corner of the crown below this cusp is more angular. At approximately the midpoint of the buccal and lingual sides of the tooth, the margins of the talonid basin are notched. The talonid basin lacks crenulations. M2 has two roots of equal size.

Comparisons

The general morphology of the dentition of Microleo (i.e. the blade-like premolar and triangular molars) is characteristic of thylacoleonids but a number of the features of its P3 are unique within the family. Its dental formula is similar to species of Lekaneleo, Wakaleo hilmeri and W. pitikantensis, retaining three premolars and four molars and is unlike most species of Wakaleo and Thylacoleo which exhibit molar loss. Microleo is also similar to Lekaneleo in the proportional size of P3 to M1. The shape of its upper molars are also most similar to those of Lekaneleo roskellyae and 215

Wakaleo hilmeri with M1 appearing to be relatively square while the succeeding molars are more triangular.

The most striking differences shown by Microleo are its small size and the morphology of P3. Microleo is the smallest species of marsupial lion. Its P3 (4.6mm) is slightly smaller than that of Lekaneleo deminutivus (5.7mm) and is approximately half the length of the P3 of Lekaneleo roskellyae (8.4mm average), one quarter the length of Wakaleo vanderleueri (17.9mm) and is one tenth the length of Thylacoleo carnifex (50.5mm).

The P3 of Microleo attenboroughi shows a general similarity to the morphology in species of Lekaneleo and Wakaleo, i.e. a blade-like tooth with two main cusps and a "W" - outline in profile. However, the P3 also exhibits some significant differences. In contrast to all other species of thylacoleonids that exhibit a lingually-bowed longitudinal blade, that of Microleo runs in a relatively straight line between the two major cusps. In addition, whereas in species of Lekaneleo the longitudinal blade is buccally deflected between the posterior cusp and the posterior margin of the tooth, in Microleo the posterior portion of the blade angles slightly lingually before connecting with the posterobuccal crest. Another difference exhibited by Microleo is the relatively steep angle at which the posterior longitudinal blade ascends to the posterior margin of the crown. In species of Lekaneleo and Wakaleo the posterior o o portion of the blade forms a relatively low angle (20 - 30 ) as it ascends to the rear margin of the tooth while in contrast, in Microleo the angle is much steeper o (approximately 60 ).

The development of the crests on the lingual and buccal sides of P3 also differs from those in other species of thylacoleonids. A strong anterolingual crest or cuspule is characteristic of all marsupial lion species. Lekaneleo roskellyae exhibits a strongly developed crest while all species of Wakaleo exhibit an anterolingual cuspule near the base of the crown, which in W. hilmeri may be connected via a crest to the anterior cusp. In W. vanderleueri and W. alcootaensis, the lingual wear facet of the longitudinal blade often approaches or appears to connect with this cuspule. In 216 species of Thylacoleo, the anterolingual crest is vertical and runs from the anterior cusp to the base of the crown. In Microleo, however, the anterolingual crest is rounded and relatively weak. In all other thylacoleonids, the lingual surface of the crown above the posterior cusp is broadly convex, but in Microleo a rounded, anteriorly-directed crest is present in this region. The lingual surface between the anterior and posterior cusps also differs from those of other thylacoleonids. In species of Lekaneleo and Wakaleo, a valley or depression that is variably developed lies directly posterior to the anterolingual crest and runs to the junction of the two roots. This valley is narrower and weakly developed in Microleo. In addition, unlike other thylacoleonids, Microleo exhibits a deep curving valley directly anterior to the posterolingual crest and the posterior cusp.

The rounded, relatively vertical anterobuccal and posterobuccal crests on Microleo’s P3 are more similar to those seen in L. roskellyae. These are unlike the buccal crests in species of Wakaleo which are more acute, shorter and less steeply inclined. The posterobuccal basin is also distinctly deeper than those of other thylacoleonids where there is a tendency for the basin to be very shallow or nearly absent.

In profile, the P3 of Microleo appears to differ from most thylacoleonids by having an anterior cusp that is lower than the posterior cusp. However, a large exposure of dentine on its anterior cusp suggests that the apex of this cusp has been broken off and/or has been subject to heavy wear. Thus, it is probable that its original height was at least similar to that of the posterior cusp, the condition seen in species of Wakaleo. The wear facet on any thylacoleonid P3 characteristically runs along the lingual edge of the longitudinal blade from the anterior cusp to the posterior cusp. In contrast, the regions of wear on the Microleo P3 lie on the posterior edges of the anterior and posterior cusps and are directed more posteriorly than lingually.

Discussion 217

The blade-like nature of the premolar and triangular shape of the molars in QM F42676 support its attribution to the Thylacoleonidae. Generic distinction of the specimen is based on the combination of its very small size and the morphological differences of its P3. Although its size difference to Lekaneleo deminutivus is not extreme (20% smaller), the morphology of its P3 is not consistent with those of species of Lekaneleo.

Initially, in attempting to identify the Neville's Garden specimen, the question of whether this small specimen simply represented a juvenile individual of a known marsupial lion (i.e. L. deminutivus or L. roskellyae) and its P3 represented a deciduous tooth, had to be addressed. The age of a marsupial can be determined by the stage of development and wear of its dentition (Kingsmill 1962, Lyne 1982). Juvenile marsupials can usually be identified by the molar position and degree of molar eruption, the molars erupting in sequence from anterior to posterior. The molars of juvenile marsupials are partially enclosed within alveolar crypts and embedded within the gum and show little or no wear on the surfaces of the crowns until fully erupted. On this basis, the Neville's Garden species represents an adult, for not only are its molars completely erupted, there are also obvious wear facets present on the crowns, especially on P3.

In all marsupial families a deciduous premolar lies anterior to the row of molars. The permanent tooth by which it is replaced is often noticeable by the bulging maxillary or dentary bone that lies either above or below the deciduous tooth. However, the Pleistocene thylacoleonid Thylacoleo carnifex seems to be the exception to the rule in this feature, for even in undisputed juvenile specimens no deciduous premolar has ever been found for this species (Finch 1971). Only the P3 as found in adult T. carnifex has been found in the juveniles. Although juvenile forms of other thylacoleonid species have yet to be discovered, it is possible that they exhibit a similar condition. In QM F41143 there is no evidence of a crypt containing a permanent tooth within the maxillary bone above the premolar and thus this tooth is most probably the permanent premolar. 218

The occlusal action of the P3s in species of Lekaneleo and Wakaleo is such that the lingual surface of the upper premolar slides down against the buccal surfaces of the lower premolar and the anterior half of M1 in a basic scissor-like action. Hence, the wear pattern on the upper premolar is normally along the lingual edge of the longitudinal blade. The uncharacteristic regions of wear on the Microleo P3 suggest that its lower premolar has quite a different morphology to the bladed P3 characteristic of thylacoleonids. If this is the case, it suggests that the slicing function normally associated with thylacoleonid premolars may not have been a predominant function of the premolar in this species. The comparatively rounded form of the anterobuccal and anterolingual crests and their lack of development into relatively sharp crests also suggest a reduced requirement for a slicing function. These morphological and hypothetical functional differences are suggestive of a diet that may have been more insectivorous or omnivorous rather than carnivorous. It is also possible, but less likely, that the wear patterns reflect a lower premolar that was malformed or an abnormal occlusal action. The phylogenetic relationships of Microleo attenboroughi to other thylacoleonids are explored in Chapter 11. 219

CHAPTER 11

INTRAFAMILY PHYLOGENETIC RELATIONSHIPS

11.1 Historical background

Previous explorations of the intrafamilial relationships of thylacoleonids have primarily involved characters of dental morphology. Clemens and Plane (1974) suggested that although species of Wakaleo exhibit many features that make them 1 likely ancestors of species of Thylacoleo, the absence of P /1 in the species of Wakaleo then known (and its presence in species of Thylacoleo) makes this hypothetical ancestor/descendant relationship improbable. For species of Thylacoleo to be directly descended from species of Wakaleo, P1 would have to have been regained in species of Thylacoleo. Such an event was regarded as improbable, although a similar phenomenon has been claimed in felids (Kurten 1963).

Archer and Dawson (1982) postulated that Wakaleo and Thylacoleo were sister- groups that comprised two temporally overlapping lineages (Fig. 11.1a). Synapomorphies identified by Archer and Dawson (1982) for Thylacoleo included 1 2 3 loss of M3, M /1 and M /2 lost or reduced in size relative to P /3 and a relatively wide I1. The primary synapomorphy of species of Wakaleo was reduction in premolar number. Within Thylacoleo, Archer and Dawson (1982) regarded T. hilli as the plesiomorphic sister-group of T. crassidentatus and T. carnifex. Within Wakaleo, they grouped W. vanderleueri and W. alcootaensis based on the presence of a large P3 and regarded W. oldfieldi as the sister-group to that pair. Priscileo was not included in the phylogeny of Archer and Dawson (1982) because it was unknown at the time. 220

Thylacoleo Thylacoleo Thylacoleo Wakaleo Wakaleo Wakaleo hilli crassidentatus carnifex oldfieldi vanderleueri alcootaensis

Fig. 11.1a Hypothesis of thylacoleonid relationships proposed by Archer and Dawson (1982).

Thylacoleoninae Wakaleoninae

Wakaleo Priscileo Thylacoleo

Fig. 11.1b Hypothesis of thylacoleonid relationships proposed by Murray et al. (1987).

Other Phalangeroids Priscileo Wakaleo Thylacoleo

Fig. 11.1c Hypothesis of thylacoleonid relationships proposed by Rauscher (1987). 221

Murray et al. (1987) placed species of Wakaleo and Thylacoleo in separate subfamilies (Fig. 11.1b): Wakaleoninae, which includes species of Wakaleo; and Thylacoleoninae, which includes species of Thylacoleo and possibly those of

Priscileo. Thylacoleonines differ from wakaleonines in: the presence of P1/1; frontal-squamosal contact on the lateral cranial wall; the squamosal tympanic wing; the course of the squamosoalisphenoid suture mesial to the tympanic wing; the presence of a complete postorbital bar; the deep incurvation anterior to P3; the reduction of upper canine; the functional shearing emphasis on P3/3; fusion of the basisphenoid and the basioccipital; weak or absent transverse palatal ridges; and the possibility that M4 in Priscileo is rotated clockwise about 20º with respect to the long axis of the tooth row. Features distinguishing wakaleonines from thylacoleonines include: absence of P1/1; lack of frontal/squamosal contact on the lateral cranial wall; the tribosphenic, conule bearing and three-rooted upper molars; squamosal/alisphenoid tympanic wing; larger alisphenoid wing; the more lateral position of the ventral squamosoalisphenoid suture; lack of fusion of the basisphenoid and basioccipital; shearing function distributed throughout the molar row; the course of premaxillary-maxillary suture from anterior of the canine alveolus to the midpoint of the anterior-palatal fenestrae; lack of a postorbital bar.

Rauscher (1987), in assessing the intrafamilial position of the monotypic Priscileo pitikantensis, found that there were no synapomorphies uniting that taxon and species of Wakaleo, or with species of Thylacoleo. In her analysis, Rauscher (1987) compared thylacoleonid postcrania and found that, for a number of postcranial character-states, species of Thylacoleo exhibited the presumed plesiomorphic condition while those of P. pitikantensis were derived. Rauscher (1987) concluded that the primitive features of Thylacoleo were secondarily derived and not a retained plesiomorphic condition. Because species of Wakaleo and Thylacoleo shared the synapomorphy of M4/4 loss, Rauscher considered them to be sister-groups and Priscileo to be the plesiomorphic sister-group to the Wakaleo/Thylacoleo clade (Fig. 11.1c). 222

Currently, although there appears to be agreement regarding the relationships of species within the thylacoleonid genera (i.e. within Wakaleo and Thylacoleo), there is disagreement regarding the interrelationships of the three genera, Priscileo, Wakaleo and Thylacoleo. In the past, analyses of intrafamilial relationships have relied heavily on dental characters but the new thylacoleonid cranial material recovered from the Riversleigh deposits enables a better and broader comparison of cranial characters and should contribute greatly to providing a clearer understanding of the relationships of these taxa.

Use of names The outcomes of the phylogenetic analyses presented in this chapter have been used to address the systematic problems arising from the revisions and comparisons of thylacoleonids taxa undertaken in Chapters 4 to 10. This has resulted in the referral of a number of previously described thylacoleonid taxa to new genera. To avoid confusion that may occur through of the use of the new generic names, thylacoleonid taxa are referred to by their specific epithets. A cladogram is presented at the end of the chapter that incorporates the nomenclatural changes supported by the results of these analyses.

11.2 Aims

The aim of this analysis is to test the previous hypotheses proposed regarding the interrelationships of marsupial lions as well as answer a number of questions arising from the discovery of the thylacoleonid material from Riversleigh. These include: (1) Are Priscileo, Wakaleo and Thylacoleo reciprocally monophyletic? (2) What are the affinities of the new forms recovered from Riversleigh? (3) Are pitikantensis (SAM P37719) and hilmeri sibling terminal taxa? Do these taxa fall within monophyletic clades or is pitikantensis more closely related to hilmeri than to roskellyae? 223

11.3 Methods

11.3.1 Ingroup and Outgroup taxa

Twelve thylacoleonid taxa comprised the ingroup. The new form sp. cf. hilmeri (QM F23446) was included as a taxon because of its possible transitional status. In an attempt to increase the data available for Thylacoleo hilli, data from the holotype was combined with that of T. sp. cf. T. hilli, following Archer and Dawson (1982). The extremely fragmentary nature of some thylacoleonid taxa, in particular attenboroughi, hilli, crassidentatus, alcootaensis, oldfieldi, pitikantensis (SAM P37719) and deminutivus (all of which lack significant cranial material), resulted in a large number of unknown character states within the data matrix: for example, for attenboroughi 30% of characters were complete and 70% were incomplete and for pitikantensis, 12.5% were complete and 87.5% were incomplete. The taxa employed as outgroups included the phascolarctids Litokoala kutjamarpensis (including specimens formerly identified as L. kanunkaensis, see Louys et al. in press) and Nimiokoala greystanesi, the phalangerids Caudipilosus vellicodens and Trichosurus vulpecula and the peramelid, Echymipera kalubu, hereafter referred to by their generic names (i.e. Litokoala, Nimiokoala, Caudipilosus, Trichosurus and Echymipera). In previous phylogenetic analyses based on either cranial or postcranial characters, the Phascolarctidae consistently appear as the basal vombatiform taxon and are the plesiomorphic sister-group of thylacoleonids (Marshall et al. 1989, Munson 1992, Louys 2004) hence their inclusion as an outgroup. Caudipilosus and Trichosurus were included because phalangerid cranial morphology is perhaps closest to the likely ancestral morphology of vombatiforms, based on a general similarity between taxa such as Wynyardia and Litokoala, and generalised phalangerids (eg. Trichosurus). Similarly, peramelids are a basal component of the Australian marsupial radiation (Retief et al. 1995, Sanchez- Villagra et al. 2002, Asher et al. 2004) and were included as an outgroup because of the relatively primitive nature of their dentitions. Specimens of all taxa examined for the character analysis are listed in Appendix 4. 224

11.3.2 Character Selection and Description

A total of 72 characters of the dentition and cranium were used in the intrafamilial analysis. Descriptions of these characters are listed below. The number of characters used by Archer and Dawson (1982) in the first cladistic study of the relationships of marsupial lions was restricted because of the fragmentary nature of the thylacoleonid specimens available at that time. The twelve dental characters used in their analysis were re-evaluated for this study. Eight of these characters were included here, some in a slightly modified version; for example, the length of P3 is compared to the length of the cheektooth row rather than to the length of M1. Murray et al. (1987, see Table 2) tabulated the character-states of a broad range of cranial and dental characters of Wakaleo and Thylacoleo but did not subject them to a parsimony-based analysis. These characters were also re-evaluated and some were included in this analysis, usually in a modified form.

Character description Characters used in ordered (Wagner) analyses are marked with an asterix (*). Characters used in previous analyses are identified and the character number used by the original authors is indicated (ch.xx). Characters and their coded states for each taxon are shown in Table 11.1.

Dentition 1. I2 and canine*: roskellyae, hilmeri, oldfieldi, vanderleueri, Caudipilosus and Trichosurus bear one alveolus between the procumbent I1 and the premolars. Whether this tooth represents I2 or a canine cannot be determined. In crassidentatus, hilli and carnifex, the two alveoli that lie between I1 and P3 are assumed to be those of the anterior premolars (Archer and Dawson 1982) hence, they are deemed to lack I2 or a canine. Nimiokoala lacks any teeth in this region and the condition for Litokoala is not known. Echymipera possesses both I2-3 and a canine. 0 = I2 and C present; 1 = I2 or C present; 2 = I2 and C absent. 225

2. Lower premolar number*: roskellyae possesses three lower premolars. hilmeri bears at least two premolars, possibly three if the anterior alveolus is considered to contain a premolar rather than a canine; oldfieldi and vanderleueri possess one premolar (the immediate post-incisor alveolus that is orientated horizontally is here regarded as being either an incisor or canine). crassidentatus and carnifex possess three premolars and the same is assumed for hilli (Archer and Dawson 1982). Lower premolar number is variable in Trichosurus; sometimes a small P2 may be present as well as P3. Nimiokoala and Litokoala possess a single premolar, P3. Three lower premolars are present in Caudipilosus vellicodens and Echymipera. (Archer and Dawson 1982, ch.1). 0 = three; 1 = two; 2 = one.

3. P3 buccally deflected: P3 is buccally deflected in Trichosurus and Caudipilosus. In all other taxa, the P3 is aligned longitudinally with the dentary. 0 = no buccal deflection; 1 = buccal deflection.

4. P3 serrated: P3 is serrated in Caudipilosus and Trichosurus. All remaining ingroup and outgroup taxa lack serrations on P3. 0 = no serrations; 1 = serrated.

5.P3 length: cheek tooth row (P3- M4)*: most thylacoleonids possess a relatively large P3 (i.e. it accounts for at least a quarter of the length of the whole cheektooth row) and show a trend of increasing size of this premolar in relation to the whole cheek tooth row (P3- M4). roskellyae has proportionately the shortest P3. For the remaining thylacoleonid taxa P3 progressively increases in size such that hilmeri < oldfieldi < vanderleueri, < crassidentatus, < carnifex. hilli is not included because the length of its full cheektooth row is not known. Phascolarctids, phalangerids and Echymipera all possess a proportionately short P3, such that it accounts for less than a quarter of the cheek tooth row. 0 = very short (<.25); 1 = short (.25-.3); 2 = medium (.3-.5); 3 = long (>.6).

6. Upper premolar number*: attenboroughi, roskellyae and pitikantensis possess three upper premolars. Most specimens of hilmeri also exhibit the maximum 226 premolar complement (three), although loss of at least P1 has also been observed in some specimens; sp cf. hilmeri possesses three upper premolars. vanderleueri exhibits one or two premolars; and alcootaensis has reduced its number to a single tooth. carnifex and crassidentatus both have three premolars. Although hilli is known to possess P3, presence or absence of more anterior premolars is unknown. Although phascolarctids and phalangerids also display a reduced premolar number, Echymipera has three upper premolars. (Archer and Dawson 1982, ch.1): 0 = three; 1 = two; 2 = one.

7. P3 posterolingual crest: A well developed posterolingual crest descends from the posterior end of the longitudinal blade in roskellyae, hilmeri, sp.cf. hilmeri, oldfieldi and vanderleueri. This crest is also present in sp. cf. T. hilli but only weakly developed in crassidentatus, and is absent in carnifex. The P3 of Nimiokoala appears to lack this crest, while that of Litokoala exhibits a small crest near the base of the crown. Within the phalangerids, this structure is absent in Caudipilosus and present in Trichosurus. 0 = absent; 1 = weakly developed; 2 = well developed.

8. Anterior crest of P3 : Archer and Dawson (1982) suggested that the bifurcated anterior crest of hilli was the apomorphic state and the simple crest present in all other thylacoleonids was the plesiomorphic condition. roskellyae exhibits both conditions (see Chapter 8). All outgroup taxa possess a simple anterior crest. (Archer and Dawson 1982, ch.8). 0 = simple; 1 = bifurcated.

9. Anterobuccal crest of P3: this crest is short and weakly developed in roskellyae, hilmeri, oldfieldi and vanderleueri and appears to be absent or very poorly developed in hilli, crassidentatus and carnifex. This crest is also poorly developed in Nimiokoala but is more strongly developed in Litokoala and in the phalangerids Caudipilosus and Trichosurus. 0 = absent; 1 = weak; 2 = moderately developed.

10. Posterobuccal crest of P3: Archer and Dawson (1982) identified this small, poorly-developed crest on the P3 of sp. cf. T. hilli and interpreted it to be an 227 apomorphic state. A small vertical posterobuccal crest also appears to be present on some specimens of hilmeri although on other specimens it appears to be absent. oldfieldi and vanderleueri appear to lack this crest. A small posterobuccal crest is also present in roskellyae although it lies slightly more anteriorly than the crest in sp. cf. T. hilli. This crest is absent in all outgroup taxa. (Archer and Dawson 1982, ch.7). 0 = absent; 1 = weak crest immediately anterior to end of tooth; 2 = at posterior end of tooth.

11. Molar morphology: The molars of Echymipera have a basic tribosphenic morphology. The molars of Litokoala and Nimiokoala are selenodont. The molars of Trichosurus and Caudipilosus are bunolophodont. Molars of thylacoleonid species are bunodont. 0 = tribosphenic; 1 = selenodont; 2 = bunolophodont; 3 = bunodont.

12. M1 trigonid height and talonid height: The trigonid is relatively tall in comparison to the talonid in roskellyae, hilmeri, oldfieldi and vanderleueri. The trigonid of crassidentatus and carnifex exhibits an even greater difference in height. Within the outgroup taxa, the trigonid of Echymipera and the phalangerids (Caudipilosus and Trichosurus) is moderately taller than the talonid. In phascolarctids the trigonid is relatively low. 0 = low; 1 = tall; 2 = very tall.

13. M1 talonid basin width between entoconid and hypoconid* : the talonid basin of roskellyae is very broad and only slightly narrower than the width of the crown. The talonid basin of hilmeri is broader than that of oldfieldi, which in turn is slightly broader than vanderleueri. The talonid basin shows further reduction in crassidentatus and is practically lost in carnifex. All the outgroup taxa display a broadly separated hypoconid and entoconid. 0 = broad; 1 = slightly narrowed; 2 = narrow; 3 = loss.

14. M1 posterior width/ anterior width ratio* : in roskellyae the talonid of M1 is broader than the trigonid. In hilmeri, oldfieldi, vanderleueri, crassidentatus and carnifex, the trigonid of M1 is broader than the talonid. Narrowing of the talonid is 228 particularly prominent in the latter two taxa. In all outgroup taxa the talonid of M1 is broader than the trigonid. 0 = broader, 1 = narrower; 2 = much narrower.

15. M2 length/ P3 length ratio* : The M2 of roskellyae is slightly shorter than its P3. hilmeri, oldfieldi and vanderleueri exhibit a proportionately shorter M2 that progressively shortens through each of these taxa, respectively. This trend of decreasing M2 size is continued in crassidentatus and carnifex. All outgroup taxa possess an M2 that is similar to or longer than their P3. (Archer and Dawson,ch.3). 0 = > 1.0, large; 1 = 1.0, similar; 2 = 0.8 > 1, slightly smaller; 3 = 0.7; 4 = 0.6 ; 5 = 0.5; 6 = 0.1, extremely small.

16. M2 trigonid height relative to talonid* : the trigonid is very similar in height to the talonid on the M2 of attenboroughi and roskellyae. hilmeri, oldfieldi and vanderleueri have a trigonid that is much taller than the talonid basin. The M2 talonid of crassidentatus and carnifex is reduced to loss. In Echymipera, the selenodont phascolarctids and bunodont phalangerids the major cusps of the trigonid (protoconid and metaconid) are slightly taller than the major cusps of the talonid. 0 = similar; 1 = slightly taller; 2 = taller.

17. M2 talonid basin width relative to crown width* : the M2 of roskellyae displays a talonid basin that is very broad and nearly as wide as the crown. hilmeri, oldfieldi and vanderleueri possess a slightly narrower talonid basin, while that of crassidentatus and carnifex is reduced to loss. All outgroup taxa exhibit a relatively broad talonid basin. 0 = broad; 1 = slightly narrowed; 2 = narrowed; 3 = loss.

18. M3 presence and occlusal outline* :M3 is present in roskellyae, hilmeri, oldfieldi and vanderleueri. This molar is lost in carnifex and crassidentatus. Evidence for the presence or absence of this molar in attenboroughi, pitikantensis, alcootaensis and hilli is not available. The outline of M3 in hilmeri is rectangular, that of oldfieldi is subrectangular and that of vanderleueri is oval. The outline of M3 of roskellyae is assumed to be rectangular on the basis of the rectangular shape of its 229 alveolus and the highly rectangular shape of the preceding molar. All outgroup taxa possess M3 that is rectangular in outline. (Archer and Dawson 1982, ch.2). 0 = present and rectangular; 1 = present and subrectangular/oval; 2 = present and oval; 3 = loss.

19. M3 talonid basin: M3 talonid basin of hilmeri and oldfieldi is a crenulated talonid. The M3 of vanderleueri displays a loss of this basin. The status of this character in attenboroughi, pitikantensis, roskellyae, alcootaensis and hilli is unknown. This character is inapplicable in carnifex and crassidentatus because M3 is lost. All outgroup taxa exhibit a talonid basin on the M3. 0 = present; 1 = loss.

20. M4 presence/absence: attenboroughi, roskellyae, pitikantensis and hilmeri retain M4 while sp. cf. hilmeri, oldfieldi, vanderleueri, crassidentatus and carnifex demonstrate loss of this molar. Status of this character is unknown for alcootaensis and hilli. All outgroup taxa possess M4. 0 = present; 1 = loss.

21. P2 position relative to P3* : in attenboroughi P2 lies directly anterior to P3. In pitikantensis, roskellyae, hilmeri, sp.cf. hilmeri, vanderleueri and crassidentatus, P2 lies anterolingually to P3. In carnifex P2 lies directly lingual to P3. The only outgroup taxon that retains P2 is Echymipera and in that species this tooth lies anterior to P3. 0 = anterior; 1 = anterolingual; 2 = predominantly lingual.

22. P3shearing facet: the shearing facet of the longitudinal crest of the P3of attenboroughi has a predominantly posterolingual orientation; the facet on the posterior part of the crest is directed posteriorly. The major shearing facets in all other thylacoleonids (except pitikantensis which is unknown) are all lingually directed. The major shearing facet in Echymipera is directed posteriorly while those of Caudipilosus and Trichosurus are directed anterolingually. The major areas of wear on the cusps of Nimiokoala are directed lingually. 0 = posterior; 1 = anterolingual; 2 = lingual. 230

23. P3 crown length relative to cheektooth row (P3- M4)* : P3 is variably developed within thylacoleonids. In attenboroughi and roskellyae it is relatively small in comparison to the cheektooth row. This premolar shows a trend of enlargement relative to the molar row through hilmeri, sp.cf. hilmeri, vanderleueri, crassidentatus and carnifex. In all the outgroup taxa the third premolar is comparatively small relative to the cheektooth row. 0 = very short (< 0.3); 1 = short (0.3-0.4); 2 = long (0.4-0.5); 3 = very long (> 0.5).

24. P3 longitudinal blade length: P3 crown length* : the length of the longitudinal blade between the anterior and posterior cusps varies within thylacoleonids. In attenboroughi and roskellyae the length of this blade is relatively short. Blade length is slightly longer in deminutivus, is moderately long in hilmeri and sp. cf. hilmeri, long in oldfieldi, vanderleueri and alcootaensis and is very long in crassidentatus and carnifex. The length of the longitudinal blade between the major cusps in the outgroup phascolarctids and phalangerids is short. 0 = short (< 0.4); 1 = medium (0.4-0.5); 2 = long (0.5-0.7); 3 = very long (> 0.7).

25. P3 longitudinal blade curvature: All species of thylacoleonid except attenboroughi have a longitudinal blade that is lingually convex. The longitudinal blade of attenboroughi, like those of the outgroup taxa, is straight. 0 = straight; 1 = lingually convex.

26. Presence and inclination of a posterior longitudinal blade on P3: a posterior longitudinal blade is present in attenboroughi and is steeply inclined. In deminutivus this blade changes its inclination at approximately its midpoint, being moderately steep in the anterior half and more gently angled in the posterior half. In roskellyae, hilmeri, oldfieldi, vanderleueri, and alcootaensis this blade tends to be bowed but roughly horizontal in its inclination. Both sp. cf. hilmeri and hilli possess a moderately-angled posterior longitudinal blade. A posterior longitudinal blade is absent in carnifex and crassidentatus. In phascolarctids and phalangerids, this blade is steeply inclined. 0 = steep; 1 = moderately angled; 2 = horizontally bowed; 3 = gently sloping; 4 = absent. 231

27. P3 mid-crown constriction: the P3 of roskellyae, hilmeri , sp.cf. hilmeri, oldfieldi, vanderleueri and alcootaensis exhibit a constriction of the crown approximately midcrown, at the junction of the anterior and posterior roots. This constriction becomes progressively weaker on the lingual side in oldfieldi, vanderleueri and alcootaensis. attenboroughi, deminutivus, crassidentatus and carnifex lack this constriction. hilli exhibits only a weak constriction lingually. All outgroup taxa lack any mid-crown constriction of P3. 0 = absent; 1 = moderate, lingual = buccal; 2 = reduced lingual constriction; 3 = absent buccal and weak lingual constriction.

28. P3 posterobuccal crest: on the posterobuccal surface of the P3 of attenboroughi a short but well developed crest curves anteriorly from the posterior end of the posterior longitudinal blade toward the base of the crown. A slightly less prominent crest is present in roskellyae and a very small, short crest is present in deminutivus. A relatively short, steeper crest is present in hilmeri, oldfieldi and vanderleueri. alcootaensis appears to lack this crest, however, the posterobuccal surface of its crown appears to have suffered damage. This crest is absent in hilli, crassidentatus, carnifex and all outgroup taxa. 0 = absent; 1 = well developed; 2 = moderately developed, short; 3 = moderately developed, oblique, long; 4 = moderately developed; steep.

29. P3 anterolingual crest/cuspule: hilmeri, sp. cf. hilmeri, oldfieldi, vanderleueri, and alcootaensis bear a moderate to large cuspule below the anterior cusp on the lingual side of the crown. hilli , crassidentatus, carnifex, deminutivus and roskellyae bear a long vertical crest in this region while attenboroughi lacks both a crest and a cuspule. This cuspule/crest is absent in the phascolarctids and Caudipilosus. Trichosurus exhibits a weak crest. 0 = absent; 1 = small crest; 2 = moderate cusp; 3 = large cusp.

30. P3 posterior width relative to anterior width: In hilli, crassidentatus and carnifex, P3 is narrower posteriorly than anteriorly (measured at the base of the crown). All remaining thylacoleonid taxa have a P3 that is broader posteriorly than 232 anteriorly. All outgroup taxa possess a P3 that is broader posteriorly. 0 = broader; 1 = slightly narrower.

31. M1 occlusal outline* : thylacoleonids display variation in the shape of the first upper molar; that of attenboroughi is relatively square while that of roskellyae is slightly subsquare. The M1 of hilmeri is slightly narrower posteriorly and is partway between being square and triangular. sp. cf. hilmeri, oldfieldi, vanderleueri and alcootaensis display a triangular M1. Archer and Rich (1982) suggest that thylacoleonids had a quadritubercular ancestor and that the subrectangular shape of M1 of crassidentatus and carnifex is probably the result of the development of a new cusp anterobuccal to the protocone. All outgroup taxa possess a square or rectangular nature of M1. 0 = square; 1 = subsquare/subtriangular; 2 = triangular.

32. M1 metaconule* : the degree of development of the metaconule on M1 varies within thylacoleonids. roskellyae bears a metaconule that is moderately developed. The metaconule is moderately weak in hilmeri and sp. cf. hilmeri, and is very weak (or possibly absent) in oldfieldi and vanderleueri. Archer and Rich (1982) postulate that carnifex and crassidentatus have lost the metaconule on M1. All outgroup taxa possess a metaconule. 0 = well developed; 1 = moderate to weak development; 2 = very weak development; 3 = loss.

33. M1 posterobuccal margin: hilmeri, sp. cf. hilmeri, oldfieldi and vanderleueri exhibit a posterior lengthening of the buccal margin of M1 that results in that molar closely abutting and impinging against the anterior margin of M2. The M1 of roskellyae has a rounded posterobuccal margin that does not lengthen posteriorly against M2. Within the outgroup taxa the M1 of Echymipera extends posteriorly; however, the angle of this tooth is such that the posterobuccal corner protrudes laterally and its posterior margin does not push up against the anterior surface of M2. None of the outgroup phascolarctids or phalangerids exhibit posterior lengthening of 233 the buccal margin. 0 = no lengthening or overlap; 1 = lengthening and overlap; 2 = lengthening and abutment against M2.

34. M1anteroposterior buccal gradient* : On the M1of roskellyae the paracone is slightly taller than the metacone which results in a very gentle anteroposterior buccal gradient. The paracone in hilmeri and sp. cf. hilmeri is relatively taller than in roskellyae, and hence, the buccal gradient is steeper. In oldfieldi and vanderleueri the paracone is much taller and the buccal gradient is relatively steep. In all outgroup taxa the paracone is lower than the metacone; marginally so in the phascolarctids and phalangerids, but distinctly so in Echymipera. 0 = paracone much shorter than metacone; 1 = paracone slightly shorter than metacone; 2 = paracone slightly taller than metacone; 3 = paracone moderately taller than metacone; 4 = paracone much taller than metacone.

35. M2 occlusal outline* : attenboroughi and roskellyae have an M2 that has a subsquare outline. M2 becomes progressively more triangular in shape through hilmeri, sp. cf. hilmeri, oldfieldi, vanderleueri and alcootaensis. All outgroup taxa have an M2 that is square or subsquare in outline. 0 = square; 1 = subsquare; 2 = subtriangular; 3 = triangular.

36. M2 buccal height relative to lingual height* : the buccal margin of M2 is variably developed in thylacoleonids. In attenboroughi the buccal and lingual margins are similar in height. In roskellyae and sp. cf. hilmeri the buccal margin is slightly taller. Buccal height progressively increases through hilmeri, oldfieldi to vanderleueri. The M2 in all outgroup taxa exhibit a slightly taller buccal margin relative to the lingual margin. 0 = similar; 1 = slightly taller; 2 = taller; 3 = much taller.

37. M2 metaconule development: the metaconule on M2 shows variable development within thylacoleonids; it is well developed in attenboroughi, moderately developed in roskellyae, very weak in hilmeri, sp. cf. hilmeri and pitikantensis, and appears to 234 be extremely reduced or lost in oldfieldi and vanderleueri. A well developed metaconule is present in all outgroup taxa. 0 = well developed; 1 = small; 2 = weak; 3 = reduced to loss.

38. Buccal margin of paracone of M2: The width of the buccal margin of the paracone of M2 of attenboroughi is similar to the width at the level of the metacone. The buccal margin of the crown above the paracone in roskellyae, sp. cf. hilmeri, hilmeri, oldfieldi and vanderleueri shows a distinct lateral bulge. No thickening of the buccal margin is evident in Echymipera, Nimiokoala or Litokoala and both Caudipilosus and Trichosurus show a very slight thickening of the buccal margin. 0 = absent; 1 = present.

39. M3 presence/absence: With the exception of crassidentatus and carnifex, all thylacoleonids possess M3. This character is unknown for deminutivus, alcootaensis and hilli. This molar is present in all outgroup taxa. 0 = present; 1 = lost.

Cranium and basicranium

40. skull shape* : the cranium of roskellyae is moderately broad: its length is 1.5 times its breadth. The crania of vanderleueri and hilmeri are relatively broader. The cranium of carnifex is relatively brachycephalic. Echymipera has a narrow skull and phascolarctids and phalangerids have moderately broad skulls. 0 = narrow, 1 = moderately broad; 2 = broad; 3 = very broad.

41. frontal sinus inflation: the dorsal surface of the frontal bones that lies between the supraorbital processes (region of the frontal sinus) in roskellyae and hilmeri form a slight depression. This region in vanderleueri is strongly inflated but in carnifex it lacks significant depression or inflation. The condition of the frontal sinus is not known for Nimiokoala and Litokoala. Both Caudipilosus and Trichosurus exhibit dorsal depression of the frontals. The frontal sinus lacks inflation or depression in Echymipera. 0 = lacks depression or inflation; 1 = slight depression; 2 = inflation. 235

42. supraorbital process* : roskellyae exhibits small, thin, triangular supraorbital processes. These processes are larger and thicker in hilmeri and vanderleueri. In carnifex the supraorbital process is developed into a postorbital bar. A very small supraorbital process is present in Litokoala but is absent in all remaining outgroup taxa. 0 = absent; 1 = small; 2 = moderate, thin; 3 = large, thick; 4 = developed into postorbital bar.

43. supraorbital crests: Priscileo roskellyae develops small rounded supraorbital crests posterior to the supraorbital processes. These crests are also present in hilmeri but are taller and thicker. These crests are absent in both vanderleueri and carnifex. Within the outgroup taxa supraorbital crests are absent (Echymipera) or small (Caudipilosus and Trichosurus). The presence of crests is indeterminate for Nimiokoala and Litokoala. 0 = absent; 1 = narrow, well developed; 2 = thickened.

44. parietal width: relative to the squamosal, the posterior width of the parietal on the braincase of roskellyae, hilmeri and vanderleueri is broad, while that of carnifex is narrow. The parietal is also relatively narrow in Trichosurus and Caudipilosus but is broad in the phascolarctids and very broad in Echymipera. 0 = broad; 1 = narrow.

45. frontal/squamosal contact: In roskellyae, hilmeri and vanderleueri the frontal fails to contact the squamosal posteriorly. In carnifex, the frontal contacts the parietal and squamosal posteriorly. In Echymipera, the squamosal contacts the frontal; however, in all outgroup diprotodontian taxa frontal/squamosal contact is absent. 0 = contact absent; 1 = contact present.

46. sagittal crest: the sagittal crest is variably developed in thylacoleonids. In roskellyae it is short, in hilmeri and vanderleueri it is tall and thick, and in carnifex it is tall. Sagittal crest development is also variable in phalangerids; it may be absent or when present is small or rudimentary. A small sagittal crest is present in phascolarctids. Echymipera lacks a sagittal crest. When present in the latter species, it is small. 0 = absent or rudimentary; 1 = small; 2 = large. 236

47. nuchal crest: roskellyae possesses a short nuchal crest which slightly overhangs the occiput. In hilmeri, vanderleueri and carnifex this crest is taller and overhangs more prominently. Echymipera exhibits a moderately developed bi-lobed nuchal crest. Within the diprotodontoid outgroup taxa a nuchal crest is absent in the phalangerids and only moderately developed in Litokoala. This region has not been preserved in Nimiokoala. 0 = absent; 1 = weak; 2 = small; 3 = large.

48. lachrymal fossa: in roskellyae and sp. cf. hilmeri, the inferior lachrymal foramen lies in a small fossa. In both hilmeri and vanderleueri this fossa is enlarged. A lachrymal fossa is absent in carnifex. Within the outgroup Caudipilosus and Trichosurus exhibit a small, shallow lachrymal fossa similar to that observed in P. roskellyae. A lachrymal fossa is absent in Echymipera and Nimiokoala and is not preserved for Litokoala. A lachrymal fossa is absent in phascolarctids and peramelids. 0 = absent; 1 = small; 2 = moderate to large.

49. pneumatisation of the squamosal: the squamosal in thylacoleonids, phascolarctids and Echymipera lacks pneumatisation. Pneumatisation is present in Caudipilosus and Trichosurus. 0 = absent; 1 = present.

50. mastoid process: the mastoid is posteroventrally-developed into a transverse, anteroposteriorly-narrowed, medially-enlarged process in roskellyae. A distinct rugose, posterolaterally-positioned mastoid process is also present in hilmeri and vanderleueri but is broader, anteroposteriorly-deeper and more rounded. In Echymipera the ventrolateral surface of the mastoid bears a tiny, peg-like process. In Litokoala, the mastoid terminates ventrally as a very narrow transverse ridge (not preserved in Nimiokoala). In Caudipilosus the ventral surface of the mastoid is broad and rounded and a very small obtuse process lies posterolaterally. A similarly small process may be present in Trichosurus. 0 = broad, rounded; 1 = transverse, thin; 2 = transverse, thickened.

51. paroccipital process: the paroccipital process in hilmeri is short, thick, and directed posteriorly. Although complete processes are not preserved for roskellyae, 237 vanderleueri and carnifex, the basal remnants of their processes indicate they were also short, thick structures that appeared to be posteroventrally directed. The paroccipital process in Litokoala, Caudipilosus, and Trichosurus is narrow, long, and ventrally directed. This process in Echymipera is narrow, very short and also ventrally-directed. 0 = short, narrow, ventrally directed; 1 = moderately long, narrow, ventrally directed; 2 = short, thick posteroventrally directed.

52. alisphenoid ventrolateral wing* : A broad alisphenoid ventrolateral wing is present in roskellyae, hilmeri, Litokoala, Trichosurus and Echymipera. vanderleueri and Caudipilosus exhibit a slightly narrower wing. This wing is relatively narrow in carnifex. 0 = broad; 1 = slightly narrowed; 2 = narrow.

53. medial glenoid process: In roskellyae, at the medial end of the glenoid fossa the squamosal forms a small, obliquely-angled ridge or medial glenoid process. This process is taller and more prominent in hilmeri but is not as strongly developed in vanderleueri. This process is very strongly developed in carnifex. A medial glenoid process is not present in Caudipilosus, Trichosurus, Litokoala (the condition for Nimiokoala is not known) and Echymipera. 0 = absent; 1 = present.

54. postglenoid process: the postglenoid process of roskellyae is broad, anteroposteriorly narrow and relatively short. This process in hilmeri and vanderleueri is similar in shape but is thicker and more robust. In carnifex the postglenoid process is short and recurved. This process in Litokoala is broad and thick (this process is not preserved in Nimiokoala). The postglenoid process in Caudipilosus and Trichosurus is broad and thin and fused to the ectotympanic. In Echymipera the postglenoid process is small being very narrow and short. 0 = short , narrow; 1 = short, broad, fused to ectotympanic; 2 = short, broad, thin; 3 = moderately long, broad, thickened; 4 = broad, recurved.

55. postglenoid cavity: the postglenoid cavity in roskellyae is broad and expands laterally into the posterior surface of the postglenoid process (pgp). The postglenoid cavity of hilmeri, vanderleueri and carnifex lies more medially and does not expand 238 laterally into the postglenoid process. The postglenoid cavity in Echymipera is relatively narrow, opens ventrally rather than posteriorly, and is not laterally expanded. The postglenoid cavity of Litokoala is large and laterally expanded into the pgp (the condition in Nimiokoala is not known). Both phalangerids lack a postglenoid cavity. 0 = narrow, opens ventrally; 1 = opens posteriorly, laterally expanded across pgp; 2 = opens posteriorly, medial end of pgp; 3 = absent.

56. position of the postglenoid foramen: the postglenoid foramina of roskellyae are small and located in the medial end of the postglenoid cavity. The foramina have a similar location in hilmeri, however, the fossa in which the foramen lies is relatively enlarged. In vanderleueri, the foramina are much enlarged and have migrated posterolaterally so as to dominate the postglenoid cavity. In carnifex the postglenoid foramen opens into a highly pneumatised postglenoid cavity. The postglenoid foramen of Litokoala is located medial to the pgp. The condition for Nimiokoala is not known. The postglenoid foramen is located anteromedial to the postglenoid process in Caudipilosus and Trichosurus. In Echymipera, this foramen is located on the posteromedial margin of the postglenoid process. 0 = posteromedial to pgp; 1= medial to pgp; 2 = anteromedial to pgp; 3 = medial end of postglenoid cavity; 4 = enlarged in postglenoid cavity.

57. tympanic wing components: Both the alisphenoid and the squamosal contribute to the tympanic wing (floor of the tympanic cavity) of roskellyae, hilmeri, vanderleueri and Litokoala. The tympanic wing of carnifex is formed only from the squamosal. The tympanic wing of Echymipera and both phalangerids is constructed from the alisphenoid. 0 = alisphenoid; 1 = alisphenoid & squamosal; 2 = squamosal.

58. tympanic wing construction: in roskellyae, hilmeri, vanderleueri and carnifex the tympanic wing is short and open posteriorly. In Litokoala (unknown for Nimiokoala) the tympanic wing extends posteriorly, nearly reaching the paroccipital process, leaving a very small posterior opening. In Caudipilosus and Trichosurus, the alisphenoid tympanic wing extends posteriorly and connects to the paroccipital process and the tympanic cavity is therefore, posteriorly enclosed. The alisphenoid 239 tympanic wing is moderately developed in Echymipera and is open posterolaterally. 0 = large posterior opening; 1 = closed or nearly closed posteriorly.

59. fusion of the ectotympanic to other skull bones: the ectotympanic is not fused to the surrounding bones of the skull in roskellyae, hilmeri, vanderleueri and carnifex. This is also the condition in Echymipera and Litokoala. The ectotympanic is fused to the mastoid and squamosal in Caudipilosus and Trichosurus. 0 = unfused; 1 = fused.

60. epitympanic wing of the petrosal: a relatively long, broad, and concave epitympanic wing extends from the lateral margin of the periotic of roskellyae. A similar structure is also associated with the periotic of hilmeri; however, it is relatively shorter, narrower and shallower. vanderleueri appears to possess a similarly narrow wing. In carnifex, medial expansion of the squamosal epitympanic sinus prevents development of this wing. An epitympanic wing is absent in Litokoala (and not known for Nimiokoala). A relatively narrow epitympanic wing is present in Echymipera. 0 = narrow, shallow; 1 = broad, deep; 2 = absent.

61. rostral tympanic process: the periotic of roskellyae lacks a rostral tympanic process. The periotic of both hilmeri and vanderleueri bear well developed rostral tympanic processes. The periotic of carnifex lacks this process. A rostral tympanic process is lacking in Echymipera. The periotic of Nimiokoala has not been preserved and within Litokoala the rostral tympanic process is well developed. In Caudipilosus this process is weak and in Trichosurus it is moderately developed. 0 = absent; 1 = weakly developed; 2 = moderately developed; 3 = well-developed.

62. development of the post-tympanic process of the squamosal: in roskellyae this process is relatively short such that the mastoid process forms the inferior posterior wall of the outer ear canal. In hilmeri, vanderleueri and carnifex the squamosal extends ventrally to completely form the posterior wall of the outer ear canal. In Echymipera the process is short but forms most of posterior wall of the outer ear canal. The mastoid process of Echymipera lies directly behind this structure and any ventral contribution to the wall is extremely small as a result of the small size of the 240 mastoid. In Litokoala the posterior wall of the outer ear canal has a superior squamosal component and inferior mastoid component (the condition for Nimiokoala is not known). In Caudipilosus and Trichosurus the mastoid also expands anteriorly, ventral to the squamosal, to form the posteroventral wall of the outer ear canal. 0 = short; 1 = long.

63. course of the stylomastoid sulcus: the stylomastoid sulcus courses laterally through the squamosal on the posterior wall of the outer ear canal in hilmeri, vanderleueri and carnifex. In roskellyae this sulcus lies in the mastoid. The posterior surface of the outer ear canal of Echymipera lacks a stylomastoid sulcus. The stylomastoid sulcus courses through the mastoid in Caudipilosus, Trichosurus and Litokoala (the condition in Nimiokoala is not known). 0 = through mastoid; 1 = through squamosal.

64. subsquamosal foramen: the subsquamosal foramen in roskellyae is large, relatively horizontal and located immediately posterior to the zygomatic limb of the squamosal. In hilmeri and vanderleueri this foramen is more obliquely oriented and has a slightly more posterodorsal location. In carnifex this foramen is very large and located posterodorsally to the zygomatic limb. The subsquamosal foramen is also located posteriorly to the zygomatic limb in Echymipera and Litokoala but is relatively small in the former and broader and more horizontal in the latter. Caudipilosus and Trichosurus lack this foramen. 0 = posterior to zygomatic limb; 1 = posterodorsal to zygomatic limb; 2 = posterodorsal to zygomatic limb, enlarged; 3 = absent.

65. alisphenopalatine lateral processes: Moderately large, laterally-projecting processes are present in the region of the palatine-alisphenoid suture in hilmeri, vanderleueri and carnifex. These processes are absent in roskellyae. This process is present in Litokoala but this region is not preserved in Nimiokoala. This process is absent in the phalangerids and Echymipera. 0 = absent; 1 = present. 241

66. infraorbital crest and fossa: the infraorbital crest in hilmeri and vanderleueri is strongly developed and is bordered ventrally by a deep concave fossa. This crest in roskellyae is weakly developed and the surface below it lacks a deep fossa. The infraorbital crest and fossa is moderately developed in carnifex. Both Nimiokoala and Litokoala possess a small infraorbital crest but lack a distinct fossa below the crest. The infraorbital crest is very weak in Echymipera and is moderately weak in Trichosurus and Caudipilosus and all three taxa lack a fossa. In the outgroup taxa, the crest is relatively weak and the fossa is absent. 0 = very weak crest, no fossa; 1 = small crest, no fossa; 2 = well developed crest and fossa.

67. anterior palatal fenestra: the palatal fenestrae of roskellyae are long and relatively wide. Those of hilmeri and vanderleueri are long and narrow, while those of carnifex are relatively short and wide. The anterior palatal fenestrae of all outgroup taxa are long and narrow. 0 = long, narrow; 1 = short, wide.

68. posterior palatal fenestra: the posterior palatal fenestra of roskellyae, hilmeri and vanderleueri is large and coalesced across the midline, extending to the level of M1 in the former and to M1/M2 in the two latter species. These fenestrae in carnifex and crassidentatus are massively developed and extend over the posterior half of the palate. The posterior palatal fenestrae are small, not coalesced, and relatively anteriorly located in Echymipera. In Litokoala these fenestrae are small and posteriorly located within the palate. In Caudipilosus and Trichosurus the fenestrae are large, coalesced and extend anteriorly to the level of M1. 0 = anterior and small; 1 = posterior, small; 2 = large, to M1; 3 = large to M1-2; 4 = extremely large.

69. lateral margins of the posterior palatal fenestrae: the palatine forms a gently rounded lateral margin of the posterior palatal fenestrae in roskellyae and carnifex. In hilmeri and vanderleueri the lateral margins of the fenestrae are sharp and distinctly protrude into the fenestra. All outgroup taxa (except Nimiokoala for which it is unknown) exhibit sharply protruding margins of their posterior palatal fenestra. 0 = sharply projecting; 1 = rounded 242

Table 11.1: Taxa character state matrix for intrafamilial parsimony analysis

1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031323334353637 Echymipera 00000000- - 010011000000000000000010010 Litokoala ?2000?00201000?10000??000000000001010 Nimiokoala 22000200101000010000- 100000?000001010 Caudipilosus 10110200202100110000- 1000000000001010 Trichosurus 11/2110110202100110000- 1000000000001010 attenboroughi ??00?0????3????0???000000001000???100 deminutivus ???0?????????????????2?1110210??????? roskellyae 10001020/11131002000?012101213100102111 pitikantensis ?????0????3????????01?1????????????22 sp.cf.hilmeri ??002?20103????????11221131?302123212 hilmeri 110010/12010/1311132100012111214201123122 oldfieldi 12002?20103121422101?2?2121430222422/33 vanderleueri 120021/22010312152221112221224302224333 alcootaensis ??00?2????3?????????- 222122?302???3?? hilli 20?0??2102???????????2??133011??????? crassidentatus 2000301012323262331112331400111304??? carnifex 200030000/12323262331122331400110304- - 3

38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 Echymipera 00000000020000000000000010000000000 Litokoala 001???1?121001003121103311111?10011 Nimiokoala 001???111?00??0????1????????10??0?1 Caudipilosus 001101210/10110110121011?121401020020 Trichosurus 0011012110/1110110131000?221400020421 attenboroughi 00????????????????????????????????? deminutivus ??????????????????????????????????? roskellyae 1011111112101?012131001101101121010 pitikantensis ?0??????????????????????????????1?? sp.cf.hilmeri 10????????1?????????????????????2?1 hilmeri 10212211232012013231000312212030100 oldfieldi 10????????????????????????????????? vanderleueri 10223011222012113241000312212020201 alcootaensis ??????????????????????????????????? hilli ??????????????????????????????????? crassidentatus ?1???????????????????????????24???0 carnifex - 1304020230012214232002012312231300 (? = character unknown: - = character not applicable) 243

70. palatal ridges: palatal ridges are absent on the palate of roskellyae. Weak ridges are present on the palate of hilmeri and pitikantensis. The palates of vanderleueri and sp. cf. hilmeri have long, well-developed ridges. carnifex demonstrates short, longitudinally-oriented ridges. Palatal ridges are absent in all outgroup taxa except Trichosurus. 0 = absent; 1 = present.

71. occiput height: relative to the size of the foramen magnum, the height of the occiput in roskellyae is moderately tall, while that of hilmeri, vanderleueri and carnifex is tall. The occiput in the phalangerids is relatively short. The occiput in Echymipera is tall and in Litokoala it is moderately. 0 = tall (ratio >2); 1 = moderately tall (ratio1.3 < 2); 2 = short (ratio < 1.3).

72. masseter process of zygoma: roskellyae, hilmeri, crassidentatus and carnifex lack a masseter process of the zygoma. A relatively short masseter process is present in vanderleueri and an incipient process is present in sp. cf. hilmeri. A masseter process is present in Litokoala, Nimiokoala and Trichosurus but is absent in Caudipilosus and Echymipera. 0 = absent; 1 = present.

11.3.3 Parsimony Analyses

Two parsimony analyses were carried out on the data matrix: 1) all-taxa included 2) reduced taxa: deminutivus excluded from the matrix

Following the initial all-taxa analysis, the area of greatest instability in the resulting trees revolved around the position of deminutivus. This instability is most likely a result of the very limited character information available for this species; it is known from a single tooth and thus most of the characters (63 out of 72) in the matrix were scored as uncertainties (?) for this taxon. In an attempt to improve the resolution of thylacoleonid relationships, a second analysis was performed with that species excluded. 244

Echymipera Fig. 11.2a. Consensus tree for 9 most Litokoala parsimonious trees obtained from Caudipilosus “unordered” parsimony analysis. Trichosurus Strict, 50 % majority-rule and Adams consensus have the same topology. Nimiokoala majority-rule frequency attenboroughi bootstrap frequency roskellyae 100 63 deminutivus 100 Fig. 11.2b Apomorphies at nodes. 100 pitikantensis 97 68 hilmeri The derived state of a character is separated from the 100 sp. cf. hilmeri character number by a point (.) Character state 100 oldfieldi reversals are indicated by a negative sign (-). 100 100 Unequivocal apomorphies are indicated in bold type. 83 vanderleueri 83 100 The consistency index and direction of change for 84 alcootaensis character states at nodes are listed in Appendix 4.

hilli Node A (Thylacoleonidae) 100 crassidentatus deltran: 11.3, 35.1 83 acctran: 5.1, 7.2, 11.3, 15.2, 32.1, 34.2, 35.1, 42.1, 50.2, 51.2, 53.1, carnifex 56.3, 57.1, 69.1

Node B deltran: 5.1, 7.2, 21.1, 22.2, 23.1, 25.1, 26.2, 29.1, 32.1, 38.1, 50.2, 53.1, 56.3, 57.1 acctran: 21.1, 22.2, 23.1, 25.1, 26.2, 29.1, 37.1, 38.1

Node C Echymipera deltran: 24.1 acctran: 13.1, 14.1, 15.3, 16.2, 17.1, 20.1, 24.1, 31.1, 34.3, 37.2, 40.2, Litokoala 42.2, 46.2, 47.3, 55.2, 62.1, 63.2, 64.1, 65.1, 66.2, 68.3, 70.1 Caudipilosus Node D Trichosurus deltran: 16.2, 46.2, 47.3, 51.2, 55.2, 62.1, 63.2, 65.1, 66.2, 68.3

Nimiokoala Node E attenboroughi deltran: 14.1, 27.1, 28.4, 33.2, 34.3, 37.2, 40.2, 48.2, 61.3, 64.1 acctran: 2.1, 27.1, 28.4, 29.2, 33.2, 48.2, -61.0, -69.0 roskellyae A Node F (pitikantensis & hilmeri) deminutivus deltran: 36.2, 70.1 B pitikantensis acctran: -20.0, 36.2, 43.2 F hilmeri C Node G E sp. cf. hilmeri deltran: 20.1, 23.2, 29.3, 31.2, 35.2, 70.2, 72.1 acctran: 2.2, 5.2, 13.2, 15.4, 17.2, 18.1, 23.2, 29.3, 31.2, 35.2, 41.2, G oldfieldi 42.3, 52.1, 56.4, 70.2, 72.1 H vanderleueri D Node H I alcootaensis deltran: 2.2, 5.2, 13.2, 17.2, 24.2, 32.2, 34.4, 36.3, 37.3 acctran: 6.2, 24.2, 32.2, 34.4, 36.3, 37.3 hilli J Node I crassidentatus deltran: 6.2, 27.2, 35.3 K carnifex acctran: 15.5, 18.2, 19.1, 27.2, 35.3

Node J (Thylacoleo) deltran: 1.2, 10.2, 30.1 acctran: 1.2, 5.3, 10.2, 12.2, 13.3, 14.2, 15.6, 17.3, 18.3, 19.1, 23.3, 24.2, 26.3, 30.1, 32.3, 34.4, 37.3, 39.1, 40.3, - 41.0, 42.4, 44.1, 45.1, Fig. 11.2b One of 9 equally parsimonious trees, 52.2, 54.4, 57.2, 60.2, 64.2, 67.2, 70.3 Unordered characters (L=194, RC=0.67, RI=0.81). Apomorphies for nodes are listed in the Node K neighbouring box. Unresolved nodes on the strict deltran: 5.3, 12.2, 13.3, 14.2, 15.6, 17.3, 18.3, 19.1, 20.1, 23.3, 24.3, 26.4, 32.3, 34.4, 39.1, 67.2 consensus tree are indicated by a diamond. acctran: -7.0, 24.3, 26.4 245

For each of the above analyses, two analyses were performed, one with all characters unordered, the second with characters ordered. The following 21 characters were ordered: 1, 2, 5, 6, 13, 14, 15, 16, 17, 18, 21, 23, 24, 31, 32, 34, 35, 36, 40, 42 and 52. The data matrix was analysed using PAUP 4.0b10 (Swofford 1998). The branch- and-bound method was used to search for trees of the shortest length. The degree of support for the resulting groupings in each analysis was determined by carrying out a bootstrap analysis (2000 replicates).

Characters were optimised using the delayed (DELTRAN) and accelerated (ACCTRAN) transformation functions in PAUP. DELTRAN optimisation delays the transformation of a character on a tree. This results in a character being moved towards the ends of the branches and favours parallelisms over reversals. ACCTRAN optimisation accelerates the transformation of a character and results in the character being moved in the opposite direction, down the tree, and favours reversals over parallelisms.

Strict consensus and 50% majority rule consensus trees were determined for each analysis. Where polytomies occurred, an Adams consensus tree was constructed to identify the most unstable taxa. Because strict consensus trees may reduce resolution and may not actually represent any of the original most-parsimonious trees, a randomly-selected tree from the group of most-parsimonious trees resulting from each analysis is also figured. Apomorphies for the nodes on this tree are listed next to this tree. Unambiguous apomorphies are noted in bold face.

11.4 Results

All taxa included Analysis of the data matrix using a branch-and-bound search (ACCTRAN optimisation) with all characters unordered resulted in nine equally-parsimonious trees (194 steps, rescaled consistency index 0.67, retention index 0.81). One character was parsimony uninformative (8: anterior crest of P3). Strict, 50% majority-rule and Adams consensus of these nine trees all had the same topology 246

11.3a strict consensus 11.3b 50% majority-rule consensus

Echymipera Echymipera Litokoala Litokoala Nimiokoala Nimiokoala Caudipilosus Caudipilosus Trichosurus Trichosurus attenboroughi attenboroughi

deminutivus roskellyae roskellyae 100 deminutivus 65 100 pitikantensis 98 pitikantensis hilmeri 86 hilmeri 79 sp.cf. hilmeri sp.cf. hilmeri 86 oldfieldi oldfieldi vanderleueri 86 100 vanderleueri 53 alcootaensis 79 86 alcootaensis hilli hilli 100 crassidentatus 86 100 crassidentatus 76 carnifex carnifex

majority rule frequency bootstrap frequency 11.3c Adams consensus

Echymipera Litokoala Fig. 11.3a-c Nimiokoala Consensus of 21 equally Caudipilosus parsimonious trees obtained from Trichosurus ordered parsimony analysis

attenboroughi (a) strict deminutivus (b) 50% majority-rule (c) Adams roskellyae pitikantensis hilmeri

sp.cf. hilmeri oldfieldi vanderleueri alcootaensis hilli crassidentatus

carnifex 247

Echymipera Litokoala Fig. 11.4 Nimiokoala One of 21 equally-parsimonious trees, Caudipilosus ordered characters. Trichosurus (L=202, RC=0.67, RI=0.84)

attenboroughi Apomorphies for nodes are listed below. A roskellyae Unresolved nodes on the strict consensus deminutivus tree are indicated by a diamond. B E pitikantensis F hilmeri C sp. cf. hilmeri

D oldfieldi vanderleueri G I alcootaensis H hilli J crassidentatus K carnifex

Apomorphies at nodes The derived state is separated from the character number by a point (.) Character state reversals are indicated by a negative sign (-). Unequivocal character changes are indicated in bold.

Node A (Thylacoleonidae) deltran: 11.3, 35.1 acctran: 5.1, 7.2, 11.3, 15.2, 16.1, 28.1, 32.1, 34.2, 35.1, 42.1, 50.2, 51.2, 53.1, 56.3, 57.1 Node B deltran: 5.1, 7.2, 15.2, 21.1, 22.2, 23.1, 25.1, 26.2, 27.1, 29.1, 32.1, 34.2, 38.1, 42.1, 50.2, 53.1, 56.3, 57.1 acctran: 21.1, 22.2, 23.1, 25.1, 26.2, 27.1, 28.3, 29.1, 37.1, 38.1 Node C deltran: 2.1, 13.1, 14.1, 15.3, 16.2, 17.1, 24.1, 28.4, 31.1, 33.2, 34.3, 37.2, 40.2, 42.2, 46.2, 47.3, 51.2, 55.2, 62.1, 63.2, 64.1, 65.1, 66.2, 68.3 acctran: 2.1, 13.1, 14.1, 15.3, 16.2, 17.1, 24.1, 28.4, 31.1, 33.2, 34.3, 36.2, 37.2, 40.2, 42.2, 46.2, 47.3, 48.2, 55.2, 61.3, 62.1, 63.2, 64.1, 65.1, 66.2, 68.3, 70.1 Node D deltran: 20.1, 23.2, 29.3, 31.2, 35.2, 70.2 acctran: 2.2, 5.2, 13.2, 15.4, 17.2, 18.1, 20.1, 23.2, 29.3, 31.2, 35.2, -41.0, 42.3, 52.1, 70.2, 72.1 Node E acctran: 43.2 Node F deltran: 36.2, 70.1 acctran: 29.2 Node G deltran: 5.2, 13.2, 15.4, 17.2, 18.1, 24.2, 32.2, 34.4, 36.2, 37.3 acctran: 24.2, 32.2, 34.4, 36.3, 37.3 Node H deltran: 15.5, 18.2, 19.1, 42.3, 52.1 acctran: 15.5, 18.2, 19.1, -27.0, 35.3 Node I deltran: 6.1, 27.2, 35.3 acctran: 6.2, 27.2, 41.2, 56.4 Node J (Thylacoleo) deltran: 1.2, -2.0, 10.2, - 28.0, -29.1, 30.1 acctran: 1.2, -2.0, 5.3, 10.2, 12.2, 13.3, 14.2, 15.6, 17.3, 18.3, 23.3, 26.3, -28.0, -29.1, 30.1, -31.1, 32.3, -33.0, 40.3, 42.4, 44.1, 45.1,- 48.0, 52.2, 54.4, 57.2, 60.2, -61.0, 64.2, 67.2, 69.1, 70.3, -72.0 Node K deltran: 5.3, 12.2, 13.3, 14.2, 15.6, 17.3, 18.3, 23.3, 24.3, 26.4, -27.0, -31.1, 32.3, -33.0, 39.1, 67.2, acctran: -7.0, 24.3, 26.4, 248

(Fig. 11.2a). The strict consensus places attenboroughi as the sister-group to a clade that includes the remaining thylacoleonid taxa. This arrangement had a high majority-rule frequency (100%) but produced only moderate bootstrap frequency of 63%. Within the latter clade roskellyae is the sister-group to a tritomy that includes the remaining species of thylacoleonids. Support for this branch was robust with a high bootstrap frequency (97%) occurring at this node. Within this tritomy, deminutivus is a sister-group to two clades, one containing hilli, crassidentatus and carnifex and the other containing pitikantensis, hilmeri, sp.cf. hilmeri, vanderleueri and alcootaensis. Within the latter grouping, hilmeri pairs with pitikantensis, and vanderleueri pairs with alcootaensis. Support for the vanderleueri-alcootaensis grouping was stronger (bootstrap 84%) than for the former (bootstrap 68%). oldfieldi lies as a sister-group to the vanderleueri-alcootaensis pairing (bootstrap 83%) and sp.cf. hilmeri is in turn the sister-group to oldfieldi. Support for the hilli- crassidentatus-carnifex clade was also high (bootstrap 83%). Apomorphies for nodes and terminal taxa for one of the nine trees (for the tree in Fig. 11.2b) are listed in Appendix 4A and 4B.

A branch-and-bound analysis with characters ordered resulted in 21 equally- parsimonious trees (202 steps, rescaled consistency index 0.67, retention index 0.84). Strict, majority-rule and Adams consensus trees for this analysis are shown in Fig. 11.3 a-c. One of the 21 equally-parsimonious trees, randomly selected and listing apomorphies for branches, is illustrated in Fig. 11.4. Ordering of characters resulted in reduced resolution of some relationships as indicated by the polytomy in the strict consensus (Fig. 11.3a). Similarities to the strict consensus resulting from the unordered analysis include the position of attenboroughi as a sister-group to the remaining thylacoleonids, the grouping of hilli, crassidentatus and carnifex, and the pairing of vanderleueri with alcootaensis. Majority-rule consensus resulted in some resolution of the polytomy with oldfieldi occurring as the sister-group to a clade that includes hilli, crassidentatus, carnifex, vanderleueri and alcootaensis; however, deminutivus, pitikantensis and hilmeri remain as a tritomy. There is strong support for the clade containing all thylacoleonids excluding attenboroughi; this branch appears in all 21 trees and has a bootstrap frequency of 98%. Other ubiquitous clades 249

Echymipera Echymipera

Litokoala Litokoala Caudipilosus Caudipilosus Trichosurus Trichosurus Nimiokoala Nimiokoala

attenboroughi attenboroughi

roskellyae 100 roskellyae 65 pitikantensis pitikantensis 100 100 98 79 hilmeri hilmeri 100 67 sp. cf. hilmeri sp.cf. hilmeri 75 oldfieldi oldfieldi 100 100 vanderleueri 82 95 100 vanderleueri 85 alcootaensis alcootaensis

hilli hilli 100 crassidentatus crassidentatus 92 carnifex carnifex

Fig. 11.5a Fig. 11.5b Strict consensus of 4 equally parsimonious trees 50% majority-rule consensus of 4 obtained from unordered parsimony analysis with trees obtained from unordered deminutivus excluded. Adams consensus has the analysis with deminutivus same topology. excluded

majority-rule frequency bootstrap frequency

Echymipera

Litokoala Nimiokoala Caudipilosus

Trichosurus

attenboroughi 100 68 roskellyae 100 Fig. 11.6 99 pitikantensis Strict, 50% majority rule and Adams hilmeri 100 consensus of 3 equally parsimonious 94 sp.cf. hilmeri trees obtained from ordered analysis with deminutivus excluded. 100 oldfieldi 78 100 vanderleueri 100 majority-rule frequency 74 80 100 alcootaensis bootstrap frequency 63 hilli 100 94 crassidentatus 79 carnifex 250 are those containing hilli-crassidentatus-carnifex (bootstrap frequency 86%) and vanderleueri-alcootaensis (bootstrap frequency 79%). The clade that includes all thylacoleonids except attenboroughi and roskellyae is relatively well supported, occurring in 86% of the shortest trees (bootstrap frequency 79%). Adams consensus results in some resolution of the polytomy with both roskellyae and deminutivus lying as sister-groups to the remaining thylacoleonid species. Within the latter, the pairing of alcootaensis and vanderleueri lies as a sister-group to the hilli- crassidentatus-carnifex grouping, although bootstrap support for this branch is weak. oldfieldi is the next inclusive taxa to that clade, followed by sp. cf. hilmeri, followed by both pitikantensis and hilmeri.

Reduced taxa - deminutivus excluded A branch-and-bound, unordered analysis (ACCTRAN) produced four equally- parsimonious trees (192 steps, rescaled consistency index 0.67, retention index, 0.81). The strict consensus tree (Fig. 11.5a) derived from these four trees has a topology that is very similar to the strict consensus tree of the all-taxa unordered analysis; however, it differs in that the taxon sp. cf. hilmeri lies as a sister-group to two groups, one containing the pitikantensis-hilmeri pairing (bootstrap frequency 79%) and the other containing oldfieldi-vanderleueri and alcootaensis (bootstrap frequency 82%) The 50% majority-rule consensus tree (Fig. 11.5b) resolves the position of sp. cf. hilmeri, placing it as the sister-group to the latter clade. Support for the clade that includes all taxa but excludes attenboroughi is robust (bootstrap frequency 98%) as is support for the clade that excludes attenboroughi and roskellyae (bootstrap frequency 95%).

When characters were ordered, a branch-and-bound search (ACCTRAN) resulted in three equally-parsimonious trees (199 steps, rescaled consistency index 0.67, retention index, 0.84). The strict, 50% majority-rule and Adams consensus (Fig. 11.6) for these trees shows a different topology to that of the unordered analysis; pitikantensis and hilmeri are both sister-groups to a large clade that includes all remaining thylacoleonids except attenboroughi and roskellyae (bootstrap frequency 94%). Within this large grouping, crassidentatus, hilli and carnifex form a clade as 251 do vanderleueri and alcootaensis (bootstrap frequency 63%). Bootstrap support for these two groups is high, 94% for the former and 80% for the latter. These two clades are united as sister-groups by a weakly supported branch (63% bootstrap frequency). The next inclusive taxon to this clade is (successively), oldfieldi (bootstrap frequency 74%) then sp. cf. hilmeri (bootstrap frequency 78%).

11.5 Discussion

The phylogenies resulting from this study show varying degrees of support for previously proposed hypotheses of thylacoleonid relationships (see Fig. 11.1). For example, the consistent grouping of the taxa hilli, crassidentatus and carnifex support the Thylacoleo clade. There is also strong support for the hypothesised relationships of the three Wakaleo species, oldfieldi, vanderleueri and alcootaensis with the former being the sister-group to the latter two. However, the relationships previously proposed for other taxa, in particular that of pitikantensis, does not appear to be supported and consequently the genera Priscileo and Wakaleo as previously known are not reciprocally monophyletic.

The taxon pitikantensis (SAM P37719) consistently lies as the sister taxon to the newly described Riversleigh species hilmeri. This pair nests within the Wakaleo clade as the sister-group to another new taxon sp. cf. hilmeri, which in turn is the sistergroup to the oldfieldi-vanderleueri-alcootaensis branch. The pitikantensis- hilmeri branch has relatively good bootstrap support. This pairing is based on dental similarities. A single unequivocal synapomorphy (unordered analysis) unites these two taxa, (36) similarity in the buccal and lingual heights of M2. The consistency of this pairing suggests that pitikantensis is more closely related to species of Wakaleo and should be referred to that genus. Such a referral results in monophyly for Wakaleo. At present, differences in the postcrania of pitikantensis and hilmeri support their specific differentiation within the genus (see Chapter 13). These two taxa may share a common ancestor or, in light of the morphoclines present within Wakaleo, it is possible that pitikantensis may be ancestral to hilmeri. The taxon 252 pitikantensis was the type species for the genus Priscileo Rauscher 1987. Referral of this taxon to Wakaleo renders Priscileo a junior synonym of Wakaleo.

Rauscher’s (1987) hypothesis that pitikantensis lies outside Wakaleo-Thylacoleo is not supported, neither is the proposal of Murray et al. (1987) that pitikantensis was possibly more closely related to species of Thylacoleo than to Wakaleo. The repositioning of pitikantensis to Wakaleo allows the subfamilial construct Wakaleoninae (Murray et al. 1987) to remain potentially monophyletic. For Thylacoleoninae to remain monophyletic it would have to contain only species of Thylacoleo.

In both unordered and ordered analyses, the taxon sp. cf. hilmeri occurs between hilmeri and oldfieldi, a position that supports the hypothesis that it is a possible transitional form between those two species (see Chapter 6).

In all analyses, all thylacoleonid taxa group together to form a monophyletic clade. Bootstrap support for this clade was moderate (64 – 66%). The single, unequivocal synapomorphy uniting this clade in unordered and ordered analyses is the presence of M2 with a subsquare outline.

All analyses indicate there is a distinct dichotomy within thylacoleonids between the new taxon attenboroughi and all other thylacoleonid species. Support for this branch is moderately high in all analyses. Apomorphies for attenboroughi indicated by the unordered analysis includes (character number in brackets): (16) M2 with trigonid and talonid basins of similar height, (28) a well-developed posterobuccal crest on P3 and (37) M2 with lingual and buccal margins of equivalent height. The latter character was the sole unambiguous apomorphy resulting from the ordered analysis. The position of this taxon relative to all other thylacoleonids and its morphological differences support its generic distinction (Microleo). Unequivocal synapomorphies for the clade containing all remaining thylacoleonid taxa (ordered and unordered analyses) include: (21) anterolingual location of P2, (22) lingual shearing facet on 253

P3, (23) P3 length: cheektooth row length value of at least 0.3, (25) a lingually- convex longitudinal blade on P3, (26) horizontally-bowed posterior longitudinal blade on P3, (29) a small anterolingual crest on P3, (39) buccal expansion of M2 adjacent to paracone, and (40) the lingual margin of M2 lacks a notch. An additional synapomorphy resulting from the ordered analysis is (27) mid-crown constriction of P3. Many of these characters relate to the morphology of P3. The relative increase in P3 length, change in direction and length of its shearing facet and increasing height of the posterior longitudinal blade suggest an increased importance of the shearing function of this tooth which is also suggestive of dietary changes. These dietary changes possibly reflect a change from insectivory to omnivory or an increasing carnivorous component in an omnivorous diet. If the P3 characters are correlated with an increased shearing function, it is possible they may not be independent.

In the large thylacoleonid clade that excludes attenboroughi, a second distinct dichotomy is also apparent between roskellyae and all the remaining thylacoleonid taxa. This branch has strong bootstrap support (97-98%). The relationship of roskellyae to other thylacoleonids and its obvious morphological distinctness from species of Wakaleo warrants its allocation to a separate genus. The synonymising of Priscileo with Wakaleo requires this species to be assigned to a new genus (Lekaneleo). Apomorphies for roskellyae common to both unordered and ordered

(and to both ACCTRAN and DELTRAN) analyses include: (10) the presence on P3 of a short, weak posterobuccal crest immediately anterior to the end of the premolar, (54) a short, broad and thin postglenoid process, (60) a broad and deep epitympanic wing of the petrosal, and (67) short and wide anterior palatal fenestrae. Additional apomorphies resulting from the unordered analysis include: (27) P3 with a moderate, lingual mid-crown constriction, (28) P3 with a moderately-developed, oblique, long posterobuccal crest, and (48) a small lachrymal fossa. A single unequivocal synapomorphy, (24) P3 with a moderately long longitudinal blade relative to P3 crown length (0.4-0.5), unites the large clade containing the remaining thylacoleonid taxa (attenboroughi excluded) in the unordered analysis. The ordered analysis resulted in a further nineteen synapomorphies for this clade (see Appendix 4C). 254

The relationship of deminutivus to other thylacoleonids is unclear. In six of the nine most-parsimonious trees (unordered analysis) the position of deminutivus is equally divided between being the sister-group of a Thylacoleo-Wakaleo (including pitikantensis) clade and being the sister-group to a Wakaleo-pitikantensis clade. In the three remaining most-parsimonious trees (unordered analysis), deminutivus occurs as the sister taxon to the Thylacoleo clade but support for this branch is weak (less than 50% bootstrap frequency). No unequivocal synapomorphies unite this clade. Of the 33 equivocal apomorphies listed for this branch, only two are applicable to deminutivus; a gently sloping posterior blade on P3 (a state shared with T. hilli) and the loss of the midcrown constriction of P3. Both character states are reversals and as such are not compelling.

The strongly supported Thylacoleo clade (83 - 94% bootstrap) is united by the following unequivocal synapomorphies (unordered analysis): (1) absence of I2 and 3 C1, (10) P posterior width narrower than anterior width and (30) presence of a posterobuccal crest at the end of P3. Two additional synapomorphies, both reversals, occur in the ordered analysis: (2) three lower premolars and (29) small anterolingual crest on P3. In unordered and ordered analyses, ACCTRAN optimisation resulted in a large number of apomorphic characters for the Thylacoleo node. However, because most of the characters for T. hilli are missing, it is uncertain whether their states are plesiomorphic or derived, hence most of the character states are equivocal for the genus.

Ordering of characters reduced the resolution of some relationships, for example those of roskellyae, deminutivus and pitikantensis, however, others persisted: attenboroughi remained as a sister-group to all remaining thylacoleonids, vanderleueri paired with alcootaensis, and hilli-crassidentatus-carnifex formed a clade. Adams consensus resulted in some improvement in resolution, producing a pairing of both deminutivus and roskellyae as sister-groups to the remaining species. Removing the relatively unstable taxon, deminutivus, from the analyses did not change the relationship of either roskellyae or pitikantensis to the remaining 255 thylacoleonid species, the former occurring as the sister to a clade containing all thylacoleonids except attenboroughi and the latter occurring as the sister taxon to hilmeri.

The Thylacoleo clade occurs as the sister-group to either a Wakaleo clade (including pitikantensis - unordered analysis), or to a clade containing W. vanderleueri and W. alcootaensis (ordered analysis). In the unordered analysis, apomorphies uniting Thylacoleo and Wakaleo only occur when optimisation is delayed and predominantly involves cranial characters (see Node D, Fig. 11.2b). In the ordered analysis, three unequivocal synapomorphies unite the vanderleueri- alcootaensis -Thylacoleo clade; (15) shortening of M2 relative to P3, (18) presence of M3 that is oval in outline and (19) loss of the talonid basin on M3. The synapomorphy previously thought to unite Wakaleo and Thylacoleo, loss of M4 (Rauscher 1987), is rejected due to the presence of this molar in W. hilmeri.

The results of this analysis present two scenarios for the ancestry of Thylacoleo. The ordered analyses suggest Thylacoleo may have evolved during the middle to late Miocene, possibly arising from the ancestor of W. vanderleueri and W. alcootaensis. Bootstrap support for this particular relationship, however, is weak. A similar scenario has also been proposed by Murray and Megirian (1990) who suggested that Thylacoleo probably evolved from a descendant of pitikantensis in the mid-late Miocene. In contrast, the unordered analyses suggest that Thylacoleo is a relatively old genus, a view also posited by Archer and Dawson (1982; see their Fig. 8), whose ancestor possibly evolved in the early Miocene. The taxon roskellyae lies as the sister-group to all thylacoleonids (excluding attenboroughi) and it, or possibly its descendents (e.g. deminutivus), may have given rise to Thylacoleo. At present, there is no pre-Pliocene record for Thylacoleo. Considering the relative paucity of Miocene vertebrate localities on the Australian continent and in New Guinea, it is highly possible that the ancestor of Thylacoleo may have evolved in a region that has not yet yielded a fossil record of appropriate age. 256

Microleo attenboroughi

Lekaneleo roskellyae

Lekaneleo deminutivus

Wakaleo pitikantensis

Wakaleo hilmeri Wakaleo sp. cf. W. hilmeri

Wakaleo oldfieldi Wakaleo vanderleueri

Wakaleo alcootaensis

Thylacoleo hilli

Thylacoleo crassidentatus

Thylacoleo carnifex

Fig. 11.7

Possible interrelationships of thylacoleonids incorporating taxonomic changes resulting from this study. 257

A cladogram showing hypothetical thylacoleonid relationships that incorporate the systematic changes supported by the outcomes of these analyses are shown in Figure 11.7.

Summary The results of this analysis indicate that, within Thylacoleonidae, there appears to be two distinct dichotomies, one in which the new taxon Microleo attenboroughi is the sister-group to all other species, and the other in which the taxon roskellyae is the sister-group to all remaining marsupial lion species. Within this clade there is strong support for the monophyly of Thylacoleo. The taxon pitikantensis consistently occurs as the sister-group to the taxon hilmeri, which lends support to a hypothesis that they are congeneric. Because of this pairing, “Priscileo” and “Wakaleo” as currently defined fail to form monophyletic clades. Referral of pitikantensis (originally the type species of Priscileo) to Wakaleo results in monophyly for this genus. Wakaleo oldfieldi consistently occurs as the sister-group to W. vanderleueri and W. alcootaensis. Wakaleo sp. cf. W. hilmeri lies between W. hilmeri and W. oldfieldi, providing support for the hypothesis that it is an intermediate stage between those taxa. Synonymy of Priscileo with Wakaleo requires the allocation of the distinctive roskellyae taxon to a new genus. The position of deminutivus is uncertain. This taxon occurs as either the sister-group to roskellyae and/or as the plesiomorphic sister- group to Wakaleo and Thylacoleo or to W. pitikantensis - W. hilmeri.

The results of this analysis strongly support the position of Wakaleo as the sister group to Thylacoleo. The positioning of pitikantensis outside the genus Wakaleo in the phylogenies of Rauscher (1987) and Murray et al. (1987) is not supported. The reallocation of the taxon pitikantensis to Wakaleo allows the subfamily Wakaleoninae (Murray et al. 1987) to remain potentially monophyletic. 258

CHAPTER 12

INTERFAMILIAL RELATIONSHIPS

12.1 Use of names

The taxonomic names used in this analysis are the published names. Hence, Priscileo roskellyae is used rather than Lekaneleo roskellyae. Ingroup and outgroup taxa are referred to by their genus name rather than species name for ease of reading.

12.2 Historical background

The relationships of the Thylacoleonidae within the Australian marsupial radiation remain unresolved. Most theories regarding the relationships of the family were based on study of Thylacoleo carnifex, the first marsupial lion species found, and the only one known until the 1970s. The consensus over time has been that the family is derived from phalangeridans. However, some recent studies suggest that thylacoleonids may be vombatomorphians.

Initially, Owen (1859) regarded the obviously carnivorous Thylacoleo to be most closely allied to the other carnivorous marsupials, the Dasyuridae, but subsequent study (Owen 1866) of more complete Thylacoleo specimens, specifically the identification of the two procumbent lower incisors, led him to conclude that the group had closer affinities to the diprotodont marsupials. However, he did not link thylacoleonids to any specific diprotodont group.

Krefft (1872), Broom (1898), and Bensley (1903) went further and considered Thylacoleo to be most closely related to the “Phalanger” family. Krefft (1872 p.177) compared the dentition of Thylacoleo with representatives of the various marsupial groups and found close similarities in the upper and especially the lower dentition of the “phalangers” - “The general form of the lower jaw of our marsupial friend (Thylacoleo) is undoubtedly that of a phalanger...". Similarly, Broom (1898; p.63) states that "...The structure of the lower jaw and the dentition render it manifest that 259 the form is more related to the Phalangers then to any other living group...". Based on the dental morphology, especially the presence of the enlarged premolar in both groups, Broom also suggested that the thylacoleonids were the sister-group of the ‘Burramyinae’ (burramyids) and had developed from a Burramys-like ancestor. Bensley (1903; p.203) stated that Thylacoleo exhibited a "normal phalangerid type" dentition," …represented by the diprotodont modification, involving enlargement of the median incisors and reduction of the posterior antemolar teeth, and by the enlargement of the posterior premolars as sectorials...". He therefore concluded that it was a derivative of the Phalangeridae and placed Thylacoleo in its own family, the Thylacoleontidae. This family was part of a polytomy which included the Phalangeridae, Macropodidae, and a combined Diprotodontidae/Vombatidae group.

Later researchers (Ride 1964, Archer 1976, Archer and Rich 1982) also expressed support for phalangeroid affinity. Ride (1964; p.118) stated that the thylacoleonids are "...obviously derived from large phalangers..." and placed them in their own subfamily, the Thylacoleoninae, within the Phalangeridae. However, he challenged Broom’s (1898) proposal of affinities with the burramyids because it is based solely on the presence of the enlarged sectorial premolar, a feature which has been shown to have developed independently in a variety of marsupials (e.g. Marshall 1982).

More recently, Murray et al. (1987) also expressed support for a phalangeroid affinity for the thylacoleonids. From comparative studies of the basicranium of Wakaleo, Thylacoleo, Trichosurus, Phalanger, Lasiorhinus, Vombatus and Cercartetus they found that Wakaleo and phalangerids are highly similar in the structure of the palate, antemolar dental formula, and in particular, the bilaminar (alisphenoid/squamosal) structure of the tympanic bulla. However, in regard to this latter character, Thylacoleo exhibits the vombatomorphian state, i.e. a fully squamosal bulla. To explain the presence of this feature in Thylacoleo they suggested that the fully squamosal bulla must have developed twice, once in Thylacoleo and once in the remaining vombatomorphians. Murray et al. (1987, p.465) concluded that thylacoleonids “… are more clearly aligned with primitive phalangeroids and not vombatoids. Their primary sister-group is the burramyidae”. A hypothesis of diprotodontian relationships proposed by these authors is shown in Fig. 12.1a. In this 260

Petaurids Fig. 12.1a Burramyids Hypothesis of thylacoleonid relationships proposed by Murray et al. (1987). Macropodids Phalangerids Diprotodontids Vombatids Thylacoleonids

Macropodidae Fig. 12.1b Phalangeridae Hypothesis of diprotodontian relationships proposed by Marshall Burramyidae et al. (1990). Petauridae

Phascolarctidae Thylacoleonidae Ilariidae Wynyardiidae Vombatidae

Palorchestidae Diprotodontidae

Phalangeriformes Fig. 12.1c Hypothesis of Phascolarctidae diprotodontian relationships Thylacoleonidae proposed by Munson (1990). Ilariidae Vombatidae Wynyardiidae

Palorchestidae Diprotodontidae 261 phylogeny, thylacoleonids are viewed as the sister-group of a combined macropodid- phalangerid/diprotodontid-vombatid clade. These groups were in turn viewed as the sister-group of the petaurids and burramyids. However, rather than supporting their contention of phalangeroid affinities for thylacoleonids, this phylogeny suggests that thylacoleonids are the sister-group to all other diprotodontians, except burramyids, and that diprotodontids and vombatids are more closely related to phalangeroids.

Rauscher (1987), although acknowledging Ride's (1964) observation that the similarities between Thylacoleo and Burramys are mainly those of the premolar, also suggested that the thylacoleonids could be derived from a burramyid lineage. This proposal was based on the loss of the hypocone in the upper molars evident in both groups. Rauscher also noted, however, that this loss may be the result of convergence due to the pressures of carnivory, as previously suggested by Archer and Rich (1982).

In contrast to the above views, Aplin and Archer (1987) proposed vombatomorphian affinities for the Thylacoleonidae. They placed the family within the Infraorder Vombatomorphia, a group containing four other extinct families (diprotodontids, palorchestids, wynyardiids, and ilariids) and the Vombatidae. This classification was based on the presence of a number of similar cranial and postcranial features within thylacoleonids and vombatomorphians (Aplin 1987). The most significant cranial feature present in both groups is the "squamosal bulla", as seen in Thylacoleo but not Wakaleo.

Marshall et al. (1990) combined dental, osteological and soft anatomical data gathered from a number of studies to produce a phylogeny of living and fossil families of marsupials. A part of the cladogram resulting from that analysis showing the position of Thylacoleonidae within Diprotodontia is shown in Fig. 12.1b. Within this phylogeny, marsupial lions are placed as part of the vombatiform radiation. Phascolarctids lie at the base of this radiation and are the plesiomorphic sister-group to thylacoleonids, which are in turn the basal member of the vombatoid (= Vombatomorphia, sensu Aplin and Archer 1987) superfamily. Rather than viewing the bilaminar bulla as a phalangeriform character, these authors viewed it (along with 262 the extension of the squamosal into the roof of the alisphenoid hypotympanic sinus) as a synapomorphy uniting the vombatoids.

Munson (1992) carried out an extensive comparative study of the known vombatiform postcrania. This study included postcrania of two thylacoleonids, Thylacoleo carnifex and Priscileo pitikantensis. Munson's phylogenetic conclusions were similar to those of Marshall et al. (1990), i.e. the vombatiforms are derived from a possum-like ancestor and thylacoleonids are the sister-group of the other vombatoids (sensu Marshall et al. 1990). A cladogram constructed by Munson (1990) showing the relationships of Vombatiformes synthesised from both postcranial and cranial data is shown in Fig. 12.1c.

Problem Areas In contrast to Murray et al. (1987), Aplin and Archer (1987, p.xlvii) viewed the bilaminar bulla found in Wakaleo as "... a phylogenetically insignificant development which might just as readily be viewed as a secondary modification of the more typical condition in Thylacoleo...". In regard to the postcranial skeleton, Aplin and Archer (1987) indicated that similarities are evident between the major limb bones of living vombatids, Wynyardia bassiana, the palorchestid Ngapakaldia tedfordi and Thylacoleo carnifex. They did, however, admit that much of the relevant morphological evidence to support the monophyly of the vombatomorphians had yet to be investigated or presented. Aplin and Archer (1987) also pointed out that due to the large proportion of fossil families within the Vombatomorphia, determination of the interfamilial relationships would have to depend solely upon the analysis of dentitions and osteology, unlike studies of relationships of living marsupials, which can also employ molecular information (for example, Horovitz et al. 2003, Asher et al. 2004). This dependence on dental and skeletal morphology is further complicated by the fact that the dentitions of most vombatomorphians are highly derived and therefore provides few useful characters to assist in determining relationships. Similarly, Munson (1992) in her phylogenetic analysis of vombatomorphian postcrania noted that determination of the relationships of thylacoleonids to other vombatiforms is difficult because of the limited postcranial remains of Priscileo and the relatively derived skeletal morphology of Thylacoleo carnifex. 263

Numerous cladistic analyses of the interrelationships of marsupials (e.g. Retief et al. 1995, Springer et al. 1997, Springer et al. 1998, Horovitz et al. 2003, Asher et al. 2004) have been carried out using algorithm-based parsimony analysis. However, very few cladistic analyses have been undertaken of the interfamilial relationships of the order Diprotodontia that include its extinct families. Munson’s (1990) study of vombatiform relationships based on postcranial characters employed parsimony analysis; however, this study did not include all extinct vombatiform families. The study of Marshall et al. (1990, p. 434) combined data from a number of cladistic and algorithm analyses but “…all data were not subjected to a single algorithm analysis”. Similarly, the diprotodontian relationships discussed by Aplin and Archer (1987) and which form the basis of their classification, are based on cladistic principles but the characters discussed were not subject to algorithmic parsimony analysis.

The aim of this analysis is to test previous hypotheses of the relationships of marsupial lions within the order Diprotodontia using parsimony analysis. These hypotheses are: (1) that thylacoleonids are derived from a burramyid-like ancestor and are more closely related to phalangeroids than vombatoids (sensu Woodburne 1984); (2) that thylacoleonids are part of the vombatiform radiation; and (3) that thylacoleonids are the most plesiomorphic members of the vombatomorphian clade. Previously, the basicranium of most extinct vombatiform taxa was either poorly known or completely lacking. The recovery from Riversleigh of well-preserved basicrania of the thylacoleonid Priscileo roskellyae, the extinct phascolarctid Litokoala kutjamarpensis and the wynyardiid Namilamadeta albivenator provide an improved database of vombatiform basicranial characters that may help to elucidate the interfamilial relationships within the suborder Vombatiformes.

Although new thylacoleonid postcranial material is described in this research (Chapter 5) an analysis similar to Munson’s (1990) was not undertaken because of the limited nature of this material. Of the 100 characters employed by Munson (1990), fewer than ten character states could have been determined. 264

12.3 Methods

12.3.1 Ingroup and Outgroup Taxa

The following species were used to represent selected marsupial families as outgroup and ingroup taxa in the analysis († indicates extinct taxon): outgroup taxa - Barinya wangala † (Dasyuridae), Echymipera kalubu (Peramelidae); ingroup taxa - Bulungamaya sp.† (Potoroidae), Caudipilosus vellicodens†, Spilocuscus maculatus (Phalangeridae), Pseudochirops cupreus (Pseudocheiridae), Cercartetus caudatus (Burramyidae), Petaurus breviceps (Petauridae), Nimiokoala greystanesi† and Litokoala kutjamarpensis† (including specimens formerly identified as L. kanunkaensis†- see Louys et al. in press; Phascolarctidae), Vombatus ursinus (Vombatidae), Priscileo roskellyae (Thylacoleonidae†), Kuterintja ngama (Ilariidae†), Namilamadeta albivenator (Wynyardiidae†), Marada arcanum (Maradidae†), Propalorchestes novaculacephalus (Palorchestidae†) and Nimbadon lavarackorum (Diprotodontidae†). In comparison to species of Wakaleo and particularly those of Thylacoleo, the cranium of P. roskellyae tends to show fewer autapomorphic characters that may mask relationships, hence the latter taxon was chosen to represent thylacoleonids in the analysis.

Specimens of the above taxa examined for the character analysis are listed in Appendix 4. Where specimens were not available, character states were determined from published descriptions for the following species: Propalorchestes novaculacephalus (Murray 1986, 1990; Black 2006), Barinya wangala (Wroe 1999) and Namilamadeta albivenator (Pledge 2005). A total of 68 dental and cranial characters were scored for the interfamilial analysis. Descriptions of these characters follow. Characters employed in previous phylogenetic studies of extinct and extant marsupials were evaluated for their applicability in this analysis. Characters included from existing analyses are cited and the character state used by those authors is indicated by “ch. xx”. (Note: that characters in Marshall et al. 1990 did not have a character number). Some previously used characters were modified; for example additional character states were recognised or separate characters were combined. 265

12.3.2 Interfamily Character Selection and Description

Dentary and dental characters 1. length of the dentary symphysis: The symphysis of Echymipera is relatively short, extending posteriorly to P1and those in Bulungamaya, Barinya, Marada, Caudipilosus, Petaurus and Cercartetus are also relatively short, terminating anterior to P3. The symphysis in Spilocuscus terminates below P3. In Priscileo, Pseudochirops, Namilamadeta, Kuterintja, Propalorchestes and Nimbadon the dentary symphysis is relatively long, extending to at least below P3/M1. In Vombatus the symphysis is extremely long extending to M3. The length of the symphysis is unknown for Nimiokoala and Litokoala. 0 = short (terminating anterior to or below the anterior half of P3); 1 = long.

2. masseteric foramen: A masseteric foramen is absent in Barinya, Echymipera, Caudipilosus, Spilocuscus and Cercartetus. A small masseteric foramen is present in Pseudochirops, Petaurus, Priscileo, Namilamadeta, Marada, and Nimbadon. A large masseteric foramen is present in Bulungamaya and in Vombatus. This character cannot be scored for Litokoala, Nimiokoala and Kuterintja. 0 = masseteric foramen absent; 1 = masseteric foramen present.

3. masseteric canal and dental canal in dentary: In Bulungamaya, the masseteric canal opens into the dental canal. In all other taxa the masseteric canal does not open into the dental canal. (Horovitz et al. 2003, ch.207). 0 = masseteric canal does not open into dental canal; 1 = masseteric canal opens into dental canal.

4. ascending ramus: The ascending ramus of Barinya, Echymipera, Pseudochirops, Namilamadeta, Kuterintja and Nimbadon form an angle to the horizontal ramus of 70o or greater. The basal portion of the coronoid crest of Propalorchestes is relatively steep, suggesting its ramus was also steeply inclined. The ascending rami of Caudipilosus (60o), Spilocuscus (60o), Cercartetus (50o), Petaurus (40o), Bulungamaya, (60o), Marada (50o) and Vombatus (65o) have an inclination of less 266 than 70o. The ascending ramus is not preserved for Priscileo, Litokoala, and Nimiokoala. 0 = ascending ramus >70 o; 1 = ascending ramus < 70o.

5. location of the mental foramen: The mental foramen in Barinya and Echymipera is low down and relatively anterior in the dentary (below P1). In Marada, the mental foramen is also located anteriorly but high in the dentary. In Bulungamaya, Caudipilosus, Spilocuscus, Kuterintja, Vombatus, Namilamadeta, and Nimbadon the mental foramen is slightly anterior to P3 and high in the dentary. In Priscileo and Nimiokoala this foramen is low in the ramus near the anterior margin of P3. In Cercartetus, Pseudochirops and Petaurus, this foramen is located ventral to the posterior root of P3. The mental foramen is not preserved in Litokoala and Propalorchestes. The more posterior location of this foramen in diprotodontians is probably related to the enlargement of the root of the procumbent incisor. 0 = anterior to P2 and low in dentary; 1 = well anterior to P3 and high in dentary; 2 = anteroventral to P3 and low in dentary; 3 = anteroventral to P3 and high in dentary; 4 = ventral to P3 and low in dentary.

6. number of lower incisors: Barinya and Echymipera exhibit three lower incisors in the dentary, i.e. the polyprotodont condition. All remaining ingroup and outgroup taxa exhibit less than three incisors, predominantly only a single procumbent incisor (i.e. the diprotodont condition). In Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, Cercartetus and Priscileo, the homology of the tooth immediately posterior to the procumbent incisor is not clear; it is either an incisor or a canine (in Caudipilosus this tooth may be absent or incipient). Horovitz et al. (2003) coded this tooth in Cercartetus as a reduced canine rather than an incisor and similar coding is employed here for these taxa (see character 10). (Marshall et al. 1990; Springer et al. 1997 ch.2; Horovitz et al. 2003 ch.150). 0 = three; 1 = less than three.

7. number of upper incisors*: Barinya and Echymipera exhibit four upper incisors. Bulungamaya, Caudipilosus, Spilocuscus, Cercartetus, Petaurus, Pseudochirops, Litokoala, Nimiokoala, Priscileo, Namilamadeta, and Nimbadon possess three upper incisors. In Vombatus, there is a single upper incisor. The condition for Kuterintja, 267

Marada, and Propalorchestes is not known. (Springer et al. 1997 ch.1; Horovitz et al. 2003 ch.151). 0 = four; 1 = three; 2 = one.

8. procumbent lower incisor: the single lower incisor present in all ingroup taxa is procumbent. (The identity of this tooth is not clear; it is often simply identified as I1, for example in Archer and Clayton p.699 (1984), however, Murray et al. (1987) identify it as I3 in phalangerids, while Springer et al. (1997) label it as I2). None of the lower incisors in Echymipera are procumbent. ( Marshall et al. 1990; Springer et al. 1997 ch.33; Horowitz et al. 2003 ch.167). 0 = absent; 1 = present.

9. lower canine: Barinya and Echymipera possess a well developed lower canine. Priscileo, Cercartetus, Caudipilosus, Spilocuscus and Petaurus exhibit a small tooth immediately posterior to the procumbent incisor, which is here regarded as a reduced canine rather than an incisor, following Horovitz et al. (2003). C1 is absent in Bulungamaya, Pseudochirops, Litokoala, Nimiokoala, Namilamadeta, Kuterintja, Marada, Vombatus, Propalorchestes and Nimbadon. In Pseudochirops, the single button-like tooth that lies in the extremely shortened ramus between the procumbent incisor and P3 is here regarded as a premolar primarily on the basis that if the ramus is shortened and the root of the anterior incisor is enlarged, loss of the more-anterior teeth seems more likely. (Springer et al. 1997, ch.39; Horovitz et al. 2003, ch.169). 0 = present; 1 = reduced or absent.

10. upper canine: Where this character can be scored, all outgroup and ingroup taxa possess an upper canine except Vombatus and Nimbadon. The condition for Kuterintja, Propalorchestes and Marada is not known. (modified; Springer et al. 1997, ch.39; Horovitz et al. 2003, ch.170). 0 = present; 1 = absent.

11. lower premolar number*: Barinya and Echymipera possess three double-rooted premolars. Caudipilosus, Cercartetus, Petaurus and Priscileo possess three premolars, but only P3 has two roots (P1 and P2 in Caudipilosus and Cercartetus are very rudimentary). Spilocuscus and Pseudochirops possess two premolars. Bulungamaya, Litokoala, Nimiokoala, Namilamadeta, Kuterintja, Marada, 268

Vombatus, Propalorchestes and Nimbadon all possess a single premolar, P3. 0 = three premolars (all double-rooted); 1 = three premolars, only P3 double-rooted; 2 = two premolars; 3 = one premolar.

12. upper premolar number*: The presence of three upper premolars is regarded as the plesiomorphic condition and is seen in Barinya, Echymipera, Priscileo, Petaurus, Pseudochirops and Cercartetus. Caudipilosus and Spilocuscus show loss of P2 (and sometimes P1). Bulungamaya, Litokoala, Nimiokoala, Namilamadeta, Vombatus, Propalorchestes and Nimbadon, show loss of P1-2. The condition for Kuterintja and Marada is unknown. 0 = three premolars; 1 = two premolars; 2 = one premolar.

13. P3 length: cheektooth row length: Priscileo and Bulungamaya have a P3 that is relatively long in comparison to the total cheektooth row. In all other taxa, P3 is relatively small in comparison to the whole cheektooth row. 0 = P3 small relative to tooth row (less than ¼ of the total cheektooth row length); 1 = P3 large relative to tooth row.

14. P3 buccally deflected: P3 of Caudipilosus and Spilocuscus is deflected buccally. In all other taxa, P3 is aligned longitudinally along the horizontal ramus. (Marshall et al.1990). 0 = no deflection; 1 = buccal deflection.

15. P3 serrated: P3 of Caudipilosus, Spilocuscus, Cercartetus and Bulungamaya is serrated. All other ingroup and outgroup taxa lack serrations on P3. Springer et al. (1997) combined this character with P3 length and coded Burramyidae as lacking serrations (contra this study). Specimens of Cercartetus caudatus examined in this analysis exhibited a small number of weak serrations (2-3) on the apical region of P3. It is likely that these small ridges are obliterated with wear. (Marshall et al. 1990; Springer et al. 1997, ch.31, modified). 0 = no serrations; 1 = serrations present.

16. P3 crown structure: Barinya and Echymipera possess a bicuspid P3 in which the anterior cusp is much taller than the posterior cusp, the latter being extremely reduced and located at the posterior edge of the premolar. Vombatus (juvenile, pre- 269 wear state), Petaurus, Marada and Nimbadon exhibit a similar structure. P3 of Pseudochirops, Nimiokoala, Litokoala and Kuterintja are multicusped. P3 of Caudipilosus, Spilocuscus and Cercartetus is that of a short blade. P3 of Priscileo, Namilamadeta and Bulungamaya forms a long blade. P3 of Propalorchestes is not preserved. 0 = bicuspid; 1 = multicusped; 2 = short blade; 3 = long blade.

17. development of a posterior cingulum on P3: Barinya, Echymipera, Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus, Petaurus, phascolarctids, Priscileo, Vombatus and Namilamadeta lack a posterior cingulum on P3. Marada and Nimbadon exhibit a well developed posterior cingulum while Kuterintja ngama exhibits a very weak cingulum (contra Myers 1994). The condition of this character in Propalorchestes novaculacephalus is not known. (Myers 1994, ch.19). 0 = absent; 1 = present.

18. development of a posterolingual crest on P3: Priscileo exhibits a well developed crest that swings lingually from the end of the longitudinal blade and then anteriorly. A much smaller posterolingual crest is present in Namilamadeta and Litokoala. A posterolingual crest is absent on P3 of all remaining ingroup and outgroup taxa (except Propalorchestes where P3 is not known). 0 = absent; 1 = present.

19. development of an anterobuccal crest on P3: An anterobuccal crest is present on P3 of Caudipilosus, Spilocuscus, Priscileo, Namilamadeta, Nimiokoala and Kuterintja. An anterobuccal crest is absent in Barinya, Echymipera, Bulungamaya, Pseudochirops, Cercartetus, Petaurus, Marada, Vombatus, and Nimbadon. It is unknown for Litokoala and Propalorchestes. 0 = absent; 1 = present.

20. development of an anterolingual crest on P3: a well-developed crest is present in Litokoala, Priscileo, Namilamadeta and Kuterintja. Nimbadon exhibits a very short and weakly developed anterolingual crest. This crest is absent in Barinya, Echymipera, Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, 270

Marada and Nimiokoala. Presence of this crest could not be determined for Vombatus and is unknown for Propalorchestes. 0 = absent; 1 = present.

21. development of a posterolingual cusp/crest on P3*: A posterolingual cusp is present in Litokoala, Nimiokoala, Kuterintja and Namilamadeta. Bulungamaya exhibits a posterolingual crest that terminates at a cuspule at the base of the crown. Nimbadon and Propalorchestes exhibit a large lingual cusp that is positioned more anteriorly, approximately midcrown. Pseudochirops exhibits a small cuspule at the midcrown level and also a small crest near the posterolingual corner of the tooth. A posterolingual cusp is absent in Barinya, Echymipera, Caudipilosus, Spilocuscus, Petaurus, Cercartetus and Priscileo. Presence of this cusp is unknown in Vombatus and Marada. 0 = absent; 1 = posterolingual; 2 = posterolingual, connected to posterolingual crest; 3 = mid-lingual.

22. development of an anterolingual crest on P3: An anterolingual crest is present on P3 of Priscileo, Litokoala, Nimiokoala, Namilamadeta, Kuterintja, Spilocuscus and Pseudochirops. This crest is absent in all other taxa, except Marada and Vombatus for which it cannot be determined. 0 = absent; 1 = present.

23. development of an anterobuccal crest on P3: An anterobuccal crest is present on P3 of Caudipilosus, Spilocuscus, Pseudochirops, Nimiokoala, Priscileo, Kuterintja and Namilamadeta. It is absent in Barinya, Echymipera, Bulungamaya, Cercartetus, Petaurus, Propalorchestes and Nimbadon and is unknown for Vombatus and Marada. 0 = absent; 1 = present.

24. molar gradient: the molar gradient in Barinya, Echymipera, Caudipilosus, Spilocuscus, Pseudochirops, phascolarctids, Namilamadeta, Kuterintja, Marada and Vombatus decreases slightly posteriorly, largely as a result of a slight decrease in the size of M4. Priscileo, Petaurus and Cercartetus display a much greater progressive reduction in the size of the posterior molars. Propalorchestes and Nimbadon exhibit an increase in size of the posterior molar (possibly reflecting an increase in the abrasiveness of the diet; Black, in press). (Marshall et al. 1990; Springer et al. 1997, 271 ch.36, modified). 0 = slight posterior decrease (mainly M4); 1 = large posterior decrease; 2 = posterior increase.

25. molar crown structure: Barinya and Echymipera exhibit a plesiomorphic, tribosphenic crown structure. Phascolarctids, Pseudochirops and Kuterintja exhibit a selenodont crown structure, as does Vombatus; however, in the latter taxon, this pattern is only evident in juveniles and is obliterated by wear, early in life. The cusps in Kuterintja are aligned to form two transverse rows and indicate a trend towards bilophodonty (see Myers 1994). The crowns of Bulungamaya, phalangerids, Namilamadeta, Marada, Propalorchestes and Nimbadon have a basic bi-lophodont structure. The lophs in phalangerids are relatively low and are regarded as bunolophodont. Those of Namilamadeta and Marada are moderately tall, and those of Bulungamaya, Nimbadon and Propalorchestes are well developed. Cercartetus, Petaurus and Priscileo exhibit a bunodont crown structure. (Springer et al. 1997, ch. 42 & ch. 43; Horovitz et al. 2003, ch. 173 & ch. 174). 0 = tribosphenic; 1 = selenodont; 2 = bunodont; 3 = bunolophodont; 4 = bilophodont.

26. development of a paraconid on M 2-4: Barinya and Echymipera exhibit a well developed paraconid on lower molars. A paraconid is also present on molars of juvenile Vombatus. This cusp is reduced or absent in Bulungamaya, Cercartetus, Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, Litokoala, Nimiokoala, Kuterintja, Priscileo, Propalorchestes, Nimbadon and Marada. (Myers 1994, ch.7; Springer et al. 1997, ch.15; Horovitz et al. 2003, ch.159). 0 = present; 1 = reduction or loss.

27. development of the entoconid and hypoconid on M1: the entoconid and hypoconid are relatively short (in comparison to the metaconid and protoconid) in Barinya, Petaurus and the juvenile Vombatus. In all other taxa except Priscileo they are moderately or well-developed. In Priscileo, these cusps are very short. 0 = short; 1 = moderate to well-developed; 2 = poorly developed.

28. M2 intersection of cristid obliqua with trigonid: in Barinya, Echymipera, Pseudochirops, Litokoala, Nimiokoala, Kuterintja and Vombatus, the cristid obliqua 272 is angled lingually and intersects the trigonid on the lingual edge of the protoconid. In Bulungamaya, Caudipilosus and Spilocuscus the cristid obliqua is kinked, initially running lingually and then anteriorly. In Bulungamaya it intersects the posterolingual edge of the protoconid and in the phalangerids it joins a short postprotocristid. In Petaurus and Cercartetus the cristid obliqua is short, lingually-directed and joins the postprotocristid to form a v-shaped crest. In Namilamadeta, Marada and Propalorchestes the cristid obliqua is lingually offset from the hypoconid and runs anteriorly to connect with a postprotocristid to form a mid-link. In Priscileo, the cristid obliqua is anteriorly-directed but is located buccally and connects with the posteriorly-directed postprotocristid. A cristid obliqua is absent in Nimbadon. (Springer et al. 1997 ch.23, modified; Horovitz et al. 2003, ch.160). 0 = lingually- directed, meets lingual edge of protoconid; 1= runs lingually, then anteriorly, meets posterolingual edge of protoconid; 2 = runs lingually then anteriorly, joins postprotocristid; 3 = lingually-directed, joins postprotocristid; 4 = anteriorly- directed, joins postprotocristid; 5 = forms midlink; 6 = absent.

29. M1 stylar cusp development: stylar cusps are moderately developed on M1 in Barinya and Echymipera. Stylar cusps are small and extremely reduced in Priscileo and phascolarctids, and lost in Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus, Petaurus, Propalorchestes and Nimbadon. Stylar cusps are well developed in Namilamadeta and Kuterintja. (Murray 1990, postulates that stylar cusps in upper molars of palorchestids and diprotodontids are not reduced but are incorporated into the buccal side of the lophs; see character 38, below). The condition could not be determined for Vombatus and is unknown for Marada. 0 = moderately developed; 1 = well developed; 2 = small; 3 = loss. 30. M1 stylar shelf: the stylar shelf of M1 is broad in Barinya, Echymipera, Litokoala, Nimiokoala, Priscileo, Namilamadeta, Kuterintja and Vombatus. The stylar shelf is narrowed or lost in Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus, Petaurus, Propalorchestes and Nimbadon. 0 = broad; 1 = narrowed or lost.

31. M1 occlusal outline*: the outline of M1 is triangular in Barinya, semi-square in Echymipera, Bulungamaya, Pseudochirops, Petaurus and Priscileo, and square in 273

Cercartetus, Litokoala, Nimiokoala and Kuterintja. M1 in Caudipilosus, Spilocuscus, Namilamadeta, Vombatus, Propalorchestes and Nimbadon is rectangular. 0 = triangular; 1 = semi-square; 2 = square; 3 = rectangular.

32. M2 occlusal outline*: the occlusal outline of M2 in Barinya is triangular while those of Echymipera, Bulungamaya, Cercartetus, Petaurus, Priscileo, Pseudochirops, Litokoala and Nimiokoala are square or subsquare in outline. M2 of Caudipilosus, Spilocuscus, Namilamadeta, Kuterintja, Vombatus, Propalorchestes and Nimbadon are rectangular. M2 is not known for Marada. (Horovitz et al. 2003, ch.153, modified). 0 = triangular; 1= semi-square or square; 2 = rectangular.

33. M3 occlusal outline*: the third upper molar of Barinya, Priscileo and Cercartetus is triangular. Echymipera, Pseudochirops and Petaurus have semi- square molars while that of Litokoala is square. Bulungamaya, Caudipilosus, Spilocuscus, Namilamadeta, Kuterintja, Vombatus, Propalorchestes and Nimbadon all exhibit a rectangular M3. M3 for Marada is not known. 0 = triangular; 1 = semi- square; 2 = square; 3 = rectangular.

34. M1 paracone and metacone placement*: the paracone and metacone on M1 are located relatively lingually in Barinya and Echymipera. In Pseudochirops, Litokoala, Nimiokoala, Kuterintja, Priscileo and Namilamadeta these cusps have a slightly more buccal location. In Bulungamaya, Caudipilosus, Spilocuscus, Cercartetus, Petaurus, Propalorchestes and Nimbadon the paracone and metacone are located near the buccal edge of the crown. (Springer et al. 1997, ch 6; Horovitz et al. 2003, ch.154). 0 = lingual; 1 = buccal; 2 = buccal margin.

35. M2 paracone and metacone placement: the paracone and metacone on M2 have a relatively lingual location in Barinya, Echymipera, Litokoala, Nimiokoala, Kuterintja and Namilamadeta. In Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, Cercartetus, Priscileo, Propalorchestes and Nimbadon the paracone and metacone have a more buccal location. Cusp placement could not be determined for Vombatus and Marada could not be scored for this character. 274

(Springer et al. 1997, ch.6; Horovitz et al. 2003, ch.154, modified). 0 = lingual; 1 = buccal.

36. M1 paracone versus metacone size*: The paracone is shorter than the metacone in Barinya and Echymipera. Diprotodontian taxa show an increase in paracone height such that the paracone and metacone are approximately equal in height in Bulungamaya, Caudipilosus, Spilocuscus, Cercartetus, Petaurus, Litokoala, Nimiokoala, Kuterintja, Propalorchestes and Nimbadon. In juvenile Vombatus, the anterior occlusal surface of M1 is lower in height than the posterior surface suggesting that the paracone is lower than the metacone (individual cusps could not be discerned). In Pseudochirops, the paracone is taller than the metacone, while in Priscileo the paracone is much taller than the metacone. This character is not known for Marada. (Springer et al. 1997, ch.7; Horovitz et al. 2003, ch.155). 0 = paracone < metacone; 1 = paracone = metacone; 2 = paracone > metacone.

37. M2 paracone versus metacone size*: the paracone is shorter than the metacone in Barinya and Echymipera. Paracone and metacone are similar in height in Bulungamaya, Petaurus, Propalorchestes, Nimbadon and Vombatus (juvenile). In Caudipilosus, Spilocuscus, Cercartetus, Pseudochirops, Litokoala, Nimiokoala, Kuterintja and Namilamadeta the paracone is slightly taller than the metacone. In Priscileo the paracone is much taller than the metacone. This character is unknown for Marada. (Springer et al. 1997, ch.7). 0 = paracone < metacone; 1 = paracone = metacone; 2 = paracone > metacone.

38. M3 paracone versus metacone size: the paracone is smaller than the metacone in Barinya and Echymipera. Diprotodontians tend to show an increase in the height of the paracone relative to the metacone. These cusps are of approximately similar height in Bulungamaya, Vombatus, Propalorchestes and Nimbadon. The paracone is taller than the metacone on M3 in Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus, Petaurus, Litokoala, Kuterintja, Nimbadon and Namilamadeta. This character is unknown for Priscileo and Marada. (Springer et al. 1997, ch.7). 0 = paracone < metacone; 1 = paracone = metacone; 2 = paracone > metacone. 275

39. M1 paraconule: a paraconule is absent in Barinya, Echymipera, Bulungamaya, Spilocuscus, Cercartetus, Namilamadeta, Kuterintja, Propalorchestes and Nimbadon. A paraconule is present on M1 in Caudipilosus, Pseudochirops, Litokoala, Nimiokoala and Priscileo. This cuspule was obvious in some specimens of Petaurus and difficult to discern in others, possibly as a result of wear. Presence or absence of this character could not be determined for Vombatus or Marada. (Argumentation herein is contra Myers (1994, ch.20). 0 = absent; 1 = present.

Cranial characters All cranial characters are unknown for Kuterintja and Marada and so are scored as a missing data (?) in the data matrix.

40. supraorbital process: A supraorbital process is absent in Echymipera, Caudipilosus, Spilocuscus, Pseudochirops and Cercartetus. A small supraorbital process is present in Petaurus, phascolarctids, Namilamadeta and Nimbadon. The supraorbital process is well developed in Barinya, Priscileo and Vombatus. 0 = absent; 1 = present.

41. parietal width: The parietal is broad posteriorly in Barinya, Echymipera, Bulungamaya, Cercartetus, Petaurus, Litokoala, Priscileo, Namilamadeta and Vombatus. In Caudipilosus, Spilocuscus, Pseudochirops, Propalorchestes and Nimbadon the parietal narrows posteriorly and the squamosal becomes broader. 0 = broad; 1 = narrowed.

42. frontal/squamosal contact: The frontal extends posteriorly to make a relatively broad contact with the squamosal in Echymipera, Petaurus, Namilamadeta and Nimbadon. This is also the predominant condition in Vombatus, however, the area of contact is smaller in this taxon and may even be absent; 20% of Vombatus skulls (N = 20) observed in the Australian Museum mammal collection lacked frontal/ squamosal contact. A smaller proportion of the Museum’s Lasiorhinus skulls (N = 10) lacked this contact (10%). In this region the four bony elements (parietal, squamosal, alisphenoid and frontal) converge and in some cases the alisphenoid 276 extends dorsally as a thin sliver, separating the squamosal from the frontal by only a few millimetres. It is obvious that only slight variations in the growth of these elements may lead to contact or non-contact. Because frontal/squamosal contact is significantly more prevalent within vombatids, it is coded for this character rather than polymorphic. The alisphenoid extends dorsally and prevents frontal/squamosal contact in Barinya, Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus, Nimiokoala and Priscileo. Murray (1986) indicates that the sutures in this region on the skull of Propalorchestes novaculacephalus are not clear and hence, this character could not be scored for this species. (Springer et al.1997, ch.44; Horovitz et al. 2003, ch.176). 0 = frontal/squamosal contact absent; 1 = frontal/squamosal contact present.

43. nuchal crest development: a well-developed nuchal crest is present in Barinya, Echymipera, Priscileo and Namilamadeta. A moderately low crest is present in Bulungamaya, Spilocuscus, Litokoala, Vombatus, Propalorchestes and Nimbadon. A nuchal crest is absent in Caudipilosus, Pseudochirops, Petaurus and Cercartetus. 0 = well developed; 1 = moderately low; 2 = absent.

44. height of occiput: The occiput (nuchal crest to foramen magnum) is tall relative to its width in Barinya, Echymipera, Namilamadeta, Vombatus, Propalorchestes, and Nimbadon. The occiput is moderately tall in Litokoala and Priscileo and is short in Caudipilosus, Spilocuscus, Pseudochirops, Petaurus and Cercartetus. 0 = tall; 1 = short. 45. interparietal on the dorsum of the braincase: An interparietal is visible on the dorsum of the skull in Barinya, Bulungamaya, Pseudochirops, Petaurus, Cercartetus and phascolarctids. An interparietal is not present Echymipera, Caudipilosus, Spilocuscus (sometimes visible in juveniles in which the sagittal crests have not yet merged in the midline), Priscileo, Namilamadeta, Vombatus, Propalorchestes and Nimbadon. 0 = present; 1 = absent.

46. presence of a masseteric process: A masseteric process is absent in Barinya, Echymipera, Bulungamaya, Caudipilosus, Cercartetus, Petaurus, Priscileo and Vombatus. Spilocuscus, Pseudochirops, Nimiokoala and Litokoala exhibit a small 277 masseteric process while Namilamadeta, Kuterintja and Nimbadon exhibit large processes. The zygomatic root on the maxilla fragment of Propalorchestes (QM F50605) projects ventrally, suggesting that a masseteric process was present in this species. 0 = absent; 1 = present.

47. fusion of ectotympanic with other bones of skull: The ectotympanic is fused to surrounding bones in Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus, Petaurus and Bulungamaya. In all remaining taxa the ectotympanic is unfused. (Springer & Woodburne 1989, ch.10; Springer et al. 1997, ch.68; Horovitz et al. 2003, ch.189).0 = unfused; 1 = fused.

48. glenoid fossa: the glenoid fossa in Barinya and Echymipera is concave. In Bulungamaya, Petaurus and Cercartetus, the articular eminence is flat and the mandibular fossa is very shallow. In Caudipilosus, the articular eminence is also flat but the mandibular fossa is moderately developed. In Pseudochirops, Spilocuscus, Litokoala and Nimbadon the articular eminence is gently convex and the mandibular fossa is moderately developed. The articular eminence of Namilamadeta is slightly flatter than the former species and its mandibular fossa is very shallow. This region is damaged in Priscileo but appears to be similar to Namilamadeta. The glenoid fossa in Propalorchestes consists of a large concave articular eminence and a broad, deep mandibular groove (see Murray 1986). 0 = simple concave fossa; 1 = flat articular eminence, shallow mandibular fossa; 2 = flat articular eminence, moderate mandibular fossa; 3 = slightly convex articular eminence, shallow mandibular fossa; 4 = gently convex articular eminence, moderate mandibular fossa; 5 = concave articular eminence, deep mandibular fossa; 6 = ridge-like articular eminence only.

49. medial glenoid process: In Priscileo, the medial end of the articular eminence is developed into a small anterolaterally-directed ridge, the medial glenoid process. The alisphenosquamosal suture courses through this process. This process is also present in Namilamadeta, Vombatus, Propalorchestes and Nimbadon. A structure very similar to that in Nimbadon is present in Spilocuscus (contra Aplin 1987). This process is absent in Barinya, Echymipera, Bulungamaya, Caudipilosus, Pseudochirops, Petaurus and Litokoala (the condition in Nimiokoala is unknown). 278

0 = absent; 1 = present.

50. position of the postglenoid foramen: the postglenoid foramen is located posteromedial to the postglenoid process in Barinya and Echymipera. In Caudipilosus, Spilocuscus, Pseudochirops, Petaurus and Cercartetus the postglenoid foramen is located anteromedial to the postglenoid process. In Bulungamaya and Litokoala (the condition is unknown for Nimiokoala) the foramen is located medial to the postglenoid process. In Priscileo, Namilamadeta, Vombatus and Nimbadon the foramen is located at the medial end of the postglenoid cavity. The condition in Propalorchestes is unknown. (Horovitz et al. 2003, ch.191, modified). 0 = posteromedial to postglenoid process; 1 = medial to postglenoid process; 2 = anteromedial to postglenoid process; 3 = within postglenoid cavity.

51. development of the postglenoid process: the postglenoid process in Barinya is moderately developed but relatively narrow. This process in both Echymipera and Bulungamaya, and Cercartetus is also relatively narrow but quite short. In Caudipilosus, Spilocuscus, Pseudochirops and Petaurus the process is broader but still relatively short. The process in Priscileo, Propalorchestes and Nimbadon is broad and moderately developed while that of Litokoala and Namilamadeta is well developed. The process in Vombatus is highly modified and is reduced to a low transverse bar. A narrow postglenoid process is deemed plesiomorphic. (Springer et al. 1997, ch.65; Horovitz et al. 2003, ch.187; modified). 0 = absent; 1 = narrow; 2 = broad and short; 3 = broad and moderately long; 4 = broad and long.

52. postglenoid cavity: the postglenoid cavity in Barinya and Echymipera is small, narrow and directed ventrally. In Bulungamaya the cavity lies on the medial half of the postglenoid process (pgp), is moderate in size and opens anteriorly, connecting with the supraglenoid region of the squamosal. In Nimbadon and Propalorchestes the cavity also lies at the medial margin of the pgp but is greatly enlarged, forming an epitympanic fenestra. In Litokoala, Priscileo, Namilamadeta and Vombatus the postglenoid cavity extends laterally across most of the postglenoid structure. Within these taxa the cavity varies in depth: in Litokoala the whole cavity is deeply excavated while in Priscileo, Namilamadeta and Vombatus the cavity becomes 279 shallower laterally. A postglenoid cavity is absent in Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus and Petaurus. 0 = absent; 1 = small, ventrally-directed; 2 = moderate, medial margin of postglenoid process; 3 = large, medial margin of pgp; 4 = extends laterally across pgp, cavity is deep; 5 = extends laterally across pgp, shallow.

53. median occipital crest: a vertical crest is present in the centre of the occiput in Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, Litokoala, Priscileo and Namilamadeta. This crest is absent in Barinya, Echymipera, and Cercartetus. The presence of this crest is variable in Vombatus and Nimbadon and its condition is unknown for Propalorchestes. 0 = absent; 1 = present.

54. tympanic cavity wall components*: the roof and floor of the tympanic cavity are formed by the alisphenoid in Barinya and Echymipera. In Bulungamaya (QM F41237 used to determine this character), Caudipilosus and Spilocuscus the squamosal forms the roof and the alisphenoid the walls and floor. In Litokoala, Pseudochirops, Cercartetus and Petaurus the alisphenoid is anteriorly inflated resulting in the anterior tympanic chamber being roofed by both the squamosal and alisphenoid. In these taxa, the tympanic cavity is floored by the alisphenoid. In Namilamadeta, the squamosal forms the roof and anterior wall of the cavity, and the alisphenoid and squamosal form a small bilaminar floor. In Propalorchestes, the squamosal forms the roof but the components of the floor are not certain (see Murray 1986). In Nimbadon, the squamosal forms the roof but the contribution of the squamosal and alisphenoid to the floor is variable; the tympanic wing may be bilaminar or consist solely of the alisphenoid. In Priscileo, the squamosal contributes to the roof and anterolateral wall and the alisphenoid forms the anteromedial wall and the floor. In Vombatus, the squamosal forms the floor and the roof. (The condition is unknown for Nimiokoala). 0 = alisphenoid roof and floor; 1 = squamosal and alisphenoid roof, alisphenoid floor; 2 = squamosal roof, alisphenoid floor; 3 = squamosal roof, alisphenoid and squamosal floor; 4 = squamosal roof, squamosal floor. 280

55. alisphenoid tympanic wing: the alisphenoid tympanic wing is developed into a bulla that is open posteriorly in Barinya and Litokoala. In Echymipera, the bulla is laterally reduced. In Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Petaurus and Cercartetus, the alisphenoid tympanic wing extends posteriorly to contact the paroccipital process. In Priscileo and Namilamadeta the wing is relatively broad and short, while in Nimbadon the wing is narrow and short. The condition for Nimiokoala and Propalorchestes is unknown. Vombatus lacks an alisphenoid tympanic wing. (Springer et al. 1997, ch.62; Horovitz et al. 2003, ch.185, modified). 0 = well-developed bulla, open posteriorly; 1 = extends posteriorly to paroccipital process; 2 = broad, short wing; 3 = narrow, short wing; 4 = absent.

56. alisphenoid ventrolateral wing: the ventrolateral wing is very broad in Barinya, Echymipera, Bulungamaya, Pseudochirops, Cercartetus, Litokoala, Priscileo and Namilamadeta. It is relatively narrow in Caudipilosus, Spilocuscus, Petaurus, Vombatus, Nimbadon and Propalorchestes. 0 = broad; 1 = narrowed.

57. rostral tympanic process of the periotic: the periotic of Barinya bears a prominent rostral tympanic process that extends anteriorly to the anterior pole. A well-developed rounded process is present in Echymipera, Bulungamaya, Petaurus and Litokoala. A small process is present in Caudipilosus. A rostral tympanic process is either absent or very weakly developed in Cercartetus, Priscileo, Namilamadeta and Vombatus. The condition in Pseudochirops could not be determined. 0 = process extends to anterior pole; 1 = well developed posterior process; 2 = process very weak or absent.

58. subsquamosal foramen: a subsquamosal foramen is present in Barinya, Echymipera, Litokoala, Priscileo, Nimbadon and Propalorchestes, Bulungamaya and Namilamadeta. A subsquamosal foramen is absent in Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, Cercartetus and Vombatus. (Springer and Woodburne 1989, ch.12). 0 = present; 1 = absent. 281

Table 12.1 Taxa character state matrix for interfamily parsimony analysis (? = character unknown; - = character not applicable) 1 2 3 4 5 6 7 8 9 1011121314151617181920212223242526272829303132 Barinya 00? 00000000000000000000000000000 Echymipera 00010000000000000000000000100011 Bulungamaya. 01103111103210130000200041113111 Caudipilosus 00003111101101120010001031123132 Spilocuscus 00003111102101120010011031123132 Pseudochirops 11104111102000010000011010/1102111 Cercartetus 00004111101000120000000121133121 Petaurus 01014111101000000000000121033111 Litokoala ? ? 0? ? ? 1? ? 0? 200010111111011102021 Nimiokoala ? 1002111103200010010111011102021 Priscileo 110? 2111101010030111011121242011 Namilamadeta 11013111103200030111111041151032 Kuterintja 1? 0131? 11? 3? 00011011111011101022 Marada 010011? 11? 3? 00001000? ? ? 04115? ? ? ? Vombatus 1101312111320000100100021000? 032 Propalorchestes 1? 01? ? ? ? ? ? 32000? ? ? ? ? 300241153132 Nimbadon 11013111113200001001300241163132

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 Barinya 00000001000000000011000000000000 Echymipera 10000000010000000011002010001010 Bulungamaya. 321111000010001101121211? 0010100 Caudipilosus 32112210102110120220121121112001 Spilocuscus 32122200101121141220121131113100 Pseudochirops 111222101021211302201110? 1112001 Cercartetus 02112200002110110210011031112011 Petaurus 12112211012110110220011111112001 Litokoala 210122110? 1011030135110000010001 Nimiokoala ?1 0 1 2 2 1 ?0 0 ?0 1 1 ?????????????????? Priscileo 01122? 11000010041334132030014101 Namilamadeta 3101220101001104134413203001012? Kuterintja 3 1 0 2 2 2 0 ??????1 ?????????????????? Marada ???????????????????????????????? Vombatus 310001? 10110200613040/144131010221 Propalorchestes 3212220? 1010? 1? 51? 33? ? ? 0? 00141? ? Nimbadon 321122011110210313330/12/33030014121 282

59. squamosal pneumatisation: the posterior part of the squamosal is pneumatised in Caudipilosus, Spilocuscus, Pseudochirops, Cercartetus and Petaurus. All other taxa lack pneumatisation. (Springer and Woodburne 1989, ch.14). 0 = not pneumatised; 1 = pneumatised.

60. medial wall of the mandibular fossa: the medial wall of the mandibular fossa is formed by the alisphenoid in Barinya and Echymipera. In the remaining taxa, the wall is formed by the squamosal. (Springer and Woodburne 1989, ch.1). 0 = alisphenoid; 1 = squamosal.

61. mastoid process: the mastoid process in Barinya, Bulungamaya, Litokoala, Namilamadeta, Vombatus, is a thin and relatively broad transverse process. The region of the mastoid process in Echymipera is relatively narrow and does not form a narrow transverse process, instead forming a smooth surface. The process in Caudipilosus, Pseudochirops, Petaurus and Cercartetus is anteroposteriorly deep and rounded. Spilocuscus has a broad, relatively flat, transverse process. The process in Priscileo, Nimbadon and Propalorchestes is relatively deep anteroposteriorly, and is narrow anteriorly, but becomes broader posteriorly. 0 = broad, thin transverse process; 1 = mastoid narrow, process poorly developed; 2 = mastoid broad and rounded; 3 = mastoid broad, flattened, angular; 4 = mastoid anteriorly narrow, posteriorly broad, transverse process.

62. position of the foramen ovale: the foramen ovale lies anterolateral to the basisphenobasioccipital suture in Barinya, Echymipera, Caudipilosus, Pseudochirops, Cercartetus, Petaurus and Litokoala. In Bulungamaya, Spilocuscus, Priscileo, Namilamadeta, Vombatus, Propalorchestes and Nimbadon the foramen ovale lies lateral to the basisphenobasioccipital suture. The condition in Nimiokoala is unknown. (differs to Horovitz et al. 2003 where the position is described in terms of the osteological composition of the foramen, ch.194). 0 = anterolateral to basisphenobasioccipital suture; 1 = lateral to basisphenobasioccipital suture.

63. lachrymal tuberosity: a small lachrymal tuberosity is present in Barinya, Bulungamaya, Caudipilosus, Spilocuscus, Pseudochirops, Petaurus, Litokoala and 283

Barinya Echymipera Bulungamaya sp. C Marada B Caudipilosus 89 Fig. 12. 2 57 Spilocuscus D Pseudochirops The most parsimonious tree obtained A Cercartetus from analysis with character-states unordered. 50 100 Petaurus (L=224, RCI=0.34, RI=0.62) Nimiokoala E Bootstrap frequencies are indicated below nodes for Litokoala branches that were supported in >50% of replicates. F Priscileo Apomorphies for major nodes are listed below. The CI and direction of character-state changes for G Namilamadeta apomorphies are listed in Appendix 5A. H Kuterintja I Vombatus J 62 Propalorchestes K 90 Nimbadon

Apomorphies at nodes: The derived state is separated from the character number by a point (.) Character state reversals are indicated by a negative sign (-). Unequivocal character states are indicated in bold type. Node A deltran: 2.1, 6.1, 7.1, 8.1, 9.1, 11.3, 25.1, 26.1, 36.1, 37.2, 38.2, 50.1, 53.1, 54.1, 60.1 acctran: 2.1, 5.1, 6.1, 7.1, 8.1, 9.1, 11.3, 12.2, 25.1, 26.1, 29.2, 33.3, 34.1, 36.1, 37.2, 38.2, 39.1, 43.1, 89.1, 50.1, - 52.1, 53.1, 54.1, 58.3, 60.1, 64.1 Node B deltran: 29.3, 30.1, 34.2, 35.1, 47.1, 48.1, 55.1 acctran: 29.3, 30.1, 34.2, 35.1, 47.1, 55.1 Node C deltran: 25.4 acctran: 21.2, 25.4, 28.1, -37.1, -38.1, -39.0, 52.2, 54.2, 62.1, -64.0 Node D deltran: 5.4, 43.2, 44.1, 50.2, -52.0, 58.1, 59.1, 61.2, 64.1 acctran: 5.4, -11.1, -12.0, 16.2, -33.1, 43.2, 44.1, 50.2, 51.2, 57.3, 58.1, 59.1, 61.2 Node E deltran: 5.2, 12.2, 16.1, 19.1, 21.1, 22.1, 23.1 29.2, 34.1, 39.1, 46.1 acctran: 1.1, 5.2, 16.1, 19.1, 21.1, 22.1, 23.1, 31.2, 40.1, 46.1, 48.3, 51.3, 52.4 Node F deltran: 18.1, 20.1, 40.1, 51.3, 64.1 acctran: 4.1, 18.1, 20.1 Node G deltran: 1.1, 45.1, 48.4, 49.1, 50.3, 52.5, 54.3, 55.2, 57.3, 62.1 acctran: 16.3, -31.1, 36.2, -43.0, 45.1, 48.4, 49.1, 50.3, 52.5, 54.3, 55.2, 57.3, 62.1 Node H deltran: 4.1, 5.3, -29.1, 31.3, 32.2, 33.3, -39.0, 42.1, 63.2 acctran: 5.3, -29.1, 31.3, 32.2, -39.0, 42.1, 63.2 Node I deltran: 17.1, -18.0 acctran: 10.1, -16.0, 17.1, -18.0, 43.1,- 48.3, 55.3 Node J deltran: 10.1, -16.0, -19.0, -22.0, -23.0, 24.2, 43.1 acctran: -19.0, -21.0, -22.0, -23.0, 24.2, 29.3 Node K deltran: 21.3, 25.4, 29.3, 30.1, 34.2, 35.1, 41.1, -52.3, 61.4 acctran: 21.3, 25.4, 28.5, 30.1, 34.2, 35.1, 41.1, -52.3, 61.4 284

Priscileo. It is completely absent in Echymipera and Cercartetus. A large tuberosity is present in Vombatus, Namilamadeta and Nimbadon. (Horovitz et al. 2003, ch.184, modified). 0 = small; 1 = absent; 2 = large.

64. nasal/frontal contact: the nasal and frontal form a v-shaped contact in Barinya, Echymipera, Bulungamaya and Spilocuscus. In Caudipilosus, Pseudochirops, Cercartetus, Petaurus, Litokoala and Priscileo the contact forms a u - or weak w shape. 0 = v shaped; 1 = u or weak w shape.

12.3.3 Parsimony Analyses Two analyses were performed on the data matrix, one in which character states were unordered and another in which character states were ordered. Of the 64 characters states present in the matrix, the following 11 were ordered: 7, 11, 12, 21, 31, 32, 33, 34, 36, 37, and 54. The data matrix was analysed using PAUP 4.0b10 (Swofford 1998). The branch-and-bound method was used to search for optimal trees. The degree of support for the resulting groupings in each analysis was determined by bootstrap analysis (2000 replicates). Characters were optimised using the delayed (DELTRAN) and accelerated (ACCTRAN) transformation optimisation, as implemented in PAUP.

12.4 Results

Unordered character-states (Fitch analysis) Analysis of the data matrix using a branch-and-bound search resulted in a single parsimonious tree (224 steps, rescaled consistency index = 0.34, retention index = 0.62).This tree and its nodal apomorphies are shown in Fig. 12.2. This cladogram shows the diprotodontian taxa split into two major clades, one containing the ‘possum’ groups (Petauridae, Pseudocheiridae, Phalangeridae and Burramyidae), macropodoids and Marada, and the other containing the remaining vombatiform taxa. The latter clade forms a step-wise branching pattern in which the phascolarctids (Nimiokoala and Litokoala) form the basal branches. Propalorchestes and Nimbadon form a clade that lies as a sister-group to Vombatus. The ilariid, Kuterintja, is the sister-group to this trio. The wynyardiid Namilamadeta, is in turn the sister-group to 285

Barinya

Echymipera Pseudochirops B Petaurus

Cercartetus

A Caudipilosus C 100 Spilocuscus 89 Bulungamaya E Marada D Nimiokoala F Litokoala Fig. 12.3 G Priscileo The most parsimonious tree resulting from H Namilamadeta analysis with character-states ordered. I Kuterintja (L = 238 steps, RCI = 0.32, RI = 0.62) J Vombatus

K Propalorchestes Bootstrap frequencies are indicated below nodes 63 for branches that were supported in >50% of L 92 Nimbadon replicates.

Apomorphies at nodes A – L from the most parsimonious tree, ordered analysis. The derived state is separated from the character number by a point (.) Character state reversals are indicated by a negative sign (-) before the character number. Unequivocal character state changes are indicated in bold. The CI value and direction of change for apomorphies is listed in Appendix 5B.

Node A deltran: 2.1, 6.1, 7.1, 8.1, 9.1, 11.2, 25.1, 26.1, 29.2, 34.1, 36.1, 37.2, 38.2, 48.1, 54.1, 60.1 acctran: 2.1, 6.1, 5.1, 7.1, 8.1, 9.1, 11.2, 25.1, 26.1, 29.2, 30.1, 34.1, 35.1, 36.1, 37.2, 38.2, 39.1, 43.1, 47.1, 48.1, 50.1, -52.0, 53.1, 54.1, 55.1, 60.1, 64.1 Node B deltran: 5.4, 30.1, 35.1, 37.2, 43.2, 44.1, 47.1, 50.2, 51.2, -52.0, 55.1, 58.1, 59.1, 61.2, 64.1 acctran: 5.4, 43.2, 44.1, 50.2, 51.2, 58.1, 59.1, 61.2 Node C deltran: -5.3, 12.1, 14.1, 19.1, 23.1, 25.3, -28.2, 31.3, 32.1, 33.3, 41.1, 53.1, 54.2, 56.1 acctran: -5.3, 12.1, 14.1, 19.1, 23.1, -24.0, 25.3, -28.2, 31.3, 32.1, 33.3, 41.1, 45.1, 48.2, 53.1, 54.2 Node D deltran: 11.3, 12.2, 21.1, 33.2, 50.1, 53.1 acctran: 11.3, 12.2, 21.1, 33.2, 52.2, 54.2, 62.1 Node E deltran: 25.4 acctran: 21.2, 25.4, 28.1, 29.3, 33.3, 34.2, -37.1, -38.1, -39.0, -64.0 Node F deltran: 5.2, 16.1, 19.1, 22.1, 23.1, 31.2, 37.2, 39.1, 46.1 acctran: -17.0, -20.0, -22.0, -23.0, -24.0, 26.4, 29.1, -32.1, -38.1, -39.1, -40.0, -47.0, -49.1, -65.0 Node G deltran: 18.1, 20.1, 40.1, 51.3, 64.1 acctran: 4.1, 18.1, 20.1 Node H deltran:1.1, 45.1, 48.4, 49.1, 50.3, 52.5, 54.3, 55.2, 57.3, 62.1 acctran: 16.3, -43.0, 45.1, 48.4, 49.1, 50.3, 52.5, 54.3, 55.2, 57.3 Node I deltran: 4.1, 5.3, -29.1, 32.2, 33.3, -39.0, 42.1, 63.2 acctran: 5.3, -29.1, 31.3, 32.2, 33.3, -39.0, 42.1, 63.2 Node J deltran: 17.1, -18.0 acctran: 10.1, -16.0, 17.1, -18.0, 43.1,-48.3, 55.3 Node K deltran: 10.1, -16.0, -19.0, -22.0, -23.0, 24.2, 31.3, 43.1 acctran: -19.0, -22.0, -23.0, 24.2, 29.3 Node L deltran: 21.3, 25.4, 29.3, 30.1, 34.2, 35.1, 41.1, -52.3, 61.4 acctran: 21.3, 25.4, 28.5, 30.1, 34.2, 35.1, 41.1, -52.3, 61.4 286 that grouping, and the thylacoleonid, Priscileo, is in turn, the sister-group to the latter clade. Only three branches received robust bootstrap support: the large bush containing all diprotodontian taxa (100%), the Propalorchestes-Nimbadon clade (90%) and the phalangerid clade of Caudipilosus-Spilocuscus (89%). Bootstrap support greater than 50% was obtained for two more clades: for the Vombatus- Propalorchestes-Nimbadon clade (62%) and the phalangeridan clade (57%). All other branches on the tree had bootstrap support values of 50% or less. The consistency index and direction of change for apomorphies at nodes are listed in Appendix 5A.

Ordered character-states (Wagner analysis) A branch-and-bound analysis of the matrix with 11 character-state changes ordered, resulted in a single tree (238 steps, rescaled consistency index = 0.32, retention index = 0.62; Fig.12.3). The tree from the ordered analysis is basically similar to that produced by the unordered analysis in splitting the diprotodontian taxa into two major clades, one containing all the “possum” species and the other containing all the vombatiform species, but differs in including Marada and the macropodoid Bulungamaya in the vombatiform clade. The step-wise arrangement of vombatiform taxa in the ordered tree is the same as in the unordered analysis. Marada forms a clade with Bulungamaya, and that pair are the sister-group to all the remaining vombatiforms. A Caudipilosus-Trichosurus clade lies as the sister group to Cercartetus and Petaurus is a sister-group to that trio. Pseudochirops is a sister- group to all other “possum” species. The diprotodontian branch received strong bootstrap support (100%), as did the Propalorchestes-Nimbadon clade (92%) and the Caudipilosus-Trichosurus clade (89%). Apomorphies occurring at nodes are listed below the cladogram in Fig. 12.3. Consistency indices and direction of change of nodal apomorphies are listed in Appendix 5B.

12.5 Discussion

Phylogenies The cladograms resulting from both the unordered and ordered analyses place Priscileo within the vombatiform radiation and therefore provide support for the 287 hypotheses of Marshall et al. (1990) and Munson (1990) which also place Thylacoleonidae within the Suborder Vombatiformes. Further similarities to those hypotheses include (see Fig. 12.2 and Fig. 12.3): 1, the grouping of the vombatiforms as the sister-group to a phalangeridan clade; 2, the position of phascolarctids (Litokoala) as the sister-group to all remaining vombatiforms (except Marada), and; 3, the position of thylacoleonids as the sister-group to all remaining vombatomorphians (except Marada).

The position of Marada as a sister-group to Bulungamaya in the vombatiform clade is unexpected. This newly described taxon is known from a single dentary and is considered to be a vombatomorphian (Black in press). The position of macropodoids within Diprotodontia is controversial. Macropodoids are traditionally regarded as being closely related to the “possum” groups on the basis of morphological similarities of the auditory region (Archer and Aplin 1987; Springer and Woodburne 1989) and certain biochemical and molecular data (eg. Kirsch et al.1997, Retief et al.1995, Amrine-Madsen et al. 2003, Kavanagh et al. 2004). Other morphological and molecular studies contradict the monophyly of Phalangerida - macropodoids appearing as a sister-group to both the “possum” clade and vombatiform clade (Springer et al. 1997) or as a sister-group to vombatiforms (Horovitz et al. 2003). In this study, the single unequivocal synapomorphy that groups Bulungamaya with Marada is lophodonty. Support for this synapomorphy is weak because the molars of phalangeridans such as Caudipilosus and Spilocuscus have a lophodont form and the presence of lophodonty in Marada and Bulungamaya is most likely homoplasious.

Unexpectedly, the phascolarctids fail to form a clade in either analysis: Litokoala is more closely aligned to vombatomorphs by the presence of a posterolingual and anterolingual crest on the P3. It is likely that the failure of the phascolarctids to form a clade results from the high number of unknown cranial characters for Nimiokoala; 20 of the 25 cranial characters could not be scored for this species.

The vombatiform phylogeny resulting from this study differs from those of Marshall et al. (1990) and Munson (1990) in the placement of ilariids, wynyardiids and vombatids: in this study, wynyardiids are the sister-group to an ilariid/vombatid/ 288 palorchestid/diprotodontid clade, unlike Marshall et al. (1990) in which ilariids are the sister-group of a wynyardiid/vombatid/palorchestid/diprotodontid clade. Munson (1990) did not include wynyardiids in her analysis but included them in a clade that contained palorchestids and diprotodontids (see Fig. 12.1c). Munson (ibid) suggested that this clade was the sister-group to ilariids and vombatids.

In the unordered analysis, unequivocal synapomorphies (character number in brackets) that unite the vombatiform clade include, (16) a P3 that is multicusped and that bears an (19) anterobuccal crest, (21) a P3 that exhibits a posterolingual crest, (22) an anterolingual crest, and (23) anterobuccal crest and, (46) presence of a masseteric process. In the ordered analysis, the character-states at Node D that unify the vombatiforms are: (11) lower premolars reduced to a single premolar, (12) upper premolars reduced to a one premolar, (21) a P3 that exhibits a posterolingual crest, and (33) M3 with a square occlusal outline.

The cladograms produced by the unordered and ordered analyses both fail to fully support the two vombatiform infraorders, Phascolarctomorphia and Vombatomorphia, proposed in the classification of Aplin and Archer (1987) primarily because of the relatively unstable vombatomorphian taxon, Marada, and because of the arrangement of the phascolarctid species. Because Litokoala and Nimiokoala fail to form a monophyletic clade in both analyses, the infraorder Phascolarctomorphia, is paraphyletic. If the affinities of Marada are regarded as uncertain and the taxon is excluded from consideration, the cladogram from the unordered analysis provides support for Vombatomorphia. In the cladogram from the ordered analysis, if the position of the macropodoid Bulungamaya is viewed as spurious, Marada is the sister-group to all remaining species of vombatiforms including phascolarctomorphs, hence making Vombatomorphia paraphyletic as well. At present, Marada is known from a single jaw which shows a combination of plesiomorphic and derived features (Black in press). A better understanding of the relationships of Marada to other vombatiforms may be possible only after the discovery of cranial material of this species. 289

In the unordered analysis, thylacoleonids are members of the vombatomorphian clade (= vombatiforms excluding phascolarctids). This clade is united by the following unequivocal synapomorphies: (45) absence of an interparietal; (49) presence of a medial glenoid process; (50) postglenoid foramen located within the postglenoid cavity; (54) squamosal roof and bilaminar (squamosal and alisphenoid) floor; (57) absence of a rostral tympanic process; and (62) the foramen ovale positioned lateral to the basisphenobasioccipital suture. As suggested by Archer and Aplin (1987, p.xlvii), although a “squamosal bulla” (present in vombatids and Thylacoleo spp.) is not a synapomorphy for vombatomorphians, other cranial features “…serve to hold Thylacoleonidae within Vombatomorphia”.

Thylacoleonid relationships to phalangeridans Murray et al. (1987) proposed that thylacoleonids are more closely related to phalangeroids than to vombatomorphians and that they are the sister-group to burramyids. Synapomorphies advanced to support the relationship of Wakaleo with phalangeroids include: 1) formula and morphology of the antemolar dentition (i.e. premolar reduction); 2) morphology of the palate (elongate rhomboid shape with palatal ridges); 3) similarity of the nasopremaxillary suture pattern and shape of rostrum; 4) bilaminar structure of the tympanic wing; and 5) broad contact of the presphenoid with the palatines and pterygoids in the interpterygoid fossa. Some of those character states - in particular, palate shape, rostrum shape and nasopremaxillary suture patterns - are also present in Priscileo and other extinct vombatiform taxa, and suggests that they are probably symplesiomorphic for diprotodontians. The shape of the palate of Priscileo is similar to that of W. hilmeri and Litokoala and, although rhomb-shaped, these palates are relatively wider than those of phalangerids. In contrast, the palates of Namilamadeta and Nimbadon are rectangular, while that of Vombatus is extremely narrow. These latter taxa are clearly more highly derived forms. The rostrum of Priscileo, like that of species of Wakaleo, is very similar in shape to those of Trichosurus and Phalanger, but is also similar in shape to Litokoala. The rostrum of other vombatiforms is highly variable, for example that of Nimbadon is elongate and anteriorly bulbous while that of Vombatus is relatively flat and broad. Priscileo, Wakaleo, Litokoala, Nimbadon and phalangerids also share a similar nasopremaxillary suture pattern but that of 290

Vombatus is extremely derived. In addition to Phalanger (i.e. Spilocuscus) and W. vanderleueri, a bilaminar tympanic wing is also present in Priscileo, Nimbadon, Namilamadeta and possibly Propalorchestes.

The partially bilaminar structure of the tympanic wing in Phalanger maculatus (i.e. Spilocuscus maculatus) and the bilaminar structure present in Wakaleo has also been considered indicative of close phalangeroid-thylacoleonid relationships (Murray et al. 1987). In contrast, Thylacoleo has a squamosal wing as do vombatids, and this similarity has been used to support a vombatomorphian affinity for thylacoleonids (Archer and Aplin 1987). The tympanic wing of Wakaleo vanderleueri is not preserved in its entirety, being broken at its base, but that of W. hilmeri is complete and indicates that the wing is bilaminar only near the anterior wall of the tympanic cavity and posteriorly it is constructed solely from the alisphenoid. A similar structure of the tympanic wing is apparent in Priscileo as well as other vombatomorphian taxa such as Nimbadon and Namilamadeta (and possibly Propalorchestes). In all these species the squamosal forms the roof and anterior wall (or part thereof) of the tympanic cavity, and the alisphenoid overlies the anterior part of the cavity and forms a relatively short tympanic wing. In many of these fossil taxa the tympanic wing is broken off, revealing its bilaminar nature. Litokoala also demonstrates a partially bilaminar wing, the squamosal being overlain by the alisphenoid at the anterolateral margin of the bulla. In both Litokoala and Priscileo the alisphenoid is present in the medial part of the tympanic roof, demonstrating an intermediate stage between the alisphenoid roof that was probably present in the ancestral diprotodontian (and present in dasyurids and perameloids) and the fully squamosal roof seen in vombatomorphians. The wide distribution of a bilaminar tympanic wing in extinct vombatiforms supports the suggestion of Springer et al. (1989, p.217) that this structure “…may represent an intermediate condition between phascolarctids and vombatids”. The partial bilaminar structure in Phalanger (i.e. Spilocuscus) is most likely to be independently derived or represents the plesiomorphic diprotodontian condition. Although Thylacoleo and vombatids both share a fully squamosal bulla, in the two phylogenies generated in this study vombatids consistently show a closer relationship to diprotodontids and palorchestids. This relationship supports the suggestion of Murray et al. (1987) that 291 the squamosal wing in these groups is independently derived and that the reduction in the alisphenoid is probably size-related.

Differences in the morphological details of other Wakaleo-phalangeroid synapomorphies listed by Murray et al. (1987) suggest that the similarities may have been independently derived. Although Wakaleo shares a reduced antemolar dentition with phalangeroids, the nature of this reduction differs. Phalangerids retain two upper premolars, P1 and P3. A small diastema lies between these two teeth and the anterior premolar is relatively large and caniniform. Although reduction of the number of upper premolars is present in Wakaleo it is variable in both W. vanderleueri and W. hilmeri. The sole anterior premolar in W. vanderleueri (when present) lies in a position similar to that seen in phalangeroids, suggesting that it is P2 that is lost, but in specimens of W. hilmeri that show premolar reduction, it is the anterior-most premolar that is lost and it is the tooth immediately anterior to P3 (i.e. P2) that is retained. The palatal ridges in thylacoleonids are very different to those in phalangeroids. Those of marsupial lions are very short and lie at an acute angle to the tooth row whereas those of phalangerids are much longer and stretch transversely across the whole palate. The differences in the size and location of the palatal ridges in these two groups suggest they have been independently derived. The broad presphenoid-palatine contact present in phalangerids and W. vanderleueri is also present in P. roskellyae as well as Nimbadon. This region is damaged or missing in other extinct vombatiform taxa and the relationships of these bones cannot be determined. It is possible that this character may also be symplesiomorphic for diprotodontians.

A number of burramyid-Wakaleo symplesiomorphies have also been cited as evidence of thylacoleonid evolution from stem phalangeroids. Strong emphasis has been placed on the similarities of the antemolar dentition of Thylacoleo and burramyids (i.e. three premolars, Murray et al. 1987) as well as the presence of tribosphenic molars in Wakaleo and burramyids (Murray et al. 1987, Rauscher 1987). However, as in Wakaleo, anterior premolar number is variable in skulls of T. carnifex, some exhibiting loss of P1. The region of the rostrum anterior to P3 in both 292

Wakaleo and Thylacoleo is clearly one morphologically variable. As indicated earlier, the tribosphenic molars of W. vanderleueri are clearly derived from a more quadritubercular ancestor and supports Rauscher’s (1987) suggestion that the loss or reduction of the metaconule in burramyids may have occurred independently. The phylogenies resulting from this study also support this view, Cercartetus (Burramyidae) occurring within the phalangeridan clade and Priscileo nesting within the vombatiform clade.

Although noted as a Wakaleo-phalangeroid symplesiomorphy, Murray et al. (1987) also argue that the sutural patterns on the occiput of thylacoleonids suggest a closer relationship to phalangeroids than to vombatoids (sensu Archer 1984). These authors indicated that the occiput of Thylacoleo and Wakaleo exhibit a very narrow lateral squamosal contribution and a broader, ventral mastoid contribution, a pattern that is similar to that of Trichosurus which also has a narrow squamosal contribution and a broad mastoid component. In contrast, vombatids were shown as having a broader lateral squamosal contribution and a relatively narrower and shorter mastoid. Skulls of Trichosurus (and Caudipilosus) commonly exhibit a transverse suture near the base of the occiput, just dorsal to the paroccipital process. In some specimens this suture cannot be discerned because of fusion between the two elements. Murray et al. (1987) noted a similar suture in Cercartetus as well as a distinct suture near the nuchal line, between the squamosal and the superior occipital element and suggested that the two elements are separate ossification areas of the mastoid. In Trichosurus, a suture cannot be observed along the nuchal line and Murray et al. (1987) assume that the squamosal fuses with a superior element of the mastoid early in development. Examination of skulls of juvenile Trichosurus present in the mammal collections of the Australian Museum indicate that the inferior occipital element is part of the mastoid, however no suture was observed to separate the large superior occipital element from the squamosal. Unless the presumed fusion of elements occurs extremely early in development, an alternative explanation for the osteological arrangement on the occiput of Trichosurus is that the broad superior element that is contiguous with the squamosal is in fact the squamosal and that the mastoid only contributes to the inferior wall of the occiput. This possibility is even more likely considering the pneumatisation of the squamosal in this region of the cranium in 293 phalangerids. It is also probable that this pneumatisation restricts the superior expansion of the mastoid. The transverse occipital suture is commonly absent in other phalangerids (Spilocuscus maculatus, Phalanger orientalis and Phalanger gymnotis). Specimens in which the exoccipital bone has been removed indicate that the periotic extends posteroventrally as the mastoid process of the petrosal. This process is delineated from the pneumaticised occipital bone by a suture that disappears towards the external surface of the skull, indicating the fusion of the occipital bone with the mastoid process. The external posterior sutural relations of the squamosal of Spilocuscus and Phalanger are difficult to discern because of the presence of a prominent ridge marking the perpendicular surface of the occiput. The occiputs of other phalangeridan taxa (Bulungamaya, Pseudochirops cupreus, Petaurus breviceps and Cercartetus caudatus) exhibit a relatively broad but tall mastoid component that is laterally flanked by the squamosal. It is possible that because the periotic occupies such a large part of the cranium in smaller possums (such as Petaurus and Cercartetus), its mastoid process dominates the relatively smaller occipital surface, hence there is a smaller squamosal contribution. A detailed developmental study of the ossification sequence of the cranial elements of Trichosurus is required to resolve this issue. Whatever the sutural relations may be in Trichosurus and other phalangerids, the narrow, lateral squamosal strip that is present in thylacoleonids is also present in other extinct vombatiforms (Namilamadeta and Litokoala) and is probably a symplesiomorphy of diprotodontians. The broader squamosal contribution evident in both Vombatus and Nimbadon is a derived condition.

Premolar and molar morphology Although the P3 morphology of Priscileo, Wakaleo and Thylacoleo is extremely derived and exhibits few similarities to other vombatiforms, that of Microleo is suggestive of closer relationships to certain vombatiforms as well as the possible morphology of the ancestral vombatiform premolar. The basic structure of the Microleo P3 consists of a lingually enlarged crown which bears a number of cusps that are aligned longitudinally along the centre of the crown. These cusps are linked by crests that result in the formation of essentially a single longitudinal crest or blade. This basic pattern is also seen in wynyardiids and phascolarctids. Although 294 the P3 of Microleo appears to have two main cusps, a rounded thickening of the longitudinal crest anterior to the posterior cusp suggests the presence of an incipient or reduced cusp and is similar to the morphology of Litokoala. The P3 morphology in more derived thylacoleonids is probably the result of enlargement or lengthening of the crown with the incipient or reduced central cusp becoming further reduced or lost, and/or being incorporated into the enlarged longitudinal blade. Considering the prevalence of a lingual cusp or cuspule in vombatiform taxa (phascolarctids, wynyardiids, ilariids and diprotodontoids), it does not seem unreasonable to postulate that the vombatiform ancestor possessed this cusp or its structural precursor, i.e. a lingually-swollen crown similar to that of Microleo.

Archer and Rich (1982) originally proposed that the tritubercular molar present in Wakaleo alcootaensis (and W. vanderleueri) is secondarily derived and results from the suppression of the hypocone (i.e. metaconule) that was present in a quadritubercular ancestor. The quadritubercular M1 present in species of Thylacoleo was interpreted to be a rotated tritubercular molar that had developed a new anterolingual cusp that resulted in a quadritubercular molar. In contrast, Murray et al. (1987) argued that the tritubercular molar present in Wakaleo is plesiomorphic and reflects an independent origin for thylacoleonids. The presence of quadritubercular molars that bear a metaconule in more plesiomorphic thylacoleonid taxa, i.e. in Microleo attenboroughi, Priscileo roskellyae and Wakaleo hilmeri, supports the original hypothesis of molar development proposed by Archer and Rich (1982). The tritubercular condition present in Wakaleo vanderleueri and W. alcootaensis is clearly a derived condition.

Summary Cladistic analyses of the interfamilial relationships of Thylacoleonidae suggests that it is a member of the vombatiform radiation and support the hypotheses of Aplin and Archer (1987), Marshall et al. (1990) and Munson (1990). The hypothesis of a thylacoleonid-burramyid sister-group relationship (Rauscher 1987, Murray et al. 1987) is not supported. Many of the features suggested by Murray et al. (1987) as evidence of a close relationship between thylacoleonids and phalangeroids are probably symplesiomorphic for diprotodontians or evolved independently. 296

CHAPTER 13

FUNCTIONAL ASSESSMENT OF THE POSTCRANIA OF Wakaleo hilmeri sp. nov. AND Wakaleo pitikantensis

13.1 Introduction

Very few functional morphological investigations have been carried out on marsupial lions primarily because little postcranial material is known for any species except the large Pleistocene species, Thylacoleo carnifex. Preservation of a near-complete skeleton of that species from Moree, NSW has resulted in functional morphological studies that have focussed on the spinal column and limb elements (Finch 1981, 1982; Finch and Freedman 1986, 1988). Although much of the surface osteology of the elements of that skeleton were obscured by a ferruginous cement, morphometric analysis using limb length indices concluded that Thylacoleo was a slow-medium cursor that may have had the capability to leap. A number of other studies investigating the functional morphology of fossil mammals have also used morphometric indices (van Valkenburgh 1985, 1987; Macleod and Rose 1993; Anyonge 1996). In contrast to T. carnifex, the limited and incomplete postcranial remains of Wakaleo hilmeri and W. pitikantensis severely restrict attempts to make similar morphometric analyses. Other functional studies of fossil mammals have focused on comparisons of the morphological features of the postcranial skeletons to those of modern taxa (Rauscher 1983, de Muizon 1998, Argot 2001, 2003; Szalay and Sargis 2001). The well preserved Riversleigh and Lake Pitikanta material allows similar comparative studies of the limb bones and their articular surfaces.

Rauscher (1983) investigated the functional morphology of Priscileo pitikantensis (SAM P37720; this specimen is referred to the genus Wakaleo in Chapter 4 and is recognised as Wakaleo pitikantensis throughout the remainder of this chapter) and compared it to that of T. carnifex. The elements studied in that assessment were 297 restricted to the humerus and radius. That study concluded that W. pitikantensis may have been fossorial or scansorial. Rauscher (ibid) based her assessment on a comparison of the osteological characteristics of the W. pitikantensis forelimb to those characteristic of three different types of locomotion, i.e. cursorial, scansorial and fossorial. Definitions of these locomotor categories were based on Hildebrand (1982), i.e. a cursor is an animal that moves far and easily on the ground, a scansor is an animal that climbs (as opposed to arboreal which means dwelling in trees), and a fossor is an animal that is adapted to dig. Rauscher (1983) notes that although the locomotor categories are rather general, many of the features used to identify particular locomotor behaviours are those of highly specialised species, for example in the fossorial category many of the features noted are those of highly specialised placental moles (Scapanus and Talpa) while features of the cheetah (Acinonyx) are used to identify a cursor. Both Hildebrand and Rauscher note that often animals do not fall into one specific category but rather show a range, or combination, of locomotory behaviours. Thus, the problem with using specialised characters to identify behaviours is that they restrict utility to highly specialised species and tend to overlook or exclude species with more generalist or broader habits.

Rauscher (1983, p.72) assessed the W. pitikantensis humerus and radius against seven cursorial characters, nine scansorial characters and seven fossorial characters. Within the scansorial category assessment was restricted by the large number of characters (four of nine) that were deemed indeterminate, while within the cursorial category only one character was indeterminate. Rauscher (1983) found that W. pitikantensis exhibited most correlation with the fossorial group although features of the radius indicated the possibility of scansorial locomotion as well. Of the seven fossorial characters assessed, W. pitikantensis was found to exhibit four and another was intermediately expressed; these characters include a relatively prominent deltopectoral crest, a wide teres major tubercle, a wide medial epicondyle, large lateral epicondylar crest, and muscle insertions that are far from the joints.

An inherent problem associated with assessing the functional morphology of fossil animals is the restricted nature of the morphological information that is available, i.e. 298 in nearly all cases only the bones are available for study because the soft tissues (muscles, ligaments and nerves) do not fossilise. Consequently, any reconstruction of the musculature of the fossil taxon must be based on assumptions either derived from the morphology of the bone (for example, the size of a crest or scar on a bone is usually correlated with the size of the muscle that attaches to this feature, and from which in turn its functional significance is then implied), or from the musculature of extant relatives (Bryant and Seymour 1990).

Five of the 23 characters (two each in the cursorial and scansorial categories and one in the fossor) in Rauscher’s (1983) assessment relate directly to muscle insertions and muscle size while a number of other osteological features are used to make indirect assumptions regarding muscle function. Rauscher noted the various difficulties in using muscle scars to predict the size, power and use of a muscle. These difficulties have been highlighted by Bryant and Seymour (1990) who emphasise that great caution must be taken when making assumptions about muscle attachments in the reconstruction of the morphology of fossil vertebrates. Their studies of the muscle attachments in Canis and Ursus found that the areal extent of a large number of muscles that attach directly to the bone cannot be ascertained because there is no scarring; however, muscles with tendinous or aponeurotic attachments are usually associated with rugose scarring ... "Therefore, for aponeurotic, and to a lesser degree for direct, attachments the location of at least part of the muscle, but not the full attachment, is inferable from osteological material" (p.113). These authors also indicate that the common assumption that an enlarged crest usually correlates with the size and functional significance of the muscle attaching to or along its edge can be erroneous and that consideration should also be given to the muscles that lie adjacent to, and which may attach to the sides of, the ridge. Those authors conclude that more reliable reconstructions need to be based on the musculature of closely related extant relatives and that as the phylogenetic relationship becomes more distant the validity of the muscular reconstruction will diminish. Unfortunately, marsupial lions have no close living relatives today and thus any comparisons with extant diprotodontians can only be with taxa that are distantly related. 299

As indicated in Chapter 4, the dental similarities between the specimen of W. pitikantensis and those of W. hilmeri suggest they may be conspecific. Although W. hilmeri is larger than W. pitikantensis, the obvious morphological similarities in their humeri probably reflect not only their close phylogenetic relationship but also similarities in their locomotory abilities.

Considering the scarcity of postcranial elements for pre-Pleistocene marsupial lions, the elements recovered for W. hilmeri from Riversleigh are important additions to the skeletal database of thylacoleonids. Together with the elements of W. pitikantensis they provide an improved picture of the forelimb of Tertiary thylacoleonids: in total, the elements now known include the humerus, proximal ulna, radius, hamatum, capitatum, trapezium, and pollex (excluding the ungual). Unfortunately, a calcaneum is the only addition to the poorly known hind limb and the paucity of elements of this limb precludes a meaningful functional assessment. New material from the lumbosacral area including posterior lumbar vertebrae, sacrum, and ilia, allows the first assessment of this region of the Wakaleo skeleton.

This study attempts to interpret the locomotor adaptations of both Wakaleo hilmeri and W. pitikantensis by comparing the morphology of their postcranial bones to those of extant marsupial groups that represent certain locomotor categories. The humerus and radius of W. pitikantensis has been included in the analysis with the aim of testing the fossorial/scansorial hypothesis that has been previously proposed for this species (Rauscher 1983).

13.2 Methods and Materials

This analysis differs from that of Rauscher’s (1983) in its use of a larger number of locomotor categories (five instead of three). On the basis of the mammalian diversity present in many of Riversleigh’s sites, Archer et al. (1989) have suggested that the Riversleigh environment was that of rainforest or that of mesic forest. Consequently, it was determined that an arboreal category (as opposed to solely scansorial) be 300 included in the range of locomotory models. In an attempt to separate more specialised fossors that live underground from those that burrow but also spend a large amount of time moving around on the surface, the fossorial category was divided into fossorial and semi-fossorial.

The extant marsupial comparative taxa were placed into one of five locomotor categories on the basis of published behavioural information: 1) arboreal, being defined as an animal that spends the majority of its time foraging and sheltering in trees; 2) scansorial, being defined as an animal that forages on the ground but is also an agile climber; 3) terrestrial, being defined as an animal that forages on the ground and rarely or never climbs; 4) semi-fossorial, being defined as an animal that digs regularly for shelter and food; and 5) fossorial, being defined as an animal that shelters and forages predominantly underground. These categories are adapted from Taylor (1974) and van Valkenburgh (1985) and are necessarily broad so as to accommodate animals that may display a range of locomotor capabilities. Definitions of ‘cursor’ such as that employed by Rauscher (1983) have traditionally been associated with the concepts of speed and distance (Stein and Casinos 1997; Carrano 1999). Because the behavioural emphasis of the categories used here is on foraging and shelter, the term ‘terrestrial’ has been preferred to that of ‘cursor’.

Of the almost 90 extant species of Australian marsupials, only six display regular burrowing activity. Within this small group, the notoryctemorphs, Notoryctes typhlops and N. caurinus are the only specialised fossorial species, while the wombats (Vombatus and Lasiorhinus spp.) and the Bilby (Macrotis lagotis) could be regarded as semi-fossorial. Approximately one third of species (30) are scansorial or arboreal, while the remainder are terrestrial (Strahan 1995). Of the terrestrial group, practically all are specialised bipedal saltators (the macropodoids) except the tree kangaroos which have secondarily become efficient climbers. To enable comparisons with terrestrial animals that are not bipedal saltators I have also included members of the Dasyuromorphia (thylacinids and dasyurids) and Peramelemorphia (peramelids). A total of 18 species, including two extinct species, of known locomotory habit were chosen. The taxa used in the comparisons are listed in Table 13.1 and include species 301

Deltopectoral crest length

Humerus length Distance between crests 35%

Lateral epicondylar crest height

Distal width

Fig. 13 .1 Humerus measurements 302 of: Thylacoleo (terrestrial; Finch and Freedman 1988), Phascolarctos (arboreal; Lee and Carrick 198; Martin and Handasyde 1995), Trichosurus and Spilocuscus (arboreal; MacLennan 1984; How and Kerle 1995; Winter and Leung 1995), Isoodon and Perameles species (terrestrial), Dasyurus (terrestrial/scansorial; Nowak 1999), Sarcophilus (terrestrial/ scansorial; Owen and Pemberton 2005), Thylacinus (terrestrial; Guiler 1985), Macropus (terrestrial; Frith and Calaby 1969), Dendrolagus (scansorial; Nowak 1999), Vombatus and Lasiorhinus (semi-fossorial; Wells 1973) and Notoryctes (fossorial; Johnson 1995). As noted above, some of these taxa do not belong exclusively to a single category, for example, species of Isoodon and Perameles are terrestrial but also partake in scratch-digging (Gordon 1983; Stodart 1983) and Dasyurus maculatus, although spending most of its time on the ground, is also an excellent climber (Edgar 1983). Adult and juvenile specimens were examined although only the bones from adults were measured unless otherwise noted. Specimens were considered to be adult if they possessed a fully erupted dentition and the epiphyses were fused to the shafts of appendicular elements.

Comparisons of the elements of the forelimb of W. hilmeri and W. pitikantensis were made with those of the extant marsupial locomotor models. Measurements were made of various morphological features of the humeri; these include humerus length, humeral distal width, deltopectoral crest length (absolute and relative), lateral epicondylar crest length (absolute and relative), and the shaft length between the deltopectoral crest and the lateral epicondylar crest (see Fig. 13.1 and Table 13.1). These features were also compared to those deemed characteristic for the different locomotor groups as described in Rauscher (1983). The morphology of the lumbar vertebrae, sacrum and ilia of W. hilmeri were also assessed in relation to those of the extant locomotory models.

In describing the surfaces and positions of the limbs, the animal is assumed to be postured in a plantigrade, four-footed position. The anatomical directions of anterior and posterior equate to cranial and caudal, respectively. In descriptions of the limb elements anterior also refers to the extensor muscle surfaces and posterior refers to the flexor muscle surfaces. The terms distal and proximal are used relative to the 303 Table 13.1: Humeral measurements of thylacoleonids and selected comparative marsupial taxa Measurements in mm. All measurements are the average of right and left elements except where indicated (*). See Figure 13.1 for illustration of measurements.

Locomotion Species Specimen No.Humerus Humerus distal Deltopec. Deltopec. lateral lateral Distance between type length distal width as crest length as epicond. epicond. deltopec. crest width % of length % of total crest crest height and lat. epicond. length length height as % of total crest ** length ?fossorial/ Wakaleo SAM P37720* 109.8 43.1 39.2 54.3 49.0 36.5 32.6 14 scansorial pitikantensis ? Wakaleo hilmeri QM F23444* 135 est. 52.9 39.4 - - 46.0 34 est. 26 terrestrial Thylacoleo AM F106768* 197.0 72.3 36.7 106.0 53.0 51.0 25.8 36 carnifex SAM P40007* 217 76.4 35.2 114.4 52.7 59.9 27.6 35 SAM P38793* 205# 76.0 37.0 - - 54.4 26.5 36 semi- Vombatus AR 7640 122.1 54.3 44.4 67.8 55.5 50.2 41.1 -2 fossorial ursinus AM P793 109.3 43.7 40.0 60.7 55.5 42.5 38.8 -2 AM M34032 108.3 48.6 44.9 58.7 54.2 40.0 36.9 6 semi- Lasiorhinus AM M13515 108.1 48.9 45.2 64.0 59.2 47.0 43.4 -6 fossorial latifrons AM M13514 105.0 49.4 47.0 63.9 60.9 42.8 40.8 -5 fossor Notoryctes SAM 637 14.4 8.8 61.1 7.6 52.7 5.5 38.1 -2 typhlops AM 16165 13.4 7.8 58.2 9.3 69.4 4.5´ 33.5 - CSIRO CM6010 15.6 10.1 64.7 11.0 70.5 6.2 39.7 - arboreal Phascolarctos UNSWZ 2464 110.6 33.9 30.6 50.3 45.4 38.2 34.5 18 cinereus AM M23550 123.7 39.8 32.2 56.3 45.5 40.2 32.5 22 AM M28154 112.0 36.3 32.4 53.1 47.4 37.6 33.5 18 arboreal Trichosurus AM 34417 74.5 21.8 29.2 36.0 48.1 23.3 31.2 16 vulpecula AM S1851 74.4 21.4 28.8 34.9 46.8 24.6 33.1 15 AM S1304 72.0 20.3 28.1 34.2 47.5 20.3 28.1 - arboreal T .v. fuliginosus AM S946 78.7 24.8 31.5 37.6 47.7 22.9 29.0 15 arboreal T. caninus AM M35406 66.7 22.2 33.2 31.5 47.2 22.7 34.0 11 AM M33831 70.8 22.5 31.7 35.2 49.7 22.5 31.7 13 AM M33832 70.2 22.9 32.6 34.1 48.6 21.6 30.8 13

est. = estimated measurement; * = single element; ** = distance between deltopectoral crest and lateral epicondylar crest rounded to nearest millimetre. Negative value indicates overlap measurement. # = dorsal articular surface missing, measurement is a minimum. ´ = broken crest. - = measurement unobtainable 304

Table 13.1 cont.: Humeral measurements of thylacoleonids and comparative marsupial taxa Measurements in mm. Measurements are the average of right and left elements except where indicated (*) See Figure 13.1 for illustration of measurements

Locomotion Species Specimen No. Humerus Humerus distal Deltopec. Deltopec. lateral lateral Distance between type length distal width as crest length as epicond. epicond. deltopec. crest width % of length % of total crest crest height and lat. epicond. length length height as % of total crest ** length arboreal Spilocuscus AM M11634 89.5 26.6 29.7 45.0 50.0 26.0 29.0 18 maculatus AM M23597* 101 26.4 26.1 51 50.4 32.4 32.0 13 AM M23627 75.3 22.3 29.6 33.7 44.7 25.7 34.1 13 scansorial Dasyurus AM S497 70.5 18.0 25.5 34.6 49.0 21.6 30.6 12 maculatus AM S879 76.6 19.0 24.8 38.6 50.3 25.4 33.1 11 AM S1787 74.7 18.8 25.1 39.5 52.8 24.0 32.1 11 scansorial Dasyurus AM P750 49.9 10.1 20.2 23.3 46.6 12.7 25.4 12 viverrinus AM M1765 55.9 12.2 21.9 27.6 49.3 15.6 27.9 10 AM S1190 69.7 19.9 28.5 31.8 45.6 18.2 26.1 19 scansorial Dendrolagus AM M25397 103.3 35.0 33.8 53.8 52.1 27.2 26.3 17 scottae terrestrial Macropus AM 9164 114.7 39.6 34.5 61.3 53.4 32.4 28.2 16 agilis terrestrial Sarcophilus AM S1605 97.7 23.8 24.3 53.8 55.0 24.7 25.3 17 harrisii AM P714 95.8 24.1 25.1 54.6 56.9 22.7 23.7 17 terrestrial Thylacinus AM 763* 189 36.4 19.2 128 67.7 45.0 23.8 8 cynocephalus terrestrial Isoodon AM M33981 57.1 15.9 27.8 38.5 67.4 11.1 19.5 5 macrourus AM M33980 51.9 16.7 32.1 33.4 64.5 10.3 19.9 7 terrestrial Perameles AM M35409 41.5 11.0 26.5 21.5 51.8 6.5 15.7 10 nasuta AM S1072 47.1 13.4 28.5 23.0 48.8 8.3 17.7 15 AM S1074 51.2 14.0 27.3 24.0 46.9 9.8 19.1 16 est. = estimated measurement; * = single element; ** = distance between deltopectoral crest and lateral epicondylar crest rounded to nearest millimetre. 305 distance from the vertebral column. The limb movements of abduction and adduction refer to the movement of the limb away and towards the central, longitudinal axis of the body. Anatomical terms are based on those from the Nomina Anatomica Veterinaria (1992), but I have replaced the Latin with the English equivalent as similarly used in Miller et al. (1964).

13.3 Results and Discussion

Humerus The general morphology of the humeri of W. hilmeri and W. pitikantensis is most similar to that of Trichosurus and Spilocuscus species. Although the humeri of W. pitikantensis and W. hilmeri have a relatively broader distal width that is reminiscent of the vombatids, overall they share a large number of similarities with the phalangerids including a relatively straight diaphysis, the size and proportional lengths of the deltopectoral and lateral epicondylar crests, a well-developed medial epicondyle that is separated from the trochlea by a moderate groove, a relatively well-developed entepicondylar foramen and most significantly, the shape and relative sizes of the capitulum and trochlea and shallow olecranon fossa (see Fig. 13.2).

The proximal portion of the humerus of W. pitikantensis shows most similarity to that of Trichosurus species. In proximal view the head of the humerus is roughly circular, the width being slightly less than the length. In lateral view the head is very rounded and relatively deep with the convexity directed both proximally and posteriorly. Similar convexity and orientation is also seen in species of Trichosurus and Phascolarctos cinereus, however, the humeral head of those species lack the proximodistal depth of W. pitikantensis. The head of the humerus of T. carnifex displays a similar depth and posterior convexity however, the proximal surface, like that of the wombats, is relatively flatter. The greater tubercle is lower than the humerus head, a feature also observed in Spilocuscus, Phascolarctos and Perameles species. The greater tubercle is slightly taller than the head in Trichosurus and 306

Dasyurus species. In T. carnifex, S. harrisii, T. cynocephalus, N. typhlops and the wombats (especially L. latifrons) the greater tubercle is well-developed and taller than the head. The humerus of arboreal mammals tends to exhibit a reduced greater tubercle and a proximally-directed head, both of which allows for a greater range of movement at the shoulder joint (de Muizon 1998).

The bicipital groove is moderately well-developed in W. pitikantensis and continues distally as a distinct sulcus along the medial border of the deltopectoral crest. Similar proximal development of this structure is seen in P. cinereus and to a lesser extent in Trichosurus species, although these taxa lack the distal sulcus. In Dasyurus, Perameles and Sarcophilus species the groove is slightly shallower and is relatively broad proximally, however, no distal extension of the sulcus is present. The bicipital groove is relatively shallow and broad in the wombats and T. carnifex. The humeri of arboreal didelphids (Argot 2001) and arboreal carnivores (Taylor 1989) tend to have a large groove for the tendon of the biceps brachii muscle and this indicates strong flexor ability of the forearm. On the scapula of W. pitikantensis, the proximal attachment sites of the biceps brachii (supraglenoid tuberosity and coracoid process) are well-developed and robust and also suggest an important role for this muscle.

On the lateral margin of the humerus of W. pitikantensis a thin line descends from near the base of the greater tuberosity and runs distally, terminating approximately 10mm proximal to the end of the deltopectoral crest. The middle portion of this line is raised into a narrow but moderate tuberosity. This tuberosity is not mentioned in Rauscher’s (1983) original description. Within the comparative taxa, the only species to exhibit a tuberosity in this region is Macropus agilis, in which it is extremely well- developed. This tuberosity was also observed in other species of macropodoids. In Dendrolagus scottae, the deltopectoral crest broadens laterally to form a thin but prominent flange in the same region. A similar prominent process is also found on the humerus of the didelphid Caluromys philander (Argot 2001). Barbour (1963) describes the deltoid muscle of Trichosurus vulpecula as attaching to the deltoid tuberosity along a line that runs proximally from the tuberosity towards the base of the greater tuberosity. In Macropus giganteus, Dendrolagus sp. (Shrivastava 1962) 307 and the didelphid, Caluromys philander (Argot 2001), the deltoid muscle attaches to the humerus in this region. More specifically, the portion of the deltoid that originates from the spine of the scapula attaches to the humerus in this particular area in Caluromys (Argot 2001) and Dendrolagus dorianus (Carlsson 1914). The development of this particular tuberosity in W. pitikantensis suggests that the deltoid muscle, and probably the spinal part in particular, were subjected to forceful contractions.

The humeral shaft of W. pitikantensis exhibits a robustness similar to that of T. carnifex and S. harrisii (especially proximally). The distal fragment of the W. hilmeri humerus is similarly robust. In profile, the shaft of W. pitikantensis is relatively straight and in this respect is similar to the vombatids. The humeral shafts of T. carnifex, Trichosurus, Spilocuscus and Macropus exhibit very slight anterior bowing and posterior concavity while in P. cinereus it is more pronounced. The greatest degree of bowing is found in the dasyurids and peramelids.

The deltopectoral crest on the humerus of W. pitikantensis is well-developed. Proximally, it is broad and distally it narrows to become a moderately narrow but strong, salient crest. The distal portion of the crest displays a slight medial bowing. The distal end of this crest is preserved on the W. hilmeri humerus and appears to have also been a moderately narrow but strongly protruding crest. The deltopectoral crest of Spilocuscus and Trichosurus species resembles that of W. pitikantensis but in a less robust fashion. The crest of P. cinereus is similar in length but is much narrower. The deltopectoral crest of T. carnifex is relatively broader especially in the distal third, and also lacks the sharp medial edge present in W. pitikantensis. Notoryctes typhlops and the wombats differ significantly in having crests that are very broad, especially distally. Although this crest in the dasyurids and Perameles nasuta exhibits a similar proximal breadth, it is much less prominent anteroposteriorly. Additionally, whereas in W. pitikantensis the crest has a slightly lateral position relative to the shaft, in these taxa the crest lies anterior to the shaft. 308

A

B

Wakaleo Wakaleo Thylacoleo Vombatus Lasiorhinus Notoryctes hilmeri pitikantensis carnifex ursinus latifrons caurinus*

A

Thylacinus cynocephalus B

Phascolarctos Spilocuscus Trichosurus Dendrolagus Macropus PeramelesDasy urus Sarcophilus cinereus maculatus vulpecula scottae agilis nasuta maculatus harrisii

Fig.13.2: Humeri of thylacoleonids and comparative marsupial taxa, anterior (A) and posterior (B) views. Note: humerus of W. hilmeri is mirrored to assist comparison. * after Warburton 2003. Not to scale. 309

Rauscher (1983) categorises the prominence of the deltopectoral crest of W. pitikantensis as an "intermediate" fossorial characteristic while that of T. carnifex as "lacking". "Prominence" in that study was interpreted as anteroposterior depth. Observation of this structure indicates that two factors contribute to its prominence, anteroposterior depth and transverse width. In both W. pitikantensis and T. carnifex, in anterior view the deltopectoral crest is moderately broad proximally and gradually narrows distally, a pattern seen in the humeri of most of the comparative species except the semi-fossorial or fossorial taxa. The crest in neither thylacoleonid species approaches the width of the crest seen in the wombats (especially Lasiorhinus latifrons) or in species of Notoryctes, in which the distal ends broaden laterally. In profile, the deltopectoral crests of W. pitikantensis and T. carnifex gradually protrude anteriorly and are similar in profile and depth to the crests of non-fossorial taxa such as Phascolarctos, Trichosurus, and Dasyurus species. The crests on the humeri of terrestrial taxa are anteroposteriorly very weak; when viewed in profile the crests usually form a weak ridge or bulge on the anterior surface of the humerus and as noted above, become narrower distally, not broader. In contrast, when the humeri of the wombats and Notoryctes typhlops are viewed in profile, the distal end of the deltopectoral crest forms an anteriorly-protruding, salient process.

Rauscher (1983) indicated that W. pitikantensis lacks the long deltopectoral crest characteristic of fossors while that of T. carnifex is intermediate. Values for the relative length of the deltopectoral crest are shown in Table 13.1. The crests of W. pitikantensis and T. carnifex are relatively similar in length, differing by only 4%. Such a small difference is probably well within the range of variation expected for such a measurement and suggests that differences due to allometry are negligible. The value for W. pitikantensis is more similar to those obtained for the arboreal species Trichosurus species and S. maculatus, the scansorial Dasyurus maculatus, and the terrestrial Perameles nasuta. The values for T. carnifex are similar to those of Macropus agilis and Dendrolagus scottae and approach those obtained for Vombatus ursinus but are less than those for Lasiorhinus latifrons and the marsupial mole Notoryctes typhlops. Within the comparative group a general trend seems to be apparent with arboreal and scansorial species having shorter crests than semi- 310 fossorial and fossorial species. The terrestrial species, i.e. T. cynocephalus and S. harrisii, also have relatively long crests. However, as noted above, these taxa differ greatly from fossorial taxa in having very narrow and anteroposteriorly-shallow crests. In all three aspects of length, width and depth, the deltopectoral crest on the humerus of W. pitikantensis is most similar to those of arboreal species as well as some scansors. The length of its deltopectoral crest is also relatively similar to that of terrestrial scratch-diggers and is least similar to those of semi-fossorial or specialist fossorial taxa.

Another fossorial feature attributed to Priscileo by Rauscher is the presence of a wide teres major tubercle. Hildebrand (1982, p. 472) notes that a "very wide flaring tubercle" for insertion of a large teres major muscle allows for more powerful internal rotation of the humerus. It should be stressed that the importance of this feature is not that it increases the area of attachment for the muscle but that it changes the angle of action of the muscle on the humerus; by moving the insertion farther away from the central axis of the shaft it increases its rotational action on the humerus (Hildebrand 1982). This structure is seen in the highly modified rotation- thrust digging placental moles such as Scapanus, Talpa and Amblysomus species and forms a large and distinctive prominence or flange on the medial margin of the humeral shaft (Hildebrand 1985). This feature is not a general characteristic of all fossorial animals, even among moles, and seems to be more of an indicator of a specialised type of digging than a general characteristic that could be applied to most fossors. No structure showing this extreme level of development is found on the humerus of any of the marsupial taxa included in this study, including the fossorial species. Within these taxa the nearest equivalent to this distinctive structure is a relatively narrow crest seen on the medial margin of the humerus of Lasiorhinus latifrons (see Scott and Richardson 1988). Posterior to this crest the shaft bears a long pit and scarring. Although a similar crest is lacking in Vombatus ursinus, that species bears a distinct, thickened, longitudinal ridge that is also posteriorly bordered by a long pit. The long trough and neighbouring rugosity seen in vombatids represents the attachment of the conjoined teres major and latissimus dorsi muscles (Murray 1998). The marsupial mole, Notoryctes typhlops, also lacks a flaring 311 tuberosity or a Lasiorhinus-like medial crest. Warburton (2003) notes a slightly raised bony area in the region of insertion for the teres major in one specimen of N. caurinus. The latissimus dorsi has no humeral insertion in Notoryctes (Wilson 1894; Warburton 2003). None of the humeri of the remaining comparative taxa bears a deep longitudinal pit as exhibited by the vombatids but most do bear a smaller scar of varying rugosity or a small round tuberosity associated with the area of attachment for these muscles. The humerus of W. pitikantensis bears a small oval tubercle on the medial edge of the shaft. This tubercle is flanked anteriorly by a longitudinal rugose area. Thylacoleo, Trichosurus and Spilocuscus species also bear a small tuberosity in the same area but differ by having the neighbouring rugosity extend distally down the posteromedial margin of the shaft. Assuming the myology of thylacoleonids is analogous with that of living diprotodontians, this small tubercle and its neighbouring rugosity is associated with the attachment of the teres major muscle and the latissimus dorsi (Barbour 1963). The lack of a flaring teres major tubercle and the similarity of the size and location of the scarring on the W. pitikantensis humerus to that seen in non-fossorial taxa, and in particular some arboreal species, do not lend strong support to Rauscher’s (1983) fossorial hypothesis for this species.

Studies have shown that muscles that are overstretched or put under continual strong exertion cause morphological responses such as necrosis and avulsions at the site of attachment which in turn results in erosion of the cortical bone and bone remodelling (Bufkin 1971; Bernasek et al. 1987; Fulton et al. 1979). Such remodelling usually results in the formation of depressions or pits in the cortical bone at the site of attachment of extremely strong muscles (Bufkin 1971; Duckler 1997). It is highly probable that the longitudinal pits seen in the area of attachment of the teres major and latissimus dorsi muscles on the humeri of Vombatus ursinus and Lasiorhinus latifrons are caused by the increased local mechanical stresses placed on these muscles which are used in the action of digging. The lack of these pits on the humerus of W. pitikantensis suggests that the forces being exerted by its teres major and latissimus dorsi muscles were not as strong or persistent as those of these semi- fossorial taxa. However, the presence of a moderate teres major tuberosity and the medial rugosity on the humerus of W. pitikantensis does suggest at least moderately 312 forceful contractions of these muscles. These muscles adduct and internally rotate the humerus, an action important not only for digging but also in climbing where there is a need to pull the body close to and maintain pressure on a branch or tree trunk (Taylor 1974).

Distally, the humeri of W. hilmeri and W. pitikantensis display a breadth similar to that of T. carnifex and V. ursinus, being broader than Trichosurus spp. and P. cinereus but narrower than L. latifrons and N. typhlops (Table 13.1). The relatively greater breadth in these taxa results from the presence of a well-developed medial epicondyle. A further similarity shared with the vombatids and T. carnifex is the relatively distal position of the condyle. However, unlike those taxa, the medial epicondyle in W. hilmeri and W. pitikantensis is separated from the trochlea by a moderate arch or groove and in this feature they are more similar to the other marsupial taxa. Argot (2001) notes similar development of the medial epicondyle in arboreal didelphids (species of Caluromys, Micoureus and Didelphis). The medial epicondyle is moderately developed in most of the arboreal and scansorial species (Trichosurus, Spilocuscus, Phascolarctos and Dasyurus species). Extreme reduction of this structure is evident in T. cynocephalus and to a lesser extent in S. harrisii. The degree of width of the medial epicondyle within marsupials shows a general trend of reduction from arboreal species to terrestrial species.

Rauscher (1983) used the characteristic of a "wide entepicondyle" (i.e. medial epicondyle) to indicate fossorial behaviour, and accredited this characteristic to the W. pitikantensis humerus. The increased width of the medial epicondyle is a feature of scratch-diggers and is the site for the origin of the forearm pronators and carpal and digital flexors (Hildebrand 1982, 1985). The width of the medial epicondyle is taken to an extreme in some species of moles, for example in Notoryctes typhlops and Amblysomus hottentotus (Hildebrand 1985). Although not exhibiting the extreme form of development seen in these species, the medial epicondyle in both the W. pitikantensis and W. hilmeri humeri is similar to those of L. latifrons and V. ursinus although the thylacoleonids lack the anteroposterior depth seen in the condyles of the vombatids. In Trichosurus vulpecula, the pronator teres, flexor digitorum profundus 313 and superficialis, flexor carpi ulnaris, and the flexor carpi radialis muscles take their origin from the front of the medial epicondyle (Barbour 1963). The latter muscle is large and includes the flexor pollicis longus, a muscular complex that is common within marsupials (Barbour 1963). Arboreal didelphids also possess a prominent medial epicondyle and strong development of the digital flexors that attach to this condyle (Argot 2001). Thus, the development of the medial epicondyle in W. pitikantensis and W. hilmeri suggests the presence of well-developed forearm pronators and digital flexors which indicate enhanced forearm pronation as well as digital flexion. These features are present both in scratch-digging mammals (Hildebrand 1985; Taylor 1989) and in climbers who also require strong digital flexors during the propulsive phase of limb movement (Argot 2001).

Strong development of the lateral epicondylar crest on the humerus is also a feature used to signify adaptation to digging (Hildebrand 1982). This crest is a site for the origin of the forearm supinator muscles and wrist extensors (Hildebrand 1982). Rauscher categorised the humerus of W. pitikantensis as possessing a large crest while that of T. carnifex was lacking such development. The lateral epicondyle forms a relatively narrow process just lateral to the capitulum on the humeri of W. pitikantensis, W. hilmeri and most of the comparative taxa with the exception of the wombats where it is quite broad. The development of the crest dorsal to the condyle also shows its broadest development in the wombats while in most other taxa the crest attenuates slightly towards its apex. Although not as broad as the crest in wombats, the crest in W. pitikantensis and W. hilmeri is relatively broader than in the comparative taxa (see Table 13.1). In terms of relative length, the lateral epicondylar crest of W. pitikantensis (32.6%) and W. hilmeri (~34%) is most similar to P. cinereus (32.5- 34.5%) and Trichosurus spp. (28.1 - 34%), is shorter than the wombats (36.8 - 43.4%), and longer than the terrestrial species, for example, S. harrisii (23.7 - 25.3%), T. carnifex (25.8%), and Thylacinus (25%). Significantly, although the lateral epicondylar crest in P. cinereus is relatively smaller than those in the wombats, the supinator muscles of the former are well-developed and comparatively larger than those of wombats (Young 1882; Sonntag 1922; Grand and Barboza 2001). The supinator muscles are also well-developed in species of 314

Trichosurus and Phalanger (Sonntag 1922; Barbour 1963). These observations highlight the caution that is required when making assumptions regarding muscle size from the size of crests (Bryant and Symonds 1990). Thus, large supinator muscles are not an exclusive feature of fossoriality. Climbers share with fossors the requirement of strong and effective supination and, as demonstrated by P. cinereus, this may be achieved by more than simply increasing the area of muscle attachment. If the moderately large lateral epicondylar crest seen in W. pitikantensis and W. hilmeri did support relatively large supinator muscles, this musculature may as much indicate arboreal locomotion as it does fossorial function. Approaching this aspect from another perspective, the level of development of these crests in W. pitikantensis and W. hilmeri speaks against them being specialised terrestrial locomotors.

The location of muscle insertions relative to the joints they act upon is also used to identify fossoriality or cursoriality (Hildebrand 1982). Hildebrand (1982) indicates that cursors have muscle insertions close to the joints. This arrangement allows a greater range of motion at the joints. Fossors, on the other hand, have their insertions far from joints because this allows stronger rotation forces on the limbs. The full length of the humeral shaft of the wombats and Notoryctes is adorned by crests (see Table 13.1), consequently the muscles that attach to these crests (pectoralis major and deltoid, forearm supinators and extensors) act upon a proportionally large part of the total shaft length. Although these crests in W. pitikantensis and W. hilmeri are moderately large, their shafts are similar to those of the non-fossorial species, especially P. cinereus, phalangerids and dasyurids, in having a relatively large central region that is devoid of crests. The muscles that attach to these crests therefore act nearer the joints compared to the semi-fossorial or fossorial taxa. The similarities of the relative humeral crest lengths and bare shaft lengths of W. pitikantensis and W. hilmeri with the arboreal and scansorial forms in comparison to those of the fossorial or terrestrial forms, lends support to an arboreoscansor niche hypothesis for these thylacoleonids.

The humeri of W. pitikantensis and W. hilmeri possess a moderate to large entepicondylar foramen. The size of the foramen in W. pitikantensis is similar to that 315 of P. cinereus, Trichosurus spp. and S. maculatus, while that of W. hilmeri is larger and approaches the size seen in the wombats. The entepicondylar foramen is reduced in Sarcophilus, Isoodon, Perameles and Macropus species and is absent in Dasyurus and Notoryctes species. Landry (1958) indicates that the presence of this foramen in mammals is associated with the ability to abduct and supinate the forearm. These movements are essential components of climbing or grasping.

Although the distal widths of the thylacoleonid humeri are similar to those of the wombats, the size and shape of their distal articular surfaces differ from those of the latter. The capitula of W. pitikantensis and W. hilmeri are well rounded and taller than their trochleae. In convexity and relative size to the trochlea, their capitula are most similar to those of Trichosurus, Spilocuscus, Phascolarctos and Macropus species. The trochleae of W. pitikantensis and W. hilmeri are relatively broad and extend distally to approximately the same level as the capitulae. The trochlea of W. pitikantensis differs from W. hilmeri in being proximodistally shallower and less convex and in this respect shows great similarity to P. cinereus. Posteriorly, both humeri possess shallow olecranon fossae, another similarity shared with P. cinereus and the phalangerids. In the dasyurids and peramelids the tendency is for the trochlea to be proximodistally larger, the capitulum to be laterally-reduced and the capitulum and trochlea to form a spindle-shaped facet. In vombatids, although the capitulum is relatively well-developed, the trochlea is also large and extends distally well beyond the level of the capitulum. Thylacoleo carnifex also demonstrates relative enlargement of the trochlea, however its capitulum is also relatively larger than that of vombatids. Taylor (1974) reports that the relatively flat, less angular trochlea and capitulum that is characteristic of climbing and arboreal walking viverrids assists the rocking of the ulna on the humerus. These mediolateral excursions aid supination of the manus, an action that is necessary for climbing (Taylor 1974). Cursorial mammals tend to restrict movement at the elbow to flexion and extension in the sagittal plane (Jenkins 1973). This restriction is achieved by deepening the trochlea and increasing the angulation between it and the capitulum. The resulting close fit of the trochlea to the trochlea notch of the ulna restricts sideways movements of the ulna on the humerus (Taylor 1974). This type of 316 morphology is most obvious in species of Dasyurus, Sarcophilus, Perameles and Thylacinus. The trochleae of these species are proximodistally deep and the margins of their facets are raised anteriorly and posteriorly to form sharply defined ridges that restricts movement to the anteroposterior plane. In comparison to W. pitikantensis, the trochlea of W. hilmeri is deeper and slightly more convex. Posteriorly, the lateral lip of the trochlea is also more prominent than in W. pitikantensis. These differences suggest that movements at the elbow joint of W. hilmeri may have been slightly more restricted to the parasagittal plane in comparison to W. pitikantensis.

The humeri of both W. pitikantensis and W. hilmeri bear very shallow olecranon fossae, a feature also exhibited by Phascolarctos, Spilocuscus, Trichosurus, Dendrolagus and Macropus species examined here. This fossa is also relatively shallow in T. carnifex but slightly deeper in vombatids. In dasyurids and peramelids the fossa perforates the distal surface of the bone. The arboreal didelphids Caluromys philander and Micoureus demererae also possess a shallow olecranon fossa, and this feature is associated with a broad but reduced beak on the olecranon of the ulna (Argot 2001). This latter feature is also present on the ulna of W. hilmeri (see Ulna, below). A deep olecranon fossa allows the forearm a greater arc of movement by allowing greater extension, however, it is associated with a stabilised and restricted elbow movement that is a requirement of cursors (Taylor 1974).

Ulna The olecranon process of the W. hilmeri ulna is very robust and in general morphology is most similar to that of species of Trichosurus, Spilocuscus, Phascolarctos and Thylacoleo, being roughly rectangular in profile, moderately deep, with a broad posterior surface (see Fig. 13.3). The olecranon of T. carnifex differs slightly in being anteroposteriorly deeper in the region of the anconeal process and tapers proximally. The broad posterior surface, large medial tuberosity and prominent muscle scarring on its proximal and lateral surfaces suggest the presence of strongly developed triceps brachii muscles. Relative to the size of the articular facet, the olecranons of the wombats and peramelids are longer and narrower both anteroposteriorly as well as mediolaterally. The olecranon process of 317

A

B

Wakaleo Thylacoleo Vombatus Lasiorhinus Phascolarctos Notoryctes hilmeri carnifex ursinus latifrons cinereus caurinus*

A

B

Spilocuscus Trichosurus Dendrolagus Macropus Sarcophilus Dasyurus Perameles maculatus vulpecula scottae agilis harrisii maculatus nasuta

Fig. 13.3 : Medial (A) and anterior (B) views of the ulna of Wakaleo hilmeri and various marsupials. * after Warburton 2003 318

W. hilmeri also lacks the proximal anterior-widening that is present in V. ursinus. The olecranon processes in species of Dendrolagus, Macropus, Sarcophilus, Dasyurus and Thylacinus are relatively short and proximally rounded. Another characteristic of diggers, especially scratch-diggers, is the medial curvature of the olecranon (Hildebrand 1985). The medial surface of the olecranon is the site of attachment for the muscles that pronate the forearm and flex the wrist and manus. Medial inflection of the olecranon changes the line of action of these muscles and increases their effectiveness. This increased medial inflection is present in wombats and taken to an extreme in species of Notoryctes. Although the ulna of W. hilmeri is incomplete its olecranon appears to lack the strong curvature observed those taxa.

The trochlea notch in W. hilmeri is proximo-distally concave and gently convex mediolaterally, and is relatively open, and shows most similarity to that of Trichosurus species. In these taxa the trochlea notch is not particularly deep and the anconeal process is weakly developed. The general shape of the facet of P. cinereus is also similar to that of W. hilmeri however the former differs in having a more strongly developed anconeal process and better developed proximo-medial margin. In contrast, this notch in the vombatids, dasyurids, peramelids and Thylacinus is very deep and their anconeal processes are very prominent. Proximal development of the proximo-medial half of the trochlea notch is commonly found in arboreal species such as species of Spilocuscus, Dendrolagus (and to a lesser extent P. cinereus) as well as arboreal didelphids (Caluromys and Micoureus; Argot 2001). The semi- fossorial and terrestrial taxa (wombats, T. carnifex, S. harrisii, P. nasuta and M. agilis) on the other hand, tend to exhibit development on the lateral side resulting in lateral protrusion of the facet. The trochlea notch of W. hilmeri is rather intermediate in this feature, lacking similar development of a particular proximal margin and in this respect is similar to species of Trichosurus and P. cinereus. Hildebrand (1985) suggests that diggers require stable joints to combat the strong forces encountered during digging. To prevent dislocation and hyperextension in these instances, their joints often have passive mechanisms such as bony stops to limit rotation. The elbow joint of the wombats show a number of such mechanisms. Firstly, the strong development and curvature of the anconeal and coronoid processes on the ulna and 319 large trochlea and deep olecranon fossa on the humerus result in a closely fitting joint. Similarly, the more terrestrial taxa, i.e. the dasyurids, peramelids, and Thylacinus, like the wombats, have very prominent anconeal processes and very deep olecranon fossae. The extremely deep olecranon fossa that is characteristic of cursors allows more extension of the ulna on the humerus and the strongly developed stops prevent hyperextension at the elbow when the animal is moving at faster speeds (Hildebrand 1982). In contrast, the trochlea notch in W. hilmeri, P. cinereus and species of Trichosurus is shallower and consequently the humerus does not sit as deeply in the ulna. Although the wombat olecranon fossa is relatively deep, its strongly developed anconeal process prevents the elbow from achieving the degree of extension that can be achieved by the fully extended forelimb of P. cinereus. In contrast, the olecranon fossa of the latter is relatively shallow and its anconeal process is not strongly developed. Both these features, coupled with straightened articular facets, allow the Koala elbow to achieve much greater extension (Grand and Barboza 2001). For arboreal animals, increased extension during the swing phase facilitates reaching for new supports (Argot 2001). The relatively small anconeal process, shallow olecranon fossa and open trochlear notch of W. hilmeri suggest a large range of flexion-extension was possible at the elbow. The moderate development of the lateral lip of the posterior surface of the trochlea indicates some stabilisation of this joint. Similar humero-ulnar articulations are evident in arboreal species.

Significantly, the remains of the articulation for the radius on the W. hilmeri ulna appears to have a flat surface, is positioned laterally relative to the shaft, and is directed anterolaterally. This orientation is similar to that in species of Thylacoleo, Trichosurus and Spilocuscus but not as lateral as that of P. cinereus. In contrast, the facet in Vombatus and Lasiorhinus species is broad, concave and anteriorly-directed. In Macropus agilis and Dendrolagus scottae the radial facet is directed anterolaterally, the facet in the latter being relatively large. In Sarcophilus harrisii and species of Dasyurus the facet has a less lateral position in relation to the diaphysis, and is directed more anterolaterally. The facet for the radius in Perameles nasuta and Thylacinus cynocephalus is directed anteriorly. The position of the radius 320

Wakaleo Thylacoleo Vombatus Lasiorhinus Phascolarctos Spilocuscus pitikantensis carnifex ursinus latifrons cinereus maculatus

Trichosurus Notoryctes Dendrolagus Macropus Sarcophilus Dasyurus Perameles vulpecula caurinus* scottae agilis harrisii maculatus nasuta

Fig.13.4 : Anterior view of radii of thylacoleonids and selected marsupials. Not to scale. * lateral view, after Warburton 2003. 321 relative to the ulna was used by Rauscher (1983) to identify cursors and scansors. Although no ulna was available for analysis in Rauscher's study, Wakaleo pitikantensis was deemed to lack the specialised cursorial characteristic (i.e. radius anterior to the ulna) based on the relatively lateral location of the capitulum. Rauscher indicated that the climbing character (a laterally placed radius with a curved notch on the ulna allowing rotation of the radius on the ulna) was indeterminate or possibly present in W. pitikantensis. The size and orientation of the radial facet on the ulna fragment of W. hilmeri suggests it was more likely a climber or scansor than a cursor.

Radius The radius of W. pitikantensis has been described by Rauscher (1983, 1987). Its relatively straight diaphysis with a slight anterior bowing most closely resembles that of Trichosurus and Dendrolagus species (see Fig. 13.4). The radii of Spilocuscus, Vombatus, Lasiorhinus, Phascolarctos and Perameles species show greater curvature of the shaft. It has been suggested that bowing of the radius probably improves the supination and pronation capabilities of the forearm muscles (Oxnard 1963). The radii of the vombatids and Phascolarctos also show an increase in anteroposterior depth distally and are more robust than W. pitikantensis. The radii of the dasyurids are straighter than W. pitikantensis but are similar in exhibiting a small muscle scar on the posterodistal margin in the region of attachment of the pronator teres muscle.

Macleod and Rose (1993) found that a circular radial head is characteristic of arboreal mammalian carnivores while scansorial carnivores may exhibit more elliptical outlines of the radial head. These authors suggest that a circular or subcircular head allows a greater range of rotation for the radius and is associated with manual dexterity. The radial head in W. pitikantensis is nearly circular and the articular surface is concave. This general morphology is seen throughout the comparative species except in P. cinereus, which has a slightly more ovoid outline. The proposal of Macleod and Rose (1993) supports an arboreal hypothesis for W. pitikantensis but it does not correlate with the locomotory modes of other marsupial species included in this study. This discrepancy suggests that this particular 322 morphological character within marsupials may reflect a stronger phylogenetic signal rather than a functional one.

The distal articular surface of the radius of Wakaleo hilmeri resembles that of W. pitikantensis; both bear a moderate styloid process and a relatively broad transverse convex ridge that runs from the base of the styloid process to the lateral edge. Posterior to this ridge, the posterior surface is slightly recessed and forms an ovoid- shaped shelf or ledge. A convex transverse distal articular surface is also found in Lasiorhinus, while that of Vombatus, Dasyurus and Sarcophilus species are slightly flatter. The distal articular surface of Wakaleo vanderleueri shows greater lateral thickness and also has a gently convex, almost flat surface similar to that in Vombatus. None of the comparative taxa displays a posterior ledge similar to that of W. pitikantensis. The distal articular surfaces in Phascolarctos, Trichosurus, Spilocuscus, Perameles, Thylacinus and Dendrolagus species lack a transverse convex ridge and tend to be concave. The greater similarity of the distal radial facet to those of more terrestrial and relatively heavier taxa may suggest that the wrist joint of both W. pitikantensis and W. hilmeri may have been subject to more weight- bearing activity.

The distal articular surface of the radius primarily articulates with the scapholunar bone. This bone is not known for Priscileo or Wakaleo and consequently the articulations at the proximal carpal joint cannot be observed. Rauscher (1983) suggests that if W. pitikantensis was a digger, the distal ledge on the posterior margin of the carpal articular surface may have been a stop preventing hyperextension of the wrist. However, in hyperextension the scapholunar would tend to move extensad, away from the posterior margin of the radius. In flexion, the scapholunar slides flexad and hence it is more likely that the ledge on the W. pitikantensis radius provided a limit to flexion.

Wrist and Manus Yalden (1972) indicates there are three main types of movement possible in the wrist of mammals: flexion-extension, or a hinge-like movement in the sagittal plane; 323 radio-ulnar deviation or medio-lateral movement in the transverse plane; and axial rotation or supination-pronation. Depending on its lifestyle, an animal may show specialisation to one particular movement at this joint (for example ungulates are mostly restricted to flexion) while others, such as climbers, may show reasonable deviations in all three directions. The movements between the carpal bones are dependent upon the shapes of their surfaces, and hence their movements relative to one another can be predicted (Yalden 1970). However, as Yalden’s (1970, 1972) studies show, particular movements at the wrist often involves separate or combined movements of the bony elements. Unfortunately, only two of the wrist bones of W. hilmeri are available for study (hamatum and trapezium) and consequently speculation regarding the movements at this joint at this stage is rather limited and tenuous. The only carpal bone known for W. pitikantensis is a left capitatum (Rauscher 1987).

Comparison of the W. hilmeri wrist elements to those of the closely related T. carnifex indicates significant differences between the two species, those of the latter being highly derived in comparison. In fact, the carpals of both T. carnifex and V. ursinus show the least similarities to those of W. hilmeri and strongly suggest that the wrist and hand of the latter functioned in a different manner to those of the former and possibly reflects a reduced emphasis on weight bearing by these structures. Rather, the strong morphological similarities to the elements of P. cinereus and species of Trichosurus suggest similar functioning of the carpal elements and actions of the wrist - i.e. the manus was being used to grasp food items and possibly branches or tree limbs.

In general, the hamatum of W. hilmeri shows a high degree of similarity to that of Trichosurus and Phascolarctos species, especially in the proportions of length to width and in the strongly convex cuneiform facet. In contrast, the hamatum of the wombats are more anteroposteriorly compressed while that of T. carnifex is strikingly so. These taxa also have a flatter cuneiform facet. Yalden's studies (1972) of the carpal bones of arboreally adapted animals showed that increased movement at the wrist is achieved by increasing the curvature of the surfaces involved. In 324

Trichosurus and Phascolarctos species, at the proximal joint the scapholunar and hamatum have a strongly convex surface and the cuneiform is reduced allowing a relatively large range of movement (60o of flexion-extension for P. cinereus; Yalden 1972). The similarly shaped cuneiform surface on the hamatum of W. hilmeri suggests that it may have also had a large range of movement and, that at least at the proximal carpal joint, had a range of movement that was greater than that of wombats and T. carnifex.

Movements of the pollex The movements of the pollex were determined by examining the shape of the articular facets and the range of movements between the bones at the carpometacarpal and metacarpophalangeal joints. The distinct dorsally-concave, ventrally-convex shape of the trapezium/MCI (first metacarpal) joint in W. hilmeri gives it a semi-saddle shape which allows flexion-extension, some medial and lateral rotation, but little or no abduction or adduction of the metacarpal. However, the medial curvature of MCI and the medial orientation of the metacarpophalangeal joint in W. hilmeri improve the degree of abduction that is possible between the thumb and the rest of the hand. With the metacarpal flexed and medially rotated, flexion of the proximal phalanx would bring the distal pollex toward the rest of the hand. Napier (1961) classifies this type of movement as pseudo-opposable. Wells and Nichol (1977) in their analysis of the manus of T. carnifex concluded that its pollex was pseudo-opposable and that its manus had the ability to grasp through the opposition of the pollex and pisiform and thus was well adapted to a climbing grasp. Interestingly, although the proximal articular facets on the first metacarpals of W. hilmeri and T. carnifex are very similar, the shape of the adjacent facet on their trapezia is quite different. In contrast to that of W. hilmeri, the articular surface in T. carnifex is more saddle-shaped allowing a large range of abduction as well as flexion-extension with limited rotation at this joint (see Wells and Nichol 1977).

Opposability is a form of prehensility which is dependent upon the presence of a ventrally-concave carpal arch (Napier 1961). Such an arch allows the digits to converge on flexion and diverge on extension (Haines 1958). The arch encloses the 325 long flexors of the digits and its form is maintained by a retinaculum which results in a tunnel-shaped structure. Napier (1961) indicates that the depth of the arch depends on the size and functional importance of the long flexors and that, at least in primates, the deeper the arch, the greater the importance of digital flexion in locomotion. Wells and Nichol (1977) found that T. carnifex possessed a deep carpal arch. An assessment of depth of this structure in W. hilmeri cannot be determined due to the incompleteness of its manus. Thus, in proposing pseudo-opposability in W. hilmeri, the assumption that it possessed a carpal arch deep enough to allow convergence of the digits is implied.

The shape of the trapezium bone of Trichosurus species is extremely similar to that of W. hilmeri, however, the convexity and concavity of the facet for the metacarpal is much weaker. The convex but slightly flattened facet on the metacarpal allows flexion-extension and abduction-adduction but little rotation at this joint. The convergent hand and pseudo-opposable pollex of Trichosurus spp. allows the effective grasping of tree branches. In functional terms, the overall similarities in the metacarpophalangeal joint of W. hilmeri to that of Trichosurus spp. and the relatively long metacarpals of W. hilmeri suggest the possibility that the former possessed a grasp similar to that of Trichosurus species. The ability to grasp, however, is not conclusive evidence of climbing ability because the grasping hand may also indicate an adaptation to hold food items. A clearer picture of the possible functions of the manus of W. hilmeri will only be possible when the morphology of the remaining wrist and phalangeal elements are known.

The MCI in P. cinereus has a strongly hemispherical proximal articular facet which lacks the medial concavity seen in W. hilmeri. Additionally, the corresponding facet on the trapezium in P. cinereus is concavely-shallow. This allows the metacarpal to not only flex and extend but also to abduct, adduct, and medially rotate resulting in an opposable pollex. In Vombatus ursinus, the carpometacarpal joint of digit I is more hinge-like and shows little evidence of an ability of the thumb to oppose the rest of the manus. This joint allows mainly abduction and adduction with some flexion while the MCI and metacarpophalangeal joint show little medial deviation. 326

Sacro-lumbar region Slijper (1946) indicates that in mammal species where the need for mobility is predominant over that for strength, the neural spines tend to be long and caudally inclined in order to provide greater leverage and give room for increased musculature. Conversely, where strength is more important than mobility, the spines are short and cranially directed. Gambaryan (1974) agrees with Slijper’s conclusion and adds that long neural spines of progressively increasing angulation tend to be held together by relatively long supraspinous ligaments which permit greater mobility.

The height and inclinations of the neural spines of the vertebrae appear to be determined by the size and demands of the epaxial musculature and ligaments that attach to them (Slijper 1946). The most favourable direction of the spines is that which is perpendicular to the forces exerted by these muscles and ligaments. The development of the epaxial muscles is in turn influenced by the size of the animal, its posture, type of locomotion, and the mobility of the vertebral column.

The anteriorly-inclined and moderately tall lumbar spines and tall sacral spines of W. hilmeri are similar to those seen in T. carnifex and some species of macropodoids (for example, Macropus agilis, Potorous, Aepyprymnus, pers. observation) although some of the latter show a straightening of the spine of L6 and a reduction in the height of the spine of S1. Although the sacral spines in Trichosurus species show a high degree of similarity to those of W. hilmeri, the neural spines of the lumbar vertebrae differ by having a slightly posterior inclination, except that of L6 which is relatively straight. On the basis of Slijper’s findings, the cranial inclination of the lumbar spines in W. hilmeri suggest they were subject to moderate epaxial (longissimus and spinalis) muscular forces. The moderate length and robustness of the neural spines also suggest an animal of medium to large size.

The inclination of the spines on mammal sacra are affected by opposing forces, i.e. those exerted by the spinal muscles (longissimus dorsi) and those of the hind limb (gluteus maximus, gluteus medius and biceps femoris) which attach to these spines 327

(Slijper 1946). The prevalence of upright or caudally inclined spines of mammalian sacral vertebrae indicate that the leg musculature is equal or of greater importance than those of the longissimus dorsi (ibid). Accordingly, the slight caudal inclination exhibited by W. hilmeri suggests more importance of the muscles of the hind leg over the spinal muscles in this region.

The structural features of the spine and the spinal ligaments and musculature also reflect the mobility of the spine. Gambaryan's (1974) studies of mammalian locomotion suggest that animals that exhibit a supple bipedal ricochet gait (e.g. kangaroos) increase the mobility of the sacro-lumbar region by narrowing the apices of L6 and S1 and reducing or eliminating the supraspinous ligament while developing the interspinal muscles in this region. Perameles nasuta exhibits a more rigid ricochet, having a more rigid lumbo-sacral region of the spine through having an enlarged spinous process on L6, anteroposterior widening of the apexes of the lumbar vertebrae, and anterior-sloping spinous processes on the posterior lumbar vertebrae that lengthens the poorly-elastic supraspinous ligament (ibid). Unfortunately, not only are the apexes of the W. hilmeri lumbar vertebrae missing, but it is impossible to know the degree of development of their spinal ligamentous and intervertebral muscular attachments. However, W. hilmeri shows a strong similarity in the size and direction and pattern of spacing of the neural spines to those of species of Trichosurus and Potorous (i.e. relatively long spines on L6 and S1 separated by a large gap) which suggests mobility of this region. In contrast, the lumbosacral spinous processes of the wombats and Koala are short and anteroposteriorly-lengthened. The lengthened spinous processes reduce the intervertebral distance over which the supraspinous ligament is stretched which results in a more rigid lumbosacral region (Gambaryan 1974). Lateral mobility of the spine is reduced by the development of large horizontal transverse processes. This is well illustrated by V. ursinus, L. latifrons, P. cinereus and ungulates in general (Gambaryan 1974). The transverse processes of L5 and L6 of W. hilmeri although incomplete, are directed anteriorly and appear to be angled anteroventrally, L6 more so than L5. The relative size and orientation of these processes are similar to those in Trichosurus, Dasyurus and Sarcophilus species, although the more anterior processes 328 in Dasyurus maculatus have a stronger ventral orientation. The orientation of the transverse processes appears to be intermediate between that of the wombats and the more ventrally-directed processes seen in some carnivores (Gambaryan 1974) and is suggestive of a relatively flexible lumbosacral region, less rigid than that of vombatids but less flexible than in some carnivores.

Both W. hilmeri lumbar vertebrae also bear anapophyses, that on L5 being particularly well-developed. These accessory processes project posteriorly and increase the stability of the vertebral column by restricting lateral flexion (Argot 2003). Similarly well-developed anapophyses are also present on the lumbar vertebrae of Trichosurus, Phascolarctos and Dasyurus species but are absent in Vombatus, Sarcophilus and Perameles species.

Elftman (1929) states that the shape of the pelvis reflects its various functions, i.e. static support, lever-arms for muscles, and the type of diet (herbivores have wider trunks and thus a wider pelvis). The shape of the ilium is "…determined chiefly by the sizes of the three muscle masses whose areas of origin form its three borders - the erector spinae mesially, the gluteus medius and the gluteus minimus dorsolaterally and the iliacus ventrolaterally" (Elftman 1929, p.221). He illustrates the differences in shape by contrasting the ilia of Phascolomys (i.e. Vombatus) against that of Perameles. Phascolomys (sic Vombatus) has small erector spinae but large and sub- equal gluteus medius and iliacus muscles which results in an ilium flattened in the horizontal plane. In contrast, the iliacus in Perameles nasuta is comparatively smaller and the gluteus medius and erector spinae are both large, which results in a flatter ilial surface but one that is more vertically oriented. When the ilia of W. hilmeri are articulated with the sacrum, like Perameles, Trichosurus, Spilocuscus and Sarcophilus species, they have a relatively vertical orientation. The ilium of T. carnifex is more horizontally oriented, and is similar in this respect to Vombatus ursinus although it lacks the anterior widening seen in the latter.

The angle the ilia make with the spinal column in the sagittal and horizontal planes is a strong indicator of the size of the erector spinae; the wider the horizontal flare of 329 the iliae and the greater the dorsal angle the iliae make with the sacrum, the greater area there is for attachment of the erector spinae (Elftman 1929). Taylor (1976) notes that the ilia of obligate terrestrial viverrids flare anteriorly while those of scansorial or arboreal species tend to be narrow and subparallel. Unlike kangaroos and bandicoots, in dorsal view, the ilia of W. hilmeri do not flare greatly with respect to the vertebral column and in lateral view are slightly dorsally inclined to the vertebral column, which suggests that the erector spinae may have been moderately developed but not to the extent seen in those groups. The angle of the W. hilmeri ilia is similar to those of Trichosurus, Spilocuscus and Sarcophilus species. The ilia of P. cinereus and the vombatids exhibit the greatest development of lateral flaring. Flaring of the ilia also partly reflects the width of the trunk, and the relatively limited flare shown by the ilia of W. hilmeri suggests that the trunk was not particularly wide. A narrow trunk may suggest that W. hilmeri was not herbivorous and adds weight to an argument for an omnivorous or carnivorous diet.

Most fossorial mammals have stout sacra which are modified to accommodate the forces involved in the bracing of the hind feet while digging with the forefeet (Hildebrand 1985). These sacra are commonly long, consist of many fused vertebrae, and have ilia that are nearly horizontal (ibid). These characteristics are readily seen in the sacra of L. latifrons and V. ursinus each of which contain four vertebrae. The sacrum of Notoryctes typhlops is quite different to that of any other marsupial and is a highly fused structure reminiscent of the synsacrum of birds. The sacrum of Wakaleo hilmeri is contrastingly short and most closely resembles that of Trichosurus spp; it consists of only two vertebrae, has a tall neural spine on S1 and a slightly shorter spine on S2 and has a deep sacral wing. Although the sacrum of Spilocuscus maculatus similarly consists of two vertebrae, its two neural spines are fused together and the ilial facets are quite dorsally-directed in comparison to those of W. hilmeri or Trichosurus species. The sacra of Phascolarctos, Dasyurus and Sarcophilus species all consist of three vertebrae and generally possess shorter neural spines. In addition, as noted above, the form of the lumbar transverse processes, the height of the neural spines, and the shape and orientation of the ilia all differ greatly 330 from that seen in vombatids and supports the notion of a more flexible sacrolumbar region, and one that does not reflect the demands of a fossor.

In conclusion, the similarity of the lumbar vertebrae, sacrum and ilia of W. hilmeri to those of Trichosurus species suggests that W. hilmeri exhibited similar movements of the trunk, but its large, anteriorly-directed lumbar spines suggest a heavier animal with stronger epaxial musculature.

Allometry and phylogeny Two other primary factors must be taken into consideration when investigating functional morphology, they are allometry (the relative growth of a part of an organism in relation to the growth of the whole) and phylogeny. Within this study, the small sample size of fossil specimens and the very limited number of variables that could be measured makes it difficult to assess the significance of allometry. Measurements of humerus distal width and deltopectoral length indicated very small relative differences (< 4%) between W. pitikantensis and T. carnifex suggesting that within thylacoleonids, as humerus length increases, the size of these features increase proportionally. This relatively isometric allometry suggests that phylogeny may be significant factor affecting thylacoleonid morphology. It is possible to interpret the slight negative allometry in the lateral epicondylar crest height of T. carnifex as indicating adaptation to a more terrestrial lifestyle, terrestrial species appear to have a relatively shorter lateral epicondylar crest (< 30%) than scansorial or arboreal species (see Table 13.1), however, larger sample sizes of both species are required to confirm this allometry and support such a hypothesis.

The common presence of morphological features within certain taxa may also be an indication of phylogenetic relationships. These features may indicate relationships at the level of a crown group or may indicate broader ancestral relationships. Munson (1990) investigated the phylogenetic relationships of vombatiforms using postcranial characters and found that they were united by characters that reflected modifications for increased size and terrestrial locomotion. The most distinctive of these was the modification of the astragulus. Munson (1990) indicates that although the arboreal 331 species Phascolarctos cinereus shares many features with the arboreal possum species, its astragulus shares features with those of other terrestrial vombatiforms. The astragulus (and many elements of the pes) for W. hilmeri and W. pitikantensis are unknown at present, however, on the basis of Munson’s findings it is likely that the Wakaleo astragulus will exhibit the vombatiform features. Munson (1990) found that the morphology of the manus, vertebrae and pes of the vombatiform Ngapakaldia tedfordi was very similar to that of Trichosurus vulpecula and suggested that this reflects the arboreal ancestry of the vombatiform clade. It is also possible that the morphological similarities of the carpal and vertebral elements of W. hilmeri and species of Trichosurus may also reflect the ancestral vombatiform condition.

Generally, the postcranial record (especially that of the manus and pes) for most extinct diprotodontians is poor. This record is also biased taphonomically, with associated postcrania for small species (i.e. small dog size or smaller) being practically absent. Consequently, knowledge of the distribution of postcranial characters in clades of fossil diprotodontians is patchy. Knowledge of the postcranial osteology and functional morphology, particularly of the manus, of many extant marsupials is also lacking. Improvement in the former coupled with an improved record for fossil species is required to provide a better understanding of the influence of phylogeny on postcranial anatomy.

Summary The morphology of most of the elements recovered for W. hilmeri and of the humerus of W. pitikantensis show a strong similarity to those of Trichosurus species. Comparison of the humerus to those of marsupials of different locomotor categories indicates a greater similarity in the proportional lengths of the major crests to those of climbers than fossors or cursors. The morphology of the elbow joint strongly resembles that of Trichosurus species and Phascolarctos cinereus and suggests a large range of flexion and extension. The position of the radius on the ulna is also more similar to that seen in climbers than cursors. The presence of well-developed biceps brachii muscles in W. pitikantensis is suggested by a well-developed bicipital 332 tuberosity on the radius, a deep bicipital groove on the humerus, and a large scapular tuberosity on the scapula. The presence of tuberosities on both the lateral and medial margins of the humeral shaft of W. pitikantensis also suggests a strong work load for muscles that abduct and adduct the arm (deltoid, pectoralis and teres major respectively). Prominent rugosities on the olecranon of W. hilmeri are suggestive of strong triceps brachii musculature. All of these muscles play important roles in moving the forelimb forward or upward, or towards the body. These movements of the forearm are all essential components in climbing. The well-developed distal structures on the humeri of both W. pitikantensis and W. hilmeri hint at strong pronation and supination abilities. The shape of the MCI/trapezium articular surfaces and the medial curvature of the MCI also indicate the possibility of a pseudo- opposable pollex for W. hilmeri, hence there is a possibility of grasping as seen in Trichosurus species.

The lumbo-sacral features of W. hilmeri are also most similar to those of Trichosurus species and show least similarity to those of the vombatids. The lumbar vertebrae are moderately large and robust with anteriorly-directed neural spines suggesting a moderate sized to large animal with strong epaxial muscles at least in the thoracic and anterior lumbar regions. The curved transverse processes on the lumbar vertebrae and tall neural spine on S1 suggest a degree of flexibility in the lumbosacral region. The anteriorly-narrow sacrum and gently curving ilia that lie relatively parallel to the spinal column suggest reduced erector spinae muscles posteriorly and also a relatively narrow trunk. The posteriorly-inclined neural spines on the sacrum indicate an increase in importance of the leg musculature over the spinal muscles in this region, however, the ilial surface and ilial position relative to the spinal column indicate that the muscular requirements were unlike those seen in wombats or the ricochetal macropodoids.

Rauscher (1983) postulated that W. pitikantensis was possibly fossorial or scansorial. The morphology of the postcranial elements presently known for W. hilmeri and W. pitikantensis strongly suggest they may have been arboreal or scansorial animals capable of terrestrial locomotion (much like Trichosurus species) but it is unlikely 333 they were specialised fossors. Although certain features of their humeri are shared with the semi-fossorial wombats, the articular surfaces of the elbow joint are not suggestive of the strongly stable or restricted joint that is typical of burrowing animals. The fragmentary nature of the postcranial remains of these species means that the inferences regarding locomotory behaviour are necessarily speculative. More definitive statements regarding their habitus will only be possible upon the discovery of more of the missing elements especially of the hind leg and pes. 334

CHAPTER 14

BIOSTRATIGRAPHY, BIOCHRONOLOGY, AND PALAEOECOLOGY OF RIVERSLEIGH THYLACOLEONIDS

14.1 Introduction

Marsupial lions have been recovered from Tertiary and Quaternary fossil localities across the breadth of the Australian continent. Many of the Tertiary mammal-bearing deposits lack precise radiometric dates and have been dependent upon “stage of evolution” biochronology or biocorrelation for an estimation of their ages. Woodburne et al. (1985) identified thylacoleonids as one of the few marsupial groups that could be utilised in “stage of evolution” biochronology.

14.2 Riversleigh Geology

Interpretations of the Tertiary limestones at Riversleigh have continued to evolve ever since R.L. Jack first named the fossil-bearing limestone outcrops, the Carl Creek Limestone (Jack 1896). Determination of a clear stratigraphic sequence of the numerous bone-bearing deposits has been difficult because of the lenticular, isolated nature of many of these sites. Brief summaries of the different interpretations of these limestone sediments follow:

R.H. Tedford, then from the University of California, Berkeley, investigated the geology and palaeontology of some of these outcrops (Sites A-D) in 1963 and postulated that these limestone sediments were fluvio-lacustrine in origin (Tedford 1967; see Fig 14.1). On the basis of the stage of evolution of the diprotodontoid marsupials recovered from Site D he proposed a late Oligocene or early Miocene age for the Carl Creek Limestone. 335

Fig. 14.1: Geographic map of the Riversleigh area showing the location of the main fossil-bearing areas: “D Site, Godthelp’s Hill, and Gag plateau. (After Megirian, 1992) 336

Since the early 1980s, investigation of the limestone plateaux surrounding Tedford’s Site D by M. Archer and his research team from the University of New South Wales has resulted in the identification of more than 200 distinct fossil localities (Archer et al. 1989, 1991, 1997). These deposits have been interpreted as representing a number of different depositional environments including fluvio-lacustrine, cave sediments and fissure fills (Archer et al. 1989). Archer et al. (1989, 1991, 1997) have allocated the majority of these fossil localities to one of three different ‘Systems’ (A, B, or C) depending upon their superposition, regional clustering, lithology, stratigraphy and biocorrelation. Each ‘System’ has been interpreted as representing a different period of geological time: System A is regarded to be late Oligocene to early Miocene in age; System B, early to mid Miocene in age; and System C is tentatively regarded as middle to early late Miocene, i.e. approximately 16 to 9 Ma (Archer et al. 1997). Archer et al. (1989) allocate Tedford’s Site D locality to System A and only apply the term ‘Carl Creek Limestone’ to the primary carbonate sediments associated with this System.

Creaser (1997) recognises three sedimentary sequences (Verdon Creek, Godthelp’s Hill, and Gag Plateau: see Fig 14.1) within Riversleigh’s Tertiary limestones and documents the topographic heights of the sediments and many of the fossil sites within these sequences. System A sites are identified as mainly occurring in the Verdon Creek sequence, System B sites in the Godthelp Hill sequence and System C sites dominate the Gag Plateau although System A sites may also be present on the latter plateau. This author has advanced a four-stage cycle of sedimentation and erosion to account for the various depositional environments (alluvial, lacustrine, travertine, cave deposits, tufa and karst) recognised within the Tertiary limestones. The stages of this cycle are: 1. uplift and or lowering of the water table; 2. erosion and development of a karst landscape; 3. subsidence or raising of the water table, and; 4. sediment accumulation within the karst environment. Creaser (1997) proposes that this cycle has occurred three times during the Oligocene-Miocene at Riversleigh. 337

On the basis of lithology, Megirian (1992) interprets the Tertiary limestones at Riversleigh as representing a single unit, the Carl Creek Limestone. He postulates that these sediments were formed in tufa and alluvial fan conditions. Megirian (1994) has criticised the ‘Systems’ terminology because it lacks a satisfactory definition, one that requires more precise lithostratigraphic and biochronologic values.

Arena (2005) has revised the interpretation of the Carl Creek limestone. He proposes another model to explain the formation of the Tertiary carbonates of Riversleigh. On the basis of geological processes that are characteristic of karst systems and lithological analyses of a number of the Riversleigh fossil localities, he proposes that there are two types of carbonate lithofacies present at Riversleigh: 1, a primary heterogeneous carbonate ‘host’ facies; and 2, a cave facies that includes speleothems and cave sediments. The heterogeneous carbonate facies is interpreted as a series of tufaceous deposits formed in a fluvial barrage system. Barrage ponds are regarded as low energy systems that would provide the depositional environment suitable for the formation of the heterogeneous carbonate facies as well as account for the unworn nature of the vertebrate fossils. To overcome the criticisms associated with the ‘Systems’ terminology and to assist in the interpretation of the complex relationships of Riversleigh’s geology and fossil faunas, Arena (2005) proposes a new biostratigraphic concept that combines two key elements, ‘depositional phase’ and ‘faunal zone’. The depositional phase represents a period of geological history diagnosed on the basis of geology, taphonomy and characteristics of the fossil faunas. Faunal zones are “….comprised of discrete local faunas (collected from individual localities) that are considered to belong to the same span of time” (Arena 2005, p.29). The Riversleigh faunal zones A, B and C proposed by Arena (2005) correspond with the time periods advanced for Systems A, B and C of Archer et al. (1989, 1994, 1997), i.e. Faunal Zone A corresponds with the late Oligocene, Faunal Zone B with the early Miocene and Faunal Zone C with the middle to early-late Miocene.

The Systems terminology is utilised in this analysis because the Systems terminology has been used throughout nearly all previous Riversleigh studies whether they be 338 systematic, phylogenetic, biocorrelative, biostratigraphic or taphonomic and facilitates comparisons with that body of research.

14.3 Riversleigh Biostratigraphy and Biocorrelation

Commonly, the relative ages of Riversleigh sediments are determined by their sequence of deposition or stratigraphy, such that older sediments are overlain by younger sediments. Although topographic data have been gathered for most of Riversleigh’s sites (Creaser 1997, Arena 2005), Arena (2005) indicates that the tufagenic nature of many of these deposits prevents the use of the laws of superposition to determine their relative ages. In tufagenic systems “…deposits of the same age can form at different levels in the profile, deposits of different age can occur at the same level and deposits of younger age can form lower in the profile than older deposits” (p.243, Arena 2005). Consequently, Arena (2005) suggests that vertebrate biocorrelation is a more reliable method of determining the relative ages of the Riversleigh deposits. The phyletic succession of species of Wakaleo, in particular, has the potential to provide a useful tool in assessing the relative ages of the Riversleigh deposits.

The distribution of marsupial lion taxa in sites from Riversleigh is shown in Table 14.1. Looking at the occurrence of thylacoleonids at Riversleigh in terms of the Systems concept and in relation to the three sedimentary sequences outlined by Creaser (1997), Wakaleo hilmeri occurs in the Verdon Creek sequence at Rackham’s Low Lion (Low Lion), Dirk’s Towers, Burnt Offering, and Neville’s Garden Sites. Low Lion Site is regarded as a System A deposit (Archer et al. 1997) but the position of Burnt Offering and Dirk’s Towers as either System A or System B sites is unresolved (Creaser 1997). Neville’s Garden Site is interpreted to be a tufa deposit and is assigned to System B on the basis of its macropodoids (Cooke 1997) and diprotodontoids (Black 1997). The possible intermediate form of Wakaleo (W. sp. cf. W. hilmeri) occurs at D Site. Creaser (1997) indicates that the Verdon Creek sequence consists of a basal conglomerate that is up to 20 meters thick and is overlain by the three metre deep, homogeneous D Site (Tedford’s Site D) limestone. 339

Table 14.1: Distribution of marsupial lion species in Riversleigh sites. Site abbreviations: LL = Rackham’s Low Lion; D = D Site; WH = White Hunter; Hia = Hiatus; BO = Burnt Offering; DT = Dirk’s Towers; CR = Creaser’s Ramparts; NG = Neville’s Garden; CS = Camel Sputum; Upp = Upper; Hel = Helicopter; KCB = Keith’s Chocky Block; COA = Cleft of Ages; JJ = Jim’s Jaw; GS = Golden Steph; HH = Henk’s Hollow; JC = Jim’s Carousel; Enc = Encore.

? ? SYSTEM A SYSTEM B SYSTEM C C + uncertain uncertain sequence VC GH VC GH VC GH GH Gag Gag Gag LL D WH Hia BO DT CR NG CS Upp Hel KCB COA JJ GS HH JC Enc L. roskellyae L. deminutiv. W. hilmeri W. cf. hilmeri

SPECIES W. oldfieldi W.vanderleu. Microleo VC = Verdon Creek sequence; GH = Godthelp Hill sequence

Within the Verdon Creek sequence, W. hilmeri appears to correlate with the thick basal conglomerate. The presence of the possible transitional species W. sp. cf. W. hilmeri in the overlying D Site sediments suggests the possibility of a slightly later deposition for those sediments.

Within the Godthelp Hill sequence (including the Hal’s Hill area) W. hilmeri occurs at Hiatus Site, White Hunter Site, Upper Site, Camel Sputum Site and Creaser’s Ramparts Site. Hiatus Site and White Hunter Site are regarded to belong to System A while Upper Site and Camel Sputum Site are regarded to be part of System B (Creaser 1997). Creaser (1997) indicates that the age of Creaser’s Ramparts Site is uncertain. The occurrence of W. hilmeri in both System A and System B sites fails to resolve the System association of either Burnt Offering, Dirk’s Towers or Creaser’s Ramparts Sites.

Within the Godthelp Hill sequence, the more derived taxon Wakaleo oldfieldi is present at Helicopter Site and indicates a period of later deposition relative to the 340

Verdon Creek sequence. Wakaleo oldfieldi is the only species of Wakaleo to have been recovered from the northern section of the Gag Plateau sequence. In this region it has been found at Jim’s Jaw Site, Golden Steph Site and Henk’s Hollow Site. All these sites are regarded to belong to System C. Creaser (1993) indicates that the stratigraphy of the southern section of the Gag Plateau is complex and is difficult to correlate with sites on the northern section of the plateau or D Site plateau. Both Cleft of Ages (C.O.A.) and Keith’s Chocky Block (K.C.B) Sites are found in this region of the Gag Plateau. Both sites have a distinctive brown lithology and are interpreted to be cave fills (Arena 2005). The presence of W. oldfieldi in both suggests they are similar in age and suggests age equivalency with Jim’s Jaw, Golden Steph and Henk’s Hollow Sites in the northern section of the Gag Plateau.

The southern end of the Gag Plateau is also the location of the single site (Encore Site) that has produced material referrable to Wakaleo vanderleueri. This site has been interpreted as a tufa deposit that has cut into the underlying Tertiary limestone (Creaser 1997) and indicates a later period of deposition.

Lekaneleo roskellyae is distributed across all three sedimentary sequences. It is present in the Verdon Creek sequence at Dirk’s Towers Site (System B) and in the Godthelp Hill sequence at White Hunter (System A), Upper and Camel Sputum Sites (System B). In the Gag Plateau sequence it is present in both Jim’s Jaw and Henk’s Hollow Sites, both of which have been allocated to System C. It is also present in the cave fill deposit, Cleft of Ages Site. This species has also been recovered from Jim’s Carousel Site, a possible tufa deposit whose age is uncertain (Creaser 1997). The occurrence of L. roskellyae in sites across all three Riversleigh Systems and in sites that have yielded both plesiomorphic and derived species of Wakaleo, suggest that it was a temporally long-ranging taxon. Within both Wakaleo and Thylacoleo there is a trend of increasing size (characteristically seen in P3 length) through time. Whether a similar trend is apparent within the long-ranging L. roskellyae is not possible to determine because of the inadequate sample size for this species. It is possible that a similar trend might occur within L. roskellyae because the three P3 specimens of this species (see Appendix 1, Table B) show an increase in size through time, that from 341 the oldest site (White Hunter) is the shortest, that from the youngest site (Jim’s Carousel) is the longest, and the P3 from the intermediate-aged site (Upper) lies in- between, being 8% longer than the White Hunter specimen and 15% shorter than the Jim’s Carousel specimen. A significant increase in sample size is needed to confirm this trend.

Lekaneleo deminutivus and Microleo attenboroughi occur at similar levels in the Verdon Creek sequence, at Dirk’s Towers Site and Neville’s Garden Site, respectively. As indicated above, Neville’s Garden Site is regarded to be a System B site while the age of Dirk’s Towers Site is uncertain (Creaser 1997). Although the singular occurrence of both these species reduces their usefulness for biocorrelation at an inter-System or broader continental level, their presence may assist in assessing the faunal diversity of Riversleigh’s sites and Systems.

In summary, Lekaneleo roskellyae has a very broad distribution and is present in sites from all three Riversleigh Systems as well as possible post-System C sites. The relatively enigmatic species L. deminutivus is confined to a single site assigned to System B. The smallest and most plesiomorphic species of Wakaleo, W. hilmeri, is present in numerous sites in System A and System B but is absent from System C sites. The larger and more derived species, W. oldfieldi, is present in System C sites and possibly some System B sites but is absent from System A. All specimens of the more derived W. vanderleueri, have been recovered from Encore Site which is estimated to be early late Miocene (9myr) in age (Myers et al. 2001). The new thylacoleonid taxon M. attenboroughi is known from a single System B site, Dirk’s Towers Site.

Does the distribution Riversleigh’s thylacoleonids support the age estimations hypothesised by Archer et al. (1989, 1991, 1997)?

On the basis of the “stage of evolution” of species of Wakaleo, the pattern of distribution of the species of Wakaleo generally shows agreement with the relative ages of the “Systems” proposed by Archer et al. (1989, 1991, 1997). The most 342 plesiomorphic species, W. hilmeri, is most common in what are regarded as the oldest deposits of System A and B (Verdon Creek sequence and Godthelp’s Hill sequence, respectively) and is absent in later deposits (System C, Gag Plateau sequence). The presence of that species in both System A and B, however, does not provide evidence in support of a possible temporal difference between those two Systems and, it is possible that sites in these two Systems, at least on the basis of marsupial lion evolution, may be similar in age. The most derived species, W. vanderleueri, is present in the youngest pre-Pliocene deposit (post-System C), Encore Site, and the intermediate species (W. oldfieldi) is most common in sites that are older (System C and possibly System B) than Encore but younger than those yielding W. hilmeri.

The presence of the transitional species, Wakaleo sp. cf. W. hilmeri, in D Site, a System A deposit, suggests there is some temporal overlap between that species and W. hilmeri. The chronology of species of Wakaleo predicts the likely presence of W. oldfieldi in sediments younger than those yielding W. sp. cf. W. hilmeri (System A), i.e. in System B sites (see Fig. 14.2).

14.4 Intracontinental correlations

Two species of marsupial lions identified from Riversleigh, Wakaleo oldfieldi and Wakaleo vanderleueri, have been recovered from other localities. Although intracontinental correlation between sites yielding W. vanderleueri is rather straightforward, correlation of sites that have yielded W. oldfieldi is more complicated because of the conflicting age estimations and variability in the faunal compositions of these sites.

Both Wakaleo oldfieldi and the vombatid Rhizophascolonus crowcrofti are shared by Riversleigh’s C.O.A. Site and the Leaf Locality of the Wipajiri Formation (Fm), South Australia, suggesting age equivalence of these deposits. The estimated age for the Kutjamarpu Local Fauna from the Leaf Locality is currently a matter of debate. On the basis of the stage of evolution of its diprotodontoids, Stirton et al. (1967) 343

Lake Eyre Basin Epoch Riversleigh sites Local Faunas Northern Territory Wakaleo sp. 6

Alcoota W. alcootaensis 8

L. Miocene L. Miocene Encore W. vanderleueri 10.4

Bullock Creek 12

14

M. Miocene M. Miocene System C 16.4

18 W. oldfieldi Kutjamarpu COA 20 Rhizophascolonus crowcrofti Wakiewakie lawsoni 22 E. Miocene Ektopodon serratus System B Emuarius gidju Neohelos tirarensis 23.5 W. sp. cf. hilmeri 24 Ngama WhiteWhite HunterHunter W. hilmeri Ngapakaldi W. pitikantensis System A 26 Ditjimanka

28

29.3 L.Oligocene

Fig. 14.2 : Correlation of Tertiary deposits based on stage of evolution of Wakaleo species, biocorrelation of other marsupial species and superposition, following Archer et al. (1997). Species shared by the Kutjamarpu Local Fauna and Riversleigh System B sites are also listed. 344 initially postulated the Wipajiri sediments to be Miocene in age. Woodburne et al. (1985) proposed a middle Miocene age. Subsequent biostratigraphic and magnetostratigraphic studies of the underlying Etadunna Formation resulted in a revision of the age of the Wipajiri Fm to possible late Oligocene (24 Ma; Woodburne et al. 1993). In contrast, Archer et al. (1993) proposed an early Miocene age for the Wipajiri Formation on the basis that it cuts unconformably into and overlies the late Oligocene Etadunna Formation and because it shares a number of species (Emuarius gidju, Rhizophascolonus crowcrofti and Wakiewakie lawsoni) with Riversleigh’s System B sites. However, as documented above, System B sites have yielded the plesiomorphic species W. hilmeri. The phyletic succession of Wakaleo predicts that W. oldfieldi is likely to occur in sites that are younger than the System B sites that yield W. hilmeri, i.e. System C sites. This prediction finds support with the occurrence of W. oldfieldi in Jim’s Jaw, Golden Steph and Henk’s Hollow Sites, all System C Sites. The presence of W. oldfieldi in System C sites and in the Kutjamarpu LF (which appears to correlate with System B) suggests that C.O.A Site may lie at the transition of Systems B and C. Using the time frames proposed by Archer et al. (1994, 1997) for these Systems, if System B sites are regarded as early Miocene and System C sites as middle Miocene in age, C.O.A. site may be late early Miocene in age (see Fig. 14.2).

The Leaf Locality shares a further five taxa with Riversleigh sites; these include Ektopodon serratus, Litokoala kutjamarpensis, Wakiewakie lawsoni, Neohelos tirarensis and Emuarius gidju (see Fig. 14.2). At least four of these species (and possibly all five) are found in System B sites while two of the species (N. tirarensis and E. gidju) also occur in System A and C sites. The relatively strong correlation of the Leaf Locality to System B sites as well as to some System C sites suggest that it lies at an intermediate stage relative to these two Systems and hence is possibly late early Miocene or early middle Miocene in age. This age estimate clearly conflicts with the late Oligocene estimate of Woodburne et al. (1993). Further support for an early Miocene age for Riversleigh’s System B sites is provided by the presence in Upper Site of Hydromops riversleighensis, a species of molossid bat whose closest 345 relative is the middle Miocene taxon H. nonhengensis from Thailand (Hand et al. 1997).

If a late Oligocene age (as proposed by Woodburne et al. 1994) is accepted for both the Leaf Locality and C.O.A. Site (and System C sites that yield W. oldfieldi), the phyletic succession of Wakaleo would suggest that the sediments of Riversleigh’s System A and B sites (yielding W. hilmeri) are early late Oligocene in age or older. A late Oligocene age for W. oldfieldi would indicate that it was a temporally long- ranging species, occurring until approximately the late middle Miocene where it was replaced by the more derived species W. vanderleueri (Encore LF and Bullock Creek LF). This age estimation for Riversleigh’s sediments would also suggest that Lekaneleo roskellyae, which is also present in C.O.A. Site and sites in all three Systems, spanned the early late Oligocene or earlier, to at least the early Miocene.

Correlations also occur between Riversleigh and the late Oligocene sediments of the Etadunna Fm of central Australian (Woodburne et al. 1993). White Hunter Site, a System A site that has yielded many specimens of W. hilmeri, also shares the ilariid Kuterintja ngama with the Ngama Local Fauna, Mammal Zone D of the Etadunna Formation, South Australia (Myers and Archer 1997; Woodburne et al. 1993). Further correlation of White Hunter Site with the Etadunna Formation is provided by the shared presence of the diprotodontoid Ngapakaldia bonythoni (Ngapakaldi Local Fauna, Mammal Zone C; K. Black pers. comm.; Woodburne et al. 1993). These correlations lend support to a late Oligocene age for White Hunter Site.

On the basis of the morphoclinal nature of the Wakaleo lineage, Wakaleo hilmeri is considered to be slightly younger than Wakaleo pitikantensis which is a component of the Ngapakaldi LF, Faunal Zone C of the Etadunna Fm (Chapter 4). A late Oligocene age (25.5 – 25 m.y.) for the Ngapakaldi LF (Woodburne et al. 1993) suggests that Riversleigh’s System A and B sites that yield W. hilmeri are slightly younger than this fauna but are older than the Kutjamarpu LF. 346

Wakaleo vanderleueri, a component of Riversleigh’s Encore LF, has also been recovered from the Bullock Creek LF in the Northern Territory. The age of the latter is estimated to be middle Miocene on the basis of stage of evolution of diprotodontoid taxa (Plane and Gatehouse 1968; Stirton et al. 1968; Woodburne et al. 1985). Other mammalian taxa from the Encore LF, for example, the palorchestid Palorchestes anulus and the macropodoid Rhizosthenurus flanneryi, are more derived than related forms from the Bullock Creek LF (e.g. Propalorchestes novaculacephalus) yet are structurally primitive in comparison to closely related species from the late Miocene Alcoota Local Fauna (i.e. Palorchestes painei and Hadronomas puckridgi, respectively: see Black 2006; Kear 2002, Kirkham 2004). On this basis, Myers et al. (2001) suggest the Encore LF is older than the Alcoota LF, (approx. 8-7 Myr; Murray and Megirian 1992) but younger than the Kutjamarpu LF and Riversleigh System C local faunas (approx. 15-10 Myr) and propose an early late Miocene (c.9 Myr) age for the fauna.

Problems in correlating Riversleigh sites and Systems Allocation of particular sites to particular Systems is problematical when they yield species that are found in other sites allocated to different Systems. For example, the position of Cleft of Ages site as either a System B or System C site is uncertain (Creaser 1997). The presence of W. oldfieldi in this site and in others considered to more certainly represent System C lends support to its allocation to that System. However, Cooke (1997) indicates that macropodoid species recovered from Cleft of Ages Site correlate with System B. The presence of a relatively plesiomorphic species of Neohelos from this site also suggests a possible System B association (Black pers. comm.). Hence, it is regarded here that there should be uncertainty regarding the age of Cleft of Ages Site.

In light of the tufagenic depositional model advanced for the limestone deposits of Riversleigh (Arena 2005), better resolution of the ages of sites of individual Riversleigh sites rather than “Systems” as a whole would be useful. Radiometric dating of some Riversleigh sites is currently being attempted and may provide a more accurate estimate of the ages of Riversleigh’s local faunas. 347 Pleistocene 1.75-10,000 5.3-1.75 Pliocene 10.4-5.3 L. Miocene 16.4-10.4 M. Miocene 23.5-16.4 E. Miocene 26-23.5 L. Oligocene sp. cf. W. hilmeri Thylacoleo crassidentatus Wakaleo alcootaensis Lekaneleo deminutivus Lekaneleo roskellyae Microleo attenboroughi Wakaleo oldfieldi Wakaleo vanderleueri Thylacoleo carnifex Thylacoleo hilli Wakaleo hilmeri Wakaleo pitikantensis Wakaleo pitikantensis Wakaleo

Thylacoleonidae

Fig. 14.3 Diversity and probable interrelationships of thylacoleonids through time. 348

14.5 Thylacoleonid diversity and palaeoenvironmental indicators

Thylacoleonid diversity appears to be conservative across time (see Fig. 14.3). Across much of the Riversleigh sequence (in particular System A and C), the diversity appears consistently low (two species). Highest thylacoleonid diversity at an individual site occurs at Dirk’s Towers Site (three species; W. hilmeri, L. roskellyae and L. deminutivus), a site allocated to System B. The recovery of Microleo attenboroughi from another System B site (Neville’s Garden) indicates a relatively high diversity for System B in comparison to other Systems. In post System C sites such as Encore, diversity decreases to a single species, W. vanderleueri. This low diversity (one or two species) continues through the late Miocene (W. alcootaensis) and into the Pliocene (T. crassidentatus and T. hilli) and Pleistocene (T. carnifex).

Australia’s climate during the Tertiary has been reconstructed using data from the marine foraminiferal record in conjunction with oxygen isotope (sea water temperatures), sea level and tectonic information, fossil pollen and macrofloras and taxic/habitat associations (Frakes et al. 1987; McGowran and Li 1997; McGowran et al. 2000). Chronological correlations of this data indicate that from the late Paleocene to the Quaternary the climate fluctuated between ‘icehouse’ conditions and ‘greenhouse’ conditions. Icehouse conditions are typified by cool, dry, continental climates with lower sea levels and relatively low biological diversity. Greenhouse conditions are typified by a warm, humid, maritime climate with relatively high sea levels and high biological diversity (Frakes et al. 1987). Icehouse conditions characterised the middle Eocene, the late Oligocene, late middle Miocene, late Miocene, mid to late Pliocene and Quaternary. Greenhouse conditions characterised the late Paleocene to early Eocene, the late middle to late Eocene, the early Miocene, early middle Miocene and the early Pliocene (Frakes et al. 1987; McGowran and Li 1997; McGowran et al. 2000).

Archer et al. (1989) interpret Riversleigh’s sites from System A to C as representing rainforest communities on the basis of the high family-level and species-level 349 mammalian diversity, high number of browsers and lack of grazers, and obligate rainforest indicator genera such as Hypsiprymnodon, Pseudocheirops, Menura and Orthonyx (Boles 1997, 2000). Further assessment of the assemblages within each System has resulted in a refinement of the palaeoenvironmental interpretation of each System such that System A has been regarded as representing mesic forests and System B assemblages as representing rainforest (Archer et al. 1997). Lower System C assemblages are interpreted as rainforest and upper System C assemblages are thought to represent relatively drier, more open forests (ibid). In contrast, on the basis of the geochemical and physical conditions required for the formation of the limestone sediments, Megirian (1992) interprets the climatic conditions at Riversleigh to have been semi-arid and postulates that the fossil assemblages represent a geographically very restricted rainforest refugium (riparian in nature), confined to perennial streams and whose diverse assemblages are a mixture of rainforest species and more distant, mesically-adapted species that travelled into the Riversleigh area to drink.

Although thylacoleonid diversity could never be regarded as particularly high, their pattern of diversity across Riversleigh’s Systems tends to correlate with the climatic model of cyclical greenhouse and icehouse conditions proposed by Frakes et al. (1987), and McGowran and Li (1997) for the Tertiary and Quaternary. In terms of the Riversleigh ‘Systems’ the highest diversity of marsupial lions at Riversleigh is associated with System B. General vertebrate diversity is also relatively high within System B sites, in particular Upper Site, Dirk’s Towers Site, Wayne’s Wok Site, Neville’s Garden Site and Camel Sputum Site (Archer et al. 1989, 1997). Crosby (2002) documents that phalangerid diversity is greatest in System B sites. Cooke (1997) notes that macropodoid diversity is high in System B sites as well as System A sites. This high diversity is suggestive of greenhouse conditions and lends support to an early Miocene age for System B sites.

The lower thylacoleonid diversity observed in System A, System C and post System C sites possibly reflects the cooler icehouse conditions correlated for the late Oligocene (System A) and late middle, and late Miocene (System C and post System 350

C). The lower thylacoleonid diversity in System A corresponds with the overall lower vertebrate diversity recorded for sites from this System in comparison to those from System B and C (Archer et al. 1997).

The recovery of both L. roskellyae and W. hilmeri from three sites at Riversleigh suggests that these species were sympatric. These sites are found in both System A and B and suggest a coexistence that may have spanned the late Oligocene (or earlier) to the middle Miocene. Lekaneleo roskellyae also exhibits temporal overlap with the larger, more-derived species of Wakaleo, W. oldfieldi, both species occurring in C.O.A. Site. Thylacoleonid sympatry has also been documented for the early to mid Pliocene species Thylacoleo crassidentatus and T. sp. cf. T. hilli, from the Bow Local Fauna, in New South Wales (Archer and Dawson 1982). Those authors postulated that such sympatry was possible because of the size difference between the two taxa, the former being much larger than the latter. The significantly smaller size of L. roskellyae in comparison to the two species of Wakaleo suggests that the same explanation may also account for their coexistence.

Niche partitioning may also be the reason why Lekaneleo roskellyae was able to persist through a relatively long period of time. This species was possibly the apical small arboreal predator in the rainforest/mesic forests that have been postulated as to have dominated Riversleigh’s palaeoenvironments (Archer et. al.1997). It is possible that while species of Wakaleo were responding to the evolutionary changes in the larger ground-dwelling prey species (i.e. increases in size), L. roskellyae may have been free to predate the smaller but diverse arboreal faunal community present in the late Oligocene to mid Miocene forests.

Dirk’s Towers Site represents a unique locality in that it has yielded three species of thylacoleonid, Wakaleo hilmeri, Lekaneleo roskellyae and L. deminutivus. This level of diversity and possible sympatry of thylacoleonids is previously unseen and suggests even more complex niche partitioning and a past environment that was resource rich. 351

An alternative hypothesis for the faunal diversity at Dirk’s Towers is proposed by Pledge (2005). The areal extent of the Dirk’s Towers Site is relatively large (10-15 metres). Pledge (2005 p.136) identifies two species of wynyardiid from Dirk’s Towers Site and suggests that this site “may represent different time periods in a shifting depositional environment, or a mixed fauna of species from two or more habitats”.

Lekaneleo roskellyae disappears from the fossil record in the early late Miocene. It is absent from the early late Miocene Encore LF, the late Miocene Alcoota LF, as well as all other post-Miocene deposits. Its disappearance coincides with the onset of an icehouse climatic event that is interpreted (McGowran and Li 1997) to have occurred at about this time. Archer et al. (1997) suggest that during this time Australia’s northern forests probably became more open. Although some faunal elements (phalangerids and browsing diprotodontoids) of the Encore LF still suggest a closed forest environment, others suggest a more open environment (a vombatid cf. Warendja sp., and a koala, Phascolarctos sp.; Myers et al. 2001). The fauna from the Alcoota LF has also been interpreted as reflecting a more open forest, subtropical savannah, environment (Murray and Megirian 1992). This major change to the forest environments of this region was probably a significant factor in the disappearance of L. roskellyae at this time. 352

CHAPTER 15

CONCLUSIONS

Study of the marsupial lion fossils recovered from the Riversleigh World Heritage Area, Lawn Hill National Park, northwest Queensland, indicates a greater diversity of this unique marsupial family during the Oligo-Miocene than previously realised. Morphological and phylogenetic studies of this material have resulted in the recognition of two new genera and five new species of thylacoleonid. These species include Microleo attenboroughi, Lekaneleo roskellyae, Lekaneleo deminutivus, Wakaleo hilmeri and Wakaleo sp. cf. W. hilmeri. Furthermore, significant additional material of two previously known species, W. oldfieldi and W. vanderleueri, has also been identified.

Re-evaluation of the fragmentary holotype of Priscileo pitikantensis Rauscher, 1984 from the Ngapakaldi Local Fauna, South Australia, indicates that this specimen most likely represents a species of Wakaleo and hence is referred to this genus, becoming W. pitikantensis (new combination). Consequently, the name Priscileo is regarded as a junior synonym of Wakaleo. The relatively small but distinctive thylacoleonid, Priscileo roskellyae Gillespie, 1997, is referred to a new genus, Lekaneleo.

On the basis of its larger size and because of the recognition of a morphocline within the Wakaleo lineage, the Riversleigh species, W. hilmeri sp. nov., is recognised as a species separate from W. pitikantensis. The relatively small size difference between the dentitions of these species suggests the possibility that they may be conspecific, however, the relatively large size differences between their postcranial elements as well as the possibility that they may be different in age, tend to refute this hypothesis. The diagnostic utility of anterior premolar loss as a synapomorphy uniting members of the genus Wakaleo is lost because W. pitikantensis, W. hilmeri and Wakaleo sp. cf. 353

W. hilmeri retain P1. This feature is also shared with Lekaneleo roskellyae indicating that it is a symplesiomorphic character within marsupial lions.

Wakaleo sp. cf. W. hilmeri from Riversleigh’s D site locality shows plesiomorphic features of W. hilmeri as well as derived features of W. oldfieldi which suggest that it may be an intermediate form.

Previously, W. oldfieldi and W. vanderleueri were distinguished primarily by relative differences in the size of P3. Dentary specimens of these two species collected from Riversleigh sites indicate that discrete differences are present in the morphology of their M3s; that of W. oldfieldi is plesiomorphic and retains a talonid basin-like structure while that of W. vanderleueri is more derived, the basin being lost.

Microleo attenboroughi represents a new plesiomorphic and diminutive taxon of marsupial lion. The P3 morphology of Microleo shows distinct differences to those of all other thylacoleonids and shares with Litokoala a number of features (including the presence of two or three longitudinally-aligned cusps that are linked by a crest and lingual swelling of the crown) which suggest a shared ancestry.

Phylogenetic studies of the intrafamilial relationships of marsupial lions using parsimony analysis indicate that M. attenboroughi is the sister-group to a clade containing the remaining thylacoleonid taxa. Similarly, Lekaneleo roskellyae is the sister-group to a clade containing all thylacoleonids excluding M. attenboroughi. In nearly all analyses, the taxon pitikantensis (SAM P37719) consistently paired with W. hilmeri, supporting referral of this specimen from the genus Priscileo to Wakaleo. Phylogenies that place pitikantensis outside of Wakaleo (Rauscher 1987; Murray et al. 1987) are not supported. Placement of pitikantensis within Wakaleo allows the subfamilial taxon Wakaleoninae (Murray et al. 1987) to remain monophyletic. The uncertain position of Lekaneleo deminutivus can be accredited to the significant amount of missing data for this taxon. 354

Lekaneleo appears to have given rise to Wakaleo probably during the late Oligocene or early Miocene. The origins of Thylacoleo remain unclear; it is possible the lineage arose from Lekaneleo or Wakaleo during the early Miocene, or alternatively from Wakaleo in the middle to late Miocene. Although Thylacoleo presently lacks a pre- Pliocene fossil record it is possible that this lineage arose in a region that is yet to yield fossil-bearing sediments of appropriate age. More complete material of T. hilli may also help to resolve the question of Thylacoleo origins.

Phylogenetic analyses of interfamilial relationships based on cranio-dental characters provide support for the hypothesis that thylacoleonids are members of the Infraorder Vombatiformes (Aplin and Archer 1987, Marshall et al. 1990, Munson 1990). The hypothesis that thylacoleonids are the sistergroup of burramyids (Murray et al. 1987, Rauscher 1987) is not supported. Within Vombatiformes, marsupial lions appear to be the sister-group to all remaining vombatomorphians, except the enigmatic taxon, Marada.

The relatively square outline of the upper molars of M. attenboroughi, L. roskellyae and more particularly W. hilmeri supports the contention of Archer and Rich (1982) that the triangular molars characteristic W. oldfieldi, W. vanderleueri and W. alcootaensis are derived and are not the plesiomorphic thylacoleonid condition.

The morphology of the postcranials of W. hilmeri and W. pitikantensis, in particular the humerus and radius, are very similar to each other and no doubt reflect phylogeny to a large degree. Morphological comparisons with the limb bones of other marsupial species of different locomotor categories suggest scansorial or arboreal abilities for these two species but do not provide strong support for a fossorial hypothesis advanced by Rauscher (1987) for W. pitikantensis.

Based on the stage of evolution of species of Wakaleo, the distribution of Wakaleo species generally supports the proposed ages of Riversleigh’s “Systems”. The most plesiomorphic species, W. hilmeri, is found in sites allocated to Systems A and B which are regarded as older than those of System C which in turn yield the more 355 derived species, W. oldfieldi. The most derived species, W. vanderleueri, is found in post-System C sites. The presence of W. hilmeri in sites allocated to Systems A and B suggests age equivalence of these Systems, however, an age difference for these two Systems as proposed by Archer et al. (1989, 1991, 1997) cannot be discounted if W. hilmeri existed over an extended period of time.

Two species of marsupial lions allow direct correlations with other localities. Wakaleo oldfieldi recovered from C.O.A. Site and a number of Riversleigh’s System C sites, is also a component of the Kutjamarpu Local Fauna in South Australia. Relatively strong species level correlations between the Kutjamarpu LF and System B sites suggest that the age of the Kutjamarpu LF lies somewhere between the ages proposed for Riversleigh’s System B and C sites. The phyletic evolution of Wakaleo suggests that Riversleigh’s System A and B sites that yield W. hilmeri are older than the Kutjamarpu LF but may be similar in age or slightly younger than the late Oligocene Ngapakaldi LF of which W. pitikantensis is a component. Although Wakaleo vanderleueri is present in Riversleigh’s Encore Site and the Bullock Creek Local Fauna in the Northern Territory, the presence of relatively more derived palorchestids and macropodoids in Encore suggests it is slightly younger in age.

In light of the information regarding palaeoclimate during the Tertiary, the overall species diversity of thylacoleonids as well as that of other mammalian families in these sites may provide a better indicator of their age. The pattern of marsupial lion diversity found at Riversleigh mirrors the diversity found in other marsupial families with greatest diversity found in System B sites. This diversity appears to reflect a species-rich rainforest environment that correlates with greenhouse climatic conditions that are believed to have prevailed in the early Miocene. The decline in thylacoleonid diversity throughout the Miocene, Pliocene and Pleistocene coincides with the drying of the Australian continent. 356

REFERENCES

Amrine-Madsen, H., Scally, M., Westerman, M., Stanhope, M.J., Krajewski, C. and Springer, M.S. 2003. Nuclear gene sequences provide evidence for the monophyly of australidelphian marsupials. Molec. Phyl. and Evol.: 28: 186-196.

Anyonge, W. 1996. Locomotor behaviour in Plio-Pleistocene sabre-toothed cats: a biomechanical analysis. J. Zool. Lond. 238:395-413.

Aplin, K.P. 1987. Basicranial anatomy of the early Miocene diprotodontian Wynyardia bassiana (Marsupialia: Wynyardiidae) and its implications for wynyardiid phylogeny and classification. Pp. 369-391 in Archer, M. (ed.), 'Possums and Opossums: Studies in Evolution'. Surrey Beatty and Sons and the Royal Zoological Society of New South Wales: Sydney.

Aplin, K.P. and Archer, M. 1987. Recent advances in marsupial systematics with a new syncretic classification. Pp. xv-lxxii in Archer, M. (ed.), 'Possums and Opossums: Studies in Evolution'. Surrey Beatty and Sons and the Royal Zoological Society of New South Wales: Sydney.

Archer, M. 1975. Abnormal dental development and its significance in dasyurids and other marsupials. Mem. Qd Mus. 17(2): 251-65.

Archer, M. 1976a. The basicranial region of marsupi-carnivores (Marsupialia), interrelationships of carnivorous marsupials, and affinities of the insectivorous peramelids. Zoo. J. Linn. Soc. Lond. 59: 217-322.

Archer, M. 1976b. Phascolarctid origins and the potential of the selenodont molar in the evolution of diprotodont marsupials. Mem. Qd Mus. 17: 367-71.

Archer, M. 1984. The Australian marsupial radiation. Pp. 634-808 in Archer, M. and Clayton, G. (eds.), 'Vertebrate zoogeography and evolution in Australasia'. (Hesperian Press: Perth).

Archer, M. and Clayton, G. 1984. ‘Vertebrate Zoogeography and Evolution in Australasia: Animals in Time and Space’. Hesperian Press. Perth.

Archer, M. and Dawson, L. 1982. Revision of marsupial lions of the genus Thylacoleo Gervais (Thylacoleonidae: Marsupialia) and thylacoleonid evolution in the late Cainozoic. Pp. 477-94 in Archer, M. (ed.), 'Carnivorous Marsupials'. (Royal Zoological Society of New South Wales: Sydney).

Archer, M., Godthelp, H., Hand, S.J. and Megirian, D. 1989. Fossil mammals of Riversleigh, northwestern Queensland: preliminary overview of biostratigraphy, correlation and environmental change. Aust Zool. 25: 29-65.

Archer, M., Hand, S. J., and Godthelp, H. 1991. Riversleigh: The story of animals in ancient rainforests of inland Australia. Reed Books, Sydney. 357

Archer, M., Hand, S.J., Godthelp, H., and Creaser, P. 1997. Correlation of the Cainozoic sediments of the Riversleigh World Heritage fossil property, Queensland, Australia. Pp. 131-152 in Aguilar, J.-P., Legendre, S. and Michaux, J. (eds.), Actes du Congrès BiochroM '97. Mémoirs et Travaux de l'École Pratique des Hautes Études, Institut de Montpellier, 21.

Archer, M. and Rich, T.H. 1982. Results of the Ray E. Lemley Expeditions. Wakaleo alcootaensis n. sp. (Thylacoleonidae: Marsupialia), a new marsupial lion from the Miocene of the Northern Territory, with a consideration of the early radiation of the family. Pp. 495-502 in Archer, M. (ed.) 'Carnivorous Marsupials'. (Royal Zoological Society of New South Wales: Sydney).

Archer, M. and Wade, M. 1976. Results of the Ray E. Lemley Expeditions, Part 1. The Allingham Formation and a new Pliocene vertebrate fauna from northern Queensland. Mem. Qd. Mus. 17: 379-97.

Arena, D. 2005. The geological history and development of the terrain at the Riversleigh World Heritage Area during the middle tertiary. PhD Thesis. University of New South Wales.

Argot, C. 2001. Functional-adaptive anatomy of the forelimb in the Didelphidae, and the paleobiology of the Paleocene marsupials Mayulestes ferox and Pucadelphys andinus. J. Morph. 247:51-79.

Argot, C. 2003. Postcranial functional adaptations in the South American Miocene borhyaenoids (Mammalia, Metatheria): Cladosictis, Pseudonotictis and Sipalocyon. Alcheringa. 27: 303-356.

Asher, R.J., Horovitz, I. and Sanchez-Villagra, M.R. 2004. First combined cladistic analysis of marsupial mammal interrelationships. Mol. Phyl. Evol. 33: 240-250.

Barbour, R.A. 1963. The musculature and limb plexuses of Trichosurus vulpecula. Aust. J. Zool. 11: 488-610.

Bartholomai, A. 1962. A new species of Thylacoleo and notes on some caudal vertebrae of Palorchestes azael. Mem. Qd. Mus. 14:33-40.

Bartholomai, A. 1975. The genus Macropus Shaw (Marsupialia: Macropodidae) in the Upper Cainozoic deposits of Queensland. Mem. Qd. Mus. 17 (2): 195-235.

Bensley, B.A. 1903. On the evolution of the Australian Marsupialia; with remarks on relationships of the marsupials in general. Trans. Linn. Soc. Lond. (Zool.) Ser. 29: 83-217.

Black, K. 1997. Diversity and biostratigraphy of the Diprotodontoidea of Riversleigh, northwestern Queensland. Mem. Qd. Mus. 41: 187-192. 358

Black, K. 2006. Description of new material for Propalorchestes novaculacephalus (Marsupialia: Palorchestidae) from the mid Miocene of Riversleigh, northwestern Queensland. Alcheringa. 30: 351-361.

Black, K. in press. Maradidae: a new family of vombatomorphian marsupial from the early Miocene of Riversleigh, northwestern Queensland.

Boles, W. E. 1997. Riversleigh birds as palaeoenvironmental indicators. Mem. Qd. Mus. 41: 241-246.

Boles, W.E. 2000. Investigations on Australian Tertiary avifauna, with an emphasis on the fossil birds of Riversleigh, northwestern Queensland. PhD. University of New South Wales.

Broom, R. 1898. On the affinities and habits of Thylacoleo. Proc. Linn. Soc. N.S.W. 22: 57-74.

Bryant, H.N. and Seymour, K.L. 1990. Observations and comments on the reliability of muscle reconstruction in fossil vertebrates. J. Morph. 206: 109-117.

Bufkin, W. J. 1971. The avulsive cortical irregularity. Amer. J. Roent., Rad. Ther. and Nuc. Med. 112: 487 – 492.

Carlsson, A. 1914. Uber Dendrolagus dorianus. Zoologische Jahrbruchen. 36(6): 547 – 620.

Carrano, M.T. 1999. What if anything, is a cursor? Categories versus continua for determining locomotor habit in mammals and dinosaurs. J. Zool. Lond. 247: 29 - 42.

Clemens, W.A. and Plane, M. 1974. Mid-Tertiary Thylacoleonidae (Marsupialia, Mammalia). J. Paleont. 48: 652-60.

Cooke, B.N. 1997. Biostratigraphic implications of the fossil kangaroos of the Riversleigh World Heritage fossil deposits, northwestern Queensland. Mem. Qd. Mus.

Creaser, P. 1997. Oligo-Miocene sediments of Riversleigh: the potential significance of topography. Mem. Qld. Mus. 41(2): 303-314.

Crosby, K. 2002. Studies in the diversity and evolution of phalangeroid possums (Marsupialia; Phalangerida; Phalangeroidea). Ph.D thesis. University of New South Wales.

Crosby, K. and Norris, C.A. 2003. Periotic morphology in the Trichosurin possums Strigocuscus celebensis and Wyulda squamicaudata (Diprotodontia, Phalangeridae) and a revised diagnosis of the tribe Trichosurini. American Museum Novitates. No. 3414. 359

Davis, D.D. 1964. The giant panda. Fieldiana, Zool. Mem. 3: 1-339.

Duckler, G.L. 1997. Parietal depressions in skulls of the extinct saber-toothed felid Smilodon fatalis: evidence of mechanical strain. J. Vert. Paleont. 17 (3): 600-609.

Edgar, R. 1983. Spotted tailed Quoll. Pp.18 in R. Strahan (ed.), ‘the Complete Book of Australian Mammals’. Angus and Robertson. London.

Elftman, H.O. 1929. Functional adaptations of the pelvis in marsupials. Bull. Am. Mus. Nat. Hist. 58: 189-232.

Flannery, T.F. 1994. Possums of the world. Geo Productions. Sydney.

Flower, W.H. 1867. On the development and succession of teeth in the Marsupialia. Phil. Trans. Roy. Soc. Lon. 157: 631-641.

Finch, E. 1971. Thylacoleo, marsupial lion or marsupial sloth? Aust. Nat. Hist. 17 (1): 7- 11.

Finch, M.E. 1981. Thylacoleo carnifex Owen. An Australian fossil phalangerid marsupial. Ph.D. Thesis. University of Western Australia.

Finch, M.E. 1982. The discovery and interpretation of Thylacoleo carnifex (Thylacoleonidae, Marsupialia). Pp. 537-572 in M. Archer (ed.), ‘Carnivorous Marsupials’, Vol. 2. Royal Zoological Society of New South Wales. Surrey Beatty and Sons.

Finch, M.E. and Freedman, L. 1982. An odontometric study of the species of Thylacoleo (Thylacoleonidae, Marsupialia). Pp. 553-561 in M. Archer (ed.), ‘Carnivorous Marsupials’, Vol. 2. Royal Zoological Society of New South Wales. Surrey Beatty and Sons.

Finch, M.E. and Freedman, L. 1986. Functional morphology of the vertebral column of Thylacoleo carnifex Owen (Thylacoleonidae: Marsupialia). Aust. J. Zool. 34: 1-16.

Finch, M.E. and Freedman, L. 1988. Functional morphology of the limbs of Thylacoleo carnifex Owen (Thylacoleonidae: Marsupialia). Aust. J. Zool. 36: 251-72.

Frakes, L.A., McGowran, B. and Bowler, J. 1987. Evolution of Australian environments. Pp. 1-16 in Dyne, G.R. and Walton, D.W. (eds), ‘Fauna of Australia’, Vol.1A, General articles. Australian Government Publishing Service, Canberra.

Frith H.J. and Calaby, J.H. 1969. Kangaroos. F.W. Cheshire Publishing. Melbourne.

Fulton, M.N., Albright, J.P., and El-Khoury, G.Y. 1979. Cortical desmoid-like lesion of the proximal humerus and its occurrence in gymnasts (ringman’s shoulder lesion). Am. J. Sports Med. 7: 57-61. 360

Gambaryan, P.P. 1974. How mammals run: anatomical adaptations. John Wiley and Sons. New York.

Gillespie, A. 1997. Priscileo roskellyae sp. nov. (Thylacoleonidae, Marsupialia) from the Oligocene-Miocene of Riversleigh, northwestern Queensland. Mem. Qd. Mus. 41(2): 321-327.

Gordon, G.1983. Northern Brown Bandicoot. Pp. 96 in R. Strahan (ed.), ‘the Complete Book of Australian Mammals’. Angus and Robertson. London.

Grand, T.I. and Barboza, P.S. 2001. Anatomy and development of the koala, Phascolarctos cinereus: an evolutionary perspective on the superfamily Vombatoidea. Anat. Embryol. 203: 211 - 223.

Guiler, E.R. 1985. ‘Thylacine: the tragedy of the Tasmanian Tiger’. Oxford University Press. Oxford.

Hand, S., Archer, M. and Godthelp, H. 1997. First record of Hydromops (Microchiroptera: Molossidae) from Australia: its biocorrelative significance. Pp 153-162 in Aguilar, J.-P., Legendre, S. and Michaux, J. (eds.), Actes du Congrès BiochroM '97. Mémoirs et Travaux de l'École Pratique des Hautes Études, Institut de Montpellier, 21.

Haines, R.W. Arboreal or terrestrial ancestry of placental mammals. Quart. Rev. Biol. 33 (1): 1-23.

Hennig, W. 1966. Phylogenetic Systematics. University of Illinois Press. Urbana.

Hildebrand, M. 1982. Analysis of vertebrate structure. 2nd ed. John Wiley and Sons. New York.

Hildebrand, M. 1985. Digging of quadrupeds. Pp. 89-109 in Hildebrand M., Bramble, D.M., Liem, K.F. and Wake, D.B. (eds), ‘Functional Vertebrate Morphology’. Belknap Press of Harvard University Press. Cambridge.

Hocknull, S. and Zhao, J. 2005. The Mount Etna menagerie – updates. 10th Conference on Australian Vertebrate Evolution, Palaeontology and Systematics and Quaternary Extinctions Symposium. Naracoorte.

Horovitz, I. and Sanchez-Villagra, M. R. 2003. A morphological analysis of marsupial mammal higher-level phylogenetic relationships. Cladistics. 19: 181 - 212.

Horton, D.R. (1984). Red Kangaroos: Last of the Australian Megafauna. Pp. 639 - 680 in Martin, P.S. and Klein, R.G., (eds.), ‘Quaternary Extinctions: A Prehistoric Revolution’. University of Arizona Press. Tuscon.

How, R.A. and Kerle, J.A. 1995. Common Brushtail Possum. Pp. 273-275 in Strahan, R. (ed.), ‘The Mammals of Australia’. Reed Books. Chatswood. 361

Jack, 1896. Stratigraphical notes on the Georgina Basin, with reference to the question of artesian water. Proc. Roy. Soc. Qld. 11: 70-74.

Jenkins, F. 1973. The functional anatomy and evolution of the mammalian humero- ulnar articulation. Am. J. Anat. 137: 281-298.

Johnson, K.A. 1995. Marsupial Mole. Pp. 409-411 in Strahan, R. (ed.), ‘The Mammals of Australia’. Reed Books. Chatswood.

Kavangh, J.R., Burk-Herrick, A., Westerman, M. and Springer, M.S. 2004. Relationships among families of Diprotodontia (Marsupialia) and the phylogenetic position of the autapomorphic honey possum (Tarsipes rostratus). J. Mamm. Evol.: 11 (3/4): 207–222.

Kear, B.J. 2002. Phylogenetic implications of macropodid (Marsupialia: Macropodoidea) postcranial remains from Miocene deposits of Riversleigh, northwestern Queensland. Alcheringa. 26: 299-318.

Kerle, A. 2001. ‘Possums: the brushtails, ringtails and greater glider’. University of New South Wales Press. Sydney.

Kingsmill, E. 1962. An investigation of criteria for estimating age in the marsupials Trichosurus vulpecula Kerr and Perameles nasuta Geoffroy. Aust. J. Zool. 12: 322- 39.

Kirkham, Z. 2004. The cranial description of the primitive macropodid Rhizosthenurus flanneryi and its phylogeny based on cranial and postcranial characters. Unpublished Ph.D Thesis, University of New South Wales.

Kirsch, J.A.W., Lapointe, F.-J. and Springer, M.S. 1997. DNA-hybridization studies of marsupials and their implications for metatherian classification. Aust. J. Zool. 45: 211-280.

Kitching, I.J., Forey, P.L., Humphries, C.J., and Williams, D.M. 1998. Cladistics. Oxford University Press.

Krefft, G. 1872. A Cuverian principle in palaeontology, tested by evidences of an extinct leonine marsupial (Thylacoleo carnifex) by Professor Owen etc. Reviewed by Krefft. Ann. and Mag. Nat. Hist. 4:169-182.

Kurten, B. 1963. Return of a lost structure in the evolution of the felid dentition. Soc. Sci. Fennica, Commentations Biol. 26:1-10.

Landry, S.O. 1958. The function of the entepicondylar foramen in mammals. Amer. Midland Nat. 60 (1): 100-112. 362

Lee, A.K. and Carrick, F.N. 1989. Phascolarctidae. Pp.740 - 767 in Walton, D.W. (ed.), ‘Fauna of Australia’ Vol. 1B Mammalia. Australian Government Publishing Service. Canberra.

Louys, J. 2004. Cranial descriptions of archaic koalas and an examination of vombatiform phylogeny based on cranial characters. Honours thesis. University of New South Wales.

Louys, J., Black, K., Archer, M., Hand, S.J. and Godthelp, H. (in Press) Descriptions of koala material from the Miocene of Riversleigh, northwestern Queensland and implications for Litokoala (Marsupialia, Phascolarctidae).

Lyne, A.G. 1982. Observations on skull growth and eruption of teeth in the marsupial bandicoot Perameles nasuta (Marsupialia: Peramelidae). Aust. Mammal. 5: 113-26.

Macleod, N. and Rose, K.D. 1993. Inferring locomotor behaviour in Paleogene mammals via eigenshape analysis. Am. J. Science. 293 A: 300-355.

MacLennan, D.G. 1984. The feeding behaviour and activity patterns of the Brushtail Possum, Trichosurus vulpecula, in an open eucalypt woodland in southeast Queensland. Pp. 155-161 in Smith, A.P. and Hume, I.D. (eds.), ‘Possums and Gliders’. Australian Mammal Society. Sydney.

McGowran, B., Archer, M., Bock, P., Darragh, T.A., Godthelp, H., Hageman, S., Hand, S., Hill, R., Li, Q., Maxwell, P.A., McNamara, K.J., MacPhail, M., Mildenhall, D., Partridge, A.D., Richardson, J., Shafik, S., Truswell, E.M. and Warne, M. 2000. Australian palaeobiogeography: the Palaeogene and Neogene record. Mem. Assoc. Aust. Palaeo. 23: 405-470.

McGowran, B. and Li, Q. 1994. The Miocene oscillation in southern Australia. Rec. S. Aust. Mus. 27: 197-212.

Marshall, L. G. 1982. Systematics of the extinct South American marsupial family Polydolopidae. Fieldiana Geol. Ns.:12: 1-109.

Marshall, L.G., Case, J.A., and Woodburne, M.O. 1990. Phylogenetic relationships of the families of the marsupials. Pp. 433-505 in Genoways, H.H. (ed.), ‘Current Mammalogy’, Vol. 2, Plenum Press, New York.

Martin, R.W. and Handasyde, K.A. 1995. Koala. Pp.196-198 in Strahan, R. (ed.), ‘The Mammals of Australia’. Reed Books. Chatswood.

Megirian, D. 1986. The dentary of Wakaleo vanderleueri (Thylacoleonidae: Marsupialia). The Beagle. 3: 71-79.

Megirian, D. 1992. Interpretation of the Miocene Carl Creek Limestone, Northwestern Queensland. The Beagle. 9 (1): 219-248. 363

Megirian, D. 1994. Approaches to marsupial biochronology in Australia and New Guinea. Alcheringa. 18: 259-274.

Miller, M.E., Christensen, G.C., and Evans, H.E. 1964. Anatomy of the Dog. W.B. Saunders Company. Philadelphia.

Muirhead, J. and Filan, S.L. 1995. Yarala burchfieldi, a plesiomorphic bandicoot (Marsupialia, Peramelimorphia) from Oligo-Miocene deposits of Riversleigh, northwestern Queensland. J. Paleont. 69 (1): 127-134.

Muizon, C. de 1998. Mayulestes ferox, a borhyaenoid (Metatheria, Mammalia) from the early Palaeocene of Bolivia. Phylogenetic and palaeobiologic implications. Geodiversitas 20 (1): 19-142.

Munson, C.J. 1992. Postcranial descriptions of Ilaria and Ngapakaldia (Vombatiformes, Marsupialia) and the phylogeny of the Vombatiforms based on postcranial morphology. University of California Publications in Zoology. Vol. 125. University of California Press.

Murray, P. 1986. Propalorchestes novaculacephalus gen. et sp. nov., a new palorchestid (Diprotodontoidea: Marsupialia) from the middle Miocene Camfield Beds, Northern Territory, Australia. The Beagle 3(1): 195-211.

Murray, P. 1990. Primitive marsupial tapirs (Propalorchestes novaculacephalus Murray and P. ponticulus sp.nov.) from the mid-Miocene of North Australia (Marsupialia: Palorchestidae). The Beagle 7(2): 39-51.

Murray, P. 1998. Palaeontology and palaeobiology of wombats. Pp. 1-33 in R.T. Wells and P.A. Pridmore (eds.), ‘Wombats’. Surrey Beatty and Sons.

Murray, P. and Megirian, D. 1990. Further observations on the morphology of Wakaleo vanderleueri (Marsupialia: Thylacoleonidae) from the mid-Miocene Camfield Beds, Northern Territory. The Beagle 7: 91-102.

Murray, P. and Megirian, D. 1992. Continuity and contrast in Middle and Late Miocene vertebrate communities from the Northern Territory. The Beagle 9 (1): 195-218.

Murray, P., Wells, R. and Plane, M. 1987. The cranium of the Miocene thylacoleonid, Wakaleo vanderleueri: click go the shears-a fresh bite at thylacoleonid systematics. Pp.433-66 in Archer, M. (ed.), 'Possums and Opossums: Studies in Evolution'. Surrey Beatty and Sons and the Royal Zoological Society of New South Wales: Sydney.

Myers, T. 1994. Kuterintja ngama (Marsupialia, Ilariidae): a revised and extended systematic and phylogenetic analysis, based on fossil material from Oligo-Miocene deposits at Riversleigh, northwestern Queensland, Australia. Honours Thesis. University of New South Wales. 364

Myers, T. and Archer, M. 1997. Kuterintja ngama (Marsupialia, Ilariidae): A revised and extended systematic analysis based on fossil material from the late Oligocene of Riversleigh, northwestern Queensland, Australia. Mem. Qd. Mus.

Myers, T., Crosby, K., Archer, M. and Tyler, M. 2001. The Encore Local Fauna, a late Miocene assemblage from Riversleigh, northwestern Queensland. Mem. Assoc. Aust. Palaeont. 25: 147-154.

Napier, J.R. 1961. Prehensility and opposability in the hands of primates. Symp. Zool. Soc. Lond. 5: 115-132.

Nomina Anatomica Veterinaria. 1994. Fourth Edition. International Committees on Veterinary Gross Anatomical Nomenclature. World Association of Veterinary Anatomists. Zurich.

Norris, C.A. 1994. The periotic bones of possums and cuscuses: cuscus polyphyly and the division of the marsupial family Phalangeridae. Zool. J. Linn. Soc. 111: 73-98.

Nowak, R.M. 1999. ‘Walker’s Mammals of the World’. The Johns Hopkins University Press. Baltimore.

Owen, D. and Pemberton, D. 2005. ‘Tasmanian Devil: a unique and threatened animal’. Allen and Unwin. Crows Nest.

Owen, R. 1858. Odontology. In ‘the Encyclopaedia Britannica, or dictionary of arts, sciences, and general literature’ (8th ed.). 16: 407-484.

Owen, R. 1859. On the fossil mammals of Australia. 1. Description of a mutilated skull of a large marsupial carnivore (Thylacoleo carnifex Owen) from a calcareous conglomerate stratum, eighty miles s.w. of Melbourne, Victoria. Phil. Trans. Roy. Soc. Lond. 149: 309-322.

Owen, R. 1866. On the fossil mammals of Australia. 2. Description of an almost entire skull of the Thylacoleo carnifex, Owen, from a freshwater deposit, Darling Downs, Queensland. Phil. Trans. Roy. Soc. Lond. 156: 73-82.

Owen, R. 1883. On the affinities of Thylacoleo. Phil. Trans. Roy. Soc. Lond. 174: 575- 582.

Owen, R. 1887. Additional evidence of the affinities of the extinct marsupial quadruped Thylacoleo carnifex (Owen). Phil. Trans. Roy. Soc. Lond. 178, B: 1-4.

Oxnard, C.E. 1963. Locomotor adaptations in the primate forelimb. Symp. Zool. Soc. Lond. 10: 165-182.

Plane, M. and Gatehouse, C.G. 1968. A new vertebrate fauna from the Tertiary of Northern Australia. Aust. J. Sci. 30 (7): 272-273. 365

Pledge, N. 1977. A new species of Thylacoleo (Marsupialia, Thylacoleonidae) with notes on the occurrence and the distribution of Thylacoleonidae in South Australia. Rec. S. Aust. Mus. 17 (16): 277-83.

Pledge, N. 2005. The Riversleigh wynyardiids. Mem. Qd. Mus. 51 (1): 135-169.

Rauscher, B. 1983. unpublished. "A new Tertiary thylacoleonid (Mammalia: Marsupialia): phylogeny and functional morphology". M.A. Thesis. University of California, Berkeley.

Rauscher, B. 1987. Priscileo pitikantensis, a new genus and species of thylacoleonid marsupial (Marsupialia: Thylacoleonidae) from the Miocene Etadunna Formation, South Australia. Pp. 423-32 in Archer, M. (ed.), 'Possums and Opossums: Studies in Evolution'. Surrey Beatty and Sons and the Royal Zoological Society of New South Wales: Sydney.

Retief, J.D., Krajewski, C., Westerman, M., Winkfein, R.J. and Dixson, G.H. 1995. Molecular phylogeny and evolution of marsupial protamine P1 genes. Proc. R. Soc.: Biol. Sci. 259 (1354): 7-14.

Ride, W.D.L. 1964. A review of Australian fossil marsupials. J. Proc. R. Soc. West. Aust. 47: 97-131.

Sanchez-Villagra, M.R. and Wible, J.R. 2002. Patterns of evolutionary transformation in the petrosal bone and some basicranial features in marsupial mammals, with special reference to didelphids. J. Zool. Syst. Evol. Res. 40: 26-45.

Scotland, R.W. 1992. Cladistic theory. Chapter 1, in Foley, P.L., Humphries, C.J., Kitching, I.J., Scotland, R.W., Siebert, D.J. and Williams, D.M., (eds.), ‘Cladistics: a practical course in systematics’. Clarendon Press. Oxford.

Scott, G.G. and Richardson, K.C. 1988. Appendicular osteological differences between Lasiorhinus latifrons (Owen, 1845) and Vombatus ursinus (Shaw, 1800) (Marsupialia: Vombatidae). Rec. S. Aust. Mus. 22(2): 95-102.

Shrivastava, R.K. 1962. The deltoid musculature of the Marsupialia. Amer. Mid. Nat. 67: 305-320.

Slijper, E.J. 1946. Comparative biologic-anatomical investigations on the vertebral column and spinal musculature of mammals. Verhand. Kon. Ned. Akad. Wet.Afd. Natuurkd. Tweede Sectie, Dl. 42 (5): 1-128.

Smith, A.B. 1994. Systematics and the fossil record: documenting evolutionary patterns. Blackwell Science. Oxford.

Sonntag, C.F. 1922. On the myology and classification of the wombat, koala, and phalangers. Proc. Zool. Soc. Lond. 1922: 863-896. 366

Springer, M. S., Kirsch, J.A.W. and Case, J.A. 1997.The chronicle of marsupial evolution. Pp.129-161 in Givnish, T.J. and Systma, K.J. (eds.), ‘Molecular Evolution and Adaptive Radiation’. Cambridge University Press, New York.

Springer, M.S., Westerman, M., Kavanagh, J.R., Burk, A., Woodburne, M.O., Kao, D.J. and Krajewski, C. 1998. The origin of the Australasian marsupial fauna and the phylogenetic affinities of the enigmatic monito del monte and marsupial mole. Proc. R. Soc. Lond. B 265: 2381-2386.

Springer M.A. and Woodburne, M.O. 1989. The distribution of some basicranial characters within the Marsupialia and a phylogeny of the Phalangeriformes. J. Vert. Paleo. 9 (2): 210-221.

Stein, B.R. and Casinos, A. 1997. What is a cursorial mammal? J. Zool. Lond. 242: 185- 192.

Stirton, R.A., Tedford, R.H. and Woodburne, M.O. 1968. Australian Tertiary deposits containing terrestrial mammals. University of California Publications in Geological Sciences. Vol. 77.

Stodart, E. 1983. Long-nosed Bandicoot. Pp.99 in Strahan R., ‘the Complete Book of Australian Mammals’. Angus and Robertson. London.

Strahan, R. 1995. The Mammals of Australia. Reed Books. Chatswood.

Swofford, 1998. PAUP: Phylogenetic Analysis Using Parsimony (and Other Methods). Version 4.0 Beta. Sinauer Associates, Sunderland. Massachusetts.

Szalay, F.S. and Sargis, E.J. 2001. Model-based analysis of postcranial osteology of marsupials from the Palaeocene of Itaborai (Brazil) and the phylogenetics and biogeography of Metatheria. Geodiversitas 23 (2): 139-302.

Taylor, M.E. 1974. The functional anatomy of the forelimb of some African Viverridae (Carnivora). J. Morphol. 143: 307-336.

Taylor, M.E. 1989. Locomotor adaptations by carnivores. Pp. 382-409 in Gittleman, J. (ed.), ‘Carnivore behaviour, ecology, and evolution’. Chapman and Hall London.

Tedford, R.H. 1967. Fossil mammal remains from the Tertiary Carl Creek Limestone, north-western Queensland. Bureau of Mineral Resources, Bulletin. 92: 217-236. van Valkenburgh, B. 1985. Locomotor diversity within past and present guilds of large predatory mammals. Paleobiology. 11(4): 406-428. van Valkenburgh, B. 1987. Skeletal indicators of locomotor behaviour in living and extinct carnivores. J. Vert. Paleo. 7(2): 162-182. 367

Warburton, N.M. 2003. Functional morphology and evolution of marsupial moles (Marsupialia; Notoryctemorphia). PhD Thesis. University of Western Australia.

Watrous, L.E. and Wheeler, Q.D. 1981. The outgroup comparison method of character analysis. Systematic Zool. 30: 1–11.

Wells, R.T. 1973. Physiological and behavioural adaptations of the hairy-nosed wombat (Lasiorhinus latifrons) to its arid environment. Ph.D Thesis. University of Adelaide.

Wells, R.T., Horton, D. R. and Rogers, P. 1982. Thylacoleo carnifex Owen (Thylacoleonidae): marsupial carnivore? Pp. 573-86 in Archer, M. (ed.), “Carnivorous Marsupials”. Royal Zoological Society of New South Wales. Sydney.

Wells, R.T. and Nichol, B. 1977. On the manus and pes of Thylacoleo carnifex Owen (Marsupialia). Trans. R.Soc.S. Aust. 101(6): 139-146.

Wiley, E.O., Siegel-Causey, D., Brooks, D.R. and Funk, V.A. 1991. The Compleat Cladist: a primer of phylogenetic procedures. The University of Kansas Museum of Natural History. Special Publication No. 19. Lawrence, Kansas.

Wilson, J.T. 1894. On the myology of Notoryctes typhlops, with comparative notes. Trans. Roy. Soc. S. Aust. 18: 3-74.

Winge, H. 1941. The interrelationships of the mammalian genera. Vol. 1. Pp xii and 418. C.A. Reitzels Forlag. Kobenhaven.

Winter, J.W. and Leung, K-P. 1995. Common Spotted Cuscus. Pp. 266-268 in Strahan, R. (ed.), ‘The Mammals of Australia’. Reed Books. Chatswood.

Wooburne, M.O. 1984. Families of marsupials: relationships, evolution and biogeography. Pp. 48-71 in Broadhead, T.W. (ed.), ‘Mammals: notes for a short course’. Univ. Tenn. Dept. Geol. Sci. Studies in Geol. 8.

Woodburne, M.O., Macfadden, B. J., Case, J.A., Springer, M.S., Pledge, N.S., Power, J.D., Woodburne, J.M., and Springer, K.B. 1993. Land mammal biostratigraphy and magnetostratigraphy of the Etadunna Formation (Late Oligocene) of South Australia. J. Vert. Paleo. 13 (4): 483-515.

Woodburne, M.O., Tedford, R.H., Archer, M., Turnbull, W.D., Plane, M.D. and Lundelius, E.L. 1985. Biochronology of the continental mammal record of Australia and New Guinea. Spec. Publ., S. Aust. Dept Mines and Energy. 5: 347-363.

Wroe, S. 1999. The geologically oldest dasyurid, from the Miocene of Riversleigh, north-west Queensland. Palaeont. 42(3): 501-527.

Yalden, D.W. 1970. The functional morphology of the carpal bones in carnivores. Acta anat. 77: 481-500. 368

Yalden, D.W. 1972. The form and function of the carpal bones in some arboreally adapted mammals. Acta anat. 82: 383-406.

Young, A.H. 1882. The muscular anatomy of the koala (Phascolarctos cinereus), with additional notes. J. Anat., Lond. 16: 217-42. 369

APPENDIX 1 : Dentition measurements of all thylacoleonid specimens from Riversleigh and of comparative specimens from other localities Table A: Measurements of upper cheekteeth for all species of Wakaleo (mm) ** = average of left and right tooth rows except M3. * = measurements from Murray and Megirian (1990). a = alveolar measurement

Species Riversleigh Site/ P3 P3 M1 M1 M2 M2 M3 M3 M4 M4 P3 -M4 P3:M1 P3 / Specimen # Local Fauna length width length width length width length width length width P3 – M4 Wakaleo pitikantensis SAM P37719 Ngapakaldi 11.0a 6.5a 5.3 3.7a 3.5a [1.69] 0.4 Wakaleo hilmeri QM F23443** Rack. Low Lion 12.2 7.1 9.4 9.2 6.0 6.3 4.5 4.9 3.5a 35.8a 1.30 0.34 QM F45200** Hiatus 11.8 7.0 8.8 8.6 6.3 6.9 4.7 5.1 34.0 1.34 0.34 QM F24680 Dirk’s Towers 12.9a 7.8a 8.9a 9.3a 6.3 7.6 4.9a 4.3a 4.4a 4.0a 35.8a 1.44a 0.36 QM F30250 Dirk’s Towers 9.7 9.9 QM F23441 Upper 9.0 9.0 QM F20573 White Hunter 5.7 7.3 QM F30378 White Hunter 6.2 7.5 QM F23444 White Hunter 4.7 4.4 QM F23801 Neville’s Garden 6.5 7.8 AR 16653 5.6 6.3 QM F24660 Creaser’s Ramp 4.7 4.7 W sp. cf . hilmeri QM F23446 D Site, River. LF 12.6 8.5 9.6 9.6 6.0 7.4 5.2 5.5 31.4 1.31 0.4 Wakaleo oldfieldi QM F23427 Helicopter 13.3 6.4 10.1 10.0 1.31 QM F11852 Jim’s Jaw 14.1 7.5 10.6 10.3 1.33 QM F31398 COA 4 7.3 8.4 QM F31294 COA 2A 5.1 5.6 QM F36466 Golden Steph 5.5 5.2 QM F 24745 KCB 7.5 8.1 AR19248 Kutjamarpu 11.6 10.6 AR 3272 Kutjamarpu 6.5 7.1 Wakaleo vanderleueri QM F30384 Encore 8.5 QM F36466 Encore 5.5 5.2 CPC 26604 * Bullock Creek 17.9 9.2 11.2 11.0 7.0 9.5 5.0 5.5 38.6 1.59 NTM P87103-9* Bullock Creek 7.5 8.4 5.6 5.8 Wakaleo alcootaensis NTM P1 Alcoota 23.3 14.7 15.7 7.2 9.0 47.3 1.58

370

APPENDIX I

Table A cont.: Measurements of lower cheekteeth for all species of Wakaleo (mm) ** = average of left and right tooth rows except M3, except W. hilmeri. * = measurements from Murray and Megirian (1990). a = alveolar measurement

Species Riversleigh site/ P3 P3 P3 M1 M1 M1 M2 M2 M2 M3 M3 P3:M1 Local Fauna length anterior posterior length anterior posterior length anterior posterior length width Specimen # width width width width width width W. hilmeri AR 19249** White Hunter 10.1 4.9 5.9 8.7 7.1 6.5 7.4 6.3 6.1 5.8 4.9 1.16 QM F30465 White Hunter 9.5 5.4 6.0 10.0 6.6 7.1 7.7 6.0 6.1 QM F31376 White Hunter 8.6 4.4 8.7 6.1 6.1 7.1 5.7 5.4 NTM P91167-3 Burnt Offering 12.2 9.3 8.5 7.3 7.3 6.2 5.7 NTM P91171-4 Burnt Offering 11.5 6.5 7.4 QM F30073 Dirk’s Towers 9.4 6.6 6.4 QM F23449 Upper 9.6 6.2 5.6 W. oldfieldi QM F20895 COA 1 12.4 5.6 6.7 10.7 8.6 7.7 8.3 6.6 5.3 5.1 4.7 1.15 QM F20304 COA 2A 10.1 8.8 6.9

QM F31295 COA4 11.6 8.2 7.4

QM F36437 COA2A 4.9 4.5

SAM P17925^ Kutjamarpu 12.4 6.2 7.8 10.4 8.7 7.2 7.5a 6.3a 6.1a 5.0a 5.5a 1.19

UCMP 102677^ Kutjamarpu 7.5 6.3

W. vanderleueri

QM F36453 Encore 14.3 6.8 7.7 12.1 8.6 7.7 8.0 6.1 5.4 5.6 3.8 1.18 QM F41204 Encore 15.5 6.4 7.7 1.30 QM F36454 Encore 8.4 5.1 QM F29712 Encore 12.5 8.5 7.6 QM F36281 Encore 7.5 - QM F51409 Encore 9.2 6.4 6.1 QM F52182 Encore 6.3 4.2 NTM P87108-5* Bullock Creek [14.5] 5.7 7.2 12.0 8.2 7.4 9.8 6.3 6.1 [1.20] NTM P87108-6* Bullock Creek [14.2] 6.3 7.6 11.3 [8.5] 7.7 9.4 6.3 6.8 [1.25] NTM P85553-4* Bullock Creek 13.9 - [7.1] 9.6 - [6.5] 7.7 5.5 5.2 [5.9] 1.44 NTM P8695-97* Bullock Creek 8.2 6.6 6.2 6.0 4.7 NTM P8695-96* Bullock Creek 5.8 4.1 CPC 13527* Bullock Creek 11.7 7.9 7.7

371

APPENDIX 1 cont.

Table B: Upper cheek dentition measurements for species of Lekaneleo (mm).

Species Riversleigh P3 P3 P3 M1 M1 M2 M2 M3 M3 M4 M4 P3:M1 P3 / Site/Local length anterior posterior length width length width length width length width P3 – M4 Specimen # Fauna width width (max.) (max.) Lekaneleo roskellyae QM F23453 (L) Upper 8.3 4.2 5.1 5.8 6.2 4.3 4.8 3.3a 3.5a 3.0a 2.5a 1.41 0.35 (R) 8.1 4.2 5.1 5.7 6.0 4.2 4.9 3.3a 3.5a 2.8a 2.5a

QM F23442 WH 7.6 3.4 4.3 5.8 6.1 4.1 4.7 - - - - 1.31 -

QM F40116 JC [9.6 ] - [5.1] 6.5 6.8 4.5 5.4 - - - - 1.47 -

Lekaneleo deminutivus

QM F24549 DT 5.7 2.5 2.7

Table C: Lower cheek dentition measurements for specimens of Lekaneleo roskellyae (mm)

Riversleigh P3 P3 P3 M1 M1 M2 M2 M3 M3 M4 M4 P3-M4 P3 / Specimen # Site length anterior posterior length width length width length width length width P3 – M4 width width QM F20191 COA 6.4 3.6 4.4 6.8 4.8 5.2 4.4 3.7a 2.6a 2.7a 2.2a [25] 0.25 QM F29623 DT 7.3 3.5 4.3 6.3 4.5 ------QM F51418 COA 4.9 3.9

Table D: Dentition measurements for specimens of Microleo attenboroughi (mm)

Riversleigh P3 P3 M1 M1 M2 M2 M3 M3 P3 – M4 M M 2 2 Specimen # Site length maximum length width length width length width length width width QM F41143 (L) NG 4.7a 2.0a 3.5a 2.9a 3.3 3.6 2.8 3.0 [16] (R) 4.6 2.7 3.6a 2.8a 3.3 3.6 - - - QM F42676 3.0 2.5

a = alveolus; [ ] = estimated measure; - = not available 372

APPENDIX 1 cont.

Table E: Comparative measurements of cranial dimensions of species of thylacoleonids (mm). See figure below for illustration of measurements.

Species Skull Skull Occiput Occiput Supraorbital Zygomatic Specimen # Length Depth Width Depth Width Width W. hilmeri QM F45200 164 63a 73.2 27.7 47.1 111.8 W. vanderleueri CPC 26604 186.8 79.3 73.4 34.5 66.4 135a L. roskellyae QM F23453 91.6 29.9 36.0 12.0 25.2 60a T. carnifex AM F4659 236.0 94.6 119.0 50.7 na 188.0 a = approximate; skull length = tip of nasals to nuchal crest; skull depth = occlusal surface (P3/M1) to maximum crown height; occiput width = breadth of braincase between lateral edges of squamosal; occiput depth = nuchal crest to dorsal margin of foramen magnum; na = not available due to presence of postorbital bar

supraorbital width

skull length zygomatic width

occiput width

373

APPENDIX 1 cont.

Table F: Comparative measurements of upper tooth rows and palate dimensions of Wakaleo hilmeri, W. vanderleueri, W. pitikantensis and Lekaneleo roskellyae (mm)

Species P3- M4 C1 - C1 P3 -P3 M4- M4 P3/M1- P3/M1 P3: M1 P3: P3 - M4 Specimen # length width width width width W. hilmeri QM F45200 34.0 19.9 39.7 33.5 41.6 1.34 0.34 QM F23443 35.8 25.1 45.7 35.5 44.3 1.30 0.34 W. vanderleueri CPC 26604 38.6* 24.4 46.4 37.3* 49.7 1.59 0.46* W. pitikantensis SAM P37719 31.4 - - - - [1.69] [0.4] L. roskellyae QM F23453 23.1 13.6 23.9 21.6 26.6 1.41 0.35 QM F23442 - - - - - 1.16 -

Note: * = measurements to M3 because W. vanderleueri lacks M4; [ ] = estimated value; - = not available; Measurements taken at the level of the alveoli, except P3/M1 taken at mid-crown; P3-M4 measure for SAM P37719 is approximate because of damage at both ends of the molar row; lengths are averages of left and right sides except for SAM P37719.

Table G: Dentary measurements of Wakaleo hilmeri, W. oldfieldi, W. vanderleueri and Lekaneleo roskellyae (mm)

Species length of molar P3: dentary dentary P3 row P3:M1 P3- M4 length depth Specimen # length P3 to base of below M2/3 P3 - M2 P3 - M4 condyle Wakaleo hilmeri QM F23443 9.5a 24.3 [32] - [0.29] [83.1] 21.8 QM F30465 9.5 27.9 35.1 1.0 0.27 - 18.2 QM F31376 8.6 24.4 - 1.0 - - 20.8 AR 19249 10.0^ 25.7 35.4^ 1.1 0.28^ 82.8 22.5 NTM P91167-3 12.2 - 37.5 1.3 0.32 - - W. oldfieldi SAM P17925 12.4 31.3 36.8* 1.1 0.33* 92.5 23.3 QM F20895 12.4 31.0 36.4* 1.1 0.34* - 24.2 W. vanderleueri NTM P8553-4 13.9 32.3 38.5* 1.4 0.36* 97.8 26.5 QM F36453 14.3 34.4 39.9* 1.2 0.36* - 24.0 Lekaneleo roskellyae QM F20191 6.4 18.2 25.0a 0.98 0.25 - 11.6 QM F29623 7.3 - - 1.15 - - -

* = no M4 in this species, measurement made to M3; a = alveolus; [ ] = estimate; ^ = average value of left and right

374

APPENDIX 2

Comparative taxa and specimens used in postcranial studies

Extinct taxa are denoted by an asterisk (*)

Family Thylacoleonidae *Thylacoleo carnifex AM F52398 SAM P38793 AM F 106768 (MF327) SAM P4007 AM F106804 SAM P13950/4 AM F106766 SAM P40009 SAM P12384 SAM P12384 SAM P39682 SAM F0/7 SAM P39678 FU 0405 SAM P39679 FU 0775 SAM P39680 FU 0427 SAM P39681 FU 0430 SAM P39256 SAM P733920 SAM P38793 FU 0396 SAM P40007 FU 0436 SAM P38795 FU 0437 SAM P39257 A6/12 FU 1125 SAM P16679b FU 0433 SAM P16679c SAM P38795 SAM P39675 SAM P18832 SAM P39676 SAM P38794 SAM P39677 SAM A6/12 SAM P39674 SAM P39259 SAM P39258

*Priscileo pitikantensis SAM P37720 *Wakaleo vanderleueri NTM P8695-95

Family Vombatidae Vombatus ursinus AM M33838 AM S1322 AM P793 AM M34032 AR 7640 Lasiorhinus latifrons AM M13514 AM M13515

Family Phascolarctidae Phascolarctos cinereus AM M35240 AM M26825 AM M23550 AM M34032 AM M P481 UNSWZ 2464 UNSWZ 2430

Family Wynyardiidae * Wynyardia bassiana (cast)

Family Diprotodontidae *Nimbadon lavarackorum QM F50714 QM F31563 375

QM F 41202a & e QM F41202 f QM F41097b QM F41104 QM F41095 QM F50608a AR 17902 AR 18242

Family Palorchestidae *Ngapakaldia tedfordi SAM P13851 (casts) *Ngapakaldia bonythoni SAM P13863 (casts)

Family Phalangeridae Trichosurus vulpecula AM M34417 AM M12686 AM M12683 AM M33806 AM S1851 AM S1304 UNSWZ 400 QVM 1984/1/239 QVM 1984/1/232 QVM 1982/1/159 T.v. fuliginosus AM S946 Trichosurus caninus AM M33831 AM M35406 Spilocuscus maculatus AM M11634 AM M23597 AM 23627 UNSWZ 181

Family Pseudocheiridae Pseudocheirops archeri AM M8950 Pseudocheirus peregrinus AM M2237 Pseudocheirus herbetensis AM M524

Family Macropodidae Macropus agilis AM M9164 Macropus rufogriseus AM A2575 Dendrolagus scottae AM M25397 Potorous tridactylus AM M23551 AM S1813 Aepyprymnus rufescens AM M9697 AM A1900

Family Thylacinidae Thylacinus cynocephalus AM 763

Family Dasyuridae Sarcophilus harrisii AM S1605 AM P714 UNSWZ 179 Dasyurus maculatus AM B5449 AM S497 AM S879 AM S1787

376

Dasyurus viverrinus UNSWZ teaching museum AM S1196

Family Peramelidae Perameles nasuta UNSWZ teaching museum AM S1074 AM S1072 AM M35409 Isoodon macrourus AM M33981

Family Notoryctidae Notoryctes typhlops SAM 637 AM 16165 CSIRO CM6010

377

APPENDIX 3

Specimens used in intrafamilial and interfamilial phylogenetic analyses

Taxa marked with an asterix (*) used in interfamilial phylogenetic analysis

Family Thylacoleonidae Microleo attenboroughi QM F411143

Priscileo pitikantensis SAM P37719

Priscileo deminutivus QM F24459

*Priscileo roskellyae QM F23453, QM F20191, QM F29623, QM F23442, QM F40116

Wakaleo hilmeri QM F45200, QM F23443, QM F31376, QM F30465, QM F24680

Wakaleo sp. cf. hilmeri QM F23446

Wakaleo oldfieldi SAM P17925, QMF20895, QM F36437, QM F23428

Wakaleo vanderleueri QM F36454, QM F41204

Wakaleo alcootaensis NTMP 1

Thylacoleo crassidentatus QM F7762, QM F10623, QM F10622

Thylacoleo carnifex AM F734, AM F739, AM F9, AM F21, AM F4660, AM F4661, AM F4662 (cast), AM F4663 (cast), AM F62525, AM F106844, V P159792

Thylacoleo hilli SAM P18621 (cast)

Thylacoleo sp. cf. T. hilli AM F63584

Family Dasyuridae *Barinya wangala QM F31408, QM F31409, QM F23889

Family Peramelidae *Echymipera kalubu AM M13765, AM M13766, AM M13769, AR 7236, AR 7250, AR 7251, AR 7258, AR 7261

Family Potoroidae *Bulungamaya sp. QM F40114, QM F41237

Family Phalangeridae *Caudipilosus vellicodens QM F42688, QM F23338, QM F36452, QM F40342

*Spilocuscus maculatus UNSWZ 38, UNSWZ 40, AR 7302, AR 11417, AM M12867, AM M5742, AM M23597, AM 23627

Family Pseudocheiridae *Pseudocheirops cupreus AR 11367, AM M15688, AM M15689, AM M15690, AM M13750

Family Petauridae *Petaurus breviceps AM M8591, AM M8375, AM M9548, AM M7941, AM M6816, UNSWZ 24 378

Family Burramyidae *Cercartetus caudatus AM M29579, AM M24341, AM M23906

Family Phascolarctidae *Litokoala kutjamarpensis QM F51382, QM F30500, QM F13079, QM F20809, AR 19361 *Nimiokoala greystanesi QM F30482, QM F30483, QM F30493, QM F30487, QM F30232, QM F29624, QM F23027

Family Wynyardiidae *Namilamadeta albivenator QM F41304, QM F41127

Family Ilariidae *Kuterintja ngama QM F20810, QM F30057, QM F30058, QM F31299, QM F23203, QM F40324

Family Vombatidae *Vombatus ursinus AM M22661, AM S1963, AM S1393, AM M2341, AR 5430, UNSWZ 3, QM J8790

Family Maradidae *Marada arcanum QM F42738

Family Palorchestidae

*Propalorchestes novaculacephalus QM F50605, NTM P8552-10, NTM P862-27, NMV P187282

Family Diprotodontidae *Nimbadon lavarackorum AR 17916, QM F50470, QM F31541, QM F40337, QM F30552, QM F31432, QM F31548

379

APPENDIX 4

4A. Intrafamilial analysis: apomorphies for nodes from Fig. 11.2 b (unordered analysis) b = equivocal transformations, => = unequivocal transformations common to ACCTRAN and DELTRAN optimisations. Transformations from DELTRAN indicated in italics, transformations from ACCTRAN are non-italicised.

Node Character CI Change A Thylacoleonidae 5 (P3 length: cheektooth row) 1.000 0 1 7 (P3 posterolingual crest) 0.500 0 2 11 (molar morphology) 1.000 0 3 0 3

15 (M2 length: P3 length) 1.000 1 2 32 (M1metaconule) 1.000 0 1 34 (M1 buccal gradient) 0.800 0 2 35 (M2 occlusal outline) 1.000 0 =>1 42 (supraorbital process) 0.800 0 1 50 (mastoid process) 1.000 0 2 51 (paroccipital process) 1.000 0 2 53 (medial glenoid process) 1.000 0 1 56 (postglenoid foramen position) 1.000 0 3 57 (tympanic wing components) 0.667 0 1 69 (lateral margins of post. palat. fenestrae) 0.500 0 1

B Priscileo, Wakaleo 5 (P3 length: cheektooth row) 1.000 0 1 & Thylacoleo 7 (P3 posterolingual crest) 0.500 0 2 21 (P2 location) 1.000 0 =>1 22 (P3 shearing facet) 1.000 0 =>2 23 (P3 length: cheektooth row) 1.000 0 =>1 25 (P3 longitudinal blade curvature) 1.000 0 =>1 26 (P3 posterior longitudinal blade) 0.800 0 =>2 29 (P3 anterolingual crest/cuspule) 1.000 0 =>1 32(M2 metaconule) 37 (M2 metaconule) 0.750 0 1 38 (M2 buccal expansion) 1.000 0 =>1 50 (mastoid process) 1.000 0 2 53 (medial glenoid process) 1.000 0 1 57 (postglenoid foramen position) 1.000 0 3 61 (tympanic wing components) 0.667 0 1

C 13 (M1width between entoconid and hypoconid)1.000 0 1 14 (M1 posterior width: anterior width) 1.000 0 1 15 (M2 length: P3 length) 1.000 2 3 16 (M2 trigonid height) 1.000 0 2 17 (M2 talonid basin width) 1.000 0 1 20 (M4 presence) 0.500 0 1 24 (P3 longitudinal blade length) 0.750 0 =>1 31 (M1 occlusal outline) 0.667 0 1 34 (M1 buccal gradient) 0.800 2 3 37 (M2 metaconule) 0.750 1 2 40 (skull shape) 1.000 1 2 42 (supraorbital process) 0.800 1 2 46 (sagittal crest) 1.000 1 2 47 (nuchal crest) 1.000 2 3 55 (postglenoid cavity) 1.000 1 2 62 (post-tympanic process of squamosal) 1.000 0 1

380

APPENDIX 4A cont.

Node Character CI Change C cont. 63 (sulcus of the stylomastoid) 1.000 1 2 64 (subsquamosal foramen) 1.000 0 1 65 (alisphenopalatine process) 0.500 0 1 66 (infraorbital crest) 0.667 1 2 68 (posterior palatine fenestra) 1.000 2 3 70 (palatal ridges) 0.750 0 1

D 16 (M2 trigonid height) 1.000 0 2 46 (sagittal crest) 1.000 1 2 47 (nuchal crest) 1.000 2 3 51 (paroccipital process) 1.000 0 2 55 (postglenoid cavity) 1.000 1 2 62 (post-tympanic process of squamosal) 1.000 0 1 63 (sulcus of stylomastoid) 1.000 1 2 65 (alisphenopalatine processes) 0.500 0 1 66 (infraorbital crest) 0.667 1 2 68 (posterior palatine fenestra) 1.000 2 3 E 2 (lower premolar number) 0.500 0 1

14 (M1 posterior width: anterior width) 1.000 0 1 27 (P3 mid-crown constriction) 0.750 0 =>1 28 (P3 posterobuccal crest) 1.000 0 =>4 29 (P3 anterolingual crest/cuspule) 1.000 1 2 33 (M1 buccal margin) 1.000 0 =>2 34 (M1 buccal gradient) 0.800 0 3 37 (M2 metaconule) 0.750 0 2 40 (skull shape) 1.000 1 2 48 (lachrymal fossa) 0.500 0 =>2 61 (rostral tympanic process) 0.750 0 3 64 (subsquamosal foramen) 1.000 0 1 69 (lateral margins of post. palatine fenestra)0.500 1 0 F 20 (M4 presence) 0.500 1 0 36 (M2 buccal height) 1.000 1 =>2 43 (supraorbital crests) 0.667 0 2 70 (palatal ridges) 0.750 0 1 G 2 (lower premolar number) 0.500 1 2 5 (P3 length: cheektooth row) 1.000 1 2

13 (M1 width between entoconid and hypoconid)1.000 1 2 15 (M2 length: P3 length) 1.000 3 4 17 (M2 talonid basin width) 1.000 1 2 18 (M3 presence and outline) 1.000 0 1 20 (M4 presence) 0.500 0 1 23 (P3 length: cheektooth row) 1.000 1 =>2 29 (P3 anterolingual crest/cuspule) 1.000 2 3 1 3 31 (M1 occlusal outline) 0.667 1 2 0 2 35 (M2 occlusal outline) 1.000 1 =>2 41 (frontal sinus) 0.667 1 2 42 (supraorbital process) 0.800 2 3 52 (alisphenoid ventrolateral wing) 0.667 0 1 56 (postglenoid foramen position) 1.000 3 4 70 (palatal ridges) 0.750 1 2 0 2 381

APPENDIX 4A cont.

Node Character CI Change G cont. 72 (masseteric process) 0.333 0 =>1 H 2 (lower premolar number) 0.500 0 2

5 (P3 length: cheektooth row) 1.000 1 2 6 (upper premolar number) 0.667 0 2

13 (M1 width between entoconid and hypoconid)1.000 0 2 17 (M2 talonid basin width) 1.000 0 2 24 (P3 longitudinal blade length) 0.570 1 =>2 32 (M1 metaconule) 1.000 1 =>2 34 (M1 buccal gradient) 0.800 3 4 3 4 36 (M2 buccal height) 1.000 1 =>3 37 (M2 metaconule) 0.750 2 3 I 6 (upper premolar number) 0.667 0 2

15 (M2 length: P3 length) 1.000 4 5 18 (M3 presence) 1.000 1 2 19 (M3 talonid basin) 0.500 0 1 27 (P3 mid-crown constriction) 0.750 1 =>2 35 (M2 occlusal outline) 1.000 2 =>3 J Thylacoleo 1 (I2 and canine) 0.667 1 =>2 5 (P3 length: cheektooth row) 1.000 1 3 10 (P3 posterobuccal crest) 1.000 0 =>2 12 (M1 trigonid height & talonid height) 0.667 1 2 13 (M1 width between entoconid and hypoconid)1.000 1 3 14 (M1 posterior width) 1.000 1 2 15 (M2 length: P3 length) 1.000 3 6 17 (M2 talonid basin width) 1.000 1 3 18 (M3 presence) 1.000 0 3 19 (M3 talonid basin) 0.500 0 1 23 (P3 length: cheektooth row) 1.000 1 3 24 (P3 longitudinal blade length) 0.750 1 2 26 (P3 posterior longitudinal blade) 0.800 2 3 30 (P3 posterior width) 1.000 0 =>1 32 (M1metaconule) 1.000 1 3 34 (M1 buccal gradient) 0.800 3 4 37 (M2 metaconule) 0.750 2 3 39 (M3) 1.000 0 1 40 (skull shape) 1.000 2 3 41 (frontal sinus) 0.667 1 0 42 (supraorbital process) 0.800 2 4 44 (parietal width) 0.500 0 1 45 (frontal/squamosal contact) 0.500 0 1 52 (alisphenoid ventrolateral wing) 0.667 0 2 54 (postglenoid process) 1.000 3 4 57 (tympanic wing components) 0.667 1 2 60 (epitympanic wing of petrosal) 0.667 0 2 64 (subsquamosal foramen) 1.000 1 2 67 (anterior palatine fenestra) 1.000 0 2 70 (palatal ridges) 0.750 1 3 K 5 (P3 length: cheektooth row) 1.000 1 3 7 (P3 posterolingual crest) 0.500 2 0 12 (M1 trigonid height) 0.667 1 2 13 (M1 width between entoconid and hypoconid)1.000 0 3 382

APPENDIX 4A cont.

Node Character CI Change

K cont. 14 (M1 posterior width: anterior width) 1.000 0 2 15 (M2 length: P3 length) 1.000 1 6 17 (M2 talonid) 1.000 0 3 18 (M3 presence) 1.000 0 3 19 (M3 talonid) 0.500 0 1 20 (M4 presence) 0.500 0 1 23 (P3 length: cheektooth row) 1.000 1 3 24 (P3 longitudinal blade length) 0.750 2 3 1 3 26 (P3 posterior longitudinal blade) 0.800 3 4 2 4 32 (M1 metaconule) 1.000 1 3 34 (M1 buccal gradient) 0.800 0 4 39 (M3 presence) 1.000 0 1 67 (anterior palatine fenestra) 1.000 0 2

______

4B. Intrafamilial analysis: Apomorphies for terminal taxa in Fig.11.2b, unordered parsimony analysis.

Microleo attenboroughi 16 (M2 trigonid height) 1.000 0 =>1 28 (P3 posterobuccal crest) 1.000 0 =>1 36 (M2 buccal height) 1.000 1 =>0

Priscileo deminutivus 26 (P3 posterior longitudinal blade) 0.800 2 =>1 28 (P3 posterobuccal crest) 1.000 0 =>2

Priscileo roskellyae 10 (P3 posterobuccal crest) 1.000 0 =>1 15 (M2 length: P3 length) 1.000 1 2 27 (P3 mid-crown constriction) 0.750 0 =>1 28 (P3 posterobuccal crest) 1.000 0 =>3 34 (M1 buccal gradient) 0.800 0 2 37 (M2 metaconule) 0.750 0 1 42 (supraorbital processes) 0.800 0 1 43 (supraorbital crests) 0.667 0 1 48 (lachrymal fossa) 0.500 0 =>1 54 (postglenoid process) 1.000 3 =>2 60 (epitympanic wing of petrosal) 0.667 0 =>1 61 (rostral tympanic process) 0.750 0 3 67 (anterior palatine fenestrae) 1.000 0 =>1 69 (lateral margin of post. palatine fenestra) 0.500 0 1

Wakaleo hilmeri 2 (lower premolar number) 0.500 0 1

13 (M1 width between entoconid and hypoconid)1.000 0 1 15 (M2 length: P3 length) 1.000 1 3 17 (M2 talonid basin width) 1.000 0 1 29 (P3 anterolingual crest/cuspule) 1.000 1 2 31 (M1 occlusal outline) 0.667 0 1 45 (supraorbital processes) 0.800 0 2 46 (supraorbital crests) 0.667 0 2

383

APPENDIX 4B cont.

Intrafamilial analysis: Apomorphies for terminal taxa in Fig. 11.2b, (unordered parsimony analysis).

Species Character CI Change Wakaleo sp. cf. hilmeri 26 (P3 posterior longitudinal blade) 0.800 2 =>3 48 (lachrymal fossa) 0.500 2 =>1

Wakaleo oldfieldi 15 (M2 length: P3 length) 1.000 1 4 18 (M3 presence) 1.000 0 1

Wakaleo vanderleueri 15 (M2 length: P3 length) 1.000 1 5 18 (M3 presence) 1.000 0 2 19 (M3 talonid basin) 0.500 0 1 41 (frontal sinus) 0.667 1 2 42 (supraorbital processes) 0.800 0 3 52 (alisphenoid ventrolateral wing) 0.667 0 1 56 (postglenoid foramen position) 1.000 3 4

Thylacoleo hilli 8 (P3 anterior crest) 1.000 0 =>1 9 (P3 anterobuccal crest) 1.000 1 =>0 24 (P3 longitudinal blade length) 0.750 1 2 26 (P3 posterior longitudinal blade) 0.800 2 3 27 (P3 mid-crown constriction) 0.750 0 =>3

Thylacoleo crassidentatus 7 (P3 posterolingual crest) 0.500 0 1 2 1 31 (M1 occlusal outline) 0.667 0 1 68 (posterior palatine fenestra) 1.000 3 =>4

Thylacoleo carnifex 7 (P3 posterolingual crest) 0.500 2 0 21 (P2 location) 1.000 1 =>2 31 (M1 occlusal outline) 0.667 1 0 35 (M2 presence) 1.000 0 =>1 37 (M2 metaconule) 0.750 0 3 40 (skull shape) 1.000 1 3 41 (frontal sinus) 0.667 1 0 42 (supraorbital process) 0.800 0 4 44 (parietal width) 0.500 0 1 45 (frontal/squamosal contact) 0.500 0 1 56 (alisphenoid ventrolateral wing) 0.667 0 2 54 (postglenoid process) 1.000 3 4 57 (tympanic wing components) 0.667 1 2 60 (epitympanic wing of petrosal) 0.667 0 2 61 (rostral tympanic process) 0.750 0 3 64 (subsquamosal foramen) 1.000 0 2 66 (infraorbital crest) 1.000 1 2 69 (lateral margin of post. palatal fenestra) 0.500 0 1 70 (palatal ridges) 0.750 0 3

384

APPENDIX 4C

4C. Intrafamilial analysis: apomorphies for nodes from Fig. 11.4 (ordered character-states) b = equivocal transformations, => = unequivocal transformations common to ACCTRAN and DELTRAN optimisations. Transformations from DELTRAN indicated in italics, transformations from ACCTRAN are non-italicised.

Node Character CI Change A Thylacoleonidae 5 (P3 length: cheektooth row) 1.000 0 1 7 (P3 posterolingual crest) 0.500 0 2 11 (molar morphology) 1.000 0 3 0 3

15 (M2 length: P3 length) 1.000 1 2 28 (P3 posterobuccal crest) 0.800 0 1 32 (M1 metaconule) 1.000 0 1 34 (M1 buccal gradient) 1.000 1 2 35 (M2 occlusal outline) 1.000 0 =>1 42 (supraorbital process) 0.800 0 1 50 (mastoid process) 1.000 0 2 51 (paroccipital process) 1.000 0 2 53 (medial glenoid process) 1.000 0 1 56 (postglenoid foramen position) 1.000 0 3 57 (tympanic wing components) 0.667 0 1 69 (lateral margins of post. palat. fenestrae) 0.500 0 1 B Priscileo, Wakaleo 5 (P3 length: cheektooth row) 1.000 0 1 & Thylacoleo 7 (P3 posterolingual crest) 0.500 0 2 15 (M2 length: P3 length) 1.000 1 2 21 (P2 location) 1.000 0 =>1 22 (P3 shearing facet) 1.000 0 =>2 23 (P3 length: cheektooth row) 1.000 0 =>1 25 (P3 longitudinal blade curvature) 1.000 0 =>1 26 (P3 posterior longitudinal blade) 0.800 0 =>2 27 (P3 midcrown constriction) 0.600 0 =>1 28 (P3 posterobuccal crest) 0.800 1 3 29 (P3 anterolingual crest/cuspule) 0.750 0 =>1 32 (M1 metaconule) 1.000 0 1 34 (M1 buccal gradient) 1.000 1 2 37 (M2 metaconule) 1.000 0 1 38 (M2 buccal expansion) 1.000 0 =>1 42 (supraorbital process) 0.800 0 1 50 (mastoid process) 1.000 0 2 53 (medial glenoid process) 1.000 0 1 56 (postglenoid foramen position) 1.000 0 3 57 (tympanic wing components) 0.667 0 1 C 2 (lower premolar number) 0.286 0 1 0 1 13 (M1width between entoconid and hypoconid)1.000 0 =>1 14 (M1 posterior width: anterior width) 1.000 0 =>1 15 (M2 length: P3 length) 1.000 2 =>3 16 (M2 trigonid height) 1.000 1 =>2 17 (M2 talonid basin width) 1.000 0 =>1 24 (P3 longitudinal blade length) 0.750 0 =>1 28 (P3 posterobuccal crest) 0.800 3 4 0 4 31 (M1 occlusal outline) 0.500 0 =>1 385

APPENDIX 4C cont.

Node Character CI Change C cont. 33 (M1 buccal margin) 0.667 0 =>2 34 (M1 buccal gradient) 1.000 2 =>3 36 (M1 buccal gradient) 0.750 1 2 37 (M2 metaconule) 1.000 1 2 0 2 40 (skull shape) 1.000 1 =>2 42 (supraorbital process) 0.800 1 =>2 46 (sagittal crest) 1.000 1 =>2 47 (nuchal crest) 1.000 2 =>3 48 (lachrymal foramen) 0.400 1 2 51 (paroccipital process) 1.000 0 2 55 (postglenoid cavity) 1.000 1 =>2 61 (rostral tympanic process) 0.600 0 3 62 (post-tympanic process of squamosal) 1.000 0 =>1 63 (sulcus of the stylomastoid) 1.000 1 =>2 64 (subsquamosal foramen) 1.000 0 =>1 65 (alisphenopalatine process) 0.500 0 =>1 66 (infraorbital crest) 0.667 1 =>2 68 (posterior palatine fenestra) 1.000 2 =>3 70 (palatal ridges) 0.750 0 1 D 2 (lower premolar number) 0.286 1 2 5 (P3 length: cheektooth row) 1.000 1 2

13 (M1 width between entoconid and hypoconid)1.000 1 2 15 (M2 length: P3 length) 1.000 3 4 17 (M2 talonid basin width) 1.000 1 2 18 (M3 presence and outline) 1.000 0 1 20 (M4 presence) 1.000 0 =>1 23 (P3 length: cheektooth row) 1.000 1 =>2 1 3 29 (P3 anterolingual crest/cuspule) 0.750 1 =>3 31 (M1 occlusal outline) 0.500 1 =>2 35 (M2 occlusal outline) 1.000 1 =>2 41 (frontal sinus) 0.667 1 0 42 (supraorbital process) 0.800 2 3 52 (alisphenoid ventrolateral wing) 0.667 0 1 56 (postglenoid foramen position) 1.000 3 4 70 (palatal ridges) 0.750 1 2 0 2 72 (masseteric process) 0.250 0 1 E 43 (supraorbital crests) 0.667 0 2 F 29 (P3 anterolingual crest/cuspule) 0.750 1 2 36 (M1 buccal gradient) 0.750 1 2 70 (palatal ridges) 0.750 0 1

G 5 (P3 length: cheektooth row) 1.000 1 2 13 (M1 width between entoconid and hypoconid)1.000 1 2 15 (M2 length: P3 length) 1.000 3 4 17 (M2 talonid basin width) 1.000 0 2 18 (M3 presence and outline) 1.000 0 1 24 (P3 longitudinal blade length) 1.000 1 =>2 32 (M1 metaconule) 1.000 1 =>2 34 (M1 buccal gradient) 1.000 3 =>4

386

APPENDIX 4C cont.

Node Character CI Change G cont. 36 (M2 buccal height) 0.750 2 3 1 2 37 (M2 metaconule) 1.000 2 =>3 H 15 (M2 length: P3 length) 1.000 4 =>5 18 (M3 presence) 1.000 1 =>2 19 (M3 talonid basin) 1.000 0 =>1 27 (P3 midcrown constriction) 0.600 1 0 35 (M2 occlusal outline) 1.000 2 3 42 (supraorbital) 0.800 2 3 52 (alisphenoid ventrolateral wing) 0.667 0 1 I 6 (upper premolar number) 0.500 0 =>2 27 (P3 mid-crown constriction) 0.600 0 2 1 2 35 (M2 occlusal outline) 1.000 2 3 41 (frontal sinus) 0.667 0 2 56 (postglenoid foramen position) 1.000 3 4 J 1 (I2 and canine) 0.667 1 =>2 2 (lower premolar number) 0.286 1 =>0 5 (P3 length: cheektooth row) 1.000 2 3 10 (P3 posterobuccal crest) 1.000 0 =>2 12 (M1 trigonid height & talonid height) 0.667 1 2 13 (M1 width between entoconid and hypoconid)1.000 2 3 14 (M1 posterior width) 1.000 1 2 15 (M2 length: P3 length) 1.000 5 6 17 (M2 talonid basin width) 1.000 2 3 18 (M3 presence) 1.000 2 3 23 (P3 length: cheektooth row) 1.000 2 3 26 (P3 posterior longitudinal blade) 0.800 2 3 28 (P3 posterobuccal crest) 0.800 4 =>0 29 (P3 anterolingual crest/cuspule) 0.750 3 =>1 30 (P3 posterior width) 1.000 0 =>1 31 (M1 occlusal outline) 0.500 2 1 32 (M1metaconule) 1.000 2 3 33 (M1 buccal margin) 0.667 2 0 39 (M3) 1.000 0 1 40 (skull shape) 1.000 2 3 42 (supraorbital process) 0.800 3 4 44 (parietal width) 0.500 0 1 45 (frontal/squamosal contact) 0.500 0 1 48 (lachrymal foramen) 0.400 2 3 52 (alisphenoid ventrolateral wing) 0.667 1 2 54 (postglenoid) 1.000 3 4 57 (tympanic wing components) 0.667 1 2 60 (epitympanic wing of petrosal) 0.667 0 2 61 (rostral tympanic process) 0.600 3 0 64 (subsquamosal foramen) 1.000 1 2 67 (anterior palatine fenestra) 1.000 0 2 69 (lateral margins of post. pal. fenestrae) 0.500 0 1 70 (palatal ridges) 0.750 2 3

K 5 (P3 length: cheektooth row) 1.000 2 3 7 (P3 posterolingual crest) 0.500 0 1 12 (M1 trigonid height) 1.000 1 2 387

APPENDIX 4C cont.

Node Character CI Change

K cont. 13 (M1 width between entoconid and hypoconid)1.000 2 3 14 (M1 posterior width: anterior width) 1.000 1 2 15 (M2 length: P3 length) 1.000 5 6 17 (M2 talonid) 1.000 2 3 18 (M3 presence) 1.000 0 3 23 (P3 length: cheektooth row) 1.000 1 3 24 (P3 longitudinal blade length) 1.000 2 =>3 26 (P3 posterior longitudinal blade) 0.800 3 4 2 4 27 (P3 midcrown constriction) 0.600 1 0 31 (M1 occlusal outline) 0.500 2 1 32 (M1 metaconule) 1.000 2 3 33 (M1 buccal margin) 0.667 2 0 39 (M3 presence) 1.000 0 1 67 (anterior palatine fenestra) 1.000 0 2

388

APPENDIX 5

A. Interfamilial analysis: nodal apomorphies, character-states unordered Transformations from DELTRAN optimisation indicated in italics, transformations from ACCTRAN are non-italicised. b = equivocal transformations => = unequivocal transformations common to ACCTRAN and DELTRAN optimisations.

Node Character CI Change A Order Diprotodontia (2) masseteric foramen 0.333 0 =>1 (5) mental foramen 0.667 0 1 (6) depth of horizontal ramus 0.500 0 => 1 (7) number of lower incisors 1.000 0 => 1 (8) number of upper incisors 1.000 0 => 1 (9) procumbent lower incisor 1.000 0 => 1 (10) lower canine 1.000 0 => 1 (12) lower premolar number 0.600 0 => 3 (13) upper premolar number 0.500 0 2 (26) molar crown structure 0.571 0 => 1 (27) paraconid 0.500 0 => 1 (30) M1 stylar cusps 0.600 0 2 (34) M3 occlusal outline 0.429 0 3 (35) M1 Pa and Me position 0.500 0 1 (37) M1 Pa vs Me size 0.286 0 => 1 (38) M2 Pa vs Me size 0.667 0 => 2 (39) M3 Pa vs Me size 0.667 0 => 2 (40) M1 paraconule 0.200 0 1 (44) nuchal crest 0.400 0 1 (49) glenoid fossa 0.667 0 1 (51) postglenoid foramen location 0.100 0 => 1 (53) postglenoid cavity 1.000 1 0 (57) postglenoid foramen 1.000 0 =>1 (54) median occipital crest 0.500 0 =>1 (55) tympanic wing components 0.800 0 =>2 (58) rostral tympanic process 0.600 0 3 (61) medial wall of mandibular fossa 1.000 0 => 1 (65) frontal/nasal contact 0.333 0 1 B Suborder (30) M1 stylar cusps 0.600 2 3 Phalangerida 0.600 0 3 (31) M1 stylar shelf 0.500 0 => 1 (35) M1 Pa and Me position 0.500 1 2 0.500 1 2 (36) M2 Pa and Me position 0.333 0 => 1 (48) fusion of ectotympanic 1.000 0 => 1 (49) glenoid fossa 0.667 0 1 (56) alisphenoid tympanic wing 0.800 0 => 1 (57) alisphenoid ventrolateral wing 0.250 0 1 C (22) P3 posterolingual crest/cusp 0.600 0 2 (26) molar crown structure 0.571 1 => 4 (29) M2 cristid oblique 0.750 0 1 (38) M2 Pa vs Me size 0.667 2 1 (39) M3 Pa vs Me size 0.667 2 1 (40) paraconule 0.200 1 0 (53) postglenoid cavity 1.000 0 2 (55) tympanic wing components 0.800 1 2 (63) foramen ovale 0.500 0 1 (65) frontal/ nasal contact 0.333 1 0

389

APPENDIX 5 cont.

A. Interfamilial analysis: nodal apomorphies, character-states unordered

Node Character CI Change D (5) mental foramen 0.667 1 4 0.667 0 4 (12) lower premolar number 0.600 3 1 (13) upper premolar number 0.500 2 0 (17) P3 crown structure 0.375 0 2 (34) M3 occlusal outline 0.429 3 1 (44) nuchal crest 0.400 1 2 0.400 0 2 (45) occiput height 1.000 0 => 1 (51) postglenoid foramen 1.000 1 => 2 (52) postglenoid process 0.800 0 1 (53) postglenoid cavity 1.000 1 0 (58) rostral tympanic process 0.600 0 3 (59) subsquamosal foramen 0.500 0 1 (60) squamosal pneumatisation 1.000 0 => 1 (62) mastoid process 0.500 0 1 E Suborder (1) dentary symphysis 0.500 0 1 Vombatiformes (5) mental foramen 0.667 1 2 0.667 0 2 (13) upper premolar number 0.500 0 2 (17) P3 crown structure 0.375 0 => 1 (20) P3 anterobuccal crest 0.333 0 => 1 (22) P3 posterolingual cusp/crest 0.600 0 => 1 (253 P3 anterolingual crest 0.250 0 => 1 (24) P3 anterobuccal crest 0.333 0 => 1 (30) M1 stylar cusps 0.600 0 2 (35) M1 paracone & metacone position 0.500 0 1 (32) M1 occlusal outline 0.429 1 2 (40) paraconule 0.200 0 1 (41) supraorbital process 0.333 0 1 (47) masseteric process 0.200 0 => 1 (49) glenoid fossa 0.667 1 3 (52) postglenoid process 0.800 0 2 (53) postglenoid cavity 1.000 1 3 F (4) ascending ramus 0.333 0 1 (19) P3 posterolingual crest 0.500 0 => 1 (21) P3 anterolingual crest 1.000 0 => 1 G Infraorder (1) dentary symphysis 0.500 0 1 Vombatomorphia (17) P3 crown structure 0.375 1 3 (32) M1 occlusal outline 0.429 2 1 (37) M1 Pa vs Me size 0.286 1 2 (44) nuchal crest 0.400 1 0 (46) interparietal 0.333 0 => 1 (49) glenoid fossa 0.667 1 4 0.667 0 4 (50) medial glenoid process 0.500 0 => 1 (51) postglenoid foramen 1.000 1 => 3 (53) postglenoid cavity 1.000 4 5 1.000 1 5 (55) tympanic wing components 0.800 1 => 3 (56) alisphenoid tympanic wing 0.800 0 2 0.800 0 2

390

APPENDIX 5 cont.

A. Interfamilial analysis: nodal apomorphies, character-states unordered.

Node Character CI Change G cont. (62) mastoid process 0.500 0 => 3 (63) foramen ovale 0.500 0 => 1 H (4) ascending ramus 0.333 0 1 (5) mental foramen 0.667 2 3 0.667 2 3 (30) M1 stylar cusps 0.600 2 => 1 (32) M1 occlusal outline 0.429 1 3 0.429 1 3 (33) M2 occlusal outline 0.667 1 => 2 (34) M3 occlusal outline 0.429 0 3 (40) paraconule 0.200 1 => 0 (43) frontal/squamosal contact 0.333 0 => 1 (64) lachrymal tuberosity 0.667 0 => 2 I (11) upper canine 1.000 0 1 (17) P3 crown structure 0.375 1 0 (18) P3 posterior cingulum 0.500 0 => 1 (19) P3 posterolingual crest 0.500 1 => 0 (44) nuchal crest 0.400 0 1 (49) glenoid fossa 0.667 4 3 (56) alisphenoid tympanic wing 0.800 2 3 J (20) P3 anterobuccal crest 0.333 1 => 0 (22) P3 posterolingual cusp/crest 0.600 1 0 (23) P3 anterolingual crest 0.250 1 => 0 (24) P3 anterobuccal crest 0.333 1 => 0 (25) molar gradient 0.667 0 => 2 (30) M1 stylar cusps 0.600 1 3 (44) nuchal crest 0.400 0 1 K (22) P3 posterolingual cusp/crest 0.600 0 3 0.600 1 3 (26) molar crown structure 0.571 1 => 4 (29) M2 cristid oblique 0.750 0 5 (30) M1 stylar cusps 0.600 1 3 (31) M1 stylar shelf 0.500 0 => 1 (35) M1 Pa and Me placement 0.500 1 => 2 (36) M2 Pa and Me placement 0.333 0 => 1 (42) parietal width 0.500 0 => 1 (53) postglenoid cavity 1.000 5 => 3 ______

5 B. Interfamilial Analysis: Nodal apomorphies, character-states ordered

Transformations from DELTRAN optimisation indicated in italics, transformations from ACCTRAN are non-italicised. b = equivocal transformations => = unequivocal transformations common to ACCTRAN and DELTRAN optimisations.

Node Character CI Change A Order Diprotodontia (5) mental foramen 0.800 0 => 4 (7) number of lower incisors 1.000 0 => 1 (8) number of upper incisors 1.000 0 => 1 (9) procumbent lower incisor 1.000 0 => 1 (10) lower canine 1.000 0 => 1 391

APPENDIX 5 cont.

B. Interfamilial analysis: Nodal apomorphies, ordered character-states

Node Character CI Change A cont. (12) lower premolar number 0.500 0 => 3 (25) molar gradient 0.500 0 1 (26) molar crown structure 0.571 0 => 1 (27) paraconid 0.500 0 => 1 (29) M2 cristid oblique 0.667 0 3 (30) M1 stylar cusps 0.600 0 => 2 (31) M1 stylar shelf 0.333 0 => 1 (35) M1 Pa and Me position 0.400 0 => 1 (36) M2 Pa and Me position 0.250 0 => 1 (37) M1 Pa vs Me 0.286 0 => 1 (38) M2 Pa vs Me 0.400 0 => 2 (39) M3 Pa vs Me 0.500 0 => 2 (44) nuchal crest 0.400 0 => 2 (45) height of occiput 0.500 0 => 1 (48) ectotympanic fusion 0.500 0 => 1 (49) glenoid fossa 0.667 0 => 1 (51) postglenoid foramen 1.000 0 => 2 (53) postglenoid cavity 1.000 1 => 0 (55) tympanic wing 0.800 0 => 1 (56) alisphenoid tympanic wing 0.667 0 1 0.667 0 1 (58) rostral tympanic process 0.600 0 3 0.600 0 3 (59) subsquamosal foramen 0.333 0 => 1 (60) squamosal pneumatisation 0.500 0 => 1 (61) medial wall of mandibular fossa 1.000 0 => 1 (65) frontal/nasal contact 0.333 0 => 1 B Suborder (2) masseteric foramen 0.500 0 => 1 Phalangerida (6) depth of horizontal ramus 1.000 0 => 1 (34) M3 occlusal outline 0.375 0 1 0.375 0 1 (40) paraconule 0.250 0 => 1 (52) postglenoid process 0.800 1 => 2 C (12) lower premolar number 0.500 1 => 2 (17) P3 crown structure 0.375 0 1 (23) P3 anterolingual crest 0.250 0 1 (24) P3 anterobuccal crest 0.333 0 => 1 (25) molar gradient 0.500 1 0 (26) molar crown structure 0.571 2 => 1 (42) parietal width 0.333 0 1 (46) interparietal 0.250 0 1 (47) masseteric process 0.200 0 1 (49) glenoid fossa 0.667 1 3 (54) median occipital crest 1.000 0 => 1 (62) mastoid process 0.500 1 => 2 D (5) mental foramen 0.800 4 => 3 (13) upper premolar number 0.500 0 => 1 (20) P3 anterobuccal crest 0.333 0 1 (32) M1 occlusal outline 0.375 1 2 (34) M3 occlusal outline 0.375 1 => 2 (55) tympanic wing 0.800 1 => 2 E (2) masseteric foramen 0.500 1 => 0 (15) P3 buccally deflected 1.00 0 => 1 392

APPENDIX 5 cont.

B. Interfamilial analysis: Nodal apomorphies, ordered character-states

Node Character CI Change E cont. (16) P3 serrated 0.333 0 => 1 (17) P3 crown structure 0.333 1 => 2 (20) P3 anterobuccal crest 0.333 0 1 (26) molar crown structure 0.571 1 => 3 (29) M2 cristid obliqua 0.667 0 => 2 (32) M1 occlusal outline 0.375 2 => 3 (33) M2 occlusal outline 0.667 1 => 2 (34) M3 occlusal outline 0.375 2 3 (42) parietal width 0.333 0 1 (57) alisphenoid ventrolateral wing 0.250 0 1 F (12) lower premolar number 0.500 2 => 3 (13) upper premolar number 0.500 1 => 2 (22) P3 posterolingual cusp/crest 0.500 0 => 1 (42) parietal width 0.333 1 0 (44) nuchal crest 0.400 2 1 (45) height of occiput 0.500 1 => 0 (46) interparietal 0.250 1 0 (51) postglenoid foramen 1.000 2 => 1 (52) postglenoid process 0.800 2 1 (53) postglenoid cavity 1.000 0 2 (59) subsquamosal foramen 0.333 1 => 0 (60) squamosal pneumatisation 0.500 1 => 0 (62) mastoid process 0.500 2 => 0 G (17) P3 crown structure 0.375 1 0 (20) P3 anterobuccal crest 0.333 1 0 (22) P3 posterlingual crest 0.500 1 2 (23) P3 anterolingual crest 0.250 1 0 (24) P3 anterobuccal crest 0.333 1 0 (26) molar crown structure 0.571 1 => 4 (29) M2 cristid oblique 0.667 0 1 (32) M1 occlusal outline 0.375 2 1 (38) M2 Pa vs Me size 0.400 2 1 (39) M3 Pa vs Me size 0.667 2 1 (40) paraconule 0.250 1 0 (47) masseteric process 0.200 1 0 (49) glenoid fossa 0.667 4 1 (65) frontal/ nasal contact 0.333 1 0 H (1) dentary symphysis 0.500 0 1 (5) mental foramen 0.800 3 2 (17) P3 crown structure 0.375 0 1 (20) P3 anterobuccal crest 0.333 0 1 (23) P 3 anterolingual crest 0.250 0 1 (30) M1 stylar cusps 0.600 3 => 2 (31) M1 stylar shelf 0.333 1 => 0 (32) M1 occlusal outline 0.375 1 2 (34) M3 occlusal outline 0.375 3 2 (35) M1 Pa and Me position 0.400 2 => 1 (36) M2 Pa and Me position 0.250 1 => 0 (41) supraorbital process 0.333 0 1 (47) masseteric process 0.200 0 1 (48) ectotympanic fusion 0.500 1 0 (52) postglenoid process 0.800 1 3 (53) postglenoid cavity 1.000 2 4 393

APPENDIX 5 cont.

B. Interfamilial analysis: nodal apomorphies, ordered character-states

Node Character CI Change H cont. (56) alisphenoid tympanic wing 0.667 1 0 (57) alisphenoid ventrolateral wing 0.250 1 0 I (4) ascending ramus 0.333 0 1 (19) P3 posterolingual crest 0.500 0 => 1 (21) P3 anterolingual crest 1.000 0 => 1 (41) supraorbital process 0.333 0 1 (48) ectotympanic fusion 0.500 1 0 (52) postglenoid process 0.800 2 3 J (1) dentary symphysis 0.500 0 1 (17) P3 crown structure 0.375 1 3 (44) nuchal crest 0.400 1 0 (46) interparietal 0.250 0 1 0.250 0 1 (49) glenoid fossa 0.667 1 4 (50) medial glenoid process 0.500 0 => 1 (51) postglenoid foramen 1.000 1 => 3 (53) postglenoid cavity 1.000 0 5 1.000 4 5 (55) tympanic wing components 0.800 2 => 3 (56) alisphenoid tympanic wing 0.667 1 2 0.667 0 2 (62) mastoid process 0.500 0 => 3 (63) foramen ovale 0.500 0 1 K (4) ascending ramus 0.333 0 1 (5) mental foramen 0.800 2 3 (30) M1 stylar cusps 0.600 2 => 1 (32) M1 occlusal outline 0.375 2 3 (33) M2 occlusal outline 0.667 1 => 2 (34) M3 occlusal outline 0.375 2 3 0.375 2 3 (40) paraconule 0.250 1 => 0 (43) frontal/squamosal contact 0.333 0 => 1 (64) lachrymal tuberosity 0.667 0 => 2 L (11) upper canine 1.000 0 1 (17) P3 crown structure 0.375 3 0 (18) posterior cingulum on P3 0.500 0 => 1 (19) P3 posterolingual crest 0.500 1 => 0 (44) nuchal crest 0.400 0 1 (49) glenoid fossa 0.667 4 3 (56) alisphenoid tympanic wing 0.667 2 3 M (11) upper canine 1.000 0 1 (17) P3 crown structure 0.375 1 0 (20) P3 anterobuccal crest 0.333 1 => 0 (23) P3 anterolingual crest 0.250 1 => 0 (24) P3 anterobuccal crest 0.333 1 => 0 (25) molar gradient 0.500 0 => 2 (30) M1 stylar cusps 0.600 1 3 (32) M1 occlusal outline 0.375 2 3 N (22) P3 posterolingual cusp/crest 0.500 1 => 3 (26) molar crown structure 0.571 1 => 4 (29) M2 cristid obliqua 0.667 0 5 (30) M1 stylar cusps 0.600 1 3 (31) M1 stylar shelf 0.333 0 => 1 394

APPENDIX 5 cont.

B. Interfamilial analysis: nodal apomorphies, ordered character-states

Node Character CI Change N cont. (35) M1 Pa and Me location 0.400 1 => 2 (36) M2 Pa and Me location 0.250 0 => 1 (42) parietal width 0.333 0 => 1 (53) postglenoid cavity 1.000 5 => 3 ______