PALAEONTOLOGIA AFRICANA PALAEONTOLOGIA AFRICANA Volume 44 Annals of the Bernard Price Institute for Palaeontological Research December 2009 ISSN 0078-8554 Supported by PALAEONTOLOGICAL SCIENTIFIC TRUST P A L A E O N T O L O G IA A F R IC A N A V O L U M E 4 4 , 2 0 0 9 PA L A E O N T O L O G I C A L S C I E N T I F I C T R U S T SCHOOL OF GEOSCIENCES BERNARD PRICE INSTITUTE FOR PALAEONTOLOGICAL RESEARCH Academic Staff Director and Chair of Palaeontology B.S. Rubidge BSc (Hons), MSc (Stell), PhD (UPE) Deputy Director M.K. Bamford BSc (Hons), MSc, PhD (Witwatersrand) Senior Research Officers F. Abdala BSc, PhD (UNT, Argentina) A.M. Yates, BSc (Adelaide), BSc (Hons), PhD (La Trobe) Research Officer L.R. Backwell BA (Hons), MSc, PhD (Witwatersrand) Collections Curator B. Zipfel NHD Pod., NHD PS Ed. (TWR), BSc (Hons) (Brighton), PhD (Witwatersrand) Post Doctoral Fellows R. Mutter BSc, MSc, PhD (Zurich, Switzerland) F. Neumann BSc, MSc, PhD (Friedrich-Wilhelms-Univer- sity, Bonn) D. Steart BSc, DipEd (La Trobe University), PhD (Victoria University of Technology, Australia) Editorial Panel M.K. Bamford: Editor L.R. Backwell: Associate Editor B.S. Rubidge: Associate Editor A.M. Yates: Associate Editor Consulting Editors Dr J.A. Clack (Museum of Zoology, University of Cambridge, Cambridge, U.K.) Dr H.C. Klinger (South African Museum, Cape Town) Dr K. Padian (University of California, Berkeley, California, U.S.A.) Dr K.M. Pigg (Arizona State University, Arizona, U.S.A.) Prof. L. Scott (University of the Free State, Bloemfontein) Dr R.M.H. Smith (South African Museum, Cape Town) Technical and Support Staff Principal Technician R. McRae-Samuel Senior Administrative Secretary S.C. Tshishonga Assistant Research Technician C.B. Dube Technician/Fossil Preparator P. Chakane S. Jirah P.R. Mukanela G. Ndlovu T. Nemavhundi S. Tshabalala Custodian, Makapansgat Sites S. Maluleke Honorary Staff Honorary Research Associates K. Angielczyk BSc (Univ of Michigan, Ann Arbor), PhD (Univ California, Berkeley) P.J. Hancox BSc(Hons), MSc, PhD (Witwatersrand) R. Reisz BSc, MSc, PhD (McGill Univ., Montreal) C.A. Sidor BSc (Trinity College) MSc, PhD (Univ. Chicago) INSTITUTE FOR HUMAN EVOLUTION Academic Staff Director J.F. Thackeray BSc (Hons), MSc (UCT), MPhil, PhD (Yale) Reader L.R. Berger BA (Hons) (GA Southern), PhD (Witwaters- rand) Research Officer K. Carlson BSc (Michigan, Ann Arbor), MSc, PhD (Indiana, Bloomington) J. Kibii BSc, (Nairobi), MSc, PhD (Witwatersrand) B. Kuhn BSc (Washington State), MSc (UCL), PhD (UP) Research Associates C. Henshilwood BA (Hons) (UCT), PhD (Cambridge) M. Lombard BA (Hons), MA (UNISA), PhD (Witwaters- rand) Administrative and Support Staff Senior Administrative Secretary Evlyn Ho Preparators T. Dingiswayo T. Makhele N. Molefe A. Mollepolle S. Motsumi L. Sekowe M. Seshoene Honorary Staff Honorary Research Associates Dr Shaw Badenhorst, Transvaal Museum Prof. Steve Churchill, Duke University, USA Prof. Francesco d'Errico, BordeauxUniversity, France Prof. Daryl de Ruiter, Texas A&M University, USA Prof. Katerina Harvati, University of Tubingen, Germany Prof. Jacopo Moggi-Cecchi, Laboratori di Antropologia, Dipartimento di Biologia Animale e Genetica, Università di Firenze, Italy Prof. Peter Schmidt, Zurich, Switzerland Prof. Himla Soodyall, Human Genomic Diversity and Disease Research Unit (HGDDRU), South African Medical Research Council in conjunction with the National Health Laboratory Service and University of the Witwatersrand Articles Kemp, T.S. — Phylogenetic interrelationships and pattern of evolution of the therapsids: testing for polytomy Nicolas, M. & Rubidge, B.S. — Assessing content and bias in South African Permo-Triassic Karoo tetrapod fossil collections Weide, D.M., Sidor, C.A., Angielczyk, K.D. & Smith, R.M.H. — A new record of Procynosuchus delaharpeae (Therapsida: Cynodontia) from the Upper Perm- ian Usili Formation, Tanzania Fourie, H. & Rubidge, B.S. — The postcranial skeleton of the basal thero- cephalian Glanosuchus macrops (Scylacosauridae) and comparison of morpho- logical and phylogenetic trends amongst the Theriodontia Govender, R. & Yates, A. — Dicynodont postcrania from the Triassic of Namibia and their implication for the systematics of Kannemeyeriiforme dicynodonts Geraads, D., Melillo, S. & Haile-Selassie, Y. — Middle Pliocene Bovidae from Hominid-bearing sites in the Woranso-Mille area, Afar region, Ethiopia Zipfel, B. & Berger, L.R. — Partial hominin tibia (StW 396) from Sterkfontein, South Africa Technical Note Zipfel, B. & Berger, L.R. — New Cenozoic fossil- bearing site abbreviations for collections of the University of the Witwatersrand Abstracts 15th Biennial Meeting of the Palaeontological Society of Southern Africa •Reviewed Extended Abstracts Abdala, F., Martinelli, A.G., Bento Soares, M., de la Fuente, M. & Ribeiro, A.M. — South American Middle Triassic continental faunas with amniotes: biostratigraphy and correlation Atayman, S., Rubidge, B.S. & Abdala, F. — Taxonomic re-evaluation of tapinocephalid dino- cephalians Backwell, L. & d’Errico, F. — Additional evidence of early hominid bone tools from South Africa. First attempt at exploring inter-site variability Bordy, E.M., Sztanó, O., Rubidge, B.S. & Bumby, A. — Tetrapod burrows in the southwestern main Karoo Basin (Lower Katberg Formation, Beaufort Group), South Africa Browning, C. — Nodular preservation of trilobite fossils from the Bokkeveld Group, Eastern Cape Province, South Africa Manegold, A. — The early fossil record of perching birds (Passeriformes) Mannion, P.D. — Review and analysis of African sauropodomorph dinosaur diversity Matthews, T., Marean, C. & Nilssen, P. — Micromammals from the Middle Stone Age (92–167 ka) at Cave PP13B, Pinnacle Point, south coast, South Africa Mostovski, M.B. — Brachyceran assemblages (Insecta: Diptera) as indicators of terrestrial palaeo- environments in the Late Mesozoic Ovechkina, M.N., Watkeys, M. & Kretzinger, W. — Nannoplankton in the manganese deposits of the Mozambique Ridge and Mozambique Basin, southwestern Indian Ocean Ovechkina, M.N., Watkeys, M. & Mostovski, M.B. — Calcareous nannofossils from the stratotype section of the Upper Cretaceous Mzamba Formation, Eastern Cape, South Africa Rubidge, B. & Angielczyk, K. — Stratigraphic ranges of Tapinocephalus Assemblage Zone dicyno- donts: implications for middle Permian continental biostratigraphy van Dijk, D.E. — Continental displacement: early lines of evidence that deserve attention •Abstracts •Poster Abstracts 1 13 21 27 41 59 71 77 83 88 91 95 100 103 108 112 121 126 129 134 136 139 190 Volume 44, December 2009 PALAEONTOLOGIA AFRICANA CONTENTS ANNALS OF THE BERNARD PRICE INSTITUTE FOR PALAEONTOLOGICAL RESEARCH UNIVERSITY OF THE WITWATERSRAND ISSN 0078-8554 © 2009 BERNARD PRICE INSTITUTE for PALAEONTOLOGICAL RESEARCH School of Geosciences University of the Witwatersrand Johannesburg ACKNOWLEDGEMENTS The Bernard Price Institute for Palaeontological Research gratefully acknowledges financial support for its programmes by THE COUNCIL’S RESEARCH COMMITTEE, UNIVERSITY OF THE WITWATERSRAND NATIONAL RESEARCH FOUNDATION (NRF) DEPARTMENT OF SCIENCE AND TECHNOLOGY (DST) and the PALAEONTOLOGICAL SCIENTIFIC TRUST (PAST) for publication of this journal Pre-press production by Isteg Scientific Publications, Irene Printed in South Africa by Ultra Litho (Pty) Ltd, Heriotdale, Johannesburg Phylogenetic interrelationships and pattern of evolution of the therapsids: testing for polytomy Tom S. Kemp Museum of Natural History and St John’s College, Oxford, OX1 3PW, U.K. E-mail: tom.kemp@sjc.ox.ac.uk Received 11 October 2008. Accepted 4 May 2009 INTRODUCTION The amniote clade Therapsida is highly significant in the history of terrestrial tetrapod life for two, no doubt interrelated reasons. One is that the anatomy indicates that from the very start therapsids were evolving the increased energy budgets and activity levels that were to culminate in the mammals, with the latter’s huge potential for physiological and anatomical diversification. The second is that it was the nonmammalian therapsids that were primarily responsible for establishing what was to become the standard structure of fully terrestrial tetrapod communities – a fauna dominated by a very large prepon- derance of diverse amniote herbivore species. This struc- ture was later repeated in essence by the dinosaur-domi- nated communities of the later Mesozoic, and by the mammalian-dominated communities of the Tertiary right up to the present day. However, the details of the early radiation of the Therapsida are shrouded in obscurity. The earliest fossil record of several derived lineages occurs approximately simultaneously, and cladistic analysis has so far led only to a number of very weakly-supported and ambiguous interrelationships amongst these lineages. Indeed, the situation is remarkably comparable to the morphological analysis of placental mammal interrelationships. Here the taxon Placentalia on the one hand, and the individual constituent orders on the other, are very well-supported clades, but to a very large extent morphological characters proved unable satisfactorily to resolve the interrelation- ships between these orders. Only with the advent of sufficient molecular sequence data over the last decade did this resolution become possible, with the by now familiar but at the time utterly unexpected results (e.g. Kemp 2005; Springer et al. 2005). Thus the Therapsida are also of interest as a paradigm for how to interpret an evolutionary pattern of a taxon where morphological based cladistic analysis does not generate a well-supported set of sister-group relationships, but where molecular evidence is not available. CURRENT VIEWS OF THE INTERRELATIONSHIPS OF THERAPSIDS As reviewed most recently by Rubidge & Sidor (2001) and Kemp (2005), there are six widely recognized undis- puted therapsid subtaxa, viz: • Biarmosuchia: medium-sized carnivores retaining several sphenacodontid characters. No clear-cut shared derived characters so this may technically be a paraphyletic group, and certainly close in structure to the hypotheti- cal therapsid ancestor. • Dinocephalia: large to very large carnivores and herbi- vores but generally quite primitive. Characterized by a tendency to pachyostosis of the skull bones. Gorgonopsia: medium to large, highly specialized carni- vores, characterized by very large canines and a jaw hinge and musculature capable of an extremely wide gape. • Anomodontia: small to large, highly specialized herbi- vores, although including some quite primitive, basal forms. Apart from the latter, dentition largely or com- pletely replaced by a horny beak, and extreme enlarge- ment of the jaw muscles. • Therocephalia: small to large-sized diverse carnivores, with a few specialized omnivores, and in the Triassic ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 1 There is little agreement on the interrelationships of the major therapsid subtaxa because none of the variously proposed sister-group relationships are supported by clearly defined, unambiguously distributed morphological characters. Rather than pursue a new cladistic analysis here, the hypothesis is explored that the lack of an agreed cladogram is because there was a polytomy at the base of the therapsid radiation that is not amenable to positive testing by conventional morphological cladistics, but that can be tested in four ways. The virtually simultaneous appearance of all the lineages except Cynodontia in the Middle Permian stratigraphic record supports the hypothesis. The palaeogeographic record, which shows a combination of taxa with first occurrences in different parts of Pangaea also supports it, though this is not strong evidence. The palaeoenvironmental record supports the polytomous hypothesis strongly by providing evidence of a coincidence between the start of the therapsid radiation and the appearance of a new suite of ecological opportunities for diversification within higher latitudes. Finally, a functional correlation analysis of the characters associated with feeding, and the reconstruction of lineages of functionally integrated organisms offers strong support by indicating that no two of the four respective lineages, Dinocephalia, Gorgonopsia, Anomodontia and Therocephalia, could have shared a functionally feasible common ancestral stage subsequent to a hypothetical ancestor at a biarmosuchian grade. The exception is Cynodontia and Therocephalia, which are inferred to have shared such a more recent common ancestral stage, and therefore to be sister-groups in the taxon Eutheriodonta. Keywords: Therapsida, Permian tetrapods, Permian palaeoecology, correlated progression. one specialized herbivorous subgroup. • Cynodontia: initially small to medium-sized carnivores, although a very diverse group in the Triassic. The most progressive taxon with many characters shared with mammals such as complex teeth, enlarged dentary and secondary palate. The first of these subtaxa, Biarmosuchia, is recognized almost completely by a number of sphenacodontid-like plesiomorphic characters. One or two minor synapo- morphies have been claimed for it (Hopson 1991; Sidor & Rubidge 2006), but the taxon may well be paraphyletic and include the ancestry of the rest of the therapsids. The rest of these therapsid subtaxa are well-supported clades. There are also a few very poorly known forms from the Middle Permian of Russia, such as Nikkasaurus, Microurania, Phthinosuchus and Niaftasuchus (Ivakhnenko 2003; Kemp 2005). These are ignored here, although with better material they may well prove important. There is also a specimen that has been claimed to represent a very basal therapsid, Tetraceratops, which is represented by a single, poorly preserved specimen from the Lower Permian of North America. Laurin & Reisz (1996) described it as possessing certain therapsid features, but several authors have subsequently rejected this claim, and interpreted it as Romer & Price (1940) originally did as an aberrant pelycosaurian grade synapsid. The most familiar cladogram of the major therapsid subtaxa (Fig. 1A) is that published in their review by Rubidge & Sidor (2001) and based mainly on the analysis and data set of Sidor & Hopson (1998). Although it is expressed as a fully resolved set of dichotomies, the number of characters supporting the major nodes is not only small or very small, but also the significance of many of these characters can be doubted on the grounds of their being either not unique to the taxon, probably non- independent, too vaguely defined for confidence that they are homologous, or based on a very small sample of species within the taxon (Table 1). Meanwhile, other authors have proposed different relationships, although on no more convincing grounds. King (1988) argued for a sister group relationship between the Dinocephalia and Anomodontia (Fig. 1B), to which combined taxon she applied the name ‘Anomodontia’ in a former usage. Cladistic analyses by Gauthier et al. (1988) and by Modesto et al. (1999) independently concluded that Anomodontia, is the sister group of Therocephalia plus Cynodontia (Fig. 1C). Recently Sidor & Rubidge (2006) published a cladistic analysis of several new and newly studied biarmosuchian grade therapsids, along with dinocephalians and basal anomodontians, but not including gorgonop- 2 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 Figure 1. Current hypotheses of the interrelationships of major therapsid sub-taxa. See Table 1 for the characters used by respective authors to define the nodes. ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 3 Table 1. The putative synapomorphies for the proposed monophyletic groups in Fig. 1, extracted from the respective cited works, with comments on ambiguousness of distribution, impreciseness of the definition, or smallness of sample size as appropriate. EUTHERAPSIDA – Fig. 1A. (all except Biarmosuchia) Zygomatic arch bowed laterally. No more so in the brithopodid dinocephalians (Orlov 1958) than in biarmosuchians (Sigogneau-Russell 1989; Ivakhnenko 1999) and not bowed but simply diverging slightly in therocephalians. No distinct ossified olecranon process of the ulna. Prominent process in gorgonopsians (Colbert 1948) and present, if short, in cynodonts (Jenkins 1971). Only three phalanges in 5th pedal digit. Extremely small sample of taxa available. NEOTHERAPSIDA – Fig. 1A. (Anomodontia, Gorgonopsia, Therocephalia and Cynodontia) Ventrally expanded squamosal hiding most of quadrate in posterior view. Difficult to accept as homologous because of the very different form and mode of attachment of the quadrate in gorgonopsians (Kemp 1969a), therocephalians (Kemp 1972b; van den Heever 1994) and anomodontians (King 1981). Epipterygoid broadly contacting underside of parietal. Contact is narrow in gorgonopsians (Kemp 1969a), and anomodontians (King 1988). Ivakhnenko (2003, figs 6 and 9) figures an epipterygoid apparently very similar to that of gorgonopsians and anomodontians in both Biarmosuchus and the dinocephalian Archaeosyodon. Epiphyses on atlas vertebra. An extremely small sample; Sigogneau-Russell (1989) states that the atlas of the biarmosuchian Hipposaurus is similar to that of gorgonopsians. Enlarged obturator foramen of pelvis. An extremely small sample of biarmosuchians. In dicynodontians wide variation is reported, from minute to large (King 1988). THERIODONTIA – Fig. 1A. (Gorgonopsia, Therocephalia and Cynodontia) Flat, low snout with dorsal surface of nasals horizontal. Short internarial process. Relatively long in the basal cynodont Procynosuchus (Kemp 1979), and short in pelycosaurs (Romer & Price 1940). Narrow temporal roof, equal or less than interorbital width. Not the case in gorgonopsians generally (Sigogneau-Russell 1989). Greater flaring of zygomatic arch. Not the case in therocephalians. Dentary with free-standing coronoid process. The gorgonopsian coronoid process differs from those of therocephalians and cynodonts in its triangular rather than flat cross-sectional shape, indicating a different pattern of muscle attachments. Dentary with masseteric fossa. Not present in gorgonopsians or therocephalians. Postdentary bones somewhat reduced in height. Not absolutely in either gorgonopsians or therocephalians, but only relative to the increased height of the dentary due to the coronoid process. Humeral head slightly dorsal. Very vague and just as true of some dicynodontians (King 1988). Deltopectoral crest more than 40% of humeral length. Also in dicynodontians (King 1988). Greater trochanter of femur still small but extends distal to head. There is little manisfest difference in the greater trochanter of biarmosuchians and gorgonopsians (Sigogneau-Russell 1989). EUTHERIODONTA – Fig. 1A. (Therocephalia and Cynodontia) Narrow intertemporal roof. No contact between postorbital and squamosal on medial margin of temporal fossa. Presumably correlated with previous character. Sagittal crest on parietal. Also presumably correlated with the first character. Antero-posterior expansion of epipterygoid. True, but only to a slight extent in most therocephalians. Loss of palatal teeth. Also in anomodontians. Postero-ventral part of dentary thickened and angular in trough. Fenestra between dentary, surangular and angular. UNNAMED TAXON – Fig. 1C. (Anomodontia, Therocephalia and Cynodontia) Frontal margins with lappet entering orbital margin. Virtually identical in most other therapsids, such as brithopian dinocephalians (Orlov 1958) and gorgonopsians, but not true of even basal cynodontians. Postfrontal small. Still substantial in early therocephalians (van den Heever 1994) and the basal anomodontians Otsheria (Chudinov 1960) and Patranomodon (Rubidge & Hopson 1996). Postorbital region of skull longer than preorbital region. Not true of more basal therocephalians, and lengths about equal in basal cynodonts. Palatine with separate palatal and choanal rami. Difficult to distinguish from the gorgonopsian condition (Kemp 1969a; Sigogneau-Russell 1989) Mandibular fenestra present. Different construction in anomodontians (King 1988) compared to therocephalians (Kemp 1972b), for example in the former the dentary extends above and below it, but only above it in the latter. Palatine teeth absent. Teeth on transverse process of pterygoid absent. Tabular separated from opisthotic by squamosal. Not the case in basal cynodonts. Odontoid fused to axis in adult. Second intercentrum fused to axis in adult. Probably functionally correlated with the previous character. Atlas neural arch separated from atlas intercentrum. Doubtful as they function together as a ring (Kemp 1969b). Clavicles narrow medially. They broaden medially in at least the therocephalian Regisaurus (Kemp 1986; Fourie & Rubidge 2007) and the basal cynodontian Procynosuchus (Kemp 1980) Humeral head articular surface bulbous and inflected. No more so in some therocephalians (Kemp 1986; Fourie & Rubidge 2007) than in gorgonopsians (Kemp 1982). Ilium more than twice height of acetabulum. Not in therocephalians (Kemp 1986). Continued on p. 4 sians, therocephalians or cynodonts (Fig. 1D). Their strict consensus tree of nine equally parsimonious trees included a fourfold polytomy of, respectively, Dinocephalia, Anomodontia, the genus Biarmosuchus alone, and the other biarmosuchians. A majority rule consensus of the nine trees generated Dinocephalia as the most basal and Anomodontia and Biarmosuchia as sister groups. It is, of course, possible that a new cladistic analysis with more thoughtful selection of characters would produce a more strongly supported set of resolved interrelation- ships. However, in the absence of radical new material, and in the light of the universal problem of how objec- tively to recognize unit morphological characters, this is unlikely, and is not attempted here. Rather, it is the purpose of this paper to propose that the failure to discover a well-supported and fully resolved tree of therapsid interrelationships lies in the inability of cladistic methodology to deal with a situation where the real evolutionary pattern may have been a virtually simulta- neously polytomous splitting of several lineages from the ancestor. In such a case, alternative, non-cladistic tests for true polytomy need to be considered. TESTING FOR POLYTOMY It must be accepted as a matter of evolutionary biology that in principle a polytomous split of several lineages from a low taxonomic level can occur. At the extreme the multiple lineages would all arise directly from a single ancestral species, although more plausibly perhaps they would arise from different respective species sharing the typical morphological disparity of a single genus, or from different genera sharing the typical disparity of a taxo- nomic family. As has always been understood, such a situation creates difficulties in principle for cladistic analysis (e.g. Maddison 1989). If there are no characters defining successive nodes on the tree, cladistic methodol- ogy is required to attribute this to a failure to find them, not to their absence. Put another way, for cladistics the null hypothesis is unresolved polytomy and therefore polytomy cannot itself be a testable hypothesis, but only an expression of ignorance of enough characters (e.g. Maddison 1989; Walsh et al. 1999). Hence the response to such a situation is to continue the search for defining characters, and in the meanwhile accept provisionally the best supported tree, however weak and unconvincing the support is. Several molecular systematists (Walsh et al. 1999; Poe & Chubb 2004; Whitfield & Lockhart 2007) have addressed the problem of resolving the interrelationships amongst lineages of modern taxa that appear to have arisen by an ‘explosive’, polytomous radiation. The issue is that the external branches (major subtaxa) of the tree are much longer (i.e. greater molecular differences between them) than the internal branches (the inferred initial diversifica- tions at the base of the tree), to the extent that the molecu- lar evidence lacks the resolving power to distinguish between a succession of finely spaced dichotomies (a ‘soft polytomy’) on the one hand (Fig. 2B) and a single poly- tomous split (a ‘hard polytomy’) on the other (Fig. 2A). The molecular situation described is clearly analogous to the morphological problem of therapsid interrelationships, and the same terminology is appropriate: therapsids exhibit a soft polytomy in that the morphology can reveal the major subtaxa, but cannot satisfactorily resolve a complete set of dichotomous nodes between them, and therefore cannot discriminate between soft and hard polytomies. In the case of molecular sequence based systematics, there are a number of possible statistical tests of the likelihood that a soft polytomy actually is a hard or true polytomy, because of the ease of recognizing a very large number of objectively definable ‘unit’ characters (i.e. nucleotides). In the case of a soft polytomy that is based solely on the morphology of fossils, other positive tests for hard polytomy need to be considered, of which there are four kinds. The stratigraphy test A polytomy implies a virtually simultaneous origin of the separate lineages. If the stratigraphic sequence covering the time of the event has both adequate temporal resolu- tion, and a dense enough fossil record it can offer support for the polytomy hypothesis. The relative dating of the first appearance of the major therapsid subtaxa has been reviewed most recently by Lucas (2006), Kemp (2006a) and Abdala et al. (2008) (Fig. 3). They are currently known from Middle Permian deposits of three areas: the lower part of the Beaufort Group of South Africa, the cis-Uralian region of Russia, and a sparse fauna from the Xidagou Formation of Dashankou, China. There are disagreements about the dating of Middle Permian continental localities relative to the standard, marine-based sequence (Rubidge 1995; Izart et al. 2003; Lucas 2004; Tverdokhlebov et al. 2005), but from the nature of the fossil faunas, two aspects of the temporal 4 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 Obturator foramen between pubis and ilium, rather than in pubis alone. Condition unclear in gorgonopsians (Sigogneau-Russell 1989). Femoral head oblong and spherical (rather than elongate, subspherical, or protuberant). Very vague and hard to exclude gorgonopsians or even biarmosuchians (Sigogneau-Russell 1989). ANOMODONTIA sensu King (1988) – Fig. 1B. (Dinocephalia and Anomodontia s.s.) Loss of coronoid bones. Non-terminal nostrils and long posterior spur of premaxillae. Also in biarmosuchians (Sigogneau-Russell 1989; Ivakhnenko 1999) and the basal cynodontian Procynosuchus (Kemp 1979). Grooved or troughed palatal exposure of vomers. Difficult to see a significant difference between biarmosuchians and dinocephalians, and complicated by the evolution of a secondary palate in anomodontians. Reduction or loss of internal trochanter of femur. Table 1 (continued) pattern of occurrence of early therapsids are generally agreed upon. First, the three formations that have so far yielded early, Guadalupian-aged therapsids are closely spaced in time. Abdala et al. (2008) believe that the South African Eodicynodon Assemblage-Zone is around the Roadian-Wordian boundary, and that both the Russian Ocher Assemblage and the Chinese Xidagou Formation fauna are Roadian, and so slightly older. The actual therapsids currently known from this time include all the major subtaxa except one. The Cynodontia did not appear until some 6–7 million years later, in the Late Permian Tatarian of Russia and the approximately contemporane- ous Tropidostoma Assemblage Zone of South Africa (Botha-Brink & Abdala 2008). Second, it is believed that there is a significant hiatus between the last occurrence of the North American pelycosaurs and these earliest therapsids. It has been termed ‘Olson’s Gap’ by Lucas (2004), after E.C. Olson who first proposed its existence (Olson 1962), and is perhaps about five million years in length. There is also a general belief that a very considerable therapsid ghost lineage of about 35 million years must have existed, from the first appearance of therapsids in the Middle Permian right back to the first appearance of their presumed sister-group, the sphenacodontid pelyco- saurs in the Late Carboniferous (Abdala et al. 2008). How- ever, as Kemp (2006a) has pointed out, on the currently available evidence the possibility cannot be ruled out that Sphenacodontidae is paraphyletic, and that Therapsida is the sister-group of a much later, Early Permian member of that taxon. If this is so, then at the minimum the ghost lineage need be scarcely any longer than that of the the five million years of Olson’s Gap. The temporal pattern of first appearances of fossil therapsids is completely compatible with a simultaneous splitting of an ancestral therapsid into all the major subtaxa except Cynodontia in Roadian times. This support for the polytomy hypothesis must nevertheless be treated with some caution because of the existence of Olson’s Gap, and because of the impos- sibility on the basis of presently known characters of knowing the precise relationships of Therapsida to the sphenacodontian pelycosaurs, and therefore how long the therapsid ghost lineage really is. The palaeobiogeography test A second implication of polytomy is that all the lineages first appeared in the same geographical region. The fossil record of a radiation could in principle be compatible with such a pattern. Alternatively, it could suggest a pattern of separate regions of origin for different combinations of taxa, implying a succession of dichotomies. The earliest therapsid fossils occurring in the three regions of mid-Permian Pangaea do not show a clear taxo- nomic differentiation between the areas. Four of the lineages occur together in South Africa, Dinocephalia, Anomodontia, Gorgonopsia and Therocephalia (Rubidge 1995). In Russia there are three, Biarmosuchia, Dino- cephalia and Anomodontia (Ivakhnenko 2003), and in China only two, Biarmosuchia and Dinocephalia (Li & Cheng 1996; Li et al. 1997). Thus the palaeogeographic pattern tends to support a polytomy rather than pointing to possible vicariant or dispersal events separating succes- sive dichotomies. However, the very small number of localities, and the paucity of fossils within them renders the pattern too weak to be regarded as particularly significant. The palaeoenvironmental test A polytomous divergence of several lineages virtually simultaneously is a realistic possibility, and such an event can be presumed to have a potentially discoverable cause. Explanations of explosive radiations have long been sought, in terms of such concepts as key innovations (Hunter 1998) and ecological opportunities (Kemp 2007b). For example, it is suggested that the Cambrian explosion is related to increase in oxygen levels (e.g. Marshall 2006), and that the Tertiary radiation of placental mammals was facilitated by the removal of competitive exclusion by dinosaurs (e.g. Kemp 2005). A hypothesis of polytomy may in principle be supported by evidence of the sudden origin in the ancestral taxon of a novel biological potential, or evidence of the coincidental origin of a series of new ecological opportunities. ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 5 Figure 2. Possible patterns of evolution; thick lines represent the fossil record, thin lines the true phylogenetic lineages. A, A ‘hard’ or true polytomy. B, A ‘soft’ polytomy in which the internal branches are too short to be resolvable by morphological characters. C, A sequence of dichotomies with long ghost lineages. In the case of the Therapsida, Kemp (2006a) proposed an explicit palaeoenvironmental model for both the origin of basal therapsids, and for their divergence in the Middle Permian into a series of subtaxa (Fig. 4). To summarize briefly, the evidence that was adduced for the model is in part functional interpretation of therapsids as tetrapods with higher energy budgets, and greater internal thermoregulatory and chemoregulatory abilities (Kemp 2007b). This supposedly allowed them to remain continu- ously active in highly seasonal environments, notably the Summerwet Biome that occupied the tropical zones of Early Permian Pangaea (Rees et al. 2002). The second line of evidence is a shift in the palaeoclimatic zones at the start of the Middle Permian. Hitherto, extensive desert zones had isolated the tropical regions from the temperate regions in both hemispheres, but at this time the Summerwet Biome appears to have expanded northwards and south- wards along the eastern edge of Pangaea (Rees et al. 2002) . For the first time it became possible for therapsids to disperse from tropical into temperate regions. According to the model, this new ecological opportunity for a taxon already adapted for fluctuating, seasonal conditions resulted in the explosive radiation of the group. Rapidly and simultaneously several lineages diverged into a series of new niches – large and small body sizes, carnivores, herbivores and omnivores. Thus the hard polytomy hypothesis of the interrelation- ships of therapsids is corroborated by this evidence of a plausible environmental cause occurring coincidentally in time and place with the fossil record of the radiation. The functional systems test However plausible an evolutionary scenario can be built on the basis of stratigraphic, biogeographical and palaeoenvironmental evidence, any hypothesis of phylo- genetic relationships of organisms must be supported 6 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 Figure 3. Stratigraphic occurrences of synapsid taxa indicating the virtually simultaneous first appearance of the major therapsid subtaxa in the Middle Permian. The earliest Dinocephalia are indicated by Anteosauridae. P indicates pelycosaurian taxa surviving into the Middle Permian. The open box labelled ‘?earliest therapsid’ refers to some extremely poorly preserved fragments from the Copper Sandstones of Russia that have been dubiously claimed to be therapsid limb bones (see Kemp 2006a). The exact stratigraphic correlation between these three regions and also the Xidagou Formation of China, which is not shown, are not yet agreed (see Izart et al. 2003; Rubidge 2005; Lucas 2006; Abdala et al. 2008). Reproduced with slight modifications from Kemp (2006a). primarily by the analysis of characters. The impasse between the realistic possibility of polytomous branching having occurred on the one hand, and the logical inability of standard cladistic methodology to provide a positive test for it on the other has been pointed out. There is, how- ever, an alternative way of treating characters that in principle at least could detect a polytomy. Cladistic analysis of morphological characters assumes that the organism can be atomized into a set of objectively recognizable discrete, independent, and initially equally weighted characters. The most parsimonious distribution of these characters amongst the organisms being analysed is then taken to indicate the best estimate of relationships – the best supported tree. However, all these assumptions are to a degree unrealistic, because of course organisms are actually integrated wholes in which the parts are structurally and/or functionally integrated, and they act together to produce the biological attributes of the pheno- type (e.g. Dullemeijer 1974; Schwenk 2001). Where large amounts of character data give strongly supported relationships, there is no reason not to accept the most parsimonious tree as a good estimate. However, when there is poor support for any one tree, and little or no agreement amongst different authors concerning which is the best supported tree, then the probability that characters are being over-interpreted is high. An alternative to the atomistic model of cladistic methodology is the much more realistic correlated pro- gression model (Kemp 2006b, 2007a,b). Here it is assumed that characters are indeed functionally interdependent on one another within an integrated organism. They evolve in loose correlation with one another, and the probability of a change in one particular character depends on what changes have already occurred in others. Mean- while the coordinated changes amongst the characters al- ways maintain a fully integrated, well-functioning organism. Applying the correlated progression model as a test for polytomy requires an understanding of the functional interrelationships between the known characters of a derived phenotype. Once this is achieved, a more primi- tive hypothetical stage can be reconstructed, in which the individual parts are more plesiomorphic but between which the functional relationships are maintained. For example, there must always be correlation between the size and orientation of the inferred site of origin and site of insertion of a muscle; between the form of the dentition and the mandibular mobility permitted by the shape of the jaw articulation; between the inferred action of the forelimb and the action of the hindlimb; between body size and various allometrically related structures, and so on. By performing such an analysis several times, a hypo- thetical sequence of successively more plesiomorphic, ancestor-like stages can be reconstructed all the way back to the inferred ancestral state of the whole lineage. The third step is to compare this sequence of reconstructed historical stages for one lineage with those of other lineages, in order to see if there is a possible coincidence of structure and function at any point. Absence of such coalescence implies that the two lineages evolved independently from the common ancestor of the whole taxon, even if in a cladistic analysis there are certain isolated characters in common. And if several lineages lack coalescence with any others, then a hypothesis of polytomy is corroborated. The most comprehensively studied functional aspect of therapsids is the feeding mechanism, which involves numerous integrated structures. The architecture of the jaw musculature is related to the size and shape of the temporal fenestrae, the posterior palate, and such structural features as a coronoid eminence, a discrete coronoid process of the dentary, fossae on the angular and dentary bones, ridges on the reflected lamina of the angular, and the form of the retroarticular process of the articular. Also integrated with the jaw muscle forces and directions are dental features such as interdigitating incisors, large, reduced or absent canines, reduced or elaborated post- ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 7 Figure 4. Model of the origin of Therapsida. Phase 1: radiation of pelycosaur-grade synapsids in an Everwet biome existing in the equatorial regions of Lower Permian Pangaea. Phase 2: evolution of adaptations for remaining active throughout the seasonal fluctuations in the Tropical Summerwet biome, leading to basal therapsids. Phase 3: retraction of the mid-latitudal desert zones along the eastern margin of Pangaea opened new dispersal routes to the northerly and southerly temperate biomes, within which therapsids rapidly radiated into a variety of new taxa. Reproduced from Kemp (2006a), which should be consulted for details. canines, and inferred keratinous beaks. The jaw articulation may vary in form and degree of robustness for stress resis- tance, and in design for size of gape, and often propalinal movements of the mandible. The nature of the attachment of the quadrate to the squamosal is surprisingly variable amongst therapsids, which reflects the vectors of the stresses generated by the jaw muscles, and mobility in some groups. There are also various structural elements of the skull associated with stress resistance, such as the epipterygoid and parocciptal process. Much less is currently understood about the mechanics of the postcranial skeleton and locomotory system of therapsids, so although in principle this part of the pheno- type could equally well be incorporated into a correlated functional analysis, for the present purpose attention is restricted to the skull and jaws. The ancestral state (Fig. 5A) The Russian biarmosuchian Biarmosuchus (Chudinov 1960; Ivakhnenko 1999) combines therapsid characters with sphenacodontid characters, and has few if any significant autapomorphies (Hopson 1991; Rubidge & Sidor 2001; Sidor & Rubidge 2006). It is therefore a good model for an ancestral therapsid-stage from which the remaining major taxa evolved. The long, convex preorbital region, relatively small temporal fenestra, modest coronoid eminence of the mandible, and well-developed postcanine dentition are all comparable to features of the sphenacodontid pelycosaurs. The more striking of the many detailed differences from the sphenacodontid are the much larger upper canine tooth, a degree of dorso- ventral expansion of the temporal fenestra, enlargement of the reflected lamina of the angular, and an anterior rotation of the occiput. Kemp (2005) attempted a simple reconstruction of the jaw musculature, concluding that the adductor mandibuli consisted mainly of a single muscle mass originating on the medial and posterior edges of the temporal fenestra, and no doubt from an aponeurotic sheet of connective tissue across it (Barghusen 1976b). It inserted in primitive sphenacodontid fashion along the dorsal and medial parts of the coronoid region of the jaw, and had not invaded the lateral surface of the jaw at all. The incisor teeth may have interdigitated, although this is not certain. The jaw articulation was a simple, roller-like hinge between the quadrate and articu- lar bones. No attention has yet been paid to the mechanical structure of the skull, but it seems likely from the estimated muscle sizes that the stresses generated by the biting action were small compared to most of the more derived therapsids. Stress transmission between skull and mandible was probably adequately accommodated by the relatively firm attachment of the large quadrate to the squamosal, and by the generally robust intertemporal and occipital regions of the skull. Dinocephalia (Fig. 5B) No comprehensive review of the jaw musculature of a dinocephalian has yet been undertaken, although Barghusen (1976a) and Kemp (1982) reconstructed the likely general features. The skull of Titanophoneus is presumed to have a structure close to that of the ancestral dinocephalian (Orlov 1958). The temporal fenestra has enlarged by dorso-ventral expansion compared to the ancestral therapsid condition, but there is no significant lateral or posterior expansion. The main distinction is an extension of the area of origin of adductor musculature onto a broadened lateral-facing surface of the intertemporal region. In this latter respect, the temporal fenestra of dinocephalians is entirely unlike that of gorgonopsians and the less specialized of the anomodontians. The form of the mandible indicates clearly that the insertion of this part of the adductor musculature had remained on the dorsal and medial parts of the lower jaw, much as in the ancestral stage, with no expansion onto the lateral jaw surface (Barghusen 1976a; Kemp 1982). A reconstruction of the functional evolution of the ancestral dinocephalians from the ancestral therapsid condition consists only of the dorso-ventral expansion of the origin of the adductor muscles, which is entirely different from what occurred in any of the other main therapsid subtaxa. Indeed, none of the integrated jaw function features characteristic of any of the other taxa are found in dinocephalians. This absence of coales- cence of derived functional organization supports the hypothesis of an independent evolutionary lineage for Dinocephalia. Anomodontia (Fig. 5C) The majority of anomodontians are the dicynodontians, which are the most modified of all therapsids, especially in terms of feeding function (Crompton & Hotton 1967; King 1981, 1988). The temporal fenestra is vastly increased in size by expansion both antero-posteriorly and, to vary- ing extents, medially in different taxa. Furthermore, the posterior part of the skull has extended ventrally, which has the effect of lowering the position of the jaw articula- tion and so increasing the torque generated by the adductor musculature. It also creates another large area of origin for this musculature. Novel areas for insertion of the muscles on the lower jaw are also present. These are a broadened dorsal surface of the jaw, and shelves on the lateral surface of the dentary, often placed very far forwards (King 1988). The jaw articulation is modified to permit propalinic shifts, correlated with the evolution of a horny tooth beak. Of the dentition, at the most only upper, tusk like canines and a few very small postdentary teeth are present. The functional evolution of anomodontians is better understood than that of other taxa because of a number of more basal anomodonts known from the Middle Permian of South Africa and Russia (King et al. 1989; Modesto et al. 1999; Reisz & Sues 2000). The cranial anatomy of Patra- nomodon (Rubidge & Hopson 1996), for example, illus- trates a stage between the hypothetical ancestral therapsid and a basal dicynodontian such as Eodicynodon (Rubidge 1990). Patranomodon has retained a relatively long pre- orbital region, and its temporal fenestra is relatively small and little expanded medially. Incisor teeth are also retained. The lack of significant medial expansion of the temporal fenestra, or of any spread of the attachment of 8 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 adductor musculature to a laterally-facing area of the edge of the intertemporal roof, but the presence already of the depression of the jaw articulation all indicate that functionally the anomodontian lineage could not have shared a common ancestry with either the dinocephalian or the therocephalian lineages, in which none of these features occur. Indeed, the reconstructed evolutionary trajectory from the ancestral anomodontian back to the hypothetical ancestral therapsid stage shows no sign of coalescence with that of any other taxon. Gorgonopsia (Fig. 5D) Kemp (1969) undertook a detailed analysis of the highly specialized jaw function system of gorgonopsians. The temporal fenestra was expanded posteriorly and laterally, but scarcely at all medially, and this is correlated with new areas of insertion on the lower jaw. There is a discrete coronoid process, which is obtusely triangular in cross- section, and part of the adductor musculature gained an insertion on the external face of the lower part. The lateral part of the adductor musculature expanded its area of ori- gin onto the inside face of the broad, outwardly bowed zygomatic arch, and acquired a unique insertion onto a strong ridge occupying the reflected lamina of the angu- lar. The part of the adductor musculature attached to the still narrow medial edge and the undersurface of the intertemporal roof corresponds to the ancestral temporalis musculature, and it still inserted on the medial side of the lower jaw. This radical reorganization of the posterior and external parts of the adductor musculature is correlated with the ability of the lower jaw to open extremely widely, by more than 90°, and this was also reflected in a very specialized jaw articulation that permitted a wide gape while retain- ing a tight connection between the articular and quadrate (Parrington 1955). Finally, notwithstanding Laurin’s (1998) claim that this was not the case, it is clear from well-preserved and fully prepared specimens that the ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 9 Figure 5. Reconstruction of the main adductor mandibuli musculature of the six major therapsid subtaxa. Outlines of skulls based on (A) Biarmosuchus (Ivakhnenko 1999); B, anteosaurid dinocephalian Titanophoneus (Orlov 1958); C, dicynodontian anomodontian Eodicynodon (Rubidge 1990); D, gorgonopsian Arctognathus (Kemp 1969a); E, therocephalian Olivierosuchus (Brink 1965); F, cynodontian Procynosuchus (Kemp 1979). quadrate was moveably attached to the squamosal in such a way as to allow propalinal movement of the lower jaw (Kemp 1969a). The jaw could shift forwards to allow the upper and lower incisor teeth to interdigitate, and back- wards to allow unencumbered energetic use of the huge opposing canines. The various structures associated with gorgonopsian jaw action were integrated with one another within a functional system designed for highly active predation. Reconstruction of hypothetical evolutionary stages lead- ing to the fully expressed gorgonopsian arrangement is constrained by the requirement that no one of the individ- ual derived elements can be fully expressed in the absence of others, and in this case the correlations are especially apparent. Of particular relevance, the gorgonopsian coronoid process can only have evolved in correlation with a simultaneous posterior expansion of the temporal fenestra, since its function is to act as the insertion of that part of the adductor jaw musculature originating from the hind region of the fenestra (Kemp 1969a). This achieved an increased length of the musculature connecting them, which prevented undue restriction of the gape, and at the same time increased the torque applied to the lower jaw to increase the velocity of jaw-closing. A simultaneous expansion of the lateralmost part of the adductor mandi- buli, was also necessary for generating a force adequate to operate the large canines. Therocephalia (Fig. 5E) The temporal fenestra was enlarged in a manner entirely unlike that of any of the previous groups. Medial expansion led to a narrowing of the intertemporal roof, but there is little development of the zygomatic arch. Kemp (1972b) interpreted the posteriormost root of the zygomatic arch as the area of origin of the homologue of the cynodontian and mammalian masseter muscle, although at this stage it did not extend anteriorly along the arch. The coronoid process is constructed differently from that of gorgo- nopsians, and is associated with medially and postero- medially directed musculature attached to the inter- temporal region of the skull, rather than with posteriorly directed musculature as in gorgonopsians. The jaw articulation (Kemp 1972b) is quite different from that of other groups, not allowing propaliny, but instead resisting a large postero-medially directed reaction from the temporalis and incipient masseter muscles. The different arrangement of the adductor musculature between gorgonopsians and therocephalians implies independent modification from the hypothetical ancestral stage. The expansion of the temporal fenestra must have occurred independently in the two taxa because it is correlated with quite different parts of the adductor mandibuli muscle. Similar, the discrete coronoid process differs in form, and in the part of the musculature attached to it. The very different form of the jaw articulation reflects different reaction force regimes between the two. There- fore there is no sign of coalescence in the reconstructed morphological sequence between the hypothetical lineages leading to the therocephalians and the gorgonopsians, respectively. Cynodontia (Fig. 5F) Several authors have published reconstructions of the basal cynodontian jaw musculature (Barghusen 1968; Kemp 1979; Abdala & Damiani 2004). Most striking is the medial expansion of the temporal fenestrae creating a deep sagittal crest, and a simultaneous lateral expansion forming a bowed zygomatic arch. The latter was associated with a masseter muscle, which inserted into the lateral fossa of the broad coronoid process. According to Kemp (1979), in the basal cynodontian Procynosuchus this fossa was for the lateralmost part of the temporalis muscle and was only invaded by masseter muscle originating along the zygomatic arch in more derived forms. Abdala & Damiani (2004) differed in believing that the fossa was for the insertion of a true masseter muscle all along. However this does not greatly affect the reconstruction. The jaw articulation of the basal cynodontians is very similar to that of therocephalians, with the antero-laterally facing condyle of the quadrate designed to resist a postero- medially directed net reaction force applied by the articular. The epipterygoid of cynodontians is very broad, and together with the narrow intertemporal roof above and the basicranial axis below formed a box-girder, to strengthen the skull against the increased stresses arising from the enlarged adductor musculature (Kemp 1972a). Thus the arrangement of the adductor musculature of the medial and posterior parts of the temporal fenestra of the basal cynodont Procynosuchus is very similar to that of therocephalians. The main difference between the two taxa lies in the absence of a muscle-bearing zygomatic arch and correlated invasion of the lateral surface of the mandible by adductor musculature in therocephalians. However, the coronoid process of both taxa is associated with a comparable medial expansion of the temporal fenestra. Therefore a functionally integrated common stage for the two lineages can be reconstructed, which is essentially therocephalian in nature. The recent description of an early and even more basal cynodontian Charasso- gnathus (Botha et al. 2007) supports the hypothesis that the cynodontian and therocephalian lineages coalesce. Unlike Procynosuchus, it lacks an adductor fossa on the lateral surface of the coronoid process, but it does have a notch in the lower edge of the process that apparently represents an incipient invasion of the outer surface of the dentary from what would have been an essentially therocephalian-like arrangement of the musculature. The structure of the jaw articulation is very similar in therocephalians and Procynosuchus, and is correlated with the greater development of the medial and posterior parts of the adductor mandibuli. Even the expansion of the epipterygoid, to a modest degree in most therocephalians but more extensively in cynodonts and the therocephalian family Whaitsiidae, corroborates the essential similarity between the two of the integrated feeding system. Functional considerations therefore support a relation- ship between these two major therapsid subgroups. Conclusion Functional analysis of the skull and inferred mandibular musculature reveals that the integrated feeding system in 10 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 four of the five major derived therapsid subtaxa differed radically from one another. Each one evolved its own unique combination of characters from an hypothetical ancestral therapsid stage, approximately represented by the highly plesiomorphic Biarmosuchus. Furthermore, on reconstructing functionally integrated back-trajectories for the four, no two of them appear to have shared a common stage – to have coalesced – at any point subsequent to the common ancestor. This evidence corroborates the hypothesis that, at least at the level of morphological reso- lution available, there was a fourfold polytomy at the base of the therapsids, consisting of the dinocephalian, gorgonopsian, anomodontian and therocephalian lineages respectively. By contrast, the Cynodontia possessed an integrated system that had several features also occurring in similar combination in the Therocephalia, and in this case evolu- tionary back-trajectories of the functional morphology of these two lineages appear to have coalesced at a common point subsequent to the biarmosuchian-grade ancestral stage. This supports the hypothesis that Therocephalia and Cynodontia are sister groups, constituting a mono- phyletic taxon Eutheriodonta. These observations indicate that certain of the characters used in formal cladistic analyses to define various interre- lationships cannot be regarded as homologous. A notable example is the presence of a coronoid process of the dentary used to support a monophyletetic Theriodontia, consisting of Gorgonopsia, Therocephalia and Cynodontia. It was concluded on functional grounds that this stucture must be convergent in the gorgonopsians and thero- cephalians because it is associated with different parts of the adductor musculature in the two. Another example is the character used to support a relationship between Anomodontia and Eutheriodontia (Fig. 2C), the increased postorbital length of the skull. Again this can hardly be regarded as a homologous character because it is associ- ated with quite different ways of enlarging the temporal fenestra, and with different associated patterns of reorga- nization of the musculature, a conclusion confirmed by the discovery of the basal anomodontian Patranomodon. CONCLUSIONS The question posed in this essay is whether the weak support for, and extensive lack of agreement on a well- resolved phylogenetic tree of the major therapsid subtaxa is because of the failure yet to discover adequately known discriminating characters, or because there was a true polytomy in which several therapsid lineages diverged virtually simultaneously from a low-level ancestral therapsid taxon. The standard test for proposed phylo- genetic relationships is formal cladistic analysis, but meth- odologically this is designed only to discover the best supported, fully dichotomous tree, however weak that support may be. Logically cladistics cannot be used posi- tively to corroborate a hypothesis of therapsid polytomy. Four valid non-cladistic, non-tautological tests of polytomy are however available in principle, and when applied to the therapsid radiation three of them offer positive support for a fourfold polytomy from a biarmosuchian-like common ancestral taxon. The remaining one, the palaeo- biogeographic test, is consistent with polytomy. Of the three positive tests, it would be disingenuous to place too much weight on the stratigraphic relationships, but nevertheless this evidence offers very clear positive support for a fourfold polytomy. Only the Cynodontia appear in the fossil record significantly later and there- fore, on this evidence, this taxon is a candidate for sharing a common ancestor with one of the other lineages. The palaeoenvironmental evidence for polytomy is impressive, with the appearance in the fossil record of all the major lineages except Cynodontia coinciding with the appearance of a potential dispersal route to higher lati- tudes, north and south, areas of the globe that would suit well a group of tetrapods adapted to remain active throughout fluctuating seasonal conditions. The most important test of polytomy is the functional correlation analysis of the morphology. In so far as this method of analysing morphology owes it origin to a much more realistic model of character evolution than that underpinning cladistics, it is potentially a more effective method for discovering true phylogenetic patterns includ- ing, critically, the power to detect polytomy. In the case of the therapsids, the correlated progression analysis unam- biguously supports the hypothesis that four major lineages diverged independently from a biarmosuchian-grade ancestor. These are Dinocephalia, Gorgonopsia, Anomo- dontia and Therocephalia. Only one major subtaxon, Cynodontia, is inferred to have shared a common ancestor with another. Cynodontia and Therocephalia constitute the monophyletic taxon Eutheriodontia. All other proposed clades are rejected. Acknowledgements to Bruce Rubidge for reading and offering a most encouraging view of an earlier version of the manuscript. I am grateful to him and to Fernando Abdala and Bernard Battail for helpful comments that improved this paper, and to St John’s College, Oxford for financial support of my research. REFERENCES ABDALA, F. & DAMIANI, R. 2004. Early development of the mammalian superficial masseter muscle in cynodonts. Palaeontologia africana 40, 23–29. ABDALA, F., RUBIDGE, B.S. & HEEVER, J.V.D. 2008. The oldest therocephalians (Therapsida, Eutheriodontia) and the early diversifi- cation of the Therapsida. Palaeontology 51, 1011–1024. BARGHUSEN, H.R. 1968. The lower jaw of cynodonts (Reptilia, Therapsida) and the evolutionary origin of the mammal-like adductor jaw musculature. Postilla for 1968, 1–49. BARGHUSEN, H.R. 1976a. Notes on the adductor musculature of Venjukovia, a primitive anomodont therapsid from the Permian of the U.S.S.R. Annals of the South African Museum 69, 249–260. BARGHUSEN, H.R. 1976b. The adductor jaw musculature of Dimetrodon (Reptilia, Pelycosauria). Journal of Paleontology 47, 823–834. BOTHA, J., ABDALA, F. & SMITH, R.M.H. 2007. The oldest cynodont: new clues on the origin and early diversification of the Cynodontia. Zoological Journal of the Linnean Society 149, 477–492. BOTHA-BRINK, J. & ABDALA, F. 2008. A new cynodont record from the Tropidostoma Assemblage Zone of the Beaufort Group: implications for the early evolution of cynodonts in South Africa. Palaeontologia africana 43, 1–6. BRINK, A.S. 1965. A new ictidosuchid (Scaloposauria) from the Lystro- saurus-zone. Palaeontologia africana 9, 129–137. CHUDINOV, P.K. 1960. Upper Permian therapsids of the Ezhovo locality. Palaeontological Journal of the Palaeontological Institute of the Academy of Sciences, U.S.S.R. 4, 81–94. COLBERT, E.H. 1948. The mammal-like reptile Lycaenops. Bulletin of the American Museum of Natural History 89, 357–404. CROMPTON, A.W. & HOTTON, N.I. 1967. Functional morphology of ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 11 the masticatory apparatus of two dicynodontians (Reptilia, Therap- sida). Postilla1967, 1–51. DULLEMEIJER, P. 1974. Concepts and Approaches in Animal Morphology. Assen, the Netherlands, Van Gorcum. FOURIE, H. & RUBIDGE, B.S. 2007. The postcranial skeletal anatomy of the therocephalian Regisaurus (Therapsida: Regisauridae) and its utilization for biostratigraphic correlation. Palaeontologia africana 42, 1–16. GAUTHIER, J.A., KLUGE, J.A. & ROWE, T. 1988. Amniote phylogeny and the importance of fossils. Cladistics 4, 105–209. HOPSON, J.A. 1991. Systematics of the nonmammalian Synapsida and implications for patterns of evolution in synapsids. In: Schultze, H-P. & Trueb, L. (eds), Origins of the Higher Groups of Tetrapods. Ithaca, Cornell University Press. HUNTER, J.P. 1998. Key innovations and the ecology of macroevolution. Trends in Ecology and Evolution 13, 31–36. IVAKHNENKO, M.F. 1999. Biarmosuches from the Ocher faunal assemblage of eastern Europe. Paleontological Journal 33, 289–296. IVAKHNENKO, M.F. 2003. Eotherapsids from the east European placket (Late Permian). Paleontological Journal 37, S339–S465. IZART, A., STEPHENSON, R., VAI, B.G., VACHARD, D., LE NINDRE, Y., VASLET, D., FAUVEL, P-J., SÜSS, P., KOSSOVAYA, O., CHEN, Z., MASLO, A. & STOVBA, S. 2003. Sequence stratigraphy and correlation of late Carboniferous and Permian in the CIS, Europe, Tethyan area, North Africa, Arabia, China, Gondwanaland and the USA. Palaeoge- ography, Palaeoclimatography, Palaeoecology 196, 59–84. JENKINS, F.A. 1971. The postcranial anatomy of African cynodonts. Bul- letin of the Peabody Museum of Natural History 36, 1–216. KEMP, T.S. 1969a. On the functional morphology of the gorgonopsid skull. Philosophical Transactions of the Royal Society B 256, 1–83. KEMP, T.S. 1969b. The atlas–axis complex of the mammal-like reptiles. Journal of Zoology 159, 223–248. KEMP, T.S. 1972a. Whaitsiid Therocephalia and the origin of cynodonts. Philosophical Transactions of the Royal Society B 264, 1–54. KEMP, T.S. 1972b. The jaw articulation and musculature of the whaitsiid Therocephalia. In: Joysey, K.A. & Kemp, T.S. (eds), Studies in Vertebrate Evolution, 213–230. Edinburgh, Oliver and Boyd. KEMP, T.S. 1979. The primitive cynodont Procynosuchus: functional anatomy of the skull and relationships. Philosophical Transactions of the Royal Society B 285B, 73–122. KEMP, T.S. 1980. The primitive cynodont Procynosuchus: structure, function and evolution of the postcranial skeleton. Philosophical Transactions of the Royal Society B 288, 217–258. KEMP, T.S. 1982. Mammal-like Reptiles and the Origin of Mammals. London, Academic Press. KEMP, T.S. 1986. The skeleton of a baurioid therocephalian therapsid from the Lower Triassic (Lystrosaurus zone) of South Africa. Journal of Vertebrate Paleontology 6, 215–232. KEMP, T.S. 2005. The Origin and Evolution of Mammals. Oxford, Oxford University Press. KEMP, T.S. 2006a. The origin and evolution of the therapsid mammal-like reptiles: a palaeobiological hypothesis. Journal of Evolutionary Biology 19, 1231–1248. KEMP, T.S. 2006b. The origin of mammalian endothermy: a paradigm for the evolution of complex biological structure. Zoological Journal of the Linnean Society 147, 473–488. KEMP, T.S. 2007a. The concept of correlated progression as the basis of a model for the evolutionary origin of major new taxa. Proceedings of the Royal Society B 274, 1667–1673. KEMP, T.S. 2007b. The origin of higher taxa: macroevolutionary processes, and the case of the mammals. Acta Zoologica 88, 3–22. KING, G.M. 1981. The functional anatomy of a Permian dicynodont. Philosophical Transactions of the Royal Society B 291, 243–322. KING, G.M. 1988. Anomodontia. Stuttgart, Gustav Fischer Verlag. KING, G.M., OELOFSEN, B.W. & RUBIDGE, B.S. 1989. The evolution of the dicynodont feeding system. Zoological Journal of the Linnean Society 96, 185–211. LAURIN, M. & REISZ, R.R. 1996. The osteology and relationships of Tetraceratops insignis, the oldest known therapsid. Journal of Vertebrate Paleontology 16, 95–102. LAURIN, M. 1998. New data on the cranial anatomy of Lycaenops (Synapsida, Gorgonopsidae), and reflections on the possible presence of streptostyly in gorgonopsians. Journal of Vertebrate Paleontology 18, 765–776. LI, J & CHENG, Z. 1997. The first discovery of eotitanosuchian (Therapsida, Synapsida) in China. Vertebrata Palasiatica 35, 268–282. LI, J., RUBIDGE, B.S. & CHENG, Z. 1996. A primitive anteosaurian dinocephalian from China – implications for the distribution of the earliest therapsid faunas. South African Journal of Science 92, 252–253. LUCAS, S.G. 2004. A global hiatus in the Middle Permian tetrapod fossil record. Stratigraphy 1, 47–64. LUCAS, S.G. 2006. Global Permian tetrapod biostratigraphy and biochronology. In: Lucas, S.G., Cassinin, G. & Schneider, J.W. (eds), Non-marine Permian Biostratigraphy and Biochronology, 65–94. London, Geological Society of London. MADDISON, W.P. 1989. Reconstructing character evolution on poly- tomous cladograms. Cladistics 5, 365–377. MARSHALL, C.R. 2006. Explaining the Cambrian ‘explosion’ of animals. Annual Review of Earth and Planetary Science 34, 355–384. MODESTO, S.P., SIDOR, C.A. & RUBIDGE, B.S. 1999. The most basal anomodont therapsid and the primacy of Gondwana in the evolution of anomodonts. Proceedings of the Royal Society of London B266, 331–337. OLSON, E.C. 1962. Late Permian terrestrial vertebrates, U.S.A. and U.S.S.R. Transactions of the Americal Philosophical Society 52, 3–224. ORLOV, Y.A. 1958. Predatory dinocephalians from the Isheevo fauna (titanosuchians) Trudy Paleontological Institute of the Academy of Sciences, USSR 72, 1–114. PARRINGTON, F.R. 1955. On the cranial anatomy of some gorgonopsids and the synapsid middle ear. Proceedings of the Zoological Society of London 125, 1–40. POE, S. & CHUBB, A.L. 2004. Birds in a bush: five genes indicate explosive evolution of avian orders. Evolution 58, 404–415. REES, P.M., ZEIGLER, A.M., GIBBS, M.T., KUTZBACH, J.E., BEHLING, P.J. & ROWLEY, D.B. 2002. Permian phytographic patters and climate data/model comparisons. Journal of Geology 110, 1–31. REISZ, R.R. & SUES, H-D. 2000. Herbivory in late Paleozoic and Triassic terrestrial vertebrates. In: Sues, H-D. (ed.), Evolution of Herbivory in Ter- restrial Vertebrates, 9–41. Cambridge, Cambridge University Press. ROMER, A.S. & PRICE, L.W. 1940. Review of the Pelycosauria. Geological Society of America Special Papers 28, 1–538. RUBIDGE, B.S. 1990. Redescription of the cranial morphology of Eodicynodon oosthuizeni (Therapsida: Dicynodontia). Navorsinge van die Nasionale Museum Bloemfontein 7, 1–25. RUBIDGE, B.S. 1995. Biostratigraphy of the Eodicynodon Assemblage Zone. In: RUBIDGE, B.S. (ed.) Biostratigraphy of the Beaufort Group (Karoo Supergroup), 3–7. Pretoria, South African Committee for Stratigraphy, Biostratigraphic Series 1. RUBIDGE, B.S. & HOPSON, J.A. 1996. A primitive anomodont therapsid from the base of the Beaufort Group (Upper Permian) of South Africa. Zoological Journal of the Linnean Society 117, 115–139. RUBIDGE, B.S. & SIDOR, C.A. 2001. Evolutionary patterns among Permo-Triassic therapsids. Annual Review of Ecology and Systematics 32, 449–480. RUBIDGE, B.S. 2005 . Re-uniting lost continents – fossil reptiles from the ancient Karoo and their wanderlust . South African Journal of Geology 108, 135–172. SCHWENK, K. 2001. Functional units and their evolution. In: Wagner, G.P. (ed.), The Character Concept in Evolutionary Biology, 167–200. San Diego, Academic Press. SIDOR, C.A. & HOPSON, J.A. 1998. Ghost lineages and ‘mammalness’: assessing the temporal pattern of character acquisition in the Synapsida. Paleobiology 24, 254–273. SIDOR, C.A. & RUBIDGE, B.S. 2006. Herpetoskylax hopsoni, a new biarmosuchian (Therapsida: Biarmosuchia) from the Beaufort Group of South Africa. In: Carrano, M.T., Gaudin, T.J., Blob, R.W. & Wible, J.R. (eds), Amniote Paleobiology: Perspectives on the Evolution of Mammals, Birds, and Reptiles, 76–113. Chicago, University of Chicago Press. SIGOGNEAU-RUSSELL, D. 1989. Theriodontia 1: Phthinosuchia, Eotitanosuchia, Gorgonopsia. Stuttgart, Gustav Fischer Verlag. SPRINGER, M.S., MURPHY, W.J., EIZIRIK, E. & O’BRIEN, S.J. 2005. Molecular evidence for the major placental clades. In: Rose, K.D. & Archibald, J.D. (eds), The Rise of the Placental Mammals, 37–49. Baltimore, Johns Hopkins University Press. TVERDOKHLEBOV, V.P., TVERDOKHLEBOVA, G.I., MINIKH, A.V., SURKOV, M.V. & BENTON, M.J. 2005. Upper Permian vertebrates and their sedimentological context in the South Urals, Russia. Earth-Science Reviews 69, 27–77. VAN DEN HEEVER, J.A. 1994. The cranial anatomy of the early Therocephalia (Amniota: Therapsida). Annals of the University of Stellenbosch for 1994, 1–59. WALSH, H.E., KIDD, M.G., MOUM, T. & FRIESEN, V.L. 1999. Polytomies and the power of phylogenetic inference. Evolution 53, 932–937. WHITFIELD, J.B. & LOCKHART, P.J. 2007. Deciphering ancient rapid radiations. Trends in Ecology and Evolution 22, 258–265. 12 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 1–12 Assessing content and bias in South African Permo-Triassic Karoo tetrapod fossil collections Merrill Nicolas & Bruce S. Rubidge* Bernard Price Institute for Palaeontological Research, University of the Witwatersrand, Johannesburg, Private Bag 3, WITS, 2050 South Africa Received 12 October 2008. Accepted 25 June 2009 INTRODUCTION The rocks of the Beaufort Group (Adelaide and Tarkastad subgroups) of the Karoo Supergroup cover a large proportion of the surface of South Africa (Smith 1990) and comprise an approximately 3000 m thick sequence of predominantly sedimentary rocks which are internationally renowned for their wealth of tetrapod fossils. These fossil-bearing strata represent one of the most complete and best preserved palaeo-ecological records of pre-mammalian terrestrial vertebrates in the world (Keyser & Smith 1979) and are pivotal in evolution- ary studies because during this period the stem lineages to both mammals and dinosaurs arose (Broom 1932; SACS 1980). The absolute age of the Beaufort Group is not yet well constrained, with current dates based mainly on faunal correlations. The oldest stratigraphic units are considered Middle Permian (Kazanian) (Rubidge 1995a), and the uppermost strata as Middle Triassic (Anisian) (Ochev & Shishkin 1989; Hancox et al. 1995; Hancox & Rubidge 1996; Hancox 1998). Because of the extensive and largely unbroken temporal record of sediment deposition, coupled with the abundance of fossils, the succession is held by many to be the global biostratigraphic standard for the non-marine Permo-Triassic (e.g. Shishkin et al. 1995; Lucas 1998). For more than a century large collections of vertebrate fossils from the Beaufort Group of South Africa have been amassed (Fig. 1) and although these are now housed in institutions around the world, by far the largest and most representative collections are in museums in South Africa. These are: Albany Museum, Grahamstown; Bernard Price Institute for Palaeontological Research, University of the Witwatersrand, Johannesburg; Council for Geoscience, Tshwane; Iziko South African Museum, Cape Town; National Museum, Bloemfontein; Rubidge Collection, Wellwood, Graaff-Reinet; and the Northern Flagship Institution (Transvaal Museum), Tshwane. As part of a project to assess Permo-Triassic tetrapod biodiversity patterns and for use in biostratigraphy and basin modelling, the catalogues of these South African collections have been amalgamated onto a single stan- dardized dataset (Beaufort Group database) as well as a Geographical Information System (GIS) of vertebrate fossil data for the Beaufort Group (Nicolas 2007). This is the first time that such datasets have been compiled for mid-Permian–mid-Triassic continental vertebrate faunas and the resource (Nicolas 2007) serves as a research tool providing standardized taxonomic, stratigraphic and locality data for all specimens. Digital acquisition, integration and application of biological collections data is increasingly viewed as funda- mental to biodiversity research (Beaman et al. 2004). In setting up the database the quality of fossil data from con- tributing collections had to be evaluated and categorized to determine the degree of compliance with digital infor- mation requirements. A summary of general spatial data quality standards (as required for optimum delivery of GIS-based data) is provided in Nicolas (2007). This assess- ment determined and evaluated the percentage of records that could potentially be used for spatial map- ping. In addition records were analysed and categorized depending on their degree of taxonomic and locality reso- lution, a process which further refined the quality of the data eventually utilized in the spatial map of the Beaufort Group (Nicolas 2007). This evaluation was a necessary precursor for the establishment of the GIS fossil database. The above procedure has been adopted to ensure compli- ance with other international natural history collection database projects such as the Global Biodiversity Facility, the Biological Collection Access Service for Europe and the European Natural History Specimen Information Network. METHOD AND RESULTS Primary analysis of original data from contributing collections The original datasets provide 26 837 specimens that are theoretically usable for spatial mapping (Table 1). This figure is the sum of the original, unaltered records of each ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 13–20 13 *Author for correspondence. E-mail: bruce.rubidge@wits.ac.za A standardized taxonomic database as well as a Geographical Information System (GIS) database of all fossil tetrapods collected from the Permo-Triassic Beaufort Group rocks of South Africa has been compiled from a number of South African museum catalogues. The data capture required rigorous evaluation of the accuracy of the original records and the degree of collecting bias. The outcome of this evaluation endorsed the accuracy of the two databases and showed no significant degree of collecting bias. This standardized database, now linked to a new GIS-based database, will be a valuable resource to scientists researching Permo-Triassic biodiversity and faunal distribution patterns. Keywords: digitized data, fossil database, Beaufort Group. contributing museum. After careful investigation of the data from each museum catalogue, criteria for the elimi- nation of unreliable data were established (Table 2). These included the removal of data pertaining to fossils not from the Beaufort Group, specimens that are not vertebrate, and records that had no identification and locality data. After this initial round of data elimination, the actual number of potentially useful vertebrate fossils from the Beaufort Group amounted to 20 968 (Table 1). Primary analysis of the quality of Beaufort Group data All recorded specimens from the Beaufort Group were categorized according to whether they had: locality data; no locality data; biozone data. This was necessary to assess the accuracy of locality information and is an exercise in quantifying the degree of compliance with digital information requirements. For this analysis no distinction was made for different methods of recording locality data (e.g. GPS coordinates or Farm Name data). However, such distinctions in locality data resolution were utilized in the establishment of the GIS database (Nicolas 2007). For accuracy it was necessary to update outdated biozone data in accordance with Rubidge (1995b), the currently accepted biozonation scheme of the Beaufort Group. The updating methodology (detailed in Nicolas (2007)) did not reconstruct biozone data for specimens with locality data. The majority of locality data from the contributing museum databases had written descriptions rather than geographic coordinates of localities. The difficulty in using these textual references to locate the provenance of specimens is problematic (Nicolas 2007) so it was decided to utilize the biozone assigned to each specimen in order to interpret biodiversity trends. The rationale for this action is that the portion of records with biozone data 14 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 13–20 Figure 1. Relative numbers of fossils of taxa that have been recovered from the rocks of the Beaufort Group. Values in brackets reflect numbers of individuals. In the databases there are 13 taxon assignments for the Beaufort Group: Amphibia; Anomodontia; Archosauromorpha; Eosuchia, Biarmosuchia; Cynodontia; Diapsida; Dinocephalia; Fish; Gorgonopsia; Parareptilia; Pelycosauria and Therocephalia. Diapsida was subdivided into Archosauromorpha, Eosuchia and unidentified Diapsida for this study. The grouping of Diapsida is still retained to accommodate five errant specimens from the National Museum. In addition there is a final grouping of 13 fossils with unknown taxon assignment. Table 1. Useable data after initial elimination from original records. The original datasets from the seven contributing museum collections (Column 1) provide a theoretical potential of 26 837 (Column 2). The theoretical potential of 26 837 is the sum of the original, unaltered records of each contribut- ing museum. Unreliable records were eliminated (see Table 2). After data elimination, the actual potential of exploitable vertebrate specimens in the Beaufort Group amounted to 20968 (Column 3). This table shows the number of records per museum collection remaining after elimination. These values are expressed as a percentage (column 4) of the records which can be utilized when compared to the original records. Database Total records Record (after initial elimination) % Potential viable data Albany Museum 588 468 79.59 Bernard Price Institute 4780 4483 93.79 Council for Geoscience 7579 5322 70.22 National Museum 3520 3171 90.09 Rubidge Collection 854 850 99.53 Iziko South African Museum 6797 5424 79.80 Transvaal Museum 2719 1250 45.97 Theoretical total Potentially usable total % Potential viable total 26837 20968 78.13 would be capable of presenting broad biodiversity trends across and within the assemblage zones of the Beaufort Group. On completion of the GIS-initiative, all records will have been geo-referenced and so a fuller and more refined picture will result (Nicolas 2007). Records with no locality and no biozone information cannot be used for either spatial mapping or biodiversity analysis. Those records with both biozone and locality data (with varying degrees of resolution) and those with only locality data may be used for spatial mapping. In so far as digital information is concerned, the optimum would be a 100% recording of fossil finds with locality information refined to thousandths of a degree, WGS datum coordinates. As it currently stands, 51% of the records from the Beaufort Group have locality informa- tion, but no biozone information; 44% have locality and biozone information; and 5% have neither locality nor biozone information. Of the 9144 Beaufort Group specimens with biozone information, 57% (5193) are identified to genus level and the remaining 43% (3951) are unidentified (Fig. 2). The 9144 records with biozone information represent 44% of the total number of records for the Beaufort Group. This means that 25% of all records from the Beaufort Group are classified to genus level and have biozone information and 19% of records are unidentified, but have biozone information. The unidentified specimens with biozone allocation (3951) (Fig. 2) were subdivided into their relative alliance to the eight biozones of the Beaufort Group. The Eodicynodon Assemblage Zone has 31 unidentified specimens, the Tapinocephalus Assemblage Zone has 140 unidentified specimens (1% of the total Beaufort Group population), the Pristerognathus Assemblage Zone lists five unidentified specimens, the Tropoidostoma Assemblage Zone lists 549 unidentified specimens (3% of the total Beaufort Group population), the Cistecephalus Assemblage Zone lists 2109 unidentified specimens (10% of the total Beaufort Group population), making it the biozone with the greatest amount of unidentified specimens, the Dicynodon Assem- blage Zone lists 368 unidentified specimens (2% of the total Beaufort Group population), the Lystrosaurus Assem- blage Zone lists 172 unidentified specimens (1% of the total Beaufort Group population) and the Cynognathus Assemblage Zone lists 259 unidentified specimens (1% of the total Beaufort Group population). The remaining 318 unidentified specimens that have biozone information are relegated to various transitional zones (Nicolas 2007). Only 71% of all the 20 968 fossil vertebrates collected from the Beaufort Group have been identified to family, and of these only 5193 (25%) are identified to genus level and have biozone data. The Eodicynodon Assemblage Zone lists 51 specimens identified to genus level. The Tapinocephalus Assemblage Zone lists 183 specimens belonging to various genera (this comprises 1% of the total Beaufort Group vertebrate fossils), the Pristerognathus Assemblage Zone lists 34 specimens belonging to various genera, the Tropidostoma Assemblage Zone lists 361 speci- mens assigned to various genera (2% of the total Beaufort Group vertebrate fossils), the Cistecephalus Assemblage Zone lists 1037 specimens classified to various genera (5% of the total Beaufort Group vertebrate fossils), the Dicynodon Assemblage Zone lists 268 specimens belong- ing to various genera (1% of the total Beaufort Group vertebrate fossils), the Lystrosaurus Assemblage Zone lists 2559 specimens classified to various genera (12% of the total Beaufort Group vertebrate fossils), the highest count of identified specimens in the Beaufort Group, and the Cynognathus Assemblage Zone lists 562 specimens belong- ing to various genera (3% of the total Beaufort Group vertebrate fossils). The remaining 143 specimens identi- fied to genus level are catalogued as being from various transition zones. The low numbers of specimens listed above are minimum values as only those records with assemblage zone data were utilized. There are 9023 speci- mens identified to genus level without a biozone allo- cated, but they do have locality data (Fig. 2). These will be allocated a biozone once the GIS database is completed. Comparative analysis of contributing museum collections A comparative analysis of different museum collections was performed by focusing on the specimen totals in each ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 13–20 15 Figure 2. Analysis of Beaufort Group data, showing the subdivision of records into those identified to genus level or those as yet unidentified. 16 ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 13–20 Table 2. Elimination of non-viable data. After investigation of the content of each museum collection, criteria for the elimination of certain records were established. This table shows those records per museum collection requiring elimination. Bold font in column 2 refers to a specific datafield within a collection. Collection name Data Eliminated (Spreadsheet Rows) Rubidge Collection Locality: Unlisted farm locality & district information; Keetmanshoop; Lady Frere. Taxon: Mesosaurus; Sysphincto- stoma BPI Collection Biozone: Equivalent of L. Cistecephalus Z. Geology: ?Stormberg; Equivalent of Beaufort. District/Country: Ficksburg (FS); Namaqualand (NC); Fouriesberg (FS); Clocolan (FS); Ceres (WC); Worcester (WC); Calvinia (NC); Prieska (NC); Marquard; Ladybrand; Ladygrey; Zambia. Genus: Tritylodon; Amniota; Dinosauria; Plants; Taxon Indet.; Squamata? ?Diapsida; Reptilia; Indet.; Tetrapoda; Trace Fossil; Worm Burrows; Indet. Amniote; Indet. Dinosaur; Indet. Reptile; Indet. Tetrapod Burrow Cast; Unidentified Bone in Lag. Locality: Unlisted Locality Information; Locality Unknown; Various Localities. National Museum Collection Taxon: Unidentified Taxon & Current Identification; Aetonyx; Saurischia; Dinosauria; Anchisaurus; Baroqueosuchus; ?; Basutodon (Large Thecodont); Bothriolepis; Invertebrata; Conchostracans; Elpistostege; Euskelosaurus; Roccosaurus; Fabrosaurus; Footprints (Bradysaurus?); Unlisted Information; Gryponyx; Herrerasaurus; Holopthychius; Hout; Foot- print (Ichnitis); Massospondylus; Melanosaurus; Mesosaurus; Orthosuchus; Mollusca (Palecypoda); Pedeticosaurus; Plant; Rauisuchid; Reptile; Riojasaurus (Prosaurosushia); Saurischia; Scaumenacia; Sysphinctostoma; Thecodontosaurus; Tritylodon; Unidentified; Unknown; Various; Worms; Dinosauria; Invertebrata; Lacertilia; Ornithischia. Country/Locality: Unlisted district/farm; Unlisted province; Quebec (Canada); Barkley East; Bethlehem; Clarens; Golden Gate; Clocolan; Elliot; Essex (England); Ficksburg; Fouriesburg; Gumtree; Herschel; Hopetown; James- town; Lady Frere; Ladybrand; Leribe; Marquard; Mafeteng; Maseru; Rosendal; Slabberts; Zastron; Wes Duitsland (Holzmaden). Geology: Non-applicable Biozone; Elliot; Clarens; Molteno; Ecca; Stormberg; Waterford; Whitehill. Zone: Euskelosaurus; Massospondylus; Trityledon Acme Zone Iziko South African Museum Collection Geology/Age/Biozone: Ecca; Stormberg; Dwyka; Witteberg; Carboniferous; Bokkeveld; Devonian; Late Jurassic; Early Cretaceous; Cretaceous; Eocene; Middle Pennsylvanian; Upper Dwyka; Sakemena; Adolphspoort; Blue Lias; Carbon- dale; Elliot; Great Oolite; Green River Shale; Greensand; Irati; Kirkwood; Kupferschiefer; Lower Chalk; Lower Elliot; Lower Greensand; Lower Sakemena; Rio Bonito; Serra Alba; Whitehill; Early Triassic; Earliest Triassic; Early Carbonifer- ous; Early Devonian; Early Jurassic; Early Permian; Jurassic; Kimmeridgian; Late Cretaceous; Late Jurassic/Early Creta- ceous; Late Triassic; Lower Devonian; Lower Permian; Middle Devonian; Oligocene; Euskelosaurus; Reptile Beds. Country: Scotland; USA; Tanzania; Madagascar; Zambia; Namaqualand; Angola; Australia; Brazil; England; France; Germany; Great Britain; Ireland; Lebanon; Lesotho; Malawi; Mozambique; Namibia; North America; Wales; Zimbabwe; Unlisted locality (district/farm). Province: Caithness; Cambridge; Dumfries; Illinois; Mocamedes; New South Wales; Parana. District: Barkley East; Boputhatswana; Calvinia; Britstown; Ceres; Clocolan; Carnarvon; Dorset; Elliot; Fouriesburg; Hay; Herbert; Herschel; Hopetown; Kirkwood; Jansenville; Lady Grey; Ladybrand; Luangwa; Maclear; Mafeteng; Messina; Mokerong; Not Known; Outshoorn; Port Elizabeth; Port Shepstone; Prieska; Swellendam; Ranohira; Quacha’s Nek; Quthing; Sebungwe; Uitenhage; Underberg; Utrecht; Will County; Williston; Wodehouse; Worcester. Genus: Beetle; Mammal; Osteolepis; Acanthodes; Acanthodian; Acanthopterygid; Acrodus; Adroichthys; Aestuarichthys; Aetonyx; Astrodon; Camarasauridae; Opisthias; Pleurocoelus; Algoasaurus; Alopias; Anaethalion; Anchisaurid; Anoxypristis; Anura; Arthrodire; Asteracanthus. Class/Subclass: Aves; Insect; Insecta; Mammalia; Placodermi; Vertebrata incertae sedis; Holocephali; Lepospondyli; Neornithes; Prototheria; Testundinata; Theria; Insectifora; Antiarchi; Anura; Arthrodira; Batoidea; Chelonia; Crocodylia; Cyprinodontiformes; Ellimmichthyformes; Elopiformes; Gadiformes; Galeomorpha; Ichthyopterygia; Lepisosteiformes; Mesosauria; Ornithischia; Sauropodamorpha; Lebias; Diplomystus; Elopoidei; Dastilbe; Anaethalion; Palaeomolva; Lamna; Carcharodon; Squalicorax; Isurus; Alopias; Heterodontus; Campylodon; Communis; Leptolepiformes; Pachycormiformes; Paleospondyliformes; Pelecaniformes; Perciformes; Salmoniformes; Saurischia; Triconodonta Council for Geo- science Collection Genus: Indet.(No identification listed); Anura; Anomodontia; Lower Jaw; Nodules; Padda; Stromatolite; Therapsida; Theriodont; Vertebrata; Indet. Anomodontida. Locality: Unlisted locality information Albany Museum Collection Plants, Unlisted Acc. No., Geology: Buntsandstein, Lower Triassic; Elgin Sandstone; Gravel Banks; Cave Sandstone; Red Beds; Elliot Formation; Dwyka; Kirkwood Formation; Uitenhage Group (Kirkwood Formation); Upper Dwyka Shales; Shale. Country and Locality: England (Shale); Lyas (Switzerland). District: Barkley East; Scotland; Farm Rietfontein; Barkley Pass; Lyme Regis, England; Bolotive; Bunter Landstein, Switzerland; Farm Glencoe (Barkley East); Ibid.; Kirkwood Cliffs, Sunday’s River; Lady Frere, Glen Gray; Mafeteng District, Lesotho; Moirosi’s Mount, Lesotho; N. Luangwa Valley, N. Malawi; Near Alice; Penhoek, Stormberg; Riechen, Switzerland; Skietnek, Kirkwood Village; Thaba-Chau (Tsueu, Chu, Cho); Unknown. Taxon: Unknown; Small Reptile; Unidentified; Massopondylus; Chelonia; Unidentified; Sauropod Dinosaur; Dinosaur Prosauropod; Anomodont; Fossil Wood – Podocarpus; Dinosaur-like Styracosaurus; Hortalotarsus; Large Prosauropod? Euskelosaurus; Mesosaurus; Saurischia – Prosauropod Dinosaur; Saurischia (perhaps Plateosaurus); Saurischia (Dinosaur); Saurischia (Euskelosaurus); Small Reptile; Thecodontia – Pseudosuchian; Therapsid; Therapsid – Unidentified; Theropod – Nqebasaurus; Unidentified; Unidentified Small Reptile; Unidentified Small Dinosaur; Unknown; Unlisted Taxon Information Transvaal Museum Collection Genus: Unidentified; Undescribed Anomodont; Bivalve Shells; Boks 140; Aristosaurus; Eozostrodon; Fossil Bone with Matrix; Fossil Reptile; Gangomopteris (a plant); Gigantoscelus; Large Anomodont; Large Mammal; Massopondylus; Reptile; Small Anomodont; Therapsid?; Theriodontia?; Theropoda; Unident.; Unlisted Genera. Biozone: Cave Sandstone; Red Beds; Rhaetic (Late Triassic). Locality: Tunnels; ‘Tunnels’ Donga; Senekal; Jame’s Donga; Provenance Unknown; St. Fort (Setsoanastad) Bethelehem District; Witjies Hoek; Milius Donga; Unlisted locality information; Letjiesbosch; Klipbank; Karoo Site; Arcadia Donga; Commander Jone’s House; Apparently East of Fossil Valley; ‘Diamond Diggings’. Province: Unlisted locality information. Country: Unlisted locality information; England collection and the diversity of genera in each collection. This exercise identified the collections which curate the largest sample of any particular taxon and will thus facilitate future research. In addition it has highlighted problems in the fossil record and where future collecting should be undertaken. We have found that the content of the different collections is dependent on: geographic position of the museum; research focus of past and present palaeontologists employed by the museum; and diligence of the palaeontologists to pursue fieldwork. Because of these variables the numbers of specimens and taxonomic diversity of the individual collections varies greatly. Specimen totals in each collection The total number of specimens within each of the collec- tions is recorded in Table 3. This total includes both unidentified specimens as well as specimens identified to genus level. Any record that had no taxonomic classifica- tion (e.g. ‘Unknown’ or ‘Unidentified’ or ‘Indet.’) or the taxonomic classification was too broad (e.g. ‘Parareptile’ or ‘Synapsid’) was relegated to the category ‘Unidentified Specimens’. Because of the very large number of records spread between the seven contributing collections, it was considered beyond the scope of the project to verify each entry from each museum for correct identification. How- ever, because of the accessibility of the BPI Palaeontology, the classification in the amphibian and anomodont categories were updated, verified and recorded in the synthesized records of the Bernard Price Institute for Palaeontological Research (Nicolas 2007). The remaining taxa were updated in so far as current taxonomy allowed. Flaws in the contributing datasets became apparent after completion of the standardization and ‘clean-up’ proce- dures (Nicolas 2007: appendix F). The contents of this Appendix F will be used as part of the data-cleaning component in Phase 1 of the GIS-initiative (Nicolas 2007) and will fast-track the taxonomic updating of all the museum collection records. Diversity of genera in each collection Iziko South African Museum has the greatest diversity of genera (163), followed by the Bernard Price Institute (124), the National Museum (88), the Rubidge Collection (85), the Transvaal Museum (73), the Council for Geo- science (63) and the Albany Museum (56) (Table 4) Extent of collecting bias of fossils from the Beaufort Group Once the amalgamated Beaufort Group dataset and the foundation GIS database of fossils from the Beaufort Group were established, it was essential to determine the ‘validity’ of the collection for biodiversity analyses by calculating the degree of collecting bias (this bias being the over-representation of fossils from a particular locality, biozone or of a specific taxon). The foundation GIS data- base was utilized only to determine the extent of collect- ing bias within the Beaufort Group (Karoo Supergroup) of South Africa caused by the specialist interest of a field collector/palaeontologist currently or in the past and is currently undergoing further refinement, with the end result that it will be an invaluable research tool. The methods and processes involved in setting up the founda- tion GIS-database are explained in Nicolas (2007). Using geo-spatial data in the form of Neighbourhood and Gap Analysis to determine collecting bias To test the reliability of the Beaufort Group data for analytical purposes it was necessary to determine ‘even- ness’ in the geographic distribution of fossil localities in the Beaufort Group. If the distribution of fossil localities were found to be highly localized or biased to specific regions, this would impact negatively on the accuracy and significance of any current or future distribution or biodiversity study. Accordingly a ‘Gap and Neighbourhood’ Analysis was undertaken to determine if there was any geographic bias in the museum records. Arcview 3.2 program with its spatial analysis extension were used to create a distance surface. The distance was calculated between the locations of all the specimen points. A contrast stretch was applied to the resultant surface to improve the visualization of the surface ISSN 0078-8554 Palaeont. afr. (December 2009) 44: 13–20 17 Table 3. Total numbers of specimens per collection. Museum collection Total number of specimens Unidentified specimens Total number of identified specimens Albany Museum 468 44 424 Bernard Price Institute 4 483 2 470 2 013 Council for Geoscience 5 322 58 5 264 National Museum 3 171 987 2 184 Rubidge Collection 850 509 341 Iziko South African Museum 5 424 1 629 3 795 Transvaal Museum 1 250 448 802 Totals 20 968 6 145 14 823 Table 4. Diversity of genera within each collection. This table lists the diversity of genera within each collection, in order of decreasing diversity. Museum collection Diversity of genera Iziko South African Museum 163 Bernard Price Institute 124 National Museum 88 Rubidge Collection 85 Transvaal Museum 73 Council for Geoscience 63 Albany Museum 56 (Cooper & Netterberg 2004). The surface was derived for an area larger than the extent of the Beaufort Group, and then masked to the extent of the sampling to remove edge effects. Figure 3 indicates the sampling distribution of fossils in the Beaufort Group. Several factors could account for a lack of sampling in certain areas. These include absence of outcrop; lack of prospecting; or paucity of fossils. Blue shading indicates a substantial distance between fossil collecting sites, while a red–orange spatial pattern indi- cates a relative closeness of fossil sites to each other. The overwhelming dominance of orange–red (Fig. 3) indicates that the majority of fossil localities are in close proximity to each other with an even distribution and points to a lack of sampling bias. This implies that the sampling of the Karoo fossil fauna as represented by the fossils in the collections will provide an accurate interpretation of the reality of faunal distribution patterns and ecological representation for the Beaufort Group. Once the equality of the distribution of fossil finds across the Beaufort Group was established, Neighbourhood Analysis Technique was applied to determine where the highest density of specimens was located (Fig. 4). The Neighbourhood Statistic used was the sum of all the points, with a circle neighbourhood within an 8-km radius (Cooper & Netterberg 2004). This was used to provide an indication of the number of specimens from a given locality. The resultant surface was ‘smoothed out’ to improve visual representation of the analysis (Cooper & Netterberg 2004). The incidence of dark red shading represents those localities which have the highest density of specimens (Fig. 4). The obvious horizontal line of red/orange dots in Figure 4 highlights that it is most certainly a collecting bias caused by the superior vertical exposures of fossil bearing strata in the Sneeuberg and Neuweveld mountain ranges along the Great Escarpment. The strata most exposed in the escarpment belong to the Tropidostoma, Cistecephalus and Dicynodon biozones. Figures 3 & 4 clearly show the geographic collecting bias along the Great Escarpment. This means that these biozones may have been over- collected compared to other biozones. As indicated by the yellow shading in Fig. 4 most fossil sites yield an average of 1–60 specimens. Localized bias on the yield of fossils at specific sites is the result of differing resolutions of locality information (i.e. GPS, Farm Name, and District) and well as the result of superior exposures. Localized biases become inconsequential when viewed across the scale of the entire Beaufort Group because the finer details of locality data specifics are blurred in the light of the larger low resolution picture of the Beaufort Group. We are satisfied that the concentratio