The Postcranial Skeleton of Temnospondyls (Tetrapoda: Temnospondyli) Phd Thesis

Pawley, K. 2006. 149 The Postcranial Skeleton of Temnospondyls (Tetrapoda: Temnospondyli) PhD Thesis. La Trobe University, Melbourne

CHAPTER 5. PHYLOGENETIC VARIATION IN THE POSTCRANIAL SKELETON OF TEMNOSPONDYLS (TETRAPODA: TEMNOSPONDYLI)

Abstract. Phylogenetically significant features of the postcranial skeleton within the Temnospondyli are described and illustrated. Considerable morphological variation is observable in vertebral structure and the configuration of the dermal pectoral girdle; otherwise, the postcranial skeleton is generally conservative. Although most temnospondyls are best described as amphibious, the most basal temnospondyls are well adapted for terrestrial locomotion, with one major derived clade, the Euskelia, displaying postcranial adaptations to the axial skeleton that may have improved the efficiency of their terrestrial locomotion. Two derived clades, the Dvinosauria and Superstereospondyli, were obligatorily aquatic. In general appearance and lifestyle, most temnospondyls superficially resemble extant crocodilians. The information provided by this overview of the postcranial skeleton is used to revise the postcranial characteristics used for classifying temnospondyls, and to create new characters for a comprehensive phylogenetic analysis of the Temnospondyli, using 57 taxa and 154 phylogenetic characters. This analysis is innovative for phylogenetic analysis of temnospondyls in that morphogenetic and phylogenetic variation in the postcranial skeleton is carefully distinguished and accounted for in the coding of the data matrix. The most basal taxa within the Temnospondyli are Caerorhachis bairdi and Edops craigi, with the Dendrerpetontidae, Cochleosauridae, Dvinosauria, and Eryopoidomorpha (Euskelia plus Stereospondylomorpha) as successively more derived clades. Comparison with the only previous computer-based analysis of the Temnospondyli indicates that the different tree topology presented here is significantly more parsimonious. Test analyses indicate that phylogenetic analyses based entirely on cranial characters produce quite different tree topologies to those including postcranial characters.

INTRODUCTION

The Temnospondyli (Zittel, 1888), named for a characteristic of their postcranial skeleton, the multipartite vertebral centra, are a large and diverse group of early tetrapods. Their known stratigraphic range spans more than 200 million years, starting in the Early Carboniferous with the basal temnospondyls Caerorhachis bairdi (Holmes and Carroll, 1977), and Balanerpeton woodi (Milner and Sequeira, 1994) from the Viséan (345-326 mya) of Scotland, and Dendrerpeton acadianum from the Namurian (313-314 mya) of Nova Scotia (Falcon-Lang et al., 2006). The latest known temnospondyl is the derived stereospondyl Koolasuchus cleelandi from the Early Cretaceous (110 mya) of Australia (Warren et al., 1997). Discovery of temnospondyl fossils on every continent, including Antarctica (e.g. Cosgriff and Hammer, 1984), emphasises their success in adapting to diverse environments. Given the extensive time range and diversification of temnospondyls, a corresponding range of phylogenetic variation in the postcranial skeleton is to be expected, thus providing a useful source of information for taxonomic purposes. 150 K. PAWLEY PHD THESIS

Previous studies of the temnospondyl postcranial skeleton

Although no previous work has attempted to assess the range of phylogenetic variation within the whole of the Temnospondyli, several partial overviews are available. Romer (1947) provided a basic outline of morphological variation of the postcranial skeleton within temnospondyls, Nilsson (1939) characterised the cleithrum and humerus, Warren and Snell (1991) provided a comprehensive overview of the range of variation in Mesozoic temnospondyls, and Shishkin (2000) described variation in cervical vertebrae. A thorough assessment of the postcranial skeleton of temnospondyls is necessary to enable comparative research. As part of long-term research on the postcranial skeleton of temnospondyls, the aim of this study is to characterise the temnospondyl postcranial skeleton, and define the extent of phylogenetic variation of the postcranial skeleton within the Temnospondyli.

Previous phylogenetic analyses

The only previous large scale, computer based analysis of temnospondyl relationships is that of Yates and Warren (2000), which also summarised previous hand generated phylogenetic analyses (Coldiron, 1978; Warren and Black, 1985; Boy, 1990; Foreman, 1990; Milner, 1990, 1991; Boy, 1993; Daly, 1994), and smaller computer based analyses (Trueb and Cloutier, 1991; Damiani and Warren, 1996; Holmes et al., 1998; Laurin and Soler-Gijón, 2006). The analysis of Yates and Warren (2000) included the ‘higher’ temnospondyls but excluded plesiomorphic temnospondyls. An analysis of early tetrapod relationships which did include several plesiomorphic temnospondyls (Ruta et al., 2003) produced a completely different tree topology to that of Yates and Warren (2000), but this analysis was still constrained because it included only a small number of plesiomorphic temnospondyls. Several recent studies of postcranial material, including the phylogenetic analysis of Pawley and Warren (2005), and the description of the postcranial skeletons of more plesiomorphic temnospondyls such as Eryops megacephalus (Pawley and Warren, 2006) and Trimerorhachis insignis (Pawley, in press), all provide evidence that their phylogenetic affinities are not congruent with those of Yates and Warren (2000).

Sources of morphological variation

Sources of morphological variation in the postcranial skeleton are both interspecific (phylogenetic) variation, and intraspecific (phenotypic and morphogenetic). None of the previous studies of the postcranial skeleton of temnospondyls has distinguished between these types of variation. Chapter 4 described the extensive morphogenetic variation of the postcranial skeleton in temnospondyls, particularly in the endochondral postcranial skeleton. Of particular concern was the finding that certain characteristics, usually considered derived, such as the height of the neural spines, supinator process and radial condyle of the humerus (to name a few), develop with morphogenesis. In practical terms, phylogenetic characters defining the presence or absence of these morphological features are sometimes simply describing the difference between morphogenetically immature and mature specimens, i.e. describing morphogenetic rather than phylogenetic variation. The problem of immaturity vs. maturity is particularly apparent when comparing taxa with paedomorphic postcranial skeletons with non paedomorphic taxa. In paedomorphic morphogenesis is delayed, so that characteristics that develop late in morphogenesis and are present in early growth stages of non paedomorphic taxa are often only present in the largest specimens of paedomorphic taxa, CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 151 if at all (Pawley and Warren, 2004; Chapter 4; Pawley, in press). Unfortunately, for many temnospondyl genera, only single specimens are available, so it is usually impossible to determine whether morphogenetically immature specimens are truly immature or actually paedomorphic adults, which also means that the phylogenetic significance of heterochronic processes such as paedomorphism are difficult to objectively assess. Steyer (2000), in a study of cranial characters, found that that using characters that pertain to morphogenetic instead of phylogenetic variation had dramatic effects on the resulting cladogram topology, and that the use of these characters resulted in extremely inconsistent tree topologies that were dependant on the morphogenetic age of the specimens used. A diagnosis of the postcranial characters used by Yates and Warren (2000), which is the only other large scale computer based analysis of temnospondyls, indicates that several characters describe morphogenetic rather than phylogenetic variation. The presence or absence states of other postcranial characters are affected by the morphogenetic stage of specimens, additionally some did not separate into discrete character states within a larger sample of temnospondyls. The same concerns are applicable to the phylogenetic analysis of early tetrapods conducted by Ruta et al. (2003). These findings prompted a revision of the postcranial characters used in phylogenetic analysis of the Temnospondyli, presented here (Appendix 8). Potentially the use of morphogenetic variation rather than phylogenetic variation for phylogenetic analysis could increase the amount of homoplasy and mask true phylogenetic signal. In order to elucidate true phylogenetic signal, consideration and assessment of all sources of morphological variation is necessary, so that the effects of non-phylogenetic variation are minimised and phylogenetic analysis is based solely on phylogenetic variation.

The main aim of this study is to summarise the extent of phylogenetic variation in the postcranial skeleton of the Temnospondyli. The additional postcranial data is used to revise existing postcranial characters used in phylogenetic analysis of temnospondyls, and to create new characters, in order to reassess the phylogenetic relationships of taxa within the Temnospondyli.

MATERIALS AND METHODS

Data sources

As noted in the introduction, this study is only concerned with characteristics pertaining to phylogenetic variation within temnospondyls. Many specimens were examined personally, but this study was mainly literature based.

Abbreviations used in text

ALM, Albany Museum, Grahamstown, South Africa; AM, Australian Museum, Sydney, New South Wales; AMNH, American Museum of Natural History, New York City, New York; BMNH, Natural History Museum, London; BP, Bernard Price Institute, Johannesburg, South Africa; M, Field abbreviation, Council for Geosciences, Pretoria, South Africa; MCZ, Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts; MGUH, Museum Geologicum Universitatis Hafniensis, Copenhagen, Denmark; MNHN, Muséum National d'Histoire Naturelle, Paris; NMV, Museum of Victoria, Melbourne, Victoria; NMQR, National Museum at Bloemfontein, Bloemfontein, 152 K. PAWLEY PHD THESIS

South Africa; PIN, Palaeontological Institute of the Russian Academy of Sciences, Moscow; QM, Queensland Museum, Brisbane, Queensland; SAM, South African Museum, Cape Town, South Africa; TM, Transvaal Museum, Pretoria, South Africa; TMM, Texas Memorial Museum, Austin, Texas; UCMP, Museum of Paleontology, University of California, Los Angeles, California; UMMP, University of Michigan Museum of Paleontology, Ann Arbor, Michigan; UMZC, University Museum of Zoology, Cambridge, England; USMN, National Museum of Natural History, Washington; WAM, Western Australian Museum, Perth, Australia.

Main phylogenetic analysis

Previous analyses of early tetrapods, including temnospondyls, have concentrated on cranial characteristics; consequently, the majority of cranial characters come from previous studies. The McClade (Macintosh) built matrix of Ruta et al. (2003) was converted into a PC format acceptable to Nexus Data Editor (NDE) (Page, 2001), and additional data was input using NDE. Many characters in this matrix were recoded due to errors and changes to character states. Additional characters were taken from Yates and Warren (2000), and various other sources (Sequeira and Milner, 1993; Carroll, 1995; Lebedev and Coates, 1995; Coates, 1996; Paton et al., 1999; Damiani, 2001a; Klembara and Ruta, 2004b; Pawley and Warren, 2004; Sequeira, 2004; Pawley and Warren, 2005). Many characters used by Ruta et al. (2003) were omitted because they were not parsimony informative for an analysis of the Temnospondyli. Some cranial characters employed by Ruta et al. (2003) were relevant, but were omitted because they are subject to character state change with progressive morphogenetic stages (Bystrow and Efremov, 1940; Milner, 1980a; Godfrey, 1989b; Milner and Sequeira, 1994; Shishkin and Rubidge, 2000; Steyer, 2000 and references therein; Steyer, 2002; Werneburg and Steyer, 2002; Steyer, 2003). Other characters, including postcranial characters, were omitted because they are proportion based (the proportional variation described is continuous rather than forming discrete character states within the taxa sample used in this analysis), or not independent (e.g. prezygapophyses cannot exist without postzygapophyses). The main analysis has 25 new characters (seven cranial and 18 postcranial). Appendix 6 lists the terminal taxa used in the main analysis, along with specimens examined and repository information. As some descriptions of postcranial material are inaccurate, reinterpretations are listed in Appendix 7. The character states (154 total, 114 cranial and 40 postcranial) are listed in Appendix 8, and the data matrix of 58 taxa and 154 characters in Appendix 9. Note that the unique character identifiers (e.g. INTCLA 1) were originally used in phylogenetic analysis of early tetrapods by Ruta et al. (2003), the unique character identifiers used in the main analysis are consistent in that they are either the same characters, or new characters with numbering additional to Ruta et al. (2003). The unique character identifiers are also consistent with those of a new analysis of the phylogenetic position of the Temnospondyli within early tetrapods (Chapter 6). Phylogenetic trees were prepared for publication using TreeView (Page, 1996).

Choice of outgroup taxa

In accordance with Bryant (1997) actual outgroup taxa were employed in the main analysis, rather than a hypothetical ancestor. The addition of extra basal temnospondyls solved initial problems encountered in resolving relationships between basal taxa. Caerorhachis bairdi and Edops craigi were set as outgroup taxa, based on the results of a CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 153 much larger, more inclusive analysis (Chapter 6) that included a substantial number of non-temnospondyl early tetrapod taxa. Casineria kiddi and Caerorhachis bairdi are sister taxa in this analysis, but the specimen of Casineria kiddi (Paton et al., 1999) is too incomplete to be used as an outgroup taxon, despite the strong statistical support for its phylogenetic position. Interestingly, any attempt to significantly modify the early tetrapod (sensu Ruta et al., 2003) lineage obtained in the more inclusive analysis (to provide a suitable non temnospondyl outgroup for this analysis) resulted in a different arrangement of basal temnospondyls to that of the more inclusive analysis. Removal of all non- temnospondyl taxa, however, resulted in an identical tree topology within the temnospondyl clade to that of the more inclusive analysis, regardless of whether outgroup taxa were set or not. It was therefore deemed unnecessary (at least in this case) to include non-temnospondyl outgroup taxa.

Choice of ingroup taxa

To reduce a priori assumptions of monophyly, species level terminal taxa were used as much as possible. Where this was not possible genus level terminal taxa were used, occasionally it was necessary to use family level taxa. Although some taxa in this analysis lack associated postcranial material, they were included for the sake of completeness and to facilitate comparison with other analyses. An analysis of early tetrapods (Chapter 6) unexpectedly placed Casineria kiddi as the sister taxa to Caerorhachis bairdi. Statistical support for this node is high, and examination of the data indicates that Casineria kiddi and Caerorhachis bairdi are indistinguishable on the basis of the available evidence, so it is very likely that Casineria kiddi is a basal temnospondyl (Chapter 6). Casineria kiddi was not included in the phylogenetic analysis because it is too incomplete, but is included in the postcranial descriptions. Taxa with postcranial skeletons of larval morphology e.g. branchiosaurs were not included, because their postcranial skeletons are so morphogenetically immature that they are almost completely devoid of identifying characteristics. Bolt (1974d) considered that Broiliellus brevis may be an immature morphogenetic stage of Dissorophus spp., pending further study Broiliellus brevis was included. Tersomius spp. was excluded, as its taxonomic status is dubious (Bolt, 1977). Using the phylogenetic trunk method of Anderson (2001), Lapillopsis nana was excluded because its phylogenetic position is unstable, and its inclusion results in a dramatic increase in tree length by up to 20 steps. The peculiar mix of plesiomorphic and derived characters interpreted in the tiny specimens of this taxon are likely to be the cause of this problem.

Selection and coding of postcranial characters

It is important that phylogenetic character states, and the coding of these states, refers to intraspecific (phylogenetic), rather than interspecific (morphogenetic or phenotypic) variation. Minimisation of the confounding effects of non-phylogenetic variation is essential to obtain a true phylogenetic signal. Consequently, the morphogenetic stage of specimens (Chapter 4) was taken into account when coding the matrix. Specimens with morphogenetically immature postcranial skeletons are denoted in the taxon list (mi). Postcranial character states known to be affected by the morphogenetic stage of the specimen are denoted ‘m’ in the character state list, and are accompanied by selection criteria to facilitate identification of immature specimens in which the derived state is indeterminate. If it was possible that a derived postcranial characteristic may be absent in 154 K. PAWLEY PHD THESIS a particular specimen due to the morphogenetic immaturity of the specimen, then the character state was coded ‘?’, to avoid confusing morphogenetic immaturity with the plesiomorphic state.

Analysis of matrices

Analysis of all matrices was performed using PAUP 4.0 beta WIN10 (Swofford, 2003), on a Pentium 4 PC. All characters were parsimony informative, unweighted, five characters are ordered. Parsimony analysis employed the heuristic search algorithm with the following settings: uninformative characters excluded, character state changes optimised with ACCTRAN, random addition sequence with ten replicates, and the TBR (tree bisection-reconnection) branch-swapping algorithm. Bootstrapping used the same options, but with 100 replicates, the number of trees held in memory per replicate was limited to 10,000 in order to limit branch swapping on suboptimal trees. Bremer node values were calculated using 10 replicates, and limited to 100,000 trees per replicate was set.

Tests of degrees of character inclusion

The effect of including postcranial characters was tested by excluding all postcranial characters. Preliminary analyses removing all cranial characters and taxa without postcranial material (Chenoprosopus milleri, Cochleosaurus spp., Saharastega moradiensis, Nigerpeton ricqlesi, Adamanterpeton ohioensis, Capetus palustris, Konzhukovia vetusta, Luzocephalus spp., Thoosuchus yakovlev, Xenobrachyops allos) resulted in a complete lack of resolution after 100,000 trees and consequently were deemed unviable. The effect of removing vertebral characters that have historical significance for temnospondyl taxonomy, but may reflect only differing degrees of ossification of an originally cartilaginous amphicoelous centra (Chapter 4), was tested by removing the relevant characters (TRU VER 8, TRU VER 8, TRU VER 14), and another analysis run where TRU VER 33 was additionally excluded. Results of these analyses are presented in Appendix 10.

Tests of phylogenetic stability

Constraint trees were built to assess alternative phylogenetic relationships for some contentious taxa. Results of these analyses are presented in Appendix 10.

Comparative phylogenetic analyses

Several tests were run in order to compare the results of the main analysis with that of Yates and Warren (2000). For all the following analyses, the same taxon list was used (including Lapillopsis nana and Tersomius spp. which were omitted from the main analysis), with a hypothetical ancestor as the outgroup, as in Yates and Warren (2000). First, the postcranial characters of the matrix of Yates and Warren (2000) were replaced, but the cranial characters and coding left unchanged, in order to test the effect of changing the postcranial characters (138 total; 102 cranial, 36 postcranial). Second, a separate data matrix was constructed using the same terminal taxa as Yates and Warren (2000), but employing the characters (134 total; 98 cranial, 36 postcranial) and coding used in the main analysis, in order to test the effect of taxon CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 155 sampling. Both these analyses were also subjected to a constrained search reproducing the tree topology of Yates and Warren (2000). Results of these analyses are presented in Appendix 10.

Comparisons of consistency indices

Expected consistency indices for the whole data sets of the new analysis and that of Yates and Warren (2000) were calculated using the method outlined by Sanderson and Donoghue (1989) for all data sets. The formula for calculating the expected consistency indices is CI = 0.90 - 0.022 (number of taxa) + 0.000213 (number of taxa)2 The resulting consistency index is then compared with the obtained consistency index from the phylogenetic analysis. If the obtained consistency index is higher than the expected consistency index, the data set contains a strong phylogenetic signal. Results of these analyses are presented in Appendix 10.

RESULTS OF MAIN PHYLOGENETIC ANALYSIS

The main analysis resulted in one tree (Figure 44) of 359 steps (58 taxa, 154 characters, CI = 0.4485, RI = 0.8386, RC = 0.3761). The most basal taxa within the Temnospondyli are Caerorhachis bairdi and Edops craigi, with the Dendrerpetontidae, Cochleosauridae, Dvinosauria, and Eryopoidomorpha as successively more derived clades. Figure 45 shows bootstrap and Bremer node values. Phylogenetic distributions of all character states are provided in the character state list (Appendix 8), figures of cladograms for comparative analyses in Appendix 10, and complete lists of synapomorphies and state changes for each clade are provided in Appendix 11.

Taxonomic definitions

Two taxonomic definitions were necessary to facilitate description and discussion of the temnospondyl taxa in this analysis. Apomorphy-based definitions (de Queiroz and Gauthier, 1990) are used.

Eryopoidomorpha

This clade includes the Euskelia and Stereospondylomorpha. The type taxon for the Eryopoidomorpha is Eryops megacephalus Cope 1877, and the two postcranial apomorphies that define this clade are: the mesial iliac ridge forms the anterior border of the ilium (ILI 8), and the adductor crest passes down the midline of the femur (FEM 8).

Superstereospondyli

This distinct clade of derived stereospondyls, united by several postcranial apomorphies, contains the only stereospondyl taxa that possess stereospondylous vertebrae. The latest definition of the Stereospondyli was provided by Yates and Warren (2000), however this definition was stem based rather than apomorphy-based, and did not mention the type apomorphy. Redefinition of the term ‘Stereospondyli’ to correctly refer to the clade of derived stereospondyls, using a strict apomorphy-based definition, would involve

156 K. PAWLEY PHD THESIS

FIGURE 44. Results of main phylogenetic analysis of the Temnospondyli. One tree of 359steps (58 taxa, 154 characters, CI = 0.4485, RI = 0.8386, RC = 0.3761). CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 157

removing the Rhinesuchidae and Lydekkerinidae to the Stereospondylomorpha, uniting them with the Archegosauridae, which would be contrary to popular usage of this term. Consequently, in the interests of taxonomic stability, the definition of the Stereospondyli (Yates and Warren, 2000) remains unaltered; and a new apomorphy based (de Queiroz and Gauthier, 1990) taxon “Superstereospondyli’ was created to refer to the distinctive clade of derived stereospondyls. The type taxon for the Superstereospondyli is Mastodonsaurus giganteus Jaeger 1828. Superstereospondyls are characterised by several postcranial apomorphies: stereospondylous vertebrae (not found in all taxa); specialised cervical vertebrae in which the dorsal tips of the atlantal neural spines are fused (m CER VER 1); and that lack the atlantal prezygapophyses (CER VER 10), atlantal postzygapophyses (CER VER 11), and axial prezygapophyses (CER VER 12); the centre of ossification of the interclavicle (INTCLA 9) is located posterior to the line of maximum width; the clavicular blades broadly contact anterior to the main body of the interclavicle (CLA 3); and the supraglenoid foramen of the scapulocoracoid forms a ventrally open notch (m SCACOR 5) at least in early morphogenetic stages (Chapter 4). The Superstereospondyli are also named for the extremely large size (in excess of six metres) attained by some taxa (Damiani, 1999; Schoch, 1999a; Steyer and Damiani, 2005).

PHYLOGENETIC VARIATION OF THE POSTCRANIAL SKELETON WITHIN THE TEMNOSPONDYLI

The following discussion is limited to postcranial characteristics subject to phylogenetic variation within the Temnospondyli. Morphogenetic variation (Chapter 4), and character states which distinguish the postcranial skeletons of temnospondyls from stem tetrapods and other early tetrapod taxa are discussed elsewhere (Chapter 6). Note that much of the previously described phylogenetic variation in the postcranial skeleton of temnospondyls is in fact morphogenetic variation, as many postcranial characteristics develop with morphogenesis, and may be absent in morphogenetically immature specimens (Chapter 4). Most of the characteristics described were included in the main phylogenetic analysis; in these cases, a unique character identifier denotes the relevant character (e.g. INTCLA 1) (Appendix 8). Character states in plain type (e.g. TRU VER 1) were previously published and their distribution discussed in Ruta et al. (2003). Characters used in other previously published analyses (see materials and methods) have only the numbers in bold (e.g. TRU VER 28). New characters, not previously used in phylogenetic analyses of early tetrapods, are entirely in bold (e.g. INTCLA 7). The letter ‘m’ prefacing a unique character identifier indicates that the characteristic is subject to morphogenetic development and variation. An analysis of the literature (Appendix 1) indicates that the postcranial skeleton of basal temnospondyls, particularly the Cochleosauridae, is not well known. Consequently, plesiomorphic states for many postcranial characteristics are unclear. High levels of missing data in the postcranial character set hampers accurate determination of the phylogenetic distribution of many characteristics. Sometimes it is difficult to allocate state changes to specific nodes, so phylogenetic distributions of postcranial characteristics listed here (also provided in the Character state list, Appendix 8) should be treated with caution. References for taxa included in the analysis are also listed in Appendix 6. Comprehensive comparison of the postcranial skeleton of temnospondyls with that of other early tetrapods is beyond the scope of this study, and is provided elsewhere (Chapter 6). 158 K. PAWLEY PHD THESIS

FIGURE 45. Bremer node and bootstrap values for main phylogenetic analysis of the Temnospondyli. Left numbers are bootstrap values, right and single numbers are Bremer node values. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 159

As is typical for tetrapods, the postcranial skeleton of temnospondyls is formed of both dermal (clavicle, interclavicle and cleithrum) and endochondral bone (all other postcranial elements). Ornament on the external surfaces of the interclavicle and clavicles is always similar to that of the skull and mandible. The pectoral girdle is composed of a median interclavicle, and paired clavicles, cleithra and scapulocoracoids (Figure 51). The pelvic girdle (Figure 58) is composed of paired ilia, ischia and, when ossified, pubis.

Vertebrae

Typically, the vertebrae of temnospondyls are rhachitomous (Figure 46.1.1, Figure 47.1), with separate vertebral elements consisting of a neural arch and wedge-shaped paired pleurocentra and a crescentric ventral intercentrum. Temnospondyl vertebrae are diplospondylous (Arratia et al., 2001) i.e. with two components per vertebral centrum. The only exception is the trunk centra of Peltobatrachus pustulatus (Panchen, 1959) which lack intercentra. Although temnospondyls are named for their vertebrae, rhachitomous vertebrae are a plesiomorphic feature of sarcopterygian fish and early tetrapods, and not an apomorphy of temnospondyls. The rhachitomous type of vertebrae is the most common, but considerable variation in centrum morphology is observable in temnospondyls, always involving an increase in the degree of ossification towards the formation of amphicoelous disks (Chapter 4). Centrum types (other than rhachitomous) include embolomerous (both pleurocentra and intercentra form complete amphicoelous rings, Figure 47.3), gastrocentrous centra (intercentra highly reduced or absent in the trunk region, Figure 47.5, Figure 47.6), and stereospondylous centra (pleurocentra highly reduced, Figure 47.7, Figure 47.8). The vast majority of temnospondyls have rhachitomous vertebrae; all of the previously listed variations are rare, and none of them is more common phylogenetically. Little regional differentiation is observable in the temnospondyl vertebral column; the atlas, axis (Figure 48), sacral (Figure 46.5) and caudal vertebrae (Figure 46.6, Figure 46.7) are easily determined, but otherwise individual vertebrae are difficult to distinguish. Hellrung (2003) demonstrated that some plagiosaur centra, previously considered anomalous (Nilsson, 1939, 1946a; Romer, 1947; Panchen, 1967; Warren and Snell, 1991), are formed by the fusion of the intercentrum and pleurocentrum, a phenomenon which occurs occasionally in other temnospondyls such as Eryops megacephalus (Moulton, 1974) and Dvinosaurus primus. (Bystrow, 1938).

Number of presacral vertebrae

Within the Temnospondyli, the number of presacral vertebrae varies from 32 in some of the more basal taxa to approximately twenty in some derived taxa. Basal temnospondyls: Caerorhachis bairdi 32 (Holmes and Carroll, 1977), Dendrerpeton acadianum 24 (Holmes et al., 1998), Balanerpeton woodi 24 (Milner and Sequeira, 1994). Dvinosauria: Trimerorhachis insignis 32 (Williston, 1916), Dvinosaurus primus 31 (Bystrow, 1938). Euskelia: Sclerocephalus haeuseri 25 (Lohmann and Sachs, 2001; Schoch, 2003), Eryops megacephalus 22 (Moulton, 1974); Dissorophidae: Cacops aspidephorus 21 (Gregory, 1950), Aspidosaurus novomexicanus 21 (Carroll, 1964a), Dissorophus angustus 26 (Carroll, 1964a), Dissorophus multicinctus 25 (de Mar, 1968), Acheloma cumminsi (Trematops milleri) 24 (Olson, 1941); Amphibamidae: Amphibamus grandiceps 20

160 K. PAWLEY PHD THESIS

FIGURE 46. Temnospondyl vertebrae- neural arch. 1, Eryops megacephalus, 1.1 articulated vertebrae in left lateral view, 1.2 neural arch in anterior view, 1.3 neural arch in posterior view, after Moulton (1974). 2, Ecolsonia cutlerensis neural arch in left lateral view, after Berman et al. (1985). 3, Acerastea wadeae neural arch in 3.1 anterior, and 3.2 left lateral views, after Warren and Hutchinson (1987). 4, Platyhystrix rugosus in left lateral view, after Lewis and Vaughn (1965). 5. Eryops megacephalus in left lateral view, after Moulton (1974). 6, Siderops kehli in 6.1 posterior, and 6.2 left lateral views, after Warren and Hutchinson (1983). 7, Eryops megacephalus in left lateral view, after Moulton (1974). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 161

(Watson, 1940; Gregory, 1950; Daly, 1994), Platyrhinops (Amphibamus) lyelli 25 (Carroll, 1964a), Doleserpeton annectans 25 (Bolt, 1969). Stereospondylomorpha: Uranocentrodon (Myriodon) senekalensis 28 (Haughton, 1915; van Hoepen, 1915), Paracyclotosaurus davidi 28 (Watson, 1958), Gerrothorax pustuloglomeratus 21-23 (Hellrung, 2003), Siderops kehli 23 (Warren and Hutchinson, 1983).

Neural arch

The neural spines of temnospondyls lack a supraneural canal (Figure 46.1.2) and are typically rectangular in lateral view (Figure 46.1.1), with pre- and postzygapophyses (Figure 46.1.2, Figure 46.1.3) on all vertebrae including the posterior caudal vertebrae. The neural arch always articulates with the pleurocentra, which are never absent, though they may be highly reduced or poorly ossified (Warren and Snell, 1991). Little variation is observable in the neural spines of temnospondyls. The neural spines are fused to the enlarged pleurocentra in Doleserpeton annectans (Bolt, 1969; Carroll et al., 2004) (Figure 47.6) and Fayella chickashaensis (Olson, 1972). Lateral tubercles (TRU VER 32) (Figure 46.2) are present in Parioxys bolli (Carroll, 1964b) and Ecolsonia cutlerensis (Berman et al., 1985). The unusual dorsal ‘sail’ of Platyhystrix rugosus (Lewis and Vaughn, 1965) (Figure 46.4) is formed of enlarged, flattened neural spines, similar to those of some pelycosaurs (Romer and Price, 1940). The ‘sail’ of Platyhystrix rugosus is not vascularized, and is unlikely to have a thermoregulatory function as proposed by Romer (1948) for Dimetrodon spp. (Vaughn, 1971). The buttresses are enlarged so that the neural spines are cross-shaped in rhytidostieds (Figure 46.3) and plagiosaurs. The trunk diapophyses are flattened rather than ovoid in the trunk neural arches of euskelians such as Eryops megacephalus (Figure 46.1.1) and Ecolsonia cutlerensis (Berman et al., 1985) (Figure 46.2).

Intercentra

The intercentrum are always centred ventrally, and characteristically form a crescent shaped half ring, wedge shaped in lateral view. Intercentra are identifiable by the presence of laterally located parapophyses for articulation with the tuberculum of the rib, but pleurocentra have dorsally located facets for articulation with the neural arch. Typically, intercentra are ridged and pitted ventrally, in plesiomorphic temnospondyls such as Trimerorhachis insignis (Pawley, in press) (Figure 47), they are usually quite thin in cross section, but in most taxa within the Eryopoidomorpha they are ventrally thickened if not stereospondylous. Posterolaterally directed tubercles are present on the cervical intercentra of Trimerorhachis insignis (Pawley, in press) (Figure 48.4), which are a retained plesiomorphy (Godfrey and Reisz, 1991). The phylogenetic distribution of this characteristic is indeterminable in other basal temnospondyls, but the tubercles are not known within the Eryopoidomorpha. Stereospondylous (m TRU VER 14) and essentially monospondylous (TRU VER 33) centra (Figure 47.7, Figure 47.8), although considered typical of derived temnospondyls, are present in the dissorophoid Parioxys ferricolus (Moustafa, 1955a) and an indeterminate temnospondyl (Warren et al., 2001). Within the Stereospondyli they are known only in some superstereospondyls including: the mastodonsaurids Mastodonsaurus giganteus (Schoch, 1999a) (Figure 47.7), M. torvus (Ochev, 1972), Eryosuchus tverdochlebovi (Ochev, 1972); all metoposaurs (e.g. Sawin, 1945; Dutuit, 1976) (Figure 47.8); the brachyopoid Gobiops desertus 162 K. PAWLEY PHD THESIS

FIGURE 47. Temnospondyl vertebrae- presacral centra. 1, Trimerorhachis insignis in 1.1 left lateral, and 1.2 anterior views, from Pawley (in press). 2, Ecolsonia cutlerensis in 2.1 anterior and 2.2 left lateral views, after Berman et al. (1985). 3, Tupilakosaurus wetlugensis in 3.1 anterior and 3.2 left lateral views, after Shishkin (1961). 4, Caerorhachis bairdi in anterior view, after Holmes and Carroll (1977). 5. Peltobatrachus pustulatus in 5.1 anterior, and 5.2 left lateral views, after Panchen (1959). 6, Doleserpeton annectans in left lateral view, after Bolt (1969). 7, Mastodonsaurus giganteus in 7.1 left lateral, and 7.2 posterior views, after Schoch (1999a). 8, Buettneria perfecta in 8.1 anterior, and 8.2, 8.3 left lateral views, after Sawin (1945). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 163

(Shishkin, 1991; Maisch and Matzke, 2005), and the chigutisaur Koolasuchus cleelandi (Warren et al., 1997). Opisthoceolous intercentra are present in some cervical vertebrae of the otherwise amphicoelous vertebrae of the metoposaur Buettneria perfecta (howardensis) (Sawin, 1945) (Figure 47.8). Note that the intercentra of another species of Eryosuchus, E. (Parotosuchus) pronus (Howie, 1970), are typically rhachitomous, rather than stereospondylous. Regional differentiation is present along the vertebral column in Mastodonsaurus giganteus (Schoch, 1999a), cervical and trunk vertebrae are stereospondylous, but the sacral and caudal vertebrae are rhachitomous (Figure 47.7.1), this variation may also be morphogenetic (Chapter 4). Embolomerous centra (TRU VER 33, m TRU VER 8, m TRU VER 9, m TRU VER 14), in which both the intercentra and pleurocentra form subequal amphicoelous disks, are characteristic of tupilakosaurs, such as Thabanchuia oomie (Warren, 1998b) and Tupilakosaurus spp. (Nielsen, 1954; Shishkin, 1961) (Figure 47.3). Proto-embolomerous centra, in which both the intercentra and pleurocentra are ventrally, but not dorsally fused, are present in Caerorhachis bairdi (Holmes and Carroll, 1977; Ruta et al., 2002) and Casineria kiddi (Paton et al., 1999).

Pleurocentra

The trunk pleurocentra are ventrally fused (m TRU VER 8) in Caerorhachis bairdi (Holmes and Carroll, 1977; Ruta et al., 2002) and Casineria kiddi (Paton et al., 1999) (Figure 47.4), tupilakosaurs (Nielsen, 1954; Shishkin, 1961; Warren, 1998b) (Figure 47.3), Peltobatrachus pustulatus (Panchen, 1959) (Figure 47.5), Fayella chickashaensis (Olson, 1972), Tersomius cf. T. texensis (Daly, 1994), and derived amphibamids (Bolt, 1969; Daly, 1994). Dorsally fused pleurocentra (TRU VER 9) (which may also be ventrally fused) are observable in tupilakosaurs (Nielsen, 1954; Shishkin, 1961; Warren, 1998b) (Figure 47.3), Peltobatrachus pustulatus (Panchen, 1959) (Figure 47.5), Fayella chickashaensis (Olson, 1972), and Doleserpeton annectans (Bolt, 1969; Carroll et al., 2004) (Figure 47.6). Essentially monospondylous (TRU VER 33), pleurocentrum dominated centra (trunk intercentra highly reduced or absent) are found in Peltobatrachus pustulatus (Panchen, 1959) (Figure 47.5); and the unrelated dissorophids Fayella chickashaensis (Olson, 1972), and Doleserpeton annectans (Bolt, 1969; Carroll et al., 2004) (Figure 47.6).

Cervical vertebrae

The atlantal neural spine differs from that of other vertebrae, and is typically is a small spinous process (Figure 48). An isolated axis is not easily distinguished from other vertebrae, except that the neural spine may be somewhat more massive. Some temnospondyls have a reduced fourth neural spine (Figure 48.1.2), as in Eryops megacephalus (Moulton, 1974) and Sclerocephalus haeuseri (Boy, 1988). An ossified proatlas is sometimes present in Sclerocephalus haeuseri (Boy, 1988) and Eryops megacephalus (Moulton, 1974) (Figure 48.1.2); the presence of a cartilaginous proatlas can be inferred in all temnospondyls except the Superstereospondyli, by the prezygapophyses for articulation with the proatlas on the atlas. Plesiomorphically, the atlas is multipartite, with a paired neural arch and separate intercentra and pleurocentra (Figure 48.1.1). The atlantal neural arch fuses to the atlantal 164 K. PAWLEY PHD THESIS

FIGURE 48. Temnospondyl cervical vertebrae. 1, Eryops megacephalus 1.1 anterior, and 1.2 left lateral views, after Moulton (1974). 2, Platyoposaurus stuckenbergi in 2.1 anterior and 2.2 left lateral views, after Gubin (1991). 3, Buettneria perfecta in 3.1 anterior and 3.2 left lateral views, after Sawin (1945). 4, Trimerorhachis insignis intercentrum in 4.1 left lateral, and 4.2 ventral views, from Pawley (in press). 5, Cacops aspidephorus in anterior view, after Williston (1910a). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 165 centrum (m CER VER 2) in the Dissorophidae (e.g. Williston, 1910a; de Mar, 1968; Daly, 1994) (Figure 48.5), and the Stereospondylomorpha (Chernin, 1977; Schoch, 1999a; Hellrung, 2003) (Figure 48.2, Figure 48.3). It is not known whether the centrum is formed by the fusion of both the pleurocentrum and intercentrum, or whether the pleurocentrum are lost and the centra is formed entirely of the intercentrum (Shishkin, 2000). Superstereospondyls have specialised cervical vertebrae (Figure 48.3), the dorsal tips of the atlantal neural spines are fused (m CER VER 1), and the atlantal prezygapophyses (CER VER 10), atlantal postzygapophyses (CER VER 11), and axial prezygapophyses (CER VER 12) are absent. An anteroposteriorly elongate atlas centrum is characteristic of brachyopids (Shishkin, 2000; Warren and Marsicano, 2000)

Sacral vertebrae

Temnospondyls typically posses only one sacral vertebra (Figure 46.5), within the Dissorophidae, two are present in Tersomius cf. T. texensis (Daly, 1994), Cacops aspidephorus (Williston, 1910a) and Parioxys spp (Moustafa, 1955a; Carroll, 1964b).

Caudal vertebrae

Few temnospondyls have articulated caudal vertebrae. Within the Euskelia, Cacops aspidephorus has 20 (de Mar, 1968), Sclerocephalus haeuseri approximately 25 (Lohmann and Sachs, 2001), and Eryops megacephalus approximately 25 (Moulton, 1974), approximately 36 are present in the basal stereospondyl Uranocentrodon senekalensis (van Hoepen, 1915). In all these taxa, the number of postsacral vertebrae is subequal to the number of presacral vertebrae. Exceptionally well-preserved larval temnospondyls have fleshy dorsal and ventral caudal fins (Milner, 1982a) (Figure 46.6), but the presence of these has not been determined in adults.

Haemal arches

Haemal arches (Figure 46.6, Figure 46.7) are present on all caudal vertebrae that lack ribs. The haemal arch always attaches to the intercentra in temnospondyls; the length of the arch corresponds to the length of the neural spine in the same vertebral segment.

Ceratobranchials

As part of the branchial skeleton, only ceratobranchials are described. As noted in Chapter 4, ceratobranchials (Figure 49) were probably present in all temnospondyls, although rarely ossified (also discussed by Witzmann, 2005). Four ceratobranchials are present on each side of the neck; each ceratobranchial curves ventrally, the ventral surface is shallowly grooved for the afferent branchial aortic arches. Preserved ceratobranchials are rare, consequently it is difficult to assess their original position. In Trimerorhachis insignis (Olson, 1979) (Figure 49.2) and Gerrothorax pustuloglomeratus (Hellrung, 2003) (Figure 49.3), the ceratobranchials are reconstructed inside the skull cavity, medial to the posterior end of the jaw, but in Dvinosaurus primus (Bystrow, 1938) (Figure 49.1); they are reconstructed alongside thecervical vertebrae. The anterior ends of the interclavicle and clavicles are located between the posterior ends of the jaws in articulated specimens (Williston, 1916; Pawley and Warren, 2005) (Figure 51); so that in life the ceratobranchials would have occupied most of the space between the skull and pectoral girdle, as in extant salamanders (Deban and Wake, 2000; Wake and Deban, 2000). 166 K. PAWLEY PHD THESIS

FIGURE 49. Temnospondyl ceratobranchials. 1, Dvinosaurus primus in ventral view (scale unknown), after Bystrow (1938). 2, Trimerorhachis insignis in ventral view, after Olson (1979). 3, Gerrothorax pustuloglomeratus in ventral view, after Hellrung (2003). 4, Micromelerpeton credneri, after Boy (1995). 5, Micromelerpeton credneri, after Witzmann and Pfretzschner (2003). 6, reconstruction of Branchiosaurus sp., after Milner (1982a). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 167

In branchiosaurs and the rhinesuchid Uranocentrodon senekalensis (de Fauw, 1989) denticulated plates (gill rakers) (Figure 49.4) were attached to the cartilaginous ceratobranchials (e.g. Boy, 1995) (Figure 49.5). Although four ceratobranchials are present, only three external gills were present in larval temnospondyls (Witzmann and Pfretzschner, 2003; Witzmann, 2004) (Figure 49.5, Figure 49.6).

Ribs

The cervical ribs of temnospondyls (Figure 50.1) are straight and expanded distally. The trunk ribs (Figure 50) gently curve ventrally (measured along the ventral edge). The expanded proximal head consists of a tuberculum that articulates with the transverse process, and a capitulum that articulates with the centrum, joined by thinner bone. Uncinate processes (m RIB 5) are unknown in basal temnospondyls, such as Balanerpeton woodi (Milner and Sequeira, 1994) and Trimerorhachis insignis (Pawley, in press), but are present on the anterior trunk ribs of all taxa within the Eryopoidomorpha, excepting amphibamids and plagiosauroids (Figure 50.2, Figure 50.3). Foramina, distal to the proximal articulation surface, are variably present in some trunk ribs of temnospondyls such as Peltobatrachus pustulatus (Panchen, 1959), rhytidosteids (Howie, 1972a; Warren and Black, 1985), and Mastodonsaurus giganteus (Schoch, 1999a) (Figure 50.3.2).

Sacral rib

The sacral rib of temnospondyls (Figure 50.4) is short and stout, with a thickened posteroventral edge, and the whole rib is posteriorly recurved. The sacral rib articulates with the vertebral column in the same way as do other ribs, and the expanded distal end lies flush against the inner surface of the ilium. Note that the description of the sacral rib articulation by Olson (1936a) cannot be correct, as the tuberculum and capitellum are dorsally positioned in Olson’s reconstruction, so that they would be unable to articulate with the sacral vertebrae.

Caudal ribs

Caudal ribs (Figure 50.5) are present on the eight most anterior postsacral vertebrae of Eryops megacephalus (Moulton, 1974), Eoscopus lockardi (Daly, 1994), Mastodonsaurus giganteus (Schoch, 1999a), and Siderops kehli (Warren and Hutchinson, 1983). Caudal ribs are short and tapered; the most posterior caudal ribs are sharply recurved.

Interclavicle

The interclavicle of temnospondyls (Figure 51, Figure 52) is typically diamond shaped, with an ornamented main body, surrounded by a thinner unornamented border. The ornament of the main body is usually of the same type of pit and groove type as on the other dermal bones. The ornament, pitted at the centre of ossification, elongates into grooves radiating towards the edges of the interclavicle. In euskelians such as Eryops megacephalus (Pawley and Warren, 2006) (Figure 52.1), Acheloma cumminsi (Olson, 1941), and Ecolsonia cutlerensis (Berman et al., 1985), the ornament is transversely oriented, with enlarged pits. The thin, unornamented border surrounding the main body of the interclavicle tapers out distally into a thin film of bone (Pawley and Warren, 2004), that is thin, easily 168 K. PAWLEY PHD THESIS

FIGURE 50. Temnospondyl ribs. 1, Balanerpeton woodi after Milner and Sequeira (1994). 2, Eryops megacephalus after Moulton (1974). 3, Mastodonsaurus giganteus after Schoch (1999a). 4, Eryops megacephalus in 4.1 anterior, 4.2 posterior, and 4.3 medial views, from Case (1911a). 5, Siderops kehli in 5.1 dorsal, and 5.2 anterolateral views, after Warren and Hutchinson (1983). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 169 damaged, and often not preserved in disarticulated specimens. The clavicular facets underlap the interclavicular border anteriorly; the ventral scutes (Figure 61.2.2) underlap the posterior border. The depth of the border is variable, but the area overlapping the clavicles is more deeply incised than the posterior border. The interclavicular border is ubiquitous among temnospondyls; examples include Capetus palustris (Sequeira and Milner, 1993), Dendrerpeton acadianum (Holmes et al., 1998), Trimerorhachis insignis (Pawley, in press), Acanthostomatops vorax (Boy, 1989), Eryops megacephalus (Pawley and Warren, 2006), Aspidosaurus (Zatrachys) apicaulis (Case, 1907), Cheliderpeton spp. (Boy, 1993; Werneburg and Steyer, 2002), Archegosaurus decheni (Meyer, 1857), an unnamed rhinesuchid (Pawley and Warren, 2004), Mastodonsaurus giganteus (Schoch, 1999a), and plagiosaurs such as Gerrothorax pustuloglomeratus (Shishkin, 1987; Hellrung, 2003). A pectinate border of the interclavicle (Figure 52.1, Figure 52.2) is typical. Preservation of this feature is rare because it is thin, fragile, easily broken and readily destroyed in disarticulated specimens. The anterior most edges of the fimbriate border extend beyond the anterior borders of the clavicles when the pectoral girdle is preserved in articulation, as in Dendrerpeton acadianum (Holmes et al., 1998), and an unnamed rhinesuchid (Pawley and Warren, 2004).

FIGURE 51. Temnospondyl pectoral girdle. 1, Lydekkerina huxleyi in 1.1 ventral view, 1.2 left lateral view, from Pawley and Warren (2005). Scale bar = 25mm. 170 K. PAWLEY PHD THESIS

FIGURE 52. Temnospondyl interclavicle. 1, Eryops megacephalus in ventral view, from Pawley and Warren (2006). 2, Trimerorhachis insignis in ventral view, from Pawley (in press). 3, Benthosuchus sushkini in dorsal view, after Bystrow and Efremov (1940). 4, Dvinosaurus primus in ventral view (scale unknown), after Bystrow (1938). 5, Mastodonsaurus giganteus in ventral view, after Schoch (1999a). 6, Siderops kehli partial articulated clavicles and complete interclavicle in ventral view, after Warren and Hutchinson (1983). 7, Gerrothorax pulcherrimus in ventral view (scale unknown), after Shishkin (1987). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 171

The posterior border of the interclavicle is broadly hemispherical (INTCLA 7) in most euskelians more derived than Onchiodon labyrinthicus (Figure 52.1). A parasternal process (INTCLA 1) is present on the posterior border of the interclavicle in the Dvinosauroidea [Dvinosaurus primus (Bystrow, 1938) (Figure 52.4) and Thabanchuia oomie (Warren, 1998b)], the dissorophid Cacops aspidephorus (Williston, 1910a), and some stereospondyls [Lydekkerina huxleyi (Pawley and Warren, 2005), Mastodonsaurusgiganteus (Schoch, 1999a), trematosaurs (e.g. Nilsson, 1943; Steyer, 2002; Schoch, 2006), and chigutisaurs (Warren and Hutchinson, 1983; Marsicano, 1993; Warren et al., 1997)] (Figure 52). In trematosaurs, the interclavicle becomes increasingly elongate with maturity (Schoch, 2006). The parasternal process in temnospondyls typically tapers gradually as it extends posteriorly, in Dvinosaurus primus the parasternal process is narrow and parallel sided (Sushkin, 1936; Bystrow, 1938) (Figure 52.4). A broadly truncate parasternal process (INTCLA 8) is diagnostic of members of the Chigutisauridae (Figure 52.6) such as Siderops kehli (Warren and Hutchinson, 1983), Pelorocephalus spp. (Marsicano, 1993), and Koolasuchus cleelandi (Warren et al., 1997). Plesiomorphically, the center of ossification of the interclavicle (INTCLA 9) is located anterior to the posterior border of the clavicular facets (e.g. Figure 52.2). In all taxa within the Stereospondylomorpha, it is located posterior to the line of maximum width (Figure 52). On the dorsal surface, the interclavicle is featureless except for the broad ridges of the lateral and sternal trabeculae. In Trimerorhachis insignis (Pawley, in press), the sternal trabecula is a narrow, steep sided ridge. In plesiomorphic temnospondyls, the clavicular blades do not contact anteriorly in articulation (Figure 52.4). An anterior extension of the ornamented main body of the interclavicle, separating the clavicular blades, is typically absent in most cases where the clavicular blades do not contact (Figure 52.2). Clavicular blades that broadly contact anterior to the main body of the interclavicle (CLA 3) are characteristic of the Superstereospondyli (Figure 52.6).

Clavicle

The clavicle of temnospondyls (Figure 53) consists of a dorsal clavicular process set onto the posterolateral corner of an ornamented ventral blade. The dorsal clavicular process is composed of a thickened anterior edge (clavicular rod) and an unormented posterior lamina (Figure 53.1.2). The ventral blade varies in the degree of anterior expansion. In more terrestrial taxa that lack lateral line sulci on their skulls, the blade is typically anteroposteriorly narrow, as in the euskelians Eryops megacephalus (Pawley and Warren, 2006) (Figure 53.1) and Micropholis stowi (Watson, 1913; Broili and Schröder, 1937; Schoch and Rubidge, 2005). Temnospondyls with lateral line sulci on their skulls, as in most taxa within the Dvinosauria and Stereospondyli, typically possess anteriorly expanded clavicular blades which are anteroposteriorly deeper than wide (Figure 53.2). The medial edges of the clavicles taper out into a narrow, thin film of bone, usually broken off in disarticulated specimens (Pawley and Warren, 2004). The medial edges of the clavicle may be straight or sinuous, with an indentation directly medial to the base of the dorsal clavicular process. In the plesiomorphic temnospondyls Balanerpeton woodi (Milner and Sequeira, 1994) and Dendrerpeton acadianum (Holmes et al., 1998), the clavicular blade is approximately as long as it is wide, and the dorsal clavicular process is set onto the clavicular blade at approximately a right angle. In derived taxa within the Euskelia, the dorsal clavicular process is smoothly set onto the clavicular blade at much greater than a right angle, so that the whole clavicle forms a rounded arc (Figure 53.1). 172 K. PAWLEY PHD THESIS

Typically, the dorsal process is the same length as the ventral blade, measured from the centre of ossification at the junction of the dorsal clavicular process and ventral blade, in some members of the Dvinosauria, such as Trimerorhachis insignis (Case, 1935; Pawley, in press), and Thabanchuia oomie (Warren, 1998b), the dorsal process is much shorter then the blade. An anterior clavicular flange (m CLA 5) is present on the dorsal clavicular process of some taxa within the Dvinosauria [Trimerorhachis insignis (Pawley, in press) and

FIGURE 53. Temnospondyl clavicle. 1, left clavicle of Eryops megacephalus in 1.1 ventral view, and 1.2 posterior views, from Pawley and Warren (2006). 2, left clavicle of Trimerorhachis insignis in ventral view, from Pawley (in press). 3, left clavicle of Eryosuchus pronus in lateral view, after Howie (1970). 4, left clavicle of Dutuitosaurus ouazzoui in posterior view, after Dutuit (1976). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 173

Thabanchuia oomie (Warren, 1998b)] and basal Superstereospondyli (Figure 53.3). The convergent development of this feature in two clades of obligatorily aquatic taxa, implies that the anterior clavicular flange is an aquatic adaptation. A clavicular groove (CLA 7) (Figure 53.1.1) is plesiomorphically present along the anterolateral surface of the clavicle; it is absent in the Dvinosauria where this characteristic is observable [Trimerorhachis insignis (Pawley, in press) and Thabanchuia oomie (Warren, 1998b)], and in some basal stereospondyls [Lydekkerina huxleyi (Pawley and Warren, 2005), mastodonsaurs (e.g. Bystrow and Efremov, 1940; Howie, 1970; Schoch, 1999a), and metoposaurs (e.g. Sawin, 1945; Dutuit, 1976)]. Note that the clavicular groove may not be obvious in figures, so its presence or absence may be difficult to determine without examination of specimens. In most temnospondyls, the clavicular blade is uniformly ornamented, in the euskelians Eryops megacephalus (Pawley and Warren, 2006), Parioxys ferricolus (Moustafa, 1955a), Ecolsonia cutlerensis (Berman et al., 1985), Anconastes vesperus (Berman et al., 1987a), and Micropholis stowi (Schoch and Rubidge, 2005), a recessed, unornamented area is present, which widens medially as it extends along the posterior surface of the clavicular blade (Figure 53.1.1). A clavicular recess (CLA 8) indents the posterior lamina in derived taxa within the Euskelia, such as Eryops megacephalus (Pawley and Warren, 2006), Ecolsonia cutlerensis (Berman et al., 1985), and Dissorophus multicinctus AMNH 4343 (Figure 53.1.1). A shelf-like, anteroposteriorly-oriented keel along the anteroventral margin of the dorsal clavicular process, separating the ornamented ventral surface from the unornamented dorsal clavicular process, is autapomorphic for the Metoposauroidea (CLA 4) (Figure 53.4). Warren and Marsicano (2000) described the more subtle morphological details that distinguish brachyopid clavicles from those of chigutisaurs.

Anocleithrum

Anocleithra are unknown within the Temnospondyli.

Cleithrum

The cleithrum of temnospondyls (Figure 54) typically consists of a long cleithral shaft capped by a posteriorly expanded dorsal cleithral process. The ventral edge of the thin, laterally flattened, dorsal cleithral process is grooved for articulation with the anterodorsal edge of the scapular blade. In cross section, the cleithral shaft is teardrop shaped, the smoothly rounded anterior edge tapering to a thin flange (scapular flange of the cleithral shaft), which articulates with the scapular blade. When preserved in articulation, in temnospondyls the cleithrum is always located between the anterior edge of the scapular blade and the dorsal clavicular process of the clavicle, which covers the ventral portion of the cleithral shaft (Figure 51). Dorsally, the cleithrum caps the anterodorsal corner of the scapular blade, the ventral tip of the cleithral shaft extends to the level of the glenoid on the scapulocoracoid. The anterodorsal surface of the dorsal cleithral process bears a sharp-edged cleithral crest (CLE 7) in plesiomorphic temnospondyls (Figure 54.1). In stereospondyls the cleithral crest is characteristically highly reduced or absent (Figure 54.3, Figure 54.4). Plesiomorphically, the ventral junction of the dorsal cleithral process and the cleithral shaft is smoothly rounded or grooved. In all taxa within the Eryopoidomorpha, a suprascapula lamina (CLE 8) (Figure 54.1), is present on the lateroventral surface of the 174 K. PAWLEY PHD THESIS

FIGURE 54. Temnospondyl cleithrum. 1, left cleithrum of Eryops megacephalus in 1.1 lateral, and 1.2 medial views, from Pawley (in press). left pectoral girdle of Cacops aspidephorus in lateral view, after Williston (1910a). 3, left cleithrum of unnamed rhinesuchid in lateral view, from Pawley and Warren (2004). 4, left cleithrum of Benthosuchus sushkini in 4.1 lateral, and 4.2 medial views, after Bystrow and Efremov (1940). 5, left cleithrum of Gerrothorax pulcherrimus in 5.1 lateral, and 5.2 medial views, 5.3 articulated cleithrum and clavicle of plagiosaur in left lateral view, after Shishkin (1987). Scale bar = 25mm CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 175 dorsal cleithral process, clasping the anterodorsal surface of scapular blade. A low, sharp- edged lateral cleithral ridge is present across the junction of the dorsal cleithral process and shaft in basal stereospondyls [rhinesuchids (Pawley and Warren, 2004), Lydekkerina huxleyi (Pawley and Warren, 2005), Mastodonsaurus giganteus (Schoch, 1999a), Cyclotosaurus intermedius (Sulej and Majer, 2005), Eryosuchus pronus (Howie, 1970), and Dutuitosaurus ouazzoui) (Dutuit, 1976)] (Figure 54.3). The cleithrum of some plagiosaurs (Figure 54.5) is unique among tetrapods because the external surface is ornamented. The ornament is pustular like that of the other dermal bones. The cleithrum of such plagiosaurs is also peculiar in its dorsoventrally compressed form and close articulation with the dorsal process of the clavicle.

Scapulocoracoid

The scapulocoracoid of temnospondyls (Figure 55) is subject to extensive morphogenetic development and is most commonly present in a morphogenetically immature stage (Chapter 4) (Figure 55.2, Figure 55.3), where only the scapular blade is present. In morphogenetically mature specimens (Figure 55.1), a large, dorsally expanded scapular blade is present, with a ventrally expanded coracoid typically set at.slightly greater than a right angle, with no visible sutures between the scapula and coracoid regions. In lateral view (Figure 55.1.1), the scapular portion is dorsally flared and narrowest just dorsal to the supraglenoid fossa. Even in the latest morphogenetic stages, the dorsal edge of the scapula is unossified, indicating the presence of a cartilaginous extension. The lateral supraglenoid ridge passes between the posterior border of the scapular blade and the posterodorsal margin of the glenoid, enclosing the supraglenoid fossa. In medial view (Figure 55.1.2), the scapular torus passes from the anterodorsal corner of the scapular blade to the supraglenoid buttress, which widens into the ventral infraglenoid buttress from above the glenoid. The anterior face of the supraglenoid buttress is smoothly rounded and dorsoventrally oriented, concealing the openings for the supraglenoid and coracoid foramina within the deep subscapular fossa All morphogenetically mature temnospondyl specimens possess three scapulocoracoid foramina, including two foramina on the coracoid (Figure 55.1.1). The coracoid foramen is located directly ventral the anterior edge of the glenoid, and the glenoid foramen is located directly ventral to the midpoint of the glenoid. The paths of these three foramina through the scapulocoracoid are similar to those observed in other early tetrapods. The external opening of the supraglenoid foramen is located in the supraglenoid fossa; the internal opening is located in the upper part of the subscapular fossa. The external openings of the coracoid foramen passes dorsally to the mid portion of the subscapular fossa, and the glenoid foramen passes to the middle of the supraglenoid buttress. In lateral view (Figure 55.1.1), the glenoid arches medially above the infraglenoid fossa, the anterior part of the glenoid is oriented posteriorly, and the posteroventral surface is oriented dorsally. Within the infraglenoid fossa, an infraglenoid recess (sensu Pawley and Warren, 2005) is present under the midpoint of the glenoid, lateral to a low infraglenoid ridge (sensu Pawley and Warren, 2005) which in turn is medial to the coracoid and glenoid foramina. Taxa in which these characteristics can be observed include Peltobatrachus pustulatus (Panchen, 1959), the euskelians Eryops megacephalus (Cope, 1884; Pawley and Warren, 2006), Acheloma cumminsi (Williston, 1909a; Olson, 1941)and Dissorophus multicinctus MCZ 4194, Sclerocephalus haeuseri (Meckert, 1993), the archegosaur Australerpeton cosgriffi (Dias and Schultz, 2003), and the basal stereospondyl Lydekkerina huxleyi (Pawley and Warren, 2005). Dendrerpeton acadianum 176 K. PAWLEY PHD THESIS

FIGURE 55. Temnospondyl scapulocoracoid. 1, left scapulocoracoid of Eryops megacephalus in 1.1 lateral, and 1.2 medial views, from Pawley and Warren (2006). 2, left scapula of Trimerorhachis insignis in 2.1 lateral, and 2.2 posterior views, from Pawley (in press). 3, left scapula of Benthosuchus sushkini in 3.1 lateral, and 3.2 posterior views, after Bystrow and Efremov (1940). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 177

(Holmes et al., 1998) was restored with only a coracoid foramen, however as noted by Carroll (1967) and Holmes et al. (1998), the area of the glenoid foramen is either not preserved or covered by gastralia in all specimens, so the presence or absence of a glenoid foramen is not determinable. A scapular tubercle (m SCACOR 9) (Figure 55.2) is present on the supraglenoid buttress of taxa within the Dvinosauria [Trimerorhachis insignis (Pawley, in press) and Dvinosaurus spp. (Nikitin, 1997)]. Plesiomorphically, the supraglenoid foramen of the scapulocoracoid is enclosed, and a ventrally open supraglenoid foramen (m SCACOR 5) (Figure 55.3), observable in at least in early morphogenetic stages, is characteristic of the Superstereospondyli.

Humerus

The humerus of temnospondyls (Figure 56) is typically of the shape described by Romer (1922), as tetrahedral, with two expanded ends set at approximately right angles to each other. In some temnospondyl taxa, the degree of torsion is phenotypically variable, with some specimens having the ends of the humerus at set at right angles; in others, the degree of torsion is much lower. Variation in the degree of humeral torsion occurs in obligatorily aquatic temnospondyls such as Trimerorhachis insignis (Pawley, in press), Dvinosaurus spp. (Nikitin, 1995), Buettneria perfecta (Olsen, 1951), and Mastodonsaurus giganteus (Schoch, 1999a). As in the scapulocoracoid, the humerus of temnospondyls is subject to extensive morphogenetic development and is most commonly present in a morphogenetically immature stage (Chapter 4), the following description applies to morphogenetically mature taxa. Typically, the humerus of temnospondyls is stout, with the proximal and distal ends approximately half the width of the length of the humerus (Figure 56.1). A slender and elongate humerus (HUM 12) that is more than three times longer than the diameter of its distal end is characteristic of the derived amphibamids Platyrhinops lyelli (Carroll, 1964a), Amphibamus grandiceps (Watson, 1940; Gregory, 1950; Daly, 1994), and Doleserpeton annectans (Gregory et al., 1956) (Figure 56.4). The large ectepicondyle and entepicondyle are the most prominent features of the extensor surface (Figure 56.1.1), the ectepicondyle is typically a steep-sided crest, projecting anteriorly beyond the radial condyle (Figure 56.1.2). The smoothly rounded posterior border of the entepicondyle is not perforated by an entepicondylar foramen (HUM 5), except in plesiomorphic taxa such as Casineria kiddi (Paton et al., 1999), Dendrerpeton acadianum (Carroll, 1967; Holmes et al., 1998) (Figure 56.4) and Balanerpeton woodi (Milner and Sequeira, 1994). An isolated humerus assigned to Edops craigi (Romer and Witter, 1942) lacks an entepicondylar foramen, which may imply the loss of this feature more than once within the Temnospondyli. Alternatively, the assignation of the humerus to Edops craigi, or the phylogenetic position of Edops craigi in this analysis may be incorrect. One specimen of Dvinosaurus campbelli (Gubin, 2004) also has an entepicondylar foramen in an unusual position, close to the proximal border of the entepicondyle. On the anterior side of the extensor surface (Figure 56.1.1), the rounded ectepicondylar ridge extends distally from the head of the humerus to the proximal portion of the ectepicondyle. On the posterior side, the entepicondylar ridge passes from below the proximal articulation surface on the entepicondyle. Between the ectepicondylar and entepicondylar ridges is a shallow groove that originates just below the articulation surface of the caput humeri, and expands distally into the deeply hollowed area between the ectepicondyle and entepicondyle.

178 K. PAWLEY PHD THESIS

FIGURE 56. Temnospondyl humerus. 1, left humerus of Eryops megacephalus in 1.1 flexor, 1.2 extensor, 1.3 anterior, and 1.4 posterior views, from Pawley and Warren (2006). 2, left humerus of Dendrerpeton acadianum in extensor view, after Carroll (1967). 3, left humerus of Trimerorhachis insignis in flexor view, from Pawley (in press). 4, left humeri attributed to Doleserpeton annectans (Bolt, 1969), in 4.1 flexor, and 4.2 flexor and extensor views, after Gregory et al. (1956). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 179

A scapulohumeral ridge (Figure 56.1.3) is present on the proximal anterior surface in Eryops megacephalus (Pawley and Warren, 2006). This feature may be present in more taxa, but it may be passed over as damage. This ridge probably divides the attachment area for the scapulohumeralis muscle from that of the more posteriorly located subcoracoscapularis muscle (after Holmes, 1977). The deltoid crest (Figure 56.1.3) is located on the flexor side of the anterior humeral surface, on the extensor side; the latissimus dorsi process is plesiomorphically a large spike of bone, flattened towards the dorsal surface of the humerus. Unfortunately the condition in the Rhinesuchidae is unknown, but in all more derived stereospondyls the latissimus dorsi process is a small inconspicuous knob or absent (HUM 25). On the flexor surface (Figure 56.1.2), the proximal humeral ridge is sharp-edged and finished in periosteal bone. It joins the proximal articulation surface with the deltopectoral area. Distal to the proximal humeral ridge, the deltopectoral crest separates into two distinct regions, the deltoid crest, and the pectoralis attachment, separated by a cancellous bone surface. In the plesiomorphic condition, the deltoid and pectoral crests of the humerus are more distally located than the latissimus dorsi process, whereas in some members of the Euskelia, [Sclerocephalus haeuseri (Meckert, 1993), Eryops megacephalus (Miner, 1925; Pawley and Warren, 2006), Ecolsonia cutlerensis (Berman et al., 1985), and Phonerpeton pricei (Dilkes, 1990)] the deltoid and pectoral crests of the humerus are located at the same level as the latissimus dorsi process (HUM 21) (Figure 56.1.3). The anterior humeral keel crosses the shaft from the distal side of the deltopectoral crest to the flexor side of the supinator process (Figure 56.1.2). The supinator process is always located proximal to the radial condyle. A distinct supinator process (HUM 2) is present in all morphogenetically mature temnospondyls except the amphibamids Micropholis stowi (Schoch and Rubidge, 2005) and Doleserpeton annectans (Gregory et al., 1956) (Figure 56.4). Where ossified, the radial condyle is always bulbous and located on the ventral surface of the humerus (Figure 56.1.2) and, in distal view, aligned with the ectepicondyle. On the posterior surface of the humerus, the pectoral crest may be highly rugose, as in Trimerorhachis insignis (Pawley, in press) (Figure 56.3), or almost indiscernible, as in Eryops megacephalus (Pawley and Warren, 2006) (Figure 56.1.4). The ventral humeral ridge is always present in temnospondyls (Figure 56.1.4, Figure 56.3), but is often difficult to observe because it is low and obscured by torsion of the humerus.

Radius

The radius of temnospondyls (Figure 57.1) is a columnar element, with gently flared proximal and distal ends. The extensor surface is convex in transverse section. The ventromesial radial ridge (sensu Warren and Ptasznik, 2002) passes down the midline of the anterior surface (Figure 57.1.2). On the flexor surface (Figure 57.1.3), the ventral radial crest is located slightly medial to the posterior edge. Originating well below the expanded humeral articulation surface, it extends into the distal border medial to the posterior edge. The shaft is concave in transverse section, a broad low ridge, the radial flexor ridge (Figure 57.1.3), originates below the midpoint of the shaft, and increases in depth distally. The humeral articulation surface is typically ovoid or subrectangular. The distal articulation surface is flattened between the extensor and flexor surfaces; in morphogenetically mature specimens, such as Eryops megacephalus (Pawley and Warren, 2006), there are three convex articulation facets for the radiale, centrale, and intermedium (Figure 57.1.3). There is no obvious phylogenetic variation in the radii of temnospondyls. 180 K. PAWLEY PHD THESIS

FIGURE 57. Temnospondyl radius, ulna, and manus. 1, left radius of Eryops megacephalus in 1.1, extensor, 1.2, anterior, 1.3 flexor, 1.4, posterior views, from Pawley and Warren (2006). 2, left ulna of Eryops megacephalus in 2.1 anterior and 2.2 posterior views, from Pawley and Warren (2006). 3, left antebrachium of Eryops megacephalus in anterolateral view, revised after Gregory et al. (1923). 4, phalanges of Trimerorhachis insignis in 4.1 extensor and 4.2 flexor views, from Pawley (in press). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 181

Ulna

The ulna (Figure 57.2) is always relatively slender when compared with the radius, is anteroposteriorly flattened, and exceeds the radius by the length of the olecranon process. The most prominent feature is the proximal portion of the ulnar extensor keel, which is typically highly rugose for the attachment to the triceps muscle narrowing distally into a steep keel, which terminates at the distal articulation surface. The most prominent feature of the posterior surface (Figure 57.2.2) is the low posterior ulnar crest, which originates below the posterior humeral articulation surface and curves distally down the midline of the posterior surface. Although the condition in most temnospondyls is not determinable, a posterolateral ulnar crest (ULNA 2) (Figure 57.2.2) is present branching out at the midpoint of the shaft from the posterior ulnar crest, in the euskelians Eryops megacephalus (Pawley and Warren, 2006), Parioxys spp. (Moustafa, 1955a; Carroll, 1964b), Phonerpeton pricei (Dilkes, 1990), Ecolsonia cutlerensis (Berman et al., 1985); and the superstereospondyls Benthosuchus sushkini (Bystrow and Efremov, 1940) and Dutuitosaurus ouazzoui (Dutuit, 1976). In proximal view, the humeral articulation surface is approximately quadrangular. In distal view, in morphogenetically mature specimens, such as Eryops megacephalus (Pawley and Warren, 2006) (Figure 57.2) and Ecolsonia cutlerensis (Berman et al., 1985), the laterally flattened articulation surface carries posterior and anterior articulation surfaces for the ulnare and intermedium, respectively.

Manus

The completely ossified manus (Figure 57.3) is rarely present in temnospondyls, as the carpal elements are one of the last parts of the postcranial skeleton to ossify during morphogenesis (Chapter 4), and often only a few are ossified. A full complement of carpal elements is only known in the euskelians Eryops megacephalus (Gregory et al., 1923) and Acheloma cumminsi (Williston, 1909a; Olson, 1941). In both these taxa, there are eleven carpals and four digits. Gregory et al. (1923) originally restored the manus of Eryops megacephalus with five digits, which was later revised by Miner (1925) to four, as there is no conclusive evidence of a fifth digit. The basal emnospondyl Casineria kiddi (Paton et al., 1999) possesses five digits, a plesiomorphic feature of early tetrapods which is not unexpected in a basal temnospondyl. All other temnospondyls in which the manus is well preserved possess only four digits in the forelimb. There are eleven carpal elements in total, consisting of a radiale, intermedium, ulnare, four centrales, and four distal carpals. The radiale, intermedium, and proximal centrale (centrale four) contact the radius; the intermedium and ulnare contact the ulna. The manus of temnospondyls was medially arched in life (Gregory et al., 1923; Miner, 1925).

Digits

A phalangeal formula of 2-2-3-3 is observable in the manus of Balanerpeton woodi (Milner and Sequeira, 1994), Platyrhinops lyelli (Carroll, 1964a), Amphibamus grandiceps (Gregory, 1950), Eoscopus lockardi (Daly, 1994), Branchiosaurus petrolei (Boy, 1972), Micromelanerpeton credneri (Boy, 1972), Sclerocephalus haeuseri (Meckert, 1993), and Uranocentrodon senekalensis (QR 1438). The phalanges of temnospondyls ossify prior to the carpal elements (Chapter 4), and consequently are more frequently preserved. Metacarpals are identifiable by their angled proximal articulation surface (in well ossified specimens). Individual phalangeal elements are swollen at the 182 K. PAWLEY PHD THESIS

FIGURE 58. Temnospondyl pelvic girdle. 1, left pelvis of Dendrerpeton acadianum in 1.1 lateral, and 1.2 medial views, after Carroll (1967). 2, left pelvis of Eryops megacephalus in 2.1 lateral, and 2.2 medial views, from Pawley and Warren (2006). 3, left ilium of Trimerorhachis insignis in medial view, from Pawley (in press). 4, left ilium of Eryosuchus pronus in medial view, after Howie (1970). 5, left ilium of Buettneria perfecta in left lateral views, after Sawin (1945) Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 183 proximal and distal ends, prominently waisted, and flattened between the extensor and flexor surfaces (Figure 57.4). Deep notches are present on the proximal and distal ends of the flexor surfaces for the flexor tendons in Casineria kiddi (Paton et al., 1999), Trimerorhachis insignis (Pawley, in press) and Eryops megacephalus (e.g. AMNH 4203). All digits terminate in a bluntly pointed, but never clawed, ungual phalanx that is proximally slightly broader than the articulated phalange.

Pelvic girdle

Typically, the dorsal iliac blade is flattened laterally and anteroposteriorly expanded. A post-iliac process (ILI 2) (Figure 58.1) is retained in the plesiomorphic temnospondyls Caerorhachis bairdi (Holmes and Carroll, 1977; Ruta et al., 2002), Casineria kiddi (Paton et al., 1999), Dendrerpeton acadianum (Carroll, 1967; Holmes et al., 1998), and Balanerpeton woodi (Milner and Sequeira, 1994). The clearly defined facet for the sacral rib on the medial surface of the dorsal iliac process (Figure 58.1.2, Figure 58.2.2) has a smooth surface, showing little evidence for muscle or ligament attachment. The area of sacral rib attachment is longitudinally grooved in Trimerorhachis insignis (Pawley, in press) (Figure 58.3), but the lateral surface is also grooved in the same fashion, so this characteristic does not indicate a strong attachment for the sacral rib. The transverse pelvic ridge is variably present in temnospondyls, either as a long ridge extending from the anterior side of the supracetabular buttress to the ventral edge of the dorsal iliac process (Figure 58.1.1), or as a short crest not extending posterior to the supracetabular buttress (Figure 58.2.1). The posteroventral border of the iliac blade extends into a sharp-edged sacral flange, which gives the iliac blade a rectangular outline in ventrolateral view in the euskelians Eryops megacephalus (Cope, 1884; Pawley and Warren, 2006) (Figure 58.2) and Platyrhinops lyelli (Carroll, 1964a; Hook and Baird, 1984). The ilium of trematosaurs is comparatively elongate and narrow (Maisch et al., 2004; Schoch, 2006), and in metoposaurs, the ilium is cylindrical in cross section (Figure 58.5). The mesial iliac ridge (ILI 8) plesiomorphically passes down the midline of the ilium (Figure 58.1.2, Figure 58.3), so that there is a recess present both anterior and posterior to the ridge. In all taxa within the Eryopoidomorpha, the mesial iliac ridge forms the anterior border of the ilium, with no anterior recess (Figure 58.2.2, Figure 58.4), and the dorsal opening of the obturator foramen opens onto the anterior surface of the pubis. The single obturator foramen typical of temnospondyls opens close to the ventral border of the acetabulum (Figure 58.1.1, Figure 58.2.1).

Femur

The femur of temnospondyls (Figure 59) is always longer than the humerus, with a relatively narrow, anteroposteriorly flattened shaft, and expanded ends. In dorsal view, the shaft is narrowest directly above the adductor blade but flares out distally towards the tibial and fibular condyles. In extensor view (Figure 59.1.1), the posterior caput femoral fossa occupies the proximoposterior surface of the extensor side of the femoral head. A small ridge may be present distal to this fossa, and is presumably the attachment for the ischiotrochantericus muscle (Romer, 1922). A sharp-edged intercondylar crest may be present on the proximal side of the intercondylar fossa, as in Eryops megacephalus (Cope, 1884; Pawley and Warren, 2006).In flexor view (Figure 59.1.2), the adductor blade borders the anterior side of the intertrochanteric fossa, and the posterior intertrochanteric ridge the posterior side, 184 K. PAWLEY PHD THESIS with the two meeting at the proximal end of the adductor crest. The ventrally oriented adductor blade has internal and fourth trochanters on the ventral surface. In morphogenetically mature specimens, a trough of periosteal bone separates the internal trochanter from the cancellous area of the anterior caput femoral fossa. The boss of the internal trochanter is always shallowly concave and filled with cancellous bone, and the surface of the fourth trochanter may be rugose or tuberculate. In

FIGURE 59. Temnospondyl femur. 1, left femur of Eryops megacephalus in 1.1 extensor, and 1.2 flexor views, from Pawley and Warren (2006). 2, left femur of Dendrerpeton acadianum in flexor view, after Carroll (1967). 3, left femur of Trimerorhachis insignis in flexor view, from Pawley (in press). 4, left femur of Acheloma cumminsi in flexor view, after Sullivan et al. (2000). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 185

Acheloma cumminsi (Williston, 1909a; Olson, 1941; Sullivan et al., 2000), the fourth trochanter is distinctively flattened and concave (Figure 59.1.4). The adductor crest is deepest and broadest immediately posterior to the junction of the posterior intertrochanteric ridge and adductor blade, it reduces in height as it passes distally to the fibular condyle. In the plesiomorphic state, the adductor crest is as deep as the femoral shaft. Characteristically in the Dvinosauria the adductor crest of the femur is reduced to a low ridge (m FEM 6), as in Neldasaurus wrightae (Chase, 1965), Trimerorhachis insignis (Case, 1911a; Pawley, in press) (Figure 59.3), and Dvinosaurus spp. (Sushkin, 1936; Bystrow, 1938; Gubin, 2004). In plesiomorphic temnospondyls, the adductor crest is aligned with the adductor blade, and runs diagonally across the femoral shaft to the fibular condyle (Figure 59.2). Within the Eryopoidomorpha, the adductor blade is set at an angle to the adductor crest, and passes down the midline of the flexor surface of the femur (FEM 8) (Figure 59.1.2), until it branches to form the anterior popliteal ridge on the proximal side of the popliteal area. The fibular fossa on the posterior surface of the femur is shallowly recessed and much narrower than the popliteal area. In Trimerorhachis insignis (Pawley, in press), a short rugose ridge, the femoral fibular ridge, forms the dorsal border of the narrow fibula fossa.

Tibia

In all temnospondyls, the tibia (Figure 60) is a wedge shaped element with an expanded proximal articulation surface tapering to a cylindrical shaft. On the flexor surface (Figure 60.1.2), the tibial flexor crests form a ‘Y’ shape down the midline. The anterior tibial flexor crest and the posterior tibial flexor crest originate in the upper third of the flexor surface and converge towards the distal end to form the distal tibial flexor crest. There may be low ridges (anterior and posterior tibial ridges) on the anterior and posterior surfaces of the shaft, as in Trimerorhachis insignis (Pawley, in press) (Figure 60.3), but the phylogenetic significance of these features is difficult to assess, as few temnospondyl tibia have been described in sufficient detail to be able to determine their presence or absence. The distal articulation surface (Figure 60.1) is ovoid and strongly convex, with continuous facets for articulation with the tibiale and intermedium. No obvious phylogenetic variation is observable in the tibia within the Temnospondyli.

Fibula

The fibula of temnospondyls (Figure 60) is a relatively slender, laterally flattened element, approximately as long as the tibia. The fibula is always waisted, with the anterior border deeply curved around the interepipodial space (Figure 60.2). A short ridge, the anterior fibular ridge, is located on the anterior proximal surface (Figure 60.5), so that the proximal head of the fibula has a rectangular outline in extensor view.The anterior fibular ridge is probably the insertion area for the medial ligament to the tibia (Romer, 1922). The flexor surface of the shaft is plesiomorphically convex along its length (Figure 60.4). The mid shaft area of the flexor surface of fibula is concave (FIB 6) in all taxa within the Eryopoidomorpha (Figure 60.5) where the condition is determinable. In lateral view, the posterior border of the fibular shaft between proximal and distal ends is plesiomorphically concave, as in Trimerorhachis insignis (Pawley, in press) (Figure 60.4). In all euskelians, it is convex (FIB 5) (Figure 60.5), as in Sclerocephalus haeuseri (Lohmann and Sachs, 2001), Onchiodon (Sclerocephalus) labyrinthicus (Credner, 1893), Eryops megacephalus (Case, 1911a; Pawley and Warren, 2006), 186 K. PAWLEY PHD THESIS

FIGURE 60. Temnospondyl tibia, fibula, and pes. 1, left tibia of Eryops megacephalus in 1.1 extensor, and 1.2 flexor views, from Pawley and Warren (2006). 2, left tibia, fibula and pes of Acheloma cumminsi in lateral view, modified (digits restored) after Schaeffer (1941). 3, left tibia of Trimerorhachis insignis in 3.1, extensor; 3.2, anterior; 3.3, flexor; 3.4, posterior views, from Pawley (in press). 4, left fibula of Trimerorhachis insignis in flexor view, from Pawley (in press). 5, left fibula of Eryops megacephalus in 5.1 extensor, and 5.2 flexor views, from Pawley and Warren (2006). Scale bar = 25mm. CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 187

Parioxys bolli (Carroll, 1964b), Ecolsonia cutlerensis (Berman et al., 1985), Dissorophus multicinctus AMNH 4343, Cacops aspidephorus (Williston, 1910a), Eoscopus lockardi (Daly, 1994), Platyrhinops lyelli (Carroll, 1964a; Hook and Baird, 1984), and Doleserpeton annectans (Gregory et al., 1956). A posterior fibular ridge passes down the posterior edge of the flexor surface; this ridge may be tuberculate in morphogenetically mature specimens of taxa such as Eryops megacephalus (Pawley and Warren, 2006) (Figure 60.5.2). In temnospondyls, a fibular sulcus is always present, across the anterodistal corner of the flexor surface. The femoral articulation surface meets the extensor surface of the bone at a steep angle in morphogenetically mature specimens (Figure 60.5.2). The distal end of the fibula is approximately twice as wide as the proximal end in plesiomorphic temnospondyls such as Dendrerpeton acadianum (Carroll, 1967), Balanerpeton woodi (Milner and Sequeira, 1994), and Trimerorhachis insignis (Pawley, in press) (Figure 60.4). Within the Euskelia, the distal end of the fibula is subequal to the proximal end rather than expanded (Figure 60.5), as in Eryops megacephalus (Case, 1911a; Pawley and Warren, 2006), Dissorophus multicinctus AMNH 4343, Ecolsonia cutlerensis (Berman et al., 1985), and the amphibamid Eoscopus lockardi (Daly, 1994). The distal articulation surface is bipartite, the medial side is convex for articulation with the intermedium, and the lateral portion is concave for articulation with the fibulare.

Pes

The complete tarsus (Figure 60.2) is rarely present in temnospondyls, as the tarsal bones are one of the last parts of the postcranial skeleton to ossify during morphogenesis (Chapter 4), and often only a few tarsal elements are ossified. A full complement of tarsal bones is only known in euskelian taxa such as Eryops megacephalus (MCZ 7555), Acheloma cumminsi (Williston, 1909a; Schaeffer, 1941), Eoscopus lockardi (Daly, 1994), and Tersomius cf. T. texensis (Daly, 1994). There are twelve tarsal bones in total, consisting of a tibiale, intermedium, fibulare, four centrales, and five distal tarsals. The tibiale, intermedium, and proximal centrale (centrale four) contact the tibia; the intermedium and fibulare contact the fibula. In Eryops megacephalus (MCZ 7555), a notch is present on the abutting ventral surfaces of the intermedium and fibulare for the perforating artery, and a groove crosses the ventral surface of the fibulare, again for the perforating artery. The plantar surface of the tarsus was medially arched in life (Daly, 1994). An intermedium with an ossified proximal edge is only known in the euskelians Eryops megacephalus (MCZ 7555), Acheloma cumminsi (Williston, 1909a; Schaeffer, 1941), and Eoscopus lockardi (Daly, 1994). In these taxa the proximal edge is notched but the intermedium is not ‘L’ shaped.

Digits

All temnospondyls in which the pes is well preserved possess five digits in the hind limb (Figure 60.2). The phalangeal formula of 2-2-3-4-3 (Figure 60.3) can be observed in Balanerpeton woodi (Milner and Sequeira, 1994), Amphibamus grandiceps (Daly, 1994), Eoscopus lockardi (Daly, 1994), Branchiosaurus petrolei (Boy, 1972), Micromelanerpeton credneri (Boy, 1972), Sclerocephalus haeuseri (Lohmann and Sachs, 2001; Schoch, 2003), and Uranocentrodon senekalensis (QR 1438). Individual phalangeal elements are not distinguishable from those of the manus (Figure 60.4). 188 K. PAWLEY PHD THESIS

Lateral lines

Evidence for lateral lines can be found in the form of pits and perforations on the ventral most dorsal scales of Dendrerpeton acadianum (Dawson, 1882) and the derived stereospondyl Gerrothorax pustuloglomeratus (Hellrung, 2003). Werneburg (2004b) found evidence of denticulate lateral lines in branchiosaurs. What can be observed of the denticles indicates that they are small and irregularly shaped, most similar to the calcified denticles in the lateral lines of orthacanthid sharks. The presence of lateral lines in both the most plesiomorphic and the most derived temnospondyls indicates that lateral lines may be widespread in temnospondyls.

Dermal ossifications

Three general types of dermal ossification are present in temnospondyls (Figure 61): cycloid dorsal scales, ventral scutes, and osteoderms, including paramedian osteoderms. Formal definitions are provided for the first time for each ossification type, as there is considerable variation in terminology in the literature. These definitions distinguish particular features of temnospondyls, and maintain consistency with definitions used in recent work on dermal ossifications in other tetrapod groups (e.g. Boy and Sues, 2000; Salisbury and Frey, 2000; Hill, 2005). Dermal ossifications are affected by morphogenetic development, as discussed in Chapter 4. Presence or absence of dermal ossifications is often difficult to determine unless specimens are exceptionally well preserved. Cycloid dorsal scales are not easily distinguished from those of fish and are often undescribed or removed without record during preparation.

Scales

Cycloid dorsal scales (Figure 61.1) are small, ovoid, plate-like structures with obvious growth rings, similar in morphology to the scales of fish, and covering the dorsal surface of the trunk and tail. Overlapping scales (Figure 61.1.2) are found in plesiomorphic temnospondyls such as Dendrerpeton acadianum (Dawson, 1882), Capetus palustris (Sequeira and Milner, 1993), Neldasaurus wrightae (Chase, 1965), Trimerorhachis insignis (Colbert, 1955), Leptorophus tener (Boy, 1986), Eryops megacephalus (Romer and Witter, 1941), Eoscopus lockardi (Daly, 1994), and Amphibamus grandiceps (Daly, 1994). Non-overlapping scales (DERM 2) are characteristic of the Stereospondylomorpha, including Archegosaurus decheni (Meyer, 1857), Uranocentrodon TM 75, TM 103, Paracyclotosaurus davidi (Watson, 1958), and Trematosaurus madagascariensis (Janvier, 1992).

Ventral scutes

Ventral scutes (Figure 61.2) are spindle shaped with a ridge along the centre (Figure 61.2.1). Growth rings are frequently observable, which indicates that ventral scutes are modified scales (Dias and Richter, 2002). Scutes are arranged in overlapping rows on the abdomen, with a specialised expanded overlap scute where the rows join in the ventral midline. The rows of scutes are articulated in an ‘X’ shape between the limbs, with the anterior most scutes forming a chevron between the interclavicle and humerus, and the majority of the rows of scutes forming an inverted posterior chevron (Figure 61.2.2). Together the dermal pectoral girdle and ventral scutes form a thick plastron that covers the ventral trunk between the limbs. Examples of ventral scutes in temnospondyls include CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 189

FIGURE 61. Temnospondyl dermal ossifications. 1, Trimerorhachis insignis, after Colbert (1955). 2.1, Uranocentrodon senekalensis, after Findlay (1968), 2.2 unnamed rhinesuchid in ventral view, from Pawley and Warren (2004). 3, Cacops aspidephorus, 3.1 articulated neural spines and paramedian osteoderms in left lateral view, after de Mar (1966), 3.2 neural spine and articulated paramedian osteoderms in anterior view, 3.3 articulated paramedian osteoderms in dorsal, and 3.4, in ventral views, after Williston (1910a). 4, Broiliellus brevis articulated neural spines and paramedian osteoderms in 4.1 dorsal, and 4.2 left lateral views, after de Mar (1966). 5, Dissorophus multicinctus, 5.1 complete carapace in dorsal view, after de Mar (1968), 5.2 articulated neural spines and paramedian osteoderms in left lateral view, and 5.3 single paramedian osteoderm in anterior view, after de Mar (1966). 6, Peltobatrachus pustulatus in dorsal view, after Panchen (1959). Scale bar = 25mm. 190 K. PAWLEY PHD THESIS

Balanerpeton woodi (Milner and Sequeira, 1994), Trimerorhachis insignis (Colbert, 1955), Peltobatrachus pustulatus (Panchen, 1959), Amphibamus grandiceps (Watson, 1940; Daly, 1994), Platyrhinops lyelli (Hook and Baird, 1984), Eryops megacephalus (Romer and Witter, 1941), Cheliderpeton vranyi (Werneburg and Steyer, 2002), Sclerocephalus haeuseri (Broili, 1926; Boy, 1988), Archegosaurus decheni (Meyer, 1857), rhinesuchids, Lydekkerina huxleyi (Pawley and Warren, 2005), Laidleria gracilis (Kitching, 1957), and Gerrothorax spp (Nilsson, 1946a; Hellrung, 2003). In Gerrothorax spp., the ventral scutes are surfaced in pustular ornament, like the dermal cranial and postcranial elements. The broad phylogenetic distribution of ventral scutes in temnospondyls strongly implies that their presence is ubiquitous.

Dermal osteoderms

Osteoderms (DERM 1) are thickened, ornamented plates of dermal bone with a vascularized internal structure, lacking obvious growth rings (Figure 61). The thickness and ornamentation type of the osteoderms is similar to that of dermal bone of the cranium and dermal pectoral girdle. In temnospondyls they are located on the dorsal surface of the trunk and tail, and vary in size from small, non articulating osteoderms in Edops craigi (Romer and Witter, 1942) and Trimerorhachis insignis (Olson, 1979) to larger articulating mosaics of osteoderms in the euskelian Ecolsonia cutlerensis (Berman et al., 1985), and plagiosauroids such as Laidleria gracilis (Kitching, 1957) and Gerrothorax spp. (Nilsson, 1946a; Hellrung, 2003). Smaller dermal osteoderms may occur with dorsal scales, as in Trimerorhachis insignis (Olson, 1979). The distribution of osteoderms in the Temnospondyli is sporadic, and they are only phylogenetic significant at lower taxonomic levels.

Paramedian osteoderms

Enlarged, laterally expanded paramedian osteoderms (new term), articulating with, or fused to the neural spines are found in euskelians, particularly members of the Dissorophidae (de Mar, 1966; Bolt, 1974d), and the unrelated Peltobatrachus pustulatus (Panchen, 1959) (Figure 61). These osteoderms differ from other types of osteoderms in temnospondyls in that they are considerably larger. They are unusual in that they are located along the median axis, articulating with the neural arches. Carroll (1964a), Lewis and Vaughn (1965) and Bolt (1974d) discussed the phylogenetic implications of the morphology of paramedian osteoderms within the Dissorophidae.

Dissorophid type paramedian osteoderms. De Mar (1966) described the various types of paramedian osteoderms associated with the neural spines of taxa within the Dissorophidae. All dissorophid type paramedian osteoderms consist of single ossifications. Cacops-type paramedian osteoderms (Figure 61.3), found in Cacops aspidephorus (Williston, 1910a; de Mar, 1966) and Alegeinosaurus aphthitos (Case, 1911a), also have two paramedian osteoderms per vertebral segment. Cacops-type paramedian osteoderms do not extend much past the lateral borders of the neural spines; in this, they differ from Broiliellus-type and Dissorophus-type paramedian osteoderms. The internal series of Cacops-type paramedian osteoderms lack ventral processes, rather they fuse to the enlarged tips of the neural spines. Aspidosaurus-type paramedian osteoderms are similar to Cacops-type paramedian osteoderms, but have only one osteoderm per vertebral segment, fused to an expanded CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 191 neural spine. Each paramedian osteoderm articulates with the adjacent paramedian osteoderms anteriorly and posteriorly. These are found in Aspidosaurus (Zatrachys) apicaulis (Case, 1907, 1911a), A. binasser (Berman and Lucas, 2003), A. chiton (Broili, 1904; Case, 1911a), A. crucifer (de Mar, 1966), A. glascocki (Case, 1911a), A. novomexicanus (Carroll, 1964a), and Fayella chickashaensis (Olson, 1972). Broiliellus-type paramedian osteoderms (Figure 61.4) have one laterally expanded paramedian osteoderm per vertebral segment, as is preserved in Broiliellus texensis (Williston, 1914; de Mar, 1966) and B. brevis (Carroll, 1964a). The articulated paramedian osteoderms curve ventrally in transverse section. The edges of adjacent osteoderms articulate with each other to form a shield that covers the entire dorsal surface of the trunk. Individual paramedian osteoderms do not overlap, and are located between the neural spines. The ventral surface of each paramedian osteoderm bears an anteriorly directed mid ventral flange that inserts between the neural spines. Dissorophus-type paramedian osteoderms (Figure 61.5) have two laterally expanded paramedian osteoderms per vertebral segment, and cover most the dorsal region of the trunk, to form a thick carapace(de Mar, 1966, 1968), as in Broiliellus. The adjacent edges of the internal and external series substantially overlap. The internal series are similar to those of Broiliellus, located between the neural spines, and bear a mid-ventral flange that inserts between the neural spines. Each individual osteoderm of the ornamented external series is located dorsal to the tip of the neural spine, and overlaps and articulates with the internal osteoderms both anteriorly and posteriorly.

Peltobatrachus type paramedian osteoderms. In Peltobatrachus pustulatus (Panchen, 1959) (Figure 61.6), the paramedian osteoderms are composite, with a number of fused individual osteoderms forming distinct ‘shields’ over the pectoral (Figure 61.6.1) and pelvic (Figure 61.6.1) regions of the spine, with transverse rows (Figure 61.6.2) between the pectoral and pelvic shield, and extending down the tail (Figure 61.6.4). The paramedian osteoderm shields and transverse rows overlap each anteriorly and posteriorly. The median osteoderm of each shield and transverse row fits between two neural spines. Laterally the paramedian osteoderm shields and transverse rows extend to the ends of the ribs and may fuse with them. The most lateral osteoderms, which articulate with the shields and transverse rows, are angled sharply ventrally.

DISCUSSION

Phylogenetic analysis

The tree topology presented in the main analysis is similar in many respects to that of Yates and Warren (2000), particularly in the arrangements at a lower taxonomic level, but differs significantly in the affinities of the major clades. The analysis of Yates and Warren (2000) was confined to the phylogenetic relationships of ‘higher’ temnospondyls. The analysis presented here differs in that it includes more plesiomorphic temnospondyls all of the extra taxa in the main analysis are either Plesiomorphic temnospondyls or members of the Euskelia. The inclusion of extra taxa is not the cause of the different tree topology, because the analysis using the same taxa as Yates and Warren (2000) but the characters and coding of the main analysis (Appendix 10: Figure 83) produced a tree topology similar to that of the main analysis (Figure 44). The major difference between the main analysis presented here and that of Yates and Warren (2000) is in the position of the Dvinosauria, which in the main analysis 192 K. PAWLEY PHD THESIS presented here, are less derived than the Euskelia. The remaining two large clades form the Eryopoidomorpha, which includes the Euskelia and the Stereospondylomorpha. The Limnarchia as defined by Yates and Warren (2000) do not exist in the main analysis, rather, there are two separate clades of obligatorily aquatic temnospondyls, the plesiomorphic Dvinosauria, and the Superstereospondyli (Figure 44). The hypothesis of the phylogenetic relationships of the major clades within the Temnospondyli presented here agrees (in the broad sense) with the hypotheses of Milner (1989; 1990), Sequeira and Milner (1993), Milner and Sequeira (1994), Holmes et al (1998), Laurin and Soler-Gijón (2001), and the informal assessments of Pawley and Warren (2006) and Pawley (in press). Interestingly, Edops craigi (the type specimen for the Edopoidea) is not united with the Cochleosauridae in the traditional Edopoidea, most recently defined by Sequeira (2004), Sidor et. al (2005), and Steyer (2006). This result is perhaps not surprising, because the Cochleosauridae lack some of the plesiomorphic cranial features possessed by Edops craigi. It is possible that the inclusion of the more plesiomorphic temnospondyl Caerorhachis bairdi, in a larger and less constrained analysis than other smaller computer based analyses of ‘edopoids’ (Sequeira, 2004; Sidor et al., 2005; Steyer et al., 2006), has allowed the plesiomorphic characteristics of Edops craigi to unite it with Caerorhachis bairdi at the base of the Temnospondyli. However, support for the phylogenetic relationships of basal temnospondyls below the Dvinosauria, as presented in this analysis, is generally lower than for the rest of the tree, and only a small number of extra steps are necessary to change the positions of the most basal taxa (Caerorhachis bairdi, Edops craigi, Dendrerpetontidae, Cochleosauridae). This is due in part to the high levels of missing data caused by the general lack of postcranial data (excepting Balanerpeton woodi and Dendrerpeton acadianum) among the basal temnospondyls. Consequently, future research on basal temnospondyls may change the hypothesis of relationships among basal temnospondyls presented here.

Comparison with previous phylogenetic analyses

The main analysis differs from that of Yates and Warren (2000) in the choice of ingroup taxa, selection of character states, and the coding of the data matrix. The results of the comparative phylogenetic analyses (Appendix 10) indicate that the different tree topology in the main analysis is due to the change in characters and coding rather than taxon sampling effects. Many of the postcranial characters used by Yates and Warren (2000) to define the Limnarchia pertain to morphogenetic, rather than phylogenetic variation (Chapter 4). Changing the postcranial characters of the matrix of Yates and Warren (2000) to a coding which minimised the effects of non-phylogenetic variation and produced a different tree topology, in which the Limnarchia do not exist (Appendix 10: Figure 86). The Limnarchia of Yates and Warren (2000) may therefore have been formed by the inclusion of postcranial characters that described the paedomorphic, morphogenetically immature condition of the endochondral postcranial skeleton (which are indistinguishable from true immaturity), rather than by characteristics that describe phylogenetic variation. Phylogenetic characters describing paedomorphic ‘adult’ features also describe subadults of non-paedomorphic taxa, resulting in lack of clear phylogenetic signal and increased homoplasy within the data set. This evidence does not refute the observation that many derived taxa within the Temnospondyli have paedomorphic postcranial skeletons, but it does imply that paedomorphism of the postcranial skeleton is convergently derived within the Temnospondyli. The postcranial skeleton of members of the Dvinosauria, which are CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 193 derived in relation to the Euskelia in the analysis of Yates and Warren (2000) is more plesiomorphic than that of taxa within the Eryopoidomorpha, which includes the Euskelia in many respects (Pawley, in press). The hypothesis of temnospondyl relationships presented here is significantly more parsimonious than that of Yates and Warren (2000) in both tree length (especially if the extra taxa are taken into account) and the obtained consistency index (Appendix 10: Table 12). The increased number of taxa in this analysis would predict an increase in the levels of homoplasy, and a subsequent decrease in the consistency index (Sanderson and Donoghue, 1989). The result obtained here is the opposite of that predicted. This indicates a stronger phylogenetic signal in the data set used here than in that of Yates and Warren (2000). As one of the aims of this analysis was to minimise the confounding effects of non-phylogenetic variation, in order to elucidate and improve true phylogenetic signal, this result is most satisfactory. The difference in tree topology is likely due to the change in postcranial characters, as simply changing the postcranial characters and coding of Yates and Warren (2000) resulted in a rearrangement of the major clades into a topology similar to that of the main analysis (Appendix 10: Figure 86). The difference in tree topology is not due to the choice of taxa or the effect of including more plesiomorphic taxa than Yates and Warren (2000), because searching with the same taxon list as Yates and Warren (2000) (Appendix 10: Figure 83) produced the same hypothesis of phylogenetic relationships as the main analysis (Figure 44). The major cause of the difference in tree topology is therefore most likely due to the choice of characters and the coding of the matrix, as outlined in the methods section. Redescription of some stereospondyl taxa, in particular rhinesuchids (Pawley and Warren, 2004), Lydekkerina huxleyi (Pawley and Warren, 2005), Mastodonsaurus giganteus (Schoch, 1999a), Thoosuchus yakovlev (Damiani and Yates, 2003), and trematosaurs (Steyer, 2002) is likely to have influenced the different phylogenetic arrangement within the basal superstereospondyls, simply by the addition of more complete and accurate data.

Phylogenetic value of postcranial characteristics

Temnospondyl researchers apparently overlook characteristics of the postcranial skeleton for a number of reasons, including collection bias towards cranial material and the general perception that there is more phylogenetic variation in cranial material. Despite the comprehensive nature of this overview of temnospondyl postcranial, and maximal inclusion of postcranial characters, only 26% of the characters used in this phylogenetic analysis refer to postcranial characteristics, highlighting the conservative nature of the postcranial skeleton in temnospondyls. The use of postcranial characters is important for phylogenetic resolution at a higher taxonomic level, simply replacing the postcranial characters of Yates and Warren (2000) (Appendix 10: Figure 86) resulted in a rearrangement of the major clades into a topology similar to that of the main analysis (Figure 44) Perhaps more interestingly, the test analyses based entirely on cranial characters (Appendix 10: Figure 82, Figure 84, Figure 85) produced results that were incongruent with those of the more inclusive analyses. In the analysis that used all the taxa of the main analysis, but excluded the postcranial characters (Appendix 10: Figure 82), the Cochleosauridae form a clade basal to the Dendrerpetontidae, unlike the analysis using a full character set. The analysis using the same taxon list as Yates and Warren (2000), but omitting postcranial characters (Appendix 10: Figure 84) produced a tree topology close 194 K. PAWLEY PHD THESIS to that of Yates and Warren (2000) in placing members of the Euskelia as basal taxa, contra all the other analyses. Omitting the postcranial characters from the original matrix of Yates and Warren (2000) placed the Dvinosauroidea as derived Superstereospondyli, again contrary to all the other analyses. Of the analyses produced here, none of the analyses based entirely on cranial characters (Appendix 10: Figure 82, Figure 84, Figure 85) agree with each other, unlike the analyses that included postcranial characters (Figure 44, Appendix 10: Figure 83, Figure 86) which are all congruent. These results indicate phylogenetic analyses of temnospondyls based entirely on cranial characters may not produce the parsimonious hypothesis of phylogenetic relationships within the Temnospondyli.

Significance of vertebral structure

The high degree of variation in vertebral structure in temnospondyls is morphogenetic (Chapter 4) as well as phylogenetic, which means vertebral structure has limited taxonomic application within the Temnospondyli. The distribution of vertebral types within the Temnospondyli is sporadic, with the various types only characteristic of clades at lower taxonomic levels. Well-ossified disc shaped amphicoelous intercentra and pleurocentra are only known in, and are characteristic of, the Tupilakosauridae. In contrast, the distribution of stereospondylous vertebrae within the Stereospondyli is so sporadic that it cannot be considered a defining synapomorphy of the clade. The occurrence of two differing vertebral types in the same genus, Eryosuchus, (Superstereospondyli) indicates that the distribution of stereospondylous vertebrae may be entirely dependant on the morphogenetic stage of the specimens. E. tverdochlebovi (Ochev, 1972; Shishkin, 1991) has stereospondylous vertebrae, but E. pronus (Howie, 1970) has rhachitomous vertebrae. The presence of stereospondylous vertebrae also varies along the vertebral column in some taxa, with the anterior trunk vertebrae being stereospondylous and the posterior trunk vertebrae rhachitomous in taxa such as Mastodonsaurus giganteus (Schoch, 1999a). Further research is necessary to clarify the situation; and determine if a correlation between morphogenetic stage and the occurrence of stereospondylous vertebrae is present. Removal of the relevant postcranial characters makes no impact on the tree topology (Appendix 10: Table 9). These results suggest that the vertebral characters do not have a strong phylogenetic signal. The historical reliance on these vertebral characteristics as the main postcranial feature for taxonomic purposes with the Temnospondyli is consequently not justified. The phylogenetic distribution of gastrocentrous, embolomerous, and stereospondylous vertebral centra is useful at a lower taxonomic level, but these characteristics are not useful or reliable at higher taxonomic levels. However, many phylogenetic characters are prone to homoplasy, so the use of vertebral characters should be continued, but treated with caution for diagnostic purposes. The comparative lack of phylogenetic variation in the postcranial skeleton may discourage the use of postcranial characters in analysis, but it is important to recognise that this lack of variation also results in a strong phylogenetic signal. The inclusion of comparatively invariant postcranial characteristics consequently establishes a strong ‘backbone’, increasing the robustness of the phylogenetic hypothesis. As an example, Yates and Warren (2000) found that constraining the Dvinosauria and Brachyopoidea as sister taxa required only one extra step. Postcranial characteristics clearly distinguish the Dvinosauria as relatively plesiomorphic temnospondyls, and the Brachyopoidea as derived stereospondyls (Pawley, in press). The phylogenetic analysis of Yates and Warren (2000) used only one of the postcranial characteristics used to distinguish the Dvinosauria and Brachyopoidea (Pawley, in press). The main analysis included all of the postcranial CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 195 characters used to distinguish the Dvinosauria and Brachyopoidea (Pawley, in press), test analyses using constraint trees found that the number of steps needed to place the Dvinosauria as sister taxa to the Brachyopoidea increased to 18, a significantly more substantial result.

Caerorhachis bairdi and Casineria kiddi as basal temnospondyls

Caerorhachis bairdi has previously been classified either as a basal temnospondyl (Holmes and Carroll, 1977; Godfrey et al., 1987) or a ‘reptiliomorph’ stem amniote (Milner and Sequeira, 1994; Ruta et al., 2002; Ruta et al., 2003). Much of the disagreement regarding its phylogenetic affinities, in particular in excluding it from the Temnospondyli, concerns ventrally joined pleurocentra present in Caerorhachis bairdi, generally considered an apomorphy of the ‘reptiliomorph’ lineage. As described above, ventrally joined pleurocentra are present in several unrelated, more derived temnospondyls, thus ventrally joined pleurocentra are not an exclusive apomorphy of ‘reptiliomorphs’. All of the other ‘reptiliomorph’ postcranial characteristics listed by Ruta et al. (2002), which are too numerous to restate here, are either actually present in other basal temnospondyls, or potentially may be present, because they are plesiomorphic for early tetrapods. Characteristics of basal temnospondyls, observable in Caerorhachis bairdi, but absent in most ‘reptiliomorphs’, include: denticle shargreen covering entire palate, reduced vomerine tusks, the presence of interpterygoid vacuities, lack of a parietal-tabular suture, the skull table firmly fused to the cheek, and an ilium with a post-iliac process, but lacking a dorsal extension. The results of this analysis are in agreement with Milner (1980b) in excluding Caerorhachis bairdi from the Dendrerpetontidae, and with Godfrey et al. (1987) in placing v as the most basal temnospondyl taxon. Caerorhachis bairdi (Holmes and Carroll, 1977; Ruta et al., 2002) and Casineria kiddi (Paton et al., 1999) both from the Carboniferous of Scotland, are indistinguishable based on the available evidence. A large phylogenetic analysis of early tetrapods placed Caerorhachis bairdi and Casineria kiddi at the base of the temnospondyl clade. The bootstrap support of 81% and Bremer node value of five for that node indicates that the phylogenetic position of Caerorhachis bairdi and Casineria kiddi is unlikely to be incorrect, despite the incomplete nature of both specimens. Chapter 6 presents a description of the characteristics that unite Caerorhachis bairdi and Casineria kiddi and that place them as basal temnospondyls, rather than rather than part of the ‘reptiliomorph’ lineage, as proposed by Ruta et al. (2002) and Paton et al (1999) respectively.

Articulation of pectoral girdle elements

Some confusion exists as to the proper articulation of the pectoral girdle elements. Articulated specimens are available for Dendrerpeton acadianum (Holmes et al., 1998), Eryops megacephalus (Pawley and Warren, 2006), Micropholis stowi (Watson, 1913; Schoch and Rubidge, 2005), rhinesuchids (Pawley and Warren, 2004), Lydekkerina huxleyi (Pawley and Warren, 2005), and Siderops kehli (Warren and Hutchinson, 1983). In lateral view (Figure 51.2), the dorsal process of the clavicle overlaps the cleithrum, which in turn articulates with the scapulocoracoid. The center of ossification of the clavicle, ventral tip of the cleithrum and glenoid align. In ventral view (Figure 51.1), the articulated interclavicle and clavicles project between the jaws. The posterior border of the clavicles and interclavicle overlaps the coracoid anteromedially; ventral scutes would cover the remainder of the coracoid in life (Figure 61.2.2). The internal surface of the 196 K. PAWLEY PHD THESIS scapular blade attaches to the expanded distal ends of the posterior cervical ribs (Watson, 1958) via the levator scapulae and serratus anterior muscles (Romer, 1922).

Phylogenetic distribution of heterochronic effects on the postcranial skeleton

A morphogenetically immature (Chapter 4), paedomorphic (sensu McNamara, 1986) postcranial skeleton is typical of the Dvinosauria and Superstereospondyli. As all taxa within these clades possess deeply incised sensory canals on their skulls, the paedomorphic condition of the postcranial skeleton is most likely an adaptation for an aquatic existence. Bone is metabolically more expensive to form and maintain than cartilage, so the delayed ossification of endochondral elements would achieve an energy saving if a high degree of ossification was not ecologically important (Pawley, in press). In contrast, the small, well ossified postcranial skeleton of the Amphibamidae is progenetic (sensu McNamara, 1986), achieving a high degree of ossification and morphogenetic maturity at a small size, as proposed by Bolt (1969).

Biology of temnospondyls

Approximately half of the temnospondyls presented in the phylogenetic analysis possess lateral lines on the skull, which indicates their lifestyle was mainly if not obligatorily aquatic. These taxa are present in the aquatic clades of the Dvinosauria and Stereospondyli (with the exception of Nigerpeton ricqlesi and Cheliderpeton spp.). Terrestrial adaptations of temnospondyls were discussed by Pawley and Warren (2006) and aquatic adaptations by Pawley (in press). De Fauw (1989) divided temnospondyls into five ecomorphic types, terrestrial, semi-aquatic freshwater, semi-aquatic euryhaline, fully aquatic freshwater, and fully aquatic euryhaline, based on morphological features and taphonomic evidence. The range of body forms displayed by temnospondyls indicates that they probably occupied a wide range of ecological niches. The most basal temnospondyls are terrestrial (Caerorhachis bairdi, Edops craigi, and the Dendrerpetontidae), and first appear in the Lower Carboniferous (Holmes, 2000). Terrestrial adaptations in temnospondyls include the following morphological features, which have been regarded as adaptations for a terrestrial lifestyle in other taxa: absence of lateral line sulci; absence of any trace of an ossified branchial system; presence of a large tympanic ear and a relatively rod like stapes implying the ability to perceive high frequency sound; short trunk and tail; limb bones with well developed articulation surfaces and a comparatively high degree of ossification; including an ossified carpus and tarsus (Olson, 1941; de Mar, 1968; Milner and Sequeira, 1994; Holmes, 2000; Carroll, 2002; Vallin and Laurin, 2004). Apart from the basal temnospondyls, the most terrestrially adapted clade is the derived Euskelia, which display not only the above adaptations, but also further adaptations that may have improved their terrestrial capabilities (below). Terrestrial taxa do not persist beyond the Lower Triassic. Obligatorily aquatic temnospondyls (Dvinosauria) first appear in the Lower Permian, but during the Lower Triassic they are replaced by the Stereospondyli, which persist into the Lower Cretaceous (Warren et al., 1997). The most obvious adaptations for an aquatic existence observed in temnospondyls are lateral line sulci, dorsoventral flattening of the skull and body, the extensive ventral area of the interclavicle and clavicles, and the poorly ossified and paedomorphic condition of the endochondral bones. Otherwise, their postcranial skeletons are most similar to those of terrestrial temnospondyls, only minor adaptations for an aquatic lifestyle can be observed, such as the anterior clavicular flange. Aquatic temnospondyls most likely had terrestrial CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 197 ancestors, and are thus secondarily rather than primarily aquatic (Pawley, in press). Bone histology of unidentified stereospondyls from the Early Triassic Arcadia Formation is consistent with an aquatic lifestyle for these taxa (Damiani, 2000). It seems most likely that many temnospondyls were amphibious, adapted for locomotion in both aquatic and terrestrial environments. There is no doubt that some temnospondyls were capable of sustained and efficient terrestrial location, particularly basal temnospondyls, and members of the Euskelia. Conversely, the small, weakly ossified limbs of the Dvinosauria, which are combined with large size in the Superstereospondyli, implies that it is unlikely that any of these taxa ever ventured onto land. Many temnospondyls cannot be so conveniently categorised, so the description of any particular temnospondyl as aquatic or terrestrial is a matter of degree, rather than an absolute. Thus most temnospondyls are best described as amphibious.

Appearance of temnospondyls

The large size and general body form of most temnospondyls is most similar to that of large amphibious extant reptiles such as crocodilians (Warren, 2000). Only amphibamids are comparable to extant salamanders in their small size and gracile limbs. The skulls and dentition of all temnospondyls (Holmes, 2000; Warren, 2000) indicate a carnivorous diet. The eyes of temnospondyls are dorsally located, Chernin and Cruickshank (1978: p 112) postulated that mastodonsaurids (capitosaurs) “lurked in neutral buoyancy with just the eyes protruding above water, as in modern crocodiles”. All temnospondyls possessed a large, laterally flattened tail suitable for aquatic propulsion. In many, if not all, temnospondyls, the skin was entirely covered in scales or osteoderms, rendering their external appearance quite unlike that of salamanders, and again more like that of crocodilians. Warren et al. (1997) hypothesised that temnospondyls occupied a similar ecological niche to that of modern eusuchian crocodilians, which out competed and replaced them.

Ceratobranchials

Reconstructions of neck musculature indicate that a large gap would be present between the muscles along the vertebral column and those of the jaw musculature (P. Trusler 2004 pers comm.). In extant salamanders, the large cartilaginous ceratobranchials and associated musculature occupy this gap, and function in feeding and respiration (e.g. Francis, 1934; Baird, 1951; Cox and Tanner, 1989; Deban and Wake, 2000; Wake and Deban, 2000). In extant salamanders, ossified ceratobranchials are present in aquatic taxa, where they provide a strong support for the hyobranchial muscles, which create negative pressure inside the buccal cavity during suction feeding (Deban and Wake, 2000). The ceratobranchials of terrestrial salamanders are typically poorly ossified or completely unossified, not only because they do not generally suction feed, but also because a more flexible neck region is advantageous for terrestrial prey capture (Wake and Deban, 2000). In extant salamanders, these gill rakers assist in preventing prey escape during suction feeding (Deban and Wake, 2000). Feeding in temnospondyls is also discussed by Damiani (2001b) and Witzmann (2005; 2006b). 198 K. PAWLEY PHD THESIS

Biomechanical interpretations of locomotor capacity in temnospondyls

Few analyses have been performed on the locomotor capacities of temnospondyls, most of these are generalised, and do not discuss the specific functions of detailed morphological characteristics.

Axial skeleton

Previous interpretations of biomechanics of rhachitomous type vertebrae (Olson, 1936b; Rockwell et al., 1938; Parrington, 1967; Holmes, 1989b) have considered that they are a unique design, adapted for a high degree of flexure which these workers considered must have been necessary for locomotion in temnospondyls. Reinterpretation of the cartilaginous parts of the rhachitomous centra in temnospondyls as amphicoelous pleurocentral and intercentral rings (Chapter 4), means that previous functional interpretations of rhachitomous-type centra are incorrect. Rhachitomous-type vertebrae are likely to have possessed similar biomechanical properties to those of other lower vertebrates (Chapter 4). The essentially monospondylous, pleurocentrum dominated centra of Peltobatrachus pustulatus (Panchen, 1959), Fayella chickashaensis (Olson, 1972), and Doleserpeton annectans (Bolt, 1969) (Figure 47) have a generally similar construction to those of extant salamanders and reptiles, indicating that this centum type had similar biomechanical properties to those of extant (plesiomorphic) reptiles and salamanders. Vaughn (1971) proposed that ‘highly terrestrial labyrinthodonts’ strengthened the vertebral column for terrestrial locomotion though various devices. These include shortening of the trunk, monospondylous (gastrocentrous) centra, paramedian osteoderms, the ‘sail’ of Platyhystrix rugosus (Lewis and Vaughn, 1965; Vaughn, 1971), and fused neural spines in Astreptorhachis ohioensis (Vaughn, 1971). The presence of these various adaptations, which reduce the degree of flexibility of the vertebral column, implies that the high degree of flexibility proposed as necessary for terrestrial locomotion by functional analyses, may in fact actually be an impediment to efficient terrestrial locomotion.

Dermal ossifications

Findlay (1968) considered that ventral scutes protected the ventral surface of the trunk, and that they were optimally designed to both prevent slipping and minimise friction during the lateral undulations of belly crawling. Paramedian osteoderms are only present in the temnospondyl taxa with adaptations for terrestriallity. De Mar (1966) considered that the ‘armour’ of dissorophids would have reduced the flexibility of the body and strengthened the trunk as an adaptation for terrestrial locomotion. Eusuchian crocodilians display efficient locomotion in both aquatic and terrestrial environments (Reilly and Elias, 1998; Frey and Salisbury, 2000), and also possess enlarged osteoderms that are confined to the dorsal surface body and articulate with the neural spines. The osteoderms (termed ‘paravertebral’) in crocodilians are of a similar, but not identical, design to those of temnospondyls. The paravertebral osteoderms in eusuchian crocodilians are integrated with the underlying epaxial musculature, and their configuration dictates the flexibility of the trunk. Isometric contraction of the epaxial muscles braces the paravertebral shield against the vertebral column, stabilising the trunk, in particular the intervertebral joints, during terrestrial locomotion. Asymmetrical contraction of the epaxial muscles also facilitates lateral undulation during aquatic CHAPTER 5: POSTCRANIAL SKELETON OF TEMNOSPONDYLS 199 locomotion (Salisbury and Frey, 2000). The paravertebral osteoderms in crocodiles thus are an adaptation that improves the efficiency of locomotion in both aquatic and terrestrial environments without major compromise to either form of locomotion. Although no biomechanical studies of the paramedian osteoderms in temnospondyl exist, they may have performed a similar function to that hypothesised for crocodilians. In extant salamanders, the loss of the fleshy dorsal and ventral caudal fins during metamorphosis reduces the efficiency of swimming in post-metamorphic individuals, and is associated with the transition to terrestrial locomotion (Frolich and Biewener, 1992). Temnospondyls do not have a definite metamorphosis (Schoch, 2002d; Witzmann and Pfretzschner, 2003; Schoch and Fröbisch, 2006), and dorsal and ventral caudal fins are only known so far in larval temnospondyls (e.g. Milner, 1982a; Werneburg, 2002). Determination of the presence or absence of dorsal and ventral caudal fins in larger specimens would be informative as to the lifestyle of temnospondyls, particularly if taxa with paedomorphic postcranial skeletons could be contrasted with morphogenetically mature terrestrial taxa.

Appendicular skeleton

The appendicular skeleton of temnospondyls resembles that of other Palaeozoic tetrapods (Holmes, 2000), so functional interpretations of locomotion in these taxa (Romer, 1922; Miner, 1925; Schaeffer, 1941; Holmes, 1977, 1984, 2003) are applicable to temnospondyls. The humerus undergoes considerable rotation during the step cycle, but there are no functional wrist or ankle joints, rather a flexible manus and pes (Holmes, 1977, 2003). These studies indicate that the limbs of well-ossified temnospondyls are load bearing and that the style of locomotion was probably similar to that of the sprawling gait of extant salamanders and plesiomorphic reptiles. All the functional interpretations summarised here are somewhat generalised, and have down played the significant morphological differences between the postcranial skeletons of temnospondyls and extant taxa. In particular, the configuration of the vertebrae, the large, rigid pectoral girdle, the structure of the humerus and femur, and the extensive heavy covering of scales, are most unlike that of any modern tetrapod. The specific effects that these different morphological characteristics may have had on the locomotor capacities of temnospondyls have yet to be explored.

CONCLUSIONS

In temnospondyls, considerable morphological variation is observable in vertebral structure and the configuration of the dermal pectoral girdle; otherwise, the postcranial skeleton is generally conservative. Phylogenetic analysis of temnospondyls has improved on previous analyses, both in reduction of tree length and reduction of homoplasy. This result is at least partially due to the separation of morphogenetic and phylogenetic variation in the postcranial skeleton, but also to the addition of new postcranial characters. Test analyses indicate that addition of postcranial characters has a significant effect the resulting hypothesis of phylogenetic relationships within the Temnospondyli. Most temnospondyls are best described as amphibious, the most basal temnospondyls are well adapted for terrestrial locomotion, with one major derived clade, the Euskelia, displaying postcranial adaptations that may have improved terrestrial locomotion. Two derived clades, the Dvinosauria and Superstereospondyli were most likely obligatorily aquatic. In general appearance and lifestyle, temnospondyls most resemble extant crocodilians. 200 K. PAWLEY PHD THESIS

ACKNOWLEDGEMENTS

This manuscript forms part of the PhD dissertation of Kat Pawley, who sincerely thanks her supervisor, Dr Anne Warren, for her support and encouragement. Funding was provided by an American Museum of Natural History collections visitation grant and an Australian Postgraduate Award to K. Pawley. For hospitality and access to collections I would like to thank Prof Bruce Rubidge and Dr Mike Raath of the Bernard Price Institute for Palaeontological Research, Johannesburg, Dr Heidi Fourie of the Transvaal Museum, Pretoria; Dr Johann Welman and Elize Potgieter of the National Museum in Bloemfontein; Dr Gene Gaffney of the American Museum of Natural History, New York; and Chuck Schaff of the Museum of Comparative Zoology, Harvard University. For assistance with the loan of specimens I would like to thank Lyndon K. Murray and Dr Tim Rowe of the Texas Memorial Museum, Austin, Texas; Dr Pat Holroyd of the Museum of Paleontology, University of California, Los Angeles, California; and Dr Greg Gunnell of the University of Michigan Museum of Paleontology, Ann Arbor, Michigan. I am grateful for the generosity of Drs Ross Damiani and Jean Sebastian Steyer in providing copies of manuscripts in press, and thanks to Dr Steve Salisbury for insightful discussion on terminology and biomechanics of dermal osteoderms. In depth discussions with Peter Trusler were very constructive in elucidating details of plausible cervical musculature. Lastly, but very much not least, I am indebted to Dr Marcello Ruta for his unstinting generosity in providing a copy of the data matrix used in Ruta et al. (2003).