School of Earth Sciences, University of Bristol, Queen s Road, Bristol BS8 1RJ, UK 2

Transcription

1 bs_bs_banner Zoological Journal of the Linnean Society, 2015, 173, With 20 figures Osteology of Rauisuchus tiradentes from the Late Triassic (Carnian) Santa Maria Formation of Brazil, and its implications for rauisuchid anatomy and phylogeny STEPHAN LAUTENSCHLAGER 1 * and OLIVER W. M. RAUHUT 2 1 School of Earth Sciences, University of Bristol, Queen s Road, Bristol BS8 1RJ, UK 2 SNSB, Bayerische Staatssammlung für Paläontologie und Geologie and Department of Earth and Environmental Sciences, Ludwig-Maximilians-University, Richard-Wagner-Str. 10, Munich, Germany Received 4 December 2013; revised 15 July 2014; accepted for publication 28 July 2014 As the eponymous type species for rauisuchians, Rauisuchus tiradentes von Huene, 1938 represents an important but inadequately known specimen. The revision of the type material from the Upper Triassic (Carnian) Santa Maria Formation of Brazil reveals new anatomical data and previously unrecognised characters, including three new possible autapomorphies: a knob-like thickening on the base of the posterior process of the premaxilla; short and ventrally keeled cervicals lacking postzygodiapophyseal laminae; and mid-caudal vertebrae with an accessory neural spine and a postspinal lamina. Several elements are re-identified, including a postorbital (originally identified as postfrontal) and a pterygoid (originally identified as a prefrontal), and additional material from a secondary fossil site, originally assigned to R. tiradentes, including a left ilium, is excluded. Based on the recovered information and new morphological data, the systematic position of R. tiradentes is tested in a comparison of two phylogenetic reanalyses. Both analyses differ in respect to the phylogenetic position of R. tiradentes, and recover rauisuchians as a paraphyletic assemblage of non-crocodylomoprh archosaurs.. doi: /zoj ADDITIONAL KEYWORDS: Archosauria archosaur phylogeny Crurotarsi Late Triassic Poposauroidea Pseudosuchia rauisuchian archosaurs Rauisuchoidea Santa Maria Formation. INTRODUCTION The Triassic period was a crucial time in the evolution and diversification of the major archosaur lineages. Among them, rauisuchians (sensu Nesbitt et al., 2013) form one of the most intriguing and yet enigmatic groups. Prior to the radiation of the dinosaurs, they were the dominant predators and an important faunal element in Middle and Upper Triassic terrestrial vertebrate ecosystems. Although their first fossils were found nearly 150 years ago (Meyer, 1861), and despite their global distribution, rauisuchian archosaurs are still poorly understood. The often fragmentary fossil *Corresponding author. glzsl@bristol.ac.uk record, a plethora of erected taxa and families, and several described, but practically unknown specimens, among them composite taxa and chimeras (Long & Murry, 1995), have led to more confusion than clarification. The monophyly of Rauisuchia is still strongly debated (Gower, 2000; Gebauer, 2004; Gower & Nesbitt, 2006; Gower & Schoch, 2009; Brusatte et al., 2010; Nesbitt, 2011), and the relationships within rauisuchians as well as their position within Archosauria are only now starting to come into focus (Nesbitt, 2011). This is at least partially linked to the fact that many taxa are based on fragmentary material, and only a few have been adequately described. A thorough and up-to-date morphological description of old and new taxa and careful diagnoses are therefore fundamental to improve our knowledge of rauisuchian anatomy 55

2 56 S. LAUTENSCHLAGER AND O. W. M. RAUHUT and systematics. Although several publications have focused on this aspect recently (e.g. Gebauer, 2004; Weinbaum & Hungerbühler, 2007; Peyer et al., 2008; Brusatte et al., 2009; Lautenschlager & Desojo, 2011; De França, Langer & Ferigolo, 2013), Rauisuchus tiradentes von Huene, 1938a has long been neglected in most works on rauisuchian anatomy and phylogeny. This is especially surprising given that this species is the name-bearing taxon of Rauisuchia as well as other taxonomic subunits (e.g. Rauisuchidae, Rauisuchoidea) terms widely used in accounts of basal crurotarsan archosaurs. Although Rauisuchia is most likely to refer to a paraphyletic assemblage of basal archosaurian taxa (Nesbitt, 2011), which are difficult to assign into discrete and well-diagnosed clades, we employ this term here in the absence of a more precise definition, and for its historical significance. We follow Nesbitt et al. (2013) in their definition of Rauisuchia, comprising all suchian taxa that are not members of Aetosauria or Crocodylomorpha. Phylogenetic definitions of more inclusive clades also follow the classification outlined in Nesbitt et al. (2013). The disarticulated and fragmentary material of R. tiradentes was found by Friedrich von Huene during an expedition in the province Rio Grande do Sul in Southern Brazil between 1928 and 1929 (von Huene, 1938a, 1942). It was collected from the Upper Triassic (Carnian) Santa Maria Formation and, together with other material, including Prestosuchus chiniquensis von Huene, 1938a, was brought to Germany, where it was described by von Huene in von Huene had already mentioned the genus Rauisuchus in a systematic listing of the Thecodontia in 1936 (von Huene, 1936), but without providing a diagnosis or description, so that the generic name remained a nomen nudum. The type species of R. tiradentes was than briefly diagnosed in a short synopsis on his findings in South America (von Huene, 1938a), before the more complete description in 1942 (von Huene, 1942). Probably because this description was published during World War II, lacking sufficient illustrations, and did not provide the detail seen in more recent works, R. tiradentes has rarely been considered in analyses of the evolutionary history of basal crurotarsan archosaurs. Only Krebs (1973) added some new information on the tarsus of R. tiradentes, and the same author later designated a lectotype for this taxon (Krebs, 1976). The scope of this contribution is a detailed description of R. tiradentes, providing the background for future work on rauisuchian anatomy and systematics. Although anatomical characters have been used to supplement phylogenetic analyses very recently (Brusatte et al., 2010; Nesbitt, 2011), they are restricted to personal observations of the respective authors and have either not been recorded in detail or are based on the partly erroneous original description. Here, we critically review the type material and provide a revised and emended diagnosis for this taxon. Furthermore, elements of uncertain assignment are removed from the type material and their taxonomic status is discussed. Morphological data derived from this description are further used to supplement two existing phylogenetic analyses to evaluate the systematic position of R. tiradentes. GEOLOGICAL SETTING The Santa Maria Formation is a terrestrial sequence consisting of conglomerates and sandstones at its base, predominantly overlain by mud- and siltstones, deposited by meandering fluvial systems. Palaeoecological and taphonomic analyses (Holz & Souto-Ribeiro, 2000; Zerfass et al., 2003) showed that the red mudstones, from which most of the Santa Maria fossils originated, represent a floodplain facies, deposited during seasonal rainfalls in an otherwise semi-arid climate. The Santa Maria Formation comprises part of the Ladinian and most of the Carnian stages (Pierini et al., 2002; Rubert & Schultz, 2004), and can be further divided into the basal Passo das Tropas Member and the Alemoa Member, with its characteristic red mud- and siltstones. Most of the material collected by von Huene was most probably derived from the fossiliferous red siltstones of the Alemoa Member (Langer, 2005). These sediments are exposed in erosional outcrops, called sangas, throughout the region. Although von Huene s (1942) description of the fossil localities is vague, a comparison of the works of Langer (2005) with the original accounts (von Huene & Stahlecker, 1931) suggests that the horizon in which R. tiradentes was discovered informally named the Zahnsanga (tooth sanga) corresponds to the Lower Carnian of the Alemoa Member. According to von Huene & Stahlecker (1931: 29), the material came from the uppermost 2.5 m of a series of approximately 8 m of brick-red to purple, slightly sandy mudstones. MATERIAL AND METHODS INSTITUTIONAL ABBREVIATIONS BSPG, Bayerische Staatssammlung für Paläontologie und Geologie, München, Germany; GPIT, Institut für Geowissenschaften, Universität Tübingen, Germany; PIMUZ/PIZ, Paläontologisches Institut und Museum der Universität Zürich, Switzerland; SMNS, Staatliches Museum für Naturkunde, Stuttgart, Germany. MATERIAL von Huene did not give precise locality information for the materials referred to R. tiradentes, but only mentioned that they were found near the road from Santa Maria to San Jose (von Huene, 1942: 198). The sketch

3 OSTEOLOGY OF RAUISUCHUS TIRADENTES 57 of a map of the area provided by von Huene & Stahlecker (1931: fig. 7) illustrates the Zahnsanga as an approximately 200 m long ravine or escarpment parallel with, and at some m distance from, the road. These authors mention that excavations were made in these sediments and have yielded materials of parasuchians (= basal crurotarsans) and cynodontians (von Huene & Stahlecker, 1931: 29). The material from excavation 15 plus two fragmentary specimens from the same Zahnsanga ( Find 1020 and Find 1025 ) were described as R. tiradentes by von Huene (1942). Excavation 15 was thus said to have yielded a disarticulated partial skeleton in a lens of 0.5 m in thickness and about 2 m in width (von Huene, 1942: 198). This specimen was later made the lectotype of R. tiradentes by Krebs (1976). Find 1025 consisted of a partial left maxilla (BSPG AS XXV 122), whereas Find 1020 comprised a posterior cervical vertebra [BSPG AS XXV 123], a rib fragment [BSPG AS XXV 124], a chevron bone [BSPG AS XXV 120] and a fragment of a second chevron bone [BSPG, no number], and a complete left ilium [BSPG AS XXV 88] (von Huene, 1942: 215). Interestingly, all the latter elements carry the field number 1025, probably because von Huene (1942: 215) noted that Find Nr was so close to 1025 that it cannot be excluded that both represent the same individual. The distance was m. Consequently, Krebs (1976: 73) listed all of this material under number 1025, which he designated as a paralectotype of R. tiradentes. The vertebra included in Find 1020 seems to represent an anterior caudal vertebra, rather than a posterior cervical vertebra. This vertebra is relatively much shorter than the anterior caudal vertebrae of the specimen from locality 15, and it seems to be much too small to belong to the same individual as the maxillary fragment of Find As the only overlapping elements between Find 1020 and the specimen from locality 15 are thus a rib fragment and a chevron bone (which also seems too large for the vertebra BSPG AS XXV 123), which do not show any diagnostic characters, this specimen cannot be referred to the same taxon with any certainty, and is thus here removed from R. tiradentes. von Huene (1942: ) furthermore referred several isolated teeth and a proximal end of an ulna (BSPG AS XXV 101) and a partial metatarsal (BSPG AS XXV 102) from different sangas to R. tiradentes; however, as the teeth are not diagnostic, and there are no ulna or metatarsals preserved in the specimen from locality 15, none of this material can be referred to R. tiradentes. This leaves the question whether all of the disarticulated elements from locality 15 represent a single individual. As noted by von Huene (1942), all of the material came from a rather small sediment lens (see above). Furthermore, there are no duplicated bones, the vertebral elements fit well in morphology, and most of the material is of comparable size. Exceptions to this are a left squamosal (BSPG AS XXV 62; originally described as a right squamosal by von Huene, 1942: ) and an element described as left atlantal neural arch by von Huene (BSPG AS XXV 71), which appear to be too large in comparison with other elements. The squamosal is rather poorly preserved and has a bloated appearance. The latter condition is frequently found in specimens from the Santa Maria Formation and might diagenetically increase the size of elements (Holz & Schultz, 1998), so this might at least partially account for the apparently disproportionate size of this bone. The element identified as atlantal neural arch by von Huene (1942) cannot be shown to represent this bone. Apart from its large size, the bone lacks an expanded ventral base that is offset from the dorsal arch by an incision (as is found in other archosaurs; e.g. Nesbitt, 2007: fig. 27) and any clearly defined postzygapophyseal facets. A possible different interpretation of this bone is that it represents an unusual, keeled anterior cervical rib, as it is present in some basal archosauriforms (e.g. Nesbitt, 2011: fig. 28J), in which case this bone would be too large to fit with the rest of the lectotype material of Rauisuchus. Given the poor preservation and lack of comparative material for many basal archosaurs, however, we consider it best to regard this element currently as an unidentified bone. In the absence of any other evidence for multiple individuals being represented, we thus retain the material from locality 15 as the lectotype of R. tiradentes, although the referral of at least these two elements should be regarded as tentative. The latter is also true for the relatively high number of isolated teeth found at this excavation. Although some of these teeth almost certainly belong to R. tiradentes, it cannot be excluded that teeth of other carnivorous archosaurs, which might have scavenged on the carcass, are also present, as is often the case even in clearly single associated or even articulated skeletons (e.g. Buffetaut & Suteethorn, 1989; Hungerbühler, 1998). SHAPE ANALYSIS For the shape analysis of the rauisuchian ilia 22 specimens were sampled, selected from first-hand observations or literature data. Incomplete and fragmentary specimens were not considered. Specimens were recorded in left lateral view. Contours of the individual ilia were digitized and saved as x y coordinates using tpsdig 2.16 (Rohlf, 2010). Fast Fourier transform (FFT) and principal component analysis (PCA) were performed on the data set using PAST (Hammer, Harper & Ryan, 2001), including the programs Hangle, Hmatch, and Hcurve, designed by Crompton & Haines (1996). Outlines were smoothed ten times to eliminate pixel

4 58 S. LAUTENSCHLAGER AND O. W. M. RAUHUT noise and 18 Fourier harmonics were found to sufficiently describe the outlines. PHYLOGENETIC ANALYSIS To test the contribution of the revised osteology of R. tiradentes, phylogenetic analyses were performed using modified versions of the data sets from Brusatte et al. (2010) and Nesbitt (2011) (as employed in an updated reanalysis in Butler et al. 2011). The character scoring for R. tiradentes for both analyses was modified, based on the morphological description in this study. Each data set was analysed in a parsimony analysis using the branch-and-bound search algorithm of PAUP 4.0b10 (Swofford, 2002). SYSTEMATIC PALAEONTOLOGY ARCHOSAURIA COPE, 1869 PSEUDOSUCHIA ZITTEL, SENSU GAUTHIER & PADIAN, 1985 RAUISUCHIA VON HUENE, 1942 SENSU BRUSATTE ET AL., 2010 RAUISUCHIDAE VON HUENE, 1942 SENSU BRUSATTE ET AL., 2010 GENUS RAUISUCHUS VON HUENE, 1938A Type and only known species Rauisuchus tiradentes von Huene, 1938a. Diagnosis As for the type species (by monotypy). surangular, prearticular and angular, odontoid(?), axis, cervical, dorsal and caudal vertebrae, ribs, chevron bones, right scapulacoracoid, left pubis, right tibia, fibula and astragalus, and body osteoderms. Type locality and horizon Zahnsanga close to Santa Maria, Province Rio Grande do Sul, Southern Brazil; Alemoa Member, Santa Maria Formation, Lower Carnian (Late Triassic). Diagnosis Rauisuchus tiradentes differs from all rauisuchian archosaurs in the following unique combination of characters (*indicates autapomorphies): main body of premaxilla quadrangular; posterior process of premaxilla longer than main body; *bulging protuberance on the base of the posterior process of the premaxilla; maxilla excluded from external naris; massive, transversely broadened articulation between the anterior margin of the ascending process of the maxilla and the posterior ascending process of the premaxilla and the nasal; anterior part of the nasal notably flexed ventrally; rugose ridge on squamosal and on main body of jugal; dorsolaterally orientated upper temporal fenestra with raised lateral margin; flattened and expanded anterior process of jugal; stepped postorbital/jugal bar; *short and ventrally keeled cervicals without postzygodiapophyseal laminae; *elongated and ventrally grooved caudals with an accessory neural spine and a postspinal lamina; chevron bones with single articular facet; pubis without pubic foot; lunate depression on distal part of fibula; body osteoderms. RAUISUCHUS TIRADENTES VON HUENE, 1938A Lectotype BSPG AS XXV 60 68, , , 121. Right premaxilla, right postorbital, left squamosal, left jugal, right pterygoid, right nasal, both splenials, left DESCRIPTION The lectotype material of R. tiradentes consists of cranial and postcranial elements (Fig. 1). The preservation is generally rather poor. Most of the material is Figure 1. Preserved skeletal elements of Rauisuchus tiradentes.

5 OSTEOLOGY OF RAUISUCHUS TIRADENTES 59 completely disarticulated, and in parts broken or distorted. Delicate surface structures and bone sutures are only rarely visible. The pronounced alteration of the size and shape of some elements is a result of replacement by calcite and haematite, giving the respective bones the swollen appearance that is typical of fossils from the Santa Maria Formation (Holz & Schultz, 1998). SKULL Premaxilla The main body of the right premaxilla (Fig. 2) is nearly rectangular and only slightly longer than high. Two processes, the anterior ascending process and the posterior process, extend from the main body. Unlike the condition in Batrachotomus kupferzellensis Gower 1999 (SMNS 52970), large nutrient foramina on the premaxillary body are absent, and the narial fossa is present only as a small inflection of the lateral wall below the external nares. On the lateral surface, at the base of the posterior process, the premaxilla has a prominently bulging protuberance, above which the bone surface seems to be more notably rugose than on the premaxillary body. A similar protuberance on the ascending process, as found in Decuriasuchus quartacolonia (De França et al., 2013), is absent in R. tiradentes. The anterior ascending process is mostly complete, and forms the anterior and anterodorsal border of the external naris. Its basal portion is more or less vertical and the distal part abruptly flexes posterodorsally at an angle of approximately 60 towards the basal part, which gives the tip of the snout a blunt shape. The process is rather thin, much longer anteroposteriorly than wide transversely, and has a flattened medial surface for the contact with the process of the other premaxilla. The distal tip of the anterior ascending process bears a small triangular facet laterally, for the articulation with the nasal. The process is very thin in this region and its distalmost portion is missing. The posterior process, forming the ventral and partly the posterior border of the external naris, is at least 1.5 times the length of the premaxillary body, although a few millimetres might be missing distally. It tapers towards its posterior end, where it meets the descending process of the nasal. Its cross section is semicircular with a medially flattened surface, wider than high, and with a flat posteroventral side that overlies the anterior margin of the ascending process of the maxilla. On its dorsomedial surface, there is a small, posterodorsally widening facet that extends over approximately 10 mm anteroventral to the distal end of the process. It obviously represents the contact with the descending process of the nasal, indicating that the maxilla was excluded from the narial margin, unlike the condition in Batrachotomus kupferzellensis (Gower, 1999). As noted by von Huene (1942), the lateral side of the process shows a roughened bone surface, whereas the anteromedial narial border is smooth. Because the maxilla of R. tiradentes is missing, the presence or absence of a subnarial fenestra and a possible kinetic premaxilla maxilla joint, which has been proposed as a characteristic feature of either Rauisuchia Figure 2. Right premaxilla (BSPG AS XXV 60) of Rauisuchus tiradentes in lateral view. Abbreviations: a.n, articulation with nasal; aap, anterior ascending process; pmxp, premaxillar protuberance; pp, posterior process; snf, subnarial fenestra.

6 60 S. LAUTENSCHLAGER AND O. W. M. RAUHUT (e.g. Parrish, 1993; Long & Murry, 1995) or a more inclusive taxonomic subunit (e.g. Benton & Clark, 1988), cannot be determined with certainty. Furthermore, the distribution of this character within Crurotarsi is not yet completely understood (Gower, 2000); however, the facet for the maxilla on the posterior process ends slightly above the base of the premaxilla, which would suggest a firm contact between premaxilla and maxilla up to this point. There is only a slight indention below the base of the posterior process on the posterior margin of the premaxilla, which might mark the anterior margin of the subnarial fenestra. In Batrachotomus kupferzellensis (SMNS 52970; Gower, 1999) and Postosuchus kirkpatricki (Weinbaum, 2002), however, the subnarial fenestra is nearly entirely formed by a small recess on the anterior margin of the maxilla, without any obvious contribution of the premaxilla. As a result, a subnarial fenestra was not necessarily absent in R. tiradentes. Despite the absence of a maxilla, Nesbitt (2011: character 12) regarded the aforementioned indentation as conclusive for a subnarial foramen in R. tiradentes. Two teeth, presumably numbers two and four, are preserved. They are comparably short and project only a short distance from the oral margin of the premaxilla, so that these may be replacement teeth that are not yet fully erupted. The anterior and posterior carinae are finely serrated with approximately 25 denticles per 5 mm. The dimensions and positions of the teeth, as well as the length of the premaxilla, indicate that there were no more than four teeth present, as is the case in most rauisuchoids (Sill, 1974; Bonaparte, 1981; Gower, 1999; Weinbaum, 2002; De França et al., 2013). As preserved, neither the shape nor the number of alveoli can be established with certainty, but a total number of six alveoli, as suggested by von Huene (1942: 203) seems highly unlikely. Subsequent preparation of the medial side revealed the remains of the median symphyseal facet. The palatal process is broken and not preserved. The area below the base of the palatal process shows two deep, nearly horizontal excavations. Chatterjee (1985) described a similar structure in Postosuchus kirkpatricki and suggested an association with olfactory nerves. Remains of a medial shelf extend over the anterior three-quarters of the bone. At its posterior end, just below the base of the posterior ascending process of the premaxilla, there is a pronounced, posteriorly opening pit of some 5 mm in diameter, which most probably marks the articulation with an anteromedial process of the maxilla, as in Batrachotomus kupferzellensis (Gower, 1999). The alveolar border is too poorly preserved to say anything about the possible presence and shape of interdental plates. The general appearance and outline of the premaxilla, with its blunt snout and long posterior process, resemble that of Saurosuchus galilei Reig, 1959 (Alcober, 2000) and Postosuchus kirkpatricki (Weinbaum, 2002). These taxa also share the nearly quadrangular morphology of the premaxilla base, which is not seen in other rauisuchian taxa such as Batrachotomus kupferzellensis (SMNS 80260, SMNS 52970; Gower, 1999) that possess a more elongated premaxilla. Nasal The posterior part of the right nasal (Fig. 3A, B) is long and straight, and faces mainly dorsally, whereas its anterior part is strongly curved anteroventrally and splits into two processes. The anterior process is much longer than the posteroventrally located process and is mostly complete, with only the anteriormost part missing. It obviously articulated with the anterior ascending process of the premaxilla, with the comparably massive anterior end laterally overlying the thin posterodorsal end of the ascending process of the premaxilla. The more posteriorly located descending process is incomplete and forms an acute angle with the anterior process, being almost parallel with the anterior process. This process is massive and considerably wider than high at its distal break. It has a flat posteroventral surface, with a slight longitudinal groove medially that is flanked by a small ridge laterally, obviously forming a stout suture with the ascending process of the maxilla. A prominent external narial fossa, located ventrally between the anterior and the descending processes, as seen in Batrachotomus kupferzellensis (SMNS 52970; Gower, 1999), cannot be recognized in R. tiradentes, although a slight depression might have been present. Because the maxilla of R. tiradentes is not preserved, the exact articulation between the nasal and the maxilla cannot be established; however, the massive shape and outline of the posterior process of the premaxilla and the small facet described above suggest that there was a strong contact between it and the descending process of the nasal. The external naris was thus probably completely framed by the premaxilla and the nasal. This morphology closely resembles the naris of Saurosuchus galilei (Alcober, 2000) in overall appearance, being considerably longer than high and tapering posteriorly. In Saurosuchus galilei the maxilla only contacts the lateroventral margin of the nasal, and the same condition can be inferred for R. tiradentes. The posterior end of the nasal is almost complete; only the posteriormost part with the suture to the frontal is missing. The dorsal surface of the nasal is wide and flattened, and has only a weakly developed lateral ridge, in contrast to the very pronounced rugose ridges in Batrachotomus kupferzellensis (SMNS 52970; Gower, 1999) and Postosuchus (Chatterjee, 1985; Peyer et al., 2008). The dorsal surface gradually widens from the anterior end posteriorly up to the suture with the lacrimal and prefrontal, which, in dorsal view, is marked

7 OSTEOLOGY OF RAUISUCHUS TIRADENTES 61 Figure 3. Right nasal (BSPG AS XXV 65) of Rauisuchus tiradentes in lateral view (A), and dorsal view (B). Abbreviations: a.l/prf, articulation with lacrimal/prefrontal; adp, anterior descending process; lr, lateral ridge; pdp, posterior descending process. as a lateral constriction of the nasal (Fig. 3B). Approximately beginning at this point, the medial side of the nasal is slightly arched dorsally, although this might be exaggerated by preservation. The main body of the nasal is very massive and forms a plane butt joint with the other nasal medially. The bone only becomes slightly thinner towards the posterior end. The lateral surface of the nasal slopes medioventrally some 30 mm posterior to the base of the descending process, forming the facet for the lateroventral

8 62 S. LAUTENSCHLAGER AND O. W. M. RAUHUT articulation with the lacrimal. As in Batrachotomus kupferzellensis (Gower, 1999: figs 6A, 7), the anteriormost part of this suture is marked as a dorsally expanded, semicircular facet, at which the laterodorsal margin of the bone is slightly raised. The lateral constriction of the nasal in dorsal view is placed immediately posterior to this expanded facet. The nasal of R. tiradentes differs from that of most other rauisuchian taxa in the marked ventral flexure of its anterior half. In Batrachotomus kupferzellensis (Gower, 1999), Saurosuchus galilei (Alcober, 2000), Postosuchus kirkpatricki (Weinbaum, 2002), and Polonosuchus silesiacus (Sulej, 2005) (compare also Brusatte et al., 2009) the bone is straight over most or all of its length in lateral view. Given the morphology of the nasal and the premaxilla, the snout of R. tiradentes must have been long and narrow, making up at least 30 to 40% of the total length of the skull. Postorbital A slender, elongate bone that von Huene (1942: 199) considered problematic was described by him as the left postfrontal, with a small part of the postorbital attached. However, the line interpreted as a suture by von Huene (1942: pl. 24, fig. 2) rather seems to be an overprepared break, judged by the continuing surface structure across this line on the posterior part of the medial side. The element is unusual in its morphology, but we interpret it as the right postorbital (Fig. 4A C). The only alternative interpretation would be that as a partial left lacrimal, but the lack of a lacrimal foramen and the lack of an expanded anterior process, as well as the tapering rather than expanded ventral end argue against such identification. The bone consists of a prominent ventral process and an indicated, although incomplete posterior process. The anterior process is completely missing, but the Figure 4. Right postorbital (BSPG AS XXV 61) of Rauisuchus tiradentes in dorsal view (A, anterior to the left), right lateral view (B), and medial view (C). Abbreviations: a.j, articulation with postorbital facet of the jugal; a.pf/f, articulation with postfrontal/frontal; a.sq, articulation with squamosal; ap, anterior process; ld, longitudinal depression; pd, posterodorsal depression; vp, ventral process.

9 OSTEOLOGY OF RAUISUCHUS TIRADENTES 63 striated anterior surface of the dorsal part of the anterior margin is reminiscent of a suture rather than a break, so that this process might have been very short, as in Arizonasaurus babbitti Nesbitt 2005, and shows a massive suture with the postfrontal [assuming that the postfrontal had a similar position as in Batrachotomus kupferzellensis (Gower, 1999), rather than the more medial position seen in some other noncrocodylomorph pseudosuchians (Weinbaum, 2002; Sulej, 2005)]. Just medial to this sutural surface, a smaller anterior process along the anterolateral margin of the supratemporal fenestra, as present in Batrachotomus kupferzellensis (Gower, 1999), might be broken off. The dorsal surface of the bone is flat and slightly depressed, with a raised medial shelf or ridge along the margin of the supratemporal fenestra. This margin is developed as a slightly dorsoventrally expanding, narrow, medially-facing surface anteriorly, which gently twists into a dorsomedially-facing facet posteriorly. As preserved, the dorsal surface is approximately trapezoidal in shape and narrows from the medial side posteriorly. Anteriorly, its medial part considerably overhangs the base of the ventral process. Because the anterior and the posterior processes are missing, no conclusions can be drawn regarding the articulations with the frontal anteriorly and the squamosal posteriorly. The ventral process tapers towards its end and is directed slightly anteriorly, although it is not kinked or stepped, to meet the ascending process of the jugal. The ventral process forms the posterior border of the orbit and the anterior margin of the infratemporal fenestra. It is relatively straight and resembles the postorbital of Saurosuchus galilei (Alcober, 2000), which also possesses a straight ventral process as opposed to the anteriorly kinked condition of the ventral process seen in Batrachotomus kupferzellensis (Gower, 1999) and Postosuchus kirkpatricki (Chatterjee, 1985). The lateral surface of the ventral process bears a thick, slightly rugose, triangular area dorsally. It becomes gradually lower ventrally and fades into the smooth surface of the transversely thin ventral part of the process at about half its length. The posterior border of this thickened area is marked by a concave, posterolaterally-facing depression along the margin of the lower temporal fenestra. It curves posterodorsally along the gradual transition, ventral to the posterior process of the bone. The lateral side of the ventral half of the ventral process is smooth and slightly convex anteroposteriorly. In medial view, the posterior margin of the ventral process is deflected medially, resulting in the formation of a longitudinal depression along the process, which becomes more prominent dorsally. The ventral end of the process abruptly narrows posteriorly some 15 mm above the ventral tip, which might be missing a few millimetres. This part of the postorbital obviously overlapped the postorbital facet of the jugal anterolaterally. Although the postorbital of R. tiradentes (if identified correctly) is highly unusual in its morphology, it shows some similarities with the postorbitals of Arizonasaurus babbitti (Nesbitt, 2005) and Batrachotomus kupferzellensis (SMNS 80260; Gower, 1999). In addition, the element is distinctly different from the lacrimal of Postosuchus kirkpatricki (Chatterjee, 1985), Batrachotomus kupferzellensis (SMNS 80260; Gower, 1999), and Teratosaurus silesiacus (Sulej, 2005) in lacking an anterior process and a pronounced anterior posterior expansion. Squamosal The preserved squamosal (Fig. 5A, B) is from the left rather than the right side (contra von Huene, 1942: 198). It has four processes directed approximately anteriorly, anteromedially, ventrally, and posteriorly. The anterior process projects slightly ventrally and laterally, and only its base is preserved. The anterior end with the socket for the reception of the postorbital is missing. The dorsolateral margin of the anterior process is strongly thickened and forms a raised, rugose ridge along the lateral margin of the supratemporal fenestra, overhanging the dorsolateral part of the squamosal. Gower (1999) described a similar structure in Batrachotomus kupferzellensis as a dorsolateral brow. Below this ridge, the body of the squamosal forms a ventrolaterally extended shelf, although this might be slightly exaggerated by deformation of the generally poorly preserved bone. As preserved, the anteromedial process turns slightly in the medial direction and is as long and prominently developed as the anterior process, the medial end of which, and thus the suture with the parietal, is missing. The poorly preserved posterior side forms a vertical to slightly dorsoposteriorly inclined, flat surface for the articulation with the lateral process of the parietal and the paroccipital process of the exoccipital/ opisthotic. The form of articulation with the parietal is unknown. Together with the anterior process, the anteromedial process forms a major part of the posterolateral border of the upper temporal fenestra. The two processes are set at an acute angle of approximately 45 to one another. Because of the positions of these two processes, the upper temporal fenestra is located further laterally than in other rauisuchians, with the exception of Saurosuchus galilei (Alcober, 2000). The ventral process is comparably thin transversely, flattened and slightly convex in lateral view, and correspondingly concave in medial view. The complete outline of the process, particularly its anterodorsal margin, is unclear. Nesbitt (2011: fig. 11) reconstructed this process as anteriorly enlarged, contacting the

10 64 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 5. Left squamosal (BSPG AS XXV 62) of Rauisuchus tiradentes in dorsolateral view (A) and ventral view (B). Abbreviations: a.q, articulation with quadrate; amp, anteromedial process; ap, anterior process; db, dorsolateral brow; pp, posterior process; sqn, squamosal notch; utf, upper temporal fenestra; vp, ventral process. ascending process of the jugal, partitioning the infratemporal fenestra. Although this condition is present in Postosuchus kirkpatricki (Chatterjee, 1985) and Polonosuchus silesiacus (Sulej, 2005), neither the jugal nor the squamosal of R. tiradentes show any indication for such morphology. In fact, the anterior and ventral processes of the squamosal are set off by a distinct notch on the posterior border of the infratemporal fenestra. A similar notch is found in Saurosuchus galilei (Alcober, 2000), where the ventral process perforates the infratemporal fenestra, but does not bisect it. The same morphology thus seems more likely for R. tiradentes. The posterior process is inconspicuous and bears a small, spike-like projection. On the ventral surface, the squamosal possesses a deep depression for the reception of the head of the quadrate. It is bordered by a three-pronged structure, formed by the anterior and posterior ends of the anteromedial process and the posterior process. The dorsal part of the quadrate was therefore covered by the lateral overhang of the squamosal. In dorsolateral view the squamosal of R. tiradentes resembles that of Batrachotomus kupferzellensis (SMNS 80260, SMNS 52970) and Postosuchus kirkpatricki (Weinbaum, 2002),

11 OSTEOLOGY OF RAUISUCHUS TIRADENTES 65 in particular in the morphology and presence of the dorsolateral brow. The squamosal of R. tiradentes is relatively enlarged regarding the other preserved cranial elements. By contrast the squamosal in almost all other known basal pseudosuchians are in comparison often very small and gracile (Weinbaum, 2002; De França et al., 2013), and make only a minor contribution to the skull roof. By comparison with the examples given by Holz & Schultz (1998) in their taphonomic study on fossils from the Santa Maria Formation, it seems likely that the squamosal is also bloated, which accounts at least in part for the increased size. The heavily deformed bone texture supports this assumption. Jugal The left jugal (Fig. 6A, B) is triradiate, presenting three distinct processes: an anterior process contacting the maxilla and/or the lacrimal; a dorsal ascending process joining the postorbital; and a posterior process articulating with the quadratojugal. The anterior process is directed slightly dorsally and is transversely flattened and dorsoventrally expanded. Laterally, it bears a large, but shallow, triangular posteriorly pointed depression, which is separated from the also depressed ventral margin by a thin, anteroposteriorly extending ridge. This structure continues on the main body and forms a low, horizontal ridge that becomes broader and more rounded posteriorly until it fades into the lateral surface of the quadratojugal process. A similar ridge is seen in the jugals of Postosuchus kirkpatricki (Long & Murry, 1995; Weinbaum, 2002) and Arizonasaurus babbitti (Nesbitt, 2005), where it was described as a rugose ridge. In both Postosuchus and Arizonasaurus this rugose ridge continues on the maxilla, which might also have been the case in R. tiradentes, although the ridge becomes notably lower towards its anterior end. The depressed facet ventral to it on the anterior process of the jugal probably represents the facet for the articulation with the maxilla, as is the case in Batrachotomus (Gower, 1999: fig. 13B). On the medial surface the anterior process of the jugal has a medially thickened dorsal margin and a notable step at about its mid-height. The area between these structures is slightly depressed, forming a subrectangular facet, possibly for the overlap of the ventral end of the lacrimal, but as neither the lacrimal nor the maxilla are preserved, the exact contact between these elements and the jugal is unknown. Gower (1999) described a corresponding, longitudinally striated shelf on the lateral surface of the anterior process of the jugal in Batrachotomus kupferzellensis, which articulates with the maxilla and a lightly striated medial area that articulates with the lateral surface of the ventral tip of the preorbital bar of the lacrimal. The dorsal process projects slightly posteriorly and tapers towards its distal end, the tip of which is missing. On the upper part of its lateral surface, the dorsal process bears an angular facet for the articulation with the ventral process of the postorbital. This facet occupies the entire breadth of the dorsal end of the process, but narrows towards the anterior margin more ventrally, where it flexes onto the anterior surface of the process and ends some 40 mm below the dorsal tip. Along its anterior margin, the dorsal process shows a characteristic step, where the facet for the postorbital ends at the anterior margin of the process a character that has traditionally been employed to define Rauisuchidae (e.g. Benton & Clark, 1988; Juul, 1994). The medial surface of the dorsal process possesses a prominently developed, thin, dorsoventrally extending ridge that arises dorsally from the medial flexure of this facet. This ridge continues to the base of the main body and separates it centrally. On the medial side of the main jugal body, this ridge defines a deep, dorsoventrally elongate, but transversally narrow, anteromedially facing depression. The ridge and thus the border of this depression flexes anteriorly some 20 mm above the ventral margin of the main body of the jugal. A small dorsoventral rib extends ventrally from the ventral margin of the depression and defines the anterior rim of a small, anteroposteriorly concave area on the ventral part of the medial side of the jugal body, below the dorsal process. This concave area probably represents part of the articulation for the ectopterygoid, as is the case in Batrachotomus (Gower, 1999) and Postosuchus (Weinbaum, 2002). The ventral margin of the main body is strongly convex ventrally, so that both the anterior and posterior processes are directed somewhat dorsally. Together, the anterior and dorsal processes form a large part of the posteroventral border of the orbit. The posterior process is curved and upturned and, compared with the rest of the jugal, is more robust and rounded. Its posterior end is missing. It forms the lower margin of the infratemporal fenestra. Laterally, the posterior process is convex and smooth. The rugose ridge on the main body of the jugal does not continue on the posterior process, but merges with the latter process. Medially the posterior process is slightly flattened, but maintains its circular cross section. Compared with other basal pseudosuchians, the jugal of R. tiradentes differs notably from that of Batrachotomus kupferzellensis. In R. tiradentes, the posterior process is tapering and curved, whereas it is massive and straight in Batrachotomus kupferzellensis (SMNS and SMNS 80260). Batrachotomus kupferzellensis also lacks the distinct rugose ridge along the lower lateral side of the main body of the jugal (SMNS and SMNS 80260). This structure is also absent in Saurosuchus galilei (Sill, 1974; Alcober, 2000),

12 66 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 6. Left jugal (BSPG AS XXV 63) of Rauisuchus tiradentes in lateral view (A) and medial view (B). Right pterygoid (BSPG AS XXV 64) of Rauisuchus tiradentes in ventral view (C) and lateral view (D). Abbreviations: a.bpt, articulation with basipterygoid; a.ec, articulation with ectopterygoid; a.mx, articulation with maxilla; a.po, articulation with postorbital; ap, anterior process; apr, anterior palatal ramus; d, depression; dp, dorsal process; epw, ectopterygoid wing; ltf, lower temporal fenestra; mcr, medial crest; o, orbit; pp, posterior process; pqr, posterior quadrate ramus/quadrate wing; rr, rugose ridge; sjb, stepped jugal bar.

13 OSTEOLOGY OF RAUISUCHUS TIRADENTES 67 but is present in Postosuchus kirkpatricki (Weinbaum, 2002) and Arizonasaurus babbitti (Nesbitt, 2005). As in R. tiradentes, this rugose ridge is restricted to the main body and parts of the anterior process of the jugal in these taxa. Pterygoid The element described as a right prefrontal by von Huene (1942: 201) is in fact a right pterygoid (Fig. 6C, D), which is somewhat distorted and poorly preserved. The triradiate bone consists of an anterior palatal ramus, the ventrolaterally directed ectopterygoid wing, and the posteriorly directed quadrate wing (which is largely missing). The anterior palatal ramus is elongated. It tapers anteriorly, but the anteriormost part is missing. In ventral view, the anterior portion is narrow transversely, but widens towards its base posteriorly. The medial edge is angled sharply in the dorsal direction and forms a high vertical medial crest, along which both pterygoids were originally joined. This crest also bears a depression running parallel with the anterior palatal ramus, which von Huene (1942: 201) erroneously interpreted as forming the dorsal margin of the orbit. The medial side of the crest is flattened anteriorly, but bears a longitudinal depression in its dorsal part posteriorly. The ventral side of the main pterygoid body and the base of the palatal ramus is slightly inclined dorsomedially. It bears a large, shallow depression posteriorly, which becomes narrower and medially sharply defined on the base of the palatal ramus, but fades more anteriorly. At the posterior end the pterygoid splits into a ventrolaterally orientated ramus joining the ectopterygoid and a posterior quadrate ramus, neither of which are completely preserved, although the ectopterygoid wing is only missing minor portions of its posterolateral extremity and anteromedial margin. The ectopterygoid wing forms a triangular flange and is as thin and flattened as the anterior palatal ramus. von Huene (1942: 201) considered this lateral ramus to represent a lateral process of the prefrontal, contacting the lacrimal. The posterior part of the ectopterygoid wing is dorsoventrally thickened, although the posterior margin forms a sharp edge. Anterolateral to the thickened part, a depression on the dorsal side of the pterygoid marks the facet for the overlap by the ectopterygoid. In ventral view, the flat ventral surface of the ectopterygoid wing is separated from the main pterygoid body by a notable step. The basipterygoid articulation expands slightly posterior from the point where the ectopterygoid wing meets the main pterygoid body. Here, the pterygoid is most massive and has a short, but stout longitudinal ventral ridge. The articulation for the basipterygoid process of the basisphenoid is developed as a narrow, posteriorly opening groove that is semicircular in outline, lateroventrally orientated, and opens dorsally. Thus, assuming that the basipterygoid process was directed lateroventrally, not or only insignificantly anteriorly, the orientation of this facet indicates that, in the articulated skull, the main body and palatal process of the pterygoid sloped considerably anterodorsally, as is the case in most archosaurs. In front of the basipterygoid articulation facet, the median crest of the main pterygoid body is drawn out dorsally into a small, triangular dorsal process. The quadrate wing of the pterygoid starts at the lateroventral end of the basipterygoid articulation. Its base is dorsoventrally low and ventrally continuous, with the ridge ventral to the basipterygoid articulation. The wing expanded posterodorsolaterally, but is mostly missing. MANDIBLE The left surangular, prearticular, and articular are preserved in their original articulation, although the articular seems to be considerably deformed and incomplete. In addition, both splenials are preserved in isolation. Splenial The splenials (Fig. 7A, B) are mostly complete, the left element more so than the right, and complement each other in the missing portions. They are elongated and form thin, laminar sheets of bone. Posteriorly the splenial is bifurcated, with a small, spike-like dorsal posterior process (missing in the left splenial) and a much longer ventral posterior process. The ventral margin is thickened and its ventrolateral surface is turned upwards, forming an overlapping ridge, which envelops the ventral border of the dentary and forms the ventral and medial wall of the Meckelian canal. In the right splenial this overlapping ridge is slightly flattened and noticeably grooved lateroventrally, possibly for the articulation with the dentary and the angular. The ridge runs all the way along the posterior process in both elements, but becomes gradually lower towards the end. Anteriorly, the ridge becomes a ventral shelf. A second depression, probably also for the reception of the dentary, is located near the dorsal edge on the lateral surface. An articular facet for the coronoid cannot be identified on either side. Unlike the condition in Postosuchus kirkpatricki (Chatterjee, 1985) and Prestosuchus chiniquensis (BSPG AS XXV 1), the splenials of R. tiradentes lack any evidence for an anterior mylohyoid foramen; however, the left splenial has a distinct notch in the anterior end, which is also seen in the splenial of Prestosuchus chiniquensis (BSPG AS XXV 1), and may be equivalent to the foramen for the mandibular ramus of the trigeminal nerve.

14 68 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 7. Mandibular elements of Rauisuchus tiradentes. Right splenial (BSPG AS XXV 66) in lateral view (A) and medial view (B). Left surangular, prearticular, and articular (BSPG AS XXV 68) in lateral view (C) and medial view (D). Abbreviations: a.an/de, articulation with angular/dentary; a.de, articulation with dentary; m. AMP, attachment site for musculus adductor mandibulae posterior; m. PTd, attachment site for musculus pterygoideus dorsalis; ar, articular; emf, external mandibular fenestra; igr, interglenoid ridge; lg, lateral glenoid fossa; mc, Meckalian canal; mg, medial glenoid fossa; pa, prearticular; pp, posterior process; sa, surangular. Compared with other taxa, the splenials of R. tiradentes are elongated and more similar to those of Batrachotomus kupferzellensis (SMNS 80260; Gower, 1999) than those of Postosuchus kirkpatricki (Chatterjee, 1985), with their triangular shape. The left splenial resembles that of Prestosuchus chiniquensis (BSPG AS XXV 1) in general shape as well as in its dimensions, although this bone is even more elongate and has a suture with the other splenial anteroventrally in the latter taxon. Surangular The surangular (Fig. 7C, D) is an elongate element with a triangular or wedge-shaped cross section above the obviously large external mandibular fenestra and a vertical plate in its posterior part. The dorsal part is thickened and forms a prominent, medially and laterally overhanging shelf, especially in the posterior part in front of the articular. Dorsally, this shelf is flattened and bears a broad depression, although this might be

15 OSTEOLOGY OF RAUISUCHUS TIRADENTES 69 amplified by the preservation. In its anterior extreme, the surangular becomes thin and sheet-like, with a thin dorsal margin. The tip has a V-shaped notch on the anterolateral surface for the reception of the posterior end of the dentary. Towards the anteroventral margin, which forms the upper border of the lateral mandibular fenestra, the posteroventral plate is mediolaterally compressed and becomes gradually thinner. The posteroventral plate has a well-developed, sharply pointing, triangular anterior process ventrally, so that it forms the posterior border and the posterior part of the ventral border of the mandibular fenestra. The medial surface of the surangular is largely featureless. A surangular foramen as described for Batrachotomus kupferzellensis (Gower, 1999) and Postosuchus kirkpatricki (Chatterjee, 1985; Weinbaum, 2002, 2013) cannot be found in R. tiradentes, although this might result from the poor preservation. Posteriorly, the surangular expands transversely towards the anterior margin of the jaw articulation. Unfortunately, no suture between the surangular and articular is visible, so to what extent the surangular participated in the jaw articulation cannot be determined. The broad and flattened dorsal surface of the surangular most probably served as an insertion for the musculus adductor mandibulae externus superficialis (m. AMES) laterally and the musculus adductor mandibulae externus medialis (m. AMEM) medially, indicating a similar myology as in Alligator mississippiensis (Busbey, 1989). The comparably large muscle attachment site suggests the presence of welldeveloped jaw adductor muscles in R. tiradentes. Prearticular The prearticular (Fig. 7C, D) is strongly curved and has a medially convex cross section similar to that of Batrachotomus kupferzellensis (SMNS 52970, SMNS 80260). The anterior part of the element is turned dorsally and has a thin, spatula-like appearance, the dorsal end of which reaches the level of the ventral margin of the anterior process of the surangular. Posteriorly it becomes considerably broader and expanded. The ventral border is thickened. The medial wall is low and contacts the articular posterodorsally, whereas the lateral wall continues dorsally to join the surangular. Between the medial and the lateral wall runs a deep groove, probably housing Meckel s cartilage, as inferred for Batrachotomus kupferzellensis (Gower, 1999). Articular The articular (Fig. 7C, D) is for the most part poorly preserved and considerably deformed, as von Huene (1942: 204) already noted. It is closely attached to the surangular and both elements have been tightly amalgamated during diagenesis, so that a suture between them is no longer visible. Thus the extent of the articular in the mediolateral direction seems to be very broad, although this is amplified by the deformation. Dorsally, the articular bears a large facet for the jaw articulation, which starts at the anteromedial margin of the articular and continues posterolaterally. The facet consists of two depressions, which are separated by a small ridge. The lateral depression is anteroposteriorly elongated and narrow, and extends further anteriorly than the medial depression. The latter is anteroposteriorly shorter, but approximately twice as wide transversely. These facets articulated with the paired condyles of the ventral part of the quadrate. Because of the poor preservation, it remains unclear whether the surangular contributed to the mandibular joint, as in Batrachotomus kupferzellensis (Gower, 1999) or Postosuchus kirkpatricki (Chatterjee, 1985). At its posterior margin, the articular facet for the quadrate bears a slightly elevated rim, which continues medially and grades into the dorsal part of the ascending process of the retroarticular region. The medial surface bears the ascending process and an adjacent posterolateral depression. The ascending process is developed as a dorsally sloping tubercle at the posterior end of the medial articular facet, and is triangular in medial view. The depression is anteroposteriorly elongate and triangular in outline, tapering anterolaterally. There seems to be one or two large foramina present in the anteromedial part of the depression, as in Batrachotomus (Gower, 1999), although some uncertainty remains because of the poor preservation. In dorsal view, the region of the ascending process and this depression form a triangular medial tubercle, as is also present in other basal pseudosuchians, although in contrast to Batrachotomus, the tubercle is less well-developed and confluent with the retroarticular process posteriorly. In medial view, there is a pronounced, rounded triangular ventral process below the ascending process, as in Postosuchus (Chatterjee, 1985; Weinbaum, 2002). A large, rounded depression is present on the anterior part of this process, its posterior rim extending vertically dorsally to the highest point of the ascending process. The ventral margin of the depression is formed by the medially thickend ventral margin of the process. The area lateral to the dorsal depression on the medial tubercle is poorly preserved, but seems to have been somewhat concave. This area bears a depression for the attachment of the musculus depressor mandibulae in Batrachotomus (Gower, 1999). In ventral view, the articular tapers to a blunt and gently rounded point. This part is offset from the main body of the articular and resembles that of Postosuchus kirkpatricki (Chatterjee, 1985). Anteriorly the articular connects to the surangular and prearticular via a diagonally inclined, prominent, and rounded ridge, which forms the posteromedial wall of the mandibular adductor

16 70 S. LAUTENSCHLAGER AND O. W. M. RAUHUT chamber. Posteriorly the articular reduces in size and becomes sagittally constricted. The strongly developed and inclined ridge on the anterior margin of the articular indicates a firm attachment of the musculus adductor mandibulae posterior (m. AMP) along its length, as well as the insertion of the musculus pterygoideus dorsalis (m. PTd) anteroventrally (compare Busbey, 1989). Dentition Apart from the two teeth in the premaxilla, 39 isolated teeth were collected and referred to R. tiradentes by von Huene (1942: 203). They are posteriorly recurved, mediolaterally compressed, and have serrated carinae anteriorly and posteriorly, as it is typical for carnivorous archosaurs. Their size ranges from 7 to 26 mm, with a varying morphology from small, broad, and blunt shapes to long, slender, and pointed forms. These teeth are similar to those preserved in the premaxilla, whereas the remaining teeth particularly the bigger specimens might be derived from the maxilla. The serrations on the carinae are very fine, with denticles per 5 mm in all teeth. The denticles are chiselshaped and perpendicular to the long axis of the carinae. The mesial carina extends to the base of the crown in many but not all specimens; in some it ends a few millimetres above the crown root junction. As demonstrated by Batrachotomus kupferzellensis (Gower, 1999), such variations in shape and size of the various teeth are not uncommon in rauisuchians. Nevertheless, none of these teeth hold any diagnostic value, so that they cannot be referred to R. tiradentes with certainty. Some might have been derived from scavenging carnivorous archosaurs, especially as many seem to have been shed, as evidenced by well-developed resorption pits at their bases. On the other hand, there are several teeth present that are complete and the preserved elements of R. tiradentes show no obvious signs of scavenging, such as bite or gnaw marks. As the respective sediments were deposited in a floodplain facies (see geological setting), an accumulation of these teeth by water currents can most probably be ruled out, also because a more diverse range of elements would then be expected. POSTCRANIAL SKELETON Vertebral column Several vertebrae, including cervicals, dorsals, and caudals, of R. tiradentes are preserved, although many of them are in such poor condition that only a few are suitable for a detailed description. Apart from the vertebrae, ribs and chevron bones have also been found. For a description of the vertebral laminae the terminology of Wilson (1999), initially proposed for sauropod dinosaurs, is used. Cervical vertebrae According to von Huene (1942: ) all eight cervical vertebrae of R. tiradentes were collected; however, we consider the supposed intercentrum of the atlas described by von Huene (1942) as a probable odontoid, and the supposed atlantal neural arch cannot currently be identified as such (see above), so no certain remains of the atlas are preserved. The element identified as the atlas intercentrum by von Huene (1942) is triangular in lateral view (Fig. 8A, B); however, if oriented as argued by von Huene (1942: , pl. 26, fig. 5), the centrum is widest dorsally and tapers ventrally, which is in contrast to the situation in other archosaurs. Thus, the identification of this element is problematic. If oriented as in other archosaurs, with the widest part ventrally and narrowing dorsally, the element has a strongly convex Figure 8. Atlas axis complex of Rauisuchus tiradentes. Possible odontoid (BSPG AS XXV 72) in dorsal view (A) and anterior view (B). Axis (BSPG AS XXV 73a) in anterior view (C) and lateral view (D). Abbreviations: diap, diapophysis; poz, postzygapophysis.

17 OSTEOLOGY OF RAUISUCHUS TIRADENTES 71 ventrolateral surface, an anterior concavity, and a flat posterior side. In this orientation, however, the dorsal side has a small dorsal spike intruding into the neural canal, in contrast to the situation in other basal archosaurs, where this area is concave transversely (e.g. Postosuchus; Chatterjee, 1985; Stagonosuchus; Gebauer, 2004; Ticinosuchus, PIZ T 2817; O.W.M. Rauhut, pers. observ.). Thus, we tentatively identify this element as possibly representing the odontoid process, which has detached from the axis. The width of the bone coincides with the anterior width of the centrum of the axis and its height is about two-thirds of that of the axial centrum. In this orientation, the convex surface faces anteriorly, whereas the flat area represents the contact with the axis. The dorsal surface is concave transversely, with the extent of this concavity coinciding with the width of the neural canal of the axis. The concave facet ends anteriorly in a small, anteriorly pointing process. Posterodorsally, there is a slightly depressed facet on either side on the posterior surface. The axis (Fig. 8C, D) consists of the centrum and attached neural arch and spine. The centrum is parallelogram-shaped and only slightly waisted ventrally in lateral view. The anterior face is poorly preserved. It is slightly convex and situated higher than the concave posterior face. On the posterior part of the centrum, a small rim is developed, surrounding the posterior face of the centrum. The lateral sides of the centrum are notably rounded dorsoventrally, and bear an elongate, dorsally placed depression on both sides. The anterior and posterior rims are connected ventrally by a flattened area with pronounced longitudinal margins, forming a double-keeled structure. A neurocentral suture is clearly visible. A small ventrolateral projection on the anterior half of the neurocentral suture represents the diapophysis. Parapophyses are not visible; however, this could be the result of the poor preservation of the anterior half of the centrum. The neural canal is round in outline, its width (10 mm) being two-fifths of the posterior width of the centrum. The neural spine is incompletely preserved, with its posterodorsal part missing. In lateral view it has a typical triangular outline, with its dorsal margin steeply rising posterodorsally, as is also seen in Fasolasuchus tenax (Bonaparte, 1981) and Batrachotomus kupferzellensis (Gower & Schoch, 2009). The anterior part is thin and strongly developed. It extends beyond the anterior rim of the centrum and curves posteriorly above the neural canal. The posterior end of the spine is strongly widened and flared laterally, giving the spine a T-shaped outline in dorsal view. The posterior side is almost flat, only very slightly concave transversely, and has a vertical ridge along its midline. Prezygapophyses are absent, although this might be linked to preservation. Posteriorly, the postzygapophyses flare laterally at a wide angle to one another and their long axes are oriented almost perpendicular to the lateral sides of the neural spine, forming a triradiate outline in dorsal view. A weak medial ridge is present above the level of the postzygapophyses on the wide posterior side of the neural spine; this ridge corresponds to the postspinal lamina of Wilson (1999). The following six cervicals were tentatively designated and arranged by von Huene (1942: ), and the gradual increase in the size of the centra and neural spines, and the position of the diapophysis and parapophysis as far as preserved justify his identification. Of these six cervicals only the fourth (BSPG AS XXV 74), sixth (BSPG AS XXV 75) (Fig. 9A C), and eighth (BSPG AS XXV 76) (Fig. 9D F) are sufficiently well-preserved, and are described exemplarily. The third cervical mentioned by von Huene could not be found in the BSPG collection. The cervical centra are amphicoelous, although only slightly concave, comparatively short, and approximately as long as tall. The articular faces are oval in outline and higher than wide. The centra are notably waisted and bear a sharp mid-line keel on their ventral surface. Just below the neurocentral suture the centra bear a shallow, elongated depression on both sides, similar to the condition in Batrachotomus (Gower & Schoch, 2009), but less deep than in the poposaurids Sillosuchus (Alcober & Parrish, 1997) and, probably, Effigia and Shuvosaurus (Nesbitt, 2007, 2011). These depressions are in the same position as pneumatic fossae in dinosaurs (Wedel, 2007), although the presence of full postcranial pneumaticity in R. tiradentes is unlikely. The parapophyses are positioned on the ventral side of the anterior margin of the lateral surface of the centrum. On the sixth cervical they face lateroventrally, but are directed more laterally on the eighth cervical; their shape changes from almost round on the sixth cervical to high oval on the eighth cervical. The diapophyses are located on the dorsolateral surface of the centrum, and gradually move from a more central position towards the anterior edge in the more caudal elements. Whereas the diapophysis is entirely placed on the centrum in the sixth cervical, and has no connection to the structures of the neural arch, it is placed lateral to the neural canal in the eighth cervical. The diapophyses are strongly downturned and face lateroventrally, but are larger and more robust than the parapophyses, from which they are distinctly separated. The neural arches are high, with large neural canals. Neural arch lamination is rudimentary in the sixth cervical, in which only spinopre- and spinopostzygapophyseal laminae are developed as distinct structures, but is well developed in the eighth cervical. In this element, the diapophyses are connected to the pre- and postzygapophyses by stout pre- and postzygadiapophyseal laminae, and furthermore are

18 72 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 9. Cervical vertebrae of Rauisuchus tiradentes. Sixth cervical vertebra (BSPG AS XXV 75) in anterior view (A), right lateral view (B), and posterior view (C). Abbreviations: acdl, anterior centrodiapophyseal lamina; cpol, centropostzygapophyseal lamina; cprl, centroprezygapophyseal lamina; d, depression; diap, diapophysis; k, keel; parp, parapophysis; pcdl, posterior centrodiapophyseal lamina; poz, postzygapophysis; prdl, prezygdiapophyseal lamina; prz, prezygapophysis; spol, spinopostzygapophyseal lamina; tpol, intrapostzygapophyseal lamina.

19 OSTEOLOGY OF RAUISUCHUS TIRADENTES 73 supported ventrally by robust anterior and posterior centrodiapophyseal laminae that enclose a small infradiapophyseal fossa. Together with a thin centroprezygapophyseal lamina, the prezygodiapophyseal laminae defines a deep, triangular, mainly anteriorly directed infraprezygapophyseal fossa. A well-developed infrapostzygapophyseal fossa is present on the posterior end of the neural arch between postzygodiapophyseal lamina, posterior centrodiapophyseal lamina, and centropostzygapophyseal lamina, and faces posterolaterally. The neural spines, only complete in the fourth and eighth cervical, are high and anteroposteriorly short (height/length ratio > 4). The spine is vertical, very slightly bent anteriorly in cervical 6 and very slightly anteriorly inclined in cervical 8. The dorsal ends of the neural spines are rounded, but not expanded. Spine tables, as seen in Prestosuchus sp. (BSPG AS XXV 4, 13) or Stagonosuchus nyassicus von Huene, 1938b (Lautenschlager & Desojo, 2011) are absent. Only the left prezygapophysis is preserved in cervicals 6 and 8 (C6 and C8). In C6, the prezygapophyseal process is anterolaterodorsally directed, whereas in C8 it is only laterodorsally directed. The articular surfaces are wider than long and steeply angled, at about 50 from the horizontal in C6 and approximately 65 in C8, although the latter value might be slightly exaggerated as a result of compression. Between the prezygapophyses, there is a deeply recessed pit at the base of the neural spine in C6; this pit is absent in C8. The postzygapophyses have a similar shape as the prezygapophyses, but open at a slightly wider angle. In C6, the spinopostzygapophyseal lamina continues onto the dorsal surface of the postzygapophysis to form a low, mainly medially and posteriorly directed epipophysis; only a slight elevation is present in this position in C8. The eighth cervical also shows a weakly developed intrapostzygapophyseal lamina. From the anterior to the posterior cervicals there is a gradual increase in the size of the centra, although their relative dimensions do not notably change. The diapophyses and the parapophyses are consistently separated, whereas the diapophyses move dorsally. Compared with other rauisuchian taxa, the cervical vertebrae of R. tiradentes correspond to a general rauisuchid morphology, with short and ventrally keeled centra that are found in a wide range of taxa, such as Batrachotomus kupferzellensis (Gower & Schoch, 2009), Postosuchus sp. (Long & Murry, 1995; Peyer et al., 2008), and Polonosuchus silesiacus (Sulej, 2005). A ventral double keel of the axis has also been described for Postosuchus kirkpatricki (Long & Murry, 1995) and Postosuchus alisonae Peyer et al., In contrast, the cervical centra of Arizonasaurus babbitti (Nesbitt, 2005), Shuvosaurus inexpectatus (= Chatterjea elegans sensu Long & Murry, 1995), Sillosuchus longicervex Alcober and Parrish 1997, and Effigia okeeffeae (Nesbitt, 2007) share elongated cervicals, at least twice as long as high. The postzygodiapophyseal laminae in R. tiradentes are autapomorphically absent, contradicting Nesbitt (2007), who postulated the presence of these laminae in all rauisuchians. Apart from their size, the posterior cervicals, in particular of R. tiradentes, Stagonosuchus nyassicus (Gebauer, 2004; Lautenschlager & Desojo, 2011), and Fasolasuchus tenax (Bonaparte, 1981), are very similar. Dorsal vertebrae Seven dorsal vertebrae of R. tiradentes were collected. Two of them are preserved in natural articulation (BSPG AS XXV 119), but they are closely fused together by diagenesis and are incomplete. Of the remaining vertebrae neither is complete, although two of them BSPG AS XXV 77 (Fig. 10A C) and AS XXV 112 are sufficiently well preserved to allow a detailed description. von Huene (1942: 207) designated the two conjoined vertebrae as either dorsals 4 and 5 or dorsals 5 and 6. Compared with Ticinosuchus ferox Krebs 1965 (PIZ T 4779), the only rauisuchian taxon, where a nearly complete and articulated vertebral series is preserved, this placement seems to be plausible, given their size, dimensions, and the position of diapophyses and parapophyses, respectively. The remaining vertebrae are identified as coming from the mid-dorsal region. Their order is as follows: BSPG AS XXV 77, BSPG AS XXV 112, BSPG AS XXV 116 (an additional vertebra, BSPG AS XXV 99, is not included because of poor preservation). All dorsals possess amphicoelous centra, with nearly uniform dimensions, although they are slightly longer than tall. As in the cervicals, the dorsals are strongly waisted and have laterally and ventrally bulging anterior and posterior margins. The lateral sides of the centra are flat to gently convex and lack any fossa or recess. The ventral surface is rounded and a sharp mid-line keel is absent. The articular faces of the centra are rounded to oval in outline. In the anterior dorsals (BSPG AS XXV 112 and BSPG AS XXV 116), the anterior faces are considerably smaller than the posterior faces. The diapophyses are located on the transverse process. In more anterior vertebrae, the transverse process is strongly developed, but it becomes notably shorter in the more posterior dorsals. In BSPG AS XXV 119 the diapophyses are still separated from the parapophyses by approximately 20 mm, and the latter structure is placed on the neurocentral suture. In the remaining dorsal vertebrae the parapophysis is placed on the lateral surface of the neural arch and is located directly anteroventral to the diapophysis. The diapophyses are positioned approximately at the level of the dorsal margin of the

20 74 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 10. Dorsal vertebra (BSPG AS XXV 77) of Rauisuchus tiradentes in anterior.view (A), right lateral view (B) and posterior view (C). Abbreviations: cpol, centropostzygapophyseal lamina; cprl, centroprezygapophyseal lamina; diap, diapophysis; hyp, hyposphene; parp, parapophysis; poz, postzygapophysis; prpl, prezygoparapophyseal lamina; prz, prezygapophysis; spol, spinopostzygapophyseal lamina. neural canal and project almost completely laterally. With the exception of BSPG AS XXV 119, where the parapophyses are placed near the anterior rim, the parapophyses migrate towards the diapophyses and face laterally and slightly posteriorly. In contrast to the circular diapophyses, the parapophyses are oval in outline. Centroprezygapophyseal laminae are visible in BSPG AS XXV 119, BSPG AS XXV 77, and BSPG AS XXV 112. BSPG AS XXV 119 also clearly shows prezygodiapophyseal and paradiapophyseal laminae. The anterior of the two fused vertebrae of BSPG AS XXV 119 clearly shows a prezygoparapophyseal lamina that forms the anteroventral margin of the large, triangular infraprezygapophyseal fossa, and defines the laterodorsal margin of a longitudinal depression below the prezygapophysis, bordered medially by the centroprezygapophyseal lamina. The area between the parapophyses and the prezygapophyses in BSPG AS

21 OSTEOLOGY OF RAUISUCHUS TIRADENTES 75 XXV 77 and BSPG AS XXV 112 shows a lateral swelling, which obviously represents a remnant of such a prezygoparapophyseal lamina. A posterior centrodiapophyseal lamina is absent in all dorsals, and a postzygodiapophyseal lamina is only indicated by a lateral swelling in most elements. Spinoprezygapophyseal and spinopostzygapophyseal laminae are limited to the lower half of the neural spine in all preserved dorsals. The neural spines, completely preserved in BSPG AS XXV 77 and BSPG AS XXV 116 only, are approximately as high as in the cervicals, but are longer anteroposteriorly. At their base, they are slender and gradually widen transversally towards the dorsal end. The anteroposterior length of the spines increases from anterior dorsals to more posterior elements. In the middle and posterior dorsals, the bases of the neural spines start at the level of the mid-length of the vertebra centrum and project beyond its posterior margin. The lateral outline of the spines is roughly rectangular with slightly curved anterior and posterior margins. The complete spines of BSPG AS XXV 77 and BSPG AS XXV 116 slightly expand anteroposteriorly in their dorsal part, and the same was probably the case in the other elements. The distal ends of the neural spines are gently rounded. The prezygapophyses are set at a wider angle than in the cervicals and are comparatively short. Their lateral surface is convex and a rounded depression is present between them. They only project very slightly beyond the anterior rim of the centrum. In contrast, the postzygapophyses project posteriorly beyond the centrum so that their entire articular surface is placed posterior to the centrum. They form thin, sheetlike processes and are steeply angled, with a deep, posteriorly opening recess at the base of the neural spine between them. In lateral view, the postzygapophyses are located slightly higher than the prezygapophyses. Along with the centropostzygapophyseal lamina, seen in BSPG AS XXV 119, BSPG AS XXV 77, and BSPG AS XXV 116, they enclose a triangular depression. In BSPG AS XXV 77 and BSPG AS XXV 116 the medial edges of the postzygapophyses are flexed ventrally to form a small, weakly developed hyposphene. As a result of the absence of additional dorsal vertebrae, a transition from the anterior to the posterior dorsals is difficult to establish. A clear grading is seen in the migration of the parapophyses towards the diapophyses, and the shortening of the transverse process. The centra and neural spines maintain their relative and absolute sizes, with the exception of the anteroposterior lengthening of the bases of the spines, as mentioned above. Compared with other taxa, R. tiradentes lacks dorsal vertebrae with strongly expanded spine tables, which are seen in Fasolasuchus tenax (Bonaparte, 1981), Stagonosuchus nyassicus (Gebauer, 2004; Lautenschlager & Desojo, 2011), Batrachotomus kupferzellensis (Gower & Schoch, 2009), and Ticinosuchus ferox (Krebs, 1965). Caudal vertebrae Eighteen caudal vertebrae of R. tiradentes are known, although only 15 are considered in this description because of the poor preservation of the remaining three. They represent elements of the anterior (BSPG AS XXV 118, BSPG AS XXV 117, BSPG AS XXV 78, BSPG AS XXV 79, BSPG AS XXV 80) (Fig. 11A C), middle (BSPG AS XXV 81, BSPG AS XXV 82, BSPG AS XXV 113) (Fig. 11D, E), and posterior part (BSPG AS XXV 108, BSPG AS XXV 107, BSPG AS XXV 88, BSPG AS XXV 109, BSPG AS XXV 83, BSPG AS XXV 110, BSPG AS XXV 84) (Fig. 11F) of the caudal series. All caudal centra are amphicoelous, with the anterior articular surface being more strongly concave than the posterior one. With the exception of the anteriormost preserved element BSPG AS XXV 118, all caudal centra are strongly elongated compared with the presacral series. In the anterior and mid-caudal vertebrae they are on average twice as long as wide, with the posterior elements reaching proportions where they are three times as long as wide. The articular faces of the centra are nearly circular and as wide as high. The anterior and posterior rims are strongly developed and thickened. On the ventral surface the anterior and posterior margins flex ventrally, in particular in the anterior and mid-caudal vertebrae, forming the attachment for the chevron bones. In several mid-caudal vertebrae, the posterior articular ends are notably offset ventrally in relation to the anterior end. A ventral keel is missing in all caudals, although the last four (BSPG AS XXV 109 BSPG AS XXV 84) bear a small groove along the ventral surface. A similar ventral groove has been described for Stagonosuchus nyassicus (Gebauer, 2004), although in this taxon the groove is present in the six anteriormost caudal vertebrae. Because no posterior caudals of Stagonosuchus nyassicus are preserved (Gebauer, 2004), it is unclear whether this feature continued through the vertebral column in this taxon. The anteriormost preserved caudal BSPG AS XXV 118 has a robust centrum, but all other caudal centra are flattened and transversally very narrow. A transverse process is present in the anterior caudals. In the mid-caudal vertebrae the transverse process becomes more and more reduced, until it is completely lost in the posterior elements, where only a thin dorsolateral ledge runs anteroposteriorly along the dorsal margin of the centrum. A gradual change can also be seen in its orientation. In BSPG AS XXV 118 and BSPG AS XXV 78 the transverse processes are directed dorsolaterally, whereas they are oriented horizontally and posteriorly in BSPG AS XXV 79 and the following two caudals. These vertebrae also bear a

22 76 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 11. Caudal vertebrae of Rauisuchus tiradentes. Anterior caudal vertebra (BSPG AS XXV 79) in left lateral view (A), and posterior view (B). Posterior caudal vertebra (BSPG AS XXV 83) in left lateral view (C). Abbreviations: ans, acessory neural spine; cpol, centropostzygapophyseal lamina; posl, postspinal lamina; poz, postzygapophysis; prz, prezygapophysis; spol, spinopostzygapophyseal lamina; tp, transverse process. depression at the point where the transverse process meets the base of the neural spine, and a thin ridge between the transverse process and the prezygapophyses. In the anterior caudals, complete prezygapophyses are only present in BSPG AS XXV 79 and BSPG AS XXV 81. As in the dorsals, they are set at a wide angle. At the position where they are attached to the neural spine a small depression is present. In the anterior and mid-caudal vertebrae the prezygapophyses project anteriorly and dorsally beyond the anterior rim, whereas the posterior caudals have reduced prezygapophyses, which are orientated parallel with the sagittal plane of the centra and barely overlap them. Dorsally, they are connected by a sharp ridge to the remaining part of the neural spine. The postzygapophyses open at an angle of approximately 45. The postzygapophyses are located higher than the prezygapophyses and do not project beyond the posterior rim, but are set off from the centra by a notch, formed between the centrum and the neural spine. A medial lamella connects the postzygapophyses to the neural spine. In the posterior caudals, the postzygapophyses are progressively reduced and the lamellae as well as the notch are lost. In the anterior caudals the neural spine is still prominently developed, and forms a thin blade that is slightly inclined posterodorsally. The spine is positioned on the posterior half of the centrum and decreases in anteroposterior length towards its dorsal end, which is gently rounded. The neural spine becomes gradually smaller and shifts towards the posterior edge in more posterior elements, until it only forms a small spur in the last caudals. As a distinct feature, BSPG AS XXV 81 shows an accessory process on the anterior part of the neural spine with a dorsal hook-like projection (Fig. 11D). It is only completely preserved in this caudal, but the respective remains are also found in BSPG AS XXV 79 and BSPG AS XXV 82. Most of the vertebral laminae are absent in the caudal vertebrae. Centroprezygapophyseal and centropostzygapophyseal laminae are weakly developed in the more anterior caudals. In those anterior caudal vertebrae in which a neural spine is preserved, spinopostzygapophyseal laminae are present.

23 OSTEOLOGY OF RAUISUCHUS TIRADENTES 77 The thin ridge between the prezygapophyses and the neural spine on the posterior caudals might be interpreted as a medial spinoprezygapophyseal lamina, although it does not continue along the neural spine to the apex. A postspinal lamina is clearly visible in BSPG AS XXV 79 and BSPG AS XXV 81, but is absent in the remaining caudals. As in Ticinosuchus ferox (Krebs, 1965), the caudals of R. tiradentes are notably elongated, in contrast to those of Stagonosuchus nyassicus (Gebauer, 2004) and Batrachotomus kupferzellensis (Gower & Schoch, 2009). A characteristic feature of the caudals of R. tiradentes is the accessory neural process. A similar structure is present in Ticinosuchus ferox (Krebs, 1965), Batrachotomus kupferzellensis (SMNS 80339), and Decuriasuchus quartacolonia (De França, Ferigolo & Langer, 2011), although there are some indications that a possible accessory neural process might also be present in Stagonosuchus nyassicus (Gebauer, 2004) and Polonosuchus silesiacus (Sulej, 2005) (compare also Lautenschlager & Desojo, 2011). Unlike in Ticinosuchus ferox (PIZ T 4779), where the process forms only a small protrusion in the anterior caudals or is completely isolated from the neural spine in the posterior caudals, the accessory process of R. tiradentes has a distinct hooklike morphology and contributes considerably to the neural spine. Ribs Along with the vertebrae, three nearly complete ribs and several rib fragments of R. tiradentes were collected. The complete ribs are all from the left side and are two-headed. von Huene (1942: ) designated them as the first, second and maybe the fifth of the thoraic region. In BSPG AS XXV 86a and BSPG AS XXV 86c the tuberculum and the capitulum are still distinctly separated, whereas they are confluent in BSPG AS XXV 87c. The tubercula are circular and slightly depressed, whereas the capitula have an oval outline and are elongated. The shafts are moderately curved and have a triangular cross section in the proximal region, which grades into a circular cross section towards the distal end. A longitudinal depression is visible on the medial surface just below the head. On the opposite side all three ribs bear a strongly developed, anteriorly projecting lateral ridge. A medial depression and a lateral ridge are present in other rauisuchian taxa such as Saurosuchus galilei (Sill, 1974), Batrachotomus kupferzellensis (SMNS 91044), and partly in Ticinosuchus ferox (Krebs, 1965), but also in aetosaurs and extant crocodylians (Mook, 1921; Krebs, 1976). A bony lamella, connecting the tuberculum and the capitulum, is known from Arganasuchus dutuiti Jalil & Peyer, 2007 and Arizonasaurus babbitti (Nesbitt, 2005), but is missing in R. tiradentes. Chevron bones The four chevron bones (Fig. 12A C) belonging to the lectotype material and one further chevron of Find 1020 all possess a Y-shaped outline in anterior view, characteristic for most archosaurs (Mook, 1921; Krebs, 1976). The dorsal branches are connected by a concave crossbar, enclosing a roughly triangular to circular haemal canal. The articular surface for the vertebrae consists of a single, saddle-like structure. The ventral part of the chevrons is gently curved posteriorly and ends in a flattened and rounded, spatula-like tip. The anteriormost of the preserved chevrons, BSPG AS XXV 85b, is approximately 85 mm long and is thus considerably longer than the longest preserved caudal centrum. This element also has the most pronounced distal expansion, with its distal end being more than twice as wide as the minimal width of the shaft. There is no significant difference between the four elements of the lectotype and the single element of Find Chevron bones are preserved in only a limited number of rauisuchian archosaurs. The spatula-like ventral expansion is also present in the chevrons of Ticinosuchus ferox (PIZ T 4779), where it covers nearly the lower half of the bone, whereas it is restricted to the ventralmost part in R. tiradentes. The chevrons in Arganasuchus dutuiti (Jalil & Peyer, 2007) and Postosuchus kirkpatricki (Chatterjee, 1985) are only slightly expanded or are completely straight. Unlike the two-headed joints in Ticinosuchus ferox (PIZ T 4779) and Batrachotomus kupferzellensis (Gower & Schoch, 2009), the attachment to the caudal vertebrae is achieved by only one enlarged articular surface in R. tiradentes; however, Sill (1974) noted that the separated arms of the chevrons join at the fifth chevron bone in Saurosuchus galilei. In crocodiles (Crocodilus americanus Cuvier, 1807), the anterior chevrons have a single articular surface, which splits into two clearly separate facets from the eighth chevron (Mook, 1921). Pectoral girdle The right scapula and part of the right coracoid (Fig. 13) are preserved in natural articulation. The scapula is a long and slender bone that expands towards its dorsal end, and is moderately constricted at its midlength. Dorsally it thins into a flat, blade-like structure, but its margin is poorly preserved and broken. Thus the full extent of the dorsal outline is unknown, but the anterior expansion appears to be larger than the posterior expansion, which would result in a rather asymmetrical form in contrast to von Huene s (1942: 211) reconstruction with a completely symmetrical dorsal end. Despite the damage it can be established that the dorsal expansion reaches at least 240% of the minimal width of the scapular shaft. The posterior edge of the scapula is thickened and rounded along its whole length and the shaft becomes gradually thinner towards the

24 78 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 12. Chevron bone (BSPG AS XXV 85b) of Rauisuchus tiradentes in anterior view (A), left lateral view (B), and posterior view (C). anterior margin. In lateral view the scapula has a dorsoventrally convex surface, with the medial surface being correspondingly concave. Whereas this flexure of the element is gradual over most of the shaft, the ventral, glenoid portion of the scapula shows a more pronounced medial flexion. The latter region is transversely thickened and only slightly expanded posteriorly. Its anterior part is missing, but was probably at least moderately expanded anterodorsally, as indicated by the broken edge and the attached coracoid. The posteroventral part of the scapular shaft bears a pronounced, elongate posterolateral tubercle, just dorsal to the supraglenoid buttress. At the supraglenoid buttress, the posterior edge of the scapula expands medially to form the dorsal (scapular) part of the glenoid cavity. The scapular glenoid facet is roughly triangular in outline, more or less flat, and has slightly raised margins. On the lateral surface, the lower half of the ventral part of the scapula bears a slight depression, which does not continue onto the coracoid. The medial surface of the scapula is for the most part poorly preserved and offers only limited details. The ventralmost part bears a medially thickened elevation, which becomes wider ventrally. It is separated from the posterior margin, which forms the scapular portion of the glenoid cavity, and has a spike-like appearance, by a dorsoventral depression. Above the glenoid part a small groove is visible. A similar groove has been described for Arizonasaurus babbitti (Nesbitt, 2005), where it covers nearly half of the scapular shaft. A similar morphology can be seen in the scapulae of Batrachotomus kupferzellensis (SMNS 80271) and Postosuchus alisonae (Peyer et al., 2008), where the scapular margin extends anteriorly. Although this part is missing in the scapula of R. tiradentes, a similar anterior expansion can probably be inferred. Compared with the shorter and more expanded scapulae of Batrachotomus kupferzellensis (SMNS 80271), Prestosuchus chiniquensis (BSPG AS XXV 12), and Ticinosuchus ferox (Krebs, 1965; Lautenschlager & Desojo, 2011), the scapula of R. tiradentes is notably long and slender. It resembles the scapula of Postosuchus kirkpatricki (Long & Murry, 1995), but it is less strongly curved posteriorly. Ventrally, the scapula is attached to the poorly preserved coracoid and both elements are firmly fused, although this is probably related to preservation, as the suture between both elements is still clearly visible. A large extent of the anterior, posterior, and ventral margin is missing. Like the scapula, the coracoid has a concave medial surface and is slightly vaulted laterally. The glenoid cavity is prominently developed and bordered by a lateral ridge. The coracoid comprises about two-thirds of the glenoid, and the articular surface is elongate oval in outline, and is slightly convex transversely and concave dorsoventrally. Anterior to the

25 OSTEOLOGY OF RAUISUCHUS TIRADENTES 79 Figure 13. Right scapulacoracoid (BSPG AS XXV 91) of Rauisuchus tiradentes in lateral view. Abbreviations: cf, coracoid foramen; cg, cavitas glenoidalis; sgb, supraglenoid buttress.

26 80 S. LAUTENSCHLAGER AND O. W. M. RAUHUT glenoid cavity, the lateral surface of the coracoid seems to be depressed, although it cannot be completely ruled out that this might be related to preservation. On the anterior margin the foramen coracoideum can still be seen, although it is bordered by the broken edge and filled by matrix. Although the extent and shape of the coracoid is unknown in R. tiradentes, comparisons with other rauisuchian taxa, including Batrachotomus kupferzellensis (Gower & Schoch, 2009) and Postosuchus kirkpatricki (Long & Murry, 1995), suggest a much larger extent in anterior and ventral directions, and a semicircular or elongated quadrangular outline. Pelvis The pelvic elements originally assigned to R. tiradentes consist of a left pubis (Fig. 14) of the lectotype, and a left ilium, which originated from the second site ( Find 1025 ) (von Huene, 1942: 215). In addition to coming from a different locality, the ilium differs considerably in size from the pubis. As the ilium cannot definitely be assigned to R. tiradentes, it is excluded from this taxon and is described in a later section. The pubis consists of a proximal, flattened base and a ventrally downturned pubic blade. Its surface is poorly preserved, so the bone structure and surface details are barely recognizable. The proximal part, which Figure 14. Left pubis (BSPG AS XXV 89) of Rauisuchus tiradentes in lateral view. Abbreviations: acr, acetabular rim.

27 OSTEOLOGY OF RAUISUCHUS TIRADENTES 81 articulates with the ilium, is missing. The posterodorsal part is slightly elevated and most likely represents the continued part of the acetabular rim from the ilium, but a clearly defined acetabular depression cannot be seen on the remaining portion of the bone. The pubis makes, therefore, only a very small contribution to the acetabulum. The large obturator foramen described by von Huene (1942: ) and Bonaparte (1984), who identified the foramen in a drawing of the pubis, cannot be confirmed with confidence. A possible outline of such a foramen is, because of poor preservation, barely visible on the pubis and is filled by matrix. The pubic blade runs downwards and is considerably torqued laterally in relation to the base, so that the transverse axis stands perpendicular to the base. The pubic blade itself is transversely wide, flattened, and internally concave in posterior view. The shaft is strongly bent posteroventrally. The distalmost part ends in a flat, deeply pitted tip. A pubic boot or foot as described for Prestosuchus sp. (Azevedo, 1991), Saurosuchus galilei (Reig, 1961), or Batrachotomus kupferzellensis (SMNS 80270) is not present in R. tiradentes. The pubis of R. tiradentes is more similar to those of Ticinosuchus ferox (Krebs, 1965), Stagonosuchus nyassicus (Gebauer, 2004), and Tikisuchus romeri Chatterjee and Majumdar 1987 in this regard. The pubes of all of these taxa lack the distal pubic foot and possess the ventrally curved, flattened morphology. Tibia and fibula A nearly complete tibia is preserved in articulation with the fibula and the astragalus (Fig. 15A, B), but all of these elements are poorly preserved. As Krebs (1973) noted, it represents the right side, not the left, as originally described by von Huene (1942: 212), who confused the proximal and distal ends. The tibia is a straight and compact bone, about 1.5 times as wide as the fibula in its proximal shaft. The proximal portion is missing, so no conclusions can be drawn concerning the length of the bone, the articulation with the femur, or the presence of a possible cnemial crest. Along the whole length, its cross section is approximately oval, with a flattened lateral side, and it increases in width towards the distal end, where it splits up to form a condyle for the reception of the astragalus, overlapping the latter medially. In this area the tibia and the astragalus have considerably been amalgamated and can hardly be distinguished. Near its proximal end, the tibia bears a deep depression on the medial surface. Peyer et al. (2008) identified a similar depression on the tibia of Postosuchus alisonae as the attachment area for the posterior musculus extensor iliotibialis. Figure 15. Right tibia, fibula, and astragalus (BSPG AS XXV 90) of Rauisuchus tiradentes in anterior view (A) and posterior view (B). Abbreviations: asn astragalar notch; ast, astragalus; fib, fibula; ift, iliofibularis trochanter; lap, laterally ascending process; ld, lunate depression; md, medial depression; tib, tibia.

28 82 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Gower & Schoch (2009) described a medially located pit on the proximal part of the tibia of Batrachotomus kupferzellensis, which might represent the insertion site for the musculus puboischiotibialis. If the depression seen on the tibia of R. tiradentes corresponds to one of these attachment areas, or if it represents a muscle scar at all, is difficult to ascertain. Because of the poor preservation, muscle scars and fine details might either be obscured or are simulated by matrix, and the bone is strongly swollen as a result of diagenetic processes. By comparison with the tibia, the right fibula is much more slender. Its proximal end is also missing, but there seems to be no significant expansion. The preserved proximal part is articulated with the expanded lateral surface of the tibia, although the exact contact between both elements is masked by the amalgamation of both. In cross section, the fibula is semi-oval along its length, with only the distal end becoming oval. On the medial side the fibula is flattened to slightly concave. Towards the distal end this concavity is more strongly developed and the fibula adopts a crescent-shaped cross section. The whole fibula is contorted and approximates a sigmoid structure, but because of diagenetic alteration the original outline is uncertain. The anterolateral side bears a weakly pronounced, elongate tubercle on the proximal half. A similar process is also known from Fasolasuchus tenax (Bonaparte, 1981) and Saurosuchus galilei (Sill, 1974), and was identified as iliofibularis trochanter by Jalil & Peyer (2007) for Arganasuchus dutuiti. This process is also present in Postosuchus kirkpatricki (Weinbaum, 2002). The distal end of the fibula is tightly fused with the tibia, but in concurrence with Krebs (1973), this seems to be a diagenetic effect during fossilization, so the articulation with either the tibia or the astragalus are unknown. An astragalar facet is not recognizable. On the lower anterior to anterolateral side, the fibula shows a deep elongated depression. Although the surface in this area is strongly altered by diagenesis, this peculiar depression seems to be natural. A similar structure is also seen in a fibula of Prestosuchus chiniquensis (BSPG AS XXV 3) and Stagonosuchus nyassicus (Lautenschlager & Desojo, 2011). In these taxa the respective depression is lunate or slightly S-shaped. Astragalus Of the tarsal complex only the astragalus (Fig. 15A, B) of R. tiradentes is known, and was originally overlooked in the description by von Huene (1942) because of the erroneous orientation of the element. The astragalus consists of a semicircular main body, a prominently developed laterally ascending process, and a laterally orientated peg. The ascending process and the astragalar peg are separated by a characteristic notch, where the calcaneal peg would have articulated. On the proximal surface the astragalus seems to be depressed and mainly articulates with the tibia, with only a small area of the lateral ascending process contacting the fibula. Despite the poor preservation, Krebs (1973) showed this configuration to be characteristic for the crurotarsal ankle joint of the crocodilenormal type. The strongly convex main body of the astragalus is considerably wider transversely than anteroposteriorly. Dermal armour Thirty-two osteoderms (Fig. 16A D) have been found together with the skeletal elements of R. tiradentes, but none of them was articulated to the vertebrae. von Huene (1942) divided them into six different morphological groups, although this seems to be rather arbitrary. Generally there are two different types of osteoderms: a group of larger, asymmetrical osteoderms (Fig. 16B, C) and a group of smaller symmetrical and leaf-shaped osteoderms (Fig. 16D). All of them share a dorsally located keel, a weak semi-concentric ornamentation, and a ventral concavity. In dorsal view most of the osteoderms are vaulted. The asymmetrical forms are either of roughly rectangular outline or pointed. Their lateral side is somewhat broadened, resulting in the asymmetrical shape. One of the rectangular osteoderms has a mirrored counterpart. Both elements articulate along their medial margin, indicating that R. tiradentes possessed a series of paired, paramedian osteoderms. These were probably restricted to the dorsal region. A single, rectangular plate with two anterior projections (Fig. 16A) originated most likely from the cervical area, where only a median row of osteoderms was developed, as seen in Ticinosuchus ferox (Krebs, 1965). Several small symmetrical and leaf- or arrowhead-shaped elements can be attributed to the tail region. As suggested for Ticinosuchus ferox (Krebs, 1965), Batrachotomus kupferzellensis (Gower & Schoch, 2009), or Decuriasuchus quartacolonia (De França et al., 2011), these covered either the dorsal or the ventral part of the tail. The dermal armour of R. tiradentes differs only slightly from other rauisuchians, with strongly keeled symmetrical and asymmetrical osteoderms. The symmetrical elements, though, are more slender laterally and do not show the expanded, heart-shaped morphology, as seen for example in Batrachotomus kupferzellensis (SMNS 90018). FURTHER MATERIAL REFERRED TO RAUISUCHUS TIRADENTES Rauisuchoid ilium The only known ilium (BSPG AS XXV 88; Fig. 17A, B) originally assigned to R. tiradentes is a left element and comes from the secondary fossil site as part of Find

29 OSTEOLOGY OF RAUISUCHUS TIRADENTES 83 Figure 16. Body osteoderms of Rauisuchus tiradentes. A, osteoderm from the cervical region (BSPG AS XXV 92), B, osteoderm (BSPG AS XXV 97) from the anterior dorsal region, C, osteoderm (BSPG AS XXV 94) from the posterior dorsal region, D, osteoderms (BSPG AS XXV 121c, f) from the caudal region Compared with the pubis of the lectotype it shows a notable difference in size and is from an individual that was at least 20 25% smaller than the lectotype. The ilium consists of an acetabular area and the iliac blade with two processes: an anterior preacetabular process and a posterior postacetabular process. The lateral surface of the main body is nearly entirely occupied by a deep acetabulum. A prominent supraacetabular rim surrounds the dorsal margin. Anteriorly, the main body possesses a spike-like, dorsally flattened projection, which forms the pubic process. In this region the supra-acetabular rim is strongly developed. Anteroventrally, below the pubic process, the ilium bears a longitudinal groove with a depression on the ventralmost part, forming the articulation socket for the pubis. The parallel acetabular rim is not interrupted, indicating that the pubis only marginally contributed, or did not contribute at all, to the acetabulum. The posteroventral part of the ilium is partly broken. The area where the ischiadic process would be located is missing. The preacetabular process is short and projects anteriorly, but does not extend anterior to the pubic process. It has a blunt outline and is only slightly separated from the main body. The postacetabular process is considerably elongated and projects posteriorly and slightly laterally, although the latter might be a preservational artefact. Its cross section is triangular with a rounded lateral side. The medial(?) side is covered by a ventromedially excavated depression. Posteriorly it tapers to a blunt tip. Above the supra-acetabular rim, located between the preacetabular and the postacetabular process, the ilium possesses a prominent supra-acetabular ridge or buttress. It is perpendicular to the longitudinal axis of the iliac blade and contacts the dorsal margin to the supra-acetabular rim. According to Colbert (1961) and Chatterjee (1985), this ridge represents the attachment point for the iliofemoralis or the iliotibialis muscle, although this inference has been debated (Gower, 2000). On the medial side of the ilium, four distinct depressions are visible, which most probably represent the articulation sockets for the two-pronged sacral ribs of the sacral vertebrae. As no corresponding or overlapping elements are preserved in the lectotype, the ilium cannot be referred to R. tiradentes, although the following characteristics support its classification as a rauisuchoid ilium: (1) the ilium contributes to the major part of the acetabulum; (2) presence of a supraacetabular rim; (3) a short preacetabular and an elongated postacetabular process; (4) presence of a supraacetabular ridge connecting the supra-acetabular rim and the dorsal margin of the ilium; (5) closed acetabulum (Gower, 2000; Nesbitt, 2011). The fact that the

30 84 S. LAUTENSCHLAGER AND O. W. M. RAUHUT Figure 17. Left ilium (BSPG AS XXV 88) of Rauisuchoidea indet. (formerly assigned to Rauisuchus tiradentes) in lateral view (A) and medial view (B). Abbreviations: a.sr, articulation with sacral ribs; ac, acetabulum; app, anterior preacetabular process; d, depression; pep, pelvic process; ppp, posterior postacetabular process; sab, supra-acetabular buttress; sar, supraacetabular rim.

31 OSTEOLOGY OF RAUISUCHUS TIRADENTES 85 acetabulum is almost entirely or even entirely placed on the ilium might indicate that it was correctly assigned to R. tiradentes, as the latter lacks an acetabular facet on the pubis; however, this can only be confirmed by additional material. DISCUSSION MORPHOLOGICAL VARIATION IN RAUISUCHIAN ILIA AND ASSIGNMENT OF INDETERMINED MATERIAL Amongst other osteological characteristics, the morphology of the ilium has often been considered to be of taxonomic importance within rauisuchian archosaurs. Distinctive features on the ilium, such as ridges or buttresses, have been treated as phylogenetically informative, although the definition, distribution, and homology of these features is often equally unclear (for detailed discussion, see Gower, 2000); however, the general shape of the ilium, as provided by the outline, appears to reflect phylogenetically distinct groups, with a majority of the variation summarized by the first two PC axes (49.7 and 13.8%; Fig. 18). The shape analysis of the ilia outlines displays a distinct morphological separation between poposauroid and rauisuchian (= noncrocodylomorph Loricata) taxa. A third group consisting of three ctenosauriscid taxa shows a morphology distinct to poposauroids, but overlaps partly with the rauisuchian morphospace. Interestingly, different specimens of the same species within these three groups show an equally large morphological disparity. Poposaurus gracilis (Mehl, 1915), Postosuchus kirkpatricki, and Prestosuchus chiniquensis each occupy a large morphospace. In particular the wide spread of the different specimens of Poposaurus gracilis is unexpected. Although taphonomic artefacts might be responsible for these results, it was ascertained that only specimens for which a complete outline could be generated were sampled. It is noteworthy, however, that Figure 18. Morphospace occupation of rauisuchian ilia, based on the first two principal components of the shape analysis of 22 specimens. PC1 contributes to 49.7% and PC2 contributes to 13.8% of the variation. Variation in shape is depicted by outlines of terminal taxa along each axis; green, Ctenosauriscidae; blue, Poposauroidea; red, noncrocodylomorph Loricata. (Colour version of figure available online.)

32 86 S. LAUTENSCHLAGER AND O. W. M. RAUHUT each of these taxa has a taxonomically complicated history (Long & Murry, 1995; Gower, 2000; Brusatte et al., 2010). Furthermore, intraspecific variability and sexual dimorphism might be responsible for moderate morphological differences, resulting in a pronounced separation in morphospace. On the other hand, Batrachotomus kupferzellensis occupies a comparably small morphospace, defined by three individual specimens, which all originated from the same locality (Gower & Schoch, 2009). Consequently, it is questionable whether the morphology of the ilium is characteristic beyond group-level taxonomic units, without more detailed knowledge about intraspecific variability. In the shape analysis, the indetermined rauisuchoid ilium (BSPG AS XXV 88) falls within the rauisuchian morphospace, but its position does not allow a more detailed assignment. The obvious difference in size between it and the pubis of the lectotype of R. tiradentes precludes the possibility that both elements were part of the same individual. Apart from R. tiradentes, three other rauisuchoid species are known from the Santa Maria Formation: Prestosuchus chiniquensis, Prestosuchus loricatus von Huene, 1942, and Decuriasuchus quartacolonia; however, these taxa either do not have an ilium preserved or, as is the case in Decuriasuchus quartacolonia, the ilium has not been described or illustrated in sufficient detail (De França et al., 2011) for comparison. Consequently, it was not possible to include the latter taxon in the shape analysis, whereas the large morphospace occupation of Prestosuchus chiniquensis makes a justifiable assignment difficult. The close stratigraphical and geographical proximity (with less than 20 km between each) of the fossil locations further complicates an exact differentiation of these four different species from the Santa Maria Formation; however, the individual taxa show a clear separation in size. Estimated from the preserved elements of the respective specimens (compare von Huene, 1942; De França et al., 2011), Prestosuchus reached a length of approximately 5 6 m, R. tiradentes reached a length of 3 4 m, and Decuriasuchus quartacolonia reached a length of 2 3 m. The latter would thus be a prime candidate for the assignment of the ilium in question, but further comparisons are impeded by the lack of a detailed description and illustration of this element, without which no clear taxonomic statement can be made. PHYLOGENETIC POSITION OF RAUISUCHUS TIRADENTES Research on rauisuchian archosaurs and their phylogenetic relationships has considerably intensified in recent years. Several new fossils and revisions of known taxa have improved our knowledge of this group. Nonetheless, there is still no conclusive and generally accepted phylogeny of Rauisuchia. In the most exhaustive phylogenetic analyses so far, Brusatte et al. (2010) and Nesbitt (2011) independently analysed the higher-level phylogeny of Archosauria. Brusatte et al. (2010) analysed 55 different taxa, including 20 rauisuchian taxa, and 187 characters, whereas Nesbitt s (2011) analysis consisted of 77 taxa (among them 18 rauisuchian taxa), with 412 cranial and postcranial characters. Although both analyses agree in some aspects, including the monophyly of several suprageneric paracrocodylomorph clades, such as Phytosauria, Aetosauria, or Ornithosuchidae, they result in different relationships for rauisuchian taxa. Whereas Brusatte et al. (2010) found a well-resolved but poorlysupported phylogeny, which places all rauisuchians in a single, monophyletic group (Brusatte et al., 2009: 229), which is divided into two major subclades Poposauroidea and Rauisuchoidea Nesbitt s (2011) analysis recovered all analysed rauisuchian taxa as a paraphyletic group basal to Crocodylomorpha. In both studies, however, the character scores for R. tiradentes are either largely based on von Huene s (1942) original description or different interpretations of anatomical structures (e.g. inclusion of the ilium in the type material and erroneous identification of elements, such as the postorbital). A reanalysis of both data sets, supplemented by updated and corrected character scorings (see Appendix S1), based on the revised osteology, generally yielded the same results and the same overall topology as the original data (Fig. 19). However, and not surprising, the removal of elements, such as the ilium, from the type material resulted in a higher number of most parsimonious trees and a reduced resolution within the rauisuchian part of the trees. In the reanalysis of Nesbitt s (2011) data set, a strict consensus tree of 1267 most parsimonious trees (MPTs; tree length, TL = 1293 steps; retention index, RI = 0.77; consistency index, CI = 0.38) finds all rauisuchian taxa in a paraphyletic assemblage with Ticinosuchus ferox as the basalmost taxon (Fig. 19A). As in the original analysis a monophyletic clade of Poposauroidea, including a monophyletic subclade Ctenosauriscidae, are recovered. Rauisuchus tiradentes is positioned in a large polytomy with Fasolasuchus tenax, CM73372 (an unidentified taxon, formerly assigned to Postosuchus kirkpatricki), Crocodylomorpha, and a clade containing Polonosuchus silesiacus, Postosuchus kirkpatricki, and Postosuchus alisonae, so that these taxa no longer form a monophyletic group (termed Rauisuchidae). As described above, there is no indication that the anteroventral process of the squamosal contacts the postorbital and bisects the lower temporal fenestra (Nesbitt, 2011: character 52), thus removing a uniting character for the aforementioned four taxa. A 50% majority rule consensus of the most

33 OSTEOLOGY OF RAUISUCHUS TIRADENTES 87 Figure 19. Phylogenetic analyses based on a reanalysis of Nesbitt s (2011) data set with 80 taxa and 415 characters. A, strict consensus tree of 1267 most-parsimonious trees (MPTs; tree length, TL = 1293 steps; retention index, RI = 0.77; consistency index, CI = 0.38). B, 50% majority rule consensus tree. Numbers denote percentage of most parsimonious trees, in which the respective clade is recovered. Colour-coding of phylogenetic units as in Figure 18. (Colour version of figure available online.) parsimonious trees shows a slightly better resolved configuration for the non-poposauroid taxa (Fig. 19B). In particular, the relationship of Polonosuchus silesiacus, Postosuchus kirkpatricki, and Postosuchus alisonae are well resolved, whereas Fasolasuchus tenax and CM73372 were recovered as sister taxa. In comparison, reanalysis of the data set from Brusatte et al. (2010) also resulted in a paraphyletic Rauisuchia as a sister taxon to Ornithosuchidae. Similarly, a strict consensus of 1911 MPTs (TL = 735 steps; RI = 0.67; CI = 0.31) recovered Ctenosauriscidae as a monophyletic subclade, but found a paraphyletic Poposauridae because

34 88 S. LAUTENSCHLAGER AND O. W. M. RAUHUT of the exclusion of Qianosuchus mixtus Li et al. (2006) (Fig. 20A). Rauisuchus tiradentes is found in a polytomy (Fig. 19B) alongside most other taxa, with the exception of two subclades (Batrachotomus kupferzellensis and Prestosuchus sp.; Postosuchus sp. and Polonosuchus silesiacus). A 50% majority rule consensus tree recovered Poposauroidea and Ctenosauriscidae as a monophyletic clade and subclade, respectively (Fig. 20B). Rauisuchus tiradentes was found in a polytomy (including Tikisuchus romeri, Yarasuchus deccanensis Sen, 2005, Poposauroidea, and two rauisuchian subgroups) basal to most rauisuchian taxa. The reanalysis of both data sets clearly shows that the phylogenetic position of R. tiradentes is not only highly variable, but also that the loss of resolution is directly related to the removal of elements from Figure 20. Phylogenetic analyses based on a reanalysis of Brusatte et al. s (2010) data set with 59 taxa and 193 characters. A, strict consensus tree of 1911 most-parsimonious trees (MPTs; tree length, TL = 735 steps; retention index, RI = 0.67; consistency index, CI = 0.31). B, 50% majority rule consensus tree. Numbers denote percentage of most parsimonious trees, in which the respective clade is recovered. Colour coding of phylogenetic units as in Figure 18. (Colour version of figure available online.)