Trilophosaurus

Nesbitt, Sterling J., Flynn, John J., Pritchard, Adam C., Parrish, J. Michael, Ranivoharimanana, Lovasoa & Wyss, André R., 2015, Postcranial Osteology Of Azendohsaurus Madagaskarensis (? Middle To Upper Triassic, Isalo Group, Madagascar) And Its Systematic Position Among Stem Archosaur Reptiles, Bulletin of the American Museum of Natural History 2015 (398), pp. 1-126 : 88-96

Trilophosaurus (including Spinosuchus

caseanus) (Bremer 5 3)

SUPPORT: Unambiguous synapomorphies include: lacrimal limited to orbital margin (11-2*); coronoid process present (79-1*); tooth shape (marginal dentition) labiolingually wider than mesiodistally long at crown base (98-2*); posterior articular surface of the centrum of the presacral vertebrae convex (102-2*); diapophysis located in the anteroposterior middle of the neural arch/centrum in the trunk vertebrae (215-1*).

Other possible synapomorphies: AC- CTRAN: Palatal teeth absent (44-1); posteri- or surface of the supraoccipital smooth (55-0); Shape of the supraoccipital pillarlike (56-1); angular exposure on lateral mandibular surface extends to the glenoid (83-1); Intercentra in the trunk vertebrae present (128-0). DELTRAN: Splenials contribute to mandibular symphysis (85-0); moderate development of posterior articular surface convexity of the presacral vertebrae (103-0); costal facets very closely appressed to one another with little or no finished bone separation in the anterior postaxial cervical vertebrae (112-1); entepicondylar crest of the humerus exhibits a prominently angled proximal margin (155-1); postzygapophyses of the anterior cervical vertebrae (presacral vertebrae 3–5) connected through a horizontal lamina (5 transpostzygapophyseal lamina) with a notch at the midline (213-1); posterior caudal vertebrae much longer than the anterior caudal vertebrae (218-1); penultimate phalanges of the pes significantly longer than the more proximal phalanges (235-1).

DISCUSSION: Because of its unusual dental anatomy, the relationships of Trilophosaurus buettneri have been hotly contested since its initial description from a partial maxilla ( Case, 1928a, 1928b; Gregory, 1945; Romer, 1956; DeMar and Bolt, 1981; Carroll, 1988). Since the advent of cladistic analyses, Trilophosaurus buettneri has been consistently placed as a non-archosauriform archosauromorph ( Gauthier et al., 1988b; Dilkes, 1998; Spielmann et al., 2008; Gottmann-Quesada and Sanders, 2009; Ezcurra et al., 2014; Pritchard et al., 2015), among various early diverging archosauromorph clades ( Dilkes, 1998). Our analysis indicates a clear relationship between Trilophosaurus (and kin) and Azendohsaurus , and a fairly well-resolved placement for this clade outside Archosauriformes . The best-known trilophosaurid, Trilophosaurus buettneri , obviously embodies a bizarre mix of cranial features. This mix includes many unique character states (e.g., tricuspid teeth with a beaklike rostrum), plesiomorphic archosauromorph character states (e.g., lack of an antorbital fenestra), and characters present in archosaurs (e.g., complete loss of palatal dentition). Our examination of character state changes in the postcrania of allokotosaurs with regard to T. buettneri yielded unexpected results, particularly that the taxon’s “lizardlike” features (e.g., long tail, elongated manus and pes, procoelous vertebrae, general “lizardlike” proportions: Gregory, 1945) all are autapomorphic, rather than retentions of plesiomorphic saurian conditions. Thus, T. buettneri and its closest relatives are even more specialized than previously suspected. In fact, most of the unique character states described originally by Gregory (1945) are now useful in grouping the closest relatives of T. buettneri into a monophyletic clade of basal archosauromorphs. We found that various character states, particularly in the teeth and skull, unite T. buettneri and T. jacobsi , and that additional postcranial character states unite T. buettneri , T. jacobsi , and Spinosuchus caseanus . Some postcranial character states shared by T. buettneri , T. jacobsi , and S. caseanus were originally detailed by Spielmann et al. (2009) but were not incorporated into a phylogenetic analysis. We tested these character states in a numeric phylogenetic analysis and found that a diapophysis located in the anteroposterior middle of the neural arch/centrum in the trunk vertebrae (215-1) was an unambiguous synapomorphy, but others detailed by Spielmann et al. (2009) (e.g., postzygapophyses of the anterior cervical vertebrae [presacral vertebrae 3–5] connected through a horizontal lamina [5 transpostzygapophyseal lamina] with a notch at the midline [213-1]) were ambiguous synapomorphies, because they could not be scored in the closest outgroup, Teraterpeton hrynewichorum . Moreover, we identified two other potential “lizardlike” synapomorphies for T. buettneri , T. jacobsi , and S. caseanus that could not be scored in Te. hrynewichorum : posterior caudal vertebrae much longer than the anterior caudal vertebrae, and penultimate phalanges of the pes significantly longer than the more proximal phalanges.

The results of our phylogenetic analysis with Trilophosaurus buettneri , Trilophosaurus jacobsi , and Spinosuchus caseanus raised the question of whether Trilophosaurus jacobsi and Spinosuchus caseanus are in fact the same animal. Collectively a comparison of the holotypes of Spinosuchus caseanus and Trilophosaurus jacobsi , their cooccurrence at the Kahle Trilophosaurus Quarry , and our phylogenetic results suggest that Trilophosaurus jacobsi is a subjective synonym of Spinosuchus caseanus , which is discussed next.

The absence of overlapping elements between the holotypes of Spinosuchus caseanus (UMMP 7507) and Trilophosaurus jacobsi (MNA V3192) complicates our proposed synonymy. The holotype of Spinosuchus consists of an axial column, and the holotype of Trilophosaurus jacobsi is an anterior portion of the maxilla with four teeth, three partial and one nearly complete (Murry, 1987). Our case for synonymy thus hinges on correct referral of skeletal elements not represented in the holotypes; these that can be compared directly to T. jacobsi , referrals are based on direct association or and all Trilophosaurus -like postcrania, hisparticular apomorphies identified over the torically have been assigned to T. jacobsi , course of this comparative anatomical and whereas vertebrae with tall neural spines have phylogenetic analysis. been assigned to S. caseanus .

Perhaps the strongest evidence that Spino- Few vertebrae from the Kahle Trilophosuchus caseanus and Trilophosaurus jacobsi saurus Quarry have been assigned to Trilorepresent the same taxon comes from their phosaurus jacobsi (see Spielmann et al., 2008) cooccurrence at Kahle Trilophosaurus Quarry whereas all the complete or nearly complete in the Tecovas Formation of the Dockum vertebrae (with tall neural spines) are as- Group (Borden County, Texas) (they also signed to Spinosuchus caseanus . On the other cooccur at other localities, e.g., at Rotten Hill hand, none of the Trilophosaurus buettneri – in the Tecovas Formation of the Dockum like postcrania have been assigned to S. Group, northern Texas [Spielmann et al., caseanus even though Spielmann et al. (2009) 2013].) Tooth-bearing elements and teeth convincingly argue that S. caseanus is closely from the Kahle Trilophosaurus Quarry were related to Trilophosaurus . Here, we posit that assigned to Trilophosaurus Heckert et al. , all of the Trilophosaurus -like crania and 2001), and subsequently to Trilophosaurus postcrania assigned to T. jacobsi , and the jacobsi (Heckert et al., 2006) . Nearly all the vertebrae assigned to S. caseanus , pertain to postcranial material from the Kahle Trilo- a single species-level taxon. In other words, phosaurus Quarry found alongside the tooth- T. jacobsi and S. caseanus represent the same bearing elements referred to Trilophosaurus species , and given the rules of priority, all this jacobsi were referred to the same taxon material from the Kahle Trilophosaurus (Heckert et al., 2006; Spielmann et al., Quarry should be assigned to S. caseanus .

2008), largely because of the similarity of To further test our proposition that the these postcrania to those of Trilophosaurus Trilophosaurus -like elements from the Kahle buettneri from the Otis Chalk quarries in Trilophosaurus Quarry represent a single Howard County, Texas (Gregory, 1945). taxon ( Spinosuchus caseanus ), we scored the Trilophosaurus jacobsi is currently known holotype of S. caseanus , and all the Kahle from hundreds of elements, including cranial Trilophosaurus Quarry elements referred to material, forelimbs, hindlimbs, pectoral and T. jacobsi by Spielmann et al. (2008) or to S. pelvic girdles, and vertebrae. Other than caseanus by Spielmann et al. (2009) as three reptile teeth clearly not belonging to T. separate terminal taxa. A similar method has jacobsi (Heckert et al., 2001) , the only been used to test the synonymy of the tetrapod elements from the Kahle Trilopho- “ Chatterjeea elegans ” and Shuvosaurus inexsaurus Quarry not identified as T. jacobsi are pectatus (Nesbitt and Norell, 2006; Nesbitt, vertebrae bearing tall neural spines. These 2007).

vertebrae were assigned to Spinosuchus case- Our phylogenetic analysis identified S. anus (Heckert et al., 2001; Spielmann et al., caseanus and T. jacobsi as nearest relatives 2009). Justification for the assignment of (fig. 72); both are scored identically in our these elements to S. caseanus is clearly spelled data matrix, sharing one unique character out by Spielmann et al. (2009). In sum, all state (spinopostzygapophyseal laminae prestooth-bearing elements from these deposits ent on the posterior edge of the neural arch of +

Pamelaria dolichotrachela (ISRI 316/9) in lateral view. (H) Anterior cervical vertebra of Trilophosaurus buettneri (TMM 31025-46) in lateral view. (I) Anterior cervical vertebra (reversed) of Azendohsaurus madagaskarensis (FMNH PR 2791) in lateral view. (J) Right scapulocoracoid of Pamelaria dolichotrachela (ISRI 316/47) in in lateral view. (K) Right scapula of Trilophosaurus buettneri (TMM 31025-68R) in lateral view. (L) Left scapula (reversed) of Azendohsaurus madagaskarensis (FMNH PR 2798) in lateral view. (M) Right coracoid of Trilophosaurus buettneri (TMM 31025-69h) in lateral view. (N) Left coracoid (reversed) of Azendohsaurus madagaskarensis (FMNH PR 3822) in lateral view. Scales 5 1 cm. Arrows indicate anterior direction. Abbreviations: a., articulates with; en, external naris; mx, maxilla; pmx, premaxilla.

most presacral vertebrae [245-1]), a feature absent in T. buettneri .

If our hypothesized synonymy is correct, Spinosuchus caseanus is more abundant in the Upper Triassic of the southwestern United States than previously recognized. The taxon would now be known from the Chinle Formation in Arizona ( Placerias Quarry UCMP A 269, holotype of T. jacobsi MNA V 3192), possibly the Chinle Formation of New Mexico (NMMNH Locality 2739, NMMNH P-34448), and potentially over a substantial stratigraphic range in the Dockum Group (MOTT VPL 3869, TTU- P10413). Even accepting this proposed range extension, secure occurrences of S. caseanus in the Revueltian land-vertebrate faunachron are lacking (Lucas, 1998).

IMPLICATIONS FOR THE DIVERGENCE OF ARCHOSAUROMORPHA

All recent phylogenetic analyses concur that the origin and lineage diversification of early archosauromorphs occurred prior to the end of the Permian (e.g., Dilkes, 1998; Muller, 2004; Nesbitt, 2011; Butler et al., 2011; Ezcurra et al., 2014). This is well supported, given that Archosauriformes , represented by Archosaurus rossicus , diverged from the rest of Archosauromorpha by the end of the Permian (Tatarinov, 1960; Nesbitt, 2011; Ezcurra et al., 2014). This divergence indicates that all successive outgroups to Archosauriformes (e.g., Rhynchosauria , Tanystropheidae ) were also present by that time. Regrettably, the fossil record of archosauromorphs prior to the Triassic is exceptionally poor ( Ezcurra et al., 2014). Not only is the record highly fragmentary, anatomical evidence of phylogenetic character transformations (or plesiomorphies) at the base of Archosauromorpha and along the spine leading to Archosauriformes is extremely limited. Consequently, it is difficult to understand whether the subgroups of archosauromorphs form successive outgroups (see Dilkes, 1998) or some of the subgroups pair with each other (e.g., the proposed affiliation of Rhynchosauria and Trilophosauridae by Merck, 1997).

This uncertainty, in turn, raises a number of other questions including: How many archosauromorph subgroups crossed the Permian-Triassic boundary and when did the highly disparate subgroups of archosauromorphs diversify and transform into the strange animals that appear in the Middle or Late Triassic? Answering these questions requires a much broader evaluation and more complete taxon sampling than is possible here, but the new anatomical and phylogenetic information gleaned from Azendohsaurus madagaskarensis provides important clues toward unraveling this puzzle. The phylogenetic conclusions presented here suggest that Tanystropheidae , Rhynchosauria , Archosauriformes , Prolacerta broomi , and Allokotosauria (containing Azendohsauridae and Trilophosauridae ) all crossed the Permian-Triassic (P-T) boundary. Although the number of crossing lineages proposed here is comparable to that suggested elsewhere (e.g., Dilkes, 1998), our study departs from its predecessors in comprehensiveness of the taxon and character analyses, as well as by placing Trilophosaurus / Trilophosauridae within a clade (Allokotosauria) that likely diversified by the end of the Middle Triassic based on current sampling of confirmed members of this clade. A recent report on the possibility of Early Triassic trilophosaurids from Russia ( Arkhangelskii and Sennikov, 2008) suggests that Trilophosauridae is much older than all other records for this clade. Although their report is intriguing, at this time we limit our analysis to the confirmed trilophosaurid forms with crania and postcrania. Nevertheless, the ghost lineage leading to Trilophosaurus originated in the Permian according to Dilkes (1998), but here we suggest that this ghost lineage is much shorter, and likely confined to the Triassic. Incorporating late Middle Triassic–aged A. madagaskarensis into the archosauromorph tree truncates the long Trilophosaurus ghost lineage hypothesized in previous analyses (e.g., Dilkes, 1998).

Moreover, the beautifully represented and almost completely sampled skeleton of A. madagaskarensis helps lessen the morphological gap between Trilophosaurus and other archosauromorphs in two respects. First, A. madagaskarensis helps demonstrate that Trilophosaurus is more closely related to the “protorosaur” Pamelaria dolichotrachela than previously thought, and, thus, that the autapomorphic body plan of Trilophosaurus does not stretch into the Permian, but instead is derived within the Triassic from a “protorosaur”-like common ancestor with other allokotosaurs. This further supports the polyphyly of “protorosaurs” (5 prolacertiforms) and illustrates that “protorosaurs” possess a plesiomorphic archosauromorph body plan rather than a derived one. Second, the close relationship of A. madagaskarensis and Trilophosaurus (with Pamelaria dolichotrachela outside this clade) indicates that the “lizardlike” morphology of the skeleton of Trilophosaurus is independently derived from the plesiomorphic condition found in Sauria (see above). This highlights the tremendous disparity in body plans among even closely related Triassic archosauromorphs; the body plan of Pamelaria dolichotrachela resembles that of Protorosaurus , the body plan of Trilophosaurus resembles generalized squamates, and the body plan of Azendohsaurus is more similar to that of a sauropodomorph dinosaur (cranial features, elongated neck, large mani) than to any other early archosauromorph.

The age, geographic distribution, anatomical features, and phylogenetic information furnished by A. madagaskarensis has the following additional implications: (1) Allokotosauria diverged from other archosauromorphs by the end of the Permian; (2) the plesiomorphic (5 Protorosaurus -like) body plan of Allokotosauria survived well into the Triassic; and (3) the sauropodomorph-like body plan of Azendohsaurus and “lizardlike” body plan of Trilophosaurus evolved by the end of the Middle Triassic. This great diversification in morphological disparity by the Middle Triassic is perhaps not unexpected given the rapid, almost simultaneous diversification of archosauriforms ( Brusatte et al., 2011; Butler et al., 2011; Nesbitt, 2011) within five million years of the end-Permian extinction.

CONVERGENCE AND THE PREVALENCE OF HERBIVORY IN TRIASSIC ARCHOSAUROMORPHA

We have previously noted the convergent similarities of skull morphology and form of the marginal teeth between A. madagaskarensis and early sauropodomorph dinosaurs ( Flynn et al., 2010), many of which have been interpreted as herbivorous adaptations in early sauropodomorphs (Weishampel and Norman, 1989; Barrett, 2000; Sues, 2000; Goswami et al., 2005). For example, both A. madagaskarensis and early sauropodomorphs share convergences in marginal tooth crown structure and hemimandible architecture (see Flynn et al., 2010). As highlighted in the descriptions above, however, homoplasious similarities between A. madagaskarensis and early sauropodomorphs extend beyond craniodental features, occurring throughout much of the skeleton. In particular, the forelimbs of Azendohsaurus and sauropodomorphs are notably robust compared with those of their closest relatives; the humerus is relatively short with greatly expanded proximal and distal ends, the ulna and radius are stocky, and the phalanges are short but bear large unguals. The presacral axial column is also similar in both taxa: the cervical series is elongate compared to the trunk series, and accessory intervertebral articulations (i.e., hypantrum-hyposphene) are present in much of the vertebral column. These strong convergences of the forelimb and axial column of Azendohsaurus and sauropodomorphs contrast with the remarkable dissimilarity of their hindlimbs. The hindlimb of A. madagaskarensis closely resembles that of lepidosauromorphs; the femur was held nearly horizontally and the complicated tarsus includes nine ossified elements. The hindlimbs of early sauropodomorphs were strikingly different in being vertically oriented directly beneath the pelvis and in having a simplified tarsus (four elements or fewer). Additionally, the tail of Azendohsaurus is much shorter (scaled for body size) than that of early sauropodomorphs (e.g., Plateosaurus, AMNH FR 6810). Thus, the convergences between A. madagaskarensis and early sauropodomorphs rest entirely within the skull and anterior half of the postcranial skeleton. The relatively large body size of Azendohsaurus is also unusual among early archosauromorphs and is even larger than most early diverging sauropodomorphs. Based on femoral length as a proxy body size ( Carrano, 2006; Sookias et al., 2012; Turner and Nesbitt, 2013), Azendohsaurus (femoral length [FL] 5 215 mm, table 7) was larger than all known early diverging sauropodomorphs from the Carnian (e.g., Saturnalia tupiniqim, FL 5 157 mm) but is clearly smaller than Norian and younger sauropodomorphs (e.g., Plateosaurus engelhardti, FL 5 750 mm). Nevertheless, the length and morphology of the elongated cervical vertebrae of Azendohsaurus are more similar to those of Norian sauropodomorphs than to the earliest diverging sauropodomorphs (e.g., Eoraptor lunensis, Sereno et al., 2013 ). Although early sauropodomorph dinosaurs were hypothesized to be the earliest high browsers, based on neck length and body size (Weishampel and Norman, 1989), the convergences with Azendohsaurus suggest that the high browser resource zone may have been already occupied by early archosauromorphs that diverged prior to the origin of dinosaurs. Alternatively, the sprawling posture of Azendohsaurus may have inhibited high browsing.

Extraordinarily, nearly all the craniodental and postcranial features identified in this study as having arisen convergently in A. madagaskarensis and early sauropodomorphs, are restricted to Azendohsauridae , and were not acquired sequentially in a stepwise pattern across more inclusive clades in Archosauromorpha. The divergent craniodental and anterior skeleton morphology in Azendohsaurus relative to other early archosauromorphs has a number of implications for the evolution of herbivory in early archosauromorphs, particularly within Allokotosauria. Three distinct craniodental character suites that occur within Allokotosauria possibly reflect divergent styles of herbivory: (1) the Azendohsaurus character suite (long, mediolaterally compressed crowns with denticles present throughout the dentary, premaxilla, and the maxilla), (2) the Trilophosaurus character suite (mediolaterally expanded teeth with distinct cusps, short edentulous portion in the anterior portion of the skull), and (3) the Teraterpeton character suite (bulbous teeth with an anterior cusp, long edentulous gap in the upper jaw and the mandible). Even within its relatively low taxonomic diversity, the craniodental disparity of Allokotosauria is extremely high, and the differences between these morphological suites are as profound as the variations in herbivorous features seen across Triassic archosaurs as a whole (e.g., aetosaurs versus silesaurids). Results from our phylogenetic analysis indicate that the three distinct craniodental suites in Allokotosauria are derived from a common herbivorous ancestor (see fig. 74). Nevertheless, the radically different dental specializations among the three terminal taxa make it unclear what pattern typified their common ancestor. This common ancestor may have borne none of the herbivorous traits present in any the three terminal taxa or it may have had a few of the character states from the distinct suites observed in the terminal taxa. However, their marked, unusual specializations have obscured the plesiomorphic anatomical condition or ancestral herbivorous character states in the common ancestor. Nevertheless, few discrete craniodental features correlated with

TABLE 12

Reinterpretation of the manus of

Trilophosaurus buettneri

an herbivorous diet are shared across the three lineages of allokotosaurs; the three distinct character suites are each unique among early archosauromorphs.

Timing of the acquisition of these suites is poorly constrained. Azendohsaurus , Trilophosaurus , and Teraterpeton appear in the fossil record during either the late Middle, or early Late Triassic (late Carnian or early Norian). The lack of terminal taxa having bridging suites of craniodental characters suggests two alternatives regarding the timing of the origin of these herbivorous specializations: (1) these lineages diverged during the Middle Triassic, and each of the three distinct craniodental suites arose quite rapidly; or (2) they diverged considerably earlier and intermediate forms simply have yet to be uncovered. Both scenarios seem equally plausible given the uncertainty surrounding the timing in the Triassic when these lineages diverged, and the currently poor sampling of early diverging, nonrhynchosaur archosauromorphs from the Early and Middle Triassic.

The considerations above regarding the origin of herbivory within Allokotosauria also shed light on shifts to herbivory in other Triassic archosauromoph subgroups. Our recovery of a clade of bizarre herbivores, Allokotosauria, demonstrates that herbivory among archosauromorphs was far more prevalent early in the Triassic than previously realized. Beyond the herbivorous groups that have been known for over a century (e.g., rhynchosaurs Huxley, 1869; Benton, 1983; aetosaurs, Agassiz, 1844, Walker, 1961; and sauropodomorphs von Meyer, 1861, Weishampel and Norman, 1989), a wealth of presumably herbivorous archosauromorphs have come to light from the Triassic in recent years (e.g., Revueltosaurus callenderi, Parker et al., 2005 ; silesaurids, Langer et al., 2010, Brusatte et al., 2010; and possibly shuvosaurids, Nesbitt and Norell, 2006), as have other postTriassic lineages (crocodylomorphs, Buckley et al., 2000). Additionally, taxa that have a typically “carnivorous” dentition such as Protorosaurus speneri have been found with plant fragments in their digestive track (Munk and Sues, 1993). We can now say with some confidence that within Archosauromorpha herbivorous specializations (see Barrett, 2000, and Zanno and Makovicky, 2011) arose at least seven times independently during the Triassic (in Rhynchosauria, Allokotosauria , Revueltosaurus + Aetosauria , silesaurids, silesaurids, sauropodomorphs, ornithischians).

The new data gleaned from A. madagaskarensis suggests that two episodes of multiple lineage shifts to herbivory may have occurred among amniotes. The first shift, during the early portion of the Middle Triassic, involved early-diverging archosauromorphs (rhynchosaurs, Allokotosauria), non-dinosaurian ornithodirans (silesaurids), and gomphodont cynodonts (Reisz and Sues, 2000). The second shift occurred by the end of the Carnian (Late Triassic), involving both dinosaurs and further diversification of herbivorous pseudosuchians (e.g., aetosaurs, Revueltosaurus ). It is noteworthy that the convergent morphologies that evolved in the taxa from the first wave overlap little in space and time those from the second wave. For example, the timing of acquisition of convergences observed between Azendohsaurus and early sauropodomorphs are completely offset in time. Even though Azendohsaurus may have temporally overlapped the earliest sauropodomorphs (e.g., Eoraptor lunensis ), although this is uncertain because of uncertainties in the age of the Makay Formation, the convergent herbivorous features shared by Azendohsaurus and sauropodomorphs occur only in younger sauropodomorphs (e.g., Plateosaurus ) from later in the Late Triassic. The proposed pattern of twin shifts to herbivory by amniotes during the Triassic is only beginning to emerge through discovery and analysis of well-preserved and complete fossils like Azendohsaurus madagaskarensis . Constraining the full extent and timing of these shifts or testing for additional episodes in early archosauromorphs history through additional analyses of existing taxa and discovery of Middle Triassic precursors represents a promising avenue of continuing research.

CONCLUSIONS AND PERSPECTIVE

Azendohsaurus madagaskarensis incorporates a seemingly incongruous mix of archosauromorph character states, some of which are found only within Archosauriformes and others within Dinosauria (discussed in Gauffre, 1993; Flynn et al., 1999, 2010). Two character states, the presence of serrated teeth and an ossified laterosphenoid in the skull, were regarded as placing A. madagaskarensis closer to Archosauriformes than Prolacerta broomi was to Archosauriformes ( Flynn et al., 2010) . Our current analysis, including of postcranial evidence for A. madagaskarensis for the first time and all available data in a comprehensive phylogenetic study, suggests a more complex distribution of these features than previously envisioned. The serrated teeth in A. madagaskarensis and other closely related taxa (i.e., Pamelaria dolichotrachela , see below) also occur in Archosauriformes , but homoplastically, as serrated teeth now appear to have evolved independently in more than Archosauriformes . In addition, an ossified laterosphenoid in A. madagaskarensis optimizes as an autapomorphy of the taxon among basal archosauromorphs, not as a character state uniting A. madagaskarensis with Archosauriformes . This is not surprising, however, given that the laterosphenoid of A. madagaskarensis has a unique morphology not shared by any other archosauriform (see Clark et al., 1993). This finding may weaken the notion that turtles are closely related to archosauriforms ( Bhullar and Bever, 2009), given that an ossified laterosphenoid is more broadly distributed within Archosauromorpha. Vertebral character states of A. madagaskarensis also hint at a more complicated distribution of character states among archosauromorphs than previously realized. The sauropodomorph-like cervical vertebrae, complex laminae of the neural arches of the presacral vertebrae, and hypantrum-hyposphene intervertebral articulation of A. madagaskarensis otherwise are more typical of members of Archosauria (see Butler et al., 2012) than of early archosauromorphs (see description above). Moreover, A. madagaskarensis lacks any postaxial intercentra, a character state that is synapomorphic well within Archosauriformes (for Vancleavea campi + Archosauria ; see Nesbitt et al., 2009). This suite of character states in a highly autapomorphic taxon illustrates a continued theme in recent discovery and descriptions of Triassic archosauromorphs (e.g., Effigia okeeffeae, Nesbitt and Norell, 2006 ; Nesbitt, 2007): rampant convergence. This pattern of frequent convergence urges caution in the identification of isolated, disarticulated remains in multitaxic bone beds or even of partial skeletons.

Deciphering early archosauromorph diversification currently lies at an interesting crossroads. Recent studies are rapidly advancing our understanding, particularly those focused on long-standing taxonomic enigmas (Proterosuchia; Ezcurra et al., 2013; Tanystropheidae ; Pritchard et al., 2015) and those detailing the anatomy of previously discovered taxa (e.g., Dilkes, 1998) as well as new taxa (Sues, 2003; our contribution here for A. madagaskarensis ) with an eye focused on distinguishing plesiomorphy from synapomorphy at various hierarchical levels of phylogenetic analysis. Despite the progress such studies have brought, a number of challenges persist, especially the dearth of “transitional” morphologies between the “stem” of the archosauromorph tree and later diverging, more autapomorphic lineages. Additionally, the apparent high rates of homoplasy in character evolution for early archosauromorphs have hindered our ability to tease apart plesiomorphy from derived character states. Targeting Lower and Middle Triassic sediments for discovery of more (and more complete specimens) of these early forms provides perhaps the only potential solution to this dilemma.