The diversity of Triassic South American sphenodontians: a new basal form, clevosaurs, and a revision of rhynchocephalian phylogeny

Rhynchocephalians are a group of lizard-like diapsid reptiles that were very diverse during the Mesozoic but are now restricted to a single extant genus in New Zealand. Recent cladistic analyses have revealed two major clades, Eusphenodontia and the more crownward Neosphenodontia, but relationships of individual taxa have remained difficult to determine because of missing data and an unrevised data matrix. Here we drastically revise the established data matrix on rhynchocephalians by reassessing, evaluating, and adding new characters and operational taxonomic units, differing from any previous analyses in our goal to consider all known rhynchocephalians. In addition, we describe a new genus and species of an early eusphenodontian taxon from the Norian of southern Brazil, with a unique mosaic of plesiomorphic and apomorphic traits, and we re-examine the craniodental anatomy of the eusphenodontian Clevosaurus brasiliensis with µCT imaging, revealing a unique form of acrodonty amongst rhynchocephalians. http://zoobank.org/urn:lsid:zoobank.org:pub:A9211C5A-D4F9-472A-B8AB-877D13ABFDD5


Introduction
Rhynchocephalia, a diapsid lizard-like group of reptiles and sister group to Squamata, was diverse and globally distributed during most of the Mesozoic but is now represented by a single genus in New Zealand (Hay et al. 2010). The earliest diverging rhynchocephalian taxa, such as Gephyrosaurus and Diphydontosaurus (Evans 1980;Whiteside 1986), had at least in part pleurodont dentition, but later forms were fully acrodont, some of the oldest of which are known from the Late Triassic of South America. There have been considerable efforts to reconstruct the phylogeny of the rhynchocephalians, but results have been mixed.
South America yields a quite remarkable record of Triassic rhynchocephalians, with specimens known from the late Carnian to early Norian of the Santa Maria Supersequence (Zerfass et al. 2003;Bonaparte & Sues 2006;Horn et al. 2014;Schultz et al. 2020) of southern Brazil, and the middle Norian to late Rhaetian of western Argentina (Mart ınez et al. 2013;Apestegu ıa 2016). The Brazilian fossils are Clevosaurus hadroprodon (Hsiou et al. 2019, but see Supplemental material) and an indeterminate genus and species based on an isolated palatine (Romo-de-Vivar-Mart ınez et al. 2021), both from the late Carnian Hyperodapedon Assemblage Zone (AZ) at the base of the Cand elaria Sequence. Two more taxa, Clevosaurus brasiliensis (Bonaparte & Sues 2006) and Lanceirosphenodon ferigoloi (Romo de Vivar et al. 2020a) from the early Norian Riograndia AZ, occur at the top of the Candel aria Sequence . The record from Argentina includes a specimen mentioned briefly from the mid-to-late Norian Los Colorados Formation, Ischigualasto-Villa Uni on Basin (Apestegu ıa 2016), and Sphenotitan leyesi from the late Rhaetian Quebrada del Barro Formation, Marayes-El Carrizal Basin (Mart ınez et al. 2013). Clevosaurus brasiliensis and Sphenotitan leyesi are the most abundant components within their respective faunal associations (Bonaparte et al. 2010;Mart ınez et al. 2013;Romo de Vivar et al. 2020a).
Here, we present two South American rhynchocephalian taxa in detail, based on specimens from the Late Triassic Riograndia AZ of the Candel aria Sequence (early Norian, Late Triassic) of southern Brazil and excellent mCT scan data. First, we describe the cranial anatomy of the early eusphenodontian Clevosaurus brasiliensis based on three skulls (including its holotype) and review the material in the literature attributed to this taxon. Second, we describe a new rhynchocephalian genus and species. The holotype and referred specimens were formerly considered to be from juveniles of Clevosaurus brasiliensis (Bonaparte & Sues 2006; Romo de Vivar Mart ınez & Soares 2015), but we reidentify them as belonging to a new taxon based on their unique traits, after a review of the large collection of rhynchocephalians from this AZ (mainly represented by Clevosaurus brasiliensis material) at the Paleovertebrate Collection of the Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil. The evidence for the validity of the new taxon came initially from a morphometric analysis of the jaws of Clevosaurus brasiliensis (Romo de Vivar Mart ınez & Soares 2015) in which the supposed juveniles grouped separately. Since then, new mCT scans and detailed study of the cranial anatomy presented below demonstrate that it is a new genus and species with significant anatomical differences and lacking numerous apomorphies of Clevosaurus.
Our detailed anatomical studies of the two taxa have allowed us to consider the wider relationships of the South American rhynchocephalians. There have been a number of phylogenetic analyses of Rhynchocephalia in recent years, but these have been dominated by a moreor-less shared data matrix (e.g. Apestegu ıa & Novas 2003;Apestegu ıa & Carballido 2014;Apestegu ıa et al. 2014;Hsiou et al. 2015;Herrera-Flores et al. 2018;Romo de Vivar et al. 2020a;Simões et al. 2020), and all analyses met similar problems of poor resolution and poor support values because of incompletely coded taxa. We were able to cross-check and substantially add to the character list and include rhynchocephalian taxa that had never been included in phylogenetic analyses before, and so develop a more comprehensive analysis than previously attempted.

Materials
The fossil material used for this study comprises four nearly complete skulls, three belonging to Clevosaurus brasiliensis (holotype UFRGS-PV-0748-T, UFRGS-PV-0974-T, UFSM-PV-0131) and the holotype specimen (UFRGS-PV-0972-T) of a new genus and species. These skulls were subjected to lCT scanning at the University of Bristol using a Nikon XTH225ST CT scanner, with scan settings of 150 kVP, 175 lA, 1 s exposure and a copper filter of 0.5 mm, except for UFRGS-PV-0748-T, which used scan settings of 150 kVP, 46 lA, 1.415 s and no filter. Each of the four fossils used 3141 projections at an average of one frame per projection acquired during a full rotation of 360 and a source to detector distance of 1176 mm. UFSM-PV-0131 had a source to object distance of 88 mm, resulting in a 15.0 lm reconstructed voxel resolution; UFRGS-PV-0748-T had a source to object distance of 91 mm, resulting in a 18.9 lm reconstructed voxel resolution; UFRGS-PV-0972-T had a source to object distance of 86 mm, resulting in a 14.6 lm reconstructed voxel resolution; and UFRGS-PV-0974-T had a source to object distance of 111 mm, resulting in a 18.9 lm reconstructed voxel resolution. The scans were processed using Avizo Lite 9 (FEI Visualisation Sciences Group) to reconstruct segmented 3D models. Other specimens (e.g. UFRGS-PV-0848-T, UFRGS-PV-0827-T) were studied by first-hand examination.

Phylogenetic analysis
The phylogenetic analysis of the relationships of rhynchocephalians is based on the largest version of the traditional data matrix of Rhynchocephalia (49 operational taxonomic units, 73 characters) as revised by Romo de Vivar et al. (2020a), but further substantially modified here in terms both of operational taxonomic units (OTUs) and characters. We modified 29 characters, removed 14 and added 82 original characters of our own, in addition to one character from Rauhut et al. (2012), 17 characters from Simões et al. (2020), and three from Dupret (2004). This resulted in a matrix with a total of 162 characters.
We paid particular attention to the inclusion and exclusion of taxa in our data matrix. We noted that many of the previous analyses excluded several rhynchocephalians represented by fragmentary material, but this had not been done in a consistent way. For example, some relatively more complete taxa (e.g. Vadasaurus, SNSB BSPG 1993 XVIII 3, Colobops noviportensis, Tingitana and the 'Kirtlington sphenodontian') have been ignored in favour of less complete taxa (e.g. Fraserosphenodon, Kawasphenodon, Pelecymala). Here, we address this problem by coding all known species of rhynchocephalians, in addition to unnamed species, that have been described in the literature. We strove to represent as many morphotypes as possible, considering all rhynchocephalians equally. We then: (1) removed those that were considered problematic (see Supplemental material); (2) checked for redundancy in coded taxa so that we still represent as many morphotypes as possible; and (3) identified and removed rogue taxa before analysis began. Unlike Romo de Vivar Mart ınez et al. (2020a), we chose not to include Homoeosaurus cf. maximiliani as there may be multiple species of Homeosaurus besides H. maximiliani, and there is not sufficient description in the literature to differentiate them. We also combined the three Chinese clevosaurs into one OTU as it is unclear whether they represent separate species (see Jones 2006). See Supplemental materials for discussion of taxa that were not included.
Our initial matrix comprises 79 taxa (76 rhynchocephalians and three outgroups), where previously the largest number in a phylogenetic analysis was 47 (Romo de Vivar Mart ınez et al. 2020a, prior to removal of rogue taxa). We studied the South American taxa (including Lanceirosphenodon ferigoloi) first-hand, in addition to seven British taxa (see Supplemental material for more information). Other taxa were re-evaluated through a combination of literature review and highresolution photography and unpublished images provided by the institutions holding certain specimens.
Prior to analysis, we checked our matrix of 79 OTUs and 162 characters in PAUP 4.0 (Swofford 2002) to identify redundant taxathose that share >99% of their characters. This eliminated 17 taxa, reducing the number of taxa to 62 OTUs and rendering characters 13 and 130 as parsimony uninformative. We then checked the reduced matrix in TNT v. 1.5 (Goloboff & Catalano 2016) to identify and remove rogue taxa using the iterPCR protocol, namely Whitakersaurus bermani, Deltadectes elvetica and Kawasphenodon expectatus. The final matrix therefore had 59 OTUs, of which 56 were rhynchocephalians and three were outgroups, and 160 parsimony-informative characters. See Supplemental material and Supplemental Fig. 3.
We rooted the trees with three outgroups, the lepidosauromorph Sophineta cracoviensis (Evans & Borsuk-Białynicka 2009) and two extant squamates, the gecko Aeluroscalabotes felinus G€ unther, 1864 and the iguanian Pristidactylus torquatus Philippi & Landeck, 1861, the latter having been used in previous phylogenetic analyses (Hsiou et al. 2015;Herrera-Flores et al. 2018). We chose to replace the previously used outgroup Eichstaettisaurus with Aeluroscalabotes as the latter is an extant taxon that can be more completely coded.
The revised taxon-character matrix was analysed under equally weighted maximum parsimony using TNT v. 1.5 (Goloboff & Catalano 2016) and Bayesian inference using MrBayes v. 3.2 (Huelsenbeck & Ronquist 2001;Ronquist et al. 2012). All characters were considered as non-additive (unordered). The TNT parsimony search strategies started using a traditional heuristic search of 5000 replicates of Wagner tree followed by tree bisection and reconnection (TBR) as the branch swapping algorithm. The best trees obtained were subjected to a final round of TBR branch swapping to find all MPTs. The possible phylogenetic positions and clade robustness of all taxa were assessed with Bremer decay indices from TBR branch swapping (Bremer 1994). Bayesian analysis was run for 10 7 generations, sampling parameters every 1000 generations and the first 25% of sampled trees were discarded as burn-in. All recovered MPTs were then summarized in a strict and a 50% consensus (MRC) tree. A maximum clade credibility (MCC) tree was also generated (Supplemental Fig. 5 Holotype. UFRGS-PV-0748-T (Figs 1A-C, 2), originally described by Bonaparte & Sues (2006), is a nearly complete, articulated skull of an adult specimen, $23 mm long and slightly dorsoventrally crushed. Many of the bones display cracks and some bones were presumably lost at the time of discovery or during preparation, with the right vomer and much of the left vomer missing, as well as the paroccipital processes, and some of the posterior-most region of the right articular.
Additional specimens used here. UFSM-PV-0131 ( Fig.  1C, D, Supplemental Fig. 1) is a nearly complete skull of length $ 31 mm, which has undergone some dorsoventral flattening. The contrast between bone and matrix is not high and details of the teeth are near impossible to distinguish. Much of the braincase is missing, but the fused left opisthotic and prootic are present, along with a fragment of the cultriform process and the right basioccipital process. UFRGS-PV-0974-T (Fig. 1E, F, Supplemental Fig. 2) is a nearcomplete skull, laterally crushed, $21 mm long, with markedly less worn teeth than the holotype and UFSM-PV-0131. Three cervical vertebrae are also present. The crushing has displaced and destroyed many of the roofing bones of this skull.
Locality and horizon. The Linha São Luiz outcrop (29 33 0 45 00 S, 53 26 0 48 00 W) is located about 1.5 km north-west of Faxinal do Soturno city, Rio Grande do Sul State, southern Brazil. Its fossil content is referred to the Riograndia AZ (Bonaparte et al. 2010;Soares et al. 2011;Martinelli et al. 2020). This location exposes the top of the Candel aria Sequence (Santa Maria Supersequence) and corresponds to the base of traditional Caturrita Formation (Andreis et al. 1980). The fossils were obtained from massive, fine-grained sandstones, interpreted as deposited in a deltaic or fluvial environment (Horn et al. 2014). Based on zircon U-Pb analyses, the maximum depositional age of the fossil-bearing layers is Norian at $225.42 ± 0.37 Mya (Langer et al. 2018). The Linha São Luiz outcrop has produced the procolophonian Soturnia caliodon (Cisneros & Schultz 2003), the rhynchocephalians Clevosaurus brasiliensis (Bonaparte & Sues 2006) and Lanceirosphenodon ferigoloi (Romo de Vivar et al. 2020a), the non-rhynchocephalian lepidosauromorph Cargninia enigmatica (Bonaparte et al. 2010;Romo de Vivar et al. 2020b), an archosaur of uncertain affinities, Faxinalipterus minima (Bonaparte et al. 2010;Soares et al. 2013), the dinosaur Guaibasaurus candelariensis (Bonaparte et al. 1999(Bonaparte et al. , 2007, and the probainognathian cynodonts Brasilodon quadrangularis (¼ Brasilitherium riograndensis, Minicynodon maieri), Riograndia guaibensis and Irajatherium hernandezi (e.g. Bonaparte et al. 2001Bonaparte et al. , 2003Bonaparte et al. , 2005Bonaparte et al. , 2010Bonaparte et al. , 2012Martinelli et al. 2005Martinelli et al. , 2017Oliveira et al. 2011;Soares et al. 2011). Bonaparte & Sues 2006;Hsiou et al. 2015). A moderately sized rhynchocephalian with a skull 21-31 mm long, markedly short robust snout and broad skull, and with the following unique combination of features: 1. Acrodont teeth that extend deeply within the premaxillae, maxillae and dentaries, and thereby remain visible along the entire length of the toothbearing portions of the jaws even when worn flat. 2. Adult maxilla with two posteriorly positioned flanged teeth, followed distally by one or two smaller, sub-conical teeth. 3. Narrow and elongate palatine, bearing anteriorly a medially positioned cluster of small teeth, in addition to a single small tooth that is laterally displaced from the main palatine tooth row. 4. Interdigitation between the edges of the jugal and postorbital bones, the prefrontal and maxilla, and the pterygoid and palatines, rather than simple overlapping facets (Supplemental Fig. 4A-D). 5. Three longitudinal rows of teeth on the pterygoid, the lateral-most row consisting of just three to four teeth. 6. A near vertical symphysis (less than 120 in lateral view between the direction of the symphysis and the longitudinal axis of the mandible). 7. A pronounced edentulous region between the posterior-most dentary tooth (ultimate additional tooth) and the coronoid process. 8. The ultimate additional tooth on the dentary is significantly larger than all other dentary teeth and is mesiodistally elongated, with an anteriorly placed cusp when unworn.

Description
Skull. Among the three specimens examined here, almost all bones of the skull can be accounted for, with the exceptions of a complete vomer and a verifiable stapes.
Premaxilla. Each premaxilla (Fig. 3A, B) bears a single tooth that extends deep into the premaxillary bone, and which is more exposed on the lingual side as the bone is worn away by the occlusion of the lower jaw, forming a chisel-like edge. The presence of a single tooth on the premaxilla is a condition shared with C. bairdi (Sues et al. 1994) but it is less prominent here and nearly in line with the maxillary teeth. The nasal process is narrow and columnar but flattens mediolaterally where it comes into contact with the nasal. The maxillary process is as tall as the nasal process and is mediolaterally flattened and broad. As in the premaxillae of other Clevosaurus (Fraser 1988;Klein et al. 2015;  Keeble et al. 2018), the maxillary process bifurcates, but this secondary maxillary process is greatly reduced in size compared to the primary maxillary process and there is also very little space between the tooth-bearing region of the premaxilla and the point where the maxillary process bifurcates. Clevosaurus convallis (S€ ail€ a 2005) lacks this spur altogether. The premaxilla is robust and forms a brace for the snout, contacting the nasal at three points: (1) lateral surface of nasal process of premaxilla and medial surface of premaxillary process of nasal (Fig. 3A, B); (2) medial surface of primary maxillary process and lateral surface of nasal   fig. 1B), there is no posterior curvature of the maxillary dorsal process forming the sub-and antero-orbital margin (Fig. 3C). Posteriorly, the maxilla wraps around the jugal ventrally, but also bears facets for the premaxilla, nasal, prefrontal and ectopterygoids along its medial surface. Dorsally, the maxilla fits into a slot on the ventral edge of the anterior face of the descending process of the prefrontal so that the latter partially overlaps the maxilla both medially and laterally, forming a brace-like structure. The posterior process of the maxilla is broad, only tapering rapidly posteriorly beyond the end of the tooth row. The maxillary teeth are all deeply embedded in the bone, but do not possess true roots. This form of acrodonty ( Fig.  4D-H, L, O) is not currently known in any other rhynchocephalian. Anteriorly, the maxilla possesses 12-13 small, presumably hatchling, teeth that are fused together and worn flat to the bone, most noticeably lingually ( Fig. 3D, E), so that they form a sharp continuous cutting surface. Behind this 'beak' of hatchling teeth, C. brasiliensis typically bears two large, flanged teeth, a condition also seen in the other Norian clevosaur Brachyrhinodon and possibly also in Polysphenodon (Fraser & Benton 1989). The Early Jurassic C. bairdi (Sues et al. 1994) may have as few as three flanged teeth upon its maxilla, fewer than in clevosaurs from the UK and China. Most posteriorly, there are one or two smaller conical teeth, similar to C. hudsoni (Fraser 1988), but fewer than the three or four of Microsphenodon (see below).
Nasal. The nasals ( Fig. 3F-H) are paired bones that are positioned nearly vertically and are much more convex than the nasals of C. cambrica (Chambi-Trowell et al. 2019) or Microsphenodon (see below), reflecting the steep short snout. Contrary to the original description (Bonaparte & Sues 2006) and the revision based on specimen MCN-PV 2852 (Hsiou et al. 2015), we find that the nasals are not fused with the frontals instead they interdigitate (Fig. 2D, Supplemental Fig. 4H). Anteriorly, they divide, forming an internarial slot where the nasal processes of the premaxillae could articulate. Laterally they bear a broad facet for the prefrontal, as well as facets for the premaxilla and maxilla. Posteroventrally, there was a facet for the frontal.
Prefrontal. The prefrontal (Fig. 3H, I) forms the anterior half of the orbital margin, it is large and lunate in shape, and steeply curved where its anterodorsal process overlaps the frontal. The ventral process is expanded mediolaterally and forms a broad contact with the dorsal surface of the palatine (Fig. 2D) and would also have contacted the jugal. On the lateral surface of the prefrontal there is a broad partial slot facet forming a partial overlap of the top of the maxilla on both its medial and lateral sides.
Postfrontal. The postfrontal (Fig. 3J, K) is a small triradiate bone with a concave ventrolateral facet for the postorbital and a medially positioned concave facet that overlaps the parietal and frontal dorsally. It forms the posterodorsal margin of the orbit, and anterodorsal corner of the supratemporal fenestra. It is relatively larger than in other species such as C. hudsoni or C. cambrica Frontal. When paired, the frontals (Fig. 3L, M) show the typical hourglass shape of Clevosaurus, and they are not fused, but interdigitate with one another along the medial axis. At the frontoparietal suture, the frontals diverge from one another posteriorly, forming two flattened processes that overlap the parietals. Anterolaterally, there is a step-like indentation indicating the prefrontal facet. The frontals contribute very little to the orbital margin.
Parietal. Articulating anterodorsally with the frontals, the parietals (Fig. 3L) are paired and not fused. Like other species of Clevosaurus (e.g. Fraser 1988, p. 174;Sues et al. 1994;Chambi-Trowell et al. 2019), they lack a parietal crest, differing from the condition seen in Sphenodon. Posteriorly the elongate posterior processes bear a distinct slot-like facet dorsally for the  Hsiou et al. (2015). The posterior process appears to be more ventrally deflected than that of the horizontal temporal bar of Clevosaurus hudsoni (Fraser 1988;O'Brien et al. 2018) or Sphenodon, and more like that of Brachyrhinodon (Fraser & Benton 1989). The anterior process of the jugal reaches the anterior orbital margin, where it contacts the prefrontal (Figs 1D, 3D), and has a broad contact with the ectopterygoid medially and slots into the medially curved posterior process of the maxilla. The anterior process Abbreviations: add, additional; alc, alveolar canal; cap, pulp cavity; d, dentary; ht, hatchling; j, jugal; m, maxilla; mc, Meckelian canal; pal, palatine; pm, premaxilla; prf, prefrontal; pt, pterygoid; th, tooth. Scale bars for A-K equal 1 mm; L-O equal 10 mm. may also have contacted the maxillary process of the palatine anteriorly. The dorsal process contacts the postorbital, with an interdigitating suture rather than simple overlapping facets as in C. hudsoni or C. cambrica (Fraser 1988;Chambi-Trowell et al. 2019). Posterodorsally, the dorsal process overlaps the squamosal. Overall, the jugals are more robust than in European Clevosaurus (Fraser 1988;Klein et al. 2015) and similar to those of C. bairdi (Sues et al. 1994).
Postorbital. A large triradiate bone, the postorbital ( Fig.  3P, Q) is similar in size to that of C. bairdi (Sues et al. 1994), though relatively smaller and with a shorter ventral process, unlike other Clevosaurus, such as C. cambrica and C. hudsoni (Fraser 1988;Chambi-Trowell et al. 2019). Uniquely for Rhynchocephalia, the postorbital does not overlap the dorsal process of the jugal but rather interdigitates with it along its ventral edge. Posteriorly, it has a broad facet that overlies the squamosal. The dorsal process is large and contacts the postfrontal ventrally, partially wrapping around its anterior margin.
Squamosal. The ventral process of the squamosal (Fig.  3R, S) is broad and curved, forming a broad contact with the quadrate. The anterior flange is large and articulates dorsally with the postorbital and jugal. The medially curved posterior process articulates with the supratemporal on its posterolateral surface. The medial surface is, for the most part, smooth and concave, but dorsal to the ventral process the squamosal widens medially, forming a robust cup-like structure (Fig. 3S) that articulates with the dorsal condyle of the quadrate. On the lateral surface there is a depression that starts from the anteroventral-most point of the squamosal and initially runs posterodorsally at a shallow angle, before curving ventrally with the ventral process and terminating half-way down the posterior margin of the ventral process.
Supratemporal. The supratemporal ( Fig. 5A-C) is a convex plate-like bone that connects the squamosal to the parietal along its medial surface and forms the posterior margin of the supratemporal fenestra. Midway along its long axis there is a parallel indentation, possibly marking the origin of the M. depressor mandibulae bs, basisphenoid; cl, cluster; epi, epipterygoid; ept, ectopterygoid; fac, facet; for, foramen; lat, lateral; m, maxilla; p, parietal; pal, palatine; pla, plate; pr, prootic; pt, pterygoid; q, quadrate; qj, quadratojugal; rid, ridge; sq, squamosal; st, supratemporal; th, tooth; ven, ventral; ver, vertical; vo, vomer. Scale bars for A-G, I-K, M equal 5 mm; H, L equal 1 mm. that inserted on the remarkably long retroarticular process. The supratemporal is roughly rectangular in shape, similar to that of other Clevosaurus (Fraser 1988;Sues et al. 1994;Jones 2006) and Microsphenodon. Posteriorly, the supratemporal curves anteriorly, forming a hook that would have articulated with the posterior edge of the dorsal condyle of the quadrate. On the posterior surface of this hook, the supratemporal articulated with the paroccipital process.
Quadrate. The quadratojugal cannot be distinguished from the quadrate. Anteriorly, the thin quadrate flange tapers to a point. The quadrate ( Fig. 5D-G) is a complex element, forming a thin lateral plate posterolaterally, with a raised, reinforced posterior margin, and posteromedially a robust strut (Fig. 5D, F) between the cephalic and articular condyles. The ventral process of the squamosal articulated to the quadrate anterior to the raised margin of the lateral plate, while most of the squamosal articulated on the cephalic condyle of the quadrate. The vertical strut and lateral plate contact one another dorsally and ventrally, forming a conch-like structure that is concave anteriorly, and with a quadratojugal foramen present anterolaterally. The quadrate flange is reinforced along its ventral margin where it curves, forming a raised rim into which the flange of the pterygoid slotted (Fig. 5G). The bone is spongy around both cephalic and articular condyles. The vertical strut is much more robust than observed in other Clevosaurus (Fraser 1988;Chambi-Trowell et al. 2019, fig. 2D, E), and this strut twists so that a raised ridge ( Fig. 5F) begins on the lateral side towards the dorsal end of the quadrate and ends up contacting the medial side ventrally.
Vomer. No complete vomers are known, but a fragment of the left vomer is preserved still articulated dorsally to the anterior of the palatine and pterygoid in both UFRGS-PV-0748-T ( Fig. 5H) and UFRGS-PV-0974-T. Known vomer fragments bear two rows of small teeth, with three small medial row teeth on a raised bony ridge, and a lateral row with another three teeth in MCN-PV 2852 (Hsiou et al. 2015). Although represented by just fragments, it is clear there are more teeth on the vomer of C. brasiliensis than in C. hudsoni or C. cambrica (Fraser 1988 (Evans 1980;Fraser 1982;Whiteside 1986), or even non-rhynchocephalians such as Kuehneosaurus (Robinson 1962).
The Clevosaurus brasiliensis palatines are more similar in shape to those of Sphenotitan, being very narrow and widening abruptly anteriorly where they bear a medial cluster of smaller teeth. However, Sphenotitan has 12 to 13 teeth in the lateral row whereas C. brasiliensis has five to seven. The palatines are elongate and narrow, forming an elevated ridge that interdigitates along the medial side with the anterior pterygoid process and widens abruptly, but not extensively, anterior to the pterygoid. The palatine bears the typical lateral row of large teeth, five to seven in a row, as in many Clevosaurus species (Fraser 1988;S€ ail€ a 2005;Jones 2006;Klein et al. 2015;Chambi-Trowell et al. 2019) and Microsphenodon, but more than the four in C. minor (Fraser 1988). There is a cluster of four to six small, medially positioned teeth in the same anterior region as the large singular medially placed tooth that is found in all other Clevosaurus; one palatine of C. hudsoni (AUP 11158;Fraser 1988, p. 137) is thought to have had a rudimentary second row of three teeth in the position of this singular tooth. Anterolaterally to the main tooth row, there is sometimes a single small, mesiodistally elongate tooth (Fig. 6B) that is fused to the lateral edge of the anterior-most tooth of the lateral row (clearly visible in UFRGS-PV-0748-T and UFRGS-PV-0974-T), forming a two-cusped tooth that is unlike anything seen in any other rhynchocephalian. The robust maxillary process bifurcates to form a very broad contact with the maxilla, possibly even reaching the premaxilla; posteriorly this process contacts the jugal, isolating the maxilla from the orbit medially. Posteriorly, the palatine ventrally overlaps the ectopterygoid. Similar to other species of Clevosaurus (cf. C. minor ;Fraser 1988, pp. 159-160), it possesses a concave vomerine facet. Dorsally, the prefrontal articulates with the anterior edge of the palatine. Unlike the curved lateral tooth row of the Chinese clevosaurs (Wu 1994;Jones 2006, fig . 1B), the row of palatine teeth can be almost completely straight or even medially concave (Wu 1994;Jones 2006).
Pterygoid. Uniquely among Clevosaurus species, but like more basal sphenodontians such as Planocephalosaurus (Fraser 1982), the pterygoids ( Fig.  5J-M) bear three rows of teeth. The lateral-most row is the shortest, with just three teeth, the same as Microsphenodon, whereas there are only two teeth in the possibly contemporary clevosaur Brachyrhinodon (SAVC-T, pers. obs.) and in Planocephalosaurus. There is some alternation in the size of teeth on both the pterygoids of UFRGS-PV-0974-T (Fig. 4J, K). As in Clevosaurus cambrica and Clevosaurus hudsoni (Fraser 1988, pp. 137-138;Chambi-Trowell et al. 2019), the ramus that runs between the tooth-bearing region of the pterygoid and the articulation point for the basipterygoid processes is relatively short compared to Sphenodon. The pterygoid flange articulates medially with the quadrate flange, which partially wraps around its ventral margin. The pterygoid lateral processes that articulate ventrally with the ectopterygoids appear typical of Clevosaurus, the processes being roughly triangular in shape with an elongated posterior end. Posteriorly, the pterygoid bends sharply as the pterygoid flanges extend posterolaterally, and there is a posteriorly positioned rounded facet for the basipterygoid process. About halfway along the flange is a small dorsally positioned spike (Fig. 5K, M) where the epipterygoid articulates. The two pterygoids articulate anteriorly with an interdigitated facet, and there is a pronounced ridge of bone on either side of this facet on the dorsal surfaces of the pterygoids.
Ectopterygoid. The ectopterygoid (Fig. 5L, M) appears typical of Clevosaurus. Anteriorly, its lateral process contacts the palatine lateral process, which has a broad contact laterally with the maxilla and jugal, as in C. bairdi (Sues et al. 1994) and C. cambrica (Chambi-Trowell et al. 2019). Medioanteriorly on its ventral side it bears a facet for the palatine, and medioposteriorly it articulates with the pterygoid. Its posterior lateral process is roughly triangular and articulates with a similar process on the pterygoid.  (Sues et al. 1994) and Microsphenodon. This shape differs markedly from C. hudsoni NHMUK PV R36832 (O'Brien et al. 2018), which suggests that the epipterygoid described there, although somewhat reminiscent of the epipterygoid of Sphenodon, might be the broken displaced right paroccipital process.
Parabasisphenoid. The parasphenoid and basisphenoid are fused (Fig. 7A-E), as in other species of Clevosaurus (Fraser 1988, fig. 18;Hsiou et al. 2015;Chambi-Trowell et al. 2019), forming the floor of the braincase and widening posteriorly. Anteriorly, there are two elongate basipterygoid processes that articulate with the pterygoids. The cultriform process has broken off in the specimens that we examined, but in MCN-PV 2852 (Hsiou et al. 2015) it is seen to be thin and elongate.  anteroventrally with the parabasisphenoid and bears anteroposteriorly elongated grooves anteriorly of the basal tubera (Fig. 7A, E), where the exoccipitals would have articulated dorsally.
Exoccipitals. Dorsally, the exoccipitals ( Fig. 7A-E) are broad and rounded, tapering rapidly towards the supraoccipital, and would have formed more than 50% of the foramen magnum, including the lateral margin, as well as a portion of the ventral and dorsal margins. Ventrally, they articulate with the basioccipital and anterodorsally with the opisthotic. Unlike the condition in basal rhynchocephalians such as Gephyrosaurus, Planocephalosaurus and Diphydontosaurus (Evans 1980;Fraser 1982;Whiteside 1986), as well as Microsphenodon, the exoccipitals are not fused to the basioccipital, a feature in common with other species of Clevosaurus (Fraser 1988;Klein et al. 2015;Chambi-Trowell et al. 2019).
supraoccipital forms part of the dorsal margin of the foramen magnum, and its dorsal surface is broad and concave with no dorsal crest. The pila antotica appears typical for Clevosaurus but also broken on both sides so that it would likely have stood more upright in life. The prootic bears a large concave facet posteriorly for the exoccipital. The paroccipital processes are elongate and articulate with the supratemporals.

Mandible.
Dentary. The dentary (Fig. 8A-E, I-K) shows a pronounced gap between the ultimate tooth and the coronoid process, which may be concave in lateral profile, as is generally the case in clevosaurs, except in some specimens of C. convallis (S€ ail€ a 2005). The coronoid process is markedly high, and the dentary is the most robust of all Clevosaurus, although it is also sturdy in C. convallis (S€ ail€ a 2005) and C. bairdi (Sues et al. 1994). In cross-section, the dentary of the holotype was observed to have three possible growth rings (Fig. 4D), indicating three cycles of growth, though it is uncertain whether these correspond with years. The symphysis is near vertical and robust, and lacks any concavity medially (Fig. 8C), differing from European clevosaurs such as C. sectumsemper and C. cambrica (Klein et al. 2015, fig. 4;Chambi-Trowell et al. 2019). The mentonian process is reduced but rounded and robust (Fig. 8D). The dentary bears 12-13 teeth in total, the ultimate additional tooth is huge (two to three times larger than any of the other teeth) and is mesiodistally elongated. Most anteriorly, there are eight to 10 teeth that become gradually smaller anteriorly and are considered here to be hatchling teeth. These are fused together (Fig. 4M) and are worn flat to the dentary, with the lateral edge of the bone worn away to expose their long 'roots' (Fig.  4G, H, O), thereby forming a continuous cutting surface like a beak. Posteriorly, there are three to four additional teeth, each circular in cross-section, except for the large mesiodistally elongated posterior-most tooth. In some specimens, there can be as few as two additional teeth (UFRGS-PV-1153-T; Romo de Vivar Mart ınez & Soares 2015).
The tooth-bearing region of the dentary is reduced relative to the length of the entire mandible and lacks the diagonal wear facets seen on the dentaries of the UK Clevosaurus (Fraser 1988;Klein et al. 2015;Chambi-Trowell et al. 2019). The wear facets of C. brasiliensis are smooth and form a continuous secondary ridge above the secondary dentine (Fig. 8D), possibly similar to those of C. mcgilli (Jones 2006). There is a pronounced lip of secondary bone along the lateral surface of the dentary, as in many rhynchocephalians. Clevosaurus brasiliensis lacks the characteristic saddleshaped additional dentary teeth of most European Clevosaurus (Fraser 1988;Klein et al. 2015;Chambi-Trowell et al. 2019) and bears triangular teeth which are heavily worn dorsally. There are no flanges on the dentary teeth, and they lack noticeable escape structures. The teeth of C. brasiliensis are acrodont but are deeply imbedded within the jaws, up to half the depth of the dentary in places (Fig. 4D, G, O), a condition not seen in any other rhynchocephalian. The teeth do not have true roots or sockets, but rather they are packed together in a long groove, many showing fusion to the next tooth, with bone surrounding them both labially and lingually. We suggest that this condition might have developed by growth of secondary bone over the bases of the teeth to such an extent that the teeth are nearly obscured with secondary bone. This is further supported by the condition in UFRGS-PV-0974-T, which we infer to be the youngest of the three specimens because its teeth are noticeably less worn than the others, where the deep acrodont implantation of the teeth is more pronounced than the posterior-most teeth (Fig. 4F).
Coronoid. The coronoid (Fig. 8A-E) is a simple inverted tear drop-shaped bone that is laterally flattened, tapering ventrally. Similar to C. bairdi, C. cambrica and C. hudsoni (Sues et al. 1994;Chambi-Trowell et al. 2019) it projects slightly above the coronoid process. It is articulated laterally to the dentary and posteriorly to the surangular. In some specimens (UFSM-PV-0131, UFRGS-PV-0974-T; Fig. 8D, E) the coronoid projects above the dentary, and appears to curve backwards as in Sphenodon and many squamates.

Articular
complex (prearticular, surangular, articular and angular). The prearticular and articular are not fused (Fig. 8A-E), unlike in most C. hudsoni, C. cambrica and C. convallis specimens (Fraser 1988; S€ ail€ a 2005; Chambi-Trowell et al. 2019), though a lack of fusion has been observed in some individuals of C. hudsoni (Fraser 1988, p. 141). The prearticular is a long, flattened bone that wraps medially beneath the articular, and laterally contacts the dentary; it becomes more rounded and complex in shape anteriorly, forming . The surangular process is moderately curved so that its dorsal surface is somewhat convex, articulating anteriorly with the posterior surface of the coronoid. The articular is a long, flattened bone that posteriorly contacts the prearticular, and tapers anteriorly, terminating below the posterior-most tooth. The dorsal surface of the surangular is also very broad mediolaterally and robust compared to other rhynchocephalians. The articular is robust with a remarkably long retroarticular process that forms a spoon-like structure that tapers posteriorly and is deepest laterally. The small mandibular foramen is formed between the surangular and dentary. The condyle is full of spongy bone.
Postcranial skeleton. Little postcranial material (Fig.  7J-N) is preserved in the specimens we examined. However, there are a few cervical vertebrae that had been dislocated from the skull in UFRGS-PV-0974-T, including a probable axis, but with the damage they have sustained dorsally, it is not possible to verify the position in the neck. UFRGS-PV-0974-T also appears to have fragments of a long bone, likely the humerus. Derivation of name. 'Micro' is derived from the Greek lijqό (mikr o) which means small, and 'Sphenodon' refers to the sole extant genus of rhynchocephalian. Like Sphenodon, the animal has caniniforms and pronounced differentiation of its teeth, but is itself a much smaller animal, therefore 'small Sphenodon'.
Derivation of name. The specific term 'bonapartei' refers to the late Argentinian palaeontologist Jos e F. Bonaparte (1928Bonaparte ( -2020, who led the main collections of fossil vertebrates from the Linha São Luis outcrop until 2005, deposited at the UFRGS, including the discovery of the holotype specimen here described. Additional referred specimens. UFRGS-PV-0613-T, UFRGS-PV-0848-T and UFRGS-PV-0827-T are referred to the new taxon, which were identified as C. brasiliensis in previous studies (Bonaparte & Sues 2006; Romo de Vivar Mart ınez & Soares 2015; Romode-Vivar-Mart ınez et al. 2021). UFRGS-PV-0613-T is a partial skull around 20 mm long, and is in similar condition to the holotype, with damage to its rostrum and the right side of the skull; it was described by Bonaparte & Sues (2006) as the paratype of C. brasiliensis, where they interpreted it to be a juvenile of the taxon. UFRGS-PV-0848-T is an isolated right mandible 24 mm long. UFRGS-PV-0827-T is an isolated right palatine.
Diagnosis. A small sphenodontian rhynchocephalian with a skull length of 19-22 mm, which has the following unique combination of features: 1. A single caniniform tooth present on both maxilla and dentary. 2. The maxillary and dentary dentition is acrodont and the largest teeth take the form of two enlarged mesiodistally elongated additional teeth at the posterior of the dentary and maxilla, posterior to a row of smaller non-hatchling teeth that alternate in size. 3. A broad flat parietal table. 4. Three to four small conical post-marginal teeth positioned behind the additional teeth of the maxilla. 5. A palatine with a single medially positioned tooth, in addition to a secondary row or ridge of two teeth that are positioned at 45 to the long axes of the main lateral tooth row. 6. Fused prearticular, articular and surangular bones. 7. Paired frontals and parietals. 8. Exoccipitals fused to basioccipital. 9. Two rows of vomerine teeth, three rows of pterygoidal teeth and two rows of palatine teeth.

Description
Skull. Nearly all bones of the skull are represented, but there is no complete vomer, any premaxillae or verifiable stapes.
Maxilla. The maxillary facial process ( Fig. 11A-C) is relatively high but not as much as in Clevosaurus brasiliensis and is broader. It bears a pronounced premaxillary process and a single caniniform (UFRGS-PV-0613-T, Romo de Vivar Mart ınez & Soares 2015, fig. 3C, 4F; Fig. 9B, C), which is followed by an edentulous region where the likely hatchling teeth have been worn to the bone, followed by around three to four smaller additional teeth that appear to alternate in size (Romo de Vivar Mart ınez & Soares 2015, fig. 4F), and ultimately two large additional teeth bearing flanges. All the teeth are acrodont (Fig. 4A-C, N). Posterior to the main tooth row are three to four small conical teeth. It is unclear whether the premaxillary process bore any successional teeth anterior to the caniniform, as this part of the maxilla is damaged. The posterior process is broad and only narrows abruptly close to its end. Unlike C. cambrica (Chambi-Trowell et al. 2019) and C. bairdi (Sues et al. 1994, fig. 1B), there is no posterior curvature of the maxillary dorsal process forming the antero-orbital margin; this shape is more akin to the maxilla of Diphydontosaurus (Whiteside 1986, fig. 4A). Unlike Diphydontosaurus, however, the suborbital process remains near uniform for much of its length rather than tapering. Posteriorly, the maxilla forms a ventrally positioned medial shelf that contacts the jugal and ectopterygoids. Dorsomedially, the maxilla is somewhat concave where it would have contacted the nasal and prefrontal.
Nasal. Only the left nasal ( Fig. 11D-F) is preserved, and the anterior portion is damaged. The nasals are large and elongated compared to those in the short robust snouts of Clevosaurus and more like those of Diphydontosaurus, Gephyrosaurus or Planocephalosaurus (Evans 1980;Fraser 1982;Whiteside 1986). The nasals are less convex than those of C. brasiliensis, reflecting a long rostrum positioned at a shallower angle.
Prefrontal. The prefrontal (Fig. 11D-F) appears relatively typical for rhynchocephalians, with a roughly lunate shape and a broad lateral contact with the maxilla, and medially with the nasal. Ventrally it contacts the palatine, but probably not the jugal unlike in C. brasiliensis. It has sustained some damage on its dorsolateral surface.
Frontal. The frontals (Fig. 11I, J) are not fused, unlike those of basal rhynchocephalians (Evans 1980;Fraser 1982;Whiteside 1986). Rather, they are paired, forming an hourglass shape, with two long posterior flattened processes that overlap the parietals dorsally. Anterodorsally there are facets for the prefrontals and nasals.
Parietal. The parietals (Fig. 11I, J) form a broad and relatively short flat parietal table, and they are paired, unlike the fused elements of some basal rhynchocephalians (Evans 1980;Fraser 1982;Whiteside 1986); the fusion is late in ontogeny in Diphydontosaurus according to Whiteside (1986). The posterior process on each parietal is short and tapers rapidly, with a distinctive slot on its dorsal surface for the supratemporal where it overlay both the parietal and quadrate.
Jugal. The jugal (Fig. 11K, L) formed a broad contact with the ectopterygoid medioventrally and with the maxilla laterally. It contacted the palatine anteriorly. The bone formed a broad contact with the postorbital along the lateral surface of its dorsal process, which also overlay the squamosal ventrally. The jugal did not interdigitate with the postorbital unlike C. brasiliensis. The posterior process is missing in all specimens.
Postorbital. The postorbital (Fig. 11M, N) is a triradiate bone that laterally overlapped the squamosal and jugal. The ventral process would have partially wrapped around the jugal along its lateral surface, a condition shared with European Clevosaurus (Fraser 1988;Chambi-Trowell et al. 2019) but differing from the interdigitating condition seen in the jugal in C. brasiliensis.
Squamosal. The squamosal (Fig. 11O, P) is generally typical of Clevosaurus (see Fraser 1988). There is some damage to the ventral process, but most of the bone is preserved. The ventral process is not curved or as broad as in C. brasiliensis. The lateral surfaces of both squamosals are damaged, so it is unclear whether there was a similar lateroventral depression as in C. brasiliensis.
Supratemporal. The supratemporal (Fig. 12A, B) is a small, convex, plate-like bone that connects the  isolated; j, jugal; lat, lateral; m, maxilla; med, medial; p, parietal; pal, palatine; proc, process; pt, pterygoid; q, quadrate; qj, quadratojugal; rid, ridge; sq, squamosal; th, tooth; ven, ventral; ver, vertical; vo, vomer squamosal to the parietal, with a ridge on its ventral side that articulates with both. It is roughly rectangular in profile, and midway along its long axes there is a parallel indentation, possibly indicating a site for attachment of the M. depressor mandibulae. Supratemporals have not been confirmed previously in any rhynchocephalians except clevosaurs (Jones 2006;Chambi-Trowell et al. 2020). The supratemporal may be damaged posteriorly, and it is unclear whether it articulated with the paroccipital process as in C. brasiliensis.
Quadrate. The overall shape of the quadrate (Fig.  12C-E) shares some similarity with C. brasiliensis, but it is far less robust, with a pronounced curved ventral condyle that is angled at roughly 45 to the vertical strut in medial view, resulting in a more posteriorly positioned dorsal condyle. The medially positioned shelf along the ventral edge of the quadrate flange is pronounced but flat and not curved dorsally as in C. brasiliensis. The bone is very porous around the condyles. A quadratojugal cannot be distinguished from the quadrate, and the quadratojugal foramen is small compared to that of Sphenodon punctatus and positioned ventrolaterally.
Vomer. No complete vomer is known. Only a fragment of the left vomer, articulated with the palatine, is preserved in the holotype (Fig. 12F). It bears two rows of teeth, one central row with three teeth preserved and one lateral with two teeth preserved. Other disarticulated teeth are found separately in the matrix. Numerous teeth on the vomer are a character typical of basal rhynchocephalians such as Gephyrosaurus, Planocephalosaurus and Diphydontosaurus (Evans 1980;Fraser 1982;Whiteside 1986), but two rows are also observed in C. brasiliensis.
C. sectumsemper (Fraser 1988;Klein et al. 2015;Chambi-Trowell et al. 2019). There is a lateral groove on the palatine between the lateral tooth row and the maxillary process, as previously noted (Romo de Vivar Mart ınez & Soares 2015; Romo-de-Vivar-Mart ınez et al. 2019). The maxillary process would likely have had a broad contact with the maxilla in life but the anterior part of the maxillary process has broken off. Posteriorly, the maxillary process contacts the ectopterygoid and jugal. The palatine articulates posteriorly on its dorsal surface with the ectopterygoid, and with the pterygoid along its medial ventral surface, excluding the pterygoid from the suborbital fenestra. This differs from the condition in non-eusphenodontians (Evans 1980;Fraser 1982;Whiteside 1986); however, the pterygoid has only the slightest contact with the suborbital fenestra in the reconstruction of Diphydontosaurus by Whiteside (1986, fig. 3b).
Pterygoid. The pterygoid (Fig. 12I, J, L) bears three rows of teeth, the third row being very short and consisting of just three teeth, as in C. brasiliensis, Brachyrhinodon and Planocephalosaurus (Fraser 1982). There is an interdigitating facet between the pterygoids anteriorly, and dorsally there is a raised ridge of bone on either side of this facet, but this is not as pronounced as in C. brasiliensis. The ramus between the start of the pterygoid flange and the lateral process is shorter than that of Sphenodon (Evans 2008) but is relatively longer and narrower than in C. brasiliensis. The pterygoid lateral processes widen laterally and articulate ventrally with the ectopterygoids. The basipterygoid facets are cup-like.
Ectopterygoid. The ectopterygoid (Fig. 12K) closely resembles that of C. brasiliensis but has a relatively longer, more gracile maxillary process and a shorter ventral process.
Epipterygoid. The epipterygoid (Fig. 12L) is rod-like dorsally and wider and more flattened ventrally where it articulates with the pterygoid flange. It is nearly identical to that of C. brasiliensis and other clevosaurs where this has been described, as well as Gephyrosaurus and Diphydontosaurus (Evans 1980;Whiteside 1986;Sues et al. 1994;Chambi-Trowell et al. 2019, fig. 8C).
Parabasisphenoid. The parasphenoid and basisphenoid are fused (Fig. 13A-F), forming the floor of the braincase, which is much narrower than in C. brasiliensis. Anteriorly, there are two elongate basipterygoid processes that widen anteriorly and articulate with the pterygoids. The cultriform process is elongate and reaches at least as far as the end of the interpterygoid vacuity.
There are two carotid foramina, one on each side of the base of the cultriform process (Fig. 13F).
Basioccipital-exoccipitals. The basioccipitals and exoccipitals are fused (Fig. 13A-F), as in Gephyrosaurus, Planocephalosaurus and Diphydontosaurus (Evans 1980;Fraser 1982;Whiteside 1986) and differing from the condition in C. brasiliensis and other clevosaurs. The exoccipitals would have formed more than 50% of the foramen magnum. The basal tubera are short and project posteroventrally. The occipital condyle is lunate in shape, with a slightly raised medial ridge positioned just anteriorly of the condyle, as in Diphydontosaurus (Whiteside 1986 , fig. 26a). The basioccipital articulates with the dorsal surface of the parabasisphenoid.
Opisthotics, prootics and supraoccipital. These bones (Fig. 13A-F) cannot be distinguished in the scans, and it is possible all three were fused, forming the roof of the braincase, as observed in C. hudsoni (Fraser 1988, p. 140;O'Brien et al. 2018;Chambi-Trowell et al. 2019) and C. brasiliensis. The roof of the braincase appears to have a much lower bone density than the rest of the skull and as a result is hard to distinguish from the matrix, a feature in common with C. cambrica and C. hudsoni (O'Brien et al. 2018;Chambi-Trowell et al. 2019) and C. brasiliensis. The supraoccipital forms part of the dorsal margin of the foramen magnum, and its dorsal surface is broad and relatively flat, with no dorsal crest present. The pila antotica is similar in shape to that of Clevosaurus, lateromedially flattened and curving dorsally, widening to a squared-off head.
Dentary. The dentaries (Fig. 8F-H, L) are relatively long and narrow, with a high coronoid process and they have the pronounced edentulous region between the ultimate tooth and coronoid process often seen in clevosaurs and in Diphydontosaurus and Planocephalosaurus (Fraser 1982;Whiteside 1986;Fraser 1988;Fraser & Benton 1989;Sues et al. 1994;Chambi-Trowell et al. 2019). As in most rhynchocephalians, there is a raised lip of secondary bone on the lateral surface of the dentary. The symphysis is long and slants anterodorsally at a low angle, and seemingly lacks the medially positioned post-symphyseal lamina seen in Lanceirosphenodon (Romo de Vivar et al. 2020a, fig.  5). The dentary has pronounced diagonal wear facets from the maxillary teeth, as often seen in Clevosaurus (Fraser 1988;Klein et al. 2015;Chambi-Trowell et al. 2019), as well as Diphydontosaurus, but not present in C. brasiliensis. The anterior-most region of the dentary appears to be edentulous (but as the scanned holotype has damage in this region of the jaw, we cannot verify this). There appears to be a single caniniform, followed by 4-5 hatchling teeth, then 4-6 teeth that alternate in size, increasing in size posteriorly (Fig. 8F Coronoid. The coronoid (Fig. 8F-H) is a small, thin bone; it is unclear whether it projected above the dentary.
Articular complex (prearticular, surangular and articular) and angular. The articular complex (Fig.  8D-F) appears to be fused, as the facets between the bones are indistinguishable from one another. The angular is flattened and reaches as far anteriorly as the posterior-most tooth of the dentary. The mandibular foramen is formed between the surangular and dentary and is large, unlike that of C. brasiliensis. The dorsal surface of the surangular is also very wide and robust compared to other rhynchocephalians, and the articular is robust, with a remarkably long retroarticular process that forms a spoon-like structure that tapers posteriorly and is deepest laterally, as in C. brasiliensis. The articular condyle is full of porous bone. The fusion of the  Abbreviations: ang, angular; art, articular; bo, basioccipital; bs, basisphenoid; cor, coronoid; d, dentary; epi, epipterygoid; ept, ectopterygoid; f, frontal; j, jugal; m, maxilla; n, nasal; p, parietal; pal, palatine; pm, premaxilla; po, postorbital; pof, postfrontal; pfr, prefrontal; pra, prearticular; ps, parasphenoid; pt, pterygoid; q, quadrate; so, supraoccipital; sq, squamosal; st, supratemporal; sur, surangular; vo, vomer. Scale bars equal 10 mm. prearticular, articular and surangular in Microsphenodon is an important difference from Clevosaurus brasiliensis but a feature shared with early diverging rhynchocephalians such as Gephyrosaurus, Diphydontosaurus and Planocephalosaurus.
Additional and postcranial material. Among the holotype material, long rod-like elements are observed beneath the skull (Fig. 10D), inferred to be remnants of the hyoid apparatus. Likewise, we also found very thin broken fragments of bone behind both orbits (Fig.  10A, C), which we suggest are fragments of the sclerotic ossicles. The axis and atlas are present (Fig.  13G-K), but the former has sustained considerable damage dorsally so that much of the neural arch is missing and only the centrum of the axis remains intact. The centrum of the atlas is fused to that of the axis, forming the odontoid process which articulates dorsally to the atlas intercentrum. The axis intercentrum is possibly preserved, but the suture lines between it and the atlas intercentrum cannot be distinguished in the scans and only the atlas intercentrum is prominent. The atlas/axis complex has a similar configuration to that in basal rhynchocephalians such as Gephyrosaurus (Evans 1981, fig. 2) and eusphenodontians like Sphenodon.

Reconstructions
The anatomy of the two Brazilian taxa has been presented in some detail based on mCT scans, and we offer standardized skull drawings of both (Fig. 14). The detail is based entirely on the fossils, and, because of their completeness, we had to make very few assumptions or restorations of missing parts. In the case of Clevosaurus brasiliensis (Fig. 14A-D), it lacks a complete vomer, as well as the stapes. In the case of Microsphenodon bonapartei (Fig. 14E-H), the posterior process of the jugal is broken off, only a small portion of the left vomer is preserved and it is missing the premaxillae and stapes. The whole-body life drawings (Fig. 15) represent the head shapes according to the fossil data, but the body shapes are speculative.

Phylogenetic results
The maximum parsimony analysis (Fig. 16A) retained four trees with lengths of 669 steps (consistency index, CI ¼ 0.297; retention index, RI ¼ 0.634), the results of which were summarized in a strict consensus tree. The Bayesian analysis output was summarized as an MRC tree, as the strict tree collapsed to form a large polytomy. The Bayesian MRC tree (Fig. 16B) generally resembles the parsimony analysis but has lower resolution within Eusphenodontia. Both analyses show good resolution of the outgroup, basal taxa, Pleurosauridae, Eilenodontinae, Opisthodontia, Eusphenodontia, Clevosauridae and Acrosphenodontia (new clade). The maximum parsimony tree also shows good resolution of Neosphenodontia, Sphenodontidae, Sphenodontinae and Sapheosauridae. Support values are fairly low for many of the nodes but are similar or higher than in previous analyses where these values were stated (e.g. Apestegu ıa et al. 2012Apestegu ıa et al. , 2014Bever & Norell 2017;Herrera-Flores et al. 2018;Romo de Vivar et al. 2020a). Most nodes have a Bremer value of at least 1, but Pleurosauridae and Clevosauridae are both supported by a Bremer value of 4, the 'Solnhofen clade' is supported by a Bremer value of 2, and several of the early diverging taxa are supported by values higher than this (Fig. 16A).
In our trees, clevosaurs form a clade that excludes Polysphenodon, in agreement with previous studies (Herrera-Flores et al. 2018;Hsiou et al. 2019;Romo de Vivar et al. 2020a). Clevosaurs are recovered as the earliest diverging eusphenodontian group (see also Hsiou et al. 2015Hsiou et al. , 2019Herrera-Flores et al. 2018;Romo de Vivar et al. 2020a;Simões et al. 2020). Among the clevosaurs, we find support for our previous proposal (Chambi-Trowell et al. 2019) that there are two morphotypes within C. hudsoni, which could indicate two separate species. Clevosaurus brasiliensis, as in previous analyses (Hsiou et al. 2015;Herrera-Flores et al. 2018), nests within Clevosaurus, but in our analyses we find it is closely related to Brachyrhinodon and C. bairdii, the three taking the earliest diverging position within Clevosauridae and forming a small subclade of their own. We found C. convallis to be the sister taxon to Sigmala sigmala, the two diverging prior to Brachyrhinodon taylori.
We recover Microsphenodon bonapartei as the earliest diverging member of Eusphenodontia after Polysphenodon.
Within Neosphenodontia, we find some results that differ from those in previous analyses. Unlike previous analyses (e.g. Bever & Norell 2017), Opisthodontia, which was defined as "all sphenodonts that are more closely related to Priosphenodon than Sphenodon" (Apestegu ıa & Novas 2003, p. 611), includes the pleurosaurs. We found the genus Opisthias to be outside Opisthodontia and closer phylogenetically to Sphenodon than to Eilenodontinae. Despite the movement of these subclades, their composition remains very similar to previous findings, and we retain the clade Opisthodontia with its original definition, even though here it excludes Opisthias.
Earlier, we also proposed the new formal name Acrosphenodontia (see clade definitions).
Our phylogenetic analysis suggests some further changes in taxon affinities compared to earlier studies: both Ankylosphenodon and Oenosaurus move out of Sphenodontidae (Herrera-Flores et al. 2018;Hsiou et al. 2019;Romo de Vivar et al. 2020a), and join the branch leading to Eilenodontinae, which differs from Simões et al. (2020), who consider the latter to be a sapheosaurid.
Uniquely, our maximum parsimony analysis also recovers a small clade of seven taxa that diverges at the base of Neosphenodontia; however, we regard this clade with some scepticism because all these taxa are represented by potentially (but not definitively) immature specimens. Though we made efforts to avoid coding characters that might be affected by ontogeny (for example, excluding characters such as skull length and limb length ratios), ontogeny might have an influence on the position of these taxa. In Sphenodon, allometric changes through ontogeny mean that juveniles possess a relatively longer antorbital region, larger orbits, and smaller adductor chambers, as in early diverging non-eusphenodontians such as Gephyrosaurus, Diphydontosaurus, Planocephalosaurus and Microsphenodon (Jones & Lappin 2009). Rebbanasaurus, Leptosaurus and Pamizinsaurus all possess an enlarged lateral premaxillary tooth, which we otherwise find to be a unique character of Sphenodontinae. But as this clade is positioned basally to Sphenodontinae, it might be a plesiomorphic trait for Neosphenodontia.

Clevosaurus brasiliensis
Our anatomical observations and phylogenetic analysis (Fig. 16) confirm that Clevosaurus brasiliensis is an early diverging clevosaur, sharing several characteristics with Brachyrhinodon taylori, and together with C. bairdii forms a small subclade at the base of Clevosauridae. Unlike later clevosaurs, both C. brasiliensis and B. taylori bear three rows of teeth on their pterygoids (SAVC-T, pers. obs.), the lateral-most of which is reduced to just two to three teeth. In addition, both taxa possess robust jaws, short snouts, broad skulls, a remarkably elongated retroarticular process, two to three large additional teeth on the maxilla, and a high coronoid process. Unfortunately, the nature of the implantation, tooth form, and numberfeatures that mark C. brasiliensis as distinct from other Clevosaurus Figure 16. Phylogeny of Rhynchocephalia, recovered from: A, strict consensus tree from the maximum parsimony analysis with Bremer values above 1 labelled beside the nodes; and B, 50% majority rule consensus tree from the Bayesian-inference analysis with clade credibility values (decimal proportions) labelled beside nodes.
cannot be verified in B. taylori as these characteristics are not preserved.
The teeth of C. brasiliensis differ from those of later clevosaurs in morphology and implantation, appearing to be acrodont but nested deep within the jaw bones so that only a small portion of the tooth is visible. Rather than forming a jagged, narrow, semi-continuous blade, posteriorly the teeth of C. brasiliensis are closely packed and fused together, and both the cusps and dentary are worn at an angle as the animal ages. This means that the anterior hatchling teeth form a continuous narrow cropping beak-like structure. Most Clevosaurus species show additional dentary teeth with a mesiodistally elongated saddle shape and the cusp asymmetrically placed posteriorly on the tooth displaying large anterolateral flanges (Chambi-Trowell et al. 2020). However, the additional teeth of C. brasiliensis are mostly conical, with only the ultimate toothwhich can be up to two to three times bigger than any of the other teethelongated mesiodistally, and with its cusp placed anteriorly rather than posteriorly (two features unique to C. brasiliensis amongst clevosaurs, but also shared with Planocephalosaurus). In contrast to other clevosaurs, the teeth on the dentary of C. brasiliensis do not appear to bear flanges at all. But despite these differences, C. brasiliensis is otherwise very similar to later clevosaurs, with a short robust snout, broad skull, supratemporal bone, premaxillary beak and an elongated retroarticular process.
Some of the skull traits of C. brasiliensis suggest that it must have been capable of a strong bite, such as its robust and short snout that seems to be further reinforced by its brace-like premaxillae and the slot-like facet on the prefrontal so that the maxilla is supported both medially and laterally. A similar slot-like facet is observed between the postorbital and jugal, with the bone interdigitating along their meeting edge rather than the simpler overlapping facets seen in later clevosaurs.

Microsphenodon bonapartei
Originally considered to be the juvenile form of Clevosaurus brasiliensis (Bonaparte & Sues 2006;Romo de Vivar Mart ınez & Soares 2015), we note the following morphological differences that indicate it represents a different genus and species: fusion of articular, surangular and prearticular; fusion of exoccipitals and basioccipital; deep diagonal wear facets on dentary; lack of an enlarged posterior-most tooth on the dentary; two rows of teeth on the palatine (albeit one is rudimentary); elongated antorbital region of the skull; broad parietal table; four-cornered postfrontal; typical acrodonty (teeth sit on the jaw crest and do not extend deeply into the jaw bones); elongated premaxillary process on maxilla; short posterior process on the parietal; relatively gracile jaw; non-interdigitating facets between jugal-postorbital and maxilla-prefrontal; and presence of (small) caniniforms. All four specimens of M. bonapartei suggest an animal with a skull around 20 mm long, which is only slightly smaller than C. brasiliensis (with specimens averaging around 25 mm).
Our phylogenetic analysis (Fig. 16) recovered Microsphenodon as the earliest diverging eusphenodontian after Polysphenodon. Plesiomorphic features include its relatively elongated snout, fusion of the exoccipitals to the basioccipital, broad flat parietal table, low gracile dentary, and its multiple rows of palatal teeth, while apomorphies include its fully acrodont dentition, a high coronoid process, differentiated dentition, and a large parietal opening. The last character-state could be interpreted as evidence the specimen is juvenile, perhaps based on the wider, flatter parietals in juvenile Sphenodon compared to adults, but some cranial sutures in Microsphenodon appear to be well fused, and the parietal opening size varies substantially among extinct rhynchocephalians, being small in Gephyrosaurus and outgroups, and large in clevosaurs and most eusphenodontians.
The palate of Microsphenodon is its key diagnostic feature, bearing multiple rows of teeth on the vomers, palatines and pterygoids. Although there are more tooth rows than in most eusphenodontians, the number of rows is still fewer than in earlier diverging forms such as Gephyrosaurus, Diphydontosaurus and Planocephalosaurus. The palatine of Microsphenodon is similar to that of Rebbanasaurus from the Early Jurassic of India (Evans et al. 2001), both of which bear two rows of palatine teeth, one reduced to just two to three teeth and positioned at roughly 45 to the main row of teeth. Medially, the palatine bears a single tooth that is otherwise an apomorphy (a single tooth or cluster of teeth placed medially on the palatine) known only within clevosaurs, suggesting this feature was plesiomorphic to Clevosaurus, though it is also arguably known in Sphenotitan from the Rhaetian of Argentina (Mart ınez et al. 2013). Whether this single tooth is present in Rebbanasaurus is unknown because this part of the palatine is missing.
The marginal teeth of Microsphenodon show some resemblances to those of derived rhynchocephalians. For example, the dentary is like that of Sphenocondor from the Middle Jurassic of Argentina (Apestegu ıa et al. 2012), with a high but elongated and blunt coronoid process, a gracile elongated dentary ramus and similar complex tooth differentiation. There are a few large, mesiodistally elongated additional teeth, and anterior to this many smaller teeth alternating in size, and successional teeth located most anteriorly, including a caniniform. Microsphenodon differs from Sphenocondor in showing no evidence of any other successional teeth anterior to the caniniform (Fig. 3C). The only known specimen of Sphenocondor has been identified as a juvenile because some of its dentary teeth alternate in size, but it is roughly the same size as all known specimens of Microsphenodon, and adult specimens of Diphydontosaurus similarly display alternation in tooth size, suggesting that this type of dentition is not restricted to juveniles. Likewise, adult specimens of Sphenodon also have additional teeth that alternate in size (Maisano 2001). Further, Microsphenodon, like Sphenocondor, also has caniniform teeth, a tooth form not known in any other sphenodontians before the Jurassic, and not outside Neosphenodontia. However, the dentary of Sphenocondor also differs from Microsphenodon in seemingly possessing a large incisiform tooth and lacking the pronounced gap between the coronoid process and teeth (see below for further explanation).
It is important to compare Microsphenodon with Lanceirosphenodon (Romo de Vivar et al. 2020a), also from the Riograndia AZ (Candel aria Sequence) of the Linha São Luiz outcrop in southern Brazil. Lanceirosphenodon resembles Microsphenodon, but there are several apomorphies that distinguish the two taxa. While the holotype and associated specimens of Microsphenodon are believed to represent adult individuals of a similar size, it is probable that Lanceirosphenodon is based on a juvenile holotype, with the distance between the symphysis to the coronoid process being $7.1 mm, making it approximately twothirds the size of Microsphenodon. The pronounced gap between the ultimate tooth and the coronoid process of Microsphenodon, seen in other early sphenodontians including Diphydontosaurus, Planocephalosaurus and Clevosaurus spp. (Fraser 1982, pl. 70 (2); Whiteside 1986, fig. 4B;Fraser 1988, fig. 19), is absent in Lanceirosphenodon. This gap does not relate to ontogeny as it is present throughout life in Clevosaurus hudsoni (Fraser 1988, fig. 23). There also appear to be two series of alternating teeth on the dentary of Lanceirosphenodon (Romo de Vivar et al. 2020a, fig.  4B), one apparently representing the hatchling teeth and the other the larger additional teeth, but there is only one such series in Microsphenodon. Lanceirosphenodon also lacks pronounced diagonal wear facets on its dentary, but this could reflect its young age. Further, though both Lanceirosphenodon and Microsphenodon share the trait of two large additional teeth most posteriorly on the dentary following a row of additional teeth that alternate in size, their teeth differ in shape. The two largest posterior teeth of Microsphenodon are mesiodistally elongated with an equidistant triangular profile, while those of Lanceirosphenodon are more conical (Romo de Vivar et al. 2020a, fig. 4). If Lanceirosphenodon is a juvenile, then it likely would have had more of these larger additional teeth as an adult. The maxilla of Clevosaurus brasiliensis possesses two large, additional teeth (Fig. 3D), so the possession of two large posterior-most teeth in the maxilla and dentary might have been a plesiomorphic trait for the rhynchocephalians of this Assemblage Zone.

Phylogenetic analysis and rhynchocephalian clade definitions
Our phylogenetic analysis (Fig. 16) confirms many broad features found in earlier studies, but also suggests some substantial rearrangements in the relationships between subclades within Neosphenodontia. The repeated discovery of key subclades within Rhynchocephalia, including Eusphenodontia, Neosphenodontia, Clevosauridae, Pleurosauridae, Opisthodontia, Eilenodontinae and Sphenodontidae, confirms some measure of stability. However, incomplete specimens and long temporal gaps pose challenges for improvement in confidence in the phylogenies.
To date there have been 10 proposed clades and subclades within Rhynchocephalia. Simões et al. (2020) reorganized clade nomenclature of rhynchocephalians, redefining several clades, but not all their definitions are compatible with our results. Below, we discuss this and their suggestion that Sphenodontia and Rhynchocephalia should be treated as synonymous. Differences in the phylogenetic topologies discovered by Simões et al. (2020) and in our analyses can be explained by differences in the character and taxon choice. When we ran our analysis with the same taxa as in Simões et al. (2020) but using our characters, we recovered a very similar topology to theirs (Supplemental Fig. 6) which is compatible with the findings below, but we did not recover Derasmosaurus as a pleurosaur. The strict maximum parsimony tree from this pruned matrix is (except for Pleurosauridae) compatible with the redefinitions they suggested. We therefore emphasise the importance of taxon selection for phylogenetic tree topology and suggest that future analyses consider all rhynchocephalians.
In the sequence of clade definitions below, we begin with the standard proposals by de Queiroz & Gauthier (2020) and , as well as some earlier definitions, all cited, and differentiate node-based and stem-based definitions.  (2020) give a very full account of the history and current nomenclature of this major diapsid clade. 2. Pan-Squamata  a. Definition: "The total clade of the crown clade Squamata" , p. 1087. This is a stem-based definition, when expressed more fully, as "Squamata and all extinct species that are more closely related to that crown clade than they are to Sphenodon punctatus" , p. 1087). b. Remarks:  give a very full account of the history and current nomenclature of crown clade Squamata, and the derivation of the name of the total clade Pan-Squamata. 3. Rhynchocephalia G€ unther, 1867 a. Definition: "the most inclusive lepidosaurian clade including Sphenodon and Gephyrosaurus, but not Iguana iguana and Gekko gecko" (Simões et al. 2020, p. 12). Stembased definition. b. Remarks: Simões et al. (2020) suggested that Sphenodontia and Rhynchocephalia should be synonyms as they found the position of Gephyrosaurus to be unstable in some of their analyses, but we find good support for Gephyrosaurus as a sister taxon to Sphenodontia in both our maximum parsimony and Bayesian analyses, and so retain the distinction between Rhynchocephalia and Sphenodontia. This stem-based clade matches Pan-Squamata, also stem-based, and together forming Lepidosauria, with a nodebased definition. 4. Sphenodontia Williston, 1925 a.  & Sues, 2006 a. Definition: "All taxa more closely related to Clevosaurus than to Sphenodon" (Hsiou et al. 2015, p. 4). Stem-based definition. b. Remarks: Clevosaurs in general have a greatly reduced number of additional teeth compared to other clades and subclades, and often have distinctive saddle-shaped additional teeth on their dentaries. They all possess supratemporal bones, but these likely existed in early noneusphenodontians, and are known in Microsphenodon. All clevosaurs possess a medially positioned isolated tooth, or cluster of teeth, on the palatine. Like earlier diverging genera, Clevosauridae (excluding C. brasiliensis) still possess a pronounced anterior alveolar foramen on their maxillae. 8. Pleurosauridae Lydekker, 1880 a. Definition: The least inclusive clade containing Vadasaurus herzogi, Palaeopleurosaurus posidoniae and Pleurosaurus goldfussi. New, node-based definition. b. Remarks: Simões et al. (2020, p. 12) redefined Pleurosauridae as "The least inclusive clade containing Palaeopleurosaurus posidoniae, Pleurosaurus goldfussi, and Derasmosaurus pietraroiae", but our analysis suggests that D. pietraroiae is not a pleurosaur. All known pleurosaurs appear to have only one premaxillary tooth per premaxilla, which in earlier diverging forms projects prominently below the jaw. All but Vadasaurus have in excess of 25 presacral vertebrae. 9. Sapheosauridae Nopcsa, 1923 a. Definition: The most inclusive clade that contains Kallimodon and Sapheosaurus. New, node-based definition. b. Remarks: Simões et al. (2020, p. 12 . Stem-based definition. b. Remarks: We note that sphenodontians within this clade lack variation in their maxillary or dentary tooth morphology other than a decrease in size anteriorly. Generally, opisthodontians lack successional teeth and tend to lack a posterior process on the jugal. The maxilla is often excluded from the naris by the lateral process of the premaxilla, and all members possess a pronounced posterior process on the ischium where this has thus far been recorded. 11. Eilenodontinae Rasmussen & Callison, 1981 a. Definition: The most inclusive clade containing Eilenodon robustus, Sphenotitan and/or Toxolophosaurus, but not Sphenodon punctatus. New, stem-based definition. b. Remarks: Rasmussen & Callison (1981) did not give a definition other than the clade should include both Eilenodon robustus and Toxolophosaurus cloudi. While the definition given by Simões et al. (2020, p. 12) is supported here ("The most inclusive clade containing Eilenodon robustus, but not Sphenodon punctatus"), we consider that Sphenotitan and Toxolophosaurus share enough similarities and are phylogenetically close enough that they should be included in this clade. 12. Sphenodontidae Cope, 1869 a. Definition: All sphenodontians that are more closely related to Sphenodon than to Priosphenodon. New, stem-based definition. b. Remarks: This is a problematic clade because it has commonly been used without a clear definition and the composition has varied greatly. However, the new definition by Simões et al. (2020, p. 12), as "The most inclusive clade containing Eilenodon robustus and Sphenodon punctatus, but not Kallimodon pulchellus or Saphaeosaurus thiollerei" is not compatible with our analysis. 13. Sphenodontinae Nopcsa, 1928 a. Definition: The most inclusive clade containing both Sphenodon punctatus and Sphenofontis velseraei. New, node-based definition.

Conclusions
We describe a new genus and species of eusphenodontian rhynchocephalian here, named Microsphenodon bonapartei, based on two remarkably well-preserved skulls and isolated cranial material, much of it previously referred to Clevosaurus brasiliensis. This new taxon shares a mosaic of features with the earliest rhynchocephalians and eusphenodontians, presenting the most complete articulated cranial remains of an early diverging sphenodontian from before the Rhaetian. We also establish a new clade name, Acrosphenodontia. In addition, we examine several remarkably well-preserved skulls of Clevosaurus brasiliensis, identifying new apomorphic features, including a unique modified form of acrodonty whereby the teeth are placed deep within the jaws in a continuous groovefuture work will investigate the nature of tooth implantation in early Rhynchocephalia. The presence of three rhynchocephalians in the Norian of southern Brazil makes this locality uniquely informative in the understanding the early evolution of this group and of great significance in the of study small sized faunal components of the Late Triassic.