Unique dentition of rhynchosaurs and their two‐phase success as herbivores in the Triassic

Rhynchosaurs were key herbivores over much of the world in the Middle and Late Triassic, often dominating their faunas ecologically, and much of their success may relate to their dentition. They show the unique ankylothecodont mode of tooth implantation, with deep roots embedded in the bone of the jaw and low crowns that were rapidly worn down in use. During growth, the main area of oral food processing, located in the middle and posterior portions of the occlusal surfaces of the jaws, moved posteriorly relative to the anterior tips of the jaws, which curved up. As the maxilla and dentary grew by addition of new bone posteriorly, the dental lamina fed in new teeth at the back of the tooth rows. CT scanning of the holotype skull of Bentonyx sidensis from the Middle Triassic of England reveals previously concealed details of the dentition. Together with new dentary material from the same location, this has enabled us to examine the tooth replacement process and elucidate ontogenetic changes in dentition and jaw morphology as the animals aged. There were major changes in rhynchosaur anatomy and function through their evolutionary history, with the early forms of the Middle Triassic dying out before or during the Carnian Pluvial Episode (233–232 Ma), and the subclade Hyperodapedontinae, with broad skulls and adaptations to chop tough vegetation, subsequently diversifying worldwide in a successful ecological expansion until their global extinction 227–225 Ma.

R H Y N C H O S A U R S were widespread archosauromorph reptiles throughout much of the Triassic, and their success as ecologically significant herbivores has been explained by their dentition (Chatterjee 1974;Benton 1984;Langer et al. 2000;Ezcurra et al. 2016). The rhynchosaur dentition appears to be unique among amniotes, comprising long-rooted teeth implanted firmly within the bone and termed ankylothecodont (Chatterjee 1974). There is no sign of gomphosis within a socket through a periodontal ligament, nor of ankylosis by means of alveolar bone, at least in teeth fully enclosed in bone. Indeed, in all microscopic cross-sections (Benton 1984;Langer et al. 2000;Scartezini & Soares 2022), there is no gap between tooth and bone. In recent overviews of amniote dentitions (e.g. Berkovitz & Shellis 2017;Bertin et al. 2018;LeBlanc et al. 2021;Salomies et al. 2021), the ankylothecodont mode of tooth implantation has not been mentioned, and, although it shares some characteristics with the acrodont mode, the deep roots and other features show that it is different.
In addition to ankylothecodonty, the rhynchosaur dentition comprises multiple tooth rows on maxilla and dentary, and the narrow, tooth-lined, dorsal margin of the mandible cut into a deeply grooved upper jaw ( Fig. 1). This provided a precision-shear bite without the possibility of any lateral or anteroposterior sawing movement, somewhat like the action of the blades of a pair of cutting shears (Huxley 1869;Chatterjee 1974;Benton 1983aBenton , 1984Benton , 1990. This unique dental arrangement evolved in stages (Ezcurra et al. 2016;Langer et al. 2017), starting simply as opposing dental plates with multiple tooth rows in early forms such as Howesia, then acquiring one or two grooves on the maxilla and one or two ridges on the dentary in Rhynchosaurus, Stenaulorhynchus and other late Anisian forms, before stabilizing finally with a single deep maxillary groove and high-sided single major dentary tooth row in the Carnian-aged Hyperodapedontinae (although these taxa show considerable variation in number of tooth rows) (Fig. 1).
A key feature of rhynchosaurs was that the occlusal surfaces of the maxilla and dentary curved. As the animal grew and new bone was laid down, the tooth-bearing regions of the maxilla and dentary were supplied with new teeth from dental laminae located posteriorly and lingually. Teeth and jaw bones acted equally as cutting surfaces, being worn down but retaining their intimate, tightly occluding contact throughout life by constant forward growth. As the animal increased in size and new bone and teeth were produced posteriorly and lingually, the curved occlusal area on maxilla and dentary expanded and moved backwards, the anterior part of those elements curving up and out of the occlusal area (Chatterjee 1974;Benton 1983a;Langer et al. 2000). The anterior maxilla curved up behind the beak-like paired premaxillae, and F I G . 1 . Key adaptations and evolution of rhynchosaurs. Time tree of rhynchosaur phylogeny, highlighting the two phases in rhynchosaur macroevolution, before and after the Carnian Pluvial Episode. Generalized skull outlines of Stenaulorhynchus (lower, dorsal view) and Hyperodapedon (upper left and right, lateral and dorsal views) and three steps in dental tooth plate evolution (darker shading) are shown. Skull images are from Benton (1984) and Benton & Kirkpatrick (1989), and jaw cross sections from Langer et al. (2000). Abbreviations: d, dentary; mx, maxilla; pmx, premaxilla. the anterior dentary curved up and laterally on either side of these elements (Fig. 1, top left). Hence the famous rhynchosaur 'grin' which became deeper and more exaggerated as the animal aged.
Rhynchosaurs arose in the Early Triassic when terrestrial ecosystems were rebuilding following the devastating end-Permian mass extinction, 252 Ma (Benton et al. 2004;Chen & Benton 2012). The ecological dominance by rhynchosaurs as herbivores in many communities is illustrated by the fact that they commonly represent 50-90% of collected tetrapod specimens in samples of hundreds of identified remains (Benton 1983b;Benton et al. 2018). New evidence suggests that rhynchosaurs underwent two separate diversifications (Fig. 1), first in the Middle Triassic, especially the Anisian , and then in the early Late Triassic (late Carnian, 232-227 Ma), each associated with different dental adaptations (Langer et al. 2000;Ezcurra et al. 2016). The first phase corresponded to tooth plates with multiple tooth rows and complex occluding grooves and ridges between the toothbearing elements, the second to a more ordered single ridge/groove jaw system with more orderly tooth organization, as seen in multiple species of the genus Hyperodapedon.
We have two aims in this paper: (1) to explore and define the features of rhynchosaurian ankylothecodonty, comparing it with acrodonty in some modern and fossil clades; and (2) to explore the two-phase diversification of rhynchosaurs and the associated dental adaptations. We provide new CT data from the British Anisian rhynchosaur Bentonyx, to supplement earlier descriptions of tooth implantation and function in Stenaulorhynchus (Benton 1984)

Specimens
We included three specimens in the study, a complete skull of the holotype specimen of Bentonyx sidensis (BRSUG 27200), as well as two isolated mandibles, one from a probable subadult (BRSUG OSF 374), and one from a probably very old individual (BRSUG OSF 327). We cannot estimate the specific ontogenetic ages of these specimens because they are incomplete, lacking suitable limb bones or ribs for comparisons of numbers of lines of arrested growth with overall body size.
These specimens were collected from near the top of the Otter Sandstone of Sidmouth, Devon, UK, a unit that is age-equivalent to the Helsby Sandstone Formation, a division of the Sherwood Sandstone Group (Ambrose et al. 2014). The Otter Sandstone comprises fluvial sandstones deposited under a semi-arid climate (Coram et al. 2019(Coram et al. , 2021 and is dated as late Anisian (Middle Triassic; c. 242 Ma) from regional geological correlations, fossils and magnetostratigraphy (Hounslow & McIntosh 2003).
Two rhynchosaur taxa have been reported from the Otter Sandstone, Fodonyx spenceri (Hone & Benton 2008) and Bentonyx sidensis (Langer et al. 2010). There is uncertainty about the postcranial skeleton of both taxa, and other cranial characters that differentiate Bentonyx from Fodonyx are in the braincase (Langer et al. 2010). The two isolated jaw bones (BRSUG OSF 327, 374) cannot be assigned confidently to either taxon, but we are comfortable to include them as comparative material because anatomical, including dental, differences between Bentonyx and Fodonyx are not extreme and adult skull lengths are similar (Langer et al. 2010;Ezcurra et al. 2016).

Scanning and segmentation
The specimens were scanned with a Nikon XTH 225 ST x-ray computed tomography (CT) scanner belonging to the Palaeobiology Research Group in the School of Earth Sciences, University of Bristol. The adult skull (BRSUG 27200) was scanned at 144 kV which produced a dataset containing 2809 slices at a voxel resolution of 50.68 lm. The two isolated jaws, BRSUG OSF 327 and 374 were scanned in two parts at 224 kV which produced a dataset containing a total of 2940 slices each at a voxel resolution of 25.59 lm.
Segmentation of skeletal elements from the CT datasets was performed using Avizo 2020.2 (Visualization Sciences Group) to generate three-dimensional models. We segmented bone from matrix, and we isolated specific elements digitally, especially discriminating the teeth within the dentary and maxillary bones. All surface models were generated in Avizo with an unconstrained smoothing factor of 2, and then exported in an ASCII Standard Triangle Language (STL) format. These files were then imported into Blender v2.93.3 (Blender Foundation) to acquire photographs and manipulate the models. Measurements were also taken in Blender, and the unit scaling was changed to 0.00137 (BRSUG 27200), 0.000996 (BRSUG OSF 327) and 0.001002 (BRSUG OSF 374) to calibrate for accurate measurements of the 3D model. The scan data and models are available in MorphoSource (Sethapanichsakul et al. 2023a).

RESULTS
The holotype of Bentonyx sidensis (BRSUG 27200), a near-complete adult skull, was described by Hone & Benton (2008) based on external evidence from the carefully prepared skull. The specimen had been described by them as an example of Fodonyx spenceri but was discriminated as a new genus and new species, Bentonyx sidensis, by Langer et al. (2010). Additional details of the palate and braincase are provided by Sethapanichsakul et al. (2023b) based on CT scanning; these authors also comment on practical issues with processing the scan data, in particular the heterogeneity of the matrix and calcitic overgrowths on many bones.
The 3D image of the skull ( Fig. 2A; Sethapanichsakul et al. 2023b, figs 1-4) shows the damaged temporal region posteriorly, the large, circular orbit, and the long curving premaxillae anteriorly. The maxillae wrap around the premaxillae, anchoring them firmly, and form the lateral surface of the snout, extending posteriorly to meet the anterior processes of the jugals beneath the orbit. The segmented dentition shows 14 teeth along the margin of the maxillary tooth plate ( Fig. 2A, C), the anteriormost three being rather narrow and short, then six wider and deeper teeth that are worn flush with the occlusal surface of the maxilla, and a posterior five teeth that still bear some or all of their pointed crowns.
Likewise, the dentaries are well preserved on both sides of the skull. They exhibit many of the expected rhynchosaurian characteristics, their anterior tips curving upwards and laterally at the beak ( Fig. 2A, B). The multiple tooth rows are best seen in medial (lingual) view (Fig. 2B). The portion of the dentary ventral to the tooth rows, and its inclusion of numerous vessel and nerve openings in more ventral portions, are typical of all rhynchosaurs. The posterior elements of the lower jaws have not been preserved on either side.

Dentition of Bentonyx holotype
The CT scans enable us to reconstruct the entire dentition of the dentary and maxilla. In life position with the jaws closed ( Fig. 2A), the occlusal margin of the dentary and all the occluding and earlier teeth are masked by the descending lateral margin of the maxillary tooth plate. The right maxilla ( Fig. 2A, C-E) has three rows of teeth and a fourth inserting in the posterior region between the lingualmost row and the middle row (Fig. 2D). In total the right maxilla has 41 teeth, with 14 in the lateral row, but is less complete than the left maxilla so is not the focus of description. The left maxilla has only two rows of teeth with a third forming in the posterior region of the tooth plate. It has only 25 teeth, far fewer than the 41 in the right maxilla, and it lacks the central tooth row seen in the right-hand maxillary tooth plate. The lateral longitudinal tooth row on the right dentary ( Fig. 2A, F-H) is well defined, but the three additional lingual longitudinal rows are less distinct. There are 14 teeth in the lateral row, matching the maxilla, and 28 in total in the lingual rows, for a total of 42. The left dentary also includes four rows of teeth for a total of 40. In occlusal view, the maxilla shows one tooth row laterally of the sulcus, and three medially (Fig. 2E).
The teeth in maxilla and dentary are deeply rooted ( Fig. 2A), as in other rhynchosaurs (Chatterjee 1974;Benton 1984;Langer et al. 2000). Individual teeth are generally three times as deep as wide, the largest measuring 12 mm deep and 4 mm wide, and somewhat cylindrical in shape, with subcircular cross sections. They stand roughly vertical, and pack into the bone of maxilla and dentary leaving a thin wall of bone between adjoining teeth. In the maxilla ( Fig. 2A-E), the posterior five teeth in the lateral row have erupted most recently, and they show conical, pointed occlusal tips, in which the dentine, forming the bulk of the material of the tooth, carries a thin enamel cap restricted just to the portion that appears above the level of the maxillary bone ( Fig. 3D, de, en). Anterior to the sixth posteriormost tooth in the lateral maxillary tooth row, the teeth come into occlusion in the area of mastication and are worn flat with the bone, losing the pointed crown and the enamel cap. Moving anteriorly, the teeth become smaller, presumably reflecting the fact they had developed when the animal was younger and smaller, and had been largely worn away. The three most anterior teeth of the lateral maxillary tooth row ( Fig. 2A-E) are separated a little from the others, are much narrower, and spaced more widely in the bone. Teeth in the more medial (more lingual) rows are smaller and restricted to the back of the jaw. It appears that the lateral row was populated with teeth in the younger animal, then the medialmost row began to be implanted later in life, and the third and fourth rows subsequently (Fig. 2E). Each new row presumably reflects an overall broadening of the maxillary tooth plate as the animal grew and aged, thereby providing empty space into which teeth could be implanted, but required also by the fact that, with aging, the occlusal area was located more posteriorly and became broader as the skull expanded overall, and so additional tooth rows could be accommodated.
The dentary teeth ( Fig. 2A, F-H) show similar features to those of the maxilla, the largest being 10 mm deep and 3 mm wide. In occlusal view, the dentary shows one lateral tooth row along the crest of the jaw, and up to three lingual rows medially (Fig. 2H). Again, the posteriormost four or five teeth in the lateral row are unworn, and the more anterior teeth are worn down to bone level, losing their pointed crowns. In the dentary, the additional F I G . 2 . Dentition of the holotype skull of Bentonyx sidensis (BRSUG 27200). A, right lateral view with highlighted dentition placed within a translucent skull model. B, medial view of the left lower jaw. C-E, isolated segmented 3D models of the teeth of the right maxilla in: C, lateral; D, medial; E, occlusal view. F-H, isolated segmented 3D models of the teeth of the right dentary in: F, lateral; G, medial; H, occlusal view. Abbreviations: ANT, anterior; lat, lateral; lin, lingual; o, orbit. Scale bars represent 10 mm. lingual tooth rows are packed more tightly against each other than in the maxilla.
In x-rays of the tooth-bearing elements ( Fig. 3), these details are exemplified. The jaws are preserved tightly shut, but the lower jaw has slipped slightly forward, so the fit between maxillary tooth plate and dentary is not perfect when viewed in transverse section (Fig. 3A). Nonetheless, this view shows how bone and teeth are worn equally and to a smooth surface, and in proper occlusion the fit between these tooth-bearing elements would be precise. Tooth-to-tooth occlusion occurs, but most occlusion is in fact between bone or bone and tooth ( Fig. 3A, C, D). In horizontal section through the tooth-bearing elements (Fig. 3B), the height differences across the occlusal faces are clear, with the lateral and medial maxillary tooth rows descending well below the crest of the dentary. This view shows the irregularity of spacing and orientation of the teeth, especially when compared with Hyperodapedon (Benton 1984, fig. 5). In Bentonyx, the teeth appear relatively evenly spaced on the worn occlusal surface (Fig. 2E, H), but deeper in the bone, they are more irregular, some nearly touching, and in places leaving spacing bone of variable thickness between adjacent teeth.
Sagittal-plane views through the posterior portions of maxilla and dentary (Fig. 3C, D) show how newly implanted teeth approach the crests of the jaws. These new teeth, located posteriorly, are shorter than those in occlusion, they bear obvious thin caps of enamel that is quickly removed as they come into occlusion, and there are cavities at the bases of the pulp cavities. The enamel caps are smaller in the maxillary teeth, little more than a C-shaped cap on the tooth, whereas the enamel caps on the dentary teeth descend further down, being V-shaped ( Fig. 3D, de, en). Whereas a pulp cavity was retained by occluding teeth in Hyperodapedon (Benton 1984, fig. 2), this rather small, basal pulp cavity in Bentonyx filled with dentine and was largely closed off, leaving only a very small pulp cavity (Fig. 3D, pc), soon after the teeth began to occlude.

Dentition of subadult
The presumed subadult specimen (BRSUG OSF 374) is a right dentary that has preserved the entirety of the tooth plate ( Fig. 4A-D). The specimen lacks the anterior tip of the dentary beyond the anterior margin of the tooth plate F I G . 3 . Individual x-rays from the CT scan of the holotype skull of Bentonyx sidensis (BRSUG 27200). A, coronal plane across maxilla (above) and dentary, showing teeth (white) and bone (darker grey). B, horizontal section through occluding maxilla (outer rows) and dentary (middle row), showing teeth and bone. C-D, two views of a vertical, sagittal-plane section through occluding maxilla (above) and dentary. Abbreviations: ANT, anterior; de, dentine; en, enamel; lat, lateral; med, medial; pc, pulp cavity. Scale bars represent 5 mm. and lacks parts of the posterior and posteroventral processes. The dentary exhibits four rows of teeth and 26 teeth in total, fewer than in the holotype adult. The lingualmost tooth row comprises only four teeth and five empty sockets (Fig. 4D), which are presumably resorption pits.
Even though the dentary shows all but the tiniest of most anterior teeth, and most in front of the vertical crack through the specimen show wear, that wear has not been intense and so most of the teeth in the occlusal area stand out from the bone surface. Wear makes the occlusal tip of each tooth somewhat chisel shaped. In longitudinal x-ray cross sections from lateral to medial sides ( Fig. 4A-C), the teeth are smaller and relatively shorter than in the adult (cf. Fig. 2), each tooth being at most 3 mm deep and 1.5 mm across. The multiple rows of teeth in section may look like successional teeth, but these are in fact cross sections through lingual teeth that are implanted at lower levels than the teeth on the jaw crest. Sequential filling of the pulp cavities by dentine varies according to position along the tooth row. A posteriorly located unworn tooth in the highest lingual tooth row is a largely hollow cone, a relatively thin dentine structure around a large pulp cavity (Fig. 4B). More anteriorly, occluding teeth along the jaw crest retain small pulp cavities at their bases (Fig. 4A), but occluding, more anteriorly placed teeth in the lingual tooth rows show entirely filled pulp cavities (Fig. 4B, C).

Dentition of older rhynchosaur
The left dentary specimen from what we interpret as an old individual (BRSUG OSF 327) preserves the entirety of the dentition (Fig. 4E-G), as well as most of the dentary itself, the anterior process of the coronoid behind the tooth row, and most of the splenial on the medial face. The dentary exhibits three tooth rows although there are indications of an incoming fourth row indicated by a single anteroventrally located tooth that does not align with the other rows. In total there are 21 teeth. Interestingly, this very large and presumably very old specimen has fewer teeth than younger examples, and they all appear to be relatively small, 4 mm deep and 2 mm wide along the crest of the jaw, and much shorter in the lingual rows. The most posterior of the teeth on the crest of the jaw, and three posteriorly located lingual teeth appear unworn, but the remainder were in occlusion at the time of death, but these occupy only the posterior 20% of the length of the dorsal margin of the dentary.
The small numbers of teeth and of tooth rows in such a large, and presumably aged individual are hard to explain other than by individual variation or, possibly, that teeth and tooth rows have been worn away. This conclusion is supported by the entire absence of teeth in the anterior two-thirds of the potentially toothed portion of the dentary. The depth of the area of occlusal wear is as much as 8 mm when measured from a curving chord marking a line from the short row of posterior teeth to the anterior dentary. In other words, the occlusal portion of the dentary has been worn down by as much as 8 mm, representing about 30% of the depth of the dentary at this point. Evidence for the extreme wear is that the crest of the jaw is stepped down in front of these few posteriorly located teeth, possibly partly highlighted by breakage, but generally the dorsal margin is original, confirming that the occlusal wear area had descended substantially, forming a deep crescent-shaped incision in old age.

Comparisons
We compare our findings about the dentition of Bentonyx with published accounts of the dentitions of other rhynchosaurs. We consider these in stratigraphic and phylogenetic sequence, first the early diverging taxa, then the coeval and closely related taxa, and then the derived genus Hyperodapedon. We could not include comparisons with all rhynchosaurian taxa because many either lack dentition, or detailed anatomical accounts of the dentitions have not been published.
Among the early diverging rhynchosaurs (Fig. 1), Noteosuchus lacks a skull, but Mesosuchus and Howesia preserve their dentitions, and they are rather different (Malan 1963). Mesosuchus bears a single row of teeth on maxilla and dentary, each row arranged in a zig-zag manner from anterior to posterior and, unusually, with two small teeth on the premaxilla. The premaxillary and anterior maxillary teeth are conical and with enamel caps, but the more posterior teeth have worn down, losing the enamel cap, but not abrading entirely flush with the bone (Malan 1963, p. 218). Dilkes (1998, p. 511) prepared the maxillary dentition and observed that there was 'a clear line of separation between the tooth and surrounding bone, thus indicating that the teeth are implanted in the jaws', rather than being ankylosed as earlier authors had said. The teeth are implanted deep within the bone, 'but it is uncertain if the mode of attachment should be described as thecodont'. Both authors agree that the zigzag tooth row can be interpreted as representing two tooth rows on the maxilla at least, and that anteriormost and posteriormost teeth are unworn, with teeth implanted posteriorly on each row. Dilkes (1998, p. 512) noted that in both maxilla and dentary, 'continuous lingual growth relocates teeth from the lingual side into occlusion and later into a more labial orientation', and pointed out similarities to the development of dentition in later rhynchosaurs.
The dentition of Howesia (Malan 1963, pp. 215-218) is more evidently comparable with that of later occurring rhynchosaurs in that both maxilla and dentary bear multiple rows of teeth, but the premaxilla is unknown so it cannot be determined whether there were premaxillary teeth or not. Maxilla and dentary lack the dentary crest and maxillary groove or grooves and, although the teeth vary in size, there is no distinction between a labial (lateral) major row and smaller teeth in lingual rows. Indeed, the teeth are not so evidently ordered in anteroposterior rows as in later taxa. Further, the occlusal and lingual faces of the maxillary tooth plate in Howesia are a continuous surface without sharp demarcation and there are two rows of small lingual teeth. Posterior teeth are unworn, confirming they had just been implanted and had not yet come into occlusion. In transverse section (Malan 1963, figs 3, 4), the teeth appear to be deeply rooted within the bone; in a broken surface she noted that one tooth was deeply embedded in the bone, and another had fallen out leaving an empty 'socket'. In discussing the dentition, Malan (1963, pp. 217-218) noted that the teeth 'are by no means thecodont' and compared them with those of Sphenodon in which 'a thick layer of secondary bone is laid down around the bases of the teeth after they have become ankylosed to the jaw.' Among Middle Triassic rhynchosaurs, the most immediate comparison is with Fodonyx spenceri, also from the Otter Sandstone. In discriminating Fodonyx and Bentonyx, Langer et al. (2010) noted anatomical differences in the braincase and in the overall posterior width of the skull, but they did not mention any differences in the dentition. Although the holotype skulls of both taxa are about the same length (140 mm), they differ greatly in posterior width; 130 mm in Bentonyx and nearly 170 mm in Fodonyx. Langer et al. (2010) regarded both as adults, and suggested that Bentonyx was not a juvenile of Fodonyx. At that time, very little detail of the Bentonyx dentition was visible because the jaws were clamped shut and CT scanning had not been done, but we can now compare the dentitions of the two taxa. In fact, the occlusal areas of maxilla and dentary do not differ much in size or shape (cf. Fodonyx holotype, EXEMS 60/1985.292; Benton 1990, fig. 29a; Langer et al. 2010, fig. 3). Even the mediolateral angling of the maxillary tooth plates is similar despite the great differences in posterior skull widths. The maxillae of Fodonyx each bear three rows of teeth, the third inserting between the lateral and medial rows, and each bearing two or three unworn teeth at the back. The dentary (Benton 1990, fig. 31) likewise bears three tooth rows, the lateral and medial being most continuous, and the third consisting of sporadic teeth inserted between those two rows. Tooth shape and depth, so far as can be seen on broken surfaces, is similar to Bentonyx.
Stenaulorhynchus from the Manda Formation of Tanzania, also probably Anisian in age, was larger, with a skull about 250 mm long and 160 mm wide (Huene 1938). Juveniles show distinctive grooves and crests on maxilla and dentary, with teeth aligned on three raised ridges on the maxilla, separated by two grooves, into which the dentary, with two dental crests, occluded. In older specimens, the grooves and crests became less pronounced, and the matching of tooth positions to the crests became less clear. As in Bentonyx, the maxilla bears two dominant tooth rows, a medial and a lateral (lingual and buccal), and the third row inserts with overall increase in size in the posterior part of the tooth plate in a less regular fashion between these two. The dentary shows a dominant buccal ridge bearing teeth at the lateral side, and numerous rows of lingual teeth, seemingly many more rows than the three seen in Bentonyx. In dental cross sections, the teeth are like those of Bentonyx, generally circular and pencil-shaped, sometimes with slightly open pulp cavities near the base (Benton 1984, pp. 752-763). As in Bentonyx, what look like successional replacement teeth in particular sections are simply roots of nearby teeth that grow close to, and sometimes even interfere with the neighbouring roots. In the Stenaulorhynchus dentary, unlike in Bentonyx, there is a broad gap between the buccal (lateral) tooth row and the lingual tooth cluster; in Stenaulorhynchus the lingual tooth rows appear to be less defined, and lingual (medial) teeth cram in as tightly as they can, forming a slightly disordered pavement. In transverse sections (Benton 1984, fig. 9a), there may be as many as four apparent rows of teeth, but neighbouring transverse sections may show just two or three, and it is hard to differentiate definite longitudinal rows of the lingual teeth. Whereas nearly all teeth in Bentonyx are more or less vertical, in Stenaulorhynchus tooth roots can be nearly horizontal at their bases, as the teeth curve up to the occlusal surface, and sometimes with interference and resorption between overlapping tooth roots (Benton 1984, fig. 12h). This sort of disorder and mutual interferenceresorption is not seen in Bentonyx, nor in Hyperodapedon.
As noted earlier, the dentition of Hyperodapedon is more orderly, with a single midline groove in the maxillary tooth plate and roughly equal medial and lateral occlusal surfaces each with 2-4 distinct tooth rows (Benton 1984, pp. 742-752). The dentary is also more regular in shape and arrangement of the teeth, with a single distinct buccal tooth row that fits directly into the maxillary groove, and a single row of lingual teeth located quite low on the medial face of the dentary. Each dentary tooth row, buccal and lingual, comprises relatively equal-sized teeth apparently budded at the back of the jaw in a regular fashion, and with very little insertion of teeth above or below the main row, as seen in Bentonyx and Stenaulorhynchus. Finally, individual teeth appear to have deeper roots than in the Middle Triassic forms and the pulp canal may have remained open for longer, so enabling addition of dentine within the pulp cavity while the tooth was in occlusion.

Ankylothecodonty and acrodonty
When Chatterjee (1974, p. 230) named the ankylothecodont mode of tooth implantation in rhynchosaurs, he specified that 'The roots of all the teeth are firmly ankylosed by bone of attachment to well-developed sockets.' However, sockets cannot be identified in his cross sections of teeth in the jaws, nor have sockets been seen in subsequent studies of different rhynchosaurian taxa (Benton 1983a(Benton , 1984Langer et al. 2000;Scartezini & Soares 2022). Chatterjee (1974, p. 230) also describes spongy bone of attachment (alveolar bone) surrounding all the teeth, which is shown in his drawings (Chatterjee 1974, fig. 9), but his photographs (Chatterjee 1974, fig. 10c, d) are at too small a scale to show this, and indeed subsequent studies have not revealed any distinct bone of attachment in other rhynchosaurian taxa (Benton 1983a(Benton , 1984Langer et al. 2000;Scartezini & Soares 2022). However, we cannot exclude the possible occurrence of alveolar bone at the point of initial implantation of the teeth when absorption pits appear on the face of the bone (Fig. 4D), and the teeth enter the bone, and are then encompassed by new bone growth.
Superficially, the rhynchosaurian dentition shares some features with the acrodonty of modern rhynchocephalians and acrodont squamates (Cooper & Poole 1973;Bertin et al. 2018;Haridy 2018;Haridy et al. 2018): the tooth is fixed in the jaw without a socket, there is no evidence of tooth replacement, and teeth are added posteriorly and lingually with new bone as the tooth-bearing elements enlarge with growth. In modern acrodont lizards, such as chameleons and the agamids Uromastyx and Pogona, as well as the rhynchocephalian Sphenodon, the acrodont teeth attach to the crest of the jaw as new bone is produced and they remain in position with respect to the bone of the jaw throughout life (Cooper & Poole 1973;Haridy 2018;Haridy et al. 2018).
There are major differences, however, between such acrodont dentitions and the ankylothecodont mode of implantation in rhynchosaurs. Rhynchosaur teeth do not sit on top of the jaw without roots, but are deep-rooted, nor do they show thickened enamel, which is an adaptation seen in modern acrodont lizards to limit the rate of wear, but the pulp cavity in both implantation modes does fill with dentine (Haridy 2018). The only adaptation in the teeth of rhynchosaurs that might have slowed the rate of wear is that several hyperodapedontines had serrations and flutes on their tooth crowns (Scartezini & Soares 2022), although such features were presumably obliterated shortly after the teeth came into occlusion.
Rhynchosaur dentitions are sometimes compared to those of the Late Carboniferous and Early Permian captorhinids, which also showed multiple tooth rows and lingual-posterior addition of teeth during growth. However, captorhinid tooth implantation has been identified as subthecodont, the teeth being implanted into shallow sockets and ankylosed to the jaw through vascularized alveolar bone that is retained throughout their lives (LeBlanc & Reisz 2015). At least one captorhinid, Opisthodontosaurus, is, however, acrodont, the oldest example of this tooth implantation mode (Haridy et al. 2018), with teeth attached to the apex of the jawbone and without sockets, but with bone of attachment. Opisthodontosaurus lost teeth as it aged because individual teeth became larger, and teeth were replaced directly from a dental lamina situated lingually of the teeth whereby new teeth resorbed tissues of the functioning teeth and attached to the jaw crest after the functioning tooth was shed. We find no evidence for this kind of replacement in rhynchosaurs.
Therefore, we confirm that the ankylothecodont mode of implantation seen in rhynchosaurs is distinct from other named modes of tooth implantation, and particularly from acrodonty, perhaps the mode it most resembles. The key differences are in the deep roots, general absence of bone of attachment in established, functioning teeth, and short span of existence of the crowns of ankylothecodont teeth after they come into wear. Also, unlike other deeply rooted teeth, such as thecodont or pleurodont teeth, there are no persistent sockets or dental grooves.

No tooth replacement
We argue here that rhynchosaurs did not show successional teeth. Sections through rhynchosaur jaws may show as many as five or six teeth in line with the roots of emergent teeth, but, as shown by Benton (1984) for Stenaulorhynchus and Hyperodapedon, these are sections through the deep roots of neighbouring teeth whose long axes are at an angle to the plane of section. In Bentonyx, new teeth were implanted only at the posterior ends of tooth rows. Further, after a tooth has been worn flat with the bone, it can only be worn down to nothing and there is no replacement.
Rhynchosaur teeth were formed from the dental lamina at the posterior and lingual ends of tooth rows (Fig. 5). This is shown by the five shallow resorption pits in the juvenile mandible (BRSUG OSF 374; Fig. 4D); these presumably delimit the position of the dental lamina because they are located along the posterior margin of the toothbearing area and represent the newest teeth in each row. Teeth were presumably about to be implanted into the bone and hollows in the bone (incipient tooth sockets) had been eroded but the teeth had failed to become ankylosed at the point of death. The fact that all five tooth rows show the same pits indicates that the dental lamina was feeding new teeth into all five rows at the same time. Similarly, Chatterjee (1974, fig. 14b) illustrated resorption pits posteriorly on a small maxilla of Hyperodapedon and Benton (1984, p. 766) noted smooth pits in the bone at the posterior limits of some tooth rows in Stenaulorhynchus. As the tooth was pushed against the bone, it resorbed bone material until the new tooth was encompassed by new bone growth. If this process was halted before completion, the nascent tooth presumably fell away, leaving the resorption pit, as also seen in captorhinids (Bolt & DeMar 1975;Ricql es & Bolt 1983;LeBlanc & Reisz 2015;Haridy et al. 2018).
Resorption between growing teeth and bone is common in the Middle Triassic rhynchosaurs such as Stenaulorhynchus (Benton 1984) and Bentonyx, but less so in Late Triassic Hyperodapedon where tooth implantation and growth were more regular. In Stenaulorhynchus, Benton (1984, pp. 760-763) noted cases where neighbouring teeth interfered with each other and one caused resorption of dentine along an arcuate front, even through to the pulp cavity, and even some cases where one tooth prevented another from reaching the jaw margin. In other cases (e.g. Benton 1984, fig. 12c, h; pl. 68, fig. 3), the crown of one tooth, generally the 'newer' tooth posterior to another, passes through the root of an occluding tooth and causes loss of dentine.
New teeth are relatively constant in width and length and are mostly circular in cross section, but their arrangement in the maxilla and dentary of Bentonyx is somewhat irregular within any identifiable row, sometimes lying a little to left or right of a straight line, suggesting haphazard implantation of teeth. The main rows on dentary and maxilla are more regular in tooth spacing and alignment than the lingual tooth rows. Hatchling and very young rhynchosaurs presumably had no lingual tooth rows, and then the first lingual teeth were implanted when the jaw reached a certain width and space was available. With more growth, third and fourth tooth rows began when sufficient space was available between the existing rows. The spacing of teeth in lingual tooth rows can be much looser than in the labial tooth row, for example as seen in our subadult (Fig. 4D).

Ontogeny of dentition and occlusion
We confirm earlier work on rhynchosaur ontogeny (Benton & Kirkpatrick 1989;Langer et al. 2000) that as the animals aged, the numbers of longitudinal tooth rows on maxilla and dentary increased, although this might have reversed later in life (see below). As body size increased with growth, the dentary and maxilla became mediolaterally wider in their posterior portions, and supplementary tooth rows were initiated, generally medially (lingually) and lower than the crest tooth row in the dentary, and potentially on both sides in the maxilla.
The contrast from small to medium-sized to large jaws of the rhynchosaur Bentonyx, which we interpret as belonging to subadult, adult and older individuals respectively, is dramatic (Fig. 5). The three specimens described here reveal a probable ontogenetic sequence from a slightly concave-down occlusal area with most teeth only slightly worn in the subadult (Figs 4A-D, 5A), through a stage of deeper curvature and thorough flattening of teeth to lie flush with the bone surface in the adult (Figs 3, 5B), and eventually deep curvature of the occlusal area and progressive loss of functionality of the teeth, in the dentary at least, in the older adult (Figs 4E-G, 5C).
Juveniles had tiny teeth, matching their size on hatching (Benton & Kirkpatrick 1989) and these sometimes survived at the very front of the adult maxillae and dentaries, perhaps because of relatively modest wear, as seen in our subadult specimen (Fig. 4A-D). Tooth width then doubled in a single increment at some early age and remained relatively constant through adulthood (Fig. 2) but unusually, in our aged specimen at least (Fig. 4E-G), tooth size is rather smaller. Whether this size reduction with great age occurs in all examples of Bentonyx, or in other rhynchosaurs, is not known, but if it were general, it could indicate some lower investment of materials through physiology in generating new teeth as the animal becomes very old.
The intensity of tooth wear apparently increases with age. In the subadult (Fig. 4A-D), the teeth in occlusion in the area of oral processing were not worn flush with F I G . 5 . Comparison of the three anterior dentary specimens, showing a possible growth series and relative movements of the zones of bone growth and tooth emplacement, as well as location of active tooth wear. The three dentary specimens of Bentonyx sidensis are: A, BRSUG OSF 374; B, BRSUG 27200; C, BRSUG OSF 327. Location of the zone of active tooth wear is marked in each case (green), together with the inferred minimal position of the dental lamina active at each time (red). As the animal grew in size, new dentary bone plus teeth were emplaced posteriorly, and the area of active tooth wear moved backwards relatively as well. Scale bar represents 10 mm. the bone but show a chisel shape in which the enamel cap and tooth tip has been flattened. In the adult (Fig. 3), the occluding teeth are worn flush with the bone, but they are still present, with an extensive portion of the length of the roots in place, even for the three or four tiny, anteriormost baby teeth that have presumably been in occlusion for longest. Finally, in the old adult ( Fig. 4E-G), the surviving teeth are few and restricted to the zone behind the occlusal field.
The fact that the functional shearing surfaces of maxilla and dentary exposed naked bone in life is relatively unusual, although seen in recent acrodont lizards such as Uromastyx (Cooper & Poole 1973) and in the rhynchocephalian Sphenodon (Kieser et al. 2009), and rhynchosaurs must have had immunological adaptations to avoid infection.

Lateralization
Preferences for the right or left dentition, termed lateralization, are commonly observed in living tetrapods (Rogers 1980;Glick 1985;MacNeilage 2013), and evidence for preferential feeding behaviour has been noted in captorhinids. In a sample of 89 specimens, Reisz et al. (2020) observed that Captorhinus aguti showed a distinct preference for chewing on the right side, with 25% of teeth worn on the left side and nearly 40% on the right.
We have noted such lateralization in the complete Bentonyx skull, with 41 teeth on the right and 25 on the left maxillary tooth plate (Sethapanichsakul et al. 2023b, figs 3B, 4B), but this is a single example, and a larger sample size would be required to test whether such a preference for wear on the left side is the norm or not. In any case, these differences between the right and left maxillary tooth plates appear to be real rather than a result of damage or pathology because both maxillary tooth plates are complete and undamaged, and they are both of equivalent widths and lengths. Therefore, both tooth plates would be expected to show about the same numbers of teeth because teeth are of similar dimensions matching the animal's age, and they appear to be budded from the dental lamina according to the availability of space in the bone (Benton 1984).
The numbers of teeth retained at any point in the life of a rhynchosaur depend on the balance of age and tooth wear: juvenile teeth are worn away entirely and so cannot be seen in the adult, and perhaps here the rhynchosaur favoured the left-hand side of its mouth in chopping its food, so the teeth on that side have worn away, leaving broad expanses of the occlusal bone surface without any trace of the teeth that were presumably once there. Maxillary variation within individual rhynchosaurs, where it occurs, could have implications for taxonomic assessments based on isolated tooth-bearing elements.

Multiple tooth rows and area of oral processing
In order to understand the key features of rhynchosaur dentition, it is worth considering their context in terms of other herbivorous taxa, especially those that also had multiple tooth rows. Herbivorous tetrapods show a variety of adaptations to processing tough plant food, including complex and sometimes constantly growing teeth in mammals. Ornithischian dinosaurs can show complex occlusal surfaces as well as frequent tooth replacement. The Permian dicynodont Endothiodon shows an unusual mode of development of its multiple tooth rows, where teeth migrate labially through the tooth-bearing bone and do not replace existing teeth (Olroyd et al. 2021). In close relatives, the newly implanted teeth move through the bone from the lingual tooth margin, and partly resorb and replace older, functional teeth. In Endothiodon, on the other hand, tooth replacement does not occur, and functional and newer teeth migrate through the bone labially, but remain separate, hence generating multiple tooth rows (Olroyd et al. 2021).
The Carboniferous to Late Permian captorhinids show multiple tooth rows (Bolt & DeMar 1975;Ricql es & Bolt 1983;LeBlanc & Reisz 2015;Haridy et al. 2018), where the number of rows increases with size and age, as in rhynchosaurs, and the teeth become implanted into the bone of the jaw and slowly drift forward as a kind of conveyor belt of teeth (Olroyd et al. 2021, p. 2). In this case, unlike in Endothiodon, the teeth are fixed in the bone and do not drift. The same is true of rhynchosaurs, which show a similar conveyor belt of teeth slowly drifting anteriorly and labially, fixed in the bone of dentary or maxilla.
In captorhinids the area of mastication increased in size through ontogeny by growth on the posterior and lingual margins and progressive relocation of older parts anteriorly (Ricql es & Bolt 1983). The same is seen in rhynchosaurs, with addition of bone and teeth along the posterior and lingual margins of the dentary and maxilla. In addition, we have shown that the area of current tooth wear also moved, relatively speaking, posteriorly through the lifetime of the individual (Fig. 5). Supporting evidence for the posterior and lingual addition of bone and teeth are the resorption pits seen in the younger dentary (Fig. 4D), and that the only teeth with enamel caps are seen behind, not below or above, the areas of active tooth wear. Evidence for the relative posteriorward movement is that juvenile-sized teeth are seen, in adult individuals of rhynchosaurs, anterior to the area of current occlusion ( Fig. 2A-E). The numbers of unworn posterior teeth apparently remain roughly constant even though new bone and teeth are being generated constantly through life, and that the dentary becomes deeply bowed in larger, presumably older, individuals ( Fig. 5A-C). This relative posterior movement of the area of mastication, and the evidence for it, are seen in other rhynchosaurs (Benton 1984(Benton , 1990Scartezini & Soares 2022).
It is not clear whether old-aged rhynchosaurs outgrew their functional dentition, in some way analogous to the manner in which very old elephants die after the last of their adult molar teeth has worn down (Sanders 2017) or retained some ability to process tough plant food using just the bony surface of the dentary, as seen in aged, and functionally edentulous individuals of Uromastyx (Cooper & Poole 1973). In the presumed aged rhynchosaur (Fig. 4E-G), the absence of teeth in the oral processing area, and indeed the deeply convex shape of the occlusal area of wear in lateral view of the dentary (Fig. 5C), confirm how the exigencies of growth and necessary mutually curving maxillary/dentary tooth plates eventually appear to make the jaw system functionally compromised with age. What apparently worked well at juvenile to adult stages of ontogeny, with the maxilla curling up gently, might well have become more extreme as the nature of the curve became more exaggerated with increasing body size.

Rhynchosaur feeding and success
Rhynchosaurs fed using a precision-shear bite, in which the upper and lower jaw dentitions cut past each other, like a pair of scissors, and with no anteroposterior movement. Further, the tight occlusion between upper and lower jaws prevented any mediolateral movement, so nothing like chewing could occur.
Rhynchosaurs show major changes in their skull morphology and dental apparatus through the two phases of their dominance as herbivores in many parts of the world. The Late Triassic hyperodapedontines were larger than the Middle Triassic rhynchosaurs, and they are characterized especially by having had remarkably broad skulls, with hugely wide adductor chambers. Ontogenetically, this exceptional posterior width appears to represent a peramorphic trend, with extreme posterior skull width expansion possibly through a longer developmental cycle, namely hypermorphosis (Benton & Kirkpatrick 1989;Langer et al. 2000). Functionally, the extreme posterior width accommodated enormous adductor chambers in the posterior part of the skull, on either side of the braincase, suggesting very powerful jaw muscles capable of exerting enormous pressure on foodstuffs. All these modifications may well have rendered hyperodapedontines more efficient at processing tough plant food than their antecedents.
The diet of rhynchosaurs is uncertain. Their dentition was presumably adapted to process tough stalks and branches, and their clawed feet and premaxillary beak-like structures suggest digging capabilities, perhaps for nutritious tubers. There is, however, no direct evidence of their diet in the form of either stomach contents or identifiable coprolites. Anisian floras typically include horsetails, ferns, tree ferns, ginkgos and herbaceous conifers belonging to extinct families (Albertiaceae, Voltziales). Floras changed substantially in the Carnian; in particular, bennetitaleans and several modern conifer and fern families diversified, although it is uncertain whether any of these led to modifications in rhynchosaur feeding mode.
Notable also was the prevalence in Gondwana of the seedfern Dicroidium, a low-growing bushy plant that had been present since the Late Permian but seems to co-occur with hyperodapedontines in many locations (Colombi et al. 2021;Pedernera et al. 2022), although would not have formed part of the diet of Laurasian rhynchosaurs. These floral changes may relate to major climatic perturbations in the mid-Carnian, although there is dating uncertainty in many cases (Kustatscher et al. 2018).
For the duration of their existence, rhynchosaurs shared the planet with other similar-sized tetrapod herbivores, including various synapsids and archosaurs, sometimes living alongside them and dividing the herbivorous niches, as seen in the Carnian Ischigualasto Formation of Argentina (Singh et al. 2021).
Elsewhere, however, herbivore distributions are more disjunct, probably due to differing physiologies and environmental preferences. Thus, Middle Triassic sites in Northern China and cis-Ural Russia, which at the time were high latitude and relatively cool and humid (Sellwood & Valdes 2006), contain no ectothermic rhynchosaurs, but yield many dicynodonts, which were probably at least partially endothermic (Benton 2021) and evidently thrived in these conditions. In contrast, rhynchosaurs are conspicuous in deposits of similar age in the UK such as the Otter Sandstone, but dicynodonts are absent, perhaps because they required more vegetation and water (to meet their higher energy needs and facilitate nitrogen excretion) than these lower latitude and more arid environments could provide (Whiteside et al. 2011;Wolvaart et al. 2023).
Herbivorous tetrapods were impacted, positively and negatively, by major environmental changes during the Carnian, the most widespread and best-documented of which was the change to humid climates during the Carnian Pluvial Episode (CPE), from 233-232 Ma. Although dating is uncertain in many cases, it is evident that the various lineages of Middle Triassic rhynchosaurs disappeared before or during the CPE, and a single lineage, the Hyperodapedontinae, survived, although perhaps low in numbers and diversity, leading to a surviving fauna that was much more unified than before, and dominated over much of the world by the single genus Hyperodapedon (Fig. 1). Nevertheless, hyperodapedontines briefly and spectacularly flowered in the generally less humid conditions following the CPE, representing up to 90% of specimens collected at some sites (Benton 1983b;Benton et al. 2018). In the well-studied and accurately dated rock sequences of South America, this time has been identified by Ezcurra et al.

CONCLUSION
CT studies of the tooth-bearing elements of the Middle Triassic rhynchosaur Bentonyx sidensis confirm earlier studies of the evolution of the remarkable dentition of rhynchosaurs. Their ankylothecodont mode of implantation is confirmed as distinct from other modes of tooth implantation, and particularly from acrodonty, because their teeth show deep roots, there is no bone of attachment throughout life, and the tooth crowns do not survive under wear for long. Further, unlike thecodont or pleurodont teeth, the ankylothecodont teeth of rhynchosaurs do not sit in persistent sockets or in a dental groove.
The multiple rows of rhynchosaur teeth were implanted throughout their lifespans, the number of rows increasing as the animal aged, before probably declining in very old individuals. Juvenile teeth were small, and tooth size increased in a sharp step during ontogeny. Teeth were implanted from a posteriorly and lingually located dental lamina, and there was no tooth replacement (i.e. successions of teeth implanted below teeth currently in occlusion). During growth, as the size of the skull increased and especially expanded laterally, the maxillary tooth plate and dentary also broadened posteriorly, enabling additional anteroposteriorly oriented tooth rows to insert between existing rows. Further, as the jaws enlarged, they changed shape to maintain occlusion between the concave-upwards dorsal margin of the dentary and the matching maxilla, which bore a deep sulcus into which the dentary cut, producing a precision-shear bite with no propalinal or lateral movement.
New dating evidence confirms that rhynchosaurs experienced two phases of diversification, first in the Anisian, and second, following the Carnian Pluvial Episode. The Middle Triassic rhynchosaurs show more chaotic tooth implantation and growth to eruption and occlusion than the Late Triassic hyperodapedontines, whose dentitions were more orderly, with more regular arrangement of teeth in rows and limited evidence of tooth-tooth interference and resorption. The probably ectothermic rhynchosaurs preferred warm and dry palaeoenvironments, and in the Late Triassic flourished especially in times of more arid palaeoclimates interspersed between humid episodes better suited to endothermic synapsids.
Where present, they were often the numerically dominant, or even sole, medium-large sized herbivores in their ecosystems.