A new parareptile with temporal fenestration from the Middle Permian of South Africa

Transcription

1 A new parareptile with temporal fenestration from the Middle Permian of South Africa 9 Sean P. Modesto, Diane M. Scott, and Robert R. Reisz Abstract: The partial skeleton of a small reptile, from the Middle Permian Tapinocephalus Assemblage Zone of South Africa, is described as a new parareptile. Australothyris smithi gen. et sp. nov. is diagnosed by contact between the postfrontal and the supratemporal, a relatively large pineal foramen, a small interpterygoid vacuity, the presence of teeth on the ventral surfaces of the basipterygoid processes, and several other autapomorphies. Phylogenetic analysis of an augmented data matrix from the literature positions A. smithi basally within Parareptilia, as the sister taxon of Ankyramorpha. Our results indicate that the presence of a lateral temporal fenestra is synapomorphic for procolophonomorphs (as regarded in this paper) and that this structure was variously modified in procolophonians. Australothyris smithi is the phylogenetically youngest member of a grade of Gondwanan parareptiles that includes mesosaurids and millerosaurs, a topology intimating that parareptiles diversified first in Gondwana and then dispersed into Laurasia. This biogeographic scenario is at odds with observations that the bolosaurids of Laurasia predate mesosaurids and other Gondwanan parareptiles by approximately 15 million years. Résumé : Le squelette partiel d un petit reptile provenant de la zone de l assemblage de Tapinocephalus, du Permien moyen d Afrique du Sud, est décrit comme étant un nouveau parareptile. Le diagnostic d Australothyris smithi gen. et sp. nov. repose sur le contact entre les os postfrontal et supratemporal, un foramen pinéal relativement imposant, une petite vacuité interptérygoïde, la présence de dents sur les surfaces ventrales des processus basiptérygoïdes ainsi que plusieurs autres autapomorphies. L analyse phylogénétique d une matrice complète des données tirées de la documentation place A. smithi en position basale au sein des Parareptilia, comme taxon frère des Ankyramorpha. Les résultats de l étude indiquent que la présence d une fenêtre temporale latérale est synapomorphique pour les procolophonomorphes (comme les définit le présent article) et que, parmi les procolophoniens, cette structure a subi diverses modifications. Du point de vue phylogénétique, Australothyris smithi est le plus jeune membre d un grade de parareptiles gondwaniens qui comprend les mésosauridés et les millérosaures, une topologie qui porte à croire que les parareptiles se sont d abord diversifiés au Gondwana pour ensuite se disperser en Laurasie. Ce scénario biogéographique va à l encontre d observations voulant que les bolosauridés de la Laurasie aient vécu environ 15 millions d années avant les mésosauridés et autres parareptiles gondwaniens. [Traduit par la Rédaction] Introduction The Tapinocephalus Assemblage Zone (AZ) of the Karoo Basin of South Africa preserves the richest fauna of Middle Permian terrestrial vertebrates in the world. Recent collecting efforts have led to the description of several new species of therapsid synapsids (Modesto et al. 1999, 2002, 2003; Rubidge and Kitching 2003; Rubidge et al. 2006) and to the discovery of important new specimens of poorly known elements of the fauna, such as varanopid synapsids (Dilkes and Reisz 1996; Modesto et al. 2001; Botha-Brink and Modesto 2007). The collection of reptiles from the Tapinocephalus AZ has not been commensurate during this period, and, apart from descriptions of new materials of known taxa Received 1 August Accepted 8 January Published on the NRC Research Press Web site at cjes.nrc.ca on 18 March Paper handled by Associate Editor H.-D. Sues. S.P. Modesto. 1 Department of Biology, Cape Breton University, 1250 Grand Lake Road, Sydney, NS B1P 6L2, Canada. D.M. Scott and R.R. Reisz. Department of Biology, university of Toronto at Mississauga, 3359 Mississauga Road, Mississauga, ON L5L 1C6, Canada. 1 Corresponding author ( sean_modesto@cbu.ca). (Gow 1997; Gow and de Klerk 1997; Rubidge et al. 1999) and the attribution of a fragmentary postcranium to Procolophonoidea (Gow and Rubidge 1997), the taxonomic diversity of Tapinocephalus AZ reptiles has changed little since Kitching s (1977) seminal overview of the Karoo vertebrate faunas. An interesting aspect of the Tapinocephalus AZ reptilian fauna is that it is formed exclusively by parareptiles, specifically by pareiasaurs (Lee 1997) and millerosaurs (Thommasen and Carroll 1981; Gow and de Klerk 1997). In striking contrast, eureptiles are completely unknown in the Tapinocephalus AZ. They appear higher in Karoo stratigraphy, with the presence of diapsids and a captorhinid in the Tropidostoma AZ, which is either uppermost Middle Permian (Smith et al. 2006) or lowermost Upper Permian (Retallack et al. 2006). Whereas eureptiles were numerically more dominant and taxonomically more diverse than parareptiles in earlier continental faunas (and again in Mesozoic faunas), they remain relatively rare in the Karoo Basin until the end of the Permian. Accordingly, it is not unexpected that the collection of a new reptilian specimen from the Tapinocephalus AZ adds to the diversity of parareptiles of the Karoo Basin. The specimen, consisting of a partial skull and fragmentary postcranial elements, was discovered in the course of 1995 field Can. J. Earth Sci. 46: 9 20 (2009) doi: /e09-001

2 10 Can. J. Earth Sci. Vol. 46, 2009 work led by Roger M.H. Smith at the Northern Cape farm Beukesplaas, which yielded a small assemblage of pareiasaurs, dicynodonts, dinocephalians, and varanopids (Botha- Brink and Modesto 2007). It was catalogued originally as Owenetta sp., perhaps because the general shape of the skull, the presence of a postfrontal supratemporal contact, and the high maxillary tooth count were suggestive of a close relationship with Owenetta rubidgei (Reisz and Scott 2002, figs. 4, 5). Further examination of the specimen, however, reveals that unequivocal procolophonoid synapomorphies are absent. More intriguingly, the skull exhibits relatively large lateral temporal fenestrae, which resemble those of lanthanosuchoids (debraga and Reisz 1996). On the basis of these and other distinctive features, it is clear that this specimen represents a new genus and species. We describe this new parareptile here and evaluate its phylogenetic position using a modified data matrix from the literature. Institutional abbreviations NM, Karoo Palaeontology, National Museum, Bloemfontein, South Africa; SAM, Iziko South African Museum, Cape Town, South Africa. Material and methods The new specimen is preserved in a hard mudstone, which necessitated mechanical preparation by airscribe and pin vise. Prior to our study, the skull was separated from the postcrania, but this resulted in some of the periosteal bone of the left opisthotic being lifted away from the occiput, and the former is now a thin film of bone on the block that preserves the postcranial skeleton. Most postcranial elements were lost to weathering and many of the remaining elements are heavily abraded, leaving only some cervical vertebrae, associated rib fragments, and the interclavicle. Systematic palaeontology Reptilia Laurenti, 1768 Parareptilia Olson, 1947 Procolophonomorpha Romer, 1964 Australothyris smithi gen. et sp. nov. DIAGNOSIS: Small parareptile with a relatively high number of maxillary tooth positions, a postfrontal supratemporal contact, a large palatine, a small interpterygoid vacuity, the presence of denticles on the ventral surfaces of the basipterygoid processes, a medial flange on the quadrate for the stapes, a quadrate process of the stapes with a small, accessory flange, and a small, ventrally positioned retroarticular process. LOCALITY AND HORIZON: A locality on the farm Beukesplaas, Northern Cape Province, Republic of South Africa. Upper levels of the Abrahamskraal Formation, Beaufort Group, Karoo Supergroup. The strata on Beukesplaas are assigned to the Tapinocephalus Assemblage Zone, which is Middle Permian in age. HOLOTYPE: SAM-PK-K8302, a skull with occluded mandible and associated postcranial elements. ETYMOLOGY: The generic name is from Latin australis, meaning the south or the south wind, and ancient Greek thyris, meaning opening, in reference to the presence of temporal fenestrae in this South African parareptile. The specific epithet honours Roger M.H. Smith in recognition of his extensive contributions to Karoo palaeontology. Description SAM-PK-K8302 consists of a nearly complete skull (Fig. 1) and a few postcranial elements (Figs. 2, 3). Weathering has removed much of the skull roof, leaving a patchwork of dermal roofing elements. Nevertheless, it is clear that lateral temporal fenestrae are present, that the pineal foramen is relatively large, and that the splenial contributes to the symphysis, a unique suite of features indicating that the specimen cannot be assigned to any described clade of parareptiles. Skull roof Only the right premaxilla is present (Figs. 1B, 1D). Weathering has removed the dorsal process, and it is not possible to determine whether this process was relatively broad at its base, as in millerettids, or transversely narrow, as in ankyramorphs (sensu debraga and Reisz 1996). The bone accommodates five teeth. The labial surfaces of the teeth are weathered, but what is present indicates that the teeth are subequal, conical structures. Both maxillae are present but incomplete, with damaged anterior ends and dorsal margins. Although it seems likely that the maxilla made some contribution to the external nares, judging from the close association with the premaxilla on the right side of the skull, the exact nature of this contribution cannot be determined. The right maxilla has an elliptical embayment along the preserved dorsal margin, at the horizontal level expected for the ventral margin of the external naris (judging from the alveolar portion of the premaxilla), so it seems that this small embayment is all that remains of the right anterior lateral maxillary foramen (sensu Laurin and Reisz 1995; see also Modesto 1999). The preserved dorsal margin of the left maxilla is nearly straight. The anterodorsal portion of the dorsal process is missing, however, and we cannot determine if the maxilla was bounded dorsally by the lacrimal or if the maxilla also made contact with the nasal. The suborbital portion of the left maxilla is perfectly preserved and shows that its dorsal margin reaches the ventral margin of the orbit. As preserved, the left and right maxillae preserve 30 and 27 teeth, respectively (Figs. 1A, 1B). Although both maxillae are damaged, the left is more complete anteriorly and, judging from the space formerly occupied by the missing right premaxilla, no anterior teeth are missing (i.e., the left maxillary dentition is complete). Taking into consideration a gap at the 13th tooth position, there are 31 maxillary tooth positions. This number falls one tooth short of that documented for Lanthanosuchus watsoni (32 teeth: debraga and Reisz 1996, fig. 4B), which exhibits the highest maxillary tooth count among parareptiles. Despite the high number of maxillary teeth in SAM-PK-K8302, the upper dentition extends no farther posteriorly than the level of the midpoint of the orbit, even though the maxilla has a 2 mm long edentulous posterior tip (equivalent to a section of alveolar margin housing 4 or 5 marginal teeth). The maxillary teeth are slender cones with slightly recurved tips. The teeth are isodont

3 Modesto et al. 11 Fig. 1. Australothyris smithi gen. et sp. nov., SAM-PK-K8302, holotype. Specimen drawings of skull in (A) right lateral, (B) left lateral, (C) occipital, (D) dorsal, and (E) palatal views. Scale bar = 10 mm. an, angular; ar, articular; bo, basioccipital; c, coronoid; d, dentary; ec, ectopterygoid; eo, exoccipital; ep, epipterygoid; f, frontal; f m, foramen magnum; j, jugal; l, lacrimal; l t f, lateral temporal fenestra; m, maxilla; m f, anterior lateral maxillary foramen; n, nasal; op, opisthotic; p, parietal; pal, palatine; pbs, parabasisphenoid; pf, postfrontal; pi, pineal foramen; po, postorbital; pop, paroccipital process of opisthotic; pp, postparietal; pra, prearticular; prf, prefrontal; prm, premaxilla; pt, pterygoid; p t f, post-temporal fenestra; q, quadrate; qj, quadratojugal; s, stapes; sa, surangular; so, supraoccipital; sq, squamosal; sp, splenial; st, supratemporal; t, tabular; v, vomer. (i.e., there are no caniniforms, nor is there a caniniform region), but the posterior maxillary teeth exhibit a slight, progressive decrease in size. Damage to some of the teeth reveals that the pulp cavity extends nearly to the apex of the tooth. The diameter of the pulp cavity is slightly greater than the thickness of the tooth wall on either side of the cavity. The external surfaces of the teeth are smooth; there is no evidence of fluting or plications (indicative of polyplocodonty sensu Schultze 1970), as documented in millerettids and a few other parareptiles (Modesto and Reisz 2008). The teeth are set into shallow pits and anchored by bone of attachment. Weathering has resulted in the loss of the right nasal and damage to the left (Fig. 1D). Only the central portion the latter remains, including a part contributing to the midline. The damage is substantial enough to preclude even general inferences of the nature of the contacts with neighbouring elements. What is preserved of the external surface suggests that the nasal exhibited light ornamentation of shallow, broad grooves. The anterior end of neither lacrimal is present, and it is not possible to determine whether or not this bone reached the external naris. Weathering has polished the external surfaces of the lacrimals, but each clearly exhibits a single lacrimal opening and a posteroventral process that extends medially along the suborbital bar to contact the jugal. Only the orbital portions of the prefrontals remain, and what is preserved shows that each forms the anterodorsal corner of the orbit and contributes to a modest antorbital buttress. Enough remains of both elements to demonstrate that there is no posterior process or spur extending from the antorbital process, as in basal procolophonoids (Reisz and Scott 2002). The anterior portions of the frontals are preserved in articulation via a slightly undulating midline suture (Fig. 1D). The available evidence suggests that the frontal must have

4 12 Can. J. Earth Sci. Vol. 46, 2009 Fig. 2. Australothyris smithi gen. et sp. nov., SAM-PK-K8302, holotype. Axial skeleton in (A) dorsal and (B) right lateral views. Scale bar = 5 mm. ant zy, anterior zygapophysis; ax, axis; ax n sp, axial neural spine; c r, cervical rib; d r, dorsal rib; n sp, neural spine; post zy, posterior zygapophysis; tr pr, transverse process. Numerals indicate presacral vertebra number. Fig. 3. Australothyris smithi gen. et sp. nov., SAM-PK-K8302, holotype. Interclavicle in dorsal view. Scale bar = 5 mm. had a broad contact with the nasal, and that it is unlikely that the prefrontal had a medial extension that interposed between them, as in some procolophonoids (Reisz and Scott 2002). The left element is also preserved enough to show that the frontal made a small contribution to the dorsal margin of the orbit. Despite some abrasion to this area of the frontal, the extent of the bone here, together with what is preserved of the dorsal surface of the left prefrontal, suggests that the frontal does not form a frontal lappet, which is present in lanthanosuchoids (debraga and Reisz 1996). The jugal, one of the best preserved elements, is a triradiate element, with an anterior suborbital process, a posterodorsally extending postorbital process, and a posterior subtemporal process, each extending from a substantial central portion. The suborbital process is narrowly triangular with a weakly curving orbital margin. Damage to the right maxilla indicates that the jugal is broadly overlapped laterally by the former bone, and that the jugal shares a short scarf joint with the lacrimal. The postorbital process is relatively broad and underlies the anterior half of the postorbital. Together with the subtemporal process, the postorbital process forms a deeply embayed contribution to the lateral temporal fenestra. The jugal is bordered ventrally by the quadratojugal and the maxilla; thus, it does not reach the ventral margin of the skull roof. The jugal is ornamented with a patchwork of broad, shallow grooves and pits. The postorbital spans the skull table and the temporal region. In dorsal view the postorbital is a moderately wide bone that maintains its breadth throughout most of its length (Fig. 1D). Anteriorly, the postorbital forms the posteromedial corner of the orbit and overlies the postorbital process of the jugal. Ventrally, it has a narrow contribution to the dorsal margin of the lateral temporal fenestra. Medially, the postorbital is bounded by the postfrontal and the supratemporal. Posteriorly, the postorbital overlies the dorsal portion of the squamosal; judging from the left side, the latter bone contacts the posterior two-fifths of the ventral surface of the former. The more complete left postorbital is ornamented with a cluster of low tuberosities anteriorly and a system of diverging furrows posteriorly. Both squamosals are present, but only the left is closely associated with neighbouring elements. The right squamosal was disarticulated and repositioned below the right half of the skull table, just posterior to the orbit. The weathering that removed large portions of the skull table on the right side has exposed the right squamosal in dorsolateral view,

5 Modesto et al. 13 revealing its polygonal shape. The otic notch is a narrow, concave shelf that spans the entire height of the squamosal. Dorsomedially and anterodorsally, there are rugose sutural surfaces for the supratemporal and the postorbital, respectively. The anteroventral margin of the squamosal descends from a sharp anterior tip to form the posterodorsal corner of the lateral temporal fenestra. Ventrally, the squamosal appears to have been overlapped slightly by the quadratojugal. In occipital view, the left squamosal exibits a notch ventrally, which probably represents the dorsal margin of the quadrate foramen (Fig. 1C). Exclusive of the otic notch, the lateral surface of the squamosal is sculpted with low, irregular mounds. In lateral view, the quadratojugal is a narrow, triangular element (Figs. 1A, 1B). It forms the slightly sigmoidal ventral margin of the temporal region. The available evidence from both sides of the skull indicates that the quadratojugal contacted the maxilla. Anterodorsally, the quadratojugal shares a narrow overlapping suture with the jugal, and posterior to this, the former bone forms the ventral margin of the lateral temporal fenestra. Posteriorly, the quadratojugal curves medially and extends a relatively short, tongue-like flange, forming a wrap-around contact with the ventral margin of the squamosal dorsally and the condylar portion of the quadrate ventrally. This posteromedial process of the quadratojugal exhibits a shallow excavation that appears to have been continuous with the otic shelf of the squamosal, indicating that the former element formed the ventral onethird of the otic notch. The quadratojugal and the quadrate form the ventral margin of the quadrate foramen. Neither postfrontal is well preserved, but together they provide a nearly complete picture of the element (Figs. 1C, 1D). The right postfrontal is preserved only as a narrow crescent of bone forming the posterodorsal margin of the orbit. The left element is preserved as two small fragments: the posterior fragment is the larger and more informative of the two, and shows that the postfrontal extended posteriorly to contact the supratemporal and thus separate the parietal and postorbital bones. The supratemporal is a polygonal bone that contacts the postorbital anterolaterally, the postfrontal anteriorly, and the parietal and the tabular medially. The dorsal surface is slightly convex and ornamented with a radiating pattern of short, slightly curving grooves. The posterior end of the supratemporal curves posteroventrally as an occipital shelf that is semilunate to crescentic in outline. Medially, the occipital edge is distinctly irregular, presumably for the suture with the tabular. The margin then angles posterolaterally to form a free margin that appears to have formed a relatively narrow post-temporal opening. The ventral tip of the bone is a triangular or narrow process that likely contacted the dorsolateral corner of the paroccipital process. Whereas only the posterior portion of the right parietal is preserved, the left appears to be missing only the anterior edge that would have contacted the frontal (Fig. 1D). The parietal is a relatively large, plate-like bone embayed medially for the pineal foramen. This appears to have been a suboval opening, with the long axis aligned sagittally. The foramen is relatively large: the long axis is approximately 11% 12% total skull length, a size matched among parareptiles by procolophonids (Carroll and Lindsay 1985, figs. 3, 5) and bolosaurids (Reisz et al. 2007, fig. 9). Weathering on the left side of the skull table reveals that the anterolateral corner of the parietal is broadly overlain by the postfrontal. The posterolateral corner of the parietal contacts the supratemporal laterally and the tabular posteriorly. The medial half of the posterior margin of the parietal makes contact with the postparietal in a deeply sigmoidal suture, suggesting that the paired postparietals fit into a median notch formed by the parietals. The parietal is entirely a skull-table element because it does not form an occipital shelf. The dorsal surface is ornamented with a system of radiating grooves of varying breadth, some of which are continuous with those on adjacent elements. The postparietal is small plate-like element bridging the skull table and the occiput (Figs. 1C, 1D). It has a small dorsal component that is integrated into the skull table medially, but most of the bone is occipital. The paired postparietals share a roughly straight midline suture and overlie the dorsomedial part of the supraoccipital. The postparietal sutures to the parietal anteriorly and the tabular laterally, although the extent of the latter contact is unclear because of damage to both tabulars. The occipital surface of the bone is weakly excavated and smoothly finished. Both tabulars have been weathered down to bone splints. Combined information from the two suggests that the tabular was a relatively small element that was not much larger than the postparietal in occipital view and would have been closely comparable in relative size and shape to the tabular of Milleretta (Carroll and Lindsay 1985, fig. 2B). The tabular of SAM-PK-K8302 is bounded by the parietal anteriorly, the postparietal medially, the supratemporal laterally, and ventrally, it overlies the dorsal portion of the supraoccipital. What remains of the left tabular suggests that its ventral margin was continuous with those of the postparietal and the supratemporal. Although difficult to determine, the tabular may have made a small contribution to the dorsal margin of the post-temporal opening. Palate As in most parareptiles, the palate is formed by paired vomers, palatines, ectopterygoids, and pterygoids (Fig. 1E). The choana is mostly obscured by the occluded mandible, but appears to be a narrow opening about one-third the length of the skull. The interpterygoid vacuity is remarkably short, being approximately as long as broad. The vomer is roughly triangular in ventral aspect, expanding in transverse breadth from a truncated anterior tip to its presumed contact with the palatine posterolaterally and then narrowing more abruptly posteriorly to form a small contact with the anterior end of the pterygoid. The approximate twothirds of the lateral margin forms most of the medial margin of the choana. The lateral edge for the choana is thickened and bears short rows of small teeth. A secondary ridge extends posteriorly and slightly medially over the ventral surface of the bone from the middle of this parachoanal ridge and bifurcates at its approximate halfway point, producing a short lateral ridge, which narrows towards the posterolateral margin of the bone, and a medial ridge, which extends posteromedially to the posterior tip of the vomer and appears to be continued along the medial margin of the pterygoid. The surface of the bifurcating ridge bears dispersed single and pairs of teeth.

6 14 Can. J. Earth Sci. Vol. 46, 2009 The palatine is distinguished by its relatively large size (Fig. 1E) and resembles the relative dimensions and the outline of the palatine of Lanthanosuchus (debraga and Reisz 1996, Fig. 4B). This bone has extensive contacts with both the vomer and the pterygoid medially, the ectopterygoid posterolaterally, and presumably with the maxilla laterally (obscured by the mandible). Anteriorly, the palatine probably forms the posterior end of the choana, but the occluded mandible and slight dissociation of the palatine from the vomer preclude mechanical preparation of this area. The ventral surface bears rows of small teeth on low ridges or swellings, of which one or two ridges continue onto the pterygoid. As seen through the left orbit, the palatine has an extensive lateral contact with the lacrimal and a shorter contact with the jugal. The dorsal surface is smoothly finished. It has a small orbitonasal ridge (Fig. 1B) that contacts the ventral tip of the prefrontal, with which it forms the medial margin of the orbitonasal foramen. The ectopterygoid is mostly covered by the occluded mandible; only the medial portion of the bone can be seen in ventral view, and it is a featureless polygon of flat bone on both sides. In dorsal view, the ectopterygoid expands laterally toward its contact with the jugal. The dorsal surface features a sharply defined, parabolic ridge. The pterygoid is remarkably broad (Fig. 1E). This is achieved, in part, by the extensive interpterygoid suture, which is longer than the interpterygoid vacuity (as measured from the anterior end of the basipterygoid articulation). From the interpterygoid suture and the anterior contact with the vomer, the pterygoid sweeps posterolaterally as a slightly curving, large sheet of bone. The ventral surface bears three ridges topped with rows of teeth. The medial margin is also thickened and bears intermittent rows of teeth, which extend as far anteriorly as the contact with the vomer. Posterolaterally, the transverse flange extends posteroventrally towards the lingual surface of the mandibular ramus. The posterior margin of the transverse flange is aligned approximately in the transverse plane. It features a thickened ridge that bears a single row of teeth, although it is possible that there are smaller anterior teeth embedded in the matrix that we have left to shore up the exposed tooth row. Whereas the row of teeth ends medially halfway across the base of the quadrate process, the ridge arcs posteriorly to merge with the base of the arcuate flange. Between the interpterygoid vacuity and the arcuate flange, the pterygoid projects a narrow collar from its medial surface to form the basicranial recess. Posteriorly, the pterygoid has a relatively short quadrate process. Its morphology is distinctive in that the anterior base of the process is positioned slightly more laterally than in millerettids (Gow 1972, fig. 3) or owenettids (Reisz and Scott 2002, fig. 2B), resulting in a conspicuous posteroventral notch between the basipterygoid region and the base of the quadrate process. In the lateral offsetting and presence of a posteroventral notch, the quadrate process in SAM-PK-K8302 resembles that seen in Lanthanosuchus (debraga and Reisz 1996, fig. 4B). A damaged triangular element positioned in the right orbit, in contact with the subtemporal process of the jugal, may be a dislodged right epipterygoid (Figs. 1A, 1D). The undamaged bone surface is smooth and featureless, and there is no preserved dorsal process. Both quadrates are preserved in contact with neighbouring skull bones, as well as in articulation with the mandible, and thus are largely obscured by the surrounding elements. What can be discerned is that the condylar region is relatively massive, forming almost the entire breadth of the element, and it supports a dicondylic articulating facet. Combining information from both elements, the articulating facet appears to be twice as broad as anteroposteriorly long. The breadth of the facet is nearly 14% the total length of the skull. The dorsal process is unusually small: it forms only approximately 23% the total transverse breadth and 35% of the total height of the bone. In these respects, the quadrate is most similar to that of the lanthanosuchoid Acleistorhinus (debraga and Reisz 1996). There is a small flange of bone on the medial surface, where the quadrate flange meets the condylar region. This medial flange may have contacted the lateral tip of the stapes (see description in following section). Braincase SAM-PK-K8302 preserves a nearly complete, articulated braincase that includes a well-preserved stapes (Figs. 1C 1E). It exhibits a relatively large and deep foramen magnum. The posterolateral corners are formed by short, but tall, paroccipital processes, resulting in a small post-temporal opening. The parasphenoid and the basisphenoid are indistinguishably fused. The supraoccipital and opisthotics are also fused together, suggesting that this individual was an adult or close to maturity. The parabasisphenoid is a broad triangular bone. The cultriform process is broadly based and relatively short. The remains of small teeth are present along its ventral midline. The basipterygoid processes are stubs that project anteroventrally from the main body of the bone. They meet the pterygoids at about a 208 angle from the sagittal plane, although the facets are not visible in ventral view. The ventral surfaces of the basipterygoid processes bear small clusters of teeth. Posterior to the basipterygoid processes, the breadth of the parabasisphenoid increases dramatically to the level of the jaw articulation and then appears to narrow abruptly, judging from the left side. Hoever, the outline of the posterior margin cannot be determined because most of the posterior region of the parabasisphenoid, including the portion that would have overlapped the basioccipital, is not preserved. The relatively large, broad basioccipital (Figs. 1C, 1E) features an anteromedial depression flanked by paired lateral tubera that are little more than swellings on the ventral surface of the bone. Its sutures with the exoccipitals are clear and indicate that the latter do not meet at the midline, with the basioccipital forming the floor of the foramen magnum. The broad, slightly reniform condyle is formed entirely by the basioccipital, and it is dimpled posteriorly by a small, circular notochordal pit. The exoccipitals extend dorsally from the posterodorsal surfaces of the basioccipital as small crescents. Their slightly thickened medial margins form the lateral walls of the foramen magnum. The exoccipitals contact the supraoccipital dorsally and the opisthotics ventrally, and are notched laterally for the metotic foramen, which lies at the level of the ventral margin of the foramen magnum. The supraoccipital can be subdivided into a low, broad dorsal process and an equally low, broad ventral portion

7 Modesto et al. 15 that forms the dorsal margin of the foramen magnum and contacts the underlying exoccipitals and opisthotics. Much of the dorsal process is obscured by the overlapping postparietals and tabulars, of which the former, at least, extend ventrally to the level of the top of the paroccipital process. The transition between the lateral margin of the dorsal process and the paroccipital process is distinguished by a small notch that demarcates the medial extent of the post-temporal opening. The ventrolateral portions of the supraoccipital swell posterolaterally from the base of the dorsal process to merge with the opisthotics, as well as (presumably, because the supraoccipital is fused to the opisthotics) making small, medial contacts with the exoccipitals. Both opisthotics are present and well preserved (Fig. 1). Surface detail is better on the right element because the periosteal bone on the occipital surface of the left element was lifted away when the skull was separated from the block containing the postcrania. Each opisthotic consists of a massive medial portion that contacts the supraoccipital, the exoccipital, and probably the prootic; a ventral process that contacts the basioccipital and forms the posterior margin of the contact for the stapes; and a short, deep paroccipital process. In its relative size and shape, the medial portion of the opisthotic is similar to that of Milleretta (Carroll and Lindsay 1985, fig. 2B). The ventral process is broader transversely than in Milleretta, being as broad as the paroccipital process is deep. The paroccipital process is short, as in Milleretta, but it is distinctly taller dorsoventrally than in that genus, being nearly as tall as the foramen magnum. Its posterior surface faces posterodorsally and is slightly convex, which is seen clearly in lateral aspect. The process thins laterally, such that the thickness of the lateral margin is about one-sixth the anteroposterior height of the process. The ventral surface of the paroccipital process is conspicuously convex, its curvature paralleling the curved contact with the stapedial footplate. A well-preserved and robustly constructed stapes is present in its expected position on the right side. It has a large footplate, a prominent, circular stapedial foramen, and a definite dorsal process. The footplate is roughly oval in outline, with the long axis aligned anteroposteriorly and nearly 40% the anteroposterior length of the braincase. The dorsal process is a small, triangular flange aligned in the transverse plane. The quadrate process (or lateral end) is a relatively complex knob of bone: the greater portion consists of a rounded boss, and about mid-way between the lateral tip and the dorsal process on the posterodorsal surface there is a small, obliquely oriented accessory flange. The footplate is in perfect articulation with the opisthotic and the prootic but not with the quadrate, presumably because of the dorsoventral compression of the skull. The quadrate process of the stapes probably made contact with the small medial flange of the quadrate, and the stapes may have served to buttress the temporal region of the skull against the braincase. Mandible This structure is preserved in tight occlusion with the skull, preventing study of the mandibular dentition and details of the lingual surface of the lower jaws. Much of the prearticular, the coronoid, and the articular is obscured by surrounding elements. The dentary is the largest element of the mandible (Figs. 1A, 1B, 1E). Anteriorly, it forms the dorsal portion of the symphysis and can be seen to support a series of tightly packed teeth. The teeth in the right dentary are too damaged for description, but the one tooth exposed on the left side shows that the dentary teeth are slightly smaller versions of opposing teeth in the upper marginal series. The dentary contacts the splenial medially and the angular and the surangular posteriorly; the nature of its presumed contacts with other post-dentary elements cannot be discerned. Both splenials are present and well preserved. They meet at the midline and form the ventral part of the symphysis. The splenial sheathes the anterior part of the lingual surface of the mandibular ramus, extending posteriorly to the approximate midpoint of the ramus. In addition to sharing a moderately long, overlapping suture with the angular along its posterolateral margin, the splenial also contacts the prearticular posteriorly. The splenial has no lateral exposure. Although the dorsal margin of the surangular is occluded by the overlying skull roof, it is clear that this bone is relatively low throughout its length. The surangular extends anteriorly to the level of the posteriormost maxillary dentition, where it is overlain by the posterior end of the dentary. Ventrally, the surangular is overlain by the broadly convex dorsal margin of the angular. Accordingly, the posterior end of the bone appears expanded dorsoventrally, where, with the posterior end of the angular, the surangular sheathes the lateral surface of the articular. The lateral surface is devoid of foramina and ornamentation. Like the surangular, the angular has a relatively low aspect in lateral view (Figs. 1A, 1B). It is also a remarkably long element, for it extends as far anteriorly as the levels of the posterior end of the choana and the antorbital margin. Anteriorly, it is sandwiched between the posterior ends of the dentary and the splenial, and dorsally it overlaps the surangular. Most of its medial contact is with the prearticular. Posteriorly, it overlaps the articular. The lateral surface is slightly concave, such that it forms a faint, ventrolaterally facing longitudinal trough. Posteriorly, along the contact with the prearticular, the ventral margin of the bone forms a low ridge that seems to overhang the prearticular in ventral aspect. Well-preserved regions of the lateral surfaces of both angulars exhibit relatively smooth bone that is very lightly sculpted with short, longitudinally aligned furrows. The coronoid is seen through the orbit in dorsal aspect, lying anteriorly in the subtemporal fenestra and posterolateral to the ectopterygoid (Fig. 1D). What is visible suggests that it forms a low coronoid eminence. The discernible parts of the prearticular do not differ greatly from those seen in other parareptiles. This bone is predominantly a lingual element that runs along the medial surface of the angular, from the posterior tip of the splenial to the ventromedial face of the articular. In the region adjacent to the subtemporal opening, the prearticular twists from a primarily vertical orientation to a more horizontal attitude where it underlies the articular. The articular, largely obscured by surrounding elements (Figs. 1A, 1B, 1E), is relatively tall and features a small, posteriorly positioned retroarticular process. The latter is formed entirely by the articular and has a gently concave posterodorsal process that is confluent with the posterior

8 16 Can. J. Earth Sci. Vol. 46, 2009 margin of the bone. The articulating facet for the quadrate is not exposed in either element, but the posterior view indicates that it formed a sigmoidal contact for the condyles of the quadrate. Postcranial skeleton The skull was preserved originally in articulation with a short series of cervical vertebrae and associated postcranial elements. Many of these elements are heavily weathered, unidentifiable bits of bone, but the neural arches of the cervical vertebrae are exposed and reasonably well preserved, as is the interclavicle. A few ribs are closely associated with the vertebrae, but they afford little anatomical information. The cervical vertebrae are represented by the axis and the two succeeding cervicals (Fig. 2). We have exposed these in dorsal aspect only to preserve the occipital impression into which the skull fits. Thus, we can present no details of the centra and cannot determine whether the neurocentral sutures are open or closed. The neural arch of the axis is a relatively robust structure, with a thick, low neural spine that rises ridge-like from the neural arch proper. Shallow lateral excavations hint at rather than demarcate the base of the spine from the rest of the arch. In dorsal view, the neural arch can be seen to be narrow in transverse breadth, and the bone here is not swollen or buttressed (Fig. 2A). The anterior part of the neural arch on both sides slopes approximately 458 from the base of the spine and presumably formed facets for articulation with the atlantal neural arches here, but details did not survive mechanical preparation of the recalcitrant matrix. Like the axis, the post-axial vertebrae have relatively narrow neural arches, which are slightly longer anteroposteriorly than they are wide, and their neural spines are distinct extensions from their respective arches. The neural spines are slightly longer anteroposteriorly than they are dorsoventrally tall (Fig. 2B). The neural spine of cervical 4 exhibits a truncated dorsal margin, and it is roughly the same relative thickness (compared to neural arch breadth) as that seen in Milleretta (Gow 1972, fig. 15A). Minor pocketing is present on the neural arch at the base of the neural spine in cervical 4, but this is not present on cervical 3. None of the zygopophyseal facets are exposed because the vertebrae are fully articulated, but it is apparent that the facets of the anterior zygopophyses would have faced mainly dorsally and would have been angled no more than a few degrees inwards from the horizontal plane. The right transverse process of cervical 4 is the only such process exposed among the three cervicals, and it projects laterally and slightly posteriorly as a narrow finger of bone. The interclavicle, exposed in dorsal aspect, is preserved partly as bone and the greater part as impression (Fig. 3). The right lateral process of the anterior plate is absent, but enough is preserved of the left lateral process to indicate that the interclavicle was anchor-shaped in outline and that the anterior head, at least, resembles that illustrated for the owenettid described by Reisz and Scott (2002, fig. 1). The lateral process is subtriangular and fin-like in dorsal view. Unlike the interclavicles of ankyramorphs, such as Procolophon (debraga 2003, fig. 3), the anterior margin is not the thickest region of the bone and instead appears to be the thinnest region of the interclavicle in SAM-PK-K8302. The bone becomes gradually thicker towards the central region of the anterior head, as indicated by the impression of the central portion of this area, suggesting that the interclavicle received the medial portions of the clavicles in broad, ventrally facing facets, much like the interclavicle morphology illustrated for Milleretta rubidgei by Gow (1972, fig. 16b). The impression of the median region of the anterior head and the anterior half of the stem of the interclavicle shows that the ventral surface of the bone is smooth and featureless. The stem is a remarkably narrow ribbon of bone and, although slightly distorted from lithostatic crushing, it appears to be slightly expanded caudally. Phylogenetic analysis To investigate the phylogenetic position of Australothyris smithi, we coded it for the 137 characters of Müller and Tsuji (2007) and ran the augmented matrix in PAUP* 4.0b10 (Swofford 2002) using the heuristic algorithm. We also ran decay and bootstrap analyses using the same algorithm to generate support values for clades. A heuristic search discovered three optimal trees, the strict consensus of which is shown in Fig. 4A. In all trees, A. smithi is the sister taxon of Ankyramorpha. A decay analysis indicates that this relationship requires four extra steps to collapse, making it one of the strongest clades in our analysis (see Fig. 4 caption for support values). Accordingly, we feel that it is appropriate to name this new clade of parareptiles. We bestow Romer s (1964) name Procolophonomorpha on the clade of A. smithi and Ankyramorpha. Lee (1995) applied this name to a clade that contained turtles, procolophonoids, pareiasaurs, lanthanosuchids, and a few other taxa, but it is equivalent (apart from the inclusion of turtles) to clades discovered in more recent studies to which the name Ankyramorpha has been applied (Reisz et al. 2007; Müller and Tsuji 2007; Modesto and Reisz 2008). Without explanation, debraga and Rieppel (1997) redefined Procolophonomorpha and restricted it to a clade that included Macroleter poezicus, pareiasaurs, and procolophonoids, but that clade is equivalent to Procolophonia sensu Laurin and Reisz (1995) in light of recent studies placing M. poezicus and pareiasaurs as sister taxa (Tsuji 2006; Müller and Tsuji 2007). Given that Procolophonomorpha has seen little use since Lee (1995, 1997) and debraga and Rieppel (1997), we resurrect it here to preserve it as a clade name. We redefine Procolophonomorpha as a stem-based group that includes Procolophon trigoniceps Owen, 1876 and all taxa related more closely to it than to Milleretta rubidgei Broom, We believe that this definition approximates the primary intent of Lee (1995, 1997) to identify a major group of parareptiles, the sister group of millerettids (more recently, millerosaurs: millerettids and Eunotosaurus africanus), as procolophonomorphs. Procolophonomorpha, as redefined here, is diagnosed by the presence of a lower temporal opening (29-1) and five other unambiguous synapomorphies (50-1, 58-1, 68-0, 72-1, 83-1), as well as eight ambiguous synapomorphies under accelerated transformation optimization (18-1, 40-0, 46-1, 93-0, 96-0, 100-1, 101-1, 116-1) and one (31-1) under delayed transformation (character numbers follow Müller and Tsuji 2007).

9 Modesto et al. 17 Fig. 4. Parareptilian phylogeny (A) with major clades identified and (B) showing the pattern of evolution of lateral temporal fenestrae (l t f) in parareptiles. Each tree represents the strict consensus of the three optimal trees discovered in a PAUP* 4.0b10 heuristic analysis of an augmented version of the data matrix in Müller and Tsuji (2007). No polytomies are shown because the disagreements among the source trees concern only the interrelationships of nycteroleterids (subsumed here in Procolophonia). Tree length = 400, consistency index = 0.43, rescaled consistency index = We recovered all of the primary clades discovered by Müller and Tsuji (2007) and so have condensed most of the terminal taxa used by those authors into the following taxa: Bolosauridae (Belebey, Eudibamus); Diadectomorpha (Diadectidae, Limnoscelidae); Eureptilia (Araeoscelidia, Captorhinidae, Paleothyris, Younginiformes); Lanthanosuchoidea (Acleistorhinus, Lanthanosuchus); Millerosauria (Eunotosaurus, Millerettidae); Procolophonia (Barasaurus, Bashkyroleter bashkyricus, Bashkyroleter mesensis, Bradysaurus, Emeroleter, Macroleter, Nycteroleter, Owenetta, Pareiasuchus, Procolophon, Scutosaurus, Tokosaurus). Australothyris is coded for the 137 characters of Müller and Tsuji (2007) as follows:?1?x01y00? ??1 0010? ? ?1? ??00 0?1000?0???????????0 0??????????????????????????????00000?, where X and Y indicate uncertainty for states 0 and 1 and for states 1 and 2, respectively. Support values for each clade (bootstrap and decay values): A (Amniota), 88/3; B (Reptilia), 87/5; C (Parareptilia), 79/4; D, 70/2; E (Procolophonomorpha), 70/4; F (Ankyramorpha), 70/4; G, 45/1; H, 42/1. Discussion The results of our phylogenetic analysis have interesting implications for elucidating the evolutionary history of lateral temporal fenestrae in parareptiles and for the biogeography of these early reptiles. Cisneros et al. (2004) inferred that lateral temporal fenestrae evolved in multiple lineages within Parareptilia, but also speculated that the presence of these openings may be synapomorphic for the group (sensu debraga and Reisz 1996). The latter possibility was not supported at the time because Cisneros et al. (2004, fig. 3) did not consider taxa that had been previously identified as parareptiles (Eunotosaurus africanus: Gow and de Klerk 1997; bolosaurids: Berman et al. 2000). Tsuji (2006) suggested that the presence of lateral temporal fenestrae is a plastic feature among parareptiles, an idea supported partly by Cisneros (2008) conclusion that temporal fenestration is a polymorphic trait in Procolophon trigoniceps. Our tree topology implies that the presence of a lateral temporal fenestra is synapomorphic not for Parareptilia, as hinted at by Cisneros et al. (2004), but that it is synapomorphic for its subclade Procolophonomorpha (Fig. 4B). According to our phylogeny, lateral temporal fenestrae appeared in the procolophonomorph ancestor as a relatively large opening bounded by (at least) the jugal, the squamosal, and the quadratojugal, and may have resembled the lateral temporal fenestra exhibited by Acleistorinus pteroticus (de- Braga and Reisz 1996, fig. 1D). However, there may have been participation by the postorbital as well, judging from the configuration seen in A. smithi and lanthanosuchids. The lateral temporal fenestra was retained but slightly modified in bolosaurs (Carroll and Gaskill 1971; Reisz et al. 2007), which exhibit an opening that is both anteroposteriorly expanded and dorsoventrally compressed compared with A. smithi and lanthanosuchoids, and the postorbital is broadly excluded from the opening by an extensive jugal squamosal contact. The lateral temporal fenestra was modified in the ancestor of procolophonians and Nyctiphruretus with the loss of the subtemporal bar, yielding the ventral temporal emargination seen in Nyctiphruretus and procolophonoids. The ventral temporal emargination was lost in the ancestor of pareiasaurs and nycteroleterids, which re-evolved the anapsid condition (i.e., no lateral temporal fenestra or ventral temporal emargination). However, a small lateral temporal fenestra, formed by the jugal, the squamosal, and the quadratojugal, reappeared in the clade of Macroleter and Tokosaurus (Müller and Tsuji 2007). The early stages of this evolutionary reconstruction are relatively robust: the position of A. smithi as the most basal procolophonomorph is stable, requiring seven extra steps to make this species form a clade with either lanthanosuchoids or bolosaurids (scenarios that would render ambiguous the evolutionary history of lateral temporal fenestration in parareptiles). An alternative tree that is five steps longer than our shortest tree positions A. smithi as the sister taxon of an ankyramorph clade that excludes lanthanosuchoids, but this topology does not alter the pattern of gains and losses of temporal fenestration outlined in the preceding paragraphs. Finally, alternative trees that position A. smithi farther basally in the phylogeny require at least six extra steps, which