2.2.1. Skull roof

In Deltavjatia, as in all pareiasaurs, the skull roof in mature individuals is fused to the palate and braincase. The skull in pareiasaurs is particularly deep, with wide, flaring cheek flanges.

The cranial sculpture (Figs 1, 2, 4A, 11, 13, 18–20) of Deltavjatia is distinctive. It involves a combination of large, irregularly rounded bosses interspersed with sparse, randomly distributed and roughly circular pits. There are no consistent raised radiating ridges as seen in Pareiasaurus and Pareiasuchus (Bystrov Reference Bystrov1957) and, unlike in Bradysaurus, there are flat areas often punctuated by round pits in between bosses. Often there are very fine radiating grooves emanating from the centre of the bone, the latter of which is usually demarcated by a prominent rounded boss on the larger specimens. The bosses themselves are not consistently circular in outline as seen in many other pareiasaur taxa, but are instead often subcircular, or even lozenge-shaped, imparting an inconsistent, pustule-like appearance to the bones of the skull roof.

The premaxilla (Figs 2, 4A, 11–13) of Deltavjatia is a triradiate structure, with a dorsal process that sutures with the nasal, a lateral process that sutures with the maxilla, and a posterior process on the palate that contacts the vomer. It also forms the anterior edge of the skull and, unlike the three premaxillary teeth illustrated by Lee (Reference Lee, Benton, Shishkin, Unwin and Kurochkin2000), the premaxillae of Deltavjatia bear two teeth (Fig. 12). The dorsal process, or internarial bar, formed by the premaxilla is relatively long and thin, so the nares face anteriorly rather than laterally. This bar is oriented vertically, consistent with the condition present in most other pareiasaurs. Dorsally, the dorsal process meets the nasal, slightly overlapping the medial process of this latter element. The flat palatal process of the premaxilla extends posteriorly to meet the anteriormost portion of the vomer. The prepalatal foramen is located posteriorly along the suture between the two premaxillae on the palatal surface (Fig. 12), near the confluence of the premaxillae and the vomers, as illustrated for Pareiasuchus nasicornis (Lee et al. Reference Lee, Gow and Kitching1997).

Posterolateral to the premaxilla, the maxilla (Figs 2, 4A, 12, 13, 19) is the other tooth-bearing bone of the skull. The maxillae of the larger specimens bear ten teeth, whereas there are fewer in the more juvenile specimens (PIN 2212/6 has nine teeth in the maxilla). The posterior spur of the maxilla and the anterior process of the quadratojugal do make contact for a short distance below the jugal, in contrast to previous descriptions of Deltavjatia (Lee Reference Lee1997b), although the prominence of this feature tends to vary between specimens and the contact is sometimes obscured by the ventral extension of the jugal. The maxilla is short compared to other closely related (non-pareiasaurian) parareptiles, the latter of which possess a maxilla that often extends past the posterior border of the orbit (see Emeroleter; Tsuji et al. Reference Tsuji, Müller and Reisz2012). In contrast, the maxilla of Deltavjatia forms a relatively long, vertical suture with the jugal, and barely extends past the anterior border of the orbit. The maxilla bears a series of foramina running parallel to the tooth row, the largest and most prominent of which is the anteriormost (Figs 4A, 13); the “anterolateral maxillary foramen” sensu Laurin & Reisz (Reference Laurin and Reisz1995), a feature that is a parareptilian synapomorphy. The element itself is relatively smooth and does not bear the sculpturing seen on the other dermal roof elements. Although there is a small boss on the maxilla close to the maxilla–lacrimal suture in some specimens, there is no prominent horn or boss as seen in Pareiasaurus. On the palatal surface, the maxilla forms the anterolateral border of the choana, and posteriorly it forms a medial suture with the palatine and the ectopterygoid.

In contrast to the closely related nyteroleters (Tsuji et al. Reference Tsuji, Müller and Reisz2012), the lacrimal (Figs 2, 4A, 11, 13, 18, 20A) of pareiasaurs participates in the posterior border of the external naris, perhaps related to their relatively shortened snout. The majority of the external surface of the lacrimal in Deltavjatia is overlapped by the dorsal process of the maxilla, thus appearing shorter externally than its full internal height. In the larger specimens, the lacrimal bears a very prominent boss just posterior to the external naris (Fig. 18). The dorsal edge of the element is bordered anteriorly by the nasal and posteriorly by the prefrontal. A process of the lacrimal extends posteroventrally, running just anterior to the ventrally directed process of the prefrontal. Posteriorly, the lacrimal bears one lacrimal foramen towards the upper half of the orbit (Fig. 13). Forming the anteroventral corner of the orbit, the ventral process of the lacrimal projects lightly posteriorly to meet the anterior process of the jugal.

The nasal (Figs 1, 2, 4A, 11, 18) meets its partner in a midline suture. It is a roughly rectangular bone that tapers towards the anterior end, where there is a slight medial excavation on the dorsal surface for the reception of the premaxilla. The nasal forms the majority of the dorsal edge of the naris. Posterior to the naris, the nasal sutures with the lacrimal anteriorly and the prefrontal posteriorly. The posterior border of the nasal is formed by a relatively straight transverse suture with the frontal. The nasal bears a large, rounded boss at the anterior end of the bone directly behind the naris.

The prefrontal (Figs 1, 2, 4A, 11, 13, 18, 20) forms the anterodorsal part of the orbital rim and, along with the nasal, makes up the dorsal border of the lacrimal. It contacts the nasal anteromedially, and the remainder of its medial limit is formed by a suture with the frontal. Posteriorly there is a short transverse suture with the postfrontal, excluding the frontal from the orbit. The ventral process of the prefrontal runs ventrally, posterior to the lacrimal, and forms a solid connection with the palate (Fig. 13). Sculpturing on the bone consists primarily of a large boss that lies close to the lateral edge (overtop of the orbit), and another, smaller boss close to the area where the prefrontal approaches the nasal-frontal suture.

The frontal (Figs 1, 2, 11, 18, 20) is a rectangular bone, meeting its pair in a straight midline suture. Anteriorly, it forms a roughly transverse suture with the nasal. Laterally, the element is bounded on its anterior half by the prefrontal and posteriorly by the postfrontal. In contrast to the condition seen in other parareptiles such as Procolophon (Laurin & Reisz Reference Laurin and Reisz1995), the frontal in pareiasaurs is excluded from making contact with the orbit. Posteriorly, the frontal appears to 'wrap around' the parietal; the fronto–parietal suture originates directed laterally, perpendicular to midline, but more laterally it is angled posterolaterally, forming a very thin tapering posterior process on each side (Fig. 11). This morphology contrasts with the reconstruction of Lee (Reference Lee, Benton, Shishkin, Unwin and Kurochkin2000), which shows this suture as straight and directed laterally. The dorsal surface of the frontal is characterised by a cluster of bosses towards the centre of the element, where the frontal, prefrontal and postfrontal meet. The rest of the bone is less sculptured, with the anterior portion bearing only shallow ridges that are directed antero-posteriorly.

The parietal (Figs 1, 2, 11, 18, 20) is a large bone that makes up most of the posterior section of the skull roof. The parietal, or pineal, foramen is contained entirely within the parietal and is positioned slightly anterior to the midpoint of the full antero–postero length of the bone, closer to the fronto–parietal suture. A posterolateral continuation of the fronto–parietal suture forms the suture with the postfrontal. The parietal also sutures laterally with the postorbital, and its posterolateral border is formed by a suture with the supratemporal. Posteriorly, the element is slightly embayed where it meets the single median postparietal. The parietal has several bosses clustered near the centre of the bone.

In some specimens of Deltavjatia the postparietal (Figs 1, 2, 4B, 11, 18, 20) features a small median boss close to the anterior edge of the bone where it sutures to the parietal (Fig. 11), but the remainder of the element is smooth and does not bear the sculpture characteristic of the other skull roof bones. As in all pareiasaurs (Lee Reference Lee1997a), the postparietal of Deltavjatia is a single, fused element, and is relatively large – over one-third of the length of the parietal itself. Ventrally, the postparietal receives the dorsal process of the supraoccipital, forming the main dorsal attachment surface of the braincase with the skull roof. This articulation takes the form of a short ventral process that is diamond-shaped in cross-section (Fig. 12). The postparietal forms a very short lateral suture with the large supratemporal, and the remainder of the lateral border is formed by the tabular or ‘supernumerary element’. The slightly concave posterior edge of the postparietal is not even with the posterior edge of the supratemporal, but is positioned more anteriorly, approximately halfway along the length of the supratemporal. Thus the median segment of the posterior edge of the skull roof is embayed significantly. This feature is so pronounced that in specimens where the braincase is present, some of the supraoccipital and the posterior part of the basioccipital can be seen in dorsal view (Fig. 20), an autapomorphy of the taxon.

The tabular (Figs 1, 2, 4B, 11, 18, 20) can be observed in all specimens of Deltavjatia where the posterior edge of the skull table is preserved, although in the larger specimens the sutures are more difficult to discern, and the element is often partially fused with the supratemporal and postparietal. In PIN 2212/6, there are clearly two smaller closely sutured, but still discrete elements on either side of the postparietal (Fig. 11). The postparietal and the tabular are strictly occipital elements in the majority of the more basal parareptiles (see Tsuji Reference Tsuji2006), which have been subsequently ‘dragged’ up onto, and integrated into, the skull roof in pareiasaurs. The extent of the tabular of Deltavjatia is typical for the majority of pareiasaurs, with the exception of Elginia, Arganaceras, and possibly Bunostegos, in which the tabulars form a median suture behind the postparietal, excluding the latter from the posterior edge of the skull table (Maxwell Reference Maxwell1991; Jalil & Janvier Reference Jalil and Janvier2005; Sidor et al. 2005).

The supratemporal (Figs 1, 2, 4A, 11, 13, 18–20) forms the posterolateral corner of the skull roof. It is a large element, bearing a very prominent round boss towards the posterior edge of the skull table, along with other smaller bosses clustered around it. The supratemporal is largely excluded from contact with the postparietal in Deltavjatia by the intervening tabular (Fig. 2), though a short suture between the former two is present in the majority of specimens. Anteriorly, the supratemporal forms a short suture with the postorbital, and ventrolaterally it sutures with the squamosal. The element also takes part in the attachment of the skull roof with the braincase, receiving the dorsal portion of the paroccipital process (Fig. 4B). The area of the supratemporal posterior to the large boss forms a noticeable posteroventrally-directed slope, and is mostly unsculptured, a morphology shared with the postparietal and tabular (Fig. 2).

The postfrontal (Figs 1, 2, 4A, 11, 13, 18, 20) of Deltavjatia is larger than reconstructed by Lee (Reference Lee, Benton, Shishkin, Unwin and Kurochkin2000), with its posterior edge extending even farther posteriorly. The postfrontal defines the posterodorsal edge of the orbit. It forms a short suture with the prefrontal anteriorly, and the majority of the median portion of the bone is delimited by the frontal. Posteromedially, the bone sutures with the parietal, and it is bounded laterally (and ventrally) by the postorbital. There is a large boss positioned at the centre of the bone along the edge of the orbit, with a number of smaller bosses clustered around the larger one (Fig. 18).

The postorbital (Figs 1, 2, 4A, 11, 13, 18, 20) is roughly triangular in shape, with its anterior edge forming the majority of the posterior rim of the orbit. It sutures with the jugal posteroventrally and the squamosal posteriorly. Sutures with the postfrontal and the frontal define the majority of the dorsal limit of the postorbital, but the element forms a short suture on its posteriormost edge with the supratemporal. The postfrontal bears a series of bosses along the orbital rim, with ridges and furrows radiating outwards from this point (Fig. 11).

The jugal (Figs 1, 2, 4A, 11, 13, 18, 20B) constitutes the majority of the ventral, and half of the posterior margin of the orbit. It is crescent-shaped in outline. The long anterior process thins anteriorly, and it forms a short suture with the lacrimal, with the majority of the anterior process bounded ventrally by the maxilla. Posteriorly, a gradually dorsally curving suture with the quadratojugal delineates the ventral edge of the bone. Posterodorsally, the jugal contacts the squamosal, and dorsally it forms an anteroventrally directed suture with the postorbital (Fig. 4A). There is a cluster of bosses located almost directly under the ventralmost extension of the orbit, a character shared with other pareiasaurs and Macroleter (Tsuji Reference Tsuji2006). In some specimens of Deltavjatia, the jugal forms a small portion of the ventral flange of the cheek, where the cheek flange meets the tooth-bearing region of the skull. This implies that the quadratojugal and the maxilla are separated, at least on the external surface of the skull in these specimens. Internally, however, these two latter bones indeed make contact, with the jugal forming a small overhang that obscures this suture externally (Fig. 4A).

The squamosal (Figs 1–4, 11–14, 18–20) is a large element, suturing dorsally and medially with the supratemporal, anteriorly with the postorbital and jugal, and ventrally with the quadratojugal. Internally, the squamosal sutures with the quadrate. A small posterior process of the pterygoid from the palate sutures with the squamosal at its dorsalmost contact with the quadrate. The squamosal makes up a large part of the cheek, and features a large oblong boss on the posterior edge towards the dorsal portion of the bone, where it sutures with the supratemporal, and a smaller, round boss just anterior to this (Fig. 19). Apart from these, the sculpturing of the squamosal consists primarily of ridges and furrows radiating outwards from the bosses towards the edges of the bone. Interiorly, the squamosal forms the ventral portion of the articulation with the paroccipital process of the braincase and is correspondingly thickened in this area, most likely to help reinforce the cheek. Although there is no apparent otic notch on the lateral (external) surface of the skull in pareiasaurs, in posterior view a small yet clear ‘notch’ facing internally can be seen enclosed primarily within the squamosal (Fig. 4B). This feature appears to be homologous to the otic notch that holds the large tympanum of the nycteroleters (Müller & Tsuji Reference Müller and Tsuji2007; Tsuji et al. Reference Tsuji, Müller and Reisz2012) but, corresponding to the more heavily ossified condition of the pareiasaur skull, it has been shifted considerably into what would have been a more protected position. The relative area occupied by this structure is smaller than the condition seen in the nycteroleters. The position of the otic notch is such that a stapes and cartilaginous extrastapes would have been directed slightly dorsolaterally from the braincase (see further discussion below).

The quadratojugal (Figs 1–4, 11–14, 18–20) comprises the entirety of the ventral margin of the cheek. As in all pareiasaurs, the element forms a prominent cheek flange, extending well below the level of the tooth row. In undistorted specimens, the anterior edge of this flange forms an angle of approximately 60 degrees with the tooth row. Although this angle is considerably lower than in pareiasaurs such as Pareiasaurus, the cheek flange of Deltavjatia is more pronounced than in Bradysaurus (Lee Reference Lee1997b). The quadratojugal delineates most of the posterior boundary of the jugal, and dorsally it forms a relatively straight horizontal suture with the squamosal. The dorsal border of the quadratojugal is approximately even with the ventral rim of the orbit. The stated autapomorphy of a long low boss on the quadratojugal (Lee Reference Lee1997b) is not consistently present, at least in the smaller two specimens (PIN 2212/6, ROM 44764). Instead, the quadratojugal ornamentation consists of two larger anterior conical spikes, with smaller bosses further posteriorly along the posterior edge of the element (Figs 4A, 20).

2.2.2. Palate

The palate of pareiasaurs is fairly conservative and, indeed, the palate of Deltavjatia differs little from the typical pattern. It is characterised by tall, denticle-bearing ridges and a reduced transverse flange of the pterygoid. Unlike in the nycteroleters, the palate, and especially the transverse flange of the pterygoid, is raised considerably above the level of the tooth row.

The vomer (Figs 3, 10, 12) of pareiasaurs is unique, typified by what is termed the ‘alar flange’ (Damiani & Modesto Reference Damiani and Modesto2001); a rounded lateral process defining the medial edge of the choana, resulting in a choana that is kidney-shaped in outline. It has been suggested that the vomer has a thin medial dorsal flange that forms a bony septum in the nasal chamber (Damiani & Modesto Reference Damiani and Modesto2001), but unfortunately none of the specimens of Deltavjatia examined for this study are preserved in a manner in which this feature could be observed.

Anteriorly, the vomer meets the premaxilla. The articulating surface of this contact is not straightforward, however. As described for an unidentified pareiasaur from the Tropidostoma Assemblage Zone of the Karoo (Damiani & Modesto Reference Damiani and Modesto2001), a channel or groove separates two articulating surfaces, with lateral processes that are larger than the median ones (Fig. 10). Posterior to the articulation with the premaxilla, a row of denticles runs along the median edge of the bone. The anteriormost denticle is the largest and is slightly recurved, and the size of the denticles decreases as the row continues posteriorly. At the level of the anteriormost extension of the vomer–palatine suture (slightly anterior to the posterior edge of the choana), the single row splits into two parallel rows. The vomer has very little sutural contact with the palatine, with this contact almost precluded by the long posterolaterally-directed suture with the pterygoid (Fig. 10).

The palatine (Figs 3, 10, 12) forms the posterolateral, the entire posterior edge, and a small portion of the medial edge of the choana. Laterally, the bone meets the maxilla in a gently sloping posterolaterally-directed suture. The posterolateral edge of the palatine sutures with the ectopterygoid, but this contact is interrupted by a large foramen, the ‘foramen palatinum posterius’ of Lee (Reference Lee1995b). This foramen is contained entirely within the suture between the palatine and the ectopterygoid (Fig. 10). It may be homologous to the ‘suborbital foramen’ of the nycteroleters (Tsuji et al. Reference Tsuji, Müller and Reisz2012); however, in nycteroleters and most other parareptiles, this foramen is contained between the ectopterygoid, maxilla and palatine. Thus, the term ‘foramen palatinum posterius’ is maintained here for pareiasaurs. Kordikova & Khlyupin (Reference Kordikova and Khlyupin2001) have suggested that the foramen palatinum posterius is proportionally much larger in neonates, almost completely separating the palatine from the maxilla. The palatine also sutures with the pterygoid posteromedially. A double row of denticles, which continues from the pterygoid, extends anterolaterally, reaching almost to the posterior edge of the choana.

The ectopterygoid (Figs 3, 10, 12) is a relatively small, roughly triangular element. The ventral surface of the bone is smooth and featureless. In ventral view, anteriorly the ectopterygoid sutures with the palatine, with the suture interrupted by the foramen palatinum posterius. The ectopterygoid sutures laterally with the maxilla, continuing the palatine–maxilla suture, and it also forms a short suture with the pterygoid posteriorly. In P. nasicornis, the ectopterygoid is reconstructed as forming a short suture with the quadratojugal (though in the description it is described as an ectopterygoid–jugal suture) (Lee et al. Reference Lee, Gow and Kitching1997), but this condition appears to be variable amongst pareiasaur taxa (Lee Reference Lee1997a). There is no contact between the ectopterygoid and the jugal or quadratojugal in the studied specimens of Deltavjatia. The lateral limit of the ectopterygoid is formed entirely by the maxilla (Fig. 12).

The pterygoid (Figs 3, 10, 12) is a large and complex bone, and is the dominant element of the palate. It is basically triangular, with an anterior process that terminates in a transverse suture with the vomer. Posterior to the suture with the vomer, the pterygoid forms a posterolaterally-directed suture with the palatine, which continues as a suture with the ectopterygoid. The pterygoid also takes part in the basipterygoid articulation, in which the basipterygoid process of the braincase fuses to the palate in mature individuals. The interpterygoid vacuity of Deltavjatia is very short and is rounded at its anterior end, more similar to the condition seen in the basal Bradysaurus than in the derived Anthodon (Lee Reference Lee1997a). The pterygoid also forms a long anterolateral suture with the quadrate, and may also form a short suture with the squamosal (Figs 3, 12).

Multiple rows of denticles run across the pterygoid. The double row of denticles continues posteriorly from the vomer and lines the edges of the interpterygoid vacuity. Another row of denticles begins approximately midway along the pterygoid–palatine suture. This set of denticles begins as a single row, but splits into two rows as it approaches the palatine. This double row, as with the other double rows on the palate of Deltavjatia, remains parallel throughout its length (Figs 3, 12, 26).

The transverse flange of the pterygoid is only slightly recurved. In outgroup parareptilian taxa, it generally points directly laterally and very closely approaches the cheek (Laurin & Reisz Reference Laurin and Reisz1995; see Macroleter, Tsuji Reference Tsuji2006). In Deltavjatia, as in other pareiasaurs, the transverse flange is directed anterolaterally and does not closely approach the cheek (Lee Reference Lee1997b). The transverse flange also bears a row of large denticles on a raised ridge on the ventral surface of the structure (Fig. 3).

No epipterygoids were found in the specimens of Deltavjatia examined in this study. The element is figured as a very small process running vertically from the cranial process of the pterygoid in Pareiasaurus serridens (Boonstra Reference Boonstra1930).

The quadrate (Figs 3, 4B, 10, 12) connects the posterior portion of the palate with the cheek and forms the upper half of the jaw articulation. The element is basically triangular and sutures with the pterygoid anteromedially, the quadratojugal ventrally, and the squamosal posteriorly. The quadrate forms an extensive suture with the quadrate flange of the pterygoid. This suture extends dorsally almost to the skull roof, close to where the paroccipital process of the braincase articulates with the squamosal and supratemporal (Fig. 12). The exact delineation of the suture between the quadrate and the pterygoid is difficult to trace in most specimens, however. On its posterior and posterolateral margins, the quadrate sutures with the quadratojugal. Further dorsally, a small quadrate foramen is found at the convergence between the quadrate, squamosal and quadratojugal (Fig. 4B). The suture between the squamosal and quadrate defines the posterior border of the otic notch in Deltavjatia.

The jaw articulation in Deltavjatia is medio-laterally broad and consists of a lateral and a medial condyle separated by a shallow groove (Fig. 12). The articulation occurs well anterior to the occiput – almost to the level of the basipterygoid processes – and ventral to the tooth row, a condition that is concordant with the herbivorous diet of pareiasaurs (Lee et al. Reference Lee, Gow and Kitching1997).

2.2.3. Braincase

The well-ossified condition of the pareiasaur skull applies in particular to the braincase, making it difficult, if not impossible, to discern the sutures between the constituent elements. However, the availability of well-preserved sub-adult Deltavjatia material allows a closer examination of the individual elements of the braincase. PIN 2212/6 includes an isolated braincase with partially disarticulated exoccipitals, allowing a detailed description of the braincase from all angles, in addition to insight into the structure of the individual elements. It is important to note, however, that this specimen is sub-adult, and thus certain features seen here may not be present in fully-grown specimens and vice-versa.

In pareiasaurs, the parasphenoid and the basisphenoid are fused into a single indistinguishable element, termed the parabasisphenoid (Figs 3, 10, 12, 16). Even in the relatively unfused braincase of the juvenile PIN 2212/6 (Fig. 16) the two elements cannot be distinguished from one another, indicating that this fusion probably occurred early in development. The parabasisphenoid has an hourglass silhouette in ventral view, with slightly flared anterior and posterior ends and a slightly constricted middle, though this constriction is less pronounced than seen in the outgroup taxa and other basal pareiasaurs such as Bradysaurus (Lee Reference Lee1997b). In pareiasaurs such as Scutosaurus, the hourglass shape is emphasised by the presence of large, paired basal tubera close to the suture with the basioccipital (Lee Reference Lee, Benton, Shishkin, Unwin and Kurochkin2000); however, these are absent in Deltavjatia. Even in the largest specimens, the ventral surface of the basisphenoid is smooth.

The basipterygoid processes project anterolaterally from the body of the parabasisphenoid, joining with the corresponding processes of the pterygoid on the palate. This joint between the braincase and the palate is fused and immobile, with no visible suture between the two. This morphology can be seen in all larger specimens of Deltavjatia, and is a pareiasaurian synapomorphy (Lee Reference Lee1997a), although an immobile basicranial articulation is also present in Macroleter (Tsuji Reference Tsuji2006) and Lanthanosuchus (deBraga & Reisz Reference deBraga and Reisz1996) within parareptiles. In the more juvenile PIN 2212/6, however, this fusion has not yet occurred, and the braincase can be completely removed from the skull. The articulating surfaces are slightly distorted and do not retain their original morphology. The ventral surface of the parabasisphenoid is relatively smooth, without the basal tubera mentioned above, with a noticeable circular indentation centered just posterior to the basicranial processes (Fig. 3). In PIN 2212/6, two small foramina can be seen towards the anterior portion of this pit (Fig. 16D), seemingly analogous to the “anterior foramen” identified by Lee et al. (Reference Lee, Gow and Kitching1997) on the ventral surface of the basisphenoid of Pareiasuchus nasicornis. This feature also appears to be present in well-preserved adult specimens, located far anteriorly on the basisphenoid, close to the cultriform process. There are two additional foramina just anterior to the suture with the basioccipital, situated within very deep, posteriorly directed grooves (Fig. 16B), possibly equivalent to the ‘posterior foramen’ of Lee et al. (Reference Lee, Gow and Kitching1997). However, these foramina are not present in larger specimens of Deltavjatia.

Due to the disarticulation of the dorsal elements of the braincase, the dorsal surface of the parabasisphenoid is exposed in PIN 2212/6. Just posterior to the basipterygoid processes, the median dorsal body, or parasphenoid rostrum, can be seen rising dorsally from the base of the parasphenoid, with a very small cultriform process projecting anteriorly (Fig. 16A, B, C). In Deltavjatia, the cultriform process is present, and extends at least as far anteriorly as the full length of the interpterygoid vacuity (Fig. 10). Its full anterior extent cannot be assessed from the specimens available, however, and in the juvenile braincase (PIN 2212/6), this feature is either very small or has been broken off. The sella turcica, a deep pit in which the pituitary body would have lain, can be seen immediately posterior to the basisphenoid rostrum (Fig. 16C). Just lateral to the sella turcica are the foramen and groove for the vidian canal (Fig. 16B, C).

The prootic (Figs 16, 17) and the opisthotic project dorsolaterally from the midline of the braincase and form the paroccipital processes, which connects the braincase with the skull roof. The prootic is sutured to the main body of the parabasisphenoid just posterior to the sella turcica. Anteriorly, the prootic is slightly embayed to accommodate the trigeminal nerve (V) (Fig. 16A, B). A small foramen for the facial nerve (VII) is present on the prootic just posterior to this embayment (Fig. 16B). The prootic then continues posteriorly, forming the anterior portion of the dorsal border of the foramen ovalis. Dorsally, it appears as if the two prootics meet at the midline dorsal to the parabasisphenoid, a morphology that differs from the condition portrayed in the procolophonid Leptopleuron, where the prootics do not meet and the parabasisphenoid is exposed dorsally (Spencer Reference Spencer2000). At its dorsalmost margin, the prootic contacts the supraoccipital. The exact length and orientation of this suture is not clear, however, as the entire suture between the prootic and the opisthotic cannot be discerned in PIN 2212/6. More posteriorly, the prootic then merges with the opisthotic to form the paroccipital process. The paroccipital process comprises the articulation of the braincase with the skull roof, but the prootic forms only the anteriormost portion of the paroccipital process and thus does not take part in the articulation.

The opisthotic (Figs 3, 4B, 16) originates posteriorly in a suture with the exoccipital, and these two bones together support the posterior part of the supraoccipital. Posteriorly, the opisthotic and the prootic form the paroccipital process, but it is the opisthotic alone that comprises the lateralmost extent of this structure, forming an articulation with the squamosal and supratemporal in the skull roof. In occipital view, the paroccipital process is u-shaped, dipping slightly ventrally before curving dorsally to fuse with the skull roof in the dorsolateral corner. The braincase in pareiasaurs is firmly fused with the skull, at both the palate and the paroccipital processes.

Anteriorly, the opisthotic contacts the parabasisphenoid via a ventral projection that delineates two large foramina (Fig. 16B). The larger, more anterior foramen can be identified as the fenestra ovalis. A semi-circular stapedial recess, which would have accommodated the footplate of the small stapes (described below), is present in the posterior edge of this larger fenestra (Fig. 16B). This feature does not appear in the braincase of Pareiasuchus nasicornis (Lee et al. Reference Lee, Gow and Kitching1997). The foramen ovalis is very open in the juvenile PIN 2212/6, but in KPM 232 there is some indication of a further division of the fenestra ovalis into a dorsal and ventral section, though they are not fully separate (Fig. 5). The lower of these can be interpreted as a ‘pressure relief window’ sensu Clack (Reference Clack1997), one of the indications of the presence of tympanic hearing, though this feature is not clearly defined on both sides of the skull. The smaller and more posterior of the two foramina is the foramen jugulare anterius of Haughton (Reference Haughton1929) and Lee et al. (Reference Lee, Gow and Kitching1997), a feature not present in the considerably less well-ossified procolophonid Leptopleuron (Spencer Reference Spencer2000).

The internal structure of the opisthotic in PIN 2212/6 consists of a very open network of pits (Fig. 16A), indicating that the otic capsule was not fully ossified in this specimen. A well-ossified state of the otic capsule is present in Macroleter (Müller & Tsuji Reference Müller and Tsuji2007), and the otic capsule of pareiasaurs is considered to be well-ossified (Lee Reference Lee1997a). The immaturity of PIN 2212/6 is likely the reason this region appears to be unossified in this case.

The dorsal exposure of the supraoccipital (Figs 4B, 17) is one of the autapomorphies of Deltavjatia. It is apparent, however, that this morphology is the result of the severe embayment of the posterior skull table, rather than the placement or morphology of the supraoccipital itself. In PIN 2212/6, the supraoccipital has been separated from the rest of the braincase and skull roof (Fig. 17) and the articular surface of the supraoccipital with this latter structure can be seen in the isolated element (Fig. 17A). The contact is roughly diamond-shaped, with the corners oriented along the axes of the skull. The ventral process of the supraoccipital then extends downwards as a thick column, before it flares to form the roof of the braincase and the dorsal edge of the foramen magnum. A tall, columnar surpaoccipital is one of the synapomorphies of pareiasaurs (Lee, Reference Lee1995b, Reference Lee1997a). Anteriorly, this column has a flattened indentation, in which is situated an anterior crest (Fig. 17A), and there is a small raised ridge on the posterior surface (Fig. 17B).

The exoccipital (Reference Benton, Newell, Khlyupin, Shumov, Price and KurkinFigs 4B, 16) is a paired element that projects dorsally and slightly laterally from the basioccipital. In pareiasaurs, it forms the dorsal half of the occipital condyle of the braincase and the lateral margins of the foramen magnum. The exoccipital consists of a basal process that sutures with the basioccipital, a vertical stem whose medial border forms the lateral edge of the foramen magnum, and a dorsal portion that connects with the opisthotic anteriorly and supports the supraoccipital dorsally. This dorsal portion consists of a medial and a lateral process. The medial process forms a significant part of the dorsal border of the foramen magnum, but does not extend to meet its pair at the midline (Fig. 4B). The long lateral process underlies the paroccipital process for a considerable distance, a character typical for pareiasaurs (Lee Reference Lee1995b). There is a series of foramina at the base of the dorsal process of the exoccipitals through which the branches of cranial nerve XII (hypoglossal) would have passed (Figs 16A, B). The base of the exoccipital meets its pair medially above the basioccipital, excluding this latter structure from forming part of the rim of the foramen magnum (Fig. 16), seemingly discordant with the condition described for Bradysaurus (Haughton Reference Haughton1929). In larger individuals of Deltavjatia, the exoccipital and basioccipital fuse indistinguishably.

The basiocciptial (Figs 4B, 16) constitutes the ventral half of the occipital condyle and the posteriormost part of the ventral surface of the braincase. In the larger specimens of Deltavjatia, the basioccipital has fused indistinguishably with the parabasisphenoid to form a single solid structure; but in the juvenile PIN 2212/6, the suture between the basioccipital and the parabasisphenoid can be discerned (Fig. 16D). The basioccipital makes up slightly more than half of the occipital condyle, with the fused exoccipitals forming the dorsal portion. The fusion of the exoccipitals excludes the basiocciptal from participation in the foramen magnum. The condyle itself is highly concave, with a relatively sharp (ie not swollen) rim in the specimens in which it is adequately preserved. This more closely resembles the condition described for Embrithosaurus than that for Bradysaurus, in which the condyle is not as concave and the rim is rounded (Haughton Reference Haughton1929).

The sphenethmoid or interorbital septum is acknowledged to be present in pareiasaurs (Romer Reference Romer1956). Primitively in tetrapods, the sphenethmoid is a large, complex bone (Clack Reference Clack2002). In amniotes, it is a solid plate that rises dorsally from the anterior portion of the parasphenoid (the cultriform process) to reach the frontal, and divides the anterior section of the skull into right and left halves (Romer Reference Romer1956). Its presence in the specimens of Deltavjatia examined in this study, however, is not certain. The area above the cultriform process (between the palate and the skull roof) is rarely prepared in complete skulls, and there is no evidence of its presence in PIN 2212/6, the only specimen in which this area is visible. The extremely short cultriform process in this specimen does not support a sphenethmoid. It is reasonable to infer, however, that PIN 2212/6 is not skeletally mature and thus the sphenethmoid is not yet ossified. This structure was certainly preformed in cartilage. Known pareiasaur sphenethmoids are described in well-ossified and presumably fully-grown individuals. The sphenethmoid of one such individual of Embrithosaurus from the Natural History Museum in London has been partially sectioned (Watson Reference Watson1914; Haughton Reference Haughton1929), and a specimen of Pareiasaurus is illustrated by Boonstra (Reference Boonstra1934a). A sphenethmoid is also present in a specimen of Bunostegos akokanensis (Tsuji et al. Reference Tsuji, Sidor, Steyer, Smith, Tabor and Ide2013).

The stapes (Fig. 21) of early tetrapods tends to be robust, with a role in helping to brace the braincase against the skull roof (Clack Reference Clack2002). In fully terrestrial animals with an acute sense of hearing, however, the structure of the stapes is quite different; instead becoming a small, relatively gracile bone that performs functions related to hearing with a tympanic ear (Clack Reference Clack2002; Müller & Tsuji Reference Müller and Tsuji2007). The stapes of pareiasaurs has been described and figured in two taxa – Arganaceras vacanti (Jalil & Janvier Reference Jalil and Janvier2005) and Scutosaurus karpinskii (Lee Reference Lee1995a) – but the morphology described varies considerably between these two accounts. Considering that both elements were found in isolation and that both of their morphologies are highly aberrant from the stapedial morphology seen in other parareptiles (Carroll & Lindsay Reference Carroll and Lindsay1985; Tsuji Reference Tsuji2006; Tsuji & Müller 2007), it is doubtful that either of these structures can be definitively identified as a stapes.

An element identified here as a stapes has been found in a specimen of Deltavjatia. In SMNS 58317, this element is visible on both sides of the skull in ventral view, and while the left element has been maintained in place, the right has been removed and prepared so it is visible in multiple views (Fig. 21). The element is approximately 1.5 cm in length; very small in comparison with the overall skull length of 25 cm. Proximally, the bone is flared, and there is a large stapedial foramen visible, best seen in ventral view. The proximal end – the footplate – possesses two articular surfaces separated by a sulcus. Distally, the stapes tapers to a point, imparting a triangular shape to the element, slightly longer than wide. This form is very similar to the stapes seen in other derived parareptile taxa such as Procolophon trigoniceps (Carroll & Lindsay Reference Carroll and Lindsay1985) and, in particular, Macroleter poezicus (Tsuji Reference Tsuji2006).

In addition to morphological similarity to the stapes of closely related parareptile groups, the stapes of SMNS 58317 is preserved in the area of the foramen ovalis (see description above) on both sides of the skull. It is clear that the stapes was only very loosely associated with the skull, and would have been preserved only in very well-preserved and undisturbed skeletons.

2.2.4. Mandible

The mandible of Deltavjatia does not deviate significantly from the typical pareiasaurian morphology. It is distinguished by a prominent angular boss extending from the ventral edge of the jaw that extends into a small, pointed horn, and by the single row of leaf-shaped teeth that mirror those on the maxilla. The lower jaw lacks the sculpturing typically seen on the external surface of the skull roof, but does bear a roughened texture on the external, or lateral, surface of the elements. It is composed of a number of overlapping sheet-like bones.

The dentary (Figs 4A, 14, 20B, 22B) is the tooth-bearing bone of the mandible and forms the majority of the anterolateral margin of this structure. It is exposed more on the external surface than the internal, where it is largely overlapped by the splenial. On the external surface of the mandible, the dentary meets the surangular posteriorly and forms a suture with the angular posteroventrally. Internally the dentary forms only the upper half of the mandibular symphysis, above the foramen intermandibularis medius (sensu Lee et al. Reference Lee, Gow and Kitching1997). The splenial and the coronoid overlap the dentary, ventrally and dorsally respectively. In the two specimens with isolated mandibles, there are 11 alveoli present in each dentary (Figs 14, 22). The bone has a number of small foramina on its external surface, but is otherwise unsculptured.

The splenial (Figs 3, 4B, 14, 22) is a platey bone that forms the anterior part of the medial surface of the jaw and, in contrast with most other parareptiles (Laurin & Reisz Reference Laurin and Reisz1995) such as Macroleter (Tsuji Reference Tsuji2006), the element does take part in the symphysis of the mandible (Figs 4B, 22B). The anterior part of the bone is slightly forked, with the upper and lower prongs forming the borders of a large foramen intermandibularis medius (Fig. 4B). The lower prong is significantly more robust than the upper. It is this ventralmost part of the splenial that meets its pair at the midline, taking part in the symphysis, as it does in stem amniotes such as Diadectes (Laurin & Reisz Reference Laurin and Reisz1995). The intermandibular foramen is formed at the confluence of the spenial, angular and the prearticular. It is almost directly dorsal to the angular boss.

Dorsal to the splenial, on the internal surface of the mandible, lies a single coronoid bone (Figs 14, 22). This element forms the dorsalmost extent of the mandible in the form of a relatively low coronoid process, which projects dorsally only to the height of the tooth crowns (Figs 14B, 22B). The anterior process of the bone does not extend much beyond the 5th tooth position. The coronoid forms the anterior rim of the adductor fossa, meeting the surangular on the external (lateral) side and the prearticular on the medial side of this structure. Anteriorly, the coronoid overlaps the posterodorsal process of the dentary and its anteriormost extent forms a short contact with the splenial.

The prearticular (Figs 14, 22) is also located on the inner (internal) surface of the mandible. The element overlaps the angular ventrally and the coronoid dorsally, where it forms the ventromedial border of the adductor fossa. Because the prearticular is very broad dorsoventrally, the adductor fossa of Deltavjatia faces predominantly dorsally. Anteroventrally, the prearticular is overlapped by the splenial.

The surangular (Figs 14, 22) lies dorsal to the angular and forms the majority of the posterodorsal portion of the mandible (on the external surface). It lies posterior to the dentary, extending to cover most of the articular laterally. It does not take part in the retroarticular process.

The angular (Figs 3, 4, 14, 20, 22) forms the ventralmost part of the mandible, and can be seen in both medial and lateral views. Medially, the majority of the dorsal border of the angular is formed by a suture with the prearticular, and it meets the splenial anteriorly and the articular posteriorly. Externally (laterally), the surangular forms the majority of its dorsal border, and it meets the dentary anterodorsally. One of the most distinctive synapomorphies of pareiasaurs is the prominent boss towards the anterior end of the angular, which often takes the form of an obvious horn. In the largest specimens of Deltavjatia, the boss is large and pointed, but still more gracile than in pareiasaurs such as Bradysaurus. In the smaller specimens of Deltavjatia, the boss is rather small, low, and laterally compressed (Figs 14, 20B). The dimensions and prominence of the angular boss increase with increasing size and, presumably, age of the individuals (Fig. 27; see also discussion below).

Figure 27

The growth and development of the angular boss in Deltavjatia rossicus (boss indicated by arrow). Specimens, from top to bottom: ROM 44764, PIN 2212/6, SMNS 58317, KPM 232.

The articular (Figs 3, 14B, 22B) has very little lateral or medial exposure and is overlapped by the prearticular internally and the angular and surangular externally. The majority of the visible part of the bone is exposed in dorsal aspect in the form of the glenoid fossa for the articulation with the quadrate of the skull. This articulation surface is concave, with a central raised and rounded anteroposteriorly-directed ridge dividing the articulation surface into two sections, the more medial of which is more concave. The articulation surface itself also tilts ventrally on its posterior end. The posterior and ventral edge of the articular forms the entirety of the retroarticular process. In Deltavjatia, and indeed in all pareiasaurs, this feature is very small, considerably smaller than in other parareptiles such as bolosaurids (Reisz et al. Reference Reisz, Müller, Tsuji and Scott2007), and bears a small slightly upturned flange (termed a ‘dorsal lump’ by Lee Reference Lee1995b) at its posteriormost end.

2.2.5. Hyoid

In general, the hyoid apparatus of parareptiles tends to be preferentially preserved (Carroll & Lindsay Reference Carroll and Lindsay1985; Reisz & Scott Reference Reisz and Scott2002; Tsuji Reference Tsuji2006). It is also known from other pareiasaurs (Lee et al. Reference Lee, Gow and Kitching1997); however, none are preserved in the specimens of Deltavjatia available for study.

2.2.6. Dentition

The teeth of Deltavjatia are leaf shaped; labio-lingually compressed, with a number of rounded cusps evenly distributed across the crown. There are 9–11 cusps in the teeth of the maxilla as well as the dentary. Contrary to what is described by Lee (Reference Lee, Benton, Shishkin, Unwin and Kurochkin2000), some of the teeth have a cingulum on the lingual surface of the tooth, the largest of which also bear small cusps (Fig. 15B). Lee has asserted that the unerupted teeth of pareiasaurs are simple and conical in shape (Lee Reference Lee1997a), but barely-erupted replacement teeth in specimens of Deltavjatia already bear a nearly complete set of cusps and their crowns are still expanded (Fig. 15) (see also Kordikova & Khlyupin Reference Kordikova and Khlyupin2001). The upper jaw has 12 alveoli (SMNS uncat), while the mandible has 11.

2.3.1. Vertebral column and ribs

The presacral vertebral count of basal pareiasaurs is lower than for those that are more derived, while the caudal vertebral count is generally higher in basal forms (Lee Reference Lee1997a). There are 17 presacral vertebrae in KPM 232, but there may be as many as 18 in UMZC T1321. There are four or five sacral vertebrae (see discussion below). The last part of the tail in KMP 232 has been grafted on from another specimen, so an exact number of caudal vertebrae cannot be determined, but it is obvious that the tail is not short. Although gastralia have been identified in Emeroleter (Tsuji et al. Reference Tsuji, Müller and Reisz2012), there is no evidence of these structures in specimens of Deltavjatia, nor in the nycteroleter Macroleter (Tsuji, pers. obs.).

Whereas they are obscured or absent in most specimens of Deltavjatia, the elements of the atlas–axis complex (Fig. 24) can be seen in UMCZ T1321 in dorsal and lateral views (Fig. 24). Although a proatlas was not found by Boonstra (Reference Boonstra1934a) in his study of Karoo pareiasaurs, an element here identified provisionally as a proatlas has drifted lateral to its natural position. The two more lateral elements are the paired atlas neural arches, which are almost square in outline when viewed in dorsal aspect. Their broad flat surfaces are tilted laterally, such that they form long, antero-posteriorly oriented ‘spines’. In lateral view, a large anterior element is identified as the atlantal intercentrum, but presumably most of the dorsal part of this structure belongs to the atlantal centrum, which is the condition present in most basal amniotes (Sumida et al. Reference Sumida, Lombard and Berman1992). The bone surface is badly preserved in this area, however, and the delineation between these two elements is not clear. Posterior to the atlantal intercentrum is a semi-lunate axial intercentrum, followed posteriorly by the axial (pleuro-) centrum, which appears to be fused with the neural arch of this element (Fig. 24B). The axial neural arch is a tall, prominent structure, whose dorsalmost tip appears to be slightly flattened, perhaps to accommodate the anteriormost osteoderms (Fig. 24A). The bone surface of these elements is not well preserved; therefore it is not possible to describe the nature of the rib articulations.

The remaining presacral vertebrae of Deltavjatia are typical for pareiasaurs. They have wide, swollen neural arches – more specifically, the dorsal edge of the postzygopophysis forms a tall, rounded arch that adjoins the neural spine. The typical dorsal vertebra has transverse processes that are approximately three times as wide as the centrum length, a condition that imparts a much broader appearance than that seen in most other basal amniotes (Sumida & Modesto 2001). The neural arches also possess a significant furrow on their posterior-facing surface, delimiting the high swollen arch from a surrounding flattened area, below which sits the flat, horizontally-oriented posterior zygopophysis for articulation with the anterior zygapophysis of the succeeding vertebra. The neural spines in Deltavjatia are short compared to other pareiasaurs such as Scutosaurus and Bradysaurus, which in well-preserved specimens extend dorsally to a height approaching that of the centrum and increase in height posteriorly (Boonstra Reference Boonstra1934b). In the articulated specimens of Deltavjatia, the neural spines are often broken. However, as evidenced by the presence of semi-articulated osteoderms, in some cases the spines are preserved in their entirety. It is clear that the neural spines never reached the height of the centra, increasing to a height of approximately 2 cm in KPM 232 as they approach the sacrum. Broken surfaces indicate that the neural spines were diamond in cross-section, and were considerably more robust posteriorly than anteriorly.

The centra are almost completely circular in outline, significantly antero-posteriorly shorter than wide, and amphicoelous. The transverse process for the articulation of the ribs consists of a thin, dorsally oriented flange that approaches the height of the centrum and is hosted by both the centrum and neural arch. The height of transverse processes increases considerably more posteriorly in the vertebral column as the size of the ribs increases (Fig. 6).

In pareiasaurs, the presacral (dorsal) ribs (Figs 6, 8, 9) form a continuous cage around the torso for the entire length of the presacral vertebral column, unlike the condition seen in the closely related nycteroleters, which do not bear ribs on their posterior dorsal vertebrae (Tsuji et al. Reference Tsuji, Müller and Reisz2012). In Deltavjatia, the presacral ribs are slightly flattened and curve forward slightly. The ribs are single-headed (holocephalus), and articulate with the extensive transverse processes (rib facets) of the vertebrae. The rib heads are very broad, but antero-posteriorly very restricted, reaching their peak size midway along the presacral vertebral column. The main body of the ribs is long and gently curved in the first two-thirds of the presacral vertebral column, and then decreases in length and curves more strongly ventrally and internally as they approach the sacrum. Overall, the ribs are slightly flattened and often possess a median groove on their posterior surface. The ribs of Deltavjatia are also widely separated, unlike the condition seen in Pumiliopareia, in which there is little separation between successive ribs (Lee Reference Lee1997b).

The sacral vertebrae (Fig. 6) in Deltavjatia have fused into a single structure. Judging from the number of sacral ribs involved in the articulation with the sacrum, this structure is likely composed of four vertebrae. The top of the neural spine expands significantly in the sacral vertebrae, providing a larger area for articulation for the osteoderms. Starting with first sacral vertebra, neural arches are greatly reduced, and the sacral ribs are fused to the body (Fig. 6).

The presence of three or more sacral ribs (Fig. 6) is a pareiasaurian synapomorphy, with some taxa in possession of as many as five. Deltavjatia appears to have four, with a fifth that, while it does not directly contact the innominate, is curved anteriorly and is fused to the rib of the vertebra immediately anterior (4th sacral vertebra), making it functionally supportive. The first two sacral ribs are the most robust, with the second slightly more so than the first. The third is relatively straight, while the fourth tapers at the midpoint, subsequent to which it flares broadly where it fuses to the sacrum (Fig. 6).

There are, unfortunately, no specimens of Deltavjatia available or prepared that preserve the complete set of caudal vertebrae (Fig. 6). Only KPM 232 shows what appears to be a long tail, but the majority of it belongs to another specimen (Fig. 6, grey). Despite this, some information can still be gleaned about the anatomy from this specimen (Fig. 6). The caudal vertebrae are considerably smaller than the presacral vertebrae. The swelling of the neural arches seen in the presacral vertebrae is absent. In the first few vertebrae posterior to the sacrum, the ribs are fused to the body of the vertebra and project laterally. These vertebrae also retain a flattened cap on the top of the neural spine for the articulation of osteoderms, though after the first five, the neural spine is much reduced. The more posterior caudal vertebrae have haemal arches, which articulate with the ventralmost portion of the centrum. The centra themselves appear to be more laterally compressed and slightly lengthened, such that the caudal vertebrae are taller than wide. There is a minimum of 26 caudal vertebrae preserved in KPM 232 (Fig. 6). Lee (Reference Lee, Benton, Shishkin, Unwin and Kurochkin2000) also mentions a long tail in the Deltavjatia. This count is lower than in Bradysaurus, but higher than in other, more derived pareiasaurs such as Scutosaurus or Anthodon (Lee Reference Lee1997b).

2.3.2. Pectoral girdle and forelimb

The scapulocoracoid (Fig. 6) of Deltavjatia is more gracile in appearance than that of the more basal pareiasaurs, but differs significantly from the condition seen the derived dwarf taxa (Boonstra Reference Boonstra1932; Lee Reference Lee1997a). The scapular blade is relatively thin, flaring only slightly at its dorsal end, and it is approximately two times as wide at the dorsal end as at the most constricted point. The blade is flattened throughout its length, however, unlike the cylindrical cross section of the blade seen among the more derived pareiasaurs such as Anthodon (Lee Reference Lee1997b). In Deltavjatia, the shape of scapular blade also contrasts with the relatively rectangular, unflaring outline of the more primitive pareiasaurs (Lee Reference Lee1997b). At the base of the scapular blade, a small flat anterolaterally-facing spur can be seen, identified as an ‘acromion-like’ process (Fig. 6), since it is not necessarily homologous with this structure in turtles (Lee Reference Lee1998). Only the dorsalmost portion of the glenoid can be seen in the specimens, with the majority of the coracoid portion remaining unprepared below the matrix of multiple specimens. In pareiasaurs, the coracoid plate of the posterior coracoid generally extends posteriorly well beyond the glenoid (Lee Reference Lee1997a).

Although the cleithrum is present in at least two pareiasaurs (Embrithosaurus schwarzi and Bradysaurus (Lee Reference Lee1997a, Reference Leeb)), it does not appear in any known specimen of Deltavjatia, and is assumed to be absent in this taxon.

The clavicles (Figs 6, 9) are relatively thin and blade-like and connect the scapulocoracoids on either side of the skeleton with the ventrally-positioned interclavicle. These elements are well preserved and in articulation in UMZC T1321 and KPM 232, and are very gracile (Fig. 6). The clavicle originates in a deep dorsally-facing facet of the interclavicle, and would have met its pair at the midline. The element then curves posterodorsally and laterally, twisting slightly where it is applied to the anterior edge of the scapular blade. The clavicle follows the edge of the scapular blade dorsally, before ending just ventral to where the latter structure increases in width – about halfway along its length. The clavicle remains about the same width throughout, tapering slightly at its dorsal end. The posterior face of the clavicle has a groove running along the length of the bone, indicating an area for the origin of the deltoid muscle.

The shape of the interclavicle (Figs 8, 23) of Deltavjatia clearly places this taxon within the Ankyramorpha – a group of more derived parareptiles (deBraga & Reisz Reference deBraga and Reisz1996). The element is vaguely anchor-shaped, with long thin laterally-projecting ‘arms’ that curve slightly dorsally at their ends (Fig. 23), and a thin stem, contrasting with the diamond-shaped interclavicle characteristic of lower tetrapods (deBraga & Reisz Reference deBraga and Reisz1996). The interclavicle of Deltavjatia would have been oriented primarly posteriorly in life, with the facets for the articulation of the interclavicles facing dorsally and slightly anteriorly, and with the relatively long stem running posteriorly. The stem also curves slightly dorsally (posteriorly) at its pointed tip. The tip of the stem in KPM 11/94 appears to be slightly forked (Fig. 8), a condition that might be considered a taphonomic artifact, were it not for a similar morphology present in a specimen of Macroleter poezicus (Tsuji pers. obs.). There is a slight ridge running along the ventral face of the stem of the interclavicle (Fig. 23A), likely related to the attachment of the pectoralis muscle. The bone does not possess any dermal sculpture, unlike the condition seen in seymouriamorphs (Romer Reference Romer1956).

In Deltavjatia the humerus (Fig. 6) retains a relatively basal form. Although the element is only exposed in dorsal view in the specimens in which it is present, certain morphologies can still be discerned. The humerus shares a fairly robust morphology with more basal pareiasaurs, but in contrast with the Tapinocephalus and Tropidostoma Assemblage Zone pareiasaurs, there is a definite, relatively gracile shaft (Fig. 6). The proximal end of the bone is flat and flared anteroposteriorly, bearing a broad surface for the articulation with the glenoid of the scapulocoracoid. The angle between the proximal and distal planes of expansion is slightly greater than 45 degrees, which is neither particularly basal nor derived. Bradysaurus seeleyi, a basal pareiasaur, has a humeral torsion of approximately 60 degrees, and the humeri of the most derived pareiasaurs such as Anthodon have an angle of around 20 degrees (Lee Reference Lee1997a). The distal end of the humerus in Deltavjatia does not flare as widely as the proximal end. A large entepicondyle foramen is present on the moderately developed entepicondyle. The end distal end of the humerus bears a broad trochlea for the reception of the large olecranon of the ulna. Although the humeri available for study are not ideally preserved, there is no evidence of an ectepicondylar foramen.

The radius (Fig. 6) and ulna are similar to those of other pareiasaurs, and indeed display a consistent morphology across most early amniote groups and are easily recognisable. The radius is slightly shorter than the ulna. It is primarily columnar, with slightly expanded proximal and distal ends. The proximal end of the bone has a slightly concave surface for articulation with the radial condyle of the humerus.

The ulna (Fig. 6) has a large expanded head and a tall and robust olecranon, but a relatively shallow sigmoidal notch for the reception of the trochlear condyle of the humerus. The head of the ulna is significantly larger than the distal end of the bone and, indeed, the radius has a larger distal end than the ulna.

The manus (Figs 6, 25) of Deltavjatia also does not differ greatly from the general pareiasaurian form. Although none of the specimens examined retain a complete manus, certain elements can be distinguished. Boonstra (Reference Boonstra1929a) describes the proximal carpus of pareiasaurs as composed of a radiale, ulnare, intermedium and centrale. In UMCZ T1321, a large, block-like element in close association with the radius can be identified as the radiale, and a slightly smaller element close to the ulna is identified as the ulnare (Fig. 25). An incomplete element located in between the radiale and ulnare is provisionally identified as an intermedium, which would also have articulated primarily with the ulna (Fig. 25). Boonstra (Reference Boonstra1929a) illustrates a centrale in the manus of Bradysaurus. A small element in between the intermedium and radiale of UMCZ T1321 is possibly a centrale, or it could also be a distal carpal that has drifted out of position. The set of distal carpals in UMCZ T1321 is incomplete, but three, or possibly four, distal tarsals are present (Fig. 25).

The metacarpals are much wider than they are long, a typically pareiasaurian feature. The terminal phalanx is robust, and takes the form of a slightly recurved claw. The phalangeal formula of the manus appears to be 2, 3, 3, 3, 2, which is typical for pareiasaurs (Lee Reference Lee1997a).

2.3.3. Pelvic girdle and hind limb

No isolated postcranial elements of Deltavjatia were available at the time of this study, as most of the skeletons including postcranial material have been prepared or mounted in situ. The pelvic girdle and portions of the hindlimb of Deltavjatia are preserved in three specimens. However, as they are mounted in situ, of the pelvic girdle only the anatomy of the ilium can be described in detail.

The ‘neck’ of the ilium (Fig. 6) of Deltavjatia is tilted anteriorly, at an angle of approximately 60 degrees to the plane of the body of the pelvic girdle and the blade is flared and slightly everted. The anterior end of the iliac blade is longer and more pointed than the posterior end. The posterior margin of the iliac blade is oriented almost vertically. The forward slant of the ‘neck’ of the ilium is typical of the more derived pareiasaurs, with a more upright blade typifying the more basal members of the clade (Lee Reference Lee1997a). It has been suggested that the anterior and dorsal expansion of the iliac blade is related to a shift in limb posture indicating a more upright stance of the hind limb, a trend which is paralleled by the contemporary synapsids (Romer Reference Romer1956).

The ischium and pubis (Fig. 6), though present in the mounted skeleton (KPM 232) are still embedded in matrix and thus features of the acetabulum and puboischiadic plate, including the obturator foramen, cannot be described for this taxon. The lateral portion of the ischium can be seen in dorsal view in KPM 232, as can the posterior edge of the pubis, which forms a relatively straight posterior margin.

The femur (Fig. 6) of Deltavjatia is relatively robust (Fig. 6). The proximal end (or head) of the bone is tilted anteriorly. It is also quite broad, with a visible and relatively large internal trochanter and corresponding adductor ridge for the attachment of the adductor musculature, that is slightly separated from the main body of the bone. The shaft of the femur is only minimally constricted, and is a flattened oval in cross-section, rather than being circular. On the distal end of the femur, the articular condyles for the tibia and fibula are well defined and separated by a significant intercondylar fossa (Fig. 6).

The tibia (Fig. 6) of Deltavjatia is not particularly distinct from the pareiasaurian and, indeed, the early amniote condition. The proximal end of the bone is expanded such that it is subtriangular in cross-section, and is divided into two articular surfaces that articulate with the femoral condyles. There is an evident cnemial crest on the dorsal side of the proximal end of the bone, bordered anteriorly by a significant furrow (Fig. 6).

The fibula (Fig. 6) is also a primitive-looking bone that is only slightly shorter, but is significantly more gracile than the tibia. The distal end is slightly expanded and slightly flattened where it articulates with the astragalocalcaneum. The shaft of the fibula curves lightly away from the tibia (Fig. 6).

The tarsus (Figs 6, 7) of pareiasaurs is distinctive. In adult animals, a single element comprises the proximal tarsus – the astragalocalcaneum – a structure resulting from the complete fusion of the astragalus and the calcaneum. This condition is also present in Macroleter (Tsuji Reference Tsuji2006), Emeroleter (Tsuji et al. Reference Tsuji, Müller and Reisz2012), and also in some procolophonoids (Ketchum & Barrett Reference Ketchum and Barrett2004). The astragalocalcaneum is only preserved in two known specimens of Deltavjatia, KPM 232 and TMP 95.95.44.10. The outline of the element can be described as sub-triangular or house-shaped, with a relatively flat posterior side and a pointed anterior end. This apex is displaced slightly laterally, such that the fibular articulation surface is shorter than the tibial. A single foramen for the perforating artery can be seen entering the element slightly posterior to the centre of the bone in KPM 232. The foramen does not run straight through the bone, entering close to the ventral (posterior) margin on the dorsal surface, but emerging closer to the centre on the ventral surface (Fig. 7A, B). The dorsal surface of the element consists of two planes that meet at a raised ridge, running down from the point dividing the articulating surfaces for the tibia and the fibula on the anterior side. On the ventral surface, there is a prominent set of grooves forming an inverted ‘y’ (Fig. 7B).

The distal tarsal elements have been described in other pareiasaurs (Broom Reference Broom1913), but are rarely preserved and prepared in situ (Boonstra Reference Boonstra1929b), making it difficult to determine the constituent elements. It is often the case that the distal tarsals are missing from specimens, and this also appears to be the case in specimens of Deltavjatia that preserve this area. Only the large astragalocalcaneum complex remains (KPM 232, TMP 95.44.10) (Fig. 6).

The metatarsals of Deltavjatia are wide and flat, approximately 1.5 times longer than wide. The phalanges of Deltavjatia, like those of all parieasaurians, are very robust, wider than long, and each digit terminates in a robust and recurved claw (Boonstra Reference Boonstra1929b). The phalangeal formula in the pes appears to be 2, 3, 4?, 4, 3.

2.3.4. Dermal armour

The large number of articulated specimens of Deltavjatia that include postcranial material and span multiple size classes permits a unique opportunity to examine the dermal armour of a pareiasaur. Deltavjatia has very little dermal armour, even in the largest specimens, approaching the condition seen in the more basal and oldest members of the clade such as Bradysaurus (Boonstra Reference Boonstra1934b). Although four of the specimens examined during this study had significant amounts of articulated postcranial material, few osteoderms were found in direct association with the vertebral column. Oddly, scutes appear to be more closely associated with each other and more likely to be found in place in the smaller, juvenile specimens of Deltavjatia (see discussion below).

The general form of a single scute in a larger individual is intermediate between the condition seen in the oldest Tapinocephalus Assemblage Zone pareiasaurs and the younger, more derived forms. Whereas the osteoderms of the basalmost (sensu Lee Reference Lee1997a) pareiasaurs such as Bradysaurus are relatively smooth and convex ovoid masses (Boonstra Reference Boonstra1934b), the osteoderms of the more derived forms such as Pareiasaurus and Nanoparia possess a raised central boss, sometimes elaborated into a horn or spike, with an extensive, flattened surrounding area bearing ridges radiating outwards from the boss (Boonstra Reference Boonstra1934b; Maxwell Reference Maxwell1991). In the case of Deltavjatia, the central boss is rounded and spherical, not approaching the horn-like form seen in the more derived pareiasaurs such as Scutosaurus and Elginia (Lee Reference Lee1997a). However, the larger osteoderms do possess a small, flat shelf surrounding the boss that is relatively featureless, with few, or only faint, ridges radiating outwards and a slightly crenulated edge in some cases (Fig. 28C). The osteoderms associated with the posterior dorsal vertebrae in a couple of specimens consist of small, almost spherical structures with a smooth surface and no associated flattened area.

Figure 28

Ontogenetic series of dermal osteoderms with anterior dorsal vertebrae in Deltavjatia rossicus. Based on specimens (top to bottom): KPM 11/94, KPM uncat/P1, KPM 232.

Where scutes are preserved, they are usually in relatively close association with the neural spines of the vertebrae. None are found in close association with the ribs or limbs. This pattern of the dermal armour in Deltavjatia is similar to that of the less-armoured forms such as Bradysaurus and Embrithosaurus, which show only a single or paired row of osteoderms associated with the neural spines (Boonstra Reference Boonstra1934b). The osteoderms are more closely associated with each other more anteriorly, beginning their close association with the neural spine of the vertebra immediately following the axis neural arch. They are particularly closely associated with each other over the cervical region of the vertebral column, even forming interdigitating sutures between them in the smaller of the specimens. They are less closely associated farther posteriorly in the vertebral column. More derived forms such as Pumiliopareia and Anthodon have a pattern of scutes that approximates a full dorsal carapace, with rows of articulated osteoderms following the ribs and bony studs also over the limbs and limb girdles (Lee Reference Lee1996, Reference Lee1997b).

In Deltavjatia the osteoderms are positioned in an offset single or double row over the neural arches of the vertebrae. A pair of osteoderms articulates with the lateral edges of the expanded and flattened dorsal surface of each neural spine, but in larger individuals, these osteoderms do not meet each other at the midline. There is also a medial row of osteoderms that is attached to the anterior side of the flattened neural spine, bridging the gap between consecutive spines in smaller individuals, similar to the configuration illustrated by Brink (Reference Brink1955) for Nanoparia pricei. The osteoderms are more closely associated to each other and to the neural spines, and are also larger, more anteriorly.

5.2.1. Sclerosaurus armatus

Sclerosaurus was found to be the sister taxon to Pareiasauria by Lee (Reference Lee1995b, Reference Lee1997a) but a recent revision revealed that it is in fact nested well within the leptopleurinine procolophonids (Sues & Reisz Reference Sues and Reisz2008), a position that is stratigraphically consistent with this taxon's Early Triassic age. Sclerosaurus has been excluded from the analysis because other closely related procolophonids – the Leptopleurinae – are not included.

5.2.2. Lanthanosuchus watsoni

Lanthanosuchus has an elaborately sculptured and highly dorso-ventrally compressed skull, with an altogether aberrant morphology. The taxon was found by Lee to have close affinities to pareiasaurs, based primarily on characters relating to the palate and braincase (Lee Reference Lee1995b, Reference Lee1997a). However, a revision of the taxon and further phylogenetic analysis showed that Lanthanosuchus was in fact most closely related to the Early Permian Oklahoman parareptile Acleistorhinus pteroticus (deBraga & Reisz Reference deBraga and Reisz1996), with this group nesting more basally within Parareptilia (deBraga & Reisz Reference deBraga and Reisz1996; Tsuji et al. Reference Tsuji, Müller and Reisz2012). Thus, as it is in fact many nodes removed from this latter group, it appears that Lanthanosuchus shares a number of convergent characters with pareiasaurs. Since the analysis in this study only involves pareiasaurs and their closest outgroups, Lanthanosuchus watsoni was not included and was eliminated as an outgroup taxon.

5.2.3. Turtles

In another major change from Lee (Reference Lee1995b, Reference Lee1997a), this analysis excludes turtles as a terminal taxon. A general consensus regarding the phylogenetic placement of turtles has emerged in the last few years (eg. deBraga & Rieppel Reference deBraga and Rieppel1997; Hedges & Poling Reference Hedges and Poling1999; Rieppel & Reisz Reference Rieppel and Reisz1999; Zardoya & Meyer Reference Zardoya and Meyer2001; Meyer & Zardoya Reference Meyer and Zardoya2003; Müller Reference Müller, Arratia, Wilson and Cloutier2004; Hill Reference Hill2005), which places them well within Eureptilia as archosauromorphs, although some still argue for their inclusion within Parareptilia (Lyson et al. Reference Lyson, Bever, Bhullar, Joyce and Gauthier2010) and even Lepidosauromorpha (Lyson et al. Reference Lyson, Sperling, Heimberg, Gauthier, King and Petersen2012). Including turtles in this analysis without other purportedly related groups would not be an adequate assessment of their position within early amniotes. The elimination of turtles resulted in a number of characters becoming phylogenetically uninformative, because some characters were solely synapomorphies uniting Anthodon with turtles, or became autapomorphic for turtles upon the exclusion of other taxa.

5.6.3. Parsimony – discussion of results

The addition of morphological information from Deltavjatia rossicus, the addition of the nycteroleters, and the elimination of outgroup taxa that have since been revised has had a definite influence on the final topology under a parsimony algorithm. This analysis strove to include as many taxa as possible and still maintain a reasonable number of most parsimonious solutions, rather than eliminating somewhat problematic taxa. The integration of the new morphological information from the nycteroleters resulted in them becoming the closest sister taxa to Pareiasauria, which differs from other analyses in which the procolophonomorphs hold this position (outside of the two other outgroups eliminated for the reasons stated above) (Lee Reference Lee1997a; Jalil & Janvier Reference Jalil and Janvier2005). The genus ‘Bradysaurus’ remains paraphyletic, as it has in all other analyses, with ‘Bradysaurus’ seeleyi retaining the basalmost position among the pareiasaurs. Indeed, the topology recovered in this analysis is broadly similar to that recovered by Lee (Reference Lee1997a) and Jalil & Janvier (Reference Jalil and Janvier2005), a fact not surprising, given the fact that the basis of the three analyses is the same.

The main differences within Pareiasauria between the current topology and those published previously include the more basal position of Parasaurus geinitzi (referred to in previous analyses as “the Kupferschiefer pareiasaur”). Following Tsuji & Müller (Reference Tsuji and Müller2008), additional morphological information and the reinterpretation of the bones of the skull roof from the state implied by the original skull description by Wild (Reference Wild1985) indicate a less derived morphology, although this position is equivocal, with placements among the more derived pareiasaurs also being among the most parsimonious topologies. More complete specimens, especially those with well-preserved cranial material, would help to further support one of these positions. The phylogenetic position of Deltavjatia rossicus, on the other hand, does not appear to be considerably altered. It remains a relatively basal taxon amongst those pareiasaurs not present in the Tapinocephalus and Pristerognathus assemblage zones, as would be expected from it possessing a mix of basal and derived characters.

The parsimony analysis has resulted in a poorly resolved cladogram, with relatively poor support for the majority of the nodes, including those within Pareiasauria. This state is similar to other analyses of pareiasaurian relationships (Lee Reference Lee1997a; Jalil & Janvier Reference Jalil and Janvier2005), suggesting that more information is needed for many of the taxa. In particular there is poor resolution amongst the basal members of the clade, and these Tapinocephalus and Pristerognathus assemblage zone taxa require more taxonomic as well as anatomical study, though due to the high degree of ossification and poor preservation and preparation of the fossils, their anatomy is notoriously difficult to discern. Whereas there are some informative postcranial characters and more information can be gleaned from these elements should they be adequately preserved, there is more potential to find new and informative characters in the skull, when and if the sutures can be determined.

One difference between the current analysis and that of Lee (Reference Lee1997a) is the exclusion in the current character matrix of ordered characters, a practice that imposes additional cost for transitions between character states that are not consecutive. The ordering of characters can in some cases be justified; certain continuous morphological characters (those that show quantitative variation along a single axis) are delimited, based on the fact that the similarity between the states is informative (Wiens et al. Reference Wiens, Fetzner, Parkinson and Reeder2005), and not ordering these types of characters implies that the character itself is not informative and thus should otherwise be excluded from the analysis. While this is true for these types of characters, for most other discrete characters the path of evolution is not as clear. What constitutes “a clear morphocline” is also not always the same to every observer, particularly when the justification is based on changes in character states that are, directly or indirectly, based on presumed phylogenetic trajectory (and hence circular reasoning), and thus the more conservative practice is to use the ordered characters only minimally (Hauser & Presch Reference Hauser and Presch1991; Slowinski Reference Slowinski1993). Lee (Reference Lee1997a) also uses an all-zero ancestor (a combination of eureptiles and millerettids were used to infer the ‘primitive’ state of the ingroup) as the outgroup, a practice which presumes the ‘basal’ condition ‘a priori’. It is the opinion of the present author that the determination of these characters is better implied by the states possessed by an outgroup taxon.

Despite the elimination of character ordering and the use of a real terminal taxon as the outgroup in this analysis, these changes have not significantly affected the topology recovered, nor have they had a significant effect on the CI or RI that result from them. Apart from the removal from the character matrix of a number of characters shared by only turtles and Pumiliopareia, there was little effect on the analysis that resulted from the removal of turtles. A proper test of the relationships of turtles requires the inclusion of a wider range of terminal taxa, in particular those of suspected eureptilian relatives, a task outside the scope of this analysis.

5.7.3. Bayesian – discussion of results

Although the basic phylogenetic signal remained the same, the considerable difference found between the different Bayesian analyses – that including autapomorphies and that without – indicates that autapomorphies have a definite effect on the analysis under a Bayesian framework. In its traditional use in molecular estimations of phylogeny, autapomorphies are naturally included in the data set, along with constant characters. However, in an analysis of morphological characters using a parsimony algorithm, these characters are traditionally excluded, because only shared derived characters are informative in determining the most parsimonious topology. Using maximum likelihood methods, however, autapomorphies are phylogenetically informative, because they influence branch lengths of terminal taxa, and estimates of the mean and variance of evolutionary rates across characters, and therefore affect the likelihood of a given tree, and thus should be included in a Bayesian analysis when possible (Lewis Reference Lewis2001). It is interesting to note that autapomorphies made up a high proportion of the total number of characters used in the analysis – 72 out of 197 characters (36.5%). It is also interesting to note that the analysis that included autapomorphies in many ways more closely approximates the parsimony tree than that without. This result suggests that the inclusion of autapomorphies does indeed have a significant, and possibly positive, effect on the topology recovered in a Bayesian analysis. However, there is no standardised technique for the identification of autapomorphies in fossil taxa, and readily available autapomorphy lists exist only for taxa that have been defined using autapomorphies, a practice that is relatively new, and by no means exhaustive. Many of the taxa that lack autapomorphies in this analysis are poorly known, or have not been reexamined in many years.

Branch lengths, as was found by Müller & Reisz (Reference Müller and Reisz2006), are not directly correlated with the posterior probability of the branch. Differences in branch lengths between the models were minimal, while between the different data sets, the internal branch lengths of the analysis that did not include autapomorphies were often noticeably longer than the same branches in the analysis with autapomorphies. This is a considerable change from the findings of Müller & Reisz (Reference Müller and Reisz2006), perhaps relating to the greater proportion of autapomorphic characters in the present analysis (0.58 vs. 0.47).

In the current analysis, the Bayes factor indicates that in the analysis including autapomorphies, the run implementing gamma was preferred over that with no model, and yet in the analysis without autapmorphies, there is little to choose between excluding and including the gamma parameters. This situation is interesting, because in other cases, including Tsuji et al. (Reference Tsuji, Müller and Reisz2012), and Müller & Reisz (Reference Müller and Reisz2006), there was a significant and obvious preference for the implementation of the gamma parameter.

Currently, Bayesian analysis using morphological characters uses the most simplistic form of character change – the Mk model – which assumes that a character can change its state at any time with equal probability for all instantaneous time intervals along the branch (Lewis Reference Lewis2001). Whether more sophisticated models of character rate substitution would have a significant effect on topologies recovered in Bayesian analysis remains to be seen. In addition, an interesting question is whether the type or definition of morphological characters created for use in parsimony analysis are also the most appropriate way of defining characters for Bayesian methods. This model of character evolution, although the most simplistic one available, was designed to model the changes in molecular sequences – those related to base substitutions. The transitions between discrete morphological characters such as those involving presence/absence can be assumed to proceed differently from continuous characters such as vertebral number, which are related to changes in serial homology. Thus, using the same type of model to represent them may not model their changes realistically. However, the use of this simplistic model seems to give relatively similar results to parsimony, so these speculations are at this time only academic in nature.