Materials

We studied material assigned to Stereognathus ooliticus and S. hebridicus, as well as Stereognathus sp. and unspecified Tritylodontidae, from the aforementioned U.K. localities. See Localities for details.

Stereognathus ooliticus material from BGS comprises the holotype BGS GSM113834, a fragment of maxilla with three postcanines collected from the Taynton Limestone Formation (Stonesfield Slate) in Oxfordshire, England. From DORCM specimens G11048 and G10828, postcanines from the Forest Marble. From GLRCM specimens, GLRCM MLR 20–22, GLRCM MLR 20–38, GLRCM 10174, GLRCM 2104, GLRCM 2105_4, GLRCM 2105_6, GLRCM G50137, GLRCM G50236, GLRCM G50505, GLRCM G50506, GLRCM G50507, GLRCM G50508, GLRCM G50647, GLRCM G50705, GLRCM G50907, GLRCM G51108, GLRCM G51221, GLRCM G51222, GLRCM G51223, GLRCM G51224, GLRCM G51243, GLRCM G51244, GLRCM G51245, GLRCM G51520, GLRCM G51521, GLRCM G51616, GLRCM G51823, GLRCM G51906, GLRCM G51907, GLRCM G52021, GLRCM G52022, GLRCM G52026, GLRCM G52027, GLRCM G52038, GLRCM G52127, GLRCM G52202, GLRCM G52204, GLRCM G52205, GLRCM G52304, GLRCM G52641, GLRCM G52643, GLRCM G52820, GLRCM G52861, GLRCM G53402, GLRCM G53403, GLRCM G53404, GLRCM G53405, GLRCM G53406, GLRCM G53407, GLRCM G53408, GLRCM G53409, GLRCM G53410, GLRCM G53411, GLRCM G53412, GLRCM G53413, GLRCM G53414, GLRCM G53415, GLRCM G53416, GLRCM G53417, GLRCM G53418, GLRCM G53419, GLRCM G53420, GLRCM G53421, GLRCM G53422, GLRCM G53423, GLRCM G53424, GLRCM G53425, GLRCM G53426, GLRCM G53427, GLRCM G53428, GLRCM G53429, GLRCM G53430, GLRCM G53431, GLRCM G53432, GLRCM G53433, GLRCM G53434, GLRCM G53435, GLRCM G53804, GLRCM G53806, GLRCM G53807, GLRCM G53809, GLRCM G53811, GLRCM G53812, GLRCM G54017, GLRCM G54018, GLRCM G54610, GLRCM G54633, GLRCM G54634, GLRCM G54635, GLRCM G54701, GLRCM G54702, GLRCM G54703, GLRCM G54810, GLRCM G54811, GLRCM G55225, GLRCM G55226, GLRCM G55227, GLRCM G55534, GLRCM G55810, GLRCM G56416, GLRCM G56424, GLRCM G56425, GLRCM G56426, GLRCM G56433, GLRCM G510202, GLRCM G510203, GLRCM G510204, GLRCM G510205, GLRCM G510206, GLRCM G510207, GLRCM G510208, GLRCM G510209, GLRCM G510210, GLRCM G510211, and GLRCM G75710, all postcanines from Hornsleasow, mostly single cusps.

At the NHMUK, the following postcanine fragments are identified as S. ooliticus: NHMUK PV M.36503, NHMUK PV M.36510, NHMUK PV M.36534, NHMUK PV M.36537, and NHMUK R.8720, and the following postcanine fragments are identified as Tritylodontidae—we identify them as Stereognathus ooliticus: NHMUK PV M.36534, NHMUK PV M.36539, NHMUK PV M.36506, NHMUK PV M.36543, NHMUK PV M.46103, NHMUK PV M.46266, NHMUK PV M.46261, NHMUK PV M.45265, NHMUK PV M.46268, NHMUK PV M.46270, NHMUK PV M.46271, NHMUK PV M.46273, NHMUK PV M.46274, NHMUK PV M.46272, NHMUK PV M.46374, NHMUK PV M.46375, NHMUK PV M.46277, NHMUK PV M.46373, NHMUK PV M.46382, NHMUK PV M.46383, NHMUK PV M.46384, NHMUK PV M.46386, NHMUK PV M.46403, and NHMUK PV M.46415. The following are also identified as Tritylodontidae, but we do not identify them as such: NHMUK PV M.46255 and NHMUK PV M.46263. All of this material is from Kirtlington Cement Quarry, Oxfordshire, England, except NHMUK R.8720, which is from Westcliff.

At the OUMNH, specimen J.21790 is an edentulous fragment of S. ooliticus maxilla from the Taynton Limestone Formation (Stonesfield Slate) and the following postcanines (mostly fragmentary) are Stereognathus sp. from Kirtlington Cement Quarry, but we consider them all to be S. ooliticus: OUMNH J.79435, OUMNH J.79439, OUMNH J.79447, OUMNH J.79448, OUMNH J.79459, OUMNH J.79466, OUMNH J.79469, OUMNH J.79470, OUMNH J.79471, OUMNH J.79477, OUMNH J.79478, OUMNH J.79480, OUMNH J.79484, OUMNH J.79492, and OUMNH J.21790.

The BRSUG material comprises Stereognathus hebridicus postcanine material: holotype postcanine BRSUG 20572; paratypes BRSUG 20573, BRSUG 20574, and BRSUG 20575; and more fragmentary specimens BRSUG 29000–29002 and BRSUG 28996–28999 (these include four to five postcanines grouped together per specimen number). All of this material was collected in the 1970s and 1980s near Elgol, from the Kilmaluag Formation, Middle Jurassic, Isle of Skye.

The NMS material comprises dental remains of Stereognathus hebridicus: NMS G.1992.47.120 (comprising two specimens in same matrix) and NMS G.2017.17.6, collected in the 1980s, and NMS G.2017.17.1, NMS G.2017.17.2, NMS G.2017.17.3, NMS G.2017.17.4, and NMS G.2017.17.5, collected between 2013 and 2016 during field work at various sites in the Kilmaluag Formation on the coast north of Elgol (Middle Jurassic, Isle of Skye). This includes some of the most intact postcanine material yet found, figured here for the first time.

Methods

Terminology—We use cusp terminology modified from Watabe et al. (2007) with additions of the PIA (posterior interlocking area) and AIA (anterior interlocking area) from Lopatin and Agadjanian (2008) (Fig. 1). We follow the convention of referring to tritylodontid molars as ‘postcanines,’ despite the absence of canines in tritylodontids. We use the cusp formula as begun by Simpson (1928), specifying buccal, medial, and lingual numbers of cusps, e.g., 2–2-2 in Stereognathus, i.e., two cusps in each longitudinal row. Postcanine can be abbreviated to PC (uppers) or pc (lowers); likewise to indicate buccal, medial, and lingual, we use uppercase for cusp terminology in the upper postcanines (B, M, and L), and lowercase for lower cusps (b, m, and l). There is debate over the homology of cusps between more basal tritylodontids (such as Oligokyphus, which is considered the most basal tritylodontid) and derived tritylodontids (such as Stereognathus, which have a reduced cusp number). Based on Oligokyphus being the most basal genus (Clark and Hopson, 1985; Setoguchi et al., 1999), it appears that cusp reduction in later tritylodontids may have taken place at the anterior of the upper postcanine and posterior of the lower. This is suggested by the presence of vestigial cusps at these loci. Therefore, the posterior-most upper medial and lingual cusps are M3 and L3, and lower, the anterior-most cusps are b1, m1, and l1. Previous authors have considered the posterior-most buccal cusp in the upper molar as B2, not B3. Although we adhere to this convention, the homology of these cusps requires further study. We consider in-depth discussions of which cusps are present, absent, or vestigial from the ancestral condition to be outside the scope of this study. For more information, see discussions in Watabe et al. (2007) and Matsuoka et al. (2016).

FIGURE 1.

Postcanine cusp terminology for Stereognathus used in this paper (modified from Watabe et al., 2007). The PIA (posterior interlocking area) and AIA (anterior interlocking area) are present on both teeth, but the PIA is not visible in occlusal view on upper postcanines and the AIA is not visible in occlusal view on lower postcanines, because they are located on the underside of the tooth.

Measurements—Measurements were taken with digital calipers where possible. For specimens still in matrix, measurements were taken from digitally reconstructed micro-computed tomography (microCT) scans in Mimics 19.0. All microCT scans were digitally reconstructed and image processed using Mimics 19.0 at the National Museum of Scotland. Specimens were also observed using conventional microscopy and morphological features recorded qualitatively based on previous literature and our observations. Maximum length and width were taken. To produce a large enough sample for statistical analysis, where minimal portions of a tooth were missing or worn, a conservative estimate of the original size was made, based on the proportions of more complete specimens.

Specimen Preparation—Specimen NMS G.1992.47.120 was prepared by coating in paraloid B72 then using 10% acetone to remove the surrounding limestone matrix. When the tooth became too fragile to continue this process, microCT data were obtained using the microCT scanner built in-house at the University of Edinburgh, School of Geosciences Experimental Geoscience Facility. The scanner comprises a Feinfocus 10–160 kV dual-transmission/reflection source, MICOS UPR-160-AIR ultra-high-precision air-bearing table, PerkinElmer XRD0822 amorphous silicon X-ray flat-panel detector and terbium-doped gadolinium oxysulfide scintillator. An 0.8 mm aluminium plate limited beam hardening, and data were acquired using a reflection source with a peak energy of 120 kV and 10W target power. Data acquisition software was written in-house, and scans were reconstructed using Octopus 8.7 software. The holotype of S. ooliticus BGS GSM113834 was also scanned in-house at Edinburgh, using a 1.6 mm aluminium plate. Historic specimens from BRSUG were mechanically prepared in the 1970s and 1980s. The S. hebridicus holotype 20572 and paratypes 20573, 20574, and 20575 were microCT-scanned at the University of Bristol using a Nikon XTH225ST scanner with a 225 kV rotating target with a peak energy of 140 kV.

Phylogenetic Analysis—Trees were analyzed using TNT version 1.5 (Goloboff et al., 2008), and the character matrix of 35 characters and 17 taxa is based on that of Velazco et al. (2017), with Stereognathus rescored to reflect our findings (Appendices 1, 2). We used the New Technology search, selecting ratchet, sectorial search, tree drift, and tree fusing. The character states were unordered, and Oligokyphus was used as the outgroup because it is considered the most basal tritylodontid (Clark and Hopson, 1985; Setoguchi et al., 1999).

Holotype S. ooliticus

BGS GSM113834—The holotype of S. ooliticus is a fragment of left maxilla 20.3 mm long, between 3.15 and 3.48 mm deep dorsoventrally, and between 3.9 and 4.8 mm wide buccolingually, although some damage to the buccal side means the original width was slightly greater (Fig. 2). It was originally thought to be a dentary (Charlesworth, 1855; Owen, 1857), then reidentified as a right maxilla (Simpson, 1928), and then correctly identified as a left maxilla (Clark and Hopson, 1985). The abrasion of the buccal maxilla surface has exposed portions of the postcanine roots. The lingual side is less damaged, and convex. The maxilla fragment sits in the original matrix, mechanically prepared out of the rock except for the dorsal surface. Digital reconstructions reveal the shallow depth of the maxilla dorsoventrally, and the lack of laminae extending upwards onto the facial part of the skull, laterally under the jugal, or medially to partially form the secondary bony palate (see Clark and Hopson, 1985:399) (Fig. 2). A ridge projects anteroposteriorly along the distal edge of the dorsal side of the maxilla, but it is broken and missing below the posterior-most postcanines.

FIGURE 3.

Stereognathus ooliticus, BGS GSM113834, holotype, postcanines only. Digitally reconstructed from microCT scans and segmented from the jaw. A, occlusal view of anterior-most postcanine; B, occlusal view of middle postcanine; C, occlusal view of posterior-most postcanine; D, original drawing by Owen (1857); E, occlusal view digital reconstruction; F, dorsal view digital reconstruction; G, anterolingual view digital reconstruction; H₁, anterior view of anterior-most postcanine digital reconstruction; H₂, posterior view of anterior-most postcanine digital reconstruction; I₁, anterior view of middle postcanine digital reconstruction; I₂, posterior view of middle postcanine digital reconstruction; J₁, anterior view of posterior-most postcanine digital reconstruction; J₂, posterior view of posterior-most postcanine digital reconstruction. Anterior direction indicated by longer black arrow, lingual by shorter arrow.

There are three empty postcanine alveoli, followed by three postcanines, and then a final, posterior-most empty alveolus. This indicates at least seven teeth in the tooth row. The postcanine cusp formula is 2–2-2. The upper postcanines are quadrate when viewed occlusally, with the medial cusp row slightly offset posteriorly from the level of the lingual one and the lingual cusp row slightly further offset posteriorly from the medial one, making the crown rhomboidal. All the postcanines are wider buccolingually than they are long anteroposteriorly. The cusps are arranged in three anteroposterior rows of two cusps each. All are broken and missing cusps. In each tooth (where cusps are intact), the anterior ridges of cusps L/M3 and B2 embrace the bases of cusps L/M2 and B1.

The anterior-most postcanine measures 5.7 mm from the tip of the broken M3 to the tip of the roots. The broken and incomplete crown of the postcanine measures 3.1 mm in length (anteroposteriorly), and what remains of the postcanine buccolingually is only 1.6 mm in width. It is the least complete postcanine in the row: only the midline cusps M2 and M3 remain, and both are heavily worn and broken (Fig. 3). The tip of the cusp of M2 is broken, revealing enamel and dentine layers. The tip and posterior slope of M3 is broken; the posterior slope is broken, with the missing portion of the posterior slope leaving a large gap between this and the subsequent tooth. The anterior of the postcanine is worn and broken, missing the M1 and AIA. Only a small portion of the intercuspal grooves remains between M2/M3 and the missing buccal and lingual cusp rows. The micro CT scans reveal that most of the roots of this postcanine are intact, although the crown is separated from the roots. The crown is also cracked between cusps M2 and M3.

The middle of the three postcanines is the most complete, possessing all cusps except L3, and with complete roots. It measures 5.1 mm from the tip of M3 to the tip of the roots, and the crown is 3.4 mm in length (anteroposteriorly) and 3.6 mm in width (buccolingually). All of the cusps are worn and/or damaged, and all are missing most of their enamel. This is the only tooth in which the vestigial cusp L1 is present and visible (Fig. 3). Vestigial cusps B0, M1, and L1 have been incorporated into the AIA, but much of the AIA is missing. There is a small cuspule posterior to, and displaced lingually from, the L2 cusp (Fig. 3E). The corresponding part of the B1 cusp is missing. Comparison with the figure of this tooth by Owen (1857: fig. 5) indicates considerable damage since it was originally discovered and figured (discussed below).

The posterior-most postcanine measures 5.5 mm from the tip of the B1 (most complete cusp, but still broken at the tip) to the tip of the roots, and the crown is approximately 3.5 mm in length (anteroposterior) and 3.1 mm in width (buccolingual)—enough remains to estimate a pre-broken width of at least 3.5 mm. This postcanine is less complete than the middle postcanine: it is missing both lingual cusps but retains the remaining cusps, although they are damaged. The tips of all cusps are broken, with B1 being the most intact, although missing enamel. The AIA and PIA are both worn, but vestigial cusps M1 and B0 remain visible, connected by the AIA ridge.

The roots of the anterior and middle postcanines in BGS GSM113834 comprise six branches, which are arranged in three anteroposterior rows of two branches each (Fig. 3). The buccal and lingual roots are larger, and thicker along their lengths, whereas the medial roots are thinner, shorter, and taper more steeply. In the posterior-most postcanine, there are three roots in the medial row (Fig. 3F), giving a total of seven roots. In all postcanines, the four larger roots are more or less the same width for most of their length (although the posterior two are slightly wider than the anterior two) and taper at the tips. The pulp cavity is hollow in all of the roots.

The AIA and PIA on all three postcanines in the holotype are badly worn, but some features remain visible. The PIA projects posteriorly, with a ridge running buccolingually along the edge of the tooth. This fits into the AIA of the next postcanine posteriorly in the tooth row, as seen between the middle and posterior-most postcanines. The AIA is also ridged buccolingually along the edge of the tooth, with vestigial cusps L1, M1, and B0 incorporated into the ridge. This is more clearly seen in the middle and posterior postcanines in the micro CT scan (Fig. 3).

It is worth noting that the holotype of S. ooliticus is in a less complete state than when first discovered and later described by Owen (1857) (Fig. 3D). Over 150 years of handling has resulted in considerable damage to the postcanine teeth. Looking especially at the most complete, middle tooth in the row: the anterolingual edge is now missing and the L3 has also gone missing since Owen's original drawing was made. The M3 is missing the tip of the cusp and posterolingual edge, and the B3 is damaged buccally and posteriorly, with sections of the tooth missing entirely. The remaining portions of the tooth appear somewhat worn at the edges since Owen's drawing was made. This makes comparisons with new material somewhat problematic.

Holotype S. hebridicus

BRSUG 20572—The holotype of S. hebridicus is a large postcanine (Table 1) with the cusp formula 2–2-2 (arranged in three anteroposterior rows of two cusps) in a small fragment of jaw (Fig. 4). It was originally described as an upper left postcanine but has since been identified as an upper right postcanine (Clark and Hopson, 1985). It is 7 mm from the tip of the M2 to the tip of the roots, and the crown is approximately 5.1 mm in length (anteroposterior) and 5.4 mm in width (buccolingual). The morphology of this specimen agrees with that of S. ooliticus.

There is a small cuspule posterior to L2 and offset lingually (Fig. 4A). A root from the next tooth posteriorly in the tooth row remains intact in the fragment of jaw. It is positioned underneath the PIA, midway between the two widely spaced posterior roots. There were originally six roots: the two posterior-most roots are still present, encased in a small amount of maxilla material (Fig. 4), but their tips are broken. The two larger anterior roots and the smaller anteromedial roots are broken where they meet the tooth base, but the hole for their pulp cavity is still visible.

Paratypes S. hebridicus

BRSUG 20573—This badly damaged upper right postcanine, originally referred to S. hebridicus, is missing much of the lingual cusps and M2, and the tips of the remaining cusps (Fig. 5A–E). The crown measures 4.8 mm in length (anteroposterior) and 5.4 mm in width (buccolingual). The morphology is congruent with the holotype of S. ooliticus. Vestigial cusps M1 and B0 are still visible, incorporated into what remains of the AIA ridge. The PIA projects strongly posteriorly, with distinct indentations and ridges along its length, and a pit in the midline to receive the M1 of the next postcanine in the tooth row. Cusp B1 lacks a posterior cuspule, whereas L2 is broken in the region where one would be found, if present. The bases of only five of the roots remain, because the posteromedial part of the tooth is damaged (Fig. 5F).

BRSUG 20574—This specimen is a lower left postcanine originally referred to S. hebridicus, but with morphology consistent with the lower postcanines of S. ooliticus. The crown measures 5.1 mm in length (anteroposterior) and 3.2 mm in width (buccolingual). The buccal side of the postcanine, including cusps, is well preserved, but the lingual side is damaged, with some enamel missing and both cusps broken (Fig. 6A–E). The anterior side of each cusp is convex. The lingual ridge extending from cusp b1 terminates at the base of b2 ventrally to the cusp apex. This feature cannot be compared with the lingual side due to damage.

The AIA is a ridged buccolingual shelf and projects »1 mm anteriorly from the crown. This area is ‘M’-shaped in occlusal view. The posterior edge of the postcanine is almost straight and is slanted due to the anteroposterior cusp rows being offset, giving the tooth a rhomboidal appearance. There is no cingulid, but the buccal edge of the tooth forms an anteroposterior ridge that we term a pseudo-cingulid (Fig. 6E). Beneath the crown the tooth pinches inwards before the single, quadrangular root extends straight ventrally. This root is broken, extending only 1–2 mm ventrally below the posterior half of the crown. Some fragments of root and matrix are embedded in the hollow of the tooth.

BRSUG 20575—This lower left postcanine is in poor condition, missing portions of enamel, all of the cusp tips, and the entire l2 cusp and portion of the tooth (Fig. 6F–J). It was originally referred to S. hebridicus but has morphology congruent with the lower postcanines of S. ooliticus. The crown measures 5.8 mm in length (anteroposterior) and 3.8 mm in width (buccolingual). The PIA is missing on the lingual side and damaged on the buccal side, but what remains retains a similar shape to that of the paratype BRSUG 20574. Specimen BRSUG 20575 has a strong pseudo-cingulid running anteroposteriorly on the buccal side of the postcanine. The quadrangular root is broken, extending only up to 2 mm, and the postcanine is hollow inside. There is a buccolingual indentation on the anterior face of the root, just less than 2 mm ventrally to the base of the crown.

TABLE 1.

Measurements of U.K. Stereognathus material.

New Material from Skye

NMS G.2017.17.2—This specimen is an upper left postcanine that we refer to S. ooliticus. It is currently the most intact upper postcanine of Stereognathus to be described, retaining most of the enamel, almost wholly intact cusps, and intact AIA and PIA (Fig. 5G–M). The morphology is as for S. ooliticus: the crown has a cusp formula of 2–2-2; deep ‘V’-shaped intercuspal grooves; longitudinal cusp rows offset anteroposteriorly; the tooth is quadrangular in occlusal view; and it is wider than it is long. The crown measures 3.8 mm in length (anteroposterior) and 4.3 mm in width (buccolingual).

Specimen NMS G.2017.17.2 has indents along the PIA and AIA for interlocking with the preceding and succeeding postcanines in the tooth row. The AIA is almost unworn and exhibits multiple cuspules and crenulations along the ridge, and in the position of the vestigial cusps (Fig. 5G, M). This specimen also has distinct cuspules posterior to L2 and B1, displaced lingually and buccally, respectively.

NMS G.1992.47.120—This specimen is a lower right postcanine that we refer to S. ooliticus. It is currently one of the most intact lower pcs of Stereognathus to be described, being almost completely intact apart from the root, retaining all cusps, and with enamel still present (Fig. 7). The crown measures 5.7 mm in length (anteroposterior) and 3.6 mm in width (buccolingual). As in S. ooliticus: cusp formula 2–2; deep ‘V’-shaped intercuspal groove; tooth is quadrangular in shape in occlusal view; has equal-sized cusps; it is longer than it is wide; and the ridges running from b1 and l1 embrace the bases of cusps b2 and l2. The anterior of each cusp is convex; the posterior of the postcanine is almost straight, slanting slightly because the longitudinal cusps are offset anteroposteriorly. The buccal side of the postcanine is straighter than the lingual side in occlusal view, and the buccal edge of the crown has a pseudo-cingulid, as in S. ooliticus lower postcanines and S. hebridicus paratypes BRSUG 20574 and BRSUG 20575.

FIGURE 4.

Stereognathus hebridicus, BRSUG 20572, holotype. A₁, occlusal view; A₂, occlusal view digital reconstruction; B, dorsal view digital reconstruction; C₁, anterior view; C₂, anterior view digital reconstruction; D, buccal view digital reconstruction; E₁, posterior view; E₂, posterior view digital reconstruction; F, lingual view digital reconstruction. Anterior direction indicated by longer black arrow, lingual by shorter arrow.

The PIA on NMS G.1992.47.120 is deep and well defined, divided almost into two by the point where the medial ridges of the l2 and b2 meet in the intercuspal groove and project posteriorly. The embayments of the PIA are pitted and ridged, containing vestigial cusps b3 and l3 (Fig. 7B and G). The AIA is most clearly seen in Fig. 7E; the anterior edges of b1 and l1 are convex, creating an ‘M’-shaped appearance in occlusal view, and project anteriorly to create a shelf that provides the point of contact with the PIA of the preceding tooth.

Measurements

The sizes of postcanine tooth specimens attributed to S. ooliticus and S. hebridicus fall along a range we interpret as ontogenetic variation (Table 1; Figs. 8, 9). The lower postcanines for each species have a similar size distribution (Table 2), with most specimens between 5.1 and 5.5 mm in length and between 3.1 and 4 mm in width.

The upper postcanines do not share the same distribution for each putative species, according to our samples. Those attributed to S. ooliticus have modes of 3.1–3.5 mm in length and 3.6–4.0 mm in width, and those attributed to S. hebridicus have modes of 4.1–4.5 mm in length and 5.1–5.5 mm in width. The distribution of S. hebridicus upper postcanines—unlike the lowers of either putative species or the uppers of S. ooliticus—are bimodal for both length and width. They have two peaks in distribution: in length 3.1–3.5 mm and 4.1–4.5 mm, and in width 2.6–3 mm and 5.1–5.5 mm.

The mean of individual measurements is similar between both species, except for the width of the uppers, which is 3.8 mm in S. ooliticus and 4.6 mm in S. hebridicus (Table 2). The sample of S. hebridicus specimens is two to three times larger than for S. ooliticus (Table 2). The largest range in the sample is among S. hebridicus lower postcanine lengths, and the smallest range is in S. ooliticus lower postcanine widths.

FIGURE 5.

Stereognathus hebridicus, BRSUG 20573, paratype, upper postcanines, and new specimen NMS G.2017.17.2, both reconstructed digitally from micro CT scans. A–F, BRSUG 20573: A₁, occlusal view; A₂, occlusal view digital reconstruction; B₁, anterior view; B₂, anterior view digital reconstruction; C₁, posterior view; C₂, posterior view digital reconstruction; D₁, lingual view; D₂, lingual view digital reconstruction; E₁, buccal view; E₂, buccal view digital reconstruction; F, dorsal view. G–M, NMS G.2017.17.2: G₁, occlusal view; G₂, occlusal view digital reconstruction; H₁, anterior view; H₂, anterior view digital reconstruction; I, posterior view digital reconstruction; J, lingual view digital reconstruction; K, buccal view digital reconstruction; L, dorsal view digital reconstruction; M, ventrolingual view digital reconstruction. Anterior direction indicated by longer black arrow, lingual by shorter arrow.

To test whether there was a difference in size between the two putative species, we carried out Mann-Whitney U tests on the length, width, and width/length ratio of upper and lower postcanines for S. ooliticus and S. hebridicus. The null hypothesis was that there was no difference in the median size of each species. We rejected the null hypothesis in only two instances: when comparing the widths of the upper postcanines and comparing the width/length ratios of the upper postcanines (which depend, in part, on widths), including incomplete specimens for which measurements were estimated. For all other measurements (lower postcanine width, length, and width/length ratio, and upper postcanine length), and when estimated lengths were removed from the data set, there was no statistically significant difference between samples.

When estimated measurements were included, S. hebridicus appeared to have wider upper postcanines than S. ooliticus. However, when estimates were excluded, there was no statistically significant difference between the putative species; therefore, we argue that this statistical result is most likely artifactual and stems from conservatively estimated measurements used to achieve larger sample sizes for analysis. There was no corresponding significant difference found in the lengths of the upper postcanines when estimated measurements were included. When estimated measurements were included, there was also no corresponding difference in the sizes of the lower postcanines, with which the uppers must occlude.

We therefore conclude that the measurement differences in upper postcanine width and upper postcanine length/width ratio between the English and Scottish teeth are an artifact of estimating measurements, and not evidence that they belong to two distinct species.

Synonymizing S. ooliticus and S. hebridicus

Recognition of the size and morphological variability within Stereognathus allows for a systematic reassessment of S. hebridicus. We find no clear diagnostic differences between S. hebridicus and S. ooliticus, either in size or morphology. We therefore synonymize S. hebridicus with S. ooliticus.

The original diagnosis for S. hebridicus stated that this second species was “1.6 times” larger in size than the type species, S. ooliticus (Waldman and Savage, 1972:122). This size difference was determined on the basis of four isolated postcanines of S. hebridicus from Skye, of which only two were upper postcanines that could be compared with the S. ooliticus holotype. No other characters distinguishing S. hebridicus from S. ooliticus were identified, as the authors awaited “full preparation of the material” from Skye before clarifying the diagnosis and anatomy of S. hebridicus (Waldman and Savage, 1972:122). However, the complete description and taxonomic assessment were never carried out. A comprehensive study of specimens assigned to Stereognathus more generally—both S. hebridicus and S. ooliticus—has not previously been undertaken.

Reexamination of all available Stereognathus postcanine material in the U.K. indicates that although the holotype postcanine of S. hebridicus is indeed larger than the holotype of S. ooliticus, when all Stereognathus postcanine tooth specimens are analyzed together, it appears that all material—including both English and Scottish specimens—comprise a spectrum of size with no discernible clustering between large and small morphs (Fig. 8). Results of the Mann-Whitney U test found no statistically significant difference in measurements, except upper postcanine width and upper postcanine width/length ratio when estimated measurements of incomplete postcanines were included (Table 2). There is no corresponding difference in lower postcanine length, width, or width/length ratio in the data including estimates, and no statistical difference at all when estimated measurements are not included. A correspondence between upper and lower postcanine size distributions would be expected, because uppers and lowers of drastically different sizes could not easily occlude with one another. We suggest that the evidence from the lower postcanines and the length of the upper postcanines of specimens attributed to each Stereognathus species indicates that size is not a diagnostic feature separating a purportedly larger species (S. hebridicus) from a smaller one (S. ooliticus). The spectrum of variation is probably best explained by ontogenetic variation, coupled with drawbacks in estimating measurements.

The mechanism for tooth replacement in tritylodontids is a ‘conveyor belt’ system in which teeth are added at the posterior end of the tooth row and lost at the anterior end at the diastema (Kühne, 1956; Matsuoka and Setoguchi, 2000). As a result, isolated tritylodontid postcanines are relatively abundant in the fossil record where tritylodontids occur in the U.K. (although they are mostly fragmentary). The advantage of this is the possibility of recovering postcanines from many ontogenetic stages, revealing information on the size range of these cynodonts. Our measurements (Table 1; Figs. 8, 9) reflect this range of ontogenetic size variation.

Although some tritylodontids are possibly sexually dimorphic (Kühne, 1956; Hopson and Kitching, 1972; Matsuoka et al., 2016), we do not see any clear clustering between possible male and female morphs in our data. However, this may be due to sample size, and such clustering could possibly become apparent if a larger sample was available to us. At the very least, the size of our sample is adequate to show that there is no clear size distinction between the English and Scottish material. For that reason, coupled with the fact that there are no discrete character diagnostic differences among them, we refer them to a single species, which is S. ooliticus by priority.

In terms of discrete characters, there is no strong evidence to support S. hebridicus as a distinct species from S. ooliticus, either in the upper or in the lower postcanine morphology. Assessing potential species-level apomorphies in all known specimens has proven difficult due to the fragmentary nature of the fossils; many features were missing due to damage or wear. This also meant that comparisons of specific characters between fossil localities cannot easily be made. However, where features are present, it appears that there is some variability, but it is not of the variety in which English specimens have one condition and Scottish specimens another. Specimens previously assigned to the two different species share characters, whereas others assigned to the same species do not. Preservation and tooth wear plays a greater role in interpretation than often acknowledged. The often damaged and fragmentary nature of the fossil record for this genus is reflected in our data, because many features could not be observed even in the most complete specimens. Moreover, our study has also revealed some new variable features. For example, the cuspules posterior to L2 and B2 in Stereognathus have not previously been identified but are present in material from multiple localities.

FIGURE 6.

Stereognathus hebridicus, BRSUG 20574 and BRSUG 20575, paratypes, lower postcanines. A–E, BRSUG 20574: A₁, occlusal view; A₂, occlusal view digital reconstruction; B₁, posterior view; B₂, posterior view digital reconstruction; C₁, anterior view; C₂, anterior view digital reconstruction; D₁, lingual view; D₂, lingual view digital reconstruction; E₁, buccal view; E₂, buccal view digital reconstruction. F–J, BRSUG 20575: F₁, occlusal view; F₂, occlusal view digital reconstruction; G₁, posterior view; G₂, posterior view digital reconstruction; H₁, anterior view; H₂, anterior view digital reconstruction; I₁, lingual view; I₂, lingual view digital reconstruction; J₁, buccal view; J₂, buccal view digital reconstruction. Anterior direction indicated by longer black arrow, lingual by shorter arrow.

FIGURE 7.

New specimen NMS G.1992.47.120, a lower postcanine, reconstructed digitally from micro CT scans. A₁, occlusal view; A₂, occlusal view digital reconstruction; B, posterior view digital reconstruction; C, anterior view digital reconstruction; D₁, lingual view; D₂, lingual view digital reconstruction; E, anterolingual view digital reconstruction; F₁, buccal view; F₂, buccal view digital reconstruction; G, posterobuccal view digital reconstruction. Anterior direction indicated by longer black arrow, lingual by shorter arrow.

Despite the lack of morphological evidence for there being two distinct species, we cannot definitively rule out that these geographically separated populations—the more northern Scottish vs. the more southern English faunas— had not undergone some degree of biological speciation that is not reflected in our tooth-based morphological comparisons. However, this is not supported by the current fossil evidence. Future discoveries may shed further light on this.

To date, very little morphological description has been carried out for lower postcanines of Stereognathus, and no formal diagnostic characters have been identified (some features, such as the interlocking areas, were described but not used diagnostically; e.g., in Ensom, 1994). Here we have formally identified several morphological characters of the lower postcanines, which are present in both English and Scottish specimens: the projection of the AIA; the PIA comprising two embayments, pitted inside; the PIA framed by the termination of the b2 and l2 cusp ridges buccally and lingually, and separated medially by the posterior projection of the meeting of the b2 and l2 cusp medial ridges in the intercuspal groove; the ‘M’-shaped anterior of the postcanine in occlusal view; and a single root that retains the quadrate shape of the crown and is indented buccolingually on the anterior face (Figs. 6, 7). Whether the pseudo-cingulid, identified in more complete lower postcanine material is a diagnostic morphological character, or whether it develops as a result of wear during occlusion, is uncertain given the incompleteness of Stereognathus material. Further investigation may reveal more about the pattern of occlusion in Tritylodontidae, particularly in more derived taxa.

FIGURE 8.

Scatterplots of postcanine measurements of Stereognathus. A, upper postcanines; B, lower postcanines. Key for B, as in A. Solid symbols denote complete specimens; open symbols denote incomplete specimens (orange square D S. ooliticus; blue diamond D S. hebridicus). Measurements in Table 1.

Comparisons

All Stereognathus possess the sharp corners and quadrangular shape in both upper and lower postcanines originally described in S. ooliticus (Owen, 1857). Stereognathus shares this feature with Polistodon (He and Cai, 1984), Xenocretosuchus (Tatarinov and Mashenko, 1999), and Montirictus (Matsuoka et al., 2016). This is in contrast to the rounder shape of all other tritylodontid genera.

Upper and lower molar cusps in Stereognathus are more or less equal in size (damage and wear permitting), which is also the case in Xenocretosuchus, Polistodon, and Montirictus, but unlike in Oligokyphus, Kayentatherium, Lufengia, Dinnebitodon, Yuanotherium, Bienotherium, Nuurtherium, or Shartegodon, in which cusp size is variable. A faint pseudo-cingulid visible on the buccal edge of the crown of the lower postcanines has also been described for Polistodon (He and Cai, 1984).

Stereognathus and the recently described genus Montirictus from Japan (Matsuoka et al., 2016) share a great number of similarities, suggesting a close relationship between these genera. Montirictus upper postcanines also possess three rows of two cusps, and well-developed anterior and posterior interlocking areas. Matsuoka et al. (2016) described the vestigial cusps as absent in larger individuals and considered the vestigial M1 cusp to be a separate feature from the AIA protrusions, located instead on the crescentic anterior cusp face of the M2. We consider the vestigial cusp M1 to be present in Montirictus, incorporated as part of the AIA ridge as in Stereognathus. The teeth of Montirictus are quadrangular in shape like in Stereognathus. Both genera have ‘V’-shaped intercuspal grooves that meet subequally (nearly equally), a character they share with Xenocretosuchus (Tatarinov and Mashenko, 1999; Lopatin and Agadjanian, 2008) and Polistodon (He and Cai, 1984) and which is often modified or removed by wear.

Stereognathus had at least seven upper postcanines in the tooth row. Tritylodontids possessed between five (Yunannodon, Bocatherium) and 13 (Polistodon) upper postcanines (the functional tooth count in Polistodon was not reported and is now difficult to determine because the holotype is glued in occlusion, with bones of the dentary and jugal obscuring the rear of the tooth row). The posterior-most postcanine was not yet, if ever, fully erupted (He and Cai, 1984). Nonfunctional posterior-most postcanines, and heavily worn and presumably soon-to-be-lost anterior-most postcanines, are present in various specimens due to the ‘conveyor belt’ mode of tooth replacement, in which teeth move anteriorly, with the oldest teeth falling out at the diastema and new replacement teeth being added at the back of the tooth row. This makes exact tooth count an unreliable character to compare among tritylodontids, because differences observed between specimens could be the result of capturing different moments in the tooth replacement process rather than a diagnostic difference in tooth count between two individuals or species. Upper tooth count can also be variable between different sides of the same animal (Young, 1982; He and Cai, 1984; Clark and Hopson, 1985; Matsuoka and Setoguchi, 2000; Watabe et al., 2007); the same is true for lowers. If there is a close relationship between Stereognathus and Polistodon (Watabe et al., 2007), it suggests the potential for a higher tooth count in the upper tooth row of Stereognathus, as recorded for Polistodon (He and Cai, 1984). More material is needed to address this issue.

The upper postcanines of Stereognathus have a cusp formula of 2–2-2. This differs from Tritylodon (Owen, 1884), Oligokyphus (Hennig, 1922), Bienotherium (Young, 1940), Lufengia (Chow and Hu, 1959), Yunnanodon (Cui, 1976), Dianzhongia (Cui, 1981), Bienotheroides (Young, 1982), Kayentatherium (Kermack, 1982; Sues, 1986), Dinnebitodon (Sues, 1986), Yuanotherium (Hu et al., 2009), Shartegodon (Velazco et al., 2017), and Nuurtherium (Velazco et al., 2017), which all have a higher number of cusps in one or more rows.

FIGURE 9.

Distributions of dimensions of Stereognathus postcanine specimens. A, upper postcanine length; B, upper postcanine width; C, lower postcanine length; D, lower postcanine width. Orange striped bars = S. ooliticus; blue bars = S. hebridicus.

The width/length ratio has been used diagnostically by other authors; we found that this ratio for the lower postcanines of Stereognathus varied from 0.60 to 0.86 and that for uppers between 0.79 and 1.41 (Table 1). These measurements for tritylodontids are open to error because many specimens are missing enamel and have varying degrees of tooth wear. However, the ratio remains useful between genera.

Previous authors have identified only five roots in Stereognathus, and this result has been repeated by subsequent authors, particularly in character analysis. We show here, based on micro CT data, that this root count is incorrect. This indicates that it may be necessary to micro CT scan and recount the root numbers in some other tritylodontid specimens. The roots of Stereognathus upper postcanines instead vary in number between six and seven, potentially connected with their position in the tooth row (with more roots in posterior postcanines). Montirictus also has six roots, but the roots of Montirictus compress inwards more tightly below the crown and orient outwards again ventrally. According to Cui and Sun (1987), Bienotherium, Lufengia, Yunnanodon, and Bienotheroides all have five roots, Nuurtherium also has five roots, and Shartegodon has four (Velazco et al., 2017). Cui and Sun (1987) observed that Lufengia has some fusion or dental laminae between the roots as in Oligokyphus (Kühne, 1956). Oligokyphus is described as having five roots connected transversely in two rows by dental laminae (Kühne, 1956). Such laminae are mostly absent in Stereognathus, although the medial roots are sometimes joined anteroposteriorly into a row (see posterior-most postcanine in BGS GSM113834, Fig. 3F). The extent of this joining appears to be variable.

TABLE 2.

The data set used for analysis, including estimated measurements. Measurements in mm.

In BRSUG 20572, there is a large root in situ between the posterior roots of the postcanine, which by morphology and position belongs to the next tooth in the tooth row (absent) (Fig. 4). However, the medial placement of this larger postcanine root does not follow the morphology of medial roots in the tooth row seen in the holotype BGS GSM113834. Because complete and well-preserved material is so rare, this unusual placement may have been more widespread in Stereognathus; it may be a postdepositional artifact or the result of the conveyor belt movement of the postcanines along the tooth row. This conveyor belt movement has been observed in other specimens of Tritylodontidae to produce an increasing curvature of the roots underneath the preceding postcanine, notably in the lower postcanines (Cui and Sun, 1987; Matsuoka et al., 2000).

The lower postcanines of Stereognathus have a cusp formula of 2–2, which differs from those of Oligokyphus (Kühne, 1956) and possibly Tritylodon (Fourie, 1963), which have the formula 3–3. The 2–2-2/2–2PC/pc cusp formula is shared with Polistodon (He and Cai, 1984), Bocatherium (Clark and Hopson, 1985), Xenocretosuchus (Tatarinov and Matschenko, 1999), Montirictus (Matsuoka et al., 2016), Shartegodon (Velazco et al., 2017), and Nuurtherium (Velazco et al., 2017). Although Bienotherium and Kayentatherium share the 2–2 cusp formula in the lower postcanines, the anterior cusps of both genera are larger than the posterior cusps (Young, 1947; Kermack, 1982), whereas in Stereognathus, Xenocretosuchus, Montirictus, Shartegodon, and Nuurtherium the cusps are equal. The lower postcanines of Stereognathus most closely resemble those of Xenocretosuchus and Montirictus in morphology, and they also bear close resemblance to the recently described Shartegodon and Nuurtherium (Velazco et al., 2017). They are all quadrangular and rhomboidal in appearance, and Stereognathus and Xenocretosuchus possess vestigial (l3) and (b3) cuspules within the AIA and PIA (Tatarinov and Mashenko, 1999). The interlocking areas in Stereognathus are especially similar to those described in Xenocretosuchus kolossovi (Lopatin and Agadjanian, 2008), because the two taxa share the same ridges of chaotic enamel inside the embayments where the next tooth in the row ‘locks’ into place. This is also described for Montirictus (Matsuoka et al., 2016). These vestigial cusps are not mentioned in Shartegodon or Nuurtherium; however, we suggest that they may be present in Shartegodon (Velazco et al., 2017: fig. 9).

The roots of the lower postcanines of Stereognathus show a similar morphology to those of Xenocretosuchus, Montirictus, Shartegodon, and Nuurtherium in being box-like and extending straight downwards from the crown. Although in Stereognathus the ventral-most section of a postcanine has not yet been recovered, the specimens we examined shared the concave ridge 2–3 mm ventrally below the AIA as present in Montirictus (Matsuoka et al., 2016:fig. 3, parts B3, B4, C3. This feature is pronounced in Montirictus: the anterior face of the root is directed posteroventrally at an angle into this concavity, before bulging anteriorly below the concavity. The root also bifurcates below the line of concavity, with the anterior-most root-half curving posteriorly at the ventral tip (Matsuoka et al., 2016). In Shartegodon (and possibly Nuurtherium), the root also curves as in Montirictus, but there is no bifurcation in the ventral part of the tooth, and the ridge identified for Stereognathus and Montirictus is not evident (Velazco et al., 2017:fig. 10). This ‘s’-shaped (referred to as ‘c-shaped’ in Velazco et al., 2017:14) curving is similar to that seen in other tritylodontid postcanines, but more pronounced and angular in appearance in Montirictus and Shartegodon, echoing the shape of the tooth crown. The lower postcanines DORCM G10828 from the Forest Marble, and GLRCM TEMP6036 from Hornsleasow, retain the most complete postcanine Stereognathus roots. They follow a Montiricus/Shartegodonlike pattern (E.P., pers. observ.) but are not complete, and therefore we cannot confirm the ventral-most morphology of the root. The same is true for as yet undescribed material recently recovered from Woodeaton (E.P., pers. observ.). More complete lower postcanine material is required.

FIGURE 10.

Trees generated by our phylogenetic analysis of tritylodontid taxa, using updated character codings for Stereognathus. A, strict consensus of the five parsimonious trees of 71 steps; B, the agreement subtree of 10 taxa.

Stereognathus lower postcanines resemble Xenocretosuchus (Lopatin and Agadjanian, 2008), Montiricutus (Matsuoka et al., 2016), Shartegodon (Velazco et al., 2017), and Nuurtherium (Velazco et al., 2017) in having the buccal cusp row slightly posteriorly offset from the lingual row, and in the morphology of the PIA: forming two embayments separated by the l2/b2 medial ridges meeting in the intercuspal groove and projecting posteriorly. In Xenocretosuchus, the ridges extending from the cusps into the intercuspal grooves were described as connecting in one intercuspal groove, but not the other (Lopatin and Agadjanian, 2008). However, this intercuspal groove is often modified by wear, and therefore this is not reliably diagnostic in either genus.

Phylogenetic Analysis

Many previous character analyses of Tritylodontidae place Stereognathus in a clade that includes Bocatherium, Polistodon, Xenocretosuchus, and Montirictus (Watabe et al., 2007). Most studies agree that Oligokyphus is the most basal member of Tritylodontidae and consider Stereognathus, Bocatherium, Bientheroides, Xenocretosuchus, and Montirictus to be ‘advanced’ tritylodontids (Clark and Hopson, 1985; Sues, 1986; Setoguchi et al., 1999; Watabe et al., 2007). However, only dental remains and two incomplete maxillae have been found and described for Stereognathus.

Clark and Hopson (1985) placed Stereognathus in a clade with Bocatherium and Bienotheroides based on the absence of facial, palatine, and zygomatic processes. Some characters were applied to Stereognathus “by inference” (Clark and Hopson, 1985:399), based on resemblances between the holotype of S. ooliticus and more complete material for Bocatherium and Bienotheroides. They describe a ‘prominent groove on the maxilla’ of S. ooliticus as an indication that it possessed an infraorbital foramen at the junction of the premaxilla, the jugal, and the lacrimal. This groove is not clear either by direct observation or in micro CTscan reconstruction, so we cannot confirm that Stereognathus possessed this character. Stereognathus, Bocatherium, Bienotheroides, Dinnebitodon, and Yuanotherium all share a uniquely reduced and cylindrical maxilla (referring to the convexity of the buccal and lingual sides, although in cross-section the maxilla of Stereognathus is somewhat rectangular), lacking laminal extensions into the face (Clark and Hopson, 1985:399). This highly derived character, along with the reduction to only two principal cusps per longitudinal row in the upper postcanines, supports grouping these tritylodontids into a clade.

The cladistic analysis by Watabe et al. (2007) used 17 taxa and 11 characters, six of which are dental. Watabe et al. (2007) scored the vestigial cusps as absent in Stereognathus, which we see from our reassessment is not the case. They also scored five cranial characters based on Clark and Hopson (1985), most of which were inferred (see above discussion). Their data were compiled secondhand from multiple sources, concluding that these data were not sufficient at that time to satisfactorily resolve the polytomies among Tritylodontidae, namely, between Stereognathus, Montirictus, Polistodon, Xenocretosuchus, and Bocatherium, and between Kayentatherium, Lufengia, and Diangzongia. They suggested that additional characters are required to do so.

The most recent character analysis was carried out by Velazco et al. (2017), using 35 characters (22 skeletal and 13 dental). Stereognathus was scored on 15 of these characters: nine dental and six skeletal. The tree presented in their paper (Velazco et al., 2017:fig. 16) was not the strict consensus tree of four most parsimonious trees of 68 steps as stated (elsewhere it is stated that there were two parsimonious trees [Velazco et al., 2017:28], but there were four). Unfortunately, there is also an error in their character matrix in the appendix for their paper. We obtained the correct matrix from Morphobank and reran the tree analysis using their methods to obtain their strict consensus tree. This places all tritylodontids in an unresolved polytomy, with the exception of Oligokyphus (outgroup) and Tritylodon as most basal, and a separate clade containing Nuurtherium, Shartegodon, and Yuanotherium in an unresolved polytomy.

Our reanalysis of Stereognathus clarifies certain characters, such as root count, and finds no support for other characters, such as a post-incisive snout constriction. We reran the Velazco et al. (2017) analysis with Stereognathus rescored (Appendices 1, 2). Five of the six skeletal characters we rescored as unknown (characters 1, 2, 7, 8, and 14). These characters were previously scored based on Clark and Hopson (1985), as mentioned previously, regarding inferences about the facial, palatine, and zygomatic processes for which we find no support. We retained the reduction of the maxilla as highly reduced (character 12), and the absence of a lateral extension of the maxilla (character 16), and added absence of the palatine contributing to the PC4 alveolus (character 14). The dental characters we retained were the cusp formula of upper postcanines as 2–2-2, the absence of M0 and L0, the large L3 cusp, and uncertainty over whether the lower postcanine bifurcates or is single-rooted (characters 24, 26, 27, 30, and 34). We rescored characters 25, 28, 29, 31, 32, 33, and 35. These are (respectively) the presence of the B0 cusp, the presence of a small M1 cusp and a small L1 cusp, six or seven roots in the upper postcanines, the presence of an anterior median root, the generalized lower cusp formula of 2–2, and the long single root in the lower postcanine with a curve in the ventral-most portion (‘s’-shaped).

Following Velazco et al. with updated characters for Stereognathus yields five parsimonious trees of 71 steps, and a strict consensus tree that is almost identical to the original, but with the polytomy between Nuurtherium, Shartegodon, and Yuanotherium resolved, finding Shartegodon and Yuanotherium more closely related to one another than to Nuurtherium (Fig. 10A). In this matrix, six taxa have >50% missing data and three >60% (Yunannodon 62.9%, Montirictus 65.7%, and Xenocretosuchus 77.2%). Stereognathus has 60% missing data. Removing Montirictus and Xenocretosuchus yields three parsimonious tree of 70 steps and results in a strict consensus tree identical to that found by Velazco et al., but with the addition of a clade formed by Polistodon and Bocatherium. Stereognathus remains part of the polytomy with most other tritylodontids.

Eliminating taxa with the most missing entries can alter the relationships among taxa, without clarifying them (Wilkinson, 2003). In order to avoid this, we ran an agreement subtree on the whole data set with Stereognathus rescored, to identify the largest subset of taxa in all of the parsimonious trees that are identically related (Goloboff et al., 2008). This resulted in a tree with 10 taxa, including Stereognathus, placing it as the nearest outgroup to Shartegodon, Nuurtherium, and Yuanotherium (Fig. 10B). Other than the addition of new taxa, this tree topology differs little from Clark and Hopson's (1985), despite including more characters and taxa.

In light of the difficulties coding only one of these tritylodontid taxa—Stereognathus—based on the previous literature, we consider that phylogenetic analyses of tritylodontids will remain problematic and cannot be further resolved until comprehensive redescriptions (to confirm or redescribe characters as necessary) of existing material are available. We also suggest that there may be more intraspecies variation in cusp shape and morphology than previously recognized—often confounded by poor preservation and degree of tooth wear—and that this variation may have occasionally been erroneously interpreted as apomorphic. We therefore consider phylogenetic analysis to be preliminary until more detailed, up-to-date information is available for the many poorly described or figured taxa, and particularly for taxa that were unresolved in our analysis, such as Polistodon, Lufengia, Beinotheroides, Diangzhongia, and Bocatherium. As is often the case, more complete material for other taxa, including Stereognathus, Xenocretosuchus, and Montirictus, would almost certainly improve the resolution of future phylogenetic analysis.