3.1.1. General features

As reconstructed (Fig. 7a–d), the skull of Eldeceeon resembles that of Silvanerpeton (Ruta & Clack Reference Ruta and Bolt2006, fig. 8) but there are several differences between these taxa (see also Supplementary Material S3 and description of individual bones below). (1) The skull of Eldeceeon has a distinct semi-elliptical outline, unlike the parabolic skull of Silvanerpeton, and is proportionally slightly wider than the latter in relation to its length. (2) Eldeceeon has proportionally slightly smaller and more rounded orbits than Silvanerpeton, and comparatively smaller external nostrils situated farther apart than those of Silvanerpeton. (3) The skull table of Eldeceeon is comparatively more abbreviated than that of Silvanerpeton and with gently convex lateral margins; in contrast, these margins are nearly straight and obliquely orientated in Silvanerpeton. (4) The suspensorium of Eldeceeon is deeper in lateral aspect than that of Silvanerpeton and with a nearly straight posterior margin (gently concave in Silvanerpeton) that is slightly embayed immediately ventral to the point where the squamosal contacts the skull table. Compared to Smithson's (Reference Smithson1994) reconstruction of the skull in lateral view, our reconstruction shows much larger orbits and external nostrils, a more abbreviated snout with a much steeper profile, and a shorter suspensorium characterised by a steeper and straighter posterior margin and a more pronounced dorsal embayment.

While acknowledging the difficulty of working from heavily disrupted material, we think the new skull reconstruction takes into account all available evidence. A possible contentious point is represented by the width of the skull at the level of the basipterygoid articulations. What little information is available on the palatines suggests that these bones were broad and, even allowing for a more oblique orientation of these bones, there appears to be no evidence that the palate was strongly vaulted. We do acknowledge, however, that the transverse flange of the pterygoid might have approached or contacted the medial surface of the jugal. This would confer a narrower projected surface area to the orbit in dorsal projection, but, even so, the orbit area would remain conspicuous, as suggested by the proportions of the circumorbital bones.

3.1.2. Dermal ornament

As in the case of Silvanerpeton, Eldeceeon shows no traces of lateral line canals (Figs 2c, 5c, d). The dermal ornament is well preserved on most skull roof bones, and is more variable and more accentuated than that of Silvanerpeton. The dominant pattern, such as is observed on the nasals, frontals, parietals, squamosals, and jugals (Fig. 7a–c), consists of small pits, slender ridges, shallow grooves, and low protuberances. Pits, ridges, and grooves mostly radiate out from the ossification centres of those bones. More peripherally, the ornament consists of fine striations, but sparse grooves and pits are also observed. Irregularly distributed foramina dot the external surface of the bones and are more densely packed on and around the ossification centres. On the squamosals, the ornament is particularly well developed (Fig. 2b, c). The ossification centre forms a slightly raised irregular area situated below the posterodorsal corner of the squamosal (in lateral aspect), from which strong ridges and grooves radiate out in a characteristic fan-like pattern. A similar, raised subtriangular area is visible on the main corpus of the jugal (Fig. 5b, d). The prefrontals and postfrontals exhibit an irregular ornament of fine, tightly appressed ridges and grooves distributed around the axis of greater anteroposterior curvature of those bones (Figs 2b, c, 5b, d). In striking contrast, the sculpture on the postorbital appears to vary in a mediolateral direction, with two distinct areas separated by a robust longitudinal ridge (Fig. 2b): the medial area carries fine longitudinal striations and shallow grooves, whereas the lateral area is occupied by deep pits and grooves. Although the squama of the lacrimal is heavily disrupted (Fig. 5b), preserved fragments show a coarse ornament of irregular, mostly longitudinal crests delimiting depressions. More posteriorly, along the suborbital ramus, the sculpture of the lacrimal changes into shallow grooves and low ridges, and this pattern continues smoothly posteriorly along the suborbital ramus and posteroventral surface of the jugal. A low relief, pustulose ornament is visible on the intertemporal, supratemporal, and tabular, consisting of irregularly spaced low tubercles and sparse low ridges (Fig. 2c). The ornament of the quadratojugal consists, for the most part, of subparallel, elongate, and anastomosed ridges and grooves (Figs 2b, c, 5b, d). The lateral surface of the maxilla shows light longitudinal striations and rugosities intercalated with slender ridges, and is more densely pitted at its anterior extremity (Fig. 2b, c).

3.1.3. Sutural patterns

Despite disruption, several skull roof sutures are traceable (Figs 2c, 5d, 7a–c). Where bones have been dislocated, the partial extension of bone overlap surfaces (underlying lamellae sensu Kathe Reference Kathe1999) and adjacent sutural seams are visible (see descriptions of individual bones below). On the skull roof, most sutures are gently sinuous. Slightly more elaborate sutural interdigitations occur between the bones of the central skull roof series (premaxillae; nasals; frontals, parietals; postparietals), both in anteroposterior succession and between antimeres, as well as between postparietals and tabulars and between parietals and tabulars. By contrast, in Silvanerpeton most sutures are gently undulating.

3.1.4. Premaxilla

Each premaxilla is divided into a subtriangular upper process and a subrectangular basal portion (Figs 2, 7a–c) delimiting, respectively, the anterior margin and anterior half of the ventral margin of the external nostril. The basal portion includes both maxillary and vomerine processes. The posterodorsally orientated upper process (= ‘nasal’ process of some authors) contacts its antimere along a weakly sinuous medial suture, and appears comparatively more robust and wider than its homologue in Silvanerpeton. Posterodorsally, it forms a strongly interdigitating suture with the nasal. The process widens rapidly ventrally before merging smoothly into the subrectangular basal portion. The abbreviated posterolateral (= ‘maxillary’) process of the basal portion contacts the anterior extremity of the maxilla along a narrow oblique suture, as in Silvanerpeton. Although the posterior (= ‘vomerine’) process of the basal portion is not visible, it is reasonable to assume it was wide and anteroposteriorly abbreviated, judging from the shape of the anterior extremity of the vomers (Fig. 2b, c), as preserved (see section 3.2.3).

3.1.5. Nasal

The broad, flat, and irregularly subpentagonal nasals (vs narrow, elongate, and rectangular in Silvanerpeton) show irregularly indented margins and occupy ~80 % of the length of the skull's preorbital region in dorsal aspect (Figs 2, 5, 7a–c). The conjoined nasals attain their greatest width at the triple sutural junctions with the prefrontals and lacrimals, along a transverse line situated slightly posterior to their mid length. The combined width of the nasals is greater than that of the frontals, unlike in Silvanerpeton where these two widths are nearly identical. In our reconstruction (Fig. 7c), the nasals are squat and foreshortened as a result of the slope of the dorsal surface of the preorbital region. However, their actual length would be slightly greater in a full plan view, as shown in NMS G.1990.7.1 (counterpart; Fig. 2c). In UMZC T.2013.3 (Figs 4, 5), both nasals are disrupted but have maintained their mutual spatial relationships. Their lateral margins are divided by an anterior portion in contact with the lacrimal and a posterior portion in contact with the prefrontal (Fig. 7a–c; see also description of lacrimal below).

3.1.6. Frontal

The shape and proportions of the frontals resemble those of Silvanerpeton, but differ from the latter in three respects (Figs 2c, 5d, 7c): (1) in both taxa the frontals decrease in width posteriorly, more gradually in Eldeceeon than in Silvanerpeton, immediately behind the triple sutural joints between frontals, prefrontals, and postfrontals; (2) the combined width of the posterior extremities of the frontals at the level of their sutures with the parietals is proportionally greater in Eldeceeon than in Silvanerpeton, and occupies more than half of the skull table width at the same transverse inter-orbital level; (3) the centres of ossifications of the frontals in Eldeceeon occur slightly posteromedial (anteromedial in Silvanerpeton) to the triple sutural joints between frontals, prefrontals, and postfrontals. As reconstructed, the frontals of Eldeceeon are only slightly longer than the parietals.

3.1.7. Parietal

The parietals delimit a subcircular pineal foramen (elongate and irregularly sub-elliptical in Silvanerpeton) situated slightly anterior to the inter-parietal suture mid length (Figs 2b, c, 5, 7c). The pre-pineal region of each parietal is more regularly trapezoidal than in Silvanerpeton, increases more gradually in width anteroposteriorly, and is comparatively wider relative to the width of the skull table at the same transverse level (this is partly due to the shape and proportions of the posterior half of the postfrontal; see section 3.1.13.). The post-pineal region is approximately square, its width changing only slightly as a result of the gentle lateral concavity of the parietal–supratemporal suture. The posterolateral area of each parietal forms a distinctly longer suture with the tabular than in Silvanerpeton. Unlike Silvanerpeton, Eldeceeon does not show a small, posterolateral rectangular ‘lappet’ wedged between the posteromedial corner of the supratemporal and the anterolateral corner of the postparietal.

3.1.8. Postparietal

The postparietals are comparatively shorter and narrower than those of Silvanerpeton. Their maximum width is about half of the width of the parietals along their posterior sutural margins (Figs 2b, c, 7c) and form strongly interdigitating sutures with the tabulars. A narrow, smooth flange extends posterior and, presumably, slightly ventral to the posterior margin of the sculptured dorsal surface of each postparietal. From its mid-point, the flange widens slightly medially before meeting its antimere, as well as laterally where it merges into a similar but narrower flange projecting from the tabular.

3.1.9. Intertemporal

The subquadrangular intertemporals are the smallest bones in the lateral temporal series, unlike in Silvanerpeton where they form conspicuous pyriform elements. In Eldeceeon, they are slightly longer than wide, with smoothly convex or weakly undulating anterior, mesial, posterior, and lateral margins, and with their greater axis orientated slightly obliquely (Figs 2c, 5d, 7a–c). The anterior extremity of each intertemporal is aligned with the mid-point of the pre-pineal tract of the inter-parietal suture, while the posterior extremity is aligned with the posterior border of the pineal foramen.

3.1.10. Supratemporal

The supratemporals extend anteroposteriorly for the entire length of the post-pineal tract of the inter-parietal suture, attaining their greatest width at the mid-point level of this suture (Figs 2c, 5d, 7a–c). The anterior half of each supratemporal narrows imperceptibly anteriorly, while the posterior half tapers rapidly, forming a blunt squarish posterior extremity. Both the mesial and the lateral margin of the bone are convex, the former more markedly so, and its sutural contact with the tabular is slightly indented. Its shallowly embayed anterior margin accommodates the intertemporal. A key difference between the supratemporals of Eldeceeon and those of Silvanerpeton is the fact that in the latter taxon, the anterior and posterior margins are anteromedially to posterolaterally oblique.

3.1.11. Tabular

The tabulars wrap around the posterolateral corners of the skull table (Figs 2c, 5d, 7a–c). They are slightly wider than long, with slightly sinuous lateral margins in dorsal aspect. Each tabular forms strongly interlocking sutures with the postparietals and parietals. The lateralmost portion of the anterior margin of the tabular accommodates the posterior extremity of the supratemporal. The posterolateral corner of the bone forms a stout process with a smoothly curved profile, projecting posterior to the level of the postparietal flanges. This process resembles the pitted dorsal portion of the tabular horn seen in various anthracosauroids (e.g., Panchen Reference Panchen1970; Holmes Reference Holmes1984, Reference Holmes1989; Clack Reference Clack1987, Reference Clack2012; Panchen & Smithson Reference Panchen, Smithson and Benton1988; Klembara Reference Klembara1997; Klembara & Ruta Reference Klembara and Ruta2004a, Reference Klembara and Ruta2005a; Ruta & Clack Reference Ruta and Bolt2006). However, it is not possible to ascertain the presence of a ‘subdermal’ component of the horn. Immediately mesial to the process, a narrow, smooth, subtriangular flange detaches from the main corpus of the tabular and contacts the postparietal flange along a short oblique suture (Fig. 2c). In all these features, the tabulars of Eldeceeon differ from those of Silvanerpeton, as follows: (1) the dorsal surface of each tabular in Silvanerpeton is less than half of that of the adjacent postparietal – in Eldeceeon, the tabulars are distinctly larger than the postparietals; (2) the tabulars of Silvanerpeton are plate-like subrectangular elements, longer than wide, without a posterolateral pronounced process, and with a slender, posterolaterally directed, spike-like horn – in Eldeceeon, the tabulars are subtriangular, wider than long, with a robust posterolateral process, and no evidence of a subdermal horn, as far as we can ascertain; and (3) in Silvanerpeton, the nearly straight posterior margins of the tabulars are obliquely orientated, while their anterior margins are divided into a longer lateral portion and a shorter mesial portion forming an obtuse angle – in Eldeceeon, the posterior margins of the tabulars are shallowly concave, while the anterior margins are approximately transversely orientated and irregularly indented.

3.1.12. Prefrontal

The prefrontals contribute to the anterodorsal and most of the anterior sections of the orbital margin (Figs 2b, c, 5b, d, 7a–c). Each consists of a slender, triangular, and posteriorly acuminate ramus (comparatively more robust than that of Silvanerpeton) and an anterior, triangular, and fan-like portion (comparatively larger, but otherwise similar to that of Silvanerpeton). Along the ramus, the lateral margin of the prefrontal (sutured with the frontal) is gently curved. More anteriorly, at the transition between the ramus and the fan-like portion, this margin becomes irregularly sinuous up to the level of the triple sutural joint between prefrontal, frontal, and nasal. More anteriorly, the margin zigzags along the sutural contact with the nasal. The fan-like portion widens rapidly anteroventrolaterally, where it contacts the preorbital squama of the lacrimal and contributes to a substantial part of the snout region between the orbit and the external nostril.

3.1.13. Postfrontal

The postfrontals are comparatively narrower than those of Silvanerpeton and contribute to the dorsal and part of the posterodorsal orbital margin (Figs 2c, d, 5, 7a–c). Unlike the semi-crescentic postfrontals of Silvanerpeton, those of Eldeceeon are sickle-shaped, with an anterior, strongly curved, and narrow ramus forming a point contact with the prefrontal and merging gradually into a wider posterior portion. The embayed posterior margin of the bone accommodates the intertemporal. Anterolateral to the embayment, the postfrontal shows a small abbreviated process sutured with the postorbital. At this level, the postfrontal reaches its maximum width, which is comparable with that of the adjacent pre-pineal portion of the parietal. Unlike in Silvanerpeton, the postfrontal–parietal suture is gently convex posteromesially. In UMZC T.2013.3 (Fig. 5c, d), the posterior portion of the (presumed) anatomically left postfrontal appears to be stouter and with less strongly curved lateral and mesial margins than its homologue in NMS G.1990.7.1 (Fig. 2c).

3.1.14. Postorbital

The large triangular postorbitals are stockier and proportionally larger than those of Silvanerpeton, and contribute to most of the posterior orbital margin (Figs 2c, d, 7a–c). Each consists of an anterodorsal, an anteroventral, and a posterior ramus. The anterodorsal ramus is poorly delimited from the main corpus of the bone and contacts the postfrontal along a short suture. The robust and subrectangular anteroventral ramus is about half as long as the total length of the bone (measured in dorsal or lateral aspects), and forms a slanting suture with the dorsal process of the jugal. This suture is proportionally longer than in Silvanerpeton, in which the anteroventral ramus, while distinct, is about one-third of the length of the postorbital. In addition, the anteroventral corner of the ramus extends into a digitiform process in Eldeceeon (absent in Silvanerpeton). The posterior ramus of the postorbital is comparatively much shorter and narrower than its homologue in Silvanerpeton and contacts the lateral margin of the supratemporal a short distance behind the lateral extremity of the intertemporal–supratemporal suture (Fig. 2c). The main corpus of the postorbital is occupied by a distinct longitudinal ridge. Posteriorly, the ridge merges smoothly into the surface of the posterior ramus. It increases slightly in depth anteriorly, before disappearing immediately behind the orbital margin. The ventral part of the ridge marks the dorsalmost portion of the anteroventral process (Fig. 7a–c).

3.1.15. Jugal

The triradiate jugals consist of a dorsal (‘postorbital’), an anterior (‘suborbital’), and a posterior ramus. Each contributes to a short tract of the posterior orbital margin, as well as to its posteroventral and ventral tracts (Fig. 7a–c). In NMS G.1990.7.1 (Fig. 2c), only a small area of the dorsal ramus of the right jugal is visible, in contact with the right postorbital. In UMZC T.2013.3 (Fig. 5c, d), most of the right jugal is visible in lateral and mesial aspects, except for the rearmost portion, which is largely incomplete. However, this portion can be reconstructed from the morphology of the surrounding elements. The anterior ramus is slender and elongate and contributes to the ventral orbital margin. Its depth is approximately constant in its anterior two-thirds, increases gently in its posterior one-third, and merges smoothly into the corpus of the bone. The dorsal ramus is broad and its posterior margin (in contact with the squamosal) slopes slightly posteroventrally. The posterior ramus is likely to have contributed to the ventral margin of the cheek region, separating the quadratojugal from the maxilla, although only its anteriormost portion is visible. The rest of the ramus is reconstructed as a short triangular flange extending posterior to the maxilla.

3.1.16. Lacrimal

The lacrimals consist of a short, posterior suborbital ramus and an anterior, elongate, subrectangular lamina (Fig. 7a–c). No specimen shows a complete lacrimal, as remnants of this bone are invariably heavily disrupted and dislocated. However, it is possible to put together a composite reconstruction from preserved fragments of the left lacrimal in UMZC T.2013.3 (Fig. 5a, b) as well as from the shape and proportions of adjacent skull bones, particularly the prefrontal and the nasal (Figs 2b, c, 5). As reconstructed, the lacrimal contributes to part of the anterior and the anteroventral orbital margin. The ramus length is approximately one-quarter of the maximum orbit length, and deepens rapidly anteriorly before merging with the lamina, as in Silvanerpeton. The lamina contributes to the posterior margin of the external nostril and appears to be comparatively less deep than the lamina of Silvanerpeton. The shape and size of the prefrontal and nasal suggest that the dorsal margin of the lamina consisted of a shorter anterior tract and a longer posterior tract meeting at an obtuse angle. The ventral margin, in contact with the maxilla, probably showed a small anterior embayment corresponding to a ‘peak’ in the anterior part of the upper margin of the maxilla (see section 3.1.19). Overall, the lamina is less deep and with a less rounded dorsal margin than that of Silvanerpeton.

3.1.17. Squamosal

The subtrapezoidal squamosals are substantial bones (Figs 1, 2b, c, 7a–c). The anterior extremity of each squamosal (at the triple sutural joint between squamosal, postorbital, and jugal) reaches anteriorly almost to the level of the intertemporal mid length. Its posteroventral corner (in lateral aspect) projects distinctly behind the transverse level of the posterolateral processes of the tabulars. The posterior free margin of the squamosal forms an angle of ~53 degrees with the horizontal and is mostly straight except in its dorsalmost tract, where it produces a small shallow embayment immediately below the skull table. At this level, the bone reaches as far posteriorly as the posteriormost part of the supratemporal. However, there is no evidence that the squamosal of Eldeceeon contacted the tabular, unlike in Silvanerpeton. Along its dorsal margin, below the lateral temporal series, the squamosal shows a longitudinal thickening representing the lateral border of the articulation surface between the cheek and the skull table, possibly suggesting the presence of a hinge-like suture (Panchen Reference Panchen1970; Smithson Reference Schoch and Milner1985; Clack Reference Clack1987). In NMS G.1990.7.1, the ventralmost part of the posterior margin of the squamosal shows a curious depressed area seemingly devoid of dermal sculpture (Fig. 2c), which probably accommodated a small process from the quadratojugal in life. Along the course of the posterior margin, the surface of the squamosal forms a shallow and smooth flange delimited by an anterior ridge.

3.1.18. Quadratojugal

Although the quadratojugals are only partially preserved (Figs 1, 2b, c, 5), it is possible to reconstruct their posterior profile and most of their lower margin in lateral aspect (Fig. 7a–c). However, the course of the quadratojugal–squamosal suture, only partly visible in the holotype (Fig. 1), and the spatial relationship between the anteriormost part of the quadratojugal and the posterior ramus of the jugal are largely conjectural. The quadratojugal contributes to more than one-third of the depth of the posterior area of the suspensorium. Its ventral margin is smoothly convex and projects slightly below the level of the rearmost portion of the maxilla. Its anteriormost part is estimated to have extended approximately to the level of the posterior process of the postorbital in lateral view. Its short posterodorsal margin is in continuity with the squamosal free margin, and the transition between the two bones at this level is marked by a subtle change in slope. Below this slope, the quadratojugal terminates in a blunt squarish extremity presumably abutting against, and partially wrapped around, the lateral surface of the quadrate (Fig. 7c). As reconstructed, the quadratojugal of Eldeceeon appears deeper and more robust than in Silvanerpeton.

3.1.19. Maxilla

The shape of the maxilla conforms to that of most other early tetrapods (Figs 2b, c, 5, 7a–c). Its upper and lower margins converge towards one another anteroposteriorly. Its anterior margin is shallowly concave and contributes to the posterior half of the ventral margin and to the lower half of the posterior margin of the external nostril. The border for the nostril is visible in NMS G.1990.7.1 (Fig. 2c). The deepest part of the maxilla occurs slightly posterior to the nostril, where the upper margin of the bone shows a low triangular ‘peak’. Posterior to this level, the upper margin has a nearly straight profile, sloping slightly posteroventrally to a point situated below the mid-length of the suborbital process of the jugal. Posterior to this point, the upper margin shows a more pronounced slope such that the maxilla decreases rapidly in depth. A further change in slope is visible more posteriorly, aligned vertically with the posterior margin of the orbit. Behind this point, the maxilla forms an acuminate triangular process separated from the quadratojugal. Silvanerpeton differs from Eldeceeon in that the upper margin of its maxilla is nearly straight.

3.2.1. General features

Comparisons between the palates of Eldeceeon and Silvanerpeton are limited due to incomplete preservation in the former taxon (except for the pterygoid and the vomer; Figs 2b, c, 7d). Most likely, the palate was either closed or exhibited narrow vacuities, and some mobility may have characterised the palate–braincase junction (e.g., Panchen Reference Panchen1970; Smithson Reference Schoch and Milner1985; Clack Reference Clack1987).

3.2.2. Pterygoid

NMS G.1990.7.1 (Fig. 2b, c) supplies most of the information on the pterygoid. Combined data from the part and counterpart of this specimen suggest that the pterygoid has a proportionally more gracile palatal ramus and a more abbreviated quadrate ramus than its homologue in Silvanerpeton. The palatal ramus is vaguely lanceolate. Anteriorly, it terminates in a narrow triangular process in contact with its antimere along a straight line. The full anteroposterior extension of this contact is estimated to have been less than one-third of the projected length of the entire palatal ramus in ventral aspect. This is indicated by a gentle change in the curvature of the mesial margin of the ramus, which parts from its antimere to follow a smoothly convex course and continues to the level of the anterior process of the basipterygoid recess (Fig. 2c). Here, the margin turns abruptly mesially, resulting in the occurrence of a deep notch in the profile of the ramus. The deep basipterygoid recess has a narrow, sub-elliptical, and asymmetric profile. It is delimited by a gracile anterior process shaped like a thin bony sliver and a sturdier, mesiolaterally shorter, and subtriangular posterior process (Fig. 2c). The lateral margin of the palatal ramus is visible only in part, and the precise nature of its sutural contact with the palatine and ectopterygoid cannot be reconstructed. The quadrate ramus forms a triangular flange tapering rapidly posteriorly, and with a smoothly convex mesial margin and a semiparabolic lateral margin. Its proportions suggest that the adjacent subtemporal fossa of Eldeceeon would have been comparatively much smaller than that of Silvanerpeton. The anterior extremity of the lateral margin turns sharply laterally to produce a distinct, subrectangular lateral flange (Fig. 2c). This flange would have conferred a constriction to the anterior profile of the subtemporal fossa, but it is unclear whether a substantial anterior extension of the fossa was present lateral to the ectopterygoid. Preservation makes it impossible to ascertain whether the pterygoid contacted an internal process of the jugal (‘insula jugalis’ sensu Bystrow & Efremov Reference Bystrow and Efremov1940), such as is observed in Silvanerpeton. Despite diagenetic compression of the quadrate ramus, there is no evidence of a transverse flange along the area where the quadrate ramus continues onto the corpus (Fig. 2c).

3.2.3. Vomer

Both vomers are preserved (Fig. 2b, c). Although they are visible in dorsal and ventral aspect, their exposed surfaces are almost featureless, except for a number of irregular longitudinal striations and shallow grooves. Toothless vomers occur in other anthracosaurs and have at times been regarded as a characteristic feature of this group (e.g., Panchen Reference Panchen1970), although exceptions are known (e.g., Silvanerpeton; Ruta & Clack Reference Ruta and Bolt2006). The elongate and subrectangular vomers of Eldeceeon have nearly straight mesial margins and shallowly concave lateral margins forming most of the lateral choanal border. Their irregularly sinuous posterior margins are orientated slightly obliquely (anterolaterally to posteromesially). In this feature, they differ from the vomers of Silvanerpeton, which terminate in robust triangular projections flanking the anterior part of the mesial margins of the palatines. In Eldeceeon, the anteriormost extremities of the palatal rami of the pterygoids were probably wedged between the posteromesial corners of the conjoined vomers in life. The spatulate anterior extremity of each vomer would have been appressed against the internal surface of the basal portion of the premaxilla.

3.2.4. Palatine

Fragments of the left and right palatines are visible in NMS G.1990.7.1 (Fig. 2b, c), allowing us to reconstruct the anteriormost part of the palatine including the course of the choanal border. The palatine probably occupied a substantial proportion of the palatal surface, but its orientation and proportions remain largely conjectural. The choanal margin is subsemicircular. On the ventral side, a sharp straight ridge occurs adjacent to the mesial part of this border. This ridge runs anteromedially to slightly posterolaterally, widening slightly and becoming increasingly sharper posteriorly, before merging into the preserved part of the surrounding surface of the bone. Irregular rugosities and depressions are visible on the larger of the two palatine fragments in NMS G.1990.7.1.

3.2.5. Ectopterygoid

No preserved bony fragment can be convincingly assigned to the ectopterygoid, as the relevant region of the palate is covered by the lower jaw rami (but see comments on the palatal dentition). As in the case of the palatine, the ectopterygoid would have formed a substantial rectangular plate, based upon general skull proportions.

3.2.6. Quadrate

A small, subrectangular bony fragment immediately posterior to, and partly overlapping, the posterior extremity of the quadratojugal in NMS G.1990.7.1 (Fig. 2b, c) is interpreted as a partially preserved quadrate. Its surface is almost featureless, except for the occurrence of weak elongate rugosities. The direction of its greatest elongation may correspond to the dorsoventral orientation of the bone, but no other details are visible.

3.4.1. Dentary

The dentary is ~70 % as long as the ramus (Fig. 2b, c, 7a, b). In lateral aspect, its anterior margin is smoothly convex. Its upper margin is very shallowly concave in its anterior half, but nearly straight and sloping gently posterodorsally in its posterior half. Combined information from the preserved portions of the upper margins of the infradentaries suggests that the ventral margin of the dentary shows a sudden change in slope at the triple sutural joint between dentary, angular, and postsplenial. Anterior to this joint, the depth of the bone is approximately constant, whereas posterior to it, it decreases rapidly. Posterior to the triple sutural joint between dentary, surangular, and angular, the dentary forms a finger-like projection extending slightly posterior to the last dentary tooth and appressed against the surangular, immediately anteroventral to the surangular crest.

3.4.2. Splenial and postsplenial

The extent of the lateral exposure of these two bones, such as is observed in NMS G.1990.7.1 (Fig. 2b, c, 7a, b), may have been exaggerated by compaction. However, it is clear that both contribute to a significant proportion of the projected lateral aspect of the ramus, and are approximately equal in length. The splenial is a narrow sliver of bone, narrowing gently to a point immediately behind the anterior extremity of the dentary and forming a nearly straight suture with the dentary, in continuity with the dentary–postsplenial suture. The lateral projection of the postsplenial is deeper than that of the splenial. Its depth increases slightly from the level of the splenial–postplenial suture to the level of the triple sutural joint between dentary, angular, and postsplenial, situated at the mid length of the postsplenial. Posterior to this level, the postsplenial decreases gradually in depth. The rearmost part of its lateral surface is aligned vertically with the triple sutural junction between dentary, surangular, and angular.

3.4.3. Angular

Most of the lateral surface of the angular (Fig. 2b, c, 7a, b) is visible, except for a small portion of its sutural contact with the surangular. The bone is deepest at the level of its triple sutural junction with the dentary and the surangular. Anterior to this point, the anterodorsal margin of the bone slopes sharply, anteroventrally forming an irregularly convex suture with the dentary. Its posterodorsal margin slopes gently posteroventrally. Its ventral margin of the angular is divided into a smoothly convex posterior half continuing seamlessly into the posterior margin of the surangular, and a nearly straight anterior half.

3.4.4. Surangular

Despite uncertainties concerning the precise course of its ventral margin, most of the surangular can be reconstructed (Figs 1, 2b, c, 7a, b). It is a sturdy subtrapezoidal bone increasing slightly in depth anteroposteriorly, with a gently convex posterior margin and a nearly straight upper margin. The anterior extremity of the upper margin ends in a small but distinct ‘step’-like surangular crest, immediately dorsal to the rearmost extremity of the dentary. A horizontal strip of the lateral surface of the surangular, immediately below the dorsal margin of the bone and extending for about one-fourth of its greatest depth, shows a poorly pronounced sculpture and is likely to have accommodated the internal surface of the quadratojugal in life (see Klembara Reference Klembara1997).

3.4.5. ?Prearticular

In the middle one-third of the left ramus of NMS G.1990.7.1 (Fig. 2b, c), the lower jaw bones appear to be dislocated. Within the gap resulting from disruption, an irregularly striated bony surface, presumably in mesial aspect, is visible. This surface is tentatively interpreted as belonging to the prearticular, but no salient features can be seen.

Coronoids cannot be observed in any specimen.

3.7.1. Pectoral girdle

We limit our description of the pectoral girdle to a few salient features of the interclavicle only. For further details, see Smithson (Reference Smithson1994). Particularly noteworthy is its shape, which readily distinguishes Eldeceeon from Silvanerpeton. In Silvanerpeton, the subrhomboidal anterior ventral plate of the interclavicle transitions smoothly into the elongate triangular parasternal process (Ruta & Clack Reference Ruta and Bolt2006, fig. 10a) In Eldeceeon, the broadly fan-shaped anterior ventral plate is sharply delimited from the narrow and rod-like parasternal process (Figs 1, 2a, 3a, 4). Although incomplete in UMZC T.2013.3 (Fig. 2a, 6), the interclavicular plate shows a large, sub-elliptical, posterior sculptured region, the surface of which shows irregular depressions, pits, ridges, and grooves. In this specimen, the anterior fringe projecting anteriorly from the smoothly convex anterior margin of the sculptured region is not preserved. The fringe, visible in the holotype (Fig. 1) and in NMS G.1990.7.1 (Fig. 2a), consists of straight radiating ridges separated by narrow sulci, and extends for approximately the same length as the sculptured region, becoming increasingly narrower laterally. The posterolateral margins of the sculptured regions are gently sinuous, with a lateral convexity and a medial concavity, and turn sharply posteromesially at the transition between the plate and the parasternal process. There is some evidence that, immediately behind this point, the anteriormost part of the parasternal process is wide and robust but tapers smoothly posteriorly, such that its straight lateral margins converge, gently posteriorly terminating in a narrowly spatulate end. The ventral surface of the process shows a delicate sculpture of fine striations and pits. The morphology of the interclavicle is unique among anthracosauroids, as far as we can tell, but does bear some similarities with the interclavicles of Ichthyostega (Jarvik Reference Jarvik1996), Westlothiana (Smithson et al. Reference Smithson1994), possibly Solenodonsaurus janenschi (Carroll Reference Carroll1970; Laurin & Reisz Reference Laurin and Reisz1999; Danto et al. Reference Danto, Witzmann and Müller2012), and especially seymouriamorphs such as Seymouria baylorensis (White Reference Swofford1939), Discosauriscus austriacus (Klembara & Bartík Reference Klembara and Bartík2000), Utegenia shpinari (Klembara & Ruta Reference Klembara and Ruta2004b), and Ariekanerpeton sigalovi (Klembara & Ruta Reference Klembara and Ruta2005b). In several seymouriamorphs, in particular, the parasternal process is elongate and often with parallel sides, and merges into a fan-shaped or rhomboid interclavicular plate. However, it often terminates in a fringe-like extremity consisting of ‘splayed-out’, small digitiform processes. In addition, a ‘swelling’ occurs in the anterior part of the process.

3.7.2. Fore limb

The humerus was thoroughly described by Smithson (Reference Smithson1994), based mostly upon NMS G.1990.7.1. A few additional remarks are provided here using information from UMZC T.1350. Although the humeri of NMS G.1990.7.1 and UMZC T.1350 (Figs 2a, 3b, 6a, c) appear slightly different, this is mostly due to incomplete preservation, which also prevents detailed observation of surface features, including processes and crests (but see remarks on possible ectepicondylar ridge below). In both specimens, part of the anterior portion of the bone and the anteriormost part of the humeral head are heavily disrupted. Each humerus shows a conspicuous and elongate proximal half with a stout and moderately elongate shaft, similar to the condition in Silvanerpeton. The posterior margin of the shaft is gently convex in its proximal two-thirds and slightly concave in its distal one-third. Its course slants slightly anterodistally before turning sharply posteriorly along a short, narrow curved edge, marking the transition to the proximal margin of the entepicondyle. This transition is smoother and comparatively broader in UMZC T.1350 (Fig. 6a, c). The entepicondyle forms a distinct subrectangular flange, conferring the well-known L-shaped profile to the humerus, such as is observed in various other early tetrapods (Clack Reference Clack2012). There is no clear evidence of an ectepicondylar ridge in any of the preserved humeri, certainly as a result of preservation, although a poorly preserved thickening appears just distal and slightly anterior to the point where the posterior margin of the shaft and the proximal margins of the entepicondyle converge in NMS G.1990.7.1 (Fig. 2a). The thickening can be followed for a short distance distally, before it disappears in the central part of the extensor surface of the entepicondylar flange. In UMZC T.1350, there is some evidence of a shallow sub-elliptical depression close to the anteriormost part of the posterior margin of the entepicondyle, in the position that would normally be occupied by the entepicondylar foramen (see also discussion in Ruta & Clack Reference Ruta and Bolt2006 and Smithson & Clack Reference Smithson, Carroll, Panchen and Andrews2018). However, whether the depression in question does correspond to a poorly preserved foramen remains a moot point. Most of the above observations are based upon NMS G.1990.7.1. The entepicondyle of UMZC T.1350 is partially preserved, presumably missing some of its posterior and posterodistal portions. In the development of a substantial shaft and in the outline and proportions of the entepicondyle, the humerus of Eldeceeon resembles closely those of Archeria and Proterogyrinus (Romer Reference Romer1957; Holmes Reference Holmes1984, Reference Holmes1989), as well as some of the tetrapod humeri from the Tournaisian of Horton Bluff in Nova Scotia (Anderson et al. Reference Anderson, Smithson, Mansky, Meyer and Clack2015), but differs somewhat from the best preserved humerus in Silvanerpeton, such as was described by Ruta & Clack (Reference Ruta and Bolt2006, fig. 10b).

The radius is poorly preserved in most specimens. In the holotype (Fig. 1), the outlines of the radii are almost complete but poorly discernible from the surrounding matrix. The best-preserved radius is observed in UMZC T.1350 (Fig. 6a, c). As preserved, it does not show obvious features that distinguish it from the radii of other anthracosauroids, although it is broadly similar to that of Archeria. The bone has a typical ‘dumbbell’-like profile, with expanded extremities and a robust shaft. The overall morphology of the bone conforms to the description provided by Panchen (Reference Panchen1970, p. 34) for the radius of Archeria, which is similar to that of Eldeceeon: ‘… a flattened cylinder with expanded ends bearing terminal articular surfaces; the one for the humerus being roughly circular, that for the carpus somewhat dorso-ventrally [sic] flattened, i.e. in the extended horizontal position.’

Ulnae are observed in all specimens (Figs 1, 2a, 3b, 4, 6a, c), albeit in various degrees of completeness and preservation. In the best-preserved examples (UMZC T.2013.3 and T.1350; Figs 4, 6a, c), the bone is characteristically slender (conforming mostly to the generalised anthracosauroid pattern) and, unlike its homologue in Silvanerpeton (Ruta & Clack Reference Ruta and Bolt2006, fig. 6a, c), it exhibits a remarkably robust olecranon process. In its general proportions, the ulna of Eldeceeon resembles those of Archeria (Romer Reference Romer1957; Holmes Reference Holmes1989) and Westlothiana (Smithson et al. Reference Smithson1994.), but is unlike that of Proterogyrinus (Holmes Reference Holmes1984), in which the bone appears sturdy and with a less pronounced olecranon process. The best-preserved ulna occurs in UMZC T.2013.3 (Fig. 4). In this specimen, the olecranon process is subparabolic in lateral aspect, with a smoothly curved proximal margin, a gently but distinctly convex posterior margin, and an imperceptibly concave anterior margin sloping anteroventrally before merging into the proximal extremity of the ulna. The distal margin of the ulna slants considerably and is divided into a small, posteroventrally straight segment presumably articulating with the ulnare, and a slightly longer, oblique, and irregularly sinusoidal anterior segment presumably articulating with the intermedium (Figs 1, 4, 6a, c).

For a complete description of the anterior autopod, the reader is referred to Smithson (Reference Smithson1994). However, the heavily disrupted autopod of UMZC T.1350 (Fig. 6a, c) shows 21–22 elements of the manus. Only some of these show a reasonably complete outline and they add little additional information to the pattern of bones in the manus. Terminal phalanges (unguals) are represented by two or three triangular elements with concave sides and pointed distal extremity. These do not show any evidence of the lateral flange-like expansions documented in the pedal unguals of Silvanerpeton (Ruta & Clack Reference Ruta and Bolt2006, fig. 10d).

3.7.3. Pelvic girdle

Combined information from all specimens provides a nearly complete picture of the morphology of the pelves (Figs 1, 2a, 4, 6a, b). In general proportions, they resemble those of Silvanerpeton, except that in that animal, the pubis appears unossified. By contrast in Eldeceeon, the puboischiadic plates are fully ossified in some specimens and show complete peripheral margins. The ilium consists of two major parts: a compact ventral corpus that sutures with the puboischiadic plate, and a robust neck that connects the corpus to a stout dorsal blade and an elongate post-iliac process.

The outline of the dorsal blade varies slightly in different specimens, but some of this variation reflects, in part, disruption and/or incomplete preservation. Its morphology appears to be unique, certainly among anthracosauroids, as far as we can tell. In the holotype (Fig. 1), the blade shows a complete outline, as reconstructed in Smithson (Reference Smithson1994). In this specimen, the transition between the anterior margin of the neck and the anterior edge of the blade is marked by a subtle change in curvature, such that the lowermost part of the edge of the blade detaches from the margin of the neck following a short vertical tract, before turning sharply posterodorsally. From this point, the anterior edge of the blade is nearly straight and can be followed along an oblique direction up to the dorsalmost point of the blade. The latter forms a small, blunt-topped, subtriangular ‘peak’. From the ‘peak’, the margin of the blade continues posteroventrally, forming a shallow embayment dorsally followed by a broadly convex tract more ventrally, followed, in turn, by a slightly deeper embayment at the transition between the blade and the dorsal margin of the post-iliac process. The morphology of the blade in UMZC T.2013.3 (Fig. 4) conforms to the pattern described above, but detailed observations are hampered by slight disruption and the fact that the blade occurs in close proximity to disrupted skeletal elements, including a possible neural arch and some gastralia.

The post-iliac process has straight and parallel upper and lower margins, and terminates in a blunt squarish extremity. This pattern is conserved across all specimens. The process is orientated subhorizontally or slightly posterodorsally, extending nearly as far posteriorly as the rearmost extremity of the ischium.

Complete puboischiadic plates are observed in UMZC T.2013.3 and UMZC T.1350 (Figs 4, 6a, b), but sutures between the pubic and ischiadic portions are not discernible. Compaction implies that such plates are slightly flattened against the bedding plane. The ischia are elongate and subtrapezoidal, narrowing slightly in an anteroposterior direction. They are preserved in plan view in the holotype (Fig. 1), where their entire outline is clearly visible. Each ischium contacts its antimere along a straight ventral margin. We are unsure as to the precise course of this margin in side view, although there is some evidence (Fig. 4) that it may have been gently convex. The posterior margin of the bone is shaped approximately like a quarter of a circle (the posterior margins of both pubes would thus delimit a ‘basin’ in life), although this curvature would appear less accentuated in lateral view and has been altered by compaction (Figs 4, 6a, b). The posterior and dorsal margins of each ischium meet at a slightly obtuse angle (Fig. 1). The dorsal margin is shallowly convex throughout most of its length in the holotype and this convexity is somewhat discernible in other specimens as well. At its anteriormost extremity, the dorsal margin turns gently upward following a smoothly concave course, before merging into the posterior margin of the corpus of the ilium. This concave tract is seen in the holotype, but is scarcely visible in other specimens (Figs 4, 6a, b).

The pubis is much shorter than the ischium, its estimated length being less than half of the length of the latter bone. Although no isolated pubes are preserved along the bedding plane, those in articulation with the rest of the pelvic girdle possess a squarish or subtrapezoidal outline, with a distinct anteroventral corner and a smoothly convex anterodorsal edge connecting the anterior and dorsal margins of the bone (Figs 4, 6a, b). Its ventral margin may have been straight or gently concave in side view (Fig. 4).

No specimen shows a clearly defined acetabulum. A small depression at the junction between the anterior part of the corpus of the ilium and the dorsal part of the ischiadic plate in UMZC T.1350 (Fig. 6b) is tentatively interpreted as the poorly preserved anterodorsal section of the acetabulum.

3.7.4. Hind limb

As noted previously (Smithson Reference Smithson1994; Clack & Milner Reference Clack and Milner2015), a distinctive feature of Eldeceeon is represented by its large and robust hind limbs. In NMS G.1990.7.1 (Fig. 2a), the length of the femur (measured as the greatest distance between the most proximal projection of its head to the most distal projection of its posterior, or fibular, condyle) is almost equal to the length of six trunk vertebrae. Similar proportions are estimated for the holotype (Fig. 1). The length of the femur exceeds that of the ischia (Fig. 1) and is comparable with the estimated length of the puboischiadic plates (Fig. 4). The best-preserved femora are observed in the holotype and in NMS G.1190.7.1 (Figs 1, 2a). In both specimens, their morphology conforms to the pattern of other anthracosauroids (e.g., Romer Reference Romer1957; Panchen Reference Panchen1970; Holmes Reference Holmes1984, Reference Holmes1989; for easily accessible comparisons of femora among selected early tetrapods, see Ruta et al. Reference Ruta, Krieger, Angielczyk and Wills2001, fig. 12). The proximal and distal extremities of the bone are greatly expanded and of similar width, and are well delimited from the shaft. The shaft is relatively stocky and abbreviated. Both the anterior and the posterior margins are embayed. The distal condyles are robust, the fibular condyle being only slightly wider and projecting slightly more distally than the tibial condyle, and there is evidence of a short intercondylar space (albeit its outline is masked by compaction). In some of these features, the shape of the femur most closely resembles that of Archeria crassidisca (Holmes Reference Holmes1989; Ruta et al. Reference Ruta, Krieger, Angielczyk and Wills2001, fig. 12j) and, to a lesser degree, Proterogyrinus scheelei (Holmes Reference Holmes1984; Ruta et al. Reference Ruta, Krieger, Angielczyk and Wills2001, fig. 12k). In both these taxa, the tibial condyles are more robust than those of Eldeceeon, and comparable in size to the fibular condyles. A final note concerns the femora of NMS G.1990.7.1 (Fig. 2a), where the internal and fourth trochanter are discernible. In both femora, as preserved, the anterior margin of the proximal extremity protrudes slightly anteriorly forming a small, proximodistally elongate rectangular flange. We interpret this flange as the anteriormost projection of the internal trochanter (e.g., Panchen Reference Panchen1970, p. 37) in the extensor–flexor plane. The most proximal part of this flange terminates in a small and blunt process, representing the fourth trochanter (see Ruta et al. Reference Ruta, Krieger, Angielczyk and Wills2001, fig. 12).

All specimens show at least one complete or nearly complete tibia (Figs 1a, 2a, 4). In NMS G.1990.7.1, the proximal extremity of a contralateral element is also visible (Fig. 2a). The tibia of Eldeceeon is broadly similar to that of Silvanerpeton (Ruta & Clack Reference Ruta and Bolt2006, figs 6a, 9a, b, 10c), but there are subtle differences between these two taxa. In Eldeceeon, the tibia has a more robust appearance with a comparatively broader and stouter proximal extremity than that of Silvanerpeton. In NMS G.1990.7.1, the complete outline of a fairly well-preserved tibia can be followed along the extensor/flexor plane of the bone. The bone shows wide proximal and distal extremities and a distinct short shaft. On the articulation surface of the proximal extremity of the tibia are two shallow condylar areas of slightly unequal extension, the anterior one appearing only marginally larger and shallower than the posterior one. The slightly raised area between the two condylar areas corresponds to the intercondylar ridge, an unremarkable low skeletal prominence in several anthracosauroids and other early tetrapods (e.g., Romer Reference Romer1957; Panchen Reference Panchen1970; Holmes Reference Holmes1984, Reference Holmes1989), but well developed in Eldeceeon. Barring preservation artefacts, the projection of this ridge in the preserved view of the tibia is divided into a smaller anterior and a larger posterior eminence or tubercle. These eminences delimit a shallow space between them, presumably corresponding to an intercondylar groove in life. The distal extremity is vaguely subtrapezoidal and its anterior margin is slightly shorter than its posterior margin. As in the case of the proximal extremity, the margins end in distinct outer and inner angles. The distal margin has two distinct sections of approximately equal extension separated by a small triangular protrusion, for the articulation with the tibiale and intermedium bones. Their course is irregular, possibly as a result of preservation, but broadly concave along the intermedium contact and vaguely sinuous along the tibiale contact.

Both fibulae are preserved in the holotype (Fig. 1), in NMS G.1990.7.1 (Fig. 2a; only one element is visible in full), and in UMZC T.2013.3. Unlike Silvanerpeton (Ruta & Clack Reference Ruta and Bolt2006, figs 6a, 9a, b, 10c), Eldeceeon possesses a comparatively more gracile fibula resembling that of anthracosauroids, such as Archeria and Proterogyrinus (e.g., Romer Reference Romer1957; Panchen Reference Panchen1970; Holmes Reference Holmes1984, Reference Holmes1989). As preserved along the extensor/flexor plane, the fibula shows a very slightly expanded proximal extremity, with an irregularly convex proximal margin, a distinct anterior extension ‘jutting out’ towards the tibia, and an inconspicuous posterior extension. The shaft is slender and elongate and its anterior and posterior margins merge indistinctly into those of the proximal and distal extremities of the bone. The distal extremity forms a flat subtriangular blade with a strongly convex posterior margin and a gently sinuous anterior margin, which would be in contact with the intermedium (anteriorly) and fibulare (posteriorly).

Although heavily dislocated, the bones of the pes are preserved in close proximity in at least two specimens (Figs 1, 2a, 3c), permitting reasonable estimates of the proportions of the pes and individual digits (Smithson Reference Smithson1994). In other specimens, the pedal elements are too disrupted to permit accurate estimates of pedal proportions (Figs 4, 6a). We agree with Smithson's (Reference Smithson1994) reconstruction of the pes as showing at least three proximal and five distal tarsal elements (e.g., Fig. 1), and we include a new reconstruction of the tarsus and pes in our new reconstruction of the skeleton (Fig. 7e). In NMS G.1990.7.1 (Figs 2a, 3c), the individual elements of a heavily disrupted pes are particularly well preserved. Scattered in close proximity to one another are five metatarsals and numerous phalanges. For a detailed account of the morphology of individual elements in this specimen, the reader is referred to the tracing of the pes (Fig. 3c). Tarsal and phalangeal elements appear to have been ‘smeared’ distally, not far from the tibia and fibula. Around each of these two bones, as well as between them, is a scatter of tarsals, three of which are distinctly larger than the rest, and are consistent with their interpretation as proximal tarsals. One of these, visible between the anterior corner of the proximal extremity of the tibia and the posterior margin of the shaft of the fibula, as preserved, is a conspicuous polygonal element, the shape of which is consistent with that of a pedal intermedium (e.g., Romer Reference Romer1957; Holmes Reference Holmes1984, Reference Holmes1989; Smithson et al. Reference Smithson1994). Overlapped by the posterior corner of the distal extremity of the fibula, and arranged almost perpendicular to the latter, is a displaced metatarsal. Further to the right in the tracing of the pes, as figured, is a series of four metatarsals, arranged in approximately anatomical succession. Assuming that metatarsal I is the smallest of these four elements, it would correspond to the leftmost bone in the series. We note a curious gap in the series of four elements towards the rightmost of these. This gap appears to be well suited for the position of the metatarsal that lies transverse to the fibula, which would thus correspond to metatarsal IV. If our interpretation is correct, then the rightmost element in the series of four is metatarsal V. Metatarsals III and IV are approximately equal in length and comparable in robustness, and are slightly longer than metatarsals II and V. As for the phalanges, we count 18 elements scattered to the right of the metatarsals, as figured. This number would be consistent with an estimated phalangeal count of 2-3-4-5-4, but there is a difficulty. In the scatter of 18 bones, two or three are easily recognisable as distal phalanges (unguals), based on their small size and shape, being characteristically subtriangular, with deeply embayed lateral and mesial margins and a pointed distal tip. This, however, implies that at least another two or three unguals are undetected, unrecognised, or not preserved. Perhaps some poorly preserved bony fragments close to the tibia may represent unguals, but their interpretation is difficult. Assuming that our interpretation of the pes of NMS G.1990.7.1 is approximately correct, this raises the possibility that the phalangeal count of the pes may have been slightly higher.