Parietal shield.

Latex casts of the two portions of the skull roof (parietal and postparietal shields), displaced from each other in the rock (Fig. 2C), are illustrated in approximate life position in Figure 3. The parietal shield (incomplete anteriorly) is preserved in part (piece a4) and counterpart (piece a11), the former (external impression) showing characteristic derived tristichopterid features (Fig. 3) such as the shape of the parietal and intertemporal, and the tear-shaped pineal complex close to the posterior margin. There is a small subsidiary pineal plate on the right side, presumably an individual variation as documented for Eusthenodon ([36]: fig. 38), although that form does not show the same pattern. Interestingly, an almost identical pineal configuration to the holotype occurs in a specimen of Cabonnichthys ([16]: fig. 4B). The pineal opening of the holotype was evidently lost in the saw-cut. The left intertemporal (IT) is slightly displaced, showing a mesial overlap area (od.IT), with another on the posterolateral corner (od.Po). Both sides show a clear overlap for the posterior supraorbital (od.So2), that straddles the suture between the parietal (Pa) and the anterior supraorbital (So1). A short margin in front of the pointed anterior end of the intertemporal (forming a distinct notch in the right lateral skull margin), demonstrates contact between the parietal and the postorbital of the cheek, as in both Eusthenodon and Mandageria, but unlike Cabonnichthys. The posterior part of a large median postrostral (Ptrp) and adjacent bones of the nasal series (Na) are preserved in position. The central part of the skull in front was lost to weathering on the surface of the outcrop, but the left anterolateral skull margin around to the midline is separately preserved in a set of small interlocking pieces, described below.

Additional information on the main shield is provided by the internal impression (a11), showing the triple junction at the posterior end of the posterior postrostral is 88 mm in front of the saw-cut, whereas it is only 48 mm in front on the external impression. This is the standard overlap relationship of these bones, but the overlap is much more extensive than in Eusthenopteron ([2]: fig. 14). The internal impression inside the right lateral margin is very similar to that preserved in Notorhizodon ([8]: fig. 22A). The point of radiation for the ossification center of the postrostral on the visceral surface is ∼12 cm from the saw-cut; in Eusthenopteron it is anteriorly placed at about one third the length of the bone [2], and a similar length is indicated in Eusthenodon ([36]: fig. 37A). Even assuming the ossification center was closer to the front of the postrostral in Edenopteron, it must have been more elongate than in Eusthenodon. This seems to be the case also in Langlieria Clément et al., 2009 [30] from Belgium, with a much larger median postrostral than the associated Eusthenodon ([37]: figs. 3B, T).

The posterior supraorbital has not been found in the holotype, even if its overlaps are clear on both sides (od.So2, Fig. 3). The bone itself is preserved in ANU V3478, which comprises a badly crushed left side of a skull with cheek attached, bisected by the main saw-cut (Figs. 11, 12). The posterior supraorbital was evidently less elongate than in Eusthenopteron, where it arched over the orbit. In Edenopteron it was positioned mainly behind the orbit, but with a somewhat different shape to the tear-shaped posterior supraorbital of Eusthenodon, with a slightly T-shaped form resulting from a lateral process projecting behind the orbit to contact the jugal. The anterior edge is lost in the saw-cut, but overlaps on the skull of the holotype show it must have had a triangular shape, like Mandageria (AMF 96855) ([15]: fig. 10a). In Cabonnichthys the supraorbital has a similar prominent lateral process, but at least in the holotype (AMF 96858) there was an extra mesial process giving a quadrilateral rather than triangular shape ([16]: fig. 3A). In that form it was restored enclosing most of the orbit, whereas in Edenopteron the orbital margin on the anterior supraorbital is clearly seen (orb.m). The lateral process of the ‘T’ on the posterior supraorbital clearly excluded the postorbital bone from the orbital margin, as in Mandageria, Cabonnichthys, and Eusthenodon, but not Eusthenopteron.

The orbital margin on both anterior supraorbitals of the holotype is only slightly concave, so the bone contributed only to the dorsal margin of the orbit. A well preserved overlap area on the right side (od.La) forms a distinct right angle, in contrast to the curved overlap area in this position for the lachrymal in Eusthenopteron. The lachrymal of Edenopteron has only a rounded anterior angle in front of the orbital notch (well preserved on the right cheek; Figs. 9, 10) so it is assumed that the anterior part of this overlap area was covered by the lateral rostral. On the counterpart the left anterior supraorbital has a well preserved margin in front of the lachrymal overlap, slightly notched where another suture seems to run anterolaterally, forming a triple junction, presumably the anteromesial corner of this bone. The right supraorbital has a pointed anterior margin, mesial to which the posterior corner of another bone is preserved, possibly the tectal (?Te), or perhaps the next element in the nasal series.

The external impression shows a clear mesial suture between the anterior supraorbital and the posterior nasal series on both sides (flattened on the left side, but retaining the lateral curvature on the right). The posterior suture of the posterior nasal is obscured by cracking, but indicates a posterior prong into the parietal on both sides, as in Eusthenodon but evidently more pronounced. The central skull region in front was completely weathered away on the holotype, the next preserved bone being the anterior part of the anterior postrostral (sandwiched between the premaxillae; see Figs. 4A, 5A). In the missing region, two smaller nasals and a larger pair meeting in the midline can be assumed after Jarvik’s [19] Eusthenodon reconstruction; there is an indication of a midline suture between nasals behind the anterior postrostral in a separate piece preserving the snout (Na, Fig. 5A).

Post-parietal shield.

The external surface is preserved mainly on piece a11, with the right tabular adjacent to the sawn edge on piece c6, and internal surfaces preserved on pieces b5–7. This unit was ∼112 mm long. The postparietals were slightly displaced (PPa, Fig. 3), their median suture showing a slight interdigitation near the posterior margin as in some other tristichopterids (e.g. Cabonnichthys from Canowindra). The well preserved left supratemporal (St) shows a deep spiracular notch (spir) at its posterior suture with the tabular (Ta). In front is a distinct lateral process that indented a notch in the dorsal cheek margin, apparently more pronounced than in Eusthenodon, but not as marked as in some Cabonnichthys (e.g., AMF 96856), where it may form a distinct anterolateral projection ([16]: fig. 4B). The right tabular is relatively complete across the saw-cut but slightly flattened, with a short transverse pitline at its center (pl). Transverse and longitudinal pitlines over the postparietal ossification centers, as in other tristichopterids, are indistinct and obscured by cracks in Edenopteron.

The posterior margin of the post-parietal shield is incomplete and partly lost in the saw-cut. It can be interpreted after the configuration in Eusthenopteron, with a central part that abutted or overlapped the median extrascapular, and lateral parts with a projecting posterior lamina ([2]: fig. 14). The same arrangement is shown in Eusthenodon ([19]: fig. 24). For Notorhizodon, the median extrascapular (unknown in Edenopteron) suggests that posterior skull projections met in the midline ([8]: fig. 26A), perhaps approaching the condition in Mandageria where the median extrascapular narrows almost to a point [16]. A specimen of Marsdenichthys recently described (NMV 179619) seems not accurately restored ([7]: fig. 3A), as the latex cast shows a central part of the posterior skull margin which evidently overlapped the median extrascapular, with a distinct process and lateral notch where the posterior pitline, which passed across onto the lateral extrascapular, the latter bone evidently overlapping the skull. For Edenopteron similar processes are suggested on the holotype (Fig. 3). The postparietals are crushed and incomplete on a second specimen (Fig. 11), with the left bone displaced back against the midline suture, and the more extensive right postparietal showing a strong posterior overlap for the lateral extrascapular (od.Esc, Fig. 12).

Snout.

Piece a8 of the holotype attaches at the front of the main palate impression (on piece b7) and preserves the internal impression of the dermal bones of the snout (Fig. 4B). Piece a8 sits ∼16 mm higher than the palate surface; both premaxillae meet in the midline (Pmx, Fig. 5), fixing the midline position of the anterior edge of the snout at about 20 cm in front of the saw-cut at the posterior edge of the parasphenoid. Noteworthy is a large tusk (t.Pmx) and circular attachment for a corresponding tusk near the midline suture between premaxillae. Premaxillary tusks are known in the Canowindra tristichopterids (Mandageria, Cabonnichthys), and various other sarcopterygians including an undescribed species of Eusthenodon [15], but they do not occur in Eusthenopteron or Platycephalichthys. These ‘pseudofangs’ also occur in Langlieria Clément et al., 2009 [30] from Belgium, and Bruehnopteron Schultze & Reed, 2012 [31] from Nevada. We use the terminology of ‘tusk’ rather than ‘fang’ [15], the latter reserved for large teeth in alternating replacement pairs as on the dermal palate bones and coronoids of the lower jaw. The snout of Edenopteron clearly had a less pointed shape than restored for Mandageria [23]. The premaxillary tusks sit on a thickened ridge just inside the anterior margin of the premaxilla (la.pal, Figs. 5B, 14A). Piece g2 (with preserved bone and tooth tissue on the counterpart h2; assumed to belong to ANU V3478) shows an impression of closely spaced tusks evidently folded back beneath the premaxillae (Fig. 14B). These also sit on a thickened ridge (la.pal), in front of which is a poorly preserved anterior margin of similar shape to that of the holotype. As preserved these resemble the premaxillary tusks figured for Langlieria ([30]: fig. 4). Behind this ridge the holotype shows a shallow v-shaped notch opening into a distinct anterior cavity (Fig. 4B), an unusual configuration because the prenasal fossa, floored by the ethmoidal ossification of the neurocranium, is usually visible in this position ([5]: fig. 5). In the Edenopteron holotype the floor of this cavity shows posterior radiating striations of the postrostral plate (Fig. 5B), indicating an anterior ossification center for this bone. The anterior postrostral has a central ossification center in Eusthenopteron, and also Eusthenodon ([19]: fig. 26B). The midline suture between premaxillae is clear internally between the tusks (Figs. 4B, 5B), but externally is obscured by a crack (junction of pieces a7, a14). It is assumed they met externally in front of the anterior postrostral, which shows a very clear anterior margin on the right side (Ptra, Figs. 4A, 5A). The posterior margin is less clear, but indicates the postrostral was about 75% as broad as long, whereas in Eusthenodon it is equilateral, and in Eusthenopteron it is broader than long. The triple junction between the postrostral, right nasal and premaxilla is also clearly preserved. Presumably the nasals met behind in the midline, but this region is poorly preserved (dashed line, Fig. 5A).

Piece a7 (left premaxilla) connects around the left skull margin with pieces a10 and e2 (Remigolepis V2378) to preserve external bone impressions for about a 20 cm distance from the midline (Fig. 7A; internal impressions preserved on a13, b6, 7). Immediately adjacent, the distinctive anterodorsal overlap of the postorbital bone (od.Pa/IT; described below) demonstrates that the left cheek was still placed against the skull, but separated and offset by a fracture. The preserved left external anterolateral margin of the skull lies ∼60 mm from the anterior preserved end of the parietal skull portion, which as noted above had been rotated anti-clockwise so that its midline is almost 90° from the palate and marginal jaw bones (Fig. 2C). In lateral view the premaxilla shows a pronounced posterior process (Pmx, Fig. 8A), slightly pulled apart from the adjacent bone, the latter with a rounded dermal process projecting anteromesially into a notch of the premaxilla (R.l, Fig. 7E). Underneath, a strong mesial lamina projecting inwards some 10 mm (pr) appears to be part of this posterior bone, although no corresponding structure is evident in restorations of the lateral rostral in Eusthenopteron ([38]: fig. 53). There is some uncertainty about the shape of the lateral rostral and adjacent bones in Edenopteron, although certain aspects are clear.

About 25 mm behind the anterior end of the lateral rostral (on piece a10) is a distinct shallow notch on the dorsal margin (nn, Figs. 7D, 8A, C), delimited behind by a smooth mesial surface projecting in at about 90° to the external surface. This notch is interpreted as lower edge of the external nasal opening. It is about 15 mm across, with the smooth mesial projection (pr.dim, Fig. 7D) about 10 mm long, and showing a patch of very fine denticulation (dent, Fig. 8C) perhaps corresponding to the special ornament inside the nasal opening of Gogonasus ([39]: fig. 13). Smooth bone or with very fine ornament also lines the external nasal opening and process dermintermedius of the lateral rostral bone in Eusthenopteron ([38]: pl. 6, fig. 2). However on the rock surface this denticulated area is separate from the lateral bone impression, so our interpretation is provisional, although the denticles are clearly seen to be separate from the adjacent impressions of scales (sc, Fig. 7A).

About 20 mm behind this nasal notch, a clear bevelled margin with a raised rim ∼15 mm long could either be the orbital margin of the lachrymal (Fig. 7A), or defining the upper (mesial) margin of an elongate lateral rostral (Figs. 7D, 8C). The latter interpretation gives a completely different shape to the lateral rostral of Eusthenodon, well preserved in the holotype ([19]: pl. 10), and demonstrating a morphology and triangular shape very similar to that of Eusthenopteron. In Edenopteron the lateral margin beneath the nasal notch forms a shallow embayment with fine rugose ornament extending back past the next crack (onto piece e2), and the preservation suggests that not much is missing (Figs. 7D, 8C). Alternatively, if part of the ornamented surface represents the lachrymal, bounding the orbit ventrally, there is no sign of a suture in the correct position (dashed line, Fig. 8A). A shallow sulcus of unknown function (sulc) just lateral to the nasal notch is in the wrong position to be part of the suture between lachrymal and lateral rostral. The adjacent ornamented area (∼25 mm wide), is narrower that the corresponding part on the much better preserved right cheek (Fig. 9A), so could be incomplete. The rock surface shows no clear edge, with slickensides on the surface where the bone impression grades into matrix, so depth of the lateral rostral as restored (Fig. 8C) may not be reliable.

Behind the level of the orbit (obscured by fractures) the bone surface of the left cheek is stepped down across a fracture onto piece a12, and evidently displaced forward. The distinct dorsal overlap belonging to the postorbital (od.Pa/IT, Fig. 7A) suggests that the region behind must include part of the squamosal, and below part of the jugal (Fig. 8A), but fractures are difficult to distinguish from sutures. A long opening crossing the bone surface behind the dorsal overlap area of the postorbital (Fig. 7A) suggests a suture or perhaps breakage along a sensory canal, but if natural is difficult to interpret in this position.

A small bone positioned immediately in front of the right supraorbitals of the rotated skull roof (Te, Fig. 7A) is interpreted as a displaced tectal, presumably also from the right side. Preserved length is 30 mm and maximum preserved width (behind the ventral notch) is 13 mm. A short section of the mesial margin is preserved about 7 mm above the nasal notch (nn, Figs. 7B, 8B). The latter is 15 mm long and 6 mm deep, corresponding in size to the notch of the lateral rostral, and implying a nasal opening (fe.exa, Fig. 23) considerably larger than in Eusthenodon (5 mm long by 3 mm high in the holotype ([19]: pl. 10). However that specimen was less than half the size of the holotype of Edenopteron. A ventral view (Fig. 7C) shows an expanded rounded thickening just behind the notch (th), inside of which a smooth concave surface continues upwards and anteromesially (pr.te), presumably the tectal process leading to the nasal cavity ([38]: fig. 53). Anterior and posterior margins of the tectal are missing, but it seems the nasal notch was relatively larger for the size of the bone than in Eusthenopteron, although positioned near the anterior end as in that form, and Eusthenodon. As restored it is more elongate than the rather equilateral tectal in the restorations of Mandageria ([15]: fig. 21).

Jarvik’s [2], [5] restorations of Eusthenopteron show the infraorbital sensory canal passing via the lateral rostral to the premaxilla, but in Bruehnopteron from Nevada [31], and re-studied specimens of Eusthenopteron from Miguasha, it passes directly from the lachrymal to the premaxilla, as is the case also in Jarvikina and Platycephalichthys (H.-P. Schultze, pers. comm. 17 Feb 2012). The restoration of Mandageria shows a lateral toothed margin on a broad lateral rostral ([15]: figs. 6c, 21b); this would require the infraorbital canal to pass through it, but according to P. Ahlberg (pers. comm., 4 July 2012), the teeth are carried on the premaxilla passing back inside the lateral rostral. The displaced maxilla in our specimen (see below) displays an ornamented surface right to the anterior tip, so the lateral rostral must have been excluded from the jaw margin, the normal arrangement in other tristichopterids apart from Mandageria.

As noted above, the relation between the premaxilla and the preserved palate indicates the anterior edge of the snout was about 20 cm of midline length in front of the posterior edge of the parasphenoid, or about 17.5 cm in front of the presumed position of the buccohypophyseal foramen (f.bhp, Fig. 5B). In Eusthenopteron, based on the restored neurocranium ([5]: fig. 11B–C), the center of the pineal cavity is 28% of total length of the anterior moiety (ethmosphenoid) anterior to the level of the buccohypophyseal foramen. In Notorhizodon the level of the buccohypophyseal foramen is about 47 mm in front of the posterior edge of the parietal bone, and approximately level with the anterior corner of the intertemporal, which on the skull pattern of Edenopteron would place it slightly in front of the pineal plate; i.e. reversing the situation of Eusthenopteron, which could be attributed to posterior migration of the pineal relative to the length of the snout. In the restored palate of Notorhizodon ([8]: fig. 37A) the buccohypophyseal foramen is well in front of the adductor fossa, whereas for Eusthenodon it was reconstructed only slightly in front, level with the posterior end of the ectopterygoid ([19]: fig. 29). Thus the previous restoration of Notorhizodon could be adjusted, but the buccohypophyseal foramen is approximately level with the anterior fang of the ectopterygoid, as seems to be the case also in the Edenopteron palate (Figs. 4B, 5B). The buccohypophyseal foramen of Edenopteron lies close to the level of the anterior end of the ectopterygoid (but there was perhaps some displacement), and about 9–10 cm in front of the adductor fossa (as preserved on the right side).

Cheek unit.

The right cheek of the holotype (Fig. 9) is preserved in part and counterpart on numerous pieces (internal surface on c1, c2, d4 behind saw-cut, b4, b5, b8 in front; external surface on a11, b1, c3, some bone of the squamosal embedded in resin on c10). As noted above, the left cheek (Figs. 7A, 8A) is less informative due to crushing; possibly part of its inner surface is preserved on b6, with numerous fragmented impressions showing external ornament on b8 and c12. The overall configuration of the cheek unit is best indicated by the right inner surface. The cheek is also preserved on the paratype (ANU V3468) and incompletely on ANU V3478 (Figs. 11, 12, 13).

The inner surface of the holotype right cheek (Fig. 9B) shows the lachrymal ossification center placed close to a gentle embayment in the ventral margin (La, Fig. 10B), inside which is an internal thickened ridge (ri), presumably underlying the infraorbital sensory canal, which thus ran just above the suture with the maxilla, its normal position. The ventral ridge continues across the slightly displaced suture with the jugal (Ju). The posterior suture of the jugal is obscured by a large crack ventrally, but is clearly inferred higher up from the radiating striations of the squamosal; its dorsal suture with the postorbital (Po) is cracked, and best located on the external surface. The smooth internal surface of the jugal shows radiating striations only on the ventral ridge; these indicate an ossification center very close to the ventral margin. The jugal clearly reached the margin of the orbit. The large postorbital (Po) has much of its outer margins obscured by fractures, but a prominent dorsal process is well preserved (pr.psp, Figs. 9, 10). In life this overlapped the posterior corner of the parietal shield (od.Po, Fig. 3).

The squamosal (Sq) has a rounded dorsal margin, showing striations radiating from the ventral ossification center, placed just above the ventral ridge which is thickest at this point. The ossification centers for the lachrymal, jugal, and squamosal all seem more ventral in position than restored for Eusthenodon ([19]: fig. 27). A triangle of bone still attached to the rock shows the external bone surface on the cast (Fig. 9B), the break representing the suture between the squamosal and preopercular, behind which the striations have a completely different orientation (Pop, Figs. 9B, 10B). There is an internal thickening inside the posterodorsal margin of the preoperculum (th), and the broken edge shows the squamosal overlapping the preoperculum in the dorsal part of their common suture, as in Eusthenopteron ([2]: fig. 9). The external surface of the squamosal (on piece c3) is retained as bone, ∼3 mm thick ventrally, and nearly 8 mm thick on the ventral ridge at the squamosal ossification center. The junction with the quadratojugal (Qj) is unclear, but like Eusthenopteron it suggests an extensive overlap area on its dorsal margin for both the squamosal and preoperculum. Radiating striations from the posteroventral preserved corner (Figs. 9B, 10B) represent the quadratojugal ossification center. The ventral edge of the quadratojugal is preserved adjacent to the maxilla on piece d1. Its posterodorsal margin runs down to the edge of the cast, inside which is a prominent internal ridge (qj.ri).

The external surface of the cheek unit is completely prepared out in front of the saw-cut (Fig. 9A), but behind much of the bone remains, although radiating striations and some bone sutures are clear. The well preserved ventral edge across the lachrymal and jugal shows a smooth zone 12–20 mm wide right along the margin (sm). The posterior edge of the lachrymal (La, Fig. 10A) is obscured by the fractured lower part of the jugal (Ju), which is displaced forward above it. The orbital margin of the lachrymal is completely preserved as an embayed and thickened edge between anterior and posterior angles (orb.m). The anterior margin of the lachrymal is missing its middle part, but clearly was much steeper than in either Eusthenopteron or Eusthenodon, in both of which the lachrymal had a different shape, with about 50% of the length of the bone in front of the orbital margin. The fractured jugal is a little displaced forward over the lachrymal, displaying its anterior margin as a rounded edge with a slight ventral notch and process, like Eusthenodon and Eusthenopteron (the process conveying the infraorbital sensory canal). In Eusthenodon the jugal reached the orbit internally, but the narrow orbital margin had an external overlap for the posterior supraorbital ([19]: fig. 27A); this is absent in Edenopteron. The upper suture with the postorbital (Po) is very clear, running back across the saw-cut (Figs. 9A, 10A), where radiating eroded bone shows a clear triple junction with the anterior edge of the squamosal (Sq). A distinctive broad dorsal overlap (od.Pa/IT) slid under the edge of the parietal shield when the cheek was in position, extending anteriorly into the orbit where it was overlapped by the posterior supraorbital (od.So2). This relationship is demonstrated on ANU V3478, a badly fractured specimen showing the left cheek slightly displaced from under the lateral edge of the skull (Fig. 11). The dorsal overlap of the postorbital (od.Pa/IT, Fig. 12) is slightly pulled apart from under the edge of the incomplete parietal in front (Pa), and intertemporal (IT) behind. The postorbital (Po) and jugal (Ju) both show clear patches of pores indicating the passage of the infraorbital sensory canal through these bones (p.ioc). The lachrymal (La) is crushed and incomplete in front of the jugal, which shows a ventral process on its anterior edge as in the holotype. The dorsal overlap of the postorbital in the holotype terminates anteriorly at the level of the jugal-postorbital suture, with a narrow selvage (od.Ju, Figs. 9A, 10A) where the jugal projected into the orbital margin, as is clearly demonstrated in the second specimen with the jugal still in position (Ju, Figs. 11, 12). The posterior extremity of the dorsal overlap is clearly seen across the saw-cut in the holotype. Eusthenodon had a similar overlap ([19]: fig. 27A), but somewhat different in shape (less embayed posteriorly, reducing to a point anteriorly) compared to Edenopteron.

Behind the postorbital the squamosal is defined mainly by radiating striations in the preserved bone; the preopercular and quadratojugal are very poorly preserved on this specimen, but more clearly seen on the paratype (Fig. 13A). This is a crushed associated left lower jaw and cheek unit (external surface on piece h8; internal cheek g5; lower jaw internal g4 and steinkern of adductor fossa g9). It is very close in size to a right Eusthenodon cheek (specimen P1480) figured by Jarvik ([19]: pl. 20). Unlike that specimen the maxilla has not remained with the cheek, being displaced inwards (see below). The lachrymal and jugal as far as preserved in ANU V3468 compare closely with the holotype. The postorbital is badly crushed dorsally, but the squamosal and quadratojugal clearly show their common suture, the latter reducing to a point anteriorly as in Eusthenodon, rather than forming a truncated edge contacting the maxilla as in Eusthenopteron. The preopercular is slightly displaced to reveal the posterodorsal overlap of the quadratojugal (od.Pop), again as in Eusthenopteron ([2]: fig. 9). Like the sqamosal in front, its dorsal margin is unclear due to fracturing.

In each of the cheek units just described the maxilla is displaced or missing, perhaps due to less complex overlap relationship compared to other tristichopterids, for example the unique overlap area for the squamosal in Eusthenodon ([36]: fig. 37C). However the almost complete right maxilla of the holotype lies adjacent to the lower jaw (Mx, Figs. 15, 16). Its anterior end is well preserved on piece b2 (internal surface; Fig. 4C), the external surface running back across the larger piece carrying the jaw symphysis (b8), then across b9 and the saw-cut onto d1 (external impression) and c1 (internal impression). Including 2–3 mm at both ends, and the saw-cut thickness (∼8 mm), the complete maxilla was at least 195 mm long (Fig. 14C). Its anterior end shows a clear dorsal process (dp.Mx, Fig. 4C), a structure well documented in Eusthenopteron, but said to be absent in Eusthenodon [36]. However, two isolated maxillae from the Famennian of Grenfell, NSW, which show this process, have been assigned to ‘Eusthenodon cf. wangsioi’, the Greenland species [18]. Posteriorly, the maxilla of Edenopteron does not expand like in Eusthenopteron, nor does it have the more complex dorsal margin of Eusthenodon (Fig. 14C). Most of the ventral margin shows fine pointed teeth, about 2 mm of length visible externally, but longer internally (5–6 mm; Fig. 14D), with a base ∼2 mm wide. By comparison the marginal teeth of Notorhizodon are more robust and widely spaced ([8]: fig. 23A–B). The dorsal surface of the Edenopteron maxilla curves gently upward before a notch for the overlap for the jugal (od.Ju, Fig. 14C), set in about 7mm from the ornamented surface (dorsal edge not complete). The maximum depth (∼18 mm) is ∼70 mm from the front, and the bone decreases in depth posteriorly (∼12 mm deep 40 mm from the posterior end). It tapers posteriorly as in Eusthenodon [36], and also anteriorly, where it curves slightly downwards apparently to a point (extremity missing; Fig. 14C), rather than curving upward or with an anterior truncation (as in Eusthenopteron and the Grenfell examples). One specimen of the Greenland Eusthenodon indicates from overlaps for the maxilla that it reached back to just project beneath the anterior edge of the quadratojugal ([19]: fig. 28). There is a more extensive contact between these bones in Eusthenopteron [2]. About 25 mm outside the posterior preserved tip of the maxilla in the Edenopteron holotype is the ventral edge of the right quadratojugal, with an embayed thin margin and clear suture just anterior to the end of the maxilla. This suggests the same arrangement as in Eusthenodon. We did not locate the left maxilla (possibly crushed or obscured beneath the preserved palate).

The paratype (ANU V3468, Fig. 13) also preserves the left maxilla, showing the same bar-like shape. Both premaxilla and maxilla are dislodged down inside the dentary, the former preserved in two portions displaced across a joint in the rock. The maxilla is preserved in three sections, the anterior with a broken anterior edge, and the posterior showing a dorsal overlap (od.Ju, Fig 13B), and again reducing in height to a posterior point at about the same level as the posterior end of the dentary on the lower jaw, which would thus have excluded the squamosal from the jaw margin.

A ‘bar-like maxilla’ was stated as a unique distinguishing feature of Marsdenichthys [7], but the above description and comparisons indicate that this is not a reliable character.

Palate.

Piece b7 (Fig. 4B), the largest preserved portion, displays the entire denticulate part of the parasphenoid, part of both vomers, dermopalatines, entopterygoids, and the left ectopterygoid. The almost complete right vomer, including the fang, has its anterior edge preserved on a8, and lateral part on b3. The left vomer on a9 includes a complete fang (at least 41 mm long measured from its root), its posterior process extending onto b7. The left ectopterygoid, entopterygoid and dermopalatine are most complete. The right side continues across the saw-cut onto c1, where ‘steinkerns’ of both adductor fossae are preserved in articulation (Fig. 17A).

The posterolateral edge of the left vomer is very clear (Fig. 4B). It is more convex than in Eusthenopteron or Eusthenodon, but similar to this margin in Platycephalichthys ([40]: fig. 23). Its overlaps with the anterior edge of the dermopalatine and entopterygoid are pulled apart (od.Vo, Figs. 4B, 5B). The denticulate part of the entopterygoid (Ent.tp) stands up with a laterally directed ridge, its surface covered with scattered fine dentition (best seen where bone adheres to the impression on the rock surface, showing that ornament was finer than in Notorhizodon). Notorhizodon also differs in its stronger ‘labial ridge’, which projects prominently towards the labial margin (r.lab, [8]: fig. 30), rather than out (downwards) from the denticulate surface as shown by Edenopteron.

The vomer of Hyneria was said to lack posterior processes [33], but more recently it has been stated that they are ‘at least 45%’ the length of the parasphenoid [41]. In Edenopteron the posterior processes are ∼54% parasphenoid length (ppr, Figs. 4B. 5B).

The parasphenoid (Psp, Figs. 4B, 5B) has a sharp anterior point, is widest (∼30 mm) just behind the level of the posterior processes of the vomers, and narrows posteriorly, being slightly waisted about 25 mm in front of the posterior margin, at the assumed level of the buccohypophysial foramen (f.bhp), with posterior bone radiations behind this level. In Notorhizodon this part of the parasphenoid is expanded posteriorly. The ventral denticulate surface in Edenopteron (not well preserved) is flat to slightly convex (with a depression at the ossification center); clearly it was different to the concave and broad shape of Notorhizodon. The left side at the back end indicates an upward projection (asc.pr, Fig. 5B). The left entopterygoid is pushed down beneath the edge of the parasphenoid, from which it is separated by matrix. On the right side a faint lineation represents the margin of a separate element corresponding to the ‘accessory vomer’ in Mandageria and Cabonnichthys (ac.Vo), which is more clearly seen on the counterpart (see below).

Both dermopalatines carry remains of one large fang (f.Dpl), the right with a base ∼20 mm in diameter (and the tip of a second fang preserved behind it). The left dermopalatine shows clearly the mesial margin pulled away from the entopterygoid, which according to Jarvik ([2]: 37) fitted into a groove. A slight but distinct mesial angle opposite the main fang notches the left entopterygoid (n, Fig. 5B). The remnant of the second dermopalatine fang was probably just erupting when the animal died. The left dermopalatine shows the contact face where the vomer has pulled away slightly at the front, a clear posterior margin where it pulled away from the entopterygoid, and the fang positioned more to the posterior, without any sign of a second fang or pit. The overall shape of the dermopalatine is similar to Eusthenopteron, and it measures about 70 mm long. The posterior margin shows no evidence of the complex interfingering on this suture seen in Notorhizodon. Although no tooth row is evident on the cast, the specimen reveals teeth within the non-removed bone, and on the sawn edge through the right dermopalatine a vertical bony lamina projects ∼12 mm downwards; i.e. as illustrated by Jarvik ([38]: fig. 55). The opposite side of the saw-cut shows a tooth projecting from the lamina (total depth ∼ 20 mm). A thickened anterolateral ridge preserved on the left dermopalatine (ri, Figs. 4B, 5B) must represent the inner edge of the choana, as in Eusthenopteron, but shows no indication of a tooth row on this part of the ridge.

Sutures between the entopterygoid, ectopterygoid, and dermopalatine are clear on the left side. The left ectopterygoid fang is near its anterior margin, with a second smaller fang at the preserved edge of b7 (f.Ect, Figs. 4B, 5B). The right ectopterygoid is behind the saw-cut on piece d7 (dermopalatine suture lost in the saw-cut), and continues back as steinkerns of the adductor fossa on piece c1 (Fig. 17A). This shows the root of a large fang on the saw-cut, and a posterior fang with a strong vertical lamina partly exposed through the rock matrix about 55 mm behind the first (the same distance between fangs on the left ectopterygoid). The upper and lower adductor fossae come together (anterior edge) about 48 mm behind this, suggesting an ectopterygoid about 116 mm long, if it reached the anterior edge of the adductor fossa as it does in Eusthenopteron. In Notorhizodon a smooth lamina (lv.Ent, 8: fig. 30] separated the adductor fossa from the denticulate part of the ectopterygoid, but this detail is not shown in Edenopteron. Mandageria is reconstructed with the entopterygoid excluding the ectopterygoid from the adductor fossa [23], and the dermopalatine is only slightly shorter (98%) than the length of the ectopterygoid, whereas in Eusthenodon it is about 83% the length, and in Eusthenopteron it is considerably shorter (64%). In Edenopteron this proportion was about 68%.

A composite latex of the counterpart (pieces b5, b6, a11) shows the upper surface of the dermal bones of the palate (Fig. 6). The parasphenoid anterior tip (visible in Fig. 3) is more oblique, the whole element having a different shape to its ventral outline. It is more elongate (preserved length 157 mm), and gently concave over most of its surface, deepening to a spatulate shape at the posterior end where a central depression evidently housed the hypophyseal fossa (fo.hyp, Fig. 6B). Parts of both vomers are in position, the more complete right vomer extending from the incomplete anterior margin (deeply embayed with a marginal ridge), back to the posterior process (ppr) some 82 mm behind. A roughened surface marks the fang position and ossification center, with clear radiating striations to the preserved extremities of the bone. Both vomerine fangs are in position on pieces b7 and a9, the latter XCT-scanned to show the presence of a tooth-bearing lamina (see Fig. 21). In anterior view the fangs (∼45 mm long) curve inwards, their tips ∼33 mm apart, with the center of bases ∼40 mm apart. The 2 mm saw-cut (Fig. 2A) exposes 30–40 mm thick bone, and a toothed lamina of the left vomer. The vomer anterior margin is not completely exposed, but rather than the transverse edge of Eusthenopteron or Jarvikina, it was evidently more deeply embayed (riVo, Fig. 6B), as in Eusthenodon ([19]: pl. 16).

Both entopterygoids were displaced upwards by compaction, such that their mesial edges (Ent, Fig. 6A) curve up above the level of the parasphenoid. On the right side of the parasphenoid, mesial to the entopterygoid, a separate bone is preserved which must be the accessory vomer (ac.Vo, Fig. 6B). It has a clear lateral margin as a rounded edge, and a sinuous central ridge posteriorly, which curves over to the mesial edge about midway along the bone (ri). The anterior end is pointed, with radiating striations, reaching just past the posterior process of the vomer (as in Cabonnichthys; in Mandageria the accessory vomer just reaches the posterior process). A groove inside the anterior extremity suggests a blood vessel (gr). An oblique edge with a roughened surface towards the posterior of the bone may have abutted against the ascending process of the parasphenoid. The posterior extremity of the bone is obscured by a crack. The inner surface of this bone has not previously been described. As noted above, an analogous or homologous bone is widespread in palaeoniscoids ([42]: 273).

Lower Jaw.

In the holotype the external surface of both lower jaws are slightly splayed out, with anterior ends still together in the jaw symphysis (Figs. 15, 16). The right jaw is rolled over with the maxilla sitting on its lateral side, and its posterior end obscured by the gular. The left jaw is displayed almost to the posterior margin, where the smooth external overlap for the quadratojugal (od.Qj, Fig. 16) is just visible in front of the overlying subopercular (Sop). This represents the fourth infradentary (Id₄, Fig. 16), with its ventral and anterior margins clearly seen. One apparently natural lineation amongst surface fractures may be an infradentary pitline (pl.Id₄). The dorsal margin in contact with the dentary is broken. The third infradentary (Id₃) seems of more oblong shape than restored for Eusthenodon, with its anteroventral suture showing a distinct inflection ventrally. An elongate overlap area along the ventral margin broadens anteriorly (odSbm), showing a slight notch (n) on the ventral margin of the second infradentary (Id₂). Its dorsal margin crosses the saw-cut to a slight angle in two other sutures, assumed to be the ventral edge of the dentary (De). Just beneath is another possible pitline (pl.Id₂), but the area is somewhat fractured. The ventral suture fades out to the anterior, and margins are unclear, so a suture between first and second infradentaries is not shown (Fig. 16). The right jaw being more rotated exposes its ventral edge, with a flat smooth margin broadening slightly anteriorly and posteriorly; a narrower middle part might represent the junction between the first two infradentaries (Id_1, Id₂), but again no suture is evident. A longitudinal lineation may be the ventral edge of the dentary, or possibly another pitline (?pl). Radiating striations in the abraded bone suggest a similar anterior position for the dentary ossification center as in Eusthenodon ([19]: fig. 26A).

The inner lower jaw surface is poorly known, the coronoids and prearticular being largely enclosed by rock matrix in the holotype. There is no evidence on whether a parasymphysial plate was present, this previously suggested to be a tristichopterid characteristic [16]. The lower surface of piece b7 has the jaw symphysis and anterior 22 cm of the left lower jaw preserved as bone (Fig. 17B). Rows of small teeth along the margin (?m.dent) could be tooth rows of the coronoids. They seem too deep to be on the dentary, but preservation is insufficient to decide this. The counterpart (external impression; piece b8) shows no sign of a marginal tooth row (Fig. 15), so either they were lost in preparation, or these small teeth do represent coronoid tooth rows. Coronoid tooth rows were interpreted as primitive ([16]: 668), as they occur in Eusthenopteron, Jarvikina and Notorhizodon, whereas marginal coronoid teeth are absent from the anterior coronoids in Eusthenodon, Mandageria and Cabonnichthys, presumably representing the more derived state.

The base of the dentary fang is visible through broken bone of the left lower jaw close to the jaw symphysis on piece b7 (f.De, Fig. 17B, C). Again there is no sign of marginal teeth between the dentary fang and the jaw symphysis, this being one of two characters by which Langlieria from Belgium was distinguished from Mandageria [30]. This is a character conflict for the sister group relationship of Mandageria and Eusthenodon to the exclusion of Cabonnichthys [18], the marginal dentition going past the dentary fang in Mandageria (and also Notorhizodon), but not in Eusthenodon or Cabonnichthys.

About 65 mm behind the dentary fang, the base of a very large first coronoid fang (diameter ∼25 mm) is exposed in a drill-hole through the bone (f.Co₁, Fig. 17B, C). Most of this fang has been prepared out from the rock matrix (length ∼42 mm), and shows a flattened cusp with cutting edges, in contrast to the rounded vomerine fangs of Edenopteron. The prearticular of the inner side of the jaw is visible only in section on the main saw-cut. Across the saw-cut both adductor fossae of the mandibular joint are preserved as bone-covered steinkerns in part and counterpart (pieces c4, d8), but these also show no sign of fangs (Fig. 17A). Noteworthy is the absence of any internal bone at the mandibular joint, the rock matrix filling the gap between thin lateral and mesial bone layers. This suggests that, as with the neurocranium, the jaw endoskeleton (quadrate-articular) was poorly ossified.

ANU V3468 also has the lower jaw preserved in articulation against the left cheek. The dentary fang impression is preserved on the main block (h8), where it is displaced ventrally across a joint (f.De, Fig. 13B). Behind this the visible fangs are as follows (measurements center to center): 60 mm behind the dentary fang is the first weathered coronoid fang, 55 mm behind this is a second fang, and 60 mm farther back is a third fang, the last being 40 mm in front of the anterior edge of the adductor fossa (all preserved on piece g4). The thickened eroded bone of the coronoids is too badly preserved to show intervening sutures. Another fang could be obscured immediately in front of the adductor fossa, but the external impression (Fig. 13) suggests that only enlarged teeth occur in this region. Thus a second fang of the posterior coronoid may not be developed in Edenopteron, which would be a shared resemblance with Notorhizodon, and some figured lower jaws of Platycephalichthys ([40]: pl. 17), although, as previously discussed ([16]: 667), this condition is variable within Platycephalichthys. The advanced condition of two fangs on the posterior coronoid occurs in Eusthenopteron, Eusthenodon, Cabonnichthys, and presumably Mandageria.

Operculum.

The left opercular and subopercular are preserved behind the lower jaw in the holotype (Op, Sop, Fig. 15). The opercular is somewhat crushed, but partly retains its dorsoventral curvature (∼80° as preserved). Its ventral edge does not correspond to the shape of the dorsal overlap area on the subopercular, so is incomplete. However it is clear that the opercular was longer than the subopercular (Fig. 18B). The dorsal overlap is completely preserved as is the sloping posterior margin. The anterior margin (Fig. 18B) is thick and slightly convex (but less so than in Eusthenopteron). In overall shape the bone seems more narrow dorsally and expanded ventrally, with the posterodorsal margin as far as preserved being clearly less convex than in Eusthenodon [19]. This is confirmed by the more complete and uncrushed right opercular (Fig. 18A), preserved as an internal impression just above the clavicle (pieces d2, d5, d9), its external impression preserved on c6, c8, c9. With compaction of the holotype this bone was evidently compressed down onto the inside of the (displaced) left clavicle, but otherwise was not much displaced. The anterodorsal corner is missing, its preserved edge showing a depressed area presumably representing the overlap area for the postspiracular, as in Eusthenopteron ([2]: fig. 9B). No other evidence of the postspiracular has been found so far in Edenopteron. The external surface of the opercular is gently concave ventrally and gently convex dorsally, and its anterior margin changes from gently concave dorsally to gently convex ventrally. The dorsal half of the anterior margin has a smooth bevelled edge ∼7 mm wide, similar to that of the subopercular. The ventral half is also bevelled, but is thinner with ornament right to the edge except for the rounded anteroventral corner, which is smooth. The adjacent part of the external surface of the subopercular is also smooth. Measured dimensions for the right opercular are ∼150 mm dorso-ventrally and 90–95 mm antero-posteriorly. Thus it was clearly higher in proportion to length than the opercular of Eusthenodon ([19]: pl. 12, fig. 2) which has length about 75% its height (∼66% in Edenopteron). The opercular is longer than the subopercular in Eusthenodon, as in Edenopteron, but in Eusthenopteron it is the opposite, with the subopercular longer than the ventral edge of the opercular.

The left subopercular of the holotype is excellently preserved (Sop, Figs. 15, 16, 18B), uncrushed and little displaced above the posterior end of the lower jaw. Its dorsal edge is obscured under a displaced scale (sc) and the overlying opercular, but the shape of the overlap area is clear (Fig. 15). The external surface is slightly convex, and ornamented except for a smooth zone extending up the anterior edge to the dorsal overlap. The concave anterior margin forms a thick (13 mm) bevelled edge that abutted against the lower jaw (a.LJ, Fig. 18B). The posterodorsal extremity is missing, but maximum preserved length (∼65 mm) indicates that the subopercular was only ∼70% the length of the opercular (it is restored as slightly shorter in Eusthenodon [19]). In contrast, Eusthenopteron has the subopercular longer than the opercular ([2]: fig. 9B). The subopercular of Edenopteron is folded around an angle of ∼110° at its anterior edge, suggesting a relatively low position in the fish, straddling the lateral and ventral laminae of the cleithrum. The adjacent posterior submandibular (Sbm₁) seems rather flat, and would have been mainly on the ventral surface and hence not visible in lateral view, this being a difference from the restoration of Eusthenodon [19]. Jarvik stated ([19]: 65) that the subopercular was ‘fairly high’ based on specimen P1473, which suggests that the subopercular was entirely on the lateral side, with the posterior submandibular inflected around the ventrolateral angle of the fish ([19]: pl. 16). A similar configuration occurs in Eusthenopteron [5], is demonstrated in a 3D prototype of the ANU Gogonasus specimen (ANU 49259), and also occurs in Mandageria (Fig. 19C). Thus Edenopteron apparently differs in this respect from all these taxa.

The subopercular is a similar size to one from Belgium that has been referred to the type species Eusthenodon waengsjoei ([37]: fig. 2B). That specimen differs from ours in having coarser ornament, and a distinct anterodorsal notch where the opercular overlapped, plus a ventrally thickened area abutting the lower jaw (dorsal in Edenopteron; Fig. 18B). The subopercular of Langlieria differs even more, being longer and lower, with the same anterodorsal notch, but with a prominent anteriorly directed process ([30]: figs. 2, 6).

Gulars, Submandibulars.

The two gulars of the holotype (Gu, Figs. 15, 16) have a maximum preserved length of 170 mm (right side) and 195 mm (left side). The shape is similar to other tristichopterids (Eusthenopteron, Eusthenodon), with a curved posteromesial margin that met the clavicles, and a straight midline margin, but the latter is proportionately much shorter in Edenopteron (80–90 mm for left bone; cf. 120 mm for clavicular margin). Both gulars are similarly preserved, neither showing the anterior end. Their lateral margins are flat and straight to convex with a smooth edge posteriorly, and slightly concave anteriorly with a more distinct overlap area. In front of the gulars as preserved is a triangle between the jaws devoid of bone, perhaps for a median gular as restored for Eusthenodon [19]. The gulars are splayed apart, and could have been displaced backward. However the left gular at its posterior end is only slightly displaced from the posterior submandibular (Sbm₁, Fig. 16), which in turn aligns closely with the subopercular (Sop), indicating little displacement from life position. By comparison with Jarvik’s Eusthenodon reconstruction, we could only fill the anterior triangle in the Edenopteron restoration by assuming the anterior ends of both gulars are incomplete.

Submandibular elements are associated with both lower jaws of the holotype (Figs. 15, 16). Some were displaced to reveal a broad overlap running along the ventral edge of each jaw (od.Sbm). Two displaced and incomplete elongate elements are adjacent to the right jaw (Sbm_2, Sbm₃), the posterior element sitting over the ventral edge of the lower jaw, with the adjacent gular (Gu) slightly displaced mesially to show its smooth lateral edge. The posterior submandibular is close to its original position on the left side (Sbm₁). A similar preservation is seen on the impression of both gulars in specimen P1473 of Eusthenodon ([19]: pl. 16, fig. 1). The left posterior submandibular extends back under the edge of the subopercular (Sop). Its anterior end projects to a point on the mesial side, not the lateral side as in Eusthenopteron [2], thus being like Eusthenodon [19], but more elongate and pointed. There are eight ‘branchiostegals’ in Eusthenopteron, but seven restored for Eusthenodon, based presumably on the holotype and P1483, where the left row is preserved, evidently with clearer sutures than preserved in Edenopteron ([19]: pl. 17, fig. 3). In Edenopteron these elements were more elongate and possibly less subdivided, as appears also to be the case in Mandageria ([16]: fig. 17). Apart from the larger posterior element (‘submandibulo-branchiostegal’ of Jarvik) there is clear evidence of only two additional elongate submandibulars (Sbm₂, Sbm₃), the posterior element missing on the left side to reveal the underlying overlap area (odSbm, Fig. 16).

The overlap of submandibulars in Eusthenopteron was fully illustrated and described [2]; they overlap the gular mesially and are overlapped by the infradentaries laterally. Thus each gular shows externally an overlap area along the lateral margin, each branchiostegal has a narrow lateral overlap, and the lower jaw in external aspect is ornamented to the ventral margin of the infradentaries. The same arrangement is seen in Osteolepis ([43]: fig. 24) suggesting it is primitive. In Eusthenodon ([19]: pls. 10, 17, fig. 3) the branchiostegals are well displayed on the holotype and a second specimen (P1483), the latter giving the best example of the gular plate. Overlap areas are not exposed, but a third specimen (P1473) shows internal impressions of the thickening at the lateral margin of the gular underlying the overlap area, with the most posterior submandibular still in position ([19]: pl. 16, fig. 1). The ventral margin of the lower jaw is not clearly illustrated in an isolated example, but Jarvik’s ([19]: fig. 27B) reconstruction reasonably assumes the same condition as previously established for Eusthenopteron [2].

The holotype of Edenopteron shows the reverse of this condition, with a clear external overlap area reaching 15+ mm wide along the ventral margin of the left lower jaw (odSbm). The same overlap relationship is also seen in ANU V3468 (Fig. 13), and is evident in some Mandageria specimens [15], where the submandibulars are displaced away to reveal the overlap along the ventral jaw margin (odSbm, Fig. 19C). Another example showing a broad overlap on the ventral jaw margin (“groove for submandibular bones”) is a specimen from Belgium referred to Eusthenodon waengsjoei [37].

Cleithrum.

The left shoulder girdle of the holotype (cleithrum, clavicle) is exposed in external view (pieces d2, d9) between the splayed-out jaws (Figs. 15, 16). The cleithrum is fractured but still preserves the inflection around a slightly obtuse angle onto its ventral lamina (Fig. 19B). About 190 mm of the lateral lamina is exposed externally. Its dorsal edge is partly obscured by the adjacent submandibular and gular, but is at least 70 mm wide at its dorsal preserved margin. The anterior preserved edge (just behind the posterior corner of the gular plate; Figs. 15, 16) clearly shows that breadth was expanding dorsally, in contrast to the subparallel shape of the lateral lamina of the cleithrum in Eusthenopteron [3], [44]. The internal impression (on piece c6) continues across a crack to piece c7, with the dorsal margin indistinct, but indicating lateral lamina length of at least 210 mm, and width of about 50 mm halfway up the lateral lamina. An incomplete cleithrum of Platycephalichthys bischoffi also suggests an increase in width dorsally, but the edges of that specimen are broken, and as illustrated the ornament is less linear than in our material ([40]: fig. 17). On the second cleithrum of Edenopteron (ANU V3479) the dorsal margin is also missing, but the ridged ornament diverges dorsally, suggesting a similar expanded shape. It is noted that a much smaller tristichopterid cleithrum from a lower level in the NSW South Coast Devonian sequence also shows a distinctly broader dorsal margin on its lateral lamina ([22]: fig. 3a).

External ornament on both cleithra of Edenopteron is distinctive, comprising elongate ridges and grooves generally aligned along the cleithrum, becoming finer and more meandering towards the anterior representing the postbranchial lamina. This is much better developed than in other tristichopterids (pbl, Fig. 19B). The ventral lamina is more elongate (only 40 mm wide to the anterior end of the ventrolateral angle), with a different shape in ventral view to that of Eusthenopteron ([5]: fig. 6C). The well-preserved posteromesial margin is straight to slightly concave, lacking the posteromesial corner of Eusthenopteron, and narrows rapidly to the midline, where the ornamented area is reduced to a point behind the distinct overlap area for the clavicle (od.Clav, Fig. 19A). There could not have been any midline contact between cleithra, in contrast to Eusthenopteron (m.ma, [3]: fig. 3E). There is no published illustration of the Eusthenodon cleithrum, but one specimen (P1481) was said to include “an almost complete cleithrum very suggestive of that in Eusthenopteron” ([19]: 68). By contrast, a previously unfigured cleithrum of Mandageria shows a similar pronounced anteromesial projection to Edenopteron (Fig. 19C), but it has a different, less triangular shape, with a rounded posteromesial angle (pa) reminiscent of Eusthenopteron. Incomplete elements representing the right shoulder girdle project off the collected slab of the holotype (Sh.g, Figs. 15, 16), and potentially more data could be obtained with further excavation.

Clavicle.

The overlap for the clavicle is clearly seen (od.Clav, Fig. 19) because the right clavicle is preserved slightly twisted and displaced forward. There is no evidence of an ascending process on the clavicle, the relevant corner on the external impression being rounded and apparently complete. If present it must have been broken off, but nor is there any indication on the internal impression (preserved on piece d5). A new clavicle of Eusthenodon waengsjoei from Greenland ([36]: fig. 36) shows a prominent but completely unornamented ascending process. The external ornament is much coarser than on the holotype clavicle of Edenopteron, even though it is a slightly smaller example. Edenopteron possibly had an internal flange of the clavicle that fitted inside the bottom edge of the cleithrum (fl, Fig. 19A); the amount of removed bone with part and counterpart in position indicates a thickness of at least 12 mm for the ventral wall of the shoulder girdle. The clavicle is about 60 mm wide, with a gently convex posterior margin. Its anteromesial extremity is obscured under the right gular plate. The anterolateral margin is rounded, and slightly convex, where it would have fitted beneath the gular. The Edenopteron clavicle was evidently less elongate than in Mandageria, based on the specimen illustrated here (Fig. 19C).

Anocleithrum, Supracleithrum, Posttemporal, Extrascapulars.

Only one of these bones can be described. A subrectangular element with indistinct overlap areas preserved just behind the left subopercular is interpreted as a supracleithrum, by comparison with Mandageria. A fracture across the middle with a sharp anterior inflection may represent the position of the sensory canal passage (llc, Fig. 18C), as preserved in Mandageria ([16]: fig. 10b). In that taxon these bones are differently developed to Eusthenopteron ([16]: 59), the supracleithrum being longer and four-sided, and the anocleithrum shorter and triangular (the reverse in Eusthenopteron). The anocleithrum and posttemporal of Edenopteron have not been identified, but can be restored after other tristichopterids (Fig. 23). The same applies to the extrascapulars, which would be expected in the holotype, given their life position immediately behind the post-parietal shield; as yet these have not been located in the prepared material. Based on the configuration of overlaps for the extrascapulars on the posterior skull margin (e.g. Figs. 11, 12) we assume the median extrascapular of Edenopteron had a broader anterior margin than the pointed shape of Mandageria.

Fin skeleton.

An assemblage of elements of various lengths (3–26 mm; 2–3 mm wide) on piece f3 (Fig. 20A) are interpreted as lepidotrichia probably from the pectoral fin, being located just caudal to the left cheek of ANU V3468. Most show a distinct central elongate ridge. The lepidotrichia of Eusthenopteron are generally simple, lacking ridges ([5]: fig. 18), but a few larger elements of the pectoral fin have been illustrated suggesting irregular ridges ([44]: fig. 7). Some ‘distal pectoral fin elements’ in Sauripterus likewise have suggestions of a longitudinal ridge ([45]: fig. 15), but the ridges illustrated here are more regular and distinct. A central ridge is also suggested on some lepidotrichia of the caudal and anal fins of Mandageria ([15]: fig. 18). The closest resemblance to the examples figured here are in pectoral fins from Canowindra for the tristichopterid Cabonnichthys ([16]: fig. 12C), and for the rhizodontid Gooloogongia ([10]: fig. 14b). Of interest is the fact that basal scutes on the fins, documented in Northern Hemisphere osteolepiforms and in Eusthenopteron, have not been found in the Canowindra tristichopterids [16].

Several other elongate elements, expanded at both ends like that illustrated from Grenfell, NSW ([17]: fig. 7), may be fin radials, for example partly exposed inside the right orbital notch of the holotype (?f.r, Figs. 9A, 10A), and on piece h4, an element 54 mm long and 9–10 mm diameter at both ends, narrowing to 4 mm diameter in the middle region.

Scales.

Isolated scales are scattered through the matrix surrounding the dermal bones of Edenopteron, mostly exposing the internal surface that shows clearly the drop-shaped central boss characteristic of tristichopterids and rhizodontids (Fig. 20C, D). The boss is situated more towards the anterior edge as in Eusthenopteron and Eusthenodon ([19]: fig. 30). One small scale (8 mm across; part and counterpart on piece h3), possibly from ANU V3468, has a notched margin where crossed by an oblique internal ridge, probably a sensory canal (llc, Fig. 20D). The external surface preserved as an impression is indicated by a few examples where portion of the scale has broken away (Fig. 20B, C). These suggest widely spaced grooves (gr), as in the distinctive ‘mandageriid’ scale morphology of the Canowindra tristichopterids previously described [20]. Another incomplete scale (on piece h2) seems to lack the grooves. We cannot attribute these isolated scales to particular specimens, so it is unclear whether this was a variable feature of Edenopteron. The grooves are much more closely spaced in Langlieria scales from Belgium ([37]: fig. 13) but these do have a relatively large boss, as in Figure 12D. In scales of Hyneria the boss is relatively smaller [33], and an indented posterior edge is characteristic of this taxon [41]. This grooved scale surface is distinctly different from the scales of Northern Hemisphere Eusthenodon [20], as exemplified by the Greenland type species ([19]: fig. 30), and also demonstrated in Jarvikina wenjukovi from Russia ([40]: pl. 8, fig. 1a). Based on that taxonomic treatment, such scale ornament is typical of several Northern Hemisphere genera, in the same way that the grooved scales typify at least four Southern Hemisphere genera: Mandageria and Cabonnichthys from Canowindra, Edenopteron as documented here, and probably also Notorhizodon from Antarctica ([20]: fig. 4C). This scale type has also been documented from a third Australian locality, Jemalong NSW ([9]: fig. 9B).