1. Introduction
The neurocrania and associated endocasts of Palaeozoic vertebrates have long been a focus of study, reflecting the abundance of anatomical features in these complex structures. For sarcopterygians, work that began in the first half of the 20th century yielded detailed accounts of both the exterior of the neurocranium and the enclosed endocranial cavity for key exemplars of major lineages (Stensiö Reference Stensiö1963): actinistians (Diplocercides kayseri; Stensiö Reference Stensiö1963); dipnomorphs (Chirodipterus wildungensis, Glyptolepis groenlandica, Youngolepis praecursor; Säve-Söderbergh Reference Säve-Söderbergh1952; Stensiö Reference Stensiö1963; Chang Reference Chang1982); and tetrapodomorphs (Ectosteorhachis nitidus, Eusthenopteron foordi; Romer Reference Romer1937; Stensiö Reference Stensiö1963). Despite a number of well-preserved neurocrania reported since, the reliance on destructive techniques (serial sectioning: Romer Reference Romer1937; ‘smash’ method: Säve-Söderbergh Reference Säve-Söderbergh1952) restricted the number of new, detailed endocast reconstructions (e.g., Chang Reference Chang1982). The widespread palaeontological application of micro-computed tomography (μCT) over the past two decades has produced a flood of complete (Rhinodipterus kimberleyensis: Clement & Ahlberg Reference Clement and Ahlberg2014; Gogonasus andrewsae: Holland Reference Holland2014; Dipterus valenciennesi: Challands Reference Challands2015; Dipnorhynchus sussmilchi: Clement et al. Reference Clement, Challands, Long and Ahlberg2016; ‘Chirodipterus’ australis: Henderson & Challands Reference Henderson and Challands2018; Cladarosymblema narrienense: Clement et al. Reference Clement, Cloutier, Lu, Perilli, Maksimenko and Long2021; Griphognathus whitei, Iowadipterus australis, Orlovichthys limnatis, Pillararhynchus longi: Clement et al. Reference Clement, Challands, Cloutier, Houle, Ahlberg, Collin and Long2022) and partial (Spodichthys buetleri: Snitting Reference Snitting2008; Powichthys spitsbergensis: Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010; Tungsenia paradoxa: Lu et al. Reference Lu, Zhu, Long, Zhao, Senden, Jia and Qiao2012; Gogodipterus paddyensis: Clement et al. Reference Clement, Challands, Cloutier, Houle, Ahlberg, Collin and Long2022; Scaumenacia curta, Pentlandia macroptera: Boirot et al. Reference Boirot, Challands and Cloutier2022) sarcopterygian endocasts that add substantial new diversity to the limited set of early examples derived from physical approaches. Ready availability of internal structure via μCT permitted the first endocast models for previously unexamined groups, most notably stem sarcopterygians (Ptyctolepis brachynotus: Lu et al. Reference Lu, Giles, Friedman and Zhu2017) and onychodonts (Qingmenodus yui: Lu et al. Reference Lu, Ahlberg, Qiao, Zhu, Zhao and Jia2016).
Despite the abundance of data for some lineages, most notably dipnoans and tetrapodomorphs, information on endocast anatomy remains sparse for one of the principal early sarcopterygian groups: porolepiforms. This modestly diverse clade of stem lungfishes contains fewer than a dozen genera of Devonian age (see summary in Mondéjar-Fernández et al. Reference Mondéjar-Fernández, Friedman and Giles2021), and reasonably complete neurocrania are known for several of these. Fossils of Porolepis sp. (Jarvik Reference Jarvik1972), Durialepis edentatus (Mondéjar-Fernández et al. Reference Mondéjar-Fernández, Friedman and Giles2021), Glyptolepis groenlandica (Stensiö Reference Stensiö1963), Laccognathus embryi (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011) and Holoptychius bergmanni (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2013) preserve the ethmosphenoid and otoccipital components of the braincase, while only the ethmosphenoid is reported for Heimenia ensis (Clément Reference Clément2001). These unambiguous porolepiforms are potentially joined by Powichthys (P. thorsteinssonni, P. spitsbergensis), an Early Devonian genus of unsettled phylogenetic placement. Powichthys preserves both neurocranial divisions (P. thorsteinssonni; Jessen Reference Jessen1980; P. spitsbergensis, Clément & Janvier, Reference Clément and Janvier2004) and is most commonly resolved as either the sister lineage of porolepiforms (Jessen Reference Jessen1980; Panchen & Smithson Reference Panchen and Smithson1987; Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010; Cui et al. Reference Cui, Friedman, Qiao, Yu and Zhu2022) or a branch crownward of porolepiforms on the lungfish stem (Ahlberg Reference Ahlberg1991; Cloutier & Ahlberg Reference Cloutier, Ahlberg, Stiassny, Parenti and Johnson1996).
Accounts of the internal structure of the porolepiform braincase remain sparse outside of early detailed treatments of the nasal capsule (Jarvik Reference Jarvik1942), with more recent studies explicitly indicating that future work should emphasise endocranial morphology in the group (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010, p. 376). Glyptolepis provides the most detailed picture for the group, based on models derived from serial grinding of specimens of G. groenlandica first shown by Stensiö (Reference Stensiö1963) but fully described by Jarvik (Reference Jarvik1972). This includes a detailed account of the endocast contained within the ethmosphenoid, complementing limited information available from natural endocasts in this species and material attributed – sometimes questionably – to Porolepis (Kulczycki Reference Kulczycki1960). Jarvik's (Reference Jarvik1972) account of the otoccipital division is limited to external reconstructions based on an incomplete grinding series. No details of the posterior half of the endocast are known, although some features can be gleaned from published sections (Jarvik Reference Jarvik1972).
This limited treatment of the otoccipital of Glyptolepis reflects a systematic lack of information on the posterior division of the braincase in porolepiforms, which is typically more poorly ossified and thus less well preserved than the comparatively robust ethmosphenoid in taxa where both are preserved (Jarvik Reference Jarvik1972; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011; Reference Downs, Daeschler, Jenkins and Shubin2013). This problem is particularly critical for ‘porolepidids’, a cosmine-bearing grade of porolepiforms branching outside of the Middle–Late Devonian, cosmine-free holoptychiids (Mondéjar-Fernández & Clément Reference Mondéjar-Fernández and Clément2012). Reasonably complete accounts of the external anatomy of the otoccipital are available for the three principal holoptychiid genera (Glyptolepis, Laccognathus and Holoptychius), but written accounts for ‘porolepidids’ were long restricted to a few incomplete and crudely prepared specimens attributed to Porolepis (Jarvik Reference Jarvik1972, pls 3, 10). This problem is made more acute by the taxonomic problems surrounding Porolepis itself. Much of this information for ‘porolepidid’ internal anatomy stems from description of individual components found in isolation rather than single, articulated individuals. Consequently, the degree to which parts gathered within Porolepis represent coherent suites of material rather than taxonomic chimaeras containing components of assorted anatomically primitive porolepiforms is unclear.
Recently, Mondéjar-Fernández et al. (Reference Mondéjar-Fernández, Friedman and Giles2021) redescribed the intact skull of the monotypic Early Devonian (Emsian) ‘porolepidid’-grade porolepiform Durialepis edentatus based on μCT data, providing a revised account relative to the external description by Otto (Reference Otto2007). This unique specimen (Geologisches Institut Köln GIK 991) is particularly important as a snapshot of early porolepiform anatomy in a single individual and thus not subject to the taxonomic uncertainties surrounding more fragmentary material reported in the literature and typically used in aggregate when scoring character matrices for phylogenetic analyses. Mondéjar-Fernández et al. (Reference Mondéjar-Fernández, Friedman and Giles2021) mention and figure both the ethmosphenoid and otoccipital divisions of the neurocranium of D. edentatus, but do not provide any revisions to Otto's (Reference Otto2007) original account. Based on the μCT dataset generated by Mondéjar-Fernández et al. we present a comprehensive description of the braincase and associated endocast of D. edentatus (Figs 1, 2). Significantly, this provides the first detailed account of an intact braincase in a ‘porolepidid’, as well as the first model of the endocast of the otic and occipital regions in any porolepiform. We compare the internal and external anatomy of the braincase to what is known from other early sarcopterygians, including novel tomographic datasets for the ethmosphenoids of two Middle Devonian Scottish porolepiforms: Glyptolepis paucidens; and an unnamed large holoptychiid. Considering these comparisons, we review features of the braincase and endocast supporting the monophyly of porolepiforms and more restrictive groups within this clade.
External neurocranial anatomy of Durialepis edentatus GIK 991. (a) Render of neurocranium in dorsal view, anterior to left. (b) Render of neurocranium in ventral view, anterior to left. (c) Render of neurocranium in right-lateral view.
Internal neurocranial anatomy of Durialepis edentatus GIK 991. (a,b) Render of cranial endocast with principal sensory canals, left post-temporal fossa and spiracular canal infilled, in dorsal view. (c,d) Renderings with addition of secondary lateral line network. Colour coding: light blue: post-temporal fossa and spiracular canal; grey: braincase and endocast; red: nerve canals and blood vessels; yellow: lateral line main sensory canal; and dark blue: lateral line canal network. Abbreviations: c.end = endolymphatic canal; c.n.V = canal for the trigeminal nerve; c.o.lat = ramus ophthalmicus lateralis; esc = external semicircular canal; eth.co = ethmoid commissural canal; ex.poc.Ext = extratemporal expansion of supraorbital lateral line canal; ex.poc.Ta = tabular expansion of supraorbital lateral line canal; gc.Ext = growth centre of the extratemporal; gc.L.Ex = growth centre of lateral extrascapular; gc.Pa = growth centre of parietal; gc.Pp = growth centre of postparietal; gc.Ta = growth centre of tabular; io.c = infraorbital canal; nca = nasal capsule; po.c = postotic sensory canal; pt.fo = post-temporal fossa; ro.tu = rostral tubuli; sc.c = supratemporal commissural canal; so.c = supraorbital sensory canal; sp.c = spiracular canal; tr.c = transverse canal; and ? = unknown canal.
2. Materials and methods
2.1. Material
Durialepis edentatus, GIK 991, Geologisches Institut Köln, Cologne, Germany. Holotype specimen preserving a near-complete and articulated individual across several blocks. Dermal skeleton of this specimen described by Mondejár-Fernández et al. (Reference Mondéjar-Fernández, Friedman and Giles2021).
Glyptolepis paucidens, NMS G 2016.32.98, National Museum of Scotland, Edinburgh, Scotland, UK. Mechanically prepared ethmosphenoid division from the Spital Flagstone Formation of Thurso, Caithness, Scotland.
Unnamed Scottish holoptychiid, NHMUK PV P.47838, Natural History Museum, London, England, UK. Incomplete acid-prepared ethmosphenoid division from Thurso, Caithness, Scotland. No formation is indicated in collection records, but the specimen is likely to come from the Spital Flagstone Formation (Newman & Dean Reference Newman and Dean2005). This specimen was mentioned and figured by Jarvik (Reference Jarvik1972, pl. 14, figs 1, 2) and briefly noted by Rosen et al. (Reference Rosen, Forey, Gardiner and Patterson1981, p. 191). It represents one of the specimens identified by Ahlberg (Reference Ahlberg1989) as representing a new genus and species of holoptychiid (Asperocephalus milleri nom. nud.).
2.2. Methods
2.2.1. μCT of D. edentatus
The anteriormost block of GIK 991 (containing the ethmosphenoid division) and the block comprising the otoccipital division were scanned using a Nikon Metrology HMX ST 225 CT scanner at the Natural History Museum, London, with the following settings: 205 kV; 160 μa; 6284 projections; and 0.5 mm copper filter. This resulted in a voxel size of 30.5 μm for the ethmosphenoid division and 34.2 μm for the otoccipital division.
2.2.2. μCT of Middle Devonian Scottish porolepiforms
NMS G 2016.32.98 was scanned using a Feinfocus 10–160 kV transmission X-ray source and a Perkin Elmer XRD0822 1 MP flat panel X-ray camera at the University of Edinburgh, Edinburgh, with the following settings: 135 kV; 18.8 W; 2000 projections; and 0.5 mm copper filter. NHMUK PV P.47838 was scanned using a Nikon XT H 225 ST industrial μCT scanner at the Natural History Museum, London.
2.2.3. Image segmentation and rendering
Data were segmented manually in Mimics Innovation Suite V.18.0 (http://biomedical.materialise.com/mimics; Materialise, Leuven, Belgium) and resultant Polygon File Format (PLY) files were exported into and rendered in Blender V.2.77a (http://www.blender.org; Blender Institute, Amsterdam, Netherlands).
3. Description
3.1. General remarks on the neurocranium
Because the entire right lateral wall and most of the ventral surface of the ethmosphenoid and otoccipital divisions were mechanically prepared, we have little to add to Otto's (Reference Otto2007) account of external anatomy. We focus on new information provided by μCT scanning, with an emphasis on the endocast.
3.2. Ethmosphenoid division
3.2.1. External anatomy
Re-examination of D. edentatus shows that Otto's (Reference Otto2007) description of the ethmosphenoid is largely accurate. A small number of features can be added, clarified, or corrected. A stout basipterygoid process (bp.pr, Fig. 3d) projects anterolaterally dorsal to the dorsal margin of the parasphenoid. The process bends dorsally at its caudal end to form a vertical lamina. A large foramen for the pituitary vein (c.v.pit, Fig. 3d) opens into the orbit dorsal to this lamina. The postnasal wall bears three major foramina, arranged in a diagonal series. The most dorsal and mesial of these pierces the junction between the orbital roof and the postnasal wall and represents the canal for the ophthalmicus lateralis (c.o.lat, Fig. 4b). A very large opening, not clearly visible in Otto's (Reference Otto2007) figures, lies ventrolateral to the foramen for the ophthalmicus lateralis and represents the canal for the medial profundus nerve (c.pr, Figs 3d, h, 4b). There are no small foramina that might represent ancillary, lateral branches of the profundus. A small orbitonasal canal lies lateral and ventral to that for the medial profundus. It opens at the bottom of a deep pit, posteroventral to the unossified ventrolateral wall of the nasal capsule (c.or, Figs 3d, f. h, 4b). Otto (2007, fig. 4d) mistakenly identified the conspicuous pit leading to the orbitonasal canal as an opening for the posterior nostril. The fenestra ventrolateralis (fe.vl, fig. 3d, f) lies anterior to this, in an area shown as a broken surface by Otto (Reference Otto2007).
Ethmosphenoid division of the neurocranium of Durialepis edentatus GIK 991. (a) Render in dorsal view and (b) interpretive drawing. (c) Render in right-lateral view and (d) interpretive drawing. (e) Render in ventral view and (f) interpretive drawing. (g) Render in posterior view and (h) interpretive drawing. Abbreviations: a.no = anterior nostril; ar.Eth = facet for the ethmoid articulation of the palate; as.pr, ascending process of the parasphenoid; bh.c = buccohypophyseal canal; bp.pr = basipterygoid process; c.a.ci = canal for the internal carotid artery; c.a.om = canal for the opthalmica magna artery; c.a.pal = canal for the palatine artery; c.a.ps = canal for the pseudobranchial artery; c.n.II = canal for the optic nerve; c.n.III = canal for the oculomotor nerve; c.or = orbitonasal canal; c.pin = canal for pineal; c.pr = canal for medial profundus nerve; c.v.pit = canal for the pituitary vein; cac = cavum cranii; fa.nc = notochordal facet; fe.vl = fenestra ventrolateralis; fo.au = autopalatine fossa; in.ca = internasal cavity; in.cr = internasal crista; io.c = infraorbital sensory canal; ov.Po = overlap area for the postorbital; ov.La = overlap area for lachrymal; ov.Prsp = overlap area for prespiracular; ov.Vo = overlap area for the vomer; pl.Pa = parietal pit-line; Pmx.t = tooth row of the premaxilla; pr.con = processus connectens; so.c = supraorbital sensory canal; Psph = parasphenoid; and su.Pa = midline suture between right and left parietals. Colour coding: grey: braincase; and red: nerve canals and blood vessels.
Postnasal wall of Durialepis edentatus GIK 991. (a) Render in posterior view and (b) interpretive drawing. Abbreviations: c.n.I = canal for olfactory nerve; c.o.lat = canal for opthalmicus lateralis nerve; c.or = oronasal canal; c.pr = canal for medial profundus nerve; fo.au = autopalatine fossa; Pmx.t = tooth row of premaxilla; Psph = parasphenoid; and so.c = supraorbital sensory canal.
On the ventral surface of the ethmosphenoid, the exact margins of the parasphenoid (Psph, Fig. 3f) are difficult to discern, even in tomograms. The parasphenoid has a narrow, blade-like anterior expansion, the lateral flanges of which were likely overlapped by denticulated parasphenotic plates. The medial region carries a small plate ornamented by a staggered row of denticles accommodated in a midline groove flanked by low ridges. These small teeth decrease in size posteriorly towards the large buccohypohyseal foramen (bh.c, Fig. 3f). Posterior to this, the parasphenoid expands into an ascending process (as.pr, Fig. 3f), which bears the spiracular groove. There are three pairs of openings that notch the preserved margin of the parasphenoid or, in the case of the most posterior pair, possibly pierce the bone. On the right side of the skull, the lateral lamina of the parasphenoid is missing – perhaps due to mechanical preparation – exposing an ‘L’-shaped trough. The anterior and lateral ends of this groove correspond to openings on the opposite side of the skull, with the groove itself roofed by the parasphenoid. We interpret the posterior branch of this groove as bearing the efferent pseudobranchial artery (c.a.ps, Fig. 3f), while its anterior limb is for the palatine artery (c.a.pal, Fig. 3f). A pair of deep, triangular furrows either pierce or notch the posterior portion of the parasphenoid and mark the foramina for the internal carotid arteries (c.a.ci, Fig. 3f). The posterior margin of the parasphenoid cannot be precisely determined due to the thinness of the bone.
3.2.2. Endocranial cavity and other aspects of internal anatomy
A near-complete endocast can be produced for the ethmosphenoid (Figs 2, 5). The nasal cavities (nca, Fig. 5b–d, f) are triangular in dorsal view, with vertices at the anterior limit, junction with the medial profundus nerve, and the fenestra ventrolateralis (Fig. 5d, f). The anterior face of the nasal capsule bears a canal for the anterior nostril (a.no, Fig. 5b), which is slot-like in cross-section. A broad groove corresponding to the crista rostrocaudalis of the nasal cavity is also present. The medial profundus nerve canal (c.pr, Fig. 5b, d, f) intersects the nasal cavity dorsolateral to the olfactory tract with a comparable diameter to the latter. The orbitonasal canal (c.or, Fig. 5d, f) is substantially smaller than either of these nerve canals, and lies midway between the medial profundus canal and the fenestra ventrolateralis. The horizontal bulge of the medial face of the nasal cavity is poorly developed and does not extend anteriorly as a distinct canal.
Ethmosphenoid endocast of Durialepis edentatus GIK 991. (a) Render in dorsal view and (b) interpretive drawing. (c) Render in left-lateral view and (d) interpretive drawing. (e) Render in ventral view and (f) interpretive drawing. Abbreviations: a.no = anterior nostril; bh.c = buccohypophyseal canal; c.a.cer = canal for cerebralis artery; c.a.ci = canal for internal carotid artery; c.a.om = canal for the opthalmica magna artery; c.a.pal = canal for the palatine artery; c.a.ps = canal for pseudobranchial artery; c.n.I = canal for olfactory nerve; c.n.II = canal for optic nerve; c.n.III = canal for the oculomotor nerve; c.or = oronasal canal; c.pr = canal for medial profundus nerve; c.v.pit = canal for pituitary vein; fe.vl = fenestra ventrolateralis; hyp = hypophyseal fossa; hypth = hypothalamic fossa; and nca = nasal capsule. Colour coding: grey: endocast; and red: nerve canals and blood vessels.
The ethmoid region of D. edentatus shows two particularly noteworthy features (Fig. 2). First, a narrow transverse canal links the right and left nasal capsules (tr.c, Fig. 2a). Second, a complex plexus of canals lies near the dorsal surface of the snout, between the nasal capsules. These canals extend from the trunk of the ophthalmic nerve. We term these rostral tubuli (ro.tu, Fig. 2a), matching the convention applied in other sarcopterygians. This extensive network of ramifying rostral tubules underlies the lateral line canal network (so.c, po.c, Fig. 2a), which itself gives rise to a complex network of smaller canals that reach the surface of the dermal skull roof via numerous small openings (Fig. 1).
Posterior to the divergence of the elongate olfactory tracts (c.n.I, Fig. 5b) the dorsal surface of the endocranial cavity is poorly defined due to weak mineralisation of the dorsal part of the neurocranial wall. The only feature clearly discernible lies immediately ventral to the pineal opening in the dermal skull roof, just anterior to the level of the bifurcation of the canals for the olfactory nerves. In this area, tomograms show a common canal with pinched lateral margins giving it a figure-eight shape in axial cross-section. This is consistent with, but not conclusive evidence for, pineal and parapineal tracts, but contrast in this area is too low for these features to be segmented. Broad, anteroventrally directed canals for the optic nerve (c.n.II, Fig. 5d) extend from the posterior half of the endocranial cavity. These are traced by smaller, but more dorsally located, canals for the oculomotor nerve (c.n.III, Fig. 5d). The cerebralis artery (c.a.cer, Fig. 5d) branches from the ventral margin of the canal for the optic nerve and travels ventrally until it intersects the horizontal component of the hypophyseal fossa. Canals for other cranial nerves exiting through the orbital wall are not apparent in the scan.
The hypophysis (hyp, Fig. 5d) is present immediately posterior to the point of exit of the optic nerves from the endocranial cavity. It is very large, with long vertical and horizontal components that meet at a near-right angle. Near the junction of these vertical and horizontal components of the fossa, a dorsolaterally oriented canal marks the position of the pituitary vein (c.v.pit, Fig. 5b, d, f). Lateral to the junction of the cerebralis artery and the horizontal component of the hypophyseal fossa, a canal for the opthalmica magna artery (c.a.om, Fig. 5d) extends from the fossa to intersect the orbital wall just dorsal to the anterior margin of the basipterygoid process.
Viewed from below, the ventral surface of the endocast of the hypophyseal fossa is expanded into two oblate lobes separated by a notch posteriorly and a shallow groove anteriorly. Canals for the internal carotid arteries (c.a.ci, Figs 3f, 5d, f) extend from their openings in the parasphenoid to join the posterior margin of each lobe. The canal for the palatine artery (c.a.pal, Figs 3f, 5d, f) is located more anteriorly and is largely in line with that for the internal carotid. Both the internal carotid and palatine artery canals are largely oriented along the anteroposterior axis of the skull, although each shows a minor degree of lateral deflection. The narrow canal for the pseudobranchial artery (c.a.ps, Figs 3f, 5b, d, f) is obliquely oriented with respect to those for the internal carotid and palatine arteries, and joins the lobe-like expansion of the endocast laterally. The hypophyseal opening (bh.c, Fig. 5d, f) lies approximately at the level of the canal for the palatine artery, but a horizontal component of the hypophyseal fossa extends some distance anterior to this point as a trough visible on the external surface of the braincase.
3.3. Otoccipital division
3.3.1. Identification of the occipital arch
Otto (Reference Otto2007, fig. 5c–d) noted a large ossification extending along the ventral and posterior surfaces of the otic capsules. He considered, but then rejected, an interpretation of this structure as the occipital region of the braincase on account of apparent asymmetry, instead tentatively suggesting it might be a pathological hyomandibula. Our scans show that the original interpretation of this ossification as the rear of the braincase was in fact correct (Figs 1, 6), with the two large pits noted by Otto (Reference Otto2007) representing the foramen magnum (f.m, Fig. 6h) and the entrance to the notochordal canal (c.nc, Fig. 6h). The asymmetry is not biological but instead represents taphonomic distortion of a structure that, like the rest of the otoccipital division, bears only sparse endochondral mineralisation. Figures 1, 6 illustrate these components restored to their approximate life positions; the unrestored position is shown in Supplementary Fig. 1 available at https://doi.org/10.1017/S1755691024000057.
3.3.2. External anatomy
The posteroventral portion of the otoccipital comprises a large parachordal plate that is united with the occipital arch. The parachordal plate appears to be incompletely joined with the occipital arch, with the two separated by symmetrical structures on both the right and left side interpreted as sutures. Complete separation of this combined basioccipital unit from the overlying otic capsules indicates the presence of a persistent otoccipital fissure (occ.f, Fig. 6h), large vestibular fontanelles (ve.fon, Fig. 6d) and basicranial fenestra (fe.ba, Figs 6f, 7a). It is unclear whether the basal plate was united with the otic capsules anteriorly by a narrow bridge of bone or lacked any ossified connection whatsoever. The basal plate is widest at mid-length, and tapers both anteriorly and posteriorly. A shallow groove visible on the left ventrolateral surface of the basal plate marks the probable course of the lateral dorsal aorta (gr.lda, Figs 6d, f, 7a). This does not join its counterpart anterior to the hind margin of the braincase, indicating that the lateral dorsal aortae joined posterior to the occiput. In posterior view, the occipital arch bears large openings for the dorsal nerve cord (f.m, Fig. 6h) and the notochord (c.nc, Fig. 6h). These are completely separated by a bridge of bone. The notochordal foramen is substantially larger than the foramen magnum (f.m, Fig. 6h). The dorsal surface of the occipital arch bears a rounded ridge (r.occ, Fig. 6h) along its dorsal midline. No foramina for spino-occipital nerves are visible.
Otoccipital division of the neurocranium of Durialepis edentatus GIK 991, with occipital region repositioned. Exact rearticulation is not possible due to distortion of both the otic and occipital components. Renders of the unrestored otoccipital division are shown in S.I. Figure 1. (a) Render in dorsal view and (b) interpretive drawing. (c) Render in right-lateral view and (d) interpretive drawing. (e) Render in ventral view and (f) interpretive drawing. (g) Render in posterior view and (h) interpretive drawing. Abbreviations: Bocc = basioccipital; c.end = endolymphatic canal; c.n.V = canal for trigeminal nerve; c.nc = notochordal canal; c.o.lat = canal for opthalmicus lateralis nerve; c.ot.VII = otic ramus of the facial nerve; cr.pa = crista parotica; f.m = foramen magnum; fa.hy.d = dorsal hyomandibular facet; fe.ba = basicranial fenestra; gr.lda = groove for lateral dorsal aorta; gr.v.ju = groove for jugular vein; Occ = occipital; occ.f = occipital fissure; ot.sh = otic shelf; ov.M.Ex = overlap area for median extrascapular; ov.Prsp = overlap area for prespiracular; pl.Pp = postparietal pit-line; pl.Ta = tabular pit-line; po.pr = postotic process; Pp = postparietal; pt.fo = post-temporal fossa; so.p = supraoccipital plug; sp.c = spiracular canal; su.Pp = suture between left and right postparietals; Ta = tabular; tec.fo = fossa tectosynotica; tr.o.p = transverse otic process; utr = utricular recess; and ve.fon = vestibular fontanelle.
Otoccipital division of the neurocranium of Durialepis edentatus GIK 991. (a) Render in right-ventrolateral view. (b) Render in left-dorsolateral view. Abbreviations: Bocc = basioccipital; c.o.lat = canal for opthalmicus lateralis nerve; c.ot.VII = otic ramus of the facial nerve; c.v.ju = canal for the jugular vein; c.n.V = canal for the trigeminal nerve; cr.pa = crista parotica; fa.hy.d = dorsal hyomandibular facet; fa.hy.v = ventral hyomandibular facet; fe.ba = basicranial fenestra; gr.lda = groove for lateral dorsal aorta; gr.v.ju = groove for jugular vein; Pp = postparietal; ov.Prsp = overlap area for prespiracular; po.c = postotic sensory canal; sp.c = spiracular canal; and Ta = tabular.
In most other respects, the dorsal portion of the otic portion of the braincase is as described by Otto (Reference Otto2007). Major features are summarised here for completeness, with updated interpretations where relevant. A small portion of the dorsal surface of the otic region extends posterior to the skull roof. This defines a broad area of overlap for the median extrascapular (ov.M.Ex, Fig. 6b). There is no well-developed posterodorsal fenestra leading to the supraoptic cavity. Instead, a wedge of bone marked on each side by deeply incised notches lies along the midline and the hind margin of the occipital unit. We interpret this as a supraoccipital plug (so.p, Fig. 6b, d, h). The notches that flank the plug are continuous with canals that extend to the supraoptic cavity (c.end, Fig. 6b, h). Otto (Reference Otto2007, fig. 5c, e) interpreted these as bearing the occipital artery. However, these foramina lead to canals that extend to the supraoptic cavity, matching structures interpreted as accommodating endolymphatic ducts (in Diplocercides; Stensiö Reference Stensiö1963) or more agnostically as a posterodorsal canal connected to the supraoptic cavity (in Eusthenopteron; Jarvik Reference Jarvik1980)
Two pairs of fossae mark the posterodorsal surface of the otic region. The fossae tectosynotica are the more mesial of these (tec.fo, Fig. 6h). Divided along the midline by the supraoccipital plug, these fossae are shallow and delimited laterally by a thickened ridge that marks the course of the posterior semicircular canal. Lateral to each fossa tectosynotica lies the much larger post-temporal fossa (fossa bridgei of some other accounts, but see Argyriou et al. Reference Argyriou, Giles, Friedman, Romano, Kogan and Sánchez-Villagra2018; pt.fo, Fig. 6h). Each fossa has a wide posterior opening, extending from the lateral edge of the fossa tectosynotica to the lateral edge of the expansive crista parotica. The fossa extends anteriorly to roughly the level of the transverse otic wall. A deep excavation, the outer margin of which is formed by a thickening containing the lateral semicircular canal, lies on the floor of the post-temporal fossa. The crista parotica (cr.pa, Figs 6f, 7b) is well developed and extends posteriorly to terminate in a pointed process. This, along with a modest postotic process (po.pr, Fig. 6f) near the junction of the horizontal and posterior semicircular canals, defines a pronounced notch on each side of the rear margin of the occipital region in dorsal view.
In its posterior half, the ventrolateral surface of the otic capsule bears a deep trough indicating the course of the jugular vein (gr.v.ju, Fig. 6f). The thickening for the horizontal semicircular canal defines a shallow fossa located immediately lateral to the jugular groove. A short transverse otic wall (lateral commissure of some other accounts, but see Giles et al. Reference Giles, Friedman and Brazeau2015c) divides the anterior and posterior halves of the lateral face of the otoccipital region. The jugular canal pierces the transverse wall. Dorsal to the jugular canal, the posterolateral face of the transverse otic wall bears the dorsal facet for the hyomandibula (fa.hy.d, Fig. 6d, f). Otto (Reference Otto2007, p. 10) indicating that a division between the upper and lower portions of the facet cannot be discerned, but virtually none of the otic sidewall is mineralised below the level of the jugular canal in the expected area of the ventral facet (Fig. 6e, f).
Anterior to the transverse otic wall, the braincase bears an expansive lateral otic shelf. This is badly broken on the right side of the neurocranium figured by Otto (Reference Otto2007) but is well-preserved on the right side that remained encased in matrix. Viewed dorsally, the shelf has a rounded, convex lateral margin. Its upper surface bears a deep trough for the jugular vein (gr.v.ju, Figs 6d, 7). The surface is deflected ventrally lateral to this groove, giving the otic shelf an arched profile in axial section. No foramina pierce the shelf, but multiple openings in the wall of the braincase are present dorsal to the shelf. The most posterior of these is a deep pit (sp.c, Figs 6d, 7a) at the dorsal limit of the anterior face of the transverse otic wall, immediately dorsal to the anterior opening of the jugular canal. This is continuous with a perichondrally lined canal (sp.c, Figs 2a, 8e) that extends posteriorly to join the post-temporal fossa, corresponding to the feature identified as accommodating the superior facial vein in Glyptolepis groenlandica (Jarvik Reference Jarvik1972, fig. 23) but generally termed a spiracular canal in tetrapodomorphs (Jarvik Reference Jarvik1980; Holland Reference Holland2014). A large foramen for the trigeminal nerve lies near the anterior margin of the otoccipital (c.n.V, Figs 6d, 7a), marking the end of a well-developed canal that extends anterolaterally from the endocranial cavity (c.n.V, Figs 2a, 8e, f). Between these two well-defined openings, near the dorsal margin of the neurocranium, is a deep pit with a foramen at its base (c.o.lat, Figs 6d, 7a). This foramen leads to a blind-ending canal that diverges within the neurocranial wall, sending branches towards the skull roof, but does not connect with the endocranial cavity. We tentatively follow previous identification of this feature as transmitting the ophthalmic branch of the facial nerve (Otto Reference Otto2007). A small foramen, not identified by Otto (Reference Otto2007), is apparent on the lateral wall of the braincase immediately anterodorsal to the anterior opening to the jugular canal (c.ot.VII, Figs 6d, 7a). This corresponds most closely in position to the foramen identified in G. groenlandica as accommodating the ramus oticus lateralis of the facial nerve (Jarvik Reference Jarvik1972, fig. 21b).
3.3.3. Endocast and other aspects of internal anatomy
Preservation only permits reconstruction of a partial endocast of the otoccipital region (Fig. 8). Viewed dorsally, the endocranial cavity can be divided into anterior and posterior components of roughly equal length. The anterior component is tubular, with lateral walls that taper posteriorly. A pair of canals of uncertain identity emerge from its dorsal surface, but do not appear to correspond with any prominent fontanelles on the roof of the endocranial chamber (anteriormost ‘?’ in Fig. 8e, f). In lateral view, the ventral half of the wall of the endocranial cavity is medially inset relative to more dorsal portions, resulting in a ‘stepped’ appearance. The roots for the trigeminal and facial nerves lie on this recessed surface, with the trigeminal canal extending anteriorly to exit the neurocranium directly and the facial canal extending posterolaterally.
Otoccipital endocast of Durialepis edentatus GIK 991. (a) Render of endocast in dorsal view, with (b) post-temporal fossa and spiracular canal, and (c) interpretive drawing. (d) Render of endocast in right-lateral view, with (e) post-temporal fossa and spiracular canal, and (f) interpretive drawing. Abbreviations: a.asc = ampulla of anterior semicircular canal; a.esc = ampulla of external semicircular canal; asc = anterior semicircular canal; c.end = endolymphatic canal; c.n.V = canal for the trigeminal nerve; c.n.X = canal for the vagus nerve; c.ot.VII = otic ramus of the facial nerve; cc = crus commune; esc = external semicircular canal; psc = posterior semicircular canal; pt.fo = post-temporal fossa; s.su = sinus superior; so.ca = supraoptic cavity; sp.c = spiracular canal; utr = utricular recess; and ? = unknown canal. Colour coding: light blue: post-temporal fossa and spiracular canal; grey: braincase and endocast; and red: nerve canals and blood vessels.
The skeletal labyrinth of the inner ear dominates the posterior half of the endocast. Three semicircular canals are present. The posterior canal is incomplete on the side of the braincase where the labyrinth is most easily segmented but is clearly the shortest of the three canals (psc, Fig. 8e, f). The junction between the anterior and posterior canals, corresponding to the sinus superior (s.su, Fig. 8f), is not well differentiated from the remainder of the cranial endocast, and the anterior canal is only modestly elevated dorsal to the cavum cranii. The anterior ends of the horizontal and anterior canals join the endocast through a well-developed utricular recess (utr, Fig. 8f). Both canals bear well-developed anterior ampullae, with that of the horizontal canal intersecting the recess posteroventral to that of the anterior canal. The horizontal and posterior canals are both incompletely preserved posteriorly. A raised midline region, located along the dorsal midline between the two posterior semicircular canals, represents the supraoptic cavity (so.ca, Fig. 8e, f). A mound-like extension extends from the main body of the endocast, separated from the posterior semicircular canal by a narrow gap. Based on position, this likely accommodated the base of the vagus nerve (c.n.X, Fig., 8e). Two canals extend (c.end, Figs 2a, 8e, f) posteriorly from the dorsal surface of the chamber. These diverge from one another to open on the posterior surface of the skull near the posterior margin of the skull roof (c.end, Fig. 6b); these canals are typically interpreted as bearing endolymphatic ducts.
4. Discussion
Neurocranial and endocast anatomy in early osteichthyan lineages (lungfishes: Friedman Reference Friedman2007; tetrapodomorphs: Coates & Friedman Reference Coates, Friedman, Elliott, Maisey, Yu and Miao2010; actinopterygians: Giles & Friedman Reference Giles and Friedman2014) provides abundant characters for inferring evolutionary relationships. Despite modest diversity and well-preserved material for many taxa, the interrelationships among porolepiform genera are little investigated and remain incompletely resolved (Schultze Reference Schultze2000). Here we review the neurocranial anatomy of D. edentatus, with an emphasis on possible patterns of trait evolution. Neurocranial anatomy has to date made virtually no contribution to our understanding of porolepiform interrelationships. The most inclusive cladistic analysis of porolepiform interrelationships includes 11 characters – representing almost 20% of the dataset – relating to braincase structure minus the parasphenoid (Schultze Reference Schultze2000: characters 40–51). However, these features bear on relationships among porolepiform outgroups rather than among porolepiform lineages themselves; excluding Powichthys, all 11 characters are invariant among the porolepiform taxa sampled. Although our review is principally focused on features that might help resolve patterns of porolepiform evolution, we also address aspects of D. edentatus that provide clarity on character evolution among dipnomorphs or sarcopterygians more broadly.
4.1. Endocranial structure in porolepiforms: a synopsis
For much of the modern interval of sarcopterygian systematics, which started with a series of influential papers ranging from the late 1980s to early 1990s (Panchen & Smithson Reference Panchen and Smithson1987; Ahlberg Reference Ahlberg1991; Cloutier & Ahlberg Reference Cloutier, Ahlberg, Stiassny, Parenti and Johnson1996), information on the porolepiform neurocranium and endocranial cavity was mostly restricted to details provided by Jarvik (Reference Jarvik1942, Reference Jarvik1972). Jarvik's coverage of porolepiform diversity was uneven, with the most detail available for mechanically prepared and serially ground material of the holoptychiid G. groenlandica (Jarvik Reference Jarvik1972, figs 16–19, 21–23; pls 17, 19, 22) that permitted complete reconstructions of the ethmosphenoid and its endocranial cavity, and incomplete restoration of the otoccipital division. By contrast, his descriptions of endocrania in both Holoptychius and Porolepis included only details of the external anatomy of the ethmosphenoid and poorly preserved otoccipitals, with information on the endocast largely limited to the nasal capsules of Porolepis (P. brevis and P. spitsbergensis). The description of the ethmosphenoid of Heimenia ensis (Clément Reference Clément2001) coupled with discoveries of well-preserved braincases of Holoptychius bergmanni (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2013) and Laccognathus embryi (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011) highlight some superficial external variation among porolepiforms, but diversity in internal structure remains unclear. This stands in contrast to the considerable detail now available for the neurocrania of other early-diverging dipnomorphs (e.g., Youngolepis praecursor, Powichthys thorsteinssoni and P. spitsbergensis; Jessen Reference Jessen1980; Chang Reference Chang1982; Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010). In this context, the braincase of Durialepis edentatus provides a refined perspective on endocranial conditions deep within porolepiform phylogeny.
4.2. Ethmoid division, including parasphenoid and endocast
4.2.1. Overall construction of the ethmosphenoid
In general structure, the ethmosphenoid of D. edentatus agrees with that of other porolepiforms, especially Porolepis (Jarvik Reference Jarvik1972). The most conspicuous pattern of morphological variation in porolepiform ethmosphenoids concerns their overall proportions. In Porolepis and D. edentatus, the anterior division of the braincase is proportionally shallower and narrower than the deeper, wider ethmosphenoids of Laccognathus and especially Holoptychius (e.g., Jarvik, Reference Jarvik1972, fig. 20). Glyptolepis, like these other holoptychiids, shows a relatively broad ethmosphenoid (Jarvik Reference Jarvik1972, fig. 23). The reconstruction of a relatively shallow ethmosphenoid of G. groenlandica based on serial grinding (Jarvik Reference Jarvik1972, fig. 21) appears to result from taphonomic compression, based on comparison with uncrushed material from other localities (e.g., Jarvik, Reference Jarvik1972: pl. 14). These contrasts in ethmosphenoid geometry reflect overall proportional differences between the crania of ‘porolepidids’ and holoptychiids apparent from external dermal bones. The conditions in ‘porolepidids’ agree closely with those of Powichthys (Jessen Reference Jessen1980) and early diverging tetrapodomorphs (Chang & Zhu Reference Chang and Zhu1993; Lu et al. Reference Lu, Zhu, Long, Zhao, Senden, Jia and Qiao2012, Reference Lu, Young, Hu, Qiao and Zhu2019), suggesting that this is the primitive rhipidistian arrangement. By contrast, the broad and deep ethmosphenoids of holoptychiids appear to be a derived trait of this subset of porolepiforms.
4.2.2. Parasphenoid
Because of its intimate association with the ethmosphenoid, we also consider the parasphenoid alongside the endoskeletal components of the braincase. The parasphenoid of porolepiforms is large and broad compared with other rhipidistians, where it usually consists of an elongate, anteriorly tapering rod of bone ornamented with denticles (as in tetrapodomorphs) or is reduced, unornamented, and posteriorly displaced (as in derived lungfishes; Jarvik Reference Jarvik1954). In D. edentatus, the parasphenoid is elongated and blade-like, with similar proportions to that of Heimenia (Clément Reference Clément2001) and Porolepis (Jarvik Reference Jarvik1972). This differs from the wider parasphenoids of the holoptychiids Holoptychius, Glyptolepis and Laccognathus (Jarvik Reference Jarvik1972; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011, Reference Downs, Daeschler, Jenkins and Shubin2013). These differences in shape reflect the contrasts in ethmosphenoid proportions already noted among porolepiforms.
The significance of differences in dentition is less clear, given the variability of fusion of dental plates to the underlying parasphenoid within several dipnomorph species (Jarvik Reference Jarvik1972; Chang Reference Chang1982). A longitudinal band of teeth or denticles, extending along the antero-posterior axis of the parasphenoid, is present in D. edentatus and some specimens identified as Porolepis. There are significant differences between these arrangements: the small denticles are more randomly distributed in D. edentatus compared to the single row of much larger teeth in the type specimen (ENS 214) of Porolepis brevis from Spitsbergen (Jarvik Reference Jarvik1972, fig. 65b) and P. ex gr. posnaniensis from Poland (Kulczycki Reference Kulczycki1960, fig. 1, specimen no.1). Other material from Spitsbergen attributed to Porolepis shows a broad denticle field (Jarvik Reference Jarvik1972, pl. 9), not unlike that found in Heimenia (Clément Reference Clément2001, fig 3, Lithuanian Institute of Geology LIG 45–2001a,b). In holoptychiids, denticles are small and not as broadly dispersed across the parasphenoid when they are present (Jarvik Reference Jarvik1972). The presence of parasphenoid teeth within holoptychiid genera is variable (Jarvik Reference Jarvik1972, p. 88), with this being attributed to the taphonomic loss of unfused denticle plates (Jarvik Reference Jarvik1972; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011). Despite this variation, there is a consistent pattern noted by Jarvik (Reference Jarvik1972, p. 88). In holoptychiids, the midline ridge that is common to other porolepiforms branches anterior to the level of the hypophysis, resulting in well-defined ridges extending along the anterior margin of each spiracular groove on the basicranium. This condition is also apparent in material of Laccognathus (Vorobyeva Reference Vorobyeva1980, Reference Vorobyeva2006; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011) described after Jarvik's (Reference Jarvik1972) account, and represents a possible holoptychiid synapomorphy.
4.2.3. Nasal cavities and rostral structures
The endocasts of the nasal cavities of D. edentatus correspond structurally to what is known in other porolepiforms: they are triangular in dorsal view, have a prominent groove for the crista rostrocaudalis, and bear the medial profundus canal with a much larger aperture than the lateral profundus and orbitonasal canals. The transverse canal uniting the nasal capsules of D. edentatus has otherwise only been reported in Powichthys spitsbergensis (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010, fig. 7a, b, Muséum national d’ Histoire naturelle, Paris MNHN SVD 2156). The rostral tubuli found in D. edentatus were previously considered to be a synapomorphy of Youngolepis plus lungfishes (Cloutier & Ahlberg Reference Cloutier, Ahlberg, Stiassny, Parenti and Johnson1996; King et al. Reference King, Hu and Long2018). However, μCT investigation indicates a wide distribution of such structures among sarcopterygians. This includes occurrences in the onychodont Qingmenodus yui (Lu et al. Reference Lu, Ahlberg, Qiao, Zhu, Zhao and Jia2016) and the tetrapodomorph Gogonasus andrewsae (Holland Reference Holland2014), suggesting that rostral tubuli might be a general sarcopterygian feature. A diffuse set of canals between the olfactory capsules of NHMUK PV P.47838 might represent traces of a similar structure. The well-preserved and relatively uncrushed ethmoids of other porolepiforms, especially Porolepis (Jarvik Reference Jarvik1972) and Heimenia ensis (Clément Reference Clément2001), should be examined to determine whether they also bear either of these internal features of the ethmoid region.
4.2.4. Main body of ethmosphenoid endocast
Like that of the nasal cavities, the more posterior portions of the endocast in D. edentatus agree with conditions previously described for porolepiforms. The endocranial chamber rises sharply dorsal to the level of the olfactory tracts, as reported in G. groenlandica (Jarvik Reference Jarvik1972, fig. 17). We can also confirm this arrangement in G. paucidens and NHMUK PV P.47838 (Figs 9, 10). This feature is variable among sarcopterygians, with a high chamber for the forebrain present in porolepiforms, Powichthys spitsbergensis (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010), and Gogonasus andrewsae (Holland Reference Holland2014), but a low roof apparent in Qingmendous yui (Lu et al. Reference Lu, Ahlberg, Qiao, Zhu, Zhao and Jia2016), Diplocercides kayseri (Stensiö Reference Stensiö1963), Tungsenia paradoxa (Lu et al. Reference Lu, Zhu, Long, Zhao, Senden, Jia and Qiao2012), Eusthenopteron foordi (Jarvik Reference Jarvik1980), Ectosteorhachis nitidus (Romer Reference Romer1937) and possibly Youngolepis praecursor (Chang Reference Chang1982). Like Glyptolepis, D. edentatus appears to have separate pineal and parapineal canals, a feature now known to have a relatively wide distribution among sarcopterygians with reports of a similar arrangement in Qingmenodus yui (Lu et al. Reference Lu, Ahlberg, Qiao, Zhu, Zhao and Jia2016), Gogonasus (Holland Reference Holland2014) and Powichthys (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010).
Ethmosphenoid endocast of Glyptolepis paucidens NMS G 2016.32.98. (a) Render in dorsal view and (b) interpretive drawing. (c) Render in left-lateral view and (d) interpretive drawing. (e) Render in ventral view and (f) interpretive drawing. Abbreviations: c.a.ci = canal for the internal carotid artery; c.a.cer = canal for cerebralis artery; c.a.pal = canal for the palatine artery; c.a.ps = canal for efferent pseudobranchial artery; c.n.I = canal for olfactory nerve; c.n.II = canal for optic nerve; c.o.lat = ramus ophthalmicus lateralis; c.or = orbitonasal canal; c.pin = pineal canal; c.ppin = parapineal canal; c.pr = canal for medial profundus nerve; c.v.pit = canal for pituitary vein; fe.vl = fenestra ventrolateralis; hyp = hypophyseal fossa; nca = nasal capsule; and rec.hypthl = hypothalamic recess.
Ethmosphenoid endocast of NHMUK PV P.47838. (a) Render in dorsal view and (b) interpretive drawing. (c) Render in left-lateral view and (d) interpretive drawing. (e) Render in ventral view and (f) interpretive drawing. Abbreviations: a.no = anterior nostril; c.a.ci = canal for the internal carotid artery; c.a.cer = canal for cerebralis artery; c.n.I = canal for olfactory nerve; c.n.II = canal for optic nerve; c.o.lat = ramus ophthalmicus lateralis; c.or = orbitonasal canal; c.pr = canal for medial profundus nerve; c.v.pit = canal for pituitary vein; fe.vl = fenestra ventrolateralis; hyp = hypophyseal fossa; nca = nasal capsule; and rec.hypthl = hypothalamic recess.
Canals for the optic nerves of D. edentatus extend ventrolaterally from the endocast at the anterior margin of the hypophyseal region, agreeing with the condition in other dipnomorphs as well as tetrapodomorphs (Romer Reference Romer1937; Jarvik Reference Jarvik1980; Holland Reference Holland2014) and coelacanths (Stensiö Reference Stensiö1963). Similarly, the canal for the oculomotor nerve originates close to the base of the canal for the optic nerve. Again, this is the condition in tetrapodomorphs (Romer Reference Romer1937; Jarvik Reference Jarvik1980; Holland Reference Holland2014), coelacanths (Stensiö Reference Stensiö1963) and most dipnomorphs (Jarvik Reference Jarvik1972; Clément Reference Clément2001; Clement et al. Reference Clement, Challands, Long and Ahlberg2016). We cannot resolve the course of the oculomotor nerve in our scans of G. paucidens and the unnamed Scottish holoptychiid. In these examples, the canals for the two nerves are approximately at the same level, but the canal of the oculomotor nerve originates dorsal to that of the optic nerve in D. edentatus. Youngolepis and Powichthys show more extreme arrangements with the oculomotor nerve lying far dorsal to the optic nerve (Chang Reference Chang1982; Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010). The onychodont Qingmenodus shows a substantial departure from the typical sarcopterygian arrangement of the optic and oculomotor nerves. As interpreted by Lu et al. (Reference Lu, Ahlberg, Qiao, Zhu, Zhao and Jia2016), the oculomotor nerve lies on the same plane as the optic nerve, but far posterior to it. This matches endocasts in the braincase attributed to the possible stem osteichthyan Ligulalepis (Clement et al. Reference Clement, King, Giles, Choo, Ahlberg, Young and Long2018) as well as in Devonian actinopterygians Mimipiscis (Giles & Friedman Reference Giles and Friedman2014) and Raynerius (Giles et al. Reference Giles, Darras, Clément, Blieck and Friedman2015b), and could represent evidence for placement of onychodonts outside the sarcopterygian crown group. The only other onychodont with a well-preserved braincase is Onychodus jandemarrai, for which there are contrasting interpretations of the placement of the oculomotor nerve: in an anterior position like most sarcopterygians (Andrews et al. Reference Andrews, Long, Ahlberg, Barwick and Campbell2006) and in a posterior position as in Qingmenodus and non-sarcopterygians (Long Reference Long2001). New information from the braincase of Grossius aragonensis may furnish key comparative data to settle these issues (pers. obs, J. M. F.).
4.2.5. Hypophyseal region of endocast
The hypophyseal fossa of D. edentatus, with its pronounced vertical and horizontal components that meet at a right angle, closely resembles that of G. groenlandica (Jarvik Reference Jarvik1972) and Powichthys spitsbergensis (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010). New data for G. paucidens and NHMUK PV P.47838 confirm a similar arrangement in these holoptychiid porolepiforms (Figs 9, 10). There are differences in proportions in these examples, with the horizontal division being larger than the vertical division in D. edentatus, P. spitsbergensis and G. paucidens (possibly due to compression of the specimen) but of roughly similar length in G. groenlandica and the NHMUK PV P.47838. The two Scottish holoptychiids share a feature not seen in the other taxa considered here: a ventral swelling or extension of the endocast near the posterior angle defined by the vertical and horizontal divisions of the hypophyseal fossa (rec.hypthl, Figs 9, 10). We term this feature, which is not shown in Jarvik's reconstructions of G. groenlandica, the hypothalamic recess, matching the name applied to a similar feature in the lungfish Dipterus valenciennesi (Challands Reference Challands2015, fig. 10). All these porolepiforms and Youngolepis praecursor also share a very long canal for the pituitary vein (c.v.pit, Figs 5d, f, 9f, 10f) with a strong vertical component. This differs from shorter, more horizontal canals of most other sarcopterygians, including lungfishes (Clement et al. Reference Clement, Challands, Long and Ahlberg2016) and Youngolepis praecursor (Chang Reference Chang1982, fig. 3, section 240). It is possible that this represents a synapomorphy uniting porolepiforms and Powichthys, adding to other endocast characters enumerated by Clément & Ahlberg (Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010, p. 376).
Dipnomorphs share a broadly similar arrangement of vessels in the plane between the ventral face of the sphenoid and the dorsal surface of the parasphenoid that intersect the hypophyseal fossa: a pair of vessels extending anteriorly; a pair of vessels directed laterally; and a pair of canals with a posterior orientation. These can be accommodated in deep grooves in the ventral surface of the braincase alone, or be covered by dermal bone to form completely enclosed canals. This closely corresponds to the arrangement found in early actinopterygians where the internal carotid extends through a parabasal canal situated between the parasphenoid and the ventral surface of the braincase (Gardiner Reference Gardiner1984). However, the condition common to dipnomorphs is unusual among sarcopterygians, including putative stem taxa (Yu Reference Yu1998), tetrapodomorphs (Jarvik Reference Jarvik1980; Holland Reference Holland2014), coelacanths (Forey Reference Forey1998) and onychodonts (Andrews et al. Reference Andrews, Long, Ahlberg, Barwick and Campbell2006). In these sarcopterygian groups, the carotid trunk and its branches do not extend through parabasal canals and generally leave few obvious traces on the ventrolateral surface of the basisphenoid apart from foramina for the internal carotids that lead directly to the hypophysis via short transversely oriented canals. This distribution suggests that the arrangement in dipnomorphs is derived within sarcopterygians, and thus convergent on the actinopterygian condition.
Conflicting interpretations surround the otherwise similar patterns of grooves and canals for the arterial system of the basisphenoid in early dipnomorphs. These canals were initially identified in G. groenlandica, from anterior to posterior, as accommodating the palatine artery, internal carotid artery and an undetermined blood vessel (Stensiö Reference Stensiö1963). Jarvik (Reference Jarvik1972, fig. 17) later omitted the most posterior of these canals from his reconstructed endocast of the hypophyseal fossa and associated vessels. However, his illustrations of the ventral face of the ethmosphenoid (e.g., Jarvik, Reference Jarvik1972, fig. 31) show a pair of posterior foramina exiting the parasphenoid, and this canal is clearly present in our scan of G. paucidens (Fig. 9). Jarvik (Reference Jarvik1972) identified those foramina – lacking canals in his endocast models but corresponding to the anticipated exit for the posterior, unnamed vessels of Stensiö – as bearing a medial branch of the internal carotid artery. Jessen's (Reference Jessen1980) account of foramina in the basisphenoid of Powichthys thorsteinssoni relies on external details. He identified paired posterior foramina as accommodating medial branches of the internal carotid, consistent with Jarvik's (Reference Jarvik1972) interpretation. Paired openings on either side of the parasphenoid, approximately at the level of the hypophyseal foramen, are illustrated but not labelled; these correspond to the openings for the internal carotid identified by Stensiö (Reference Stensiö1963) and Jarvik (Reference Jarvik1972). Parallel grooves lie exposed on the ventral surface of the basisphenoid anterior to the parasphenoid. These are interpreted as accommodating ‘artery vessels’, but they appear to correspond to the more completely covered canals for the palatine arteries in G. groenlandica. In Youngolepis, Chang (Reference Chang1982) retained identification of the anterior canal as accommodating the palatine artery but offered contrasting interpretations for the more posterior canals; based on comparison with Mimipiscis and Moythomasia she identified the middle, laterally directed canal as bearing the efferent pseudobranchial artery, with the posteriorly directed canal accommodating the internal carotid itself. Clément & Ahlberg (Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010) maintain the terminology applied by Chang (Reference Chang1982) in interpreting P. spitsbergensis, while also accepting the identifications given by Jarvik (Reference Jarvik1972) for G. groenlandica. However, we argue that patterns across these dipnomorphs are more anatomically similar than these contrasting naming conventions suggest. We regard the posterior openings in porolepiforms as bearing the internal carotids, and the lateral openings as accommodating the efferent pseudobranchial arteries, in agreement with the interpretation favoured by Gardiner (Reference Gardiner1984, p.275).
Durialepis edentatus disagrees with G. groenlandica as described by Jarvik (Reference Jarvik1972) in at least three features of the hypophyseal region, instead showing a closer resemblance to Powichthys and Youngolepis (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010). First, the pituitary vein canal of D. edentatus extends from the posteroventral corner of the hypophyseal fossa, not the lateral wall of the chamber as in G. groenlandica. Second, D. edentatus lacks an anterior canal linking the anterior of the hypophyseal fossa with the main body of the endocranial cavity, identified by Jarvik as bearing the pars tuberalis. Third, there is no indication of a ventral midline separation of the hypophyseal fossa in D. edentatus, whereas a median hypophyseal crest creates such a division in G. groenlandica (Jarvik Reference Jarvik1972, fig. 17). Both G. paucidens and NHMUK PV P.47838 have a pituitary vein canal that originates on the lateral wall of the hypophyseal fossa, suggesting that this arrangement – which is also found in other sarcopterygian groups (e.g., Jarvik Reference Jarvik1980, fig. 87) – might be typical of holoptychiids. The distribution of other features among holoptychiids is less clear. Glyptolepis paucidens does not appear to have a canal for the pars tuberalis and NHMUK PV P.47838 is too incompletely preserved to assess its condition. Likewise, the presence or absence of a median hypophyseal crest cannot be determined in either G. paucidens or NHMUK PV P.47838.
In sum, these observations on the hypophyseal region in D. edentatus suggest additional support for the clade including Powichthys plus Porolepiformes (long, vertically oriented canals for the pituitary vein), as well as apparent peculiarities of G. groenlandica (canals for the pars tuberalis). μCT examination of the well-preserved ethmosphenoids of other porolepiforms (e.g., Holoptychius, Laccognathus; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011, Reference Downs, Daeschler, Jenkins and Shubin2013) is necessary to determine the significance of the posterior recess apparent in NHMUK PV P.47838 and G. paucidens.
4.3. Otoccipital division and endocast
4.3.1. Overall construction of the otoccipital
Comparisons between D. edentatus and other porolepiforms is restricted by limited information available on the otoccipital for most members of the group. As with the ethmosphenoid, there is substantial variation in the shape of the otoccipital among different porolepiforms that is reflected in the geometry of the overlying postparietal shield. Most porolepiforms show proportions like those of Powichthys (Jessen Reference Jessen1980) and early tetrapodomorphs (Chang & Zhu Reference Chang and Zhu1993; Lu et al. Reference Lu, Young, Hu, Qiao and Zhu2019), where the width of the otic component of the otoccipital is similar to its length. Durialepis edentatus is unusual among porolepiforms in having an otic region that is longer than wide. Holoptychius also represents an exception, but in the opposite direction: it has an otic region nearly 50% wider than it is long (H. bergmanni, Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2013).
Aside from proportions, the otoccipital of D. edentatus confirms some common structural patterns among porolepiforms, most notably the distinct separation of the basioccipital plate and remainder of the otic region. In other sarcopterygians with a persistent fissure between the otic capsules and occipital region, there is typically a bony bridge linking the basioccipital plate to the otic region at approximately the level of the transverse otic wall (Jarvik Reference Jarvik1980; Chang Reference Chang1982; Holland Reference Holland2014). A similar bridge is present in early actinopterygians that bear a persistent otoccipital fissure and vestibular fontanelle (Gardiner Reference Gardiner1984; Giles et al. Reference Giles, Coates, Garwood, Brazeau, Atwood, Johanson and Friedman2015a, Reference Giles, Darras, Clément, Blieck and Friedmanb). By contrast, most descriptive accounts of porolepiform braincases show no evidence of either the basioccipital plate or occipital arch even where the otic capsules are well preserved (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011, Reference Downs, Daeschler, Jenkins and Shubin2013). Outside of D. edentatus, the few reported examples are poorly preserved and difficult to interpret (Jarvik Reference Jarvik1972, p. 64). In some ways, the condition in porolepiforms resembles that of the stem or deeply diverging crown osteichthyans Psarolepis and ‘Ligulalepis’, which likewise seem to have complete separation between the basioccipital and otic capsules (Yu Reference Yu1998; Clement et al. Reference Clement, King, Giles, Choo, Ahlberg, Young and Long2018). However, the condition in porolepiforms is likely convergently derived within the sarcopterygian crown based on the presence of ossified bridges between the otic capsules and basioccipital plate enclosing the anterior margins of vestibular fontanelles in Powichthys (Jessen Reference Jessen1980), Youngolepis (Chang Reference Chang1982) and tetrapodomorphs. Within porolepiforms, there is variation in this region of the basicranium. In both Porolepis (Jarvik Reference Jarvik1972, pl. 3, pl. 10.3) and D. edentatus, the undersides of the lateral otic shelf contribute exclusively to the lateral margins of the basicranial fenestra, with evidence for a basicranial plate in both taxa that encloses the posterior margin of the fenestra. By contrast, mesial extensions from the lateral otic shelves meet each other along the midline in Holoptychius (Jarvik Reference Jarvik1972, pl. 24.2; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2013) and Laccognathus (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011). These are arched and described as accommodating the dorsal surface of the notochord. Neither taxon shows evidence of a basicranial plate, and it was probably unmineralised. The conditions in Glyptolepis are difficult to discern between Jarvik's reconstructions (1972, fig. 92a) and specimen photographs (Jarvik Reference Jarvik1972, pl. 22.1), although his reconstructed section through the otoccipital (Jarvik Reference Jarvik1972, fig. 24f) suggests an arrangement matching that of Holoptychius and Laccognathus. The combination in these holoptychiids of a well-developed bony roof for the notochordal canal formed by mesial extensions from the lateral otic shelves with the absence of an ossified basicranial plate is reminiscent of the arrangement in the derived tristichopterid Mandageria fairfaxi (Johanson et al. Reference Johanson, Ahlberg and Ritchie2003). That condition is interpreted as permitting additional flexibility between the neurocranium and axial column, thereby increasing gape. Outgroup comparison with Powichthys (Jessen Reference Jessen1980) and early diverging tetrapodomorphs (Holland Reference Holland2014) indicate that the arrangement in Holoptychius and Laccognathus is derived relative to that in D. edentatus and, probably, Porolepis.
Durialepis edentatus highlights additional features of the otoccipital that vary among porolepiforms, some of which might be phylogenetically relevant. First, like Holoptychius (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2013), but unlike Laccognathus and Glyptolepis (Jarvik Reference Jarvik1972), D. edentatus lacks a large dorsal fontanelle in the otoccipital above the level of the basicranial fenestra. Outgroup comparison with tetrapodomorphs (Jarvik Reference Jarvik1980; Holland Reference Holland2014) suggests that the presence of a fenestra is derived. This could be useful in providing some clarity on the relationships of major holoptychiid lineages, which are currently unresolved (Schultze Reference Schultze2000). Second, in D. edentatus and Porolepis, the crista parotica extends as a spur-like extension that terminates posterior to the level of the postotic process. This feature is particularly long in D. edentatus, possibly representing an autapomorphy. Holoptychiids appear to show a contrasting condition, lacking a pronounced posterior extension to the crista parotica. Jarvik's (Reference Jarvik1972, fig. 23) reconstruction of G. groenlandica shows a modest extension, but it falls short of the level of the postotic process. The long extension of the crista parotica in ‘porolepidids’ appears to be derived, based on the absence of the feature in Youngolepis (Chang Reference Chang1982), Powichthys (Jessen Reference Jessen1980), early diverging tetrapodomorphs (Holland Reference Holland2014; Lu et al. Reference Lu, Young, Hu, Qiao and Zhu2019) and coelacanths (Jarvik Reference Jarvik1980). Third, D. edentatus and Glyptolepis are the only porolepiforms in which the morphology of the post-temporal fossae is known in any reasonable detail, but the two taxa show important contrasts. The fossa is substantially longer than wide in D. edentatus, contrasting with the broader, shorter fossa of Glyptolepis; this is consistent with proportional differences between the two as already apparent in the gross geometry of the otoccipital division. Both taxa share a long canal extending from the post-temporal fossa and opening on the outside of the neurocranium anteriomesial to the transverse otic process and dorsal to the otic shelf. In D. edentatus, this canal extends from the anterolateral margin of the fossa. By contrast, Glyptolepis has a complex arrangement of canals leading from the post-temporal fossa. A transverse canal, absent in D. edentatus, connects the mesial margins of the right and left fossae. A pair of canals intersect this transverse passage and exit the lateral wall of the braincase dorsal to the otic shelf, corresponding to the spiracular canal of D. edentatus. Each spiracular canal bears a mesial branch that intersects the fontanelle in the roof of the anterior otic region. The significance of these canals in Glyptolepis is unclear, although it is possible that they represent the co-mingling of a spiracular canal with an arterial network like that described on the dorsal surface of the skull in Eusthenopteron (Jarvik Reference Jarvik1980). Comparisons with immediate outgroups are challenging. Powichthys has a spiracular canal with an uncertain interior course (Jessen Reference Jessen1980), and Youngolepis lacks a canal (Chang Reference Chang1982). The most detailed comparisons can be made with tetrapodomorphs (Romer Reference Romer1937; Jarvik Reference Jarvik1980; Holland Reference Holland2014), and suggest that the termination of the spiracular canal on the mesial surface of the post-temporal fossa in Glyptolepis is derived. Durialepis edentatus shares with tetrapodomorphs a canal that intersects the post-temporal fossa on its anterior or anterolateral face. Whether the arrangement in Glyptolepis is restricted to that taxon or characteristic of some or all holoptychiids is unclear; tomographic study of well-preserved otoccipitals of Laccognathus (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011) and Holoptychius (Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2013) should provide clarity on the evolution of spiracular canal geometry within porolepiforms.
Apart from these patterns of variation within porolepiforms, conditions of the otoccipital in D. edentatus reinforce the presence of several features previously reported in porolepiforms that underpin current hypotheses of sarcopterygian relationships. Most notable are features representing probable rhipidistian synapomorphies: expansive post-temporal fossae that open via a so-called spiracular canal anterior to the lateral otic wall; a supraoptic chamber; and exit of the profundus through the intracranial joint and the trigeminal through the otoccipital. Additionally, the well-preserved otoccipital of D. edentatus also corroborates the absence of a canal piercing the otic shelf for the palatine branch of the facial nerve. This branch also appears to be extramural in other porolepiforms (Jarvik Reference Jarvik1972; Downs et al. Reference Downs, Daeschler, Jenkins and Shubin2011, Reference Downs, Daeschler, Jenkins and Shubin2013), Powichthys (Jessen Reference Jessen1980) and coelacanths (Jarvik Reference Jarvik1980), but is present in tetrapodomorphs (Romer Reference Romer1937; Jarvik Reference Jarvik1980; Holland Reference Holland2014). The condition in Tungsenia – the sister lineage of all other tetrapodomorphs – is unclear due to deficient preservation (Lu et al. Reference Lu, Young, Hu, Qiao and Zhu2019). Thus, while a canal for the palatine branch that pierces the otic shelf is clearly a derived feature of tetrapodomorphs, it is uncertain whether it unites all known members of the group or instead a large subset of them.
4.4. Overview of neurocranial characters bearing on porolepiform phylogeny
In sum, our examination of D. edentatus, in combination with past accounts of endocranial anatomy in porolepiforms and character lists, suggests the following character scheme:
Possible synapomorphies of Dipnomorpha
(1) Grooves for internal carotids, efferent pseudobranchial, and palatine arteries in surface of basisphenoid. These may be overlain by the parasphenoid to form parabasal canals.
Possible synapomorphies of Porolepiformes plus Powichthys
(2) Hypophysis with pronounced vertical and horizontal components meeting at right angle (Clément & Ahlberg Reference Clément, Ahlberg, Elliott, Maisey, Yu and Miao2010). Present in Powichthys, D. edentatus, and Glyptolepis. Condition unknown in other porolepiforms.
(3) Long, vertically oriented canals for the pituitary vein. Present in Powichthys, D. edentatus, Glyptolepis (G. groenlandica, G. paucidens) and NHMUK PV P.47838. Condition unknown in other porolepiforms.
(4) Transverse canal between nasal capsules. This character is present in both D. edentatus and Powichthys spitsbergensis. A transverse canal is not apparent in physical tomographic sections of G. groenlandica (Jarvik Reference Jarvik1972) or Youngolepis (Chang Reference Chang1982), or μCT sections of G. paucidens and NHMUK PV P.47838. Other well-preserved early dipnomorphs (e.g., Porolepis, Heimenia, Powichthys thorsteinssoni) should be examined with μCT to better establish the distribution of this trait.
Possible synapomorphies of Porolepiformes
(5) Parasphenoid pierced ventrally by canals for the internal carotid and efferent pseudobranchial artery.
(6) Loss of the ossified bridge between the basioccipital plate and the anterior of otic region. This characterises D. edentatus and Porolepis. The lack of any obvious basioccipital ossification in Glyptolepis, Laccognathus and Holoptychius (see below) is considered here an elaboration of this condition.
Possible synapomorphy of D. edentatus plus Porolepis
(7) Long, spur-like extension of crista parotica.
Possible autapomorphy of D. edentatus
(8) Otic component of the otoccipital substantially longer than wide.
Possible synapomorphies of Holoptychiidae
(9) Prespiracular ridges on the parasphenoid (Jarvik Reference Jarvik1972).
(10) Unossified basicranial plate (Jarvik Reference Jarvik1972).
(11) Mesial extensions of the otic shelf embracing the dorsal surface of the notochord.
Characters of unclear significance
(12) Canal linking the anterodistal tip of the hypophyseal fossa with main body of the neural endocranial cavity (canal for the pars tuberalis of adenohypophysis of Jarvik Reference Jarvik1972). Among porolepiforms, this character is only apparent in G. groenlandica. It is absent in D. edentatus and G. paucidens. This character is derived based on its absence in outgroups to Porolepiformes.
(13) Median hypophyseal crest dividing the anteroventral component of hypophyseal fossa into paired lobes. The condition is only known within porolepiforms for D. edentatus (absent) and G groenlandica (present). Preservation of this region in G. paucidens is not sufficient to establish the state in this species.
5. Conclusions
μCT scanning of the braincase of Durialepis edentatus provides the first detailed model of the endocast of the ethmosphenoid division in a ‘porolepidid’-grade porolepiform, and the only reasonably intact endocast of the otoccipital division in any non-dipnoan dipnomorph apart from Youngolepis. These new data reinforce past interpretations of the anatomy of porolepiform neurocrania and endocasts derived principally from the holoptychiid Glyptolepis groenlandica. The ethmoid region of D. edentatus contains an extensive network corresponding to the rostral tubuli of some other sarcopterygians. The otoccipital is broadly similar in structure to that of tetrapodomorphs, reflecting features of the last common ancestor of rhipidistians, highlighting unusual features of holoptychiid otoccipitals that might represent derived features uniting that subset of porolepiforms, including the lack of an ossified basicranial plate and the presence of an extension of the otic shelf that embraces the dorsal surface of the notochord. Although neurocranial remains are known for most porolepiform genera, detailed information on their structure is lacking. Our study of the well-preserved braincase and endocast of D. edentatus provides a template for studies of other porolepiforms, with taxa such as Porolepis, Heimenia, Holoptychius and Laccognathus representing priorities for future work. Additional information on neurocrania and endocasts may provide data to help refine porolepiform interrelationships, which remain poorly resolved.
6. Data availability
Raw data (.TIFF stacks), Mimics files and three-dimensional (3D) surface files (zipped .PLY files) for each of the D. edentatus scans, as well as raw data (.TIFF stack) and 3D surface files (.STL) files for NHMUK PV P.47838 and the 3D surface (.PLY) file for G. paucidens NMS G 2016.32.98 are deposited in Zenodo (10.5281/zenodo.10647957).
7. Supplementary material
Supplementary material is available online at https://doi.org/10.1017/S1755691024000057.
8. Acknowledgements
It is our pleasure to present this paper in recognition of the continuing palaeontological contributions of our friend and colleague Tim Smithson. Tim's early work with Alec Panchen critically reviewing lobe-finned fish interrelationships was central to the renaissance in sarcopterygian research in the late 1980s and early 1990s. Much of that work drew on the reconsideration of Early Devonian members of Dipnomorpha; we present this new information on Durialepis as a tribute to those major efforts. This work began while M.F. and S.G. were in the Department of Earth Sciences, University of Oxford. Access to material was facilitated by Michael Amler (GIK), Emma Bernard (NHM) and Stig Walsh (NMS). We also thank the two reviewers for their comments on an earlier version of the manuscript.
9. Financial support
S.G. was supported by a Royal Society Dorothy Hodgkin Research Fellowship (no. DH160098). J.M.F was supported by the Louis Gentil-Jacques Bourcart prize of the French Academy of Sciences. Support for the scanning of NMS G. 2016.32.98 was provided to T.J.C. by Callidus Services Ltd.
10. Competing interest
The authors declare no competing interests.
Open access
