Institutional abbreviations: NHMUK, Zoological collections, Natural History Museum, London; NHMUK PV P, Palaeontological collections, Natural History Museum, London; NMS, National Museums of Scotland, Edinburgh. Anatomical abbreviations: Ancl, anocleithrum; Cla, clavicle; Cle, cleithrum; Hu, humerus; la.pbl, postbranchial lamina; lep, lepidotrichia; pr.ax, axial process of mesomere; ra.prax, preaxial radials; ra.ptax, postaxial radials; ri.ot, outturned ridge of the cleithrum; sc, scale; Scc, scapulocoracoid; Ule, ulnare; Uln, ulna; Uln + Rad, cartilage formed by fusion of radius and ulna. Mesomeres are indicated by “m” followed by a number indicating their position in a proximo-distal sequence.

NMS 1875.29.45 consists of a left cleithrum and clavicle preserved in lateral (external) view and an associated pectoral fin with a series of endoskeletal components. The specimen is well preserved with most of the fin unobscured by matrix. EDX images indicate the presence of scale cover, particularly distally (Figure 5), obscuring internal fin structure, as is the case in most lungfish fossil specimens (Figure 1). The pectoral fin of NMS 1875.29.45 measures roughly 28 mm in length, and based on comparisons with more complete specimens we estimate of total length of the individual as approximately 150 mm. Pentlandia macroptera rarely exceeds sizes greater than approximately 170 mm in total length (pers. obs., EJ of material in NHM fossil fish collection), so we assume that NMS 1875.29.45 preserves the remains of an adult individual.

Shoulder girdle. We suggest that there has been only minor post-mortem disturbance to the pectoral girdle and fin, with loss of some scales from the latter (Figures 4, 5). The cleithrum is a splint-shaped bone with a broad, rounded dorsal margin. A thickened ridge extends along the anterior margin of the exposed portion of the bone, and represents the remains of the outturned ridge of the lateral lamina of the cleithrum that is characteristic of many Paleozoic lungfishes (e.g., Johanson, 2003; Campbell et al., 2006: Figure 6; Friedman, 2010: Figure 8). Small bony buttresses extend from the posterior margin of the outturned ridge. The remainder of the lateral lamina of the cleithrum is smooth, suggesting that it was deeply buried within soft tissue. The postbranchial lamina is largely exposed, and has a rough surface texture. The cleithrum of Pentlandia appears proportionally narrower than that in probable close relatives like Barwickia (Long and Clement, 2009: Figures 3a,b), Howidipterus (Long and Clement, 2009: Figures 3c,d), and Scaumenacia (Jarvik, 1980: Figure 335). The posterior margin of the cleithrum is gently concave. A shallow excavation along the anteroventral margin of the cleithrum accommodates the posterior margin of the clavicle. The clavicle is leaf-shaped. A broad, shallow trough extends along the ventral half of this bone. Its lateral surface bears a field of irregularly spaced pores. No scapulocoracoid is visible, suggesting it either was unmineralized or remains covered posteriorly by matrix. However, the wide gap between the dermal girdle and the articulated fin skeleton suggests that the scapulocoracoid might have been large and posteriorly extensive as in Dipterus (Campbell et al., 2006: Figure 3a).

Pectoral fin. The articulated pectoral fin in NMS 1875.29.45 is posterior to, and separated from, the pectoral girdle, indicating either a large scapulocoracoid, some taphonomic displacement, or a combination of the two. The outline of the pectoral fin itself is largely preserved, including lepidotrichia along the preaxial and postaxial margins, and is similar in shape to that of Dipterus (Figures 1, 3A) and Neoceratodus (Figures 6, 7). The preaxial margin is directed ventrally. The visible face of the pectoral-fin endoskeleton therefore corresponds to the surface designated as “dorsal” in many accounts of sarcopterygian appendicular anatomy (e.g., Ahlberg, 1989; Coates et al., 2008). Here we refer to this as the upper surface in order to avoid confusion with directional terminology applied to the positional arrangements of skeletal components as they are preserved.

The fin endoskeleton comprises a series of distinct units, or mesomeres, that extend distally and form the central axis of the internal fin skeleton. Radial ossifications are visible in association with these mesomeres, and are described below and identified by comparison with the pectoral fin skeleton of Neoceratodus (Figures 6, 7). In some cases, bones of the fin endoskeleton are covered by, or in close proximity to, scales encasing the fin lobe. However, components of the fin endoskeleton are easily distinguished from the more weakly mineralized, ornamented scales in EDX images (e.g., Figure 5H). The fin skeleton appears largely undisturbed within the fin itself, permitting identification of fin mesomeres and radials when they are preserved in line with surrounding elements.

The proximal fin mesomeres are preserved as thin sheaths of perichondral bone, but more distal examples are not visible due to scale cover or what was likely weak mineralization in this region of the fin. More proximal radials are also preserved, but more distal radials cannot readily be observed.

The first mesomere, or humerus, is a squarish, flat bone. The proximal margin (articulating with the scapulocoracoid) is straight, while the distal margin of the bone is curved. The humeral mesomere is approximately twice as long as the second mesomere, or ulna. The exposed surface of the humerus carries a large axial process (“dorsal process” of Ahlberg, 1989), the long axis of which is aligned with that of the fin (Figure 4C). Endochondral bone is exposed at the distal end of the humerus, indicating either an unfinished articular facet that would have been capped in cartilage in life or, alternatively, damage to the overlying perichondral ossification. The second mesomere, or ulna, is wider anteroposteriorly with curved margins. It carries an irregularly-shaped axial process, though smaller and less distinct than that on the humerus. There is no evidence that the ulna bore either pre- (i.e., an intermedium) or post-axial radials.

As preserved, the ulna lies dorsal to a bone of slightly smaller size that we interpret as the radius. The radius does not carry an axial process like that found on the ulna, and its distal surface is aligned with a chain of two radial ossifications (Figure 3C, connected by gray lines). The radius designates the preaxial part of the pectoral fin, and is directed ventrally in NMS 1875.29.45 (i.e., toward the bottom of Figures 3B, 4, 5). This contrasts with the life orientation in Neoceratodus, where the dorsal edge of the pectoral fin corresponds to the preaxial margin. In this genus the radius and associated radials are dorsally directed in life position (Figures 6, 7). An equivalent “rotated” geometry has also been identified for Dipterus where the specimen is preserved in true lateral view (Ahlberg and Trewin, 1995); in other Dipterus shown in Figure 1, the skull and pectoral girdle and fin have been rotated slightly relative to the rest of the skeleton, so that the pectoral fin is held away from the body and rotated with the preaxial surface oriented ventrally. We cannot distinguish whether the arrangement apparent in Pentlandia represents post-mortem displacement comparable to specimens of Dipterus in Figure 1 (the skull and postcranial skeleton are not preserved) or whether that the fin was held with its preaxial margin directed ventrally in life.

Four more mesomeres are clearly visible distal to the ulna or second mesomere, with suggestions of at least two more toward the distal end of the fin that are largely obscured by overlying scales. These final preserved mesomeres are not clear under reflected light, but are apparent as higher abundances of phosphorus and calcium in EDX images (Figures 5D–F). Mesomeres three through six are all comparable in size to the ulna. Ossified pre- and post-axial radials are preserved in articulation with mesomeres three and four (Figure 4C, gray lines). The radial ossifications are roughly subquadrate and do not contact one another or their supporting mesomeres, suggesting the presence of extensive cartilaginous caps. The preaxial radial of mesomere three bears three ossified segments, while only a single ossification of the postaxial radial is preserved. Mesomere four bears a preaxial radial that consists of two ossifications in series. The postaxial radial of the mesomere is similarly represented by a chain of two ossifications. The most distal ossification is not apparent in specimen photographs, but is clear in false-color images derived from EDX for calcium and phosphorus (turquoise in combination, indicated by arrowhead in Figure 5F). A second associated region of high phosphorus and calcium concentration might be misinterpreted as another fin radial element, but targeted s.e.m. imaging of this ossification shows thin ridges of ornament suggesting that it is a scale. Radials are not preserved or visible in the more distal regions of the fin.

There is no definitive evidence for branching radials of the sort found in Neoceratodus (Figures 6, 7). Two bone fragments are preserved immediately distal to and in line with the proximal preaxial radial segment of the ulnare. These could either represent portions of a single damaged radial, or two separate radials representing a branching structure.

The bone of individual mesomeres is thin and appears to preserve a pattern of concentric circles on the surface (Figure 4). This suggests that ossification of the mesomeres progressed from the bone center both laterally and proximodistally, and that these cartilages had unmineralized proximal and distal surfaces. The bony mesomeres as preserved therefore do not accurately indicate the size of the fin cartilage. This conclusion is supported by two further observations: the absence of abutting relationships between successive mesomeres, and the lack of clear articular facets on the mesomeres to which the fin radials would have attached (e.g., compared to the distinct articular facets on the cartilaginous mesomeres of the fin of Neoceratodus; Figure 7) with the possible exception of a distal facet on the humerus.

Lepidotrichia are present along both sides of the fin. Their proximal extent is greatest along the postaxial margin as in Neoceratodus (Figure 7A), but unlike Glyptolepis where both the preaxial and postaxial series of lepidotrichia insert at a similar distance from the proximal end of the fin (Ahlberg, 1989: Figure 5B). All lepidotrichia have long unjointed bases. The lepidotrichia are segmented distally, but do not show any evidence of branching (Figure 4).

The pectoral fin of Neoceratodus has been described several times previously, most notably by Günther (1871), Semon (1898), Braus (1906), Holmgren (1933: Figures 9–16), Goodrich (1958: Figure 152), Jarvik (1980: Figure 334), Rosen et al. (1981: Figures 30, 31), Shubin and Alberch (1986: Figure 18), Schultze (1987), Joss and Longhurst (2001) and Johanson et al. (2004, 2007). Holmgren (1933) was one of the first to describe the ontogeny of fin development in detail, including the mesomeres comprising the fin axis and pre- and postaxial radials, followed by Shubin and Alberch (1986), Joss and Longhurst (2001) and Johanson et al. (2004, 2007).

One of the earliest stages described in Neoceratodus shows the presence of the cartilaginous scapulocoracoid and humerus (first mesomere), as well as the second mesomere (Semon, 1898; Braus, 1906: Figure 206; Holmgren, 1933: Figures 10, 11; Joss and Longhurst, 2001: Figure 21.1b; Johanson et al., 2004: Figure 2). In the next stages, a small cartilaginous element begins to form in apparent association with the humerus (e.g., Holmgren, 1933; Ahlberg, 1989; Joss and Longhurst, 2001), representing the anlage of the radius (Figure 6A). The main axis of the fin continues to develop cartilaginous mesomeres distally, while more distal segments develop associated with the radius (Holmgren, 1933: Figures 11, 13; Joss and Longhurst, 2001: Figures 21.1d; Figure 6B, arrowhead). This is contrary to description of Shubin and Alberch (1986: Figure15), which suggests that the cartilaginous mesomeres continue to condense distally while branching to produce preaxial fin radials, with postaxial radials developing subsequently. In other words, Shubin and Alberch (1986: Figure15) suggest that the mesomeres of the axis and the preaxial radials developed concurrently from a sequence of branching events. As noted by Holmgren (1933: Figure 13), preaxial fin radials do develop in association with more distal mesomeres, as suggested by Shubin and Alberch (1986). However, postaxial fin radials begin to develop soon after, and from this point, fin radials condense and develop both pre- and postaxially (Joss and Longhurst, 2001: Figure 21.1e; Figure 6C). Fin radials also begin to develop in association with the ulna on the postaxial side of the fin, although this is a later ontogenetic event (Holmgren, 1933: Figures 14, 15; Figure 6C). Radials develop independently from the fin mesomeres (Holmgren, 1933; Joss and Longhurst, 2001; Johanson et al., 2007; Figure 6C), and only later do pronounced articular surfaces for radials develop on distal radials (Figure 6D, white arrowheads).

In larger specimens and adults (Figures 6D, 7A,B), the first two mesomeres develop small axial processes on the ventral surface, with the process on the humerus being the larger (Figures 6D, 7A,B). Processes are also developed on the ventral surface of these mesomeres. The preaxial radials have a one-to-one correspondence with the mesomeres (beginning with the ulnare), while postaxially, the numbers are more variable, with three radials associated with the ulna in and two with the ulnare (Figures 6D, 7A,B). The most proximal radials associated with the ulna form a complex structure, with the main radial and associated cartilages, which become partially fused in the adult to varying degrees (Günther, 1871; Holmgren, 1933; Rosen et al., 1981; Figures 6A,B). Interestingly, in Figure 6C (also Joss and Longhurst, 2001: Figure 21.1e), there are additional preaxial fin radials associated with the ulnare and the fourth mesomere, but these must be lost during growth (compare these mesomeres in Figure 6C with those in Figures 6D, 7A,B). In the adult pectoral fin, the mesomeres and radials are very irregular distally, with successive radials becoming fused (Figure 7C). This is likely due to individual variation, and suggests that patterning of radial development is not tightly constrained in distal parts of the fin. This extensive distal fusion is not readily apparent in existing figures of the pectoral-fin endoskeleton of Neoceratodus, many of which are highly diagrammatic (e.g., Günther, 1871; Rosen et al., 1981).

Further aspects of the anatomy of the pectoral fin of Pentlandia can be clarified by comparison to the pectoral fin of Neoceratodus, particularly in earlier ontogenetic stages. The most relevant comparison can be drawn between the structures immediately distal to the humerus. In Pentlandia, the fin skeleton bears two separate elements in this area. This pattern is not observed in the adult Neoceratodus, but matches the arrangement, position, and relative sizes of the ulna (larger) and radius (smaller element) present early in the ontogeny of the living genus (Figures 6A–C). As discussed above, NMS 1875.29.45 is considered to belong to an adult Pentlandia, so we consider two alternative interpretations of the radial and ulnar ossifications in this genus that are consistent with these observations. First, the radius and ulna might have been separate cartilages in adult Pentlandia, as is primitive for the sarcopterygian fin (Ahlberg, 1989; Friedman et al., 2007). Second, the constituent cartilages of the humerus and ulna might have fused in adult Pentlandia but maintained separate ossification centers. We prefer the former interpretation, as described in more detail below.

Two extra radial bones are associated distally with the radius in the fin of Pentlandia (Figures 3, 4), which compares well with the smaller specimens of Neoceratodus (Figures 6B–D). There is one preaxial and one postaxial radial associated with the ulnare (mesomere three) in Pentlandia, with one and three segments, respectively (Figures 4, 5). The next axial mesomere also has one pre- and postaxial radial, with two segments in each. Fin radials are not visible in the more distal parts of the fin. By comparison with Neoceratodus, these more distal radials develop later in ontogeny, after the main fin axis (e.g., Figure 6C). In Pentlandia, the cartilaginous fin radials in this region may have not yet condensed, or, more probably, not yet ossified.

Because of the rarity of pectoral-fin endoskeletons among fossil dipnomorphs, even the incomplete example in Pentlandia adds considerably to our understanding of structural diversity and evolution in this group (Figure 8). Our interpretations of trait evolution are informed by well-constrained hypotheses of deep divergences within sarcopterygians (Maisey, 1986; Ahlberg, 1991; Cloutier and Ahlberg, 1996; Friedman, 2007; Friedman et al., 2007; Zhu et al., 2009), along with placement of Glyptolepis, Dipterus, Pentlandia, and Conchopoma as successively more crownward members of the dipnoan stem lineage (Schultze, 2001; Lloyd et al., 2012). We assume these relationships to be fixed for the purposes of our character analysis. Following Friedman and Brazeau (2010), we regard the character state shown by the ingroup as derived when a contrasting arrangement is shared by two or more successive outgroups (the “first doublet” rule of Maddison et al., 1984).

As preserved, the pectoral-fin endoskeleton of Dipterus is represented only by a series of mesomeres (perhaps as many as eight; Ahlberg and Trewin, 1995: Figure 2), multiple unbranched but jointed preaxial radials, and a single postaxial radial. In Conchopoma, pre- and postaxial radials are unjointed, and the large number preserved suggest that a large number of mesomeres would have been present (Schultze, 1975: Figure 9; Figure 3F). There are no major structural disagreements between the arrangement found in Dipterus and that reported here for Pentlandia, although only the coarsest comparisons can be made due to the poor state of preservation in Dipterus. The most conspicuous difference between these two lungfishes is the apparently greater degree of mineralization of the pectoral-fin endoskeleton in Pentlandia relative to Dipterus, where the corresponding structure appears principally cartilaginous (Ahlberg and Trewin, 1995, p. 160).

More detailed comparisons can be drawn with the porolepiform Glyptolepis, which is represented by the most complete suite of pectoral-fin endoskeleton material of any fossil dipnomorph (Ahlberg, 1989; Figures 2A,C). This genus shares with Pentlandia a long chain of axial mesomeres and extensive series of both pre- and postaxial radials (also predicted for Conchopoma, Figure 8, characters 1–2). Outgroup comparison with tetrapodomorphs and coelacanths suggests that both represent derived dipnomorph features. In both Pentlandia and Glyptolepis, radials are unbranched but jointed (unjointed in Conchopoma), mesomeres articulate with only single pre- and post-axial radials, and there is no postaxial radial associated with the second mesomere (unknown in Conchopoma). This disagrees with conditions in Neoceratodus, where extensive branching of radials is present and some mesomeres, including the second, articulate with multiple postaxial radials (Figures 6, 7). We suggest that the arrangement common to Pentlandia and Glyptolepis is primitive for dipnomorphs, and that the features apparent in Neoceratodus are specialized within the group (cf. Friedman et al., 2007; Figure 8, characters 8–10).

Pentlandia shares with Glyptolepis the absence of a preaxial radial, or intermedium, on the second mesomere. By contrast, a distally segmented radial articulated with the second mesomere is present in adult specimens of Neoceratodus (Figure 7). However, developmental sequences for Neoceratodus clearly show that this chain of cartilages is associated with the radius, which fuses ontogenetically with the ulna to form the second mesomere in Neoceratodus. These same sequences also show that the ulna bears no preaxial radials at any developmental stage in Neoceratodus (Joss and Longhurst, 2001; Johanson et al., 2007; Figure 6). The condition shown by these dipnomorphs contrasts strongly with that seen in outgroups, which consistently bear an intermedium (Friedman et al., 2007; Coates et al., 2008). We regard loss of the intermedium as a potential dipnomorph synapomorphy (character 3). The condition in Conchopoma is unknown.

The most prominent differences between the pectoral-fin endoskeletons of Glyptolepis and Pentlandia relate to the structure of mesomeres, their relationships with associated radials, and symmetry of the fin itself. First, all well-preserved mesomeres in Glyptolepis bear prominent axial processes. By contrast, only the humerus and ulna bear well-developed axial processes in Pentlandia, a condition shared with Neoceratodus. Development of axial processes on distal mesomeres characterizes both fossil and living coelacanths (Millot and Anthony, 1958; Friedman et al., 2007; well-developed axial processes are not present in tetrapodomorphs), so we regard the condition in Glyptolepis as primitive relative to that common to Pentlandia and Neoceratodus (Figure 8, character 6). The state of the axial processes in Conchopoma is unknown, but we infer that they would be restricted to proximal mesomeres as in Pentlandia and Neoceratodus.

Second, there is no independent ossification representing the radius in Glyptolepis, but instead a single bone articulating with the distal surface of the humerus. By contrast, Pentlandia bears two separate ossification centers in articulation with the distal surface of the humerus and which correspond to the radius and ulna of other sarcopterygians. Neoceratodus has a single cartilaginous element formed by developmental fusion of separate radial and ulnar cartilages, but lacks any ossification. The arrangement in Conchopoma cannot be determined. These contrasting conditions make it difficult to infer possible sequences of character change. The presence of separate ossification centers for the radius and ulna in Pentlandia is clearly primitive based on comparison with other sarcopterygians (Ahlberg, 1989; Friedman et al., 2007). In Neoceratodus, the fused ulna and radius are noticeably broader than more distal mesomeres (Figures 6D, 7A,B), while in Glyptolepis, the second mesomere articulating with the humerus is the same width as more distal mesomeres (Ahlberg, 1989: Figures 4, 5; Figure 2C). This agreement in size suggests that the second mesomere of Glyptolepis does not represent a coalesced radius and ulna. Since it appears that the cartilaginous mesomere representing the ulna condenses before the radius (Joss and Longhurst, 2001: Figure 21.1a; Johanson et al., 2004: Figure 4; Figure 6A), it is possible that the mesomere in Glyptolepis represents the ulna and a radius did not form in this taxon, or was lost during ontogeny (Figure 8, character 5). In any event, this is a very different condition from that seen in Pentlandia and Neoceratodus. We therefore suggest that the condition of a single bone or cartilage articulating with the distal surface of the humerus in Glyptolepis and Neoceratodus is likely not homologous between these two taxa.

Third, Pentlandia bears an ossification center in line with and distal to the radius. Distal segments of the radius remain independent in adult Neoceratodus, and articulate with, but are not incorporated into, the fused radial + ulnar cartilage. Conditions in Conchopoma are not known. In Glyptolepis, no radials articulate with the second mesomere, further supporting hypothesized loss of the radius rather than its fusion with the ulna as in Neoceratodus. The presence of additional bones or cartilages that articulate with the distal end of the radius is a distinctive feature of Neoceratodus and Pentlandia, but the polarity of this character is difficult to establish. The absence of a discrete radius in Glyptolepis means direct comparisons cannot be made with this taxon. Conditions in more distant outgroups are variable. Most non-digited tetrapodomorphs show no ossifications associated with the distal surface of the radius, but such additional segments are present in some rhizodonts (Coates et al., 2008: Figure 4). Modern coelacanths have a greatly reduced radius, but early fossil examples show long preaxial radials with no indication of segmentation (Friedman et al., 2007: Figure 1). At present, it is equally parsimonious to consider distal segmentation of the radius as: a feature of dipnomorphs and tetrapodomorphs that is secondarily lost in tetrapodomorphs crownward of rhizodonts, or a character derived independently in rhizodonts and dipnomorphs.

Fourth, the insertion of the proximal lepidotrichia of the preaxial finweb is distal to the equivalent insertion of the postaxial finweb in Pentlandia. This agrees with the condition in Neoceratodus (Figure 8, character 7), but differs from the arrangement in Glyptolepis. This porolepiform genus shows the primitive sarcopterygian arrangement (Ahlberg, 1989), with the insertion for the proximal insertion for the preaxial finweb lying proximal to that for the postaxial finweb (Friedman et al., 2007).

Among living lungfishes, it is clear that the pectoral-fin endoskeleton of Neoceratodus is more primitive in overall architecture than those of Lepidosiren and Protopterus, where this structure is reduced to a chain of mesomeres. However, it is also clear that the fin of Neoceratodus is modified relative to primitive dipnomorph conditions (cf. Friedman et al., 2007). When considered alongside that of other fossil dipnomorphs, the material of Pentlandia described here places additional constraints on the evolutionary timing of major anatomical changes that characterize Neoceratodus. The high number of axial mesomeres, extensive series of pre- and post-axial radials, and absence of the intermedium characteristic of porolepiforms and lungfishes must have arisen no later than the minimum time of divergence between those groups, which is indicated by the Lockhovian Powichthys, Youngolepis, and Diabolepis (Jessen, 1980; Chang and Yu, 1984). Pentlandia indicates that reduction of axial processes on distal mesomeres and expansion of the extent of the postaxial finweb relative to the preaxial finweb occurred prior to the Givetian. Given the paucity of available material for dipnomorphs in general and lungfish more specifically, it is not possible to provide a minimum date for the origin of four additional derived features of Neoceratodus: branching radials (Figure 8, character 8; also absent in Conchopoma); multiple postaxial radials in articulation with single mesomeres (Figure 8, character 9); the presence of postaxial radials in association with the ulna (Figure 8, character 10); and fusion of the radial and ulnar cartilages. However, it is clear that the first three characters, and likely the fourth, arose on the lungfish stem crownward of Pentlandia. Glyptolepis is characterized by its own specializations: the presence of radials that do not bear lepidotrichia (Friedman et al., 2007; convergent examples of “naked” radials are found in Latimeria and digit-bearing tetrapods), and possibly the loss of the radius (see above; Figure 8, characters 4–5).