Rhomboid scale: Francillon-Vieillot et al. (1990) compiled the characteristics of scales. They distinguished between rhomboid (rhombic) and elasmoid scales within osteichthyans. Rhomboid scales (Fig. 2) are thick, possess peg-and-socket articulation and the longest (diagonal) axis of the scales is oblique to the long axis of the fish (Gross, 1966 and Schultze, 1966), whereas elasmoid scales are thin, overlap each other widely, and their long axis is parallel to the long axis of the fish. The rhomboid scale possesses a free exposed field, which is covered by ganoine, enamel or exposes bone, whereas the anterior and dorsal areas surrounding the free area, are covered by surrounding scales and formed by bone; they slope towards the margin. The ‘anterior ledge’ (Qu et al., 2013) is the anterior overlapped area of a rhomboid scale. The overlapped area extends from the anterior margin of the scale (Qu et al., 2013, fig. 4a, b; Schultze, 1966, fig. 1a) over to the dorsal margin (Schultze, 1966, fig. 1a), where it is not marked in Qu et al. (2013, fig. 4a, b). The anterior margin is covered by scales of the preceding vertical scale row and the dorsal margin by the dorsal following scale in the same vertical row. The rhomboid scale is the basic structure of osteichthyans; it is divided based on histology into cosmoid and ganoid scales, and within the ganoid scale into palaeoniscoid and lepidosteoid scales.

Elasmoid scale: Elasmoid scales appear in different lineages of ganoid and cosmoid scales. The elasmoid scale can be divided in two types, the amioid scale with radial ridges in the covered part of the scale (Fig. 3A), and the cycloid scale with concentric rings parallel to the margin (Fig. 3B) in the covered part of the scale (Schultze, 1966 and Schultze, 1996). Amioid scales occur in different lineages of sarcopterygians and actinopterygians (Fig. 4; Schultze, 1977 and Schultze, 1996, fig. 3), whereas cycloid scales are restricted to teleosts (= Leptolepis and all more advanced teleosts). Ctenoid scales are derivations from cycloid scales in different lineages of euteleosts (Roberts, 1993).

Distribution of scale morphology and histology: Rhomboid (rhombic) scales Osteichthyes  Cosmoid scales Sarcopterygii   Bony scales Elpistostegalia  Ganoid scales Actinopterygii   Palaeoniscoid scales Palaeoniscimorpha   Lepidosteoid sales Holostei, most Teleosteomorpha   Bony scales/scutes/shields Chondrostei, Aspidorhynchus, Pycnodonti, diverse Teleostei  Elasmoid scales   Amioid scales Actinistia, advanced Dipnoi, Onychodontida Holoptychiida, Tristichopterida   some Palaeoniscimorpha, Amiidae, Caturidae, one teleost (Eurycormus)  Cycloid scales (including ctenoid scales) Teleostei (Leptolepis coryphaenoides + more advanced teleosts)

Enamel: Enamel is a highly mineralized tissue deposited by basal epidermal cells, called ameloblasts. Hexagonal imprints of epidermal cells (e.g., Dipterus: Bemis and Northcutt, 1992, fig. 35; Schultze, 1977) indicate a similar prismatic structure and formation as in enamel of tetrapods. The hexagonal imprints alone do not prove the existence of enamel (Märss, 2006). Enamel is deposited centrifugally to underlying dentine, so that there is a sharp boundary between these two tissues. There is a vast amount of literature to describe the formation and structures of enamel, so that there is no need to discuss it here.

Smith and Hall (1990) interpreted the tissue of the outer layer of tubercles in an Ordovician agnathan as enamel. This is the clear transparent tissue, which Denison (1967, fig. 26) described as durodentine (= enameloid) in scales of a “Vertebrate indet. A”. The dentine tubules reach into the transparent tissue, which does not show the properties of enamel under crossed Nichols.

Friedman and Brazeau (2010) repeated the proposition that conodonts have enamel, whereas it has not only been shown that the tissue in question (lamellar tissue) has not the properties of enamel under crossed Nichols, and also chemical analyses indicate no similarities with enamel (Kemp, 2002, Kemp and Nicoll, 1995, Kemp and Nicoll, 1996 and Trotter et al., 2007; see further discussion in Turner et al., 2010).

Ganoine: Ganoine is a multi-layered enamel (Fig. 1A, B; see Richter and Smith, 1995 for an extensive review). There appears always a sharp boundary to the underlying dentine. Ganoine was described by Williamson (1849) in the extant Lepisosteus and the fossil Lepidotes, where it overlays bone (isopedin) in many layers. This kind of rhomboid scale within actinopterygians was named lepidosteoid scale (Goodrich, 1907, Goodrich, 1909, Gross, 1966 and Schultze, 1966: “Lepidosteus-Typus” after the old term for Lepisosteus; Francillon-Vieillot et al., 1990: lepisosteid scales). In contrast, dentine and a vascular layer are intercalated between ganoine and lamellated bone (= isopedine) in the palaeoniscoid scale (Fig. 1A; Aldinger, 1937, Francillon-Vieillot et al., 1990, Goodrich, 1907, Goodrich, 1909 and Gross, 1966: “Palaeoniscus-Typus”). The canals of the vascular layer house blood vessels (Sewertzoff, 1932), which serve the odontoblasts of the dentine. One can compare these canals with the pulp cavity of a tooth. Sometimes the ganoine of overlying ridges does not superimpose the earlier ganoine layer; it is single layered in such actinopterygian scales. Ørvig (1978b) wrote about “odontodes” in Boreosomus (Ørvig, 1978b, figs. 10–17), Gyrolepis (Ørvig, 1978b, figs. 27, 28), and Plegmolepis (Ørvig, 1978b, figs. 23–26), where single layered ganoine is covered by dentine of the next younger ridge. Multi-layered ganoine occurs in these scales too (Ørvig, 1978b, figs. 25–29). Multilayered ganoine is the common situation in palaeoniscimorph and holostean scales as seen also in Devonian actinopterygians like Moythomasia (Fig. 5; Jessen, 1968, fig. 6) and Orvikuina (Fig. 6; Gross, 1953, figs. 9, 12, 13; Schultze, 1968, fig. 18).

The ganoine surface (Fig. 7; Reissner, 1859: Polypterus, Lepisosteus; Kerr, 1952: Polypterus: very small grooves; Schultze, 1966: Eugnathus; Schultze, 1977: Dapedium; Ørvig, 1967: Colobodus, Nephrotus; Ermin et al., 1971: Polypterus; Gayet and Meunier, 1986: Lepidotes, Polypterus, Lepisosteus, Atractosteus; Meunier et al., 1988: Lepisosteus, Erpetoichthys, Polypterus; Gayet et al., 1988: Lepisosteus, polypterid; Gayet and Meunier, 1992: Polypterus, Dajetella, Lepisosteus, Atractosteus; Daget et al., 2001: Polypterus; Märss, 2006: Dapedium) can be marked by small tubercles, representing the center of the epidermal cells. Reissner (1859) was the first to describe small “hills” on the surface of the ganoine of scales of Lepisosteus and Polypterus, which are separable only microscopically. These small tubercles rediscovered by Schultze (1966), Ørvig (1967) and Ermin et al. (1971) were used by Gayet and Meunier (1986), Meunier et al. (1988), and Meunier and Gayet, 1992 and Meunier and Gayet, 1996 to distinguish species. They are typical for the ganoine surface and indicate a different secretion process of the basal layer of the epidermis in actinopterygians (Sire et al., 1986 and Sire et al., 1987). The tubercles on the surface of ganoine (Fig. 7) have been observed on Mesozoic and Recent holostean and polypterid scales.

A similar surface ornamentation like the small tubercles of actinopterygian ganoine occurs in few acanthodians (Derycke and Chancogne-Weber, 1995 and Märss, 2006: Acanthodes-type). The surfaces of acanthodian scales show either different ornamentation or no structure (smooth), reaching from hexagonal arrangements in Cheiracanthus, more irregular structures to small tubercles (Märss, 2006: fig. 11). It means that one has to be careful when using the surface structure to characterize a special tissue. In contrast to enamel (ganoine or true enamel), the clear outer zone that sometimes occurs around each box in acanthodian scales is not separated from the underlying dentine below, as can be seen in figure 3A, B of Richter and Smith (1995); the dentine tubuli reach to the upper limit of the higher mineralized layer. It may represent enameloid (“durodentine” of different authors; see Ørvig, 1951 and Ørvig, 1967). Märss (2006) showed too that a hexagonal surface structure similar to that of enamel appears on the surface of scales of chondrichthyans, thelodonts and acanthodians, even though these scales possess no enamel but only enameloid.

Sire et al., 1986 and Sire et al., 1987 and Sire, 1994 and Sire, 1995 have shown that ganoine is enamel with successive deposition of layers by the epidermis. In earlier time, two opinions contradicted each other: ganoine was considered either enamel (e.g., Ermin et al., 1971, Hertwig, 1879, Reissner, 1859, Schultze, 1966, Scupin, 1896 and Sewertzoff, 1932) or durodentine, an enameloid (e.g., Gross, 1966, Ørvig, 1967), a highly mineralized dentine. Ørvig (1978a, fig. 5) labels as ‘collariform ganoine’ enamel on the lateral side of teeth of palaeoniscimorph actinopterygians (e.g., Birgeria). That tissue occurs widely in actinopterygians (Thomasset, 1930, “émail du collet,” Schmidt and Keil, 1958, Kragendurodentin; Peyer, 1968, enamel) and shows an outside deposition of enamel (Peyer, 1968, pl. 27, fig. c; Schmidt and Keil, 1958, fig. 137). Birefringence and deposition on the outside is taken as indication of “true” enamel (Peyer, 1968 and Schmidt and Keil, 1958), whereas Ørvig (1978a) identified it as ganoine based on the layered deposition. A study of its growth in extant actinopterygians would be desirable.

In conclusion, ganoine is a special kind of enamel of actinopterygians secreted in a different way than “true” enamel in sarcopterygians. In contrast to “true” enamel, ganoine grows by superposition, so that partial overlap of the enamel surface layer indicates ganoine as seen in Mesozoic forms (Ørvig, 1978b) and in Devonian (Fig. 5 and Fig. 6) and Silurian forms. To my best knowledge the typical surface structure of ganoid scales of Mesozoic and Recent actinopterygians ganoid scales has not been observed in Paleozoic actinopterygians; there is thus a difficulty to identify ganoine in early actinopterygians on the basis of the surface structure, as Friedman and Brazeau (2010) pointed out. Multilayered enamel occurs only in ganoine, even partial overlap of enamel (Fig. 5 and Fig. 6) is found only in ganoine. Thus, ganoine can be defined as multilayered or partially overlapping enamel with a tuberculated surface in Mesozoic and Recent ganoid scales.

Cosmine: Cosmine (Fig. 1 and Fig. 8) is a combination of tissues and one structure: a single thin layer of “true” enamel covers dentine interrupted by pores of the pore–canal system (Megalichthys: Goodrich, 1907, Goodrich, 1909 and Ørvig, 1969, p. 241; Williamson, 1849). One has to use cosmine in this sense and not confuse it with dentine as has been done often in the literature. It does not occur in any recent fish, only in fossil sarcopterygians. The pores open to flask-like cavities, which are connected by a mesh of canals. The mesh canals (“Maschenkanal” of Gross, 1956) form a second canal system besides the dentinal canal system (Fig. 9) as has been shown by Williamson (1849, fig. 17) and Goodrich (1907, fig. 197A; here Fig. 1C). Both systems are connected in some places, generally in the deeper spongious part of the scale.

Williamson (1849) used interchangeable dentine and cosmine. Goodrich, 1907 and Goodrich, 1909 defined the cosmoid scale as built by three layers: a layer of dentine (for him cosmine) + pore–canal system underlain by a vascular layer and isopedin, and covered by a thin layer of enamel. Rauther (1929) and many others followed Goodrich, 1907 and Goodrich, 1909. Gross (1956) dealt extensively with cosmine in different sarcopterygians (Dipterus, Rhinodipterus, Porolepis, osteolepids), and showed differences between taxa and differences to cosmine-like structures in agnathans (Tremataspis). Enamel extends into the pores in dipnoans and porolepiforms (Fig. 9 and Fig. 10), whereas it ends at the border of the pores in osteolepiforms (Fig. 8 and Fig. 10). The pore of the pore–canal system widens to a single flask-like space in dipnoans and porolepiforms (Fig. 10B, C) in contrast to osteolepiforms, where a bony plate divides the flask-like space into an upper and lower division (Fig. 10A).

Because there is no extant fish with cosmine, growth and function of cosmine is difficult to access. Gross (1956, p. 96, fig. 83A) compared cosmine with the “flaschenförmigen Sinnesorganen” (flask-like sensory organs) of the South American lungfish Lepidosiren (Fahrenholz, 1929, fig. 2). Bemis and Northcutt (1992) compared it with blood vessel structures in the snout of the Australian lungfish Neoceratodus forsteri.

In contrast to ganoid scales of actinopterygians, one does not observe any resorption in cosmoid scales of sarcopterygians except in dipnoan scales with Westoll-lines. Therefore, it is difficult to understand the marginal enlargement of the scales. Thomson (1977) argued that the whole surface is resorbed and formed new. However, I checked and re-studied the material described by Thomson (1977) and found but no sign of resorption in the scales of Ectosteorhachis (Thomson, 1977, fig. 2; Thomson, 1975, figs. 2, 15, 23) and Megalichthys (Thomson, 1977, fig. 3). Borgen (1989) showed reabsorbed grooves on osteolepidid jaws, which look like postmortem or even post-fossilization etchings and have nothing to do with the growth process of these fishes.

Zhu et al. (2010) placed Meemannia together with Andreolepis and actinopterygians as a group without resorption at the base of their cladogram. Resorption is common in actinopterygian scales (e.g., Aldinger, 1937, Ørvig, 1977, Ørvig, 1978a and Ørvig, 1978b) even in basal forms like Ligulalepis und Orvikuina (Schultze, 1968, figs. 5, 19). Meemannia shows resorption (see below) like Psarolepis (Zhu et al., 2010, fig. 8: “partial resorption”). Zhu et al. (2010, fig. 8) accepted Thomson's (1977) concept of complete resorption in rhipidistians, for which I cannot find any evidence. Resorption is often difficult to detect, nevertheless it occurs undoubtedly in scales of actinopterygians and may be in scales and bones of Meemannia and Psarolepis.

In the past, cosmine was considered an electrosensory system (Thomson, 1977, fig. 9), but there is no fish with such an extensive distribution of electric organs, therefore Bemis and Northcutt (1992) compared it with the vascular system (vertical capillary loops) in Neoceratodus and found good concordance with the pore–canal system. Campbell et al. (2010) doubted the comparison; they argued that the canals in the snout of Devonian dipnoans are occupied by nerves, whereas the blood vessels are positioned outside the bony canals. They argued that there is no space for a returning vessel in the canals, which cannot be the case in the expanded spaces of the pores and the mesh canals of the pore–canal system. The mesh canals are even wider than the dentinal canals (Fig. 9). Campbell et al. (2010) were mostly concerned with the vascularisation of the snout, whereas Bemis and Northcutt (1992) dealt with the smaller units of dermal-epidermal interface (dermal papillae, ampullary organs, etc.). Gross (1965) described the distribution of pores in Devonian dipnoans and distinguished between pores of the pore–canal system and larger pores, which reach to a deeper system; the pores get larger towards the front and have a larger distance between each other. The same situation can be seen in figures of Campbell et al. (2010, figs. 2, 11A–D, 29), where the fine pores of the pore–canal system are gone. Thus the interpretation by Bemis and Northcutt (1992) refers to the pore–canal system not to the vascular system of the snout in general and appears to be the most likely explanation for the pore–canal system.

Canal system: The canal system in ganoid scales, which serves the dentine, has been intensively studied and figured since Williamson (1849, pl. 40 fig. 7, pl. 41, fig. 8: Palaeoniscum). It has been described and figured for the extant Polypterus by Reissner (1859), Hertwig (1879, pl. 3, fig. 5, bone or dentine around canals to ganoine surface), Goodrich (1907), Sewertzoff (1932), Kerr (1952, and fig. 11), and for the extant Erpetoichthys [Calamoichthys] by Rauther (1929, fig. 136). Goodrich (1907: Eurynotus, Gonatodus, Cheirolepis), Aldinger (1937: Cheirolepis, Boreolepis, Palaeoniscum, Plegmolepis, Acropholis, Glaucolepis, Ganolepis, Gonatodus, Elonichthys, Platysomus, Scanilepis), Wilson (1953: Lawnia), Cavender (1963: Elonichthys, Gonatodus, Pteronisculus, Holurus, Rhadinichthys, Eurynothus, Eurylepidoides, Haplolepis, Cryphiolepis, Watsonichthys, Acrolepis, “Aetheretmon-type,’ Amblypterus, Lawnia, Brachydegma, Aeduella, Cheirolepis) and Ørvig (1978a: Gyrolepis, Reticolepis, Ganolepis, Ptycholepis, Eurynotus, Aetheretmon, Canobius) showed it in scales of many lower actinopterygians. Aldinger (1937) dealt extensively with the horizontal canal system, he distinguished longitudinal, cross, and radial canals. The canal system has connections to the surface, the margin and to the inner side of the scale; on the surface the canals open with pores between ganoine ridges (e.g., Gayet and Meunier, 1992; Dajetella) but also on the ganoine surface (Fig. 11; see also Goodrich, 1907, fig. 198B; here Fig. 1A). The canals can be wide or narrow, and this changes from genus to genus. Considering the early actinopterygians, one finds wide canals in Cheirolepis (Aldinger, 1937, fig. 51A–D) and in Ligulalepis (Schultze, 1968, fig. 4), but also in the stem osteichthyan Lophosteus (Gross, 1969, fig. 11). Narrow dentinal canals are more common in palaeoniscimorphs (Aldinger, 1937, many figs.) and are present in scales of Dialipina, Orvikuina (Schultze, 1968, figs. 11, 17) and Andreolepis (Gross, 1968, fig. 12A). They are commonly placed below sculptures as pulpa canals of the dentine (Fig. 12: l.ca). The canals to the surface branch off dorsolaterally (always very clear in cross sections). The dentinal canal system is variable (Fig. 12), and it nourishes the dentine. It seems to be the primitive state in osteichthyans.

Gross (1956) dealt specifically with the canal systems in cosmoid scales and put special emphasis on the canal systems. There is the dentinal system, which forms the pulpa canals below the dentine (Fig. 10: pul.k changing into bl.k), and in addition the pore–canal system with its mesh canals (Fig. 10: mk; Gross, 1956, “Maschenkanal”). Both systems form horizontal canals.

Acrodin: Actinopterygian teeth have a tip of highly mineralized dentine named acrodin by Ørvig (1973). Acrodin is a unique feature of actinopterygians accepted by all, including Friedman and Brazeau (2010). Not all actinopterygian possess teeth with acrodin. It is missing, e.g., in teeth of Cheirolepis, a taxon, which possesses multi-layered ganoine in its scales. On the other side, Ligulalepis, placed by Friedman and Brazeau (2010) outside the actinopterygians, possesses acrodin (Fig. 13).

Dentine: The tissue termed dentine in this paper is orthodentine in the sense of Ørvig, 1951 and Ørvig, 1967. The hard tissue encompasses the parallel running dentinal tubules, which branch distally, whereas the cell body, the odontoblast, remains in the pulp cavity or canal.

Enamel is a synapomorphy of osteichthyans. It does not occur in acanthodians (contrary to published accounts: Richter and Smith, 1995 and Richter et al., 1999). It occurs in two forms, the multilayered ganoine on scales and bones of actinopterygians (even in those with a tuberculated surface) and the single layered “true” enamel (Francillon-Vieillot et al., 1990 and Peyer, 1968) on bones and scales of sarcopterygians. Lateral superposition (partial overlap) of enamel is considered here as indication of ganoine, because it occurs frequently in actinopterygians, and not in sarcopterygians.

Ganoine is limited to actinopterygians (Goodrich, 1909, p. 217: “the true ganoid scale”). Usually multilayered enamel is accepted as ganoine. The lateral overlapping enamel in tubercles and ridges occurs in rhomboid scales of Paleozoic (e.g., Aldinger, 1937, figs. 7, 90, pl. 18, fig. 1) and Mesozoic (e.g., Ørvig, 1978b, figs. 10–12, 18, 19, 23, 24) actinopterygians. It represents the first step towards multilayered ganoine. This feature can be seen, sometimes, even in scales of the extant Polypterus (Poole, 1967, figs. 15, 16). Friedman and Brazeau (2010) accept the presence of ganoine in acanthodians, as published by Richter and Smith (1995), Richter et al. (1999), and Derycke and Chancogne-Weber (1995). A thin surface layer covers the subsequent growth addition in some acanthodians (“Acanthodes-Typ” of Gross, 1973, with enameloid, ‘Durodentin’). However the layer is not separated from the underlying dentine. This is an enameloid. The surface micro-ornament of acanthodian scales is variable, from hexagonal to tuberculated or smooth (Märss, 2006), and gives no indication of enamel or ganoine.

The canal system in ganoid scales serves the dentine (emphasized by Ørvig, 1978b). That can be compared with the pulp cavity of teeth. The canal system in ganoid scales and its variations were intensively described by Aldinger (1937). Cavender (1963) did a doctoral dissertation on the histology of palaeoniscoid scales and showed variation in the distribution of the canal system similar to the results of Aldinger (1937). Cavender (1963, pl. 5, fig. A = Haplolepis, B = Cryphiolepis, and F = “Aetheretmon-type”) showed that lateral pores extend from the main canals and open in the bony furrows between the ganoine ridges [also described and figured by Aldinger, 1937: fig. 13 ‘breiter Kanal in einer Kosmin (= dentine) lamelle erster Art’]. Gross has shown that the pores open lateral to the sculpture ridges in Lophosteus (Gross, 1969, figs. 1A, 11; Schultze and Märss, 2004, figs. 4, 5) and Andreolepis (Gross, 1968, figs. 1–6, 12A; Janvier, 1978, pl. 2, figs. 1, 6, 7, 10, 12; also Märss, 2001, pl. 1, figs. 4, 10, 11, 13, pl. 2, 5). That is the common situation in actinopterygians, as shown for Cheirolepis (Aldinger, 1937, fig. 50A, 51; p. 201: no canals to scale surface). Long and Trinajstic (2000) showed the canal openings in grooves between ridges (pl. 3, figs. 1, 3) and pores in a closed ganoine surface (pl. 3 figs. 2, 4) of scales of Moythomasia.

Cosmine is limited to sarcopterygians (Megalichthys: Goodrich, 1909, p. 217: “the cosmoid scale”). In the cephalaspid Tremataspis a distantly comparable pore–canal system lies in the bony layer (Gross, 1956, figs. 84, 88, 89, 93; Schultze, 1969, fig. 42a), a sieve plate extends along the whole canal system and a superficial enamel layer is missing; mesodentine forms the superficial layer. Gross (1956) described a pore–canal system in Poracanthus and Radioporacanthodes (Gross, 1956, fig. 99, 104, and arrangement of canals in figs. 105–114, pl. 14, fig. 6, pl. 15, pl. 16, figs. 1, 2). It is a canal system with longitudinal or circular (corresponding to growth additions) canals, which open vertically on the surface by many pores. The canal system is independent from the dentine like pore–canal system in cosmine, only its complexity of narrow and wide longitudinal canals (Gross, 1956, figs. 99, 104) and in some cases antler-like branching of the pores into many small pores (Gross, 1956, figs. 114A, B, 118) is very unlike cosmine.

Scales of the extant polypterids show pores on the ganoine and bone surface, which are connected to a dentinal vascular system (e.g., Kerr, 1952 and Rauther, 1929) and do not represent an independent pore–canal system like in cosmine (Fig. 11). The vascular canals as pulp canals are shown in vertical sections (e.g., Goodrich, 1909, fig. 265; Rauther, 1929, fig. 137).

Cosmine has been described and figured for the osteolepidids Megalichthys (Goodrich, 1907, Goodrich, 1909, Gross, 1956 and Williamson, 1849), Osteolepis (Gross, 1966) and Ectosteorhachis (Thomson, 1975), the porolepidid Porolepis (Gross, 1956 and Gross, 1966), the dipnoan Rhinodipterus, Dipterus (Bemis and Northcutt, 1992 and Gross, 1956), Ganorhynchus (Gross, 1956 and Gross, 1965) and Chirodipterus (Bemis and Northcutt, 1992) and the basal sarcopterygian Styloichthys (Zhu et al., 2006). Cosmine has not been recorded in the sarcopterygian onychodonts and actinistians, where only elasmoid scales are known. This may be an indication that cosmine occurs only above onychodonts and actinistians in sarcopterygians: Styloichthys, Dipnomorpha and Tetrapodomorpha (Long et al., 2014 and Lu and Zhu, 2008).

Teeth of actinopterygians possess a tip of highly mineralized dentine named acrodin by Ørvig (1973). Friedman and Brazeau (2010) accepted acrodin as a feature of actinopterygians, even though some actinopterygians do not possess it. Nevertheless Ligulalepis, which possesses acrodin (Fig. 13), is placed by the two authors outside the actinopterygians.

Acrodin has not been described from teeth of Dialipina and Andreolepis. Lophosteus does not possess acrodin. It is lacking in Cheirolepis as far as known, but known in other Palaeozoic actinopterygians.

Occurrence of tissues: Enamel Osteichthyes  Ganoine Actinopterygii  “true” enamel Sarcopterygii Acrodin Actinopterygii Dentine (Orthodentine) Gnathostomata above Placodermi and some ‘Agnatha’ Cosmine Sarcopterygii (except Actinistia and Onychodontida)

Lophosteus: Gross, 1969 and Gross, 1971 described in detail the morphology and histology of the scales of the genus, which was first published by Pander (1856). In contrast to Friedman and Brazeau's (2010, p. 50) statement, scales of Lophosteus have a peg-and-socket articulation combined with a keel on the inner side of the scales; a socket for the peg is present ventral to the keel (Fig. 14; Gross, 1971, fig. 1A, B and Gross, 1969, fig. 1). The peg is broad-based as on scales of sarcopterygians, therefore Schultze (1977, fig. 1) indicated the possibility of placing the genus with the sarcopterygians. The peg-and-socket articulation places Lophosteus within osteichthyans. The posterior margin of the scales is straight even though the free surface is covered with ridges, which can reach posteriad above the posterior margin. In cross sections, scales of Lophosteus show a sequence of each other overlying units of dentine without enamel (Gross, 1969, figs. 12A, B, 13C, 14A, 15E). In osteichthyans, the overlying units possess enamel, therefore Lophosteus has to be placed below the branching point of actinopterygians and sarcopterygians.

Lophosteus possesses broad sensory grooves (Gross, 1969, figs. 5H, 6D, 7 A, B; Märss, 1986, pl. 35, figs. 6, 7), whereas the lateral line system is enclosed in canals in osteichthyans. The lateral line system lies in open grooves in acanthodians, chondrichthyans and most placoderms. Thus this primitive feature places Lophosteus below the branching point of actinopterygians and sarcopterygians. The preserved piece of a maxilla described by Botella et al. (2007) shows similarities to that of onychodonts, an indication that the genus is close to osteichthyans, whereas the fin spines look like those of acanthodians (Otto, 1991) and thus are a primitive feature compared to osteichthyans. Burrow (1995b) compared the scale ornamentation with that of placoderms, a feature, which was considered superficial by Märss (2001). A detailed discussion of the systematic position of Lophosteus can be found in Schultze and Märss (2004).

In conclusion, morphology and histology of the scales of Lophosteus look like scales that one would expect from a stem osteichthyan. In agreement, Lophosteus appears in most recent phylogenetic analyses (e.g., Dupret et al., 2014, Long et al., 2014, Zhu et al., 2013 and Zhu et al., 2009; Fig. 15) as stem osteichthyan.

Andreolepis: Andreolepis was first described by Gross (1968) and placed within actinopterygians. That placement was accepted by Schultze, 1977 and Schultze, 1992 and Janvier (1978). Janvier (1996) and Schultze and Märss (2004) placed Andreolepis as sister group of all other actinopterygians, whereas Botella et al. (2007), Zhu et al., 2010 and Zhu et al., 2009, and Chen et al. (2012) placed the genus as basal osteichthyan (Fig. 15A). An extended anterodorsal process and ganoine (= overlapping enamel) places the genus at the base of the actinopterygians (Schultze and Märss, 2004). The broad based peg of Lophosteus may represent the primitive osteichthyan condition. The posterior margin of the scales is straight, except in the anterior deeper flank scales (‘Type 1’ in Chen et al., 2012). The histology of the scales is similar to that of scales of the actinopterygian Moythomasia; the histology together with the extended anterodorsal process points towards a placement of the genus within the actinopterygians.

Dialipina: Scales of Dialipina salgueiroensis were first described by Schultze (1968), and additional material by Schultze (1992) and Schultze and Cumbaa (2001). Schultze, 1968 and Schultze, 1992, Schultze and Cumbaa (2001), Zhu et al. (2009), Janvier (2007) and Long et al. (2014) placed Dialipina as basal actinopterygian (Fig. 15A, C). Friedman and Brazeau (2010) and Dupret et al. (2014) considered Dialipina as stem osteichthyan (Fig. 15B). The shape of the scales (narrow dorsal peg, anterodorsal extended corner, serrated posterior margin), histology (ridges with overlapping enamel = ganoine) and the bone arrangement on the skull roof indicate a placement within actinopterygians. The presence of two dorsal fins is a primitive feature; the triphycercal tail is an autapomorphy occurring independently in similar fashion within different groups of sarcopterygians. A short-based peg, an extended anterodorsal process and ganoine (= overlapping enamel) place the genus within actinopterygians above Andreolepis (Schultze and Märss, 2004). The lateral line runs inside the dermal bones, as in all osteichthyans, and in contrast to Lophosteus.

Ligulalepis: Ligulalepis was first described by Schultze (1968) and placed within actinopterygians. This was followed by Wang and Dong (1989: Silurian Ligulalepis from China), Burrow (1994), Janvier (1996), Märss (2001) and others. Zhu et al. (2006) placed Ligulalepis with Dialipina as stem actinopterygian or as “higher” actinopterygian with Osorioichthys, whereas Zhu et al. (2009) considered Ligulalepis as a stem sarcopterygian together with the clade (Meemannia (Guiyu (Psarolepis + Achoania))) (Fig. 15A). Friedman and Brazeau (2010) and Dupret et al. (2014) considered the genus to be a stem osteichthyan (Fig. 15B). In contrast, Ligulalepis appears with Osorioichthys within the actinopterygians in Long et al. (2014), one option) already proposed in Zhu et al. (2006).

The scales of Ligulalepis possess a narrow based peg, a well developed anterodorsal process, ganoine (= overlapping enamel) forming ridges, deep scales otherwise known only in actinopterygians and a serrated posterior border. Sarcopterygian scales possess a straight posterior margin with one exception (osteolepidid indet., Young and Schultze, 2005). The best indication for a placement of Ligulalepis within actinopterygians is the presence of acrodin in the teeth (Fig. 13).

Naxilepis: Wang and Dong (1989) described actinopterygian scales from the Silurian of China. The low and elongated scales occur usually in the ventral region of the body. The anterodorsal process is strongly extended, and the posterior margin serrated. Wang and Dong (1989) compared the scales with those of Orvikuina. Naxilepis appears in the phylogenetic analysis of Schultze and Märss (2004) within the actinopterygians crownward to Andreolepis. Lu and Zhu (2008) considered the genus as a basal osteichthyan. Strongly extended anterodorsal process and ganoine (= overlapping to multilayered enamel) point to a placement within actinopterygians.

Terenolepis: Scales of this genus were described by Burrow (1995a) as actinopterygian scales from the Lower Devonian of Australia. Terenolepis appears in the phylogenetic analysis of Schultze and Märss (2004) within the actinopterygians above Andreolepis, only the inclusion of the genus results in an unresolved interrelationship within actinopterygians. None of the scales figured by Burrow (1995a) seems to be complete so that it is not clear how strongly the anterodorsal process is developed. There is an internal keel (Burrow, 1995a, pl. 1, fig. 3 and fig. 3d, f), but peg and socket are not preserved. The posterior margin may be serrated (Burrow, 1995a, pl. 1, figs. 3, 5, and fig. 2b). Ganoine is sporadically developed; it is described as single layered Burrow (1995a, fig. 2b, explanation as fig. 3b), and it shows overlaying enamel layers. Despite insufficient information, the interpretation of the genus as an actinopterygian seems the only acceptable one.

Orvikuina: Gross (1953) described ventral scales of an actinopterygian from the Middle Devonian of Estonia. The scales possess an anterodorsal extended corner, but a dorsal peg is not formed as is typical for elongated ventral scales. A keel is present on the inner side. Schultze (1968) described shorter and deeper scales of Orvikuina from the Middle Devonian of Spitsbergen. They show a small peg and a shallow socket ventral of the keel (Schultze, 1968, fig. 13, pl. 3, fig. 7a, b). The dorsal surface is covered with long ridges formed by dentine and covered by enamel. The enamel of the ridges overlaps laterally (Gross, 1953, figs. 9A, B, 13; Schultze, 1968, fig. 16) and sometimes for a longer distance (Gross, 1953, fig. 12; Schultze, 1968, fig. 17), that is a feature of ganoine and not found in “true” enamel of sarcopterygians. The scales have all characters of an actinopterygian, so that it is difficult to understand why Friedman and Brazeau (2010) placed the genus as stem osteichthyan. Orvikuina appears in the phylogenetic analysis of Schultze and Märss (2004) within the actinopterygians above Andreolepis based on the presence of an anterodorsal process and ganoine (= overlapping and multilayered enamel; Gross, 1953, figs. 9, 12, 13, pl. 7 figs. 3, 4, Ørvig, 1957, fig. 3A, B; Schultze, 1968, figs. 16, 18).

In recent years a number of basal osteichthyans have been described from the Upper Silurian and Lower Devonian of China. Their position within osteichthyans has varied much; they have been placed with the Sarcopterygii initially. I will discuss here the position mainly based on scale characters.

Guiyu: Zhu et al. (2009) described the Silurian osteichthyan, Guiyu, which they placed with Ligulalepis and Meemannia as stem sarcopterygian [Ligulalepis (Meemannia (Guiyu (Psarolepis + Achoania) (Sarcopterygii)))] (Fig. 15A), while they stated it displays a mixture of actinopterygian and sarcopterygian features. Qiao and Zhu (2010) and Dupret et al. (2014) placed the clade (Guiyu (Psarolepis + Achoania)) as basal sarcopterygians (Fig. 15B), whereas Long et al. (2014) moved the three genera to a basal actinopterygian position (Fig. 15C).

Guiyu has the snout of an actinopterygian with ganoine ridges (Qiao and Zhu, 2010, fig. 4a, b), which expose the bone sutures. The premaxillaries meet in the middle in front of the rostral as in Mimipiscis and Moythomasia. The position of both nares between the bones of the snout compares with Cheirolepis and Paratarrasius. Guiyu possesses many branchiostegals and a large preoperculum, as in actinopterygians and in contrast to sarcopterygians. The similarities in the cheek region between Guiyu and actinopterygians are interpreted as primitive osteichthyan characters by Zhu et al. (2009, p. 473). In contrast, the skull roof shows an intracranial joint between parietal and postparietal (Qiao and Zhu, 2010, fig. 1). Taking the scales (Zhu et al., 2009: fig. 4i and suppl. info. figs. 1, 9b) into consideration, these show only actinopterygian features with ridges of ganoine (Qiao and Zhu, 2010, p. 1837; Zhu et al., 2009, p. 473), an anterodorsal process, a narrow based peg, and a serrated posterior margin. The double ridged keel is reminiscent of that in scales of Ligulalepis. The ridges are covered by “a glossy outer dermal coating of single or multilayer enamel” (Qiao and Zhu, 2010, p. 1837), typical for ganoine of actinopterygians.

Achoania: Zhu et al. (2001) and Zhu and Yu (2002) placed Achoania close to Psarolepis as stem sarcopterygian (Fig. 15A). Long et al. (2014) moved it to a basal actinopterygian position together with Psarolepis and Guiyu (Fig. 15C). Only the anterior cranial portion is known from the Lochkovian of China, but no scales. The snout is very similar to that of Psarolepis (Qiao and Zhu, 2010, figs. 4c, d, 9).

Psarolepis: Yu (1998) described the new genus Psarolepis from the Silurian and Lower Devonian of China as porolepiform sarcopterygian. In Zhu et al. (1999), and Zhu and Schultze (2001), the taxon appears as a stem osteichthyan, but in Zhu et al., 2010 and Zhu et al., 2006 as a stem sarcopterygian, Zhu et al. (2012) and Qu et al. (2013) offer both variants, either as stem osteichthyan or stem sarcopterygian. In Long et al. (2014), Psarolepis appears with Achoania at the base of the Actinopterygii (Fig. 15C): clade (Guiyu (Achoania + Psarolepis)) (all other actinopterygians).

Qu et al. (2013) described the histology of the scales of Psarolepis romeri; they argued that Psarolepis possesses cosmine with a developing pore–canal system. The surface of the scales shows large pores arranged in longitudinal rows (Qu et al., 2013, figs. 2A–G, 3A, 4A, B) as it can be found in actinopterygians, a very similar case are the scales of the Carboniferous Acrolepis wilsoni (Fig. 16; Traquair, 1909, pl. 25, figs. 12, 13). The surface of the scales of Psarolepis romeri does not show the fine pore system of cosmoid scales. The sections (Qu et al., 2013, figs. 5A–D, 7A, B, 8A, D, E, 10A) show a canal system with few large openings to the surface. There is only one canal system, which serves the dentine. The presence of one canal system only is especially clear in the two horizontal sections (Qu et al., 2013, fig. 9A, B). Only longitudinal canals are visible with a vertical connection near the posterior border and radial canals near the lower (ventral) margin. That is the typical situation in actinopterygian scales as shown many times by Aldinger (1937, see above) and is figured here (Fig. 5) for Moythomasia, where the dentine tubuli are visible to demonstrate the connection of dentine with the longitudinal canals. A second canal system, a mesh canal system as it appears in cosmine (see Fig. 9), is missing completely in the scales of Psarolepis. The arrangement of enamel corresponds to that of actinopterygians also, overlapping and even multilayered ganoine is present (Qu et al., 2013, fig. 7E). The sequence of additional growth (Qu et al., 2013, figs. 5B, D, 6A, 7B, F) looks like that in actinopterygian scales (Fig. 5 and Fig. 6). Zhu et al. (2006, fig. 3a) considered such growth correctly as typical for actinopterygians. Thus, Qu et al. (2013, fig. 12) compare the scales in morphology and histology – in my opinion correctly – with those of Ligulalepis and Andreolepis, two basal actinopterygians. All scale characters point towards a position as sister group to Achoania and Guiyu and all other actinopterygians (Long et al., 2014).

Meemannia: Zhu et al. (2006) placed the new taxon, Meemannia, at the base of the Sarcopterygii despite similarities in the arrangement of skull roofing bones with the pattern in actinopterygians (Zhu et al., 2006, compare fig. 1c with fig. 1d [Dialipina], g [Cheirolepis]) and only a cosmine-like dermal surface. Zhu et al. (2010) and Janvier (2007) placed the genus as stem sarcopterygian below Psarolepis (Fig. 15A). In contrast, Meemannia is placed with Ligulalepis and Osorioichthys in higher actinopterygians in Long et al. (2014; Fig. 15C).

Zhu et al. (2006, fig. 2c) give a free reconstruction of a pore–canal system in Meemannia. The sections (Zhu et al., 2006, fig. 2a, b) show sequential overgrowth (“odontodes”) of dentine + single enamel layers, with partially overlapping enamel like ganoine in actinopterygian scales, and no pore–canal system. Scales are unknown for the genus, nevertheless skull roof pattern and histology point to a placement close to actinopterygians.

Of the four genera, Guiyu, Psarolepis and Achoania are placed in close relationship to each other in recent phylogenies, whereas the position of Meemannia is uncertain. Meemannia may be an actinopterygian, but the position of Guiyu, Psarolepis and Achoania is more uncertain. The characters discussed here, seem to support the placement of Long et al. (2014). Nevertheless Guiyu, Psarolepis and Achoania possess an intracranial joint considered a feature of sarcopterygians (lost in dipnoans and tetrapods), but also discussed as a primitive structure; an old controversy.

Arquatichthys: Lu and Zhu (2008) described a new Early Devonian taxon, Arquatichthys, based on a lower jaw and detached scales. The scales possess an anteroventral process, which occurs also in Powichthys and Youngolepis after Lu and Zhu (2008). The authors placed the genus between Dipnoi and Porolepiformes. The presence of cosmine, a broad based peg and a straight posterior margin support a placement within sarcopterygians.

Styloichthys: Styloichthys was first described by Zhu and Yu (2002); they placed the genus within the sarcopterygians as sister taxon to Porolepiformes, Dipnoi and Tetrapodomorpha. The scales of the genus show typical cosmine with a single enamel layer and pore–canal system (Zhu and Yu, 2002, supplementary information); the enamel layer reaches into the pores (Zhu et al., 2006). These characters place the genus clearly within sarcopterygians supporting studies by Zhu and Yu (2002), Zhu et al. (2006), Lu and Zhu (2008), Long et al. (2014) and Giles et al. (2015). The scales of Styloichthys and Youngolepis possess an anterodorsal process (Lu and Zhu, 2008), an appearance of the character in parallel to that in actinopterygians.

Placement in cladogram (Fig. 17A, B; for character list, data matrix and apomorphy list see supplementary data): Schultze and Märss (2004) presented an analysis based mainly on scale characters. Using that analysis with addition of the four new taxa Styloichthys, Arquatichthys, Meemannia, and Guiyu and two additional characters, there appears a clear division between actinopterygians and sarcopterygians. Osteichthyes possess two common characters, the lateral line canal within dermal bone and no paired fin spines (with a reversal in Guiyu). The actinopterygians are characterized by overlapping enamel, here accepted as indication of ganoine. Multilayered ganoine occurs only in two taxa included in this analysis, Cheirolepis and Moythomasia. “True” enamel (= single layered enamel) occurs only together with cosmine in sarcopterygians. Guiyu, Psarolepis and Meemannia appear within the actinopterygians together with basal actinopterygians, Terenolepis, Orvikuina and Naxilepis similar to Long et al. (2014). The sister group of these basal actinopterygians is the clade (Moythomasia (Ligulalepis + Dialipina)) and all higher actinopterygians. The synapomorphy for sarcopterygians is the broad-based peg and cosmine with the occurrence of single layered enamel, “true” enamel. The sarcopterygians are artificially separated into forms with rounded scales (including Achoania) and those with rhombic scales. Lophosteus appears outside the osteichthyans together with Climatius. The division into the two taxa Actinopterygii and Sarcopterygii is well supported and is in congruence with Long et al. (2014), nevertheless the distribution of taxa within each group should not be taken too seriously because the number of characters is too small for the number of taxa.

The phylogenetic analysis of Long et al. (2014) is based on the data set of Dupret et al. (2014), which is based on earlier data sets like that of Zhu et al. (2013) and others. Nevertheless the tree topologies of Long et al. (2014) are quite different from earlier published ones based on the same data sets (and different from Giles et al., 2015). Long et al. (2014) reanalyzed the data set of Dupret et al. (2014) and retrieved the same tree topologies as for their data set enlarged by seven taxa (total of 85 taxa). It is interesting that the analysis of the much smaller data set of scale characters with few taxa (27) results in a similar tree topology like that of Long et al. (2014).