January 2, 2007

A New Actinopterygian From the Famennian of East Greenland and the Interrelationships of Devonian Ray-Finned Fishes

By Friedman, Matt; Blom, Henning


A new actinopterygian, Cuneognathus gardineri new genus and species, is described from the Devonian (Famennian) Obrutschew Bjerg Formation of East Greenland on the basis of multiple incomplete specimens. Cuneognathus most closely resembles Limnomis from the Famennian Catskill Formation of Pennsylvania, and, like that taxon, is known exclusively from freshwater deposits. A cladistic analysis with an ingroup of 13 actinopterygians and an outgroup of five sarcopterygians explores the relationships between the new genus and some of its better-known Devonian contemporaries, and recovers the same four topologies regardless of the implementation of limited character ordering. Cheirolepis is resolved as the most basal of well-known Devonian actinopterygians, consistent with a majority of previous studies. A novel sister-group relationship between Howqualepis and Tegeolepis is found in all trees. Disagreement between the most parsimonious cladograms is concentrated in a clade whose members are often informally referred to as 'stegotrachelids.' Cuneognathus and Limnomis are resolved as sister taxa within this large radiation along with the pairings of Moythomasia dugaringa plus M. nitida and Krasnoyarichthys plus Stegotrachelus. The arrangement of taxa is conserved when the enigmatic Dialipina is added to the analysis, although the reconstructed position of that genus above both Cheirolepis and Osorioichthys seems improbable. Our scheme of relationships suggests that actinopterygians invaded freshwater environments at least four times during the Devonian, while age constraints indicate that many of the cladogenic events between ingroup taxa included in this study occurred during or before the Givetian.


ARTICULATED DEVONIAN actitiopterygians (ray-finned fishes) are relatively uncommon; only 10 genera have been described on the basis of such material. The oldest fossils attributed to actinopterygians are isolated scales and bone fragments assigned to Andreolepis Gross, 1968 (Janvier, 1971, 1978; Mrss, 2001) and Naxilepis Wang and Dong. 1989, which ostensibly extend the record of the group into the Silurian (but see Discussion). The incompletely described Dialipinu Schultze. 1968 (Schultze, 1992; Schultze and Cumbaa, 2001) has been advanced as the oldest articulated actinopterygian (Schultze and Cumbaa, 2001), but the position of this Early Devonian (Emsian) genus among early osteichthyans remains unclear (Zhu and Schultze, 2001, p. 296; Cloutier and Arratia, 2004, p. 247) and empirical support for its actinopterygian affinities is equivocal. Cheirolepis Agassiz, 1835. the first Devonian actinopterygian to be described, is the oldest uncontested ray-fin represented by articulated remains. Known primarily from the Eifelian Old Red Sandstone of Scotland and the Frasnian lagersttte of Miguasha, Canada (Arratia and CIoutier. 1996), remains of this genus have also been reported from Nevada (Reed, 1992; Arratia and CIoutier, 2004) and Germany (Gross, 1973). Studies of exceptionally preserved specimens of Mimia toombsi Gardiner and Bartram, 1977 and Moythomasia durgaringa Gardiner and Bartram, 1977 from the Frasnian Gogo Formation, Western Australia, have made these species the most thoroughly documented Devonian actinopterygians (Gardiner, 1984). Articulated material of Moyilwmasia Gross, 1950 is also known from the latest Givetian- earliest Frasnian of Bergisch-Gladbach, Germany (Jessen. 1968).

Several additional Devonian actinopterygians are known in less detail. Apart from Cheirolepis and Moythomasia, the Devonian deposits of Europe have yielded articulated material of two additional genera: Osorioichthys marginis (Casier, 1952) from the Famennian of Belgium and Stegotracheliisfinlayi Woodward and White, 1926 from the Givetian of Shetland Island, Scotland. The single incomplete specimen of Kmsnoyarichthyx jesseni Prokofiev, 2002 from the Famennian of western Siberia is the sole occurrence of articulated Devonian actinopterygian material in Asia. Howqualepis rostridens Long, 1988 is known from the Givetian of Mt. Howitt, central Victoria. Australia, while undescribed material from the Aztec Siltstone of southern Victoria Land, Antarctica (Young, 1989, 1991), marks an additional find in Gondwana.

Late Devonian deposits of the USA yield a morphologically diverse array of actinopterygians. The partially articulated remains of at least two species are known from the Frasnian (Rotondo and Over, 2000; contra Gardiner, 1993, p. 612) Rhinestreet Shale of western New York, and have been referred to the genus Moythomusia by Gardiner (1963). Tegeolepis clarki (Newberry, 1888) and Kentuckiu hlavini Dunkle, 1964 have been described from the Famennian Cleveland Member of the Ohio Shale (Gardiner, 1963; Dunkle and Schaeffer, 1973), while maxillae reminiscent of those of the Carboniferous genus Canobius Traquair, 1881 are found in the same deposits (M. I. Coates, personal commun., 2004). Two forms are known from the Famennian of Pennsylvania: Limnomis delanevi Daeschler, 2000 from the Catskill Formation and a single articulated specimen from the marine Venago Formation reported by Eastman (1907), which does not appear assignable to any established taxon and may represent a new genus.

Here we describe a new genus and species of actinopterygian from the Upper Devonian (Famennian) of East Greenland. This marks the first ray-finned fish from these deposits, which otherwise yield a diverse assemblage of gnathostomes. including placoderms, sarcopterygian fishes, and the early tetrapods Acanthostegci Jarvik, 1952 and Ichthyoxteaa Sa've-Soderbergh, 1932 (Jarvik, 1961; Bendix- Ahlmgreen. 1976). The synthesis of information from numerous incomplete specimens permits description of much of the morphology of the new taxon. These data are incorporated into a cladistic analysis that examines the interrelationships of Devonian actinopterygians. The relationships of the extant cladistians Polypterus Lacepede, 1803 and Erpetoichthys Smith, 1865 to these Paleozoic forms (Gardiner, 1984; Gardiner and Schaeffer, 1989; Lund et al., 1995; Taverne, 1996, 1997; Coates, 1999; Schultze and Cumbaa, 2001; Zhu and Schultze, 2001; Lund and Poplin, 2002; Cloutier and Arratia, 2004) are beyond the scope of our study, and we therefore do not address the limits of crown-group Actinopterygii and total-group Actinopteri in our analysis. Instead, our proposed hypothesis of interrelationships is used to examine the environmental context of early actinopterygian evolution. Abbreviations for the collections housing material discussed in this paper are: Cleveland Museum of Natural History. Cleveland (CMNH) and Geological Museum, University of Copenhagen, Copenhagen, Denmark (MGUH).


Devonian sedimentary rocks in East Greenland are exposed along a nearly 100 km wide belt, stretching approximately 200 km from Hudson Land in the north to Traill in the south (Fig. 1). The entire succession is over 8 km thick and was deposited in a Middle to Late Devonian continental basin that was characterized by a broad range of depositional environments (Olsen, 1993; Olsen and Larsen, 1993). Fossil vertebrates are known throughout much of the succession, but are particularly common in its uppermost strata, which are Famennian in age (Jarvik, 1961; Bendix-Ahlmgreen, 1976).

The actinopterygian fossils described here were collected in 1954 by the sedimentologist H. Butler from "a dark calcareous shale with fish and plant remains at about 1,200 m on the western ridge on the south side of Celsius Bjerg" (letter to E. Jarvik, dated 1954; translated from German), a description consistent with the lithology and location of the Obrutschew Bjerg Formation (Nicholson and Friend, 1976; Olsen and Larsen, 1993). This formation is 4-6 m thick and consists of black shales in association with dark limestones, grey mudstones, and red sandstones. The depositional environment has been interpreted as lacustrine characterized by anaerobic conditions (Olsen. 1993; Olsen and Larsen, 1993). The Obrutschew Bjerg Formation represents the uppermost depositional complex of the Celsius Bjerg Group, which is dominated by a combination of floodplain siltstones and shales and point bar sandstones. While the other formations of the Celsius Bjerg Group are rich in fossil vertebrates (Jarvik, 1961: BendixAhlmgreen, 1976). fossils have rarely been reported from the Obrutschew Bjerg Formation.

The paucity of fossils from the Obrutschew Bjerg Formation has led to uncertainty about its age (Olsen, 1993: Olsen and Larsen, 1993). However, recent palynological analyses of the Obrutschew Bjerg Formation on Stensi Bjerg, Gauss Halv0. reveal a major turnover in the spore faunas within the unit, demonstrating that this formation straddles the Devonian-Carboniferous boundary (J. E. A. Marshall, personal commun., 2003). Between the lower and upper portions of the formation, which yield diagnostically Devonian and Carboniferous palynomorph assemblages, respectively, is an interval rich in amorphous organic matter (AOM) but lacking spores (Fig. 1). Analyses of rock matrix containing the actinopterygian specimens reported here yield only AOM and no palynomorphs, indicatin\g that these fossils originate from the barren portion of the formation (J. E. A. Marshall, personal commun., 2003). The palynomorph Reuspora lepidophyta (Kedo, 1957) Playford, 1976. the last appearance of which marks the end of the Devonian, is found on both sides of this AOMrich zone, indicating that this interval and the actinopterygian fossils which it contains are latest Famenniun in age.


Chemical preparation was carried out on one specimen (MGUH VP 27638) using weak (ca. 10%) hydrochloric acid (HCl) to remove weathered bone. Acid-degraded bone was prepared away under a binocular microscope using a mounted entomological needle in order to ensure that delicate details were left intact on the negative. In some cases, specimens were dusted with a sublimate of ammonium chloride (NH^sub 4^CI) in order to enhance contrast for photography.

Dermal bone nomenclature follows Gardiner (1984) and Gardiner and Schaeffer (1989). Although we accept that actinopterygian frontals and parietals are the homologues of sarcopterygian parietals and postparietals, respectively (Schultze. 1993), we nevertheless utilize the traditional actinopterygian nomenclature for ease of comparison with existing literature.


Class OSTEICHTHYES Huxley, 1880

Subclass ACTINOITKRYGII Woodward, 1891


Genus CUNEOGNATHUS new genus

Type species.-Cuneognathus gardineri n. sp.

Diagnosis.-As that of the species.

Etymology.-Latin cuneus, meaning wedge, and New Latin gnathus, derived from Greek gnathos, meaning jaw. The combination of these roots refers to the wedge-shaped ganoine ridges that ornament the lower jaw of the new taxon.


Figures 2-7

Palaeonisciformes JARVIK, 1961, p. 199, table 1.

Palaeonisciformes gen. and sp. indet. BENDIX-AHLMGREEN, 1976, p. 542.

Diagnosis.-Rostral ornamented with transverse ganoine ridges; frontals twice as long as parietals; parietals subquadrate; suture between intertemporal and supratemporal anterior to that between frontals and parietals: dermosphenotic contacts frontal, excluding intertemporal from contact with nasal; maxilla ornamented with vertical ganoine ridges, with a largely unornamented postorbital expansion; dermohyal with conspicuous anterior boss; caudal fin strongly asymmetrical; fringing fulcra on leading margin of dorsal fin and caudal tin (pectoral and anal tins insufficiently known); pelvic tins very small or absent; ca. 40 scale rows anterior to the caudal inversion; dorsal articular peg absent from scales; ridge coursing along anterior margin of individual scales that truncates horizontal ornament ridges; at least two basal fulcra preceding the hypochordal lobe of the caudal fin; elongated caudal fulcra, with largest being half the length of hypochordal lobe of the caudal fin; keel scales absent along ventral midline of abdomen.

Etymology.-Specific name in honor of Professor Brian G. Gardiner, in recognition of his considerable contributions to the study of Paleozoic actinopterygians.

Types.-The material described here is deposited in the Geological Museum, Copenhagen, and includes a skull preserved in left-lateral view (MGUH VP 27637; Fig. 2) that is designated as the holotype. Additional specimens include a dorsoventrally compressed skull with scattered trunk scales and bones of the pectoral girdle (MGUH VP 27638; Fig. 3), a nearly complete postcranium in part and counterpart that preserves median fins and an external mold of the dermal pectoral girdle (MGUH VP 27639; Fig. 4), and a caudal tin in part and counterpart (MGUH VP 27640; Fig. 5), and are designated as paratypes. All of these specimens share an identical scale ornament, and can therefore be assigned to the same taxon with confidence. Additional material from the same locality consist of isolated, fragmentary scales and dermal bones that may be attributable to the new actinopterygian, but these are not described here.

Occurrence.-From the Upper Devonian (latest Famennian) Obrutschew Bjerg Formation at an elevation of 1,200 m on the south face of Celsius Bjerg, East Greenland.


Dermal bones of the snout.-The rostral is the largest dermal ossification contributing to the snout of Cuneognathus n. gen., and is clearly visible in both specimens that preserve portions of the skull (Figs. 2, 3). Unlike that of Cheirolepis (Pearson and Westoll, 1979; Arratia and Cloutier, 1996), but similar to those of other Devonian actinopterygians. the rostral of Cuneognathus is a large median bone that contacts the premaxillae ventrally, the nasals laterally, and the frontals posteriorly. No specimen preserves the ventral border of the rostral. However, since the bone expands below the anterior nostrils (Fig. 3). it is possible that it contributed to the margin of the upper jaw as in Howqualepis Long, 1988 (Long, 1988), Limnomis Daeschler, 2000 (Daeschler, 2000), and Moythomasia durgaringa (Gardiner, 1984). A series of transverse ridges of ganoine ornaments the rostral. This distinctive pattern is limited to Cuneognathus and Limnoinis (Daeschler, 2000) among Devonian actinopterygians, but is found in many Carboniferous taxa (e.g., Lund and Poplin, 1997). Dorsally the transverse ornamentation breaks down into a series of abbreviated anteroposteriorly oriented ridges, while it grades ventrally into large, irregular tubercles.

A portion of premaxilla is preserved on the right side of one specimen (Fig. 3). Its ornament consists of irregular pustules similar to those found on the adjoining nasal. As in most Devonian actinopterygians, with the exception of Osorioichthys Casier, 1954 (Taverne, 1997) and some specimens of Mimia Gardiner and Bartram, 1977 (Gardiner, 1984), the premaxilla does not appear to contribute to the margin of the anterior nostril.

The nasal is narrow and extends along the lateral margin of the rostral, defining the anterodorsal margin of the orbit (Figs. 2, 3). It is ornamented with a series of discontinuous ganoine ridges that run parallel to the long axis of the bone. The nasal extends beyond the posterior margin of the rostral, contacting the frontal mesially and the dermosphenotic posteriorly.

Dermal hones of the cheek.-The maxilla is the largest bone contributing to the cheek of Cuneognathus and is similar in shape to that of most early actinopterygians (Janvier, 1996), with a low infraorbital ramus and a deep postorbital expansion. The infraorbital projection and posterior expansion are of roughly equal length, similar to the condition in most Devonian actinopterygians. However, these proportions are unlike those of Cheirolepis (Pearson and Westoll, 1979; Arratia and Cloutier, 1996), Tegeolepis Miller, 1892 (Dunkle and Schaeffer, 1973). and Howqualepis (Long, 1988), in which an elongated postorbital expansion reflects a relatively more oblique Suspensorium (Gardiner and Schaeffer, 1989). Anteriorly inclined to subvertical ridges of ganoine ornament both the anterior ramus and the posterior expansion of the maxilla. This pattern is unusual among Devonian actinopterygians, in which the ornament typically consists of horizontal ridges or isolated tubercles (Gardiner, 1984). The ridges break down into small pustules toward the jaw margin, while the dorsalmost portions of the postorbital expansion are largely unornamented.

The maxilla is bounded on its posterior and dorsal margins by the preopercle (Figs. 2, 3), which consists of a deep anterodorsal limb and a nearly vertical stem. The surface of the preopercle is not well preserved, although it is clear that the ventral stem is ornamented with vertical ridges similar to those on the maxilla, while the anterodorsal expansion of the bone is smooth. A small bone, interpreted as the quadratojugal, is visible in one specimen posterior to the preopercular stem (Fig. 3). This teardrop-shaped ossification is ornamented with a series of divergent ganoine ridges. The oblong dermohyal is located just above the dorsal margin of the preopercle. It is covered with a series of longitudinal ridges, and its anterior angle is marked by a large, striated boss.

The jugal is a relatively large, crescentic bone that contacts the anterodorsal margin of the maxilla (Fig. 3). It is unclear if this bone contacted the preopercle as well, or if there were additional ossifications between these two elements as in Osorioichthys (Taverne, 1997), Moythomasia nitida Gross, 1953 (Jessen, 1968), Kentuckia hlavini (CMNH 8060; Hansen, 1996: fig. 21- 8.2) and many post-Devonian forms (Gardiner and Schaeffer, 1989). As with the other major dermal bones of the cheek, the jugal is ornamented with vertical ridges of gunoine. The lacrimal and the sclerotic ring are not preserved in any of the available material, and there is no evidence for any supraorbitals.

Dermal bones of the skull roof.-The frontals are the largest bones of the dermal skull roof in Cuneognathus (Figs. 2, 3). These elongate ossifications are ornamented with longitudinally oriented and concentric ridges of gunoine. A pineal foramen appears to be present (Fig. 2). The division between the frontuls and the parietals is irregular, and is marked by a region of isolated tubercles. The parietals are subquadrate in shape and less than half the length of the frontuls (Figs. 2, 3). This appears to be a derived condition among Devonian actinopterygians (Schulue, 1992: Schnitze and Cumbaa, 2001). The ornamentation of the parietal is composed of anteroposteriorly oriented ridges of gunoine. which is interrupted by well-defined anterior, middle, and posterior pit lines (Fig. 3). The posterior margin of the parietal consists of a smooth, shelflike flange for the overlap of the extrascapular series.

Flanking the median series of paired rooting bones are the intertemporuls and supratemporals (Figs. 2, 3). The intertemporal is roughly triangular in shape, with its anterior vertex wedged between the frontal and the dermosphenotic. \It does not contact the nasal. This condition is plesiomorphic for actinopterygians (Gardiner and Schaeffer, 1989), and is primitively shared with all other Devonian forms with the exception of Kentuckia hlavini (Dunkle, 1964), Stegotrachelus Woodward and White, 1926 (Gardiner, 1963), and Moythomasia durgaringa (Gardiner, 1984), which exhibit the derived contact between the nasal and the intertemporal. The supratemporal is a four-sided bone that extends beyond the posterior margin of the parietals. The division between the intertemporal and supratemporal lies anterior to the transverse suture between the frontuls and parietals. The orientation of ornament differs between the two bones; the ridges on the intertemporal are transverse, while those on the supratemporal are anteroposteriorly directed. Only fragments of the dermosphenotic are preserved (Figs. 2, 3). This ossification clearly contacts the nasal anteriorly, but its other margins cannot be discerned. The dermosphenotic is ornamented with a series of longitudinal ridges. This region is too poorly preserved to comment on the relationships of these bones to the spiracular opening.

A fragment of an extrascapular, ornamented with irregular pustules of gunoine, is preserved in one specimen (Fig. 3). The width of this incomplete bone is nearly that of the parietal, strongly suggesting that only a single pair of extrascapulars were present in Cuneognathus.

Lower jaw.-Individual bones of the lower jaw are indistinct, but it is reasonable to assume that the dentary is the dominant dermal ossification of the mandible as in other actinopterygians. Ornament consists of a series of nested chevron-shaped ganoine ridges with their apices pointed posteriorly. Dorsally and posteriorly, this ornamentation breaks down into isolated pustules. Ornament appears to be absent from the region of the lower jaw overlapped by the posteroventral extremity of the maxilla. Large, widely spaced primary teeth are separated by deep alveoli and are flanked laterally by closely spaced, smaller laniaries on the dorsal margin of the dentary. It is unclear if the teeth are topped with acrodin caps.

Operculogular series.-The preserved portions of the operculogular series show the conventional configuration for early actinopterygians (Figs. 2, 3; Gardiner, 1984). The large opercle resembles a parallelogram with rounded corners. Ornament consists of ridges oriented subparallel to the long axis of the bone, with isolated pustules located dorsally in one specimen (Fig. 3). A smooth, unornamented strip runs along the anterior margin of the opercle (Fig. 2). The subopercle lies ventral to the opercle, and is approximately square in outline. There is no evidence of an additional ossification lying between the dorsalmost members of the operculogular series (opercle and subopercle) and the preopercle. This bone, which has variously been referred to as an accessory operculum (Gardiner, 1963: Pearson and Westoll, 1979; Arratia and Cloutier, 2004), epipreopercle (Prokofiev, 2002), or extension of the dermohyal (Gardiner and Schaeffer, 1989), is sporadically distributed among Devonian actinopterygians, having been identified in Krasnoyarichthys Prokofiev, 2002 (Prokofiev, 2002). Cheirolepis (Pearson and Westoll, 1979; Arratia and Cloutier, 1996), and Moythomasia nitida (Jessen, 1968).

Pectoral girdle and fin.-The rhomboidal posttemporal (Figs. 2, 3) contacts the posterior of the dermal skull roof anteriorly and overlaps the rounded dorsal margin of the supracleithrum posteriorly (Fig. 3). The supracleithrum is an elongate bone that tapers ventrally (Figs. 2-4). The small presupracleithrum (Fig. 3) lies in the window framed by the posttemporal, supracleithrum, and opercle. All of these bones are ornamented with line, vermiform ridges of ganoine.

The cleithrum and clavicle are preserved only as poorly defined impressions on one specimen (Fig. 4), and the precise division between these two ossifications cannot be determined. The unit formed by these two bones is crescentic in shape, with acutely pointed dorsal and anterior apices. The posterior margin of the cleithrum is deeply excavated, and would have accommodated the pectoral fin in life. There are no remains of the pectoral fin itself in any of the specimens. Neither the interclavicle nor postcleithrum is preserved, but their presence cannot be ruled out.

Pelvic girdle and fin.-Although one specimen preserves most of the ventral surface of the body, there is no indication of the pelvic fins or their internal skeleton (Fig. 4). This is unusual among Devonian actinopterygians; Cheirolepis (Pearson and Westoll, 1979) and Howqualepis (Long, 1988) have pelvic fins with exceptionally long insertions, while most other forms have well- developed pelvics with relatively shorter bases. Limnomis, however, appears to lack pelvic fins entirely (Daeschler, 2000), and in this respect is similar to Cuneognathus.

Median fins.-The single dorsal fin is located on the posterior half of the body (Fig. 4) and is anteriorly shifted relative to the anal fin. This most closely resembles the condition reconstructed for Limnomis (Daeschler, 2000) and Krasnoyarichthys (Prokofiev, 2002). and Kentuckia hlavini (CMNH 8061; Hansen, 1996: fig. 21- 8.1). Fringing fulcra are present along the anterior margin of the dorsal fin, but the anal fin is too poorly preserved to determine if these structures were present there as well. The state of preservation also precludes precise counts of the number of lepidotrichia contributing to these fins.

The caudal fin is deeply notched and is composed of approximately 60 segmented, distally branched lepidotrichia (Fig. 5). Fringing fulcra are present on the anterior margin of the hypochordal lobe. The chordal lobe is considerably longer than the hypochordal lobe, giving the tail a strongly heterocercal profile.

Squamation.-As in other Devonian actinopterygians, with the exception of Cheirolepis (Pearson and Westoll, 1979) and Tegeolepis (Dunkle and Schaeffer, 1973). the squamation of Cuneognathus is macromeric. Scale rows anterior to the caudal inversion are anteriorly inclined. The rhombic flank scales are squat, unlike the vertically elongated scales of Mimia (Gardiner. 1984; Trinajstic, 1999) and Krasnoyarichthys (Prokofiev, 2002). Well-developed articular pegs have not been observed on scales located on any part of the body of any specimen, and are assumed to be absent. Two or three subparallel ridges of ganoine trace the anterior and ventral margins of each flank scale (Fig. 6). This ornament truncates the anterior extremities of wedge-shaped ganoine ridges that ornament the body of the scale. The posterior margins of the trunk scales are serrated, forming approximately six distinct, caudally oriented points. The overall morphology of the flank scales is closest to Limnomis (Daeschler, 2000) among Devonian forms, but broadly similar patterns of ornamentation are also found in Dialipina (Schultze, 1968) as well as isolated scales reported from Iowa (Storrs, 1987) and Siberia (Schultze, 1992).

There are approximately 40 scale rows between the skull and the level of the caudal inversion (Fig. 4); this figure agrees well with that for Stegotrachelus (ca. 48; Woodward and White, 1926). Limnomis (40-45: Daeschler. 2000), and the species of Moythomasia (44-52: Jessen, 1968; Gardiner, 1984). Scales posterior to the caudal inversion are diamond-shaped, and are relatively shallow in comparison to more anteriorly located flank scales. Ornament on scales in this region consists of a few longitudinal ridges of ganoine. The poor preservation of available material precludes a thorough investigation of scale histology.

At least 14 fulcra extend along the dorsal margin of the caudal fin, forming a rigid cutwater (Figs. 4, 5). Each of these elements is rounded anteriorly where it is overlapped by its more anterior fellow, and tapers to a narrow point posteriorly. The ornament of the fulcra is distinctive, with a series of low ridges projecting outwards from a median, triangular eminence. This raised area bears a shallow depression on its midline that is overlapped by the posterior extremity of the preceding fulcrum. The fulcra are of considerable size, with the largest being approximately half the length of the hypochordal lobe of the caudal fin. Anteriorly, the fulcra appear to shorten, and continue forward as ridge scales. It is unclear if these ridge scales reached the base of the dorsal fin, or if they terminated far posterior to it as in Cheirolepis (Pearson and Westoll, 1979), Limnomiis (Daeschler, 2000), and numerous post- Devonian genera (e.g., Traquair, 1877-1914). It is not known if Cuneognathus possessed ridge scales anterior to the dorsal fin. The remains of at least two basal fulcra precede the base of the hypochordal lobe of the caudal fin, but it is impossible to determine if this series of ridge scales extended further anteriorly due to the incompleteness of the specimen. An extensive row of ventral ridge scales anterior to the anal fin of the kind seen in Stegotrachelus (Woodward and White. 1926) and Krasnoyarichthys (Prokofiev, 2002) is absent in Cuneognathus (Fig. 4). However, the presence of a limited set of basal fulcra immediately anterior to the insertion of the anal fin cannot be ruled out.

Body form and reconstruction.-Although no single fossil of Cuneognathus is complete, its morphology is sufficiently well represented in available specimens to permit a preliminary reconstruction (Fig. 7). The relative proportions of Cuneognathus most closely resemble those of the Carboniferous Melanecta Coates, 1998 and, to a lesser extent, the Devonian Limnomis. The dorsoventral compression of the postcranial specimen (Fig. 4) is characteristic of fishes with a subcircular cross section (Weigelt, 1989), and agrees with the inferred body form of both Melanecta and Limnomis. The postcranial specimen of Cuneognathus (F\ig. 4) indicates a fish on the order of 4 cm in total length, while the large cranial specimen (Fig. 3) extrapolates to an individual of perhaps three times that size. Assuming this estimate is reliable, it places Cuneognathus in a similar size class with better known taxa such as Moythomasia and Mimia (Gardiner, 1984). With the meter- long Tegeolepis (Dunkle and Schaeffer, 1973) at one end of the spectrum, and the diminutive Limnomis at only a few centimeters in length at the other (Daeschler, 2000), Devonian actinopterygians span over an order-of-magnitude range in size, highlighting the early ecological proliferation of this group.


Although early actinopterygian phylogeny has been the subject of a series of studies (Patterson, 1982; Gardiner, 1984; Gardiner and Schaeffer, 1989; Coates, 1993, 1998, 1999; Lund et al., 1995; Lund and Poplin, 2002; Cloutier and Arratia, 2004), much of this research has focused on the placement of Paleozoic taxa relative to extant clades (Cladistia, Chondrostei, Neopterygii) and has therefore only superficially addressed Devonian forms. The analyses of Taverne (1997) and Schnitze and Cumbaa (2001) have been exceptions to this trend, and have emphasized the interrelationships of Devonian actinopterygians. Although there are some areas of agreement between their phylogenies, such as the basal or near-basal position of Cheirolepis, a consensus remains elusive. Studies with broader taxonomic and character samples also fail to retrieve consistent patterns of interrelationships. Cloutier and Arratia (2004) have conducted the most extensive cladistic survey of early actinopterygians, which includes a broad sample of Devonian forms. Their results are sensitive to changes in taxon and character sets, and their published cladograms are characterized by considerable discord. Consistent signals that were recovered by their study represent a notable departure from earlier hypotheses of actinopterygian interrelationships. Particularly remarkable is that all but one of their trees that incorporate extant taxa fail to place any Paleozoic form within the actinopterygian crown, a result that is strongly at odds with previous analyses (Patterson, 1982; Gardiner, 1984; Gardiner and Schaeffer, 1989; Coates, 1993, 1998, 1999). There is still conflict among their hypotheses even when only the relationships among Devonian taxa are considered. Apart from the monophyly of Cheirolepis, the only relationship conserved across all trees is a sister-group pairing between Moythomasia and Mimia to the exclusion of all other Devonian actinopterygians. This marks an additional departure from many earlier cladograms, where Mimia and Moythomasia are resolved as successive or near-successive plesions along the actinopteran stem, with the latter genus considered to be more closely related to Stegotrachelus and Kentuckia (Gardiner, 1984; Gardiner and Schaeffer, 1989; Taverne, 1997; Coates, 1999). Support for the sister-group relationship between Mimia and Moythomasia is weak, and many of the characters that ostensibly underpin this clade are problematic. Some are miscoded (character 43 for both genera, each of which has a single nasal on either side of the head; character 123, which is unknown in Moythomasia nitida but scored at odds with the observed condition in M. durgaringa; Jessen, 1968; Gardiner, 1984), while others make troublingly liberal assessments of homology (character 10, which equates the moderately pointed apex of the platelike premaxillae of Mimia and Moythomasia with the complicated nasal processes of neopterygians and ascending processes of teleosts; Patterson, 1973) or are so vaguely formulated (character 166, which fails to specify how pelvic fins with "short" and "reduced" insertion bases are distinguished from one another) as to be of questionable systematic value.

Given the uncertainty surrounding early actinopterygian phylogeny, reconsideration of the relationships among Devonian forms is necessary in order to establish the systematic placement of Cuneognathus n. gen. Our intention is neither to test current hypotheses of osteichthyan interrelationships nor to infer the primitive suite of actinopterygian characters. Rather, we seek to highlight the relationships among the well-circumscribed set of articulated Devonian actinopterygians, with the central goal of placing Cuneognathus within a phylogenetic framework that relates it to its contemporaries. We are somewhat hesitant to include Dialipina (Schultze, 1968) and the material subsequently attributed to it (Schultze, 1992; Schultze and Cumbaa, 2001) until they are more fully described, despite assertions by Schultze and Cumbaa (2001) that this genus represents the most primitive actinopterygian known from articulated material. Although Dialipina may indeed be an early actinopterygian, we find the current evidence for this interpretation equivocal.

Schultze and Cumbaa (2001) have claimed their cladistic analysis demonstrates that Dialipina occupies a basal position within Actinopterygii. However, we are not convinced by this conclusion as a consequence of the methodological approaches used to reach it. Acknowledging that placement of Dialipina within an all- actinopterygian ingroup would have represented an a priori phylogenetic assessment and thus fail to test the status of this genus as an actinopterygian. Schultze and Cumbaa (2001) chose not to specify an outgroup in their preliminary analyses. They then proceeded to root their tree such that a monophyletic Sarcopterygii fell as sister to the clade Dialipina plus Actinopterygii. justifying this procedure by noting that "the sarcopterygians clustered together; therefore they were placed in the outgroup to find the arrangement of Dialipina within the actinopterygians" (Schultze and Cumbaa, 2001, p. 321). However, there are three possible resolutions of this unrooted network that preserve the monophyly of Sarcopterygii as classically defined: Dialipina as the most basal actinopterygian (the hypothesis advocated by Schultze and Cumbaa, 2001), Dialipina as the most basal sarcopterygian, and Dialipina as sister to the clade Actinopterygii plus Sarcopterygii (thus making this genus a stem osteichthyan). Schultze and Cumbaa (2001) do not defend their rooting by reference to outgroups to Osteichthyes, but instead cite a series of similarities-not necessarily synapomorphies-linking Dialipina with actinopterygians. Outgroup comparison with acanthodians (Denison, 1979; Richter and Smith, 1995; Richter et al., 1999, p. 739-741) strongly suggests that two of these characters (long-based pelvic fins, as opposed to lobate pelvic fins of sarcopterygians; presence of ganoine) are osteichthyan symplesiomorphies. The polarity of two additional characters (the presence of a dermosphenotic: presence of a consolidated nasal) cannot be logically assessed due to the lack of a macromeric dermatocranium in the most proximal outgroups to Osteichthyes: Acanthodii and Chondrichthyes. Similar difficulties are encountered in determining the polarity of two scale characters (narrow dorsal peg; anterodorsal process) due to the micromeric squamation of acanthodians and chondrichthyans.

Regardless of its precise position relative to the two osteichthyan crown-clades, it nevertheless seems clear that Dialipina occupies a very deep position within osteichthyan phylogeny. An unusual mosaic of characters otherwise typically associated with either actinopterygians (e.g., scales with a narrow dorsal peg) or sarcopterygians (e.g., triphycercal tail), but not both, would suggest that Dialipina is perhaps more proximal to the last common ancestor of osteichthyans than any other known well- known genus with the possible exceptions of Psarolepis Yu, 1998 and AMF101607 (referred to the genus Ligulalepis Schultze, 1968 by Basden and Young, 2001), both of which are known primarily from neurocranial remains. In order to test the position of Dialipina in osteichthyan phylogeny, outgroup comparison with nonosteichthyans is necessary. Only two studies have taken this approach, and both have yielded ambiguous results. Zhu and Schultze (2001) failed to resolve the relationship between Dialipina, Sarcopterygii, and Actinopterygii. Cloutier and Arratia (2004, p. 247) recovered Dialipina as a stem sarcopterygian when polarizing their analysis of actinopterygian relationships with sarcopterygian and acanthodian outgroups. Additional trees presented by these authors that reconstruct Dialipina as sister to all other actinopterygians do not use outgroup comparison with non-osteichthyans, thus failing to provide a critical test of the position of Dialipina among early bony fishes (cf. Schultze and Cumbaa, 2001).

Given the uncertainties regarding the phylogenetic position of Dialipina, we have decided to carry out our analysis in two parts. The first of these excludes Dialipina, while the second includes Dialipina to assess what impact, if any, it has on hypotheses of interrelationships among unequivocal actinopterygians. We have chosen to polarize our characters through outgroup comparison with members of each of the major sarcopterygian radiations: Glyptolepis Agassiz, 1844 (porolepiforms); Miguashaia Schultze, 1973 (coelacanths); Osteolepis Agassiz, 1835 (tetrapodomorphs); Strunius Jessen, 1966 (onychodonts): and Uranolophus Denison, 1968 (lungfishes). The relationships among the major sarcopterygian clades were left free to vary, thereby accommodating a wide range of conflicting phylogenetic hypotheses (Cloutier and Ahlberg, 1996; Chang and Yu, 1997; Zhu and Schultze, 2001). As our study does not include non-osteichthyans, it is impossible for us to address the position of Dialipina within Osteichthyes.

Phylogenetic analysis excluding Dialipina Schultze, 1968.-There are four most parsimonious trees (L = 112; CI = 0.5357; RI = 0.7189; RCI = 0.3857) for the data set when Dialipina is excluded and selectcharacters are ordered along morphoclines (see Appendix). Identical topologies are found when no character ordering is implemented, although tree statistics vary slightly (L = 110; CI = 0.5364; RI = 0.7167; RCI = 0.3844). Unless stated otherwise, all support metrics that follow refer to the solution incorporating character ordering. All ingroup nodes are completely resolved, with the exception of a polytomy between a series of Middle to Late Devonian forms, many of which have been referred to Stegotrachelidae Gardiner, 1963 (Gardiner, 1984, 1993) or Moythomasiidae Kazantseva, 1971 (Prokofiev, 2002). Both Adams and majority-rule consensus trees place Kentuckia as sister to Limnomis plus Cuneognathus n. gen., and leave this clade in an unresolved polytomy with Moythomusia and Krasnoyarichthys plus Stegotrachelus (Fig. 8).

Cheirolepis occupies a basal position among actinoplerygians, consistent with the findings of most previous studies, cladistic and otherwise (Gardiner. 1967; Pearson and Westoll, 1979; Patterson, 1982; Pearson, 1982; Gardiner, 1984; Gardiner and Schaeffer, 1989; Taverne, 1997; Coates, 1999; Cloutier and Arratia, 2004; but see Lund et al., 1995 and Schultze and Cumbaa, 2001). The monophyly of Cheirolepis is well supported, with a bootstrap value of 98% and a Bremer decay index of five. The branch leading to Cheirolepis is comparatively long, with five unambiguous character changes; it seems that the isolated use of this apomorphic genus for inferring morphological aspects of the ancestral osteichthyan or actinopterygian may be misleading (contra Pearson, 1982). The monophyly of Devonian actinopterygians above Cheirolepis is well supported, with a bootstrap value of 88% and a Bremer decay index of four. Several characters unambiguously unite this clade to the exclusion of Cheirolepis, and include the consolidation of the nasal and median rostral.

Oxorioichthys lies above Cheirolepis, as found by Taverne (1997) and Coates (1999), and is the only ingroup taxon apart from Cheirolepis that retains pectoral fins that insert into fleshy basal lobes and a posterior nostril that lacks complete communication with the orbital fenestra. This genus also lacks a presupracleithrum, as do C. canadensis Whiteaves, 1881 (Arratia and Cloutier, 1996) and all outgroup taxa. Pearson and Westoll (1979, p. 366, fig. 11c) reported that the presupracleithrum ('postspiracular' in their terminology) is variably present in C. trailli Agassiz 1835, but it is unclear if this ossification represents a discrete bone or is merely a broken extension of the supracleithrum in the specimen they figure.

A novel clade uniting Tegeolepis and Howqualepis is resolved above Osorioichthys. Tegeolepis is bracketed by genera with macromeric scales, suggesting that micromeric squamation in this genus was derived independently of that in Cheirolepis. This inference is corroborated by inconsistencies in scale morphology between these two genera. While the scales of Cheirolepis superficially resemble those of acanthodians and lack both an anterodorsal process and an articular peg, those of Tegeolepis have both of these features (Dunkle and Schaeffer, 1973; CMNH 5518) and essentially resemble miniaturized versions of the scales of other early actinopterygians.

Although a close relationship between Tegeolepis and Howquailepis has not been entertained previously, our results reveal a series of characters relating to the lower jaw and pectoral fin that link these two genera. Uniquely among Devonian actinopterygians, all of the pectoral lepidotrichia of Howqualepis and Tegeolepis are unsegmented along much of their length (Dunkle and Schaeffer, 1973; Long, 1988). The pectoral fin is similarly constructed in the Carboniferous genera Huanghelepis Lu, 2002 and Melanecta, the latter of which Schultze and Cumbaa (2001) recovered as the sister taxon of Howqualepis. Young's (1989) suggestion that the undescribed Aztec actinopterygian may be related to Howqualepis seems unlikely in light of these findings, as the Antarctic form has extensively segmented pectoral lepidotrichia (Young, 1989: fig. 3; confirmed in photographs provided to the senior author by M. I. Coates).

Perhaps the most compelling series of synapomorphies linking Tegeolepis and Howqualepis relate to the structure of the lower jaw. In both genera, the anterior portion of the dentary is reflexed, and bears a series of greatly enlarged teeth. Unlike the condition in early sarcopterygians, where similar tooth whorls lie on separate plates (Jarvik, 1980; Zhu and Yu, 2004), the parasymphysial teeth of Howqualepis and Tegeolepis are located on the dentary itself. Our preferred hypothesis indicates that it is most parsimonious to regard this dental pattern as neomorphic rather than a modification of the arrangement found in sarcopterygians, which would require independent losses in the lineages leading to Cheirolepix and Osorioichthys. Absence of the mentomeckelian ossification also appears to unite Tegeolepis and Howqualepis, but the state of this character is unknown in both Cheirolepis and Osorioichthys; interpretation of this character as a synapomorphy must be taken cautiously. Apart from these synapomorphies, Tegeolepis and Howqualepis share additional characters of uncertain polarity which we have not included in our analysis. The exceptional size of these genera (Tegeolepis, 1 m: Howqualepis, 50 cm) is particularly noteworthy, but the relatively large size of Osorioichthys and Cheirolepis suggests that this may be primitive for actinopterygians. An additional similarity relates to the gross morphology of the rostrum, which is conspicuously elongated in Tegeolepis. Long (1988, p. 11) noted the pointed snout of Howqualepis and drew comparisons with Tegeolepis, but dismissed these similarities as insignificant due to the different arrangement of the median rostral and premaxillae of these genera.

The recognition of mandibular characters shared by Tegeolepis and Howqualepis elucidates the possible systematic placement of a large actinopterygian dentary fragment from the Famennian of Turkey (Janvier et al., 1984: pl. 2.5; Lelivre et al., 1994: fig. 7.7). This specimen not only is characterized by a pattern of ornamentation similar to that of Howqualepis and Tegeolepis, but also lacks a mentomeckelian ossification and has a reflexed distal tip. Unfortunately, it is impossible to determine the state of the parasymphysial dentition from the published figures. The attribution of this jaw to Howqualepis by Lelivre et al. (1994) seems unlikely due to the large stratigraphic gap (Givetian to Famennian) between it and the only other known occurrence of the genus. However, our results suggest the spirit of this identification is essentially correct, and that the specimen can be assigned to a Givetian- Famennian clade that includes Tegeolepis in addition to Howqualepis.

A large radiation whose members are often informally referred to as 'stegotrachelids' or 'moythomasiids' is reconstructed as sister to the clade Tegeolepis plus Howqualepis. Apart from changes in a number of discrete morphological features, this node is also characterized by an apparent shift in general habitus from large and elongate forms such as Cheirolepis and Howqualepis to smaller, relatively deep-bodied taxa.

Mimia is the most basal member of this radiation of small-bodied taxa. This arrangement is consistent with that found in most previous studies, in which this genus falls outside the least- inclusive clade containing both Kentuckia Rayner, 1951 and Moythomasia (Gardiner, 1984; Gardiner and Schaeffer, 1989; Taverne, 1997; Coates, 1999). Recovery of this topology is particularly notable in light of the restricted set of characters we have incorporated in our study. In order to include as broad a sample of Devonian actinopterygians as possible, our data matrix necessarily focuses upon details of the dermal skeleton, while the characters that have been cited as linking Moythomasia and Kentuckia to the exclusion of Mimia are largely neurocranial. Our analysis reveals two additional, unambiguous synapomorphies related to the dermal skeleton for the clade above Mimia: a premaxilla that does not contribute to the margin of the posterior nostril and a reduction in the number of scale rows. The reliability of the second character is unclear; it is homoplastic in our study, and the phylogenetic significance of scale counts has been called into question by Coates (1998, p. 50). Subquadrate parietals may also link these taxa, but optimization of this character is ambiguous in two of the four most parsimonious trees. An additional dermal character, contact between the nasal and intertemporal, has been proposed as a feature uniting Kentuckia and Moythomasia (Gardiner and Schaeffer, 1989; Coates, 1999), but this character is variable among the nominal species of both genera and our study cannot reliably identify it as synapomorphic for the clade above Mimia.

Due to its exclusively Devonian scope, our analysis cannot address the neurocranial characters that are claimed to place both Moythomasia and Kentuckia relatively more proximal to the actinopteran erown than other Devonian actinopterygians. The braincase is unknown in the Devonian Kentuckia hlavini, and character coding for this genus has typically relied on the type species, the early Carboniferous K. deani (Eastman, 1908). Kentuckia is a heterogeneous genus whose two nominal species differ substantially in both cranial dermal bone patterns (dermosphenotic does not contact nasal in K. deani, contact present in K. hlavini; suture between intertemporal and supratemporal at level of suture between frontals and parietals in K. deani, well posterior to that level in K. hlavini; intertemporal shorter than supratemporal in K. deani, the reverse in K. hlavini; Rayner, 1951; Dunkle, 1964) and scale ornament (concentric ganoine ridges in K. deani, lacking orn\ament ridges in K. hlavini; Rayner, 1951; Dunkle, 1964). In light of these morphological inconsistencies and the absence of any uniquely derived characters defining the genus, we see no reason to regard Kentuckia as monophyletic. Although K. deani is known almost exclusively from cranial remains, K. hlavini is represented by multiple complete specimens, as well as many isolated dermal bones (including skull roofs, maxillae, lower jaws, and parasphenoids). The latter species is in need of reinvestigation and, if necessary, formal removal from the genus Kentuckia.

The cladogram is incompletely resolved above Mimia. The sister- taxon pairs Stegotrachelus plus Krasnoyarichthys, Moythomasia durgaringa plus M. nitida, and Cuneognathus plus Limnomis are placed in a polytomy with Kentuckia in the strict consensus topology. Majority-rule and Adams consensus trees are slightly more resolved, placing Kentuckia as the immediate outgroup to the pairing of Cuneognathus and Limnomis (found in 75% of trees; Fig. 8). The node subtending Cuneognathus and Limnomis is the best supported in this radiation, with a bootstrap value of 97% and a decay index of three. This clade is underpinned by a series of unambiguous synapomorphies relating to ornamentation of the scales and median rostral, as well as loss of the pelvic fins. The link between Cuneognathus and Limnomis is intriguing from both a biogeographical and environmental standpoint; these two genera are found in the freshwater Famennian deposits of the Old Red Sandstone continent, and are associated with broadly similar vertebrate assemblages (Jarvik, 1961; Bendix- Ahlmgreen, 1976; Daeschler et al., 2003). The position of Stegotrachelus in our preferred tree draws many of the cladogenic events between ingroup taxa into the Givetian or earlier, suggesting that the Middle Devonian might have been an important interval in actinopterygian diversification (Fig. 9).

Phylogenetic analysis including Dialipina Schultze, 1968.- Dialipina was included in our data matrix to examine what effect it might have on our hypothesis of relationships. Our inclusion of this genus is not intended as a test of its phylogenetic position, but rather as an assessment of the robustness of our previous results to the inclusion of this taxon. Incorporation of Dialipina in our analysis raises two major difficulties. First, it is unclear what a priori assumptions, particularly those relating to the composition of the ingroup and outgroup, are appropriate. We therefore chose not to specify an outgroup, and report our results as unrooted networks where necessary. Second, the unusual morphology of Dialipina makes coding some characters difficult for this genus. For example, Schultze and Cumbaa (2001) reported that Dialipina lacks bones of the cheek and lower jaw generally thought to be homologous between actinopterygians and sarcopterygians, and suggested that this provides evidence for the parallel consolidation of much of the dermal exocranium in the two osteichthyan radiations. However, this inference is clearly predicated on the belief that Dialipina is a stem actinopterygian. There is a more parsimonious alternative that is consistent with the (unrooted) cladistic hypothesis of these authors: numerous small ossifications making up the skull and jaw is the primitive osteichthyan condition (a surmise supported by outgroup comparison with acanthodians; Denison, 1979; Gagnier and Wilson, 1996), consolidation of the cranial dermal skeleton is a synapomorphy underpinning the clade Actinopterygii plus Sarcopterygii, and Dialipina is a stem osteichthyan which retains a plesiomorphic cranial morphology. Regardless of its possible phylogenetic significance, the dermal bone pattern of Dialipina as described by Schultze and Cumbaa (2001) is truly exceptional among osteichthyans and is in need of further investigation.

Inclusion of Dialipina in analyses yields a topology essentially identical to that recovered from the data set which excludes this genus (select characters ordered: N = 4: L = 117; CI = 0.5128; RI = 0.7031; RCI = 0.3606: all characters unordered: N = 4; L = 114; CI = 0.5175; RI = 0.7043; RCI = 0.3645). If rooted to preserve the monophyly of the least inclusive clade containing all ingroup taxa from previous analyses, Dialipina inserts above Cheirolepis and Osorioichthys, but does not otherwise alter the pattern of interrelationships depicted in our preferred cladogram (Fig. 8). An identical arrangement is recovered if three additional characters widely cited as actinopterygian synapomorphies but absent in Dialipina are included in the analysis (select characters ordered: N = 4; L = 126; CI = 0.5000; CI = 0.6986; RCI = 0.3493; all characters unordered: N = 4; L = 123; CI = 0.5041; RI = 0.6995; RCI = 0.3526).

We do not claim that the reconstructed position of Dialapina in our analyses is accurate. The placement of this genus so high within actinopterygian phylogeny is almost certainly an artifact of a data matrix not designed for resolving deep splits within Osteichthyes. If the position of Dialipina is constrained such that it lies outside of a clade containing unequivocal actinopterygians, there is a slight loss of resolution relative to our preferred solution, with Cheirolepis, Oxorioichthys, and the clade uniting all other Devonian actinopterygians placed in a polytomy. This same arrangement is found regardless of character ordering or the inclusion of putative actinopterygian synapomorphies. Although these exercises do not provide new insights on the phylogenetic affinities of Dialipina, they do indicate that our hypothesis of interrelationships among Devonian actinopterygians (Fig. 8) is robust to the inclusion of this genus in analyses.

The inability of previous studies (Zhu and Schultze, 2001; Cloutier and Arratia, 2004) to establish convincingly the phylogenetic position of Dialipina within Osteichthyes raises serious doubts about the assignment of the apparently more primitive (Schultze, 1992) and largely scale-based genera Naxilepis, Lophosteus Pander, 1856, and Andreolepis to Actinopterygii (cf. Schultze, 1977). The systematic position of these taxa is unclear (Janvier, 1996, p. 184), and it seems possible that at least some of them may fall outside the osteichthyan crown group. The absence of a peg-and-socket articulation between the scales assigned to these taxa, which is found in both early sarcopterygians ('osteolepiforms': Jarvik, 1980; lungfishes and basal dipnomorphs: Denison, 1968; Jessen, 1980) and actinopterygians, casts some doubt on their placement within crown-group Osteichthyes. If the similarities between Lophosteus and placoderms noted by Burrow (1995) and Janvier (1996) are not merely superficial but are instead indicative of phylogenetic affinity, then that genus may fall outside of crown-group Gnathostomata. Unfortunately, the limited taxonomic and morphological scope of our study does not permit us to make any further contributions to the debate on the phylogenetic position of Dialipina and other putative early actinopterygians. It is hoped, however, that the issues raised here will provide fertile ground for future research.

The status of Devonian actinopterygian taxonomy.-Classification schemes of early actinopterygians are cluttered with numerous ill- defined families, the monophyly of many of which is doubtful. This has led some authors (e.g., Long, 1988) to refrain from placing Devonian forms in established families until a steady set of relationships emerge, while others (e.g., Prokofiev, 2002) have strongly defended existing family-level taxonomy. Our cladistic solution allows us to review the families of Devonian actinopterygians identified in Gardiner (1993), the most recent synopsis of the family-level classification of early actinopterygians (see also Gardiner, 1963; Obruchev, 1964; Kazantseva, 1971; Schaeffer, 1973). However, since our cladogram is artificially truncated at the end of the Devonian, we cannot assure that families that appear monophyletic in our study will remain so when post-Devonian taxa are considered. Our review should therefore be regarded as preliminary.

Of the six Devonian families identified by Gardiner (1993), three (Cheirolepididae Pander, 1860; Osorioichthyidae Gardiner, 1967; Tegeolepididae Romer, 1945) contain single genera. Of the remaining three families, two (Mimiidae Gardiner, 1993; Stegotrachelidae Gardiner, 1963) are probably nonmonophyletic. As defined by Gardiner (1993), Mimiidae comprises the genera Mimia and Howqualepis. However, our results indicate that Howqualepis is more closely related to Tegeolepis than it is to Mimia, rendering Mimiidae polyphyletic. The shortest tree that resolves this family as monophyletic is seven steps longer than the most parsimonious solution (119 steps vs. 112 steps). We suggest that Howqualepis should be removed from Mimiidae, and placed either in Tegeolepididae or in a monotypic family of its own. Gardiner (1993) placed both Stegutrachelus and Moythomasia within Stegotrachelidae, and our results dictate that Krasnoyarichthys must be included in this family as well. Broadly speaking, Stegotrachelidae corresponds in content to the 'Moythomasia' group of Gardiner and Schaeffer (1989), who considered its monophyly doubtful. We agree with their assessment. Although Stegotrachelidae is monophyletic in one of the most parsimonious solutions, it is paraphyletic in the remaining three trees and there are no compelling synapomorphies that would seem to underpin this group. Only one species of Kentuckiidae, K. hlavini, has been included in this analysis, making it impossible to assess the status of this family. However, given the uncertainties surrounding the coherency of Kentuckia (see above), the monophyly of Kentuckiidae should be viewed skeptically. Although the sister- group relationship between Limnomis and Cuneognathus appears robust, we are hesitant to eitherinclude these taxa in a preexisting family (e.g., Kentuckiidae) or erect a new family for them pending more thorough studies that include post-Devonian forms.

The environmental context of earlv actinoptervgian evolution.- The environmental context for the origins and evolution of early vertebrates has been a subject of considerable interest and debate (Romer and Grove, 1935; Thomson, 1980; Halstead, 1985). Despite the development of geochemical techniques that complement traditional sedimentological, paleontological, and ichnological methods of paleoenvironmental inference, the paleosalinity of many fossil fish habitats remains a point of contention (Schultze and Cloutier, 1996). Uncertainties are compounded when examining Paleozoic assemblages, where constituent taxa have no extant phyletic analogues that might permit estimation of paleoenvironmental conditions based upon uniformitarian assumptions of ecology within clades (Elder and Smith, 1988). Although the paleosalinity of particular depositional environments and, more specifically, the salinity tolerances of individual fossil taxa are likely to remain controversial, it nevertheless seems worthwhile to consider the implications that the hypothesis of Devonian actinopterygian relationships presented here has for the environmental evolution of the group.

The earliest putative actinopterygian remains suggest a marine origin for the clade. The Upper Silurian marine deposits of Gotland, Sweden, yield remains of Andreolepis (Gross, 1968; Janvier, 1971, 1978; Mrss, 2001), while Naxilepis has been recorded from similarly aged sediments in China (Wang and Dong, 1989). The early Devonian saw an apparent proliferation of actinopterygian-like taxa, but the group appears to have remained confined to marine environments (Schultze, 1968, 1992; Basden et al., 2000; Basden and Young, 2001; Schultze and Cumbaa, 2001). There is no evidence of actinopterygians in Lower Devonian (Pragian-Emsian) estuarine facies that yield sarcopterygians, acanthodians, placoderms, osteostracans, and heterostracans (Elliott and Johnson, 1997). Even if many or all of these remains are shown not to belong to total-group Actinopterygii, a marine origin for that clade seems probable in light of the ecology of primitive members of the major sarcopterygian radiations (Thomson, 1980; Janvier, 1996).

Middle Devonian deposits not only yield the oldest universally recognized articulated actinopterygians, but also record the environmental expansion of the clade, with representatives found in sediments believed to have been deposited in freshwater ecosystems. Cheirolepis is found in the lacustrine (Rayner, 1963; Trewin, 1986) Caithness Flagstones of the Middle Old Red Sandstone in Scotland. Classical sedimentological interpretations of the Orcadian Lakes as freshwater have been corroborated by isotopic analyses of fossil bone apatite (Schmilz et al., 1991). However, Cheirolepis is scarce at Achanarras, and is only found in those beds deposited during times of maximum transgression. During these intervals it is believed that the Orcadian Lakes were in communication with the sea. raising the distinct possibility that this genus was only an occasional migrant into lacustrine environments. However, finds of Cheirolepis in the similarly aged deposits of Tynet Burn, which are thought to have been deposited in fluvial and lacustrine settings (Hamilton and Trewin. 1988), suggest a freshwater ecology for this taxon. The lacustrine Givetian Exanboe Fish Bed of Shetland, Scotland, yields specimens of Stegotrachelus, while Howqualepis is known from the late Givetian shales of Mt. Howitt, Australia, which have been interpreted as freshwater in origin (Long, 1982, 1988). The Givetian Aztec Siltstone of southern Victoria Land, Antarctica, has yielded remains of a yet undescribed actinopterygian (Young, 1989, 1991), and is believed to have been deposited in an alluvial plain setting, complete with representative point bar, back swamp, and lacustrine environments (McPherson, 1978).

Cheirolepis is also present in the rich Late Devonian (Frasnian) ichthyofauna of the Escuminac Formation, Miguasha, Canada. Although these sediments had long been interpreted as freshwater in origin, multiple lines of evidence now suggest that they were deposited in an estuarine setting (Chidiac, 1996; Prichonnet et al., 1996). Based on the absence of small specimens of Cheirolepis at Miguasha, Arratia and Cloutier (1996) hypothesized that this genus may have been anadromous, with young individuals living in freshwater environments. However, this inference does not agree well with the distribution of Cheirolepis in Scotland, where only large individuals are found in putatively freshwater deposits (Trewin, 1986).

In the Famennian, freshwater actinopterygians are represented by Limnomix from oxbow lake and river channel deposits of the Catskill Formation of Pennsylvania. The closely related Cuneognathus marks an additional freshwater occurrence of actinopterygians in the latest Devonian. Unlike Limnomix, which is known both from overbank and channel facies (Daeschler, 2000), Cuneognathus has thus far only been recovered from lake deposits. It is unclear if the absence of this taxon from higher-energy channel deposits is a taphonomic artifact caused by preferential destruction of small, fragile actinopterygian elements, or instead represents a significant absence indicative of habitat selectivity.

Assuming that actinopterygians are primitively marine, the hypothesis of interrelationships presented above (Fig. 8) requires at least four independent invasions of freshwater habitats during the Devonian (Cheirolepis trailli, Stegotrachelus, Howqualepis, and the clade Limnomis plus Cuneognathus; Fig. 9). This suggests that the assembly of the earliest freshwater ecosystems was dominated not by unique, isolated 'seedings' of these novel environments by primitively marine clades, but instead by iterative and relatively frequent colonization events (cf. Thomson. 1980).


In a cladogram accompanying their description of the Early Devonian (Lochkov