Osteology and systematic affinities of Palaeonotopterus greenwoodi Forey 1997 (Teleostei: Osteoglossomorpha)

Transcription

1 Zoological Journal of the Linnean Society (2001), 133: With 14 figures doi:lo.1006/zjls , available online at httpj/~r~~.idealibrary.com on ID E C Osteology and systematic affinities of Palaeonotopterus greenwoodi Forey 1997 (Teleostei: Osteoglossomorpha) :LIONEL CAVIN (C;IS "PalBontologie et sedimentologie continentales" (Toulouse-Esperaza), Musee des Dinosaures, Fsperaza, France PETER L. FOREY* Department of Palaeontology, The Natural History Museum, Cromwell Road, London SW7 5BD Received October 2000; accepted for publication November 2000 The anatomy of the osteoglossomorph Palaeonotopterus greenwoodi Forey is redescribed on the basis of more complete material than used in the original description. This new material shows large rounded and attached parasphenoid tooth plates with a Plethodus-like histology. We are therefore able to associate some species of lnlethodus with a skull. Not all species of Plethodus may belong to the same kind of fish: specifically, Plethodus oblongus Dixon, which is known from articulated skull material, appears to be a very different kind of fish. ~~alaeonotopterus shows a mixture of notopterid and monnyroid characters. With notopterids Palaeonotopterus shares an elongate foramen for V +VII straddling the suture between prootic and pterosphenoid in the orbital wall, an auditory fenestra between the prootic and basioccipital (homoplastic with Hiodon), a sagitta with a prominent anterior process (inferred in Palaeonotopterus from the shape of the labyrinth cavity). With primitive morrnyroids 13alaeonotopterus shares a similar and distinctively shaped supraoccipital crest and a suture between the parasphenoid and autosphenotic. A supraorbital branch of the otic sensory canal, formerly thought to be a character of notopterids, is present in morrnyroids and therefore is interpreted as a character of notopterids + mormyroids. % 2001 The Linnean Society of London ADDITIONAL KEY WORDS: mormyroids - trigemino-facialis chamber - sensory canals - pectoral girdle. INTRODUCTION In 1997 Forey described an incomplete braincase from the Albiaflenomanian Kem Kem Beds of southern Morocco under the name Palaeonotopterus greenwoodi, the generic name implying immediate relationships with modern notopterids. Notopterids (knifefishes) are a small group of fresh and brackish water fishes known hy four genera distributed in Africa (Papymcranus and.xenomystus), Asia (Notopterus) and southeast Asia (Chitala). Taverne & Maisey (1999) described an alrnost identical braincase from the same locality, the chief difference being that their specimen has the * Corresponding author. p.forey@nhm.ac.uk C /0l/ $35.00/0 25 dorsal part of the basisphenoid intact and a betterpreserved occiput. Both authors I-ecognized the notopterid affinities based on the possession of a groove within the frontal above the eye which they interpreted as having carried an anterior branch of the otic sensory canal which has been recognized as a synapomorphy of notopterids. Forey's reason for erecting a new genus and species centred on the fact that the basioccipital is expanded ventrally as paired processes although he could not ascribe a function to these structures. We now have new specimens that show that these processes supported a greatly enlarged parasphenoid tooth plate that has the histology matching that seen in species of Plethodus, a genus originally recognized on tooth plates and showing an unusual type of tubular histology. We are now able to associate some of these tooth plates with a skull. This association has been a 2001 'I'he Llnnran Society of 1,ondon

2 26 L. CAVIN and F? L. FOREY noticed by Taverne (2000~) who has studied newly found material of Palaeonotopterus, as well as by Dutheil(1999: 557, as crossognathiforms). Our new specimens also show much more of the braincase anatomy, such as the parasphenoid, basisphenoid, cranial cavities as well as anterior vertebrae, parts of the shoulder girdle and extrascapulars. We therefore choose this opportunity to describe these additional features and to reassess the systematic position of Palaeonotopterus as well as comment on the relationship between notopterids and mormyroids (in this paper we use mormyroids to include the families Gymnarchidae and Mormyridae and therefore follow Taverne (1998). MATERIAL AND METHODS During this study the following specimens were examined. Fossil specimens are prefixed with a superscript dagger. Recent material is suffixed by the following symbols: C&S, cleared and stained; SK, dry skeleton; SF', spirit specimen. The specimens studied belong in the collections of Musee des Dinosaures, Esperaza, The Museum of Comparative Zoology, Harvard, The Natural History Museum, London, and Booth Museum, Brighton and are prefixed by MDE, MCZ, BMNH and BOM respectively. Elopidae Elops saurus Linnaeus BMNH [SKI Elops saurus Linnaeus MCZ [C&S] Elops saurus Linnaeus MCZ [C&S] Elops affinis Regan BMNH [SKI Megalopidae Megalops cyprinoides (Broussonet) BMNH [SKI, BMNH [SKI Albulidae A1 bula vulpes (Linnaeus) BMNH [SKI Albula vulpes (Linnaeus) BMNH [SKI Albula vulpes (Linnaeus) MCZ [C&S] Hiodontidae Hiodon alosoides (Rafinesque) BMNH [SKI Hiodon alosoides (Rafinesque) BMNH [SKI Osteoglossidae Hetemtis niloticus (Cuvier) BMNH [SKI Hetemtis niloticus (Cuvier) BMNH [SKI Hetemtis niloticus (Cuvier) BMNH [SP] Osteoglossum bicirrhosum Agassiz in Spix & Agassiz BMNH [SKI Sclempages leichardti Gunther BMNH [SKI tphareodus encaustus (Cope). Eocene, Green River Formation, Wyoming, USA. BMNH P Notopteridae Notopterus notopterus (Lacepede) BMNH [SKI Notopterus notopterus (Lacepede) BMNH [SKI Notopterus notopterus (Lacepede) BMNH [SKI Xenomystus nigri Giinther BMNH [SKI Papyrocranus afer (Gunther) BMNH [SKI Papymranus afer (Giinther) BMNH [SKI 'Palaeonotopterus greenwoodi Forey. All specimens of this species are from Albian/Cenomanian Kem Kem Beds of Taouz area, southern Morocco. BMNH P.64799: holotype, incomplete braincase BMNH P XII: single individual in 12 pieces showing posterior part of skull, anterior portion of vertebral column and partial shoulder girdles MDE F6: posterior part of braincase with attached parasphenoid tooth plate MDE F7I-111: posterior part of braincase with attached parasphenoid tooth plate (in three pieces) MDE F8: posterior part of braincase MDE F9: posterior part of a skull roof MDE F10: dorsal part of a hyomandibular Two tooth plates can be referred to Palaeonotopterus with reservation: MDE F11: lower dental plate (basihyal and/or basibranchial) MDE F12: lower dental plate (basihyal and/or basibranchial) Mormyridae Petmcephalus bane (Lacepede) BMNH [SKI Petmcephalus bane (Lacepede) BMNH [SKI Petmcephalus bovei (Valenciennes) BMNH [SP] Mormymps anguilloides (Linnaeus) BMNH [SKI Mormymps anguilloides (Linnaeus) BMNH [SKI Mormyrus caschive Linnaeus BMNH [SKI Mormyrus caballus Boulenger BMNH [SKI Mormyrus rume Valenciennes in Cuvier & Valenciennes BMNH [SP] Hypemspisus bebe (Lacepede) BMNH [SKI Plethodontidae iplethodus expansus Dixon BMNH P.7 (lower tooth plate), Albian, Kent, England 'Plethodus expansus Dixon BMNH P.8 (lower tooth plate), Albian, Kent, England 'Plethodus expansus Dixon BMNH (lower tooth plate), Albian, Kent, England

3 'Plethodus expansus Dixon BMNH P.7277 (upper tooth plate), Cenomanian, Cambridge, England tplethodus expansus Dixon BMNH (lower tooth plate), Cenomanian, Cambridge, England 'Plethodus expansus Dixon BMNH P.7280 (upper tooth plate), Cenomanian, Cambridge, England iplethodus expansus Dixon BMNH P.7281 (upper tooth plate), Cenomanian, Cambridge, England tplethodus expansus Dixon BMNH (upper tooth plate), Cenomanian, Kent, England 'Plethodus expansus Dixon BMNH (lower tooth plate), Cenomanian, Kent, England Plethodus expansus Dixon BMNH (upper tooth ]plate), Cenomanian,?Kent, England Plethodus expansus Dixon BMNH (lower tooth plate),?turonian, Wiltshire, England Pentanogmius pentagon (Woodward) BMNH a 1 holotype] (lower tooth plate), Cenomanian, Kent, Englland Pentanogmius pentagon (Woodward) BMNH (lower tooth plate), Cenomanian, Kent, England Dixonanogmius oblongus (Dixon) BOM [holotype] (lower tooth plate), Cenomanian, Sussex, Engl and ldzxonanogmius oblongus (Dixon) BOM (head), Cenomanian, Sussex, England idixonanogmius oblongus (Dixon) BMNH (head), Cenomanian, Surrey, England 13sp (21 (;or 13pi 13po 13x0 13xtsc 171 I-Im I c I, IVIco 1V.a (Irs ]'a 13.0 l'sp ]'st s 1iad.d 1tad.p 12b Sca Soc ABBREVIATIONS USED IN FIGURES basisphenoid cleithrum coracoid epineural epiotic exoccipital extrascapular frontal hyomandibular intercalar lepidotrich mesocoracoid neural arch orbitosphenoid parietal prootic parasphenoid posttemporal pterotic pterosphenoid distal radial proximal radial rib scapula supraoccipital art. con art.fac art.rad asc.pro.psp c.v + VII c.1 c.x d.f f.cor fen f.i.c h0r.sem.c hy.f inf.s.c j.c lam med.lam.cl my.p 0t.s.c ped.bsp p0p.s.c AFFINITIES OF I? GREENWOOD1 27 supraoccipital crest autosphenotic first free vertebra anterior ampulla exterior ampulla posterior ampulla anterior opening of the jugular canal articular condyle articular facet articular facet for the radial ascending process of the parasphenoid bony bridge basipterygoid process of the parasphenoid chamber for the intracranial diverticulum of the swimbladder canal for the ti-igeminal and facial nerves canal for the olfactory nerve canal for the vagus nerve dilatator fossa coracoid foramen fenestra foramen for internal carotid artery scapular foramen foramen for the olfactory nerve foramen for the optic nerve foramen for the trigeminal and facial nerves groove for the lateral line groove for the mucous canal groove for the otic sensory canal groove for the preopercular sensory canal groove for the antero-dorsal process of the sagitta otolith groove for the supraorbital branch of the otic sensory canal groove for the supraorbital sensory canal groove for the hyomandibular ramus of VII horizontal semicircular canal hypophysial foramen infraorbital sensory canal jugular canal lamella medial lamina of the cleithrum posterior myodome otic sensory canal pedicel of the basisphenoid preopercular sensory canal

4 28 L. CAVIN and P. L. FOREY S0.S.C stt.com temp.fen tube X ut.ch posterior opening of the jugular canal posterior vertical semicircular canal prootic bridge sacculo-lagenar chamber supraorbital branch of the otic sensory canal supraorbital sensory canal supratemporal cornmissure temporal fenestra tube lying parallel to so.br utricular chamber SYSTEMATIC PALAEONTOLOGY SUPERORDER OSTEOGLOSSOMORPHA GREENWOOD ETAL., 1966 ORDER OSTEOGLOSSIFORMES REGAN 1909 FAMILY NOTOPTERIDAE BLEEKER 1859 PALAEONOTOPTERUS FOREY 1997 Qpe and only species. Palaeonotopterus greenwoodi Forey 1997 Palaeonotopterus greenwoodi 1997 Palaeonotopterus greenwoodi Forey: pl. 1 and figs Palaeonotopterus greenwoodi Forey; Taverne and Maisey: figs c Palaeonotopterus greenwoodi Forey; Taverne: figs 1-8 Diagnosis (emended). Notopterid in which the supraorbital sensory canal is partially roofed; supratemporal commissure contained entirely within an enclosed canal in the supratemporals (extrascapulars) which meet in the midline; no foramen between the supraorbital groove for the supraorbital branch of the otic sensory canal; basioccipital bearing prominent ventrolateral processes and, with the parasphenoid, supporting a large circular tooth plate with complex histology resembling fused teeth; ascending process of the parasphenoid sutured with autosphenotic. Holotype. BMNH P.64799, an incomplete braincase preserved in three dimensions. Horizon and locality. Late Albian or early Cenomanian, Kem Kem Beds, near Taouz, southeastern Morocco. Material. See above. COMPARATIVE DESCRIPTION The general morphology of the new braincase specimens agrees with that included in the previous descriptions of Palaeonotopterusgreenwoodi (Forey, 1997; Taverne & Maisey, 1999). As in the original material (BMNH P and AMNH 19559) the anterior part of the braincase (ethrnoid region and anterior extremities of the frontals) of all available specimens is not preserved. BRAINCASE BMNH P and AMNH lack the parasphenoid and most of the basisphenoid (only the dorsal part is preserved in AMNH where it forms a transverse bridge separating the optic foramen anteriorly from an hypophysial fenestra posteriorly). In BMNH P the postero-dorsal part of the basisphenoid is present (Fig. l, Bsp): it is expanded ventrolaterally and thus completely encloses the hypophysial fenestra which forms an antero-posteriorly elongated foramen which is orientated ventrally. In Notopterus the hypophysial foramen is a large opening bordered anteriorly by the basisphenoid and posteriorly by the prootics and which accommodates the ventral extremity of the conical hypophysis (Taverne, 1976). In Xenomystus the hypophysial foramen is small because of the development of the parasphenoid which obstructs the posterior myodome and no longer accommodates the hypophysis. In Papycmcranus, which lacks the basisphenoid, the hypophysial foramen is minute and visible as a small pore between the prootics (Taverne, 1976). The new specimens (MDE F6, MDE F7II) reveal another important difference between the extant notopterids and Palaeonotopterus greenwoodi. In the former the basisphenoid is small and fails to contact the parasphenoid (Notopterus and Xenomystus) or is even absent (Papymcranus), whereas in Palaeonotopterus the basisphenoid forms a median, anteroventrally inclined bridge which is sutured with the parasphenoid by a large contact (Fig. 2A, ped.bsp; Fig. 3B, Bsp). The parasphenoid is remarkable for its size and shape (Figs la, 2B, Psp): its ventral face forms a massive and ellipsoidal concave smooth biting surface braced against the braincase posteriorly by the basioccipital (basioccipital processes of Forey, 1997), anteriorly by the basisphenoid and laterally by the autosphenotic and well-developed lateral extensions of the prootics. Although this area is not well preserved on all the available specimens, the ascending process of the parasphenoid seems to be well developed and forms the ventrolateral wall of the anterior portion of the posterior myodome and extends dorsally, covering part of the lateral face of the prootic and suturing with the autosphenotic (BMNH P , Fig. 1, asc.pro.psp). The internal structure of the parasphenoid tooth plate is composed of dense matter (dentine?) with tubular structures (Taverne, 2000c: fig. 8C), very similar to those observed by Woodward (1907) on

5 AFFINITIES OF i? GREENWOODI 29 - Figure 1. Palaeonotopterus greenwood2 Forey Orbltal vlew of braincase (BMNH P ) A, terpretlve line drawing. Scale bar = 10 mm photograph. B, in- Plethodus tooth plates. When complete (MDE F6, MDE 1'7II), the anterior rim of the tooth plate is regularly curved (Fig. 2B) and ends at the anterior margin of t.he suture with the basisphenoid, further posteriorly than the expected anterior extreinity of a 'normal' parasphenoid dentition. It is assumed that the ellipsoidal concave surface corresponds to the biting portion of the parasphenoid strengthened by the fusion

6 30 L. CAVIN and F? L. FOREY Figure 2. A photograph; A2, interpretive line drawing of floor braincase and parasphenoid in dorsal view (MDE F7I). Scale bar = 10 mm. B, Palaeonotoptems greenwoodi ~orey. Parasphenoid in ventral view (MDE F7I). C, Plethodus expansus Dixon. Dorsal view of parasphenoid (BMNH 38585). D, Palaeomtopterus greenwoodi Forey: Dl, photograph; D2, interpretive line drawing of the internal view of otic region of the left side of the braincase showing cavities of the inner ear (MDE F7I). Scale bar = 10 mm.

7 AFFINITIES OF fl GREENWOODI 31 Figure 3. Palaeonotopterusgreenwoodi Forey. Right lateral view of braincase (MDE F6). A, photograph. B, interpretive line drawing. Scale bar = 20 mm. of teeth and that the supposed fragile, probably toothless, anterior extremity of the parasphenoid as well as the vomer were generally detached before fossilization. No Recent notopterids bear such a huge biting portion of the parasphenoid but in the extant mormyrid Hypertopisus, the biting area is very large and indeed extends forward to the ethmoid region (Taverne, 1972: fig. 35), almost triangular in shape, and bears numerous -bulbous and flattened teeth which are compressed.against each other, forming an almost continuous hard tenameloid surface. In our collection there are two ellipsoidal bones ICMDE F11, MDE F12) which show a convex occlusal :surface composed of dental tissue (Fig. 4A), suggesting -that they are lower tooth plates (basihyals or ba- ~sibranchials) which may have occluded with their cor- :responding parasphenoid tooth plate. One (MDE F11) lhas an eroded aboral surface, while the other (MDE 6'12) shows four paired pits and a median longitudinal ipoove. It is possible that these paired pits are for the ;articulation of hypohyals and hypobranchials and the median groove is for the basihyal/basibranchials. Without articulated material we cannot be certain as to which gill arches this plate belongs. However, we do ]note the remarkable similarity shown by the de- 13ression on this plate and those shown by Schaal (1984: fig. 28B) on the aboral surface of Plethodus tibinensis despite the overall shape of the plate being different. Since these plates have not been found in situ we can only refer these specimens (MDE F11, MDE F12) to Palaeonotopterus with reservation. Taverne (2000~) also describes such a tooth plate associated with a Palaeonotopterus braincase and interprets an isolated lower dental plate as the opposing basihyay basibranchial tooth plate. The structure of the trigemino-facialis chamber and associated foramina is influenced by the lateral expansion of the prootic which forms a pillar reinforcing laterally the attachment of the large parasphenoid. This part of the skull is damaged on the two previously described specimens and it was thought that the jugular canal was covered by a very narrow bridge (Forey, 1997; Taverne & Maisey, 1999). BMNH P and MDE F6 show that the jugular canal is actually deeply enclosed in the prootic (Fig. 2A, j c) and covered laterally by the parasphenoid and autosphenotic (Fig. 1, ant.0p.j.c). A foramen for the hyomandibular branch of VlI opens into the posterior part of the jugular canal (BMNH P and MDE F6). Posteriorly the jugular vein left the pars jugularis by a large round opening and ran on the lateral face of the prootic in a groove fonned ventrally by a thin lamina of bone (MDE F6). The dorsal margin of the lamina is broken in MDE F6 but suggests from comparison with Recent notopterids that this may represent part of a canal: it probably contained a foramen for the exit of the hyomandibular branch of VII. Below this lamina a ventro-posteriorly oriented canal opens for the course of the orbital artery (the course of this artery is visible in section on the left side of BMNH P.64799). The orbital face of the prootic was described by Forey (1997) and Taverne & Maisey (1999). All specimens showing this part of the skull (BMNH P and MDE F6) have a foramen for V and VII which is large and slitlike and straddles the suture between prootic and pterosphenoid (Fig. 1, f.v + VI [). The right side of BMNH P shows a thin lamella extending antero-ventrally from the trigeminal foramen and applied against the lateral face of the prootic and the basisphenoid (Fig. 1, lam). Its dorsal extremity is not

8 32 L. CAVIN and I? L. FOREY Figure 4. Lower tooth plate referred to Palaeonotopterus greenwoodi. A, dorsal view (MDE F11). Scale bar= 10 mm. B, ventral view (MDE F12). Scale bar= 10mm. preserved and it is unclear whether it is an outgrowth of the pterosphenoid or of the prootic. Its dorsal margin is rolled over and forms a delicate groove. This groove could have accommodated rami of the V nerve (possibly the maxillaris and mandibularis rami), these having left the brain cavity from the ventral part of the slitlike foramen, whereas another set of rami (possibly the ophthalmic rami of the V and VII) passed directly from the dorsal part of the slitlike foramen. On the right side of the same specimen, antero-medially to the opening into the jugular canal there is a thin lamina of bone pierced by a fenestra (Fig. 1, lam) which we interpret conveyed the orbital artery. These delicate laminae, which are situated ventromedially to the anterior opening of the jugular canal, are not preserved on other specimens we have examined: in these, the pattern appears to be much simpler with only two large pores. It is very likely that these differences are due to preservation. Ventral to the major opening for V and VII a pore opens, probably for the exit of a secondary blood vessel as in Notopterus (Taverne, 1978). The trigemino-facialis chamber of Palaeonotopterus agrees with the general morphology of this chamber in notopterids: in particular in showing a single, large, antero-dorsally situated foramen for V and VII and a short jugular canal. The new specimens described herein bring new characters which complicate the interpretation of the structure. The jugular canal is deeply enclosed in prootic which, in turn, is covered laterally by the parasphenoid and sphenotic (Fig. 5H). Such a pattern is unknown in notopterids where the lateral wall of the jugular canal is formed by the prootic (Fig. 5D,F) and this is probably the plesiomorphic teleost condition. However, the structure in Palaeonotopterus shares similarities with that observed in mormyrids. In mormyroids the jugular canal is covered laterally by a bridge of bone formed by a ventral expansion of the autosphenotic meeting the well-developed ascending process of the parasphenoid (Fig. 5E,G; Taverne, 1971). The lateral wall of the jugular canal in Palaeonotopterus may be regarded as structurally intermediate in the development of a mormyroid pattern: the lateral wall is already formed by the parasphenoid and sphenotic but the prootic component of this wall is still present. In Petmphulus bane (BMNH ), however, the margin of the posterior opening of the jugular canal is still formed by the prootic. In Osteoglossum bicirrhosum, there is a contact between the elongated ascending process of the parasphenoid and the autosphenotic, forming a bridge under which run the jugular vein, the orbital artery and the ophthalmic rami of V and VII (Fig. 5B; Taverne, 1977). However, this bridge is probably not homologous with that observed in mormyroids because it is situated entirely anteriorly to the jugular canal and medially to the course of the ramus buccalis of the VII and of the V. We consider this parasphenoidautosphenotic bridge to have arisen independently from that in Palaeonotopterus and mormyroids. The anterior roof of the posterior myodome is formed

9 AFFrNITIES OF P GREENWOODI 33 A Hiodon alosoides 6 Osteoglossum bicirrhosum C Heterotis niloticus - - D Notopterus notoptems E Petrocephalus bane G Mormyrops deliciosus H Palaeonotopterus greenwoodi Figure 5. Semischematic drawings in anterolateral view of the trigeminofacialis region of various osteoglossomorphs. by the prootic bridge (Fig. 2A, pr.br). On each side a foramen for the abducens opens (BMNH P.64799, ElMNH P ). The opening of VI within the roof of the myodome is a primitive feature compared with Recent notopterids in which the foramen is located more anteriorly, at the level of the suture between prootic and basisphenoid (Xemmystus), or between prootic, pterosphenoid and sphenotic (Papymcranus) or entirely in the pterosphenoid (Notopterus). The autosphenotic, partially preserved on AMNH 19559, was originally described as being very small (Taverne & Maisey, 1999) as in modern notopterids. BMNH P has a well-preserved right autosphenotic and it can be seen that this forms the dorsal part of a pillar-like support for the parasphenoid (see also Taverne, 2000~). In anterior view (Fig. 1, Sph), the autosphenotic has a long vertical suture with the prootic and ventro-laterally with the ascending process of the parasphenoid (the suture between parasphenoid and prootic is damaged in this area). The orbital face of

10 34 L. CAVIN and I? L. FOREY B PS~ \ Figure 6. Palaeonotopterus greenwoodi Forey. Right lateral view of braincase (BMNH P ). A, photograph. B, interpretive line drawing. Scale bar = 10 mm. the autosphenotic extends as a well-developed anterolateral process which is rounded in anterior view. The postero-ventral margin of the lateral face of the sphenotic is deeply excavated (Fig. 6, Sph) and receives the anterior articulatory head for the hyomandibular (still in place on BMNH P ). A hyomandibular articulating in a deep concavity of the autosphenotic is found in Hiodon and Osteoglossum, whereas in notopterids and mormyroids the autosphenotic is barely involved with or excluded from the articular facet. The dorsolateral face of the autosphenotic is elongated antero-posteriorly and almost triangular in lateral view and forms the floor of the deep dilatator fossa (Fig. 6, d.0. This fossa is roofed by the pterotic

11 AFFINITIES OF f! GREENWOODI 35 Figure 7. Palaeonotopterusgreenwoodi Forey. Dorsal view of braincase (BMNH P ). A, photograph. B, interpretive line drawing. Scale bar = 10 mm. (Figs 6,7, Pto) which is marked dorsally by the grooves for sensory canals as described on BMNH P (Forey, 1997). Few details can be added to the descriptions already given of the posterior part of the braincase (Forey, 1997; Taverne & Maisey, 1999). An opening for the internal carotid artery connecting with the posterior myodome, previously suspected by the shape of the 'basioccipital processes' (Forey, 1997), is present between the basioccipital, prootic and parasphenoid (Fig. 3, f.i.c) as in Notopterus and Papymcranus (Taverne, 1978: fig. 64). Unfortunately, as with previously described specimens the intercalars are not completely preserved so we are still unable to say whether the jugular vein is or is not enclosed in this bone as in modern notopterids. In Recent notopterids the shape of the intercalar is highly modified and extended forward, anterior to the foramina for the supratemporal rami of IX and X. In BMNH P and BMNH P (Figs 6, 7, Ic) the intercalar is attached to the posterior margin of the exoccipital, postero-dorsally to both foramina as in most lower teleosts. This position and structure imply that the jugular vein probably did not penetrate the intercalar in Palaeonotopterus. The relationships between the intercalar and prootic vary in Recent notopterids. There is a broad contact In Xenomystus, a narrow bridge in Notopterus, an ~nterrupted bridge in Papymcranus and probably no contact in Palaeonotopterus. A contact or a bridge IS present in most other osteoglossomorphs (Hlodon, Osteoglossum, Sclempages, Hetemtis) but not in the mormyroids which have no intercalar. The absence of contact between intercalar and prootic is regarded as a derived character among teleosts but the distribution of this character among the different clades of osteoglossomorphs may possibly indicate that the contact between prootic and intercalar is secondarily acquired in notopterids. In Osteoglossum b~cirrhosum, the bridge is largely formed by the posterior growth of the prootic whereas in Notopterus it is formed by forward growth of the intercalar. Thus, the bridge may not be homologous in osteoglossids and Notopterus. The temporal fenestra of Palaeonotopterus is very well developed. The inclusion of the epiotic in the margin of the temporal fenestra has been used as a character to distinguish some osteoglossomorphs (Taverne, 1998) (see below p. 44). The small braincase of Notopterus notopterus figured by Taverne (1978 fig. 62), 27mm in length, has an epiotic included in the margin of the fenestra whereas in a specimen of almost the same size, 25mm in length (BMNH ), the contribution of the epiotic to the margin of the fenestra is tiny. Also, in a large specimen, 46 mm in length (BMNH ), the epiotic is not included in the fenestra margin but a Lateral expansion of the supraoccipital is now visible above the fenestra, between the epiotic dorsally, the exoccipital postero-ventrally and the pterotic anteriorly. Thus this character may be size dependent and therefore we regard this as ambiguous. Epiotics, exoccipitals and basioccipital have been

12 36 L. CAVIN and P. L. FOREY described previously by Forey (1997) and Taverne & Maisey (1999). What is of interest in the new specimen BMNH P is that the supraoccipital crest is preserved. This is a posteriorly expanded crest which has a concave postero-ventral margin, giving a shape almost identical to that seen in Genyomyrus and Petmcephalus (Taverne, 1969: fig. 30) and many other mormyroids (Taverne, 1972). A supraoccipital crest is absent from most osteoglossomorphs but is present in Papymcranus and Xenomystus (Taverne, 1978). However, the shape of the crests is very different, being without the concave postero-ventral margin. The right hyomandibular is still articulated in the specimen BMNH P (Figs 1, 6, Hm). The posterior margin is damaged and the ventral extremity is not preserved but, using the hyomandibular of a Recent Notopterus notopterus as a model, less than one-third of the entire length is missing. The articular facet with the neurocranium is elongated and its anterior extremity forms a large hemispherical condyle articulating in a deep concavity of the autosphenotic (see above). The canal for the hyomandibular ramus of VII which runs through the hyomandibular is short and opens laterally into a deep groove (Fig. 6, gr.vii.hm). The groove shows a very pronounced ridge along the anterior border and a strong ridge posteriorly which diverges posteriorly from the path of the groove. We are uncertain of the significance of these ridges but the morphology is strikingly like that of notopterids and short-snouted (i.e. primitive) mormyroids. An extrascapular is still present above each temporal fenestra of BMNH P (Figs 1, 6, 7, Extsc). The extrascapular is very well developed although very thin. It covers the entire surface of the fenestra and extends anteriorly beyond the suture between the parietal and the frontal (Fig. 7). The posterior and medial margins of the extrascapular of both sides are damaged but each extrascapular seems to have extended backward beyond the hinder level of the neurocranium to cover partially the posttemporal and to have reached its antimere on the top of the skull. The extrascapular bears a transverse ridge at the level of the posterior margin of the temporal fenestra marking the course of the sensory canal of the supratemporal commissure (Figs 6, 7, stt.com). It appears as though the entire supratemporal commissure is enclosed in bone. A large extrascapular covering the entirety of posterior skull roof and meeting its fellow at the midline of the skull is present in old specimens of Hiodon (Fig. 8A) (Taverne, 1977). In mormyroids (Fig. 8G) and Notopterus (Fig. 8D) the extrascapulars are large but fail to contact each other in the midline. In Papyrocranus the extrascapular is very small and irregular in shape, and in Xenomystus it is restricted to a small sheet of bone in the postero-dorsal edge of the temporal fenestra. Li & Wilson (1996: fig. 1) regarded an extrascapular expanded and more or less square or irregularly triangular as plesiomorphic for osteoglossomorphs and an extrascapular slender and distinctly angular or branched as apomorphic and only observed in the fossil genera Phareodus and Brychaetus. However, Osteoglossum (Fig. 8B), Sclempages, Hetemtis (Fig. BC), Arapaima (Taverne, 1977: figs 44, 73, 95, 125) all have slender, tubular extrascapulars which are variously angled. The reduced extrascapular which incompletely covers the temporal fenestra in Papymcranus and in Xenomystus appears to be a separate state derived from the Hiodon and mormyroid condition. The sensory canal only passes within a tube through part of the extrascapular in Papymcranus and Xenomystus and lay entirely superficially in Notopterus. A sensory canal running above the extrascapular is derived compared with the complete bone-enclosed canal in other osteoglossomorphs. It appears therefore that the plesiomorphic condition for the extrascapulars within osteoglossomorphs is to have large sheet-like extrascapulars which meet in the midline and contain the supratemporal commissure within a bony tube. From this condition two forms of reduction take place. First, the extrascapulars fail to meet in the midline and the central part of the supratemporal comrnissure crosses the parietals. This occurs in osteoglossoids (sensu Li & Wilson, 1996) and Recent notopterids. Second, the reduced extrascapular may be tubular (osteoglossoids) or small and irregularly shaped (Xenomystus and Papymcranus) and the canal runs entirely above bones in the central area of the skull roof (mormyroids) or partly within tubes in the parietals (Recent notopterids). The type specimen and the specimen MDE F71 provide details of the morphology of the brain cavity. MDE F71 most clearly shows the internal surface. In this specimen a small depression is present in the internal face of the exoccipital at the postero-ventral edge of the internal margin of the temporal fenestra. This depression represents the location of the ampulla at the base of the posterior vertical semicircular canal (Fig. 2D, amp.post). The foramen for the vagus (Fig. 2D, c.x) opens at the inner margin of this depression and two foramina, one above the other, open more laterally. The dorsal of the latter formina leads into a vertical canal which runs through the epioccipital and opens within the supraoccipital in the postero-dorsal edge of the internal margin of the temporal fenestra (Fig. 2D, post.vert.sem.c). The ventral one leads into a horizontal canal which opens in the pterotic below the anterior edge of the internal margin of the temporal fenestra (Fig. 2D, h0r.sem.c). The former housed the posterior vertical and the latter the horizontal semicircular canals of the inner ear. In most lower teleosts the posterior vertical semicircular canal lies in a groove in the ventral surface of the supraoccipital (Maisey,

13 AFFINITIES OF l? GREENWOODI 37 inf.s.c 0t.s.c A H~odon alosoides \ 1 P0P.S.C B Osteoglossum bicirrhosum C ~eterotis niloticus D Notopterus notopterus - E Petroce~halus bane F ~a~aeonoto~terus greenwoodi Figure 8. Semischematic drawings of the skull roof of various osteoglossomorphs in dorsal view to show the paths of the sensory canals (right side). Thin lines =grooves; dotted lines =enclosed tubes; bold line =soft tissue sensory canals. In G the path of the sensory canal was determined from Mormymps rume. Scale bar = 10 mm. :1999), but here the dorsal part of the canal is completely enclosed within a tube in the supraoccipital and this is similar to conditions in Petmcephalus bane (Taverne, 1969), Notopterus notopterus (Taverne, 1.978), adults of Arapaimn gigas (Maisey, 1999) and old individuals of Hiodon (Taverne, 1977). The anterior opening of the horizontal semicircular canal opens into EL shallow elongated depression in the pterotic which probably accommodated the anterior and exterior ampullae of the inner ear (Fig. 2D, amp.ant, amp.ext). 1L large depression, representing the sacculo-lagenar chamber (Fig. 2A, s.l.ch) lies antero-medially to the canal for the horizontal semicircular canal and is in continuity with the auditory fenestra. This lies mainly in prootic. This chamber extends anteriorly as a welldeveloped groove (Fig. 2A, gr.sag) which may have accommodated an elongated antero-dorsal process of the sagitta otolith. Such an otolith shape is a synapomorphy of Recent notopterids (Nolf, 1985: fig. 33F; Taverne, 1998; Singh Rana, 1988). Medial to this depression is a larger one which is the utricular chamber (Fig. 2A,D, ut.ch). In BMNH P there is a rounded opening situated antero-dorsally to the utricular chamber in the pterotic. It probably leads to a canal which accommodated the otic ramus of the facial nerve. Anterior to this opening is another, lateral to the median canal accommodating the olfactory nerves and spanning the

14 38 L. CAVIN and P. L. FOREY Figure 9. Palaeonotopterus greenwoodi Forey. Posterior view of supraorbital region (MDE F7III) to show large paired chamber within the pterosphenoids which presumably housed anterior extensions of the swimbladder. A, photograph. B, interpretive line drawing. Scale bar = 10 mm. pterotic and pterosphenoid. This leads to an anteroposterior-oriented cavity extending into the pterosphenoid. This cavity (Fig. 9A, ch.int.swb) ends blindly anteriorly and is roofed by the frontal in its anterior part and is most clearly displayed in MDE F7III on each side of the olfactory canal. These paired spaces may have accommodated part of the intracranial diverticulum of the swimbladder. The sacculo-lagenar chamber is atrophied in mormyroids because of the development of intracranial diverticulae of the swimbladder between the semicircular canals of the inner ear. In Recent notopterids the sacculo-lagenar chamber is hypertrophied affecting the basioccipital and prootic which are inflated (Ta- Verne, 1998). The latter character is probably linked with the presence of an auditory fenestra. In Palaeonotopterusgreenwoodi an auditory fenestra is present and the sacculo-lagenar chamber is well developed but not hypertrophied. If the groove extending anteriorly from the sacculo-lagenar chamber is correctly interpreted, E? greenwoodi could have borne the typical notopterid sagitta which has an elongated anterodorsal process (Nolf, 1985: fig. 33F). The paired spaces excavated within the pterosphenoids are of uncertain interpretation. Xenomystus and Papymcranus have intracranial diverticulae of the swimbladder. In Xenomystus the bony support for the intracranial diverticulum involved the postero-ventral part of the braincase and involves either modification of the intercalar (Taverne, 1978: fig. 94) or ossification of the tunica externa of the swimbladder (Greenwood, 1963: 399). In either case they are not the same as in I? greenwoodi. In Papyrocranus, however, where the swimbladder enters the skull through two paired openings and expands internally as two diverticulae (Greenwood, 1963: fig. 2), the inferior diverticulum extends anteriorly into a space excavated in orbitosphenoid, pterosphenoid and prootic (Taverne, 1978: figs ). The spaces located in the pterosphenoids of f? greenwoodi are provisionally regarded as intracranial diverticulae of the swimbladder, even if the precise paths of the swimbladder diverticulum cannot be established. In mormyroids, the intracranial diverticulae of the swimbladder are closed off from the main swimbladder early in ontogeny and there is no auditory fenestra in the adult. When fully grown the intracranial diverticulae are located between the semicircular canals of the inner ear and thus are not immediately comparable with those of notopterids (Orts, 1967). The mormyroid condition would appear to be apomorphic in comparison with the notopterid condition to the extent that a continuation of the swimbladder ear connection is lost in ontogeny. In adult Gymnarchus there is still a canal in the exoccipital accommodating an extension of the intracranial diverticulum but this canal does not have any connection with the main swimbladder (Taverne, 1973). CEPHALIC SENSORY CANALS A partially roofed supraorbital sensory canal on the frontal and a supraorbital branch of the otic sensory canal extending above the orbit and running parallel to the supraorbital sensory canal are present in Ti greenwoodi (Forey, 1997; Taverne & Maisey, 1999). A connection between the supraorbital branch of the otic and supraorbital canals was presumed absent in P. greenwoodi (Forey, 1997). On the frontals of BMNH P.64799, one ridge extends postero-medially from each groove for the supraorbital canal and may mark the course of the roofed posterior part of this canal. Such

15 AFFINITIES OF P GREENWOODI 39 a branch exists in Recent notopterids and mormyroids. A lateral branch of the supraorbital canal opens at the junction of the frontal and pterotic (Forey, 1997: fig. 1). In Recent notopterids the posterior portion of the supraorbital canal is no longer roofed; instead the broad canals lie in wide grooves, and the juxtaposition between the supraorbital canal and supraorbital branch of the otic canal occurs via a large fenestra. 'Thus, the tube which opens postero-laterally from the $supraorbital canal at the junction of the frontal and pterotic described on BMNH P (Forey, 1997) may be homologous with the fenestra present in Recent ~notopterids (see below). As explained below it is important to discriminate between a lumen connection between canals in bone and a soft tissue continuity of ibe sensory canals. Of course this is impossible to verify in fossils. The presence of a supraorbital branch of the otic sensory canal was regarded as a plesiomorphic teleost character retained in notopterids (Taverne, 1978) or as a synapomorphy of the family (Taverne, 1979, 1998; Forey, 1997). In notopterids this canal and the supraorbital canal run side by side in parallel grooves separated by a bony ridge and are in contact through ti fenestra (Fig. 8D). However, the lumen of both canals remains independent, being separated from each other by a thin membrane in both larval (Omarkhan, 1949) and adult (Kapoor, 1964) specimens. The supraorbital branch of the otic sensory canal has not been previously recognized in momyroids. In Petmcephalus bane the supraorbital sensory canal rests above the orbit in a wide groove (Fig. 10B, gr.so.s.c), separated from its smtimere by a high and thin median ridge, reminiscent clf the notopterid pattern. A narrow bridge of bone overhangs the canal (Fig. 10, b.br). It is incomplete in some specimens (Taverne, 1969). The skull of Petmcephalus bane (BMNH ) shows a narrow groove on the lateral margin of the frontal in continuation with the groove for the pterotic portion of the otic canal and extending to above the centre of the orbit (Fig. 10B, gr.so.br). A small bone-enclosed tube (Fig. 10B, tubex) leads postero-laterally from the groove for the supraorbital sensory canal and opens just above the groove for the supraorbital branch of the otic canal close to the frontal pterotic suture. Ink injection in the otic canal of one specimen of E'etmcephalus bouei (BMNH ) and one of E'etmcephalus bane (BMNH ) shows, after removal of the skin, that a short and thin supraorbital branch of the otic canal lies in the groove for the supraorbital branch of the otic canal described above (Pig. 10A, so.br). There is no continuation between the supraorbital and the otic sensory canals through tubex as assumed by Taverne (1969). Some basal teleosts (for instance Leptolepis normand~ca Nybelin, 1962: fig. 1B) have a small branch of a sensory canal running in the frontal above the orbit a.nd parallel to the supraorbital canal. However, this branch comes from the tri-radiate pattern of sensory canals within the dermosphenotic and represents an antero-dorsal extension of the infraorbital canal, not an extension of the otic canal. In notopterids and Petmcephalus the demosphenotic is reduced to a tubular bone which carries the most dorsal part of the infraorbital canal and the otic canal runs directly from the pterotic onto the frontal. The infraorbital canal joins the otic canal through a fenestra (Notopterus) or a notch (Xenomystus, Papymcranus) lying between the frontal and the pterotic, or by passing entirely laterally to the frontal margin (Petmcephalus). Consequently the pattern of the otic canal present in notopterids and Petmcephalus is unlikely to be honlologous with that observed in basal teleosts and is regarded as an apomorphic character within osteoglossomorphs. In a recent papel-, Taverne (1998) regarded the fusion of the supraorbital and otic canals (his postorbital canal, character 57) as a synapomorphy of the Osteoglossiformes and Mormyriformes (or Osteoglossiformes sensu Li & Wilson, 1996). We checked this character in different osteoglossomorph species. In Hetemtis niloticus (BMNH ), which shows bone-enclosed canals, the supraorbital canal bends sharply lateral at the level of the posterior margin of the orbit and joins with both the otic and infraorbital s,ensory canals (Fig. 8C). It is possible that this posterior portion of the supraorbital, which lies almost transversely, represents the connecting tube between otic and supraorbital canals as described above in Palaeonotopterus and Petmcephalus (Fig. 8E,F). This pattern does not fully correspond to those illustrated by Taverne (1977) in that the lateral bending of the canal is not illustrated by him. In Osteoglossu~n bicirrhosum (BMNH ), a portion of the tube connecting both canals is slightly antero-laterally oriented fig. 8B). The skull of both Hetemtis and Osteoglossum bears no groove for a supraorbital branch of the otic canal. Injection of methylene blue dye into the sensory canal of specimens kept in alcohol of E-letemtis niloticus and Osteoglossum bicirrhosum shows that there is a connection of the lumen of the otic and supraorbital canals. In skulls of Mormymps deliciosus (BMNH ) and Mornzyrops anguilloides (BMNH ), the supraorbital sensory canal lies within a groove on the anterior pard, of the frontal and extends posteriorly within a bone-enclosed canal which is oriented postero-medially to end near the midline (Fig. 8G). A tube connects the lumen of the otic and supraorbital canals. However, injection of methylene blue in a spirit specimen of Mormy,rus rume (Fig. 8G) shows that there is no connection between the soft tissue lumen of the supraorbital a.nd otic. As Figure 8G shows, the otic. canal reaches antero-medially to

16 40 L. CAVIN and P. L. FOREY Figure 10. Supraorbital branch of the otic sensory canal in Petrocephalus. A, anterior part of head of Petrocephalus bouei (BMNH ) with skin removed showing path of canal revealed by injecting canal with methylene blue. B, skeleton of same region in Petmcephalus bane (BMNH ). Scale bar =5 mm. stop just short of the supraorbital sensory canal. A large skull (175 mm in length) of Mormyrus caschive (BMNH ) bears a narrow groove within the lateral margin of the frontal extending anteriorly beyond the junction between the otic and the infraorbital canals and running nearly parallel to the supraorbital canal. This corresponds to the groove for the supraorbital branch of the otic sensory canal as described in Palaeonotopterus, Petmcephalus and notopterids. On this large specimen, unlike conditions in Mormyrus rume, no connection is visible in the bone between the otic and supraorbital canals. It is thus unclear whether this pattern is diagnostic for Mormyrus caschive or whether it is size dependent. Further work on the cephalic sensory canals of osteoglossomorphs is necessary to draw up a comprehensive understanding of the cephalic sensory system. However, from this short examination we propose the following. (1) A connection between the otic and the supraorbital sensory canals by a bone-enclosed canal, emerging from the otic canal posteriorly to the connection with the infraorbital canal, is probably a synapomorphy of all osteoglossomorphs minus Hiodon. This canal is well developed in Palaeonotopterus, reduced in Petmcephalus and shortened and fenestra-like in Recent notopterids. (2) A connection between the soft tissue lumen of the otic and the supraorbital sensory canals is a synapomorphy of Osteoglossum and Hetemtis. This would correspond to the osteoglossoids sensu Li & Wilson (1996) but we have not examined many of the other taxa referable to this clade. (3) The presence of a supraorbital branch of the otic sensory canal, extending anteriorly from the junction between the otic and infraorbital canals, is a synapomorphy of the notopterids and Petmcephalus. SHOULDER GIRDLE, VERTEBRAL COLUMN AND SCALES BMNH P contains two pieces belonging to the shoulder girdle. BMNH P consists of a partial left cleithrum and associated scapula, mesocoracoid and partial coracoid. Small portions of three ribs are also attached (Fig. 11, Rb). BMNH P shows a jumble of mesocoracoid, scapula and scales from the right side which we have found difficult to interpret.

17 AFFINITIES OF P GREENWOOD1 41 :Figure 11. Palaeonotopterus greenwoodi Forey. Pectoral girdle of right side: A, lateral view; B, medial view; C, posterior view. BMNH P Scale bar = 20 mm. BMNH P shows a small portion of the dorsal,limb of the posttemporal (Figs 6, 7, Pst) but little else ]nay be said about its shape. Our comments on the :shoulder girdle are based entirely on BMNH P The cleithrum is missing dorsal and ventral exiiremities but shows a well-preserved middle portion with a very broad postbranchial lamina (Fig. 11A). At i;he ventral end of the preserved portion there is a broad medial flange (Fig. 11A,B, med.lam.cl), the pos- 1,erior end of which supports the scapula. Despite the fact that this medial flange is incomplete the relative width agrees with conditions in mormyroids examined here (Fig. 12D,E) but differs from other osteoglossomorphs, including notopterids, where a medial flange is either absent (notopterids) or relatively narrow (Hiodon, osteoglossifonns, Pantodon). The mesocoracoid (Fig. 11B.C, Mco) is very robust and amongst osteoglossomorphs is comparable with that of Osteoglossum (Fig. 12B; Taverne, 1977: fig. 57), Scleropages (ibid.: fig. 861, Arapaima (ibid.: fig. 137) and Pantodon (Taverne, 1979: fig. 44) rather than the very slender mesocoracoid seen in notopterids, mormyroids and Hetemtis. The ventral end is broken in BMNH P but that which remains suggests that it

18 42 L. CAVIN and P. L. FOREY A Hiodon alosoides D PetrocephaIus bane Figure 12. Drawings of the lower part of the right shoulder girdle in representative osteoglossomorphs in medial view to show some potential characters of osteoglossomorph subgroups. A, Hiodon alosoides-bmnh B, Osteoglossum bicirrhosum-bmnh : B1, medial view; B2, posterior view. C, Notopterus notoptems-bmnh D, Petmcephalus bane (LacepedeGBMNH E, Hypempisus bebe (Lacepede)-BMNH Note especially in Notopterus the anterior fan-shaped expansion of the coracoid; in Notoptems, Petmcephalus and Hypempisus the expanded innermost proximal radial; in Petmcephalus and Hypempisus the broad medial lamina on the coracoid and interdigitating suture with the coracoid; in Osteoglossum the very large coracoid foramen. must have sutured with both the scapula and the coracoid. The scapula completely encloses the scapular foramen (Fig. 11B,C, f.sc) and is developed posteriorly as the condylar surface for the articulation of the pectoral fin. This condylar surface is formed by two saddleshaped convex surfaces and a deep medial concavity (Fig. 11C, art.fac). This limited information prevents any detailed phylogenetic conclusions but comparisons can be made by suggesting that the broad mesocoracoid is like that of osteoglossoids (except Hetentis) and the inferred broad medial flange of the cleithrum is very similar to that of mormyroids. While comparing specimens of osteoglossomorphs with those of Palaeonotopterus we noticed several features of the shoulder girdle which may have relevance to the recognition of osteoglossomorph clades. Hitherto, most of these potential characters have not been discussed in previous papers dealing with interrelationships of osteoglossomorph subgroups (Taverne, 1979,1998; Greenwood, 1973; Li &Wilson, 1996; Shen, 1996; Li, 1994; Bonde, 1996). Characters of the shoulder girdle and fin have been used by several authors. Shen (1996) recognizes two conditions of the posttemporal: unforked and large vs. forked and small. According to Shen most osteoglossomorphs with the exception of two early Cretaceous forms (TKuntulunia and thuashia) show the second condition. Li & Wilson (1996) distinguish two types of forked posttemporal dependent on the length of the dorsal arm relative to the ventral arm. Members of the family Hiodontidae (fyanbiania, fplesiolycoptera, TEohiodon and Hiodon) show a posttemporal with a long dorsal limb: the remainder of the osteoglossomorphs show a (relatively) short dorsal limb. Gymnarchus is exceptional in that the posttemporal is reduced to a tiny tube carrying part of the lateral line. Taverne (1998) lists some 15 characters related to the morphology of the shoulder girdle. Three (his numbered ) are autapomorphic for Gymnarchus; seven are autapomorphic for numbered , ) and most of these are associated with the long and stout fin designed for 'flying'. One further [331] is autapomorphic for Ara-

19 AFFINITIES OF l? GREENWOOD1 43 paima. This leaves four characters which may be useful at a higher hierarchical level. Two of these are opposites [46, 841 and refer to the hypertrophy of the coracoid 1461 in association with the elongation of the outermost pectoral ray which Taverne optimized as characterizing ti group osteoglossiforms + mormyriforms (notopterids and mormyroids), subsequently disappearing [84] in his mormyriforms. Ignoring TSingida, which Li & \Nilson (1996) place deep within osteoglossiforms, the hypertrophied coracoid is probably a character of a s~ubgroup of osteoglossiforms (Li & Wilson's Osteoglossinae). Of the two remaining characters one [I031 describes a supracleithrum which no longer encloses the lateral line, a feature which is present in osteoglossiforms and notopterids - a homoplasous distribution in all current classifications. The other is the absence of a postcleithrum [I661- a character of mormyroids. In addition to the above characters we have identified five further potential synapomorphies. In Hiodon the coracoid is an entire unfenestrated bone (Fig. 12A). This stands in contrast to the remainder of the Recent c~steoglossomorphs in which there is a coracoid fenestra cleveloped to a greater (Osteoglossum, Fig. 12B) or lesser (mormyroids, Fig. 12D,E) degree. We have not had the opportunity to examine many fossil osteog;lossomorphs for this feature but note that in tphareodus encaustus (BMNH P.64636) and ts~ngida (Greenwood & Patterson, 1967: fig. 2B) it is developed als in Osteoglossum. The polarity of this character is unclear. Elops and Ptemthrissus lack such a fenestra vvhile Albula and Megalops cyprinoides show a tiny foramen. Thus we are tempted to propose that a cora~coid fenestra is a synapormorphy of a group o steoglossiforms + mormyriforms. The apposition of the much expanded coracoid over the entire medial surface may be a synapomorphy of the subfamily Osteoglossinae (sensu Li & Wilson, 1996). In notopterids the coracoids are very distinctively shaped as the anterior end is much expanded into a t~road fan shape (Fig. 12C) totally unlike the shape in other osteoglossomorphs and other lower teleosts. In rnormyroids (Fig. 12D,E, med.lam.cl) the cleithrum hears a broad medial lamina which is in an interdigitate sutural contact with the coracoid throughout its length. This unusual feature may have been present in Palaeonotopterus (see above) and we suggest this may be a character of mormyroids and possibly I'alaeonotopterus. In Hiodon and osteoglossiforms the proximal radial series consists of (usually) four rod-shaped radials. In r~otopterids and the mormyroids we have examined there are usually three radials, the innermost of which i,s expanded as a thin membranous flange and we suggest this expansion may be a synapomorphy of r~otopterids + mormyroids. BMNH P.65643IV, V, VI shows a few pieces of vertebral column which collectively cover the first 15 free vertebrae (the first is attached to the skull fragment I). Each of the centra is about as half as long as deep and the anterior six are marked by fine fibrous ornament. Behind this level the centra are marked by deep lateral pits. The parapophyses are autogenous. Those upon the anterior five are placed anteroventrally; thereafter they are closely centred on the ventral midline. This is of some interest because in Notopterus the anterior nine parapophyses are positioned laterally, and from here {,hey descend more ventrally until about the 15th vertebra when the very short abdominal region gives way to the long caudal region. The similar shift in pos~tion of the parapophyses in Palaeonotopterus might suggest a similar very short abdominal region. BMNH P.65643X shows four partial ribs. As in many osteoglossomorphs (except Hiodon) the ribs are highly compressed antero-posteriorly. Also there is a strong similarity with the ribs in notopterids in that each is nearly straight except dorsally where they are markedly incurved. This might suggest that the body was highly laterally compressed. BMNH P.65643IV shows very long epineurals (as in Notopterus) but their precise point of attachment to the vertebral column remains unknown. The scales are very large, at least 20mm in depth (scales just above and behind the pectoral fin). Unfortunately the preservation is poor but it can be seen that the exposed portion is marked by coarse vermiform ornament. The scales are badly eroded so it is not possible to determine whether the many cracks denote true squamules or preservational artifacts. PHYLOGENETIC POSI'I'ION OF PAL,4EONOTOPTER US Both Forey (1997) and Taverne & IMaisey (1999) considered Palaeonotopterus to be a stem species notopterid based on the synapomorphy of a supraorbital branch of the otic canal. However, we have shown above that this character is probably a synapomorphy of a more inclusive group - notopterids +mormyroids (Mormyridae + Gyrnnarchus). Therefore we need to reevaluate the characters we can observe and reasonably infer in Palaeonotopterus with characters in osteoglossomorphs in general. There have been several recent computer cladistic studies of osteoglossomorph interrelationships (Li & Wilson, 1994, 1996, 1999; Li, Grande & Wilson, 1997a; Li, Wilson & Grande, 1997b) and the overlapping authorship of these papers reflect$; overlapping data sets and similar phylogenetic conclusions. The chief conclusions relative to the interrelationships of Recent osteoglossomorphs are that notopterids and mormyroids are sister groups and that Hiodon is the

20 44 L. CAVIN and P. L. FOREY most primitive living osteoglossomorph. That is, the following classification is supported: (Hiodon, (Osteoglossiformes, (Notopteridae, Mormyroidei))). Ta- Verne (1998), in his comprehensive survey of osteoglossomorphs, reaches the same phylogenetic conclusions concerning the Recent taxa but differs considerably in the placement of some of the fossils, in particular Singida and Ostariostoma. Taverne's (1998) study was a hand-wrought cladistic study and therefore cannot be guaranteed to be globally parsimonious and he does differ slightly in the higher category names. However, his study uses some 333 characters (although some of these separate characters are in fact opposite states and some others are considered as losses of previous characters) and absorbs most of the characters used by Li and co-workers. It therefore provides a very good starting point to discuss the phylogenetic position of Palaeonotopterus. Within the context of characters observable in Palaeonotopterus we have no reason to dispute the suggested sister group Notopteridae +Mormyroidei. Another recent study of osteoglossomorphs by Bonde (1996) presents a phylogenetic tree in which notopterids are regarded, perhaps in combination with mormyroids, as the sister group of Hiodon. No characters are given in evidence but, to the extent that a notopterid fmormyroid clade is not contradicted, all phylogenetic hypotheses are compatible at this level. Of course, the very different position of Hiodon and some of the fossils will have implications for theories of homology and theories of character polarity. We will discuss the phylogenetic position of Palaeonotopterus using Taverne's (1998) characters (using his numbers labelled as [xx]) with some additions of our own. We are only concerned with those characters about which we can make comparative statements in Palaeonotopterus and because this is limited by the material we will not conduct a computer cladistic analysis. The presence of a temporal fenestra has been used by previous investigators. Taverne listed four characters relating to this feature [4, 72, 248, In osteoglossomorphs a temporal fenestra is present in Hiodon, Pantodon, notopterids and mormyroids. The size and position of the fenestrae vary between these taxa, raising questions of character polarity and even homology with temporal fenestrae within osteoglossomorph clades and within other lower teleosts. Outside of osteoglossomorphs a temporal fenestra has been described in clupeoids (e.g. Grande, 1985). In Hiodon, notopterids and mormyroids the temporal fenestra is very large and laterally directed. In Hiodon the fenestra is bordered by the parietal, pterotic and epiotic. In Palaeonotopterus, notopterids and mormyroids the margin of the fenestra lacks the parietal contribution and is usually formed by the pterotic, exoccipital and epiotic. As mentioned above Xenomystus and Papymcranus lack the epiotic contribution and Notopterus is polymorphic (p. 35). Taverne (1998: 117) suggests that the condition in notopterids and mormyroids is plesiomorphic because he believes that the temporal fenestra is a transformation of the posttemporal fossa which is located between the same three bones in elopids and albulids (Forey, 1973), Pholidophorus and Leptolepis (Patterson, 1975) and receives epaxial musculature. Taverne's argument about the polarity of transformation is weakened by the fact that in Pantodon the temporal fenestra is bordered by the parietal, pterotic and epiotic - as in Hiodon - the chief difference being the relatively small size of the fenestra in Pantodon. In the phylogeny of Taverne (1998: fig. 22) and Li & Wilson (1996: fig. 4) Pantodon is nested deep within osteoglossiformes. Thus, in both hypotheses Hiodon and Pantodon form a paraphyletic relationship relative to notopterids and mormyroids. Thus, it is more parsimonious to assume that the Hiodon/Pantodon condition is plesiomorphic relative to the notopterid/mormyroid condition, with perhaps the Xenomystus/Papymcranus condition as being further derived (lacking the epiotic contribution). The evaluation of the temporal fenestra is further complicated by the studies of Li et al. (199713) and Shen (1996). Shen does not differentiate between types of temporal fenestrae but simply records presence/ absence. However, Notopterus is scored as lacking a temporal fenestra (mormyroids and Pantodon were not considered in that study). Li et al. (199713) similarly record a temporal fenestra as being absent in notopterids, mormyroids and Ostariostoma (although Li & Wilson, 1996: fig. 3A labelled a temporal fenestra in this taxon as did Grande & Cavender, 1991). Li et al. (199713: table 4) record the presence of a temporal fenestra in Hiodon and Pantodon but not in Ostariostoma, notopterids or mormyroids. The discrepancy between the codings and labelled figures suggest some simple coding mistakes. However, the consistent absence codings in notopterids and mormyroids are more difficult to explain and may imply that those authors consider the Hiodon/Pantodon temporal fenestra as non-homologous with that in notopterids/mormyroids. It is difficult to believe that this could be so but rather these fenestrae are homologous at some hierarchical level, e.g. Osteoglossomorpha. In both the fenestrae lie opposite the cranial cavity (closed over by cartilage in Hiodon; Taverne, 1977) between the anterior and posterior semicircular canals. They differ only in the make-up of the bordering bones. Li et al. (199713: appendix 1) suggest homology between the temporal fenestrae (of at least Hiodon) and the pre-epiotic fossa. The pre-epiotic fossa is found in clupeomorphs - perhaps as a synapomorphy (Patterson, 1970: 176). If this

21 AFFINITIES OF P GREENWOODI 45 i,ransfomational homology is accepted the plesioinorphic condition of the bordering bones will be shown by Hiodon, Pantodon and Ostariostoma (Li & Wilson, 1996: fig. 3A) since the pre-epiotic fossa is bordered by parietal, epiotic and pterotic. Temporal fossae are assumed to be absent from osteoglossiforms (except Pantodon). These fishes have a roofed posttemporal fossa. We note, however, that in our specimens of Osteoglossu m (BMNH ) and Sclempages (BMNH ) the medial wall of the posttemporal fossa is incomplete. Instead there is a very large vacuity left between the pterotic, epiotic and parietal which matches precisely that seen in IJantodon (Kershaw, 1970: fig. I), the latter being regarded as a temporal fenestra (Li et al., 1997b; Taverne, 1978). In our specimens of Osteoglossum and Sclempages this fenestra was probably closed by membrane (it is difficult to be certain since these are dry skeletons). It is therefore probable that osteoglossiforms retain a temporal fossa. It also means that it is unlikely that the temporal fenestra is homologous with, or a transformation from, the posttemporal fossa since osteoglossids would have both. We summarize this discussion by suggesting that the temporal fenestra is a plesiomorphic character for osteoglossomorphs where it originally occurred on the medial wall of a roofed posttemporal fossa, as in osteoglossifoms minus Pantodon, and was bordered by at least the parietal, pterotic and epiotic. In Hiodon, Pantodon, notopterids+momyroids the roof of the posttemporal fossa was lost, probably on three separate occasions, so exposing the temporal fenestra (except ill those cases where very large extrascapulars are present). In notopterids +mormyroids a derived condition occurs where the parietal no longer contributes to the fenestra border. Taverne (1998) used two characters of the ext~rascapular within his phylogenetic hypothesis. The first [26] differentiates the Hiodon condition, where the extrascapular is very large and meets its antimere in the midline, from the remainder of the ostc?oglossomorphs where extrascapulars are reduced and fail to meet one another in the midline. Paltxeonotopterus shows the Hiodon condition which is plesiomorphic for teleosts. The second character [97] relates to the enclosure or non-enclosure of the supratemporal commissure within a bony tube in the exti-ascapular irrespective of the size of the bone (see p. 36). Again Palaeonotopterus shows a plesiomorphic enclosed condition. Only Recent notopterids show a partially non-enclosed supratemporal commissure within the extrascapular. Li & Wilson (1996: character 27) and Li et al. (1997b: character 40) recognized two shapes of the extrascapular, the derived shape being distinctly angled and branched (see p. 12): Paltaeonotopterus is plesiomorphic in this respect. Li ~t al. (1997a: character 54) maintain that inclusion of at least part of the supratemporal canal within the parietals is a derived feature. We would agree with the polarity of this character but not with their coded distribution (Li et al., 1997a: table lob) where they record the derived state as present in notopterids and morrnyroids. In the mormyroids we have examined the central part of the supratemporal commissure passes entirely in skin whereas it clearly passes within a bony tube through the parietals of Recent notopterids and osteoglossiforms (except Pantodon). This is a potential synapomorphy of notopterids + osteoglossiforms (minus Pantodon). It has arisen independently in clupeiforms. The presence or absence [88] of a basipterygoid process has been used as a character by Shen (1996: character 8) and Li et al. (199713: character 5) as well as by Taverne. Taverne lists absence of a basipterygoid process as a character of notopterids but acknowledges parallelism within mormyroids and in Gymnarchus. We note here that it is also absent in Hiodon. Li et al. (199%) record the absence of a basipterygoid process in momyroids but this is only true of some taxa. For example, the process is well developed in Hyperopisus (Taverne, 1972: fig. 33) and poorly developed in Petmcephalus (Taverne, 1969: fig. 30; although our specimen BMNH seems not to show a process) and Stomatorhinus (Taveine, 1972: fig. 76). Shen (1996: table 1) also records the absence in Osteoglossum, contra most other workers, but this assessment depends on his acceptance of the arguments of Arratia & Schultze (1991) that the process in Osteoglossum is non-homologous with that in Leptolepis. We are uncertain about the presence or absence of a basipterygoid process in Palaeonotopterus but since the process appears to have been lost independently so many times in lower teleosts it must have low significance for phylogeny reconstn~ction. The fusion of the supraorbital and otic (postorbital of Taverne) canals is regarded as a. synapomorphy of all osteoglossomorphs except the liiodontiforms and three extinct families (Huashiidae, Kipalaichthyidae, Singididae) [57]. The discussion above has shown that this character should be slightly redefined: in all osteoglossomorphs except the hiodontifoms, there is a connection between the otic and the supraorbital sensory canals through a bone-enclosed canal emerging from the otic canal posteriorly to the connection with the infraorbital canal or through a fenestra situated at the level of the connection with the infraorbital canal. It is not possible to observe this character on Kipalaichthys according to the figures of Taverne (1976). or in the Huashiidae according to the figures of Zhang (1998) or in Singidia according to the figures of Greenwood & Patterson (1967). Moreover, another character may be added: a lumen. connection between the otic and

22 46 L. CAVIN and P. L. FOREY the supraorbital sensory canals which is a synapomorphy of the osteoglossoids sensu Li & Wilson (1996). In Palaeonotopterus, the first character is present and the second is unknown. Cephalic sensory canals without pores to the exterior and without grooves for the mucous canals are regarded as a synapomorphy of mormyroids plus notopterids [83]. This character should be split into two: the cephalic sensory canals without pores to the exterior is a synapomorphy of mormyroids plus notopterids and is unknown in Palaeonotopterus; the presence of grooves for mucous canals which lie alongside parts of the sensory canals is a derived character present in Recent osteoglossoids sensu Li & Wilson (1996) except in Pantodon in which the supraorbital sensory canal lies in a shallow and wide groove on the frontal and the extrascapular comrnissure lies in the skin (Taverne, 1977). In the fossil forms of osteoglossoids, there are apparently no grooves for the mucous canals (see for instance Li, 1994) except in Singidia (Greenwood & Patterson, 1967). This character is absent in Palaeonotopterus. The presence of a supraorbital branch of the otic canal (frontal part of the supraorbital sensory canal in two grooves according to Taverne's terminology) is a synapomorphy of notopterids [87]. According to the discussion above, we state that the presence of a supraorbital branch of the otic sensory canal, extending anteriorly from the junction between the otic and infraorbital canals, is a synapomorphy of the notopterids (including Palaeonotopterus), Mormymps and Petmcephalus. This character was probably lost in other mormyroids. Another apomorphic character of the skull roof sensory canal is the fact that the otic and supraorbital sensory canals lie partially or completely in grooves. This character is present in mormyroids and notopterids and in the osteoglossiform 'Singidia (Greenwood & Patterson, 1967). A jugular canal which is not enclosed within the prootic is regarded as a synapomorphy of mormyroids [204]. The structure observed on Palaeonotopterus implies a redefinition of this character. The ascending process of the parasphenoid is sutured with the autosphenotic lateral to the jugular canal in Palaeonotopterus and mormyroids. In Osteoglossum and Sclempages, the parasphenoid is sutured with the autosphenotic and pterosphenoid as well, but this contact is situated anteriorly to the anterior opening of the jugular canal and is thus not homologous with the pattern observed in E! greenwoodi and mormyroids. A pars jugularis which is no longer enclosed in the prootic is another derived character observed in mormyroids. In Petmcephalus, there is still a tiny participation of the prootic to the posterior opening of the jugular canal and, in Palaeonotopterus, the jugular canal is completely enclosed in the prootic. The increased length of the jugular canal such that it includes the openings for V and VII and the orbital artery is regarded as a synapomorphy of pantodontids and osteoglossids [217]. However, this character is probably primitive for teleosts (Patterson, 1975). Ta- Verne (1977) described the trigemino-facialis chamber of Hiodon tergisus with an opening for the V and VII situated anterodorsally to the anterior opening of the jugular canal. The examination of a skull of Hiodon alosoides (BMNH ) shows that the opening for the V and VII is situated in the anterior part of the jugular canal, which is the primitive condition for teleosts and osteoglossomorphs. The derived character, present in notopterids and I? greenwoodi, is the presence of a large opening for the V and VII situated dorsally and straddling the suture between prootic and pterosphenoid. An intercalar is present in Palaeonotopterus, absent in mormyroids [157]. However, in Palaeonotopterus the intercalar is incompletely known and it is impossible to determine whether it formed a vertical wing covering in part the cranial diverticulum of the swimbladder [92] or whether it enclosed the jugular vein as in Recent notopterids. The polarity of the enclosure or non-enclosure of the jugular vein within the intercalar is unclear according to its distribution among osteoglossomorphs. Taverne (1977) stated that the intercalar-prootic bridge present in Osteoglossum is homologous with those of Elops. In Elops, the intercalar-prootic bridge is situated above the posterior opening of the jugular canal and the jugular vein runs below the subtemporal fossa, on the lateral faces of the prootic and exoccipital (Forey, 1973). Skulls of Osteoglossum bicirrhosum (BMNH ) and Sclempages leichardti (BMNH ) show that in these genera the posterior process of the prootic forming the intercalar-prootic bridge is situated ventrally to the posterior opening of the jugular canal and floors the groove for the jugular vein. The bridge itself is formed mainly by the intercalar and, in posterior view, the jugular vein appears completely enclosed by the intercalar as in notopterids. The situation seems to be similar in Phareodus testis and Brychaetus muelleri (Taverne, 1978). In Hiodon, there is no intercalarprootic bridge but the intercalar forms a ventrally oriented wing which overhangs the jugular vein. In a large skull of Hetemtis niloticus (68mm in length, BMNH ), the intercalar and the lateral face of the prootic are simpler in shape and do not cover the jugular vein laterally. However, a small skull (11 mm in length) illustrated by Taverne (1977: fig. 94) shows a short posterodorsally oriented process on the prootic overhanging the posterior opening of the jugular canal

23 AFFINITIES OF 19 GREENWOODI 47 and reminiscent of the intercalar-prootic bridge present in Osteoglossum and Sclempages. Consequent1 y the relationships between the jugular vein and the intercalar indicate that the prootic-intercalar bridge in Osteoglossum and Scleropages is probably nonhomologous with those of Elops and basal teleosts and that a jugular vein which is enclosed by the intercalar may be a primitive character of osteoglossomorphs retained in some osteoglossids and in notopterids. The loss of the posterior myodome is regarded by Taverne as having arisen in parallel in monnyroids, Papymcranus and Xenomystus [127], and the loss of the basisphenoid as occurring in parallel in mormyrins (mormyroids except Palaeonotopterus), Papymcranus and osteoglossiforms ( = osteoglossoids of Li & Wilson, 1996) [126]. We agree that the loss of the myodome and of the basisphenoid occurs in several teleost lin- (-ages and of little phylogenetic use because of its high level of incongruence with other characters. Palaeo-.notopterus shows the primitive condition with a broad ;myodome and a well-developed basisphenoid sutured.with the parasphenoid through a pedicel. The opening for the VI in front of the prootic bridge, instead of through the prootic bridge, is a syniapomorphy of notopterids + mormyroids and is absent jn Palaeonotopterus. Two characters relating to the bone pattern making up the articular facet for the hyomandibular have been used to discuss the relationships among osteoglossomorphs. Greenwood (1973) observed that, in ~Yiodon and notopterids, the intercalar contributes to 1,he posterior and medial borders of the facet and 'Paverne (1998) stated that a facet without a par- 1,icipation of the autosphenotic is a synapomorphy of ~normyroids fnotopterids (except Xenomystus) [77]. The first character is inconsistent: on the one hand, a skull of Papymcranus afer (BMNH ) and a11 the notopterids illustrated by Taverne (1978) show tbat the intercalar extends anteriorly below the articular facet but is never included. On the other hand, skulls of Osteoglossum bicirrhosum (BMNH ) and of Scleropages leichardti (BMNH ' ) show a tiny participation of the intercalar a t the posterior margin of the facet. The second character also seems ambiguous because a skull of Mor-?nyrus caballus (BMNH ) and one of imormymps anguilloides (BMNH ) show a participation of the autosphenotic to the anterodorsal part of the articular facet. The sacculo-lagenar bulla is hypertrophied in Recent notopterids [go] and inflates the lateral wall of the basioccipital. This character is not present in Palaeonotopterus. The saccular bulla is atrophied in morrnyroids [I601 and this character is probably linked to the presence of intracranial diverticula of the swimbladder situated between the three semicircular ca- nals. The sagitta of notopterids shows an unusual anterior process (941 which was probably present in Palaeonotopterus as suggested by the anterolaterally oriented groove of the saccular chamber. We suspect the presence of intracranial diverticula of the swimbladder in Palaeonotopterus, entering through the auditory fenestra [91] and differing from the mormyroid intracranial diverticula [I591 because of the presence of lateral chambers in the orbitosphenoid. However, it is not clear whether the structure observed in Palaeonotopterus shares similarities with either Xenomystus [I371 or Papymcranus [125] (in Hiodon, Notopterus and C'hitala, there is a simple otophysic connection by apposition of the swimbladder diverticulum to an auditory fenestra (Greenwood, 1973)). The presence of a prootic-basioccipital chamber which houses the vascular portion of the posthypophysis (Taverne, 1976) is regarded as a synapomorphy of Recent notopterids [93]. This character is absent in Palaeonotopterus. A paired orbitosphenoid is regarded as an apomorphic character acquired in parallel in mormyrins and osteoglossiforms ( = osteoglossoids sensu Li & Wilson, 1996) [208]. This character is absent in Palaeonotopterus. A well-developed lateral expansion of the anterior part of the frontals is regarded as a synapomorphy of osteoglossids and pantodontids but reduced in the Recent forms of these two families [214]. It is more relevant to retain this character as a synapomorphy of Phaerodus and Rqchaetus only (Li &Wilson, 1996). This character is absent in Palaeor~otopterus. As a result of this survey of characters which may be assessed in Palaeonotopterus we may discuss the relationships of this genus. We assume that notopterids and mormyroids are sister groups. There are four possible relationships entertained here and as can be seen there are two hypotheses nearly equally supported (Fig. 13A,B). The characters used to support the various groupings are as follows (numbers in square brackets refer to numbers used by Taverne (1998)). (1) Sensory canals partially or completely within grooves. Independently arisen in Singzda. (2) Supraorbital branch of the otic. sensory canal. (3) Temporal fenestra surrounded by pterotic and epiotic (and occasionally the exoccipital). (4) Anterior process on the sagitta otolith (inferred in Palaeonotopterus from the shape of the labyrinth cavity) [94]. (5) Foramen for \'+VII in orbital wall enlarged and straddling suture between prootic and pterosphenoid. (6) Auditory fenestra between prootic and basioccipital [91]. Independently ai-isen in Hiodon. (7) Series of ossified ligaments attached to rear of

24 48 L. CAVIN and P. L. FOREY Figure 13. Four phylogenetic hypotheses showing alternative relationships of Palaeonotopterus supported by putative synapomorphies discussed in the text. The top two solutions represent the optimal (left) and nearest suboptimal (right). See text for discussion. braincase and reaching posteriorly into the epaxial musculature [36]. As Taverne (1998: 112) notes this character is also present in the Cenomanian Kipalaichthys sekirsky Casier which is considered by Taverne to be a stem-lineage taxon to the grouping Osteoglossiformes + Mormyriformes. (8) Hypertrophy of the sacculo-lagenar chamber [go]. (9) Development of a prootic-basioccipital chamber to house a vascular sac associated with the hypophysis [93]. (10) Optic fenestra no longer present between orbitosphenoid, pterosphenoid and parasphenoid [156]. (11) Obliteration of the saccular chamber due to the expansion of the intracranial diverticulae of the swimbladder [160]. (12) Loss of intercalar [157]. (13) Supraoccipital crest which overhangs first vertebra. (14) Broad median lamina of cleithrum contacting coracoid through a sinusoidal suture. The question mark denotes incomplete knowledge in Palaeonotoptems. (1 5) Suture between parasphenoid and autosphenotic. (16) Extrascapulars failing to meet in midline. Independently arisen in Pantodon. (17) Abducens opening anteriorly through prootic bridge. (18) Orbitosphenoid closed dorsally as well as ventrally [175]. Gymnarchus shows many autapomorphies summarized by Taverne (1998: 124) among which we mention posttemporal without a dorsal limb [181],

25 AFFINITIES OF P GREENWOODI 49 supracleithrum with a condylar articulation with the epiotic [I821 and absence of an anal fin [194]. Mormyroids show fused premaxillae [19], atrophied vomer [203] and unique shape of the cerebellum [205]. Thus our conclusions are that either Pa-!aeonotopterus is the sister group to notopterids (as suggested by Forey, 1997; Taverne & Maisey, 1999; 'I'averne, 2000c) and as we continue to classify it) or -- as suboptimal solution - it is the sister group to inormyroids. We are convinced that, were more parts of the anatomy known, resolution between these two alternatives would be relatively easy. ASSOCIATION OF PALAEONOTOPTERUS WITH PLETHODUS th we have indicated above (p. 4) we can now associate skulls with tooth plates described under the generic name Plethodus. Whether all species of Plethodus can tie referred to a Palaeonotopterus-like osteog;lossomorph is a moot point. Plethodus is a genus tlased on Plethodus expansus Dixon, a species named for a median dental plate from the Turonian of England and which is leaf shaped with a convex occlusal surface and probably associated with the basihyal and/or basibranchial. Subsequent finds of both lower and concave, and therefore presumed upper occluding, tooth plates of the same shape extend this species from the Albian to Senonian (Woodward, 1901). Woodward (1899: 26, pl. 13, fig. 2) described a specimen of an upper tooth plate as having a median "fibrous azygous bony bar for attachment of some kind". This bony bar would clx-respond to the portion of the upper tooth plate of E'alaeonotopterus which is anchored to the basioccipital. Woodward (1899) described a further species Plethodus pentagon for very differently shaped plates from the Cenomanian of England and he suggested that Plethodus belonged to an osteoglossomorph fish because the association of upper and lower midline plates suggested a fish in which the primary bite was between the basihyal/basibranchial and the parasphenoid. A number of species from outside of England have blsen described. Weiler (1935: 38, pl. 3, fig. 21) described Plethodus libycus from the Cenomanian of Egypt. This is based on a nearly circular presumed upper tooth plate with a concave occlusal surface. Schaal (1984: 56, figs 28, 29; pl. 9, figs 11-15) added another species - Plethodus tibnienszs -from the Cenomanian of Egypt based on a lower and approximately pentagonal plate. Among all of these species E! lzbycus is so similar to the a1,tached parasphenoid tooth plate of Palaeonotopterus that we have little hesitation in referring both to the same kind of fish. Taverne (2000~) reached the same cc~nclusion and went further by suggesting that l? libycus is a synonym of Palaeonotopterus greenwoodl. I? libycus was described a lot earlier than l? greenwoodi and therefore the rules of nomenclature would suggest that the later should be a junior synonym. However, as Taverne noted, the type of F? libycus has been lost and hence it would be inadvisable to synonymize species with one with no holotype specimen. We would not go as far as Taverne (2000~) because we feel that absolute comparisons between the parasphenoid tooth plate of I? greenwoodi and the type specimen of P libycus can no longer be made. However, we are uncertain about the status of the other species. This uncertainty is raised because of another species originally described by Dixon (1850) as Plethodus oblongus. This species was erected for an elongate and convex tooth plate (HOM Fig. 14A) which has the same gross histology of other species. Woodwarcl (1899) was able to associate this specimen of l? oblongus with a small fragment of presumed basihyal tooth plate lying within a partial skull (BOM ) and further with another specimen of a skull (BMNH Fig. 14B). The cranial anatomy of these two skulls is completely different from that of Palaeonotopterus. For instance, the frontals are much expanded and with the parietals and pterotics are marked with fibrous and coarsely pitted ornament, there is no temporal fossa and the various proportions of the bones are very different and yet at the same time very similar to the skull roof of Ananogmius (Banunogmius, Taverne, 2000b). Furthermore, the parasphenoid of Plethodus oblongus carries a slender, elongate tooth plate extending entirely under the orbit. The inner margin of the entopterygoid also appears to support a narrow tooth plate with typical Plethodus histology and it is probable that the basihyal/t~asibranchial tooth plate occluded against both parasphenoid and entopterygoid. This type of occlusion is more similar to that of an albulid than that of an osteoglossomorph. Taverne (2000b) has recently revised the species of Plethodus, noting that the skull anatomy is completely different from Palaeonotopterus, and he considered that they belonged to at least three separate genera. As a result of his work Plethodus expansus Dixon is maintained within the genus Plethodus, Plethodus oblongus Dixon is referred to Dixonanogmius new genus and Plethodus pentagon Woodward is referred to a new genus Pentanogmius. In tur-n all three genera are placed with the Tselfatiiformes. Although we note the similarity of Dixonangmius oblongus with Anwnogmius (Banaizogmius) we also note that the latter genus is itself a sy stematic problem. Its relationships have been thought to lie with elopiforms (Nelson, 19731, osteoglossomorphs (Patterson, 1967) or clupeocephalans (Taverne, 2000a). Since Ananogmius is currently the subject of current revision (Taverne, submitted) we prefer not to comment

26 50 L. CAVIN and P. L. FOREY Figure 14. Articulated skulls of Dixonanogmius oblongus (Dixon) ip left lateral views. A, BOM B, BMNH P Scale bars = 20 mm. further at this point except to say that there may be at least two very different kinds of fishes bearing Plethodus-like tooth plates. We feel confident in associating Plethodus libycus with Palaeonotopterus as coming from the same kind of fish (even though we refrain from placing it in synonymy). SUMMARY Palaeonotopterus is an osteoglossomorph related as the primitive sister species to either notopteridae or mormyroids. The braincase anatomy shows features of both modern groups. With notopterids Palaeonotopterus shares an enlarged foramen for V +VII which straddles the prootic-pterosphenoid suture, an auditory fenestra, an inferred anterior process of the sagitta otolith. With primitive mormyroids Palaeonotopterus shares a characteristically shaped supraoccipital crest, a suture between the parasphenoid and the autosphenotic and possibly a broad medial lamina on the cleithrum. Palaeonotopterus shows an expanded parasphenoid tooth plate with a histology matching that of some Plethodus species. The overall morphology of the parasphenoid tooth plate is very similar to Plethodus libycus, a species known only by isolated tooth plates. The only other previously named species of Plethodus known by articulated cranial material - Plethodus oblongus (now Dixonanagmius oblongus) - looks very different from Palaeonotopterus and we agree with Taverne (2000~) that it is not closely related to Palaeonotopterus. ACKNOWLEDGEMENTS In the preparation of this manuscript we would like to thank Dr Louis Taverne, Universite Libre de Bruxelles, for discussion and his critical reading of our manuscript. We thank Mr William Lindsay (Natural History Museum, Palaeontological Conservation Unit) for preparing some of the specimens used in this paper and Mr Phil Crabb (Natural History Museum, Photostudio) for preparing most of the photographs. We extend our thanks to Ms Alison Longbottom (Natural History Museum) for help with preparing the illustrations. We thank Karsten E. Hartel (Museum of Comparative Zoology, Harvard) for the loan of comparative material. Mr John Cooper, Curator, Booth Museum of Natural History, Brighton, made the type material of Plethodus available and L.C. would like to thank the administrators of the Bioresource LSF Project for the opportunity to travel to London to carry out this joint work and the Federal Office for Education and Science (OFES) and the Swiss National Funds (FNS) for funding this work.

27 AFFINITIES OF rl GREENWOODI 51 REFERENCES.Arratia G, Schultze H-P Palatoquadrate and its ossifications: development and homology within osteichthyans. Journal of Morphology 208: 1-81, Bonde N Osteoglossids (Teleostei: Osteoglossomorpha) of the Mesozoic. Comments on their interrelationships. In: Arattia G, Viohl G, eds. Mesozoic Fishes - Systematics and Palaeoecology. Munchen: Friedrich Pfeil, Dkon F The Geology and Fossils of the Tertiary and Cretaceous Formations of Sussex. London: Richard & John Edward Taylor. llutheil DB An overview of the freshwater fish from the Kem Kem beds (late Cretaceous: Cenomanian) of southeast Morocco. In: Arratia G, Schultze H-P, eds. Mesozoic Fishes-Systematics and Fossil Record. Miinchen: Friedrich Pfeil, I'orey PL A revision of the elopiform fishes, fossil and Recent. Bulletin of the British Museum (Natural History), Geology (suppl. 10): Forey PL A Cretaceous notopterid (Pisces: Osteoglossomorpha) from Morocco. South African Journal of Science 93: Grande L Recent and fossil clupeomorph fishes with materials for revision of the subgroups of clupeoids. Bulletin of the American Museum of Natural History 181: Cirande L, Cavender TM Description and phylogenetic reassessment of the monotypic Ostariostomidae (Teleostei). Journal of Vertebrate Paleontology 11: Crreenwood PH The swimbladder in African Notopteridae (Pisces) and its bearing on the taxonomy of the family. Bulletin of the British Museum (Natural History), Zoology 11: Cireenwood PH Interrelationships of osteoglossomorphs. In: Greenwood PH, Miles RS, Patterson C, eds. Interrelationships of Fishes. London: Academic Press, Ch-eenwood PH, Patterson C A fossil osteoglossoid fish from Tanzania (E. Africa). Journal of the Linnean Society, Zoology 47: Gkeenwood PH, &sen DE, Weitzman SH, Myers GS Phyletic studies of teleostean fishes, with a provisional classification of living forms. Bulletin of the American Museum of Natural History 131: K.apoor AS Functional morphology of the laterosensory canals in the Notopteridae (Pisces). Acta Zoologica 45: Kershaw DR The cranial osteology of the 'Butterfly Fish', Pantodon buchholri Peters. Zoological Journal of the Linnean Society 49: Li G-Q New Osteoglossomorphs (Teleostei) from the Upper Cretaceous and Lower Tertiary of North America and their phylogenetic significance. Unpublished PhD thesis, University of Alberta, Canada. Li G-Q, Wilson MVH An Eocene species of Hiodon from Montana, its phylogenetic relationships, and the evolution of the postcranial skeleton in the Hiodontidae (Teleostei). Journal of Vertebrate Paleontolog?' 14: Li G-Q, Wilson MVH Phylogeny of Osteoglossomorpha. In: Stiassny MU, Parenti L, Johnson GD, eds. Interrelationships of Fishes. San Diego, California: Academic Press, Li G-Q, Wilson MVH Early divergence of Hiodontiformes sensu stricto in East Asia and phylogeny of some Late Mesozoic teleosts from China. In: Arratia G, Schultze H-P, eds. Mesozoic Fishes 2 - Systematics and Fossil Record. Miinchen: Freidrich Pfeil, Li G-Q, Grande L, Wilson MVH. 1997a. The species of TPhareodus (Teleostei: Osteoglossidne) from the Eocene of North America and their phylogenetic relationships. Journal of Vertebrate Paleontology 17: Li G-Q, Wilson MVH, Grande L. 1997b. Review of Eohiodon (Teleostei: Osteoglossomorpha) from western North America, with a phylogenetic reassessment of Hiodontidae. Journal of Paleontology 71: Maisey JG The supraotic bone in neopterygian fishes (Osteichthyes, Actinopterygii). American Museum hrouitates 3267: Nelson G Notes on the structure and relationships of certain Cretaceous and Eocene teleostean fishes. American Museum Novitates 2524: Nolf D Otolithi Piscium. In: Schultze H-P, ed. Handbook of Paleoichthyology. Stuttgart: Gustav Fischer, Nybelin Preliminary note on two species previously named Leptolepis bmnni Agassiz. Arkiv for Zoologi 15: Omarkhan M The lateral sensory canals of larval Notopterus. Proceedings of the Zoological Society of London 118: Orts S Contribution a l'anatorr~ie comparee et B la systematique des Monnyroldes. Acadkmie myale des Sciences d'outre-mer 3: Patterson C A second specimen of the Cretaceous teleost Pmtobrama and the relationship of the sub-order Tselfatioidei. Arkic for Zoologi 19: 21, Patterson C A clupeomorph fish from the Gault (Lower Cretaceous). Zoological Journal of the Linnean Society 49: Patterson C The braincase of pholidophorid and leptolepid fishes, with a review of the actinopterygian braincase. Philosophical Transactions of the Royal Society of London 269: Regan CT The classification of teleostean fishes. Annals and Magazine of Natural History, Series 8 7: Schaal S Oberkretazische Osteichthyes (Knochenfische) aus dem Bereich von Bahariya und Kharga, ;igypten, und ihre Aussagen zur Palijkologie und Stratigraphie. Berliner geouissenschaften ilbhandlungen A 53: Shen M Foss~l "osteoglossomorphs" from East Asia and their implications for teleostean phylogeny. In: Arattia G, Viohl G, eds. Mesozoic Fishes - Systematics and Paleoecolog3: Miinchen: Freidrich Pfeil, Singh Rana R Freshwater fish otoliths from the Deccan Trap associated sedimentary (Cretaceous-Tertiary

28 52 L. CAVIN and P. L. FOREY transition) Beds of Rangapur, Hyderabad District, Andhra Pradesh, India. Geobios 21: Taverne L ~tude osteologique des genres Boulengemmyrus Taverne et Gery Genyomyrus Boulenger Petmcephalus Marcusen (Pisces Mormyriformes). Annales, Musk Royal de I'Afrique Centrale, Serie in 80, Sciences Zoologiques 174: Taverne L Le complexe de la chambre trigeminofaciale des Teleosteens et le cas des poissons Mormyriformes. Revue Zoologique et Botanique Africaine 84: Taverne L Osteologie des genres Mormyrus Linne, Mormymps Muller, Hypempisus Gill, Isichthys Gill, Myomyrus Boulenger, Stomatorhinus Boulenger et Gymnarchus Cuvier. Considerations generales sur la systematique des poissons de I'ordre des Mormyriformes. Annales, Musee Royal de 1'Afrique Centrale, Serie in 8v0, Sciences Zoologiques 200: Taverne L Le connexion otophysaire de Gymnarchus (Monnyriformes) et de Papyrocranus (Osteoglossiformes) et la parente des Osteoglossomorphes et des Clupeomorphes. Etablissement d'une nouvelle systematique des poissons teleosteens. Revue Zoologique et Botanique Africaine 87: Taverne L Sur l'existence d'une chambre osseuse prootico-basioccipitale pour le sac vasculaire posthypophysaire chez les Notopteridae adultes (Pisces Osteoglossomorpha). Revue Zoologique Africaine 90: Taverne L Osteologie, phylogenese et systematique des teleosteens fossiles et actuels du super-ordre des osteoglossomorphes. Premiere partie. Osteologie des genres Hiodon, Eohiodon, Lycoptera, Osteoglossum, Sclempages, Heterotis et Arapaima. Memoires, Acadkmie Royale de Belgique, Classe des Sciences, Serie in 8uo 42: Taverne L Osteologie, phylogenese et systematique des Teleosteens fossiles et actuels du super-order des Osteoglossomorphes. Deuxieme partie. Osteologie des genres Phareodus, Phareoides, Brychaetus, Musperia, Pantodon, Singida, Notopterus, Xenomystus et Papymcranus. Memoires, Academie Royale de Belgique, Classe des Sciences, Serie in 8vo 42: Taverne L Osteologie, phylogenese et systematique des Teleosteens fossiles et actuels du super-order des Osteoglossomorphes. Troisieme partie. Evolution des structures osteologiques et conclusions generales relatives a la phylogenese et a la systematique du super-ordre. Addendum. Memoires, Acadkmie Royale de Belgique, Classe des Sciences, Serie in 8vo 43: Taverne L Les osteoglossomorphes marins de I'Eocene du Monte Bolca (Italie): Monopterus Volta 1796, Thrissopterus Heckel, 1856 et Foreyichthys Taverne Considerations sur la phylogenie des teleosteens osteoglossomorphes. Studi e Ricerche sui Giacimenti Terziari di Bolca 7: Taverne L. 2000a. Tselfatia formosa, teleosteen marin du Cretace (Pisces, Actinopterygii), et la position systematique des Tselfatiiformes ou Bananogmiiformes. Geodiversitas Taverne L. 2000b. Osteologie et position systematique du genre Plethodus et des nouveaux genres Dimnanogmius et Pentanogmius, poissions marins du Cretace (Teleostei, Tselfatiiformes). Biologisch Jaarboek Dodonaea 67: Taverne L. 2000~. Nouvelles donnees osteologique et phylogenetiques sur Palaeonotopterus greenwoodi, notopteride (Teleostei, Osteoglossomorpha) du Cenomanian inferieur continental (Cretace) du Maroc. Stuttgarter Beitrage zur Naturkunde B (Geologic und Palaontologie) 293: A Taverne L, Maisey JG A notopterid skull (Teleostei, Osteoglossomorpha) from the continental early Cretaceous of southern Morocco. American Museum Novitates 3260: Weiler W Ergebnisse der Forschungsreisen Prof. E. Stromers in den Wusten Agyptens. 11. Wirbeltierreste der Baharije-Stufe (unterstes Cenoman). 16. Neue Untersuchungen an den Fischresten. Abhandlungen der Bayerischen Akademis der Wissenschaften Mathematischnaturwissenschaftliche Abteilung, Neue Folge 32: Woodward AS On the Cretaceous fish Plethodus. Annals and Magazine of Natural History 3: Woodward AS Catalogue of the Fossil Fishes in the British Museum (Natural History). London: British Museum (Natural History). Woodward AS The Fossil Fishes of the English Chalk. Part 3. London: Palaeontographical Society. Zhang J-Y Morphology and phylogenetic relationships of TKuntulunia (Teleostei: Osteoglossomorpha). Journal of Vertebrate Paleontology 18: