Redescription of †Ellimma Branneri and †Diplomystus Shengliensis, and Relationships of Some Basal Clupeomorphs Author(S): MEE-MANN CHANG and JOHN G

Redescription of †Ellimma branneri and †Diplomystus shengliensis, and Relationships of Some Basal Clupeomorphs Author(s): MEE-MANN CHANG and JOHN G. MAISEY Source: American Museum Novitates, Number 3404:1-35. 2003. Published By: American Museum of Natural History DOI: http://dx.doi.org/10.1206/0003-0082(2003)404<0001:ROEBAD>2.0.CO;2 URL: http://www.bioone.org/doi/full/10.1206/0003-0082%282003%29404%3C0001%3AROEBAD %3E2.0.CO%3B2

BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research. PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, NY 10024 Number 3404, 35 pp., 13 ®gures, 2 tables May 22, 2003

Redescription of ²Ellimma branneri and ²Diplomystus shengliensis, and Relationships of Some Basal Clupeomorphs

MEE-MANN CHANG1 AND JOHN G. MAISEY2

ABSTRACT

Two extinct clupeomorphs, ²Ellimma branneri from the Cretaceous of Brazil and ²Diplo- mystus shengliensis from the Eocene of China, are redescribed. ²Ellimma branneri was for- merly classi®ed within the Clupeiformes, but it lacks derived characters of clupeiforms and clupeoids. Dorsal scute ``wings'' are expanded and subrectangular in ²Ellimma and other members of the family ²Paraclupeidae Chang and Chou (1977), approximately equal to ²El- limmichthyidae Grande (1982a). Consequently, ²Ellimma branneri is classi®ed here within the family ²Paraclupeidae. ²Paraclupeidae are known from the Lower Cretaceous to the middle Eocene. In the present work, two monophyletic groups are identi®ed within the ²Paraclupeidae. One group (subfamily ²Paraclupeinae of Chang and Grande, 1997), known only from the Lower Cretaceous (Hauterivian±Albian), includes ²Paraclupea,²Ellimmichthys, and ²Ellimma. These taxa are united by strongly sculptured, skull-roo®ng bones with ridges radiating from the growth center, and a dorsal scute ornament of prominent ridges. ²Scutatuspinosus may also belong in this group. The other group includes ²Diplomystus (Upper Cretaceous±Eocene) and ²Armigatus (Upper Cretaceous), which are united by a single homoplaseous character (presence of a posteriorly expanded third hypural, leaving no gap between hypurals 2 and 4): this character also occurs in pristigasteroids, ²Erichalcis, osteoglossids, some elopomorphs (²Le- banichthys lewisi, and most Albula spp.), and a number of ostariophysans not included in our analysis. ²Paraclupeines are customarily regarded as being more closely related to the Clu-

1 Research Associate, Division of Paleontology, American Museum of Natural History; Research Professor, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing 100044, People's Republic of China. e-mail: [email protected] 2 Curator and Axelrod Research Chair, Division of Paleontology, American Museum of Natural History. e-mail: [email protected]

peiformes than to other teleosts (i.e., as clupeomorphs), although no derived characters are uniquely shared by ²Ellimma branneri and modern Clupeiformes. The relationships of ²Ellim- ma and certain other extinct herring-like teleosts (including other ²paraclupeines) with the Clupeiformes are unclear, and they may collectively form a paraphyletic assemblage. No biogeographical hypothesis satisfactorily explains the known distribution of nonmarine ²paraclupeine ®shes in the Cretaceous. A substantial portion of their nonmarine fossil record is missing (as evidenced by the recent discovery of a possible ²paraclupeine, ²Ezkutuberezi carmeni Poyato-Ariza et al., 2000, in Spain), and some aspects of their early distribution pattern may have involved marine dispersal. Eocene ²Diplomystus occurs on both sides of the Paci®c Ocean, but the ``Paci®ca'' hypothesis (which lacks empirical support) is abandoned as an explanation for such Eocene (and younger) trans-Paci®c distribution patterns of nonmarine ®shes. Instead, a ``freshwater Arctic Ocean'' hypothesis is favored. According to this hypothesis (for which there are several independent lines of geological evidence), temporary desalination of the Arctic Ocean occurred during the Paleocene and early Eocene, which may have permitted freshwater ®shes to move unimpeded by salt-water barriers between Asia and North America; this temporary desalination event may eventually become recognized as a signi®cant factor in the holarctic distribution patterns of various Tertiary-Recent freshwater ®shes.

INTRODUCTION pramaxillae and a complex pattern of sculpture on the surface of its dorsal scutes. He The primary purpose of our study is to re- restored the generic name ²Ellimma and re- examine ²Ellimma branneri, an Early Cre- ferred it to the family Clupeidae, subfamily taceous (Upper Aptian) clupeomorph from Clupeinae. Riacho Doce, in the Sergipe Basin of Ala- Our further preparation and observation of goas, Brazil. The original material was col- the original specimens of ²E. branneri, from lected by J.C. Branner in 1907 and was de- both the American Museum and Carnegie posited in the Carnegie Museum, Pittsburgh. Museum collections, has provided new in- It formed the basis for Jordan's (1910) description under the name Ellipes, where it formation supporting the view that ²Ellimma was referred to the family Clupeidae. Unfor- is a distinct taxon, but also suggesting that it tunately the name Ellipes is preoccupied by does not belong in the Clupeidae. Instead, we a genus of orthopteran (Scudder, 1902), re- propose that it is a more primitive clupeo- quiring that another name be created for the morph, closely related to three other Early ®sh taxon (²Ellimma; Jordan, 1913). Cretaceous taxa, ²Ellimmichthys longicosta- ²E. branneri was restudied by Schaeffer tus (Cope, 1886), ²E. goodi (Eastman, 1912), (1947) using additional material collected by and ²Paraclupea chetungensis Sun (1956). Euphrasio Borges in 1917±1918. In 1931 Eu- The collective geographical distribution of zebio Paulo de Oliveira donated this material these taxa is of considerable interest. ²Ellim- to the Department of Vertebrate Paleontology michthys longicostatus is from the Mar®m at the American Museum of Natural History Formation (late Hauterivian±early Barremi- from the Geological Survey of Brazil. an, RecoÃncavo Basin) of Brazil (Maisey, Schaeffer (1947) transferred the species to 2000), and ²Ellimmichthys goodi is from Ap- the clupeid genus ²Knightia, effectively tian±Albian strata of Equatorial Guinea, West making ²Ellimma a synonym of that genus. Africa (Eastman, 1912), while ²Paraclupea His proposal implied a signi®cant strati- chetungensis is from the Lower Cretaceous graphic range extension for ²Knightia, as the of southeastern China (Sun, 1956). A close genus had previously been recognized only relationship between ²Ellimmichthys and in the Eocene deposits of western North ²Paraclupea was established by Chang and America (Jordan, 1907). Schaeffer's proposal Grande (1997), who erected the subfamily was subsequently reversed by Grande ²Paraclupeinae for these two genera, within (1982a, 1982b, 1985a), who excluded ²E. the family ²Paraclupeidae of Chang and branneri from the genus ²Knightia, noting Chou (1977). The family was de®ned by a that ²Ellimma can be distinguished from oth- single character (dorsal scutes broader than er clupeid taxa by the presence of two su- long), which also occurs in the genus ²Di- 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 3 plomytus. Grande (1982a) had previously re- Cretaceous, Aptian±Albian?) from the Cabo lated ²Ellimmichthys and ²Diplomystus with- Basin of northeastern Brazil. This form was in his family ²Ellimmichthyidae on the basis identi®ed originally as ²Ellimmichthys lon- of a similar character (``lateral wings of the gicostatus (Costa et al., 1979), but it was dorsal scute elongated and blunted at the lat- subsequently referred to ²Ellimma and eral edges, giving the scute a subrectangular named ²E. cruzi by Silva Santos (1990). The outline''), and Chang and Grande (1997) also same horizon has produced a small gonor- included ²Diplomystus within the ²Paraclu- hynchiform which was referred to ²Dastilbe. peidae. The Cretaceous and Eocene paraclupeids The pattern of the dorsal scutes in ²Ellim- each present a puzzling biogeographic pat- ma branneri only partially agrees with this tern for freshwater ®shes. The Early Creta- family-level character, because the anterior ceous ``southern transatlantic'' pattern (i.e., dorsal scutes are longer than broad in this spe- restricted to western Gondwana, and repre- cies, whereas the posterior ones are broader sented by ²Ellimmichthys longicostatus and than long. Inclusion of this form within the ²Ellimma branneri from northeastern Brazil ²Paraclupeidae on the basis of scute shape is and ²Ellimmichthys goodi from West Africa) therefore not straightforward and necessitates is now extended into southeastern China a reevaluation of characters supporting para- (²Paraclupea chetungensis) and will perhaps clupeid monophyly. In searching for charac- be expanded farther by the new Spanish dis- ters other than scute shape, we addressed wid- covery (Poyato-Ariza et al., 2000). The Eo- er issues such as the relationships between cene forms have a ``transpaci®c'' distribu- these taxa and other basal clupeomorphs, in- tion, with the two sister species (²Diplomys- cluding ²Santanaclupea,²Armigatus,²Diplo- tus shengliensis and ²D. dentatus) on oppo- mystus, and ²Knightia. site sides of the northern Paci®c. Possible A redescription of the Chinese species of origins of this pattern are reexamined below ²Diplomystus, that is, ²D. shengliensis in the light of recent discoveries concerning Zhang et al. (1985), is also incorporated in the history of the Arctic region in the early this paper. According to its original descrip- Cenozoic. tion (in Chinese), ²Diplomystus shengliensis is an Eocene teleost ®rst collected from dril- ABBREVIATIONS ling cores of the Sheng-li Oil Field situated on the southern coast of Bohai Gulf along Anatomical the western coast of the Paci®c Ocean. The AA angulo-articular species is very similar to ²Diplomystus den- br.r branchiostegal rays tatus from the Eocene Green River Forma- BT? basibranchial toothplate? tion of North America, and hardly any dif- Ch ceratohyal ferences exist between them except for a few Cs caudal scute minor meristic characters (Zhang et al., D dentary Enpt entopterygoid 1985), a remarkable situation considering the Ep epural widely separated occurrences of these spe- Fr frontal cies on opposite sides of the Paci®c. H3 hypural 3 Besides the taxa considered here, there are Hm hyomandibular a few other records of Cretaceous ``double- Io2,Io3 infraorbital 2, infraorbital 3 armoured'' clupeomorphs, such as ²Scuta- Iop interopercle tuspinosus itapagipensis from the Neocomi- msc mandibular sensory canal an of Brazil (Silva Santos and Correa, 1985), Mx maxilla an undescribed species of ²Ellimmichthys Op opercle from the Tlayua Formation of southern Mex- P parasphenoid Pa parietal ico (Chang and Grande, 1997: ®g. 7e, f), and Pd predorsal bones ²Ezkutuberezi carmeni from Spain, said to be Ph parhypural related to paraclupeids (Poyato-Ariza et al., Pmx premaxilla 2000). There are also records of a supposed Pop preopercle paraclupeid in the Cabo Formation (Lower Psp parasphenoid 4 AMERICAN MUSEUM NOVITATES NO. 3404

Pt posttemporal observe without some cleaning of the mate- Pu1 preural centrum 1 rial. Fortunately the bituminous matrix has a Q quadrate very low carbonate content, and so for the S symplectic present study we were able to use dilute HCl Sc sclerotic bone Smxa anterior supramaxilla to dissolve and clean broken bone from the Smxp posterior supramaxilla matrix in order to obtain clean impressions Soc supraoccipital of the skeleton. Latex peels were made from

U1 ural centrum 1 these cleaned impressions, and both reveal Ur urohyal much more detail then unprepared speci-

Un1 uroneural 1 mens.

Institutional SYSTEMATICS AMNH American Museum of Natural History, SUBCOHORT CLUPEOMORPHA GREENWOOD New York ET AL., 1966 CM Carnegie Museum of Natural History, Pittsburgh, PA ORDER ²ELLIMMICHTHYIFORMES GRANDE, FMNH Field Museum of Natural History, Chi- 1982 cago FAMILY ²PARACLUPEIDAE CHANG AND CHOU, IVPP Institute of Vertebrate Paleontology 1977 and Paleoanthropology, Beijing SOF Shengli Oil Field, Dongying, Shandong ϭ ²ELLIMMICHTHYIDAE GRANDE, 1982 Province, China SUBFAMILY ²PARACLUPEINAE CHANG AND GRANDE, 1997 MATERIALS AND METHODS ²ELLIMMA JORDAN, 1913 This study primarily involves specimens of ²Ellimma branneri from the collections of ²Diplomystus Cope, 1877: 811 (in part). the AMNH and CM, and of ²Diplomystus ²Knightia Jordan, 1907: 136. shengliensis from the Shengli Oil Field. ²Ellipes Jordan, 1910: 24 (Brazilian species only). ²Ellimma Jordan, 1913: 79 (new name for Ellipes Specimens of ²Paraclupea chetungensis Jordan, 1910, preoccupied). from the IVPP, as well as those of ²Armi- ²Ellimma, Jordan and Gilbert, 1919: 26. gatus brevissimus,²Diplomystus dentatus,a ²Knightia, Schaeffer, 1947: 17. few species of ²Knightia, plus dried skele- ²Ellimma, Grande, 1982a: 22. tons and cleared-and-stained specimens of ²Ellimma, Grande, 1982b: 13. Clupea harengus and Onchorhynchus mykiss ²Ellimma, Grande, 1985a: 250. from the AMNH, were compared, as were some specimens of ²Diplomystus dentatus EMENDED DIAGNOIS: Paraclupeine with from the FMNH. comparatively low body depth. Surface of The specimens of ²E. branneri examined opercle with striations. Teeth on jaws and are all preserved in bituminous shales of the palate very weakly developed. Anterior dor- Muribeca Formation and are in a laterally sal scutes slightly longer than broad and or- compressed and ¯attened state. In all these namented with ridges; posterior ones broader specimens the braincase is badly crushed, than long and ornamented with tubercles or and their study was further impaired by bro- tubercles plus ridges. Sharp spine extending ken surfaces and weathering of exposed sur- from median keel of posterior few dorsal faces, which frequently makes it dif®cult to scutes pointing posteriorly rather than pos- examine features such as sutures between terodorsally. Origin of pelvic ®n opposite to bones and other structures. The poor state of middle point of base of dorsal ®n. Procurrent preservation also helps explain why features rays 5 and 6 on each side of base of caudal such as dorsal scute ornamentation, the struc- ®n. ture of the caudal skeleton, and various other TYPE SPECIES:²Ellipes branneri Jordan, anatomical details have not been described 1910, the only species (based on page pri- previously, for they are almost impossible to ority). 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 5

²Ellimma branneri (Jordan), 1910 namentation, dorsal scute morphology, and Figures 1±8 entopterygoid teeth, that are not discernible in the holotype of ²E. branneri. There is in- ²Ellimma branneri Jordan, 1910: 25, pl. 8, ®g. 3. evitably some circularity in our inclusion of ²Ellimma riacensis Jordan, 1910: 28, pl. 10. ²Ellimma branneri Jordan, 1913: 79. these features in our description of ²E. bran- ²Knightia branneri, Schaeffer, 1947: 17. neri, but we found no evidence to retain ²El- ²Ellimma branneri, Grande, 1982a: 22; 1982b: limma riacensis as a distinct species, and this 13; 1985a: 250. taxon is regarded here as a subjective synonym of ²E. branneri. Thus, we conclude HOLOTYPE: CM 5249/1 and CM 5249/2. there is only one species of ²Ellimma rep- Note that only CM 5249/1 was designated resented in the Riaco Doce material, rather the type by Schaeffer (1947), and this was than two sympatric species. also the type specimen chosen by Jordan for ²Ellimma branneri. In fact, CM 5249/2 is the DESCRIPTION counterpart to CM 5249/1, although this was not noticed in previous descriptions. The ho- GENERAL SHAPE: The overall outline of ²E. lotype is a small ®sh, with a standard length branneri resembles that of ²Paraclupea che- of 38.5 mm (measured from the counterpart tungensis (Chang and Chou, 1977; Chang CM 5249/2, which is more complete than and Grande, 1997), with a markedly convex CM 5249/1). ventral outline and a gently curved dorsal ADDITIONAL MATERIAL: AMNH 10046± outline rising to an apex at the origin of the 10062; CM 5248/10, 11, 21, 36, 39; 5249/2 dorsal ®n (®gs. 1A, B, 2A±C). The maxi- (counterpart of 5249/1), 3, 25, 27, 33, 49, 52, mum body depth occurs at the origin of the 54, 55, 56 (counterpart of 5248/11), 125, 130 dorsal ®n. The standard length of individuals (counterpart of 5249/125), 175 (counterpart in our samples varies from 20.8 (CM 5249/ of 5249/52). 125) to 102.2 mm (CM 5248/4); in other DIAGNOSIS: As for genus. words, our largest specimens are ®ve times HORIZON AND LOCALITIES: From Riacho longer than the smallest. The proportion be- Doce, Sergipe Basin, Brazil (Muribeca For- tween the body depth and the standard length mation, Lower Cretaceous, Aptian±Albian, is therefore somewhat variable, usually being ϭ ``Alagoan'' local stage). The age of this approximately 41±44% standard length, but unit is well established from palynomorphs rarely 36 or 50%. and foraminifers (FeijoÂ, 1994). SKULL ROOF: The bones of the braincase COMMENTS: The type specimen of ²Ellim- in ²E. branneri are usually so badly crushed ma riacensis (CM 5248/4) has a standard that it is impossible to recognize bone mar- length of 102.2 mm and is therefore a some- gins or even various recesses, fossae, and what larger individual than the type of ²E. bullae within the bones; importantly, it has branneri. It is also much better preserved, not been possible to determine whether a re- and it consequently shows more morpholog- cessus lateralis was present (an important ical details than the type of ²E. branneri.We clupeiform character absent in primitive clu- were however unable to identify any un- peomorphs; Grande, 1985a). Few features of equivocally unique features of ²Ellimma ri- the braincase can be described. In the ante- acensis, and we consequently agree with rior part of the skull roof, the posterior por- Schaeffer (1947) that ²Ellimma riacensis is tion of the dermethmoid (seen in CM 5249/ a synonym of ²E. branneri. We regard ²E. 49) bears a lateral process on both sides and branneri as the type species of the genus on a pair of processes extending posteriorly the basis of page priority. Our observations above the frontals. In larger specimens the led us to the conclusion that all the ²Ellimma posterior one-third of the frontals, as well as specimens studied by Schaeffer indeed rep- the parietals, are strongly ornamented with resent a single species, but we must also ad- ridges radiating from their growth centers. mit that the holotype of ²E. riacensis (CM The dorsal part of the supraoccipital (seen in 5248/4) provides some morphological infor- CM 5249/56) is small and triangular and mation, including details of the skull roof or- shows indications of ridges radiating from 6 AMERICAN MUSEUM NOVITATES NO. 3404

Fig. 1. ²Ellimma branneri, Muribeca Formation, Riaco Doce, Sergipe Basin, Brazil. (A) CM 5248/ 4; (B) CM 5249/2, the holotype. Both scale bars ϭ 1 cm. the posteromedian point of the bone antero- rietal. The supratemporal commissure passes laterally. In smaller individuals (e.g., CM through the parietals. 5248/39, standard length approximately 51 A comparatively broad and ¯at ridge runs mm) the skull roof bones are smooth and the length of the posttemporal along its mid- without sculpture (®g. 3). Unlike in ²Para- line, and its broader posterior part is orna- clupea chetungensis there is no anterior fon- mented with ridges in larger individuals. The tanelle between the frontals. The parietals dorsal (epiotic) limb of the posttemporal is meet at the midline and are not separated by long and narrow anteriorly but becomes the supraoccipital (in clupeiforms the supra- broader posteriorly, while the ventral limb of occipital separates the parietals; Grande, this bone is thin. The lateral line sensory ca- 1985a). As far as can be determined the su- nal runs diagonally through the supracleith- praoccipital crest is small and low (CM rum. 5249/56, 5248/11, and 5248/39). The supra- ORBITAL REGION: Faint impressions of orbital sensory canal is enclosed in a bony three infraorbital bones can be seen in a few ridge and extends from the frontal to the pa- specimens, for example, AMNH 10048 (®g. 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 7

Fig. 2. ²Ellimma branneri, Muribeca Formation, Riaco Doce, Sergipe Basin, Brazil. Silicone peels prepared from AMNH specimens. (A) AMNH 10048; (B) AMNH 10057; (C) AMNH 10060. 8 AMERICAN MUSEUM NOVITATES NO. 3404

Fig. 3. ²Ellimma branneri, CM 5248/39 to show skull roof.

2A) and 10050. Two detached infraorbitals coronoid process is present on the dentary. (perhaps the second and third) are exposed The mandibular sensory canal runs above the on CM 5249/52 and 5249/175 (part and lower margin of both the dentary and the an- counterpart; ®g. 4). The ®rst two infraorbitals gulo-articular (®g. 5B). The maxilla is long seem to be narrow and elongated, but the and extends behind the jaw joint. In most third is broader and shorter. The infraorbital specimens of ²Ellimma branneri the jaw sensory canal runs along the orbital margin teeth are not visible, but a few ®ne, conical of all three bones. Parts of the sclerotic ring teeth are recognizable on the oral margin of were seen in several specimens, although no the premaxilla and dentary in the latex peel details were evident. of AMNH 10057. Very ®ne serrations are PARASPHENOID AND ENTOPTERYGOID:No also present on the oral margin in one de- teeth are present on the parasphenoid. A bas- tached maxilla (CM 5248/11 and CM5249/ ipterygoid process is recognizable on the la- 56, part and counterpart; ®g. 6C, D). tex cast of AMNH 10051. In most specimens Collectively, these observations suggest of ²Ellimma branneri no teeth were ob- that the entire jaw dentition of ²Ellimma served on the entopterygoid, although faint branneri consists only of very small teeth, traces of tiny broken tips of teeth are some- times present (e.g., AMNH 10047, 10048 and is much reduced in comparison with and CM 5248/4). In contrast, ²Ellimmichthys both ²Ellimmichthys longicostatus and ²Par- and ²Paraclupea have an entopterygoid den- aclupea chetungensis, where a well-devel- tition that is relatively better developed, and oped dentition is present both on the jaws the reduced dentition in ²Ellimma is a distin- and the palatal surface of the entopterygoid guishing feature of the genus within paraclu- (Chang and Chou, 1977; Chang and Grande, peids. 1997). The presence of very small entopter- JAWS:In²Ellimma branneri the articula- ygoid teeth in ²Ellimma branneri may be re- tion of the lower jaw is positioned more or lated to a microphagous diet, as in many Re- less below the posterior margin of the orbit cent clupeomorphs. ²Ellimma branneri pos- (®gs. 2A±C, 5B). The dentary, angulo-artic- sesses two supramaxillae (a primitive tele- ular, and retroarticular are seen clearly in ostean state), of which the posterior one is AMNH 10048 (®g. 5A). A well-developed larger. Both supramaxillae are ornamented by 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 9

Fig. 4. ²Ellimma branneri, to show detached opercle, subopercle (at bottom) and two small infraorbitals. (A) CM 5249/52 (part) and (B) CM 5249/175 (counterpart).

®ne shallow grooves on their external sur- which contains three or four branches of the face. preopercular sensory canal (®g. 5B). OPERCULAR SERIES AND HYPOBRANCHIAL The hyomandibular shaft is narrow, with APPARATUS: A detached but well-preserved a single head and no anterodorsal process. opercle with an almost complete margin is The quadrate is thick, with the symplectic present in CM 5249/52 and 175. It is deep inserted in a notch above its posteroventral and rectangular in shape, with a protruding margin, as in teleosts generally (®g. 5A, B). anteroventral corner. The ventral part of the The anterior ceratohyal is rectangular in opercle in most specimens of ²Ellimma shape, longer than deep, and contains an branneri is ornamented on its lateral surface elongated oval bericiform foramen (AMNH by striations which radiate ventrally from the 10048, 10050; ®g. 5A). This opening is tra- area of attachment (®g. 4A, B), but the sur- versed by a groove for the hyoidean artery face of the opercle in the smallest specimen on the outer face of the anterior ceratohyal. is smooth, with no striations (CM 5249/125 Absence of the bericiform foramen is con- and 130, standard length 20.8 mm). The su- sidered to be a synapomorphic character of bopercle shows the anterior ascending pro- clupeiforms by Grande (1985a). There are cess which is usually covered by the opercle approximately 10 branchiostegal rays (CM 5249/52; ®g. 4A). The preopercle has (AMNH 10048; ®g. 5A). An outline of the two arms forming an obtuse angle; its verti- urohyal was detected on AMNH 10048, but cal arm is longer than the horizontal one, few details are discernible because the bone 10 AMERICAN MUSEUM NOVITATES NO. 3404

Fig. 5. ²Ellimma branneri, silicone peels showing detail of the head in (A) AMNH 10048; (B) AMNH 10057. Both scale bars ϭ 1 cm. 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 11 (C, D) CM 5248/11, both to same scale; (B) CM 5249/56 and its counterpart (A) , detached jaw elements in Ellimma branneri Fig. 6. ² enlarged details of maxilla in CM 5249/56, showing the ®nely serrated oral margin. 12 AMERICAN MUSEUM NOVITATES NO. 3404

Fig. 7. ²Ellimma branneri. Detail of dorsal scutes in (A) AMNH 10048 and (B) AMNH 10057 (both from silicone peels). Anterior scutes are better preserved in A, whilst the posterior scutes are clearer in B. Ornamentation of anterior scutes is mostly ridges, whereas the posterior scutes have ridges and tubercles. Both scale bars ϭ 1 mm.

lies beneath the branchiostegal rays; it is nar- dal. There are eight predorsal bones with thin row and triangular, with a long, narrow an- anterior and posterior bony expansions (®g. terior process. 5A, B). VERTEBRAL COLUMN AND FINS: Two lon- In ²Ellimma branneri the dorsal and anal gitudinal ridges are present along the lateral ®ns are of more or less equal size (®gs. 1, side of each vertebra (®g. 2). The total ver- 2). The dorsal ®n contains 2 unbranched and tebral count in ²Ellimma branneri could only 14 branched rays, and is supported by 14± be estimated because the anteriormost centra 16 pterygiophores. In both AMNH 10048 are covered by the opercle in all the available and 10050 the anal ®n has 15 rays and 14± specimens. We estimate that the vertebral 15 pterygiophores. The pelvic ®n is small column originally comprised 36±38 centra, and contains about six or seven rays (seven of which 20±22 are abdominal (based on the are present in CM 5249/27). Its insertion is number of pairs of ribs) and 15±16 are cau- approximately opposite to the middle point 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 13

Fig. 8. ²Ellimma branneri caudal ®n skeleton, silicone peels. (A) AMNH 10048; (B) AMNH 10060. Both scale bars ϭ 1 cm. 14 AMERICAN MUSEUM NOVITATES NO. 3404 of the base of the dorsal ®n. The pectoral ®n smaller than in ²Ellimmichthys longicostatus is large but does not reach the origin of the (see Jordan, 1910: pl. XI). pelvic, and has at least 12 ®n rays. The SCALES: The scales are thin, with ®ne con- cleithrum is large and curved. No postcleith- centric growth lines, although no vertical cir- ra have been recognized in the available culi were observed like those in ²Diplomys- specimens. tus and many other clupeiforms. From the SCUTES:²Ellimma branneri possesses a material available it could not be determined complete dorsal scute series (unlike in ²Ar- whether lateral line scales are present. migatus), extending from the back of the CAUDAL SKELETON AND TAIL (FIG. 8A, B): head to the origin of the dorsal ®n (®gs. 1A, The structure of the caudal skeleton in ²E. 2A, C, 7). The entire series includes 12±14 branneri closely resembles that of ²Ellim- scutes, each of which is overlapped anteri- michthys and ²Paraclupea. Collectively, all orly by the next (AMNH 10048, 10060, and these taxa have a caudal ®n morphology that CM 5249/2). The more anterior scutes are differs from the clupeiform condition. There generally rounded in shape and are slightly are two free ural centra, the ®rst ural centrum longer than broad (like those illustrated by is approximately equal in size to the ®rst Grande, 1982a: ®g. 23), but those farther preural centrum, and there are six hypurals. posteriorly tend to be broader than long. The The ®rst hypural is in close contact with (but last three or four scutes in the dorsal series not fused to) the ®rst ural centrum, although are expanded laterally and are much broader the second hypural is fused to this centrum. than those farther anteriorly. All the dorsal The third hypural is comparatively narrow, scutes in ²Ellimma branneri have a median not expanded posteriorly, and there is a gap keel, but scute margins are not always well between the second and third hypurals. The preserved, and it is therefore uncertain parhypural is fused with the ®rst preural cen- whether the keel of every scute was extended trum, which has a short neural arch. As many as a spine (as in ²Ellimmichthys and ²Par- as three epurals are present (e.g., AMNH aclupea; Chang and Grande, 1997). The last 10060), although in most specimens only one few scutes in ²Ellimma branneri are extend- or two are visible. There are three uroneurals, ed into a sharp spine pointing posteriorly. all of which are autogenous. The ®rst uro- This arrangement is slightly different from neural is long and thick, extending to the that seen in ²Ellimmichthys and ²Paraclu- posterodorsolateral side of the second preural pea, where the spine projects from the scute centrum, but the second and third uroneurals in a more posterodorsal direction. In ²E. are much smaller. branneri most of the scutes are ornamented The caudal ®n is deeply forked (®gs. 1, 2, with ridges, which arise laterally from the 8). Its upper lobe contains one unbranched posterior end of the keel and then turn pos- and nine branched rays, and its lower lobe teriorly to pass almost parallel to the keel. has one unbranched and eight branched rays. The ornamentation of the posteriormost The proximal end of the lowermost branched scutes is rather different, however, as some ray from the upper lobe is bifurcated into up- are covered with tubercles arranged in rows, per and lower branches, as is the uppermost and others have both tubercles and ridges branched ray from the lower lobe. In both (®g. 7A, B). Almost the entire upper surface cases, the upper branches are longer, broader, of each scute is ornamented in ²E. branneri; and more ¯attened than the lower ones. by contrast, in ²Paraclupea chetungensis, There are ®ve or six procurrent rays on both only the posterior half or third of the scute the upper and lower margins of the peduncle upper surface is ornamented. in front of the caudal ®n, and one caudal The abdominal scutes in ²E. branneri scute is present on the upper margin of the form a continuous series extending from the peduncle anterior to the procurrent rays. lower end of the cleithrum to just in front of ²PARACLUPEIDAE INCERTAE SEDIS the origin of the anal ®n (®g. 2A±C). There are 27±30 abdominal scutes, of which 8±10 ²Diplomystus Cope, 1877 lie behind the origin of the pelvic ®n. The EMENDED DIAGNOSIS: Differs from other abdominal scutes behind the pelvic ®n are genera of ²Paraclupeidae in having a high 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 15 number of dorsal scutes; posterior border of holotype SOF 790001 (®g. 9), which fortu- dorsal scutes pectinate; supraoccipital crest itously lies entirely within the diameter of a very well developed and high; posttemporal drilling core. This example is nevertheless with long and slender epiotic branch; entop- somewhat distorted in the abdominal region, terygoid with robust teeth along dorsal mar- and its scale cover is missing. Data from this gin of buccal surface; seven or eight predor- specimen are supplemented by SOF 790002, sal bones; H3 much expanded posteriorly. No in which the anterior portion of vertebral col- gap between H2 and H3, lowermost ray of umn and ribs is better preserved; and SOF upper caudal ®n lobe bifurcates proximally, 790003, which has a well preserved caudal with two branches of more or less equal size, skeleton and ®n rays. Some skeletal features uppermost ray of lower caudal ®n lobe cannot be observed in any of the available slightly bent and enlarged. material, and our description is therefore in- TYPE SPECIES:²Diplomystus dentatus complete. Cope, 1877 The total length of the holotype of ²D. OTHER INCLUDED SPECIES:²D. shengliensis shengliensis (SOF 790001, the only complete Zhang et al., 1985; ²D. birdi Woodward, specimen) is 5.5 cm. The other fragmentary 1895; ²D. dubertreti Signeux, 1951. specimens evidently represent larger individuals, although their original size is uncertain. ²Diplomystus shengliensis Zhang et al., In the holotype, the body is elongate and fu- 1985 siform, reaching a maximum depth (approx- Figures 9±12 imately 40% of standard length) a short way behind the head (®g. 9). The dorsal margin HOLOTYPE: SOF 790001, ®rst illustrated in of the body curves upward a short distance Zhang et al., 1985: pl. 1, ®g. 1 (see ®gs. 9, behind the head, and then descends gradually 10A, 11, 12A here). to the caudal peduncle, without an abrupt an- ADDITIONAL MATERIAL: SOF 790002 (®g. gle at the origin of the dorsal ®n. The ventral 10B), 790003 (®g. 12B). outline is moderately convex. HORIZON AND LOCALITIES: Top of series 4 SKULL ROOF: The head can be studied in to bottom of series 3 of the Shahejie For- the holotype and in SOF 790002 (®g. 10). It mation, Middle Eocene; SOF 790001 and is not easy to determine the arrangement of 790002 from Bore Lai 38±8, depth from the the bones along the midline of the skull roof ground surface 2628m and 2668m, respec- in ²D. shengliensis because the specimens tively, SOF 790003 from Bore Lai 1±9, are all laterally compressed. The supraoccip- depth from the ground surface 2622 m, Kenli ital is situated far posteriorly and probably county, Shandong Province, East China. did not separate the parietals, which therefore EMENDED DIAGNOSIS:²Diplomystus with may have met at the dorsal midline. The ex- elongate fusiform body, depth and standard ternal surfaces of the frontals and parietals length ratio of about 40%; with number of are generally smooth, apart from ridges con- dorsal scutes (41) and dorsal ®n rays (1, 13) taining the sensory canals. The supraorbital slightly higher than in ²D. dentatus (highest sensory canal extends from the frontal into number of dorsal scutes 40 and that of dorsal the parietal, where it meets a prominent ®n rays 1, 12), but number of ribs (15 pairs) curved ridge which appears to contain part and number of vertebrae (around 42), namely of the supratemporal commissure. The supra- number of abdominal vertebrae (around 17), occipital crest is well-developed, being high lower than in the latter (17±18 pairs of ribs and triangular in shape (®g. 10B), but it can- and 20 abdominal vertebrae). not be determined whether this bone con- ETYMOLOGY: sheng-li-, transliteration of tained part of the supratemporal commissure. the Chinese word ``victory'', the species The posttemporal has a long and slender epi- named after Sheng-li Oil Field. otic limb (®g. 11). ORBITAL REGION: Nothing is preserved ex- DESCRIPTION cept part of the sclerotic ring (SOF 790002; GENERAL SHAPE: The only more or less ®g. 10B). complete specimen of ²D. shengliensis is the PARASPHENOID AND ENTOPTERYGOID:No 16 AMERICAN MUSEUM NOVITATES NO. 3404 . Lower Cretaceous, Sheng-li Oil Field, China. The holotype, SOF 790001. Diplomystus shengliensis Fig. 9. ² 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 17

Fig. 10. ²Diplomystus shengliensis. Detail of the head in (A) the holotype, SOF 790001; and (B) SOF 790002. Both scale bars ϭ 1cm. 18 AMERICAN MUSEUM NOVITATES NO. 3404

Fig. 11. ²Diplomystus shengliensis. Stereophotographs showing the predorsal bones and dorsal scute series in the holotype, SOF 790001 (specimen oriented vertically). teeth were observed on the parasphenoid, Zhang et al. (1985). The buccal surface of and it is apparently edentulous in ²D. shen- the entopterygoid bears numerous teeth, of gliensis. A basipterygoid process is present which the posterodorsal ones adjacent to the in the holotype SOF 790001 (®g, 10A), al- parasphenoid are much stouter than the rest. though no features are descernible. In this JAWS: The dentary and premaxilla bear specimen, the left entopterygoid is observed small conical teeth, and the oral margin of in situ against the parasphenoid margin, the maxilla is ®nely serrated. There are two while the right entopterygoid has its buccal supramaxillae, but unlike in ²Paraclupea side turned outward and its posterior end dis- and ²Ellimma these bones both have smooth placed downward. This bone therefore oc- surfaces. cupies an oblique position, and it was mis- OPERCULAR SERIES AND HYPOBRANCHIAL takenly identi®ed as the ectopterygoid by APPARATUS (FIG. 10A, B): The bones of the 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 19

Fig. 12. ²Diplomystus shengliensis. Detail of caudal ®n skeleton in (A) the holotype, SOF 790001; and (B) SOF 790003. Both to same scale. 20 AMERICAN MUSEUM NOVITATES NO. 3404 opercular series are smooth in ²D. shenglien- and the second is fused to it. The third hy- sis. The vertical branch of the preopercle is pural ®lls the entire space between the sec- much longer than the horizontal branch. In ond and the fourth hypurals, unlike in ²Par- SOF 790001 there is a bone covered with aclupea,²Ellimma, and ²Ellimmichthys, stout teeth, in occlusion with the left entop- where there is a gap. The parhypural in ²D. terygoid; we interpret this as the basibran- shengliensis is apparently fused with the ®rst chial bone and its associated toothplate. preural. The latter carries a short neural arch VERTEBRAL COLUMN AND FINS: The verte- protruding backward into a very sharp spine. bral column is fairly well preserved in the There are three epurals and three uroneurals holotype of ²D. shengliensis, but the exact in ²D. shengliensis, the ®rst of which extends number of abdominal vertebrae cannot be de- anteriorly to the second preural. termined because the anterior portion of the The caudal ®n is deeply forked in ²D. vertebral column is distorted (®g. 9). We es- shengliensis, and the lower lobe is slightly timate that there were 42 vertebrae, 25 of longer than the upper. There are 19 principal which are caudal. In SOF 790002 (where the ®n rays (as in primitive teleosts generally), anterior portion of the vertebral column is of which the 2 lateral ones are unbranched. better preserved) there are 17 abdominal ver- The proximal end of the lowermost ray from tebrae and 15 paired ribs. There are seven the upper lobe bifurcates into two short predorsal bones on the holotype of ²D. shen- branches of approximately equal size. The gliensis, with both anterior and posterior corresponding end of the uppermost ray of bony expansions (®g. 11). the lower lobe is slightly bent and spatulate. The dorsal ®n of ²D. shengliensis contains one unbranched and 13 branched ®n rays PHYLOGENETIC ANALYSIS (®g. 9). The anal ®n is very long and contains at least 39 ®n rays. The pterygophores The data we assembled (see table 1 and of both dorsal and anal ®ns are badly pre- appendix 1) are mainly based on characters served in all the available specimens. from Grande (1985a), Forey (1975), Patter- SCUTES: The dorsal scute series in the ho- son and Rosen (1977), Maisey (1993), Ar- lotype, SOF 790001 is shown in ®gure 11. ratia (1996), and Chang and Grande (1997), As the stereophotographs illustrate, the dor- plus a few new observations. Our phyloge- sal scutes of ²D. shengliensis are much netic analysis involved 30 osteological char- broader than long, with a prominent median acters and 11 taxa, 7 of which are extinct keel and a pectinate posterior margin, but the (²Armigatus,²Diplomystus,²Ellimma,²El- scute surface is smooth and lacks ornamen- limmichthys,²Knightia,²Paraclupea and tation. There are 41dorsal scutes on the ho- ²Santanaclupea). Two Recent clupeomorphs lotype of ²D. shengliensis, but the number of were included (Clupea and Denticeps), abdominal scutes is uncertain because of which, together with ²Knightia, are represen- poor preservation. These scutes are quite tatives of either Clupeoidei or Denticepito- similar in shape to those of other clupei- idei. Two other Recent taxa (Elops and On- forms, and as in ²D. dentatus the ventral part corhynchus) represent successive outgroups of the posterior border is pectinate in the an- (Elopomorpha and Euteleostei, respectively). teriormost few scutes. Taxa only known from very incomplete fos- SCALES: Scale morphology cannot be deter- sils such as ²Spratticeps (Patterson, 1970) mined in any of the specimens of ²D. sheng- are not included in our analysis because data liensis, but detached scales associated with are too limited. Characters for ²Diplomystus them resemble those referred to ²Knightia bo- are taken from the type species ²D. dentatus haiensis by Zhang et al. (1985: ®g. 16). and ²D. shengliensis. The Late Cretaceous CAUDAL SKELETON AND TAIL: Only parts of species ²D. birdi Woodward and ²D. dub- the caudal skeleton and ®n are preserved in ertreti Signeux need further investigation be- the holotype, SOF 790001, but more can be fore they can be meaningfully included in a seen in SOF 790003 (®g. 12). There are sev- phylogenetic analysis. en hypurals in ²D. shengliensis, of which the The dataset was analyzed by using Swof- ®rst is in contact with the ®rst ural centrum ford's (2000) Phylogenetic Analysis Using 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 21

TABLE 1 form two monophyletic sister groups. Our re- Data Matrix sults differ from his in three respects: (1) the position of ²Armigatus; according to Grande (1985a), this forms an unresolved trichotomy with two monophyletic groups (²Ellimmich- thyiformes, Clupeiformes), whereas here it is grouped with ²Diplomystus; (2) ²Paraclupea and ²Ellimma were not included in Grande's (1985a) analysis, although ²Paraclupea was later placed in the subfamily ²Paraclupeinae along with ²Ellimmichthys by Chang and Grande (1997); (3) ²Santanaclupea (a clupeomorph from Santana Formation, North- east Brazil; Maisey, 1993) is also included in our analysis and appears to be closer to Clu- peiformes than to ²Paraclupeidae (ϭ ²Ellim- michthyiformes). It may form a sister taxon Parsimony (PAUP Version 4.0). All charac- to Clupeoidei, represented by ²Clupea and ters are unordered and unweighted. Our anal- ²Knightia in our analysis (®g. 13A), or it ysis (using the branch-and-bound search) may be a sister taxon to Clupeiformes, rep- generated two equally short trees (®g. 13A, resented by (²Clupea ϩ ²Knightia) ϩ ²Den- B), each 45 steps long and with a consistency ticeps (®g. 13B). index (excluding uninformative characters) Despite some uncertainties both in the data of 0.6585, a homoplasy index (excluding un- and in our phylogenetic analysis, the present informative characters) of 0.3111, and a re- study con®rms the monophyly of Clupeo- tention index of 0.7255 (with rescaled con- morpha and also its subdivisions into two sistency index of 0.4998). The two trees major groups: (1) ²Paraclupeidae (ϭ ²Ellim- agree in recognizing two groups within the michthyiformes) including (²Armigatus ϩ Clupeomorpha. One consists of (²Armigatus ²Diplomystus) ϩ ²Paraclupeinae; and (2) ϩ ²Diplomystus) ϩ (²Ellimma ϩ ²Ellimmi- Clupeiformes. Our analysis also supports chthys ϩ ²Paraclupea), which corresponds monophyly of the Denticipitoidei and Clu- approximately to the ²Paraclupeidae of peoidei within the Clupeiformes, as proposed Chang and Grande (1997) or ²Ellimmi- by Grande (1982a, 1985a). The Clupeomor- chthyiformes (Grande, 1982a). The other pha, Clupeiformes and Clupeoidei are well group comprises (Clupea ϩ ²Knightia) ϩ supported by shared derived characters (e.g., Denticeps and is equivalent to Clupeiformes. Clupeomorpha by characters 6, 16, 20, 22, In both trees ²Ellimma,²Ellimmichthys, and node 1 in ®g. 13A, B; Clupeiformes by char- ²Paraclupea collectively form a monophy- acters 2, 4, 25, node 5; and Clupeoidei by letic group, but their interrelationships are characters 1, 5, 7, 18, 20, 24, 28, 29, node unresolved. This group is referred to below 7). The monophyly of the ²Paraclupeidae as the subfamily ²Paraclupeinae. Differences (²Ellimmichthyiformes) is less robust, being between the two most parsimonious trees supported only by characters with low con- mainly involve the position of ²Santanaclu- sistency (i.e., characters which are shared pea. In tree 1 (®g. 13A), ²Santanaclupea is with taxa outside the group, or which are grouped with ²Clupea ϩ ²Knightia (i.e., lacking in some of the in-group taxa, 11, 12, Clupeoidei), but in tree 2 (®g. 13B) ²Santan- 14, node 2). In Grande's (1985a) analysis, aclupea is the closest extinct stem taxon to ²Armigatus, ²Ellimmichthyiformes (includ- Clupeiformes. ing ²Diplomystus), and Clupeiformes formed The overall grouping of taxa within these an unresolved trichotomy. In our analysis trees largely agrees with the phylogeny pro- this trichotomy is resolved so that ²Armiga- posed by Grande (1982a: ®g. 20; 1985a: tus and ²Diplomystus form a sister pair (node ®g.1A) in which ²Diplomystus ϩ ²Ellimmi- 3 in ®g. 13A, B). Among the characters unit- chthys and Denticipitoidei ϩ Clupeoidei ing these taxa (3, 9, 10, 26) only 26 (pres- 22 AMERICAN MUSEUM NOVITATES NO. 3404

Fig. 13. Results of the phylogenetic analysis discussed in the text, with the two most parsimonious trees generated by our characters. (A) ²Santanaclupea is grouped with Clupeoidei. Characters supporting nodes: 1 (6, 16, 20, 22); 2(11, 12, 14); 3(3, 9, 10, 26); 4 (15); 5 (2, 4, 25); 6 (3, 27); 7 (1, 5, 7, 18, 20, 24, 28, 29). (B) ²Santanaclupea is grouped more generally with Clupeiformes. Characters supporting nodes: 1 (3, 6, 16, 20, 22); 2 (11, 12, 14); 3 (9, 10, 26); 4 (15); 5 (2, 4, 25); 6 (1); 7(5, 7, 18, 20, 24, 27, 28, 29). 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 23 ence of a posteriorlyexpanded third hypural local stage), and ²Ellimmichthys from the leaving no gap between hypurals 2 and 4) is sedimentary sequence in Bahia (Mar®m For- uniquely shared by two of the 11 taxa under mation) is even older (late Hauterivian±early our analysis. Among clupeoids it also occurs Barremian) (for references, see Maisey, in pristigasteroids (Grande, 1985a). Our prin- 2000). cipal ®ndings are discussed further below. Grande (1982a, 1982b, 1985a) restored the generic name ²Ellimma and removed ²E. DISCUSSION branneri from the genus ²Knightia on morphological grounds: (1) ²Ellimma has two PHYLOGENETIC COMMENTS. Jordan (1910: supramaxillae, whereas ²Knightia has only 23) noted that the herringlike ®shes he rec- one; and (2) the dorsal scutes of ²Ellimma ognized among fossils collected in 1907 by show a complex pattern of sculpture, where- John Casper Branner from Riaco Doce, Bra- as in ²Knightia they do not. Grande (1982a, zil ``. . . have the general traits of ²Diplo- 1985a) also resurrected the genus ²Ellim- mystus dentatus, with the short anal and few- michthys, again on morphological grounds, er vertebrae of ²Knightia, while at the same and thus he essentially restored the system- time their squamatiom seems to be different atic distinctions recognized by Jordan (1910, from both ...''. 1913), albeit within a far more comprehen- Schaeffer (1947) placed the genus ²Ellim- sive and re®ned phylogenetic framework for ma into synonymy with ²Knightia, which clupeomorph ®shes. In Grande's (1985a) considerably extended the stratigraphic range phylogenetic analysis ²E. branneri is not of the latter (until then it had only been dis- closely related to ²Ellimmichthys or ²Diplo- covered in Eocene strata of western North mystus, and is placed instead within Clupei- America; Jordan, 1907). He also placed ²El- dae (i.e., within Clupeiformes, rather than limmichthys into synonymy with ²Diplomys- ²Ellimmichthyiformes); as such, ²Ellimma tus, despite admitting there were similarites would be the only clupeine having dorsal between ²Ellimma branneri and ²Diplomys- scutes. tus in the ovate shape of the dorsal scutes Unfortunately, there is little evidence in and the absence of scute lateral wings (Cope, support of that conclusion. Grande's (1985a) 1877), as well as similarities in the size and single morphological character supporting shape of body scales. Schaeffer (1947) also monophyly of the Clupeidae (presence of concurred with Jordan (1910) that the dorsal two long, rodlike postcleithra; Grande, scutes of ²Knightia branneri differ from 1985a: character 23) has not been identi®ed those of North American ²Knightia in being in ²E. branneri. Clupeids are nested within wider and in lacking a posterior median the Clupeoidea, whose single synapomorphy spine. (increase in pleural rib to preural vertebrae It is possible that such morphological dif- ratio; Grande, 1985a: character 22) is cer- ferences, however slight, might have been tainly present in ²E. branneri, but is also a given greater emphasis had it been realized feature of ²Ellimmichthys (where it was re- that the Brazilian taxa were in fact far older garded as convergent by Grande, 1985a: than those from North America. Jordan 263). ²Ellimma branneri also lacks three of (1910) commented that the age of the Riacho the derived characters of clupeoids (fusion of Doce shales is ``. . . Lower Eocene, possibly the ®rst uroneural with the ®rst preural cen- but not probably Upper Cretaceous'', and trum, reduction in relative size of the ®rst Schaeffer (1947) also considered them to be ural centrum, separation of the parhypural Eocene. ²Ellimmichthys from Bahia was from the ®rst preural centrum; Grande, nevertheless considered to be of Upper Cre- 1985a: characters 14, 15 17) and is unknown taceous age by both Jordan (1910) and with respect to the fourth (absence of lateral Schaeffer (1947). In fact, the fossils from line scales; Grande, 1985a: character 16). Fi- both Alagoas and Bahia are Lower Creta- nally, ²Ellimma branneri lacks two of the ceous in age; ²Ellimma from the Riacho three synapomorphies shared by the Clupei- Doce shales of Alagoas (Muribeca Forma- formes (parietals separated by supraoccipital, tion) dates from the Aptian±Albian (Alagoan beryciform foramen lost; Grande, 1985a: 24 AMERICAN MUSEUM NOVITATES NO. 3404 characters 10, 11) and the third (presence of in ²Ellimmichthys and ²Paraclupea the or- a recessus lateralis, Grande, 1985a: character namentation of all dorsal scutes consists only 9) is unknown. of ridges. Additionally, the jaw and palatal Thus, none of Grande's (1985a) characters dentition of ²Ellimma is reduced in compar- support the inclusion of ²Ellimma within the ison with ²Ellimmichthys and ²Paraclupea. Clupeidae, Clupeoidei, or Clupeiformes, and The origin of the pelvic ®n also differs. In its inclusion within Clupeoidea is only sup- ²Ellimma it is opposite to the middle point ported by a single ambiguous character of the dorsal ®n base, whereas in ²Ellim- which could be convergent with clupeoids, michthys longicostatus it is in advance of the as in ²Ellimmichthys. Grande (1985a) found dorsal ®n, and in ²Ellimmichthys goodi the a single character to unite his ²Ellim- pelvic ®n inserts opposite to the posterior michthyidae (lateral expansion of dorsal third of the base of the dorsal ®n; in ²Par- scute ``wings'' which give scute a subrect- aclupea chetungensis the pelvic ®n lies be- angular shape; Grande, 1985a: character 7; hind the origin of the dorsal ®n. There are this feature is also shared with ²Diplomys- also fewer procurrent rays on each side of tus). In ²Ellimmichthys and ²Paraclupea, all the peduncle (approximately ®ve in ²Ellim- the dorsal scutes are broader than long, as ma;in²Ellimmichthys and ²Paraclupea are the posterior dorsal scutes in ²Ellimma there are eight or nine rays on the upper side branneri (although its anterior dorsal scutes and four to six on the lower side). These dif- are slightly longer than broad); thus, lateral ferences suggest that ²Ellimma is a distinct expansion of dorsal scutes remains a syna- genus from ²Ellimmichthys and ²Paraclu- pomorphy of ² Paraclupeidae (ϭ ²Ellimmi- pea. chthyidae), including ²Ellimma, but the con- Observed meristic characters of ²Ellimma dition is clearly variable. In addition, ²Ellim- branneri agree more closely with those of the ma shares several generalized clupeomorph two species of ²Ellimmichthys than with characters with ²Ellimmichthys and ²Para- ²Paraclupea (see table 2) The number of clupea, such as presence of abdominal and vertebrae, dorsal ®n rays, dorsal scutes and dorsal scutes, enclosure of the supratemporal abdominal scutes in ²Ellimma are respective- commissure within the parietals, and fusion ly, 36±38, ii, 12±14, 12±14, and 27±30. of the second hypural with the ®rst ural cen- Those in ²Ellimmichthys are 35±38, 12±15, trum, although these similarities do not help 12±13, and 28±32. In ²Paraclupea the num- elucidate its relationship within clupeo- ber is noticeably higher (41, 17±18, 18, and morphs. 38±43). Furthermore, in ²Ellimma and ²El- Two derived characters are shared by ²El- limmichthys almost the entire surface of each limma,²Ellimmichthys, and ²Paraclupea and dorsal scute is ornamented, whereas in ²Par- distinguish them collectively from ²Diplo- aclupea only the posterior third or half is or- mystus: strongly sculptured skull roo®ng namented. bones, with ridges radiating from the growth ²Ellimna guineensis from Equatorial center, and dorsal scute ornament of promi- Guinea, West Africa (apparently misspelled nent ridges. It is concluded from this that as ``²Ellimna''; Gayet, 1989), was referred ²Ellimma branneri (Jordan, 1910) is closely to Clupeidae and has been considered close related to ²Ellimmichthys longicostatus to ²Diplomystus goodi. Grande (1982a) pre- (Cope, 1886), from the Hauterivian±Barre- viously noted that the characters of ²''Dip- miam of Brazil, ²E. goodi (Eastman, 1912), lomystus'' goodi mentioned by Taverne from the late Aptian of Equatorial Guinea, (1975) are all primitive for clupeomorphs, and ²Paraclupea chetungensis Sun, 1956, however, and do not clearly place this taxon from the Lower Cretaceous of southeastern in ²Diplomystus; he consequently removed it China. All these taxa are grouped here in the from ²Diplomystus and placed it in Clupeo- subfamily ²Paraclupeinae. morpha, incertae sedis. Gayet (1989) regard- ²Ellimma can be distinguished at generic ed two characters (presence of two supra- level from these other taxa by the ornamen- maxillae and ornamented dorsal scutes) as tation on the last few dorsal scutes, where diagnostic of the genus ²Ellimna, but the ®rst there are tubercles (or tubercles plus ridges); is a primitive teleost feature and the other is 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 25

TABLE 2 Comparison of Characters Between ²Diplomystus shengliensis and ²D. dentatus 26 AMERICAN MUSEUM NOVITATES NO. 3404 present in ²Ellimmichthys and ²Paraclupea. these three ²Diplomystus species by the pres- Chang and Grande (1997) considered that ence of dorsal scutes with a pectinate pos- both ²Ellimna guineensis and ²''Diplom- terior border, more numerous dorsal scutes ystus'' goodi should be placed within ²Ellim- (22±36) than in ²Ellimmichthys, and reduc- michthys. tion or loss of a median recess in the poste- Based on the original description by Silva rior edge of the scute (present in ²Ellimmi- Santos and Correa (1985), ²Scutatuspinosus chthys). Grande (1982a) regarded the two itapagipensis displays some features sug- Cretaceous species from Lebanon as closer gesting paraclupeine af®nity. Its dorsal scutes to each other than to ²D. dentatus, because are progressively larger and wider caudally they both have more dorsal ®n rays and few- (as in ²Ellimmichthys,²Paraclupea, and er vertebrae. ²Ellimma). They are also ornamented with Zhang et al. (1985) noted that ²D. shen- ridges (a paraclupeine character), but lack tu- gliensis is very similar to ²Diplomystus den- bercles (as in ²Ellimmichthys and ²Paraclu- tatus from the Green River shales. In fact, pea). It is unknown whether the skull roo®ng they are so similar that it is dif®cult to dis- bones are sculptured with ridges, as in par- tinguish them as separate species, despite aclupeines. The origin of the pelvic ®n is an- their widely separated occurrences today on terior to the dorsal ®n in ²Scutatuspinosus opposite sides of the Paci®c. A list of char- itapagipensis,asin²Ellimmichthys longicos- acters comparing ²D. shengliensis and ²D. tatus, not beneath it, as in ²Ellimma bran- dentatus is provided in table 2. Despite the neri. Unfortunately few of the described fea- small sample size for ²D. shengliensis (with tures in this taxon are phylogenetically in- merely three specimens), information is lack- formative, and it would bene®t from a revi- ing in only 4 out of 32 characters (data are sion, but our observations suggest that missing for the pelvic ®n, the number of ab- ²Scutatuspinosus is also a paraclupeine. Its dominal scutes, and scale morphology). caudally-expanded dorsal scute series is an ²Diplomystus shengliensis resembles ²D. unusual feature and may represent a syna- dentatus in 22 of the 28 characters that can pomorphy with ²Ellimma. According to Sil- be compared. The ratio of depth/standard va Santos and Correa (1985), ²Scutatuspi- length is slightly greater in ²D. shengliensis nosus itapagipensis has 32 vertebral centra (40%) than that in ²D. dentatus (35±37%), and 12 dorsal ®n rays (comparable with par- but this may be a taphonomic rather than aclupeines other than ²Paraclupea itself), 10 morphological distinction related to defor- dorsal scutes, and 25 ventral scutes (the low- mation of the ventral portion of the body in est number of each for any paraclupeine). available specimens of ²D. shengliensis. As far as ²Diplomystus is concerned, al- There are 41 dorsal scutes in ²D. shengli- though several nominal North American spe- gensis, only one more than the maximum cies have been recognized, it has long been number observed in ²D. dentatus, although suspected that their taxonomic diversity is the majority of specimens have fewer (be- exaggerated. Woodward (1901) suggested tween 33 and 36; Grande, 1982a). There are that ²D. analis,²D. pectorosus, and ²D. the- 13 branched rays in the dorsal ®n of ²D. ta were synonyms of ²D. dentatus, while shengligensis, again one more than the max- Eastman (1912) pointed out that species of imum count in ²D. dentatus. An estimated ²Diplomystus from the Green River were al- 42 vertebrae are present in ²D. shengligensis, most indistinguishable. In his revision of only one fewer than the minimum count ²Diplomystus, Grande (1982a) also conclud- from ²D. dentatus (most individuals have ed that the only valid species from the Green more than 44). Greater apparent differences River shales is ²D. dentatus, and that the oth- are found in the number of abdominal ver- ers are all junior synonyms of that taxon. tebrae (17 in ²D. shengligensis,20in²D. Two other species were retained by Grande dentatus) and ribs (15 pairs in ²D. shengli- (1982a) in the genus ²Diplomystus:²D. birdi gensis, 17±18 in ²D. dentatus. Given the ex- and ²D. dubertreti from the Upper Creta- tent to which these features vary in ²D. den- ceous of Lebanon (Woodward, 1895; Sig- tatus, and the very small available sample of neux, 1951). Grande (1982a, 1985a) united ²D. shengliensis, the meristic distinctions be- 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 27 tween them are best regarded as provisional. ²Diplomystus brevissimus (Patterson, 1967; Whether ²D. shengliensis is a distinct spe- now ²Armigatus, Grande, 1982a). cies, however, the presence of ²Diplomystus The relationships between the two Recent in China during the Eocene is con®rmed by clupeiform suborders (Denticepitoidei, Clu- the present observations. peioidei) and some putative clupeomorph The morphological similarity of ²Diplo- fossils have been investigated by Forey mystus dentatus and ²D. shengliensis is fur- (1975) and Grande (1982a, 1985a), while the ther emphasized by comparison with the two problem of clupeomorph relationships gen- Lebanese Cretaceous species ²D. birdi and erally within teleosts has been discussed by ²D. dubertreti, both of which have dorsal Patterson and Rosen (1977) and Arratia scutes like those of ²D. dentatus in their (1996). ²Diplomystus,²Armigatus, and ²El- adult morphology and presumed ontogenetic limmichthys all seem to be close extinct rel- development, but which differ from ²D. den- atives of modern clupeomorphs (i.e., Clupei- tatus in having many more dorsal ®n rays formes sensu Grande, 1985a; in his phylog- and far fewer dorsal scutes, abdominal eny there is an unresolved trichotomy be- scutes, anal ®n rays, vertebrae, and ribs tween ²Armigatus, the group ²Diplomystus (Grande, 1982a). There is also a gap between ϩ ²Ellimmichthys, and Clupeiformes). There the second and third hypurals in the two Cre- are differences of opinion regarding the in- taceous species, unlike in ²D. dentatus and clusion of certain other extinct taxa within ²D. shengliensis. Some other potentially use- clupeomorphs. ²Ornategulum was included ful systematic features, such as the supraoc- in clupeomorphs by Forey (1975), Patterson cipital crest and entopterygoid dentition, are and Rosen (1977), and Grande (1982a), but unknown in the Lebanese species. While was excluded by Grande (1985a). ²Erichal- both species agree in most respects with the cis was placed in the Clupeomorpha by For- generic diagnosis of ²Diplomystus presented ey (1975) and Grande (1985a), but was re- salmo- earlier, their relationships are left unresolved garded as the sister taxon of Esox ϩ nids by Arratia (1996). Part of the uncertain- and require further analysis. ty is due to the incomplete preservation of According to Greenwood et al. (1966), the material, and in the case of ²Erichalcis Patterson (1970), and Nelson (1973), the the type series may include more than one modern Clupeomorpha are monophyletic and taxon (Mark Wilson and Lance Grande, per- share derived features of the ear-swimbladder sonal commun., 2000). ²Santanaclupea connection, the recessus lateralis, the cranial shares at least two synapomorphies with clu- morphology, the caudal skeleton, the abdom- peiforms, but cannot be placed unambigu- inal scutes, and the structure of gill arches ously within any of the modern clupeiform (see also Nelson, 1967, 1970). With the in- groups (Maisey, 1993). clusion of extinct taxa (especially at basal In Grande's (1985a) revision of Recent levels), however, monophyly of the Clupeo- and fossil clupeomorphs, ²Diplomystus and morpha becomes increasingly tenuous. Forey ²Ellimmichthys form a monophyletic group (1975) conducted a comprehensive analysis (order ²Ellimmichthyiformes) supported by a of clupeomorph characters while studying single character (dorsal scutes broader than ²Erichalcis, and paid particular attention to long or of rectangular shape). Clearly this the morphology of the recessus lateralis, the character also unites ²Ellimma branneri with position and size of the dermosphenotic, and these genera, although it is developed only the sensory canal arrangement around the re- in the posterior part of the scute series in this cessus. He found that presence of an upper form. division of the levator arcus palatini muscle Chang and Grande (1997) referred ²Par- in clupeoids can be correlated with presence aclupea chetungensis to an equivalent mono- of a ridge on the frontal and an anterodorsal phyletic group (family Paraclupeidae ϭ El- process on the hyomandibular. Both features limmichthyidae). They concluded that ²Par- are present in ²Knightia, leading Forey aclupea is more closely related to ²Ellim- (1975) to include it in clupeoids, but they are michthys than to ²Diplomystus, and should absent in ²Denticeps,²Ornategulum, and be included with ²Ellimmichthys in the sub- 28 AMERICAN MUSEUM NOVITATES NO. 3404 family ²Paraclupeinae. Our investigation of According to our analysis, paraclupeines and ²Ellimma branneri from Brazil suggests that ²Diplomystus are closely related (as mem- it is also a paraclupeine, in which case the bers of the family ²Paraclupeidae), but their ornamentation of dorsal scutes becomes a respective distribution patterns are funda- synapomorphy of the subfamily. ²Ellimma, mentally different and may be the result of ²Ellimmichthys, and ²Paraclupea form an quite unrelated biogeographic histories. In unresolved trichotomy in our phylogenetic both cases the fossil record is undoubtedly analysis, but meristic data for ²Ellimma and incomplete, and the known distributions of ²Ellimmichthys are similar whereas those for these forms may be so strongly biased as to ²Paraclupea are different. Our conclusions make them biogeographically unintelligible. regarding ²Diplomystus shengliensis agree We restrict our remarks to a few salient with those of Chang and Chow (1978) and points that may be worthy of further inves- Zhang et al. (1985) that it is most closely tigation. related to ²D. dentatus from North America. BIOGEOGRAPHIC OBSERVATIONS. All the Our analysis (®g. 13) supports the main genera included here within the ²Paraclupei- conclusions reached by previous workers, es- nae (²Ellimmichthys,²Ellimma,²Paraclu- pecially those of Grande (1985a), although pea, and perhaps ²Scutatospinosus) are Early some aspects of our results require critical Cretaceous in age. ²Ellimmichthys and ²El- comment. First, our morphological evidence limma both occur in Brazil, but ²Ellim- that the ²Ellimmichthyiformes form a mono- michthys is also known from Africa and phyletic group is not as strong as we would Mexico. The two better known species of wish. Second, some extinct ``double-ar- ²Ellimmichthys are ²E. longicostatus (Hau- mored'' clupeomorph taxa (e.g., ²Scutatos- terivian±Barremian, Mar®m Formation, Re- pinosus,²Ezkutuberezi carmeni) were ex- coÃncavo Basin, Brazil) and ²E. goodi (late cluded from our analysis, because we were Aptian±Albian, West Africa). The genus ²El- unable to examine material or retrieve suf®- limma contains three nominal species, all cient data from the published descriptions; from Brazil: ²E. branneri and ²E. riacensis their inclusion in future phylogenetic analy- (considered synonymous; Schaeffer, 1947; ses is clearly desirable and may provide a Grande, 1985a) from the late Aptian (Muri- useful test of the conclusions reached here. beca Formation, Sergipe Basin; Jordan, Third, although the trichotomy formed by 1910); and ²E. cruzi from the Aptian±early ²Armigatus, ²Ellimmichthyiformes, and Clu- Albian (Cabo Formation, Cabo Basin; Silva peiformes in Grande's (1985a) cladogram Santos, 1990; previously identi®ed as ²El- seems to be resolved here, our grouping of limmichthys by Costa et al., 1979). The un- ²Armigatus and ²Diplomystus is actually described Albian species of ²Ellimmichthys supported by ambiguous characters (3, 9, 10, from Mexico mentioned by Chang and and 26 in our character list). Even the most Grande (1997) represents a western Tethyan convincing synapomorphy (a posteriorly ex- (Caribbean) rather than Gondwanan occur- panded third hypural, leaving no gap be- rence, and is the only marine paraclupeine tween hypurals 2 and 4) also occurs in some record. Another double-armored clupeo- clupeiforms not included in our analysis. For morph, ²Scutatospinosus itapagipensis, oc- all these reasons the phylogenetic hypothesis curs along with ²Ellimma branneri in the presented here is considered tentative, as it Mar®m Formation of Brazil (Silva Santos re¯ects current uncertainties concerning par- and Correa, 1985). aclupeine relationships. Today there is little doubt that the margins Two interesting paleobiogeographical puz- of Brazil and West Africa were contiguous zles involving China have emerged from this during the Early Cretaceous (Pitman et al., work. First, an apparent sister-group relation- 1993). The earliest record of ²Ellimmichthys ship has emerged between ²Paraclupea and (and consequently the minimum postulated Early Cretaceous paraclupeines of western age of divergence between ²Ellimma and Gondwanan. Second, the presence of very ²Ellimmichthys) is Hauterivian±Barremian, similar ²Diplomystus fossils in China and which predates the formation of an equatorial North America during the Eocene is curious. seaway by a considerable margin, although 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 29

²Ellimmichthys goodi from Africa and all re- 1980; Yabumoto, 1994) is rather similar to cords of ²Ellimma are Aptian or younger the Chinese form and may extend the range (approximately contemporary with perma- of ²Paraclupea. These discoveries consid- nent emplacement of the seaway in the late erably expand the distribution of paraclu- Aptian; Maisey, 2000). All Gondwanan par- peines beyond western Gondwana and near- aclupeines are from strata deposited within by western Tethys, far into eastern Asia. At rift basins that were involved in the ®nal tec- this stage we can only speculate about the tonic separation of Africa and South Amer- overall paraclupeine distribution pattern in ica. At that time these basins were intracon- the Cretaceous. tinental, but subsequent rifting, crustal exten- There is little geologic evidence to support sion, and drifting transformed them into parts an Early Cretaceous nonmarine paleogeo- of the new continental margins. Their distri- graphic connection between the eastern Asi- bution pattern differs from those of some atic margin and western Gondwana that other Early Cretaceous ®shes from western would account for the observed distribution Gondwana, such as ²Mawsonia and ²Cala- of paraclupeine ®shes. We could plead for mopleurus, which occur more widely across more widespread distribution and earlier oc- interior localities of Africa and Brazil (in- currences of freshwater paraclupeine than is cluding nonrift settings), although these oc- presently realized, especially since the ²Di- currences probably also represent former re- plomystus ϩ ²Armigatus clade represents the gions of endemism which were disrupted by sister group to paraclupeines and is presum- formation of the Equatorial Atlantic seaway ably of equal antiquity (in which case there (Maisey, 2000). should be relatives of that clade in the Bar- The habitats and history of Gondwanan remian±Hauterivian). Another possibility is paraclupeine ®shes were undoubtedly affect- that paraclupeines may have marine origins ed by the early (predrift) phases of these mo- and their distribution may re¯ect a largely mentous tectonic processes. The distribution unknown marine history. Unfortunately, both of Gondwanan ²Ellimmichthys (and ²Scuta- scenarios involve sweeping ad hoc assump- tuspinosus?) may re¯ect vicariant isolation of tions, for example that the remains of a wide- lacustrine populations along the tectonically spread fauna are still to be found elsewhere, active RecoÃncavo-Tucano-JatobaÂ and Gabon- or that their traces in other areas have been Sergipe-Alagoas (GSA) rift trends (Maisey, obliterated by later geological events. 2000), and ²Ellimma may be an endemic ²Diplomystus shengliensis from Bohai South American genus which perhaps Gulf and ²D. dentatus from Green River are evolved as a result of vicariant isolation morphologically so similar that there is little within lakes of the GSA trend (Sergipe and doubt they represent closely related sister Cabo basins). Albian occurrences of ²Ellim- species. Interestingly, besides ²Diplomystus, michthys are restricted to freshwater deposits four other genera of ®shes (²Eohiodon, within African marginal rift basins, but have ²Knightia,²Amyzon, and Esox) supposedly not been documented from the marine strata are shared by the Green River and Bohai overlying them. The genus has nevertheless Gulf faunas (Chang and Chow, 1986; Chang been identi®ed in marine sediments from and Zhou, 1993, 2002; Chang et al., 2001), Mexico, suggesting that a salt-tolerant form and several other lineages from the Bohai reached western (Caribbean) Tethys by the Gulf deposits are also represented in the Albian±Cenomanian. Green River ichthyofaunas (e.g., ``Dasyati- The distribution of ²Paraclupea is a bio- dae'', Amiidae). As pointed out by Grande geographic conundrum (Chang and Chow, and Bemis (1998: 337), the Chinese amiid is 1986; Chang and Chen, 2000). ²Paraclupea still inadequately known and cannot yet be is not known from any Gondwanan locality; assigned either to Amia or ²Cyclurus, al- ²P. chetungensis is found in freshwater de- though it is almost certainly an amiine. ²Di- posits near the western Paci®c coast of Chi- plomystus,²Eohiodon,²Knightia,²Amyzon na, but an Early Cretaceous freshwater par- and Esox have an Eocene transpaci®c distri- aclupeid from Japan (described as ²Diplo- bution pattern, but amiines and dasyatids are mystus; Uyeno, 1979; Uyeno and Yabumoto, more broadly distributed. 30 AMERICAN MUSEUM NOVITATES NO. 3404

Interestingly, similar faunal distribution between ²Diplomystus from China and North patterns have been noticed among Eocene America. It would also offer a partial expla- terrestrial vertebrates of Asia and North nation for some Eocene ®sh distribution pat- America. These have been used to suggest terns, and perhaps even the holarctic distri- that a land bridge existed between these areas bution of some modern ®shes (e.g., paddle- in the vicinity of the Bering Straits, permit- ®shes). ²Eohiodon falcatus occurs in the ting mammals and other tetrapods to pass Green River Formation, and hiodontids are freely between these regions (Russell and represented in eastern China and East Kha- Zhai, 1987). McKenna (1975, 1980, 1983, zakstan (Central Asia). Eocene catostomids 1984) has also proposed that arctic connec- also occur in both these areas and are widely tions between the continents have played a spread in the vast plains area of East Asia, signi®cant role in such dispersals. from Mongolia in the north to Guangdong Grande (1985b, 1989, 1994) noted this re- Province at the very south of China (Chang peating pattern of area relationship in the ear- et al., 2001). The osteoglossid ²Phareodus ly Tertiary between western North America occurs in the Green River Formation, as well and east Asia, based on the phylogenetic re- as in Australia (²Phareodus queenslandicus; lationships of teleosts from Green River and Hills, 1934; Li, 1994). Eocene osteoglossids the west Paci®c region (including China, In- are found in Sichuan Province (²Sinoglossus, donesia, and Australia). He characterized this Su, 1986) and Hubei provinces, South China distribution as a ``transpaci®c'' pattern that (Song, in prep.) and Eocene or Oligocene of could be explained by the Paci®ca hypothesis Indonesia (²Musperia radiata; Sanders, (Grande, 1994). While we do not question 1934). Eocene hiodontids, catostomids, and these observations, note that much of the osteoglossids are also reported from British similarity in ichthyofaunas is restricted to the Columbia (Wilson, 1977). The wide boreal Eocene fossil assemblages of Bohai Gulf and distribution of amiines during the Eocene Green River, whereas other areas on the (including North America, northern China, western side of the Paci®c share far fewer East Khazakstan, and Europe) is remarkably taxa with Green River (known examples in- similar to the pan-Arctic distribution of many clude osteoglossids of Australia and Indo- riparian and terrestrial vertebrates during the nesia; osteoglossids and catostomids of Late Cretaceous and early Cenozoic (Estes South China; amiines and catostomids of in- and Hutchison, 1980). Freshwater Arctic land North China; amiines, hiodontids and connections between the northern continents catostomids of East Khazakstan). may therefore have played a signi®cant role More recent geophysical and geochemical in amiine dispersal. studies (some still in progress) have raised a On the other hand, in the vast area be- startling and radically different alternative to tween the Bohai Gulf and Green River oc- the Paci®ca hypothesis. There is mounting currences there are several other Eocene evidence that, for a brief period (approxi- freshwater deposits with abundant fossil ®sh- mately 2 million years, during the late Paleo- es, yet these ichthyofaunas do not yet include cene and early Eocene), the Arctic Ocean dasyatids, ²Diplomystus, or clupeids (e.g., was almost completely landlocked, with a middle Eocene of Washington State, U.S., terrestrial corridor between North America, and British Columbia, Canada; inland areas Iceland, and Europe and another through of South China and East Khazakstan; Wil- Beringia, and a single outlet to the sea via son, 1977, 1978; Tang, 1959; Liu et al., the Turgai Strait (McKenna, 1998). With sev- 1962; Wang et al., 1981; Sytchevskaya, eral major river systems draining northward 1986). In addition, some taxa occur in one into the Arctic, net runoff/evaporation may region but not the other (e.g., Eocene lepi- have caused dramatic lowering in salinity, sosteids in Green River and western Canada suf®cient perhaps to allow freshwater ®shes but not Asia; Eocene cyprinids in Bohai to move unimpeded by a saltwater barrier be- Gulf, South China, and East Khazakstan but tween Asia and North America. not in North America or Europe). Such ab- Such a rapid dispersal event could certain- sences from the fossil record may re¯ect a ly account for the sister-group relationship preservational bias against certain taxa in one 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 31 area or the other, or it could indicate a real Either a substantial portion of their nonmar- absence, suggesting that some taxa did not ine fossil record is missing (as is evidenced participate in any holarctic dispersal event by the recent discovery of a possible para- during the late Paleocene and early Eocene. clupeine Ezkutuberezi from Spain; Poyato- Quite probably, there is no single factor Ariza et al., 2000), or their distribution in- responsible for the Eocene ``transpaci®c'' volved marine dispersal. distribution pattern. A broad connection be- 6. A ``freshwater Arctic Ocean'' hypothe- tween Asia and North America in the Bering sis (supported by several lines of geological Strait area could have played a role, as could evidence) is preferred over the ``Paci®ca'' temporary desalination of the Arctic Ocean. hypothesis (which lacks empirical support Both hypotheses provide equally plausible from geological data) to account for Eocene biogeographic alternatives to the Paci®ca (and younger) trans-Paci®c distribution pat- model and are certainly more in accordance terns of nonmarine ®shes (the hypothesis with geological data than the latter. At pre- may also explain distribution patterns among sent the only common link between the Early certain Recent ®shes). Cretaceous distribution of paraclupeines and the Eocene distribution of ²Diplomystus is ACKNOWLEDGMENTS China, for only there have both taxa been We thank Lorraine Meeker and Chester discovered. Tarka for preparing the illustrations and Rob- ert Evander for preparing the silicone peels. CONCLUSIONS We also thank Jin Meng for useful phylogenetic discussions. Mary Dawson and Chris 1. No derived characters are uniquely Beard kindly arranged for the loan of the shared by ²Ellimma branneri and the Clu- Carnegie Museum specimens. Support in the peiformes. Thus, while modern Clupeiformes United States for Chang was generously pro- may be monophyletic, the inclusion of ex- vided by the H. and E. Axelrod Research tinct forms such as ²Ellimma and other ²par- Chair in Vertebrate Paleontology, American aclupeines may potentially create a paraphy- Museum of Natural History, and by Lance letic assemblage. Grande, Field Museum of Natural History. 2. One derived character (lateral expansion Support for her in China is from the National of dorsal scute ``wings'' which give scutes a Science Foundation of China, project subrectangular shape) is shared by ²Ellimma 48932010. and other members of the family ²Paraclu- peidae Chang and Chou (1977); ϭ ²Ellim- REFERENCES michthyidae Grande (1982a). 3. The Paraclupeidae is divided into two Arratia, G. 1996. Reassessment of the phyloge- subfamilies: Paraclupeinae Chang and netic relationships of certain Jurassic teleosts Grande (1997) and an unnamed sistergroup. and their implications on teleostean phylogeny. Paraclupeids are known from the Lower Cre- In G. Arratia and G. Viohl (editors), Mesozoic ®shesÐsystematics and paleoecology: 219± taceous±middle Eocene. 242. MuÈnchen: Verlag Dr. F. Pfeil. 4. The Paraclupeinae include ²Paraclu- Chang, M.M., and Y.Y. Chen. 2000. Late Meso- pea,²Ellimmichthys, and ²Ellimma. These zoic and Tertiary ichthyofaunas from China and taxa are united by strongly sculptured skull some puzzling patterns of distribution. Verte- roo®ng bones, with ridges radiating from the brata PalAsiatica 38(3): 161±175. growth center, and by dorsal scute ornament Chang, M.M., and J.J. Chou. 1977. On Late Me- of prominent ridges. ²Scutatuspinosus may sozoic fossil ®shes from Zhejiang Province, also belong in the ²Paraclupeinae. Paraclu- China. Institute of Vertebrate Paleontology and peine ®shes are at present only recognized Paleoanthropology, Academia Sinica Memoirs 12: 1±60. [in Chinese with English summary] from the Lower Cretaceous (Hauterivian±Al- Chang, M.M., and J.J. Chou. 1978. On the fossil bian). ®shes in Mesozoic and Cenozoic oil-bearing 5. No biogeographical hypothesis satisfac- strata from east China and their sedimentary torily explains the known distribution of non- environment. Vertebrata PalAsiatica 16(4): marine paraclupeine ®shes in the Cretaceous. 229±237. [in Chinese] 32 AMERICAN MUSEUM NOVITATES NO. 3404

Chang, M.M., and J.J. Chou. 1986. Stratigraphic Gayet, M. 1989. Note preÂliminaire sur le mateÂrial and geographic distributions of the late Meso- paleÂoichthyologique eÂocreÂtacique du Rio Ben- zoic and Cenozoic ®shes of China. Indo-Paci®c ito (sud de Bata, GuineÂe EÂ quatoriale). Bulletin ®sh biology. In T. Uyeno, R. Arai, T. Taniuchi, du MuseÂum national d'Histoire Naturelle, Sec- and K. Matsuura (editors), Proceedings of the tion C, Sciences de la Terre Paleontologie Geo- Second International Conference of Indo-Pacif- logie Mineralogie 11(1): 21±31. ic Fishes: 529±539. Tokyo: Ichthyological So- Grande, L. 1982a. A revision of the fossil genus ciety of Japan. ²Diplomystus, with comments on the interre- Chang, M.M., and L. Grande. 1997. Redesciption lationships of clupeomorph ®shes. American of ²Paraclupea chetungensis, an early clupeo- Museum Novitates 2728: 1±34. morph from the Lower Cretaceous of south- Grande, L. 1982b. A revision of the fossil genus eastern China. Fieldiana Geology New Series, ²Knightia, with a description of a new genus (37) 1489: 1±19. from the Green River Formation (Teleostei, Chang, M.M., D.S. Miao, Y.Y. Chen, J.J. Zhou, Clupeidae). American Museum Novitates 2731: and P.F. Chen. 2001. Suckers (®sh, Catostomi- 1±22. dae) from the Eocene of China account for the Grande, L. 1985a. Recent and fossil clupeomorph family's current disjunct distributions. Science ®shes with materials for revision of the sub- in China 31: 577±586. groups of clupeoids. Bulletin of the American Chang, M.M., and J.J. Zhou. 1993. A brief survey Museum of Natural History 181(2): 231±372. of the Chinese Eocene ichthyofauna. DarmstaÈd- Grande, L. 1985b. The use of paleontology in sys- ter BeitraÈge zur Naturgeschichte Heft 2: 157± tematics and biogeography, and a time control 162. re®nement for historical biogeography. Paleo- Chang, M.M., and J.J. Zhou. 2002. First discov- biology 11(2): 234±243. ery of Esox (Teleostei) in China. Vertebrata Grande, L. 1989. The Eocene Green River Lake PalAsiatica 39(2): 81±96. [in Chinese with En- system, Fossil Lake, and the history of the glish summary] North American ®sh fauna. In J. Flynn (editor), Cope, E.D. 1874. Supplementary notices of ®shes Mesozoic/Cenozoic vertebrate paleontology: from the freshwater Tertiaries of the Rocky classic localities, contemporary approaches. Mountains. Bulletin of the U.S. Geological and 28th International Geological Congress Field- Geographical Survey of the Territories 1(2): trip Guidebook T 322: 18±28. Washington DC: 49±51. American Geophysical Union. Cope, E.D. 1877. A contribution to the knowledge Grande, L. 1994. Repeating patterns in nature, of the ichthyological fauna of the Green River predictability, and ``impact'' in science. In L. shales. Bulletin of United States Geological and Grande and O. Rieppel (editors), Interpreting Geographical Survey 3(34): 807±819. the hierarchy of nature: from systematic pat- Cope, E.D. 1886. A contribution to the vertebrate terns to evolutionary process theories: 61±84. paleontology of Brazil. Proceedings of the New York: Academic Press. American Philosophical Society 23(121): 3±4. Grande, L., and W.E. Bemis. 1998. A comprehen- Costa, A.C., V.A. Mendes, and D. E.G. Rocha. sive phylogenetic study of amiid ®shes (Ami- 1979. Sintese preliminar da Geologia do Pro- idae) based on comparative skeletal anatomy. jecto Extremo Nordeste do Brasil. Boletim do An empirical search for interconnected patterns NuÂcleo do NE Sociedade Brasileira de Geolo- of natural history. Society of Vertebrate Pale- gia 7: 421±440. ontology Memoir 4: i±x, 1±690; Supplement to Eastman, C.R. 1912. Tertiary ®sh-remains from Journal of Vertebrate Paleontology 18(1). Spanish Guinea in West Africa. Annals of Car- Greenwood, P.H., D.E. Rosen, S.H. Weitsman, negie Museum 8(2): 370±378. and G.S. Myers. 1966. Phyletic studies of tel- Estes, R., and J.H. Hutchison. 1980. Eocene lower eostean ®shes with a provisional classi®cation vertebrates from Ellesmere Island, Canadian of living forms. Bulletin of the American Mu- Arctic Archipelago. Palaeogeography, Palaeo- seum of Natural History 131: 339±456. climatology, Palaeoecology 30: 325±347. Hills, E.S. 1934. Tertiary fresh water ®shes from FeijoÂ, F.J. 1994. Bacias de Sergipe e Alagoas. southern Queensland. Memoirs of the Queen- Boletim GeocieÃncias da PETROBRAÂ S, Rio de land Museum 10: 157±174. Janeiro 8(1): 149±161. Jordan, D.S. 1907. The fossil ®shes of California, Forey, P.L. 1975. A fossil clupeomorph ®sh from with supplementary notes on other species of the Albian of the Northwest Territories of Can- extinct ®shes. University of California Publi- ada, with notes on cladistic relationships of clu- cations in Geological Sciences 5: 95±144. peomorphs. Journal of Zoology (London) 175: Jordan, D.S. 1910. Description of a collection of 151±177. fossil ®shes from the bituminous shales at Ria- 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 33

co Doce, State of Alagoas, Brazil. Annals of Gault (Lower Cretaceous). Zoological Journal Carnegie Museum 7(1): 23±34. of Linnean Society 49(3): 161±182. Jordan, D.S. 1913. Ellimma, a genus of fossil her- Patterson, C., and D.E. Rosen. 1977. A review of rings. Proceedings of the Biological Society of ichthyodectiform and other Mesozoic teleost Washington 26: 79. ®shes, and the theory and practice of classify- Li, G.Q. 1994. Systematic position of the Austra- ing fossils. Bulletin of the American Museum lian fossil osteoglossid ®sh ²Phareodus of Natural History 158(2): 81±172. (ϭPhareoides) queenslandicus Hills. Memoirs Pitman, W.C., S. Cande, J. LaBreque, and J. Pin- of the Queensland Museum 37(1): 287±300. dell. 1993. Fragmentation of Gondwana: the Liu, T.S., H.T. Liu, and X. Tang. 1962. A new separation of Africa from South America. In P. pecoid ®sh from South China. Vertebrata Goldblatt (editor), Biological relationships be- PalAsiatica 6: 121±127. [in Chinese with En- tween Africa and South America: 15±34. New glish summary] Haven, CT: Yale University Press. Maisey, J.G. 1993. A new clupeomorph ®sh from Poyato-Ariza, F.J., M.A. LoÂpez-Horgue, and F. the Santana Formation (Albian) of NE Brazil. Garcia-Garmilla. 2000. A new early Cretaceous American Museum Novitates 3076: 1±15. clupeomorph ®sh from the Arratia Valley, Maisey, J.G. 2000. Continental break up and the Basque County, Spain. Cretaceous Research distribution of ®shes of Western Gondwana 21: 573±585. during the Early Cretaceous. Cretaceous Re- Russell, D.E., and R.J. Zhai. 1987. The paleogene search 21: 281±314. of Asia: mammals and stratigraphy. Memoires McKenna, M.C. 1975. Fossil mammals and early du MuseÂum National d'Histoire Naturelle, Ser- Eocene North Atlantic land continuity. Annals ie C, Sciences de la Terre, 52: 1±488. of Missouri Botanical Garden 62: 335±353. Sanders, M. 1934. Die fossilen Fische der alter- McKenna, M.C. 1980. Eocene paleolatitude, cli- tiaÈren SuÈsswasserablagerungen aus Mittel-Su- mate, and mammals of Ellemere Island. Palaeo- matra. Verhandelingen van het Geologisch-Mi- geography, Palaeoclimatology, Palaeoecology jnbouwkundig Genootschap voor Nederland en 30: 349±362. KolonieÈn Geologische Serie 11(1): 1±143. McKenna, M.C. 1983. Cenozoic paleogeography Schaeffer, B. 1947. Cretaceous and Tertiary actin- of North Atlantic land bridges. In M.H.P. Bott, opterygian ®shes from Brazil. Bulletin of the S. Saxov, M. Talwani, and J. Thiede (editors), American Museum of Natural History 89(1): Structure and development of the Greenland- Scotland Ridge: 351±399. New York: Plenum. 1±40. McKenna, M.C. 1984. Holarctic landmass rear- Scudder, S.H. 1902. On the United States Orthop- rangement, cosmic events, and Cenozoic terras- tera which have been referred to the genus Tri- trial organisms. Annals of the Missouri Botan- dactylus. Psyche, A Journal of Entomology. 9: ical Garden 70: 459±489. 308±310. McKenna, M.C. 1998. Semi-isolation of the Arc- Signeux, J. 1951. Notes paleÂoichthyologiques. V. tic Ocean in the late Thanetian to earliest Yp- Diplomystus dubertreti, une nouvelle espeÁce du resian time. Geological Society of America Ab- SeÂnonien du Liban. Bulletin du MuseÂum Na- stracts with Programs 30(7): 118. tional d'Histoire Naturelle 23(2): 692±693. Nelson, G.J. 1967. Gill arches of teleostean ®shes Silva Santos, R.Da. 1990. Clupeiformes e Gon- of the family Clupeidae. Copeia 1967: 389± orhynchiformes de CretaÂceo Inferior (Aptiano) 399. da FormacËaÄo Cabo, Nordeste do Brasil. Anais Nelson, G.J. 1970. The hyobranchial apparatus of da Academia brasileira de CieÃncias 62(3): 261± teleostean ®shes of the families Engraulidae 268. and Chirocentridae. American Museum Novi- Silva Santos, R.Da., and V.L.S. Correa. 1985. tates 2410: 1±30. ContribucËaÄo ao conhecimento da paleoictio- Nelson, G.J. 1973. Relatiohships of clupeo- faÂunula do CretaÂceo no Brasil. In D. De A. morphs, with remarks on the structure of the Campos, C.S. Ferreira, I.M. Brito, and C.F. Vi- lower jaw in ®shes. In P.H. Greenwood, R.S. ana (editors), ColetaÃnea de Trabalhos Paleon- Miles, and C. Patterson (editors), Interrelation- toloÂgicos SeÂrie Geologia 27(2): 169±174. Rio ships of ®shes. Zoological Journal of Linnean de Janeiro: MinisteÂrio das Minas e Energia±De- Society 53(Suppl. 1): 333±349. partamento Nacional de ProducËaÄo Mineral. Patterson, C. 1967. Are the teleosts a polyphyletic Su, D.Z. 1986. The discovery of a fossil osteo- group? Colloques Internationaux du Centre Na- glossid ®sh in China. Vertebrata PalAsiatica tional de la Recherche Scienti®que 163: 93± 24(1): 10±19. [in Chinese with English sum- 109. mary] Patterson, C. 1970. A clupeomorph ®sh from the Sun, A.L. 1956. ParaclupeaÐa genus of double- 34 AMERICAN MUSEUM NOVITATES NO. 3404

armored herrings from Chekiang. Acta Pa- Wilson, M.V.H. 1977. Middle Eocene Freshwater laeontologica Sinica 4: 413±418. ®shes from British Columbia. Royal Ontario Swofford, D.L. 2000. PAUP*. Phylogenetic anal- Museum Life Sciences Contributions 113: 1± ysis using parsimony (*and other methods). 61. Version 4. Sunderland, MA: Sinauer Associ- Wilson, M.V.H. 1978. Eohiodon woodruf® n. sp. ates. (Teleostei, Hiodontidae), from the Middle Eo- Tang, X. 1959. A new fossil ®sh from Linli, Hu- cene Klondike Mountain Formation near Re- nan. Vertebrata PalAsiatica 1: 211±213. [in Chinese] public, Washington. Canadian Journal of Earth Taverne, 1975. A propos du genre fossile Knightia Sciences 15(5): 679±686. Jordan, D.S., 1907, ClupeÂideÂaÁ double careÁne Woodward, A.S. 1895. On two deep-bodied spe- de l'EoceÁne des Etats-Unis et du BreÂsil. Biol- cies of the clupeoid genus Diplomystus. Annals ogisch Jaarboek Dodonaea 43: 246±251. and Magazine of Natural History ser. 6, 15: 3. Uyeno, T. 1979. Early Cretaceous freshwater ®sh- Woodward, A.S. 1901. Catalogue of fossil ®shes es from Northern Kyushu, Japan. 1. Description in the British Museum (Natural History). Vol- of two new species of the clupeid genus Diplo- ume 4: 696 pp. British Museum (Natural His- mystus. Bulletin of the Kitakyushu Museum of tory), London. Natural History 1: 11±24. Yabumoto, Y. 1994. Early Cretaceous freshwater Uyeno, T., and Y. Yabumoto. 1980. Early Creta- ®sh fauna in Kyushu, Japan. Bulletin of the Ki- ceous freshwater ®shes from northern Kyushu, takyushu Museum of Natural History 13: 107± Japan. Bulletin of the Kitakyushu Museum of Natural History 2: 25±31. 254. Wang, J.K., G.F. Li, and J.S. Wang. 1981. The Zhang, M.M., J.J. Zhou and D.R. Qing. 1985. Ter- Early Tertiary fossil ®shes from Sanshui and its tiary ®sh fauna from coastal region of Bohai adjacent basin, Guangdong. Palaeontologia Sin- Sea. Academia Sinica Institute of Vertebrate ica 160; New Series C 22: 1±90. [in Chinese Paleontology and Paleoanthropology Memoirs with English summary] 17: 1±60.

APPENDIX 1

CHARACTER LIST 14. Dorsal scutes, absent (0), scute series incomplete (1) or complete (2). 1. Recessus lateralis absent (0), present (1). 15. Ornamentation on dorsal scutes, absent (0), 2. Otophysic connection involving a divertic- present (1). ulum of the swimbladder that penetrates 16. Abdominal scutes absent (0), present (1). the exoccipital and extends into the prootic 17. Caudal scutes present (0), absent (1). within the lateral wall of the braincase, ab- 18. First ural centrum about equal in size with sent (0), present (1). ®rst preural centrum (0), much smaller (1). 3. Preepiotic fossa absent (0), present (1). 19. Neural spine of ®rst preural centrum short 4. Parietals meeting at midline (0), separated or lacking (0), long (1). completely by supraoccipital (1). 20. Parhypural not fused to ®rst preural cen- 5. Frontal fontanelle absent (0), present (1). trum (0), fused to ®rst preural centrum (1). 6. Supratemporal commissure not passing 21. Number of hypurals: six or more (0), less through parietals or parietals and supraoc- than six (1). 22. Second hypural not fused with ®rst ural cen- cipital (0), passing through these bones (1). trum (0), fused with ®rst ural centrum (1). 7. Anterodorsal process of hyomandibular ab- 23. Number of epurals: three (0), less than sent (0), present (1). three (1). 8. Supramaxillae, two (0), less than two (1). 24. First uroneural independent (0), fused with 9. Teeth on entopterygoid ®ne or absent (0), ®rst preural (1). strong (1). 25. First uroneural extending forward to sec- 10. Tooth patch on posterior part of parasphen- ond preural centrum (0), failed to extend to oid present (0), absent (1). second preural centrum (1). 11. Basipterygoid process present (0), absent (1). 26. Third hypural not expanded posteriorly, 12. Bericiform foramen in anterior ceratohyal leaving a gap or notch between second and present (0), absent (1). third hypurals (0), expanded posteriorly, 13. Angular fused with retroarticular (0), with leaving no gap or notch between second articular (1). and third hypurals (1). 2003 CHANG AND MAISEY: TWO EXTINCT CLUPEOMORPHS 35

27. First hypural articulating with ®rst ural 29. Lateral line scales present (0), absent (1). centrum (0), not articulating with (1). 30. Neural spines of a few posterior preurals 28. Semicircular to almost vertical ®ne surface without laminar outgrowth (0), with lami- ridges (circuli) on scales absent (0), present nar outgrowth and lea¯ike (1). (1). Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates and Bulletin published during the last ®ve years are available at World Wide Web site http://library.amnh.org. Or address mail orders to: American Museum of Natural History Library, Central Park West at 79th St., New York, NY 10024. TEL: (212) 769-5545. FAX: (212) 769- 5009. E-MAIL: [email protected]

a This paper meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).