AMERICAN MUSEUM Novitates PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y

Home , Articular bone, Myctophiformes

AMERICAN MUSEUM Novitates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2827, pp. 1-57, figs. 1-45 August 13, 1985

An Essay on Euteleostean Classification

DONN E. ROSEN' ABSTRACT The anatomy of the occipital region and rostral letic and raises questions about the monophyly of cartilage in euteleostean fishes is reviewed in some the fishes formerly grouped in the Osmeroidei. detail. These data, in combination with other an- Evidence is presented on how the occipital region atomical features taken from the literature, have might be used in acanthomorph systematics, and led to a reassessment of interrelationships within includes reasons for rejecting the concept of the the Euteleostei. This review supports the notions Paracanthopterygii, as this group was formerly that the Salmoniformes, Aulopiformes, Mycto- constituted. phiformes, and Beryciformes are nonmonophy- INTRODUCTION The earliest general classification of fishes dan (1923), A. Smith-Woodward (1932), in which it is possible to pick out many of Norman (1934), Berg (1940) and its various the main components of the Euteleostei is translations, reprinted editions and slightly that ofJohannes Muller (1844), in which the modified versions, and lastly Greenwood et teleosts as a whole were presented as a ver- al. (1966), McAllister (1968), and J. Nelson tebrate subclass, and their components as or- (1984)] went through an evolution from the ders. These orders of Muller's bore names early nonsubordinated, ordinal classifica- that may seem strange and unfamiliar to to- tions of Muller (1844), Agassiz (1858), Gill day's student, but the etymological charac- (1872), Boulenger (1904), and Regan (1909, teristics of many of them were preserved for 1929), to the complex subordinated, hierar- some time, a few even to the present. chical system of Goodrich (1909). It was Some of these names and their cross- Goodrich (1909) who introduced the use of equivalents in different classifications were uniform group endings, a practice that was reviewed and explained in detail by Myers adopted by Berg (1940) and his successors, (1958) and now it remains only to emphasize to ease recognition of the hierarchical posi- that these many different classifications tion of a group in the general classification. [Muller (1844), Agassiz (1858), Gunther During all ofthis history some ofthe fishes (1859-1870), Gill (1872), Boulenger (1904), now assigned to the Euteleostei were distrib- Regan (1909, 1929), Goodrich (1909), Jor- uted throughout most of the higher taxa rec-

1 Curator, Department of Ichthyology, American Museum of Natural History.

MULLER 1844

REGAN 1906 1929

FIG. 1. Branching diagrams representing the degree of resolution of teleostean classifications from 1844 (Muller) to 1909 (Goodrich) and 1929 (Regan). 1985 ROSEN: EUTELEOSTEANS 3

ognized by earlier authors. For example, in ment is a sign ofhealth and vigor in the field MiUller's (1844) classification of fishlike ver- that should serve as an example for other tebrates, the subclass Teleostei included six vertebrate systematists, some of whom have orders, each of which included taxa now en- abandoned the search for hierarchical order compassed by the Euteleostei. Thus, his in favor ofgeneral ecological research or lab- Acanthopteri included "perciform" fishes; the oratory studies of the behavior of selected Anacanthini, the codfishes, cusk eels, and species. flatfishes; Pharyngognathi, the labrids, and This paper is premised on certain new ob- their immediate allies; the Phystomi, the oto- servations and interpretations of teleostean physan Ostariophysi, percopsiforms, myc- anatomy and the cladistic notions derived tophids, salmonids, galaxiids, synbranchid from them. One of these notions, in agree- eels, pikes, and mudminnows; the Plectog- ment with Fink and Weitzman (1982) and nathi, members of the modem Tetraodon- Fink (1984), is that the Salmoniformes ofRo- tiformes; and the Lophobranchii, the pipe- sen (1974) is not monophyletic. The second fishes, and seahorses. The Physostomi of is that the Aulopiformes of Rosen (1973) is Muller also included some osteoglosso- also not monophyletic. A third is that Poly- morphs and the former is understood to be mixia and the acanthomorph neoteleosteans more or less equivalent to the Malacopteryg- are defined, in part, by the absence of a re- ii, Isospondyli, and Clupeiformes in classi- sidual neural arch between the first vertebra fications as recent as Berg's (1940). The three and the occiput that is primitively present in latter "groups" were unnatural assemblages most halecostomes, many primitive eute- of primitive teleosts, which included numer- leosts, and in neoscopelids but not mycto- ous taxa later treated by Greenwood et al. phids among the Myctophiformes, as rede- (1967) as euteleosteans in another unnatural fined by Rosen (1973) to include the assemblage that Greenwood et al. (1966) had myctophids and neoscopelids.2 A fourth is previously termed the Protacanthopterygii. Progress has been slow, teleost classifications 2 This and other characters conflict with four shared, going through a long period when all teleosts derived features for myctophids and neoscopelids given were assembled into one ofthree main kinds: by Stiassny (Ms), but are consistent with the presence of lower, intermediate, or higher even as re- a subocular shelf, and single, medial, rostral cartilage in cently as Gosline (1971). most polymixiids and acanthomorphs (absent in neo- The first comprehensive modem attempt scopelids), and three or fewer predorsal bones in myc- at detailed hierarchical synthesis since Good- tophids, polymixiids, and acanthomorphs (four in neo- rich (1909) was that of Greenwood et al. scopelids). One of Stiassny's characters linking the two (1966), closely followed by that ofMcAllister families is the cone-shaped ventrally directed parapoph- (1968). This history is best appreciated by yses for Baudelot's ligament that will, I believe, prove a of dia- to be the primitive state of a similar structure in Poly- examining sprinkling branching mixia and other acanthomorphs (fig. 18). Another of grams extracted from the main components Stiassny's reasons for linking myctophids and neosco- of each of the systems proposed since 1844 pelids is described by Lauder (1983). In Lauder's paper, (figs. 1 and 2). he stated that "all myctophiforms (including neoscope- It might be guessed that many major taxo- lids; Rosen, 1973) possess a unique attachment of the nomic problems have remained unsolved branchial skeleton to the urohyal." In myctophids, how- once an interest in cladistic methods ofanal- ever, the third hypobranchials have long anteroventral ysis and classification was adopted by the ich- tips that clasp the urohyal laterally, whereas in neosco- thyological community and most of these pelids, the anteroventral tips ofthe third hypobranchials problems might be expected to be within the extend forward above the urohyal to the dorsal edge as the of all where they are attached by ligament. The latter condition Euteleostei, largest recognized is similar to that for primitive acanthomorphs (e.g., Poly- teleostean assemblages. I do not regard these mixia) except that in the latter, the ligaments from the problems as close to solution since there ex- third hypobranchials extend forward to contact the dor- ists, still, significant disagreement amongst sal edge of the urohyal more anteriorly. In any event, I ichthyologists on the interpretation of char- see no character here that clearly aligns myctophids with acter information and the delimitation of neoscopelids. This leaves two ligament characters of natural groups (Fink, 1984). This disagree- Stiassny's to align those taxa, as compared with the eight 4 AMERICAN MUSEUM NOVITATES NO. 2827 that the Neoteleostei is characterized by a Certain subsidiary notions also emerge as tripartite occipital condyle (the basioccipital consequences of this work, (1) that the "os- and two exoccipital condyles), as described meroids" and the Salmonidae might not be and illustrated by Rosen and Patterson (1969), monophyletic groups. Pefiaz (1983, p. 370) and which unites stomiiforms with them, as recently attempted an ontogenetic diagnosis elaborated by Fink and Weitzman (1982) and ofthe Salmonidae, but pointed out that some Fink (1984). A fifth is that, in disagreement of the diagnostic features may also be found with Fink and Weitzman (1982), the presence in other fish groups, (2) that Aulopus, and ofa well-defined triplejoint that incorporates perhaps some closely allied forms excluding two large exoccipital condyles is not evidence Chlorophthalmus, mightjointly form the sis- for linking the Salmonidae with the neote- ter group to the Ctenosquamata based, in part, leosts since this type of joint has a limited on the anatomy of the rostral cartilage, and distribution only in Recent salmonines and (3) that within the groups that remain, the is, therefore, probably convergent. A sixth is old Paracanthopterygii toadfishes and their that neoteleosts primitively show a cervical immediate allies might be more closely linked gap between the occiput and first vertebra. A to cods and theirs than the cods are to cusk seventh is that the Acanthomorpha are de- eels and brotulas based on both neurocranial, fined by complete closure of the cervical gap vertebral, and gill arch evidence.3 via two prezygapophyseal exoccipital facets and a basioccipital facet from the body ofthe ACKNOWLEDGMENTS vertebral centrum. An eighth is that Poly- I am grateful to the following for com- mixia is the sister group to the Acantho- mentaries during the work and specimens for morpha, thus defined because it possesses ex- study, especially Daniel M. Cohen, S. and W. occipital facets but retains a part ofthe cervical Fink, P. H. Greenwood, George Lauder, Co- gap in the basioccipital position. Other "be- lin Patterson, D. Siebert, K. Sulak, Richard ryciforms" have a more derived "percoid- Vari, and S. Weitzman. Part ofthis work was like" condition. begun with Colin Patterson and Melanie Stiassny in 1983, both ofwhom supplied me with many helpful comments and drawings. features that relate myctophids, but not neoscopelids, to To the others, many thanks: R. Bailey, F. the acanthomorphs. (1) a large subocular shelf (like that L. G. Nelson, and in Polymixia), (2) rostral cartilage a simple median struc- Berry, B. Collette, Grande, ture (with relic pairs of lateral cartilages, or no paired V. Springer. I thank Mary Rauchenberger, cartilages), (3) an interarcual cartilage between the first Christine Rossi, and Janice Lillien for much and second gill arches (small, when present, and absent technical assistance and specimen prepara- in some species including Polymixia and some primitive tion. This work was supported by NSF grant acanthomorphs), (4) only three predorsal bones, (5) par- BSR 8100103. tial closure ofthe cervical gap, (6) absence ofan accessory neural arch in cervical region, (7) presence ofneural arch ANATOMICAL ABBREVIATIONS prezygapophyses on the first vertebra, (8) direct connection (via ligaments) of the autocentrum of first vertebra ACCNA, accessory neural arch with the exoccipital condyles. Character 3 is questionable ABAUDLIG, attachment for Baudelot's ligament because several groups ofprimitive acanthomorphs lack ANA, ankylosed neural arch an interarcual cartilage and characters 6 through 8 might ARTPR, articular process be manifestations of only a single developmental shift. ASCPR, ascending process Even allowing for the latter two ambiguities, there are AUTLIG, autocentral ligament four trenchant features suggesting nonmonophyly ofthe AUTNA, autogenous neural arch myctophids plus neoscopelids (characters 1, 2, 4, and 5- BAUDLIG, Baudelot's ligament 8) as contrasted with the two remaining ligament features BO, basioccipital proposed by Stiassny. This state of affairs indicates to BOC, basioccipital condyle me that myctophids are the sister group of acantho- BOF, basioccipital facet morphs (in a restricted ctenosquamata), but that neo- CG, cervical gap scopelids, for reasons discussed in this paper, are best regarded, at present, as part of major polychotomy immediately preceding the ctenosquamata in the cladogram 3The gill arch evidence will be discussed in a subse- (fig. 45). quent paper by Patterson and Rosen. 1985 ROSEN: EUTELEOSTEANS 5

A. SMITH WOODWARD 1932

MCALLISTER 1 968

FIG. 2. Branching diagrams representing the degree of resolution of teleostean classifications from 1932 (A. Smith-Woodward) to the present time as summarized by Rosen in 1982. 6 AMERICAN MUSEUM NOVITATES NO. 2827

CTNOT, connective tissue sheath of notochord 1964; Rosen and Patterson, 1969; and Fink ?CART, questionably cartilaginous and Weitzman, 1982) that advanced eute- DNA, depression in autocentrum for neural arch leosts can be characterized by the presence base in the posterior neurocranium ofan inverted ENR, epineural rib EP, epural Y-shaped junction between the basioccipital EXO, exoccipital and the exoccipitals. This configuration can EXOC, exoccipital condyle be seen in primitive myctophids, stomi- EXOF, exoccipital facet iforms, tCtenothrissa radians, and Polymix- EXO FRAGMENTS, exoccipital fragments from ia (Patterson, 1964). This condition appears site of attachment of autocentral ligament to differ from that of primitive teleosts in HYPI-6, hypural 1 to 6 which the basioccipital occupies the entire LEXO, left exoccipital area ofthese three bones for contact with the MEDCART, medial upper jaw cartilage centrum ofthe first vertebra. When that basi- MX, maxilla NA, neural arch occipital contact is reduced, the exoccipitals NAPZYG, prezygapophyseal neural arch enter the posterior neurocranial surface to NOT, notochord form a tripartite condylar surface for contact NPU2, 3, neural spines on second and third pleural with the first vertebra. centrum Fink and Weitzman (1982), citing the ear- PAL, palatine lier paper by Rosen and Patterson (1969), PAR, parapophysis called attention to the tripartite occipital con- ?PAR, questionably a parapophysis dyle as a synapomorphy ofneoteleosts. Find- PARHYP, parhypural ing a similar occipital joint in the gonosto- PD 1, first predorsal bone matid stomiiform, Diplophos, they proposed PMX, premaxilla POPMYO, posterior opening of posterior myo- that this is one of two features that unites dome stomiiforms with neoteleosts. The other fea- PMXCART, premaxillary cartilage ture, a rostral premaxillary cartilage, is dis- PR, pleural rib cussed below. Fink and Weitzman also used PTMXPR, postmaxillary process of premaxilla the tripartite condyle and rostral cartilage to PU1, 2, first or second preural centrum propose a sister-group relationship ofthe Sal- PZYG, prezygapophysis monidae with the neoteleosts plus stomi- RCART, rostral cartilage iforms. RETDORS TENDON, tendinous origin of re- The nature of the occipital joint with the tractor dorsalis muscle first vertebra has attracted the attention of REXO, right exoccipital RV1NA, right halfofneural arch on first vertebra several investigators. Ridewood (1904, 1905) RVINSP, right half of neural arch and spine on held the view that a tripartite joint is prim- first vertebra itive for teleosts, but is masked by the fusion SACBUL, saccular bulla of the first vertebral centrum (V1) to the oc- SEXOF, site for development of exoccipital facet ciput so as to exclude the exoccipitals from STEG, stegural the joint surface. In that view, what is iden- U 1, 2, first or second ural centrum tified in most teleosts as the basioccipital is UN 1, 2, first or second uroneural actually a vertebra fused to the braincase. V1, 2, 3, first, second or third vertebra Removal of this vertebra should, therefore, VIANA, ankylosed neural arch of first vertebra reveal the primitive tripartite arrangement. V1NSP, neural arch and spine of first vertebra Patterson (1975, p. 318) proposed that the X, foramen for vagus nerve basioccipital condyle, rather than being a ver- INSTITUTIONAL ABBREVIATIONS tebra, is made up of a plug of osteoid tissue AMNH, American Museum of Natural History representing the ossification of the small an- MCZ, Museum ofComparative Zoology, Harvard terior part of the notochord that penetrates University the basioccipital bone, and that it is the growth of this osteoid plug that excludes the exoc- ANATOMICAL EVIDENCE cipitals from the posterior face ofthe occiput. THE NEUROCRANIAL JOINT WITH THE FIRST Cavender and Miller (1972) also reviewed VERTEBRA: It has been proposed (Patterson, the origin ofthe salmonid occipital joint and 1985 ROSEN: EUTELEOSTEANS 7

FIG. 3. Salmonine occipital regions show character and disposition of the condyles that articulate with the first vertebra. A. Salmo salar Linnaeus, AMNH 39098, posterior view ofskull ofadult specimen. B and C. Oncorhynchus tshawytscha (Walbaum), AMNH 21719. B. Posterior view of exoccipital and basioccipital bones. C. As in B, but in lateral view with first vertebra in place (see fig. 4).

concluded, correctly in my view, that the tri- Cavender and Miller took an interest in this partite condition of salmonids is present or matter after finding and describing a large not as a consequence of whether a vertebra middle Pliocene salmonid from western fuses or does not fuse with the braincase. North America (tSmilodonichthys rastro- 8 AMERICAN MUSEUM NOVITATES NO. 2827

EXOF

FIG. 4. The first vertebra of Oncorhynchus tshawytscha (Walbaum), AMNH 21719, in posterior (A) and three-quarter posterodorsal (B) views. This vertebra contains a pair of wells on its dorsal surface underlying a poorly developed (accessory) neural arch that does not bear a neural spine.

sus); they proposed that it has a close relationship to the species of Oncorhynchus. Af- can therefore be used as evidence to link the salmonids ter noting that Oncorhynchus has a tripartite with the neoteleosts. But in my material, exoccipital par- joint (figs. 3B, C, 4, 5) and tSmilodonichthys ticipation is not true of Prosopium williamsoni, P. cy- but a simple basioccipital condyle, they re- lindraceum, or any other coregonine examined (figs. 7, viewed the distribution ofthese different kinds 8, 13) and the condition in Thymallus is hardly different 9) and Pterothrissus, among of articulations in a variety of teleosts. Ca- from that of Albula (fig. other teleosts, in which only a small extension of the vender and Miller concluded that there is evi- exoccipital is visible posteriorly without noticeably af- dence of a vertebra fusing with the braincase fecting the shape of the basioccipital. The tripartite oc- in the Pliocene fossil and that such fusion is cipital condyle of salmonines is more derived than the by no means unusual or restricted to just a simple, inverted Y-shaped morphology in primitive few taxa. In fact, they report that "in Core- ctenosquamates (except during early ontogeny, fig. 12A), gonus two centra may be fused with the basi- resembling in the adult state that of an advanced per- occipital" and that the "condition in Proso- comorph (e.g., Lutjanus, cf. figs. 3B and 24, 25). Thus pium williamsoni and [Thymallus, Norden the resemblance is probably secondary. This conclusion (1961)] is somewhat intermediate" between predicts a different ontogeny for the salmonine and cten- if would indicate the tripartite condyle and the single one in osquamate conditions which, found, write in- their nonhomology (see below, p. 54). Perhaps the prob- Coregonus.4 They also that "close lem is, as stated by Fink (1984), that the monophyly of the Salmonidae "is based primarily on a single character, 4 Fink (1984) argued that because Prosopium has the apparent polyploidy of the karyotype . . ." and that, as exoccipitals participating in the occipital condyle along enunciated by him, many of the salmoniform taxa are with Thymallus and the Salmoninae that the tripartite unnatural (e.g., "salmonids," "osmeroids," and "eso- condition is primitive for the Salmonidae in general, and coids"). 1985 ROSEN: EUTELEOSTEANS 9

DNA

EXOF krv

B C FIG. 5. First vertebra of Salmo gairdneri Richardson, AMNH 40268, parr stage (ca. 10 cm total length). A. Dorsal view to show wells for accessory neural arch. B. Three-quarter posterior view. C. Anterior view to show extent of development of facets that articulate with exoccipital condyle. Note large notochordal canal in centrum. spection of the condyle [in tSmilodonich- [appears to have] become trapped by flanges thys] shows that it is ... a fused centrum that from the exoccipital and basioccipital . . . ," supported a neural arch [in] a pair of inden- thus agreeing with Ridewood's (1904) inter- tations ... on its dorsal surface" (cf. figs. 3- pretation of the clupeid occipital region (and 8). And further, that this "basicranial-verte- see Greenwood, 1968, on Denticeps). Addi- bral joint is similar to that found in Tarpon tional circumstantial support for the primi- [sic] atlanticus, Megalops cyprinoides tive ontogenetic incorporation of vertebrae (Greenwood, 1970), [and] Albula vulpes...." with the braincase in modem halecostomes Their claim is problematical because onto- comes from the correlation ofthe occurrence genetic data illustrating the course of verte- of accessory, free-floating neural arches with bral fusion are lacking for most cited exam- fusion of vertebrae to the occiput [e.g., in ples. Such data are available for Megalops Amia (Jollie, 1984a, p. 431)]. atlanticus, however (fig. 9A, B). Strong cir- The clupeocephalan first neural arch is re- cumstantial support for the idea can be found duced and incomplete (figs. 11 A, 12B) as in other elopomorphs, as illustrated by Forey compared with the neural arch associated with (1973), where the part claimed to be a fused the basioccipital in some elopomorphs (fig. centrum not only bears a neural arch but par- 10), and might, therefore, be another syn- apophyses as well (Forey, 1973, figs. 3, 5, 21- apomorphy of the Clupeocephali. This line 23, 31). of argument depends on an assumption that In Elops, which appears to have a vertebra the position of Baudelot's ligament is a reli- that is ontogenetically a part of the basioc- able landmark for the identification ofa given cipital (fig. 10 and illustrations in Forey, vertebra. In osteoglossomorphs there is no 1973), Baudelot's ligament is attached to the evidence, direct or circumstantial, that a cen- ventrolateral aspect of the first free vertebra. trum is primitively fused with the braincase Whitehead and Teugels (in press) describe a (e.g., in Hiodon or Scleropages), but in a more situation much like that of Elops in a fresh- derived condition several vertebrae either are water herring, Sierrathrissa. They state that closed adherent to the basioccipital (Osteo- "the posterior halfofa first vertebral centrum glossum) or are included in a complex an- lo AMERICAN MUSEUM NOVITATES NO. 2827

DNA

AR A B

PAR

SEXOF

C D FIG. 6. First vertebra ofjuvenile lake trout, Cristivomer namaycush (Walbaum), AMNH 39269 (ca. 6 cm total length). A. Lateral view (anterior to left). B. Dorsal view (anterior up) to show position of wells for accessory neural arch. C. Posterior view. D. Anterior view to show areas where exoccipital facets develop in larger specimens. Note large notochordal canal in centrum. kylosed structure (Arapaima, Notopterus). In uation in the osteoglossomorphs and elopo- Hiodon, Scleropages, and Osteoglossum, cephalans is that the ontogenetic mechanism Baudelot's ligament arises on VI as opposed for moving the ligament between V1 and the to its attachment to the basicranium in taxa basicranium is unknown. But if the position with fused centra. That correlation is not true of Baudelot's ligament in elopocephalans is of ostariophysans, however, where there is correctly judged to be primitively V1 and is no evidence of a centrum fused to the brain- stable, then two possible explanations for its case; yet, as pointed out by Fink and Fink variable attachment are: (1) the nonhomol- (1981), the ligament attaches to V1 in cypri- ogy of the vertebral and basicranial liga- noids, and to the basicranium in the other ments,4 and (2) the presence or absence of a otophysans.S All one can say about this sit- centrum fused to the braincase, as in tSmi- lodonichthys and some coregonines (according to Cavender and Miller, 1972). 5 All one can say for the Otophysi is that the onto- In a recent account of the development of genetic mechanism for transferring the ligament from the of VI to the basioccipital is unknown but the character syncranium salmonines, Jollie (1984b) appears to be consistent and therefore usable taxonom- states that there is no vertebra fused with the ically. The alternative is that there is no such implied braincase, but he did not comment on Ca- character transformation because the two kinds of lig- vender and Miller's (1972) paper, nor did he aments in the Otophysi are not homologous. In fact, if illustrate the presence of an accessory neural one envisions a whole series of ligaments arising prim- arch between the braincase and the first cer- itively on the basioccipital and V1, and inserting on the vical vertebra (bearing Baudelot's ligament), shoulder girdle-and the disappearance of one or more which appears to be a general feature of my of these in the ontogeny of different taxa-the ligament material. I cannot resolve these inconsis- that is left becomes a retained primitive character and tencies. the absence ofligaments from certain areas, the derived An accessory neural arch is in condition. Under such circumstances, comprehensive present survey of shoulder girdle support ligaments would have primitive neoteleosts (figs. 14B, 15), and most to be made before one could use the character in a cla- primitive neoteleosts exhibit a gap between distic sense. Nevertheless, I am inclined to treat the lig- the braincase and the first vertebra. This cer- aments as homologous when the insertion on the shoul- vical gap is progressively smaller in more de- der girdle is as precisely similar as illustrated by Fink rived neoteleosts (figs. 14-17), being taken and Fink (1981). up by ligament or bony facets from the au- 1985 ROSEN: EUTELEOSTEANS I1I

EXO

BO B

FIG. 7. Occipital views ofneurocranium. A. Coregonus artedi Lesueur, AMNH 20096 (ca. 7 cm total length). B. Thymallus arcticus (Pallas), after Norden (1961). Note postenor opening of posterior myo- dome in A and compare with fig. 3A and apparent absence of same in B. Note also that the exoccipitals are masked posteriorly by basioccipital condyle in A and are almost excluded from the condylar surface in B. Compare with figures 3 and 8. tocentrum in ctenosquamates (myctophoids cipital and the growth dorsally of this cen- plus acanthomorphs). The complete closure trumlike disc to exclude the exoccipitals. His of this gap is correlated with (1) the absence argument carries with it the implication that ofthe accessory neural arch (figs. 16, 17), (2) the tripartite arrangement of bones on the the formation of vertebral facets for the ex- posterior face of the braincase is a primitive occipitals, and (3) the presence of prezyg- feature that has been restored in neoteleosts apophyses on the neural arches. These three and some primitive euteleosts by a means so features are synapomorphies ofthe Acantho- far unknown. morpha (cf. figs. 14-20). The question ofwhich is the first and which Patterson (1975, p. 318) argued that the the second centrum in elopocephalans is, as appearance of a small centrum (or part of a Whitehead and Teugels (in press) remarked, centrum) fused to the braincase is merely a "left open." Detailed histological investiga- false impression produced by the centrumlike tions of early ontogeny might "close" the ossification of the notochord in the basioc- question. 12 AMERICAN MUSEUM NOVITATES NO. 2827

EXO

Bu a 0

FIG. 8. Occipital region. A. Prosopium cylindraceum (Pallas), AMNH 31044 (ca. 10 cm total length), shown in three-quarter view with the arrow representing the anteroposterior axis. B. Prosopium williamsoni (Girard), AMNH 37967 (ca. 27 cm total length), in posterior view, showing position ofaccessory neural arch above basioccipital condyle. Note that, in both, the exoccipitals are masked posteriorly by basioccipital condyle. Compare with figures 3, 7, and 13B. 1985 ROSEN: EUTELEOSTEANS 13

BAUDLIG BO DNA

ACCNA EXO

FIG. 9. Occipital regions in the elopomorphs, Megalops atlanticus Valenciennes (A, B) and Albula vulpes (Linnaeus) (C). A. A 10 cm total length juvenile (AMNH uncataloged). B. An 80 cm subadult, AMNH 55321. C. Subadult specimen, AMNH 21516. Note in B the ankylosis ofthe first vertebra with the basioccipital and the presence of wells on its dorsal margin for the small neural arch shown in A. In C, note especially the presence of an accessory neural arch articulating between the exposed tips of the exoccipitals which resemble those of Thymallus (fig. 7B). Compare also with figure 10.

Neoscopelids, chlorophthalmids, and au- accessory neural arch (Rosen and Patterson, lopids have a large notochordal gap and an 1969, figs. 61-63). (Rosen and Patterson mis- 14 AMERICAN MUSEUM NOVITATES NO. 2827

ACCNA

FIG. 10. Occipital region and first vertebra of a 6 cm Elops saurus Linnaeus, AMNH 51485. The left exoccipital is separated from the basioccipital to illustrate the latter's resemblance to a foreshortened cervical vertebra (demarcated by the sculpturing around the ventral half). The accessory neural arch corresponds with a pair of dorsal indentations or wells as in Megalops, figure 9B. takenly labeled the first of four predorsal within the immense acanthomorph assem- bones in their fig. 61A as a neural arch in blage. Neoscopelus, however.) Myctophids have a Thus, the tripartite joint of neoteleosts is remnant of the gap which is being closed by a very old feature and the primitive and wide- ligaments and bone from the autocentrum spread presence of an accessory neural arch (figs. 16-17) in the position ofacanthomorph is inferred to be the remains of an ontogeny prezygapophyses (figs. 18-20) and they and that had incorporated vertebral fusion with acanthomorphs lack an accessory neural arch. the occiput. In the fossil salmonine described These shared derived states of myctophids by Cavender and Miller (1972), V1 is fused and acanthomorphs, and the absence ofsame to the occiput to produce a Coregonus-like in Neoscopelus suggest that the Ctenosqua- single condylar articulation. A situation like mata should be restricted (as noted above) to that in some species of Salmo and Onco- the Myctophidae and the Acanthomorpha.2 rhynchus was previously described by Gos- The primitive position of Baudelot's liga- line (1969, fig. 7) in a species of the argen- ment on the first cervical (fig. 17) is retained tinoid genus Alepocephalus, but I know of in some primitive acanthomorphs such as none of the above described conditions in amblyopsids (Woods and Inger, 1957), but esocoids or ostariophysans, except to note Baudelot's ligament has made its way onto that in Esox the back end ofthe basioccipital the basicranium ofmany acanthomorphs, informs a disclike ossification resembling the cluding atherinomorphs (Woods and Inger, articular surface of a centrum (rings of acel- 1957). A survey needs to be conducted to lular bone that constrict the notochord), which discover the anatomical position and possi- grows rapidly, occluding the exoccipitals from ble ontogenetic correlatives of this ligament thejoint. This disc resembles that ofOsmerus 1985 ROSEN: EUTELEOSTEANS 15

x ACCNA ol

0 o

0 0~~~~~~~ a

0 / " X1 '0 0 0 0 00a BAUDLIG SACBU

EXO BO

B SACBUL FIG. 11. Osmeroid occipital regions. A. Spirinchus thaleichthys (Ayres), ca. 8 cm, AMNH 51363, showing the extent ofthe cervical gapjust below an accessory neural arch. B. Osmerus mordax (Mitchill), AMNH 21727; occipital region in specimens ranging in size from 1 to 3 cm in length. In B, the dimensions of the saccular bullae and their relation to the basioccipital condyle in a 2 cm individual in ventral view are shown at left; in the upper row, the middle figure shows the conjunction of the exoccipital and basioccipital; the same area appears at right in dorsal view in which articulation points for an accessory neural arch are also shown; the same area is again shown at lower right to illustrate where the basioccipital condyle is deformed by the neural arch bearing exoccipitals. Compare with figure 1 2B ofa 3 cm individual. 16 AMERICAN MUSEUM NOVITATES NO. 2827

FIG. 12. Occipital region in a salmonine and two osmeroids. A. Salmo gairdneri Richardson, AMNH 40308 (ca. 6 cm). B. Osmerus mordax (Mitchill), AMNH 108093 (ca. 3 cm). C. Spirinchus thaleichthys (Ayres), AMNH 51363 (ca. 7 cm). An accessory neural arch is shown in B. In A, B, and C, there is a cervical gap between the occiput and first vertebra occupied by unconstricted notochord and its connective tissue sheath, as indicated. The basioccipital part of the vertebral joint is incompletely developed in each. In B and C, the posterior cartilage cores of the exoccipitals are visible through the still poorly ossified, rounded basioccipital condyle; bits of the exoccipitals are visible above it and remain so to produce an effect like that in some coregonines. 1985 ROSEN: EUTELEOSTEANS 17

VINSP

V2 - A

PAR

\- PR x

C-000: ACCNA 000 o 00 00 0. 00aoo &00 o, oo

0 0a o INSRNARV ~000.0

.0 lb eo * 000

FIG. 13. A. Neural arches and spines of first three vertebrae of Salmo gairdneri Richardson, AMNH 40508, to illustrate the absence of neural arch prezygapophyses and the extent of the ventral cartilage base seated in dorsal vertebral wells; vertebral body of first vertebra and left half of its arch and spine omitted. B. Occipital region and cervical vertebral elements in Prosopium williamsoni (Girard), AMNH 37967 (ca. 7 cm) to show a vertebralike basioccipital condyle and its association with a differentiated neural arch without a spine. This neural arch resembles the accessory arch in other primitive euteleosts. Compare with figure 8 and figure 10 of Elops. 18 AMERICAN MUSEUM NOVITATES NO. 2827

FIG. 14. Occipital regions. A. Chlorophthalmus agassizi Bonaparte, AMNH 27402. B. Aulopus japonicus Gunther, AMNH 28635. The cervical gap between the occiput and the first vertebra is occupied by an unconstricted notochord and its connective tissue sheath, as indicated. In both species, the exoccipitals are exposed posteriorly above the basioccipital. mordax, described below. Its removal in of the exoccipitals (i.e., a dual, rather than either Osmerus or Esox would expose, not a tripartite, articular surface). tripartite joint, but the empty interior of the Study of the occipital region of Osmerus basioccipital and the formerly occluded ends mordax at different sizes (figs. 1 1B, 12B) (1- 1985 ROSEN: EUTELEOSTEANS 19

ACCNA

/ X ~~~~~~~~CTNOT

CG PAR~~

'I ~~~~~~~~~~~~B

FIG. 15. The occipital region of Aulopus japonicus Gunther, AMNH 28635, to show the extent of the cervical gap between the occiput and the first vertebra and its relation to the accessory neural arch. A. Posterior quartering view. B. Lateral view. The connective tissue in the region of the gap shows a slight degree of staining with alizarine dye, indicating the presence of some calcification or ossification. This slightly stained, transparent connective tissue gap is hypothesized here to be what remains of a single basioccipital facet that would normally occlude the exoccipitals from the posterior face of the occiput as in Elops and Prosopium (figs. 8, 10, and 13).

5 cm in standard length) suggests how the At 2 cm, neural arches are well developed neoteleostean condition might have arisen. along the vertebral axis and a small less well- 20 AMERICAN MUSEUM NOVITATES NO. 2827

BAUDLIG

ENR

EXO FRAGMENTSA ATITT T TN-'

BAUDLIG

FIG. 16. Anterior vertebrae in myctophids. A. Rhinoscopelus tenuiculus (Garman), AMNH 1915. B. Myctophum nitidulum Garman, AMNH 25022. In B, separation of Vl from the occiput caused the dense autocentral ligaments to break away with fragments from the exoccipitals. In A, a bit of neural arch base of V1 has grown forward into such ligaments (not shown here, but see fig. 17).

formed arch is present between the occiput and it articulates ventrally with a notochor- and VI. At 5 cm, the accessory neural arch dal-connective tissue plug in a distinct gap is ossified except at its dorsal and ventral tips between the occiput and the first cervical ver- 1985 ROSEN: EUTELEOSTEANS 21

ENR

EXO

? PAR

EXO

ABAUD LIG ? PAR

FIG. 17. Myctophid occipital regions to show the extent of the cervical gap and the autocentral ligaments to the exoccipitals just below prezygapophyses on the neural arches. These ligaments are hypothesized to be the primitive state of exoccipital facets (as prezygapophyses) in more derived ctenosquamate conditions. A. Myctophum obtusirostreTining, AMNH 25022. B. M. aurolaternatum Garman, AMNH 15975. tebra (fig. 1 1, 12B). At 2 cm, the small arch The anteroposterior dimension ofthis disc is articulates ventrally with a flat disc that is very small (fig. 11 B), indicating that if it is a adherent to the basioccipital and is somewhat centrum, that centrum has formed only indented dorsolaterally by the exoccipitals. around a narrow anterior part of the noto- 22 AMERICAN MUSEUM NOVITATES NO. 2827

NAPZYG

, ENR

BAUDLIG

EXO

CG ? PAR (CTNOT) (ABAUD HG) FIG. 18. The cervical region of Polymixia lowei Gunther, AMNH 37335. A. First two vertebrae showing the exoccipital facets and a persisting notochordal plug on V1. B. The vertebrae in A in their articulated position. Note that the cervical gap is closed dorsally by autocentral prezygapophyses to the exoccipitals, but still open ventrally. chordal sheath which is easily deformed by The origin ofthe ctenosquamate triplejoint the exoccipitals. appears to follow a direct course involving As hypothesized above, the presence of an the following steps: (1) formation of a gap accessory neural arch, in the absence of an between the occiput and VI and exposure of underlying centrum, is inferred to represent the basi- and exoccipitals as attachment or a retention of the neural arch component of articular surfaces (figs. 14, 15); (2) loss ofthe a vertebral segment that either is incomplete accessory neural arch in all myctophids and or had been incorporated indistinguishably acanthomorphs (figs. 16-20); (3) attachment into the braincase. of the dorsolateral part of V 1 with the ex- 1985 ROSEN: EUTELEOSTEANS 23

AUTNA

..- EXOC

B11 BOF

FIG. 19. Holocentrid cervical anatomy. A. Occiput and first vertebra ofHolocentrus rufus (Walbaum), AMNH 35497. B. Anterior quartering view of the first vertebra of H. ascensionis (Osbeck), AMNH 22006, showing the single, continuous exoccipital facet and autogenous neural arch and spine. occipital, initially by ligament (figs. 16, 17); (figs. 19, 20); and (5) finally the full devel- (4) the growth posteriorly of the exoccipital opment of bone-to-bone condylar articula- into this ligamentous network as in Chlo- tions between the occipital region and facets rophthalmus and at least one myctophid (fig. on VI. 16) accompanied by the growth of autocen- Primitively, V1 in ctenosquamates has a tral prezygapophyses toward the exoccipitals neural arch with a ventral cartilage tip seated (fig. 18) and the closure of the gap by the in dorsal depressions in the centrum (figs. 4- basioccipital and the body of the centrum 6), or is at least sutured, rather than anky- 24 AMERICAN MUSEUM NOVITATES NO. 2827

ENR

BOC \ %RET DORS BOF TENDON FIG. 20. Holocentrid cervical anatomy. A. Holocentrus rufus (Walbaum), AMNH 35456, first two vertebrae, showing, in lateral view, the exoccipital facet and autogenous neural arch and spine on V1. B. H. rufus, AMNH 35477, showing the first two vertebrae in normal articulation with the occiput. The tendon for the retractor dorsalis muscle is shown on V2 in A and B. losed with the autocentrum. There is some leucopsarus shows the arch to be completely question about the generality of that last ankylosed with the centrum. Yet another statement, since my specimen of adult Myc- myctophid, Rhinoscopelus, is consistent with tophum aurolaternatum (fig. 17B) shows the Jollie's figure ofLampanyctus (fig. 16A). The suture line clearly, whereas Jollie's (1954) fig- neural arch of V1 is unfused in holocentrids ure 22 of the first vertebra of Lampanyctus (figs. 19, 20) and primitive "percoids" such 1985 ROSEN: EUTELEOSTEANS 25

FIG. 21. Occipital region and first two vertebrae of Haemulon album Cuvier, AMNH 30827. In A, the occipital condyle is shown from a slightly ventral orientation (inset) to clarify condyle shape. Vertebrae (C and D) in anterior view. Neural arch on VI autogenous. as haemulids, gerreids, lutjanids, and lethri- (1964) earlier supposition that the acanthop- nids (figs. 21-28), whereas it is ankylosed in terygians might not represent a monophyletic stephanoberycoids and many apomorph group. groups of "perciforms," recalling Patterson's The taxonomic implications of occipital 26 AMERICAN MUSEUM NOVITATES NO. 2827

EXO

VI V2 FIG. 22. Occipital region and first two vertebrae of Gerres cinereus (Walbaum), AMNH 21732. In A, the occipital condyle (inset) is shown from a slightly ventral orientation to clarify condyle shape. B. Lateral view as in A. Vertebrae (C and D) in anterior view. Neural arch on V1 autogenous. anatomy do not stop there, however. Prim- delicately interdigitating suture (fig. 29A). A itively, the exoccipital condyles meet in the similar arrangement has also been found in midline just above the notochordal canal. In trachichthyids and berycids. Reference to holocentrids (fig. 29A), they are united by a Patterson's (1964) account of the acanthop- 1985 ROSEN: EUTELEOSTEANS 27

AUITNA

EXOF C

BOF FIG. 23. Occipital region (A, B) and first vertebra (C) of Pomatomus saltatrix (Linnaeus). Neural arch on Vl autogenous.

terygian fishes of the English Chalk shows as an elongate rectangle that forms a concave that the pattern formed by these two condyles arc posteriorly. A single convex facet on the is very old and occurs in such primitive crea- first cervical vertebra of similar shape artic- tures as tAulolepus typus. The arrangement ulates with these medially united condyles. produced by these condyles is best described The exoccipital condyles retain this basic form 28 AMERICAN MUSEUM NOVITATES NO. 2827

EXO

EXOC

A BO

PZYGF (EXOF) FIG. 24. Occipital region and first two vertebrae ofLutjanus campechanus (Poey). A. AMNH 21632. B. AMNH 21688. In B, the autogenous neural arch ofVI had fallen offand was lost during preparation. and remain medially united in a variety of features that are shared with at least some fishes that have been regarded as "basal per- "berycoids" (e.g., holocentrids): the medially coids" (see figs. 21-28, 29B, C). Thus, the united condyles forming a long concave rect- "percoid" condition includes two primitive angle and an autogenous neural arch on the 1985 ROSEN: EUTELEOSTEANS 29

DNA DNA

BOF PZYG FIG. 25. Anterior (A) and lateral (B) views of the first vertebra of Lutjanus campechanus (Poey), AMNH 21632. The autogenous neural arch was removed to show the articular wells on the dorsal surface.

first cervical. Derived transformations of the premaxillae (fig. 40A), and hypothesized these conditions involve lateral displacement that the elements are a synapomorphy of sal- ofthe condyles accompanied by an alteration monids and neoteleosts. But not all salmo- of their rectangular shape and the shape and nids, or even salmonines, have such carti- position of the articular facets on the first lages (I have not found one in any species of cervical vertebra (figs. 29D, 30-32), and an- Salvelinus) and at least one osmeroid, Os- kylosis of the neural arch with the centrum merus mordax, does have them (fig. 41A). of this vertebra (figs. 33, 34). A comprehen- Among neoteleosts, however, Chlorophthal- sive survey needs to be made before the taxo- mus has them well developed in connection nomic usefulness of this area can be judged with a series of premaxillary processes (figs. fully, but, in at least two cases, modifications 40C, 41B) almost certainly synapomorphic ofthe occiput and first vertebra suggest affin- for a group defined previously as the Euryp- ities between major fish groups that were ma- terygii (Rosen, 1973). One stomioid also has jor components of the Paracanthopterygii a Chlorophthalmus-like upper jaw with a (figs. 35-39). In each of the batrachoid-lo- similar, though slight, indication of cartilage phiiform and ophidiiform-gadiform groups, development. By this, I mean that it stains the exoccipital condyles primitively receded prominently for mucopolysaccharides with from the posterior occipital margin and con- alcian blue, as does cartilage, and the stained sist ofwidely separated, cartilage-filled tubes area appears to contain a few cells as well as to which prezygapophyses from the first cer- connective tissue fibers (fig. 40B). This sto- vical articulate. In all ofthese fishes the neu- mioid is Maurolicus muelleri, but Fink and ral arch and spine ofthe first cervical vertebra Weitzman (1982) treat this as homoplasious is ankylosed to the centrum and both are on the grounds that Maurolicus is a derived firmly joined to the back end of the skull. member of an apomorph group of stomioids THE ROSTRAL CARTILAGE: The rostral car- (the sternoptychids)-ruling out the possi- tilage is another feature that has been given bility of a retained primitive feature. prominent attention as a neoteleostean syn- They claimed, however, that in Diplophos apomorphy. Here again, Fink and Weitzman there is a rostral cartilage which, in their view, (1982) have found paired elements in some corroborates the placement ofstomioids with salmonines and one coregonine that are small neoteleosts. What they illustrated in this in- discs of cartilage affixed to the inner face of stance, though, is a flat, median domino- 30 AMERICAN MUSEUM NOVITATES NO. 2827

EXO

EXOC

BOC

NAPZYG

FIG. 26. Occipital region and anterior two vertebrae ofLethrinus sp., AMNH 30872. In A the condyles are shown from a slightly ventral orientation (inset) to clarify condyle shape. Neural arch of VI autogenous.

shaped element not firmly bound to the a long cylinder surrounded by connective tis- premaxillaries or oriented like the median sue, so that as redescribed it still does not fit element in neoteleosts. But S. Weitzman in- the anatomical requirements of a neoteleos- forms me (in litt.) that the cartilage is actually tean rostral cartilage. 1985 ROSEN: EUTELEOSTEANS 31

- EXOC BOC

EXO

EXOF

BOF

FIG. 27. Occipital region and anterior two vertebrae ofSeriola sp., AMNH 30856. In A, the condyles are shown from a slightly ventral orientation (inset) to clarify condyle shape. Neural arch of VI autogenous.

Nevertheless, Fink and Weitzman (1982) cartilages transform into the median element deserve much credit for making these initial ofneoteleosts is suggested by a series ofstruc- observations ofthe cartilaginous skeleton and tures found amongst neoscopelids and some thereby opening up new avenues for profit- ctenosquamates (figs. 24-25). (K. Sulak, per- able research. sonal commun., informs me that paired car- The manner in which paired premaxillary tilages occur also in Ateleopus, a fish that has 32 AMERICAN MUSEUM NOVITATES NO. 2827

AUTNA NAPZYG

ENR

FIG. 28. Occipital region and anterior two vertebrae of Archosargus probatocephalus (Walbaum), AMNH 21663. In A, the condyles are shown from a slightly ventral orientation (inset) to clarify condyle shape. Neural arch of VI autogenous. defied, placement so far-see comments in appears to form as an extension of chondri- Rosen, 1973.) But within neoscopelids (figs. fication into the interpremaxillary ligament, 42, 43A), the paired elements retain their firm or perhaps is best described as a sesamoid association with the premaxillary ascending cartilage that develops in response to the processes and a median element with which stresses on that ligament. In all events, in they become fused in Neoscopelus (fig. 42B) Neoscopelus, what appear to be three distinct 1985 ROSEN: EUTELEOSTEANS 33

FIG. 29. Comparison of primitive and derived acanthopterygian occipital regions. A. Holocentrus ascensionis (Osbeck), AMNH 22086, showing the planar exoccipital condyles suturally united medially. B. Centropomus undecimalis (Bloch), AMNH 28058, showing the laterally expanded exoccipital condyles retaining the medial suture. C. Morone chrysops (Rafinesque), AMNH 22528, showing condition as in B. D. Sebastes sp., AMNH 36935, showing a more derived condition in which the exoccipital condyles are displaced laterally. cartilages become one, and apparent evi- terior view resembles a somewhat reduced dence of a tripartite history may still be de- version ofthe Neoscopelus structure (fig. 43A). tected in the structure of some other cteno- If the foregoing analysis is correct, a new squamates (figs. 43B-D). Aulopus (fig. 41C) cladogram for the Euteleostei is needed. Fink appears to have a rostral cartilage that in pos- and Weitzman wished to extend the resolved 34 AMERICAN MUSEUM NOVITATES NO. 2827

AUTNA

EXOF

FIG. 30. Occipital region and first vertebra of Rachycentron canadum (Linnaeus), AMNH 22135, showing laterally displaced exoccipital condyles. Neural arch on V1 is still autogenous, however.

part ofthis character-state tree to include the good case can be made for the Osmeridae, sister group of the Neoteleostei, which they which incidentally show the same spottiness identify as the Salmonidae. But they base this of character distribution. For example, Spi- conclusion on features not diagnostic of all rinchus shows the primitive neoteleostean salmonids. The difficulty is that an equally cervical gap between the occiput and the first 1 985 ROSEN: EUTELEOSTEANS 35

EXOF

BOF FIG. 31. Occipital region and first two vertebrae in Arripis trutta (Bloch and Schneider), AMNH 21632, showing laterally expanded exoccipital condyles just barely in contact in midline. Neural arch on Vl autogenous. vertebra (figs. 1 1A, 12C) and has, as a con- egonine I have seen that shows the condi- sequence, exoccipitals that have exposed pos- tion). But if we must make a choice from terior condyle-like faces, and Osmerus mor- amongst the old "salmoniforms" for a neo- dax has the paired premaxillary cartilages (fig. teleostean sister group, then the Osmeroidei, 41A) developed to the same extent as they or at least Spirinchus, is a better choice than are in Prosopium williamsoni (the only cor- salmonids on three counts: 36 AMERICAN MUSEUM NOVITATES NO. 2827

EXO

NAPZYG

EXOF

FIG. 32. Occipital region and first two vertebrae in Coryphaena hippurus Linnaeus, AMNH 21750, showing laterally displaced exoccipital condyles and a still autogenous neural arch on V1.

1. All osmeroids have a modified neural (see Vladykov, 1962, figs. 1-3) when the as- spine definitely associated with the second sociation ofspine and centrum is ofa definite preural centrum which is reduced in height sort [although the spine is short in corego- in virtually all cases and fitted with laminar nines (fig. 44) and might be primitive for the bone fore and aft, resembling this spine in Salmonidae if that group is monophyletic]. aulopids. Salmonines, in contrast, have a full, IfFink and Weitzman (1982) and Fink (1 984) strong spine on the second preural centrum are correct in aligning galaxioids with os- 1985 ROSEN: EUTELEOSTEANS 37 meroids, then a similar problem arises in the of inconsistencies, some more obvious than osmeroid complex since galaxioids, but not others, involving the rostral cartilage, the oc- osmeroids, have a full spine on PU2. Which cipital region ofthe skull, the caudal skeleton, state is primitive for such an osmeroid com- and the muscle that retracts the dorsal gill plex?6 arch elements (discussed below). 2. The premaxillary has an alveolar process ROSTRAL CARTILAGE: Among primitive of sorts (fig. 41A) under which there are no ctenosquamates there is a conflict arising from maxillary teeth (in other words, they show a the fact that myctophids, but not neoscope- simple tandem arrangement of these bones lids, always have a single median cartilage as opposed to their more primitive serial and a subocular shelf, as in Acanthomorpha. alignment in salmonids (fig. 40A). This conflict is resolved by simply admitting 3. The osmeroids (as defined by Fink and that the group Myctophidae + Neoscopeli- Weitzman) so far studied (a species of Os- dae = Myctophiformes, is unnatural, and merus and Galaxias) have acellular bone in should be so represented in the cladogram common with neoteleosts (Parenti, MS), (fig. 45). Stiassny (Ms) treats the group as nat- whereas the salmonids studied have a mix- ural based on several synapomorphies, how- ture of primitive cellular and some acellular ever.2 bone. Thus, the relatively smaller, usually def- OCCIPITAL REGION: The capricious manner initely shortened spine in osmeroids, a sim- in which the tripartite occipital joint is pres- ple form oftandem upperjaw bone alignment ent or not (except within the Neoteleostei and acellular bone, are synapomorphies that where it is diagnostic) has already been al- osmeroids, but not salmonines, share with luded to, and has several possible taxonomic the Neoteleostei. implications: (1) the Salmonidae is nonmonophyletic; (2) the Osmeridae is nonmon- CHARACTER CONFLICrS ophyletic; (3) the Clupeiformes and/or Clu- peomorpha are nonmonophyletic, although The evidence gathered so far for overall the significance of a triple joint in Denticeps euteleostean classification involves a number is overriden by the congruence ofmany other characters (Grande, MS); (4) the presence of a tripartite joint in Lepidogalaxias (Fink, 6 Fink and Weitzman (1982) and Fink (1984) argued for a linkage between "osmeroids" and galaxiids mainly 1984) might be one more example of char- on the grounds that both possess a distinct row of teeth acter capriciousness ifRosen (1974) was cor- dorsomedially on the mesopterygoid. Pterygoid teeth are rect in linking that genus with the pikes and primitive for teleosts, and when the mesopterygoid patch mudminnows.7 is reduced, it usually leaves such a row in the dorso- THE RETRACTORES ARCUUM BRANCHI- medial position (e.g., in Pterothrissusgisu and the Eocene ALIUM MUSCLE: It has been known for some tDiplomystus dentatus). Such teeth also are present in some primitive neoteleosts such as synodontids (K. Su- lak, personal commun.). Galaxiids show other plesio- 7Fink (1984) supports his argument that Lepidoga- morphous features not occurring in "osmeroids." These laxias is allied to neoteleosts by his discovery ofa dorsal include three, and in some cases, four small, accessory retractor muscle in this fish. And Fink and Weitzman rays in advance of the dorsal fin. And, in the caudal (1982) disagree with my earlier (1974) alignment of it skeleton, neural arches, and often spines, are associated with esocoids, partly, and I think correctly, on the grounds with the first ural and first preural centra (cf. Rosen, that my comparison of anterodorsal outgrowths of its 1974, figs. 18, 19, and illustrations in Patterson and Ro- first uroneural with those ofesocoids leaves a great deal sen, 1977). Small, but well-ossified, posteriorly cocked, to the imagination. An alternative placement of Lepi- neural arches also occur in some "salmonid" species, dogalaxias is suggested by the resemblance ofits cephalic especially coregonines, directly over or adjacent to the sensory pit-lines with those of Dallia (Nelson, 1972, p. first preural centrum and on both the first ural and first 38). A case might also be made for its original placement preural in a variety of primitive teleosts. These are part as a galaxiine (see Rosen, 1974) based on dorsal and of a derived caudal feature of tailed elopomorphs (Pat- caudal fin anatomy and position, or even with its align- terson and Rosen, 1977, cf. figs. 23, 27, 35, 36). Fink's ment to a Novumbra-Dallia esocoid subgroup defined (1984) summary ofthe problems of"lower" euteleostean by having but a single epural and uroneural in the caudal classification is an exemplary statement illustrating the skeleton. Fink's (1984) statement that the "position of need for comprehensive character surveys. Lepidogalaxias is controversial" is unarguable. 38 AMERICAN MUSEUM NOVITATES NO. 2827

ANA

EXOF

FIG. 33. Occipital region and first two vertebrae in Echeneis naucrates Linnaeus, AMNH 21844, showing enlarged lateral exoccipital condyles and the neural arch on VI ankylosed. time that the retractor dorsalis, as this muscle muscles has been investigated in some detail has been called, has a sporadic occurrence in relation to the monophyly at the Aulopi- among halecomorph fishes, which Rosen formes by Stiassny (work in progress), whose (1973) commented upon earlier. Most re- conclusions are consistent with my own pres- cently, Fink (1984) has identified one in Lep- ent and prior observations. A difficulty in idogalaxias salamandroides. The problem of character interpretation arises when this fea- the homology of these different retractor ture is used if one considers that Aulopus, 1985 ROSEN: EUTELEOSTEANS 39

BOC

a_~~~1J ANA

BOF FIG. 34. Occipital region and first vertebra ofStrongylura marina (Walbaum), AMNH 27805, showing laterally displaced exoccipital condyles and the neural arch on VI ankylosed.

Chlorophthalmus, and neoteleosts, in gener- acter in more detail. It appears possible that al, are characterized by a very short retractor stomiiforms and alepisauroids are linked by that inserts on the inner faces ofthe third and a peculiar type of dorsal retractor, although fourth pharyngobranchials, as opposed to the each of these groups are individually diag- more posterior insertion in some stomi- nosable, the stomiiforms by the peculiar type iforms and alepisauroids on the fourth pha- of photophores (Fink and Weitzman, 1982) ryngobranchial, which often bears a single and at least some alepisauroids by the ex- large toothplate with recurved, fanglike teeth. tremely attenuated second pharyngobranchi- Stomiiforms have a variety ofspecialized re- al described by Rosen (1973). tractor origins, including the shoulder girdle THE POSITION OF AuLoPus AND IMMEDI- and ribs as in sternoptychids but the condi- ATELY ALLIED FoRMs: The title of this sub- tion in Aulopus, Chlorophthalmus, and myc- section is phrased somewhat ambiguously for tophids with the anterior origin and double a good reason-no one knows with any de- pharyngeal insertion appears to be an ad- gree of certainty what those immediately al- vanced neoteleostean synapomorphy. Here I lied forms might be, but perhaps they are simply defer to Stiassny's work in progress bathysaurids, bathypteroids, and ipnopids. which will deal with the analysis ofthis char- But whoever they are, later collectively termed 40 AMERICAN MUSEUM NOVITATES NO. 2827

F VIANA

N Y XPZYG VI ANA -EXO' /~~~~

VIANA ~ ViAN

NAPZYG ~ ~ ~ ~ I N EXO ~ ~ ~ ~ X

BO~ ~ ~ ~ ~ ~ B

E=0

FIG. 35. Comparison ofthe occipital region and anterior vertebrae in a gadiform and some members of the batrachoid-lophioid clade. A. Gadus morhua Linnaeus, AMNH 21680. B. Porichthys notatus Girard, AMNH 22432. C. Opsanus tau (Linnaeus), AMNH 21564. D. Lophius americanus Valenciennes, AMNH 22129. Note integration of neural arch on Vi with the occipital bones, the sharply angled autocentral prezygapophyses, and their integration with a complex exoccipital condyle. Both the V1 prezygapophysis (exoccipital facet) and the exoccipital condyle are formed around cores of cartilage so that the condyle-facet contact is cartilage-to-cartilage. Compare with figures 36 to 38.

"aulopoids," at least Aulopus may be treated basis of its (1) high-set pectorals, (2) subtho- as a sister group to the Ctenosquamata on the racic pelvics, (3) ctenoid scales, (4) three pre- 1985 ROSEN: EUTELEOSTEANS 41

ENR

PZYG FIG. 36. Occipital region and anterior vertebra in "ophidiiforms." A. Ophidion holbrooki (Putnam), after Rose (1961). B. Ogilbia cayorum Evermann and Kendall, AMNH 26098. A, an ophidioid has a relatively primitive, planar exoccipital facet whereas B, a bithytoid, has paired, codlike facets. Both show the more posterior position of the basioccipital facet.

dorsal (or supraneural) bones, (5) a reduced of spiny-finned euteleosts, and of advancing spine on NPU2, (6) derived premaxillary a specific proposal that could be addressed, morphology (see Rosen and Patterson, 1969), or even openly attacked. There was not long (7) type of dorsal retractor muscle, (8) a to wait after the initial publication ofthe 1966 toothplate fused to the third epibranchial general classification of modem teleosts. In (Rosen, 1973), and (9) median rostral carti- fact, its first reviewer, Carl Hubbs, referred lage. to it as a bizarre collection ofodd bedfellows. STATUS OF THE PARACANTHOPTERYGII: Of Rosen and Patterson (1969) rooted the all the proposals for subdividing the old problem firmly by producing a monograph Acanthopterygii or Percomorpha, none has whose sole, and now seemingly unfortunate been more controversial, or, in my present purpose, was to entrench the problem by col- view, ill-fated, than the Paracanthopterygii. lecting "confirmatory" evidence of paracan- The taxon was proposed partly in an effort thopterygian substance. Now, the spirit ofthe to narrow the enormous scope of the taxo- times has changed, and neither Patterson nor nomic problem posed by 6000 or so species I would consider that a worthy purpose. In 42 AMERICAN MUSEUM NOVITATES NO. 2827

N. W EXOC

BOC A

BOC FIG. 37. Occipital regions of a gadiform and members of the batrachoid-lophiiform clade. A. Gadus morhua Linnaeus, AMNH 21680. B. Porichthys notatus Girard, AMNH 22432. C. Opsanus tau (Lin- naeus), AMNH 21564. D. Lophius americanus Valenciennes, AMNH 22129. Note the small, laterally displaced, cartilage-filled, tubelike exoccipital condyles. In D, the cartilage core is hidden by lateral wings of bone. fact, my object now is to show that Hubbs' that some of its constituents can no longer criticism was well taken, and not only that be accepted as natural. I refer particularly to the paracanthopterygians no longer can be the percopsiforms that have never had as de- accepted as a natural group, but to point out fining traits anything but shared primitive 1985 ROSEN: EUTELEOSTEANS 43

BOF

EXOF C BOF

FIG. 38. The first vertebra in a gadiform and two batrachoids to illustrate the very derived autocentral prezygapophyses. A. Gadus morhua Linnaeus, AMNH 21680. B. Porichthys notatus Girard, AMNH 22432. C. Opsanus tau (Linnaeus), AMNH 21564. features. Aphredoderids do align themselves Percopsids are, of course, euteleosts be- with amblyopsids on the basis ofthe thoracic cause they possess that group's single defining anus and segmented premaxilla, but nothing trait, the adipose fin. And they are, presum- of which I am aware properly unites them ably, some form of primitive ctenosquamate with percopsids. In fact, I find it difficult even because of their premaxillary processes, ros- to state features which link the living and few tral cartilage, pectoral fin position, and fin fossil percopsids-apart from the broadly spines. At least one of them, Percopsis omis- arched alveolar premaxillary process. comaycus, is known to have a complex series 44 AMERICAN MUSEUM NOVITATES NO. 2827

ENR

PZYG

FIG. 39. Comparison of first two vertebrae in a gadiform and two acanthopterygians. A. Gadus morhua Linnaeus, AMNH 21680. B. Centropomus undecimalis (Bloch), AMNH 28058. C. Sebastes sp., AMNH 36935. Note that exoccipital facets are ofautocentral origin, following the angle ofa neural arch prezygapophysis on either VI or V2. ofjaw muscles found also in some gadoids, frequency that investigators would be fool- but as Dietz (1914) pointed out, so do some hardy to base major taxonomic judgments liparids. The one thing they share that is de- upon it unless we could formulate an argu- rived in relation to current pharacter inter- ment involving a unique ontogeny that doc- pretation is, a full spine on PU2, but that uments the redevelopment of a full NPU2. character seems to come and go with such No such empirical ontogenetic data yet exist. 1985 ROSEN: EUTELEOSTEANS 45

PMX CART

PMX .* B

PMX CART ASC PR ARTPR

/ PTMX PR

PMX MX FIG. 40. Euteleostean upper jaw bones in medial view. A. Salmo gairdneri Richardson, AMNH 40268, to show small, adherent, premaxillary cartilage. B. Maurolicus muelleri (Gmelin), AMNH 37329. C. Chlorophthalmus agassizi Bonaparte, AMNH 40812. Note the serial (qnd-to-end) alignment of the two bones in A and their tandem (overlapping) alignment in B and C. Note also the distribution and shape of premaxillary processes in B and C. Cartilage, showni by the presence of black dots, is inferred in B because ofspecific alcian blue staining for mucopolysaccharides and the presence ofsome cellulanrty within the fibrous tissues engulfing the ascending process. Compare C with figure 4113.

All "percopsiforms" are primitive with re-, rived in 'having enlarged infraorbital canal spect to the presence of an unconsolidated bones not supporting a subocular shelf, but second ural centrum, an adipose fin, and more the latter might also be primitive or simply than 15 branched caudal rays. They are de- homoplasious since similar infraorbitals oc- 46 AMERICAN MUSEUM NOVITATES NO. 2827

PMX CART

- I/

R CART

PAL

FIG. 41. Euteleostean upperjaw bones in dorsal view. A. Osmerus mordax (Mitchill), AMNH 40726. B. Chlorophthalmus agassizi Bonaparte, AMNH 40892. C. Aulopusjaponicus Gunther, AMNH 28635. Note paired premaxillary cartilages in A and B; those in A are similar to ones observed in Prosopium williamsoni and P. cylindraceum, except that in the latter two taxa the cartilages are only loosely associated with the premaxillae rather than firmly adherent to them as in S. gairdneri (fig. 40A) or 0. mordax (A). The cartilage depicted in C is similar to that shown in figure 43A. cur in stephanoberycoids. In their dorsal gill underused thesis ofconsiderable breadth and arches they lack an interarcual cartilage, orig- insight. The interarcual cartilage has since inally shown sometime ago to be present in been found to be absent in other myctophids, other primitive ctenosquamates such as myc- however (Stiassny, MS). As for the polymix- tophids by Malcolm Jollie (1954) in a sadly iids, the most interesting thing I can think of 1985 ROSEN: EUTELEOSTEANS 47

PMX

MED CART- MED CART

A B

PMX CART

MED CART - MED CART

FIG. 42. Rostral cartilages ofneoscopelids. Anterior to left in A, right in C; B, three-quarters posterior view; D, posterior view. A and B. Scopelengys dispar Garman, AMNH 12841. C and D. Neoscopelus macrolepidotus Johnson, MCZ (uncat.). B and D are posterior views, B tumed slightly to the right. In A and B the paired and median elements are joined by connective tissues with only the slightest amount of chondric invasion. In C and D the necks that join the three elements are fully chondrified. C and D based, in part, on information from M. Stiassny.

saying about them is that they look like huge ed subocular shelf like that of myctophids. spiny myctophids but are closer to the orig- They are autapomorphic (i.e., diagnosable) inal percomorph assemblage because the an- in possessing numerous dorsal fin spines, a terior vertebra has developed ossified auto- full spine on PU2, and a pair of long mental central prezygapophyses that articulate cirri that give them their name ofbeardfishes. directly with the exoccipital condyles. They Clingfishes are clearly autapomorphic in also are primitive in retaining a free second many traits and appear to have dorsal gill ural centrum, a first abdominal vertebral cen- arches (fig. 58B in Rosen and Patterson, 1969) trum that does not contact the basioccipital, like those oftropical blennies. This latter fea- many branched pelvic rays, more than 15 ture is difficult to use decisively because it branched caudal rays, and an undifferentiat- involves the loss and reduction of so many 48 AMERICAN MUSEUM NOVITATES NO. 2827

R CART

A (

R CART

A D

FIG. 43. Ctenosquamate rostral cartilages. A. Neoscopelus macrolepidotus Johnson, MCZ (uncat.) (from a sketch by M. Stiassny). B. Notoscopelus resplendens Richardson, AMNH 29528. C. Polymixia lowei Gunther, AMNH 49674. D. Scopeloberyx sp., AMNH 40268. Note that indications of a tripartite origin (as per fig. 42) are present in all. A, B, C (left), and D (right) are posterior views. C (right), a dorsal view, anterior to left. D (left), a lateral view, anterior to left. Fink (1984) has reported that in a cichlid species (Acanthopterygii), the rostral cartilage has only a dual origin ontogenetically from paired premaxillary cartilages.

gill arch elements; their relationship to dra- gested by Gosline (1971). The latter (i.e., cal- conettids and callionymids has been sug- lionymids) have similar dorsal gill arches. The 1985 ROSEN: EUTELEOSTEANS 49

UN 1 EP

HYP 6 ,P HYP 5

FIG. 44. Caudal skeleton of Prosopium williamsoni (Girard), AMNH 37967. The small, recurved, neural arch and spine element on PU2 appears to be a derived feature that occurs in coregonines, and some salmonines and is variably associated with the first or second preural centrum (PU1 or PU2). Within the Salmoninae a usual condition is to have a full spine associated with PU2, whether or not a reduced, recurved element joins it on that centrum. The primitive eurypterygian neoteleost condition, however, is to have a spatulate or broadly lance-shaped, somewhat or greatly reduced spine on PU2 (Rosen, 1973, figs. 46 to 48) and this is true also of osmeroids (excluding salangids) (e.g., Rosen, 1974, figs. 25, 26C, 27 and Greenwood and Rosen, 1971, fig. 16). More primitive neoteleosts such as stomiiforms have usually either a full spine on PU2 (based on a random survey of eight gonostomatids and sternoptychids) or, if somewhat shorter than that on PU3 (e.g., Diplophos, as figured by Fink and Weitzman, 1982), it is a narrow spine without platelike expansions. Spines like that of Diplophos occur also in some of the more derived paralepidids and evermanellids and full spines on PU2 occur in a variety ofacanthomorphs (e.g., in the "paracanthopterygii" and some acanthopterygians). The condition shown here for Prosopium might, therefore, serve as a synapomorphy for salmonids.

similarity does not, however, include the occipital region discussed above, and some closely united third and fourth epibranchials, features of the dorsal gill arches to be pre- the converging ventral gill arches, and unos- sented in a forthcoming paper. sified copula that clingfishes share with lo- Features that appear to unite batrachoids, phiiforms or, at least, lophiids and anten- lophiiformes, gadiforms, and bythitoids are nariids (fig. 58C in Rosen and Patterson, the relation between the occipital region and 1969), but not with callionymids. the first vertebra. The contact between the What appear to unite the batrachoid clade exoccipital facets and the prezygapophyses of with the gadiforms are derived features ofthe the first vertebra is between the cartilage cores 50 AMERICAN MUSEUM NOVITATES NO. 2827

37 - 38

12 - 14

11-5

FIG. 45. Proposed interrelationships of the main groups of clupeocephalans, based on the synapomorphies as numbered in the text. Uncertainties about the placement of a number of groups are represented by unresolved polychotomies. This scheme differs from that proposed by Fink and Weitzman (1982) and Fink (1984) mainly in the position of salmonids and "osmeroids" and in excluding Lepi- dogalaxias. or tips of these elements (figs. 35, 36B, 37, socids from the assemblage since they have 38, 39A), as compared with the bone-to-bone no derived chondrification associated with contacts in acanthopterygians and the bery- the exoccipitals and vertebral prezygapoph- coid-like arrangement in ophidioids (cf. figs. yses. 19, 20, and 29A with 36A). The other char- As mentioned earlier, it is primitive for acter which also includes ophidioids as well neoteleosts to lack a direct vertebral contact as bythitoids, is the articulation of pleural with the occiput, and it is this spatial sepa- ribs with ventrolateral cavities in the verte- ration that seems most closely correlated with brae normally occupied by parapophyses. the presence of the exoccipitals in the occip- The feature that might be synapomor- ital joint region. What characterizes the phous for gadiforms and at least some ophid- ctenosquamates is a gradual closing of this iiform subgroups is the position of the ex- space so that acanthomorph fishes, excluding occipital facets anterior to the basioccipital Polymixia, have the first vertebra firmly and the corresponding anterior extension of united with the occiput by one of two means the first vertebral prezygapophyses onto the (see, e.g., an atherinomorph, fig. 34). One is back of the occiput to meet the exoccipitals. unique to acanthomorphs and this is the for- Neural arches are carried forward along with mation of autocentral prezygapophyses that the prezygapophyses, in some cases firmly grow forward to contact the exoccipital con- incorporated into the exoccipitals and the su- dyles. The other is typical ofseveral primitive praoccipital (figs. 47-49A). These vertebral acanthomorphs (holocentrids, berycids, and features, alone, exclude zoarcids and gobie- ophidioids, for example). Here, the articu- 1985 ROSEN: EUTELEOSTEANS 51 latory surface on the first vertebra for the The significance ofthis conclusion is that the exoccipitals is a more-or-less continuous, monophyly of both the Salmonidae and Os- planar surface (figs. 19B, 36A) that fits neatly meroidei should be reevaluated, and that against an opposing exoccipital surface clear- much more detailed character surveys are ly divided by suture into right and left halves needed-undertakings that lie outside the (fig. 29A). scope of this study. Paired rostral cartilages In fishes that have been previously called in their transformed or untransformed states acanthopterygians or percomorphs, this type are present consistently in members of Ro- ofcontact has been altered by the lateral dis- sen's (1973) Aulopiformes, even though that placement of the two exoccipital facets and taxon must now be abandoned as monophy- their growth backward over the basioccipital letic since Aulopus appears to have a rostral (as in Centropomus and Sebastes, fig. 29B, cartilage synapomorphous with that ofcteno- D). What characterizes the batrachoids, gad- squamates that is lacking in Chlorophthal- iforms, and some ophidiiforms is that (1) the mus. However, I can find no obstacle to mak- exoccipital facets have moved laterally to an ing a transformation sequence between the exceptional degree, as noted by Rosen and salmonine occipital region and that of Patterson (1969), and usually have a deep "higher" euteleosteans-unless the Osme- core ofcartilage (fig. 37); and (2) the vertebral roidei is nonmonophyletic and Spirinchus is prezygapophyses extend well forward onto the a more appropriate immediate sister group braincase, in many cases, carrying a neural to the neoteleosts. The latter possibility will arch and spine component with them to meet depend on two kinds ofevidence: (1) that the the epioccipitals, supraoccipital, and the dis- gap between the occiput and the first cervical placed exoccipitals (fig. 35). vertebra is truly diagnostic for primitive neo- Even in the ophidiiforms that retain the teleosts, including stomiiforms, and (2) that primitive, continuous, planar surface for ex- the postoccipital gap in Spirinchus is synapo- occipital contact, these prezygapophyses ex- morphous, rather than homoplasious, with tend well forward over the centrum as clearly that of neoteleosts. illustrated by Rose (1961) in Ophidion hol- If the occipital anatomy of Spirinchus is brooki (fig. 36A). Some of the other ophidi- primitive for osmeroids, as Fink and Weitz- iforms (the bythitoids) are much more cod- man (1982) claim the anatomy of Diplophos like (see fig. 36B, which shows separate right is for stomiiforms, then I would be forced to and left vertebral facets). And, since ophid- place osmeroids, stomiiforms, and neote- iiforms fall readily into two classes based on leosts in an unresolved trichotomy and to caudal anatomy, gill arches, viviparity and, exclude the salmonids on two grounds (cau- to some degree, fin structure (Cohen and dal skeleton and premaxillary anatomy), ex- Nielsen, 1978, and Patterson and Rosen, work cept perhaps as the sister group to those three in progress), there is an implication that the if the Salmonidae is monophyletic and the ophidiiforms might be nonmonophyletic. One salmonine premaxillary cartilages and occip- group, the more derived in anatomy (bythi- ital anatomy are synapomorphies. Mean- toids) might be linked to cods and batra- while, however, I conclude that the type of choids, and the other (ophidioids) could be jaw, caudal skeleton, and bone histology earn the sister group to the whole lot. the Osmeroidei a closer linkage with stomiiforms and eurypterygians than do these same SUMMARY AND CONCLUSIONS features in salmonines. According to evidence presented here, the Below is a synapomorphy scheme of the significance of the paired cartilages and tri- Clupeocephali representing the data pre- partite occipital condyles found in some sal- sented here and by other workers which sum- monids is ambiguous because of the uneven marize anatomical findings relevant to the distribution of the characters, the lack of a ctenosquamates and the included acantho- good theory ofrelationships among the taxa, morphs. the occurrence of both features in some os- Clupeomorphs share the following derived meroids, and a parsimony argument that fa- features with euteleosts (Patterson and Ro- vors osmeroids as a neoteleost sister group. sen, 1977): 52 AMERICAN MUSEUM NOVITATES NO. 2827

1. Articular bone in the lower jaw co-os- completely so in galaxioids (see character sified with the angular. conflicts above). 2. Retroarticular in lowerjaw excluded from 11. In the caudal skeleton, NPU2 is shorter the joint surface. than NPU3 and bladelike. [NPU2 is also 3. A median (in clupeomorphs) or paired shorter than NPU3 in the stomiiform, (in euteleosts) anteriorly directed mem- Diplophos, but is not bladelike according branous outgrowth on the anterodorsal to Fink and Weitzman (1982, fig. 16).] margin of the first uroneural (the stegural). Stomiiforms and alepisauroids share with 4. Neural arch over U1, when present, re- other neoteleosts: duced, rudimentary, lying free over U1 12. The exoccipitals and basioccipital ex- or nestled against the posterior face of posed posteriorly and joined by an in- the neural arch ofPU 1, or ankylosed with verted Y-shaped suture. Presence of a other neural arch or uroneural com- cervical gap between the occiput and first plexes, or absent. vertebra. [When the first vertebra is re- 5. Toothplates, when present, fused with mote from the occipital region, it is cor- first three pharyngobranchials and fifth related with the exposure ofthe exoccip- ceratobranchial. ital surfaces as parts of the posterior Euteleosts, primitively, have but a single, occipital outline-the basioccipital hav- unambiguous, derived feature: ing dorsolateral depressions to accom- modate the exoccipitals in juvenile Os- 6. An adipose dorsal fin (not present in eso- merus (prevertebral space being occupied coids). The stegural feature (character 3) by the notochord and its connective tis- of the first uroneural has been used as a sue sheath and by a small, spineless neu- euteleost synapomorphy, but this feature ral arch).] The "osmeroid" Spirinchus seems properly to define only a subgroup also has the exoccipitals entering the oc- of euteleosts. It does not allow decisive cipital region posteriorly and future study inclusion ofthe esocoids [which, at best, may demonstrate that osmeroids are have an outgrowth of the first uroneural nonmonophyletic. of doubtful homology with the stegural 13. An interoperculohyoid ligament, present of other euteleosts (see illustrations in but feebly developed in stomiiforms (see Rosen, 1974)]. Lauder and Liem, 1983, p. 34). Within the Euteleostei, the Salmonidae (but 14. A retractores dorsalis branchialium not argentinoids or ostariophysans) share with muscle [in most stomiiforms and alepi- all other groups only: sauroids originating far back in the abdominal region, with fibers inserting 7. Paired stegural outgrowths of the first principally or only on the fourth gill arch uroneural. (Rosen, 1973), and perhaps nonho- 8. At least some parts of the endoskeleton mologous with that in eurypterygians, with acellular bone (Parenti, MS). according to work in progress by M. Stiassny].2 Osmeroids share with stomiiforms and primitive neoteleosts: Neoscopelids share with primitive ctenosquamates (e.g., myctophids): 9. Acellular endoskeletal bone (but this statement is based only on study of Os- 15. Ctenoid scales and median fin spines mems mordax. However, K6lliker (1859) (well developed in Cretaceous forms; see reported acellular bone in a galaxioid, Rosen, 1973, pp. 456-459). which is consistent with Fink and Weitz- 16. Premaxilla with ascending, articular, and man's (1982) realignment of galaxioids postmaxillary processes. with osmeroids (and see Parenti, MS). 17. Pectoral fins arising laterally rather than 10. A toothed alveolar process on the pre- ventrally. maxilla which lies under the maxilla, 18. Pelvic fins arising anteriorly in a sub- 1985 ROSEN: EUTELEOSTEANS 53

abdominal, rather than in the more pos- ture of apparently tripartite origin in terior abdominal, position. some species (and with a relic pair of 19. A hyoid bar and branchiostegals ofacan- lateral cartilages in a few species). thomorph type (see McAllister, 1968). 29. An interarcual cartilage present between 20. Clamshell-shaped saccular otolith with the first epibranchial and second pha- definitive cauda and ostium in sulcus ryngobranchial (small, when present, and (Nolf, MS). absent (according to N. Stiassny, person- 21. A pair of small, lateral cartilages joined al commun.) in some taxa. to the premaxillary ascending processes 30. No more than three predorsal bones. and to a larger median cartilage. 31. Partial closure of the cervical gap be- 22. Retractores dorsalis muscle short, orig- tween the occiput and first vertebra. inating anteriorly on basioccipital or cer- 32. Absence of an accessory neural arch in vical vertebrae and inserting on connec- the cervical region. tive tissues on dorsomedial edges ofthird 33. Presence ofneural arch prezygapophyses and fourth pharyngobranchials. This as- on the anterior vertebrae. sumes character 22 to be a more derived 34. Direct connection (via ligaments) of the (or simply nonhomologous) condition of autocentrum ofthe first vertebra with the character 14. exoccipital condyles. "Aulopoids" share with ctenosquamates: Polymixiids share with other acantho- 23. A median rostral cartilage without lat- morphs: eral components (Stiassny, personal commun., informs me that some myc- 35. Autocentral prezygapophyses between tophids retain small, lateral components, first vertebra and exoccipital condyles. so that this character would be stated 36. A pelvic fin spine. more accurately as: distinct lateral com- Other acanthomorphs share: ponents greatly reduced or absent.) 24. Pelvic fins more anterior in a subthoracic 37. Complete closure of the notochordal- position. (Fin spines are unknown in connective tissue space between the ba- modem "aulopoids," which could be in- sioccipital and the centrum of the first terpreted as an autapomorphic loss in vertebra. "aulopoids" or gain in neoscopelids.) 38. Primitively, a rodlike interarcual carti- 25. Three or fewer predorsal bones (four in lage between the first and second gill neoscopelids). arches [in some beryciforms (e.g. an- omalopids) and in most "percoids"]. Chlorophthalmids share with ctenosquamates: Within the above scheme I recognize three main areas ofuncertainty involving the align- 26. A subocular shelf (said by Stiassny, per- ment of alepisauroids and other members of sonal commun., to be very narrow and the Aulopiformes recognized by Rosen thus interpreted here as primitive rela- (1973). Unexpectedly, the neoscopelids seem tive to the much larger, subocular shelves to me now of uncertain status, although they in ctenosquamates). and the myctophids are usually closely linked. Chlorophthalmids exhibit an additional con- Lauder (1983) described a character, the flicting character in this alignment because, urohyal-third hypobranchial ligament, that in the upper jaw, they lack a median rostral he thought was present uniquely in members cartilage, but have well-developed paired ones of these two families, but the neoscopelid on the premaxillary ascending processes. ligamentous connection is more like that of Myctophids share with acanthomorphs: Polymixia.2 I have not searched elsewhere, and I reserve judgment on precise realign- 27. A subocular shelfthat extends well under ment ofneoscopelids until all relevant groups the eyeball (Rosen and Patterson, 1969, have been more thoroughly reviewed. fig. 11). The various character conflicts noted above 28. Rostral cartilage-a single median struc- extend to other components of the Aulopi- 54 AMERICAN MUSEUM NOVITATES NO. 2827 formes, and each ofthese is represented either myctophids (Travers, 1 98 1), but not oth- by an unresolved tri- or tetrachotomy. Per- ers and absent in neoscopelids and po- haps others (Fink and Weitzman, 1982; Fink, lymixiids. 1984) would prefer an additional unresolved 2. Acanthomorph type of hyoid bar and point involving salmonids and osmeroids, or branchiostegals (see McAllister, 1968) in simply to transpose the two. I cannot, how- chlorophthalmids but not in aulopids or ever, accept their evidence of the salmonine synodontids. occipital joint for the principal reason that 3. Presence of three occipital condyles in the salmonine occiput most closely resembles myctophids and neoscopelids, and, an advanced acanthomorph occiput rather among synodontids, in Synodus, but not than the basic inverted Y-junction of prim- in Saurida. itive neoteleosts and thus appears to be an 4. A cervical gap in primitive cteno- independent specialization of salmonines, squamates, but not in synodontids. which is nonhomologous with the basic neo- 5. Acanthomorph type of rostral cartilage teleostean arrangement. (fig. 43) in some myctophids, but not Chlorophthalmids and aulopids appear not others, and not in neoscopelids (fig. 42). to form parts of a monophyletic group be- 6. Paired premaxillary components of ros- cause the former have two features (charac- tral cartilage present in some salmonines ters 21 and 22) linking them with cteno- and coregonines, and not others; present squamates, and at least one alepisauroid, a in at least one osmerid. synodontid, also lacks character 21, and also 7. Cervical gap present, and exoccipitals character 22 according to McAllister (1968, entering occipital joint surface in some plate 12). Furthermore, some alepisauroids osmerids. have ethmoid cartilages resembling those of 8. In the caudal skeleton, a full spine on the stomiiforms (Stiassny, MS) and aulopids have second preural centrum in salmonines, a rostral cartilage ofacanthomorph, or at least but not coregonines. ctenosquamate, type. Such data effectively 9. Large, marginal basihyal fangs in sal- dissolve the Aulopiformes of Rosen (1973), monines and galaxioids, but not in thy- and Rosen's (1973) more inclusive taxon, the mallines or coregonines. Eurypterygii, will come to include the Cteno- 10. Row ofdorsomedial mesopterygoid teeth squamata and perhaps the Chlorophthalmi- in osmeroids, galaxioids, and also in dae and some "aulopoid." I have not restud- Pterothrissus, in the primitive clupeo- ied the alepisauroid fishes, however, and so morph tDiplomystus, and in the cteno- these are represented as unresolved in the squamate, Synodus. synapomorphy scheme and summary clado- 11. Neural spine on second preural centrum gram. reduced to a low crest except in Poly- There are numerous derived anatomical mixia and fishes formerly united as par- features of euteleostean fishes that appear in acanthopterygians and a few subgroups a confusing array of taxa which suggest the of acanthopterygians. need for much additional study and the re- In this extended essay on the taxonomy of definition oftaxonomic boundaries. Much of the single largest recognized group ofteleosts, the confusion has resulted from the assign- the Euteleostei, I have adopted a scheme of ment of "groups," such as "osmeroids" and relationships ofits subgroups differing some- "aulopiforms" that never were defined prop- what from my own and those of other recent erly to begin with. Within the Acanthomor- workers. My proposal raises the general ques- pha, improperly defined taxa are probably tion of why taxonomists periodically adopt the rule rather than the exception. At least novel taxonomies. Many years ago it was be- ten ambiguous features come immediately to cause taxonomists used symplesiomorphies mind among fishes that have been associated to describe groups, e.g., the Clupeiformes, at some time or another with the Neoteleos- Malacopterygii, Isospondyli, and Mesich- tei. thys. But the recent flux offish classifications has other causes, since most ichthyologists 1. Interarcual cartilage present in some have abandoned symplesiomorphy in favor 1985 ROSEN: EUTELEOSTEANS 55 of synapomorphy as a basis for defining resembles). But ambiguous or inconsistently groups. distributed characters still plague fish system- To me, at least, there appear to be three atics and are used again and again to support major reasons for this flux. One is that dif- some proposed relationship. Having been ferent taxonomists perceive characters dif- guilty of the crime myself, I am naturally ferently. Since our experiences do not coin- tolerant of others who commit it, allowing cide, neither do our perceptions. This is a that it is a far better thing to have proposed phenomenological problem of what each of and been rejected on grounds of nonhomol- us is able to see and how we interpret what ogy than never to have proposed at all. we perceive. And it is aggravated by a linguistic problem ofhow we describe those perceptions. A second main reason for the steady LITERATURE CITED flow of new taxonomic proposals has to do Agassiz, J. L. R. with the willingness of some taxonomists to 1858. The classification offishes. Proc. Amer. align groups based on a trait peculiar to one Acad. Sci., vol. 4, p. 108. or only a few ofits members, a trait that must Berg, L. S. be suspected ofbeing homoplasious unless it 1940. Classification offishes, both Recent and can be argued successfully to be primitive for fossil. Russian and English, Ann Arbor, the entire group. The tripartite occipital con- Michigan, Edwards Bros. dyle and paired rostral cartilages of salmon- Boulenger, G. A. ines would fall in this category (vide Fink and 1904. A synopsis ofthe suborders and families Weitzman, 1982, and Fink, 1984) as would ofthe teleostean fishes. Ann. Mag. Nat. Hist., vol. 7, pp. 161-190. the basihyal dentition of salmonines (vide Cavender, T. M., and R. R. Miller Rosen, 1974). And evaluation of an ambig- 1972. Smilodonichthys rastrosus, a new Plio- uous trait, too, is aggravated by a linguistic cene salmonid fish from western United problem ofhow the trait is described. A third States. Mus. Nat. Hist., Univ. Oregon, main reason is a genuinely new discovery of Bull. 18, 44 pp. such a major character that it seems obvious Cohen, D. M., and J. G. Nielsen and significant to one and all. An example of 1978. Guide to the identification of genera of this type would be the Weberian apparatus the fish order Ophidiiformes with a ten- ofotophysans or the recessus lateralis of clu- tative classification ofthe order. NOAA peiforms. The retractor dorsalis muscle Tech. Rept. NMFS Circular 417,72 pp. one of these characters when it Dietz, P. A. seemed like 1914. Beitrage zur Kenntnis der Kiefer- und was first reported and described in acantho- Kiemen-bogenmuskulatur der Teleos- morphs by Dietz in a series of papers (see tier. I. Die Kiefer- und Kiemen-bogen- Rosen, 1973) and by Holtzvoogd (1965), but muskeln der Acanthopterygier. Mitt. the retractor muscle appears widely amongst Zool. Staz. Neapel, vol. 22, pp. 99-162. halecomorph neopterygians and its homo- Fink, S. V., and W. L. Fink plasious nature made application ofthe char- 1981. Interrelationships of the osteriophysan acter ambiguous at various hierarchical levels fishes (Teleostei). Jour. Linn. Soc. (Nelson, 1967). There appear to be only two (Zool.), vol. 72, no. 4, pp. 297-353. of with this One is Fink, W. L. ways dealing ambiguity. 1984. Basal euteleosts: relationships. In Spe- to conduct an empirical study of the feature cial Publ. no. 1, Amer. Soc. Ichthyol. in different taxa showing that its ontogeny is and Herp. Ontogeny and Systematics of unique in some ofthose taxa and that its final Fishes, pp. 202-206. expression is, therefore, homoplasious (i.e., Fink, W. L., and S. H. Weitzman irrelevant because of nonhomology). The 1982. Relationships of the stomiiform fishes other is to study many other features that (Teleostei), with a description of Diplo- phos. Bull. Mus. Comp. Zool., vol. 150, yield congruent hierarchical arrangements- no. 2, pp. 31-93. a congruence which the ambiguous feature Forey, P. L. does not share (i.e., again, because it is an 1973. A revision ofthe elopiform fishes, fossil expression of a different ontogeny and is, and recent. Bull. British Mus. (Nat. therefore, nonhomologous with a trait that it Hist.), Geol., Suppl. 10, 222 pp. 56 AMERICAN MUSEUM NOVITATES NO. 2827

Gill, T. N. 1 984a. Development of the head and pectoral 1872. Arrangement of the families of fishes, skeleton of Amia with a note on the or classes Pisces, Marsipobranchii, and scales. Gegen baurs morph. Jahrb., Leptocardii. Smithsonian Misc. Coll., Leipzig, vol. 130, no. 2, pp. 315-351. vol. 11, no. 247, pp. i-1-49. 1984b. Development of the head skeleton and Goodrich, E. S. pectoral girdle of salmons, with a note 1909. Studies on the structure and develop- on the scales. Canadian Jour. Zool., vol. ment of vertebrates. Dover Publ. re- 62, pp. 1757-1778. print 1958, 2 vols. Jordan, D. S. Gosline, W. A. 1923. A classification offishes. Stanford Univ. 1969. The morphology and systematic posi- Publ., Univ. Ser., Biol. Sci., vol. 3, no. tion of the alepocephaloid fishes. Bull. 2, pp. 1-243 + x. British Mus. (Nat. Hist.), vol. 18, no. 6, K6lliker, A. pp.183-218. 1859. On the different types in the microscop- 1971. Functional morphology and classifica- ic structure of the skeleton of osseous tion of teleostean fishes. Univ. Hawaii fishes. Proc. Roy. Soc. London, vol. 9, Press, 208 pp. pp. 1-13. Grande, L. Lauder, G. V. [in press] Recent and fossil clupeomorph fishes 1983. Functional design and evolution of the with materials for a revision of the pharyngeal jaw apparatus in euteleos- subgroup ofclupeoids. Bull. Amer. Mus. tean fishes. Zool. Jour. Linn. Soc., vol. Nat. Hist. 77, no. 1, pp. 1-38. Greenwood, P. H. Lauder, G. V., and K. F. Liem 1968. The osteology and relationships of the 1983. The evolution and interrelationships of Denticipitidae, a family of clupeo- the actinopterygian fishes. Bull. Mus. morph fishes. Bull. Brit. Mus. (Nat. Comp. Zool., vol. 150, no. 3, pp. 95- Hist.), Zool., vol. 16, pp. 13-273. 197. 1970. Skull and swimbladder connections in McAllister, D. E. fishes of the family Megalopidae. Bull. 1968. Evolution of branchiostegals and clas- Brit. Mus. (Nat. Hist.), Zool., vol. 19, sification ofteleostome fishes. Natl. Mus. no. 1, pp. 121-135. Canada, Bull. 221, Biol. Ser., no. 77, Greenwood, P. H., G. S. Myers, D. E. Rosen, and 239 pp. S. H. Weitzman Mead, G. W. 1967. Named main divisions of teleostean 1966. Family Chlorophthalmidae. In Fishes fishes. Proc. Biol. Soc. Washington, vol. of the Western North Atlantic. Mem. 80, pp. 227-228. Sears Found. Mar. Res., no. 1, pt. 5, pp. Greenwood, P. H., and D. E. Rosen 162-189. 1971. Notes on the structure and relationships Muller, J. ofthe alepocephaloid fishes. Amer. Mus. 1844. Uber den Ban und die Grenzen der Novitates 2473, pp. 1-41. Ganoiden und uiber das naturliche sys- Greenwood, P. H., D. E. Rosen, S. H. Weitzman, tem der Fische. Abhandl. Akad. Wiss. and G. S. Myers Berlin, phys.-math., pp. 201-204. 1966. Phyletic studies ofteleostean fishes, with Myers, G. S. a provisional classification of living 1958. Nomenclator of certain terms used for forms. Bull. Amer. Mus. Nat. Hist., vol. higher categories offishes. Stanford Ich- 131, pp.339-456. thyol. Bull., vol. 7, no. 3, pp. 31-40. Gunther, A. C. L. G. Nelson, G. 1859-1870. Catalogue of fishes of the British 1967. Branchial muscles in some generalized Museum, vols. 1-8, Brit. Mus. Nat. Hist. teleostean fishes. Acta Zool., Stock- London. holm, vol. 48, pp. 277-288. Holtzvoogd, C. Nelson, J. 1965. The pharyngeal bones and muscles in 1984. Fishes of the world. New York, John Teleostei, a taxonomic study. Proc. K. Wiley, 2nd ed. ned. Akad. Wet. (Zool.) (Ser. C), vol. Nolf, D. 68, no. 3, pp. 209-218. 1974. De Teleostei-otolieten uit het Eoceen Jollie, M. T. van het Belgisch Bekken-reconstructie 1954. The general anatomy of Lampanyctus van de fauna en biostratigrafische toe- leucopsarus (Eigenmann and Eigen- passing. Doctoral dissertation, Rijks- mann). Univ. Microfilms, 239 pp. universiteit, Gent. 1985 ROSEN: EUTELEOSTEANS 57

Norden, C. R. Rosen, D. E. 1961. Comparative osteology of representa- 1973. Interrelationships of higher euteleos- tive salmonid fishes, with particular ref- tean fishes. In Greenwood, P. H., R. S. erence to the grayling (Thymallus arc- Miles, and C. Patterson (eds.), Interre- ticus) and its phylogeny. Jour. Fish. Res. lationships infishes. London, Academic Bd. Canada, vol. 18, no.5, pp. 679-791. Press, pp. 397-513. Norman, J. R. 1974. Phylogeny and zoogeography of sal- 1934. A draft synopsis of the orders, families moniform fishes and relationships of and genera of Recent fishes and fishlike Lepidogalaxias salamandroides. Bull. vertebrates. 1957 photo offset by British Amer. Mus. Nat. Hist., vol. 153, art. 2, Mus. Nat. Hist. pp. 265-326. Parenti, L. R. 1982. Teleostean interrelationships, morpho- [MS] The phylogenetic significance of bone logical function and evolutionary infer- types in euteleost fishes. Submitted to ence. Amer. Zool., vol. 22, no. 2, pp. Linn. Soc. London. 261-273. Patterson, C. Rosen, D. E., and C. Patterson 1964. A review ofMesozoic acanthopterygian 1969. The structure and relationships of the fishes, with special reference to those of paracanthopterygian fishes. Bull. Amer. the English Chalk. Phil. Trans. Roy. Soc. Mus. Nat. Hist., vol. 141, art. 3, pp. London, ser. B, vol. 247, pp. 213-482. 357-474. 1975. The braincase of pholidophorid and Smith, C. L., and R. M. Bailey leptolepid fishes, with a review of the 1961. Evolution of the dorsal-fin supports of percoid fishes. Papers Mich. Acad. Sci., actinopterygian braincase. Ibid., Biol. Arts and Letters, vol. 46, pp. 345-363. Sci., vol. 269, no. 899, pp. 275-579. Smith-Woodward, A. Patterson, C., and D. E. Rosen 1932. Textbook ofPalaeontology. 2nd English 1977. Review of ichthyodectiform and other ed. revised with addition of C. East- Mesozoic teleost fishes and the theory man's English translation of vol. 2 by and practice of classifying fossils. Bull. Karl A. von Zittel. Amer. Mus. Nat. Hist., vol. 158, art. 2, Stiassny, M. L. J. pp. 81-172. [MS] The limits and relationships ofthe acan- Peiiaz, M. thomorph teleosts. 1983. Ecomorphological laws and saltation in Travers, R. A. the early ontogeny of Salmonoidei. Fo- 1981. The interarcual cartilage: a review ofits lia Zool., vol. 32, no. 4, pp. 365-373. development, distribution and value as Regan, C. T. an indicator ofphyletic relationships in 1909. The classification of teleostean fishes. euteleostean fishes. Jour. Nat. Hist., vol. Ann. Mag. Nat. Hist., ser. 7, vol. 11, 15, pp. 853-871. pp. 459-466. Vladykov, V. D. 1929. Fishes. In Encyclopedia Britannica, 14th 1962. Osteological studies on Pacific salmon ed., London and New York, vol. 9, pp. of the genus Oncorhynchus. Bull. Fish. 305-329. Res. Bd. Canada, no. 136, pp. 1-172. Weitzman, S. H. Ridewood, W. G. 1974. Osteology and evolutionary relation- 1904. On the cranial osteology of the fishes of ships ofthe Sternoptychidae, with a new the families Elopidae and Albulidae, classification of stomatiatoid families. with remarks on the morphology of the Bull. Amer. Mus. Nat. Hist., vol. 153, skull in the lower teleostean fishes gen- art. 3, pp. 327-478. erally. Proc. Zool. Soc. London, pp. 35- Whitehead, P. J. P., and G. G. Teugels 81. [in press] The West African pygmy herring 1905. On the cranial osteology ofthe clupeoid Sierrathrissa leonensis: general fea- fishes. Ibid., pt. 1, pp. 448-493. tures, visceral anatomy and osteology. Rose, J. A. Amer. Mus. Novitates. 1961. Anatomy and sexual dimorphism ofthe Woods, L. P., and R. F. Inger swimbladder and vertebral column in 1957. The cave, spring, and swampfishes of Ophidion holbrooki (Pisces: Ophidi- the family Amblyopsidae ofcentral and idae). Bull. Mar. Sci. Gulf and Carib., eastern United States. Amer. Midl. Nat., vol. 11, no. 2, pp. 280-308. vol. 58, no. 1, pp. 232-256.