phylogenetic relationships of the families curimatidae, … · 2017-08-20 · smithsonian...

Phylogenetic Relationshipsof the Families Curimatidae,

Prochilodontidae, Anostomidae,and Ghilodontidae

(Pisces: Characiformes)

RICHARD P. VARI

SMITHSONIAN CONTRIBUTIONS TO ZOOLOGY • NUMBER 378

SERIES PUBLICATIONS OF THE SMITHSONIAN INSTITUTION

Emphasis upon publication as a means of "diffusing knowledge" was expressedby the first Secretary of the Smithsonian. In his formal plan for the Institution, JosephHenry outlined a program that included the following statement: "It is proposed topublish a series of reports, giving an account of the new discoveries in science, andof the changes made from year to year in all branches of knowledge." This themeof basic research has been adhered to through the years by thousands of titles issuedin series publications under the Smithsonian imprint, commencing with SmithsonianContributions to Knowledge in 1848 and continuing with the following active series:

Smithsonian Contributions to AnthropologySmithsonian Contributions to Astrophysics

Smithsonian Contributions to BotanySmithsonian Contributions to the Earth SciencesSmithsonian Contributions to the Marine Sciences

Smithsonian Contributions to PaleobiologySmithsonian Contributions to Zoo/ogySmithsonian Studies in Air and Space

Smithsonian Studies in History and Technology

In these series, the Institution publishes small papers and full-scale monographsthat report the research and collections of its various museums and bureaux or ofprofessional colleagues in the world of science and scholarship. The publications aredistributed by mailing lists to libraries, universities, and similar institutions throughoutthe world.

Papers or monographs submitted for series publication are received by theSmithsonian Institution Press, subject to its own review for format and style, onlythrough departments of the various Smithsonian museums or bureaux, where themanuscripts are given substantive review. Press requirements for manuscript and artpreparation are outlined on the inside back cover.

S. Dillon RipleySecretarySmithsonian Institution

S M I T H S O N I A N C O N T R I B U T I O N S T O Z O O L O G Y • N U M B E R 3 7 8

Phylogenetic Relationships of theFamilies Curimatidae, Prochilodontidae,

Anostomidae, and Chilodontidae(Pisces: Characiformes)

Richard P. Vari

SMITHSONIAN INSTITUTION PRESS

City of Washington

1983

A B S T R A C T

Vari, Richard P. Phylogenetic Relationships of the Families Curimatidae,Prochilodontidae, Anostomidae, and Chilodontidae (Pisces: Characi-formes). Smithsonian Contributions to Zoology, number 378, 60 pages, 41 figures,1 table, 1983.—A series of osteological and soft anatomical systems in thefamilies Curimatidae, Prochilodontidae, Anostomidae, and Chilodontidaewere examined to reconstruct the phylogenetic relationships of these taxa anddetermine whether they are a monophyletic assemblage. Numerous adapta-tions associated with the food gathering and manipulation systems along withthose in other portions of the body provided data useful in the reconstructionof the phylogenetic relationships of these taxa. The evidence of this study iscongruent with the hypothesis that the four families constitute a monophyleticassemblage definable by shared derived characters. Results of the analysisindicate that the Curimatidae and Prochilodontidae are most parsimoniouslyinterpreted to be sister groups. The assemblage formed by those two familiesis considered to be the sister group of the monophyletic lineage that consistsof the Anostomidae and Chilodontidae. Synapomorphies define the membersof each family as monophyletic units.

The well-developed epibranchial organs in the Chilodontidae, Citharinidae,and the Curimatidae plus Prochilodontidae, were found to represent threedistinct, nonhomologous kinds of pharyngeal outpocketings. The most parsi-monious interpretation of those data and other available information onostariophysan phylogeny is that such diverticuli have been acquired severaltimes independently among characiforms and their near relatives. This con-clusion agrees with the hypothesis proposed by Nelson (1967) but indicatesthat Bertmar, Kapoor, and Miller's concept (1969) of epibranchial organs asthe ancestral condition for teleosts is not parsimonious, at least when appliedto ostariophysan fishes.

OFFICIAL PUBLICATION DATE is handstamped in a limited number of initial copies and is recordedin the Institution's annual report, Smithsonian Year. SERIES COVER DESIGN: The coral Montastreacavemosa (Linnaeus).

Library of Congress Cataloging in Publication DataVari, Richard P.Phylogenetic relationships of the families Curimatidae, Prochilodontidae, Anostomidae, and

Chilodontidae (Pisces, Characiformes)(Smithsonian contributions to zoology ; no. 378)Bibliography: p.Supt. of Docs, no.: SI 1.27:3781. Curimatidae—Classification. 2. Prochilodontidae—Classification. 3. Anostomidae—Clas-

sification. 4. Chilodontidae—Classification. 5. Fishes—Classification. 6. Fishes—LatinAmerica—Classification. I. Title. II. Series.

QL1.S54 no. 378 [QL638.C89] 591s [597'.53] 82-600338

Contents

Page

Introduction 1Acknowledgments 3

Systematic Procedures 3Methods and Materials 4

Terminology 6Abbreviations 6

Character Description and Analysis 7Teeth and Jaws 7Gill Arches 11Epibranchial Organs 21Hyoid Apparatus 24Suspensorium and Circumorbital Series 25Pectoral Girdle 33Neurocranium 36Vertebral Column and Ribs 41Myology 42

Phylogenetic Reconstruction 46The Four-Family Assemblage 46Curimatidae and Prochilodontidae Clade 47Family Curimatidae 48Family Prochilodontidae 49Anostomidae and Chilodontidae Clade 50Family Anostomidae 50Family Chilodontidae 51

Discussion 52Comparisons with Previous Classifications 54Resumen 56Literature Cited 58

in

Phylogenetic Relationships of theFamilies Curimatidae, Prochilodontidae,

Anostomidae, and Chilodontidae(Pisces: Characiformes)

Richard P. Vari

Introduction

The Curimatidae, Prochilodontidae, Anostom-idae, and Chilodontidae form a speciose, morpho-logically diverse assemblage of characiform fisheswidely distributed in Neotropical freshwaters.Members of these families occur in the Atlanticdrainages of South America from northern Co-lombia to the Rio Negro of Argentina and inhabitthe rivers and streams of the Pacific drainages ofthe region from Puntarenas Province of CostaRica south to central Peru. Curimatids, prochil-odontids, anostomids, and chilodontids inhabit abroad variety of freshwater ecosystems rangingfrom stagnant ox-bow lakes through sluggish riv-ers to rapidly flowing streams. In certain ecologi-cal settings they represent a significant proportionof the fish biomass (Lowe-McConnell, 1975:109),and many members of these taxa are exploitedby commercial and subsistence fisheries through-out the Neotropics (Lowe-McConnell, 1975:74;Smith, 1981:140; Goulding, 1981:60).

The results reported in this paper are an out-growth of phylogenetic and revisionary studiesoriginally centered on the Curimatidae. An effec-

Richard P. Vari, Department of Vertebrate Zoology, National Museumof Natural History, Smithsonian Institution, Washington, D.C.20560.

tive analysis of phylogenetic relationships withinthat family necessitated the recognition of anappropriate outgroup, thereby allowing the po-larization of intrafamilial character variation. Aprerequisite for the determination of the sistergroup to curimatids was an explicit, corraboratedhypothesis of the phylogenetic placement of cur-imatids within a taxonomically more encompass-ing assemblage of characiforms.

Widely divergent opinions have been advancedon the exact relationships of the Curimatidae toother characiforms (Table 1). Giinther (1864:288)united the then-known members of the Curima-tidae (Curimatus), Prochilodontidae {Prochilodus),Chilodontidae (Caenotropus), Hemiodontidae(Hemiodus), and Parodontidae (Saccodon, Parodon)(families sensu Greenwood et al., 1966) in hissuprageneric group Curimatina, and placed thedescribed genera of the present day Anostomidae(Anostomus, Rhytiodus, Leporinus) in the Anostoma-tina group. Boulenger (1904:576), in a significantshift, incorporated the Neotropical curimatids(Curimatus) and prochilodontids (Prochilodus) intohis subfamily Citharininae, together with the OldWorld characiform genera Citharinus and Cithari-dium. His Anostominae included the present dayanostomids (Anostomus, Leporinus, Nanognathus[? = Schizodon]) and genera today considered

SMITHSONIAN CONTRIBUTIONS TO ZOOLOGY

TABLE 1.—Classifications of the four-family assemblagediscussed in the text

Author Classification

Gunther, 1864

Boulenger, 1904

Regan, 1911

Gregory and Conrad,1938

Greenwood et al.,1966

Gery, 1977b

Curimatina groupCurimatus, Prochilodus,Caenotropus, Hemiodus, Saccodon,Parodon

Anostomatina groupAnostomus, Rhytiodus, Leporinus

Characinidae [in part]Citharininae

Citharinus, Citharidium,Curimatus, Prochilodus

AnostominaeAnostomus, Leporinus,Characidium, Chorimycterus,Nanostomus, Nanognathus

HemiodontinaeCaenotropus, Hemiodus, Saccodon,Parodon

AnostomidaeCurimatinae

Curimatus, AnodusProchilodontinae

ProchilodusAnostominae

Anostomus, Rhytiodus, Leporinus,Leporellus, Caenotropus

Anostominae (=Anostomidae ofRegan, 1911)

Curimatidae, Prochilodontidae,Anostomidae, Chilodontidae

AnostomidaeAnostominae

Schizodon, Leporinus, Abramites,Rhytiodus, Anostomoides, Anosto-mus, Synaptolaemus, Sartor,Gnathodolus

LeporellinaeLeporellus

CurimatidaeChilodinae

Caenotropus, ChilodusProchilodinae

Semaprochilodus, Prochilodus,Ichthyoelephas

CurimatinaeCurimata, Curimatella, Curima-topsis

AnodinaeAnodus

either characids {Characidium, Chorimycterus [ =Characidium]) or lebiasinids (Nanostomus [ = Nan-nostomus]). The chilodontid taxa described at thattime (Caenotropus with Chilodus as a synonym) wereunited by Boulenger in the Hemiodontidae to-gether with various hemiodontids (Hemiodus) andparodontids (Saccodon, Parodon). Regan (1911:20)rejected Boulenger's transatlantic grouping ofNeotropical curimatids and prochilodontids withsome Old World characiforms. Furthermore, hedisagreed with Giinther's recognition of a sepa-rate taxon solely for anostomoids and that au-thor's alignment of the chilodontids with hemio-dontids. Rather, he incorporated the present dayCurimatidae (Curimatus), Prochilodontidae (Pro-chilodus), Anostomidae (Anostomus, Rhytiodus, Le-porinus, Leporellus) and Chilodontidae (Caenotropus)in his family Anostomidae with the last twogroups combined in his subfamily Anostominae.Gregory and Conrad (1938:347) followed Reganin considering these four families to be a singletaxon, but at the subfamily level. Eigenmann(1917:38-39), in contrast, placed the Anostomi-nae, Chilodinae, Prochilodinae, Hemiodontinae,Elopomorphinae, and Curimatinae (= the Anos-tomidae, Chilodontidae, Prochilodontidae, Hem-iodontidae, Anodinae, and Curimatidae ofGreenwood et al., 1966) as "an offshoot of theCheirodontinae" without any explicit commenton the relationships of the individual components.

Gery (1961:108) presented a dendrogram of hishypothesized phytogeny of anostomids and theirpresumed relatives. That outgroup included cur-imatids, prochilodontids, hemiodontids and lebi-asinids but not chilodontids. However, exact con-cepts of phylogenetic relationship cannot be re-trieved from Gery's figure (1961, fig. 24). Morerecently Gery (1977b:210) united curimatids, pro-chilodontids, chilodontids, and anodontines in hisfamily Curimatidae but retained a separate An-ostomidae for anostomids (see Table 1 for in-cluded genera). No explicit statement on thedegree to which that classification reflects under-lying phylogenetic concepts was provided, norwere the characters that support this grouping oftaxa discussed.

NUMBER 378

The classifications summarized above gener-ally were advanced without a discussion of thebasis for the different groupings of taxa or withonly a cursory presentation of the anatomicalinformation underlying the taxonomic decisionsof the different authors. The absence of such datamade it impossible to determine the extent towhich those classifications reflected phylogeneticrather than phenetic concepts. Thus, an evalua-tion of the relative merits of the classifications asphylogenetic indicators was difficult, if not im-possible.

Roberts (1973:221) was the only author toexplicitly state his concepts on the relationshipsof a subgroup of these four families and to detailthe characters associated with his hypotheses.Contrary to his predecessors, Roberts rejected thehypothesis of a close phylogenetic relationship ofthe Curimatidae and Prochilodontidae. Ratherhe stated that "the evidence favoring relationshipbetween Prochilodontidae and Anostomidae isrelatively strong. . . ." He did not, however, com-ment on what relationship he envisioned for cur-imatids relative to the lineage purportedly formedby those two families or propose an alternativesister group for the Curimatidae.

Curimatids have, therefore, been inconsistentlyassociated with a broad range of Neotropical andOld World characiform groups. Similarly, hy-potheses on the relationships of prochilodontids,anostomids, and chilodontids to each other andcharaciform outgroups have differed in explicit-ness and inclusiveness. Furthermore, during thecentury-old discussion on the relationships ofthese families, no phylogenetic hypothesis hasbeen put forward that uses criteria consideredacceptable in this study: the possession of sharedderived characters. The conjunction of these fac-tors made it impossible to evaluate which, if any,of the previous classifications was an accuraterepresentation of the phylogenetic history of cur-imatids and their close relatives. It was in anattempt to resolve this question that this studywas undertaken.

ACKNOWLEDGMENTS.—I am greatly indebted tothe following individuals and institutions for the

loan and exchange of specimens, information andother assistance: D.E. Rosen and C.L. Smith,American Museum of Natural History; P.H.Greenwood and G.J. Howes, British Museum(Natural History); N. Menezes, Museu do Zool-ogia, Universidad de Sao Paulo; and F. MagoLeccia and A. Machado Allison, Museo de Biol-ogia, Universidad Central de Venezuela. I wouldlike to thank G. J. Howes, A. Machado Allisonand particularly S.H. Weitzman for spendingmany hours discussing questions of characiformanatomy and phylogeny. This study was partiallysupported by the Smithsonian Institution Neo-tropical Lowland Research Program. S.L. Jewett,J.R. Gomon, and K. Bruwelheide provided tech-nical assistance, which facilitated various aspectsof the study. The comments and criticisms of S.H.Weitzman, J.C. Tyler, and R. Winterbottom con-tributed greatly to the improvement of this paper.Antonio Machado Allison generously providedthe Spanish translation of the resumen.

Systematic Procedures

The relationships of the Curimatidae, Prochil-odontidae, Anostomidae, and Chilodontidae areanalyzed using the methodology of phylogeneticreconstruction first formalized by Hennig (1950,1966). The goal of phylogenetic systematics (al-ternatively termed cladistics or cladism) is thegrouping of taxa in a series of nested units thatreflect the best estimate of the natural hierarchi-cal history of groups of organisms.

Taxa are grouped on the basis of the possessionof shared derived characters (synapomorphies),which are considered the ooly valid basis forhypotheses of common ancestry. Hypotheses ofrelationship derived from the possession of sharedprimitive characters (symplesiomorphies) andphylogenetic speculations based on concepts ofoverall phenetic similarity or degrees of differenceare either not useful tests of phylogenetic hy-potheses or incongruent with the aim of thisstudy, that is, the advancement of hypotheses ofthe phylogenetic histories of the taxa under ex-amination.


In keeping with the general scientific principleof parsimony, the hypothesis of the phylogenetichistory of a group that necessitates the fewest adhoc assumptions about character transformationsis preferred. Monophyletic groups consist of alldescendants of a common ancestor, exclusive ofany species not descended from that commonancestor. The traditional alternative definitionsof monophyly are of such generality as to includeall possible combinations of taxa or are dependenton arbitrary taxonomic ranks, or both, and are,therefore, often meaningless in application andinformation content.

The polarity of character transformation series(plesiomorphy vs apomorphy: primitive vs de-rived) is determined by outgroup comparison ordata from ontogenetic transformations. Outgroupcomparisons to characiforms and if necessaryother otophysan or ostariophysan groups wereused to determine the polarity of characters thatvary within the assemblage under examination.The ontogenetic and phylogenetic polarities areconsidered equivalent for any character that un-dergoes developmental changes in one of twosister groups. The ontogenetically later stage inthe transforming lineage is, therefore, consideredapomorphous relative to its homologue in thenontransforming group. This procedure is moreparsimonious than assuming that the transitionwas primitively present in the common ancestorof both groups and was secondarily lost in onelineage (detailed discussions of the theory behind,and methods of application of, the above meth-odologies can be found in Nelson and Platnick,1981, and Wiley, 1981).

Methods and Materials

Osteological and cartilaginous skeletal systemswere examined in cleared and counterstainedspecimens prepared in a modified version of thealizarin Red S-alcian blue method of Dingerkusand Uhler (1977). Previously cleared specimensstained solely in alizarin Red S, along with dryskeletal materials, were supplemental sources ofosteological data. Drawings were reversed into

conventional orientation if drawn from the rightside of the specimen. Anatomical illustrationswere prepared using a Zeiss microscopic cameralucida. The osteological material and myologicalpreparations examined are deposited in theUSNM collections of the National Museum ofNatural History, Smithsonian Institution; theAmerican Museum of Natural History (AMNH);and Museo de Biologia, Universidad Central deVenezuela (MBUCV). It was impossible to ex-amine each character in all members of the as-semblage under discussion given the extremelylarge number of species involved, the rarity ofsome nominal forms, and the myriad taxonomicproblems in this assemblage. A wide spectrum ofgenera and species were examined, but it is pos-sible that in some cases the hypothesized distri-bution of a character may differ from that notedherein.

The following specimens are the basis for illus-trations or observations noted in the text. (Meas-urement in mm is standard length (SL)).

CURIMATIDAE

Curimata cyprinoides (Linneaus), USNM 231433; 2 specimens,17.5-21.7 mm; Surinam, Corantijn River.

Curimata vitatta Kner, USNM 231434; 1 specimen, 81.7 mm;Brazil, Rio Negro (Figures 5, 6, 15-17).

Curimatopsis evelynae Gery, USNM 214794; 4 specimens, 25.7-34.9 mm; Brazil, Rio Negro.

Curimatopsis macrolepis (Steindachner), USNM 231436; 1specimen, 51.3 mm; Peru.

Curimatopsis macrolepis (Steindachner), USNM 190285; 1specimen, 34.1 mm; Peru, Iquitos.

Potamorhina laticeps (Valenciennes), USNM 121325; 1 speci-men, 129.3 mm; Venezuela, Lake Maracaibo Basin (Fig-ures 1, 2, 25).

Potamorhina latior (Spix), AMNH 48677; 1 specimen, 73.1mm; Bolivia, Rio Mamore (Figure 27).

Psectrogaster amazonica Eigenmann and Eigenmann, AMNH40088SD; 1 specimen, 139.2 mm; Bolivia, Rio Guapore(Figures 7, 8, 33, 35).

PROCHILODONTIDAE

hhthyoelephas species, USNM 231437; 1 specimen, 110.2 mm;Ecuador (Figures 15, 28).

Prochilodus mgricans Agassiz, USNM 231438; 1 specimen,144.3 mm; Bolivia, Tumpasa.

Prochilodus rubrotaeniatus Schomburgk, USNM 225419; 1 spec-

NUMBER 378

imen, 108.7 mm; Surinam, Corantijn River (Figures 1, 2,22, 39, 40).

Prochilodus species, USNM 231538; 3 specimens, 29.3-35.2mm; Amazon River.

Semaprochilodus taemurus (Steindachner), USNM 231536; 1specimen, 74.7 mm; Brazil, Rio Negro, vicinity of Manaus(Figure 9).

Semaprochilodus taemurus (Steindachner), USNM 231537; 1specimen, 34.7 mm; Brazil, Amazon River.

ANOSTOMIDAE

Anostomus plicatus Eigenmann, USNM 225396; 2 specimens,75.8 95.3 mm; Surinam, Nickerie District, MatappiCreek.

Anostomus species, USNM 231540; 1 specimen, 81.3 mm; nodata (Figures 20, 26).

Gnathodolus bidens Myers, USNM 231539; 1 specimen, 76.2mm; aquarium material.

Laemolyta species, USNM 179514; 2 specimens, 75.3-77.0mm; Brazil, Rio Urubu (Figure 11).

Upormus fasctatus (Bloch), USNM 103847; 1 specimen, 73.2mm; aquarium material (Figure 29).

Leporinus megalepis Giinther, USNM 231541; 1 specimen, 68.2mm; no data (Figure 23).

Leporinus reinhardti Lutken, AMNH 40104SD; 1 specimen,170.0 mm; Bolivia, Rio Guapore (Figures 33, 34).

Leporinus striatus Kner, USNM 231948; 26 specimens 68.5-178.5 mm; Colombia, Rio Salado (Figure 38).

Rhytiodus microlepis Kner, USNM 163850; 1 specimen, 131.7mm; Peru, Iquitos.

Schizodon fasciatum Agassiz, USNM 179507; 1 specimen, 49.3mm; Brazil, Rio Urubu (figures 1, 2, 15).

Synaptolaemus cingulatus Myers and Fernandez-Yepez,MBUCV V-4252; 1 specimen, 71.2 mm; Venezuela, Ri'oParagua.

CHILODONTIDAE

Caenotropus labyrinthicus (Kner), USNM 231543; 1 specimen,58.5 mm; Brazil, Rio Negro.

Caenotropus labyrinthicus (Kner), USNM 231544; 1 specimen,64.2 mm; Venezuela, upper Rio Negro (Figure 11).

Caenotropus maculosus (Eigenmann), USNM 231545; 2 speci-mens, 42.7-46.3 mm; Guyana (Figures 1, 2, 10, 12, 18, 19,21,24, 30-32).

Caenotropus maculosus (Eigenmann), USNM 231546; 1 speci-men, 97.5 mm; Surinam, Saramaca River.

Chtlodus punctatus Miiller and Troschel, USNM 231542; 13specimens, 27.1-38.2 mm; Peru, Rio Nanay (Figures 11,36, 37).

CHARACIDAE

Acestrorhynchus falcatus (Bloch), USNM 225614; 2 specimens,81.0-85.1 mm; Surinam, Nickerie District, Sisa Creek.

Brycon falcatus Miiller and Troschel, USNM 226161; 2 spec-

imens, 71.3-78.3 mm; Surinam, Corantijn River (Figures3,4, 13-15).

Camegiella strigata (Giinther), USNM 225245; 5 specimens,23.4-25.6 mm; Surinam, Corantijn River.

Chalceus macrolepidotus Cuvier, USNM 231547; 1 specimen,39.9 mm; aquarium material.

Crenuchus spilurus Giinther, USNM 225630; 4 specimens,21.7-28.2 mm; Surinam, Nickerie District, Lana Creek.

Gasteropelecus stemicla (Linneaus), USNM 226337; 5 speci-mens, 29.7-32.5 mm; Surinam, Corantijn River.

Hydrolycus pectoralis (Giinther), USNM 231548; 1 specimen,167.1 mm; Peru, Rio Ucayali.

Rhaphiodon vulpinnis Agassiz, USNM 231549; 4 specimens,41.7-44.3 mm; Brazil, Rio Solimoes.

Salminus maxillosus Valenciennes, USNM 194213; 1 specimen,214.7 mm; Venezuela, Rio Nitiado-Seco.

HEMIODONTIDAE

Anodus elongatus Spix, USNM 231550; 1 specimen, 120.3 mm;Peru, Rio Ucayali.

Bivibranchia protractila Eigenmann, USNM 194363; 1 speci-men, 62.8 mm; Brazil, Mato Grosso, upper Rio Juruena.

Hemiodopsis ocellata Vari, USNM 225593; 1 specimen, 99.6mm; Surinam, Corantijn River.

Hemiodus species, USNM 231551; 2 specimens, 55.7-57.1mm; Brazil, Mato Grosso, Rio Arinos.

Micromischodus sugillatus Roberts, USNM 205527; 1 specimen,96.1 mm; Brazil, Para.

PARODONTIDAE

Parodon suborbitalis Valenciennes, USNM 231552; 2 speci-mens, 55.0-58.1 mm; Colombia, Rio Salado.

Saccodon dariensis (Meek and Hildebrand), USNM 208505; 1specimen, 73.3 mm; Panama, Rio Membrillo.

HEPSETIDAE

Hepsetus odoe (Bloch), USNM 231553; 1 specimen, 96.2 mm;Togo, Kama.

ClTHARINIDAE

Citharinus citharus Geoffrey, USNM 52146; 1 specimen, 218.7mm; Egypt, Nile River.

Citharinus species, USNM 231554; 2 specimens, 56.3-64.5mm; Volta, Black Volta River.

DlSTICHODONTIDAE

Nannocharax inlermedius Boulenger, USNM 231555; 2 speci-mens, 50.7-63.4 mm; West Africa.

Neolebias olbrechtsi Poll, USNM 227394; 4 specimens, 25.3-29.7 mm; Zaire.

Paradistichodus dimidiatus Pellegrin, USNM 231556; 2 speci-mens, 45.6-47.3 mm; Ghana, Dayi River.


In addition to the listed specimens, a largenumber of dry skeletons, stained and clearedglycerine preparations and alcohol preservedspecimens in the collections of AMNH andUSNM were examined in the comparative studiesassociated with the present analysis.

TERMINOLOGY.—Osteological nomenclaturefollows Weitzman (1962) with the following mod-ifications. As suggested by Roberts (1969:402),"vomer" is substituted for "prevomer," and "in-tercalar" for "opisthotic." The element tradition-ally termed "epihyal" is instead referred to as"posterior ceratohyal." The "ceratohyal" of manyprevious authors is more correctly termed "ante-rior ceratohyal." This shift in terminology is aconsequence of the lack of serial homology be-tween the so-called epihyal and the epibranchialsand the resultant misleading inference of homol-ogy inherent in the continued use of "epihyal"(Nelson, 1969:480-481). The use of "epioccipital"rather than "epiotic" follows Patterson (1975).Nelson's substitution (1973) of "angulo-articular"for "articular" and "retroarticular" for "angular"is more reflective of the homologies of these ele-ments among teleosts than previous terminologiesand is utilized in this paper. Myological termi-nology is that of Winterbottom (1974).

Unless noted otherwise, the concepts of thecharaciform families used in this paper are thoseof Greenwood et al. (1966) with the followingmodifications. The Cynodontidae of those au-thors is considered a tribe in the Characidaerather than a distinct family in keeping with theresults of Howes (1976). The Ichthyboridae ofGreenwood et al. is placed within the Disticho-dontidae following Vari (1979:339). The termi-nology utilized for the major groups of osatrio-physan fishes is that of Fink and Fink (1981).

ABBREVIATIONS.—The following abbreviationsare used in the text and illustrations.

A

AAPAAACBBTP

division of the adductor mandibulae(1 to 3 and w)

adductor arcus palatiniangulo-articularanterior ceratohyalbasihyalbasihyal tooth plate

BBBOCBRCCENCLCORCRCSICTSDENDHDOEEAFECTEOMEPIEREXTFRFMFSHHAHIHYOIINTLAPLELIGLOCMESMETMETHMXNCNSOPOPFPAPALPAPPARPBPCPCLPMXPOSTPPPRPREPTEPZ

basibranchial (1 to 4)basioccipitalbranchiostegal rayceratobranchial (1 to 5)centrumcleithrumcoracoidceratobranchial ridgecavum sinus imparisconnective tissue sheetdentarydorsal hypohyaldilatator operculiepibranchial (1 to 5)efferent artery foramenectopterygoidepibranchial organ muscleepioccipitalepibranchial ridgeextrascapularfrontalforamen magnumforamen for scaphiumhypobranchial (1 to 3)hyohyoidei abductoreshyohydoides inferiorhyomandibulainterhyalinteroperclelevator arcus palatinilateral ethmoidligamentlateral occipital foramenmesopterygoidmetapterygoidmesethmoidmaxillaneural canalneural spineopercleopercular flangeparietalpalatinepterotic articular processparasphenoidinfrapharyngobranchial (1 to 4)posterior ceratohyalpostcleithrum (1 to 3)premaxillaposttemporalparapophysispleural ribpreoperclepteroticprezygapophysis

NUMBER

QUARASSAPSCSCLSMXSPHSPOSOPSTSYMUPVVH

378

quadrateretroarticularscaphiumsupracleithral articular processscapulasupracleithrumsupramaxillasphenoticsuprapreoperclesuboperclesternohyoideussymplecticupper pharyngeal tooth plate (4 and 5)vomerventral hypophyal

Character Description and Analysis

The phylogenetic analysis of the present studyhas three primary goals: first, to advance anexplicit hypothesis of the phylogenetic relation-ship of the Curimatidae within the Characi-formes; second, to examine the hypothesizedmonophyly of the Curmatidae and closely relatedgroups; and third, to evaluate the utility of pre-vious alternate classifications as indicators of thephylogenetic history of the Curimatidae and as-sociated lineages.

In this section the characters that show phylo-genetically significant variation in the anatomicalsystems examined in the present study are dis-cussed, along with data that bear on polaritydeterminations for each character transition se-ries. The evidence from the analyzed charactersis brought together in the "Phylogenetic Recon-struction" to advance the most parsimonious hy-pothesis of the phylogenetic history of the taxaunder consideration. The resultant phylogenyserves as the basis for the evaluation of the utilityof previously advanced classifications as indica-tors of phylogenetic relationships among curi-matids and closely related taxa (see "Compari-sons with Previous Classifications").

The purpose of this study is the advancementof a phylogenetic hypothesis rather than a de-tailed description of the osteology and soft anat-omy of the involved taxa. Thus, only anatomicalsystems that show variation relevant to the phy-

logenetic reconstruction are discussed. Similarly,only intrafamilial character differences pertinentto the phylogenetic reconstruction at the familiallevel are detailed. Osteological and soft anatom-ical systems are discussed separately other thanwhere the functional association between the twosystems would make a divided discussion lessefficacious.

Among the hypothesized derived charactersdiscovered during the study, a minority werefound to have distributions incongruent with themost parsimonious phylogenetic reconstructionbased on the overall distribution of synapomor-phies. Such homoplasies are recognizable only inthe context of the final phylogenetic hypothesisand will be summarized as a group in the "Dis-cussion" section. In order to simplify the evalua-tion of homoplasies, such characters and theirdistributional incongruities relative to the arrivedat phylogeny will also be discussed at appropriatepoints in the character descriptions.

TEETH AND JAWS

The feeding habits of curimatids, prochilodon-tids, anostomids, and chilodontids range frommicrophagous filtration of food items from thesubstrate and water column to micropredationand macrovegetation cropping. The diversity ofjaw forms in this assemblage reflects the differingrequirements for dealing with that spectrum offood items.

Characiforms most commonly have a singletooth row on the maxilla and one or two rows ofmoderate-sized teeth on the dentary and premax-illa. The maxilla of chilodontids, curimatids, andanostomids differs from the above pattern inbeing edentulous throughout ontogeny. Thequestion of the possible presence of maxillaryteeth in prochilodontids is difficult to resolve. Theprochilodontid suctorial mouth bears multiplerows of numerous small teeth on hypertrophiedlips (Roberts, 1973:217). Portions of the dentitionoverlap the maxilla but are associated with thefleshy lips rather than being attached to the boneitself. Neither phylogenetic nor ontogenetic data

8 SMITHSONIAN CONTRIBUTIONS TO ZOOLOGY

permit a determination of whether the oral diskdentition proximal to the maxilla is homologouswith the primitively present maxillary dentition.An assumption of the nonhomology of the mul-tiple tooth rows near the maxilla with the primi-tive, attached maxillary teeth would make anedentulous maxilla synapomorphous for prochil-odontids, anostomids, chilodontids, and curima-tids although not unique to these groups amongcharaciforms. Alternatively, it is possible that theoral disk dentition proximal to the prochilodontidmaxilla is derived from the plesiomorphous max-illary teeth. If the latter situation approximatesthe actual phylogenetic history of portions of theprochilodontid oral disk dentition, then the ab-sence of maxillary dentition is not a shared de-rived character for the four taxon unit but wouldrather be a synapomorphy for the Anostomidae,Curimatidae, and Chilodontidae. That singlecharacter phylogenetic scheme is incongruentwith the pattern of relationships considered mostparsimonious, based on the overall distribution ofall synapomorphies. Within the context of thephylogenetic hypothesis incorporating all exam-ined characters (p. 46), two equally parsimoniousexplanations could account for the phylogeneticdistribution of edentulous maxillae among thesetaxa. First, the maxillary teeth were indepen-dently lost in the ancestral curimatid and thecommon ancestor of chilodontids and anostom-ids. Second, maxillary dentition was lost, or al-ready absent, in the common ancestor of the fourfamilies but re-acquired in prochilodontids. Theconjunction of the uncertain homology of por-tions of the prochilodontid dentition with ques-tions about character polarity in the systemmakes it impossible to appropriately use the ab-sence of attached maxillary dentition in the res-olution of phylogenetic relationships within theunit that consists of curimatids, prochilodontids,chilodontids, and anostomids.

Departures from the generalized characiformpattern of premaxillary and dentary dentitiondistinguish each of the families under discussion.Juvenile curimatids (Curimatopsis macrolepis, Curi-mata cyprinoides, and various undetermined Curi-mata species) have a single row of attached uni-

cuspidate premaxillary and dentary teeth (seealso Gery, I977b:231). Excepting the absence ofmaxillary teeth, this dentition pattern agrees inform and distribution with that of various gen-eralized characiforms. Ontogenetically, however,the curimatid premaxilla and dentary undergo aprogressive loss of dentition, with adult curima-tids being endentulous (Figure ID). This ontoge-netic information and the broad distribution ofjaw teeth in characiforms and more inclusiveostariophysan groups indicate that this absenceof jaw dentition is derived within characiforms.

Totally edentulous jaws are limited in chara-ciforms to the Curimatidae and the genus Anodus,considered by some authors (Eigenmann and Ei-genmann, 1889; Fernandez-Yepez, 1948) as amember of that family. Roberts (1974) demon-strated that Anodus actually shares numerous de-rived characters with, and is most closely relatedto, the Hemiodontidae. The other members ofthe Hemiodontidae possess well-developed teeth;therefore, the absence of jaw dentition in Anodus

FIGURE 1.—Upper jaws: A, Caenotropus maculosus, USNM231545 (row of reduced teeth not shown); B, Schizodon fascia-turn, U S N M 179507; c, Prochilodus rubrotaeniatus, U S N M

225419 (multiple rows of reduced teeth not shown); D,Potamorhina laticeps, USNM 121325, left side, lateral view.

NUMBER 378

is most parsimoniously considered a loss achievedindependent of that in curimatids.

The teeth of chilodontids and prochilodontidsare significantly smaller than those in most char-aciforms and attach to the fleshy covering of thejaws rather than onto the premaxilla and dentary.The significance of these similarities in these twotaxa is difficult to evaluate. Differences in themorphology and distribution of jaw dentition inthe two families raise questions as to the homol-ogy of the tooth size reduction and the absence ofdirect tooth to jaw articulation in the lineages.The absence of jaw dentition in adult curimatidscould also be considered the terminal state of atransition series whose intermediate stage mighthave included the precursor of the detached jawteeth in prochilodontids and chilodontids. Avail-able ontogenetic data does not assist in the eval-uation of the preferability of either of these twoscenarios. These uncertanties render it impossibleto appropriately use the dentition morphology inthe Chilodontidae, Prochilodontidae, and Curi-matidae for the resolution of phylogenetic rela-tionships in the four-family unit. Neither thehypothesis of the homology of prochilodontid-chilodontid dentitional form nor the hypothesisthat considers that condition to be a precursor tothe edentulous jaws of curimatids is congruentwith the arrived at phylogeny (see "PhylogeneticReconstruction").

Chilodontids retain the primitive single toothrow on each jaw, whereas prochilodontids have aunique, pronounced increase in the number offunctional teeth and replacement tooth rows.Roberts (1973:217) reported about 600 functionalteeth in an Ichthyoelephas specimen of 97.2 mm SLand 12 replacement tooth rows in larger prochil-odontid specimens. That dentition pattern con-trasts with a maximum of less than 60 teeth and2 or 3 replacement tooth rows elsewhere in theorder. The pronounced increases in the numberof teeth and replacement tooth series are, thus,considered shared derived characters for themembers of the Prochilodontidae.

Anostomids retain the single row of premaxil-lary and dentary teeth common to many chara-ciforms. The teeth of anostomids are, nonetheless,

distinctive in their thickness, multiserrate non-symmetrical cutting edges, increased relativesizes, and the congruent reduction in tooth num-ber (Figure 1B). These shifts from the moderate-sized, more nearly symmetrical, and more nu-merous teeth that evidently are primitive for theCharaciformes are considered synapomorphousfor the subunits of the Anostomidae and are verylikely correlated with their leaf cropping andmicropredatory food habits.

The various distinctive dentitional synapomor-phies of the Curimatidae, Prochilodontidae, Chil-odontidae, and Anostomidae are reflected in theirjaw forms. The overall morphology of the curi-matid upper jaw is little altered relative to thatof other characiforms (Figure ID). The derivedreduction in the overall strength of the involvedbones, the lessening of the medial interconnec-tions between the premaxillae, and the looseningof the articulation between the maxilla and pre-maxilla are all congruent with the absence of jawdentition in adult curimatids. These changes arepresumably correlated with the lack of firm ma-nipulation of food items by the jaws in this mi-crophagous family.

Unique types of highly derived upper jawscharacterize each of the other three families. Thechilodontid maxilla (Figure 1A) is greatly en-larged with the well-developed dorsal maxillaryprocess ligamentously attached medially to themesethmoid's anterolateral surface. The chilo-dontid premaxilla is small relative to the mor-phology typical for characiforms and especiallyso in comparison to the enlarged maxilla in thefamily. The reduced chilodontid premaxillaforms a limited portion of the upper jaw marginwith the paired bones not in contact medially.Rather, the distal portion of the anteroventrallyenlarged mesethmoid lies between the premaxil-lae. In addition to separating those primitivelyconjoined elements, that portion of the meseth-moid also forms the midsection of the arc of theupper jaw (Figure 32). The dentition associatedwith the chilodontid upper jaw is reduced in sizeand not in direct contact with the premaxilla andmaxilla, but rather attached to the fleshy lips.The latter modification permits the maintenance


of a continuous row of teeth along the upper jawdespite the intercalation of the mesethmoid be-tween the premaxillae. Chilodontids also have asupernumerary upper jaw ossification: the supra-maxilla, situated along the posterodorsal marginof the maxilla (Figure 1A). The association of therobust, enlarged maxilla and a reduced premax-illa is not approximated among characiformsother than chilodontids; neither did outgroupcomparisons in that order uncover a mesethmoidthat so distinctly separates the premaxillae. Adiscrete supramaxilla is known elsewhere in theorder only in the characid Chalceus macrolepidotus.Present concepts on phylogenetic relationshipswithin characiforms indicates that the relation-ships of Chalceus lie with the Neotropical familyCharacidae, an assemblage that is not an imme-diate sister group to the Chilodontidae. The com-mon possession of a supramaxilla in Chalceus andthe Chilodontidae is, therefore, considered to rep-resent two independent acquisitions of that ossi-fication.

The Anostomidae, the hypothesized sistergroup of the Chilodontidae, has the relative sizesof the upper jaw elements reversed. Anostomidshave a very robust, triangular, tooth-bearing pre-maxilla with a relatively smaller maxilla locatedalong the posteroventral margin of the premax-illa. The overall form of the premaxilla is theresult of the apomorphous expansion of the as-cending premaxillary process that results in adistinctly triangular element (Figure 1B).

The highly specialized, fleshy suctorial mouthof prochilodontids is supported in part by distinc-tively structured upper jaw bones (Figure lc).The outwardly curved premaxilla and maxillaform rounded surfaces which serve as expandedattachment areas for the multilayered fleshy lipsthat bear the numerous rows of functional andreplacement teeth. The pronounced fenestrationof the maxilla is especially unusual, with one ormore large fenestra along the lateral surface ofthe element and a number of smaller aperturesoften arranged in posteroventrally aligned linearseries. Neither the outward curvature of the pre-maxilla and maxilla nor the high degree of fenes-tration of the latter are found among characi-

forms outside of the family. Equally unusual isthe strong bony process that extends posteroven-trally from the medial surface of the maxilla.That spur serves as the attachment area for themaxillo-mandibular ligament, which terminatesventrally on the dentary.

Congruent with the modifications of the upperjaw among the four families under discussion arevarious changes in the form of the lower jawelements. As in the case of the upper jaw, themandibular elements in curimatids are the leastmodified relative to those of most characiforms.The curimatid lower jaw (Figure 2D) is distin-guished from the generalized characiform condi-tion primarily in the lack of teeth and toothassociated structures, most notably the dentaryreplacement tooth trench. The dentary of curi-matids is relatively elongate and overlaps thetriangular angulo-articular. The retroarticular isa small lateral ossification along the ventral mar-gin of the latter bone. The other three familieshave the lower jaw foreshortened to differingdegrees compared to the relatively elongate man-dible of curimatids. Such a foreshortening, al-though hypothesized to be derived, has a phylo-genetic distribution incongruent with the phylog-eny arrived at here (p. 46); thereby representingat least two separate trends towards lower-jawshortening.

The lower jaw of chilodontids is little modifiedother than in being foreshortened (Figure 2A).The absence in chilodontids of direct contact ofthe teeth with the dentary is correlated with theabsence of a replacement tooth trench. As notedabove, that attribute also occurs, evidently inde-pendently, in the edentulous curimatids. Anos-tomids, in contradistinction, have a transverselywidened dentary that reflects the expansion ofthe replacement tooth trench. The widened re-placement tooth trench has both an increasedattachment area for the enlarged functional teeth(Figure 2B) and space for development of therelatively large replacement dentition. Anotherunusual aspect of the anostomid dentary replace-ment tooth trench is the large fenestra on thebone's ventral surface. A total ventral enclosureof the trench is typical for characiforms, and the

NUMBER 378 11

DEN

FIGURE 2.—Lower jaws: A, Caenotropus maculosus, USNM231545 (row of reduced teeth not shown); B, Schizodon fascia-turn, USNM 179507; c, Prochilodus rubrotaeniatus, USNM225419 (multiple rows of reduced teeth not shown); D,Potamorhina laticeps, USNM 121325, left side, lateral view.

anostomid form of aperture is not approximatedin any other characiform examined for the pres-ent study. The anostomid retroarticular lies in adistinct pocket formed by the angulo-articularand dentary (Figure 2B), contrary to the gener-alized condition in which enclosure by the bor-dering bones is minimal.

The prochilodontid lower jaw is particularlynoteworthy for the alteration of its retroarticularand dentary (Figure 2c). The prochilodontid re-troarticular is greatly reduced relative to the typ-ical characiform condition. Equally distinctive isthe repositioning of the retroarticular along theanteromedial surface of the angulo-articular,which results in its being barely visible in lateralview. This is a significant shift in size and positionfrom the larger, laterally located retroarticular inother characiforms. The numerous rows of re-placement teeth on the lower jaw of prochilodon-tids are contained in a greatly widened, laterallyrotated replacement trench with a large medialfenestra. A rotation of the dentary replacementtooth trench onto the anterior surface of thedentary also characterizes the Citharinidae (Vari,1979:267). The trench size in that family does notapproximate the condition in prochilodontids,

and more inclusive hypotheses of relationshipsindicate that these rotations in trench positionwere achieved independently (page 46; Vari,1979). The medial fenestra opening into the re-placement tooth trench in the Prochilodontidaehas not been encountered elsewhere in the order.

GILL ARCHES

The hypothesized plesiomorphous condition ofthe characiform gill arches' ventral portion iscomparable to that illustrated in Figures 3 and 4.Medially there are three ossified basibranchials,each bordered by paired hypobranchials. Theunossified posteriormost fourth basibranchial isan anteriorly wider, overall triangular element.The three pairs of hypobranchials are associatedwith the first through third basibranchials. Thehypobranchials typically are unelaborated, withthe exception of hypobranchial 3 which often hasan anteroventral, prong-shaped process. Whenpresent, that protrusion extends lateral to theventral aorta and is variously developed in differ-ent characiform groups. Ceratobranchials 1 to 3attach to the lateral margins of the associatedhypobranchials, whereas the fourth and fifth cer-atobranchials are in direct contact with the unos-sified fourth basibranchial. The first throughthird ceratobranchials are elongate, dorsoven-trally flattened bones that lack any pronounceddorsal ridges or processes. Ceratobranchial 4 hasa comparable overall form and a definite ridgealong its ventromedial edge, which serves as theattachment area for a connective tissue band.Ceratobranchial 5 is expanded anteromediallyinto a flattened, largely horizontal plate, whichbears a patch of moderately developed, unicus-pidate, dorsally, or posterodorsally directed teeth.

This hypothesized primitive condition of thegill arches among characiforms is modified, oftenmarkedly, among members of the families Curi-matidae, Prochilodontidae, Anostomidae, andChilodontidae. The phylogenetic distribution ofthe various, derived restructurings are congruentwith the recognition of the four-taxon assemblageas a natural evolutionary unit and with its divi-sion into two monophyletic subunits. One subunit


BB,

BB-

FIGURE A.—Bryconfakatus, USNM 226161, ventral portion ofFIGURE 3.—Btyconfalcalus, USNM 226161, ventral portion of gill arches, ventral view (denser stippling represents carti-gill arches, dorsal view (denser stippling represents cartilage). lage).

H,

BB. BB,

BB

BB

FIGURE b.—Cunmata vittata, USNM 231434, ventral portion FIGURE 6.—Curimata vittata, USNM 231434, ventral portionof gill arches, dorsal view (denser stippling represents carti- of gill arches, ventral view (denser stippling represents car-lage). tilage).

NUMBER 378 13

consists of curimatids and prochilodontids, theother of anostomids and chilodontids.

The most readily apparent difference betweenthe basic form of the ventral portion of the gillarches common to prochilodontids and curima-tids and the hypothesized plesiomorphous condi-tion of these elements for characiforms is themarked reduction or absence of dentition on thelower pharyngeal (fifth ceratobranchial) in theadults of these families (Figure 5). Fifth cerato-branchial dentition is present in postlarvae ofboth families. In specimens of Prochilodus of ~ 35mm SL, six to eight unicuspidate teeth are ar-ranged in two parallel rows along the postero-medial portion of the bone. The progressive on-togenetic loss in lower pharyngeal teeth results inthe fifth ceratobranchials of prochilodontidsbeing edentulous in — 85 mm SL specimens.Similarly curimatids under ~ 25 mm SL com-monly have a patch of unicuspidate teeth on thefifth ceratobranchial. An ontogenetic loss of pha-ryngeal dentition that leads to toothless fifth cer-atobranchials is common to the vast majority ofcurimatid species, particularly those that attainan adult standard length over 50 mm. In some ofthe smaller species, however, the reduction is notcarried to its ultimate limit. Fully grown Curima-topsis species (i.e., ~ 25 mm SL) retain a distinct,albeit greatly reduced, patch of teeth on the fifthceratobranchial.

Within the Characiformes a reduction or elim-ination of lower pharyngeal dentition also occursin the Neotropical genus Anodus (Roberts,1974:223) and the African family Citharinidae(Vari, 1979:303). Various shared derived char-acters unite these taxa respectively to the Hem-iodontidae (Roberts, 1974) and Distichodontidae(Vari, 1979). The latter families otherwise havea full complement of lower pharyngeal dentition.The reduction or elimination of fifth ceratobran-chial dentition in Anodus and the Citharinidae isconsequently most parsimoniously considered ahomoplasy relative to the comparable adaptationcommon to curimatids and prochilodontids.

The lower portion of the gill arches also pro-vides some other phylogenetically interestingcharacters. A prominent, anteriorly directed proc-

-< ' Ef , i

FIGURE 7.—Psectrogaster amazomca, AMNH 40088SD, cerato-branchial 4, left side, lateral view (arrow points to ventralprocess).

ess extends from the fourth certobranchial's an-teromedial surface in prochilodontids and curi-matids (Figures 6, 7). This process is an apomor-phous elaboration of the ridge on the medialportion of the fourth ceratobranchial in othercharaciforms. In a similar fashion the anteriorcorner of the third hypobranchial is often antero-ventrally prolonged in characiforms into a prong-shaped process that extends lateral to the ventralaorta. Curimatids have the ventral portion of thatprocess apomorphously further expanded ven-trally and posteriorly to form a vertical wall thatextends lateral to, and evidently partially sup-ports, the ventral aorta (Figures 6, 8). Anteriorand posterior extensions of this process' ventralmargin are positioned along the lateral wall ofthe blood vessel. The degree of development ofthese secondary processes vary among differentsubunits of the family.

Certain hemiodontids {Anodus, Hemiodus, Hem-iodopsis (Vari, 1982a: 191), Micromischodus and Bi-vibranchia) and chilodontids (Caenotropus) alsohave prominent ventral processes of the third

FIGURE 8.—Psectrogaster amazomca, AMNH 40088SD, basi-branchial 3 and hypobranchial 3, left side, lateral view,showing ventral process of hypobranchial 3.


hypobranchial. The hypobranchial projections inthose taxa arise from the ventrolateral margin ofthe main body of the hypobranchial rather thanthe ventromedial portion of the element as incurimatids. Those structures are consequentlyconsidered nonhomologous with the adaptationsin the latter family; a hypothesis congruent withthe arrived at hypothesis of relationships.

The first basibranchial is the final phylogenet-ically significant character discovered in the ven-tral portion of the gill arches in curimatids andprochilodontids. Most characiforms have a rela-tively well-developed conical or elongate first ba-sibranchial that extends distinctly anterior to thefirst hypobranchial (Figures 3, 10). In prochilo-dontids, however, basibranchial 1 is a small, tri-angular element approximate to the anterior mar-gin of basibranchial 2 and barely extends ante-riorly to the level of the anterior terminus of thefirst hypobranchial (Figure 9). This reductionaltrend is carried further in the Curimatidae, whichlack a separate, ossified first basibranchial (Figure5). Available evidence does not permit a decisionas to whether the conical cartilagenous processanterior to the ossified second basibranchial ofcurimatids represents the unossified remnant of

Hv *V

BB2

BB-

CR

FIGURE 10.—Caenotropus maculosus, USNM 231545, ventralportion of gill arches, dorsal view (denser stippling representscartilage).

BB-

FIGURE 9.—Semaprochilodus taeniurus, USNM 231536, ventralportion of gill arches, anterior region, dorsal view (denserstippling represents cartilage).

the reduced first basibranchial of prochilodontids.Nonetheless, an ossified first basibranchial iswidespread in otophysans and its reduction andultimate loss are, thus, considered synapomor-phies of different levels of universality within theprochilodontid-curimatid lineage. The lack of anunossified first basibranchial also occurs homo-plasiously in the examined adults of at least someanostomids {Rhytiodus microlepis, Pseudanos irinae

(Winterbottom, 1980:40)) and parodontids (Par-odon suborbitalis).

The primary shared derived character for an-ostomids and chilodontids in the gill arches' ven-tral portion involves the fifth ceratobranchialdentition (Figure 10). Chilodontids and anostom-ids have enlarged bi- or multicuspidate teethrather than the relatively small or moderatelydeveloped unicuspidate pharyngeal dentitioncommon to most characiforms. Bicuspidate pha-ryngeal teeth with unequally developed cusps are

NUMBER 378 15

universal among anostomids (Figure 11A). Theelaboration of the distal portion of the fifth cer-atobranchial dentition progresses further in Caen-otropus and Chilodus in which the lower pharyngealteeth have three cusps of markedly different sizealigned along the longitudinal body axis. Thisdentition is comparable to that of the upperportions of the Caenotropus gill arches, as illus-trated in Figure UB.

Lower pharyngeal dentition enlarged to thedegree found in chilodontids and anostomids hasnot been reported or observed elsewhere amongcharaciforms (Roberts, 1969:424; Winterbottom,1980:2). Consequently, its possession is most ap-propriately considered a shared derived charac-ter. Non-unicuspidate pharyngeal dentition isknown to occur in only one other characiform,the hemiodontid Bivibranchia, which has tricuspi-date teeth on the fifth ceratobranchial (Roberts,1974, fig. 37). The equally developed, transverselyor obliquely transversely oriented, tooth cusps ofBivibranchia do not appear to be homologous withthe anteroposteriorly oriented cusps of markedlydifferent size found in chilodontids and anostom-ids. Furthermore, as discussed by Roberts, Bivi-branchia is evidently most closely related to hem-iodontids, a family otherwise characterized byunicuspidate pharyngeal teeth. The anatomicaland phylogenetic data concordantly indicate thatthe tricuspidate dentition of Bivibranchia is a hom-oplasy relative to the apomorphous bi- or multi-cuspidate tooth forms common to the Chilodon-tidae and Anostomidae.

Although numerous fish groups with promi-nent epibranchial organs have associated altera-tions of the gill arches' dorsal portion, chilodon-tids are unusual, if not unique, in also incorpo-rating elements of the ventral portion of the

FIGURE 11.—Form of fifth up-per pharyngeal tooth plate den-tition, left side, lateral view:A, Laemolyta species, USNM179514; B , Caenotropus labyrinlhi-

cus, USNM 231544; c, Chiloduspunctatus, USNM 231542. (Scale= 0.1 mm.)

FIGURE 12.—Caenotropus maculosus, USNM 231545, fifthceratobranchial, left side, dorsal view.

branchial apparatus into the system. Most no-table are the modifications of the fourth and fifthceratobranchials. The enlarged fourth cerato-branchial (Figures 10, 21) has greatly expandedposterior and posteroventral surfaces that form abroad curved plate which closely approximatesneighboring portions of the enlarged fifth cera-tobranchial (Figures 10, 12, 21). The increasedsize and modified form of the fourth ceratobran-chial are seemingly unique to chilodontids amongostariophysans. The lower pharyngeal (fifth cer-atobranchial) of chilodontids is similarly totallyrestructured (Figures 10, 12) relative to the con-dition basic for characiforms (Figure 3). Mostdistinctive is the great expansion of the ventralportion of ceratobranchial 5 into a broad, cup-shaped plate that extends under and closely ap-proximates the posterior and ventral surfaces ofthe enlarged fourth ceratobranchial (Figures 10,21). Dorsomedially the fifth ceratobranchial is aflattened, fenestrated plate that extends ante-riorly along the dorsal surface of the cartilagenousfourth basibranchial. Posteriorly the tooth-bear-ing portion of the bone is expanded dorsally androtated anteriorly. This results in the associateddentition being oriented directly anteriorly orhaving only a slight dorsal inclination contraryto the plesiomorphous dorsal or moderately an-terodorsal orientation of these teeth in other char-aciforms. The chilodontid gill arches also haveprominent dorsal ridges that arise from the main


PB2

FIGURE 13.—Bryconfalcatus, USNM 226161, dorsal portion ofgill arches, left side, ventral view (denser stippling representscartilage).

body of ceratobranchials 1, 2, and 3 (Figure 10).Such elaborations are distinctive relative to theprimitive, flattened dorsal surface of these ele-ments in other characiforms. Whether, and how,these ridges are functionally related to the epi-branchial organs is unknown. Ceratobranchialridges also occur in the hemiodontid Bivibranchia.In that genus the margins of the ridges arestraight and the main body of the entire cerato-branchial is bowed ventrally; a system differentfrom the chilodontid form of ceratobranchialridge. Furthermore, Bivibranchia apparently is notclosely related to the assemblage formed by chil-odontids and anostomids. The conjunction ofthese factors leads to an assumption that theceratobranchial elaborations in Bivibranchia andthe Chilodontidae are homoplasious.

Characiforms as a whole more typically dem-onstrate alterations of the gill arches' dorsal por-tion than restructurings of the branchial appara-tus' ventral region. This dichotomy is apparentamong curimatids, prochilodontids, anostomids,and chilodontids, all of which have some of themost pronounced alterations of the dorsal por-tions of the gill arches among characiforms. Onceagain some modifications are synapomorphousfor the four-family unit, while others delineatetwo subgroups within the assemblage. The firstsubgroup consists of the Curimatidae and Pro-

chilodontidae and the second of the Anostomidaeand Chilodontidae.

The hypothesized plesiomorphous characiformmorphology of the gill arches' dorsal portion iscomparable to that illustrated for Brycon (Figures13, 14). Four infrapharyngobranchials occuralong the midline. The first, a small, rod-like ortriangular element associated with the first epi-branchial, is ligamentously attached dorsally tothe neurocranium. Infrapharyngobranchials 2and 3 are triangular, anteriorly directed elementsthat contact their respective epibranchials lat-erally. The fourth infrapharyngobranchial is aflattened cartilage that extends between the thirdinfrapharyngobranchial and the fourth epibran-chial. Epibranchials 1 through 3 are elongate,flattened structures unelaborated other than forproximal uncinate processes. The fourth epibran-chial is a more complex triangular bone with awell-developed, dorsally directed process abovethe fifth upper pharyngeal tooth plate. Postero-laterally, the fourth epibranchial articulates withthe relatively small, rod-shaped, cartilaginousfifth epibranchial (Figures 13-15). The fourthand fifth epibranchials diverge dorsally, with anintervening groove that serves as the passage forthe fifth efferent branchial artery. Upper pharyn-geal tooth plate 4 is attached to the ventralsurface of the fourth infrapharyngobranchial,

PEL,

FIGURE 14.—Bryconfalcatus, USNM 226161, dorsal portion ofgill arches, left side, dorsal view (denser stippling representscartilage).

EAF

UP,

FICURE 15.—Fourth and fifth epibranchials, fifth upperpharyngeal tooth plate, and associated dentition, right side,medial view: A, Brycon falcatus, USNM 226161; B, Curimatavittala, USNM 231434; c, Ichthyoeltphas species, USNM231437; D, Schizodon fasciatum, USNM 179507. (Scales = 1.0mm; stippling represents cartilage).

while upper pharyngeal tooth plate 5 is joineddorsally to the fourth epibranchial's ventral sec-tion. The two tooth plates together form a contin-uous, flattened bony sheet that bears a patch ofsmall unicuspidate teeth, which oppose those onthe fifth ceratobranchial. Comparable small uni-cuspidate dentition often also occurs in smallpatches on the ventral surfaces of the second andthird infrapharyngobranchials.

Curimatids, prochilodontids, anostomids, andchilodontids have in common a pronounced re-duction in the amount and distribution of pha-ryngeal dentition. Most notable is their lack ofteeth on the fourth upper pharyngeal tooth plate(UP4), a condition here considered to be derivedwithin characiforms relative to the previouslydescribed presumed primitive condition. Re-duced or absent dentition on the fourth upperpharyngeal tooth plate is not unique to this par-ticular assemblage within the order. Among OldWorld characiforms, an edentulous UP4 occurs inall citharinids and in some, but not all, species ofthe distichodontid genera Hemigrammocharax, Nan-

17

nocharax and Neolebias (Vari, 1979:303-305). Re-duced dentition on the fourth upper pharyngealtooth plate also characterizes the Neotropicalhemiodontid genus Anodus. Although derived, thereduction or loss of dentition on the fourth upperpharyngeal tooth plate does not delimit a naturalassemblage. The species of Hemigrammocharax,Nannocharax, and Neolebias that have reduced den-tition on the fourth upper pharyngeal tooth plateare all relatively small forms with congruent re-ductions in the overall ossification of other headparts. Furthermore, the data on relationshipswithin and between these distichodontid generamake it most parsimonious to conclude that thisloss of dentition was achieved independently ineach lineage and that a complete pattern of pha-ryngeal dentition was ancestral for distichodon-tids (Vari, 1979:305). The citharinids Citharidiumand Citharinus, which also lack teeth on the fourthupper pharyngeal tooth plate, are hypothesizedto form the sister assemblage to the Distichodon-tidae (Vari, 1979:324). The presence of dentitionon the fourth upper pharyngeal tooth plate incharaciform outgroups and most members of theDistichodontidae indicates that the edentulousnature of the element in citharinids, althoughmost simply considered a synapomorphy for themembers of that family, is a homoplasy relativeto the comparable condition in curimatids, pro-chilodontids, anostomids, and chilodontids. Anumber of derived characters place Anodus, inwhich UP4 lacks teeth, within the Hemiodontidae(Roberts, 1974:429). The remaining members ofthat family have the more typical plesiomorphouscharaciform pattern of complete pharyngeal den-tition, which is consequently considered the prim-itive condition for the Hemiodontidae.

Curimatids, prochilodontids, anostomids, andchilodontids or subunits of that assemblage sharerestructurings of the fourth upper pharyngealtooth plate. All four families lack the plesiomor-phous close contact of the fourth and fifth upperpharyngeal tooth plates (Figures 16-19). The ab-sence of that immediate association of those ele-ments is thus hypothesized to be a shared derivedcharacter for the quadrifamilial grouping. Curi-matids and prochilodontids have a somewhat

18

PB,


'5PB3

UP,

FIGURE 16.—Curimata vitlata, USNM 231434, dorsal portionof gill arches, left side, ventral view (denser stippling repre-sents cartilage).

PB,PB

FIGURE \9.—Caenotropus maculosus, USNM 231545, dorsalportion of gill arches, left side, dorsal view (denser stipplingrepresents cartilage).

FIGURE 17.—Curimata vitlata, USNM 231434, dorsal portionof gill arches, left side, dorsal view (denser stippling repre-sents cartilage).

ER

FIGURE \8.—Caenotropus maculosus, USNM 231545, dorsalportion of gill arches, left side, ventral view (denser stipplingrepresents cartilage).

FIGURE 20.—Anostomus species, USNM 231540, dorsal por-tion of gill arches, left side, dorsal view (denser stipplingrepresents cartilage).

reduced fourth upper pharyngeal tooth plate thatno longer retains its primitive direct contact pos-teriorly with the elements of the fifth gill arch(Figure 16). In addition to its separation from thefifth upper pharyngeal tooth plate, UP4 is alsoapomorphously altered into a curved sheet thatsurrounds the ventral, lateral, and sometimes por-tions of the dorsal surfaces of the fourth infra-pharyngobranchial (Figures 16, 17); a markedshift from the flat tooth plate form typical forcharaciforms. Anostomids and chilodontids re-tain the flattened form of the fourth upper pha-

NUMBER 378 19

ryngeal tooth plate hypothesized to be primitivebut have the posterior portion of the elementshifted laterally (Figure 18). Consequently, theposterior portion of UP4 now contacts the antero-lateral surface of the main body of the fourthepibranchial rather than the anterior margin ofthe fifth upper pharyngeal tooth plate. Themarked reorientation of the tooth plate apparentin anostomids is more pronounced in chilodon-tids.

Subunits of the four-family assemblage underconsideration have the form of the fifth upperpharyngeal tooth plate (UP5) modified in differ-ent ways relative to the hypothesized plesiomor-phous flattened plate common to most characi-forms. Prochilodontids have a transversely com-pressed fifth upper pharyngeal tooth plate mov-ably articulated with the fourth epibranchial(Figure 15c; Roberts, 1974, fig. 22). In the An-ostomidae and Chilodontidae the fifth upper pha-ryngeal tooth plate is vertically thickened relativeto the condition generalized for characiforms.Among chilodontids, this thickened plate is ad-ditionally rotated posterodorsally. As a conse-quence, the primitively ventral surface of UP5 isoriented to face posteriorly (Figures 18, 19). Thevarious restructurings of the fifth upper pharyn-geal tooth plate in prochilodontids, anostomids,and chilodontids are not known to occur else-where among characiforms. Therefore, they areconsidered derived characters of differing levelsof universality. In the Curimatidae, UP5 is apo-morphously expanded posteriorly and postero-ventrally into a curved, somewhat convolutedossification (Figures 15B, 16) that contacts andmatches the curved dorsal surface of the fifthceratobranchial. A posterior extension of UP5somewhat comparable to that in curimatids oc-curs in the Old World citharinids (Vari, 1979, fig.28). The pronounced difference in the morphol-ogy of the bony plates in the two families and theoverall distribution of derived characters withincharaciforms indicates that these are analogousrather than homologous changes.

Distinctive modifications of the fifth upperpharyngeal tooth plate dentition distinguish sub-units of the four-family assemblage. Prochilodon-tids have reduced teeth with swollen bases on

that element (Roberts, 1973, fig. 22). Comparableteeth occur at some point in the ontogeny ofvarious curimatids. The majority of curimatidspecies have carried the trend towards tooth re-duction further and have very few teeth on thefifth upper pharyngeal tooth plate, or have anedentulous bone. Roberts (1973:219) reportedthat "in Curimatidae third and fourth epibran-chial and fifth ceratobranchial toothplates bearnumerous conical teeth . . .." Observations asso-ciated with the present study have shown, how-ever, that curimatids are characterized by themarked reduction or absence of teeth on the fifthupper pharyngeal tooth plate, the absence ofteeth on the fifth ceratobranchial, and an eden-tulous fourth upper pharyngeal tooth plate. Else-where among characiforms, a reduction in theUP5 dentition comparable to that in curimatidsis known only in the Old World family Cithar-inidae (Vari, 1979:303). However, numerous syn-apomorphies unite the Curimatidae to the Pro-chilodontidae (see "Phylogenetic Reconstruc-tion") and various derived characters indicatethat the Citharinidae forms the sister group tothe Distichodontidae (Vari, 1979:324). Giventhose more inclusive hypotheses of relationship, itis simplest to consider the edentulous nature ofthe fifth upper pharyngeal tooth plate in cithar-inids to be a homoplasy relative to the reduceddentition on that bone in curimatids. This con-vergent reduction in the UP5 dentition in themicrophagous curimatids and citharinids may bea consequence of the decreased utility of pharyn-geal teeth in fishes with that feeding habit. Thecongruent, hypothesized homoplasious expansionof the fifth upper pharyngeal tooth plate is pos-sibly associated with the food concentrating func-tion of the epibranchial organs.

The enlarged bi- or multicuspidate dentitionon the fifth upper pharyngeal tooth plate ofanostomids and chilodontids is comparable tothat on the fifth ceratobranchial. Anostomidshave totally bicuspidate pharyngeal dentition(Figures 11A, 15D). Caenotropus (family Chilodon-tidae) has tricuspidate pharyngeal teeth on bothventral and dorsal portions of the gill arches(Figure 11B). The dentition on the fifth upperpharyngeal tooth plate of Chilodus (Chilodonti-


dae) is quadricuspidate (Figure 1 lc) and opposesthe tricuspidate teeth on the fifth ceratobranchial.Enlarged non-unicuspidate UP5 dentition is lim-ited to the preceding families among characiformsexamined, with the single exception of the hem-iodontid Bivibranchia (Roberts, 1974, fig. 37). Asin the case of the lower pharyngeal dentition, thesupernumerary tooth cusps on the teeth in theupper portion of the gill arches in this hemiodon-tid are evidently nonhomologous with those ofanostomids and chilodontids.

An overall restructuring of many other ele-ments in the dorsal portions of the gill archesoccurs in anostomids and chilodontids. The formof the anteromedial process of the third epibran-chial of anostomids is distinctive among characi-forms. This transversely widened portion of thebone has a distinct process that is somewhatdorsally recurved, thereby extending over thedorsal surface of the fourth infrapharyngobran-chial (Figure 20). The third infrapharyngobran-chial of anostomids and chilodontids is trans-versely widened posteriorly relative to the ple-siomorphous characiform condition. This expan-sion is most pronounced in chilodontids (Figures18, 19), in which the associated third epibranchialhas an elongate slender medial process that ex-tends dorsally to the fourth infrapharyngobran-chial. Chilodontids have pronounced ridges onthe ventral surfaces of epibranchial 1, 2, and 3that oppose the dorsal ridges on the three ante-riormost ceratobranchials (Figure 18). Mosthighly developed is the process on epibranchial3. Once again the ridge form in the Chilodontidaediffers in detail from that of the hemiodontidBivibranchia. The anatomical differences and thedata associating Bivibranchia phylogeneticallywith the Hemiodontidae (Roberts, 1974) indicatethat the ridges in that genus and the Chilo-dontidae are evidently analogous rather thanhomologous elaborations of the involved cerato-branchials.

Epibranchial 4 is notably modified in the Cur-imatidae, Prochilodontidae, Anostomidae, andChilodontidae. In curimatids and prochilodon-tids, the primitively dorsally directed process ofthe fourth epibranchial (the dorsal extension of

Nelson (1967) and the suprapharyngobranchialprocess of Bertmar et al. (1969)) is expanded andreoriented anteriorly, thereby extending over thedorsal surface of the fourth infrapharyngobran-chial (Figures 15B,C, 17). The resultant sheet ofbone and cartilage runs parallel to, and is closelyassociated with, the lateral surface of the pharyn-geal outpocketing of the epibranchial organ. Thisvertical plate in these taxa is also extended pos-teriorly and posterodorsally by a significant an-terodorsal expansion of the cartilaginous fifthepibranchial, which attaches to the rear of thefourth epibranchial (Figures 15B,C). The enlargedform of the cartilage contrasts with the usualcharaciform condition of a relatively small fifthepibranchial dorsally separated from the fourth.The dorsal contact between the fourth and fifthepibranchials in curimatids and prochilodontidsresults in the encirclement of fifth efferent bran-chial artery. Comparable expansions of the fourthand fifth epibranchials with a resultant enclosureof the efferent artery are associated with well-developed epibranchial organs in various fishgroups (Nelson, 1967:76, 83). However, otherthan curimatids and prochilodontids, no chara-ciform with well-developed epibranchial organsis known to have such an association of theseelements.

In anostomids (Figure 15D) and chilodontids,the form of the fourth epibranchial is dramati-cally altered relative to the condition hypothe-sized to be plesiomorphous for characiforms. Thedorsal process of the bone in these families isconsiderably thickened transversely and foreshor-tened longitudinally. Furthermore, the primaryaxis of the process also has a posterodorsal align-ment rather than the dorsal or anterodorsal ori-entation that characterizes other characiforms.The anterior portion of the fourth epibranchial,which articulates with the cartilaginous fourthinfrapharyngobranchial, has undergone a consid-erable vertical expansion in anostomids and chil-odontids. That expanded portion of the bone andthe posteriorly reoriented dorsal process delimita distinct, transversely aligned notch along thedorsal surface of the bone (Figures 15D, 19, 20).The notch serves as an attachment area for the

NUMBER 378 21

very large obliquus dorsal is muscle that extendsanteriorly to the fourth infrapharyngobranchial.In no other characiform examined has such ahighly developed form of this upper pharyngealmuscle been found. The alteration of the overallform of the fourth epibranchial is carried furtherin chilodontids (Figures 18, 19), which have theventromedial surface of that element rotated pos-terodorsally. As a consequence the dentition onthe associated fifth upper pharyngeal tooth platepoints posteroventrally rather than ventrally asin other characiforms.

Chilodontids also have a number of modifica-tions of the upper portions of the gill arches thatare associated with their unique form of epibran-chial organ.

EPIBRANCHIAL ORGANS

Epibranchial organs have been reported in anumber of fish groups including clupeiods, osteo-glossiforms, and ostariophysans. Within the Os-tariophysi (sensu Rosen and Greenwood, 1970,and Fink and Fink, 1981), such a system is knownin some gonorhynchiforms, characiforms, and cy-priniforms. The typical teleostean epibranchialorgan is primarily associated with the fourth epi-branchial and consists of paired dorsal diverticuliof the branchial chamber. Each diverticulumcommunicates posteroventrally with the pharynxvia an aperture in the posterodorsal region of thebranchial chamber. Most early students studyingthe system suggested that the epibranchial organsin characiforms had a respiratory function (e.g.,Sagemehl, 1887; Rauther, 1910). Subsequent re-search by Heim (1935) on the Old World chara-ciforms Distichodus (Distichodontidae) and Cith-arinus (Citharinidae) and the New World generaCurimatus (Curimatidae) and Prochilodous (Prochil-odontidae), followed by Anglescu and Gneri's(1949) research on Prochilodus and Bertmar's study(1961) on Citharinus indicate that epibranchialorgans in the examined taxa are not accessoriesof the respiratory system. Rather they serve tosense food items in, and concentrate them from,the water column and/or substrate (see Bertmaret al., 1969, for a detailed discussion).

Moderately developed pharyngeal outpocket-ings associated with the fourth and fifth epibran-chials have been reported in some African char-acids and distichodontids (Daget, 1958, 1959,1960; Bertmar, 1961:156). Highly developed epi-branchial organs, however, are present within thesuborder only in the African Citharinidae (Sage-mehl, 1885, 1887; Daget, 1962a, 1962b) and theNeotropical Chilodontidae, Prochilodontidae,and Curimatidae (Kner, 1861; Heim, 1935) (forHemiodontidae, see p. 22). Although well-devel-oped epibranchial organs are hypothesized to bederived for characiforms, they apparently are nota shared derived character for all the taxa pos-sessing them. Pronounced differences between thethree types of highly developed epibranchial or-gans in the group cast doubt on the homology ofthese adaptations, as do present concepts ofhigher level phylogenetic relationships amongcharaciforms.

The epibranchial organ in the citharinids Cith-aridium and Citharinus is an elaborate system offoliate lobes and dichotomously branching inter-nal tubes supported by numerous bony spicules(Daget, 1962a, figs. 6-8). Epibranchial organs ofthat type are presently unknown elsewhereamong teleostean fishes (Bertmar et al., 1969,table 1). This unique anatomy suggests that thecitharinid epibranchial organ is a synapomorphyfor the members of that family, analogous ratherthan homologous with the well-developed pha-ryngeal diverticuli in some other characiforms.That hypothesis is congruent with more inclusiveconcepts of relationships based on shared derivedcharacters and the lack or poor development ofepibranchial diverticuli in distichodontids, thehypothesized sister group to citharinids (see Vari,1979:322, for a further discussion).

The epibranchial organs of curimatids and pro-chilodontids are elongate, longitudinally alignedsacs with both longitudinal and circular musclesin their walls (Heim, 1935:64, 71). Commencingat the posterodorsal limits of the branchial cham-ber, the diverticuli then extend dorsally and an-terodorsally over the fourth infrapharyngobran-chial, the fourth and fifth upper pharyngeal toothplates, and part of the tooth epibranchial. The


medial surfaces of the sacs are in contact orslightly separated, with each diverticulum en-closed laterally by the enlarged cartilaginous fifthepibranchial and anteriorly expanded osseousand cartilaginous dorsal process of the fourthepibranchial. Although Bertmar et al. (1969:20)state that the hemiodontid Hemiodus and the chil-odontid Caenotropus have large, sac-like epibran-chial diverticuli, the present study has not foundsuch pharyngeal outpocketings in those genera.The only hemiodontid with a noticeable poster-odorsal epibranchial pouch is Anodus, which wasnot placed in the Hemiodontidae at the time ofthe study by Bertmar et al. The diverticulum inAnodus is relatively small outpocketing at the pos-terodorsal limit of the branchial apparatus and isformed largely by a cartilage associated with thefifth ceratobranchial. It, furthermore, does notextend dorsal of the fourth epibranchial andfourth infrapharyngobranchial. The Anodus pha-ryngeal diverticulum thus differs in both positionand structure relative to the sac-like, muscular,dorsally located epibranchial organs of curima-tids and prochilodontids; the two forms of diver-ticuli are consequently considered nonhomolo-gous.

The cited presence of sac-like epibranchial or-gans in Caenotropus is a misinterpretation of thesystem in that genus. The pharyngeal diverticu-lum of chilodontids, including Caenotropus, is com-posed largely of cartilage and connective tissuesheets and differs in numerous details from thatof curimatids and prochilodontids. Those differ-ences and available data on phylogenetic rela-tionships of chilodontids support the hypothesisthat the pharyngeal diverticulum in the Chilo-dontidae is nonhomologous with the sac-like mus-cular epibranchial organ of some other characi-forms. Thus, the sac-like epibranchial organ formof the Curimatidae and Prochilodontidae is ap-parently synapomorphously limited to those fam-ilies among characiforms, with the reports byBertmar et al. (1969) of large sac-like epibranchialorgans in the Hemiodontidae and Chilodontidaebeing evidently erroneous.

The final highly developed epibranchial organform within the Characiformes is common to all

chilodontids. Bertmar et al. (1969:20) erroneouslyconsidered the epibranchial organ of the chilo-dontid Caenotropus to be equivalent to the ex-panded, sac-like epibranchial organs of curima-tids and prochilodontids. They also described theepibranchial organ of the chilodontid Microdus as"more convoluted and heliciform . . . than that ofother characoids." As alluded to above, the epi-branchial organ of Caenotropus is nonhomologouswith that in curimatids and prochilodontids.Rather it is like that described by Bertmar et al.for Microdus, which is an older, although preoc-cupied, generic name for species now placed inCaenotropus (Gery, 1964:6).

Several features of the chilodontid epibranchialorgan are unknown among other characiformgroups. Foremost among these are the restructur-ings of the fourth and fifth ceratobranchials andtheir incorporation into the epibranchial organsystem (see "Gill Arches"). The exact functionalsignificance of the expansion of the apparatusinto the ventral portions of the gill arches isuncertain. However, the adjoining surfaces of theexpanded ceratobranchials form two broadlycurved, matching plates (Figures 10, 21) with theadjacent surfaces in close contact and covered bysheets of smooth connective tissue. Laterally thiscovering layer is developed into a series of radiallyarranged, interdigitating ridges (Gery, 1964, fig.2), which presumably filter food particles fromthe water passed from the oral cavity.

The dorsal portions of the epibranchial organsalso incorporate atypical elements. The postero-dorsal margin of the fifth ceratobranchial isjoined to a broad, shell-shaped connective tissuesheet (Figure 21). That structure extends upwardabove the level of the dorsal portions of theepibranchials and forms the posterior wall of apartially helical, medially spiraling epibranchialchamber. The anterior wall of the transverselyoriented chamber is largely formed by the carti-laginous fifth epibranchial, which is significantlyexpanded vertically and transversely relative tothe condition in other characiforms (Figure 21).The fifth epibranchial's ventral margin is at-tached by a connective tissue sheet to the poster-oventral margin of the fourth epibranchial. That

NUMBER H78 23

EOM

CTS

FIGURE 21.—Caenotropus maculosus, USNM 231545, fourthand fifth branchial arches and anterior portion of epibran-chial organ muscle, right side, anterolateral view.

common connective tissue band then continuesventrally to attach to the fourth ceratobranchial'sdorsolateral surface. This soft tissue complex in-terconnects the dorsal cul-de-sac with the spacedelimited by the expanded fourth and fifth cera-tobranchials; an association between the dorsaland ventral portions of the branchial apparatuslimited to chilodontids among fishes.

Chilodontids lack the circular and longitudinalmusculature found in the sac-like epibranchialorgans of curimatids and prochilodontids. Rathertheir epibranchial organs are formed largely ofnonmuscular tissue with only a single, narrow,vertical muscle band that extends along the ex-ternal surface of the anterior and posterior wallsof the epibranchial portion of the diverticulum(Figure 21). The muscle segment on the anteriorface of the epibranchial organ arises from theprimitively dorsal surface of the fourth epibran-chial. It extends dorsally from its origin to passover the upper edge of the fifth epibranchial and

joins its counterpart on the posterodorsal surfaceof the epibranchial organ. The posterior portionof the muscle band arises in part on the postero-ventral margin of the fifth ceratobranchial andpartially on the posterior surface of the shell-shaped connective tissue band surmounting thatbone. The anterior and posterior muscle bundlesjoin on the posterior surface of the epibranchialorgan slightly ventral to the dorsal border of theepibranchial organ. The resultant common mus-cle band extends dorsally from this point of con-tact to insert on the neurocranium. Although theexact homology of these epibranchial organ mus-cles is uncertain, they are considered a synapo-morphy for all chilodontids because of the lack ofany comparable musculature in the branchialapparatus of other characiforms.

The concurrent presence in chilodontids ofwell-developed pharyngeal dentition and largeepibranchial organs runs counter to the generalpattern noted by Nelson (1967:81), in which thedevelopment of large epibranchial diverticuli istypically accompanied by a reduction or elimi-nation of pharyngeal dentition. The functionalsignificance of the association of such seeminglydisparate adaptations—a large epibranchial or-gan correlated with filter-feeding and enlargedpharyngeal teeth usually associated with shred-ding of larger food items—is unknown. An in-depth examination of food habits in the groupmay resolve the question.

The possession of epibranchial organs has beenhypothesized to be the ancestral condition forvarious groups of fishes by Bertmar et al.(1969:44). They consequently suggested that theabsence of these adaptations in members of var-ious lineages was achieved by repeated indepen-dent losses. Nelson (1967), in contrast, advancedthe suggestion that the presence of such epibran-chial diverticuli in different groups of fishes rep-resented independent acquisitions and that theabsence of these pharyngeal outpocketings con-stituted the ancestral state for teleostean fishes.The scope of this study is too limited to analyzewhich of these hypotheses is most parsimoniousfor all subunits of the Teleostei. However, themorphology of well-developed epibranchial or-


gans in characiforms indicates that the threeforms of the diverticuli in chilodontids, citharin-ids, and the unit formed by curimatids and pro-chilodontids are analogous, homoplasious devel-opments. The hypothesis that these diverticuliarose independently in each lineage is also themost parismonious within the context of our pres-ent knowledge of relationships within the Char-aciformes. Similarly, the epibranchial organ inthe cyprinid Hypopthalmichthys is an analog of thevarious systems in characiforms, with that genusmost closely related to other cyprinids that lackthe adaptation (Howes, 1981). Comparable ar-guments apply to the pharyngeal diverticuli ofvarious gonorhynchiformes. Therefore, the ana-tomical and phylogenetic evidence concordantlyindicates that, for the Ostariophysi, the hypoth-esis that large epibranchial organs are apomor-phous independent acquisitions as stated by Nel-son, is more parsimonious than the hypothesis ofBertmar et al., who considered such outpocket-ings ancestral for all teleosts.

HYOID APPARATUS

The four families under discussion possess aseries of hyoid arch alterations congruent withadaptations of other portions of the oral appara-tus. The generalized and presumed plesiomor-phous hyoid arch form for characiforms is com-parable to that in Brycon (Weitzman, 1962, fig.1 lc). The arch's lateral portion consists of ventraland dorsal hypohyals followed posteriorly by an-terior and posterior ceratohyals. The hypohyalsare transversely wide, complex bones, whereas theceratohyals are distinctly compressed, laterallyunelaborated elements. The anterior ceratohyalis approximately twice as long as the posterior(e.g., Prochilodus, Figure 22). The hyoid series iscompleted posterodorsally by a rod-like interhyal,which attaches ligamentously to the dorsal sur-face of the posterior ceratohyal (e.g., Leporinus,Figure 23). In the majority of characiforms thebasihyal is a relatively elongate element, ossifiedapproximate to the hypohyals and cartilaginousanteriorly. The ossified and cartilaginous portionsof the element are both partially overlain dorsally

BTP

B

FIGURE 22.—Prochilodus rubrotaematus, USNM 225419, hyoidarch: A, left side, lateral view (branchiostegal rays removed);B, left side, dorsal view. (Stippling represents cartilage.)

by the edentulous basihyal tooth plate (e.g., Le-porinus, Figure 23). Four branchiostegal rays arecommon to most characiforms, three attached tothe anterior ceratohyal and one to the posterior.The relative width of these elements is variablewithin the order, but the degree of overlap be-tween neighboring rays is usually not extensive.The median urohyal typically has a moderatelydeveloped laminar vertical process and horizontallateral wings.

The hyoid arches of curimatids differ fromthose just described only in the lateral expansionof the anterior portions of the basihyal and basi-hyal tooth plate, which results in a distinctlytriangular basihyal complex. The anterolateralexpansion of these elements is carried further inprochilodontids (Figure 22B) in which the ante-rior margin of the complex is apomorphouslygreatly developed transversely. This enlargementprogresses ontogenetically to result in a squat T-shaped basihyal complex in larger individuals.Several other alterations of the hyoid arches char-acterize the Prochilodontidae. Smaller individ-uals of the family have a distinctive ventrallynotched interhyal straddling a corresponding de-pression in the dorsal surface of the posteriorceratohyal (Roberts, 1973:219). Large specimensthat were examined have the interhyal notch lesspronounced, but there is often (always?) a distinctsesamoid cartilage in the ligament that joins the

NUMBER 378 25

interhyal to the posterior ceratohyal. Neither anequivalent interhyal form nor a comparable in-dependent ossification in the ligament have beenencountered in the characiform outgroups ex-amined. Prochilodontids also have a markedbroadening of the branchiostegal rays, particu-larly the most lateral element, and a unique formof urohyal with highly developed, laterally di-rected ventral wings. Once again these apomor-phous modifications are not found in curimatids,anostomids, and chilodontids, but are present,evidently as homoplasies, in some distichodontids(Nannocharax) and parodontids (Parodon).

The overall hyoid arch form of anostomids andchilodontids is unique within characiforms. Sin-gularly notable is the longitudinal foreshorteningand transverse expansion of the lateral portionsof the system (compare Figures 23 and 24 with22). Anostomids and chilodontids also share aposteroventrally slanting joint between the ante-rior and posterior ceratohyals. The degree ofobliqueness differs significantly within anostom-ids, but a distinctly nonvertical joint is by far themost common condition in both families. Mostother characiforms have a vertical articulationbetween these elements. However, an obliquearea of contact between these elements occurssporadically in some other characiform groups,

BTP

PC

FIGURE 23.—Leporinus megalepis, USNM 231541, hyoid arch:A, left side, lateral view (branchiostegal rays removed); B, leftside, dorsal view. (Stippling represents cartilage.)

VH

FIGURE 24.—Caenotropus maculosus, USNM 231545, hyoidarch: A, left side, lateral view (branchiostegal rays removed);B, left side, dorsal view. (Stippling represents cartilage.)

most notably in a subunit of the Distichodontidae(e.g., Paradistichodus), in which it was apparentlyachieved independently.

The thickening of the lateral elements in thehyoid arch, particularly the anterior and posteriorceratohyals, is most pronounced in the Chilodon-tidae. Another notable feature of the hyoid archof that family is the prominent horizontal ridgeextending from the posterior portion of the ante-rior ceratohyal onto the adjoining region of theposterior ceratohyal. Chilodontids also have athickened, complex interhyal with a distinct me-dial process that extends onto a depression on thedorsomedial surface of the posterior ceratohyal.This process serves as the point of attachment forthe ligament that joins the interhyal to the me-tapterygoid and quadrate. Such an attachment isa pronounced shift from the usual insertion of theligament on the dorsal portion of the main bodyof the interhyal and is not known in the othercharaciform groups examined. This medial proc-ess of chilodontids is somewhat similar to themedial portion of the ventrally notched prochil-odontid interhyal. However, the Hgamentous at-tachments of the interhyal in prochilodontids areof the type generalized for characiforms ratherthan comparable to those of chilodontids.

SUSPENSORIUM AND ClRCUMORBITAL SERIES

Many of the derived characters in the suspen-sorium of curimatids, prochilodontids, anostom-ids, and chilodontids are associated with trophic


specializations. Functionally, the majority ofthese modifications are related either to differingdegrees of mobility between the suspensoriumand neurocranium or to relative motion of sub-units of the suspensory apparatus. Two distinctivesuspensorium modifications, one in curimatidsand the other in anostomids involve the relation-ship of the suspensorium and the neurocraniumand serve to further reduce mobility betweenthose systems.

Characiforms typically have a single, antero-dorsal articular surface on the palatine. This car-tilagenous portion of the bone contacts both thevomerine region of the neurocranium anterodor-sally and the cartilage associated with the upperarm of the maxilla laterally (the submaxillarycartilage of Daget (1964, fig. 23) and ethmopa-latine cartilage of Fink and Fink (1981:311)). Thepalatine in curimatids is apomorphously ex-panded posterodorsally to form a second, carti-lage-topped articular surface. That cap of carti-lage is continuous ventrolaterally with the rela-tively thick cartilaginous mass that overlies thelateral portions of the metapterygoid and ectop-terygoid (Figure 25). A corresponding cartilage-capped articular process formed by an anteroven-

METH

MET

MX

FIGURE 25.—Potamorhina laticeps, USNM 121325, upper jawand anterior portion of suspensorium and neurocranium, leftside, lateral view (denser stippling represents cartilage).

tral expansion of the distal margin of the lateralethmoid's ventral wing contacts the palatine'sposterodorsal process (Figure 25). Although de-rived for characiforms, a close articulation of thelateral ethmoid and palatine, per se, is not uniqueto curimatids within the order. Various forms ofcontact between those elements are also found inthe Old World hepsetid characiform Hepsetus andin the Neotropical characid genera Salminus, Aces-trorhynchns, Rhaphiodon, Hydrolycus, and Crenuchus.The Hepsetus lateral ethmoid-palatine joint differsfrom that in curimatids in having two distinctcontact points between the bones. Furthermore,the lateral ethmoid's anterior process serves as thepoint of articulation rather than the bone's ven-tral wing as in curimatids. The section of thepalatine that contacts the lateral ethmoid in Sal-minus is a posterior extension of the originallyanterodorsal articular surface of the bone, a situ-ation also not comparable to that in curimatids.The common possession of various derived char-acters (e.g., presence of an ossified rhinosphenoid)indicates that Acestrorhynchus, Rhaphiodon, and Hy-drolycus are most closely related to taxa that lackany lateral ethmoid-palatine contact. Therefore,it is most parsimonious to consider the presenceof the articulation in those noncurimatid generaas homoplasious relative to the condition in theCurimatidae; a hypothesis also congruent withthe distribution of shared derived characters thatunite the Curimatidae and its close relatives (see"Phylogenetic Reconstruction").

The form of the lateral ethmoid-palatine con-tact in curimatids and the characid genus Crenu-chus does not differ notably. However, Crenuchushas a median orbital ossification which may rep-resent an autapomorphous form of rhinosphen-oid. If correct, that homology would support thehypothesis that Crenuchus is more closely relatedto characiforms with a rhinosphenoid than tothose, such as curimatids, that lack it. Rhino-sphenoid-bearing taxa generally lack the second-ary palatinejoint. Therefore, if the median orbitalossification of Crenuchus is homologous to the rhi-nosphenoid, the secondary palatine joint of thatgenus would be presumed homoplasious relativeto the articulation in curimatids. In the absence

NUMBER 378 27

of a homology between the median orbital ossi-fication of Crenuchus and the rhinosphenoid, itcan, nonetheless, be noted that Crenuchus lacks thenumerous synapomorphies shared by curimatidsand their close relatives. That situation is incon-gruent with the union of curimatids and Crenuchusin a monophyletic assemblage as would be sug-gested by their common possession of indistin-guishable forms of lateral ethmoid-palatine joints.Under either alternative the lateral ethmoid-pal-atine articulations in curimatids and Crenuchus(and perhaps its close relatives) are most parsi-moniously considered independent acquisitions.

In anostomids and chilodontids the contactbetween the suspensorium and lateral ethmoid istightened by an alternate system. Chilodontidshave a discrete cord-like ligamentous band thatarises from the unelaborated dorsolateral surfaceof the ectopterygoid. That connective tissue bandextends posterodorsally lateral to the palatine andattaches to the anteroventral surface of the lateralethmoid's ventral wing. Anostomids have themodifications carried further (Figures 26, 38). In

METH

ECT

PAR

QUA

FIGURE 26.—Anostomus species, USNM 231540, lateral eth-moid and anterodorsal portion of suspensorium, left side,lateral view (denser stippling represents cartilage).

most members of that family a well-developedlateral process on the posterodorsal portion of theectopterygoid serves as the attachment point fora cord-like ligament both thicker and strongerthan that in chilodontids. As in chilodontids, theligament passes lateral to the palatine, but ratherthan terminating on the anteroventral margin ofthe lateral ethmoid's ventral process, it attachesin a pit-like depression on the anteroventral por-tion of the bone.

A broad connective tissue sheet between theneurocranium, particularly the parasphenoid,and the medial and dorsal margins of at least themetapterygoid and mesopterygoid portions of thesuspensorium is ubiquitous among characiforms.In some members of the order, portions of theconnective tissue sheet are thickened into one ormore cord- or strap-like ligamentous bands thatjoin the medial margins of the suspensorium tothe vomer or adjoining areas (e.g., prochilodon-tids and some curimatids). However, examinationof characiform outgroups has not revealed taxaother than anostomids and chilodontids with athick ligament that extends between the suspen-sorium's lateral surface and the lateral ethmoid.The presence of some form of discrete lateralethmoid-ectopterygoid ligament is, therefore,considered a shared derived character for chilo-dontids and anostomids. The posterolateral proc-ess on the ectopterygoid, which serves as an at-tachment point of the ligament, and the thick-ening of that connective tissue band are furtherelaborations of the system synapomorphous forthe members of the Anostomidae.

No traces of an ectopterygoid process or theassociated ligamentous band are found in theanostomid Gnathodolus, contrary to the broad dis-tribution of those characters in anostomids. InGnathodolus, the elongate, horizontally aligned ec-topterygoid is very distant from the lateral eth-moid and is flexibly attached to the dorsal marginof the quadrate (Winterbottom, 1980, figs. 41,58). The primary function of the lateral ethmoid-ectopterygoid ligament in other anostomids ap-pears to be in reducing motion between the neu-rocranium and suspensorium. The wide separa-tion of the ectopterygoid from the lateral ethmoid


and its mobility relative to the quadrate woulddrastically diminish, if not eliminate, the stabiliz-ing action of the ligament in Gnathodolus. Theabsence of the connective tissue band and of theassociated ectopterygoid modifications is conse-quently not unexpected from a functional view-point. As noted, the ligamentous attachment ispresent in chilodontids, the sister group of theAnostomidae. Furthermore, the other anostomidsexamined, perhaps including Synaptolaemus thesister genus to the lineage containing Gnathodolus(Winterbottom, 1980), have the lateral elabora-tion of the ectopterygoid and an ectopterygoid-lateral ethmoid ligament. (A definite statementon the presence or absence of the ligament inSynaptolaemus cannot be made based on availablematerial, but the dorsolateral process on the ec-toptyergoid is present, although not prominent.)The lack of the ligament in Gnathodolus (andperhaps its sister genus Sartor) is most parsimoni-ously considered a secondary loss rather than theprimitive absence of the band given the presenceof the ligament in sister groups to Gnathodolus ofincreasing universality.

The preceding discussion has dealt with mod-ifications that involve the functional relationshipof the suspensorium with the neurocranium. Anumber of other characters are found in subunitsof the suspensorium. Curimatids have the leastmodified form of suspensorium (Figure 27)among the four families that are the primaryfocus of this study. Nonetheless, they have severalderived alterations of the system, including thesecondary articular facet on the palatine dis-cussed above. Curimatids are also distinguishedby a horizontal shelf along the dorsomedial sur-face of the metapterygoid and sometimes meso-pterygoid (Figure 27B). This evidently apomor-phous elaboration of the metapterygoid does notoccur in the family's close relatives. The some-what similar horizontal shelf on the medial sur-face of the metapterygoid described by Winter-bottom (1980:50) in various anostomids is formedby the ventral bending of the dorsal margin ofthe metapterygoid rather than via an elaborationof the medial surface of that element.

Roberts (1974:429) used the presence of astrong, posteroventrally sloping flange located on

BMET

QUA

SYMPRE

OP

SOP

INT

FIGURE 27.—Potamorhina latior, AMNH 48677: A, suspensorium, left side, lateral view; B, detailof region of metapterygoid-quadrate fenestra, left side, medial view.

NUMBER 378 29

OP

A

QUA

SYMSOP

PREFIGURE 28.—Ichthyoelephas species, USNM 231437: A, suspensorium, left side, lateral view;

B, quadrate, left side, ventral view.

the opercle's lateral surface just above the hy-omandibula-opercle joint in the Curimatidae asa distinguishing feature of that family (Figure27). In an earlier paper Roberts (1973:213) re-jected previous hypotheses of a close relationshipof curimatids and prochilodontids, citing in sup-port of his argument the absence of an opercularflange in prochilodontids, among other differ-ences. Such an absence would, in and of itself, beactually irrelevant to the question of the interfa-milial relationships since, if unique to curimatids,the flange would be a synapomorphy for themembers of that family with no bearing on thequestion of curimatid-prochilodontid relation-ships. That point is moot, however, since prochil-odontid juveniles have an opercular flange iden-tical to that in curimatids. The prochilodontidflange undergoes a unique ontogenetic modifica-tion that leads to the condition Roberts equatedwith the plesiomorphous, laterally unelaboratedopercle. Juvenile prochilodontids of ~ 20 to 40mm SL have a posteroventrally slanting, dorsallyundercut opercular flange. With increasing bodysize, the members of the family develop a series

of bony ridges that extend posterodorsally atapproximately right angles from the main bodyof the process. Although similar secondary pro-cesses also occur in larger adult curimatids, inprochilodontids the processes' number and degreeof development increases ontogenetically, leadingto a progressive masking of the originally discreteflange. The flange is still apparent in — 80 mmSL specimens but indistinguishable in individualsof over 100 mm SL in which the ridges havecoalesced into a gently sloping, thickened area onthe opercle's lateral surface. In large specimensthe thickening is not apparent in lateral view(Figure 28A), but it is readily visible when thebone is examined in transmitted light. The oper-cular modifications of curimatids and prochilo-dontids presumably serve to strengthen the op-ercle in these microphagous fishes that use asuctorial feeding mode that requires the move-ment of relatively large water volumes throughthe oral chamber. These characters are derived attwo levels of universality. The presence of anopercular flange or a further derived condition ofthat process is synapomorphous for the Curima-


tidae and the Prochilodontidae, while the furtherelaboration of the flange is a shared derivedcharacter for the members of the Prochilodonti-dae. A distinct opercular flange also characterizesthe African characiform family Citharindae(Vari, 1979:295), which also consists of micropha-gous filter feeders. However, the flange in thattaxon differs in several features from the processin curimatids and juvenile prochilodontids. Fur-thermore, citharinids share numerous, oftenunique, synapomorphies with, and are evidentlymost closely related to, distichodontids, whichlack a lateral elaboration of the opercle. Thus,the opercular flange of the African family isconsidered homoplasious with respect to that inthe Curimatidae and Prochilodontidae.

The prochilodontid "suctorial" mouth isunique among characiforms with its functionalmodifications reflected in the numerous changesin the form and relationships of the elements ofthe suspensorium's anterior portion. Roberts(1974) has previously discussed the osteology ofprochilodontids, and it is only necessary to reviewthe adaptations of phylogenetic interest. In lateralview the prochilodontid ectopterygoid is tripar-tite, with a definite anteroventral process (Figure28A) that contrasts with the relatively straightmargin of the ectopterygoid in most characiforms.Equally distinctive is the high mobility of theectopterygoid that results from its loose ligamen-tous attachments to the quadrate and mesopter-ygoid. Ectopterygoid-quadrate mobility is alsopresent in hemiodontids and parodontids. How-ever, the relationships of the involved elements inthose taxa differ from the condition in prochilo-dontids, and neither family has the ectopterygoid-mesopterygoid mobility characteristic of prochil-odontids. Roberts (1973:218) also reported ecto-pterygoid-quadrate mobility in the Anostomidae.The anostomids on which that observation wasbased were not specifically noted, although Schi-zodon fasciatus and an unidentified Leporinus speciesare figured or cited in the text. However, theectopterygoid and quadrate are tightly, immov-ably joined in the Schizodon and Leporinus materialexamined in the present study. Indeed, the onlyanostomid studied with a mobile ectopterygoid-

quadrate joint is the highly specialized Gnathodo-lus, in which the ectopterygoid is shifted awayfrom the anterior margin of the remaining ele-ments of the pterygoid series. Reference to thereconstruction of anostomin phylogeny arrived atby Winterbottom (1980) shows that Gnathodolus isa member of a lineage whose sister species, Syn-aptolaemus cingulatus, lacks ectopterygoid-quadratemobility. That concept of relationships in con-junction with the immobile contact of these ele-ments in other anostomids results in the mostparsimonious hypothesis: Gnatholodus achieved ec-topterygoid-quadrate mobility independently ofprochilodontids.

Pronounced quadrate-preopercle mobility isanother character unique to prochilodontids inthe Characiformes. The prochilodontid quadrateis laterally expanded into a prominent horizontalshelf that serves as an expanded attachment areafor portions of the adductor mandibulae muscles.Medially the quadrate has a comparable, thoughless pronounced process that extends along theinner surface of the preopercle (Figure 28B). Thehorizontal notch delimited by the medial andlateral quadrate processes fits into the correspond-ing vertical notch bordered by dorsal and ventralsubdivisions of the preopercle's anterior portion(Figure 28). This distinctive joint permits signifi-cant vertical motion of the quadrate on the pre-opercle. A prominent lateral quadrate shelf alsooccurs among characiforms in chilodontids andanostomids. The phylogenetic hypotheses of re-lationships arrived at in this study indicate, how-ever, that these three families do not form amonophyletic group (see "Phylogenetic Recon-struction") but that, rather, prochilodontids aremore closely related to curimatids than to theunit that consists of anostomids and chilodontids.Within that hypothesis of relationships there existtwo equally parisomonious explanations for thedistribution of this derived character. First, thatthe quadrate shelf arose in the common ancestorof the four families and was secondarily elimi-nated in curimatids. Second, that the quadrateshelf was independently acquired in the ancestorof the Prochilodontidae and in the commonancestor of the Anostomidae and Chilodontidae.

NUMBER 378 31

Anostomids have further expanded the attach-ment area for the adductor musculature via alateral shelf on the preopercle's ventral arm. Thatprocess forms part of the floor of the bony troughwhose lateral wall is the posterolateral process ofthe quadrate. The preopercular shelf continuesposteriorly beyond the terminus of the quadratein the form of a gradually diminishing ridge.Although not unique to anostomids, a well-de-veloped lateral preopercular shelf is consideredderived within characiforms. A comparable pre-opercular elaboration also occurs in a subunit ofthe Distichodontidae (Vari, 1979:293). The dis-tichodontid subunit that has the preopercularexpansion is evidently most closely related toother members of that family, which lack a lat-erally elaborated preopercle. The Distichodonti-dae, in turn, shares a variety of derived characterswith the Citharinidae in which the preopercularshelf is also absent. Thus, the presence of thepreopercular shelf in some distichodontids is evi-dently an independent acquisition relative to theprocess in anostomids.

Anostomids and prochilodontids may sharetwo other derived suspensorium characters. Sev-eral questions exist, however, on the homology of

the modifications in the two taxa. A preoperclewith an enclosed laterosensory canal that extendsas a continuous ossification to the level of thequadrate's articular facet is typical for characi-forms (e.g., the curimatid, Potamorhina latior, Fig-ure 27). The Anostomidae and Prochilodontidae,in contradistinction, have the anterior portion ofthe preopercular laterosensory canal enclosed bytwo or three discrete ossified tubes aligned alongthe ventral surface of the quadrate (Figures 28,29). These separate ossifications were termed thesubpreopercles by Roberts (1973:218), whereasWinterbottom (1980:37) considered them to bepart of the preopercle. The homology of theseossified tubes in the two families is questionable.Even if homologous, the overall most parsimoni-ous hypothesis of relationships that includes thetwo families indicates that the separate ossifica-tions in anostomids and prochilodontids representindependently achieved adaptations to functionalrequirements that can only be speculated upon.At least for the Prochilodontidae, Winterbottom'sterminology is preferred, because it is consideredmore reflective of the probable homology of theelements. The autogenous ossifications in thatfamily occur in the region plesiomorphously oc-

SPO

OP

SOP

PRE INT

FIGURE 29.—Leporinus fasciatus, USNM 103847: A, suspensorium, left side, lateral view;B, quadrate, left side, ventral view.


cupied by the anterior portions of the preopercle(Figure 28). An anterior subdivision of the preo-percle in prochilodontids is necessary for verticalmotion of the quadrate on the preopercle. Suchmobility would be impossible or greatly restrictedif prochilodontids retained the primitive rigidpreopercle that extends along the anteroventralsurface of the quadrate. A hypothesis that theseparate ossifications arose via a fragmentation ofthe attenuate anterior portions of the preopercleis, thus, the simplest. The alternate assumption,inherent in Robert's terminology, that the tubesare nonhomologous with portions of the preoper-cle, would necessitate the loss of the anteriormostsection of the preopercle ventral to the quadratefollowed by the development of two or three new,autogenous elements.

The quadrate-preopercle joint in anostomids isimmobile. Thus, the anterior subdivision of thepreopercle in that family cannot be ascribed tothe same underlying functional explanation ad-vanced for prochilodontids, moreso because thephylogenetic reconstruction indicates that themodifications are homoplasious. The anostomidsuspensorium (Figure 29) has undergone a pro-nounced anterior elongation with a consequentwide separation of the anterior terminus of thequadrate and, thereby, the lower jaw from themain body of the preopercle. Under such circum-stances the maintenance of continuity betweenthe laterosensory system of the preopercle andlower jaw necessitates a pronounced lengtheningof the preopercular laterosensory canal along theventral margin of the quadrate. If that extensiontook place via a progressive ossification of theanterior portion of the preopercular sensorycanal, then the extremely attenuate ossified tubethat results might subsequently subdivide. Alter-natively, the laterosensory system might extendforward as an unossified canal, which is subse-quently partially strengthened by autogenous os-sifications that developed anterior to the preoper-cle per se. Data relevant to a selection betweenthese alternative possibilities is not available atthis time. The most parsimonious hypothesis ofrelationships for anostomids, prochilodontids,and their close relatives indicates that under

either scenario the autogenous preopercularlaterosensory canal ossifications are homoplasiesin prochilodontids and anostomids.

The highly modified genus Gnathodolus differsfrom other anostomids in its lack of the separateanterior preopercular ossifications (Winterbot-tom, 1980, fig. 41). The absence of the ossifica-tions in that genus is most parsimoniously consid-ered a secondary loss in light of the phylogeneticdistribution of the ossifications and our presentinformation on relationships within the Anostom-inae (Winterbottom, 1980).

Most anostomids have the horizontal portionof the quadrate posteriorly bifurcate, as amongprochilodontids (Figures 28B, 29B). However, inanostomids the relationships of the quadrate'smedial process with the preopercle differ from theprochilodontid condition, thereby, casting doubton the homology of the modifications in the twofamilies. The quadrate's medial process in pro-chilodontids runs internal to the preopercle withthe longitudinal midline of the resulting quadratenotch aligned with the anteroposterior axis of theventral portion of the preopercle. In conjunctionwith other previously discussed suspensoriummodifications, this restructuring of the quadratepermits vertical mobility of that bone on thepreopercle. In anostomids, in contrast, the distinctposterior notch in the preopercle, when present,is largely a consequence of the posterior expansionof the bone's lateral portion rather than the com-mon posterior development of its medial andlateral processes. The midline of the notch definedby the medial and lateral processes of the anos-tomid quadrate's posterior region is not alignedwith the longitudinal axis of the preopercle but,rather, lateral to it. Consequently, the medialquadrate process does not extend internal to thepreopercle, but, rather, often inserts into a notchin that bone to form an immobile joint. In lightof these differences, the equivalence of the poste-rior quadrate bifurcation in prochilodontids andanostomids is questionable, moreso since the mostparsimonious hypothesis of relationships of theinvolved families indicates that these are homo-plasies.

A final suspensorium character of note involves

NUMBER 378 33

the hyomandibula in anostomids and chilodon-tids. In both families, that bone has a variouslydeveloped process that extends over the postero-dorsal portion of the metapterygoid. The posses-sion of such a process is evidently derived withincharaciforms and is not known in prochilodontidsand curimatids. It is found, evidently as an in-dependent acquisition, in some serrasalmines (Co-lossoma), parodontids (Parodon, Saccodon), hemio-dontids (Bivibranchia), and characids (Acestrorhyn-chus).

Gregory and Conrad (1938), followed by Gery(1961), apparently misinterpreted the identity ofcertain elements in the anostomid suspensorium.Comparisons of the illustration of the osteologyof an unspecified Leporinus species in Gregory andConrad's figure 28 with Figure 29 of this studyshows their metapterygoid is actually a portionof the hyomandibula. The entopterygoid (= me-sopterygoid) of those authors should be moreproperly identified as the metapterygoid. Theunlabelled ectopterygoid and palatine are illus-trated so as to appear as portions of the quadrateand lateral ethmoid, respectively. Gery's moredetailed illustrations of Leporinus' head osteology(1961, figs. 12, 14) have similar problems. Theelement labelled as the palatine in his drawing ofL. friderici is actually a composite of that elementand the mesopterygoid. This error results in theremaining elements in the series being sequen-tially misidentified. His entopterygoid (= meso-pterygoid) should be identified as the metapter-ygoid. The element labelled by Gery as the me-tapterygoid is actually the anteroventral sectionof the hyomandibula. Gery illustrates a jointbetween a posterodorsal, triangular hyomandi-bula, and a vertically oriented metapterygoid.That line of demarcation actually represents thelocation of a distinct ridge along the lateral sur-face of the hyomandibula, rather than a point ofcontact between two bones. The actual hyoman-dibula of Leporinus is equivalent to the hyoman-dibula plus metapterygoid of Gery's illustration.As a consequence of these misinterpretations,Gery's osteological description of various portionsof the suspensorium (1961:104) is misleading,particularly the comments on the relative size of

the hyomandibula in anostomids.The synapomorphous form of the second in-

fraorbital of prochilodontids is the only phyloge-netically significant character within the cir-cumorbital series at the family level of the four-family group. All prochilodontids examined havethe ventral margin of the bone anteroventrallyexpanded into a distinct triangular process. Theanterior border of that process, the ventral marginof the enlarged first infraorbital and the antero-ventral edge of the second infraorbital delimit atriangular notch that borders the posterior mar-gin of the large fleshy lips (Roberts, 1973:229).No similar alteration of the second infraorbitalwas found in the characiform outgroups exam-ined in the present study.

Gregory and Conrad (1938:348) reported thata supraorbital is absent in the anostomid genusLeporinus, an erroroneous observation correctedby Gery (1961:103). Gery's illustrations of thecircumorbital series and the dermal ossificationsanterior to the orbit (1960, fig. 2; 1961, figs. 12,15) utilized osteological terminology inconsistentwith that of previous and subsequent authors.The ossification he labels as the lachrymal is thebone that Weitzman (1962:28) termed theantorbital. No other author dealing with chara-ciforms has identified the antorbital as the lach-rymal and such a homology appears to be erro-neous. The lachrymal of teleosts has traditionallybeen equated with the anteriormost (first) infraor-bital, whereas Gery (1960, fig. 2; 1961, figs. 12,15) labelled the anteriormost infraorbital as thejugal in his figures. The use of jugal for a subunitof the infraorbital series in ostariophysan fishesapparently follows Gregory and Conrad (1938).Those authors, however, used jugal and infraor-bital 2 interchangeably, whereas Gery appliedthe term to the first infraorbital.

PECTORAL GIRDLE

The common name for chilodontids, headstan-ders, refers to their oblique, head-down swimmingorientation (Gery, 1977b:212-213). This highlyunusual swimming position and the resultantatypical forces that act on the body perhaps


-POST

EXT

PCL-2

PCL-3

scCOR

FIGURE 30.—Caenotropus maculosus, USNM 231545, pectoralgirdle, left side, lateral view (radials and pectoral fin raysremoved).

account for most of the numerous unique pectoralgirdle modifications in the family. Examinationof the overall form of the girdle shows that theproportions of the major elements differ markedlyfrom those encountered in other characiforms.Unlike many characiforms, chilodontids have acleithrum with a distinct posterior developmentof the lamina above the pectoral fin base (Figure30). More notable is the ventral expansion of thesupracleithrum's main shaft over the cleithrum'slateral surface. Consequently, the supraclei-thrum's length is well over half of the girdle'stotal vertical span. The increased overlap of thecleithrum on the supracleithrum increases theinternal rigidity of the pectoral girdle, with theflexibility of that complex on the neurocraniumreduced by alterations of the supracleithrum,posttemporal, and pterotic. With few exceptions,characiforms have the pectoral girdle attached tothe neurocranium via a relatively complex post-temporal. In the typical condition the posttem-poral's main body is triangular and gives rise to

two primary processes (Weitzman, 1962, fig. 19).The dorsal process of the posttemporal extendsdorsomedially along the posterior edge of theparietal, to which it is tightly attached. A secondsmaller, strut-like process extends from the ven-tromedial surface of the posttemporal to contactthe rear of the neurocranium. The laterosensorycanal segment in the post temporal's ventral por-tion communicates anteriorly with the sensorycanal of the extrascapular and posteroventrallywith that in the dorsal portion of the supracleith-rum. The typical characiform extrascapular istriangular with a tripartite laterosensory systemthat contacts the canals of the pterotic anteriorly,posttemporal posteriorly and parietal dorsally.

The chilodontid posttemporal is greatly alteredrelative to the hypothesized plesiomorphous mor-phology described above. The bone is reducedventrally, with the ventromedial posttemporalprocess and the laterosensory canal segment ab-sent. The dorsal, relatively unaltered, portion ofthe bone is elongate, tapers dorsally and attachesfirmly to the posterior margin of the parietal(Figure 31). The reduction of the posttemporal's

PTESPH

SAP SCL

FIGURE 31.—Caenotropus maculosus, USNM 231545, dorsalportion of pectoral girdle and posteroventral portion ofneurocranium, left side, lateral view.

NUMBER 378 35

laterosensory canal segment and the supraclei-thrum's dorsal expansion result in direct contactof the extrascapular and supracleithral laterosen-sory canal segments. A reduced posttemporal sen-sory canal typically occurs in various smallercharaciforms that have an overall reduction inhead ossification. In those taxa, however, thisreduction results in a gap in the laterosensorycanal system rather than direct communicationbetween the canals of the supracleithrum andextrascapular as in chilodontids. The plesiomor-phously present ventromedial posttemporal proc-ess is also absent in some parodontids and invarious less well-ossified characiforms assigned toa number of different families. The informationof this study and other data indicates that it ismost parsimonious to consider the chilodontidposttemporal reduction as nonhomologous withthose in the outgroups. The chilodontid extra-scapular is less elaborate than that common tomost characiforms. The anterior laterosensorycanal is absent as a distinct process, being repre-sented only by an anterior opening midway alongthe margin of the remaining ossified tube.

The marked reduction in the relative size andcomplexity of the chilodontid posttemporal elim-inates that element as the primary junction of thepectoral girdle and neurocranium. Rather, inchilodontids the primary dorsal articulation be-tween the pectoral girdle and the neurocraniumis achieved via the insertion of the dorsal tip ofthe supracleithrum into a distinct pocket in theposteroventral corner of the pterotic (Figure 31).The expanded dorsal portion of the supraclei-thrum is overlain dorsolaterally by the reducedextrascapular and is bordered dorsally by theremaining dorsomedial portion of the posttem-poral. No comparable direct pterotic-supraclei-thral contact has been found in the other chara-ciforms examined, nor has a similar associationof the supracleithrum with the posttemporal andextrascapular been encountered in the order.

Chilodontids also have a second area of pter-otic-supracleithral articulation. The posteroven-tral corner of the characiform pterotic typicallybears a distinct process that ranges in form froma short spur to a distinct spine that approaches,

or even contacts, the medial surface of the supra-cleithrum. Chilodus and Caenotropus have that pter-otic process posterolaterally expanded into anobliquely aligned, flattened plate. That plate con-tacts the medial surface of, and corresponds inform to, an associated anterior process of thesupracleithrum (Figure 31). This articulation, to-gether with the above-described dorsal joint be-tween the pterotic and dorsal tip of the supra-cleithrum, renders the pectoral girdle immobilerelative to the neurocranium. Although a contactof the supracleithrum with the posteroventralprocess of the pterotic occurs in various characi-forms, such as articulation via a distinct distallywidened posteroventral pterotic process has beenencountered in only one group of nonchilodontidcharaciforms. The African family Citharinidaehas the posteroventral portion of the pteroticflared outwards into a slightly bi-pronged platethat contacts the medial surface of the supra-cleithrum's dorsal portion. Comparison of thesefunctional complexes in chilodontids and cithar-inids reveals that the pterotic expansion in theCitharinidae differs in overall form from the dis-tally ovoid supporting process of the Chilodonti-dae. The citharinid supracleithrum also lacks thewell-developed anterior process that articulateswith the pterotic in chilodontids. Citharinids,furthermore, do not possess the other distinctivepectoral girdle modifications of chilodontids andthe Citharinidae shares uniquely derived char-acters with, and is evidently most closely relatedto, the Old World Distichodontidae (Vari,1979:324). These factors support a hypothesis ofthe independent acquisition of the more ventralof the pterotic-supracleithral contacts in the Chil-odontidae and Citharinidae.

Three postcleithra along the posterior marginof the pectoral girdle are typical for characiforms.The dorsalmost (postcleithrum 1) is a flattened,round ossification located at the posterior marginof the contact area of the supracleithrum andcleithrum. Postcleithrum 2 is usually a flattened,ovoid bone at the rear of the cleithrum's posteriorlamina. The straight or slightly curved, rod-likepostcleithrum 3 attaches to the medial surface ofpostcleithrum 2 and extends ventral of the latter


element. Postcleithra of this type are common tomost curimatids, prochilodontids and anostom-ids. Chilodontids, however, have that systemmodified both at the familial and intrafamiliallevels. In all members of that family, postclei-thrum 1 is lacking (Figure 30). Although theabsence of the element is considered derived, itslack is not unique to this family within the Char-aciformes. No postcleithra are present in the gas-teropelecins of the family Characidae (Weitzman,1954:226). Postcleithrum 1 is also missing in theAfrican hepsetid characoid Hepsetus (Roberts,1969:426), the distichodontids Nannocharax andHemigrammocharax (Vari, 1979:311) and the NewWorld anostomid genera Synaptolaemus and Sartor(Winterbottom, 1980:46) among others. Nonethe-less, the results of this study and our presentunderstanding of phylogenetic relationships inthe order indicate that the losses in those taxa arehomoplasious relative to the absence of the ele-ment in chilodontids. Within the Chilodontidae,two different conditions of the third postclei-thrum exist. The bone is altered into a highlycurved, anteromedially shifted ossification inCaenotropus (Figure 30), but is totally lacking inChilodus. The utility of these postcleithral adap-tations in phylogenetic reconstruction is reducedby the existence of two equally parsimoniousexplanations for the distribution of these charac-ters. First, the curved third postcleithrum actuallyrepresents a synapomorphy for all chilodontids,with the bone having been subsequently lost isChilodus. Second, the form of the bone in Caenotro-pus and the absence of the ossification in Chiloduseach represent synapomorphies for their respec-tive genera independently modified from the gen-eralized characiform type of third postcleithrum.Present data do not permit a decision as to whichof these explanations is preferable.

The numerous pectoral girdle modifications inthe Chilodontidae contrast with the relativelyinvariant forms of that system, at least at thefamily level, in the Curimatidae, Prochilodonti-dae, and Anostomidae. The single exception in-volves the anostomid extrascapula. The extra-scapular laterosensory system of characiforms istypically tripartite with a single posterior aper-

ture that communicates with the sensory canal ofthe supracleithrum. The anostomid extrascapularhas an additional posterodorsal sensory canal seg-ment opening onto the posterior or posterodorsalmargin of the bone. The resultant quadripartitesystem is not encountered in the other familiesunder discussion. Gregory and Conrad (1938, fig.28) mistakenly illustrate the extrascapular (theirscalebone) of Leporinus as carrying a tripartitelaterosensory canal; an error repeated by Gery(1961, figs. 12, 16) in his discussion of Leporinus

friderici. However, all specimens of L. friderici andother Leporinus species examined have the quad-ripartite extrascapular laterosensory canal systemherein considered typical for anostomids.

NEUROCRANIUM

Pronounced differences in aspects of neuro-cranial morphology occur in each of these fami-lies. Many, including subtle differences or contin-uous variation in the form or proportions of var-ious bones, are difficult to quantify and/or polar-ize. The characters noted in the following discus-sion, consequently, tend to be discrete apomor-phies, many of which were previously noted inassociation with changes in other body systems,with which they are functionally associated. Suchpreviously discussed characters are only cited inpassing in this section.

The mesethmoid is very broad in curimatids,with the large associated cartilagenous ethmoidblock particularly distinctive. That cartilage massfills the space between the mesethmoid dorsallyand the vomer ventrally and has posteriorly di-rected lateral processes that contact the corre-sponding anterior projections of the lateral eth-moid. This cartilage mass undergoes a progressivelaminar ossification of its exposed surfaces, butthe major portion remains as an unossified blockeven in adults. Such a large cartilagenous eth-moid block has also been found only in theAfrican Citharinidae among characiforms exam-ined in the present study. The similarities in theethmoid region morphology may be a reflectionof their common microphagous feeding habits,but these similarities are homoplasious under our

NUMBER 378 37

present concepts of characiform phylogeny. Theseconcepts indicate that the Citharinidae and Cur-imatidae are not each others closest relatives (see"Phylogenetic Reconstruction" and Vari, 1979).

The mesethmoid in the Chilodontidae, in con-tradistinction to that in curimatids, is laterallycompressed and anteroventrally developed into adistinctly angled process that extends betweenthe medial margins of the premaxillae (Figure32). The ventral expansion of the mesethmoidresults in the total separation of the premaxillaeand, more interestingly, in the mesethmoid con-tributing to an appreciable portion of the marginof the upper jaw (sec "Teeth and Jaws"). Asnoted on page 10, this ethmoid region morphol-ogy is evidently synapomorphous for the membersof the Chilodontidae.

The Curimatidae has reduced the flexibility ofthe suspensorium on the neurocranium via adistinct direct articulation between the lateralethmoid and palatine (Figure 25). The typicalcharaciform lateral ethmoid has a discrete narrowedge along the entire lower margin of its ventralwing. In curimatids, there is instead a distinctlongitudinal expansion of the ventral midsectionof the wing into a cartilage-capped articular facet,which contacts a similar palatine process. Asdiscussed in the description of the suspensorium,this is evidently a synapomorphy for all curima-tids.

Associated with the overall changes of the pec-

toral girdle in the Chilodontidae are two adap-tations of the pterotic, which serve to reduce themobility of the pectoral girdle relative to theneurocranium. These were discussed at length inthe section on the pectoral girdle (p. 34).

Variation exists within the Characiformes bothin the number of posttemporal fossae and in thebones that border some of the apertures. The vastmajority of characiforms are characterized by adorsal and posterodorsal pair of fossae on eitherside of the neurocranium. Curimatids possessthose apertures and an additional third smallround fossa entirely within the epioccipital. Thistype of third posttemporal fossa also occurs in theHemiodontidae (Roberts, 1974:416, fig. 5) andParodontidae (Roberts, 1974:425, fig. 59). A moreextensive, vertically ovate third posttemporalfossa bordered by both the epioccipital and ex-occipital is found in the Old World characiformfamilies Citharinidae and Distichodontidae(Vari, 1979, fig. 15) arid the Neotropical characidtribe Cynodontini (Vari, 1979:289).

The presence of an epioccipital or epioccipital-exoccipital posttemporal fossa is apparently de-rived within the Ostariophysi with the epioccipi-tal portions of each aperture presumably homol-ogous. Comparable apertures have not been re-ported or discovered in characiforms outside ofthese groups or in other otophysans. The posses-sion of such openings would, thus, appear to beapomorphous within the order. The available

METH

METH

FIGURE 32.—Caenotropus maculosus, USNM 231545: A, anterior portion of neurocranium, left side,lateral view (large stippling represents cartilage); B, anterior portion of neurocranium andupper jaw, dorsal view.


information does not, however, permit a polaritydetermination on the presumed transition seriesbetween the epioccipital limited and epioccipital-exoccipital bordered forms of third posttemporalfossae.

The exact phylogenetic relationships of the twoother characiform families with the curimatidform of epioccipital fossa have not been rigorouslyexamined. Hemiodontids (sensu Roberts, 1974)have a median orbital ossification, the rhino-sphenoid, unique to a subunit of the Characi-formes among fishes. The possession of that ele-ment is, thus, considered a synapomorphy unitingthe components of that assemblage. That char-acter and others unite the Hemiodontidae mostclosely to characiforms other than curimatids,which lack the bone. The phylogenetic relation-ships of the Parodontidae are enigmatic. Roberts(1974:429) noted various characters common tothe Parodontidae and Hemiodontidae. However,he felt that traditional concepts of a close rela-tionship of the two families were not well corra-borated. In the absence of a tested hypothesis ofthe relationships of parodontids to a group ofnoncurimatid characiforms, I can only note thatthe series of characters that unite curimatids withprochilodontids, anostomids, and chilodontids,which lack the opening, would indicate that thecommon possession of a epioccipital posttemporalfossa in curimatids and parodontids is homopla-sious.

Curimatids and prochilodontids share a dis-tinctive modification of the basic characiformbauplan of the neurocranium's posterior region.The most widespread and hypothesized plesio-morphous characiform condition of the portionsof the exoccipitals proximal to the foramen mag-num is illustrated in Figures 33A and 34. Theexoccipitals are complex, paired elements in con-tact medially. Each exoccipital articulates ven-trally with the median basioccipital and along itsdorsal border contacts the unpaired supraoccipi-tal. The dorsal portions of the exoccipitals meetmedially along a vertical joint that extends ven-trally from the articulation of the exoccipitalswith the supraoccipital to the foramen magnum.Ventral to that area of midline contact, the pos-

FIGURE 33.—Scaphium and rear of neurocranium, left side,lateral view: A, Leporinus reinhardti, AMNH 4O1O4SD; B,Psectrogaster amazonica, AMNH 40088SD. (Scale = 5.0 mm.)

teromedial margins of the exoccipitals divergelaterally to form a median arch. The resultantopening encompasses the lateral borders of theforamen magnum dorsally, and the cavum sinusimparis ventrally. A horizontal sheet of boneextends from the medial surface of the main bodyof each exoccipital to meet its counterpart at themidline. The shelf formed by those processesserves both as the floor of the forman mangumand roof of the cavum sinus imparis. In lateralview (Figure 33A), the posterior margin of theexoccipital proximal to the foramen magnum isdistinctly concave. Viewed from the rear (Figure34) the region of concavity is seen as a bony pillarbounded by the lateral occipital foramen andforamen magnum, with an indentation, the fossafor the scaphium, present on its posterior surface.The concave medial lip of this scaphial fossacorresponds in shape to, and closely approxi-mates, the anterior margin of the first Weberianossicle, the scaphium, to which it is ligamentouslyconnected. In the majority of characiforms out-

NUMBER 378 39

LOC

CSI

FlCURE 34.-

BOC-Leporinus reinhardti, AMNH 40104SD, region of foramen

magnum, posterior view.

side of curimatids and prochilodontids, the fossaranges from a negligible depression in some an-ostomids (Figure 34) to a distinct cul-de-sac incharacids such as Brycon (Weitzman, 1962, fig. 5).A foramen, the lateral occipital fenestra, is typi-cally found along the midlateral surface of theexoccipital. That aperture's degree of develop-ment differs within the order, ranging from abarely apparent foramen to a large opening oc-cupying most of the surface of the exoccipitallateral to the foramen magnum. The lateral oc-cipital fenestra typically has a nearly verticalorientation and serves, at least in part, as a pas-sage for nerves.

Curimatids and prochilodontids share a dis-tinctive restructuring of the portion of the exoc-cipital bounded by the lateral occipital foramenand the foramen magnum. In both families theplesiomorphous cul-de-sac fossa for the scaphiumis greatly expanded laterally and anteriorly (Fig-ure 35). As a consequence the slightly or moder-ately developed scaphial fossa common to mostcharaciforms is altered in curimatids and pro-chilodontids into a large foramen that is contin-uous anterolaterally with the lateral occipital for-

amen and communicates anteriorly with the neu-rocranium's interior. This expanded fossa for thescaphium subdivides the primitively single verti-cal pillar bounded by the foramen magnum andlateral occipital foramen into two parts (Figure35). The medial portion that borders the foramenmagnum and cavum sinus imparis is deeply in-dented to accommodate the anterior margin ofthe scaphium to which it is ligamentously at-tached. A marked outwards shift of the primi-tively lateral edge of the fossa results from theexpansion of that aperture. Thus, the associatedportion of the exoccipital's lateral surface is nowmore obliquely aligned than in the hypothesizedprimitive condition. With this change in inclina-tion the opening of the lateral occipital fossa facesmore posterolaterally in curimatids and prochil-odontids than in other characiforms.

The rear margin of the exoccipital lateral tothe enlarged fossa for the scaphium is also poste-riorly expanded. This expansion results in a singleinclusive opening leading to the foramen mag-num, cavum sinus imparis, and the paired sca-phial fossae. The foramen magnum and sinuscavum imparis apertures are consequently in-


LOC

BOCFIGURE 35.—Psectrogaster amazomca, AMNH 40088SD, region of foramen

magnum, posterior view.

eluded in the neurocranium rather than locatedalong its posterior margin as in the primitivecondition. The posterior extension of the lateralmargin of the exoccipital also results in the pos-terior portion of that element overlapping thescaphium's anterior section to the level of theattachment point of the interosseus ligament (Fig-ure 33B).

Examination of a wide variety of other chara-ciforms and Ostariophysans has revealed only asingle group, the Gasteropelecinae, with compa-rable exoccipital modifications (Weitzman,1954:218, figs. 5, 10). The gasteropelecin exoccip-ital, as that in curimatids and prochilodontids,has the deep foramen form of fossa for the sca-phium. This results in the inclusion of the fora-men magnum, cavum sinus imparis, and pairedscaphial fossae in a large common opening andthe anterior overlap of the scaphium by the ex-occipital's margin. The only significant differencebetween the gasteropelecin exoccipital and thatcommon to curimatids and prochilodontids oc-curs in the position of the lateral occipital fora-

men. In gasteropelecins the foramen's lateralopening is shifted ventrally to a deep fossa barelyposterodorsal to the lagenar capsule. With thisrepositioning, the foramen's opening is now di-rected ventrolaterally, whereas a lateral orienta-tion is common to the majority of characiforms,and a slightly dorsolateral alignment is shared bycurimatids and prochilodontids. The ventral po-sition and ventrolateral orientation of the fora-men in gasteropelecins, contrary to the moredorsal location and slightly dorsolateral align-ment of the aperture in prochilodontids and cur-imatids, is not, in itself, evidence of the nonho-mology of the other shared exoccipital modifica-tions. The variation in lateral occipital foramenposition might rather represent different apomor-phous adaptations of a shared derived ancestralcondition common to curimatids, prochilodon-tids, and gasteropelecins.

A hypothesis of the homoplasy of the exoccip-ital alterations must, therefore, depend on a phy-logenetic reconstruction of greater generalitywithin characiforms. The results of this study

NUMBER 378 41

indicate that the phylogenetic relationships ofcurimatids and prochilodontids lie with characi-forms other than gasteropelecins (see "Phylo-genetic Reconstruction"). The question of thepossible relatives of gasteropelecines within char-aciforms has not been critically analyzed, usingmore recently formalized systematic procedures.Nonetheless, Weitzman's conclusion (1954:231)that "the Gasteropelecinae probably arose fromsome generalized characid somewhat like As-tyanax, Brycon, and Bryconamericus" is consistent

with our present knowledge of character polaritiesand phylogenetic relationships within the Char-aciformes. Although we are presently unable torigorously test a hypothesis of the probable hom-oplasy of the exoccipital modifications of gaster-opelecins on the one hand and those in the assem-blage consisting of curimatids and prochilodon-tids on the other, the available data, nonetheless,indicates that these seemingly equivalent char-acters represent convergencies.

A final neurocranial character of note is thesize of the lagenar capsule of curimatids. In theGurimatidae, the capsule is significantly ex-panded into a large bulbous chamber. The apo-morphous enlargement is notable both relative tothe condition in the other three families underconsideration and the various characiform out-groups examined.

VERTEBRAL COLUMN AND RIBS

A variety of derived characters that involve theanteriormost pleural ribs and the associated ver-tebrae occur in anostomids and chilodontids. Oneof the more distinctive characters common tothese families are the two or more discrete, well-developed intercostal ligaments that join the mid-sections of three or more of the anteriormost fullpleural ribs. These intercostal ligaments presum-ably are thickened sections of the broad connec-tive tissue sheet that typically joins the medialmargins of the pleural ribs. The intercostal liga-ments in chilodontids are thicker than those inanostomids and span a maximum of three ribs.The ventral ligament in chilodontids arises fromthe posterior margin of the first full pleural rib,

extends obliquely posterodorsally to attach to themedial surface of the second rib, and terminatesposteriorly on the anterior border of the third rib(Figure 36). A second, shorter ligament, dorsaland parallel to the first, extends between the firstand second ribs. Anostomids have a greater num-ber of ribs incorporated into the system, althoughthe individual intercostal ligaments are not asdeveloped as in chilodontids. The ventral liga-ment in anostomids commences on the posteriormargin of the first full pleural rib, attaches to themedial margins of the second and third ribs, andterminates posterodorsally on the anterior borderof the fourth rib. A second, more dorsal, ligamentextends between the first and third ribs, with anattachment to the inner surface of the second rib.The described pattern is the simplest within thefamily, with the number of ribs and intercostalligaments in the system further increased in someanostomids. The ligaments' exact function is un-

UG

PR

FIGURE 36.—Chiloduspunctatus, USNM 231542, fifth to eighthvertebrae, dorsal portions of anteriormost full pleural ribsand intercostal ligaments, left side, lateral view.


certain, but the interconnected anterior pleuralrib complex in these fishes may be associated withtheir oblique, head-down swimming position.

The oblique body orientation during locomo-tion is most pronounced in the Chilodontidae(Gery, 1977b:212-213), which have the thickestintercostal ligaments. The first three full pleuralribs spanned by these heavy bands in chilodontidshave their proximal portions and the parapo-physes, neural spines, and arches of the associatedvertebrae extensively restructured. Among char-aciforms the typical moderately developed para-pophyses on the first three full pleural ribs insertinto circular or ovoid articular fossae limited tothe respective centrum's lateral surface (Weitz-man, 1962, fig. 12). However, the chilodontidsCaenotropus and Chilodus, have the proximal por-tions of the first three full pleural ribs expandeddorsally, which results in the proximal portionsof the ribs having a triangular form with maxi-mum vertical development in the region adjoin-ing the parapophyses (Figure 37). The parapo-physes are also expanded vertically relative to theossification in other characiforms. The dorsally

NS

PR

PR

CENFIGURE 37.—Chilodus punctatus, USNM 231542, sixth verte-bra, parapophyses, and dorsal portion of pleural ribs, ante-

rior view.

elongate medial surface of the parapophyses ar-ticulates with a vertically elongate fossa. Thatarticular surface incorporates ventrally the ple-siomorphous small fossa on the lateral face of thecentrum and a vertically elongate flange on thelateral surface of the neural arch and spine (Fig-ure 37). Such lateral elaborations of the neuralarch and spine are unknown elsewhere in theorder. This degree of dorsal development of thecomponents of this complex is greatest on thesecond full pleural rib and associated portions ofthe sixth vertebrae. This development is also ap-parent, though not as pronounced, on the firstand third full pleural ribs. The consolidation thatresults from these adaptations eliminates the mo-tion of the dorsal portions of the ribs relative tothe vertebral column, particularly given the moredistal interconnections of these ribs via the afore-mentioned intercostal ligaments.

MYOLOGY

Considerable variation exists in the morphol-ogy of the jaws, suspensorium, gill arches, and theparts of the neurocranium associated with thosesystems among curimatids, prochilodontids, an-ostomids, and chilodontids. Congruent with theosteological changes are a series of adaptations ofthe associated musculature.

Two myological apomorphies have alreadybeen described in the discussions of the gill archesand epibranchial organs. The first is the distinc-tive vertical muscle band that extends along theanterior and posterior surfaces of the chilodontidepibranchial organ's dorsal portions (see "Epi-branchial Organs"). The other is the greatlythickened obliquues dorsalis associated with thehighly modified fourth gill arch in the Anostom-idae and Chilodontidae (see "Gill Arches").

The form of the adductor mandibulae is hy-pothesized to be synapomorphous for the four-family unit. In his discussion of the cheek mus-culature in Leporinus, Alexander (1964:183-184)noted the presence of a discrete medial portion ofthe adductor mandibulae, which he suggestedwas the A3 section of that muscle. That homologyis reasonable within the context of our present

NUMBER 378 43

AAP DO

LIG

PMX

LAP

HYO

FIGURE 38.—Leporinus striatus, USNM 231948, cheek and suspensoriummusculature, left side, lateral view.

understanding of characiform myology. A dis-tinct A3 portion of the adductor mandibulae thatarises from the lateral surface of at least themesopterygoid and metapterygoid is common tocurimatids, chilodontids, prochilodontids, andanostomids, with further pronounced apomor-phous changes unique to each of the latter twofamilies (Figures 38, 39). In its simplest condition,as m curimatids and chilodontids, the A3 isformed by a series of anterolaterally slanting fi-bers that have an origin on the dorsolateral sur-face of the mesopterygoid and metapterygoid.These muscle slips attach laterally to the tendi-nous band located along the medial surface of theA2 portion of the adductor mandibulae. No com-parable form of A3 has been encountered in thebroad, but by no means exhaustive, outgroupcomparisons carried out among other characi-form taxa, nor has it been reported in the litera-ture. The A3 reported in citharinids and disticho-dontids (Vari, 1979:316) differs in several fea-tures. That muscle arises from the anteromedialsurface of the hyomandibula and extends forward

as a discrete muscle band margined medially bya tendinous sheet, which coalesces anteriorly withthe tendon of A2. A comparable situation existsin serrasalmins. Thus, the A3 of these taxa differsin origin and mode of insertion relative to thecondition of the A3 form in the Curimatidae,Prochilodontidae, Anostomidae, and Chilodonti-dae. Consequently the characteristic morphologyof the A3 section of the adductor mandibulae ishypothesized to be a synapomorphy for the latterfour-family clade.

Two sets of apomorphous modifications of thecheek musculature are characteristic for the spe-cies of anostomids on one hand and those ofprochilodontids on the other. One character inthe Anostomidae is particularly noteworthy. Theanostomids available for myological study(Abramites, Anostomus, Leporellus, Leporinus, Rhy-

tiodus, and Schizodon) have a triangular muscle inthe posterodorsal portion of the orbital chamber(Figure 38). The V-shaped ventral portion of themuscle is capped by an aponeurotic sheet thatattaches onto a discrete process on the anterodor-


AAPDO

PMX

MX

LAP

DEN

FIGURE 39.—Prochilodus rubrotaeniatus, USNM 225419, cheek and suspensoriummusculature, left side, lateral view.

sal margin of the hyomandibula. The muscle fansout dorsally with an origin on the anteroventralface of the sphenotic and orbital surface of thefrontal. The phylogenetic derivation of the mus-cle is uncertain, but the levator arcus palatini isthe most likely source. That homology hypothesisis advanced primarily on the evidently commonfunction of the two muscles and their generalproximity. Furthermore, in some characiformsthe medial portion of the levator arcus palatinioverlaps the anterior margin of the hyomandi-bula; the medial portion also has a limited medialorigin along the anteroventral face of the sphen-otic spine comparable to, though not as extensiveas, that of the discrete triangular muscle in an-ostomids. Outgroup comparisons have failed toreveal another characiform group with a similardiscrete muscle.

The adductor mandibulae of anostomids isunusually altered. The Ai section of the muscle issubdivided into two portions. The first section islocated along the anteroventral portion of thecheek and extends anterodorsally to attach di-

rectly to the maxilla (Figure 38). A second sectionof the Ai, medial to the above, arises from theentire ventral surface of the trough formed by thelateral shelf of the quadrate and preopercle. Com-mencing posteriorly at the terminus of the quad-rate, the muscle extends anteriorly with a tendi-nous band along its dorsomedial margin. Thatconnective tissue sheet progressively thickens an-teriorly to form a distinct tendon, which attachesto the maxilla. The A2 portion of the adductormandibulae is significantly more highly devel-oped transversely in anostomids (Figure 38) thanin chilodontids, curimatids, and prochilodontids(Figure 39). In those families, the posterior por-tion of the A2 section of the adductor mandibulaelies lateral to the levator arcus palatini with anorigin solely from the angle of the preopercle.Such a form of the muscle is also common tonumerous generalized characiforms and evidentlyprimitive in the order. In the Anostomidae, bycontrast, the posterior portion of A2 has an addi-tional medial section that extends internally tothe ventral portion of the levator arcus palatini

NUMBER 378 45

and has an apomorphous partial origin from adistinctly concave region of the anteroventralportion of the hyomandibula.

The junction between the posteriorly separatedlateral and medial sections of A2 in anostomids isdemarcated by a tendinous band, which progres-sively thickens anteriorly to form a distinct ten-don that extends forward to an insertion on thelower jaw. A considerable portion of the medianmargin of that tendon serves as the insertion sitefor the A3 portion of the adductor mandibulae.That muscle section is greatly developed relativeto the condition in curimatids, prochilodontids,and chilodontids, with this expansion particularlypronounced in the region of the ventral wing ofthe lateral ethmoid (Figure 38). This dorsallybulging muscle section attaches to the medial anddorsal surfaces of the ligament and joins ante-riorly with the medial surface of the dorsallyexpanded anterior portion of A2. The dorsal con-tact between these two muscle sections results inthe common tendon on the A2 and A3 runninginternal to the muscle masses that attach to thelower jaw, rather than along their dorsal margin.Such a morphology of the musculature is consid-ered derived in light of the absence of comparableadaptations in outgroups examined in the presentstudy. The Aw portion of the adductor mandi-bulae, present in most characiforms, was notfound as a separate muscle in anostomids; anapomorphous absence previously noted by Alex-ander (1964:183).

Not unexpectedly, the highly unusual suctorialmouth of prochilodontids is reflected in the myol-ogy of the cheek region (Figure 39). The mostdistinctive character of the prochilodontid adduc-tor mandibulae is the great expansion posteriorlyof the muscle's Ai section. As a consequence theAi arises from across the entire horizontal arm ofthe quadrate and preopercle with a partial originon the ventral portion of its vertical arm. Ante-riorly the muscle has a broad direct insertion onthe angulo-articular's posterior margin and aninsertion on the maxilla via the primordial liga-ment. The posterior expansion of the Ai hasexcluded the A2 from most of its plesiomorphous

HI

FIGURE 40.—Prochilodus rubrotaeniatus, U S N M 225419, mus-culature of the ventral surface of the head, ventral view(branchiostegal rays of right side removed).

area of origin along the horizontal arm of thepreopercle and ventral section of the vertical limbof that element. The A2 is consequently reducedto a relatively small, horizontally aligned bandwith a limited origin on the vertical arm of thepreopercle in the region ventral to the insertionof the levator arcus palatini. The muscle sectionbecomes increasingly attenuate anteriorly, givingrise to a tendinous band. The relatively large A3attaches to the medial margin of that tendon,which then passes forward to the lower jaw. Thisreversal of the usual proportions of the Ai and A2sections of the adductor mandibulae is distinctivefor prochilodontids among characiforms exam-ined.

A myological synapomorphy for the four-fam-ily clade occurs in the musculature of the ventralsurface of the head. Characiforms typically havea hyohyoidei abductores with a relatively broadinsertion on the dorsal surface of the branchio-stegal rays. The anterior portion of the muscleprogressively tapers anteriorly, with a discreteorigin from the anteroventral margin of the uro-

46

hyal and the anteromedial surface of the hyoidarch. The posteroventral portion of the urohyalis covered by the sternohyoideus, which is readilyvisible in ventral view and only partially over-lapped, primarily anteriorly, by the hyohyoideiadductores. In the Curimatidae, Prochilodonti-dae, Anostomidae, and Chilodontidae, in con-trast, the hyohyoidei abductores has a broadinsertion across the entire ventral surface andlateral margins of the ventrolateral wings of theurohyal, contrary to the narrow attachment tothe anteriormost portion of the bone found inother characiforms (Figure 40). This broad at-tachment results in both the near total ventraloverlap of the sternohyoideus by the hyohyoideiabductores and drastic restriction of the apertureof the gill slits. A reduction in the gill slit aperturealso occurs in parodontids and some distichodon-tids and characids. In none of the examinedoutgroups is the constriction of the gill openingsarrived at in a manner comparable to that inthese four families. The mode in which the con-striction is achieved in the Curimatidae, Prochil-odontidae, Anostomidae, and Chilodontidae isconsequently considered a shared derived char-acter for the assemblage.

Phylogenetic Reconstruction

Synapomorphies within the osteological andmyological systems described in the previous sec-tion provide information relevant to a hypothesisof familial and suprafamilial relationships amongthe Curimatidae, Prochilodontidae, Anostomi-dae, and Chilodontidae. These same systems alsodemonstrate synapomorphies for the members ofeach family. The following discussion first detailsthe synapomorphies for the four-family unit, fol-lowed by those derived characters that distinguishclades of decreasing universality within that as-semblage. Subsequent to the reconstruction of themost parsimonious hypothesis of phylogenetic re-lationships among these families, there is a dis-cussion of those derived characters that have aphylogenetic distribution incongruent with thearrived at hypothesis of relationships. The discus-sion deals both with homoplasies internal to the

SMITHSONIAN CONTRIBUTIONS T O ZOOLOGY

CURIMATIDAE RIOCHIIJOOONTIDAE ANOSTOMIDAE CMLOOONTIOAE

74 99

-58

FIGURE 41.—Cladogram of the most parsimonious hypothe-sis of relationships of the families Curimatidae, Prochilodon-tidae, Anostomidae, and Chilodontidae (numbered synapo-morphies correspond to those of the text).

four-family assemblage and those that involve asubunit of that lineage and a characiform out-group. In conjunction with the phylogenetic re-construction, that discussion provides the basisfor the evaluation of previous classificatoryschemes as reflectors of the phylogenetic historyof the four-family assemblage and its subunits.

Figure 41 presents the most parsimonious phy-logenetic hypothesis that incorporates the previ-ously discussed synapomorphies. The apomor-phous characters defining the various suprafam-ilial assemblages and families are numbered se-quentially within clades. That procedure simpli-fies the visualization of character distribution andfamilial relationships. The numbering of thecharacters in the following text corresponds to thenumbered synapomorphies of the cladogram inFigure 41.

THE FOUR-FAMILY ASSEMBLAGE

The hypothesized monophyly of the assem-blage formed by the families Curimatidae, Pro-chilodontidae, Anostomidae, and Chilodontidaeis supported by the following synapomorphies forthe four-family unit.

1. The elimination of the direct contact between

NUMBER 378 47

the fourth and fifth upper pharyngeal toothplates.

2. The absence of dentition of the fourth upperpharyngeal tooth plate.

3. The A3 portion of the adductor mandibulaewith an extensive origin along the lateral sur-face of the mesopterygoid and metapterygoidand a broad insertion on the tendon of A2.

4. The longitudinally expanded attachment ofthe hyohyoidei abductores on the urohyal.

Although this hypothesis of relationships is theleast robust in the phytogeny when evaluated interms of the number of synapomorphies, it is notchallenged by any equally parsimonious alterna-tive hypothesis derivable from the characters ex-amined. The convergencies shared by the varioussubunits of this assemblage with diverse characi-form outgroups typically involve a single familyand an outgroup rather than a monophyletic twofamily unit and the outgroup. Thus, hypothesesof the homoplasious nature of those charactersare supported both by the synapomorphies forthe four-taxon unit and the shared derived char-acters that unite each pair of family level sistergroups. The homoplasious nature of the occur-rence of such characters in and outside of thefour-taxon unit under consideration is also indi-cated by the available data on the phylogeneticplacement of the characiform outgroups thatshare the convergent characters. A supramaxilla,for example, is evidently limited to the Chilodon-tidae and the characid Chalceus among characi-forms and is considered an independently ac-quired apomorphy in each of the two lineages.That hypothesis is in agreement with a parsimonyevaluation, based both on the synapomorphiesfor the four-family unit and the more numerousshared derived characters that unite chilodontidswith anostomids that lack a supramaxilla. Addi-tional support for a hypothesis of the homopla-sious distribution of the ossification comes fromthe various shared derived characters (e.g., com-mon possession of a rhinosphenoid) that joinChalceus to the Characidae, in which a supramax-illa otherwise is absent. Considering the presenceof a supramaxilla in Chalceus and the Chilodon-

tidae as a convergence is, thus, the simplest ex-planation. Comparable parsimony arguments ap-ply to other characters homoplasiously presentwithin the four-family assemblage and in one ormore characiform outgroups. The overall distri-bution of such homoplasious characters does notform any other pattern consistent with an alter-nate concept of relationships as, or more paris-monious than, that advanced herein.

CURIMATIDAE AND PROCHILODONTIDAE C L A D E

The less inclusive familial and suprafamilialclades among the Curimatidae, Prochilodontidae,Anostomidae, and Chilodontidae are defined bynumerous advanced characters. Synapomorphiesof the Curimatidae and Prochilodontidae pri-marily are associated with the food gathering andmanipulating systems, with a lesser number beingexoccipital modifications. In summary theseshared derived characters are as follows.

5. The reorientation of the dorsal process of thefourth epibranchial anteriorly with its con-sequent extension over the dorsal surface ofthe fourth infrapharyngobranchial.

6. The anterodorsal expansion of the cartilagi-nous fifth epibranchial, its attachment to theposterodorsal margin of the fourth epibran-chial, and the resultant encirclement of thefifth efferent branchial artery.

7. The large sac-like, muscular epibranchial or-gan that extends dorsal to the medial ele-ments of the dorsal portion of the gill arches.

8. The conversion of the plesiomorphously flatfourth upper pharyngeal tooth plate into acurved ossification wrapped around thefourth infrapharyngobranchial.

9. The reduction or loss of dentition on the fifthupper pharyngeal tooth plate.

10. The reduction or loss of an ossified first bas-ibranchial.

11. The absence of dentition on the fifth cerato-branchial.

12. The anteromedially directed process on thefourth ceratobranchial's ventral surface.

13. The distinct, posteroventrally aligned flange


on the lateral surface of the opercle or afurther derived condition of that process.

14. The increase in the depth and width of thefossa for the scaphium that results in inter-connections of the fossa with the lateral oc-cipital foramen laterally and the interior ofthe cranium anteriorly.

15. The posterior development of the lateralmargin of the exoccipital lateral to the fora-men magnum, thereby forming a commonaperture for the foramen magnum, cavumsinus imparis, and paired fossae for the sca-phium and a cover laterally for the antero-lateral surface of the scaphium.

The phylogenetic hypothesis based on theabove characters (i.e., the close relationships ofthe Curimatidae and Prochilodontidae) is inagreement, in whole or in part, with some previ-ous classificatory schemes, but noteably differentfrom those of Boulenger (1904) and Roberts(1973) (see "Comparisons with Previous Classifi-cations"). Examination of the characters that tra-ditionally distinguish these families shows thatthey are largely external synapomorphies for themembers of the Prochilodontidae, particularlythose in mouth form, rather than any discreteexternal features of curimatids. Indeed, only thefirst listed shared derived character for the mem-bers of the latter family (see below), the toothlessjaws, is apparent in external examination. Twosubunits of the assemblage characterized by char-acters 5 to 15 are, in turn, definable by a series ofsynapomorphies unique to the members of eachfamily.

FAMILY CURIMATIDAE

The Curimatidae is the least derived taxonamong the four families under consideration interms of the overall number of synapomorphiesfound during the present study. The shared de-rived characters congruent with the hypothesisthat the Curimatidae forms a monophyletic lin-eage are the following.

16. The lack of dentary and premaxillary denti-tion.

17. The third posttemporal fossa bordered solelyby the epioccipital.

18. The anteroventral process on the third hy-pobranchial that extends parallel to the ven-tral aorta.

19. The absence of an ossified first basibranchial.20. The posteroventral expansion of the fifth up-

per pharyngeal tooth plate into a curved,convoluted process.

21. The secondary posterodorsal, cartilage-capped process on the palatine.

22. The cartilage-capped articular process alongthe ventral edge of the lateral ethmoid's ven-tral wing.

23. The horizontal shelf on medial surface of themetapterygoid.

24. The large cartilagenous ethmoid block andwidened mesethmoid.

25. The enlarged lagenar capsule.

The absence of dentary and premaxillary den-tition is the only externally apparent discretesynapomorphy for the members of the Curima-tidae. Most members of the family, nonetheless,share a distinctive bauplan that permits theiridentification at the family level without the ne-cessity for examination of the jaws. Some curi-matids, however, differ dramatically from thetypical curimatid body plan and are not readilyrecognized as members of the family. Curimataocellata is distinctly fusiform and can be rathereasily confused with some hemiodontids, which itclosely parallels in body form and pigmentationpattern (Gery, 1977b, fig. 3). Alternatively, somecurimatid species are extremely deep-bodied withlarge individuals of Curimata abramoides (in whichthe body depth exceeds 50% of standard length)representing the extreme in that direction (Kner,1859, pi. 2).

The intrafamilial taxonomy of the Curimati-dae is highly confused, with many of the nominalgenera consisting of evidently unnatural group-ings of species. The number of recognizable formsamong the described species is uncertain. Studiesto date (e.g., Vari, 1982b) have shown that abouttwenty of the ten dozen presently recognizedspecies are invalid, but also that numerous pre-viously unrecognized distinct species exist.

NUMBER 378 49

The taxonomic limits of the Curimatidae ofthis study differ from those utilized by the vastmajority of prior researchers. Traditionally, curi-matids and Anodus have been associated on thebasis of their edentulous jaws. Some authors orig-inally described various curimatid species in An-odus (e.g., Anodus latior Spix, in Spix and Agassiz,1829) or used Anodus as a genus in curimatids(e.g., Eigenmann and Allen, 1942:300). Otherresearchers (e.g., Eigenmann and Eigenmann,1889; Fernandez-Yepez, 1948) have reserved An-odus for a limited number of fusiform speciesknown from the Orinoco and Amazon river ba-sins. Those authors, nonetheless, either associatedthe genus in some unspecified manner with curi-matids (Eigenmann and Eigenmann, 1889) orexplicitly placed it in the Curimatidae, as thesubfamily that formed the sister lineage to curi-matids (sensu stricto) (Fernandez-Yepez, 1948:19). However, the phylogenetic affiliations of An-odus appear to lie with the Hemiodontidae ratherthan the Curimatidae (sensu stricto), a hypothesisfirst advanced by Roberts (1974) and confirmedby this study. That larger assemblage, in turn,shows various synapomorphies with some, butnot all characids, to which it is apparently moreclosely related.

FAMILY PROCHILODONTIDAE

The family Prochilodontidae is a very distinc-tive assemblage having the following numerousshared derived characters.

26. The increased number of functional and re-placement tooth rows and teeth per row.

27. The highly developed fleshy lips that form asuctorial oral disk.

28. The bulbous form of the premaxilla.29. The bulbous form of the maxilla, the numer-

ous maxillary fenestrae, and the posterome-dial bony maxillary process that serves forthe attachment of the maxillo-mandibularligament.

30. The foreshortened dentary and the ex-panded, laterally rotated dentary replace-ment tooth trench with a large medial fenes-tra.

31. The reduction in the overall relative size ofthe retroarticular and its shift onto the me-dial surface of the angulo-articular.

32. The transverse compression of the fifth upperpharyngeal tooth plate and the mobility ofthat element with respect to the fourth epi-branchial.

33. The transverse expansion of the basihyal'santerior portion.

34. The ventrally notched interhyal and thesesamoid ossification in the ligament thatjoins the interhyal to the posterior ceratohyal.

35. The transverse broadening of the branchio-stegal rays.

36. The transverse widening of the urohyal's ven-tral wings.

37. The ontogenetic expansion of the opercularflange into a broad, flat, thickened region onthe opercle's lateral surface.

38. The tripartite ectopterygoid that is mobilerelative to the quadrate and preopercle.

39. The posteriorly notched quadrate that is ver-tically mobile on the preopercle and the lat-eral shelf on the quadrate.

40. The subdivision of the anterior portion of thepreopercular laterosensory canal into two ormore ossified tubes.

41. The form of the second infraorbital.42. The posterior extension of the Ai portion of

the adductor mandibulae.43. The reduction of the A2 portion of the ad-

ductor mandibulae.

Prochilodontids are most readily recognizableexternally by their enlarged fleshy lips, whichevert into a suctorial oral disk. That mouth formis unique to the family among characiforms andthe listed synapomorphies for the members of thefamily are all associated with oral apparatus mod-ifications. Three genera (Ichthyoclephas, Prochilodus,and Semaprochilodus) with some 30 to 40 nominalspecies, presently are recognized, although anunderstanding of the actual number of valid spe-cies must await a thorough revision of the family.Many of the species are economically importantand undertake large-scale, long-distance, spawn-ing-associated migrations. Roberts (1973) has de-scribed the osteological variation in the Prochil-


odontidae and provided a partial synopsis of theavailable life history information for the family.

ANOSTOMIDAE AND CHILODONTIDAE CLADE

The Anostomidae and Chilodontidae havebeen the subject of divergent opinions concerningtheir interrelationships and the appropriate waysto reflect those concepts taxonomically. Some ofthe adaptations common to the two families arequite unusual, particularly the various restructur-ings of the branchial apparatus. Those gill archmodifications and the synapomorphies in otherbody systems are as follows.

44. The longitudinal foreshortening of the man-dible.

45. The pronounced enlargement of the upperand lower pharyngeal dentition.

46. The presence of two or more cusps on allpharyngeal teeth.

47. The shift in the alignment of the fourth upperpharyngeal tooth plate that results in itscontact with the fourth epibranchial ratherthan the fifth upper pharyngeal tooth plate.

48. The vertical thickening of the fifth upperpharyngeal tooth plate.

49. The transverse expansion of the posteriorregion of the third infrapharyngobranchial.

50. The thickening and posterodorsal reorienta-tion of the dorsal process of the fourth epi-branchial.

51. The vertical expansion of the anterior processof the fourth epibranchial.

52. The highly developed obliquues dorsalis as-sociated with the fourth infrapharyngobran-chial.

53. The longitudinal foreshortening of the hyoidarch.

54. The oblique angle of the articulation be-tween the anterior and posterior ceratohyals.

55. The distinct cord-like ligament that joins thelateral surface of the ectopterygoid and ven-tral wing of the lateral ethmoid.

56. The prominent lateral shelf on the quadrate.57. The hyomandibular process that extends

over the metapterygoid's posterodorsal re-gion.

58. The presence of two or more intercostal lig-aments that join the three or more of theanteriormost full pleural ribs.

The two subunits of the assemblage defined bycharacters 44 to 58 are, in turn, distinguished bytheir less inclusive sets of shared derived charac-ters. These clades are (1) the morphologicallydiverse, speciose family Anostomidae; and (2) theChilodontidae, a small group both in terms ofgenera and species, that is, characterized by nu-merous synapomorphies and relatively little in-trafamilial variation.

FAMILY ANOSTOMIDAE

The family Anostomidae is largely character-ized by branchial apparatus and suspensoriummodifications that reflect the unique overall formof those systems in the family. The synapomor-phies for the species of anostomids are as follows.

59. The expansion of the ascending process ofthe premaxillary resulting in a heavy trian-gular bone.

60. The pronounced expansion of the dentaryreplacement tooth trench, and the distinctfenestra along the ventral portion of thetrench.

61. The enlargement of the dentary and premax-illary teeth.

62. The expansion of the third epibranchial'smedial portion to form a curved process thatextends medially over the fourth infraphar-yngobranchial's dorsal surface.

63. The very well-developed cord-like ligamentthat runs between the lateral surface of theectopterygoid and the ventral wing of thelateral ethmoid.

64. The distinct process on the posterolateralsurface of the ectopterygoid.

65. The expansion of the preopercle's lateral sur-face to form a shelf for the origin of theadductor mandibulae.

66. The enclosure of the anterior portion of thepreopercular laterosensory canal by two orthree autogenous ossified tubes.

67. The quadripartite laterosensory canal systemin the extrascapular.

NUMBER 378 51

68. The two or more intercostal ligaments thatunite four or more of the anteriormost fullpleural ribs.

69. The triangular, autogenous section of thelevator arcus palatini that arises from theposterodorsal portion of the orbital cavityand inserts on the anterodorsal margin of thehyomandibula.

70. The subdivision of the Ai portion of theadductor mandibulae.

71. The expansion of the posterior part of the A2portion of the adductor mandibulae medialof the levator arcus palatini and its partialorigin on the hyomandibula.

72. The dorsal expansion anteriorly of the A3and A2 portions of the adductor mandibulaethat results in the contact of these muscleswith each other dorsal to the tendon thatattaches to the main body of A2.

73. The absence of the Aw portion of the adduc-tor mandibulae.

Anostomids are a distinctive group of someten genera, a number of which are monotypic,but with Leporinus containing some 70 nominalspecies. Phylogenetic relationships within thesubfamily Anostominae, which is characterizedby some very unusual jaw adaptations, were an-alyzed by Winterbottom (1980), who also revisedthe contained species. Most of the remaininggenera, particularly Leporinus, are poorly under-stood phylogenetically and taxonomically.

FAMILY CHILODONTIDAE

The members of the Chilodontidae have incommon a profusion of synapomorphous modifi-cations not encountered elsewhere in the familiesanalyzed or often in any other fishes. Many ofthese characters are associated with the form ofthe epibranchial organ, which is not approxi-mated in outgroups reported on in the literatureor examined during this study. A second systemthat shows marked modifications is the pectoralgirdle; perhaps as a consequence of the unusualhead-down body orientation utilized by chilodon-tids while swimming. The various synapomor-phies for members of the family are as follows.

74. The reduction in the relative size of the pre-m axilla.

75. The expansion in the relative size and thick-ness of the maxilla.

76. The ossified supramaxilla located along theposterodorsal margin of the maxilla.

77. The anteroventral expansion of the meseth-moid into an angled process that extendsbetween the premaxillae.

78. The three or more cusps on all pharyngealteeth.

79. The posteroventral expansion of ceratobran-chial four into a broad, curved surface.

80. The expansion of the anterior portions ofceratobranchial five into a cup-shaped plate.

81. The rotation anteriorly of the tooth-bearingportion of ceratobranchial five.

82. The ridges on the dorsal surfaces of cerato-branchials 1, 2, and 3.

83. The reorientation posteriorly of the fifth up-per pharyngeal tooth plate.

84. The pronounced transverse expansion of thethird infrapharyngobranchial's posterior por-tion.

85. The ridges on the ventral surfaces of epibran-chials 1, 2, and 3.

86. The posterior rotation of the ventral portionof the fourth epibranchial.

87. The lateral and dorsal expansion of the car-tilaginous fifth epibranchial to form the ma-jor part of the anterior wall of the epibran-chial organ.

88. The shell-shaped connective tissue sheet thatextends dorsal to the fifth ceratobranchialand forms the posterior wall of the epibran-chial organ.

89. The distinct ridges on the soft tissue layerthat covers the adjoining surfaces of thefourth and fifth ceratobranchials.

90. The distinct band-like muscle on the anteriorand posterior surfaces of the epibranchialorgan.

91. The laterally thickened anterior and poste-rior ceratohyals with pronounced ridgesalong their lateral surfaces.

92. The complex, thickened interhyal with theligament that extends to the metapterygoid


and quadrate attaching to the discrete me-dial process of the interhyal.

93. The ventral expansion of the main shaft ofthe supracleithrum.

94. The elimination of the ventral process andlaterosensory canal bearing portion of theposttemporal and the anterior section of theextrascapular.

95. The direct articulation of the dorsal tip ofthe extrascapular with the pterotic.

96. The expansion of the posteroventral spine ofthe pterotic into a flat articular surface thatcontacts a corresponding supracleithralprocess.

97. The absence of postcleithrum 1.98. The dorsal expansion of the proximal por-

tions of the first three full pleural ribs.99. The vertical expansion of the parapophyses

and articular fossae associated with the firstthree full pleural ribs.

Very few of the shared derived characters com-mon to the members of the family are readilyvisible externally. Consequently, chilodontids donot appear as highly derived superficially as domembers of some of the other families in thesuprafamilial lineage under discussion. Thoughvery distinct anatomically, the group has notundergone any great intrafamilial differentiationor phyletic radiation. Two genera and three orfour species presently are recognized (Gery,1977b).

Discussion

The advanced hypothesis of relationships is themost parsimonious derivable, given the availableinformation on the polarity and distribution ofcharacters in the examined body systems of curi-matids, prochilodontids, anostomids, and chilo-dontids. The size and complexity of the Chara-ciformes and scarcity of specimens of various taxalimits the practical outgroup comparisons. Thus,the occurrence of characters may in some casesbe wider than noted here.

Two types of homoplasies were encountered:those internal to the four-family clade and con-vergencies between a subsection of that lineage

and one or more characiform outgroups. Themajority of these incongruities are loss characters,while others are sufficiently different in the var-ious groups that possess them to cause doubtsabout their homology.

Characters hypothesized to be homoplasiouslydistributed within the four-family unit are asfollows.

1. The distinct lateral shelf on the quadrate inthe Prochilodontidae and the unit that consistsof the Anostomidae and Chilodontidae.

2. The autogenous ossified tubes associated withthe anterior section of the preopercular latero-sensory canal system in the Prochilodontidaeand Anostomidae.

3. The absence of a dentary replacement toothtrench in the Chilodontidae and Curimatidae.

4. The longitudinal foreshortening of the lowerjaw in the Prochilodontidae and the unit thatconsists of the Chilodontidae and Anostomi-dae.

5. The reduced teeth not in direct contact withthe jaws in the Chilodontidae and Prochilo-dontidae.

Roberts (1973) also advanced the similaritiesin ethmoid (= mesethmoid) form and the enclo-sure of the angular (= retroarticular) by thedentary as indicators of close anostomid-prochil-odontid relationships. The homology of thesecharacters is difficult to evaluate. If they arehomologous apomorphies they would be consid-ered homoplasies under the results of this study.That decision and the comparable treatment ofthe five characters listed above are based onparsimony criteria. Including the two characterslisted by Roberts, we find that three characterssupport a hypothesis of a sister group relationshipbetween the Prochilodontidae and the Anostom-idae (the two characters proposed by Roberts andcharacter 2 above). Two characters (1 and 4above) are congruent with the concept of theProchilodontidae being most closely related tothe lineage that consists of the Anostomidae andthe Chilodontidae. One character (number 3above) unites the Chilodontidae and Curimati-dae. Finally, one character (number 5 above)

NUMBER 378 53

unites the Prochilodontidae and Chilodontidae.If we consider the tooth reduction of character 5to be an intermediate state to the total loss of jawdentition in the adults of the Curimatidae, areduction or loss of jaw teeth would rather be asynapomorphy for the assemblage formed by theProchilodontidae, Chilodontidae, and Curimati-dae.

The acceptance of even the most corraboratedof these phylogenetic alignments (that is, unitingthe Prochilodontidae and the Anostomidae)would require the hypothesis that all 11 charac-ters considered synapomorphous for the Curi-matidae and Prochilodontidae arose indepen-dently in those two lineages and that a similarsituation exists with respect to the 13 characterssynapomorphous for the Anostomidae and Chil-odontidae. The selection of a hypothesis of a sistergroup relationship between the Prochilodontidaeand Anostomidae would indeed eliminate thethree homoplasies noted above but would neces-sitate a hypothesis that 24 other characters aroseindependently; as such, it would be obviously lessparsimonious. Comparable arguments are appli-cable to the other characters considered homopla-sious in the present study.

Such criteria are also applicable to the questionof the absence of the first postcleithrum in chilo-dontids and a subunit of anostomids, and theabsence of an ossified first basibranchial in curi-matids and some anostomids. Several charactersconsidered synapomorphous for the members ofthe Anostomidae are evidently secondarily lost inGnathodolus and perhaps its sister genus Sartor,which was not available for study (see "Suspen-sorium and Circumorbital Series"). Once again,parsimony arguments based on the available dataon intra-anostomid relationships suggest that theabsence of those characters within the Anostom-idae is a secondary loss.

A variety of other characters were noted whichevidently occur homoplasiously in the group thatconsists of the Curimatidae, Prochilodontidae,Anostomidae, and Chilodontidae or subunits ofthat clade and various characiforms outgroups.These characiform outgroups include the Cith-arinidae, Parodontidae, and Hepsetidae, various

subunits of the Distichodontidae, the Hemiodon-tidae (particularly Bivibranchia and Anodus) andwithin the Characidae Acestrorhynchus, Chalceus,Cotossoma, Crenuchus, Hydrolycus, Rhaphiodon, andthe Gasteropelecinae. The characters involved,comments on their possible homology, and morespecific details on their phylogenetic distributionsare discussed in the preceeding "Character De-scription and Analysis". Neither the ingrouphomoplasies nor those with characiform out-groups show a repeated pattern of phylogeneticdistribution consistent with an alternative equallyor more parsimonious hypothesis of relationships.

The majority of the noted outgroups have onlyone shared derived character in common withsome subunit of the four-family assemblage ana-lyzed. Hypotheses of relationships based on suchcharacters are obviously less robust than thatbased on the more numerous characters listed inthe previous section. Only two taxa, the OldWorld characiform family Citharinidae and theNeotropical genus Anodus show any multiple, de-rived similarities with subunits of the clade con-sisting of the Curimatidae, Prochilodontidae, An-ostomidae, and Chilodontidae.

The loss of dentition on the fourth upper pha-ryngeal tooth plate is a characteristic of the Cith-arinidae and the entire four-family assemblage.The Citharinidae has a reduction or loss of lowerpharyngeal dentition and the presence of anopercular flange in common with the Curimati-dae plus Prochilodontidae. Both the Citharinidaeand the Curimatidae have a third posttemporalfossa and possess a well-developed cartilagenousethmoid block. An outwardly rotated dentaryreplacement tooth trench is shared by the Cith-arinidae and the Prochilodontidae. Finally, theCitharinidae and Chilodontidae have in commona secondary contact of the supracleithrum with aposteroventrally expanded process of the pterotic.Any hypothesis of a sister group relationship be-tween the Citharinidae and any family or multi-familial grouping just listed would require theassumption that the various hypothesized syna-pomorphies for the Citharinidae and Disticho-dontidae (Vari, 1979) are actually homoplasies.In the majority of cases it would also necessitate


similar assumptions relative to the numerous syn-apomorphies for the unit that consists of theCurimatidae, Prochilodontidae, Anostomidae,and Chilodontidae or subgroups of that clade.Thus, for example, the recognition of the Cith-arinidae and Curimatidae as sister groups on thebasis of their synapomorphous possession of athird postemporal fossa and a large cartilagenousethmoid block would necessitate the assumptionthat characters 1 through 15 are homoplasies.That procedure results in obviously less parsi-monious hypotheses.

The second taxon, the genus Anodus, shares theabsence of jaw teeth with the Curimatidae, theabsence of lower pharyngeal teeth with the Cur-imatidae and Prochilodontidae, and the absenceof teeth on the fourth upper pharyngeal toothplate with the Curimatidae, Prochilodontidae,Anostomidae, and Chilodontidae. The hypothesisof a sister group relationship between Anodus andthe four-family assemblage is incongruent withthe various synapomorphies that unite the genusand the Hemiodontidae (Roberts, 1974). Anyhypothesis of a close relationship between Anodusand subunits of the four-family clade is in addi-tion incongruent with the various synapomor-phies for and within that quadri-familial assem-blage.

Uncertainties about the homology of the teethon the prochilodontid suctorial mouth make itimpossible to unambiguously evaluate the properphylogenetic level at which it was appropriate touse the absence of maxillary teeth as a characterin the phylogenetic reconstruction. The absenceof teeth on that bone may be either a synapo-morphy for the four-family unit or at broader ornarrower levels of taxonomic unversality. Thereduced size of the jaw dentition in chilodontidsand prochilodontids may be homologous, perhapsalso representing an intermediate stage in a re-ductional trend, which achieves its terminal con-dition in the edentulous jaws of curimatids. Theambiguity associated with the question of homol-ogy in the character makes it impossible to deter-mine the appropriate level at which to use themorphological information. Similarly, it is notpossible to determine at what phylogenetic level

it is correct to use the various adaptations of thethird postcleithrum in the Chilodontidae. Theabove characters, the analyses of polarities, andalternate scenarios consistent with their phyloge-netic distribution were discussed in greater detailin preceeding sections of this paper.

Comparisons with Previous Classifications

A number of alternate taxonomic conceptshave been proposed for the families Curimatidae,Prochilodontidae, Anostomidae, and Chilodonti-dae. Those treatments typically lack an explicitstatement of the methodology of analysis and thephylogenetic assumptions underlying the classi-fication. Rather, each of the previously proposedclassifications is an amalgam of phenetic andquasi-phyletic concepts, with a consequent lowretrievability of the underlying phylogenetic as-sumptions. Two exceptions, at least with respectto the last factor, were the studies of Eigenmann(1917) and Roberts (1973).

The differing classificatory schemes or phylo-genetic hypotheses fall into five major patterns,arranged chronologically by first presentation(family concepts follow Greenwood et al., 1966).

1. The union of the Curimatidae, Prochilodon-tidae, Chilodontidae, and Parodontidae in onetaxon and the separation of the Anostomidaein a second taxon of equivalent rank (Giinther,1864).

2. The placement of the Curimatidae and Pro-chilodontidae with the Citharinidae; the An-ostomidae with some members of the Chara-cidae and Lebiasinidae; and the Chilodonti-dae with the Hemiodontidae and Parodonti-dae (Boulenger, 1904).

3. The incorporation of the Curimatidae, Pro-chilodontidae, Anostomidae, and Chilodonti-dae into a single taxon (Regan, 1911; Gregoryand Conrad, 1938).

4. The recognition of the Curimatidae, Pro-chilodontidae, Anostomidae, Chilodontidae,Hemiodontidae, and Anodinae as a discretelineage (Eigenmann, 1917).

5. The union of the Curimatidae, Prochilodon-tidae and Chilodontidae as a taxon, distinct

NUMBER 378 55

from, but equivalent in rank to, that contain-ing the Anostomidae (Gery, 1977b).

6. The hypothesis that the Prochilodontidae isnot closely related to the Curimatidae butrather to the Anostomidae (? and Chilodonti-dae) (Roberts, 1973).

The degree to which phylogenetic concepts, ifany, were incorporated into the majority of theabove classificatory schemes is uncertain. Thus,it is only possible to note the utility, or lackthereof, of the previously suggested taxonomicsubdivisions as appropriate indicators of the hy-pothesis of relationships arrived at in this study,given the criterion that all recognized taxa mustbe monophyletic. Eigenmann's (1917) hypothesiscan reasonably be interpreted as a phyletic con-cept, although with a limited degree of resolutionof relationships within his proposed lineage. Sim-ilarly, Roberts (1973) dealt with questions ofrelationship, although there is some question asto the exact hypothesis advanced.

Giinther's (1864) Anostomina group, whichconsisted of what are now considered anostomids,is monophyletic according to the findings of thisstudy. Prochilodontids, curimatids, and chilodon-tids, together with hemiodontids and parodon-tids, all were included in his Curimatina. Giventhe separation of anostomids into a separategroup, as the Anostomina, Giinther's Gurimatinagroup is nonmonophyletic, in that it does notinclude all descendents of the hypothesized com-mon ancestor of the lineage. Thus, Giinther'sclassification is not an appropriate vehicle toconvey the hypothesis of the relationships arrivedat herein for the Curimatidae, Prochilodontidae,Anostomidae, and Chilodontidae. Furthermore,the Hemiodontidae, placed by Giinther in theCurimatina group, shares various derived char-acters, including the presence of a rhinosphenoidand tooth form, with subunits of the Characidae,with which it is presumably most closely related.The exact phylogenetic placement of parodontidsis not resolved, but available data fails to supportthe hypothesis that they are more closely relatedto the remaining members of Giinther's Curima-tina group than those taxa are to anostomids.

Boulenger's taxonomic scheme (1904) placessubunits of the monophyletic four-family unit inthree different taxa that each contain other char-aciform groups. His proposed classification is,thus, highly incongruent with the proposed phy-logenetic history of the clade. Temporarily put-ting aside the works of Regan and of Gregory andConrad, we can see that Eigenmann's more ex-plicit phyletic statement agrees with that pre-sented here in uniting curimatids, prochilodon-tids, chilodontids, and anostomids into a singlelineage. Eigenmann did not, however, suggestwhat were the relationships among these fourfamilies. The limited internal resolution of hishypothesis restricts comparisons at lower levels ofuniversality. Our present understanding of phy-logenetic relationships within characiforms, indi-cates, however, that the Hemiodontidae and An-odinae, which he also aligned with those taxa, areactually more closely related to other phyleticsubunits of the order.

Gery (1977b), in recognizing an Anostomidaefor a group that consists of anostomids, and aCurimatidae formed by curimatids, chilodontids,and prochilodontids, nearly approximatesGunther's classification; differing only in separat-ing the Parodontidae into a separate taxon. Giventhe fact that his Curimatidae does not satisfy theprimary criterion of this analysis—that it containsall the descendents of its hypothesized commonancestor—it is considered an invalid indicator ofthe hypothesis of relationships arrived at in thisstudy.

Returning to the earlier classifications of Regan(1911) and Gregory and Conrad (1938), we seethat their union of what are presently recognizedas the Curimatidae, Anostomidae, Chilodontidae,and Prochilodontidae into a single taxon accu-rately reflects the monophyly of the group otherthan for the inclusion of the genus Anodus, whichis apparently more closely related to the Hemio-dontidae (Roberts, 1974). These works are theonly previous studies that utilized a classificatoryscheme under which nearly all recognized taxaconstitute monophyletic assemblages under theresults of this study. Gregory and Conrad (1938)were not explicit about their concepts of relation-


ships within the four-family lineage. Therefore,comparisons cannot be made at lower levels ofuniversality between their scheme and the hy-pothesis advanced in this study. Other than forthe aforementioned inclusion of Anodus in theCurimatinae by Regan, his classification agreeswith the phylogenetic hypothesis advanced hereinboth at the level of the four-family unit and atlower levels of universality. Regan did not com-ment on the possible relationships of the four-taxon group to other characiforms, whereas Gre-gory and Conrad made a qualified suggestionthat the assemblage formed by these four familieswas most closely related to the Citharinidae(sensu Greenwood et al., 1966). The numeroussynapomorphies for citharinids and distichodon-tids (Vari, 1979) indicate that Gregory and Con-rad's suggestion is incongruent with availabledata on characiform phylogeny.

Regan united the taxa as a family with threesubfamilies, whereas Gregory and Conrad consid-ered them to form a subfamily. More recentlyeach of the components has been recognized as afull family (Greenwood et al., 1966). It is, ofcourse, presently impossible to be non-arbitraryin the decision as to which taxonomic level is themost appropriate for the entire assemblage or itssubunits given a possible range between a singlesubfamily and four families. Such a decision mustawait a phylogenetic hypothesis of this clade'splacement within characiforms. The recognitionof four separate families in this study is in keepingwith the general practice of most recent works.Suprafamilial taxa for more inclusive groupingsare not proposed at this time because of our stillpreliminary concepts of higher level relationshipswithin the Characiformes.

It was noted earlier that Roberts' treatment(1973) of the Prochilodontidae was the one workto advance phylogenetic concepts for at leastportions of the four-family assemblage. Roberts,without specific cross-reference, commented thatearlier researchers had proposed a close relation-ship between the Curimatidae and Prochilodon-tidae. His primary premise, in contrast, was thatthe relationships of prochilodontids lie withgroups other than curimatids. Some ambiguity

exists on the identity of the specific group orgroups he considered the closest relatives to theprochilodontids. In the abstract to his work(1973:213) he stated that prochilodontids "areclosely related to Anostomidae and Chilodonti-dae," whereas in his discussion of relationships(1973:221) he stated that "the evidence favoringrelationship between Prochilodontidae and An-ostomidae is relatively strong." The latter, morelimited concept of the sister group to prochilo-dontids is that discussed at greatest length in thepaper and, thus, presumably more in keepingwith the ideas of the author. Both concepts ofoutgroup relationship run counter to the resultsof this study. The hypothesis that prochilodontidsare phylogenetically aligned with the lineage thatconsists of anostomids and chilodontids wouldnecessitate the especially unparsimonious hypoth-esis of repeated convergent acquisitions in theCurimatidae and Prochilodontidae of all elevenof the characters herein considered synapomor-phies for the members of the curimatid-prochilo-dontid clade. The more restricted hypothesis ofrelationships that uses the Anostomidae as thesister group to the Prochilodontidae would in-volve both those concepts of convergence and theindependent acquisition in the Chilodontidaeand Anostomidae of the numerous derived char-acters considered synapomorphies for that bifam-ilial lineage in this study. Thus, both of the twopossible sister group concepts mentioned by Rob-erts are significantly less parsimonious than thehypothesis of relationship arrived at in this study.

Resumen

Una hipotesis acerca de las relaciones filoge-neticas de las familias Curimatidae, Prochilodon-tidae, Anostomidae, y Chilodontidae pertene-cientes al Orden Characiformes es presentada,utilizando la metodologia de la sistematica filo-genetica (cladistica) propuesta por Hennig(1950). Numerosos caracteres derivados (apomor-ficos), compartidos por estas cuatro familias, fu-eron encontrados en las mandibulas, arcos bran-quiales, suspensorium, neurocraneo, columna ver-tebral, y musculatura de la cabeza. Los resultados

NUMBER 378 57

de este estudio indican que los Curimatidae estanmas cercanamente relacionados con los Prochil-odontidae. La unidad formada por estas dos fam-ilias es a su vez el grupo hermano (sister group)de la unidad formada por las familias Anostomi-dae y Chilodontidae. Se encontraron tambienuna serie de carcateres unicos (autapomorfias),que definen a estas familias como unidades mon-ofileticas.

La hipotesis de relaciones desarrollada en estesentido rechaza la sugerencia presentada porRoberts (1973:221) la cual establece que los Pro-chilodontidae estan mas cercanamente relacion-ados con la familia Anostomidae, que a los Cur-imatidae. Boulenger (1904) reunio los Curimati-dae y Prochilodontidae con la familia de chara-ciformes africanos Citharinidae. Los datos obten-idos en este estudio y el trabajo previo de Vari(1979) indican que el criterio de Boulenger creaun taxon no-natural. Giinther (1864) y Gery(1977a) erigieron un taxon para incluir Anostom-idae y otro taxon de rango equivalente para losCurimatidae, Prochilodontidae, y Chilodontidae.Segun el criterio de este estudio, el ultimo taxonpropuesto por los citados autores es no-monofile-tico y no refleja la historia evolutiva de este grupode familias. El concepto de relaciones filogeneti-cas llevadas adelante por Eigenmann (1917) estade acuerdo con la filogenia hipotetica propuestaen este estudio. Sin embargo, Eigenmann incluyea los Hemiodontidae en su discusion, grupo quees excluido en este trabajo. Roberts (1974) sugiereque el genero Anodus (un characiforme sin dientes)no esta cercanamente relacionado con los Curi-

matidae, a pesar de la larga tradicion taxonomicala cual asocia estos taxa. En este estudio, lainclusion de Anodus en la familia Curimatidaeesta apoyada por varios caracteres derivados com-partidos (sinapomorfias).

Tres formas diferentes de organos epibranqui-ales (bien desarrollados) han sido determinadosen los Characiformes. Los Curimatidae y Pro-chilodontidae tienen un organo epibranquial pa-recido a un gran saco muscular, mientras que losde la familia africana Citharinidae son altamentedivididos y soportados por una serie de pequenasosificaciones. El organo epibranquial en Chilo-dontidae esta formado por una extension cartila-ginosa de quinto epibranquial, una banda detejido conectivo asociado al cuarto epibranquialy partes del cuarto y quinto ceratobranquiales.Las diferencias morfologicas entre estos tres tiposde paquetes faringeos y los datos disponibles sobrelas interrelaciones de los Characiformes indicanque organos epibranquilaes bien desarrollados,han evolucionado separadamente, al menos tresveces, dentro de este Orden. Esta conclusion estade acuerdo con la hipotesis previa de Nelson(1967) quien sugirio que los organos epibranqui-ales desarrollados aparecieron independiente-mente en diferentes grupos de peces. Bertmar,Kapoor, y Miller (1969) en contraste, considera-ron que estos organos son ancestrales para lospeces teleosteos, cuya ausencia en muchas lineasevolutivas seria debido a perdidas secundarias.Los resultados de este estudio indican que estaultima hipotesis no es parsimoniosa cuando esaplicada al Orden Characiformes.

Literature Cited

Alexander, R. McN.1964. Adaptation in the Skulls and Cranial Muscles of

South American Characinoid Fish. Journal of theLimean Society (Zoology), 45(305): 169-190.

Angelescu, V., and F.S. Gneri1949. Adaptaciones del aparato digestivo al regimen

alimenticio en algunos peces del Rio Uruguay yde Rio de la Plata. Revista de la Institute National deInvestacion de las Ciencias Naturales, 1(6): 161-272.

Bertmar, G.1961. Are the Accessory Branchial Organs in Characi-

dean Fishes Modified Fifth Gills or RudimentaryUltimobranchial Bodies? Acta Zoologica, 42:151-162.

Bertmar, G., B.G. Kapoor, and R.V. Miller1969. Epibranchial Organs in Lower Teleostean

Fishes—An Example of Structural Adaptation. InW.J.L. Felts, and RJ. Harrison, editors. Interna-tional Review of General and Experimental Zoology, 4:1-48.

Boulenger, G.A.1904. Fishes (Systematic Account of the Teleostei). In

S.F. Harmer and A.E. Shipley, editors, The Cam-bridge Natural History, 7:475-727. London: Mac-millan and Co. Ltd.

Daget.J.1958. Le genre Paradistichodus (Poissons, Characiformes).

Bulletin d'Institute Francois de I'Afrique Noire, 20:1360-1378.

1959. Note sur les Distichodus (Poissons, Characiformes)de 1'Ouest African. Bulletin d'Institute Francois deI'Afrique Noire, 21:1275-1303.

1960. Le genre Xenocharax (Poissons, Characiformes). Re-vue de Zoologie et de Botanique Africaines, 61:35-48.

1962a. Le genre Citharinus. Revue de Zoologie et de BotaniqueAfricaines, 66:81-106, figures 1-12.

1962b. Le genre Citharidium. Bulletin d'Institute Francois deI'Afrique Noire, 24(2):505-522.

1964. Le Crane des Teleosteens. Memoires du MuseumNational d'Histoire Naturelle, 31(2): 163-342, figures1-53.

Dingerkus, G., and L. Uhler1977. Enzyme Clearing of Alcian Blue Stained Whole

Vertebrates for Demonstration of Cartilage. StainTechnology, 52:229-232.

Eigenmann, C.H.1917. The American Characidae. Memoirs of the Museum

of Comparative Zoology, 43(1): 1-102.

Eigenmann, C.H., and W.R. Allen1942. Fishes of Western South America, I: The Inlercordilleran

and Amazonian Lowlands of Peru. II: The High Pampasof Peru, Bolivia, and Northern Chile, with a Revision ofthe Peruvian Gymnotidae and of the Genus Orestias. 494pages. Lexington: the University of Kentucky.

Eigenmann, C.H., and R.S. Eigenmann1889. A Revision of the Edentulous Genera of the Cur-

imatinae. Annals of the New York Academy of Sciences,4:409-440.

Fernandez-Yepez, A.1948. Los Curimatidos (Peces fluviales de Sur America):

Catalogo descriptivo con nuevas adiciones gener-icas y especificas. Boletin Taxonomico del Laboratorwde Pesqueria de Caiquire, 1:1 -86.

Fink, S.V., and W.L. Fink1981. Interrelationships of the Ostariophysan Fishes

(Teleostei). Zoological Journal of the Linnean Society,72(4): 297-353.

Gery.J.1960. Contributions a l'etude des Poissons Characoides

(no. 7): Validite de Leporinus despaxi Puyo et dusous-genre Hypomasticus Borodin. Bulletin du Mu-seum d'Histoire Naturelle, 32(3):222-229, figures 1-4.

1961. Contributions a l'etude des Poissons Charcoides,13: Structures et evolution des Anostominae. Bul-letin of Aquatic Biology, 2(19):93-112, figures 1-24.

1964. A Review of the Chilodinae, with a Key to theSpecies. Tropical Fish Hobbyist, 12(9):5-10, 63-67,9 figures.

1977a. Deux nouveaux cas de mimetisme chez les pois-sons: curimatides mimetiques d'hemiodides(Characoides). Revue Francois de Aquariologie, 4:103-106.

1977b. Characoids of the World. 672 pages. Neptune City,New Jersey: TFH Publications.

Goulding, M.1981. Man and Fisheries on an Amazon Frontier, xiii + 137

pages. The Hague: W. Junk Publishers.Greenwood, P.H., D.E. Rosen, S.H. Weitzman, and G.S.

Myers1966. Phyletic Studies of Teleostean Fishes with a Pro-

visional Classification of Living Forms. Bulletin ofthe American Museum of Natural History, 131(4): 339-456.

Gregory, W.K., and G.M. Conrad1938. The Phylogeny of the Characin Fishes. Zoologica

(New York), 23(4):319-360, figures 1-37.

58

NUMBER 378 59

Giinther, A.1864. Catalogue of the Fishes in the Collection of the British

Museum. Volume 5, xii + 455 pages. London.Heim, W.

1935. Uber die Rachensacke der Characiniden und iiberverwandte akzessorische Organe bei andern Te-leosteen. Zoologische Jahrbucher, 60:61-106.

Hennig, W.1950. Grundizuge einer Theorie der phylogenetischen Systerna-

tike. 370 pages. Berlin: Deutscher Zentralverlag.1966. Phylogenetic Systematics. 263 pages. Urbana: Univer-

sity of Illinois Press.Howes, G.

1976. The Cranial Musculature and Taxonomy of Char-acoid Fishes of the Tribes Cynodontini and Char-acini. Bulletin of the British Museum (Natural History),Zoology, 29:203-248.

1981. Anatomy and Phylogeny of the Chinese MajorCarps Clenopharyhgodon Steind., 1866 and Hypo-phthalmichthys Blkr., 1860. Bulletin of the BritishMuseum (Natural History), Zoology, 41(1): 1-52.

Kner, R.1859. Zur Familie der Characinen, III: Folge der ichthy-

ologischen Beitrage. Denkschriften der Akademit derWissenschafien (Wien), 17:137-182, plates 1-9.

1861. Ueber Kieman-Anhange bei Characinen. Verhan-dlungen der Zoologisch-Botanischen Gessellschaft inWien, 11:189-192.

Lowe-McConnell, R.1975. Fish Communities in Tropical FreshwaUrs. 337 pages.

New York: Longman Publishing.Nelson, GJ.

1967. Epibranchial Organs in Lower Teleostean Fishes.Journal of Zoology (London), 153:71-89.

1969. Gill Arches and the Phylogeny of Fishes withNotes on the Classification of Vertebrates. Bulletinof the American Museum of Natural History,141(4):475-522.

1973. Relationships of Clupeiomorphs, with Remarks onthe Structure of the Lower Jaw in Fishes. Zoological

Journal of the Linnean Society, 53 (supplement 1):333-349.

Nelson, G.J., and N. Platnick1981. Systematics and Biogeography: Cladistics and Vicariance.

567 pages. New York: Columbia University Press.Patterson, C.

1975. The Braincase of Pholodophorid and LeptolepidFishes with a Review of the ActinopterygianBraincase. Philosophical Transactions of the Royal So-ciety, series B, 269:275-579.

Rauther, M.1910. Die Akzessorischen Atmungsorgane der Knock-

enfische. Ergebnisse wide Fortschritte der Zoologie,2:517-585.

Regan, C. T.1911. The Classification of the Teleostean Fishes of the

Order Ostariophysi, I: Cyprinoidea. Annals andMagazine of Natural History, series 8, 8:13-32.

Roberts, T.R.1969. Osteology and Relationships of Characoid Fishes,

Particularly the Genera Hepsetus, Salminus, Hoplias,Ctenolucius and Acestrorhynchus. Proceedings of the Cal-ifornia Academy of Sciences, 36(15):391-500.

1973. Osteology and Relationships of the Prochilodon-tidae, a South American Family of CharacoidFishes. Bulletin of the Museum of Comparative Zoology,145(4):213-235, figures 1-29.

1974. Osteology and Classification of the NeotropicalCharacoid Fishes of the Families Hemiodontidae(including Anodontinae) and Parodontidae. Bul-letin of the Museum of Comparative Zoology, 146(9):411-472, figures 1-78.

Rosen, D.E., and P.H. Greenwood1970. Origin of the Weberian Apparatus and the Rela-

tionships of the Ostariophysan and Gonorhynchi-form Fishes. American Museum Novitates, 2428:1-25.

Sagemehl, M.1885. Uber die Pharyngealtaschen der Scarinen und das

"Wiederkauen" dieser Fische. Morphologischesjahr-buch, 10:192-203.

1887. Die Accessorischen Branchialorgane von Cithari-nus. Morphologisches Jahrbuch, 12:307-323.

Smith, N.1981. Man, Fishes and the Amazon. 180 pages. New York:

Colombia University Press.Spix, J.B. von, and L. Agassiz

1829. Selecta Genera et Species Pisdum Quos in Itinere perBraziliam Amis 1817-20 Perecato Collegit J.B. deSpix. 138 pages. Munich.

Van, R.P.1979. Anatomy, Relationships and Classification of the

Families Citharinidae and Distichodontidae(Pisces, Characoidea). Bulletin of the British Museum(Natural History), Zoology, 36(5):261-344.

1982a. Hemiodopsis ocellata, a New Hemiodontid Chara-coid Fish (Pisces: Characoidea) from Western Sur-inam. Proceedings of the Biological Society of Washing-ton, 95(\)A88-\93.

1982b. Systematics of the Neotropical Characoid GenusCwrimatopsis (Pisces: Characoidei). Smithsonian Con-tributions to Zoology. 373: 28 pages.

Weitzman, S.H.1954. The Osteology and the Relationships of the South

American Characid Fishes of the Subfamily Gas-teropelecinae. Stanford Ichthyological Bulletin, 4(4):213-263.

1962. The Osteology of Brycon meeki, a Generalized Char-acid Fish, with an Osteological Definition of theFamily. Stanford Ichthyological Bulletin, 8(1): 1-77,figures 1-21.

Wiley, E.O.1981. Phylogenetics: The Theory and Practice of Phylogenetic


Systematics. vi + 439 pages. New York: John Wileyand Sons.

Winterbottom, R.1974. A Descriptive Synonymy of the Striated Muscles

of the Teleostei. Proceedings of the Academy of Natural

Sciences of Philadelphia, 125:225-317.

1980. Systematics, Osteology and Phylogenetic Rela-

tionships of Fishes of the Ostariophysan Subfamily

Anostominae (Characoidei, Anostomidae). Life

Sciences Contribution, Royal Ontario Museum, 123:1

112, figures 1-68.

REQUIREMENTS FOR SMITHSONIAN SERIES PUBLICATION

Manuscripts intended for series publication receive substan-tive review within their originating Smithsonian museums oroffices and are submitted to the Smithsonian Institution Presswith Form SI-36, which must show the approval of the appropri-ate authority designated by the sponsoring organizational unit.Requests for special treatment—use of color, foldouts, case-bound covers, etc.—require, on the same form, the addedapproval of the sponsoring authority.

Review of manuscripts and art by the Press for requirementsof series format and style, completeness and clarity of copy, andarrangement of all material, as outlined below, will govern, withinthe judgment of the Press, acceptance or rejection of manu-scripts and art.

Copy must be prepared on typewriter or word processor,double-spaced, on one side of standard white bond paper (noterasable), with VU" margins, submitted as ribbon copy (notcarbon or xerox), in loose sheets (not stapled or bound), andaccompanied by original art. Minimum acceptable length is 30pages.

Front matter (preceding the text) should include: title pagewith only title and author and no other information; abstractpage with author, title, series, etc., following the establishedformat; table of contents with indents reflecting the hierarchy ofheads in the paper; also, foreword and/or preface, if appropri-ate.

First page of text should carry the title and author at the topof the page; second page should have only the author's nameand professional mailing address, to be used as an unnumberedfootnote on the first page of printed text.

Center heads of whatever level should be typed with initialcaps of major words, with extra space above and below thehead, but with no other preparation (such as all caps or under-line, except for the underline necessary for generic and specificepithets). Run-in paragraph heads should use period/dashes orcolons as necessary.

Tabulations within text (lists of data, often in parallel columns)can be typed on the text page where they occur, but they shouldnot contain rules or numbered table captions.

Formal tables (numbered, with captions, boxheads, stubs,rules) should be submitted as carefully typed, double-spacedcopy separate from the text; they will be typeset unless other-wise requested. If camera-copy use is anticipated, do not drawrules on manuscript copy.

Taxonomic keys in natural history papers should use thealigned-couplet form for zoology and may use the multi-levelindent form for botany. If cross referencing is required betweenkey and text, do not include page references within the key, butnumber the keyed-out taxa, using the same numbers with theircorresponding heads in the text.

Synonymy in zoology must use the short form (taxon, author,yearpage), with full reference at the end of the paper under

Literature Cited.' For botany, the long form (taxon, author,abbreviated journal or book title, volume, page, year, with noreference in Literature Cited") is optional.

Text-reference system (author, yearpage used within thetext, with full citation in Literature Cited" at the end of the text)must be used in place of bibliographic footnotes in all Contri-butions Series and is strongly recommended in the StudiesSeries: (Jones, 1910:122) or . . .Jones (1910:122). Ifbibliographic footnotes are required, use the short form (author,

brief title, page) with the full citation in the bibliography.Footnotes, when few in number, whether annotative or bib-

liographic, should be typed on separate sheets and insertedimmediately after the text pages on which the references occur.Extensive notes must be gathered together and placed at theend of the text in a notes section.

Bibliography, depending upon use, is termed "LiteratureCited," "References," or "Bibliography." Spell out titles ofbooks, articles, journals, and monographic series. For book andarticle titles use sentence-style capitalization according to therules of the language employed (exception: capitalize all majorwords in English). For journal and series titles, capitalize theinitial word and all subsequent words except articles, conjunc-tions, and prepositions. Transliterate languages that use a non-Roman alphabet according to the Library of Congress system.Underline (for italics) titles of journals and series and titles ofbooks that are not part of a series. Use the parentheses/colonsystem for volume(number):pagination: "10(2):5-9. " For align-ment and arrangement of elements, follow the format of recentpublications in the series for which the manuscript is intended.Guidelines for preparing bibliography may be secured fromSeries Section, SI Press.

Legends for illustrations must be submitted at the end of themanuscript, with as many legends typed, double-spaced, to apage as convenient.

Illustrations must be submitted as original art (not copies)accompanying, but separate from, the manuscript. Guidelinesfor preparing art may be secured from Series Section, SI Press.All types of illustrations (photographs, line drawings, maps, etc.)may be intermixed throughout the printed text. They should betermed Figures and should be numbered consecutively as theywill appear in the monograph. If several illustrations are treatedas components of a single composite figure, they should bedesignated by lowercase italic letters on the illustration; also, inthe legend and in text references the italic letters (underlined incopy) should be used: "Figure 9b. " Illustrations that are in-tended to follow the printed text may be termed Plates, and anycomponents should be similarly lettered and referenced: "Plate9b." Keys to any symbols within an illustration should appearon the art rather than in the legend.

Some points of style: Do not use periods after such abbre-viations as "mm, ft, USNM, NNE." Spell out numbers "one"through "nine in expository text, but use digits in all othercases if possible. Use of the metric system of measurement ispreferable; where use of the English system is unavoidable,supply metric equivalents in parentheses. Use the decimal sys-tem for precise measurements and relationships, common frac-tions for approximations. Use day/month/year sequence fordates: 9 April 1976." For months in tabular listings or datasections, use three-letter abbreviations with no periods: Jan,Mar. Jun," etc. Omit space between initials of a personal name:"J.B. Jones."

Arrange and paginate sequentially every sheet of manu-script in the following order: (1) title page, (2) abstract, (3)contents, (4) foreword and/or preface, (5) text, (6) appendixes.(7) notes section, (8) glossary, (9) bibliography, (10) legends,(11) tables. Index copv may be submitted at page proof stage,but plans for an index should be indicated when manuscript issubmitted.

«•

,111/.,

phylogenetic relationships of the families curimatidae, … · 2017-08-20 · smithsonian...

Documents