A PHYLOGENETIC AND BIOGEOGRAPHIC ANALYSIS OF CYPRINODONTIFORM FISHES (TELEOSTEI, ATHERINOMORPHA) LYNNE R. PARENT] BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY VOLUME 168 : ARTICLE 1 NEW YORK : 1981 A PHYLOGENETIC AND BIOGEOGRAPHIC ANALYSIS OF CYPRINODONTIFORM FISHES (TELEOSTEI, ATHERINOMORPHA) LYNNE R. PARENTI Department of Ichthyology American Museum of Natural History SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN THE FACULTY IN BIOLOGY OF THE CITY UNIVERSITY OF NEW YORK BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY VOLUME 168 : ARTICLE 4 NEW YORK : 1981 BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY Volume 168, article 4, pages 335-557, figures 1-99, tables 1-3 Issued September 3, 1981 Price: $14.35 a copy ISSN 0003-0090 This article completes Volume 168. Copyright © American Museum of Natural History 1981 CONTENTS Abstract Introduction Acknowledgments Methods Overview of Past Internal Classifications of Cyprinodontiform Fishes Derived Characters of Cyprinodontiforms Phylogenetic Analysis Aplocheiloids (Group B) Neotropical Aplocheiloids Cladistic Summary of Neotropical Aplocheiloids Old World Aplocheiloids The Aphyosemion-Nothobranchius Group The Aplocheilus-Pachypanchax-Epiplatys Group Cladistic Summary of Old World Aplocheiloids Cyprinodontoids (Group C) Relationships of the Cyprinodontoids Internal Fertilization and Viviparity Cladistic Summary of the Cyprinodontoids Group D Fundulines Group E Group F Group H Jenynsia, Anableps, and Oxyzygonectes Poeciliids, Fluviphylax, Pantanodon, and the Procatopines Group G Empetrichthys, Crenichthys, and the Goodeidae Group I Cubanichthys and Chriopeoides Group J Orestias and the Anatolian Cyprinodontines New World Cyprinodontines Classification Key to Genera and Suprageneric Categories Systematic Accounts Order Cyprinodontiformes Suborder Aplocheiloidei Family Aplocheilidae Genus Aplocheilus Genus Pachypanchax Genus Epiplatys Subgenus Lycocyprinus Subgenus Parepiplatys Subgenus Pseudepiplatys Subgenus Aphyoplatys 341 341 344 346 349 354 365 367 375 385 386 389 394 396 397 404 429 443 443 443 447 447 447 448 450 ,454 454 457 457 457 457 458 461 465 470 470 471 471 471 473 474 475 475 475 475 Genus Aphyosemion Subgenus Archiaphyosemion Subgenus Chromaphyosemion Subgenus Diapteron Subgenus Kathetys Subgenus Mesoaphyosemion Genus Fundulopanchax Subgenus Paludopanchax Subgenus Paraphyosemion Subgenus Gularopanchax Subgenus Callopanchax Subgenus Raddaella Genus Adamas Genus Nothobranchius Family Rivulidae Genus Rivulus Genus "Rivulus" Genus Trigonectes Genus Pterolebias Genus Rachovia Genus Austrofundulus Genus Neofundulus Genus "Neofundulus" Genus Cynolebias Suborder Cyprinodontoidei Section 1 Family Profundulidae Genus Profundulus Section 2 Division 1 Family Fundulidae Genus Plancterus Genus Fundulus Genus Lucania Genus Leptolucania Genus Adinia Division 2 Sept 1 Family Valenciidae, New Family Genus Valencia Sept 2 Superfamily Poecilioidea Family Anablepidae Subfamily Anablepinae Genus Anableps Genus Jenynsia 476 477 477 477 477 477 477 479 479 479 479 479 479 479 481 481 483 484 485 486 487 489 489 490 491 492 492 492 493 493 493 493 494 496 497 498 499 499 499 500 501 501 501 501 501 503 Subfamily Oxyzygonectinae, New Subfamily Genus Oxyzygonectes Family Poeciliidae Subfamily Poeciliinae Subfamily Aplocheilichthyinae Genus Aplocheilichthys Subgenus Micropanchax Subgenus Lacustricola Subgenus Poropanchax Subgenus Congopanchax Genus Lamprichthys Genus Pantanodon Genus "Aplocheilichthys" Genus Procatopus Genus Cynopanchax Genus Plataplochilus Genus Hypsopanchax Subfamily Fluviphylacinae Genus Fluviphylax Superfamily Cyprinodontoidea Family Goodeidae Subfamily Empetrichthyinae Genus Empetrichthys Genus Crenichthys Subfamily Goodeinae Family Cyprinodontidae Subfamily Cubanichthyinae, New Subfamily Genus Cubanichthys Subfamily Cyprinodontinae Tribe Orestiini Genus Aphanius Genus "Aphanius" Genus Kosswigichthys Genus Orestias Tribe Cyprinodontini Genus Cyprinodon Genus Jordanella Genus Cualac Genus Floridichthys Genus Megupsilon Historical Biogeography Summary Literature Cited 504 504 505 505 507 507 508 508 508 508 508 509 510 511 513 513 513 514 514 515 515 516 516 517 518 519 519 520 521 521 521 522 524 525 526 526 528 529 530 531 533 544 547 ABSTRACT The cyprinodontiforms, or killifishes, are a large and diverse group of 900 fresh- and brackishwater species with a pantropical and temperate Laurasian distribution. Traditionally, it has been classified in five families: the worldwide, oviparous Cyprinodontidae, and four New World viviparous families: the Poeciliidae, Anablepidae, Jenynsiidae, and Goodeidae. Fishes of the diverse Cyprinodontidae, in turn, have been divided into as many as eight subfamilies. The objectives of the present study are to: (1) determine if the cyprinodontiform fishes as a whole form a monophyletic group; (2) determine if each of the five families is monophyletic; (3) define the major subgroups of cyprinodontiforms, concentrating on the genera of the Cyprinodontidae; (4) determine the interrelationships of the subgroups; (5) present a comprehensive classification of the cyprinodontiforms that reflects the interrelationships; and (6) provide a hypothesis for the distribution of the group. The following general results were obtained by using the methods of phylogenetic systematics and vicariance biogeography: (1) the cyprinodontiforms are considered to be monophyletic by their sharing derived characters of the caudal skeleton, upper jaw, gill arches, position of the first pleura! rib, pectoral girdle, and aspects of breeding and development; (2) the family Cyprinodontidae is nonmonophyletic as it contains some of the most primitive and derived cyprinodonti- forms; (3) each of the four viviparous families is monophyletic; however, their previous definitions in terms of uniquely derived characters have been altered; (4) the development of an annual habit, exhibited by members of the aplocheiloid killifishes and possibly some cyprinodontoids, includes derived reproductive traits exhibited to some degree by all killifishes; therefore, the annual habit does not define a monophyletic group of killifishes; (5) similarly, viviparity is not hypothesized to be a uniquely derived character, but has apparently arisen at least three times within the group; and (6) the interrelationships of cyprinodontiforms correspond, in part, with a pattern of the break-up of Pangea, except for an Andean-Eurasian sister group pair. A scheme of interrelationships of cyprinodontiforms as well as of monophyletic subgroups is presented in the form of cladograms, of which the former is transformed into a comprehensive classification of the group. The fishes under study are recognized as comprising the order Cyprinodontiformes Berg and divided into two suborders, the Aplocheiloidei (which previously comprised, in part, the Cyprinodontidae), and the Cyprinodontoidei (comprising all other cyprinodontiforms as well as the four viviparous families). In order to minimize the number of named empty categories, a numbering system is incorporated into a traditional naming system to create the new classification. INTRODUCTION The cyprinodontiforms, commonly known as killifishes, top minnows, or toothcarps, are a large and diverse group of teleostean fishes distributed nearly worldwide in temperate and tropical freshwaters (fig. 1), with some members regularly entering brackish water. The term cyprinodontiforms as used in this paper refers to fishes of the five families of the superfamily Cyprinodontoidea, order Atheriniformes (Rosen, 1964). These are the cosmopolitan and oviparous Cyprinodontidae, and four New World viviparous families, the Anablepidae, Goodeidae, Jenynsiidae, and Poeciliidae. The Cyprinodontidae are the largest and most diverse family containing over 650 nominal species in approximately 80 nominal genera. Included are the popular aquarium fishes, including the annual killifishes of tropical South America and Africa of the subfamily Rivulinae (Myers, 1955), and widely used experimental fish such as those of the genus Fundulus. Of the four viviparous families, two, the Mexican, Central American, and northern South American Anablepidae, and the southeastern South American Jenynsiidae, contain just one genus each with several species. The Goodeidae are diverse, comprising approximately 35 species in 16 genera, all of which are restricted to the Mexican Plateau (Miller and Fitzsimons, 1971). The Neotrop- 342 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 1. Present day distribution of cyprinodontiforms. Dotted and dashed line approximates Wallace's Line. ical and temperate Poeciliidae comprise approximately 200 species in 19 genera (Rosen and Bailey, 1963; Rosen, 1979), and the guppy (Poecilia reticulata) and mosquito-fish (Gambusia affinis) are included. Knowledge of the relationships of the cyprinodontiforms to other fishes has advanced considerably, whereas the proposed interrelationships of cyprinodontiforms has progressed little since Carman's (1895) outline of the major subgroups. Killifishes are typically soft-rayed, and, as such, have historically been aligned with the more primitive teleost groups. Gill (1874) aligned the cyprinodontiforms with the esocoids, which together comprised the order Haplomi. Starks (1904) divided the Haplomi into three suborders: the Esocoidei (including the mud-minnow Umbra, and the pike Esox), the Amblyopsioidei (the cavefishes), and the Poecilioidei (=Cyprinodontoidei of Rosen, 1964). Yet, he admitted there were no important [unique] characters which defined the order. Regan (1911) remarked on the killifish and cavefish relationship to more derived teleost groups while noting that the esocoids were relatively primitive. He included as evidence for this distinction the fact that in esocoids the maxilla enters the gape, whereas the maxilla is excluded from the gape in killifishes (including the adrianichthyoids) and cavefishes. Regan (1909) separated the last two groups from the rest of the Haplomi and constructed for them a new order Microcyprini. This action was supported by Hubbs (1919) who reported that the Microcyprini (including the phallostethoids after Regan, 1913) have a derived branchiostegal number and arrangement, comparable to those of the acanthopterygians; whereas, the Haplomi, sensu Regan, are primitive in this regard. Myers (1928a) removed the phallostethoid fishes from the Microcyprini, and suggested their close relationship to the Atherinidae, then in the order Percesoces. The alignment of the amblyopsoids with the cyprinodontiforms and adrianichthyoids was never more than tentative; yet it remained unchallenged until Rosen (1962) removed the cavefishes from the Microcyprini, referred to as the Cyprinodontiformes following Berg (1940), and placed them in the newly created Amblyopsiformes which he 1981 PARENTI: CYPRINODONTIFORM FISHES claimed was more closely related to the Percopsiformes. The cyprinodontiforms and adrianichthyoids remained as the sole constituents of the order Cyprinodontiformes. Gosline (1963) continued to support the naturalness of the order Microcyprini, and criticized Rosen (1962) for separating the order into two groups while giving no hint as to the placement of the Cyprinodontiformes in a higher classification of teleost fishes. An answer to this was provided by Rosen (1964) when he created the order Atheriniformes to include the cyprinodontiforms, adrianichthyoids, atherinoids, phallostethoids, exocoetoids, and scomberesocoids. Rosen's (1964) classification is summarized in table 1. The alignment of these fishes had casually been suggested earlier by several workers, although, this was done with little formal taxonomic treatment. Cope (1870) first remarked on the possible close relationship of atherinids and cyprinodonts. In addition, Myers (1928a) commented that the structure of the ethmoid region and the mouth suggested the affinity of cyprinodontiforms and members of the Percesoces. Furthermore, Regan (1911, p. 321), commenting on the possible alignment of his new order, said: "Whereas the Haplomi show relationship to the most generalized isospondylous fishes, the Microcyprini bear more resemblance to the Salmopercae and Synentognathi, especially the latter." The monophyly (in the sense of Hennig, 1966) of the order Atheriniformes, and the monophyly and interrelationships of its subgroups were not rigorously defined by Rosen (1964). However, recent evidence indicates that the Atheriniformes is monophyletic, and problems of its higher order interrelationships may easily be summarized (Rosen and Parent!, MS). Rosen (1964, p. 260) suggested that the atherinomorph fishes: "arose from a group that stood somewhere in the ancestry of the order Perciformes." This point, which may be restated as fishes of the Atherinomorpha and Percomorpha share a common ancestor, was reiterated in the classifications of Greenwood, Rosen, Weitzman and Myers (1966), Rosen and Patterson (1969), Rosen (1973a), 343 TABLE 1 Classification of Fishes of the Order Atheriniformes (Rosen, 1964) Superorder Acanthopterygii Series Atherinomorpha Order Atheriniformes Suborder Atherinoidei Superfamily Atherinoidea Superfamily Phallostethoidea Suborder Cyprinodontoidei Superfamily Adrianichthyoidea Superfamily Cyprinodontoidea Family Cyprinodontidae Family Anablepidae Family Jenynsiidae Family Goodeidae Family Poeciliidae Suborder Exocoetoidei Superfamily Exocoetoidea Superfamily Scomberesocoidea Series Percomorpha Patterson and Rosen (1977) and Rosen and Parent! (MS), and is supported by derived features of the gill arches and the jaws and jaw suspensorium. Thus, with increased knowledge of interrelationships of teleosts, cyprinodontiforms have progressed from a primary alignment with the primitive esocoids to a hypothesized close relationship with the advanced percomorph fishes. Yet, as stated previously, our knowledge of the interrelationships of members of the superfamily Cyprinodontoidea has undergone little comparable progress. Relationships among the" families and among the included genera, have been presented as speculation. Workers have either dealt with the primary groups of oviparous cyprinodontiforms alone (e.g., Myers, 1931, 1955; Sethi, 1960; Uyeno and Miller, 1962), or one of the four viviparous families (e.g., Rosen and Bailey, 1963; Hubbs and Turner, 1939; Miller, 1979), never more than casually discussing the relationship of one family to another or to a group of oviparous cyprinodontids. However, aside from discussions and repeated speculation on the affinity of one group of killifishes to another, there has been 344 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY no formal treatment of the interrelationships of the five families, no statement supporting or refuting the monophyly of each of the five families, and no formal definition of the superfamily Cyprinodontoidea. Knowledge of such interrelationships could serve as a basis for biogeographic hypotheses concerning the history of the pantropical and temperate Laurasian regions, could be an invaluable reference for research scientists and aquarists alike, and form the framework for an understanding of the variety of reproductive modes found within the group. Thus, the objectives of this study are to: (1) determine the monophyly of the superfamily Cyprinodontoidea; (2) determine the monophyly of each of the five families; (3) define the major subgroups of cyprinodontiforms, with a concentration on the genera of the family Cyprinodontidae; (4) determine the interrelationships of the subgroups; (5) present a comprehensive classification of the cyprinodontiforms which reflects the interrelationships; and (6) provide a hypothesis for the distribution of the group. ACKNOWLEDGMENTS I thank Dr. Donn E. Rosen for his advice and encouragement throughout the course of this study as well as directing my interests toward many of the problems in cyprinodontiform systematics. I also thank Dr. James S. Farris, and especially Dr. M. F. Mickevich for their early encouragement and support of my systematic endeavors. For helpful discussions of aspects of the research in progress, I thank Mr. Guido Dingerkus, Dr. Mickevich, Ms. Nancy A. Neff and Dr. Richard P. Vari. Ms. M. Norma Feinberg provided invaluable technical advice and assistance for which I am most grateful. Financial and logistical support was provided primarily by the Department of Ichthyology, American Museum of Natural History for which I extend sincere thanks. Additional support was provided by a grant from the Theodore Roosevelt Memorial VOL. 168 Fund, American Museum of Natural History, the City University of New York and the Department of Ichthyology, California Academy of Sciences. For loan of specimens and accommodations during visits, I particularly thank Drs. William N. Eschmeyer and Tomio Iwamoto, Ms. Pearl M. Sonoda, Mr. William C. Ruark and Mr. James Gordon (California Academy of Sciences, San Francisco), as well as Dr. James Bohlke and Mr. William Saul (Academy of Natural Sciences, Philadelphia), Dr. P. Humphry Greenwood, Mr. James Chambers and Mr. Gordon Howes (British Museum [Natural History], London), Dr. Robert K. Johnson and Mr. Garrett S. Glodek (Field Museum of Natural History, Chicago), Dr. Gianna Arbocco (Museo Civico di Storia Naturale, Genova), Dr. William L. Fink and Mr. Karsten Hartel (Museum of Comparative Zoology, Harvard University, Cambridge), Dr. Robert R. Miller (University of Michigan, Museum of Zoology, Ann Arbor), Drs. Richard P. Vari and Stanley H. Weitzman (National Museum of Natural History, Washington, D.C.), and Dr. Raul VazFerreira (Uruguay). For the gift of live or preserved specimens, I thank Mr. John S. Brill, Jr., Dr. A. Casinos, Mr. Thomas Coon, Mr. J. Dry an, Mr. Daniel Fromm, Mr. Tim Hardin, Mr. Leonard Mackowiak and Mr. Jerry Shapiro. For access to a text-editor and computing funds for preparation of the typescript, I thank Drs. Robert F. Rockwell, the City College, and Leslie F. Marcus, Queens College, both of the City University of New York. For information on systematic literature, as well as personal observations, I thank Dr. James W. Atz, Mr. Fromm, Drs. Kenneth J. Lazara, F. Douglas Martin, Mr. Eric Quinter, Drs. Alfred C. Radda, Jamie Thomerson, Bruce J. Turner, Stanley H. Weitzman, E. O. Wiley and John P. Wourms. VERNACULAR NAMES: In a systematic study that ends with a reclassification, names must be used in the discussion of interrelationships and character distributions which are at once familiar to most workers on the group, and which unambiguously refer to a 1981 PARENTI: CYPRINODONTIFORM FISHES given group of genera or families. Thus, I use the following vernacular names throughout the text. The term acanthopterygian refers to fishes of the superorder Acanthopterygii, which includes the two series Atherinomorpha and Percomorpha (Rosen, 1973a). The series are termed atherinomorph and percomorph, respectively. The series Atherinomorpha contains a sole order, the Atheriniformes. Thus, the two categories are equivalent, and the term atherinomorph describes the membership of both. Atheriniform is therefore not used herein to avoid confusion. Within the order Atheriniformes, vernacular names are used for the major subdivisions listed in the classification of table 1. Fishes of the suborder Atherinoidei are referred to as the silversides. The terms atherinoid and phallostethoid are reserved for members of the superfamilies Atherinoidea and Phallostethoidea, respectively. There is no vernacular reference for the suborder Cyprinodontoidei of Rosen; components are referred to separately. The term adrianichthyoid refers to fishes of the superfamily Adrianichthyoidea. The fishes of the superfamily Cyprinodontoidea, the subject of this revision, are referred to alternately as the cyprinodontiforms, cyprinodonts, or killifishes. They are reclassified in this study as the order Cyprinodontiformes. Two suborders are named, the Aplocheiloidei, comprising those fishes of the Rivulinae, and the Cyprinodontoidei, comprising all other cyprinodontiforms. These groups will be referred to as the aplocheiloids and cyprinodontoids, respectively. There is no vernacular reference for the suborder Exocoetoidei, and its two superfamilies are referred to as the exocoetoids (superfamily Exocoetoidea) and the scomberesocoids (superfamily Scomberesocoidea). Within the Cyprinodontoidea, members of the five families are normally referred to as the cyprinodontid, anablepid, jenynsiid, goodeid or poeciliid fishes. The Cyprinodontidae are also referenced as the oviparous 345 TABLE 2 Current Comprehensive Classification of the Cyprinodontid Fishes Family Cyprinodontidae Subfamily Fundulinae Genus Fundulus, Lucania, Leptolucania, Oxyzygonectes, Cubanichthys, Chriopeoides, Valencia, Empetrichthys, Crenichthys, Profundulus, Hubbsichthys,11 Adinia Subfamily Cyprinodontinae Genus Cyprinodon, Megupsilon, Floridichthys, Jordanella, Cualac, Aphanius, Tellia, Kosswigichthys, Anatolichthys Subfamily Lamprichthyinae Genus Lamprichthys Subfamily Orestiatinae Genus Orestias Subfamily Pantanodontinae Genus Pantanodon Subfamily Procatopodinae Genus Aplocheilichthys, Procatopus, Hypsopanchax, Micropanchax, Cynopanchax, Plataplochilus, Platypanchax, Hylopanchax, Congopanchax, Poropanchax Subfamily Rivulinae Genus Rivulus, Trigonectes, Rivulichthys, Pterolebias, Rachovia, Austrofundulus, Terranotus, Cynolebias, Cynopoecilus, Campellolebias, Simpsonichthys, Aphyosemion, Nothobranchius, Adamas, Epiplalys, Aplocheilus, Pachypanchax, Fundulosoma, Callopanchax Subfamily Fluviphylacinae Genus Fluviphylax " Schultz (1949) described Hubbsichthys laurae, new genus and species of cyprinodontid. The holotype (USNM 120999), the only recorded specimen, was examined and determined as a female poeciliid, and most likely of the species Poecilia caucana (Steindachner). I propose Hubbsichthys be dropped from the subfamily and placed in synonymy of the genus Poecilia. killifishes, whereas the four remaining families are collectively referred to as the viviparous killifishes. Vernacular names for the various groups found within the Cyprinodontidae follow the current classification of the family, listed in table 2. The subfamily Cyprinodontinae is referred to as the cyprinodontines, which are further divided into the New World cyprinodontines, comprising Cyprinodon and its imme- 346 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY diate relatives, and the Anatolian or Old World cyprinodontines, comprising Aphanius and its immediate relatives. The subfamilies Fluviphylacinae, Orestiatinae, and Pantanodontinae are referred to as the genera Fluviphylax, Orestias, and Pantanodon, respectively. The Procatopodinae and Lamprichthyinae are collectively referred to as the procatopines. Members of the subfamily Fundulinae, VOL. 168 which are no longer considered to be members of a monophyletic group containing Fundulus, are referred to by their formal generic names (e.g., Oxyzygonectes, Cubanichthys, Chriopeoides, Empetrichthys, Crenichthys, and Profundulus). The term funduline refers to Fundulus, Adinia, Leptolucania, Lucania and their nominal subgenera. Various other groups of teleosts are discussed using conventional terminology. METHODS PHYLOGENETIC ANALYSIS The method of phylogenetic analysis adopted here is that put forth formally by Hennig (1950, 1966), alternately referred to as cladistics, cladism, or phylogenetic systematics. Within a cladistic scheme, taxa are grouped hierarchically on the basis of their sharing derived characters (termed synapomorphies), rather than on their overall similarity. This method of analysis is preferred over those of evolutionary taxonomy (e.g., Mayr, 1969, 1974; Simpson, 1961) and phenetics (e.g., Sokal and Sneath, 1963; Sneath and Sokal, 1973) if the goal is the hierarchical grouping of taxa based solely on a hypothesis of common ancestry. Given the assumption that nature is structured hierarchically, a cladogram that reflects increasing levels of generality of character distributions is concluded to be the best estimate of the one, true phylogeny. Recognized taxa are those which can be defined as monophyletic groups in the sense of Hennig. That is, a monophyletic group contains all the descendants, and only the descendants of a common ancestor. Monophyletic groups, therefore, are defined by their members sharing derived characters. Such groups are assembled into more inclusive monophyletic groups until a hierarchical arrangement of all the members is achieved. It is not the intention of a cladistic analysis to recognize paraphyletic groups. However, since this study is done primarily at the ge- neric level, genera that are not monophyletic may have groups of species assignable to other monophyletic groups, leaving the remainder as a paraphyletic assemblage at the most plesiomorph position of the more inclusive monophyletic group. In these cases, the traditional generic name will be retained, and recommendations for a species-level revision will be made. When character conflicts occur at any level in the analysis, the principle of parsimony is invoked to choose among alternative explanations of the data. The assumption is not made that evolution always, or ever, must proceed along a parsimonious course; however, it is concluded that our explanation of a hypothesized phylogeny should be the most parsimonious one since, by definition, it is the one which requires that we invoke the fewest assumptions about character transformations. Similarly, characters are not weighted in the analysis since no objective criteria for weighting could be determined. Character transformation series are constructed among states of homologous characters. Characters are hypothesized to be homologous if they are comparable in shape and position, or present in different forms, but exhibiting the same ontogenetic sequence. A homology, therefore, is, at one level, comparable to a derived character or apomorphy (Wiley, 1975). 1981 PARENTI: CYPRINODONTIFORM FISHES The polarity of a transformation series is initially determined by comparison to an outgroup (as discussed by Lundberg, 1972), or by comparison with an ontogenetic transformation (as discussed in Nelson, 1973). In the latter procedure, an ontogenetic change in one of two taxa that are hypothesized to share a common ancestor, hence termed sister groups, must logically be considered derived if the principle of parsimony is applied. That is, one need only make the assumption that the transformation was gained by one taxon, rather than the assumptions that the character was present in the common ancestor, and that it was subsequently lost in the other. In the former procedure, a character state is analyzed as being primitive or derived by comparing it to the state within other groups of atherinomorph fishes, within the percomorphs, or to the teleosts as a whole. Characters or character complexes recognized at once as being unique are analyzed as derived. The general state of a character in an outgroup or in the cyprinodontiforms is initially assessed as primitive. However, a character which may be described in the same manner as that of the general state may be termed secondarily derived within a group of cyprinodontiforms if this interpretation is consistent with the most parsimonious interpretation of all the data. That is, the polarity of a transformation series is not always determined by the constraint that the general state represents the primitive condition. Transformation series treated in this manner are discussed in detail. BIOGEOGRAPHIC ANALYSIS: A hypothesis of the historical distribution of cyprinodontiform fishes is constructed upon completion of the phylogenetic analysis. The distribution of monophyletic groups should reflect the history of the areas of distribution if we accept the premise inherent in the works of Croizat (1958, 1964) that the world and its biota evolved together. This concept forms the basis of vicariance biogeography as put forth by Croizat, and Croizat, Nelson and Rosen (1974), Platnick and Nelson (1978) and Rosen (1976, 1978). The cladogram of cyprinodontiforms may 347 readily be transformed into a cladogram of areas occupied by monophyletic groups (Rosen, 1978). A pattern of earth history is suggested by the interrelationships of the areas. The generality of this pattern and those of monophyletic groups will be tested by comparison to other established patterns as well as to each other. DISPOSITION OF SPECIMENS AND COLLECTION OF DATA: Counterstained specimens of cyprinodontiforms were prepared according to the alcian blue-alizarin Red S method of Dingerkus and Uhler (1977) to facilitate the examination of cartilage as well as bone. When possible, at least two males and two females of several species in a genus were prepared. In some cases, just one pair was prepared. When lots were only large enough for the preparation of one specimen a male was chosen since cyprinodontiforms are markedly sexually dimorphic with males typically exhibiting a greater degree of variation than females. Additional specimens, which were cleared and solely alizarin-stained, were available from the collection of fishes in the Department of Ichthyology, AMNH. Radiographs were prepared primarily of species represented only by the type material in order to facilitate a cursory examination of the osteological details. Anatomical illustrations were prepared from sketches of structures as viewed through a camera lucida mounted on a dissecting microscope. Primarily, dissected cleared and stained material was used for this purpose; however, alcohol specimens were partially dissected when necessary. Most illustrations and descriptions of states of cartilaginous elements are of counterstained preparations. Developmental series of available aquarium representatives of several genera of atherinomorphs were bred and reared in the vivarium of the Department of Ichthyology, AMNH. Details of development, including structure of the egg, period of time from spawning to hatching, and age at first spawning were observed for several genera. Adults were also observed for details of reproductive behavior. All preserved aquarium spec- BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 348 imens were catalogued in the department's collection. Measurements and counts were made according to the procedure outlined by Miller (1948) for cyprinodont fishes, except as noted. As Miller pointed out, killifishes do not possess a complete lateral line; therefore, it is customary to count scales in a lateral series starting from the shoulder girdle to the end of the hypural plate, ascertained by bending the caudal fin. Names for skeletal structures are those traditionally used in a description of teleost anatomy, as updated by Patterson (1975). Details of the gonopodium of poeciliid fishes are described using the terminology of Rosen and Bailey (1963). Patterns of head scales and sensory pores and canals are described according to the conventions established by Hoedeman (1958) and Gosline (1949), respectively, to facilitate comparisons among the results of this and other studies. Estimates of number of species in groups currently classified in the Cyprinodontidae are from Lazara (1979) unless otherwise stated. Specimens examined and their catalog numbers appear in the systematic section following each generic and family diagnosis. Catalog numbers followed by an asterisk (*) indicate lots from which counterstained specimens were prepared; those followed by a cross (+) are lots from which solely alizarin-stained preparations had been made. The number of such specimens prepared is given in both cases as a fraction of the total specimens in the lot (e.g., 4 out of 20 is given as 4/20). Catalog numbers with no designation are of alcohol lots from which no special preparations were made. ABBREVIATIONS INSTITUTIONAL AMNH, American Museum of Natural History, New York ANSP, Academy of Natural Sciences, Philadelphia BMNH, British Museum (Natural History), London VOL. 168 CAS, California Academy of Sciences, San Francisco FMNH, Field Museum of Natural History, Chicago IU, Indiana University (now at California Academy of Sciences, San Francisco) MCSN, Museo Civico di Storia Naturale, Genova MCZ, Museum of Comparative Zoology, Cambridge MNHN, Museum National d'Histoire Naturelle, Paris SU, Stanford University (now at California Academy of Sciences, San Francisco) UMMZ, University of Michigan, Museum of Zoology, Ann Arbor USNM, National Museum of Natural History, Washington, DC. ZVC, Zoologia Vertebrados, de la Facultad Ciencias, Montevideo, Uruguay ANATOMICAL AC, anterior ceratohyal ALV, alveolar arm of premaxilla AMR, middle anal radial APL, autopalatine APR, proximal anal radial AR 1, anal ray 1 ART, articular ASC, ascending process of premaxilla BOC, basioccipital BR, branchiostegal ray CL, cleithrum COR, coracoid DEN, dentary DHH, dorsal hypohyal DMX, dorsal process of maxilla DPR, proximal dorsal radial DR1, first dorsal ray E 1-4, epibranchial 1-4 END, endopterygoid EP, epural EPL, epipleural rib EPO, epiotic EPO-PRO, epiotic processes EXO, exoccipital FRO, frontal HY 1-5, hypural 1-5 HYO, hyomandibula HYP, hypaxial musculature ICARM, infracarnalis medius ICARP, infracarnalis posterior IF, inferior pharyngeals IH, interhyal INCLA, inclinatores anales IS, ischial process 1981 PARENTI: CYPRINODONTIFORM FISHES K, kidney MAX, maxilla MDN, medial process of dentary MET, metapterygoid NA-1, neural arch 1 NL, nasal NS, neural spine PAR, parietal PAS, parasphenoid PB 1-3, pharyngobranchial 1-3 PC, posterior ceratohyal PCL1, 3, postcleithrum 1, 3 PHY, parhypural PL, pleural rib PMX, premaxilla POP, preopercle PRO, prootic PSP, pseudophallus PTT, posttemporal PU2, preural centrum 2 349 QUA, quadrate RAD, radials RC, rostral cartilage RET, retroarticular SAC, subautopalatine cartilage SCL, supracleithrum SOC, supraoccipital SOC-PRO, supraoccipital processes SPH, sphenotic SYM, symplectic T, testis TPB 1-4, pharyngobranchial toothplate 1-4 UB, urinary bladder UG, urogenital opening UN, uroneural UR, ureter VHH, ventral hypohyal VMX, ventral process of maxilla VO, vomer OVERVIEW OF PAST INTERNAL CLASSIFICATIONS OF CYPRINODONTIFORM FISHES Cyprinodontiform fishes, as a whole or in part, have been the subject of various revisionary studies, many of which have included a formal reclassification. Together, these works may be characterized as studies in recognition of diversity, rather than in elucidation of interrelationships. From Garman's (1895) summary of all members to Sethi's (1960) discussion of primarily the oviparous cyprinodontids, reclassifications have focused on the description of differences, rather than of the derived similarities, among groups of cyprinodontiforms. The ma-, jor classifications are summarized in table 3. Garman attempted a synopsis of cyprinodontiforms; however, he had included the characin Neolebias and the cyprinoid Fundulichthys in the group. As a result, his diagnosis was general enough to apply to almost any group of soft-rayed fishes with a single dorsal fin. Aside from these shortcomings, however, Carman's summary of cyprinodontiform subgroups has remained little changed in subsequent reclassifications. Garman divided the Cyprinodontes Gill (1865) (=Cyprinodontoidea of Rosen, 1964) into eight subfamilies. The known genera of goodeid fishes were included in the subfamily Cyprinodontinae, along with Neolebias. The poeciliid fishes were the sole constituents of the subfamilies Poeciliinae and Belonesocinae. Jenynsia and Anableps were each placed in their own subfamilies, Jenynsiinae and Anablepinae, respectively. Cyprinodontids were divided among the Cyprinodontinae and the remaining three subfamilies. The monotypic Orestiasinae contained the genus Orestias; the known cyprinodontines constituted the remainder of the Cyprinodontinae. Carman's Nothobranchiinae consisted of two African genera, Haplochilichthys and the aplocheiloid Nothobranchius. The remainder of the cyprinodontids, including the fundulines and South American aplocheiloids and Fundulichthys, together formed the subfamily Haplochilinae. The name of the type genus, Haplochilus, was a corrected spelling of Aplocheilus McClelland. This spelling change was not valid under the International Code of Zoological Nomenclature and was not used in subsequent revisions. However, the names Haplochilus and Haplochilichthys persist as identifications in ca 73 33 4> = 73 ta 2 = ° _ .a ^ = s 12 o •- u u u '5 >, a u < •o o N o B o Xi § D. & o. ° ft £ fc < O < < O S a «> o •3 J3 S e C ca S J3 S J3 CB ca 8 .2 3 c ca X o _ .5 =a .S ft"% 13 •a c !•§ |ft I § .> £• o a o tt ai O a. a. 0 u eo •ee C8 '5 73 O o O a. 3<~ § .S .S .2 i 1.11 5 '5 '5 -C .c oo oo I & § <2 * u -g u •a 'S •a o o O h O c •c ft >> CO 73 73 cB 73 C o V O 73 73 O O C •a 73 a o e "C ft >> u c c C8 S S3 O 73 O C •c ft >> u s c e < 8 c -S 73 :s 5. I C8 73 ##^ u O 0- < u e o o c ca •c ft ««c V as .g 3 3 73 73 C C 3 3 a. ft. o o £ o O a. <u — 2 •2 CB 2 | <u o <o ca ca C3 a e a a - 2 c = 2 •Q, 8 * 3 - %ft 73 g ou O Cw uov 73 O ."S 'E c 3 >> •^ O u 2 o ca ft _ o ,. ca •a c8 £ u J O o c 3 73 C u ca 8 &a u o 73 O £ § •£ O X z ca c u CB e 4> O PH 8 " = Z >> ca c ££ < V '5 o c '3 _o 2 13 O k m ca a e 'ft c JO >> ca c c < 1981 PARENTI: CYPRINODONTIFORM FISHES many collections. The name Haplochilus has been used to refer to fishes in such genera as Epiplatys, Pachypanchax, Fundulus, Aphyosemion, and Oxyzygonectes as well as Aplocheilus. Similarly, Haplochilichthys has been used to reference fishes in any of the procatopine genera, not solely Aplocheilichthys. In Regan's (1911) reclassification of the order Microcyprini, the two subfamilies of poeciliids were united into the subfamily Poeciliinae. He separated the goodeids from the cyprinodontines and placed them in their own subfamily, the Characodontinae. Thus, with Regan's work, the four viviparous families were separated from the oviparous cyprinodontids. The precedent for treating oviparous and viviparous cyprinodontiforms separately in systematic revisions was established, and has remained virtually unchallenged until the present study. The only concrete statements regarding the interrelationships of the five currently recognized families were made by Hubbs (1924) and Regan (1929), and these were in direct opposition. Hubbs grouped Jordan's Fitzroyidae (which he corrected to Jenynsiidae) and Anablepidae together into the family Anablepidae, and Jordan's Characodontidae and Goodeidae into the family Goodeidae. His action concerning the Goodeidae has remained unchallenged. However, the grouping together of Anableps and Jenynsia was rejected by Regan (1929) who placed Anableps in the Poeciliidae. Myers (1931) effectively avoided the problem by reverting to the placement of the two genera in monotypic families. Most recently, Miller (1979) has criticized the inferred relationship of the two, and considered Anableps to be more closely related to the Poeciliidae than to any other group of cyprinodontiforms. Since Hubbs s work, reclassification of cyprinodontiforms has focused on the division of the Cyprinodontidae (encompassing the Cyprinodontinae and Fundulinae) into a variety of subgroups which have undergone elevations or reductions in rank. Myers (1931) treated the oviparous cyprinodontids and Oryzias. His definition of the 351 family consisted of the following characters: occipital condyles present on both basioccipitals and exoccipitals, no modifications of the anal fin into an intromittent organ, oviparous, premaxillaries distinct from maxillaries and protractile or not, and never more than 65 scales in a lateral series. The definition, however, is not consistent with the distribution of characters among all cyprinodontids. More importantly, all the characteristics are primitive, and are found generally among teleost fishes. Thus, Myers (1931) effectively described the fishes of the family Cyprinodontidae as cyprinodontiforms that lack the prominent specializations of the viviparous groups. That is, no derived [unique] characters of the family were given to unambiguously define it as a monophyletic group. Myers (1931) divided the family into four subfamilies, and further divided the largest, the Fundulinae, into four tribes. The tribe Fundulini was restricted to the North American fundulines and their presumed relatives. The tribe Rivulini is coextensive with the aplocheiloids as discussed throughout the present paper. Together, the tribe Aplocheilichthyini and subfamily Lamprichthyinae are coextensive with the procatopines. Similarly, the subfamily Orestiatinae consisted solely of the genus Orestias. Eigenmann (1920) suggested that the North American Empetrichthys was the closest relative of Orestias based on the fact that both lack pelvic fins and fin supports, and have fleshy bases of the dorsal and anal fins. As a result, the two genera constituted the membership of the Orestiidae and Orestiinae of Jordan (1923) and Hubbs (1924), respectively. Myers (1931), however, supported the idea that Empetrichthys was more closely related to Fundulus, and placed it in his tribe Fundulini where it has remained until this study. The final tribe, Aplocheilini Bleeker, consisted of a single genus, the ricefish Oryzias which Myers and contemporaries referred to as Aplocheilus. Fishes now commonly referred to the genus Aplocheilus were re- 352 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY ferred to as members of the genus Panchax, in the Rivulini. The confusion over the available names for these genera, all of the IndoMalaysian region, was eliminated by Smith (1938) who demonstrated that Panchax was an objective synonym of Aplocheilus, and that Oryzias was the proper name for the ricefish. However, there was enough time for the name Panchax to become established as a common name for most of the aplocheiloid fishes, and it is still casually employed. Earlier, Myers (1924c) pointed out that Fundulichthys Bleeker, a name applied to a specimen known only from an illustration, referred to a cyprinoid. Berg (1940) substituted the name Cyprinodontiformes for the Microcyprini. He divided the order into two superfamilies, an oviparous Cyprinodontoidea including the families Cyprinodontidae and Adrianichthyidae, and a viviparous Poeciloidea, including the Goodeidae, Poeciliidae, Jenynsiidae, and Anablepidae. The adrianichthyid fishes, comprising the genera Adrianichthys and Xenopoecilus, were associated with the Cyprinodontiformes since their transfer from the Beloniformes (=Exocoetoidei) by Weber and de Beaufort (1922). However, because of its poor representation in collections (the monotypic Adrianichthys known only until recently from the single holotype), the family has gone virtually ignored in revisions. Rosen (1964) placed it along with Oryzias, which he elevated to family rank, and the Horaich thy idae, in the superfamily Adrianichthyoidea. Myers (1955), again treating the oviparous killifishes and Oryzias, elevated each of his tribes of 1931 to subfamily rank. He acknowledged the correction of the use of the name Aplocheilus for Oryzias by elevating the rank of the tribe Aplocheilini to the subfamily Oryziatinae. A new subfamily Pantanodontinae appeared to include the single genus and species Pantanodon podoxys described in Myers (1955) by name only. An eighth subfamily, the Lamprichthyinae, was omitted from the list presumably inadvertently. The oviparous killifishes were treated again by Sethi (1960) who elevated the ranks VOL. 168 of these groups yet again. The oviparous cyprinodontids were classified in six families. Oryzias, included in the study, was also placed in its own family. The procatopines and Lamprichthys were grouped together in the family Aplocheilichthyidae. Aphanius and its allies were removed from the subfamily Cyprinodontinae of Myers and placed in their own family, the Aphaniidae. Thus the Cyprinodontidae of Sethi consisted solely of Cyprinodon and its New World relatives. The Pantanodontinae was inexplicably omitted from Sethi's study, as were other unique cyprinodontid genera such as Oxyzygonectes, Chriopeoides, Cubanichthys, and Rivulichthys. Also in 1970, Roberts created yet another subfamily, the Fluviphylacinae, to include a single genus and species, Fluviphylax pygmaeus (Myers and Carvalho). This genus was also disregarded by Sethi. Therefore, the most comprehensive and also most widely accepted classification of oviparous killifishes is that listed in table 2. The eight subfamilies are grouped together in a single family, the Cyprinodontidae. The relationship of the subfamilies to each other and to the families of viviparous killifishes has never been formally treated before this study. In fact, Sethi's grouping of the Lamprichthyinae and procatopines represented the only alignment of subgroups of cyprinodontids since Garman (1895). Revisions of the viviparous killifishes have focused on the interrelationships of included genera or species, but have presented no more than informal remarks about the relationship of the considered family to another viviparous family or to the cyprinodontids. Hubbs and Turner (1939), in a revision of the Goodeidae, emphasized the structural differences between the family and other cyprinodontiforms. More recently, Miller and Fitzsimons (1971) proposed several defining characters of the family (some of which are found among the oviparous cyprinodonts), and synonymies of several genera, yet made no statement as to the relationship of the goodeids to other cyprinodontiforms. Rosen and Bailey (1963) provided a comprehensive discussion of the relationships of 1981 PARENTI: CYPRINODONTIFORM FISHES the family Poeciliidae to other cyprinodontiforms, yet came to no firm conclusions. They suggested that the closest relative of the poeciliids is perhaps another viviparous killifish; however, they stressed the fact that modifications for viviparity among the four families were not alike, except for the similar gonopodial structure of Jenynsia and Anableps, the intrafollicular development of poeciliids and Anableps, and the presence of trophic processes in goodeids and Jenynsia. In addition, Rosen and Bailey maintained that poeciliids were more like some oviparous than viviparous cyprinodontiforms in general body form and osteology; however, they did not suggest a group of cyprinodontids which could possibly be a close relative of the poeciliids. Rosen and Bailey classified the Poeciliidae in three subfamilies: the Tomeurinae Eigenmann, containing just one genus and species, Tomeurus gracilis; the Xenodexiinae Hubbs, also containing just one genus and species, Xenodexia ctenolepis; and, the Poeciliinae, containing all other members of the family. Structurally, Tomeurus is much like other poeciliids; however, it diverges strongly in the elaborate modifications of the gonopodium, and also in that it is the only oviparous poeciliid. (Internal fertilization results in the laying of a fertilized egg.) These differences, and also the remarkable similarity of the form of the gonopodium to that of the oviparous Horaichthys, an adrianichthyoid, led Nikol'skii (1954) to propose that the two genera, each classified in its own family, be united into one superfamily, the Tomeuroidea. Kulkarni (1948) however, suggested that on the basis of overall osteological similarity, Horaichthys was closer to Oryzias than Tomeurus. This conclusion has been supported by all recent workers on both poeciliids and adrianichthyoids (e.g., Rosen, 1964, 1973a; Rosen and Parent!, MS). The alignment of Tomeurus with the poeciliids is also supported by the present study. Miller (1979), in discussing the relationships of Anableps dowi, supported the alignment of Anableps with the poeciliids, citing as evidence of close relationship the retention of the embryos in modified ovarian fol- 353 licles during the entire developmental period, and the fact that the first three anal rays are unbranched in both. However, he made no formal reclassification and maintained that Anableps was so distinct that it should remain in its own family. Alignments of one subfamily of cyprinodontids to another have been suggested by a number of workers (e.g., Ahl, 1924, 1928; Hoedeman and Bronner, 1951; Miller, 1955a; Uyeno and Miller, 1962). Ahl (1924, 1928) considered solely the African cyprinodontid genera of the Rivulinae and Procatopodinae, which he believed formed a natural group. Hoedeman and Bronner (1951, p. 1) made recommendations for the alteration of cyprinodontiform classification. They constructed the tribe Profundulidi to include the Old World genera Kosswigichthys and Valencia, and the North and Central American fundulines Profundulus and Adinia. The tribe was regarded as unnatural by Miller (1955a) who suggested that the Fundulinae and Cyprinodontinae be merged. Each of these four genera is regarded as a member of a different subgroup in the phylogenetic analysis of the present study. Miller (1956), in describing Cualac, a new genus of cyprinodontids, stated that it was intermediate between the Fundulinae and North American Cyprinodontinae. He reiterated his previous suggestion that the two groups together comprise the subfamily Cyprinodontinae, claiming that they were probably artificially separated on the basis of dental morphology. Later, however, Uyeno and Miller (1962) supported Sethi's conclusion that the two groups remain separated in a classification. They listed a series of characters from Sethi (1960) by which the fundulines could be distinguished from the cyprinodontines: the presence of parietals; neural arches of the first vertebra not fused to skull, and therefore, taking no part in the articulation of the vertebral column to the skull; the presence of occipital condyles; and, the lack of a gap between the first and second vertebrae. These characters are, however, as Myers's defining characters of the Cyprinodontidae, 354 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY primitive for atherinomorph fishes. Thus, it is not just the superfamily and families which are poorly defined, but the subfamilies as well that lack precise definitions and, therefore, need to be supported or refuted as monophyletic groups of genera. Foster (1967) presented a summary of his conclusions regarding atherinomorph phylogeny in a branching diagram. The Pantanodontinae was removed from the cyprinodontiforms and placed as a close relative of the adrianichthyoids. Within the cyprinodontids, he recognized six subfamilies (the Fluviphylacinae Roberts obviously omitted), concluding that the Rivulinae and Orestiatinae were sister groups primitive to other cyprinodontiforms, excluding Pantanodon. The Procatopodinae and Lamprichthyinae were similarly depicted as sister groups primitive to the four viviparous families and the Fundulinae and Cyprinodontinae. Foster considered the fundulines to be most closely related to the Anablepidae (presumably including Jenynsia) which together formed the sister group of the poeciliids. This subgroup, in turn, was assessed as being most closely related to the Cyprinodontinae and Goodeidae, represented as sister groups. This analysis was based on an assessment largely of overall similarity for a group of 16 or more characters, including those of the osteology, behavior and development. Foster presented no formal reclassification of the atherinomorph fishes. He therefore retained the subfamily rank for Pantanodon even though he considered it to be more VOL. 168 closely related to the adrianichthyoids. However, in spite of such inconsistencies related to the level at which he was approaching the problem, this work represented for the first time a precise although informal statement about the interrelationships of the four viviparous families and their relationship to the subfamilies of the Cyprinodontidae was presented. It is noteworthy also for including the implicit statement of the nonmonophyletic nature of the family Cyprinodontidae. However, Foster's treatment, like the others, grouped the families and subfamilies mainly on overall similarity without regard to the primitive or derived nature of characters. The family Cyprinodontidae and also the cyprinodontiform fishes as a whole, as indicated by Foster's removal of Pantanodon, are left to be formally defined as monophyletic groups or to have their monophyly refuted. Similarly, the viviparous families remain to be unambiguously defined on the basis of derived characters unique to them and not found in other cyprinodontiforms. Thus, to accomplish the stated objectives of this study, the four viviparous families are treated as four more genera of cyprinodontiform fishes, the monophyly of each being supported or rejected. Furthermore, the reclassification of cyprinodontiforms presented in this study is based on an attempt to define the major groups of genera and represent their hierarchical relationship, rather than to obscure such relationship by basing the rank of a taxon on subjective criteria of uniqueness. DERIVED CHARACTERS OF CYPRINODONTIFORMS The Cyprinodontoidea has been recognized since the definition of the family Cyprinodontidae by Gill (1865). However, the failure by him and subsequent workers to define the Cyprinodontidae rigorously has resulted in the uncritical inclusion with them of the ricefish genus Oryzias until Rosen (1964) placed it in its own family and suggested its close relationship to the adrian- ichthyoid fishes. Inadequate definition of the Cyprinodontidae also is responsible for the unsupported placement of Pantanodon with adrianichthyoids by Foster (1967). Previous workers attempting to define the superfamily (e.g., Regan, 1911; Hubbs, 1924) have included characters either primitive for atherinomorph fishes or shared by a number of its subgroups. As a result, the superfamily 1981 PARENTI: CYPRINODONTIFORM FISHES has never been unambiguously denned as a monophyletic group (in the sense of Hennig, 1966). The current study has revealed that all fishes of the superfamily Cyprinodontoidea may be distinguished from all other teleost fishes by the following derived features. CAUDAL FIN: A series of derived characters within the caudal fin is found relatively unmodified in all cyprinodontiforms. Externally, the fin is rounded or truncate, although in males of several aplocheiloid genera, procatopines, and the South American Orestias, there are often extensions of the dorsal and ventral caudal rays. In no case are there incipient lobes; although, Miller (1979) reports that in males of Anableps microlepis, lower caudal rays are often grouped together forming a lobelike structure. Branched caudal rays typically number eight or more. Internally, the supports of the caudal fin are symmetrical (fig. 2E). There are two hypural plates, one above and one below corresponding to fused hypurals 3, 4, and 5 and 1 and 2, respectively. In some species of Epiplatys and Aplocheilus, the upper hypural plate is divided in two, apparently representing the unfused hypurals 4 and 5 (fig. 2D). Within the cyprinodontiforms, fusion of the hypural plates into a so-called hypural fan (following the terminology of Rosen, 1964) occurs within several monophyletic groups of genera (e.g., fig. 2F). There is just one epural which mirrors in shape and position the autogenous parhypural. There are no separate ural centra. The hypochordal musculature is also absent (Rosen, 1964). This formation of a symmetrical caudal fin in unique among teleost fishes. The esocoid Umbra limi has a caudal fin which is externally unlobed and rounded. Yet, an examination of the internal structure reveals that the external symmetry is effected by a complex of two epurals, one uroneural, five unfused hypurals, and two separate ural centra, the second of which is dorsally offset to the first (fig. 2A). In addition, there are fewer than eight branched caudal rays. 355 Among other groups of atherinomorphs, there are lobate caudal fins exclusively. The atherinoid Menidia beryllina (fig. 2B) has a caudal skeleton which is asymmetrical in having two epurals which are relatively smaller than the opposing parhypural. The divided hypural plate has a larger dorsal segment. Oryzias (fig. 2C) has an asymmetrical caudal fin support in which two small epurals oppose the single, large parhypural. The hypural plate is divided into subequal dorsal and ventral segments. FIRST PLEURAL RIB: Typically among the atherinomorph fishes, the first pleural rib arises on the parapophysis of the third vertebra. Occasionally the rib is borne on the parapophysis of the fourth vertebra in males of the genus Ceratostethus, and both males and females of the genus Gulaphallus, both phallostethoid fishes in the suborder Atherinoidei (Roberts, 1971). Within the cyprinodontiforms, the first pleural rib is borne on the parapophysis of the second vertebra. In the funduline genus Adinia there is a pleural rib on the parapophysis of the first vertebra; this condition is considered to be apomorphic for the genus. Rosen (1964) followed Myers (1928a) in stating that the first pleural rib of phallostethoids arose on the parapophysis of the second vertebra, and therefore suggested a close affinity between the phallostethoids and cyprinodontiforms. However, Roberts (1971) has shown this to be a misidentification of the state in phallostethoids. Examination of several genera of phallostethoid fishes as part of this and other studies has supported Roberts' contention. Therefore, the first rib arising on the parapophysis of the second vertebra is a characteristic unique to cyprinodontiforms. JAW STRUCTURE: The protrusible upper jaw of the atherinomorph fishes differs from that of other acanthopterygians in the lack of a ball and socket joint between the autopalatine and the maxilla, and the absence of crossed rostral ligaments (Rosen, 1964). The absence of a ball and socket joint prevents the premaxillaries from being locked in the protruded position by the autopala- 356 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 1981 PARENTI: CYPRINODONTIFORM FISHES tines upon the opening of the mouth. However, the premaxillaries may still be held protracted by contraction of the superficial division of the adductor mandibulae (Al) (Alexander, 1967a, 1967b), which inserts on the middle of the distal arm of the maxilla. Crossed rostral ligaments run from the left autopalatine and the right autopalatine to the heads of the right and left premaxillaries, respectively. These ligaments, along with a pair of ethmomaxillary ligaments, typify the mechanism of the protrusible upper jaw of acanthopterygians (Schaeffer and Rosen, 1961). The atherinomorphs lack crossed rostral ligaments; thus, the forward movement of the premaxillaries is limited by contact with the maxilla. Among the atherinomorphs, Alexander (1967b) reports the presence of an ethmomaxillary ligament in the atherinoids Atherina and Melanotaenia, and the aplocheiloid Aplocheilus. He notes its absence in Fundulus and the poeciliid Xiphophorus. The mechanism of protrusion of the upper jaw of cyprinodontiforms has been described in detail by Rosen (1964) and Alexander (1967a, 1967b). It is characterized by a twopart alveolar process of the premaxillaries. A distal part of the process is joined to an offset proximal part of the process, thus creating a wide bow, as illustrated in the aplocheiloid Austrofundulus (fig. 3A). In all cyprinodontoids, the process is primitively S-shaped (fig. 3B) as a result of the distal part of the process being strongly indented posteriorly. At the posterior tip of the ascending processes of the premaxillaries is a large, free rostral cartilage. Among acanthopterygians, the rostral cartilage is typically firmly attached to the ventral surface of the tips of the ascending processes by connective tissue fibers, and, in addition, is sometimes wrapped around the tips of the processes. The median process of the maxillary head is 357 bound by connective tissue fibers to the anterior end of the rostral cartilage. The presumed function of the rostral cartilage is to prevent the independent movement of the premaxillaries, and also to prevent their rolling off the cranium when the mouth is opened (Alexander, 1967a, 1967b). However, since the maxilla functions as a brace during the forward movement of the premaxillaries in atherinomorphs, and the rostral cartilage is not present in all cyprinodontiforms, it appears that the rostral cartilage serves mainly as a restrainer of the independent movement of the premaxillaries. Primitively, within most acanthopterygians and most atherinomorphs, in addition to being bound to the ascending processes of the premaxillaries, the rostral cartilage is attached to the median processes of the maxillary heads by connective tissue fibers. Alexander (1967b) reports that in atherinoids there is also often an attachment of the cartilage to the vomer and to the ethmoid cartilage. Among cyprinodontiforms, variability in the states of degree of attachment of the rostral cartilage, size of the cartilage, its presence or absence, length of the ascending processes of the premaxillaries and presence or absence of the ethmomaxillary ligaments allows for the description of transition series of these characters and the delimitation of at least three distinct mouth forms within the group. (Sethi, 1960, referred to both the rostral cartilage and the mesethmoid as the mesethmoid; therefore, his descriptions of states of the mesethmoid are unreliable since they refer to either one or the other.) The aplocheiloids share some similar upper jaw characteristics with the atherinoids. These are the presence of an ethmomaxillary ligament, the presence of a ligament from the internal hooks of the maxillaries to the rostral cartilage, and the presence of a meniscus between the premaxilla and maxilla. These FIG. 2. Diagrammatic representation of the caudal skeleton of A. Umbra limi (after Rosen, 1974); B. Menidia heryllina; C. Oryzias latipes (after Rosen, 1964); D. Aplocheilus panchax; E. Aphyosemion gardneri; F. Fundulus heteroclitus. 358 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 ASC ALV B FIG. 3. Diagrammatic representation of premaxillary alveolar arm, lateral view, of A. Austrofundulus transilis; B. Profundulus punctatus. Arrows point to areas indented in cyprinodontoids to form S-shaped arm. ligaments and the meniscus are absent in all cyprinodontoids. The rostral cartilage in cyprinodontiforms, as stated, is free and not wrapped around the ascending processes of the premaxillaries as it is in most acanthopterygians. This permits the movement of the rostral cartilage relative to, rather than with, the ascending processes. In all aplocheiloids, the rostral cartilage is a large, disc-shaped element lying beneath the flat and broad ascending processes (fig. 4). In one cyprinodontoid genus, Profundulus (fig. 5B), the ascending processes are broad and the cartilage large, yet somewhat reduced relative to the aplocheiloid condition. It is further reduced in the fundulines (fig. 5C), and Valencia (fig. 5D). Among remaining cyprinodontiforms (figs. 35, 39), the cartilage is present as a minute disc, or absent, whereas the ascending processes are shortened, or nearly absent as in Anableps, and held together by connective tissue fibers. Thus, within the cyprinodontiforms there are three basic forms of the upper jaw and jaw suspension. The first, and apparently most primitive, is that of the aplocheiloids. The rostral cartilage is large and firmly attached to the broad premaxillary ascending processes. There are ligamentous attachments of the head of the maxilla to the rostral cartilage and of the maxilla to the ethmoid. The large size of the rostral cartilage and the presence of ligaments are assessed as primitive by comparison with an outgroup, the atherinoids, and to the percomorph fishes, in which these states are present. There are three distinct states of the premaxillary ascending processes in cyprinodontiforms: flat and broad in aplocheiloids and Profundulus, long and narrow in Fundulus and related genera, and short and 1981 PARENTI: CYPRINODONTIFORM FISHES 359 opal FIG. 4. Diagrammatic representation of the upper jaw in A. Aplocheilus panchax; B. Pachypanchax playfairi; C. Aphyosemion petersi; D. Cynolehias whitei. pointed or triangular in all remaining cyprinodontiforms. Exceptions occur in Anableps, as mentioned, and in Oxyzygonectes in which the processes are enlarged. These exceptions are discussed in the phylogenetic analyses and generic diagnoses. The size of the ascending processes in other atherinomorphs and acanthopterygians is variable; however, the general or most common state is for the processes to be long and narrow. This would suggest that the transition series for ascending processes is from a primitive state of long and narrow to flat and broad in one lineage, and to short and narrow in another. However, information from other systems clearly indicates that Profundulus is more closely related to cyprinodontoids than to aplocheiloids. Therefore, the flat and broad ascending processes are most parsimoniously assessed as the primitive state within the cyprinodontiforms. The short and pointed or triangular processes coupled with an extremely reduced, or in some cases absent, rostral cartilage, are defining characters of a large group of cyprinodontiforms encompassing the poeciliids, goodeids, Jenynsia, Anableps, Oxyzygonectes, the cyprinodontines, Orestias, Cubanichthys, Chriopeoides, and the procatopines. In addition, in this group as well as in the 360 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 5. Diagrammatic representation of the upper jaw in A. Rivulus harti; B. Profundulus punctatus; C. Fundulus diaphanus; D. Valencia hispanica. Mediterranean Valencia, there is a dorsal extension of the maxilla over the ascending processes, which forms a cuplike process with the ventral extension. The short ascending premaxillary processes slide in and out of this maxillary cup. In primitive cyprinodontiforms the twisted maxilla extends ventrally under the premaxilla but not dorsally. GILL ARCHES: Interarcual cartilages are found among the percomorph fishes (Rosen and Greenwood, 1976). The general condition is that found in the atherinoid genus Melanotaenia. A rod of cartilage extends between an uncinate process of the first epi- branchial and the second pharyngobranchial. In contrast, the uncinate process is lacking in all cyprinodontiforms. The cartilage is subequal to the epibranchials in aplocheiloids, and extends between the posterior base of the first epibranchial and the second pharyngobranchial (fig. 6A). In cyprinodontoids, the cartilage and the first epibranchial are both present in the same position; yet, they are reduced to approximately half their length relative to the size of these elements in the aplocheiloids (fig. 6B). The interarcual cartilage is present in all cyprinodontiforms except groups of procatopine and of aplocheiloid genera in which PARENTI: CYPRINODONTIFORM FISHES 1981 361 B PB2 TPB4 FIG. 6. Diagrammatic representation of dorsal gill arches, ventral view, A. Austrofundulus transilis; B. Profundulus punctatus. other elements of the dorsal gill arches are present in a typical arrangement, and the interarcual cartilage is assumed to have been lost. PECTORAL GIRDLE: The cyprinodontiform shoulder girdle typically has a first postcleithrum which is large and scale-shaped (figs. 7C, 8A, B). This is in contrast to the condition in atherinoids and exocoetoids in which the first postcleithrum is typically a slender bone (figs. 7A, B). Among cyprinodontiforms the first postcleithrum is absent, and therefore presumed lost, in poeciliids, most procatopines, Leptolucania, Orestias, Rivulus and its South American relatives and one species of Anableps, A. dowi. There is another postcleithrum situated medial to the scapula and radials, and extending ventrally beyond the coracoid. This long, slender element is present in cyprinodontiforms. Rosen and Bailey (1963) interpreted this as a "secondary postcleithrum" in the poeciliids without discussing its homology to the second postcleithrum of lower teleosts. Roberts (1970) referred to it as the "first rib?" in his description of the osteology of the South American Fluviphylax pygmaeus. Sethi (1960) refers to the element only in an illustration in which it is labeled "PCL [Postcleithrum] 2." Weitzman (1962) illustrated the shoulder girdle of the characin Brycon meeki which has three postcleithral elements. The third postcleithrum is comparable in shape and position to the so-called secondary postcleithrum of cyprinodontiforms. Therefore, I interpret these structures as homologues and conclude that the general condition for a cyprinodontiform is to have a large first postcleithrum, and a narrow third postcleithrum, with the second postcleithrum always lacking. Also, the lowset pectoral fin, effected by a ventral position of the radials (e.g., fig. 7A, D), primitively distinguishes the cyprinodontiforms from all other atherinomorphs which have highset pectoral fins and more dorsally situated radials (fig. 7A, B). 362 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 pc-3 FIG. 7. Diagrammatic representation of left shoulder girdle of A. Oryzias javanicus; B. Menidia menidia; C. Aplocheilus panchax; D. Rivulus harti. Cartilage is stippled. Highset pectoral fins within cyprinodontiforms occur in the poeciliids and procatopines (fig. 8C, D), and are interpreted as being secondarily derived. This is the most parsimonious interpretation of the condition of the shoulder girdle based on (1) a series of uniquely derived characters that indicate a close relationship between the poeciliids and procatopines, and (2) a series of derived characters that indicate that these two are 1981 PARENTI: CYPRINODONTIFORM FISHES 363 D FIG. 8. Diagrammatic representation of left shoulder girdle of A. Profundulus punctatus; B. Cyprinodon variegatus; C. Procatopus gracilis; D. Tomeurus gracilis. Cartilage is stippled. together more closely related to one group of cyprinodontiforms with lowset pectorals than to another, which also possesses lowset pectorals. 364 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY BREEDING AND DEVELOPMENT: Eggs of the oviparous atherinomorph fishes are distinguished by having long, chorionic filaments by which they attach to the spawning substrate, and conspicuous oil droplets (inferred to be secondarily lost in the suborder Exocoetoidei) (Foster, 1967). Cyprinodontoid eggs, in turn, are distinguished by their relatively longer development time and thickened chorion, the outermost egg membrane. Thus, the egg of a typical oviparous cyprinodontiform may be characterized as large (some over 2.0 mm in diameter), containing several oil droplets and surrounded by a thick, filamentous chorion. A typical nonannual cyprinodont egg has a development time of 12 days or longer. Eggs of the atherinoid Bedotia geayi were observed in the laboratory to have a development time of about nine days. Development time is up to six months for eggs of fishes in true annual genera such as the South American Cynolebias and the African Nothobranchius. The thick chorion permits survival under conditions of desiccation during an extended development period, such as those typical of Fundulus. The annual habit (first reported by Myers, 1942; then described in detail by Peters, 1965, and Wourms, 1963, 1964, 1967, 1972a, 1972b, and 1972c) is exhibited by a minority of the aplocheiloid species of tropical South America and Africa. Adults live for no more than one rainy season during which time they spawn. The eggs enter diapause and survive the dry period buried in the substrate. The fertilized eggs normally hatch at the onset of the subsequent rainy season; however, they have been known to survive dry periods of several years. Foster (1967) reports that in the atherinoid Melanotaenia and in Oryzias spawning normally takes place without direct contact with a substrate. In addition, a large number of eggs are extruded at once. In contrast, all killifishes, with a few possible exceptions, spawn in contact with a substrate, and eggs are extruded one at a time. Spawning in a typical annual occurs daily from the onset of VOL. 168 sexual maturity, which occurs as early as four to six weeks, until death. Annual fish eggs enter three diapause stages prior to hatching (Wourms, 1972a). The first, termed Diapause I, occurs during the pre-embryonic stage. The cells of the blastodisc separate and disperse around the surface of the yolk sphere. Arrest lasts until the cells reaggregate to form the embryonic shield when the anterior-posterior axis of the embryo is established for the first time. Diapause II occurs during the mid-somite stage about the time of formation of the heart tube. Diapause III occurs just prior to hatching. The embryo is fully formed and capable of hatching, yet does not. The embryo remains quiescent; its heart beat slows down and the characteristic turning of the embryo and associated beating of the pectoral fins within the chorion are slowed or cease. The duration of each of the diapause stages is controlled either by genetic or environmental factors, or an interplay of the two. Embryos in stage III have remained quiescent for more than six months (Wourms, 1964). Previous workers have considered the annual habit to be uniquely derived within the annual killifish genera therefore suggesting that these fishes form a monophyletic group. The present study disagrees with this conclusion for two reasons: (1) On the basis of anatomical characters, certain true annuals are assessed as being more closely related to nonannuals than they are to other true annuals. (2) All cyprinodontiforms have a prolonged development time, and within genera that are not closely related to the aplocheiloids, survival of eggs through periods of desiccation has been demonstrated. The first of these reasons is discussed fully in the phylogenetic analysis of cyprinodontiform genera. The second is given in support of the contention that the annual habit is no more than an exaggeration, due to extreme environmental fluctuations, of a capability of all cyprinodontiforms to survive stress that involves desiccation. Foster (1967, p. 538) summarized the habitats of killifishes as: "If any generalization could be made about the ecology of killifish- 1981 PARENTI: CYPRINODONTIFORM FISHES es, it is that they exploit niches mostly in ephemeral waters, places which are temporarily submerged by tides, floods from heavy rains, or similar causes." The ability of nonannual killifish embryos to survive desiccation has been reported for a number of species within the North American genera Fundulus and Cyprinodon. Harrington (1959) reported that populations of Fundulus confluentus in Florida have survived hatching delays of up to three months. Areas of the salt marsh habitat of this killifish are exposed to the air during the months of October through December. Also, Taylor, DiMichele and Leach (1977) observed that another estuarine Fundulus, F. heteroclitus, often spawns during the high night tide. Eggs are thus stranded in the substrate up to a week after expected hatching time; hatching is delayed until reimmersion occurs. In an effort to test the generality of the ability of cyprinodont eggs to survive desiccation, F. Douglas Martin (personal commun.) exposed eggs of Cyprinodon to the air and found that they can survive through hatching; although, on the average, they are less successful at hatching than Fundulus species. This may indicate that the ability to survive periods of desiccation is primitive for cyprinodontiforms and is lost in the more advanced genera such as Cyprinodon. The first two diapause stages have not been demonstrated in nonannuals; however, the ability to survive pre-hatching desicca- 365 tion in the nonannuals appears to be comparable to Diapause III. Turner (1966) reports that a collection of Pantanodon podoxys has been made in Africa from stagnant pools. The common cyprinodontiforms in the vicinity were two species of the annual Nothobranchius which were present in similar pools. No permanent body of fresh water was found that could be inferred to have originally formed the pools. Such circumstances suggest that Pantanodon may be an annual, and therefore that annualism among cyprinodontiforms is not restricted to the aplocheiloids, but is perhaps a general characteristic of those cyprinodontiforms which inhabit ephemeral waters. The use of a potential annual lifestyle as a defining character of the cyprinodontiforms is confounded by the fact that the atherinoid Leuresthes tenuis spawns in conjunction with the tidal water level fluctuations so that the eggs are incubated while exposed to the air (Clark, 1925). This suggests that the ability to survive desiccation is a derived character for the atherinomorph or some larger group of fishes. However, the generality of this condition and its concordant developmental alterations for other groups of fishes awaits further description. Therefore, the early and regular breeding habit and long developmental period, coupled with the ability to survive desiccation, is considered to describe a unique developmental pattern of cyprinodontiform fishes. PHYLOGENETIC ANALYSIS The monophyly of each of the five families of cyprinodontiform fishes has been tested. A preliminary examination revealed that each of the four viviparous families is monophyletic and can be unambiguously defined, although not with all of the characters previously used to define them. The Cyprinodontidae, however, as currently constituted, cannot be defined as a monophyletic group. The alternative hypoth- esis is that some oviparous cyprinodontiforms are more closely related to the viviparous cyprinodontiforms than they are to other oviparous forms. Therefore, the genera of cyprinodontid fishes are used as a basis for a cladogram of all cyprinodontiforms. The poeciliids, goodeids, Jenynsia, and Anableps are treated as additional genera incorporated into the overall scheme. 366 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 9. Cladogram of major groups of cyprinodontiforms. Derived characters: Node A: symmetrical caudal fin, externally rounded or truncate, internally with one epural opposing a similarly shaped parhypural, hypural plates symmetrical; alveolar arm of premaxilla; interarcual cartilage from base of first epibranchial to second pharyngobranchial; first pleural rib on parapophysis of second vertebra; lowset pectoral girdle with scale-shaped first postcleithrum; early breeding habit and long developmental period; Node B: attached orbital rim; cartilaginous mesethmoid; close-set pelvic fin supports; narrow and twisted lacrimal; broad anterior end of basihyal; tubular anterior naris; reduced cephalic sensory pore pattern; pigmentation pattern (see text); Node C: two ossified basibranchials; loss of the dorsal hypohyal; reduced interarcual cartilage; head of autopalatine offset to main axis and with posterior flange; anterior ventral extension of the autopalatine; loss of the metapterygoid; premaxilla with a posterior indentation of the alveolar arm; dentary expanded medially; loss of the first dorsal fin ray; loss of an ethmomaxillary ligament; loss of a ligament from the maxillaries to the rostral cartilage; loss of a meniscus from between the premaxilla and maxilla; Node D: premaxillary ascending processes narrow or greatly reduced in adults; rostral cartilage reduced or absent; inner arms of maxillaries not abutting the rostral cartilage; lateral ethmoid with reduced facet for articulation of autopalatine; Node E: maxilla with straight proximal arm; large dorsal process of the maxilla directed over the premaxillary ascending process; Node F: ascending processes of the premaxillaries short and narrow; dorsal processes of maxillaries rounded or reduced; nasal expanded medially (or secondarily reduced); Node G: lateral ethmoid expanded medially and lying perpendicular to the frontal; reduced autopterotic fossa; enlarged inclinators of the anal; Node H: maxilla with expanded distal arm (or secondarily reduced); parasphenoid with expanded anterior arm; dorsal process of maxilla with a distinct lateral indentation; elongate retroarticular; pouch created by scales surrounding urogenital opening of females. Node I: dorsal processes of the maxillaries expanded medially, nearly meeting in the midline and possessing a distinct groove; lateral arm of maxilla robust; toothplate of fourth pharyngobranchial reduced. Node J: uniserial outer teeth: second pharyngobranchial offset to third; parietal absent; Meekers cartilage expanded posteriorly; transverse processes of vertebrae reduced and cup-shaped. For defining characters and relationships within terminal taxa, see text and figures 20, 25, 75, 81, 83, 87, and 89. In this discussion of phylogenetic relationships, the vernacular names as summarized previously are used for suprageneric cate- gories. At the conclusion of the phylogenetic analysis, a reclassification is presented. New group names in the classification are used in 1981 PARENTI: CYPRINODONTIFORM FISHES the systematic account and biogeographic analysis which follow. The results of the phylogenetic analysis are best presented in a cladogram (fig. 9). Limits and definitions of the recognized genera are presented in this discussion and formally in the systematic accounts. The cladogram is a hierarchic representation of the relationships among genera and suprageneric categories which are being proposed. The representation is of the most parsimonious distribution of derived characters and character states. Hypothesized convergences are discussed along with the proposed derived characters. The characters for the most inclusive node are the derived characters of the cyprinodontiforms (Group A of fig. 9) discussed in 367 the previous section. These are the unique formation of a symmetrical caudal fin; the first pleural rib arising on the parapophysis of the second vertebra, rather than on that of the third; a derived type of protrusible jaw; a unique form and position of the interarcual cartilage; the lowset pectoral fins with a large, scale-shaped first postcleithrum; and a unique pattern of breeding and development. The cyprinodontiforms are readily divided into two subgroups, the currently recognized cyprinodontid subfamily Rivulinae (the aplocheiloids), and all other cyprinodontiforms, termed the cyprinodontoids. Since these groups are both large and quite distinct, their interrelationships are discussed separately. APLOCHEILOIDS (GROUP B) The* aplocheiloid killifishes comprise over 500 species in 44 nominal genera and subgenera. Within the aplocheiloids, there are two groups of genera, the Old World aplocheiloids comprising Epiplatys, Aplocheilus, Pachypanchax, Nothobranchius, Aphyosemion and their included subgenera; and, the New World or Neotropical aplocheiloids comprising the genera Rivulus, Trigonectes, Rivulichthys, Rachovia, Pterolebias, Simpsonichthys, Campellolebias, Cynolebias, Austrofundulus, Cynopoecilus, Terranotus and their included subgenera. CHARACTER ANALYSIS: Orbital rim: The orbital rim is attached to some degree in all aplocheiloids. In all the Neotropical aplocheiloids and in the African genera Aphyosemion, Fundulosoma, Nothobranchius and Epiplatys, the covering of the eye is continuous with that of the head along the perimeter of the orbit. In the remaining aplocheiloids, those species of the genera Aplocheilus and Pachypanchax, the rim is attached on the lower half of the orbit, and is apparently folded under the expansion of the orbit dorsally. In the cyprinodontoids, and other atherinomorph fishes, the orbital rim is free all along its perimeter. The partially attached rim of Aplocheilus and Pachypanchax initially appears to be an intermediate state between the completely free rim and the fully attached rim. However, because of its apparent folding under the frontals, and also because the Old World aplocheiloids are assessed as a monophyletic group on the basis of a series of other characters, the partially attached rim is most parsimoniously assessed as secondarily derived in Aplocheilus and Pachypanchax. Mesethmoid: The atherinomorph fishes exhibit a derived condition of the ethmoid region as described by Rosen (1964). The mesethmoid is typically represented by two ossified discs which are angled toward each other at their anterior limit to create a wedge. In all aplocheiloids examined, the mesethmoid is totally cartilaginous, except for the presence of some small ossification centers in several larger specimens of Cynolebias. The mesethmoid is generally a large ossified structure in most other cyprinodontiforms; however, it is cartilaginous in the procatopines, and in the cyprinodontines of the Anatolian region (e.g., in Aphanius and Kosswigichthys). Because the last two groups are members of the well defined, monophyletic cyprinodontoids, the cartilaginous mesethmoid is hypothesized to be independently 368 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY FIG. 10. Diagrammatic representation of pelvic girdles of A. Aplocheilus panchax; B. Jenynsia lineata; C. Aphanius fasciatus. VOL. 168 derived several times within the evolution of cyprinodontiform fishes. Pelvic Girdle: The pelvic girdles of atherinomorphs are found united, as in Menidia and other atherinoids, or widely separated, as in Oryzias. When united, the pelvic bones are joined medially by overlapping processes. In cyprinodontiforms, they are so united. In the cyprinodontoids, as well as in Menidia, the anterior part of the girdle is perpendicular to the medial overlapping processes (fig. 10B, C). In contrast, the aplocheiloids (fig. 10A) have pelvic bones which are set close together as a result of the medial processes being reduced. Gill Arches: Aplocheiloids generally exhibit the primitive state of the gill arch characters for cyprinodontiform fishes. That is, there is a large interarcual cartilage running from the base of the first epibranchial to the side of the second pharyngobranchial, to which it attaches by a ligament. There are also rosette-shaped gill rakers which have been described by Myers (1927) as being unique to the aplocheiloids. This type of gill raker, however, is found in Menidia and other atherinoids, as well as many cyprinodontoids. Therefore, it is hypothesized to be a primitive character of a group larger than the cyprinodontiforms and is not a defining character of the aplocheiloids. One characteristic of the gill arches unique to the aplocheiloids is the broad anterior end of the basihyal (fig. 11 A). This is typically a slender bone with a cartilaginous cap which is only slightly flared in cyprinodontoids (fig. 11B) and other atherinomorphs. In aplocheiloids, however, the bone and its associated cartilage approach the shape of an equilateral triangle, especially in the Old World aplocheiloids which have just a small ossified basihyal and large cartilaginous segment (fig. 11 A). In the Neotropical aplocheiloids, the ossified segment is much larger. The large cartilaginous anterior end of the basihyal gives the aplocheiloids their characteristic large "tongue" which is readily visible upon opening the mouth. Lacrimal: Another derived character that defines the aplocheiloids as a monophyletic group is the shape of the lacrimal. In all aplocheiloids, the lacrimal is a narrow and 1981 PARENTI: CYPRINODONTIFORM FISHES 369 FIG. 11. Diagrammatic representation of ventral gill arches of A. Nothobranchius melanospilus; B. Cubanichthys cubensis. Cartilage is stippled. twisted bone which often carries a distinct sensory canal (fig. 12A). This is in contrast to the wide and flat lacrimal typically found in the atherinomorph fishes. In Menidia (fig. 12B) the lacrimal is flattened and expanded ventrally. In Poecilia (fig. 12C) the lacrimal is also flat and wide. The lacrimal is reduced among the cyprinodontoids in the genus Pantanodon; however, the preorbital distance is wide compared to that of the aplocheiloids, and the narrow lacrimal of Pantanodon is considered secondarily derived. Among the aplocheiloids, the canal of the lacrimal is apparently reduced along with the sensory canals of other dermal bones in the Neotropical aplocheiloids, as discussed below. Anterior Naris: In all atherinomorph fishes minus the exocoetoids, the anterior and posterior naris are represented by two separate openings. The anterior naris is typically a small opening just posterior to the fold of skin surrounding the maxilla, whereas the posterior naris is typically a small slit just anterior and dorsal to the orbit. The anterior naris of all aplocheiloids is surrounded by a distinct tube of skin which projects anteriorly over the upper jaw (fig. 13). A distinct tubular naris is found among the cyprinodontoids in Cubanichthys and Anableps, and is considered to be independently derived in the aplocheiloids and in each of these advanced genera. In all other cyprinodontiforms, the anterior naris never has such a fleshy extension. Cephalic Sensory Pores and Squamation: Gosline (1949) surveyed sensory pore patterns in the cyprinodontiforms as a whole with an emphasis on the goodeids and Fundulus. Hoedeman (1958, 1974) in his surveys of head-scale patterns among cyprinodontiforms was responsible for bringing attention to this associated character. Rosen and Mendelsohn (1960) surveyed the sensory pore as well as head-scale patterns among the poeciliids. It has been observed that both sensory pore and head-scale patterns are correlated. That is, their separate analyses 370 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 12. Diagrammatic representation of left lacrimal of A. Rivutus harti; B. Menidia menidia; C. Poecilia vivipara. Anterior is to the left. could introduce redundancy into this study. Therefore, the two systems are discussed together. There is much taxic variability as well as ontogenetic variation within these systems. The cephalic sensory system starts out in ontogeny as a series of open grooves over which skin grows leaving the grooves open to the surface only by a series of pores. Gosline also noted for three species of Fundulus that neuromasts of the juvenile pattern are covered over as the individual grows and replaced by a covered canal with pores. Hoedeman (1958) referred to replacement of two scales by one larger scale as a fusion; however, Roberts (1970) suggested instead that it was the formation of fewer scale precursors that resulted in such changes. Without developmental studies on comparative scale formation, this difference seems moot; however, such replacements will not be referred to as fusions because this implies a process for which there is no evidence. As a consequence of variability within these systems, it is only the most general characteristics which may be incorporated into a higher level phylogenetic analysis. Also, because of the ontogenetic changes, it is usually the maximum development of pores and scales that is reported as the general state for a species or genus. The generalized sensory pore and squamation pattern for cyprinodontiforms is as in the cyprinodontoid, Jenynsia lineata (fig. 14A, B, C). There are seven preopercular pores, three lacrimal pores, four mandibular pores, and seven supraorbital pores. This pattern is judged to be the primitive pattern for cyprinodontid fishes based on personal observation, and data from Gosline (1949) and Wiley (personal commun.). The numbering system for the pores follows Gosline. Departures from the general pattern (fig. 14) occur both with respect to the continuity of canals between pores, and therefore, the number of pores, and the number and position of scales. Among the aplocheiloids, the sensory pores of the head are reduced to a series of neuromasts, as in Rivulus harti (fig. 13C), a Neotropical aplocheiloid, and in Epiplatys sexfasciatus (fig. 13B), and Old World species. Among the cyprinodontoids, the sensory pores of the head are often reduced, but not replaced by neuromasts. Sexual Dimorphism: All cyprinodontiforms show marked sexual dimorphism. Males typically are elaborately pigmented and frequently have elongated rays in the unpaired and the pelvic fins. All cyprinodontiforms exhibit sexual dimorphism in size. In aplocheiloids, males are larger than females. In the cyprinodontoids, the reverse is true: females are larger than males. The exception in the cyprinodontoid killifishes occurs within the procatopines and Pantanodon and Fluviphylax in which the males are always larger than the females. Among the silversides, in genera such as Bedotia, Melanotaenia, Gulaphallus and 1981 PARENTI: CYPRINODONTIFORM FISHES 371 FIG. 13. Diagrammatic representation of cephalic sensory pores and squamation. Epiplatys sexfasciatus: A. lateral view of lacrimal and preopercular pores, B. dorsal view of neuromasts and squamation, C. ventral view of mandibular and preopercular pores; Rivulus harti: D. lateral view of reduced pore pattern, E. dorsal view of neuromast and squamation pattern (lettering of scales corresponds with b.), F. ventral view of covered branchiostegal region; Cynolebias elongates: G. lateral view of minute neuromast pattern, H. dorsal view of neuromast pattern, I. ventral view of neuromast pattern and branchiostegal region. others, the males are larger than the females. In at least one species related to Menidia beryllina, females are larger than males. In exocoetoids the males are also larger than females, or of approximately the same size. Thus, the most general condition among the atherinomorph fishes appears to be that males are larger than females. If so, the primitive state for the cyprinodontiforms is for males to be larger than females. In this case, the aplocheiloids retain the primitive condi- tion and the small size of the males is a derived character of cyprinodontoids. The large male in procatopines is, therefore, a secondarily derived condition. No general pattern of pigmentation has been discovered among aplocheiloids. Some pigmentation patterns do, however, occur in some of the more primitive genera of both the Old World and Neotropical groups, and these are assessed as derived for the aplocheiloid fishes. These include a caudal ocel- 372 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 14. Diagrammatic representation of cephalic sensory pores and squamation. Jenynsia tineata: A. lateral view of preopercular and lacrimal pores, B. dorsal view of supraorbital pores and squamation, C. ventral view of mandibular and preopercular pores; A generalized poeciliid (after Rosen and Mendelsohn, 1960): D. lateral view of preopercular and lacrimal pores, E. dorsal view of supraorbital pores and squamation pattern, F. ventral view of mandibular and preopercular pores; Orestias cuvieri: G. lateral view of minute neuromast pattern, H. dorsal view of minute neuromast pattern, I. ventral view of neuromast pattern and embedded urohyal. lus in females; bars running across the ventral surface of the head, commonly called "throat bars"; and a band, normally of white or yellow, on the dorsal and ventral base of the caudal. The caudal ocellus, or "Rivulus spot" has been used to diagnose fishes of the genus Rivulus. In females of all nominal species of the genus, there is a black blotch (fig. 54), or sometimes discrete spot (e.g., as in R. marmoratus) dorsally on the caudal fin base. Such an ocellus, however, is also found in some females of the genus Aphyosemion. Assuming that the spots are homologous, the character is a synapomorphy of the aplocheiloids, and no longer defines Rivulus as a monophyletic genus. Throat bars are most prominent in species 1981 PARENTI: CYPRINODONTIFORM FISHES 373 FIG. 15. Diagrammatic representation of ventral view of head to show throat bars in A. Epiplatys dageti, B. Epiplatys chaperi, C. Aplocheilus panchax. of the genus Epiplatys (fig. 15A, B), although they are also a conspicuous component of the pigmentation patterns of species of Aphyosemion and Aplocheilus (fig. 15C), and to a lesser extent Rivulus and Pachypanchax. These are often found in conjunction with a line of pigment on the lower lip; however, this line of pigment is found among many cyprinodontoids, as well as in many aplocheiloids which do not possess conspicuous throat bars (e.g., in Nothobranchius). A horizontal band of yellow or white, rarely blue, on the dorsal and ventral bases of the caudal fin is a distinctive component of the pigmentation patterns of the males of most species of Rivulus, Aphyosemion, Fun- A " " C 16. Diagrammatic representation of a ventral view of skull in A. Fundulus heteroclitus, B. Tomeurus gracilis, C. Aplocheilichthys johnstoni. Lateral ethmoid is cross-hatched, dermosphenotic blackened, autopterotic stippled. FIG. 374 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 17. Diagrammatic representation of a ventral view of skull in A. Aphyosemion occidentale, B. Aplocheilus panchax, C. Rivulus hard. Lateral ethmoid is cross-hatched, dermosphenotic blackened, autopterotic stippled. dulosoma, and to a lesser extent in species of the genus Epiplatys (e.g., in Epiplatys fasciolatus). Such a band is found only on the ventral base of the caudal in species of the nominal genera Rachovia and Neofundulus. In at least two species of Aphyosemion, A. celiae and cinnamomeum, the light yellow bars extend along the posterior margin of the fin to form one continuous band (Radda, 1979). The difficulty in determining the primitive or advanced state of these pigment patterns is analogous to determining whether sexual dimorphism involving large males or large females is primitive. The general color patterns just described are normally retained in preserved specimens since the majority involve dark pigments, rather than the unstable or water and alcohol soluble pigments of yellow, blue and red. In an analysis done at the level of the current study, intrageneric variation has not been the subject of concentrated study. However, elaborate color patterns of large males, distinguished by the derived characters listed here, distinguish aplocheiloids from all other cyprinodontiform fishes. Other types of elaborate male color patterns occur within the poeciliid fishes among cyprinodontoids. Elaborate male coloration does not occur in the hypothesized primitive poeciliid, Tomeurus gracilis, however. Position of the Vomer: The posterior extension of the vomer typically lies ventral to the anterior arm of the parasphenoid in all cyprinodontoids (fig. 16), as well as other atherinomorphs. In contrast, the posterior extension of the vomer lies dorsal to the anterior arm of the parasphenoid in all aplocheiloids (fig. 17), a clearly derived condition. SUMMARY OF DERIVED CHARACTERS 1. 2. 3. 4. 5. 6. 7. 8. Attached orbital rim. Cartilaginous mesethmoid. Close set pelvic girdles. Broad anterior end of the basihyal. Narrow and twisted lacrimal. Tubular anterior nans. Reduced cephalic sensory pores. Males more elaborately pigmented than females. 9. Posterior extension of the vomer dorsal to the anterior arm of the parasphenoid. NEOTROPICAL APLOCHEILOIDS The aplocheiloids are hypothesized to comprise two monophyletic groups of genera: one a solely Old World group, and the second Neotropical and north temperate. The latter range from southern Florida and the Bahamas, through Central America, and southward to Paraguay (fig. 18). They are currently classified in 13 nominal genera and subgenera: Rivulus, Rachovia, Austrofundulus, Trigonectes, Rivulichthys, Pterolebias, Terranotus, Neofundulus, Leptolebias, Campellolebias, Cynopoecilus, Simpsonichthys, and Cynolebias, five of which are monotypic. In all revisions or discussions of all or part of the Neotropical aplocheiloids, they have been treated as a monophyletic group (e.g., Regan, 1912; Myers, 1927; Weitzman and Wourms, 1967; and Taphorn and Thomerson, 1978), although the group has never been formally defined. Regan (1912) listed the following characters to distinguish the three then known genera Rivulus, Pterolebias, and Cynolebias, and his newly named Cynopoecilus from other fishes then classified in the Fundulinae: snout short, margin of eyes not free, gill membranes separate, mouth wide and transverse with the premaxillaries protractile, lower jaw prominent and oblique, teeth subconical and arranged in bands, with one large outer row, pectorals placed low, and pelvics not far in advance of the anal. The majority of these characters are either derived for the aplocheiloids as a group, or for all cyprinodontiforms. Others, such as "gill membranes separate" may refer to certain characters inferred in this study to be derived. Myers (1927) in a revision of Neotropical aplocheiloids cited the attached orbital rim and rosette-shaped gill rakers as defining characters, characters which clearly define a larger group of fishes. Weitzman and Wourms (1967) emphasized the ambiguity of the definitions of the genera of Neotropical aplocheiloids, and remarked on the states of certain characters, such as the close-set pelvic fins in all members, but did not compare these with other cyprinodontids in an effort to present defining characters of the group. Similarly, Taphorn and Thomerson (1978) who treated only those species of the genera Rachovia and Austrofundulus, describing Terranotus as new, concentrated on enumerating differences among nominal genera, rather than describing derived similarities in an effort to determine interrelationships of the genera. The question still remains, therefore, if the Neotropical and Old World aplocheiloids form distinct monophyletic groups of genera, or whether some Neotropical genera are more closely related to some Old World genera than to other Neotropical genera. References to the overall similarity of the African Nothobranchius to the South American Cynolebias have been made continuously in the aquarium literature (e.g., Stoye, 1947; Scheel, 1968). Furthermore, and perhaps more importantly, the idea that annualism is a uniquely derived character (Wourms, 1972a) suggests that Neotropical annuals are more closely related to the Old World annuals than to the fishes of the predominantly nonannual genus Rivulus. Taphorn and Thomerson (1978) stated they agreed with Weitzman and Wourms's (1967) conclusion that the Neotropical aplocheiloids such as Austrofundulus and Rachovia were derived from a Rivulus-like ancestor, but Cynolebias and Terranotus were not included in this endorsement of Weitzman and Wourms' position. Thus, the definition of the Neotropical aplocheiloids as a distinct monophyletic group is a step toward our understanding of the evolution of the annual lifestyle, as well as toward a definition of the problems that remain in aplocheiloid systematics. CHARACTER ANALYSIS: Shoulder Girdle: The shoulder girdle of Neotropical aplocheiloids is distinguished from that of other aplocheiloids by lacking the first postcleithrum (fig. 7D). Other cyprinodontiforms have a large, scale-shaped first postcleithrum, as in BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 376 V 1 VOL. 168 1— J~^SS—»- 20°- - ^B^o°- - ^ 20°- 90° L FIG. 75° V\ /60° 40° 1 18. Distributional limits of Neotropical aplocheiloids. Aplocheilus (fig. 7C). The first postcleithrum is present in all Old World aplocheiloids examined. Cephalic Sensory Pores and Squamation: A series of derived characters related to the sensory canal system of the head, including squamation patterns, opercular and branchiostegal membranes, and development of dermal bones carrying sensory canals serve to define the Neotropical aplocheiloids as a monophyletic group. For example, the typical condition of the branchiostegal and opercular membranes in atherinomorph fishes is that exhibited by the Old World aplocheiloid Epiplatys (fig. 13C). Several branchiostegal rays are visible through a clear membrane which is separate from that overlying the opercular region. The preoperculum, which is not united by a membrane to the operculum ventrally, carries a distinct sensory canal which is open externally as a series of pores. There is a fold of skin covering the throat region between the dentary and the urohyal. The throat region, including the branchiostegal membranes, is generally not covered with scales. In contrast, the membranes covering the opercular region and the branchiostegal rays are totally united in the Neotropical aplocheiloids. In the generalized state (e.g., in Rivulus, fig. 13F) the branchiostegal rays are not conspicuous externally. Scales usually extend over this continuous covering of the ventral surface of the head which obscures the separation between the preoperculum and operculum. There are no preopercular pores, which is 1981 PARENTI: CYPRINODONTIFORM FISHES concordant with the weakly developed or absent sensory canal in the preoperculum (fig. 13D). No Neotropical aplocheiloid examined has a complete canal in the preoperculum; although, there is a short canal in its dorsal extension in all except Terranotus, Cynolebias, Campellolebias, Simpsonichthys, Cynopoecilus, and Leptolebias. The canal is visible externally as an obsolescent canal, rather than as a pore or series of pores. Similarly, the dermosphenotic is reduced (fig. 17C) and carries just a small canal. In the Old World aplocheiloids, the dermosphenotic is more deeply concave. Also, as noted in the previous section, the lacrimal is a narrow and twisted bone in all aplocheiloids; it carries a distinct canal in the Old World aplocheiloids, whereas in those of the Neotropics, the canal is obsolescent. As expected, a reduction in the sensory canal system in the opercular and infraorbital region is correlated with that on the dorsal surface of the head. The reduction involves the substitution of enclosed canals which open to the surface by pores for neuromasts or pit organs. Gosline (1949) examined just one species of Neotropical aplocheiloid, Rivulus holmiae, and concluded that it had not developed supraorbital canals, or preopercular, mandibular, or preorbital canals. This was found to be true for all Neotropical aplocheiloids; however, there is not a progressive reduction of neuromast patterns within the group, but rather a further elaboration of the neuromast pattern in more derived genera. The pattern of neuromasts is the simplest in primitive members of the genus Rivulus. For example, in R. marmoratus, there are four supraorbital neuromasts, which may correspond to the pores 1, 2a, 2b, and 5 or 6 of the general pattern. The lacrimal, mandibular and preorbital pores typically are represented by neuromasts. The neuromast pattern is progressively more elaborate in the more derived species of Rivulus (e.g., R. harti) and in the other genera of Neotropical aplocheiloids. In a review of species of the nominal genera Rachovia and Austrofundulus, Taphorn and Thomerson (1978) illustrated a variety of 377 neuromast patterns found among the species currently placed in those genera. There are a series of neuromasts running posteriorly from a position medial to the anterior naris and extending to the posterior limit of the orbit. The pattern continues in the preorbital region. A derived form of this pattern is exhibited by members of the genus Cynolebias (fig. 13G, H, I). There are a series of minute neuromasts which resemble perforations of the skin. These neuromasts run posteriorly along the dorsal surface of the head from a position medial to the anterior naris to a position posterior to the orbit where the line turns abruptly back toward the orbit, then posteriorly again for a short distance, curving back to the path of the postorbital canal, continuing around the eye, and ending near the posterior naris. A similar pattern is exhibited by some cyprinodontoids, e.g., the South American genus Orestias (fig. 14G, H, I) and the nominal Anatolian genus Anatolichthys and some species of Aphanius. This pattern is considered to be secondarily derived in the cyprinodontoids based on the fact that the last three mentioned genera share a series of derived characters with other cyprinodontiforms that are not shared by the aplocheiloids. The significance of the pattern exhibited by these genera will be evaluated in the discussion of their interrelationships. The head squamation pattern of figure 14B is hypothesized to be primitive for cyprinodontiform fishes. Following the convention established by Hoedeman (1958), there is a single A scale, which is identified as the median scale lying posterior to a line drawn through the posterior limits of the orbits. It is preceded by two E scales, one of which overlaps the other. A single G scale precedes these, and there are often two or more F scales laterally. The terminology of the head scales used to describe the general pattern found in cyprinodontiforms was not developed to describe this pattern, but for that typically found in Rivulus and other Neotropical aplocheiloids. Hoedeman considered the pattern of Rivulus (fig. 13B) to be a funda- 378 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY mental arrangement of head scales. The central A scale is surrounded by a series of scales which are coded B through E proceeding counterclockwise from the scale just posterior to A. This pattern is not fundamental for cyprinodontiforms, but is unique to the Neotropical genera. The unique component of the pattern in more derived species of Rivulus and the other Neotropical genera is the inclusion of the G scale in series surrounding the A, thus preventing the overlap of the E scales. Also, scales B through E are all approximately the same size, whereas, generally in cyprinodontiforms, the A, E, and G scales are prominent. Hoedeman (1974) illustrated a pattern for Cynopoecilus melanotaenia in which there are two small G scales, rather than one, and these are not in series around the A. I have not observed this pattern in C. melanotaenia, and conclude that it is possibly part of intraspecific variation. There is a great deal of such variation with head squamation patterns. The single A scale is not always present, and when present can either overlie the circular arrangement of the B through E scales, or lie beneath them. There are also a series of smaller scales anterior to the lettered scales, covering the dorsal surface of the head from a position over the middle of the eye to the margin of the frontals. Taphorn and Thomerson (1978) reported much individual variation in the named scale patterns of Hoedeman and therefore questioned the value of this character to distinguish among species groups within the Neotropical aplocheiloids. Their judgment was borne out by this study, and it is concluded that the sensory neuromast patterns may be more readily characterized and incorporated into a phylogenetic analysis. Lateral Ethmoid: Another series of characters uniting the Neotropical aplocheiloids into a monophyletic group concerns the relative shape and position of the lateral ethmoid and the vomer. The orientation and relative size of the lateral ethmoid varies among cyprinodontiforms, but the general condition among aplocheiloids is that exhibited by the Old World genus Aphyosemion (fig. 17A). VOL. 168 The vomer is broad and typically bears a patch of teeth. Its lateral processes extend anteriorly just under the lateral ethmoids. There is a distinct facet for articulation of the autopalatine on the anterior edge of the lateral ethmoid. This process, which is prominent in all species of Profundulus (fig. 57A), is not strongly formed in any other cyprinodontoids. It is considered to be a retained primitive character in Profundulus. The medial face of the lateral ethmoid in Old World aplocheiloids is not expanded toward the parasphenoid. Within the Neotropical aplocheiloids, however, the medial processes of the lateral ethmoids are greatly expanded and lie just lateral or dorsal to the anterior arm of the parasphenoid (fig. 17C). Maxilla: The maxilla among all aplocheiloids, Profundulus and fundulines is narrow and twisted (fig. 5B, C). The anterior arms of the maxilla in aplocheiloids and Profundulus extend medially toward the large rostral cartilage to which they are affixed by connective tissue fibers. In the Neotropical aplocheiloids alone, the arm has a process on its anterior face (fig. 5A), rather than being gently curved as in all other cyprinodontiforms with pronounced anterior arms. SUMMARY OF DERIVED CHARACTERS 1. First postcleithrum absent. 2. Branchiostegal and opercular membranes united. 3. Obsolescent preopercular and lacrimal canals. 4. Lacrimal, preopercular and mandibular canals represented by neuromasts. 5. Head scales arranged in circular pattern. 6. Medial process of lateral ethmoid expanded. 7. Process on ventral arm of maxilla. RELATIONSHIPS OF NEOTROPICAL APLOCHEILOIDS: The interrelationships of Neo- tropical aplocheiloids have been discussed most recently by Weitzman and Wourms (1967) and Taphorn and Thomerson (1978). There are currently 13 nominal genera and subgenera in the group as listed above which together are defined as monophyletic by the characters just discussed. 1981 PARENTI: CYPRINODONTIFORM FISHES 379 FIG. 19. Sketch of body form and fin formation in a male Cynolebias (Austrofundulus) dolichopterus. Dotted line approximates base of hypural plates. (After Weitzman and Wourms, 1967.) Treatments of the interrelationships of these genera have focused on the overall similarity of genera without regard to the primitive or derived nature of characters, and, also, on the failure of current generic definitions to distinguish the included species from those of other genera. The latter problem is perhaps the most serious barrier to recognizing the monophyletic groups of species. Because no single character consistently distinguished all the species of Pterolebias, Austrofundulus, and Rachovia, Weitzman and Wourms suggested that they be included in one genus, although no formal taxonomic decisions were made. In the same paper, they described a new species of South American aplocheiloid which they hesitantly placed in the genus Austrofundulus on the basis of overall body shape and coloration. The species, A. dolichopterus (fig. 19) is readily distinguished from all other members of the group by its small size and extremely elongate fin rays. Taphorn and Thomerson removed dolichopterus from Austrofundulus and placed it in a new monotypic genus, Terranotus. They stressed that it had no clear relationship to any known genus, but that it might be more closely related to Cynolebias than to either Austrofundulus or Rachovia. The creation of monotypic genera for the reception of species whose placement in a taxonomic scheme is not readily apparent has been the trend in aplocheiloid systematics. Five of the 13 genera of Neotropical aplocheiloids are monotypic: Terranotus, Simpsonichthys, Cynopoecilus, Campellolebias, and Trigonectes. Redefinitions of aplocheiloid genera based on derived characters rather than on unique combinations of characters will eliminate the ambiguous placement of species such as dolichopterus. The following phylogenetic analysis does not involve the revision of all species of the genera of Neotropical aplo- 380 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 20. Cladogram of relationships of Neotropical aplocheiloids. Node A: lack of the first postcleithrum; branchiostegal rays covered; reduced cephalic sensory pore pattern; reduced dermosphenotic; reduced lacrimal; reduced preopercular; medial processes of the lateral ethmoids expanded; vomer triangular posteriorly; maxilla with anterior process; Node B: cartilaginous interhyal; pelvic fin rays seven or more; Node C: elongate rostral cartilage; extension of the pectoral fin rays to the base of the pelvics; Node D: reduction of anterior ramus of the alveolar arm; derived pigmentation pattern; Node E: lack of the interarcual cartilage; Node F: vertical bar through the eye; thickened anal rays in females; tendency to develop a fatty predorsal ridge in older males; Node G: derived color pattern; Node H: reduced number of scales in lateral series; Node I: dorsal fin rays number 14 or more; Node J: no derived characters; Node K: first proximal radial absent; enlarged spine on first vertebra; reduced fourth ceratobranchial dentition; Node L: pigmented anal papilla; Node M: caudal fin not scaled; preopercular canal closed. cheiloids which comprise over 110 species (Lazara, 1979). Instead, interrelationships based primarily on the type species of nominal genera are estimated to produce a working model of the phylogeny of this group as in the cladogram of figure 20. No new generic names are proposed for species excluded from nominal genera, since it is believed that new generic names should be proposed only for definable monophyletic groups of species. Rivulus currently contains over 60 species and as such is the largest genus within the group. Traditionally it has been defined on the basis of the presence of a caudal ocellus in females (fig. 54). The presence of such an ocellus in females of the African genus Aphyosemion suggests that this character is primitive for all aplocheiloid fishes. Annualism: Among the Neotropical aplocheiloids, only the species of the genus Rivulus, except for R. stellifer Thomerson and Turner, are reportedly nonannual. Rivulus stellifer, in this scheme, is hypothesized to be more closely related to the more derived genera of Neotropical aplocheiloids. The nature of the development of many species of Rivulus is unknown, and there are probably more annual species included. Furthermore, the annualism of the nominal genera Trigonectes and Rivulichthys is merely inferred by their locality of capture. Thus, it has not been determined whether the annual Neotropical aplocheiloids form a monophyletic group. This possibility remains to be tested by a revision of the nominal species of Rivulus in light of the present findings. Jaw Structure: All Neotropical aplocheiloids excluding the species of Rivulus examined, have an elongate rostral cartilage (fig. 4) extending for at least half its length beyond the tips of the premaxillary ascending processes. The primitive state of the rostral cartilage for cyprinodontiforms as evidenced by its occurrence in all other aplocheiloids, and in a slightly modified state in Profundulus, is as a pentagonal block of 1981 381 PARENTI: CYPRINODONTIFORM FISHES cartilage, the posterior end of which extends just slightly beyond the tips of the ascending processes (e.g., in R. hard, fig. 5A). The monotypic Trigonectes and its included species strigabundus and the genus Rivulichthys are distinguished from all other aplocheiloids by their sharply angled mouth cleft. This is produced by a foreshortening of the anterior ramus of the arm of the premaxilla (fig. 21). Such an oblique cleft and reduction of the anterior ramus also occur in a nominal species of Rivulus, R. rogoaguae, considered herein to be a close relative of strigabundus and rondoni. Coloration: Species of Trigonectes and RivulicHthys, as well as R. rogoaguae, also have a derived color pattern consisting of four rows of brown or red reticulations along the sides of the body. The dorsal, anal, caudal, and pectoral fins also have two or three rows of reticulations. A similar color pattern occurs in one other species of South American aplocheiloids, Neofundulus paraguayensis. Neofundulus paraguayensis on the basis of other characters, which will be discussed, is apparently not closely related to strigabundus and rondoni. Rivulus rogoaguae, on the other hand, most likely belongs in a monophyletic group with strigabundus and rondoni, although it is considered to be a synonym of neither. A vertical bar through the eye is a prominent component of the color pattern of species of the genera Rachovia, Austrofundulus, Cynolebias, as well as its included subgenera, and the genus Neofundulus (fig. 19). This bar is not present in species of Pterolebias, Rivulichthys, Trigonectes, nor any species of Rivulus examined. A bar occurs in just one other species of aplocheiloids, Nothobranchius microlepis of Somalia. The occurrence of such a derived pattern in what have otherwise been evaluated as two unrelated groups of aplocheiloids suggests that this pattern has been independently derived twice within the aplocheiloids. The two species of Austrofundulus, A. limnaeus and transilis, are distinguished from all other Neotropical aplocheiloids by having heavily pigmented anal papillae. The ASC FIG. 21. Diagrammatic representation of premaxilla of Trigonectes {Rivulichthys) rondoni, lateral view. anal papillae are only slightly pigmented or bare in other genera. Predorsal Ridge: Taphorn and Thomerson (1978) stated that only members of the genera Rachovia and Austrofundulus as they defined them, developed a fatty predorsal ridge, and that this ridge did not develop in Terranotus. However, in all male Terranotus examined, including types, a prominent dorsal ridge was found comparable to those in Rachovia and Austrofundulus. Therefore, it is proposed that this character is derived for this inclusive group. Spine on First Vertebra: The inferred more primitive genera, including Rachovia, Pterolebias, Trigonectes, Rivulichthys, and Rivulus, have a relatively short spine on the first vertebra and a correspondingly straighter dorsal profile than in Austrofundulus, Terranotus and Cynolebias. In these latter genera, the dorsal profile is greatly arched (figs. 382 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 13G, 19), apparently as a result of an enlarged spine on the first vertebra. Gill Arches: The species of the genus Rivulus may be divided into two groups. One includes the smaller forms such as cylindraceus, marmoratus, and heyei. These are all characterized by having an interhyal which is ossified; that is, there is a perichondrally ossified element between the posterior ceratohyal and hyomandibula. This is the condition typical of the interhyal of teleost fishes. The larger species of Rivulus, of which hard may be considered typical, have an unossified interhyal instead. The character of an unossified interhyal, however, does not define a monophyletic group of Rivulus species, for it is found in all species examined of all remaining Neotropical aplocheiloids. Thus, it indicates that some species of Rivulus are more closely related to the other genera than to some species of Rivulus. Hence, the genus as currently defined is paraphyletic. Rachovia has fourth ceratobranchials which are covered with teeth. In Austrofundulus and Terranotus, there are no teeth on the medial expansion of the gill arch elements, but the posterior arm does possess at least one tooth. In Cynolebias, the fourth ceratobranchials are devoid of teeth. Therefore, the reduction of teeth on the fourth ceratobranchial is assessed as a defining character of a group including Cynolebias, Austrofundulus, and Terranotus. All characters could not be determined for the sole specimens of ornatipinnis and Paraguay ensis studied; however, they do have the primitive character of fourth ceratobranchials covered with teeth. The interarcual cartilage is present primitively in the aplocheiloids (e.g., as in Austrofundulus transilis, fig. 6A) and in all Neotropical aplocheiloids except those of the nominal genus Pterolebias, in which it is absent. The cartilage is reduced in one species of Rachovia, R. maculipinnis. Fins: All species of Rivulus examined (except rogoaguae) have all fins rounded. There are never any caudal, pelvic, dorsal or anal fin extensions as found in the other Neotropical aplocheiloids. Typically, the pectoral VOL. 168 fins are elongate and reach the base of the pelvic fins in rogoaguae and all non-Rivulus species. Cynolebias and its relatives have a caudal which is rounded as a result of a unique orientation of the procurrent rays perpendicular to the vertebral column. There are usually six pelvic fin rays in atherinomorph fishes. There are six in all Old World aplocheiloids as well as members of the genus Rivulus with an ossified interhyal (e.g., R. marmoratus). In the larger species of Rivulus examined, e.g., harti and stellifer, the number of pelvic fin rays is increased to seven or eight as it is in all other Neotropical aplocheiloids. This increase is interpreted as a derived character at this level. The pelvic fin rays rarely are increased to seven among the cyprinodontoids in some species of the North American cyprinodontines. The nominal species of Cynolebias and Terranotus lack scales on the caudal fin. As mentioned, the scaled caudal fin is apparently primitive for aplocheiloids, and is present in all other members. Taphorn and Thomerson suggested that Terranotus may be more closely related to Cynolebias because its anterior proximal anal radials "articulate" with ribs rather than hemal spines. However, within actinopterygian fishes, the proximal anal radials never properly articulate with either ribs or hemal spines, but are often found intercalated with the latter. The character to which Taphorn and Thomerson referred could more appropriately be described in terms of the position of the anal fin relative to the abdominal and precaudal vertebrae. In dolichopterus, there are 14 abdominal vertebrae and 12 precaudal; the first proximal anal radial (equivalent to the second of Rivulus since the first is lost) extends between the pectoral ribs of the ninth abdominal vertebra. In the one species of Pterolebias which exhibits this character, there are 15 abdominal vertebrae, 18 precaudal, and the first proximal anal radial extends just behind the ribs of the twelfth abdominal. Among species of Cynolebias, the vertebral number is quite variable. In C. melano- 1981 PARENTI: CYPRINODONTIFORM FISHES taenia, there are 12 abdominal, 17 precaudal, and the first proximal anal radial extends just posterior to the ribs of the tenth abdominal vertebra. In C. whitei, the counts are 14 plus 15, with the anal radial at the rib of the eighth abdominal. In C. elongatus, the counts are 16 plus 21, with the anal radial at the rib of the fourteenth abdominal. Perhaps a derived character is the extreme anterior position of the anal fin which would indicate that dolichopterus is closely related to a group of Cynolebias species that includes whitei. However, in addition to contradicting the derived characters of Cynolebias not shared by dolichopterus, it conflicts with another character that may be of significance in Cynolebias interrelationships; that is, the number of anal fin rays. In both elongatus and whitei there are more than 20; in dolichopterus they range from 15 to 18 (Weitzman and Wourms, 1967). The usefulness of this character can only be determined by a survey of its states among all species of Cynolebias. The genus Neofundulus currently contains two species, paraguayensis Myers, the type, and ornatipinnis Myers. Both species are known from only a few specimens, and only the holotype of each has been examined as part of this study. Arambaru, Arambaru, and Ringuelit (in de Souza, 1979) suggested the two species be synonymized. This is opposed by evidence that they are not even closest relatives. De Souza (1979) reported meristic data for a recent collection of paraguayensis, and listed these along with values for the holotypes of each species. The number of dorsal fin rays in paraguayensis ranges from 10 to 13; in ornatipinnis it is 15. The number of anal rays is 12 to 16, and 18, respectively. The number of scales in the lateral series ranges from 34 to 38 in paraguayensis, and is 37 in ornatipinnis. The holotype of paraguayensis is a female, and that of ornatipinnis a male. Therefore, from the type material alone, it is difficult to tell whether the differences are due to sexual dimorphism alone. (In a group of species in the genus Cynolebias, males have higher dorsal and anal fin ray numbers.) 383 However, de Souza (1979) reported meristic data for what were referred to as nine randomly chosen individuals of paraguayensis, that included two juveniles. The holotype is a female, and it can reasonably be assumed that at least one of the nine specimens was a male; therefore, I conclude that ornatipinnis and paraguayensis are two species which may be easily distinguished on the basis of meristic characters. Furthermore, the disparate meristic data indicate the nonmonophyletic nature of the genus. Taphorn and Thomerson separated Rachovia from Austrofundulus the basis of, among other characters, fewer dorsal fin rays which are generally less than 13 (range 9-14) in Rachovia, as compared with generally 14 or more (range 12-18) in Austrofundulus. Although there is range overlap, the increase in dorsal fin ray numbers in Austrofundulus and N. ornatipinnis is consistent with an increase in dorsal fin ray numbers to 15 or more in Cynolebias and Terranotus. The first proximal anal radial is often fused to the second at its base. Nonetheless it is present in the genus Rachovia, as well as in the other more primitive Neotropical genera. The radial is absent (fig. 22) in the genus Austrofundulus as well as in Terranotus and Cynolebias and its included subgenera. Scales in a Lateral Series: The Neotropical aplocheiloids typically have a high number of scales in a lateral series. Determining the polarity of scale number, however, presents certain problems since the scale count is not high in all members. Within primitive members of the group, such as R. marmoratus, the scale count is over 40, greater than the more typical number of 30 for cyprinodontiforms as a group. The scale count is high also in some members of the derived genus Cynolebias, such as C. elongatus, in which the scales number over 60. There are members of the genus Cynolebias which have lower counts in the thirties and forties. Among all Neotropical aplocheiloids, however, the count is low only in members of the genus Rachovia, as the genus is delimited by Taphorn and Thomerson. In fact, they gave as a defining character of the genus a lateral series scale count of less than 32, 384 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 22. Diagrammatic representation of anal fin of Auslrofundulus transilis. Unbranched anal rays are blackened, middle and distal anal radials are stippled. lower than any other Neotropical aplocheiloid. Taking all other characters into consideration, the high scale count is derived for some group, the limits of which cannot be readily determined at the level of this study. Therefore, it is logical to conclude that the reduced scale count of Rachovia is a derived character, or else the increased number of scales has occurred several times which is an unparsimonious assessment of this character. SUBGROUP DEFINITION AND COMPOSITION: The assignment of species currently in the genus Rivulus to one group or another requires examination of all species for the state of the interhyal and number of pelvic rays. Such a survey is out of the scope of the present study. Rivulus cylindraceus is the type species of the genus; therefore, it will retain the name. This leaves those species of Rivulus with a cartilaginous interhyal without a name. I suggest they be referred to as "Rivulus" rather than propose a new name at this time since the relationship of all species of "Rivulus''' to the remaining Neotropical genera is uncertain, as is, therefore, the monophyly of the genus. The species of Rivulichthys, Trigonectes strigabundus, and Rivulus rogoaguae share a pointed snout (caused by reduction of the anterior ramus of the premaxilla) and a derived color pattern. I propose, therefore, that the two genera and R. rogoaguae be grouped in Trigonectes, of which Rivulichthys is a junior synonym. The nominal genus Neofundulus is considered to contain one species paraguayensis. The band of pigment on the ventral edge of the caudal fin in males is a character it shares with Rachovia. However, this character may very well be part of a transition series from the pigmentation pattern derived for aplocheiloid fishes; that is, a band of white or yellow pigment on both dorsal and ventral margins of the caudal. Therefore, although Rachovia and Neofundulus are considered to be included in a monophyletic group with Auslrofundulus, Terranotus, and Cynolebias, this relationship is represented as an unresolved trichotomy (fig. 20). Neofundulus ornatipinnis does not have the pigmentation pattern, yet, as stated, has an elongate dorsal fin. The condition of the holotype precludes its inclusion in any of the 1981 PARENTI: CYPRINODONTIFORM FISHES recognized genera. Therefore, I propose that it be referred to as "Neofundulus" ornatipinnis until additional specimens are available. Six nominal species have been described in the genus Pterolebias: longipinnis Garman, zonatus Myers, peruensis Myers, bockermanni Travassos, hoignei Thomerson, and maculipinnis (Weibezahn). They all possess a dorsal fin of 10 to 12 rays which is set back on the body, normally the first ray being over the second half of the anal fin; the caudal fin is finely scaled for at least onethird its length; and, the caudal peduncle is strongly compressed laterally. These characters, however, are typical for all species of Neotropical aplocheiloids discussed so far, or, as in the case of the compressed caudal peduncle, may be an artifact of preservation (Weitzman and Wourms, 1967). As such, the genus has never been defined as a monophyletic group. Taphorn and Thomerson (1978) removed maculipinnis from Pterolebias and placed it in the genus Rachovia, after Thomerson (1974) stated that three species {bockermanni, maculipinnis, and peruensis) are probably not closely related to the other members of Pterolebias. The removal of maculipinnis from a close relationship to longipinnis seems to be a valid decision based on the following characters: Pterolebias longipinnis and zonatus both lack the interarcual cartilage, a lack that may be considered a defining character of Pterolebias. This element is present in maculipinnis, although apparently reduced relative to the generalized state for aplocheiloids as previously discussed. The species maculipinnis shares with the remaining Neotropical aplocheiloids (species assigned to Rachovia, Austrofundulus, Terranotus, Neofundulus, and Cynolebias): a vertical bar running from below the eye to near the dorsal surface of the head; thickened anal rays in females; and the tendency of males to develop a fatty predorsal ridge. Provisionally, Rachovia may remain as delimited by Taphorn and Thomerson; however, the four included species brevis, maculipinnis, pyropunctata, and hummelincki might not constitute a monophyletic group. 385 They are distinguished from other Neotropical aplocheiloids with a vertical bar through the eye and a fourth ceratobranchial covered by teeth by having a lateral series scale count of less than 32 (Taphorn and Thomerson, 1978). Terranotus, with an unsealed caudal fin and a closed preopercular canal, is hypothesized to be the sister group of an assemblage of four genera: Cynolebias, Simpsonichthys, Cynopoecilus, and Campellolebias. The members of these four genera all possess rounded caudal fins in both sexes and have fourth ceratobranchials without teeth. They all have been suggested as synonyms at one time or another (Myers, 1942; Lazara, 1979), and on the basis of their shared characters, I unite them (along with Terranotus) within their senior synonym, Cynolebias. CLADISTIC SUMMARY OF NEOTROPICAL APLOCHEILOIDS: Rivulus and other Neo- tropical aplocheiloids share the derived characters outlined in the previous section: absence of a first postcleithrum; a unique head squamation pattern; the uniting of the opercular and branchiostegal membranes; reduction of the preopercular and dermosphenotic canals; the medial expansion of the lateral ethmoids and posterior extension of the vomer; and, the triangular process on the anterior face of the medial arms of the maxilla. The genus Rivulus is paraphyletic, and its species are referenced to two genera Rivulus and "Rivulus." Members of "Rivulus" share with other Neotropical aplocheiloids, excluding Rivulus, a cartilaginous interhyal and seven or more pelvic fin rays. In addition, all species in the genus Rivulus are nonannual, whereas at least one of the genus "Rivulus," stellifer Thomerson and Turner (1973), is annual. "Rivulus" and Rivulus are excluded from a larger group which is defined by an elongate rostral cartilage, and extensions of the pectoral fin rays to the base of the pelvics. Within this larger group, three subgroups are recognized: 1. Trigonectes, of which Rivulichthys is considered to be a junior synonym, is defined by an oblique mouth cleft formed prin- 386 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY cipally by the reduction of the anterior ramus of the premaxilla. 2. Pterolebias is defined here by its lack of the interarcual cartilage. 3. A group including Rachovia, Neofundulus, Austrofundulus, Terranotus and Cynolebias is defined by the following characters: a vertical bar through the eye often extending on to the top of the head; thickened anal rays in females; and the tendency to develop a fatty predorsal ridge in older males. Rachovia may not be definable as a monophyletic group, but is retained here to reference its four included species which may be distinguished from other members of group 3 that have teeth on the fourth ceratobranchial by its low number of scales in a lateral series. As presently defined, Neofundulus is polyphyletic. Neofundulus ornatipinnis is considered to be more closely related to the more derived aplocheiloids Cynolebias, Austrofundulus, and Terranotus on the basis of its increase in dorsal fin rays. It is the sole constituent of the genus "Neofundulus." Terranotus, Austrofundulus, and Cynolebias lack the first proximal anal radial; have a large spine on the first vertebra; and have reduced dentition on the fourth ceratobranchial. Austrofundulus is defined by its darkly pigmented anal papilla. Terranotus and Cynolebias have a caudal fin which is not finely scaled and lack a preopercular canal. A group of species has a unique head neuromast pattern and an increase in the number of dorsal and anal fin rays in males. These two characters define subgroups of Cynolebias. Since there is no longer reason to maintain Terranotus as a monotypic genus, I propose that this genus be considered synonymous with Cynolebias. OLD WORLD APLOCHEILOIDS Members of the aplocheiloids of the Old World have often been referred to as the most primitive of all cyprinodontiforms. Myers (1958) stated that the genus Aplocheilus represents the most basic characteristics VOL. 168 of cyprinodontiform fishes which have either been lost or become more derived in other genera. These characters include the arm of the premaxilla being free rather than embedded in the skin on the side of the head, and the swimbladder extending through several hemal arches rather than ending at the first hemal arch. My own analysis of derived characters of the aplocheiloids and of those which define the Old World aplocheiloids, indicates that Aplocheilus is a relatively derived aplocheiloid genus. Aplocheiloids of the Old World are currently classified in 29 nominal genera and subgenera. These include Aplocheilus of the Indian subcontinent and Laurasia extending along the Indo-Australian archipelago to Java; Pachypanchax of Madagascar and the Seychelles; and the African genera Epiplatys and four proposed subgenera; Aphyosemion and 11 subgenera; Adamas; Nothobranchius and four included subgenera; and Fundulosoma (fig. 23). Together they comprise nearly 300 species, over 100 of which are referred to the genus Aphyosemion or one of its subgenera. The definition of the Old World aplocheiloids as a monophyletic group again supports the contention that annualism is a lifestyle which has either arisen at least twice within cyprinodontiform fishes, or is a characteristic that in some sense is basic to all members. CHARACTER ANALYSIS: Shoulder Girdle: The Old World aplocheiloids are distinguished from all other killifishes by the fusion of the posttemporal and supracleithrum (fig. 7C) to form one slender bone connecting the shoulder girdle to the skull. The fusion is complete, and no joint lines are visible. In the Neotropical aplocheiloids, the posttemporal and supracleithrum are similarly shaped, however, the two bones may always be separated easily. Gill Arches: Two derived characters of the gill arches distinguish the Old World aplocheiloids. One is the reduction of the basihyal to a small triangular-shaped bone which is capped by a large cartilaginous wedge (fig. 11 A). As stated, in all aplocheiloids, the basihyal is very wide, and forms the basis of a wide "tongue" which is visible upon opening PARENTI: CYPRINODONTIFORM FISHES 1981 FIG. 23. 387 Distributional limits of Old World aplocheiloids. the mouth. In the Neotropical aplocheiloids, as well as in the cyprinodontoids and other atherinomorphs, the basihyal is ossified for more than half its length. A second character concerns the attachment of the interarcual cartilage to the second pharyngobranchial. The Neotropical aplocheiloids exhibit the primitive state for cyprinodontiforms; that is, the cartilage attaches to a small flange of bone lateral to the cartilaginous extension of the pharyngobranchial (fig. 6A) as it does in the aplocheiloids. In the Old World aplocheiloids, the cartilage attaches in the same position, but the bony flange is absent; thus, the cartilage nearly abuts the cartilaginous articulation point of the pharyngobranchial (fig. 24A). In the genus Nothobranchius, the cartilage attaches directly to the cartilaginous head of the pharyngobranchial (fig. 24B). This character is apomorphic for the genus. Premaxillary: The premaxillary ascending processes are flat and broad in the Neotropical aplocheiloids and in Profundulus. This state is most parsimoniously assessed as the most primitive among cyprinodontiforms as discussed in the section on derived characters of the group. Within the Old World aplocheiloids, the premaxillary ascending processes are tapered posteriorly to form, in the most derived state, the greatly expanded triangular processes of Aplocheilus (fig. 4A). In Pachypanchax (fig. 4B) and Epiplatys and its included subgenera, the processes are also expanded, although not to the degree exhibited by Aplocheilus. In Aphyosemion (fig. 4C), Nothobranchius and Fundulosoma, the processes are tapered posteriorly, but never as widely expanded as in any of the three above mentioned genera. SUMMARY OF DERIVED CHARACTERS 1. Supracleithrum fused to posttemporal. 2. Small, triangular basihyal capped by a wedge of cartilage. 3. Interarcual cartilage attached directly to lateral face of the second pharyngobranchial. 4. Premaxillary ascending processes tapered. RELATIONSHIPS OF OLD WORLD APPLOCHEILOIDS: The Old World aplochei- loids, like the Neotropical aplocheiloids, 388 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 PB3 FIG. 24. Diagrammatic representation of dorsal view of dorsal gill arches of A. Aplocheilus panchax, B. Nothobranchius melanospilus. Cartilage is stippled. have been treated as a natural group of fishes (e.g., Scheel, 1968) although their monophyly had never been tested. In addition to considering Aplocheilus as the most primitive cyprinodontiform (e.g., Myers, 1958), various authors suggest that the annual Old World aplocheiloids are the closest relatives of the Neotropical aplocheiloids. Furthermore, the Old World aplocheiloids were considered by many workers to be closely related to the procatopines, the other group of cyprinodontiforms which inhabits subsaharan Africa. Ahl (1924, 1928), for example, grouped the two without questioning their monophyly; more recently, Huber (1979) described a new genus and species, Adamas formosus, which he considered to be intermediate between Old World aplocheiloids of the genus Aphyosemion and the procatopines. Such conclusions were a product of the ambiguous definitions previously put forth for the subfamilies Rivulinae and Aplocheilichthyinae. Nevertheless, as understood here, these two distributionally similar groups have little more than primitive characters in common. The procatopines do not possess any of the derived characters of Old World aplocheiloids summarized above; they possess only one of the derived characters for aplocheiloids as a whole, the cartilaginous mesethmoid. This condition is considered to be independently derived in these two groups, as well as in Anatolian cyprinodonts. Procatopines, however, do share a series of unique features with the rest of the cyprinodontoids. Thus, procatopines and aplocheiloids are not considered together further. Taxonomic revisions of Old World aplocheiloids, like those of New World forms, place emphasis on the recognition of differences among taxa rather than on the discovery and description of derived characters shared among taxa. Emphasis on differences has led to the naming of four subgenera of Epiplatys, four subgenera of Nothobranchius, and the division of Aphyosemion into 13 genera and subgenera. In a recent paper by Radda (1977), four new subgenera are named, each to encompass a group of species referable to Aphyosemion. Radda included a phylogenetic tree 1981 PARENTI: CYPRINODONTIFORM FISHES which purports to summarize the relationships of subgenera within the genus. The monophyly of Aphyosemion is doubtful, following Radda's diagram, for the subgenus Pronothobranchius and the genus Fundulosoma, considered here as close relatives, are included as more closely related to some subgenera of Aphyosemion than they are to each other. Furthermore, the genus Nothobranchius is not considered at all; therefore, it is unclear whether the implication is that Nothobranchius is in turn most closely related to Fundulosoma or Pronothobranchius, to some subgroup of Aphyosemion, or whether it need be considered at all in a revision of Aphyosemion. The genus Aphyosemion is large, currently comprising over 110 species; if some members are more closely related to Nothobranchius species then the two genera must be considered together in a phylogenetic analysis. If the two genera do not form a monophyletic group, then Nothobranchius need not be considered in a study of the interrelationships of Aphyosemion and its subgenera. The problem of defining monophyletic genera and subgenera extends to all members of the Old World aplocheiloids. Scheel (1972) has recommended that the genera Epiplatys and Aplocheilus be synonymized. Clausen (1967) has named a new subgenus of Epiplatys to included the species E. duboisi; a subgenus Aphyoplatys is named to indicate that this species is intermediate between Aphyosemion and Epiplatys. Wildekamp (1977) has recently named a new subgenus of Nothobranchius to include the species N. janpapi; it is named Aphyobranchius to reflect its intermediacy between Aphyosemion and Nothobranchius. It is of little use, however, to know that all the nominal genera of Old World aplocheiloids grade into one another from the Aplocheilus type to the Nothobranchius type. Logically, there is no reason not to classify all the species in one genus; however, that too would be avoiding the problem of the interrelationships of the included species as much as if each species were put into its own genus. Unambiguous definitions of monophyletic groups of nominal genera 389 would allow for the reference of a particular species to one monophyletic group or another, and avoid the confusion created by the current generic limits. For example, Aphyobranchius janpapi is either more closely related to species of Nothobranchius or to some group of Aphyosemion. A concise definition of each group would allow such a decision to be made. As for the Neotropical aplocheiloids, this analysis does not include the revision of all species of each genus. Rather, it is an attempt to identify and define on the basis of shared derived characters the major monophyletic groups of species and their proposed interrelationships (given in the cladogram of fig. 25) which will eventually lead to the definition of monophyletic genera. Supraspecific categories of Old World aplocheiloids may be divided into two major monophyletic groups. One is referred to as the Aphyosemion-Nothobranchius group; and the second is referred to as the Aplocheilus-Pachypanchax-Epiplatys group. The interrelationships of the members of the two are discussed separately with reference to the states of characters in the other, in the Neotropical aplocheiloids, and in cyprinodontiforms as a whole. THE Aphyosemion-Nothobranchius GROUP Species in this group are currently classified in four genera and fifteen subgenera. These are Aphyosemion Myers with eleven subgenera: Archiaphyosemion Radda, Mesoaphyosemion Radda, Kathetys Huber, Fundulopanchax Myers, Paludopanchax Radda, Chromaphyosemion Radda, Callopanchax Myers, Raddaella Huber, Diapteron Huber and Seegers, Paraphyosemion Kottelat and Gularopanchax Radda; Nothobranchius Peters, with four subgenera: Adiniops Myers, Pronothobranchius Radda, Zonothobranchius Radda, and Aphyobranchius Wildekamp; and Fundulosoma Ahl and Adamas Huber, two monotypic genera. Most of these names are unfamiliar to the majority of workers on cyprinodontiform fishes since nearly all have been just recently described in journals that are not widely dis- 390 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 25. Cladogram of relationships of Old World aplocheiloids. Taxa treated as subgenera not included in diagram (see text for further information). Node A: posttemporal fused to supracleithrum; reduction of basihyal to small, triangular wedge; interarcual cartilage attached to second pharyngobranchial which lacks a bony flange; premaxillary processes tapered posteriorly; Node B: broad, flattened upper jaw caused by expanded premaxillary ascending processes; expanded coronoid process on the dentary; bifurcate upper hypural plate in juveniles and some adults; loss of the uncinate process on fourth epibranchial; Node C: lower limb of posttemporal represented by ligament; teeth on third and fourth hypobranchials; dorsal ocellus in females; orbital rim attached only ventrally; darkened caudal fin margin; Node D: premaxillary ascending processes expanded medially and overlapping in the midline; attenuate lower jaw; Node E: hypural fan in adults; lateral scales of male angled away from body; Node F; bifid epipleural ribs; reduced chromosome number; attenuate posterior extension of the vomer; Node G: dorsal fin rays fourteen or more; dorsal origin opposite anal origin; swimbladder not expanded past first arch; Node H: preopercular canal open, not represented by pores; attachment of the interarcual cartilage directly to the second pharyngobranchial; oval eggs. tributed publicly. Nonetheless, they represent available supraspecific categories within the Aphyosemion-Nothobranchius group; therefore, their references are summarized in the systematic accounts, and they are considered herein. CHARACTER ANALYSIS: Bifid Epipleural Ribs: Bifid epipleural ribs have been used to distinguish species of the genus Aphyosemion from those of the genus Nothobranchius. Typically, in Aphyosemion, the first five or six epipleural ribs are strongly bifid distally (fig. 26). This derived character is unambiguously present in all species of Aphyosemion examined, including A. petersi, a species which has alternately been placed in the genus Aphyosemion, and in Epiplatys. On this basis, petersi should properly be placed in the AphyosemionNothobranchius group. In some species of Aphyosemion, for example, A. gulare and sjoestedti, and in Nothobranchius, the epipleural ribs are often not as strongly bifid as in most of the species of Aphyosemion; however, on close examination, they are easily determined as bifid. For example, in Nothobranchius orthonotus, the type of the genus, the first six epipleural ribs are unambiguously bifid. This character does not appear to be related to the size or age of specimens, nor is any sexual dimorphism apparent. Vomer: The vomer typically has a broad posterior extension in cyprinodontiform fishes (e.g., in Aplocheilus panchax, fig. 17B). In contrast, in members of the Aphyosemion-Nothobranchius group the posterior extension of the vomer is narrow (fig. 17A). This sets off the anterior extension of the vomer as a large rectangular element. Chromosome Number: Fishes of this group exhibit some of the lowest chromosome numbers known for teleost fishes. Teleosts generally have a haploid chromosome number of 24, and therefore a diploid number of 48. Gyldenholm and Scheel (1971) listed haploid and diploid chromosome numbers of temperate and tropical freshwater fishes in 19 families. Included were representatives of the percomorph, ostariophysan, atherino- 1981 PARENTI: CYPRINODONTIFORM FISHES 391 FIG. 26. Diagrammatic representation of the posterior region of the skull and attachment of first vertebra, lateral view, of Aphyosemion occidentale. Posttemporal removed. morph, and paracanthopterygian lineages. Karyotypes of 53 species within the family Cyprinodontidae (including Oryzias latipes) and those of 19 species of poeciliid fishes were listed. Within the cyprinodontiforms, as well as all teleosts, the usual haploid chromosome number is 24. In poeciliids the number ranges from 23 to 25; in cyprinodontiforms from nine to 25. Scheel (1968) stated that only among the aplocheiloid fishes within the family Cyprinodontidae did the haploid chromosome number reach 25; therefore, he concluded that this is perhaps the primitive number for aplocheiloids. In light of present findings concerning the nonmonophyletic nature of the cyprinodontids, inclusion of the poeciliids indicates that (1) the haploid number of 25 is attained in cyprinodontiforms other than aplocheiloids, and that (2) it is logical to conclude on the basis of outgroup comparison that n = 24 is basic for aplocheiloids, as well as cyprinodontiforms as a whole. In species of Aphyosemion (which includes the species placed in Roloffia in Gyldenholm and Scheel) and Nothobranchius, the number ranges from nine to 23 (Scheel, 1968). Among the nominal species of Epiplatys, the haploid chromosome number ranges from 17 to 25; it is 24 for the type species of the genus, E. sexfasciatus. Among the nominal species of Aplocheilus, the haploid number ranges from 18 to 25; it is 18 for the type of the genus, A. panchax. Pachypanchax playfairi is reported to have 24 haploid chromosomes, as does Aphyoplatys duboisi. Scheel (1968) maintains that the type of chromosome reduction differs in the genera Epiplatys and Aplocheilus from that in the Aphyosemion-Nothobranchius group. That is, in the former genera, the reduction involves the production of large metacentric elements with a subsequent loss of the smaller metacentrics. In the latter genera, the reduction involves the production of large ac- 392 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY rocentric elements, accompanied by loss of the smaller elements. White (1968 in Scheel) maintained that such so-called superacrocentrics could arise from pericentric inversions in metacentric elements from the same size. While this could account for the difference in karyological morphology in these two groups of Old World aplocheiloids, Scheel correctly maintains that there are no indications the superacrocentrics were produced in this manner. In one species of Neotropical aplocheiloids, Cynolebias (Cynopoecilus) melanotaenia, the superacrocentrics occur and the haploid chromosome number of the current aquarium strain is 22 (Scheel, 1968). The occurrence of superacrocentrics within another group of aplocheiloid fishes indicates that they are not unique to fishes of the Aphyosemion-Nothobranchius group, and also that chromosome reduction is not limited to the Old World aplocheiloids. Therefore, the division of Old World aplocheiloids into two groups on the inferred mode of reduction is suspect since the polarity of this reduction cannot be determined. We are left with the character of reduction of chromosomes which is useful in a phylogenetic analysis if it can be correlated with characters from a presumed independent source. Such is the case with the bifid epipleural ribs and slender posterior extension of the vomer used here. SUMMARY OF DERIVED CHARACTERS 1. Bifid epipleural ribs. 2. Attenuate posterior process of the vomer. 3. Reduced chromosome number. RELATIONSHIPS OF THE Aphyosemion-Nothobranchius GROUP Among those species of Epiplatys with low haploid numbers, the following have been examined and possess weakly bifid epipleural ribs: E. bifasciatus (n = 20) and E. spilargyreia (n = 17). Thus, it appears that on the basis of these two characters certain species of Epiplatys may be more closely related to the Aphyosemion-Notho- VOL. 168 branchius group than to Epiplatys, or else the character of bifid epipleural ribs is derived for Old World aplocheiloids. I suggest that no synonymies of Old World genera be undertaken unless the generic limits are formally defined in terms of shared derived characters drawn from a survey of all species for such characters. Adamas formosus (new genus and species described by Huber, 1979) has not been examined. It is placed in this group on the basis of overall external morphology, color pattern and sexual dimorphism as noted from a photograph included in the description. It appears to be most closely related to the primitive species of Aphyosemion as described below. Thus, I cannot accept Huber's suggestion that Adamas is intermediate between the procatopines and Old World aplocheiloids. Its precise placement within one of the existing supraspecific subdivisions of Aphyosemion will require an examination of material. Annualism: Annualism means that the fertilized egg and embryo exhibit diapause. The species included in the Aphyosemion-Nothobranchius group are both annual and nonannual. All members of the Aplocheilus-Pachypanchax-Epiplatys group are nonannual. The nonannual members of Aphyosemion are Archiaphyosemion, Mesoaphyosemion, Kathetys, Chromaphyosemion, Diapteron, Aphyosemion, and Parepiplatys. The annual species are in Raddaella, Paraphyosemion, Paludopanchax, Gularopanchax, Callopanchax, Fundulopanchax, as well as the subgenera of Nothobranchius and in Fundulosoma. Members of the genus Chromaphyosemion have been referred to as semiannual (e.g., Radda, 1979) because the eggs were observed to tolerate partial drying in the field. However, since all annuals, including socalled true annuals such as Austrofundulus transilis can be water-incubated (Wourms, 1972a) it is perhaps more appropriate to refer to the semiannual species as nonannual unless diapause can be demonstrated. Otherwise nothing more than the tolerance of desiccation has been demonstrated. 1981 PARENTI: CYPRINODONTIFORM FISHES In addition to being annual, species of the genera Fundulosotna and Nothobranchius have oval rather than the more typical round eggs of other aplocheiloids (Scheel, 1968). Swimbladder: The swimbladder of the aplocheiloids typically extends posteriorly beyond several hemal arches, as in the Aplocheilus-Pachypanchax-Epiplatys group and Aphyosemion petersi. Failure of the swimbladder to extend beyond more than the first pair of hemal arches in Paludopanchax, Gularopanchax, Callopanchax, Fundulopanchax, Raddaella, Paraphyosemion, Nothobranchius, and Fundulosotna is interpreted as a derived character of the included species. Dorsal Fin Position and Ray Number: The most primitive position of the dorsal fin for aplocheiloids is inferred to be the general condition for Old and New World genera in which the dorsal fin is set back on the body approximately opposite the last third of the anal fin. Dorsal fin rays typically number seven to 10, though the number can be slightly higher. These primitive conditions occur in the following subgenera of Aphyosemion: Archiaphyosemion, Mesoaphyosemion, Kathetys, and Aphyosemion, as well as Adamas. A second group of subgenera have a dorsal fin which is slightly elongate (generally from 10-14 rays) and situated over the first quarter of the anal fin. In this group are the subgenera Chromaphyosemion and Diapteron. A third group of Aphyosemion subgenera (Paludopanchax, Gularopanchax, Callopanchax, Fundulopanchax, Raddaella, and Paraphyosemion) and Fundulosoma and Nothobranchius have an elongate dorsal fin of over 14 dorsal fin rays the origin of which is opposite the anal origin. Cephalic Sensory Pores and Squamation: One character which does not seem to be useful in separating these subgenera into groups is the open versus closed frontal neuromas! pattern. Clausen (1966) first used the closed pattern as a defining character of a new genus Roloffia (^Callopanchax Myers). In the closed pattern, the two frontal neuromasts are encircled by a rim of epidermis, 393 whereas in the closed pattern the two neuromasts lie separated by a ridge of epidermis. Scheel (1968) published photographs of the two conditions, and used this character to place Aphyosemion petersi in the genus Callopanchax, along with the more apomorphic occidentale. Radda (1977) recognized the apparent unnatural status of Callopanchax as thus constituted and placed petersi in his Archiaphyosemion, where on the basis of the above characters, it more properly belongs. The closed frontal neuromast pattern was used again by Clausen (1967) to separate the species of his Parepiplatys from the rest of the Epiplatys species. Scheel (1968) reports that in a brood of Pachypanchax playfairi, some individuals developed with the open pattern and some with the closed. Thus, the character seems of doubtful significance for a phylogenetic study, and therefore, fails as a defining character of Callopanchax. The preopercular canal is present in all Old World aplocheiloids and typically opens to the outside by a series of pores (fig. 13A). In the subgenera of Nothobranchius and in Fundulosoma thierryi, there are no pores, as the canal is open to the outside all along the margin of the preoperculum. SUBGROUP DEFINITION AND COMPOSITION: I conclude that (1) the genus Aphy- osemion as currently constituted is not monophyletic; and (2) the annual species previously assigned to Aphyosemion are most closely related to the species of Nothobranchius and Fundulosoma. Annualism is thus postulated to have arisen just once within the Old World aplocheiloids, as may also be true of Neotropical aplocheiloids. Division of species of the AphyosemionNothobranchius group on the basis of dorsal fin position and ray number is problematic. The position of the dorsal fin is variable even among individuals of the same species. However, the species of Aphyosemion may be grouped artificially into two categories; those with from seven to 14 dorsal fin rays and the dorsal situated no farther forward than opposite the first quarter of the anal fin; and those with more than 14 dorsal fin rays and 394 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY the dorsal situated over the anal fin origin or just slightly before or after. Species groups with a posterior dorsal fin and low dorsal-fin ray number are nonannual and have a swimbladder extending beyond the first hemal arch. They include Aphyosemion, Archiaphyosemion, Mesoaphyosemion, Chromaphyosemion, Diapteron, and Kathetys. Since they share only primitive characters, these subgenera are not considered to form a monophyletic group. I suggest that they be referred to the genus Aphyosemion, however, they will not be formally synonymized with the genus since the monophyletic nature of the group is not implied. Species groups with an anterior dorsal fin and high fin ray number are annuals that have a swimbladder which does not extend posteriorly beyond the first hemal arch. They include Raddaella, Paludopanchax, Gularopanchax, Callopanchax, Fundulopanchax, Paraphyosemion, Fundulosoma, and Nothobranchius. Species of the genera Fundulosoma and Nothobranchius both share the derived state of the interarcual cartilage as described, oval eggs, and an open preopercular canal. The sole species of Fundulosoma may be distinguished from all species included in Nothobranchius by the forked posttemporal, and the caudal fin extensions of the males. However, since Fundulosoma is monotypic, there is no reason to separate it from the rest of the Nothobranchius species. Therefore, I consider it to be a junior synonym of Nothobranchius. It may be considered as the most primitive Nothobranchius species. The remaining subgenera of Aphyosemion have no generic reference if they are excluded from Aphyosemion and hypothesized to be more closely related to the species of Nothobranchius as defined above. Among the names of subgenera within this group, Fundulopanchax Myers is the oldest and therefore the name which will be used to reference the annual Aphyosemion species. However, it is not implied that this group itself is monophyletic since some members may be more closely related to Nothobranchius than to each other. Therefore, no syn- VOL. 168 onymy of the subgenera is suggested at this time. THE Aplocheilus-PachypanchaxEpiplatys GROUP Species in this group are currently classified in three genera (Aplocheilus, Pachypanchax and Epiplatys). Epiplatys, in turn, is divided into four subgenera (Lycocyprinus, Parepiplatys, Aphyoplatys, and Pseudepiplatys). CHARACTER ANALYSIS: Jaw: Typically, the head is greatly flattened, as is the upper jaw, resulting in a dorsal profile which has been referred to as pikelike. (When first described, Aplocheilus panchax was placed in the genus Esox.) The flattened upper jaw is represented internally by broadly expanded premaxillary ascending processes (fig. 4A). In addition to this upper jaw characteristic there is a concordant feature of the lower jaw which contributes to the flattened appearance of the mouth. As illustrated for Aplocheilus panchax (fig. 27) there is a unique, large coronoid process on the dentary which overlaps the dorsal extension of the articular. There is no such process in the Aphyosemion-Nothobranchius group (e.g., fig. 31C). Caudal fin: The internal supports of the caudal fin differ among adults of the three genera although they are similar in juveniles. In Aplocheilus, the upper hypural plate is divided in two (fig. 2D). In Epiplatys, the upper and lower hypural plates are separate and never fused together to form an hypural fan. In at least one species, E. sexfasciatus, there is evidence of a line of division in the upper hypural plate, suggesting the division seen in species of Aplocheilus. In adult Pachypanchax, the hypural plates are fused to form an hypural fan, as is the case in Nothobranchius, Fundulosoma, and some species of Aphyosemion, a group of Neotropical aplocheiloids, and most, but not all, the cyprinodontoids. However, in juvenile lab-reared Pachypanchax playfairi the dorsal and ventral hypural plates have an evident joint, and there is also such a suture between the dorsal and ventral portions of the upper hypural plate. In Fundulus majal- PARENTI: CYPRINODONTIFORM FISHES 1981 is, a funduline, and in Nothobranchius guentheri, species in which adults have a hypural fan, the juveniles possess a hypural fan, even at the stage of a cartilaginous precursor of the hypural elements. Therefore, given that the aplocheiloids form a monophyletic group, the separate hypurals of the Aplocheilus -Pachypanchax-Epiplatys group are an indication of a secondarily derived condition. Dorsal Gill Arches: The AplocheilusPachypanchax-Epiplatys group exhibits a derived feature of the dorsal gill arches. Typically among cyprinodontiforms, an uncinate process from the third epibranchial articulates via a cartilage with a corresponding process on the fourth epibranchial. The uncinate process of the fourth epibranchial, however, is absent in these three genera. The fourth epibranchial is present as a slender element (fig. 24A) which has no point of articulation to the third. SUMMARY OF DERIVED CHARACTERS 1. Broadly expanded premaxillary ascending processes. 2. Coronoid process on dentary overlaps dorsal extension of articular. 3. Separate upper hypurals at least in juveniles. 4. Loss of the uncinate process on the fourth epibranchial. Aplocheilus-Pachypanchax-Epiplatys On the basis of the following characters, I conclude that Pachypanchax and Aplocheilus are more closely related to each other than either is to Epiplatys; therefore, placing Epiplatys in synonymy with Aplocheilus (Scheel, 1972; Radda, 1973) and excluding Pachypanchax would create a paraphyletic genus. Posttemporal: The posttemporal is typically a forked bone attaching distally to the supracleithrum and proximally to the epiotic dorsally and the exoccipital ventrally. In Old World aplocheiloids, the supracleithrum is not a distinct element, thus the posterior extension of the posttemporal-supracleithrum attaches directly to the cleithrum. Among GROUP RELATIONSHIPS: 395 several groups of cyprinodontiforms, the lower limb of the posttemporal extending to the exoccipital is unossified, and represented only by a ligament. Within the aplocheiloids, this occurs in the genera Aplocheilus and Pachypanchax and in Nothobranchius. It is fully forked in all species of Epiplatys and Aphyosemion examined, as well as in Fundulosoma thierryi. The lower limb being represented by an unossified ligament is most parsimoniously assessed as independently derived in Aplocheilus and Pachypanchax and Nothobranchius. Hypobranchial Teeth: Both Aplocheilus and Pachypanchax have patches of teeth on the second and third pair of hypobranchials, as well as on the fourth ceratobranchials. Such teeth are typically found on the fourth ceratobranchials of atherinomorphs except when lost or reduced as in a group of the Neotropical aplocheiloids. Teeth on the hypobranchial elements, however, have not been found except in these two genera of aplocheiloids and in two cyprinodontoid genera Anableps and Oxyzygonectes. Thus, the presence of hypobranchial teeth is considered to be independently derived in these two cases. Dorsal Ocellus: A dorsal ocellus is present in all females of Aplocheilus and Pachypanchax playfairi. The ocellus is developed also in males of several species of Aplocheilus such as in A. panchax. The dorsal ocellus is absent in all other Old World aplocheiloids. The genus Pachypanchax contains two species, playfairi and homalanotus. The dorsal ocellus is reported to be absent from both males and females of the latter species (Scheel, 1968). Only one specimen of homalanotus was examined, and the species' continued placement in Pachypanchax should perhaps be investigated. Orbital Rim: As discussed for the defining characters of the aplocheiloids, Aplocheilus and Pachypanchax have an orbital rim which is attached ventrally and folded under the frontals dorsally. This is in contrast to the condition in all other aplocheiloids in which the orbital rim is attached all along its perimeter. 396 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 27. Diagrammatic representation of upper and lower jaw of Aplocheilus panchax, lateral view. Maxilla is stippled. Caudal Fin Margin: In Pachypanchax playfairi and a number of species of Aplocheilus, including panchax, there is a dark line of pigment on the caudal fin. Such a margin is not found elsewhere within the aplocheiloids, and as such it is considered uniquely derived. SUBGROUP DEFINITION AND COMPOSITION: Pachypanchax may be distinguished from the species of Aplocheilus and Epiplatys by lateral scales in males which are angled away from the body, and by the fusion of the hypural plates in adults into an hypural fan. The former character refers to a long-known feature of playfairi. The scales stand away from the body in live adult males and give the impression that the individual is suffering from dropsy. Aplocheilus may be distinguished from the species of Pachypanchax and Epiplatys by an attenuate lower jaw and medially greatly expanded premaxillary ascending processes. The species of Epiplatys considered to be part of this monophyletic group may be distinguished only by its lack of the derived characters present in Pachypanchax and Aplocheilus. Epiplatys has a forked posttemporal, a completely attached orbital rim, and lacks a dorsal ocellus, darkened caudal fin margin and teeth on the second and third hypobranchials. Epiplatys, therefore is not definable as a monophyletic group; it may eventually be restricted to the type species, sexfasciatus Gill, and closely allied species. CLADISTIC SUMMARY OF OLD WORLD APLOCHEILOIDS: Old World aplocheiloids are divisible into two groups. The Aplocheilus-Pachypanchax-Epiplatys group is distinguished from the other Old World aplocheiloids by the following derived characters: a broad, flattened, upper jaw effected by expanded premaxillary ascending processes and an expanded coronoid process on the dentary, a bifurcate upper hypural plate 1981 PARENTI: CYPRINODONTIFORM FISHES in juveniles and some adults; and the loss of the uncinate process on the fourth epibranchial. Aplocheilus and Pachypanchax are assessed as sister genera on the basis of the following derived characters: teeth on the second and third hypobranchials; lower limb of the posttemporal represented by an unossified ligament; a dorsal ocellus in females; an orbital rim attached only ventrally; and a dark caudal fin margin. Pachypanchax is defined by two derived characters: fusion of the upper and lower hypural plates into a hypural fan in adults; and, lateral scales of males angled away from the body. Aplocheilus is defined by an attenuate lower jaw and premaxillary ascending processes expanded medially and overlapping. The genus Epiplatys as recognized here cannot be defined as a monophyletic group. Some species currently referred to the group may prove to be more closely related to forms of Aphyosemion and Nothobranchius. The Aphyosemion-Nothobranchius group is defined by the following derived characters: bifid epipleural ribs; attenuate posterior expansion of the vomer; and a reduced chromosome number. The subgenera of Aphyosemion may be grouped into the following two categories: The Aphyosemion group comprising the subgenera Aphyosemion, Archiaphyosemion, Mesoaphyosemion, Kathetys, Diapteron, and Chromaphyosemion, and the genus Adamas. They are all nonannual, 397 possess a dorsal fin of from seven to 14 rays which is situated no farther anterior than opposite the first quarter of the anal fin origin, and have a swimbladder which extends posteriorly to the first one or two hemal spines. The Fundulopanchax group comprising the subgenera Fundulopanchax, Gularopanchax, Raddaella, Callopanchax, Paraphyosemion, and Paludopanchax which shares with the species of Nothobranchius and Fundulosoma the following derived characters: dorsal fin rays increased to 14 or more; dorsal situated opposite the anal fin origin or just slightly in front or behind the origin; and swimbladder not expanded past the first hemal arch. All included species are annual. Monophyly of Aphyosemion and Fundulopanchax is not implied. Fundulosoma and Nothobranchius share the following derived characters: preopercular canal open, not represented by pores; a derived position of the interarcual cartilage and oval eggs. The species of Nothobranchius and its included subgenera may be separated from Fundulosoma thierryi on the basis of the following derived characters: lower limb of posttemporal represented only by an unossified ligament and all fins rounded with no caudal fin extensions. However, thierryi is considered to be the primitive member of the genus Nothobranchius since the recognition of a monotypic genus at this position in the phylogenetic analysis is uninformative with respect to the interrelationships of included species. CYPRINODONTOIDS (GROUP C) The cyprinodontoids as the term is used in this study refers to the fishes of the four viviparous families, the Poeciliidae, Goodeidae, Jenynsiidae and Anablepidae, and the cyprinodontid subfamilies Fundulinae, Lamprichthyinae, Fluviphylacinae, Cyprinodontinae, Aplocheilichthyinae, Orestiatinae and Pantanodontinae (see table 2). The subgroups are referred to using the vernacular names as defined previously. Prior to this study, these fishes have not been considered to- gether as a group without including the aplocheiloids. However, together they form one of the most well-corroborated monophyletic groups of fishes. Together these groups comprise nearly 400 species, slightly less than its sister group, the aplocheiloids. Their diversity includes oviparity, ovoviviparity to viviparity; unicuspid, bicuspid, tricuspid, or no teeth in the jaws; and a size range from the diminutive male Heterandria formosa of the poeciliid 398 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY fishes which matures at a standard length of approximately 8 mm. (Rosen and Bailey, 1963) to the large females of the viviparous Anableps which reach a standard length of over 300 mm. (Miller, 1979). They are found in fresh, brackish and salt water, and are distributed pantropically as well as in temperate Laurasia from North America and as far east as Iran. CHARACTER ANALYSIS: Gill Arches: Several derived characteristics of the gill arches distinguish the cyprinodontoids from the aplocheiloids and all other atherinomorph fishes. The first of these is the presence of just two basibranchials in the ventral gill arch skeleton. In aplocheiloids, as in all other atherinomorphs, as well as most other acanthopterygian fishes, there are three ossified basibranchials. These lie medially in a straight line behind the basihyal and extend posteriorly to the angle created by the fifth ceratobranchials (fig. 11). The basihyal and basibranchials are initially represented in ontogeny by a continuous rod of cartilage known as the copula. This precursor is replaced in ontogeny by separate basihyal and basibranchial ossifications. In the aplocheiloids, as illustrated for Nothobranchius melanospilus (fig. 11 A), the basihyal is followed posteriorly by three ossified basibranchials. In contrast, within all cyprinodontoids, the first ossified basibranchial is absent, whereas the second and third are present in much the same position as those of the aplocheiloids (fig. 11B). Two ossified basibranchials occur elsewhere in the acanthopterygian fishes, notably in synbranchid eels. Rosen and Greenwood (1976) report that the condition of two ossified basibranchials is effected by the fusion of the first basibranchial with the basihyal. In the cyprinodontoids, however, there is no such apparent fusion of the first basibranchial to either the basihyal or the second basibranchial. In addition, the section of the cartilaginous precursor of the first basibranchial is absent in adult cyprinodontoids; thus, the condition of the two ossified basibranchials may be described as the loss of the first basibranchial. Typically among atherinomorphs, as for VOL. 168 most teleosts, the hyoid bar is composed anteriorly of two hypohyals. The two elements, a dorsal and a ventral hypohyal, articulate with the anterior process of the anterior ceratohyal. Typically among aplocheiloids, and most other cyprinodontiforms, there is an extension of the anterior ceratohyal under the ventral hypohyal. This is the case as illustrated for Pachypanchax playfairi (fig. 28 A). In all cyprinodontoids, the dorsal hypohyal is absent. The anterior ceratohyal typically retains its anterior extension under the ventral hypohyal, as in Oxyzygonectes dowi (fig. 28B). However, in the poeciliid fishes, Fluviphylax and procatopines there is no distinct anterior extension of the anterior ceratohyal, and the remaining ventral hypohyal is present as a cap of bone over the end of the anterior ceratohyal as in Procatopus gracilis (fig. 28C). In Pantanodon madagascariensis, there is no extension of the anterior ceratohyal under the ventral hypohyal; however, there appear to be two ossification centers in the cap of cartilage present on its anterior face. These would probably be interpreted as a dorsal and a ventral hypohyal; however, in the light of the other evidence which clearly places Pantanodon as a member of the cyprinodontoids with a derived state of the anterior ceratohyal, I interpret the cartilaginous cap with its two ossification centers as a secondarily derived condition which is most like that described for the poeciliids, procatopines and Fluviphylax. In the aplocheiloids, the typical state of the interarcual cartilage is as an elongate rod approximately equal in length to the epibranchials. It is absent among aplocheiloids in the genus Pterolebias and was found reduced in Rachovia maculipinnis. In all cyprinodontoids, the interarcual cartilage is reduced to approximately one half the length of the epibranchials (fig. 6B). (The reduced condition in maculipinnis is considered secondarily derived within the aplocheiloids.) Jaw and Jaw Suspensorium: In cyprinodontiforms as a whole, as is true for many other, but not all atherinomorphs, there are no dermal jaw suspensorium elements. Sim- 1981 PARENTI: CYPRINODONTIFORM FISHES 399 A FIG. 28. Diagrammatic representation of hyoid bar of A. Rivulus harti; B. Oxyzygonectes dowi; C. Procatopus gracilis. Cartilage is stippled. ilarly, the ectopterygoid is also lacking in many atherinomorphs including cyprinodontiforms, although the identification of this state has not been made consistently in atherinomorph studies. Rosen (1964) illustrated a section of the jaw suspensorium of Xiphophorus helleri, a poeciliid, and identified the ventral extension of the autopalatine as the ectopterygoid, although it is not present as a distinct bone, and no joint lines are visible between it and the autopalatine. In some exocoetoids in which the ectopterygoid is a separate bone (e.g., in Parexocoetus brachypterus) there is also a ven- BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 400 VOL. 168 HYO FIG. 29. Diagrammatic representation of jaw suspensorium of Cynolebias whitei. tral extension of the autopalatine. Thus the ectopterygoid is considered to be lost in certain atherinomorphs including the cyprinodontiforms. The degree of the extension on the autopalatine varies among cyprinodontiforms. In the aplocheiloids, as in most other atherinomorphs, the autopalatine extension is short and does not reach the quadrate. In contrast, in the cyprinodontoids, as illustrated for Procatopus gracilis (fig. 30), the extension of the autopalatine is enlarged and covers part of the quadrate. In addition to having an enlarged ventral process, the head of the autopalatine is set at an angle to its arm. In the aplocheiloids (fig. 29), as is true generally for atherinomorphs, the head of the autopalatine is straight, whereas in the cyprinodontoids (fig. 30) the head of the autopalatine is distinctly offset. There is also a bony flange which extends posteriorly giving the anterior extension of the autopalatine the shape of a ham- merhead. In a group of the cyprinodontoids, which include goodeids, Empetrichthys, and Crenichthys, the head of the autopalatine is reduced to a nubbin, a condition considered to be secondarily derived. It is readily distinguished from the aplocheiloid condition in that the head is blunt, rather than slender. A third derived character of the jaw suspensorium of the cyprinodontoids is the loss of the metapterygoid (see figs. 29 and 30). The metapterygoid is also absent in the adrianichthyoids Oryzias and Horaichthys; but, since they possess none of the derived characters for cyprinodontiforms, their loss of the metapterygoid is inferred to be independent. The cyprinodontiforms exhibit a derived state of the premaxilla characterized by the two-part alveolar process. The superficial division of the adductor mandibulae inserts via a tendon to the middle of the arm of the 401 PARENTI: CYPRINODONTIFORM FISHES 1981 t-HYO POP FIG. 30. Diagrammatic representation of jaw suspensorium of Procatopus gracilis. maxilla, whereas the more anterior layers insert on the posterior extension of the alveolar arm. In the cyprinodontoids, the alveolar arm is distinctly S-shaped (fig. 3B), as a result of bending and enlarging of the post-maxillary process. This is the condition typical of cyprinodontoids, and although the arm undergoes modifications in several of its subgroups, it can always be distinguished from that of the aplocheiloids by the posterior indentation. In the aplocheiloids, the dentary is a relatively thin bone, which carries a distinct sensory canal (fig. 31C). In Menidia (fig. 3IB), as in many other atherinoids, there is a large coronoid process on the dentary; yet, ventrally, the bone is unexpanded as in the aplocheiloids. Similarly, in Oryzias (fig. 31 A) the dentary is unexpanded. In all cyprinodontoids, the dentary is a robust bone (fig. 33) expanded medially, and therefore, carrying the sensory canal along its midline. There are no ethmomaxillary ligaments present in cyprinodontoids as there are in aplocheiloids. Similarly, there are no ligaments extending from the interior arms of the maxillaries to the middle of the rostral cartilage. In addition, there is no meniscus between the premaxilla and the maxilla. These elements are present, however, in the aplocheiloids and atherinoids (Alexander, 1967b). Hence, their absence in cyprinodontoids is considered derived. Vomer: The vomer bears teeth in all aplocheiloid species. The state of this character is variable, however. When present, the teeth are usually in a round patch at the anteromedial extension of the vomer, as in Rivulus (fig. 17C). In Aplocheilus (fig. 17B), the teeth extend across the anterior edge of the vomer. In cyprinodontoids and in ather- 402 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 31. Diagrammatic representation of lower jaw, lateral view: A. Oryzias javanicus; B. Menidia menidia; C. Aphyosemion occidentale. Cartilage is stippled, Meekers cartilage is the elongate central element. inoids the vomer does not possess a medial extension and there are never any teeth on the vomer. Vomerine teeth might, therefore, be derived for aplocheiloids and lost independently in some aplocheiloids and in cyprinodontoids. The usefulness of this char- acter, however, is dubious because its distribution coincides with no other known character. Loss of First Dorsal Fin Ray: Another derived character which defines the cyprinodontoids as a monophyletic group pertains to 1981 PARENTI: CYPRINODONTIFORM FISHES 403 FIG. 32. Diagrammatic representation of first two proximal dorsal radials articulating with two dorsal fin rays in A. Pachypanchax playfairi; with one dorsal fin ray in B. Adinia xenica. the number of dorsal fin rays. In all aplocheiloids, there is one dorsal fin ray articulating with each of the first two dorsal radials (fig. 32A). The first dorsal ray is often rudimentary; nonetheless, it is present. In all cyprinodontoids (fig. 32B) the first dorsal ray is apparently lost, and the second remaining ray articulates with the first two proximal dorsal radials. SUMMARY OF DERIVED CHARACTERS 1. Two basibranchials in the ventral gill arch skeleton. 2. Loss of the dorsal hypohyal. 3. Reduction of interarcual cartilage to one half its length, relative to that of the aplocheiloids, and the associated place- 4. 5. 6. 7. 8. 9. 10. 11. 12. ment of the first epibranchial closer to the second pharyngobranchial. Autopalatine with its anterior extension bent sharply and hammer-shaped. Extension of the autopalatine ventrally forming an anterior covering of the quadrate. Metapterygoid absent. Alveolar arm of premaxilla S-shaped. Dentary expanded medially and robust. Loss of an ethmomaxillary ligament. Loss of a ligament from the interior arms of the maxillaries to the middle of the rostral cartilage. Loss of a meniscus from between the premaxilla and the maxilla. Loss of an anterior dorsal fin ray result- 404 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 33. Diagrammatic representation of lower jaw, lateral view, A. Profundulus punctatus; B. Characodon lateralis; C. Crenichthys baileyi; D. Empetrichthys latos pahrump. ing in the articulation of the first dorsal fin ray with first two proximal radials. RELATIONSHIPS OF THE CYPRINODONTOIDS: Jaw and jaw suspensorium: The most primitive type of jaw structure within cyprinodontoids is that found in the Central American genus Profundulus. The alveolar arm of the premaxilla is indented posteriorly, forming an S-shaped distal process. The dentary (fig. 33A) is expanded medially forming a robust lower jaw. There are no large processes on the dentary or the articular, and the retroarticular is of moderate size. Premaxillary ascending processes in Profundulus are flat and broad (fig. 5B). At their tips sits the large, rectangular rostral cartilage. The interior arms of the twisted maxillaries abut the rostral cartilage and are bound to it by collagen fibers. No ligament from the interior arms to the cartilage has been found, as present in the aplocheiloids and atherinoids, as reported by Alexander (1967b). Similarly, there is no ethmomaxillary ligament, nor is there a meniscus between the premaxilla and maxilla. In other cyprinodontoids, the rostral cartilage is re- duced relative to the condition found in Profundulus. Alexander (1967b) stated that in Fundulus the rostral cartilage is Y-shaped and therefore comes in contact with the hooks on the interior arms of the maxillaries. With the benefit of the counterstaining technique employed throughout this study, it has been determined that the rostral cartilage is not Y-shaped but is represented by, at most, four small discs of cartilage in the fundulines; one is situated posterior, and two smaller elements anterior, to a larger medial cartilage located between the internal hooks of the maxillaries (fig. 5C). These bits of cartilage are held together and to the maxillary by connective tissue fibers, forming what is presumably the "Y-shaped" rostral cartilage of Alexander. Thus, in fundulines as well as in all other cyprinodontoids (excluding Profundulus) there has been a loss of contact between the inner arms of the maxillaries and the rostral cartilage, a condition associated with reduction of the cartilage. In both the fundulines and the Mediterranean genus Valencia which has heretofore been classified in the same subfamily as the 1981 PARENTI: CYPRINODONTIFORM FISHES 405 FIG. 34. Diagrammatic representation of upper and lower jaw, lateral view, of Fundulus diaphanus. Maxilla is stippled. fundulines, the premaxillary ascending processes are narrow and elongate (fig. 5D). Narrow premaxillary processes are characteristic of the group of cyprinodontoids excluding Profundulus. The elongate processes of fundulines and Valencia are considered as stages in a transition series from the broad processes typical of Profundulus and the aplocheiloids to the short and narrow processes of the large subgroup of cyprinodontoids comprising the following: Jenynsia, Anableps, Oxyzygonectes, the poeciliids, procatopines, Pantanodon, Fluviphylax, the goodeids, Empetrichthys, Crenichthys, Orestias, the cyprinodontines, Cubanichthys and Chriopeoides. This large group, plus 406 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 35. Diagrammatic representation of the upper jaw in A. Anableps dowi; B. Oxyzygonectes dowi; C. Aplocheilichthys johnstoni; D. Jenynsia lineata. Valencia, has attenuate interior arms of the maxillaries, rather than the broad tips associated with Profundulus and the aplocheiloids. In addition, they have a maxilla which is straight rather than characteristically twisted as in fundulines, aplocheiloids, and Profundulus. The arm does not have a pronounced bend anterior to the autopalatine (as in Fundulus diaphanus, fig. 5C), but, it is rather straight and often has a pronounced flat dorsal process which extends anteriorly over the premaxillary ascending processes (fig. 5D). The fundulines are unique among cyprinodontiforms in having pronounced hooks on the interior arms of the maxillaries (figs. 5C, 34). In addition, the interior arms are directed anteriorly, rather than medially as in other cyprinodontoids. These characters may be considered derived for the fundulines, and therefore define them as a monophyletic group. Valencia shares the derived characters of the rest of the cyprinodontoids as described above; that is, a straight maxilla with attenuate interior arms, and the development of a dorsal extension over the premaxillary ascending processes. Valencia is unique among cyprinodontiforms in having very long attenuate dorsal processes of the maxillaries (fig. 5D). In other cyprinodontoids of the large group delimited above, the dorsal processes are rounded when present. Since there are no such dorsal processes in the fundulines, Profundulus or aplocheiloids, the polarity of the character is ambiguous. The elongate dorsal processes of Valencia may represent the primitive state of the processes 1981 PARENTI: CYPRINODONTIFORM FISHES 407 FIG. 36. Diagrammatic representation of upper and lower jaw, lateral view, of Oxyzygonectes dowi. Maxilla is stippled. which are further reduced in the large subgroup; or, the reverse may be true. The polarity of this character may be resolvable with an ontogenetic series of Valencia. The large subgroup minus Valencia is defined by having narrow and shortened premaxillary ascending processes, and the rostral cartilage reduced or absent. The dorsal processes of the maxilla are rounded when present. This subgroup may itself be subdivided into two monophyletic groups. The poeciliids, procatopines, Fluviphylax, Pantanodon, and Oxyzygonectes, Jenynsia and Anableps form a monophyletic group based on three derived jaw characters: the dorsal processes of the maxillaries are indented laterally to form nearly fan-shaped processes; the distal arm of the maxilla is expanded; and the retroarticular is enlarged. The dorsal process of the maxilla, as in the procatopine Aplocheilichthys johnstoni (fig. 35C) and for Jenynsia lineata (fig. 35D), has a distinct lateral indentation. The result is a distinct fan-shaped process which projects over the triangular premaxillary ascending processes. The dorsal process is found in this state, as well, in Anableps (fig. 35A), Oxyzygonectes (fig. 35B), the remaining procatopine genera, and the majority of the poeciliids (e.g., as illustrated in Rosen and Bailey, 1963). In both Pantanodon and Fluviphylax the dorsal processes are weakly formed; yet, on the basis of characters to be discussed they are considered to be part of this monophyletic group, and their weakly formed processes are considered to be secondarily derived. The distal arm of the maxilla is enlarged at its most ventral extension (e.g., as in Oxyzygonectes dowi, fig. 36) in all members of the group excluding the procatopine genera Procatopus and Hypsopanchax, and a group of species of Aplocheilichthys (e.g., as in Procatopus gracilis, fig. 37), in which the distal arm of the maxilla is shortened relative to its condition in the other members of this group. Similarly, the retroarticular is extremely elongate in Anableps, Jenynsia (fig. 38B), Oxyzygonectes (fig. 36), and the procatopine 408 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 37. Diagrammatic representation of upper and lower jaw, lateral view, of Procatopus gracilis. Maxilla is stippled. genera Aplocheilichthys (fig. 38C) and Lamprichthys, and moderately elongate in Tomeurus, the presumed primitive poeciliid. Within Pantanodon, Fluviphylax, and Procatopus (fig. 37) and Hypsopanchax, the retroarticular is reduced. Thus, the premaxillary and retroarticular characters appear to be correlated. The elongate retroarticular and expanded arm of the premaxilla are a general characteristic of the group, but both of these elements are sec- ondarily reduced in Hypsopanchax, Procatopus, Pantanodon, and Fluviphylax. Within this large group, the poeciliids, Fluviphylax, Pantanodon and the procatopines are distinguished by the formation of a greatly enlarged dentary. The less expanded dentary of Jenynsia (fig. 38B), Oxyzygonectes (fig. 36) and Anableps is the condition primitive for all cyprinodontoids. In the poeciliids, procatopines, Pantanodon and Fluviphylax, the dentary is much 1981 PARENTI: CYPRINODONTIFORM FISHES 409 FIG. 38. Diagrammatic representation of lower jaw, lateral view, of A. Valencia hispanica; B. Jenynsia lineata; C. Aplocheilichthys johnstoni. Cartilage is stippled, Meckel's cartilage is elongate medial element. more expanded, especially at its most anterior end, e.g., in Procatopus gracilis (fig. 37). The dentary in this case continues to carry a sensory canal; however, the ossified enclosure of the canal is reduced relative to that in other cyprinodontiforms. The second division of these cyprinodontoids comprises the genera Empetrichthys, Crenichthys, Cubanichthys, Chriopeoides, Orestias, the goodeids, and cyprinodontines. The mouth of this group is smaller than that in any other group of cyprinodontiforms. The premaxillary ascending processes are short and attenuate, rather than triangular as in the former group. In the goodeids and the two North American genera Empetrichthys and Crenichthys, the dorsal process of the maxilla is present yet weakly formed (fig. 39). The result is a maxilla which has a small cup-shaped process medially to receive the premaxillary ascending process. Because these three taxa share other characters with the more derived cyprinodontiforms, the dorsal process is inferred to be reduced rather than primitively unformed. 410 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 ASC ALV FIG. 40. Diagrammatic representation of premaxilla, lateral view in A. Characodon lateralis; B. Crenichthys baileyi; C. Empetrichthys latos pahrump. 39. Diagrammatic representation of the upper jaw in A. Empetrichthys merriami; B. Crenichthys baileyi; C. Characodon lateralis. FIG. Among these three, the distal arm of the premaxilla is straight, rather than S-shaped, although the posterior indentation of the alveolar arm is well-formed (fig. 40). A third unique jaw characteristic of Empetrichthys, Crenichthys, and the goodeids is the greatly reduced articular that possesses no medial extension to carry the sensory canal (fig. 33B, C, D). A fourth unique character, mentioned previously, is the reduction of the anterior arm of the autopalatine, with no anterior or posterior extensions. These four jaw and jaw suspensorium characters of goodeids, Empetrichthys, and 1981 FIG. 41. fasciatus. PARENTI: CYPRINODONTIFORM FISHES 411 Diagrammatic representation of the upper jaw in A. Cubanichthys cubensis; B. Aphanius Crenichthys, along with characteristics of other systems, unite them into a monophyletic group. In the nominal genera Cubanichthys, Chriopeoid.es, Orestias and in the cyprinodontines, the dorsal processes of the maxillaries are expanded medially, and nearly meet in the midline (fig. 41). There is also a distinct groove running down the middle of the dorsal process. The large distal arm of the maxilla is correlated in this group (fig. 42) with the development of a robust upper jaw. Cubanichthys and Chriopeoides are hypothesized to be the primitive members of this assemblage because they possess two primitive characteristics of the jaws found modified in the remaining members. Both nominal genera possess several rows of teeth on the upper and lower jaw; there is a prominent outer row with smaller, scattered inner jaw teeth not forming regular rows. Also, Meekers cartilage is narrow posteriorly where it inserts into the medial articular process (e.g., as in fig. 38). In Orestias and the cyprinodontines, the teeth are present in a single outer row on both the upper and lower jaws. These teeth are unicuspid and bicuspid in Orestias, unicuspid in Kosswigichthys, and tricuspid in remaining cyprinodontines. Teeth occur in a single outer row independently in one other species of cyprinodontiform, the funduline Lucania parva. In addition to a single row of outer teeth, the cyprinodontines and Orestias also have a derived lower jaw which is characterized by the posterior expansion of Meckel's cartilage (e.g., as in Aphanius fasciatus, fig. 43A). The cartilage is expanded so that it covers a large portion of the articular, in contrast to the state of the cartilage in other cyprinodontiforms (e.g., figs. 27, 38) in which the cartilage is present as a rod of uniform width. In Anatolian cyprinodontines (e.g., Aphanius, fig. 43A) there is a medial extension of the dentary which projects anteriorly. In the South American genus Orestias (fig. 43B) the dentary is even further expanded to form a medial shield of bone. The condition in Orestias is considered to be the most derived condition of this transition series (i.e., from the typical condition in Chriopeoides, fig. 42, to the expanded condition of Aphanius, to the fully expanded condition of Orestias). The dentary characteristics are correlated with those of the gill arches, shoulder girdle and pattern of squamation, to be discussed. The characteristic lower jaw of these two groups is characterized by a robust dentary and recession of the urohyal and branch- 412 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 42. Diagrammatic representation of upper and lower jaw, lateral view of Cubanichthys (Chriopeoides) pengelleyi. iostegal rays (fig. 141). In lateral view, the mouth cleft is nearly vertical (fig. 14G). Aphanius fasciatus has tricuspid jaw teeth, as do other Old World and all New World cyprinodontines. The character transformation series described above for the dentary indicates that if tricuspid teeth can be used as a derived character, it is only at the level of defining the cyprinodontines and Orestias as a monophyletic group, with a reversion to unicuspid teeth in some Orestias and in Kosswigichthys. Tricuspid outer teeth occur in one other group of cyprinodontiforms, fishes of the genera Jenynsia, Anableps, and Oxyzygonectes. The teeth of Jenynsia are distributed in one large outer row and several smaller scattered in indistinct inner rows. All the jaw teeth are tricuspid. However, the shape of the outer teeth of Jenynsia varies from distinctly tricuspidate with the inner cusp just slightly longer than the middle (fig. 44B) to a faintly tricuspidate form in which the lateral shoulders are only weakly formed (fig. 44 A). In Oxyzygonectes (fig. 36) adults have a 1981 PARENTI: CYPRINODONTIFORM FISHES 413 mdn FIG. 43. Diagrammatic representation of lower jaw, lateral view, of A. Aphanius fasciatus; B. Orestias sp. Cartilage is stippled, Meckel's cartilage is the enlarged medial element. row of very large recurved unicuspid teeth, and a dense inner patch of teeth which appear to be distributed in about five or six rows. These inner jaw teeth are all tricuspid in both juveniles and adults. The teeth are so closely packed that on a cursory examination they appear to be villiform. The outer teeth of juvenile Oxyzygonectes are weakly tricuspidate. Thus, the jaw dentition of Oxyzygonectes and Jenynsia is apparently very much the same, with Oxyzygonectes losing the lateral cusps of the outer row, and having more inner jaw teeth. The jaw dentition of an adult Anableps consists of one large outer row of recurved teeth and several smaller scattered inner rows of unicuspid teeth. The inner jaw teeth have what appear to be weakly formed lateral shoulders. The upper jaw of an adult Anableps is very derived (fig. 35A). There are only weakly formed premaxillary ascending processes and the premaxillaries form an arc. The maxilla is elongated medially, however, the dorsal process of the maxilla and expanded dis- 414 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 44. Sketch of two forms of tricuspid teeth in the genus Jenynsia (see text for discussion). tal arm distinctive of the monophyletic group to which it is assigned are prominent. Another unique feature is the dumbbell shape of the rostral cartilage, unknown in other cyprinodontiforms. Also, a block of cartilage sits between the autopalatine and the maxilla (fig. 35A) termed here the subautopalatine cartilage. Such a block is often found in this position in atherinomorph fishes; its presence in Anableps is therefore considered primitive. Juvenile Anableps show all the specializations of the adults, therefore, an embryo of Anableps dowi, the presumed most primitive species of the genus (Miller, 1979) was examined. The yolk sac was removed and the specimen counterstained. The outer teeth of the embryo have distinct lateral shoulders and there is a very narrow medial cusp, they differ little from the weakly tricuspidate teeth in Jenynsia. Furthermore, in the embryo, triangular-shaped ascending processes like those in Jenynsia and Oxyzygonectes are present on the premaxillaries. Hence, on the basis of dentition, the three genera, Jenynsia, Anableps, and Oxyzygonectes are hypothesized to form a monophyletic group. GILL ARCHES: The structure of the branchial skeleton has been used in recent years to deduce phylogenetic relationships because it is both constant within large groups FIG. 45. Diagrammatic representation of dorsal gill arches, ventral view, of Pantanodon madagascariensis. Cartilage is stippled. and quite variable among them (Nelson, 1969; Rosen, 1973). Except in the unusual Pantanodon (fig. 45) and certain poeciliids, dorsal gill arch anatomy among cyprinodontoids varies little from the basic structure exhibited by Profundulus punctatus (fig. 6B). In Profundulus, the interarcual cartilage is reduced relative to the condition in aplocheiloids. The three pharyngobranchial 1981 PARENTI: CYPRINODONTIFORM FISHES 415 FIG. 46. Diagrammatic representation of dorsal gill arches, ventral view, of A. Fundulus diaphanus; B. Lucania parva. Cartilage is stippled. toothplates (associated with pharyngobranchials 2,3, and 4) are separate elements. The cartilaginous points of articulation are relatively narrow. Species in the subgenera Zygonectes, Xenisma, and Fundulus of the genus Fundulus differ from Profundulus in having the cartilaginous point of articulation of the second pharyngobranchial toothplate greatly expanded laterally to produce a broad head for the articulation of the interarcual cartilage (fig. 46A). In the subgenus Plancterus, and in the funduline genera Adinia, Leptolucania and Lucania (fig. 46B), the cartilaginous point of articulation is not enlarged. This is also the case in Valencia (fig. 47A) which exhibits the primitive condition for the cyprinodontoids. Among the more derived cyprinodontoids, the structure of the dorsal gill arches differs most from the general condition in Pantanodon and some derived poeciliid genera. In Pantanodon madagascariensis (fig. 45), the second pharyngobranchial toothplate is greatly expanded into a sheet of bone. There are no teeth in sockets on the toothplate; however, toothlike structures lie above it suspended in connective tissue. The third and fourth pharyngobranchial toothplates are fused into one large toothbearing element. The teeth are arranged in discrete rows, with tricuspid teeth being found on the posterior five rows. Epibranchials one through three are absent, as is the interarcual cartilage. (Also, the hypobranchials of Pantanodon are reduced or absent, as illustrated by Rosen, 1965.) The expanded second pharyngobranchial toothplate has been found in no other atherinomorph genus examined. Fusion of the third and fourth pharyngobranchial toothplates occurs within a group of cyprinodontines, but otherwise their structure is basically that of the general form. Among the poeciliids, teeth on the third and fourth pharyngobranchial toothplates are often arranged in discrete rows, and even tricuspid teeth are present. In these poeciliids however, the epibranchials and interarcual cartilage are present, as is a more primitively shaped second pharyngobranchial toothplate. In this study, Pantanodon is BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 416 VOL. 168 FIG. 47. Diagrammatic representation of dorsal gill arches, ventral view, of A. Valencia hispanica; B. Empetrichthys latos pahrump. Cartilage is stippled. PB3 FIG. 48. Diagrammatic representation of dorsal gill arches, ventral view, of A. Procatopus gracilis; B. Tomeurus gracilis. Cartilage is stippled. 1981 PARENTI: CYPRINODONTIFORM FISHES considered to be a close relative of the poeciliids; however, its close relationship is not based on gill arch morphology. Tomeurus has dorsal gill arches of the structure primitive for cyprinodontoids (fig. 48B), although the epibranchials are reduced and no interarcual cartilage has been found. These characters are considered derived for the genus, since the cartilage and more robust epibranchials are found in more derived poeciliids. Therefore, either Pantanodon is a poeciliid which lost its gonopodium, or the similar gill arch structure of the poeciliids and Pantanodon are independently derived. On the basis of the distribution of all derived characters, the latter hypothesis is accepted here. Empetrichthys (fig. 47B) and Crenichthys exhibit a peculiar shape of the first epibranchial. The bone is nearly Y-shaped resulting from an indentation at its base. This type of first epibranchial has not been found elsewhere within cyprinodontiforms. In the cyprinodontines and Orestias, the second pharyngobranchial is offset to the third, as in Cyprinodon variegatus (fig. 49). This change in orientation of the pharyngobranchial excludes the cartilaginous point of articulation from the ventral toothed surface of the pharyngobranchial toothplates. In addition, the fourth pharyngobranchial toothplate is reduced. However, such a reduction is not unique to this group, as the toothplate is also reduced in Procatopus (fig. 48A) and other procatopines. In Cyprinodon (fig. 49) the third and fourth pharyngobranchial toothplates are fused into a single toothbearing element. The teeth are arranged in rather discrete rows. Such fusion occurs in many, perhaps most, individuals of the New World cyprinodontine genera Cyprinodon, Jordanella, Garmanella, Megupsilon, Floridichthys, and Cualac, and in Orestias. Although the occurrence of some individuals with unfused toothplates makes the upper pharyngeal character difficult to use, the regular arrangement of the teeth is a constant defining character of all of these genera. In Floridichthys carpio (fig. 50A) there is a distinct first pharyngobranchial cartilage as well as a toothplate which bears a patch of 417 FIG. 49. Diagrammatic representation, dorsal gill arches, ventral view, of Cyprinodon variegatus. Cartilage is stippled. teeth. In Cualac tessellatus (fig. 50B) there is no toothplate yet there is a distinct separate cartilage which sits at the anterior tip of the first epibranchial cartilage. An element in this position has been found in only one other species of cyprinodontiform, Cynolebias elongatus. In this Neotropical aplocheiloid, there is a distinct cartilage as well as a bony toothbearing element. Its condition is comparable to that of Floridichthys. Among atherinomorph fishes, no first pharyngobranchial toothplate is found except in the genera Cynolebias and Floridichthys. An ossified first pharyngobranchial is present among some atherinoid fishes, including species of the genera Melanotaenia and Menidia. Otherwise among atherinomorphs, first pharyngobranchial elements are absent. Their appearance within these two rather unrelated groups of cyprinodontiforms poses a problem for interpretation. If the condition is a retained primitive character, then the most parsimonious interpre- 418 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 TPBl? S^) FIG. 50. Diagrammatic representation of dorsal gill arches, dorsal view, of A. Floridichthys carpio; B. Cualac tessellatus. Cartilage is stippled. tation in light of all other data would be that the elements are lost individually in all other groups of cyprinodontiforms. So many independent losses, however, presuppose far more evolutionary events than if it is assumed that these elements are uniquely derived twice among cyprinodontiforms, once in C. elongatus and again in Floridichthys and Cualac, thereby supporting a sister group relationship of the last two genera. The condition may also define a subgroup of Cynolebias species. PECTORAL GIRDLE: The pectoral fins of cyprinodontiforms are described as typically lowset, with the corresponding radials situated ventrally rather than dorsally. There is a large, scale-shaped first postcleithrum and a thin third postcleithrum. The posttemporal may have an ossified lower limb, or a limb represented solely by a ligament; this character is used only at the lower levels of phytogeny reconstruction since it is not correlated with larger sets of characters used to delimit major groups. Shoulder girdles are lowset in all cyprinodontiforms except the poeciliids, Fluviphylax, Pantanodon, and the procatopines. Within this group, the pectoral fins are distinctly highset (e.g., as in Tomeurus gracilis, fig. 51; and Heterandria bimaculata, fig. 52; and the procatopines Procatopus glaucicaudus, fig. 53) as opposed to the lowset fins of, for example, Rivulus (fig. 54), Cualac (fig. 55) and Lucania (fig. 56). The highset pectoral fins are related to the placement of the radials in a dorsal position on the scapulocoracoid, and a gently arched dorsal limit of the scapula and cleithrum (fig. 8C, D). This is correlated with a loss of the first postcleithrum which is wanting in all members of the group except for some nominal species of the genus Aplocheilichthys. The structure of the shoulder girdle of Profundulus (fig. 8A) is the general condition, as in the aplocheiloids with the radials situated ventrally. The pectoral fins are distinctly lowset in the genera Anableps, Jenynsia, and Oxyzygonectes, as well as other cyprinodontiforms; however, as stated, the pectoral fins are generally highset in most other groups of afherinomorph fishes. Since the derived form of the pectoral fins has been interpreted as lowset within cyprinodontiforms, the highset pectoral fins may only be interpreted as sec- 1981 PARENTI: CYPRINODONTIFORM FISHES 419 FIG. 51. Sketch of body form and fin position of Tomeurus gracilis, male above, female below. Dotted line approximates base of hypural plate. (After Rosen and Bailey, 1963.) ondarily derived in poeciliids, procatopines, Pantanodon and Fluviphylax. SKULL ANATOMY: A constant feature of the skull of aplocheiloids is the presence of a lateral facet on the anterior surface of the lateral ethmoid which articulates with the head of the autopalatine (fig. 17). Such an extension is present in the cyprinodontoids only in the genus Profundulus (fig. 57A). This character was one of several Farris (1968) used to separate the species of Fundulus from Profundulus; he reported the process as absent in all species of Fundulus ex- amined. The presence of this character in all aplocheiloids and Profundulus suggests that it is a primitive character for cyprinodontiforms. Thus, its absence or reduction in all other cyprinodontoids is evaluated as a derived character supporting their monophyly. The generalized state of the size and position of the lateral ethmoid is exemplified by Tomeurus gracilis (fig. 16B). Among procatopines, the lateral ethmoid is expanded medially under the broad arm of the parasphenoid, as in Aplocheilichthys johnstoni (fig. 16C). (Compare this expan- 420 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 52. Sketch of body form and fin position of Heterandria bimaculata, male above, female below. Dotted line approximates base of hypural plate. (After Rosen, 1979.) sion of the lateral ethmoid with that of the Neotropical aplocheiloids, fig. 17C.) Medial expansion of the lateral ethmoid is accompanied by a change in its orientation relative to the frontal bones in Empetrichthys, Crenichthys, the goodeids, Cubanichthys, and Chriopeoides, the cyprinodontines and Orestias. As in the goodeid, Characodon lateralis (fig. 57D), the lateral ethmoid is oriented such that the greater part of the element lies anterior to the limit of the frontals. This may be compared with the general condition in cyprinodontoids as in Profundulus punctatus (fig. 57A), in which the outer flange of the lateral ethmoid is expanded, rather than narrow as in Characodon. Among the fundulines, the lateral ethmoid is also expanded under the parasphenoid (fig. 16A). However, the lateral ethmoid not only lacks the facet for articulation of the autopalatine, but the autopalatine does not come in contact with the lateral ethmoid. Rather, the fundulines' pronounced snout is effected not only by the anteriorly projecting ventral arms of the maxillaries, but by the extension of the autopalatines to a position lateral to the enlarged vomer as well. The mesethmoid is cartilaginous in aplocheiloids. In addition, it is cartilaginous among the cyprinodontoids in Pantanodon and the procatopines (it is ossified in Fluviphylax) as well as in the Anatolian cyprinodontines. The cartilaginous mesethmoid is considered a derived condition defining the aplocheiloids; among the cyprinodontoids, its independent occurrence within two unrelated groups is convergent. The group consisting of Empetrichthys, Crenichthys, the goodeids, Cubanichthys, and Chriopeoides, the cyprinodontines and Orestias possess another derived feature of the skull; viz., a reduced autopterotic fossa (fig. 57B, C, D). Uyeno and Miller (1962) used the narrow fossa to separate Empetrichthys and Crenichthys from Profundulus which has an extremely wide fossa (fig. 57A). 1981 PARENTI: CYPRINODONTIFORM FISHES 421 FIG. 53. Sketch of body form and fin position of Procatopus glaucicaudus, male above, female below. Dotted line approximates base of hypural plate. (After Clausen, 1959.) However, they did not compare this condition to its state in other cyprinodontiforms. The fossa of Profundulus is wider than in any other cyprinodontiform and may be considered an autapomorphy of the genus. Enlarged supraoccipital and epiotic processes occur among many groups of acanthopterygian fishes. The general condition of the supraoccipital crests among atherinomorph fishes is paired (Rosen, 1964); among cyprinodontiforms this is the case except in the two monotypic genera Cubanichthys and Chriopeoides. In these genera the supraoccipital crest is a large, single process which extends above the dorsal profile (fig. 58). Thus, the sister group relationship of these two genera is supported again. Another unique form of the supraoccipital processes is shared by Anableps, Oxyzygo- nectes and Jenynsia. As illustrated for Oxyzygonectes dowi (fig. 59) the crests are greatly elongate and are separated by a distinct notch from the dome over the foramen magnum. In contrast, supraoccipital crests are present in Profundulus (fig. 60), yet they abut the dorsal wall of the foramen magnum, rather than being separated from it by a notch. The states of the first vertebra in oviparous cyprinodontiforms have been described, although somewhat erroneously, by Sethi (1960). All the aplocheiloids have a complete neural spine on the first vertebra (fig. 26). Among the cyprinodontoids, the neural arch of the first vertebra is open, and therefore, does not form a neural spine (fig. 60) in Profundulus, Valencia, Empetrichthys, Crenichthys, the fundulines, and goodeids. 422 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 54. Sketch of body form and fin position of Rivulus beniensis, male above, female below. Dotted line approximates base of hypural plate. (After Klee, 1965.) Among the procatopines, as well as in Fluviphylax, the neurapophyses meet in the midline; no distinct spine is formed, however. The condition is interpreted as a reduction in the neurapophyses and a secondarily derived medial fusion. In this group as well as in the aforementioned cyprinodontiforms, the basioccipital and exoccipital condyles are all well-formed. In all poeciliids, as well as Pantanodon, there are no exoccipital condyles. The attachment of the first vertebra to the skull in Tomeurus (fig. 61) involves the forward expansion of the neurapophyses around the base of the foramen magnum. The arch is open and the first vertebra articulates with the skull only via the basioccipital condyle. In Pantanodon, as well as some of the more derived poeciliids such as those of the genus Poecilia, the neurapophyses are even more expanded anteriorly and applied to, and fused with, the skull. This characteristic attachment of the first vertebra must be considered independently derived in both Pantanodon and the poeciliids if the monophyly of the poeciliids based on the presence of a gonopodium and other reproductive specializations is accepted. A superficially similar condition of the attachment of the first vertebra to the skull occurs in the New World cyprinodontines of the nominal genera Cyprinodon, Megupsilon, Jordanella, Floridichthys, Cualac, and Garmanella. There is no spine formed by the neurapophyses of the first vertebra. Instead, the neurapophyses are slightly expanded, brought forward, and applied to the skull (fig. 62). Th.i exoccipital condyles are lacking 1981 PARENTI: CYPRINODONTIFORM FISHES 423 FIG. 55. Sketch of body form and fin position of Cualac tessellatus, male above, female below. Dotted line approximates base of hypural plate. (After Miller, 1956.) and, in addition, the supraoccipital forms the roof of the foramen magnum. In all other cyprinodontiforms, as well as in the poeciliids and Pantanodon, the supraoccipital is excluded from formation of the foramen magnum. Also, the form and position of the neurapophyses is quite different between the poeciliids and New World cyprinodontines. In poeciliids, they are greatly expanded and form a trough in which the supraoccipital region of the skull sits; in cyprinodontines, the neurapophyses are simply applied to the skull and provide reinforcement yet form no trough similar to that of the poeciliids. In Orestias and the Anatolian cyprinodontines the exoccipital condyles are present as in Profundulus, yet reduced. The neurapophyses of the first vertebra are also reduced, as in the New World cyprinodontines, and may or may not meet in the midline. The vomer is absent in Pantanodon, Fluviphylax, the procatopines (except Poropanchax, Lamprichthys and species of Aplocheilichthys), and the South American Orestias. The vomer is hypothesized to be lost independently at least twice among cyprinodontiforms, once in Orestias, and once in the procatopines, Fluviphylax and Pantanodon. The significance of its distribution is discussed in the following section. Parietals are absent in two groups of cyprinodontoids. They are lacking in Orestias and the cyprinodontines, as well as the procatopines, Fluviphylax and Pantanodon. Their absence in these two groups is considered to be an independent loss. In addition, their absence from more derived members of the poeciliids is secondarily derived since parietals are present in Tomeurus. Axial Skeleton: The first pleural rib arising on the parapophyses of the second vertebra 424 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 56. Sketch of body form and fin position of Lucania parva, male above, female below. Dotted line approximates base of hypural plate. (After Hubbs and Miller, 1965.) has been described as a derived character of cyprinodontiforms. This state occurs in all members of the group except the funduline genus Adinia in which the rib arises on the parapophyses of the first vertebra. Since the general state among acanthopterygians is for the rib to be on the third vertebra, this case in Adinia is hypothesized to be a further derived state in the transition series, and serves as a defining character of the genus. The parapophyses themselves are generally robust, with the pleural rib inserting into a furrow in the posterior face of the process. Within Orestias and the cyprinodontines, the transverse processes are reduced to cupshaped processes (Sethi, 1960) into which the pleural ribs insert. This reduction is considered as another derived character of Orestias and the cyprinodontines. Pleural ribs by definition occur only on parapophyses of abdominal vertebrae and not on caudal vertebrae. However, within Pantanodon, the procatopines and some poeciliids including Xiphophorus and Poe- FIG. 57. Diagrammatic representation of the skull, ventral view, in A. Profundulus punctatus; B. Cyprinodon variegatus; C. Empetrichthys latos pahrump; D. Characodon lateralis. Lateral ethmoid is cross-hatched; lacrimal is blackened; autopterotic stippled. 1981 PARENTI: CYPRINODONTIFORM FISHES C D 426 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 58. Diagrammatic representation of the posterior region of the skull and attachment of first vertebra, lateral view, of Cubanichthys {Chriopeoides) pengelleyi. Posttemporal removed. cilia, at least one or two pleural ribs are found on the first and second caudal vertebrae. This character, absent in Fluviphylax and certain poeciliids, is ambiguous. Sensory Pores and Cephalic Squamation: The general pattern of the head scales and sensory pore patterns of cyprinodontiform fishes is exhibited by the genus Jenynsia (fig. 14B). There are seven preorbital, three or four lacrimal, four mandibular, and six or seven supraorbital pores. In Jenynsia, there is a break between the anterior and posterior section of supraorbital pore 2, termed 2a and 2b. A series of three pores (2b, 3, and 4) follow the section formed by pores 1 and 2a. There is another break between sections of pore 4 referred to as pores 4a and 4b. The section 4b through 7 completes the supraorbital series. Gosline (1949) figured an identical pattern for Fundulus chrysotus and stated that this was the common pattern among Fundulus species. It was also observed that such a pattern is typical of cyprinodontoids such as Profundulus (also reported by Miller, 1955a), Oxyzygo- nectes, and many but not all goodeids (see Miller and Fitzsimons, 1971). In Anableps, the central row of pores (2b through 4a) is reduced to two pores which are referred to as pores 3 and 4a. Departure from the general squamation pattern also occurs in Anableps in which there are many scales arranged in a scattered pattern which cannot readily be interpreted using Hoedeman's terminology. Since the pattern of Jenynsia is postulated as the plesiomorphic sensory pore pattern, departures from this pattern are of interest in defining monophyletic groups of cyprinodontoids. However, patterns discussed here are only the most common ones found within a group of genera. A rigorous analysis of head pore and scale patterns requires a survey of inter- and intraspecific variation that is outside the scope of this study. Supraorbital pores of the poeciliids, procatopines, Fluviphylax and Pantanodon show an apparently unique modification. The maximum development of sensory pores of poeciliids (fig. 14D, E, F) was based on a 1981 PARENTI: CYPRINODONTIFORM FISHES soc-pro 427 epo-pro FIG. 59. Diagrammatic representation of the posterior region of the skull and attachment of first vertebra, lateral view, of Oxyzygonectes dowi. Posttemporal removed. FIG. 60. Diagrammatic representation of the posterior region of the skull and attachment of first vertebra, lateral view, of Profundulus punctatus. Posttemporal removed. 428 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 soc-pro FIG. 61. Diagrammatic representation of the posterior region of the skull and attachment of first vertebra, lateral view, of Tomeurus gracilis. Posttemporal removed. survey of such patterns in all major supraspecific categories of the family (Rosen and Mendelsohn, 1960). The supraorbital pores lie in groupings similar to the plesiomorphic pattern for cyprinodontiforms. The unique feature is the recessed neuromasts in the middle section (pores 2b through 4a) forming a small trough. {Poropanchax was defined on the basis of its embedded neuromasts which open as a series of pores.) The pattern is only weakly shown by the diminutive Fluviphylax and by Pantanodon. The connection of the canal between pores 4a and 4b (forming just one pore 4) occurs in Empetrichthys and Crenichthys, both of which retain the disrupted canal between pores 2a and 2b. The pattern Gosline termed the simplest among cyprinodontiforms (in the cyprinodontines Cyprinodon, Floridichthys, and Garmanella, and the funduline Lucania) in addition to connection of canals between 4a and 4b involves a connection between 2a and 2b (resulting in one pore 2). Thus, the canal is continuous between pores 1 through 7. In the New World cyprinodontine Jordanella and some goodeids (see Gosline, 1949), the canal is continuous except for a break between pores 4a and 4b. This pattern is considered to be independently derived in Jordanella and among a group of goodeids. Reduction of the pore system to pores 6 and 7 only, occurs in the fundulines Adinia and Leptolucania. A cephalic sensory pore system is absent in the monotypic New World cyprinodontine, Megupsilon. In Cubanichthys a canal is present between what appear to be pores 1 and 3 only, although Gosline stated they were present between pores 2 and 3 as well as 6 and 7. There are no pores posterior to what is identified here as pore 3 in Cubanichthys; however, since Gosline reports pores 6 and 7 present they must be considered part of the maximally developed pattern. Because of their position, I interpret the first two pores as 1 and 3, even though by definition the pore anterior to 3 should be 2b or 2. The ambiguity of the numbering system is evident in such a case. 1981 PARENTI: CYPRINODONTIFORM FISHES 429 FIG. 62. Diagrammatic representation of the posterior region of the skull and attachment of first vertebra, lateral view, of Cyprinodon variegatus. Posttemporal removed. Pore 3 of Cubanichthys is large, as it is in Chriopeoides, and is considered a synapomorphy of the two monotypic genera. Replacement of the two E scales by one large E scale also occurs within the New World cyprinodontines, Lucania, Cubanichthys, Chriopeoides, Empetrichthys, and Crenichthys. Reduction of the number of pores is apparently correlated at some level with the reduction in the number of head scales. Gosline reported that Aphanius dispar, an Anatolian cyprinodontine, has a canal between pores 2 and 4 and 6 and 7, and also noted the lack of mandibular canals. Specimens of A. dispar also possess pore 1, and three neuromasts apparently corresponding to pores 5 through 7. Another species, A. mento, lacks cephalic sensory pores and has a series of minute neuromasts arranged in a lyre-shaped pattern. Neuromasts ring the orbit and a line of minute neuromasts replaces the preopercular and mandibular canals. The entire system is strikingly like that of the genus Orestias (fig. 14) and of the aplocheiloid Cynolebias (fig. 13). This character within cyprinodontoids cannot always be distinguished from that in Cynolebias (except for the fact that the preorbital area is smaller in the aplocheiloids; yet this character is independent of the preorbital line of neuromasts). A line of preorbital neuromasts is not peculiar to these genera, as it is also found among the fundulines; therefore, the generality of the pattern cannot yet be determined. Consequently, it is assessed as a convergence between members of the genus Cynolebias and Orestias and a group of Anatolian cyprinodontines. INTERNAL FERTILIZATION AND VIVIPARITY: Previous workers (e.g., Rosen and Bai- ley, 1963; Miller, 1979) have assumed that viviparity defines a monophyletic group of cyprinodontiforms, and have therefore focused on describing the similarities and differences of adaptations for viviparity among the families in an effort to determine which viviparous family was more closely related to which other such family. In the present study, this presumption was discarded at the 430 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 64. Diagrammatic representation of first several rays of the anal fin of a male Cynolebias (Campellolebias) brucei. External view. FIG. 63. Diagrammatic representation of first several rays of the anal fin of a male Cynolebias (Cynopoecilus) melanotaenia. External view. outset. The simple division of cyprinodontiforms into oviparous and viviparous groups is an artificial one, and grossly oversimplifies the question of cyprinodontiform interrelationships. Internal Fertilization and Anal Fin Modification: Internal fertilization occurs in groups of atherinomorph fishes with and without an anal fin modified into a gonopodium. Developing embryos have been found in the body cavity of the ricefish Oryzias (Amemiya and Murayama, 1931), yet no modifications in the anal fin structure of this genus have been reported. Among aplocheiloids, a group of Neotropical genera are distinguished by the thickening of the anal rays of the females. All included species are annual, and this has been suggested as an aid to the depositing of eggs in the substrate during fertilization (Weitzman and Wourms, 1967). The anal fin of aplocheiloids is typically unmodified, except in two species placed here in the genus Cynolebias: melanotaenia (Regan) and brucei (Vaz-Ferreira and Sierra). In melanotaenia, the first six anal rays of the male are crowded together (fig. 63) and slightly offset from the rest of the fin. The rays are covered with small contact organs. In brucei (fig. 64) the first three anal rays are drawn together to form what is effectively a true gonopodium. Both cases are inferred to represent modifications for internal fertilization, yet both species are oviparous as well as annual. Females of brucei isolated after being in contact with males have laid fertilized eggs (Vaz-Ferreira and Sierra, 1974). Presumably, once the eggs are laid, they develop in a fashion typical of their annual relatives. One other case of internal fertilization occurs within aplocheiloids in Rivulus marmoratus. Populations of this species have been found consisting of self-fertilizing hermaphrodites and possess color patterns indistinguishable from females of the species (Harrington, 1961). This self-fertilization, of course, involves no modification of the anal fin. The fertilized eggs of marmoratus are laid as in C. brucei and melanotaenia, thus there are no known cases of embryo retention among the aplocheiloids. Among cyprinodontoids, internal fertilization has been demonstrated only among the viviparous families; however, its discovery in an oviparous cyprinodontoid would not be surprising, considering the generality of the condition. Structure of the gonopodium: Among the viviparous families there are three basic types of anal fin modifications of the male which effect internal fertilization. These are the gonopodia of poeciliids, the tubular gonopodia of Jenynsia and Anableps, and the muscular internal organ and slightly modified anal fin of the goodeids. The structure and development of the gonopodium of the poeciliid fishes has been discussed in detail (Rosen and Bailey, 1963; Rosen and Gordon, 1953; Rosen and Kail- 1981 PARENTI: CYPRINODONTIFORM FISHES 431 FIG. 65. Diagrammatic representation of gonopodium and gonopodial suspensorium of Poecilia vivipara. Anal radials are blackened. man, 1959). The gonopodia and gonopodial suspensoria vary among taxonomic groups of poeciliid fishes; it is primarily on these structures that such groups are defined. 432 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 66. Sketch of body form and fin position of Anableps microlepis, male above, female below. Dotted line approximates base of hypural plate. (After Rosen, 1973b.) The poeciliid gonopodium (fig. 65) is formed principally from the third, fourth, and fifth anal rays. Transformation from an undifferentiated anal fin begins with a thickening of the third anal ray. In all poeciliids the first three anal rays are unbranched. There is a rapid growth of rays three through five to form the so-called 3-4-5 complex. Further elaboration and growth occur, resulting in a gonopodium which is often adorned with various spicules, hooks, and spines. Internal supports are modified within poeciliids to a greater degree than in the other viviparous families. Again, as in Poecilia vivipara (fig. 65), the proximal anal radials two through five are elongated. Histolysis of the first hemal arch results in an ossified remnant termed the ligastyle, which migrates anteriorly. In addition, the second, third, and sometimes fourth hemal arches are expanded, the distal tips of which project anteriorly to meet the anteriorly projecting tips of the proximal anal radials. Anableps and Jenynsia have a tubular gonopodium formed from enlarged anal fin rays covered anteriorly with a fleshy sheath (fig. 66). In Anableps, the sheath is covered with scales; in Jenynsia it is bare. Internally, the structures are similar in that the anal rays are twisted around each other (figs. 67, 68). Similarly, there is an enlargement of the proximal anal radials, as well as an elongation of the hemal spines. Gonopodial development in Jenynsia and Anableps differs considerably; neither resembles that of the 3-4-5 complex typical of poeciliids, however. In Anableps, the gonopodium is formed from the 12 anal fin rays, counting each ray separately. The first ray is rudimentary, nonetheless it will be referred to as ray 1, contrary to the convention established by Turner (1950). The first four rays are unbranched. Rays three through six are enlarged and twisted around each other (fig. 67), whereas seven through nine are also enlarged but lie straight. Rays 10 through 12 are drawn forward in the formation of the tubular sheath, but otherwise undergo little differentiation. The proximal radials are also enlarged, drawn together and angled anteriorly. The first four or five proximal radials are offset 1981 PARENTI: CYPRINODONTIFORM FISHES 433 ^ FIG. 67. Diagrammatic representation of gonopodium and associated elements of Anableps. Anal radials are blackened. (After Turner, 1950.) FIG. 68. Diagrammatic representation of gonopodium and associated elements of Jenynsia lineata. Sixth middle anal radial is stippled; all other radials are blackened. 434 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 69. Diagrammatic representation of the anal fin of a female Jenynsia lineata. Sixth middle anal radial is stippled; all other radials are blackened. to either the left or the right of the midline in sinistral or dextral males, respectively; they extend to just beneath the vertebral column. Typically, there is histolysis of the last several pleural ribs. The hemal spines, especially the first three, extend ventrally and are situated between the proximal radials. At their bases these radials typically have bony flanges which project dorsally. Similarly, the bases of the anal fin rays are greatly enlarged and have similar flanges which overlap on adjacent rays. A full complement of middle and basal radials appears to be present, although Turner (1950) did not illustrate all of these for his specimen. In Jenynsia (fig. 68) the gonopodium is formed from 10 anal rays; however, its development involves primarily reductions of some elements found typically in the female anal fin. Therefore, the anal fin of the female will be described first so that a comparison with the structures within the gonopodium may be made easily. In an adult female Jenynsia (fig. 69) there are 10 anal rays, counting the last two separately. The first two rays are unbranched. The first six rays are crowded together; there is a corresponding crowding and reduction of the first five middle anal radials. The first two proximal radials are fused at their base. 1981 PARENTI: CYPRINODONTIFORM FISHES 435 FIG. 70. Diagrammatic representation of the internal structure of the abdominal cavity, anal fin rays and vertebral column of Characodon lateralis. There appears to be no separate proximal radial for the first anal ray; however, it is possible that the radial has become fused to the base of the recognized first radial which has a small bony knob projecting anteriorly. This interpretation is supported by the fact that there are three middle radials present corresponding to the large proximal radial. Five proximal radials all lie anterior to the elongate first hemal spine. In an adult male Jenynsia the sixth middle anal radial is the first unreduced radial as it is in the female (fig. 69), and as in the female, five rays precede and four rays follow this radial. Of the first six rays, all but rays 3 and 6 are extremely reduced. The seventh and eighth rays, as well as 3 and 6, are elongate and thickened; together with the relatively unelaborated segments of the ninth ray, they constitute the principal rays of the gonopo- 436 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 168 FIG. 71. Anterior rays and supports of the anal fin of a male A. Characodon lateralis; B. Crenichthys bailey/; C. Empetrichthys merriami. Cartilage is stippled. 1981 PARENTI: CYPRINODONTIFORM FISHES dium. Ray 6, the thickest and the longest, is hooked at its tip. The proximal radials, especially those of rays 2 through 6, are crowded together more so than in the female. These radials appear to be fused in part, although they have not been observed completely fused. All radials appear to be present, but identification of individual segments is difficult. In male and female Jenynsia, there are bony flanges on the base of the rays and the proximal radials. The proximal radials are offset to the midline in males, corresponding to the laterality of the individual as in Anableps; a dextral male is illustrated. A ligastyle has been found in Jenynsia males, as in poeciliid males. In addition, the gonopodium of Jenynsia is similar to that of Anableps and differs from that of the poeciliids in having the proximal radials enlarged and angled forward to the left or right; there are never enlarged hemal arches which project anteriorly to meet the proximal radials of the anal fin which migrates anteriorly in its development within the poeciliids. The structure of the jenynsiid gonopodium differs from that described by Turner (1950) who stated (p. 352): "most of the rays in the anterior part of the fin undergo absorption . . . ." It agrees more with that of de Gil (1949) who illustrated several variations in the formation of the fin, but in each case indicated that all the fin rays were present. The anal fin of the goodeid males is relatively unmodified compared with those of the poeciliid, jenynsiid, and anablepid fishes. The structure is not properly termed a gonopodium, as the modifications of the anal fin elements themselves appear to have little to do directly with the transfer of sperm. Goodeid males, however, are diagnosable on the basis of anal fin structure. The first six or seven fin rays are shortened and unbranched, and offset from the rest of the fin rays (fig. 70). The first anal fin ray is rudimentary, and the middle radials of the first six or seven rays are fused to the base of the proximal radials. Taken together with the presence of trophotaeniae in embryos, these characters were used by Miller and Fitzsimons (1971) to define the Goodeidae. 437 The rudimentary anal ray is formed to varying degrees among members of the family (Miller and Fitzsimons, 1971). The first four to seven middle anal radials are not present as distinct structures in all goodeids examined (e.g., as in Characodon lateralis, fig. 71A). However, among cyprinodontiform fishes, middle anal radials are fused to the proximal radials in the two North American genera suggested as close relatives of the goodeids, Crenichthys and Empetrichthys. In Crenichthys baileyi (fig. 76), the first five middle radials are lacking. In Empetrichthys merriami (fig. 71C) the first proximal radial is fused and there is no first or second middle radial; the third middle radial is represented by a minute ossification at the base of the third proximal radial. The proximal radials corresponding to the shortened anal fin rays are greatly elongate in goodeids (fig. 70). They are not fused together as in the other viviparous families, however. The proximal radials of Empetrichthys and Crenichthys are slightly elongate; however, they differ little from that of a typical oviparous cyprinodont (fig. 22). In the ontogeny of anal fin rays, all are formed unbranched and then successively become branched. In both Oryzias and Menidia, the number of unbranched anal rays is two, as it is in many cyprinodontoids. Among the aplocheiloid fishes which have lost the first proximal anal radial, there are often three unbranched anal rays. The number of unbranched rays varies among cyprinodontiforms from no rays unbranched in an occasional specimen of Profundulus (Miller, 1955a) and in members of the genus Orestias to all but one unbranched in some fundulines and cyprinodontines. In all poeciliids, as well as the genus Pantanodon and at least one nominal species of Aplocheilichthys, A. johnstoni, there are three unbranched anal rays. In Anableps there are four in males and three in females, whereas there are two in both males and females of Jenynsia. Among the goodeids, the unbranched anal fin rays typically number more than four. Considering the ontogeny of anal fin rays, it could be argued that a high number of unbranched rays is primitive, while successive- BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY 438 VOL. 168 HYP HYP INCLA INCL CARP 1CARM ICARP ICARM FIG. 72. Diagrammatic representation of the generalized primitive state of the anal fin musculature in cyprinodontiforms. FIG. 73. Diagrammatic representation of the derived anal fin musculature of cyprinodontiforms. ly lower numbers are derived. Conversely, a certain number of rays could be primitive for a group, and the suppression of branching a derived modification. I accept the latter viewpoint since it is the description of characters in their adult form whose distribution must be analyzed without recourse to presumptions about varying ontogenies. Thus, the increase in unbranched anal fin rays above one or two is hypothesized to be a derived character among cyprinodontiforms. Anal Fin Musculature: The anal fin musculature of the aplocheiloids which exhibit internal fertilization, C. melanotaenia and C. brucei, is of the primitive type for cyprinodontiforms (fig. 72). That is, there is a set of external inclinators as well as erectors and depressors corresponding to the individual anal fin rays. One broad band of muscle, the infracarnalis medius runs from the base of the pelvic fins to the first anal radial; a second, the infracarnalis posterior runs from the last anal radial to the distal tip of the last hemal spine (Winterbottom, 1974). In male poeciliids, the anal inclinators are drawn together into a fan-shaped mass of muscle. In Anableps and Jenynsia, there are no such fan-shaped masses of muscles. The inclinators are thickened, otherwise the muscles differ little from those of the generalized type. Nelson (1975) discussed the mechanism of sperm transfer in the goodeids, and illustrated the anal fin muscles. Such muscles in goodeids diverge most from the generalized type in forming a large muscular mass surrounding the vas deferens and urinary tract in a structure which was termed a pseudophallus by Mohsen (1961a). Along with this urogenital organ, is an elaboration of the inclinators of the anal fin. The inclinators, which arise between the hypaxial musculature and insert on the bases of the anal fin rays distal to the insertion of the erectors and depressors (Winterbottom, 1974), arise just below the division of the epaxial and hypaxial musculature (fig. 73). Such elaborate inclinators, however, are not restricted to the goodeids. They are found also in a group defined above on the basis of skull and jaw specializations, viz., Empetrichthys, Crenichthys, Cubanichthys, Chriopeoides, Orestias, and the cyprinodontines. SPERM TRANSFER: Sperm transfer occurs in poeciliids when the gonopodium is swung forward and folds over to one side or the other to form a groove. Sperm pass down the groove in unencapsulated bundles termed spermatozeugmata and are transferred to the female by application of the gonopodial tip to, or within, the genital pore. In some poeciliids the pelvic fins are also 1981 PARENTI: CYPRINODONTIFORM FISHES modified in the males. Clark and Kamrin (1951) report that in a number of poeciliids tested, the pelvic fin of one side is swung forward together with the gonopodium. They speculated that the pelvics in such species contribute to the formation of the sperm groove. In the poeciliids, procatopines, Pantanodon and Fluviphylax, pelvic fins are set far forward and are often under the pectoral fin bases (e.g., in the poeciliid Heterandria bimaculata, fig. 52 and in the procatopine Procatopus glaucicaudus, fig. 53). The thoracic or subthoracic position of the pelvic fins in these groups results from an ontogenetic forward migration during sexual differentiation. This fact has been well known within the poeciliids for some time (Clark and Kamrin, 1951; Rosen and Kallman, 1959); however, the phenomenon in the procatopines and other genera is little known. Trewavas (1974) reported data for several species of Procatopus from West Cameroon to support the fact that the pelvic fins indeed migrate forward in ontogeny. The extent of this phenomenon among procatopines is not known and could properly be explored with laboratory developmental series of representatives of all procatopine genera as well as Fluviphylax and Pantanodon. Within Anableps and Jenynsia the tubular gonopodium is associated with a distinctive mode of sperm transfer. In both genera, all the rays of the anal fin are brought close together and surrounded by a fleshy tube (fig. 66). The sperm duct enters the tube at the base of the first anal ray and follows the ventral edge of the tube to its tip. Sperm do not travel down the tube in sperm bundles, but individually. Grier, Burns and Flores (MS) report that partial sperm bundles are formed in Anableps dowi, the presumed primitive species of the genus (Miller, 1979) but break down before they enter the sperm duct; only free spermatozoa were observed in both the efferent and main testes ducts of A. anableps and Jenynsia lineata. The abdominal pelvics of Jenynsia and Anableps are inferred to have no function in sperm transfer. Since Carman (1895) described the presence of sexual lefts and rights in Anableps 439 and Jenynsia the phenomenon of dextral males pairing with sinistral females (and vice versa) has been reported tentatively in the literature (e.g., Rosen and Bailey, 1963) although its occurrence is still doubtful (Miller, 1979). In theory, there are two kinds of males, sinistral and dextral and corresponding types of females. A dextral male supposedly has a gonopodium which is offset to the right; therefore, he can only copulate with a sinistral female. The sidedness of a female is determined by the placement of one or two scales over one side of the urogenital opening; hence a scale covering the left side of the opening defines a female as dextral. Individuals, although they may easily have their laterality determined, have not been observed to be either exclusively dextral or sinistral in their mating (Miller, 1979). Thus, the significance of the anatomical modifications related to sidedness remain speculative. Although laterality is not evident among young males of Anableps examined, adult males could easily be classified as left or right. Females' sidedness is also generally easy to determine; however, some large adult female Anableps seem to be neither left nor right. In the oviparous Oxyzygonectes, a hypothesized close relative of Jenynsia and Anableps, males have a distinct anal papilla which in preserved specimens has been observed to be offset to the left or to the right. The significance of this character is equivocal since it may simply be an artifact of preservation. Females offer no clue since they possess a fleshy pouch surrounding the genital opening. Both male and female Oxyzygonectes, however, have scales around the anterior region of the anal fin much like the pocket of scales surrounding the anus and first several anal fin ray bases considered to be a diagnostic character of the procatopines (Clausen, 1967). This pocket of scales apparently is a derived character of the larger group including the poeciliids, Jenynsia, Anableps, the procatopines, Fluviphylax and Pantanodon as well as Oxyzygonectes. If so, these scales have been modified or reduced in the viviparous forms. 440 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY Sperm bundles are formed in the goodeids (Grier, Fitzsimons and Linton, 1978), although the precise mechanism of sperm transfer is still unknown. Mohsen (1961a, 1961b) described a muscular organ surrounding the vas deferens and urinary canal and believed that sperm bundles were ejected during copulation by a contraction of the organ. Nelson (1975) reported that during a copulation attempt, the male clasps the female: the anterior portion of the anal fin formed by the shortened rays is wrapped around the genital opening of the female while the notch in the fin is placed near the anterior margin of the anal fin of the female. Thus, the urogenital organ of the male comes very close to the female's urogenital opening. True spermatophores, that is, encapsulated sperm bundles, occur only in the adrianichthyoid Horaichthys (Kulkarni, 1940). Sperm bundles, however, occur among several teleost groups including the cyprinodontiforms just mentioned and the exocoetoid Dermogenys whose sperm bundles are indistinguishable from those of the poeciliids (Grier, Burns and Flores, MS). Free spermatogonia, like oviparity, must be considered a primitive character. However, the occurrence of spermatozeugmata with viviparity among different relatively unrelated groups suggests that spermatozeugmata have arisen independently along with viviparity. Fertilization and Development: The distribution of characteristics related to egg retention and maternal contribution to development precludes the ready division of members of the four so-called viviparous families into oviparous, viviparous and ovoviviparous groups. Among the poeciliids, one genus and species, Tomeurus gracilis, is oviparous and just facultatively viviparous (Rosen and Bailey, 1963). Fertilization takes place as in all other poeciliids, within the follicle; however, the developing embryo is quickly released from the ovary into the oviduct and then passed to the outside for the remainder of the developmental period. The egg of Tomeurus has a thick chorion with adhesive filaments like that in oviparous cyprinodonti- VOL. 168 forms. Among other poeciliids, the egg retains an extremely reduced chorion (Zahnd and Porte, 1962; Flegler, 1977). Also among the poeciliids are found some of the smallest vertebrate eggs (Scrimshaw, 1946). Ovoviviparity may be identified in certain groups of poeciliids, for example Brachyrhaphis episcopi (Turner, 1938). The yolk sac is relatively large, and although development is internal, nutritional support is derived primarily from the yolk. Development of fertilized eggs in the rest of the true viviparous poeciliids is of two major types (Turner, 1939). In Heterandria formosa, for example, a so-called pseudochorion and pseudoamnion are formed by the folding around the embryo of extraembryonic somatopleure in the pericardia! region. The outer membrane thus formed is termed the pseudochorion; it is highly vascularized, whereas the inner pseudoamnion is nonvascular. The pseudochorion is used as an organ for respiration and nutrition throughout development. In species of Poeciliopsis, which also have small yolk reserves, the lateral region of the somatopleure is poorly developed. The ventral region becomes highly vascularized and invades the region between villi formed in the follicular membrane; together they form what is termed a follicular pseudoplacenta. Thus, the embryos of Heterandria formosa and Poeciliopsis and related species derive the greater part of their nutrition from maternal tissues rather than from stored yolk. Poeciliids also exhibit the phenomenon of superfetation, that is, the overlapping of developing broods in the ovary (Scrimshaw, 1944). Anablepid, jenynsiid, and goodeid fishes exhibit more elaborate adaptations for viviparity. In Anableps, both fertilization and development are intrafollicular as in poeciliids. In addition, a follicular-pseudoplacenta more elaborate than that of the poeciliids is also developed. The ventral portion of the somatopleure expands to form highly vascularized projections; the surrounding follicle is covered with vascular villi. In Anableps dowi, the species presumed to be most prim- 1981 PARENTI: CYPRINODONTIFORM FISHES itive, the large intestine expands and nearly fills this sac. Follicular fluid is absorbed by the embryo across intestinal villi. At birth, the follicle ruptures and the expanded belly sac eventually undergoes shortening. In jenynsiids, although fertilization is intrafollicular the embryo is evacuated from the follicle and development takes place within the ovary. Development is viviparous rather than ovoviviparous since the yolk supply is consumed at an early stage, and nutrition is derived mainly from maternal fluids. Respiration occurs across an expanded ventral somatopleural sac as in the anablepids. Maternal fluids enter the developing embryo through its mouth or through the opercular openings (Turner, 1940b). Flaps grow out from the wall of the ovary and invade the opercular region and an intimate connection between embryo and mother is provided as the flaps of tissue invade the pharyngeal and buccal cavities. In goodeids, as in jenynsiids, the eggs are fertilized in the follicles and then released into the ovarian cavity for development. Goodeids are characterized by the possession of trophotaeniae (Turner, 1937), elaborate outgrowths in the perianal region, the epithelium of which possesses villi and is indistinguishable from intestinal epithelium (Wourms and Cohen, 1975). Their function as absorbers of nutritive ovarian fluids is inferred from their structure. Morphology of the trophotaeniae and of the ovary served as the principal character for the last general revision of the Goodeidae by Hubbs and Turner. However, more recently, Miller and Fitzsimons (1971) have reviewed the classification and concluded that the great degree of variability among these structures makes them of little importance in phylogenetic studies. Miller and Fitzsimons did not propose a reclassification. Goodeid ovaries (median organs formed by fusion of right and left anlagen) fall into one of two main types (Hubbs and Turner, 1939): (1) an ovarian septum and outer wall composed of ovigerous tissues, the inner septum is often folded down the middle of the joint ovarian cavity; and (2) an ovarian septum and outer wall which is devoid of 441 ovigerous tissue, the structure of which is as two folded masses, one in each section of the ovary. The first of these is apparently the primitive state of the fused ovaries, and the second the more derived state with ovigerous tissue excluded from the walls and septum and confined to the middle of the ovarian cavities. In one genus and species, Characodon lateralis, there is an intermediate type of ovary which has ovigerous tissue both in a short section of the septum and in weakly formed tissue extensions into each of the ovarian cavities. This so-called intermediate condition, however, may be more accurately assessed as more closely related to the derived type 2; that is, it forms a transition between the distinctly primitive and derived types. Trophotaeniae occur in three types: rosette or ribbon-like, and when ribbon-like, sheathed or unsheathed. In a sheathed process the external epithelium is separated from the internal connective ti