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

Novitatesx ^AMERICAN MUSEUM PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2779, pp. 1-12, figs. 1-7, table 1 February 8, 1984

On the Relationships of Phallostethid Fishes (Atherinomorpha), With Notes on the Anatomy of Phallostethus dunckeri Regan, 1913

LYNNE R. PARENT!'

ABSTRACT The Phallostethidae (including Neostethidae) is western Pacific Dentatherina Patten and Ivantsoff a family comprised of approximately 20 species is proposed as the sister group of the Phallosteth- of small, fresh, brackish, and occasionally salt- idae. water atherinomorph fishes of Indo-Australia. The anatomy of Phallostethus Regan, the type Phallostethids have variously been suggested as genus, is poorly known because of the scarcity and closest relatives of the atherinoid or cyprinodon- unsatisfactory condition of available material. A tiform fishes among the atherinomorphs, or of the report on the anatomy of Phallostethus dunckeri polynemids or gobioids among the percomorphs. Regan, the sole species in the genus, based on ex- Phallostethids uniquely share several derived amination of the syntypes and on unpublished characters of the jaws and the anal fin with a group notes and sketches is also included. of Indo-Australian and Pacific atherinoids. The

INTRODUCTION The phallostethids (Atherinomorpha, pium (Regan, 1913, 1916), and among oth- Phallostethidae) are a little-known group of ers, a series of derived characters related to coastal fishes distributed throughout the Phil- reproduction as, for example, the anterior ippines, Borneo, Java, Malay Peninsula, and placement of the urogenital opening and re- Southeast Asian mainland.2 They are defined duction and/or modification of the pelvic fins as monophyletic by a complex subcephalic and fin girdles (see Roberts, 1971b). copulatory organ in males, termed the pria- The primary purpose of this report is to

1 Research Associate, Department of Ichthyology, American Museum of Natural History. 2 Report of a collection of phallostethids from northwestern Sumatra (Aurich, 1937) is considered to be unconfirmed.

Copyright © American Museum of Natural History 1984 ISSN 0003-0082 / Price $1.25 AMERICAN MUSEUM NOVITATES NO. 2779 present the hypothesis of close relationship coetoids (flying fishes, halfbeaks, sauries, and among the phallostethids and several atheri- needlefishes) into the newly named series noid genera of Indo-Australia and the Pacific. Atherinomorpha, which he considered to be The definition and relationships of phallo- most closely related to the series Percomor- stethid species and genera are the subjects of pha. Rosen and Parenti (1981) formally de- an ongoing study (Parenti, in prep.). How- fined the atherinomorph fishes, giving as sev- ever, such a study could not be carried out eral of their derived characters specializations without a detailed, well-corroborated hy- of the egg, embryo, sperm formation, rostral pothesis of the relationship of phallostethids cartilage and association of the premaxillary to other fishes. ascending processes, and dorsal gill arches. When phallostethids were first discovered, Phallostethids remained in the Atherino- it was assumed that they were viviparous cy- morpha based on their possession of several prinodontiforms (then Microcyprini) be- of these derived characters. Rosen and Par- cause the priapium superficially resembles enti (1981) concluded that the atherinoid intromittent organs found among some fishes are not currently definable as mono- members of that group, for example, poecili- phyletic and simply listed the families of ath- ids (Regan, 1913, 1916). Smith (1927) re- erinoids in their Division I of the Atherino- ported that phallostethids which he observed morpha, abandoning the use of the term in Thailand were oviparous. The priapium is atherinoid in formal classification to empha- apparently used by males for passing sperm size uncertainty in our knowledge of rela- bundles to females who subsequently lay fer- tionships of these fishes (table 1). tilized eggs, although details of priapial func- The phallostethids have undergone eleva- tion are unknown. tions and reductions in taxonomic rank since Both Herre (1925) and Myers (1928) point- their discovery; however, no formal state- ed out that some phallostethids have a spi- ment concerning their relationship to another nous first dorsal fin which is lacking in the group of fishes has been made previously. cyprinodontiforms. On this basis, and be- Myers (1935) created a new suborder within cause of the overall resemblance of phallo- the Percesoces, what he termed Phallosteth- stethids to atherinid fishes (the silversides or oidea, which served to emphasize further the hardyheads), Myers transferred the phallo- unique characters of these fishes. Rosen (1964) stethids to the order Percesoces, which then placed the phallostethids in the superfamily contained the silversides, mullets, barracu- Phallostethoidea, suborder Atherinoidei, but das, and threadfins. Bailey (1936) proposed suggested that they may be more closely re- that phallostethids were close relatives of the lated to the cyprinodontiforms. Roberts threadfins (the polynemids) based on a su- (1971b) speculated that phallostethids are perficially similar association of the pectoral closely related to the atherinid subfamily and pelvic fins. Taeniomembradinae of Schultz (1948, 1950). Hubbs (1944) concurred with Myers that However, the Taeniomembradinae was not the phallostethids were more closely related defined by Schultz as monophyletic. He stat- to the atherinids than to the cyprinodonti- ed only that these atherinids possess the forms; however, whereas Myers considered primitive state of the swimbladder character the cyprinodontiforms and the atherinids to used to define a relatively more derived be closely related, Hubbs considered the subfamily. Even the idea that phallostethids percesocans (including the phallostethids and are atherinomorph fishes has been ques- atherinids) to have a relatively more derived tioned recently by V. G. Springer (personal fin structure than that of the cyprinodonti- commun.) who suggests that they may be forms. Thus, Hubbs supported an alignment closely related to the gobioid fishes. of the percesocans closer to some of the per- This report was prompted by: (1) the recent ciform fishes. discovery and diagnosis of a new genus and Rosen (1964) placed the phallostethids, species of atherinid, Dentatherina merceri by atherinids, cyprinodontiforms, along with the Patten and Ivantsoff (1983), which I propose adrianichthyoids (the ricefishes), and the exo- as the closest living relative of the phallo- 1984 PARENTI: PHALLOSTETHID FISHES stethid fishes; (2) the opportunity to examine TABLE 1 Regan's syntypes of Phallostethus dunckeri Classification of Atherinomorph Fishes at the British Museum (Natural History); and, (From Rosen and Parenti, 1981) (3) most important, by the gift of notes and sketches prepared by Dr. Ethelwynn Trewa- Series Atherinomorpha vas in the 1930s as part of her planned re- Division I vision of phallostethid fishes. Both because Family Atherinidae Family Bedotiidae of the scarcity of material and present poor Family Isonidae condition, which precludes a formal rede- Family Melanotaeniidae scription, Dr. Trewavas's notes and sketches Family Phallostethidae of P. dunckeri contain some of the only data Family Telmatherinidae available on the anatomy of this unique Division II species. Order Cyprinodontiformes Order Beloniformes ACKNOWLEDGMENTS Dr. Donn E. Rosen, American Museum of Drs. Ivantsoff, John Patten, and Rosen read Natural History, New York, has frequently and commented on the manuscript. discussed aspects of atherinomorph anatomy and systematics with me. Dr. Tyson R. Rob- Phallostethus dunckeri Regan erts, California Academy of Sciences, San Francisco has, on numerous occasions, freely Phallostethid anatomy has been the focus given information on the anatomy and dis- of numerous studies published since Regan's tribution of phallostethids. Drs. Walter (1913) description of Phallostethus dunckeri Ivantsoff, Macquarie University, Australia, (e.g., Regan, 1916; Myers, 1928; Bailey, 1936; and Brian White, Los Angeles County Mu- Aurich, 1937; TeWinkle, 1939; Herre, 1942; seum of Natural History, Los Angeles, kindly Woltereck, 1942a, 1942b; Hubbs, 1944). shared information on atherinoids. The re- Roberts (1971a, 1971b) presented the most port on the anatomy of Phallostethus dunck- recent reviews of some problems in phallo- eri Regan would not have been possible with- stethid anatomy and systematics. He stated out the gift of notes and sketches made by that basic comparative data of Phallostethus Dr. Ethelwynn Trewavas, Curator Emerita, are unknown because Regan (1913) did not British Museum (Natural History) as well as report states of characters which we now con- her kind permission to use them here. sider to be useful in defining phallostethid I thank Drs. P. Humphry Greenwood, Brit- relationships, and because P. dunckeri is ish Museum (Natural History), London known today only from syntypes (Roberts, (BMNH), Gareth Nelson, American Mu- 1971b). For example, Roberts stated that the seum of Natural History, New York (AMNH), number of branchiostegal rays, and whether and Richard P. Vari and Ms. Susan L. Jewett, or not P. dunckeri has a first dorsal fin, were National Museum of Natural History, Smith- unknown. sonian Institution, Washington, DC. Regan (1913) based his description of (USNM), for lending or allowing me to ex- Phallostethus dunckeri on seven specimens amine specimens in their care. Dr. Clyde Bar- collected from Johore, on the Malay Penin- bour, Wright State University, Dayton, and sula. The description centered on the struc- Dr. Roberts, kindly donated specimens to the ture of the priapium and included other char- American Museum of Natural History which acters that are still considered to distinguish were used in this study. This project was ini- P. dunckeri from all other phallostethid tiated during the tenure of a NATO Post- species. One such character is a high number doctoral Fellowship at the British Museum of anal fin rays, ranging from 26 to 28 as and completed while I was a Research As- opposed to 14 to 15 in Phenacostethus, the sociate at the American Museum of Natural presumed closest relative (Roberts, 1971a). History. Regan (1913, pp. 548-555) stated that his AMERICAN MUSEUM NOVITATES NO. 2779

epioccipital supraoccipital frontal

pterotic prootic omandibula

opercle dentary-

subopercle anguloarticular and interopercle retroart icular preopercle quadrate symplectic FIG. 1. Lateral view of neurocranium, jaws and jaw suspensorium, and opercular series, Phallostethtis dunckeri Regan, Syntype, BMNH 1913.5.24:22. From a sketch prepared by Dr. Trewavas. Bone stippled; cartilage open circles.

description was based on seven specimens, Basisphenoid absent. Exoccipital and basioc- two of which were sectioned for study of in- cipital (not shown in fig. 1) with condyles well ternal anatomy. Five specimens, four alcohol developed. Infraorbital series represented by (BMNH 1913.5.24:18-20, BMNH 1913.5.24: a dermosphenotic and a preorbital bone. 21) and one cleared and stained (BMNH Mesethmoid an ossified triangular plate, the 1913.5.24:22), are present in the Recent fish base anterior, with a notch for passage of ol- collection of the British Museum (Natural factory nerve. Vomer, with small toothplate History) and hence, are treated as the re- and patch of teeth, ventral to ethmoid car- maining syntypes. The syntypes, augmented tilage. by notes and sketches of Trewavas, constitute JAWS AND JAW SUSPENSORIUM (figs. 1, 2): the study material on which the following Upper jaw represented by premaxilla with anatomical description is based. ascending process long and narrow; maxilla NEUROCRANIUM (figs. 1, 2): Supraoccipital with a process dorsal and a process ventral overlapped anteriorly by frontals. Frontals to premaxilla. Lower jaw represented by den- forming convex roof of orbits, a lateral limb tary, paradentary, articular and retroarticular posterior to nasal capsule, articulating with bones, the last two bones not necessarily dis- preorbital, and anterior limb reaching an- tinct. Premaxilla, dentary, and paradentary terolateral surface of lateral ethmoid. Pari- with small, unicuspid teeth. Submaxillary etals absent. Temporal region concave. Os- bone between maxilla and vomer. Rostral sified epioccipital and pterotic present. cartilage long and of moderate width. Hyo- Intercalar absent. Pterosphenoid small, not mandibula with a large, single dorsal head meeting sphenotic and just meeting frontal. articulating with sphenotic anteriorly and 1984 PARENTI: PHALLOSTETHID FISHES

premaxilla

_maxilla 1 basibrabasibranchial 1

-vomer basibranchial 2 submaxillary bone -rostral cartilage

ceratobranchial 1 asibranchial 3 ^-lateral ethrnoid haryngobranchial 2

.epibranchial 1 |:'f\. '—-——frontal ceratobranchial 2, pibranchial 2 ceratobranchial 3. ^ethmoid cartilage pibranchial 3 pibranchial 4 FIG. 2. Dorsal view of upper jaw, ethmoid region ceratobranchial 4 and anterior portion of neurocranium, Phalloste- ceratobranchial 5 pharyngobranchial 3 thus dunckeri Regan, Syntype, BMNH 1913.5.24: 22. From a sketch prepared by Dr. Trewavas. Bone FIG. 3. Dorsal view of gill arches, with left dorsal stippled; cartilage open circles. portion removed, Phallostethus dunckeri Regan, Syntype, BMNH 1913.5.24:22. From a sketch prepared by Dr. Trewavas. See text for discussion on pterotic posteriorly. Symplectic a long, nar- identification of structures. row bone. Metapterygoid present as a small bone at hyomandibular-symplectic junction. anterior (pharyngobranchial 2) articulating Quadrate with slender posterior ramus. En- with the second epibranchial, the posterior dopterygoid narrow and elongate; ectopter- (pharyngobrancial 3) with the third and fourth ygoid minute. Autopalatine reaching maxilla epibranchials. anteriorly. VERTEBRAL COLUMN: Forty vertebrae, 13 OPERCULAR SERIES (fig. 1): Opercle oval abdominal, 27 caudal. First pleural rib on with posteroventral indentation, lacking ser- fourth vertebra. No epineurals or epipleurals. rations. Preopercle, subopercle and inter- CAUDAL SKELETON (fig. 4): Last vertebra opercle narrow. consisting of a half-centrum (PU2) with which HYOBRANCHIAL APPARATUS (fig. 3): Hyoid are united a dorsal and a ventral hypural plate. bar represented by a single hypohyal, and an- Parhypural autogenous. Two epurals. terior and posterior ceratohyal. Four branchiostegal rays. Interhyal ossified. Ossified portion of basihyal narrow and elongate. procurrent rays Three ossified basibranchials.3 Hypobran- chials, if present, not ossified. First ceratobranchial with 12 slender gill rakers, second ceratobranchial without gill rakers, the third and fourth ceratobranchials each with a toothplate and patch of teeth, fifth ceratobranchial expanded medially and with a parhypural toothplate bearing a patch of curved, pointed teeth. Four epibranchial bones, the first with gill rakers. Two upper pharyngeal bones, the

3 In her notes, Dr. Trewavas indicated that Phallo- stethus dunckeri has three ossified basibranchials. In her procurrent rays original sketch, there are just two ossified basibranchials FIG. 4. Lateral view of caudal skeleton, Phal- posterior to an elongate ossified basihyal, however. I lostethus dunckeri Regan, Syntype, BMNH interpret the elongate basihyal as a basihyal and the first 1913.5.24:22. From a sketch prepared by Dr. Tre- basibranchial. wavas. Epurals are blackened. AMERICAN MUSEUM NOVITATES NO. 2779

papillary bones

toxactinium

"ctenactinium

FIG. 5. Schematic diagram of internal priapial structure of Phenacostethus smithi Myers, BMNH 1927.12.29:1-10. From a sketch prepared by Dr. Trewavas.

FINS: NO spinous first dorsal fin; second a toxactinium, rather than one or two cten- soft-rayed dorsal fin with eight to 10 rays. actinia and a pulvinulus. The currently ac- Caudal fin truncate, seven procurrent rays cepted homologies of priapial parts requires dorsally, 16 branched caudal rays, and 11 that the toxactinium of Phallostethus and procurrent caudal rays ventrally. Minute pel- Phenacostethus be viewed as a modified pul- vic fins and fin girdles in females; males with vinulus (Roberts, 1971b). The ctenactinium pelvic fins and fin girdles modified into pria- of Phenacostethus (fig. 5) is rudimentary and pium (see below). not as well developed as that of Phallostethus PRIAPIUM: The present report does not re- in which it is serrated (see also Regan, 1913; quire a comprehensive review of priapial Bailey, 1936; Roberts, 1971a). anatomy, which will be considered in detail, in a revision of the Phallostethidae sensu lato RELATIONSHIPS OF THE (Parenti, in prep.). However, certain details PHALLOSTETHIDS are pertinent to this discussion. Characters in addition to the priapium and The priapium primitively has two con- associated reproductive traits have been pro- spicuous externalized bones, the ctenactin- posed as unique to phallostethids: a submax- ium and pulvinulus, both of which are de- illary bone (fig. 2), and a paradentary bone rived from pelvic fin structures (Regan, 1913, (fig. 1) (Roberts, 1971b). Rosen and Parent! 1916). Phallostethus Regan and Phenacos- (1981) treated the two families (Phallosteth- tethus Regan are distinguished from other idae and Neostethidae) of phallostethoids phallostethids by the type of externalized sensu Myers as one, the Phallostethidae. The priapial bones; they have a ctenactinium and division of phallostethids into two families 1984 PARENTI: PHALLOSTETHID FISHES was based on differences in priapial mor- niomembradinae, I have determined that phology, and by the identification of sub- most, but not all, members of these groups maxillary and paradentary bones in genera have submaxillary cartilages or bones. For that had been assigned to the Neostethidae example, the Western Australian isonid, Iso (Roberts, 1971b), such as Ceratostethus Myers rhothophilus (AMNH 55027SW), as well as and Neostethus Regan. Roberts (1971b) called the New World taeniomembradine Atherino- these bones "neomorphs." morus stipes (AMNH 52025SW), have no The recent discovery and diagnosis by Pat- such cartilages or bones. The maxilla and ten and Ivantsoff (1983) of Dentatherina, a vomer are separated by a small connective western Pacific marine atherinid, allows for tissue meniscus, which is a primitive char- a reinterpretation of these as well as other acter for acanthopterygian fishes. characters used to hypothesize the relation- Furthermore, other atherinoids, such as ships of phallostethids and other Indo-Aus- Quirichthys stramineus (AMNH 20571SW), tralian and Pacific atherinoids. Telmatherina ladigesi (AMNH 35378SW), SUBMAXILLARY BONES:4 Submaxillary and Pseudomugil tenellus (AMNH 36598SW) bones are prominent endochondral bones that have prominent submaxillary cartilages. Each lie between the medial ramus of the maxilla of these genera was placed in a different fam- and the anterolateral portion of the vomer ily of atherinoid fishes by Allen (1980). Ac- (figs. 2, 6). Bony elements in this position are cessory cartilages and bones in the ethmoid found in Dentatherina (Patten and Ivantsoff, region are not uncommon among euteleosts, 1983, figs. 4, 5) and phallostethids in both nor are elements in the position of the sub- the Phallostethidae and Neostethidae sensu maxillary bones. I judge these to be derived stricto, contra Roberts (1971b), who stated at one level within the atherinoid fishes, dis- they occur only in the latter. He did not know cussed below. the condition in Phallostethus. In some phal- PARADENTARY BONES: A separate, slender lostethids (e.g., Gulaphallus mirabilis, BMNH bone, termed the paradentary, lies lateral to 1933.3.11:179-186) and taeniomembradine the dentary in both Dentatherina and the atherinids (e.g., Craterocephalus cuneiceps, phallostethids. Patten and Ivantsoff (1983) AMNH 43184SW, fig. 7C) a submaxillary noted what they called "calcified nodules" in element is present as a large cartilage rather the lower jaw ligaments, preferring not to than a bone. Roberts (1971b) postulated that consider them homologues of the paraden- the submaxillary bones of neostethids sensu tary bones of phallostethids. However, for stricto contributed toward the extremely pro- two reasons I take the view that the structures tractile mouths of these fishes. are homologous. It is not my purpose here to present a hy- First, the suggestion that they are fortui- pothesis of the relationships of all atherinid tous ossifications of jaw ligaments is con- fishes, or even atherinid fishes of the sub- tradicted by the presence of a paradentary family Taeniomembradinae (comprising the bone with a single row of small, unicuspid genera Taeniomembras, Craterocephalus, teeth in Phallostethus dunckeri (fig. 1). As far Stenatherina, Alepidomus, Hypoatherina, as known, these bones are unique to Denta- Atherinomorus, and tentatively Quirichthys). therina and phallostethids. In some closely Therefore, I have not surveyed all the genera related taeniomembradine atherinids, such of any such group to determine whether or as Craterocephalus cuneiceps (AMNH not each has a submaxillary bone or cartilage. 43184SW), there is a mass of connective tis- However, from the limited survey of Indo- sue in the position of the paradentary. Pacific atherinoids and the subfamily Tae- Second, the homology of paradentary bones in Dentatherina and phallostethids is supported by parsimony. Reasons for their in- 4 This element should not be confused with one be- clusion in yet a larger group of atherinoids, tween the maxilla and autopalatine, called a subauto- is discussed below. The same ontogenetic se- palatine cartilage by Parent! (1981, p. 406, fig. 35A). quences of paradentary formation in both Such an element, found in some cyprinodontiforms, is groups would support this homology, how- apparently primitive for acanthopterygian fishes. ever, no such data are available. Histological AMERICAN MUSEUM NOVITATES NO. 2779

therina) and Pseudomugil, the first proximal anal radial is expanded anteriorly. In phallostethids (fig. 6A) and Dentatherina (fig. 6B) the first proximal anal radial is a long, blade- pleura! rib of like element. The main shaft of the radial is vertebra 13 oriented dorsally. That is, the radial is expanded anteriorly, rather than being recum- bent. first proximal anal radia anal fin rays In the other taxa listed above, the first proximal anal radial is expanded anteriorly hemal spine of and may be bladelike, as in Atherinomorus vertebra 22 stipes (AMNH 52025SW), or may be expanded just slightly, as in Pseudomugil tenellus (fig. 6C). In each case, the main shaft of the radial is evident. In other atherinoids (e.g., Bedotia geayi, AMNH 28132SW) the first proximal anal radial has no anterior ex- pansion; it is represented solely by the main, dorsally directed shaft. pleural rib of vertebra 11 ~ ADDITIONAL CHARACTERS: TWO other derived characters, the absence of parietal bones from the skull, and the absence of a dorsal or first postcleithral bone from the pectoral first proximal anal radia skeleton support a close relationship between Dentatherina and phallostethids. FIG. 6. Internal view of anal fin, anterior to left, of A. Gulaphallus mirabilis, BMNH RELATIONSHIPS OF 1933.3.11:179-186; B. Dentatherina merceri, ATHERINOMORPH FISHES USNM 230374; C. Pseudomugil tenellus, AMNH 36598SW. Bone stippled; cartilage open circles. It is hypothesized that the Phallostethidae and Dentatherina are sister groups, and that they in turn are members of a group of Indo- study is also needed to determine whether Australian and Pacific, and some New World, these are dermal or endochondral bones. atherinoids that include the taeniomembra- PREMAXILLA AND ROSTRAL CARTILAGE: In dines as far as I have examined them. Thus, phallostethids (fig. 2), Dentatherina and some Roberts's (1971b) speculation that phallo- taeniomembradine atherinids, such as Cra- stethids and taeniomembradines are closely tewcephalus cuneiceps (fig. 7C), the ascend- related is corroborated. Patten and Ivantsoff ing processes of the premaxillae, as well as (1983) named a new subfamily, the Denta- the rostral cartilage, are thin and elongate, as therininae, for their new genus, because they opposed to being short in many other ath- could not place it in any other known subfam- erinids. There are exceptions among New ily of the Atherinidae. The definitions and World menidiine atherinids such as Melan- relationships of the six families of Division orhinus microps (AMNH 25878SW) in which I, the atherinoids, of the Atherinomorpha there are elongate premaxillary ascending (table 1), are poorly known. As stated above, processes; however, I view these as indepen- genera that I believe to be closely related to dently derived in Melanorhinus because oth- phallostethids and Dentatherina have been er characters indicate that it is distantly re- placed in several different families by recent lated to the phallostethids. workers. One such genus, Quirichthys, en- ANAL FIN: In phallostethids, Dentatherina, demic to three river systems in northern Aus- taeniomembradine atherinids (of the genera tralia, has been placed most recently in the Craterocephalus, Atherinomorus and Quir- Atherinidae by Allen (1980, p. 483); how- ichthys), telmatherinids (the genus Telma- ever, previously it has been considered a close 1984 PARENTI: PHALLOSTETHID FISHES relative of the Telmatherinidae, which now ascending process contains a single genus, Telmatherina. Allen articular process (1980) also placed Pseudomugil in the Me- ^alveolar process lanotaeniidae. The Isonidae (or Notocheiri- rostral cartilage^ dae) has been considered by Patten (1978, in Patten and Ivantsoff, 1983) to constitute a autopalatine- subfamily of the Atherinidae. These changes, or suggested changes, in classification serve to point out where additional work is needed. For example, Allen (1980, p. 465) claimed ascending process rostral cartilage-. that melanotaeniids, as he defined them, are articular process distinguished from atherinids in having males alveolar process with elongate dorsal, anal, and pelvic fin rays, and more brightly colored than females. However, not only does Pseudomugil share autopalatine- this characteristic, it is found also in other atherinids such as, for example, Telma- therina. No definition of the Atherinidae in terms of unambiguous, derived characters has ever been proposed. It is inherent in the present argument concerning the relationship of rostral cartilage- the phallostethids, that one does not exist. That is, the Atherinidae is not monophyletic. submaxillary cartilage More important, I believe that there is ad- esethmoid ditional evidence to support Rosen and Par- ascending process ent's (1981) claim that the atherinoids cannot be defined as monophyletic. Two of the FIG. 7. Dorsal view of upper jaw and ethmoid characters for the monophyly of the atherino- region of A. Holocentrus rufus, AMNH 27118SW; morph fishes as a group, of the 10 listed by B. Bedotia geayi AMNH 28132SW; C. Cratero- cephalus cuneiceps, AMNH 43184SW. Bone stip- Rosen and Parent! (1981, pp. 20-21) are the pled; cartilage open circles. derived ethmoid region (their character 9), and the decoupling of the rostral cartilage from the ascending processes of the premaxillae (their character 7). With these charac- oped mesethmoid bone. This is not the case ters, Rosen and Parenti (1981) supported the in atherinomorphs (with the exception of the monophyly of the Atherinomorpha, but they apomorphic Iso) in which the mesethmoid is could not make a firm statement concerning represented by a small wedge or disc of bone its relationship to other acanthopterygians. (fig. 7B-C) or may be cartilaginous as in some They stated only that it was the sister group aplocheilichthyine and orestiine killifishes of the Percomorpha, a group that is itself not (Parenti, 1981). definable as monophyletic. To hypothesize Furthermore, acanthopterygian fishes the polarity of particular characters within primitively have distinct ascending and ar- the Atherinomorpha, therefore, particular ticular processes of the premaxillae, with the groups of percomorphs may be chosen for ascending processes intimately in contact with outgroup comparison. The holocentrid be- the rostral cartilage, and often wrapped ryciforms, which in some characters are around the cartilage so that the processes and primitive and in others derived relative to the cartilage move together as one unit upon atherinomorphs, comprise such a group cho- opening and closing of the mouth (fig. 7A) sen to hypothesize polarity of characters of (see Alexander, 1967). the upper jaw and ethmoid region. Primitively for atherinomorphs, as in the Primitively for acanthopterygian fishes, as bedotiid Bedotia geayi (fig. 7B), the rostral in the holocentrid Holocentrus rufus (fig. 7 A), cartilage is not as firmly attached to the as- the ethmoid region consists of a well-devel- cending processes of the premaxillae, and yet 10 AMERICAN MUSEUM NOVITATES NO. 2779 there is what I identify here as a distinct ar- conflict with two proposed by White, Lav- ticular process that comes in contact with the enberg, and McGowen (in press) who claim maxilla. the atherinoid fishes can be defined as mono- Generally, the ascending is identified as that phyletic by a short preanal length at flexure, process of the premaxilla in contact with the and a unique dorsal pigmentation pattern. rostral cartilage or connected to it via liga- They propose using the ordinal term Athe- ments, and the articular as that process in riniformes for the atherinoids, the fishes of contact with the maxilla, either directly or Division I (table 1). Parsimony does not al- via a connective tissue meniscus. Greenwood low us to choose between the hypothesis of et al. (1966) claimed that atherinomorphs do White, Lavenberg, and McGowen and that not have true ascending processes of the pre- proposed here. However, I feel it is perhaps maxillae. Alternatively, Alexander (1967) premature to treat the fishes of Division I as claimed that atherinomorphs have no artic- monophyletic. Much comparative anatomi- ular processes of the premaxillae, only as- cal work needs to be done, particularly with cending processes. Rosen and Parent! (1981) regard to fin spine development, to hypoth- called the processes in atherinomorphs as- esize the polarity of characters that exhibit cending. I claim that there is a distinct artic- different states among the atherinoids. It has ular as well as ascending process in some been stated repeatedly (e.g., Myers, 1928; Ro- primitive atherinomorphs and that the artic- sen, 1964; Rosen and Parenti, 1981) that ath- ular process is reduced or lost in relatively erinomorph fishes are acanthopterygians in more derived members. In most atherino- which the spinous first dorsal fin is reduced morphs of table 1, minus the Bedotiidae and phylogenetically, from the strongly devel- the Melanotaeniidae sensu stricto, there is no oped spinous dorsal of the bedotiids, to the distinct articular process of the premaxilla, reduction or loss in phallostethids and nu- only a distinct ascending process, as in Cra- merous atherinids, and to its eventual loss terocephalus cuneiceps (fig. 7C). Further- (absence) in cyprinodontiforms and beloni- more, there is less contact between the rostral forms. Yet, the ontogenetic sequence of this cartilage and the ascending processes, and the reduction and loss, as well as other characters two often move rather independently during with which it may be correlated, have never jaw movement (Alexander, 1967). been stated clearly. Therefore, I propose the following classification of atherinomorphs to reflect some of DISCUSSION the findings of this paper: Rosen and Parent! (1981, p. 14) suggested Series Atherinomorpha that the Bedotiidae and Melanotaeniidae Division I sensu stricto, may be primitive relative to all Family Atherinidae other atherinomorphs (that is, the atherinids, Family Bedotiidae telmatherinids, pseudomugilids sensu stric- Family Isonidae to, phallostethids, cyprinodontiforms, and Family Melanotaeniidae beloniforms). Characters suggested to sup- Family Telmatherinidae port this alignment, although just briefly ex- Superfamily Phallostethoidea plained, were conditions of the state of the Family Phallostethidae (includ- dorsal fins, pelvics, extent of spine develop- ing Neostethidae) ment, and number of vertebrae. Family Dentatherinidae The two characters proposed here, reduc- Division II tion of the articular process on the premaxilla Order Cyprinodontiformes and the further decoupling of the rostral car- Order Beloniformes tilage from the ascending processes of the premaxillae, support the hypothesis that all oth- Patten and Ivantsoff (1983) placed their er atherinomorphs are derived relative to the new genus in its own subfamily, the Den- bedotiids and melanotaeniids (minus Pseu- tatherininae; I raise the rank to family. The domugit). order of the taxa is arbitrary, as in the clas- The two characters proposed here are in sification of Rosen and Parenti (1981). Any 1984 PARENTI: PHALLOSTETHID FISHES 11 classification of atherinomorphs must, in my phalus and Quirichthys, the Australian me- opinion, reflect the sister group relationship lanotaeniid or pseudomugilid Pseudomugil, of the phallostethids and Dentatherina. I have and Telmatherina, endemic to Sulawesi. chosen not to place the Dentatherininae in Other taeniomembradines examined, for ex- the Phallostethidae solely for reasons of tra- ample the New World Atherinomorus, have dition. However, it is critical to recognize the the primitive state of a connective tissue me- phallostethids and Dentatherina as belonging niscus. to a group distinct from other members of 3. An enlarged first proximal of the anal Division I. To list all the families without radial fin is present in phallostethids, Denta- this indication would represent a loss of in- therina, Craterocephalus, Quirichthys, Pseu- formation in the printed classification. Should domugil, Telmatherina, and Atherinomorus, we wish to include some of the taeniomem- the taxa listed in 2, above. It has not been bradine atherinids in the group including found in bedotiids, melanotaeniids, and in phallostethids and Dentatherina, they may other nontaeniomembradine atherinids ex- be included in the Phallostethidae, Denta- amined. therinidae, or a third family to be placed in 4. The atherinoid fishes, those of Division the superfamily Phallostethoidea. I of Rosen and Parent! (1981), are not cur- On the other hand, I cannot support the rently definable as monophyletic. monophyly of the fishes of Division I, and 5. The classification of atherinomorph fish- do not use the ordinal term Atheriniformes es proposed above represents the finding of for them. this paper that phallostethids are most closely I tentatively accept the definition of the related to the western Pacific atherinoid Den- Melanotaeniidae of Allen (1980) to include tatherina. Other families listed in the clas- Pseudomugil and its presumed close relative sification, such as the Atherinidae, cannot Popondetta Allen; however, the evidence currently be defined as monophyletic. They herein suggests that Pseudomugil (and per- are currently under revision by other work- haps Popondetta) is not closely related to others. er melanotaeniids, but is rather more closely 6. At least some of the taeniomembradine related to a group that includes phallosteth- atherinids appear to be most closely related ids, Dentatherina, some taeniomembradines to the phallostethids and Dentatherina. In and Telmatherina. particular, the western Australian Cratero- The primary goal of the present paper is cephalus shares the derived character of elon- the clear definition of the relationship of gate premaxillary ascending processes and phallostethids to other atherinomorphs for rostral cartilage with these two groups. Other the purpose of carrying out a taxonomic re- taeniomembradines most likely share this as vision of the included species. The definitions well as other characters with the phallosteth- and relationships of the families not treated ids and Dentatherina. A clearer statement of in detail here are currently under study by the relationship of all taeniomembradines other workers. awaits a revision of the subfamily, as well as the rest of the Atherinidae. SUMMARY 1. The Indo-Australian fresh, brackish, and LITERATURE CITED occasionally saltwater fishes of the family Phallostethidae and the western Pacific ma- Alexander, R. McN. rine atherinid Dentatherina are hypothesized 1967. Mechanisms of the jaws of some ath- to be sister groups that share three derived eriniform fish. Jour. Zool. London, vol. characters: a paradentary bone, absence of 151, pp. 233-255. Allen, G. R. parietals, and absence of a dorsal (first) post- 1980. A generic classification of the rainbow- cleithrum from the pectoral skeleton. fishes (family Melanotaeniidae). Rec. 2. A submaxillary bone or cartilage be- Western Australian Mus., vol. 8, no. 3, tween the maxilla and vomer is present in pp. 449-490. phallostethids, Dentatherina, the Australian Aurich, H. taeniomembradine atherinids Crateroce- 1937. Die Phallostethiden (Unterordnung 12 AMERICAN MUSEUM NOVITATES NO. 2779

Phallostethoidea Myers). Internatl. Rev. Roberts, T. R. Ges. Hydrobiol. Hydrogr., vol. 34, pp. 1971a. The fishes of the Malaysian family Phal- 263-286. lostethidae (Atheriniformes). Breviora, Bailey, R. G. no. 374, 27 pp. 1936. The osteology and relationships of the 1971b. Osteology of the Malaysian phallosteth- phallostethoid fishes. Jour. Morph., vol. oid fish Ceratostethus bicornis, with a 59, no. 3, pp. 453-483. discussion of the evolution of remark- Greenwood, P. H., D. E. Rosen, S. H. Weitzman, able structural novelties in its jaws and and G. S. Myers external genitalia. Bull. Mus. Comp. 1966. Phyletic studies of teleostean fishes, with ZooL, vol. 142, no. 4, pp. 393-418. a provisional classification of living Rosen, D. E. forms. Bull. Amer. Mus. Nat. Hist., vol. 1964. The relationships and taxonomic posi- 131, art. 4, pp. 339-456. tion of the halfbeaks, killifishes, silver- Herre, A. W. sides and their relatives. Bull. Amer. 1925. Two strange new fishes from Luzon. Mus. Nat. Hist., vol. 127, art. 5, pp. Philippine Jour. Sci., vol. 27, no. 4, pp. 217-268. 507-514. Rosen, D. E., and L. R. Parent! 1942. New and little known phallostethids with 1981. Relationships of Oryzias, and the groups keys to the genera and Philippine species. of atherinomorph fishes. Amer. Mus. Stanford Ichth. Bull., vol. 2, no. 5, pp. Novitates, no. 2719, 25 pp. 137-156. Schultz, L. P. Hubbs, C. L. 1948. A revision of six subfamilies of atherine 1944. Fin structure and relationships of phal- fishes, with descriptions of new genera lostethid fishes. Copeia, pp. 69-79. and species. Proc. U.S. Natl. Mus., vol. Myers, G. S. 98, no. 3220, 48 pp. 1928. The systematic position of the phallo- 1950. Correction for "A revision of six stethid fishes, with diagnosis of a new subfamilies of atherine fishes, with de- genus from Siam. Amer. Mus. Novi- scriptions of new genera and species." tates, no. 195, 12 pp. Copeia, no. 2, p. 150. 1935. A new phallostethid fish from Palawan. Smith, H. M. Proc. Biol. Soc. Washington, vol. 48, 1927. The fish Neostethus in Siam. Science, pp. 5-6. vol. 65, pp. 353-355. Parent!, L. R. TeWinkle, L. E. 1981. A phylogenetic and biogeographic anal- 1939. The internal anatomy of two phallo- ysis of cyprinodontiform fishes (Te- stethid fishes. Biol. Bull. Woods Hole, leostei, Atherinomorpha). Bull. Amer. vol. 76, no. 1, pp. 59-69. Mus. Nat. Hist., vol. 168, art. 4, pp. White, B. N., R. J. Lavenberg, and G. E. Mc- 335-557. Gowen Patten, J. M., and W. Ivantsoff [In press] Ontogeny and Systematics of the Ath- 1983. A new genus and species of atherinid eriniformes. In, Ontogeny and system- fish, Dentatherina merceri from the atics of fishes. E. H. Ahlstrom Memorial Western Pacific. Japanese Jour. Ichthy., Symposium, American Society of Ich- vol. 29, no. 4, pp. 329-339. thyologists and Herpetologists. Regan, C. T. Woltereck, R. 1913. Phallostethus dunckeri, a remarkable 1942a. Stufen der Ontogenese und der Evolu- new cyprinodont fish from Johore. Ann. tion von Kopulationsorganen bei Neo- Mag. Nat. Hist., vol. 12, pp. 548-555. stethiden (Percesoces, Teleostei). Inter- 1916. The morphology of the cyprinodont natl. Rev. Ges. Hydrobiol. Hydrogr., fishes of the subfamily Phallostethinae, vol. 42, pp. 253-268. with description of a new genus and two 1942b. Neue Organe, durch postembryonale new species. Proc. London ZooL Soc, Umkonstruction aus Fischflossen en- 26 pp. tstehend. Ibid., vol. 42, pp. 317-355.