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

NovitatesAMERICAN MUSEUM PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2951, 74 pp., 61 figs., 6 tables August 10, 1989

Systematics and Biogeography of the Genus Gambusia (Cyprinodontiformes: Poecilidae)

MARY RAUCHENBERGER

CONTENTS Abstract ...... 2 Introduction ...... 2 Historical Review ...... 3 Methods ...... 6 Abbreviations ...... 8 Materials Examined ...... 9 Acknowledgments ...... 12 Osteology ...... 13 Systematics ...... 32 Outgroup Relationships ...... 32 Diagnosis of the Genus Gambusia ...... 35 Diagnoses of Subgenera ...... 36 Relationships Among Subgenera ...... 40 Relationships Within the Subgenus Heterophallina ...... 42 Relationships Within the Subgenus Arthrophallus ...... 43 Relationships Within the Subgenus Gambusia ...... 49 Biogeography ...... 54 Distribution of the Subgenus Heterophallina ...... 54 Distribution of the Subgenus Arthrophallus ...... 56 Distribution of the Subgenus Gambusia ...... 59 Geographic Areas ...... 63 References ...... 67 Appendix ...... 71

1 Postdoctoral Fellow, Division of Fishes, National Museum of Natural History, Smithsonian Institution.

ABSTRACT The systematic and biogeographic relationships ofGambusia hispaniolae also is included, for com- of the poeciliid genus Gambusia, a monophyletic parison and reference. group of 45 species, are examined in detail. As it Biogeographic analysis of the species groups of focuses primarily on the interrelationships ofsub- Gambusia concentrates on three areas. In southern genera and species groups, this work is not in- Texas and northern Mexico, members of the af- tended as a formal revision ofthe genus; however, finis, nobilis, and senilis species groups overlap; areas where revisionary work are needed are point- their area cladograms are compared. The Rio Pa- ed out. The genus has a sister-group relationship nuco basin of eastern Mexico is a meeting point with the genus Belonesox, and these two in turn for members of the subgenus Heterophallina, the are related to the genus Brachyrhaphis. Gambusia senilis species group, and the afjinis species group. vittata and the members of the rachowi species A repeated pattern indicated a relationship be- group, each previously assigned to monotypic gen- tween the Panuco basin and the Cuatro Cienegas era, do possess the synapomorphies described for basin of northern Mexico, with further ties to the genus Gambusia, and are thus included. Three southern Mexico. Finally, links between Central subgenera within Gambusia are diagnosed: the America and the Greater Antilles are explored subgenus Heterophallina, the subgenus Arthro- through the distribution patterns and area clado- phallus, and the subgenus Gambusia. The subgen- grams of the nicaraguensis, puncticulata, and era Arthrophallus and Gambusia are sister groups. punctata species groups. Repeated patterns in- Within the subgenus Heterophallina, the panuco clude a relationship between Nuclear Central and rachowi species groups are more closely re- America (specifically Belize) and the Antilles, be- lated to each other than either is to G. vittata. In tween Cuba and the Cayman Islands, and between the subgenus Arthrophallus, the nobilis and senilis eastern Cuba, Hispaniola, and the Bahamas. species groups are sister taxa, and in the subgenus Whereas most of the Caribbean patterns do not Gambusia, the puncticulata and punctata species strictly reiterate each other, the patterns are all groups are sister taxa. Relationships within species consistent with each other. It is inferred that they groups also are described. This scheme of rela- represent different fragments of the complex his- tionships, and the new classification derived from tory of Caribbean biogeography, and are com- it, is a product ofcladistic analysis ofmorpholog- pared with some published distributional infor- ical characters. A complete descriptive osteology mation for other taxa.

INTRODUCTION Much attention was drawn to the poeciliid levels; however, areas of the genus that are genus Gambusia Poey, with around 45 in need of revision are noted. species, in 1963, when two independent works Poeciliid taxonomy traditionally has fo- were published dealing with the systematics, cused on the specialization ofthe male's anal and to a lesser extent, the biogeography, of fin, the gonopodium, often to the exclusion this group (Rivas, 1963; Rosen and Bailey, of other characters. In fact, detailed descrip- 1963). Since that time, the conflicts between tions of other parts of the anatomy of any the two schemes have never been resolved, poeciliid generally are not available, with and describers have had difficulty classifying some notable exceptions, such as Rama- their new Gambusia species (Miller, 1975). swami's 1945 description ofthe chondrocra- Systematic studies have emerged that focus nium of Gambusia affinis, and Weisel's simon higher (Parenti, 1981) and lower (Rivas, ilar account ofthe guppy, Poecilia reticulata, 1969; Fink, 197 la, 197 lb; Greenfield, 1985; in 1967. My osteological account ofthe mem- Echelle and Echelle, 1986) systematic levels; bers of the genus Gambusia uses a male G. this work aims to classify the entire genus, hispaniolae as an exemplar, but discusses the and to establish its close relatives, through range of variation in the genus for all char- the principles and methods of cladistic sys- acters. It will serve both as an introduction tematics. This paper is not intended as a for- to the characters important for systematic mal revision of the genus, as it concentrates analysis of the genus, and as a general ref- primarily on the subgenus and species group erence for the complete osteology of a poe- 1989 RAUCHENBERGER: GAMBUSIA 3

TABLE 1 TABLE 1-(Continued) Proposed Classification of Gambusia GAMBUSIA Poey, 1854 G. wrayi Regan, 1913 Genus puncticulata species group: Gambusia Poey, 1854: 382 (type species Gambusia G. yucatana Regan, 1914 punctata Poey, by subsequent designation by Jordan G. hispaniolae Fink, 197 lb and Copeland, 1876: 142) G. manni Hubbs, 1927 Heterophallus Regan, 1914: 66 (type species Hetero- G. hubbsi Breder, 1934 phallus rachovii, by monotypy) G. bucheri Rivas, 1944 Flexipenis Turner, 1940: 89 (nomen nudem for Gam- G. baracoana Rivas, 1944 busia vittata Hubbs, which was subsequently desig- G. monticola Rivas, 1971 nated as the type species ofFlexipenis, by C. L. Hubbs, G. puncticulata Poey, 1854 in Rivas, 1963: 334) G. oligosticta Regan, 1913 Dicerophallus Alvarez, 1952: 95-97 (type species Dicero- G. caymanensis Regan, 1913 phallus echeagayari, by original designation) G. howelli Rivas, 1944 Subgenus Heterophallina Hubbs, 1926:26 (type species punctata species group: G. regani, by original designation) G. luma Rosen and Bailey, 1963 G. vittata Hubbs, 1926 G. beebei Myers, 1935 panuco species group: G. pseudopunctata Rivas, 1969 G. marshi Minckley and Craddock, in Minkley, G. punctata Poey, 1854 1962 G. rhizophorae Rivas, 1969 G. panuco Hubbs, 1926 G. xanthosoma Greenfield, 1983 G. regani Hubbs, 1926 rachowi species group: G. rachowi (Regan, 1914) ciliid, valuable for higher-level comparative G. echeagayari (Alvarez, 1952) work. Subgenus Arthrophallus Hubbs, 1926 (type species The species of Gambusia live in fresh and Heterandriapatreulis Baird and Girard, a subjective brackish water, and range from the eastern synonym of Heterandria affinis Baird and Girard; by original designation) and southern United States through Mexico affinis species group: and Central America, with a single species G. affinis (Baird and Girard, 1854) along the South American coast. Particularly G. holbrooki Girard, 1859b interesting are the 16 taxa found on islands G. speciosa Girard, 1859c in the West Indies. Speculations about how G. lemaitrei Fowler, 1950 these fishes found their island homes, invok- G. aurata Miller and Minckley, 1970 ing phenomena such as land bridges, migra- nobilis species group: tion, and hurricanes, have abounded in the G. nobilis (Baird and Girard, 1854) past, but subsequent developments in the sci- G. heterochir Hubbs, 1957 have made it G. krumholzi Minckley, 1963 ence of biogeography possible G. georgei Hubbs and Peden, 1969 to critically assess hypotheses concerning dis- G. eurystoma Miller, 1975 tribution patterns. A necessary prerequisite G. sexradiata Hubbs, 1936 for such a biogeographic analysis is a sound senilis species group: systematic hypothesis for the fishes involved. G. senilis Girard, 1 859c Only then can the nature of the relationship G. gaigei Hubbs, 1929 between the organisms and the areas they G. alvarezi Hubbs and Springer, 1957 inhabit be determined. The major objective G. hurtadoi Hubbs and Springer, 1957 of this work, therefore, is to describe rela- G. amistadensis Peden, 1973b tionships within the genus, which are sum- G. geiseri Hubbs and Hubbs, in Hubbs and Spring- marized in the classification (table er, 1957 proposed G. longispinis Minckley, 1962 1), and to discuss their biogeographic impli- G. atrora Rosen and Bailey, 1963 cations. Subgenus Gambusia nicaraguensis species group: HISTORICAL REVIEW G. nicaraguensis Giinther, 1866 The genus Gambusia was described by Poey G. melapleura (Gosse, 1851) in 1854. Although he realized that the two 4 AMERICAN MUSEUM NOVITATES NO. 2951

TABLE 2 evinced by the names assigned to subgenera. Classification of Gambusia, Based on Subgenera Arthrophallus and Schizophallus, Rivas (1963) both monotypic, contained G. patreulis and Genus Flexipenis G. holbrooki, respectively. Subgenus Het- F. vittata erophallina was created for three newly de- Genus Heterophallus scribed species, G. vittata, G. regani, and G. H. rachovii panuco, and the remaining 11 species in the Genus Dicerophallus genus were placed in the subgenus Gambusia. D. echeagayari Hubbs also created a tribe, the Gambusiini, Genus Gambusia to contain allied genera -the monotypic Het- Subgenus Orthophallus erophallus and Belonesox. G. lemaitrei By the time ofthe next review ofthe genus Subgenus Arthrophallus (1963), 30 species bore the name Gambusia. G. affinis Subgenus Heterophallina Rivas (1963) removed Hubbs' G. vittata to a G. panuco G. regani G. marshi monotypic genus, Flexipenis, in the tribe Subgenus Gambusia Gambusiini, primarily because it lacks the senilis species group: distinctive elbow on ray 4a of the gonopo- G. geiseri G. senilis G. gaigei dium. He also listed four subgenera; whereas G. hurtadoi G. alvarezi G. longispinis he synonymized Schizophallus with Arthro- nobilis species group: phallus, a new subgenus, Orthophallus, was G. nobilis G. heterochir G. sexradiata created for G. lemaitrei, from Colombia. nicaraguensis species group: Within the subgenus Gambusia, he described G. melapleura G. nicaraguensis G. wrayi G. gracilior G. dominicensis five species groups: the senilis species group, puncticulata species group: with 6 members; nobilis, 3 members; nica- G. bucheri G. puncticulata G. howelli raguensis, 5 members; puncticulata, 9 mem- G. baracoana G. manni G. hubbsi bers; and punctata, 2 members (table 2). The G. oligosticta G. caymanensis G. yucatana criteria he employed in grouping were gon- punctata species group: opodial characters and geographic distribu- G. punctata G. beebei tion. Also in 1963, Rosen and Bailey published a review ofthe Poeciliidae. Their tribe Gam- species he assigned to the genus (G. punctata, busiini contained Brachyrhaphis, Belonesox, type species, and G. puncticulata) were poe- and Gambusia; their genus Gambusia listed ciliids because of the presence of the male's 34 species, including two described as new. gonopodium, he distinguished Gambusia They also included Flexipenis vittata and from other genera of poeciliids that he en- Heterophallus rachovii in the genus Gambu- countered in his studies of Cuban fishes on sia. While emphasizing the need for a thor- the basis of body shape, tooth structure, and ough study ofinterrelationships in the genus, carnivorous habit. Many more species now they recognized six species groups: affinis referred to the genus Gambusia were de- species group, 17 species; punctata, 4 species; scribed in the years that followed, but there nobilis, 7 species; panuco, 3 species; vittata, was much confusion in generic assignment. monotypic; and rachowi, 2 species (table 3). It was Regan, in 1913, who realized that "dif- These groupings primarily were based on ferences in structure ofthe intromittent organ gonopodial specializations, but such char- are ofgreat systematic importance" (p. 978), acters as pigmentation, vertebral number, and and he described Gambusia as being char- body and fin configuration were also consid- acterized by the spines on ray 3, elbow in ray ered. 4a, and hooks on rays 4p and 5a of the gon- These two treatments in 1963 were con- opodium. He listed 17 species, eight ofwhich current and incongruent, both in delineation he described as new. of groups and assignment to groups, and the Hubbs (1924, 1926) was the first to sub- overall status of relationships within the ge- divide the genus. By this time, the heavy em- nus, or of generic limits, has not been ad- phasis placed upon gonopodial structure was dressed since. 1989 RAUCHENBERGER: GAMBUSIA 5

Several studies have concentrated on re- TABLE 3 lationships within species groups, following Classification of Gambusia, Based on Rosen and one scheme or the other. Rivas (1969) revised Bailey (1963: 92-107) the punctata species group, describing two Genus Gambusia new species. Here he sharply disagreed with affinis species group: Rosen and Bailey's (1963) delineation ofthis G. nicaraguensis G. aestiputius G. lemaitrei group, and reiterated his belief that "only G. affinis G. melapleura G. wrayi gonopodial characters (including the suspen- G. gracilior G. dominicensis G. myersi sorium) appear to be of value in recognizing G. yucatana G. puncticulata G. howelli and characterizing subgenera and species G. baracoana G. bucheri G. manni groups" (p. 779). In this revision, as in his G. oligosticta G. caymanensis 1963 work, Rivas did not explicitly describe punctata species group: G. sexradiata G. luma G. punctata relationships, although one is tempted to in- G. beebei fer them from his classification. nobilis species group: Revisions of the nicaraguensis and punc- G. atrora G. longispinis G. nobilis ticulata species groups, again as defined by G. senilis G. gaigei Rivas, appeared in 1971 (Fink, 1971a, panuco species group: 1971b). These were detailed studies docu- G. panuco G. regani G. marshi menting the range of variation within each vittata species group: of these groups, particularly in reference to G. vittata meristics and gonopodial features, but no rachowi species group: schemes of relationships were presented. G. rachowi G. echeagayari Further work on the puncticulata species group (Greenfield et al., 1982; Greenfield and Wildrick, 1984; Greenfield, 1985) has in- place G. eurystoma in either the nobilis or cluded other types ofcharacters, such as pig- punctata group, suggested that "species groups mentation, body proportions, and electro- of Gambusia be employed with caution until phoretic analysis. One of these studies phyletic studies are completed" (1975: 1). (Greenfield and Wildrick, 1984) did present The taxonomic problems that need to be a branching diagram to summarize its results, addressed include an assessment ofthe limits and its implications are discussed in the Sys- of the genus. C. L. Hubbs, and later Rivas, tematics section. Unfortunately, these work- tended to remove any anomalous fish from ers lacked an overall scheme ofrelationships the main body ofthe genus, either to a mono- for the genus, necessary for outgroup com- typic genus, as with Flexipenis vittata, Het- parisons, and their work has been criticized erophallus rachovii, and Dicerophallus echea- because of small sample sizes. gayari, or to a monotypic or small subgenus, New species described since 1963 some- Orthophallus for G. lemaitrei, Arthrophallus times have been easy to classify, such as G. for G. affinis, and Heterophallina for the three- xanthosoma in the punctata species group member panuco group. (Greenfield, 1983) or G. amistadensis in the Ifclassification is not necessarily meant to senilis species group (Peden, 1973b), but reflect relationship, this tactic ofsequestering often, especially with taxa described from the odd fishes is a convenient way to deal with southern Texas-northern Mexico area, the them; however, I believe the most useful clas- new species simply did not fit neatly into sifications are those that directly reflect ideas either of the Rivas or Rosen and Bailey ofrelationship, and the sister-group relation- schemes. Hubbs and Peden (1969) felt that ships implied by that classification with the G. georgei might lie either in the affinis or monotypic taxa have not been substantiated. nobilis species group. Minckley's (1963) G. Similarly, the delineation of subgenera and krumholzi showed characters both of the species groups should be arranged according nicaraguensis species group and the nobilis to relationship, not convenience or geog- species group, and Miller and Minckley's raphy. (1970) G. aurata also falls between the afJinis The relationship ofGambusia to other poe- and nobilis species groups. Miller, trying to ciliids has not been addressed since 1963, 6 AMERICAN MUSEUM NOVITATES NO. 2951 when Rosen and Bailey allied the genus with problems of circularity when one wishes to Belonesox and Brachyrhaphis. Beyond that, make general statements about biogeography relationships among poeciliids are yet to be based on systematics. critically examined. The position of poecil- Croizat (1958, 1964) observed that groups iids within the order Cyprinodontiformes was of related organisms (even secondary-divi- recently reappraised by Parenti (1981), and sion freshwater fishes) generally are not dis- her classification of the order is employed tributed randomly; in fact, he documented herein. an amazing degree of congruence in distri- Interpretation of Gambusia distribution, bution patterns when surveying many groups. particularly in the Caribbean, historically has Rather than postulating that the distribution been heavily influenced by a paper by Myers pattern of each group requires a separate ex- (1937) entitled "Fresh-water fishes and West planation, as would be needed in a scenario Indian zoogeography." In this, he differen- based on chance dispersal, Croizat proposed tiated two kinds of freshwater fishes: pri- that these congruent patterns (generalized mary-division fishes, such as ostariophysans, tracks) have a single cause, that "life and earth that are "very strictly confined to fresh water" evolve together." Rosen (1975) applied Croi- due to "an ancient physiological inability to zat's method to the problem of Caribbean survive in salt sea water" (pp. 342-343), in- biogeography, and concluded that the distri- ferred because nearly all members of such bution of Gambusia in the Antilles and Cen- groups are never found in even brackish water; tral America was but one example of a gen- and secondary-division fishes, including cy- eralized track relating North America to the prinodontiforms, that are more tolerant to Caribbean. saltwater conditions. "It is evident that many Obviously, to discern the interdependence species of this secondary group might easily oforganismic relationships and biogeograph- survive a short seajourney. This is borne out ic relationships, the groups utilized must be by distributional fact'' (p. 345). Regardless of natural units; that is, monophyletic groups. the fact that these fishes do not habitually As expounded by Nelson and Platnick (1981), migrate from island to island, such "coloniz- the employment of branching diagrams to ing flights," of the type so maligned by Croi- portray levels of relationship is useful not zat (1964), are exactly the mechanism in- only in systematics but also in biogeography. voked to explain the distribution ofpoeciliids With such a tool, not only can one discern in the Greater Antilles, simply because some broad patterns of distributional congruence members ofthe family do not die in salt water. (Croizat's generalized tracks) but also finer Biogeography has often, until recently, been patterns ofcongruence within the tracks, cor- relegated to the last few pages of systematic responding to the components of the clado- revisions, and Gambusia revisers have tried grams. The classic example ofthis method is to explain the distribution of the taxa based found in Rosen (1979), also discussed in Nel- on their systematic conclusions (Hubbs and son and Platnick (1981), Humphries and Pa- Springer, 1957; Rosen and Bailey, 1963; Ri- renti (1986), and many other sources. The vas, 1963, 1969; Fink, 1971a, 1971b). Two method is underutilized, mostly because of obstacles have prevented this endeavor from the prerequisite for resolved cladograms of being as fruitful as it might have been. First, related taxa that are endemic to certain areas. because of the influence of Myers' judgment The ultimate purpose of this work is to pro- that random dispersal by poeciliids might not vide such cladograms, to be used in biogeo- be that uncommon and ofthe prevailing no- graphic analysis, for monophyletic groups tions of stabilist geography, related species within Gambusia. were uncritically assumed to have achieved their current stations through migration. Sec- METHODS ondly, oftentimes notions of "geographical cohesiveness" were used to justify certain The descriptive osteology follows the style systematic arrangements, i.e., using geo- of a similar work for a characin (Weitzman, graphic proximity to indicate close system- 1962); thus I have chosen a single specimen, atic relationship. This obviously leads to a mature male Gambusia hispaniolae, to il- 1989 RAUCHENBERGER: GAMBUSIA 7 lustrate and describe. However, the descrip- well described, nor, in Gregory's case, very tion also is comparative, in the sense that accurate. Only works dealing with higher level those characters that vary in the genus also groups have addressed other parts ofthe skel- are discussed. The systematic implications of eton. these characters are treated in the Systematics Osteological observations were made on section. cleared and stained specimens, following the With characters that do vary, I have tried method ofDingerkus and Uhler (1977), where to indicate the level of generality that they bone is stained with alizarin red and cartilage describe. For features that characterize higher stained with alcian blue. Numbers of speci- groups that include Gambusia, I have relied mens prepared per lot are listed in the Ma- on several sources-Rosen and Bailey (1963) terials Examined section; in general two for poeciliids, Rosen (1964) and Rosen and males, two females, two juveniles, and em- Parenti (1981) for atherinomorphs, and, par- bryos taken from gravid females were pre- ticularly, Parenti (1981) for cyprinodonti- pared, collection size permitting. Several forms. I also have noted where my obser- Gambusia species were raised in the lab to vations conflict with the predictions ofthose provide ontogenetic series. Drawings are ink works. These apparent contradictions are due, tracings of pencil renderings, using a camera in large part, to the probably necessary, but lucida attached to a Leitz binocular dissecting unfortunately untrue, assumption that a few microscope. Meristic, pigmentary, and sen- poeciliids, say, Tomeurus or Xiphophorus, are sory canal characters were observed on well- representative of all poeciliids. Although preserved alcoholic specimens. poeciliids certainly are well corroborated as Systematic analysis of character distribu- a monophyletic group, there is a great deal tions follows the method of Hennig (1966) of diversity among them. as set forth by Nelson and Platnick (1981). As noted in the introduction, few osteo- The aim is to discover and describe mono- logical treatises on poeciliids have appeared. phyletic groups, and characterize them by Rosen and Bailey (1963), of course, figured synapomorphies. Groups characterized by and discussed many gonopodia and gono- primitive characters, or by the lack ofderived podial suspensoria. They also figured many characters, are disallowed. Determining the dorsicrania and radiographs of whole fish, level ofgenerality at which a particular char- but left these undiscussed. The gonopodium acter is to be considered derived, i.e., what has received much attention with regard to group it is a synapomorphy of, is the most its structure, vascularization, innervation, difficult and at the same time the most chal- functional morphology (Rosen and Gordon, lenging part of systematics. Helpful in this 1953), behavior (Rosen and Tucker, 1961; effort are the techniques of outgroup com- Peden, 1972b), and development (Turner, parison and ontogenetic investigation, both 1941, 1942a) and similar attention has fo- ofwhich aid in discovering homologies (sen- cused on the other sexually dimorphic sys- su Patterson, 1982). A common pitfall in- tems of poeciliids, such as the gonopodial volves too heavy a reliance on a single out- suspensorium (Turner, 1942b; Rosen and group, rather than surveying a wide range of Kallman, 1959), pectoral fin (Turner, 1942c; taxa of ever-more-encompassing groups; I Hubbs and Reynolds, 1957), and female gen- have tried to avoid this danger but undoubt- italia (Peden, 1972a, 1973a). Other areas have edly have not escaped it completely. received relatively little attention, the only Character distributions are not always con- exceptions as follows. Cranial development gruent; parsimony is the criterion employed in Gambusia affinis was examined by Ra- to choose between rival schemes, in that the maswami (1945), and in Poecilia reticulata most parsimonious diagram requires the few- by Weisel (1967). Hollister (1940) described est ad hoc assumptions. Parsimony analysis the caudal skeletons of G. holbrooki and Poe- was aided by Swofford's PAUP program (ver- cilia latipinna. And although de Beer (1937) sion 2.4.1), run on an IBM-PC. The classi- included a figure of G. affinis and Gregory fication presented in this work is a direct re- (1933) figures ofBelonesox belizanus in their flection of the systematic conclusions based respective treatises on fish skulls, neither is on parsimony analysis, such that all taxa are 8 AMERICAN MUSEUM NOVITATES NO. 2951 monophyletic, and the convention of se- CBB4 Cartilaginous basibranchial 4 quencing specifies interrelationships. CH Ceratohyal Biogeographic analysis uses the result of CL Cleithrum the systematics analyses, the cladograms of CLP Cleithral process CMK Coronomeckelian bone monophyletic taxa, as input data. Instead of COR Coracoid simply implying relationships among taxa, DEN Dentary these branching diagrams now also imply re- DOR Dorsal process of maxilla lationships among areas. A comparison of DR Distal radial these area cladograms will reveal any con- DSP Dermosphenotic gruence ofdistribution patterns among them, EB 1-4 Epibranchials 1-4 to the extent that the groups studied overlap ECT? Ectopterygoid? in their areas of endemism. A simple inter- ELB Elbow pretation is that congruence implies shared END Endopterygoid history, incongruence independent dispersal EP Epural EPI Epioccipital events. However, Rosen (1985) recently dis- EPIP Epioccipital process cussed the meaning of incongruence in area EPR Epipleural rib cladograms. Ifan area, such as the Caribbean, EXO Exoccipital has had a complex history, more than one FM Foramen magnum generalized track may reflect that history. In FR Frontal terms of area cladograms, there will be in- GPl-3 Gonapophyses 1-3 congruent components that nevertheless re- GPUN Uncini of gonapophyses flect the same history; however, each shows GNA Gonactinosts just a part ofthe complete picture, a part that HBI-3 Hypobranchials 1-3 may be temporally and/or spatially displaced HHC Hypohyal cartilage HYM Hyomandibula from other parts. HYP Hypural A fuller use of these area cladograms for HYPP Hypural plate biogeographic analysis of the regions occu- IAC Interarcual cartilage pied by Gambusia species, therefore, will only IH Interhyal come when cladograms are available for many lOP Interopercle other taxa that also occupy those areas; this IPB2,3 Infrapharyngobranchials 2, 3 work is a step toward that end. IPR Inner process LAC Lacrimal ABBREVIATIONS LET Lateral ethmoid LIG Ligastyle Institutional: MAX Maxilla MES Mesethmoid AMNH American Museum ofNatural History MKC Meckel's cartilage ANSP Academy ofNatural Sciences ofPhila- MR Medial radial delphia NAS Nasal UMMZ University of Michigan, Museum of NP Neuropophysis Zoology NS Neural spine USNM United States National Museum (Na- OP Opercle tional Museum of Natural History, PAR Parietal Smithsonian Institution) PAS Parasphenoid PC Postcleithrum Anatomical: PH Parhypural AAR Anguloarticular PMX Premaxilla ACC Accessory cartilage POP Preopercle APL Autopalatine PQC Palatoquadrate cartilage ASC Ascending process of premaxilla PR Pleural rib BB2,3 Basibranchial 2, 3 PRO Prootic BH Basihyal PRR Proximal radial BOC Basioccipital PTT Posttemporal BSR Branchiostegal rays PU2 Preural centrum 2 CBI-5 Ceratobranchials 1-5 Q Quadrate 1989 RAUCHENBERGER: GAMBUSIA 9

RAD Radials UMMZ 198934 (177). Rio Mesquites, Cuatro RAR Retroarticular Cienegas, Coahuila, Mexico. SC Scapula G. panuco: SCF Scapular foramen AMNH 32390 (9/29). Rio Soto la Marina SCL Supracleithrum system, Tamaulipas, Mexico. SER Serrae AMNH 78131 (4/150). Rio Coy, San Luis SOC Supraoccipital Potosi, Mexico. SOCP Supraoccipital process UMMZ 198801 (143). Rio Verde, San Luis SOP Subopercle Potosi, Mexico. SP Spines G. regani: SPH Sphenotic AMNH 78132 (7/100). Rio Sabinas system, SYM Symplectic Tamaulipas, Mexico. THC Terminal half-centrum UMMZ 164734 (48). Arroyo del Encino, En- TP-IPB4 Fourth infrapharyngobranchial tooth- cino, Tamaulipas, Mexico. plate rachowi species group UH Urohyal G. rachowi: US Urostyle AMNH 78133 (4/15). Mexico. V Vomer UMMZ 187761 (164). Rio Chiquito, Te- VHH Ventral hypohyal nochtitla, Veracruz, Mexico. UMMZ 167098 (1) holotype of G. atzi (= G. rachowi). Jesus Carranza, Veracruz, Mex- MATERIALS EXAMINED ico. G. echeagayari: For species of Gambusia examined, local- UMMZ 191731 (3/40). Palenque, Chiapas, ity information is provided and the species Mexico. are arranged according to the new classification proposed herein. Other comparative Subgenus Arthrophallus affinis species group material is arranged alphabetically within hi- G. affinis: erarchical groupings; i.e., other poeciliids, UMMZ 122760 (3/349). Wheeler Reservoir, other cyprinodontiforms, other atherino- Alabama. morphs. Locality information is not provid- UMMZ 129784 (1580). Limola Lake, Brazos ed for these materials. Numbers ofspecimens Co., Texas. are indicated in parentheses, first the number G. holbrooki: of cleared and stained specimens, followed AMNH 52245 (5/30). Green Cove Spring, St. by the total number in the lot. When only John Co., Florida. one number is given, the lot consists of only AMNH 16255 (1/3). Cochabamba, Bolivia. alcoholic specimens. UMMZ 163475 (61). Jackson Co., Florida. G. speciosa: Gambusia: AMNH 77947 (4/30). Melchor Musquiz, Rio Subgenus Heterophallina Grande drainage, Coahuila, Mexico. G. vittata: UMMZ 120320 (684). Devils River, Verde AMNH 32391 (1/12). Rio Soto la Marina Co., Texas. system, Tamaulipas, Mexico. UMMZ 196738 (177). Arroyo La Salada, AMNH 38498 (1/5). Rio Axtla, San Luis Po- Coahuila, Mexico. tosi, Mexico. G. aurata: UMMZ 202820 (126). Rio Choy, San Luis AMNH 75821 (7/62). Nacimiento del Rio Potosi, Mexico. Mante, Tamaulipas, Mexico. AMNH 75822 (3/50). Rio Mante, Tamauli- UMMZ 188736 (1) holotype. Rio Mante, Ta- pas, Mexico. maulipas, Mexico. AMNH 75754 (6/79). Rio San Rafael, Ta- nobilis species group maulipas, Mexico. G. nobilis: panuco species group AMNH 52233 (4/200). Dexter National Fish G. marshi: Hatchery, Eddy Co., Texas. AMNH 78130 (7/50). Cuatro Cienegas, Coa- UMMZ 179801 (681). Phantom Cave, Jeff huila, Mexico. Davis Co., Texas. UMMZ 179167 (1) holotype. Rio Salado, 2 UMMZ 132268 (27). Tunis Spring, Pecos Co., mi south Hermanos, Coahuila, Mexico. Texas. 10 AMERICAN MUSEUM NOVITATES NO. 2951

G. krumholzi: UMMZ 190407 (1) holotype. Goodenough UMMZ 180322 (3/87) paratopotypes. Rio de Springs, Val Verde Co., Texas. Mara, Coahuila, Mexico. UMMZ 190408 (3/65) paratopotypes. Good- UMMZ 180320 (1) holotype. Nava, Coahui- enough Springs, Val Verde Co., Texas. la, Mexico. G. geiseri: G. georgei: UMMZ 120247 (3/130) paratypes. San Mar- UMMZ 187488 (3/43) paratopotypes. San cos River, Hays Co., Texas. Marcos River, Hays Co., Texas. UMMZ 168974 (1) holotype. San Marcos UMMZ 187447 (1) holotype. San Marcos River, Hays Co., Texas. River, Hays Co., Texas. G. longispinis: G. heterochir: UMMZ 130382 (2/50). Cuatro Cienegas, UMMZ 170937 (6) paratopotypes. Clear Coahuila, Mexico. Creek, Menard Co., Texas. UMMZ 179620 (2) holotype. Cuatro Ciene- UMMZ 170936 (1) holotype. Clear Creek, gas, Coahuila, Mexico. Menard Co., Texas. G. atrora: G. eurystoma: AMNH 40812 (30/728). Rio Axtla, San Luis UMMZ 184717 (3/360) paratypes. Arroyo del Potosi, Mexico. Azufre, Teapa, Tabasco, Mexico. UMMZ 179990 (1) holotype. Rio Axtla, San UMMZ 197600 (1) holotype. Arroyo del Luis Potosi, Mexico. Azufre, Teapa, Tabasco, Mexico. G. sexradiata: Subgenus Gambusia AMNH 24529 (17/122). Rio de la Pasi6n, nicaraguensis species group Alta Verapaz, Guatemala. G. nicaraguensis: AMNH 25615 (1/13). Rio de la Pasion, Pe- AMNH 20849 (10/64). Great Corn Island, ten, Guatemala. Nicaragua. AMNH 32025 (5/57). Rio de la Pasion, Pe- AMNH 18830 (2/7). Guatemala. ten, Guatemala. AMNH 28612 (2/8). Rio Chagres, Panama. AMNH 24516 (4/31). Rio de la Pasi6n, Alta UMMZ 199461 (73). Port Royal, Honduras. Verapaz, Guatemala. G. melapleura: UMMZ 102989 (1) holotype (of G. nicara- USNM 205559 (372). Shrewsberry River, Ja- guensis sexradiata). Rio Papaloapan, Oa- maica. xaca, Mexico. USNM 205555 (242). Bluefields River, Ja- senilis species group maica. G. senilis: G. wrayi: UMMZ 209026 (3/270). Rancho Houmigas, USNM 104338 (1/12). Roaring stream, Ja- Chihuahua, Mexico. maica. UMMZ 166712 (135). Rio San Juan, Rio AMNH 74080 (8/104). Kingston Lagoon, Ja- Concho, Chihuahua, Mexico. maica. G. cf. senilis: UMMZ 196801 (4). 30 mi east of Montego UMMZ 211115 (3/89 1). Balnearia, San Die- Bay, Jamaica. go, Chihuahua, Mexico. USNM 205578 (193). Patterson Spring, G. gaigei: Kingston, Jamaica. UMMZ 84527 (1) holotype. Rio Grande at G. gracilior: Boquillas, Brewster Co., Texas. USNM 206286 (36). Stony River, Chapelton, G. hurtadoi: Jamaica (cat. as G. wrayi). UMMZ 211112 (3/255). Ojo (Hacienda) Do- puncticulata species group lores, Chihuahua, Mexico. G. yucatana: UMMZ 168975 (1) holotype. Ojo (Hacienda) AMNH 32261 (9/79). Laguna Yaxja, Peten, Dolores, Guatemala. Chihuahua, Mexico. AMNH 32364 (1/4). Naranjal Reservoir, Pe- G. alvarezi: ten, Guatemala. AMNH 27492 (4/4). El Ojo de San Gregorio, AMNH 32389 (1/38). Laguna Yaxja, Peten, Chihuahua, Mexico (cat. as G. gaigei). Guatemala. UMMZ 211110 (3/295). El Ojo de San Gre- G. hubbsi: gorio, Chihuahua, Mexico. AMNH 12455 (8) paratypes. South Andros UMMZ 168979 (1) holotype. El Ojo de San Island, Bahamas. Gregorio, Chihuahua, Mexico. AMNH 12979 (1/89). Long Island, Bahamas G. amistadensis: (cat. as G. puncticulata). 1989 RAUCHENBERGER: GAMBUSIA I11

G. manni (AMNH lots cat. as G. manni, but British West Indies (cat. as G. puncticula- more probably G. hubbsi): ta). AMNH 28804 (10/142). Grand Bahama, Ba- G. howelli: hamas. UMMZ 143374 (9) paratypes. Punta del Este, AMNH 28680 (1/10). Berry Island, Baha- Isla de la Juventud, Cuba (cat. as G. punc- mas. ticulata). AMNH 12459 (1/3). Andros Island, Baha- punctata species group mas. G. luma: AMNH 34353 (1/3). Conception Island, Ba- AMNH 35161 (3/5). Rio Sarst(un system, Iza- hamas. bal, Guatemala. UMMZ 64246 (5). Lake Kilkarney, New AMNH 35096 (1/29). Rio Dulce, Izabal, Providence Island, Bahamas. Guatemala. UMMZ 72183 (1) holotype. New Providence UMMZ 143565 (1) holotype. Puerto Barrios, Island, Bahamas. Izabal, Guatemala. G. hispaniolae: G. beebei: AMNH 37344 (8/37). Between Port-au-Prince USNM 203162 (163). Lake Miragoane, Haiti. and Lake Miragoane, Haiti. USNM 203161 (1) neotype. Lake Miragoane, UMMZ 136377 (11) paratypes. Locality mis- Haiti. labeled as Cuba. G. pseudopunctata: AMNH 78134 (1/1). Source Trou Caiman, AMNH 78136 (1/1). Les Cayes, Haiti. Haiti. Exemplar used in descriptive os- ANSP 158843 (31). Stream 2 km west Les teology. Cayes, Haiti. AMNH 78135 (6/16). Aquarium material, UMMZ 208139 (25) paratypes. Roseaux, descendants offishes caught at Source Trou Dept. de Sud, Haiti. Caiman, Haiti. USNM 203163 (1) holotype. Roseaux, Dept. de Sud, Haiti. AMNH 37346 (2/11). East ofPort-au-Prince, G. punctata: Dominican Republic. AMNH 3081 (1/2). Santa Clara, Cuba. USNM 204865 (1) holotype. Source Trou USNM 203165 (1) neotype. Havana, Cuba. Caiman, Haiti. G. xanthosoma: G. bucheri: UMMZ 210200 (1) paratype. West Bay, UMMZ 143373 (3/6) paratypes. Rio Jicotea, Grand Cayman Island, British West Indies. Moa system, Oriente, Cuba. UMMZ 213616 (17). Grand Cayman Island, USNM 203149 (1) holotype. Rio Jicotea, Moa British West Indies. system, Oriente, Cuba. G. G. rhizophorae: baracoana: AMNH 52329 UMMZ 143375 (3/6) paratypes. Rio Miel, (1/3). Crawl Key, Monroe Co., Baracoa, Oriente, Cuba. Florida. USNM 203150 (1) holotype. Rio Miel, Ba- UMMZ 213650 (100). Monroe Co., Florida. racoa, Oriente, Cuba. USNM 203223 (1) holotype. Dade Co., Flor- G. puncticulata: ida. AMNH 9192 (2/2). Santa Clara, Cuba. UMMZ 103336 (191). Artemisa, Pifiar del Other Poeciliids Rio, Cuba. Alfaro cultratus (Regan, 1908) - AMNH 10589 USNM 204402 (41). Rio Jaimanitas, Ha- (2/6) vana, Cuba. Brachyrhaphis cascajalensis (Meek and Hilde- USNM 167691 (50). Cienfuegos Soledad, brand, 1913) - AMNH 37810 (1/43); AMNH Santa Clara, Cuba. 28591 (4/12) G. oligosticta: Brachyrhaphis episcopi (Steindachner, 1878) - USNM 104339 (2/8). Spanish Town, Jamai- AMNH 20592 (4) ca. Brachyrhaphis terrabensis (Regan 1907) - AMNH UMMZ 190129 (92). Port Henderson, Ja- 20503 (3/9) maica (cat. as G. puncticulata). Belonesox belizanus Kner, 1860 - AMNH 32027 G. caymanensis: (1/22); AMNH 27493 (5/5) USNM 089789 (2/7). Grand Cayman Island, Carlhubbsia stuarti Rosen and Bailey, 1959 - British West Indies. AMNH 35103 (6/83) UMMZ 213615 (20). Grand Cayman Island, Cnesterodon sp. - AMNH 35367 (4/30) 12 AMERICAN MUSEUM NOVITATES NO. 2951

Girardinus uninotatus Poey, 1860 - AMNH 37749 Girardinichthys viviparus (Bustamante, 1837) - (1/4) UMMZ 192379 (55/55) Heterandria attenuata Rosen and Bailey, 1979 - Kosswigichthys asquamatus S6zer, 1942 - ANSP AMNH 3291 1SW (4/76) 89883 (1/1) Heterandria jonesi (Gunther, 1874) - AMNH Megupsilon aporus Miller and Walters, 1972 - 3678 1SW (4/4) AMNH 38405SW (2/10) Limia nigrofasciata Regan, 1913 - AMNH 37349 Orestias forgeti Lauzanne, 1981 - AMNH (6/82) 52113SW (1/4) Neoheterandria umbritalis (Meek, 1912) - AMNH Pantanodon madagascarensis (Arnoult, 1963) - 37665 (4/20) AMNH 20526 (1/4) Phallichthys amates (N. Miller, 1907) - AMNH Profundulus labialis (Gunther, 1866) - AMNH 20593 (2/6) 24567 (2/2) Poecilia caucana (Steindachner, 1880) - AMNH Profundulus punctatus (Gunther, 1866) - AMNH 28137 (19/99) 24432 (2/2) Poecilia latipinna (LeSueur, 182 1) - AMNH 51452 Rachovia sp. - AMNH 22477 (3/3) (6/82) Rivulus hartii (Boulenger, 1890) - AMNH 8354 Poecilia minor (Garman, 1895) - AMNH 35359 (2/2) (3/11) Skiffia bilineata (Bean, 1887) - UMMZ 189050 Poeciliopsisgracilis (Heckel, 1848) - AMNH 24462 (42/42) (3/20) Priapella intermedia Alvarez, 1952 - AMNH Other Atherinomorphs 78021 (5/40) Priapichthys dariensis (Meek and Hildebrand, Dermogenys sumatranus (Bleeker) - AMNH 9584 1913) - AMNH 28601 (3/28) (2/2) Scolichthys greenwayi Rosen, 1967 - AMNH Menidia beryllina (Cope) - AMNH 35721SW 32887 (5/108) (4/12) Tomeurus gracilis Eigenmann, 1909 - AMNH Menidia menidia (Moenkhaus) - AMNH 72910 (5/31) 55066SW (2/2) Xenodexia ctenolepis Hubbs, 1950-AMNH 24578 Oryzias latipes (Temminck and Schlegel) - AMNH (5/187) 57106SW (8/8) Xiphophorus alvarezi Rosen, 1960 - AMNH Parexocoetus brachypterus (Richardson) - AMNH 36800SW (4/148) 14133SW (1/2) Xiphophorus couchianus (Girard, 1859) - AMNH Strongylura timucu (Walbaum) - AMNH 27409 22644 (14/14) (1/1) Xiphophorus milleri Rosen, 1960 - AMNH 27486 (15/15) ACKNOWLEDGMENTS Other Cyprinodontiforms An earlier version of this manuscript was submitted to the City University of New Adinia xenica (Jordan and Gilbert, 1882) - AMNH 21836SW (2/2) York, Biology Program, as a Ph.D. disser- Aphanius ginaonis (Holly, 1929) - AMNH tation. The project began under the guidance 28677SW (2/2) and inspiration of the late Dr. Donn E. Ro- Aplocheilichthys johnstoni (Gunther, 1893) - sen, and I dedicate the final product to him. AMNH 41591 (20) I also thank my other committee members, Aplocheilus panchax (Hamilton-Buchanan, 1822) Drs. K. D. Kallman, R. T. Schuh, C. Simon, - AMNH 21957 (4/4) M. L. Smith, and especially G. Nelson, who Cubanichthys cubensis (Eigenmann, 1903) - ANSP later became my committee chairman, for 60286 (1/1) their encouragement and helpful comments. Cynolebias adloffi Ahl, 1922 - AMNH 22148SW Others whom I thank for discussion of this (1/1) work include Drs. L. R. and Cynolebias whitei Myers, 1942 - AMNH 22168 J. Atz, Parenti, (3/3) especially, D. J. Siebert. Drs. W. Fink, D. W. Cyprinodon variegatus Lacepede, 1803 - AMNH Greenfield, and C. Hubbs generously provid- 28100 (1/1) ed reviews or comments of the manuscript Fundulusgrandis Baird and Girard, 1854- AMNH for publication. For loans of materials and/ 21915 (1/1) or accommodation during visits to collec- 1989 RAUCHENBERGER: GAMBUSIA 13 tions, I thank D. W. Nelson (UMMZ), R. P. dian cartilage, serves as articulation point for Vari (USNM), and B. Chernoff (ANSP). the autopalatine. Two wings of each lateral The Biology Department of City College, ethmoid expand anterolaterally, although CUNY, has provided me with teaching sup- these two are sometimes not separated com- port throughout my graduate career, and pletely. CUNY granted me a University Fellowship The only elements corresponding to an in- one year, and a Block grant for OTS study. fraorbital series in Gambusia are the lacrimal The AMNH also has awarded me a half-year (infraorbital 1) and the dermosphenotic (in- predoctoral fellowship; the Hodgekiss Fund, fraorbital 6). The lacrimal is found lateral and of the Ichthyology Department of AMNH anterior to the lateral ethmoid. Sensory nerve provided support to attend meetings to pre- foramina are seen in its lateral surface; some- sent this work; and the Giles Meade Fund, times a distinct canal is present too. On its also of the Ichthyology Department of medial surface is a small transverse shelf, and AMNH, provided part of the publication posteromedially a small knob. The lateral funds for this manuscript. These forms of posterior border attaches to tissues covering support have been greatly appreciated. the eye. This border sometimes is broken into several sections. OSTEOLOGY The dermosphenotic is small and teardrop- shaped. It is found posterior to the postero- The following descriptive osteology is based lateral edge of the frontal, which ends in a on AMNH 78134, a mature male Gambusia notch, and anterior to a lateral projection of hispaniolae, standard length 32.0 mm, and the sphenotic. all illustrations in this section are ofthis spec- The skull is roofed by three pairs ofdermal imen except where noted. Comments on bones, the nasals, frontals, and parietals, and variability of characteristics refer to vari- the median supraoccipital. The nasals are ability within the species or among the species broad, flat, and roughly triangular. Nerve fo- within the genus; the groups delineated by ramina can be seen along a median crest. systematically important variabilities are The frontals are very large, covering more specifically noted. than half of the skull. They overlap irregularly at midline, until they are separated by CRANIUM (figs. 1, 2, 3) the supraoccipital. A bony tube, representing The large, cartilaginous mesethmoid block the sensory canal, runs in an anterior-pos- fills the front of the skull; a small disclike terior direction, dividing the anterior portion ossification can be seen within. The paired of the frontal into lateral (supraorbital) and lateral ethmoid bones are affixed to the lateral medial sections. This bony tube also has nerve edges ofthe cartilage block. Dorsally, the car- foramina piercing the frontal associated with tilage projects anteriorly under the nasals and it, just lateral to the tube on the supraorbital posteriorly along the underside ofthe roofing side. The lateral portion of the frontal is the bones (frontals) at midline. Ventrally, the car- dorsal border ofthe orbit; anterior to it is the tilage extends posteriorly on the dorsal bor- lateral ethmoid, posterior the dermosphen- der ofthe parasphenoid. Anteriorly, the ven- otic. A large notch in the posterior section of tral surface of the cartilage is covered by the the frontal accommodates the dermosphen- vomer. The edentulous vomer is triangular, otic. The bony tube and associated foramina and although broad, the ossification is quite are deflected from their straight path toward thin. the dermosphenotic. Although the tube does The lateral ethmoids are twisted, complex not continue onto the dermosphenotic, more bones. On each, the posterodorsal extension foramina are seen on this bone. meets the anterior edge of the supraorbital Parietals come and go among poeciliids; section of the frontal, curving underneath it. however, all species of Gambusia examined This section medially attaches to the mes- possess these thin, elongate bones. Inciden- ethmoid block. An anteromedial projection tally, Belonesox also has well-developed pa- of the lateral ethmoid, again along the me- rietals, although they are quite odd in form, 14 AMERICAN MUSEUM NOVITATES NO. 2951

A B Fig. 1. Cranium of Gambusia hispaniolae. A. Dorsal view. B. Ventral view. being dumbbell-shaped. Gambusia parietals sally, and the exoccipital ventrally and are found at the posterior edge ofthe frontals, posteriorly. The upper arm of the posttem- which overlie their anterior edge. Medially, poral contacts the epioccipital near the lateral the parietals extend further than the frontals, edge of the base of the epioccipital process. onto the supraoccipital. The pterotic (autopterotic; there is no der- The supraoccipital dorsally covers the pos- mal component) forms a lateral shelflike terior half of the skull. Its anterior edges are bulge. It is bordered anterolaterally by the underneath the frontals. Short processes of sphenotic, anteromedially by the parietal, the supraoccipital project laterally at the an- posteromedially by the epioccipital, and posterior edges of the parietals, again ventral to teroventrally by the exoccipital in dorsal view. the frontals. Two winglike supraoccipital In ventral view, most of its medial border processes, joined medially, project posterior- contacts the prootic. As a lateral shelf, the ly over the end ofthe skull. The posterolateral pterotic creates a shallow, open fossa above extension of the supraoccipital contacts the it and a deeper, also open, fossa below it. An epioccipital, and, although the supraoccipital elliptical autopterotic fossa, the thin flange of does not participate in the border of the fo- bone separating the course of the horizontal ramen magnum, a thin vertical plate extends semicircular canal, conspicuously traverses down from the supraoccipital processes to the pterotic also. Laterally, the posterior fork contact the exoccipitals at the top of the fo- of the hyomandibula contacts the pterotic. ramen magnum. The anterior arm ofthe hyomandibula touch- The epioccipitals also send off elongate es the sphenotic, which is wedged between posterior processes, paralleling the the pterotic, parietal, dermosphenotic, and supraoccipital processes but slightly below frontal dorsally; in ventral view, the sphen- them. The epioccipital bulges posterolater- otic is just anterior and lateral to the prootic. ally, then slopes back in to contact the pter- The exoccipitals form most of the border otic anteroventrally, the parietal anterodor- ofthe foramen magnum, excluded only from 1989 RAUCHENBERGER: GAMBUSIA 15

Soc

DOR

Fig. 2. Oblique dorsolateral posterior view of the cranium of G. hispaniolae. the ventral portion at the basioccipital condyle. Ventrally, the exoccipitals are medially bordered by the basioccipital, laterally by the -MAX pterotics, and anteriorly by the prootics. A large jugular foramen is visible. There are no exoccipital condyles; however, the neurapophyses of the first vertebra attach laterally to the exoccipitals at the edges of the foramen magnum (see separate discussion of verte- B brae). The basioccipital ends posteriorly in a cen- Fig. 4. Upperjaw of G. hispaniolae. A. Dorsal trumlike condyle, which serves as articular view. B. Ventral view. point for the first vertebra and, in ventral view, is seen to narrow to a point anteriorly. The basioccipital forms a ventral dome for terior edge of each prootic is riddled with the basicranium. Its anterior edge is covered small notches and bridges of bone. On each by a posterior extension ofthe parasphenoid. side, the prootic laterally contacts the sphen- The prootics are large domed bones, form- otic and pterotic, posteriorly the exoccipital, ing much of the floor of the basicranium. and posteromedially the basioccipital. Ventrally, they contact each other postero- The parasphenoid ventrally covers the me- medially, but are covered and anteriorly sep- dian part of the basioccipital and prootics. arated medially by the parasphenoid. The an- Just anterior to the prootics, two small lateral tabs project dorsally. Anterior to this point, the parasphenoid is a thin median bar that FR angles dorsally to contact the mesethmoid NAS i \S H block anterior to the lateral ethmoid, medial to the lacrimal, and just ventral to the nasals. This long, anterior extension of the para- DSP sphenoid has a thin vertical laminar flange. UPPER JAW (fig. 4) LAC Two paired bones comprise the upperjaw, the maxillae and premaxillae. These overlap in both their dorsal and lateral aspects. Dor- Fig. 3. Lateral view ofthe ethmoid and orbital sally the ascending processes of the premax- regions of G. hispaniolae. illae are fairly long, broad, and triangular. 16 AMERICAN MUSEUM NOVITATES NO. 2951

They parallel each other along the midline; swelling, the maxilla narrows again, until it in most species each ascending process ends crosses the ventral margin of the premaxilla. in a pointed tip. Some variation is seen, how- Here there is another widening, more pro- ever; the ascending processes may be round- nounced in some species than in others. The ed5 squared, or medially pointed. Rounded anteroventral tip is usually drawn out ven- ascending processes are seen in Brachyrha- trally to a point. The dorsal extension of the phis and Heterandria (Belonesox, because of maxilla also is said to characterize the su- its elongatejaws, is quite apomorphic for pre- perfamily Poecilioidea (Parenti, 1981). maxilla structure, and is not particularly use- Upper jaw teeth consist ofan outer row of ful as an outgroup for this character). large recurved teeth and several inner rows Capping the ascending process is the dorsal of smaller pointed teeth. They are found on process ofthe maxilla, which has both dorsal the premaxilla from the anterior edge back and ventral components. The dorsal com- along the alveolar process to just where the ponent is a thumb-shaped laminar sheet of maxilla laterally crosses the premaxilla. Tooth bone, anteromedially directed, that usually structure is constant for the genus and im- partly covers the ascending process. The ven- mediate outgroups, although a wide range of tral component runs below the ascending tooth morphologies is observed among poe- process, is much narrower than the dorsal ciliids. component, and usually more medially directed, approaching its fellow along the midline. The posteromedial junction of the dor- LOWER JAW (fig. 5) sal and ventral components forms a large The lower jaw is quite robust, consisting bump; just lateral to this bump is a depres- ofdentary, anguloarticular, retroarticular, and sion, andjust lateral to the depression, on the coronomeckelian (sesamoid articular) bones. posterior side ofthe maxilla, one head ofthe Two areas of Meckel's cartilage are promi- autopalatine articulates. No rostral cartilage nent; the long thin mentomeckelian cartilage, was seen in any poeciliid examined. surrounded by the dentary anteriorly and the The thumb-shaped dorsal process of the anguloarticular posteriorly; and the cone- maxilla is a characteristic of the superfamily shaped articular cartilage, mostly surrounded Poecilioidea. This group is part of a larger by the retroarticular but dorsally covered by group, Sept. 2 characterized in part by short the anguloarticular. The articular ossification and narrow ascending processes (see table 5, ofMeckel's cartilage is presumed to be fused from Parenti, 1981). Although some poecil- to the angular, as is general for clupeoceph- iids do have short and narrow ascending pro- alans (Nelson, 1973), hence the name an- cesses (e.g., Xiphophorus), all Gambusia guloarticular. Within the retroarticular, the species have ascending processes that are well cartilage is usually completely replaced with developed. ossified tissue, leaving a hollow cone in the In lateral view, the alveolar process of the bone. Sometimes, however, a cone of carti- premaxilla bears teeth back to the level where lage persists in the adult. the maxilla crosses it. Posteroventral to this The dentary bears teeth quite similar to the point, the premaxilla is roughly rectangular, upper jaw teeth: an outer row of large re- varying from this shape sometimes to form curved teeth and several irregularly spaced an S-shape (thought to be primitive for cy- inner rows of smaller teeth. Teeth continue prinodontoids; Parenti, 1 981) by drawing out posteriorly on the dentary to the beginning the posterodorsal corner posteriorly, the pos- of the coronoid process, which reaches posteroventral corner ventrally, and the antero- teriorly nearly to the posterior border of the ventral corner curving anterodorsally. premaxilla, which overlies it laterally. A The maxilla laterally covers the premax- prominent foramen is seen just below the illa. It varies from being a straight bar at two tooth row, and a sensory canal runs below points. First, where it crosses the end of the the level of Meckel's cartilage. A small pos- tooth-bearing portion ofthe premaxilla, there terior projection of the dentary carries the is an anterior swelling, usually rounded in sensory canal posteriorly, between the two profile but sometimes pointed. Ventral to this limbs of the anguloarticular. Below the sen- 1989199RAUCHENBERGER: GAMBUSIA 17

-AAR Fig. 6. Opercular apparatus of G. hispaniolae, lateral view.

mentomeckelian cartilage is surrounded by a RARo\7 cup of bone, an anteromedial extension of the anguloarticular. Just anterior to this cup is a small sesamoid bone, the coronomeck- MKC \@ DEN elian, that lies dorsally on the mentomeckelian cartilage. The retroarticular, as noted above, ossifies about the ventral portion ofthe articular car- B tilage. It is wedge-shaped, fitting into the posteroventral corner ofthe anguloarticular, and is Fig. 5. Lowerjaw of G. hispaniolae. A. Lateral moderately sized. view (anterior to left). B. Medial view (anterior to bottom right corner). OPERCULAR APPARATUS (fig. 6) The opercle is a large, fairly flat bone with sory canal, the dentary is angled medially. In no spines or serrations. It articulates with the a medial view (fig. 5A), it can be seen that lateral side of the hyomandibula, just below this ventral portion of the dentary forms a the posterior fork on the dorsal part of that fossa into which Meckel's cartilage neatly fits. bone. The articulation is effected by a raised In this view it can also be seen that the den- knob on the opercle. tary expands slightly at the anterior midline Lying ventral and medial to the opercle is symphysis. the subopercle, which projects beyond the The anguloarticular dorsally reaches as high limits ofthe opercle in both its posterior and as the coronoid process on the dentary. An- ventral aspects. It also is a flat, unserrated teriorly it is forked; the dorsal, larger limb bone. runs medial to the dentary, and it is this limb The dorsal limit of the preopercle is near that surrounds the mentomeckelian cartilage. the anterior fork of the hyomandibula. Its A smaller ventral limb also extends toward posterior margin parallels the anterior mar- the medial process of the dentary, but is not gin ofthe opercle, and then it swings forward, overlain by it. The ventral portion of the an- narrowing anteriorly. Its anterior extension guloarticular also angles medially, similar to fits into the medial face of the quadrate, just the dentary, continuing the fossa for Meckel's short of the quadrate-anguloarticular joint. cartilage. The posteroventral corner of the The preopercle also has an entire margin, and anguloarticular forms a fossa for the articu- carries a distinct sensory canal, running dor- lation with the condylar head ofthe quadrate. soventrally near the posterior margin of the The retroarticular is not involved in the joint bone, then turning at a right angle to run surface, but the dorsal edge of the articular anteriorly to the end of the preopercle. cartilage abuts the ventral margin ofthejoint. The interopercle is found closely appressed In medial view, the posterior end of the to the medial side ofthe preopercle, although 18 AMERICAN MUSEUM NOVITATES NO. 2951

goid closely appressed, ankylosed, or fused to the autopalatine (a line between them is APL. visible, but they are neither separable nor independently movable), or whether it rep- END E' resents a laminar perichondral ossification of ECT?END the autopalatine could not be determined. IH The endopterygoid (= mesopterygoid) be- ~~SYM gins at the posterior head ofthe autopalatine, PQC and runs along the posterior edge of the palatoquadrate cartilage, investing far into the A B medial face of the quadrate. A projection of the endopterygoid just ventral to Fig. 7. A. Jaw suspensorium ofG. hispaniolae, the auto- lateral view. B. Embryonic view of anterior half palatine heads is connected via ligamentous of jaw suspensornum. tissue to the lateral ethmoid. At its posteroventral edge, the endopterygoid is expanded, projecting past the quadrate to the symplec- it slightly exceeds the borders of the pre- tic. opercle in its posterior and ventral aspects. The quadrate has two sections, the dorsal Anteriorly, the interopercle narrows, crosses part that ossifies in the palatoquadrate car- the preopercle dorsomedially, and via liga- tilage, reaching toward the autopalatine, and ment is connected to the retroarticular. a ventral section that extends posteriorly under the symplectic. The anteroventral corner JAW SUSPENSORIUM (fig. 7) ofthe quadrate forms a condylar head, articulating with the anguloarticular. The anterior The palatoquadrate cartilage ofthe embryo ends of the symplectic and preopercle, both ossifies in two centers, the autopalatine dor- rodlike, terminate along the medial face of sally and the quadrate ventrally. The dorsal the ventral part of the quadrate. Ossification end of the autopalatine is hammerhead- ofthe palatoquadrate cartilage is never com- shaped. The large anterior projection of the plete; in adults, cartilage still can be seen be- hammerhead is cartilage-capped and articu- tween the autopalatine and the quadrate, as lates with the maxilla. The hammerhead is well as between the quadrate and the sym- somewhat twisted in the middle, so that the plectic. posterior head also faces slightly medially. It The hyosymplectic cartilage ossifies into articulates with a facet on the lateral ethmoid. two bones, the hyomandibula and the sym- This facet either lies against the lateral edge plectic; no dermal bones are associated with ofthe cartilaginous mesethmoid, or is a small this region of the suspensorium. The sym- lateral projection from the same position. plectic, as mentioned above, ends in a rod- There is no dermopalatine, and according like, cartilage-capped terminus along the me- to many authors there is no ectopterygoid dial face of the quadrate, near the articular (Ramaswami, 1945; Parenti, 198 1); how- condyle. Posterior to the quadrate and en- ever, there is a laminar extension of bone dopterygoid, a thin dorsal flange is developed along the anterior edge ofthe palatoquadrate on the symplectic, narrowing again as it ap- arch, running from the dorsal tip of the arch proaches the cartilage-capped junction with ventrally to cover most ofthe anterior surface the hyomandibula. Also articulating at this ofthe autopalatine and quadrate. This is con- hyosymplectic juncture is the interhyal. sidered to be an extension ofthe autopalatine At its ventral edge, the hyomandibula also by Parenti (1981), a characteristic of cyprin- bears a cartilage-capped head to articulate odontoids. This anterior flange of bone, with the symplectic and interhyal. Dorsally, however, does not preform in cartilage (de- the hyomandibula is forked, again with two termined by examination of cleared and cartilage-capped heads. The anterior head ar- stained embryos); its development mirrors ticulates with the sphenotic, the posterior with that of the endopterygoid; however, whether the pterotic. A facet to receive the operculum this flange of bone represents an ectoptery- is ventral to the posterior fork. 1 989 RAUCHENBERGER: GAMBUSIA 19

VHH A

BSR Fig. 8. Lateral view of hyoid arch of G. hispaniolae.

HYOID ARCH (figs. 8, 9) The small ossified interhyal is connected via cartilage tips to the hyomandibula dor- B t====3ZZII sally and to the posterior end of the ceratohyal ventrally. The large ceratohyal can be visualized in two parts, anterior and posterior. The posterior section (posterior ceratohyal; posterohyal of Nelson, 1969; epihyal ofmany authors, including Rosen, 1964, and C = Weitzman, 1962) serves as the attachment site for the last ofthe branchiostegal rays along Fig. 9. A. Anterior part of the hyoid arch of its lower lateral surface. It is separated from G. hispaniolae, ventral view. B. Urohyal of G. the anterior section (anterohyal of Nelson, hispaniolae, dorsal view. C. Urohyal of G. his- 1969; ceratohyal ofmany authors) by a deep, paniolae, lateral view. cartilaginous intrusion. Dorsally, however, the anterior and posterior sections ofthe cer- ceratohyal. An ossification around the ante- atohyal are firmly joined by a bridge ofbone. rior, anteroventral, anteromedial, and an- The anterior section of the ceratohyal is dis- terolateral aspects of this cartilage is inter- cussed in two sections. Its posterior half is preted as the ventral hypohyal. Among wide, and bears the branchiostegal rays along Gambusia species, the extent of the ventral its lateral surface. The third of these rays ac- hypohyal ossification is variable. Some tually articulates sometimes on the cartilage species show the ventral hypohyal ossifica- that separates the anterior and posterior section covering most of the ventral surface, tions of the ceratohyal.The anterior half of meeting the anterior ceratohyal in a relatively the anterior section ofthe ceratohyal narrows straight line; fishes of the subgenera Hetero- sharply, then widens to form a cup around phallina and Arthrophallus show this condi- the cartilage of the hypohyal. Two slender tion. In other species, it is the anterior cer- branchiostegals articulate medially along this atohyal that covers most ofthe ventral surface narrowed section; sometimes they merely ap- of the hypohyal cartilage, projecting an an- proach the anterior ceratohyal at its ventral teriorly directed ventrolateral lobe and a margin, articulating neither medially nor lat- smaller anteromedially directed lobe. This erally. bilobed head of the anterior ceratohyal is a According to Parenti (1981), cyprinodon- more conventional arrangement, reminiscent toids have no dorsal hypohyal, and no ossi- of the situation in many other fishes, where fied dorsal hypohyal was found in poeciliid the anteromedially directed lobe contacts the materials examined; however, the hypohyal dorsal hypohyal. However, these fishes have region is occupied by a large block ofcartilage no dorsal hypohyal. The dorsal aspect ofthis that is posteriorly delimited by the anterior region reveals a large expanse ofcartilage with 20 AMERICAN MUSEUM NOVITATES NO. 2951 only the anteromedial portion ossified as the The basihyal is a long, triangular bone, medial section ofthe ventral hypohyal. When comparable in length to the sum of the rest the anterior ceratohyal extends forward thus, of the basibranchial series. Its narrow pos- the ventral hypohyal is sometimes reduced terior end articulates with basibranchial 2, to a triangular shape, not covering the an- and it is ossified for only one-third to one- terolateral corner of the hypohyal cartilage. half of its length, the broader anterior end This condition typifies the nicaraguensis remaining cartilaginous. species group. Other members of the subge- The branchiostegal rays, as alluded to ear- nus Gambusia have the ventral hypohyal lier, are consistently six in number, with two covering the anterolateral corner of the car- thin anterior rays that attach to (or approach) tilage. the narrow part of the anterior ceratohyal, A group containing poeciliids, procato- and four larger rays attached to the lateral pines, Fluviphylax, and Pantanodon is de- face of the wide part of the ceratohyal. Al- scribed by Parenti (1981) as having the ven- though the dorsal heads ofthese four rays are tral hypohyal form a cap of bone over the roughly identical, the rays expand in width end of the anterior ceratohyal (see her fig. ventral to those heads, appearing to have an 28C, p. 399). This condition is found in To- anterior bulge. This is slight in the first of meurus; however, Gambusia and other poe- these rays and most pronounced in the more ciliids examined possess cartilage where the posterior rays. dorsal hypohyal would be, so that the ossi- fications appear more similar to her figure 28B, the condition found in most cyprino- DORSAL GILL ARCHES (fig. 10) dontoids. There are no suprapharyngobranchials, as In ventral view, the ventral hypohyals face is true of most modern teleosts. There also each other anteromedially in distinct cup- is no first infrapharyngobranchial; its absence shaped condyles. These are connected to each is the general condition for Division II ath- other by tough connective tissue in which a erinomorphs-Beloniformes and Cyprino- median ovoid cartilage or ossified bone dontiformes (Rosen and Parenti, 1981). In- sometimes appears. This is most probably an frapharyngobranchials 2 and 3 are ossified adventitious structure, as its appearance is and have toothplates fused to them. There is not constant within species, and it is seen on no fourth infrapharyngobranchial in ather- occasion throughout cyprinodontiforms. inomorphs, but there is a fourth infrapha- The anterior end of the urohyal lies be- ryngobranchial toothplate. tween the posterior ends of the two hypo- Infrapharyngobranchial 2, with its associ- hyals, and is somewhat expanded relative to ated toothplate, is rather small, partially its posterior shaft. The anterior head varies overlain by infrapharyngobranchial 3. It has from being broadly triangular to more com- two cartilage-tipped articular heads, one on monly being slightly cleft, with a median key- the lateral side that is approached by epi- hole and two slightly angled heads, each lig- branchial 2, and an usually larger one on the amentously connected to a ventral hypohyal. anterolateral corner of infrapharyngobran- Posteriorly, the urohyal extends to the level chial 2, which is approached by a similar pro- of basibranchial 4. Thin laminar plates ex- cess on infrapharyngobranchial 3. The re- tend slightly in a lateral plane. curved teeth ofthe infrapharyngobranchial 2 Laterally the urohyal is seen as a broad toothplate face slightly more anteriad than median plate, wider posteriorly. Anteriorly is those ofthe infrapharyngobranchial 3 tooth- a posterodorsally directed spike, with a small plate, which are more ventrally directed. knob that approaches the anterior end ofba- The large infrapharyngobranchial 3's and sibranchial 2, and a usually spatulate exten- their toothplates are the major elements in sion posterior to that knob that runs under the dorsal gill arches. Infrapharyngobranchi- basibranchial 2 and is approached by the anal 3 has three cartilage-tipped articular heads. terior forks of the first hypobranchials. In Anteriorly, one approaches its counterpart on some Gambusia species, this extension is bi- infrapharyngobranchial 2. Near the postero- fid, each side associated with its respective lateral edge of infrapharyngobranchial 3, a hypobranchial 1 anterior fork. small articular head contacts epibranchial 3. 1 989 RAUCHENBERGER: GAMBUSIA 21

Finally, the posterior edge of infrapharyngobranchial 3 is dominated by a broad articular head that receives the equally broad epibranchial 4. On the infrapharyngobranchial 3 toothplate, the large teeth curve medially. IPB3 In Gambusia, but not in immediate outgroups, these teeth are bicuspid or serrate along their inner edge. The medial teeth in members ofthe subgenus Heterophallina are exaggeratedly deformed from a simple curve cone; they have a serrate bump or pad along the medial side of the tooth. The infrapharyngobranchial 4 toothplate is fairly small, beginning at the posterior edge of infrapharyngobranchial 3 and extending just slightly beyond epibranchial 4. Its teeth are smaller than those associated with infrapharyngobranchial 3. EB4 There are four epibranchials. Generally in .4V cyprinodontiforms, epibranchial 1 has no Fig. 10. Dorsal gill arches of G. hispaniolae, distinct uncinate process (Parenti, 1981). dorsal view, right side only. Rather, the base ofthis element is broadened, and its posterior edge serves as articular point process of infrapharyngobranchial 2. In cy- for the interarcual cartilage. This type ofepi- prinodontoids, the interarcual cartilage is re- branchial 1 is seen in outgroups studied and duced in size, and rarely approaches infra- many Gambusia species as well. However, pharyngobranchial 2. In Gambusia, the members ofthe subgenus Heterophallina and maximum length ofinterarcual cartilage seen of the nobilis species group of the subgenus is roughly one-third ofthe length of epibran- Arthrophallus regularly have uncinate pro- chial 1. Often it is reduced to a small ball of cesses on their first epibranchials. Other cartilage, and sometimes it is absent alto- members ofthe subgenus Arthrophallus usu- gether. ally do not have such an uncinate process. Epibranchial 2 is a slender, slightly twisted And, more confusingly, whereas most mem- bone that medially approaches infrapharyn- bers of the subgenus Gambusia do not typi- gobranchial 2. It never has an uncinate pro- cally have these uncinate processes, they oc- cess in Gambusia; the generality of that lack casionally do, even varying from the left and is unclear, as it is absent in cyprinodonti- right sides of the same fish. A distinct unci- forms and some other atherinomorphs, but nate process on epibranchial 1, leading to the not in beloniforms. interarcual cartilage, can be considered a de- Epibranchial 3 is similarly slender, but it rived condition at this level, but the character does bear an uncinate process, fairly close to is variable in the subgenus Gambusia. The its base. (In fact, this bone looks quite similar orientation of epibranchial 1 is interesting to epibranchial 1 when it bears an uncinate also. It does not run parallel to the other epi- process.) The medial tip of epibranchial 3 branchials, but points almost directly ante- contacts a small cartilage-tipped process on riorly. This condition is common to many infrapharyngobranchial 3. The uncinate pro- cyprinodontiforms. cess, directed posterodorsally, approaches the The interarcual cartilage, when present, uncinate process of epibranchial 4. arises near the posteromedial edge of epi- Epibranchial 4 is much larger than the oth- branchial 1 when epibranchial 1 has no un- er epibranchials, forming the major support cinate process, or near the uncinate process for the dorsal gill arch elements, as is the of epibranchial 1 when present. The general general condition for atherinomorphs (Rosen acanthomorph condition is for the interar- and Parenti, 1981). Its broad medial head cual cartilage to run from the uncinate pro- contacts infrapharyngobranchial 3 and the cess of epibranchial 1 to the cartilage-tipped toothplate of infrapharyngobranchial 4. The 22 AMERICAN MUSEUM NOVITATES NO. 2951

dorsally, but as there is no hypobranchial 4, ventrally it approaches the cartilaginous fourth basibranchial. The medial head ofcer- atobranchial 4 is cartilage-tipped. Just distal to this is an anterolaterally directed flange. In all species of Gambusia examined as well as Belonesox, Brachyrhaphis, and some other poeciliids, there are teeth on ceratobranchial 4, most heavily concentrated near the medial BH head but extending along one-third of the bone's length. These teeth have been hypothesized to represent fused gill rakers (Gos- line, 1963; Nelson, 1969). This a character HBI ofspotty distribution within poeciliids, within cyprinodontiforms, and within atherinomorphs. Ceratobranchial 5 is a much larger bone, bearing only teeth, no gill rakers. Anteriorly, the cartilage-tipped head approaches the medial head of ceratobranchial 4. The two ceratobranchial 5's are triangular-shaped and, although they are separate bones, they are close together along the midline, until each curves dorsolaterally toward its respective side. Ceratobranchial 5 is covered with small to moderately sized conical teeth, and larger, somewhat recurved teeth are found along the inner curve of the bone. Posteriorly, ceratobranchial 5 tapers to a cartilage-tipped point that approaches epibranchial 4. On the medial side ofthis point is a posteriorly directed process, shaped like a shark's tail. Ventrally, each ceratobranchial 5 has a large median flange, extending directly ventrad, with a hook at the anterior end. Fig. 11. Ventral gill arches of G. hispaniolae, There are three pairs of hypobranchials. dorsal view. There is a good deal ofvariability in the form of hypobranchial 1. It can be described as having two parts: a medial uncinate process arises from the anterior side triangular part of the bone, pointing dorsolaterally toward that points laterally so that the broad flat base of the articulates with a narrow the uncinate process of 3. Lat- triangle part epibranchial ofbasibranchial and to eral to the uncinate process, 2, lateral this triangle, the bone extends a square or diamond-shaped base. Such a a thin then to contact cer- flange, tapers the form of hypobranchial 1 is found in many atobranchials. species of Brachyrhaphis and Heterandria. Other species ofthese genera, as well as most VENTRAL GILL ARCHES (fig. 11) species of Gambusia, have hypobranchial 1 Ceratobranchials 1, 2, and 3 are slender, modified in that the triangular part is me- rodlike elements running from their respec- sially cleft, resulting in a forked hypobran- tive hypobranchials ventrally to their respec- chial 1 with two medial articular heads. The tive epibranchials dorsally. They are heavily posterior, usually wider, head articulates with adorned with gill rakers. the constricted part of basibranchial 2. The Ceratobranchial 4 contacts epibranchial 4 anterior head is angled somewhat ventrally, 1 989 RAUCHENBERGER: GAMBUSIA 23 and approaches the dorsal, ventrally directed Is flange of the urohyal. Variation occurs with respect to the depth of the cleft, the degree of separation of the two heads, and the relative sizes ofthe two heads, compared to each other. In addition, there is variation in the lateral part of hypobranchial 1. Some species of Gambusia (particularly those in the subgenus 'EPR Gambusia) have an anteriorly directed pro- Pi cess along the lateral edge of hypobranchial 1. In other species, and outgroups examined, Fig. 12. Lateral view of the anterior vertebrae this process is indistinct or absent. of G. hispaniolae. Only the first pleural and epi- Forked first hypobranchials are not unique pleural ribs are shown. to Gambusia or its close relatives. Some, but not all, poeciliids have them, and I have no- branchial 4, but is only cartilaginous in this ticed the character in some other cyprino- region. dontoids (Aphanius, Cubanichthys; Orestias, A cartilaginous basibranchial 4 occupies a in Parenti, 1984), some aplocheloids (Ra- central position to the third hypobranchials chovia, Rivulus), and atherinoids such as Ath- anteriorly and the fourth ceratobranchials erinella (Chemoff, 1986); however, the forked posteriorly. Its shape is variable among hypobranchial 1 with a distinct lateral pro- species, ranging from diamond to hexagon to cess seems to be unique to Gambusia. octagon to spade. Dense, blue-staining con- Hypobranchial 2 is similarly composed of nective tissue often connects the two third a lateral box and a medial triangle, but these 4, and the two shapes are less distinct, and the element is hypobranchials, basibranchial neither forked nor does it have a lateral pro- fourth ceratobranchials into one unit. cess. Hypobranchial 2 medially articulates with a constricted part of basibranchial 3. VERTEBRAE (figs. 12, 13, 17A, 18) Hypobranchial 3 is closer to being a simple Total vertebral number in the genus ranges rectangle in shape. Medially it articulates with from 29 to 33, usually 30 or 31, with 12 to a second narrowing in basibranchial 3 at the 14 precaudal vertebrae and 15 to 19 caudal end of its ossified portion. All three pairs of vertebrae (not counting the terminal half- hypobranchials primarily stain blue, al- centrum of the caudal skeleton). though clear red ossified portions are seen The centrum ofthe first vertebra fits neatly along the lateral and anterior edges. against the centrumlike basioccipital con- Cyprinodontoids have two ossified basi- dyle; like other poeciliids, Gambusia has no branchials. Parenti (1981) hypothesized that exoccipital condyles. The neurapophyses of this represents a loss ofbasibranchial 1, rath- the first vertebra are applied to the lateral er than a fusion event. Therefore, basibran- sides of the skull near the centrum. As the chial 2 anteriorly contacts the basihyal, and neural arch continues dorsally, the neur- is posteriorly flared where it contacts basi- apophyses ofeach side retreat from the skull, branchial 3. Similarly, the anterior head of meeting at midline. Once joined, the neural basibranchial 3 is flared, then the bone nar- arch is now directed sharply forward, point- rows at the spot where the second hypobran- ing anterodorsally, to again contact the skull. chials articulate with it. Just beyond that This projection ofthe first neural arch lies on point, the bone flares out laterally, forming top ofthe exoccipitals at the dorsal border of wings. These wings are a characteristic ofthe the foramen magnum. This intriguing form subgenus Gambusia; other subgenera in the of the first vertebra is found in all members genus have a relatively straight basibranchial ofthe tribe Gambusiini and some ofthe Het- 3. Posterior to its articulation with hypo- erandriini; however, other groups of poecil- branchial 3, basibranchial 3 extends further iids have a completely different form of the back, ventral to basibranchial 4 and cerato- first vertebra. 24 AMERICAN MUSEUM NOVITATES NO. 2951

vertebrae are modified for gonopodial support. In females, these vertebrae also are un- conventional; they are discussed in conjunc- tion with the female anal fin. The form ofthe caudal vertebrae posterior to the anal fin can be seen in figure 13. The neural arches and spines are generally as described for precaudal vertebrae, although the neural prezygapophyses usually do not extend far enough to meet in midline; rather, they approach the postzygapophyses of the preceding vertebra. The hemal arch and spine mirror the neural side, although with smaller zygapophyseal processes. A hemal canal, beginning at the first caudal vertebra (not before) runs through the hemal arches.

CAUDAL FIN (fig. 13) Fig. 13. Caudal skeleton of G. hispaniolae, lat- The terminal half-centrum, representing the eral view. Only the right halves of the fin rays are shown. first preural centrum plus the first two ural centra, is fused to the single hypural plate, which is a fusion of hypurals 1-5. The half- The second, third, and fourth vertebrae centrum curves dorsally, and in some small have neural spines that are greatly expanded specimens it extends a short way along the in the anterior-posterior plane. The par- dorsal edge ofthe hypural plate. There seems apophyses are found rather high on these an- to be some neural arch material (presumably terior vertebrae, although the parapophyses of the first ural vertebra) incorporated here gradually move to a more ventral position in at the posterodorsal edge ofthe half-centrum subsequent vertebrae. A pleural rib articu- in embryos; this suggests possible homology lates along the posterior edge of each with the urostyle of many other fishes (also parapophysis. The first pleural rib is borne noted by Hollister, 1940). However, in most upon the second vertebra, a distinctive fea- fully formed adults, the "urostyle" material ture ofcyprinodontiforms (Rosen, 1964; Par- is completely incorporated into the hypural enti, 1981). Epipleural ribs, extending pos- plate, and the half-centrum ends abruptly, terolaterally from each pleural rib, also begin with a small hook projecting posterolaterally here, and are found through to about the ninth on each side. Anteriorly on the half-centrum, or tenth vertebra. neural arch material (presumably of the first The neural arch of the fifth vertebra is preural vertebra) forms a dorsal projection sometimes slightly expanded, but usually it on each side; these usually meet at the mid- and always the sixth vertebra assume a more line. No extension into a neural spine is seen. conventional form, seen throughout the pre- All hypurals arise from the half-centrum. caudal vertebrae. The neural prezygapophy- Varying degrees offusion ofthe hypurals are ses rise anterodorsally and meet in midline, seen in the genus; however, within any par- forming a complete arch. The elongate neu- ticular species, the pattern observed is fairly ral spine, angled posteriorly, forms when constant. Belonesox belizanus has two sym- neural arch material from each side of the metrical hypural plates, the lower plate pre- centrum extends dorsally to midline, then sumed to represent hypurals 1 and 2, the up- lengthens as a single unit. A small neural per hypurals 3, 4, and 5. The medial edges postzygapophysis usually is present also. of the two plates are straight and parallel. A few anterior caudal vertebrae, i.e., those Gambusia luma exhibits a pattern common vertebrae bearing hemal arches (and spines, in the genus, where there are two hypural here), also bear pleural ribs. In males, these plates, as in Belonesox, but their medial edges 1989 RAUCHENBERGER: GAMBUSIA 25 are not straight. Instead, there is an indented Preural vertebra 2 bears a full neural arch notch, starting at the half-centrum and ex- and spine, as well as a full hemal arch and tending about two-thirds of the length of the spine. The neural and hemal spines of the hypural plate. These symmetrical notches second and third preural vertebrae have thin look like a keyhole. The two plates are sep- laminar extensions, as do the epural and par- arate distal to the keyhole. This pattern is hypural. typical of most of the subgenus Gambusia, Principal caudal fin rays are borne on the but appears occasionally elsewhere in the ge- hypural plate, the epural, the parhypural, and nus. Gambusia hispaniolae shows the other the neural and hemal spines of the second, popular pattern for Gambusia. Here, the key- third, and often fourth preural vertebrae. The hole notch is present, but the hypurals are neural and hemal spines of the fourth and fused distal to the keyhole, forming a single fifth preural vertebrae also are involved in hypural plate with a hole in it. Sometimes the support of procurrent fin rays. Lying be- there is a small notch in the posterior edge tween the distal tips of the hemal spines of ofthe plate, directly distal to the keyhole. The the sixth and fifth, fifth and fourth, and fourth size of the keyhole also varies. The final pat- and third preural vertebrae are small blocks tern seen occasionally in Gambusia has no of accessory cartilages that also are involved keyhole at all; instead a solid hypural fan is in fin-ray support. Three blocks of accessory fused to the half-centrum. This is found only cartilage also are found between the distal tips in several derived members of the senilis of the neural spines of the same vertebrae. species group. Although these varying degrees of fusion might easily be envisioned as a transforma- DORSAL FIN (fig. 14) tion series, the polarity of such a series is not The number ofrays in the dorsal fin ranges clear. All four states described above are found from 7 to 10 in the genus. In a young spec- throughout the Poeciliidae, and though Bel- imen, each fin ray is supported by a long onesox has the unfused state, species of proximal radial and a small round distal ra- Brachyrhaphis exhibit all states of fusion. dial. In the adult fish, a medial radial has Ontogeny offers some help, as embryos gen- separated off from the distal end of each of erally have two separate hypural plates. The the proximal radials except the first. The shape of these two early plates (e.g., see fig. proximal and medial radials often remain 44D) is common to most species of Gam- connected by intervening cartilage. busia, but does not represent any ofthe adult The proximal radials extend between the stages. No species has been observed to pass neural spines, sometimes abutting their dor- through several different stages to reach the sal tips. When this happens, the dorsal part adult state; instead, the adult state is formed ofthe neural spine often becomes cup-shaped, directly from the embryonic state. sometimes with a median hole in the cup. The parhypural articulates with the ter- The first proximal radial has an enlarged minal half-centrum, but is not fused with it. anterior keel. Subsequent proximal radials It ends posterodorsally with the hypurapoph- have thin laminar flanges also, but not as ysis. extensive as the first. The single epural dorsally mirrors the po- There is an extra element at the posterior sition ofthe parhypural; however, it does not end of the dorsal fin (the "end piece" of articulate with the terminal half-centrum. The Weitzman, 1962). Rarely does it completely form and development ofthis epural suggest ossify. Its proximal end is fingerlike, mim- that it represents a detached neural spine, icking a small proximal radial. It is found presumably of the first preural vertebra, se- close to the medial radial of the last dorsal rially homologous to the neural spines (not fin ray; sometimes it is inseparable from it. arches) ofpreceding centra. A similar relation Its serial homology is unknown. between the parhypural and the preceding Gambusia, and all other cyprinodontoids, hemal spines is also suggested, although the have no rudimentary rays anterior to the first fate of the hemal arch material for the first dorsal fin ray; i.e., anterior to the ray that is preural vertebra is unclear. associated with the first proximal and distal 26 AMERICAN MUSEUM NOVITATES NO. 2951

PRR ' I Fig. 14. Dorsal fin skeleton of G. hispaniolae, lateral view. Only the proximal ends ofthe right halves of the fin rays are shown. radials. Parenti (1981) calls this character with the scapula it forms most of the border "loss offirst dorsal fin ray"; strictly speaking, ofthe scapular foramen. The dorsal extension the character should be "loss of ray(s) ante- ofthis face meets the dorsal part ofthe medial rior to the first dorsal fin ray"; although the wing, the supracleithum, and the posttem- groups described by these two characters are poral. Posteriorly, the dorsal edge of the me- the same. dial face forms a thumb-shaped projection. PECTORAL GIRDLE AND FiN (fig. 15) The posttemporal is forked. The longer dorsal arm contacts the epioccipital ante- A riorly; the lower arm, that points anteroventrally, is about half the length of the upper arm, and does not contact the skull. The lower arm is often reduced or fragmented, but always present in this genus. The ventral base of the posttemporal laterally covers the dorsal part of the supracleithrum. *RAD The small supracleithrum is shaped like an upside-down teardrop, and is sandwiched between the posttemporal laterally and the cleithrum medially. The cleithrum consists of an anterior face, a lateral wing, a medial wing, and a medial face. The anterior face is roughly diamond- shaped. The lateral and medial edges of the B anterior face are raised and thickened, and the ventral extension of this face meets its counterpart at the midline. The lateral wing begins dorsally at the top ofthe anterior face, and extends laterally almost to the scapular foramen. It narrows abruptly ventral to the scapula. The medial wing mirrors the lateral wing but extends further-dorsally to the posttemporal and supracleithrum, and ven- Fig. 15. Pectoral girdle and fin of G. hispanio- trally to the coracoid. The medial face is firm- lae. A. Lateral view, fin removed. B. Pectoral fin, ly affixed to the medial side of the scapula lateral view. Modified extension on rays 3-5 are and upper part of the coracoid, and together strippled darkly. 1989 RAUCHENBERGER: GAMBUSIA 27

This is similar in shape and position to the opodium during copulation (Hubbs and "scapular process" of Cyprinodon, although Reynolds, 1957). The second, third, fourth, not as large. Similar projections are seen in and fifth fin rays, particularly, are elongated, Profundulus, some aplocheiloids, and most and their distal third curves upwards. This poeciliids. This cleithral projection is partic- curve becomes quite pronounced in some ularly elongate in the rachowi species group, species, most spectacularly in G. heterochir. of the subgenus Heterophallina. There is a bit of variation in the degree of The scapula is anteriorly affixed to the lat- modification along these rays, but the basic eral edge ofthe medial face ofthe cleithrum. pattern is for each segment in the middle third It is a roughly triangular flat bone, its most of the fin ray to thicken and produce a lam- acute angle posteriad. The scapular foramen inar dorsal extension. Sometimes the exten- is oval, with its long axis running anterior- sion of each segment is kept separate, giving posterior. When the pectoral girdle is fully the fin ray a scalloped appearance; other times, formed, the scapular foramen is confined to these extensions, and sometimes the seg- the anteror third of the scapula; however, in ments themselves, anastomose. Turner juveniles, the foramen more closely resem- (1 942c) described the morphogenesis of the bles a cleft, running posteriorly almost to the pectoral fin modification in the male ofGam- border ofthe scapula. The medial face ofthe busia affinis, as well as partial induction of cleithrum also participates in the border of the character in females when treated with the scapular foramen; this usually signals the androgenic hormones. This pectoral fin mod- posterior limit of the cleithrum. ification is seen in all species of Gambusia, Ventrally, the scapula is bordered by the including G. vittata, although its develop- coracoid. Its major flat surface is comma- ment is very weak in G. luma. It is not seen shaped, running from the scapula to meet the in any other poeciliid. According to Turner ventral tip ofthe cleithrum. A second surface, (1942c), the final length ofthe elongated pec- projecting laterally, runs from the anteroven- toral fin rays 2, 3, 4, and 5 is dependent at tral edge, about three-quarters of the way up least in part on the length of time that male along the coracoid. reproductive hormones are maintained at Four radials fit neatly into a recess in the proper concentrations. This, in turn, depends posterior edge of the borders of the scapula on the size at which the fish matures, more and coracoid. The dorsal three radials are rapidly at a small size or more slowly at a roughly cube-shaped; the fourth is somewhat larger size. larger than the others and more irregularly shaped. An elongate postcleithrum arises along the PELVIC GIRDLE AND FIN (fig. 16) medial side ofthe coracoid, just anterior and The two pelvic bones lie parallel, with their ventral to the radials. Parenti (1981) labeled posteromedial extensions overlapping. In this as postcleithrum 3, following Weitzman ventral view, the long anterior extension of (1962). It tapers posteroventrally, often abut- the bone has a bony tube in its center. A ting the first rib. There is no first postclei- vertical flange projects ventrally from this thrum in poeciliids, although most other cy- tube on the anterior third ofthe bone. At the prinodonts have this scale-shaped bone just posterior terminus of the tube is a tablike posterior to the thumb-shaped process ofthe projection, near which the ventral halfof the cleithrum. No cyprinodont has a bone in the lateralmost pelvic ray articulates. The dorsal position ofWeitzman's (1962) postcleithrum halfofthat ray articulates near a notch in the 2, and, as is true for all neoteleosts, there is posterolateral border of the bone. The other no mesocoracoid. five pelvic fin rays articulate along the pos- The pectoral fin of females is quite con- terior border of the bone; this area is angled ventional, with 12 to 14 pectoral fin rays, the slightly dorsally. No radials were observed, top three and bottom three unbranched. In nor was a significant ischial process noted. males, however, the second, third, fourth, In lateral view, the posterodorsal portion fifth, and sometimes sixth pectoral fin rays of the bone rises in a dorsal process. The are modified, presumably to support the gon- shape of this process is variable, sometimes 28 AMERICAN MUSEUM NOVITATES NO. 2951

angled anteriorly or posteriorly, sometimes flat or squared. The aforementioned tab also is evident in this view, at the posteroventral corner.

FEMALE ANAL FIN (fig. 17) The proximal radials of the first five anal- fin rays are anterior to and lean on the hemal arch and spine ofthe 14th vertebra. This vertebra also bears a parapophysis and pleural rib. The next caudal vertebra also bears a parapophysis; sometimes a small rib articulates with it too. Parenti (1981) noted that many poeciliids and some of their relatives show this condition. There are 11 anal-fin rays. Except for the first and last, each is supported by proximal, medial, and distal radials. The proximal and distal radials are always joined by a large ball of cartilage; the medial radial never completely separates off. The ventral ends of the proximal radials often develop small anterior and posterior wings to grab the cartilage. The distal radial is a small ball of cartilage that has not been seen ossified. It is at least partly obscured by the anal-fin ray, as it is held between the anterodorsal tips ofthe two halves of the fin ray. Fig. 16. Pelvic girdle of G. hispaniolae. A. The first anal-fin ray lacks any ossification Ventral view. Only the outer two fin rays are shown. ofa medial radial, as does the last. The prox- B. Lateral view. imal radial ofthe last anal-fin ray is reduced,

A B Fig. 17. Anal fin skeleton of a female G. hispaniolae. A. Lateral view of hemal spines and proximal anal radials. B. Closer lateral view of proximal, medial, and distal anal radials. Right side only shown for proximal radials and fin rays. 1989 RAUCHENBERGER: GAMBUSIA 29 irregular in shape, and often not completely ossified.

GONOPODIAL SUPPORT (fig. 18) The skeletal support of the gonopodium involves modification of two types of bones. The hemal arches and spines of the 14th, 15th, and 16th vertebrae are changed into gonapophyses, and the anterior proximal ra- GPI dials of the anal fin are modified into the gonactinostal complex. GP2 The first gonapophysis projects anteroventrally from the anteroventral edge ofthe centrum of the 14th vertebra at approximately a 450 angle. Near the base of the first gonapophysis, a small parapophysis extends backwards, almost always bearing a small rib. The first gonapophysis extends anteriorly the length ofthree centra, where it meets the gon- GNA 2+3+4 actinostal complex. This gonapophysis is Fig. 18. Gonopodial suspensorium of G. his- sometimes thickened along its posteroventral paniolae, lateral view. edge, and sometimes bowed. The second gonapophysis extends forward Kallman (1959), in a series of experiments, just behind the first. It often has a small par- investigated the possibility of an inductive apophysis, but usually does not bear a rib. In relationship between the gonactinosts and Gambusia luma, the second gonapophysis is gonapophyses. identical to the first, but in all other species, The gonactinostal complex is of constant the second gonapophysis develops posterior- form in the genus and in its sister taxon, Bel- ly directed uncini on each side. (The lack of onesox. The first proximal anal radial is not such an uncinate process on the first gon- involved and rather reduced. The second, apophysis is a synapomorphy of the genus.) third, and fourth proximal anal radials grow The ventral outline of the second gonapoph- together and elongate, forming a column. The ysis, from the anterior extension to the end column has a small anterior wing, and ex- ofthe uncini, varies from being softly round- pands dorsally to form a cup, open at the top. ed to almost horizontal. In some taxa, par- The fifth proximal anal radial also is elongate ticularly in G. hispaniolae, the posterior ends but separate. It lies along the posterior edge of the uncini are slightly forked. of the column, and projects into the dorsal Some Gambusia species usually have only cup. In other poeciliids, the fifth proximal these two gonapophyses; this condition unites anal radial is incorporated into the gonacti- the nobilis and senilis species groups. Most, nostal complex and does not stand freely. however, develop a third, again with uncini, From the dorsal tip of the fifth radial, a although the uncini ofthe third gonapophysis ligament connects the gonactinostal complex are usually not as long as those ofthe second. to the ventral side of the tenth vertebra (the The anterior projection of the third gon- fourth vertebra anterior to the first gonapo- apophysis is bowed slightly dorsally, whereas physis). In this ligament is a small bone called the first and second gonapophyses are bowed the ligastyle, that varies from being a long slightly ventrally. The third gonapophysis thin splint, to a teardrop, to a round ball, or does not generally bear parapophyses. it may be absent altogether. Turner (1942b) described the formation of Turner (1942b) has shown that the 13th the gonapophyses in development, and the vertebra is, in fact, the first caudal vertebra, partial induction ofgonapophyses in females upon which a hemal arch and spine form. following hormone treatment. Rosen and The hemal arch is subsequently histolyzed, 30 AMERICAN MUSEUM NOVITATES NO. 2951

Fig. 19. Lateral view of the proximal part of the anal fin of G. hispaniolae, rays 1-7. and the hemal spine ofthe 13th vertebra be- to a lesser degree, eighth anal-fin rays, while comes associated with the ligament that at- they do not elongate, are specialized at their taches to the anterior vertebra as the ligastyle. distal tips. These branched rays become He showed that the same process ofhistolysis thickened and swollen, due to increased seg- ofthe first hemal arch occurs in females also, mentation distal to the branching point. There and a small remnant of the hemal spine is is sometimes a bit of anastomosis of these sometimes seen in small females; however, segments, making these fin ray tips look scal- the female "ligastyle" never becomes asso- loped or paddlelike. ciated with a suspensory ligament or migrates The third ray is very thick, from its base forward, and in larger females, it too is com- to just before the terminal specializations. pletely histolyzed. This is not a branched fin ray. At the distal The sixth and seventh proximal anal radials parallel the gonactinostal complex, but are dorsally limited by the first gonapophysis. Similarly, the second gonapophysis provides the dorsal border for the eighth and ninth proximal anal radials. The medial radials of the anal fin also are modified for gonopodial support. The medial radials of the second, third, and fourth anal- fin rays are fused to the proximal radials, as part of the columnar gonactinostal complex. The medial radials of the fifth, sixth, and seventh anal-fin rays, however, are separate units and highly developed. Each produces an anterior spike, and expanded lateral processes, facing posterodorsally. GONOPODIUM (figs. 19, 20) The main units ofthe gonopodium are anal- fin rays 3, 4, and 5, which, at the onset of sexual maturity, elongate, thicken, and dif- ferentiate. The first anal-fin ray is reduced to a small splint, and the second, although thickened at its base, also is relatively short. The thickening at the base of ray 2, however, is Sa 4p 4a 3 unique to Gambusia. The sixth, seventh, and, Fig. 20. Gonopodium of G. hispaniolae. 1989 RAUCHENBERGER: GAMBUSIA 31 tip ofthe gonopodium, segments ofthe third bow, and the punctata species group of that ray develop spines along the lateral (anterior) subgenus modifies the simple triangle into a edge. (A note on directional terminology: spatulate form. It may be composed ofa sin- when the gonopodium is being used, the male gle segment, or up to eight segments. It often swings it around so that ray 3 is relatively has small lateral processes arising from its anterior, ray 5 posterior. This anterior-pos- sides (in the affinis and puncticulata species terior labeling is used in differentiating po- groups). Distal to the elbow, ray 4a some- sitions on the gonopodium, whether it is in times is quite constricted, sometimes fore- use or resting.) The number of spines varies shortened, such that ray 4a does not reach from 8 to 17; high spine number characterizes the end of the gonopodium, or sometimes the panuco species group, whereas very low rather curved, due to an increased number spine number is a feature ofthe senilis species of segments distal to the elbow. Proximal to group. The length of the spines also varies, the elbow, ray 4a also is constricted some- from quite short and stubby to long and thin, times, forming a distinct notch in the ray (in or long and straplike. In most groups, these the nobilis and senilis species groups). spines are directed primarily toward the dis- Ray 4p has segments that bear posteriorly tal tip, with only the panuco and rachowi directed serrae. Serrae on ray 4p are common species groups possessing many recurved to most poeciliids. Characteristics ofthe ser- spines. In all species, however, the most prox- rae vary within the genus, however. The imal spine may be recurved. The outline of number varies from as few as three to as many the spines, and thus the anterior outline of as eight. High serrae numbers are found in the gonopodium, varies among the species the panuco species group, and unusually low groups: fanlike (panuco species group); point- serrae numbers in the senilis species group. ed mediodistally (vittata, affinis species group, Their position relative to the elbow varies- nicaraguensis species group); rounded (no- distal, opposite, or proximal; generally among bilis and puncticulata species groups); point- poeciliids, the serrae are not very close to the ed anterodistally (senilis species group); or distal tip, therefore, species with serrae op- lobelike (punctata species group). posite the elbow (the nobilis species group) The segments ofray 3 that bear spines may or distal to the elbow (the rachowi species be further modified to develop proximally group, and within the affinis species group) directed inner processes along their posterior are considered to be specialized in this trait. (medial) edges. Inner processes most often The shape of the serrae in most poeciliids is are associated with long spines, as in the no- for them to be pointing sharply proximally. bilis and senilis species groups of the subge- This general shape is modified in some groups nus Arthrophallus, and the puncticulata and to serrae that project horizontally from the punctata species groups of the subgenus segment (such as in the puncticulata species Gambusia. group) to highly antrorse recurved serrae (as The fourth anal-fin ray is relatively un- seen in G. sexradiata and G. eurystoma of modified near the base, but the termini ofits the nobilis species group). two branches (ray 4a and ray 4p) are highly Just distal to the serrae on ray 4p, some specialized. Ray 4a has an anterior swelling, groups in Gambusia are characterized by a termed the elbow, that is possessed by all large segment or two, as seen in the punctic- Gambusia species except G. vittata and by no ulata and senilis species groups. Beyond this, other poeciliid; G. vittata and Belonesox be- the ray is narrow, with small segments. Dis- lizanus do exhibit a swollen, ankylosed ray tally, a hooked terminus is common to the 4a that is bent anteriorly at the spot where genus. It may end in a rounded hook, curved an elbow would be, and this is interpreted as posteriorly, that is composed of a single ele- being homologous to the elbow, character- ment. It sometimes is distinguished further istic of the genera Gambusia and Belonesox. by being either a rather weak hook, pointing The shape ofthe elbow varies among groups; posterodistally, or by being enlarged and claw- it is most commonly falcate; the subgenus shaped. Alternatively, the terminal modifi- Gambusia is distinguished by a triangular el- cation of ray 4p might be acuminate, the 32 AMERICAN MUSEUM NOVITATES NO. 2951 anterior point reaching toward (or even izations ofrays 6, 7, and 8, the proximal bulge touching) the distal tip of ray 4a, and the ofray 5, and the thickening ofray 3 are active, posterior tip hooked toward ray 5a. Acumi- followed by the appearance of spines on ray nate hooks are found in the subgenera Arthro- 3. In the differentiation period, first the serrae phallus and Gambusia, although some mem- of ray 4p appear, followed by the terminal bers of those groups have the more general hooks of ray 4p and ray 5a, and lastly, the rounded hooks. The sharpness of the distal elbow. This order of appearance seems ste- point varies; it might be softened, and in G. reotyped in the genus, as I observed the same wrayi is recurved. The acuminate tip of ray thing in many species. Beyond this ordering, 4p often is multisegmented; hooks with sev- the gonopodium, like most fin rays, develops eral distinct segments are found in the affinis along a proximal-to-distal gradient. There- species group, and some members of the fore, although it is observed that, in the hook puncticulata species group. at the end of ray 4p, for example, the single- Ray 5 also is a branched ray, but ray 5p segment, posteriorly directed hook ontoge- does not extend all the way to the tip of the netically precedes the multisegment, ante- gonopodium, and therefore is not involved riorly directed part of the acuminate hook, in the terminal specializations. At the prox- this simply may be a consequence ofthe fact imal end ofthe anal fin, ray 5 has a prominent that the anteriorly directed hook is distal to bulge, characteristic ofthis genus alone. Dis- the posteriorly directed hook. A similar distal to this bulge, ray 5 is split open posteriorly, cussion can be construed for the single-seg- with serrate edges. Ray 5p ends just before ment elbow versus the multisegment elbow. the level ofthe serrae on ray 4p. Ray 5a con- I have considered ontogenetic information tinues, often rather straight but sometimes where possible, given this caveat, and ulti- curving sharply anteriorly, to end in a pos- mately incorporated it into an overall par- teriorly directed hook, found also in the sister simony argument. group to the genus, Belonesox. Like the hook on ray 4p, this may be rather weak, pointing posterodistally, or enlarged and clawlike. SYSTEMATICS Most often, however, it is simply rounded, although a small knob sometimes appears on In the discussion ofcharacters pertinent to its anterodorsal edge. The terminal segment the study of relationships that follows, each of ray 5a sometimes has a long base, sug- character is followed by a number in brackets gesting that it has fused with the penultimate [ ]. The systematic results ofeach section are segment. The spotty distribution ofthis char- summarized with branching diagrams con- acter, however, has limited its usefulness. taining character numbers. These cladograms The interpretation of gonopodial charac- were derived using PAUP; the character ma- ters is discussed in the systematic section. In trices (using the same numbering system for general, outgroup comparison is less helpful the characters) and tree statistics are included here than with other types of characters be- in the Appendix. cause outgroups simply do not possess comparable specializations. Ontogenetic criteria can be employed, but with a degree of cau- OUTGROUP RELATIONSHIPS (fig. 21) tion. Turner (1941) described the morpho- Tables 4, 5, and 6 show the placement of genesis of the gonopodium of Gambusia af- Gambusia within a hierarchical classifica- finis, concentrating on a comparison of tion. Following Rosen and Parenti (1981), segmentation, bifurcation, growth, and an- atherinomorphs are the sister group to the kylosis of segments between males and fe- other group of acanthopterygians, the per- males. In a later work (1942a), he described comorphs. The Series Atherinomorpha con- gonopodial formation in terms of develop- tains two divisions. It is within Division II mental fields, with a growth period followed that the order Cyprinodontiformes lies, with by a differentiation period. During the growth the order Beloniformes as its sister group. period, fields corresponding to the special- Parenti's (1981) classification of the order 1989 RAUCHENBERGER: GAMBUSIA 33

TABLE 4 TABLE 6 Classification of Atherinomorphs, from Rosen Classification of Poeciliids, from Rosen and Bailey and Parenti (1981: 23) (1963: 10) with the Addition of Scolichthys from Rosen (1967) Series Atherinomorpha Division I Family Poeciliidae (= Subfamily Poeciliinae of Paren- Family Atherinidae ti, 1981) Family Bedotiidae Subfamily Tomeurinae Family Isonidae Tomeurus Family Melanotaeniidae Subfamily Poeciliinae Family Phallostethidae Tribe Poeciliini Family Telmatherinidae Alfaro Priapella Division II Poecilia Xiphophorus Order Cyprinodontiformes (see table 5 for classifi- Tribe Cnesterodontini cation of this order) Phallotorhynus Phalloptychus Order Beloniformes Phalloceros Cnesterodon Suborder Adrianichthyoidei Tribe Gambusiini Family Adrianichthyidae Bachyrhaphis Gambusia Suborder Exocoetoidei Belonesox Superfamily Exocoetoidea Tribe Girardiini Family Hemiramphidae Girardinus Quintana Family Exocoetidae Carlhubbsia Superfamily Scomberesocoidea Tribe Heterandriini Family Belonidae Priapichthys Heterandria Family Scomberesocidae Neoheterandria Poeciliopsis Phallichthys Tribe Scolichthyini Scolichthys TABLE 5 Subfamily Xenodexinae Classification of Cyprinodontiformes, from Xenodexia Parenti (1981: 462-463) Order Cyprinodontiformes Suborder Aplocheiloidei Cyprinodontiformes is presented in table 5. Family Aplocheilidae I Family Rivulidae Following convention, refer to members of Suborder Cyprinodontoidei the order vernacularly as cyprinodontiforms, Section 1 and of the suborder Cyprinodontoidei as cy- Family Profundulidae prinodontoids. Parenti expanded the mem- Section 2 bership of the family Poeciliidae to include Division 1 procatopines, Fluviphylax, and Pantanodon, Family Fundulidae and the fishes vernacularly referred to as poe- Division 2 ciliids now constitute the subfamily Poecili- Sept 1 inae, as it was originally defined by Regan Family Valenciidae (1913). Although I accept her classification, Sept 2 I have to use the vernacular term Superfamily Poecilioidea continued Family Anablepidae poeciliid to refer to members ofthis subfam- Subfamily Anablepinae ily. Subfamily Oxyzygonectinae A classification of poeciliids is presented Family Poeciliidae in table 6, based primarily on Rosen and Bai- Subfamily Poeciliinae ley (1963) with the addition of a more re- Subfamily Fluviphylacinae cently described tribe (Rosen, 1967). Al- Subfamily Aplocheilichthyinae though this classification does not imply Superfamily Cyprinodontoidea sister-group relationships as the two previous Family Goodeidae tables do, there is evidence that the tribe Family Cyprinodontidae Gambusiini is a monophyletic group. Its sis- 34 AMERICAN MUSEUM NOVITATES NO. 2951

_I%~~~ 0

A.- %-

/1,34

shipsintetribe Gmbusiinae 2,5,6 A B 1,394 Fig. 21. Cladogram depicting interrelationships in the tribe Gambusiinae. ter group, however, remains unknown. Mem- bers of the tribe Gambusiini are similar in many ways to some members of the tribe Heterandriini, but it has not been determined if these similarities are primitive or derived. Members of the tribe Gambusiini- Brachyrhaphis, Belonesox, and Gambusia - share specializations of the gonactinostal complex. In figure 22A, Heterandria jonesi shows the general condition for many poeciliids. The first five gonactinosts are united by a laminar plate in the anterior-posterior plane. The gonactinosts are spread out on this plate like spokes in a fan. The plate divides and curves posterior to the fourth gonactinost, extending as lateral wings. In Brachy- C D rhaphis cascajalensis (fig. 22B), the anterior Fig. 22. Lateral view ofthe gonactinostal com- part ofthe plate is somewhat reduced but still plex of: A. Heterandria jonesi, B. Brachyrhaphis present. The lateral wings are modified to- cascajalensis, C. Belonesox belizanus, D. Gam- ward embracing gonactinost 5. Gonactinost busia holbrooki. 5 stands freely, rather than being incorporated into the plate, and projects dorsally Turner, 1942b) represents the lateral wings, through the lateral wings [1]. Belonesox bel- that are expanded to form a cup [2]. This izanus (fig. 22C) and all Gambusia species distinctive form ofthe gonactinostal complex (G. holbrooki in fig. 22D; also see G. hispan- is quite consistent in Belonesox and all of iolae in fig. 18) also have a free gonactinost Gambusia; it is known nowhere else. 5, projecting dorsally through the modified These three genera share some pigmenta- superior lateral wings. In these two genera, tion characters, particularly an axillary blotch, the anterior plate is much reduced, and gon- anterodorsal to the pectoral fin insertion [3]. actinosts 2, 3, and 4 form a column. The Hildebrand (1938) and Rosen and Bailey dorsal part of this column (the "collar" of (1963) have commented on the striking re- 1989 RAUCHENBERGER: GAMBUSIA 35

A B Fig. 23. Gonopodia of: A. Brachyrhaphis cascajalensis, B. Belonesox belizanus.

semblance in body form and pigmentation terior bend where the elbow would be [6]. that exists between Brachyrhaphis and Gam- The same sort of ray 4a is seen in Gambusia busia, particularly in the anal-fin area of the vittata (see fig. 33A); all the other Gambusia female. Brachyrhaphis species have a very species, although they have well-formed el- dark band of pigment at the base of the anal bows, also show these coalesced segments and fin. Anal-fin pigment is not quite as dark in anterior bending to some degree. Gambusia; it ranges from very dusky to very faint, usually heaviest on the first few rays. Females in both genera have dark pigment DIAGNOSIS OF THE GENUS GAMBUSIA around the genital opening, its expression The genus Gambusia, as described here, varying from species to species and depen- includes G. vittata, G. rachowi, and G. echea- dent on the female's physiological state [4]. gayari, fishes that have each been assigned Belonesox shows no anal pigment, either to monotypic genera by some authors. A se- around the genital opening or on the anal fin, ries ofderived characters places these species and these pigments are seen only rarely else- within Gambusia, and therefore their ban- where in the subfamily. ishment from the genus is unwarranted, as Gambusia and Belonesox share some gon- that classification would not reflect sister- opodial features, in particular, a well-formed group relationships. hook on ray 5a [5] (fig. 20; see also fig. 23). Males of all species of Gambusia have Although the termini of rays 4p and 5a do modified pectoral fins [7], as described in the curve over in some Brachyrhaphis species, Osteology section (fig. 1 5B), and, contrary to no hook is formed. The pointed anterior tip the claims ofHubbs and Reynolds (1957) and ofthe Belonesox hook is seen in several Gam- others, this modification is not seen, to any busia species (see G. nobilis, fig. 40A) al- degree, in Belonesox, or, in fact, any other though most are more rounded. Belonesox poeciliid. Details ofthis modification and its does not have an elbow on ray 4a; however, degree ofdevelopment vary within the genus. this ray is modified from the straight ray that The proximal part ofthe male anal fin (fig. is seen in Brachyrhaphis, to be thickened, 19) shows two synapomorphies for the genus, somewhat coalesced, and with a sharp an- a thickening of ray 2 at its base [8], and a 36 AMERICAN MUSEUM NOVITATES NO. 2951 prominent bump in ray 5 [9]. Again, these for members of the panuco species group. characters are consistent within the genus, Gambusia rachowi, of the rachowi species and not found elsewhere. group, has often been assigned to a mono- In the gonopodial suspensorium, there are typic genus, Heterophallus. The subgenus several reductive trends in the form of the name, Heterophallina, was chosen by Hubbs gonapophyses. Although there are some poe- to indicate some degree of affinity between ciliids that have no gonapophyses at all, the the groups. most common condition is for there to be Gambusia vittata and the members ofthese three gonapophyses, each with full uncinate two species groups share a gonopodial spe- processes. Within the genera Gambusia and cialization (see fig. 33). The hooks on rays 4p Poecilia, there are subgroups that have only and 5a are distinctively shaped [12]. They two gonapophyses; on all other bases, these are rounded anteriorly, and the posterior point two genera do not seem to be very closely does not hook over completely but rather related. In all members of Gambusia, there points at an angle. The hooks are also small, is no uncinate process on the first gonapoph- each consisting ofonly a single segment. This ysis [10]. This process is quite prominent in distinctive shape is considered a derived con- Belonesoxand Brachyraphis; however, as with dition for this subgenus, rather than a prim- the reduction in number of gonapophyses, itive state, because the hooks of Belonesox this reductive character is seen in some dis- show a different shape, tending toward an- tantly related poeciliids. terior points. The hypothesis that the Belo- In the gonopodium, all Gambusia species nesox state is primitive for the genus is sup- except G. vittata have an elbow in ray 4a [1 1]. ported by the fact that it does appear Because G. vittata does possess the other syn- occasionally in some members of the other apomorphies for the genus [7-10] and those subgenera of Gambusia. for its subgenus, its lack of a well-formed A unique morphology ofthe median teeth elbow implies the retention of a primitive of the toothplate of the third infrapharyn- character, not a basis for grouping. gobranchial also characterizes this group [13]. These large teeth are recurved at their tips, as they are in several species of Gambusia, DLAGNOSES OF SUBGENERA but on the medial surface, below the tip, is a As shown in table 1, I recognize three sub- serrate lump or pad (fig. 24). In other subgenera-Heterophallina, Arthrophallus, and genera, these median teeth are usually smooth Gambusia. Each ofthese names has been pro- along their inner borders, or havejust a slight posed for a subgenus before, by Hubbs in hint of a shoulder along the medial edge. 1926. Note, however, that although the des- Arthrophallus: Hubbs (1926) created this ignated type species for each subgenus, of subgenus for G. patreulis, a name often syn- course, remains within it, the constituency of onymized with G. affinis (see discussion by each subgenus is changed from that originally Geiser, 1923, and in the footnote about the proposed. confusion over these names2). Arthrophallus Heterophallina: Hubbs (1926) used this name for a new subgenus to include three newly described species, G. vittata, G. pa- 2A nomenclatural digression is necessary to clarify the nuco, and G. regani, designating G. regani as status of some of the members of this group. Baird and the type species. At the time, he knew G. Girard (1854) described two Texas mosquitofishes as vittata only from females, and when he later Heterandria affinis and H. patreulis, which differ slightly obtained G. vittata males, he removed it from in head shape and dorsal-fin ray number. After Poey's Gambusia altogether (in Rivas, 1963). How- 1854 description ofthe genus Gambusia, Girard (1859a) referred these two species to the new genus, adding G. ever, I have reinstated vittata to Gambusia, holbrooki of Florida (1859b), which he cited as being and placed it in the subgenus where it was described by Agassiz in the genus Heterandria, and later originally described. In addition to G. vittata, G. speciosa of the Rio San Juan in Nuevo Leon and G. this subgenus also includes two species gracilis of Matamoras (1859c). Jordan and Copeland groups-panuco and rachowi. Rivas (1963) (1876) listed G. holbrooki and G. affinis as valid, with used the subgenus name Heterophallina only patreulis, gracilis, and speciosa as doubtful species. How- 1989 RAUCHENBERGER: GAMBUSIA 37 has page priority over Schizophallus, another Hubbs (1926) subgenus, this one created for G. holbrooki, which is also often synonymized with G. affinis. Rivas (1963) retained Arthrophallus as a monotypic subgenus. As described below, the subgenus Arthro- A B C phallus is not monotypic; rather, it contains Fig. 24. Median teeth of third pharyngobran- three species groups- affinis (including, chial toothplate: A. Gambusia vittata, B. Gam- among others, both G. affinis and G. hol- busia yucatana, C. Gambusia luma. brooki), nobilis, and senilis. The cephalic sensory canal system provides a consistent syn- 2a; 2b, 3, 4a), infraorbital, posterior section apomorphy for this subgenus [14]. Rosen and only (4b, 5, 6a; 6b, 7), preopercular (7 neu- Mendelson (1960) surveyed cephalic sensory romasts), preorbital (7 neuromasts), and canals in a range ofpoeciliids, noting that the mandibular (4 neuromasts). Of interest here carnivorous top-feeding species, e.g., Brachy- is the infraorbital section. In all members of rhaphis, Belonesox, and Gambusia, have open the subgenusArthrophallus, there is a distinct grooves rather than closed canals, with the break in this canal, leaving two discontinuous following sets ofneuromasts: supraorbital (1, grooves (4b, 5, 6a; 6b, 7) (fig. 25A). In ever, Jordan and Gilbert (1886) recognized G. patreulis However, the ability to interbreed is a primitive char- with G. gracilis now synonymized under this name. Ev- acter (Rosen, 1979), and shared primitive characters are ermann and Kendall (1894), however, synonymized G. not valid for describing groups, even if the level of the patreulis, G. gracilis, and G. speciosa under the name G. group is a single species. Secondly, all affinis x holbrooki affinis, and in 1896, G. holbrooki was also placed in this hybrids are not viable; a chromosomal difference in- synonymy (Jordan and Evermann, 1896-1900). volving female heterogameity in one form (holbrooki) However, Regan (1913) recognized three valid species, but not the other results in only the offspring ofholbrooki G. holbrooki, G. patreulis, and G. affinis, the latter con- males with affinis females surviving (Black and Howell, taining speciosa as well, differentiated again by slight 1979). Furthermore, natural hybridization occurs else- variations in dorsal-fin ray number. Geiser (1923) dem- where in the genus: between G. nobilis and G. affinis onstrated that G. holbrooki is gonopodially distinct from (Hubbs, 1955); G. heterochirand G. affinis (Hubbs, 1957); all other species under question, being the only species G. georgei and G. affinis, G. nobilis and G. geiseri, and in the genus to have denticulations on the posterior edges G. hurtadoi and G. heterochir (the last in artificial pools) of the ray three segments proximal to the spines. He (Hubbs and Peden, 1969). Clearly if the interbreeding recommended full species status for holbrooki, with all criterion does not apply in these cases, it need not apply other forms named affinis. in the affinis-holbrooki case either. Like any symplei- Adding to the confusion, Hubbs (1926) decided to siomorphy, the ability to interbreed is certainly indica- limit the name G. affinis to a new species, related to G. tive of relationship at some (higher) level. All reported senilis, from San Marcos, Texas, and to use the name cases ofnatural hybridization occur within the subgenus G. patreulis for the widespread species usually called Arthrophallus, and therefore this ability, although ap- affinis. Under G. patreulis, he included G. speciosa, G. parently lost in some apomorphic taxa, may characterize modesta, and G. myersi (the latter two names proposed this large group. Finally, Wooten et al. (1988) have re- by Ahl for a fish ofthe Pfanuco region in eastern Mexico). cently examined the affinis-holbrooki matter biochemi- As a separate subgenus, Hubbs recognized G. holbrooki, cally, and have found the two taxa to be electrophoret- following Geiser. Hubbs and Springer (1957), after ex- ically distinct. They also recommend the recognition of amining the co-types for Heterandria affinis, described G. holbrooki as a valid species. the San Marcos mosquitofish as G. geiseri, and the wide- Gambusia speciosa, as a valid species, has appeared spread fish was again renamed G. affinis, this time the on two recent checklists, without comment as to its status name G. patreulis finally disappearing. (Miller, 1986; Smith and Miller, 1986). I have examined Krumholz(1948), C. L. Hubbs(1955), C. Hubbs(1957), some new material, from Melchor Musquiz, Mexico, and and Hubbs and Springer (1957) have all cited two un- have found it to be gonopodially distinct from other published manuscripts that reported natural introgres- forms of Gambusia, and therefore provisionally treat it sion between G. affinis and G. holbrooki, and therefore as a full species. However, I have not yet been successful have recognized these forms as only subspecifically dif- in locating the type material; this will be necessary for ferent. (Hubbs et al., 1953, also listed G. affinis speciosa, full resolution of this problem, which clearly needs fur- from Comal Springs.) ther investigation. 38 AMERICAN MUSEUM NOVITATES NO. 2951

A B Fig. 25. Dorsal views of the cephalic sensory canal systems of: A. Gambusia senilis, B. Gam- busia pseudopunctata. A B Fig. 26. Epibranchial 1: A. Gambusia regani, Brachyrhaphis, Belonesox, and gambusiins B. Gambusia pseudopunctata. in the subgenera Heterophallina and Gam- busia, the two sections ofthis canal are joined into one continuous groove (fig. 25B). At this to be a separate character. The metacentric level, disjunct canals must be considered a chromosome in females only is inferred to be synapomorphy of the subgenus Arthrophal- a synapomorphy of the subgenus Arthro- lus; it is consistent within the group. The gen- phallus, but this is only tentative, and must erality of this character-continuous versus be tested by karyotyping more species. discontinuous-above this level, for exam- Gambusia: Seven characters [15-21] de- ple, among all poeciliids, was not examined. scribe the subgenus Gambusia. The only gon- It was not discussed by Rosen and Mendelson opodial character refers to the shape of the (1960), and unfortunately their figures do not elbow of ray 4a. In the other two subgenera, consistently distinguish the two conditions. the elbow is distinctly falcate and hooked. In Karyotypes are known for several mem- the subgenus Gambusia, the proximal surface bers of this subgenus (G. affinis of the affinis of the elbow is flat; rather than falcate, the species group, G. nobilis ofthe nobilis species elbow is triangular in shape [15]. In the punc- group, and G. hurtadoi of the senilis species tata species group, the triangular shape is fur- group, for examples) (Campos and Hubbs, ther modified to a more spatulate outline; 1971; Chen and Ebeling, 1968). The species these fish still have a flat, unarched, proximal mentioned have a diploid number of48, with surface to the elbow. a moderate to large metacentric chromosome In the dorsal gill arches, the first epibran- in females only [59]. This karyotype is dif- chial has no uncinate process for species of ferent from that known for Belonesox, for this subgenus [16] (fig. 26). Although this rep- members ofthe subgenus Gambusia (G. luma, resents a reversion to the condition general G. rhizophorae, G. xanthosoma- Wildrick et for atherinomorphs, the rest ofthe genus does al., 1985), and for members of the subgenus have such a process. In the ventral gill arches, Heterophallina (G. vittata, G. regani; G. mar- the members ofthe subgenus Gambusia pos- shi is autapomorphic with a reduced number sess very well-developed anteriorly directed of chromosomes). Except for the subgenus lateral prongs on the first hypobranchials [1 7] Arthrophallus, the general karyological con- (fig. 27). Such processes are sometimes seen dition seems to be a diploid number of 48, in Heterophallina and Arthrophallus, but with no metacentric chromosomes. merely as thickenings, not prongs. The third A subgroup of the punctata species group, basibranchial also is modified in this group in the subgenus Gambusia, has a different from the relatively straight rod that is seen karyotype, with one pair ofmetacentric chro- in the other two subgenera to an hourglass- mosomes in both sexes, but this is inferred shaped bone (fig. 27). The bone is slender at 1989 RAUCHENBERGER: GAMBUSIA 39

A B Fig. 27. Anterior part ofventral gill arches, dorsal view: A. Gambusiapanuco, B. Gambusia yucatana, C. Gambusia puncticulata. the point where the two second hypobran- plate, usually with a keyhole near the base. chials articulate with it, but just posterior to Most members of the subgenus Gambusia this point the bone flares out in lateral wings have two hypural plates, as shown for G. luma [18]. (see fig. 44), also with a keyhole opening [20]. The ventral aspect ofthe ventral hypohyal This condition is seen occasionally in the oth- is distinctive in this group as well. As dis- er subgenera, for example the rachowi species cussed earlier (p. 20), a group containing poe- group, but best characterizes this level. ciliids is described in part by a ventral hy- Another character that fits best at this level pohyal that contacts the anterior ceratohyal but is not completely consistent has to do squarely, in a straight line. This condition is with pigment patterns on the dorsal and cau- seen in the other two subgenera. In the sub- dal fins. Most but not all members of Het- genus Gambusia, the anterior ceratohyal has erophallina and Arthrophallus have a basal an anterior lobe that invades much of the row of spots on these fins, with dusky to very space of the ventral hypohyal; therefore, the dark fringe at the edge ofthe fins. In the sub- juncture between these two bones is not at genus Gambusia, the fin pattern consists of all square [19] (fig. 28). three or four rows ofspots (fig. 29). This spot- The pattern of hypural fusion in the genus ted pattern is seen occasionally in the other is complex. Generally, adults ofHeterophal- subgenera (e.g., G. holbrooki of Arthrophal- lina and Arthrophallus have a single hypural lus, and G. panuco of Heterophallina) [21].

Fig. 28. Ventral hypohyal and anterior ceratohyal, ventral view: A. Gambusia panuco, B. Gambusia wrayi, C. Gambusia punctata. 40 AMERICAN MUSEUM NOVITATES NO. 2951

Fig. 31. Suborbital bar, shown on Gambusia holbrooki.

varies from being just a few melanophores running in a curved line below the eye to a very dark streak in the same location (fig. 3 1). The degree ofexpression ofthe suborbital bar B also varies, depending on the age, sex, and = physiological state of the organism. In the gonopodium, members of these two subgen- Fig. 29. Dorsal and caudal fin patterns: A. era develop an acuminate terminal hook on Gambusia alvarezi, B. Gambusia pseudopunctata. ray 4p, drawn out distally and sometimes arching anteriorly [23]. In many cases, the RELATIONSHIPS AMONG acuminate hook is multisegmented. This SUBGENERA (fig. 30) condition is thought to be more specialized The subgenera Arthrophallus and Gam- than the single-segment, rounded hook be- busia are sister groups, with the subgenus cause during the morphogenesis of the gon- Heterophallina as the sister group to that as- opodium, the proximal segment of the hook semblage. Members of both Arthrophallus forms first, and it much resembles a rounded and Gambusia have a suborbital bar [22] that hook; it is not yet acuminate. The distal portion is next formed, producing the acuminate tip. This ontogenetic sequence, however, also

2,13 Fig. 30. Cladogram depicting relationships Fig. 32. Cladogram depicting relationships in among subgenera. the subgenus Heterophallina. A

Fig. 33. Gonopodia of members of the subgenus Heterophallina: A. Gambusia vittata, B. Gambusia marshi, C. Gambusia panuco, D. Gambusia regani, E. Gambusia rachowi, F. Gambusia echeagayari. 41 42 AMERICAN MUSEUM NOVITATES NO. 2951 reflects the general prc ximal-to-distal direc- is a high number of slender spines on ray 3, tion of specialization. typically 14-17 [28]; the number of spines in Peden (1973a) described some differences other groups usually ranges from 8 to 13. among species in the form of the anal spot Some members ofthe punctata species group in the female. Members ofthepanuco species have up to 15 spines, but these are broad and group, as well as Brachyrhaphis species, have straight with long inner processes. In the pa- a median anal spot. Members of the subgen- nuco species group, the spines curve like a era Arthrophallus and Gambusia that he de- fan, and their distal tips do not end pointedly, scribed have paired crescent-shaped spots but gently merge into the membrane of the around the anus and genital pore instead [24], gonopodial margin [29]. Spines in other except for three members ofthe senilis species groups have distinct pointed edges. The ter- group, whose median spots are considered minal segment of ray 5a in this group, that derived at that level. Although this synapo- bears the hook at its distal tip, is proximally morphy seems straightforward enough based elongated, covering the distance of two seg- on Peden's work, the character could not be ments [30]. The terminal segment may have examined in the many species that he did not fused with the penultimate one; however, no cite, because, as he pointed out, the anal spot evidence of fusion was seen. This particular is often not seen well in preserved specimens character is found occasionally in other groups (his observations were of live material). in the genus. Within the panuco species group, G. panuco and G. regani are more closely related RELATIONSHIPS WITHIN THE to each other than either is to G. marshi. SUBGENUS HETEROPHALLINA (fig. 32) Gambusia marshi, however, is polymorphic The panuco and rachowi species groups to- with regard to pigmentation patterns and gether form a monophyletic group. Three morphometrics (Minckley, 1962 and person- gonopodial peculiarities unite them (fig. 33): al commun.); it may represent a complex of retrorse spines on ray 3, as opposed to all species, but variation at that level was not spines basically pointing forward [25]; an in- investigated here. The gonopodia of G. pa- crease in the number of segments to six or nuco and G. regani have more serrae on ray more in ray 4a distal to the elbow, with a 4p than any others in the genus [31]. These concomitant reduction in size [26]; and a two species have eight or nine serrae, G. mar- similar increase in the number of segments shi usually five, the rachowi species group distal to the serrae of ray 4p often up to nine and G. vittata two to four, and other groups or more [27]. The normal range in the num- in the genus typically four to six. These serrae ber of these segments in the genus is from extend directly perpendicular from ray 4p four to eight, usually four or five. [32]. This modification is seen again in some G. vittata: As noted earlier, this species lacks other groups (see the puncticulata species the well-formed elbow on ray 4a of the gon- group, and within the nobilis species group) opodium. It also is remarkable in that it is but must be considered a more advanced the only Gambusia, ofthose for which stom- condition than the retrorse serrae that are ach content analyses have been performed, common not only in the genus but to most that is reported to eat much algae (specifically poeciliids. Gambusia panuco and G. regani filamentous Spirogyra), which accounted for can be distinguished from each other on me- up to 50 percent ofthe stomach contents. The ristic (G. regani has fewer dorsal-fin rays) and remainder consisted of arthropod parts, a pigmentary (G. regani has no spotting on the more conventional diet for a Gambusia (Dar- body) grounds. Otherwise, these sister species nell, 1962). are very similar. panuco species group: The three members rachowi species group: This group contains of the panuco species group, G. marshi, G. two species that previously have been as- panuco, and G. regani, have a distinctive signed to two monotypic genera. Hetero- gonopodium (fig. 33B-D); this group has been phallus rachovii was inadvertently rede- recognized at some level as a cohesive unit scribed as Gambusia atzi by Rosen and by all previous students of the genus. There Gordon (1951); while recognizing its differ- 1989 RAUCHENBERGER: GAMBUSIA 43

SCL 4;l k 4; CLP CLP a0 a0 a04

0! .4;~

A B Fig. 34. Cleithral process: A. Gambusia luma, B. Gambusia echeagayari. ;,38,39,61 ences from the other members of Gambusia, they stressed commonalities, especially the 14,59 elbow of ray 4a in the gonopodium. When describing Dicerophallus echeagayari, Alva- Fig. 35. Cladogram depicting relationships in rez (1952) had not seen any material of G. the subgenus Arthrophallus. rachowi; the two species are actually quite close and do not belong in separate genera. Rather Rosen and Bailey (1963), following Rosen ofthe genus (or allied genera) (fig. 34). and Gordon (1951), retained both of these than being a simple, rounded process, it is species within Gambusia, as the rachowi more like a drawn-out ellipse. species group, as I have done here. Rivas (1963), however, recognized the monotypic RELATIONSHIPS WITHIN THE taxa and described them as "allied genera." SUBGENUS ARTHROPHALLUS (fig. 35) These two species do have a very strange The nobilis and senilis species groups form gonopodium that clearly distinguishes them a monophyletic unit. In their gonopodium, as a monophyletic group (fig. 33E, F). There ray 4a has a distinct notch proximal to the are no forward-pointing spines on ray 3 [33]. elbow [38]. Although obviously all species There is an odd buckling ofray 4a [34], distal with elbows narrow in ray 4a just proximal to the elbow, that results in the elbow, which to the elbow, in these groups this narrowing is quite prominent and falcate, being dis- soon widens and ray 4a resumes its normal placed proximally. A fleshy palp grows lat- width, with the transition from narrow to wide erally from the gonopodium between rays 3 being rather abrupt. In other groups, ray 4a and 4a, just distal to the elbow [35]. Present also widens, but does so gradually, with no in both species, this palp is developed to a noticeable demarcation of narrow and wide greater degree in G. echeagayari. Both species zones (compare figs. 41 and 43 with those of are very slender fish, as are all members of other groups, e.g., figs. 33, 47, 49, 51). These this subgenus, but these two narrow to a steep two groups also share long spines on ray 3, ventral keel; all other species are more round- with well-developed inner processes [39]. This ed ventrally [36]. Additionally, these two particular character, as just described, also fishes have a caudal-fin skeleton that consists will characterize a monophyletic subunit of of two unfused hypural plates [20]. Within the subgenus Gambusia (see p. 49); however, this subgenus, all other fishes have hypurals the width, shape, and outline of the spines fused into one plate, with a keyhole in the and inner processes are different in the two middle. This condition is not unique within groups. Most members of these two species the genus, however; the subgenus Gambusia groups typically have only two gonapophyses also is characterized by two unfused hypurals (fig. 37), instead of the more general number in the caudal skeleton. Finally, these two of three, often with a slight swelling on the species have a cleithral process that extends third hemal spine being the only indication further posteriorly than in any other species of a third gonapophysis [61]. Also, the mem- 44 AMERICAN MUSEUM NOVITATES NO. 2951

0 Go)4,

,44

Fig. 37. Gonopodial suspensorium of Gam- ,57 busia georgei. Compare with fig. 18.

Fig. 36. Cladogram depicting relationships slightly toward one side. A lateral projection within the affinis species group. from the gonopodium is seen in the rachowi species group, but the projection is a large, fleshy palp, dissociated from (or not associ- bers ofthese groups have no uncinate process ated with, at any rate) the elbow. In the affinis on the first epibranchial [16]. The difficulty species group, the lateral projection is a bony in interpreting the generality ofthis character prominence on the elbow. Finally, these fish- was discussed earlier (p. 21); although the es all have a particularly impressive pectoral character may not be a synapomorphy, it does fin in males, with a unique modification [57]. describe the condition found in these two The fifth pectoral ray has a distinct scalloped subgroups and is congruent with other syn- portion near its distal tip, something more apomorphies. than the blade that is developed above this affinis species group (fig. 36): This consti- ray in most species ofthe genus (fig. 39). Ex- tution of this group is the most novel prop- cept for the autapomorphic G. heterochir of osition of this work. Although Rosen and the nobilis species group, this group has the Bailey (1963) designated an "affinis species most pronounced pectoral-fin modifications group," theirs was a rather uncharacterizable in the genus. assemblage of some fishes not only of this Gambusia aurata and G. lemaitrei both subgenus but also some ofthe subgenus Gam- have very high serrae on ray 4p of the gon- busia. The group as here described contains opodium, extending to distal to the elbow five members, sharing the following special- [43]. In these two species, ray 3 is somewhat izations (fig. 38). The elbow of ray 4a of the foreshortened, such that it does not reach the gonopodium consists of many segments, up distal tip of the gonopodium [44]. And in G. to ten, which are coalesced along their an- affinis and G. holbrooki, the gonopodia are terior borders [40]. Elbows in other groups exaggeratedly drawn out at their distal tips, are most often unisegmental, or consist of at coming to a much more acute point than in most three segments. The spines of ray 3 are any other species in the genus [62]. small and stout, and their distal tips are bifid nobilis species group (fig. 40): A group with [41]. The small size might be considered a this name was implied in a key by Hubbs in retention ofthe condition found in Belonesox 1926, and in 1929 he formalized it with the and Brachyrhaphis, whose spine-bearing seg- description of G. gaigei. He grouped G. no- ments of ray 3 do not have true spines but bilis, G. senilis, and "G. afJinis" with this new are slightly drawn out anteriorly; however, species, but the characters describing the these "spines" are not bifid. A further pe- group are actually applicable to larger groups culiarity ofthis group is a lateral bulge at the in Gambusia. Hubbs and Springer (1957) anteroproximal tip of the elbow [42]. It is clarified that the "G. affinis" ofHubbs (1926, usually more pronounced on one side of the 1929) was not G. affinis, but a new species, gonopodium than the other, so that in the G. geiseri. They added another two species resting gonopodium, ray 4a is displaced to the group, G. alvarezi and G. hurtadoi. 1989 RAUCHENBERGER: GAMBUSIA 45

C C

Fig. 39. Male pectoral fins. Modified parts of pectoral fins are stippled darkly: A. Gambusia vittata, B. Gambusia holbrooki, C. Gambusia pseudopunctata. Note: Compare with fig. 15B for G. hispaniolae.

Fig. 38. Gonopodia of members of the affinis species group: A. Gambusia affinis, B. Gambusia kV holbrooki, C. Gambusia speciosa, D. Gambusia 6) aurata. Note: G. lemaitrei also belongs to this 4.0 group. See Rivas (1963) for a figure of its gonopodium.

Their group definition again consisted primarily of primitive characters, although the composition corresponds roughly to the com- bination of the nobilis and senilis species groups as described herein. The same applies to the Rosen and Bailey (1963) nobilis species group; they also added G. heterochir, G. Ion- gispinis, and G. atrora. Rivas (1963) did distinguish the nobilis and senilis species groups; however, his nobilis species group is again Fig. 40. Cladogram depicting relationships not characterized by derived traits, but rather within the nobilis species group. 46 AMERICAN MUSEUM NOVITATES NO. 2951

Fig. 41. Gonopodia of members of the nobilis species group: A. Gambusia nobilis, B. Gambusia georgei, C. Gambusia heterochir, D. Gambusia krumholzi, E. Gambusia eurystoma, F. Gambusia sexradiata. by the lack of the derived traits of the senilis opposite the elbow rather than proximal to species group. it [43], and by the rounded gonopodial out- Indeed, members of the nobilis species line [47], unique to this group within the subgroup do share several derived characters with genus (fig. 41). the senilis species group, indicating a sister- Gambusia nobilis, G. heterochir, G. geor- group relationship (see p. 43); however, the gei, G. krumholzi, G. eurystoma, and G. nobilis species group can be characterized by sexradiata comprise this species group. Gam- the position of the serrae on ray 4p, being busia eurystoma and G. sexradiata clearly are 1989 RAUCHENBERGER: GAMBUSIA 47 E F

Fig. 41-Continued. sister taxa, sharing recurved serrae on ray 4p nilis, G. gaigei, G. alvarezi, G. hurtadoi, and of the gonopodium [32], spotted dorsal and G. amistadensis, is characterized by a median caudal fins [21], continuous grooves in the anal spot in females (Peden, 1973a), rather posterior section of the infraorbital latero- than the paired crescents more common in sensory canals [14], and three gonapophyses the subgenera Arthrophallus and Gambusia [61]. No other resolution of relationships [24], and a transversely enlarged segment on within this group was possible, however. Each ray 4a distal to the serrae [48]. Transversely of the four remaining species has certain au- enlarged segments also are seen in the punc- tapomorphies, for example the exaggerated ticulata and punctata species groups of the pectoral-fin modification of G. heterochir subgenus Gambusia, but in those fishes, the males, but no synapomorphies uniting any segments extend primarily toward ray 4a, and subgroup of these were discernable. in the members of the senilis species group, senilis species group (fig. 42): The senilis the extension is toward ray 5a only. Fur- species group was first proposed by Rivas (1963), to contain G. senilis, G. geiseri, G. gaigei, G. alvarezi, and G. hurtadoi. To his list I add G. longispinis of his nobilis species group, G. atrora, and G. amistadensis. Three gonopodial features distinguish this group (fig. %v 43). Ray 4a is foreshortened distal to the elbow, such that it does not extend to the end ofthe gonopodium [45]. Consequently, there are very few segments in ray 4a distal to the elbow. There also is a reduced number of spines on ray 3, typically eight [46]. Finally, these fishes have a very distinctive gonopodial outline [47]; the furthest extension ofthe gonopodial tip is drawn out to a point on the anterior edge, and the medial distal edge actually dips a little proximally before being drawn out to that tip. This group can be divided into two mono- Fig. 42. Cladogram depicting relationships phyletic subgroups. One, containing G. se- within the senilis species group. 48 AMERICAN MUSEUM NOVITATES NO. 2951

A C

Fig. 43. Gonopodia ofmembers ofthe senilis species group: A. Gambusia senilis, B. Gambusia gaigei, C. Gambusia alvarezi, D. Gambusia hurtadoi, E. Gambusia amistadensis, F. Gambusia geiseri, G. Gambusia longispinis, H. Gambusia atrora. thermore, these senilis-type species usually should be emphasized that all five members have only one expanded segment, with an of this clade are quite similar. abrupt transition to more narrow segments The sister clade to this group consists of just distal to it. In the puncticulata and punc- G. geiseri, G. longispinis, and G. atrora. These tata species groups, there often are two, three, three are united by a reduction in the number or four expanded segments that gradually of serrae on ray 4p, to fewer than four [31], narrow distally. Within this small group, G. no suborbital bar [22], and a form ofhypural hurtadoi and G. amistadensis share a small fusion unique within the genus (fig. 44). There lateral projection on the elbow ofray 4a [42], is only one hypural plate, a complete fusion and G. alvarezi shares with these two a sim- of the hypural elements into a fan, with no ilar hypobranchial 1 shape [60]; however, it keyhole remaining [20]. Gambusia atrora and 1989 RAUCHENBERGER: GAMBUSIA 49

4:0

A B

/715-21 Fig. 45. Cladogram depicting relationships among species groups in the subgenus Gambusia.

curve anteriorly as they extend proximally, c V ' D and they narrow to fingerlike proximal edges. A. Fig. 44. Caudal fin skeletons of: Gambusia Similarly, the shapes of the spines, and par- luma, B. Gambusia hispaniolae, C. Gambusia of the as atrora, D. embryonic caudal skeleton of G. his- ticularly the outline gonopodia paniolae. formed by the spines, are distinctive for each

G. longispinis are sister species within this clade, as these two species have three gonapophyses, whereas other members ofthe nobilis and senilis species groups typically have only two [61]. 4.~~ ~ ~ ~ ~~~.

RELATIONSHIPS WITHIN THE SUBGENUS GAMBUSIA (fig. 45) Within this subgenus, a group consisting of the puncticulata and punctata species groups can be described. These fishes share several synapomorphies in gonopodial structure. In ray 4p, just distal to the serrae, one or two segments (or more) are transversely enlarged [48] in all species except G. luma. As in the nobilis and senilis species groups, the spines of ray 3 are elongate and develop inner processes [39]; however, the inner pro- 19 cesses of this group are wider and flatter at their proximal edges. In the nobilis and senilis Fig. 46. Cladogram depicting relationships species groups, the inner processes tend to within the nicaraguensis species group. 50 AMERICAN MUSEUM NOVITATES NO. 2951 A B EU EU~~~~~C ~ ~~~CL

\ /17, 23, 24 , '

449

428 Fig. 47. Gonopodia of members of the nica- Fig. 48. Cladogram depicting relationships raguensis species group: A. Gambusia nicara- within the puncticulata species group. guensis, B. Gambusia wrayi. Note: Gambusia melapleura also belongs to this group; see Fink (1 97 lb) for a figure of its gonopodium. novation is not shown in either Fink's (197 ib) or Regan's (1913) figures. This is one of several areas in the genus where additional taxo- species group. These two groups are further nomic work at the alpha level would be fruit- separated from the nicaraguensis species ful. group by the fact that their serrae on ray 4p The nicaraguensis species group, as herein are straight, not retrorse [32]. diagnosed, contains G. nicaraguensis, G. nicaraguensis species group (fig. 46): Rivas wrayi, and G. melapleura; present evidence (1963) described a nicaraguensis species group places G. hispaniolae in the puncticulata to contain G. nicaraguensis, G. wrayi, G. mel- species group (fig. 47). apleura, G. gracilior, and G. dominicensis, These fishes have a more triangular ventral but did not use derived characters for its de- hypohyal than that found elsewhere in the scription. Fink (197 lb) revised Rivas' group, genus [19]. In figure 28, different conditions adding G. hispaniolae. He also discussed the of the ventral aspect of the ventral hypohyal status ofG. dominicensis, described by Regan are portrayed. The general condition for the from "Haiti," which has never been subse- genus is for the ventral hypohyal to be rough- quently collected there or anywhere. I have ly square, meeting the anterior ceratohyal in not examined Regan's syntypes, but Fink's a straight line. In the subgenus Gambusia, figures clearly indicate that the common the anterior ceraohyal has an anterior lobe, Gambusia ofHispaniola, which is often iden- so that the connection ofthe two bones is no tified as G. dominicensis, is actually G. his- longer square. In the puncticulata and punc- paniolae, and that, as Fink suggests, Regan's tata species groups, however, the anterolat- locality information for G. dominicensis is eral edge of the ventral hypohyal is still probably incorrect. Fink (1971b) also syn- squared. In the nicaraguensis species group, onymized G. gracilior under G. wrayi, after the anterolateral corner of the ventral hy- examining Regan's syntypes ofboth and con- pohyal is no longer squared; rather, the edge cluding that the proportional differences cited of the bone cuts directly from the antero- by Regan (19 13) were indistinct. Not having medial edge to the posterolateral edge, re- examined these syntypes myself, I accept the sulting in a triangular bone. synonymy, but cautiously. Material I have Gambusia wrayi and G. melapleura have examined of G. wrayi shows a dimorphism odd segments in ray 4a distal to the elbow, in gonopodial structure, with a unique re- each with two or three serrations in its an- curved form of the terminal hook on ray 5a, terior edge [63]. and also a distinctive pelvic girdle in these puncticulata species group (fig. 48): The same samples. This particular gonopodial in- puncticulata species group of Rivas (1963) 1 989 RAUCHENBERGER: GAMBUSIA 51

Fig. 49. Gonopodia of members of the puncticulata species group: A. Gambusia yucatana, B. Gam- busia hubbsi, C. Gambusia baracoana, D. Gambusia bucheri, E. Gambusia puncticulata, F. Gambusia oligosticta, G. Gambusia caymanensis. Note: G. manni, G. monticola, and G. howelli also belong to this group, but are not figured. See Fink (1971a) for some of these figures. See fig. 20 for G. hispaniolae. was revised by Fink (1971a). In that work, pigmentation, and gonopodial counts and he synonymized the eight forms in the group measures, and the rationalization for com- under the name G. puncticulata, with G. bining all these forms under one species name puncticulata puncticulata representing the was that they were not gonopodially distinct. former G. puncticulata, G. oligosticta, G. cay- Additionally, the forms combined in the manensis, G. hubbsi, and G. howelli. Gam- nominal subspecies had overlapping ranges busia manni, G. bucheri, G. baracoana, and for all of the characters studied. G. yucatana were reduced to subspecific sta- Greenfield and Wildrick (1984) reexam- tus, along with G. puncticulata monticola. ined the group, using additional characters This study was primarily based on meristics, including electrophoretic analysis. They pro- 52 AMERICAN MUSEUM NOVITATES NO. 2951

bow in ray 4a of the gonopodium [42], reminiscent of the lateral knob of the affinis species group; however, in that group, the lateral projection is a knob projecting from the anteroproximal tip ofthe elbow, while in this group the thickening is along the flat proximal edge, seeming more like a shelfthan a knob. The members of this species group also share a somewhat different pectoral-fin modification. In other groups, the blades on rays 2 through 5 are developed primarily on the dorsal side of the ray. In this group (fig. 39), the blades are equally well developed on both the dorsal and ventral sides [58]. Greenfield and Wildrick (1984) united all species in the group except G. yucatana by several electrophoretic characters. They also Fig. 50. Cladogram depicting relationships noted that all island species ofthis group show within the punctata species group. red and yellow coloration in the dorsal and caudal fins, especially in breeding males [49]. Gambusia yucatana shows no such colors. posed that G. yucatana be reinstated to full Another character they use at this level is the species rank and be the sister group to the longer serrae on ray 4p possessed by the is- other members ofthe group. In fact, subspe- land species as compared to G. yucatana; al- cies G. yucatana yucatana and G. yucatana though this is true, the sister group of the australis were later described (Greenfield, puncticulata species group, the punctata 1985). The island members of the species species group, also has long serrae. group were divided into two sections: one Gambusia hubbsi, G. manni, G. hispanio- containing G. puncticulata from Cuba, G. lae, G. baracoana, G. bucheri, and G. mon- howelli, G. caymanensis, and G. oligosticta; ticola constitute a monophyletic group, cor- the other with G. puncticulata from the Ba- responding (except for the addition of G. hamas (= G. hubbsi), G. manni, G. bucheri, hispaniolae) to the "hubbsi group" designat- G. baracoana, and G. monticola. These groups ed by Greenfield and Wildrick (1984). Within were found to be electrophoretically distinct the group, G. hubbsi and G. hispaniolae have from each other and from G. yucatana. hypurals fused into a single hypural plate, Clearly, further alpha-level work needs to with a keyhole [20]. Gambusia bucheri, G. be done to address the species problems in baracoana, and G. monticola share an in- this group. It is beyond the scope of the cur- creased number of lateral-line scales and rent work to evaluate the relative merits of vertebrae [50]; Fink (197 la) documented the these two studies; however, it is my intent to range ofvariation in these meristic characters try to determine which taxa form groups, and for the species group. which groups are related to which. Although In Greenfield and Wildrick's (1984) "punc- specimens of some of the endemic taxa are ticulata group," consisting ofG. puncticulata, quite scarce, I have chosen to include all G. oligosticta, G. caymanensis, and G. how- nominal forms in the analysis; taxa shown to elli, the anteriorly directed lateral prongs of form a monophyletic group might at some the first hypobranchials in the ventral gill later point be synonymized if their differ- arches, characteristic of the subgenus Gam- ences prove to be inconsequential. Further- busia, are reduced to very small bumps [17] more, I wish to have all the taxa available (fig. 26C). This group also is characterized by for use in the biogeographic analysis. a permanent median anal spot in the females; As described herein, the members of the Greenfield and Wildrick (1984) reported no puncticulata group (fig. 49) have a lateral anal spot either in G. yucatana or in the hubb- thickening along the proximal base of the el- si forms, although Peden (1973a) indicated 1989 RAUCHENBERGER: GAMBUSIA 53

Fig. 51. Gonopodia of members of the punctata species group: A. Gambusia luma, B. Gambusia pseudopunctata, C. Gambusia punctata, D. Gambusia rhizophorae, E. Gambusia xanthosoma. Note: G. beebei also belongs to this group. See Rivas (1969) for a figure of its gonopodium. 54 AMERICAN MUSEUM NOVITATES NO. 2951 that some of these forms have paired cres- space between ray 4p and ray Sa, sometimes cents. with small anterior projections from ray 5a punctata species group (fig. 50): This species filling in the gaps [48]. group was recognized by both Rivas and Ro- Gambusia xanthosoma and G. rhizophorae sen and Bailey, although these authors dis- share a unique karyotype, reported by Wild- agreed with a few members. The gonopodium rick et al. (1985). One pair of metacentric is quite distinctive (fig. 51), with a spatulate chromosomes and two pairs of submetacen- elbow [15]. The hooks on both rays 5a [51] tric chromosomes are found in both sexes and 4p [52] are enlarged and scythelike. The [54]. Gambusia luma also was karyotyped by spines on ray 3 are very long and straplike the same authors; it has no metacentrics or [39], as are their inner processes, which reach submetacentrics in either sex, inferred to be far in, almost touching ray 4a. These spines the general condition for the genus (see dis- form a lobelike outline to the gonopodium, cussion, p. 38). Gambusia beebei, G. pseu- not pointed at all [47]. In the ventral gill arch- dopunctata, and G. punctata were not tested, es, the wings on the third basibranchial are so it is unclear how large a group this char- reduced [18] in all species examined except acter might specify. In the gonopodium of G. G. pseudopunctata; this character was not ex- xanthosoma and G. rhizophorae, ray 4a distal amined in G. xanthosoma or G. beebei. to the elbow curves in an S-shape anteriorly As diagnosed above, this species group in- over the first of the ray 3 spines [55]. These cludes six members: G. luma, G. beebei, G. two species also share a pigmentation char- pseudopunctata, G. punctata, G. xanthoso- acter, the development of a distinct midlat- ma, and G. rhizophorae. All of these species eral stripe, and they are both the normal small except G. luma have the suborbital bar re- size of most gambusias. Other members of duced or absent [22], and several morpho- this species group attain large sizes, with G. metric and meristic characters described by beebei at up to 59 mm standard length (Rivas, Greenfield (1983) apply at this level, such as 1969) being the largest species of the genus. a more anterior placement of the dorsal fin Gambusia rhizophorae and G. xanthosoma and higher dorsal-fin ray counts. All mem- are the only species of the group that thrive bers ofthis species group except G. luma have in brackish water; they are both found as- a unique pectoral-fin modification [57] in sociated with mangrove swamps. males, where the blade on ray 2 thickens abruptly near its proximal end, and the seg- BIOGEOGRAPHY ments are anastomosed in this area. Gam- Distributional information presented busia luma does not have the transversely herein is compiled from various sources, in- enlarged segments on ray 4p just distal to the cluding the materials examined, the original serrae, but otherwise its placement in this descriptions of the species, Geiser (1923), group is obvious, contrary to the arguments Hubbs and Springer (1957), Rosen and Bai- of Rivas (1969). ley (1963), Rivas (1963, 1969, 1971), Fink Except for the suborbital bar of G. luma (197 1 a, 197 1b), and Greenfield et al. (1982), and the postocular bar that is prominent in as well as other sources below. all species ofthis group, pigmentation is very mentioned plain in G. luma and G. beebei. The other DISTRIBUTION OF THE four species ofthe group are united by a well- SUBGENUS HETEROPHALLINA (fig. 52) developed pigmentation pattern, of four to six rows of large dark spots along the sides Members of this subgenus are known only ofthe body [53], a striking pattern that earned from strictly freshwater habitats, most often G. punctata its name. the headwaters and swiftly flowing streams Gambusia punctata, G. xanthosoma, and that drain the Sierra Madre Oriental. G. rhizophorae have especially pronounced Gambusia vittata: This species is limited enlargements of the posterior edges of the to the Rio P'anuco and Rio Tamesi systems segments just distal to the serrae on ray 4p of Tamaulipas and San Luis Potosi, Mexico. of the gonopodium. These extensions fill the It is sympatric with G. aurata and G. affinis 19891RAUCHENBERGER: GAMBUSIA 55

7-G.vittata A -G.panuco *-G. regani O-G. marshi -G echeagayari O-G. rachowi Fig. 52. Distribution of members of the subgenus Heterophallina.

in the Rio Mante, and with G. atrora in the P'anuco valley. It is found in the Rio Tamesi Rio Axtla system. It also is sympatric with system, except at the extreme headwaters of G. panuco throughout much of its range; al- the Rio Sabinas, where it is replaced by G. though G. vittata is not found in the Rio Ojo regani. In the Rio Panuco system, it is the Frio system, it is found in P'anuco tributaries only Gambusia in the Rio Ojo Frio system, further east and south than G. panuco (such and it coexists with both G. vittata and G. as the Rio Atlapexco). It may extend south atrora in the Rio Axtla. It is found as far into the Rio Tuxpan system. south as the Rio Calabozo, but drops out after panuco species group: Two of the species that. It also is found in headwater areas of of this group are confined to the Rio P'anuco the Rio Soto La Marina, to the north. and Rio Tamesi systems; the third is isolated Gambusia regani is confined to the head- from them in northern Mexico. All three waters ofthe Rio Sabinas system, Rio Tamesi species in this group are allopatric with re- drainage, Tamaulipas, Mexico. Although spect to one another. some other localities in the Tamesi and Soto Gambusia marshi is widespread in the Rio La Marina systems are cataloged in some col- Salado system and in parts ofthe Cuatro Cie- lections as G. regani, these are in fact G. pa- negas basin in Coahuila, Mexico. It shares nuco. the Cuatro Cienegas area with G. longispinis, rachowi species group: Each of the two and may be sympatric in part of the Salado species ofthis small group is known only from drainage with G. speciosa. the area of the type locality, both in south- Gambusia panuco, as described above un- eastern Mexico. Both are sympatric with G. der G. vittata, is widespread throughout the sexradiata. 56 AMERICAN MUSEUM NOVITATES NO. 2951

0 0 0 0 0 0 0 0 o 030 0 00 0 0 o 0 0 0 0 0 0 0 00 a 0 0

X. 9'. . X, C_,;

0.G holbrooki 0- G.affinis *' Gspeciosa 0* U- Gaurata CG. lemaitrei ..

Fig. 53. Distribution of members of the afJinis species group.

Gambusia rachowi is found at Jesus Car- some limited success; Gambusia apparently ranza, Veracruz, in the Rio Coatzocoalcos prefers to eat larval fish when available, and system. the introduction of these species has often Gambusia echeagayari inhabits a small had devastating effects on the native fish fau- stream near the town of Palenque, Chiapas, na, including other species of Gambusia. It which drains into the Rio Usumacinta sys- is difficult to know certainly the natural ranges tem. of these two taxa, but careful compilations and record-keeping allow a reasonable ap- DISTRIBUTION OF THE SUBGENUS ARTHROPHALLUS proximation. Gambusia holbrooki is found in fresh water This subgenus is distributed over conti- along the Eastern Seaboard of the United nental and coastal habitats, from eastern States, from the southern tip of New Jersey North America to southern Mexico, with a through to the Florida Keys, and in western single species in South America. It has no drainages from Georgia into the GulfofMex- members either in lower Central America or ico. It is replaced by G. affinis further west; the Antilles; the greatest diversity of taxa oc- the exact dividing line or the nature of a zone curs in Texas and northern Mexico. of intergradation was not studied here. It is affinis species group (fig. 53): This species broadly sympatric with G. rhizophorae in ex- group is the most far-ranging of the genus, tremely southern Florida. both as a group, from New Jersey to Colom- Gambusia affinis, in addition to the Gulf bia, and as individual species, as G. afijnis drainages of Louisiana, Alabama, Mississip- and G. holbrooki are quite widespread. These pi, and Texas, ranges further north in the two species have been introduced worldwide United States (to Illinois) and is common to control mosquitos. These efforts have had throughout the Rio Grande and Pecos River 1989 RAUCHENBERGER: GAMBUSIA 57

°- G. sexradiata *- G.eurystoma A- G.nobilis *- G.georgei - G.heterochir A- G. krumholzi

Fig. 54. Distribution of members of the nobilis species group.

drainages. It also is reported to occur in Gulf (personal commun.) also reports it from the drainages along the east coast of Mexico as Devil's River, Texas. The full extent of the far south as Tampico. However, such reports range ofthis species also needs further study. were not checked by the author, and many As with G. affinis and G. holbrooki, locality of these southern localities may be referable records for this species are confusing because to G. speciosa. In Texas, G. affinis is sym- of the years of nomenclatural confusion. patric (but seldom syntopic) with many Gambusia aurata resides in the Rio Mante, members of the nobilis and senilis species ofRio Tamesi drainage, in Tamaulipas, Mex- groups. ico. It has been taken, both in the nacimiento Gambusia speciosa was described from the and canals through Ciudad Mante, with three Rio San Juan, in Nuevo Leon, and a popu- other Gambusia species: G. affinis, G. pa- lation recently has been found at Melchor nuco, and G. vittata. Darnell (1962) and Mil- Musquiz, in a stream of the Rio Salada-Rio ler and Minckley (1970) described the mi- Grande drainage, in Coahuila. Clark Hubbs crohabitats of these sympatric species. 58 AMERICAN MUSEUM NOVITATES NO. 2951

Gambusia lemaitrei offers the oddest dis- a small stream in the Rio Grande drainage, tributional note. Known only from Totumo, of Coahuila, Mexico. Colombia, it is the only South American rep- Gambusia eurystoma was described from resentative of the genus. However, it is not a sulfurous, spring-fed stream, Arroyo del well collected, and no further details of its Azufre, at Bai-nos de Azufre, 10 km west of distribution are available. Teapa, Tabasco, Mexico, and is found only nobilis species group (fig. 54): Most of the at that locality (Miller, 1975). species of this group are endemic to fresh- Gambusia sexradiata, first described as a water springheads. The precarious nature of subspecies of G. nicaraguensis, inhabits the these habitats, in the face of development, coastal drainages of Mexico from the Rio has warranted three of these species to be Nautla south to the Yucatan peninsula. It federally listed as endangered: G. nobilis, G. also is found throughout the Usumacinta heterochir, and G. georgei. The Mexican G. drainage of Guatemala. In these areas, it is krumholzi is limited to a small area, but is primarily a freshwater form. In Belize, how- not threatened; G. eurystoma also is a strict ever, Greenfield et al. (1983) reported that it endemic whose status is still safe. Only G. also is found in brackish and sea water. In sexradiata ofthis group has a widespread dis- Belize, G. sexradiata shares its habitat with tribution at present, and it is the only mem- G. luma and G. yucatana. ber found in other than a strictly freshwater senilis species group (fig. 55): As with the situation. nobilis species group, many members of this Gambusia nobilis is endemic to the Pecos group are strictly limited to small, threatened River drainage in Texas and New Mexico, springhead habitats. Gambusia gaigei and G. found primarily in small springs and sink- amistadensis are listed as endangered; G. holes. Geographic variation among the iso- amistadensis is reported to be extinct. lated springhead populations was recently Gambusia senilis is most common in the studied by Echelle and Echelle (1986), who Rio Conchos system of Chihuahua and Du- concluded that subspecific designations were rango, Mexico; it also occurred in the Devil's unwarranted. Historical records indicate that River of Texas, but is now extinct there. It the species was previously both more abun- inhabits clear springs, turbid water, and ir- dant and widespread, and C. Hubbs (personal rigation ditches, and although widespread, commun.) reports hybridization of this much of its habitat dries up periodically. species with G. affinis. Gambusia gaigei, the Big Bend Gambusia, Gambusia georgei, of the upper San Mar- was described from Boquillas Spring, Rio cos River, Hays County, Texas, is sympatric Grande drainage, Brewster County, Texas. with G. geiseri, of the senilis species group, Hubbs and Brodrick (1963) report that the and G. affinis, of the affinis species group. population was extirpated at the type locality, According to Hubbs and Peden (1969), these but that the fish was introduced into Croton three species partition the environment, with Springs, is in Graham Ranch Springs, and is G. geiseri most common in midwater, G. af- found in an artificial pool in Big Bend Na- finis in shallow vegetated areas, and G. geor- tional Park. gei restricted to shallow areas with no vege- Within the range of G. senilis are two iso- tation. This rare (and possibly extinct) species lated areas, each containing an endemic also hybridizes somewhat with G. affinis, re- species. Gambusia alvarezi is known only ported by Hubbs and Peden (1969). from El Ojo de San Gregorio (a warm spring) Gambusia heterochir is restricted to the and adjacent creeks. Gambusia hurtadoi is headwaters of Clear Creek, Menard County, described from El Ojo de la Hacienda Do- Texas. Clear Creek flows into the San Saba lores, another warm spring, and associated River, where G. affinis occurs, and hybrid- irrigation ditches. Both are in Chihuahua, ization between the two species, in Clear Mexico. Creek, has been reported by Hubbs (1957). Gambusia amistadensis was described from Gambusia krumholzi is known only from Goodenough Spring and its outflow to the the type locality, the Rio de Nava, which is Rio Grande in Val Verde County, Texas (Pe- 1989 RAUCHENBERGER: GAMBUSIA 59

00 0

0 00 a 0

A G. atrora A - G. longispinis °' G.geiseri v - G. amistadensis * - G. hurtadoi * - G. alvarezi ° G. senilis V- G. gaigei

Fig. 55. Distribution of members of the senilis species group. den, 1973b), based on material collected in Gambusia longispinis is restricted to the April 1968. In July of that same year, the Cuatro Cienegas basin of Coahuila, Mexico. construction ofthe Amistad Reservoir caused Gambusia marshi, of the subgenus Hetero- permanent flooding of the spring, destroying phallina, is more widespread, occurring in the habitat necessary for this species. Al- parts of the Rio Salado system. though some stocks were maintained in ar- Gambusia atrora is rather far removed from tificial aquaria, Peden (1 973b) surmised that the remainder ofthe species group. It is found the indigenous population of this species is in the Rio Axtla and its tributaries, in the Rio extinct. Panuco drainage, San Luis Potosi, Mexico. Gambusia geiseri is found in several sys- It is sympatric here with G. panuco and G. tems in Texas, always near the emergence of vittata. cold, clear springs. It occurs in the San Mar- cos, Guadelupe, Conchos, and Devil's rivers, and is sympatric with G. affinis. Hubbs and DISTRIBUTION OF THE Springer (1957) suspected that it is native SUBGENUS GAMBUSIA only to the San Marcos Springs and Comal Each ofthe species groups in this subgenus Springs and introduced at the other Texas is primarily Caribbean, but each also has one springs, based on earlier collection records. mainland member, in Central America. 60 AMERICAN MUSEUM NOVITATES NO. 2951

0 0 0 o

O - G. nicaraguensis *- G melapleura v- G. wrayi

Fig. 56. Distribution of members of the nicaraguensis species group.

nicaraguensis species group (fig. 56): Gam- Gambusia melapleura is restricted to the busia nicaraguensis is found in Atlantic Coast type locality, freshwater springs of the drainages of Central America, as far south as Shrewsberry River, Bluefields, Jamaica. the Rio Chagres, Panama. It generally is found puncticulata species group (fig. 57)3: Gam- in fresh water in lower Central America, where busia yucatana, the only mainland member it is the only Gambusia present, although Bel- of this species group, is found in Mexico, onesox and Brachyrhaphis are common here; Guatemala, and Belize. Greenfield (1985) in the northern extremity of its range, in Be- distinguished two subspecies, G. yucatana lize, it is most often found in high-salinity yucatana from the northern Yucatan pen- river mouths or in full strength sea water. In insula, and G. yucatana australis further Belize, it is sympatric with G. luma, but G. south, based on multivariate analysis ofmor- luma prefers less saline waters; in saline phometric data. The species dwells in a va- waters, G. nicaraguensis sometimes occurs riety ofhabitats, from inland freshwater lakes with G. yucatana. As with G. yucatana (see to brackish coastal streams to nearly full- below), slender- and deep-bodied forms can be described, with deeper-bodied forms occurring at more inland, freshwater localities 3As noted on p. 52, I have not made any formal (Greenfield et 1982). recommendations as to the status ofthe taxa in the punc- al., ticulata species group. Recent workers (Fink, 1971a; as Fink Gambusia wrayi, redescribed by Greenfield and Wildrick, 1984) disagree as to the number (197 lb), is common in the freshwater drain- of species-level taxa in this group. Not having looked at ages ofJamaica, most populous in the south- enough material to make alpha-level decisions, I have ern drainages. In brackish and salt waters, it simply included all the named taxa provisionally, in or- is replaced by the G. oligosticta form of the der that statements about relationships and biogeogra- puncticulata species group. phy could be made. 1989 RAUCHENBERGER: GAMBUSIA 61

I@ t. a ., -G. yucatana % DO 0 O-G. puncticulata G)-G. howelli A -G caymanensis (D -G. oligosticta A-G. hubbsi *-G. bucheri A-G. baracoana *- G. monticola V- G. manni V-G hispaniolae Fig. 57. Distribution of members of the puncticulata species group. strength sea water. Greenfield (1985) also maican form, is occasionally found with G. noted that, in both subspecies, the freshwater wrayi, but in general G. oligosticta occupies inland forms are more slender-bodied than brackish water and G. wrayi fresh. As with their coastal counterparts. Gambusia yuca- G. howelli, however, it must be emphasized tana is sympatric with members ofthree oth- that the taxonomic distinction of G. oligo- er species groups, G. luma, G. nicaraguensis, sticta and G. caymanensis from G. puncti- and G. sexradiata. culata is not clear at this point. The group corresponding to Greenfield and The "hubbsi group" also has members in Wildrick's (1984) "puncticulata group" has Cuba, but these are restricted to purely fresh- members throughout Cuba and the Isla de la water habitats, narrowly endemic to localities Juventud (Isle of Pines), Grand Cayman Is- on the extreme east end of the island. Other land and Cayman Brac, and Jamaica. Gam- members are found in the Bahamas and His- busia puncticulata (sensu stricto) is quite paniola. In Cuba, G. monticola inhabits a common across Cuba, sympatric with both swift mountain stream, the Rio Yao, Rio G. punctata and G. rhizophorae, but G. punc- Cauto system, Province of Oriente; G. ba- ticulata prefers more brackish water than G. racoana is described from a freshwater pond punctata. Gambusia howelli, the form from at the mouth of the Rio Miel, near Baracoa, the Isla de la Juventud, is not readily distin- Oriente; and G. bucheri is found in Jicotea guishable from the species ofthe main island Creek, Moa system, also in the Province of based on materials available, and should Oriente. Each of these is known only from probably remain in synonymy with G. punc- the type locality. Another narrow endemic in ticulata. Gambusia caymanensis is the form this group is G. manni, whose distribution found on Grand Cayman, Little Cayman, and Fink (1971a) described as limited to two Cayman Brac. Gambusia oligosticta, the Ja- freshwater lakes on New Providence Island 62 AMERICAN MUSEUM NOVITATES NO. 2951

y.V , %

I9*

O - G. luma * - G beebei

o - G. pseudopunctata v- G punctata v- G.xanthosomna .- . 0

. G rhizophorae

Fig. 58. Distribution of members of the punctata species group. in the Bahamas, Lake Cunningham and Lake patrically with introduced G. holbrooki, but Kilkarney. Other Bahamian Gambusia spec- otherwise its widespread distribution does imens are often referred to as G. manni in seem to restrict the other Gambusia species many collections, but Fink (1971a) relegated of the island, which are narrowly endemic the much more common Bahamian form to (G. pseudopunctata), rarely caught (G. bee- G. puncticulata puncticulata, which was orig- bei), or virtually unknown except for the type inally known as (and herein referred to as, at species (G. dominicensis). least provisionally) G. hubbsi. Rosen and punctata species group (fig. 58): Gambusia Bailey (1963), feeling that populations from luma ranges through Belize and Guatemala, South Andros Island, from where G. hubbsi from Rio Hondo in the north to Puerto Bar- was described (Breder, 1934), and popula- rios and the Belize River in the south; it is tions from other Bahamian islands were in- primarily coastal but is found inland distinguishable, synonymized G. hubbsi un- throughout Lago de Izabal. It is both sym- der the name G. manni. However, ifthe name patric and syntopic with G. sexradiata G. manni applies only to the fishes ofthe two throughout the northern part ofthe range; G. lakes ofNew Providence Island, the name G. luma, however, prefers rivers and streams hubbsi should be used for the other Bahamian whereas G. sexradiata is more likely to be populations. found in standing water. When together, G. Gambusia hispaniolae is common in both luma is often in the middle of streams, in fresh and brackish water habitats throughout open water, and G. sexradiata is near shore, Haiti and the southern Dominican Republic. amid vegetation (Greenfield et al., 1983). In Fink (1971b) called it the dominant Gam- southern Belize, G. luma is sympatric with busia of the island; it has been taken sym- G. nicaraguensis, but G. nicaraguensis is usu- 1989 RAUCHENBERGER: GAMBUSIA 63 ally in more brackish water. In the northern part of its range, a similar relationship exists A W E E E w E with G. yucatana, which may be found occasionally in full-strength sea water (Green- field et al., 1982). Although not commonly found in salty waters, Carter (1981) has shown that G. luma does have the physiological tol- erance to adapt to saline conditions. Gambusia beebei is endemic to Lake Mira- goane, Haiti, on the Tiburon peninsula. Ap- parently a rare fish, it is known in U.S. collections only from the neotype and "neo- paratype" series designated by Rivas in 1969. Several recent collections in Haiti have re- covered only G. hispaniolae from Lake Mir- agoane; however, a more thorough study of that lake is needed to assess the status ofthis species. Gambusia pseudopunctata is also endemic B to Haiti, again on the Tiburon peninsula. It is known from the type locality springs at Roseaux, as well as several nearby springs, but it is not widespread. As Fink (197 lb) speculated, G. hispaniolae may well have confined both G. beebei and G. pseudopunctata to their limited ranges. Gambusia punctata is known from many Fig. 59. Area cladograms for groups found in localities across Cuba and the Isla de la Ju- Texas and northern Mexico: A. nobilis species ventud; it is common and abundant, and may group, B. senilis species group. Legend: W = wide- be found up to elevations of 2000 ft (Rivas, spread, E = endemic. 1969). The type locality is from the moats of the Jardin Botanico in the city ofHavana. As is true also for G. beebei and G. pseudopunc- GEOGRAPHIC AREAS tata, G. punctata is not at all salt tolerant; in Texas and Northern Mexico (fig. 59): brackish areas it is replaced by G. puncticu- Members of the nobilis and senilis species lata. It is sympatric but not syntopic with G. groups overlap in this area, although they rhizophorae along the northwest coast of the occupy different areas of endemism, making island. direct comparison of their distribution pat- Gambusia rhizophorae, unlike the pre- terns impossible. In each of these groups, vious species, shows a great deal of salt tol- however, one species is fairly widespread (G. erance. As the trivial name implies, it lives nobilis in the nobilis species group, although in mangrove swamps, sometimes in full- its range is currently limited and diminishing; strength sea water. It occurs only in such G. senilis and G. geiseri in each of the sub- swamps, along the northwest coast of Cuba, clades of the senilis species group), and all extreme southern Florida, and the Florida others are restricted to endemic habitats, usu- Keys. ally springheads. The recurrent pattern finds Gambusia xanthosoma, a distinctive yel- the widespread species unresolved with re- low species, is confined to Grand Cayman spect to some ofthe endemic members ofthe Island, which it shares with a member of the clade. In other words, a widespread species puncticulata species group, G. caymanensis. does not appear to be more closely related to The type locality is a brackish-water mos- a particular endemic species than to the other quito control ditch, adjacent to a mangrove endemic species of the group. Hubbs and swamp. Springer (1957) have suggested that the en- 64 AMERICAN MUSEUM NOVITATES NO. 2951

p sm s c p p members of a group that includes a widespread member. Such a scheme ofperipheral isolation might also account for the lack of resolution in these groups. Rio Panuco Basin, Mexico (fig. 60): Al- A though physically not very different from similar river basins in eastern Mexico, the Rio Panuco basin is something ofa focal point in Gambusia biogeography, as members of three species groups, and G. vittata, overlap there. Two sister species ofthe panuco species group, G. panuco and G. regani, are endemic there, with G. panuco being fairly widespread in the system and G. regani endemic in a northern tributary. These two taxa are, in turn, related to G. marshi, of the Cuatro Cienegas B basin. A relationship between the Panuco basin and Cuatro Cienegas area is repeated in the senilis species group, where G. atrora (Rio Axtla-Rio Panuco) and G. longispinis are sister species. These two, however, are related to G. geiseri, of central Texas. The panuco group is related to the rachowi group, which inhabits drainages further south in Mexico (Rio Coatzocoalcos, Rio Usumacinta). The sister group to that entire assemblage, G. vittata, is again a Panuco endemic. A very similar track is exhibited by the poeciliid genus Xiphophorus, which has members in Cuatro Cienegas and northern Mexico, much diversification in the Panuco basin, and representatives southwards, through the drainages mentioned above to Guatemala. Cladistic relationships within the genus have been proposed (Rosen, 1979) and are in the process ofbeing updated, with the incorporation ofnewly discovered taxa; when complete, the degree of congruence between Fig. 60. Area cladograms for groups found in Gambusia and Xiphophorus patterns can be the Rio P'anuco basin: A. subgenus Heterophalli- explored. na, B. affinis species group, C. senilis species group. A most curious member ofthe Pfanuco ba- Legend: P = Panuco basin, C = Cuatro Cienegas sin fauna is G. aurata, endemic to the Rio basin, SM = Southern Mexico, NM = Northern Mante, in the northern Tamesi drainage, Mexico, SA = South America. which also is the home of another endemic poeciliid, Poecilia latipunctata. Gambusia demic species of these groups are differen- aurata is a member ofthe afjinis species group, tiated relicts of formerly widespread species, whose namesake, G. affinis, is a widespread isolated due to desert conditions and com- species that reaches just to the Panuco basin; petition with G. affinis. The cladograms for however, by virtue of gonopodial evidence, these groups, superimposed with distribu- G. aurata seems to be most closely related tional information, do not contradict the in- not to G. affinis but to another member of terpretation of Hubbs and Springer (1957), the species group, G. lemaitrei, from northern as, in each case, endemics are (often derived) Colombia! The affinities ofthe Panuco basin, 1 989 RAUCHENBERGER: GAMBUSIA 65 which seemingly relate to three habitats -to NCA j J Cuatro Cienegas, northern Mexico, and Tex- A as, to Cuatro Cienegas and points south in Mexico, and to South America-are complex. Seemingly unintelligible tracks, however, should not be dismissed simply because they do not agree with a priori notions, and simply because there is incongruence does not mean that these tracks are uninformative. They do imply that the Panuco basin is a much more complex area than one might ex- pect at first glance, and, as in the Caribbean, no one group ofspecies will provide the entire story. This area will be an exciting one for future research; in addition to Gambusia, Xi- phophorus, and Poecilia, already mentioned, groups of minnows and cichlids also show endemic patterns in the Panuco basin. B Central America and the Antilles (fig. 61): Perhaps the most interesting generalizations about biogeographic patterns within Gam- busia might be drawn from an examination of the tracks of the three species groups that have members in the West Indies and a comparison of those tracks to each other and to other published distributional records. The nicaraguensis species group, although it contains three members, is oflimited usefulness here as an area cladogram. Gambusia wrayi NCA H and G. melapleura are sister species, and because they both inhabit the island ofJamaica, the area cladogram for the nicaraguensis species group is reduced to a two-taxon state- C ment: a relationship between Central Amer- ica, specifically the area from Belize to Costa Rica, and Jamaica; however, it is intriguing to note that one ofthe generalized tracks doc- Fig. 61. Area cladograms for groups found in umented by Croizat in the Caribbean por- Central America and the Caribbean: A. nicara- trays the same relationship. His figure 110 guensis species group, B. puncticulata species (p. 782, 1958, I) summarizes the distribu- group, C. punctata species group. Legend: NCA = tions of the bird Mimus Neso- Nuclear Central America, J = Jamaica, H = His- genera and paniola, B = Bahamas, EC = Eastern Cuba, CI = mimus, stretching from Central America, at Cayman Islands, C = Cuba, F = Florida. the Honduras-Nicaragua bulge, to Jamaica. Some of the taxa he included in this generalized track also have representatives in the of the group implies the same relationship Galapagos. between Nuclear Central America (specifi- The other two species groups of the sub- cally the Yucatan Peninsula, Guatemala, and genus Gambusia, the puncticulata and punc- Belize; Rosen, 1975) and the major Antillean tata, offer more detailed information about islands. Within the Antillean clade are two the interrelationships among the Antillean is- monophyletic subgroups. The "puncticulata lands and Central America. Within the punc- forms," which include many ofthe taxa syn- ticulata species, a sister-group relationship onymized by Fink (1971a), show a relation- between G. yucatana and all other members ship between Cuba, the Isla de la Juventud, 66 AMERICAN MUSEUM NOVITATES NO. 2951 the Cayman Islands, and Jamaica. (Relation- the pattern described by the puncticulata ships within this clade were not resolved, species group, where a group with Cuba-Cay- however; perhaps the synonymy ofthese un- man Island members is related to a group der the name G. puncticulata would be ap- including a Hispaniolan species. Following propriate.) The second Antillean clade, the this view, one would predict that any mem- "hubbsi forms," shows a different pattern, ber of the punctata species group discovered with a greater degree ofresolution. The mem- in the future in extreme eastern Cuba or the bers of this clade inhabit the Bahamas, east- Bahamas might be likely to be more closely ern Cuba, and Hispaniola. The three Cuban related to the Hispaniolan species than to G. taxa, G. monticola, G. bucheri, and G. ba- punctata and its close associates. On the other racoana, forming a monophyletic group, are hand, the distributional data for the Hispan- each confined to restricted habitats on the iolan Gambusia species also support the pos- extreme portion ofCuba. These taxa are part sibility of a hybrid Hispaniola. The two of a group encompassing the Bahamas (G. members of the punctata species group are hubbsi and G. manni) and Hispaniola (G. his- only found on the Tiburon Peninsula, where- paniolae). Within this group, evidence sug- as the puncticulata species group represen- gests that G. hispaniolae and G. hubbsi are tative seems to be widespread across the is- closely related, implying the same relation- land. Biogeographic patterns, therefore, can ship between the Bahamas and Hispaniola, be proposed as evidence for certain scenarios with these two related to eastern Cuba. of historical geology. The puncticulata species group, therefore, One place to compare the patterns dis- lends some support to the notion of a hybrid played by Gambusia is in the compendia of Cuba, as the extreme eastern part of the is- Croizat (1958, 1964). His figure 37 (p. 146, land acts as a separate area of endemism, 1964) ofthe distribution ofcertain liverworts different from the actions ofthe species wide- indicates two tracks: one connecting Jamaica spread throughout the island and its western with Cuba (as in the puncticulata species adjunct (Isla de la Juventud). group); and a separate track uniting Belize The distribution of G. puncticulata across with southeastern Hispaniola (as in the punc- Cuba is repeated by G. punctata, which also tata group). And, his figure 96 (p. 687, 1958, is widespread in Cuba and the Isla de la Ju- I) portrays a relationship between Hispaniola ventud. Another repeated pattern is the sis- and the Bahamas (as in the puncticulata ter-group relationship between a species in- species group) for a lizard genus, Aristelliger, habiting Guatemala and Belize (G. luma) and and a passeriform, Loxigilla. This relation- the species in the Antilles. Details of the re- ship is again reflected in his figure 35 (p. 141, lationships within the Antilles for the punc- 1964) for the genus Lycium. Finally, the re- tata species group, however, reveal another lationship between Cuba, the Cayman Is- facet of the complex story of Caribbean bio- lands, and the Florida Keys is reiterated in geography. Gambusia punctata, G. rhizopho- his figure 86 (p. 633, 1958, I) for species of rae, and G. xanthosoma, a monophyletic the bird genus Columba. This selective list clade, depict the relationship between Cuba of examples is provided merely to testify to (with the Isla de la Juventud), the Florida the fact that distribution patterns of Gam- Keys, and Grand Cayman Island. This is busia in the Caribbean are not peculiar sim- reminiscent of the relationship portrayed by ply to Gambusia, but are merely instances in the "puncticulata forms," particularly in a larger scheme. terms ofthe affinities ofthe Cayman Islands. Rosen (1985), in an attempt to synthesize The next two sister taxa for this group, G. current ideas of historical biogeography in pseudopunctata and G. beebei, indicate a re- the Caribbean, derived a series of four-taxon lationship between the areas just mentioned area cladograms that would be consistent with above and southwestern Hispaniola (the Ti- that history. He predicted that groups of or- buron Peninsula), to which these two strictly ganisms with Antillean endemics should pro- endemic species are confined. Such a rela- duce area cladograms congruent with his de- tionship can be interpreted in two ways. On rived from geology. The close relationship of one hand, this pattern is not incongruent with the puncticulata species group members of 1989 RAUCHENBERGER: GAMBUSIA 67

Hispaniola, eastern Cuba, and southern Ba- coastal lagoons in East-Central Mexico. hamas is evident in his scheme, as is the re- Publ. Inst. Marine Sci. 8: 300-365. lationship between Cuba and Jamaica. By ex- de Beer, G. R. relationships based on 1937. The development ofthe vertebrate skull. amining predicted London: Oxford Univ. Press. progressive slices of geological time, he Dingerkus, G., and L. D. Uhler showed that 18 different four-taxon state- 1977. Enzyme clearing of alcian blue stained ments are consistent with this historically whole small vertebrates for demonstra- complex area. As other groups are studied, tion of cartilage. Stain Technol. 52(4): their information will add, as that of Gam- 229-232. busia has done, to a greater understanding of Echelle, A. F., and A. A. Echelle the area. 1986. Geographic variation in morphology of a spring-dwelling desert fish, Gambusia REFERENCES nobilis (Poeciliidae). Southwestern Nat- uralist 31(4): 459-468. Alvarez, J. Evermann, B. W., and W. C. Kendall 1952. Dicerophallini, nueva tribu de Poecili- 1894. The fishes ofTexas and the Rio Grande, idae de Chiapas. Ciencia 12(3-4): 95- considered chiefly with reference to their 97. geographical distribution. Bull. U.S. Baird, S. F., and C. Girard Fish. Comm. 12(for 1892): 57-126. 1854. Descriptions of new species of fishes Fink, W. L. collected by Mr. John H. Clark, on the 1971a. A revision of the Gambusia puncticu- U.S. and Mexican Boundary Survey, lata complex (Pisces: Poeciliidae). Publ. under Lt. Col. Jas. D. Graham. Proc. Gulf Coast Res. Lab. Mus. 2: 11-46. Acad. Nat. Sci. Philadelphia 6(1852- 1971b. A revision of the Gambusia nicara- 1853): 387-390. guensis species group (Pisces: Poecili- Black, D. A., and W. M. Howell idae). Publ. Gulf Coast. Res. Lab. Mus. 1979. The North American mosquitofish, 2: 47-77. Gambusia affinis: a unique case in sex Fowler, H. W. chromosome evolution. Copeia 1979(3): 1950. Colombian zoological survey. Part VI.- 509-513. Fishes obtained at Totumo, Colombia, Breder, C. M., Jr. with descriptions of two new species. 1934. A new Gambusia from Andros Island, Notulae Nat. 222: 8 pp. Bahamas. Am. Mus. Novitates 719: 3 Geiser, S. W. PP. 1923. Notes relative to the species of Gam- Campos, H. H., and C. Hubbs busia in the Unites States. Am. Midl. 1971. Cytomorphology of six species of gam- Nat. 8(8-9): 175-188. busiine fishes. Copeia 1971(3): 566-569. Girard, C. Carter, H. J. 1859a. Ichthyology ofthe boundary. In United 1981. Aspects of the physiological ecology of States and Mexican boundary survey, species of Gambusia from Belize, Cen- under the order of Lieut. Col. W. H. tral America. Copeia 1981: 694-700. Emory, Major First Cavalry, and U. S. Chen, T. R., and A. W. Ebeling Commissioner. Washington, 2(2): 85 pp. 1968. Karyological evidence offemale hetero- 1859b. Ichthyological notices. Proc. Acad. Nat. gameity in the riosquitofish, Gambusia Sci. Philadelphia 11: 56-68. affinis. Copeia 1968(1): 70-75. 1859c. Ichthyological notices. Proc. Acad. Nat. Chernoff, B. Sci. Philadelphia 11: 113-122. 1986s. Systematics ofAmerican atherinid fish- Gosline, W. A. es of the genus Atherinella. I. The sub- 1963. Considerations regarding the relation- genus Atherinella. Proc. Acad. Nat. Sci. ships of the percopsiform, cyprinodon- Philadelphia 138(1): 86-188. tiform, and gadiform fishes. Occas. Pap. Croizat, L. Mus. Zool. Univ. Mich. 629: 38 pp. 1958. Panbiogeography. Caracas: publ. by the Gosse, P. H. author. 1851. A naturalist's sojourn in Jamaica. Lon- 1964. Space, time, form: the biological syn- don: Longman, Brown, Green, and thesis. Caracas: publ. by the author. Longmans. Darnell, R. M. Greenfield, D. W. 1962. Fishes of the Rio Tamesi and related 1983. Gambusia xanthosoma, a new species 68 AMERICAN MUSEUM NOVITATES NO. 2951

of poeciliid fish from Grand Cayman Hubbs, C., and A. E. Peden Island, BWI. Copeia 1983(2): 457-464. 1969. Gambusia georgei sp. nov. from San 1985. Review of the Gambusia yucatana Marcos, Texas. Copeia 1969(2): 357- complex (Pisces: Poeciliidae) ofMexico 364. and Central America. Copeia 1985(2): Hubbs, C., and R. A. Reynolds 368-378. 1957. Copulatory function of the modified Greenfield, D. W., T. A. Greenfield, and D. M. pectoral fin of gambusiin fishes. Am. Wildrick Nat. 91(860): 333-335. 1982. The taxonomy and distribution of Hubbs, C., and V. G. Springer species of Gambusia (Pisces: Poecili- 1957. A revision of the Gambusia nobilis idae) in Belize, Central America. Copeia species group, with descriptions ofthree 1982(1): 128-147. new species, and notes on their varia- Greenfield, D. W., C. F. Rakocinski, and T. A. tion, ecology, and evolution. Texas J. Greenfield Sci. 9(3): 279-327. 1983. Spatial and trophic interactions in wet Hubbs, C. L. and dry seasons between Gambusia 1924. Studies ofthe fishes ofthe order Cyprin- luma and Gambusia sexradiata (Pisces: odontes. I-IV. Misc. Publ. Mus. Zool. Poeciliidae) from the G. yucatana com- Univ. Michigan 13: 32 pp. plex of Mexico and Central America. 1926. Studies ofthe fishes ofthe order Cyprin- Copeia 1984(4): 921-933. odontes. VI. Misc. Publ. Mus. Zool. Greenfield, D. W., and D. M. Wildrick Univ. Michigan 16: 86 pp. 1984. Taxonomic distinction of the Antilles 1927. Studies ofthe fishes ofthe order Cyprin- Gambusia puncticulata complex (Pisces: odontes. VII. G. manni, a new species Poeciliidae) from the G. yucatana com- from the Bahamas. Copeia 164: 61-65. plex of Mexico and Central America. 1929. Studies ofthe fishes ofthe order Cyprin- Copeia 1984(4): 921-933. odontes. VIII. Gambusia gaigei, a new Gregory, W. K. species from the Rio Grande. Occas. 1933. Fish skulls: a study of the evolution of Pap. Mus. Zool. Univ. Michigan 198: natural mechanisms. Philadelphia: Am. 11 pp. Philos. Soc. 1936. Fishes of the Yucatan Peninsula. Publ. Gunther, A. Carnegie Inst. Washington 457: 157- 1866. A catalogue of the fishes in the British 287. Museum. London, 6: 368 pp. 1955. Hybridization between fish species in Hennig, W. nature. Syst. Zool 4(1): 1-20. 1966. Phylogenetic systematics. (translated) Humphries, C. J., and L. R. Parenti Urbana: Univ. Illinois Press. 1986. Cladistic biogeography. Oxford: Clar- Hildebrand, S. F. endon Press. 1938. A new catalogue ofthe fresh-water fish- Jordan, D. S, and H. E. Copeland es of Panama. Field Mus. Nat. Hist., Zool. Ser. 22(4): 217-359. 1876. Checklist ofthe fishes ofthe freshwaters Hollister, G. ofNorth America. Bull. Buffalo Soc. Nat. 1940. Caudal skeleton of Bermuda shallow Sci. 3: 133-164. water fishes. IV. Order Cyprinodontes: Jordan, D. S., and B. W. Evermann Cyprinodontidae, Poeciliidae. Zoologi- 1896-1900. The fishes of North and Middle ca 25(1): 97-112. America: a descriptive catalogue of the Hubbs, C. species of fish-like vertebrates found in 1957. Gambusia heterochir, a new poeciliid fish the waters of North America, north of from Texas, with an account of its hy- the Isthmus ofPanama. Bull. U.S. Natl. bridization with G. affinis. Tulane Stud. Mus. 1-4: 1-3313. Zool. 5(1): 3-16. Jordan, D. S., and C. H. Gilbert Hubbs, C., and H. J. Brodrick 1886. List of the fishes collected in Arkansas, 1963. Current abundance ofGambusia gaigei, Indian Territory, and Texas, in Septem- an endangered fish species. Southwest. ber, 1884, with notes and description. Nat. 8(1): 46-48. Proc. U.S. Natl. Mus. 9: 1-25. Hubbs, C., R. A. Kuehne, and J. C. Ball Krumholz, L. A. 1953. The fishes ofthe upper Guadalupe Riv- 1948. Reproduction in the western mosqui- er, Texas. Texas J. Sci. 5: 216-244. tofish, Gambusia affinis affinis (Baird 1989 RAUCHENBERGER: GAMBUSIA 69

and Girard), and its use in mosquito struction, Systematic Assoc. Special Vol. control. Ecol. Monogr. 18: 43 pp. 21: 21-74. London: Academic Press. Miller, R. R. Peden, A. E. 1975. Five new species of Mexican poeciliid 1972a. Differences in the external genitalia of fishes ofthe genera Poecilia, Gambusia, female gambusiin fishes. Southwest. Nat. and Poeciliopsis. Occas. Pap. Mus. Zool. 17(3): 265-272. Univ. Michigan 672: 44 pp. 1972b. The function of gonopodial parts and 1986. Composition and derivation of the behavioral pattern during copulation of freshwater fish fauna ofMexico. An. Esc. Gambusia. Can. J. Zool. 50(7): 955-968. Cienc. Biol. Mexico 30(1-4): 121-153. 1973a. Variation in anal spot expression of Miller, R. R., and W. L. Minckley gambusiin females and its effect on male 1970. Gambusia aurata, a new species ofpoe- courtship. Copeia 1973(2): 250-263. ciliid fish from northeastern Mexico. 1973b. Virtual extinction of Gambusia amis- Southwest. Nat. 15(2): 249-259. tadensis n. sp., a poeciliid fish from Tex- Minckley, W. L. as. Copeia 1973(2): 210-221. 1962. Two new species of fishes of the genus Poey, F. Gambusia (Poeciliidae) from north- 1854. Los guajacones, pecesillos de aqua dulce. eastern Mexico. Copeia 1962(2): 391- In F. Poey, Memorias sobre la historia 396. natural de la Isla de Cuba. Havana, 1: 1963. A new poeciliid fish (genus Gambusia) 374-392. from the Rio Grande drainage of Coa- Ramaswami, L. S. huila, Mexico. Southwest. Nat. 8(3): 1945. The chondrocranium of Gambusia (Cy- 154-161. prinodontes) with an account of the os- Myers, G. S. teocranium of the adult. Half-yearly J. 1935. An annotated list of the cyprinodont Mysore Univ., Sec. B, 6(1): 273-299. fishes of Hispaniola, with descriptions Regan, C. T. of two new species. Zoologica 10(3): 1913. A revision of the cyprinodont fishes of 301-316. the subfamily Poeciliinae. Proc. Zool. 1937. Fresh-water fishes and West Indian zoo- Soc. London 11: 977-1018. geography. Ann. Rep. Smithsonian Inst. 1914. Description of two new cyprinodont 1937: 339-364. fishes from Mexico, presented to the Nelson, G. British Museum by Herr A. Rachow. 1969. Gill arches and the phylogeny of fishes, Annu. Mag. Nat. Hist. 8(14): 65-67. with notes on the classification of ver- Rivas, L. R. tebrates. Bull. Am. Mus. Nat. Hist. 1944. Contributions to the study of the poe- 141(4): 475-552. ciliid fishes of Cuba. I. Descriptions of 1973. Relationships of clupeomorphs, with six new species of the subfamily Gam- remarks on the structure of the lower busiinae. Proc. New England Zool. Club jaw in fishes. In P. H. Greenwood et al. 23: 41-53. (eds.), Interrelationships of fishes, 333- 1963. Subgenera and species groups in the 349. London: Academic Press. poeciliid fish genus Gambusia Poey. Nelson, G., and N. Platnick Copeia 1963(2): 331-347. 1981. Systematics and biogeography. New 1969. A revision of the poeciliid fishes of the York: Columbia Univ. Press. Gambusia punctata species group, with Parenti, L. R. descriptions oftwo new species. Copeia 1981. A phylogenetic and biogeographic anal- 1969(4): 778-795. ysis ofcyprinodontiform fishes (Teleos- 1971. A new subspecies of poeciliid fish of tei, Atherinomorpha). Bull. Am. Mus. the genus Gambusia from eastern Nat. Hist. 168(4): 335-557. Cuba. Publ. Gulf Coast Res. Lab. Mus. 1984. A taxonomic revision ofthe Andean kil- 2: 5-9. lifish genus Orestias (Cyprinodonti- Rosen, D. E. formes, Cyprinodontidae). Bull. Am. 1964. The relationships and taxonomic posi- Mus. Nat. Hist. 178(2): 107-214. tion of the halfbeaks, killifishes, silver- Patterson, C. sides, and their relatives. Bull. Am. Mus. 1982. Morphological characters and homol- Nat. Hist. 127(5): 217-268. ogy. In K. A. Joysey and A. E. Friday 1967. New poeciliid fishes from Guatemala, (eds.), Problems of phylogenetic recon- with comments on the origins of some 70 AMERICAN MUSEUM NOVITATES NO. 2951

South and Central American forms. Am. Smith, M. L., and R. R. Miller Mus. Novitates 2303: 15 pp. 1986. The evolution ofthe Rio Grande Basin 1975. A vicariance model of Caribbean bio- as inferred from its fish fauna. In C. H. geography. Syst. Zool. 24(4): 431-464. Hocutt and E. 0. Wiley (eds.), The zoo- 1979. Fishes from the uplands and intermon- geography of North American fresh- tane basins of Guatemala: revisionary water fishes, 487-517. New York: Wi- studies and comparative geography. ley. Bull. Am. Mus. Nat. Hist. 162(5): 267- Turner, C. L. 376. 1940. Superfetation in viviparous cyprino- 1985. Geological hierarchies and biogeo- dont fishes. Copeia 1940(2): 88-91. graphic congruence in the Caribbean. 1941. Morphogenesis of the gonopodium in Ann. Missouri Bot. Gard. 72: 636-659. Gambusia affinis affinis. J. Morphol. Rosen, D. E., and R. M. Bailey 69(1): 161-186. 1959. Middle-American poeciliid fishes ofthe 1942a. A quantitative study of the effects of genera Carlhubbsia and Phallichthys, different concentrations of ethynyl tes- with descriptions of two new species. tosterone and methyl testosterone in the Zoologica 44(1): 1-44. production of gonopodia in females of 1963. The poeciliid fishes (Cyprinodonti- Gambusia affinis. Physiol. Zool. 15(3): formes), their structure, zoogeography, 262-279. and systematics. Bull. Am. Mus. Nat. 1942b. Morphogenesis of the gonopodial sus- Hist. 126(1): 1-176. pensorium in Gambusia affinis and the Rosen, D. E., and M. Gordon induction ofthe male suspensorial char- 1951. A new fish ofthe genus Gambusia from acters in the female by androgenic hor- southern Veracruz, Mexico, with a dis- mones. J. Exp. Zool. 91(2): 167-193. cussion ofthe tribe Gambusiini Hubbs. 1942c. Sexual dimorphism in the pectoral fin Zoologica 36(4): 267-272. of Gambusia and the induction of the 1953. Functional anatomy and evolution of male character in the female by andro- male genitalia in poeciliid fishes. Zoo- genic hormones. Biol. Bull. 83(3): 389- logica 38(1): 1-47. 400. Rosen, D. E., and K. D. Kallman Weisel, G. F. 1959. Development and evolution of skeletal 1967. Early ossification in the skeleton of the deletions in a family ofviviparous fishes sucker (Catostomus macrocheilus) and (Cyprinodontiformes, Poeciliidae). Q. the guppy (Poecilia reticulata). J. Mor- J. Florida Acad. Sci. 22(4): 169-190. phol. 121(1): 1-18. Rosen, D. E., and J. R. Mendelson Weitzman, S. H. 1960. The sensory canals of the head in poe- 1962. The osteology of Brycon meeki, a gen- ciliid fishes (Cyprinodontiformes), with eralized characid fish, with an osteolog- reference to dentitional types. Copeia ical definition of the family. Stanford 1960(3): 203-210. Ichthyol. Bull. 8(1): 77 pp. Rosen, D. E., and L. R. Parenti Wildrick, D. M., I-S Kim, and D. W. Greenfield 1981. Relationships ofOryzias, and the groups 1985. A unique gambusiine karyotype and its of atherinomorph fishes. Am. Mus. relevance to the systematics of the Novitates 2719: 25 pp. Gambusia punctata species group. Co- Rosen, D. E., and A. Tucker peia 1985(4): 1053-1056. 1961. Evolution of secondary sexual charac- Wooten, M. C., K. T. Scribner, and M. H. Smith ters and sexual behavior patterns in a 1988. Genetic variability and systematics of family of viviparous fishes (Cyprino- Gambusia in the southeastern United dontiformes, Poeciliidae). Copeia States. Copeia 1988(2): 283-289. 1961(2): 201-212. 1989 RAUCHENBERGER: GAMBUSIA 71

APPENDIX

In the following character list, each character is 38. Notch proximal to elbow. merely summarized and abbreviated here; the dis- 39. Long ray 3 spines, with long inner processes. cussion ofthe full nature ofeach character and its 40. Multisegmented elbow (to 7 segments). level ofgenerality is found in the text. Most char- 41. Small bifid ray 3 spines. acters listed are scored absent = 0, present = 1. 42. Lateral projection on elbow. Scoring for multistate characters is given in the 43. Serrae on ray 4p: proximal to elbow = 0; op- list. posite elbow = 1; distal to elbow = 2. 44. Ray 3 foreshortened. 45. Ray 4a foreshortened. 1. Gonactinost 5 stands freely. 46. Eight or fewer ray 3 spines. 2. Gonactinosts 2, 3, and 4 fuse into a column. 47. Gonopodial outline: median pointed = 0; 3. Axillary blotch. rounded = 1; lobelike = 2 within subgenus 4. Female anal pigment. Gambusia; pointed at anterolateral edge = 2 5. Hook on ray 5a. within subgenus Arthrophallus. 6. Ray 4a with coalesced segments, bent ante- 48. Transversely enlarged segment(s) distal to riorly. serrae on ray 4p = 1; same segments greatly 7. Male pectoral fin modification. enlarged, filling gaps between ray 4p and 5a 8. Thickened ray 2, proximally. = 2. 9. Bump in ray 5 proximally. 49. Electrophoretic characters; red and yellow 10. No uncinate process on gonapophysis 1. caudal and dorsal fins (Greenfield et al., 1982). 11. Elbow in ray 4a. 50. High meristic characters (Fink, 197 la). 12. Hooks on rays 4p and Sa pointing upwards. 51. Ray Sa hook large, scythelike. 13. IPB3 teeth with medial serrate pad. 52. Ray 4p hook large, scythelike. 14. Disconnected infraorbital sensory canals. 53. Body with 4-6 rows of large spots. 15. Elbow shape: falcate = 0; triangular = 1; spat- 54. One pair metacentric and two pairs submeta- ulate = 2. centric chromosomes in both sexes. 16. No uncinate process on EP1. 55. Ray 4a S-shaped, curving back over ray 3 17. Lateral prong on HB 1. spines. 18. Wings on BB3. 56. Pectoral ray 5 with distal scallop. 19. Shape ofVHH: flat = 0; with central processes 57. Pectoral ray 2 with enlargedblade proximally. = 1; triangular with central processes = 2. 58. Pectoral rays with both dorsal and ventral 20. Hypurals: unfused = 0; fused, with keyhole blades. = 1; fused completely = 2. 59. One pair of large metacentric chromosomes 21. Spotted fin patterns. in female only. 22. Suborbital bar. HB no = = 60. Shape ofmedial part of 1: square, cleft 23. Ray 4p hook: round 0; acuminate 1; mul- = 0; with shallow cleft not split = 1; deep tisegmented acuminate = 2. = = = = cleft, not split 2; forked, widely split 3. 24. Anal spots: none 0; median 1; paired 61. Number of gonapophyses: three = 0; some- crescents = 2. times or always two = 1. 25. Retrorse spines on ray 3. 62. Exaggerated distal point of gonopodium. 26. High number (more than 5) ofsegments distal 63. Serrated segments on ray 4a distal to elbow. to the elbow on ray 4a. 27. High number (up to 9 or more) of segments distal to the serrae on ray 4p. 28. High number (16 or more) of spines on The following character matrices were used as ray 3. PAUP input. The analyses were performed on an 29. Ray 3 spines fanlike, not pointed. IBM-PC, which has a limit of 100 equally parsi- 30. Terminal segment of ray Sa with long base. monious trees. Because of the lack of resolution 31. High number (8 or more) of serrae on ray 4p. in several areas of the cladogram, this limit was 32. Serrae on ray 4p: pointing proximally = 0; quickly exceeded when all taxa and all characters perpendicular = 1; reaching distally and re- were run together. Therefore, I chose to subdivide curved = 2. the data set into the following matrices, using only 33. No forward-directed spines on ray 3. the characters relative to that particular level and 34. Buckling in ray 4a, with displaced elbow. group. The following options were used: MISS- 35. Lateral palp on gonopodium. ING = 9, SW = GL, MULPARS, MAX- 36. Body with ventral keel. TREE = 100, ROOT = OUTGROUP, OPT = 37. Elongate cleithral process. FARRIS. 72 AMERICAN MUSEUM NOVITATES NO. 2951

1. For figure 21. Outgroup relationships.

CHAR. NO. 1 2 3 4 5 6 OUTGROUP 0 0 0 0 0 0 BRACHYRH 1 0 1 1 0 0 BELONESO 1 1 1 0 1 1 GAMBUSIA 1 1 1 1 1 1 One tree was found. Length = 7.000 Consistency Index = 0.857 2. For figure 30. Subgeneric relationships. CHAR. NO. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 OUTGROUP 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 HETEROPH 1 1 1 1 1 1 1 0 0 1 0 0 0 0 0 0 1 ARTHROPH 1 1 1 1 1 0 0 1 0 0 0 0 0 1 1 1 2 GAMBUSIA 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 One tree was found. Length = 20.000 Consistency Index = 0.950 3. For figure 32. Relationships within the subgenus Heterophallina. CHAR. NO. 12 13 20 21 25 26 27 28 29 30 31 32 33 34 35 36 37 OUTGROUP 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 VITTATA 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 MARSHI 1 1 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 PANUCO 1 0 0 1 1 1 1 1 1 1 1 1 0 0 0 0 0 REGANI 1 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 RACHOWI 1 1 1 0 1 1 1 0 0 0 0 1 1 1 1 1 1 ECHEAGAY 1 0 1 0 1 1 1 0 0 0 0 1 1 1 1 1 1 One tree was found. Length = 20.000 Consistency Index = 0.850 1989 RAUCHENBERGER: GAMBUSIA 73

4. For figures 35, 36, 40, and 42. Relationships within the subgenus Arthrophallus.

CHAR. NO. 14 16 17 18 20 21 22 23 24 30 31 32 38 39 40 41 42 43 44 45 46 47 48 57 60 61 62 OUTGROUP 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 AFFINIS 1 0 0 0 1 1 1 2 2 0 0 0 0 0 1 1 1 0 0 0 0 0 0 1 2 0 1 HOLBROOK 1 0 0 0 1 1 1 2 2 0 0 0 0 0 1 1 1 0 0 0 0 0 0 1 1 0 1 SPECIOSA 1 0 0 0 0 1 0 2 9 0 0 1 0 0 1 1 1 0 0 0 0 0 0 1 3 0 0 AURATA 1 0 1 0 1 1 1 2 9 1 0 0 0 0 1 1 0 2 1 0 0 0 0 1 3 1 0 LEMAITRE 1 0 9 0 9 1 9 2 9 1 0 0 0 0 1 1 0 2 1 0 0 0 0 9 9 9 0 NOBILIS 1 1 0 0 0 0 1 0 2 1 0 0 1 1 0 0 0 1 0 0 0 1 0 0 1 1 0 GEORGEI 1 1 0 0 0 0 1 1 2 0 0 1 1 1 0 0 0 1 0 0 0 1 0 0 0 1 0 KRUMHOLZ 0 1 0 0 0 0 0 0 9 1 0 1 1 1 0 0 0 1 0 0 0 1 0 0 3 1 0 HETEROCH 1 9 0 1 9 0 1 1 2 1 2 1 1 1 0 0 0 2 0 0 0 1 0 0 9 9 0 EURYSTOM 0 9 0 0 0 1 1 1 9 0 0 2 1 1 0 0 0 1 0 0 0 1 0 0 0 0 0 SEXRADIA 0 1 0 0 0 1 1 0 9 0 0 2 1 1 0 0 0 1 0 0 0 1 0 0 2 0 0 SENILIS 1 1 0 0 1 0 1 0 1 1 0 0 1 1 0 0 0 0 0 1 1 2 1 0 1 1 0 GAIGEI 1 1 0 0 9 0 1 0 1 1 0 0 1 1 0 0 0 0 0 1 1 2 1 0 3 1 0 GEISERI 1 1 0 0 2 0 0 1 2 1 1 0 1 1 0 0 0 0 0 1 1 2 0 0 3 1 0 ALVAREZI 1 1 0 0 1 0 1 0 1 0 0 0 1 1 0 0 0 0 0 1 1 2 1 0 0 1 0 HURTADOI 1 1 0 0 1 0 1 0 1 0 0 0 1 1 0 0 1 0 0 1 1 2 1 0 0 1 0 LONGISPI 1 1 0 0 2 0 0 0 1 1 1 0 1 1 0 0 1 0 0 0 1 2 0 0 1 0 0 AMISTADE 1 1 0 0 1 0 1 0 1 0 0 0 1 1 0 0 1 0 0 0 1 2 0 0 1 1 0 ATRORA 1 1 0 0 2 0 0 0 2 0 1 0 1 1 0 0 0 0 0 1 1 2 1 0 0 0 0 15 trees were found. Length = 7 1.000 Consistency Index = 0.507 Figures 35, 36, 40, and 42 represent parts of the strict consensus tree of the 15 trees found. 74 AMERICAN MUSEUM NOVITATES NO. 2951

5. For figures 45, 46, 48, and 50. Relationships within the subgenus Gambusia.

CHAR. NO. 15 16 17 18 19 20 21 22 23 24 30 32 39 42 47 48 49 50 51 52 53 54 55 59 60 61 63 OUTGROUP 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 NICARAGU 1 0 0 1 2 0 1 1 2 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 WRAYI 1 1 1 1 2 0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 1 MELAPLEU 1 9 1 1 2 9 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 9 0 1 YUCATANA 1 0 1 1 0 0 1 1 1 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 3 0 0 HISPANIO 1 0 1 1 1 1 1 0 1 0 0 1 0 1 1 1 1 0 0 0 0 0 0 1 1 0 0 MANNI 1 9 1 1 1 1 1 0 1 0 0 1 0 1 1 1 1 0 0 0 0 0 0 1 9 0 0 HUBBSI 1 11 1 1 1 1 1 1 0 0 1 1 1 1 1 1 0 0 0 0 0 0 1 1 1 0 MONTICOL 1 9 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 0 0 0 1 9 0 0 BUCHERI 1 9 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 0 0 0 1 1 1 0 BARACOAN 1 9 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 0 0 0 1 9 0 0 PUNCTICU 1 1 1 1 1 0 1 1 2 1 0 1 1 1 1 1 1 0 0 0 0 0 0 1 2 1 0 OLIGOSTI 1 1 1 1 1 0 1 1 2 1 0 1 1 1 1 1 1 0 0 0 0 0 0 1 9 1 0 CAYMANEN 1 9 1 1 1 0 1 1 2 1 0 1 1 1 1 1 1 0 0 0 0 0 0 1 9 1 0 HOWELLI 1 9 1 1 9 0 1 1 2 1 0 1 1 1 1 1 1 1 0 0 0 0 0 1 9 1 0 LUMA 2 0 1 0 0 0 1 1 0 0 0 2 1 0 2 0 0 0 1 1 0 0 0 0 1 0 0 BEEBEI 2 0 1 1 1 0 1 0 0 0 1 1 1 0 2 9 0 0 1 1 0 9 0 0 9 0 0 PSEUDOPU 2 0 1 1 1 0 1 0 0 0 1 1 1 0 2 1 0 0 1 1 1 9 0 0 3 0 0 PUNCTATA 2 0 1 1 1 0 1 0 0 0 0 1 1 0 2 2 0 0 1 1 1 9 0 0 2 0 0 RHIZOPHO 2 1 1 0 1 0 1 0 0 0 0 2 1 0 2 2 0 0 1 1 1 2 1 0 2 0 0 XANTHOSO 2 9 1 9 1 0 1 0 0 0 0 1 1 0 2 2 0 0 1 1 1 2 1 0 9 0 0 For relationships among species groups in the subgenus, and for relationships within the nicaraguensis and punctata species groups, G. oligosticta, G. caymanensis, and G. howelli were dropped (except for missing data, these taxa are coded identically to G. puncticulata, and to each other). This produced 80 trees. Length = 62.000 Consistency Index = 0.613 Figures 45, 46, and 50 are parts of a strict consensus tree of the 80 equally parsimonious trees found. For relationships within the puncticulata species group, the deleted taxa were restored and members of other species groups deleted. This produced 45 trees. Length = 32.000 Consistency Index = 0.812 Figure 48 is a strict consensus tree of the 45 equally parsimonious trees found.

Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates, Bulletin, and Anthropological Papers published during the last five years are free of charge. available Address orders to: American Museum of Natural History Library, Department D, Central Park West at 79th St., New York, N.Y. 10024.

THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER. a