Systematics of the subfamily Poeciliinae Bonaparte ( Cyprinodontiformes : Poeciliidae ) , with an emphasis on the tribe Cnesterodontini Hubbs

* Laboratório de Ictiologia Sistemática, Universidade Federal do Tocantins, Campus de Porto Nacional, rua 3, Quadra 17, s/n, Caixa Postal 136, 77500-000 Porto Nacional, TO, Brazil. e-mail:lucinda@uft.edu.br ** Laboratório de Ictiologia, Pontifícia Universidade Católica do Rio Grande do Sul. Av. Ipiranga, 6681, Caixa Postal 1429, 90619-900 Porto Alegre, RS, Brazil. e-mail: reis@pucrs.br Systematics of the subfamily Poeciliinae Bonaparte (Cyprinodontiformes: Poeciliidae), with an emphasis on the tribe Cnesterodontini Hubbs


Introduction
This paper is resultant from a project that intended to perform the taxonomic revision of the tribe Cnesterodontini, as well as to propose a phylogenetic hypothesis of relationships among its members. Testing the monophyly of the Cnesterodontini was the first step attempting to achieve our aims. Thus, the original project was broadened to embrace a phylogenetic hypothesis of relationships and the proposal of a provisional classification for the subfamily Poeciliinae. The resultant hypothesis included representatives of all poeciliine genera, and all described Cnesterodontini species, as well as 24 new species revealed by the taxonomic revisions of genera Cnesterodon Garman, Phallotorynus Henn, and Phalloceros Eigenmann. Intrageneric relationships of Phalloptychus Eigenmann, Cnesterodon, Phalloceros, and Phallotorynus are provided herein, but will be discussed in the aforementioned taxonomic revisions.

Nomenclatural and Taxonomic History
Poeciliinae. The subfamily Poeciliinae is a cyprinodontiform group widely distributed throughout the Americas. Poeciliinae is the sister group of the Procatopodinae, a group composed of the South-American Fluviphylax Whitley and the African procatopodines. The clade Poeciliinae plus Procatopodinae is the sister group of the Aplocheilichthyinae (Costa, 1996;Ghedotti, 2000). These three subfamilies compose the family Poeciliidae. The Poeciliinae embraces approximately two hundred twenty species currently allocated in approximately twenty-eight genera (Lucinda, 2003). Theses fishes are characterized by the uniquely derived possession of a gonopodium formed by the modified male anal-fin rays 3, 4, and 5 (Parenti, 1981).
The Poeciliinae includes well-known aquarium fishes such as the guppies, mosquito fishes, swordtails, platys, and mollies. Poeciliines are well known subjects of study for ecologists, anatomists, embryologists, and other research- ers. Notwithstanding, this fish assemblage is only superficially studied from the perspective of systematics. Intrageneric diversity and intergeneric relationships of the Poeciliinae are poorly known, regardless of its huge distribution, range, and notoriety. Similarly, phylogenetic hypotheses for most genera are still wanting. Taxonomic revisions and phylogenetic analyses have provided some insight into the relationships of smaller groups of the Poeciliinae (e.g., Rosen, 1967Rosen, , 1979Rauchenberger, 1989;Rosa & Costa, 1993;Meyer et al., 1994;Mojica et al., 1997;Rodriguez, 1997;Poeser, 2003) or have dealt with higher taxa (Rosen, 1964;Parenti, 1981;Costa, 1996Costa, , 1998Ghedotti, 2000). The only comprehensive study is the classic revision of "Poeciliidae" by Rosen & Bailey (1963), which did not deal with cladistic methodology. Nonetheless, Rosen & Bailey (1963) is the basis for the current internal classification of Poeciliinae. Later, Parenti & Rauchenberger (1989) modified the classification of Rosen & Bailey (1963) in order to accommodate it into the taxonomic rank of subfamily proposed by Parenti (1981) (Table  1). Following Rosen & Bailey (1963) and Parenti & Rauchenberger (1989), Tomeurus Eigenmann alone is the supertribe Tomeurini. The remaining genera form the supertribe Poeciliini, which is subdivided in the tribes Poeciliini, Cnesterodontini, Gambusiini, Scolichthyini, Girardinini, Heterandriini, and Xenodexini. Later, Ghedotti (2000) proposed another classification for the Poeciliinae ( Table 2) based in his phylogenetic study of the Poecilioidea despite the fact that only 12 genera were examined in his cladistic analysis.
The history of the subfamily began in 1801 with the description of Poecilia vivipara Bloch & Schneider as new genus and new species. Before establishment as a distinct family, the history of the Poeciliidae is merged with that of other cyprinodontiform families and with the Cyprinidae. Table 1. Classification of Poeciliinae Bailey, 1963 andParenti &Rauchenberger, 1989).
Howell Rivero & Hubbs (1936) recognized Alfaro as distinct from both Tomeurus and the Poeciliinae and classified this genus it in its own subfamily, Alfarinae. This taxonomic decision was supported by Rosen (1952) and Rosen & Gordon (1953).
From 1940 Rosen (1967) erected the poeciliid genus Scolichthys Rosen. Rivas (1980) removed Limia from the synonym of Poecilia and erected the subgenus Odontolimia Rivas, splitting Limia in two subgenera, L. (Limia) and L. (Odontolimia). Poeser (2002) created the monotypic genus Pseudolimia Poeser for Limia heterandria Regan. Some taxonomic and phylogenetic studies have provided some progress into the relationships of smaller groups of the Poeciliinae. The genera Heterandria and Xiphophorus were reviewed by Rosen (1979), a classic paper concerning methods of biogeographical analysis. Rauchenberger (1989) put forward hypotheses of systematic and biogeographic relationships among the species of the genus Gambusia. Rosa & Costa (1993) made a taxonomic revision of the genus Cnesterodon, describing two new species and proposing some putative synapomorphies for Cnesterodon, in the absence of a phylogenetic analysis. The phylogenetic relationships of Xiphophorus species have been surveyed by Rosen (1979), Rauchenberger et al. (1990), Meyer et al. (1994), Marcus & McCune (1999) and Kallmann et al. (2004). Mojica et al. (1997) proposed a hypothesis of relationships among Brachyrhaphis species on the basis on mitochondrial DNA evidence. Rodriguez (1997) studied the relationships among genera of the tribe Poeciliini sensu Rosen & Bailey, re-defining the tribe Poeciliini as comprehending the genera Alfaro, Priapella, Xiphophorus, Poecilia, Limia and Pamphorichthys. Ptaceck & Breden (1998) proposed a molecular phylogeny for Poecilia, focusing on the species of the subgenus Mollienesia. Breden et al. (1999) performed a phylogenetic analysis for part of the species of the genus Poecilia based on mitochondrial DNA evidence. Hamilton

P R O O
F S (2001) proposed a phylogenetic hypothesis for Limia species based on the mitochondrial genes sequences. Mateos et al. (2002) proposed a historical biogeography hypothesis for Poeciliopsis species based on sequence variation in two mitochondrial genes. Poeser (2003) carried out a taxonomic revision of Poecilia and proposed a phylogenetic hypothesis for this genus. Recent phylogenetic studies have dealt with higher taxa but none has tackle the relationships among members of the subfamily Poeciliinae as a whole. The only comprehensive study is Rosen & Bailey (1963), but these authors proposed a non-cladistic classification. In 1963, Rosen and Bailey published their classic revision of poeciliid fishes. This work separate the Poeciliidae into tree subfamilies, Tomeurinae for Tomeurus, Xenodexiinae for Xenodexia, and Poeciliinae for all other poeciliids including Alfaro, which was recognized in the tribe Poeciliini. Parenti (1981) recognized the Poeciliidae of Rosen & Bailey as a subfamily (Poeciliinae) within a more inclusive family (Poeciliidae) including the oviparous Aplocheilichthyinae and Fluviphylacinae. Parenti & Rauchenberger (1989) modified the classification of Rosen & Bailey (1963) to reflect the change in taxonomic rank proposed by Parenti (1981) (Table  1). Meyer e Lydeard (1993) put forward a molecular phylogeny for Cyprinodontiformes including four poeciliine genera. Costa (1996) presented new evidence concerning the monophyly of the subfamilies of Poeciliidae, and proposed a hypothesis of phylogenetic interrelationships among them. Costa (1998) put forward a new phylogenetic framework for the Cyprinodontiformes, differing from Parenti's (1981) hypothesis. Nonetheless, Parenti (1981) and Costa (1996Costa ( , 1998 did not tackle the inter-and intrageneric relationships. Recently, Ghedotti (2000) recognized the monophyly of the family Poeciliidae, with three monophyletic subfamilies: (1) the Aplocheilichthyinae containing solely Aplocheilichthys spilauchen, (2) the Procatopodinae containing Fluviphylax (Fluviphylacini), and the African lamp-eyed killifishes (Procatopodini), and (3) the Poeciliinae. He also resurrected the tribe Alfarini and proposed a new tribe, the Priapellini for Priapella.
Cnesterodontini. The tribe Cnesterodontini as originally erected by Hubbs (1924) was composed of genera Phalloceros, Cnesterodon, Phallotorynus, and Diphyacantha Henn. The Cnesterodontins were defined as poeciliines bearing "terminal segment of ray 3 forming a more or less specialized process" (Hubbs, 1924: 9). Hubbs (1926) added Darienichthys to the Cnesterodontini. Later, Rosen & Bailey (1963) recognized Diphyacantha and Darienichthys as junior synonyms of Priapichthtys and removed them from the Cnesterodontini, placing it in the tribe Heterandriini. Rosen & Bailey (1963) also added Phalloptychus to the Cnesterodontini. More recently, Ghedotti (2000) based on his phylogenetic study of the Poeciloidea recognized Tomeurus as a member of the tribe Cnesterodontini and provided a diagnosis for the group anchored in unique and unreversed synapomorphies. Thus, as currently defined the Cnesterodontini comprises five genera: Cnesterodon, Phalloceros, Phallotorynus, Phalloptychus, and Tomeurus. Cnesterodontines are disappointingly ill-studied from the perspective of systematics. Except for Cnesterodon, this group of fishes has received very little attention. The history of the genus Cnesterodon began with Jenyns' (1842) description of Poecilia decemmaculata Jenyns, the first described species currently placed in the genus. The genus Cnesterodon was erected by Garman, with Poecilia decemmaculata as typespecies, for it differed from the remaining genera so far assigned to Poeciliinae: Garman in the same paper also described a second species for the genus: C. scalpridens Garman. A third nominal species, C. carnegiei Haseman was described from the rio Iguaçu drainage. Regan (1913) removed C. scalpridens from Cnesterodon and erected the genus Pamphoria [= Pamphorichthys] for this species. Rosa & Costa (1993) recognized the validity of C. decemmaculatus and C. carnegiei, and described C. brevirostratus Rosa & Costa from the upper rio Uruguay and rio Jacuí drainages, and C. septentrionalis Rosa & Costa from the rio Araguaia drainage. Rosa & Costa (1993) also reported nine putative synapomorphies for the genus. Latter, Lucinda & Garavello (2001) described C. hypselurus Lucinda & Garavello from the rio Paranapanema basin, and C. omorgmatos Lucinda & Garavello, a second species from the rio Iguaçu basin. Cnesterodon raddai Meyer & Etzel was described from the lower portions of rio Paraná.
Studies concerning Phallotorynus species are notably scarce, and are mostly confined to original descriptions. Henn (1916) erected the genus Phallotorynus, based on gonopodium structure, for his new species, P. fasciolatus Henn, from the rio Paraíba do Sul. Later, Ihering (1930) described P. jucundus Ihering from a tributary of rio Mogi-Guaçu in the upper rio Paraná drainage. Rosen & Bailey (1963) redefined the genus by osteological characters on the basis of specimens of P. fasciolatus and specimens from the neighborhood of Asunción, expanding the distribution range for P. jucundus to the Paraguay drainage. Finally, Oliveros (1983) described P. victoriae Oliveros from the lower portions of rio Paraná basin in Argentina.
Papers concerning Phalloptychus are also extremely rare in systematic literature, being confined to original descriptions. The history of Phalloptychus began with the first described species currently in the genus: Girardinus januarius Hensel. A second species, G. iheringii Boulenger was described from Rio Grande do Sul. Eigenmann (1907) created the genus Phalloptychus for Girardinus januarius. Henn (1916) described a third species, P. eigenmanni, from the rio Catu at Alagoinhas, Bahia.
Phalloceros and Tomeurus are monotypic genera and disappointingly ill-studied from the perspective of systematics. Both genera were erected by Eigenmann (1907 and1909, respectively) for Girardinus caudimaculatus Hensel and Tomeurus gracilis, respectively.
Clearing and staining followed the method of Taylor & Van Dyke (1985). Anatomical illustrations were prepared from sketches of structures from cleared and stained specimens as viewed through a camera lucida mounted on a dissecting stereomicroscope. External characters, e.g. color pattern, were also examined.
Proposed hypotheses of phylogenetic relationships among studied taxa followed the phylogenetic method formally put forward by Hennig (1966). The ingroup included representatives of all poeciliine genera, and all species of the tribe Cnesterodontini sensu Rosen & Bailey (1963). The data matrix of 71 taxa and 144 characters (Appendix II) includes 24 new species of Cnesterodontini, whose descriptions will be provided by Lucinda (in prep.) and Lucinda et al. (in prep.). Intrageneric relationships and synapomorphy lists for subclades of Phalloptychus, Cnesterodon, Phalloceros, and Phallotorynus are provided herein, but are discussed in Lucinda (in prep.) and Lucinda et al. (in prep.). Question marks were used to indicate when a character state could not be checked due to lacking of available specimens. Dashes were employed for both inapplicable coding and polymorphisms. The phylogenetic analysis aimed to test the monophyly of the subfamily Poeciliinae as well and its tribes (sensu Rosen & Bailey, 1963). Fundulus heteroclitus (Linnaeus), Cyprinodon macularius Baird & Girard, Jenynsia unitaenia Ghedotti & Weitzman, Aplocheilichthys spilauchen (Duméril), Fluviphylax pygmaeus (Myers & Carvalho), and Procatopus gracilis Clausen were included as outgroup taxa. Phylogenetic analyses included all 71 taxa simultaneously and were performed with Hennig86 (Farris, 1988) coupled with Tree Gardener (Ramos, 1997). Examination of more specimens suggested that there were problems with homology concerning some characters of Ghedotti's (2000) analysis (e.g. fusion of the dorsal-most proximal pectoral radial to the scapula) or these characters states could not be confirmed on existing specimens. Therefore these characters were excluded from the analysis.
All transformation series were considered unordered. Maximum parsimony analyses were undertaken using the mh*; bb* algorithm of Hennig86. Character optimization followed accelerated transformation model (ACCTRAN) for it is more consistent with the concepts of homology and synapomorphy (de Pinna, 1991). The numbers on the branches of the strict consensus tree (Fig. 1, 2, and 3) corresponds to tree nodes and to clade number. In the diagnoses and synapomorphy list uniquely derived and unreversed features are indicated by two asterisks (e.g. 47-2**); uniquely derived features are indicated by one asterisk (e.g. 24-1*). An asterisk indicates uniquely derived autapomorphies. Transformation series analysis (TSA) is presented in Appendix III. Fits of individual characters are shown in Appendix IV.
Long epiotic process extending beyond first pleural rib is present in Fundulus, Jenynsia, and Aplocheilichthys and is hypothesized as plesiomorphic (state 0). Among poeciliines, enlarged epiotic processes are present in Alfaro, Phallichthys, Xenophallus, Poeciliopsis, Quintana, and Carlhubbsia. Parenti (1981) hypothesized enlarged epiotic processes as a synapomorphy for anablepids. However, Ghedotti (1998) did not recognize expanded epiotic processes as uniquely synapomorphic of the Anablepidae and recorded their presence in Anableps, Oxyzygonectes Fowler, three species of Jenynsia, and Aplocheilichthys spilauchen. Ghedotti (2000) also observed the presence of long epiotic processes in Fundulus chrysotus Günther as well as in some poeciliines.
Following the hypothesis presented here state [3-1] appeared once among poeciliines: in the ancestor of members of Clade 125. Condition [3-2] appeared independently four times in Heterandria, Pseudopoecilia, Xenodexia, and "Poecilia" reticulata, whereas the loss of epiotic processes occurred independently in Tomeurus, Priapella, Phalloptychus, in the ancestor of Clade 92, and in cnesterodontines.
In most cyprinodontiforms halves of supraoccipital process are simple (state 0; Fig. 4a). Girardinus and Pamphorichthys possess bifid halves of supraoccipital process with a minute external half (state 1; Fig. 4b), which is interpreted as independently acquired. Cnesterodontines possess bifid halves of supraoccipital process with a large external half (state 2; Fig. 4c). A reversal to state 0 occurs in Phallotorynus fasciolatus.

Cephalic sensory system
Nomenclature follows Gosline (1949) and Rosen & Mendelson (1960). We refer the reader to figures depicted in Gosline (1949) and Rosen & Mendelson (1960) and in the articles cited below for a detailed comprehension of characters 5 to 10.
Posterior supraorbital canal varies among members of the outgroup and among members of the ingroup. In Aplocheilichthys, Fluviphylax, and poeciliines, except Alfaro, Brachyrhaphis, Priapichthys, and Priapella, anterior section of posterior remnant of infraorbital system is lacking or is opened, forming a shallow groove (state 0). Anterior section of posterior remnant of infraorbital system forms a major sinuous depression above and slightly behind the orbit (state 1) in Procatopus, Alfaro, Brachyrhaphis, Priapichthys, and Priapella. Anterior section of posterior remnant of infraorbital system is closed (state 2) in Jenynsia, Cyprinodon, and Fundulus.
Most cyprinodontiforms possess a closed posterior section of posterior remnant of infraorbital system (state 0). Posterior section of posterior remnant of infraorbital system opened into a groove (state 1) is herein hypothesized as apomorphic and to have been independently acquired by Tomeurus, Gambusia + Belonesox [clade 118], Scolichthys, Phalloptychus, Pamphorichthys, and cnesterodontines (with a reversal in Phallotorynus).
Among studied taxa, mandibular canal is entirely closed; bearing four pores (state 0) in Aplocheilichthys, Jenynsia, Cyprinodon, and Fundulus. Mandibular canal is absent or opened, forming a very shallow groove (state 1) in Fluviphylax, Tomeurus, and members of the supertribe Poeciliini with the exception of Girardinus, Xenodexia, and Poecilia. This condition is hypothesized to have been independently acquired in these groups. Procatopus and Priapella possess a mandibular canal entirely closed, bearing five pores (state 2) and the presence of this feature in these taxa is interpreted as homoplastic. Mandibular canal is partially closed bearing six pores (state 3) in Priapichthys. Mandibular canal is present and partially closed; canal connecting pores X and W closed; canal connecting pores Y and Z absent (state 4) in Gambusia. Mandibular canal is entirely closed bearing three pores, pore Z absent (state 5) in Girardinus. Mandibular canal is opened, formed by two deep grooves in Xenodexia (state 6). States 3, 4, 5, and 6 are interpreted as autapomorphic for Priapichthys, Gambusia, Girardinus, and Xenodexia, respectively.

Suspensorium and mandibular arch
Character 11 -Medial surface of ascending process of premaxilla (Ghedotti, 2000: fig. 3): (0) approximately straight; (1) slightly angled laterally; (2) angled laterally at proximal end, forming a triangle space between proximal ends of ascending processes. Ghedotti (2000) reported a laterally angled medial surface of the ascending processes as independently acquired by Micropanchax and the common ancestor of a clade composed of Cnesterodon, Phalloceros, Phallotorynus, Girardinus, Poecilia, Phallichthys, and Tomeurus with a reversal in Tomeurus. Ghedotti (2000) recognized only two character states whereas this study recognizes a third intermediate state (state 1).
Medial surface of ascending process of premaxilla is approximately straight (state 0) in all studied outgroup taxa and Scolichthys, Neoheterandria, Pseudopoecilia, Cnesterodon n. sp. A, C. hypselurus, C. brevirostratus, C. septentrionalis, and Phallotorynus fasciolatus. Alfaro, Brachyrhaphis, Priapichthys, Priapella, and Heterandria possess a slightly laterally angled medial surface (state 1). Medial surface of ascending process of premaxilla is laterally angled at proximal end, forming a triangular space between proximal ends of ascending processes (state 2) in the remaining studied taxa.
An elongate ascending process of premaxilla, with a pointed distal tip (state 1; Fig. 5b) is hypothesized as synapomorphic and independently acquired in the Priapichthyini and the Gambusiini (with a reversal to state 0 in Pseudopoecilia). A short and pointed ascending process of premaxilla (state 2; Fig. 5c) is interpreted as a uniquely derived and unreversed synapomorphy for the supertribe Poeciliini [Clade 119], with subsequent transformations to states 3 (Quintana and Phalloceros), 4 (Girardinus), and 5 (Cnesterodon n. sp. B).
The contact area between premaxillae is plain in almost all cyprinodontoids (state 0). An elevated contact area between premaxillae (state 1) is hypothesized as a synapomorphy for the members of the supertribe Poeciliini, with a reversal to the plesiomorphic condition in the Phallotorynus + Phalloceros The anterior border of ventral maxilla is straight (state 0; Fig. 6a) in outgroup taxa, as well as in Priapella, Gambusia, Belonesox, Neoheterandria, and Scolichthys and all members of the supertribe Poeciliini with the exception of Girardinus. Anterior border of ventral maxilla is concave (state 1; Fig. 6b Character 15 -Ventral surface of dentary (Costa, 1991: fig. 4G, 5E andGhedotti, 2000: fig. 5): (0) straight or bearing a tiny straight process; (1) bearing a curved and forward directed process.
Most cyprinodontiforms possess the ventral surface of dentary straight or bearing a tiny straight process (state 0). Costa (1991) suggested the presence of a curved and forward directed process on ventral surface of dentary (state 1) as a putative synapomorphy for a group embracing Pamphorichthys, Poecilia, Limia, Xiphophorus, Cnesterodon, Phalloceros, Phallotorynus, Phalloptychus, Priapichthys, Poeciliopsis, Priapella, Quintana, Carlhubbsia, Xenodexia, and Phallichthys. This is partially corroborated by our results. The current phylogenetic analysis supports this feature as a uniquely derived and unreversed synapomorphy for the supertribe Poeciliini [Clade 119]. In addition to the genera above (except Priapichthys and Priapella), this group comprises Girardinus, Xenophallus, and Micropoecilia.
Most poeciliids lacks a notch on dentary (state 0). A notch on dentary is present in Phallichthys, Poeciliopsis, Carlhubbsia, Xiphophorus, Xenodexia, Poecilia, Limia, and "Poecilia" reticulata. This feature is interpreted as apomorphic (state 1) and independently acquired by Phallichthys, Poeciliopsis and by the common and exclusive ancestor of members of the tribe Poeciliini, with a reversal in Pamphorichthys + Micropoecilia + "Poecilia" (Clade 92). A return to state 1 occurs in "Poecilia" reticulata. Character 18 -Ventral process of anguloarticular (Ghedotti, 2000: fig. 5): (0) long, extending anterior to where anguloarticular overlaps dentary; (1) short, not extending anterior to where anguloarticular overlaps dentary; (2) absent. Parenti (1981) recognized an elongate retroarticular as synapomorphic for the superfamily Poecilioidea. Costa (1998) modified the character state description used by Parenti (1981) to recognize the co-occurrence of a long retroarticular and a long ventral process of the anguloarticular as synapomorphic of Poecilioidea. Ghedotti (2000) treated the ventral process of the anguloarticular and the retroarticular as separate transformation series because they vary independently. Ghedotti (2000) reported a long ventral process of the anguloarticular in all poecilioids examined by him.
Among studied taxa, Aplocheilichthys, Jenynsia, Procatopus and almost all poeciliines the ventral process of anguloarticular is long, extending anterior to where anguloarticular overlaps dentary (state 0). This process is short, not extending anterior to where anguloarticular overlaps dentary (state 1) in Cyprinodon, Fundulus, Poeciliopsis, and Phalloptychus. State 1 is herein interpreted as synapomorphic for the clade [Poeciliopsis + Phalloptychus]. Its presence in Fundulus and Cyprinodon is considered homoplastic.
Ventral process of anguloarticular is absent (state 2) and interpreted as synapomorphic for the clade [Phalloceros n. sp. Among studied taxa, Phalloceros n. sp. B, Phalloceros n. sp. C, and Phalloceros n. sp. V possesses a ventral invagination on anguloarticular, which is interpreted as synapomorphic for Phalloceros n. sp. C + Phalloceros n. sp. V and independently acquired in Phalloceros n. sp. B.

Hyoid arch
Character 22 -Number of branchiostegal rays (Ghedotti, 2000: fig. 7): (0) five; (1) six. This character has been discussed by Ghedotti (2000). In individuals with six branchiostegal rays, the anterior two branchiostegal rays are in contact with the slender anterior portion of the anterior ceratohyal, three branchiostegal rays are in contact with the ventromedially expanded portion of the anterior ceratohyal, and the posterior branchiostegal ray is in contact with the posterior ceratohyal (state 1). In individuals with five branchiostegal rays, one of the anterior two branchiostegal rays is absent (state 0). The possession of six branchiostegal rays is interpreted as a synapomorphy for poeciliines with a reversal to plesiomorphic condition in the ancestor of members of the supertribe Poeciliini [Clade 119]. Inside this clade a change to state 1 occurred independently in Xenophallus and Limia.  Character 23 -First and second branchiostegal rays: (0) free from each other; (1) united at the base.
In most atherinomorphs first and second branchiostegal rays are free from each other (state 0). In Neoheterandria and Scolichthys first and second branchiostegal rays are united at the base (state 1). This condition is interpreted as synapomorphic for Neoheterandria + Scolichthys [Clade 114].
We observed this process in Aplocheilichthys, Jenynsia, Tomeurus, and Brachyrhaphis. All remaining studied taxa lack the anterior process of anterior ceratohyal extending ventral to ventral hypohyal (state 1). According to present hypothesis the absence of this process is synapomorphic for the subfamily Poeciliinae with reversals in Tomeurus, Brachyrhaphis, and Heterandria.
The interarcual cartilage is present (state 0) in all taxa examined, except in Tomeurus and Priapella. The absence of an interarcual cartilage (state 1) in these two taxa is interpreted as independently acquired.
Anterior margin of first hypobranchial is mostly straight (state 0) in Fluviphylax, Procatopus, and Fundulus. Anterior margin of first hypobranchial is concave, forming distinct anterolateral point (state 1) in Aplocheilichthys, Jenynsia, Cyprinodon, and all poeciliines. This feature was useless for poeciliine relationships and probably is a synapomorphy for a more inclusive clade. Ghedotti (2000) reported a concave anterior margin of first hypobranchial in Oxyzygonectes dovii, Cyprinodon variegatus, Fundulus chrysotus, and all poeciliines examined except some individuals of Tomeurus gracilis (coded as polymorphic).
The position of the pectoral fins is low, with dorsal insertion below midline in most cyprinodontiforms (state 0). In poeciliids, the position of the pectoral fins is high, with dorsal insertion at or above midline (state 1). With the exception of Jenynsia, Fundulus, and Cyprinodon, all studied taxa present high pectoral fins. This feature probably is synapomorphic for the family Poeciliidae. In fact, as pointed out by Ghedotti (2000: 25): " Parenti (1981) and Costa (1998) recognized high pectoral fins as a synapomorphic reversal in poeciliids to the condition in non-cyprinodontiform atherinomorphs".

P R O O F S
Males specimens of Phalloceros and Phallotorynus species as well as Cnesterodon n. sp. A, Cnesterodon n. sp. B, and C. hypselurus possess five pelvic-fin rays (state 1). Males of Cnesterodon brevirostratus, C. septentrionalis, C. carnegiei, and C. omorgmatos, Phalloptychus januarius and P. iheringii exhibit four pelvic-fin rays (state 2). Males of Cnesterodon raddai Meyer & Etzel and Tomeurus possess three pelvic-fin rays (state 3). Cnesterodon decemmaculatus (Jenyns) was coded "-" for it is polymorphic, males having four or five pelvic-fin rays. Phalloptychus eigenmanni Henn was coded "?" for character state could not be checked due to poor condition of the material studied. According to the present phylogenetic hypothesis state 1 is interpreted as synapomorphic for a clade containing Cnesterodon, Phallotorynus, and Phalloceros [Clade 111]. State 2 is hypothesized as independently acquired and synapomorphic for Phalloptychus species and for the clade [Cnesterodon brevirostratus + C. septentrionalis + C. carnegiei + C. omorgmatos]. The presence of state 3 in C. raddai and Tomeurus is considered homoplastic.
Character 34 -Pelvic-fin length in adult males: (0) short; second ray not surpassing the end of anal-fin base; (1) long, second ray surpassing the end of anal-fin base.
In most cyprinodontiform fishes the second pelvic-fin ray does not surpass the end of anal-fin base (state 0). In Xenophallus, Xiphophorus, Xenodexia, Poecilia, Limia, Pamphorichthys, Micropoecilia, and "Poecilia" the second pelvic-fin ray surpasses the end of anal-fin base (state 1). Following our phylogenetic study, second pelvic-fin ray surpassing the end of anal-fin base is hypothesized to have been independently acquired by Xenophallus and by the ancestor of a clade comprising Xiphophorus, Xenodexia, Poecilia, Limia, Pamphorichthys, Micropoecilia, and "Poecilia" [Clade 108]. Rodriguez (1997) reported this feature as synapomorphic for Xiphophorus, Poecilia, Limia, and Pamphorichthys. In Neoheterandria, Pamphorichthys hollandi, and Phalloceros species excepting Phalloceros n. sp. B, and Phalloceros n. sp. D pelvic girdle is not very anteriorly located, with posterior border of basipterygium aligned with posterior border of cleithrum (state 2). In Tomeurus and Cnesterodon the basipterygium is very anterior, located below pectoral girdle; posterior border of basipterygium anterior to posterior border of cleithrum (state 3). In Procatopus, Scolichthys, Xenophallus, and Limia posterior border of cleithrum is approximately aligned with anterior border of basipterygium (state 4). Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes. Parenti (1981), Costa (1998), and Ghedotti (2000) recognized an anteriorly positioned pelvic girdle as synapomorphic for poeciliids and also recognized the presence of an anterior pelvic girdle in cyprinodontids, and some individuals of Jenynsia. Ghedotti (2000) also described a pelvic girdle under the pectoral girdle in males as synapomorphic for Tomeurus, Cnesterodon, Phalloceros, and Phallotorynus. In most cyprinodontiform fishes a dorsolateral process of basipterygium in adult males is hardly developed or lacking (state 0, Fig. 8a). In Tomeurus, Heterandria, Neoheterandria, Belonesox, Poeciliopsis, and Phalloceros n. sp. U, Phalloceros n. sp. T, and Phalloceros n. sp. S this process is large (state 1, Fig. 8b). In Phalloptychus this process is enormous; it is as long as the remaining basipterygium (state 2, Fig. 8c; 15). Based on present analysis of relationships, state 1 is interpreted as independently evolved in Tomeurus, Heterandria, Neoheterandria, Belonesox, Poeciliopsis, and in the ancestor of Clade 82, with a reversal in Clade 78. State 2 is hypothesized as synapomorphic for Phalloptychus species.
Character 38 -Lateral keel of basipterygium in adult males: (0) absent; (1) present. Most cyprinodontiforms lack a lateral keel of basipterygium in adult males. This derived feature appeared to have been independently acquired by Scolichthys, Pamphorichthys, and by the ancestor of Poeciliopsis and Phalloptychus.
Character 39 -Narrowing of lateral surface of basipterygium base in adult males ( Fig. 9): (0) absent; (1) present.  Cnesterodon is unique among cyprinodontiforms by the narrowing of lateral surface of basipterygium base in adult males (state 1; Fig. 9). This condition is absent in remaining atherinomorphs (state 0) and is interpreted as synapomorphic for Cnesterodon.
Character 40 -First ray of left and right pelvic fins in adult males ( Fig. 9, 10): (0) similar to each other; (1) different from each other.
All cyprinodontiform fishes possess first ray of left and right pelvic fins similar to each other in adult males (state 0, Fig. 9), except for Phalloptychus species, in which first ray of left pelvic fin is much wider and more specialized than right one (state 1; Fig. 10). This condition is hypothesized as synapomorphic for Phalloptychus.
In Neoheterandria, Heterandria, Girardinus, Xenophallus, Poeciliopsis, Poecilia, Limia, Phalloceros, (except Phalloceros n. sp. R) width of first pelvic-fin ray in adult males decreases abruptly at distal portion, and distal slender portion is long (state 1, Fig. 11). In adult males of Tomeurus and Cnesterodon, the first pelvic-fin ray decreases abruptly at distal portion, and distal slender portion is short (state 2, Fig. 9), which is synapomorphic for the genus and independently acquired in Tomeurus. In Phalloptychus the first pelvic-fin ray in adult males is very wide, especially the right (state 3, Fig. 10). This condition is interpreted as synapomorphic for Phalloptychus.
Second pelvic-fin ray is branched (state 0) in adult males of cyprinodontiforms, with the exception of Tomeurus and Cnesterodon species, which possess unbranched second pelvic-fin ray (state 1). Following the present hypothesis, Tomeurus and Cnesterodon independently acquired an unbranched second pelvic-fin ray in adult males. A reversal occurs in C. septentrionalis.
Phalloptychus is unique among cyprinodontiform fishes by the possession of a lateral projection near the bifurcation of second right pelvic-fin ray in adult males (Fig. 10a). This condition is interpreted as synapomorphic for Phalloptychus. Character 44 -Number of pelvic-fin rays in females: (0) six or seven; (1) five; (2) three. Females of most cyprinodontiforms possess six or seven pelvic-fin rays (state 0). Ghedotti (2000) reported less than six pelvic-fin rays as synapomorphic for Phallotorynus, Phalloceros, Cnesterodon and Tomeurus.
Females of Phalloptychus, Cnesterodon, Phallotorynus, and Phalloceros possess five anal-fin rays (state 1). Females of Tomeurus exhibit three pelvic-fin rays (state 2). Tracking the present historical hypothesis for poeciliines, state 1 is proposed as synapomorphic for Phalloptychus januarius and P. iheringii. This derived feature is independently acquired by and also synapomorphic for a clade embracing Cnesterodon, Phallotorynus, and Phalloceros [Clade 111].
Phalloptychus is unique among cyprinodontiform fishes by the possession of a callosity at the distal portion of right pelvic fin in adult males. This condition is interpreted as synapomorphic for Phalloptychus (state 1; Fig. 12).
Ghedotti (2000) Fig. 13c). Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.    (4) Cnesterodon shows a hypotrophy of gonapophyses to a vestigial stage. Rosen & Bailey (1963) described the structure and distribution of gonapophyses in many poeciliines. Ghedotti (2000) reported the presence of gonapophyses in poeciliines, except in Tomeurus and Cnesterodon.
Among studied taxa, members of the outgroup, Tomeurus, Alfaro, and Cnesterodon species lack well-developed gonapophyses (state 0). Poecilia, Limia, Pamphorichthys, Micropoecilia, and "Poecilia" possess two (rarely one) welldeveloped gonapophyses (state 2). Xenodexia exhibit four well-developed gonapophyses (state 3). The remaining studied taxa present three well-developed gonapophyses (state 1). Results of the present phylogenetic analysis support the following interpretation of character evolution: (1) state 1 appeared once in the history of poeciliines in the ancestor of members of Clade 125; (2) this state is posteriorly modified to state 2 in the ancestor of members of Clade 99; (3) state 3 is interpreted as autapomorphic for Xenodexia; and (4) a reversal to state 0 occurred in Cnesterodon.
The curvature of gonapophyses (in general) was employed by Rodriguez (1997), however, this author codified only two character states: gonapophyses forming an acute angle relative to vertebral column and perpendicular to vertebral column.
Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.
The present phylogenetic analysis indicates that state 1 appeared at the ancestor of Clade 124. State 2 is supposed to have been acquired by the ancestor of members of Clade 122, with reversals to state 1 in Belonesox, Phallichthys, and in Clade 112. A reversal to state 0 occurs in Xiphophorus. State 3 is interpreted as synapomorphic for Phalloptychus.
Adult males of cyprinodontiform fishes (with the exception of Cnesterodon species) exhibit the distal portions of pleural ribs 6, 7, and 8 not expanded (state 0). Rosa & Costa (1993) recognized the presence of winglike expansions on distal portions of male pleural ribs, as synapomorphic for Cnesterodon species. Cnesterodon is unique among cyprinodontiforms by the expansion of distal portion of pleural ribs 6, 7, and 8 in adult males (state 1). This condition is herein also interpreted as synapomorphic for Cnesterodon species.
In most cyprinodontiforms the distal tip of pleural rib 7 in adult males does not surpass that of pleural rib 8 (state 0). Cnesterodon is unique among cyprinodontiforms by the length of pleural rib 7 in adult males, which surpasses that of pleural rib 8. This condition is interpreted as synapomorphic for Cnesterodon (state 1). The possession of this state by Phalloceros n. sp. I is assumed as homoplastic.
Character 58 -Pleural rib 9 in adult males ( Fig. 15): (0) normally developed; (1)  The curvature of pleural ribs in adult males was described and discussed by Rosen & Bailey (1963) and Rosa & Costa (1993). Ghedotti (2000: 24) discussed the rib condition in adult male. However, the rib condition of Tomeurus as described by Ghedotti (2000) himself is far autapomorphically peculiar and different from states 1 and 2 of the present character, therefore Tomeurus was coded 0 for this character.
Results of the present phylogenetic analysis support the following interpretation of character evolution: (1)  Character 60 -Pleural ribs association with haemal arches in males: (0) absent; (1) present.
There is no consensus among authors concerning this character. Basal atherinomorph fishes do not exhibit pleural ribs associated with haemal arches (state 0). Parenti (1981) recognized pleural ribs on haemal arches as synapomorphic for Poeciliidae. However, Costa (1998)  We opted to examine this character in males and females separately, i.e. we split this character in two, because the atherinomorphs (state 0). Phalloptychus is unique among cyprinodontiform fishes by having pleural rib 9 well-developed in adult males, i.e. longer than remaining pleural ribs, curved forward and expanded at distal tip (state 1). This condition is interpreted as synapomorphic for Phalloptychus.  Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.

Dorsal Fin
Character 62 -Position of the first proximal radial of dorsal fin in adult males: Located between neural arches of vertebrae: (0) 12 and 13; (1) 13 and 14; (2) 8 and 9; (3) 23 and 24 or 24 and 25; (4) 10 and 11; (5) 15 and 16 or 16 and 17; (6) 11 and 12; (7) 14 and 15; (8) 7 and 8. Ghedotti (2000) studied the position of dorsal-fin origin relative to the origin of anal fin. We opted to examine this character by means of the position of the first proximal radial of dorsal fin in adult males relative to neural arches of vertebrae. We also examined males and females separately, i.e. we split this character in two, because the position of dorsaland anal-fin origins may vary between sexes.
Neoheterandria, Carlhubbsia, and Xiphophorus exhibit this structure located between neural spines of vertebrae 8 and 9 (state 2). In Tomeurus the first proximal radial of dorsal fin in adult males is located between neural spines of vertebrae 23 and 24 or 24 and 25 (state 3). Jenynsia, Phallichthys, Xenophallus, Limia, and Pamphorichthys present this element located between neural spines of vertebrae 10 and 11 (state 4) In Fluviphylax, this radial is located between neural spines of vertebrae 15 and 16 or 16 and 17 (state 5). Brachyrhaphis, Heterandria, Quintana, Xenodexia, Poecilia, and Cnesterodon septentrionalis possess the first proximal radial of dorsal fin in adult males located between neural spines of vertebrae 11 and 12 (state 6). In Phalloceros n. sp. U, Phalloceros n. sp. R, and Phalloceros n. sp. O, this structure is located between neural spines of vertebrae 14 and 15 (state 7) and in Cyprinodon it is located between neural spines of vertebrae 7 and 8 (state 8).
Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals (24 steps) during the history of the Cyprinodontiformes.
Cyprinodon, Phallichthys, Xenophallus, Xenodexia, Limia, and Pamphorichthys exhibit this structure located between neural spines of vertebrae 10 and 11 (state 2). In Brachyrhaphis, Quintana, "Poecilia", and C. septentrionalis the first proximal radial of dorsal fin in adult females is located between neural spines of vertebrae 11 and 12 (state 3). Alfaro, Gambusia, Neoheterandria, Phallotorynus n. sp. A, Phalloceros n. sp. P, Phalloceros n. sp. M, and Phalloceros n. sp. J present this element located between neural spines of vertebrae 14 and 15 (state 4). In Fluviphylax and Phallotorynus fasciolatus this radial is located between neural spines of vertebrae 15 and 16 (state 5). Tomeurus possess the first proximal radial of dorsal fin in adult females located between neural spines of vertebrae 23 and 24 or 24 and 25 (state 6). In Carlhubbsia and Xiphophorus this structure is located between neural spines of vertebrae 8 and 9 (state 7). Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.
Second, third, and fourth gonactinosts are fused into a gonactinost complex in all poeciliines except Xenodexia. This feature is therefore hypothesized as a synapomorphy for Poeciliinae with a reversal in Xenodexia.
Character 68 -Inclination of gonactinost complex relative to body longitudinal axis: (0) very inclined backwards, forming a less than 45 degrees angle with body longitudinal axis; (1) gonactinost complex approximately perpendicular to body longitudinal axis; (2) inclined forward, forming a more than 90 degrees angle with body longitudinal axis; (3) little inclined backwards, forming a angle between 45 and 90 degrees with body longitudinal axis. Ghedotti (2000) employed this character, however, recognizing two states "(0) anteriorly inclined or vertical and (1) posteriorly inclined". This author reported a posteriorly inclined gonactinost complex as synapomorphic for Cnesterodon and Tomeurus and independently acquired in Alfaro.
In Tomeurus and Cnesterodon the gonactinost complex is very inclined backwards to an angle smaller than 45 º relative to the body longitudinal axis (state 0). On the basis of the present hypothesis of relationships, the condition in Cnesterodon is interpreted as a synapomorphic reversal and as plesiomorphic in Tomeurus.
In Phalloptychus the gonactinost complex is a little inclined backwards, forming an angle of 45 º and 90 º with body longitudinal axis (state 3); this is hypothesized as a uniquely derived and unreversed feature of Phalloptychus species.

P R O O F S
Character 69 -Basal process on first gonactinost: (0) absent; (1) small; (2) 29A). This derived feature presented several independent acquisitions and reversals during the history of the Cyprinodontiformes. In Cnesterodon species this process is enlarged (state 2, Rosen & Bailey, 1963: fig. 30) and is synapomorphic for the species of Cnesterodon. Remaining studied taxa lack such process (state 0).
The anterior border of second proximal radial of anal fin in adult males of most atherinomorphs fishes is straight (state 0). However, in Priapichthys, Heterandria, Pseudopoecilia, Neoheterandria, Girardinus, Phallichthys, Xenophallus, Poeciliopsis, Carlhubbsia, Pamphorichthys hollandi, Micropoecilia, "Poecilia", and Phallotorynus the anterior border of second gonactinost exhibits a convex expansion in adult males (state 1, Rosen & Bailey, 1963: fig. 29A). Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.
In most cyprinodontiform fishes the distal portion of second and third anal-fin proximal radials in adult males is separate (state 0, Fig. 16a Fig. 16b, e). This feature presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.
In Phallotorynus n. sp. A and Phallotorynus n. sp. B these are coalescent only at distal tip, forming an oblong aperture (state 2, Fig. 16c), with is interpreted as a synapomorphy for these two species.
In outgroup taxa, Xenodexia, and Micropoecilia sp. second and third gonactinosts are free in adult males (state 0, Fig. 16a). These elements are completed fused (state 1, Fig.  16b) Fig. 16c, e). Although this character contributed to the resolution of the present topology, it pre-sented several independent acquisitions and reversals during the history of the Cyprinodontiformes.
In most poeciliines the gonactinostal complex is expanded in a laminar plate in the anterior-posterior plane; gonactinosts are spread out in this plate-like spokes in a fan (state 0). Belonesox and Gambusia possess second, third, and fourth gonactinosts fused into a column (state 1). This is interpreted as a synapomorphic for a clade composed by both genera. This feature has been proposed by Rauchenberger (1989).
Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the poeciliines. Ghedotti (2000) recognized 3 states of this character. We add a fourth state (75-3). This author reported third and fourth proximal anal-fin radials completely fused in Gambusia affinis, Poeciliopsis latidens, and Heterandria formosa and partially fused, forming an oblong opening in Phallichthys amates and Poecilia sphenops. This character was employed by Ghedotti (2000). Most cyprinodontiforms lack lateral flanges on ventral portion of fourth anal-fin radial in adult males (state 0). Lateral flanges on ventral portion of fourth gonactinost are continuous, without dorsal cleft (state 1) in Gambusia and Belonesox and are hypothesized as a synapomorphy uniting both. In Girardinus, Phallotorynus, and Phalloceros lateral flanges are present and cleaved dorsally forming separate dorsally directed processes (state 2). The topology of our strict consensus tree supports the hypothesis that state 2 appeared independently in Girardinus and in the ancestor of Phallotorynus and Phalloceros.
Phallotorynus species are unique among cyprinodontiforms by having a posteroventral projection of ventral flange of fourth gonactinost (Fig. 17). The absence of this structure in remaining cyprinodontiform taxa is consid-ered plesiomorphic (state 0). In Phallotorynus fasciolatus and P. victoriae this projection is large (state 1, Fig. 17a). In Phallotorynus jucundus, Phallotorynus n. sp. A, and Phallotorynus n. sp. B the projection is small (state 2; Fig.  17b). The presence of a posteroventral projection of ventral flange of fourth gonactinost is interpreted as synapomorphic for Phallotorynus with a derived size reduction in Phallotorynus jucundus, Phallotorynus n. sp. A, and Phallotorynus n. sp. B.
In most cyprinodontiforms middle anal-fin radials 5, 6, and 7 are symmetrical in adult males, with right and left lateral projections similar in shape and size (state 0). Middle anal-fin radials 5, 6, and 7 are asymmetrical in adult males, with right lateral projection more compressed and much larger than left one in Phallichthys, Xenophallus, Poeciliopsis, Phalloptychus, and Xenodexia (state 1, Fig. 19). This feature is hypothesized as independently acquired by Xenodexia and in the ancestor of a clade containing Phallichthys, Xenophallus, Poeciliopsis, and Phalloptychus [Clade 113]. Gonactinost 5 is free in almost all poeciliines (state 0). In Limia, "Poecilia" reticulata, and Micropoecilia gonactinost 5 is fused to gonactinostal complex (state 1). This condition is interpreted as derived and independently acquired by the aforementioned genera.
Most cyprinodontiforms possess the ventral portion of proximal anal-fin radials 6-10 in adult males laterally compressed, with anterior and posterior flanges (state 0). In poeciliines the ventral portion of gonactinosts 6-10 is not

P R O O F S
laterally compressed, without anterior and posterior flanges (state 1), which is proposed as synapomorphic for the subfamily Poeciliinae. This character has been discussed by Ghedotti (2000).
Character 84 -Ninth anal-fin gonactinost (Rosen & Bailey, 1963: fig. 53): (0) straight; (1) bearing wing-like lateral projections. Rosen & Bailey (1963) reported the presence of wing-like projections in Brachyrhaphis, Priapichthys, and Neoheterandria (including Xenophallus) and employed this character as diagnostic (in combination with other characters) for the genera above. Meyer & Etzel (1996) reported the presence of "plate-like outgrowths" in the ninth gonactinost of Neoheterandria tridentiger (Garman) and in the eighth gonactinost of Xenophallus umbratilis (Meek). These authors stated that the presence of wing-like lateral projections as a "primitive character in poeciliid fishes" (Meyer & Etzel, 1996: 3), arguing that this condition is very common among poeciliines: "because plate-like outgrowths of the gonactinostal system are found in several taxa" (Meyer & Etzel, 1996: 3). However, this statement is made in the absence of a cladistic analysis, which prevents the evaluation of character polarity. Besides, the fact that a character state is very common does not guarantee it is the plesiomorphic state.
Brachyrhaphis, Priapichthys, Neoheterandria, and Xenophallus, are unique among poeciliines by the possession of wing-like lateral projections on the ninth gonactinost. The presence of this structure in each of four taxa above is interpreted as apomorphic (state 1) and homoplastic (independently derived). On the other hand, the absence of these wing-like lateral projections in remaining atherinomorphs is viewed as the plesiomorphic condition (state 0).
Anal-fin rays 3, 4, and 5 in adult males are normally developed, i.e. are similar to remaining anal-fin rays in most atherinomorphs (state 0). Poeciliines are unique by the possession of a copulatory structure (gonopodium) formed by modified anal-fin rays 3, 4, and 5 in adult males (state 1). This condition is hypothesized as a uniquely derived and unreversed synapomorphy for the subfamily.
Most cyprinodontiform fishes exhibit a symmetrical anal fin in adult males (state 0). The anal fin in adult males specimens is asymmetrical in Phallichthys, Xenophallus, Poeciliopsis, and Phalloptychus, Quintana, Carlhubbsia, and Xenodexia (state 1). Rosen & Bailey (1959) discussed the gonopodium asymmetry of Phalloptychus, Poeciliopsis, Phallichthys, Xenophallus, Carlhubbsia, Quintana, and Girardinus. They believed that characters related to gonopodium folding were highly adaptive and may had evolved independently more than once within the subfamily.

P R O O F S
Most cyprinodontiforms do not exhibit a palp in subdistal segments of R3 (state 0), whereas this structure is present (state 1) in some poeciliids. Rosen & Bailey (1963) used the presence of a palp in Alfaro and Poecilia as evidence that Alfaro should be placed in the tribe Poeciliini. Rodriguez (1997) suggested that the presence of a palp was independently derived in Alfaro and in the ancestor of a clade including Poecilia, Pamphorichthys, and Limia. Ghedotti (2000) reported the presence of a palp in Alfaro and Poecilia. Among the taxa studied, a palp was observed in Alfaro, Poecilia, Limia, Pamphorichthys, Micropoecilia, and "Poecilia". Our results support the hypothesis that a palp in subdistal segments of R3 was independently derived in Alfaro and the common and exclusive ancestor of a clade including Poecilia, Limia, Pamphorichthys, Micropoecilia, and "Poecilia".
Phallotorynus species are unique among cyprinodontiforms by the possession of a V-shaped ventral projection at distal portion of R3. This condition is hypothesized as apomorphic (state 1) and exclusively shared by Phallotorynus species.
Most cyprinodontiform fishes lack a pedicel in R3 united to R4 (state 0). Phallotorynus, Phalloceros, and Cnesterodon species are unique among cyprinodontiforms by the possession of a pedicel in R3 united to R4 (state 1). This condition is hypothesized as a synapomorphy for the tribe Cnesterodontini.
The presence of a pedicle at tip of R3 long been reported by different authors (Hubbs, 1924;Rosen & Bailey, 1963). Phallotorynus, Phalloceros and Cnesterodon species are unique among cyprinodontiforms by the possession of a pedicel at tip of R3 (state 1). This condition is hypothesized as a synapomorphy for the tribe Cnesterodontini.
The presence of a membranous appendix at tip of R3 long been reported by different authors (Hubbs, 1924;Rosen & Bailey, 1963;Ghedotti, 2000). Phallotorynus, Phalloceros, and Cnesterodon species are unique among cyprinodontiforms by the possession of a membranous appendix at tip of R3 (state 1). This condition is hypothesized as a synapomorphy for the tribe Cnesterodontini.
Paired appendix at tip of R3 has long been reported as diagnostic for Phalloceros (e.g. Eigenmann, 1907;Rosen & Bailey, 1963) Phalloceros species are unique among cyprinodontiforms by the possession of a paired appendix at tip of R3 (state 1). This condition is hypothesized as synapomorphic for the genus Phalloceros.  Fig. 22), which is interpreted as synapomorphic for the ancestor of a clade containing these species plus Phalloceros n. sp. M, with a reversal in the latter.
In most Phalloceros species hooks are small, anterodorsally directed and located nearer the base of the gonopodial appendix (Fig. 21). Phalloceros n. sp. I, Phalloceros n. sp. R, and Phalloceros n. sp. N, are unique among poeciliines by the possession of large hooks, downward directed and located in the corner of the gonopodial appendix (Fig. 23). This feature is hypothesized as synapomorphic for these three species.      Rosen & Bailey (1959) suggested that Girardinus and Quintana were closely related and share a homologous minute paired terminal hook on R3. However our results support that Girardinus and Quintana are not closely related and that the possession of a minute paired terminal hook on R3 was independently acquired by both taxa. Thus this derived feature is regarded as non-homologous in Girardinus and Quintana.

P R O O F S
Character 100 -Trowel-like appendix at tip of R3 (Rosen & Bailey, 1963: fig. 31 A, B): (0) absent; (1) present. Rosen & Bailey (1963) reported a trowel-like appendix at tip of R3 as synapomorphic for Phallotorynus species. Phallotorynus is unique among cyprinodontiforms by the possession of a trowel-like appendix at tip of R3. This condition (state 1) is hypothesized as a synapomorphy for the genus, while the absence of such structure in remaining cyprinodontiform fishes is interpreted as plesiomorphic (state 0).
Unpaired appendix at tip of R3 was suggested as a putative synapomorphy for Cnesterodon species by Rosen & Bailey (1963) and Rose & Costa (1993). Cnesterodon species are unique among cyprinodontiforms by the possession of an unpaired appendix at tip of R3. The possession of this structure (state 1) is interpreted as synapomorphic for Cnesterodon.
Character 102 -Large membrane on pedicel of gonopodium: (0) absent; (1) present: Phallotorynus species are unique among cyprinodontiforms by the possession of a large membrane on pedicel of the gonopodium. This condition is herein hypothesized as a synapomorphy for the genus Phallotorynus (state 1).
Phallotorynus species are unique among cyprinodontiforms by the possession of a trowel-like appendix at tip of R3. In the light of the present evidence, the absence of this appendix in cyprinodontiform fishes other than Phallotorynus is interpreted as plesiomorphic (state 0). In Phallotorynus victoriae, P. jucundus, and Phallotorynus n. sp. B the trowel-like appendix is narrow and elongate (state 1;Figs. 24,27). In Phallotorynus n. sp. A the appendix is wide and short (state 2; Fig. 25). The presence of a trowel-like appendix at tip of R3 is interpreted as synapomorphic for Phallotorynus with a derived widening and reduction in length in Phallotorynus n. sp. A. Phallotorynus fasciolatus was coded "?" for character state could not be checked due to poor condition of the material studied.
Phallotorynus victoriae, Phallotorynus n. sp. A, and Phallotorynus n. sp. B possess small lateral processes on the trowel-like appendix (state 0; Figs. 25, 27). Phallotorynus jucundus presents large lateral processes (state 1; Fig. 24), and this feature is interpreted as autapomorphic for P. jucundus. Phallotorynus fasciolatus was coded "?" for character state could not be checked due to poor condition of the material studied.
Character 106 -Profile of lateral border of left and right halves of trowel-like appendix: (0) straight; (1) concave.
Phallotorynus victoriae, Phallotorynus jucundus, and Phallotorynus n. sp. B possess the profile of lateral border of left and right halves of trowel-like appendix straight (state 0; Figs. 24, 27). Phallotorynus n. sp. A presents a concave profile (state 1; Fig. 25), and this feature is interpreted as autapomorphic for Phallotorynus n. sp. A. Phallotorynus fasciolatus was coded "?" for character state could not be checked due to poor condition of the material studied.
Character 107 -Level of separation/union of left and right halves of the trowel-like appendix: (0) halves separate by a gap along two thirds of its extension; (1) halves united along its whole extension; (2) halves separate along approximately five sixths of its whole extension.
In Phallotorynus victoriae and Phallotorynus n. sp. B the halves of trowel-like appendix are separate by a gap along two thirds of its extension (state 0; Fig. 26). In Phallotorynus jucundus halves are united along its full extension (state 1; Fig.  24). Phallotorynus n. sp. A presents halves separate along approximately five sixths of its whole extension (state 2; Fig. 25). Phallotorynus species are unique among cyprinodontiforms by the possession of a trowel-like appendix at tip of R3. States 1 and 2 are interpreted as autapomorphic for Phallotorynus jucundus and Phallotorynus n. sp. A, respectively.
The present phylogenetic study support the hypothesis that state 1 was independently acquired by Heterandria, Limia, and by the ancestor of Cnesterodon, Phallotorynus, and Phalloceros, with a reversal to state 0 in C. decemmaculatus. State 2 is interpreted as synapomorphic for Phallotorynus n. sp. A, and Phallotorynus n. sp. B, whereas state 3 is interpreted as synapomorphic for Phalloceros species with a reversal in Phalloceros n. sp. D.
In most poeciliines distal segments of R4p posterior to serrae are wider than deep (state 0). Rodriguez (1997) proposed distal segments of R4p posterior to serrae deeper than wide (state 1) as a synapomorphy uniting Limia and Pamphorichthys. This was confirmed herein. Additionally, this feature was also observed to be present in Xenodexia. Our results support the assumption of this derived condition as synapomorphic for a clade containing Xenodexia, Poecilia, Limia, Pamphorichthys, Micropoecilia, and Character 116 -Number of subdistal retrorse spines on R4p: (0) zero; (1) eight or more; (2) four to seven.
Phalloptychus species are unique among cyprinodontiforms by the possession of a short, dorsal protuberance close to base of R4p (Fig. 27). Our phylogenetic analysis supports the hypothesis that this feature represents a synapomorphy for Phalloptychus species (state 1).
An elongate, dorsal protuberance just behind retrorse spines series of R4p (state 1, Rosen, 1979: fig. 6, 26) in present in Priapella, Heterandria, Girardinus, Phallichthys, Quintana, and Phalloceros. Remaining cyprinodontiforms lack this structure. The phylogenetic analysis supports assuming the absence of such protuberance as plesiomorphic (state 0) and its presence as apomorphic (state 1). The presence of this protuberance is herein interpreted as synapomorphic for the members of Clade 123, with subsequent reversals at nodes 120, 110 and 115. Within Clade 115, Quintana and Phalloceros independently reacquired state 1 condition. (1) present. Rodriguez (1997) reported a keel on posterior ventral surface of R5 formed by the projection of R5 toward R4p as synapomorphic for a clade comprising Poecilia, Pamphorichthys, and Limia. We observed this keel in these taxa as well as in Xenodexia. Our phylogenetic framework supports the absence of this keel in most cyprinodontiforms as plesiomorphic (state 0), and its presence as apomorphic (state 1). Thus, the presence of such keel is interpreted as a synapomorphy for a clade embracing Xenodexia, Poecilia,Limia,Pamphorichthys,Micropoecilia,and "Poecilia" [Clade 104] with a reversal in Micropoecilia + "Poecilia" clade.
Xenodexia, Poecilia, Limia, Pamphorichthys, Micropoecilia, and "Poecilia" share the presence of a wide groove dorsal to R5. Our phylogenetic framework supports the absence of this groove in most cyprinodontiforms as plesiomorphic (state 0), and its presence as apomorphic (state 1). Thus, the presence of a groove dorsal to R5 is interpreted as a uniquely derived and unreversed synapomorphy for a clade containing Xenodexia, Poecilia,Limia,Pamphorichthys,Micropoecilia,and "Poecilia" [Clade 104] in the light of present evidence.
Character 122 -Distal segment at tip of R5a: (0) normal; (1) transformed in retrorse triangular spine; (2) hook-like. Rosa & Costa (1993) suggested the distal segment at tip of R5a modified into a retrorse triangular spine as a putative synapomorphy for Cnesterodon species, whereas Rauchenberger (1989) recognized a hook on tip of R5a as a synapomorphy shared by Gambusia and Belonesox. In most cyprinodontiforms, the distal segment at tip of R5a is similar in shape to remaining segments (state 0). Cnesterodon spe-cies are unique among poeciliines by having the distal segment at tip of R5a modified into a retrorse triangular spine (state 1; Ghedotti, 2000: fig. 15A), which is herein hypothesized as synapomorphic for this genus. In Gambusia and Xiphophorus this segment is modified into a hook (state 2; Rauchenberger, 1989: fig. 20). However, following our hypothesis of relationships, these derived features are interpreted as independently acquired in Gambusia and Xiphophorus.
Character 123 -Hook on R5a contacting the segments of R4p (Rauchenberger, 1989: fig. 20, 23): (0) absent; (1) present. Rauchenberger (1989) proposed the presence a hook on R5a contacting the segments of 4p as a synapomorphy uniting Gambusia and Belonesox. We have confirmed this and also observed this derived feature in Xiphophorus. Thus, the presence of a hook on 5a contacting the segments of R4p is interpreted as independent acquisitions by Xiphophorus and by the ancestor of Gambusia and Belonesox.
Most cyprinodontiforms lack serrae on R5p (state 0). Girardinus, Quintana, and Carlhubbsia are unique among poeciliines by the possession of serrae on R5p (state 1). This derived feature is interpreted as independent acquisitions by Girardinus and by the ancestor of Quintana and Carlhubbsia.

P R O O F S
In members of the outgroup, Alfaro, Brachyrhaphis, Pseudopoecilia, Pamphorichthys, Micropoecilia, and " Poecilia" R6a and R6p are free from each other (state 0). Remaining poeciliines present various degrees of partial fusion between these elements (state 1). These branches are totally fused in Poeciliopsis and Phalloptychus (state 2). Our results allow the assumption that state 1 appeared independently in Tomeurus and in the ancestor of members of the Clade 124, with two subsequent reversals: (1) in Pseudopoecilia and (2)  Character 127 -Degree of fusion between more distal elements of branches of sixth anal-fin ray in adult males (R6): (0) not fused; segmented; (1) partially fused; (2) totally fused.
In members of the outgroup, Alfaro, Brachyrhaphis, Belonesox, Pseudopoecilia, Pamphorichthys, Micropoecilia, and "Poecilia" the more distal elements of R6 branches are not fused (state 0). In males of Tomeurus, Priapella, Priapichthys, Heterandria, Gambusia, Phallichthys, Quintana, Carlhubbsia, Xiphophorus, Poecilia, Limia, Phallotorynus, and some species of Phalloceros more distal elements of R6 branches are partially fused (state 1). These elements are totally fused (state 2) in remaining taxa studied. Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.
The distal portion of R6 of most cyprinodontiform fishes is not expanded (state 0). Neoheterandria, Phallichthys, Carlhubbsia, Cnesterodon, and Phallotorynus are unique among poeciliines by the possession of an expanded distal portion of R6 (state 1; Ghedotti, 2000: fig. 14D). Following our results, this feature is interpreted as independently acquired in all genera above.
Character 129 -Size of lower branch of R6: (0) longer than upper branch; (1) as long as upper branch.
In most cyprinodontiforms the lower branch of R6 is longer than the upper branch (state 0). In Phalloptychus species the lower branch is as long as the upper (state 1), which is assumed as synapomorphic for Phalloptychus species Character 130 -Distal portion of R6 and seventh anal-fin ray in adult males (R7): (0) not fused; (1) fused.
In most cyprinodontiforms distal portion of R6 and R7 are not free (state 0). In Phallotorynus n. sp. A and Phallotorynus n. sp. B distal portion of R6 and R7 are fused to each other (state 1). This is hypothesized as synapomorphic for a clade containing these two species.
Character 132 -Number of caudal-fin rays in contact with the hypural plate: (0) less than nine; (1) nine or more.
In Aplocheilichthys, Fluviphylax, Procatopus, Tomeurus, Brachyrhaphis, Belonesox, Gambusia, Xenophallus, Phalloptychus eigenmanni, Pamphorichthys, Micropoecilia, "Poecilia", Cnesterodon septentrionalis, Phallotorynus, and Phalloceros (except Phalloceros n. sp. G and Phalloceros n. sp. I) the number of caudal-fin rays in contact with the hypural plate is less than nine (state 0). In remaining taxa studied there is nine or more (state 1) caudal-fin rays in contact with the hypural plate. Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.

Pigmentation
Character 133 -Elongate vertical bars on lateral surface of body (Fig. 29) Fig. 30b). An elliptical spot appeared independently in Phalloceros and in the clade Neoheterandria + Scolichthys. Phalloceros n. sp. P possesses an autapomorphic-squared spot (state 3, Fig. 30c). In Phalloceros n. sp. H the spot is elongate forming a bar vertical reaching dorsal and ventral profiles (state 4, Fig. 30d) and it is autapomorphic for this species. Phalloceros n. sp. F also possesses an autapomorphic spot, which is typically densely pigmented rectangle-like lateral spot more anterior located (on the 14 th or 15 th , very rarely 16 th , scale of longitudinal line) (state 5; Fig. 30e).
Phallotorynus species, with the exception of P. fasciolatus are unique among cyprinodontiforms by the possession of round to elliptical dark blotches along ventral half of flanks. Therefore, this feature is interpreted as synapomorphic for a clade comprising Phallotorynus jucundus, Phallotorynus victoriae, Phallotorynus n. sp. A, and Phallotorynus n. sp. B. Phallotorynus jucundus exhibits four to seven blotches, and it is considered as an autapomorphy. Phallotorynus n. sp. A presents two blotches, and this state is considered autapomorphic for this species.
Most cyprinodontiforms possess the dorsal fin slightly pigmented with black (state 0, Fig. 32a). In Cyprinodon, Brachyrhaphis, Poecilia, Xenophallus, Micropoecilia, and Phallotorynus dorsal fin is moderately pigmented with black (state 1, Fig. 32b), and this feature seems to be independently acquired in each of these genera. An autapomorphic dorsal fin densely pigmented with black (state 2, Fig. 32c) is present in Phallotorynus jucundus. Most cyprinodontiforms lack a dark stripe on median portion of dorsal fin (state 0,Figs. 29,34). Such a stripe is present in Brachyrhaphis, Priapichthys,Heterandria,Gambusia,Pseudopoecilia,Neoheterandria,Scolichthys,Xiphophorus,Poecilia,Limia,Pamphorichthys,Micropoecilia,"Poecilia",Phallotorynus,and Phalloceros (state 1, Fig. 32a). Our results support the hypothesis that the derived condition appeared at the ancestor of Clade 125, with subsequent reversals in Belonesox, Priapella, Xenodexia and re-acquisitions in Clades 112, 106 and Quintana. Character 138 -Dark patch of pigmentation along R3: (0) absent; (1) present. Most cyprinodontiforms lack a dark patch of pigmentation along R3 (state 0, Fig. 34). Scolichthys, Girardinus, Phallichthys, Xenophallus, Phalloptychus, Xiphophorus, Phallotorynus, and Phalloceros exhibit a dark patch of pigmentation along R3 (state 1, Fig. 33). According to the present analysis, this feature evolved independently in (1)  Character 139 -Dark spot posterior to anal-fin base of males continuous ventrally side by side and continuous with ventral median line of caudal peduncle (Fig. 34): (0) absent; (1) present. Rosa & Costa (1993) suggested that the presence of a dark spot posterior to anal-fin base of males continuous ventrally side by side and continuous with ventral median line of caudal peduncle could be a synapomorphy for the species of Cnesterodon. This is confirmed by our phylogenetic study. Character 140 -Ground pigmentation of anal fin of females: (0) slightly speckled with black, or hyaline, not forming a distinct stripe on first rays; (1) moderately speckled with black, with chromatophores more concentrated anteriorly and forming a dark stripe on first rays; (2)  Most female cyprinodontiforms possess anal fin slightly speckled with black, or hyaline, not forming a distinct stripe on first rays (state 0, Fig. 32a). Females of Phallotorynus species are unique among cyprinodontiforms by the possession of a anal fin moderately to densely speckled with black, with chromatophores more concentrated anteriorly and forming a dark stripe on first rays (states 1 and 2, Fig.  32b,c). This condition is herein interpreted as synapomorphic for the genus and the possession of a black anal fin (state 2, Fig. 32c) is considered as a derived autapomorphy for Phallotorynus jucundus.
Viviparity among cyprinodontiform fishes has long been discussed (e.g. Rosen & Bailey, 1963;Parenti, 1981;Meyer & Lydeard, 1993;Ghedotti, 2000). Among cyprinodontiforms viviparity evolved independently in the Goodeidae, Anablepidae, and Poeciliinae. Tomeurus was coded "-" because it exhibits facultative viviparity. Since Tomeurus is the most basal poeciliine according to the present phylogenetic hypothesis, the facultative viviparity could be viewed in two different ways: (1) as a preliminary "testing" stage of viviparity towards "true viviparity" which was achieved by the ancestor of remaining poeciliines; or (2)  In Fluviphylax, Fundulus, Jenynsia, Brachyrhaphis, Tomeurus, Phalloptychus, Xenodexia, and Cnesterodon (except Cnesterodon n. sp. A) orbital osseous plates are absent (state 0). Pseudopoecilia and Quintana exhibit one anterior orbital bones (state 2). Remaining taxa studied possess two orbital osseous plates (anterior and posterior ones) (state 1). Although this character contributed to the resolution of the present topology, it presented several independent acquisitions and reversals during the history of the Cyprinodontiformes.

Phylogenetic reconstruction and synapomorphy list
The phylogenetic analysis yielded 96 equally most parsimonious trees with length (L) = 758 steps (including TSA in the outgroup), consistency index (CI) = 0.35, and retention index (RI) = 0.75. A strict consensus tree is shown in Figs. 1, 2, and 3. Synapomorphy list is presented in the "Taxonomic Account" section and in the Appendix III. Fits of individual characters are summarized in Appendix IV.

Taxonomic Account
The current phylogenetic study supports the proposal of a new classification for the subfamily Poeciliinae. These modifications are necessary in order to make groups natural (monophyletic). Diagnoses are provided for suprageneric clades. Nevertheless, diagnoses for monotypic tribes are only preliminary because this study focused on the search for derived features uniting genera rather than on generic autapomorphies. Thus, generic diagnoses were not fully surveyed during this study. Oncoming studies will possibly reveal several other diagnostic features for these monotypic tribes. Some authors (e.g. Regan, 1913;Hubbs, 1924; Rosen,  Rosen, 1979;Rosen & Bailey, 1963;Rauchenberger, 1989;Rodriguez, 1997) already provided some insight on diagnostic characters for some of these tribes but not necessarily on the light of a phylogenetic framework. Rather than writing down the diagnoses and discussions of these monotypic clades, we refer the reader to the articles above for a detailed study. Thus, the following classification is proposed (summarized in Table 3):  Remarks. The family name Poecilini has already been used by Bonaparte (1831), however it appeared to be preoccupied in Coleoptera. The family-group name based on Poecilus Bonelli (Carabidae) was created by Bonelli in 1810. He called the group "Poecilii", which is typically taken to be a familygroup name. When it is used these days, it is either as a tribe (Poecilini) or a subtribe (Poecilina) (David Maddison, in litt.). Later, Bonaparte (1846) added one "i" to the name differing it from the Coleoptera family-group name (even if the differences between two family-group names is only one letter they are not homonyms -article 55.4 of the ICZN, 1999).
However, Poeciliini Bonaparte, 1846 is an unjustified emendation for Poecilini Bonaparte, 1831 (article 32.5.3 of the ICZN, 1999) but it is in prevailing usage. So, it is attributed to the original author and date and is deemed to be a justified emendation following the article 33.2.3.1 of the ICZN (1999).

Composition. Tomeurus gracilis Eigenmann.
Distribution. Tomeurus gracilis occurs in small coastal drainages of the Venezuelan departments Delta Amacuro, Monagas, Territorio Federal, and in Brazilian states of Amapá, and Pará. The species also inhabits the drainages of rio Guamá and rio Tocantins in Brazil, the drainages of the Cuyuni, Mazaruni, and Essequibo Rivers in the Guyana and Courantyne rivier drainage in Suriname.
Distribution. As for Alfaro.

Xiphophorus
Distribution. Drainage of the rio Amazonas, rio Guamá basin, coastal drainages of Brazil, French Guyana, Guyana, and Suriname; Trinidad and Tobago.
Remarks. The tree topology placed "Poecilia" reticulata as the sister group of the Micropoecilia clade and far from the type-species of the genus, Poecilia vivipara. This fact brings taxonomic and nomenclatural implications, since the generic name Poecilia cannot be applied to the reticulata species. The most parsimonious action should be the merging of "Poecilia" reticulata with the genus Micropoecilia Hubbs, 1926. "Poecilia" reticulata was originally described as Poecilia reticulata Peters, 1859 and Lebistes poecilioides De Fillipi, 1861 is considered a junior synonym. Lebistes poecilioides, whose types are lost, was described as a new genus and a new species. Since the name Lebistes is older than Micropoecilia it has priority and Micropoecilia should be considered a junior synonym of Lebistes. However, Poeser & Isbrücker (2002) suggested, based on evidence on the original description of De Fillipi, that L. poecilioides is not equal to Poecilia reticulata. If it is the case, Lebistes cannot be a synonym of Micropoecilia. However, Eigenmann (1907) erected the genus Acanthophacelus for Poecilia reticulata, thus if Micropoecilia species and "Poecilia" reticulata are merged under the same generic name, Acanthophacelus Eigenmann, 1907 has priority over Micropoecilia Hubbs, 1926 and therefore should be resurrected and revalidated.
Contrarily to us, Rosen & Bailey (1959) sustained the idea that Poeciliopsiinae is not a natural group and should therefore be dissolved. They discussed the gonopodium asymmetry of the Poeciliopsinae and believed that characters related to gonopodium folding were highly adaptive and may had evolved independently more than once within the subfamily. Theses authors also supposed that Poeciliopsis and Phallichthys form a natural group and that Carlhubbsia is related to Quintana and Giradinus. Our results are contrary to Rosen & Bailey's (1959) assumptions, except for the statement that Carlhubbsia is related to Quintana. The dissolution of the Poeciliopsinae was formally put forward by Rosen & Bailey's (1963) new poeciliin classification. Members of this group were allocated in two different tribes: Phalloptychus in Cnesterodontini, Poeciliopsis, Phallichthys, and Xenophallus in Heterandriini. Expectedly Carlhubbsia, Quintana, and Giradinus were unified in a tribe Girardinini.
The classification of Rosen & Bailey (1963) also differs from ours in the position of genera Alfaro and Priapella in the tribe Poeciliini and the position of Brachyrhaphis in the tribe Gambusiini. We proposed Alfaro, Brachyrhaphis, and Priapella as type-genera of monotypic, basal tribes. Additionally, the tribes Girardinini and Heterandriini of Rosen & Bailey (1963) differ substantially from ours. This incongruence can be explained by the fact that diagnoses of supraspecific groups by Rosen & Bailey (1963) (and also the statements of Rosen & Bailey, 1959) were made in the absence of cladistic methodology and therefore not necessarily reflect phylogenetic relationships. Rosen (1979: 278-279) was aware of this and recognized the fragility of Rosen and Bailey's classification, when stated: " (…) it should be noted that the diagnoses of genera and other supra-specific groups in Rosen and Bailey were designed as phenetic statements of overall similarity. In short, little attention was paid to find unique characters for defining the taxa and only an implicit effort was made to interpret the different states of a character as primitive or derived. (…)." The advent of cladistic methodology has brought some improvement towards the comprehension of poeciliine relationships. Some authors addressed the relationships of smaller groups of the Poeciliinae (e.g. Rosen, 1979;Rauchenberger, 1989;Rosa & Costa, 1993;Rauchenberger et al., 1990;Meyer et al., 1994;Mojica et al., 1997;Rodriguez, 1997;Ptaceck & Breden, 1998;Marcus & McCune, 1999;Breden et al., 1999;Hamilton, 2001;Mateos et al., 2002;Poeser, 2003;Kallmann et al., 2004), whereas others coped with higher taxa (Parenti, 1981;Meyer & Lydeard, 1993;Costa, 1996Costa, , 1998Ghedotti, 2000). However, no phylogenetic study has simultaneously analyzed the relationships among representatives of all poeciliine genera. The only comprehensive study is the classic revision of the "Poeciliidae" by Rosen & Bailey (1963), which did not deal with cladistic methodology.
The proposed phylogeny and classification attempted to include representatives of all poeciliine genera. They are consistent with the results of the phylogenies proposed by Breden et al. (1999), andFigueiredo (1997) for Poecilia and its allies. On the other hand, the phylogenetic hypothesis of Rauchenberger (1989), Rodriguez (1997), and Ghedotti (2000) are partially congruent with the present results. Probably part of the incongruence can be explained by the fact that these phylogenetic studies had been performed for different subunits of the Poeciliinae. Rauchenberger (1989) proposed: (1) Gambusia and Belonesox as sister groups and (2) that clade as the sistergroup of Brachyrhaphis. The former proposal, but not the P R O O F S later was corroborated by our results. Our hypothesis support Gambusia and Belonesox more closely related to Pseudopoecilia, Neoheterandria, and Scolichthys than to Brachyrhaphis. Thus, we modified the usage of the name Gambusiini sensu Rauchenberger (1989), employing it to refer to a clade composed of Gambusia, Belonesox, Pseudopoecilia, Neoheterandria, and Scolichthys. Costa (1991) suggested the monophyly of the group embracing Pamphorichthys, Poecilia, Limia, Xiphophorus, Cnesterodon, Phalloceros, Phallotorynus, Phalloptychus, Priapichthys, Poeciliopsis, Priapella, Quintana, Carlhubbsia, Xenodexia, and Phallichthys supported by four putative synapomorphies (see our characters 12, 15, 26 and 28). This is partially corroborated by our results. The current phylogenetic analysis supports these features as a uniquely derived and unreversed synapomorphies for the supertribe Poeciliini [Clade 119], which in addition to the genera above (except Priapichthys and Priapella), also comprises Girardinus, Xenophallus, and Micropoecilia. Rodriguez's (1997) conclusions are also consistent with ours. This author was unable (like us) to find shared derived characters uniting Alfaro and Priapella together with Xiphophorus, Poecilia (including Micropoecilia), Limia, and Pamphorichthys as suggested by Rosen & Bailey (1963). Rodriguez (1997) hypothesized a monophyletic Poeciliini composed of Xiphophorus, Poecilia (including Micropoecilia), Limia, and Pamphorichthys on the basis of three synapomorphies: (1) a wide groove dorsal to R5; (2) long pelvic-fin in adult males, second ray surpassing the end of anal-fin base; and (3) compressed external teeth. However, none of the synapomorphies above are exclusive to this fish assemblage and are not useful as diagnostic characters. The presence of a wide groove dorsal to R5 is also observed in Xenodexia and herein interpreted as a uniquely derived and unreversed synapomorphy for a broader clade containing Xenodexia, Poecilia,Limia,Pamphorichthys,Micropoecilia,and "Poecilia" [Clade 104]. A long second pelvic-fin ray surpassing the end of anal-fin base is also present in Xenophallus and therefore is hypothesized to have been independently acquired by Xenophallus and by the ancestor of a clade comprising Xiphophorus, Xenodexia,Poecilia,Limia,Pamphorichthys,Micropoecilia,and "Poecilia" [Clade 108]. Presence of compressed teeth is herein interpreted as synapomorphic for a more encompassing clade, i. e. the supertribe Poeciliini [Clade 119] (with a reversal in Cnesterodon brevirostratus + C. septentrionalis clade). Ghedotti (2000) was the first to propose a phylogenetic classification scheme for the group, despite relying his conclusions on a limited sample (representatives of twelve poeciliine genera). This is understandable, however, given that the main focus of Ghedotti's paper was the relationships among members of a more encompassing group: the superfamily Poecilioidea. Except for the recognition of Alfaro and Priapella as basal taxa, Ghedotti's (2000) hypothesis is only partially harmonious with ours. It differs mainly by the placement of Tomeurus as a highly derived poeciliine (in the tribe Cnesterodontini), whereas our results support the assump-tion that Tomeurus is the most basal poeciliine; the sistergroup of the remaining members of the subfamily. The phylogenetic position of Tomeurus will be discussed in more detail below in the next section.
The goal of phylogenetic analyses is to continuously improve our knowledge of relationships. The purpose of a written classification is to communicate phylogenetic relationships. The proposed hypothesis of relationships and classification is only preliminary. Much of continued effort on taxonomy and phylogeny are still required towards a fully understanding of poeciliine history.
The Cnesterodontini. Questions related to the definition of the Cnesterodontini have long been based on the pre-cladistic diagnoses proposed by Hubbs (1924; and Rosen & Bailey (1963). The tribe Cnesterodontini as originally erected by Hubbs (1924) was composed of the genera Phalloceros, Cnesterodon, Phallotorynus, and Diphyacantha. The cnesterodontins were defined as poeciliines bearing "terminal segment of ray 3 forming a more or less specialized process" (Hubbs, 1924: 9). Hubbs (1926) added Darienichthys to the tribe. Later, Rosen & Bailey (1963) recognized Diphyacantha and Darienichthys as junior synonyms of Priapichthtys and removed them from the Cnesterodontini, placing them in the tribe Heterandriini. Rosen & Bailey (1963) also added Phalloptychus to the Cnesterodontini. These authors provided a diagnosis for the tribe based on the following characters: (1) absence of parietals; (2) unforked posttemporal; (3) long gonopodium; (4) unique type of gonopodial appendix at tip of R3; and (5) sexually dimorphic pleural ribs. Among these characters only "the unique type of gonopodium appendix" revealed useful in diagnosing the tribe. The loss of parietals is not unique to Cnesterodontini (sensu Rosen & Bailey, 1963); it also occurs in Pseudopoecilia, Xenodexia, Pamphorichthys, Micropoecilia, and Poecilia reticulata. Besides Phalloceros do possess parietals. An unbranched post-temporal is exhibited by Scolichthys, Cnesterodon, Phallotorynus, and Phalloceros, but Phalloptychus possess a bifid post-temporal. Long gonopodium and sexually dimorphic pleural ribs are also not uniquely derived features shared by the Cnesterodontini sensu Rosen & Bailey, 1963. Ghedotti (2000 was the first to attempt a solution for the recognition or diagnosis of a monophyletic Cnesterodontini. He recognized Tomeurus as a member of the tribe Cnesterodontini. Ghedotti (2000: 53) diagnosed the Cnesterodontini by the following unique and unreversed synapomorphies: "(1) less than six pelvic-fin rays; (2) male pelvic girdle far anterior, under pectoral girdle; (3) paired bony cirri on the third anal-fin ray in males; and (4) dorsoposterior region of lachrymal very narrow. " Ghedotti (2000: 42) argued that "Most of this evidence for a basal position of Tomeurus is based on the morphology of the oral and pharyngeal jaws (T.S. 11, 13, 21, 38, 41) and most of the evidence for placement in the Cnesterodontini is based on the morphology of the gonopodium and gonopodial suspensorium (T.S. 45, 65, 72, 79, 85). The principle of parsimony supports the latter hypothesis and the homoplastic reversals P R O O F S in jaw structure possibly may be explained as a developmental by-product of reduced size. " However, our results (also based on global parsimony) support a rather different view, i. e. the morphology of the oral and pharyngeal jaws supports the assumption that Tomeurus is the most basal poeciliine, the sister-group of remaining members of the subfamily. We have different interpretations for Ghedotti's (2000) above evidence for Tomeurus placement in the Cnesterodontini, which are discussed below. Ghedotti (2000;character 45) argued that absence of gonapophyses in Cnesterodon and Tomeurus are evidence of common exclusive ancestry. However, Cnesterodon species do exhibit gonapophyses, although rudimentary (see character 47). Following Ghedotti (2000;character 79), another putative synapomorphy for Tomeurus + Cnesterodon clade would be the posteriorly inclined position of the proximal anal-fin radials in males. Actually, in Tomeurus and Cnesterodon the gonactinost complex is very inclined backwards to an angle smaller than 45 º relative to the body longitudinal axis (state 68-0). However, on the basis of the present hypothesis of relationships, the condition in Cnesterodon is interpreted as a synapomorphic reversal and as plesiomorphic in Tomeurus. Ghedotti (2000;character 85) also proposed the lateral processes on ventral portions of sixth, seventh, and eighth proximal anal-fin radials in adult males contacting each other as synapomorphic for a clade Tomeurus + Cnesterodon. We believe these structures are non-homologous in both genera. These processes are entirely fused in Tomeurus forming a co-ossified structure, whereas in Cnesterodon lateral processes are present in posterior middle anal-fin rays of males (and not in the proximal ones), which are fused to their respective proximal radials. Ghedotti (2000;character 65) proposed the male pelvic girdle far anterior, under pectoral girdle, as synapomorphic for the clade ((Tomeurus, Cnesterodon) Phalloceros, Phallotorynus)). Our analysis revealed the position of the pelvic girdle is not a useful character in diagnosing the Cnesterodontini. We identified five states of this character, which presented several independent acquisitions and reversals during the history of the Cyprinodontiformes (see character 35).
According to Ghedotti (2000;character 72), the presence of paired bony cirri on tip of R3 is a putative synapomorphy for a Cnesterodon + Phalloceros + Phallotorynus + Tomeurus clade. However, the bony cirri of our Cnesterodontini and Tomeurus seem to represent nonhomologous structures. The bony cirrus or pedicel of R3 of cnesterodontins is attached to R4 (character 90), whereas in Tomeurus, such structure is only attached to R3. Besides the pedicel of cnesterodontins is associated with a membranous appendix, whereas a membranous appendix associated with R3 is lacking in Tomeurus. Ghedotti (2000) recognized the reduced number of pelvic fin-rays (less than six) as synapomorphic for Phalloceros, Phallotorynus, Cnesterodon, and Tomeurus. Our results support the hypothesis that such a reduction independently ap-peared in Tomeurus and in the ancestor of Phalloceros, Phallotorynus, and Cnesterodon.
The naturalness of a tribe Cnesterodontini composed of Phalloceros, Phallotorynus, and Cnesterodon, has already been suggested by Rosen (1959: 496): "(…) the morphological details of the osteocranium and some post-cranial bones of Cnesterodon nevertheless show conclusively that this genus and two others, Phalloceros and Phallotorynus, form a tightly knit group (…)." Rosen & Kallman (1959) suggested that axial and appendicular skeleton resemblances between Tomeurus and Cnesterodon may be attributed to parallel evolution. These authors experimentally demonstrated that the loss or extreme reduction of gonapophyses in Tomeurus and Cnesterodon, respectively, is ontogenetically related to (and can be attributed) the far anterior position of the male pelvic girdle. These conclusions are very congruent to our results, which interpreted the position of the male pelvic girdle and the loss or extreme reduction of gonapophyses in Tomeurus and Cnesterodon as independently acquired.
The phylogenetic position of Tomeurus as the sister-group of remaining poeciliines brings to discussion the evolution of viviparity in the subfamily. Facultative viviparity in Tomeurus could be viewed as (1) a preliminary stage of viviparity towards true viviparity, a stage achieved by the ancestor of all remaining poeciliines; or (2) as an autapomorphic specialized condition of viviparity adaptable for different environmental conditions.
Most generic diagnostic characters for Cnesterodon, Phalloceros, and Phallotorynus provided by different authors (Eigenmann, 1907;Henn, 1916;Rosen & Bailey, 1963;Oliveros, 1983;Rosa & Costa, 1993) have been corroborated by this study. These generic characters as well as the relationships within the genera of the Cnesterodontini will be discussed in detail on oncoming studies that include the taxonomic revisions and phylogeny of the genera Cnesterodon, Phalloceros, Phallotorynus, and Phalloptychus (see Introduction for further details).

Biogeography of poeciliines.
A clear biogeographic explanation for poeciliine distribution is not available, however, few considerations can be speculated. The subfamily Poeciliinae is distributed throughout the Americas. The most basal poeciliine Tomeurus gracilis occurs in coastal drainages of northern South America. The tribes Alfarini, Brachyrhaphini, Priapichthyini, and Priapellini have a mainly Central American distribution, whereas the Heterandriini, and Gambusiini have also invaded North American drainages. The genera Priapichthys, Pseudopoecilia, and Neoheterandria are distributed along Pacific drainages of Panama, Ecuador, Peru, and Colombia, which may represent an endemism area in South America. These genera may have differentiated in isolation provided by the elevation of the Andes. Girardinins are Central American fishes, except for the South American Phalloptychus. An explanation for the distribution of the Girardinini requires the assumption of either a broader preterit distribution, followed by extinction events or a dispersionist hypothesis. Pamphorichthys is distributed in rio Paraguai basin and northern drainages of Brazil (Tocantins, Xingu, São Francisco, Parnaíba, Amazonas, and Tapajós). Since Micropoecilia + "Poecilia" reticulata and Pamphorichthys are sister groups, it can be hypothesized that the ancestor of these fishes inhabit a huge ancient area in northern South American which was split in two areas, which represent the current distribution of Micropoecilia and Pamphorichthys. Cnesterodontins are endemic to southern South America. The pattern of distribution of subsets of the subfamily Poeciliinae may help to tentatively identify putative areas of endemism for poeciliines in the American continent. Those areas are: (1) southern North America; (2) Cuba; (3) the Caribbean and Venezuela; (4) inland Central America; (5) pacific drainages of Panama, Colombia, Ecuador and Peru; (6) coastal drainages of northern South America along the Venezuela, Guyana, Suriname, French Guyana, and Brazil (Amapá and Pará States); (7) Paraguay River basin and northern drainages of Brazil (Tocantins, Xingu, São Francisco, Parnaíba, Amazonas, and Tapajós); and (8)