Phylogeny of the family Characidae (Teleostei: Characiformes): from characters to taxonomy

385 Neotropical Ichthyology, 8(3):385-568, 2010 Copyright © 2010 Sociedade Brasileira de Ictiologia Phylogeny of the family Characidae (Teleostei: Characiformes): from characters to taxonomy Juan Marcos Mirande The family Characidae is the most diverse among Neotropical fishes. Systematics of this family are mainly based on precladistic papers, and only recently a phylogenetic hypothesis for Characidae was proposed by the author. That phylogeny was based on 360 morphological characters studied for 160 species, including representatives of families related to Characidae. This paper is based on that phylogenetic analysis, with the analyzed characters described herein and documented, accompanied by comparisons of their definition and coding in previous papers. Synapomorphies of each node of the proposed phylogeny are listed, comparisons with previous classifications provided, and autapomorphies of the analyzed species listed. Taxonomic implications of the proposed classification and the position of the incertae sedis genera within Characidae are discussed. A discussion of the phylogenetic information of the characters used in the classical systematics of the Characidae is provided. La familia Characidae es la más diversa entre los peces neotropicales. La sistemática de esta familia está basada principalmente en trabajos pre-cladísticos, y sólo recientemente una hipótesis filogenética para Characidae fue propuesta por el autor. Esa filogenia estaba basada en 360 caracteres estudiados en 160 especies, incluyendo representantes de familias relacionadas a Characidae. Este trabajo está basado en ese análisis filogenético, y los caracteres analizados son aquí descriptos y documentados, comparando su definición y codificación con trabajos previos. Las sinapomorfías de cada nodo de la filogenia propuesta son listadas, y se incluyen comparaciones con clasificaciones previas; también se listan las autapomorfías de las especies analizadas. Las implicancias taxonómicas de la clasificación propuesta y la posición de los géneros incertae sedis dentro de Characidae son discutidas. También se presenta una discusión de la información filogenética de los caracteres usados en la sistemática clásica de Characidae. Key words: Ostariophysi, Osteology, Morphology, Classification, Systematics. CONICET-Fundación Miguel Lillo, Miguel Lillo 251, 4000 San Miguel de Tucumán, Argentina. [email protected] Introduction The order Characiformes includes more than 1,800 species, of which the family Characidae is the most diverse, with approximately 1,200 species (Reis et al., 2003); indeed, the Characidae is the fourth most diverse family of fishes, after the Cyprinidae, Cichlidae and Gobiidae (Eschmeyer & Fricke, 2009). Members of the Characidae occur from southern portions of the USA to northern Patagonia in Argentina, being especially diverse in the Amazon, Orinoco, and La Plata River basins. According to the currently accepted phylogenetic hypotheses, based on both morphological and molecular data, the Cypriniformes constitute the sister group of (Characiformes (Siluriformes + Gymnotiformes)) (Fink & Fink, 1981, 1996; Dimmick & Larson, 1996). According to classifications prior to Mirande (2009), the Characiformes consisted of three African families (Citharinidae, Distichodontidae and Hepsetidae) (Géry, 1977, Calcagnotto et al., 2005), 14 Neotropical families (Acestrorhynchidae, Anostomidae, Characidae, Chilodontidae, Crenuchidae, Curimatidae, Ctenoluciidae, Cynodontidae, Erythrinidae, Gasteropelecidae, Hemiodontidae, Lebiasinidae, Parodontidae, Prochilodontidae, and Serrasalmidae) (Reis et al., 2003; Calcagnotto et al., 2005), and one trans-Atlantic family (Alestidae) (Zanata & Vari, 2005). The families Citharinidae and Distichodontidae constitute the suborder Citharinoidei, considered as the sister group of the Characoidei, which includes all the remaining Characiformes (Vari, 1979; Fink & Fink, 1981, 1996; Buckup, 1998; Calcagnotto et al., 2005). Among the Characoidei, the monophyly of a clade composed of the families Anostomidae, Chilodontidae, Curimatidae, and Prochilodontidae (Anostomoidea; Vari, 1983; Buckup, 1998), and a clade formed by the Neotropical families Ctenoluciidae, Erythrinidae and Lebiasinidae and the African family Hepsetidae (Erythrinoidea; Vari, 1995; Buckup, 1998) had been proposed. Relationships between these suprafamilial groups and the remaining Characiformes were unclear, and some hypotheses that conflicted in varying degrees were proposed Phylogeny of the family Characidae 386 (Uj, 1990; Ortí & Meyer, 1997; Buckup, 1998; Calcagnotto et al., 2005, Hubert et al., 2005). Most families of the Characiformes have evidences of monophyly (Weitzman, 1954; Roberts, 1973, 1974; Vari, 1979, 1983, 1995; Buckup, 1998; Toledo-Piza, 2000; Zanata & Vari, 2005), whereas there are no consensus on the monophyly and composition of the Characidae. Most currently recognized subfamilial and generic groups in the Characidae are based on the pre-cladistic papers of Eigenmann (e. g. 1912, 1915, 1917, 1918, 1921, 1927) and Eigenmann & Myers (1929). Eigenmann (1917) was highly influential in terms of our present concepts of relationships within the Characidae. Eigenmann defined 17 characters with discrete alternative states, and used them in different combinations to diagnose the genera in the Characidae, considering the most frequent states as being primitive. The genus Astyanax Baird & Girard has the combination of the most frequent states of all these characters, and it was consequently considered by Eigenmann (1917) as primitive within the family. Eigenmann, however, recognized that the less frequent states could have independent origins in different species of the same genus, producing “polyphyletic” [sic] genera. Given the impossibility to classify the Characidae in a branching scheme, Eigenmann (1917) presented a radial pattern, identifying a “nucleus” of generalized morphology (represented by the genus Astyanax) and different lines of evolution diverging from it. Eigenmann’s classification was followed by Greenwood et al. (1966) and particularly by Géry (e. g. Géry, 1977). Géry also recognized the polyphyletic nature of this classification, and that this systematic scheme failed to reflect the phylogeny (Géry, 1972). In this pre-cladistic systematic classification, most genera of the Characidae, especially those with “generalized” morphology, were included in the subfamily Tetragonopterinae. The remaining genera were distributed across several subfamilies defined by the presence of somewhat arbitrarily chosen characters. Géry (1977), following the general classification of Eigenmann recognized the subfamilies (number of genera in parentheses) Agoniatinae (1), Rhaphiodontinae (2), Characinae (14), Bryconinae (6), Clupeacharacinae (1), Paragoniatinae (6), Aphyocharacinae (1), Glandulocaudinae (18), Stethaprioninae (3), Tetragonopterinae (49), Rhoadsiinae (2) and Cheirodontinae (13 genera sensu stricto and 36 sensu lato). The first genus of the Characidae explicitly diagnosed by shared presumably apomorphic features was Bramocharax Gill (Rosen, 1972). Later, Vari (1977) presented evidence based on shared presumably derived features, supporting the monophyly of the subfamily Iguanodectinae. Weitzman & Fink (1983) explicitly explained the problems related with some generic characters used for the systematic schemes of Eigenmann and Géry, and the needing of a classification reflecting the phylogeny of the Characidae. A series of contributions proposing or corroborating the monophyly of some genera and subfamilies of the Characidae were published subsequently [Serrasalminae (Machado-Allison, 1983), Stethaprioninae (Reis, 1989), Glandulocaudinae (Weitzman & Fink, 1985; Weitzman & Menezes, 1998), Cheirodontinae (Malabarba, 1998a) and Paracheirodon Géry (Weitzman & Fink, 1983), Charax Scopoli (Lucena, 1987), Jupiaba Zanata (Zanata, 1997), Roestes Günther and Gilbertolus Eigenmann (Lucena & Menezes, 1998), Spintherobolus Eigenmann (Weitzman & Malabarba, 1999), Creagrutus Günther and Piabina Reinhardt (Vari & Harold, 1998, 2001), Deuterodon Eigenmann (Lucena & Lucena, 2002), Cyanocharax Malabarba & Weitzman (Malabarba & Weitzman, 2003), Attonitus Vari & Ortega (Vari & Ortega, 2000), and Bryconadenos Weitzman, Menezes, Evers & Burns (Weitzman et al., 2005)]. Most of these papers were focused on particular groups of the Characidae, without enough exploration of their relationships with the remaining Characidae. Malabarba (1998a) restricted the subfamily Cheirodontinae to a subset of the genera recognized in this subfamily by Géry (1977), leaving 33 genera as incertae sedis. Later, Reis (2003a) restricted the Tetragonopterinae to its type genus Tetragonopterus Cuvier, leaving many genera as incertae sedis within the Characidae. Lima et al. (2003) classified also several genera previously included in the subfamilies Bryconinae, Characinae, Cheirodontinae, and Paragoniatinae (Géry, 1977) as incertae sedis within the Characidae. The subfamilies and incertae sedis genera recognized in the last revision of the Characidae are as follows (number of genera in each group in parentheses): incertae sedis (88) (Lima et al., 2003), Agoniatinae (1) (Lima & Zanata, 2003), Clupeacharacinae (1) (Lima, 2003a), Iguanodectinae (2) (Moreira, 2003), Bryconinae (3) (Lima, 2003b), Serrasalminae (15) (Jégu, 2003), Aphyocharacinae (1) (Lima, 2003c), Characinae (12) (Lucena & Menezes, 2003), Stethaprioninae (4) (Reis, 2003a), Tetragonopterinae (1) (Reis, 2003b), Rhoadsiinae (3) (Cardoso, 2003a), Cheirodontinae (15) (Malabarba, 2003), and Glandulocaudinae (19) (Weitzman, 2003). The subfamily Rhaphiodontinae (sensu Géry, 1977) was included into the Cynodontidae (Lucena & Menezes, 1998; Toledo-Piza, 2003), and all the genera assigned to the Paragoniatinae by Géry (1977) were included in the incertae sedis-group (Lima et al., 2003). The Acestrorhynchidae was considered as a separate family (Lucena & Menezes, 1998, 2003). Malabarba & Weitzman (2003) described Cyanocharax and proposed the monophyly of a group of genera (their clade A) including all members of Glandulocaudinae plus several incertae sedis genera. Later the glandulocaudin Lophiobrycon Castro, Ribeiro, Benine & Melo (Castro et al., 2003), and the incertae sedis genera Myxiops Zanata & Akama (Zanata & Akama, 2004), Nantis Mirande, Aguilera & Azpelicueta (Mirande et al., 2004, 2006a), Dectobrycon Zarske & Géry (Zarske & Géry, 2006), and Phallobrycon Menezes, Ferreira & Netto-Ferreira (Menezes et al., 2009) were described. The phylogeny of Calcagnotto et al. (2005) implicitly raised the Serrasalminae to the family level. Weitzman et al. (2005) described the incertae sedis genus Bryconadenos and restricted the Glandulocaudinae to Glandulocauda Eigenmann, Lophiobrycon and Mimagoniates Regan, shifting J. M. Mirande 387 the remaining genera previously in the Glandulocaudinae to the subfamily Stevardiinae. Quevedo (2006) phylogenetically diagnosed the subfamily Paragoniatinae with a composition very similar to that proposed by Géry (1977) (the results of this and other recently completed theses are not discussed, pending their eventual publications). There is no consensus about the phylogenetic relationships among subfamilies of the Characidae. Some phylogenetic analyses of different scope were performed, both from morphological and molecular data. The phylogenies of Uj (1990), Buckup (1991, 1998) and Lucena (1993) were based on morphological data. Uj (1990) did not perform a cladistic analysis; he just mapped character transformations on a “phylogenetic” tree obtained without specific criteria. This unpublished thesis, however, was an advance on compared morphological knowledge of the Characidae. The doctoral theses of Buckup (1991) and Lucena (1993) were the first cladistic analyses of the Characidae, with the main phylogenetic results of Buckup published later (Buckup, 1998). These analyses shared a high proportion of characters, but they had different objectives and, consequently, different taxon sampling. Most of the characters in those theses were analyzed by Mirande (2008, 2009) and discussed in the present paper. As the aim of Buckup (1991, 1998) was to obtain a hypothesis of relationships of the members of the Crenuchidae, he included only six genera of the Characidae. Given that the main objective of Lucena (1993) was to recover the phylogenetic relationships of the Characidae, the taxon sampling reflected the morphological diversity of the family, in a scope more similar to that of Mirande (2008, 2009). Most of the conclusions of Buckup (1991, 1998) and Lucena (1993) are included on the last classification of the Neotropical members of the Characiformes (Reis et al., 2003). Subsequently, several unpublished theses focused on phylogenies of specific groups within the Characidae (Moreira, 2002; Bertaco, 2003; Cardoso, 2003b; Serra, 2003; Benine, 2004; Bührnheim, 2006; Lima, 2006; Quevedo, 2006; Bertaco, 2008). Some of the characters used on these theses were also analyzed by Mirande (2008, 2009). Molecular phylogenies of the Characiformes were proposed by Ortí & Meyer (1997), Hubert et al. (2005), and Calcagnotto et al. (2005). As in the cited morphological studies, the objectives of these analyses differed, and this affected the taxon and gene samplings and the methodologies used. Calcagnotto et al. (2005) published the most comprehensive molecular phylogeny of the Characiformes, including 27 taxa of the Characidae. Mirande (2009) proposed the monophyly and a classification of the Characidae based on a phylogenetic analysis. Most incertae sedis genera were assigned to a subfamily or subfamiliallevel clade, at least tentatively. The paper of Mirande (2009) was, however, mainly concerned with analytical issues, leaving most morphological descriptions and discussions, and comments on the taxonomic implications or the phylogeny for the present contribution. In the present study some characters were redefined or added from Mirande (2009) and the results herein obtained slightly differ to those of that paper. Material and Methods Osteological preparations Osteological preparations were made following Taylor & van Dyke (1985) on one to five specimens of each species included in the analysis, according to their availability and observed intraspecific variability. Some characters involving musculature and soft tissues were observed with the aid of non-permanent Methylene Blue staining. A total of 23 species of 14 characiform families and one cypriniform form the outgroup, while 137 species of the Characidae form the ingroup of this study. Figures 1-124 illustrate most characters and character-states. Most figures are stacks of pictures at different focal depth, constructed with CombineZM software (Hadley, 2006), running under Linux through Wine software. Taxon sampling The taxonomic nomenclature of the Characoidea used in the present paper follows Mirande (2009), while that of remaining Characiformes follows Buckup (1998). Terminal taxa were included in the data matrix at species-level. The only exceptions are the root, and the superfamily Citharinoidei, which actually are compound taxa based primarily on Puntius tetrazona (Bleeker) and Distichodus maculatus Boulenger but allowing for documented variations within the Cyprinidae and Citharinoidei, respectively. Taxon sampling was done considering the inclusion of members of recognizedly monophyletic groups, representatives of the morphological variation within the family, members of the incertae sedis genera, species with special taxonomic interest (e. g. type species of the most diverse genera), and an outgroup including members of most families in the Characiformes. The taxon sampling focused in the inclusion of as many species as possible, with studies of intraspecific variations beyond the scope of this paper. The analyses are rooted on the compound terminal taxon based on Puntius tetrazona (Cypriniformes, Cyprinidae). Cases in which the states observed in this species differed from those considered as plesiomorphic for Cypriniformes (Howes, 1978, 1979, 1980; Vari, 1979; Fink & Fink, 1981, 1996) were coded as polymorphic. Although there is enough consensus on the position of the characiform families Citharinidae and Distichodontidae (Citharinoidei) as the sister group of the remaining Characiformes (Characoidei) (Vari, 1979; Fink & Fink, 1981, 1996, Buckup, 1991, 1998; Calcagnotto et al., 2005), a root external to the Characiformes was used to test also such hypothesis. A rather broad sampling of families related to the Characidae was carried out, to correctly optimize the characters and to test as rigorously as possible the monophyly of the Characidae. This test was improved analyzing some members of the families morphologically closer to Characidae, or included historically in this family, such as the Alestidae, Gasteropelecidae, and Serrasalmidae (Weitzman, 1954; Géry, 1977; Machado-Allison, 1983). Phylogeny of the family Characidae 388 The ingroup is composed of members of all the subfamilies recognized in Reis et al. (2003) excepting the monotypic Clupeacharacinae; however, most of the effort was oriented towards sampling the incertae sedis genera which represented approximately two-thirds of the diversity of the Characidae, prior to Mirande (2009). The taxon termed “undescribed n. gen. and sp.” by Mirande (2009) proved to be an undescribed species of Oligosarcus Günther. Thus, in this paper this species is named as Oligosarcus sp. leaving its description and discussion of relationships to be published elsewhere. The species named as Bryconamericus beta Eigenmann by Mirande (2009) is referred to as B. alpha Eigenmann in this paper, following the synonymy proposed by Román-Valencia (2003). The species named as Roeboides bonariensis (Steindachner) and R. paranensis by Mirande (2009) are referred to as R. microlepis (Reinhardt) and R. descalvadensis Fowler, following the synonymies proposed by Lucena (2003, 2007). Finally, the specimens referred to as Hemigrammus cf. rhodostomus Ahl by Mirande (2009) proved to be H. bleheri Géry & Mahnert. The list of examined material is shown in the Appendix 1. All species (with the exception of Brycon meeki Eigenmann & Hildebrand that was coded following Weitzman, 1962) were observed by the author. The coding of each species was made upon all its available information. If a particular state was observed in a species, but published data indicate the alternative condition in that species, such species were coded as polymorphic. Meristic characters (e. g. anal-fin rays counts) were coded according to the ranges cited in the literature. Nomenclature and Abbreviations Abbreviations mentioned on the list of examined material are as following: AI (Asociación Ictiológica, La Plata), ANSP (Academy of Natural Sciences of Philadelphia), CI-FML (Colección Ictiológica de la Fundación Miguel Lillo, Tucumán), LACMNH (Los Angeles County Museum of Natural History), MCNi (Colección ictiológica del Museo de Ciencias Naturales, Salta), MCP (Museu de Ciências e Tecnologia da Pontifícia Universidade Católica do Rio Grande do Sul, Porto Alegre), MHNG (Muséum d’histoire naturelle, Genève), MNHN (Muséum national d’histoire naturelle, Paris), and MZUSP (Museu de Zoologia da Universidade de São Paulo). Osteological nomenclature follows Weitzman (1962) with the modifications adopted by Zanata & Vari (2005), which are based principally on Nelson (1969), Patterson (1975), and Fink & Fink (1981, 1996). Abbreviations in character definitions are references to the following papers: EI (Eigenmann, 1917), FF (Fink & Fink, 1981, 1996), AM (Machado-Allison, 1983), UJ (Uj, 1990), VA (Vari, 1995), BU (Buckup, 1998), LU (Lucena, 1993), LC (Lucena, 1998), MA (Malabarba, 1998a), WM (Weitzman & Menezes, 1998), LM (Lucena & Menezes, 1998), CM (Malabarba, 1998b), TP (Toledo-Piza, 2000), VH (Vari & Harold, 2001), MO (Moreira, 2002), VB (Bertaco, 2003), CA (Cardoso, 2003b), SE (Serra, 2003), BE (Benine, 2004), ZV (Zanata & Vari, 2005), BÜ (Bührnheim, 2006), LI (Lima, 2006), QU (Quevedo, 2006), PZ (Toledo-Piza, 2007), MW(Menezes & Weitzman, 2009). The number following these abbreviations refers to the character number as used on the cited analysis; those cases in which the character states were modified from the cited paper are indicated with a “m”, and those instances where the ordering of states were inverted are indicated with an “i”. Some numbers of the list of characters of Malabarba (1998b) do not correspond with those of the data matrix; instead their correspondence was deduced from the information given in the text. In these cases the number in parentheses corresponds to the one deduced to have each particular character in the data matrix. The principal objective of the proposed taxonomic nomenclature is to classify members of the Characidae in monophyletic units. The proposed nomenclature is as conservative as possible concerning to the creation of new names for taxonomic groupings, with all the names used in the recent literature which are compatible with the obtained phylogeny retained other than in cases when their preservation necessitates the creation of a number of new taxa. The new suprageneric names are rooted on the first described genus included within the clade. An evaluation of the monophyly and phylogeny of all genera is beyond the scope of this paper; therefore, new generic names are not proposed nor are species reassigned between genera. Cladistic methodology Additive characters were recoded as binaries and are represented by two or more character numbers; this improve greatly the efficiency of searches under self-weighting optimization (Goloboff, 1997) in terms of time and optimality. Binary coding of the additive characters has no effects on the results obtained under implied weighting and relatively small influence to the results under self-weighted optimization (see Mirande, 2009 for details). Conditions that resulted as intermediate between the defined states were coded as polymorphisms; although both situations are conceptually different, it was preferred over coding them as inapplicable or missing entries. Analyses were performed by parsimony, following the methods described by Hennig (1966) and developed by Farris (e. g. 1969, 1970, 1983) among others. Analyses under implied weighting (Goloboff, 1993) and selfweighting optimization (Goloboff, 1997) were performed with TNT software (Goloboff et al., 2003a, 2008). Details of this analysis were described elsewhere (Mirande, 2008, 2009), and they are not treated here. In this analysis the number of explored conditions were almost duplicated from Mirande (2009). In the present study, 21 values of k were used under each of the weighting schemes (vs. 11 in Mirande, 2009). Measures of stability and support are expressed in the discussion of each node. Stability measures consider all the range of explored parameters (see Mirande, 2009), while support measures were calculated for k = 13, under implied weighting. Those measures are, respectively, GC values as stability measures, relative frequencies, GC values as support measures (Goloboff et al., 2003b), and relative Bremer support (Bremer, 1994; Goloboff & Farris, 2001). Cases in which the support measures are (artificially) negative are indicated with a dash (–), whereas stability measures are indicated as negative. J. M. Mirande 389 Results and Discussion Description of phylogenetic characters Most analyzed characters are osteological (90%), while the remaining ones come from coloration, external features and reproductive biology. Of these, 135 were not described previously in the literature (published or not), and represent new definitions. Some characters about bony hooks on fins of adult males were redefined from Mirande (2009). The characters proposed by Menezes & Weitzman (2009) to be evidence for the monophyly of their Glandulocaudinae and Stevardiinae are herein analyzed together with the characters from Mirande (2009). Also, several missing entries in the analysis of Mirande (2009) were coded for this study. With that modifications, the data matrix herein analyzed has 365 characters and is provided as Appendix 2. Neurocranium Epiphyseal bar: 1. Posterior laminar expansion of epiphyseal bar: (0) absent; (1) present. (LU13i, LC4). The frontals articulate each other via the epiphyseal bar, which transversely crosses the medial cranial fontanel. In most examined species, the epiphyseal bar is slender and approximately cylindrical in cross-section (state 0; Fig. 1), while a group of species has a laminar projection on the posterior margin of the epiphyseal bar, making it proportionally broader in dorsal view (state 1; Fig. 2). Lucena (1993, 1998) coded the presence of this expansion in several characins and Rhaphiodon vulpinus Agassiz. A small laminar expansion was also observed in examined specimens of Charax stenopterus (Cope) and Galeocharax humeralis (Valenciennes). Since Lucena (1998) noted the absence of such an expansion in these species, they are coded as polymorphic. Serrasalmus maculatus Kner has a broad epiphyseal bar that apparently lacks the laminar expansion. This condition is considered to be different from the states herein defined for this character, and this character is coded as inapplicable to S. maculatus. This character is also considered as inapplicable to the species in which the fontanel is completely covered by the frontals. Basioccipital: 2. Ventral longitudinal lamellae of basioccipital: (0) falling short of posterior border of basioccipital; (1) reaching posterior border of cranium. (PZ24). The prootic and basioccipital have two bilateral lamellae articulating with two longitudinal dorsal processes of the parasphenoid, forming the limits of the posterior myodome where part of the extrinsic musculature of the eye attaches, as described by Weitzman (1962: 24). In most examined members of the Characidae, these lamellae are restricted to the area of contact between the basioccipital and parasphenoid, and the surface of the basioccipital lacks any bilateral lamellae or ridges posterior to that region (state 0; Fig. 3). In most members of the outgroup and some of the Characidae these lamellae extend posteriorly to the parasphenoid in the region ventral to the lagenar capsules and reach the posterior margin of the cranium (state 1; Fig. 4). Lagenar capsule: 3. Ventral projection of lagenar capsule: (0) not extending ventrally to horizontal through articulation between basioccipital and parasphenoid; (1) extending ventrally to articulation between basioccipital and parasphenoid. (UJ5m, ZV47m, PZ25m). The extension of the lagenar capsules lateral to the cranial condyle is a synapomorphy of the Characiformes according to Fink & Fink (1981) and was observed in all the characiforms herein examined. The ventral extension of these capsules, in contrast, is variable among the examined species. In most species of the outgroup, the lagenar capsules do not extend ventrally to the articulation between the basioccipital and parasphenoid (state Fig. 1. Neurocranium of Bario steindachneri, MHNG 2184.46, 62.0 mm SL, dorsal view, anterior to left. EPB: epiphyseal bar, FRO: frontal, MES: mesethmoid. Scale bar = 1 mm. Fig. 2. Cranium of Roeboides microlepis, CI-FML 3881, 82.9 mm SL, dorsal view, anterior to left. EPB: epiphyseal bar, ETC: ethmoid cartilage, FRF: frontal fontanel, FRO: frontal, IO6: sixth infraorbital, LET: lateral ethmoid, MES: mesethmoid, NAS: nasal. Scale bar = 1 mm. Phylogeny of the family Characidae 390 0; Fig. 5), while in most species of the Characidae and some members of the outgroup these capsules are conspicuously extended, continuing ventrally beyond the area of articulation of those bones (state 1; Figs. 6 and 7). 4. Epioccipital bridge over posttemporal fossa: (0) absent; (1) present. (BÜ7i). The posttemporal fossa in Characiformes is longitudinally crossed by the epioccipital bridge (state 1; Fig. 6) except in some miniature species of the Characidae and Crenuchidae (state 0; Weitzman & Fink, 1983: figs. 6, 8, 15, and 17). Although this character was considered to be related with miniaturization, its phylogenetic value has to be tested. The absence of an epioccipital bridge was herein observed only in Hasemania nana (Lütken) and Pyrrhulina australis Eigenmann & Kennedy. This character, however, is variable in the two examined specimens of the latter species, which is coded as polymorphic. 5. Form of epioccipital bridge: (0) cylindrical or vertically expanded in transverse section; (1) depressed in its middle region. The epioccipital bridge over the posttemporal fossa is usually cylindrical or slightly expanded dorsally (state 0; Fig. 6). In the studied specimens of Bramocharax bransfordii Gill, Oligosarcus bolivianus (Fowler), O. cf. jenynsii (Günther), and O. sp., the middle region of this bridge is dorsoventrally depressed (state 1; Fig. 8). 6. Anterior articulation of epioccipital bridge: (0) with both parietal and pterotic; (1) only with parietal. In most examined species the anterior region of the epioccipital bridge articulates with the parietal and pterotic (state 0; Fig. 6). In Grundulus cochae (Humboldt) and Paracheirodon axelrodi (Schultz), the epioccipital bridge is displaced dorsally and its anterior portion articulates only with the parietal (state 1; Weitzman & Fink, 1983: figs. 4, 5, and 7). 7. Posteriorly-oriented epioccipital spine: (0) present; (1) absent. (LU21m, LC3i). Most examined species lack projections on the posterior surface of the epioccipital (state 1; Fig. 5). A posterior projection of the epioccipital aligned with the epioccipital bridge that serves as a point of attachment of the epaxial musculature was observed in some species (state 0; Fig. 7). In Brycon orbignyanus (Valenciennes) and Salminus Fig. 4. Posterior region of neurocranium of Triportheus nematurus, CI-FML 3948, 82.1 mm SL, ventral view, anterior to left. BOC: basioccipital, PSP: parasphenoid, VLB: ventral lamellae of basioccipital. Scale bar = 1 mm. Fig. 5. Posterior region of neurocranium and anterior vertebrae of Piabucus melanostomus, CI-FML 3894, 67.0 mm SL, lateral view, anterior to left. BOC: Basioccipital, EPO: epioccipital, HYO: hyomandibula, NCO: neural complex of Weberian apparatus, PSP: parasphenoid, PTO: pterotic, SOC: supraoccipital. Scale bar = 1 mm. Fig. 3. Neurocranium of Bario steindachneri, MHNG 2184.46, 62.0 mm SL, dorsal view, anterior to left. BOC: basioccipital, EXO: exoccipital, INT: intercalar, PTO: pterotic. Scale bar = 1 mm. J. M. Mirande 391 opening was described in the Citharinidae and Distichodontidae by Vari (1979). This opening was later referred as ventromedial opening of the posttemporal fossa by Buckup (1991, 1998) and Lucena (1993). This opening is situated posteriorly on the cranium, and is margined by the epioccipital and exoccipital or completely contained by the epioccipital (state 1; Zanata & Vari, 2005: fig. 10). The ventromedial opening of the posttemporal fossa was observed by Buckup (1998) and Lucena (1993) in members of the Alestidae, Crenuchidae, Curimatidae, Cynodontidae, Hemiodontidae, and Parodontidae. Vari & Harold (2001) and Zanata & Vari (2005) defined the different positions of this opening as two separate states, which is treated in the following character. Benine (2004) cited the presence of this opening in Moenkhausia barbouri Eigenmann, M. dichroura (Kner), and M. intermedia Eigenmann; however, this opening is absent in the examined species of Moenkhausia Eigenmann, and they are herein coded as state 0. According to Lucena (1993) this opening is present in Acestrorhynchus pantaneiro Menezes, although in the examined specimen it is absent, and the species is consequently coded as polymorphic. 9. Position of ventromedial opening of posttemporal fossa: (0) between epioccipital and exoccipital; (1) bordered entirely by epioccipital. (ZV43m, VH45m). As previously mentioned, the ventrolateral opening of the posttemporal fossa is limited by the epioccipital and exoccipital in some species (state 0; Vari, 1979: fig. 15) while it is completely contained within the epioccipital in others (state 1; Roberts, 1974: figs. 5 and 59; Zanata & Vari, 2005: fig. 10). This opening Fig. 6. Posterior region of neurocranium and anterior vertebrae of Astyanax chico, CI-FML 3913, 60.2 mm SL, lateral view, anterior to left. BOC: basioccipital, EPO: epioccipital, LET: lateral ethmoid, OBB: olfactory bulb of brain, OLN: olfactory nerve, OSP: orbitosphenoid, PAR: parietal, PTO: pterotic, PSP: parasphenoid, PTS: pterosphenoid, RSP: rhinosphenoid, SPH: sphenotic, TP3: transverse process of third neural arch, TRI: tripus. Scale bar = 1 mm. Fig. 7. Posterior region of neurocranium and anterior vertebrae of Triportheus nematurus, CI-FML 3948, 82.1 mm SL, lateral view, anterior to left. BOC: basioccipital, EPO: epioccipital, HYL: hyomandibular ligament insertion site, NP3: neural pedicle of third vertebra, NS4: neural spine of fourth vertebra, NCO: neural complex of Weberian apparatus, PSP: parasphenoid, SN1: first supraneural, TP3: transversal process of third neural arch, TRI: tripus. Scale bar = 1 mm. brasiliensis (Cuvier), the posterior tip of this process is rounded, differing from most species with state 0, in which it is pointed; these species are coded with state 0 regardless this difference. Weitzman (1962: fig. 3) illustrated a small lobe in Brycon meeki, similar to that herein observed in B. falcatus Müller & Troschel, B. pesu Müller & Troschel, Bryconexodon juruenae Géry, Cynopotamus argenteus (Valenciennes), Hemibrycon dariensis Meek & Hildebrand, Markiana nigripinnis (Perugia) and Moenkhausia xinguensis (Steindachner). These species are coded as polymorphic. Contrary to the observations of Lucena (1993), in the examined species of Aphyocharax Günther this process is absent, and they are coded as state 1. Puntius tetrazona, as in all other Cypriniformes, lacks a posttemporal fossa, and the form of the epioccipital differs slightly. This character was coded as inapplicable to the root of this analysis. In Carnegiella strigata (Günther) and Thoracocharax stellatus (Kner), the epineurals extend anteriorly to the cranium, reaching a position occupied by this spine when present. Indeed, the anteriormost epineurals are fused with the epioccipital. This character is also coded as inapplicable to these species. 8. Ventromedial opening of posttemporal fossa: (0) absent; (1) present. (UJ24, BU19, LU20, VH45m, TP23, BE17, ZV43m, LI35). Most examined species have only two openings of the posttemporal fossa situated posterolateral in the cranium, with these separated by the epioccipital bridge (state 0). A third Phylogeny of the family Characidae 392 is completely enclosed by the epioccipital in some species of Creagrutus (Vari & Harold, 2001) and most members of the Alestidae (except Chalceus Cuvier, among taxa examined here), Curimatidae, Hemiodontidae, and Parodontidae (Roberts, 1974; Zanata & Vari, 2005). A third opening partially margined by the exoccipital was cited for members of the Citharinidae, Crenuchidae, Cynodontidae, and Distichodontidae (Vari, 1979; Zanata & Vari, 2005). Species in which this opening is absent are coded as inapplicable to this character. Sphenotic: 10. Length of sphenotic spine: (0) not extending ventrally to articulation between sphenotic and hyomandibula; (1) extending ventrally to articulation between sphenotic and hyomandibula. (VB28m, VB29m). The sphenotic spine extends ventrally to the fossa for the dilator operculi, being bordered posteriorly by that muscle and partially by the levator arcus palatini. The ventral development of this spine is variable among the examined species. In some species it falls short of the ventral margin of main body of the sphenotic (state 0), whereas in other species this spine is longer, anteriorly bordering the levator arcus palatini and ventrally extending past the region of articulation of the sphenotic with the hyomandibula (state 1; Fig. 9). This character was only coded with states 0 or 1 in the species in which the sphenotic spine is clearly either not reaching or surpassing the articulation with the hyomandibula, respectively. In the examined specimens of Distichodus maculatus, Acestrorhynchus pantaneiro, Carlana eigenmanni (Meek), Mimagoniates rheocharis Menezes & Weitzman, Bryconamericus alpha, Cyanocharax alburnus (Hensel), Hemibrycon dariensis, Knodus breviceps Eigenmann, Odontostoechus lethostigmus Gomes, Aphyocharacidium bolivianum Géry, Aulixidens eugeniae Böhlke, Axelrodia lindeae Géry, Exodon paradoxus Müller & Troschel, Hemigrammus bleheri, Hollandichthys multifasciatus Eigenmann & Norris, Nematobrycon palmeri Eigenmann, Oligosarcus bolivianus, Parecbasis cyclolepis Eigenmann, Probolodus heterostomus Eigenmann, and Thayeria obliqua Eigenmann, the sphenotic spine is hardly reaching the articulation between this bone with the hyomandibula, and they coded as polymorphic. In Puntius tetrazona this spine reaches the ventral limit of the sphenotic, but it is variable in the Cypriniformes (Howes, 1978) and consequently the root of this analysis is also coded as polymorphic. In Astyanax lineatus (Perugia) this character is apparently variable during the growth. Examined juvenile specimens have state 1, while the adults have state 0, and this species is coded as polymorphic. 11. Position of sphenotic spine relative to hyomandibula: (0) rather aligned with anterior margin of hyomandibula; (1) displaced anteriorly relative to anterior margin of hyomandibula. In most examined species, the sphenotic spine is aligned or slightly anterior to the anterior margin of the hyomandibula (state 0; Fig. 9), while in a group of species, such spine is Fig. 9. Cranium of Hyphessobrycon socolofi, CI-FML 3933, 37.7 mm SL, lateral view, anterior to left. HYO: hyomandibula, IO2-3: second and third infraorbitals, SPH: sphenotic. Scale bar = 1 mm. Fig. 8. Posterior region of neurocranium and anterior vertebrae of Oligosarcus bolivianus, CI-FML 3366, 83.4 mm SL, lateral view, anterior to left. EPO: epioccipital, NA4: neural arch of fourth vertebra, NCO: neural complex of Weberian apparatus, NP3: neural pedicle of third vertebra, SOC: supraoccipital. Scale bar = 1 mm. J. M. Mirande 393 Lateral ethmoid: 14. Form of anterior process of lateral ethmoid: (0) broad in ventral view, contacting proximal region of vomer in its entire length; (1) slender and separated from vomer. (CM10m). In most examined species the lateral ethmoid has an anterior process oriented in the direction of the vomer. In many species of the outgroup and some examined members of the Characidae, this process is broad in ventral view and contacts the entire length of the parasphenoid and vomer in the region anterior to the main body of the lateral ethmoid (state 0). In most members of the ingroup this process is, in contrast, comparatively more slender and, as a consequence, leaves a broad space between the lateral ethmoid process and the lateral margin of the posterior portion of the vomer (state 1; Fig. 12). In Gymnocharacinus bergii Steindachner the process is much reduced, and this character is coded as inapplicable. In Aulixidens eugeniae and Engraulisoma taeniatum Castro this process is displaced medially, and it contacts the parasphenoid and vomer in its entire length. Although the origin of this condition seems to be different, resulting in a contact due to a different mechanism (medial displacement, rather than broadening of the process), these species are tentatively coded with state 0. 15. Lateral opening between ventral diverging lamellae of mesethmoid and anterior process of lateral ethmoid: (0) broad; (1) small, ovate and partially occluded by diverging lamellae of mesethmoid and anterior process of lateral ethmoid. (LU2). The anterior process of the lateral ethmoid is situated approximately in the same plane as the corresponding diverging lamella of the mesethmoid (Weitzman, 1962) leaving, in most cases, a broad space between these structures, which is evident in lateral view (state 0; Fig. 13). In a few examined species, the diverging lamellae are much developed ventrally and both the vomer and the anterior process of the lateral ethmoid are expanded dorsally, with both articulating broadly with the mesethmoid. As a result the space delimited by these structures has an ovate shape and is much reduced compared with state 0 (state 1; Fig. 14). This character is coded as inapplicable in species lacking ventral lamellae of the mesethmoid. 16. Dorsal margin of lateral ethmoids: (0) aligned; (1) situated obliquely in dorsal view, converging in an anteriorly directed angle. The dorsal margin of lateral ethmoids articulate with the frontals and, usually, with the ventral diverging lamellae of the mesethmoid. The medial region of the lateral ethmoids articulate with the roof of the mesethmoid through a cartilage. In most examined species the medial portion of the lateral ethmoids form a rather straight line from dorsal view, with its margin visible through the frontals and/or mesethmoid (state 0; Figs. 15 and 16). In a relatively small group of species the medial portions of the lateral ethmoids meet each other along an anterior angle (state 1; Figs. 17 and 18). Apparently such configuration of the lateral ethmoids allows an anterior displacement of the extrinsic musculature of the eye, which inserts in the anterior Fig. 10. Posterior region of cranium of Galeocharax humeralis, CI-FML 3951, 94.6 mm SL, lateral view, anterior to left. HYO: hyomandibula, IO3-5: third to fifth infraorbitals, SPH: sphenotic. Scale bar = 1 mm. anterior to the margin of the hyomandibula (state 1; Fig. 10). Bryconexodon juruenae and Hollandichthys multifasciatus have intermediate states that are coded as polymorphisms. 12. Position of sphenotic spine relative to the orbit: (0) bordering orbit posteriorly and aligned with anterior border of fourth and fifth infraorbitals; (1) distinctly posterior to orbital margin. As stated in the previous character, the sphenotic spine is usually aligned with the anterior margin of the hyomandibula, thereby forming the posterior margin of the orbit (state 0). In a group of species the sphenotic spine is situated posterior to the anterior margin of the fourth and fifth infraorbitals and distant from the posterior margin of the orbit (state 1). 13. Temporal fossa: (0) well developed; (1) absent or much reduced. (VH41, LI27). In most examined species the entire anterior margin of the pterotic articulates with the sphenotic, without an intervening space between these bones, or only a small pore (state 1; Figs. 6 and 11). The temporal fossa is an opening limited anteriorly by the sphenotic and posteriorly by the pterotic, and is present in some examined species (state 0; Weitzman, 1962: fig. 3). Vari & Harold (2001) reported the presence of this fossa in Piabina argentea Reinhardt and several species of Creagrutus not analyzed herein. Although Lima (2006) mentioned its absence in Brycon falcatus, among other species of the genus, this fossa is present in the examined specimen of this species, and is coded as polymorphic. In the examined specimens of Astyanax troya Azpelicueta, Casciotta & Almirón and Piabina argentea, this fossa has a size intermediate to the defined character states, and these species are coded as polymorphic for this character. In the examined specimens of Markiana nigripinnis the presence of this fossa is variable and this species is also coded as polymorphic. Phylogeny of the family Characidae 400 Paracheirodon axelrodi, Parecbasis cyclolepis, and Phenagoniates macrolepis, the ventral diverging lamellae converge with the nasal septum, but do so anteriorly rather than posteriorly, and the olfactory capsules are separated each other, at least partially, by the medial longitudinal nasal septum of the mesethmoid. This situation is coded as state 0. This character is inapplicable for species in which the ventral diverging lamellae of the mesethmoid are reduced or absent. 32. Nasal septum of mesethmoid: (0) single longitudinal lamella; (1) two parallel lamellae apparently formed, in part, by ventral diverging lamellae. As previously mentioned, in most examined species the ventral diverging lamellae of the mesethmoid are independent each other, at least posteriorly, and the nasal septum is formed by a medial single lamella attached dorsally to the ventral surface of the mesethmoid (state 0; Figs. 1 and 28). In some species the nasal septum is formed by two closely-positioned and parallel lamellae that articulate posteriorly through cartilages with the medial region of lateral ethmoids. This condition can be observed dorsally through the somewhat transparent dorsal lamella of the mesethmoid (state 1; Fig. 18). As mentioned under character 30, in these species the ventral diverging lamellae of the mesethmoid are absent as separate structures, but probably partially form the composite nasal septum. Given that the identity of the ventral diverging lamellae as part of this nasal septum was not corroborated herein, this character is considered different from character 31. This character is coded as inapplicable to Aphyocharax nattereri in which the whole nasal septum is much reduced and Phenagoniates macrolepis in which the posterior portion of the lamellae forming the nasal septum diverge slightly at their posterior tips resulting in a not directly comparable condition. In Heterocharax macrolepis, Hoplocharax goethei, and Lonchogenys ilisha, the nasal septum resembles state 1, but the ventral diverging lamellae are present as separate structures. This character is coded as inapplicable to these species, pending future studies. The origin and homologies of the different structures forming the nasal septum and their relationships with the olfactory capsules remain to be studied in greater detail. Nasal: 33. Nasal: (0) present; (1) absent. (ZV17). The nasal bone is present in almost all the Characiformes as a tubular bone lateral to the mesethmoid (state 0; Figs. 2 and 21); its absence was cited among the examined phylogenies only in the alestid Lepidarchus adonis Roberts (Zanata & Vari, 2005) (state 1; Fig. 27). This bone is present in all the examined species except for Coptobrycon bilineatus (Ellis) and Hyphessobrycon elachys Weitzman. Absence of an ossified nasal is probably associated with miniaturization, although this bone is present in species of smaller adult sizes than Coptobrycon bilineatus. 34. Bony lamellae bordering sensory canal of nasal: (0) absent or more slender than tubular region; (1) wider at some point than tubular region. (VA17, LU31, MO48m, LI7, PZ10). In most examined species the nasal is reduced to a tubular bone, lacking or with distinctly small associated lamellae (state 0; Figs. 2 and 21). Vari (1995) reported the presence of lamellae bordering the sensory canal of the nasal dorsally and ventrally (state 1) in members of the families Ctenoluciidae, Erythrinidae, Hepsetidae, and Lebiasinidae. Orbitosphenoid: 35. Synchondral articulation between lateral ethmoid and anterodorsal border of orbitosphenoid: (0) present; (1) absent, with orbitosphenoid distant from lateral ethmoid. The anterodorsal tip of the orbitosphenoid (the orbitosphenoid wing, sensu Weitzman, 1962: 20) is usually distant from the lateral ethmoid, and the area between these bones is filled by the posterior projection of the ethmoid cartilage (sensu Weitzman, 1962: 20), which also limits anterolaterally the anterior myodome (state 1; Fig. 6). In some species, in contrast, the anterior margin of the orbitosphenoid is much closer to the lateral ethmoid and these bones articulate synchondrally (state 0; Fig. 29). This character is coded as inapplicable for Puntius tetrazona which, as is usual in the Cypriniformes, has an extensive contact between the orbitosphenoid and lateral ethmoid along the entire anterior margin of the orbitosphenoid. In Alestes cf. macrophthalmus Günther, Brycinus carolinae (Paugy & Levéquè), and Bryconaethiops macrops Boulenger, the orbitosphenoid articulates anteriorly with the lateral ethmoid, but contact in the region lateral of the olfactory nerve and ventral to the region considered in this character. These species are correspondingly coded as state 1. In the examined specimens of Bryconamericus cf. rubropictus this character is variable and is coded here as polymorphic. Fig. 27. Anterior region of cranium of Coptobrycon bilineatus, MCP 39051, 28.9 mm SL, dorsal view, anterior to left. MES: mesethmoid. Scale bar = 0.5 mm. J. M. Mirande 401 36. Lateral bony coverage of olfactory nerve: (0) absent; (1) covered by posterior expansion of lateral ethmoid; (2) covered by an anterior tubular projection of orbitosphenoid; (3) covered laterally and ventrally by orbitosphenoid and lateral ethmoid, which do not form canal. (UJ13m, ZV29m). The orbitosphenoid usually has, from anterior view, an anterior concavity partially containing the olfactory bulb of the brain, with the olfactory nerve directed anteriorly from the bulb to the olfactory capsule through a foramen in the lateral ethmoid. In most examined species, the olfactory bulb and nerve have no lateral bony coverage and they are visible through the orbit after the eye is removed (state 0; Fig. 13). In some species, the olfactory nerve is instead covered laterally in different modes. In the Parodontidae the lateral ethmoid has a posterior projection that laterally covers the olfactory nerve (state 1; Roberts, 1974: figs. 57, 61, and 63). In most species of the Alestidae, the olfactory nerves are covered by a tubular anterior projection of the orbitosphenoid that reaches the lateral ethmoid (state 2; Zanata & Vari, 2005: fig. 9). In Puntius tetrazona there is an extensive articulation between the orbitosphenoid and the lateral ethmoid that completely covers the olfactory bulb and tract. Distichodus Müller & Troschel and Xenocharax Günther, as is general in the Citharinoidei (Vari, 1979; Zanata & Vari, 2005), have a similar condition, but the lateral coverage of the anterior portion of the brain is not complete (state 3). The lateral coverage of the olfactory nerve by a tubular projection of the orbitosphenoid was considered as typical of the Alestidae by Géry (1977), and it was proposed as a synapomorphy for this family by Murray & Stewart (2002), and a synapomorphy for the African alestids by Zanata & Vari (2005). 37. Form of orbitosphenoid: (0) slender, relatively small and separate from parasphenoid; (1) massive, almost reaching parasphenoid ventrally. (UJ12, UJ36, LC1, TP18, BE25, BÜ9). In most examined species of the Characidae the orbitosphenoid is slender, and its ventral margin is distant from the parasphenoid (state 0; Fig. 25), while in most members of the outgroup and some species of the Characidae, the orbitosphenoid is relatively massive and its ventral margin is close to the parasphenoid (state 1; Fig. 14). In the examined specimens of Acestrocephalus sardina, Chalceus macrolepidotus Cuvier, Hollandichthys multifasciatus, and Salminus brasiliensis the orbitosphenoid has an intermediate size and these species are coded as polymorphic. 38. Distance between posterodorsal margin of ethmoid cartilage and lateral ethmoids: (0) contacting, or almost contacting, lateral ethmoids; (1) distant from lateral ethmoids. The dorsal region of the orbitosphenoid is margined anteriorly by a cartilage, the ethmoid cartilage of Weitzman (1962), which is arched in dorsal view and anteriorly limits the olfactory bulb of the brain and posteriorly the anterior myodome. This myodome contains part of the extrinsic musculature of the eye which attaches principally to the posterior wall of the lateral ethmoid. The position of the olfactory bulb relative to the lateral ethmoid and, consequently, the development and position of the anterior myodome varies among the examined species. The most obvious feature reflecting these differences is the position of the ethmoid cartilage, which is visible dorsally through the frontals. In most examined species the arch formed by this cartilage contacts, or almost contacts, the medial region of the lateral ethmoids (state 0; Fig. 2). In other species the ethmoid cartilage is distant from the lateral ethmoids and it is, instead, connected to the medial region of the lateral ethmoids through a longitudinal cartilage extending dorsally from the rhinosphenoid, when present (state 1; Fig. 15). In the Cypriniformes and Siluriformes the olfactory bulbs are broadly separated from the telencephalon and situated just posterior to the olfactory organ (Harder, 1975). Although this condition is not congruent with any of the states defined herein, it represents an opposite state to that described on state 1. Thus, the root is herein coded as state 0. State 1 was Fig. 28. Mesethmoid of Bryconamericus cf. rubropictus, CI-FML 3902, 47.3 mm SL, ventral view, anterior to left. Scale bar = 0.5 mm. Fig. 29. Cranium of Aphyocharax dentatus, CI-FML 3035, 53.2 mm SL, ventrolateral view, anterior to left. IO2-4: second to fourth infraorbitals, LET: lateral ethmoid, MAX: maxilla, OSP: orbitosphenoid, POP: Preopercle. Scale bar = 1 mm. Phylogeny of the family Characidae 402 observed in a juvenile specimen of Salminus brasiliensis; however, Zanata & Vari (2005) described an anterior displacement of the olfactory bulb in Salminus Agassiz during the growth, and this character is coded as polymorphic for this species. 39. Opening between orbitosphenoid and pterosphenoid: (0) present, rounded or ovate, usually margined by frontal dorsally; (1) absent. (AM15m). In most members of the outgroup the posterior margin of the orbitosphenoid is broadly articulated with the anterior margin of the pterosphenoid, leaving no gaps between these bones (state 1). Almost all the examined characids instead have incomplete articulations between these bones, resulting in an opening margined anteriorly by the orbitosphenoid, posteriorly by the pterosphenoid, and dorsally by the orbital lamella of the frontal (state 0; Figs. 6 and 30). Weitzman (1962) described a small foramen situated between the orbitosphenoid and pterosphenoid for Brycon meeki, which serves as passage for the trochlear nerve. The correspondence of the opening discussed herein and that described by Weitzman (1962) was not corroborated, and Brycon meeki is coded as a missing entry. A similar opening between the orbitosphenoid and pterosphenoid (i. e. not limited by the frontal), is present in Markiana nigripinnis and the examined species of Triportheus Cope. These species are coded with the state 0. The presence of this opening is variable in the examined specimens of Hoplocharax goethei and this species is coded as polymorphic. Parasphenoid: 40. Anterior paired projections of parasphenoid: (0) absent; (1) present. (BE26). The anterior region of the parasphenoid articulates with the posterior lamella of the vomer and the ventral margin of the lateral ethmoids. Usually, the lateral edges of the parasphenoid are entire and parallel in all their extent across the orbit, lacking any projections (state 0). The parasphenoid of some examined species instead has paired acute processes situated near its anterior end, oriented towards the posteroventral margin of each lateral ethmoid (state 1; Fig. 12). Benine (2004) defined three states for this character, considering an intermediate situation, in which such processes are present but reduced in size. Only the species in which these processes are present and well developed in all the examined specimens are herein coded as state 1, leaving as polymorphic those species in which these process are of variable occurrence or much reduced. Parietal: 41. Parietal fontanel: (0) present in adults; (1) absent in adults. (BU15, LU12, ZV37, PZ16). The parietal fontanel, when present, is limited laterally by the frontals and parietals, anteriorly by the epiphyseal bar, and posteriorly by the supraoccipital (state 0; Fig. 26). In some species, the contralateral frontals and parietals meet medially, occluding the parietal fontanel (state 1; Fig. 23). Among the members of the Characidae, Lucena (1998) mentioned the absence of this fontanel only in Brycon pesu, while Buckup (1998) coded it as present in B. guatemalensis Regan. According to Zanata & Vari (2005) this fontanel is present in specimens of B. pesu of 27.0 mm SL, and absent in specimens of 49.7 mm SL, indicating that at least in some species this character is variable during the growth. The parietal fontanel is present in specimens of 256 mm SL of B. meeki (Weitzman, 1962), and it is coded as present for this species. In the examined specimens of B. falcatus and B. orbignyanus this fontanel is present; as there is no published evidence indicating ontogenetic elimination of the parietal fontanel in these species, this character is also coded with state 0. Prootic: 42. Trigemino-facialis foramen: (0) broad, largely limited by sphenotic dorsally; (1) narrow, as a cleft with sphenotic almost excluded from its margin. In most examined species the trigemino-facialis foramen is situated in an approximately triangular chamber on the anterior surface of the prootic, with the chamber limited dorsally by the sphenotic (state 0; Figs. 11 and 31). In the examined species of Aphyocharax the trigemino-facialis foramen is reduced to a cleft, narrow in anterolateral view and almost completely limited by the prootic and pterosphenoid (state 1; Fig. 30); in this state the dorsal margin of the trigemino-facialis foramen is much reduced and the sphenotic, which limits dorsally the foramen, is almost excluded from its margin. Pterosphenoid: 43. Large foramen on pterosphenoid: (0) absent; (1) present, well developed. In most examined species the lateral surface of the pterosphenoid is flat or has a shallow ridge ventrally limiting Fig. 30. Posterior region of neurocranium of Aphyocharax anisitsi, CI-FML 2951, 29.6 mm SL, lateral view, anterior to left. OSP: orbitosphenoid, PRO: prootic, PTS: pterosphenoid, SPH: sphenotic. Scale bar = 0.5 mm. J. M. Mirande 403 the supraorbital nerve, which is directed to the trigeminofacialis foramen (state 0; Figs. 11, 30, and 31). In the examined specimens of Aphyocharacidium bolivianum and Axelrodia lindeae there is a large foramen situated in the middle of the pterosphenoid (state 1; Fig. 32). The association of this foramen with blood vessels or nerves was not confirmed. The presence of this foramen is variable among the examined specimens of Microschemobrycon casiquiare and this species is coded as polymorphic. 44. Small foramen near posterior margin of pterosphenoid: (0) absent, or not pierced by nerves; (1) present, pierced by a branch of supraorbital nerve. The supraorbital nerve runs through the orbit in the region of articulation between the frontal and orbitosphenoid/pterosphenoid, entering to the braincase through the trigemino-facialis foramen in most examined species (state 0). In some species, mostly of the Stevardiinae, a branch of this nerve enters to the braincase through a small foramen near the posteroventral margin of the pterosphenoid (state 1; Fig. 31). The examined alcoholpreserved specimens of Astyanax cf. eigenmanniorum (Cope), Cynopotamus argenteus, Hemibrycon surinamensis Géry, Hyphessobrycon bifasciatus Ellis, Moenkhausia sanctaefilomenae (Steindachner), Oligosarcus spp., and Roeboides microlepis have small pores in a similar position, but these pores are not pierced by nerves or blood vessels; therefore these species are coded with state 0. Pterotic: 45. Dorsal process of pterotic where tendon from epaxial musculature attach: (0) absent; (1) present, projecting dorsally from tube for semicircular canal. The examined species usually have a tendon from the epaxial musculature attached to the lateral margin of the pterotic tube for the horizontal semicircular canal, or to a process ventral to this tube (state 0). In a group of species, there is a small process directed dorsally from the tube for the semicircular canal, onto which this tendon attaches (state 1; Fig. 33). The presence of this process is variable among the examined specimens of Mimagoniates rheocharis, Nematobrycon palmeri, and Phenagoniates macrolepis which are coded as polymorphic. 46. Relative length of pterotic spine: (0) projected more posteriorly than attachment site of ligament from hyomandibula; (1) restricted to attachment region of hyomandibular ligament. In all the examined members of the outgroup and some characids, the pterotic spine is projected posterior to the attachment site of a ligament from the hyomandibula (state 0; Figs. 5 and 7). In most examined members of the Characidae the pterotic spine is relatively reduced and its extension is limited to the posterior extension of the attachment site for the hyomandibular ligament (state 1). Rhinosphenoid: 47. Rhinosphenoid: (0) absent; (1) present. (UJ35, BU7, LU8i, CM11(13), TP16i, VB15, SE15, BE20i, ZV31i, LI28, BÜ8, QU21i, PZ13). The rhinosphenoid is a medial bone situated anterior to the orbitosphenoid, which is present in most characiforms (state 1; Figs. 6, 24, and 34) albeit is absent in many groups in the Characiformes and the remaining Ostariophysi (state 0; Fig. 14). Bührnheim (2006) mentioned the presence of rhinosphenoid, among others, in Charax stenopterus; this bone is, however, absent in the specimens of this species herein examined, and this species is coded as polymorphic. The presence of this bone is intraspecifically variable among the examined specimens Fig. 31. Posterior region of neurocranium of Bryconamericus alpha, ANSP 130512, 42.8 mm SL, anterolateral view, anterior to left. PRO: prootic, PTS: pterosphenoid, SPH: sphenotic. Scale bar = 1 mm. Fig. 32. Posterior region of neurocranium of Axelrodia lindeae, MCP 37314, 18.7 mm SL, lateral view, anterior to left. PRO: prootic, PTS: pterosphenoid, SPH: sphenotic, TSC: tube for semicircular canal. Scale bar = 1 mm. Phylogeny of the family Characidae 404 of Astyanax correntinus (Holmberg), A. latens Mirande, Aguilera & Azpelicueta, Diapoma terofali (Géry), and Hyphessobrycon bifasciatus, in which this character is coded as polymorphic. 48. Dorsal expansion of rhinosphenoid: (0) absent; (1) present and forming a bony wall between olfactory nerves. (SE17, BE21). The form of the rhinosphenoid is variable among the examined species. In some species this bone has a rectangular form and is situated completely ventral to the olfactory nerves (state 0; Figs. 6 and 34). In Characidium spp., Hemiodus cf. thayeria, and many species of the Characidae, the rhinosphenoid is expanded dorsally and so forming a bony wall between the olfactory nerves (state 1; Fig. 24). Benine (2004) coded, among others, Bario steindachneri (Eigenmann) and Tetragonopterus argenteus Cuvier as lacking a dorsally expanded rhinosphenoid, however a well developed process was observed in B. steindachneri and a relatively smaller process was found in T. argenteus and these two species are coded as polymorphic. A relatively reduced or intraspecifically variable dorsal projection of the rhinosphenoid was also observed in Aphyodite grammica Eigenmann, Hemigrammus unilineatus, Hyphessobrycon elachys, H. pulchripinnis Ahl, Microschemobrycon casiquiare, and Moenkhausia sanctaefilomenae which are coded as polymorphic. This character is coded as inapplicable to species lacking an ossified rhinosphenoid. 49. Posterior extension of rhinosphenoid cartilage: (0) projected only to middle horizontal length of orbitosphenoid, or less; (1) extended to vertical through region of articulation between orbitosphenoid and pterosphenoid. The rhinosphenoid cartilage is situated between the rhinosphenoid, lateral ethmoid, and parasphenoid. In almost all of the examined species, this cartilage extends posteriorly along the dorsal margin of the parasphenoid to a point not surpassing the vertical through middle horizontal length of the orbitosphenoid (state 0; Fig. 35). In Inpaichthys kerri and Xenagoniates bondi, this cartilage extends more posteriorly and reaches the vertical through the region of articulation between the orbitosphenoid and pterosphenoid (state 1). The examined specimens of Hollandichthys multifasciatus have an intermediate situation that is coded as polymorphic. 50. Ventral border of rhinosphenoid: (0) distinctly separate from parasphenoid; (1) almost contacting parasphenoid. In most examined species the rhinosphenoid is variably, but distinctly separated from the parasphenoid (state 0; Fig. 25). In a small group of species, the rhinosphenoid is situated near the parasphenoid, and its ventral margin is parallel to and almost contacts the dorsal margin of the parasphenoid (state 1). This character is coded as inapplicable in species without an ossified rhinosphenoid. Supraoccipital: 51. Anterior margin of supraoccipital: (0) situated completely behind vertical through posterior orbital margin; (1) situated anterior to vertical through posterior orbital margin. The anterior margin of the supraoccipital that forms the the posterior margin of the parietal fontanel, when that opening is present, is located posterior to the vertical through the posterior margin of the orbit in most examined species (state 0). In some species, in contrast, the anterior margin of the supraoccipital is situated anterior to the vertical through the posterior margin of the orbit (state 1; Fig. 24). The examined Fig. 33. Cranium of Phenacogaster tegatus, CI-FML 3880, 35.3 mm SL, lateral view, anterior to left. IO2-3: second and third infraorbitals, MAX: maxilla, POP: Preopercle, PTO: pterotic. Scale bar = 1 mm. Fig. 34. Posterior region of cranium of Oligosarcus bolivianus, CI-FML 3366, 83.4 mm SL, lateral view, anterior to left. IO4-6: fourth to sixth infraorbitals, FRO: frontal, OPE: opercle, POP: preopercle, PTO: pterotic, RSP: rhinosphenoid, SPP: suprapreopercle. Scale bar = 1 mm. J. M. Mirande 405 specimens of Hyphessobrycon socolofi Weitzman have an intermediate situation that is coded as polymorphic. 52. Length of supraoccipital spine: (0) extends dorsal of entire neural complex of Weberian apparatus; (1) extends dorsal of approximately one half extent of neural complex. (AM16m, CM8(9), MO34m, VB24m, SE18m, BE19m, QU25m). 53. Length of supraoccipital spine: (0) extends posteriorly to, at least, middle length of neural complex of Weberian apparatus; (1) extends only to anterior limit of neural complex. (LU23m, MO34m, VB24m, SE18m, BE19m, QU25m). When it is present, the supraoccipital spine forms the posterior margin of the parietal fontanel and extends dorsal to the modified vertebrae of the Weberian apparatus to different degrees among the examined species. The length of the supraoccipital spine is herein compared to that of the neural complex of the Weberian apparatus. In some species the supraoccipital spine is greatly developed and extends dorsal to the entire neural complex (character 52, state 0; Fig. 24). In an intermediate state, it extends onto the middle of the length of the neural complex (character 52, state 1; character 53, state 0; Figs. 5 and 8), whereas in some species the supraoccipital spine covers only the anterior vertical portion of the neural complex (character 53, state 1; Fig. 35). Intermediate states of the character 52 were observed in Astyanax abramis (Jenyns), A. cf. abramis, A. correntinus, Brycon orbignyanus, Markiana nigripinnis, Piaractus mesopotamicus, Pseudochalceus kyburzi, and Rhoadsia altipinna, which are coded as polymorphic. The examined specimens of Hyphessobrycon socolofi and Roeboexodon geryi have intermediate states of the character 53, and this character is also coded as polymorphic in these species. Vomer: 54. Dorsolateral processes of vomer: (0) absent; (1) present. (PZ7). The vomer, in most species, has a medial longitudinal ridge in the dorsal surface aligned with a ridge on the ventral surface of the mesethmoid, together with there forming the nasal septum (state 0). In addition to this medial lamella, some species have paired dorsal lamellae on the vomer aligned with the ventral diverging lamellae of the mesethmoid (state 1; ToledoPiza, 2007: fig. 5). These ridges were called dorsolateral processes of vomer by Toledo-Piza (2007). Orbital region Antorbital: 55. Antorbital: (0) present; (1) absent or fused with first infraorbital. (VA1, BU20, LU25, ZV1). The antorbital is present as an independent ossification in almost all the Characiformes (state 0; Fig. 37), whereas it is presumably fused with the first infraorbital, forming the lachrymal, in the Cypriniformes and Siluriformes (Weitzman, 1962). The absence of antorbital (state 1) is not usual in Characiformes; it was mentioned for several genera (Lucena, 1993; Buckup, 1998) and proposed by Vari (1995) as a synapomorphy of a clade composed of the Ctenoluciidae and Erythrinidae. Hoplias cf. malabaricus (Bloch), Inpaichthys kerri, and the root of this analysis are coded with state 1, although based on its position and extension this bone appears to be fused in Hoplias cf. malabaricus and cyprinids and absent in I. kerri. 56. Position of antorbital relative to lateral ethmoid in lateral view: (0) antorbital entirely anterior to lateral ethmoid; (1) antorbital overlapping lateral ethmoid. (TP2). In most examined species the antorbital is situated lateral to the olfactory capsules and bordering the nasal openings posteriorly but is positioned entirely anterior to the lateral ethmoid (state 0; Figs. 35 and 36). In some species the antorbital is relatively broader, extensively overlapping the lateral margin of the lateral ethmoid (state 1; Fig. 37). This broadening of the antorbital was mentioned by Castro (1984) as a possible synapomorphy of a proposed clade formed by Clupeacharax Pearson and Engraulisoma Castro. Infraorbitals: 57. Relative position of anterior margin of antorbital and first infraorbital: (0) anterior margin of antorbital either aligned with or anterior to first infraorbital; (1) anterior margin of antorbital posterior to first infraorbital. (ZV6i). The first infraorbital is located just ventral to the antorbital and extends to a varying degree along the longitudinal length of this bone. In most species the anterior margin of the first infraorbital reaches a point between the middle and anterior margin of the antorbital, but not projects anterior to this bone (state 0; Fig. 37). In several outgroups and Galeocharax humeralis the first infraorbital conspicuously projects anterior to the antorbital (state 1). Fig. 35. Cranium of Microschemobrycon casiquiare, ANSP 159704, 26.0 mm SL, lateral view, anterior to left. MAX: maxilla, MEC: Meckelian cartilage, NCO: neural complex of Weberian apparatus, OSP: orbitosphenoid, PTS: pterosphenoid, RHC: rhinosphenoid cartilage, SOC: supraoccipital. Scale bar = 1 mm. Phylogeny of the family Characidae 406 58. Bony lamellae bordering laterosensory canal of first infraorbital: (0) present; (1) absent. (ZV9m). In most examined species the first infraorbital laterosensory canal is margined by a bony lamella both dorsally and ventrally, (state 0; Figs. 38 and 39). In some species the first infraorbital is reduced to the laterosensory canal, and these lamellae are lacking (state 1; Fig. 40). The lamellae of all the infraorbitals are much reduced in Gymnocharacinus bergii differing from other species lacking bony lamellae on the first infraorbital, in which the lamellae of the remaining infraorbitals are well developed. However, G. bergii is coded as state 1. The first infraorbital is absent in Carnegiella strigata and Coptobrycon bilineatus, and this character is coded as inapplicable to these species. 59. Extent of expansion of first infraorbital lateral to maxilla: (0) covering less than one half length of maxilla; (1) covering most of maxilla. (PZ30m). The ventral margin of the first infraorbital forms a groove that receives the dorsal margin of the maxilla in most examined species. This grooves usually laterally covers as much as one half the length of the maxillary lamella when the mouth is closed (state 0). In Acestrorhynchus pantaneiro the first infraorbital entirely covers the maxilla in lateral view (state 1; e. g. Menezes, 1969: fig. 23), and the maxillary lamella is also significantly covered by the first infraorbital in Agoniates anchovia, Bryconexodon juruenae, Engraulisoma taeniatum, Exodon paradoxus, Heterocharax macrolepis, and Roeboexodon geryi, all of which are coded as state 1. This character is coded as inapplicable to Carnegiella strigata and Coptobrycon bilineatus, in which the infraorbitals are much reduced. 60. Lateral overlap of first infraorbital by anterior margin of second infraorbital: (0) absent; (1) present. In most examined species the posterior margin of the first infraorbital and the anterior margin of the second infraorbital either are not contacting each other or the first infraorbital is overlapping laterally the second infraorbital (state 0; Figs. 40 and 41). The examined species of Bryconops Kner and Triportheus instead have the anterior margin of the second infraorbital extensively overlapping the posterior margin of the first infraorbital, especially in the region of the laterosensory canal, and covering it from lateral view (state 1; Fig. 39). This character is coded as inapplicable to Carnegiella strigata and Coptobrycon bilineatus in which the infraorbitals are reduced. This character differs with that described by Lucena & Menezes (1998: fig. 4) in which only the laterosensory canal of the second infraorbital overlaps laterally the lamella of the first one, whereas its lamella is situated clearly medial to that of the first infraorbital. In the state 1 of this character both the laterosensory canal and the lamella of the second infraorbital overlap laterally the lamella of the first infraorbital. 61. Overlap of maxilla by second infraorbital: (0) absent; (1) present. (LM5). In most species of the ingroup the first infraorbital overlaps the maxilla to some degree, while the ventral margin of second infraorbital is situated dorsally and does not cover the maxilla laterally (state 0). In some examined species, both the first and second infraorbitals form part of a sheath that receives the dorsal margin of the maxilla when the mouth is closed (state 1; Zanata & Vari, 2005: fig. 3). In Apareiodon affinis (Steindachner), Characidium spp., Distichodus maculatus, Hemiodus cf. thayeria, Iguanodectes geisleri Géry, Leporinus striatus, Metynnis maculatus, Parodon nasus (Kner), Piabucus melanostomus, and Puntius tetrazona, the posterior tip of the maxilla falls short of the second infraorbital and this character is coded as inapplicable. Fig. 36. Anterior region of cranium of Rhoadsia altipinna, MHNG 2173.31, 73.5 mm SL, lateral view, anterior to left. AOR: antorbital, LET: lateral ethmoid. Scale bar = 1 mm. Fig. 37. Anterior region of cranium of Agoniates anchovia, MCP 16969, 127.7 mm SL, lateral view, anterior to left. AOR: antorbital, IO1: first infraorbital, LET: lateral ethmoid. Scale bar = 5 mm. J. M. Mirande 407 62. Articulation between second and third infraorbitals: (0) vertical; (1) anteroventrally angled; (2) posteroventrally angled. The articulation between the second and third infraorbitals lacks interdigitations among the examined species. The main variation in this articulation is the relative angle of contact between these bones. In the most common situation the posterior margin of the second infraorbital and the anterior margin of the third infraorbital describe an approximately vertical line perpendicular to the horizontal arm of the preopercle (state 0; Fig. 41). In a relatively diverse group of species the second infraorbital has an approximately triangular shape, with an anteroventrally oblique posterior margin. In this state the third infraorbital partially borders the ventral edge of the second infraorbital (state 1; Fig. 40). The inverse situation is found in some species in which the posterior margin of the second infraorbital is posteroventrally oblique and the second infraorbital ventrally borders the anterior region of the third infraorbital (state 2; Fig. 39). The latter state is usually found in species with a long maxilla, although some species with a long maxilla have state 0 or even state 1, such as Paragoniates alburnus. 63. Anterior region of third infraorbital: (0) not much expanded relative to posterior region of second infraorbital; (1) abruptly expanded relative to posterior region of second infraorbital. In most examined species the ventral margins of the second and third infraorbitals form a nearly continuous line, without conspicuous expansions of these bones (state 0; Fig. 41). In Fig. 38. Infraorbitals of Chalceus macrolepidotus, MHNG 2189.13, 64.0 mm SL, anterior to left. IO1: first infraorbital, IO4: fourth infraorbital, IO6: sixth infraorbital. Scale bar = 1 mm. Fig. 39. First to fifth infraorbitals of Bryconops melanurus, MCP 15807, 83.4 mm SL, anterior to left. IO1-4: first to fourth infraorbitals. Scale bar = 1 mm. Fig. 40. Infraorbital series of Paragoniates alburnus, MHNG 2370.12, 65.1 mm SL, anterior to left. IO1-4: first to fourth infraorbitals. Scale bar = 1 mm. Fig. 41. Infraorbital series of Astyanax cf. asuncionensis, CIFML 3912, 61.1 mm SL, anterior to left. IO1-6: first to sixth infraorbitals. Scale bar = 1 mm. Phylogeny of the family Characidae 408 Hyphessobrycon eques (Boulenger), H. socolofi, and Pristella maxillaris, the anterior margin of the third infraorbital is much expanded ventrally relative to the posterior margin of the second infraorbital. Thus, the line formed by the ventral margins of these bones is not straight but rather has a sinusoidal shape in the region of articulation between these elements (state 1; Fig. 9). 64. Ventral extent of third infraorbital: (0) reaching horizontal arm of preopercle; (1) not reaching horizontal arm of preopercle, at least anteriorly. (EI5mi, AM23, UJ51, UJ53, BÜ19m). The ventral extension of the third infraorbital was used in traditional characid systematics (Eigenmann, 1917) to discriminate some genera (e. g. Astyanax from Bryconamericus). In some species the third infraorbital is comparatively more developed, covering completely the cheek and reaching the horizontal arm of the preopercle (state 0; Fig. 29). In other species instead it is less developed and there is a “naked” area between the anterior region of this infraorbital and the preopercle (state 1; Fig. 33). 65. Posterior extent of third infraorbital: (0) covering angle of preopercle; (1) relatively reduced, angle of preopercle covered partially by fourth infraorbital. (MO44). The posteroventral angle of the preopercle is bordered dorsally by the third infraorbital in most examined species (state 0; Weitzman, 1962: fig. 8, Zanata & Vari, 2005: fig. 3). In Hemiodus cf. thayeria, Micralestes stormsi, and Piabucus melanostomus, the third infraorbital is relatively less developed, articulating with the fourth infraorbital just dorsal to the angle of the preopercle (state 1; Roberts, 1974: fig. 6). Moreira (2002) mentioned the presence of state 1 as a synapomorphy for a clade including all the species of the Iguanodectinae except Iguanodectes geisleri and cited a parallel occurrence of this state in Micralestes acutidens (Peters). This character is coded as inapplicable in species with reduced infraorbital lamellae. 66. Fourth infraorbital: (0) present, well developed; (1) absent or much reduced and bordered posteriorly by third and fifth infraorbitals. (VA12, LU28m, LC7m, VB14, ZV12, QU16m, PZ34). Most Characiformes have six infraorbitals (state 0; Figs. 40 and 41), with several reductions in different groups, being the fourth infraorbital most frequently the absent or reduced infraorbital (state 1; Fig. 29). Uj (1990) proposed the reduction or absence of such infraorbital as diagnostic for his Aphyocharacidae (including the genera Aphyocharax and Prionobrama), and Vari (1995) mentioned it as a synapomorphy of the Ctenoluciidae. The presence of the fourth infraorbital is variable among the examined specimens of Charax stenopterus, Mimagoniates rheocharis, Hyphessobrycon eques, and Nantis indefessus (Mirande, Aguilera & Azpelicueta), which are coded as polymorphic. The absence and the extreme reduction of the fourth infraorbital are variable within some species of Aphyocharax, and both conditions are very different to the presence of six well developed infraorbitals that is usually present in the Characidae. Therefore it is preferred to include both conditions in the same state, instead of coding this character for presence or absence of the fourth infraorbital. This character is coded as inapplicable to the species in which all the infraorbitals are reduced. 67. Form of fourth infraorbital: (0) approximately square or more developed longitudinally than dorsoventrally; (1) longer dorsoventrally than longitudinally. (SE25m, BE2, BÜ20). The form of the fourth infraorbital is variable among the examined species. The fourth infraorbital in some species is more developed longitudinally than dorsoventrally or it is approximately square (state 0; Fig. 40), while in others it is more elongate dorsoventrally (state 1; Figs. 39 and 41). Contrary to the observations of Benine (2004), the examined specimens of Bario steindachneri and Bryconops melanurus (Bloch) have state 0 and these species are coded as polymorphic given possible intraspecific variability. The examined specimen of Knodus breviceps shows an intermediate state and this species is coded as polymorphic. The examined specimens of Moenkhausia sanctaefilomenae are variable in this character, and this species is also coded as polymorphic. 68. Posterior dorsoventral expansion of fourth infraorbital: (0) absent; (1) present. (PZ33). In most examined species, the fourth infraorbital is approximately square or rectangular and bordered dorsally by the fifth infraorbital, whose posterior margin is situated just anterior to the exposed portion of the vertical arm of the preopercle (state 0). In Acestrocephalus sardina, Acestrorhynchus pantaneiro, Cynopotamus argenteus, Galeocharax humeralis, Oligosarcus spp., and Rhaphiodon vulpinus among the examined species, the posterior region of the fourth infraorbital is dorsoventrally expanded and the fifth infraorbital is displaced dorsally and relatively distant from the vertical arm of the preopercle (state 1; Figs. 10 and 34). 69. Lateral coverage of dilator fossa by sixth infraorbital: (0) almost complete, at least in its ventral border; (1) leaving a conspicuous naked area in anterior region of dilator fossa. (UJ49m, SE28m, PZ36m). In most outgroups the sixth infraorbital laterally covers the fossa for the dilator operculi muscle (state 0). A group of species with well developed infraorbitals instead have the sixth infraorbital leaving a “naked” area in the anterior region of the dilator fossa (state 1). Uj (1990) mentioned the reduced size of the dermosphenotic (=sixth infraorbital) in a clade including his Aphyocharacidae (genera Aphyocharax and Prionobrama), Piabucidae (=Iguanodectinae), Paragoniatidae (=Paragoniatinae), and Tetragonopteridae (=Tetragonopterinae). This character is coded as inapplicable to the species in which the dilator fossa is absent or there is a reduction of the infraorbitals. J. M. Mirande 409 Supraorbital: 70. Supraorbital: (0) present; (1) absent. (UJ48, VA9, BU21, LU26, MO39, VB13, SE21, BE5, ZV2, QU15, PZ37). The supraorbital, when present, is situated dorsal or anterodorsal to the orbit (state 0: Fig. 13; Zanata & Vari: figs. 1, 2, and 4). The absence of a supraorbital (state 1) was considered by Uj (1990) as a synapomorphy for the clade composed of his Aphyocharacidae (Aphyocharax and Prionobrama), Piabucidae (=Iguanodectinae), Paragoniatidae (=Paragoniatinae), and Tetragonopteridae (=Tetragonopterinae); however this character has an ambiguous optimization for that node according to the tree proposed by Uj. In addition, Vari (1995) mentioned the absence of the supraorbital as a parallelism for the Erythrinidae and Lebiasinidae. In the phylogenetic hypothesis of Buckup (1998), the absence of the supraorbital is a synapomorphy for a clade composed of Charax, Cynopotamus Valenciennes, Oligosarcus, Phenacogaster Eigenmann, and Tetragonopterus. According to Lucena (1993), the absence of this bone is a synapomorphy independently supporting a clade composed of Gnathocharax Fowler, Heterocharax Eigenmann, and Hoplocharax Géry, a clade including most members of the Characidae, and a node composed of the Erythrinidae and Lebiasinidae. Moreira (2002) mentioned the presence of the supraorbital in all known species of the Iguanodectinae, differing in that observation from Lucena (1993) and Malabarba & Weitzman (2003). Malabarba & Weitzman (2003) proposed a clade of characids lacking the supraorbital composed of the subfamilies Aphyocharacinae, Characinae, Cheirodontinae, Iguanodectinae, Rhoadsiinae, Stethaprioninae, Tetragonopterinae, some incertae sedis genera, and their clade A. 71. Contact between supraorbital and sixth infraorbital: (0) absent; (1) present. (AM19, UJ40, CM6(7)i, LU27i, MO40m, ZV5, LI4). When present, the supraorbital is variably developed across the Characiformes. In some species it is longitudinally reduced and restricted to the anterior region of the dorsal margin of the orbit, not contacting the sixth infraorbital. In this state the orbit is margined dorsally by the frontal (state 0; Zanata & Vari, 2005: fig. 1A). In some species the posterior region of the supraorbital is relatively more developed, reaching to the anterior margin of the sixth infraorbital. In these cases the dorsal margin of the orbit is bordered by these two bones, which exclude the frontal from this margin (state 1; Fig. 13, Zanata & Vari, 2005: figs. 1B, 2, and 4). State 0 is coded only in the cases in which the supraorbital is present but not contacting the sixth infraorbital. This character is coded as inapplicable for species in which the supraorbital is absent. Laterosensory system Infraorbital canal: 72. Laterosensory canal in antorbital: (0) absent; (1) present. (UJ41, CM4(5), MO38, SE22m, BE1m). In the examined species the infraorbitals usually bear laterosensory canal segments. Such canals are absent in the antorbital (state 0). The presence of a laterosensory canal in the antorbital (state 1) was proposed as a synapomorphy of Bryconops by Lucena (1993), and cited also for some species of Brycon Müller & Troschel and Iguanodectes Cope (Malabarba, 1998b; Moreira, 2002). This character is coded as inapplicable to the species lacking the antorbital. 73. Laterosensory canal of first infraorbital: (0) projects dorsally from main body of first infraorbital; (1) absent or does not projects dorsally. (LU30, CM2(3)i). In most examined species the first infraorbital bears a laterosensory canal that extends dorsally to a point near the dorsal margin of the lamellar region of the bone (state 1). In some species this canal projects dorsal to the lamella of the first infraorbital (state 0; Fig. 13). This character is herein considered as inapplicable to Hoplias cf. malabaricus, in which the first infraorbital appears to be fused with the antorbital. Malabarba (1998b) coded this character with state 0, among others, in Brycon pesu and Salminus sp. In the examined specimens of B. pesu and S. brasiliensis this canal does not project dorsally. This character is coded as polymorphic in B. pesu, while S. brasiliensis is coded with state 0, given that the observations of Malabarba (1998b) are from an unidentified species of the genus, and the only available data about this character for S. brasiliensis is based on the observations made for the present paper. 74. Branching of laterosensory canals of fourth or fifth infraorbitals: (0) absent; (1) present. (BE3i, LI53m, PZ92). The infraorbital laterosensory canal usually has a posterior branch oriented towards the dorsal end of the laterosensory canal in the preopercle or a pore in the vertical arm of the preopercle. In most examined species this canal is not ossified, but rather formed by soft tissue and it is situated superficially on the infraorbitals (state 0). In some species this canal is partially contained in the fourth or fifth infraorbitals and is evident as an ossified longitudinal branch of the laterosensory canal (state 1; Figs. 38, 39, and 41). This character is coded as polymorphic in Acestrorhynchus pantaneiro, Acestrocephalus sardina, Astyanax chico Casciotta & Almirón, Cyphocharax spilotus Vari, Deuterodon iguape, Hyphessobrycon pulchripinnis, Moenkhausia xinguensis, and Probolodus heterostomus, in which this branch is variably present. This character is coded as inapplicable for species in which the fourth and/or fifth infraorbitals are reduced or absent. 75. Direction of posterior branch of laterosensory canal of fourth or fifth infraorbital: (0) to a pore on preopercle near hyomandibular condyle; (1) to a pore conspicuously ventral to hyomandibular condyle. In most examined species the posterior branch of the laterosensory canal of the fourth or fifth infraorbitals is directed towards the dorsal end of the preopercle near the region of articulation between the hyomandibula and opercle (state 0). In Chalceus macrolepidotus and Hemiodus cf. thayeria this branch is oriented to a point approximating half way down the vertical arm of the preopercle (state 1). Phylogeny of the family Characidae 416 Figs. 35 and 56). This character is coded as polymorphic in Distichodus maculatus, which has state 1, whereas Xenocharax spilurus, a basal distichodontid, has state 0. This character is also coded as polymorphic in Carlana eigenmanni and Rhoadsia altipinna, which have state 1 when young and state 0 when adults. 101. Ontogenetic lengthening of maxilla: (0) absent; (1) present. In most examined species growths of the maxilla during ontogeny is proportional with that of the remaining bones of the head, and its length is approximately constant in relation to the other bones (state 0). In members of the subfamily Rhoadsiinae, the proportional length of the maxilla increases during the ontogeny such that it becomes proportionally longer in larger individuals (state 1; Eigenmann, 1927: figs. 10, 13, and 14). 102. Dorsal projection of maxilla: (0) not overlaps second infraorbital; (1) overlaps second infraorbital. The dorsal margin of the maxilla usually fits to a sheath formed only by the first infraorbital and, in some species, first and second infraorbitals when the mouth is closed (state 0). The inverse situation is present in the examined species of Aphyocharax, in which the maxilla laterally overlaps the second infraorbital when the mouth is closed (state 1; Fig. 29). Premaxilla: 103. Interdigitations between premaxillae: (0) present; (1) absent. (LU37, ZV55i). The mesethmoid spine usually separates almost completely the premaxillae. The premaxillae, in turn, are usually joined to each other anterior of the mesethmoid solely by connective tissue and lack bony interdigitations (state 1). Weitzman (1962) described the presence of bony interdigitations between the premaxillae of Brycon meeki (state 0; Weitzman, 1962: fig. 2). Vari (1979) later proposed the presence of bony interdigitations between premaxillae as a synapomorphy of the Citharinoidei. 104. Length of ascending process of premaxilla: (0) reaching at least one-third of length of nasal; (1) reaching just anterior end of nasal. (LU34mi). The premaxillary ascending process usually reaches one third or more the length of the nasal, as described by Weitzman (1962) for Brycon meeki (state 0; Fig. 21). This condition is broadly distributed among the members of the Characidae. In most outgroups and a diverse group within the Characidae, the ascending process of the premaxilla is relatively shorter, reaching only the anterior end of the nasal (state 1; Figs. 18 and 26). Intermediate states were observed in Astyanax latens, Bryconamericus mennii, Diapoma terofali, Hasemania nana, and Pseudocorynopoma doriae, which are coded as polymorphic. 105. Alignment of ascending process of premaxilla: (0) aligned with medial margin of nasal; (1) medially shifted and separated from nasal. The anterior process of the mesethmoid usually has a triangular shape with its posterior region as wide as the dorsal lamella of the mesethmoid, which is in turn situated posterior to the lateral wings. Since the anterior process medially borders the premaxilla and the dorsal lamella borders the nasal, the ascending process of premaxilla is usually aligned with the nasal, or contacts to it medially (state 0; Fig. 21). In Aphyocharax spp., Carnegiella strigata, Prionobrama paraguayensis, and Thoracocharax stellatus, the anterior process of the mesethmoid is much slender relative to the dorsal lamella of the mesethmoid. In these cases the nasal is consequently distinctly displaced laterally relative to the ascending process of the premaxilla (state 1; Fig. 18). 106. Form of posterolateral portion of premaxilla: (0) with notch; (1) with pedicle expanded laterally to maxilla. (LU36i, ZV52). The posterolateral tip of the premaxillary alveolar arm in the members of the Characidae has a concave surface that receives the ventral margin of the ascending process of the maxilla (state 0; Zanata & Vari, 2005: fig. 14). Zanata & Vari (2005) described a projection of the posterior region of the premaxillary alveolar arm that borders the lateral surface of the proximal region of the maxilla (state 1; Zanata & Vari, 2005: fig. 13). These authors cited the presence of this pedicle-like process in most members of the Alestidae and in the serrasalmids Colossoma macropomum (Cuvier) and Piaractus mesopotamicus. In the examined specimen of Piaractus mesopotamicus the posterolateral end of the premaxilla has a concave surface where the maxilla articulates; the external lobe of this structure is relatively more developed but it does not laterally borders the maxilla and P. mesopotamicus is consequently coded as state 0 for this character. This character is not directly comparable in Puntius tetrazona and the root is coded as inapplicable. Fig. 56. Oral region of cranium of Odontostilbe microcephala, CI-FML 3886, 50.6 mm SL, lateral view, anterior to left. DEN: dentary, IO1-2: first and second infraorbitals, MAX: maxilla, MEC: Meckelian cartilage. Scale bar = 1 mm. J. M. Mirande 417 Lower jaw Anguloarticular: 107. Lateral ridge of anguloarticular: (0) absent; (1) present. (BÜ57m). The anguloarticular has two anterior processes, a horizontal process and an oblique anterodorsal process. The horizontal process is situated along the longitudinal axis of the dentary and forms the posterior attachment site for the Meckelian cartilage; it is attached laterally to that cartilage and the coronomeckelian bone. The oblique process projects towards the posterodorsal margin of the dentary and usually is smooth or has longitudinal striae on its lateral surface (state 0; Fig. 57). In addition to these striae, some species have a bony ridge parallel to the anteroventral margin of the oblique process between its border and the ventral margin of the primordial ligament (state 1; Fig. 42). 108. Horizontal process of anguloarticular: (0) laterally covered by dentary only anteriorly; (1) broadly covered by dentary which reaches posterior border of Meckelian cartilage. The anguloarticular processes articulate with the dentary such that the oblique dorsal process overlaps the dentary laterally and the horizontal process overlaps it medially. Both processes are visible laterally in most species with the oblique process completely visible, but the horizontal process only partially visible due to the overlap of the dentary which covers it laterally to different degrees. In most species the dentary laterally covers a relatively small portion of the horizontal process, which does not reach the vertical through posterior tip of the Meckelian cartilage (state 0; Figs. 42 and 58). In some species the dentary overlaps a longer portion of the horizontal process of the anguloarticular and reaches the posterior end of the Meckelian cartilage. In this state the Meckelian cartilage appears to be completely contained within the dentary from a lateral view, although it is clear in medial view that its posterior portion is in contact with the anguloarticular (state 1; Figs. 59 and 60). 109. Ventral margin of horizontal process of anguloarticular: (0) posteroventrally angled relative to laterosensory canal of dentary from medial view; (1) perpendicular to laterosensory canal of dentary from medial view. The horizontal process of the anguloarticular articulates medially within the dentary in the Meckelian fossa, which in turns serves as the attachment area for the Aw section of the adductor mandibulae (Winterbottom, 1974; Howes, 1976). Fibers from the Aw section usually inserts posteriorly on a tendon from the A2 and A3 sections of the adductor mandibulae. The anterior portion of this tendon usually divides in two smaller tendons that attach to the retroarticular and the medial face of the dentary, respectively (Zanata & Vari, 2005). The tendon to the dentary inserts anteriorly or anteroventrally to the anteroventral margin of the anguloarticular at a point just ventral to the laterosensory canal of the dentary. In most examined species the anteroventral margin of the anguloarticular is angled with respect to the laterosensory canal of the dentary. In these cases the tendon from the adductor mandibulae attaches anteroventral to the anguloarticular (state 0; Fig. 58 and 60). In some species the anteroventral margin of the anguloarticular has a sinusoidal shape, bordering posteriorly the site of attachment of this tendon and crossing perpendicularly the laterosensory canal of the dentary (state 1; Figs. 59 and 61). This character is variable in Astyanax latens and A. cf. rutilus (Jenyns) which are coded as polymorphic. Coronomeckelian: 110. Position of coronomeckelian: (0) situated mainly lateral to Meckelian cartilage; (1) situated mainly dorsal to Meckelian cartilage. (SE44). The coronomeckelian serves as the attachment site of one of the anterior tendons of the adductor mandibulae muscle. The size and longitudinal position of the coronomeckelian are related with the degree of development of this tendon and the length of the lower jaw. The position of the coronomeckelian along the vertical is variable among the examined species. In some species the coronomeckelian bone is situated almost completely dorsal to the Meckelian cartilage (state 1; Fig. 61), while in others the bone is located mainly lateral to that cartilage (state 0; Figs. 57 and 58). Several intermediate cases were observed among the examined species which are coded as polymorphisms. Dentary: 111. Interdigitations between dentaries: (0) absent; (1) present. (UJ18, BU35, LU46m, MA4, ZV82, LI10). In the Cypriniformes, Citharinoidei, and some families of the Characoidei the articulation between the dentaries lacks interdigitations and these bones are joined together only by connective tissues (Vari, 1979) (state 0). This condition was considered as plesiomorphic for the Characiformes by Fink & Fink (1981). According to the phylogenetic hypothesis of Fig. 57. Lower jaw of Roeboides descalvadensis, CI-FML 3859, 61.4 mm SL, lateral view, anterior to left. AAR: anguloarticular, CME: coronomeckelian, DEN: dentary, LCD: laterosensory canal of dentary, MEC: Meckelian cartilage. Scale bar = 1 mm. Phylogeny of the family Characidae 418 Buckup (1998) the presence of bony interdigitations between dentaries (state 1; Figs. 60 and 61) is a synapomorphy for a clade composed of the Acestrorhynchidae, Alestidae, Characidae, Ctenoluciidae, Erythrinidae, Hepsetidae, and Lebiasinidae, with a reversal in Nannostomus Günther. The absence of interdigitations between dentaries was later proposed as a synapomorphy of a cheirodontin clade composed of Amazonspinther Bührnheim, Carvalho, Malabarba & Weitzman and Spintherobolus by Bührnheim et al. (2008). 112. Form of interdigitations between dentaries: (0) simple bony lamellae; (1) undulate lamellae. The number of interdigitations between dentaries varies ontogenetically (Miquelarena, 1986), but some differences have been observed in the form of these lamellae. In most of the examined species the lamellae forming these interdigitations are simple, disposed horizontally, and parallel each other (state 0). Zanata & Vari (2005) described the undulation of margins of these lamellae (state 1; Zanata & Vari, 2005: fig. 22) for some African members of the family Alestidae. This character is coded as inapplicable in species lacking interdigitations between the dentaries. 113. Form and dentition of anterior region of dentary: (0) toothed and not depressed anteriorly; (1) edentulous and much depressed anteriorly. The dentary is usually toothed and not depressed anteriorly in characiforms (state 0), while in the Parodontidae, the dentary is edentulous and much depressed anteriorly (state 1; e. g. Roberts, 1974: fig. 61). Some members of this family, however, have teeth along the lateral margins of the dentary. In Puntius tetrazona and the Cypriniformes in general, the dentary lacks teeth, but that bone is not anteriorly depressed as in the Parodontidae. The members of the Curimatidae lack dentary teeth when adults, but teeth are present in juveniles. Hemiodus cf. thayeria and almost all the genera of the Hemiodontidae lack dentary teeth, but the anterior margin of the dentary is not depressed as it is in the members of the Parodontidae (Roberts, 1974). All of these taxa were coded as state 0. 114. Medial anteroventral notch of dentary: (0) absent; (1) present. (BU37m, LU48m). The presence of a notch along the anteroventral medial border of the dentary (state 1; Fig. 62) was observed in Iguanodectes geisleri, Phenagoniates macrolepis and Xenagoniates bondi, whereas in most examined species the ventral margin of the dentary is straight or slightly curved in this area (state 0). An anteroventral notch in the dentary was illustrated by Weitzman (1964) for Poecilobrycon harrisoni and this character was considered typical of the Pyrrhulinini (=Pyrrhulininae, Lebiasinidae) by him. However, as described by Weitzman (1964) this notch is continuous with the dentary foramen for the mental ramus of the mandibular branch of trigeminus nerve, while the notch herein considered is not related with the dentary foramen. 115. Medial process of dentary bordering Meckelian cartilage dorsally and medially: (0) absent; (1) present. (MO9). According to Moreira (2002, 2003) the presence of a medial process in the dentary constitutes a synapomorphy for the Iguanodectinae (state 1; Fig. 62), with this process absent (state 0; Fig. 60) in the remaining species examined by him. This process was herein observed only in Iguanodectes geisleri and Piabucus melanostomus. In these species such a process medially borders a portion of the Meckelian cartilage and serves as an area of attachment for a tendon from the adductor mandibulae which inserts in the dentary, ventral to the Meckelian cartilage in other species. Fig. 58. Lower jaw of Stichonodon insignis, MHNG 2173.85, 54.5 mm SL, medial view, anterior to left. AAR: anguloarticular, CME: coronomeckelian, DEN: dentary, LCD: laterosensory canal of dentary, MEC: Meckelian cartilage. Scale bar = 1 mm. Fig. 59. Lower jaw of Knodus breviceps, MHNG 2184.97, 43.5 mm SL, lateral view, anterior to left. AAR: anguloarticular, DEN: dentary, LCD: laterosensory canal of dentary, MEC: Meckelian cartilage. Scale bar = 1 mm. J. M. Mirande 419 116. Bony lamella covering dentary foramen laterally: (0) absent; (1) present. The dentary foramen serves as a passage for nerves and blood vessels and is situated just dorsal and anterior to the anterior tip of the Meckelian cartilage. This foramen is usually evident in lateral view (state 0; Figs. 43 and 63). In Acestrocephalus sardina, Cynopotamus argenteus, and Galeocharax humeralis, among the examined species, this foramen is covered laterally by a bony lamella and is not evident from a lateral view (state 1; Fig. 44). 117. Longitudinal ridge covering laterosensory pores of dentary: (0) absent; (1) present. The laterosensory canal of the dentary has a variable number of pores. These pores are evident in ventrolateral view in most examined species (state 0). In Charax stenopterus and the examined species of Roeboides Günther, there is a longitudinal bony ridge covering partially these pores which are thus visible only from a ventral view (state 1; Fig. 57). Dentition Generalities: 118. Morphology of premaxillary, maxillary, and dentary teeth: (0) all teeth conical, caniniform, or mamilliform; (1) some teeth multicuspidate or molariform. (FF44m, UJ2, BU72m, LU47m, LC23m, LU53m). According to Uj (1990), the presence of multicuspidate teeth (state 1) is a synapomorphy of the Characiformes, although his conclusion was not based in a cladistic analysis but in the optimization of certain features on a tree arrived at without an explicit phylogenetic method. Lucena (1993) considered this variation under two different characters, involving the maxillary and dentary teeth respectively. According to the phylogeny proposed by Lucena (1993) the possession of multicuspidate teeth constitute independent synapomorphies for the Lebiasinidae and a clade including most of the Characidae, with a reversal to conical teeth in a clade composed of the genera Aphyocharax and Phenacogaster. 119. Premaxillary, maxillary, and dentary teeth: (0) not pedunculate, or pedunculate only in some of these bones; (1) pedunculate and uniformly shaped. (MA55m, MA56m, BÜ70m). Multicuspidate teeth of most species, when present, have a broad base and somewhat variable form and size on the premaxilla, maxilla, and dentary (state 0). The presence of pedunculate teeth of similar form in the upper and lower jaws (state 1; Figs. 42 and 53) was mentioned as two independent synapomorphies for the Cheirodontinae by Malabarba (1998a). The presence of pedunculate teeth on the premaxilla, maxilla, and dentary was observed in this study in most of the examined members of the Cheirodontinae and in Gymnocharacinus bergii and Odontostoechus lethostigmus. Although the teeth of the latter two species are less compressed that those in members of the Cheirodontinae, they are coded with the state 1 based on overall form. The teeth in both jaws of Odontostilbe pequira are pedunculate, but having a slightly different morphology; this species, however, is also coded as state 1 given the small degree of that difference. Premaxillary teeth: 120. Mamilliform teeth outside mouth: (0) absent; (1) present. (LU41mi). The presence of teeth outside the mouth (state 1; Figs. 16 and 21) associated with lepidophagous habits has been cited for several genera in the Characiformes. Such dentition is absent in most examined species (state 0). The presence of three mamilliform teeth on the premaxilla oriented outside the oral cavity was proposed by Lucena (1993) as a synapomorphy of Roeboides. Fig. 60. Lower jaw of Triportheus nematurus, CI-FML 3948, 82.1 mm SL, medial view, anterior to left. AAR: anguloarticular, DEN: dentary, LCD: laterosensory canal of dentary, MEC: Meckelian cartilage. Scale bar = 1 mm. Fig. 61. Lower jaw of Astyanax pelegrini, CI-FML 3847, 62.5 mm SL, medial view, anterior to left. AAR: anguloarticular, CME: coronomeckelian, DEN: dentary, LCD: laterosensory canal of dentary, MEC: Meckelian cartilage. Scale bar = 1 mm. Phylogeny of the family Characidae 420 Exodon paradoxus and Roeboexodon geryi have only one or two mamilliform teeth in the premaxilla. Since the morphology of such teeth is very similar to those of Roeboides, these species are coded with the state 1 of this character. The teeth of Probolodus heterostomus have a similar overall shape to those present in Roeboides, but are tricuspidate. In Bryconexodon juruenae there is variation between unicuspidate and tricuspidate mamilliform teeth which are much similar to those of Exodon Müller & Troschel and Roeboexodon Géry and Probolodus Eigenmann, respectively. Given the extra-oral position of the teeth of Bryconexodon juruenae and Probolodus heterostomus and the rather similar shape of these teeth with those of the species of Roeboides both species are coded as state 1. 121. A pair of large conical teeth in premaxilla: (0) absent; (1) present. (LU40i). In most examined characids there are one to three rows of teeth in the premaxilla; the teeth of each row are usually rather homogeneous in morphology (state 0). Lucena (1993) mentioned the presence of two conical teeth much larger than the remaining ones, situated near the tips of the premaxillary alveolar ramus (state 1; Fig. 51) in several characids. In the examined specimens of Charax stenopterus these teeth are slightly larger than the remaining ones, and this species is coded with state 1. 122. Number of rows of premaxillary teeth: (0) one; (1) two or three. (AM27m, LU38m, LU39m, BU32m, MA55m, CM37, MO3, VB75i, BE27, ZV57, LI11m, BÜ35m, QU80m). 123. Number of rows of premaxillary teeth: (0) one or two; (1) three. (EI4m, AM27m, LU38m, LU39m, BU32m, MA55m, CM37, MO3, VB75i, BE27, ZV57, LI11m, BÜ35m, QU80m). The homology between the premaxillary rows of teeth among different species is not always easy to establish, especially when entire rows are lacking. The main criterion used herein to recognize, at least tentatively, teeth from different rows is developmental. The premaxillary teeth from the anterior two rows have intraosseous development (sensu Trapani, 2001), while those from the inner row have extraosseous development, growing in the soft connective tissue situated posterior to the alveolar ramus of the premaxilla. However, in some cases the discrimination between teeth from different rows could only be done following a topological criterion. Three states can be recognized among the examined species. In some species only one row of premaxillary teeth, with extraosseous development, is present (character 122, state 0). An anterior row of teeth, with intraosseous development, is present in most of the Characidae resulting in two rows of premaxillary teeth (character 122, state 1; character 123, state 0; Fig. 64). In a relatively small group of species there is an additional row of teeth, which also has intraosseous development but whose homology is difficult to establish and remains to be studied in detail. In the latter state there are three rows of premaxillary teeth (character 123, state 1; Zanata & Vari, 2005: fig. 15). As noted above, according to the hypothesis of Lucena (1993) the presence of two premaxillary rows of teeth is a synapomorphy for a clade corresponding to the Characidae, with parallel reversals to a single row in a clade composed of the genera Roestes, Lonchogenys Myers, Heterocharax, Gnathocharax, and Hoplocharax, and the clade composed of Hemigrammopetersius Pellegrin and Hydrocynus Cuvier. According to Lucena (1993) the possession of three rows of premaxillary teeth constitutes two independent synapomorphies for the clades composed of Creagrutus and Piabina and Brycon and Chalceus, respectively. In the hypothesis of Buckup (1998) the presence of two or three premaxillary rows of teeth is a synapomorphy for a group composed of the Alestidae, Characidae, Ctenoluciidae, Hepsetidae, Lebiasinidae, Erythrinidae, and Acestrorhynchidae, with a reversion in a subclade formed by the last five families. This is, however, just one of the possible most parsimonious optimizations for this character [the optimization produced by ACCTRAN (de Pinna, 1991)]. It is equally parsimonious to propose parallel acquisitions of two or more premaxillary rows of teeth in the Alestidae and Characidae without reversals. Malabarba (1998a) proposed the presence of only one functional row of smoothly aligned and similarly shaped premaxillary teeth as a synapomorphy of the subfamily Cheirodontinae. According to the phylogenetic hypothesis of Zanata & Vari (2005) the presence of two rows of premaxillary teeth is plesiomorphic for the Characiformes. The first of these characters is coded herein as inapplicable to Acestrorhynchus pantaneiro, Bryconexodon juruenae, Charax stenopterus, Cynopotamus argenteus, Exodon paradoxus, Galeocharax humeralis, Heterocharax macrolepis, Hoplias cf. malabaricus, Hoplocharax goethei, Lonchogenys ilisha, Oligosarcus cf. jenynsii, O. bolivianus, Rhaphiodon vulpinus, Roeboexodon geryi, Roeboides spp., and Serrasalmus maculatus. Although some of these species have a single premaxillary tooth row, these teeth are variable in size and/or morphology, probably corresponding to teeth from both the outer and the inner rows of other species. In Apareiodon affinis and Parodon nasus the premaxillary teeth originate from Fig. 62. Lower jaw of Iguanodectes geisleri, MHNG 2177.10, 49.0 mm SL, medial view, anterior to left. AAT: anterior adductor mandibulae tendon. DEN: dentary, MEC: Meckelian cartilage. Scale bar = 1 mm. J. M. Mirande 421 deep grooves separated by vertical lamellae; this condition is not comparable to that of the outer row of other species, and the character 122 is coded as inapplicable for these species. The single premaxillary teeth row of Leporinus striatus has an intraosseous development and it was thus considered as nonhomologous with the outer row of other species. Quevedo (2006) coded Paragoniates alburnus as having just one premaxillary tooth row; however, in the specimen examined here, there are two teeth situated distinctly anterior of the remaining five teeth, as if they were in a separate anterior row. The first of these characters is coded as polymorphic for P. alburnus, considering possible intraspecific variations. A variable presence of one or two premaxillary tooth rows was mentioned in the description of Inpaichthys kerri (Géry & Junk, 1977). In the observed specimens all the premaxillary teeth have extraosseous development and would be homologous to the teeth from the inner row of other species. Inpaichthys kerri is consequently coded as state 0 of the character 122. The teeth of the genera Creagrutus and Piabina were described by Vari & Harold (2001) as composed of a triangular cluster of medial teeth (or triad), a primary row of teeth, and a single lateral tooth, which may be either present or absent. Although this characterization is useful to compare between species of these two genera, it is not easy to homologize the teeth of these genera with those of other characids based on these definitions. The primary row, as described by Vari & Harold (2001) is composed of teeth that have both intraosseous (those situated medially) and extraosseous development (those situated just lateral to the posterior teeth of the triad), which differs from the criteria used herein to recognize homologies between tooth rows. These definitions can be seen as homology statements which are useful at different levels of analysis; however, there is clearly still much uncertainty relative to the homologies between characid tooth rows. The analyzed species of Creagrutus are coded with the state 1 of the character 123, while Piabina argentea is coded as polymorphic for this character, given that its teeth could be interpreted as disposed in three rows or in two rows, with the anteriormost in a zigzag pattern. These characters are considered as inapplicable to species lacking premaxillary teeth. Outer row of premaxillary teeth: 124. Alignment of teeth on anterior premaxillary row: (0) aligned; (1) not aligned, with one or two teeth situated anterior to remaining teeth. In most species with two rows of premaxillary teeth, the anterior row is composed of teeth aligned in a shallow arch (state 0; Figs. 65). In Bryconamericus exodon and B. cf. exodon, there are one or two teeth relatively displaced anteriorly, giving to the outer row a zigzag-shape (state 1). The correspondence between these anteriorly displaced teeth with those of the outer row in species with three rows of premaxillary teeth is unclear; therefore, this character is considered inapplicable to species coded as having three premaxillary rows of teeth. The examined specimens of Bryconamericus scleroparius (Regan), Hemibrycon dariensis, H. surinamensis, and Knodus breviceps have intermediate conditions that are coded as polymorphisms. This character is coded as inapplicable to species lacking an outer row of premaxillary teeth with intraosseous development. 125. Cusps of teeth on outer premaxillary row: (0) one to three cusps; (1) five or more cusps. (BE28). The premaxillary teeth of the outer tooth row are relatively slender and tricuspidate in most examined species (state 0; Fig. 55). In a relatively small group of species these teeth are instead expanded distally and have five or more cusps (state 1; Fig. 66). In the examined specimens of Astyanax cf. rutilus, Bryconamericus alpha, and Hemibrycon dariensis the teeth of the outer premaxillary row have five cusps, but the lateral ones are minute and the distal portion of teeth is not expanded. This condition is thus considered to be intermediate, and these species are coded as polymorphic. This character is coded as inapplicable to species with only one row of premaxillary teeth. Fig. 63. Lower jaw of Serrasalmus maculatus, CI-FML 3873, 69.2 mm SL, lateral view, anterior to left. DEN: dentary. Scale bar = 1 mm. Fig. 64. Premaxilla of Astyanax chico, CI-FML 3913, 60.2 mm SL, ventral view, anterior to bottom. Scale bar = 0.5 mm. Phylogeny of the family Characidae 422 Inner row of premaxillary teeth: 126. Teeth on inner premaxillary row: (0) molariform; (1) with aligned in straight series or anteriorly concave pattern of cusps; (2) with anteriorly concave pattern plus anterior cusps. (ZV64m). The cusps of teeth of the inner premaxillary row, when present, are either aligned in a straight line or form an anteriorly concave arch (state 1; Figs. 64, 65, and 67). In Alestes cf. macrophthalmus, Brycinus carolinae, and Bryconaethiops macrops, among the examined species, there is, in addition, a series of anterior cusps separated by a concave surface from the posterior cusps. The anterior series of cusps and the concave surface jointly give to these teeth a molariform aspect (state 2; Zanata & Vari, 2005: fig. 15). The teeth of the inner premaxillary row in Metynnis maculatus and Piaractus mesopotamicus have a molariform form, but their anterior margin is not formed by a series of cusps, but rather an even margin (state 0). This character is considered as inapplicable to species lacking a defined inner row of premaxillary teeth. 127. Alignment of cusps of medial teeth on inner premaxillary row: (0) forming anteriorly concave semicircle from ventral view; (1) forming shallow arch or aligned in straight series from ventral view. (ZV63i). As mentioned above, the cusps of the inner premaxillary teeth row have an anteriorly concave arrangement from ventral view. In most examined species this concavity is relatively shallow, describing approximately one-fourth of a circle, or less, from ventral view (state 1; Figs. 64 and 67). Zanata & Vari (2005) mentioned that this concavity is much pronounced (state 0; Fig. 65) in some alestids and in the characid Triportheus albus. An intermediate state was herein observed in Bryconops affinis Günther, which is coded as polymorphic. This character is coded as inapplicable to species without a definite inner row of premaxillary teeth. 128. Form of teeth of inner premaxillary tooth row: (0) with cusps forming anteriorly concave arch; (1) with cusps aligned in straight series and without anterior concavity. (MO4). As mentioned above the anterior surface of the teeth of the inner premaxillary row has an anteriorly oriented concavity in ventral view resulting from the arched position of the cusps on these teeth (state 0; Figs. 64 and 65). In a relatively small group of species these teeth are much compressed anteroposteriorly and, from a ventral view, their cusps follow an almost straight line (state 1; Fig. 67). This character and the preceding one represent a series of nested homologies and they can be seen as a single additive character with three states. This character is considered as inapplicable to those species lacking a defined inner row of premaxillary teeth. Both this character and the preceding one are coded as inapplicable to Agoniates anchovia, Aphyocharax spp., Aphyodite grammica, Axelrodia lindeae, Grundulus cochae, Inpaichthys kerri, Microschemobrycon casiquiare, Parecbasis cyclolepis, Prionobrama paraguayensis, Roeboides microlepis, and Salminus brasiliensis which have conical teeth or with minute lateral cusps. 129. Number of teeth in inner premaxillary row: (0) four or fewer; (1) five or more. (VB76, BE29i, ZV61i, BÜ41m). 130. Number of teeth in inner premaxillary row: (0) seven or fewer; (1) eight or more. (BÜ41m). The presence of four teeth on the inner premaxillary row (character 129, state 0) was used in the generic key of Eigenmann (1917) to distinguish Argopleura Eigenmann, Bryconamericus, Ceratobranchia, Creagrutus, Hemibrycon, Knodus, Microgenys, Nematobrycon Eigenmann, and Piabina from the remaining characids, which usually have five teeth in that series (character 129, state 1; character 130, state 0; Fig. 64). Géry (1977) included most genera with four Fig. 65. Premaxilla of Astyanax pelegrini, CI-FML 3847, 62.5 mm SL, ventral view, anterior to bottom. Scale bar = 0.5 mm. Fig. 66. Premaxilla of Astyanax chico, CI-FML 3913, 60.2 mm SL, anterior view, anterior to left. Scale bar = 1 mm. J. M. Mirande 423 teeth on the inner premaxillary row in his tribe Tetragonopterini, within a subgroup therein called “genus Hemibrycon and allied genera” which included Boehlkea, Bryconacidnus, Bryconamericus, Carlastyanax Géry, Ceratobranchia, Coptobrycon Géry, Creagrudite Myers, Hemibrycon, Knodus, Microgenys, Nematobrycon, Piabarchus, Rhinobrycon, and Rhinopetitia. Although the phylogenetic importance of the possession of four premaxillary teeth on the inner row was not adequately tested, this character was later proposed by Malabarba & Weitzman (2003) to be a putative synapomorphy of their clade A. The presence of eight or more teeth on the inner premaxillary row (character 130, state 1; Fig. 68) was mentioned by Bührnheim (2006) for several small species of the Characidae. The first of these characters is coded as polymorphic in Bryconamericus cf. rubropictus, Cyanocharax alburnus, Diapoma terofali, and Hemibrycon surinamensis, and the second one in Aphyocharax dentatus Eigenmann & Kennedy, Inpaichthys kerri, and Pseudocorynopoma doriae. 131. Polymorphism of teeth on inner premaxillary row: (0) absent; (1) present, with two medial teeth somewhat larger and usually separated from remaining ones by a gap. In most examined species the two medial teeth of the inner premaxillary tooth row are slightly larger than the remaining teeth, although the homology of these teeth in different species is unclear. These medial teeth are probably homologous to the posterior teeth of the triad described by Vari & Harold (2001) for Creagrutus and Piabina. Apparently this identification was followed by Zanata & Vari (2005: 34), who considered Brycon and Triportheus to have two teeth on the posteriormost premaxillary row of teeth, corresponding to the medial teeth of the inner row. Although the homology of the premaxillary teeth of different species remains to be studied, the teeth from the posteriormost premaxillary row of the examined species are usually concave anteriorly and have extraosseous development. In Brycon spp., Metynnis maculatus, Piaractus mesopotamicus, and Triportheus, among the examined species, these medial teeth are slightly larger and separated from the lateral ones by a gap (state 1). In other species the two medial teeth are not distinctly larger than the third tooth and the second and third tooth are separated each other in the same extent than the remaining pairs of teeth of the inner premaxillary tooth row (state 0). 132. Number of replacement tooth rows on premaxilla: (0) one; (1) two or more. Most examined species have a single series of inner premaxillary row replacement teeth that demonstrate extraosseous development and which usually become functional simultaneously (state 0). Hemiodus cf. thayeria and the examined species of the Parodontidae have two or more rows of replacement teeth which apparently become functional sequentially (state 1; e. g. Roberts, 1974: fig. 7). Although the homology of particular teeth among Characiformes is uncertain, the simultaneous presence of two or more rows of replacement premaxillary teeth with, at least partially, extraosseous development is tentatively coded as present in the members of the Hemiodontidae and Parodontidae and as absent in the remaining species. 133. Fossa for inner row of replacement premaxillary teeth: (0) absent; (1) present. The teeth of the inner premaxillary row in most examined species undergo a completely extraosseous development being formed in the soft connective tissue just posterior to the premaxilla (state 0). In a small group of species these teeth rather form within shallow cavities or fossae situated on the posterior surface of the premaxilla (state 1; Fig. 68). Fig. 67. Premaxilla of Deuterodon iguape, MHNG 2183.6, 52.0 mm SL, ventral view, anterior to bottom. Scale bar = 0.5 mm. Fig. 68. Upper jaw of Aphyocharax dentatus, CI-FML 3771, 53.2 mm SL, anterolateral view, anterior to left. MAX: maxilla, PMX: premaxilla. Scale bar = 1 mm. Phylogeny of the family Characidae 424 Maxillary teeth: 134. Maxillary teeth: (0) absent; (1) present. (LU44m, MO6m, SE46m, BE34m, ZV78, LI12, BÜ54m). The presence (state 1) or absence (state 0) of maxillary teeth was longly considered as significant in the classification of the Characidae. This character is coded as polymorphic in Hasemania nana, Hemigrammus bleheri, Moenkhausia dichroura, and Thayeria obliqua, in which there occurs either cited or herein observed variation in this character. 135. Number of maxillary teeth: (0) only one, or absent; (1) two or more. (LU45m, WM42m, CM38m, BE34m, VH7m, BÜ54m). 136. Number of maxillary teeth: (0) up to three; (1) four or more. (LU45m, WM42m, CM38m, BE34m, VH7m, BÜ54m). The number of maxillary teeth has been considered in almost all published phylogenies of genera and subfamilies of the Characidae. The variation in number of maxillary teeth, however, is almost continuous among the examined species, and any definition of discrete states within this character would have some degree of subjectivity. The number of maxillary teeth is not distributed normally among the examined species, and many species have none (Fig. 54) to four teeth, while relatively few have more than ten teeth. In species with low number of maxillary teeth, this number is relatively more stable intraspecifically, and differences between one or two teeth could be phylogenetically informative, while species with high numbers of maxillary teeth also have relatively higher intraspecific variations in this feature, and the gaining or loss of one teeth has comparatively less correlation with phylogeny and species discrimination. The option of analyzing this character as a lineal continuous character, implemented in TNT (Goloboff et al., 2003, 2006), tends to overestimate the phylogenetic information of the transformations in groups with higher and variable number of maxillary teeth, in relation to groups with lower number of teeth (e. g. a transformation between one to three teeth will have lower weight than a transformation between 20 to 25 teeth). The coding of this character takes into account both personal observations and published information about the number of maxillary teeth in order to consider intraspecific variations, which in several species results in polymorphisms. 137. Extent of implantation of teeth along maxilla: (0) not reaching middle of maxillary lamella; (1) extending across almost entire maxillary lamella. (EI3, LU44m, MO6, VB79, SE46m, QU82). The posterior extent of the area of implantation of the maxillary teeth is one of the “generic” level characters proposed by Eigenmann (1917). Eigenmann noted the presence of teeth along the entire anterior margin of the maxilla (state 1; Figs. 51 and 69) in Hollandichthys, Nematobrycon, Phenacogaster, and Pseudochalceus Kner. Lucena (1993) divided this character to three states, considering the absence of maxillary teeth, the presence of teeth only in the anterior half of the maxillary margin (state 0; Fig. 68), and the presence of teeth along the entire anterior margin of the maxilla. The examined specimens of Odontostoechus lethostigmus and Probolodus heterostomus have teeth only in the anterior half of the maxillary lamella, and these species are coded with state 0. In Carlana Strand and Rhoadsia Fowler, the maxilla is progressively longer and bears an increasingly number of teeth during ontogeny, with the dentition extending along the entire maxillary margin in adult specimens. Given that young specimens of Carlana eigenmanni and Rhoadsia altipinna have state 0 and adults exhibit state 1, these species are coded as polymorphic. 138. Number of cusps of anterior maxillary teeth: (0) conical, a single cusp; (1) three or more cusps. (LU43m, MO7m, VB81m, SE48m, BE35, BÜ55i, QU81m). 139. Number of cusps of anterior maxillary teeth: (0) up to three; (1) five or more cusps. (LU43m, MO7m, VB81m, SE48m, BE35, BÜ55i, QU81m). Maxillary teeth usually have a similar morphology to that from the other jaw bones; however, some species with five or more cusps in the premaxillary and dentary teeth, have only conical (character 138, state 0; Fig. 69) or tricuspidate (character 138, state 1; character 139, state 0; Fig. 50) teeth on the maxilla. Although the number of cusps on the maxillary teeth is somewhat correlated with the number of cusps of other teeth, it is coded as a separate character as a function of the lack of correlation in some species. Five cusps were herein observed Bryconops melanurus instead of three, as reported by Benine (2004), and this species is coded as polymorphic. Benine (2004) mentioned the presence of five or more cusps in the maxillary teeth (character 139, state 1; Fig. 52), among others, in Bario steindachneri, Hemigrammus unilineatus, Hyphessobrycon eques, and Poptella paraguayensis (Eigenmann). This character is variable among the examined specimens of Bario steindachneri, Hyphessobrycon eques, and Poptella paraguayensis, which are coded as polymorphic for character 139. The examined specimen of Hemigrammus unilineatus has only three cusps on the maxillary teeth. This character is also coded as polymorphic for this species following the observations of Benine (2004). The maxillary teeth of Salminus brasiliensis are essentially conical but have lateral projections that appear to be rudimentary cusps. This condition is herein considered to be intermediate between the states defined for the character 138, and Salminus brasiliensis is coded as polymorphic for this character. Variation in the character 139 was observed in Astyanax chico, A. cf. eigenmanniorum, A. endy Mirande, Aguilera & Azpelicueta, A. lineatus, Bryconops affinis, Hyphessobrycon anisitsi (Eigenmann), H. eques, Prodontocharax melanotus, and Pseudocorynopoma doriae and this character is coded as polymorphic for these species. 140. Ontogenetic acquisition of conical teeth on maxilla: (0) absent; (1) present. J. M. Mirande 425 The proportional lengthening of the maxilla and the acquisition of conical maxillary teeth during the ontogeny (state 1; Fig. 70), were mentioned above as characteristic for members of the subfamily Rhoadsiinae by Fink & Weitzman (1974). Notwithstanding those observations, Fink & Weitzman classified the genus Carlana within the “so-called Cheirodontin fishes” due to the presence of only one row of premaxillary teeth. Carlana eigenmanni and Rhoadsia altipinna, the only analyzed members of Rhoadsiinae, are herein coded as state 1. This character is coded as unknown for Nematocharax venustus Weitzman, Menezes & Britski; although the maxillary dentition of adults of this species bear certain similarities with that of the Rhoadsiinae, an ontogenetic series of this species was unavailable for study and this character could not be analyzed. Dentary teeth: 141. Orientation of anterior dentary teeth: (0) oriented dorsally or anterodorsally; (1) oriented anteriorly, almost parallel to main axis of dentary. In most examined species the anterior teeth of dentary are situated perpendicular to or slightly oblique to the longitudinal axis of the bone, (state 0; Fig. 60). Some species instead have the anteriormost dentary teeth oriented anteriorly, in a plane almost parallel to the longitudinal axis of the dentary (state 1; Fig. 45). 142. Size and number of anterior dentary teeth: (0) four or five relatively broad teeth at front of dentary; (1) eight or more small and slender teeth at front of dentary. Most examined species have four or five relatively broad teeth situated in the anterior region of the dentary (state 0; Fig. 59). In a relatively small group of species, the anterior dentary teeth are more slender, and eight or more teeth are situated at the anterior portion of the dentary (state 1). This character is coded as polymorphic in Grundulus cochae which has six or seven teeth in the anterior region of the dentary. 143. Inner row of dentary teeth: (0) present; (1) absent. (UJ50, BU36i, LU49i, CM41m, MO10, SE63m, LI13m). The inner row of dentary teeth, when present, is situated just posterior to the replacement fossa for the anterior dentary teeth (Buckup, 1998) (state 0; Figs. 60 and 71). This inner row of dentary teeth is absent in most members of the Characidae (state 1). Distichodus and Triportheus were coded as state 1 by Buckup (1998) and Lucena (1993) respectively; however, an inner row of dentary teeth was observed in the examined species of both genera and they are herein coded as state 0. The members of the Crenuchidae analyzed by Buckup (1998) have two rows of teeth in the dentary; however the examined species of Characidium have only one row of dentary teeth and they are herein coded as state 1. The examined specimen of Hemiodus cf. thayeria lack dentary teeth. Juvenile specimens of this genus have teeth which are lost ontogenetically; although no juveniles of H. cf. thayeria were examined for this paper, the inner tooth row of dentary is absent in juveniles according to Langeani (1998), and this character is coded as state 1 for this species. 144. Symphyseal dentary teeth: (0) absent; (1) present. (UJ46, LU50i, CM41m, SE63m, ZV88, LI14). In most species with two rows of dentary teeth, the posterior row is composed of numerous and minute conical teeth situated along the ridge posterior to the replacement fossa of the dentary (state 0; Fig. 71). In a group of species, the inner row of dentary teeth has a symphyseal tooth much larger than the remaining teeth, a broad diastema, and a posterior series of teeth similar to those present in state 0 (state 1; Fig. 60). This character is coded as inapplicable in species lacking the entire inner row of dentary teeth. 145. Articulation between dentary teeth: (0) absent; (1) present with associated processes and fossae. (AM6). Fig. 69. Upper jaw of Phenacogaster tegatus, CI-FML 3880, 35.3 mm SL, anterolateral view, anterior to left. MAX: maxilla, PMX: premaxilla. Scale bar = 0.5 mm. Fig. 70. Maxilla of Carlana eigenmanni, LACMNH 9230.020, 57.6 mm SL, medial view, anterior to left. Scale bar = 1 mm. Phylogeny of the family Characidae 432 character is coded herein as polymorphic. Intraspecific variation was observed in Astyanax troya, Creagrutus anary Fowler, Microschemobrycon casiquiare, and Oligosarcus sp. in which this character is similarly coded as polymorphic. 176. Bony lamellae bordering laterosensory canal of suprapreopercle: (0) absent; (1) present. The suprapreopercle is usually limited to a tubular laterosensory canal without associated lamellae (state 0). Variably developed bony lamellae associated with the laterosensory canal of the suprapreopercle were observed in Markiana nigripinnis and Oligosarcus spp., (state 1; Fig. 34). Branchial and hyoid arches Anterior ceratohyal: 177. Anterior projection of anterior ceratohyal articulating laterally with hypohyals: (0) absent or much reduced; (1) present and achieving half length of hypohyals. (UJ69). In most examined species, the proximal portion of the anterior ceratohyal articulates synchondrally with the hypohyals without bony contact between these bones (state 0). Uj (1990) mentioned the presence of a process laterally bordering the hypohyals in the African alestid Hydrocynus and the South American genera Acestrorhynchus, Agoniates, and Rhaphiodon (state 1). An intermediate state was observed in Heterocharax macrolepis among the examined species, which is coded as polymorphic. 178. Hyoid artery: (0) completely contained within anterior ceratohyal in passage from posterior ceratohyal to hypohyals; (1) emerging from anterior ceratohyal near its articulation with posterior ceratohyal. (BU25, BE50, ZV111). A segment of the hyoid artery enters in the posterior ceratohyal and is oriented towards the dorsal hypohyal through the anterior ceratohyal. Castro (1984) and Buckup (1998) noted that in some taxa this artery is contained completely within the anterior ceratohyal as it passes to the dorsal hypohyal (state 0; Figs. 80 and 81), whereas in other species it exits by a pore near the posterodorsal margin of the anterior ceratohyal and then passes into an opening between the anterior ceratohyal and the dorsal hypohyal (state 1; Fig. 82). This state is present in most examined species of the Characidae, while state 0 is present in all members from the outgroup and some genera of the Characidae. A morphologically intermediate state was observed in Hoplocharax goethei and Piabucus melanostomus which are coded as polymorphic. In these species, the hyoid artery emerges at middle length of the anterior ceratohyal, but is contained in a deep open canal along the dorsal margin of the bone. 179. Ventral margin of anterior ceratohyal: (0) smooth and without notches; (1) with notches for articulation of branchiostegal rays. (LU68, BÜ79m). The branchiostegal rays articulate with the ventral margin of the ceratohyals. Lucena (1993) mentioned the presence of notches along the margin of the anterior ceratohyal in most species of the Characidae (state 1; Figs. 81 and 82). These notches are absent (state 0; Fig. 80) in several characids and in most members of the outgroup. A series of small cartilages coincident in shape and position with the margins of the notches of other species were observed in the members of the Serrasalmidae. These cartilages are probably homologous with portions of the anterior ceratohyal forming these notches. Metynnis maculatus, Piaractus mesopotamicus, and Serrasalmus maculatus are thus coded as polymorphic for this character. Much reduced notches are present in the examined specimens of Hoplocharax goethei which is also coded as polymorphic. Fig. 80. Detail of hyoid skeleton of Agoniates anchovia, MHNG 2388.93, 158.5 mm SL, anterolateral view, anterior to left. ACH: anterior ceratohyal, BRR: branchiostegal rays, PCH: posterior ceratohyal. Scale bar = 5 mm. Fig. 81. Detail of hyoid and branchial skeleton of Salminus brasiliensis, CI-FML 3784, 131.4 mm SL, anterolateral view, anterior to left. ACH: anterior ceratohyal, PCH: posterior ceratohyal. Scale bar = 1 mm. J. M. Mirande 433 180. Number of notches along ventral border of anterior ceratohyal: (0) zero to two; (1) three. Species in the Characiformes usually have three branchiostegal rays articulating with the ventral margin of the anterior ceratohyal. In these species, the anterior two branchiostegal rays have a rather pedunculate anterior portion, which articulates with the notches of the ventral margin of the anterior ceratohyal. The third branchiostegal ray lacks a basal pedicle and articulates with the posteroventral surface of the anterior ceratohyal (state 0). In some members of the Cheirodontinae there is a third notch in the ventral margin of the anterior ceratohyal where the third branchiostegal ray articulates. This ray, however, lacks a developed anterior pedicle (state 1). The presence and development of this third notch is variable among the examined specimens of Aphyocharax dentatus and this species is coded as polymorphic. 181. Articulation between anterior and posterior ceratohyals: (0) synchondral, without bony interdigitations; (1) with bony interdigitations. (VA57, LI37). In most examined species, the anterior ceratohyal articulates synchondrally with the posterior ceratohyal without any bony interdigitations between the bones (state 0; Figs. 80 and 82). Uj (1990) mentioned the presence of interdigitations between these bones (state 1; Fig. 83) as a synapomorphy of his Cynopotaminae (=Characinae, in part). Vari (1995) mentioned the presence of such interdigitations as a synapomorphy of a clade including the families Ctenoluciidae and Erythrinidae, and independently occurring in the lebiasinid genera Lebiasina Valenciennes and Piabucina Valenciennes. Basibranchials: 182. First basibranchial: (0) absent or much reduced, not articulating anteriorly with basihyal; (1) well developed and articulating anteriorly with basihyal. Members in the Characiformes have three or four ossified basibranchials situated medial to the hypobranchials. The anterior margin of the first basibranchial articulates with the posterior margin of the basihyal in almost all the examined taxa (state 1; Fig. 84). In some members of the outgroup, the first basibranchial is absent or much reduced, and it does not contact the posterior margin of the basihyal (state 0). 183. Contact between lamella on anterior portion of first basibranchial with lamella on posterior portion of second basibranchial: (0) absent; (1) present. Basibranchials usually bear, anteriorly and/or posteriorly, thin bony lamellae which project dorsal to the cartilages situated between their main portions. These lamellae apparently develop as autogenous ossifications and usually fuses to the main portion of the basibranchials during ontogeny, as suggested by the presence of autogenous lamellae in some examined specimens. The bony lamellae are usually situated between the first and second and the second and third basibranchials. Some species have also a bony lamella situated dorsal to the fourth basibranchial, which is usually completely cartilaginous. The bony lamellae situated between the first and second basibranchials are usually absent or not contacting each other (state 0; Fig. 86), whereas in some species these lamellae form a bony bridge between the main portions of the first and second basibranchials (state 1; Fig. 85). 184. Bony lamellae between second and third basibranchials: (0) absent; (1) present. (PZ66m). The presence of bony lamellae between the basibranchials is variable among the examined species and often independent in terms of occurrence between pairs of basibranchials; this character is therefore considered as not Fig. 82. Detail of hyoid skeleton of Axelrodia lindeae, MCP 37314, 18.7 mm SL, anterolateral view, anterior to left. ACH: anterior ceratohyal, BRR: branchiostegal ray, PCH: posterior ceratohyal. Scale bar = 0.1 mm. Fig. 83. Hyoid and branchial skeleton of Galeocharax humeralis, CI-FML 3951, 94.6 mm SL, lateral view, anterior to left. ACH: anterior ceratohyal, BRR: branchiostegal ray, PCH: posterior ceratohyal. Scale bar = 1 mm. Phylogeny of the family Characidae 434 correlated with the preceding one. The bony lamellae between the second and third basibranchials are present in most examined species (state 1; Fig. 84). In some species these lamellae are completely absent, and the space between the main portion of the second and third basibranchials is filled only by cartilage (state 0; Fig. 86). 185. Bony lamella dorsal to fourth basibranchial: (0) present; (1) absent. (LU74, VH32, MO56, BE53i, LI43i, BÜ82i, PZ67i). The fourth basibranchial is completely cartilaginous in most examined species. Lucena (1993) mentioned the presence of an ossified fourth basibranchial in many species of the Characidae; however, based on his coding, it is likely that Lucena was instead referring to the bony lamellae situated just dorsal to the fourth basibranchial (state 0; Figs. 84 and 87). In the phylogenetic hypothesis of Lucena (1993), the presence of this lamella is a synapomorphy of a clade including most characids, with a reversal in a clade composed of Creagrutus, Ctenobrycon Eigenmann, Iguanodectes, and Piabina. Contrary to the observations of Lucena (1993) and Benine (2004), this ossification is present in the examined specimens of Bryconamericus exodon, Deuterodon iguape, Hemigrammus unilineatus, and Poptella paraguayensis; these species are coded as polymorphic, in light of probable intraspecific variations. The presence of this bone is variable among the examined specimens of Astyanax cf. eigenmanniorum1 and Probolodus heterostomus and this character is also coded as polymorphic in these taxa. The bony lamella dorsal to the fourth basibranchial is absent (state 1; Fig. 88) in the examined specimens of Exodon paradoxus and Hyphessobrycon herbertaxelrodi Géry. Since Lucena (1993) and Benine (2004), respectively, noted the presence of this lamella in these species this character is coded as polymorphic for these taxa. In the examined specimen of Prionobrama paraguayensis the fourth basibranchial is ossified and bordered dorsally by the bony lamella, and this species is therefore coded as state 1. 186. Main portion of fourth basibranchial: (0) completely cartilaginous; (1) ossified. As noted above, the fourth basibranchial is usually completely cartilaginous or has only a dorsal bony lamella (state 0; Figs. 87 and 88). In Phenagoniates macrolepis, Prionobrama paraguayensis, and Xenagoniates bondi among the examined species, the main portion of the fourth basibranchial is ossified (state 1). This character could not be examined in Paragoniates alburnus which is coded as a missing entry. 187. Teeth on lamella dorsal to fourth basibranchial: (0) absent; (1) present. (LM36, PZ68). Lucena (1993) proposed the presence of teeth on the fourth basibranchial (state 1) as independent autapomorphies of Acestrorhynchus and Agoniates. Such teeth are usually absent (state 0) in the Characiformes. In the phylogenetic hypothesis of Lucena & Menezes (1998) the presence of teeth on the lamella dorsal to the fourth basibranchial is also optimized as a parallelism between Acestrorhynchus and Agoniates. Basihyal: 188. Cartilages anterior to basihyal: (0) one or two blocks of cartilage, but anterior block much smaller; (1) two well developed blocks of cartilage. (VA56). The anterior margin of the basihyal has a cartilaginous margin bordered anteriorly by an autogenous block of cartilage which supports the anterior portion of the primary tongue (state 0; Figs. 84 and 88). Some species have two independent autogenous blocks of cartilage anterior to the basihyal; these blocks are approximately of the same width and as wide as the basihyal (state 1; Fig. 87). In some species the anterior block of cartilage is much smaller and these cases were coded state 0. Both the homology between these cartilages and possible ontogenetic variations should be assessed in more focused studies. Intraspecific variation in this character were observed in Astyanax cf. eigenmanniorum2, A. endy, A. latens, A. troya, Cheirodon interruptus, Hollandichthys multifasciatus, Hyphessobrycon socolofi, and Psellogrammus kennedyi, which are coded as polymorphic. 189. Edentulous basihyal lamella: (0) absent; (1) present. (ZV109i, LI42). Vari (1983: 24) described an “edentulous basihyal tooth plate” (state 1; Fig. 85; Vari, 1983: fig. 22) as a lamella situated just anterior and dorsal to the basihyal. This lamella is probably a serial homologous to the lamellae situated between the basibranchials. As this lamella, when present, lacks teeth in the Characiformes, is here referred as edentulous basihyal lamella. Such a lamella is absent in most examined species (state 0; Fig. 88). Zanata & Vari (2005) mentioned the presence of this lamella Fig. 84. Branchial skeleton of Bryconamericus exodon, CIFML 3897, 39.0 mm SL, dorsal view, anterior to left. BB1-3: first to third basibranchials, BHY: basihyal, VHH: ventral hypohyal. Scale bar = 1 mm. J. M. Mirande 435 in Hepsetus odoe (Bloch), Hoplias microlepis, Triportheus albus, Xenocharax spilurus, and all of the alestids examined by them. 190. Anterior development of basihyal: (0) broadly extending beyond anterior margin of hypohyals; (1) slightly surpassing anterior margin of hypohyals. The anterior margin of the basihyal usually extends distinctly anterior of the hypohyals, being the anteriormost element of the hyoid arch (state 0; Fig. 84). In the examined specimens of Carnegiella strigata, Thoracocharax stellatus, and Triportheus spp., the basihyal is relatively reduced, not extending anteriorly beyond the hypohyals, or at most slightly surpassing that elements (state 1; Fig. 85). 191. Form of anterior expansion of basihyal: (0) slender, with anterior margin less than two-thirds of its length; (1) expanded, with anterior margin with two-thirds or more of its length. (SE76m). The anterior margin of the basihyal supports the anterior portion of the primary tongue. The basihyal expands laterally as it progresses anteriorly from its relatively slender articulation with the first basibranchial. In most species this expansion is limited, and the anterior margin of the basihyal have less than two-thirds of its length (state 0; Fig. 88). In some species the anterior region of the basihyal is much expanded laterally and the anterior margin have two-thirds or Fig. 85. Anterior region of branchial skeleton of Triportheus nematurus, CI-FML 3948, 82.1 mm SL, dorsal view, anterior to left. BB1-2: first and second basibranchial, BHY: basihyal, EBL: edentulous basihyal lamella. Scale bar = 1 mm. Fig. 86. Detail of branchial skeleton of Galeocharax humeralis, CI-FML 3951, 94.6 mm SL, dorsal view, anterior to left. BB1 3: first to third basibranchial, CB1: first ceratobranchial, HB1: first hypobranchial. Scale bar = 1 mm. Fig. 87. Branchial skeleton of Aphyocharacidium bolivianum, MCP 37960, 26.3 mm SL, dorsal view, anterior to left. BB4: cartilaginous fourth basibranchial, BHY: basihyal. Scale bar = 1 mm. Fig. 88. Detail of branchial skeleton of Nematocharax venustus, MCP 17987, 34.3 mm SL, dorsal view, anterior to left. BB4: cartilaginous fourth basibranchial, BHY: basihyal. Scale bar = 1 mm. Phylogeny of the family Characidae 436 more of the length of this bone (state 1; Fig. 89). The basihyal in Bryconamericus cf. exodon, Cheirodon interruptus, Diapoma speculiferum, and Probolodus heterostomus has an intermediate state which is coded as polymorphic. Gill rakers: 192. Rows of gill rakers on first ceratobranchial: (0) one; (1) two. (LU73i, LC34, SE83m, BÜ85, PZ61i). 193. Rows of gill rakers on second ceratobranchial: (0) one; (1) two. (LC35, SE83m, PZ62mi). 194. Rows of gill rakers on third and fourth ceratobranchials: (0) one; (1) two. (PZ63mi, PZ64i). Most members of the outgroup have two rows of gill rakers on the first ceratobranchial (character 192, state 1), while most characids have only one row of gill rakers in this bone (character 192, state 0). The first row of gill rakers is situated along the leading margin of the ceratobranchial and the second row is located along the trailing margin. In almost all the examined species, there are two rows of gill rakers on the third and fourth ceratobranchials (character 194, state 1). The number of rows of gill rakers on the different ceratobranchials is not independent, according to the observations done for this paper. All species having two rows on the first ceratobranchial, have also two rows on the remaining ceratobranchials; correspondingly, all the species with two rows on the second ceratobranchial (character 193, state 1) have two rows on the third and fourth ceratobranchial, although not necessarily on the first ceratobranchial. In this way, this character can be seen as an ordered character with four states. The examined specimen of Cyanocharax alburnus has only one minute gill raker on the second row of the first ceratobranchial and this character is coded as polymorphic. 195. Number of gill rakers on first hypobranchial and ceratobranchial: (0) 16 or more; (1) 15 or fewer. (CM16(18)m, SE82m, BÜ90m). 196. Number of gill rakers on first hypobranchial and ceratobranchial: (0) 11 or more; (1) ten or fewer. (CM16(18)m, SE82m, BÜ90m). The number of gill rakers is continuously variable among the examined taxa. Therefore, the limits of the states used in characters 195 and 196 are rather subjective and they were defined, in part, considering their congruence with the remaining characters under preliminary analyses. Malabarba (1998b) defined the states of this character with different ranges. According to her phylogeny, a high number of gill rakers (more than 20) (character 195, state 0) is a synapomorphy of Triportheus. Intraspecific variations in the character 196 was observed in Astyanax abramis, A. cf. abramis, and Prodontocharax melanotus, and in the character 195 in Astyanax troya, Coptobrycon bilineatus, Cynopotamus argenteus, Galeocharax humeralis, Grundulus cochae, Hasemania nana, Hemibrycon surinamensis, Hollandichthys multifasciatus, Hoplocharax goethei, Jupiaba mucronata, Micralestes stormsi, Mimagoniates rheocharis, Nematobrycon palmeri, Odontostoechus lethostigmus, Roeboides descalvadensis, and Thayeria boehlkei. These species are coded as polymorphic for the variable character. 197. Shape of first ceratobranchial gill rakers: (0) pointed and not anteroposteriorly compressed; (1) laminar and much compressed perpendicular to ceratobranchial; (2) short, broad and strongly denticulate. (UJ70m, LU71m, LM35m, TP38, PZ65m). The gill rakers are morphologically rather homogeneous within the Characiformes, with a broad proximal region which articulates with the corresponding ceratobranchial and a progressively slender distal region (state 0; Fig. 83). Among the examined species of the Parodontidae the gill rakers are laminar, much reduced anteroposteriorly, and situated perpendicularly to the main axis of the ceratobranchial (state 1). Lucena (1993) mentioned the presence of short broad gill rakers bearing strong denticles in their dorsal surface (state 2; Toledo-Piza, 2000: fig. 12) in Acestrorhynchus pantaneiro, Cynodon gibbus (Agassiz), and Rhaphiodon vulpinus. This state is a synapomorphy of a clade composed on these genera in his phylogeny. According to Lucena & Menezes (1998), the optimization of this character is ambiguous, and this state could be a synapomorphy for their Acestrorhynchidae plus Cynodontidae, with a reversal in their Roestinae, or a parallelism between the Acestrorhynchidae and Cynodontidae. According to the phylogeny of Toledo-Piza (2007), the presence of the state 2 is a synapomorphy for a clade consisting of Acestrorhynchus, Cynodon, Hydrolycus Müller & Troschel, and Rhaphiodon. 198. Form of anterior gill rakers on first ceratobranchial: (0) not fused; (1) with fused bases forming plates extensively articulated with ceratobranchial. Fig. 89. Anterior region of branchial skeleton of Serrapinnus calliurus, CI-FML 3889, 23.1 mm SL, dorsal view, anterior to left. BHY: basihyal. Scale bar = 0.5 mm. J. M. Mirande 437 The gill rakers are lanceolate and similar in form each other in most examined species (state 0). In Cynopotamus argenteus and Galeocharax humeralis the posterior gill rakers of the first ceratobranchial are similar to those of other examined species, while the anterior ones are progressively shorter and broader in these species. The gill rakers situated along the anterior one-third of the first ceratobranchial and on the first hypobranchial are fused to each other, thereby forming strongly denticulate plates (state 1; Fig. 86). 199. Lateral base of gill rakers on first ceratobranchial: (0) slender; (1) broad and laminar at least on anteriormost gill rakers. The gill rakers are usually articulated with the first ceratobranchial by two slender anterolaterally and posteromedially oriented bases and aligned along an approximately transverse line relative to the main axis of the ceratobranchial (state 0; Fig. 83). In some species the anterolateral base of the gill rakers is expanded, forming a lamella extensively articulated with the first ceratobranchial (state 1; Fig. 81). This lamella is usually notched basally and appears to be composed of two parallel bony platelets. As mentioned under the preceding character, the anterior gill rakers of Cynopotamus argenteus and Galeocharax humeralis are expanded basally. This state, however, is not comparable to the states defined in this character, which is coded as inapplicable for those species. In Characidium spp., Carnegiella strigata, Puntius tetrazona, and Thoracocharax stellatus the anterolateral base of the gill rakers is reduced and this character is coded as inapplicable. 200. Form and degree of ossification of first ceratobranchial gill rakers: (0) laminar and not ossified distally; (1) rather thick and completely ossified distal region. In most examined species, the gill rakers are slender and are not ossified distally (state 0; Mirande et al., 2007: fig. 3). In some species the gill rakers are relatively stronger and completely ossified (state 1; Fig. 83). This character has an intermediate state in Hollandichthys multifasciatus which is coded as polymorphic. 201. Denticles on gill rakers: (0) present; (1) absent. (CM17(19)m, SE80m, BÜ93m). Denticles on the gill rakers (state 0; Mirande et al., 2007: fig. 3) are broadly distributed in the Characiformes, especially among the members of the Characidae. The presence and distribution of these denticles were considered by Bührnheim & Malabarba (2006) in the systematics of the genus Odontostilbe. The absence of these denticles was observed in most members of the outgroup and some characids (state 1). 202. Distribution of denticles on gill rakers: (0) restricted to margins, or absent; (1) along entire surface of gill rakers. The denticles of the gill rakers, when present, are usually more densely distributed along the anterior and posterior margins of the rakers, where they are approximately aligned into a row along each margin (state 0). In some species these denticles are more densely distributed and situated also on the lateral surfaces of the gill rakers, (state 1), as mentioned by Mirande et al. (2007) for Astyanax chico, A. puka Mirande, Aguilera & Azpelicueta, and A. troya. This character, along with the preceding one can be considered as an additive character with three states; therefore the absence of gill rakers is coded in both characters. In Jupiaba mucronata and Moenkhausia xinguensis the denticles are largely restricted to margins of the gill-rakers, but some isolated denticles occur on their lateral surface and these species are coded as polymorphic. 203. Rows of gill rakers on first epibranchial: (0) one; (1) two. (LC38i). In most examined species each of the four epibranchials has two rows of gill rakers (state 1). A single row of gill rakers on the first epibranchial (state 0) was observed in this study only in Acestrorhynchus pantaneiro, Agoniates anchovia, and Salminus brasiliensis. Ceratobranchials: 204. Shape of dentigerous plate of fifth ceratobranchial: (0) rounded, with posterior notch; (1) elongated, without posterior notch. (VH33i, MO58i). The fifth ceratobranchial bears a dentigerous plate in its posterior margin. This plate in most examined species is rather elongate, with obtuse angles with the main body of the fifth ceratobranchial both in their anterior and posterior regions (state 1; Vari & Harold, 2001: fig. 13A). In a small number of species the dentigerous plate has a rounded shape and its posterior region forms a straight or acute angle with the main body of the fifth ceratobranchial (state 0; Vari & Harold, 2001: fig. 13B). Uj (1990) mentioned the rounded shape of the dentigerous plate of the fifth ceratobranchial as a synapomorphy of his Piabucidae (= Iguanodectinae). Uj cited also this character state for the genus Creagrutus. 205. Teeth on fifth ceratobranchial: (0) present; (1) absent. Most species in the Characiformes have teeth on the fifth ceratobranchial (state 0; Vari & Harold, 2001: fig. 13). The absence of such teeth (state 1) was proposed as a synapomorphy of the Curimatidae plus Prochilodontidae by Vari (1983) and it was observed in this study only in Cyphocharax spilotus and Prochilodus lineatus. Pharyngobranchials: 206. Teeth on third pharyngobranchial: (0) present; (1) absent. (LC37, VH34, MO57, SE87m, BÜ84). The third pharyngobranchial bears teeth in most members of the Characidae (state 0) and these teeth are absent in a relatively small group of examined species (state 1). Contrary to the reported by Moreira (2002), teeth on the third pharyngobranchial are present in the examined specimens of Brycon pesu and Pseudocorynopoma doriae, which are coded as polymorphic. Iguanodectes geisleri is also coded as polymorphic following the observations of Moreira (2002) for this species. Phylogeny of the family Characidae 438 207. Teeth on fourth pharyngobranchial: (0) present; (1) absent. (BU41). The fourth pharyngobranchial bears teeth in most species of the Characiformes (state 0). The absence of teeth in this bone (state 1) was proposed as a synapomorphy of the Anostomoidea (Anostomidae, Chilodontidae, Curimatidae, and Prochilodontidae) by Vari (1983). The ancestral condition for the Characiformes is unclear. The Cypriniformes lack dentigerous plates on the fourth and fifth pharyngobranchials (Fink & Fink, 1981, 1996) while Diplomystes, considered a basal siluriform, bears teeth on the third and fourth pharyngobranchials (Azpelicueta, 1994). This character is coded as missing for the root of this analysis. 208. Teeth on fifth pharyngobranchial: (0) present; (1) absent. The fifth pharyngobranchial bears teeth in most taxa in the Characiformes (state 0). The absence or reduction of such teeth (state 1) is a synapomorphy of a clade composed of the Curimatidae and Prochilodontidae according to Vari (1983). For the same reasons as in character 207, the root is coded as missing. 209. Contact between fourth and fifth pharyngobranchial dentigerous plates: (0) absent; (1) present. (UJ81, BU42i). In most characiforms the dentigerous plates of the fourth and fifth pharyngobranchials contact each other and form a continuous surface (state 0; Vari, 1983: fig. 13). The absence of contact between these pharyngobranchials (state 1; Vari, 1983: figs. 16-19) was considered by Vari (1983) as a synapomorphy of the Anostomoidea. Interhyal: 210. Interhyal: (0) present; (1) absent. The interhyal joins the suspensorium with the hyoid arch in almost all the examined species. This bone contacts dorsally with the symplectic and hyomandibula by means of cartilages; the interhyal contacts ventrally with the posterior ceratohyal (state 0). The interhyal is absent (state 1) in Aulixidens eugeniae and Engraulisoma taeniatum among the examined species. 211. Length of interhyal: (0) shorter than one-third of symplectic length; (1) equal to or longer than one-half of symplectic length. The interhyal is a cylindrical bone much shorter than the symplectic in most examined species (state 0; Fig. 72). A configuration of the jaws and suspensorium including a shortening of the horizontal arm of the preopercle, the lengthening of the dentary, and a dorsal displacement of the posterior end of the symplectic was observed in some species. The dorsal displacement of the symplectic in this arrangement is correlated with an elongation of the interhyal to a relatively longer form (state 1). Although this character could be associated with predation, some apparently predatory genera such as Bramocharax and Oligosarcus have state 0. Branchiostegal rays: 212. Number of branchiostegal rays: (0) three; (1) four or five. (VA61m, LU70, LM33m, TP33m, ZV112i, LI39). 213. Number of branchiostegal rays: (0) three or four; (1) five. (VA61m, LU70, LM33m, TP33m, ZV112i, LI39, PZ57). Most characiforms have four branchiostegal rays (character 212, state 1; Character 213, state 0). Vari (1995) mentioned the presence of five branchiostegal rays (character 213, state 1) in Erythrinus Scopoli, Hoplerythrinus Gill and Hoplias and considered this state as a synapomorphy of the Erythrinidae. Lucena & Menezes (1998) considered the possession of five branchiostegal rays as a synapomorphy of the Cynodontinae. Variation between four and five branchiostegal rays was observed in Astyanax cf. asuncionensis, Cyanocharax alburnus, Knodus breviceps, and Nantis cf. indefessus with these species coded as polymorphic for character 213. The Cypriniformes has three branchiostegal rays, but both the Gymnotiformes and Siluriformes have more than five and the ancestral state for the Characiformes is unknown. Therefore, the root of this analysis is coded as missing. 214. Anterior portions of branchiostegal rays: (0) broad near their articulation with ceratohyals; (1) slender near their articulation with ceratohyals. (BE56). In most examined species the branchiostegal rays are anteriorly expanded near their articulation with the ceratohyals (state 0; Fig. 80). In some characids, especially in members of the Characinae, the branchiostegal rays are slender along their entire length and lack the cited expansion (state 1; Fig. 82). 215. Attachment of first branchiostegal ray: (0) on proximal one-half length of anterior ceratohyal or anterior to that; (1) posterior to one-half length of anterior ceratohyal. Brycon meeki has three branchiostegal rays attached to the ventral margin of the anterior ceratohyal (Weitzman, 1962). The first branchiostegal ray articulates to some point on the proximal one-half of the anterior ceratohyal, usually along its proximal one-third (state 0; Fig. 80). In the examined specimens of Triportheus spp. the attachment site of the branchiostegal rays is displaced posteriorly to the posterior half of the ceratohyal (state 1). 216. Distance between attachment site of first and second branchiostegal rays: (0) equal or shorter than distance between second and third rays; (1) longer than distance between second and third rays. The first and second branchiostegal rays are usually closer to each other than are the second and third rays (state 0; Fig. 82). In some species, however, the first and second branchiostegal rays are relatively more distant from each other, being separated by a distance equal to or greater than the distance between the second and third rays (state 1; Fig. 83). J. M. Mirande 439 217. Number of branchiostegal rays attached to posterior ceratohyal: (0) one; (1) two. (VA60, LU69i, LM34, TP34, PZ58). Most examined species share the possession of three branchiostegal rays articulated with the anterior ceratohyal, plus one attached to the posterior ceratohyal as described by Weitzman (1962) (state 0; Fig. 82). Vari (1995) mentioned the presence of two branchiostegal rays articulating with the posterior ceratohyal (state 1) in Ctenolucius spp, while Lucena (1993) also coded this state in Acestrorhynchus pantaneiro and Rhaphiodon vulpinus and considered the presence of two branchiostegal rays articulating with the posterior ceratohyal as a synapomorphy of a monophyletic clade composed of Acestrorhynchus, Cynodon, and Rhaphiodon, with a parallel occurrence in Ctenolucius. Lucena & Menezes (1998) proposed the inclusion of Gilbertolus and Roestes (subfamily Roestinae) in their family Cynodontidae, as sister group of Cynodon and Rhaphiodon. Since Gilbertolus and Roestes have state 0, the optimization of this character is ambiguous in the analysis of Lucena & Menezes (1998) for the clade containing the Acestrorhynchidae and Cynodontidae. This character could be optimized both as a synapomorphy of the Acestrorhynchidae and Cynodontidae, with a reversion in the Roestinae, or as parallelisms in the Acestrorhynchidae and Cynodontinae. Vertebrae, ribs and Weberian apparatus Weberian apparatus: 218. Form and articulation of neural pedicle of third vertebra: (0) pedicle well developed and articulating synchondrally with neural complex; (1) pedicle much smaller and without an articular surface with neural complex. In most ingroup species the neural pedicle of the third vertebra is smaller relative to the condition described by Weitzman (1962) for Brycon meeki. In that species the posterior margin of the neural pedicle articulates with the neural arch of the fourth vertebra, and its dorsal margin articulates with the neural complex (state 0; Fig. 7; Weitzman, 1962: fig. 12). This situation was observed herein in most members of the outgroup and some representatives of the Characidae. In the remaining species the neural pedicle of the third vertebra is reduced to a furca extending dorsally to the neural complex but not articulated synchondrally with that bone. Consequently the ventral margin of the neural complex is articulated solely with the neural arch of the fourth vertebra (state 1; Fig. 8; Fink & Fink, 1981: fig. 16). In Puntius tetrazona and Opsariichthys (Fink & Fink, 1981), the neural arch of the third vertebra is well developed and broadly articulates with the anterior supraneurals. This situation is coded herein as state 0. In the gymnotiform Sternopygus Müller & Troschel, the situation is similar to that of Xenocharax and Brycon meeki (Fink & Fink, 1981; Weitzman, 1962), which constitutes additional evidence to consider state 0 as ancestral for the Characiformes. This character is coded as polymorphic in Galeocharax humeralis which demonstrates an intermediate state. Apparently, the neural arches of the third and fourth vertebrae of Engraulisoma taeniatum are fused each other, and this character is coded as inapplicable. 219. Development of transverse process of neural arch of third vertebra: (0) not reaching anterior margin of tripus; (1) well developed and extending beyond anterior margin of tripus. (VA71m, ZV114m). The transverse process of third neural arch projects anteriorly or anterodorsally from the main body of the bone and receives the posterior tip of the intercalarium. Uj (1990) considered the great development of this process as diagnostic of his Agoniatidae (=Agoniatinae). Vari (1995) mentioned that this process is relatively reduced in the Ctenoluciidae and not reaching the posterior end of the intercalarium. In most species this process does not reach the anterior end of the tripus (state 0; Fig. 6) while in others it extends beyond the tripus anteriorly (state 1; Fig. 7). Intermediate states or intraspecific variability were observed in Bryconexodon juruenae, Diapoma speculiferum, Hoplocharax goethei, Lonchogenys ilisha, Moenkhausia xinguensis, and Odontostoechus lethostigmus which are coded as polymorphic. 220. Ascending process of neural pedicle of third vertebra: (0) absent; (1) present. (UJ9m, LU76m). In most examined species, the neural pedicle of the third vertebra has a variably developed ascending process directed towards the neural complex (state 1; Fig. 8). In several outgroups this process is completely absent (state 0). A reduced dorsal expansion of the neural pedicle was observed in Characidium rachovii and Distichodus maculatus, and these species are coded as polymorphic. 221. Dorsal development of dorsal process of neural pedicle of third vertebra: (0) not broadly overlapping neural complex; (1) broadly overlapping neural complex. (UJ37m, BU45m, LU76m). The dorsal process of the neural pedicle of the third vertebra, when present, is directed dorsally towards the neural complex. In most species this process falls short of the ventral margin of the neural complex, or only overlaps the margin slightly (state 0; Fig. 8), whereas in some species this process extensively overlaps the neural complex (state 1). Intermediate states were observed in Astyanax cf. rutilus, Carlana eigenmanni, Creagrutus anary, Hollandichthys multifasciatus, Hyphessobrycon luetkenii, Moenkhausia dichroura, and Oligosarcus sp. These species are coded as polymorphic. The cases in which this process is completely absent are coded as inapplicable. 222. Neural arch and vertebral centrum of fourth vertebra: (0) not fused and with autogenous fourth neural arch; (1) fused. (BU49). Phylogeny of the family Characidae 440 Fink & Fink (1981) mentioned that the neural arch of the fourth vertebra is autogenous (state 0; Fink & Fink, 1981: figs. 14-15) in the Citharinidae and Distichodontidae and considered this state as plesiomorphic for the Characiformes, due to its presence in the Cypriniformes and Gymnotiformes. Among the examined species, the neural arch of the fourth vertebra is only autogenous in Puntius tetrazona and Distichodus maculatus, corroborating the observations of those authors. This neural arch is fused to the corresponding vertebral centrum in the remaining examined species (state 1; Weitzman, 1962: fig. 12). Ribs: 223. Anteriorly directed spine at base of first rib: (0) absent; (1) present. (LU81m). In most examined species, the base of the first rib posterior to the Weberian apparatus bears an expansion onto which a ligament directed towards the following rib and parapophysis attaches. This medially directed expansion is approximately triangular (state 0). In some species there is, additionally, a well-developed spine projected anteriorly from that expansion (state 1; Fig. 90). Zanata & Vari (2005) described a process in Acestrorhynchus microlepis (Jardine), Brycon spp., and Chalceus spp., which is medially directed and projects from the main body of the rib. That feature is non-homologous with the process described in the state 1 of this character. 224. Laminar bony ridge on dorsal margin of abdominal ribs: (0) absent; (1) present. (BE59). The dorsal margin of the abdominal ribs lacks conspicuous projections in most examined species (state 0). Lucena (1993) mentioned the presence of dorsal bony ridges in the third to sixth abdominal ribs (state 1; Fig. 91) as parallel autapomorphies of Moenkhausia lepidura (Kner) and Parecbasis cyclolepis. Benine (2004) proposed the presence of bony ridges on abdominal ribs as a synapomorphy of a clade within Moenkhausia including M. lepidura. 225. Abdominal ribs on anterior caudal vertebrae: (0) absent; (1) present, associated to first and occasionally second caudal vertebrae. In most species, the ribs are articulated to the abdominal and transitional vertebrae both of which lack haemal spines (state 0). In the examined species of the Serrasalmidae, the posterior ribs are associated with the first, or first and second, caudal vertebrae (state 1). Vertebrae: 226. Relative number of precaudal vertebrae: (0) exceeding caudal vertebrae in two or more elements; (1) equal or less numerous than caudal vertebrae. (LU80m, MO87m). The precaudal vertebrae include the abdominal and transitional vertebrae. In most examined species of the Characidae the caudal vertebrae are as numerous or less numerous than the precaudal vertebrae (state 1). In some members of the outgroup, in contrast, the precaudal vertebrae are more numerous than the caudal vertebrae (state 0). 227. Total number of vertebrae: (0) 40 or fewer; (1) 41 or more. The number of vertebrae is almost continuously variable across the examined species. The limits of the states used herein are, therefore, rather subjective. Most examined species have a vertebral number comprised of between 35 and 38 vertebrae (state 0), while in some species a higher number was observed, exceeding 40 vertebrae (state 1). This character is coded as polymorphic for Distichodus maculatus, Galeocharax humeralis, and Hoplias cf. malabaricus which have ranges of vertebral counts overlapping the defined states. The examined specimen of Puntius tetrazona has 29 vertebrae, but the root of this analysis is coded as polymorphic because Opsariichthys, considered to be a generalized cypriniform by Fink & Fink (1981) has more than 40 vertebrae (Howes, 1978). 228. Total number of transitional vertebrae: (0) four or more; (1) three or fewer. (UJ47m, ZV121m). Haemal processes are ventral projections of the vertebrae that form part of the haemal arch and the haemal spine of the caudal vertebrae. Transitional vertebrae have bilateral haemal processes but lack a haemal spine (Weitzman, 1962: fig. 14b). Most examined species have fewer than four transitional vertebrae, most often two (state 1), while some species have four or more (state 0). Variation between these states was observed in Acrobrycon tarijae and Iguanodectes geisleri, which are coded as polymorphic. 229. Transitional vertebrae with haemal canal: (0) present; (1) absent. Additionally to the haemal processes, the posteriormost transitional vertebrae usually bear a haemal canal which is also present in the caudal vertebrae. This canal is formed by a transverse bony bridge between the contralateral haemal processes (state 0; Weitzman, 1962: figs. 14b and 14c). In some species this type of Fig. 90. Posterior portion of neurocranium and first vertebrae of Triportheus nematurus, CI-FML 3948, 82.1 mm SL, ventral view, anterior to left. RB1: first rib posterior to Weberian apparatus. Scale bar = 1 mm. J. M. Mirande 441 transitional vertebrae is absent, and the first vertebra that bears a haemal canal also has a haemal spine, being therefore the first caudal vertebra (state 1). This character is coded as polymorphic in Aphyocharacidium bolivianum, Astyanax latens, Bryconamericus cf. rubropictus, Carlana eigenmanni, Coptobrycon bilineatus, Hyphessobrycon pulchripinnis, and Psellogrammus kennedyi, which show variation between the defined states. Pectoral girdle Pectoral fins: 230. Margin of first pectoral ray in adult specimens: (0) not serrated; (1) conspicuously serrated. (VB41). The external margin of the first pectoral and pelvic fins is formed by a cartilage situated between the hemiradii that constitute such ray. This margin is usually not serrated (state 0). Bertaco (2003) reported the presence of a serrated margin of this cartilage in Hollandichthys and Pseudochalceus as a synapomorphy for a clade composed of those genera (state 1). This condition was also observed in this study in Apareiodon affinis, Parodon nasus, and Characidium spp. 231. Base of second pectoral ray: (0) large and partially overlapping base of first pectoral ray from medial view; (1) similar in form and size to base of posterior rays. (ZV137i). The base of each pectoral-fin ray has an expansion where several branches of the adductor radialis muscle attach (Winterbottom, 1974). In most examined species the base of the first pectoral-fin ray, usually the single unbranched ray, is much expanded in comparison to the bases of the posterior rays, all of which are of similar size (state 1). In a relatively small group of species, the base of the second pectoral-fin ray is much expanded and conspicuously laterally overlaps the base of the first fin ray (state 0; Fig. 92). Zanata & Vari (2005) mentioned that state 0 is broadly distributed in the Alestidae, being also present in several other characiforms. Zanata & Vari (2005) considered the state present in Hoplias microlepis as inapplicable, because the second pectoral-fin base is not conspicuously larger that those of the posterior rays, and it is oriented anteromedially rather than dorsally. A similar condition was observed in this study in Hoplias cf. malabaricus and Pyrrhulina australis in which this character is coded as inapplicable. Cleithrum: 232. Anterior margin of cleithrum: (0) slightly sinuous; (1) with anterior pointed projection. (UJ64m, LU88, LC42m, PZ77m). Lucena (1987) described a pointed projection of the anterior margin of the cleithrum lateral lamella (state 1; Fig. 93) of Charax. This projection extends laterally and posteriorly the surface where the sternohyoideus muscle attaches to the cleithrum. Uj (1990) mentioned the presence of such projection in his Cynopotamidae (genera Cynopotamus and Galeocharax Fowler) and Characidae (most of the Characinae). According to Lucena (1998), such a process is absent (state 0; Fig. 94) in, among others, Cynopotamus argenteus. The anterior process of the cleithrum was, however, observed in this species in this study and C. argenteus is coded as polymorphic for this character. 233. Form of posterior margin of cleithrum: (0) convex or slightly sinuous just dorsal to pectoral-fin insertion; (1) with notch just anterior to pectoral-fin insertion. (UJ64m, BU57, LU88, LC43, MO65, PZ78). The lateral lamella of the cleithrum is situated just dorsal to the pectoral-fin insertion and usually has a slightly sinusoidal form that partially follows the anterior margin of this fin (state 0; Fig. 94). Lucena (1987) described the Fig. 91. Detail of axial skeleton of Parecbasis cyclolepis, MHNG 2228, 60.0 mm SL, lateral view, anterior to left. PRB: pleural ribs. Scale bar = 1 mm. Fig. 92. Detail of pectoral fin and radials of Piabucus melanostomus, CI-FML 3894, 67.0 mm SL, dorsolateral view, anterior to left. Scale bar = 1 mm. Phylogeny of the family Characidae 448 also in Clupeacharax anchoveoides and Engraulisoma taeniatum; proposing it as a putative synapomorphy of a clade composed of these two taxa. Castro considered the state of these genera as different to that of Agoniates, in which such processes are articulated by way of a cartilage. The condition in Agoniates, Clupeacharax, and Engraulisoma are herein all coded as state 1. 262. Anterior extension of pelvic-bone along main axis: (0) not projecting anterior of lateral and medial lamellae; (1) projecting anterior of lateral and medial lamellae of pelvic bone. (CM26, ZV140, PZ86i). The medial margin of the pelvic bone is formed by a cylindrical process, whose anterior tip terminates in a small cartilage in almost all the examined species. The anterior portion of the pelvic bone along its primary axis is situated near the anterior confluence of the lateral and medial pelvic bony lamellae. In these cases the primary axis of the pelvic bone does not extend anteriorly (state 0). Malabarba (1998b) considered the anterior projection of the primary axis of the pelvic bone beyond the lateral and medial lamellae (state 1; Malabarba, 1998b: fig. 12) as a synapomorphy for a clade composed of Lignobrycon and Triportheus. In Jupiaba spp., the anterior portion of the pelvic bone is modified as an acute point, and this character is coded as inapplicable. This character is also coded as inapplicable for the species in which the pelvic bone is anteriorly bifurcate. 263. Anterior tip of pelvic bone: (0) rounded and capped by a small cartilage; (1) pointed, lacking associated cartilage and frequently projecting outside body wall. The anterior end of the pelvic-bone primary axis is rounded and capped by a small cartilage in most species in the Characiformes (state 0). In the species of Jupiaba this bone is very acute anteriorly and lacks this cartilage (Zanata, 1997) (state 1; Zanata, 1997: fig. 4A). In some species of this genus, the anterior tip of the pelvic bone pierces the body wall and projects outside as an externally visible spine. State 1 was observed only in the examined species of Jupiaba. 264. Dorsal longitudinal ridge on medial lamella of pelvic bone: (0) present; (1) absent. The principal axis of the pelvic bone is bordered medially by a variably developed lamella. This lamella is usually uniformly thin (state 1; Fig. 103), whereas in some species the medial lamella has instead a dorsal longitudinal ridge that forms a secondary axis which diverges anteriorly from the primary axis (state 0; Fig. 104); this character is coded as inapplicable to species with an anteriorly bifurcate pelvic bone. Epineurals, supraneurals, and dorsal fin skeleton Dorsal fin: 265. Relative position of dorsal-fin anterior insertion: (0) anterior to or at vertical through pelvic-fin origin; (1) posterior to vertical through pelvic-fin origin. (CM29m). In most examined characids, the dorsal fin is situated posterior to the vertical through the pelvic-fin origin (state 0). In some species, mostly in the outgroup, the dorsal-fin origin is situated at the vertical through the pelvic-fin insertion or slightly anterior of that line (state 1). This character varies among the examined specimens of Bryconops affinis and the A. eugeniae is coded as polymorphic. The cyprinids Opsariichthys spp., Puntius tetrazona, and Zacco spp. and, usually, the siluriforms have also the state 0. The root of this analysis was consequently coded as this state, although many species in the Cypriniformes have state 1 (Pflieger, 1997). 266. Dorsal-fin rays articulating with first dorsal pterygiophore: (0) two; (1) three or four. (LU98i, SE75, ZV124i). The unbranched dorsal-fin rays articulate with the first proximal dorsal-fin pterygiophore with the number of rays articulating with that pterygiophore variable among the examined species. Lucena (1993) coded his functional outgroup and several species as having three or more rays articulating with this pterygiophore (state 1). Zanata & Vari (2005) mentioned the presence of only two rays associated with the first dorsal pterygiophore (state 0) in the members of the Alestidae and considered the possession of three rays as a common condition among the Neotropical characiforms. The presence of only two unbranched and eight branched dorsalfin rays (ii+8) was considered by Malabarba & Weitzman (2003) as a diagnostic character of their clade A. This character is variable among the examined specimens of Aphyodite grammica, Astyanax paris, Brycon pesu, Carlana eigenmanni, Hollandichthys multifasciatus, Hyphessobrycon herbertaxelrodi, Inpaichthys kerri, Markiana nigripinnis, Moenkhausia cf. intermedia, Oligosarcus cf. jenynsii, Parecbasis cyclolepis, and Roeboides descalvadensis, which are coded as polymorphic. Fig. 103. Pelvic girdle and fins of a male Bryconamericus cf. iheringii, CI-FML 3898, 50.2 mm SL, ventral view, anterior to left. Scale bar = 1 mm. J. M. Mirande 449 267. Anteriorly oriented spine formed by first dorsal-fin ray: (0) absent; (1) present. Among the species having three unbranched dorsal-fin rays, the first ray is usually similar to the posterior ones, but smaller (state 0). The Stethaprioninae was diagnosed by Géry (1964a) as having the first dorsal-fin ray oriented anteriorly to form a spine (state 1; Fig. 105); Reis (1989) considered this state to be a synapomorphy of this subfamily. Géry (1977: 367) illustrated an anterior projection of the first dorsal-fin ray of Prochilodus; however, as noted by this author, this projection have paired anterior processes, rather than a medial spine as in the Stethaprioninae of Reis (1989) and the situation in Prochilodus is considered non-homologous to that described in the state 1. 268. Anterior rays of dorsal fin of adult males: (0) not elongate; (1) elongate and reaching posteriorly to position close to adipose fin. (VB36, CA28). The dorsal fin in most examined species is not sexually dimorphic, or when elongate it does not demonstrate the overall shape in males described for state 1. Cardoso (2003b) reported the posterior elongation of the first dorsal-fin rays (state 1) in adult males of Carlana eigenmanni, Nematocharax venustus, Parastremma pulchrum Dahl, Pseudochalceus spp., Pseudocorynopoma doriae, and Rhoadsia spp. In this condition, the last unbranched and first branched dorsal-fin rays reach posteriorly almost to the adipose fin. In Odontostilbe spp. only the last unbranched dorsal-fin ray is extended as a filament. This situation is herein considered as non-homologous to state 1 but rather treated in the following character. 269. Last unbranched dorsal-fin ray of adult males: (0) approximately as long as first branched ray; (1) distinctly longer than first branched ray and in the form of filament. (MA11, BÜ102). In most examined species, the last unbranched and the first branched dorsal-fin rays are the longest rays with the dorsal-fin rays gradually decreasing in length posteriorly (state 0). Malabarba (1998a) noted that the posteriormost unbranched dorsal-fin ray is much longer than the remaining rays and extended as a filament (state 1; Bührnheim & Malabarba, 2007: fig. 23) in adult males of Odontostilbe fugitiva and Holoshesthes (=Odontostilbe) pequira. According to Malabarba (1998a), the elongation of this ray is a synapomorphy for the species of Odontostilbe. 270. Number of branched-rays on dorsal-fin: (0) eight or fewer; (1) nine or more. (AM2m, VB38m, BE64m, QU38m). The number of branched dorsal-fin rays is rather uniform across the Characiformes with most examined species having nine rays (state 1). Malabarba & Weitzman (2003) partially diagnosed their clade A by the shared possession of only eight branched dorsal-fin rays (state 0) in almost all the members of that clade, with this state also present is Clupeacharax and Engraulisoma and variably in Paracheirodon, Piabucus, Serrabrycon Vari, and Tyttobrycon Géry. Malabarba & Weitzman (2003) mentioned the presence of nine branched dorsal-fin rays in an examined paratype of Aulixidens eugeniae; however the four examined specimens of this species have only eight rays and the species is coded here as polymorphic. 271. Relative length of anterior dorsal-fin rays: (0) not reaching tip of posterior rays when adpressed; (1) reaching tip of posterior rays when adpressed. In Puntius tetrazona, as usual in the Cypriniformes, the length of the posterior dorsal-fin rays is similar to that of the anterior rays, consequently reaching more posteriorly in the adpressed fin (state 1). In most characid species, the anterior rays are proportionally longer and extend posteriorly to the tip of posterior rays, when the fin is adpressed (state 1). Fig. 104. Pelvic girdle and fins of a male Characidium borellii, CI-FML 3865, 50.9 mm SL, ventral view, anterior to left. Scale bar = 1 mm. Fig. 105. Detail of anterior dorsal-fin pterygiophores and rays of Stethaprion erythrops, MHNG 2187.33, 44.5 mm SL, lateral view, anterior to left. Scale bar = 1 mm. Phylogeny of the family Characidae 450 272. Number of dorsal-fin rays on last pterygiophore: (0) one; (1) two, adnate. (BU64i, LU100i, MO78i, ZV126i). Most examined characids have only one ray in the last dorsalfin pterygiophore (state 0), while the last two dorsal-fin rays articulate with the last dorsal-fin pterygiophore (state 1) in several species of the outgroup. According to the phylogenetic hypothesis of Buckup (1998), the presence of only one ray articulating with the last pterygiophore is a synapomorphy of a clade composed of the Acestrorhynchidae, Characidae, Ctenoluciidae, Erythrinidae, Hepsetidae, and Lebiasinidae. Under the phylogeny of Lucena (1993), the presence of two rays on the last dorsal-fin pterygiophore is a synapomorphy of the Cynodontidae, that independently occurs in a clade composed of Alestes, Brycon, Chalceus, Rhabdalestes, Hydrocynus, and Serrasalmus Lacépède, with a reversal in Brycon. Intermuscular bones: 273. Dorsal myorhabdoi: (0) absent; (1) present. (TP67). The dorsal myorhabdoi are slender intermuscular bones situated dorsal to the epineurals (Chapman, 1944; Weitzman, 1954). These bones are present (state 1; Weitzman, 1954: fig. 2) in the Gasteropelecidae (Weitzman, 1954), the species of Citharinus Cuvier, and Rhaphiodon vulpinus (Toledo-Piza, 2000), but absent (state 0) in most members of the Characiformes. 274. Position of anteriormost epineurals: (0) lateral to fourth or fifth vertebrae; (1) reaching to cranium. (LM38, MO85m, ZV123m, PZ74). In most examined species the anteriormost epineurals are situated posterior to the fourth or fifth vertebrae (state 0). In some taxa including species of Carnegiella Eigenmann, Thoracocharax Fowler (Weitzman, 1954), Clupeacharax (Castro, 1984), Acestrorhynchus, Agoniates, Chalceus, Cynodon, Gilbertolus, Lignobrycon, Rhaphiodon, Triportheus (Lucena & Menezes, 1998), Brycinus, Hydrocynus, and Micralestes Boulenger (Zanata & Vari, 2005), the anterior epineurals are situated along the posterior margin of the cranium, and in some of these species they originate in the posttemporal fossa (state 1; Fig. 5). Dorsal pterygiophores: 275. Predorsal spine formed by first dorsal pterygiophore: (0) absent; (1) present. (AM21). The presence of a predorsal spine formed by the first dorsal-fin pterygiophore was proposed by Machado-Allison (1983) as a synapomorphy of the Serrasalmidae (state 1). This spine is absent (state 0) in some Serrasalmidae and all non-serrasalmid Characiformes. 276. Number of dorsal pterygiophores: (0) nine; (1) 10 or more. (LU97m, VB38m, ZV125m). 277. Number of dorsal pterygiophores: (0) 10 or less; (1) 11 or more. (LU97m, VB38m, ZV125m). 278. Number of dorsal pterygiophores: (0) 11 or less; (1) 12 or more. (LU97m, VB38m, ZV125m). Most examined species have 10 dorsal-fin pterygiophores and nine branched dorsal-fin rays (character 276, state 1). Lucena (1993) mentioned the presence of 11 or more dorsal pterygiophores (character 277, state 1) in Chalceus sp., Cynodon gibbus, Hoplias malabaricus, Rhaphiodon vulpinus, and Serrasalmus (=Pristobrycon) striolatus (Steindachner). Zanata & Vari (2005) mentioned the presence of 18 pterygiophores in Crenuchus Günther, 13 in Hoplias, 14 in Piaractus Eigenmann, 16 in Serrasalmus, and 18 in Xenocharax (character 278, state 1). Variations between 10 and 11 pterygiophores were herein observed in Hyphessobrycon herbertaxelrodi and Inpaichthys kerri, which are coded as polymorphic for character 277. Supraneurals: 279. Supraneural anterior to neural spine of fourth vertebra: (0) absent or small; (1) present and vertically elongate. (BU47, LU96, MO74, SE98i, BE63i, ZV122, LI45, QU90, PZ76). The supraneurals are situated between the neural spines anterior to the dorsal-fin pterygiophores. In most examined species the anteriormost supraneural is situated between the neural spines of the fourth and fifth vertebrae (state 1), whereas in some species there is a supraneural situated more anteriorly, anterior to the neural spine of the fourth vertebra (state 0; Fig. 7). According to Buckup (1998) the loss of this supraneural would be a synapomorphy of a clade composed of the genera Oligosarcus, Tetragonopterus, Phenacogaster, Charax, and Cynopotamus. Under the phylogenetic hypothesis of Lucena (1993) this supraneural was independently lost in four clades and a synapomorphy for a clade including most of the Characidae except for Agoniates, Brycon, Bryconops, Hemibrycon, Roeboexodon, and Triportheus. Zanata & Vari (2005) observed that in the Alestidae the absence of this supraneural is rather correlated with events of miniaturization. This supraneural is present in the examined specimen of Acestrorhynchus pantaneiro, contrary to the observations of Lucena (1993). This species is coded as polymorphic in light of possible intraspecific variability. In Apareiodon affinis, Parodon nasus, and Prochilodus lineatus, there is a large supraneural bounding anteriorly, dorsally, and posteriorly the neural spine of the fourth vertebra. Since the homology of this supraneural with the one treated here is uncertain, this character is coded as inapplicable for these species. In the examined specimens of Hasemania nana, there is an additional supraneural situated just posterior to the neural spine of the fourth vertebra. Given that the homology of this additional supraneural is unclear, this character is also coded as inapplicable for H. nana. 280. Number of supraneurals: (0) four or fewer; (1) five or more. (MO76m, SE95m, BE61m). 281. Number of supraneurals: (0) seven or fewer; (1) eight or more. (MO76m, SE95m, LI46, QU78m). This character is related with the dorsal-fin position and the number of vertebrae; however, it is included as a separate feature J. M. Mirande 451 given its potential phylogenetic informativeness at different levels. This character is coded considering the intraspecific variations mentioned in the original species descriptions. The examined specimen of Puntius tetrazona has only four supraneurals, but Barilius, a genus related with Opsariichthys has 13 (Howes, 1978), covering the entire range of variation considered for this character between these species. Therefore, the root of this analysis is coded as polymorphic for both these characters. 282. Bony lamellae associated with supraneurals: (0) absent or small; (1) wider than primary axis of supraneurals. (MO75m, SE96m, BE62, LI44). The supraneurals usually have a more or less vertical, cylindrical, body, and variably developed anterior and posterior bony lamellae, which when present are wider dorsally. The degree of development of these lamellae is difficult to define as discrete states. The herein recognized states are lamellae that are absent or are narrower than the main cylindrical body of the supraneurals (state 0), and that in which the lamellae are wider than the body (state 1). This character is coded as polymorphic in the species in which these lamellae are variably present. 283. Position of last supraneural: (0) located two or fewer vertebrae in front of first dorsal pterygiophore; (1) located more than two vertebrae in front of first dorsal pterygiophore. The posteriormost supraneural in most examined species is situated between the neural spines of the vertebrae immediately anterior to the first dorsal-fin pterygiophore (state 0). In a few species the posteriormost supraneural is distant three or more neural spines from the first dorsal pterygiophore (state 1). Anal fin and pterygiophores Anal fin: 284. Anal-fin position: (0) posterior or almost posterior to vertical through last dorsal-fin ray. (1) extended anteriorly ventral to dorsal fin. (EI16, MO77). The anal-fin origin, in most examined species, is situated at or posterior to the vertical through the base of the posteriormost dorsal-fin ray (state 0), whereas in some species the anal fin extends anteriorly below the middle of the dorsal fin (state 1). 285. Number of unbranched anal-fin rays: (0) three or fewer; (1) four or more (ZV142m, LI51m). Most examined species have four to six unbranched analfin rays (state 1), with a high intraspecific variation within this range. Zanata & Vari (2005) mentioned that most alestids have only three unbranched anal-fin rays (state 0), sharing this state with Crenuchus spilurus, Hepsetus odoe, and Hoplias microlepis, among the species examined by them, with most characids having four or five of such rays (state 1; Fig. 106). This character is herein coded as polymorphic in Aphyocharacidium bolivianum, in which a variation between three or four rays was observed. 286. Number of branched anal-fin rays: (0) 10 or less; (1) 11 or more. (UJ76m, BU65m, LU103m, SE92m, BE78m, LI52m). 287. Number of branched anal-fin rays: (0) 17 or less; (1) 18 or more. (LU103m, WM35i, MO80m, SE92m, BE78m, LI52m). 288. Number of branched anal-fin rays: (0) 24 or less; (1) 25 or more. (CM30m, MO80m, SE92m, BE78m). 289. Number of branched anal-fin rays: (0) 34 or less; (1) 35 or more. (BU66m, LU104m, CM30m, MO80m, SE92m, BE78m, QU61m). The number of anal-fin rays is used in species-level systematic studies of many genera of the Characidae. However, this number is highly variable within some evidently monophyletic clades and its phylogenetic utility may be mostly restricted to the resolution of rather small clades. Notably, however, supposedly basal Cypriniformes and Siluriformes and most non-characid Characiformes have a relatively low number of branched anal-fin rays (character 286, state 0), while most species of the Characidae have usually more than 15 rays (character 286, state 1). Buckup (1998) proposed the presence of 19 or more total anal-fin rays as a synapomorphy of a clade including the Alestidae, Characidae, Acestrorhynchidae, Erythrinidae, Lebiasinidae, Hepsetidae, and Ctenoluciidae, with a reversal in a clade formed by the four latter families. According to the hypothesis of Lucena (1993) the presence of 14 or more anal-fin rays is a synapomorphy of a clade composed of the Acestrorhynchidae, Alestidae, Characidae, Cynodontidae, and Serrasalmidae. This character involves only the branched analfin rays, instead of total rays, as in Lucena (1993) and Buckup (1998). This coding system has advantages in terms of homology assessment, avoiding the coding of dissimilar arrangements as the same character state (e. g. ii+15 rays vs. v+12 rays). Coding of these four characters includes ranges of intraspecific variation taken from the original descriptions and systematic and faunistic revisions (Eigenmann, 1912, 1915, 1917, 1918, 1921, 1927; Eigenmann & Myers, 1929; Géry, 1977; Ringuelet et al., Fig. 106. Anal fin and pterygiophores of an adult male of Axelrodia lindeae, MCP 37314, 22.5 mm SL, lateral view, anterior to left. Scale bar = 1 mm. Phylogeny of the family Characidae 452 1967; Reis, 1989; Vari & Harold, 2001; Zanata & Toledo-Piza, 2004). Most examined species have 18 to 35 branched anal-fin rays. The species having ranges of variation comprising more than one state were coded as polymorphic. 290. Form and length of anterior anal-fin rays: (0) similar to posterior rays; (1) longer and more compressed laterally than posterior rays. (BÜ109m, BÜ112m, BÜ113m). In most examined species the anterior anal-fin rays are rather similar in form to the posterior rays (state 0). Bührnheim (2006) mentioned that the adult males of Axelrodia lindeae, the species of Heterocheirodon Malabarba, the species of Serrapinnus Malabarba, and two species of Spintherobolus have a series of modifications in the anterior anal-fin rays of males. These involve the presence of a lobe formed by rays bearing bony hooks, with the rays much compressed laterally and whose proximal portions are joined by strong ligaments (state 1; Fig. 106). These three characters are here considered a single feature, because they are functionally correlated, and given that the presence or absence of a lobe and the degree of development of the ligaments between the anal-fin rays are difficult to define as discrete states. In Acestrorhynchus pantaneiro the anterior anal-fin rays are much elongated relative to the posterior rays, forming a pronounced lobe. This state is not, however, exclusive to the males and it is not considered homologous with state 1 of this character. An intermediate condition between the defined states was observed in Aphyocharacidium bolivianum, which is coded as polymorphic. 291. Number of rays on last anal pterygiophore: (0) two; (1) one. Most examined species have two rays articulating with the posteriormost anal-fin pterygiophore (state 0; Fig. 106). The examined specimens of Coptobrycon bilineatus and Hasemania nana instead have only one ray articulating with that pterygiophore (state 1). Anal-fin pterygiophores: 292. Anterior notch on first anal pterygiophore: (0) absent; (1) present. The cylindrical main body of the anteriormost anal pterygiophore, among the examined species, is limited anteriorly by a variably developed bony lamella. The anterior margin of this lamella lacks notches in most taxa (state 0). The examined members of the Parodontidae and Markiana nigripinnis instead have a marked notch (state 1) along the dorsal margin of the lamella. 293. Number of anal pterygiophores anterior to first haemal spine: (0) three or fewer; (1) four or more. (LU101, LC41). In most species the dorsal portion of the first proximal anal-fin pterygiophore is situated between the anteriormost haemal spines (state 0). Lucena (1998) proposed that the anterior displacement of the anal fin relative to the haemal spines, with at least four anal pterygiophores anterior to the first haemal spine (state 1) is a synapomorphy of a clade composed of Charax and Roeboides. Lucena coded Cynopotamus spp. as state 0, but the specimen of C. argenteus herein examined has seven pterygiophores anterior to the first haemal spine, and is coded as state 1. 294. Proximal and medial radials of anal fins: (0) fused on anterior five pterygiophores; (1) fused in most pterygiophores; (2) medial radials absent or completely fused with proximal ones. (BE82m, ZV190m). The anal pterygiophores are composed of three radials of which the proximal is the longest, extending dorsally between the haemal spines of caudal vertebrae. The medial and distal radials are much smaller and participate in the articulation of the pterygiophore with the corresponding anal-fin ray. In most characids the four or five anterior proximal anal pterygiophores are fused with the corresponding medial ones, while in the remaining pterygiophores the three radials are independent ossifications (state 0). In some species most pterygiophores, instead, have the proximal and medial pterygiophores fused and only a few posteriormost pterygiophores have three independent radials (state 1). In the examined specimens of Carnegiella strigata, Characidium spp., Hoplias cf. malabaricus, Metynnis maculatus, Piaractus mesopotamicus, Pyrrhulina australis, Serrasalmus maculatus, and Thoracocharax stellatus the medial radials are completely absent or fused with the proximal radials in all the pterygiophores (state 2). Zanata & Vari (2005) mentioned that in the Ctenoluciidae and Erythrinidae (as also herein observed in Characidium and Pyrrhulina) the form of the proximal radials suggests their fusion with the medial radials, while in the Serrasalmidae (as also herein observed in Carnegiella and Thoracocharax) the medial radials appear to be primarily absent. Both conditions are coded with the state 2, pending further studies. 295. Lateral lamellae on anterior anal pterygiophores: (0) absent; (1) present. (ZV144m). In most examined species, the anal pterygiophores have two associated bony lamellae positioned anteriorly and posteriorly to their main body (state 0). Some species of the outgroup have, in addition, two lateral lamellae giving the pterygiophores a cross-shaped transverse section (state 1). Caudal skeleton Epurals: 296. Number of epurals: (0) one; (1) two or three. (BU68, LU106i, VH55, MO88m, BE85i, PZ90m). 297. Number of epurals: (0) one or two; (1) three. (BU69, LU107, CM33, MO88m, ZV148, PZ90m). The number of epurals in the Characiformes varies from one to three, but most species have two epurals (character 296, state 1, character 297, state 0; Fig. 107). The presence of only one epural (character 296, state 0; Fig. 108) was mentioned for several species in the Characiformes (Lucena, 1993; Buckup, 1998; Moreira, 2002, Benine, 2004; Lima, 2006). Two epurals were observed herein in Acestrorhynchus pantaneiro, Gymnocorymbus ternetzi (Boulenger), Hemigrammus unilineatus, Hyphessobrycon eques, Parecbasis cyclolepis, J. M. Mirande 453 and Piabina argentea, differing with observations of Lucena (1993), Moreira (2002), Benine (2004), and Lima (2006). These species are coded as polymorphic for character 296, in light of possible intraspecific variability. Variation between one and two epurals was observed in Pristella maxillaris which is coded as polymorphic for this character. The presence of three epurals (character 297, state 1; Fig. 109) was mentioned for several mostly non-characid Characiformes (Buckup, 1998; Moreira, 2002; Zanata & Vari, 2005; Lima, 2006). Hypurals: 298. Fusion of hypural 2 to compound centrum: (0) absent; (1) present. (UJ28m). The fusion of the hypural 2 with the compound centrum (state 1; Fig. 108) was proposed by Fink & Fink (1981, 1996) as a synapomorphy of the Otophysi. Zanata & Vari (2005: 122) mentioned that this fusion is present in most Characiformes, being secondarily absent (state 0; Zanata & Vari, 2005: fig. 30) in a few taxa which usually have the hypurals 1 and 2 fused to each other. According to Roberts (1969: figs. 56-58 and 60), the hypural 2 is not fused to the compound centrum in Ctenolucius, Hepsetus, Hoplias, and Hydrocynus. Roberts (1974: figs. 18, 38, and 39) illustrated such a condition also in the hemiodontids Argonectes, Bivibranchia Eigenmann, and Hemiodus Müller. Miquelarena (1982) mentioned that in Rhaphiodon the hypurals 2 and 3 are fused and articulated, but not fused, with the compound centrum. Zanata & Vari (2005), in the discussion on the phylogenetic position of †Mahengecharax carrolli Murray, indicated that the hypural 2 is fused with the compound centrum in the Alestidae with the exception of the miniature species Lepidarchus adonis. In this species both hypurals are fused each other but separated from the compound centrum; however they (Zanata & Vari, 2005: fig. 30) illustrated an autogenous hypural 2 in Brycinus macrolepidotus, and Murray (2004: fig. 13) illustrated six autogenous hypurals in Alestes stuhlmannii (Pfeffer). The absence of fusion between the hypural 2 and the compound centrum is usually correlated with the fusion of the hypurals 1 and 2 (Zanata & Vari, 2005); however, some exceptions were herein observed, and these characters are coded as separate in this paper. Intraspecific variability was observed in Heterocharax macrolepis, which is coded as polymorphic. 299. Fusion between hypurals 1 and 2: (0) absent; (1) present. (UJ28m, BU71, LU109). As mentioned above, the fusion between the hypurals 1 and 2 (state 1) is often correlated with the lack of fusion between the hypural 2 and the compound centrum, but as this correlation is not perfect, these characters are coded as separate. 300. Posterior margin of hypural 3: (0) equal to or narrower than posterior margin of hypural 4; (1) deeper than posterior margin of hypural 4. In the Characiformes and especially in the Characidae the posterior margin of the hypural 3 is usually deeper than that of the hypural 4 (state 1; Fig. 107), while in a group of species of the outgroup the posterior margin of the hypural 3 is equal or more slender than the hypural 4 (state 0; Fig. 109). Procurrent rays: 301. Ventral procurrent caudal-fin rays of adult males: (0) slender; (1) laminar. (MA45m, BÜ127m, BÜ135). In the generalized condition described by Weitzman (1962) for Brycon meeki, the lepidotrichia that form each procurrent ray are autogenous or are fused only at their ventral portion. In these cases the procurrent rays are similar to the principal caudal-fin rays but shorter or, if corresponding lepidotrichia are fused each other, they are as slender as the principal caudal-fin rays (state 0; Fig. 107). Malabarba (1998a) proposed the fusion and anteroventral Fig. 107. Caudal skeleton of Carlana eigenmanni, LACMNH 9230.020, 45.2 mm SL, lateral view, anterior to left. EPU: epurals, HY3 4: hypurals 3 and 4. PRR: procurrent ray. Scale bar = 1 mm. Fig. 108. Caudal skeleton of Thoracocharax stellatus, CIFML 3869, 45.2 mm SL, lateral view, anterior to left. EPU: epural, HY2: hypural 2. Scale bar = 1 mm. Phylogeny of the family Characidae 454 expansion of the corresponding lepidotrichia of all the procurrent rays along their entire length, forming laminar structures (state 1; Fig. 110), as a synapomorphy of the tribe Cheirodontini. 302. Number of ventral procurrent caudal-fin rays: (0) 11 or fewer; (1) 12 or more. (MA42m, VB74m, BÜ124m). The number of ventral procurrent rays is 11 or fewer in most examined species (state 0). Malabarba (1998a) considered the presence of more than 11 ventral procurrent rays (state 1) as a synapomorphy of the Cheirodontini (see previous character). Miquelarena (1982) mentioned a higher number of rays in Roeboides bonariensis (Steindachner) (=R. microlepis), Acestrorhynchus altus Menezes (=A. pantaneiro), Moenkhausia dichroura, and Tetragonopterus argenteus. Up to 11 ventral procurrent rays were observed in Roeboides microlepis (11 in 1 ex.), Acestrorhynchus pantaneiro (11 in 1 ex.), Moenkhausia dichroura (11 in 2 ex.), and Tetragonopterus argenteus (8-9 in 2 ex.) and these species are thus coded as polymorphic for this character, in light of possible intraspecific variability. Variation between the defined states was observed in Acestrocephalus sardina, Aphyocharacidium bolivianum, Axelrodia lindeae, Bryconamericus thomasi, and Gymnocharacinus bergii, which are coded as polymorphic. 303. Ventral procurrent caudal-fin rays of adult males: (0) not projecting through musculature and skin of peduncle; (1) projecting ventrally through peduncle musculature and skin. (MA47, BÜ129). Most examined species have their anteriormost ventral procurrent rays contained completely within the skin and musculature of the caudal peduncle and not projecting ventrally as a keel (state 0). Malabarba (1998a) proposed as a synapomorphy of the Cheirodontini the ventral projection of such procurrent rays through the musculature and skin forming an externally visible keel (state 1; Fig. 110). 304. Caudal-fin bony stays: (0) absent; (1) present. (VA82, BU67, LU105, ZV150). The presence of caudal stays (sensu Roberts, 1969) was mentioned several times in some studies treating the Ctenoluciidae (Vari, 1995) and Alestidae (Murray & Stewart, 2002; Zanata & Vari, 2005), in which they are particularly developed (state 1; Zanata & Vari, 2005: fig. 30). Bony stays are absent in the examined characids (state 0). Stays are longitudinal median bones situated at the base of the procurrent rays. whose homology with the procurrent rays is uncertain. Therefore, the presence or absence of stays is herein treated as different from the following character. 305. Anterior ventral procurrent caudal-fin rays: (0) paired, only distally fused; (1) fused in laminar medial bones. Zanata & Vari (2005) mentioned that in the Characidae there are medial bones, similar to stays, but much anteriorly situated. These procurrent rays are situated in parallel; the posteriormost procurrent rays have separated lepidotrichia, which are progressively fused in the anterior ones. In this condition the anterior ventral procurrent rays are medial plates without remnants of separated lepidotrichia, articulated or situated between haemal spines of two or more vertebrae anterior to the compound centrum (state 1; Fig. 49). These plates have a different position than the caudal stays, which are longitudinal, perpendicular to the procurrent rays. In Puntius tetrazona and most representatives of the outgroup, the lepidotrichia of the anterior ventral procurrent rays are not fused to each other and do not reach the haemal spines (state 0). Uroneurals: 306. Uroneurals: (0) absent or just one pair; (1) two pairs. (UJ54, BU70, LU108, MO89, ZV149, LI50). The uroneurals are small paired bones aligned along the urostyle. Most examined taxa have only one pair of uroneurals Fig. 110. Caudal skeleton of an adult male of Axelrodia lindeae, MCP 37314, 22.5 mm SL, lateral view, anterior to left. PRR: procurrent ray. Scale bar = 0.5 mm. Fig. 109. Caudal skeleton of Chalceus macrolepidotus, MNHG 2189.13: 67.0 mm SL, lateral view, anterior to left. EPU: epurals, HY3-4: hypurals 3 and 4. Scale bar = 1 mm. J. M. Mirande 455 (state 0), whereas the presence of two pairs of uroneurals (state 1; Weitzman, 1962: fig. 15) was observed in several mostly noncharacid Characiformes. Among the examined species, the size of the second pair of uroneurals is variable when present, and in some cases they are much reduced. Nonetheless, all the species with a second pair of uroneurals present, regardless it size, are coded as state 1. In the examined specimens of Acestrocephalus sardina, Bryconamericus exodon, Carlana eigenmanni, Coptobrycon bilineatus, Moenkhausia cf. intermedia, M. sanctaefilomenae, Odontostoechus lethostigmus, and Roeboides descalvadensis, the second pair of uroneurals has a variable occurrence and these species are coded as polymorphic for this character. The complete lack of uroneurals was observed in Characidium rachovii, and this character is coded as inapplicable for that species. Bony hooks Distribution: 307. Bony hooks on fin rays: (0) absent; (1) present in adult males. (CM35m, CM36m, VB58, SE91i, LI60). The presence of bony hooks on the fin rays (usually on the anal and pelvic fins) is a secondary sexual structure of males, that is broadly distributed in the Characidae and also present in the Gasteropelecidae and Serrasalmidae (Malabarba & Weitzman, 2003) (state 1; Figs. 103 and 104). Malabarba & Weitzman (2003) proposed the presence of bony hooks in adult males as a synapomorphy of a clade including the Gasteropelecidae, Serrasalmidae, and Characidae, excepting Agoniates, Clupeacharax, and Engraulisoma. Most species examined for this paper have bony hooks; however, the absence of bony hooks can not be assumed to be typical for a species based on the observation of a limited number of specimens of such species. Rather it should be coded only for species for which a good sample of individuals and/or undoubtedly adult males were examined. Serra (2003) coded Salminus as lacking bony hooks; although in the examined material no hooks were observed for this genus, S. brasiliensis is coded with the state 1 following Morais Filho & Schubart (1955). Weitzman & Malabarba (2003) mentioned the presence of bony hooks in Microschemobrycon Eigenmann citing Böhlke (1953a) and Géry (1973); however, Géry (1973) explicitly mentioned the absence of hooks in M. casiquiare and this character is coded as missing for this species. Additionally to personal observations, the absence of bony hooks is coded for Rhoadsia altipinna (Cardoso, 2002), Nematobrycon palmeri, Pseudochalceus kyburzi (Bertaco, 2003), Coptobrycon bilineatus, Hasemania nana (Serra, 2003), Paragoniates alburnus, Xenagoniates bondi (Quevedo, 2006) and their presence for Creagrutus anary, C. cf. taphorni, Piabina argentea (Vari & Harold, 2001), Roeboides microlepis (Lucena, 2003), Bryconexodon juruenae (Lima, 2006), Roeboides descalvadensis (Lucena, 2007), Hemibrycon dariensis and H. surinamensis (Bertaco, 2008) following literature. 308. Anal-fin bony hooks in adult males of species bearing hooks on fins: (0) absent; (1) present. (CM35m, CM36m). Bony hooks, when present, are usually restricted to the anal and/or pelvic fins. A small group of species has bony hooks also on the other fins and this variation is considered below. Usually, there is an association between the presence of bony hooks in the pelvic and anal fins but this correlation is not perfect, and their presence in each of these fins is analyzed as separate characters. In most species whose mature males bear bony hooks, they are present in the anal fin (state 1), whereas in Characidium borellii and Nantis indefessus, among the examined species, these bony hooks are absent (state 0). This character is coded as inapplicable to species lacking bony hooks. 309. Pelvic-fin bony hooks in adult males of species bearing hooks on fins: (0) absent; (1) present. (CM35m, CM36m). Pelvic-fin bony hooks are present (state 1) in most species having secondary sexual bony hooks. These hooks are absent, however, (state 0) in several species in which the anal-fin hooks are present. The absence of pelvic-fin bony hooks is somewhat related with the compressed body of some species, which is in turn usually associated with a small size of the pelvic fin. 310. Pectoral-fin bony hooks in adult males of species bearing hooks on fins: (0) absent; (1) present. (MA39m, MA41m, BÜ138m). Bony hooks are usually absent in the pectoral (state 0), dorsal, and caudal fins. As the presence or absence of hooks on these fins is somewhat independent from each other, this character and the following two are coded as separate. Among the species bearing secondary sexual bony hooks, these structures are present in the pectoral fin in some species (state 1). All the examined species having pectoralfin bony hooks have also hooks on the dorsal fin, excepting Phenacogaster tegatus. 311. Dorsal-fin bony hooks in adult males of species bearing hooks on fins: (0) absent; (1) present. (MA39m, MA41m, VB37, BÜ138m). The presence of dorsal-fin bony hooks (state 1) is usually associated with the presence of hooks on the pectoral fin, as herein observed. However, the presence of bony hooks on the dorsal fin but their absence on the pectoral fin was herein observed in Bryconamericus thomasi and Hemigrammus erythrozonus. This character is coded as state 1 in Nematocharax venustus following Bertaco (2003). 312. Caudal-fin bony hooks in adult males of species bearing hooks on fins: (0) absent; (1) present.(MA39m, MA41m, BÜ138m). The presence of caudal-fin bony hooks (state 1) is usually associated with the presence of these structures on the pectoral and dorsal fins. However, several species with bony hooks on those fins lack bony hooks on the caudal fin (state 0). The presence of caudal-fin bony hooks but their absence on pectoral and dorsal fins was observed herein only in Acrobrycon tarijae. 313. Bony hooks on base of pelvic-fin rays of adult males: (0) absent, or in small number compared to on segmented portion of rays; (1) as numerous as on segmented portion of rays. Phylogeny of the family Characidae 456 Pelvic-fin bony hooks are usually more abundant on the branched region of the rays and are absent or much less concentrated in the basal portions of the rays (state 0). In some species the bony hooks are more densely concentrated along the basal portions of the pelvic-fin rays (state 1). Intermediate conditions were observed in Bryconamericus alpha, Hemigrammus erythrozonus, Knodus breviceps, Oligosarcus cf. jenynsii, and Prodontocharax melanotus, which are coded as polymorphic. 314. Bony hooks on last pelvic-fin ray of adult males: (0) absent or reduced in number; (1) as numerous as in other rays. Bony hooks are usually absent or much less concentrated in the last pelvic-fin ray than on the remaining rays (state 0). In some species, bony hooks are also present and relatively abundant on the last pelvic-fin ray (state 1; Fig. 103). An intermediate situation was observed in Odontostoechus lethostigmus, which is coded as polymorphic. 315. Bony hooks on first pelvic-fin ray of adult males: (0) absent; (1) present. The unbranched first pelvic-fin ray usually lacks bony hooks, even in species in which the hooks are abundant on other rays (state 0). The first pelvic-fin ray in a few species have, instead, relatively abundant bony hooks (state 1). This character is variable in Odontostoechus lethostigmus which is coded as polymorphic. 316. Position of anal-fin bony hooks of adult males: (0) paired and ordered laterally or posterolaterally; (1) medially positioned and oriented posteriorly; (2) asymmetrically disposed and irregularly arranged. (MA25m, BÜ145). The anal-fin bony hooks are usually paired and oriented laterally or posterolaterally (state 0). Malabarba (1998a) proposed the presence of medial, posteriorly-oriented bony hooks, as a synapomorphy of a clade of the Cheirodontinae (state 1; Malabarba, 1998a: fig. 17). The presence of irregularly arranged and asymmetrically placed anal-fin bony hooks was proposed as a synapomorphy of the Stethaprioninae of Reis (1989) (state 2; Reis, 1989: fig. 13). Scales Morphology: 317. Scales: (0) cycloid; (1) ctenoid; (2) spinoid; (3) crenate. (EI10, LU119m). Most characiforms have cycloid scales, without projections on their posterior field (state 0; Fig. 111). Vari (1979) reported the presence of ctenii along the posterior margin of the scales in the Distichodontidae and Citharidium Boulenger, and Lucena (1993) in Acestrocephalus sardina, Ctenobrycon hauxwellianus (Cope), Cynopotamus kincaidi (Schultz), and Galeocharax knerii (Steindachner). Following the classification of Roberts (1993), the scales of the Citharinidae and Distichodontidae are ctenoid (state 1), while those of Acestrocephalus, Cynopotamus and Galeocharax are spinoid (state 2; Fig. 112). The scales of Ctenobrycon (not analyzed here) and Psellogrammus are referred to as ctenoid in the literature; Roberts (1993) did not find ctenii in Ctenobrycon, differing with Lucena (1993), and did not examine specimens of Psellogrammus; thus, the type of squamation of that genera remained undiscussed by Lucena. The scales of the belly of Psellogrammus kennedyi have simple flattened serrations restricted to the margin of the scales similar to that of crenate scales in the classification of Roberts (1993) (state 3: e. g. Roberts, 1998: fig. 4G). 318. Anterior margin of scales: (0) uniformly curved or slightly undulated; (1) with conspicuous undulations. (LU121m). The shape of the anterior margin of the scales is variable among the examined species. In many species this margin is almost straight, somewhat rounded, or slightly expanded along its central portion (state 0; Fig. 112). As the shape of the anterior margin of the scales is variable even in different regions of the same specimen, it is only considered here the case in which the undulations are much evident and regularly present on scales of different regions of the body (state 1; Fig. 111). Eigenmann (1917) mentioned the presence of crenate scales in Entomolepis (=Bario) steindachneri, while Géry (1977) indicated that the scales of B. steindachneri become undulated when specimens reach 55 mm SL. In the specimen examined for this paper (62 mm SL), the scales are only slightly undulated anteriorly; given this variation during the growth of this species, Bario steindachneri is coded as polymorphic for this character. 319. Circulii on posterior field of scales: (0) present; (1) absent. (BU80, LU118, ZV171m). The circulii are concentric striae from the focus to the margins of the scales. In most characids the circulii are absent on the posterior field of the scales (state 1; Fig. 111), whereas most species of the outgroup and some characids have complete circulii reaching the posterior field of scales (state 0). The root of this analysis is coded as polymorphic because the circulii are absent in Puntius tetrazona but are present in other members of the Cypriniformes (Pflieger, 1997). 320. Radii on scales: (0) absent or reduced in number; (1) present and numerous on most scales. The radii are radially disposed grooves on the surface of scales extending from the focus to the margins. In most species the radii are uniformly present in all regions of the body, at least in the posterior field of scales (state 1; Fig. 113). In some species such radii are either absent, much reduced in number, or variably present in different regions of body (state 0; Fig. 111). 321. Radii oriented towards anterior field of scales: (0) present; (1) only as longitudinal groove without defined margins; (2) absent. (ZV172m). In most examined species the radii are only present on the posterior field of the scales (state 2), and do not converge towards the focus of the scales. Uj (1990) illustrated scales with radii oriented anteriorly (state 0) in some members of the Alestidae, Anostomidae, Ctenoluciidae, Erythrinidae, J. M. Mirande 457 Hepsetidae, and Parodontidae. Zanata & Vari (2005) mentioned that in most alestids, along with the genera Hepsetus, Hoplias, and Triportheus, the radii are oriented from the focus in all directions, even to the anterior field. A much wider anteriorly-oriented groove without defined margins, (state 1, Fig. 113) was observed in Bario steindachneri and Moenkhausia sanctaefilomenae. This character is coded as inapplicable for species in which the radii are absent. 322. Radii of scales: (0) not converging at focus; (1) converging at focus. Although the radii are oriented towards the focus, in most species they do not converge centrally, being independent each other (state 0). In some species the radii are in contact at the focus of the scales (state 1; Fig. 113). In Astyanax cf. abramis, A. asuncionensis, and A. cf. asuncionensis, this character is variable in different regions of the body of the same specimen, and this character is coded as polymorphic. This character is coded as inapplicable for species in which the radii are absent or reduced in number. 323. Semicircular grooves on posterior field of scales: (0) absent; (1) present. As mentioned above, the posterior field of the scales usually has radii and/or circulii (state 0). Uj (1990) illustrated a few grooves similar to radii but semicircular in shape, situated on the posterior field on the scales of Stethaprion erythrops Cope (state 1). The presence of these grooves was corroborated in this paper. Although this character is autapomorphic for S. erythrops in this analysis, this character is included as potentially informative in the future. Scale distribution: 324. Scales covering supraoccipital spine: (0) absent; (1) present and completely covering supraoccipital spine. (VA74, BU18, LU24, ZV168, LI61m, PZ102). In most examined species the dorsal scales extend anteriorly to the posterior margin of the parietals and supraoccipital but do not cover the supraoccipital spine, which projects posteriorly between rows of scales (state 0). Vari (1995) mentioned that the anteriormost dorsal scales extend anteriorly dorsal to the supraoccipital spine and cover it completely from dorsal view (state 1) in the Ctenoluciidae, Erythrinidae, Hepsetidae, and Lebiasinidae; this state was considered by him as a synapomorphy for a clade composed of these four families. Vari (1995) mentioned a similar coverage of the supraoccipital spine in the Alestidae and Parodontidae, although he considered these cases as not comparable given the lack of a supraoccipital crest in the latter two families. This character is considered inapplicable to species lacking median predorsal scales. Fig. 111. Lateral-line scale of Markiana nigripinnis, CI-FML 3936, 75.3 mm SL, anterior to top. Scale bar = 1 mm. Fig. 112. Scale of Galeocharax humeralis, CI-FML 3951, 94.6 mm SL, anterior to top. Scale bar = 0.5 mm. Fig. 113. Scale of Moenkhausia sanctaefilomenae, CI-FML 3939, 29.1 mm SL, anterior to top. Scale bar = 0.1 mm. Phylogeny of the family Characidae 464 354. Hypertrophied ventral caudal-peduncle squamation: (0) absent; (1) present. (MW6) The presence of a glandular fold on the ventral lobe of the caudal fin (state 1; Weitzman & Menezes, 1998: figs. 10-16) was considered as diagnostic for the Glandulocaudinae (Géry, 1977; Weitzman & Fink, 1985; Weitzman & Menezes, 1998). The monophyly of this subfamily was later questioned and the Stevardiinae was proposed to include most species of the former glandulocaudins (Weitzman et al., 2005). Weitzman & Menezes (2009) analyzed the glandular fold in two different characters, considering the origin of the scales involved in such fold. They considered the hypertrophy of the scales of the dorsal lobe of the caudal fin and that of the ventral lobe as different characters and proposed the first character to be diagnostic of their Glandulocaudinae and the second character to be synapomorphic for their Stevardiinae. The presence (state 1; e. g. Weitzman & Menezes, 1998: fig. 14) or absence of hypertrophy of scales of the caudal-fin ventral lobe is herein coded. This character was not included as such by Mirande (2009), who instead coded the presence or absence of a caudal fold of scales, without considering if the scales forming this fold were from the dorsal or ventral caudal-fin lobe. 355. Caudal gland cells consisting of modified mucous cells: (0) absent; (1) present. (MW5). The presence of caudal gland cells formed by modified mucous cells (state 1) was proposed by Menezes & Weitzman (2009) as a synapomorphy of their Stevardiinae. No histological examination were made for this paper and the coding of this character is exclusively based on Menezes & Weitzman (2009). The presence or absence (state 0) of modified mucous cells was coded only at generic or suprageneric levels by Menezes & Weitzman (2009) and their presence or absence should be corroborated in each species. This character was not included by Mirande (2009). 356. Adipose fin: (0) present; (1) absent. (EI15, VA81, MA12, VB39, SE90m, QU41). The presence of an adipose fin (state 0) was considered as a plesiomorphy of the Ostariophysi (Fink & Fink, 1981) that was independently lost in the Gonorynchiformes, Cypriniformes, and Gymnotiformes. However, according to the phylogenetic hypothesis of Fink & Fink (1981), the correct optimization of this character is ambiguous, and their interpretation of character-state evolution is equally parsimonious to the absence of the adipose fin in the ancestor of the Ostariophysi and its acquisition in the common ancestor of the Characiformes, Gymnotiformes, and Siluriformes. This character, thus, is coded as polymorphic to the root of this analysis. In most members of the Characiformes the adipose fin is present, while it is absent (state 1) in the Erythrinidae, Lebiasinidae, and some characids. 357. Papillae on tongue: (0) not aligned; (1) forming longitudinal rows anteriorly. The epithelial papillae situated in the dorsal surface of the primary tongue are irregularly arranged in most examined species (state 0; Fig. 123). In some species these papillae are instead aligned, forming four to six conspicuous longitudinal rows on the anterior region of dorsal surface of the primary tongue (state 1; Fig. 124). 358. Insemination: (0) absent; (1) present. (MA70, WM2, BÜ168, QU92, MW1). Although a detailed study of the reproductive biology of the species included in this analysis is beyond the scope of this Fig. 123. Primary tongue and anterior branchial skeleton of Oligosarcus bolivianus, CI-FML 3366, 83.4 mm SL, dorsal view, anterior to left. Non-permanently stained with methylene blue. Scale bar = 1 mm. Fig. 124. Primary tongue and anterior branchial skeleton of Astyanax cf. asuncionensis, CI-FML 3912, 61.1 mm SL, dorsal view, anterior to left. Non-permanently stained with methylene blue. Scale bar = 0.5 mm. J. M. Mirande 465 paper, the available data about insemination is herein analyzed in a familial phylogenetic context. Although the list of characid inseminating species is surely far from complete, the published list of species in which insemination is known to be absent is even more incomplete. Thus, this and the following character, which is also related with reproductive biology, have a relatively high proportion of missing entries. Most characids have external fertilization (state 0), while the presence of insemination (state 1) was reported for several genera of the Cheirodontinae and Stevardiinae and the genus Hollandichthys (Bertaco, 2003; Burns et al., 1995, 1997, 1998; Burns & Weitzman, 2005; Castro et al., 2003; Menezes et al., 2003; Weitzman et al., 2005). This character is coded exclusively from literature. 359. Type of spermatozoa: (0) aquasperm; (1) introsperm. (MA71, WM47m, BÜ169, QU91). The spermatozoa in externally fertilizing species are usually of the aquasperm type, with a rounded nucleus (state 0), while in the inseminating species they are usually of the introsperm type, with an elongated nucleus (state 1) (Burns & Weitzman, 2005). Although there is a correlation between the presence of insemination and introsperms, there are exceptions to such relationships (Burns & Weitzman, 2005). This character was coded only from literature, and many species are coded as missing entries. 360. Sperm storage area on testes: (0) absent or small; (1) present, as broad as spermatogenic area. (QU93). The presence of a developed aspermatogenic region in the testicle, which serves as storage area (state 1) was proposed by Weitzman & Menezes (1998) as a synapomorphy of their Glandulocaudinae. A similar storage area was observed by Bertaco (2003) in the species of Hollandichthys. This storage area is absent or extremely reduced in most of the Characidae (state 0). This character was coded exclusively from literature. 361. Number of 2n chromosomes: (0) 36 to 40; (1) 46 or more. 362. Number of 2n chromosomes: (0) 48 or less; (1) 50 or more. 363. Number of 2n chromosomes: (0) 50 or less; (1) 52 or more. 364. Number of 2n chromosomes: (0) 52 or less; (1) 54 or more. 365. Number of 2n chromosomes: (0) 56 or less; (1) 58 or more. Although several species analyzed herein have published information as to their chromosome number, this information was not previously included in phylogenetic analyses. The root of this analysis is coded as polymorphic for the characters 363, 364, and 365 because the number of chromosomes in the Cypriniformes usually ranges between 48 and 52. This character was coded from the literature (Arefjev, 1990a, b; Artoni & Bertollo, 2002; Bellafronte et al., 2005; Bertollo et al., 1986; Carvalho et al., 2002; Centofante et al., 2003; Cestari & Galetti, 1992; Falcão & Bertollo, 1985; Foresti et al., 1989; Galetti Jr. et al., 1981; ; Hinegardner & Rosen, 1972 ; Kirby et al., 1977; Paintner-Marques et al., 2002; Pauls & Bertollo, 1984; Portela et al., 1988; Porto et al., 1992; Silva & Maistro, 2006). Different numbers of chromosomes were reported for several species, which might indicate identification problems. These usually occurs in non-systematic papers; thus, several species are coded as polymorphic denoting some uncertainty as to identifications rather than polymorphisms per se. Phylogenetic Results The final hypothesis is the strict consensus between the most parsimonious trees obtained in a range of K-values under implied weighting (Goloboff, 1993). Details about the analysis itself, including the criteria by which this final hypothesis was constructed were already published (Mirande, 2009). The final hypothesis is presented in the Figs. 125-129. Diagnosis of the obtained clades The common synapomorphies of the individual trees are listed for each node of the final hypothesis. Those synapomorphic changes that are present only in some of the original trees are listed under “some trees”. The node numbers correspond to those obtained from the TNT software and presented in the Figs. 125-129. GC values and relative frequencies as measures of stability, and GC values and relative Bremer support as measures of support are expressed between parentheses in each node (see Material and Methods, and Mirande, 2009). OUTGROUP Node 172: Characoidea (100 / 100 / 76 / 24) Families Alestidae, Characidae, Gasteropelecidae, and Serrasalmidae. The superfamily Characoidea was proposed by Buckup (1998) to include the Characidae (containing the Serrasalmidae) and Gasteropelecidae. That author did not include the Gasteropelecidae in his analysis, but classified that family with the Characidae based mainly on Géry (1977), who stated that the gasteropelecids have most characters in common with certain characids. The Characoidea is redefined in this paper to include, in addition, the former families Acestrorhynchidae and Cynodontidae (herein considered as subfamilies of the Characidae) and the Alestidae. In the publication by Buckup (1998) both the Acestrorhynchidae and Cynodontidae are included in the superfamily Cynodontoidea, while the Alestidae was classified in its own superfamily, the Alestoidea. Although no serrasalmids were included in the phylogeny of Buckup (1998), that author preferred to maintain this group as a subfamily of the Characidae in light of the lack of phylogenetic information refuting a close relationship of these families. In the molecular hypothesis of Calcagnotto et al. (2005) the Serrasalmidae, however, forms a clade separate from the Characidae. The hypothesis herein proposed also supports a familial level status for serrasalmids. Phylogeny of the family Characidae 466 Synapomorphies: 1. Relative position of anterior margin of antorbital and first infraorbital (57): (1 > 0) anterior margin of antorbital either aligned with or anterior to first infraorbital. Reversed in node 184 and in Galeocharax humeralis. 2. Form of lateral line (89): (0 > 1) curved ventrally in abdominal region. Paralleled in Hemiodus cf. thayeria. Reversed in Rhaphiodon vulpinus. 3. Relative number of precaudal vertebrae (226): (0 > 1) equal or less numerous than caudal vertebrae. Reversed in Brycinus carolinae, Chalceus macrolepidotus, and Gymnocharacinus bergii. 4. Dorsal longitudinal ridge on medial lamella of pelvic bone (264): (0 > 1) absent. Reversed in the node 184. 5. Number of branched anal-fin rays (286): (0 > 1) 11 or more. Reversed in Chalceus macrolepidotus. 6. Number of branched anal-fin rays (287): (0 > 1) 18 or more. Reversed in the Alestidae, in nodes 280 and 290, and in Attonitus ephimeros and Prodontocharax melanotus. Some trees: Reversed in Hasemania nana. Paralleled in node 182. 7. Lateral lamellae on anterior anal pterygiophores (295): (1 > 0) absent. Reversed in the Alestidae. Node 171: Gasteropelecidae (100 / 100 / 100 / 89) Genera Carnegiella, Engraulisoma, Gasteropelecus Scopoli, and Thoracocharax; genus Clupeacharax? Although the monophyly of the Gasteropelecidae, as traditionally defined, has long been obvious (Weitzman, 1954, 1960; Buckup, 1998), the relationships of this family with the remaining Characiformes had not been adequately tested. The great morphological divergence of the gasteropelecids probably complicated their inclusion in previous morphological phylogenies; however, no members of this family were included also in the molecular phylogeny of Calcagnotto et al. (2005). As herein proposed this family includes all genera traditionally considered in the Gasteropelecidae plus Engraulisoma (and tentatively also Clupeacharax). Castro (1984) proposed a close relationship between Engraulisoma taeniatum and Clupeacharax anchoveoides, the only member of the characid subfamily Clupeacharacinae (Géry, 1977; Lima, 2003a), based on seven putative synapomorphies: the possession of a foramen in the ventral surface of the pterotic, the short supraorbital, the fusion of the third and fourth infraorbitals (*), the antorbital with unique form and size within Characidae (*), the third branchiostegal ray articulated in the suture between the anterior and posterior ceratohyals, the absence of the second and third postcleithra (*), and the presence of interdigitations between the ischiatic processes of the pelvic bones. Synapomorphies marked with (*) are comparable to those defining the Gasteropelecidae in this analysis. The fusion of the posttemporal and supracleithrum, and the possession of only one epural, mentioned as autapomorphies of Engraulisoma taeniatum by Castro (1984), are shared with all the former Gasteropelecidae and with Thoracocharax stellatus, respectively. Similarly, the posteriorly situated dorsal fin is shared between Clupeacharax and the Gasteropelecidae sensu Weitzman (1954). Thus, although Clupeacharax anchoveoides was not included in this analysis, it can be tentatively included as incertae sedis within this clade, given the observations of Castro (1984). With conservativeness as a criterion, the genera Clupeacharax and Engraulisoma should be included in the Gasteropelecidae whether if the former genus is actually the sister group of Engraulisoma, as proposed by Castro (1984), or if Clupeacharax is basal to Engraulisoma and the remaining Gasteropelecidae. Under all possible scenarios Engraulisoma taeniatum and probably also Clupeacharax anchoveoides are removed from the Characidae and included in the Gasteropelecidae. Fig. 125. Cladogram showing relationships of the Characidae with remaining characiforms. Node numbers correspond to those in the text. J. M. Mirande 467 Synapomorphies: 1. Position of antorbital relative to lateral ethmoid in lateral view (56): (0 > 1) antorbital overlapping lateral ethmoid. Paralleled in node 302. 2. Fourth infraorbital (66): (0 > 1) absent or much reduced and bordered posteriorly by third and fifth infraorbitals. Paralleled in the Aphyocharacinae, in node 186, and in Aphyodite grammica, Hasemania nana, Hemigrammus erythrozonus, Hoplocharax goethei, Hyphessobrycon pulchripinnis, and Nematobrycon palmeri. 3. Laterosensory canal of first infraorbital (73): (1 > 0) projects dorsally from main body of first infraorbital. Paralleled in the Heterocharacinae. 4. Alignment of ascending process of premaxilla (105): (0 > 1) medially shifted and separated from nasal. Paralleled in node 193. 5. Position of coronomeckelian (110): (1 > 0) situated mainly lateral to Meckelian cartilage. Paralleled in nodes 176 and 206, and in Hoplias cf. malabaricus and Prochilodus lineatus. 6. Inner row of dentary teeth (143): (0 > 1) absent. Paralleled in nodes 166, 168, and 189, and in Rhaphiodon vulpinus and Serrasalmus maculatus. 7. Teeth on third pharyngobranchial (206): (0 > 1) absent. Paralleled in Piaractus mesopotamicus. 8. Second postcleithrum (248): (0 > 1) absent. Paralleled in node 302 and in Pseudocorynopoma doriae and Rhaphiodon vulpinus. 9. Third postcleithrum (249): (0 > 1) absent. Paralleled in node 302, and in Piabucus melanostomus, Pyrrhulina australis, Rhaphiodon vulpinus, and Xenagoniates bondi. 10. Fusion between posttemporal and supracleithrum (255): (0 > 1) present. 11. Anterior extension of pelvic-bone along main axis (262): (0 > 1) projecting anterior of lateral and medial lamellae of pelvic bone. Paralleled in node 302 and in Hoplias cf. malabaricus, Piabucus melanostomus, Rhaphiodon vulpinus, and Stethaprion erythrops. 12. Circulii on posterior field of scales (319): (0 > 1) absent. Paralleled in nodes 168 and 206, and in Agoniates anchovia. Some trees: Paralleled in node 302. Autapomorphies of Engraulisoma taeniatum: 1. Extent of expansion of first infraorbital lateral to maxilla (59): (0 > 1) covering most of maxilla. Paralleled in node 277 and in Heterocharax macrolepis. 2. Shape of ectopterygoid (156): (0 > 2) approximately square. 3. Contact between ectopterygoid and anterodorsal region of quadrate (162): (0 > 1) absent. Paralleled in nodes 184 and 242, and in Aphyocharax dentatus, Prionobrama paraguayensis, and Stichonodon insignis. Some trees: Paralleled in the Cheirodontinae and in Microschemobrycon casiquiare. 4. Interhyal (210): (0 > 1) absent. Paralleled in Aulixidens eugeniae. 5. Number of branchiostegal rays (212): (1 > 0) three. Paralleled in Apareiodon affinis, Leporinus striatus, and Pyrrhulina australis. 6. Development of transverse process of neural arch of third vertebra (219): (0 > 1) well developed and extending beyond anterior margin of tripus. Paralleled in node 302, and in Agoniates anchovia, Cyanocharax alburnus, Deuterodon langei, Hemiodus cf. thayeria, Roeboexodon geryi, and Thayeria obliqua. Some trees: Paralleled in Microschemobrycon casiquiare and Parecbasis cyclolepis. 7. Transitional vertebrae with haemal canal (229): (0 > 1) absent. Paralleled in nodes 195 and 212, and in Aulixidens eugeniae, Metynnis maculatus, and Piabina argentea. Some trees: Paralleled in node 247 and in Bryconamericus alpha and Paracheirodon axelrodi. 8. Base of second pectoral ray (231): (0 > 1) similar in form and size to base of posterior rays. Paralleled in node 175. Some trees: Paralleled in node 302. 9. Bony ridge of coracoid between base of mesocoracoid and ventral margin of interosseous space (239): (1 > 0) absent. Paralleled in node 204. 10. Articulation between pelvic bones (261): (0 > 1) with bony interdigitations between ischiatic processes. Paralleled in node 302 and in Agoniates anchovia. Fig. 126. Cladogram showing relationships between basal Characidae. Node numbers correspond to those in the text. Phylogeny of the family Characidae 468 11. Relative length of anterior dorsal-fin rays (271): (0 > 1) reaching tip of posterior rays when adpressed. Paralleled in nodes 163 and 179 and in Pyrrhulina australis. 12. Position of last supraneural (283): (0 > 1) located more than two vertebrae in front of first dorsal pterygiophore. Paralleled in nodes 174 and 244 and in Gymnocharacinus bergii and Xenagoniates bondi. 13. Proximal and medial radials of anal fins (294): (2 > 0) fused on anterior five pterygiophores. Paralleled in node 179. 14. Posterior margin of hypural 3 (300): (1 > 0) equal to or narrower than posterior margin of hypural 4. Paralleled in node 181 and in the Alestidae. 15. Longitudinal position of insertion of adductor mandibulae tendon on dentary (330): (0 > 1) on vertical through middle or anterior half of Meckelian cartilage. Paralleled in the Iguanodectinae, in nodes 184, 186, 209, 241, 261, and 270, and in Gymnocharacinus bergii. Node 170: (100 / 100 / 100 / 65) Genera Carnegiella, Gasteropelecus, and Thoracocharax. This node includes the Gasteropelecidae as defined by Weitzman (1954), and recognized in subsequent papers (e. g. Géry, 1977). As Gasteropelecus was not analyzed, it is possible that some synapomorphies of this node actually correspond to a more restricted clade. Although the monophyly of a clade composed of Carnegiella, Gasteropelecus, and Thoracocharax was never tested in a phylogenetic context, it never was questioned, given the high resemblance of these genera and their morphological divergence from the remaining Characiformes (Weitzman, 1954; Weitzman & Palmer, 2003). The phylogenetic position of this clade, however, was unknown, because no previous higher-level phylogenies included species of the Gasteropelecidae. According to Géry (1977: 243), the members of this node “have most (characters) in common with the Characidae, chiefly with certain tetras”; however, this author maintained the Gasteropelecidae as a family, separated from the Characidae. Malabarba & Weitzman (2003) proposed a clade supported by the presence of bony hooks on the fin rays, which included the Characidae, Gasteropelecidae, and Fig. 127. Cladogram showing relationships between Characidae lacking a supraorbital bone (Bramocharax clade, Pseudochalceus clade, Characinae, Rhoadsiinae, and Tetragonopterinae). Node numbers correspond to those in the text. J. M. Mirande 469 Serrasalmidae with the exception of the genera Agoniates, Clupeacharax, and Engraulisoma. According to their hypothesis both the Serrasalmidae and Gasteropelecidae should be included in the Characidae. However, such hypothesis was rather speculative, and not based on any published phylogenetic analysis. Synapomorphies: 1. Ventral longitudinal lamellae of basioccipital (2): (1 > 0) falling short of posterior border of basioccipital. Paralleled in node 205 and in Serrasalmus maculatus. 2. Ventral projection of lagenar capsule (3): (0 > 1) extending ventrally to articulation between basioccipital and parasphenoid. Paralleled in the Serrasalmidae, in nodes 205, and 302, and in Cyphocharax spilotus and Micralestes stormsi. 3. Articulation between medial region of lateral ethmoid and frontal or mesethmoid (17): (0 > 1) extensive articulation of entire lateral ethmoid dorsal margin. Paralleled in node 193. 4. Form of orbitosphenoid (37): (1 > 0) slender, relatively small and separate from parasphenoid. Paralleled in the Characidae, in node 168, and in Hemiodus cf. thayeria. 5. Parietal fontanel (41): (0 > 1) absent in adults. Paralleled in nodes 162 and 181 and in Brycinus carolinae and Brycon pesu. 6. Supraorbital (70): (0 > 1) absent. Paralleled in nodes 185 and 205 and in Micralestes stormsi. 7. Lateral surface of vertical canal of preopercle (81): (1 > 0) canal uncovered and situated posteriorly to musculature and infraorbitals. Paralleled in nodes 175 and 204 and in Chalceus macrolepidotus and Metynnis maculatus. 8. Number of teeth in inner premaxillary row (130): (0 > 1) eight or more. Paralleled in the Aphyoditeinae and in Brycon orbignyanus, Grundulus cochae, Phenacogaster tegatus, Prionobrama paraguayensis, and Salminus brasiliensis. 9. Form of quadrate (150): (0 > 1) with anterodorsal portion equal or longer than ventral region. Paralleled in the Pseudochalceus clade, in nodes 176, 211, and 299, and in Exodon paradoxus and Hoplias cf. malabaricus. 10. Form of anterior portion of ectopterygoid (157): (0 > 1) slender and articulating only to lateral margin of palatine, and lacking ligaments to neurocranium. Paralleled in the Alestidae and in Agoniates anchovia and Attonitus ephimeros. 11. Bony lamella dorsal to fourth basibranchial (185): (1 > 0) present. Paralleled in nodes 168, 203, and 302 and in Phenacogaster tegatus. 12. Anterior development of basihyal (190): (1 > 0) broadly extending beyond anterior margin of hypohyals. Paralleled in node 302. 13. Rows of gill rakers on first ceratobranchial (192): (1 > 0) one. Paralleled in node 179. 14. Development of medial lamella of coracoid (238): (0 > 1) expanded as a keel. Paralleled in node 302 and in Paragoniates alburnus, Piabucus melanostomus, Pseudocorynopoma doriae, and Rhaphiodon vulpinus. 15. First postcleithrum (247): (0 > 1) absent. Paralleled in node 236. 16. Dorsal myorhabdoi (273): (0 > 1) present. Paralleled in Rhaphiodon vulpinus. 17. Position of anteriormost epineurals (274): (0 > 1) reaching to cranium. Paralleled in nodes 175 and 302 and in Distichodus maculatus and Piabucus melanostomus. 18. Number of supraneurals (281): (0 > 1) eight or more. Paralleled in node 207 and in Hemiodus cf. thayeria and Pyrrhulina australis. 19. Anal-fin position (284): (0 > 1) extended anteriorly ventral to dorsal fin. Paralleled in nodes 208, 212, and 236 and in Piabucus melanostomus. Fig. 128. Cladogram showing relationships between Characidae lacking a supraorbital bone (Astyanax clade, Astyanax paris clade, Bryconamericus scleroparius clade, Hyphessobrycon luetkenii clade, and Gymnocharacinae). Node numbers correspond to those in the text. Phylogeny of the family Characidae 470 Autapomorphies of Thoracocharax stellatus: 1. Opening between orbitosphenoid and pterosphenoid (39): (1 > 0) present, rounded or ovate, usually margined by frontal dorsally. Paralleled in nodes 205 and 302. 2. Ventral extent of third infraorbital (64): (0 > 1) not reaching horizontal arm of preopercle, at least anteriorly. Paralleled in nodes 168 and 180. 3. Number of maxillary teeth (135): (0 > 1) two or more. 4. Articulation between ventral margin of metapterygoid and posterodorsal margin of quadrate (155): (0 > 1) present. Paralleled in nodes 176 and 299 and in Deuterodon langei, Heterocharax macrolepis, Pristella maxillaris, and Roeboides descalvadensis. 5. First basibranchial (182): (1 > 0) absent or much reduced, not articulating anteriorly with basihyal. Paralleled in node 163. 6. Number of branchiostegal rays (213): (0 > 1) five. Paralleled in Characidium borellii, Hoplias cf. malabaricus, Piaractus mesopotamicus, and Rhaphiodon vulpinus. 7. Number of dorsal pterygiophores (278): (0 > 1) 12 or more. Paralleled in the Serrasalmidae and in Characidium rachovii, Distichodus maculatus, and Hoplias cf. malabaricus. 8. Number of branched anal-fin rays (289): (0 > 1) 35 or more. Paralleled in nodes 207 and 212 and in Gymnocorymbus ternetzi, Metynnis maculatus, Piabucus melanostomus, Pseudocorynopoma doriae, Rhaphiodon vulpinus, and Stethaprion erythrops. Some trees: Paralleled in node 261 and in Markiana nigripinnis. 9. Scales covering anal-fin base (327): (0 > 1) several rows covering basal third of anal fin. Paralleled in the Serrasalmidae, in nodes 210 and 221, and in Bario steindachneri, Markiana nigripinnis, Paragoniates alburnus, Rhaphiodon vulpinus, and Roeboides microlepis. 10. Ventral union of gill membranes (349): (0 > 1) joined along length of isthmus but not attached to isthmus. Paralleled in the Iguanodectinae and in node 162. Autapomorphies of Carnegiella strigata: 1. Number of rows of premaxillary teeth (122): (1 > 0) one. Paralleled in node 195 and in Aulixidens eugeniae, Carlana eigenmanni, Grundulus cochae, Odontostoechus lethostigmus, Paracheirodon axelrodi, Piabucus melanostomus, and Probolodus heterostomus. 2. Edentulous basihyal lamella (189): (0 > 1) present. Paralleled in the Alestidae and Heterocharacinae, in nodes 175 and 302, and in Bryconops affinis. 3. Adipose fin (356): (0 > 1) absent. Paralleled in the Gymnocharacinae, in node 181, and in Phenagoniates macrolepis. Fig. 129. Cladogram showing relationships between Characidae lacking a supraorbital bone (Aphyocharacinae, Aphyoditeinae, Cheirodontinae, and Stevardiinae). Node numbers correspond to those in the text. J. M. Mirande 471 Node 180: (100 / 100 / – / 12) Families Alestidae, Characidae, and Serrasalmidae. The monophyly of a clade equivalent to this grouping was not proposed in previous phylogenies. Uj (1990) proposed the monophyly of an African clade composed of the Alestidae, Citharinidae, and Distichodontidae, which was repeatedly contradicted in all subsequent phylogenies. The hypothesis of Buckup (1998) included the Erythrinoidea in a node similar to this one, making his results incompatible with those obtained in this study. In the hypothesis of Calcagnotto et al. (2005) the Alestidae are more related to the Crenuchidae and Erythrinoidea than to the Characidae. The families forming this node formed a trichotomy in the final hypothesis proposed by Mirande (2009), whereas the Serrasalmidae are herein proposed to be the sister group of the Alestidae and Characidae. Synapomorphies: 1. Length of sphenotic spine (10): (0 > 1) extending ventrally to articulation between sphenotic and hyomandibula. Paralleled in Prochilodus lineatus. Reversed in the Characinae and in nodes 176 and 197. 2. Ventral diverging lamellae of mesethmoid (30): (0 > 1) present. Reversed in node 184. 3. Ventral extent of third infraorbital (64): (0 > 1) not reaching horizontal arm of preopercle, at least anteriorly. Paralleled in node 168 and in Thoracocharax stellatus. Reversed in the Heterocharacinae and Iguanodectinae, in nodes 184, 198, 225, and 302, and in Agoniates anchovia, Brycon pesu, Hasemania nana, Markiana nigripinnis, Moenkhausia sanctaefilomenae, Pseudochalceus kyburzi, Roeboexodon geryi, and Stichonodon insignis. Some trees: Reversed in node 292. 4. Tubules for passage of blood vessels on lamellar portion of maxilla (98): (0 > 2) anastomosed tubules. 5. Horizontal process of anguloarticular (108): (0 > 1) broadly covered by dentary which reaches posterior border of Meckelian cartilage. Reversed in nodes 176 and 206. 6. Ascending process of neural pedicle of third vertebra (220): (0 > 1) present. Paralleled in Hemiodus cf. thayeria. Reversed in Rhaphiodon vulpinus. 7. Supraneural anterior to neural spine of fourth vertebra (279): (0 > 1) present and vertically elongate. Reversed in the Iguanodectinae, in node 204, and in Micralestes stormsi. 8. Uroneurals (306): (0 > 1) two pairs. Paralleled in node 163. Reversed in node 205. Node 187: Serrasalmidae (100 / 100 / 100 / 59) Genera Acnodon Eigenmann, Catoprion Müller & Troschel, Colossoma Eigenmann & Kennedy, †Megapiranha Cione, Dahdul, Lundberg & Machado-Allison, Metynnis Cope, Mylesinus Valenciennes, Myleus Müller & Troschel, Mylossoma Eigenmann & Kennedy, Ossubtus Jégu, Piaractus, Pristobrycon Eigenmann, Pygocentrus Müller & Troschel, Pygopristis Müller & Troschel, Serrasalmus, Tometes Valenciennes, and Utiaritichthys Miranda Ribeiro. The monophyly of this clade was proposed by Machado-Allison (1983, 1986), supported by 27 synapomorphies, and subsequently corroborated by the molecular phylogenies by Ortí et al. (1996) and Calcagnotto et al. (2005). The taxonomic level of this group has been longly debated. Géry (1977) treated it as a family, while Machado-Allison (1982, 1983, 1985, 1986) and Jégu (2003) considered this clade as a subfamily of the Characidae. According to the phylogenetic hypothesis of Calcagnotto et al. (2005), this clade should be classified as a family, as herein proposed. As only a small sample of the serrasalmids are herein analyzed, some synapomorphies could correspond to more inclusive clades. Synapomorphies: 1. Ventral projection of lagenar capsule (3): (0 > 1) extending ventrally to articulation between basioccipital and parasphenoid. Paralleled in nodes 170, 205, and 302 and in Cyphocharax spilotus and Micralestes stormsi. 2. Length of supraoccipital spine (53): (1 > 0) extends posteriorly to, at least, middle length of neural complex of Weberian apparatus. Paralleled in nodes 182 and 205 and in Piabucus melanostomus and Prochilodus lineatus. Some trees: Paralleled in Brycon meeki and B. orbignyanus. 3. Teeth on inner premaxillary row (126): (1 > 0) molariform. 4. Polymorphism of teeth on inner premaxillary row (131): (0 > 1) present, with two medial teeth somewhat larger and usually separated from remaining ones by a gap. Paralleled in the Bryconinae. 5. Posterior directed radial striae from articular region of opercle (169): (0 > 1) present. Paralleled in Prochilodus lineatus. 6. Length of medial bony ridge of opercle (170): (1 > 0) 60% or greater than opercular length. Paralleled in node 210 and in Astyanax abramis, Creagrutus cf. taphorni, Hoplias cf. malabaricus, and Roeboides microlepis. Some trees: Paralleled in Acestrorhynchus pantaneiro and Salminus brasiliensis. 7. Bony lamellae between second and third basibranchials (184): (1 > 0) absent. Paralleled in Attonitus ephimeros, Axelrodia lindeae, Hollandichthys multifasciatus, Hoplocharax goethei, Jupiaba scologaster, Piabucus melanostomus, Pyrrhulina australis, Rhaphiodon vulpinus, and Xenagoniates bondi. 8. Lateral base of gill rakers on first ceratobranchial (199): (0 > 1) broad and laminar at least on anteriormost gill rakers. Paralleled in in nodes 166 and 177 and in Hoplias cf. malabaricus. Reversed in Agoniates anchovia. 9. Abdominal ribs on anterior caudal vertebrae (225): (0 > 1) present, associated to first and occasionally second caudal vertebrae. 10. Number of dorsal pterygiophores (278): (0 > 1) 12 or more. Paralleled in Characidium rachovii, Distichodus maculatus, Hoplias cf. malabaricus, and Thoracocharax stellatus. 11. Bony lamellae associated with supraneurals (282): (0 > 1) wider than primary axis of supraneurals. Paralleled in Micralestes stormsi. Phylogeny of the family Characidae 472 12. Fusion of hypural 2 to compound centrum (298): (1 > 0) absent. Paralleled in nodes 174 and 184 and in Distichodus maculatus, Hemiodus cf. thayeria, and Hoplias cf. malabaricus. 13. Radii on scales (320): (1 > 0) absent or reduced in number. Paralleled in the Iguanodectinae, in node 174, and in Cyphocharax spilotus, Distichodus maculatus, Markiana nigripinnis, and Phenagoniates macrolepis. 14. Median predorsal scales (325): (0 > 1) leaving naked area anterior to dorsal fin. Paralleled in node 284 and in Lonchogenys ilisha. 15. Ventral serrae (326): (0 > 1) present. 16. Scales covering anal-fin base (327): (0 > 1) several rows covering basal third of anal fin. Paralleled in nodes 210 and 221 and in Bario steindachneri, Markiana nigripinnis, Paragoniates alburnus, Rhaphiodon vulpinus, Roeboides microlepis, and Thoracocharax stellatus. 17. Diffuse spots on flanks (346): (0 > 1) present, especially in young specimens. 18. Number of 2n chromosomes (363): (0 > 1) 52 or more. Paralleled in the Characinae, in node 196, and in Chalceus macrolepidotus, Hyphessobrycon herbertaxelrodi, Markiana nigripinnis, and Rhaphiodon vulpinus. Some trees: Paralleled in Hemigrammus unilineatus. 19. Number of 2n chromosomes (364): (0 > 1) 54 or more. Paralleled in Rhaphiodon vulpinus. Autapomorphies of Piaractus mesopotamicus: 1. Branching of laterosensory canals of fourth or fifth infraorbitals (74): (0 > 1) present. Paralleled in the Bryconops clade, in nodes 167, 177, 218, 260, and 276, and in Bryconamericus scleroparius and Chalceus macrolepidotus. 2. Canal of lateral line on caudal-fin membrane (92): (0 > 1) present. Paralleled in the Characidae. 3. Interdigitations between premaxillae (103): (1 > 0) present. Paralleled in the Bryconinae and in Distichodus maculatus. 4. Form of interdigitations between dentaries (112): (0 > 1) undulate lamellae. Paralleled in node 183. 5. Anterior extension of interopercle (163): (0 > 1) not extending anteriorly beyond terminus of horizontal arm of preopercle. Paralleled in the Heterocharacinae, in nodes 162, 174, and 212, and in Hoplias cf. malabaricus. 6. Number of gill rakers on first hypobranchial and ceratobranchial (195): (1 > 0) 16 or more. Paralleled in nodes 177 and 183 and in Astyanax latens, A. cf. rutilus, A. pelegrini Eigenmann, Hoplias cf. malabaricus, Hyphessobrycon socolofi, Moenkhausia dichroura, Parecbasis cyclolepis, and Stichonodon insignis. 7. Teeth on third pharyngobranchial (206): (0 > 1) absent. Paralleled in the Gasteropelecidae. 8. Number of branchiostegal rays (213): (0 > 1) five. Paralleled in Characidium borellii, Hoplias cf. malabaricus, Rhaphiodon vulpinus, and Thoracocharax stellatus. Node 186: (100 / 100 / 90 / 54) Genera Acnodon?, Catoprion?, Colossoma?, †Megapiranha?, Metynnis, Mylesinus?, Myleus?, Mylossoma?, Ossubtus?, Pristobrycon?, Pygocentrus?, Pygopristis?, Serrasalmus, Tometes?, and Utiaritichthys? The internal relationships of the Serrasalmidae are unresolved, and several hypotheses has been proposed regarding this issue. Machado-Allison (1983) proposed the monophyly of a clade including the genera Catoprion, Metynnis, Pygocentrus, Pygopristis, Pristobrycon, and Serrasalmus. In the hypothesis of Calcagnotto et al. (2005), Piaractus is the sister group of the remaining analyzed members of the family, which include species of Metynnis and Serrasalmus. Both hypotheses are compatible with the one herein obtained, although the synapomorphies of this node should be further evaluated, because it is likely that some of those features are diagnostic of more or less inclusive clades. The fossil genus †Megapiranha was proposed to be the sister group of Pygopristis, Pristobrycon, Pygocentrus, and Serrasalmus (Cione et al., 2009), and is tentatively included at this node. Synapomorphies: 1. Fourth infraorbital (66): (0 > 1) absent or much reduced and bordered posteriorly by third and fifth infraorbitals. Paralleled in the Aphyocharacinae and Gasteropelecidae, and in Aphyodite grammica, Hasemania nana, Hemigrammus erythrozonus, Hoplocharax goethei, Hyphessobrycon pulchripinnis, and Nematobrycon palmeri. 2. Articulation between dentary teeth (145): (0 > 1) present with associated processes and fossae. 3. Predorsal spine formed by first dorsal pterygiophore (275): (0 > 1) present. 4. Number of branched anal-fin rays (288): (0 > 1) 25 or more. Paralleled in the Characidae. 5. Longitudinal position of insertion of adductor mandibulae tendon on dentary (330): (0 > 1) on vertical through middle or anterior half of Meckelian cartilage. Paralleled in the Iguanodectinae, in nodes 184, 209, 241, 261, and 270, and in Engraulisoma taeniatum and Gymnocharacinus bergii. 6. Insertion of adductor mandibulae tendon on dentary (331): (0 > 1) anterior to Meckelian cartilage. Paralleled in nodes 183 and 253 and in Xenagoniates bondi. 7. Number of 2n chromosomes (365): (0 > 1) 58 or more. Autapomorphies of Serrasalmus maculatus: 1. Ventral longitudinal lamellae of basioccipital (2): (1 > 0) falling short of posterior border of basioccipital. Paralleled in nodes 170 and 205. 2. Position of sphenotic spine relative to hyomandibula (11): (0 > 1) displaced anteriorly relative to anterior margin of hyomandibula. Paralleled in nodes 162 and 211 and in Acestrorhynchus pantaneiro, Piabina argentea, and Salminus brasiliensis. 3. Dorsal process of pterotic where tendon from epaxial musculature attach (45): (0 > 1) present, projecting dorsally from tube for semicircular canal. Paralleled in the Heterocharacinae, in node 193, and in Rhoadsia altipinna. 4. Posterior branch of posttemporal laterosensory canal (88): (0 > 1) absent. Paralleled in Characidium rachovii. J. M. Mirande 473 5. Inner row of dentary teeth (143): (0 > 1) absent. Paralleled in the Gasteropelecidae, in nodes 166, 168, and 189, and in Rhaphiodon vulpinus. 6. Ectopterygoid teeth row (159): (0 > 1) present. Paralleled in nodes 168 and 300 and in Acestrorhynchus pantaneiro, Distichodus maculatus, Hoplias cf. malabaricus, and Xenagoniates bondi. 7. Fusion between hypurals 1 and 2 (299): (0 > 1) present. Paralleled in Distichodus maculatus and Hemiodus cf. thayeria. Autapomorphies of Metynnis maculatus: 1. Contact between frontals anteriorly to frontal fontanel (21): (1 > 0) absent. Paralleled in node 167. Some trees: Paralleled in nodes 175 and 206 and in Brycon meeki. 2. Anterior margin of supraoccipital (51): (0 > 1) situated anterior to vertical through posterior orbital margin. Paralleled in node 220 and in Cynopotamus argenteus. 3. Lateral surface of vertical canal of preopercle (81): (1 > 0) canal uncovered and situated posteriorly to musculature and infraorbitals. Paralleled in nodes 170, 175, and 204 and in Chalceus macrolepidotus. 4. Posterior extent of ventral process of quadrate (151): (0 > 1) falling short of posterior margin of symplectic. Paralleled in the Hyphessobrycon luetkenii clade and Iguanodectinae, in nodes 284, 289, and 298, and in Astyanax mexicanus, A. cf. rutilus, Aulixidens eugeniae, Cyphocharax spilotus, Diapoma speculiferum, Hemiodus cf. thayeria, Micralestes stormsi, Moenkhausia sanctaefilomenae, Nematocharax venustus, Probolodus heterostomus, Psellogrammus kennedyi, and Pseudocorynopoma doriae. Some trees: Paralleled in node 302. 5. Cartilages anterior to basihyal (188): (0 > 1) two well developed blocks of cartilage. Paralleled in nodes 244 and 299 and in Hasemania nana, Hyphessobrycon bifasciatus, Odontostilbe microcephala Eigenmann, and Roeboides descalvadensis. Some trees: Paralleled in node 265. 6. Transitional vertebrae with haemal canal (229): (0 > 1) absent. Paralleled in nodes 195 and 212, and in Aulixidens eugeniae, Engraulisoma taeniatum, and Piabina argentea. Some trees: Paralleled in node 247 and in Bryconamericus alpha and Paracheirodon axelrodi. 7. Dorsal-fin rays articulating with first dorsal pterygiophore (266): (0 > 1) three or four. Paralleled in nodes 203 and 276 and in Salminus brasiliensis. Some trees: Paralleled in Brycon orbignyanus. 8. Number of branched anal-fin rays (289): (0 > 1) 35 or more. Paralleled in nodes 207 and 212 and in Gymnocorymbus ternetzi, Piabucus melanostomus, Pseudocorynopoma doriae, Rhaphiodon vulpinus, Stethaprion erythrops, and Thoracocharax stellatus. Some trees: Paralleled in node 261 and in Markiana nigripinnis. Node 179: (-16 / 90 / – / 21) Families Alestidae and Characidae. The family Alestidae was longly considered as related with the Characidae or even included in it (e. g. Greenwood et al., 1966; Weitzman & Malabarba, 1998). In the hypothesis of Buckup (1998), the Alestidae is the sister group of a clade composed of the Acestrorhynchidae, Characidae, and Erythrinoidea. In the phylogenetic hypothesis of Lucena (1993), the Alestidae is included in the Characidae, being related with Brycon and Serrasalmus (the only serrasalmid included in his analysis). Murray & Stewart (2002) proposed both the monophyly and the familial level of the Alestidae. In the phylogeny of Calcagnotto et al. (2005) the African Alestidae are monophyletic and included in a node also containing the Erythrinoidea and Crenuchidae. Zanata & Vari (2005) published a comprehensive phylogeny of the Alestidae including several members of other characiform families in their analysis. The sister-group relationships of the Alestidae, however, were not hypothesized by Zanata & Vari (2005). In the present analysis the Alestidae are proposed to be the sister group of the Characidae. However, this hypothesis has low stability across the different analyses made in this study given that under most conditions, but not in the globally more stable hypotheses, the alestids are instead the sister group of the serrasalmids. Synapomorphies: 1. Form of anterior process of lateral ethmoid (14): (0 > 1) slender and separated from vomer. Paralleled in Distichodus maculatus and Hemiodus cf. thayeria. Reversed in Aulixidens eugeniae, Rhaphiodon vulpinus, and Salminus brasiliensis. 2. Distance between cartilage anterior to orbitosphenoid and lateral ethmoids (38): (1 > 0) contacting, or almost contacting, lateral ethmoids. Paralleled in node 168 and in Hemiodus cf. thayeria and Hoplias cf. malabaricus. Reversed in node 204. 3. Rows of gill rakers on first ceratobranchial (192): (1 > 0) one. Paralleled in node 170. Reversed in the Iguanodectinae, in node 280, and in Brycon orbignyanus, Bryconaethiops macrops, Carlana eigenmanni, Cheirodon interruptus, Hoplocharax goethei, Hyphessobrycon elachys, Odontostilbe microcephala, Parecbasis cyclolepis, Prodontocharax melanotus, and Rhaphiodon vulpinus. Some trees: Reversed in node 249 and in Attonitus ephimeros. 4. Coracoid foramen (243): (0 > 1) well developed. Reversed in node 205 and in Salminus brasiliensis. Some trees: Reversed in node 302 and in Brycon orbignyanus. 5. Relative length of anterior dorsal-fin rays (271): (0 > 1) reaching tip of posterior rays when adpressed. Paralleled in node 163 and in Engraulisoma taeniatum and Pyrrhulina australis. Reversed in Mimagoniates rheocharis. 6. Proximal and medial radials of anal fins (294): (2 > 0) fused on anterior five pterygiophores. Paralleled in Engraulisoma taeniatum. Transformed to state 1 in nodes 184, 208, 218, and 221 and in Psellogrammus kennedyi and Pseudocorynopoma doriae. Some trees: Transformed to state 1 in node 295. Phylogeny of the family Characidae 480 6. Form of anterior portion of ectopterygoid (157): (0 > 1) slender and articulating only to lateral margin of palatine, and lacking ligaments to neurocranium. Paralleled in the Alestidae, in node 170, and in Attonitus ephimeros. 7. Contact between lamella on anterior portion of first basibranchial with lamella on posterior portion of second basibranchial (183): (1 > 0) absent. Reversal of synapomorphy 7 of node 177. 8. Lateral base of gill rakers on first ceratobranchial (199): (1 > 0) slender. Reversal of synapomorphy 9 of node 177. 9. Development of transverse process of neural arch of third vertebra (219): (0 > 1) well developed and extending beyond anterior margin of tripus. Paralleled in node 302, and in Cyanocharax alburnus, Deuterodon langei, Engraulisoma taeniatum, Hemiodus cf. thayeria, Roeboexodon geryi, and Thayeria obliqua. Some trees: Paralleled in Microschemobrycon casiquiare and Parecbasis cyclolepis. 10. Anteriorly directed spine at base of first rib (223): (0 > 1) present. Paralleled in the Heterocharacinae. Some trees: Paralleled in node 302. 11. Posterior margin of cleithrum (235): (0 > 1) with markedly concave margin, almost forming straight angle. Paralleled in nodes 162, 247, and 253 and in Attonitus ephimeros, Characidium borellii, Iguanodectes geisleri, Moenkhausia cf. intermedia, Prionobrama paraguayensis, and Xenagoniates bondi. 12. Dorsal development of third postcleithrum (251): (0 > 1) not projects dorsally to posterior region of scapula. Paralleled in node 192 and in Gymnocharacinus bergii. 13. Articulation between pelvic bones (261): (0 > 1) with bony interdigitations between ischiatic processes. Paralleled in node 302 and in Engraulisoma taeniatum. 14. Number of epurals (296): (1 > 0) one. Paralleled in Prionobrama paraguayensis. 15. Circulii on posterior field of scales (319): (0 > 1) absent. Paralleled in the Gasteropelecidae and in nodes 168 and 206. Some trees: Paralleled in node 302. 16. Posterior attachment of A1 section of adductor mandibulae (332): (0 > 1) restricted or almost restricted to horizontal arm of preopercle. Paralleled in the Iguanodectinae, in node 211, and in Aphyodite grammica and Pyrrhulina australis. Node 174: (100 / 100 / 99 / 73) Subfamilies Acestrorhynchinae and Cynodontinae. The monophyly of a clade composed of the Acestrorhynchinae and Cynodontinae was proposed in several morphological phylogenies (Uj, 1990; Lucena, 1993; Lucena & Menezes, 1998); however, this clade was not resolved as a monophyletic unit in the molecular analyses of Ortí & Meyer (1990) and Calcagnotto et al. (2005). This analysis corroborates a sister-group relationship between these two subfamilies, which is supported by numerous synapomorphies. Synapomorphies: 1. Posteriorly-oriented epioccipital spine (7): (0 > 1) absent. Reversal of synapomorphy 1 of node 177. 2. Dilator fossa on lateral surface of frontal (24): (1 > 0) absent. Paralleled in node 181. 3. Bony lamellae bordering sensory canal of nasal (34): (0 > 1) wider at some point than tubular region. Paralleled in node 181 and in Galeocharax humeralis, Leporinus striatus, and Roeboexodon geryi. 4. Ventral border of rhinosphenoid (50): (0 > 1) almost contacting parasphenoid. Paralleled in node 193. 5. Posterior dorsoventral expansion of fourth infraorbital (68): (0 > 1) present. Paralleled in nodes 210 and 299. 6. Morphology of premaxillary, maxillary, and dentary teeth (118): (1 > 0) all teeth conical, caniniform or mamilliform. Paralleled in the Heterocharacinae, in nodes 181 and 211, and in Axelrodia lindeae, Grundulus cochae, and Exodon paradoxus. 7. A pair of large conical teeth in premaxilla (121): (0 > 1) present. Paralleled in node 299. 8. Anterior extension of interopercle (163): (0 > 1) not extending anteriorly beyond terminus of horizontal arm of preopercle. Paralleled in the Heterocharacinae, in nodes 162 and 212, and in Hoplias cf. malabaricus and Piaractus mesopotamicus. 9. Mesopterygoid teeth (165): (0 > 1) present. 10. Shape of first ceratobranchial gill rakers (197): (0 > 2) short, broad and strongly denticulated. 11. Length of interhyal (211): (0 > 1) equal to or longer than one-half of symplectic length. Paralleled in node 211 and in Hoplias cf. malabaricus and Pseudochalceus kyburzi. 12. Number of branchiostegal rays attached to posterior ceratohyal (217): (0 > 1) two. 13. Position of last supraneural (283): (0 > 1) located more than two vertebrae in front of first dorsal pterygiophore. Paralleled in node 244 and in Engraulisoma taeniatum, Gymnocharacinus bergii, and Xenagoniates bondi. 14. Fusion of hypural 2 to compound centrum (298): (1 > 0) absent. Paralleled in the Serrasalmidae, in node 184, and in Distichodus maculatus, Hemiodus cf. thayeria, and Hoplias cf. malabaricus. 15. Radii on scales (320): (1 > 0) absent or reduced in number. Paralleled in the Iguanodectinae and Serrasalminae and in Cyphocharax spilotus, Distichodus maculatus, Markiana nigripinnis, and Phenagoniates macrolepis. Cynodontinae: Genera Cynodon, Gilbertolus?, Hydrolycus, Rhaphiodon, and Roestes? Hydrolycus and Rhaphiodon (including Cynodon) were included in the characid subfamily Rhaphiodontinae by Géry (1977). Lucena & Menezes (1998) proposed this clade as the sister group of the Acestrorhynchidae, and the familial name Cynodontidae (already used by Greenwood et al., 1966) was given to this clade plus the genera Gilbertolus and Roestes (not analyzed here). The phylogeny of Acestrorhynchus undertaken by Toledo-Piza (2007) J. M. Mirande 481 included the genera of the Cynodontinae as part of the outgroup. In her paper Gilbertolus and Roestes are not closely related with Cynodon, Hydrolycus, and Rhaphiodon, but rather related with other members of the Characidae. Therefore, the position of Gilbertolus and Roestes is currently uncertain. Toledo-Piza (2007), however, argued that not all the known evidence relating these five genera was used in her analysis. Thus, these genera are herein tentatively maintained within the Cynodontinae. The monophyly of the Cynodontinae (as Cynodontidae) was corroborated by Lucena & Menezes (1998) and Toledo-Piza (2000). Autapomorphies of Rhaphiodon vulpinus: 1. Posterior laminar expansion of epiphyseal bar (1): (0 > 1) present. Paralleled in Piabucus melanostomus. 2. Form of anterior process of lateral ethmoid (14): (1 > 0) broad in ventral view, contacting proximal region of vomer in its entire length. Reversal of synapomorphy 1 of node 179. Paralleled in Aulixidens eugeniae and Salminus brasiliensis. 3. Synchondral articulation between lateral ethmoid and anterodorsal border of orbitosphenoid (35): (1 > 0) present. Paralleled in the Aphyocharacinae and in Leporinus striatus, Mimagoniates rheocharis, Pristella maxillaris, and Rhaphiodon vulpinus. 4. Form of orbitosphenoid (37): (0 > 1) massive, almost reaching parasphenoid ventrally. Reversal of synapomorphy 1 of the Characidae. Paralleled in node 193 and in Markiana nigripinnis and Roeboides microlepis. 5. Pores of laterosensory canal of lower jaw (80): (0 > 1) seven or more. Paralleled in nodes 211 and 299. 6. Form of lateral line (89): (1 > 0) approximately straight. Reversal of synapomorphy 2 of the Characoidea. 7. Position of coronomeckelian (110): (0 > 1) situated mainly dorsal to Meckelian cartilage. Reversal of synapomorphy 4 of node 176. 8. Inner row of dentary teeth (143): (0 > 1) absent. Paralleled in the Gasteropelecidae, in nodes 166, 168, and 189, and in Serrasalmus maculatus. 9. Patch of ectopterygoid teeth (160): (0 > 1) present. Paralleled in Hoplias cf. malabaricus. 10. Suprapreopercle (175): (0 > 1) autogenous, separated from preopercle. Paralleled in nodes 210 and 302 and in Markiana nigripinnis and Roeboides microlepis. 11. Articulation between anterior and posterior ceratohyals (181): (0 > 1) with bony interdigitations. Paralleled in node 210 and in Hoplias cf. malabaricus and Salminus brasiliensis. 12. Bony lamellae between second and third basibranchials (184): (1 > 0) absent. Paralleled in the Serrasalmidae and in Attonitus ephimeros, Axelrodia lindeae, Hollandichthys multifasciatus, Hoplocharax goethei, Jupiaba scologaster, Piabucus melanostomus, Pyrrhulina australis, and Xenagoniates bondi. 13. Rows of gill rakers on first ceratobranchial (192): (0 > 1) two. Reversal of synapomorphy 3 of node 179. Paralleled in the Iguanodectinae, in node 280, and in Brycon orbignyanus, Bryconaethiops macrops, Carlana eigenmanni, Cheirodon interruptus, Hoplocharax goethei, Hyphessobrycon elachys, Odontostilbe microcephala, Parecbasis cyclolepis, and Prodontocharax melanotus. Some trees: Paralleled in node 249 and in Attonitus ephimeros. 14. Rows of gill rakers on second ceratobranchial (193): (0 > 1) two. Paralleled in nodes 210, 225, 276, and 297 and in Brycon orbignyanus, Hoplocharax goethei, and Hyphessobrycon elachys. 15. Rows of gill rakers on first epibranchial (203): (0 > 1) two. Reversal of synapomorphy 9 of node 176. 16. Number of branchiostegal rays (213): (0 > 1) five. Paralleled in Characidium borellii, Hoplias cf. malabaricus, Piaractus mesopotamicus, and Thoracocharax stellatus. 17. Anterior portions of branchiostegal rays (214): (0 > 1) slender near their articulation with ceratohyals. Paralleled in nodes 212 and 256. 18. Ascending process of neural pedicle of third vertebra (220): (1 > 0) absent. Reversal of synapomorphy 6 of node 180. 19. Dorsal development of cleithrum (237): (0 > 1) ending in a position just dorsal of tip of mesocoracoid. Paralleled in node 302. 20. Development of medial lamella of coracoid (238): (0 > 1) expanded as a keel. Paralleled in nodes 170 and 302 and in Paragoniates alburnus, Piabucus melanostomus, and Pseudocorynopoma doriae. 21. Second postcleithrum (248): (0 > 1) absent. Paralleled in the Gasteropelecidae, in node 302, and in Pseudocorynopoma doriae. 22. Third postcleithrum (249): (0 > 1) absent. Paralleled in the Gasteropelecidae, in node 302, and in Piabucus melanostomus, Pyrrhulina australis, and Xenagoniates bondi. 23. Ventral exit of laterosensory canal of supracleithrum (254): (0 > 1) ventral to lamella of supracleithrum and exiting on posterior margin of this bone. Paralleled in node 205. 24. Anterior extension of pelvic-bone along main axis (262): (0 > 1) projecting anterior of lateral and medial lamellae of pelvic bone. Paralleled in the Gasteropelecidae, in node 302, and in Hoplias cf. malabaricus, Piabucus melanostomus, and Stethaprion erythrops. 25. Number of dorsal-fin rays on last pterygiophore (272): (0 > 1) two, adnate. 26. Dorsal myorhabdoi (273): (0 > 1) present. Paralleled in node 170. 27. Number of branched anal-fin rays (289): (0 > 1) 35 or more. Paralleled in nodes 207 and 212 and in Gymnocorymbus ternetzi, Metynnis maculatus, Piabucus melanostomus, Pseudocorynopoma doriae, Stethaprion erythrops, and Thoracocharax stellatus. Some trees: Paralleled in node 261 and in Markiana nigripinnis. 28. Scales covering anal-fin base (327): (0 > 1) several rows covering basal third of anal fin. Paralleled in the Serrasalmidae, in nodes 210 and 221, and in Bario steindachneri, Markiana nigripinnis, Paragoniates alburnus, Roeboides microlepis, and Thoracocharax stellatus. 29. Posterior region of levator arcus palatini (337): (0 > 1) limited lateral and medially by A2 and A3 sections of adductor mandibulae. Paralleled in Piabina argentea. Phylogeny of the family Characidae 482 30. Number of 2n chromosomes (363): (0 > 1) 52 or more. Paralleled in the Characinae and Serrasalmidae, in node 196, and in Chalceus macrolepidotus, Hyphessobrycon herbertaxelrodi, and Markiana nigripinnis. Some trees: Paralleled in Hemigrammus unilineatus. 31. Number of 2n chromosomes (364): (0 > 1) 54 or more. Paralleled in the Serrasalmidae. Acestrorhynchinae: Genus Acestrorhynchus The subfamily Acestrorhynchinae was originally proposed to include Acestrorhynchus and Oligosarcus by Menezes (1969). Géry (1977) classified these two genera plus Bramocharax in the tribe Acestrorhynchini, as part of the subfamily Characinae. That author even stated that the Acestrorhynchini was “one of the best-known characid groups” (Géry, 1977: 323). This hypothesis of a close relationship was challenged by Menezes & Géry (1983) and refuted by Buckup (1998). The latter author found that Oligosarcus and Acestrorhynchus are only distantly related. In his phylogeny Oligosarcus is deeply nested in the Characidae, while Acestrorhynchus was not included in the Characidae, but rather in its own family, the Acestrorhynchidae. The Acestrorhynchidae was also considered as valid by Lucena & Menezes (1998) and Menezes (2003), among others. The monophyly of Acestrorhynchus was tested by Toledo-Piza (2007), who obtained a sister-group relationship between this genus and the Cynodontidae as previously proposed by Lucena & Menezes (1998). This clade containing Acestrorhynchus and Rhaphiodon is nested in a group composed of the Agoniatinae and Salmininae, which is the sister group of the Bryconinae. All of these were traditionally included in the Characidae. Consequently, both the Acestrorhynchinae and Cynodontinae are herein proposed to be subfamilies of the Characidae. Autapomorphies of Acestrorhynchus pantaneiro: 1. Position of sphenotic spine relative to hyomandibula (11): (0 > 1) displaced anteriorly relative to anterior margin of hyomandibula. Paralleled in nodes 162 and 211 and in Piabina argentea, Salminus brasiliensis, and Serrasalmus maculatus. 2. Position of sphenotic spine relative to the orbit (12): (0 > 1) distinctly posterior to orbital margin. Paralleled in nodes 193 and 299 and in Attonitus ephimeros, Cynopotamus argenteus, and Gymnocharacinus bergii. 3. Anterior region of laterosensory canal of frontal (83): (0 > 1) opens into a chamber limited. dorsally by frontal and ventrally by lateral ethmoid. 4. Ectopterygoid teeth row (159): (0 > 1) present. Paralleled in nodes 168 and 300 and in Distichodus maculatus, Hoplias cf. malabaricus, Serrasalmus maculatus, and Xenagoniates bondi. 5. Rows of gill rakers on third and fourth ceratobranchials (194): (1 > 0) one. 6. Total number of transitional vertebrae (228): (0 > 1) three or fewer. Paralleled in node 205 and in Characidium rachovii, Cyphocharax spilotus, Micralestes stormsi, and Triportheus pantanensis. 7. Number of branched anal-fin rays (288): (1 > 0) 24 or less. Reversal of synapomorphy 5 of the Characidae. Paralleled in nodes 200, 277, and 300 and in Iguanodectes geisleri. 8. Humeral spot (341): (0 > 1) horizontally-ovate. Paralleled in node 259 and in Brycon orbignyanus, Jupiaba mucronata, and Roeboides microlepis. Some trees: 9. Length of medial bony ridge of opercle (170): (1 > 0) 60% or greater than opercular length. (k11-14). Paralleled in the Serrasalmidae, in node 210, and in Astyanax abramis, Creagrutus cf. taphorni, Hoplias cf. malabaricus, and Roeboides microlepis. Some trees: Paralleled in Salminus brasiliensis. Node 190: Bryconinae (86 / 97 / 1 / 10) Genera Brycon, Chilobrycon Géry & de Rham, Henochilus Garman, Lignobrycon, and Triportheus. The subfamily Bryconinae was considered by Géry (1977) to be composed of the tribes Bryconini, Salminini, and Triportheini. Although a close relationship between Brycon and Triportheus has been suggested several times in the literature (e. g. Géry, 1977), only Malabarba (1998b) obtained evidence supporting this hypothesis. In her phylogeny, Lignobrycon is the sister group of Triportheus, supported by eight synapomorphies. Lima (2003b), mainly based in the analysis of Zanata (2000) restricted the Bryconinae to Brycon, Chilobrycon, and Henochilus, leaving Lignobrycon, Salminus, and Triportheus as incertae sedis within the Characidae (Lima et al., 2003). This subfamily is herein redefined to include the previously incertae sedis genera Lignobrycon and Triportheus. The included species of Brycon form a polytomy in this node, and the monophyly of the genus is not supported in this analysis. However, only a small sample of this genus was analyzed, and a study focused on the question of the monophyly of Brycon lies far beyond the scope of this paper. Synapomorphies: 1. Interdigitations between premaxillae (103): (1 > 0) present. Paralleled in Distichodus maculatus and Piaractus mesopotamicus. 2. Number of rows of premaxillary teeth (123): (0 > 1) three. Paralleled in Bryconaethiops macrops and Chalceus macrolepidotus. 3. Polymorphism of teeth on inner premaxillary row (131): (0 > 1) present, with two medial teeth somewhat larger and usually separated from remaining ones by a gap. Paralleled in the Serrasalmidae. Autapomorphies of Brycon pesu: 1. Frontal fontanel (22): (0 > 1) totally occluded by frontals. Paralleled in Brycinus carolinae. 2. Parietal fontanel (41): (0 > 1) absent in adults. Paralleled in nodes 162, 170, and 181 and in Brycinus carolinae. J. M. Mirande 483 3. Ventral extent of third infraorbital (64): (1 > 0) reaching horizontal arm of preopercle. Reversal of synapomorphy 3 of node 180. Paralleled in the Heterocharacinae and Iguanodectinae, in nodes 184, 198, 225, and 302, and in Agoniates anchovia, Hasemania nana, Markiana nigripinnis, Moenkhausia sanctaefilomenae, Pseudochalceus kyburzi, Roeboexodon geryi, and Stichonodon insignis. Some trees: Reversed in node 292. 4. Number of gill rakers on first hypobranchial and ceratobranchial (195): (0 > 1) 15 or fewer. Reversal of synapomorphy 8 of node 177. 5. Total number of vertebrae (227): (1 > 0) 40 or fewer. Reversal of synapmorphy 4 of the Characidae. Paralleled in nodes 205 and 302. 6. Position of ventral end of posttemporal (253): (1 > 0) anterior or lateral to lateral margin of epioccipital. Reversal of synapomorphy 10 of node 177. 7. Scales covering supraoccipital spine (324): (0 > 1) present and completely covering supraoccipital spine. Paralleled in Prochilodus lineatus. Autapomorphies of Brycon orbignyanus: 1. Rhinosphenoid (47): (1 > 0) absent. Reversal of synapomorphy 2 of the Characidae. Paralleled in nodes 207, 260, 280, and 298 and in Aphyocharax nattereri, Attonitus ephimeros, Bryconamericus scleroparius, Hollandichthys multifasciatus, Pseudocorynopoma doriae, and Salminus brasiliensis. 2. Cusps of teeth on outer premaxillary row (125): (0 > 1) five or more cusps. Paralleled in nodes 265 and 294 and in Bryconops melanurus, Gymnocharacinus bergii, Micralestes stormsi, and Nematocharax venustus. Some trees: Paralleled in Bryconamericus agna. 3. Number of teeth in inner premaxillary row (130): (0 > 1) eight or more. Paralleled in the Aphyoditeinae, in node 170, and in Grundulus cochae, Phenacogaster tegatus, Prionobrama paraguayensis, and Salminus brasiliensis. 4. Number of cusps of anterior maxillary teeth (139): (0 > 1) five or more cusps. Paralleled in the Rhoadsiinae, in nodes 273, 283, and 294, and in Bramocharax bransfordii, Gymnocharacinus bergii, Hemibrycon dariensis, Hyphessobrycon pulchripinnis, and Odontostoechus lethostigmus. Some trees: Paralleled in node 246. 5. Ventral margin of anterior ceratohyal (179): (0 > 1) with notches for articulation of branchiostegal rays. Paralleled in nodes 167 and 189 and in Micralestes stormsi and Salminus brasiliensis. 6. Rows of gill rakers on first ceratobranchial (192): (0 > 1) two. Reversal of synapomorphy 3 of node 179. Paralleled in the Iguanodectinae, in node 280, and in Bryconaethiops macrops, Carlana eigenmanni, Cheirodon interruptus, Hoplocharax goethei, Hyphessobrycon elachys, Odontostilbe microcephala, Parecbasis cyclolepis, Prodontocharax melanotus, and Rhaphiodon vulpinus. Some trees: Paralleled in node 249 and in Attonitus ephimeros. 7. Rows of gill rakers on second ceratobranchial (193): (0 > 1) two. Paralleled in nodes 210, 225, 276, and 297 and in Hoplocharax goethei, Hyphessobrycon elachys, and Rhaphiodon vulpinus. 8. Humeral spot (341): (0 > 1) horizontally-ovate. Paralleled in node 259 and in Acestrorhynchus pantaneiro, Jupiaba mucronata, and Roeboides microlepis. Some trees: 9. Length of supraoccipital spine (53): (1 > 0) extends posteriorly to, at least, middle length of neural complex of Weberian apparatus. (k11-14). Paralleled in the Serrasalmidae, in nodes 182 and 205, and in Piabucus melanostomus and Prochilodus lineatus. Some trees: Paralleled in Brycon meeki. 10. Coracoid foramen (243): (1 > 0) absent or reduced to small pore. (k11-14). Reversal of synapomorphy 4 of node 179. Paralleled in node 205 and in Salminus brasiliensis. Some trees: Paralleled in node 302. 11. Dorsal-fin rays articulating with first dorsal pterygiophore (266): (0 > 1) three or four. (k9-10). Paralleled in nodes 203 and 276 and in Metynnis maculatus and Salminus brasiliensis. 12. Anterior extension of adductor arcus palatini (336): (1 > 0) covering most of dorsal surface of mesopterygoid. (k9-10). Paralleled in node 166 and in Creagrutus anary, Markiana nigripinnis, and Salminus brasiliensis. Autapomorphies of Brycon meeki: 1. Temporal fossa (13): (1 > 0) well developed. Paralleled in node 300 and in Bryconexodon juruenae and Salminus brasiliensis. 2. Number of epurals (297): (0 > 1) three. Paralleled in the Alestidae and in node 167. Some trees: 3. Contact between frontals anteriorly to frontal fontanel (21): (1 > 0) absent. (k11-14). Paralleled in node 167 and in Metynnis maculatus. Some trees: Paralleled in nodes 175 and 206. 4. Length of supraoccipital spine (53): (1 > 0) extends posteriorly to, at least, middle length of neural complex of Weberian apparatus. (k11-14). Paralleled in the Serrasalmidae, in nodes 182 and 205, and in Piabucus melanostomus and Prochilodus lineatus. Some trees: Paralleled in Brycon orbignyanus. Autapomorphy of Brycon falcatus: 1. Laterosensory canal in antorbital (72): (0 > 1) present. Paralleled in the Bryconops clade and in Chalceus macrolepidotus and Iguanodectes geisleri. Node 302: (100 / 100 / 100 / 47) Genera Lignobrycon and Triportheus. The monophyly of Triportheus was generally accepted even before the cladistic analysis of Malabarba (1998b), who identified four synapomorphies for Triportheus and proposed Lignobrycon as the sister group of the genus. Some of the synapomorphies found here for the two analyzed species of Triportheus could be applicable to more or less restricted clades within this genus, or to the node composed of Lignobrycon plus Triportheus. Synapomorphies: 1. Ventral projection of lagenar capsule (3): (0 > 1) extending ventrally to articulation between basioccipital and parasphenoid. Paralleled in the Serrasalmidae, in nodes 170 and 205, and in Cyphocharax spilotus and Micralestes stormsi. Phylogeny of the family Characidae 484 2. Opening between orbitosphenoid and pterosphenoid (39): (1 > 0) present, rounded or ovate, usually margined by frontal dorsally. Paralleled in node 205 and in Thoracocharax stellatus. 3. Position of antorbital relative to lateral ethmoid in lateral view (56): (0 > 1) antorbital overlapping lateral ethmoid. Paralleled in the Gasteropelecidae. 4. Lateral overlap of first infraorbital by anterior margin of second infraorbital (60): (0 > 1) present. Paralleled in the Bryconops clade. 5. Ventral extent of third infraorbital (64): (1 > 0) reaching horizontal arm of preopercle. Reversal of synapomorphy 3 of node 180. Paralleled in the Heterocharacinae and Iguanodectinae, in nodes 184, 198, and 225, and in Agoniates anchovia, Brycon pesu, Hasemania nana, Markiana nigripinnis, Moenkhausia sanctaefilomenae, Pseudochalceus kyburzi, Roeboexodon geryi, and Stichonodon insignis. Some trees: Reversed in node 292. 6. Form of fourth infraorbital (67): (0 > 1) longer dorsoventrally than longitudinally. Paralleled in node 189. 7. Branching of laterosensory canals of fourth or fifth infraorbitals (74): (1 > 0) absent. Reversal of synapomorphy 3 of node 177. 8. Epiphyseal branch of corresponding supraorbital canals (85): (0 > 1) oriented obliquely, opening posteriorly to epiphyseal bar. Paralleled in Heterocharax macrolepis. 9. Number of maxillary teeth (136): (1 > 0) up to three. Reversal of synapomorphy 5 of node 177. 10. Extent of implantation of teeth along maxilla (137): (1 > 0) not reaching middle of maxillary lamella. Reversal of synapomorphy 6 of node 177. 11. Anterodorsal lobe of metapterygoid oriented towards mesopterygoid (166): (0 > 1) present, conspicuous and anteriorly oriented. 12. Relative length of palatine (172): (0 > 1) distinctly longer than one-half length of ectopterygoid. Paralleled in nodes 197 and 261 and in Hyphessobrycon pulchripinnis and Paracheirodon axelrodi. 13. Suprapreopercle (175): (0 > 1) autogenous, separated from preopercle. Paralleled in node 210 and in Markiana nigripinnis, Roeboides microlepis, and Rhaphiodon vulpinus. 14. Bony lamella dorsal to fourth basibranchial (185): (1 > 0) present. Paralleled in nodes 168, 170, and 203 and in Phenacogaster tegatus. 15. Edentulous basihyal lamella (189): (0 > 1) present. Paralleled in the Alestidae and Heterocharacinae, in node 175, and in Bryconops affinis and Carnegiella strigata. 16. Anterior development of basihyal (190): (1 > 0) broadly extending beyond anterior margin of hypohyals. Paralleled in node 170. 17. Distribution of denticles on gill rakers (202): (1 > 0) restricted to margins, or absent. 18. Attachment of first branchiostegal ray (215): (0 > 1) posterior to one-half length of anterior ceratohyal. 19. Development of transverse process of neural arch of third vertebra (219): (0 > 1) well developed and extending beyond anterior margin of tripus. Paralleled in Agoniates anchovia, Cyanocharax alburnus, Deuterodon langei, Engraulisoma taeniatum, Hemiodus cf. thayeria, Roeboexodon geryi, and Thayeria obliqua. Some trees: Paralleled in Microschemobrycon casiquiare and Parecbasis cyclolepis. 20. Total number of vertebrae (227): (1 > 0) 40 or fewer. Reversal of synapmorphy 4 of the Characidae. Paralleled in node 302 and in Brycon pesu. 21. Medial laminar expansion at dorsal tip of cleithrum (236): (0 > 1) present. 22. Dorsal development of cleithrum (237): (0 > 1) ending in a position just dorsal of tip of mesocoracoid. Paralleled in Rhaphiodon vulpinus. 23. Development of medial lamella of coracoid (238): (0 > 1) expanded as a keel. Paralleled in node 170 and in Paragoniates alburnus, Piabucus melanostomus, Pseudocorynopoma doriae, and Rhaphiodon vulpinus. 24. Second postcleithrum (248): (0 > 1) absent. Paralleled in the Gasteropelecidae and in Pseudocorynopoma doriae and Rhaphiodon vulpinus. 25. Third postcleithrum (249): (0 > 1) absent. Paralleled in the Gasteropelecidae and in Piabucus melanostomus, Pyrrhulina australis, Rhaphiodon vulpinus, and Xenagoniates bondi. 26. Number of branched pelvic-fin rays (258): (1 > 0) six or less. Paralleled in the Aphyocharacinae, in nodes 220, 236, and 280, and in Axelrodia lindeae, Cheirodon interruptus, Cyanocharax alburnus, Hollandichthys multifasciatus, Hoplocharax goethei, and Hyphessobrycon luetkenii. Some trees: Paralleled in Hasemania nana and Hyphessobrycon elachys. 27. Articulation between pelvic bones (261): (0 > 1) with bony interdigitations between ischiatic processes. Paralleled in Agoniates anchovia and Engraulisoma taeniatum. 28. Anterior extension of pelvic-bone along main axis (262): (0 > 1) projecting anterior of lateral and medial lamellae of pelvic bone. Paralleled in the Gasteropelecidae and in Hoplias cf. malabaricus, Piabucus melanostomus, Rhaphiodon vulpinus, and Stethaprion erythrops. 29. Position of anteriormost epineurals (274): (0 > 1) reaching to cranium. Paralleled in nodes 170 and 175 and in Distichodus maculatus and Piabucus melanostomus. Some trees: 30. Overlap of maxilla by second infraorbital (61): (1 > 0) absent. (k9-10). Paralleled in the Alestidae and in node 205. 31. Alignment of cusps of medial teeth on inner premaxillary row (127): (1 > 0) forming anteriorly concave semicircle from ventral view. (k9-10). Paralleled in nodes 183 and 262 and in Moenkhausia dichroura. 32. Posterior extent of ventral process of quadrate (151): (0 > 1) falling short of posterior margin of symplectic. (k 9-10). Paralleled in the Hyphessobrycon luetkenii clade and Iguanodectinae, in nodes 284, 289, and 298, and in Astyanax mexicanus, A. cf. rutilus, Aulixidens eugeniae, Cyphocharax spilotus, Diapoma speculiferum, Hemiodus cf. thayeria, Metynnis maculatus, Micralestes stormsi, Moenkhausia sanctaefilomenae, Nematocharax venustus, Probolodus heterostomus, Psellogrammus kennedyi, and Pseudocorynopoma doriae. 33. Anteriorly directed spine at base of first rib (223): (0 > 1) present. (k9-10). Paralleled in the Heterocharacinae and in Agoniates anchovia. J. M. Mirande 485 34. Base of second pectoral ray (231): (0 > 1) similar in form and size to base of posterior rays. (k9-10). Paralleled in node 175 and in Engraulisoma taeniatum. 35. Coracoid foramen (243): (1 > 0) absent or reduced to small pore. (k11-14). Reversal of synapomorphy 4 of node 179. Paralleled in node 205 and in Salminus brasiliensis. Some trees: Paralleled in Brycon orbignyanus. 36. Circulii on posterior field of scales (319): (0 > 1) absent. (k9-10). Paralleled in the Gasteropelecidae, in nodes 168 and 206, and in Agoniates anchovia. 37. Radii oriented towards anterior field of scales (321): (2 > 0) present. (k9-10). Reversal of synapomorphy 9 of the Characidae. 38. Radii of scales (322): (0 > 1) converging at focus. (k9-10). Reversal of synapomorphy 7 of the Characidae. Paralleled in node 273 and in Stichonodon insignis and Tetragonopterus argenteus. Some trees: Paralleled in Microschemobrycon casiquiare. Autapomorphy of Triportheus pantanensis: 1. Total number of transitional vertebrae (228): (0 > 1) three or fewer. Paralleled in node 205 and in Acestrorhynchus pantaneiro, Characidium rachovii, Cyphocharax spilotus, and Micralestes stormsi. No autapomorphies found for Triportheus nematurus. Node 189: (100 / 100 / 61 / 20) Subfamilies Aphyocharacinae, Aphyoditeinae, Characinae, Cheirodontinae, Gymnocharacinae, Heterocharacinae, Iguanodectinae, Rhoadsiinae, Stethaprioninae, Stevardiinae, and Tetragonopterinae; Astyanax clade, Astyanax paris clade, Bramocharax clade, Bryconamericus scleroparius clade, Bryconops clade, Hyphessobrycon anisitsi clade, and Pseudochalceus clade. This node is congruent with clade 15 of Buckup (1998), which included the genera Bryconops, Charax, Cynopotamus, Oligosarcus, Phenacogaster, and Tetragonopterus. According to that author, this clade is supported by two synapomorphies: the frontal expanded lateral to the supraorbital laterosensory canal and the absence of an inner dentary row of teeth. The former of these synapomorphies is related with the presence and size of the supraorbital bone, and it is coded differently in this study. The second synapomorphy is corroborated here. Synapomorphies: 1. Form of fourth infraorbital (67): (0 > 1) longer dorsoventrally than longitudinally. Paralleled in node 302. Reversed in nodes 200, 210, 228, 277, and 282. 2. Inner row of dentary teeth (143): (0 > 1) absent. Paralleled in the Gasteropelecidae, in nodes 166 and 168, and in Rhaphiodon vulpinus and Serrasalmus maculatus. Reversed in the Heterocharacinae, in node 276, and in Aphyocharacidium bolivianum. 3. Ventral margin of anterior ceratohyal (179): (0 > 1) with notches for articulation of branchiostegal rays. Paralleled in node 167 and in Brycon orbignyanus, Micralestes stormsi, and Salminus brasiliensis. Reversed in node 211 and in Stichonodon insignis. 4. Form and articulation of neural pedicle of third vertebra (218): (0 > 1) pedicle much smaller and without an articular surface with neural complex. 5. Dorsal development of dorsal process of neural pedicle of third vertebra (221): (1 > 0) not broadly overlapping neural complex. Reversed in node 212. 6. Posterior margin of cleithrum (234): (0 > 1) with concavity ventral to first postcleithrum. Paralleled in Micralestes stormsi. Reversed in node 204. Node 188: Iguanodectinae (100 / 100 / 100 / 57) Genera Iguanodectes and Piabucus. The monophyly of the Iguanodectinae was proposed by Vari (1977), based in three characters of the gas-bladder; several new synapomorphies are added in this paper, although some of them could correspond to inner nodes within this subfamily. The monophyly of the Iguanodectinae was corroborated by Moreira (2002), who also studied the composition and internal relationships of this clade. Synapomorphies: 1. Ventral extent of third infraorbital (64): (1 > 0) reaching horizontal arm of preopercle. Reversal of synapomorphy 3 of node 180. Paralleled in the Heterocharacinae, in nodes 184, 198, 225, and 302, and in Agoniates anchovia, Brycon pesu, Hasemania nana, Markiana nigripinnis, Moenkhausia sanctaefilomenae, Pseudochalceus kyburzi, Roeboexodon geryi, and Stichonodon insignis. Some trees: Reversed in node 292. 2. Lateral coverage of dilator fossa by sixth infraorbital (69): (0 > 1) leaving a conspicuous naked area in anterior region of dilator fossa. Paralleled in the Iguanodetinae, in node 197, and in Charax stenopterus, Hoplocharax goethei, Phenacogaster tegatus, and Psellogrammus kennedyi. 3. Laterosensory canal of sixth infraorbital (76): (1 > 0) not branched. Paralleled in node 203 and in Charax stenopterus, Cyphocharax stellatus, Micralestes stormsi, and Phenacogaster tegatus. 4. Posterior extent of maxilla (99): (1 > 0) not reaching second infraorbital. 5. Medial process of dentary bordering Meckelian cartilage dorsally and medially (115): (0 > 1) present. 6. Posterior extent of ventral process of quadrate (151): (0 > 1) falling short of posterior margin of symplectic. Paralleled in the Hyphessobrycon luetkenii clade, in nodes 284, 289, and 298, and in Astyanax mexicanus, A. cf. rutilus, Aulixidens eugeniae, Cyphocharax spilotus, Diapoma speculiferum, Hemiodus cf. thayeria, Metynnis maculatus, Micralestes stormsi, Moenkhausia sanctaefilomenae, Nematocharax venustus, Probolodus heterostomus, Psellogrammus kennedyi, and Pseudocorynopoma doriae. Some trees: Paralleled in node 302. [Document text truncated for crawler view.]