Morphology-based phylogeny of Eigenmanniinae Mago-Leccia, 1978 (Teleostei: Gymnotiformes: Sternopygidae), with a new classification

1 INTRODUCTION

Sternopygidae comprises seven genera and 55 valid species of glass knifefishes included in Sternopyginae and Eigenmanniinae (Ferraris et al., 2017; Fricke et al., 2020). The family is characterized by fishes exhibiting multiple rows of teeth in both jaws, enlarged lateral line canals, and that produce monophasic weak electric organ discharges, ranging from 100 to 780 Hz (Crampton & Albert, 2006; Mago-Leccia, 1978). As also characteristic for other Gymnotiformes, glass knifefishes use their electrogenic-electrosensorial system to communicate and navigate in a variety of aquatic environments in the Neotropical region (Moller, 1995), including floodplains, rapids, river channels, and subterranean drainages (Crampton, 2007; Peixoto et al., 2015; Triques, 1996). Within Sternopygidae, the subfamily Sternopyginae (sensu Albert, 2001) was created to include Sternopygus Müller & Troschel, and †Humboldtichthys (Gayet & Meunier) [formerly proposed as Ellisella Gayet & Meunier, preoccupied by Ellisella Gray, 1858 (Cnidaria)]. In turn, Eigenmanniinae (sensu Meunier et al., 2011) was raised by Mago-Leccia (1978) to comprises five valid living genera (Archolaemus Korringa, Distocyclus Mago-Leccia, Eigenmannia Jordan & Evermann, Japigny Meunier, Jegú & Keith, and Rhabdolichops Eigenmann & Allen—Figure 1), with 45 species and one incertae sedis species—“Eigenmannia” goajira Schultz (Table 1; Ferraris et al., 2017).

Eigenmanniinae has been historically characterized by the presence of a scapular foramen entirely included within the scapula, the fusion of the post-temporal and the supracleithrum into a single ossification, and the presence of 11 to 15 precaudal vertebrae (e.g., Lundberg & Mago-Leccia, 1986; Mago-Leccia, 1978; Vari et al., 2012—Table S1). Recently, Tagliacollo et al. (2016b) proposed additional synapomorphies for Eigenmanniinae (Table S1).

Mago-Leccia (1978) proposed the first phylogenetic hypothesis on interrelationships within Eigenmanniinae (Figure 2). Therein, Rhabdolichops is the sister group of all remaining genera based on the retention of putative plesiomorphic conditions, such as the presence of ossified first basibranchial, larger gill rakers on first gill arch, and of posttemporal fossa. Mago-Leccia’s hypothesis supported a clade compounded by Archolaemus, Distocyclus, and Eigenmannia characterized by sharing cartilaginous gill rakers, mouth and gill opening reduced, and a variable position of the anus. Fink and Fink (1981) discussing the interrelationships within Sternopygidae based on Korringa (1970) and Mago-Leccia (1978), offered an alternative hypothesis, considering Archolaemus as the sister group of the remaining Eigenmanniinae, and Rhabdolichops as the sister group of Eigenmannia + Distocyclus. The monophyly of Eigenmannia, Distocyclus, and Rhabdolichops would be supported by the presence of a reduced number of pleural ribs, length of the anterior two to three pleural ribs subequal to the depth of the abdominal cavity, the fusion of the two posterior pectoral proximal radials, and a subcutaneous eye (Fink & Fink, 1981; Lundberg & Mago-Leccia, 1986). Conversely, Triques (1993) rejected the monophyly of Eigenmanniinae based on Rhabdolichops’s position as the sister group of all Sternopygids (including Sternopygus), sustained by the forward displacement of maxilla, and the elongation of palatine and maxillary cartilages. Nevertheless, Triques (1993) emphasized that his hypothesis was incongruent with that of Lundberg and Mago-Leccia (1986), and it would require further analysis.

Alves-Gomes et al. (1995) provided the first Gymnotiformes’ hypothesis based on molecular evidence (mitochondrial DNA—mtDNA). The monophyly of Sternopygidae was not recovered with Sternopygus in an uncertain position within Gymnotiformes, and Eigenmanniinae (therein, Eigenmanniidae) as sister group of Apteronotidae. Differently, the relationships of Distocyclus, Eigenmannia, and Rhabdolichops were corroborated as in Fink and Fink (1981), where Rhabdolichops was considered the sister group of Distocyclus plus Eigenmannia. It is noteworthy that Alves-Gomes et al. (1995) did not include Archolaemus in their analyses. A year later, Albert and Fink (1996) also corroborated the hypothesis of Fink and Fink (1981) but Eigenmannia emerged as an unnatural group. Alves-Gomes (1998) used mtDNA to investigate the position of Archolaemus and his results corroborated those of Mago-Leccia (1978), in which Archolaemus emerged as the sister group of Distocyclus plus Eigenmannia, and Rhabdolichops as the sister group of that clade. Like in Alves-Gomes et al. (1995), Alves-Gomes (1998) did not recover the monophyly of Sternopygidae.

Albert and Campos-da-Paz (1998) proposed Rhabdolichops as the sister group of Eigenmannia based on the presence of a short snout, curved frontally in lateral view, the anterior displaced hemal spine as large and straight as the remaining hemal spines, and the lengthy pterygiophores of the anal fin. Eigenmannia virecens, was the sole species of the genus included in the analysis, and none autapomorphy was found to define the genus. A few years later, Albert (2001) added supplementary taxa and characters in the matrix of Albert and Campos-da-Paz (1998) and the monophyly of the clade Eigenmannia plus Rhabdolichops was corroborated by the presence of two additional synapomorphies: fifth epibranchial with a short ascending process, and developmental origin of an adult electric organ from portions of both the hypaxial and anal-fin pterygiophore muscles. The monophyly of Eigenmannia was tested by the inclusion of E. humboldtii in the analysis of Albert (2001) but rejected because of the close relationship between the E. virescens clade and Rhabdolichops, rather than with other members of Eigenmannia. An elongated body, presence of 160 to 199 anal-fin rays, and anal-fin pterygiophores longer than hemal spines supported this relationship. Hulen et al. (2005) during the investigation of the species-level relationships in Sternopygus, corroborated Albert’s (2001) hypothesis of relationships of Eigenmanniinae. Later, Albert and Fink (2007) working on the phylogenetic position of Humboldtichthys also supported Albert (2001) hypothesis.

In the year following the description of Japigny (Meunier et al., 2011), Vari et al. (2012) discussed its phylogenetic position and suggested it to be the sister group of all Eigenmanniinae primarily based on the absence of three synapomorphies that are present in all other genera of the subfamily: the angulo-articular with a distinct socket to receive the condyle of the quadrate; the parapophysis of the second vertebra straight and contacting the parapophysis of the fourth vertebra; and the parapophysis of the fourth vertebra curved ventrally. Tagliacollo et al. (2016a) rejected such hypothesis and proposed Japigny as a member of Eigenmannia despite the lack of genetic information of this taxon. Also, they recovered Archolaemus as the sister group of Distocyclus, as proposed by Triques (1993), and Rhabdolichops as the sister group of all Eigenmanniinae, as proposed by Mago-Leccia (1978) and Alves-Gomes (1995). The monophyly of Eigenmannia was recovered by the first time by Alda et al. (2019) based on the analyses of ultra-conserved elements; however, the authors did not make comments on these results.

Contrasting to its taxa delimitations and relationships (Figure 2), the monophyly of Eigenmanniinae has proven to be well supported by both morphological and molecular datasets (Albert, 2001; Albert & Campos-da-Paz, 1998; Albert & Fink, 1996; Alda et al., 2019; Alves-Gomes, 1998; Alves-Gomes et al., 1995; Fink & Fink, 1981; Hulen et al., 2005; Lundberg & Mago-Leccia, 1986; Mago-Leccia, 1978; Tagliacollo et al., 2016a; Triques, 1993). Thus, by examining most of its species diversity and using characters of external anatomy, osteology, myology, and neuroanatomy we​ tested the phylogenetic relationships in Eigenmanniinae and redifined its classification.

2 MATERIAL AND METHODS

3 RESULTS

3.1 Character description

The characters described and discussed below are arranged by morphological complexes following whenever possible overall an anterior to posterior pattern. Each character entry includes the name of the character, description of character states and, the consistency and retention indices on the final phylogenetic hypothesis. In the Neuroanatomy section, numbers after each structure name correspond to those indicated in Figure 45. Unutilized characters that present great variations were not included in the analysis. Explanations for inapplicability of each character are discussed in Data S1.

3.1.1 Quantitative characters

• 1. Number of premaxillary teeth: (0) minimum eight; (1) maximum 75 (CI = 0.201, RI = 0.372).

This character is inapplicable in examined members of Hypopomidae due to absence of teeth in adult of this species.

• 2. Number of teeth rows on premaxilla: (0) minimum one; (1) maximum nine (CI = 0.286, RI = 0.535).

This character is inapplicable in examined members of Hypopomidae (see character 1).

• 3. Number of dentary teeth: (0) minimum one; (1) maximum 88 (CI = 0.247, RI = 0.505).

Although D. conirostris has been illustrated with a toothless dentary (i.e., Hulen et al., 2005: fig. 12i), all examined specimens herein examined have at least one dentary tooth (see also Dutra et al., 2014). This character is inapplicable in examined members of Hypopomidae (see character 1).

• 4. Number of teeth rows on dentary: (0) minimum one; (1) maximum six (Albert, 2001, char. 39; Albert & Campos-da-Paz, 1998, char. 40) (CI = 0.227, RI = 0.514).

This character is inapplicable in examined species of Hypopomidae (see character 1).

• 5. Number of endopterygoid teeth: (0) minimum two; (1) maximum 20 (Albert, 2001, char. 133 in part; Albert & Campos-da-Paz, 1998, char. 134 in part; CI = 0.266, RI = 0.242).

This character is inapplicable in Apteronotidae, Gymnotidae, Hypopomidae, D. conirostris, R. caviceps, R. troscheli, and R. zareti and examined species of Apteronotidae, Gymnotidae, Hypopomidae, which have no endopterygoid teeth.

• 6. Number of gill rakers on first branchial arch: (0) minimum seven; (1) maximum 35 (CI = 0.469, RI = 0.446).
• 7. Number of teeth on upper pharyngeal plate: (0) minimum five; (1) maximum 38 (CI = 0.250, RI = 0.484).
• 8. Number of teeth on lower pharyngeal plate: (0) minimum six; (1) maximum 24 (CI = 0.197, RI = 0.361).
• 9. Number of pectoral-fin rays: (0) minimum 11; (1) maximum 24 (Hulen et al., 2005, char. 56; CI = 0.290, RI = 0.600).

Pectoral-fin rays’ counts include both unbranched and branched rays.

• 10. Number of anal-fin rays: (0) minimum 132; (1) maximum 302 (Albert, 2001, char. 196; Albert & Campos-da-Paz, 1998, char. 197; Albert & Fink, 1996, char. 48; Correa et al., 2006, char. 38; Hulen et al., 2005, char. 64) (CI = 0.225, RI = 0.349).

Anal-fin rays’ counts include both unbranched and branched rays.

• 11. Number of precaudal vertebrae: (0) minimum 11; (1) maximum 44 (Albert, 2001, char. 205–206; Albert & Campos-da-Paz, 1998, char. 206–207; Albert & Fink, 1996, char. 40–41; Correa et al., 2006, char. 34; Hulen et al., 2005, char. 62; Lundberg & Mago-Leccia, 1986, char. 5 in part; Mago-Leccia, 1978; CI = 0.624, RI = 0.739).

Precaudal vertebrae count includes the fourth vertebrae of Weberian complex and all vertebrae that lacks hemal spines.

• 12. Number of transitional vertebrae: (0) minimum one; (1) maximum 11 (Albert & Campos-da-Paz, 1998, char. 208; Albert, 2001, char. 207) (CI = 0.321, RI = 0.424).

Transitional vertebrae are recognized by the precaudal vertebrae that not belongs to Weberian apparatus and lacks pleural ribs.

• 13. Number of pleural ribs: (0) minimum three; (1) maximum 37 (Albert, 2001, char. 209; Albert & Campos-da-Paz, 1998, char. 210; Albert & Fink, 1996, char. 42; Correa et al., 2006, char. 35; Hulen et al., 2005, char. 60; Lundberg & Mago-Leccia, 1986, char. 5 in part and 13; CI = 0.499, RI = 0.613).
• 14. Number of displaced hemal spines: (0) minimum one; (1) maximum five (Albert, 2001, char. 191; Albert & Campos-da-Paz, 1998, char. 192; Correa et al., 2006, char. 40; Lundberg & Mago-Leccia, 1986, char. 21; CI = 0.250, RI = 0.250).

Displaced hemal spines are the rib-like bone on the ventral body wall, anterior and sometimes attached to the first hemal spine (e.g., Albert, 2001).

• 15. Number of lateral line scales: (0) minimum 77; (1) maximum 225 (CI = 0.450, RI = 0.602).
• 16. Number of scale rows above lateral line: (0) minimum four; (1) maximum 25 (CI = 0.324, RI = 0.521).

This character is inapplicable in Rhabdolichops due to the absence of scales on the anterodorsal portion of body.

• 17. Number of visible rows of electrocytes over end of anal-fin base: (0) minimum one; (1) maximum nine (Correa et al., 2006; char. 41; Lundberg & Mago-Leccia, 1986, char. 27 in part; CI = 1.000, RI = not calculated).

The deepness of the electric organ can be expressed by the number of electrocytes organized over the end of anal-fin base (Lundberg & Mago-Leccia, 1986: 67). When present, rows of visible electrocytes over the end of anal-fin base range from one to nine in Rhabdolichops. The codification in all the other analyzed species is uncertain because electrocytes are not externally visible.

3.1.2 Mouth

• 18. Position of mouth: (0) superior (Figure 3a); (1) terminal (Figure 3b); (2) subterminal (Figure 3c; Albert, 2001, char. 21; Albert & Campos-da-Paz, 1998, char. 21–22; Correa et al., 2006, char. 7; Lundberg & Mago-Leccia, 1986, char. 26 and 28; Tagliacollo et al., 2016b; char. 8; CI = 0.222; RI = 0.611).

The superior mouth is defined by the lower jaw distinctly longer than upper jaw. On the contrary, in the terminal mouth both jaws have approximately the same length. As for the subterminal mouth, upper jaw is distinctly longer than lower jaw.

• 19. Extent of mouth gape: (0) shorter than its width (Figure 3a–b); (1) longer than its width (Figure 3c; Albert, 2001, char. 19; Albert & Campos-da-Paz, 1998, char. 19–20; Albert & Fink, 1996, char. 3; Hulen et al., 2005, char. 14; Tagliacollo et al., 2016b, char. 6–7; Vari et al., 2012; CI = 0.500; RI = 0.800).

The extension of mouth gape is treated here as a ratio between the gape length and mouth width, in which mouth gape was measured from the tip of snout to the mouth rictus, whereas the mouth width is the distance between contralateral rictus.

• 20. Upper lip: (0) without teeth; (1) with teeth (CI = 1.000; RI = AUT).

A toothed upper lip presents well-separated extra teeth located just anterior to the premaxilla (see Peixoto et al., 2019: fig. 14).

• 21. Oral valve: (0) without teeth; (1) with teeth (Figure 4; Vari et al., 2012; CI = 1.000; RI = AUT).

A toothed oral valve presents teeth on the soft tissue located at the anterior portion of the vomer.

3.1.3 Nares

• 22. Internarial distance: (0) equivalent to at least two times the diameter of posterior nostril; (1) equivalent diameter of posterior nostril (Figure 3d; Dutra et al., 2014; CI = 1.000; RI = AUT).

Short internarial distance is a reflect of the anterior position of posterior nostril, and the horizontally oriented nasal bone that shifts anterior nostril to a more posterior position (Dutra et al., 2014), a condition present only in D. conirostris (state 1—Figure 3d).

3.1.4 Orbital region

• 23. Orbital rim: (0) eye subcutaneous, attached to orbital rim; (1) orbital rim free (Figure 3c; Albert, 2001, char. 104; Albert & Campos-da-Paz, 1998, char. 105; Albert & Fink, 1996, char. 14; Correa et al., 2006, char. 6; Fink & Fink, 1981; Hulen et al., 2005, char. 16; Korringa, 1970; Lundberg & Mago-Leccia, 1986, char. 16; Mago-Leccia & Zaret, 1978; Vari et al., 2012; CI = 0.500; RI = 0.857).

3.1.5 Upper jaw

• 24. Relative length of premaxilla: (0) twice or less as long as its width (Figure 5a–c); (1) four times or more as long as its width (Figure 5d; Albert, 2001, char. 25 in part; Albert & Campos-da-Paz, 1998, char. 26; Lundberg & Mago-Leccia, 1986, char. 20 in part; Mago-Leccia, 1978; Tagliacollo et al., 2016b, char. 39; CI = 1.000; RI = 1.000).

The relative length of premaxilla is treated here as a ratio between the premaxilla length and premaxilla width, in which its length corresponds to the distance between medial and lateral margins, whereas its width corresponds to the distance between anterior and posterior margins.

• 25. Relative width of premaxilla: (0) half or less than half of its length (Figure 5b–d); (1) as long as its length (Figure 5a; CI = 0.333; RI = 0.667).

The relative width of premaxilla is treated here as a ratio between the premaxilla width and premaxilla length. The condition in Japigny was misinterpreted by Dutra et al. (2014), who considered the presence of a broad premaxilla in the genus.

• 26. Anterolateral process on premaxilla: (0) absent (Figure 5a,d); (1) present (Figure 5b–c; CI = 0.250; RI = 0.883).
• 27. Teeth on both jaws in adults: (0) present; (1) absent. (Albert, 2001, char. 22; Albert & Campos-da-Paz, 1998, char. 24; Tagliacollo et al., 2016b, char. 36; Triques, 1993, char. 23 and 24; CI = 1.000, RI = 1.000).

The absence of teeth on both jaws has been historically proposed as a synapomorphy of Rhamphichthyoidea (state 1; e.g., Albert, 2001; Tagliacollo et al., 2016b). Recently, Sullivan et al. (2013) described two Brachyhypopomus (Hypopomidae) species that possess vestigial teeth on premaxilla. Few years later, Crampton et al. (2016) also observed teeth on juveniles of B. brevirostris, B. diazi, B. palenque, and B. walteri but not in adults.

• 28. Shape of teeth: (0) conical; (1) villiform (Albert, 2001, char. 24; Albert & Fink, 1996, char. 9; Correa et al., 2006: char. 9; Hulen et al., 2005, char. 34; Lundberg & Mago-Leccia, 1986, char. 2; Mago-Leccia, 1978; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 37; CI = 0.500, RI = 0.500).

This character is inapplicable in the examined Hypopomidae (see character 1).

• 29. Attachment area of anterior row of premaxillary teeth: (0) teeth firmly, but not fully ankilosed, attached to ventral surface of premaxilla; (1) teeth attached to ventral surface of premaxilla only by their anterobasal margin (Dutra et al., 2018; Fink, 1981; Vari et al., 2012; CI = 0.500; RI = 0.857).

Teeth are firmly, but not fully ankilosed, attached to premaxilla in most Gymnotiformes (Type 2 attachment of Fink, 1981; state 0). In contrast, teeth of anterior row of premaxilla are attached only by their anterobasal margins in Archolaemus, D. guchereauae, and E. oradens (state 1). As a consequence, these teeth are variably mobile relative to the premaxilla [Type 3 attachment of Dutra et al. (2018), Fink (1981) and Vari et al. (2012)].

• 30. Relative width of descending blade of maxilla: (0) narrower than its dorsal most portion (Figure 6b–d); (1) as wide as its anterior most portion (Figure 6a; Albert, 2001, char. 30 in part; Albert & Campos-da-Paz, 1998, char. 37; Correa et al., 2006, char. 11; Hulen et al., 2005, char. 36; Lundberg & Mago-Leccia, 1986, char. 20 in part; CI = 1.000; RI = 1.000).

The maxilla in Gymnotiformes can be splitted in two portions, the head of the maxilla which frequently possess an anterior hook-like process (see character 31), and a descending blade which correspond to the longer portion of this bone, that extends from a point just below the insertion of anterior hook-like process to the ventral tip of this bone. The relative width of descending blade of the maxilla is treated here as a ratio between the width of the dorsal portion of the descending blade and its width at middle length. When the anterior hook-like process is absent, and the head of maxilla is not well limited, the dorsal portion of anterior cartilage of maxilla was used as a landmark to establish the dorsal portion of the descending blade.

• 31. Anterior hook-like process on maxilla: (0) absent (Figure 6a); (1) present (Figure 6b–d; Albert, 2001, char. 32; CI = 0.500; RI = 0.900; Albert & Campos-da-Paz, 1998, char. 34; Albert & Fink, 1996, char. 8; Lundberg & Mago-Leccia, 1986, char. 4).

Rhabdolichops caviceps, R. eastwardi, R. electrogrammus, R. troscheli, R. zareti, and Sternopygus lack the anterior hook-like process on maxilla. Such process is present in small (less than 75 mm) Rhabdolichops species and reduces with growth (Lundberg & Mago-Leccia, 1986). Considering this ontogenetic variation, this character was codified only in adult specimens.

3.1.6 Lower jaw

• 32. Arrangement of teeth rows on dentary: (0) reaching at least the anterior limit of Meckel’s cartilage (Figures 8c, 9c, 10b, 11b); (1) not reaching the anterior limit of Meckel’s cartilage (Figure 7; Dutra et al., 2014; CI = 1.000; RI = AUT).

Teeth rows on the dentary in Gymnotiformes extend from its symphysis to a position at, or posterior to the anterior limit of Meckel’s cartilage. In contrast, the single teeth row in D. conirostris is restricted to the anterior portion of the dentary and never extends beyond the anterior limit of the coronomeckelian cartilage, a condition present only in this genus (Dutra et al., 2014). This character is inapplicable in the examined Hypopomidae (see Character 1).

• 33. Size of teeth along dentigerous surface of dentary: (0) similar in size along dentigerous surface (Figures 7-10); (1) increasing in size along dentigerous surface (Figure 11; Peixoto et al., 2015; de Santana & Vari, 2013; CI = 0.500; RI = 0.750).
• 34. Shape of dentary teeth: (0) medially curved (Figures 8-11); (1) straight (Figure 7; CI = 0.143; RI = 0.400).

This character is inapplicable in the examined Hypopomidae (see character 1).

• 35. Attachment area of teeth on dentary: (0) teeth attached along dentary rim (Figures 7, 10, 11); (1) teeth attached on a dorsolateral flange of dentary (Figures 8b and 9b; Dutra et al., 2018; CI = 0.333; RI = 0.750).

When present, the dentary teeth are attached along the dentary rim in most Gymnotiformes. Conversely, they are attached to an unossified dorsolateral flange of this bone in Archoalemus, E. oradens, D. guchereauae, and Japigny. This apomorphic condition was proposed as a diagnostic character for E. oradens by Dutra et al. (2018), who also commented on the presence of this condition in all aforementioned taxa.

• 36. Orientation of coronoid process: (0) ventrally curved in its posterior end (Figures 7 and 10); (1) nearly straight (Figures 8, 9 and 11; CI = 0.167; IR = 0.545).
• 37. Composition of coronoid process tip: (0) ossified; (1) cartilaginous (Vari et al., 2012; CI = 1.000; RI = 1.000).
• 38. Form of Meckel's cartilage: (0) rectangular shaped, its anterior portion approximately as wide as posterior portion (Figures 7, 9-11); (1) triangular-shaped, its anterior portion approximately half as wide as posterior portion (Figure 8; CI = 1.000; RI = 1.000).

Meckel's cartilage is rectangular shaped in most Gymnotiformes. In contrast, it is posteriorly expanded in Archolaemus due to its large attachment area in the angulo-articular. This configuration confers to it a triangular shape.

• 39. Relative length of coronomeckelian bone: (0) coronomeckelian bone corresponds to 45% or more of length of Meckel's cartilage (Figures 8, 9); (1) coronomeckelian bone corresponds to 30% or less of length of Meckel's cartilage (Figures 7, 10 and 11; Peixoto et al., 2015; Vari et al., 2012; CI = 0.200; RI = 0.692).

This character is polymorphic in A. blax due to the variation present from specimen to specimen, neither related to sexual dimorfism nor with ontogeny.

• 40. Association between retroarticular and socket of angulo-articular that receives condyle of quadrate (Figure 7): (0) retroarticular included in socket; (1) retroarticular not included in socket (Figures 8-11; Vari et al., 2012; CI = 0.500; RI = 0.500).

Vari et al. (2012: 672) proposed the retroarticular (misinterpreted as “angulo-articular”) not included in socket to receive the condyle of the quadrate as an evidence to consider Japigny as sister group of all the other Eigenmanniinae, a condition that also occurs in Sternopygus. In these taxa, the retroarticular shifts ventrally, and the socket that receives the condyle of quadrate is formed only by the angulo-articular. In all the other Gymnotiformes, the retroarticular is included in this socket.

• 41. Ventralmost lower jaw bone: (0) retroarticular (Figures 7 and 10); (1) dentary (Figures 8, 9; CI = 0.250; RI = 0.727).
• 42. Relative length of dentary: (0) as long as lower jaw depth (Figures 7 and 9-11); (1) 1.5 times as long as lower jaw depth (Figure 8; CI = 0.333; RI = 0.667).

Snout elongation in Gymnotiformes is not homologous within the order, since different bones are involved in the elongation process (e.g., de Santana & Vari, 2010: 267–670). One of the bones may be involved to the snout elongation is the dentary. The relative length of dentary is a ratio between the dentary length and dentary depth, in which dentary length corresponds to the distance between symphysis and coronoid process, whereas the lower jaw depth corresponds to distance between dentary rim and ventral margin of retroarticular.

3.1.7 Suspensorium

• 43. Ventral surface of endopterygoid: (0) edentulous (Figures 12, 13); (1) toothed (Figure 14; Albert, 2001, char. 133; Albert & Campos-da-Paz, 1998, char. 134 in part; Albert & Fink, 1996, char 22; Correa et al., 2006, char. 16; Dutra et al., 2014; Hulen et al., 2005, char. 42; Lundberg & Mago-Leccia, 1986, char. 25; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 125; CI = 0.250–0.333; RI = 0.700–0.800).

The presence of teeth on the ventral surface of endopterygoid has been proposed as a synapomorphy for the Sternopygidae by several authors (e.g., Albert & Fink, 1996; Mago-Leccia, 1978, 1994), and its absence recently proposed as an autapomorphy for D. conirostris (Dutra et al., 2014).

• 44. Relative length of anterior portion of endopterygoid: (0) as long as its depth (Figures 12, 13); (1) twice as long as its depth (Figure 14; CI = 0.500; RI = 0.667).

The relative length of the endopterygoid anterior portion was taken in comparison with its depth. It extends from the anterior tip of this bone until the origin of its ascending process, whereas its depth was taken at the insertion of the same process. This character is inapplicable in Sternarchorhamphus due to the absence of the ascending process of endopterygoid.

• 45. Relative length of metapterygoid: (0) as long as its largest depth (Figures 12, 13); (1) at least twice as long as its largest depth (Figure 14; Albert, 2001, char. 136; CI = 0.500; RI = 0.667).

The length of metapterygoid corresponds to the distance between its anteriormost and posteriormost margins, whereas its depth corresponds to distance between its dorsal most and ventral most margins.

• 46. Relative length of symplectic: (0) one to two thirds as long as hyomandibula (Figures 12, 13); (1) as long as hyomandibula (Figure 14; Albert, 2001, char. 137; Albert & Campos-da-Paz, 1998, char. 138; Tagliacollo et al., 2016b, char. 129; CI = 0.500; RI = 0.667).
• 47. Orientation of hyomandibula: (0) main axis of hyomandibula forming an angle of nearly 40° in relation to its dorsal margin (Figure 12): (1) main axis of hyomandibula forming an angle of nearly 60° in relation to its dorsal margin (Figures 13, 14; Albert, 2001, char. 138; Tagliacollo et al., 2016b, char. 130; CI = 0.200; RI = 0.556).
• 48. Shape of posterodorsal portion of hyomandibula: (0) extending posteriorly to condyle that receives opercular socket (Figures 13, 14); (1) ending at same level of condyle that receives opercular socket (Figure 12; CI = 0.500; RI = 0.857).
• 49. Foramen on posterodorsal margin of hyomandibula: (0) present (Figure 15a); (1) absent (Figure 15b; CI = 0.333; RI = 0.714).

The recurrent ramus of the anteroventral part of the anterior lateral line nerve (R-Avn) extends through the postotic foramen of the neurocranium. In most of analyzed Gymnotiformes, this ramus passes through the foramen in the posterodorsal margin of the hyomandibula, before extending to the body. In contrast, the posterodorsal foramen of hyomandibula is absent in R. caviceps, R. electrogrammus, R. eastwardi, R. troscheli, R. zareti, Sternarchorhamphus, and Sternopygus, and the R-Avn passes next to the hyomandibula.

3.1.8 Cephalic lateral line system

• 50. Shape of tubules of mandibular and preopercular canals: (0) slender and tubular; (1) enlarged and half-pipe shaped (Figure 16; see Albert, 2001, char. 91; Albert & Campos-da-Paz, 1998, char. 92 in part; Peixoto et al., 2021: fig. 5; Triques, 1993, char. 26; CI = 1.000; RI = 1.000).

In general, the cephalic lateral line system in Gymnotiformes are formed by slender tubules. However, members of Sternopygidae possess most of the cephalic lateral line system, including the mandibular and preopercular canals, compounded by enlarged and poorly-ossified half-pipe shaped tubules. The presence of enlarged cephalic lateral line system has been historically proposed as a diagnostic feature for Sternopygidae (e.g., Albert, 2001; de la Hoz & Chardon, 1984; Triques, 1993). However, the enlargement of some sections of the cephalic lateral line system (e.g., infraorbitals, otic, supratemporal) varies within Eigenmanniinae; thus, each of these transformations is presented as independent characters.

• 51. Association of mandibular canal and dentary: (0) not fused; (1) fused (Albert, 2001, char. 90; Albert & Campos-da-Paz, 1998, char. 91; Triques, 1993, char. 25; CI = 1.000; RI = 1.000).

The mandibular canal of the cephalic lateral line system is embedded in the dermis, sometimes partially fused through its anteriormost portion to dentary. In contrast, the mandibular canals in Sternopygidae are completely incorporated to the dentary.

• 52. Shape of nasal: (0) slender and tubular (see Peixoto et al., 2019: fig. 4b); (1) enlarged and half-pipe shaped (Figures 17-20; Albert, 2001, char. 67; Albert & Campos-da-Paz, 1998, char. 68; Triques, 1993, char. 22; CI = 1.000; RI = 1.000).
• 53. Shape of supraorbital canal: (0) slender and tubular; (1) enlarged and half-pipe shaped (Figures 17-20; Triques, 1993, char. 5; CI = 1.000; RI = 1.000).

• 54. Shape of antorbital and infraorbitals 1–4: (0) slender and tubular; (1) enlarged and half-pipe shaped (Figure 21—Albert, 2001: char. 87; Albert & Campos-da-Paz, 1998, char. 88; Albert & Fink, 1996, char. 17; Correa et al., 2006, char. 12; Lundberg & Mago-Leccia, 1986: char. 1; Mago-Leccia, 1978; Mago-Leccia & Zaret, 1978; Triques, 1993: char. 21; Tagliacollo et al., 2016b, char. 97; CI = 1.000; RI = 1.000).

• 55. Posterodorsal laminar expansion on infraorbital 1 + 2: (0) absent (Figure 21a–b); (1) present (Figure 21c–e; CI = 0.333; RI = 0.857).

According to Mago-Leccia (1978) and Lundberg and Mago-Leccia (1986), infraorbitals 1 and 2 are fused as a single osseous element that usually has a posterodorsal laminar expansion. This expansion is absent in Apteronotidae, Gymnotidae, Hypopomidae, Archolaemus, D. conirostris, and R. nigrimans. In contrast, that expansion is present in D. guchereauae, Eigenmannia, Japigny, R. caviceps, R. eastward, R. electrogrammus, R. lundbergi, R. troscheli, R. zareti, and Sternopygus.

• 56. Depth of posterodorsal laminar expansion of infraorbital 1 + 2: (0) half as long as infraorbital 1+2 length (Figure 21d–e); (1) as long as infraorbital 1+2 length or more (Figure 21c; Peixoto et al., 2015; CI = 0.167; RI = 0.615).

The depth of posterodorsal lateral expansion corresponds to the distance between its insertion on infraorbital 1 + 2 and its distal limit (Peixoto et al., 2015). This character is inapplicable in Gymnotidae, Hypopomidae, Apteronotidae, Archolaemus, D. conirostris, and R. nigrimans due to the absence of such expansion in species of these taxa.

• 57. Association between infraorbital 1 + 2 and infraorbital 3 (Figure 21a–d): (0) not in contact; (1) contacting each other (Figure 21e; CI = 1.000; RI = 1.000).
• 58. Shape of infraorbitals 5 and 6: (0) slender and tubular (Figure 21a–d); (1) enlarged and half-pipe shaped (Figure 21e; Correa et al., 2006, char. 13; Lundberg & Mago-Leccia, 1986, char. 19 in part; CI = 1.000; RI = 1.000).
• 59. Site of connection between infraorbital and supraorbital canals: (0) anterior to sphenotic process; (1) on sphenotic process (Correa et al., 2006, char. 15; Lundberg & Mago-Leccia, 1986, char. 19 in part; CI = 0.333; CI = 0.800).
• 60. Shape of otic canal: (0) slender and tubular (Figures 17, 18); (1) enlarged and half-pipe shaped (Figures 19, 20; CI = 0.500; RI = 0.875).
• 61. Association between extrascapular and neurocranium: (0) extrascapular at joint of parietal, pterotic and epioccipital bones (Figures 17-19); (1) extrascapular at posterodorsal border of posttemporal fossa (Figure 20; Albert, 2001, char. 170; Albert & Campos-da-Paz, 1998, char. 171; Correa et al., 2006, char. 20; Lundberg & Mago-Leccia, 1986, char. 18 in part; CI = 0.500; RI = 0.833).

The extrascapular located on the posterior border of posttemporal fossa was previously proposed as a synapomorphy for Rhabdolichops (Correa et al., 2006; Lundberg & Mago-Leccia, 1986); however, in R. lundbergi and R. nigrimans, the extrascapular is located at joint of parietal, pterotic, and epioccipital bones.

• 62. Shape of extrascapular: (0) slender and tubular (Figures 17, 18); (1) enlarged and half-pipe shaped (Figures 19, 20; Lundberg & Mago-Leccia, 1986, char. 18 in part; Mago-Leccia, 1994; CI = 1.000; RI = 1.000).

3.1.9 Hyoid Arch

• 63. Shape of urohyal ridge: (0) ovoid (Figure 22a); (1) diamond-shaped (Figure 22b); (2) triangular (Figure 22c; Albert, 2001, char. 166; Tagliacollo et al., 2016b, char. 160; CI = 0.333; RI = 0.692).

The shape of the urohyal ridge is delimited by the width variation along its length. In Gymnotidae, Brachyhypopomus, and Sternopygus, its posterior most edge is rounded, resulting in an ovoid shape. On the contrary, in Apteronotus, Archolaemus, D. guchereauae, Eigenmannia, “E.” goajira, and Japigny, the urohyal ridge is narrow on its anterior portion becoming wider until its midlength, and then narrowing to its posteriormost portion, resulting in a diamond shape formed by its contralateral ridges. A third condition is present in D. conirostris, Microsternarchus, Rhabdolichops, and Sternarchorhamphus, where the largest width of the contralateral ridges is located at its posteriormost margin, resulting in a triangular shape.

• 64. Relative length of free urohyal blade: (0) shorter than urohyal ridge (Figure 22a); (1) as long as urohyal ridge (Figure 22b–c; Albert, 2001, char. 167–168; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 161–162; CI = 0.500; RI = 0.833).

The length of the urohyal varies within Gymnotiformes (for diversity of urohyal in Gymnotiformes see Albert, 2001: fig. 32). The free urohyal blade corresponds to the portion of this bone posterior to the posterior margin of the urohyal ridge.

• 65. Relative length of posterior ceratohyal: (0) as long as ventral hypohyal; (Figure 23a; 1) 1.5 times as long as ventral hypohyal (Figure 23b; Vari et al., 2012; CI = 0.333; RI = 0.600).

• 66. Shape of first branchiostegal: (0) thin (Figure 23a); (1) spatulated (Figure 23b; Albert, 2001, char. 145 in part; Albert & Campos-da-Paz, 1998, char. 146 in part; Tagliacollo et al., 2016b; char. 140 in part; CI = 0.500; RI = 0.500).

Gymnotiformes possess up to five branchiostegals (Albert, 2001; de Santana & Vari, 2010). Three branchiostegals are attached to the anterior ceratohyal, the fourth is variable attached from posterior portion of anterior ceratohyal to anterior portion of posterior ceratohyal, and the fifth, when present, is attached at joint of the anterior and posterior ceratohyal (de Santana & Vari, 2010). Each of these elements varies independently in shape from a thin to a spatulated bone. Thus, we codified each variation as an independent character, except for the fourth branchiostegal which is broad in all examined specimens. de Santana and Vari, (2010), also suggested, based on positional information, that branchiostegal loss in anterior ceratohyal proceeds in an anterior–posterior sequence. Thus, when only two branchiostegal are attached in the anterior ceratohyal, we interpreted that first branchiostegal is absent, a condition present in examined Apteronotidae, Gymnotidae, and Microsternarchus. Thus, this character is inapplicable in these taxa.

• 67. Shape of second branchiostegal: (0) thin (Figure 23a); (1) spatuladed (Figure 23b; Albert, 2001, char. 145 in part; Albert & Campos-da-Paz, 1998, char. 146 in part; Tagliacollo et al., 2016b; char. 140 in part; CI = 0.333; RI = 0.600).
• 68. Shape of third branchiostegal: (0) spatulated (Figure 23b); (1) thin (Figure 22a; Albert, 2001, char. 145 in part; Albert & Campos-da-Paz, 1998, char. 146 in part; Tagliacollo et al., 2016b; char. 140 in part; CI = 0.167; RI = 0.167).
• 69. Shape of fifth branchiostegal: (0) spatulated (Figure 23b); (1) thin (Figure 23a; Albert, 2001, char. 145 in part, Tagliacollo et al., 2016b; char. 140 in part; CI = 0.333; RI = 0.000).

3.1.10 Branchial arches

• 70. Form of gill rakers: (0) short and unossified; (1) long and ossified (Figure 24; Albert, 2001; char. 147; Albert & Campos-da-Paz, 1998, char. 148; Albert & Fink, 1996, char. 34; Correa et al., 2006, char. 21 and 23; Lundberg & Mago-Leccia, 1986, char. 24 in part; Fink & Fink, 1981; Mago-Leccia, 1978; Mago-Leccia & Zaret, 1978; CI = 1.000; RI = 1.000).

The presence of elongate and well-ossified gill rakers is related to the planktivorous feeding habits in fishes (Mago-Leccia & Zaret, 1978). In most Gymnotiformes, the gill rakers are short and unossified (see Hilton et al., 2007: fig. 14). However, elongate and ossified gill rakers are present in R. caviceps, R. eastwardi, R. electrogrammus, R. troscheli, and R. zareti. Mago-Leccia & Zaret (1978) initially proposed this unique gill morphology as a primitive condition within the order. Later, Fink and Fink (1981: 309) and Lundberg and Mago-Leccia (1986: 64), however, argued that presence of well-developed gill rakers is, in fact, an apomorphy within Gymnotiformes. Thus, this apomorphic condition was proposed as a synapomorphy for Rhabdolichops by several authors (e.g., Lundberg & Mago-Leccia, 1986). In spite of that, after the descriptions of R. lundbergi and R. nigrimans, in which the gill rakers are short and unossified, the presence of long and ossified gill rakers was restricted to a less inclusive clade in Rhabdolichops (Correa et al., 2006).

The gill rakers firmly attached on gill arches were proposed by Albert and Campos-da-Paz (1998) and Albert (2001) as a synapomorphy for Rhabdolichops. The morphology of the gill rakers in Gymnotiformes is well discussed by Lundberg and Mago-Leccia (1986: 64), which allow us to interpret that type of attachment in Rhabdolichops as non-independent condition from the presence of long and ossified gill rakers in this genus.

• 71. Patches of teeth associated with gill rakers: (0) absent (Figures 24, 25); (1) present (Figure 26; Albert & Fink, 1996, char. 34; Lundberg & Mago-Leccia, 1986, char. 24 in part; Mago-Leccia, 1978; CI = 1.000; RI = 1.000).

In Sternopygus, there are paired patches that support small denticles at the base of all gill rakers on branchial arches 1 to 4, and a single patch laterally to the base of each gill raker of ceratobranchial 5. This condition was previously proposed as a synapomorphy for this genus by several authors (e.g., Albert & Fink, 1996; Hulen et al., 2005).

• 72. Basibranchial 1: (0) unossified (Figures 25, 26); (1) ossified (Figure 24; Albert, 2001, char. 163 in part; Albert & Campos-da-Paz, 1998, char. 164 in part; Fink & Fink, 1981; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 158 in part; CI = 0.200; RI = 0.200).

• 73. Basibranchial 3: (0) unossified (Figure 25); (1) ossified (Figures 24 and 26; Albert, 2001, char. 163 in part; Albert & Campos-da-Paz, 1998, char. 164 in part; Tagliacollo et al., 2016b, char. 158 in part; CI = 0.500, IR =0.750).

3.1.11 Neurocranium

• 74. Relative length of mesethmoid: (0) 1.5 times as long as nasal length (Figures 18-20); (1) twice as long as nasal length; (2) three times as long as nasal length (Figure 17; Albert, 2001, char. 53; Albert & Campos-da-Paz, 1998, char. 54; Tagliacollo et al., 2016b, char. 61; CI = 0.222; RI = 0.682).

The length of anterior portion of frontal was proposed by Albert (2001) and followed by Tagliacollo et al. (2016b) to compare the relative length of the mesethmoid. However, considering the elongation of the anterior portion of frontal in some taxa (see character 81), mesethmoid length was herein estimated in comparison with nasal length that does not vary as much within Eigenmanniinae.

• 75. Posterodorsal process of lateral ethmoid: (0) absent (Figure 27); (1) present (Figures 17-20; CI = 0.500; RI = 0.667).

The lateral ethmoid is connected to frontal only through its dorsal portion, which is linked to the anteroventral portion of this bone in A. orientalis, Brachyhypopomus, and Apteronotidae. In most Gymnotiformes, however, the lateral ethmoid has a posterodorsal process that provides an additional connection of this bone near the antorbital process of frontal. Because of the absence of the lateral ethmoid, this character is inapplicable in Gymnotidae and Microsternarchus.

• 76. Relative length of posterodorsal process of lateral ethmoid: (0) longer than main axis of this bone (Figure 17); (1) shorter than main axis of this bone (Figures 18-20; CI = 0.500; RI = 0.875).

This character is inapplicable in A. orientalis, Brachyhypopomus, and Apteronotidae due to the absence of the process, and Gymnotidae and Microsternarchus (see character 75).

• 77. Relative length of posterior portion of vomer: (0) shorter than its anterior portion (Figure 28a); (1) as long as its anterior portion (Figure 28b; CI = 0.200; RI = 0.556).

The anterior portion of vomer extends from its anterior margin to the anteriormost contact with the endopterygoid, whereas its posterior portion extends from that point until posterior margin of this bone. This character was not coded in Hypopomidae due to the absence of dermal portion of vomer in species of these taxa.

• 78. Relative length of posterior process of vomer: (0) shorter than vomer largest width (Figure 28a); (1) longer than vomer largest width (Figure 28b; CI = 0.333; RI = 0.500).

The vomer in Sternopygidae has two posterior processes. The length of these processes was compared with the broadest portion of vomer, which correspond to the width at base of these processes. This character is inapplicable in Apteronotidae and S. macrurus due to the absence of the posterior process of the vomer. It is also inapplicable in Hypopomidae because of the absence of the dermal portion of vomer.

• 79. Antorbital process of frontal: (0) absent; (1) present (Figures 17-20; Aguilera, 1986; Albert, 2001, char. 69; Albert & Campos-da-Paz, 1998, char. 70; Albert & Fink, 1996, char. 20; Hulen et al., 2005, char. 27; Tagliacollo et al., 2016b, char. 77; Triques, 1993; char. 3; CI = 1.000; RI = 1.000).
• 80. Degree of elongation of antorbital portion of parasphenoid: (0) shorter than its orbital portion (Figures 18-20); (1) longer than its orbital portion (Figure 17; CI = 0.333; RI = 0.600).

The parasphenoid antorbital portion extends from its anterior margin to a vertical through the frontal antorbital process, when this is present, or through the orbital anterior margin, whereas the parasphenoid orbital portion extends from that point to the vertical through the sphenotic process.

• 81. Relative length of antorbital portion of frontal: (0) shorter than half-length of orbit (Figure 18); (1) longer than half-length of orbit (Figure 17; CI = 0.200; RI = 0.429).

The frontal antorbital portion extends from its anterior margin to a vertical through frontal antorbital process or through the orbit anterior margin.

• 82. Number of prootic facial nerve foramina: (0) single foramen for two main rami of facial nerve (Figures 17-19); (1) two foramina, each rami of facial nerve with their own foramina (Figure 20; Correa et al., 2006, char. 18; Lundberg & Mago-Leccia, 1986, char. 17; CI = 1.000; RI = 1.000).

The two main rami of the facial nerve leave the braincase through a single foramen located at the prootic middorsal portion in most of Gymnotiformes. Conversely, R. caviceps, R. eastwardi, R. electrogrammus, R. troscheli, and R. zareti have two foramina in the prootic. The apomorphic condition was previously presented as a synapomorphy for Rhabdolichops (Correa et al., 2006; Lundberg & Mago-Leccia, 1986). However, Correa et al. (2006) incorrectly described the presence of two prootic foramina in R. lundbergi and R. nigrimans, which actually have only a single prootic foramen (Figure 18).

• 83. Posttemporal fossa: (0) absent (Figures 17-19); (1) present (Figure 20; Albert, 2001, char. 81; Albert & Campos-da-Paz, 1998, char. 82; Albert & Fink, 1996, char. 25; Correa et al., 2006, char. 19; Fink & Fink, 1981, char. 12; Lundberg & Mago-Leccia, 1986, char. 18 in part; Mago-Leccia, 1978; Tagliacollo et al., 2016b, char. 88; CI = 0.500; RI = 0.833).

In most Gymnotiformes, the parietal, epioccipital, and pterotic bones are in contact; consequently, the posttemporal fossa is absent. However, the parietal, epioccipital and pterotic bones do not contact their mutual margins, and therefore, the posttemporal fossa is visible in Rhabdolichops caviceps, R. eastwardi, R. electrogrammus, R. troscheli, R. zareti, and Sternopygus.

3.1.12 Pectoral fin and girdle

• 84. Medial border of cleithrum: (0) limited by a medial expansion of this bone (Figures 29-31); (1) limited by a cleithral ridge (Figure 32; Correa et al., 2006; char. 30; CI = 1.000; RI = 1.000).

The cleithral ridge is the crest to which the coracoid is dorsally attached. This ridge extends anteriorly toward the cleithrum anterior portion. In most Gymnotiformes, the cleithrum is medially expanded; consequently, the cleithral ridge is located on the middle of this bone. In R. caviceps, R. eastwardi, R. electrogrammus, R. troscheli, and R. zareti, the cleithral ridge limits the medial border of this bone.

• 85. Association between anteroventral process of coracoid and cleithral ridge: (0) anterior process of coracoid short, not contacting the cleithral ridge on cleithrum anterior most portion (Figures 29, 30); (1) anteroventral process of coracoid long, contacting the cleithral ridge on cleithrum anterior most portion (state 0—Figures 31, 32; Albert, 2001, char. 175; Vari et al., 2012) (CI = 1.000; RI = 1.000).

The coracoid in most Gymnotiformes has an anteroventral process that extends toward the cleithral ridge anterior portion. In Apteronotidae, Japigny, and Sternopygus, the coracoid anteroventral process is short and does not contact the cleithral ridge anteriormost portion. In Archolaemus, D. conirostris, D. guchereauae, Eigenmannia (polymorphic in E. sayona) and “E.” goajira and Rhabdolichops, the coracoid anteroventral process is elongated and contacts the cleithral ridge anteriormost portion. This character is inapplicable in Gymnotidae and Hypopomidae due to the absence of the coracoid anteroventral process.

• 86. Shape of cleithrum posterior process: (0) approximately triangular (Figures 29-31); (1) narrow, almost digitiform (Figure 32; Correa et al., 2006, char. 29; Lundberg & Mago-Leccia, 1986, char. 22; I = 1.000; RI = 1.000).

The presence of a narrow, almost digitiform posterior process of cleithrum was proposed as a synapomorphy for Rhabdolichops (Correa et al., 2006; Lundberg & Mago-Leccia, 1986). However, a cleithrum posterior process approximately triangular occurs in all examined specimens of R. lundbergi and R. nigrimans, a condition also present in most Gymnotiformes. A narrow, rather digitiform, posterior process of cleithrum is present in R. caviceps, R. eastwardi, R. electrogrammus, R. troscheli, and R. zareti.

• 87. Postcleithrum 1: (0) present; (1) absent (Hulen et al., 2005; char. 61; Lundberg & Mago-Leccia, 1986, char. 11 in part; CI = 0.500; RI = 0.667).
• 88. Shape of postcleithrum 1: (0) thin and discoid; (1) robust with posterior margin straight (Albert, 2001, char. 171; Lundberg & Mago-Leccia, 1986, char. 11 in part; CI = 1.000; RI = AUT).

This character is inapplicable in Sternopygus and Hypopomidae due to the absence of first postcleithrum in these taxa.

• 89. Association between supracleithrum and posttemporal: (0) not fused (Figure 33a); (1) fused (Figure 33b; Albert, 2001, char. 169; Albert & Campos-da-Paz, 1998, char. 170; Correa et al., 2006, char. 24; Fink & Fink, 1981, char. 95 in part; Hulen et al., 2005, char. 51; Lundberg & Mago-Leccia, 1986, char. 10; Mago-Leccia, 1978; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 163; CI = 1.000; RI = 1.000).

• 90. Baudelot's ligament: (0) unossified (Figures 19, 20); (1) partially ossified (Figures 17, 18; Fink & Fink, 1981, char. 95 in part; CI = 0.333; RI = 0.833).
• 91. Position of scapular foramen: (0) between scapula and coracoid (Figure 29); (1) entirely included within the scapula (Figures 30-32; Albert, 2001, char. 173; Albert & Campos-da-Paz, 1998, char. 174; Albert & Fink, 1996, char. 26; Correa et al., 2006, char. 27; Hulen et al., 2005, char. 49; Lundberg & Mago-Leccia, 1986, char. 9; Mago-Leccia, 1978; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 164; CI = 1.000; RI = 1.000).
• 92. Mesocoracoid: (0) present; (1) absent (Figures 29-32; Albert, 2001, char. 174; Albert & Campos-da-Paz, 1998, char. 175; Fink & Fink, 1981; Lundberg & Mago-Leccia, 1986, char. 12; Mago-Leccia & Zaret, 1978; Tagliacollo et al., 2016b, char. 165; Triques, 1993, char. 46; CI = 1.000; RI = 1.000).
• 93. Association between proximal radials 3 and 4: (0) not fused (Figure 29); (1) fused (Figures 31, 32; Albert, 2001, char. 176; Albert & Fink, 1996, char. 28; Correa et al., 2006: char. 37; Hulen et al., 2005, char. 54; Lundberg & Mago-Leccia, 1986, char. 15; Mago-Leccia, 1978; Mago-Leccia & Zaret, 1978; Meunier et al., 2011; Tagliacollo et al., 2016b, char. 167; Triques, 1993, char. 47; Vari et al., 2012; I = 1.000; RI = 1.000).
• 94. Pectoral-fin length: (0) pectoral-fin tip not beyond abdominal cavity (Figure 34a–c); (1) pectoral-fin tip extending beyond abdominal cavity (Figure 34d; modified from Albert, 2001, char. 177; Albert & Campos-da-Paz, 1998, char. 178; Albert & Fink, 1996, char. 29; Correa et al., 2006, char. 28; Hulen et al., 2005, char. 55; Tagliacollo et al., 2016b, char. 168; CI = 0.333, RI = 0.750).

The pectoral-fin length was previously determined based in relation to head length (Correa et al., 2006). Because of the large variation of the snout length in Eigenmanniinae, pectoral fin elongation was instead estimated in function of the abdominal cavity length that does not vary as much within Eigenmanniinae, and could be easily estimated through X-ray images of specimens.

3.1.13 Weberian complex

• 95. Shape of parapophysis of second vertebra: (0) laterally projected (Figure 35); (1) ventrally curved (Figures 36-38; Vari et al., 2012; CI = 0.500, RI = 0.800).

Vari et al. (2012: 673) argued that the presence of a laterally projected parapophysis of second vertebra suggests that Japigny is the sister group of all the other Eigenmanniinae. It is noteworthy, however, that the second vertebra parapophysis is also ventrally curved in Japigny.

• 96. Association between parapophyses of second and fourth vertebrae: (0) distinctly separated (Figures 35, 36); (1) in contact (Figures 37, 38; Albert, 2001, char. 181; Correa et al., 2006, char. 32; Fink & Fink, 1981, char. 88 in part; Lundberg & Mago-Leccia, 1986, char. 8; Vari et al., 2012; CI = 0.500, RI = 0.889).
• 97. Shape of neural spine of fourth vertebra of Weberian complex: (0) long spine-shaped (Figure 37); (1) tiny spine-shaped (Figure 38); (2) nearly rectangular (Figures 35, 36; CI = 0.200; RI = 0.467).

The neural spine of fourth vertebra of Weberian complex when long and spine-shaped, it is almost as long as neural spine of fifth vertebra. In contrast, a tiny and spine-shaped neural spine of fourth vertebra is distinctly smaller than neural the spine of fifth vertebra. A third condition occurs when neural spine of fourth vertebra of Weberian complex is nearly rectangular with distal portion horizontally oriented.

3.1.14 Axial skeleton

• 98. Relative length of two anteriormost post-Weberian ribs: (0) approximately 80% as long as abdominal cavity depth (Figure 34a–b); (1) approximately as long as abdominal cavity depth (Figure 34c–d; Albert, 2001, char. 210; Albert & Fink, 1996, char. 43; Albert & Campos-da-Paz, 1998, char. 211; Correa et al., 2006, char. 36; Hulen et al., 2005, char. 59; Lundberg & Mago-Leccia, 1986, chars. 6 and 14; CI = 0.200, RI = 0.333).
• 99. Relative size of hemal spine of 26th to 30th vertebrae (modified from Albert, 2001, char. 198): (0) shorter than its associated pterygiophore (Figure 39a); (1) longer than its associated pterygiophore (Figure 39b; CI = 0.200, RI = 0.765).

3.1.15 Intermuscular bones

• 100. Rami arrangement of epineurals at 7–9th vertebrae: (0) all rami in same direction (Figure 40a); (1) rami arranged in several directions (Figure 40b; Albert, 2001, char. 184 in part; Albert & Campos-da-Paz, 1998, char. 185 in part; Albert & Fink, 1996, char. 38; Correa et al., 2006, char. 33 in part; Hulen et al., 2005, char. 57 in part; Lundberg & Mago-Leccia, 1986, char. 7 in part; CI = 0.250, RI = 0.800).

The intermuscular bones extend along body length in Gymnotiformes and are variably present in other fishes (de Santana & Vari, 2010). The highly branched intermuscular bones were proposed as a synapomorphy for Eigenmanniinae (Lundberg & Mago-Leccia, 1986). However, the ramification of the three series of intermuscular bones (epineurals, epicentrals, and “epipleurals”) vary independently in the subfamily.

Epineurals are located laterally to the neural spines of the vertebrae and dorsally to epicentrals. The epineural bones at region of 7–9th vertebrae present all rami in same direction in most Gymnotiformes. Some specimens of E. limbata and E. macrops present few rami upright oriented to main axis of epineurals. However, these rami are not connected to the epineurals, and consequently, this condition was codified as state 0. On the contrary, the epineurals could present a highly branched structure, in which some rami are upright to the main axis of the epineurals resulting in a star-shaped appearance.

• 101. Shape of epicentrals at 7–9th vertebrae: (0) unbranched (Figure 40b); (1) branched (Figure 40a; Albert, 2001, char. 184 in part; Albert & Campos-da-Paz, 1998, char. 185 in part; Albert & Fink, 1996, char. 37 in part; Correa et al., 2006, char. 33 in part; Hulen et al., 2005, char. 57 in part; Lundberg & Mago-Leccia, 1986, char. 7 in part; CI = 0.333, RI = 0.882).

The epicentrals are located laterally to vertebral centra, between the epineurals and “epipleurals”.

• 102. “Epipleurals” at 7–9th vertebrae: (0) absent (Figure 40a); (1) present (Figure 40b; Albert, 2001, char. 184 in part; Albert & Campos-da-Paz, 1998, char. 185 in part; Albert & Fink, 1996, char. 37 in part; Correa et al., 2006, char. 33 in part; Hulen et al., 2005, char. 57 in part; Lundberg & Mago-Leccia, 1986, char. 7 in part; CI = 0.500, RI = 0.941).

Lundberg and Mago-Leccia (1986) indicated the presence of epipleurals in some Sternopygidae members as the ossified structure that links rib to epicentrals. Such definition contrast to Gemballa and Britz (1998), whose described the epipleurlas as fiber bundle that ossify into epipleural originates on the rib or the dorsal part of the hemal arch and runs posterodorsally to the integument. Considering the inconclusive homology of these structures located between epicentrals and pleural ribs, we called they here as “epipleurals”. The “epipleurals” are included in the figure depicting intermuscular bones in R. electrogrammus (Lundberg & Mago-Leccia, 1986: fig. 6D), but none of our analyzed specimens of R. electrogrammus possess any “epipleural” bone. Consequently, the character was codified as polymorphic for this taxon.

3.1.16 Unpaired fins

• 103. Branched anal-fin rays: (0) present along all the fin except for anterior 6 to 60 fin rays; (1) only 10 to 20 branched anal-fin rays present at midfin (Albert, 2001, char. 197; Albert & Campos-da-Paz, 1998, char. 198; Albert & Fink, 1996, char. 49; Hulen et al., 2005, char. 65; CI = 1.000, RI = 1.000).

The anal fin in most Gymnotiformes anteriorly has between 6 and 60 unbranched fin rays followed by more than 100 branched fin rays. Among the examined taxa, most of the anal-fin rays in Sternopygus are unbranched, except for the presence of 10 to 20 branched fin rays at midfin.

• 104. Caudal fin: (0) absent, (1) present (see Albert, 2001, char. 222; de Santana et al., 2013: fig. 3; Fink & Fink, 1981, char. 109 in part; Triques, 1993, char. 50; CI = 1.000, RI = 1.000).

3.1.17 Pigmentation

• 105. Vertical bands on body: (0) absent (Figure 1a–c,e); (1) present (see Meunier et al., 2011: fig. 2; Vari et al., 2012; CI = 0.500, RI = 0.000).
• 106. Lateral line stripe: (0) absent (Figure 1d–e); (1) present (Figure 1a–c; Albert, 2001, char. 8 in part; Albert & Campos-da-Paz, 1998, char. 8 in part; Peixoto et al., 2015; Tagliacollo et al., 2016b, char. 20 in part; Vari et al., 2012; CI = 0.250, RI = 0.857).

The presence of two to four dark longitudinal lines was previously used in phylogenetic analyses by some authors (Albert, 2001; Albert & Campos-da-Paz, 1998; Tagliacollo et al., 2016b). The presence of stripes is herein analyzed independently to infer correct hypotheses of homology as proposed in Vari et al. (2012) and Peixoto et al. (2015). This character is polymorphic in D. conirostris (Dutra et al., 2014: 350).

• 107. Superior midlateral stripe: (0) absent (Figure 1a–b,d–e); (1) present (Figure 1c; Albert, 2001, char. 8 in part; Albert & Campos-da-Paz, 1998, char. 8 in part; Peixoto et al., 2015; Tagliacollo et al., 2016b, char. 20 in part; CI = 0.333, RI = 0.883).

The superior midlateral stripe is a diffuse group of chromatophores between the lateral line and the proximal portion of pterygiophores in form of a stripe (Peixoto et al., 2015; Peixoto & Wosiacki, 2016).

• 108. Inferior midlateral stripe: (0) absent (Figure 1d–e); (1) present (Figure 1a–c; Albert, 2001, char. 8 in part; Albert & Campos-da-Paz, 1998, char. 8 in part; Peixoto et al., 2015; Tagliacollo et al., 2016b, char. 20 in part; CI = 0.143, RI = 0.667).

The inferior midlateral stripe is formed by a band of dusky to dark pigmentation overlying the proximal portion of pterygiophores of the anal fin (Peixoto et al., 2015).

• 109. Anal-fin base stripe: (0) absent; (1) present (Figure 1; Albert, 2001, char. 8 in part; Albert & Campos-da-Paz, 1998, char. 8 in part; Vari et al., 2012; Peixoto et al., 2015; Tagliacollo et al., 2016b, char. 20 in part; CI = 0.333, RI = 0.800).

The anal-fin base stripe is formed by a band of dusky to dark pigmentation overlying the distal portion of the anal fin pterygiophores (Peixoto et al., 2015).

• 110. Coloration along distal margin of anal fin: (0) fin hyaline to dark but lacking a distinct band along the fin distal portions (Figure 1a,c-e); (1) distal margin dark (see Hilton & Cox-Fernandes, 2017: fig. 1d and 2; Correa et al., 2006, char. 2; CI = 0.250, RI = 0.400).
• 111. Humeral spot: (0) absent; (1) present (see Albert & Fink, 1996: fig. 7; Albert & Fink, 1996, char. 35; Hulen et al., 2005, char. 4; CI = 1.000, RI = 1.000).

3.1.18 Miscellaneous

• 112. Scales on anterodorsal portion of body: (0) present; (1) absent (Albert, 2001, char. 15; Albert & Campos-da-Paz, 1998, char. 15; Correa et al., 2006, char. 4; Lundberg & Mago-Leccia, 1986, char. 23; Mago-Leccia, 1978; Tagliacollo et al., 2016b, char. 30) (CI = 0.500, RI = 0.857).
• 113. Dorsal filament: (0) absent; (1) present (Albert, 2001, char. 192; Albert & Campos-da-Paz, 1998, char. 193; Tagliacollo et al., 2016b, char. 183; CI = 1.000, RI = 1.000).

The dorsal filament, also called as dorsal fleshy organ (Albert, 2001; Tagliacollo et al., 2016b), is a tapering structure that is attached to a middorsal groove on the posterior half of body and extends parallel to the dorsal margin of epaxial (Franchina & Hopkins, 1996: figs. 1 and 3). Its homology still in debate, the first hypothesis suggests that it is homologous to the adipose fin of Characiformes and Siluriformes; on the contrary, considering that adipose fin is absent in all other Gymnotiformes, the competitive hypothesis suggests that it is a unique structure present in Apteronotidae (Albert, 2001; Franchina & Hopkins, 1996).

• 114. Visible rows of rectangular electrocytes above end of anal-fin base: (0) absent (Figure 1a–d); (1) present (Figure 1e; Albert, 2001, char. 229; Albert & Campos-da-Paz, 1998, char. 229; Lundberg & Mago-Leccia, 1986, char. 27; CI = 1.000, RI = 1.000).

The electric organ in Gymnotiformes comprises a series of electrocytes on the ventral portion of body. Lundberg and Mago-Leccia (1986) proposed the presence of visible rectangular electric organ subunits (electrocytes) above the end of anal-fin base, as a synapomorphy for a clade that includes: R. caviceps, R. eastwardi, R. stewardi, and R. troscheli. Further up, the authors recognized the presence of these visible electrocytes in other species of the genus (R. electrogrammus and R. zareti). However, the exposition areas of electrocytes are less visible in these species (as present in character 17).

The shape of these electrocytes also presents a puzzling variation in Eigenmannia. As discussed by Lundberg and Mago-Leccia (1986) and Albert (2001), the shape of these electrocytes in Eigenmannia is somehow different from those present in Rhabdolichops [e.g., Lundberg and Mago-Leccia (1986) described the electrocytes in Eigenmannia as hexagonal shaped]. However, although recognize the potential information in the electrocytes morphology, we did not prepare specimens to properly address this information (e.g., via histological preparation—see Giora & Carvalho, 2018). Therefore, the potential variation in electrocytes shape was not explored in the present analysis. Consequently, this character refers only to the presence of visible rectangular shaped electrocytes above the end of anal-fin base. This character is polymorphic in R. electrogrammus (see Lundberg & Mago-Leccia, 1986: Table 1).

3.1.19 Myology

• 115. Attachment of adductor mandibulae: (0) pars malaris not inserted to infraorbital 1 + 2; (1) pars malaris inserted to infraorbital 1 + 2 (see Peixoto & Ohara, 2019: fig. 11; modified from Albert, 2001: char. 45; Albert & Campos-da-Paz, 1998: char. 46; Albert et al., 2005: char. 82; Peixoto & Ohara, 2019; Tagliacollo et al., 2016b: char. 57; CI = 1.000, RI = 1.000).

Gymnotiformes exhibit a wide variation related to the pars malaris insertion, running from the lower jaw (e.g., Gymnotidae) to only to the maxilla (e.g., Apteronotidae; Aguilera, 1986). Only in Sternopygidae, the pars malaris inserts primarily onto the mesial face of the infraorbital 1 + 2, with the medial most fibers converging into a small endomaxilar ligament, which, in turn, is attached to the mesial face of the maxillary bone posterodorsal portion. Albert and Campos-da-Paz (1998) and Albert (2001) proposed the apomorphic condition as a synapomorphy for the Sternopygidae. However, in these studies, the pars malaris was proposed as attached to the infraorbital as a homoplastic condition found in Rhamphichthys. After myological analysis in Rhamphichthys, it was possible to detect that the pars malaris converges into an elongated endomaxilar ligament and is inserted on the antorbital and maxillary bones (Peixoto & Ohara, 2019). As result, it became clear that the pars malaris occurs in different sites in the Rhamphichthyidae and Sternopygidae, and thus, has non-homologous sites of insertion. Albert et al. (2005; char. 82) proposed a character comparable to the one described herein. However, their codification comprises two distinct variables (insertion versus muscular division), and thus, it is not followed in the present contribution. This is also true for character 57 of Tagliacollo et al. (2016b), who include variations related to two adductor mandibulae subsections: adductor mandibulae, pars malaris (= A1) and pars, rictalis (= A2). De la Hoz and Chardon (1984: 13) described two subsections (A.1a and A1b) to the adductor mandibulae, pars malaris. Our observations corroborate the myological pattern found by other several authors (e.g., Aguilera, 1986; Datovo & Vari, 2014), in which the pars malaris is compounded of a single mass of fibers, without differentiation in subsections.

• 116. Topographical relationship among adductor mandibulae, pars malaris and adductor mandibulae, pars stegalis: (0) pars stegalis partially located medially to pars malaris (see Peixoto & Ohara, 2019: fig. 11); (1) pars stegalis fully located medially to pars malaris (Figure 16; CI = 1.000, RI = 1.000).

In most Gymnotiformes, the pars malaris is located laterally only to the mid-ventral portion of the pars stegalis, resulting in a partial overlap between these muscles, with pars stegalis visible in lateral view. Conversely, Archolaemus (except A. orientalis) exhibits the pars stegalis compressed dorsoventrally, resulting in a fully medial position in relation to the pars malaris.

• 117. Origin of the adductor mandibulae, pars stegalis: (0) including sphenotic; (1) not including sphenotic (Peixoto & Ohara, 2019; CI = 0.500, RI = 0.875).

The adductor mandibulae, pars stegalis in Gymnotiformes originates mostly from the suspensorium and, frequently, osteological elements of the neurocranium (e.g., sphenotic, pterosphenoid, and parasphenoid). In Eigenmanniinae (except E. correntes), the pars stegalis has its origin shifted ventrally, and restricted to the bones of the suspensorium, not including the sphenotic as a site of its origin.

• 118. Association between pterosphenoid and origin of the adductor mandibulae, pars stegalis: (0) not including the pterosphenoid; (1) including the pterosphenoid (CI = 0.333, RI = 0.833).

In Gymnotus, the adductor mandibulae, pars stegalis, and pars malaris are largely continuous with one another (Datovo & Vari, 2014; fig. 19), forming a compound adductor mandibulae, pars stego-malaris, where the lateral most fibers correspond to the pars malaris, and the medial most fibers are presumed homologous to the pars stegalis (Datovo & Vari, 2013). The medial most fibers of the adductor mandibulae, thus the pars stegalis, in Gymnotidae and Eigenmanniinae (except Japigny and Archolaemus) originates from the suspensorium and the neurocranium but does not include the pterosphenoid as a site of origin. In contrast, the pars stegalis possesses the central axis posteromedially oriented toward its origin on the posteroventral margin of the pterosphenoid in the Apteronotidae, Hypopomidae, Archolaemus, Japigny, and Sternopygus.

• 119. Topographical relationship among adductor mandibulae, pars stegalis, and adductor arcus palatini: (0) pars stegalis fully lateral to adductor arcus palatini (see Peixoto & Ohara, 2019: fig. 13B); (1) pars stegalis lateral only to posterior portion of adductor arcus palatini (Figure 41); (2) pars stegalis lateral only to region near insertion of the adductor arcus palatini (Figure 16; CI = 0.500, RI = 0.818).

In the generalized gymnotiformes, the adductor mandibulae, pars stegalis is located laterally to the adductor arcus palatini, resulting in a complete overlapping between these muscles, with the adductor arcus palatini not visible in lateral view. In most Eigenmanniinae species and Sternarchorhamphus, the pars stegalis is only located laterally to the posterior portion of the adductor arcus palatini, which exhibits its anterior portion visible in lateral view. In D. conirostris and Archolaemus (except A. orientalis), the pars stegalis is elongated and shifted ventrally, and as the result, the pars stegalis is located laterally just near the insertion of the adductor arcus palatini, which, in turn, is almost totally visible.

• 120. Adductor mandibulae, segmentum mandibularis: (0) absent; (1) present (CI = 0.250, RI = 0.571).

The portion of the adductor mandibulae inserted on the medial face of the lower jaw is named segmentum mandibularis (Datovo & Vari, 2013). It originates from a mandibular tendon and, when present in Gymnotiformes, is characterized as a single element, without subsections.

• 121. Segmentum mandibularis extension in relation to Meckel's cartilage: (0) contact between segmentum mandibularis and Meckel's cartilage occurring up to 80% of dorsal margin of the cartilage (Figure 42a); (1) contact between segmentum mandibularis and Meckel's cartilage occurring more than 95% of dorsal margin of the cartilage (Figure 42b; CI = 1.000, RI = 1.000).

When present, the segmentum mandibularis in Gymnotiformes is relatively small, contacting up to 80% of the dorsal margin of Meckel's cartilage. In Rhabdolichops (except R. lundbergi and R. nigrimans), the segmentum mandibularis is hypertrophied and contact more than 95% of Meckel's cartilage. This character is inapplicable in Gymnotus, Sternarchorhamphus, and Archolaemus (except A. luciae), due to the absence of the segmentum mandibularis in members of these taxa.

• 122. Insertion of segmentum mandibularis: (0) including angulo-articular and dentary (Figure 42a–b); (1) restricted to angulo-articular (CI = 0.333, RI = 0.000).

In most Gymnotiformes, the sites of the segmentum mandibularis insertion include the medial face of the angulo-articular and dentary. In contrast, the segmentum mandibularis is reduced, and its insertion occurs only on the angulo-articular, not including the dentary or other osteological elements of the lower jaw.

• 123. Origin of the adductor arcus palatini (Figure 43): (0) not including orbitosphenoid; (1) including orbitosphenoid (CI = 0.500, RI = 0.000).

In most Gymnotiformes, the adductor arcus palatini usually originates from the ventrolateral surface of the parasphenoid and the anteroventral portion of the prootic (Aguilera, 1986). Conversely, in E. oradens and Japigny, the origin of the adductor arcus palatini is shifted anterodorsally, resulting in a prominent expansion of its anterodorsal portion, including the lateroventral face of the orbitosphenoid as a site of origin.

• 124. Adductor arcus palatini: (0) insertion occurring at base of endopterygoid, not including ascending process; (1) insertion occurring at base of endopterygoid and including ascending process (Figure 43; CI = 0.500, RI = 0.000).

The ascending process of the endopterygoid is an ossification of the pterygocranial ligament (Albert, 2001). This structure is present in the anterolateral face of the endopterygoid and connects the base of this bone to the neurocranium anteroventral surface. In most Gymnotiformes, the adductor arcus palatini inserts only on the endopterygoid base. In contrast, the adductor arcus palatini shows a slightly anterolateral expansion and, as result, this muscle inserts in the medial face of the median portion of the ascending process of the endopterygoid in A. luciae and Japigny. This character was coded as non-applicable in Apteronotus and Sternarchorhamphus, due to the absence of ossification of the pterygocranial ligament.

• 125. Origin of levator arcus palatini: (0) including frontal; (1) not including frontal (CI = 0.167, RI = 0.737).

In Gymnotiformes, the levator arcus palatini originates from the sphenotic, usually including the frontal, and, less frequently, the pterosphenoid (Aguilera, 1986).

• 126. Association between pterosphenoid and origin of levator arcus palatini: (0) not including pterosphenoid; (1) including pterosphenoid (CI = 0.250, RI = 0.750).

In most Gymnotiformes, the origin of the levator arcus palatini does not include the pterosphenoid. In contrast, the main axis of the levator arcus palatini anterior portion is expanded medially, consequently, medial fibers arise from the pterosphenoid posterodorsal portion.

• 127. Origin of levator arcus palatini: (0) width as broad as or broader than at its insertion; (1) narrower than its insertion (CI = 0.250–0.333, RI = 0.700–0.800).

The origin of the levator arcus palatini is as broad as or broader than its insertion in most Gymnotiformes. Conversely, in some Eigenmanniinae and Sternarchorhamphus, the levator arcus palatini originates mostly from the ventral portion of the sphenotic spine, becoming broader toward its insertion, resulting an origin narrower than its portion of attachment with the hyomandibula.

• 128. Anterolateral fibers of main axis of levator arcus palatini (Peixoto & Ohara, 2019): (0) oblique in relation to vertical arm of preopercle (see Peixoto & Ohara, 2019: fig. 13B); (1) approximately straight in relation to vertical arm of preopercle (Figure 41; CI = 0.250, RI = 0.667).

In the general condition found in Gymnotiformes, the levator arcus palatini has its position oblique in relation to the vertical arm of the preopercle, forming an angle ca. 45° with the longitudinal axis of the head. In contrast, in most Eigenmanniinae and Sternarchorhamphus, this muscle has its anterior portion approximately straight in relation to the vertical arm of the preopercle, resulting an angle ca. 90° with the longitudinal axis of the head.

• 129. Levator arcus palatini: (0) wider at origin; (1) origin and insertion with similar width (see Peixoto & Ohara, 2019: fig. 13B); (2) wider at insertion (Figures 15 and 40; CI = 1.000, RI = 1.000).
• 130. Relationship among R-Avn nerve and levator operculi anterior (Peixoto & Ohara, 2019): (0) R-Avn nerve passing laterally to levator operculi anterior (see Peixoto & Ohara, 2019: fig. 13B); (1) R-Avn nerve passing medially to levator operculi anterior (Figures 16 and 41; CI = 0.500, RI = 0.889).

The levator operculi is a laminar and superficial muscle, located immediately posterior to dilatator operculi and laterally to adductor operculi. The general configuration found in the Teleostei consists of a muscle as a single myological element, without subsections, arising from the neurocranium posterior bones and attaching in the opercle (Winterbottom, 1974). In Gymnotiformes, the levator operculi is characterized by two sections, designated as the levator operculi anterior and the levator operculi posterior (Aguilera, 1986), and the R-Avn nerve is located laterally to levator operculi anterior and medially to the levator operculi posterior. Distinct from the generalized pattern found in Gymnotiformes, Eigenmanniinae species, except in R. lundbergi and R. nigrimans, where the R-Avn nerve is located medially to the levator operculi anterior.

• 131. Origin of levator operculi: (0) not including hyomandibula; (1) including hyomandibula (CI = 0.500 RI = 0.889).

The anterior fibers of the levator operculi, corresponding to the levator operculi anterior in all Gymnotiformes analyzed herein, originate from the pterotic (Aguilera, 1986; de La Hoz & Chardon, 1984). In Eigenmanniinae species, except R. lundbergi and R. nigrimans, the levator operculi anterior has an anterodorsal expansion, and, consequently, the hyomandibula posterodorsal margin is considered as a site of origin.

• 132. Muscle forming posteroventral limit of muscular hiatus between first and second rib (Dutra et al., 2015, char. 3): (0) obliquus inferioris (see Dutra et al., 2015: fig. 1); (1) obliquus superioris (see Dutra et al., 2015: fig. 2E–F for comparable condition in Steatogenys; CI = 0.200, RI = 0.636).

The pseudotympanum is formed by the reduction of the hypaxialis muscle in the body wall lateral to the anterior portion of the swim bladder. This structure is present in all Gymnotiformes, and some informative characters were recently pointed out by Dutra et al. (2015). The delimitation of the hiatus h3 posteroventral margin varies due to the degree of reduction of the obliquus superioris in that area exposing or not the obliquus inferioris.

3.1.20 Neuroanatomy

• 133. Position of anterior portion of the lobus vagi (#2): (0) convergent (Figure 45a–c,e–g); (1) parallel (Figure 45d; CI = 0.500, RI = 0.500).

The lobus vagi is compounded by two rod-shaped somewhat curved structures (Figure 44). Such lobe is located in the dorsal portion of the rhombencephalon and positioned posterior to the corpus cerebelli (ventral to posterior portion of it, in some species). The posterior portion of the lobus vagi has a tip-shaped ending, whereas its anterior portion can be convergent or parallel, without contact between its counterparts (Figure 44). In the convergent condition, the lobus vagi is somewhat horseshoe-shaped, resulting in a greater proximity between the anterior portion of its counterparts. In the parallel condition, such lobe is somewhat U-shaped, resulting in a parallel position between its counterparts.

• 134. Posterior elongation of electrosensory lateral line lobe (#3) in relation to lobus vagi (#2): (0) exceeding posteriorly (Figure 45a,c); (1) in posterior margin (Figure 45b,e); (2) in half length (Figure 45d,f,g; CI = 0.500, RI = 0.800).

The electrosensory lateral line lobe (ELL) is located in the dorsal region of the rhombencephalon and positioned lateroventrally to the eminentia granularis (Figure 44). In the most elongated condition, the vertical line through the posterior margin of the ELL is positioned posteriorly in relation to the lobus vagi. In the intermediate condition, the posterior margin of the ELL is aligned to the posterior margin of the lobus vagi. In the less elongated condition of the ELL, its posterior margin reaches only the half length of the lobus vagi.

• 135. Posterior elongation of eminentia granularis (#4) in relation to lobus vagi (#2): (0) exceeding posteriorly (Figure 45e); (1) in posterior margin (Figure 45a,c); (2) until half length (Figure 45b); (3) positioned anteriorly (Figure 45d,f–g; CI = 0.600, RI = 0.778).

The eminentia granularis is also located in the dorsal region of the rhombencephalon and positioned latermedially to the ELL (Figure 44). In the most elongated condition, the vertical line through the posterior margin of the eminentia granularis is positioned posteriorly in relation to the lobus vagi. In the intermediate conditions, the posterior margin of the eminentia granularis extends until to posterior margin, or to the half length of the lobus vagi. In the shorter condition, the eminentia granularis extends only until the anterior portion of the lobus vagi.

• 136. Posterior elongation of corpus cerebelli (#6) in relation to lobus vagi (#2): (0) positioned anteriorly (Figure 45d); (1) until anterior margin (Figure 45f–g); (2) in half length (Figure 45a–c); (3) positioned posteriorly (Figure 45e; CI = 0.750, RI = 0.917).

The corpus cerebelli is located in the dorsal region of the rhombencephalon, positioned medially to the eminentia granularis (Figure 44). In the shorter condition, the posterior margin of the corpus cerebelli is positioned anteriorly to the lobus vagi. In the intermediate conditions, the posterior margin of the corpus cerebelli reaches the anterior margin, or the half length of the lobus vagi. In the most elongated condition, the posterior margin of the corpus cerebelli extends until to the posterior region of the lobus vagi.

• 137. Topographical relationship among corpus cerebelli (#6), tectum mesencephali (#7) and telencephalon (#14): (0) corpus cerebelli positioned above posterior portion of telencephalon (Figure 45a); (1) corpus cerebelli positioned dorsal to anterior margin of tectum mesencephali (Figure 45d–e); (2) corpus cerebelli positioned dorsal to anterior portion of tectum mesencephali (Figure 45f–g); (3) corpus cerebelli positioned dorsal to half length of tectum mesencephali (Figure 44b; CI = 0.600, RI = 0.818).

In the most elongated condition, the anterior margin of the corpus cerebelli reaches the posterior portion of the telencephalon. In the intermediate conditions observed among examined species, the anterior margin of the corpus cerebelli is positioned at the anterior margin, or at the anterior portion of the tectum mesencephali. In the shorter condition, the anterior margin of the corpus cerebelli is elongated until the half length of the tectum mesencephali.

• 138. Valvula cerebelli (#5): (0) not topographically visible between corpus cerebelli (#6) and tectum mesencephali (#7; Figure 45a,d); (1) topographically visible between corpus cerebelli and tectum mesencephali (Figure 45b–c,e–g; CI = 0.333, RI = 0.500).

The valvula cerebelli is located in the dorsal region of the rhombencephalon and positioned, when visible, laterally to the corpus cerebelli and dorsally to the tectum mesencephali (Figure 44).

• 139. Shape of valvula cerebelli in dorsal view (#5): (0) rod-shaped (Figure 45b); (1) ovoid-shaped (Figure 45c); (2) comma-shaped (Figure 45e–g; CI = 1.000, RI = 1.000).

This character is inapplicable in Gymnotus, Microsternarchus, and Sternopygus in which the valvula cerebelli is not visible in dorsal view.

• 140. Position of valvula cerebelli (#5) in relation to tectum mesencephali (#7) in lateral view: (0) covering posterior portion of tectum mesencephali (Figure 45e–g); (1) covering posteromedial portion tectum mesencephali (Figure 45b); (2) covering hole tectum mesencephali (Figure 45c; CI = 1.000, RI = 1.000).

This character is inapplicable in Gymnotus, Microsternarchus, and Sternopygus in which the valvula cerebelli is not visible in lateral view.

• 141. Position of nucleus electrosensorius (#8): (0) on vertical line through anterior most portion of tectum mesencephali (#7; Figure 45a); (1) extending beyond vertical line through anterior most portion of tectum mesencephali (Figure 45b–g; CI = 1.000, RI = 1.000).

The nucleus electrosensorius is located in the anterior margin of the tectum mesencephali and positioned ventrally to it, and posterior to the telencephalon (#14; Figure 44). This pretectal structure is normally short, round-shaped, and extends until the vertical line through the anteriormost portion of the tectum. When elongated, the nucleus electrosensorius reaches the posterior portion of the telencephalon ventrally, extending beyond the vertical line through the anteriormost portion of the tectum mesencephali.

• 142. Position of anterior margin of lateral preglomerular nucleus (#12): (0) beneath telencephalon (#14; Figure 45a–b); (1) posterior to vertical line through posterior margin of telencephalon (Figure 45c–g; CI = 0.500, RI = 0.750).

The lateral preglomerular nucleus is located in the anterior portion of the diencephalon and positioned ventral to the nucleus electrosensorius and anterior to the lobus inferior hypothalami (Figure 44). When elongated, the anterior margin of the lateral preglomerular nucleus reaches the posterior portion of the telencephalon, ventrally. In the shorter condition, the lateral preglomerular nucleus is more reduced and its anterior portion does not reach the posterior portion of the telencephalon ventrally.

• 143. Position of telencephalon (#14): (0) above posterior portion of bulbus olfactorius; (1) above bulbus olfactorius. (CI = 0.200, RI = 0.600).

The telencephalon is the anteriormost part of the encephalon (Figure 44). The telencephalon is normally short, covering only the posterior portion of the bulbus olfactorius. When elongated, the telencephalon is positioned over the entire bulbus olfactorius, covering entirely its dorsal surface.

• 144. Thickness of nervus opticus: (0) thinner than nervus olfatorius (Figure 46a–b); (1) thicker than nervus olfatorius (Figure 46c; CI = 1.000, RI = 1.000).

The nervus opticus connects the ventral portions of the tectum mesencephali to the retina in the eyes, whereas the nervus olfactorius connects the bulbus olfactorius to the olfactory organ.

3.2 Phylogenetic reconstruction

The analysis of the phylogenetic relationships among the species of Eigenmanniinae, based on 144 morphological characters of 45 taxa (37 representing species of Eigenmanniinae and eight outgroups representing other gymnotiform groups—Tables S5–S6, Data S2), resulted in two most parsimonious trees, with 389.152 score, and CI of 0.399 and RI of 0.754 (Figure 47).

3.2.1 Monophyly of Eigenmanniinae Mago Leccia, 1978 (Clade 1)

Eigenmanninae Mago-Leccia, 1978: 14 (type genus: Eigenmannia Jordan & Evermann, 1896). Albert & Campos-da-Paz, 1998: 439 (phylogeny). Albert, 2001: 71 (phylogeny). Tagliacollo et al., 2006a: 31 (phylogeny).

Eigenmanniidae Alves-Gomes et al., 1995: 312 (phylogeny).

Eigenmanniinae Dutra et al., 2014: 346 (position of Distocyclus). Peixoto & Ohara, 2019:18 (discussion on myology).

Included genera

Archolaemus, Distocyclus, Eigenmannia, Japigny, Rhabdolichops, and Rhinosternarchus gen. nov.

Phylogenetic diagnosis

Eigenmanniinae is supported by the following unambiguous synapomorphies: (1) decrease of number of teeth in upper pharyngeal plate (char. 7, 13–15>12); (2) increase of number of pectoral-fin rays (char. 9, 15–17>18–20); (3) decrease of number of precaudal vertebrae, 14 or 15 (char. 11, 17–18>14–15); (4) decrease of number of pleural ribs (char. 13, 12–13>8); (5) site of connection between infraorbital and supraorbital canals on sphenotic process (char. 59, 0>1); (6) free urohyal blade as long as urohyal ridge (char. 64, 0>1); (7) supracleithrum and posttemporal fused (char. 89, 0>1); (8) scapular foramen entirely included within scapula (char. 91, 0>1). (9) parapophysis of second vertebra ventrally curved (char. 95, 0>1); (10) epicentrals at 7–9^th vertebrae branched (char. 101, 0>1); (11) presence of anal-fin base stripe (char. 109, 0>1); (12) origin of the adductor mandibulae, pars stegalis not including the sphenotic (char. 117, 0>1); (13) Adductor mandibulae, pars stegalis lateral only to posterior portion of adductor arcus palatini (char. 119, 0>1); (14) origin of the levator arcus palatini narrower than its insertion (char. 127, 0>1); (15) R-Avn nerve passing medially to levator operculi anterior (char. 130, 0>1); (16) origin of the levator operculi including the hyomandibula (char. 131, 0>1); (17) Telencephalon located above bulbus olfactorius (char. 143, 0>1).

3.2.2 Japigny Meunier et al., 2011

Japigny Meunier et al., 2011: 48 [type species: Japigny kirschbaum Meunier et al., 2011; type by original designation (also monotypic). Gender feminine].

Included species

Japigny kirschbaum Meunier et al., 2011.

Phylogenetic diagnosis

Japigny is diagnosed by fourteen autapomorphies: (1) decrease of number of endopterygoid teeth (char. 5, 7–11>4); (2) decrease of number of anal-fin rays (char. 10, 216–217>132–164); (3) increase of number of displaced hemal spines (char. 14, 3>4); (4) mouth subterminal (char. 18, 1>2); (5) presence of teeth on oral valve (char. 21, 0>1); (6) dentary teeth attached on a dorsolateral flange of this bone (char. 35, 0>1); (7) third branchiostegal slender (char. 68, 0>1); (8) fifth branchiostegal slender (char. 69, 0>1); (9) basibranchial 3 unossified (char. 73, 1>0); (10) rami arrangement of epineurals at 7–9^th in several directions (char. 100, 0>1); (11) presence of vertical bands on body (char. 105, 0>1); (12) Segmentum mandibularis insertion restricted to angulo-articular (char. 122, 0>1); (13) origin of the adductor arcus palatini including the orbitosphenoid (char. 123, 0>1); and (14) Adductor arcus palatini insertion on endopterygoid occurring at the base of the endopterygoid, and including the ascending process (char. 124, 0>1).

Taxonomic diagnosis

In addition to the aforementioned characters under the phylogenetic diagnosis, Japigny differs from Archolaemus by having the eye covered by skin (versus free orbital rim), and anteriormost row of the premaxilla teeth completely attached to the ventral surface of the premaxilla (versus teeth attached only along their anterobasal margins). It differs from Distocyclus by the presence of endopterygoid teeth (versus absence), mouth subterminal (versus terminal), and dentary teeth row extending posteriorly beyond anterior limit of Meckel’s cartilage (versus limited to anterior portion of dentary). Japigny can be distinguished from “E.” goajira by having the mouth subterminal (versus terminal), the snout elongate with the upper jaw distinctly longer than the lower jaw (versus the snout pointed but short turning it about equally developed as the dentary), and 132–164 (versus 259–263) anal-fin rays. It can be diagnosed from Rhabdolichops by the presence scales on anterodorsal portion of body (versus absence of such scales), the tip of the pectoral fin never extending beyond (versus extending beyond) the end of body cavity and the absence (versus presence) of visible rows of electrocytes above the end of anal-fin base.

3.2.3 Clade 2

Included genera

Archolaemus, Distocyclus, Eigenmannia, Rhabdolichops and Rhinosternarchus.

Phylogenetic diagnosis

The Clade 2 is diagnosed by nine unambiguous synapomorphies: (1) coronomeckelian bone corresponding to 20% or less of the length of Meckel’s cartilage (char. 39, 0>1); (2) anterior process of the coracoid contacting the main crest of cleithrum on its anterior portion (char. 85, 0>1); (3) parapophyses of the second and fourth vertebrae in contact (char. 96, 0>1); (4) hemal spine of 26th to 30th vertebrae longer than its associated pterygiophore (char. 99, 0>1); (5) origin of the adductor mandibulae, pars stegalis not including the pterosphenoid (char. 118, 1>0); (6) orientation of the anterolateral fibers of main axis of the levator arcus palatini approximately straight in relation to the vertical arm of the preopercle (char. 128, 0>1); (7) posterior portion of the corpus cerebelli extending to the anterior margin of the lobus vagi (char. 136, 0>1); (8) corpus cerebelli positioned dorsal to the anterior portion of the tectum mesencephali (char. 137, 1>2); and (9) Valvula cerebelli comma-shaped in dorsal view (char. 139, 0>2).

3.2.4 Rhabdolichops Eigenmann & Allen, 1942

Rhabdolichops Eigenmann & Allen, 1942: 316 (type species: Rhabdolichops longicaudatus Eigenmann & Allen, 1942; type by monotypy. Gender masculine).

Included species

Rhabdolichops caviceps (Fernández-Yépez 1968), R. eastwardi Lundberg & Mago-Leccia, 1986, R. electrogrammus Lundberg & Mago-Leccia, 1986, R. jegui Keith & Meunier, 2000, R. lundbergi Correa et al., 2006, R. navalha Correa et al., 2006, R. nigrimans Correa et al., 2006, R. stewarti Lundberg & Mago-Leccia, 1986, R. troscheli (Kaup, 1856), and R. zareti Lundberg & Mago-Leccia, 1986.

Phylogenetic diagnosis

Rhabdolichops is diagnosed by twelve unambiguous synapomorphies: (1) increase of number of gill rakers on first branchial arch (char. 6: 11>12–14); (2) decrease of number of precaudal vertebrae (char. 11: 14>12–13); (3) length of premaxilla corresponding to four times as long as its width (char. 24: 0>1); (4) absence of an anterolateral process on premaxilla (char. 26: 1>0); (5) posterodorsal margin of hyomandibula ending at same level of condyle that receives the opercle socket (char. 48: 0>1); (6) depth of the posterodorsal laminar expansion of infraorbital 1+2 half or less as long as infraorbital 1+2 length (char. 56: 1>0); (7) otic canal enlarged and half-pipe shaped (char. 60: 0>1); (8) extrascapular enlarged and half-pipe shaped (char. 62: 0>1); (9) urohyal ridge triangular (char. 63: 1>2); (10) mesethmoid 1.5 times as long as nasal length (char. 74: 1>0); (11) tip of pectoral fin surpassing the end of abdominal cavity (char. 94: 0>1); and (12) scales on anterodorsal portion of body absent (char. 112: 0>1).

Taxonomic diagnosis

In addition to the aforementioned characters under the phylogenetic diagnosis, Rhabdolichops differs from Archolaemus by having the eye covered by skin (versus orbital rim free). It differs from Distocyclus by having the snout rounded (versus conical), the internarial distance equal to or at least twice the diameter of the nostril, and the dentary teeth row extending posteriorly beyond the anterior limit of Meckel’s cartilage (versus limited to dentary anterior portion). Rhabdolichops further differs from Japigny by the absence (versus presence) of eight dark, vertical bands along the body, and the absence (versus presence) of teeth associated with the oral valve. It further differs from Rhinosternarchus by having a round (versus subconical) snout.

3.2.5 Clade 3

Included genera

Archolaemus, Distocyclus, Eigenmannia and Rhinosternarchus.

Phylogenetic diagnosis

The Clade 3 is diagnosed by five unambiguous synapomorphies: (1) increase of number of dentary teeth (char. 3: 24–40>23); (2) increase of number of teeth in lower pharyngeal plate (char. 8: 9>10–11); (3) Baudelot’ ligament partially ossified (char. 90: 0>1); (4) neural spine of four vertebra of weberian apparatus long spine-shaped (char. 97: 2>0); and (5) Levator arcus palatini wider at insertion (char. 126: 1>2).

3.2.6 Eigenmannia Jordan & Evermann, 1896

Eigenmannia Jordan & Evermann, 1896: 340 (type species: Sternopygus humboldtii Steindachner, 1878; type by being a replacement name of Cryptops Eigenmann, 1894, preoccupied by Cryptops Leach 1814 in Myriopoda, Cryptops Schoenherr, 1823 and Cryptops Solier 1851 in Coleoptera. Gender feminine).

Included species

Eigenmannia antonioi Peixoto et al., 2015, E. besouro Peixoto & Wosiacki, 2016, E. camposi Herrera-Collazos et al., 2020, E. correntes Campos-da-Paz & Queiroz, 2017, E. desantanai Peixoto et al., 2015, E. dutrai Peixoto et al., 2021, E. guchereauae (Meunier et al., 2014), E. guairaca Peixoto et al., 2015, E. humboldtii (Steindachner, 1878), E. limbata (Schreiner & Miranda-Ribeiro, 1903), E. loretana Waltz & Albert, 2018, E. macrops (Boulenger, 1897), E. magoi Herrera-Collazos et al., 2020, E. matintapereira Peixoto et al., 2015, E. meeki Dutra et al., 2017, E. microstoma (Reinhardt, 1852), E. muirapinima Peixoto et al., 2015, E. nigra Mago-Leccia, 1994, E. oradens Dutra et al., 2018, E. pavulagem Peixoto et al., 2015, E. sayona Peixoto & Waltz, 2017, E. sirius Peixoto & Ohara, 2019, E. trilineata López & Castello, 1966, E. vicentespelaea Triques, 1996, E. virescens (Valenciennes, 1936), E. waiwai Peixoto et al., 2015, and E. zenuensis Herrera-Collazos et al., 2020.

Phylogenetic diagnosis

Eigenmannia is diagnosed by four unambiguous synapomorphies: (1) increase in number of premaxillary teeth, 30–37 (char. 1: 21–25>30–37); (2) increase in number of teeth rows on premaxilla (char. 2: 4>5); (3) presence of “epipleurals” at 7–9th vertebrae (char. 102: 0>1); and (4) Nervus opticus thicker than nervus olfactorius (char. 144: 0>1).

Taxonomic diagnosis

In addition to the aforementioned characters under the phylogenetic diagnosis, Eigenmannia differs from Archolaemus by having the eye covered by skin (versus free orbital rim), and the teeth of premaxilla anteriormost teeth row completely attached to ventral surface of premaxilla (versus attached only along their anterobasal margins—also present in E. guchereauae and E. oradens). It can be distinguished from Distocyclus by the presence (versus absence) of endopterygoid teeth, and the dentary teeth row extending posteriorly beyond anterior limit of Meckel’s cartilage (versus limited to anterior portion of dentary). Eigenmannia is diagnosed from Japigny by not having teeth associated with the oral valve (versus teeth present), and the absence (versus presence) of eight dark, vertical bands along the body. It can be also diagnosed from Rhabdolichops by the presence scales on anterodorsal portion of body (versus absence of such scales), the tip of pectoral fin never extending beyond the end of body cavity (versus extending beyond). Eigenmannia can be further distinguished from Rhinosternarchus by having a round or pointed (versus subconical—see Dutra et al., 2014) snout.

3.2.7 Clade 4

Included genera

Archolaemus, Distocyclus and Rhinosternarchus.

Phylogenetic diagnosis

The clade 4 is diagnosed by five unambiguous synapomorphies: (1) absence of the posterodorsal laminar expansion in Infraorbital 1+2 (char. 55: 1>0); (2) mesethmoid length corresponds to at least three times of length of nasal bone (char. 73: 1>2); (3) posterodorsal process of lateral ethmoid longer than main axis of ethmoid lateral (char. 75: 1>0); (4) posterior margin of the ELL extends to posterior limit of the lobus vagi (char. 131: 2>1); and (5) posterior portion of the corpus cerebelli surpassing the posterior portion of the lobus vagi (char. 133: 2>3).

3.2.8 Distocyclus Mago-Leccia, 1978

Distocyclus Mago-Leccia, 1978: 267 (type species: Eigenmannia conirostris Eigenmann & Allen, 1942; type by original designation. Gender masculine).

Included species

Distocyclus conirostris (Eigenmann & Allen, 1942).

Phylogenetic diagnosis

Distocyclus is diagnosed by eleven autapomorphies: (1) decrease of number of premaxillary teeth, 14–19 (char. 1: 20–25>14–19); (2) decrease of number of teeth rows on premaxilla (char. 2: 4>3); (3) decrease of number of dentary teeth (char. 3: 23>1–6); (4) decrease of number of teeth rows on dentary (char. 4: 2–3>1); (5) internarial distance equivalent to diameter of posterior nostril (char. 22: 0>1); (6) teeth row of dentary not reaching anterior limit of Meckel’s cartilage (char. 32: 0>1); (7) coronoid process ventrally curved (char. 36: 1>0); (8) retroarticular limiting the ventralmost margin of lower jaw (char. 41: 1>0); (9) endopterygoid edentulous (char. 43: 1>0); (10) urohyal ridge triangular (char. 63: 1>2); and (11) distal margin of anal fin dark (char. 110: 0>1).

Taxonomic diagnosis

Dutra et al. (2014) provided a taxonomic diagnosis of Distocyclus based on external characters.

3.2.9 Clade 5

Included genera

Archolaemus and Rhinosternarchus.

Phylogenetic diagnosis

The clade 5 is diagnosed by three unambiguous synapomorphies: (1) premaxillary bone compact, longitudinal length approximately equal to its transverse width (char. 25: 0>1); (2) absence of anterolateral process on the premaxillary bone (char. 26: 1>0); and (3) hemal spine of 26th to 30th vertebrae longer than its associated pterygiophore (char. 99: 1>0).

3.2.10 Rhinosternarchus gen. nov.

urn:lsid:zoobank.org:act:B43E0006-FE9E-40DC-8753-A394A89B5BE9.

Type species. Eigenmannia goajira Schultz, 1949, by monotypy.

Included species

Rhinosternarchus goajira (Schultz, 1949; Figure 48).

Phylogenetic diagnosis

Rhinosternarchus is diagnosed by seven autapomorphies: (1) decrease of number of endopterygoid teeth (char. 5: 5–8>3–4); (2) increase of number of anal-fin rays (char. 10: 216–217>259–263); (3) decrease of number of pleural ribs (char. 13: 8–9>6); (4) decrease of number of displaced hemal spines (char. 14: 3>2); (5) anterior portion of endopterygoid approximately twice its width at ascending process (char. 44: 0>1); (6) antorbital portion of the parasphenoid longer than its orbital portion (char. 80: 0>1); and (7) anal-fin base stripe absent (char. 109: 1>0).

Taxonomic diagnosis

In addition to the aforementioned characters under the phylogenetic diagnosis, Rhinosternarchus can be diagnosed from Archolaemus by possessing the mouth terminal (versus subterminal), and the eye covered by skin (versus with a free orbital rim). It differs from Distocyclus by having a subconical (versus conical) snout, the internarial distance equal to at least twice the diameter (versus equivalent to a diameter of the posterior nostril), presence (versus absence) of teeth on the endopterygoid, and dentary with four rows (versus a single row of teeth). Rhinosternarchus is diagnosed from Eigenmannia by having a subconical (versus rounded or pointed) snout. It is diagnosed from Japigny by having the mouth terminal (versus subterminal), teeth associated with the oral valve absent (versus teeth present), and the absence (versus presence) of eight dark, vertical bands along the body. Rhinosternarchus also differs from Rhabdolichops by having a subconical (versus round) snout, the presence scales on the anterodorsal portion of body (versus absence of such scales), and the tip of pectoral fin never extending beyond (versus extending beyond) end of the body cavity.

Etymology

The genus name is from the Greek “rhino,” meaning nose, in reference to the elongated snout, and “sternarchus,” a commonly scientific suffix used in Gymnotiformes nomenclature, from the Greek “sternon,” chest, and “archus,” rectum, an allusion to the anterior position of the anus, a common feature in the order.

3.2.11 Archolaemus Korringa, 1970

Archolaemus Korringa, 1970: 267 [type species: Archolaemus blax Korringa, 1970; type by original designation (also monotypic). Gender masculine].

Included species

Archolaemus blax Korringa, 1970, A. ferreirai Vari et al., 2012, A. janeae Vari et al., 2012, A. luciae Vari et al., 2012, A. orientalis Stewart, Vari et al., 2012, A. santosi Vari et al., 2012.

Phylogenetic diagnosis

Archolaemus is diagnosed by five unambiguous synapomorphies: (1) decrease of number of anal-fin rays (char. 10: 216–217>205–213); (2) increase of number of precaudal vertebrae, 15 (char. 11: 14>15); (3) increase of number of scale rows above lateral line (char. 16: 13>14–15); (4) mouth subterminal (char. 18: 1>2); and (5) free orbital rim (char. 23: 0>1).

Taxonomic diagnosis

In addition to the aforementioned characters under the phylogenetic diagnosis, Archolaemus differs from all the other Eigenmanniinae genera by the teeth of the premaxilla anterior most row attached only along their anterobasal margins (versus anteriormost teeth completely attached to the ventral surface of the premaxilla in all Eigenmanniinae, except E. guchereauae and E. oradens). It further differs from Distocyclus by the presence of endopterygoid teeth (versus absence), and the dentary teeth row extending posteriorly beyond anterior limit of Meckel’s cartilage (versus limited to anterior portion of dentary). Archolaemus is also diagnosed from Japigny by the absence (versus presence) of eight dark, vertical bands along the body, and the absence (versus presence) of teeth associated with the oral valve. It can be also diagnosed from Rhabdolichops by the presence (versus absence) of scales on anterodorsal portion of body, the tip of pectoral fin never extending beyond the end of body cavity (versus pectoral-fin tip beyond that point), and presence of visible rows of electrocytes above end of anal-fin base (versus absence).

4 DISCUSSION

Eigenmanniinae was proposed by Mago-Leccia (1978) to include Eigenmannia, Archolaemus, and Distocyclus. Those three genera shared the scapular foramen entirely included within the scapula, the fusion of the post-temporal and the supracleithrum into a single ossification, and by having 11–15 precaudal vertebrae (Table S1). Subsequently, Rhabdolichops and Japigny were included in Eigenmanniinae (Albert, 2001; Albert & Campos-da-Paz, 1998; Fink & Fink, 1981; Mago-Leccia, 1994; Maldonado-Ocampo, 2011; Meunier et al., 2011; Tagliacollo et al., 2016a; Vari et al., 2012). In the present study, we recovered all synapomorphies proposed by Mago-Leccia (1978), as well as all three putative synapomorphies based on musculature proposed by Peixoto and Ohara (2019). Additionally, eleven new synapomorphies were found to support the monophyly of the subfamily (see in Phylogenetic diagnosis of Eigenmanniinae).

Japigny was previously proposed as the sister group of all the other Eigenmanniinae by molecular and morphological datasets (Maldonado-Ocampo, 2011; Vari et al., 2012). Vari et al. (2012) presented three characters shared by all Eigenmanniinae except Japigny (Clade 2), to justify their hypothesis of Japigny as the sister group of all other members of the subfamily (Table S7). Among them, the parapophyses of the second and fourth vertebrae in contact was recovered here as a synapomorphy of Clade 2, while the retroarticular included in the socket to receive the quadrate condyle can be ambiguously optimized due its presence in Japigny and Sternopygus. The parapophysis of the fourth vertebra curved was not utilized in this study. Reasons are discussed in Data S1 (U22). Later, Tagliacollo et al. (2016a) recovered Japigny intermingled within Eigenmannia. The species J. kirschbaum was then recently proposed as a member of Eigenmannia by Waltz and Albert (2017) based in the Tagliacollo’ et al. (2016a) hypothesis, according to which this species was recovered as sister group of E. macrops by sharing a subterminal mouth. Tagliacollo et al. (2016a), however, highlighted that the inclusion of Japigny in their total evidence analysis was based only on morphological characters, suggesting that result could be influenced by the absence of molecular data for this species. Yet, it is important to note that the characters proposed by Vari et al. (2012) to consider Japigny as sister group of the remaining Eigenmanniinae were not included in the analysis by Tagliacollo et al. (2016a). Thus, considering Japigny’s position as the sister group of all Eigenmanniinae as recovered herein, we believe that location of Japigny in Tagliacollo et al. (2016a) was also influenced by the absence of informative morphological characters for this species rather than just by the missing data of molecular traits only.

Several authors recognized the genus Rhabdolichops as monophyletic based on both morphology (Correa et al., 2006; Lundberg & Mago-Leccia, 1986) and molecules (Tagliacollo et al., 2016a). On the contrary, Maldonado-Ocampo (2011) rejected that hypothesis and emphasized that R. lundbergi Correa et al., 2006 and R. nigrimans Correa et al., 2006 more closely related to Eigenmannia than others Rhabdolichops. The monophyly of Rhabdolichops was recovered herein contrasting with Maldonado-Ocampo (2011) hypothesis, who considered these two species as sister group of E. humboldtii + E. limbata and suggested that R. lundbergi and R. nigrimans should be placed to Eigenmannia. Thirteen synapomorphies supported the monophyly of Rhabdolichops, and the clade R. lundbergi + R. nigrimans (Clade R2) was recovered as the sister group of the other species of the genus as proposed by Correa et al. (2006). Most of the synapomorphies proposed by Lundberg and Mago-Leccia (1986) for Rhabdolichops, support a less inclusive group within the genus (Clade R3), which includes all species except R. lundbergi and R. nigrimans, as discussed by Correa et al. (2006). These last two species were only described in 2006 and were not available for the study by Lundberg and Mago-Leccia (1986). Correa et al. (2006) defined Rhabdolichops based on six synapomorphies: (1) scales absent above lateral line on anterior portion of body without scales above lateral line; (2) premaxilla elongate; (3) extrascapular characterized as an independent ossification; (4) two prootic foramina; (5) gill rakers long and bony; and (6) gill rakers ossified (Table S8). Among the proposed synapomorphies, the absence of scales above the lateral line on anterior portion of body, and an elongate premaxilla were recovered as synapomorphies for the genus herein, and the remaining characters as synapomorphies for a less inclusive group (clade R3). Lundberg and Mago-Leccia (1986) and Correa et al. (2006) proposed R. electrogrammus as sister group of all remaining species for clade R3. We did not recover this hypothesis herein; instead, we recovered an unresolved trichotomy with R. electrogrammus, R. zareti, and Clade R4 (R. eastwardi + R. caviceps + R. troscheli).

The monophyly of Clade 3 (Archolaemus +Distocyclus + Eigenmannia +Rhinosternarchus), which is recovered herein, was previously recovered by both morphological (Mago-Leccia, 1978), and molecular datasets (Alda et al., 2019; Alves-Gomes, 1995, 1998; Maldonado-Ocampo, 2011). Mago-Leccia’s (1978) proposed four characters to support the monophyly of this clade: (1) gill rakers cartilaginous without osseous bases; (2) branchial opening reduced or only moderately so; (3) mouth reduced; and (4) variable position of the anus. However, none of these characters supported Clade 3 in the present study. The gill rakers in most Gymnotiformes are unossified, except in some Rhabdolichops species that present elongated and osseous gill rakers, a clear derived condition within members of the genus (char. 70, state 1, see topic 3.2.4.2 for Rhabdolichops phylogenetic diagnosis). On the contrary, the size of the branchial opening did not present a conspicuous variation across the analyzed specimens. The mouth reduced, and the variable position of the anus was not utilized in this study. Reasons are discussed in Data S1 (U4 and U8).

The monophyly of Eigenmannia is obtained based on morphological characters for the first time herein. We studied over 80% of the valid species (21 of the 26 valid species—Table 1)—that represents the most comprehensive analysis yet done. The monophyly of Eigenmannia was recovered based on the four aforementioned synapomorphies (see Phylogenetic diagnosis of Eigenmannia), of which only one is unique for the genus, nervus opticus thicker than the nervus olfactorius. Mago-Leccia (1994) provided two diagnostic features for Eigenmannia: (1) five branchiostegal, two anterior slender, last three broad; (2) anterior intermusculars highly branched (Table S9). We evaluated the shape of each branchiostegal, and each intermuscular bone series independently. The shape branchiostegal were not recovered as a synapomorphy for Eigenmannia, but the presence of “epipleurals” at 7–9th vertebrae is recovered as a synapomorphy for the genus.

The two groups of species recognized by Peixoto and Ohara (2019) are recovered as monophyletic: E. humboldtii species group and E. trilineata species group. Also, E. macrops do not belong to any of them. Eigenmannia macrops was recovered as the sister group of the other congeners. Waltz and Albert (2017, 2018) included this species in the so-called Eigenmannia macrops species group, a suggestion not followed herein because it is not necessary to name a group to include a single species (Peixoto & Ohara, 2019). Waltz and Albert (2017) proposed Eigenmannia humboldtii species group (Clade E3) to include E. humboldtii, E. limbata, and E. nigra. It was diagnosed by the presence of distal margin of anal fin black, largest body sizes of the species of the genus (>45 cm TL), and by a relatively deep body in mature individuals (body depth >11% TL). Characters based on the total length are problematic, as the specimens are frequently damaged or regenerated in the posterior portion of the body, as discussed by Mago-Leccia et al. (1985; see Peixoto & Ohara, 2019 for additional comments on E. humboldtii species group). Thus, these characters were not included in the present study. Five synapomorphies recovered the monophyly of this group (clade E3): (1) number of premaxillary teeth, 38–54; (2) number of teeth on lower pharyngeal plate; (3) number of anal-fin rays, 226; (4) distal margin of anal fin dark (char. 110: 0>1); and (5) origin of the levator arcus palatini not including the frontal (char. 125: 1>0). This group is proposed as sister group of clade E5 (the E. trilineata species group). Recently, Waltz and Albert (2018) proposed Eigenmannia meeki as a member of the E. trilineata species group. Later on, Peixoto and Ohara (2019) removed E. meeki from this group due to the absence of the superior midlateral stripe. In the present study, we expanded the definition of Eigenmannia trilineata species group to include E. meeki, as proposed by Waltz and Albert (2018). This group, (Clade E5), is now defined by the presence of lateral line stripe (char. 106: 0>1). This character is absent in E. virescens; however, based on our comprehensive analysis, we considered the absence of this character in this species as a reversion of the condition present in Clade E5.

The monophyly of Clade 4 (Archolaemus + Distocyclus + Rhinosternarchus), or the close relationship between Archolaemus and Distocyclus—considering that Rhinosternarchus was not included in most previous studies—was proposed by Triques (1993), Maldonado-Ocampo (2011), Tagliacollo et al. (2016a), and Alda et al. (2019). Triques’ (1993) proposed the elongation of the cartilaginous palatine, and the sinuous anterior portion of fourth epibranchial as synapomorphies for this clade. Later, Clade 4 was additionally supported by Tagliacollo et al. (2016b) who provided six other synapomorphies: (1) preorbital length about one-third as long as total head length in mature specimens; (2) mesethmoid anterior portion horizontal in relation to ventral ethmoid; (3) fourth epibranchial with an elongate ascending process; (4) fifth epibranchial sinuous; (5) proximal surface of first DHS as broad as its descending blade; and (6) 200–299 anal-fin rays. The cartilaginous palatine is clearly more elongated in Archolaemus and Distocyclus than in the other members of Eigenmaniinae, however, considering the difficulties in establishing a precise way to evaluate this enlargement, this character was not included in the present analysis. Other elements responsible for the snout elongation have been coded as independent characters and some of them support these relationships. Characters on epibranchial shape (Tagliacollo et al., 2016b; Triques, 1993) and shape of first DHS (Tagliacollo et al., 2016b) were not included in the present analysis as discussed in Data S1 (U12, U14, and U21, respectively). The number of anal-fin rays was treated here as a quantitative character without discretization, which did not recover the relationships between Archolaemus and Distocyclus as sister groups, contrasting with the findings of Tagliacollo et al. (2016b).

Distocyclus was established to include the nominal species E. conirostris and E. goajira (Mago-Leccia, 1978). Albert and Fink (1996) recovered the close relationships between these species by the presence of a conical snout and a small nasal capsule (Table S10). Dutra et al. (2014), however, redefined both characters and restricted the genus to its type species, D. conirostris, not supporting Albert & Fink’ hypothesis. Meunier et al. (2014) described D. guchereauae and assigned it to Distocyclus based on the shape of the snout. However, the snout of this species differs in shape when compared with that of D. conirostris. Its snout is more triangular than conical as defined by Dutra et al. (2014). Distocyclus guchereauae also differs from D. conirostris by having more teeth on both jaws and the presence of endopterygoid teeth. In the present analysis, D. conirostris is the sister group of Archolaemus plus Rhinosternarchus, while D. guchereauae is the sister group of E. oradens and both members of a distal clade in Eigenmannia. Thus, D. guchereauae is allocated in Eigenmannia, and Distocyclus is again recognized as a monotypic genus.

In the present study, the nominal species “Eigenmannia” goajira was included in the analysis only using external characters and osteological features accessed via radiographs or available in the literature, representing 53.5% of the character list (77 of 144), and 46.5% missing data. In spite of that, as discussed by Dillman et al. (2015), the inclusion of taxa with large amount of missing data in morphological-based phylogenies has provided valuable results to recover the phylogenetic position of taxa with uncertain relationships. These authors, however, reinforced that additional analyses are necessary to determine the threshold of missing data that allows the recovery of a reasonable phylogenetic position. “Eigenmannia” goajira was recovered as the sister group of Archolaemus in the present analysis. Further, considering that Archolaemus have been historically diagnosed by the presence of a free orbital rim (e.g., Korringa, 1970; Vari et al., 2012—see Table S11), we opted to keep traditional characters used to recognize this genus and raise Rhinosternarchus to include “E.” goajira.

In the most recent revisionary study of Archolaemus, Vari et al. (2012) proposed four putative synapomorphies for the genus: (1) free orbital rim; (2) first row of teeth attached to premaxilla only by the anterobasal margin; (3) a pronounced gap between lower lip posterior margin and premaxilla anterior margin; and (4) upper lip ventral surface sponge-like with pores and fleshy raised papillae (Table S11). The presence of a free orbital rim was recovered as a synapomorphy for the genus in this study. The first row of mobile teeth mobile presented an ambiguous optimization due to the missing data in “E.” goajira. Other proposed characters were not included as discussed in the Data S1 (U2 and U3). The presence of the superior midlateral stripe was proposed as a synapomorphy for a more inclusive clade within Archolaemus by Vari et al. (2012); however, a more detailed examination of the type series of A. orientalis revealed the presence of a weak dusky to dark pigmentation on the same position. This character was recovered as a synapomorphy for the genus. The monophyly of clade A2 (Archolaemus except A. orientalis) was supported by 11 synapomorphies (see Data S3). Vari et al. (2012) proposed an unresolved trichotomy between A. santosi, A. blax plus A. janeae, and A. ferreirai plus A. luciae. In the present study, derived features define a series of successive sister groups (clades A2–A5) in which most of clades are supported by characters related to the elongation of snout. Vari et al. (2012) presented two putative synapomorphies to support the monophyly of A. janeae plus A. blax: (1) posterior ceratohyal approximately 1.5 times as long as ventral hypohyal; and (2) presence of a single row of teeth on the posterior portion of the dentary. The first character is also present in A. luciae and consequently represents a synapomorphy for clade A3 with a reversal in A. ferreirai. While the other character was treated as a quantitative character without discretization, it was not recovered as a synapomorphy for this clade. Thus, the hypothesis of monophyly of A. janeae plus A. blax was not corroborated; instead, A. janeae is proposed as sister group of A. blax + A. ferreirai + A. luciae. The presence of a coronomeckelian bone approximately 50% or more as long as Meckel's cartilage was presented as a synapomorphy for A. ferreirai plus A. luciae by Vari et al. (2012). This condition, however, is ambiguous due to the polymorphism in A. blax. The clade A5 (A. ferreirai + A. luciae) is supported herein by three additional synapomorphies (see Data S3).

Traditional morphological-based phylogenies on Neotropical fishes are mostly based on external anatomy and osteology (e.g., de Santana & Vari, 2010). Consequently, some anatomical complexes that are putative source of phylogenetic information, such as myology and neuroanatomy, are frequently overlooked [e.g., Mago-Leccia, 1978 (Sternopygidae); Triques, 1993 (Gymnotiformes); Buckup, 1998 (Characiformes); Birindelli, 2014 (Doradoidea—Siluriformes)]. Information on myology and neuroanatomy is available in Gymnotiformes literature (Aguilera, 1986; Albert, 2001; Albert et al., 1998; Dutra et al., 2015; de La Hoz & Chardon, 1984; Peixoto & Ohara, 2019) and then used in studies on the phylogenetic reconstruction of the group. Albert (2001) included 30 characters of sensorial and nervous complexes, representing approximately 12% of the total listed characters, being most of them also used by Tagliacollo et al. (2016b). On the contrary, characters derived from myology represent less than 0.2% of the entire universe explored of the morphological traits in cladistics studies in Gymnotiformes (Peixoto & Ohara, 2019). In the present study, 18 myological and 12 neuroanatomical characters were included, representing approximately 20% of the total of analyzed characters (12% and 8%, respectively). Informative characters in these anatomical complexes are represented by few intraspecific modifications, as discussed by Abrahão et al. (2018), and consequently, their contribution to the phylogenetic reconstruction are noticed in more inclusive taxonomic levels. For example, these characters represented six synapomorphies supporting the monophyly of Eigenmanniinae (five myological and one neuroanatomical).

Quantitative characters, such as number of anal-fin rays, have historically been employed in phylogenetic reconstructions of Gymnotiformes (e.g., Albert, 2001). However, in all previous studies, this information was treated through a discretization of range of variation, representing less than 0.5% of the entire universe explored in recent contributions (Albert, 2001; Tagliacollo et al., 2016a, 2016b). It is important to note that characters that were used as a diagnostic features within Eigenmanniinae species, such as the number of premaxilla and dentary teeth (e.g., Peixoto et al., 2015) were never addressed in a phylogenetic context. Consequently, information on quantitative characters was restricted to shallower clades. For instance, the variation in the number of precaudal vertebrae, which is historically recovered as a synapomorphy for Eigenmanniinae (Table S1), has never been properly explored to resolve relationships at the species level. In the present study, 17 quantitative characters without discretization were included for the first time in a phylogenetic study of Gymnotiformes, representing approximately 12% of the total of analyzed characters. As a result, quantitative characters have proven to be a valuable source of information for resolving relationships in all levels of inclusiveness. For comparison, the analysis excluding quantitative characters resulted in 310 most parsimonious trees [344 score, CI of 0.352 and RI of 0.707 (Figure S1)]. Where Eigenmannia's monophyly is not recovered and most relationships within Clade 2 are unsolved, except for the hypothesis of Rhinosternarchus as Archolaemus's sister group. Ferrer et al. (2014) found a similar scenario, in which analysis including quantitative characters without discretization resulted in a better-resolved topology.

Thus, the sum of poorly explored anatomical complexes (myology and neuroanatomy) and quantitative characters without discretization associated with morphological characters (external anatomy and osteology) traditionally employed in phylogenetic analyses resulted in a set of informative characters at all taxonomic levels within Eigenmanniinae, resulting in a robust hypothesis presented herein. Also, it provides a solid basis for future studies involving taxonomy and evolution of the group.

5 EXAMINED MATERIAL