Redescription of the type species of Odontostilbe Cope, 1870 (Teleostei: Characidae: Cheirodontinae), and description of three new species from the Amazon basin

The widespread Neotropical cheirodontine Odontostilbe fugitiva is reviewed, and three new species of the genus are described from the Amazon basin. A neotype is designated for O. fugitiva , from the mouth of the río Mazán, upper río Amazonas basin, Peru. Three nominal species are herein considered as synonyms of O. fugitiva : O. madeirae , O. drepanon , and O. caquetae . The three new species are respectively O. ecuadorensis from the río Napo basin, O. nareuda from the rio Madeira basin, and O. parecis from the rio Guaporé basin (rio Madeira basin). Among other diagnostic characters, gill rakers are especially informative in distinguishing species. Description of gill raker denticulation is provided for the first time for species of the Cheirodontinae. Secondary sexually dimorphic characters are described for all species. O queirodontíneo , espécies descritas o bacia Amazônica. neótipo designado O. fugitiva , boca do rio Mazán, rio espécies nominais são consideradas sinônimas O. fugitiva : O. madeirae , O. drepanon O. caquetae espécies respectivamente, O. ecuadorensis bacia Napo, nareuda Madeira parecis bacia Guaporé (bacia do rio Madeira). caracteres diagnósticos, branquiais são especialmente informativos para distinguir as espécies. Uma descrição da denticulação dos rastros branquiais é apresentada pela primeira vez para espécies de Cheirodontinae. Caracteres de dimorfismo sexual secundário são descritos para todas as espécies.


Introduction
The Cheirodontinae contains about 15 genera and 47 species of small Neotropical characid fishes, usually attaining 30-40 mm in standard length (Malabarba, 2003;Malabarba et al., 2004). Odontostilbe Cope, 1870 was originally defined based on the presence of a complete lateral line, versus interrupted in Cheirodon Girard, 1855. The genus initially comprised only Odontostilbe fugitiva Cope, 1870 from "Pebas, Eastern Equador". Currently, Pebas is a city in the Departamento de Loreto in Peru, not Ecuador. The taxonomic history of the discussion of the validity of Odontostilbe versus Cheirodon dates back to the work of the danish zoologist Christian F. Lütken. While redescribing Poecilurichthys pulcher Gill, 1858, Lütken (1875 noted that "technically the species of Trinidad is properly an Odontostilbe Cope, but this genus only differs from the known [Chirodon] G. in the lateral line which is not complete …" [our translation], referring the species as Chirodon (Odontostilbe) pulcher (Gill, 1858). Subsequently, Steindachner (1882aSteindachner ( , 1882b described Chirodon pequira, and designated the type species of Odontostilbe as Chirodon (Odontostilbe) fugitiva. Subsequent works alternated usage of Odontostilbe as valid (e.g. Eigenmann & Eigenmann, 1892, in a preliminary list of freshwater fish of South American listed Odontostilbe fugitiva and recognized Odontostilbe pulcher), or as synonym of Cheirodon (e.g. Ulrey, 1895, when examining "Characinidae" collected by Charles F. Hartt in Brazil listed Cheirodon fugitiva and Cheirodon pulcher, along with five other species of Cheirodon).
Herein we provide a redescription of the type species of Odontostilbe, O. fugitiva, including the designation of a neotype, a discussion of its synomyms, and morphological variation of populations across the Amazon basin. In order to present a better diagnosis for the type-species of the genus, we also describe herein three new species putatively related to O. fugitiva from major tributaries of the Amazon basin, one from upper río Napo and río Putumayo, in Ecuador, and two from rio Madeira basin, in Brazil and Peru. This is part of a major systematic study of Odontostilbe.

Material and Methods
Counts and measurements follow Fink & Weitzman (1974), primarily on the left side of the specimen. Head length is the distance between the tip of the snout and the posterior end of subopercle, which is slight posterior to the margin of the opercle. Total vertebrae number includes the four vertebrae of the Webberian apparatus, and the terminal "half centrum" as outlined by Malabarba & Weitzman (1999). The gill raker at the junction of the ceratobranchial and the epibranchial is referred as the posteriormost gill raker of the lower branch of the arch, and counted jointly with the gill rakers on the lower branch. Specimens were cleared and stained according Taylor & Van Dyke (1985), and/or radiographs were used for counting vertebrae, teeth, and gill rakers. Scanning Electron Microscope (SEM) images were obtained from teeth and denticulation of gill rakers. Values of the holotypes and neotype of O. fugitiva are marked in bold in descriptions.
The following institutions provided material for the study: As demonstrated by Gonçalves et al. (2005) in the characid Aphyocharax, sexual dimorphism may start development at different body sizes, according to gonadal maturation, and so, body size itself does not allow assuming a given specimen lacking male sexually dimorphic features is a female. So, males and females are recognized only when dissected, or when showing sexual dimorphism in specimens of similar size in the same collecting lot. Specimens smaller than the smallest specimen showing sexual dimorphism in each lot are termed as unsexed. Specimens of any size belonging to lots where sexual dimorphism is absent are termed as unsexed.
Statistical analyses. Principal component analysis (PCA) was used to check overall variation among samples, looking for differences in morphometrics among species or between sexes, being an input to multiple regressions (Johnson & Wichern, 1998). PCA on covariances of logarithmically transformed measurements were obtained using Past version 1.28 2004, and Sigma Plot version 6.10 2000. The first principal component was partitioned out, considering that it mostly accounts for size variation (Strauss, 1985).
Multiple linear regressions were applied to describe morphometric differences among species or between sexes. An "overall test for coincidental regressions" as described by Zar (1996: 368,431), and Souza (1998:72-77) was used to determine whether the regressions of two set of data are estimating two species or sexes (H1) or the same population regression (Ho). The test was applied when regression lines showed overlapping in confidence limits. The null hypothesis is that there is a single population underlying all k regressions. The following formulae were used: F= ((SSR1 -SSR2 / 2) / (SSR2 / n-4)), where n= number of x, y cases, and n-4 means the sum of the two residual degrees of freedom. The F value is obtained by calculating sum of squared residuals (SSR) of each linear regression of any two supposed different species (or between males and females) (SSRsp1+SSRsp2=SSR2), and the pooled sum of squared residuals (SSR1) obtained of a unique linear regression for the same data. The SSR2 < SSR1 indicates two regression lines are better adjusted to the linear model. Considering two separated regression lines would better describe the function between dependent and independent variables of two hypothesized different species than an unique regression line for the same data, the F statistic would be higher than Fisher's F (alpha= 0.05), rejecting Ho. Associated p values were obtained using a F-table.
Males with acute, elongate, retrorse hooks on posterior margin of pelvic-and anal-fin rays. One or two (rarely 3) paired and unpaired hooks per segment of lepidotrichia, on last unbranched anal-fin ray, and 1 st to 8 th (mostly 6 th -7 th ) anal-fin branched rays, positioned at distal half-length and not extending to ray tip; tiny hooks can be present distally up to 19 th anal-fin branched rays in very developed males (e.g. INPA 18506, ANSP 178908, ANSP 15006, and neotype). One or two (rarely 3) unpaired and long hooks per segment of lepidotrichia along almost all length of 1 st to 7 th branched pelvic-fin rays, reaching distal tip of fin ray (expect on 1 st branched pelvic-fin ray, on midlength portion). No hooks found on 1 st unbranched pelvic-fin ray. One to 3 unpaired hooks sometimes present on anterior border of anal-and pelvic-fin rays branches.

Color in alcohol.
General ground body color brownish or pale yellow. Dark dorsum pigmentation from head to caudal peduncle, dark chromatophores mostly on border of scales forming reticulated pattern. Pigmented scales extend laterally on body below longitudinal stripe, also on pseudotympanum area or not. Dorsal fin entirely scattered with dark chromatophores along fin rays, densely along 1 st -2 nd unbranched dorsal-fin rays. Dark chromatophores along 1 st -9 th dorsal branched fin rays, except almost hyaline 9 th branched fin ray. Pectoral and pelvic fins with scattered dark chromatophores, more numerous on anteriormost portions of fins. Fresh specimens only with 7 th branched pelvic-fin ray hyaline. Neotype with 1-5 th branched pelvic-fin rays pigmented. Pelvic fin mainly hyaline in old-preserved specimens. Anal fin with dark chromatophores spread on all fin rays or on more central branched anal-fin rays, except at tips of last unbranched fin rays and 1 st branched fin ray. Adipose fin not pigmented. Caudal fin almost entirely covered with diffuse dark chromatophores along fin rays, except clear areas on base of caudal-fin lobes just behind caudal-fin spot. Rounded black to brown spot on base of caudal fin, reaching upper and lower border of peduncle, or upper border only, sometimes extending little onto proximal portion of 7 th to 12 th central principal caudal-fin rays.
Numerous dark chromatophores, somewhat contiguous, on snout, upper and lower lips, anteriormost portion of maxilla, and 1 st , 2 nd , and 6 th infraorbitals. Top of head on frontals and parietals pigmented, with deep-lying dark chromatophores over brain membranes below frontals and parietals, and fontanel. Body with faint dark and silver midlateral stripe, sometimes forming black line along middle longitudinal body axis, beginning behind or on pseudotympanum, above lateral line, and reaching caudal spot. Guanine on eye iris, opercle, isthmus, and most of 3 rd infraorbital. Below lateral line, belly faint almost without pigmentation, chromatophores above anal-fin base forming 12 to 15 chevron shaped markings.

Color in life.
Life color of specimen from lower Ucayali drainage (Fig. 4). One red spot on base of each caudal fin lobe, just posterior to caudal fin black blotch. Red and yellow pigments most on dorsal and anal fins. White pigment on tips of last unbranched and 1 st -2 nd branched anal-fin rays, and of unbranched and first branched pelvic-fin rays.
Sexual dimorphism. Males with hooks on anal-and pelvicfin rays, and elongate dorsal-and pelvic-fin rays, in contrast to absence of these features in females. A gill gland (Burns & Weitzman, 1996) present in mature males on first gill arch, covering about 8 to 10 anterior branchial filaments (Fig. 3). Sexual dimorphism in general body shape is also recognized by principal component analysis performed on measurements of males and females (Fig. 5). Principal component 2 (PC2) grouped males and females separately. It was affected strongly and negatively by pelvic-fin length, snout length, and dorsal-fin length; and positively by depth at dorsal-fin origin. Males are particularly distinct from females in all of these characters, having higher values of pelvic-fin length, snout length, and dorsal-fin length than females, while females have higher depth values at dorsal-fin origin than males (Table 1). Principal component 3 (PC3) was affected strongly and positively by snout length and upper jaw length, and negatively by pelvic-fin length and dorsal-fin length. Variation on PC3 seems to be related to the level of development of pelvic-and dorsal-fin rays, snout protusion (snout length, and upper jaw length) in males, because males with higher values for these features are located upper on PC3 axis (Fig. 5).
Remarks on the holotype of Odontostilbe fugitiva. The short description of O. fugitiva by Cope (1870) makes no reference to type specimens. Most type-specimens of fishes described by Cope were deposited at ANSP, although some at USNM and some are missing (Böhlke, 1984;Vari & Howe, 1991). Henry W. Fowler was the first full-time curator of the fish collection at ANSP, and one of his first tasks was the cataloging of the extensive collection of Edward D. Cope (Böhlke, 1984). During his stay at ANSP, Fowler (1913: 529, 532) Fowler (1943: 232) stated that the type of O. fugitiva was lost. Géry (1972a) also affirmed the holotype of O. fugitiva was lost after visiting ASNP to study Fowler's type material. Vari & Howe (1991) did not list O. fugitiva among characiform type specimens at USNM. Böhlke (1984: 46) indicated ANSP 8059 as the holotype of O. fugitiva, and noticed that "original Cope label and catalog indicate this to be type of O. fugitiva". However, in the same publication, Böhlke (1984: 46) remarked the same lot (ANSP 8059) was listed by Fowler (1906) as the holotype of Aphyocharax filigerus, valid as Prionobrama filigera (Cope, 1870). Examination of ANSP 8059 confirmed this is a specimen of P. filigera. Both authors have fruitlessly searched the ANSP and USNM fish collections after the type material of O. fugitiva [including the lot ANSP 8073 listed by Fowler (1906) as cotypes of Aphyocharax filigerus].

Neotype designation.
Odontostilbe fugitiva is the type species of Odontostilbe, and the designation of a neotype, as well as establishment of the identity of the type species is essential in solving nomenclatural questions among cheirodontines. The neotype was not caught at Pebas, the original type locality of O. fugitiva. Pebas is located at the confluence of río Ampyiacu with río Amazonas, Loreto, Peru. The neotype is a mature male from an affluent of lower río Napo basin, río Mazán, upper rio Amazonas, Loreto, Peru. Río Mazán is to the west of the type locality, being the nearest site (Fig. 6) with the best preserved mature male ( Fig. 1) available, being in conformity with ICZN (1999). Since secondary sexually dimorphic characters are important to distinguish Odontostilbe and other Cheirodontinae genera, a mature male is preferred as the neotype. The chosen male has entire elongate 2 nd unbranched dorsal-fin ray and unbranched  pelvic-fin rays, well-developed acute retrorse hooks on pelvic-and anal-fin rays, preserved scales, and well-preserved color pattern (Fig. 1). Cope (1870: 566) referred to a "transverse line [of scales] at vent 11; at ventral fins 5-5". Although distinct from the counts of the neotype, these counts are included in the range found herein for the species (5-7 between lateral line and dorsal-fin origin and 4-5 between lateral line and anal-fin origin).
The status of Odontostilbe madeirae and Odontostilbe drepanon. Both species were described in the same paper by Fowler (1913). The holotype (Fig. 7) and all paratypes but one of O. drepanon have the dimorphic characters unique to males of Odontostilbe: elongation of 2 nd unbranched dorsal-fin ray and 1 st unbranched pelvic-fin rays, and presence of hooks on pelvic-fin rays and anal-fin rays. On the other hand, the holotype (Fig. 7) and all paratypes of O. madeirae lack these characters, and have higher body depth, the diagnostic features for females (see Table 1 and description under Sexual Dimorphism above). The two species described by Fowler (1913) overall match the O. fugitiva diagnosis, and this is a case of misunderstanding dimorphic males and females as two different species. Fowler (1913: 529) remarked that the absence of "the dark-edged silver lateral band" in O. madeirae is a character to distinguish it from O. fugitiva. The holotype and paratypes of O. madeirae actually have a faded longitudinal line of chromatophores posterior to the pseudotympanum.
The holotype of O. madeirae has few dentary teeth remaining, but laterally 4 teeth remain, each with 7 cusps. Premaxillary teeth 5 with 7-8 cusps and maxillary teeth 2 with 7 cusps. The holotype of O. drepanon also lacks most dentary teeth, but on the right side four teeth remain each with about 7 cusps. Premaxillary teeth 5 with 7-9 cusps, and maxillary teeth 2 with 7 cusps. The 2 nd unbranched dorsal-fin ray is broken.
The status of Odontostilbe caquetae. Odontostilbe caquetae was described from a single immature specimen (28.0 mm SL). Fowler diagnosed O. caquetae comparing the holotype with data available in the short redescription of O. fugitiva by Eigenmann (1915:93), which was based on four specimens examined from "San Antonio de Rio Madeira", measuring "largest 41 mm". Following Fowler's diagnosis, O. caquetae would have a smaller head, smaller eye and larger snout than O. fugitiva. However, measurements and counts of the holotype of O. caquetae overlap those of O. fugitiva (Table 1), not supporting any of Fowler's diagnostic characters. For that reason, we consider O. caquetae as a synonym of O. fugitiva.
The holotype is now dark colored and damaged with most dorsal-fin rays and caudal-fin rays broken (Fig. 8). Infraorbitals 2-5 are lacking on the left side. Premaxilla probably would have five teeth, but just two of them remain on each left and right sides, bearing 8-7 cusps. Right maxilla with 2 teeth bearing 8-9 cusps, and left maxilla with only one tooth with 8 cusps. Four dentary teeth remain on each side bearing 5-6 cusps. This dentition is similar to that found in smaller speci-  (Fowler, 1945b). The main reason for not considering O. Fig. 6. Distribution of Odontostilbe fugitiva in central and northen South America. Circles (material examined, symbols can represent more than one lot, and more than one locality), and stars (type locality of neotype on the rigth and of the lost holotype on the left).  caquetae as synonym of O. euspilura is the terminal mouth in the former, against the inferior mouth in the latter.
The first right gill arch of the holotype was found cut off from branchial apparatus, but kept under the operculum. Both first gill arches have 6 gill rakers on upper branch and 11 on lower branch. Fowler (1943: 232) described 7+11 gill rakers and it is possible the first gill raker was lost in the damaged upper branches of both gill arches. Specimens of Odontostilbe examined from rio Japurá, the Brazilian name for mid and lower río Caqueta, match O. fugitiva in measurements and counts, including 6, 7 gill rakers on upper gill arch and 12 on lower (one among 10 examined specimens of rio Japurá has 6+11 gill rakers as the holotype of O. caquetae). Therefore we consider O. caquetae as synonym of O. fugitiva. However, we consider this synonym provisional since no mature males of Odontostilbe were available from upper río Caqueta or rio Japurá.

Distribution and geographical variation.
Odontostilbe fugitiva is the most widespred Odontostilbe species in the Amazon basin, occurring from the major western tributaries to the lower rio Amazonas (Fig. 6). More extense records are from the río Ucayali, río Marañon, rio Amazonas and rio Madeira basins. Records from western major tributaries of the Amazon, Purus and Juruá systems, are relatively scarce. Predominantly, the species occurs in major whitewater rivers of the Amazon basin, being not so far identified from the rio Negro, rio Trombetas, rio Tapajós, rio Xingu, rio Tocantins-Araguaia. The well-sampled fish fauna of the rio Negro did not reveal O. fugitiva (Goulding et al., 1988), though they listed 12 specimens of "Cheirodon sp." (not found at INPA collection). Chao (2001) (Fig. 9), and more pronounced snouts than O. fugitiva, including the Marañon population.
Males of río Ucayali basin also had differences in snout and upper jaw length when contrasted with males from the main rio Amazonas basin, having snout length difference statistically significant in the coincidental regression test (F (2, 74) = 11.8, p<0.00) (Fig. 10). When compared to O. fugitiva, O. dierythrura and Odontostilbe n. sp. from Madre de Dios, the upper jaw length of the population from río Urubamba, upper Ucayali, remains at an intermediate range between O. fugitiva and the other two species (Fig. 11). Again, considering there  were no additional characters differing these populations, and the overlaps in statistics for snout length and upper jaw length, the Ucayali population is considered to be O. fugitiva. A significant statistical difference in measurements was also found when comparing rio Solimões samplings from the region near Manaus to overall O. fugitiva material from upper rio Amazonas basin, including río Ucayali and río Marañon systems. Not only males, but females and unsexed specimens jointly, show differences in snout to pectoral-fin length (F (2, 200) = 85.6, p<0.00) (Fig. 12). This difference is also considered a variation among O. fugitiva populations. However, there is a lack of mature males and females among specimens available from an extensive portion of the rio Solimões between Letícia, Colombia, and Manaus, Brazil. Most samples from Solimões consisted of material caught near Manaus by Peter Bayley in 1970s or more recently by INPA researchers. Examined specimens from rio Juruá basin are unsexed, probably juveniles. Exceptionally, two specimens among 8, have 8 upper gill rakers, and 3 specimens have 14 lower gill rakers (Fig. 13a). Higher gill raker counts occur in O. dierythrura and in an undescribed species from Madre de Dios. However, the upper jaw lengths of the avaliable specimens of Juruá fit that of O. fugitiva. One specimen of O. fugitiva from the rio Madeira also has 14 lower gill rakers (Fig. 13b). Similarly, specimens examined from rio Purus contain only unsexed specimens. In the lack of fully developed, mature specimens, we tentatively assign Juruá and Purus specimens to O. fugitiva.     (2001) results pointed out a high abundance of the cheirodontine O. fugitiva and at least three Serrapinnus spp. (Serrapinnus micropterus, Serrapinnus gr. microdon, and Serrapinnus n. sp., our identifications) in lakes from the rio Solimões floodplain. Peter Bayley fish collections undertaken during the late 1970's from the same lakes were also extensively examined (material at ANSP, MCP, and USNM), confirming that O. fugitiva is very common in these floodplain lakes. At the same localities, Marchantaria Isle (lago Camaleão), Janauacá, and Janauari, O. fugitiva was caught associated with macrophytes Echinochloa polystachya, E. crassipes, Paspalum repens, Panicum sp., or unvegetated places, and considered as omnivorous (Petry et al., 2003).
Two new cheirodontine species provisionally treated as "Odontostilbe", but not described herein, are sympatric and have been found mixed with O. fugitiva in some jars. One of these new species mostly occurs in the río Ucayali basin and the second in the río Marañon basin. Diagnosis. Characters that diagnose Odontostilbe ecuadorensis are: (1) upper gill rakers 5-6, lower 9-11, mostly 10-11 (vs. upper gill rakers 6-8, mostly 6-7, and lower 11-14, usually 12 in O. fugitiva ( Fig. 13a-b); (2)  Description. Morphometric data given in Table 2. Slender and large Odontostilbe species, largest male (holotype) 45.7 mm SL, largest female 47.7 mm SL. Body elongate and compressed, more compressed in males than females. Greatest body depth at dorsal-fin origin. Snout pointed in males, rounded in females. Head profile slightly convex from snout to dorsal-fin origin; profile descends from that point almost straight to caudal peduncle. Ventral profile convex from lower jaw to caudal peduncle. Anal-fin base straight. Caudal peduncle slightly longer than deep.
Males with robust, elongate, retrorse hooks on posterior border of pelvic-and anal-fin rays. One to 3 paired or unpaired hooks per segment on last unbranched anal-fin ray, and 1 st to 16 th branched anal-fin rays, more developed from 1 st to usually 7 th rays and very small like knots usually on 10 th branched fin ray and following rays. Hooks on middle portion of anterior rays and on distal portion after 3 rd branched anal-fin ray. One to 3 unpaired acute, elongate hooks per segment of lepidotrichia along almost entire length of 1 st to 7 th branched pelvic-fin rays, and reaching tip of fin rays (except on 1 st branched pelvic-fin ray, on midlenght portion). No hooks found on 1 st unbranched pelvic-fin ray. One or two hooks on anterior border of branches of anal-fin rays. One or two hooks on anterior border of branches of pelvic-fin rays.

Color in alcohol.
General ground body color brownish yellow. Dark dorsum pigmentation from head to caudal peduncle, chromatophores mostly on border of scales. Pigmented scales extend laterally on body above longitudinal stripe, sometimes on pseudotympanum area in well-pigmented specimens. Dark chromatophores on all dorsal-fin rays, along 1 st and 2 nd unbranched fin rays and faint on 1 st -8 th branched dorsal-fin rays, almost hyaline on 9 th branched ray. Dark chromatophores on 1 st unbranched pectoral-fin ray, less numerous on 1 st and 2 nd branched fin rays, following branched rays practically hyaline. Pelvic fins usually hyaline, with some dark chromato-phores on unbranched pelvic-fin ray. Anal fin mostly clear with few dark chromatophores. Some specimens bearing dark chromatophores on distal portion of middle branched analfin rays. Adipose fin not pigmented. Caudal fin almost entirely covered with diffuse dark chromatophores along fin rays, except light areas on base of caudal-fin lobes just behind caudal-fin spot. Somewhat rounded brownish caudal spot on base of caudal fin, reaching upper border of peduncle, but never reaching lower border, sometimes extending on proximal portion of 7 th to 12 th central caudal-fin rays (specimens from río Cuyabeno). Holotype is not well-pigmented, with pelvic and anal fins hyaline.
Black chromatophores on snout, upper lip, sometimes on lower lip, anteriormost portion of maxilla, and 1 st , 2 nd , and 6 th infraorbitals. Pigmentation on top of head on frontals, and parietals with deep-lying dark chromatophores over brain membranes below frontals and parietals, and fontanel. Body with faint dark and silver midlateral stripe, sometimes reduced to one or two parallel lines, along middle longitudinal body axis, beginning above pseudotympanum, above lateral line, and reaching caudal spot. Guanine on eye iris, opercle, isthmus, and most of 3 rd , 4 th , and 5 th infraorbitals. Below lateral line, belly faint, chromatophores just above anal fin-base forming 6 to 10 chevron shaped markings.

Sexual dimorphism.
Males with hooks on anal-and pelvicfin rays, elongate dorsal-and pelvic-fin rays, in contrast to absence of these features in females (see description). Gill gland (Burns & Weitzman, 1996) presents on first gill arch in mature males, covering about 8 to 12 anterior branchial filaments (Fig. 16). Principal component analysis of morphometric data of males and females denotes sexual dimorphism in general body shape (Fig. 17). Principal component 2 (PC2) grouped males and females separately, being affected strongly and negatively by snout length, pelvic-fin length, and caudal peduncle length; and positively by depth at dorsal-fin origin. Males are distinguished from females in all these characters, showing higher values of snout length, pelvic-fin length, and caudal peduncle length than females, while females have higher values of body depth at dorsal-fin origin than males (Table 2). Principal component 3 (PC3) was affected strongly and positively by snout length and upper jaw length, and negatively by pelvic-fin length and dorsal-fin length. Some males have well-developed dorsal and pelvic-fin rays, and protuding snouts (greater snout length, and upper jaw length). This correlates with their heaviest loadings in PC3 forming a group on upper left corner of the graph (Fig. 17).
Distribution. Río Napo, río Putumayo, and río Pastaza basins (a single lot known for the latter) (Fig. 18). Nearly all material from Ecuador, and some lots from Peru on border with Ecuador. The species is also recorded on lower río Napo basin in sympatry with O. fugitiva.
Etymology. The epithet ecuadorensis refers to the country where the species was first discovered. Ecological notes. Saul (1975) identified Odontostilbe cf. madeirae Fowler, 1913, andHoloshesthes heterodon Eigenmann, 1915 as two cheirodontine species commonly taken together in the río Aguarico, the former primarily feeding on algae and plant material, and the latter on insects and insect larvae. However, examination of Saul's (1975) material proved the species named O. cf. madeirae to be a mix of O. ecuadorensis described herein and O. euspilura. The material he named H. heterodon consist of an undescribed cheirodontine. Part of the specimens caugth by W. Saul was selected herein as type material for O. ecuadorensis. These Ecuadorian species are reported to occur in clearwater, blackwater or turbid waters, riverine sandy-beaches (Stewart et al., 2002), or lagoons and associated tributaries (Galacatos et al., 1996). All the material examined for this study came from altitudes between 200-500 m. Galacatos et al. (1996) recognized two altitudinal regions in Ecuadorian Amazon, lowland between 200 m and 220 m, and piedmont between 235 m and 295 m, where most Odontostilbe species were caught, preferentially in lagoons rather than streams.   Fig. 19 Holotype.  for all other species). Adittionaly, gill raker counts on upper branch 5-6 and lower branch 11-12 of O. nareuda differ from sympatric species O. fugitiva with mostly 6-7 on upper and 12-13 on lower (Figs. 13a-b).

Odontostilbe nareuda, new species
Description. Morphometric data given in Table 3. Largest male 33.1 mm SL; female 32.3 mm SL. Body elongate and compressed. Greatest body depth at dorsal-fin origin. Snout usually pointed; not protuded in males. Dorsal profile ascends slightly convex from snout to dorsal-fin origin, and descends straight from that point to caudal peduncle. Ventral profile convex from lower jaw to anal-fin origin, and straight along anal-fin base. Caudal peduncle slightly longer than deep. Head relatively small. Posterior margin of opercle sinusoidal with upper portion concave and lower portion convex. Mouth terminal. Maxilla short, oblique; posterior tip reaching near or to vertical at anterior eye border, and ending at level of inferior eye border. Cleared and stained specimens (2). Premaxillary teeth 5, bearing 7-9 cusps; midcentral cusp longer than lateral cusps; smaller lateral cusps of each tooth overlap cusps of adjacent teeth, except anteriormost teeth (Fig. 21).
Holotype and paratype males of FMNH 106433 are unique, all having not totally developed hooks. Males with acute, retrorse hooks on posterior margin of pelvic and anal-fin rays. One or two paired or unpaired hooks per segment of lepidotrichia on last unbranched anal-fin ray, and 1 st to 9 th anal-fin branched rays, positioned at middistal length of rays. One or two unpaired hooks per segment of lepidotrichia along almost entire length of 1 st to 7 th branched pelvic-fin rays, and reaching tip of fin rays (except on 1 st branched pelvic-fin ray, on midlenght portion).
Odontostilbe nareuda seems to have fewer dark chromatophores than similar species such as O. fugitiva, O. dierythrura, and O. parecis, and another new species from río Madre de Dios (in manuscript), all of these occur in the rio Madeira basin. Except for O. parecis from rio Guaporé basin, the other three species were found to be sympatric with O. nareuda.

Sexual dimorphism.
Only three males available, holotype and two paratypes, but these are not fully mature, without completely developed hooks. These males have longer pectoralfin rays in contrast with specimens with undetermined sex, statistically significant in test for coincidental regressions (F (2, 276) = 28.5, p<0.00) (Fig. 23). The snout profile of holotype is slightly protruded in comparison with the unsexed paratype FMNH 106433 (Fig. 19). Gill gland not found in males of O. nareuda. Absence of this gill gland probably explained by lack of mature males among examined specimens.

Distribution.
Odontostilbe nareuda is known from the lower río Beni basin, in the ríos Nareuda and Muyumanu, of the ríos Tahuamanu-Orthon drainage that runs parallel to the río Madre de Dios, and the middle rio Madeira (Fig. 18). However, it may be widespread in the rio Madeira basin. Some specimens from the upper rio Mamoré basin were tentatively identified as O. nareuda. Also, three specimens from the rio Purus match the larger anal-fin base diagnostic of O. nareuda. A somewhat similar distribution was observed for Creagrutus occidaneus that occurs in río Madre de Dios and upper rio Purus (Vari & Harold, 2001). Further samples of these regions including mature males and females are needed to assure these distributions.
Etymology. In allusion to the type-locality, in the río Nareuda.
Ecological notes. FMNH field notes for O. nareuda indicate that it occurs in oxbow lakes, river or small river habitats in forest, having whitewater or turbid-white water. The bottom was sandy or muddy substrate with sticks, logs, and leaves. Nevertheless the río Nareuda is mostly a blackwater river as noted by Machado-Alllison et al. (1999a), all the locality notes of FMNH lots refer to whitewater, except FMNH 106430 which mentions clearwater of brown color. The lower río Nareuda seems to be influenced by whitewaters of the río Tahuamanu -río Orthon basin. The FMNH specimens all came from the AquaRAP expedition in the upper río Orthon basin, Bolívia. These specimens were originally identified as Cheirodon fugitiva, Odontostilbe fugitiva or Odontostilbe sp. Sarmiento et al., 1999). Specimens identified for the AquaRAP expedition as O. paraguayensis, and listed as that species by Willink et al. (1999) as "believed to be" a new record for the Bolivian Amazon, actually are O. nareuda or an undescribed species of cheirodontine. Both species were collected together, being probably syntopic. Machado-Allison et al. (1999b) detailed the abundance of species regarding the field station that corresponds to FMNH lots of the holotype and some paratypes (FMNH 106433) and cited "O. paraguayensis" as the most abundant species, and "Cheirodon fugitiva" the fifth ranked, being "typically from cochas or flooded lakes". fin origin. Snout blunt, larger in males. Head profile pointed, gently convex from snout to posterior tip of supraoccipital bone. Dorsal profile ascends almost straight from that point to dorsal-fin origin, sometimes with sligth depression at posterior tip of supraoccipital; from dorsal-fin origin descends almost straight to caudal peduncle. Ventral profile convex from lower jaw to anal-fin origin, and straight along anal-fin base. Caudal peduncle slightly longer than deep. Head relatively small. Posterior margin of opercle sinusoidal with upper portion concave and lower portion convex. Mouth terminal. Maxilla short, terminating posteriorly at vertical near anterior eye border, and ending at level equal to inferior eye border, positioned at 45 degrees angle relative to longitudinal body axis. Cleared and stained specimens (2). Premaxillary teeth 5, bearing 8-10 cusps; midcentral cusp longer than lateral cusps; smaller lateral cusps of each tooth overlap cusps of adjacent teeth (Fig. 27). Premaxillary teeth juxtaposed externally to dentary teeth, leaving premaxillary tooth cusps exposed when mouth closed. Maxilla with 2-3 teeth bearing 7-9 cusps, usually decreasing cusp number to posterior tip of maxilla. Dentary teeth 8-9 bearing 2-7 cusps, gradually decreasing in size posteriorly, first 6-7 teeth large with 6-7 cusps, following 2 small teeth with 2-5 cusps. Smaller cusps of all dentary teeth overlaping adjacent teeth cusps, usually not in posteriormost teeth. All dentary tooth cusps slightly recurved towards interior of mouth.
Males with acute, retrorse hooks on posterior margin of pelvic and anal-fin rays. One or two paired and unpaired hooks per segment of lepidotrichia, on last unbranched anal-fin ray, and 1 st to 8 th branched anal-fin rays, positioned at middistal length of rays and usually not extending to ray tip; tiny hooks can be present up to 16 th anal-fin branched ray, occurring distally on branches (up to 17 th anal-fin branched ray in holotype). One or two unpaired hooks per segment of lepidotrichia on 1 st to 7 th branched pelvic-fin rays, extending almost entirely to distal ray tip (along midlenght of 1 st branched pelvicfin ray). Few mature males available for description of hooks. Type material from INPA contains males with more developed hooks. Holotype with hooks on 2 nd to 5 th branched pelvic-fin rays. Because available males are not fully mature, this species could have typical elongation of unbranched dorsal and pelvic-fin rays present in of Odontostilbe species.

Color in alcohol.
General ground body color brownish or pale yellow. Dorsum dark from head to caudal peduncle, chromatophores mostly on scale borders forming reticulated pattern. Pigmented scales extend laterally on body below longitudinal stripe, also on pseudotympanum area. Dorsal fin with scattered chromatophores along all fin rays, more densely pigmented along 1 st and 2 nd unbranched dorsal-fin rays. Pectoral and pelvic fins with scattered dark chromatophores, more numerous on anteriormost portions of fins. Holotype with pelvic fin hyaline. Anal fin mostly pigmented with chromatophores spread on all fin rays, except at distal tips of last unbranched fin ray and 1st branched fin ray. Adipose fin hyaline. Caudal fin almost entirely covered with diffuse chromatophores along fin rays, clear areas on base of caudal-fin lobes just posterior to caudal-fin spot. Rounded black to brown spot on base of caudal fin, usually light near lower border, sometimes extending little onto proximal portion of 7 th to 12 th central caudal-fin rays. Numerous chromatophores on snout, upper and lower lips, anteriormost portion of maxilla and lower jaw, and 1 st -2 nd infraorbitals. Pigmentation on dorsal surface of head on frontals and parietals, and deep-lying dark chromatophores over brain membranes below frontals and parietals, and fontanel. Body with faint dark midlateral stripe, sometimes reduced to line, along midlongitudinal body axis, beginning posterior to or on pseudotympanum, above lateral line, reaching caudal spot. Below lateral line, ventral body faint almost without chromatophores. Pigmented area above anal-fin base forming 10 to 11 chevron-shaped markings. Guanine surrounds eye iris, interopercle, opercle, ishtmus, and most 3 rd infraorbital. Holotype with some remaining guanine, but fresher specimens of MCP without any guanine.
Sexual dimorphism. Males with anal-and pelvic-fin hooks on rays as previously described, such hooks absent in females. Gill gland (Burns & Weitzman, 1996) present in three mature males of Odontostilbe parecis, holotype, and two paratypes (Fig. 28). Holotype of O. parecis with gill gland present on first gill arch, covering about 9 anterior branchial filaments. Paratype INPA 21425, a male 30.6 mm SL with a gland including about 7 anterior branchial filaments and paratype MCP 37319, 39.0 mm SL with gland including about 6 anterior branchial filaments. Principal component analysis of morphometric data of males and females plus unsexed specimens shows sexual dimorphism in general body shape (Fig. 29). Principal component 2 (PC2) grouped males and females separately. This affected strongly and negatively by caudal peduncle length, pectoral-fin length, pelvic-fin length, and snout length; and positively by depth at dorsal-fin origin. Principal component 3 (PC3) affected strongly and positively by snout length and caudal peduncle length, negatively by anal-fin base and pectoral-fin length. Males differed from females in all of these characters. Usually males with higher mean values for peduncle length, pectoral-fin length, pelvic-fin length, and snout length, while females and unsexed specimens have higher depth at dorsal-fin origin than males (Table 3). Unsexed specimens were not assumed to be females, because they probably also include immature males. All examined males are probably not fully mature as mentioned in description. Additional material may reinforce sexual morphometric characters described here.   Distribution. Upper rio Guaporé drainage, rio Galera and rio Novo.
Etymology. The epithet parecis refers to the Chapada dos Parecis, a plateau situated on the east of Guaporé tributaries where the species was discovered.
Ecological notes. The Guaporé-Iténez is a clearwater rightbank tributary of the major rio Madeira (Goulding et al., 2003). Field notes record of MCP material: translucent water of quiet to medium flow, sandy bottom with pebbles, and a depth of 1-1.2 m. This was species collected along with Serrapinnus sp.

Discussion
Previous phylogenetic diagnosis for the genus Odontostilbe among cheirodontines was based on two synapomorphies described in Malabarba (1998), and related to the elongation of the second unbranched dorsal-fin ray and the unbranched pelvic-fin ray of males. These characters were found homoplastic in "Odontostilbe" mitoptera of the Compsurini. Recognition of Odontostilbe separately from genera assigned to the tribes Cheirodontini and Compsurini was also supported by the lack in Odontostilbe species of the several synapomorphies dignosing those tribes and their included genera. These synapomorphies consist of primary and secondary sexual characters. The genera Pseudocheirodon, Prodontocharax, Aphyocheirodon, and Cheirodontops also lack specialized sexually dimorphic features present in the Compsurini and Cheirodontini and were diagnosed based on jaws and teeth modifications, absent in Odontostilbe.
The diagnosis of Odontostilbe in the present contribution follows that of Malabarba (1998), based on the analysis of two Odontostilbe species only, O. fugitiva and O. pequira, respectively the type species of the genera Odontostilbe and Holoshesthes. Holoshesthes was placed in the synonym of Odontostilbe. We found morphological diversity in Odontostilbe much larger than previously known. We recognize in Odontostilbe six species of the ten species referred to the genus by Malabarba (2003): O. fugitiva, O. dierythrura Fowler, 1940, O. microcephala Eigenmann, 1907, O. paraguayensis Eigenmann & Kennedy, 1903, O. pequira (Steindachner, 1882, and O. roloffi Géry, 1972 (actually a synonym of O. euspilura, see below). This excludes the two species of Central American compsurins provisionally kept in Odontostilbe: "O." dialeptura and "O." mitoptera. Also, the generic assignment of the two nominal species described by Géry (1960) as Cheirodon (C. gracilis and C. gracilis littoris) and latter transferred to Odontostilbe by Planquette et. al. (1996), deserve a separate revision and are not included in Odontostilbe.
In a comprehensive taxonomic revision of Odontostilbe in progress by the first author, 10 additional species of cheirodontines were found related to this genus, in addition to the six species listed above and the three new species described herein. Among these 10 species, 8 are undescribed, one is Odontostilbe pulchra (Gill, 1858) [left as species inquirenda among Cheirodontinae by Malabarba (2003)], and the last one is Cheirodontops geayi Schultz, 1944, type species of Cheirodontops. Cheirodontops geayi was not previously hypothesized as related to Odontostilbe, although Böhlke (1954) pointed out that Cheirodontops Schultz, 1944 was solely distinguished from Holoshesthes Eigenmann, 1903 by the absence of teeth on the maxillary. The examination of type material of C. geayi, and an extensive revision of the species, showed that the teeth are present on the maxilla. Also it was found that the elongation of dorsal-and pelvic-fin rays in mature males that diagnoses Odontostilbe is present in Cheirodontops. Schultz (1944) observed the resemblance of teeth between Cheirodontops and Holoshesthes. Malabarba (1998) based on the phylogenetic analysis of Cheirodontinae placed Holoshesthes in synonym of Odontostilbe, arguing that the character of the presence of dentary teeth with at least five cusps, with the three central cusps larger, compressed and in a row forming a sharp cutting edge, was found to be ambiguous, and independently acquired in four cheirodontine lineages. Eigenmann (1915), and Géry (1977) used this character alone to distinguish Holoshesthes of Odontostilbe. Additional phylogenetic analysis of Odontostilbe is needed to clarify the relationships and taxonomic status of Holoshesthes and Cheirodontops. Malabarba (2003) listed Deuterodon (Lobodeuterodon) euspilurus Fowler, 1945, described from upper río Caquetá basin, Colombia, as a junior synonym of O. fugitiva. According to work in progress by the senior author, O. euspilura (Fowler, 1945) is a valid species, and a senior synonym of O. roloffi Géry, 1972 described from upper río Napo, Ecuador. A full redescription of this species will be provided in a forthcoming paper.
Almost all the mature males of the examined Odontostilbe related species share both elongations of the unbranched dorsal-and pelvic-fin rays, excepting Odontostilbe euspilura and an undescribed species from rio Uruguay basin that lack the elongation of the second unbranched dorsal-fin ray. Two other undescribed species from upper rio Paraná basin lack both elongations, of unbranched dorsal-and pelvic-fin rays, but there are few mature males available of them. Odontostilbe pulchra has a slight elongation of the second unbranched dorsal-fin ray only in some specimens, but has the unbranched pelvic-fin ray elongate.
A reevaluation of the diagnosis of Odontostilbe is still needed because the discovery of several new species not previously used in a phylogenetic analysis of the genus by Malabarba (1998). The discovery of new data regarding Cheirodontops geayi indicating that it shares characters diagnosing Odontostilbe (elongation of the dorsal-and pelvicfin rays) requires a reanalysis of the phylogeny of these genera of the Cheirodontinae. Also, such an analysis should include Lobodeuterodon (its type species) and Holoshesthes, apparent junior synonyms of Odontostilbe.