Two new species and a new subgenus of toothed Brachyhypopomus electric knifefishes (Gymnotiformes, Hypopomidae) from the central Amazon and considerations pertaining to the evolution of a monophasic electric organ discharge

Abstract We describe two new, closely related species of toothed Brachyhypopomus (Hypopomidae: Gymnotiformes: Teleostei) from the central Amazon basin and create a new subgenus for them. Odontohypopomus, new subgenus of Brachyhypopomus, is diagnosed by (1) small teeth present on premaxillae; (2) medialmost two branchiostegal rays thin with blades oriented more vertically than remaining three rays; (3) background color in life (and to lesser extent in preservation) distinctly yellowish with head and sides peppered with small, widely spaced, very dark brown stellate chromatophores that greatly contrast with light background coloration; (4) a dark blotch or bar of subcutaneous pigment below the eye; (5) electric organ discharge waveform of very long duration (head-positive phase approx. 2 milliseconds or longer, head-negative phase shorter or absent) and slow pulse repetition rate (3–16 Hz). The type species of the new subgenus, Brachyhypopomus (Odontohypopomus) walteri sp. n., is diagnosed by the following additional character states: (1) subcutaneous dark pigment at base of orbit particularly prominent, (2) body semi-translucent and nearly bright yellow background coloration in life, (3) a biphasic electric organ discharge (EOD) waveform of very long duration (between 3.5 and 4 milliseconds at 25° C) with head-positive first phase significantly longer than second head-negative phase in both sexes. Brachyhypopomus (Odontohypopomus) bennetti sp. n. is diagnosed by two character states in addition to those used to diagnose the subgenus Odontohypopomus: (1) a deep electric organ, visible as large semi-transparent area, occupying approximately 14–17% body depth directly posterior to the abdominal cavity in combination with a short, but deep, caudal filament, and (2) a monophasic, head-positive EOD waveform, approximately 2.1 milliseconds in duration in both sexes. These are the only described rhamphichthyoid gymnotiforms with oral teeth, and Brachyhypopomus bennetti is the first Brachyhypopomus reported to have a monophasic (head-positive) EOD waveform. Unlike biphasic species, the waveform of its EOD is largely unaffected by tail damage from predators. Such injuries are common among specimens in our collections. This species’ preference for floating meadow habitat along the major channels of the Amazon River basin may put it at particularly high risk of predation and “tail grazing.”

the medial and lateral portions of the upper jaw form a continuous curve with little to no inflection point, viewed externally. However, this form of maxilla also occurs in the short-snouted rhamphichthyoid genus Steatogenys and in the Family Sternopygidae and may just be a concomitant feature of short (as opposed to more tubular) snouts. Albert (2001) recognized a monophyletic group consisting of the Brachyhypopmus species recognized by Mago-Leccia (1994) and several undescribed forms on the basis of four synapomorphies: (1) premaxilla gracile with a curved anterior margin and forming a distinct angle with the maxilla in lateral view, (2) dentary gracile, (3) body cavity with 16 or 17 precaudal vertebrae, and (4) a single transitional vertebrae. We regard these characters in combination with those enumerated by Mago-Leccia (1994) as provisionally sufficient to diagnose Brachyhypopomus, with the exception that pre-caudal vertebrae may be fewer than indicated by Albert: B. bullocki Sullivan & Hopkins (2009) has a short abdominal cavity with only 11-13 precaudal vertebrae. An unpublished phylogenetic analysis of mitochondrial DNA sequences in Sullivan (1997) indicated monophyly of eleven species, seven of which are now treated as valid Brachyhypopomus, with four additional undescribed forms, two of which are those treated here. This group of eleven species is monophyletic with respect to Hypopomus artedi, Microsternarchus bilineatus, and species of Hypopygus, Steatogenys, Rhamphichthys and Gymnorhamphichthys (Sullivan 1997).
Within gymnotiforms the complete absence of oral teeth is a character state unique to the Hypopomidae and Rhamphichthyidae and is among those used to unite these two families into the superfamily Rhamphichthyoidea (Mago-Leccia 1976, 1978, 1994, Triques 1993, Sullivan 1997, Albert 2001, Albert and Crampton 2003. The two species described here are remarkable for being the only rhamphichthyoids known to bear premaxillary teeth.

Materials and methods
Fishes were collected during day trips from Manaus, Brazil in a motorboat between March and May 1993; others were collected during the Calhamazon Project (Cox Fernandes et al. 2004) in November and December of the same year. The primary collection site for the type material is a few kilometers due south of Manaus in a series of channels, shallow lakes and islands that lie between the blackwater Rio Negro and the whitewater Rio Solimões, close to their confluence, as well as the Ilha da Marchantaria, a large, seasonally flooded island in the Solimões itself. We transported freshly captured individuals to a laboratory at the Instituto Nacional de Pesquisas da Amazônia (INPA) in Manaus with water from their capture locality. We recorded their EODs in a 10 cm x 40 cm x 12 cm aquarium with silver/silver-chloride electrodes positioned at the head (positive electrode) and tail (negative electrode) of the fish and a reference electrode in the center. EODs were amplified using a CWE Corporation bio-amplifier with filters set to 0.1 Hz to 50,000 Hz using low gain and captured with a Tektronix 222 digital storage oscilloscope (512 point/8-bit resolution). Longer recordings of EOD trains from the specimens here designated as holotypes were recorded on a Sony Walkman Pro cassette tape recorder, later digitized at 48 kHz on an Edirol FA-66 (Roland Corporation, Los Angeles, CA) and written to wav files with Audacity 2.0 software. Specimens were overanesthetized in MS-222, photographed in a photo aquarium, tagged, and fixed in 10% buffered formalin. Procedures for handling, euthanizing and preserving fishes followed guidelines for the use of fishes in research (AFS/ASIH/ AI-FRB 1987).
We examined the type material for all described Brachyhypopomus species with the exception of those recently described from southern Brazil and Uruguay (B. jureiae, B. bombilla, B. draco, and B. guaderio) for which we consulted the published descriptions. We also examined a large quantity of non-type material of both described and undescribed forms (see Comparative Material Examined, below). Measurements were taken with a digital, needle-point caliper to within 0.1 mm under low power magnification. All measurements were taken point-to-point, i.e. not orthogonal to the main body axis. Counts of anal-fin rays and vertebrae were made from film radiographs of the specimens, observed under magnification. All vertebral counts began with C5, the first vertebra to bear a neural spine. "Precaudal vertebrae" include all anterior vertebrae bearing neural spines up to the first vertebra to bear a hemal spine. Vertebrae bearing hemal spines are termed "caudal vertebrae." Counts of pectoral-fin rays, made with the aid of dissecting microscope and strong transmitted light, include all elements. Measurements were taken on the left side unless otherwise specified.
Anatomical measurements and abbreviations follow Hubbs and Lagler (1958). Three measurements require explanation. (1) LEA is the length from the tip of the snout to the posterior end of the anal-fin base. This measurement is generally used as standard length in descriptions of gymnotiforms, since most lack caudal fins and often have regenerated a portion of their caudal filament after injury from predators. Specimens that had suffered such damage anterior to the terminus of the anal fin were identified from radiographs, and excluded from measurements involving the length of the body reported for the type series. LEAs reported on damaged individuals are noted as such.
(2) Head length (HL) is measured from the tip of the snout to the end of the opercle bone, not to the uppermost limit of the branchial membrane. (3) Interorbital width is the distance between the upper margins of the eyes. The term "branched" pectoral-fin rays refers to all rays posterior to the anterior unbranched rays, even if the posterior terminal ray is unbranched. The abbreviation "alc" is used to indicate specimens that are preserved in alcohol, "cs" for those that have been cleared and stained. Institutional abbreviations follow Sabaj Pérez (2012).
Because measuring body depth is problematic due to lack of external landmarks on these fishes, distance from the tip of the snout to the 1 st , 20 th and 40 th caudal vertebrae were obtained from film radiographs of these specimens. These distances were then measured off on the specimens themselves and the depth of the body at each of these three points was obtained with digital needle point calipers. Measurements are presented as percentages of LEA except for those within the head that are presented as percentages of HL.
Staining protocols for bone and cartilage followed Potthoff (1984). In order to count columns of electrocytes in preserved specimens, skin was peeled back from caudal filaments and electrocytes were observed with strong transmitted light. "Electrocyte columns" refers to the number of bilateral bands of electrocytes along the longitudinal axis of the fish that begin under the head and continue to the tip of the caudal filament. These bands are most visible just above the posterior anal-fin base and on the caudal filament. These columns can often be counted without special preparation by viewing the area with strong transmitted light. Diagnosis. This subgenus of Brachyhypopomus is diagnosed by (1) teeth present on premaxillae: usually one to five small, needle-like teeth on ventral surface of each ( Fig.  1); (2) medialmost two branchiostegal rays thin with blades oriented more vertically than remaining three rays; (3) background color in life and to lesser extent in alcohol distinctly yellowish, head and sides peppered with small, widely spaced, dark brown stellate chromatophores that greatly contrast with background color of skin; bands along sides poorly defined, saddles of pigment mostly incomplete over dorsum; (4) a diffuse blotch of subcutaneous pigment directly beneath orbit, suggestive of a teardrop; (5) EOD pulse waveform of very long duration (head-positive phase approx. 2 milliseconds or longer, head-negative phase shorter or absent; Fig. 2) and slow repetition rate (3-16 Hz).
Teeth are absent from the premaxillae in all other rhamphichthyoid species, but present in all other gymnotiform lineages. (The teeth in preserved Odontohypopomus tend to be obscured by overlying tissue and are only easily visible in cleared and stained specimens.) In other Brachyhypopomus, the first (more medial) one or two branchiostegal rays are wide and oriented nearly horizontally, EODs are of shorter duration and faster repetition rates with second head-negative phases nearly equal in amplitude or of greater duration than the head-positive first phase, teardrop-like pigment below the orbit is absent, background color is less yellowish and chromatophores not as dark. In several other Brachyhypopomus species the pigment on the anterior flanks is arranged into distinct bands.
In all other Brachyhypopomus with biphasic EODs, the second, head-negative phase of the EOD is nearly equal in amplitude or of greater duration than the headpositive first phase. Microsternarchus bilineatus has an EOD waveform of similar duration, but the second phase is roughly equal or longer than the first (and the repetition rate is far faster). No other species of Brachyhypopomus is as distinctly yellow in color, particularly in life.
This species can be distinguished from the similar B. bennetti sp. n. by a shorter body (depth quickly tapers posteriorly: depth of body at 40 th post-abdominal vertebra 36-41% of depth at first abdominal vertebra vs. 46-57% in B. bennetti), fewer anal fin rays (198-216 rays vs. 227-255 rays in B. bennetti), a shallower electric organ, and a long, fine caudal filament (length 20-32% of LEA vs. 10-19% of LEA in B. bennetti) with three or four bilateral columns of electrocytes at base of caudal filament (vs. six columns in B. bennetti sp. n.). Subcutaneous pigment below eye is absent in other hypopomids and usually less conspicuous in the sister species B. bennetti sp. n. The EOD waveform of B. walteri sp. n. is biphasic in contrast to B. bennetti's monophasic EOD waveform. Brachyhypopomus bennetti sp. n. tends to be more darkly pigmented and less yellow and translucent.
Description. Morphometric and meristic data are presented in Tables 1, 3 and 4. A Brachyhypopomus of moderate to small adult size for a hypopomid; largest specimen examined measures 175 mm TL, 125 mm LEA. Body very compressed, depth at posterior end of abdominal cavity 2.7-3.1 times body width. Body more compressed posteriorly, sides of body with only slight curvature posterior to abdominal cavity. Dorsal profile gently convex. Depth quickly tapers posteriorly: depth of body at 40th post-abdominal vertebra 36-41% depth at first abdominal vertebra. Head short in comparison to body length, deep and wide: HL 11.2-12.6% LEA, head depth at occiput 72-81% HL, head width at opercle 54-63% HL. Head triangular in lateral view, dorsal profile of head straight from occiput to point of downturn of snout, ventral profile of head straight from lower jaw to opercular area with little if any concavity between opercular area and tip of lower jaw. Eye moderate in size, 12.4-14.5% HL. Mouth small, terminal, jaws equal, gape 20-23% HL. Closed lips meet ventral to a horizontal through ventral margin of eye. One to five small needle-like conical teeth present on each premaxilla ( Fig. 1), lower jaw edentate. Maxilla moderate in length, thin, with slight curvature. Snout moderate in length, 26-29% HL, edge of upper lip close to farthest anterior extent of snout. Posterior naris close to eye, posterior nariseye 1.8-3.7% HL. Lateral ethmoid present. Round ossification present in anterior portion of palatine cartilage (Fig. 1). Infraorbital portion of cephalic lateralis system nostril, pores inconspicuous. Preopercular lateral-line canal embedded in preopercle, canals radiating out to pores. Pores of lateral-line canal immediately behind head without downward pointing tubes. Discernible lateral scales terminate along caudal filament. Five branchiostegal rays, medialmost two thin with blades oriented nearly vertically compared to outer three (see diagnosis of Odontohypopomus). Gill rakers robust for genus, some with weakly ossified cores, on anterior faces of first four gill arches. Rakers subtended on ceratohyals one to four by small trough-shaped ossicles. Approximately 40 gill filaments on arch one. Three pectoral radials, all partially fused together at proximal end. Mesocoracoid bridge absent. Pectoral fin broad, 12-15 branched plus unbranched rays, length 5.3-7.0% LEA. 198-216 anal fin rays, longest rays 4.0-4.9% LEA. Precaudal vertebrae 13-16, up to 75 caudal vertebrae in advance of regenerated portion of caudal filament. Body excluding head and fins covered with thin cycloid scales, small dorsally, larger posterolaterally, partially obscured by skin. Twelve scale rows above, 13 scale rows below lateral line at farthest extent of pectoral fin. Anal-fin origin slightly posterior to vertical at midpoint of extended pectoral fin. Caudal filaments long and fine in intact mature specimens, 20-32% of TL. Sexual dimorphism of caudal filaments not observed. Three or four bilateral columns of electrocytes along caudal filament, number often alternating along length of caudal filament; 38-63 rows of electrocytes. Electrocytes do not extend farther anteriorly than base of urogenital pore. No accessory electric organs on head or humeral region.
Coloration. Background color yellow in life, yellowish-tan in preservation. In life, body semi-translucent, with gill filaments appearing cherry red through opercle, gut dark, and swim bladder whitish through abdominal wall. Pigmentation variable: poorly to moderately developed irregular bands along sides, darker and wide above lateral line, often with a spot of darker intensity on lateral line itself. Bands either restricted to anterior portion of body above lateral line or connected to fainter bands below. Some bands connect to eight to 12 irregular saddles across dorsum. Saddles more regular in smaller individuals. Dorsal rami of the anterior lateral line nerve visible when viewed from above as two thin, dark parallel lines running along upper back beginning a short distance behind head and continuing to mid-point of the back. Cheeks, underside of head and sides of body below lateral line peppered with prominent dark brown stellate chromatophores that greatly contrast with background color of skin and that do not form part of a larger pattern. Diffuse pigment below eye resembling a teardrop is more prominent in live specimens as overlying tissue becomes opaque upon preservation. Pectoral and anal fin with irregular brown pigment along rays; interradial membranes hyaline.
Distribution and ecology. See distribution map (Fig. 5). Brachyhypopomus walteri sp. n. is known only from the Amazon basin where it appears to be common in floating meadow habitats, (mostly composed of the grass Paspalum repens, Poaceae), on the margins of the Amazonas/Solimões and its tributaries. It has been collected predominantly in white water, but also in areas near the confluence of black water rivers with the Amazonas/Solimões ranging from low to medium conductivity. Apart from one collection very near Manaus and the white water Rio Branco, it is absent from collections in the Rio Negro system. It is frequently taken with B. bennetti sp. n. and sometimes with B. brevirostris. Species of Eigenmannia, Gymnotus, the apteronotid Parapteronotus hasemani and the electric eel, Electrophorus electricus, frequently co-occur in the floating meadow habitats preferred by this species.
Etymology. This species is named for Walter Heiligenberg  in honor of his discoveries in electric fish neurophysiology and behavior made at the Scripps Institute of Oceanography. These notably include the "jamming avoidance response" in Eigenmannia, often described as the best-understood vertebrate behavior.    Diagnosis. Brachyhypopomus (Odontohypopomus) bennetti sp. n. is diagnosed by two character states in addition to those used to diagnose the subgenus Odontohypopomus above: (1) electric organ along caudal filament and along body above anal fin exceedingly deep and visible as large semi-translucent area, occupying approximately 14-17% body depth directly posterior to abdominal cavity; (2) monophasic, headpositive EOD, approximately 2.1 milliseconds in duration in both sexes at 25°C. The appearance of the electric organ in this species when backlit (Fig. 8) is significantly larger than in any other species of Brachyhypopomus. No other described Brachyhypopomus has a monophasic EOD waveform.
This species can be distinguished from the similar B. walteri sp. n. by a longer body (depth gradually tapers posteriorly: depth of body at 40 th post-abdominal vertebra 46-57% vs. 36-41% of depth at first abdominal vertebra vs. in B. walteri), more numerous anal fin rays (227-255 vs. 198-216 in B. walteri), a deeper electric organ along the body and a short, deep caudal filament (10-19% of LEA vs. 20-32% of LEA in B. bennetti) with six bilateral columns of electrocytes at its base (vs. three or four columns in B. bennetti sp. n.). Subcutaneous pigment suggestive of a teardrop below eye is usually less conspicuous than in the sister species B. walteri sp. n., although often present. The EOD waveform of B bennetti sp. n. is monophasic in contrast to B. walteri's biphasic EOD waveform. B. walteri sp. n. tends to be less darkly pigmented and more translucent and yellowish in life than B. bennetti sp. n.
Description. Morphometric and meristic data are presented in Tables 2-4. A Brachyhypopomus of moderate adult size for a hypopomid; largest specimen examined measures 232 mm TL, 189 mm LEA. Body very long and compressed, depth at posterior end of abdominal cavity 2.6 to 2.9 times body width. Body more compressed posteriorly, but sides of body with only slight curvature posterior to abdominal cavity. Dorsal profile gently convex. Depth gradually tapers posteriorly: depth of body at 40th post-abdominal vertebra 46-57% of depth at first abdominal vertebra. Head short in comparison to body length, deep and wide: HL 10.3-12.3% LEA, head depth at occiput 76-80% HL, head width at opercle 58-65% HL. Head triangular in lateral view: dorsal profile of head straight from occiput to point of downturn of snout, ventral profile of head straight from lower jaw to opercular area with little if any concavity between opercular area and tip of lower jaw. Eye moderate in size, 11-14% HL. Mouth small, terminal, jaws equal, gape 21-26% HL. Closed lips meet ventral to a horizontal through ventral margin of eye. One or more small needle-like conical teeth present on each premaxilla. This feature is variable, in one case observed only unilaterally. Lower jaw edentate. Maxilla moderate in length, thin, with slight curvature. Snout length moderate, 26-30% HL, edge of upper lip close to farthest anterior extent of snout.
Posterior nostril particularly small and close to eye: posterior naris-eye 2.3-4.3% HL. Lateral ethmoids present. Round ossification present in anterior of palatine cartilage. Infraorbital portion of cephalic lateralis system incomplete, lacking recurrent anterodorsal segment and associated pores beneath and anterior to the posterior nares present  Electric organ discharge. The EOD has a simple, head-positive monophasic waveform with a total duration 1.9-2.4 milliseconds at 25°C (Fig. 2). No sexual dimorphism has been observed. Resting EOD rate is very slow (2.0-8.9 Hz, mean 4.7 Hz, median 4.9 Hz, at 21-25° C, n=31). See Appendix III.
Coloration. Background color yellowish-tan in life, brownish-tan in preservation. Pigmentation variable: poorly to moderately developed irregular bands along sides, darker and wide above lateral line, often with a spot of darker intensity on lateral line itself. Bands either restricted to anterior portion of body above lateral line or connected to fainter bands below. Some bands connect to 8-12 irregular saddles across dorsum. Saddles more regular in smaller individuals. Dorsal rami of the anterior lateral line nerve visible when viewed from above as two thin, dark parallel lines running along upper back beginning a short distance behind head and continuing to mid-point of the back. Cheeks, underside of head and sides of body below lateral line peppered with prominent dark brown stellate chromatophores that greatly contrast with background color of skin and that do not form part of a larger pattern. Pectoral and anal fins with irregular brown pigment along rays, interradial membranes hyaline.
Distribution and ecology. See distribution map, Fig. 5. Brachyhypopomus bennetti sp. n. is known only from Amazon Basin where it appears to be common in floating meadow habitats on the margins of the Amazonas/Solimões River and its tributaries. Its distribution and habitat preference seems very similar to that of its sister species, Brachyhypopomus walteri sp. n., with which it is often collected.  Etymology. This species is named for Michael V.L. Bennett of the Albert Einstein College of Medicine of Yeshiva University, Bronx, New York, in honor of his pioneering work on electric fish neurophysiology. Bennett (1961Bennett ( , 1971 reported studying a Brachyhypopomus (therein Hypopomus) with a monophasic EOD likely to have been this species.

Affinities of the new subgenus
Together, several shared character states unobserved in other Hypopomidae and so presumably derived (teeth on the premaxillae, branchiostegal-ray orientation, similar pigmentation and long duration EODs with slow repetition rates) are strong evidence that these two new species are closest relatives among the described Brachyhypopomus. An unpublished analysis of mitochondrial DNA sequences in Sullivan (1997) further supports this conclusion. We take the step of erecting the new subgenus Odontohypopomus to provide a name to this distinctive, toothed subgroup of Brachyhypopomus that will serve to unite them should the genus be reorganized in the future. Other subgroups of Brachyhypopomus have already been recognized, although this one is the first to be formerly named. (Until others are, all other Brachyhypopomus species belong to the nominotypical subgenus Brachyhypopomus by default.) While phylogenetic relationships among all the described species of hypopomids remain to be determined, the available morphological and molecular data in Sullivan (1997) Loureiro & Silva, 2006 within Brachyhypopomus is difficult to assess from its description alone. We leave the task of recognizing additional subgenera of Brachyhypopomus to other authors currently revising the group.
When first collected, we considered whether these two morphotypes represented sexual dimorphism within a single species (indeed the two are frequently lumped in existing museum lots), but the observation of unambiguous males and females within each type dispelled this possibility. The two species are most easily distinguished from each other by differences in the thickness of the electric organ, the length of the caudal filament and their EOD, as well as by anal-fin ray counts. Brachyhypopomus bennetti sp. n. has a short but very deep caudal filament (length less than 20% TL) and a monophasic EOD waveform. By contrast, caudal filaments of B. walteri sp. n. measure 20-30% TL in individuals with sufficiently regenerated caudal filaments and this species has a very long, biphasic EOD waveform. Damage to the caudal portion of the body is common in both species and many specimens show incomplete regeneration. The electric organ of B. bennetti has six columns of electrocytes at the base of the caudal filament, that of B. walteri only three or four, and B. bennetti's organ viewed with transmitted light appears to occupy much more tissue above the anal fin musculature, along length of the body, than in B. walteri (Fig. 8). The EODs of B. bennetti are monophasic, while those of B. walteri are biphasic (Fig. 2). Brachyhypopomus bennetti have 227-255 anal fin rays while B. walteri have considerably fewer, 198-216. Furthermore, the teardrop-like pigment below the eye is usually more prominent in B. walteri, although also present to varying degrees in individuals of both species. We have examined an insufficient sample of cleared and stained specimens to judge if there are differences in the number of premaxillary teeth between the species, although in a sample of three cleared and stained specimens we found no more than two teeth per premaxilla in B. bennetti while up to five teeth per premaxilla in a sample of four cleared and stained B. walteri. In both species, the number of teeth is variable. Unfortunately, it is difficult to see these teeth in alcoholic specimens as they are obscured by thick, opaque tissue lining the roof of the mouth.

Ecological and evolutionary considerations
The probable sister-species status of B. walteri and B. bennetti is especially interesting given that they are frequently collected together (including at their common type locality) and have a broadly overlapping geographic distribution. Both species are known exclusively from the Amazonas/Solimões basin and seem to prefer the root tangle of large, floating grass meadows that are common along the margins of the Amazon's riverfloodplain system (Fig. 9). Why the ancestor of these two species would have regained oral teeth remains an interesting question, one that could perhaps be addressed by future studies of the diet and feeding behavior of these two species relative to their congeners. Brachyhypopomus bennetti is unusual both for its remarkably large electric organ (as proportion of body depth occupied, depth of the caudal filament and a high number of horizontal columns of electrocytes) and an unusual EOD waveform that consists of a simple, head-positive pulse of long duration. Hopkins (1999) suggested that the differences in electric organ structure among species of Brachyhypopomus may often be adaptations to the conductivity of water in a species' preferred environment. Other species with five or six parallel columns of electrocytes and short caudal filaments such as B. diazi and B. occidentalis tend to be found in high conductivity environments (above 150 µS/cm), while species with three columns and extended caudal filaments such as B. brevirostris and B. bullocki are most commonly found in lower conductivity environments (often well below 100 µS/cm). The first type of organ, with more electrocytes in parallel, but fewer in series, has low internal resistance and is adapted to generating current in water with low resistivity (high conductivity). The latter type, with more electrocytes in series, but fewer in parallel, has higher internal resistance and is capable of generating the higher voltages necessary for passing current through highly resistive (low conductivity) water. Thus, species-specific differences in electric organ structure may often reflect impedance-matching to water conductivity regimes (Hopkins 1999).
Conductivity in the white water floating meadow habitat where B. bennetti and B. walteri were collected is intermediate for Neotropical freshwater habitats: between 60 and 100 µS/cm (pers. obs.). The characteristics of B. walteri's electric organ (3 or 4 columns, intermediate length caudal filament) are similar to those seen in the probable sister clade to Odontohypopomus, B. pinnicaudatus + B. beebei. (Although these latter two species similarly occupy intermediate resistivity environments and are sometimes collected within floating meadows, they are primarily found farther from large river channels in "terra firme" creeks and lagoons.) Given this phylogenetic assumption, it seems probable that B. bennetti's enlarged electric organ with five or six horizontal electrocyte columns evolved from an ancestor with an electric organ similar to that seen in B. walteri, B. pinnicaudatus and B. beebei, but as an adaptation to something other than high water conductivity.
In a study that considered selective pressures on gymnotiform EOD waveform evolution, and in which B. bennetti and B. walteri were identified as "sp. 1" and "sp. 2," respectively, Stoddard (1999) reported that amplitude-calibrated recordings of B. bennetti's monophasic EOD show them to be very much more powerful than those of other, similarly sized Brachyhypopomus species and between five and ten times greater amplitude than those of B. walteri. The unusual EOD waveform, hypertrophied electric organ and high-amplitude EOD of Brachyhypopomus bennetti invites inquiry into the possible adaptive value of these features.
Biphasic EODs in pulse gymnotiforms may have evolved from primitive monophasic EODs as a means to reduce the low frequency/direct current component of the signal to which electroreceptive predators (other gymnotiforms and catfishes equipped with ampullary-type electroreceptors) are sensitive (Stoddard 1999, Stoddard 2002, Stoddard and Markham 2008. Monophasic EODs are rare among modern adult gymnotiforms with pulse-type EODs: the only species reported to have them apart from B. bennetti, Electrophorus electricus (itself an electroreceptive predator) and one species of Gymnotus (G. cylindricus) from Guatemala. The first is protected from predation by its strong electric discharge and the second is geographically isolated from electric eels and other electroreceptive predators such as pimelodid catfishes. Noting that B. bennetti's monophasic waveform is nearly identical in duration and shape to that of the electric eel, Stoddard (1999) suggested that the convergence may be a form of Batesian mimicry to deter predation by electroreceptive predators that associate monophasic EODs with electric eels.
Our fieldwork confirms that electric eels are common at the Odontohypopomus collection localities, as are piscivorous Gymnotus species that likely account for some of the tail damage observed in our specimens. Preference for floating meadow habitat near deep water also make these species vulnerable to predation from large pimelodid catfishes, such as Pseudoplatystoma tigrinum, a species that may specialize on gymnotiforms (Reid 1983, Zuanon 1990). Exposure to the high predation pressure that likely characterizes Amazonian floating meadow habitat might favor the evolution of an EOD mimic of E. electricus and Stoddard's hypothesis makes predictions with respect to the behavior of electroreceptive predators that should be tested. However, in our collections we note similar proportions (>60%) of both B. bennetti and B. walteri that exhibit regenerated caudal filaments and posterior anal fin rays from earlier predation. We do not know the identity of these "tail grazers" and what proportion of them are electroreceptive, but the monophasic EOD of B. bennetti clearly does not prevent a high proportion of individuals from suffering such injuries.
An alternative (or additional) advantage of EOD monophasy in Brachyhypopomus bennetti may be related to the fact that, in contrast to its biphasic relatives, its EOD waveform remains largely unaltered after tail predation (Fig. 10). In a typical biphasic Brachyhypopomus EOD, the anterior and caudal portions of the electric organ do not contribute equally to the head-positive and head-negative phases. Only the posterior portion of the electric organ that includes the caudal filament produces the head-negative second phase to the EOD waveform, while the anterior electric organ produces a mostly head-positive, monophasic pulse (Bennett 1961, 1971, Caputi 1999, Stoddard 1999, Stoddard and Markham 2008. For this reason, individuals suffering predation injuries to the caudal filament and caudal portion of the body produce EODs with attenuated head-negative second phases (Fig. 10 A, B). Electrical crypsis by biphasy may be effectively impossible under conditions of heavy "tail grazing," in which case selection may favor other adaptive solutions. EOD monophasy for electric eel mimicry is one interesting possibility, but monophasy for stability of the waveform is another.
Active electroreception and electric communication rely upon "tuberous"-type dermal electroreceptors some populations of which are narrowly tuned to the peak frequency of the fishes' own EOD (Bastian 1976, 1977, Hopkins 1976 and to the EODs of conspecific individuals detected at a distance (Hopkins and Heiligenberg 1978). Any mismatch between a fishes' EOD frequency spectrum and the frequency sensitivity of its own electroreceptors (and those of conspecific individuals) will be deleterious. In species with biphasic EODs, tail predation not only decreases the amplitude of the EOD, but alters its waveform and thus its frequency spectrum ( Fig. 10 A, B), whereas such injury in Brachyhypopomus bennetti only affects amplitude (Fig 10 C). Hence it is worth considering that the monophasic EOD of Brachyhypopomus bennetti may have evolved to provide its electrosensory system greater robustness to tail injuries. Likewise, the positioning of more of the electric organ rostrally on the body (as opposed to along an exposed caudal filament) and a high amplitude EOD would increase resiliency of electrosensory function in fish that regularly lose caudal electrocytes to predation. These two hypotheses to account for the distinctive characteristics of B. bennetti's EOD and electric organ are not incompatible: this species could, in theory, enjoy advantages from both electric eel mimicry and EOD waveform stability simultaneously.
Sister species Brachyhypopomus walteri lacks these characteristics of the electric organ and EOD despite also persisting in floating meadows. However, differences in behavior or feeding ecology of these two species might result in exposure to different selective pressures, even in the same habitat, and future study may yet indicate differ-ences in niche breadth and distribution between B. walteri and B. bennetti. The very different EOD waveforms of these two species may mediate reproductive isolation between them and reproductive character displacement may have also played a role in the divergence of these signals, as has been suggested for sympatric Gymnotus species .