Chirping response of weakly electric knife fish (Apteronotus leptorhynchus) to low-frequency electric signals and to heterospecific electric fish

In many species, sensory systems are tuned to be most sensitive to particular spectral or temporal features of its species-specific communication signal. Matching between the reception and production of signals ensures that animals communicate primarily with members of their own species (Bradbury and Vehrencamp, 1998). However, in some cases, heterospecifics can elicit the same kind of communication signals as those commonly used in intraspecific interactions. Such responses to heterospecifics have been widely documented in animals that communicate in acoustic, olfactory and visual modalities (Ord and Stamps, 2009), but they have been examined very little in animals that communicate in the electric modality. Here, we explore how an electric fish, Apteronotus leptorhynchus, responds to heterospecific electric fish and to low-frequency electric stimuli that lie far below the peak sensitivities of the electroreceptors they typically use for intraspecific communication.

Brown ghost knife fish (Apteronotus leptorhynchus) generate continuous, high-frequency (700–1100 Hz) electric signals that convey information about species and sexual identity. During social interactions, they modulate the continuous electric organ discharge (EOD) by briefly (10–100 ms) increasing the frequency of their EOD to produce signals termed ‘chirps’ (Larimer and MacDonald, 1968; Hagedorn and Heiligenberg, 1985; Zupanc and Maler, 1993). The precise communicatory function of chirps is not fully understood, but they are emitted at particularly high rates when male fish engage in agonistic interactions with other male fish of similar EOD frequency (within ∼100 Hz) (Dunlap, 2002; Hupé and Lewis, 2008; Triefenbach and Zakon, 2008). Chirps are also emitted, albeit at much lower rates, during sexual interactions (Hagedorn and Heiligenberg, 1985; Dunlap, 2002). Because males have higher EOD frequencies than females (Zupanc and Maler, 1993; Dunlap et al., 1998), the frequency differences in male–female pairings are typically larger (100–400 Hz) than in male–male interactions. During social interactions, or during experimental presentation of high-frequency sine waves that mimic another fish, the electrical stimuli activate tuberous electroreceptors, whose peak sensitivities closely match the fish's own EOD frequency (Hopkins, 1976). Although the high-frequency signals within the species range are the most effective stimulus for eliciting chirps, we noticed that low-frequency signals (10–300 Hz) that are 600–900 Hz lower than the Apteronotus species range also elicit chirping. This observation caused us to hypothesize that low frequencies contained within the EOD of other electric fish species could stimulate chirping and suggested the possibility that chirping might serve some function in encounters with heterospecifics.

There are several potential sources of low-frequency stimuli in the waters surrounding these fish (Bodznick and Montgomery, 2005). Many aquatic organisms produce low-frequency (0.5–100 Hz) electric fields resulting from respiratory movements and nervous and muscular activity. These fields are almost always in the microvolt range and are detected primarily by ampullary electroreceptors, which are tuned to low frequencies (1–100 Hz). In addition, some species of gymnotiform fish that are sympatric with Apteronotus generate continuous EODs that contain low-frequency components. For example, Sternopygus macrurus emits a wave-type discharge with a low dominant frequency (50–150 Hz) (Hopkins, 1974a). Other species, such as Brachyhypopomus gauderio, emit a pulse-type discharge that contains a broad spectrum of frequencies, and, in some cases (e.g. mature males), the spectrum includes substantial energy in the low-frequency range (Franchina et al., 2001). These low-frequency components of heterospecific EODs are in the millivolt range. Finally, some gymnotiforms make low-frequency electric fields when they modulate their EOD. For example, Eigenmannia produce low-frequency stimuli when they briefly interrupt their EODs during social interactions (Naruse and Kawasaki, 1998).

Given these potential heterospecific sources of low-frequency electric fields, we sought first to characterize the chirping response of Apteronotus leptorhynchus to artificial low-frequency stimuli. We then examined the chirping response to the low-frequency energies contained in the EOD of two other electric fish, S. macrurus and B. gauderio, and to the passive low-frequency emissions from a non-electric fish, Carassius auratus. We found artificial low-frequency stimuli (10–300 Hz) directly elicited chirping in A. leptorhynchus and potentiated the response to high-frequency stimuli. Moreover, fish chirped in response to the two other electric fish species, suggesting that chirps might function in communicating to conspecifics about the presence of other electric fish species or in interspecific interaction.

All the subject fish were adult male Apteronotus leptorhynchus that were 12–19 cm in length. Males of this species emit a wave-type EOD with a dominant frequency of 850–1100 Hz (Fig. 1A). Fish were obtained from commercial dealers and housed either in 285-litre group tanks or 38-litre individual tanks that were part of a 1235-litre system. Water temperature (26–28°C), conductivity (400–500 μS cm^–1) and pH (6.0–6.5) and the light:dark cycle (12 h:12 h) were held constant. All fish were housed in isolation for at least 7 d before experimental testing. Each experiment described below was conducted with a different set of subject fish. All procedures in this study were approved by the Trinity College Animal Use and Care Committee and adhere to the guidelines specified by the National Institutes of Health (DHEW Publication 80–23).

To determine the effect of stimulus frequency on chirp rate, we used a chirp testing apparatus termed a ‘chirp chamber’ identical to that described in previous studies (Zupanc and Maler, 1993; Dunlap et al., 1998). In brief, the chirp chamber consists of a PVC tube (length=17.7 cm, inner diameter=3.8 cm) fitted at each end with electrodes that recorded the EOD of the fish. The recording was amplified (100–1000×) with an amplifier (Grass Instruments, West Warick RI, P-55) that was connected to an oscilloscope and audio speaker. Stimuli were presented through two carbon rods attached to the sides of the tube and connected to a function generator (Pasco Scientific Roseville, CA, PI–9587C). The stimulus strength for all stimuli was 1–2 mV cm^–1, measured with paired electrodes spread 1 cm and placed at the midpoint between stimulus electrodes.

Male fish (N=12) were placed individually in a chirp chamber, and their EOD frequency was measured. After acclimating for 10 min, the fish were presented for 30 s with a sine-wave stimulus at one of five stimulus frequencies: 10, 30, 100, 300 Hz or 10 Hz below their own EOD frequency. Fish were presented with all five stimuli in random sequence, with each subject receiving only one stimulus per day and tested every 2–3 d. For this and all other experiments, the chirps were detected by listening to audio transformations of the electrical signal while the behavioral test was occurring.

To determine the effect of stimulus amplitude on chirp rate, we tested male fish (N=6) using the chirp chamber described above at five different stimulus amplitudes (0.3, 1, 3, 10 or 30 mV cm^–1) while keeping the stimulus frequency constant at either 30 or 150 Hz. Fish were tested in a chirp chamber and stimulated with a sine wave from a function generator, as above. We counted the number of chirps every 30 s for 1 min before stimulus onset to record the rate of spontaneous chirping and for 5 min after stimulus onset to record stimulus-evoked chirping. Fish were given a different stimulus amplitude once per day for 5 d, with a random stimulus order.

To test whether low-frequency stimuli alter the response to subsequent presentation of high-frequency stimuli, we used the chirp chamber described above and a protocol depicted in Fig. 2A. Male fish (N=8) were allowed to acclimate in the chamber for 10 min and then presented with one of three treatments. In all cases, the stimulus was a sine-wave signal of constant duration (30 s) and strength (2–3 mV cm^–1). In the high-high treatment (HH), fish were presented with a stimulus at 10 Hz below their own frequency, a so-called ‘jamming stimulus’, followed by a 3 min period without stimulus and then another identical stimulus (10 Hz below own EOD frequency). In the low-high treatment (LH), fish were given a 30 Hz stimulus, followed by a 3 min period without stimulus and then another stimulus at 10 Hz below their own EOD frequency. In the control treatment, fish remained unstimulated for the first 3 min and were then presented with a single stimulus at 10 Hz below their own EOD frequency. Each fish was tested once per day for 3 d, with a different stimulus on each day; the stimulus order was randomized.

To determine whether A. leptorhynchus chirps in response to heterospecific fish that have low-frequency components in their EOD, we recorded the response of A. leptorhynchus (N=6) when placed in the same aquarium with either S. macrurus or B. gauderio. To determine whether non-electrical heterospecifics also elicit chirping, we examined the response to a goldfish, C. auratus.

Sternopygus macrurus emits a wave-type discharge, with the dominant frequency varying among individuals from ∼50–150 Hz (Fig. 1B). The signal also has lower-amplitude components at higher harmonic frequencies. For the present experiment, two S. macrurus, 32 and 39 cm in length, were obtained from a commercial dealer and housed together in a 285-litre tank. At 28°C, these two individuals had EOD frequencies of 145 and 151 Hz and created a maximum field strength of approximately 50 mV cm^–1 at 1 cm from the fish.

The test tank consisted of a 76-litre aquarium that was divided into compartments with rigid nylon-mesh barriers (Fig. 2B). A single A. leptorhynchus and a single S. macrurus were placed on opposite sides of these barriers. The barriers were moved to vary the field strength experienced by the subject fish. In one test, the S. macrurus was placed in a compartment (length×width×depth: 25×30×35 cm) that was 24 cm from the compartment (13×30×35 cm) with the A. leptorhynchus (Fig. 2B). This created a relatively low stimulus strength (1–5 mV cm^–1) in the compartment occupied by the A. leptorhynchus. In another test, the S. macrurus was placed in a compartment (22×30×35 cm) that was directly adjacent to the compartment (13×30×35 cm) housing the A. leptorhynchus (Fig. 2B). This created a relatively high field strength (30–50 mV cm^–1) experienced by the A. leptorhynchus.

At the beginning of the trial, the A. leptorhynchus was allowed to acclimate to the test aquarium for 10 min. The total number of spontaneously emitted chirps was counted for 5 min. Then the S. macrurus was introduced into its compartment, and the total number of chirps was counted for an additional 5 min. All A. leptorhynchus were tested with each S. macrurus, and each A. leptorhynchus was tested only once per day.

Brachyhypopomus gauderio (formerly B. pinnicaudatus) produces a pulse-type EOD with a broad spectrum of frequencies and a variable pulse rate (15–70 pulses s^–1; Fig. 1C). At night, the low-frequency component of the EOD increases in amplitude and the pulse rate speeds up (Franchina and Stoddard, 1998). In the present study, B. gauderio were reared in the laboratory of P. Stoddard at Florida International University and were 14 and 15 cm in length. In our laboratory, they were housed initially in group tanks, and then, one week before the experiment, they were housed in pairs in 38-litre aquaria. For measuring the chirp response of Apteronotus leptorhynchus to B. gauderio, we used a 38-litre test aquarium that was subdivided into two equal compartments, each of which measured 10 cm×10 cm×12 cm. Apteronotus leptorhynchus were placed individually into the test compartment, and the number of spontaneous chirps emitted during the first 5 min was recorded. Then a B. gauderio was introduced into the adjacent compartment, and the chirp rate of A. leptorhynchus was recorded for an additional 5 min. Each A. leptorhynchus was tested with one of two B. gauderio on successive days. The field strength of the B. gauderio in the compartment of the A. leptorhynchus was 4–9 mV cm^–1. The pulse rate of the B. gauderio was 25–29 pulses s^–1 when alone and 36–40 pulses s^–1 when placed in the compartment adjacent to an A. leptorhynchus.

To determine whether a non-electric fish affected the chirp rate in A. leptorhynchus, we tested A. leptorhynchus with goldfish (C. auratus), using the method identical to the experiment described above with B. gauderio. The goldfish were purchased from commercial dealers and weighed 11 and 13 g.

To determine the threshold amplitude for chirping response by A. leptorhynchus to B. gauderio, we recorded the EOD of a B. gauderio and played it back to A. leptorhynchus in a chirp chamber. B. gauderio were enclosed in a PVC tube. The EOD was amplified, collected for 5 min through head-to-tail silver electrodes located at either end of the tube, sampled at 44 kHz using Canary software (version 1.2.4) and recorded onto the sound card. The EOD recording had a stable inter-pulse interval with a discharge rate of 27–29 pulses s^–1. To vary the amplitude of the EOD playback, we passed the signal through an adjustable attenuator to yield peak-to-peak amplitudes of 1.5, 3.0, 10 and 30 mV cm^–1 measured with paired electrodes separated by 1 cm, with the midpoint centered on the midpoint between the stimulus electrodes. Male A. leptorhynchus (N=6) were placed individually in a chirp chamber, allowed to acclimate for 10 min and presented with a B. gauderio EOD playback through carbon electrodes placed on either side of the fish. Each A. leptorhynchus was presented with one of the four amplitudes on four successive days. The order of stimulus was randomized among fish and across days. We recorded the number of chirps in each 30 s bin for 1 min before stimulus presentation and for 5 min during stimulus presentation.

To determine whether the low-frequency component contained within the EOD of B. gauderio influences the chirping response of A. leptorhynchus, we recorded the EOD of a B. gauderio, filtered out the low frequencies (<150 Hz), and compared the responses of A. leptorhynchus to playbacks in which the low frequencies were absent to those in which they were present (Fig. 1C). The B. gauderio EOD was recorded for 3 min at night, when the low-frequency component is highest (Franchina and Stoddard, 1998), using the method described above. To test whether Canary software could accurately record and deliver frequencies at the low end of the spectrum, we used a function generator to generate sine-wave signals with a range of frequencies (1–100 Hz) and recorded these signals in Canary. We then played these signals through the playback apparatus while measuring the resulting amplitude on an oscilloscope. We found that the signal attenuated very little (<5%) above 10 Hz, but the signal dropped off considerably below 10 Hz.

From the initial EOD recording, we generated to two playback signals: an unfiltered signal that contained the full spectrum of frequencies (high+low) and a filtered signal in which all the frequencies below 150 Hz were greatly reduced (high only; Fig. 1C). We filtered the signal using the DFT projection method contained within the Canary software. For both filtered and unfiltered stimuli, the pulse rate was 29 pulses s^–1 and the peak frequency was 851 Hz. The amplitude of the playback was adjusted so that the stimulus was always 30 mV cm^–1 at the midpoint between the stimulus electrodes. This represents a stimulus strength that a fish would experience at approximately 5 cm from an adult Brachyhypopomus (Stoddard et al., 1999).

Male A. leptorhynchus (N=8) were placed individually in the chirp chamber, allowed to acclimate for 10 min, presented one stimulus type (high only or high+low), given an additional 10 min period of rest, and then presented with the alternative stimulus. We counted chirps in 30 s bins for 1 min before stimulus presentation and during each 3 min stimulus presentation. Fish were tested on two days, with the order of stimulus presentation reversed on the second day.

To compensate for small variations in temperature (26–28°C), all chirp rates were adjusted to 28°C using a Q₁₀ of 3.16 (Dunlap et al., 2000). To compare chirp rates between treatments and across time, we used Prism 5.0 software to conduct repeated measures ANOVA, with stimulus treatment (amplitude, frequency or filtered vs unfiltered playback) as dependent variables and time as the repeated measure. Bonferroni post-tests enabled us to determine at what time-points chirp rates were statistically elevated above the pre-stimulus period or statistically different between treatment groups. In heterospecific pairings with two individual stimulus fish, we used a two-way repeated measures ANOVA, with the stimulus fish as a cofactor. In all cases, chirp rate was not statistically different between the two stimulus fish, so the data from these trials were pooled. Data are presented as means ± s.e.m. P<0.05 was considered statistically significant.

When presented with a stimulus of constant amplitude (1–2 mV cm^–1), A. leptorhynchus chirped at significantly higher rates towards stimuli signals of 100 and 300 Hz than to signals of 10 and 30 Hz (Fig. 3; F=24.3, d.f.=3, P<0.0001). Chirp rates towards these low-frequency signals (10–300 Hz) were reduced by a factor of 4–8 compared with the chirp rates of fish presented with a jamming stimulus 10 Hz below EOD frequency. Of 12 fish tested, only 25% made any chirps to 10 Hz stimuli, whereas 75% of fish chirped to stimuli at 30–300 Hz, indicating that there is considerable individual variability in the threshold response to stimuli of <30 Hz. All fish chirped and produced a jamming avoidance response to a jamming stimulus. However, we did not see a jamming avoidance response to any low-frequency (≤300 Hz) stimuli.

For both 30 and 150 Hz stimuli, fish chirped at rates above the pre-stimulus period at 1, 3, 10 and 30 mV cm^–1 (Fig. 4; P<0.01), but not 0.3 mV cm^–1 (P>0.05), indicating that the threshold for response at these frequencies is between 0.3 and 1 mV cm^–1. For both stimulus frequencies, chirp rates were elevated above pre-stimulus values for the entire 5 min stimulus duration at 10 and 30 mV cm^–1. For all these stimuli, the chirp rate peaked at 0.5–1.5 min and then declined for the following 3.5–4.5 min. At 3 mV cm^–1, the chirp rate became elevated above that of the pre-stimulus period only at 1–1.5 min of stimulation (P>0.005), indicating that the fish are slower to respond to weaker stimuli. For the 30 Hz stimulation at 3 mV cm^–1, the fish chirped at baseline rates after 3 min of stimulation. At 1 mV cm^–1, fish responded above baseline at only a single time bin, 1–1.5 min after stimulus onset. The chirp rate was significantly elevated (P<0.001) above baseline for the entire stimulus duration at 10 and 30 mV cm^–1. For all stimuli except 30 Hz at 10 mV cm^–1, fish decreased their chirp rate after reaching a peak, indicating that habituation occurred over the 5 min period. At 30 Hz and 30 mV cm^–1, fish showed a second but small peak in chirp rate in the final 1.5 min of the trial. As above, we did not observe a jamming avoidance response to any stimuli.

When fish were presented with high-frequency (jamming) stimuli twice within 3 min, they chirped at higher rates to the second stimulus than to the first stimulus or a single stimulus at the same time-point (F=4.4, d.f.=1, P<0.01; Fig. 5). This indicates that a high-frequency stimulus potentiated the response to the same stimulus. When presented with a low-frequency stimulus, the fish chirped at rates lower than those when presented with high-frequency stimuli (F=16.1, d.f.=1, P<0.0001). However, when fish were presented with a high-frequency signal 3 min after a low-frequency stimulus, they chirped at higher rates than those when presented with the control stimulus (F=4.1, d.f.=1, P<0.05) and equivalent to those at the second presentation of a high-frequency signal (F=1.0, d.f.=1, P>0.05). Thus, although low-frequency stimuli did not elicit as much chirping as high-frequency stimuli, they were equally effective in potentiating the response to high-frequency stimuli.

Apteronotus leptorhynchus significantly increased its chirp rate above baseline levels when exposed to S. macrurus and B. gauderio but not to goldfish (Carassius) (Table 1). When A. leptorhynchus were placed at long distance from S. macrurus and experienced a relatively low-amplitude field strength (1–5 mV cm^–1), A.leptorhynchus chirped at rates comparable to the pre-pairing period (F=0.82, d.f.=1, P>0.05). However, when they were placed adjacent to S. macrurus and experienced a relatively high field strength (30-50 mV cm^–1), they significantly increased their chirp rate (F=9.8, d.f.=1, P<0.0001). There were no differences in the response of A. leptorhynchus to each of the S. macrurus or across test days (P>0.05). A. leptorhynchus occasionally swam near the barrier, scanning back and forth as though investigating the S. macrurus. More commonly, A. leptorhynchus appeared agitated and avoided spending time near the S. macrurus stimulus fish. The S. macrurus displayed no apparent reaction to the A. leptorhynchus.

Apteronotus leptorhynchus exposed to B. gauderio increased chirp rate above the pre-stimulus period on both test days (F=12.9, d.f.=1, P<0.0001) but chirped significantly more on the first test day than on the second day (F=2.8, d.f.=1, P<0.05). The response did not differ according to which individual B. gauderio fish was used (P>0.05). During the behavioral trials, the A. leptorhynchus usually swam actively throughout the test compartment, showing neither a preference nor aversion to the area near the B. gauderio. However, A. leptorhynchus occasionally lunged towards the B. gauderio that occupied the adjacent compartment. Brachyhypopomus gauderio had no discernable motor reaction to the A. leptorhynchus, but increased its EOD discharge rate from ∼27 to ∼38 pulses per second.

When we presented A. leptorhynchus with B. gauderio EOD playbacks that varied in amplitude, the mean chirp rates differed from baseline levels when the stimulus was 10 and 30 mV cm^–1 (Fig. 6; P<0.001), but not 1.5 or 3.0 mV cm^–1 (P>0.05), indicating that, for most A. leptorhynchus, the threshold for chirping to a B. gauderio is between 3.0 and 10 mV cm^–1. Although four fish had a threshold at this level, two fish chirped significantly above baseline at 1.5 mV, and thus some individuals appear sensitive to considerably lower signal strengths. At 10 and 30 mV cm^–1, chirp rates decreased by about 50% compared with the peak over the 5 min stimulus (Fig. 6).

Apteronotus leptorhynchus chirped significantly more to a B. gauderio EOD that contained low-frequency components than to the same EOD in which the low-frequency components were filtered out (Fig. 7; F=3.1, d.f.=1, P<0.05). The effect was present only in the first min.

Most animals use species-specific signals in courtship and territorial interactions. This specificity prevents animals from making mistakes in mate selection and incurring the costs (e.g. energetic or predation risks) of unnecessary signaling to inappropriate receivers (Bradbury and Vehrebcamp, 1998). Nevertheless, a recent meta-analysis showed that, in a surprising number of cases (approximately one-third), animals respond similarly to heterospecific and conspecific communication signals (Ord and Stamps, 2009). Such interspecific communication can have considerable fitness consequences (Gröning and Hochkirch, 2008) and is the subject of a growing body of theory in the field of animal communication (Ord and Stamps, 2009; Wiley, 2006).

In electric fish, chirping has always been considered a signal for electrocommunication among conspecifics because it can be readily evoked by the presence of conspecifics or by artificial signals within the frequency range of the species (Hagedorn and Heiligenberg, 1985; Zupanc and Maler, 1993; Zupanc, 2002). We show here that heterospecific electric fish and artificial signals that contain low-frequency energy far below the Apteronotus frequency range also elicit chirping. Fish chirp less to heterospecific stimuli than to conspecific stimuli, but our results nevertheless broaden our notion of the kinds of stimuli that provoke chirping responses and cause us to ask whether its expression extends beyond conspecific aggression and courtship.

Chirp rates are significantly elevated above background rates when presented with sine waves that range from 30–300 Hz, with some fish responsive to frequencies as low as 10 Hz. Chirp rates to these low-frequency stimuli (8–15 chirps min^–1) are only about one-fifth of those to a ‘jamming’ signal that mimics conspecific males and that maximally elicits chirping (∼60 chirps min^–1). We did not measure the response to frequencies that mimic the presence of a female (∼700–825 Hz), but the chirp rates towards females recorded under similar conditions in other studies [25 chirps min^–1 (Kolodziejski et al., 2007)] were about twice the rates we recorded towards low frequencies. So clearly, intraspecific high-frequency signals are the most important activators of chirping. Nevertheless, chirp rates towards low-frequency stimuli are over ten times the spontaneous chirp rate, indicating that low-frequency stimuli can elicit a significant response.

Fish respond consistently to low-frequency sine waves that are above 1 mV cm^–1 in field strength. The chirp response threshold is much higher (about 1000 times) than the threshold for a behavioral response in a conditioning task (1–2 μV cm^–1) (Knudsen, 1974). So, it appears that fish do not simply reflexively chirp to the minimal detectable signal, and instead the response likely involves higher-order brain processing. The idea that low-frequency stimuli are processed in higher levels of the electrocommunication circuitry is also supported by our finding that low-frequency stimuli are able to potentiate the chirping response to subsequent high-frequency stimuli. Low- and high-frequency stimuli are equally effective in potentiating chirping, indicating that the underlying sensory pathways that encode these stimuli have equal weight in the higher-order ‘memory’ of the electromotor system.

We found that A. leptorhynchus increased its chirp rate when exposed to nearby Sternopygus or Brachyhypopomus. No response was made towards Carassius, indicating that the mere presence of another fish and the low-amplitude field emanating from a non-electric organism does not elicit this response. Apteronotus also chirped to heterospecific (Brachyhypopomus) EOD playbacks or sinusoidal stimuli mimicking a heterospecific EOD (Sternopygus), demonstrating that the electric stimuli created by heterospecific electric fish are sufficient for eliciting chirps.

Although our study is the first to document electric signaling between two live heterospecific gymnotiforms, Hopkins (Hopkins, 1974b) observed the behavioral response of Eigenmannia to EOD playbacks of four heterospecific electric fish (A. albifrons, S. macrurus, Platyurosternarchus macrostomus and Gymnorhamphichthys hypostomus). Eigenmannia responded to conspecifics much more than to heterospecifics, yet they nonetheless made EOD interruptions towards the signals of other species, whose dominant EOD frequencies lay outside the range of Eigenmannia. This suggests that other gymnotiforms might also make electrocommunication responses to heterospecific electric fish.

Apteronotus leptorhynchus is sympatric with S. macrurus throughout much of its range in the neotropics (Albert and Crampton, 2005). The distribution of A. leptorhynchus likely does not overlap with that of B. gauderio, but it overlaps with other species in the Brachyhypopomus genus that have an EOD similar to that of B. gauderio. Thus, it is plausible that A. leptorhynchus encounters fish in these genera in their natural habitat (W. Crampton, personal communication). There have been few observations of interactions among gymnotiform species in the wild, but they are known to prey on each other (Cox Fernandes, 1999). Sternopygus can grow to over 100 cm in length, and we once observed a Sternopygus (∼40 cm) in the laboratory eating a small Apteronotus (∼5 cm), suggesting that they pose a predatory threat to Apteronotus in the wild. Given the small size of Brachyhypopomus (<24 cm), it is unlikely that it threatens Apteronotus, but other large pulse-type fish cohabit waters with Apteronotus and could easily threaten and perhaps elicit a chirping response from Apteronotus.

Our data indicate the A. leptorhynchus chirp response towards heterospecific fish can be influenced by the distance and duration of the encounter. When placed together with Sternopygus, A. leptorhynchus chirped above background rates only when the two fish were within 35 cm, where the field strength of the stimulus fish was 30–50 mV cm^–1. We did not directly measure the effect of spacing in the interaction between A. leptorhynchus and B. gauderio, but, extrapolating from the response threshold to B. gauderio EOD playbacks (3–10 mV cm^–1) and the measured field strength of closely related Brachyhypopomus species (Stoddard et al., 1999), A. leptorhynchus would need to be within ∼5 cm of a B. gauderio to respond with chirps.

The chirp rate towards B. gauderio EOD playbacks and sine-wave stimuli (150 Hz) that mimic Sternopygus EOD peaks in the first 0.5–1 min and diminishes by about 50% over the remaining stimulus period (3–5 min). However, except for the lowest effective stimulus amplitude (1 mV cm^–1) at 150 Hz, chirp rates at the end of the stimulus period were significantly higher than the pre-stimulus period, indicating that fish never fully habituated within this time-frame.

The Apteronotus chirp rate is enhanced by the low-frequency components of Brachyhypopomus gauderio EOD, but only in the initial portion of the stimulus presentation. During the first minute of stimulus presentation, fish chirped more to an unfiltered EOD playback than to the same playback in which the low-frequency components were filtered out experimentally. The responses to the two stimuli were the same after one minute. This habituation to the low-frequency component is consistent with the response to low-frequency sine waves, in which fish habituate to low-amplitude stimuli (1 mV cm^–1) after the first 1–1.5 min of stimulus presentation. The strength of the low-frequency component of B. gauderio EOD is greater in males than in females, particularly at night (Franchina et al., 2001; Franchina and Stoddard, 1998). Thus, the heterospecific chirping response of A. leptorhynchus would likely depend on the sex of the B. gauderio it encountered and the time of day. Stoddard (Stoddard, 1999) has shown that these low-frequency components of the male B. gauderio EOD also elicit a behavioral response in the electric eel (Electrophorus electricus), a potential electropredator of B. gauderio. Given that the chirp response of A. leptorhynchus towards heterospecifics is influenced by many factors – distance, duration, time of day and sex of stimulus fish – it will be crucial to map spatially and temporally the distributions of these fish in their habitat to assess the relevance of our present study for understanding behavioral interactions in the field.

We can only speculate about the function of heterospecific-evoked chirping behavior. It is possible that this chirping has no communicative function at all and is simply a manifestation of electrosensory arousal. Apteronotus emit chirps at a low rate (∼0.2–1 chirp min^–1) in the absence of any electrosensory stimuli (Engler et al., 2000), and low-frequency- and heterospecific-evoked chirping might simply be an increase in that background rate in response to generalized electrosensory activation.

Alternatively, this chirping could serve as an alarm signal to alert conspecific fish to the presence of a heterospecific electric fish. In many social species, the arrival of a heterospecific predator or competitor triggers the production of an alarm signal that provokes a defensive or aggressive response in conspecifics (Marler, 1955). A long history of research has shown that, among diverse vertebrates, these signals are communicated through the auditory, visual or olfactory modalities (Bradbury and Vehrebcamp, 1998). Recent work on mormyriform electric fish in Africa suggest that alarm signals might also occur in the electric modality. Scheffel and Kramer (Scheffel and Kramer, 2006) observed interspecific interactions among three sympatric electric fish species in large semi-natural tanks and noticed that the fish change their interdischarge interval (IDI) as if it functioned as an alarm signal. When Petrocephalus catosoma encountered a Hippopotamyrus szaboi entering its territory, it decreased the interdischarge interval and made its discharges much more regular. P. catosoma in neighboring territories subsequently responded with similar short and regular IDIs, indicating that the presence of the heterospecific intruder was communicated to nearby fish. Scheffel and Kramer hypothesized that heterospecific-induced changes in IDI could serve as an electrical warning signal to alert conspecifics to avoid intruding predators. To evaluate whether chirping could similarly act as an alarm signal, one would need to show that Apteronotus modify their behavior when exposed to heterospecific-induced chirps from a conspecific and can discriminate between heterospecific- and conspecific-induced chirps.

Finally, the chirps we described could serve in interspecific aggression or competition. Apteronotus commonly live sympatrically with many other electric fish that have similar preferences for prey and retreat sites (Cox Fernandes, 1999; Albert and Crampton, 2005). Heterospecific evoked chirps might signal to competing fish to stay away. Again, to test this possibility, it would be necessary to determine whether heterospecifics can detect and respond behaviorally to Apteronotus chirps.

Tuberous receptors are commonly considered the initial step in the sensory pathway involved in electrocommunication because they are most responsive to the frequency range of the species (Hopkins, 1976; Bullock, 1982; Metzner and Viete, 1996). However, we have shown that the production of electrocommunication signals (i.e. chirps) is also elicited by low-frequency signals that lie far below the best frequency of tuberous receptor sensitivity. How then are these low frequencies detected and transmitted to the electromotor system underlying chirping? One possibility is that tuberous receptors transmit the signal even though they specialize in higher-frequency detection. Hopkins (Hopkins, 1976) examined the response of tuberous receptors in A. albifrons to a broad range of stimulus frequencies using a stimulus strength (1.1 mV cm^–1) similar to the chirp response threshold we found in A. leptorhynchus. He found that tuberous receptors had a bimodal response as a function of stimulus frequency, with one peak in receptor firing rate near the EOD frequency of the fish and another in the low-frequency (30–100 Hz) range. The response rate to low-frequency stimuli was almost as strong (∼85%) as the response to the fish's own EOD frequency. If the same bimodal sensitivity is present in A. leptorhynchus, its chirping response to low-frequency stimuli could be activated through the tuberous system. P-type tuberous receptors encode low-frequency amplitude modulations (Nelson et al., 1997) and might be able to also transmit the low-frequency stimuli we presented.

Alternatively, low-frequency stimuli could be conveyed to the electrocommunication circuitry through ampullary receptors, which have particular sensitivity to stimuli in the low-frequency range (Bodznick and Montgomery, 2005). The dominant frequency of the Sternopygus EOD and the low-frequency components of the Brachyhypopomus EOD are certainly within the frequency range that can activate ampullary receptors. Ampullary receptor sensitivity has not been measured in A. leptorhynchus, but threshold amplitude for ampullary receptor activation in another gymnotiform fish (Gymnotus) is ∼0.1–3 mV cm^–1 at low frequencies (30–150 Hz) (Dunning, 1973). This is below or near the threshold for eliciting chirps in Apteronotus, suggesting that ampullary receptors are sufficiently sensitive to participate in low-frequency-induced chirping.

Ampullary receptors are commonly considered a separate system serving passive electroreception rather than electrocommunication (Bullock, 1982). Indeed, ampullary activation is processed in a separate pathway through many of the electrosensory processing regions of the brain. However, in Eigenmannia, ampullary pathways functionally converge with tuberous pathways at certain neurons of the torus semicircularis (Metzner and Heiligenberg, 1991; Rose and Call, 1992; Fortune and Rose, 1997), an important premotor processing area, and could thereby influence the production of chirps. Moreover, ampullary receptors are likely involved in the detection of low-frequency components of chirps in Eigenmannia, and such ampullary activation can influence behavior (Naruse and Kawasaki, 1998). Thus, the specialization of ampullary receptors for low-frequency detection combined with evidence that they can send information to pre-electromotor regions of the brain suggest that ampullary receptors contribute to low-frequency-evoked chirping.

Our study shows that chirping is not confined to intraspecific aggression and courtship but rather is a behavior associated with a continuum of stimulus conditions. Fish chirp at the highest rates to conspecific males, followed by conspecific females, then heterospecific electric fish and finally at low spontaneous rates in the absence of stimuli. Determining whether heterospecific-evoked chirps have any specific communicative value requires examining the response of conspecifics and heterospecifics to playbacks of these chirps. In some cases (e.g. the response to conspecific males vs females), fish change chirp structure as well as chirp rate, and this variation in chirp structure likely conveys an important meaning in intraspecific interactions (Zupanc, 2002). Examining the structure of heterospecific-evoked chirps might help clarify their function in communication.

We thank V. Salvador, S. St Jean and J. Nord for help with animal care, J. Larkins-Ford, M. Chung, P. Garberg, K. Pappas for data collection, and P. Stoddard, M. Chung and A. Silva for comments on previous drafts.