Journal of J Comp Physiol A (1986) 159:297-310 Sensory, Neural, Comparative and Physiology A ~..,ora,Physiology Springer-Verlag 1986

Acoustic communication in an electric fish, Pollim yrus isidori (Mormyridae)*

John D. Crawford, Mary Hagedorn, and Carl D. Hopkins Section of Neurobiology and Behavior, Seeley Mudd Hall, Cornell University, Ithaca, New York 14853, USA

Accepted June 5, 1986

Summary. It has been known since von Frisch's kins 1986; Bell and Szabo 1986). Like many other work in the 1930's that mormyrid electric fishes teleosts, the mormyrids also exhibit specializations are quite sensitive to sound. We now describe a for detecting acoustic signals. In the labyrinth, a repertoire of natural sounds produced by the mor- small gas bladder coupled to the sacculus probably myrid, Pollimyrus isidori, during breeding and ag- enhances sensitivity to sound pressure (Stipetic gression; reception of communication sounds is 1939). This anatomy was described over 150 years probably a major function for mormyrid audition. ago by Heusinger (1826). In contrast to the electric 1. In aquaria, Pollimyrus isidori produce modality, very little is known about biologically ' grunts ', ' moans', ' growls ', ' pops' and ' hoots' at significant acoustic signals. various phases during nesting, courtship, and terri- Based. on behavioral observations, von Frisch tory defense. (1938a) concluded that mormyrids were quite sen- 2. All five sounds are produced primarily at sitive to sound although he did not know the sig- night. Territorial males produce grunts, moans and nificance of this since he thought the fish's world growls during courtship. Vocalizing is stimulated to be silent. Since von Frisch, a number of studies by the presence of a gravid female on the male's have supported and extended his conclusions con- territory and decreases with the onset of spawning. cerning mormyrid audition (Stipetic t939; Kramer Hoots and pops are given during agonistic behav- et al. 1981; McCormick and Popper 1984). The ior. only reported mormyrid sounds are the 'clicks' of 3. Grunts are bursts of acoustic pulses, stereo- Gnathonemus petersii (Rigley and Marshall 1973), typed for an individual, with the potential as indi- and nothing else is known about the natural acous- vidual signatures. tic environment of any mormyrid. 4. The electric organ is silent during grunts and Following techniques developed by Birkholtz moans and is discharged at a reduced rate during (1969 and 1970) and by Kirschbaum (1975, 1979, growls. 1982, 1984) for inducing breeding in mormyrids, 5. The courtship and spawning of Pollimyrus we have studied breeding behavior in Pollimyrus isidori is described. isidori. We have discovered that acoustic signals play a critical role in breeding behavior. In this paper we describe the repertoire of acoustic signals and the breeding and courtship behavior of Polli- Introduction myrus isidori. A preliminary report of our findings has been published in abstract form (Crawford The Mormyridae is a diverse family of African et al. 1985). freshwater fishes, of over two hundred species, dis- tinguished by their electrogenic capabilities and Materials and methods highly developed electric sense (reviewed in Hop- 1. Animals and animal care. We established a community of Abbreviation: EOD electric organ discharge 2 male and 3 female Pollimyrus isidori (Cuvier and Valenciennes 1846) in a 4801 aquarium (L=183cmxW=38cmxH= * Sequence of Authors determined by a toss of a coin 54 cm; Fig. 1). Males were distinguished from females by not- 298 J.D. Crawford et al. : Acoustic communication in an electric fish ./ / / / \

....',. .:;, " ~ ! ... ,.. t~i" ~ ...... '/ " ..

A B C D E "~ Fig. 1. Breeding aquarium equipped with two hydrophones and five electrodes (only three electrodes are visible in drawing). On the left, male AB tends a nest in a large tube. Two subordinate females are shown hovering high in the water column within sections B and D. The large gravid female hovers near the bottom in section D. On the far right, the DE male approaches his nest

ing the ventral fin margin: males have a concave indentation EOD recording in different parts of the aquarium, a permanent where the anal fin inserts along the ventral body surface where- electrode was placed in each of its 5 compartments (Fig. 1). as the insertion is straight in females (Iles 1960). Males and A switch outside the aquarium allowed one to select the most females were also distinguished by noting differences in the appropriate electrode. electric organ discharge (EOD) waveform (Westby and Kirsch- baum 1982). The fish were free to swim between five semi- 3. Analysis of acoustic" signals. The sound pressure of acoustic isolated compartments, created by four opaque plastic parti- signals was monitored with a pair of calibrated hydrophones tions, but soon adopted preferred areas within the aquarium. (Celesco Transducer Products Model LC-34 and Celesco pre- Cement cinder blocks and sections of plastic pipe were also amplifiers LG-1344) and further amplified with a Princeton provided for shelter. Aquatic vegetation (Ceratopteris thalic- Applied Research amplifier (Model 113; filter settings: low cut- troides and Leptodictyum riparum) and filter floss were added off=0.i Hz; high cut-off=30 kHz). A Nagra IV-SJ tape re- for increased heterogeneity and to serve as nesting material. corder was used for recording acoustic signals and for estimat- Water temperature was maintained at 27 _+ 2 ~ Adult fish were ing sound pressure levels (SPL); SPL was calculated from read- fed on Tubifex worms daily. Larvae were free to graze on algae ings of the tape recorder's calibrated meter, on the 'peak' scale, and zooplankton, and were provided with finely chopped Tubi- while a fish produced a sound at a known position. For SPL fex. The fish were maintained on a photoperiod regime with measurements, a hydrophone was positioned in the middle of 11.5 h light and 11.5 h dark phases separated by 30 min dim section B (Fig. 1) about 12 cm from the bottom; water depth white light 'dawn' and 'dusk' periods. Five 3 W red lights was approximately 35 cm. From this reading, the input to the were illuminated behind the aquarium at all times so that during tape recorder was determined (RMS V) and the output of the the dark phase there was sufficient light for behavioral observa- hydrophone was calculated. SPL (0 dB referred to 1.0 gbar = 0.1 tions. N/m 2) was determined from the calibration curve for the hydro- To induce spawning, water conductivity was steadily re- phone. Due to the complex acoustics of aquaria (Parvulescu duced from about 1,000 gMhos/cm to 40 gMhos/cm by daily 1966; Hawkins and Maclennan 1976), the SPLs determined additions of approximately 351 of deionized water. When in this way should be taken only as approximations of the aquarium capacity was reached, the water level was decreased SPLs that might be measured at similar distances under natural by 50%. Rain was simulated by pumping aquarium water or free-field conditions. through overhead sprinklers for 3 h periods separated by 3 h Signals recorded at 38.1 cm/s were used for sound analysis. intervals. Sonograms were made with a Kay Electric Co. Sound Spectro- graph (7029A) or with a Spectral-Dynamics 301C spectrum 2. Analysis of electric signals. EODs were detected with silver/ analyzer. Sounds were filtered at 100 to 10,000 Hz and digitized silver-chloride wire electrodes, amplified with a Grass P-15 dif- and stored directly on a computer disk, at a sampling rate ferential preamplifier (filter settings: high cut-off= 50 kHz, low of 20 kHz using a 12-bit A/D converter. Waveforms were Four- cut-off=0.1 Hz), and digitized with a Tektronix 5233 digital ier-analyzed as above. In the complex sound field of the aquar- oscilloscope (1.0 MHz sampling rate). Fast Fourier Transforms ium, sound waveforms undoubtedly depended on the precise of digitized EODs were done on a PDP-I1/34 computer using location of the fish and hydrophone. Where possible, several a Digital Equipment Corporation FFT routine. examples of recorded waveforms have been provided to illus- To facilitate simultaneous behavioral observations and trate variability. J.D. Crawford et al. : Acoustic communication in an electric fish 299

4. Simultaneous EOD and sound analysis. With an electrode are more common. This sequence terminates with input to one channel of the tape recorder and a hydrophone a long growl. Sound pressure levels, at distances input to a second, we could record EODs and acoustic signals of 1-3 cm, were on the order of + 28 dB for these of a single fish simultaneously. Temporary barriers were used to prevent other fish from approaching the recording site so three types of vocalizations (Fig. 5). EOD contamination was minimal. Acoustic contamination was eliminated by selecting nights when only a single animal was 1. Grunts. Grunts were the most stereotyped of vocalizing. Tapes made at 38.1 cm/s were played back at 3.81 the vocalizations. Lasting approximately 300 ms, cm/s or 1.905 cm/s and displayed on a two channel oscillograph chart recorder (Gould Brush Chart Recorder, Model 220). they consisted of a series of about 15 pulses of increasing amplitude (Fig. 2B). The pulse repeti- 5. Behavioral observations. Three sounds (grunts, moans, and tion rate was approximately 50 Hz. Figure 3A growls) were produced by male fish that had been isolated; shows examples of single pulses and their power we never recorded these three sounds from similarly isolated spectra. The peak of the power spectrum was in females. The remaining two vocalizations were not heard from isolated animals (see Results). To determine the location of the 500 to 1000 Hz range. Pulse waveform varied vocalizing fishes, hydrophones were placed at each end of the as the fish moved around in the aquarium but was long aquarium, and by listening in stereo, an observer could constant from cycle to cycle within a grunt (see determine the end of the tank from which the sound was made consistency in successive pulses in Fig. 3 A, 1 and (presumably this discrimination was based mainly on intensity cues). We confirmed our ability to locate sound sources by differences between examples 3A, 1-3). Single gently tapping the aquarium wall in different places. The identi- pulses were followed by a damped oscillation of fication of vocalizing individuals was reliable since there were about 500 Hz in our tank (Fig. 6). just two males vocalizing from opposite ends of the aquarium. Stereo tape recordings, made with the Nagra and analyzed on 2. Moans. Moans were the most tonal of the five the real-time spectrum analyzer, confirmed the sound source direction by intensity method. vocalization types. They had a nearly sinusoidal To correlate vocal behavior with other behaviors, including waveform that began softly and steadily increased overt motor patterns, position, and proximity to other fish, in amplitude (Figs. 2B and 3 B). They varied in we recorded sounds in stereo while coding behavioral data si- duration from 250 ms to about 3 s. The waveform multaneously on a third track of the tape recorder with a key- operated tone source. Behavioral events were then correlated of the moan was also stereotyped from cycle to with ongoing acoustic signals with 1 2 s resolution. Detailed cycle (Fig. 3B). The sonograms (Fig. 2A) and data on courtship maneuvers, egg deposition, and vocal activity power spectra (Fig. 3 B) reveal two harmonically- were obtained from 9 h of video recorded spawning (AB male, related peaks, a fundamental at about 220 Hz and 19 March, 1985). its harmonic at 440 Hz. Many moans were fre- Individual fish were recognized by natural markings and differences in standard length (SL). The two males were called quency modulated upward or downward by 8-18 'AB' (83 mm SL, 13.5 g) and 'DE' (85 mm SL, 13.4 g) accord- Hz. Frequency modulations were examined on real ing to the locations of their nesting areas. The three females time spectrograms with analysis bandwidth of were named large (82 mm SL, 13.0 g), medium (78 mm SL, 3 Hz. Downward sweeps were 2-3 times as com- 9.9 g), and small (74 mm SL, 7.0 g), according to size. mon as upward. Direction or presence of frequency modulation did not appear to be dependent upon Results behavioral context. A. Description of vocalizations 3. Growls. Growls, like grunts, were composed of Sounds were first recorded when males AB and pulses, but each pulse had characteristic ringing DE began building nests within their respective ter- waveform, producing a broad spectral peak at ritories (Fig. 1). A rich repertoire of acoustic sig- about 210 Hz (Fig. 3C). Growls were highly vari- nals, including ' grunts', 'moans', and ' growls' able in duration with the longest bouts (no silences (Figs. 2 and 3) was heard at night from males sta- > 5.0 s) lasting about 40 s and the shortest, less tioned on their territories. ' Hoots' and' pops' were than 2 s. The pulse repetition rate was about 25 Hz observed only during agonistic encounters (Fig. 4). or about half that of the grunt (Fig. 3 C). These vocalizations were relatively infrequent and we have not recorded them from isolated animals. 4. Pops. Pops had a sharp onset, very short dura- Our behavioral observations of interactions be- tion and thus a broad power spectrum (Fig. 4A). tween animals suggest that they are produced by These click-like sounds were only heard during males and females during aggression. fights, chases or other movements involving rapid Figure 2 shows a typical sequence of acoustic maneuvers. signals recorded from the AB male. Grunts and growls may be produced singly but sequences be- 5. Hoots. Hoots were the most infrequent sounds ginning with several grunts separated by moans and also the most variable (Fig. 4 B and C). Hoots 300 J.D. Crawford et al. : Acoustic communication in an electric fish

o.. 04

Lt3 N 'grint' "l- v 'giwl' o. 'moan' A) I g

0

B)

Time(s) Fig. 2A, B. A typical courtship vocalization sequence of Pollimyrus #idori is composed of grunts, moans and a growl. The sonogram (A) was made with a Kay Sonagraph in the 20-2,000 I-Iz frequency range (wide band: time resolution=13 ms, filter bandwidth = 75 Hz). The oscillogram (B) was made from a digitized record of the same sound

A) Grunt B) Moan C) Growl , , , %. . .,...,

q o N , .,.-., 0.0 25.6 51.2 ' 101 102 103 104 O.O 25.6 51,2 % 101 102 103 104 6.0 25.6 51.2 ~ 101 102 103 104 ms Frequency (Hz) ms Frequency (Hz) ms Frequency (Hz)

Fig. 3. Waveforms and power spectra of grunts (A), moans (B), and growls (C). To illustrate variability, three different samples have been analyzed for each sound type. Digitized waveforms are shown on the left and their power spectra on the right. The power spectra represent the average of the FFTs of 5 different single periods within the same vocalization (e.g. 5 of the 15 pulses comprising a single grunt)

often had a tonal quality with rapid frequency B. Individual specificity in temporal parameters modulation. Figure 4B shows a single 1.0 s burst of vocalizations with a peak power frequency of about 125 Hz. Figure 4C shows a hoot consisting of 4 shorter 1. Grunts. Grunts were highly stereotyped for each tonal bursts (each lasting about 200 ms). The wa- of the two males. Male AB produced grunts with veform plots for several of the bursts in Fig. 4C a mean pulse rate of 45.3 Hz (n = 14, SD = 1.2 Hz) (2 and 3) reveal a higher frequency, about 200 Hz, while male DE produced grunts at a mean rate than seen in Fig. 4 B. of 54.8 Hz (n= 14, SD= 1.8 Hz). The pulse rate J.D. Crawford et al. : Acoustic communication in an electric fish 301

Fig. 4A-C. Sonograms, waveforms and power spectra of pops and hoots. In each example, a sonogram is shown with its digitized waveform below on the same time base. In A and B, part of the waveform is expanded to the right and the corresponding power spectrum is shown. C An amplitude-modulated hoot is shown with its compressed (below) and expanded (right) time domain waveform. Frequency bandwidth, time and frequency resolution for each wide-band sound-spectrogram, A 80-8,000 Hz, 3.3 ms and 300 Hz; B and C 20-2,000 Hz, 13 ms and 75 Hz 302 J.D. Crawford et al. Acoustic communication in an electric fish

F~ 40 A TO~L ~ ~ CAMS ~ I

35 Grunts

Moans 30 Growls [] :!

O [] [] Pops 25 O~ 0 0 B 9 12 15 18 21 24 3 6 0 0 0 35 20 0 OO ,.Q 30 8[] o, 25 15 20 15 10 1 I 10 5 1 10 100 0 9 12 15 18 21 24 3 6 Distance from Fish (cm) Fig. 5. Sound pressure levels recorded at different distances from fish. Distance estimates were 1-3 cm, 4-6 cm, 7-10 cm, dusk nigtlt dawn 11-14 cm and 15-20 cm. SPLs were plotted according to the "riME (hrs) log of the midpoint of each of these ranges. Different vocaliza- Fig. 7A, B. Twenty-four hour periodicity in acoustic behavior. tion types were displaced slightly along the distance axis for One five-minute sample was taken every 30 min over a 24 h clarity period. The frequency of grunts, moans and growls is shown in A and hoots and pops are shown in B. Vocal activity is primarily restricted to the evening with peaks during dusk and pulse dawn

Sequence of Spawns for Two Males] and Large Female male (JAB: I I I I I C~ r DE: I I I I I ~LG" I I [I I I I I II I I [ I I I I I I I I I I 0 50 100 150 200 I I I I I I I I I NIGHTS I I I I I Fig. 8. Sequence of spawns for males AB and DE and the large female. Each vertical bar indicates a night when the large female spawned with one of the two males. The mean interval between spawns was 17.0_+ 10.6 nights for the large female, 39.5_+9.3 nights for male AB, and 31.5_+15.7 nights for male DE. The first spawn was observed on 16 December 1984 and the last 27 May 1985 (the animals continued to breed after this last "DE" ~~~~~t~ date but the data presented here are based upon this sequence male of 10 spawns)

of the two males was statistically different (Stu- dent's t-test, P<0.001). Figure 6 illustrates indi- vidual stereotypy and the repetition rate difference i I between individuals. Grunts displayed in Fig. 6 100 ms and used in the between-male comparison were re- Fig. 6. The pulse repetition rate of the AB male's grunt was corded on the same night to insure the same tem- 45 Hz, while that of the DE male was 55 Hz. Digitized portions perature (pulse repetition rate has a Qlo of approx- (130 ms) of three different grunts are shown for each male fish, imately 2.8 over the range 24-26 ~ A third male illustrating the high degree of stereotypy in repetition rate for was induced to vocalize in a sep- each individual and the consistent difference (dashed lines) be- Pollimyrus isidori tween the two animals. Note that each pulse is followed by arate aquarium and produced the same repertoire high frequency oscillations of vocalizations as males AB and DE. He also J.D. Crawford et al. : Acoustic communication in an electric fish 303

Fig. 9. Heightened vocal activity elicited by the gravid large female. During this continuous 250 s sound spectrogram, the large female swam onto the male's territory twice (dotted lines). The male responded by giving grunts and moans in alternation: we have called these sequences grunt-moan bouts (see Fig. 10). When the female left the territory, males usually terminated grunt-moan bouts and produced growls. This transition to growling is clear in the second trace when the female leaves (second arrow). In the last trace, when the female was on the male's territory again, the grunt-moan bout seemed to be initiated by the female's entrance but continued after she left. This was less typical. Real time spectra were made in the 0-3,000 Hz band (frequency and time resolution 30 Hz and 50 ms respectively). The two gaps in the record (lines 1, 3), were inserted to synchronize recording channels. The four clicks at 130 s were room noises showed little variability in his grunt pulse repeti- 2. Growls. More variability existed in the temporal tion rate (7(=55.1 Hz, SD=1.2 Hz, n=14, at parameters of growls than grunts. The AB male 26 ~ Unfortunately we cannot compare rates be- produced growls with a pulse repetition rate of tween males in different aquaria since we do not 25.6 Hz (n=12, SD=2.4) whereas DE's growls have the necessary precise temperature data. Grunt had a mean rate of 23.0 Hz (n= 11, SD =2.3). Thus duration was more variable than the pulse rate the mean pulse rates of the males differed (Stu- and there was no significant difference between in- dent's t-test, P < 0.02) but the ranges of pulse rates dividuals (Student's t-test, P>0.05) (AB male overlapped. Growl duration (a growling 'bout' 310+_ 58 ms, n = 14; DE male" 280+ 57 ms, n = 14). was defined as a period of growling not interrupted 304 J.D. Crawford et al. : Acoustic communication in an electric fish

Female influence on the duration of a C. Behavioral context of vocalizations male's Grunt-Moan calling bouts 1. Twenty-four hour pattern in vocal activity. Vocal activity was much greater at night than during the day. There was a peak in vocal activity at dusk and a second but less pronounced peak at dawn (Fig. 7). The swimming and feeding activity pat- 7-_ [l AB Ma,el terns were similar to the vocal activity pattern. 2. Agonistic interactions and territoriality. Figure 1 shows the fish in typical postures and locations 8 within the aquarium. The two males occupied and

4 defended areas at opposite ends of the aquarium, m rarely moving into the center section or encounter- ing each other. Males usually swam near the bot- DOMINANT SUBORDINATE NO FEMALE in presence of: FEMALE FEMALE tom and aggressively chased females from their ter- Fig. 10. Only the dominant (large) female elicited long grunt- ritories. The largest female also swam close to the moan bouts from males. The duration of grunt-moan bouts bottom. She occupied the central region of the were measured for each male under three different conditions: aquarium (section C and the large tube in D). She 1) while the dominant female (large) was on the territory, 2) was the only female that was observed to spawn. while one of the two subordinate females (medium or small) was on the territory, 3) while no female was on the territory. The smaller females were variable in their posi- From left to right, sample sizes (= number of bouts recorded) tions in the aquarium and were restricted to the are AB: 7 bouts, 40 bouts, 321 bouts and DE: 7 bouts, 9 upper regions of the water column by the other bouts, 49 bouts. Bars represent mean + 1.0 SD. See Fig. 9 for animals. These subordinate females generally ho- example of a grunt-moan bout and text for statistics vered within crevices or floating vegetation. They were frequently displaced by rapid chases initiated by males or the large female. Such chases were by other vocalizations or silences greater than commonly accompanied by a single pop. We ob- 5.0 s) was 11.2 s (n=20, SD=9.7) for male AB served several cases with pops in which the chasing and 4.3 s (n=20, SD=3.3) for male DE. These fish did not make contact with the subordinate. measures indicate high variability, and the two The outcomes of chases were scored during 25 males did not differ statistically (Mann-Whitney different 10 min observation trials. 'Winners' ag- U-test, U=229, P>0.05). gressively displaced 'losers'. Since the males were

I-3 Night of DE male spawn (N = 184 vocalizations/44 m)

1 Night of AB male spawn (N = 630 vocalizations/44m)

40 30 20 10 0 10

Percent Total Vocalizations on Given Night Fig. 11. Vocal activity was elevated during courtship but both males became relatively quiet when spawning began. Bars represent the number of each type of vocalization as a percent of the total number of vocalizations recorded in 44 rain. Half of each sample (22 rain) was taken during courtship, approximately 4 h before spawning began, and the other half during spawning (the calls of both males have been combined). Data were taken on one night when the DE male spawned (white bars: total number of vocalizations = 630) and on another night when the AB male spawned (black bars: total number of vocalizations = 184). Note that the courtship phase usually occurred early in the night when there was a peak in vocal activity (Fig. 7), even when spawning did not occur. This doubtless contributed to the trend shown here, but the change in vocal activity seen with the onset of spawning was much greater than that seen on non-spawning nights J.D. Crawford et al. : Acoustic communication in an electric fish 305

an opaque barrier (e.g., one wall of a plastic tube, see Fig. 1), each fish would lunge, in alternation, toward the barrier as if to approach the fish on the opposite side. Each lunge was accompanied by a pop: sequences of 2 to 6 lunge/pops were heard often and such exchanges commonly termin- ated with one or two hoots. A lunge typically be- gan with a short dart forward and terminated with a rapid turn. Although the interposed barrier elimi- G nated visual contact between the fishes, they re- mained in acoustic and electric continuity during these exchanges. These interactions occurred at the male-female territory border and were associated with an increase in EOD rate.

3. Nesting, courtship and spawning. Both males D built nests on their territories and vocalized during the nights preceding spawning. On a given spawn- ing night, however, the large female deposited her eggs with only one of the males. The average peri- od of the spawning cycle was 40 days for the AB male and 32 days for the DE male. The large fe- male spawned, on average, every 18 days (Fig. 8). F Early in a male's breeding cycle he actively pa- trolled his territory during the evening, and vigor- ously excluded all females. As the cycle progressed, the frequency of grunts, moans and growls in- creased. In the middle of the cycle, he began build- G ing nests from vegetation and filter floss within spaces provided by tubing and cement structures. Fig. 12A-G. Courtship and spawning behavior in Pollimyrus Small pieces of nesting material were carried to isidori. Before spawning, there are periods of courtship lasting as long as 4 h (A-E). A Female swims into the male's territory, the nest site in the male's mouth and forced into the male approaches her from behind, and they engage in head- confined spaces. Nests were approximately ovoid to-tail circling behavior. B and C The female positions herself and about 3x3x6cm. Nests from previous at an oblique angle to the vertical while the male moves under spawns were often destroyed and males commonly her, ventral-side-up ; at this point, the longitudinal axes of their built 2 new nests (particularly male DE) although heads are perpendicular. D The pair becomes coupled vent-to- vent and begins rotating while quivering. E After a complete only one was used. Nests were always elevated rotation, the pair separates and the female swims out of the from the bottom by about 5 to 10 cm. territory. When spawning begins, the pair engages only briefly The male's vocalizing rate steadily increased as in head-to-tail circling as in A, the female positions herself at the spawning date neared; males usually called an oblique angle again but the male swims along her ventral from a particular site within their territories. Data side in a head-to-head parallel fashion, illustrated in F. The male stimulates the female's cloaca, and the female deposits on vocalizing as a function of position within the eggs. G The male gathers the eggs into his mouth and transports territory were taken on several occasions for each them to his nest male. The total observation time for males was subdivided into time spent in different sections of on stable non-adjacent territories, they interacted the territory. Using data on male position at time minimally; males frequently displaced females. The of vocalizing, it was then possible to compute rates smallest female lost every encounter with the two of sound production in different sections. In two other females; the large breeding female always 0.5 h samples taken from each of two males, the displaced the smaller females. The dominance rela- rate of vocalizing in the preferred area was about tionship amongst the females suggested that rank twice that in other regions. Thus, each male tended was determined by size. to vocalize more from one location than from We frequently recorded pops and hoots during others but did not vocalize exclusively from one agonistic interactions between the large female and area. Over the course of the study, preferred voca- the territorial males. While on opposite sides of lizing areas varied from one breeding cycle to the 306 J.D, Crawford et al.: Acoustic communication in an electric fish

EODs

r T ~"m" ~ .r t d41~NIt, ...... T !' "~ll~ r~ ~ -- ~ -- ~

SOUND

; i i ~ I II .... i

Growl /@tun ' \ Grotwl Growlt of ts Growl Growl Moan distant fish Fig. 13. The electric organ is suppressed during acoustic signalling. Top trace shows EODs of a male fish while bottom trace shows the simultaneous vocalizations of the same male. During this 11.75 s sample, there were 3 episodes of calling, and the number of coincident EODs is much less than would be expected if the activity of the electric organ was independent of the sonic organ's activity. EODs underlined show area of overlap with growls; note the reduced pulse rate of the EODs. EOD suppression is strongest during grunts and moans

next as did nest position. Preferred vocalizing areas (Fig. 12C): the female positioned herself at an ob- were not on the nest site but were generally within lique angle to the vertical and the male rolled under 25 cm of the nest. Males did not preferentially vo- her so that their ventral surfaces were apposed. calize at the borders of their territories. In vent-to-vent coupling, the fish appeared to be As the male increased his vocalization rate, he tightly joined; the couple rotated while quivering also became increasingly tolerant of the large fe- (Fig. 12D). The pair then separated (Fig. 12E), male; she began to visit his territory more often. and the female swam from the territory. This se- Upon entering the territory, she circled above the quence of events lasted about 10 s and was re- male several times and left within 10 s of entering. peated approximately every 15 min. Occasionally The male often followed her through his territory the female left the territory without engaging the as she retreated. Such visits were tightly correlated male. The male vocalized between interactions, with heightened male vocal activity. The male typi- when the female was off his territory, but became cally responded to the large female with a long quiet when she returned for another courtship ritu- sequence of grunts and moans (a grunt-moan bout) al. while she was on his territory and then produced When spawning began, male vocalizations a long growl when she swam off the territory ceased and interactions on the territory became (Fig. 9). simplified. Head-to-tail circling, vent-to-vent cou- The durations of grunt-moan bouts (Fig. 9) pling, quivering were bypassed. The female contin- averaged 2 to 3 times longer when the large female ued to enter as before and position herself at an was on a male's territory compared to when a non- oblique angle. The male moved into position gravid female or no female was present (Fig. 10). against her ventral surface (Fig. 12F). While posi- This difference was highly significant (Student's t- tioned over the male, the female released a group test, P<0.0001). of 2 to 20 eggs. The female left the territory, the Acoustic signalling reached a peak during the male picked up the eggs in his mouth (Fig. 12G) hours preceding spawning, but when spawning be- and deposited them in his nearby nest. Each gan, there was a precipitous drop in the vocal activ- spawning event, from female entry to departure, ity of both male fish in the aquarium (Fig. 11), lasted only 10-15 s. Spawns occurred for a period though only one male spawned on a given night of about 6 hours, at intervals of ] to several min. (Fig. 8). Spawning was preceded by a period of On two occasions, milky white clouds (presumably courtship that lasted as long as 4 hours. sperm released by the male) were observed in the During courtship, the large female repeatedly area where the eggs had been deposited. entered the male's territory and the two engaged After spawning, the male patrolled his territory in a courtship ritual (Fig. 12A-E). The female excluding all other fish including his mate. He swam into the territory above the male (about spent some time close to his nest, gently nudging 15 cm) and the male then approached her from the nest with his nose and occasionally fanning behind. The two engaged in head-to-tail circling the nest with his caudal fin. By 24 h, development (Fig. 12A). From head-to-tail circling, a transition of the eggs was clearly visible and at 48 h larvae occurred (Fig. 12B) to vent-to-vent coupling began wiggling within the nest. After 8-10 days, J.D. Crawford et al. : Acoustic communication in an electric fish 307

18- kins 1986) and now we can add vocalizations to O the mormyrid repertoire of communication signals. LU 15- CO average rate The acoustic repertoire of Pollimyrus isidori in- 12 cludes 5 sounds that can be separated into two O LU 9- AB MALE categories according to behavioral context: 1) 6- NDEMALE three sounds produced by males in the context of courtship and 2) two sounds produced during 3- agonistic encounters. Although acoustic signaling ~ Grunt Moan Growl Silence is seen in many groups of teleosts (Fine et al. 1977; I t Myrberg 1981; Hawkins and Myrberg 1983), to ACOUSTIC BEHAVIORS our knowledge, this is the first report of complex Fig. 14. Electric organ discharge rates during differentacoustic acoustic signaling in mormyrids. behaviors. A bi-modal sample was analyzedfor each male. The overall average rate of EOD was the same for the two males (12 EODs/s) and is plotted as a horizontal line for comparison with rates shown by bars. For both males the rate of EODs A. Mechanisms of acoustic signal generation was higher than average during acousticallysilent periods and and reception much lower than average during each of the three vocalizations. Note that male DE produced no EODs during grunts or moans. Characteristics of the grunt, moan, and growl sug- A sample 8.76 min in duration was analyzed for male AB, gest a common mechanism for these three vocaliza- including 37 grunts, 37 moans and 48 growls. A sample 2.72 tions. These vocalizations are often produced to- min in duration was analyzedfor male DE, including 14 grunts, gether during a single vocalizing episode and we 11 moans and 27 growls suggest that they may be produced by a single structure driven by three distinct motor com- young fish began to disperse from the nest (see mands. The moan is spectrally quite pure and the also Kirschbaum and Westby 1975; Kirschbaum waveform is quasi sinusoidal with two main spec- 1984). tral peaks. Moans often grade into, and sometimes out of, growls. The sound spectrograms of these D. Motor correlates of vocalizations vocalizations are similar (Figs. 2 and 3). Examina- 1. Locomotory behaviors and vocalizing. Grunts, tion of waveforms suggests that the moan is gener- moans and growls were all produced during active ated by a continuous train of muscle contractions swimming or hovering. None of these vocalizations at 220 Hz whereas the growl is generated by short was closely associated with any particularly overt bursts of 220 Hz contractions separated by 40 ms. motor pattern. The pop was the only sound that The grunt could be produced by contractions at was tightly linked to overt motor patterns: aggres- 50 Hz acting on the same sonic structure but in sive chases and lunges. a highly damped condition. It seems likely that, as in many other teleosts, 2. Electric organ discharge and acoustic vocalizing. the mormyrid swim bladder is involved in sound Coincident with the onset of breeding under our production. Orts (1967) has examined the anatomy rain-regime, we noted a sex difference in the EODs of the swim bladders of a number of mormyrid of PolIimyrus isidori, confirming Westby and species; it occupies 45 to 50% of the volume of Kirschbaum (1982). The EOD was suppressed the body cavity and a thin sheet of muscle fits when males produced grunts, moans and growls. around its ventral half. The swim bladder of PolIi- Figure 13 illustrates the suppression of electric dis- myrus isidori might be driven into resonance at charges during vocalization. Data from longer re- 220 Hz to produce the moan. Data from other cords are presented graphically for each male swim bladder sonic systems show, however, that (Fig. 14). Both males produced an average of 12 teleost swimbladders can be driven continuously EOD's per s, but while vocalizing the rate was at frequencies as high as 220 Hz by sonic muscle much lower than this. Inhibition was not as strong to produce a moan-like vocalization without reso- during the growl as it was during the other two nance (see Tower 1908; Tavolga 1964; Cohen and calls. During silent periods, EODs were produced Winn 1967; Schneider 1967; Tavolga 1971). at a higher rate of about 17 EOD's per s. It is likely that the inner-ear air bladder-cou- pled saccular system discussed by von Frisch Discussion (1938b) and Stipetic (1939) is critical in reception of the courtship sounds of Pollimyrus isidori (see The electric discharges of mormyrids are well also Werns and Howland 1976). Recent behavioral known to subserve social communication (Hop- audiograms for the mormyrid Gnathonemus peter- 308 J.D. Crawford et al. : Acoustic communicationin an electricfish sii show best sensitivity in the 0.3 to 1.0 kHz range Brenowitz 1981) in these two modalities is under- (McCormick and Popper 1984). The match be- way. tween these data and the acoustic repertoire of PoI- We suspect that the acoustic signals of territori- limyrus isidori is good: all of the sounds for which al males provide distant conspecifics with informa- we were able to make SPL estimates should be tion about the presence of and/or location of terri- at least 30 dB above threshold (at close range) ac- tories and potential breeding sites. In this sense, cording to the tuning curves of McCormick and we think it is likely that these signals serve as ad- Popper (p. 756: McCormick and Popper 1984). vertisement calls. Play-back experiments point to this sort of function for male vocalization in a B. Relationship between electric number of other teleost fishes (Tavolga 1956; Ger- and acoustic signaling ald 1971 ; Horch and Salmon 1973; Myrberg 1972). The acoustic signals of Pollimyrus isidori may The only courtship-specific electric behavior we also play a role in individual recognition. We have observed in PoIlimyrus isidori was the cessation of seen clear differences in temporal parameters of the electric organ during vocalization (Figs. 13 and the calls of three male Pollimyrus isidori, and these 14). This contrasts with the wealth of electric sig- could be used in mate selection by females and/or nals observed in other species of electric fish (Hop- in inter-male communication. Acoustically-me- kins 1986; Hopkins and Bass 1981; Hagedorn and diated inter-male recognition is known in bicolor Heiligenberg 1985). The significance of the inhibi- damselfish (Pomacentrus partitus), although this is tion of EODs during acoustic signals is not yet probably based on spectral cues (Myrberg and clear, but suggests interaction at the level of CNS Riggio 1985). Temporal parameters are, however, command nuclei for electric and acoustic motor important in species recognition in damselfish output; the control of sonic and electromotor or- (Myrberg et al. 1978; Spanier 1979) and could sub- gans resides in the caudal medulla (Bell et al. 1983). serve intraspecific individual recognition in Polli- High-frequency bursts of EODs are produced by myrus isidori. The variability in pulse repetition several species of electric fish and apparently func- rate within the grunts of our individual males was tion as courtship displays (Hopkins and Bass 1981 ; remarkably small, the standard deviation being Hagedorn 1986). We have not observed bursts in only 3 % of the mean in each case. courting Pollimyrus isidori. Pollimyrus isidori do, Two of our observations indicate that females however, exhibit more subtle electrical behaviors are the targets of male acoustic signals. First, males such as the preferred latency avoidance or echo dramatically increase the duration of their grunt- responses (Lfiker and Kramer 1981), which may moan bouts when a gravid female is nearby be important in courtship. Our analysis of the be- (Figs. 9 and 10) and second, males stop vocalizing havior of free-swimming fish has not allowed us when they have successfully courted a female to to examine this sort of electrical behavior. the point of spawning. A related observation is that even the unsuccessful male stops vocalizing C. Mating behavior and function of acoustic signals when spawning is initiated, suggesting that he is Acoustic signals may allow mormyrids to commu- attending to a signal emitted by a member of the nicate over longer distances than is possible in the spawning pair. The nature of this signal is not yet electric modality. Although estimates of the range known. of electric communication have not yet been made Of the two sounds heard during aggression, the in nature, under free-field conditions, the magni- pop was most closely correlated with lunges and tude of an electric field decays with the cube of chases. 'Pops' and 'knocking' sounds have also distance from a dipole source. This rapid rate of been reported for other teleosts in aggressive inter- signal attenuation may lead to an effectively short actions (Dijkgraaf 1947; Myrberg 1972, 1981). It communication range. In the best case, the acous- seems possible, as others have suggested (Myrberg tic signal (SPL) will decay with the inverse first 1981; Schwarz 1974; Lanzig 1974), that these power of distance (under free-field conditions). sounds might enhance other aspects of aggressive Acoustic signals may thus be detectable over displays. In the male-female interactions we ob- greater distances, depending on a variety of condi- served in Pollimyrus isidori, however, pops were tions, including water depth, the nature of the fish often heard while the individuals were separated as a source (monoplar or dipolar; see Kalmijn by a visually opaque plastic barrier. Possibly the 1986), ambient noise and the parameter of the sig- pop augments an electrical display. nal to which receivers are sensitive. A comparative Male Pollimyrus isidori became highly territori- analysis of active space (Wilson and Bossert 1963; al during breeding. They invested a substantial J.D. Crawford et al. : Acoustic communication in an electric fish 309 amount of time in territorial defense and nest Birkholz J (1969) Zuf~illige Nachzucht bei Petrocephalus bovei. building prior to spawning and continued to de- Das Aquarium 3:201-203 fend the territory and tend nests after eggs were Birkholz J (1970) Nachwuchs bei Petrocephalus bovei. Das Aquarium 4: 340-342 deposited. Female parental investment was limited Brenowitz EA (1982) The active space of red-winged blackbird to the production and deposition of ova. Under song. J Comp Physiol 147:511-522 such conditions in which males make a relatively Cohen MJ, Winn HE (1967) Electrophysiological observations heavy investment in nesting and parental care, fe- on hearing and sound production in the fish, Poriehthys notatus. J Exp Zool 165:355 370 male competition for males is likely (see Emlen Crawford JD, Hagedorn MM, Hopkins CD (1985) Acoustic and Oring 1977; Li and Owings 1978a). In our song in an electric fish. Neurosci Abstr 11 : 270 laboratory system, there was a great deal of inter- Cuvier G, Valenciennes A (1946) Histoire naturelle des pois- female aggression and all of the eggs laid (in both sons. 19:214-286 males nests) were eggs of a single female. This sug- Dijkgraaf S (1947) Ein TSne erzeugender Fisch im Neapler Aquarium. Experientia 3:493-494 gests that females were competing for males but Emlen ST, Oring LW (1977) Ecology, sexual selection and the this should be confirmed in the field. evolution of mating systems. Science 197:215-223 Fine M, Winn HE, Olla BL (1977) Communication in fishes. D. Conclusion In: Sebeok TA (ed) How animals communicate. Indiana University Press, Bloomington, pp 472-518 The courtship sounds of male Pollimyrus isidori Fernald R (1975) Fast body turns in a cichlid fish. Nature are among the most elaborate acoustic signals we 258 : 228-229 Ficken RW, Tienhoven A van, Ficken MS, Sibley FC (1960) know of for fish. It is interesting that a fish with Effect of visual and vocal stimuli on breeding in the budgeri- a sophisticated electrosensory system has also gar (Melopsittacus undulatus). Animal Behav 8:104-106 evolved a relatively complex acoustic repertoire. Frisch K von (19381) f.)ber die Bedeutung des Sacculus und Considering the physics of signal transmission in der Lagena ffir den Geh6rsinn der Fische. Z Vergl Physiot the two modalities, acoustic signalling may extend 25 : 703-747 Frisch K von (1938b) The sense of hearing in fish. Nature the mormyrid's communication range or active 141 : 8-11 space. Gerald JW (1971) Sound production during courtship in six species of sun fish (Centrarchidae). Evolution 25 : 75-87 Acknowledgements. We would like to thank S. Arnesen, B. Hagedorn M (1986) The ecology, courtship and mating of gym- Baldwin, R. Capranica, J. Conner, S. Coombs, R. Fay, C. notiform electric fish. In : Bullock TH, Heiligenberg W (eds) McCormick and A. Myrberg for commenting on drafts of the Eleetroreception. John Wiley & Sons, New York, pp manuscript. R. Capranica, C. Clark, R. Hoy, S. Volman and 497-525 D. Yager were generous with their equipment and technical Hagedorn M, Heiligenberg W (1985) Court and spark: Electric expertise. J Gittleman, D. Maddox, C. McCulloch and R. Rob- signals in the courtship and mating of gymnotoid fish. Ani- bins provided statistical consultation. P. Bensadouin contrib- mal Behav 33 : 254-265 uted the biological drawings and B. Baldwin drafted many of Hawkins AD, Maclennan DN (1976) An acoustic tank for hear- the figures. We are indebted to B. Baldwin, A. Collazo, C. ing studies on fish. In: Schuijf A, Hawkins AD (eds) Sound Marler, and B. McFall for maintenance of the fish and the reception in fish. Elsevier, Amsterdam, pp 149-170 success of our breeding project. We thank S. Mancil and T. Hawkins AD, Myrberg AA (1983) Hearing and sound commu- Natoli for typing and editorial work. This research has been nication under water. In: Lewis B (ed) Bioacoustics. Aca- supported by NIMH grant MH37972 to CDH; NIMH training demic Press, New York, pp 347-406 grant MH15793 to JDC; NIMH post-doctoral fellowship 5F32 Heusinger CF von (1826) Bemerkungen fiber das Geh6rwerk- NSO7381 to MMH. zeug des Mormyrus eyprinoides, Gastroblecus compressus und Pimelodus synodontis. Arch Anat Physiol Meckel 1 : 324-327 References Hirata NR, Fernald RD (1975) Non-intentionalsound produc- tion in cichlid fish (Haplochromis burtoni, Gunther). Exper- Bass AH, Hopkins CD (1983) Hormonal control of sexual dif- ientia 31:299-300 ferentiation: changes in electric organ discharge waveform. Hopkins CD (1986) Behavior of Mormyridae. In: Bullock TH, Science 220:971-974 Heitigenberg W (eds) Eleetroreception. John Wiley & Sons, Bass AH, Hopkins CD (1985) Hormonal control of sex differ- New York, pp 527 576 ences in the electric organ discharge (EOD) of mormyrid Hopkins CD, Bass AH (1981) Temporal coding of species rec- fishes. J Comp Physiol A 156:582604 ognition signals in an electric fish. Science 212:85-87 Bell CC (1986) Electroreception in mormyrid fish. Central Horch K, Salmon M (1973) Adaptations to the acoustic envi- Physiology. In: Bullock TH, Heiligenberg W (eds) Electro- ronment by the squirrelfishes, Myripristis violaceus and M. reception. John Wiley & Sons, New York, pp 423-452 pralinius. Mar Behav Physiol 2:121-139 Bell CC, Szabo T (1986) Electroreception in mormyrid fish. Iles RB (1960) External sexual differences and their significance Central Anatomy. In: Bullock TH, Heiligenberg W (eds) in Mormyrus kannume Forskal 1775. Nature 188:516 Electroreception. John Wiley & Sons, New York, pp 315- Kalmijn A (1986) Hydrodynamic and acoustic detection in 421 aquatic vertebrates. In: Atema J, Fay R, Popper AN, Ta- Bell CC, Libouban S, Szabo T (1983) Pathways of the electric volga WN (eds) Proc Conference Mote Marine Laboratory, organ discharge command and its corollary discharges in Springer, Berlin Heidelberg New York (in press) mormyrid fish. J Comp Neurol 216:327-338 Kirschbaum F (1975) Environmentalfactors control the period- 310 J.D. Crawford et al. : Acoustic communication in an electric fish

ical reproduction of tropical electric fish. Experientia Myrberg AA, Spanier E, Ha S (1978) Temporal patterning in 31:1159-1160 acoustical communication. In: Reese ES, Lighter FJ (eds) Kirschbaum F (1979) Reproduction of the weakly electric fish Contrast in behavior. John Wiley, New York, pp 137 179 Eigenmannia vireseens (Rhamphichthyidae, Teleostei) in Orts S (1967) Contribution/t l'anatomie compar6e et ~i la syst~- captivity. I. Control of gonadal recrudescence and regres- matique des Mormyroides. M~m Acad R Sci Outre-mer sion by environmental factors. Behav Ecol Sociobiol (Brussels) (C1 Sci Nat Med 8~ (NS) 17:1-90 4:331-355 Parvulescu A (1966) The acoustics of small tanks. In: Tavolga Kirschbaum F (1982) Reproduction of the mormyrid Pollimyr- WN (ed) Marine bioacoustics, vol 2. Pergamon Press, Lon- us isidori in captivity. Fourth Congress of European Ich- don, pp 7-13 thyologists. Hamburg, Abstract 149 Rigley L, Marshall J (1973) Sound production by the elephant- Kirschbaum F (1984) Reproduction of weakly electric teleosts: nose fish, Gnathonemus petersii (Pisces, Mormyridae). Co- just another example of convergent development? Environ peia 1973 : 134~135 Biol Fishes 10:3 14 Schneider H (1967) Morphology and physiology of sound-pro- Kirschbaum F, Westby GWM (1975) Development of the elec- ducing mechanisms in teleost fishes. In: Tavolga WN (ed) tric discharge in mormyrid and gymnotid fish (Mareusenius Marine bio-acoustics. Pergamon Press, Oxford, pp 135- sp. and Eigenmannia virescens). Experientia 31:1290-1293 158 Kramer B, Tautz J, Markl H (1981) The EOD sound response Schwarz A (1974) Sound production and associated behavior of weakly electric fish. J Comp Physiol 143:435-441 in a cichlid fish, Cichlasoma centrarchus. Z Tierpsychol Lanzing WJR (1974) Sound production in the cichlid Tilapia 35:147~156 mossambica Peters. J Fish Biol 6:341-347 Spanier E (1979) Aspects of species recognition by sound in Li SK, Owings DH (1978 a) Sexual selection in the three-spined four species of damselfishes, genus Eupomaeentrus (Pisces: stickelback. I. Normative observations. Z Tierpsychol Pomcentriae). 46: 359-371 Stipetic E (1939) Uber das Geh6rorgan der Mormyriden. Z Li SK, Owings DH (1978 b) Sexual selection in the three-spined Vergl Physiol 26 : 740-752 stickelback. II. Nest raiding during the courtship phase. Tavolga WN (1956) Visual, chemical and sound stimuli as cues Behaviour 64:298-304 in the sex discriminatory behavior of the gobiid fish Bathy- Lficker H, Kramer B (1981) Development of a sex difference gobius soporator. Zoologica 41:49-64 in the preferred latency in the weakly electric fish, Pollimyr- Tavolga WN (1964) Sonic characteristics and mechanisms in us isidori (Cuvier et Valenciennes) (Mormyridae, Teleostei). marine fishes. In: Tavolga WN (ed) Marine bio-acoustics. Behav Ecol Sociobiol 9:103-109 Pergamon Press, Oxford, pp 195-211 McCormick CA, Popper AN (1984) Auditory sensitivity and Tavolga WN (1971) Sound production and detection. In: Hoar psychophysical tuning curves in the elephant nose fish, WS, Randall DJ (eds) Fish physiology, vol 5. Academic Gnathonemus petersii. J Comp Physiol A 155:753-761 Press, New York, pp 135-205 Myrberg AA (1972) Using sound to influence the behavior of Tower RW (1908) The production of sound in the drumfishes, free-ranging marine animals. In: Winn HE, Olla BL (eds) the sea robin and the toadfish. Ann NY Acad Sci Behavior of marine animals, vol. 2, Plenum Press, New 18:149-180 York, pp 435-468 Werns S, Howland HC (1976) Size and allometry of the saccular Myrberg AA (1981) Sound communication and interception air bladder of Gnathonemus petersii (Pisces : Mormyridae) : in fishes. In: Tavolga WN, Popper AN (eds) Hearing and Implications for hearing. Copeia 1976:200-202 sound communication in fishes. Springer, Berlin Heidelberg Westby GW, Kirschbaum F (1982) Sex differences in the wave- New York, pp 395-426 form of the pulse-type electric fish, Pollimyrus isidori (Mor- Myrberg AA, Riggio R (1985) Acoustically mediated individual myridae) J Comp Physiol 145:399-403 recognition by a coral reef fish (Pomacentrus partitus). Ani- Wilson EO, Bossert WH (1963) Chemical communication mal Behav 33:411416 amongst animals. Rec Prog Hormone Res 19:673-716