ORIGINAL PAPER

T. N. Kenyon á F. Ladich á H. Y. Yan

A comparative study of hearing ability in ®shes: the auditory brainstemresponse approach

Accepted: 8 August 1997

Abstract Auditory brainstem response (ABR) tech-niques, an electrophysiological far-®eld recordingmethod widely used in clinical evaluation of humanhearing, were adapted for ®shes to overcome the majorlimitations of traditional behavioral and electrophysio-logical methods (e.g., invasive surgery, lengthy trainingof ®shes, etc.) used for ®sh hearing research. Responsesto clicks and tone bursts of dierent frequencies andamplitudes were recorded with cutaneous electrodes. Toevaluate the eectiveness of this method, the auditorysensitivity of a hearing specialist (gold®sh, Carassiusauratus) and a hearing generalist (oscar, Astronotusocellatus) was investigated and compared to audiogramsobtained through psychophysical methods. The ABRscould be obtained between 100 Hz and 2000 Hz (oscar),and up to 5000 Hz (gold®sh). The ABR audiograms aresimilar to those obtained by behavioral methods in bothspecies. The ABR audiogram of curarized (i.e., Flaxedil-treated) gold®sh did not dier signi®cantly from twopreviously published behavioral curves but was lowerthan that obtained from uncurarized ®sh. In the oscar,ABR audiometry resulted in lower thresholds and alarger bandwidth than observed in behavioral tests.Comparison between methods revealed the advantagesof this technique: rapid evaluation of hearing in un-trained ®shes, and no limitations on repeated testing ofanimals.

Key words Fishes á Audiometry á Auditory brainstemresponse á Evoked potentials á Hearing sensitivity

Introduction

Hearing in ®shes was ®rst demonstrated at the beginningof this century in cyprinids (Parker 1903; Bigelow 1904).Quantitative work on the range of frequencies overwhich ®sh hear and on representatives of several dier-ent families was later carried out by von Frisch and hiscolleagues (see review by von Frisch 1936). Since thennumerous methods have been used to test auditorysensitivity in ®shes which has resulted in audiograms ofmore than 50 species (see review by Fay 1988). Thesedata reveal an enormous diversity of hearing abilities in®shes, mostly owing to various peripheral modes ofcoupling the ear to some internal structures. Thesestructures include the swimbladder (with or withoutspecial hearing-related modi®cations such as diverticulaeand Weberian ossicles), and air-®lled bullae (Hawkinsand Myrberg 1983; Hawkins 1993). However, in the fewspecies that have been tested repeatedly using dierentmethods, large intraspeci®c variations in auditorythresholds appeared which may be due to dierences intesting conditions (Popper et al. 1973; Hawkins 1981) orsimply cannot be explained (Popper and Fay 1973). Todate, no standardized method has been establishedwhich is suciently convenient to apply to a largenumber of species in a short time. Consequently, thegeneral auditory capabilities of less than 0.2% of ®shspecies (Nelson 1984; Fay 1988) are known so far.

In the past, both behavioral and electrophysiologicalmethods have been used to investigate ®sh audition.Behavioral methods usually involve training ®sh byusing electric shock or food rewards to respond uponhearing a sound. In the classical conditioning procedure,®sh respond with innate behavior such as stereotypeddefense responses (Myrberg and Spires 1980; Kenyon1996), cardiac suppression (Chapman and Sand 1974;Hawkins and Johnstone 1978; Jerkù et al. 1989; Kojimaet al. 1992) or ventilatory suppression (e.g., Banner1967; Popper et al. 1973; Fay 1995). The classical con-ditioning procedure has so far generated the majority of

J Comp Physiol A (1998) 182: 307±318 Ó Springer-Verlag 1998

T.N.Kenyon áH.Y. Yan (&)T.H. Morgan School of Biological Sciences,University of Kentucky, Lexington,KY 40506-0225, USAFax: + 1-606 257-7410; e-mail: hyyan00@pop.uky.edu

F. LadichUniversity of Vienna, Institute of Zoology,Althanstrasse 14, A-1090 Wien, Austria

®sh behavioral audiogram data (see Fay 1988). How-ever, this conditioning method is potentially stressfuldue to electric shock and is not applicable to thosespecies that cannot be conditioned to shock ± for ex-ample, a shark (Nelson 1967), and some cichlid species(Tavolga 1974; Allen and Fernald 1985).

A second behavioral method is instrumental avoid-ance conditioning in which ®sh learn to cross a barrierin the tank upon hearing a sound to avoid electric shock(e.g., Behrend and Bitterman 1962; Tavolga andWodinsky 1963; Popper 1971). The advantage of thismethod is that the response is unambiguous. This par-adigm, however, makes precise calibration of soundperceived by ®sh dicult and may require excessivelylong training times in some species, e.g., 20±30 days forthe cichlid Tilapia macrocephala (Tavolga 1974), and 25or more days for the cat®sh Ictalurus nebulosus (Weisset al. 1969). Operant conditioning methods involvingpositive reinforcement have been infrequently applied tothe study of acoustic sensitivity in ®shes. A paradigmwhere subjects were trained to peck paddles in responseto sound was used successfully in measuring audiogramsin gold®sh (Yan and Popper 1991; Yan 1995), the oscarAstronotus ocellatus (Yan and Popper 1992; Yan 1995)and even used for intensity discrimination study (Yanand Popper 1993). Despite the advantages of using foodas reward, free movement of subjects and precise cali-bration of sound pressure levels perceived by the sub-ject, this method also suers some drawbacks includinglong duration of conditioning, (gold®sh 1±2 days, oscar10±14 day), and high degrees of variation among indi-viduals of a species. In addition, the ®sh must also belarge and responsive enough to perform the paddlepecking task. As with any operant method, the condi-tioned response must closely approximate a part of theanimal's natural behavioral repertoires. Therefore, for®sh not using striking mode for food (prey) capture, thisoperant conditioning method may not work well (Yan1995).

Electrophysiological methods do not have the limi-tations associated with training subjects. One widelyused technique is the measurement of microphonics (a.c.currents) from auditory end organs while presentingacoustic stimuli to the tested subjects (Enger and An-derson 1967; Furukawa et al. 1972; Fay and Popper1974; Sand 1974; Saidel and Popper 1987). Data canpotentially be obtained more quickly than in behavioralmethods, but preparation can often be complex andgenerally require invasive surgery, precluding use of thesame animal for repeated tests later. Furthermore,placement of electrodes is restricted to speci®c endor-gans and thus responses recorded do not necessarilyrepresent the whole auditory pathway including hearingend organs, eighth nerves, auditory brainstem and themidbrain. Even single-unit recordings where single nerve®ber discharge patterns are measured (e.g., Enger andAndersen 1967) experience similar limitations.

A third electrophysiological method widely used inaudition studies of many mammalian species, is the

auditory brainstem response (ABR). The ABR is a non-invasive far-®eld recording of synchronous neural ac-tivity in the eighth nerve and brainstem auditory nucleielicited by acoustic stimuli (Jewett 1970; Jewett andWilliston 1971; Jacobson 1985). ABR recording hasproven invaluable in clinical evaluations of humanhearing (Jacobson 1985; Hall 1992) and allows thresholddeterminations from uncooperative or inattentive sub-jects where behavioral methods generally fail. However,the ABR techniques have been applied to only a few ®shspecies. Corwin et al. (1982) showed that ABRs can beobtained from a variety of ®sh species, when using in-tracranial as well as cutaneous electrodes. Waveformsclearly indicated a similarity between ®sh and highervertebrates (Corwin 1981). Bullock (1981) also espousedthe great potential of ABR recording for the study of®sh audition and its advantages over conventionalmethods, including rapid whole-animal measurementswithout time-consuming training, and the potential forrepeated use of the same animals. Nevertheless thismethod has not been used to study the auditory capa-bilities of ®shes.

The aim of the present paper is to demonstrate thatABR techniques can be used to study hearing in ®shes,and provide an ecient method for obtaining audio-grams. Comparison of ABR audiograms with thoseobtained by psychophysical methods clearly demon-strate the advantages and the eectiveness of thismethod, and the great potential for its use in futureinvestigations of ®sh audition.

Materials and methods

Experimental animals

Eight specimens each of Carassius auratus [59±78 mm standardlength (SL); 6.1±18.2 g] and A. ocellatus (62.1±81 mm SL, 8.6±18.9 g) were obtained from a local aquarium ®sh supplier. Animalswere maintained in ®ltered aquaria at 251 °C, and were fedArtemia and commercially prepared foods (TetraMin, TetraWerke,Germany). The animal-use protocol used in this study wasapproved by the University of Kentucky IACUC (93005L).

ABR recording setup

Test subjects were secured inside a rectangular 15-l plastic tub(38 cm ´ 24.5 cm ´ 14.5 cm) ®lled with water. Fine-mesh nylonscreen was wrapped around the animal and held in a metal clampattached to a glass rod that was ®xed in a micromanipulator(M3301, World Precision Instruments, Sarasota, Fla., USA).Respiration was achieved through a temperature-controlled(25 1 °C) gravity-feed aerated water system to keep ®sh aliveduring recording. In order to reduce myogenic noise levels, animalswere temporarily immobilized with an injection of gallaminetriethiodide (Flaxedil; Sigma G-8134, St. Louis, Mo., USA). Dos-ages were adjusted so that animals were generally still capable ofslight opercular movements, but unable to initiate gross musclecontractions. The dosage required to obtain this level of immobilitywas species dependent, varying from 2 to 4 lg g)1 body weight forgold®sh and 6±10 lg g)1 body weight for oscars. Tests were alsorun on animals without Flaxedil, to determine if it had an eect onthreshold determinations.

308

The test animal position was adjusted so that the nape of thehead was just 1 mm above the surface of the water, and a respi-ration pipette was inserted into the subject's mouth (Fig. 1). Theplastic tub rested on a vibration-free air table (Kinetic Systemsmodel 1201), and the entire setup was enclosed in a walk-in sound-proof room (2 m ´ 3 m ´ 2 m, Industrial Acoustics).

A small piece of Kimwipes tissue paper (10 mm ´ 2 mm) wasplaced on the exposed top of the head region to prevent skin fromdrying. The recording electrode was placed on the midline of theskull over the medulla region. The reference electrode was placed5 mm anterior to the recording electrode. Both electrodes werepressed ®rmly against the skin through the tissue paper. The elec-trodes consisted of Te¯on-insulated silver wire (0.25 mm diameter)with ca. 1 mm of exposed tip. Wires were ®xed with epoxy and werehoused inside plastic pipettes, and clamped in micromanipulators.Shielded electrode leads (ca. 40 cm in length) were twisted togetherto reduce potential noise, and attached to the dierential inputs ofan a.c. preampli®er (Grass P-15, 40 dB gain, high-pass at 30 Hz,low-pass at 3000 Hz). The ground terminal of the preampli®er wasconnected via a wire to the water in the tub. A hydrophone (Cel-esco LC-10), placed adjacent to the exterior of presumed inner earregion of the test subject, was used to monitor stimulus soundpressure. A second Grass P-15 preampli®er (40 dB gain, high-passat 10 Hz, low-pass at 10 kHz) was used to amplify the hydrophoneoutput. Speakers, suspended in air, were mounted 1 m above thetest subject. For frequencies less than 3000 Hz, a 30-cm ``woofer''(Pioneer; frequency response 19±5 kHz) was used, while for higherfrequencies (>3000 Hz) a 12-cm midrange speaker (Pyle MR 516;frequency response 500±11 kHz) was used. Output terminals of thepreampli®ers were hooked to shielded leads that passed through aport in the wall of the chamber. Speaker leads also passed throughthis port.

ABR recording apparatus and stimulus presentation

Both sound stimuli presentation and ABR waveform recordingwere accomplished via a Tucker-Davis Technologies (Gainesville,Fla., USA) modular rack-mount system controlled by an opticalcable-linked 66-MHz 486 PC containing a TDT AP2 Digital SignalProcess board and running TDT ``Bio-Sig'' software. Sound stimuliwaveforms were generated using TDT ``Sig-Gen'' software, and fedthrough a DA1 digital-analog converter, a PA4 programmableattenuator, and a power ampli®er (QSC Audio Products, ModelUSA 370) which drove the speaker. The hydrophone preampli®eroutput cable was fed to one channel of an AD1 analog-digitalconverter, while the electrode preampli®er output was ®rst passedthrough a PC1 spike conditioner (which provided an additional60 dB gain and 3000 Hz low-pass ®lter) before reaching the AD1.Both preampli®er outputs were also fed to an oscilloscope for real-time assessment of electrode noise levels. A TG6 timing generatorwas used to synchronize A/D and D/A conversion (see Fig. 1 fordetails).

Both tone bursts and clicks were presented to test subjects.Clicks were 0.1 ms in duration, and presented at a rate of 38.2 s)1

Fig. 1 Schematic diagram of the ABR-recording setup. (AD1 analogto digital converter; DA1 digital to analog converter; DSP digitalsignal processor; mic microphone; MA1 microphone ampli®er; MS1monitor speaker with ampli®er; PA4 programmable attenuator; PC1spike pre-conditioner; power amp power ampli®er for speaker; preampGrass P-15 preampli®er; sump sump pump; TG6 six-channel timinggenerator)

309

(to prevent phase locking with any 60-Hz noise). Clicks exhibited anearly ¯at power spectrum between 1 and 4000 Hz (Fig. 2). Thenumber of cycles in a tone burst was adjusted according to fre-quency in order to get the best compromise between stimulus ra-pidity (i.e., greater rapidity of onset greater ecacy at generatingABRs) and peak-frequency bandwidth (longer duration sharperspectral peak; Silman and Silverman 1991). All bursts were gatedusing a Blackman window to reduce spectral ``side lobes'', and toprovide a ramped onset/decay, thereby preventing speaker tran-sients (Silman and Silverman 1991). See Fig. 2 for spectra of broad-band clicks and tone bursts of dierent frequencies. The ABRtraces of opposing polarities (one thousand sweeps each) were av-eraged together forming a 2000-sweep trace to eliminate anystimulus artifact. At each tested frequency, this was done twice andoverlaid to examine if traces were repeatable. The lowest soundpressure level where a repeatable ABR trace could be obtained, asdetermined by overlaying replicate traces, was considered thethreshold. Sound pressure level was attenuated in 5-dB steps untilrecognizable and repeatable ABR waveforms could no longer beproduced. Smaller steps of 3 dB were used at frequencies above2000 Hz, as ®ner threshold resolution was needed at the high soundlevels required at these frequencies. This method of visual inspec-tion/correlation is the traditional means of determining thresholdin ABR audiometry (Jacobson 1985; Kileny and Shea 1986; Gorgaet al. 1988; Hall 1992; Song and Schacht 1996). Some ABR systemsemploy an automatic threshold-seeking algorithm (OÈ zdamar et al.1994) with accuracy similar to that of human examiners, but thisoption was not available in TDT BioSig system. Once the thresholdlevel was determined, the hydrophone recording was analyzed todetermine root mean square (RMS) sound pressure level, based onthe method of Burkard (1984). Using the capabilities of the Bio-Sigsoftware, cursors were placed one cycle apart on either side of thelargest (i.e., center) sinusoid of a particular tone-burst recordingfrom the hydrophone. The software then calculated the RMS of thewaveform between the cursors, and calibration factors were appliedto determine actual sound pressure level in dB re 1 lPa. No onsettransients were evident in any of the hydrophone traces. Animalswere tested at frequencies of 100 Hz (the lowest frequency where anABR-like waveform was obtained), 200, 300, 400, 500, 600, 800,1000, 1200 (oscar), 1500, 2000, 3000, 4000, and 5000 Hz (thehighest frequency where consistent results were obtained under theacoustic conditions present in the chamber). Animals typically

regained full mobility within 12 h of completion of testing afterFlaxedil was metabolized.

Ambient noise levels in the test tank were also measured usingthe hydrophone and preampli®er. Samples of full-spectrum am-bient noise were recorded into the TDT system and ®ltered withdigital band-pass ®lters. Filtered noise RMS levels were measuredusing the BioSig software, and spectrum levels were calculated byapplying appropriate ®lter corrections and calibration factors.

Data analysis

Threshold values from all individuals were averaged to produceaudiograms for each species. Mean threshold values of entire au-diograms were compared using one-way ANOVA. Flaxedil-treated®sh were compared with animals without Flaxedil treatment andwith previously published behavioral data.

Results

ABR waveform characteristics

ABR waveforms were obtained in all animals tested, andshowed almost identical characteristics at suprathresh-old levels in both species. The ABR traces of signalspresented at dierent polarities do not cancel out eachother when averaged. This is contrary to sound pressurewaveforms when averaged (Fig. 3). A typical supra-threshold ABR consisted of a series of four to nine rapiddownward peaks superimposed over a slow negativede¯ection lasting approximately 8 ms at low frequencies,to around 2.5 ms in response to clicks and high-fre-quency tone bursts (Fig. 4). The series of rapid peakswas generally followed by a slow positive de¯ection. Theonset latency of the ABR varied with frequency, rangingfrom 7.5 ms after stimulus onset at 100 Hz to as little as0.3 ms with clicks and 5000-Hz tone bursts. Onset la-

Fig. 2 Power spectra of abroad-band click and twotone bursts (500 Hz and1000 Hz) of acoustic signalspresented to tested subjectsas recorded from hydro-phone. dB in relative scale

310

tency also increased as sound pressure decreased, at arate of approximately 0.05±0.10 ms per 5-dB decrease insound level.

Waveforms were clear at levels 20±30 dB abovethreshold, and could be observed with less than 100averages (sweeps). As sound pressure levels approachedthreshold, 2000 sweeps were required to distinguishABRs from background noise. Threshold was de®ned asthe lowest sound pressure level that elicited an observ-able, repeatable ABR waveform (Jacobson 1985; Kilenyand Shea 1986; Gorga et al. 1988; Warren 1989; Hall1992; Hall 1992; Song and Schacht 1996). Superimpos-ing at least two replicate runs, each one representingaveraged traces at a particular level, and comparingthem to clear responses at higher sound pressure levelsfacilitated recognition of near-threshold ABRs (Fig. 5).

The number of cycles in the tone-burst stimuli af-fected the clarity of the elicited ABR waveform. Thus,cycle number was adjusted to provide the best ABRresponse while still providing acceptable power spectra(i.e., sharp peaks at the dominant frequency, as veri®edby FFT analysis using the BioSig software; see Fig. 2).Results indicated that shorter duration tone burstselicited the clearest ABRs, especially at low frequencies,so bursts of only two cycles in duration were used at100±300 Hz. Middle frequencies (400±2000 Hz) werepresented using ®ve cycle tone bursts, with two-cycle riseand fall. High frequency (3000±5000 Hz) tone burstswere eight cycles long, also with two-cycle rise and fall(see Silman and Silverman 1991 for technical details).

Audiograms

Audiograms of gold®sh were obtained from eightgold®sh treated with Flaxedil, and from three untreated

specimens. Auditory thresholds were signi®cantly lower(Table 1) in treated animals (Fig. 6), hence furthercomparisons will deal with audiograms from Flaxedil-treated individuals only. The overall shape of the au-diogram, as well as the individual sound pressurethresholds were quite similar to those obtained usingbehavioral methods (Table 1, Fig. 7). There was nosigni®cant dierence between the ABR audiogram andthe behavioral audiograms obtained by Jacobs andTavolga (1967) and Popper (1971). However, the ABRthresholds were generally higher than behavioralthresholds below 1500 Hz, and lower than behavioralvalues above 1500 Hz. Ambient noise levels were atleast 13 dB below threshold at all frequencies tested(Fig. 6).

Audiograms were also examined for eight oscars.Three of the animals were treated with Flaxedil, butbecause there was no signi®cant dierence between theseand untreated animals, data were pooled. As observed ingold®sh, the ABR audiogram had a similar shape to theonly available behavioral data (Fig. 7: Yan and Popper1992). However, it was signi®cantly dierent, with allthresholds well below behavioral limens (Table 2).

Discussion

ABR recording is a proven method of assessing auditoryfunction in a wide range of human clinical applications(for reviews see Jacobson 1985; Hall 1992). It has alsogained acceptance among many animal researchers, andhas been applied in a variety of studies with mammals(e.g., cetaceans, Ridgeway et al. 1981; Supin et al. 1993;Dolphin 1996; cats, Gorga et al. 1983; Mair and Laukli1985; rats, Overbeck and Church 1992; guinea pigs,Song and Schacht 1996 ), and birds (e.g., chickens, Tucci

Fig. 3 ABR waveforms ofgold®sh, Carassius auratus(upper two traces) and oscar,Astronotus ocellatus (middletwo traces), obtained in re-sponse to tone bursts(300 Hz) of opposite polari-ties, 90° and 270° (lower twotraces). Light traces: ABRwaveforms from tone burstspresented at 90°. Dark trac-es: ABR waveforms fromtone bursts presented at270°. Note that despite thechanges of polarities ofstimuli, the polarities ofABR waveforms remain thesame

311

and Rubel 1990). Furthermore, Corwin (1981) andCorwin et al. (1982) demonstrated that it was possible toobtain ABRs from poikilothermic vertebrates, includingreptiles, amphibians, osteichthyes (bony ®shes) and elas-mobranchs (cartilaginous ®shes). The ABR waveformexhibits remarkable similarity across vertebrate groups,which indicates a general consistency in the ascendingauditory pathways of all vertebrates (Corwin et al.1982).

The general features of the ABR as recorded fromgold®sh and oscars are in agreement with those de-scribed for other lower vertebrates (Corwin et al. 1982),and hence amniotes as well. Besides having similarwaveform characteristics, the response of the evoked

potential to varying stimulus parameters is also consis-tent. The polarity of the ABR waveform is independentof the polarity of the sound stimulus (Fig. 3), andincreases in latency as the stimulus level decreases. Thereis also a frequency-dependent delay in the onset andduration of the ABR. The present study shows that ®shABRs can be recorded in response to stimulusfrequencies ranging from at least 100 Hz to 5 kHz.Ongoing experiments in our laboratory have alsodemonstrated that such responses can be obtained froma variety of ®sh species (e.g., cichlids, centrarchids,minnows, gouramis, cat®sh) and are eective means ofobtaining complete audiograms within a short timeframe.

Fig. 4 ABR waveforms ofC. auratus (solid lines) andA. ocellatus (dotted lines)obtained in response to tonebursts and clicks of dierentfrequencies presented atsound pressure levels 30 dBabove hearing threshold(3 kHz not tested in theoscar). Averaged traces oftwo dierent polarities (1000sweeps each, a total of 2000sweeps) are shown. Notedecreasing onset latencywith increase of frequency

312

Fig. 5 ABR waveforms ofC. auratus in response to600 Hz tone bursts attenu-ated in 5-dB steps. Averagedtraces of two dierent runs(2000 sweeps each) for eachlevel are overlaid. Hydro-phone recording of 600 Hztoneburst is also shown. dBin relative scale

Table 1 Comparison of Carassius auratus audiograms obtained bydierent methods and treatments. P values give the signi®cancelevel of the dierence between ABR Flaxedil data, and ABR no

Flaxedil data as well as behavioral data from Jacobs and Tavolga(1967) and Popper (1971), tested by one-way ANOVA

Frequency ABRFlaxedil

ABRno Flaxedil

Jacobs andTavolga (1967)

Popper (1971)

Mean SD n Mean SD n Mean SD n Mean SD n

100 85.8 3.3 8 88 1 3 71.6 6.1 4 73.8 5.9 4200 73.3 4.3 8 79.3 2.1 3 58.7 6 4300 68.8 3.3 8 75.3 2.1 3 53.8 7.2 3400 63.9 2.9 8 74 3.5 3500 64 4 8 73.7 4.9 3 54.4 7.7 4 51.8 6.1 3600 64.1 4.2 8 71.3 3.8 3800 64 2.7 8 70 1 3 55.5 5.9 41000 64.6 3 8 66 3 3 54.9 7.6 4 60.1 7.4 31500 71.5 3.1 8 78.7 3.5 3 72.1 6.9 4 73.6 5.8 32000 80 2 8 84.3 4 3 98.2 6 4 94.6 6.7 33000 96.4 4.5 8 102.3 4.9 3 122.3 5.7 44000 107.4 4.3 8 113.3 4.9 35000 119.5 3.4 8 122.7 5.1 3

One-way ANOVA F = 66.33 F = 0.02 F = 0.36P < 0.001 P = 0.87 P = 0.56

313

Auditory thresholds and comparisonswith behavioral audiograms

Work with humans has demonstrated that auditorythresholds determined using tone-burst ABR audiome-try (e.g., frequency-speci®c measurements) are generallyhigher, by 10±20 dB or more, than those obtained usingbehavioral methods (Gorga et al. 1988). However, ABRthresholds in response to click (i.e., broad-band) stimulican approach behavioral levels in a limited frequencyrange between 2000 and 4000 Hz (Warren 1989). Aswith any electrophysiological technique, the minimumdetectable response is limited by the level of the ever-present ``noise ¯oor''. In order to reduce backgroundnoise (general brainwaves and myogenic signals) andhence increase the sensitivity of measurements, ABRs

are generally recorded from humans while the subjectsare in a resting or sedated state. Our data indicates thatuse of a curariform agent (e.g., Flaxedil) with somegold®sh also lowers its ABR thresholds which certainlyindicates the eect of reduced background noise onlowering thresholds.

Although generally accepted as the most sensitivetechniques for determining sensory limens, behavioralmethods do not always provide consistent results, andare not universally applicable. Behavioral testing canprovide varying results caused by dierent experimentalsetups, changes in subject motivation, etc. In addition,some subjects are even impossible to test behaviorally.Furthermore, in the two species of ®shes where multiplebehavioral audiograms have been performed, thresholdsvary greatly between studies. As reviewed by Hawkins(1981), the three audiograms available for the Atlanticcod, Gadus morhua, show discrepancies of as much as30 dB or more at some frequencies. These dierencescould be clearly attributed to varying ambient noiselevels during testing. Such environmental variabilitywould aect thresholds regardless of the method used toevaluate them. On the other hand, it is not so easy toexplain the extreme variability (greater than 60 dB atsome frequencies) among gold®sh audiograms. A varietyof causes have been suggested (Popper et al. 1973;Popper and Fay 1973; Hawkins 1981), including near-®eld eects, strain dierences, ambient noise, and dif-ferent conditioning techniques. It should be noted thatno statistical comparisons between these audiogramshave been made (except Popper 1971, for two dierentsize groups), and thus it is dicult to draw any con-clusions from simple observations of the curves. Clearly,the closest agreement among gold®sh audiograms arebetween those of Jacobs and Tavolga (1967) and Popper(1971). Both studies were conducted using the same

Fig. 6 Comparison (meanSD) of ABR audiograms of the cichlidA. ocellatus (closed circles) and the cyprinid C. auratus, the lattershowing in¯uence of Flaxedil treatment (open and closed triangles). Nodierences resulting from Flaxedil treatment were found in As-tronotus. Ambient noise level is indicated by the dashed line

Fig. 7 Comparison of ABRaudiograms (solid lines;closed circles: Astronotus;closed triangles: Carassius)and audiograms obtainedthrough dierent behavioralmethods (dotted lines). EEnger 1966 (n 6), F Fay1969 (n 4), J Jacobs andTavolga 1967 (n 4), PPopper 1971 (n 6), Y Yanand Popper 1992 (n 3).ABR: Astronotus (n 8),Carassius (n 8)

314

method of instrumental shock-avoidance conditioningin a small tank with a speaker in air. A well-knownproblem in underwater acoustic experiments is thecomplexity and high particle motion of sound ®eldsgenerated by an underwater speaker inside small tankssurrounded by air (Parvulescu 1967). Although not aperfect solution, placing the sound source in air andusing a small container of water helps to reduce theseeects. The similar acoustic conditions of our setup andthose of Popper (1971), and Jacobs and Tavolga (1967),allowed us to use these audiograms as benchmarks for acomparison of ABR and behavioral thresholds. It isimportant to note that the behavioral or electrophysio-logical responses are not only caused by the pressurecomponent of the sounds, but also can be caused by theparticle component. However, this should not detractfrom the merit of comparing dierent methods.

Our gold®sh ABR audiograms show shape andbandwidth characteristics similar to those of behavioralcurves, and were not signi®cantly dierent on a whole-audiogram basis. However, at lower frequenciesthresholds were generally higher, while at higher fre-quencies they were lower. Unlike that observed in hu-mans (Gorga et al. 1988), there was no simple upwardshift of the ABR audiogram versus behavioral data. Onecould argue that the lower thresholds at high frequenciesmight have been caused by broad power spectrums ofthe tone bursts, i.e., a particular threshold was a re-sponse to a broad band of frequencies. However, fast-Fourier transformation analysis of the tone-burst stimuli(Fig. 2) presented indicated that this is unlikely, as en-ergy on either side of the spectral peak dropped orapidly enough (owing to the Blackman gating) to pre-clude such a possibility.

In contrast to the gold®sh, little hearing data areavailable for the oscar. Cichlids are hearing generalistsand in general are dicult or impossible to train in somebehavioral paradigms, especially when electric shock isused as the stimulus (Tavolga 1974; H.Y. Yan, unpub-lished observations). Hence, the only audiogram dataavailable for the oscar is a sound pressure audiogramwhich was obtained through positive reward operant

conditioning (Yan and Popper 1992), and even this op-erant conditioning method proved quite dicult andtime consuming. Our oscar ABR audiogram, whilesimilar in shape and slope to this behavioral curve, isbelow it at all frequencies, and hence signi®cantly dif-ferent. Therefore, in a species like the oscar which isdicult to train, ABR threshold determination may wellbe superior to behavioral methods. The unusual ramp-like shape of the audiogram, combined with the poorsensitivity at frequencies above 100 Hz, suggests thatthis animal probably has a best frequency below 100 Hz.Testing frequencies below 100 Hz turned out to be dif-®cult with the ABR technique due to the diculty cre-ating short tone bursts at lower frequencies (Silman andSilverman 1991; Hall 1992). Another cichlid, T. macro-cephala (Tavolga 1974) showed a similar audiogrampro®le like that of the oscar when tested behaviorally. Inhearing generalists which are known to be especiallysensitive to particle motion at low frequencies (Popperand Fay 1973; Hawkins 1993), the kinetic component ofsound may additionally be measured in order to get aparticle displacement hearing curve. Our intention wasto evaluate our methods and results by comparing themwith previously established techniques. Because themajority of investigators published sound pressure au-diograms (see Fay 1988), we compared only soundpressure thresholds. In addition, testing of frequenciesbelow 100 Hz is dicult in the laboratory due to po-tential lateral line stimulation (MuÈ nz 1989). However,we ruled out a lateral line contribution at 100 Hz andabove, at least in the gold®sh. A gold®sh treated for 24 hwith 0.1 mmol l)1 cobalt chloride (a lateral line functionblocker, Karlsen and Sand 1987) showed no changes inABR thresholds (T. N. Kenyon et al., unpublished ob-servations).

The present data clearly show that ABR audiometryprovides results comparable to those obtained behav-iorally. An obvious and important advantage is the ra-pidity with which a complete audiogram can beobtained. Instead of requiring days or weeks to obtainaudiograms by behavioral methods, an ABR audiogramcan be completed in as little as a few hours. For exam-

Table 2 Comparison ofAstronotus ocellatus audio-grams obtained by dierentmethods. P value gives the sig-ni®cance level of the dierencebetween ABR data and beha-vioral data from Yan and Pop-per (1992) tested by one-wayANOVA

Frequency ABR Yan and Popper 1992

Mean SD n Mean SD n

100 100.5 4.6 8200 105.9 5.8 8 118.4 2.8 3300 106.4 1.9 8 120.5 3 3400 112.3 1.8 8 120.7 3.9 3500 116.3 1.8 8 125.1 3.5 3600 116.4 3.2 8 129.6 1.8 3700 131.4 2.1 3800 117.8 2.6 8 134 1.9 31000 118.3 2.9 81200 124.8 2.0 81500 130.3 3.5 82000 134.8 4.9 8

One-way ANOVA F = 13.84 P < 0.001

315

ple, we were able to measure the temporary thresholdshift in a few gold®sh caused by overnight exposure toloud pump/air stone noise, over the entire hearing rangewithin 3 h of removing the animals from the noisy tank.Slower methods such as behavioral training may not beable to measure such a relatively short-lived phenome-non. As pointed out above, ABR techniques may pro-vide a way to study hearing in previously untestablespecies. Because it does not rely on the motivationalstate of the subject, it could well be employed in exper-imental situations where behavioral testing would beimpossible or dicult. We are currently using ABRaudiometry to measure hearing impairment caused byaminoglycoside antibiotic (e.g., gentamicin)-inducedototoxicity in oscars. It would be impossible to use be-havioral methods with these animals, since the drugtreatments render them lethargic, induce loss of vestib-ular function and make them unsuitable for behavioraltesting. Several advantages are also apparent whencomparing ABR audiometry to other electrophysiolog-ical methods: (1) no invasive surgery is required, savingtime and animals, (2) individuals can easily be subjectedto repeated testing with no apparent ill eects, (3) thepreparation is far easier than with invasive methods in-volving surgical operation, allowing de®nite results to beobtained within minutes of removing the subject from itsholding tank. Unlike microphonic or single-unit re-cording techniques, the sensitivity of the entire ascend-ing auditory pathway is measured, giving a goodrepresentation of the sensitivity of the entire system in-cluding the brainstem and higher brain regions (Corwin1981; Jacobson 1985; Silman and Silverman 1991; Hall1992). However, one should be aware of the potentialdisadvantage of subjectivity of threshold determinationin ABR audiometry. ABR is frequently viewed as anobjective procedure because acquisition of the test resultis not dependent upon the subject's conscious coopera-tion. Interpretation of the results, however, is highlysubjective. The examiner must look at pairs of wavylines and make a number of very subjective decisionsregarding the presence of or absence of a response.Though the ABR is highly repeatable and easily recog-nized in high-quality recording conditions, the responsecan be very illusive when a hearing-impaired subjectis being evaluated (see Weber 1983, 1985; Silman andSilverman 1991 for details). In order to overcomesubjective pitfall, many automated threshold-seekingalgorithms have been developed for more objectivedetermination of ABR thresholds (Weber and Fletcher1980; Salvi et al. 1987; OÈ zdamar et al. 1994). With theavailability of more sophisticated computers and auto-mated threshold-seeking software, the problem of sub-jective determination of ABR should be solved in thenear future.

Clearly, ABR recording is a viable and eectivemethod for studying audition in ®shes. Its advantages ±rapidity and repeated use of animals ± might help toincrease the small number of ®sh species in whichhearing sensitivity is known so far. To date, audiograms

of about 50 species are published which is barely ade-quate to describe the diversity of hearing abilities arisingfrom numerous morphological adaptations, e.g., We-berian ossicles in ostariophysans (von Frisch 1936),swimbladder diverticulae in holocentrids (Coombs andPopper 1979), or gas-®lled bullae in mormyrids (Stipetic1939; McCormick and Popper 1984). ABR techniquescould be useful in systematic studies when a largenumber of species has to be analyzed, e.g., for investi-gating dierent adaptations for hearing or correlationsbetween hearing sensitivity and sounds produced. Basedon our ongoing experiments, and according to Corwinet al. (1982), there are no systematic limitations or majorsize limitations for this method. We have successfullytested a few juvenile bluegill (Lepomis macrochirus; av-erage SL 18 mm, mean body weight 0.09 g). Therefore,this method should be applicable to studies of ontoge-netic changes of hearing in ®shes, an area poorly studiedto date (Kenyon 1996), largely owing to the diculty ofemploying behavioral methods with juveniles. Additionof ABR audiometry to the traditional repertoire ofmethods available for studying hearing in ®shes shouldprovide a tool that expedites the answers to many of theremaining questions in ®sh audition.

Acknowledgements Initial discussion with William M. Saidel of theBiology Department, Rutgers University, led to the development ofthis project. Jerey T. Corwin of University of Virginia providedinvaluable advice and encouragement on the setup of the system.Victor N. Rush provided much needed assistance during earlyphase of development of the protocol. William M. Saidel, ArthurN. Popper and three anonymous referees oered useful commentson earlier drafts of the manuscript. This study was supported bygrants from the NIH-DC 01729, Deafness Research Foundation,National Organization for Hearing Research, Center for Ecology,Evolution and Behavior of the University of Kentucky, KentuckyWater Resources Research Institute and the University of Ken-tucky Vice Chancellor oce for Research and Graduate Studies toHYY. FL was supported by a grant from the Austrian ScienceFoundation (FWF P10295). These experiments comply with the``Principles of animal care'', publication No. 86-23, revised 1985 ofthe National Institute of Health, and were approved by the Uni-versity of Kentucky IACUC (93005L).

References

Allen EE, Fernald RD (1985) Spectral sensitivity of the Africancichlid ®sh, Haplochromis burtoni. J Comp Physiol A 157: 247±253

Banner A (1967) Evidence of sensitivity to acoustic displacementsin the lemon shark, Negaprion brevirostris (Poey). In: Cahn PE(ed) Lateral line detectors. Indian University Press, Blooming-ton, pp 265±273

Behrend ER, Bitterman ME (1962) Avoidance conditioning in thegold®sh: exploratory studies of the CS-US interval. Am J Psy-chol 75: 18±34

Bigelow, HB (1904) The sense of hearing in the gold®sh Carassiusauratus L. Am Nat 38: 275±284

Bullock TH (1981) Neuroethology deserves more study of evokedresponses. Neuroscience 6: 1203±1215

Burkhard R (1984) Sound pressure level measurement and spectralanalysis of brief acoustic transients. Electroencephalogr ClinNeurophysiol 57: 83±91

Chapman CJ, Sand O (1974) Field studies of hearing in twospecies of ¯at®sh Pleuronectes platessa (L.) and Limanda

316

limanda (L.) (family Pleuronectidae). Comp Biochem PhysiolA 47: 371±385

Coombs S, Popper AN (1979) Hearing dierences among Hawai-ian squirrel®sh (family Holocentridae) related to dierences inthe peripheral auditory system. J Comp Physiol A 132: 203±207

Corwin JT (1981) Audition in elasmobranchs. In: Tavolga WN,Popper AN, Fay RR (eds) Hearing and sound communicationin ®shes. Springer, Berlin Heidelberg New York, pp 81±105

Corwin JT, Bullock TH, Schweitzer J (1982) The auditory brainstem response in ®ve vertebrate classes. Electroencephalogr ClinNeurophysiol 54: 629±641

Dolphin WF (1996) Auditory evoked responses to amplitudemodulated stimuli consisting of multiple envelope components.J Comp Physiol A 179: 113±121

Enger PS, Anderson R (1967) An electrophysiological ®eld study ofhearing in ®sh. Comp Biochem Physiol 22: 517±525

Fay RR (1969) Behavioral audiogram for the gold®sh. J Aud Res9: 112±121

Fay RR (1988) Hearing in vertebrates: a psychophysics databook.Hill-Fay, Winetka, Illinois

Fay RR (1995) Psychoacoustical studies of the sense of hearing ingold®sh using conditioned respiratory suppression. In: KlumpGM, Dooling RJ, Fay RR, Stebbins WC (eds) Methods incomparative psychoacoustics. BirkhaÈ user, Basel, pp 249±261

Fay RR, Popper AN (1974) Acoustic stimulation of the ear of thegold®sh (Carassius auratus). J Exp Biol 61: 243±260

Frisch K von (1936) UÈ ber den GehoÈ rsinn der Fische. Biol Rev 11:210±246

Furukawa T, Ishii T, Matsuura S (1972) An analysis of micro-phonic potentials of the sacculus of gold®sh. Jpn J Physiol 22:603±616

Gorga MP, McGee J, Walsh EJ, Javel E, Farley GR (1983) ABRmeasurement in the cat using a forward masking paradigm.J Acoust Soc Am 73: 255±261

Gorga MP, Kaminski JR, Beauchaine KA, Jesteadt W (1988)Auditory brainstem response to tone bursts in normally hearingsubjects. J Speech Hear Res 31: 87±97

Hall JW (1992) Handbook of auditory evoked responses. Allyn andBacon, Boston

Hawkins AD (1981) The hearing abilities of ®sh. In: Tavolga WN,Popper AN, Fay RR (eds) Hearing and sound communicationin ®shes. Springer, Berlin Heidelberg New York pp 109±137

Hawkins AD (1993) Underwater sound and ®sh behavior. In:Pitcher TJ (ed) Behavior of teleost ®shes. Chapman and Hall,London, pp 129±169

Hawkins AD, Johnstone ADF (1978) The hearing of the Atlanticsalmon, Salmo salar. J Fish Biol 13: 655±673

Hawkins AD, Myrberg AA (1983) Hearing and sound communi-cation underwater. In: Lewis B (ed) Bioacoustics, a comparativeapproach. Academic Press, London, pp 347±405

Jacobs DW, Tavolga WN (1967) Acoustic intensity limens in thegold®sh. Anim Behav 15: 324±335

Jacobson JT (1985) An overview of the auditory brainstem re-sponse. In: Jacobson JT (ed) The auditory brainstem response.College-Hill Press, San Diego, pp 3±12

Jerkù H, Turunen-Rise I, Enger PS, Sand O (1989) Hearing in theeel (Anguilla anguilla). J Comp Physiol A 165: 455±469

Jewett DL (1970) Volume conducted potentials in response to au-ditory stimuli as detected by averaging in the cat. Electro-encephalogr Clin Neurophysiol 28: 609±618

Jewett DL, Williston JS (1971) Auditory evoked far ®elds averagedfrom the scalp of humans. Brain 94: 681±696

Karlsen HE, Sand O (1987) Selective and reversible blocking of thelateral line in freshwater ®sh. J Exp Biol 133: 249±262

Kenyon TN (1996) Ontogenetic changes in the auditory sensitivityof the bicolor damsel®sh, Pomacentrus partitus (Poey). J CompPhysiol A 179: 553±561

Kileny P, Shea SL (1986) Middle-latency and 40-Hz auditoryevoked responses in normal-hearing subjects: clicks and 500-Hzthresholds. J Speech Hear Res 29: 20±28

Kojima T, Shimamura T, Kazumasa Y, Okumoto N, HatakeyamaY, Soeda H (1992) W-shaped auditory threshold curves of masu

salmon Onchorhynchus masou. Nippon Suisan Gakkaishi 58:1447±1452

Mair IWS, Laukli E (1985) Frequency speci®city of the auditorybrainstem response in the cat. Acta Otolaryngol 99: 377±383

McCormick CA, Popper AN (1984) Auditory sensitivity andpsychophysical tuning curves in the elephant nose ®sh,Gnathonemus petersii. J Comp Physiol A 155: 753±761

MuÈ nz H (1989) Functional organization of the lateral lineperiphery. In: Coombs S, GoÈ rner P, MuÈ nz H (eds) The mech-anosensory lateral line: neurobiology and evolution. Springer,Berlin Heidelberg New York, pp 285±297

Myrberg AA, Spires JY (1980) Hearing in damsel®shes: an analysisof signal detection among closely related species. J CompPhysiol A 140: 135±144

Nelson DR (1967) Cardiac responses to sound in the lemon shark,Negaprion brevirostris. In: Gilbert GW, Mathewson RF, RallDP (eds) Sharks, skates, and rays. Johns Hopkins UniversityPress, Baltimore, pp 533±544

Nelson JS (1984) Fishes of the world, 2nd edn. Wiley, New YorkOverbeck GW, Church MW (1992) Eects of tone burst frequency

and intensity on the auditory brainstem response (ABR) fromalbino and pigmented rats. Hear Res 59: 129±137

OÈ zdamar OÈ , Delgado RE, Eilers RE, Urbano RC (1994) Auto-mated electrophysiologic hearing testing using a threshold-seeking algorithm. J Am Acad Audiol 5: 77±88

Parker GH (1903) The sense of hearing in ®shes. Am Nat 37: 185±204

Parvulescu A (1967) The acoustics of small tanks. In: Tavolga WN(ed) Marine bio-acoustics, vol 2. Pergamon Press, Oxford, pp7±13

Popper AN (1971) The eects of size on the auditory capacities ofthe gold®sh. J Aud Res 11: 239±247

Popper AN, Fay RR (1973) Sound detection and processing byteleost ®shes: a critical review. J Acoust Soc Am 53: 1515±1529

Popper AN, Chan AT, Clarke NL (1973) An evaluation of meth-ods for behavioral investigations of teleost audition. Behav ResMethods Instrum 5: 470±472

Ridgeway SH, Bullock TN, Carder DA, Seeley RL, Woods D,Galambos R (1981) Auditory brainstem response in dolphin.Proc Natl Acad Sci USA 78: 1943±1947

Saidel WM, Popper AN (1987) Sound perception in two anabantid®shes. Comp Biochem Physiol A 88: 37±44

Salvi RJ, Ahroon W, Saunders SS, Arnold SA (1987) Evoked po-tentials: computer-automated threshold tracking procedureusing an objective detection criterion. Ear Hear 8: 151±156

Sand O (1974) Directional sensitivity of microphonic potentialsfrom the perch ear. J Exp Biol 60: 881±899

Silman S, Silverman CA (1991) Auditory diagnosis: principles andapplications. Academic Press San Diego

Song BB, Schacht J (1996) Variable ecacy of radical scavengersand iron chelators to attenuate gentamicin ototoxicity in guineapig in vivo. Hear Res 94: 87±93

Stipetic E (1939) UÈ ber das GehoÈ rorgan der Mormyriden. Z VerglPhysiol 26: 740±752

Supin AY, Popov VV, Klishin VO (1993) ABR frequency tuningcurves in dolphins. J Comp Physiol A 173: 649±656

Tavolga WN (1974) Signal/noise ratio and the critical band in®shes. J Acoust Soc Am 55: 1323±1333

Tavolga WN, Wodinsky J (1963) Auditory capacities in ®shes. Puretone thresholds in nine species of marine teleosts. Bull Am MusNat Hist 126: 177±240

Tucci DL, Rubel EW 1990. Physiologic status of regenerated haircells in the avian inner ear following aminoglycoside ototoxic-ity. Otolaryngol Head Neck Surg 103: 443±450

Warren MP (1989) The auditory brainstem response in pediatrics.Pediatr Otolaryngol 22: 473±500

Weber BA (1983) Pitfalls in auditory brain stem response audio-metry. Ear Hear 4: 179±184

Weber BA (1985) Interpretation: problems and pitfalls. In: Ja-cobson JT (ed) The auditory brainstem response. College Hill,San Diego, pp 99±112

317

Weber BA, Fletcher GL (1980) A computerized scoring procedurefor auditory brainstem response audiometry. Ear Hear 1: 233±236

Weiss BA, Strother WF, Hartig GM (1969) Auditory sensitivity inthe bullhead cat®sh (Ictalurus nebulosis). Proc Natl Acad SciUSA 64: 552±554

Yan HY (1995) Investigations of ®sh hearing ability using an au-tomated reward method. In: Klump GM, Dooling RJ, Fay RR,Stebbins WC (eds) Methods in comparative psychoacoustics.BirkhaÈ user, Basel, pp 263±276

Yan HY, Popper AN (1991) An automated positive reward methodfor measuring acoustic sensitivity in ®sh. Behav Res MethodsInstrum 23: 351±356

Yan HY, Popper AN (1992) Auditory sensitivity of the cichlid ®shAstronotus ocellatus (Cuvier). J Comp Physiol A 171: 105±109

Yan HY, Popper AN (1993) Acoustic intensity discrimination bythe cichlid ®sh Astronotus ocellatus (Cuvier). J Comp Physiol A173: 347±351

318