Diversity in frequency response properties of saccular afferents of the toadfish, Opsanus tau

Hearing Research 113 (1997) 235^246

Diversity in frequency response properties of saccular a¡erents of the toad¢sh, Opsanus tau Richard R. Fay a

a ;b ;

*, Peggy L. Edds-Walton

b

Parmly Hearing Institute and Department of Psychology, Loyola University Chicago, 6525 North Sheridan Rd., Chicago, IL 60626, USA

b

Marine Biological Laboratory, Woods Hole, MA 02543, USA

Received 17 March 1997; revised 22 July 1997; accepted 29 July 1997

Abstract

The frequency response of primary saccular afferents of toadfish (Opsanus tau) was studied in the time and frequency domains using the reverse correlation (revcor) method. Stimuli were noise bands with flat acceleration spectra delivered as whole-body motion. The recorded acceleration waveform was averaged over epochs preceding and following each spike. This average, termed the revcor, is an estimate of the response of an equivalent linear filter intervening between body motion and spike initiation. The spectrum of the revcor estimates the shape of the equivalent linear filter. Revcor responses were brief, damped oscillations indicative of relatively broadly tuned filters. Filter shapes were generally band-pass and differed in bandwidth, band edge slope, and characteristic frequency (74 Hz to 140 Hz). Filter shapes tend to be independent of stimulus level. Afferents can be placed into two groups with respect to characteristic frequency (74^88 Hz and 140 Hz). Some high-frequency afferents share a secondary peak at the characteristic frequency of low-frequency afferents, suggesting that an afferent may receive differently tuned peripheral inputs. For some afferents having similar filter shapes, revcor responses often differ only in polarity, probably reflecting inputs from hair cells oriented in opposite directions. The origin of frequency selectivity and its diversity among saccular afferents may arise from a combination of hair cell resonance and micromechanical processes. The resulting frequency analysis is the simplest yet observed among vertebrate animals. During courtship, male toadfish produce the `boatwhistle' call, a periodic vocalization having several harmonics of a 130 Hz fundamental frequency. The saccule encodes the waveform of acoustic particle acceleration between 50 and about 250 Hz. Thus, the fundamental frequency component of the boatwhistle is well encoded, but the successive higher harmonics are filtered out. The boatwhistle is thus encoded as a time-domain representation of its fundamental frequency or pulse repetition rate.

6

Keywords :

Fish; Hearing; Auditory nerve; Reverse correlation

1. Introduction

Fishes share a primitive mode of hearing in which one or more of the otolith organs respond directly, as inertial accelerometers, to the acoustic particle motion that accelerates the ¢sh's body in a sound ¢eld (Fay and Edds-Walton, 1997). In addition, some ¢shes have a more specialized mode of hearing in which sound pressure £uctuations cause motions of the swimbladder (or other gas bladder) that are transmitted to the inner ears (von Frisch, 1938). For most ¢shes, the saccule func* Corresponding author. Tel.: +1 (773) 508-2714; Fax: +1 (773) 508-2719; E-mail: [email protected]

tions as the auditory organ with a bandwidth from several hundred Hertz for the primitive mode of hearing, to several thousand Hertz for species specialized to detect sound pressure (Popper and Fay, 1993). The psychophysics (Fay, 1988) and neurophysiology (e.g. Fay et al., 1996) of hearing among ¢shes have been quantitatively studied primarily in the pressure-sensitive `hearing specialists' such as the gold¢sh (Carassius auratus) because the sound pressure stimulus can be easily controlled and measured. Sound encoding in the gold¢sh saccule begins with a crude peripheral frequency analysis (Fay, 1996) that is enhanced and diversi¢ed by inhibition in the central auditory system (Lu and Fay, 1993, 1996). This analysis is revealed in psycho-

0378-5955 / 97 / $17.00 ß 1997 Published by Elsevier Science B.V. All rights reserved PII S 0 3 7 8 - 5 9 5 5 ( 9 7 ) 0 0 1 4 8 - 2

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

236

physical studies of frequency processing (e.g. Fay et al.,

and to promote oxygenation of the tissues. Recordings

1978 ; Fay, 1995). These and other results for gold¢sh

were made from the saccular nerve of the left ear.

suggest that ¢shes share the strategy of peripheral audi-

After the saccular nerve was exposed to view, the ¢sh

tory frequency analysis that has been observed among

was placed in an aluminum cylinder (diameter = 23 cm,

all vertebrate species investigated.

height = 8 cm, wall thickness = 0.5 cm) ¢lled with fresh

Most species of ¢sh do not have specialized links

seawater. The cylinder was mounted on a three-dimen-

between the swimbladder (or other gas bubble) and

sional shaker table (described below). A rigid holder

the ears, and thus may have only the primitive, accel-

was used to secure the head of the ¢sh in the cylinder.

erometer mode of hearing. This mode of hearing has

Water was changed when necessary to maintain temper-

only rarely been studied quantitatively (e.g. Chapman

ature and appropriate dissolved oxygen levels.

and Sand, 1974 ; Sand and Karlsen, 1986), primarily because it is di¤cult to manipulate, measure, or calcu-

2.2. Directional stimulation

late the magnitude and direction of the adequate stimulus : acoustic particle motion. Psychophysical studies

Gadus morhua)

The shaker table (Fay, 1984 ; Lu et al., 1996 ; Fay

have

and Edds-Walton, 1997) uses a system of three, mov-

demonstrated that frequency analysis occurs at a behav-

ing-coil shaker channels to create motion along one of

ioral level (e.g. Hawkins and Chapman, 1975), but there

three orthogonal directions. Signals were 30 digitally

are no quantitative neurophysiological studies on the

synthesized noise bands, each 0.81 s in duration, having

frequency

£at acceleration spectra between 50 and 1000 Hz. Ten

on unspecialized species (e.g. cod,

selectivity

of

primary

a¡erents

(but

see

noise samples were created for each of three orthogonal

Fine, 1981 ; Horner et al., 1981). We report experiments on frequency selectivity in saccular a¡erents of the toad¢sh (Opsanus

tau),

directions. These were read out of Tucker Davis Tech-

a spe-

nologies (TDT) 16-bit digital-to-analog converters (10

cies that is not specialized for sound pressure detection.

kHz sample rate), low-pass ¢ltered at 2 kHz, attenuated

Our goal was to quantitatively describe peripheral fre-

using TDT PA4 programmable attenuators, and ampli-

quency selectivity in this species for comparison with

¢ed (Techron, model 5507). Fixed resistors at the power

the gold¢sh, the only other ¢sh species for which we

ampli¢er outputs were used to attenuate the signals to

have comparable data, to help evaluate the notion that

the shakers by 32 dB. In combination with increased

peripheral frequency analysis may be common among

signal levels input to the power ampli¢ers, this attenu-

¢shes, and thus among all vertebrate classes. Since the

ation improved the signal-to-ampli¢er noise ratios. The

toad¢sh is not specialized to detect sound pressure, a

shaker table was supported by a pneumatic vibration

motional stimulus was used for which acceleration mag-

isolation system. Cylinder movement was calibrated by

nitudes and directions were manipulated and measured

monitoring the output of three accelerometers (Piezo-

along three orthogonal axes (see also Fay, 1984 ; Fay

tronics model 002A10, Flexcel), orthogonally oriented

and Edds-Walton, 1997).

on the up-down, side-side, and front-back axes. For reverse correlation (revcor) data collection, an a¡erent was stimulated using motion along the one of these

2. Materials and methods

three axes that produced the greatest spike rate.

2.1. Preparation of the toad¢sh

2.3. Extracellular recording

Toad¢sh 15^25 cm in total length were obtained

Extracellular recordings were made using 3 M KCl-

6.

from the Marine Biological Laboratory (MBL) Depart-

¢lled pipettes pulled to a tip resistance of 60^120 M

ment of Marine Resources. The toad¢sh were sexually

A 100 Hz search stimulus was synthesized to produce

mature and were maintained at 10^20³C in fresh sea-

equal amplitudes of motion along the three orthogonal

water. The care and use of toad¢sh were carried out

axes. The electrode was advanced through the saccular

using protocols approved by The Marine Biological

nerve after contact with the nerve was con¢rmed visu-

Laboratory and Loyola University Chicago.

ally. Spikes were discriminated using a voltage criterion

Toad¢sh were anesthetized (immersed in a bath of 1 :4000 solution of 3-aminobenzoic acid, methanesulfo-

under software control, and spike times were recorded with 0.1 ms resolution.

nate salt, Sigma) and immobilized (an intramuscular

After an a¡erent was contacted, its three-dimension-

injection of 0.1^0.2 ml of pancuronium bromide in a

al directional sensitivity was measured to sinusoidal

2 mg/ml solution ; Sigma) and placed in a saltwater-

stimuli as described in a previous report by Fay and

¢lled dissection box. The cranium was entered dorsally

Edds-Walton (1997). Following this, two repetitions

and the cerebrospinal £uids surrounding the brain were

of 10, 0.81-s noise stimuli (16 s total stimulus time)

gradually replaced with a £uoroinert liquid (FC-77, 3-

were presented successively from the three orthogonal

M Corp.) to provide a clearer view of the saccular nerve

directional channels and the numbers of spikes were

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

237

Fig. 1. Schematic rationale for the reverse correlation (revcor) method used to obtain revcors and ¢lter functions. From the bottom, spike times de¢ne 102.4 ms time epochs of the acceleration stimulus waveform that are averaged to form the revcor. The amplitude spectrum of the revcor de¢nes the ¢lter shape (see text).

recorded.

The

duced the

directional

stimulus

greatest number of

channel

spikes

that

pro-

was chosen

for

smoothed with a 10-point (1 ms) moving average window.

further tests on the assumption that its axis of motion was closest to the characteristic axis (CA) of the a¡erent

under

study

CA

is

axis

the

(Fay in

and

Edds-Walton,

spherical

coordinates

1997).

along

The

3. Results

which

an a¡erent has its lowest threshold. An a¡erent's fre-

Revcor analyses were carried out for 119 saccular

quency response was estimated using the reverse corre-

a¡erents from 22 toad¢sh. Fig. 2 shows ¢lter shapes

lation or revcor method (see below) by presenting the

for 19 saccular a¡erents (right) and selected revcor re-

20 noise samples along the appropriate axis at several

sponses

levels.

frequencies (CF) and sharpness of tuning observed. Fil-

(left)

representing

the

range

of

characteristic

ter shapes are generally band-pass and di¡er primarily

2.4. Revcor analysis

in bandwidth, slope of the high-frequency band-edge, and CF (74 Hz to 140 Hz). Revcors are damped oscil-

Segments of the recorded accelerometer waveforms (51.2 ms preceding and following each spike) were averaged for all spikes that occurred at least 51.2 ms after

lations di¡ering in polarity, oscillation rate, duration, and in the temporal asymmetry of their envelopes. A¡erents of

are

best

grouped

response

according in

Fig.

2.

to

the

Panels

frequency

stimulus onset and 51.2 ms before stimulus o¡set, as

range

A

illustrated in Fig. 1. This 1024-point averaged wave-

show 11 representative low-frequency a¡erents having

and

B

form (the revcor) estimates the linear ¢ltering that pre-

CFs ranging between 74 and 88 Hz. Panel B a¡erents

cedes

(de

have a wider bandwidth that those in panel A. The top

Boer and de Jongh, 1978). A fast Fourier transform

four revcors come from some of these low-frequency

(Matlab)

¢lter

a¡erents. Panel C shows eight a¡erents having the high-

shape between 50 and 1000 Hz within about a 20 dB

est CFs (140 Hz) and sharpest tuning observed. Some

dynamic

for

of the high-frequency a¡erents share a secondary peak

dynamic

at the CF of the low-frequency a¡erents. The bottom

range. Before computing the spectrum the revcor was

three revcors of Fig. 2 are from the most sharply tuned

spike

several

generation

of

the

range. noise

revcor

Revcor

levels

for

phase-locked

estimated

responses

within

each

an

a¡erents

a¡erent's

were

obtained

a¡erent's

HEARES 2903 28-11-97

238

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

Fig. 2. Right panels : Filter functions for 19 representative saccular a¡erents obtained in response to £at-spectrum noise using the revcor method. The ordinate is acceleration in dB with an arbitrary reference. Left panels : selected revcors for some of the ¢lter functions. The animal and a¡erent are identi¢ed on the left, and the number of spikes used in forming the revcor is on the right. The ¢lter functions for revcors M10, V14, and V18 are indicated with dotted lines. The thin vertical line through all functions indicates the spike time.

3

high-CF a¡erents observed (M10, V14, V18), whose

than

¢lter functions are shown in panel C as dotted lines.

order,

36 dB/oct, indicating ¢ltering of high dynamic

In these cases, sharp tuning is primarily determined

The low-frequency roll-o¡ rate is about 12 dB per oc-

by a steep high-pass characteristic.

tave.

according

to

the

de¢nition

of

Lewis

(1992).

Panel D shows the averaged ¢lter functions for the

A¡erents having similar ¢lter shapes sometimes have

a¡erents in panels A, B, and C. The average high-fre-

revcors that di¡er only in polarity (e.g. compare H2 to

quency roll-o¡ rate for these cells is near or steeper

H18, E1 to E2, and D8 to D11). Polarity di¡erences

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

239

Fig. 3. Revcors (A), ¢lter functions (B), and phase response functions (C) for a¡erents E1 and E2. Panel D plots thresholds to sinusoidal stimulation (100 Hz) in polar coordinates for directional stimulation in two planes (see Fay and Edds-Walton (1997) for a description of how these directional data were obtained and interpreted). These two a¡erents have nearly identical response properties with the exception that they are excited by opposite directions of motion.

probably arise from inputs from hair cells that are ori-

excited by movement in the opposite direction along the

ented in opposite directions on the saccular epithelium

same axis.

(Edds-Walton and Popper, 1995). Fig. 3 illustrates this

Revcors and ¢lter shapes tend to be independent of

for two a¡erents encountered close to one another in

stimulus level within a given a¡erent. Fig. 4 illustrates

one electrode penetration. The right panel (D) shows

this for three a¡erents having di¡erent ¢lter functions.

that the two cells have essentially identical 3-dimension-

A¡erent H20 (spontaneous rate of 1 spike/s) has a CF

al directionality as de¢ned in Fay and Edds-Walton

of 74 Hz and a low-pass characteristic of about

(1997). The most excitatory axis of motion is on a

per octave. Within the dynamic range of this a¡erent

line from up-right-front to down-left-back. Panel B

(between 10 and 41 spikes/s), the ¢lter functions are

shows that the ¢lter shapes are nearly identical. How-

nearly identical. A¡erent J10 (spontaneous rate of 56

ever, the revcors are inverted versions of one another

spikes/s) has a CF of 140 Hz and has a secondary peak

(panel A), and the phase functions of frequency are

at 74 Hz. The frequency selectivity of this a¡erent has

separated by about 180³ throughout the frequency re-

both low- and high-pass characteristics (about 12 and

sponse area (panel C). This simple polarity di¡erence indicates that one of these cells is excited by movement in one direction along a single axis, and the other is

3

3

18 dB

24 dB per octave, respectively). A¡erent V18 (sponta-

neous rate of 112 spikes/s) has a CF of 140 Hz and has relatively steep roll-o¡s for both the low- and high-pass

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

240

Fig. 4. Revcors (left panels) and ¢lter functions (right panels) for three representative saccular a¡erents illustrating the e¡ects of overall stimulus level within the dynamic range. For each revcor, the relative stimulus level is indicated on the left as an attenuation in dB with an arbitrary reference. The number of spikes used in forming the revcor is on the right. For the ¢lter functions, the ordinate is acceleration in dB with an arbitrary reference.

3

36 dB per octave, respectively).

per dB, respectively). The phase response of a¡erent J10

Narrowly tuned, high-CF a¡erents similar to V18 are

resembles that of H20 at low frequencies (up to 120 Hz)

the only ones to show a suggestion of decelerating ¢lter

and that of V18 at higher frequencies. As shown in Fig.

slopes (at frequencies below CF) and highly temporally

4, the ¢lter function for J10 also resembles that for H20

asymmetric revcor envelopes. All other saccular a¡er-

at low frequencies, and that for V18 at high frequencies.

ents tend to have a ¢lter shape with convex skirts and a

A¡erent J10 appears to receive inputs from two, dif-

characteristics (18 and

revcor that tends to be relatively more symmetric.

ferently tuned peripheral processes. Fig. 6A shows that

Fig. 5 shows the phase functions of frequency for the

J10's two-peaked ¢lter function can be understood as

three a¡erents of Fig. 4. All three a¡erents exhibit fre-

resulting from inputs from one process having a 74 Hz

quency-dependent phase lags varying over at least one

CF

cycle (360³) throughout their frequency response range.

tuned at 140 Hz (the ¢lter function of V18). The dotted

A¡erent H20 shows a progressive phase lag as stimulus

line

level declines within a 15 dB range (mean of 4.3³ þ 0.34

3) +AV18 )

S.E.M. per dB). The phase of a¡erents V18 and J10

amplitudes for the two a¡erents in linear units of accel-

vary little or inconsistently with stimulus level (mean

eration. The reasonably good ¢t to the J10 ¢lter func-

of 1.6³ þ 0.76 S.E.M. per dB, and 0.14³ þ 0.41 S.E.M.

tion indicates that the inputs to J10 are nearly equiva-

2

(the

in

HEARES 2903 28-11-97

¢lter

Fig.

2 0:5

function

6

was

of

H20)

computed

and

as

a

20

second

input

log10 (((AH20 )/

where AH20 and AV18 are ¢lter function

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

241

Fig. 5. Phase functions of frequency for the three a¡erents shown in Fig. 4 studied at four overall levels in a 15 dB range. These functions illustrate that phase is independent of level for a¡erent J10, but is systematically dependent on level for a¡erent H20. For a¡erent V18, level has a relatively small and inconsistent e¡ect on phase.

lent to a weighted sum of the inputs to H20 and V18,

E2 (Fig. 2), and H20 (Fig. 4) have features that occur

with H20 contributing about one third the in£uence of

following spikes (i.e. to the right of the spike time line),

V18. Panel B shows that another a¡erent having two

indicating that these stimulus features could not have

peaks in the ¢lter function (E2) can also be ¢t with a

caused the spikes, but appear to have been correlated

weighted sum of H20 and V18. In this case, the relative

with those that did. These non-causal revcor features

contribution of the H20 response is greater.

were

In general, the 119 saccular a¡erents recorded had

evident

only

for

a

subset

of

a¡erents

tuned

at

the lowest frequencies. These e¡ects may have arisen

revcors and ¢lter shapes similar to those illustrated in

from

Figs. 1^6. These data show that the saccule of the toad-

o¡ the axis at which the accelerometer selected for the

¢sh encodes the waveform of acoustic particle acceler-

revcor analysis was most sensitive. The three noise stim-

ation between

50 and 250 Hz, but that this range is

uli were synthesized to produce a £at acceleration spec-

divided among frequency-selective a¡erents having CFs

trum along one of the three orthogonal accelerometer

at about 74 Hz, 140 Hz, or both, in variably weighted

axes (up-down, side-side, or front-back). However, cor-

combinations.

related motion along the remaining two axes, although

6

an

a¡erent's

response

to

motion

that

occurred

lower in amplitude, could not be controlled in amplitude or spectral shape and may have played a role in causing spikes in some cases. A¡erents E1 and E2, for

4. Discussion

example, were nearly as sensitive to front-back motion

4.1. Limitations of the revcor method for directional stimuli and responses

as they were to up-down motion (see Fig. 3D). It is possible motion

The revcor functions of a¡erents H2, H18, E1 and

that

correlated

contributed

to

but the

uncontrolled response,

and

front-back thus

that

the vertical accelerometer recording used to form the

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

242

¢lter functions. Toad¢sh (a `hearing generalist') share peripheral

frequency

analysis

as

a

sound

processing

strategy with gold¢sh (a `hearing specialist') (Furukawa and Ishii, 1967 ; dence

for

Fay, 1996). Direct and Indirect evi-

peripheral

frequency

analysis

has

been

re-

ported for several other taxonomically diverse ¢sh species including two mormyrid species (Crawford, 1993 ; McCormick and Popper, 1984), the cod (Hawkins and

Osteo-

Chapman, 1975 ; Horner et al., 1981), arawana (

glossum bicirrhosum ; the

Hawaiian

Coombs and Popper, 1981) and

Adioryx xantherythrus ;

squirrel¢sh

(

Coombs and Popper, 1981). Thus, peripheral frequency analysis

may

Although

the

be sort

common of

among

peripheral

teleost

frequency

¢shes.

selectivity

demonstrated for teleost ¢shes is crude and is the simplest yet observed among vertebrate animals, its existence

demonstrates

itself

is

species

a

that

characteristic

of

¢shes,

peripheral shared

amphibians

frequency

among (Lewis,

at

analysis

least

1992),

some

reptiles

(Ko ë ppl and Manley, 1992), birds (Manley and Gleich, 1992), and mammals (Fay, 1992). Since no comparable data are available for non-teleost ¢shes (e.g. superclass Agnatha, and class Chondrichthyes), it is not yet clear whether peripheral frequency analysis is characteristic of all vertebrates. It should be noted, however, that both gold¢sh and toad¢sh di¡er from species of other vertebrate classes studied in having a small number of di¡erently tuned channels (two to three), and questions arise about the utility of such a simple system compared with the continuously variable tuning observed in the auditory nerve of

anuran

amphibians

(from

the

amphibian

papilla),

reptiles, birds, and mammals. Many cells of the audiFig. 6. Complex ¢lter functions having two peaks can be modeled as the sum of two simple ¢lter functions, one with a peak at 74 Hz, and the other with a peak at 140 Hz. In A, a¡erent J10 (light solid line) has a two-peaked ¢lter function. This can be matched by the

tory midbrain in gold¢sh show sharper tuning and a more continuous distribution of CFs than observed in primary a¡erents (Lu and Fay, 1993). Sharpening and

weighted sum (dotted line, see text) of the ¢lter functions for H20

the dispersion in CF is created by inhibitory interac-

and

similarly

tions observable at the level of the midbrain (Lu and

matched by a di¡erently weighted sum of the same a¡erents' ¢lter

Fay, 1996). For the gold¢sh, then, a set of auditory

V18

(heavy

solid

lines).

In

B,

a¡erent

E2

can

be

functions.

¢lters with continuously distributed CF is synthesized in the brain using only two or three di¡erently tuned

revcors for E1 and E2 did not represent all of the causal

peripheral inputs. We suggest that similar processing

stimulus information. This problem could possibly be

occurs in the toad¢sh and other ¢shes as well. Indirect

solved by restricting stimulus motion to a single axis

evidence that such processing occurs in hearing general-

congruent with the most excitatory axis of each a¡erent

ists comes from the demonstration in cod (

studied and then orienting the recording accelerometer

rhua)

to measure the acceleration waveform along this axis.

auditory ¢lters de¢ned in behavioral experiments (Haw-

However, this was not feasible in the present experi-

kins

ment.

quantitative

of

and

sharply

tuned

Chapman, data

on

and

1975). tuning

continuously

Although in

Gadus modistributed

there

saccular

are

little

a¡erents

of

the cod (see Horner et al., 1981), they appear to resem-

4.2. Peripheral frequency analysis

ble those reported here for the toad¢sh and thus cannot explain behavioral frequency analysis without further

The present results show that peripheral frequency

neural computation. By analogy with the central neuro-

and is

physiological data for gold¢sh (Lu and Fay, 1993, 1995,

transmitted to the ascending auditory pathway among

1996), we expect that central studies on toad¢sh will

saccular a¡erents having a diverse array of frequency

reveal similar sharpening of frequency selectivity and

analysis

occurs

in

the

saccule

of

the

toad¢sh

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

243

rotated around the 400 Hz point by 15 dB/octave and plotted as a dotted line that overlies the low-CF ¢lter. This rotation reproduces the low-CF ¢lter accurately between about 170 and 1100 Hz. One

interpretation

of

this

correspondence

is

that

low-CF a¡erents innervate hair cells responding to otolith displacement while high-CF a¡erents contact hair cells responding to otolith acceleration. These di¡erences could arise from di¡erences between hair cells in hair bundle sti¡ness and the friction of coupling between

the

hair

bundle

and

the

otolithic

membrane

(Rogers and Cox, 1988). It has been noted that a similar spectral tilt operates in the mechanosensory lateral line system to give the acceleration-coupled canal neuromasts a response to higher frequencies of hydrodynamic £ow than the velocity-coupled super¢cial neuromasts (Denton and Gray, 1983). Fig. 7B illustrates that the low- and high-CF ¢lter functions for toad¢sh saccular a¡erents are related in a similar way. In this case, the averaged high-CF ¢lter function was rotated by

3

24 dB per octave around a

pivot point of about 100 Hz, yielding the dotted line. Although we cannot presently interpret the degree of spectral tilt required for a match, it is intriguing that in both gold¢sh and toad¢sh, one of the a¡erent ¢lter functions seems to be equivalent to the other with the addition of a simple spectral tilt. Thus, the combination Fig. 7. A : Averaged ¢lter functions for the two major categories of saccular a¡erents in the gold¢sh (thick and thin solid lines) and the average

for

dB/octave

the

tilt

high-frequency

around

a

pivot

a¡erents point

of

(dotted about

line)

400

after

Hz

a

(from

3

of a single hair cell resonance (see below) and micromechanical factors such as the sti¡ness and friction of

15

coupling between hair cell cilia and the otolith appar-

Fay,

ently can lead to a crude peripheral frequency analysis

1996). B : A similar comparison of the two categories of saccular af-

that can be enhanced through central computation.

ferents in the toad¢sh (from Fig. 3D). For the toad¢sh, the highfrequency ¢lter function (thick solid line) matches the low-frequency function (thin solid line) best after a pivot

point

of

about

100

Hz

3

(dotted

24 dB/octave tilt around a

line).

Although

the

4.3. Origin of a¡erent frequency selectivity

gold¢sh

hears in a wide frequency range compared with toad¢sh, the hearing range is similarly divided between two a¡erent types in both species.

The

origin

of

frequency

selectivity

in

the

toad¢sh

saccule may arise, at least in part, from hair cell resonance. In studies on isolated hair cells of the toad¢sh

dispersion of the CF distribution. We hypothesize that

saccule, Steinacker and Romero (1992) observed elec-

like the gold¢sh (e.g. Fay et al., 1978), cod (Hawkins

trical resonances in response to current steps, ranging in

and

(re-

frequency between 107 and 175 Hz with a mean of 142

viewed by Fay, 1988), toad¢sh are able to use this fre-

Hz. It therefore seems likely that the ¢lter functions

quency analysis in behavioral detection, masking, and

with peaks at 140 Hz observed among primary a¡erents

discrimination tasks. These new neurophysiological and

in the present study (Figs. 2^4 and 6) are the result of

behavioral experiments are in the planning stage.

inputs from these resonant hair cells. Steinacker and

Chapman,

1975),

and

several

other

species

be classi¢ed

Romero (1992) also observed saccular hair cells that

into two CF groups based on ¢lter functions derived

exhibited long-duration spikes in response to current

from revcor analysis (Fay, 1996). The low-CF a¡erents

steps, rather than resonance. These hair cells may con-

have a CF near 170 Hz and the high-CF a¡erents show

tribute to the broad, low-CF tuning (74 Hz) in many

a peak near 600 Hz. As illustrated in Fig. 7, the aver-

a¡erents of the present study. However, it is not yet

aged

are

clear how hair cell spikes would be expected to be rep-

related to one another in a simple way. The low-CF

resented in the acoustic response of a¡erents in vivo.

¢lter function is nearly identical to the high-CF func-

A¡erents

tion after it has been spectrally `tilted' by

15 dB per

and 140 Hz : J10 in Figs. 4 and 6 ; E1 and E2 in Figs.

octave, pivoting around the ¢lter amplitude at about

2, 3 and 6 ; and several other examples in Fig. 2, panel

400

C) may innervate both resonant hair cells and an addi-

Saccular

¢lter

Hz.

a¡erents of

functions

To

the gold¢sh can

for these

illustrate

this,

the

two a¡erent types

high-CF

3

average

was

HEARES 2903 28-11-97

showing

two-peaked

¢lter

functions

(at

74

244

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

tional class of hair cells. Individual saccular a¡erents of

monotonically increasing phase lags as a function of

the toad¢sh have been shown to innervate up to at least

frequency, and little indication of the asymptotes ex-

100 hair cells (Edds-Walton et al., 1996).

pected from second-order resonant systems. In this re-

While it appears that hair cell resonance plays a role

spect, the phase functions of toad¢sh saccular a¡erents

in the tuning of a¡erents, simple second-order reso-

resemble those presented by Lewis (1992) for a¡erents

nance alone cannot explain the ¢lter shapes we have

from a variety of auditory organs having resonant hair

observed. As Lewis (1992) has discussed, second-order

cells (in an amphibian, reptile, and a mammal), illus-

resonances typical of hair cells result in ¢lter functions

trating the high dynamic order of these systems. For

having decelerating rollo¡ slopes above and below the

linear ¢lters, the range of phase shift can be used to

center frequency, and in impulse responses having tem-

characterize the ¢lter. For the present results, however,

porally asymmetric envelopes that rise rapidly and de-

the relatively small number of spikes used to form the

cay more slowly. Yet, as Lewis (1992) has pointed out,

revcors constrains the dynamic range of the analysis to

primary a¡erents from hearing organs with resonant

about 20 dB (Evans, 1989), constraining, in turn, the

hair cells tend to have ¢lter functions with accelerating

frequency range over which the phase functions can be

slopes above and below the center frequency (i.e. have

reliably interpreted. In addition, unknown time delays

convex ¢lter skirts), and revcor envelopes that tend to

between the recorded acceleration and the a¡erent re-

rise and decay at similar rates (i.e. tend to show tem-

sponse probably contribute to the slopes of the phase

poral symmetry). In the present results, most a¡erents

functions. Thus, the phase functions of the present re-

do not show the hallmarks of second-order resonance :

sults cannot be clearly interpreted to further reveal

They tend to have convex ¢lter skirts (e.g. solid-line

quantitative characteristics of the intervening ¢lters.

¢lter functions of Fig. 2) and substantially symmetric

Lewis (1992) has suggested that electrical elements of

revcor envelopes (e.g. revcors for the solid-line ¢lter

the variously tuned hair cells could interact with each

functions of Fig. 2). Thus, most saccular a¡erents of

other through a mechanical linkage if the hair cells were

the toad¢sh behave similarly to auditory a¡erents of

bidirectional transducers (mechanical to electrical, and

amphibians, reptiles, and mammals, and saccular a¡er-

the reverse). Such an interaction could produce a re-

ents of the bullfrog (Lewis, 1992). However, the present

sponse in which the electrical resonances of hair cells

results also show that some of the high-CF saccular

were absorbed into the complex dynamics of the entire

a¡erents show responses more consistent with a sec-

system. So far, there are no data for teleosts indicating

ond-order resonance, including clearly asymmetric rev-

that the hair cells are bidirectional transducers. How-

cor envelopes (e.g. a¡erents M10, V14, and V18 in Fig.

ever, Lewis (1992) cites direct and indirect evidence that

2) and the suggestion of decelerating slopes of the high-

hair cells of the saccule of the bullfrog (Lewis, 1988)

pass segments of the ¢lter functions. Even in these

and of most anamniote hearing organs operate bidirec-

cases, however, resonance alone cannot account for

tionally. Based on the present data, we hypothesize that

the ¢lter functions since the low-pass segments show

saccular hair cells of the toad¢sh may be bidirectional

accelerating slopes above the center frequency.

transducers as well.

The phase functions of Fig. 3 and 5 illustrate the

In gold¢sh, the saccule appears to be crudely tono-

di¡erences among a¡erents in their phase responses.

topically organized with high-CF a¡erents originating

For example, the functions for a¡erents E1 and E2

primarily from the rostral region (Furukawa and Ishii,

(Fig. 3) are parallel but separated by 180³, indicating

1967). So far there is no evidence that the saccule of the

an essential similarity of the ¢lters but with an opposite

toad¢sh is tonotopically organized. A previous study

polarity. The polarity di¡erence is probably due to

that crudely characterized the frequency response func-

these a¡erents receiving inputs from hair cells that are

tions of a¡erents from the rostral, middle, and caudal

oppositely oriented on the saccular epithelium

(See

regions of the saccular epithelium found no tendency

Edds-Walton and Popper, 1995). In Fig. 5, the phase

for tonotopy (Fay and Edds-Walton, 1997). A similar

functions illustrate a level dependency for the nonspon-

failure to demonstrate tonotopy was reported for the

taneous a¡erent H20, but not for the highly spontane-

cod saccule (Horner et al., 1981). Furthermore, Stein-

ous J10. This di¡erence is in accord with observations

acker and Romero (1992) found that resonant, nonres-

using single tone stimuli that phase may advance up to

onant, and spiking hair cells were isolated from all re-

90³ with increasing level only for a¡erents having very

gions

low spontaneous rates (Fay and Edds-Walton, 1997).

epithelium.

over

the

length

of

the

toad¢sh

saccular

Finally, the in£ections of the phase functions for J10 suggest inputs from at least two di¡erent ¢lters, one

4.4. The saccule and sound communication

tuned similarly to H20 (having a ¢lter function peak near 70 Hz), and the other tuned similarly to V18 (hav-

The toad¢sh communicates using at least two vocal-

ing a ¢lter peak near 140 Hz), as illustrated in Fig. 6A.

ization sounds ; the grunt (and growl consisting of mul-

The phase functions of Figs. 3 and 5 generally show

tiple grunts) and the boatwhistle. The behavioral con-

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

245

texts for these sounds are well known (e.g. Gray and

frequency tuning di¡erences may be due, in part, to

Winn,

resonances shown for isolated saccular hair cells and

1961 ;

Fish,

1972).

The

boatwhistle

is

a

loud,

long-duration sound consisting of a fundamental fre-

to

quency between 130 and 250 Hz and successive higher

their

micromechanical

harmonics. Females approach males that emit the boat-

This sort of frequency analysis has been suggested as

whistle call from at least 2 m away (Fish, 1972), and

the most primitive basis for peripheral frequency anal-

males avoid the territories of vocalizing males. Watkins

ysis yet indicated for vertebrates (Fay, 1996). The over-

(1967) reported that the fundamental frequency of the

all frequency range is adequate to encode the funda-

boatwhistle is 130 Hz in the Woods Hole, MA region.

mental frequency of the `boatwhistle' vocalization, but

This fundamental frequency is very close to the high-

is not wide enough to encode the amplitudes and phases

frequency CFs observed in the present experiment (140

of the higher harmonics of this call. Thus, we conclude

Hz). Thus, the fundamental frequency component of

that the fundamentally important aspect of this court-

the boatwhistle call is well represented in primary a¡er-

ship vocalization is its pulse repetition rate, or the fun-

ents, but the higher harmonics are not. The most robust

damental frequency. In the context of studies on a wide

neural code for the boatwhistle is a representation of

variety of vertebrate species (Webster et al., 1992), these

the fundamental frequency or pulse repetition rate.

results suggest that peripheral frequency analysis may

ciliary

processes

attachments

to

local

the

to

hair

otolithic

cells

and

membrane.

Our previous work has shown that the most sensitive

be a common characteristic among vertebrates. Bony

saccular a¡erents respond to displacements of 0.1 nm

¢shes were probably the ¢rst vertebrates to solve prob-

(rms)

lems

at

100

Hz

(Fay

and

Edds-Walton,

1997).

The

of

frequency

analysis,

and

these

solutions

may

behavioral audiogram for the toad¢sh (Fish and O¡utt,

have formed a model for those of their tetrapod de-

1972) shows sound pressure thresholds to frequencies

scendants.

WPa. In the acous(usually greater than 1/2Z

below 150 Hz of about 100 dB re : 1 tic far¢eld of a sound source

wavelengths from the source (Rogers and Cox, 1988)),

Acknowledgments

acoustic particle motion at the threshold for toad¢sh Thus,

This research was primarily supported by a Program

these results indicate that the normal mode of acoustic

Project Grant from the NIH, NIDCD to Washington

communication for the toad¢sh probably involves the

University

direct

particle

came from a Program Project Grant from the NIH,

motion by the saccule's response to particle accelera-

NIDCD to the Parmly Hearing Institute, Loyola Uni-

tion.

versity Chicago, and from an ONR Grant (N00014-94-

hearing

(at

100

detection

Hz)

of

is

approximately

the

boatwhistle's

0.1

nm.

acoustic

School

of

Medicine.

Additional

support

Finally, we wish to make a comment about the only

10410) to the University of Maryland. Thanks to Steph-

other report in the literature on the frequency response

en Highstein for his instruction in surgery and electro-

of saccular a¡erents of toad¢sh. Fine (1981) reported a

physiology. Thanks also to Bill Yost for his support.

`mismatch'

compo-

The software used for calibration, data acquisition, and

nents of the boatwhistle call and the best frequencies

analysis was written by R. Fay while he was supported

of

frequencies

on a sabbatical leave by Loyola University Chicago.

were too low. The present results indicate a much better

The shaker system used was originally constructed for

`match'. The reason for the discrepancy is that Fine

research supported by the NSF.

saccular

between

the

a¡erents ;

dominant

the

frequency

a¡erent's

best

assumed that the toad¢sh auditory system responded in

proportion

to

sound

pressure.

We

have

assumed

that acoustic particle acceleration is the adequate stimulus. We have shown that the saccule responds directly to

acoustic

particle

motion

with

su¤cient

sensitivity

and bandwidth to account for the detection and processing of the boatwhistle call. As yet, there is no evidence that sound pressure is an adequate stimulus for

References de Boer, E., de Jongh, H.R., 1978. On cochlear coding : Potentialities and limitations of the reverse correlation technique. J. Acoust. Soc. Am. 63, 115^135. Chapman, C.J., Sand, O., 1974. Field studies of hearing in two species of £at¢sh,

the toad¢sh auditory system.

Pleuronectes platessa

(L.) and

Limanda limanda

(L.)

(family Pleuronectidae). Comp. Biochem. Physiol. 47, 371^385. Coombs, S.L., Popper, A.N., 1981. Comparative frequency selectivity

4.5. Summary and conclusions

in ¢shes : Simultaneously and forward-masked psychophysical tuning curves. J. Acoust. Soc. Am. 71, 133^141.

Saccular a¡erents of the toad¢sh are frequency selective with highest e¡ective frequencies ranging to about 250

Hz,

and

with

individual

a¡erents

dividing

this

range among frequency-selective a¡erents having ¢lter function peaks at about 74 Hz, 140 Hz, or both. These

Crawford, J.D., 1993. Central auditory neurophysiology of a soundproducing mormyrid ¢sh ; the mesencephalon of

Pollimyrus isidori.

J. Comp. Physiol. A 172, 1^14. Denton, E.J., Gray, J.A.B., 1983. Mechanical factors in the excitation of clupeid lateral lines. Proc. R. Soc. 218, 1^26. Edds-Walton, P.L., Popper, A.N., 1995. Hair cell orientation patterns

HEARES 2903 28-11-97

R.R. Fay, P.L. Edds-Walton / Hearing Research 113 (1997) 235^246

246

on the saccules of juvenile and adult toad¢sh,

Opsanus tau.

Acta

Zool. 76, 257^265.

Horner, K., Hawkins, A.D., Fraser, P.J., 1981. Frequency characteristics of primary auditory neurons from the ear of the cod,

Edds-Walton, P.L., Highstein, S.M., Fay, R.R., 1996. Saccular innervation patterns in physiologically characterized primary a¡erents of the toad¢sh. Soc. Neurosci. Abst., 178.4.

morhua.

Gadus

In : Tavolga, W.N., Popper, A.N., Fay, R.R. (Eds.),

Hearing and Sound Communication in Fishes. Springer-Verlag, New York, pp. 223^241.

Evans, E.F., 1989. Cochlear ¢ltering : a view seen through the tempo-

Ko ë ppl, C., Manley, G., 1992. Functional consequences of morpho-

ral discharge patterns of single cochlear nerve ¢bers. In : Wilson,

logical trends in the evolution of lizard hearing organs. In : Web-

J.P., Kemp, D.T. (Eds.) Cochlear Mechanisms. Plenum Press,

ster, D.B., Fay, R.R., Popper, A.N. (Eds.), The Evolutionary Bi-

New York, pp. 341^414.

ology of Hearing. Springer-Verlag, New York, pp. 489^510.

Fay, R.R., 1984. The gold¢sh ear codes the axis of particle motion in three dimensions. Science 225, 951^953.

447.

Fay, R.R., 1988. Hearing in Vertebrates : A Psychophysics Databook. Hill-Fay Associates, Winnetka, IL. among vertebrates. In : Webster, D., Fay, R., Popper, A. (Eds.), Evolutionary

Biology

of

Hearing.

Springer-Verlag,

New

York, pp. 229^263. sounds by the gold¢sh (Carassius

auratus). Hear. Res. 89, 146^154.

Fay, R.R., 1996. Frequency selectivity of saccular a¡erents of the revealed

Evolutionary Biology of Hearing. Springer-Verlag, New York, pp. 163^184. Lu, Z., Fay, R.R., 1993. Acoustic response properties of single units in the torus semicircularis of the gold¢sh,

Fay, R.R., 1995. Perception of spectrally and temporally complex

gold¢sh

Lewis, E., 1992. Convergence and design in vertebrate acoustic sensors. In : Webster, D.B., Fay, R.R., Popper, A.N. (Eds.), The

Fay, R.R., 1992. Structure and function in sound discrimination The

Lewis, E., 1988. Tuning in the bullfrog ear. Biophys. J. 53, 441-

by

revcor

analysis.

In :

Lewis,

E.R.,

Long,

G.R., Lyon, R.F., Narins, P.M., Steele, C.R., Hecht-Poinar, E.

Lu, Z., Fay, R.R., 1995. Acoustic response properties of single units of the central posterior nucleus of the thalamus in gold¢sh (Ca-

rassius auratus).

J. Comp. Physiol. 176, 747^760.

Lu, Z., Fay, R.R., 1996. Two-tone interaction in auditory nerve ¢bers

(Eds.), Diversity in Auditory Mechanics. World Scienti¢c Publish-

and midbrain neurons of the gold¢sh,

ers, Singapore, pp. 69^75.

Neurosci. 2, 257^273.

Fay, R.R., Ahroon, W.A., Orawski, A.T., 1978. Auditory masking

Carassius auratus.

J. Comp. Physiol 173, 33^48.

Carassius auratus.

Audit.

Lu, Z., Popper, A., Fay, R.R., 1996. Behavioral detection of acoustic

patterns in the gold¢sh : Psychophysical tuning curves. J. Exp.

particle motion by a teleost ¢sh (Astronotus

Biol. 74, 83^100.

and directionality. J. Comp. Physiol. 179, 227^234.

ocellatus) :

sensitivity

Fay, R.R., Chronopolous, M., Patterson, R., 1996. The sound of a

Manley, G., Gleich, O., 1992. Evolution and specialization of function

sinusoid : Perception and neural representations in the gold¢sh

in the avian auditory periphery. In : Webster, D.B., Fay, R.R.,

(Carassius

Popper,

auratus).

Audit. Neurosci. 2, 377^392.

Fay, R.R., Edds-Walton, P.L., 1997. Directional sensitivity of saccular a¡erents of the toad¢sh,

Opsanus tau.

Hearing Research (in

press).

A.N.

(Eds.),

The

Evolutionary

Biology

Hearing.

McCormick, C.A., Popper, A.N., 1984. Auditory sensitivity and psychophysical tuning curves in the elephant nose ¢sh,

Fine, M., 1981. Mismatch between sound production and hearing in

of

Springer-Verlag, New York, pp. 561^580.

petersii.

Gnathonemus

J. Comp. Physiol. 155, 753^761.

the oyster toad¢sh. In : Tavolga, W.N., Popper, A.N., Fay, R.R.

Popper, A.N., Fay, R.R., 1993. Sound detection and processing by

(Eds.), Hearing and Sound Communication in Fishes. Springer-

¢sh : Critical review and major research questions. Brain Behav.

Verlag, New York, pp. 257^263.

Evol. 41, 14^38.

Fish, J.F., 1972. The e¡ect of sound playback on the toad¢sh. In :

Rogers, P., Cox, M., 1988. Underwater sound as a biological stimu-

Winn, H.E., Olla, B.L. (Eds.), Behavior of Marine Animals, Vol.

lus. In : Atema, J., Fay, R.R., Popper, A.N., Tavolga, W.N.

2. Plenum, New York, pp. 386^434.

(Eds.), Sensory Biology of Aquatic Animals. Springer-Verlag,

Fish, J.F., O¡utt, G., 1972. Hearing thresholds from toad¢sh,

nus tau,

Opsa-

measured in the laboratory and ¢eld. J. Acoust. Soc. Am.

51, 1314^1321.

New York, pp. 131^149. Sand, O., Karlsen, H.E., 1986. Detection of infrasound by the Atlantic cod. J. Exp. Biol. 125, 197^204.

ë ber die Bedeutung des Sacculus und der Lavon Frisch, K., 1938. U

Steinacker, A., Romero, A., 1992. Voltage-gated potassium current

gena fu ë r den Gehorsinn der Fische. Z. Vergl. Physiol. 25, 703^747.

and resonance in toad¢sh saccular hair cells. Brain Res. 574,

Furukawa, T., Ishii, Y., 1967. Neurophysiological studies on hearing in gold¢sh. J. Neurophysiol. 30, 1377^1403. Gray, G.A., Winn, H.E., 1961. Reproductive ecology and sound production of the toad¢sh,

Opsanus tau.

Ecology 42, 274^282.

Hawkins, A.D., Chapman, C.J., 1975. Masked auditory thresholds in the cod,

Gadus morhua

229^236. Watkins, W., 1967. Spectral analysis of pulse trains. In Tavolga, W.N.

L. . J. Comp. Physiol. A 103, 209^226.

(Ed.) Marine Bio-Acoustics, Vol. 2. Pergamon Press, Oxford, pp. 15-44. Webster, D.B., Fay, R.R., Popper, A.N. (Eds.) 1992. The Evolutionary Biology of Hearing. Springer-Verlag, New York.

HEARES 2903 28-11-97