J. eomp. Physiol. 103, 209 226 (1975) 9 by Springer-Verlag 1975

Masked Auditory Thresholds in the Cod, Gadus morhua L. A. D. H a w M n s a n d C. J. C h a p m a n Marine Laboratory, Aberdeen, Scotland Received May 26, 1975

Summary. Auditory thresholds were determined for cod (Gadus morhua) using a cardiac conditioning technique. Pure tone stimuli were presented against a background of high level white noise. As the bandwidth of the masking noise was reduced the threshold:noise ratio remained approximately constant until a critical bandwidth was reached, when the threshold:noise ratio declined. The width of the critical band was investigated by experiments where pure tone thresholds were determined in the presence of narrow bands of noise centred at different frequencies. Thresholds declined with increasing frequency separation between the stimulus and noise band. The masking function obtained can be described in terms of an equivalent rectangular filter of finite width. Values for the critical bandwidth derived in this way show a general increase with stimulus frequency, ranging from 59.0 Hz wide at 40 Hz, to 165 I-Iz wide ag 380 Hz. These values are close to calculated values (critical ratios) determined from the results of earlier experiments, where pure tone stimuli were masked by broad band noise. Introduction W h e n t h e d e t e c t i o n of one sound is i m p a i r e d in t h e presence of a n o t h e r t h e former is said to be m a s k e d b y t h e latter. T a v o l g a (1967) f i r s t d e m o n s t r a t e d t h a t a m b i e n t noise could m a s k t h e d e t e c t i o n of sounds b y fish. Two m a r i n e species m a i n t a i n e d in a q u a r i u m t a n k s showed changes in t h e i r d e t e c t i o n thresholds for p u r e t o n e signals when s u b j e c t e d to changes in t h e level of b a c k g r o u n d noise. B u e r k l e (1968) l a t e r confirmed t h a t t h e hearing thresholds of cod were influenced b y t h e p r e v a i l i n g level of noise in t h e a q u a r i u m . Because noise levels in l a b o r a t o r y t a n k s are often m u c h higher t h a n noise levels in t h e sea, these e a r l y e x p e r i m e n t s d i d n o t establish w h e t h e r s o u n d stimuli are m a s k e d b y b a c k g r o u n d noise u n d e r n a t u r a l conditions. Our e x p e r i m e n t s ( C h a p m a n a n d Hawkins, 1973; C h a p m a n , 1973) c o n d u c t e d o n cod a n d o t h e r m a r i n e teleosts a t a m i d w a t e r location in a sheltered sea loch h a v e confirmed t h a t m a s k i n g can occur even u n d e r r e l a t i v e l y quiet sea conditions. This suggests t h a t for m o s t m a r i n e fish t h e a b s o l u t e sens i t i v i t y of t h e a u d i t o r y s y s t e m is less i m p o r t a n t t h a n its a b i l i t y to d i s c r i m i n a t e b e t w e e n a s o u n d stimulus a n d t h e b a c k g r o u n d noise. Masking implies an i n a b i l i t y to s e p a r a t e signal a n d noise. I-Iowever, n o t all f r e q u e n c y c o m p o n e n t s of t h e noise necessarily p r o m o t e masking. F o r h u m a n subjects i t has long been k n o w n t h a t a p u r e t o n e signal is m a s k e d m o s t effectively b y noise c o m p o n e n t s a t t h e s a m e a n d i m m e d i a t e l y a d j a c e n t frequencies. F l e t c h e r (1940) a p p l i e d t h e t e r m " c r i t i c a l b a n d " to t h e f r e q u e n c y s p a n of noise which is effective. H e i n t r o d u c e d t h e a n a l o g y of a n a u d i t o r y filter, which can be t u n e d to t h e f r e q u e n c y of t h e stimulus a n d effectively eliminates r e m o t e frequencies.

210

A.D. Hawkins and C. J. Chapman

Tavolga, in his experiments (lot. tit), a t t e m p t e d to demonstrate the existence of a critical band for masking in fish. His approach was to determine pure tone thresholds in the presence of bands of noise of different width, centred at the same frequency. I n such an experiment, the thresholds should remain constant for noise bands wider than the critical band, but decline significantly for narrower bands where the total noise energy integrated over the span of the auditory filter should be less. I n practice, varying the width of the noise band did not produce a difference in threshold. Tavolga nevertheless concluded t h a t a critical band for masking did exist and suggested t h a t its width was narrower t h a n the narrowest noise bandwidth employed. Later, Buerkle (1969) performed a series of rather different experiments upon cod. Thresholds were determined for signals consisting of bands of noise, centred at various frequencies, in the presence of a stationary band of noise. The signals were masked most effectively when they were centred at the same frequency as the noise band, and masking diminished as the frequency separation between signal and noise increased. More recently, Tavolga (1974) has repeated his earlier experiments, using goldfish as subjects. The fish were presented with pure tones in the presence of noise of varying bandwidth. I n this case there was a clear decline in threshold as the noise band was narrowed beyond a certain width, providing strong evidence of a critical band for masking in fish. I n mammals and birds the critical band has been interpreted in terms of the mechanical response of the cochlea of the inner ear. Von B@k6sy (1960) demonstrated t h a t a sinusoidal signal sets up a travelling wave along the cochlear partition and t h a t there is a point along the basilar membrane where movement is maximal for each particular frequency. The subjective sensation of frequency, or pitch, is held to depend on the position of the maximum, and with stimulation of a localised group of the sensory cells distributed along the membrane. However, it is evident t h a t the motion generated b y a pure tone stimulus does not occur at a single locus, but is spread over a much wider area. The critical band has been associated with the distance along the basilar membrane over which the neural response is integrated (Zwicker, Flottorp and Stevens, 1957). On this basis, components of the background noise which are closely related in frequency to a particular pure tone signal will stimulate the same part of the basilar membrane, and will therefore be indistinguishable from the signal. This explanation is not entirely adequate. 1Yiost psychophysical measurements of the critical band have indicated a rather narrower bandwidth than can be accounted for solely in terms of the mechanical response of the cochlea. Both central and peripheral neural sharpening mechanisms have been postulated to account for this discrepancy (Keidel, 1970). I n seeking a morphological correlate of an auditory filter within the fish ear, we are faced with the difficulty t h a t there is no obvious form of mechanical frequency analyser analogous to the cochlea. I t is therefore possible t h a t the filtering is performed neurally, rather t h a n mechanically. However, before attempting to elucidate the physical processes involved it is desirable to re-affirm t h a t frequency selective masking does occur, and to determine the characteristics of any auditory filtering mechanism employed. I n this paper we present further observations on the hearing of cod in the presence of masking noise of differing character.

Masked Auditory Thresholds in

Gadus

211

~Iethods Pure tone hearing thresholds were determined for cod in the presence of high level background noise. The thresholds were obtained using a classical conditioning m e t h o d in which a change in cardiac r h y t h m was established in response to the pure tone (the conditioned stimulus) b y t h e delayed administration of a mild DC electric shock (the unconditioned stimulus). The response consisted of a n inhibition of the h e a r t rate, occurring after onset of the sound b u t before receipt of the shock. Once this response was firmly established (generally within the first i 0 trials) a threshold was obtained b y gradually lowering a n d raising the stimulus level in 3 dB steps (see Fig. 1). The experiments were performed on 12 immature cod (length 27-42 era) caught at depths of less t h a n 10 m in Upper Loch Torridon, on the west coast of Scotland. A n electro-cardiograph electrode was inserted into each fish under MS 222 anaesthetic, a n d the animal allowed to recover before being transferred to a cylindrical plastic mesh cage incorporating a pair of stainless steel shock electrodes. The cage and fish were t h e n t a k e n b y diver to a n underwater acoustic range. The cage was mounted on top of a water-filled plastic piping framework standing 6 m above the seabed a t a d e p t h of 20 m. Sound stimuli were t r a n s m i t t e d through a single projector (Dyna-Empire type J9) moored 2 m from the fish and at the same depth. The stimuli were monitored on a calibrated hydrophone positioned b e n e a t h the head of the animal. Cables from the underwater site connected the electrodes and other apparatus to a shore laboratory. During each masking experiment noise was continually t r a n s m i t t e d from the sound projector. Bands of noise of different width, centred a t different frequencies were derived from a sine-random generator (Bruel and Kjaer, type 1024) used in conjunction with a passive band-pass filter (Dawe, type 1462) a n d attenuator. The pure tone stimulus was presented as a short pulse, lasting 8 sec, with a rise time of 300 msec and fall time of 600 msec. I t was generated b y a beat frequency oscillator (Bruel and Kjaer, type 1022) feeding a transient-free gating circuit and attenuator. The frequency of the tone was monitored on a digital frequency meter. The tone pulse was superimposed upon the continuous noise b y a n adding circuit inserted before common power amplifier a n d sound projector. A t the end of the gate period, a relay was operated feeding a 12 v DC pulse of 0.2 see duration to the shock electrodes on the fish cage. The characteristics of the noise were determined b y frequency anMysis of the projected sound. The hydrophone signal was fed to a narrow b a n d frequency analyser a n d logarithmic level recorder (Bruel and Kjaer, types 2107 a n d 2305). Sound levels are expressed in the paper as decibels relative to a reference sound pressure of 1 mierobar 1 (i.e. dB/~bar). The width of the noise bands are given for the 3 dB down points. The shape factors for the noise bands varied from 20 dB per octave for the narrowest band, to 60 dB per octave for the widest band. A more detailed account of the acoustic range, the conditioning procedure, and the apparatus employed is given b y Chapman and Hawkins (1973). The masking effect of the background noise upon pure tone stimuli was studied in two ways :

1. Band-narrowing Experiments Each fish was first allowed to adjust to the experimental situation (indicated b y a slowing of the heart-beat from a post-operative rate of approximately 1 beat per sec to a rate of 1 beat every 1.4-2.2 see). Once adjusted, it was conditioned to a pure tone stimulus against a background of a m b i e n t sea noise and a threshold determined. Then, broad b a n d random noise (bandwidth 30-1000 Hz) was t r a n s m i t t e d through the sound projector at a spectrum level 20-30 dB above the spectrum level of ambient sea noise, and a new pure tone threshold determined. Subsequently, thresholds were re-determined as the bandwidth of the noise was progressively reduced, while maintaining the spectrum level of noise within the band-pass constant. Three pure tone frequencies were tested (60, 160 and 380 l~z). I n each case the noise b a n d centre frequency was the same as the tone frequency. The results were expressed in terms of the ratio in dB between the sound pressure threshold for the tone, a n d the spectrum level of noise a t the tone frequency (the threshold:noise ratio). A similar band-narrowing 1 1 [zbar ~ 0.1 Newton/m 2.

212

A.D. Hawkins and C. J. Chapman

technique has been applied to human subjects by Hamilton (1957) and to fish by Tavolga (1967, 1974).

2. Band-shi/ting Experiments Initially, we intended to examine the masking of one pure tone by another, as described for human subjects by Wegel and Lane (1924). After determining the threshold to a pure tone against sea noise the threshold was re-determined in the presence of a series of tones of different frequency at a fixed level of -~ 10 dB/tzbar (3040 dB above the spectrum level of sea noise). However, when using this method it became clear that signal detection--at least for the human observer--was enhanced by the presence of "beats" when the frequency of the masking tone was close to that of the signal. After one series of experiments at 160 Hz this method was therefore abandoned in favour of a technique which employed a narrow band of masking noise (10 Hz wide). Thus after determining the threshold for a pure tone against sea noise the threshold was re-determined in the presence of a successive series of narrow bands of noise. At first, the noise was centred at a frequency remote from that of the tone. Later it was progressively moved closer to the tone frequency. The overall level of noise was maintained constant at ~ 10 dB/~zbar (i.e. the spectrum level within the noise band was 0 dB/Izbar/Hz). The results were expressed in terms of the difference (in decibels) between the threshold at a given noise band centre frequency and the threshold obtained when the noise band centre frequency was tuned to the frequency of the tone. Buerkle (1969) conducted rather similar masking experiments on cod, bat used narrow bands of noise as both signal and masker, and moved the signal band whilst keeping the noise band stationary. Egan and Hake (1950), Schafer et al. (1950) and Bilger and Hirsh (1956) have employed similar methods with human subjects.

Results

1. Band-narrowing Experiments T h e thresholds o b t a i n e d from cod i n the presence of a m b i e n t sea noise were similar to those previously reported ( C h a p m a n a n d Hawkins, 1973). W h e n thresholds were d e t e r m i n e d against a high level of b r o a d - b a n d noise t h e y were prop o r t i o n a l l y higher, confirming t h a t m a s k i n g h a d occurred. This is i l l u s t r a t e d b y Fig. 1, which presents a sequence of thresholds for a t o n e of 380 Hz. As b r o a d b a n d noise was a d d e d to a m b i e n t sea noise the threshold rose to a n e w value, t h o u g h the t h r e s h o l d : n o i s e ratio r e m a i n e d constant. Then, as the b a n d w i d t h of the m a s k i n g noise was reduced a p o i n t was reached where the t h r e s h o l d : n o i s e ratio declined. T h e c o m b i n e d results of several such sequences, a t tone frequencies of 60, 160 a n d 380 Hz, are s u m m a r i s e d i n Fig. 2. I u order to minimise v a r i a b i l i t y bet w e e n i n d i v i d u a l fish the m a s k i n g effect of each b a n d (indicated b y the threshold: noise ratio) is expressed relative to the effect of broad b a n d noise. Fig. 2A shows t h a t t h e m a s k i n g of a 380 Hz tone is a p p r o x i m a t e l y the same for b a n d w i d t h s of 100 t I z or more, b u t below this it is directly proportional to the b a n d w i d t h . Thus, there can be said to be a critical m a s k i n g b a n d w i d t h for cod. I t lies b e t w e e n 30-100 Hz i n m a g n i t u d e , since the a d d i t i o n of noise outside this range does n o t cause a n increase i n masking. A t 160 Hz (Fig. 2B) t h e degree of m a s k i n g r e m a i n e d a p p r o x i m a t e l y c o n s t a n t for b a n d w i d t h s wider t h a n 50 Hz, b u t was progressively less for b a n d w i d t h s of 30 a n d 10 ttz, suggesting t h a t the critical m a s k i n g b a n d w i d t h lies b e t w e e n 30 a n d 50 Hz. A t 60 Hz (Fig. 2C) m a s k i n g was a p p r o x i m a t e l y the same for all noise b a n d widths, showing only a small decline at 10 Hz; This suggests t h a t the critical b a n d -

Masked Auditory Thresholds in Gadus +20

I'

I

~

I

I

I

~ z~ f=lO00

213 I

I

I

/, f~lO0

+10

.a ~ -lO oJ

mm -20 -

24'8 ~ o

25.3 (0)

-30

27,1 (1.8)

19 .t {-6-2]

16-1 (-9-2)

U3

-40 . . . . . . . . . . . . 1 I 0 10

I

!

20

I

30 40 Trial Number

I

I

I

I

50

50

70

80

Fig. 1. Sequence of threshold determinations (solid lines) obtained firstly in the presence of sea noise, and then against a background of masking noise of differing bandwidth. The dashed lines represent the spectrum level of noise at the frequency of the tone (380 Hz). The figure above each threshold curve denotes the noise bandwidth, while the lower figure gives the threshold:noise ratio and (in brackets) the difference in masking level relative to wide band noise (z] ] ~ 1000)

A

380Hz

B

160Hz

Ill

-4 -6

-8 -10

---12

~ -14 or >o -2 F

~

"

"

/

9

9

J" -2/,_

-6

0'

9 10

I 20

9

I 40 Noise

I 60

I

I 100

Bond - Width

1 200

I

l

830

I

J

1000

{ Hz }

Fig. 2. The relative masking effect of different noise bands upon 380, 160 and 60 Hz pure tone signals, expressed as the difference in masking level relative to wideband noise (A / ~ 1000)

214

A.D. Hawkins and C. J. Chapman

/,% -10 "10

.~. -20 /o

.~ -30 "5

-40

o--

[

l

LO

l

I

I

l

60 80 100 200 Centre Frequency of Noise Bond (Hz) Frequency of Secondary Tone (Hz)

I

I

400

600

Fig. 3. The relative masking effect of secondary tones (o.... o) and 10 Hz wide noise bands (.--.), centred at different frequencies, upon detection of a 160 Itz pure tone signal

width at this frequency is rather narrow. However, at 60 Hz a technical problem was encountered which m a y have led to misleading results. The J9 sound projector gave a reduced output at very low frequencies (below 60 Hz). Thus the spectrum within a given pass band was not uniform, but showed a decrease at low frequencies. Alterations in the noise bandwidth were therefore only effective at the upper end of the pass band, and this m a y have reduced the expected differences in the threshold:noise ratio. The results of the band narrowing experiments confirm that there is a critical bandwidth for masking. Progressive reduction of the bandwidth of a masking noise beyond a critical value reduces the threshold of a pure tone located at the centre of the band. However, it would seem that the band narrowing type of experiment is not ideal for determining the actual width of the critical band. The resultant changes in threshold:noise ratio as the bandwidth is reduced are only gradual. Greenwood (1961) has noted this for human subjects. De Boer (1962) has pointed out that by narrowing the bandwidth of the noise a higher degree of amplitude fluctuation is introduced. This may influence the degree of masking, and casts doubt upon the efficacy of band-narrowing techniques. Though actual values for the critical bandwidth cannot be given for the band narrowing experiments it does appear from Fig. 2 that the width of the band is greater for higher stimulus frequencies.

2. Band-shi/ting Experiments The masking effects of both secondary tones and 10 Hz wide noise bands were first examined at a signal frequency of 160 Hz. The results are illustrated in Fig. 3. Masking was most pronounced when the centre frequency of the noise band (or secondary tone) coincided with the signal. A small shift in the centre frequency above or below the signal resulted in a sharp decline in masking. This effect was most pronounced when secondary tones were used, rather than narrow

Masked Auditory Thresholds in Gadus

o

./"-\

O\o k/"

-20 -

0 -3o

I

o

-ot

I

6CHz/

"\

/

9

\

l

1

o

/f~

215


I

I

250Hz

i//

-10

9

o/oo o O

o/.L.

o

2 -20 ~- -30

9

o i

1

I

I

N9 o\

I

t

110

i

380Hz

-10

/ -30 -

9

O

@@

0

~0

-40 20

40

I 60 Centre

I 80

I 100

[ 200

~O'1,400

I 600

Frequency of Noise Band (Hz)

Fig. 4. The relative masking effect of i0 Hz wide noise bands, centred ~t different frequencies, upon 40, 60, 110, 200, 250 and 380 Hz pure tone signals

noise bands. This is to be expected, for when the centre frequency of a noise band is shifted just outside the critical band some energy remains at frequencies inside the critical band. This will tend to limit the precision with which the critical band can be defined using noise bands. The problem does not arise with a masking tone, since energy exists only at a single frequency. However, another difficulty arises when a tone is used. The masking effect may be reduced at frequencies close to the signal because of " b e a t s " between the signal and masker. Thus, Wegel and Lane (1924)and Egan and Hake (1950), working with human subjects, have reported that signal detection is enhanced when the masking tone

216

A.D. Hawkins and C. J. Chapman

-10

~-20

-

5

-30 -

-z,O

20

I

40

I

60

I

80

Frequency

1

100

I 200

I 400

I 600

1 800

(Hz)

Fig. 5. A comparison of the relative masking effect of 10 Hz wide noise bands upon pure tone signals at different frequencies within the hearing range of the cod. The audiogram is derived from Chapman and Hawkins (1973) is very close to the signal tone, creating a notch in the relative masking curve. Though similar effects were not noted with cod at 160 Hz, they could conceivably occur. I n addition to the observations made at 160 Hz, the masking effects of 10 Hz wide noise bands were examined at signal frequencies of 40, 60, 110, 200, 250 and 380 Hz. Curves showing the relative masking effect of noise bands centred at different frequencies are given in Fig. 4. These confirm t h a t masking is most pronounced when the masker and signal are close in frequency. Moving the noise band centre frequency one octave from the signal reduced masking b y 20-35 dB. At most signal frequencies (60, ]10, 160, 200, 250 Hz) the relative masking curves obtained w e r e approximately symmetrical when the noise band centre frequency was plotted on a logarithmic scale. At the lowest frequency tested (40 Hz) it was not possible to examine the s y m m e t r y of the curve, since only noise band centre frequencies of 40 Hz and above could be tested (owing to deficiencies in the sound projectors at very low frequencies). At 380 Hz, the masking curve was clearly asymmetrical, showing a steeper decline in masking on the high frequency side. This is to be expected, since the auditory system of the cod shows a sharp decrease in sensitivity to sounds above 400 Hz (Chapman and Hawkins, 1973) as illustrated in Fig. 5.

Masked Auditory Thresholds in Gadus

217

Table 1. Values for the critical ratio, calculated from threshold: noise data for the detection of pure tone signals against a background of white noise (Chapman and Hawkins, 1973). The second set of figures (B) are those obtained on the assumption that the power of the noise integrated over the critical band exceeds the power of the pure tone itself by 1 dB Frequency (Hz)

Threshold:noise ratio (dB) A. Critical ratio (ttz) B. Critical ratio (Hz)

50

60

110

160

200

310

380

17.1 51.3 64.5

15.9 38.9 48.9

15.6 36.3 45.7

18.3 67.6 85.1

18.9 77.6 97.7

20.8 120.2 151.3

20.6 114.8 144.5

3. The Characteristics o / t h e A u d i t o r y Filter

The existence of a critical band for masking was first demonstrated for human subjects by Fletcher (1940), who determined thresholds to pure tones in the presence of bands of noise of various width. As with our own band narrowing experiments, the results showed that for bands wider than some critical value the threshold of the tone was constant; however, as the bands were made progressively narrower the threshold for the tone decreased. Fletcher also proposed a method for calculating the width of the critical band indirectly, from experimentally determined threshold:noise ratios. This procedure was based on the assumption that when " w h i t e " noise just masks a tone, the relatively narrow band performing the masking is equal in power to the tone itself. A similar assumption can be applied to masking data obtained for cod by Chapman and Hawkins (1973). Thus, given a threshold:noise ratio of -~15.9 dB for detection of a 60 ttz tone in the presence of b r o a d band noise, then the effective power of the noise integrated over the critical band is equal to the power of the tone itself, i.e it is 15.9 dB above the spectrum level of the noise. Since for white noise the level within a band of width A / Hz is equal to the spectrum level plus 10 log10 A/, then: 10 log10 z]/c = 15.9, where zJ/c is the critical bandwidth. T h u s / I / c = 38.9 Hz. Values for the critical band determined in this way are termed critical ratios, to distinguish them from critical bandwidths derived empirically. Calculated critical ratios for the cod are presented in Table 1 A. Attempts to measure the width of the critical band empirically for humans have yielded results which conflict with the calculated critical ratios. Zwicker, Flottorp and Stevens (1957) have reported values up to 21/2 times wider than those given by Fletcher (1940) and similar discrepancies have been reported by Hamilton (1957) and Greenwood (1961). Scharf (1961) considered that these discrepancies were more apparent than real, and showed that slight modification of Fletcher's equal power assumption could give values of the critical ratio equal to measured values of the critical band. However, it is clear for human subjects that the measured values obtained can also vary, depending on the experimental method employed. For example, de Boer (1962) has criticised band narrowing techniques on the grounds that very narrow bands of noise are characterised by a high degree of amplitude fluctuation, which may influence their effectiveness in masking a signal. Perhaps the most serious criticism has, however, come from Swets and Green (1962). These authors have pointed out that the critical band

218

A.D. Hawkins and C. J. Chapman 180 '-"

160

9

J

140

120~ loo

8 i

80

6 40 2O 0

I 100

I 200

I 300

I ZOO

Frequency [Hz}

~ig. 6. Values of the critical b a n d w i d t h for different stimulus frequencies (o). The fitted line represents A ~ ~ 38.853 -t- 0.291f. Values of the critical ratio, t a k e n from Table 1 B are provided for comparison ( 9

is unlikely to constitute a fixed property of the auditory system, but is likely to v a r y from one sensory task to another. I n addition, they have stressed t h a t the value obtained for the width from a particular experiment will v a r y with a n y assumptions made concerning the shape of the auditory filter. Thus, in estimating the width of the critical band from our own data one important factor to be considered is the likely shape of the filter function. The relative masking curves (Figs. 3 and 4) provide evidence of this, but their shape is likely to be affected b y the actual characteristics of the masking noise employed. I t is most appropriate to examine the curves derived using secondary tones, as shown for a frequency of 160 ttz in Fig. 3. This curve clearly indicates t h a t the auditory filter is not rectangular. When the secondary tone is shifted to either side of the signal there is no sudden cessation of masking. I n this event it is not possible to describe the critical band entirely adequately in terms only of a finite bandwidth. However, to enable comparison of the data at different frequencies, and with data from other species we have attempted to devise an acceptable arbitrary measure. I n electronics or acoustics a complex filter function is commonly described in terms of an effective bandwidth. This is based on the concept of a filter function of rectangular shape, giving a uniform response to a narrow range of frequencies, but with a response which is equivalent in total energy terms to the response of

Masked Auditory Thresholds in @adus

219

Table 2. Values for the critical bandwidth, derived from band shifting experiments with 10 ttz noise bands as the masker Frequency (ttz) 40 Lower frequency limit Upper frequency limit Critical bandwidth

60

110

20.3b 37.6 79.3 102.0 59.0 64.4

160

160a

200

250

380

86.3 129.5 185.0 159.9 204.0 246.5 157.7 208.5 121.0 253.3 300.1 411.5 71.4 72.2 64.0 93.4 96.1 165.0

a Values derived from experiments with secondary tones as masker. b This value was not determined directly, but estimated on the assumption that the left part of the masking function extended only over half the frequency span of the right part.

the more complex function. Our colleagues C. Johnstone and M. Nieholson have taken our data on the relative masking effect of a 10 Hz wide noise band centred at different frequencies and have fitted appropriate curves. This has enabled them to calculate the area beneath each curve and then to construct a rectangular curve of equivalent area. The bandwidth of this rectangular function gives an effective value for the critical bandwidth. Values obtained by this procedure are given in Table 2, and are shown in Fig. 6. The detailed procedure is described in an appendix to this paper. Since the effective values are calculated for data based on narrow bands of masking noise, they are clearly likely to be rather larger than those obtained from pure tone masking experiments. At 160 Hz, the only frequency where secondary tones were employed, the discrepancy was 8.2 Hz. The effective width of the critical band increases for stimuli of higher frequency, as shown by Fig. 6.

Discussion 1. Critical Bands in Fish Our experiments have confirmed that for cod a pure tone auditory stimulus is masked most effectively by components of the background noise centred at the same and immediately adjacent frequencies. However, it is evident that masking is not confined to a discrete band of frequencies as implied by the use of the term critical band. t~ather, it shows a progressive decline with increasing frequency separation between signal and masker. Nevertheless, there are advantages to be gained by describing the masking function in terms of an effective band of finite width. This form of description is easily prepared, and readily allows comparison of data for different species. I n addition it can facilitate computation of the detectability of sounds under various noise conditions using the methods outlined by Urick (1967) and others. The values we have assigned in Table 2 to the critical bandwidth can be compared with the critical ratios we have derived from wide band masking experiments (Table 1 A). The former are rather large, perhaps because the empirical values are slight overestimates (since they are based on masking experiments using 10 Hz wide noise bands, and not pure tones). However, a similar but rather larger discrepancy between the critical ratio and critical bandwidth has been noted 7

J. eomp. Physiol.

220

A.D. Hawkins and C. J. Chapman

for h u m a n subjects, where the critical band m a y be up to 21/2 times greater (Zwicker et al., 1957). Scharf (1961) has pointed out that a simple modification of Fletcher's equal power assumption can eliminate this discrepancy. We could assume in the case of the cod, for example, t h a t when white noise just masks a pure tone the power of the noise integrated over the band performing the masking is not equal in power to the tone itself but to 1 dB more than this. Thus, given a threshold:noise ratio of 15.9 dB for detection of a 60 Hz tone against white noise then the effective power of the noise integrated over the critical band is equal to (15.9 ~-1) dB, corresponding to a bandwidth of 48.9 Hz (compared to 38.9 Hz obtained from the equal power assumption). Values of the critical ratio obtained using this 1 dB modification of the equal power assumption are given in Table 1 B. They show close correspondence with the empirically derived values (Table 2) and the same dependence on signal frequency (Fig. 6). Comparison of our data with t h a t presented b y other workers is rather difficult at present, mainly because the methods used to derive the critical bandwidth have varied widely. Buerkle (1969) has provided data for the cod based essentially on experiments using relatively wide bands of noise as both signal and masker. At 35 Hz the width is given as approximately 40 Hz, at 70 Hz it is approximately 60 Hz, and at 140 Hz it is rather greater t h a n 90 Hz. These values resemble our own, and show a similar, though rather steeper, upward trend with frequency. Cahn, Siler and Wodinsky (1969) have presented critical ratios for two species of Haemulon. Their values are very large, however, ranging from a width of 320 Hz at 200 Hz, to a width of 40000 Hz at 400 Itz. These authors have clearly misunderstood the concept of the critical ratio. As a pre-requisite for calculating critical ratios from threshold:noise data it is necessary to demonstrate t h a t the thresholds really are masked b y the noise. This is achieved b y showing t h a t as the noise level is raised the threshold:noise ratio remains approximately constant. Cahn et al. did not adopt this precautionary procedure and it is evident t h a t for some frequencies at least the thresholds were not masked. Tavolga (1974), following his earlier inconclusive experiments (Tavolga, 1967), has recently investigated masking in the goldfish. A band narrowing experiment showed t h a t at 500 Hz the critical bandwidth lay between 100 and 200 Hz. However, much smaller critical bandwidths were indicated b y a series of pure tone masking experiments, where extremely sharp relative masking curves were obtained. Thus, for a 500 Hz signal, masking declined b y over 25 dB when secondary tones of 490 and 510 ttz were used (as opposed to a secondary tone at 500 Hz). I t is difficult to reconcile this striking degree of frequency selectivity with the relatively wide critical bandwidth indicated b y Tavolga's b a n d narrowing experiment. Tavolga (1974) has also examined two other teleosts. Pinfish (Lagodon rhomboides), trained to a 500 Hz tone, were tested in the presence of wide bands of noise, centred first at 300 and then at 500 Hz. The resultant thresholds were not significantly different. Similarly, for Tilapia mac~vcephala, a threshold to a 300 Hz tone in the presence of a noise band centred at 300 Hz was not significantly different to the threshold obtained against a noise band centred at 590 Hz.

Masked Auditory Thresholds in Gadus

221

Tavolga tentatively concluded that these two species were incapable of frequency discrimination and that critical bands were absent from their auditory systems. In view of the rather high thresholds obtained, and the low noise levels employed, we believe that it is more likely that the thresholds were simply not masked. Varying the characteristics of the noise could not b e expected to have any effect in such circumstances. I t would appear that band shifting experiments offer the most precise method for examining the characteristics of the auditory filter in fish, especially if secondary tones are employed. With fish, unlike humans, complications due to the subjective sensation of " b e a t s " (occurring either near the stimulus frequency or its harmonics) are not apparent when tones are used. This does not mean that beats are not experienced by fish. The fish may simply not associate them with the presence of the tone stimulus.

2. The Auditory Filtering Mechanism Fish, like mammals, evidently use an auditory filter to improve the detection of pure tone signals in the presence of wide band noise. Many species are also known to be able to discriminate between pure tones of differing frequency (for recent reviews see Tavolga, 1971; Hawkins, 1973). Yet there is no obvious form of frequency analyser within the fish ear analogous to the cochlea. Both van Bergeijk (1967) a M Tavolga (1974) have nevertheless concluded that the existence of critical bands demands some form of mechanical frequency analyser. Van Bergeijk suggested that the sensory membrane or macule supporting the otolith behaved as a bounded membrane with sufficient asymmetry to respond differently to different frequencies. Sand (1974), on the other hand, has argued that the vibration pattern of the otoliths themselves may be frequency dependent, and that the part of the macula stimulated by the otolith changes with frequency. More recently, Tavolga (1974) has suggested that in the goldfish the Weberian ossicles connecting the ear to the swimbladder may play a role in frequency analysis. There is little morphological evidence to support any of these proposals. In mammals, strong evidence that the cochlea constitutes a mechanical frequency analyser comes from the observation that the microphonic potentials generated in response to a particular pure tone stimulus are located at a particular place along the cochlear partition (Tasaki etal., 1952). Andersen and Enger (1968) recorded microphonic potentials from the saccular otolith organ of the bullhead, Cottus scorpius, and reported rather small but consistent local differences in the response to higher frequencies. In addition, Sand (1974) showed a rather broad loealisation of mierophonics in the sacculns of the perch, Perca ]luviatilis. At present however there is little evidence for sufficiently sharp localisation to justify full acceptance of a mechanical theory. There is more convincing evidence that frequency selectivity is mediated within the central nervous system, the waveform of an auditory signal being conveyed from the otolith organ to the brain with only a minimum of filtering. Enger (1963) studied the activity of primary auditory neurons in the bullhead and revealed that there were three kinds of neuron, those with no spontaneous discharge responding to frequencies up to 200-300 Hz, those with an irregular

222

A.D. Hawkins and C. J. Chapman

spontaneous discharge responding to frequencies up to 300-500 Hz, and those with a burst-like spontaneous discharge responding with phase-locked discharges up to frequencies of 200-300 Hz. Enger concluded that frequency discrimination might be performed centrally, based on a stimulus following neural response. However, he also suggested that an additional peripheral neural mechanism might be employed, based on the existence of 2 types of neuron, one responding to low frequencies, and the other responding to stimuli over the whole frequency range perceived by the fish. A study of primary neuron activity in the goldfish by Furukawa and Ishii (1967) also provided strong evidence of a stimulus following response, whereby the waveform of the stimulus is conveyed directly to the central nervous system. I t would certainly seem that for fish, unlike mammals, there is only a very limited differentiation between primary units responding to different frequencies. Sharply tuned units, like those reported by Kiang (1965) for the cat have not been found. Furthermore, the coding of potential information on stimulus waveform in terms of intervals between successive nerve impulses may also be maintained at lower levels within the central nervous system. GrSzinger (1967) working on the tench, Tinca tinca, has reported that electrical responses at all levels in the central auditory pathway show a dependence of excitation pattern upon stimulus waveform, with no evidence for any topographical separation of frequencies. On the other hand, both Enger (1967) and Page (1970) showed that some central frequency selectivity is practised. Enger studied multi and single unit activity in the acoustic area of the medulla oblongata of the herring, Clupea harengus, and demonstrated that the range of frequencies to which single units would respond varied, though the tuning curves were rather broad. Page (1970) found that in the medulla of the goldfish less than half the units gave a stimulus following response, and that the range of frequencies over which these units followed phasically was more limited than for primary units. At a higher level, in the torus semieircularis, none of the units observed showed a stimulus following response and most gave much narrower tuning curves than the medullary units. Page suggested that though fibers in the auditory nerve responded with a patterned impulse discharge conveying frequency information, in the central nervous system this was transformed into a code based on the frequency specificity of individual neurons. Units at each succeeding level were more ilk ely to respond briefly with a discharge pattern not correlated with the stimulus, and to have a narrower tuning curve. Recently, Enger (1973) has investigated the masking of auditory responses in the medulla oblongata of the goldfish. The response of individual neurons was similar to that reported b y Page (1970) and consisted of an increase in discharge rate above a low spontaneous level. The response given to pure tones and to third octave and 1 octave bands of noise of similar power, centred at the same frequency, was essentially the same. Enger therefore suggested that the response of each unit was associated with an integration of acoustic power over a frequency span of about an octave. However, Enger also compared the degree of masking of the response to a pure tone stimulus in the presence of third octave bands of noise at different amplitudes and centre frequencies. Masking was strongest when the centre frequencies of signal and noise coincided, and declined by 20-22 dB/octave

Masked Auditory Thresholds in Gadus

223

as the frequency separation between them increased. There was negligible masking of a 250 Hz tone by a noise band centred 1 octave away. These results indicate that some medullary auditory units integrate acoustic power over a frequency range of much less than an octave. Enger's results for single unit activity in the medulla conform strikingly with our own measurements. This is convincing evidence that a mechanism capable of generating the relatively narrow masking bandwidths of fish exists within the central nervous system. These observations do not entirely preclude the possibility of a crude peripheral mechanical filtering being performed within the otolith organ. Indeed, Sand (1974) has pointed out that a rather coarse mechanical filter. ing might facilitate the operation of a central neural analyser. We are grateful to A.D.F. Johnstone and C. Robb for their assistance with the experiments; to our colleagues J. Foster and B.B. Parrish for their critical reading of the manuscript; and to A. Tumarkin for his valuable advice.

Appendix: Derivation of an Effective Value for the Critical Bandwidth Christine Johnston and M.D. Nicholson For a given pure tone signal (f), we were given a series of values expressing the relative masking effect of a number of 10 Hz wide noise bands centred at different frequencies. These have been plotted in Fig. 3 and 4 for various values of f. We have defined the critical bandwidth as the width of a rectangular function equal in area and height to the region bounded by an integrable curve fitted through the data and a suitably chosen baseline. A curve composed of two segments, each of which is a quadratic constrained to pass through (f, o) provides a suitable integrable function, and gives an adequate fit to the data. The baseline was taken to be --30 dB, since below this level ambient noise can be expected to affect the data. The limits of integration were then taken to be either the points where the quadratics cut the baseline, or the turning points of the quadratics, whichever were the less extreme. Each equivalent rectangular curve thus had a height of 30 dB, and its width, the effective bandwidth, could readily be calculated. The position of the effective bandwidth was chosen to make the area to the left (and right) of the pure tone signal the same for both the quadratic and rectangular curves. This method can be illustrated by the calculation of an effective value for the critical bandwidth at a signal frequency of 200 Hz. The data are given in Table 3. Table 3. Values for the relative masking effect of 10 Hz wide noise bands centred at different frequencies upon a 200 Hz pure tone signal Centre frequency of noise band (Hz) 60 140 160 180 190 200

220

Relative masking (dB) --29.4 --21.6 --17.8 --8.6

--11.4 --13.6 --15.6 --21.4 --33.0 --25.1

--5.3 0.0

250

250

310

310

380

224

A.D. Hawkins and C. J. Chapman Effective Critical Band- width

<

>I I il I I

/

~f

c~ -20 .~ - 3 o

-40 ~-

|

-'-i

.......

,

',

,,.\,

I

--"'.4\---]

e/ ....

,z,, \

I. . . .

9

,

!I

r

200

t00 Centre

Frequency

of Noise Bond

II II ]

I l

I

400

300 {Hz)

Fig. 7. Derivation of an effective value for the critical bandwidth at a signal frequency of 200 Hz. The curve is composed of 2 segments, each representing a quadratic equation fitted to the relative masking data. The equivalent rectangular function is also shown

Two q u a d r a t i c s were f i t t e d b y l e a s t squares; one to t h e 5 p o i n t s w i t h abscissa less t h a n 200 Hz, t h e o t h e r t o t h e 6 p o i n t s w i t h abscissa g r e a t e r t h a n 200 Hz. B o t h q u a d r a t i c s were c o n s t r a i n e d t h r o u g h t h e p o i n t (200.0). T h e f i t t e d s e g m e n t e d curve was y=/(x)=O.OO206x~--O.326x--17.2 for x _<200 0.00143x2--0.972x + 1 3 7 . 1

for

x > 200

where x is t h e centre f r e q u e n c y of t h e 10 H z wide noise b a n d a n d y is t h e r e l a t i v e masking. T h e left h a n d s e g m e n t of t h e curve has a v a l u e of - - 3 0 d B a t a f r e q u e n c y of 85.6 H z w h i c h was t a k e n as t h e lower l i m i t of i n t e g r a t i o n . T h e m i n i m u m v a l u e of t h e r i g h t h a n d s e g m e n t is --27.8 d B a t t a i n e d a t a f r e q u e n c y of 339.2 Hz. T h e u p p e r l i m i t of i n t e g r a t i o n was t h e r e f o r e t a k e n to be 339.2 Hz. T h e a r e a b o u n d e d b y t h e s e g m e n t e d curve a n d t h e baseline of - - 3 0 d B is 2801.4. T h u s t h e w i d t h of a n e q u i v a l e n t r e c t a n g l e is 93.4 Hz. T h e a r e a b o u n d e d b y t h e left h a n d segment, t h e baseline a n d t h e t o n e freq u e n c y 200 H z is 1201.9. H e n c e t h e b a s e of t h e r e c t a n g l e is (159.9, 253.3). Fig. 7 shows t h e s e g m e n t e d curve f i t t e d t o t h e d a t a a n d t h e e q u i v a l e n t rectang u l a r function.

References Andersen, R.A., Enger, P. S. : ]%~ierophonic potentials from the saeculus of a teleost fish. Comp. Biochem. Physiol. 27, 879-881 (1968) Bergeijk, W.N. van: Discussion. In: Marine Bio-acoustics, vol. 2 (ed. W.N. Tavolga), p. 244. Oxford: Pergamon Press 1967 B6k~sy, G. yon: Experiments in hearing. New York: McGraw-Hill 1960 Bilger, R.C., Hirsh, L J. : Masking of tones by bands of noise. J. acoust. Soc. Amer. 28, 623-630 (1956)

Masked Auditory Thresholds in Gadus

225

Boer, E. De: Note on the critical bandwidth. J. acoust. Soc. Amer. 84, 985-986 (1962) Buerkle, U.: Relation of pure tone thresholds to background noise level in the Atlantic cod. J. Fish. Res. Bd. Can. 25, 1155-1160 (1968) Buerkle, U. : Auditory masking and the critical band in Atlantic cod. J. Fish. ges. Bd. Can. 2g, 1113-1!19 (1969) Cahn, P.IL, Siler, W., Wodinsky, J. : Acoustico-lateralis system of fishes: tests of pressure and particle-velocity sensitivity in Grunts, Haemulon sciu~'us and H. par,rai. J. acoust. See. Amer. 46, 1572-1578 (1969) Chapman, C. J. : Field studies of hearing in teleost fish. Helgoli~nder wiss. Meeresunters. 24, 371-390 (1973) Chapman, C.J., Hawkins, A.D.: A field study of hearing in the cod. J. comp. Physiol. 85, 147-167 (i973) Egan, ff.P., Hake, H.W. : On the masking pattern of a simple auditory stimulus. J. ucoust. Soc. Amer. 22, 622-630 (1950) Enger, P. S. : Single unit activity in the peripheral auditory system of a teleost fish. Acta physiol, scand. 69, suppl, i-48 (i963) Enger, P. S. : Hearing in herring. Comp. Biochem. Physiol. 22, 527-538 (1967) Enger, P. S. : Masking of auditory responses in the medulla oblongata of goldfish. J. exp. Biol. 69, 415424 (1973) Fletcher, H. : Auditory patterns. Rev. rood. Phys. 12, 47-65 (1940) Furakawa, T., Ishii, Y. : Neurophysiological studies on hearing in goldfish. J. Neurophysiol. 80, 1377-1403 (1967) Greenwood, D.D. : Auditory masking and the Critical band. ft. acoust. See. Amer. 33, 484-502 (1961) Grhzinger, B.: Elektro~physiologische Untersuchungen an der Hhrbahn der Schleie (Tinca tinca [L.]). Z. vergl. Physiol. 57, 44-76 (1967) Hamilton, P. M. : Noise masked thresholds as a function of tonal duration and masking noise bandwidth. J. acoust. Soc. Amer. 29, 506-511 (1957) Hawkins, A.D. : The sensitivity of fish to sounds. Oceanogr. Mar. Biol. Ann. Roy. 11, 291-340 (i973) Keidel, W.D.: Biophysics, mechanics and electrophysiology of the human cochlea. In: Frequency analysis and periodicity detection in hearing (ed. R. Plomp, G.F. Smoorenburg), p. 60-79. Leiden: A.W. Sijthoff 1970 Ki~ng, N.Y.: Discharge patterns of single fibres in the cat's auditory nerve. M.I.T. Research Monograph l~o. 35, Cambridge Mass. : M.I.T. Press 1965 Page, C. H. : Electrophysiological study of auditory response in the goldfish brain. J. Neurophysiol. 88, 116-128 (1970) Sand, O. : Directional sensitivity of microphonic potentials from the perch ear. J. exp. Biol. 60, 881-899 (1974) Schafer, T.tt., Gales, C.A., Shewmaker, C.A., Thompson, P.O. : The frequency selectivity of the ear as determined by masking experiments. J. aeoust. Soc. Amer. 22, 490-496 (1950) Seharf, B. : Complex sounds and critical bands. Psychol. Bull. 68, 205-217 (1961) Seharf, B.: Critical bands. In: Foundations of modern auditory theory, vol. 2 (ed. T.V. Tobias), p. 159-202. New York: Academic Press 1970 Swots, O. Green, D.M.: On the width of critical bands. J. acoust. Soc. Amer. 84, 108-113 (1962) Tasaki, I., Davis H., Legouix, J.: The space-time pattern of the cochlear microphonics (guinea pig), as recorded by differential electrodes. J. acoust. Soc. Amer. 24, 502-519 (1952) Tavolga, W.iN. : Masked auditory thresholds in teleost fishes. In: Marine Bio-acoustics, vol. 2 (ed. W.N. Tavolga), p. 233-243. Oxford: Pergamon Press 1967 Tavolga, W. 1%.: Sound production and detection. In: Fish Physiology, vol. 5 (ed. W. S. Hoar, D.J. R~ndall), p. 135-205. iYew York: Academic Press 1971 Tavolga, W.N. : Signal-noise ratio and the critical band in fishes. J. acoust. See. Amer. 55, 1323-1333 (1974) Urick, R. J. : Principles of underwater sound for engineers. 342 pp. 5Tow York: McGraw-Hill 1967

226

A.D. Hawkins and C. J. Chapman

Wegel, R.L., Lane, C. E. : The auditory masking of one pure tone by another and its probable relation to the dynamics of the inner ear. Phys. Rev. 23, 266-285 (1924) Zwicker, E., Flottorp, G., Stevens, S.S. : Critical bandwidth in loudness summation. J. acoust. Soc. Amer. 29, 548-557 (1957) Dr. A. D. Hawkins Mr. C. J. Chapman Marine Laboratory P.O. Box 101, Victoria Road Aberdeen AB9 8DB, Scotland