Hearing Research, 8 (1982) 157- 178 Elsevier Biomedical Press
157
AP unmasking and AP tuning in normal and pathological human ears W.L.C. Rutten and P. Kuper ENT Department,
University (Received
Hospital, Rijnsburgerweg 3 February
IO, 2333 AA L.eiden, ,The Netherlands
1982; accepted
21 May 1982)
In a groupof seven normal and eight abnormal hearing subjects three-tone AP unmasking experiments and/or AP tuning experiments were performed during electrocochleography. In the unmasking experiments the Shannon forward masking paradigm was applied, i.e. two simultaneous tones (the masker and the suppressor) influence a test tone in a forward masking procedure. Frequency and intensity of the suppressor were varied. It appeared that unmasking effects are clearly present, i.e. the suppressor stimulus can reduce masking of the AP. This effect, however, is very variable, in normal ears as well as in pathological ears. AP tuning quality deteriorated with increasing hearing loss but no correlation was found between AP unmasking parameters and hearing loss. It seems that AP suppression areas shift less upward (or not at all) than the AP tuning curve does upon increasing the test tone level. These results raise some questions about the intercorrelation of the triad: hearing loss, quality of tuning and suppression effects, as single fibre experiments and psychophysical investigations suggest. Key words: compound action potential (AP); tuning; auditory nonlinearity; suppression.
AP unmasking;
AP suppression;
forward
masking;
AP
I. Introduction In this paper we present some results of experiments on auditory unmasking and tuning in man on the basis of compound action potential (AP) measurements. Unmasking is the release from masking. In a normal ear it can be brought about by using a non-simultaneous forward masking paradigm in which an appropriate third tone (besides masker and test tone) is presented simultaneously with the masking tone, the third tone being of the same duration as the masking tone. This third tone will be further called ‘the suppressor’. Unmasking has been measured by psychophysical methods in man [ 12,241 and in single fibre and whole nerve responses in chinchilla [lo]. AP unmasking has been shown to be related to two-tone rate suppression [lo]. Two-tone rate suppression is the phenomenon that the response of an auditory nerve fibre to an excitatory tone may be reduced by a second tone which itself need not be excitatory. Two-tone rate suppression is an aspect of the so-called essential non-linear behaviour [9] of the peripheral hearing organ. The expression ‘essential’ points to the fact that this non-linear behaviour already exists at stimulus levels near threshold and cannot be described by a classical 0378-5955/82/0000-OOOOooo/sO2.75
@ 1982 Elsevier
Biomedical
Press
1%
polynomial non-linearity [9,13]. Two-tone rate suppression and in general essential auditory non-linearity is rather susceptible to cochlear pathology. This has been shown in many electrophysiological animal experiments [3,19,20,22,23]. In man the effect of cochlear pathology on essential non-linear behaviour has been studied by psychophysical methods. We mention here the studies of Wightman et al. [28], Leshowitz and Lindstrom [15] and Penner [ 181 who studied unmasking in subjects with a hearing loss of cochlear origin and Smoorenburg [25] who investigated the effect of a ‘dip’ in the audiogram on combination tone generation. From both animal and human studies it can be concluded that auditory non-linear behaviour is reduced by pathology. However, the extent to which aspects of non-linear behaviour are affected will depend on the nature, degree and spatial extent of the particular injury. Moreover, different aspects of non-linear behaviour may behave differently under the same cochlear pathology. In this respect we mention the study of Smoorenburg [26] who found a different effect of tone-induced temporary threshold shift (ITS) on psychophysical combination tone generation and psychophysical unmasking. He concluded unmasking to be more resistant to TTS than the generation of the combination tone 2 f, - f2. In animal experiments a different effect was reported on low- and high-frequency suppression (‘low’ and ‘high’ refer to the position of the suppressing frequency with respect to the characteristic frequency (CF) of the fibre) by Dallos et al. [3], Evans [8] and Schmiedt and Zwislocki [22]. The same cochlear pathology can also have a different effect on non-linearities on the one hand and other features of cochlear functioning, such as threshold and tuning, on the other hand. In this context we mention the studies of Schmiedt and Zwislocki [22] and Schmiedt et al. [23] who found near-normal tuning in the absence of high-frequency two-tone rate suppression in kanamycin-damaged cochleas of gerbils. Similar findings were reported by Dallos et al. [3], Robertson and Johnstone [ 191 and Evans [S]. Since AP unmasking is believed to reflect some aspects of cochlear non-linear behaviour, in particular two-tone rate suppression, it can be used to study these methods is aspects, especially in man, where the choice of electrophysiological limited. AP unmasking seems to be more suited in this respect than psychophysical unmasking, because psychophysical methods are subject to certain confusing central effects such as “quality difference cues” [ 17,271. Moreover, in a clinical setting subjects are not always able to receive the necessary training for psychophysical experiments. In view of the fact that auditory non-linearities show an intricate dependence on cochlear pathology, it is of interest to study the AP-unmasking effect in normal as well as in pathological human ears and to compare the unmasking behaviour with other measures of cochlear functioning such as pure tone audiogram, AP thresholds and AP tuning curves. In conclusion, in this study we address ourselves to finding an answer to the following questions: (1) Is it possible to show AP unmasking in normal as well as in abnormal human ears? (2) Being presumably the reflection of a highly vulnerable cochlear non-linearity, how does AP unmasking behave in pathological ears?
159
(3) Is there a relation between some measure of the AP other measures of cochlear functioning such as hearing loss (4) Does the dependence of unmasking on suppressor psychophysical unmasking, manifest itself analogously in AP
unmasking effect and and AP tuning curve? level, as observed in responses?
II. Methods
Experimental unmasking and tuning data were collected during standard combined electrocochleography (ECoG)/Brainstem Evoked Response (BER) test procedure. The applied standard Leiden ECoG procedure is basically identical to that used by Eggermont et al. [5]. Briefly, it consists of recording by transtympanic electrode AP responses to short tone burst stimuli at 0.5 kHz, 1, 2, 4 and 8 kHz. Tone bursts have a trapezoidal envelope with 2 periods rise and fall time and a 4 ms plateau. They are delivered at a rate of about 8/set. Input-output curves for AP amplitude and latency are determined, yielding AP threshold and slope of the input/output (I/O) curve. At a standard intensity level SP amplitudes are measured (for 2,4 and 8 kHz at 85, 80 and 95 dB HL, respectively). Narrow band APs (NAPS) are derived for 90 dB pe SPL (re 20 PPa), 70 ps rectangular clicks using a simultaneous high pass masking/subtraction paradigm [6]. Recordings are made in a sound-treated room, stimuli are delivered free-field by a Vitavox GPl pressure unit, attached to a circular exponential horn (Vitavox, type 190). Distance between entrance to the ear canal and pressure unit is 1.2 m. The response is amplified 20000-50000 times, band pass filtered between 10 Hz and 8 kHz (slopes 12 dB/octave, PAR 129) and averaged 5 12 times by a DATALAB 4000 system. Analysis time is 15 ms (including 3.5 ms prestimulus time), sampling rate is 16.66 kHz. Averaged responses are stored on flexible discs in a PDP 1 l/O3 minicomputer system for computer-aided read-out and further analysis. For details about spectra of the stimuli see [6]. Additionally to the standard ECoG procedure one of the following types of experiments was performed on each subject: (1) AP unmasking, (2) AP tuning curves, (3) both unmasking and tuning. First, in the unmasking experiments. (nine subjects) a test-tone burst, with the same envelope as given above, with frequency fT was presented. The level was chosen such that a well defined AP response was elicited, usually at lo-30 dB above threshold. A second tone burst, the masker, was then added in a forward masking configuration, having the same frequency and (usually) the same intensity (in dB SPL during the plateau) as the test tone. Rise and fall time of the masker burst was 5 ms, the plateau length was 100 ms. Setting the masker level equal to the test tone level resulted usually in a 50% reduction of the test tone response. The test tone started 10 ms after the end of the masker. Repetition rate of the masker test-tone combination is 4/set. This rather slow rate is necessary in order to avoid adaptation effects [5]. A third tone burst, to be called the suppressor, was presented simultaneously with the masker. In one type of unmasking experiment, the frequency of the suppressor, f,, was varied between 250 and 12 000 Hz, its level was adjusted 20 dB
160
above the masker level. This type of unmasking experiment will be called the U( A\f). In the other type, unmasking was measured as a function of suppressor intensity; this type of experiment will be called U(Al). In U(Al),f, is below, at or above the test frequency fr. Test tone, masker and suppressor were generated and attenuated separately, mixed electrically and delivered through the same Vitavox loudspeaker. There was no phase lock between the three tones. Besides the above described types of experiment U(Af) and U(AI), other parameters have been varied additionally in a few cases. In one subject U( Af) was measured at a low as well as a high fr. In two subjects U( Al) was performed at two test-tone levels, 25-30 dB apart. In one subject, both U( Af ) and U( Al) experiments were done. Secondly, AP tuning curves [7,11] were measured in three subjects. Tuning curves (TC) were measured by a forward masking or a simultaneous paradigm. In the forward masking TC experiment (one subject) the relative timing of test to maskertone was the same as during unmasking experiments. Test tone level was adjusted lo-40 dB above AP threshold at the test frequency. Masker frequency and intensity were varied. At each masker frequency the masker intensity corresponding to 50% reduction of test tone response amplitude was determined. The 50% point was found by interpolation in a response amplitude (linear) versus intensity (logarithmic) plot between at least three data points. In the simultaneous masking paradigm (two
TABLE I OVERVIEW Subjects
OF THE EXPERIMENTS Tuning
Unmas~ng
Mean ECoG loss a (dB)
WAf 1 121180 181280 021280 251180 011080 130181 210181 03028 1 301080
0 0 0 0 0 00 00
071180 171080 250280
0 0
150279a 150279b 180880
r/(AO
Simultaneous masking
Forward masking 42 17 18 24 42 23 15 29 60
00 0
0
0
0 0 0
I3 44 49
0 0
34 2 46
* Average hearing loss as obtained by electrocochieography.
0
161
ClA
fm
B
IS
Fig. I. A. Highly schematic plot of a tuning curve with flanking suppression areas (hatched areas). Horizontal arrow indicates how suppression and excitation areas are traversed in a V( Af) experiment, in which the only variable is frequency of suppressor I,. The vertical arrow indicates how in a (I( A I) experiment suppressor intensity was varied, the other parameters being held constant. In the lower left corner, the stimulus configuration of an unmasking experiment has been shown schematically, i.e. two simultaneous stimuli (masker M and suppressor S) both influence the test tone (T) response in a forward masking configuration. For further details see Methods. B. Schematic qualitative plots of AP/AP,,, ratios versus!, in a V( A/) experiment or versus I, in fJ(AI) experiments, as expected on the basis of Fig. IA. Where the suppressor ‘crosses’ suppression areas, AP/AP,,, ratios will exceeed the 100s reference level, i.e. unmasking is observed (hatched areas).
subjects) the only difference with the forward masking situation was that the masking tone was a continuous sinusoid. Thirdly, in three subjects both unmasking and tuning were measured. In one of them both simultaneous- and forward-tuning curves were determined, and in two subjects only simultaneously masked tuning curves. An overview of the experiments is given in Table I. A highly schematic illustration summarizing the unmasking and tuning experiments is given in Fig. 1A. It serves to illustrate how the third tone, the suppressor, ‘cross-sections’ the tuning curve in the two types of experiments U( Af ) and U( AI). The hatched area in Fig. lA, flanking the (simultaneous) tuning curve indicates the suppression area [ 1,10,21]. It is in this area that we expect the suppressor tone to reduce the masker activity and thus to unmask the test tone response. Depending on the choice of the variable, f, or I,, one expects results like those of Fig. 1B. The hatched area indicates where unmasking is present, i.e. where the response to test plus masker plus suppressor exceeds the response to test plus masker alone (the 100%reference level).
III. Subjects
Subjects were positioned in a supine position and sedated with 150 mg Nembutal, 25 mg Pethidine chloride and 25 mg Phenergan. Recording sessions took two to three hours.
162
In the course of routine electrocochleography it was decided to do unmasking or tuning experiments (or both) in a total of 15 adult patients. Decision criteria were (1) sedation had to be highly effective, i.e. the patient should sleep long and deep enough, (2) the hearing loss had to be cochlear or of the conduction type and (3) AP latencies and amplitudes had to be unambiguous. All subjects were extensively analysed before ECoG by conventional audiometry, i.e. pure tone audiometry, speech audiometry, BCkCsy audiometry, Alternate Binaural Loudness Balance test, tone decay test, stapedius reflex threshold- and decay test. Also vestibular function was examined and X-ray petrosum photographs were taken. It turned out that 13 patients had a hearing loss of cochlear origin in the ear tested by ECoG. Two patients (who were difficult to test conventionally) had a moderate conduction loss. In the group of 13 patients seven were clinically diagnosed as early or advanced Menitre patients, two as sudden deafness cases (one had recovered nearly completely at the time of ECoG, one recovered after ECoG). In the remaining four patients the cochlear hearing loss could not be further classified clinically. Typical ECoG findings in these four patients were desynchronization (broad AP waveform) or functional haircell loss (very small or zero SP at 2, 4 and 8 kHz). ECoG AP thresholds nearly always were close (within 15 dB) to pure tone audiometric losses between 0.5 and 8 kHz. Classified according to hearing loss seven patients showed almost flat, moderate loss ECoG ‘audiograms’, with an average loss (0.5 to 8 kHz) less than 25 dB (including the two conduction loss patients). We will refer to this group as ‘normals’. In the remaining 8 ears average losses were higher, up to 60 dB (mean 43 dB) while audiograms could be flat, sloping (positively or negatively, up to slopes of 20 dB/octave) or irregular. Pure tone audiogram losses of
Fig. 2. Overview of pure-tone-audiogram hearing losses versus frequency in a three-dimensional plot in which subjects are positioned along the third axis, divided into a normal (7 subjects) and an abnormal group (7 subjects). The criterion for normality was an average hearing loss (between 0.5 and 8 kHr) of less than 25 dB. The two groups are separated by four flat horizontal lines.
163
the two groups are collected in a three-dimensional plot (Fig. 2) in which the normal group is positioned in front of the abnormals. One abnormal, malingering subject (250280) has been omitted.
IV. Results - Experimental
data
A. U(A f) experiments
Fig. 3 shows a typical series of AP responses in an U(Af) experiment (subject 121180). The top response is the response to a test tone burst alone (3 kHz, 65 dB HL). The two curves drawn in the second recording show the responses to test tone plus masker (with masker frequency and intensity equal to those of test tone). The two traces were collected at the beginning and end of the experiment. The forward masker reduced the test tone response by 50%. The remaining traces in Fig. 3 indicate the effect of adding the suppressor to the masker. Frequency of the suppressor was varied between 12 and 0.25 kHz, the level was held constant at 90 dB SPL. Dependent on suppressor frequency f,, the AP response amplitude was enhanced (unmasked) or reduced (extra masking) with respect to the test tone plus masker condition. At the end of the experiment the suppressor was switched off. Now the response to test tone plus masker was again determined (see the superimposed dotted Curve in the response pair, second from top): a good reproduction has been found. Definitions of AP amplitude and latency have been indicated in the f, = 2.0 kHz trace. AP amplitudes and latencies of Fig. 3 are plotted as functions off, in Fig. 4 (left part), Amplitudes are expressed as AP/AP,,, ratios where AP,,, is the response amplitude to test plus masker alone. Values in excess of 100% indicate unmasking. For clarity, the resulting unmasking area has been shaded (Fig. 4). The right part of the figure shows the pure-tone audiogram (drawn curve), together with ECoG thresholds (circles). The plus-symbol marks frequency and level of the test tone and the forward masker. Unmasking is very prominent for this subject both for f, >fr and for f, < fT (in the course of this article we shall abbreviate these two conditions as ‘high-frequency’ and ‘low-frequency’ unmasking), in spite of the appreciable low frequency cochlear losses. Unmas~ng amplitude is 186% at 0.5 kHz at the lowfrequency side and 152% at 10 kHz in the high-frequency region. As the masker-alone reduced the test tone responses by 50%, an AP/AP, value of 186% means nearly complete unmasking. Unmasking areas are very broad. Latency is rather constant, except for a prolongation in the 3.5 kHz region. Note that latency maximum and amplitude minimum do not occur at the same suppressor frequency. Immediate remeasurement of U(A f) in this and other subjects showed that AP/AP,,, values could have been changed by at most *to%. On the average this intrasubject variation, i.e. the ‘reproducibility’, was about * 10%. In Fig. 5A-D we present V( Af ) results in four subjects with various audiograms. We plot these (and following) results in separate figures, as there are large inter-individual unmasking differences. Averaging over subjects, even in subgroups, would markedly obscure these differences.
121180
TESTALONE
TEST+ MASKER
TEST+ MA SKER + SUPPRESSOR Fs E 12 kHz l0 6
6
L 3.5
3 2.5 2 1.5
1
0.75
0.5 0.25
f
* . . .
’
10 5 4RT TESTTONE
’
’
*
’
’ ’ 15 ms
Fig. 3. Typical example of AP responses in a U( Af) experiment (subject 121180). Responses to test stimulus alone, test plus masker (solid line, at the start; dotted line, at the end of the suppressor series) and test plus masker plus suppressor are shown sequentially from top to bottom. Test tone and masker: 3 kHz, 65 dB HL; suppressor: 90 dB SPL. Definitions of AP ampiitude and latency are given in thel, = 2.0 kHz response.
165
0
2
..1..'.1 .5
WFpressm
1
fr*grfKy 1, CT 5 10 kHz
Fig. 4. Subject 121180. AP/AP, ratios versus suppressor frequency (lower left part). The AP/AP, for the test plus masker alone condition has been chosen as the 100% reference level. Hatched areas indicate unmasking areas. Corresponding latencies are shown in the upper left part. The audiogram is given on the right. Open circlesdenote AP thresholds; the solid line is the pure-tone subjective audiogram. The plus symbol indicates frequency and level of the test tone. Test tone and masker: 3 kHz, 65 dB HL; suppressor: 90 dB SPL.
Subject 181280(Fig. 5A) shows clearly confined unmasking areas with maxima at 2 and 4 kHz (AP/AP,,, = 190% and 1408, respectively). Because of the rather low hearing loss, stimulus levels could be lower than in Fig. 4. Latency is prolonged at 3.5 kHz, as in the previous case. In Fig, 5B (subject 021280) U(Af) results are presented in a pure conductive loss patient. Response quality was such that test-tone level had to be chosen 30 dB above threshold. Unmasking areas are shifted outwards, compared to Figs. 4 and 5A, i.e. to higher frequencies for f, > fT and to lower frequencies for f, < fT. The low-frequency part probably has its maximum below 0.7 kHz. The minimum amplitude is very low, in fact there was no response for f, = 2 kHz and f, = 2.5 kHz. Contrary to Figs. 5A and 4 there is no sharp latency maximum. No low frequency unmasking was found in a moderate-loss Meniere patient (subject 25 1180, Fig. 5C). High-frequency unmasking is broad, its maximum seems to be situated at 10 kHz. Latency is almost constant and extra masking for f, - fT cannot be observed. Fig. 5D displays U(A f) results in subject 011080, who has high-frequency losses. Rather low high-frequency unmasking is observed, whereas at the low-frequency side unmasking is broad and large and has two maxima. This case is, to a certain extent, the reverse of the previous case, both for the audiogram and the V( Af ) outcomes. Most remarkably, some high-frequency unmasking is observed, although for f, above 4 kHz the suppressor will not be above sensation threshold. Latencies are prolonged not only around 2 kHz, as observed earlier, but also below 0.4 kHz.
B. lJ(AjJ experiments, level effects In order to explore the effect of variation of test tone level and frequency,
three
Fig. 5. A. As in Fig. 4, for subject 181280. Test tone: 3 kHz, 40 dB HL. Masker: 3 kHz, 30 dB HL; suppressor: 60 dB SPL. B. As in Fig. 4, for subject 021280. Test tone: 3 kHz, 50 dB HL. Masker: 3 kHz, 55 dB HL; suppressor: 85 dB SPL. C. As in Fig. 4, for subject 251180. Test tone and masker: 3 kHz, 45 dB HL; suppressor: 75 dB SPL. D. As in Fig. 4, for subject 011080. Test tone and masker: 2 kHz, 45 dB HL; suppressor: 70 dB SPL.
U(Af) experiments were undertaken at two different test levels (2 subjects) or two test frequencies (one subject). Figs. 6A and B show the results for test level variation in 2 subjects ( 130181, conductive losses and 210181, low &one perceptive loss). Test tone levels were W/75 dB. HL and 30/60 dB HL, respectively (+ and X symbols in the audiograms). Masker and suppressor levels were varied accordingly (see Methods section). Both subjects fall in the normal group. For’the test level of 50 dB HL (drawn line in Fig. 6A) there is a clear unmasking effect, the low- and ham-frequency regions being separated by a sharp 3 kHz ‘dip’. The higher test level of 75 dB (dashed curve) shows a broadened ‘dip’ and a beginning of unmasking above 10 kHz and probably below 0.7 kHz. These results will be discussed in Section V. The second case (subject 210181, Fig. 6B) is qualitatively identical to Fig. 6A, except that at the higher level unmasking areas are much less shifted apart. Moreover, unmasking areas in Fig. 6B are smaller and more sharply confined than in Fig. 6A. Latencies are almost independent off,, in both cases. These two cases illustrate the inter-subject unmask-
.I
5
lOkH(r
C
Fig. 6. A. Subject 130181, as in Fig. 4, for two stimulus levels of test tone. Solid curves are for the low-level situation (+symbol in the audiogram; test tone and masker: 3 kHz, 50 dB HL; suppressor: 80 dB SPL). Dashed curves are for the high-level situation ((X) symbol in the audiogram; test tone and masker: 3 kHz, 75 dB HL; suppressor 105 dB SPL). B. AS in A, for subject 210181. Solid curves: low level U( A/) experiment (+ symbol in the audiogram; test tone and masker 3 kHz, 30 dB HL; suppressor: 70 dB SPL). Dashed curves: high level U(Af) experiment (X symbol in the audiogram; test tone and masker 3 kHz, 60 dB HL; suppressor: 90 dB SPL). C. Subject 030281. As in Figs. 6A and B, but now test tone level and frequency have been varied. Solid curves: low level/high fT situation (-t symbol in the audiogram; test tone and masker: 3 kHz, 35 dB HL; suppressor 65 dB HL). Dashed curves: high level/low-frequency condition (X symbol in the audiogram; test tone: I kHz, 70 dB HL; masker: 1 kHz, 80 dB HL; suppressor: 110 dB SPL).
unmasking variability, considering the fact that test tone levels were not essentially different. In subject 030281 (Fig. 6C), who belongs to the non-normal audiogram group, two U(Afj experiments have been performed, thereby varying test tone frequency. Because of the strongly sloping audiogram a change of the test frequency implies a change of test level. Test tone parameters were 3 kHz/35 dB HL (+ symbol, drawn curve) and 1 kHz/70 dB HL (X symbol, dashed curve). In both situations the
168
high-frequency unmasking is much more pronounced than its low-frequency counterpart, the 1 kHz situation yielding enormous unmasking (dashed curve). Note that the high-frequency unmasking maximum in the latter case is situated at a higher frequency than the corresponding maximum in the fT = 3 kHz case. In the f, = 1 kHz situation there is a large latency increase at f, = 3 kHz (dashed latency curve). This increase is relatively larger than in any of the previous fT,= 3 kHz cases.
C, U(A f) as well as U(AI) experiments Fig. 7 demonstrates U( Af >results (left part, fT = 4 kHz) as well as U( AI) results (right part, fT = 4 kHz, f, = 1 kHz) in one subject (301080). In spite of very large hearing losses unmasking is pronounced under both conditions. The U( AI) experiment reveals AP/AP,,, values in excess of 100% between Z, = 15 dB and I, = 90 dB. Above 90 dB the suppressor acts as a second masker, as AP/AP,,, falls below 100%. Note that in the U(Al) case sub-hearing-threshold suppressor intensities bring about very convincing unmasking effects (0 dB in the audiogram corresponds to 10 dB SPL of the 1 kHz tone burst). In other words, large hearing losses at the suppressor frequency do not prevent suppressive action at other sites along the cochlear partition. D. U(Afl, U(AI) and tuning In a recovered sudden deafness subject (071180, Fig. 8A) we performed a combined unmasking/tuning experiment. The U(Af ) experiment (upper left part of Fig. 8A) demonstrates the absence of high-frequency unmasking. Inspection of the AP tuning curves (lower left part) shows differences in tuning between the forward masking condition (open circles) and the simultaneous masking condition (closed circles), differences being most pronounced at the low-frequency side. Dallos and Cheatham [2] interpret such difference-areas as ‘suppression areas’. According to this
301080
&
MMCY
I2ms
0
d0 LO 60
.5 kHz
i
Fig.7. Subject 301080. A U(Af) as well as a U(Al) experiment have been performed. Left: U(A/) experiment (test tone and masker: 4 kHs 65 dB HL; suppressor: 95 dB SPL). Right: U(AI) experiment (test tone and masker: 4 kHz, 75 dB HL; suppressor: 1 kW.
169
interpretation, the unmasking should indeed be most pronounced at the lowfrequency side. See also [28]. Fig. 8B summarizes U( A f )-unmasking (upper part of Fig. 8B) and (simultaneous masked) tuning results (lower part) in a sudden deafness case (subject 171080), suffering from very large low-frequency hearing losses. In spite of these losses the
.5
10 kWz
5
1
TUNING
.5
1
2
A
250260
5wz
to
.I
1
5
10 kWz
B
IJNMAwNt
C
Fig. 8. A. Subject 071180. Top: U(Af) experiment (lest tone and masker: 3 kHz, 40 dB HL; suppressor: 75 dB SPL). Bottom: AP tuning curve. 0, forward masking; 0, simultaneous masking condition. Test tone: 3 kHt, 50 dB SPL (+ symbol). The horizontal arrow indicates the level of the suppressor tone in the U(Af) experiment. B. As in A, for subject 171080. AP tuning curve was obtained for the simultaneous masking condition alone. U( Af) stimuli were: test tone and masker: 3 kHz, 70 dB HL; suppressor: 95 dB SPL. Filled circle in the tuning plot indicates test tone level and frequency. C. U(AI) and tuning in subject 250280. U(AI) parameters: test tone: 2 kHz, 60 dB HL; masker: ZkHz, 70 dB HL; suppressor frequencies: 2.5 kHz (+), 3 kHz (0), 3.5 kHz (X). Vertical arrows in the tuning curve denote frequency and intensity variation of suppressor tones in the &AI) experiment. Tuning was measured with simultaneous procedures.
masker level
AUDIOGRAM
1 oo4:.
.i
i
010
50.
7 fm 0
,...I 0.5
I 1
2
.“I
5 kHz
10
Fig. 9. Left: AP tuning curves of Figs. 8A, 8, C (0, V, A) together with those of three other subjects (150279a (0). 150279b (X), 180880 (+)); QIwe values of these curves have been plotted versus hearing loss (at j7) in the lower right part of the figure. The figures combine data obtained with forward and simultaneous masking procedures.
tuning curve has an identifiable, albeit broad, tip and also there are high- as well as low-frequency unmasking effects. A combination of (simultaneous masked) tuning and U(AZ)-unmasking is shown in Fig. K, subject 250280. The ECoC audiogram has been drawn in between, losses are 50-60 dB, except at 2 kHz where threshold elevation is only 25 dB. Tuning at 2 kHz is rather well preserved. The three vertical arrows in the tuning plot indicate how in the U(AZ) experiment suppressor frequency and intensity were varied. The effect of this variation is shown in the lower right part: no U(AZ)-unmasking has been found. Latencies are almost independent of suppressor intensity and frequency.
E. Tuning cumes
AP tuning curves of the latter three cases are summarized in the left part of Fig. 9, together with three tuning curves in subjects in which no unmasking experiments were performed. Forward masking was applied in subjects 071180 and 180880, simultaneous masking in the remai~ng four subjects. Audiograms are collected in the right upper part of Fig. 9. In the lower right part of Fig. 9 QIodB values (i.e. ratio of test frequency and frequency-width at 10 dB above the tip) are plotted as a function of hearing loss at the test frequency. On the average, AP tuning is observed to become progressively less sharp with increasing hearing loss. Qlode varies between 8 (subject I50279a, hearing loss at f,,, is -5 dB) and about 0 (subject 180880,45 dB loss at /,,,). The dashed line in the lower right of the figure illustrates the gradual decrease in between these two extremes. V. Discussion
The foregoing results allow the following answers to the first two questions, put forward in the Introduction: (I) Unmasking can be observed in compound AP
171
responses in normal as well as in abnormal ears. (2) Upon inspection of the unmasking/audiogram figures, one observes that AP unmasking is not essentially different in normals, compared to abnormals. Also, the degree of pathology, taking the audiogram as an indicator for this, does not seem to be of decisive importance for the amplitude and extent of AP unmasking. Peaked, restricted unmasking curves as well as flat, broad ones, have been measured in normals as well as abnormals. Ten out of twelve subjects show bilateral, i.e. both low-frequency as high-frequency, unmasking. Only in two subjects (‘normals’ 071180 and 251180) unmasking was observed unilaterally. Most interesting, their unmasking maxima (237 and 200%) are higher than in any of the bilateral cases. Figs. 6A and B show that an increase of the test tone level (and consequently also of the masker levels) dramatically shifts the position of the low-frequency unmasking area towards lower frequencies and that of the high-frequency area towards higher frequencies. This can be understood (see also Fig. 1A) by assuming (1) that tuning curves shift shape-invariantly, i.e. retaining the same shape in the frequency-intensity plane, towards higher level upon increasing the test tone level, and (2) at the same time, that flanking suppression areas do not shift shape invariantly towards higher levels, but maintain more or less their absolute position. These two suppositions being valid, cross-sectioning horizontally the tuning curve at two suppressor levels implies that at the higher level suppression areas will be left and entered at lower and higher frequencies at the low- and high-frequency side of the tuning curve, respectively, compared to the situation at the lower suppressor level. The first supposition seems valid on basis of earlier AP tuning work [2] (psychophysically the same was observed by Wightman et al. [28]). We are not aware of reports in the literature about test tone level variation in AP-unmasking experiments which might offer arguments in favour of or against the second supposition. Thus, our results indicate that in humans AP suppression areas shift less upward (or not at all) than the AP tuning does upon increasing the test tone level. Possibly, this means that AP suppression and tuning measure different aspects of cochlear auditory analysis and will be not necessarily correlated (see also the Appendix).
VI. Results - parameter analysis
For further analysis of the data we focus on (1) relationships between unmasking and hearing ‘loss, and (2) dependence of unmaskmg on stimulus parameters (level difference Z, - I,,,, absolute and sensation levels Z and I,). Unmasking
versus hearing loss
From the literature the notion appears that non-linear behaviour is reduced by pathology (see Introduction). Taking hearing loss as a gross indicator for pathology one might expect that unmasking will be progressively reduced as a function of increasing hearing loss. Figs. 1OA and B relate AP/AP,,, values to hearing loss. Data have been taken from V( Af ) figures presented above (12 subjects). Not included are the f, =fr data points. As is evident from Figs. 10A and B, no correlation between
172
AP/AP,,, and hearing loss can be observed, neither at low nor at high frequencies. Upon considering exclusively the unmasking maxima (closed circles in Figs. lOA, B) again no correlation is found. Scattering is about as wide as for the total ensemble. Unmasking versus stimulus parameters Shannon [24] states that (1) for & f, unmasking depends on Z, - I,. Duifhuis (1980) points at the limited validity of these conclusions and also stresses the large interindividual differences in (psychophysical) unmasking behaviour. Suppressor level variation was brought about in two V( A I) experiments, of which only one (subject 311080, Fig. 7) yielded unmasking. Testing statement (1) would have required studying the effect of Z, variation for different I,,, levels. In general, it is impossible in clinical ECoG procedures to perform such extensive mappings, as forward masking procedures are very time consuming. So, for a parameter evaluation one has to resort to interindividual comparison. Then it is possible to examine Shannon’s first conclusion in our material (as we kept Z, - I,,., always at 20 dB, we are not able to verify his second statement). To this end we plotted low-frequency AP/AP,,, maxima versus suppressor level Z, in Fig. 11. Again, no correlation can be detected, although between the two subgroups there is a positive correlation between suppressor level and unmasking maximum. Instead of relating AP/AP, data to absolute suppressor levels, one may consider AP/AP, versus suppressor sensation level [28,18]. Then, again wide scattering is found at both frequency sides and in both subgroups (results not shown). Mean values In Table II average hearing losses, U(Af )-maxima and summarized for each hearing loss group. Data in parentheses
fscmax)/fT ratios are standard
are devia-
Fig. 10. AP/AP,,, for specific& frequencies (from (I( A/) experiments) versus hearing loss at that specific f,. Left: for frequencies f, < fT. Unmasking maxima have been drawn as filled circles. Right: as left, but forf,>/,.
VALUES
OF WA/)
18.3 (4.5) 43.4 (11) 29.7 (15.2)
ECoG
Mean heating
19.5 (6.7) 42.6 (15.5) 30 (16.3)
Audiogram
loss (dB)
172 (44.7) 162 (34.5) 168 (37.4)
Low freq.
Mean f_!(A/)
maxima
149 (33.9) 143 (9.9) 146 (28.3)
High freq.
(a)
0.54 (0.26) 0.27 (0.13) 0.40 (0.24)
Low freq.
Mean /smaX)/fT
ratio
2.05 (0.81) 2.5 (1.1) 2.25 (0.93)
High freq.
columns).
Average values in two subgroups, as well as in the total group, are given of (1) hearing losses, (2) unmasking maxima, (3) ratio of suppressor frequency at Both hearing losses obtained by electrocochleography unmasking maximum ( Ltm**) ) and test frequency (jr). Standard deviations are given in parentheses. and by pure tone audiometry are given (first two columns). Unmasking maxima are given at the low-frequency side of the test frequency as well as at the high-frequency side (third and fourth columns). Such a low- and high-frequency subdivision has also been made for the f!max)/fT ratio (fifth and sixth
Normals ( n = 6) Abnormals (n = 5) Total (n = I 1)
MEAN
TABLE II
tions. With respect to mean hearing loss, the two groups are well separated. Mean ECoG loss matches mean audiogram loss very well. Statistically, no significant difference between the two subgroup means exists for the U(Af) maxima (t-test).
Fig. 11. Plot of the low-frequency normals; e, abnormals.
unmasking
maxima
versus
corresponding
suppressor
levels.
0,
The same holds for the high frequencyf,‘ma”)/fT ratio. The low frequency_&(max)/ff ratio is a factor of two lower for abnormafs than for normals (however, this is statistically significant only for rr Z 0. I).
VII. General discussion This study shows that the phenomenon of AP unmasking, observed in the chinchilla [IO] and guinea pig [14], also exists in humans. This opens a new way to study suppression phenomena in man electrophysiologically. The method offers advantages to psychophysical methods in which disturbing effects occur, such as ‘off-frequency listening’ or ‘quality difference cues’ [17]. On the other hand, an AP response is an on-effect response, its quality depending on synchronized firing of a number of neurons. This requires the use of stimufi, such as - spectrally rather (but not too) broad - tone bursts. As a consequence, the excited area along basilar membrane is broader than in single neuron or psychophysical experiments. Moreover, upon varying suppressor frequency in an U( Af ) experiment, one is not certain if the excited population of fibers, contributing to the AP, does not change. However, AP latencies can serve as a gross indicator for this. As in the majority of subjects AP later&es were found to show no appreciable change under & variation, we fee1 rather confident that the test tone bursts were appropriate probes. A. Human psychophysics
In the majority of subjects we observed low- as well as high-frequency AP unmasking. Test tone frequencies were nearly always set at 3 kHz. Probe and masker levels were IO-30 dB above threshold, suppressor level was always 20 dB higher.
175
O’Malley and Feth [ 161 investigated psychophysically relationships between forward masked tuning and unmasking in the 3 kHz frequency region in normal ears. Their stimulus ensemble (in the ‘quiet’ condition) is comparable to ours (except no time interval between masker and probe). The authors find qualitatively a positive covariation between high-frequency tuning curve slope and high-frequency unmasking maximum. In their opinion this tendency is in line with Wightman’s et al. [28] unmasking results in hearing impaired subjects. They observed in two subjects the absence of unmasking in elevated thresholds regions, accompanied by reduced difference (on the high-frequency side) between forward and simultaneous tuning curves. This leads them to the firm conclusion: “the suppression is rendered totally ineffective by the hearing loss”, which is obviously in contrast with our AP findings. Their observations parallel ours only in two subjects, i.e. subject 071180 (Fig. 8A, no high-frequency unmasking and very small high-frequency side tuning differences) and subject 251180 (Fig. 5C, no low-frequency unmasking). Possibly, there are a few reasons why Wightman et al. failed to observe unmasking. First their abnormal subjects have large high-frequency losses, while many of our subjects have predominantly low-frequency sloping audiograms. Secondly, their stimulus levels have been chosen such that in the elevated threshold region the suppressor was hardly or not excitatory. Wightman et al. did recognize this and increased suppressor level from 60 to 100 dB SPL. Again, no unmasking was observed. As our subject 130181 (Fig. 6A) demonstrated, such an increase might result in an outward frequency shift of unmasking regions, in their case possibly beyond the applied frequency range. So, even if there was unmasking, it would not have been detected. Thirdly, compound AP responses are reflections of neural on-effect responses, while psychophysically one measures the subjective reflection of the (centrally processed) neural steady state. Suppression mechanisms may not have the same effect for these two temporal stages in case of pathology, resulting in different outcomes. B. Animal AP unmasking
Comparison across species is a dangerous matter. Nevertheless, our human data resemble the normal chinchilla data [IO]. Low- and high-frequency unmasking have been observed in both species - in somecases up to ‘complete unmasking’, although in the chinchilla high-side unmasking is strongest. Also, f,(“‘/fr ratios are comparable (for comparable stimuli). In one abnormal chinchilla, having a broad 20 dB/3 kHz ‘dip’, Harris [lo] observed only low-side unmasking (his Fig. 15) parallelled by the same observation in single fibre experiments. In normal guinea pigs [14] low-frequency unmasking is always less (or absent) than high-frequency unmasking. Complete unmasking is seldom observed in this animal. C. Animal single fibre data
In the single fibre literature there is conflicting evidence about the correlation between experimentally induced hearing loss,’ tuning and suppression, although the
176
general trend seems to be that auditory non-linearity is rather susceptible to cochlear pathology (see also Introduction). For example, Dallos et al. [3] report a tendency for reduced suppression in chinchillas with outer hair cell lesions due to kanamycin ad~nistration, while the high-frequency slope of the tuning curve is not greatly affected. Robertson and Johnstone [19] report an analogous finding in noise-damaged guinea pig cochleas. Schmiedt and Zwislocki (221, however, find that complete loss of high-frequency two-tone suppression is associated with shallow high-frequency slopes of the tuning curve after kanamycin poisoning. They also find a concurrent diminishing of low-frequency suppression effects. Evans [S] points at the non-identical behaviour of the two ‘side-bands’ of suppression. Low-frequency bands appear to be much more resistant to pathology, in cat and guinea pig, than high-frequency bands. They “may even be unchanged under conditions of elevation in CF threshold by over 40 dB”. The latter finding supports our suggestion (section V) about suppression regions m~nt~ning their absolute position in frequency-intensity coordinates. From the reports cited above and our own data, it becomes clear that aspects of non-linear behaviour are rather intricately related to each other, depending on which type of damage was introduced and in which species. Also, within a species, degree and spatial extent of cochlear lesions are important determinants. Therefore, as we do not know type and extent of cochlear lesions in our subjects, comparison with these animal data is difficult. However, as far as we could trace, none of our subjects had a history of ototoxic damage, which could imply that the cited animal results do not really contrast with our findings.
(1) AP unmasking has been observed in man, both in the normal as well as in the abnormal hearing threshold group. (2) In the majority of subjects U( Al) unmasking is present both forf, maxima, f,/fT ratios, size and shape of the unmasking areas is large, in both subgroups. (4) Quality of AP tuning deteriorates with increasing hearing loss. As unmasking maxima do not correlate with hearing loss, AP tuning and AP unmasking are not correlated. This might imply that suppression and tuning measure different aspects of cochlear non-linearity, (5) On an individual basis (and for the applied stimulus conditions) it was found that reduced difference between forward and simultaneously masked tuning curves (on the high-frequency side) went together with absence of high-frequency unmasking.
171
Appendix - Note added in manuscript
After preparation of this article we became aware of results of R.A. Schmiedt (J. Acoust. Sot. Am. 70, S9 (1981) abstract) on two-tone suppression boundaries of cochlear nerve fibers in gerbil. His data suggest that “..... the tuning curve is a locally determined phenomenon along the cochlear duct, whereas two-tone rate suppression is determined more globally and to a large extent independent of the tuning curve” (op. cit.). We regard these conclusions as an argument in favour of our second supposition in section V, i.e. about flanking suppression areas maintaining their absolute position in the frequency-intensity plane.
Acknowledgements
The critical reading of the manuscript by Professors J.J. Eggermont and E. de Boer is gratefully acknowledged. This work was supported by the Heinsius Houbolt fund and the Dutch Organization for the Advancement of Pure Research (ZWO).
References 1 Arthur, R.M.. Pfeiffer, R.R. and Suga, N. (1971): Properties of ‘two-tone inhibition’ in primary auditory neurones. J. Physiol. (London) 212, 593-609. 2 Dallos, P. and Cheatham, M.A. (1976): Compound action potential (AP) tuning curves. J. Acoust. Sot. Am. 59, 591-597. 3 Dallos, P., Harris, D.M. and Cheatham, M.A. (1980): Two-tone suppression and intermodulation distortion in the cochlea: effect of outer haircell lesions. In: Psychophysical, Physiological and Behavioural Studies in Hearing, pp. 242-252. Editors: C. v.d. Brink and F.A. Bilsen. Delft University Press, Delft, The Netherlands. 4 Duifhuis, H. (1980): Level effects in psychophysical two-tone suppression. J. Acoust. Sot. Am. 67, 914-927. 5 Eggermont, J.J., Odenthal, D.W., Schmidt, P.H. and Spoor, A. (1974): Electrocochleography. Basic principles and clinical application. Acta Otolaryngol. Suppl. 316. 6 Eggermont, J.J. (1976): Analysis of compound action potential responses to tone burst in the human and guinea pig cochlea. J. Acoust. Sot. Am. 60, 1132- 1139. 7 Eggermont, J.J. (1977): Compound action potential tuning curves in normal and pathological human ears. J. Acoust. See. Am. 62, 1247-1251. 8 Evans, E.F. (1980): Comment on: [3]. In: Psychophysical, Physiological and Behavioural Studies in Hearing, pp. 242-252. Editors: G. v.d. Brink and F.A. Bilsen. Delft University Press, Delft. 9 Goldstein, J.L. (1967): Auditory nonlinearity. J. Acoust. Sot. Am. 41, 676-689. 10 Harris, D.M. (1979): Action potential suppression, tuning curves and thresholds: comparison with single fiber data. Hearing Res. 1, 133- 154. 1 I Harrison, R.V., Aran, J.M. and Erre, J.P. (1981): AP tuning curves from normal and pathological human and guinea pig cochleas. J. Acoust. Sot. Am. 69, 1374- 1385. 12 Houtgast, T. (1974): Lateral suppression in hearing. Thesis, Amsterdam. 13 Humes, L.E. (1980): On the nature of two-tone aural nonlinearity. J. Acoust. Sot. Am. 67,2073-2083. 14 Kuper and Rutten, to be published. 15 Leshowitz, B. and Lindstrom, R. (1977): Measurement of non-linearities in listeners with sensorineural hearing loss. In: Psychophysics and Physiology of Hearing, pp. 283-293. Editors: E.F. Evans and J.P. Wilson. Academic Press, London.
178 16 O’Malley, H. and Feth, L.L. (1979): Relationship between psychophysical tuning curves and suppression. J. Acoust. Sot. Am. 66, 1075-1087. 17 Moore, B.C.J. (1980): Detection cues in forward masking. In: Psychophysical, Physiological and Behavioural Studies in Hearing, pp. 222-229. Editors: G. v.d. Brink and F.A. B&en, Delft University press, Delft. 18 Penner, M.J. (1980): Two-tone forward masking patterns and tinnitus. J. Speech Hearing Res. 23, 779-786. 19 Robertson, D. and Johnstone, B.M. (1981): Primary auditory neurons: Non-linear responses altered without changes in sharp tuning. J. Acoust. Sot. Am. 69, 1096- 1098. 20 Robertson, D. (1976): Correspondence between sharp tuning and two-tone inhibition in primary auditory neuronea. Nature (London) 259.477-478. 21 Sachs, M.B _and Kiang, N.Y.-S. (1968): Two-tone in~bition in auditory nerve fibers. J. ACOW. f&. Am. 43, 1120-I 128. 22 Schmiedt, R.A. and Zwislocki, fibers. II. Single and two-tone
J.J. (1980): Effects of hair cell lesions on responses of cochlear nerve intensity functions in relation to tuning curves. J. Neurophysiol. 43,
1390-1405. 23 Schmiedt, R.A., Zwislocki. J.J. and Hamernik, R.P. (1980): Effects of hair cell lesions on responses of cochlear nerve fibres. I. Lesions, tuning curves, two-tone inhibition and responses to trapezoidal wave patterns. J. Neur~hysjol. 43, 1367- 1389. 24 Shannon, R.V. (1976): Two-tone unmasking and suppression in a forward masking situation. J. Acoust. Sot. Am. 59, 1460-1470. 25 Smoorenburg, G.F. (1972): Combination tones and their origin. J. Acoust. Sot. Am. 52, 615-632. 26 Smoorenburg, G.F. (1980): Effects of temporary threshold shift on combination tone generation and or two-tone suppression. 27 W&r, D.L. and Moore,
Hearing Res. 2, 347-355. B.C.J. (1981): Forward masking
by sinusoidaf
and noise maskers.
J. Acoust.
Sot. Am. 59, 1402-1409. 28 Wightman, F., McGee, T. and Kramer, M. (1977): Factors influencing frequency selectivity in normal and hearing impaired listeners. In: Psychophysics and Physiology of Hearing, pp. 295-308. Editors: E.F. Evans and J.P. Wilson. Academic Press, London.