Recent auditory brain stem response measurements in tinnitus subjects with normal audiograms indicate the presence of hidden hearing loss that manifests as reduced neural output from the cochlea at high sound intensities, and results from mice suggest a link to deafferentation of auditory nerve fibers. As deafferentation would lead to deficits in hearing performance, the present study investigates whether tinnitus patients with normal hearing thresholds show impairment in intensity discrimination compared to an audiometrically matched control group. Intensity discrimination thresholds were significantly increased in the tinnitus frequency range, consistent with the hypothesis that auditory nerve fiber deafferentation is associated with tinnitus.

Normal hearing thresholds have long been regarded as an indicator for the absence of cochlear damage. However, it has recently been shown in mice that sound exposure at nightclub noise levels, which leads to a temporary shift of the hearing thresholds, also causes permanent deafferentation of a large fraction of the auditory nerve (AN) fibers in the high-frequency range (Kujawa and Liberman, 2009). The effect of this cochlear damage also manifested in a marked reduction of the amplitude of the wave I potential (generated through the activation of AN fibers) of the auditory brain stem response (ABR) at high sound intensities, whereas responses at low sound intensities were almost unchanged, suggesting that the deafferentation predominantly affected high-threshold fibers. A similar reduction of ABR-wave I at high sound intensities has also been observed in subjects with tinnitus and a normal audiogram (Schaette and McAlpine, 2011). Moreover, normal-hearing tinnitus subjects also showed increased tone-detection thresholds in threshold-equalizing noise at high intensities, which has been interpreted as evidence for central deafferentation despite normal hearing thresholds (Weisz et al., 2006).

The wide dynamic range of sounds in the acoustic environment is covered by the different types of AN fibers: The dynamic range of the response of high spontaneous rate fibers covers low sound intensities, whereas fibers with medium and low spontaneous firing rates are sensitive to higher sound intensities (Liberman and Kiang, 1978; Yates et al., 1990). Deafferentation of AN fibers might thus impair the ability to discriminate sound intensities as deafferentation reduces the number of fibers that are available to resolve intensity differences. Specifically, when high-threshold fibers are deafferented, as it seems to be the case in noise-exposed mice, one would expect to see a deficit at medium to high sound intensities.

If tinnitus patients with normal audiograms do have deafferentation of AN fibers, as suggested by ABR measurements (Schaette and McAlpine, 2011) and hearing tests in background noise (Weisz et al., 2006), they should thus also show an impairment in an intensity discrimination task. Moreover, if the deafferentation is linked to the development of tinnitus, as indicated by computational modeling (Schaette and McAlpine, 2011), the effect should be most pronounced in the tinnitus frequency range. To test this hypothesis, just-noticeable intensity differences (intensity JNDs) for pure tones were measured in normal hearing listeners reporting tinnitus and in control participants with normal hearing. In a first experiment, pure-tone detection thresholds were measured at the relevant probe frequencies, including a frequency in the peak region of the average tinnitus spectrum of the participants. A band-pass noise was simultaneously presented to reduce off-frequency listening in adjacent auditory filters. In a second experiment, intensity JNDs were measured for the same frequencies in the presence of the same maskers.

For each listener, hearing thresholds in quiet were measured using a calibrated clinical audiometer for frequencies of 125–8000 Hz. Subsequently, masked thresholds and pure tone intensity JNDs at frequencies of 1000, 2450, and 6000 Hz were assessed. Measurements were done monaurally for both ears of each listener. For both measures, an adaptive three-alternative forced-choice procedure with a one-up-two down rule was used, resulting in the 70.7% point of the psychometric function (Levitt, 1971). Signals were generated in matlab with a sampling rate of 96000 Hz and 32-bit resolution. Signals were played back via headphones (Sennheiser HDA200) using an external soundcard (RME fireface 400) in a sound-proof booth. Each interval had a duration of 500 ms, and the intervals were separated by 500 ms of silence. Visual feedback was given after each trial. Off-frequency listening was reduced by presenting two additional one-half octave wide noise bands with a total level of 50 dB SPL above and below the signal of the tone. The noise bands were generated by bandpass-filtering a white noise using a eighth-order Butterworth bandpass filter at center frequencies one octave above and below the target signal frequency.

Three intervals were presented to the listener, each containing an independent realization of the masker. One randomly chosen interval also contained the target signal. The listener had to indicate which of the intervals contained the target signal. The initial level of the target signal was 8.5 dB SPL. The step size of the adaptive procedure was initially 4 dB and was halved after each upper reversal of the tracking procedure until the smallest step size of 1 dB was reached. The arithmetic mean of six reversals using the smallest step size was used to estimate the threshold. Individual thresholds were calculated as the arithmetic mean of three repetitions.

Three intervals were presented to the listener, each containing an independent realization of the masker and the tone at a fixed level (in the following referred to as reference). The reference tone had levels of 30, 50, or 70 dB SPL (in the following referred to as reference level). In one randomly chosen interval, another tone with same the frequency was added in phase to the reference tone, i.e., the amplitude of the resulting tone was equal to the sum of the amplitudes of the reference and the added tone. The listener had to indicate in which interval the intensity of the tone was higher. The step size of the intensity increment calculated as ΔI/I was initially set to 0.3 and was halved after each upper reversal until the smallest increment of 0.0375 was reached. The arithmetic mean of six reversals using the smallest step size was used to estimate the JND. Individual JNDs were calculated as the arithmetic mean of three repetitions.

Tinnitus pitch was characterized using a modified version of the tinnitus spectrum approach (Norena et al., 2002). Comparison sounds (pure tones of 250, 500, 1000, 1500, 2000, 3000, 4000, 6000, 8000, 12 000, and 16 000 Hz) were generated using custom-made matlab software and were presented via Sennheiser HD600 headphones. The comparison tones were first matched to the loudness of the tinnitus using a single-interval adaptive procedure (Lecluyse and Meddis, 2009). During loudness matching, sound intensity was limited to levels ≤100 dB SPL. To obtain tinnitus pitch similarity ratings, the loudness-matched tones were presented in random order, and following each presentation, participants were asked to rate the similarity between the pitch of the comparison tone and their tinnitus on a scale of 0–10, with 0 for completely different and 10 for extremely similar. Each tone was presented and rated three times. For comparison tones of 16 000 Hz, a loudness match could not be achieved for all participants, and thus the results for this frequency were excluded from further analysis.

Fourteen listeners without tinnitus (control group, mean age 26 ± 2 yr, 10 female) and 11 listeners who reported a tinnitus (tinnitus group, mean age 38 ± 4 yr, all female) participated in the experiment. All listeners had pure-tone thresholds in quiet ≤20 dB HL in the frequency range from 250 to 8000 Hz. One of the listeners of the tinnitus group reported a tinnitus in the left ear only, and there were two listeners with tinnitus dominant in the left ear and two listeners with tinnitus dominant in the right ear. The remaining tinnitus subjects reported a tinnitus in either both ears or located within the head. For the participant with unilateral tinnitus, only the tinnitus ear was included in the analysis.

To test for significant differences, the Mann–Whitney U-test was employed. For audiometric and masked thresholds, two-sided tests were used. For comparing intensity JNDs between the tinnitus and the control group, one-sided tests were used because it was expected to see higher intensity JNDs in the tinnitus group. Differences were considered to be significant for P < 0.05. All data analysis was carried out using matlab.

The mean audiograms of the listener groups are shown in Fig. 1(a) (gray, control; black, tinnitus). There were no significant differences in hearing thresholds of the two groups at any frequency. Tinnitus spectrum measurements showed that pure tones of 6 and 8 kHz were rated as most similar to the tinnitus pitch [Fig. 1(b)].

FIG. 1.

Audiograms and tinnitus spectra. (a) Mean audiograms of all tinnitus (black) and control participants (gray). (b) Average tinnitus spectrum of the tinnitus participants, errorbars indicate ±1 standard error of the mean. Tinnitus pitch similarity ratings, on a scale from 0 to 10 with 0 for completely different and 10 for extremely similar, were obtained using pure tone stimuli.

FIG. 1.

Audiograms and tinnitus spectra. (a) Mean audiograms of all tinnitus (black) and control participants (gray). (b) Average tinnitus spectrum of the tinnitus participants, errorbars indicate ±1 standard error of the mean. Tinnitus pitch similarity ratings, on a scale from 0 to 10 with 0 for completely different and 10 for extremely similar, were obtained using pure tone stimuli.

Close modal

Masked thresholds and intensity JNDs at 1000, 2450, and 6000 Hz were measured (see Sec. II). The highest frequency was chosen to be in the peak region of the average tinnitus spectrum [Fig. 1(b)]. The average masked thresholds and intensity JNDs for control (n = 28 ears) and tinnitus participants (n = 21 ears) are shown in Fig. 2 (gray lines, control; black lines, tinnitus).

FIG. 2.

Masked thresholds (vertical lines) and intensity JNDs, grand mean over all control (gray) and tinnitus (black) participants. Results for target signal frequencies of 1000 Hz are shown in (a), 2450 Hz in (b), and 6000 Hz in (c). Intensity JNDs are shown as intensity increments Δ I/I as a function of the reference level. An asterisk indicates a statistical significant difference between control and tinnitus group (P < 0.05). Errorbars indicate ±1 standard error of the mean.

FIG. 2.

Masked thresholds (vertical lines) and intensity JNDs, grand mean over all control (gray) and tinnitus (black) participants. Results for target signal frequencies of 1000 Hz are shown in (a), 2450 Hz in (b), and 6000 Hz in (c). Intensity JNDs are shown as intensity increments Δ I/I as a function of the reference level. An asterisk indicates a statistical significant difference between control and tinnitus group (P < 0.05). Errorbars indicate ±1 standard error of the mean.

Close modal

For signal frequencies of 1000 and 2450 Hz, the intensity JNDs were rather similar for reference levels of 30 and 50 dB SPL and showed a tendency to decrease for a reference level of 70 dB SPL. This pattern was observed in both the tinnitus and the control group. At 1000 Hz, there were no significant differences in JNDs between the control and the tinnitus group at all levels. At 2450 Hz, the average JND of the tinnitus group was significantly higher for a reference level of 70 dB SPL. For a signal frequency of 6000 Hz, the pattern of intensity JNDs vs sound intensity differed between the tinnitus and the control group with the tinnitus group displaying a more pronounced “mid-level hump.” JNDs did not differ significantly at reference levels of 30 and 70 dB SPL but were significantly higher in the tinnitus group at a reference level of 50 dB SPL.

The presence of increased intensity JNDs for 6000 Hz at 50 dB SPL but not at 70 dB SPL in the tinnitus group might seem at odds with the hypothesized deafferentation of high-threshold AN fibers as one would expect this kind of cochlear damage to lead to intensity discrimination deficits at medium and high sound intensities. However, a comparison of the present results to ABR data from noise-exposed mice with histologically confirmed deafferentation of AN fibers (Kujawa and Liberman, 2009) shows a high similarity that offers a putative explanation: Eight weeks after exposure to an octave band noise at 100 dB SPL, the ABR-wave I response of the mice showed a substantial decrease in the slope of the amplitude growth function at levels around 40–60 dB SPL, whereas the slope was similar to control mice for higher sound intensities [Kujawa and Liberman, 2009, see also Fig. 3(b) of the present study]. Assuming that the detection of an intensity increment is coded as an intensity-dependent increment in neural activity at the level of the auditory nerve, a shallower slope in the input-output (I/O) function would require a larger input increment to result in the same activity increment in the auditory nerve. The JNDs of tinnitus and control listeners differ for a reference level of 50 dB SPL, where the ABR I/O function of the noise-exposed mice is shallower (Fig. 3), while at 70 dB SPL, where the slope of the ABR I/O functions of noise-exposed and control mice is the same (Fig. 3), there is no significant difference between the control group and the tinnitus group. Therefore our intensity JND results are generally consistent with electrophysiological data from mice with deafferentation of AN fibers.

FIG. 3.

Comparison of psychoacoustical data with mouse auditory brain stem response (ABR) data. (a) Intensity JNDs of tinnitus and control listeners for a frequency of 6000 Hz. (b) ABR-wave I amplitude (tone pip stimuli, 32 000 Hz) of control mice (gray) and of noise-exposed mice eight weeks after exposure to octave-band noise (8000–16 000 Hz) at 100 dB SPL for 2 h (black), data courtesy of Kujawa and Liberman (2009). ABR-wave I, reflecting the summed activity of AN fibers, showed a particularly shallow growth around 50 dB SPL in the exposed mice.

FIG. 3.

Comparison of psychoacoustical data with mouse auditory brain stem response (ABR) data. (a) Intensity JNDs of tinnitus and control listeners for a frequency of 6000 Hz. (b) ABR-wave I amplitude (tone pip stimuli, 32 000 Hz) of control mice (gray) and of noise-exposed mice eight weeks after exposure to octave-band noise (8000–16 000 Hz) at 100 dB SPL for 2 h (black), data courtesy of Kujawa and Liberman (2009). ABR-wave I, reflecting the summed activity of AN fibers, showed a particularly shallow growth around 50 dB SPL in the exposed mice.

Close modal

Nine of the 11 tinnitus listeners also participated in a previous study by Schaette and McAlpine (2011), where a reduced wave I amplitude of the ABR was found in listeners with tinnitus and a normal audiogram. The deficit in intensity JNDs found in the present study in the tinnitus listeners might thus be linked to reduced ABR-wave I amplitudes as found in mice with AN fiber deafferentation (Kujawa and Liberman, 2009). Taken together, these two results indicate that deafferentation of AN fibers could underlie tinnitus with a normal audiogram, which is further supported by the finding that listeners with tinnitus and normal hearing thresholds also show increased tone detection thresholds in high-level background noise in the tinnitus frequency range (Weisz et al., 2006).

The two groups of subjects were matched with respect to their audiograms, but the medium age was higher in the tinnitus group than in the control group. He et al. (1998) found higher intensity JNDs for low (40 dB) and high (80 dB) reference levels for older subjects (68–77 yr) than for younger subjects (22–33 yr). This may account for the (in most cases not significant) tendency of slightly higher JNDs in the tinnitus group for the 1000 - and 2450-Hz data. It is, however, unlikely that age effects account for 6000-Hz data, because (i) the effect of age in He et al. (1998) is considerably smaller for their highest frequency (4000 Hz) and (ii) the JNDs of the present study are the same for both groups of subjects at the low and high reference levels and only significantly higher at 50 dB.

Apart from peripheral damage, changes in central auditory processing could have influenced intensity discrimination thresholds in the tinnitus group. For example, sounds that are similar to the tinnitus might be processed by the same neuronal networks that generate the tinnitus, and thus they could also be affected by altered neuronal processing. Computational modeling studies have provided indications of how tinnitus might alter information processing in the central auditory system. For tinnitus after hearing loss (Schaette and Kempter, 2009) or AN fiber deafferentation with normal hearing thresholds (Schaette and McAlpine, 2011), the models predict that plasticity trying to compensate for the resulting loss of AN signal will lead to an increase in neuronal response gain in the central auditory system. Increased gain in the tinnitus frequency range could also influence intensity discrimination. Future studies could try to address this in more detail by combining psychophysical and evoked potential measurements to study the relation between AN fiber deafferentation, psychophysical performance and neuronal responses in the central auditory system.

We would like to thank Sharon Kujawa and Charles Liberman for providing us with the mouse ABR data used for Fig. 3. This work was supported by the British Tinnitus Association and the Deutsche Forschungsgemeinschaft (SFB/TRR 31).

1.
He
,
N.
,
Dubno
,
J. R.
, and
Mills
,
J.
(
1998
). “
Frequency and intensity discrimination measured in a maximum-likelihood procedure from young and aged normal-hearing subjects
,”
J. Acoust. Soc. Am.
103
,
553
565
.
2.
Kujawa
,
S. G.
, and
Liberman
,
M. C.
(
2009
). “
Adding insult to injury: cochlear nerve degeneration after temporary noise-induced hearing loss
,”
J. Neurosci.
29
,
14077
14085
.
3.
Lecluyse
,
W.
, and
Meddis
,
R.
(
2009
). “
A simple single-interval adaptive procedure for estimating thresholds in normal and impaired listeners
,”
J. Acoust. Soc. Am.
126
,
2570
2579
.
4.
Levitt
,
H.
(
1971
).
“Transformed up-down methods in psychoacoustics”
J. Acoust. Soc. Am.
49
,
467
477
.
5.
Liberman
,
M. C.
, and
Kiang
,
N. Y. S.
(
1978
). “
Acoustic trauma in cats—cochlear pathology and auditory-nerve activity
,”
Acta Otolaryngol.
358
,
5
63
.
6.
Norena
,
A. J.
,
Micheyl
,
C.
,
Chery-Croze
,
S.
, and
Collet
,
L.
(
2002
).
“Psychoacoustic characterization of the tinnitus spectrum: Implications for the underlying mechanisms of tinnitus”
Audiol. Neurootol.
7
,
358
369
.
7.
Schaette
,
R.
, and
Kempter
,
R.
(
2009
). “
Predicting tinnitus pitch from patients’ audiograms with a computational model for the development of neuronal hyperactivity
,”
J. Neurophysiol.
101
,
3042
3052
.
8.
Schaette
,
R.
, and
McAlpine
,
D.
(
2011
). “
Tinnitus with a normal audiogram: Physiological evidence for hidden hearing loss and computational model
,”
J. Neurosci.
31
,
13452
13457
.
9.
Weisz
,
N.
,
Hartmann
,
T.
,
Dohrmann
,
K.
,
Schlee
,
W.
, and
Norena
,
A. J.
(
2006
). “
High-frequency tinnitus without hearing loss does not mean absence of deafferentation
,”
Hear. Res.
222
,
108
114
.
10.
Yates
,
G. K.
,
Winter
,
I. M.
, and
Robertson
,
D.
(
1990
). “
Basilar membrane nonlinearity determines auditory nerve rate-intensity functions and cochlear dynamic range
,”
Hear. Res.
45
,
203
219
.