Temporal modulation transfer functions obtained using sinusoidal carriers with normally hearing and hearing-impaired listeners.

Temporal modulation transfer functions were obtained using sinusoidal carriers for four normally hearing subjects and three subjects with mild to moderate cochlear hearing loss. Carrier frequencies were 1000, 2000 and 5000 Hz, and modulation frequencies ranged from 10 to 640 Hz in one-octave steps. The normally hearing subjects were tested using levels of 30 and 80 dB SPL. For the higher level, modulation detection thresholds varied only slightly with modulation frequency for frequencies up to 80 Hz, but decreased for high modulation frequencies. The decrease can be attributed to the detection of spectral sidebands. For the lower level, thresholds varied little with modulation frequency for all three carrier frequencies. The absence of a decrease in the threshold for large modulation frequencies can be explained by the low sensation level of the spectral sidebands. The hearing-impaired subjects were tested at 80 dB SPL, except for two cases where the absolute threshold at the carrier frequency was greater than 70 dB SPL; in these cases a level of 90 dB was used. The results were consistent with the idea that spectral sidebands were less detectable for the hearing-impaired than for the normally hearing subjects. For the two lower carrier frequencies, there were no large decreases in threshold with increasing modulation frequency, and where decreases did occur, this happened only between 320 and 640 Hz. For the 5000-Hz carrier, thresholds were roughly constant for modulation frequencies from 10 to 80 or 160 Hz, and then increased monotonically, becoming unmeasurable at 640 Hz. The results for this carrier may reflect "pure" effects of temporal resolution, without any influence from the detection of spectral sidebands. The results suggest that temporal resolution for deterministic stimuli is similar for normally hearing and hearing-impaired listeners.

The influence of external and internal noise on the detection of increments and decrements in the level of sinusoids.

This paper examines the influence of external and internal noise on the detection of increments and decrements in the level of sinusoidal pedestals. In experiment 1, the pedestals were presented either 18 dB above the masked threshold in broadband noise (condition 18-Masked) or 18 dB above the absolute threshold (condition 18-Abs). Pedestal frequencies were 250, 1000 or 4000 Hz, and increment/decrement durations ranged from 5 to 200 ms. For condition 18-Masked, thresholds decreased with increasing pedestal frequency, while for condition 18-Abs, thresholds did not change significantly with pedestal frequency. These results are consistent with the idea that, in condition 18-Masked, thresholds were influenced by the inherent fluctuations produced by the background noise at the output of the auditory filter centred at the pedestal frequency. These fluctuations would decrease in rate with decreasing centre frequency, and this might have a greater deleterious effect on performance. In contrast, the characteristics of the internal noise that presumably limited performance in condition 18-Abs do not appear to vary with pedestal frequency. In experiment 2, a 4000 Hz pedestal was used. It was presented either in quiet or in the presence of narrowband noise centred at 4000, or 7000 Hz, or both. The noise bandwidth ranged from 50 to 400 Hz. The increment/decrement duration ranged from 5 to 100 ms. The noise centred at 7000 Hz produced only a small deterioration in performance relative to that measured in quiet. The noise centred at 4000 Hz had a larger effect, and the effect increased with decreasing noise bandwidth. This is consistent with the idea that slow fluctuations at the output of the auditory filter impair increment and decrement detection more than rapid fluctuations. A model is proposed to account for the results, based on a simulated auditory filter, a compressive non-linearity, a sliding temporal integrator, a logarithmic transform and a template mechanism. Analysis using the model suggests that the effect of centre frequency observed in experiment 1, when background noise was present, cannot be explained entirely in terms of the fluctuations produced by the background noise at the output of the auditory filter centred at the pedestal frequency.

Effects of slow- and fast-acting compression on the detection of gaps in narrow bands of noise.

The inherent amplitude fluctuations in narrow bands of noise may limit the ability to detect gaps in the noise; 'dips' in the noise may be confused with the gap to be detected. For people with cochlear hearing loss, loudness recruitment may effectively magnify the fluctuations and this could partly account for the reduced ability to detect gaps in noise bands that is usually found for such people. Previously, we tested these ideas by processing the envelopes of noise bands to alter the amount of envelope fluctuation. We showed that instantaneous compression, implemented via processing of the Hilbert envelope, led to smaller (that is, better) gap detection thresholds for subjects with cochlear hearing loss. In the present experiment, we determined whether fast-acting compression of the type sometimes used in hearing aids could also lead to improved gap detection. A behind-the-ear (BTE) digital hearing aid was programmed to implement multi-band compression, either fast-acting or slow-acting (control condition). A reference condition using unaided listening was also used. Stimuli were delivered via an earphone placed over the hearing aid. Overall stimulus levels at the output of the hearing aid were similar across conditions. Thresholds for detecting gaps in noise bands centred at 4 kHz were measured as a function of noise bandwidth (10-500 Hz). To prevent the detection of spectral changes introduced by the gap, stimuli were presented in a broad-band background noise. Three normally hearing subjects and three subjects with bilateral cochlear hearing loss were tested. Gap thresholds varied non-monotonically with noise bandwidth, being maximal around 50 Hz. Gap thresholds were generally higher for the hearing-impaired than for the normally hearing subjects. For the latter, gap thresholds were similar for the three conditions. For the hearing-impaired subjects, gap thresholds were similar for the unaided condition and the condition using slow compression. However, fast compression led to smaller gap thresholds, especially for noise bandwidths up to 50 Hz. The results show that fast compression can improve the ability of hearing-impaired subjects to detect gaps in sounds with slowly fluctuating envelopes.

Frequency selectivity as a function of level and frequency measured with uniformly exciting notched noise.

Thresholds for detecting sinusoidal signals were measured as a function of the spectral width of a notch in a noise masker. The notch was positioned both symmetrically and asymmetrically around the signal frequency. The noise was designed to create equal excitation per ERB within its passbands (uniformly exciting noise), after allowing for the transfer function of the headphone and the middle ear. For a signal frequency of 250 Hz, the level per ERB ranged from 35 to 80 dB in 15-dB steps. For signal frequencies of 500, 1,000, 2,000, and 4,000 Hz, the level per ERB ranged from 40 to 70 dB per ERB in 15-dB steps. Auditory filter shapes were derived from the data by modeling the auditory filter as the sum of a sharply tuned tip filter and a broader tail filter. The gain of the tip filter was assumed to be a function of level. The shape of the tip filter and the gain and shape of the tail filter were assumed to be level independent. The data for all levels were fitted simultaneously. The data were fitted best when the gain of the tip filter was assumed to be a function of the signal level (as opposed to the masker level per ERB). The filter shapes showed a level dependence that qualitatively resembled the level dependence of filtering on the basilar membrane. The maximum gain of the tip filter tended to increase with increasing center frequency up to 1 kHz, but to remain roughly constant for higher frequencies.

A test for the diagnosis of dead regions in the cochlea.

Hearing impairment may sometimes be associated with complete loss of inner hair cells (IHCs) over a certain region of the basilar membrane. We call this a 'dead region'. Amplification (using a hearing aid) over a frequency range corresponding to a dead region may not be beneficial and may even impair speech intelligibility. However, diagnosis of dead regions is not easily done from the audiogram. This paper reports the design and evaluation of a method for detecting and delimiting dead regions. A noise, called 'threshold equalizing noise' (TEN), was spectrally shaped so that, for normally hearing subjects, it would give equal masked thresholds for pure tone signals at all frequencies within the range 250-10,000 Hz. Its level is specified as the level in a one-ERB (132 Hz) wide band centred at 1000 Hz. Measurements obtained from 22 normal-hearing subjects and TEN levels of 30, 50 and 70 dB/ERB confirmed that the signal level at masked threshold was approximately equal to the noise level/ERB and was almost independent of signal frequency. Masked thresholds were measured for 20 ears of 14 subjects with sensorineural hearing loss, using TEN levels of 30, 50 and 70 dB/ERB. Psychophysical tuning curves (PTCs) were measured for the same subjects. When there are surviving IHCs corresponding to a frequency region with elevated absolute thresholds, a signal in that frequency region is detected via IHCs with characteristic frequencies (CFs) close to that region. In such a case, threshold in the TEN is close to that for normal-hearing listeners, provided that the noise intensity is sufficient to produce significant masking. Also, the tip of the PTC lies close to the signal frequency. When a dead region is present, the signal is detected via IHCs with CFs different from that of the signal frequency. In such a case, threshold in the TEN is markedly higher than normal, and the tip of the PTC is shifted away from the signal frequency. Generally, there was a very good correspondence between the results obtained using the TEN and the PTCs. We conclude that the measurement of masked thresholds in TEN provides a quick and simple method for the diagnosis of dead regions.

Comparison of the NAL(R) and Cambridge formulae for the fitting of linear hearing aids.

This paper describes a laboratory-based comparison of the effectiveness of two formulae for fitting linear hearing aids, the NAL(R) formula and the Cambridge formula. The formulae prescribe the desired insertion gain as a function of frequency, based on the audiometric threshold. The two formulae have a similar rationale; both are based on the goal that, for speech with a moderate level, all frequency bands should be equally loud (equal loudness per critical band) over the frequency range important for speech (400-5000 Hz), and the overall loudness should be comfortable. However, the formulae differ; generally the Cambridge formula leads to slightly more high-frequency gain (above 2 kHz) and slightly less mid-frequency gain (between 500 Hz and 2000 Hz) than the NAL(R) formula. The two formulae were implemented using an experimental digital hearing aid whose frequency-gain characteristic could be controlled very precisely. A loudness model (Moore and Glasberg, 1997) was used to adjust the overall gains for each subject and each formula so that a speech-shaped noise with an overall level of 65 dB SPL would give the same loudness as for a normally hearing person (according to the model). The adjustments were, on average, smaller for the Cambridge than for the NAL(R) formula. A condition was also used with all insertion gains set to zero, simulating unaided listening. Evaluation was based on: (1) subjective ratings of the loudness, intelligibility and quality of continuous discourse presented in quiet at levels of 45, 55, 65 and 75 dB SPL and in babble at an 0-dB speech-to-babble ratio, using speech levels of 55, 65 and 75 dB SPL; (2) measures of the speech reception threshold (SRT) in background noise for two noise levels (65 and 75 dB SPL) and four types of background noise. Neither the subjective ratings nor the measures of the SRTs revealed any consistent difference between the results obtained using the two formulae, although both formulae led to lower (better) SRTs than for simulated unaided listening. It is concluded that the differences between the NAL(R) formula and the Cambridge formula are too small to have measurable effects, at least in a laboratory setting.

Detection and intensity discrimination of Gaussian-shaped tone pulses as a function of duration.

Van Schijndel et al. [J. Acoust. Soc. Am. 105, 3425-3435 (1999)] proposed that the auditory system partitions the spectro-temporal domain into frequency-time (f-t) windows and that the characteristics of these windows could be explored by measuring intensity discrimination for Gaussian-shaped tone pulses presented just above their detection threshold in noise. They reasoned that for a long-duration tone pulse, the auditory representation would be maximally compact in the frequency domain, but would spread across several f-t windows in the time domain. For a very-short-duration tone pulse, the auditory representation would be maximally compact in the time domain, but would spread across several f-t windows in the frequency domain. There should be some intermediate duration at which the auditory representation is compact in both the time and frequency domains and for which intensity-discrimination performance should worsen, due to the limited opportunity for multiple looks. Their data for signal frequencies of 1 and 4 kHz were consistent with this expectation; intensity discrimination was poorest at a duration of about 3-5 ms at 1 kHz and 1 ms at 4 kHz (durations are specified between 6.8-dB-down points on the envelope). This experiment attempted to replicate those results and to extend them to a wider range of frequencies and levels. Intensity discrimination of Gaussian-shaped tone pulses was measured at three levels: 10 dB above absolute threshold or above masked threshold in a pink noise with a spectrum level of either 15 or 40 dB at 1 kHz. The signal frequency was 0.25 kHz (durations from 2 to 320 ms), 1 kHz (durations from 0.5 to 80 ms), or 4 kHz (durations from 0.1 to 20 ms). Three normally hearing subjects were tested. At 1 and 4 kHz, performance was poorest overall for the 15-dB pink noise level, and thresholds showed a peak at intermediate durations (about 3-5 ms at 1 kHz and 1 ms at 4 kHz). Such peaks were still apparent, but smaller in the no-noise condition and were almost absent at the higher noise level. For the 0.25-kHz signal frequency, peaks were not observed consistently at any level, although two subjects showed small peaks for durations around 10 ms. An explanation is offered for the results in terms of the level and frequency dependence of basilar-membrane input-output functions.

Use of a loudness model for hearing aid fitting: III. A general method for deriving initial fittings for hearing aids with multi-channel compression.

A model for predicting loudness for people with cochlear hearing loss is applied to the problem of the initial fitting of multi-channel fast-acting compression hearing aids. The fitting is based entirely on the pure tone audiogram, and does not require measures of loudness growth. One constraint is always applied: the specific loudness pattern evoked by speech of a moderate level (65 dB SPL) should be reasonably flat (equal loudness per critical band), and the overall loudness should be similar to that evoked in a normal listener by 65-dB speech. This is achieved using the 'Cambridge' formula. For hearing aids where the compression threshold in each channel can be set to a very low value, an additional constraint is used: speech with an overall level of 45 dB SPL should be audible over its entire dynamic range in all frequency channels from 500 Hz up to about 4 kHz. For hearing aids where the compression thresholds cannot be set to very low values, a different additional constraint is used: the specific loudness pattern evoked by speech of a high level (85 dB SPL, and with the spectral characteristics of shouted speech) should be reasonably flat, and the overall loudness should be similar to that evoked in a normal listener by 85-dB speech. For both cases, compression ratios are limited to values below 3. For each of these two cases, we show how to derive compression ratios and gains, and for the first case, compression thresholds, for each channel. The derivations apply to systems with any number of channels. A computer program implementing the derivations is described. The program also calculates target insertion gains at the centre frequency of each channel for input levels of 50, 65 and 80 dB SPL, and target gains at the eardrum measured relative to the level at the reference microphone of a probe microphone system.

Further evaluation of a model of loudness perception applied to cochlear hearing loss.

This paper describes further tests of a model for loudness perception in people with cochlear hearing loss. It is assumed that the hearing loss (the elevation in absolute threshold) at each audiometric frequency can be partitioned into a loss due to damage to outer hair cells (OHCs) and a loss due to damage to inner hair cells (IHCs) and/or neurons. The former affects primarily the active mechanism that amplifies the basilar membrane (BM) response to weak sounds. It is modeled by increasing the excitation level required for threshold, which results in a steeper growth of specific loudness with increasing excitation level. Loss of frequency selectivity, which results in broader excitation patterns, is also assumed to be directly related to the OHC loss. IHC damage is modeled by an attenuation of the calculated excitation level at each frequency. The model also allows for the possibility of complete loss of IHCs or functional neurons at certain places within the cochlea ("dead" regions). The parameters of the model (OHC loss at each audiometric frequency, plus frequency limits of the dead regions) were determined for three subjects with unilateral cochlear hearing loss, using data on loudness matches between sinusoids presented alternately to their two ears. Further experiments used bands of noise that were either 1-equivalent rectangular bandwidth (ERB) wide or 6-ERBs wide, centered at 1 kHz. Subjects made loudness matches for these bands of noise both within ears and across ears. The model was reasonably accurate in predicting the results of these matches without any further adjustment of the parameters.

Modulation masking produced by beating modulators.

This study examined whether "modulation masking" could be produced by temporal similarity of the probe and masker envelopes, even when the masker envelope did not contain a spectral component close to the probe frequency. Both masker and probe amplitude modulation were applied to a single 4-kHz sinusoidal or narrow-band noise carrier with a level of 70 dB SPL. The threshold for detecting 5-Hz probe modulation was affected by the presence of a pair of masker modulators beating at a 5-Hz rate (40 and 45 Hz, 50 and 55 Hz, or 60 and 65 Hz). The threshold was dependent on the phase of the probe modulation relative to the beat cycle of the masker modulators; the threshold elevation was greatest (12-15 dB for the sinusoidal carrier and 9-11 dB for the noise carrier, expressed as 20 log m) when the peak amplitude of the probe modulation coincided with a peak in the beat cycle. The maximum threshold elevation of the 5-Hz probe produced by the beating masker modulators was 7-12 dB greater than that produced by the individual components of the masker modulators. The threshold elevation produced by the beating masker modulators was 2-10 dB greater for 5-Hz probe modulation than for 3- or 7-Hz probe modulation. These results cannot be explained in terms of the spectra of the envelopes of the stimuli, as the beating masker modulators did not produce a 5-Hz component in the spectra of the envelopes. The threshold for detecting 5-Hz probe modulation in the presence of 5-Hz masker modulation varied with the relative phase of the probe and masker modulation. The pattern of results was similar to that found with the beating two-component modulators, except that thresholds were highest when the masker and probe were 180 degrees out of phase. The results are consistent with the idea that nonlinearities within the auditory system introduce distortion in the internal representation of the envelopes of the stimuli. In the case of two-component beating modulators, a weak component is introduced at the beat rate, and it has an amplitude minimum when the beat cycle is at its maximum. The results could be fitted well using two models, one based on the concept of a sliding temporal integrator and one based on the concept of a modulation filter bank.

Use of a loudness model for hearing aid fitting: II. Hearing aids with multi-channel compression.

A model for predicting loudness for people with cochlear hearing loss was applied to the problem of the initial fitting of a multi-channel compression hearing aid. The fitting was based on two constraints: (1) The specific loudness pattern evoked by speech of a moderate level (65 dB SPL) should be reasonably flat (equal loudness per critical band), and the overall loudness should be similar to that evoked in a normal listener by 65-dB speech (about 23 sones for binaural listening); (2) Speech with an overall level of 45 dB SPL should just be audible in all frequency bands from 500 Hz up to about 4 kHz, provided that this does not require compression ratios exceeding about 3. These two constraints were used to determine initial values for the gain, compression ratio and compression threshold in each channel of a multi-channel compression system. This initial fitting was based entirely on audiometric thresholds; it does not require suprathreshold loudness measures. The fitting method was evaluated using an experimental fast-acting four-channel compression system. The initial fitting was followed by an adaptive procedure to 'fine tune' the fitting, and the aids were then used in everyday life. Performance was evaluated by use of questionnaires and by measures of speech intelligibility. Although the fine tuning resulted in modest changes in the fitting parameters for some subjects, on average the frequency response shapes and compression ratios were similar before and after the fine tuning. The fittings led to satisfactory loudness impressions in everyday life and to high speech intelligibility over a wide range of levels. It was concluded that the initial fitting method gives reasonable starting values for the fine tuning.

Effects of frequency and duration on psychometric functions for detection of increments and decrements in sinusoids in noise.

Psychometric functions for detecting increments or decrements in level of sinusoidal pedestals were measured for increment and decrement durations of 5, 10, 20, 50, 100, and 200 ms and for frequencies of 250, 1000, and 4000 Hz. The sinusoids were presented in background noise intended to mask spectral splatter. A three-interval, three-alternative procedure was used. The results indicated that, for increments, the detectability index d' was approximately proportional to delta I/I. For decrements, d' was approximately proportional to delta L. The slopes of the psychometric functions increased (indicating better performance) with increasing frequency for both increments and decrements. For increments, the slopes increased with increasing increment duration up to 200 ms at 250 and 1000 Hz, but at 4000 Hz they increased only up to 50 ms. For decrements, the slopes increased for durations up to 50 ms, and then remained roughly constant, for all frequencies. For a center frequency of 250 Hz, the slopes of the psychometric functions for increment detection increased with duration more rapidly than predicted by a "multiple-looks" hypothesis, i.e., more rapidly than the square root of duration, for durations up to 50 ms. For center frequencies of 1000 and 4000 Hz, the slopes increased less rapidly than predicted by a multiple-looks hypothesis, for durations greater than about 20 ms. The slopes of the psychometric functions for decrement detection increased with decrement duration at a rate slightly greater than the square root of duration, for durations up to 50 ms, at all three frequencies. For greater durations, the increase in slope was less than proportional to the square root of duration. The results were analyzed using a model incorporating a simulated auditory filter, a compressive nonlinearity, a sliding temporal integrator, and a decision device based on a template mechanism. The model took into account the effects of both the external noise and an assumed internal noise. The model was able to account for the major features of the data for both increment and decrement detection.

Comparison of different forms of compression using wearable digital hearing aids.

Four different compression algorithms were implemented in wearable digital hearing aids: (1) The slow-acting dual-front-end automatic gain control (AGC) system [B. C. J. Moore, B. R. Glasberg, and M. A. Stone, Br. J. Audiol. 25, 171-182 (1991)], combined with appropriate frequency response equalization, with a compression threshold of 63 dB sound pressure level (SPL) and with a compression ratio of 30 (DUAL-HI); (2) The dual-front-end AGC system combined with appropriate frequency response equalization, with a compression threshold of 55 dB SPL and with a compression ratio of 3 (DUAL-LO). This was intended to give some impression of the levels of sounds in the environment; (3) Fast-acting full dynamic range compression in four channels (FULL-4). The compression was designed to minimize envelope distortion due to overshoots and undershoots; (4) A combination of (2) and (3) above, where each applied less compression than when used alone (DUAL-4). Initial fitting was partly based on the concept of giving a flat specific-loudness pattern for a 65-dB SPL speech-shaped noise input, and this was followed by fine tuning using an adaptive procedure with speech stimuli. Eight subjects with moderate to severe cochlear hearing loss were tested in a counter-balanced design. Subjects had at least 2 weeks experience with each system in everyday life before evaluation using the Abbreviated Profile of Hearing Aid Benefit (APHAB) test and measures of speech intelligibility in quiet (AB word lists at 50 and 80 dB SPL) and noise (adoptive sentence lists in speech-shaped noise, or that same noise amplitude modulated with the envelope of speech from a single talker). The APHAB scores did not indicate clear differences between the four systems. Scores for the AB words in quiet were high for all four systems at both 50 and 80 dB SPL. The speech-to-noise ratios required for 50% intelligibility were low (indicating good performance) and similar for all the systems, but there was a slight trend for better performance in modulated noise with the DUAL-4 system than with the other systems. A subsequent trial where three subjects directly compared each of the four systems in their everyday lives indicated a slight preference for the DUAL-LO system. Overall, the results suggest that it is not necessary to compress fast modulations of the input signal.

Use of a loudness model for hearing-aid fitting. I. Linear hearing aids.

A model for predicting loudness for people with cochlear hearing loss is applied to the problem of prescribing the frequency-gain characteristic of a linear hearing aid. It is argued that a reasonable goal is to make all frequency bands of speech equally loud while achieving a comfortable overall loudness; this can maximize the proportion of the speech spectrum that is above the absolute threshold for a given loudness. In terms of the model this means that the specific loudness pattern evoked by speech of a moderate level (65 dB SPL) should be reasonably flat (equal loudness per critical band), and the overall loudness should be similar to that evoked in a normal listener by 65 dB speech (about 23 sones). The model is used to develop a new formula - the 'Cambridge formula' - for prescribing insertion gain from audiometric thresholds. It is shown that, for a fixed overall loudness of 23 sones, the Cambridge formula leads to a higher calculated articulation index than three other commonly used prescriptive methods: NAL(R), FIG6 and DSL.

Development and evaluation of a procedure for fitting multi-channel compression hearing aids.

Hearing aids with multi-channel compression are often fitted on the basis of loudness scaling data obtained using narrow bands of noise or tones. Here, we report the development and evaluation of an alternative fitting procedure based on the use of speech signals. The parameters of the hearing aid (the gains in each channel for high and low input levels) are adjusted adaptively under computer control on the basis of the listener's responses. The goal is that speech at 85 dB SPL should be judged as 'loud', speech at 60 dB SPL should be judged as 'quiet', and speech at both levels should be judged as 'neither tinny nor boomy'. The procedure was evaluated using a two-channel compression hearing aid, the remote control of which allowed two programs to be stored. One program was based on our fitting procedure. The other was either based on the manufacturer's recommended full fitting procedure (which included loudness scaling with bands of noise), or was based on the audiogram alone, using the manufacturer's algorithm. After an acclimatization period of at least two weeks, subjects were then asked to fill in a questionnaire about their experiences with the two programs in different listening situations. The results generally indicated a preference for the program based on our adaptive fitting procedure. We also conducted laboratory measurements of speech intelligibility, in quiet and in a background of a single competing talker. These showed no clear difference between programs, although scores overall were very high. We conclude that our adaptive procedure gives very satisfactory results in everyday life. Parameter values giving good comfort also give good intelligibility. The procedure typically takes between five and 10 minutes per ear, which is quicker than most loudness scaling procedures.

Comparison of real and simulated hearing impairment in subjects with unilateral and bilateral cochlear hearing loss.

Simulations of hearing impairment were presented to the normal ears of subjects with moderate to severe unilateral cochlear hearing loss. The intelligibility of speech in quiet and in background sounds was compared with that obtained for the impaired ears using unprocessed stimuli. The results of loudness matches between the two ears were used to tailor a simulation of threshold elevation combined with loudness recruitment individually for each subject. This was assessed either alone, or in combination with a simulation of reduced frequency selectivity, performed by spectral smearing. Finally, we included a simulation of 'dead' regions in the cochlea, where there are assumed to be no functioning inner hair cells and/or neurones, by band-stop filtering over the frequency range corresponding to the dead region. Performance for the impaired ears was markedly worse than for the normal ears using the simulation of threshold elevation and loudness recruitment. The addition of the simulation of reduced frequency selectivity caused performance to worsen, but it remained above that for the impaired ears. The additional simulation of a dead region had little effect, except for one subject, for whom it produced performance comparable to that for the impaired ear in quiet but not when background sounds were present. It is suggested that the relatively poor results for the impaired ears may be caused partly by a form of 'neglect' which is specific to subjects with unilateral or asymmetric loss. This idea was supported by results obtained using bilaterally hearing-impaired subjects, which were markedly better than for the impaired ears of the unilaterally hearing-impaired subjects, and comparable to those for the normal ears listening to the combined simulation of threshold elevation, loudness recruitment and reduced frequency selectivity.

Detection of decrements and increments in sinusoids at high overall levels.

Thresholds for the detection of decrements in level of sinusoidal signals were measured as a function of duration (2, 4, 6, 10, and 14 ms), level (70, 80, and 90 dB SPL) and frequency (250, 500, 1000, 2000, and 4000 Hz). Seven normally hearing listeners were tested at each frequency (with different subjects for each frequency). Thresholds for detecting a 10-ms increment in level were also measured. The sinusoids were presented in a background noise low-pass filtered at 5 kHz, which was intended to mask spectral splatter associated with the decrement or increment. Performance improved with increasing frequency for all decrement and increment durations. Performance also tended to improve with increasing level at 2000 and 4000 Hz. The results were analyzed using a four-stage model consisting of an auditory filter centered on the signal frequency, a compressive nonlinearity, a sliding temporal integrator and a decision mechanism. The analysis indicated that the improved performance with increasing frequency and increasing level could be attributed partly to off-frequency listening; for the two highest center frequencies, subjects probably made use of the output of an auditory filter centered above the signal frequency, where changes in excitation level associated with an increment or decrement were magnified. The measurements at 4000 Hz were repeated using a broadband background noise (15-kHz bandwidth), which would prevent the use of information from auditory filters centered far above the signal frequency. Performance was poorer than when low-pass noise was used, but still improved somewhat with increasing level. The slight improvement in performance with increasing level can be accounted for by a reduced compressive linearity at high levels. A good fit to the data could be obtained by assuming that the equivalent rectangular duration (ERD) of the temporal integrator was invariant with level, but that the compressive nonlinearity varied with level in a similar way to basilar-membrane input-output functions. The nonlinearity appears to be somewhat less compressive at 250 Hz than at higher center frequencies. The ERD is about 7 ms regardless of center frequency.

The probe-signal method and auditory-filter shape: results from normal- and hearing-impaired subjects.

In the probe-signal method, subjects are required to detect a signal in noise that is presented on the majority of trials at an "expected" (target) frequency but on a minority of trials at an "unexpected" probe frequency. Detection of the probes worsens with increasing separation between the target and probe frequencies. This result has often been interpreted as indicating that subjects monitor the output of a single auditory filter centered at the target frequency. To test this idea, a two-stage experiment was conducted. In the first stage, auditory-filter shapes were estimated using the notched-noise method at center frequencies of 1000, 1259, 1585, and 2000 Hz. These were the frequencies that were used for the targets in the second stage of the experiment. In the second stage, low-pass filtered white noise was presented continuously. On each trial, a cue tone was presented at one of the four possible target frequencies. The specific frequency was selected randomly on each trial. This was followed by two observation intervals during one of which a further sinusoidal tone was presented. This tone was either a target (the same as the cue frequency) (on 60% of trials), or had one of four possible probe frequencies corresponding to that target. The four probe frequencies were chosen to correspond to specific points on the estimated response curve of the auditory filter centered at the target frequency. The percentage of correct detections of a given probe was compared with that obtained in a separate condition where the frequency of the tone was fixed throughout, the cue frequency always equaled the target frequency, and the target was attenuated by an amount corresponding to the attenuation of the auditory filter at the probe frequency. Two subjects with normal hearing and two subjects with unilateral cochlear hearing loss were used. Comparison of the results for the normal and impaired ears suggests that the detectability of the probes is governed more by the selectivity of the auditory filters than by the ratios of the expected and probe frequencies. However, detection of the probes was generally better than would occur if subjects monitored the output of a single auditory filter centered at the target frequency.

Simulation of the effects of loudness recruitment on the intelligibility of speech in noise.

This experiment simulated the threshold elevation and loudness recruitment associated with three different types of hearing loss: moderate flat (condition R2), severe flat (condition R3), and moderate-to-severe sloping (condition RX). This was done to allow an examination of the effects of these factors on the intelligibility of speech, in isolation from other factors that are normally associated with cochlear hearing loss, such as reduced frequency selectivity. The speech was presented at a fixed input level of 65 dB SPL, against a background of a noise whose spectrum was shaped to match the long-term average spectrum of the speech. The level of the background noise varied from 65 to 74 dB SPL. The simulation was performed by splitting the input signal into 13 frequency bands, and processing the envelope in each band so as to create loudness sensations in a normal ear that would resemble those produced in an impaired ear with recruitment. The bands were then recombined. All tests were performed using subjects with normal hearing. The simulation of hearing loss produced decrements in performance. The speech in condition R3 was inaudible. For conditions R2 and RX, the speech-to-noise ratios had to be up to 6 dB higher than in the control condition (R1, unprocessed stimuli) to achieve similar levels of performance. When linear amplification according to the NAL prescription was applied before the simulation, performance improved markedly for conditions R2 and RX, and did not differ significantly from that for R1. For condition R3, performance with simulated NAL amplification remained below that for condition R1; the decrement in performance was equivalent to about a 1 dB change in speech-to-noise ratio. The results of the present experiment show much smaller decrements in performance than those of an earlier experiment using a single talker as the interfering sound (Moore and Glasberg, 1993). It appears that loudness recruitment and threshold elevation have larger effects for a fluctuating background sound than for a steady background sound, and linear amplification is more effective in the latter case.

Effects of level and frequency on the detection of decrements and increments in sinusoids.

Thresholds for the detection of decrements in level of sinusoidal signals were measured as a function of decrement duration, level (25, 40, 55, and 70 dB SPL) and frequency (250, 1000, and 4000 Hz) in eleven normally hearing subjects. Thresholds for detecting a brief increment in level were also measured. The sinusoids were presented in a background noise intended to mask spectral splatter associated with the decrement or increment. Performance tended to worsen with decreasing frequency, for all decrement durations and for increment detection. Performance also worsened with decreasing level. The results were analyzed using a model consisting of a compressive nonlinearity, a sliding temporal integrator, and a decision device. The analysis indicated that the worsening in performance with decreasing frequency and decreasing level can be attributed partly to increases in the equivalent rectangular duration (ERD) of the temporal integrator, but mainly to changes in the efficiency of the detection process following the temporal integrator; at lower frequencies and levels a larger change is required at the output of the integrator for threshold to be reached. At each frequency, the ERD was relatively invariant with level for levels more than about 20 dB above the absolute threshold.