The intensity-difference limen for Gaussian-enveloped stimuli as a function of level: tones and broadband noise.

Van Schijndel et al. [J. Acoust. Soc. Am. 105, 3425-3435 (1999)] have proposed that the internal excitation evoked by an auditory stimulus is segmented into "windows" according to the stimulus spectrum and stimulus length. This "multiple looks" model accounts for the mid-duration hump they observed in plots of intensity-difference limens (DLs) versus pip duration for Gaussian-shaped 1- and 4-kHz tones, an effect replicated by Baer et al. [J. Acoust. Soc. Am. 106, 1907-1916 (1999)]. However, van Schijndel et al. and Baer et al. used few levels. A greater number of levels were used by Nizami (1999) for Gaussian-shaped 2-kHz tone-pips whose equivalent rectangular duration (D) was 1.25 ms. The DLs show the mid-level hump known for clicks [Raab and Taub, J. Acoust. Soc. Am. 46, 965-968 (1969)]. At some duration this pattern must become the "near-miss to Weber's law." To determine this duration, as well as the level-dependence of the mid-duration hump, DLs were established for Gaussian-shaped 2-kHz tone-pips of D = 1.25, 2.51, and 10.03 ms at levels of 30-90 dB SPL. The across-subject average DLs for the tone-pips rise up at mid-levels for D= 1.25 and D = 2.51 ms. The DLs for D=2.51 ms are larger, creating the mid-duration hump. At all durations, the new DLs are smaller at high levels than at low levels, consistent with the near-miss to Weber's law. DLs were also obtained here for Gaussian-shaped broadband-noise pips of D=0.63, 1.25, 2.51, 5.02, and 10.03 ms. The DLs for the noise-pip show a mid-level hump for all pip durations. The noise-pip DLs decrease as the pip lengthens, such that the plot of DL versus log duration shows a linear decline, with no mid-duration hump. Analysis of variance reveals that the mid-level hump coexists with the classical patterns of level-dependence, perhaps reflecting the existence of two level-encoding mechanisms, one that depends on firing-rates counted over single neurons and which is responsible for the classical patterns, and one that depends on the initial coordinated burst of neuronal spikes caused by rapid ramping, and which presumably causes the mid-level hump.

Effect of relative amplitude, presentation level, and vowel duration on perception of voiceless stop consonants by normal and hearing-impaired listeners.

Previous work on the influence of relative amplitude and presentation level in listeners with normal hearing and those with sensorineural hearing loss was extended to include (1) a comparison of relative amplitude manipulation in the presence of neutral formant transitions versus manipulation of both formant transition and relative amplitude and (2) the additional variable of vowel duration. Synthetic consonant-vowel (CV) stimuli were used, and the amplitude of the burst relative to the vowel in the F4-F5 frequency range was varied across a 20-dB range using a /p-t/ contrast. In experiment I, some stimuli had neutral formant transition values and relative amplitude manipulations; other stimuli had both formant transition and relative amplitude manipulations. For stimuli in experiment II, neutral formant transition values were used, relative amplitude was manipulated, and vowel duration ranged from 14 to 200 ms. Results from experiment I showed no significant difference between listener groups when only relative amplitude information was manipulated, but significant differences when both relative amplitude and formant transition information was present. These results suggest that the listeners with normal hearing weighted the two acoustic cues differently from listeners with sensorineural hearing loss. Results from experiment II indicated that increasing vowel duration generally increased the number of labial responses from listeners with normal hearing, but did not always increase the number of labial responses from listeners with sensorineural hearing loss.

Intensity discrimination in the presence of random-frequency, multicomponent maskers and broadband noise.

This study examined the effects of multicomponent, random-frequency maskers and broadband-noise maskers on intensity discrimination at 1000 Hz. Maskers and signals were 200 ms, presented simultaneously. In the first set of conditions, thresholds were measured for the detection of a 1000-Hz tone in the presence of 40 or 60 dB SPL random-frequency or noise maskers, with extensive training of listeners with the random-frequency masker to assure stable effects of masker-frequency uncertainty. The random-frequency maskers had two, six, or ten components chosen at random from a large frequency range (300-3000 Hz, excluding a 160-Hz band around 1000 Hz). For these maskers, performance across the four listeners was very similar, showing large effects of masker-frequency uncertainty. For noise maskers, performance matched predictions for energy-based masking. In the second and third sets of conditions, intensity discrimination was measured at 1000 Hz for pedestals ranging from 40 to 80 dB SPL, first in isolation and then in the presence of the maskers. The pattern of results for intensity discrimination in quiet showed the expected near miss to Weber's Law, but poorer performance than typically observed. The addition of broadband-noise maskers had little effect on performance. However, random-frequency maskers degraded performance in nearly all conditions, with the size of the effect dependent on the level of the pedestal relative to the masker. Considering the pedestal as a tonal masker, the data were fitted with various models of combined masking. A simple power-law model provided excellent fits, with exponents ranging from 0.24 to 0.35 for the multicomponent maskers, but 1.0 (linear) for the noise. The results support models positing that the effects of individual maskers undergo nonlinear transformation before they are added, independent of the mechanisms which produce these effects. Because random-frequency maskers presumably produce informational (uncertainty-based) masking, the nonlinearity in this case appears central rather than peripheral.

Effect of relative and overall amplitude on perception of voiceless stop consonants by listeners with normal and impaired hearing.

Previous studies of the /p/-/t/ contrast for normal-hearing listeners have shown that both manipulation of the amplitude of the burst relative to the vowel in the F4-F5 frequency region and overall presentation level can influence the perception of place of articulation [R. N. Ohde and K. N. Stevens, J. Acoust. Soc. Am. 74, 706-714 (1983); Gravel and Ohde, Asha 25, 101 (1983)], such that greater burst amplitude in the high frequencies and higher presentation levels result in more alveolar responses. The influence of relative amplitude and presentation level was tested for both normal-hearing (NH) and hearing-impaired (HI) listeners in the present study. Synthetic CV stimuli were used, and the amplitude of the burst relative to vowel-onset amplitude in the F4-F5 frequency region was manipulated across a 20-dB range. In addition, overall presentation level was varied across a 45-dB range. The findings revealed that the hearing-impaired listeners selected more alveolar responses than listeners with normal hearing when tested at equivalent SPLs. A group of five normal-hearing listeners were then presented the synthetic stimuli in a background of broadband noise at a level that produced thresholds at 4 kHz equivalent to the thresholds of five hearing-impaired listeners. Results from the noise-masked normal-hearing listeners did not consistently show more alveolar responses as presentation level of the stimuli was increased, thus failing to mimic the responses from the hearing-impaired listeners in quiet.

Informational masking for multicomponent maskers with spectral gaps.

Simultaneous maskers comprised of a few random-frequency sinusoids can produce considerable informational (uncertainty-based) masking if the component frequencies are drawn from a wide range and changed with each stimulus presentation. The present experiments examined the effect on informational masking of removing masker energy from large frequency regions around the signal. Threshold for a 1000-Hz signal was measured in the presence of maskers comprised of 2, 4, 6, 10, 50, or 100 random-frequency sinusoids, notched-noise, or two fixed-frequency sinusoids. The multicomponent maskers had a maximum frequency range of 300-3000 Hz, typically excluding a 160-Hz band around the signal. In comparison conditions, masker frequencies were limited to the high or low side of the signal, or the gap around the signal was progressively widened. Four listeners showed substantial informational masking which was not eliminated even by extreme spectral gaps in the maskers. Four other listeners showed much smaller effects of masker uncertainty across all conditions. Notched-noise measures of auditory-filter width did not distinguish the two subject groups, but indices of processing efficiency were typically poorer for the high-threshold listeners, as were measures of both the width and processing efficiency of presumed "attentional filters" under conditions of masker-frequency uncertainty.

A comparison of transient-evoked and distortion product otoacoustic emissions in normal-hearing and hearing-impaired subjects.

The ability of transient-evoked otoacoustic emissions (TEOAEs) and distortion product otoacoustic emissions (DPOAEs) to distinguish normal hearing from hearing impairment was evaluated in 180 subjects. TEOAEs were analyzed into octave or one-third octave bands for frequencies ranging from 500 to 4000 Hz. Decision theory was used to generate receiver operating characteristic (ROC) curves for each of three measurements (OAE amplitude, OAE/noise, reproducibility) for each OAE measure (octave TEOAEs, 1/3 octave TEOAEs, DPOAEs), for octave frequencies from 500 to 4000 Hz, and for seven audiometric criteria ranging from 10 to 40 dB HL. At 500 Hz, TEOAEs and DPOAEs were unable to separate normal from impaired ears. At 1000 Hz, both TEOAE measures were more accurate in identifying hearing status than DPOAEs. At 2000 Hz, all OAE measures performed equally well. At 4000 Hz, DPOAEs were better able to distinguish normal from impaired ears. Almost without exception, measurements of OAE/noise and reproducibility performed comparably and were superior to measurements of OAE amplitude, although the differences were small. TEOAEs analyzed into octave bands showed better performance than TEOAEs analyzed into 1/3 octaves. Under standard test conditions, OAE test performance appears to be limited by background noise, especially for the low frequencies.

Analysis of transient-evoked otoacoustic emissions in normal-hearing and hearing-impaired ears.

Transient-evoked otoacoustic emissions (TEOAEs) were measured in 113 normal-hearing and hearing-impaired ears to examine repeatability within a test session, which TEOAE parameter (level, TEOAE level-to-noise or reproducibility) best identified hearing loss and if the TEOAE separated into frequency-specific bands identified hearing loss in a corresponding frequency region. TEOAEs and stimulus levels were found to be very repeatable. For broadband TEOAEs, TEOAE level, TEOAE-to-noise, and % reproducibility were found to identify hearing loss equally well, based on measurement of the area underlying relative operator characteristic curves. Analysis for frequency-specific bands showed that separation of normal-hearing and hearing-impaired ears depended on frequency, with best identification at 2000 and 4000 Hz, identification at 1000 Hz slightly worse, and virtually no separation between normal-hearing and hearing-impaired ears at 500 Hz. Again, all three parameters were essentially equal in identifying hearing loss. Subjective evaluations of presence or absence of TEOAEs was highly correlated between two judges, with good agreement for TEOAEs at 1000, 2000, and 4000 Hz. The findings from this study suggest that TEOAEs will be valuable for clinical use because of their repeatability and identification of hearing-impaired ears.

Otoacoustic emissions from normal-hearing and hearing-impaired subjects: distortion product responses.

Distortion product otoacoustic emissions (DPOAE) were measured in normal-hearing and hearing-impaired human subjects. Analyses based on decision theory were used to evaluate DPOAE test performance. Specifically, relative operating characteristic (ROC) curves were constructed and the areas under these curves were used to estimate the extent to which normal and impaired ears could be correctly identified by these measures. DPOAE amplitude and DPOAE/noise measurements were able to distinguish between normal and impaired subjects at 4000, 8000, and, to a lesser extent, at 2000 Hz. The ability of these measures to distinguish between groups decreased, however, as frequency and audiometric criterion used to separate normal and hearing-impaired ears decreased. At 500 Hz, performance was no better than chance, regardless of the audiometric criterion for normal hearing. Cumulative distributions of misses (hearing-impaired ears incorrectly identified as normal hearing) and false alarms (normal-hearing ears identified as hearing impaired) were constructed and used to evaluate test performance for a range of hit rates (i.e., the percentage of correctly identified hearing-impaired ears). Depending on the desired hit rate, criterion values of -5 to -12 dB SPL for DPOAE amplitudes and 8 to 15 dB for DPOAE/noise accurately distinguished normal-hearing ears from those with thresholds greater than 20 dB HL for the two frequencies at which performance was best (4000 and 8000 Hz). It would appear that DPOAE measurements can be used to accurately identify the presence of high-frequency hearing loss, but are not accurate predictors of hearing status at lower frequencies, at least for the conditions of the present measurements.

Two experiments on the spectral boundary conditions for nonlinear additivity of simultaneous masking.

The present paper describes the results from two experiments which explored the spectral boundaries for the nonlinear additivity of simultaneous masking. The first experiment measured the threshold for detection of a 2-kHz tone in the presence of two 800-Hz-wide bands of noise that had varying degrees of spectral overlap with each other and the 2-kHz signal. Results revealed an abrupt transition from linear to nonlinear additivity of masking as the spectral separation between the two maskers varied from some overlap to none. The second experiment examined alternative explanations for the data. Explanations based on restricted-listening or distortion-product-detection hypotheses were not supported by the results of this experiment. These data indicate that nonlinear additivity of simultaneous masking holds for maskers that do not overlap within the critical band centered on the signal frequency. This interpretation is also consistent with a large body of data on the monaural and binaural summation (additivity) of loudness.

Models of the effects of threshold on loudness growth and summation.

The present paper extends previous efforts to compare models of additivity of masking in normal and impaired listeners to measures of loudness. A model that provides a good account of additivity of masking for normal and impaired listeners, the modified power-law model with compressed internal noise, also provides a good description of basic loudness data from normal and impaired listeners. This model has been discussed frequently in the literature on loudness as one means of correcting loudness functions near threshold, but has not been widely accepted. The applicability of this model to loudness growth and the summation of loudness, both monaurally and binaurally, is evaluated and compared to the more generally accepted alternative. The ability of the model reported here to account for growth of loudness near threshold, as well as the additivity of masking, suggests a need to reconsider its application to loudness data, as well as a need to reconsider the relation between loudness and masking.

Modeling the interactions between noise exposure and other variables.

The interaction of noise exposure with other variables is reviewed. For the case of the interaction of noise with other variables that produce behavioral threshold shifts, the application of a newly developed model is described and demonstrated. This model, referred to as the modified power-law model, provides an accurate prediction of the combined effects of two threshold-elevating factors. The model accounts for the interaction of post-exposure a pre-existing pre-existing permanent loss or a pre-existing temporary loss. The model's application is demonstrated for multiple exposures to steady-state noise in which each exposure lasts as short as 12 min or as long as 6 h. Finally, implications of the model's application to the interaction long as 6 h. Finally, implications of the model's application to the interaction of noise with other ototraumatic agents are reviewed.

Speech perception in low-pass filtered noise for normal and hearing-impaired listeners.

Two experiments were conducted concerning speech perception in noise. In Experiment 1, a comparison was made between adaptive and fixed-level procedures to estimate the S/N ratio at which 50% correct performance occurred for nonsense syllables for normal-hearing listeners. The two methods yield similar S/N ratio estimates, but the consonant confusions found with the fixed-level method could not be predicted accurately from the adaptive procedure. In Experiment 2, the adaptive procedure was used to estimate the S/N ratio for a 50% performance level in low-pass filtered noise with a range of cutoff frequencies. Data were obtained from 5 normal-hearing listeners at two speech levels (50 and 75 dB SPL) and 4 hearing-impaired listeners at one speech level (75 dB SPL). The hearing-impaired listeners required a better S/N ratio than the normal listeners at either presentation level for all except the widest bandwidth, where their S/N ratios began to converge with the normal values. In addition, the S/N ratios for the hearing-impaired listeners plateaued at relatively narrow bandwidths (0.75 to 2.5 kHz) compared to the normal-hearing group (3.0 to 5.0 kHz). That is, the addition of high-frequency components to the noise did not alter performance. These findings suggest that the hearing-impaired listeners may have relied upon either low-frequency cues or prosodic cues in the perception of these test items.

Normative thresholds in the 8- to 20-kHz range as a function of age.

Using a prototype high-frequency audiometer, auditory thresholds in the 8- to 20-kHz range were obtained from 240 subjects ranging in age from 10-60 years. These measurements were obtained in interest of developing a normative database for frequencies above 8 kHz, and to evaluate intersubject variability as a function of age. An analysis of variance (ANOVA) revealed significant effects of frequency, age, and sex, and a significant frequency-by-age interaction. The largest changes in sensitivity with age occurred between 40 to 59 years. Below approximately 15 kHz, the intersubject variability of threshold estimates increased as a function of both age and frequency. Further analysis revealed that the age-related changes in variability were related to absolute thresholds rather than to age per se. When data are converted to dB HL (relative to the youngest group tested), the region of maximum hearing loss shifts to lower frequencies with increasing age, and threshold shifts with age are greatest in the 13- to 17-kHz range.

Auditory brainstem responses from children three months to three years of age: normal patterns of response. II.

Auditory brainstem responses (ABR) were measured in 535 children from 3 months to 3 years of age. The latencies reported in this paper should be unaffected by peripheral hearing loss because each child had bilateral wave V responses at 20 dB HLn. Wave V latencies decreased as age increased, at least to 18 months of age, while little or no change was noted in wave I latencies over the same age range. Thus, interpeak latency differences followed the same developmental time course as wave V. The shapes of wave V latency-level functions were comparable across age groups. These results suggest that changes in wave V latency with age are due to central (neural) factors and that age-appropriate norms should be used in evaluations of ABR latencies in children. Interaural differences in absolute wave V latencies and interpeak latency differences were similar to those observed in infants and adults, indicating that response symmetry is independent of age. Statistical analyses suggested that the distributions of absolute and relative latency measurements are normal, making it possible to describe norms in terms of means and standard deviations. A simple model is described that accounts accurately for changes in mean wave V latencies as function of age from preterm through the first three years of life.

Models of the additivity of masking.

Three models of masking additivity are reviewed, which are referred to as the high-compression model [M. J. Penner, J. Acoust. Soc. Am. 67, 608-616 (1980); M. J. Penner and R. M. Shiffrin, J. Acoust. Soc. Am. 67, 617-627 (1980)], the power-law model [R. A. Lutfi, J. Acoust. Soc. Am. 73, 262-267 (1983); 80, 422-428 (1986)], and the modified power-law model with compressed internal noise [Humes et al., J. Acoust. Soc. Am. 83, 188-202 (1988)]. While the high-compression model was derived from data for two or more nonsimultaneous maskers and the power-law model was derived from data for two or more simultaneous maskers, the modified power-law model can be applied to both cases. The modified power-law model assumes that the threshold in quiet is equivalent to a masked threshold resulting from an internal noise that is continually present. Additional assumptions concern the interaction of two maskers prior to the addition of the masking effects. Most of the data on the additivity of masking are well described by the modified power-law model, regardless of the nature of the maskers. Thus the model provides a good description of data for combined simultaneous maskers and combined nonsimultaneous maskers, a task heretofore requiring the use of at least two separate and independently developed models.

High-frequency audiometry: test reliability and procedural considerations.

This study compared the reliability of a recently developed high-frequency audiometer (HFA) [Stevens et al., J. Acoust. Soc. Am. 81, 470-484 (1987)] with a less complicated system that uses supraaural earphones (Koss system). The new approach permits calibration on an individual basis, making it possible to express thresholds at high frequencies in dB SPL. Data obtained from 50 normal-hearing subjects, ranging in age from 10-60 years, were used to evaluate the effects on reliability of threshold variance, earpiece/earphone fitting variance, and the variance associated with the HFA calibration process. Without earpiece/earphone replacement, the reliability of thresholds for the two systems is similar. With replacement, the HFA showed poorer reliability than the Koss system above 11 kHz, largely due to errors in estimating the calibration function. HFA reliability is greater for subjects with valid calibration functions over the entire frequency range. When average correction factors are applied to the Koss data in an effort to convert threshold estimates to dB SPL, individual transfer functions are not represented accurately. Thus the benefit of being able to express thresholds at high frequencies in dB SPL must be weighed against the additional source of variability introduced by the HFA calibration process.

The reliability of auditory thresholds in the 8- to 20-kHz range using a prototype audiometer.

This study was designed to evaluate both intra- and intertester reliability of auditory thresholds in the 8- to 20-kHz range using a recently developed high-frequency audiometer [Stevens et al., J. Acoust. Soc. Am. 81, 470-484 (1987)]. With this device, signals from a high-frequency transducer are introduced into the ear canal via a plastic tube. A calibration function is calculated for each ear and used to estimate the sound-pressure level (SPL) at the tympanic membrane. Twenty normal-hearing listeners were tested four times, twice by each of two examiners. In the higher frequencies, accurate calibration functions could not be obtained for many subjects; in these cases, values extrapolated from lower frequencies were used to estimate SPL. Findings reveal that the standard error of measurement for both intra- and intertester measures increases as a function of frequency. Intertester variability was only slightly higher than intratester variability. In most cases, variability of threshold estimates in dB SPL was higher than that observed for the uncorrected attenuator settings. Exclusion of extrapolated values improved reliability substantially.

Auditory brainstem responses to tone bursts in normally hearing subjects.

Auditory brainstem responses were recorded from 20 normally hearing subjects using tone-burst stimuli that were gated with cosine-squared functions. Clear responses were observed over a wide range of frequencies and levels. These responses were highly reproducible within individual subjects and were reliably measured by two independent examiners. ABR thresholds were higher than behavioral thresholds for all frequencies, especially for lower frequencies. Intersubject variability also was greater for lower frequencies. Wave-V latencies decreased with increases in both frequency and level for frequencies from 250 to 8000 Hz and for levels from 20 to 100 dB SPL. The standard deviations seldom exceeded 10% of the mean wave-V latency for any combination of level and frequency. These latencies can be viewed as the sum of both a peripheral and a central component. Assuming that the central component is relatively independent of both frequency and level, changes of wave V latency must be related to peripheral factors, such as travel time along the cochlear partition, and to stimulus characteristics, such as rise time.

Latency of auditory brain-stem responses and otoacoustic emissions using tone-burst stimuli.

A comparison of the latency of auditory brain-stem responses (ABR) and evoked otoacoustic emissions (EOAE) has led to an interpretation for the travel of transients in the peripheral auditory system that is consistent with both sets of data. The "cochlear echo" theory for the origin of the EOAE indicates that the latency of a particular frequency component back to the ear canal should be twice the forward latency of its characteristic place in the cochlea. The latency of wave V of the ABR to tone-burst stimuli can be described as the sum of two components: (1) a component that varies with intensity and frequency in an orderly and predictable manner and (2) a component that is independent of both intensity and frequency. Because the EOAE data can be predicted by taking twice the value of component (1) of the ABR latency, this component is interpreted to be due to mechanical travel through the cochlea. A consequence of this interpretation is that the remaining neural component of the ABR latency must be relatively independent of frequency and intensity.

Effects of pure-tone forward masker duration on psychophysical measures of frequency selectivity.

The effects of forward masker duration on psychophysical measures of frequency selectivity were investigated in two experiments. In both experiments, masker duration was 50 or 400 ms, signal duration was 20 ms, and there was no delay between masker offset and signal onset. In the first experiment, growth-of-masking functions were measured for a masker whose frequency was below, at, or above the 1000-Hz signal frequency. From those data, input filter patterns (IFPs) were plotted for masker levels from 40-90 dB SPL. In the second experiment, masking patterns (MPs) were measured for a 1000-Hz masker presented at 50, 70, and 90 dB SPL. Both measures of frequency selectivity (IFPs and MPs) indicate that frequency selectivity is greater for the 400-ms masker. These data suggest that there may be a sharpening of frequency selectivity with time at a stage prior to the adaptation observed in forward masking.