Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Feb 17.
Published in final edited form as: J Am Acad Audiol. 2022 Feb 17;32(9):596–605. doi: 10.1055/s-0041-1732333

Audibility and spectral-ripple discrimination thresholds as predictors of word recognition with nonlinear frequency compression

Marc A Brennan 1, Ryan W McCreery 2
PMCID: PMC9112840  NIHMSID: NIHMS1800146  PMID: 35176803

Abstract

Background:

Nonlinear frequency compression (NFC) lowers high-frequency sounds to a lower frequency and is used to improve high-frequency audibility. However, the efficacy of NFC varies widely—while some individuals benefit from NFC, many do not. Spectral resolution is one factor that might explain individual benefit from NFC. Because individuals with better spectral resolution understand more speech than those with poorer spectral resolution, it was hypothesized that individual benefit from NFC could be predicted from the change in spectral resolution measured with NFC relative to a condition without NFC.

Purpose:

To determine the impact of NFC on access to spectral information and whether these changes predict individual benefit from NFC for adults with sensorineural hearing loss (SNHL).

Research Design:

A quasi-experimental cohort study. Participants used a pair of hearing aids set to the Desired Sensation Level algorithm (DSL m[i/o]).

Study Sample:

Participants were 19 adults with SNHL, recruited from the Boys Town National Research Hospital Participant Registry.

Data Collection and Analysis:

Participants were seated in a sound-attenuating booth and then percent-correct recognition of words and spectral-ripple discrimination thresholds were measured for two different conditions, with and without NFC. Because audibility is known to influence spectral-ripple thresholds and benefit from NFC, audibility was quantified using the aided speech intelligibility index (SII). Linear mixed models were generated to predict word recognition using the aided SII and spectral-ripple discrimination thresholds.

Results:

While NFC did not influence percent-correct word recognition, participants with higher (better) aided SII and spectral-ripple discrimination thresholds understood more words than those with either a lower aided SII or spectral-ripple discrimination threshold. Benefit from NFC was not predictable from a participant’s aided SII or spectral-ripple discrimination threshold.

Conclusions:

We have extended previous work on the effect of audibility on benefit from NFC to include a measure of spectral resolution, the spectral-ripple discrimination threshold. Clinically, these results suggest that patients with better audibility and spectral resolution will understand speech better than those with poorer audibility or spectral resolution; however, these results are inconsistent with the notion that individual benefit from NFC is predictable from aided audibility or spectral resolution.

Keywords: Hearing aids and assistive devices, Speech perception, Hearing science, Frequency lowering, Spectral resolution

Introduction

Hearing loss afflicts 15% of middle-aged adults, rising to 60% for senior ages (Agrawal, Platz, and Niparko 2008). While the most common form of rehabilitation incorporates hearing-aid amplification, hearing aid use remains at 25% of those with a hearing loss (Kochkin 2005). Although hearing aids improve speech recognition by increasing audibility, word recognition for listeners with sensorineural hearing loss (SNHL) are still poorer than for listeners with normal hearing by 30% or more (Souza et al. 2007) in quiet and the presence of background noise poses an even greater problem on speech recognition, especially for listeners with SNHL (Gordon-Salant and Fitzgibbons 1993; Brennan et al. 2016). To understand speech, listeners need to decode level differences across speech (spectral information) that are contained in speech (Boothroyd et al. 1996; Swanepoel, Oosthuizen, and Hanekom 2012). Because individuals discriminate differences in the frequency content between noise and speech to parse speech from background noise, a poorer ability to encode spectral information underlie some of these deficits in speech recognition that are observed in listeners with SNHL (Zwicker and Schorn 1978; Bernstein et al. 2013; Davies-Venn, Nelson, and Souza 2015). While much work has been completed on spectral resolution in listeners with SNHL, little attention has been paid to the impact of hearing-aid signal processing on access to spectral cues—as such, the purpose of this study was to examine the influence of a type of hearing-aid signal processing, nonlinear frequency compression (NFC), on access to spectral cues and speech recognition.

The primary goal of amplification for listeners with sensorineural hearing loss (SNHL) is to restore audibility for speech sounds to optimize communication. The ability to amplify the speech spectrum can be limited at higher frequencies by hearing aid bandwidth and the listener’s degree of hearing loss (Kimlinger, McCreery, and Lewis 2015). Frequency lowering signal processing algorithms, such as nonlinear frequency compression (NFC), have been used to increase the audibility of high-frequency speech sounds for listeners with limited audible bandwidth by representing these sounds from the source band at a lower and narrower frequency region (i.e., destination band). Sounds below the start frequency are not lowered in frequency and sounds above the start frequency are compressed on a log scale by the compression ratio, with higher ratios resulting in a narrower destination region. Despite this aim of improving access to high-frequency sounds, the benefits of NFC for improving speech recognition have been mixed in previous studies with adults (Simpson, Hersbach, and McDermott 2005; Simpson, Hersbach, and McDermott 2006; Glista et al. 2009; Bohnert, Nyffeler, and Keilmann 2010; Alexander, Kopun, and Stelmachowicz 2014; McCreery et al. 2014; Picou, Marcrum, and Ricketts 2015). Numerous factors have been identified as contributing to individual differences in speech recognition with NFC including the listener’s age, degree of hearing loss, the method used to select the NFC parameters, and degree to which high-frequency audibility is increased with the processing.

The variability in speech recognition across participants fit with NFC is consistent regardless of the method used to set the NFC parameters to provide audibility, including live speech presentation of high-frequency phonemes, such as /s/ and /ʃ/ (Glista et al. 2009; Park et al. 2012; Picou, Marcrum, and Ricketts 2015), calibrated presentation of filtered speech stimuli with energy peaks at specified frequencies (Wolfe et al. 2010; Wolfe et al. 2011; Hillock-Dunn et al. 2014), or based on the listener’s maximum audible frequency (Hillock-Dunn et al. 2014; McCreery et al. 2014; Brennan et al. 2017). While a benefit of NFC has been observed for a wide variety of speech stimuli, including fricatives, words, and sentences presented both in quiet and in differing noise backgrounds (speech-shaped noise and multitalker babble: Alexander, Kopun, and Stelmachowicz 2014; Ellis and Munro 2015; Shehorn, Marrone, and Muller 2018), positive effects for NFC compared to conventional processing are often only observed for a subset of participants (Simpson, Hersbach, and McDermott 2005; Simpson, Hersbach, and McDermott 2006; Glista et al. 2009; Wolfe et al. 2010; Wolfe et al. 2011; Glista, Scollie, and Sulkers 2012; Park et al. 2012; Hillock-Dunn et al. 2014; McCreery et al. 2014; Picou, Marcrum, and Ricketts 2015). Collectively, these results suggest that while improvements in audibility are necessary to achieve improvements in speech recognition with NFC, other factors may mediate the listener’s ability to use these cues to support speech recognition.

Factors identified as contributing towards benefit from NFC include greater degree of hearing loss (Souza et al. 2013; Shehorn, Marrone, and Muller 2018; but see Brennan et al. 2014) and the extent to which audibility, as quantified with the Speech Intelligibility Index (SII), is increased for sounds from the source to destination region with NFC (McCreery et al. 2014). One potential factor that has not been explored is the ability to resolve changes in the spectrum over time (i.e., spectral resolution) with NFC. Spectral information can be resolved using the frequency-to-place map of the auditory system (i.e. auditory filtering), phase locking to the individual spectral components or envelope, and other mechanisms such as cochlear propagation and suppression—all of which are negatively impacted by SNHL (Vogten 1978; Ruggero 1994; Shamma and Klein 2000; Robles and Ruggero 2001). Spectral resolution can be measured using a spectral-ripple discrimination task where a listener has to discriminate broadband noises with contrasting distributions of spectral peaks and nulls—i.e. spectral ripples (Supin, Popov, and Milekhina 1994). Listeners appear to use excitation-based cues (i.e., spectral resolution) to detect these ripples, but also likely use temporal cues when the ripples per octave (RPO) are equal to or greater than 8 (Narne et al. 2016; Buss and Grose 2018). The ability to perceive rippled stimuli is thought to be influenced by degree of hearing loss, auditory filter bandwidth, and audibility—quantified by the pure-tone average, sensation level, or SII of the stimuli (Bernstein et al. 2013; Mehraei et al. 2014; Davies-Venn, Nelson, and Souza 2015; Kirby et al. 2015; Jorgensen et al. 2020). Individuals with greater degrees of hearing loss have wider auditory filters and can discriminate fewer RPOs between spectrally-modulated stimuli (Henry, Turner, and Behrens 2005; Davies-Venn, Nelson, and Souza 2015). When RPO thresholds are measured with hearing-aid amplification, children with a higher SII tend to have better RPO thresholds (Kirby et al. 2015). Due to inaudibility of the dips or portions of the noise spectrum, the ability to perceive spectral ripples degrades at low sensation levels and, presumably due to widening of the auditory filters, degrades again at high sensation levels (Isarangura et al. 2019; Jorgensen et al. 2020).

Varying measures of spectral-ripple discrimination have been shown to predict speech recognition in listeners with normal hearing and hearing loss (Henry, Turner, and Behrens 2005), cochlear implants (Henry and Turner 2003; Litvak et al. 2007; Won, Drennan, and Rubinstein 2007) and hearing aids (Davies-Venn, Nelson, and Souza 2015). Listeners with better spectral resolution abilities generally have better speech recognition than listeners with poorer spectral resolution abilities, as they are presumably better able to resolve differences in the speech spectrum that are crucial for identifying speech. The benefits of spectral resolution for speech recognition have been documented in quiet (Henry and Turner 2003; Henry, Turner, and Behrens 2005) and in steady-state noise (Won, Drennan, and Rubinstein 2007). Despite the known relationship of audibility to both spectral-ripple discrimination and speech recognition, few studies to date have determined the individual contribution of audibility and spectral-ripple discrimination to speech recognition and instead only compared spectral-ripple discrimination to speech recognition (e.g. Henry, Turner, and Behrens 2005). Possibly, individuals with better audibility can better detect spectral ripples and understand speech, simply due to having better audibility. Alternatively, individuals with both better audibility and better perception of spectral ripples better perceive speech relative to those with less audibility or poorer perception of spectral ripples. While Bernstein et al. (2013) and Mehraei et al. (2014) observed that both the SII and spectro-temporal modulation sensitivity predicted sentence recognition in noise, the relationship of spectro-temporal modulation sensitivity to ripple discrimination is unclear. Due to the ability of spectral-ripple thresholds to predict speech performance, there has been interest in adopting spectral-ripple testing to the clinic (Drennan et al. 2014; Gifford, Hedley-Williams, and Spahr 2014; Landsberger, Stupak, and Aronoff 2019).

Understanding how the perception of spectral ripples is influenced by NFC could help clinicians determine who is likely a good candidate for NFC. While previous research has suggested that the benefits from amplification may be mediated by listeners’ spectral resolution abilities (Davies-Venn, Nelson, and Souza 2015) and that spectral resolution is poorer with NFC (Kirby and Brown 2015), the impact of NFC on the ability of listeners to resolve spectral information is unclear. By compressing the speech spectrum to increase audibility, the spectral distinctiveness of the signal is reduced and this reduction in spectral distinctiveness may have a negative impact on a listener’s ability to resolve spectral cues—resulting in poorer speech recognition. Listeners with sensorineural hearing loss who are better able to resolve spectral cues with NFC may show a larger benefit for speech signals that have been spectrally compressed. However, these predictions have not been directly investigated in listeners while using amplification processed with NFC.

Despite the extensive literature linking spectral resolution with speech recognition, interest in adopting measures of spectral resolution to the clinic, and known impact of hearing-aid processing on access to temporal and spectral cues (Brennan, McCreery, and Jesteadt 2015; Jürgens et al. 2016; Brennan et al. 2018), there is only one study to date that documented the effects of frequency lowering on spectral-ripple discrimination thresholds and speech recognition (Kirby and Brown 2015). Using a start frequency of 1.5 kHz and compression ratios of 1.5, 2, 3, and 4, Kirby and Brown measured the identification of vowels, consonants, and sentences, along with spectral-ripple discrimination thresholds and the acoustic change complex for spectral-ripple stimuli. For their 10 adults with SNHL, Kirby and Brown observed poorer speech recognition and spectral-ripple discrimination thresholds with concomitant increases in the compression ratio, suggesting that spectral-ripple discrimination thresholds might be useful for predicting the effects of different NFC compression ratios on speech recognition. However, spectral-ripple discrimination thresholds were not used to predict performance with NFC and because the NFC start frequency and compression ratios were not fit individually to each participants degree of hearing loss, it is unclear if the same pattern observed by Kirby and Brown would occur when NFC is set to prescriptive targets for each individual’s hearing loss.

The purpose of this study was to evaluate the effect of NFC amplification, by comparing a condition with NFC (NFC-on) to a condition without NFC (NFC-off), on spectral resolution and word recognition in speech-shaped noise for adults with mild-to-severe sensorineural hearing losses. Word recognition was measured using speech-shaped noise because a) difficulty hearing in noisy situations is a common complaint for hearing-aid users and b) perceiving speech in noise necessitates being able to spectrally separate speech from background noise. RPO threshold and word recognition were measured for two conditions: NFC-off and NFC-on. We predicted that listeners with better RPO thresholds would have better word recognition with both NFC-off and NFC-on than listeners with poorer thresholds and that variance in individual benefit from NFC would be predictable from individual changes in RPO threshold between NFC-off and NFC-on. Because higher aided audibility has been identified as a factor that improves the perception of spectral ripples (Kirby et al. 2015) and speech recognition (Humes and Riker 1992; Brennan and Souza 2009), it was also hypothesized that listeners with higher aided audibility would have better RPO thresholds and word recognition. Observation of a relationship of SII to RPO threshold would support the argument that overall audibility of the ripple stimulus contributes to ripple perception and observation of a relationship of SII and RPO threshold to speech recognition would support the argument that both audibility and spectral resolution promote speech recognition.

Method

Participants

Twenty-four potential participants assented or consented to participate in the study but two child participants (ages 13 and 16 years), an adult participant with missing spectral-ripple discrimination thresholds, and two adult participants who did not improve in audibility with NFC-on (due to having maximum audible frequencies less than 1.5 kHz) were excluded. For the remaining nineteen participants, mean age was 52 years (range 20–73 years), right-ear pure-tone average (PTA, .5, 1, and 2 kHz) = 56 dB HL, and left-ear PTA = 53 dB HL. Participants had mild-to-severe sensorineural hearing loss, with Figure 1 displaying the minimum, mean and maximum audiometric thresholds. Audiometric thresholds were measured with insert earphones (ER3A) or supra-aural earphones (TDH-50P) following the procedures of American Speech-Language-Hearing Association (2005), except that threshold at 6 kHz was measured for each participant. One participant wore monaural amplification, and the remaining eighteen participants wore binaural amplification. The mean self-reported use time was 14.4 hours per day. Only one participant had previous experience with NFC in their own hearing aids. All participants were paid $15/hour for their participation. The study and consenting procedures were approved and overseen by the Boys Town National Research Hospital Institutional Review Board.

Figure 1:

Figure 1:

Open in a new tab

Audiometric thresholds (color online). Mean audiometric thresholds (dB HL) as a function of frequency (Hz) for the right (O) and left (X) ears. Hatched region represents the range of thresholds across participants.

Equipment

Equipment consisted of a sound-attenuating booth, GSI-1 Audiometer (Grason-Stadler, Eden Prairie, MN), Sennheiser SD25–1 Headphones (Wedemark, Germany), Verifit 1 hearing-aid electroacoustic measurement system (AudioScan, Dorchester, Ontario), a PC with a touchscreen monitor and running MATLAB (version 2012a; The MathWorks, Natick, MA), Lynx Two B soundcard (Lynx Studio Technology, Costa Mesa, CA), MiniMon Mon800 sound mixer (Behringer, Kirchardt, Germany), HP4 amplifier (PreSonus, Baton Rouge, LA), JBL Professional LSR2300 loudspeaker (Harmon, Northridge, CA), four Unitron Max 20 SP hearing aids (Waterloo, Canada), and System 824 sound level meter (Larson Davis, Pravo, UT).

Amplification

Listening conditions were wide-dynamic range compression with NFC-off or with NFC-on. A pair of Unitron Max 20 SP hearing aids were programmed with all advanced signal processing features deactivated, except feedback suppression. Eleven participants wore their personal earmolds and nine participants used temporary earmolds with a foam tip for the experiment. Hearing aid output for a 55 and 65 dB SPL speech input and 85 dB SPL swept pure tone, measured using a probe tube placed 25–30 mm from the inter-tragal notch, were set to match Desired Sensation Level (DSL m[i/o]-Adult: Scollie et al. 2005) prescriptive targets with NFC-off. The left panel of Figure 2 documents the difference between the output and target levels. Root mean square error (RMSE) for the fit to target (.25 to 6 kHz) was computed for each participant and was a mean of 8 dB and ranged from 3.6 to 16.9. The poor fits to target at 4 and 6 kHz contributed to these high RMSE values. Audibility was then quantified for the NFC-off condition, with the 65 dB SPL speech input, by the Verifit 1 using the one-third octave band procedure of the speech intelligibility index (SII: ANSI 1997) but without the 160 Hz band or hearing loss desensitization. A 30 dB (−15, +15 dB) range of speech was assumed and thresholds in dB HL were converted to dB SPL using the average real-ear-to-dial difference. The highest frequency where the average speech spectrum intersected the audiometric thresholds was noted as the maximum audible output frequency. The NFC settings were determined using the SoundRecover Fitting Assistant with a method described in McCreery et al. (2014), where the combination of start frequency and compression ratio that yield the greatest audible bandwidth were selected for each participant and ear. Figure 2 displays the range of NFC parameters that were selected, along with the range of maximum audible output and input frequencies. Because the long-term average speech spectrum decreases with higher frequency (Cox and Moore 1988), lower output with NFC-on than off can occur because the input level for the source region can be lower than the input level for the original sounds from the destination region. For that reason, hearing-aid output for a 65 dB SPL speech input was measured with NFC-on and if necessary, output was increased with NFC-on to match output with NFC-off. Otherwise, the same gain and compression parameters were used for NFC-off and NFC-on.

Figure 2:

Figure 2:

Open in a new tab

Fit to target and nonlinear frequency compression (NFC) settings. Target level was subtracted from output level, averaged for the two ears, and plotted (left panel). Due to a data collection error, target and output levels are missing for two of the participants. The NFC compression ratio (middle panel) and start frequency (SF), maximum audible output frequency with NFC-off (MAOF), and maximum audible input frequency with NFC-on (MAIF, right panel) for right (unshaded) and left (shaded) ears. Each box plot represents the median (horizontal line), mean (circle) and interquartile range (box). The error bars are the 5th and 95th percentiles.

Word Recognition Task

Three hundred monosyllabic consonant-vowel-consonant words from (McCreery and Stelmachowicz 2011) were used for the word recognition task. Each word contained at least one fricative or affricate sound (/s/, /z/, /f/, /v/, /ʧ/, /ʤ/, /ʃ/, /ʒ/, /θ/) and each vowel consisted of /a/, /i/, /l/, /ɛ/, /u/, or /ʌ/. These words were presented at 60 dBA to each participant, seated at 0-degrees azimuth and 1 meter from the loudspeaker, with unmodulated speech-shaped noise (accomplished using one-third octave filters) at 54 dBA for a 6 dB signal-to-noise ratio. A practice list of 10 words was used to familiarize the participants with the task, followed by two trials of each processing condition (NFC-off and NFC-on, in random order) with 75 words per trial. The practice and trial word lists were randomly generated for each participant.

Spectral-ripple discrimination task

Following the procedure of Won, Drennan, and Rubinstein (2007) spectrally rippled noise with a sampling frequency of 44.1 kHz, 13-dB depth, and 5-dB level rove (±2.5 dB) were generated using MATLAB software. The ripple depth of 13-dB, instead of the nominal 30-dB, was chosen to better fit within the dynamic range of individual participants with hearing loss. Two-thousand tones with random starting phase were spaced equally on a logarithmic frequency scale from 1,500 – 9,991 Hz, chosen to correspond to the lowest start frequency and highest maximum input frequency with the study hearing aid. For the reference ripple stimulus, the spectral modulation starting phase of the full-wave rectified sinusoidal spectral envelope was randomly selected from a uniform distribution (0 to 2π radian), and for each corresponding “oddball” stimulus, the phase was determined by adding π/2 to the phase of the reference ripple stimulus. This random starting phase of the spectral modulation was used to limit the ability of participants to rely exclusively on local intensity cues that fixed-phase stimuli might create (Aronoff and Landsberger 2013). The stimuli had 500 ms total duration, ramped with 150 ms rise/fall times, filtered to the long-term average speech spectrum, and were presented at 60 dBA through the loudspeaker.

A three-interval forced choice adaptive procedure was used to determine the 71% spectral-ripple discrimination threshold. Two reference stimuli and one oddball stimulus were presented for each trial. The ripple density step size was a ratio of 1.414. Each participant identified the interval that sounded different by pressing on a virtual button on the touchscreen monitor. Feedback was not provided. The threshold for a single adaptive track was estimated by averaging the ripple density (RPO) for the final 8 of 13 reversals. Here, higher spectral-ripple thresholds indicate better discrimination performance. One track was completed as practice, using a randomly chosen processing condition, followed by three adaptive tracks per processing condition, with the order of processing condition randomized for each participant. The final threshold was the arithmetic mean of these three adaptive tracks for each condition.

Analysis

Data consisted of percent-correct word recognition and RPO threshold, each with and without NFC, and better-ear aided SII (i.e., the ear with the highest SII value) for NFC-off. To examine the relationship between variables, the bivariate correlations among the dependent and independent variables were reported. To analyze mean differences in proportion correct and RPO threshold by processing condition (NFC-off, NFC-on), linear mixed models with random intercepts for each participant were conducted; followed by a linear mixed model that also included aided SII and RPO threshold as predictor variables and percent correct as the dependent variable.

Results

Figure 3 shows results for word recognition performance for each processing condition. The main effect of processing (p = 0.065, CI [−5.1, 0.2]), with mean scores of 36% with NFC-off (SD=16) and 34% with NFC-on (SD=14) was not significant. Figure 4 shows the results for spectral-ripple discrimination. The effect of processing condition on spectral-ripple discrimination threshold was significant (p = .010, CI [−1.1, −.2]) with mean 2.3 RPO for NFC-off (SD = 1.3) compared to 1.8 RPO for NFC-on (SD = 1.5).

Figure 3:

Figure 3:

Open in a new tab

Word recognition (color online). Percent correct word recognition for nonlinear frequency compression (NFC) off and on.

Figure 4:

Figure 4:

Open in a new tab

Spectral-ripple discrimination (color online). Circles connected with solid lines represent the change in spectral-ripple discrimination between nonlinear frequency compression (NFC) off (open) and on (closed) for individual participants.

Table 1 provides the correlations of the dependent and independent variables and Figure 5 depicts word recognition as a function of aided SII (left panel) and RPO (right panel) in both NFC-off and NFC-on conditions. While the aided SII was not significantly associated with word recognition (p = 0.732, CI [−2.2, 1.5]), the RPO threshold was significantly associated with word recognition (p = .049, CI [−75.5, −0.2]) and both the aided SII and the RPO threshold interacted significantly to influence word recognition (p = .042, CI [0.03, 1.26]). This significant interaction was due to a synergetic effect of aided SII and RPO, with word recognition being better for those with both higher aided SII and RPO than those with lower aided SII or RPO and this effect is illustrated in Figure 6. Notice that RPO increased as aided SII increased and that proportion correct was greater for participants with both a higher aided SII and RPO than participants with either a lower aided SII or RPO. Regarding the hypothesis that listeners with a smaller decrease in RPO from NFC-off to NFC-on would experience a larger benefit of NFC, RPO threshold did not interact significantly with processing condition (p = .848, CI [−24.4, 29.5]). In addition, the interactions of aided SII with processing condition (p = .614, CI [−0.8, 1.3]) and RPO threshold, aided SII, and processing condition (p = .833, CI [−0.5, 0.4]) were not significant.

Table 1:

Correlation of dependent and independent variables.

Aided SII WR NFC-off WR NFC-on RPO NFC-off
Aided SII
WR NFC-off 0.45
WR NFC-on 0.45 0.93***
RPO NFC-off 0.25 0.45 0.52*
RPO NFC-on 0.40 0.65** 0.72*** 0.76***
Open in a new tab
*

p<.05;

**

p<.01;

***

p<.001.

NFC = nonlinear frequency compression; RPO = ripples per octave; SII = speech intelligibility index; WR = word recognition.

Figure 5:

Figure 5:

Open in a new tab

Word recognition as a function of aided speech intelligibility index (SII, left panel) and ripples per octave (RPO, right panel) (color online).

Figure 6:

Figure 6:

Open in a new tab

Aided speech intelligibility index (SII) as a function of spectral-ripple discrimination. Each circle represents an individual participant, and the size of each circle represents the word-recognition score, with larger circles representing higher percent-correct word recognition.

Discussion

The primary goal of NFC processing is to increase audibility for high-frequency speech sounds. Yet, even when improvements in audibility can be made relative to a condition without NFC, some listeners do not appear to benefit from NFC processing (e.g. McCreery et al. 2014). It was hypothesized that this variability in benefit from NFC processing could be predicated on a participant’s change in RPO threshold from NFC-off to NFC-on. While listeners with a higher (better) RPO threshold understood more words than those with a lower RPO threshold, individual changes in word recognition between NFC-off and NFC-on were not predictable from a participant’s RPO threshold. It was also observed that participants with a higher aided SII obtained a higher RPO threshold, both aided SII and RPO threshold predicted word recognition, and while mean RPO threshold was lower by 0.5 with NFC-on than NFC-off, percent-correct word recognition did not change significantly between NFC-on and NFC-off. Here, we discuss the contributions of audibility and auditory filter bandwidth to ripple perception and the relationship of spectral-ripple thresholds with amplification to speech recognition.

Contribution of audibility to ripple perception

Factors posited to contribute to the perception of spectral ripples include audibility, the increased auditory filter bandwidths associated with sensorineural hearing loss, and other factors including temporal resolution and local loudness cues (Aronoff and Landsberger 2013). Evidence supporting the influence of audibility to ripple perception include the observations that RPO thresholds decrease at low sensation levels and individuals with higher aided SII values have better RPO thresholds. Specifically, the perception of rippled noise is known to degrade at low sensation levels (Isarangura et al. 2019; Jorgensen et al. 2020), ostensibly due to a combination of inaudibility of ripple nulls or portions of the spectrum at low sensation levels; consequently, the participants with a lower aided SII in this study and in Kirby et al. (2015) might have been less able to resolve spectral peaks due to inaudibility. While Kirby et al. (2019) did not find a relationship between aided SII and RPO thresholds, the use of participants with high levels of audibility (Mean aided SII=78) may have limited the observed relationship of RPO to aided SII in that study. The mean aided SII values were lower for this study (Mean aided SII=56) and Kirby et al. (2015; Mean SII=60). While the aided SII values predicted the RPO thresholds, note that both the better-ear pure-tone average (.5, 1, and 2 kHz; r = −.25, p = .128) and better-ear high-frequency pure-tone average (2, 4, and 6 kHz; r = −.23, p = .164) were not significantly correlated with RPO threshold. By taking into account the variations in sensation level across frequency, the SII is probably a better predictor for aided ripple detection than the PTA (which only reflects degree of hearing loss grossly).

While the work presented herein provides supporting evidence that the perception of spectral ripples is associated with audibility, individual differences in auditory filter bandwidth and presentation level are also likely contributing factors. Auditory filter bandwidth increases as degree of hearing loss increases, in turn leading to reduced distinctness of spectral peaks and poorer perception of spectral ripples (Summers and Leek 1994; Davies-Venn, Nelson, and Souza 2015; Isarangura et al. 2019; Jorgensen et al. 2020). Because auditory filter bandwidths are wider at high presentation levels, RPO thresholds decline for elevated presentation levels (Isarangura et al. 2019; Jorgensen et al. 2020). Consider that the prescribed gain was higher for those with greater hearing loss, resulting in a higher presentation level than for those with lessor hearing loss. Therefore, the use of a higher presentation level may have contributed to poorer ripple thresholds for those with poorer audibility. In summary, individual variations in audibility, auditory filter bandwidth, and presentation level contributed towards the poorer perception of spectral ripples observed for those with a lower aided SII.

Relationship of aided SII and RPO to speech recognition

Individuals with greater unaided perception of spectral ripples (Henry, Turner, and Behrens 2005; Won, Drennan, and Rubinstein 2007; Davies-Venn, Nelson, and Souza 2015) or higher aided SII (Humes and Riker 1992) demonstrate greater understanding of speech and both better spectral-temporal modulation sensitivity and higher SII contribute to sentence recognition (Bernstein et al. 2013; Mehraei et al. 2014). Consistent with this prior work, our participants with greater access to spectral information (i.e., higher aided RPO thresholds) and aided SII values had better speech recognition than participants with lower access to spectral information or aided SII. Note that the work presented herein suggests that both aided SII and ability to resolve spectral information contribute, with word recognition being best for those with both higher aided SII and better RPO thresholds. This work highlights the importance of quantifying both audibility and aided access to spectral information when relating measures of spectral resolution to speech recognition.

Less clear is how to adopt spectral ripple tests for the clinic. For example, it has been argued that aided RPO thresholds could be used to predict speech recognition outcomes for individuals whose speech recognition ability cannot be tested, such as young children or patients who cannot be tested in their native language (Kirby and Brown 2015). The positive relationship of RPO thresholds to speech recognition supports this notion; but this argument is tempered by the fact that while the mean RPO threshold decreased, significantly, from 2.3 with NFC-off to 1.8 with NFC-on, the proportion-correct word recognition did not change significantly between NFC-off (M=34%) and NFC-on (M=31%) and individual changes in RPO threshold between NFC-off and NFC-on did not correspond to a similar change in word recognition. The establishment of a relationship between changes in NFC settings, RPO thresholds, and speech recognition would strengthen the argument for incorporating a measure of ripple perception into the clinic.

Our observation of less access to spectral information with NFC-on relative to NFC-off is consistent with Kirby and Brown (2015), who also observed a reduction in RPO threshold with NFC. By compressing spectral peaks and nulls, NFC reduces spectral distinctness leading to a poorer ability to resolve spectral cues. By using a noise bandwidth that corresponded to the potential bandwidth of NFC processing, we were able to capture the influence of NFC processing on spectral resolution. However, by not including frequencies below the start frequency, but known to contribute to speech recognition (ANSI 1997), this measure may have overestimated the potential effect of NFC processing on access to spectral cues in speech and consequently on speech recognition. Consider also that the change in RPO threshold and hence access to spectral information (M=0.5 RPO) may have been inconsequential to the perception of speech. Future work using a wider range of NFC settings might reveal a relationship between changes in RPO thresholds and speech recognition with NFC that wasn’t observed for this study.

Because there was not a significant change in speech recognition with NFC, this may have limited the ability to observe a relationship between RPO threshold and NFC benefit. Benefit of NFC has varied across studies, with some studies finding an overall benefit (Bohnert, Nyffeler, and Keilmann 2010; Alexander, Kopun, and Stelmachowicz 2014; McCreery et al. 2014; Ellis and Munro 2015; Brennan et al. 2017; Shehorn, Marrone, and Muller 2018) and others finding no difference in speech recognition between NFC-off and NFC-on (Park et al. 2012; Picou, Marcrum, and Ricketts 2015). While some studies documented negative effects of NFC, these studies are unique in that—instead of documenting improvements in audibility—excessive compression (increased distortion without any concomitant improvements in the audibility of speech) were used (Arehart et al. 2013; Souza et al. 2013; Kirby and Brown 2015). Reasons for a lack of benefit in this study are unclear but here we consider the potential contributions of choice of fitting method, stimuli, and degree of hearing loss.

The NFC settings were adjusted for each participant using a method previously documented to improve speech recognition with NFC (McCreery et al. 2014; Brennan et al. 2017). Because inability to understand speech in background noise is a common complaint amongst hearing aid users, this study presented the words in background noise. While the background noise may have prevented the participants from differentiating the background noise from frication noise for the words with fricatives, a similar set of stimuli to the stimuli used here were used by McCreery et al. (2014), wherein a benefit of NFC was observed for words presented in speech-shaped background noise. While some studies did not observe a benefit of NFC for speech presented in background noise (Simpson, Hersbach, and McDermott 2006; Hopkins et al. 2014; Kokx-Ryan et al. 2015; Picou, Marcrum, and Ricketts 2015; Miller, Bates, and Brennan 2016), other studies instead observed a benefit of NFC in background noise, including vowel-consonant-vowels and low context sentences in babble noise (Ellis and Munro 2015; Shehorn, Marrone, and Muller 2018) and fricatives presented in speech shaped noise (Alexander, Kopun, and Stelmachowicz 2014). These mixed finding suggest that the inclusion of background noise in the study design was not a contributing factor to the lack of a benefit observed for NFC.

Because some studies observed that individuals with greater hearing loss benefited more from NFC (Souza et al. 2013; Shehorn, Marrone, and Muller 2018), possibly including individuals with a wider range of hearing losses than used here would have revealed a benefit that was not observed. The mean maximum audible output frequency (i.e., frequency where LTASS cross threshold of hearing to become inaudible) was 4300 Hz and ranged from 2300 to 6700 Hz (see figure 2); while some of the participants with a higher MAOF might not traditionally be considered a candidate for NFC, consider too that there is disagreement about the relationship of high-frequency audibility to benefit from frequency lowering. Both Ellis and Munro (2015) and Picou, Marcrum, and Ricketts (2015) did not observe a relationship between degree of hearing loss and benefit from NFC and Brennan et al. (2017) observed that listeners with less hearing loss experienced greater benefit—possibly due to a better ability, secondary to a less damaged system, to use the lowered frequency components.

Conclusion

The primary aim of this study was to evaluate the impact of spectral resolution and audibility on word recognition with NFC for adults with mild-to-severe high-frequency hearing loss. Although word recognition did not change significantly between NFC-off and NFC-on, the mean RPO threshold was significantly poorer, by 0.5 RPO, with NFC. While prior work has examined the contributions of audibility or spectral-ripple discrimination to speech recognition, this study examined the contribution of both together and observed that together RPO threshold and aided SII predicted word recognition. However, neither predicted benefit from NFC; thus, it remains unclear how to utilize measures of spectral-ripple discrimination to set the NFC start frequency or compression ratio or to determine candidacy for NFC. These results support the notion that individual variability in spectral resolution contribute to a listener’s ability to resolve speech information. More research is needed to clarify the role of audibility in the discrimination of ripples for listeners with SNHL and the relationship of RPO threshold to an individual’s candidacy for frequency lowering.

Acknowledgements

The authors wish to thank Jong Ho Won for providing the spectral-ripple discrimination program and Brianna Byllesby, Sarah Garvey, and Kris Fernau for assistance with data collection and data preparation.

Sources of Funding

This work was supported by Unitron who provided the hearing aids and by NIH/NIDCD R01 DC004300 (Patricia G. Stelmachowicz), R01 DC013591 (Ryan McCreery), P30 DC004662 (Boys Town National Research Hospital), R21DC017588 (Marc Brennan).

Abbreviations:

MAOF

maximum audible output frequency with wide-dynamic-range compression

MAIF

maximum audible input frequency with NFC

NFC

nonlinear frequency compression

PTA

pure-tone average

RPO

ripples per octave

RMSE

root mean square error

SNHL

sensorineural hearing loss

SII

speech intelligibility index

WDRC

wide-dynamic-range compression

Footnotes

This project was previously presented as a poster. McCreery, R.W., Brennan, M.A., Won, J.H., Lewis, D.E. & Kopun, J. (2014) Audibility and spectral resolution as predictors of speech recognition with nonlinear frequency compression. International Hearing Aid Conference, Tahoe City, CA – August.

Conflicts of Interest

There are no other conflicts of interest, financial, or otherwise.

References

  1. American National Standards Institute (ANSI). (1997). Methods for calculation of speech intelligibility index (ANSI S3.5) New York: American National Standards Institute. [Google Scholar]
  2. Agrawal Y, Platz E, Niparko J. (2008) Prevalence of hearing loss and differences by demographic characteristics among us adults: Data from the national health and nutrition examination survey, 1999–2004. Arch Intern Med 168: 1522–1530. [DOI] [PubMed] [Google Scholar]
  3. Alexander JM, Kopun JG, Stelmachowicz PG. (2014) Effects of frequency compression and frequency transposition on fricative and affricate perception in listeners with normal hearing and mild to moderate hearing loss. Ear Hear 35: 519–532. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. American Speech-Language-Hearing Association. (2005) Guidelines for manual pure-tone threshold audiometry Rockville, MD: Author. [Google Scholar]
  5. Arehart KH, Souza P, Baca R, Kates JM. (2013) Working memory, age, and hearing loss: Susceptibility to hearing aid distortion. Ear Hear 34: 251–260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Aronoff JM, Landsberger DM. (2013) The development of a modified spectral ripple test. J Acoust Soc Am 134: EL217–EL222. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bernstein JGW, Mehraei G, Shamma S, Gallun FJ, Theodoroff SM, Leek MR. (2013) Spectrotemporal modulation sensitivity as a predictor of speech intelligibility for hearing-impaired listeners. J Am Acad Audiol 24: 293–306. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bohnert A, Nyffeler M, Keilmann A. (2010) Advantages of a non-linear frequency compression algorithm in noise. Eur Arch Otorhinolaryngol 267: 1045–1053. [DOI] [PubMed] [Google Scholar]
  9. Boothroyd A, Mulhearn B, Gong J, Ostroff J. (1996) Effects of spectral smearing on phoneme and word recognition. J Acoust Soc Am 100: 1807–1818. [DOI] [PubMed] [Google Scholar]
  10. Brennan M, Souza P. (2009) Effects of expansion on consonant recognition and consonant audibility. J Am Acad Audiol 20: 119–127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Brennan MA, McCreery R, Kopun J, Hoover B, Alexander J, Lewis D, Stelmachowicz PG. (2014) Paired comparisons of nonlinear frequency compression, extended bandwidth, and restricted bandwidth hearing aid processing for children and adults with hearing loss. J Am Acad Audiol 25: 983–998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Brennan MA, McCreery RW, Jesteadt W. (2015) The influence of hearing-aid compression on forward-masked thresholds for adults with hearing loss. J Acoust Soc Am 138: 2589–2597. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Brennan MA, McCreery RW, Kopun J, Lewis D, Alexander JM, Stelmachowicz PG. (2016) Masking release in children and adults with hearing loss when using amplification. J Speech Lang Hear Res 59: 101–121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Brennan MA, Lewis D, McCreery R, Kopun J, Alexander JM. (2017) Listening effort and speech recognition with frequency compression amplification for children and adults with hearing loss. J Am Acad Audiol 28: 823–837. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Brennan MA, McCreery RW, Buss E, Jesteadt W. (2018) The influence of hearing aid gain on gap-detection thresholds for children and adults with hearing loss. Ear Hear 39: 969–979. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Buss E, Grose J. (2018) Auditory sensitivity to spectral modulation phase reversal as a function of modulation depth. PLoS One 13: e0195686. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Cox RM, Moore JN. (1988) Composite speech spectrum for hearing aid gain prescriptions. J Speech Hear Res 31: 102–107. [DOI] [PubMed] [Google Scholar]
  18. Davies-Venn E, Nelson P, Souza P. (2015) Comparing auditory filter bandwidths, spectral ripple modulation detection, spectral ripple discrimination, and speech recognition: Normal and impaired hearinga). J Acoust Soc Am 138: 492–503. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Drennan WR, Anderson ES, Won JH, Rubinstein JT. (2014) Validation of a clinical assessment of spectral-ripple resolution for cochlear implant users. Ear Hear 35: e92–e98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Ellis RJ, Munro KJ. (2015) Predictors of aided speech recognition, with and without frequency compression, in older adults. Int J Audiol: 1–9. [DOI] [PubMed]
  21. Ellis RJ, Munro KJ. (2015) Benefit from, and acclimatization to, frequency compression hearing aids in experienced adult hearing-aid users. Int J Audiol 54: 37–47. [DOI] [PubMed] [Google Scholar]
  22. Gifford RH, Hedley-Williams A, Spahr AJ. (2014) Clinical assessment of spectral modulation detection for adult cochlear implant recipients: A non-language based measure of performance outcomes. Int J Audiol 53: 159–164. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Glista D, Scollie S, Bagatto M, Seewald R, Parsa V, Johnson A. (2009) Evaluation of nonlinear frequency compression: Clinical outcomes. Int J Audiol 48: 632–644. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Glista D, Scollie S, Sulkers J. (2012) Perceptual acclimatization post nonlinear frequency compression hearing aid fitting in older children. J Speech Lang Hear Res 55: 1765–1787. [DOI] [PubMed] [Google Scholar]
  25. Gordon-Salant S, Fitzgibbons PJ. (1993) Temporal factors and speech recognition performance in young and elderly listeners. J Speech Hear Res 36: 1276–1285. [DOI] [PubMed] [Google Scholar]
  26. Henry BA, Turner CW. (2003) The resolution of complex spectral patterns by cochlear implant and normal-hearing listeners. J Acoust Soc Am 113: 2861–2873. [DOI] [PubMed] [Google Scholar]
  27. Henry BA, Turner CW, Behrens A. (2005) Spectral peak resolution and speech recognition in quiet: Normal hearing, hearing impaired, and cochlear implant listeners. J Acoust Soc Am 118: 1111–1121. [DOI] [PubMed] [Google Scholar]
  28. Hillock-Dunn A, Buss E, Duncan N, Roush PA, Leibold LJ. (2014) Effects of nonlinear frequency compression on speech identification in children with hearing loss. Ear Hear 35: 353–365. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Hopkins K, Khanom M, Dickinson A-M, Munro KJ. (2014) Benefit from non-linear frequency compression hearing aids in a clinical setting: The effects of duration of experience and severity of high-frequency hearing loss. Int J Audiol 53: 219–28. [DOI] [PubMed] [Google Scholar]
  30. Humes LE, Riker S. (1992) Evaluation of two clinical versions of the articulation index. Ear Hear 13: 406–409. [DOI] [PubMed] [Google Scholar]
  31. Isarangura S, Eddins AC, Ozmeral EJ, Eddins DA. (2019) The effects of duration and level on spectral modulation perception. J Speech Lang Hear Res 62: 3876–3886. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Jorgensen EJ, McCreery RW, Kirby BJ, Brennan M. (2020) Effect of level on spectral-ripple detection threshold for listeners with normal hearing and hearing loss. J Acoust Soc Am 148: 908–917. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Jürgens T, Clark NR, Lecluyse W, Meddis R. (2016) Exploration of a physiologically-inspired hearing-aid algorithm using a computer model mimicking impaired hearing. Int J Audiol 55: 346–357. [DOI] [PubMed] [Google Scholar]
  34. Kimlinger C, McCreery R, Lewis D. (2015) High-frequency audibility: The effects of audiometric configuration, stimulus type, and device. J Am Acad Audiol 26: 128–137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Kirby BJ, Brown CJ. (2015) Effects of nonlinear frequency compression on acc amplitude and listener performance. Ear Hear 36: e261–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Kirby BJ, Browning JM, Brennan MA, Spratford M, McCreery RW. (2015) Spectro-temporal modulation detection in children. J Acoust Soc Am 138: EL465–EL468. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Kirby BJ, Spratford M, Klein KE, McCreery RW. (2019) Cognitive abilities contribute to spectro-temporal discrimination in children who are hard of hearing. Ear Hear 40: 645–650. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Kochkin S (2005) Marketrak vii: Customer satisfaction with hearing instruments in the digital age. Hear J 58: 30–39. [Google Scholar]
  39. Kokx-Ryan M, Cohen J, Cord MT, Walden TC, Makashay MJ, Sheffield BM, Brungart DS. (2015) Benefits of nonlinear frequency compression in adult hearing aid users. J Am Acad Audiol 26: 838–55. [DOI] [PubMed] [Google Scholar]
  40. Landsberger DM, Stupak N, Aronoff JM. (2019) Spectral-temporally modulated ripple test lite for computeRless measurement (SLRM): A nonlinguistic test for audiology clinics. Ear Hear 40: 1253–1255. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Litvak LM, Spahr AJ, Saoji AA, Fridman GY. (2007) Relationship between perception of spectral ripple and speech recognition in cochlear implant and vocoder listeners. J Acoust Soc Am 122: 982–991. [DOI] [PubMed] [Google Scholar]
  42. McCreery RW, Stelmachowicz PG. (2011) Audibility-based predictions of speech recognition for children and adults with normal hearing. J Acoust Soc Am 130: 4070–4081. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. McCreery RW, Alexander J, Brennan MA, Hoover B, Kopun J, Stelmachowicz PG. (2014) The influence of audibility on speech recognition with nonlinear frequency compression for children and adults with hearing loss. Ear Hear 35: 440–447. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Mehraei G, Gallun FJ, Leek MR, Bernstein JGW. (2014) Spectrotemporal modulation sensitivity for hearing-impaired listeners: Dependence on carrier center frequency and the relationship to speech intelligibility. J Acoust Soc Am 136: 301–316. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Miller CW, Bates E, Brennan M. (2016) The effects of frequency lowering on speech perception in noise with adult hearing aid users. Int J Audiol 55: 305–312. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Narne VK, Sharma M, Van Dun B, Bansal S, Prabhu L, Moore BC. (2016) Effects of spectral smearing on performance of the spectral ripple and spectro-temporal ripple tests. J Acoust Soc Am 140: 4298–4306. [DOI] [PubMed] [Google Scholar]
  47. Park LR, Teagle HFB, Buss E, Roush PA, Buchman CA. (2012) Effects of frequency compression hearing aids for unilaterally implanted children with acoustically amplified residual hearing in the nonimplanted ear. Ear Hear 33: e1–e12. [DOI] [PubMed] [Google Scholar]
  48. Picou EM, Marcrum SC, Ricketts TA. (2015) Evaluation of the effects of nonlinear frequency compression on speech recognition and sound quality for adults with mild to moderate hearing loss. Int J Audiol 54: 162–169. [DOI] [PubMed] [Google Scholar]
  49. Robles L, Ruggero MA. (2001) Mechanics of the mammalian cochlea. Physiol Rev 81: 1305–1352. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Ruggero MA. (1994) Cochlear delays and traveling waves: Comments on ‘experimental look at cochlear mechanics’:[a. Dancer, audiology 1992; 31: 301–312] ruggero. Audiology 33: 131–142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Scollie S, Seewald R, Cornelisse L, Moodie S, Bagatto M, Laurnagaray D, Beaulac S, Pumford J. (2005) The desired sensation level multistage input/output algorithm. Trends Amplif 9: 159–197. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Shamma S, Klein D. (2000) The case of the missing pitch templates: How harmonic templates emerge in the early auditory system. J Acoust Soc Am 107: 2631–2644. [DOI] [PubMed] [Google Scholar]
  53. Shehorn J, Marrone N, Muller T. (2018) Speech perception in noise and listening effort of older adults with nonlinear frequency compression hearing aids. Ear Hear 39: 215–225. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Simpson A, Hersbach AA, McDermott HJ. (2005) Improvements in speech perception with an experimental nonlinear frequency compression hearing device. Int J Audiol 44: 281–292. [DOI] [PubMed] [Google Scholar]
  55. Simpson A, Hersbach AA, McDermott HJ. (2006) Frequency-compression outcomes in listeners with steeply sloping audiograms. Int J Audiol 45: 619–29. [DOI] [PubMed] [Google Scholar]
  56. Souza PE, Boike KT, Witherell K, Tremblay K. (2007) Predictions of speech recognition from audibility in older listeners with hearing loss: Effects of age, amplification, and background noise. J Am Acad Audiol 18: 54–65. [DOI] [PubMed] [Google Scholar]
  57. Souza PE, Arehart KH, Kates JM, Croghan NBH, Gehani N. (2013) Exploring the limits of frequency lowering. J Speech Lang Hear Res 56: 1349–1363. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Summers V, Leek MR. (1994) The internal representation of spectral contrast in hearing-impaired listeners. J Acoust Soc Am 95: 3518–3528. [DOI] [PubMed] [Google Scholar]
  59. Supin AY, Popov VV, Milekhina ON. (1994) Frequency resolving power of the human’s hearing. Neurosci Lett 165: 195–198. [DOI] [PubMed] [Google Scholar]
  60. Swanepoel R, Oosthuizen DJJ, Hanekom JJ. (2012) The relative importance of spectral cues for vowel recognition in severe noise. J Acoust Soc Am 132: 2652–2662. [DOI] [PubMed] [Google Scholar]
  61. Vogten L (1978) Low-level pure-tone masking: A comparison of’’tuning curves’’obtained with simultaneous and forward masking. J Acoust Soc Am 63: 1520–1527. [DOI] [PubMed] [Google Scholar]
  62. Wolfe J, John A, Schafer E, Nyffeler M, Boretzki M, Caraway T, Hudson M. (2010) Evaluation of nonlinear frequency compression for school-age children with moderate to moderately severe hearing loss. J Am Acad Audiol 21: 618–628. [DOI] [PubMed] [Google Scholar]
  63. Wolfe J, John A, Schafer E, Nyffeler M, Boretzki M, Caraway T, Hudson M. (2011) Long-term effects of non-linear frequency compression for children with moderate hearing loss. Int J Audiol 50: 396–404. [DOI] [PubMed] [Google Scholar]
  64. Won JH, Drennan WR, Rubinstein JT. (2007) Spectral-ripple resolution correlates with speech reception in noise in cochlear implant users. Journal of the Association for Research in Otolaryngology 8: 384–392. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Zwicker E, Schorn K. (1978) Psychoacoustical tuning curves in audiology. Audiology 17: 120–140. [DOI] [PubMed] [Google Scholar]

RESOURCES