Featured ArticleResearch Articles, Systems/Circuits

Distorted Tonotopy Severely Degrades Neural Representations of Connected Speech in Noise following Acoustic Trauma

Satyabrata Parida and Michael G. Heinz
Journal of Neuroscience 23 February 2022, 42 (8) 1477-1490; DOI: https://doi.org/10.1523/JNEUROSCI.1268-21.2021

Abstract

Listeners with sensorineural hearing loss (SNHL) struggle to understand speech, especially in noise, despite audibility compensation. These real-world suprathreshold deficits are hypothesized to arise from degraded frequency tuning and reduced temporal-coding precision; however, peripheral neurophysiological studies testing these hypotheses have been largely limited to in-quiet artificial vowels. Here, we measured single auditory-nerve-fiber responses to a connected speech sentence in noise from anesthetized male chinchillas with normal hearing (NH) or noise-induced hearing loss (NIHL). Our results demonstrated that temporal precision was not degraded following acoustic trauma, and furthermore that sharpness of cochlear frequency tuning was not the major factor affecting impaired peripheral coding of connected speech in noise. Rather, the loss of cochlear tonotopy, a hallmark of NH, contributed the most to both consonant-coding and vowel-coding degradations. Because distorted tonotopy varies in degree across etiologies (e.g., noise exposure, age), these results have important implications for understanding and treating individual differences in speech perception for people suffering from SNHL.

SIGNIFICANCE STATEMENT Difficulty understanding speech in noise is the primary complaint in audiology clinics and can leave people with sensorineural hearing loss (SNHL) suffering from communication difficulties that affect their professional, social, and family lives, as well as their mental health. We measured single-neuron responses from a preclinical SNHL animal model to characterize salient neural-coding deficits for naturally spoken speech in noise. We found the major mechanism affecting neural coding was not a commonly assumed factor, but rather a disruption of tonotopicity, the systematic mapping of acoustic frequency to cochlear place that is a hallmark of normal hearing. Because the degree of distorted tonotopy varies across hearing-loss etiologies, these results have important implications for precision audiology approaches to diagnosis and treatment of SNHL.

Introduction

Individuals with sensorineural hearing loss (SNHL) demonstrate speech-perception deficits, especially in noise, which are often not resolved, even with state-of-the-art hearing-aid strategies and noise-reduction algorithms (McCormack and Fortnum, 2013; Lesica, 2018). In fact, difficulty understanding speech in noise is the number-one complaint in audiology clinics (Chung, 2004). Although audibility is a factor contributing to these difficulties (Phatak and Grant, 2014), suprathreshold deficits associated with dynamic spectrotemporal cues substantially contribute as well, especially in noise (Festen and Plomp, 1983; Zeng and Turner, 1990). Unfortunately, neurophysiological studies of impaired speech coding in the auditory nerve (AN), the first neural site affected by cochlear components of SNHL (Trevino et al., 2019), have been primarily limited to synthetic vowels in quiet and have not included everyday sounds with natural dynamics.

Common suprathreshold factors hypothesized to underlie speech-perception-in-noise deficits following SNHL are reduced sharpness of cochlear frequency tuning, degraded temporal-coding precision, and cochlear synaptopathy. Reduced frequency-tuning sharpness, often observed in listeners with SNHL (Glasberg and Moore, 1986), limits the ability to resolve spectral components of speech and allows more noise into auditory filters (Horst, 1987; Moore, 2007). Reduced perceptual frequency selectivity parallels broader tuning observed in physiological responses following outer-hair-cell dysfunction (Liberman and Dodds, 1984a; Ruggero and Rich, 1991). A second hypothesized suprathreshold deficit is a loss of temporal-coding precision (Lorenzi et al., 2006; Moore, 2008; Halliday et al., 2019). While one study has reported a decrease in AN phase-locking following SNHL (Woolf et al., 1981), others did not find a degradation (Harrison and Evans, 1979; Kale and Heinz, 2010); however, these studies have been limited to laboratory stimuli (e.g., tones). Finally, cochlear synaptopathy progresses with age and noise exposure (Kujawa and Liberman, 2009; Fernandez et al., 2015) and has been hypothesized to underlie perceptual deficits in listeners even without audiometric loss.

Another neural suprathreshold factor that may contribute to speech-coding degradations is the change in on-frequency versus off-frequency sensitivity (i.e., reduced tip-to-tail ratio; TTR) observed in AN fibers following noise-induced hearing loss (NIHL). This phenomenon, which results from reduced tip sensitivity and/or tail hypersensitivity, has been well characterized with tonal frequency-tuning curves (FTCs; Liberman and Dodds, 1984a). These effects can produce tonotopic distortions with important implications for complex-sound processing but have only recently begun to be explored for nontonal stimuli such as broadband noise (Henry et al., 2016, 2019).

Several neurophysiological studies have investigated speech coding following SNHL but not with connected speech in noise, for which listeners with SNHL struggle the most (Young, 2012; Sayles and Heinz, 2017). NH AN speech-coding studies have predominantly used short-duration synthesized vowel-like and consonant-like stimuli in quiet (Young and Sachs, 1979; Sinex and Geisler, 1983; Delgutte and Kiang, 1984b,c), with the few exploring background noise limited to vowels (Sachs et al., 1983; Delgutte and Kiang, 1984a; Geisler and Gamble, 1989; Geisler and Silkes, 1991; Silkes and Geisler, 1991). A few studies of NH coding have used connected speech sentences but not in noise (Kiang and Moxon, 1974; Delgutte et al., 1998; Young, 2008). Speech-coding studies of hearing-impaired (HI) animals have been limited to vowel-like stimuli in quiet (e.g., Miller et al., 1997; Schilling et al., 1998). Connected speech differs from synthetic speech tokens in its dynamic structure, with most information contained in its time-varying properties, like formant transitions and spectrotemporal modulations (Elliott and Theunissen, 2009; Elhilali, 2019). Therefore, accurate characterization of how hearing loss affects the neural coding of the time-varying properties in connected speech is important but missing from our current understanding of the underlying mechanisms for why listeners with hearing loss struggle to understand speech in noise.

Here, we used a connected speech sentence and collected AN-fiber spike trains from anesthetized chinchillas with either NH or NIHL. The sentence was also mixed with speech-shaped noise at perceptually relevant signal-to-noise ratios (SNRs). We analyzed dynamic formant transitions in vowel responses using new nonstationary analyses based on frequency demodulation of alternating-polarity peristimulus-time histograms (PSTHs; Parida et al., 2021), and analyzed onset and sustained responses to fricatives and stop consonants. Our results provide insights into the physiological suprathreshold mechanisms that do and do not contribute to degraded connected-speech coding in noise, specifically highlighting the important role distorted tonotopy plays in increased noise susceptibility following SNHL. These findings have important implications for better understanding individual differences in speech perception in people with SNHL.

Materials and Methods

All code and data accompanying this paper can be found at the following link: https://github.com/HeinzLabPurdue/DT_SpINcoding-paper.

Ethics statement

All procedures described below followed PHS-issued guidelines and were approved by Purdue University Animal Care and Use Committee (Protocol No. 1111000123).

Animal model

Young male chinchillas (less than one year old, weighing between 400 and 700 g) were used in all experiments. A focus of the present study was the effect of noise exposure on the spread of AN-fiber thresholds, which was expected to affect the relative audibility of vowels and consonants. Since accurate quantification of the AN-fiber threshold distribution requires pooling across animals, only male chinchillas were used here since sex differences in noise susceptibility (with females generally less susceptible than males) have been shown in chinchillas (Trevino et al., 2019). This experimental design was chosen to avoid a known sex-related factor that would confound our estimate of the AN-fiber threshold distribution following noise exposure (by adding an unrelated source of variability) if we were to pool across animals of different sexes. While this approach allowed us to address one of our specific hypotheses more accurately for the noise-exposure paradigm used here, it did not allow us address the effects of variation in the degree of noise-induced damage (including because of sex differences).

Animals were socially housed in groups of two until they underwent any anesthetized procedure, after which they recovered in their own cage. All animals received daily nonauditory environmental enrichment including dietary treats and chewing toys. The animal facility was maintained in a 12/12 h light/dark cycle.

Noise exposure and noninvasive physiological recordings

Detailed descriptions of the noise-exposure and noninvasive physiological recording procedures are provided elsewhere (Parida and Heinz, 2021), with brief descriptions provided here. A single discrete noise exposure [116 dB SPL (C-weighted), 2-h duration, octave-band noise centered at 500 Hz] using an enclosed subwoofer (Selenium 10PW3, Harman; placed ∼30 cm above the animal's head) was used to induce NIHL. Noise levels were calibrated at the entrance of the ear canal using a sound-level meter (886-2, Simpson). Animals were allowed at least two weeks to recover following noise exposure before any physiological recordings were made to minimize temporary threshold shift effects (Miller et al., 1963). Animals were anesthetized using xylazine (2–3 mg/kg, s.c.) and ketamine (30–40 mg/kg, i.p.) for data recordings and noise exposure. The rectal temperature of all anesthetized animals was maintained at 37°C using a feedback-controlled heating pad (50–7053F, Harvard Apparatus). Atipamezole (0.4–0.5 mg/kg, i.p.) was used to facilitate faster recovery from anesthesia following noninvasive experiments.

Auditory brainstem responses (ABRs) to tone pips and speech-evoked ABRs (sABRs) were recorded using three subdermal needle electrodes in a vertical montage (vertex to mastoid, differential mode, common ground near animals' nose; Henry et al., 2011; Zhong et al., 2014). ABRs (frequency range = 0.3–3 kHz) and sABRs (30 Hz to 1 kHz) were bandlimited using analog filters (ISO–80, World Precision Instruments; 2400A, Dagan). Distortion-product otoacoustic emission (DPOAEs) level was measured using an in-ear microphone (Etymotic ER–10B, Etymotic Research), following in-ear calibration for each animal. Calibrated sound was presented using ER2 loudspeakers (Etymotic Research) for electrophysiological recordings. Sound presentation and data recordings were controlled using a custom-integrated system of hardware (Tucker-Davis Technologies; National Instruments) and software (MATLAB, The MathWorks). For the HI animals, these physiological recordings (i.e., ABR, DPOAE, and sABR) were obtained before as well as after (range = 14–42 d) the noise exposure.

Surgical preparation and single-unit recordings

Detailed surgical-preparation and neurophysiological-recording procedures are described by Henry et al. (2016) and are only briefly described here. Anesthesia was induced with the same doses of xylazine/ketamine as used for ABRs and was maintained with sodium pentobarbital (∼7.5 mg/kg/h, i.p.). Animals were supplemented with lactated Ringer's solution during experiments (∼1 ml/h), which typically lasted 18–24 h. A posterior fossa approach was employed for the craniotomy in the right ear, following venting of the bulla with 30 cm of polyethylene tubing to maintain middle-ear pressure.

Spike trains were recorded from single AN fibers using glass micropipettes (impedance between 10 and 50 MΩ). Recordings were amplified (2400A, Dagan) and filtered from 0.03 to 6 kHz (3550, Krohn-Hite). Isolated spikes were identified using a time-amplitude window discriminator (BAK Electronics) and stored digitally with 10-µs resolution. Experiments were terminated if sudden shifts in FTC threshold and tuning were observed for two or more AN fibers, following which a lethal dose of Euthasol (2 ml, i.p.; Virbac AH) was administered. Single-unit data are from 15 NH (286 AN fibers) and 6 HI (119 AN fibers) chinchillas. For HI animals, AN experiments were done between 20 and 61 d following noise exposure for five animals and after 339 d for one animal (with no obvious difference in responses compared with other HI animals). These intervals following noise exposure are all expected to be long enough to avoid any temporary threshold shifts (Miller et al., 1963).

Stimuli

Screening (ABR and DPOAE) experiments

For ABRs, tone pips (5-ms duration, 1-ms on and off ramps) ranging from 0.5 to 8 kHz (octave spaced) were played at 0 dB SPL to 80 dB SPL in 10-dB steps. A total of 500 repetitions each of both positive and negative polarities were played for each condition. ABR threshold was calculated based on a cross-correlation analysis (Henry et al., 2011). Another intermediate (odd multiple of 5 dB) step was used near the preliminary ABR threshold estimate to fine-tune the final estimate. DPOAEs were measured for pairs of tones (f1, f2) presented simultaneously with f2/f1 =1.2 at 75 (f1) and 65 (f2) dB SPL.

sABR experiments

A naturally spoken speech sentence (list #3, sentence #1) from the Danish speech intelligibility test (CLUE; Nielsen and Dau, 2009) was used for sABR experiments (Fig. 1). The sentence was: “Den gamle mand smilede stort” (IPA phonetic transcription: dɛn kamlə mæːːˀn smiːləð̠˕ˠə s). The overall level was set to 70 dB SPL for both groups by adjusting the rms of the entire utterance. Both positive and negative polarities (500 repetitions/polarity) of the stimulus were used to allow estimation of envelope and temporal fine structure components from the sABR. Both envelope (sABRENV = average of sABRs to opposite polarities) and temporal fine structure (half the difference between sABRs to opposite polarities) amplitudes were comparable between the two groups (also see Parida and Heinz, 2021). This similarity in amplitude at 70 dB SPL despite hearing loss likely results from a loss of cochlear compression because of outer-hair-cell dysfunction (Ruggero and Rich, 1991) as well as central-gain related changes in the midbrain (Auerbach et al., 2014), which substantially contribute to the brainstem responses (King et al., 2016).

Figure 1.
Figure 1.

The speech sentence used in this study. A, Time-domain signal (left y-axis) with orthographic representation and time-varying stimulus-level function (in dB SPL, right y-axis) computed over 40-ms moving windows with 50% overlap. B, Spectrogram of the sentences in A. Red lines represent the first three formant trajectories (i.e., F1, F2, F3, left y-axis); black line represents the fundamental frequency (F0) trajectory (right y-axis).

sABRENV was used to analyze sABR onset responses. Only the onset response was considered to evaluate consonant coding because evoked responses like the sABR require synchronous activity (e.g., the onset) across populations of neurons. As sustained responses to fricatives lack a clear temporal pattern, they are rather weakly represented in the sABR (Skoe and Kraus, 2010). Onset strength was quantified as the peak-to-peak sABR amplitude in an onset window (see Fig. 8E). Note that sABRs consist of both transient ABR and sustained FFR components and are primarily of subcortical origin (Chandrasekaran and Kraus, 2010; Skoe and Kraus, 2010).

AN experiments

Monaural sound was delivered via a custom closed-field acoustic system. A dynamic loudspeaker (DT–48, Beyerdynamic) was connected to a hollow ear bar inserted into the right ear canal to deliver acoustic stimuli that were calibrated near the tympanic membrane. Calibration was done at the beginning of the experiment using a probe-tube microphone (ER–7C, Etymōtic) that was placed within a few millimeters of the tympanic membrane.

Single AN fibers were isolated by advancing the electrode while playing broadband noise (20–30 dB re 20 μPa/√Hz; higher as needed for noise-exposed animals) as the search stimulus. Monopolar action potential waveform shape was used to confirm that recordings were from AN-fiber axons (as opposed to bipolar shapes exhibited by cell bodies in the cochlear nucleus). Before collecting spike-train data in response to speech and/or noise, all AN fibers were characterized as follows. An automated algorithm was used to estimate the FTC (Chintanpalli and Heinz, 2007). FTCs were smoothed by a three-point triangular window before estimating parameters such as characteristic-frequency (CF), threshold, 10-dB quality factor (Q10=CF/10-dB bandwidth), and TTR. For HI fibers, the CF was determined as the local minimum closest to the steep high-frequency-side FTC slope, which closely matches the basilar membrane frequency-place map following NIHL (Liberman, 1984; Miller et al., 1997). The threshold at CF was considered the threshold of the AN fiber. The FTC 10-dB bandwidth was estimated as the linearly interpolated bandwidth near the CF corresponding to an FTC threshold equal to 10 dB above the CF threshold. For W-shaped FTCs (i.e., when thresholds at the tail and tip were comparable), the lower cut off for the 10-dB bandwidth was set to the highest frequency below CF at which FTC threshold crossed the “threshold + 10 dB” level criteria. This approach to estimating 10-dB bandwidth (and Q10) differs from previous studies that used the lowest frequency at which FTC threshold crossed the criterion level (i.e., including both tip and tail effects), and enabled us to decouple tuning-sharpness effects at the tip from tonotopic-distortion effects (Miller et al., 1997; Kale and Heinz, 2010; Henry and Heinz, 2012; Henry et al., 2016). Therefore, this local Q10 captures narrowband aspects of frequency selectivity near CF (i.e., tuning sharpness), whereas TTR captures broadband aspects of frequency selectivity. For a minority of HI AN fibers, the FTC threshold at the lowest or highest frequency measured was below the 10-dB bandwidth criterion (i.e., CF-threshold + 10 dB); in those cases, the 10-dB bandwidth was based on the lowest or highest frequency measured, respectively. This likely led to a slightly overestimated Q10 value for a few very-broad HI AN fibers. Tonotopic-distortion effects were quantified using TTR, which was estimated as the difference (in dB) between the threshold at CF (i.e., the tip) and the lowest threshold for frequencies at least 1.5 octaves below CF (i.e., the tail). Spontaneous rate (SR) was measured over a 30-s silence period for each AN fiber.

For AN experiments, the same speech sentence was used as for the sABRs. The overall level was set to 65 dB SPL for NH chinchillas and 80 dB SPL for HI chinchillas (by normalizing the rms over the entire utterance). The spectrally flat gain of 15 dB was roughly based on the half-gain rule (Lybarger, 1978), which has been used in AN studies of NIHL (Schilling et al., 1998), and was confirmed to restore driven rates (i.e., total number of spikes within a stimulus window divided by window duration) for voiced portions of the sentence in preliminary experiments. Speech was also presented after mixing with frozen steady speech-shaped noise at three different perceptually relevant SNRs (−10, −5, and 0 dB). Frozen noise was used instead of running noise to allow for the same population-level analyses for both speech and noise. The speech-shaped noise was spectrally matched to 10 sentences spoken by the same speaker using autoregressive modeling. The two polarities of stimuli were presented in an interleaved manner (25 trials per polarity for most AN fibers). It should be noted that, in general, our sABR results and AN-fiber results do not provide an exact comparison as these data were collected at somewhat different stimulus levels. However, as shown later, results regarding selective audiometric deficits for low-intensity (e.g., consonants) but not for high-intensity (e.g., vowels) sounds still hold for both datasets despite these stimulus-level differences.

Spike-train temporal-precision analysis

The temporal precision of spike trains was measured using the Victor–Purpura (VP) distance metric (Victor and Purpura, 1996), which quantifies the dissimilarity between two sets of spike trains. If a fiber responds precisely and consistently across different stimulus trials, the VP distance between the two spike trains will be small. The VP distance between spike trains from two different stimulus trials, X (total spikes = NX) and Y (NY), is defined as the total cost involved in transforming X so that it matches Y. This transformation allows three operations: (1) addition of spikes; (2) deletion of spikes; and (3) shifting of spikes. The cost of adding or deleting a spike is set to 1 by convention. The cost of shifting a spike by Δt seconds is proportional to a time-shifting cost parameter (q, in cost/s), which controls the temporal resolution of the analysis. This is because when |Δt|≥2/q, the cost of shifting the spike is greater than that of simply deleting the spike and adding a new spike. Note that for q=0, this analysis simplifies to a rate code where the cost is simply the difference in the number of spikes |NX−NY|. As q increases, the emphasis on temporal coding becomes greater as the temporal resolution improves. Specifically, all time shifts |Δt|≥2/q, are maximally costly (cost = 2) in this analysis, which can be interpreted as saying that this analysis is most sensitive to frequencies less than q/2. Another way to think about this is that for a temporal resolution (sampling rate) of 1/q, the Nyquist rate of frequencies considered is q/2 Hz. We have used this time/frequency relation to pick the specific q values considered here (see Fig. 4) with respect to the three main temporal resolutions of importance for speech (Rosen, 1992).

For a given temporal resolution 1/q, the optimal (smallest total cost) set of operations to transform X to Y is determined based on dynamic programming; the VP distance between X and Y is computed as the total cost of all operations involved in this optimal transformation. For each pair of spike trains X and Y, the maximum possible cost (VPmax) of transforming X into Y is equal to the sum of the number of spikes in each spike train, NX+NY (corresponding to removing all spikes in X and adding all spikes in Y). The minimum value of the VP distance is VPmin=|NX–NY|, corresponding to the condition where q=0, and the only cost consists of equalizing the number of spikes. To minimize the basic dependence of VP distance on the number of spikes, we computed a normalized VP distance: VPnorm=VP−VPminVPmax−VPmin, where, by definition, VPnorm ranges from 0 to 1. Note that despite this normalization, some dependence on discharge rate remains because with more spikes in each spike train, it is more likely to find a nearby spike (i.e., the expected time shift between spikes is less when there are more spikes). As such, we compare VPnorm distance (or precision) between groups as a function of discharge rate in Figure 4.

To compute the VPnorm distance for a given AN fiber and shifting cost parameter q, VPnorm was first computed for all pairwise combinations of spike trains (across different stimulus trials) and then all these VPnorm estimates were averaged to form the final estimate. Temporal precision was then computed as ln(1/VPnorm), where the natural logarithm was used to linearize the data for statistical analyses.

Spectral analyses of alternating-polarity PSTHs

Temporal and spectral analyses of spike-train responses were based on alternating-polarity PSTH analyses (Parida et al., 2021), with a PSTH bin width of 200 µs in all cases. By alternating-polarity PSTHs, we refer to the collection of PSTHs that can be derived from PSTHs corresponding to positive and negative polarities, as defined in Parida et al. (2021). This collection includes the compound (Goblick and Pfeiffer, 1969) as well as the sum PSTH, which were used here to study TFS and onset coding, respectively. To emphasize temporal-fine-structure responses for voiced speech coding in Figures 5–7, the difference PSTH, d[n], was constructed by halving the difference between the PSTHs to opposite stimulus polarities. The difference-PSTH spectrum, D(f), was computed using multitaper (n = 3) analyses in Figure 6 (Thomson, 1982). For each AN-fiber response, near-CF fractional power (in dB re: total response power) was quantified based on the D(f) power in a spectral window centered at CF. The spectral window was fourth order, and its 3-dB bandwidth was set to the 50th percentile fit for AN-FTC 3-dB bandwidth for NH chinchillas (Kale and Heinz, 2010). The 3-dB bandwidth was estimated from the 10-dB bandwidth using the following formula, which provided a reasonable approximation for fourth order Butterworth filters: 3-dB bandwidth=10-dB bandwidth×10−7/80/2.

Low-frequency fractional power was quantified using a low-pass spectral window (cutoff = 400 Hz, 10th order).

Quantifying formant coding strength using the harmonicgram

Formant coding is traditionally quantified using the Fourier spectrum of the period histogram or the difference PSTH (Young and Sachs, 1979; Sinex and Geisler, 1983). These analyses provide sufficient spectrotemporal resolution for analyzing responses to stationary speech tokens like those used in many previous studies. To quantify power along dynamic formant trajectories in a nonstationary speech stimulus, the harmonicgram can be used, as it offers higher spectrotemporal resolution than the spectrogram (Parida et al., 2021).

Briefly, the harmonicgram was constructed as follows. The fundamental frequency contour (F0[n], where n is the discrete index of time; Fig. 1) was estimated from the sentence stimulus using Praat (Boersma, 2001). Response components (HG[k,n] for sampling frequency fs) along the kth-harmonic of F0[n] were estimated using frequency demodulation of the difference PSTH d[n] followed by low-pass filtering (LPF). The low-pass filter was a sixth order zero-phase IIR filter with 10-Hz cutoff frequency. ϕtrajk[n]=1fs∑m=1nkF0[m] ddemod[n]=d[n]e−j2πϕtrajk[n] HG[k,n]=LPF(ddemod[n]).

To evaluate the coding strength of a formant, FX (where X = 1, 2, or 3), in the response d[n], the fractional power (FracPower) of three harmonics near F0-normalized FX was calculated as: FracPower(FX)=∑n=1N∑k=α[n]−1α[n]+1(HG[k,n])2∑n=1N∑k=1K(HG[k,n])2, where α[n]=round(FX[n]F0[n]), N is the length of d[n], round is the nearest-integer operator, and K is the total number of F0 harmonics considered. Because the lowest value for F0[n] was ∼100 Hz and phase-locking for chinchillas is substantially degraded beyond 3500 Hz, K was set to 35 in these analyses. Formants were also estimated using Praat (Fig. 1).

For a given SNR condition, fractional power metrics were computed for responses to noisy speech [FracPowerSN(FX)] and noise alone [FracPowerN(FX)] for individual AN fibers. The difference in these two power metrics [i.e., FracPowerSN(FX)–FracPowerN(FX)] was used as the strength of FX coding for each fiber for that SNR condition. For the quiet condition, FracPowerN(FX) was set to 0.

Driven-rate analysis to evaluate consonant coding in quiet

Coding of a fricative (/s/) and two stop consonants (/d/ and /g/) was analyzed based on driven-rate analysis within different stimulus windows. For the fricative (see Fig. 8), two metrics were used: onset and sustained driven rates. Onset driven rate was estimated as the weighted rate over a window that had a value of 1 for the first 10 ms of the fricative and linearly dropped to 0 over the next 15 ms (i.e., total window duration = 25 ms). Sustained driven rate was estimated using a trapezoidal window that had a central duration of 30 ms and linearly dropped to 0 at the beginning and end of the fricative. These windows were based on a previous study with modifications to account for differences in fricative duration (previous study: 200 ms, current study: ∼100 ms; Delgutte and Kiang, 1984c).

Responses to two stop consonants, /d/ and /g/, were also analyzed (see Fig. 9). A stop consonant consists of a closure or “stop,” a release burst, and a voiced segment that includes formant transitions. Here, we analyzed responses only to the release burst. The driven rate was estimated over the burst period, which was ∼25 ms for both stop consonants.

Coding-fidelity metrics for fricatives in noise based on correlation and driven-rate analyses

To evaluate the fidelity of the neural representation of the fricative /s/ in noise, correlations of the slowly varying envelope responses were quantified between speech-alone and noisy-speech conditions during the fricative segment. To confirm that these correlation values were not spurious correlations between speech and noise, the speech-alone and noisy-speech correlations were corrected by subtracting the correlation between response envelopes of speech-alone and noise-alone conditions for the same fricative window.

Response envelopes were obtained from single-polarity PSTHs using a low-pass filter (fourth order, cutoff = 32 Hz). Let the response envelope to speech-alone, noisy-speech, and noise-alone be denoted by RS,RSN, and RN, respectively. Then, a speech-in-noise coding fidelity metric was quantified by computing the corrected correlation between RS and RSN as: corrcorrected(RS,RSN)=corr(RS,RSN)−corr(RS,RN), where corr(X,Y)=E(XY) and E is the expectation operator. The minimum value of corrcorrected(RS,RSN) was set to 0.001 for display purposes in Figure 10D–F.

Consonant coding in noise was also evaluated using a driven-rate analysis. Driven rate for each SNR condition was estimated within the whole 100-ms fricative window for noisy speech (DRSN) and noise alone (DRN). Then, the speech-in-noise coding fidelity was quantified using the difference in these driven rates (i.e., DRSN−DRN).

Plosive coding in noise was not considered because plosives were completely masked by the noise, even at the highest SNR (0 dB) used here. Instead, inferences about plosive coding in noise based on formant transitions (as quantified in Fig. 7) are considered in Discussion.

Statistical analyses

Statistical analyses of group effects (i.e., hearing status) were performed in R (version 4.0.3) using linear mixed-effects models (lme4 package; Bates et al., 2014) and the car package (Fox and Weisberg, 2018). Both p-value and F-value were estimated using Type II Wald F tests (Kenward and Roger, 1997). Note that these p-value and F-value correspond to partial regression coefficients of the predictors (i.e., they represent the significance of each predictor's contribution after partialling out all other predictors); this is important because these predictors (e.g., Q10 and TTR) can be correlated. Log-transformed CF and its interaction with hearing status were included in all statistical models for AN data. CF was log-transformed because of the roughly logarithmic spacing of frequency in the cochlea. Appropriate CF ranges for statistical analyses were determined based on stimulus spectral bandwidth and are described in each figure caption. Driven rate and SR were included as predictors in the temporal-precision analyses in Figure 4. Spontaneous-rate group (low and medium SR, <18 spikes/s; high SR, >18 spikes/s) was included as a binomial variable in Figure 10. Interactions were included between fixed effects, with main and interaction effects judged to be significant when p < 0.05.

The effects of various physiological mechanistic factors on speech-coding fidelity in Figure 6 (tonotopic coding), Figure 7 (formant coding in noise), and Figure 10 (fricative coding in noise) were also evaluated by including the following fixed-effect predictors: log-transformed CF, FTC threshold (in dB SPL), TTR (in dB), and local Q10. Animal identifier was treated as a random effect. The dequivalent or deq metric was used to indicate the effect size for significant and close-to-significant individual factors, i.e., deq is reported for all effects where p < 0.1 (Rosenthal and Rubin, 2003).

Results

Chinchilla model of NIHL captures reduced audibility and degraded frequency selectivity

Mild-to-moderate hearing loss is the most prevalent type of hearing loss (Goman and Lin, 2016). To investigate the neural coding deficits these patients likely experience, we used an established noise-exposure protocol to induce mild-to-moderate SNHL in a chinchilla model (Kale and Heinz, 2010). Thresholds for ABRs to tone bursts increased by ∼20 dB following NIHL (Fig. 2A, statistics in legends). Similarly, DPOAE levels decreased by ∼15 dB (Fig. 2B), indicating the presence of substantial outer-hair-cell damage. These electrophysiological changes indicate a mild-to-moderate permanent hearing-loss model (Clark, 1981).

Figure 2.
Figure 2.

Chinchilla model of NIHL captures reduced audibility, reduced frequency selectivity, and expanded AN-fiber threshold distribution. A, ABR thresholds for HI chinchillas (red) were elevated by ∼20 dB relative to NH (blue). Thin lines with symbols represent individual animals (n = 9/6, NH/HI); thick lines represent group averages (main effect of group, F = 276.4, p < 2.2 × 10−16). B, DPOAE levels were reduced by ∼15 dB (F = 943.6, p < 2 × 10−16). C, AN-fiber thresholds were elevated for the HI group (F = 646.36, p < 2 × 10−16; n = 286/119, NH/HI). Symbols represent individual AN fibers; solid and dashed lines represent 50th and 10th percentiles, respectively, computed within octave bands for which there were greater than or equal to seven fibers in the group. D, Q10 tuning-sharpness values were reduced for HI AN fibers (F = 53.87, p = 1.2 × 10−12). Thick lines represent octave-band averages. E, Shifts in 50th percentile chinchilla AN-fiber thresholds (squares) following NIHL were greater than shifts in the 10th percentile (triangles). ABR-threshold shifts are shown with plusses. Data are mean ± SEM based on bootstrapping (10 repetitions of 80% random sampling with replacement). Data are slightly shifted along the x-axis for display. F, Same format as E but showing the same threshold-distribution expansion for a more extensive noise-exposed data set from cat (n = 180/558, NH/HI), reanalyzed from Heinz et al. (2005). CF, characteristic frequency; DPOAE, distortion product otoacoustic emissions; ANF, AN fiber.

Single-fiber thresholds were elevated by ∼25–35 dB for the most-sensitive AN fibers in the population (Fig. 2C,E). This threshold shift was accompanied by substantial near-CF broadening of tuning as quantified by reductions in local Q10 values for individual-fiber FTCs (Fig. 2D). These physiological results were similar to previous results from our laboratory (Kale and Heinz, 2010; Henry et al., 2016).

NIHL expands AN-fiber threshold distribution: audiometric threshold shift underestimates average fiber-audibility deficit

Although audiometric thresholds are probably determined by the most-sensitive AN fibers in the population, suprathreshold speech perception in complex environments likely involves many fibers spanning a range of thresholds and not just the most sensitive ones (Young and Barta, 1986; Bharadwaj et al., 2014; but see Carney et al., 2016). Therefore, changes in AN-fiber threshold distribution could affect the coding of complex sounds. A traditional hypothesis, which assumes that following NIHL the high-threshold fibers function normally and the most-sensitive (low-threshold) fibers have elevated thresholds, predicts a compressed AN-threshold distribution (Moore et al., 1985; Ngan and May, 2001). Alternatively, across-fiber heterogeneity in susceptibility to acoustic trauma could lead to an expanded AN-fiber-threshold distribution following NIHL (Heinz et al., 2005), with a greater shift in the distribution mean than the shift in distribution minimum. If acoustic gain were based on the most-sensitive fibers (i.e., using the audiogram to compensate for hearing loss), then an expanded threshold distribution would lead to a selective “audibility” deficit for low-intensity stimuli (e.g., stop consonants) for many AN fibers, whereas a compressed distribution would not cause such a deficit (Fig. 3).

Figure 3.
Figure 3.

A schematic illustration shows that an expanded AN-fiber threshold distribution can lead to selective deficits in audibility for low-intensity stimuli (e.g., stop consonants). Thick lines represent threshold histograms for populations of 30,000 AN fibers under NH (blue) and HI (red) conditions. Top and bottom panels correspond to two hypothesized effects of NIHL on AN-fiber threshold distributions: compressed and expanded, respectively. These schematized data represent a 30-dB hearing loss (in all panels), as indicated by a 30-dB shift in the HI distribution minimum threshold. Triangles represent the level of a low-intensity stimulus (left) and high-intensity stimulus (right) for the NH and HI groups. For the HI group, a 30-dB gain was applied to compensate for hearing loss. Shaded regions represent the subpopulation in each group that is driven by the stimulus (i.e., “audible”). While there was no “audibility” deficit for the group with a compressed distribution for either stimulus, a substantial portion of the HI group with an expanded distribution was not driven by the low-intensity stimulus (unshaded region).

To evaluate these hypotheses, we estimated the 10th and 50th percentiles for our NH and HI threshold distributions in octave-wide CF bands (Fig. 2E). Our results showed a greater shift for the 50th percentiles than for the 10th percentiles following NIHL for all bands. We reanalyzed a more extensive previously published data set from cats (Heinz et al., 2005), which also showed an expansion (not compression) in AN-fiber threshold distribution following NIHL (Fig. 2F). These consistent results suggest that any audiometric indicator (i.e., based on the most-sensitive AN fibers) will underestimate audibility deficits for many AN fibers across the population.

Temporal-coding precision for connected speech in quiet was not degraded by NIHL

To test whether there was any degradation in the ability of AN fibers to precisely encode temporal information in response to connected speech, trial-to-trial precision was quantified using the VP distance metric (Victor and Purpura, 1996). Temporal precision is inversely related to VP distance and was computed for a range of temporal resolutions using the time-shift cost parameter (q) of the VP analysis to span the syllabic, voice-pitch, and formant time scales of speech (Rosen, 1992). The temporal precision of connected speech responses was not degraded following NIHL for any of the temporal resolutions considered (Fig. 4). In fact, there was a small but significant increase in precision for all three time-scale conditions. This increase in precision may arise because of overrepresentation of lower-frequency information associated with distorted tonotopy, where synchrony is stronger than at higher frequencies. Overall, these data provide no evidence for a degradation in the fundamental ability of AN fibers to precisely phase-lock to the temporal features in connected speech following NIHL.

Figure 4.
Figure 4.

Temporal-coding precision for connected speech in quiet was not degraded by NIHL. A, Across-trial precision is plotted versus discharge rate in response to a connected speech sentence in quiet at conversational levels. A time-shift cost (q) corresponding to 250-ms time scale was used to represent syllabic rate. Symbols represent individual AN fibers across all CFs. B, 10-ms time scale to emphasize voice-pitch coding. C, 0.67-ms time scale to emphasize speech-formant coding. A small but significant increase in temporal precision is observed for HI responses (group, F = 30.8, p = 5.4 × 10−8), which likely derives from increased responses to lower-frequency energy because of distorted tonotopy (CF × group interaction, F = 5.0, p = 0.025).

Changes in AN-fiber tuning following NIHL distort the normal tonotopic representation of naturally spoken vowels

Normal hearing (NH) AN fibers are characterized by high sensitivity (low threshold) and spectral specificity (sharp tuning), which allows for selective responses to stimulus energy near their CF. These NH properties produce tonotopic responses to complex sounds, as demonstrated previously for synthetic steady-state vowels (Young and Sachs, 1979; Miller et al., 1997). Here, we evaluated the effects of NIHL on natural-vowel responses (Figs. 5, 6) by examining spectral estimates [D(f)] computed from difference PSTHs (which minimize rectifier-distortion artifacts; Parida et al., 2021). The exemplar NH AN-fiber response shown in Figure 5 was dominated by the second formant F2, the spectral feature closest to fiber CF for the segment considered (Fig. 5B). This effect can be seen in the NH D(f) (Fig. 5C), which shows a clear peak near F2 (blue arrow), representative of a tonotopic response. In contrast, the exemplar HI fiber with similar CF had a D(f) with a peak near F1 (∼650 Hz), which is well below its CF (∼1.5 kHz; Fig. 5B,C). This HI nontonotopic response could arise from some combination of its elevated tip threshold, broadened tuning, and hypersensitive tail, which are typical for AN fibers following NIHL (Liberman, 1984; Liberman and Dodds, 1984a). These physiological changes lead to a reduced TTR in the FTC (Fig. 5B, double arrows), which is a consistent trend in our HI population (Fig. 5D).

Figure 5.
Figure 5.

Changes in AN-fiber tuning following NIHL distort the normal tonotopic representation of naturally spoken vowels. A, Alternating-polarity PSTHs in response to a “quasi-stationary” segment from a natural sentence in quiet (stimulus: dark gray). Darker (lighter) color shades represent PSTHs for positive (negative, reflected across x-axis for display) stimulus polarity. Stimulus scaled arbitrarily. B, Spectrum of the stimulus segment (gray, left y-axis) with fundamental frequency (F0) and first and second formants (F1 and F2) labeled. Tuning curves (right y-axis) from the two example AN fibers in A, C. Stars indicate CF (where the HI CF was chosen based on convention from anatomic labeling studies; Liberman and Dodds, 1984a); double-arrows indicate TTR in dB. Both AN fibers have CF near F2. C, Spectra of the difference PSTHs in A [D(f), representing NH and HI TFS responses]. NH D(f) shows a tonotopic response (i.e., near CF), with a peak near F2 (blue arrow); HI D(f) has a peak near F1 (red arrow), despite having CF near F2 (i.e., nontonotopic response). D, TTR distributions for NH and HI fibers. TTR was consistently (group × CF interaction, F = 2.9, p = 0.09) and significantly (group, F = 59.1, p = 3.4 × 10−6) reduced following NIHL. PSD, power spectral density; TC, tuning curve.

Figure 6.
Figure 6.

Distorted tonotopy enhances below-CF coding of voiced speech at the expense of near-CF representations, even when audibility is restored. A, Driven rates (DRs; solid trend lines) for NH and HI AN fibers were comparable in voiced portions of the connected speech sentence in quiet (group, F = 0.6, p = 0.44). SRs (dashed lines) were reduced for HI (F = 44.1, p = 1.5 × 10−10). Triangular-weighted trend lines (here and in the following figures) were computed for 2/3-octave averages. B, Fractional response power near CF [in dB re: total power based on the difference-PSTH spectrum, D(f)] was significantly reduced for HI primarily at lower (e.g., <3 kHz) CFs (group, F = 84.14, p < 2 × 10−16; group × CF interaction, F = 13.8, p = 2.4 × 10−4). C, Fractional response power in a LF band (<400 Hz) was enhanced for HI (group, F = 10.5, p = 1.3 × 10−3; group × CF interaction, F = 23.0, p = 2.5 × 10−6). D, Ratio of power near CF to power in the low-frequency band was significantly reduced for HI fibers (group, F = 36.86, p = 3.9 × 10−9; group × CF interaction, F = 26.7, p = 4.5 × 10−7). CF range: 0.6–5 kHz in all panels.

Distorted tonotopy enhances below-CF coding of voiced speech at the expense of near-CF representations, even when audibility is restored

To quantify the effects of distorted tonotopy on voiced-speech coding at the population level, fractional power near-CF and at low frequency (LF) was quantified as metrics related to tonotopic coding (Fig. 6). Fractional power was used to minimize the effects of overall power and rate differences, if any. Despite the expected reduction in SR (Liberman and Dodds, 1984b), driven rates during voiced segments were similar for the two groups likely because most AN fibers in both groups were saturated at the stimulus levels used (Fig. 6A). Therefore, there was no deficit in the total spike counts (i.e., driven rates) in response to voiced speech in our data.

Near-CF power was significantly lower for HI fibers, particularly at lower (<3 kHz) CFs (Fig. 6B). To quantify the susceptibility of fibers with higher CFs (>0.6 kHz) to very LF stimulus energy, the power in D(f) was computed in a low-pass spectral window (400-Hz 3-dB cutoff). As expected, LF power was significantly higher for HI fibers than for NH fibers (Fig. 6C). The ratio of power near CF to power at LF, which indicates the strength of tonotopic coding (relative to LF coding), was >0 dB on average for NH fibers with CF ≤3 kHz (Fig. 6D). In contrast, this relative-power metric was <0 dB on average for HI fibers in the same CF region. Overall, these results demonstrate a disruption of near-CF stimulus energy coding at the expense of very LF stimulus energy coding at a single AN-fiber level. These deficits are more severe than previously thought (Henry et al., 2016), likely because of the steep negatively sloping spectrum of connected speech.

What physiological factors contribute to this distortion in tonotopic coding? To address this question, a mechanistic linear mixed-effects model was constructed with relative near-CF-to-LF power (Fig. 6D) as the response variable, and threshold, local Q10, and TTR included as predictors (in addition to the significant dependence on CF, F = 11.7, p = 7.8 × 10−4). Whereas traditionally hypothesized factors underlying speech-coding deficits include elevated threshold and broader bandwidth, the model suggested that threshold was a significant factor (deq= 0.35, p = 0.01), but Q10 was not (p = 0.46). Interestingly, TTR was the major factor underlying this distorted tonotopic coding (deq= 0.42, p = 3 × 10−3).

Temporal-place formant representation was more susceptible to background noise following NIHL

The effects of distorted tonotopy on formant coding were evaluated for connected speech in quiet and in noise using harmonicgram analyses (Parida et al., 2021). Coding strength for the first three formants was quantified as the D(f)-based power estimate along the dynamic formant trajectories at different SNRs (Fig. 7). For NH speech-in-quiet responses, fractional power for individual formant trajectories peaked at CFs near or slightly above the formants' mean frequencies (Fig. 7A,D,G), a result consistent with previous reports of tonotopic formant coding (Young and Sachs, 1979; Delgutte and Kiang, 1984a). As expected, F1 representation was expanded for the HI group extending to CFs substantially above F1 (i.e., CFs in the F2 and F3 ranges). Coding of F2 and F3 was diminished for the HI group, and the corresponding peaks were shifted to higher CFs than expected based on the formant frequencies. This distorted tonotopic coding for connected speech is consistent with system characterization studies that report a lowering of best frequency following NIHL (Liberman, 1984; Henry et al., 2016), and with previous reports using synthesized vowels (Miller et al., 1997).

Figure 7.
Figure 7.

Temporal-place formant representation was more susceptible to background noise following NIHL. A–C, Temporal-coding strength for the first formant (F1) is shown for NH (blue) and HI (red) AN-fiber responses to the connected speech sentence in quiet (A) and in noise at SNRs of 0 and −5 dB (B, C). Symbols represent harmonicgram-based fractional power estimates (left y-axis) along the corresponding formant trajectory (thin black lines, time indicated on right y-axis) for the speech response relative to the fractional power for the noise-alone response for single AN fibers. F1 coding was significantly enhanced for HI fibers (F = 23.7, p = 1.5 × 10−6; CF range: 0.5–5 kHz). D–F, Same as A–C but for F2. For the NH population, trend lines peak near formants, as expected for tonotopic coding. In contrast, HI trend lines peaked at frequencies well above F2. Compared with NH, F2 coding was degraded for HI (group, F = 21.1, p = 8.0 × 10−6; CF range: 1–2.5 kHz). G–I, Same layout for F3 coding, which was significantly degraded for HI (F = 8.1, p = 0.007; CF range: 2–3 kHz).

Noise had a strong detrimental effect on the already degraded (reduced strength and nontonotopic) representations of F2 and F3 for the HI population. At 0 dB SNR, F2 and F3 peaks were still strong and tonotopic for the NH group (Fig. 7E,H). For HI fibers, the F2 peak was still discernible and centered at a CF location much higher than F2, but F3 coding was almost nonexistent. Similarly, at −5 dB SNR, the NH group still showed a discernible and tonotopic peak for F2 coding, whereas F2 coding for the HI pool was severely degraded.

To identify the contribution of various physiological factors to these formant-coding degradations, mixed-effects models were constructed for each formant fractional power response with TTR, Q10, and threshold as fixed-effect predictors, in addition to log-CF and SNR. For F1 enhancement in the HI representation, TTR (deq= 0.21, p = 0.034) was the major contributing factor and not threshold (deq= 0.19, p = 0.051) or Q10 (p = 0.26). Similarly, for the degraded F2 response, TTR (deq= 0.78, p = 2.1 × 10−6), but not threshold (deq= 0.33, p = 0.053) or Q10 (p = 0.17), was the major factor. For F3, TTR (deq= 0.80, p = 0.022) was the only contributing factor to the degraded response (threshold: p = 0.20; Q10: p = 0.71).

Overall, these results show a degradation of higher-formant (F2 and F3) coding, which is known to be important for the perception of vowels as well as consonants, in HI-fiber responses following acoustic insult. Moreover, these degradations are more susceptible to the presence of noise, which is consistent with the stronger perceptual deficits that listeners with SNHL experience in noisy environments (Horst, 1987).

Unlike voiced segments, driven rates for the fricative /s/ were not restored in HI fibers despite compensating for overall audibility loss

Fricatives constitute a substantial portion of phoneme confusions among listeners with SNHL (Bilger and Wang, 1976; Van de Grift Turek et al., 1980; Dubno et al., 1982). To explore potential neural bases of these deficits, neural responses to a fricative (/s/) were analyzed. Previous studies have reported robust fricative coding by NH AN fibers in terms of onset and sustained responses (Delgutte and Kiang, 1984c). NH fibers with higher CFs (i.e., near frequencies where /s/ has strong energy; Fig. 8A) showed a sharp onset response, followed by a robust sustained response (Fig. 8B). In contrast, HI fibers showed a substantial reduction in onset response (Fig. 8B), with less of an effect on the sustained response. This reduction in onset response contrasts with previous studies that reported increased onset responses to tones following NIHL; however, those results were for equal sensation level (Scheidt et al., 2010). In the high-frequency region (3–8 kHz) where the fricative had significant energy, driven onset rates were significantly lower for the HI population (Fig. 8C), whereas sustained rates were slightly, but significantly, reduced for the HI group (Fig. 8D). While both onset and sustained neural representations for both groups resembled the fricative spectrum, reduced rates for the HI group may lead to weaker salience particularly because saturation rates (data not shown) and driven rates for voiced speech were similar for the two groups.

Figure 8.
Figure 8.

Unlike voiced segments, driven rates for the fricative /s/ were not restored in HI fibers despite compensating for overall audibility loss. A, Spectrum of /s/ (gray) from the connected sentence and exemplar tuning curves. B, Time-domain waveforms of /s/ (gray) and alternating-polarity PSTHs of AN fibers shown in A. Same format and data-analysis parameters as Figure 5A. Cyan and magenta temporal windows denote the masks used to compute onset and sustained response rates. While these example AN fibers had comparable sustained rates, the HI onset response was substantially degraded. C, Driven onset rates for the HI population were significantly reduced compared with NH for all CFs. Same format as Figure 6A (group, F = 90.5, p < 2 × 10−16; group × CF interaction, F = 10−3, p = 0.98). D, Driven sustained rates were also reduced for the HI group for all CFs, including in the high-CF (>2.5 kHz) region where /s/ had substantial energy (F = 4.4, p = 0.037). SRs were subtracted from driven onset and sustained rates in panels C, D for display. E, Exemplar sABR data (sABRENV) from a NH and a HI chinchilla demonstrate a reduced onset response for /s/ following NIHL. Cyan window (20-ms) indicates the onset window in which peak-to-peak amplitude was computed. F, Distributions of peak-to-peak onset amplitudes show a significant reduction in sABR onset response for HI chinchillas (F = 5.8, p = 0.039). Asterisk denotes p<0.05.

To evaluate the noninvasive correlates of these single-fiber degradations in consonant coding, sABRs were recorded in response to the speech stimulus from the same cohort of animals. Representative sABRs from two animals (one from each group) replicate these fricative onset-response degradations (∼0.75 s in Fig. 8E). While the NH sABR had a sharp onset, the HI sABR lacked any clear onset response. In contrast, responses during voiced speech were comparable for the NH and HI examples (e.g., ∼0.6 s), mirroring the AN result that driven rates to voiced speech were similar for the two groups (Fig. 6A). Thus, despite comparable responses to voiced speech, there was a significant reduction in fricative onset response in the HI group (Fig. 8F).

Driven rates for stop consonants (/d/ and /g/) were also not restored despite overall audibility compensation

Stop consonants are among the most confused phonemes for listeners with SNHL (Bilger and Wang, 1976; Van de Grift Turek et al., 1980). The neural representations of two stop consonants (/d/ and /g/) present in the speech stimulus (Fig. 9A) were also evaluated using driven-rate analysis during the release burst. In response to /d/ and /g/, the NH AN fiber showed a strong onset response, followed by sustained activity that was well-above spontaneous activity (Fig. 9B,C). In contrast, for the HI AN fiber, driven activity was substantially reduced. Population results showed that driven rates during the burst portion for /d/ and /g/ were significantly reduced for the HI population relative to the NH population (Fig. 9D,E).

Figure 9.
Figure 9.

Driven rates for low-intensity stop consonants (/d/ and /g/) were also not restored despite overall audibility compensation. A, Spectra for stop consonants /d/ (dark gray) and /g/ (light gray) from the connected sentence in quiet. FTCs of representative NH and HI AN fibers. B, Alternating-polarity PSTHs for the two fibers in A in response to /d/. Same format and analysis parameters as Figure 8B. Driven burst rate was reduced for the HI fiber. Burst window (green) was 25 ms. C, Same layout but for /g/, which shows the same general effects as /d/. Burst window (green) was 25 ms. D, Driven rates in response to /d/ burst were significantly reduced in the HI population compared with the NH population (F = 64.6, p = 1.3 × 10−14). Same format as Figure 8C. E, Same layout and results as D but for /g/ (F = 34.5, p = 1.8 × 10−9). SRs were subtracted from driven onset and sustained rates in panels D, E for display. F, Difference in the power spectral density of /d/ and /g/. Dashed gray = zero line. G, Difference in the driven rates during the release burst of /d/ and /g/ (i.e., rates in D minus rates in E) for both groups. Dashed gray = zero line. H, Distributions of sABR peak-to-peak onset amplitudes in response to /d/ for chinchillas in both groups show a significant reduction for the HI group (F = 5.97, p = 0.037). I, Same layout as H but for /g/. The observed reduction in sABR onset amplitudes was not significant (F = 0.58, p = 0.47). CF range for statistical analyses: 0.5–8 kHz for /d/ and 0.5–3 kHz for /g/. Asterisk denotes p<0.05. ns, not significant.

A reduction in driven rates by itself does not necessarily mean poorer coding of the stimulus spectrum for the two phonemes. Therefore, we investigated how different the rate-CF profiles for /d/ and /g/ were for each group, and whether this difference followed the spectral properties of the stimuli. We estimated the difference in rates in response to /d/ and /g/ for individual AN fibers (Fig. 9G). While the difference profile for the NH group closely matched the difference in the power spectra of the two stimuli (Fig. 9F), the difference profile for the HI group was relatively less salient (i.e., closer to the zero line). Furthermore, the steeply changing spectral edge in the spectral difference (near 2.5 kHz in Fig. 9F) was well represented in the NH profile at the appropriate CF region. In contrast, it was shifted to higher CFs in the HI representation, consistent with distorted tonotopy effects.

Next, we evaluated the representation of the two stop consonants in sABRs. The sABR onset was significantly reduced for/d/following NIHL (Fig. 9H), consistent with the universal reduction in burst rate across the whole CF range for HI fibers (Fig. 9D). The sABR onset response to /g/, however, was only slightly reduced (not significant) for the HI group (Fig. 9I). Overall, these sABR data (and those from Fig. 8) suggest that degraded consonant representations persist even when vowel audibility is restored, despite central-gain related changes that can occur in the midbrain (Auerbach et al., 2014).

Changes in tuning following NIHL eliminate the noise-resilient benefits of AN fibers with lower SR for fricative coding

As previously described, listeners with SNHL often struggle in noisy environments in identifying consonants, more so than for vowels. Here, we investigated the effect of background noise on the coding of the fricative /s/, which elicited robust sustained activity for both groups in the quiet condition (Fig. 8D). When the sentence is mixed with noise at a particular SNR, even negative SNRs, the resultant signal often has specific spectrotemporal regions with a favorable SNR (e.g., high-frequency region for /s/; Fig. 10A). These high-SNR regions likely mediate robust speech perception in noise (Cooke, 2006). In our data, NH AN fibers that were narrowly tuned near this high-SNR region responded selectively to the fricative energy (Fig. 10B). In contrast, HI AN fibers showed reduced TTR (Fig. 10A, and 5D). As a result, HI fibers tuned to higher frequencies responded poorly to fricative energy and strongly to LF energy in either the speech (e.g., voiced segments) or noise (Fig. 10C).

Figure 10.
Figure 10.

Changes in tuning following NIHL eliminate the noise-resilient benefits of AN fibers with lower SR for fricative coding. A, Spectra for /s/ from the connected-speech sentence and the concurrent speech-shaped noise (N) segment for 0 dB overall SNR. Although overall SNR was 0 dB, a local high-SNR (10–15 dB) region occurs at high frequencies (>3 kHz). FTCs of two example AN fibers that schematize the increased deleterious effect of speech-shaped background noise following NIHL, especially compared with pink noise (pink line). B, PSTHs (gray) in response to speech-alone (S), noisy-speech (SN), and noise-alone for the NH fiber (SR = 0.2/s) in A. Thick black curves represent response envelopes. Dashed magenta lines indicate temporal region in the stimulus (green) containing /s/. C, Same layout as B but for the HI fiber (SR = 1.1/s). D–F, Speech-in-noise coding fidelity for /s/ at perceptually important SNRs, as quantified by the corrected correlation between responses to S and SN (minimum value set to 0.001 for display). Squares and asterisks correspond to AN fibers with low/medium SR (<18/s) and high SR (>18/s), respectively. For NH, AN fibers with low/medium SR show better coding in noise than high-SR fibers; however, the opposite is true following NIHL because the noise resilience of low/medium SR fibers was lost resulting in overall degraded fricative fidelity (SR, F = 16.5, p = 6.2 × 10−5; group, F = 3.3, p = 0.07; group × SR interaction, F = 18.8, p = 2.0 × 10−5).

Fricative-coding fidelity was quantified by the corrcorrected metric, defined as the difference between the response-envelope correlations for speech-alone and noisy-speech, and speech-alone and noise-alone. For the NH group, responses of AN fibers with low/medium SR, hypothesized to drive robust human speech perception in noise (Costalupes et al., 1984; Silkes and Geisler, 1991; Bharadwaj et al., 2014), were particularly resistant to noise compared with high-SR fibers (Fig. 10D,E). In contrast, coding fidelity for low/medium-SR fibers was significantly lower than for high-SR fibers following NIHL.

When fricative coding in noise was evaluated using the difference in driven rates for noisy-speech and noise-alone as the metric, similar results were observed (data not shown; SR, F = 19.4, p = 1.5 × 10−5; group, F = 4.6, p = 0.03; group × SR interaction, F = 11.9, p = 6.5 × 10−4). Note that these deficits in rate-place coding for the HI group do not simply reflect reduced rates in response to noisy-speech and noise-alone because these driven rates were actually slightly greater for the HI group (data not shown; noisy speech: group, F = 9.7, p = 0.002; SR, F = 69.5, p = 2.8 × 10−15; noise alone: group, F = 15.5, p = 10−4; SR, F = 91.4, p < 2.2 × 10−16), despite reduced SR (Fig. 6A) and reduced rate responses to the fricative alone (Fig. 8C,D). Rather, these deficits are consistent with increased masking by noise for HI high-CF AN fibers (including low SR fibers) because of distorted tonotopy.

To identify the contribution of various physiological factors to fricative coding in noise, a mixed-effects model was constructed with corrcorrected as the response variable and with TTR, Q10, and threshold as fixed-effect predictors, in addition to log-CF and SNR. Results indicated that TTR (deq= 0.34, p = 0.028) had a greater effect than Q10 (deq= 0.29, p = 0.053), whereas threshold did not contribute significantly (p = 0.54).

Discussion

A common reaction from patients with SNHL is “I can hear you, but I cannot understand you,” especially in background noise, for both hearing-aid users and nonusers (Lesica, 2018). Several neurophysiological mechanisms are hypothesized to underlie these suprathreshold deficits. Here, we describe the first data, to our knowledge, characterizing SNHL effects on AN-fiber responses to a connected speech sentence in noise. These data elucidate physiological mechanisms that do and do not contribute to deficits in the peripheral coding of speech following SNHL. In particular, these data highlight the prominent role distorted tonotopy plays in degrading the coding and increasing the noise susceptibility of vowels and consonants.

Commonly hypothesized suprathreshold deficits were not the major factors degrading neural coding of connected speech following NIHL

Two suprathreshold deficits commonly speculated to affect speech perception in noise are reduced sharpness of cochlear frequency tuning (local Q10) and diminished temporal-coding precision. Our data suggest that local Q10 was not a significant factor in degraded voiced-speech coding in quiet (Fig. 6), consistent with perceptual studies suggesting a minor influence of degraded frequency selectivity on degraded speech-perception in quiet (Festen and Plomp, 1983; Dubno and Dirks, 1989). Importantly, we found that TTR, a global measure of frequency selectivity, and not local Q10, was the dominant factor in explaining the increased noise susceptibility of formant (Fig. 7) and fricative coding (Fig. 10).

We found no degradation in temporal-coding precision of AN-fiber responses to connected speech (Fig. 4), consistent with the majority of neurophysiological studies that report intact phase-locking to pure tones in quiet following SNHL (Harrison and Evans, 1979; Miller et al., 1997). Thus, these connected-speech-coding data, along with most previous studies using laboratory stimuli, provide no evidence for a reduction in the fundamental ability of AN fibers to encode rapid acoustic fluctuations.

That being said, changes in phase-locking strength to complex sounds can be observed following SNHL [e.g., decreased balance of fine-structure-to-envelope coding resulting from enhanced envelope coding (Kale and Heinz, 2010), or decreased tonal phase-locking in noise because of broadened tuning (Henry and Heinz, 2012)]. Although these changes in the temporal-coding strength of complex signals may be perceptually relevant, these temporal-coding degradations occur because of factors other than a reduction in the fundamental ability of AN fibers to follow rapid fluctuations. It is also possible that the effects of synaptopathy and/or inner hair cell damage, i.e., fewer neurons coding temporal information (Kujawa and Liberman, 2009), could lead to poorer temporal information at the perceptual level that would not be apparent in peripheral responses.

Distorted tonotopy was the major factor in degraded coding and increased noise susceptibility for both vowels and consonants following NIHL

The primary factor affecting degraded coding of connected speech in noise was the loss of tonotopy, a hallmark of normal cochlear processing. For voiced-speech segments, where AN-fiber driven rates were comparable for the two groups, the spectral content in the responses differed substantially. TTR was the dominant physiological factor explaining the increased representation of LF energy at the expense of near-CF energy for voiced speech (Fig. 6). Our metric for dynamic formant coding required a technical advance in neural analyses using frequency demodulation applied to alternating-polarity PSTHs (Parida et al., 2021), because standard Fourier analysis blurs the coding of dynamic formants in speech. These single-fiber results are also consistent with our recent sABRs to connected speech, which showed overrepresented LF energy in HI sABRs (Parida and Heinz, 2021).

Another perceptually relevant effect of distorted tonotopy was the increased susceptibility to background noise following NIHL, which was seen for both voiced-speech (Fig. 7) and fricative coding (Fig. 10). Our data on fricative coding in speech-shaped noise provide insight into the real-world significance of this effect. The fricative /s/, which has primarily high-frequency energy, normally has a better-than-average SNR because of the steep spectral decay of speech-shaped noise and the tonotopic coding in the NH cochlea. Without this normal tonotopicity following SNHL, the SNR in AN fibers with CFs within the spectral band of the fricative is greatly reduced because of the substantial LF noise now driving the neural response. Based on our previous sABR results illustrating the dependence of distorted tonotopy degree on spectral shape (Parida and Heinz, 2021), this deleterious effect on fricative coding is expected to be greater for speech-shaped noise than for white or even pink background noise. While many environmental sounds have a pink spectrum (3 dB/octave drop in spectrum level), long-term speech spectrum has a downward spectral slope that is two to three times steeper (i.e., 6 to 9 dB/octave drop) than for pink noise (Fig. 10A; French and Steinberg, 1947; Byrne et al., 1994). Thus, these results have important implications for listening amid multiple talkers, which is a condition of great difficulty for many listeners with SNHL (Festen and Plomp, 1990; Le Prell and Clavier, 2017).

It should be noted that the relatively flat hearing-loss pattern observed here is not typical of the clinical population, which usually has a high-frequency sloping component in addition to a flat hearing loss (Dubno et al., 2013; Parthasarathy et al., 2020). Therefore, these results may underestimate real-world deficits. On the other hand, the head-related transfer function for humans can boost the level at higher frequencies (>1 kHz) relative to lower frequencies, and may therefore partially offset the effects seen here (Moore et al., 2008). Finally, our data support hearing-aid algorithms that prescribe more gain at high frequencies because they counteract the effects of distorted tonotopy (Moore et al., 2010).

Noise-resistant ability of low/medium SR fibers for fricative coding was substantially degraded following NIHL

The statistical analyses of our fricative-in-noise data (Fig. 10) showed a significant effect of SR, and an interaction between group and SR. The main SR effect was not surprising given previous NH studies showing the superior coding-in-noise of low-SR fibers (Costalupes et al., 1984; Young and Barta, 1986), including for consonant-vowel tokens (Silkes and Geisler, 1991). The new insight provided by our data relates to the interaction between group and SR, which arises because NIHL caused a larger deficit in speech-coding fidelity in the low/medium-SR fibers than in the normally poorer high-SR fibers. Note that raising the fricative level may partly mitigate the deficits seen for low-SR fibers; however, it would not help with two key factors: (1) tuning sharpness, which is important for noise-resistant representation; and (2) increased forward masking by preceding low-frequency energy in both speech and noise (Glasberg et al., 1987; Fogerty et al., 2017). Overall, these data demonstrate that the normally robust low/medium-SR fibers can be preferentially degraded by hearing loss, similar to recent studies on cochlear synaptopathy with age and moderate noise exposure where low/medium-SR fibers are preferentially lost (Fernandez et al., 2015, 2020; Wu et al., 2019).

Audibility restoration is not the same for consonants and vowels because of expanded AN-fiber threshold distribution following NIHL

Psychoacoustic studies suggest that degraded fricative perception for listeners with SNHL depends on both a reduced ability to use formant transitions and reduced audibility of the frication cue (Zeng and Turner, 1990). As discussed for vowels, dynamic representations of transitions for higher formants (F2 and F3) were substantially degraded with HI, especially in noise (Fig. 7). Formant transitions are also important for speech-in-noise perception of low-intensity stop consonants that are easily masked by noise; thus, the increased noise susceptibility of these transitions likely contributes perceptually.

The “audibility” of consonants, as indexed by driven rates (Figs. 8, 9), was often not restored, in contrast to vowels (Fig. 6A). Such divergent audiometric effects are consistent with psychoacoustic studies reporting similar audibility differences for consonants and vowels even after compensating for overall audibility (Phatak and Grant, 2014), and likely contribute to consonant confusions, which are common for both fricatives and stop consonants (Bilger and Wang, 1976).

Our data suggest that the expanded AN-fiber threshold distribution following NIHL (Fig. 2E,F) may contribute to differences in the audibility of consonants and vowels for amplified speech. Although across-animal variability in susceptibility to noise exposure could have partly contributed to this expansion, similarly expanded threshold distributions are observed following NIHL in individual cats (Heinz et al., 2005; their Fig. 9B). Thus, compensating for audiometric loss only guarantees restoration of audibility for the most sensitive AN fibers. Because elevation in the 10th percentile thresholds underestimated elevation in the 50th percentile, lower-amplitude consonants activated a smaller percentage of the population following NIHL, and thus were not as “audible” despite restoration of vowel audibility. It should be noted that the reduction in driven rates for consonants likely depends on the choice of sound levels used here, and these neural representations can be improved by using louder levels. For example, near-normal AN-fiber onset rates are seen following NIHL for tones at equal sensation levels (Scheidt et al., 2010). These results are consistent with psychoacoustic studies that report improved speech recognition by listeners with HI when consonants are selectively amplified (Kennedy et al., 1998; Shobha et al., 2008) and highlight the importance of compressive gain control in hearing aids (Jenstad et al., 2000; Souza, 2016).

Variations in the degree of distorted tonotopy across etiologies may contribute to individual differences in real-world speech perception

There is general consensus among researchers regarding the inadequacies of the audiogram in accounting for real-life perceptual deficits. Individual variability in speech perception, even among listeners with similar audiograms, likely stems from variations in suprathreshold deficits. Here, we demonstrate that distorted tonotopy is a significant factor in degraded coding and increased noise susceptibility of connected speech following NIHL. Because the degree of distorted tonotopy appears to vary across different etiologies (e.g., NIHL, age-related) even for similar hearing-loss degrees (Henry et al., 2019), it is likely that this variation contributes to individual differences in speech perception. The development of noninvasive diagnostics to identify distorted tonotopy (e.g., Parida and Heinz, 2021) is critical for determining the extent and perceptual relevance of this important physiological mechanism affecting the neural coding of speech.

Footnotes

  • This work was supported by the International Project Grant G72 from Action on Hearing Loss (United Kingdom) and by the National Institutes of Health Grant R01-DC009838. We thank Kenneth Henry and Hari Bharadwaj for their valuable feedback on an earlier version of this manuscript, Josh Alexander for stimulating discussions on human speech perception, and Suyash Joshi for help with phonetic transcription of the speech sentence.

  • The authors declare no competing financial interests.

  • Correspondence should be addressed to Michael G. Heinz at mheinz{at}purdue.edu

References

  1. ↵
    2. Auerbach BD,
    3. Rodrigues PV,
    4. Salvi RJ
    (2014) Central gain control in tinnitus and hyperacusis. Front Neurol 5:206. doi:10.3389/fneur.2014.00206 pmid:25386157
    OpenUrlCrossRefPubMed
  2. ↵
    2. Bates D,
    3. Mächler M,
    4. Bolker B,
    5. Walker S
    (2014) Fitting linear mixed-effects models using lme4. arXiv 1406.5823.
  3. ↵
    2. Bharadwaj HM,
    3. Verhulst S,
    4. Shaheen L,
    5. Liberman MC,
    6. Shinn-Cunningham BG
    (2014) Cochlear neuropathy and the coding of supra-threshold sound. Front Syst Neurosci 8:26. doi:10.3389/fnsys.2014.00026 pmid:24600357
    OpenUrlCrossRefPubMed
  4. ↵
    2. Bilger RC,
    3. Wang MD
    (1976) Consonant confusions in patients with sensorineural hearing loss. J Speech Hear Res 19:718–748. doi:10.1044/jshr.1904.718 pmid:1003952
    OpenUrlCrossRefPubMed
  5. ↵
    2. Boersma P
    (2001) Praat, a system for doing phonetics by computer. Glot Int 5:341–345.
    OpenUrl
  6. ↵
    2. Byrne D,
    3. Dillon H,
    4. Tran K,
    5. Arlinger S,
    6. Wilbraham K,
    7. Cox R,
    8. Hagerman B,
    9. Hetu R,
    10. Kei J,
    11. Lui C,
    12. Kiessling J,
    13. Kotby MN,
    14. Nasser NHA,
    15. El Kholy WAH,
    16. Nakanishi Y,
    17. Oyer H,
    18. Powell R,
    19. Stephens D,
    20. Meredith R,
    21. Sirimanna T, et al
    . (1994) An international comparison of long-term average speech spectra. J Acoust Soc Am 96:2108–2120. doi:10.1121/1.410152
    OpenUrlCrossRef
  7. ↵
    2. Carney LH,
    3. Kim DO,
    4. Kuwada S
    (2016) Speech coding in the midbrain: effects of sensorineural hearing loss. In: Physiology, psychoacoustics and cognition in normal and impaired hearing, pp 427–435. Cham: Springer.
  8. ↵
    2. Chandrasekaran B,
    3. Kraus N
    (2010) The scalp-recorded brainstem response to speech: neural origins and plasticity. Psychophysiology 47:236–246. doi:10.1111/j.1469-8986.2009.00928.x pmid:19824950
    OpenUrlCrossRefPubMed
  9. ↵
    2. Chintanpalli A,
    3. Heinz MG
    (2007) Effect of auditory-nerve response variability on estimates of tuning curves. J Acoust Soc Am 122:EL203–EL209. doi:10.1121/1.2794880 pmid:18247642
    OpenUrlCrossRefPubMed
  10. ↵
    2. Chung K
    (2004) Challenges and recent developments in hearing aids: part I. Speech understanding in noise, microphone technologies and noise reduction algorithms. Trends Amplif 8:83–124. doi:10.1177/108471380400800302 pmid:15678225
    OpenUrlCrossRefPubMed
  11. ↵
    2. Clark JG
    (1981) Uses and abuses of hearing loss classification. Asha 23:493–500. pmid:7052898
    OpenUrlPubMed
  12. ↵
    2. Cooke M
    (2006) A glimpsing model of speech perception in noise. J Acoust Soc Am 119:1562–1573. doi:10.1121/1.2166600 pmid:16583901
    OpenUrlCrossRefPubMed
  13. ↵
    2. Costalupes JA,
    3. Young ED,
    4. Gibson DJ
    (1984) Effects of continuous noise backgrounds on rate response of auditory nerve fibers in cat. J Neurophysiol 51:1326–1344. doi:10.1152/jn.1984.51.6.1326 pmid:6737033
    OpenUrlCrossRefPubMed
  14. ↵
    2. Delgutte B,
    3. Kiang NY
    (1984a) Speech coding in the auditory nerve: V. Vowels in background noise. J Acoust Soc Am 75:908–918. doi:10.1121/1.390537 pmid:6707320
    OpenUrlCrossRefPubMed
  15. ↵
    2. Delgutte B,
    3. Kiang NYS
    (1984b) Speech coding in the auditory nerve: I. Vowel-like sounds. J Acoust Soc Am 75:866–878. doi:10.1121/1.390596 pmid:6707316
    OpenUrlCrossRefPubMed
  16. ↵
    2. Delgutte B,
    3. Kiang NYS
    (1984c) Speech coding in the auditory nerve: III. Voiceless fricative consonants. J Acoust Soc Am 75:887–896. doi:10.1121/1.390598 pmid:6707318
    OpenUrlCrossRefPubMed
  17. ↵
    2. Delgutte B,
    3. Hammond BM,
    4. Cariani PA
    (1998) Neural coding of the temporal envelope of speech: relation to modulation transfer functions. In: Psychophysical and Physiological Advances in Hearing (Palmer AR, Reese A, Summerfield AQ, Meddis R, eds), pp 595–603. London: Whurr.
  18. ↵
    2. Dubno JR,
    3. Dirks DD
    (1989) Auditory filter characteristics and consonant recognition for hearing-impaired listeners. J Acoust Soc Am 85:1666–1675. doi:10.1121/1.397955 pmid:2708682
    OpenUrlCrossRefPubMed
  19. ↵
    2. Dubno JR,
    3. Dirks DD,
    4. Langhofer LR
    (1982) Evaluation of hearing-impaired listeners using a nonsense-syllable test II. Syllable recognition and consonant confusion patterns. J Speech Hear Res 25:141–148. doi:10.1044/jshr.2501.141 pmid:7087417
    OpenUrlCrossRefPubMed
  20. ↵
    2. Dubno JR,
    3. Eckert MA,
    4. Lee FS,
    5. Matthews LJ,
    6. Schmiedt RA
    (2013) Classifying human audiometric phenotypes of age-related hearing loss from animal models. J Assoc Res Otolaryngol 14:687–701. doi:10.1007/s10162-013-0396-x pmid:23740184
    OpenUrlCrossRefPubMed
  21. ↵
    2. Elhilali M
    (2019) Modulation representations for speech and music. In: Timbre: acoustics, perception, and cognition (Siedenburg K, Saitis C, McAdams S, Popper AN, and Fay RR, eds), pp 335–359. Cham: Springer International Publishing.
  22. ↵
    2. Elliott TM,
    3. Theunissen FE
    (2009) The modulation transfer function for speech intelligibility. PLoS Comput Biol 5:e1000302. doi:10.1371/journal.pcbi.1000302
    OpenUrlCrossRefPubMed
  23. ↵
    2. Fernandez KA,
    3. Jeffers PWC,
    4. Lall K,
    5. Liberman MC,
    6. Kujawa SG
    (2015) Aging after noise exposure: acceleration of cochlear synaptopathy in “recovered” ears. J Neurosci 35:7509–7520. doi:10.1523/JNEUROSCI.5138-14.2015 pmid:25972177
    OpenUrlAbstract/FREE Full Text
  24. ↵
    2. Fernandez KA,
    3. Guo D,
    4. Micucci S,
    5. De Gruttola V,
    6. Liberman MC,
    7. Kujawa SG
    (2020) Noise-induced cochlear synaptopathy with and without sensory cell loss. Neuroscience 427:43–57. doi:10.1016/j.neuroscience.2019.11.051 pmid:31887361
    OpenUrlCrossRefPubMed
  25. ↵
    2. Festen JM,
    3. Plomp R
    (1983) Relations between auditory functions in impaired hearing. J Acoust Soc Am 73:652–662. doi:10.1121/1.388957 pmid:6841805
    OpenUrlCrossRefPubMed
  26. ↵
    2. Festen JM,
    3. Plomp R
    (1990) Effects of fluctuating noise and interfering speech on the speech-reception threshold for impaired and normal hearing. J Acoust Soc Am 88:1725–1736. doi:10.1121/1.400247 pmid:2262629
    OpenUrlCrossRefPubMed
  27. ↵
    2. Fogerty D,
    3. Bologna WJ,
    4. Ahlstrom JB,
    5. Dubno JR
    (2017) Simultaneous and forward masking of vowels and stop consonants: effects of age, hearing loss, and spectral shaping. J Acoust Soc Am 141:1133–1143. doi:10.1121/1.4976082 pmid:28253707
    OpenUrlCrossRefPubMed
  28. ↵
    2. Fox J,
    3. Weisberg S
    (2018) An R companion to applied regression. Thousand Oaks: SAGE Publications.
  29. ↵
    2. French NR,
    3. Steinberg JC
    (1947) Factors governing the intelligibility of speech sounds. J Acoust Soc Am 19:90–119. doi:10.1121/1.1916407
    OpenUrlCrossRef
  30. ↵
    2. Geisler CD,
    3. Gamble T
    (1989) Responses of “high-spontaneous” auditory-nerve fibers to consonant–vowel syllables in noise. J Acoust Soc Am 85:1639–1652. doi:10.1121/1.397952 pmid:2708680
    OpenUrlCrossRefPubMed
  31. ↵
    2. Geisler CD,
    3. Silkes SM
    (1991) Responses of “lower-spontaneous-rate” auditory-nerve fibers to speech syllables presented in noise. II: glottal-pulse periodicities. J Acoust Soc Am 90:3140–3148. doi:10.1121/1.401422
    OpenUrlCrossRefPubMed
  32. ↵
    2. Glasberg BR,
    3. Moore BCJ
    (1986) Auditory filter shapes in subjects with unilateral and bilateral cochlear impairments. J Acoust Soc Am 79:1020–1033. doi:10.1121/1.393374 pmid:3700857
    OpenUrlCrossRefPubMed
  33. ↵
    2. Glasberg BR,
    3. Moore BCJ,
    4. Bacon SP
    (1987) Gap detection and masking in hearing-impaired and normal-hearing subjects. J Acoust Soc Am 81:1546–1556. doi:10.1121/1.394507 pmid:3584692
    OpenUrlCrossRefPubMed
  34. ↵
    2. Goblick TJ,
    3. Pfeiffer RR
    (1969) Time-domain measurements of cochlear nonlinearities using combination click stimuli. J Acoust Soc Am 46:924–938. doi:10.1121/1.1911812 pmid:4309951
    OpenUrlCrossRefPubMed
  35. ↵
    2. Goman AM,
    3. Lin FR
    (2016) Prevalence of hearing loss by severity in the United States. Am J Public Health 106:1820–1822. doi:10.2105/AJPH.2016.303299 pmid:27552261
    OpenUrlCrossRefPubMed
  36. ↵
    2. Halliday LF,
    3. Rosen S,
    4. Tuomainen O,
    5. Calcus A
    (2019) Impaired frequency selectivity and sensitivity to temporal fine structure, but not envelope cues, in children with mild-to-moderate sensorineural hearing loss. J Acoust Soc Am 146:4299–4314. doi:10.1121/1.5134059 pmid:31893709
    OpenUrlCrossRefPubMed
  37. ↵
    2. Harrison RV,
    3. Evans EF
    (1979) Some aspects of temporal coding by single cochlear fibres from regions of cochlear hair cell degeneration in the guinea pig. Arch Otorhinolaryngol 224:71–78. doi:10.1007/BF00455226 pmid:485951
    OpenUrlCrossRefPubMed
  38. ↵
    2. Heinz MG,
    3. Issa JB,
    4. Young ED
    (2005) Auditory-nerve rate responses are inconsistent with common hypotheses for the neural correlates of loudness recruitment. J Assoc Res Otolaryngol 6:91–105. doi:10.1007/s10162-004-5043-0 pmid:15952047
    OpenUrlCrossRefPubMed
  39. ↵
    2. Henry KS,
    3. Heinz MG
    (2012) Diminished temporal coding with sensorineural hearing loss emerges in background noise. Nat Neurosci 15:1362–1364. doi:10.1038/nn.3216 pmid:22960931
    OpenUrlCrossRefPubMed
  40. ↵
    2. Henry KS,
    3. Kale S,
    4. Scheidt RE,
    5. Heinz MG
    (2011) Auditory brainstem responses predict auditory nerve fiber thresholds and frequency selectivity in hearing impaired chinchillas. Hear Res 280:236–244. doi:10.1016/j.heares.2011.06.002 pmid:21699970
    OpenUrlCrossRefPubMed
  41. ↵
    2. Henry KS,
    3. Kale S,
    4. Heinz MG
    (2016) Distorted tonotopic coding of temporal envelope and fine structure with noise-induced hearing loss. J Neurosci 36:2227–2237. doi:10.1523/JNEUROSCI.3944-15.2016 pmid:26888932
    OpenUrlAbstract/FREE Full Text
  42. ↵
    2. Henry KS,
    3. Sayles M,
    4. Hickox AE,
    5. Heinz MG
    (2019) Divergent auditory-nerve encoding deficits between two common etiologies of sensorineural hearing loss. J Neurosci 39:6879–6887. doi:10.1523/JNEUROSCI.0038-19.2019 pmid:31285299
    OpenUrlAbstract/FREE Full Text
  43. ↵
    2. Horst JW
    (1987) Frequency discrimination of complex signals, frequency selectivity, and speech perception in hearing-impaired subjects. J Acoust Soc Am 82:874–885. doi:10.1121/1.395286 pmid:3655121
    OpenUrlCrossRefPubMed
  44. ↵
    2. Jenstad LM,
    3. Pumford J,
    4. Seewald RC,
    5. Cornelisse LE
    (2000) Comparison of linear gain and wide dynamic range compression hearing aid circuits II: aided loudness measures. Ear Hear 21:32–44. pmid:10708072
    OpenUrlCrossRefPubMed
  45. ↵
    2. Kale S,
    3. Heinz MG
    (2010) Envelope coding in auditory nerve fibers following noise-induced hearing loss. J Assoc Res Otolaryngol 11:657–673. doi:10.1007/s10162-010-0223-6 pmid:20556628
    OpenUrlCrossRefPubMed
  46. ↵
    2. Kennedy E,
    3. Levitt H,
    4. Neuman AC,
    5. Weiss M
    (1998) Consonant–vowel intensity ratios for maximizing consonant recognition by hearing-impaired listeners. J Acoust Soc Am 103:1098–1114. doi:10.1121/1.423108 pmid:9479764
    OpenUrlCrossRefPubMed
  47. ↵
    2. Kenward MG,
    3. Roger JH
    (1997) Small sample inference for fixed effects from restricted maximum likelihood. Biometrics 53:983–997. doi:10.2307/2533558 pmid:9333350
    OpenUrlCrossRefPubMed
  48. ↵
    2. Kiang NYS,
    3. Moxon EC
    (1974) Tails of tuning curves of auditory-nerve fibers. J Acoust Soc Am 55:620–630. doi:10.1121/1.1914572 pmid:4819862
    OpenUrlCrossRefPubMed
  49. ↵
    2. King A,
    3. Hopkins K,
    4. Plack CJ
    (2016) Differential group delay of the frequency following response measured vertically and horizontally. J Assoc Res Otolaryngol 17:133–143. doi:10.1007/s10162-016-0556-x pmid:26920344
    OpenUrlCrossRefPubMed
  50. ↵
    2. Kujawa SG,
    3. Liberman MC
    (2009) Adding insult to injury: cochlear nerve degeneration after “temporary” noise-induced hearing loss. J Neurosci 29:14077–14085. doi:10.1523/JNEUROSCI.2845-09.2009 pmid:19906956
    OpenUrlAbstract/FREE Full Text
  51. ↵
    2. Le Prell CG,
    3. Clavier OH
    (2017) Effects of noise on speech recognition: challenges for communication by service members. Hear Res 349:76–89. doi:10.1016/j.heares.2016.10.004 pmid:27743882
    OpenUrlCrossRefPubMed
  52. ↵
    2. Lesica NA
    (2018) Why do hearing aids fail to restore normal auditory perception? Trends Neurosci 41:174–185.
    OpenUrl
  53. ↵
    2. Liberman MC
    (1984) Single-neuron labeling and chronic cochlear pathology. I. Threshold shift and characteristic-frequency shift. Hear Res 16:33–41. doi:10.1016/0378-5955(84)90023-6
    OpenUrlCrossRefPubMed
  54. ↵
    2. Liberman MC,
    3. Dodds LW
    (1984a) Single-neuron labeling and chronic cochlear pathology. III. Stereocilia damage and alterations of threshold tuning curves. Hear Res 16:55–74. doi:10.1016/0378-5955(84)90025-X pmid:6511673
    OpenUrlCrossRefPubMed
  55. ↵
    2. Liberman MC,
    3. Dodds LW
    (1984b) Single-neuron labeling and chronic cochlear pathology. II. Stereocilia damage and alterations of spontaneous discharge rates. Hear Res 16:43–53. doi:10.1016/0378-5955(84)90024-8 pmid:6511672
    OpenUrlCrossRefPubMed
  56. ↵
    2. Lorenzi C,
    3. Gilbert G,
    4. Carn H,
    5. Garnier S,
    6. Moore BCJ
    (2006) Speech perception problems of the hearing impaired reflect inability to use temporal fine structure. Proc Natl Acad Sci USA 103:18866–18869. doi:10.1073/pnas.0607364103 pmid:17116863
    OpenUrlAbstract/FREE Full Text
  57. ↵
    2. Lybarger SF
    (1978) Selective amplification—a review and evaluation. Ear Hear 3:258.
    OpenUrl
  58. ↵
    2. McCormack A,
    3. Fortnum H
    (2013) Why do people fitted with hearing aids not wear them? Int J Audiol 52:360–368. doi:10.3109/14992027.2013.769066 pmid:23473329
    OpenUrlCrossRefPubMed
  59. ↵
    2. Miller JD,
    3. Watson CS,
    4. Covell WP
    (1963) Deafening effects of noise on the cat. Acta Otolaryngol Suppl 176:1–81.
    OpenUrl
  60. ↵
    2. Miller RL,
    3. Schilling JR,
    4. Franck KR,
    5. Young ED
    (1997) Effects of acoustic trauma on the representation of the vowel/ε/in cat auditory nerve fibers. J Acoust Soc Am 101:3602–3616. doi:10.1121/1.418321 pmid:9193048
    OpenUrlCrossRefPubMed
  61. ↵
    2. Moore BCJ
    (2007) Cochlear hearing loss: physiological, psychological, and technical issues, Ed 2. Chichester: Wiley.
  62. ↵
    2. Moore BCJ
    (2008) The role of temporal fine structure processing in pitch perception, masking, and speech perception for normal-hearing and hearing-impaired people. J Assoc Res Otolaryngol 9:399–406. doi:10.1007/s10162-008-0143-x pmid:18855069
    OpenUrlCrossRefPubMed
  63. ↵
    2. Moore BCJ,
    3. Glasberg BR,
    4. Hess RF,
    5. Birchall JP
    (1985) Effects of flanking noise bands on the rate of growth of loudness of tones in normal and recruiting ears. J Acoust Soc Am 77:1505–1513. doi:10.1121/1.392045 pmid:3989105
    OpenUrlCrossRefPubMed
  64. ↵
    2. Moore BCJ,
    3. Stone MA,
    4. Füllgrabe C,
    5. Glasberg BR,
    6. Puria S
    (2008) Spectro-temporal characteristics of speech at high frequencies, and the potential for restoration of audibility to people with mild-to-moderate hearing loss. Ear Hear 29:907–922. doi:10.1097/AUD.0b013e31818246f6 pmid:18685497
    OpenUrlCrossRefPubMed
  65. ↵
    2. Moore BCJ,
    3. Glasberg BR,
    4. Stone MA
    (2010) Development of a new method for deriving initial fittings for hearing aids with multi-channel compression: CAMEQ2-HF. Int J Audiol 49:216–227. doi:10.3109/14992020903296746 pmid:20151930
    OpenUrlCrossRefPubMed
  66. ↵
    2. Ngan EM,
    3. May BJ
    (2001) Relationship between the auditory brainstem response and auditory nerve thresholds in cats with hearing loss. Hear Res 156:44–52. doi:10.1016/S0378-5955(01)00264-7 pmid:11377881
    OpenUrlCrossRefPubMed
  67. ↵
    2. Nielsen JB,
    3. Dau T
    (2009) Development of a Danish speech intelligibility test. Int J Audiol 48:729–741. doi:10.1080/14992020903019312 pmid:19626512
    OpenUrlCrossRefPubMed
  68. ↵
    2. Parida S,
    3. Heinz MG
    (2021) Noninvasive measures of distorted tonotopic speech coding following noise-induced hearing loss. J Assoc Res Otolaryngol 22:51–66. doi:10.1007/s10162-020-00755-2 pmid:33188506
    OpenUrlCrossRefPubMed
  69. ↵
    2. Parida S,
    3. Bharadwaj H,
    4. Heinz MG
    (2021) Spectrally specific temporal analyses of spike-train responses to complex sounds: a unifying framework. PLoS Comput Biol 17:e1008155. doi:10.1371/journal.pcbi.1008155 pmid:33617548
    OpenUrlCrossRefPubMed
  70. ↵
    2. Parthasarathy A,
    3. Romero Pinto S,
    4. Lewis RM,
    5. Goedicke W,
    6. Polley DB
    (2020) Data-driven segmentation of audiometric phenotypes across a large clinical cohort. Sci Rep 10:6704. pmid:32317648
    OpenUrlPubMed
  71. ↵
    2. Phatak SA,
    3. Grant KW
    (2014) Phoneme recognition in vocoded maskers by normal-hearing and aided hearing-impaired listeners. J Acoust Soc Am 136:859–866. doi:10.1121/1.4889863 pmid:25096119
    OpenUrlCrossRefPubMed
  72. ↵
    2. Rosen S
    (1992) Temporal information in speech: acoustic, auditory, and linguistic aspects. Phil Trans R Soc Lond B 336:367–373.
    OpenUrlCrossRefPubMed
  73. ↵
    2. Rosenthal R,
    3. Rubin DB
    (2003) r(equivalent:) a simple effect size indicator. Psychol Methods 8:492–496. doi:10.1037/1082-989X.8.4.492 pmid:14664684
    OpenUrlCrossRefPubMed
  74. ↵
    2. Ruggero MA,
    3. Rich NC
    (1991) Furosemide alters organ of Corti mechanics: evidence for feedback of outer hair cells upon the basilar membrane. J Neurosci 11:1057–1067. doi:10.1523/JNEUROSCI.11-04-01057.1991 pmid:2010805
    OpenUrlAbstract/FREE Full Text
  75. ↵
    2. Sachs MB,
    3. Voigt HF,
    4. Young ED
    (1983) Auditory nerve representation of vowels in background noise. J Neurophysiol 50:27–45. doi:10.1152/jn.1983.50.1.27 pmid:6875649
    OpenUrlCrossRefPubMed
  76. ↵
    2. Sayles M,
    3. Heinz MG
    (2017) Afferent coding and efferent control in the normal and impaired cochlea. In: Understanding the cochlea (Manley GA, Gummer AW, Popper AN, and Fay RR, eds), pp 215–252. Cham: Springer.
  77. ↵
    2. Scheidt RE,
    3. Kale S,
    4. Heinz MG
    (2010) Noise-induced hearing loss alters the temporal dynamics of auditory-nerve responses. Hear Res 269:23–33. doi:10.1016/j.heares.2010.07.009 pmid:20696230
    OpenUrlCrossRefPubMed
  78. ↵
    2. Schilling JR,
    3. Miller RL,
    4. Sachs MB,
    5. Young ED
    (1998) Frequency-shaped amplification changes the neural representation of speech with noise-induced hearing loss. Hear Res 117:57–70. doi:10.1016/S0378-5955(98)00003-3 pmid:9557978
    OpenUrlCrossRefPubMed
  79. ↵
    2. Shobha NH,
    3. Thomas TG,
    4. Subbarao K
    (2008) Influence of consonant-vowel intensity ratio on speech perception for hearing impaired listeners. Proc. Congress in Image and Signal processing, Hainan, China, pp 342–346.
  80. ↵
    2. Silkes SM,
    3. Geisler CD
    (1991) Responses of “lower-spontaneous-rate” auditory-nerve fibers to speech syllables presented in noise. I: general characteristics. J Acoust Soc Am 90:3122–3139. doi:10.1121/1.401421 pmid:1787250
    OpenUrlCrossRefPubMed
  81. ↵
    2. Sinex DG,
    3. Geisler CD
    (1983) Responses of auditory-nerve fibers to consonant–vowel syllables. J Acoust Soc Am 73:602–615. doi:10.1121/1.389007 pmid:6841800
    OpenUrlCrossRefPubMed
  82. ↵
    2. Skoe E,
    3. Kraus N
    (2010) Auditory brain stem response to complex sounds: a tutorial. Ear Hear 31:302–324. doi:10.1097/AUD.0b013e3181cdb272 pmid:20084007
    OpenUrlCrossRefPubMed
  83. ↵
    2. Souza P
    (2016) Speech perception and hearing aids. In: Hearing aids (Popelka GR, Moore BCJ, Fay RR, and Popper AN, eds), pp 151–180. New York: Springer International Publishing.
  84. ↵
    2. Thomson DJ
    (1982) Spectrum estimation and harmonic analysis. Proc IEEE 70:1055–1096. doi:10.1109/PROC.1982.12433
    OpenUrlCrossRef
  85. ↵
    2. Trevino M,
    3. Lobarinas E,
    4. Maulden AC,
    5. Heinz MG
    (2019) The chinchilla animal model for hearing science and noise-induced hearing loss. J Acoust Soc Am 146:3710–3732. doi:10.1121/1.5132950 pmid:31795699
    OpenUrlCrossRefPubMed
  86. ↵
    2. Van de Grift Turek S,
    3. Dorman MF,
    4. Franks JR,
    5. Summerfield Q
    (1980) Identification of synthetic /bdg/ by hearing-impaired listeners under monotic and dichotic formant presentation. J Acoust Soc Am 67:1031–1040. doi:10.1121/1.384070 pmid:7358910
    OpenUrlCrossRefPubMed
  87. ↵
    2. Victor JD,
    3. Purpura KP
    (1996) Nature and precision of temporal coding in visual cortex: a metric-space analysis. J Neurophysiol 76:1310–1326. doi:10.1152/jn.1996.76.2.1310 pmid:8871238
    OpenUrlCrossRefPubMed
  88. ↵
    2. Woolf NK,
    3. Ryan AF,
    4. Bone RC
    (1981) Neural phase-locking properties in the absence of cochlear outer hair cells. Hear Res 4:335–346. doi:10.1016/0378-5955(81)90017-4 pmid:7263519
    OpenUrlCrossRefPubMed
  89. ↵
    2. Wu PZ,
    3. Liberman LD,
    4. Bennett K,
    5. de Gruttola V,
    6. O'Malley JT,
    7. Liberman MC
    (2019) Primary neural degeneration in the human cochlea: evidence for hidden hearing loss in the aging ear. Neuroscience 407:8–20. doi:10.1016/j.neuroscience.2018.07.053 pmid:30099118
    OpenUrlCrossRefPubMed
  90. ↵
    2. Young ED
    (2008) Neural representation of spectral and temporal information in speech. Philos Trans R Soc Lond B Biol Sci 363:923–945. doi:10.1098/rstb.2007.2151 pmid:17827107
    OpenUrlCrossRefPubMed
  91. ↵
    2. Young ED
    (2012) Neural coding of sound with cochlear damage. In: Noise-induced hearing loss: scientific advances (LePrell C and Henderson D, eds), pp 87–135. New York: Springer.
  92. ↵
    2. Young ED,
    3. Sachs MB
    (1979) Representation of steady-state vowels in the temporal aspects of the discharge patterns of populations of auditory-nerve fibers. J Acoust Soc Am 66:1381–1403. doi:10.1121/1.383532 pmid:500976
    OpenUrlCrossRefPubMed
  93. ↵
    2. Young ED,
    3. Barta PE
    (1986) Rate responses of auditory nerve fibers to tones in noise near masked threshold. J Acoust Soc Am 79:426–442. doi:10.1121/1.393530 pmid:3950195
    OpenUrlCrossRefPubMed
  94. ↵
    2. Zeng FG,
    3. Turner CW
    (1990) Recognition of voiceless fricatives by normal and hearing-impaired subjects. J Speech Hear Res 33:440–449. doi:10.1044/jshr.3303.440 pmid:2232762
    OpenUrlCrossRefPubMed
  95. ↵
    2. Zhong Z,
    3. Henry KS,
    4. Heinz MG
    (2014) Sensorineural hearing loss amplifies neural coding of envelope information in the central auditory system of chinchillas. Hear Res 309:55–62. doi:10.1016/j.heares.2013.11.006 pmid:24315815
    OpenUrlCrossRefPubMed
Back to top

In this issue

Email
Print
View Full Page PDF
Citation Tools
Respond to this article
Distorted Tonotopy Severely Degrades Neural Representations of Connected Speech in Noise following Acoustic Trauma
Satyabrata Parida, Michael G. Heinz
Journal of Neuroscience 23 February 2022, 42 (8) 1477-1490; DOI: 10.1523/JNEUROSCI.1268-21.2021
del.icio.us logo Digg logo Reddit logo Twitter logo CiteULike logo Facebook logo Google logo Mendeley logo

Jump to section

Keywords

Responses to this article

Respond to this article

Jump to comment:

No eLetters have been published for this article.