Effects of aging and hearing loss on perceptual and electrophysiological measures of pulse-rate discrimination

Abstract

Auditory temporal processing declines with age, leading to potential deleterious effects on communication. In young normal-hearing listeners, perceptual rate discrimination is rate limited around 300 Hz. It is not known whether this rate limitation is similar in older listeners with hearing loss. The purpose of this study was to investigate age- and hearing-loss-related rate limitations on perceptual rate discrimination, and age- and hearing-loss-related effects on neural representation of these stimuli. Younger normal-hearing, older normal-hearing, and older hearing-impaired listeners performed a pulse-rate discrimination task at rates of 100, 200, 300, and 400 Hz. Neural phase locking was assessed using the auditory steady-state response. Finally, a battery of non-auditory cognitive tests was administered. Younger listeners had better rate discrimination, higher phase locking, and higher cognitive scores compared to both groups of older listeners. Aging, but not hearing loss, diminished neural-rate encoding and perceptual performance; however, there was no relationship between the perceptual and neural measures. Higher cognitive scores were correlated with improved perceptual performance, but not with neural phase locking. This study shows that aging, rather than hearing loss, may be a stronger contributor to poorer temporal processing, and cognitive factors such as processing speed and inhibitory control may be related to these declines.

I. INTRODUCTION

As people age, temporal processing abilities tend to degrade; this has been demonstrated in both neural encoding [e.g., Anderson et al. (2012)] and perceptual performance [e.g., Gordon-Salant and Fitzgibbons (1993), Gordon-Salant et al. (2006), Gordon-Salant et al. (2008), and Füllgrabe et al. (2015)] experiments. These changes likely originate from a combination of peripheral, central, and cognitive factors, which are difficult to separate. Willott (1991) posited two hypotheses about age-related central processing changes, which were derived from a rodent model. The first hypothesis, the central effect of biological aging (CEBA), suggests that poorer central temporal processing is a result of deteriorating neural structures that occur independent of peripheral structure damage. The second hypothesis, the central effect due to peripheral pathology (CEPP), suggests that peripheral process degradation (e.g., hair cell loss, spiral ganglion neuron deterioration) causes an upstream effect on central auditory structures, leading to altered auditory processing in older people.

Despite numerous investigations, the relative contributions of peripheral and central changes associated with aging effects on perception remains elusive. One approach that may be partially successful in separating the effects of age and hearing loss is to measure temporal processing in younger normal-hearing (YNH), older normal-hearing (ONH), and older hearing-impaired (OHI) listeners using non-speech stimuli. The choice of non-speech stimuli reduces the influence of higher-level language abilities that may affect speech perception (Owens, 1961; Humes et al., 2012). One such temporal processing measure that can be assessed is pulse-rate discrimination. Age-related deficits in pulse-rate discrimination are associated with difficulties with perception of suprasegmental aspects of speech, such as gender identification [e.g., Grimault (2003)] and prosodic cues (Mitchell and Kingston, 2014).

The auditory system is rate limited in YNH listeners. Specifically, temporal information is processed monaurally up to 100 Hz in auditory cortex and up to approximately 4000 Hz in the periphery, though there is debate on the rate limit in the high frequency range [for a review, see Verschooten et al. (2019)]. The auditory system low-pass filters temporal information through ascending stages in the auditory pathway. YNH listeners are able to adequately identify differences in band limited pulse trains up to about 200 to 300 Hz, with the best performance at 100 Hz (Carlyon et al., 2008). A similar 200- to 300-Hz rate limitation has also been observed for interaural-time-difference sensitivity using high-frequency modulated stimuli [e.g., Bernstein and Trahiotis (2002) and Goupell et al. (2009)].

Temporal processing in the aging auditory system has been assessed using the auditory steady-state response (ASSR) (a scalp-recorded evoked potential that measures synchronization to the temporal envelope of amplitude-modulated stimuli) in ONH listeners for amplitude-modulation rates up to 128 Hz (Grose et al., 2009; Goossens et al., 2016). In Goossens et al. (2019), ASSR was measured in normal-hearing and hearing-impaired younger (20–30 years), middle-aged (50–60 years), and older adults (70–80 years). They were presented octave-bands of white noise centered at 1 kHz that were sinusoidally amplitude modulated at 4, 20, 40, and 80 Hz. The different rates were used to target both lower and higher levels of the auditory system. Stimuli were presented both at a constant level of 70 dB SPL and at individually adjusted comfortable levels. While the investigators did not explicitly analyze age as a factor, they found differences in the effects of hearing loss between age groups. Specifically, they found that hearing loss was associated with enhanced temporal envelope synchronization for the 80-Hz modulation rate representing subcortical regions and the 4-Hz modulation rate representing cortical regions in the younger and middle-aged listeners. The increased synchronization may have been a result of neural compensation associated with hearing loss or loss of cochlear compression. However, this enhanced synchronization was not found in the older listeners. These studies did not analyze whether independent effects of age and hearing existed in their neural data, nor did they evaluate perceptual and cognitive performance. To better understand the effects of aging and hearing loss on simple temporal processing abilities, combining perceptual, neural, and cognitive measures may help elucidate whether independent contributions of aging and hearing loss exist.

The literature evaluating age-related rate limitations with simple temporal stimuli both in the perceptual and electrophysiological domains is limited. One reason for this lack of research is a stimulus confound; changes in temporal aspects of an acoustic signal have concomitant spectral changes that can be problematic when those changes interact with cochlear filtering. To avoid these confounds, investigators have evaluated rate processing in listeners with cochlear implants (Carlyon and Deeks, 2002; Macherey and Carlyon, 2014; Johnson et al., 2021); however, fundamental differences between acoustic and electric hearing make it difficult to parse potential independent contributions of the aging process. Furthermore, electrophysiological recording with cochlear implants is often complicated by electrical artifacts (Luke et al., 2015).

In one study using both neural and perceptual measures, Gaskins et al. (2019) showed that ONH listeners had lower (worse) ASSR SNRs at 400 Hz than those observed in YNH listeners. However, ONH listeners' perceptual rate discrimination performance was similar to that of YNH listeners from 80 to 400 Hz. Nonetheless, within each listener group, the strength of ASSR spectral energy at 400 Hz was predictive of perceptual performance at 400 Hz, as was a non-auditory measure of speed of processing. This relationship between neural encoding and perception suggests variability in neural representation of pulse trains is a factor in perceptual performance. The Gaskins study, however, was limited to only normal-hearing listeners and there were relatively small sample sizes (15 YNH and 15 ONH listeners).

The purpose of the current study is to expand upon Gaskins et al. (2019) by testing not only YNH and ONH listeners, but also older listeners with mild-to-moderate sensorineural hearing loss (OHI). This allows for better disambiguation between: (1) effects of age and hearing loss and (2) contributions of peripheral and central sources of rate limitations. ASSR was measured in response to pulse trains presented at relatively high rates, allowing for measures of auditory encoding at lower levels of the auditory system (i.e., brainstem and midbrain), and was compared to judgments of identical stimuli in the perceptual domain.

Finally, cognitive measures were included in this study as potential contributors to differences in perceptual measures. It has been well-established that cognitive function decreases with age [see Lipnicki et al. (2017) for review]. Cognitive measures (e.g., executive function, speed of processing, working memory, inhibitory control) have been shown to contribute to differences in temporal processing in older adults using non-speech tasks such as gap detection (Harris et al., 2010) and temporal fine structure sensitivity (Füllgrabe et al., 2015). This study included a comprehensive set of cognitive measures to evaluate the potential contribution of these factors on pulse-rate discrimination in older listeners, irrespective of hearing status.

It was hypothesized that: (1) ONH listeners will have better neural phase locking and perceptual rate discrimination performance than OHI listeners; (2) YNH listeners will have better neural phase locking and perceptual rate discrimination performance than both ONH and OHI listeners; (3) neural phase locking to pulse trains will correlate with perceptual pulse-rate discrimination within and across groups; and (4) cognitive factors, particularly speed of processing, will impact perceptual measures in all older adults.

II. METHODS

A. Listeners

This study included three groups: 25 YNH (age: 18–30 years, M = 20.8 years, SD = 2.7 years), 38 ONH (age: 64–77 years, M = 68.8 years, SD = 3.7 years, and 22 OHI (age: 65–84 years, M = 73.6 years, SD = 6.3 years) listeners. It should be noted that the mean and maximum age represented in the OHI group is higher than that of the ONH group. For the YNH and ONH groups, normal hearing was defined as pure-tone thresholds ≤25 dB HL (ANSI, 2018) at octave frequencies from 250 to 4000 Hz. OHI listeners had mild to moderate sensorineural hearing loss, with high-frequency pure-tone averages (1, 2, and 4 kHz) >30 dB HL and thresholds at 2 and 4 kHz <70 dB HL. The hearing thresholds across ears were within 10 dB at each frequency. Individual audiograms (right and left ears) for each listener group are shown in Fig. 1. A two-way ANOVA was performed to assess differences in threshold across frequencies and groups. There was a main effect of listener group (p < 0.001), because YNH listeners had lower (better) thresholds than both ONH (p < 0.001) and OHI listeners (p < 0.001), and ONH listeners had lower thresholds than OHI listeners (p < 0.001). There was a significant main effect of frequency (p < 0.001). There was also a significant frequency × group interaction (p < 0.001), where there were higher (poorer) thresholds for the OHI listeners at the higher frequencies compared to the two normal-hearing groups, as expected given the inclusion criteria. Listeners also underwent an auditory brainstem response screening to ensure appropriate wave V morphology as determined visually by a research audiologist, per standard clinical practice. In addition, wave V latency was required to be in the normal range (< 6.8 ms) and there was minimal interaural latency asymmetry (< 0.2 ms).

FIG. 1.

(Color online) Individual and average hearing thresholds and 95% confidence intervals from 0.25 to 8 kHz for YNH (triangles), ONH (squares), and OHI (circles) groups. Thinner lines represent individual data and thicker lines represent the means for each group.

All listeners passed a visual acuity screening prior to cognitive testing. Prior to enrollment, listeners passed a cognitive screen with a score of ≥ 26 in the Montreal Cognitive Assessment (MoCA) [version 7.1; Dupuis et al. (2015) and Nasreddine et al. (2005)] before continuing with the experiments (YNH: M = 28.7, SD = 1.1; ONH: M = 27.7, SD = 1.4; OHI: M = 27, SD = 1.2). A one-way ANOVA was conducted to detect differences in MoCA scores across groups. There was a main effect of group (p < 0.001). Post hoc testing shows that YNH listeners had higher average MoCA scores than the ONH (p = 0.01) and OHI (p = 0.0002) listeners. ONH and OHI listener MoCA scores were not significantly different (p = 0.18). All listeners were native speakers of American English and had at least a high school education. Each listener provided written consent, and the experiment was approved by the University of Maryland Institutional Review Board.

B. Stimuli

The stimuli consisted of 300-ms bandpass-filtered pulse trains with a 1000-Hz bandwidth (3500–4500 Hz). A 4000-Hz center frequency was used to avoid the effect of low-frequency temporal fine structure. The 1000-Hz bandwidth was used to maintain fully modulated pulse trains up to the maximum 400-Hz pulse rate. The filtering was done with a 5th-order forward-backward bandpass Butterworth filter, resulting in 10th-order filter slopes of 60 dB/oct. Pulse trains were presented at four pulse rates: 100, 200, 300, or 400 Hz for both perceptual and neural measures. Pulse rates below 100 Hz were not used to ensure that the perceptual task was challenging enough for listeners in all groups, though lower rates (e.g., 40 or 80 Hz) might be preferred to measure higher levels of the auditory system with the ASSR. The highest pulse rate presented, 400 Hz, is likely the highest rate that can be used and still elicit a measurable neural response. Thus, these rates were chosen to allow for direct comparison between perceptual and neural measures, while still allowing for reliable responses for both tasks.

For the electrophysiological experiment, stimuli were presented at a sampling frequency of 10 kHz and were recorded with online filtering from 1 to 5 kHz. Each trial had a presentation rate of 1.66 Hz, or 600 ms duration, with stimuli presented in the first 300 ms followed by 300 ms of silence.

In the perceptual experiment, a low-frequency masking noise presented at an overall level of 61 dB SPL was mixed with the pulse train stimuli to eliminate the use of low-frequency distortion products during the task. Wideband masking noise was low-pass filtered using a 200-Hz cutoff with a −3 dB/octave filter, and then was low-pass filtered using a 1000-Hz cutoff with a −5 dB/octave filter. A Tukey window was applied to the masking noise with a rise/fall time of 10 ms, the noise onset commenced 300 ms before the stimulus onset, and the noise offset occurred 300 ms after the offset of the last stimulus. Stimuli included ±10 dB level roving that was applied such that the level varied randomly across the three intervals on a stimulus trial. This was done to avoid the use of loudness cues, rather than temporal cues, during the experiment.

C. Procedure

1. ASSR

The ASSR was used to assess the neural response to rapid, repetitive auditory stimuli (Presacco et al., 2010). Testing was conducted in a double-walled electrically shielded, sound-attenuating chamber (IAC Acoustics, Naperville, IL). Listeners were seated in a reclining chair where they watched a silent, captioned movie to maintain a relaxed but awake state. Stimuli were presented through the Intelligent Hearing Systems Smart EP Continuous Acquisition Module (Miami, FL). A vertical montage of three electrodes (Cz active, forehead ground, and right earlobe reference) was used. Stimuli were delivered monaurally through electromagnetically shielded insert earphones (ER-3A; Etymotic, Elk Grove Village, IL) to the right ear at 75 dB SPL. One thousand artifact-free sweeps (responses <30 μV) were collected for each rate (1000 sweeps × 4 rates = 4000 sweeps) and analyzed offline using custom matlab programs (Mathworks, Natick, MA). ASSR measurement took about 60–90 min to complete.

2. Perceptual pulse-rate discrimination

Listeners were seated in a double-walled sound-attenuating booth during the perceptual pulse-rate discrimination experiment. Pulse trains were presented through a personal computer running a custom matlab program connected to a real-time processer (RP2.1; Tucker-Davis Technologies, Alachua, FL), headphone buffer (HB7), and programmable attenuator (PA4). All pulse trains were presented monaurally in the right ear using an insert earphone (ER-2; Etymotic, Elk Grove Village, IL) at a level of 75 dBA. Stimuli were presented monaurally to facilitate comparison with other studies measuring auditory temporal processing [e.g., Gaskins et al. (2019)].

Frequency difference limens (ΔF) were measured with a 3-interval-2-alternative forced choice (3I-2AFC) procedure with a 2-down-1-up adaptive rule to target 71% correct on the psychometric function (Levitt, 1971). There was a fixed number of 60 trials in each staircase. The order of frequency presentation was randomized. Three testing blocks were obtained per condition, totaling 720 trials (60 trials × 4 pulse rates × 3 blocks) for each listener, taking approximately 45–60 min to complete. The reference rate was fixed on each staircase and the target rate was adapted; the first interval was always the reference rate. The initial rate difference between the reference and the target rate was 40%. The maximum allowable rate difference was 40% and the minimum allowable rate difference was 0% (i.e., adaptive tracks could not go below the reference rate). The adaptation step size was then decreased by a factor of 2 until the listener reached three reversals, after which the step size decreased by a factor of √2. The DLs in percent for an individual adaptive track was found by calculating the geometric mean over all of the reversals in the adaptive procedure except the first two. The arithmetic mean of the discrimination thresholds measured in the second and third tracks was used to calculate the final DL for each listener and condition. The first track was omitted to decrease the effects of learning from the first track. The DLs were log-transformed due to a negative skew in the data prior to conducting the statistical analysis.

A trial was initiated by clicking a button on the testing interface. Each trial consisted of three 300-ms intervals with a 300-ms inter-stimulus interval. The first interval was always a reference interval. The target interval was the second or third interval, randomly chosen on each trial with equal probability. Each interval was visually indicated on the screen using one of three boxes. The listener was instructed to click on the second or third box that sounded different from the first reference interval. Correct answer feedback was provided after each trial. Since the listener initiated each trial, the trials were self-paced. Listeners were also able to take breaks between blocks to avoid fatigue. Responses were recorded using custom matlab software and frequency difference limens were calculated offline for each condition.

3. Cognitive testing

Several cognitive measures were administered, as they may account for additional variance in perceptual measurements, particularly in older adults. Using the NIH Toolbox (Gershon et al., 2013), measures of processing speed (Pattern Comparison Test), inhibitory control and attention (Flanker Test), working memory (List Sorting task), and executive function (Dimensional Card Sort task) were collected. The NIH Toolbox has been validated as a reliable tool (relative to other “gold standard” tests of cognition) across different age groups (Heaton et al., 2014). These tests were administered using the NIH Toolbox application on an Apple iPad (Apple Inc., Cupertino, CA) under the supervision of a research audiologist. This testing is visual in nature with minimal auditory cues from the software. The use of non-auditory cognitive testing was used to diminish the possible effects of hearing status on the cognitive scores (Füllgrabe, 2020; Lentz et al., 2022). Testing was completed in 30–45 min and was self-paced by the listener. Uncorrected standardized scores (M = 100, SD = 15) were used, which are derived from the NIH Toolbox normative sample across the age range. These standardized scores were used to allow for comparison across tests and listener groups.

D. Data analysis

1. ASSR

ASSR phase-locking factor (PLF) was used to assess the neural synchronization to pulse trains, an identical analysis to that used in previous studies (Jenkins et al., 2018; Roque et al., 2019a; Roque et al., 2019b). Morlet wavelets were used to decompose the signal from 50 to 500 Hz (Tallon-Baudry et al., 1996). ASSR phase-locking factor (PLF) was used to assess the neural synchronization to pulse trains, an identical analysis to that used in previous studies run in this lab (Jenkins et al., 2018; Roque et al., 2019a; Roque et al., 2019b). Complex Morlet wavelets with a Gaussian shape both in the time (SD σt) and in the frequency (SD = σf) domains were used to decompose the signal from 50 to 500 Hz at a 1-Hz step (Tallon-Baudry et al., 1996). The wavelet family used is defined by f0/σf = 7 (wavelet duration σt of about two periods of oscillatory activity at f0), where f0 is the central frequency and σf = 1/2πσt. As reported by Tallon-Baudry and colleagues, this configuration leads to a wavelet duration (2σt) of 111.4 ms at 20 Hz (spectral bandwidth = 5.8 Hz) and of 22.2 ms at 100 Hz (spectral bandwidth = 28.6 Hz). Therefore, the time resolution increases with frequency, whereas the frequency resolution decreases. Wavelets are also normalized to ensure that their total energy is 1. More details can be found in Tallon-Baudry et al. (1996). The PLF at each frequency step was calculated in a 20-Hz bin around each rate in the 10- to 300-ms time range. That means that for each rate, there are 5800 values (one for each sample in the selected time range of 290 ms) for each of the 21 frequencies selected (e.g., 90 to 110 Hz for the fundamental frequency of 100 Hz). The 5800 × 21 PLF calculated are then averaged to get a final value for each rate and each subject.

Individual ASSR PLF values were calculated for 20-Hz bins corresponding to each rate in the response region of 10 to 300 ms. These values were then averaged for each listener group. For analysis, all ASSR PLF values were log-transformed due to the negative skew in the data distribution.

ASSR PLF values were analyzed with a generalized linear mixed-effects regression (GLMER) (R Core Team, 2020), using the lmer package (Kuznetsova et al., 2017). The approach described by Hox et al. (2017) (pp. 161–172) builds the GLMER as follows: (1) a starting intercept-only model was constructed and used as a benchmark, (2) main effects and interactions between level-1 predictor variables (pulse rate and listener group) were added to the fixed effects structure, and non-significant interactions that did not result in any improvement in model fit (evaluated with an ANOVA significance test) were removed from the model, (3) random effects were introduced and kept based on model comparisons and successful model convergence, and (4) cross-level interactions were added to the model structure.

The final models contained the level-1 fixed effects of pulse rate (100, 200, 300, and 400 Hz) and listener group (YNH, ONH, and OHI), as well as their interactions. The referent rate category was 100 Hz, as this condition often had the most robust ASSR responses. Two models were necessary to interpret interactions with listener group: one using YNH as the referent listener category (model 1a), and one using ONH as the referent listener category (model 1b). The models also contained a random intercept of listener. A random intercept of rate was added, but this model did not converge, likely due to sample size, and was subsequently removed. Because the model was run twice with different referents, the α level was corrected for multiple analyses such that p < 0.025 (=0.05/2) were considered significant.

2. Perceptual pulse-rate discrimination

Pulse-rate discrimination for each condition (100, 200, 300, and 400 Hz) was quantified using relative DLs in percent. For example, the relative DL for a listener whose final averaged DL is 120 Hz for a reference rate of 100 Hz would have a relative DL of 20%. Relative DLs were analyzed with a GLMER, built similarly to that for the ASSR PLF values. The final models contained the level-1 fixed effects of pulse rate (100, 200, 300, and 400 Hz) and listener group (YNH, ONH, and OHI), as well as their interactions. The referent rate category was 100 Hz, consistent with the ASSR modeling. Two models were necessary to interpret interactions with listener group: one using YNH as the referent listener category (model 2a), and one using ONH as the referent listener category (model 2b). The model also contained a random intercept of listener. A random intercept of rate was added but this model did not converge, likely due to sample size, and was subsequently removed. Because the model was run twice with different referents, the α level was corrected for multiple analyses such that p < 0.025 (=0.05/2) was considered significant.

3. Relationships between electrophysiological, perceptual, and cognitive measures

A third GLMER was conducted to assess the combined effects of electrophysiological and cognitive measures on perceptual responses to pulse trains. As before, two similar models were run, one with YNH as the referent and one with ONH listeners as the referent category. Perceptual pulse rates at 100 Hz were used as the referent category for both models. Listener Group was not significant in either of these models and therefore only the YNH referent model was necessary. Multicollinearity diagnostics were run among cognitive measures for both models. The variance inflation factor was acceptable for the model with YNH as the referent (highest = 1.7) and with ONH as the referent (highest = 1.7) listener group.

III. RESULTS

A. Effects of aging and hearing loss on ASSR

Figure 2 displays ASSR PLFs for YNH, ONH, and OHI listeners in the time-frequency domain. Figure 3 displays mean ASSR PLF values for each rate and listener group. The results for the YNH listeners and 100-Hz rate referents are shown in Table I (middle columns). The intercept of model 1a represents the expected ASSR PLF value for a YNH listener at a 100-Hz rate. ASSR PLFs did not differ across listener group. ASSR PLFs at 300 Hz were significantly poorer than ASSR PLFs at 100 Hz (p = 0.001), but no other comparisons were significant.

FIG. 2.

(Color online) Auditory steady-state response (ASSR) phase locking factor (PLF) for each rate tested. The top row is the YNH group, the middle row is the ONH group, and the bottom row is the OHI group.

FIG. 3.

(Color online) Log-transformed ASSR PLF as a function of rate for YNH (triangles), ONH (squares), and OHI (circles) listeners. Symbol color and shape signify the group means and error bars are±1 standard error in length.

TABLE I.

GLMER for ASSR PLF. The reference rate was 100 Hz. The reference group was either YNH (model 1a) or ONH (model 1b). Since two models were performed, the significant α level was p < 0.025. Significant p-values are bold. The final model was: ASSR PLF ∼ Group × Rate + (1 | Subject).

	Model 1a (referenced to YNH)			Model 1b (referenced to ONH)
Fixed effects	Coefficients	SE	p	Coefficients	SE	p
(Intercept)	0.0589	0.0040	<0.001	0.0614	0.0033	<0.001
ONH	0.0025	0.0052	0.6364
OHI	−0.0050	0.0059	0.3958	−0.0075	0.0054	0.1673
200 Hz	−0.0008	0.0046	0.8542	−0.0161	0.0037	<0.001
300 Hz	−0.0160	0.0046	0.0005	−0.0286	0.0037	<0.001
400 Hz	−0.0078	0.0046	0.0902	−0.0251	0.0037	<0.001
ONH × 200 Hz	−0.0152	0.0059	0.0102
OHI × 200 Hz	−0.0139	0.0067	0.0386	0.0013	0.0062	0.8320
ONH × 300 Hz	−0.0126	0.0059	0.0331
OHI × 300 Hz	−0.0074	0.0067	0.2732	0.0053	0.0062	0.3937
ONH × 400 Hz	−0.0173	0.01	0.003
OHI × 400 Hz	−0.0111	0.01	0.100	0.01	0.01	0.308
YNH				0.001	0.01	0.636
YNH × 200 Hz				0.02	0.01	0.010
YNH × 300 Hz				0.01	0.01	0.033
YNH × 400 Hz				0.02	0.01	0.003
Random effects
Variance	0.0001450			0.01204
Residual	0.0002643 listener			0.01626 listener
Observations	340			340
Marginal R2/Conditional R2	0.215 / 0.493			0.215 / 0.493

Table I shows that there were significant interactions between ONH × 200 Hz (p < 0.02) and ONH × 400 Hz (p = 0.004) None of the other interactions were significant. These results indicate that the ONH group had poorer phase locking than YNH listeners that was rate specific at 200 and 400 Hz (Fig. 3). The OHI listeners did not differ significantly from YNH listeners.

For model 1b, the referent categories were 100 Hz and the ONH listener group. This model allows for direct comparisons between the ONH and OHI listener groups, thus isolating the potential effects of hearing loss on ASSR PLF values and allowing for interpretation of the significant interactions in model 1a. Since the intercept of model 1b represents the expected ASSR PLF value for an ONH listener at a 100 Hz rate, results may differ when comparing the models directly in Table I. The main effect of Rate was significant for 200 Hz (p < 0.001), 300 Hz (p < 0.001), and 400 Hz (p < 0.001), unlike model 1a that differed only for 300 Hz. There were significant interactions between YNH × 200 Hz (p = 0.01), and YNH × 400 Hz (p = 0.003).

In summary, models 1a and 1b indicate that ASSR PLFs in OHI and ONH listeners do not differ, but are worse than those of YNH listeners, most notably at 200 Hz. This supports the conclusion that older listeners have poorer neural temporal processing than younger listeners, and that differences in ASSR PLF are a result of age and not hearing loss.

B. Effects of aging and hearing loss on perceptual pulse-rate discrimination

Figure 4 displays relative DLs for YNH, ONH, and OHI groups. The results for the YNH listeners and 100-Hz rate referents are shown in model 2a of Table II (middle columns). The intercept of model 2a represents the expected relative DL value for a YNH listener at the 100-Hz rate. The model shows that the main effect of Listener Group was significant across rates for ONH (p < 0.001) and OHI (p < 0.001) listeners, meaning that both groups of older listeners had poorer relative DLs compared to the YNH group. The main of effect of Rate was significant for both older groups compared to the YNH group at 300 Hz (p = 0.001) and 400 Hz (p < 0.001); at these rates, older listeners had poorer DLs than younger listeners. The only significant interaction was OHI × 300 Hz (p < 0.001) when compared to YNH listeners. In summary, this model indicates that ONH and OHI listeners had poorer pulse-rate discrimination than YNH listeners, as expected. The combined effect of age and hearing loss was only apparent at 300 Hz, where OHI listeners exhibited significantly poorer rate discrimination than YNH listeners.

FIG. 4.

(Color online) Pulse-rate discrimination relative difference limens (DLs) as a function of rate for YNH (triangles), ONH (squares), and OHI (circles) listeners. Symbol color and shape signify the group means and error bars are±1 standard error in length.

TABLE II.

GLMER for relative DLs (%). The reference rate was 100 Hz. The reference group was either YNH (model 2a) or ONH (model 2b). Since two models were performed, the significant α level was p < 0.025. Significant p-values are bold. The final model was: Pulse Rate Discrimination ∼ Participant Group × Rate + (1 | Subject).

	Model 2a (referenced to YNH)			Model 2b (referenced to ONH)
Factors	Coefficients	SE	p	Coefficients	SE	p
(Intercept)	5.76	1.43	<0.001	12.68	1.1627	<0.001
ONH	6.92	1.85	<0.001
OHI	7.99	2.10	<0.001	1.07	1.9202	0.577
200 Hz	−0.99	1.34	0.459	−2.60	1.0855	0.017
300 Hz	−3.67	1.34	0.006	−2.49	1.7927	0.023
400 Hz	−4.29	1.34	0.001	−3.71	1.7927	0.001
ONH × 200 Hz	−1.61	1.72	0.351
OHI × 200 Hz	−0.76	1.96	0.698	0.85	1.7927	0.635
ONH × 300 Hz	1.18	1.72	0.492
OHI × 300 Hz	5.44	1.96	0.006	4.26	1.7927	0.018
ONH × 400 Hz	0.58	1.72	0.736
OHI × 400 Hz	1.76	1.96	0.368	1.18	1.7927	0.510
YNH				−6.92	1.8458	<0.001
YNH × 200 Hz				1.61	1.7232	0.351
YNH × 300 Hz				−1.18	1.7232	0.492
YNH × 400 Hz				−0.58	1.7232	0.736
Random effects
Variance	22.39			22.39
Residual	28.98 listener			28.98 listener
Observations	0.56			0.56
Marginal R2/Conditional R2	85 listener			85 listener

For model 2b in Table II, the referent categories were 100 Hz and the ONH listener group, which allows for direct comparisons between the ONH and OHI listener groups. The only significant main effect was for Rate at 200 Hz (p = 0.019), 300 Hz (p = 0.017), and 400 Hz (p < 0.001), similar to findings reported for model 1b. The only significant interaction was OHI × 300 Hz (p = 0.008). In summary, this model indicates no main effect of listener group in pulse-rate discrimination between OHI and ONH listeners. There was also significantly poorer rate discrimination for OHI listeners compared to ONH listeners at 300 Hz.

These results indicate an effect of aging on rate discrimination, where younger listeners have better perceptual rate discrimination than older listeners, similar to the findings observed with the ASSR PLF measures. Moreover, ONH and OHI listeners perform similarly on this task. Hearing impairment produced a selective decrement in performance at 300 Hz compared to both NH groups.

C. Comparison of electrophysiological, perceptual, and cognitive outcomes

Figure 5 shows cognitive scores for the list sort (working memory), pattern comparison (speed of processing), dimension card sorting (executive function), and the Flanker tasks (inhibitory control). A series of one-way ANOVAs with Listener Group as a fixed factor and Tukey adjustment were conducted for each cognitive test. For all tests except dimension card sorting, the YNH listeners had higher standardized scores compared to ONH and OHI listeners (p < 0.001). There were no differences between the ONH and OHI listeners on any test (p > 0.05), and no difference among all groups for the dimension card sort task (p > 0.05).

FIG. 5.

(Color online) Boxplots with individual data points for each listener group and cognitive test. (A) List sorting (working memory); (B) pattern comparison (speed of processing); (C) dimension card sort (executive function); and (D) flanker (inhibitory control) task. Note that y axes ranges differ due to variable ranges in the standardized scores for each cognitive test. The boxes represent the inter-quartile range for each cognitive test. The whiskers for each plot represent the minimum and maximum values.

GLMER model 3 was run to investigate the effects of the combination of cognitive and electrophysiological factors on behavioral pulse-rate discrimination. Cognitive values were each centered at their mean values so that coefficients for these variables are predictive of mean pulse-rate discrimination values. Results from model 3 are shown in Table III. For this model, the referent categories were YNH (for listener group) and 100 Hz (for pulse rate). The intercept of model 3 represents the expected relative DL value for a YNH listener at the 100-Hz rate. Backward elimination revealed no significant main effects or interactions with ASSR PLFs, list sorting task, or listener group. These terms were removed systematically to achieve the most parsimonious model; therefore, this is the only model where results are collapsed across listener groups. The main effect of pulse rate was significant for all rates, relative to 100 Hz (200 Hz, p = 0.011; 300 Hz, p = 0.006; 400 Hz, p < 0.001), similar to models 1 and 2. These coefficients are all negative, indicating that relative difference limens improve compared to 100 Hz when accounting for cognitive variables. The main effects of pattern comparison (p = 0.013), and Flanker (p < 0.001) were all significant. This model suggests that speed of processing, and inhibitory control are contributing factors to perceptual rate discrimination performance, regardless of age or hearing status.

TABLE III.

GLMER for relative DLs including electrophysiological and cognitive factors. The final model was: Pulse Rate Discrimination ∼1 + Rate + Pattern Comparison + Dimension Card Sort + Flanker + (1 | Subject).

	Model 3
Factors	Coefficients	SE	p
(Intercept)	10.99	0.76	<0.001
200 Hz (referenced to 100 Hz)	−1.86	0.75	0.014
300 Hz (referenced to 100 Hz)	−1.86	0.75	0.014
400 Hz (referenced to 100 Hz)	−3.56	0.75	<0.001
Pattern Comparison	−1.76	0.84	0.036
Flanker	−3.46	0.84	<0.001
Random Effects
Variance	23.61
Residual	24.30
Observations	336
Marginal R2/Conditional R2	0.303/0.656

IV. DISCUSSION

The purpose of this study was to disambiguate the effects of aging and hearing loss on relationships between neural encoding and pulse-rate perception of band limited pulse trains, and cognition. Our results show that compared to YNH listeners, ONH and OHI listeners have poorer neural representation of pulse trains (Figs. 2 and 3) and perceptual rate-discrimination performance (Fig. 4), and that perceptual performance is impacted by speed of processing and inhibitory control across all groups (Table III).

A. Effects of aging and hearing loss on neural encoding

Neural representation of pulse trains was poorer in the ONH and OHI groups as compared to the YNH group, as hypothesized (Fig. 2). Rate-specific age effects did emerge, in which all older listeners had poorer phase locking than younger listeners to 200- and 400-Hz rates. However, there was no difference in ASSR PLF between ONH and OHI listeners (model 1b), in contrast to our hypothesis that OHI listeners would have poorer neural encoding than ONH listeners.

By comparing YNH, ONH, and OHI groups, we are better able to differentiate between the effects of aging and those of hearing loss to the auditory system. There was no significant effect of hearing loss on ASSR PLF when comparing ONH and OHI listeners, at any rate tested. These results are similar to those of Goossens et al. (2019), who found no effects of hearing loss in older listeners on neural envelope encoding. In the Goossens et al. (2019) study, a 1000-Hz center frequency was used, whereas a 4000-Hz center frequency was used in the current study. The higher center frequency in our study was used to achieve a large bandwidth of 1000 Hz and maintain the modulation depth across pulse trains; however, this decision could have also resulted in reduced audibility for the OHI listeners (Fig. 1), which may have offset enhancement of the neural response in these listeners (Presacco et al., 2019).

Figure 3 shows that as the pulse rate increases from 100 Hz, phase locking significantly decreases at 300 Hz. This decrease in ASSR PLF may be due to (1) higher rates capturing responses from progressively lower levels in the brainstem, which are physically farther from the scalp electrodes and (2) fewer neurons firing synchronously at higher rates. In addition, as rate increases above 200 Hz, phase locking degrades precipitously in older listeners, regardless of hearing status. This rate-specific aging effect emerges, potentially sourced at brainstem (400 Hz) levels of the auditory system (Rees et al., 1986). These results are in line with previous studies that observed a Rate × Age interaction with varying ASSR rates (Gaskins et al., 2019; Leigh-Paffenroth and Fowler, 2006) and with higher modulation frequencies (Grose et al., 2009; Leigh-Paffenroth and Fowler, 2006). The results partially replicate Gaskins et al. (2019), who found that the ASSR signal-to-noise ratio (SNR) at 400 Hz in ONH listeners was poorer than that of the YNH group at 400 Hz (but not at 20, 40, 80, or 200 Hz). Gaskins et al. observed an overall decrease in SNR at higher rates, but only in the ONH listeners.

B. Effects of aging and hearing loss on perceptual pulse-rate discrimination

YNH listeners had better pulse-rate discrimination abilities compared to either ONH or OHI listeners at all rates (Fig. 4), consistent with our second hypothesis. ONH listeners had similar rate discrimination compared to OHI listeners, except at 300 Hz. Because no other rate-specific effects emerged related to hearing loss, our findings suggest a primary effect of aging on pulse-rate discrimination (Fig. 4). Nonetheless, these results partially support our hypothesis that the ONH group would have better perceptual rate discrimination than the OHI group, although this is only true at 300 Hz.

Studies testing perceptual rate discrimination in YNH, ONH, and OHI listeners are limited. Gaskins et al. (2019) measured pulse-rate discrimination in YNH and ONH listeners at 80, 200, and 400 Hz (in addition to ASSR SNRs). As with the ASSR data, they found that perceptual DLs increased (i.e., worsened) for both groups at faster rates, and that the older listeners (both ONH and OHI) exhibited higher pulse-rate DLs than YNH listeners, but only at the 400 Hz rate. In contrast, the present study showed that ONH listeners performed more poorly than YNH listeners at all rates. This may be due to differences in sample size or modulation rates used between the two studies (i.e., Gaskins et al. did not test 100 or 300 Hz).

Though we did not observe within-group differences in relative DLs with increasing rate as in the Gaskins et al. study, there was an effect of hearing loss observed at 300 Hz. Further, OHI listeners worsened markedly at 300 Hz, compared to ONH listeners, with a significant improvement in DLs at 400 Hz. This may be due to a local minimum around 300 Hz, wherein phase-locked signals arising from different neural generators are not phase-aligned (Tichko and Skoe, 2017). Carlyon and Deeks (2002) suggested that decreased relative DLs at higher rates may be due to resolved harmonics for 400-Hz pulse trains, thus confounding interpretation of effects of aging and hearing loss at higher rates. Poorer pulse rate discrimination at higher rates may be more pronounced in OHI listeners due to a broader critical bandwidth at these frequencies compared to ONH listeners.

Despite the relative lack of literature examining the effects of aging and hearing loss on acoustic rate discrimination in adults with hearing loss, several seminal studies have studied other temporal processing measures in this context [for an early review, see Fitzgibbons and Gordon-Salant (1996)]. These studies measured gap detection (Moore et al., 1992), duration discrimination (Fitzgibbons and Gordon-Salant, 1994), and discrimination of complex tonal sequences (Fitzgibbons and Gordon-Salant, 1995; Fitzgibbons and Gordon-Salant, 2001). Results showed that whether the stimuli were simple (e.g., gap detection) or complex (e.g., tonal sequence discrimination), effects of aging were observed with no effect of age-related hearing loss. The current study adds to these findings using rate discrimination, though more data are needed at a wider range of rates in both ONH and OHI listeners for a fuller picture to emerge.

C. Relationships between neural and perceptual measures and cognitive factors

Cognitive tests of working memory, speed of processing, executive function, and inhibitory control were conducted. YNH listeners have higher scores than both older listener groups, as expected (Fig. 5). Poorer scores on tests measuring these cognitive domains in older listeners compared to younger listeners are well-documented (Park and Reuter-Lorenz 2009). As the cognitive testing performed in this study was visual in nature, it is unsurprising that hearing loss did not have an effect on cognitive scores in the OHI group. However, it should be noted there is ongoing debate as to how the modality of cognitive testing affects construct validity and test scores (e.g., Shen et al., 2020).

Correlations between neural, perceptual, and cognitive performance were conducted and shown in model 3 (Table III). This model did not show any relationship between ASSR PLF and relative DLs, contrary to our hypothesis. This indicates that phase locking at these rates, as measured by ASSR, does not bear on perceptual performance. However, speed of processing, and inhibitory control—but not working memory and executive function—contributed to differences across groups in perceptual pulse-rate discrimination. In other words, individuals with higher scores on the aforementioned tests tended to have better rate discrimination scores, whereas those with lower cognitive scores exhibited poorer rate discrimination performance. This is in alignment with another study that measured gap detection in noise, wherein higher scores on speed of processing related to better gap detection across younger and older adults (Harris et al., 2010). These findings add to the evidence that age-related decline in cognitive abilities negatively affects performance on several temporal discrimination tasks, regardless of hearing status.

The lack of a relationship between perceptual and neural measures in the current study differs from the relationship found by Gaskins et al. (2019), who conducted a similar analysis for 80, 200, and 400 Hz rates. They found that ASSR SNRs at 400 Hz correlated with better (lower) DLs across groups; that is, age did not affect the relationship between the neurophysiologic response and behavioral performance. In the current study, however, the relationship between ASSR SNRs and perceptual DLs was not significant. The discrepant findings between studies could be due to differences in the analyses. In Gaskins et al., ASSR spectral energy was calculated on the average response, as opposed to ASSR PLF in the current study, in which the phase coherence between individual trials was calculated. It remains to be seen if a relationship exists between perceptual and neural measures of rate discrimination, and whether effects of aging and hearing loss emerge in future studies.

Gaskins et al. (2019) conducted a correlational analysis between speed of processing (using the Digit Symbol Coding test) and perceptual pulse-rate discrimination in YNH and ONH listeners. They found that speed of processing significantly predicted variance in perceptual performance for the 400-Hz rate regardless of age. This agrees with our results, despite the fact that we used the Pattern Comparison task (NIH Toolbox) for speed of processing, and Gaskins used the Digit Symbol Coding and Digit Symbol Search task (WAIS-III). The Gaskins et al. study did find that speed of processing correlated with the strength of ASSR responses at 200 and 400 Hz, though this could be spurious due to the small sample size. Harris et al. (2010) found that cortical-evoked potentials in response to gaps in noise were correlated with gap detection thresholds and speed of processing, consistent with the findings of Gaskins et al. (2019). These findings contrast with those in this study, where there was no effect of cognitive variables on ASSR phase locking at any rate tested.

Finally, we showed that lower scores on inhibitory control contributed to poorer perceptual rate discrimination across all listeners. In general, it is hypothesized that decreased inhibitory control with aging suppresses the ability to filter out unwanted or distracting information (Hasher and Zacks, 1988). Hasher et al. (1991) tested the extent and time-course of inhibition in older adults with a letter-naming selective attention task. They found that younger adults tended to exhibit inhibition, while older adults did not, leading to poorer performance on their task. While the current study was not designed as a selective attention task, it is possible that poorer inhibitory control in the older adults limited their ability to suppress the incorrect auditory stimuli presented during the listening task, thereby contributing to relatively poor performance, as in the Hasher studies.

In summary, our results agree with prior studies that evaluated cognitive factors in relation to temporal processing abilities in general [e.g., Füllgrabe et al. (2015)] and rate discrimination in particular (Gaskins et al., 2019). In contrast to previous studies, we did not find a relationship between ASSR measures and cognitive variables at any rate tested. The source of this disparity is currently unknown, but may be due to differences in cognitive tests, variation in EEG measures and/or statistical approaches. Further, studies that measured neural encoding using complex speech signals such as time-compressed speech scores [e.g., Dias et al. (2019)] may invoke increased cognitive resources, such as language abilities, that have a downstream effect on the auditory evoked potential (or vice versa).

V. CONCLUSIONS

We evaluated the effects of aging and hearing loss on perceptual measures of rate limitation and neural phase locking in YNH and ONH listeners, and OHI listeners with mild to moderate sensorineural hearing loss. We also investigated the potential contribution of cognitive factors to these measures. YNH listeners had better neural encoding of rate, perceptual rate discrimination, and cognitive scores compared to the ONH and OHI listeners, as expected. There were effects of aging, and not hearing loss in the older groups for the ASSR measure. Results were similar between YNH, ONH, and OHI groups for the perceptual rate discrimination task, except at 300 Hz, where a significant hearing loss effect was observed for the OHI group. There was no relationship between perceptual and neural measures. Finally, cognitive factors such as speed of processing and inhibitory control contributed significantly to the variance in perceptual scores for all groups, but not to the variance in neural measures arising from subcortical sources.

This study adds to the body of evidence that aging is a stronger contributor to poorer temporal processing than hearing loss. Future studies employing OHI listeners would benefit from increased group sample sizes, and perhaps a wider range of hearing loss severity. It may be that the listeners in this study with relatively mild hearing loss are able to compensate with higher level processes; indeed, we show that some cognitive factors significantly correlated with the behavioral data and appear to play an unclarified role in perceptual performance.