80 Hz auditory steady state responses (ASSR) elicited by silent gaps embedded within a broadband noise

Introduction Although auditory temporal processing plays an important role in speech comprehension, it cannot be measured by pure tone audiometry. Auditory temporal resolution is often assessed by behavioral gaps-in-noise test. To evaluate whether auditory temporal resolution could be objectively assessed, we measured the auditory steady state response (ASSR) elicited by silent gaps embedded within broadband noises at 80 Hz. Methods We prepared six sound types as test stimuli. One was a continuous broadband noise without a silent interval as a control stimulus and the others were broadband noises with 80 Hz silent intervals of 0.4, 0.8, 1.6, 3.1, and 6.3 ms. Results Significant ASSRs were recorded only when the gap length was longer than the behavioral thresholds and the ASSR amplitude increased as the gap length increased. Conclusion Eighty Hertz gap-evoked ASSR appears to reflect the neural activity related to the auditory gap processing and may be used as an objective measure of auditory temporal resolution in humans.


. Introduction
Time is a very important variable in hearing because all sounds vary in frequency and amplitude over time. Complex natural sound signals such as speech and music can be decomposed into slowly varying "envelope" and rapidly oscillating "fine structure." Out of the two, envelope plays a more important role in speech comprehension (1). Auditory temporal resolution is the ability to detect temporal changes of sound stimuli and to correctly recognize the envelope of sound signals. The temporal resolution of the auditory system is often assessed by the gap detection test. This test uses sound stimuli with silent gaps to estimate the minimum perceivable gap length (2,3). It should be noted that this method requires the participant to take an action in response to the sound signal whenever they perceived the gap. As such, this method is not objective since the participant's personality, varying levels of concentration and dexterity necessary for the physical action of pressing the button may affect the results to some extent. Therefore, many studies currently aim to find objective measures of the temporal resolution using neuroimaging techniques. Previous studies have shown that gaps embedded in continuous tones elicit various types of event related potentials. In an experiment by Uther et al. (4), continuous 2,000 Hz pure tones inserted with silent gaps of 3, 5, and 7 ms evoked mismatch negativity (MMN) and the MMN amplitude was proportional to the gap length. Electrophysiological studies have shown that the main sources of MMN are located close to the primary and secondary auditory cortex (5). Furthermore, Werner et al. (6) found that the behavioral gap detection threshold is similar to the neurophysiological gap detection threshold which was estimated by the sensitivity of the V wave of the auditory brainstem response elicited by silent gaps embedded within a broadband noise. The neural generator of V wave is thought to be the inferior colliculus (7). It remains elusive whether it is the cortex or the brainstem that plays a more important role in auditory gap detection.
In our previous study (8), we have obtained clear 40 Hz auditory steady state responses (ASSRs) elicited by the silent gaps of 3.125, 6.25, and 12.5 ms embedded within a broadband noise. The ASSR is one of the objective hearing tests used in clinical practice. Its neural sources and sensitivity depend on the modulation frequency. The 40 Hz ASSR appears to originate in the region spanning from the primary auditory cortex (9). It is considered to have a higher signal-to-noise ratio while the subject is awake (10). On the other hand, the 80 Hz ASSR appears to originate primarily from subcortical sources (11), and sleep has little effect on the response (12).
In the previous study, we demonstrated that gaps embedded within a broadband noise elicited clear 40 Hz ASSR; however, there has been no report that investigated the ASSR elicited by gap stimulus presented at the 80 Hz rate and compared the gap-evoked ASSR and the behavioral gap detection thresholds. The 80 Hz ASSR has the advantage of being an objective measure used in clinical practice because it can be measured reliably even during sleep (13). In this study, we conducted an experiment to verify whether the ASSR can be elicited by 80 Hz silent gaps of different lengths that are below and above the behavioral threshold. The results would contribute to developing a non-invasive objective measure of auditory temporal resolution in humans.

. . Participants
Twenty-one students (10 males) were recruited at the International University of Health and Welfare for this experiment. Their ages ranged from 18 to 31 (median 20). Eighteen participants were right-handed and three were left-handed, and all had normal hearing and no neurological or psychiatric disorders. They were fully informed about the study and gave written informed consent for their participation. The present study was approved by the Ethics Committee of the International University of Health and Welfare, School of Medicine and conformed to The Code of the World Medical Association (Declaration of Helsinki).

. . Stimuli and experimental design for electroencephalography recording
Six types of white noise of 1 min duration were used as sound stimuli (sampling rate: 48,000 Hz). One type was a white noise without a silent interval as a control stimulus (GAP_0), and the others were white noises with 80 Hz silent intervals of 0.396 (GAP_0.4), 0.792 (GAP_0.8), 1.563 (GAP_1.6), 3.125 (GAP_3.1), and 6.25 ms (GAP_6.3; Figure 1 and Supplementary Audios 1-6). The stimuli were randomly played for 30 min, resulting in 5 min of each GAP condition (1 min × 5 times × 6 noise types). All auditory stimuli were designed in the MATLAB environment (The MathWorks Inc., MA, USA). Participants were presented with stimuli at an intensity of 70 dBA SPL via ER-3A insert earphones (Etymotic Research Inc., IL, USA). Figure 2 displays the amplitude spectra of the sound stimuli measured at the earpiece by using Ear Simulator TYPE 4157 (Bruel & Kjaer Sound & Vibration Measurement A/S, Denmark). EEG recordings were performed with participants seated comfortably in a silent electromagnetically shielded room. They were instructed to watch a silent movie with captions during experiments.

. . Behavioral data
The detection threshold of 80 Hz gap inserted sound stimuli and single gap inserted stimuli were measured in all participants. An adaptive, three-alternative forced-choice, two-down, one-up procedure was used to track the 70.7% correct rate for gap detection threshold (GDT) determination as described in detail in the previous studies (14, 15). The threshold for the length of detectable gap was tested at an intensity of 70 dBA SPL via ER-3A insert earphones, same as in the ASSR measurements. The duration of the white noise (sampling rate: 48,000 Hz) was set to 500 ms. In the detection threshold condition for a single gap inserted stimulus, each test sound contains only one silent gap in the middle, whereas in the detection threshold condition for an 80 Hz gap inserted sound stimulus, each test sound contains 40 silent gaps because the gaps are embedded at 80 Hz. The inter-stimulus interval between two successive test sounds was 500 ms. The gap length started from 7 ms. The step size was set to 1 ms in the first four reversals and 0.5 ms thereafter. The measurements were continued for 12 reversals, and the threshold was estimated as the mean of the values for the last eight reversals. Thresholds were measured twice, and the mean of the two measurements was used as the detection threshold.

. . Data acquisition and analysis
Sound stimuli were presented via Multi Trigger System Ver.2 (MTS0410, Medical Try System, Co., Ltd., Japan), which simultaneously sent triggers to Neurofax EEG1200 (Nihon Koden, Co., Ltd., Japan). We used six types of triggers (GAP_0, GAP_0.4, GAP_0.8, GAP_1.6, GAP_3.1, and GAP_6.3) and, apart from the control condition (GAP_0), the triggers were synchronized with the gap offset (or the onset of the noise segment). The EEG signals were recorded using a Neurofax EEG1200 system at a sampling   rate of 1,000 Hz. The recording electrodes (Ag/AgCl) were located at Cz according to the international 10-20 system. The averaged signal of two electrodes placed on both mastoids was used as a reference electrode and the ground electrode was located around the forehead midpoint. Electrode impedance was maintained below 15 kΩ. Recorded EEG data were exported as ASCII files and were analyzed offline using Matlab R2020a and EEGLAB (16). For EEG waveforms, a fast Fourier transform (FFT) was computed in each condition and amplitude spectra were extracted after removing the power line fluctuations at 50 Hz using the Clean-Line plugin for EEGLAB. In order to obtain ASSR, an epoching procedure was applied to the EEG signals. While the 80 Hz cycle necessitates the triggers to be placed at 12.5 ms intervals, this was difficult due to 1,000 Hz sampling rate. This problem was resolved by placing a marker every 25 ms in a 40 Hz cycle and calculating two cycles as one epoch. One minute sound stimulus contained 2,400 epochs, and for each GAP condition, the sound stimulus was presented 5 times, resulting in 12,000 epochs. A total of 60,000 epochs were labeled in all GAP conditions of each participant. EEG waveforms were bandpass filtered (70-90 Hz) offline and epochs of 0-24 ms (25 sampling points) from the markers were separately averaged after artifact rejection (set to a . /fneur. . threshold of ± 20 µV) for each GAP condition (GAP_0, GAP_0.4, GAP_0.8, GAP_1.6, GAP_3.1, and GAP_6.3) for each participant. The obtained ASSR waveforms were fitted into the 80 Hz sinusoidal curves and the amplitudes were used for the statistical analysis. Next, we calculated component synchrony measure (CSM) (17) and estimated GDT from the CSM obtained at Cz. The acquired EEG was filtered by a bandpass filter with cutoff frequencies of 79 and 81 Hz. Then the filtered EEG was divided into 600 epochs of 500 ms length and grouped into 10 groups, each containing 60 epochs. We obtained ten averaged waveforms based on those 60 epochs and the CSM is calculated using the following formula: Where, ψ denotes phase of ten averaged waveforms (k = 1, 2, 3, . . . 10) and m denotes frequency. CSM value varies from 0 to 1. The value is equal to 1 if the phases of epochs are the same and approaches 0 when the phases change randomly. The criterion for the presence of response is set at M + 3SD (= 0.385), where M denotes the mean of CSM value for non-response and SD denotes standard deviation. The obtained CSM functions as silent interval lengths were approximated by a sigmoid function, and the silent interval length whose obtained sigmoid function exceeded 0.385 was defined as GDT estimated from the CSM values.
The relationships between GDT estimated from the CSM values, single gap behavioral GDT, and 80 Hz gap behavioral GDT were evaluated based on correlation analysis.

. Results
After the measurements were obtained from all participants, the collected data were analyzed. The average of the behavioral data of single gap GDT was 2.76 and the 95% confidence interval obtained by boot-strap resampling tests (iteration = 100,000) was 2.60-2.92. The average of the behavioral 80 Hz gap GDT was 0.994 and the 95% confidence interval obtained by boot-strap resampling tests (iteration = 100,000) was 0.902-1.101. For the EEG analysis, the mean rejection rate of the obtained epochs was 6.6%. Figure 3 shows grand averaged FFT waveforms under each condition. No clear response at 80 Hz was observed under GAP_0, GAP_0.4, and GAP_0.8 conditions in contrast to GAP_1.6, GAP_3.1, and GAP_6.3 conditions, where prominent responses at 80 Hz were observed. Figure 4 shows grand averaged EEG waveforms under each condition. Similar to the FFT results, no significant ASSR was evoked in the GAP_0, GAP_0.4, and GAP_0.8 conditions. In the GAP_1.6, GAP_3.1, and GAP_6.3 conditions, clear EEG was evoked and the amplitude increased as the gap length increased. Figure 5 shows the mean normalized ASSR amplitude in each GAP condition together with the corresponding 95% confidence intervals obtained by boot-strap resampling tests (iteration = 100 000). A one-way ANOVA applied to the normalized ASSR amplitude revealed a significant main effect for gap length [F (5,120) = 299.66, p < 0.0001]. As shown in Table 1 Figure 6 shows the mean CSMs calculated from EEG in each GAP condition together with the corresponding 95% confidence intervals obtained by boot-strap resampling tests (iteration = 100 000). For the results in GAP_1.6, GAP_3.1, and GAP_6.25, the mean CSM gradually increased as GAP length increased. A oneway ANOVA applied to the CSM revealed a significant main effect for gap length [F (5, 120) = 41.33, p < 0.0001]. As shown in Table 1, the post-hoc multiple comparison revealed significant differences between GAP_0 and GAP_1.6 (p = 0.011), GAP_0 and GAP_3.1

. Discussion
The results of the present study demonstrated that 80 Hz ASSR can be elicited by silent gaps embedded within a broadband noise in people with normal hearing. Significant ASSRs were elicited only by test sound stimuli with silent intervals longer than 1 ms, which was the gap detection threshold derived from behavioral data. The ASSRs elicited by GAP_0.4 and GAP_0.8, which were below the behavioral threshold, were similar to those elicited in the GAP_0 condition, in which a continuous broadband noise was used as a sound stimulus. These results suggest that

Latency [ms]
ASSR time-locked to the Gap-offset
significant ASSRs were elicited only when the length of the silent gaps exceeded the behavioral threshold. This implies that auditory temporal resolution may be objectively measured using the 80 Hz gap-evoked ASSR.
It is known that stimuli differing in modulation frequencies elicit ASSR originating in different brain areas. Previous studies compared cortical and subcortical neural activity with 40   hemisphere infarction and normal subjects. All participants had normal pure tone audiograms; however, the results indicated that auditory temporal processing was impaired when the GIN test was performed on the ear contralateral to the infarction site. Lavasani et al. (24) found that the GDT became longer in patients with temporal epilepsy with normal pure tone audiograms, indicating that auditory temporal processing was impaired in those people. Moreover, research done by Bamiou et al. (25) suggested that insula plays an important role in auditory temporal processing, since patients with insular hemispheric infarction showed prolonged GDT.
In the present study, we used gap sounds that are thought to be processed in the auditory cortex, yet we obtained clear 80 Hz ASSR which are mostly associated with the brainstem. The results obtained could be interpreted as follows.  the 80 Hz ASSR is generally believed to originate primarily from the brainstem, the signal appears to be contaminated with the neural activity in the auditory cortex (18,26). Therefore, the results obtained in the present study might reflect a portion of 80 Hz ASSR derived from the auditory cortex. Alternatively, the present results could imply that the auditory temporal processing involved in the gap detection occurs at least partially in the brainstem. Galambos et al. (10) first reported that the ASSR amplitude became maximal at the modulation rate of 40 Hz. The 40 Hz ASSR was found to become smaller with sleep, sedation, and anesthesia (27). On the other hand, a higher modulation frequency (70-110 Hz) provided a stable ASSR even during sleep or under sedation (13, 28). The ASSRs elicited by 80 Hz amplitude modulated tones are clinically used as an objective audiometry and are often measured during sleep since the signal-to-noise ratio of 80 Hz ASSR improves during sleep (29). In the present study, the participants were awake during the EEG recording; however, the 80 Hz gap-evoked ASSR may become more prominent during sleep. Eighty hertz ASSR recording during sleep would be especially useful for infants and children who could not stay still during the EEG recording. Interestingly, the detection threshold of 80 Hz gap inserted sound stimuli was significantly shorter than the normal single gap GDT. While the single gap GDT (mean: 2.76 ms) in the present study was similar to those obtained in the previous studies (30,31), the regular insertion of gaps at 80 Hz appears to shorten the GDT. Bacon et al. (32) measured detection thresholds of sinusoidal amplitude modulation on a broadband noise and found that normal hearing participants could detect up to 1,024 Hz modulation frequency. The results indicated that the participants could detect the envelope fluctuations of about 1 ms. Moreover, Ross and Pantev (33) measured behavioral GDT and the auditory evoked fields elicited by gaps embedded within 500 Hz tones with 40 Hz amplitude modulation. Normally, the gap detection of pure tones with frequencies between 400 and 1,000 Hz was estimated to be between 6 and 8 ms (34); however, Ross and Pantev's (33) results showed that the detection threshold for the gap embedded within the 40 Hz AM tone was 3 ms and the gap duration of 3 ms elicited significant auditory evoked fields. Similar to the above results, the present study also demonstrated that the detection thresholds for gaps embedded within sounds with repetitive envelope fluctuations became shorter than those embedded within continuous sounds with no repetitive fluctuation (35). One might argue that the insertion of periodic gaps added spectral components corresponding to the stimulation rate (80 Hz) and its harmonics, and consequently the participants might have detected the corresponding spectral components. In the present study, we used white noise segments longer than and equal to the half (6.25 ms) of one 80 Hz cycle and thus the 80 Hz gap inserted sound stimuli had spectral components similar to the control white noise as shown in Figure 2. Another possibility is that the participants might have detected the spectral splatter caused by the steep onset and offset sound envelopes of the gap. However, we used white noise as sound stimuli and the onset and offset of white noise did not give a spectral cue since the spectral splatter was masked by the white noise. Therefore, it is less likely that the participants in the present study detected the spectral changes of test sound stimuli instead of temporal ones.
We found no significant relationship between the GDT estimated from the CSM and the behavioral GDT. All the participants in the present study were young adults and had normal hearing. This could explain why there was little variance in the behavioral GDTs among the participants. Moreover, alertness and motivation of the participants appears to have a stronger impact on the behavioral results than the interindividual difference. It is necessary to conduct similar studies on people with suspected deterioration of auditory temporal resolution, such as auditory neuropathy patients (36,37) and the elderly (38).
Recent studies focused their attention on people who have normal pure tone audiogram but struggle to listen to speech signals especially in noisy environments (39,40). This symptom may derive from impaired temporal processing in the auditory system (41). It remains elusive whether speech perception is significantly correlated with the within-channel gap detection threshold (42). According to Tyler et al. (43) and Snell et al. (44) who used noise bursts as stimulus sounds, GDT and speech perception under noise are significantly correlated, while Strouse et al. (45) and Snell et al. (46) found no significant correlation. Although most of those results were obtained in crosssectional studies, a longitudinal study by Babkoff and Fostick (38) showed a significant relationship between temporal processing and speech perception even after corrections for auditory level and cognitive ability.
Therefore, the gap-evoked ASSR obtained in the present study may reflect the auditory temporal processing and speech comprehension ability. Many ailments thought to be related to impaired auditory temporal processing, such as hidden hearing loss, auditory processing disorder, and other listening difficulties, are currently diagnosed mainly on the basis of the patients' subjective symptoms. The gap-evoked ASSR can be introduced as an objective diagnostic measure for such cases. Moreover, objective measure of auditory temporal processing can be used as a screening test for children with language and speech developmental delays, allowing early detection and early therapeutic intervention for them. Additionally, in cases of elderly patients with communication problems, measurement of gap-evoked ASSR could potentially help discriminate between patients suffering from dementia and those with impaired auditory temporal processing.

. Conclusions
Significant ASSRs and CSMs were elicited only by test stimuli with gap lengths above the behavioral threshold. The test stimuli with gap lengths below the behavioral threshold elicited ASSRs similar to those elicited by the continuous broadband noise. Eighty Hertz gap-evoked ASSR may provide insights into the neural mechanisms of auditory temporal processing and may be applied to objectively and non-invasively measure auditory temporal resolution in humans.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of the International University of Health and Welfare, School of Medicine. The patients/participants provided their written informed consent to participate in this study.