The effects of broadband elicitor duration on a psychoacoustic measure of cochlear gain reduction

Abstract

Physiological and psychoacoustic studies of the medial olivocochlear reflex (MOCR) in humans have often relied on long duration elicitors (>100 ms). This is largely due to previous research using otoacoustic emissions (OAEs) that found multiple MOCR time constants, including time constants in the 100s of milliseconds, when elicited by broadband noise. However, the effect of the duration of a broadband noise elicitor on similar psychoacoustic tasks is currently unknown. The current study measured the effects of ipsilateral broadband noise elicitor duration on psychoacoustic gain reduction estimated from a forward-masking paradigm. Analysis showed that both masker type and elicitor duration were significant main effects, but no interaction was found. Gain reduction time constants were ∼46 ms for the masker present condition and ∼78 ms for the masker absent condition (ranging from ∼29 to 172 ms), both similar to the fast time constants reported in the OAE literature (70–100 ms). Maximum gain reduction was seen for elicitor durations of ∼200 ms. This is longer than the 50-ms duration which was found to produce maximum gain reduction with a tonal on-frequency elicitor. Future studies of gain reduction may use 150–200 ms broadband elicitors to maximally or near-maximally stimulate the MOCR.

I. INTRODUCTION

The human auditory system can adjust to different sound environments, from quiet to noisy backgrounds. One system that may aid in the dynamic adjustment to sound is the medial olivocochlear reflex (MOCR), a bilateral sound-activated feedback loop in the brainstem, with fibers that project to both cochleae and synapse directly onto outer hair cells (OHCs) (Guinan, 1996). The MOCR decreases the gain produced by the OHCs in a frequency-specific manner relative to an elicitor sound, thereby decreasing the basilar membrane (BM) movement (Murugasu and Russell, 1996; Cooper and Guinan, 2006a). It has been posited that gain reduction via the MOCR can shift the dynamic range in everyday listening conditions. This is supported by the fact that MOCR activation can enhance auditory nerve responses to transient sound in a noisy background in animals (Winslow and Sachs, 1988; Kawase et al., 1993) or increase sensitivity to changes in intensity in auditory perception (Almishaal et al., 2017; Strickland et al., 2018). It may enhance the fluctuation profile for complex sounds (Carney, 2018) and has been implicated in supporting speech intelligibility in noise (Giraud et al., 1997). Despite this growing body of neurophysiological and psychoacoustic evidence that adjustment of cochlear gain may be beneficial for auditory perception, there are aspects of MOCR gain reduction that are not fully understood.

One aspect of MOCR gain reduction that has drawn significant interest from researchers is the time-course of activation and decay. For example, because cochlear gain reduction has been theorized to translate to an improvement in speech perception in noise (Giraud et al., 1997; Brown et al., 2010; Clark et al., 2012), it would be generally useful to know how fast gain adjustment by the MOCR would need to be in order to be beneficial. This information may have potential translational utility in designing hearing aid (Jürgens et al., 2016) and cochlear implant algorithms (Lopez-Poveda et al., 2016), as there is modelling data that supports an improvement of speech intelligibility in noise with efferent-inspired contrast between the signal and noise (Yasin et al., 2018; Liu and Demosthenous, 2020; Yasin et al., 2020). Another reason to study the time course of gain reduction is that it can provide insight about the appropriate stimuli for studying the MOCR physiologically and psychoacoustically, in addition to a general understanding of the system. The time course of gain reduction has not been studied as extensively psychoacoustically as it has using physiological measures [neural responses and otoacoustic emissions (OAEs)]. The pertinent psychoacoustic and physiological literature on the time course of MOCR gain reduction will be reviewed here.

Early psychoacoustic evidence that would be consistent with gain reduction by sound came from a phenomenon called overshoot (Zwicker, 1965), also known as the temporal effect (Hicks and Bacon, 1992), and its time course has been measured. In the temporal effect, a signal presented at the onset of a masker is more detectible if it is preceded by an additional sound. This additional sound can either be an extension of the masker or a separate sound, which has been referred to as a precursor or an elicitor. The term elicitor will be used in this study. Therefore, the temporal effect is the difference in signal (or masker) threshold when the signal is at the onset of the masker (i.e., no elicitor present; the reference condition), and the threshold for when the signal and masker follow an elicitor (i.e., elicitor present condition). The time course of the buildup of the temporal effect is the decrease in signal threshold (or increase in masker level) as the signal's onset is delayed relative to the onset of the elicitor. The buildup of the temporal effect plateaus at around 200 ms (Zwicker, 1965; Bacon and Moore, 1986; McFadden et al., 2010). Another study estimated the time constant of the temporal effect both psychoacoustically and with a modified stimulus-frequency OAE (SFOAE) paradigm using very similar stimuli in the two measurements (Walsh et al., 2010b). They found that the average time constant measured psychoacoustically (64.8 ms) was very similar to the average time constant measured from the SFOAEs (72 ms), which supports the idea that the temporal effect may be mediated by the MOCR. One thing to note about the temporal effect in simultaneous masking paradigm is that the results may include the effects of two-tone suppression. Two-tone suppression also reduces gain, but on a much faster time scale (nearly instantaneous; Kiang et al., 1965) compared to the relatively sluggish MOCR. This complicates interpretation of the results as there may be interactions between the two mechanisms (Hegland and Strickland, 2018).

The time course of gain reduction has also been studied psychoacoustically with forward masking paradigms, which avoids the possibility of two-tone suppression. The general methodology of gain reduction using forward masking will be described first. Most studies have used short duration tonal signals and maskers to measure functions hypothesized to reflect cochlear processing without the influence of gain reduction. As will be discussed below, studies using OAEs have shown that there is a 20–25 ms delay between the onset (or offset) of an elicitor and the onset (or offset) of the MOCR. These functions may then be measured with an elicitor before the signal and masker. If the masker frequency is approximately an octave below the signal frequency, the growth of masking is hypothesized to reflect the cochlear input-output function (Oxenham and Plack, 1997). Using this paradigm, several experiments have shown that an ipsilateral elicitor shifts the linear lower-input sound level portion of the input-output function to higher signal levels, consistent with a decrease in gain (Krull and Strickland, 2008; Jennings et al., 2009; Roverud and Strickland, 2010; Jennings and Strickland, 2012; Yasin et al., 2014; DeRoy Milvae and Strickland, 2018). These forward masking techniques rely on using maskers at the signal frequency (on-frequency) and maskers approximately an octave below the signal frequency (off-frequency). Under the classic power spectrum model of masking, it is assumed that the listener attends to the auditory filter with the best signal-to-masker ratio, which will typically be at or near the signal frequency (Fletcher, 1940). Gain reduction at the signal frequency place is expected for both the signal and the on-frequency masker, but not the off-frequency masker. Therefore, the change in signal threshold with an off-frequency masker following an elicitor can provide an estimate of gain reduction. This differential processing between on- and off-frequency maskers is the basis for studying behavioral gain reduction (Yasin et al., 2014; DeRoy Milvae and Strickland, 2018; Salloom and Strickland, 2021). It is not consistent with other mechanisms such as temporal integration of the masker and elicitor, which has been referred to as additivity of masking [e.g., Penner and Shiffrin (1980)], where an elicitor would be expected to produce the same signal threshold shift for an on-frequency and off-frequency masker. If multiple masker frequencies are used to trace out a psychoacoustic tuning curve, adding an elicitor decreases frequency selectivity (Jennings et al., 2009; Jennings and Strickland, 2012), which is also consistent with a reduction in gain. Schematics based on the power-spectrum model of masking using input-output functions at the signal frequency place for the gain reduction conditions described in this section can be found in Fig. 1. Specifically, Fig. 1 outlines how gain reduction is measured on the linear low-input sound level portion of the cochlear input-output function, how the effect of the elicitor differentially affects signal threshold for the on- and off-frequency masked conditions, and how the elicitor affects signal threshold when no masker is present.

FIG. 1.

(Color online) Schematic of gain reduction effects on the signal and masker for each listening condition depicted with cochlear input-output functions. Baseline conditions are the panels to the left, with the corresponding elicitor present conditions in the panels to the right. The solid lines in these schematics represent the responses to the signal (denoted bold “S”) within a filter at or near the signal frequency. Responses to the signal with gain reduction are indicated by the dashed line directly underneath the solid line. The double-headed (yellow) arrow on the y axis shows the threshold signal-to-masker ratio, which is assumed to be constant. Off-frequency conditions are in panels (A) and (B), on-frequency conditions are in panels (C) and (D), and masker absent conditions are in (E) and (F). The gray dashed line in panels (E) and (F) indicates absolute threshold.

Several forward masking studies have estimated gain reduction using tonal elicitors of various durations in order to generate a time constant, at a signal frequency of 4 kHz. Roverud and Strickland (2010) measured the time course of forward masking gain reduction by manipulating the duration of the on-frequency elicitor and the delay between elicitor offset and masker onset. The elicitor duration ranged from 5 to 100 ms. For some subjects, gain reduction increased with elicitor duration up to about 50 ms, but then decreased or rolled over with a 100-ms elicitor. These findings are consistent with findings from Krull and Strickland (2008), which showed that for some subjects, an on-frequency elicitor with a 40-ms duration resulted in a larger reduction in gain than a 160-ms duration elicitor of the same intensity. In a follow-up study to Roverud and Strickland (2010), Roverud and Strickland (2014) used a psychoacoustic measure of gain reduction and found differential effects of duration for on- vs, off-frequency tonal elicitors. This study used a wider range of 10 to 150 ms and finer steps (20–30 ms) between elicitor durations. For the on-frequency elicitor, thresholds increased with increasing duration up to about 50 ms, and then plateaued or oscillated. In contrast, thresholds with off-frequency elicitors continued to increase with elicitor duration. Time constants were fitted to the data as part of a model that also included a temporal integration window and gain reduction. The time constants ranged from approximately 28 ms to approximately 76 ms. These results are consistent with cochlear gain reduction, possibly by the MOCR, in which the on-frequency elicitor is affected by gain reduction at the signal frequency place, but the off-frequency elicitor is not (Roverud and Strickland, 2014). One psychoacoustic study examined the effects of broadband elicitor duration on forward masking (Oxenham and Plack, 2000). In one condition there was a 20-ms gap between masker offset and signal offset, which is identical to the masker duration in the Roverud and Strickland (2014) study. For a masker set a 0-dB spectrum level (approximately 40 dB SPL), the maximum masking was for durations of 30 ms or less (i.e., nonmonotonic). For a masker set at 40 dB spectrum level (approximately 80 dB SPL), two listeners showed the same masking for 30 ms as for 200 ms maskers (i.e., flat), and two showed an increase with a 200-ms masker. Durations between 30 and 200 ms were not tested, so it is not clear whether the optimum duration is closer to the duration found with the on-frequency tonal masker or the off-frequency tonal masker in the Roverud and Strickland (2014) study. Previous research has shown that for this psychoacoustic forward masking paradigm, the effects of an elicitor are consistent with the elicitor being processed through a filter centered at or near the signal frequency place. Tuning for elicitor frequency is slightly broader than for a short forward masker (Jennings and Strickland, 2010). Growth of gain reduction with signal level is compressive for tonal elicitors at the signal frequency, but approximately linear with a slope of one for tonal elicitors well below the signal frequency (Roverud and Strickland, 2014; DeRoy Milvae and Strickland, 2021). For most listeners, noise is about as effective as an on-frequency tone, if the level of the noise is calculated as the level passing through a filter centered at the signal frequency (DeRoy Milvae and Strickland, 2021), when the elicitor is 50 ms in duration. If a noise elicitor is increased in duration, the decrease in gain will turn down components near the signal frequency, but will also widen the filter. In addition, self-suppression within the noise may change with duration [see, e.g., Weber (1978) for evidence of suppression as a noise band is widened]. Therefore, the most effective duration of broadband noise might be longer than for an on-frequency tone. This also suggests that the time-course of MOCR gain reduction may vary depending on the frequency content of the elicitor and its relationship to the signal frequency.

As stated earlier, the time course of the MOCR has been tested most extensively using OAE paradigms. OAEs are sounds that are produced by the amplification of the OHCs and can be recorded in the ear canal (Guinan, 2006). Generally, two OAE methods have been used to measure the time course of the MOCR: changes in OAE level with an elicitor present, and OAE adaptation paradigms. OAE adaptation may reflect the MOCR response if the stimuli used to elicit the OAE also elicit the MOCR (Liberman et al., 1996). Liberman et al. (1996) measured adaptation of the distortion product OAEs (DPOAEs; 2f₁-f₂) in anesthetized cats by presenting primary tones (f₁ and f₂) either ipsilaterally or bilaterally, as well an ipsilateral primary tone presentation with a contralateral broadband elicitor present. They found that the amplitude of the DPOAE decreased after the onset of the primary tones over two different time scales: a more “rapid” component with a time constant of 60–100 ms, and a second “slower” component time constant of approximately 1000 ms. Time constants were estimated by fitting data with a two-exponential function. While the effects of monoaural versus bilateral stimulus presentation on the DPOAE differed in terms of magnitude, the time constants were similar. Importantly, the time course of the rapid adaptation component of the DPOAE disappeared after the crossed (i.e., ipsilateral) pathway was severed, indicating that the rapid adaptation effect is likely caused by ipsilateral MOCR activation by the primary tones. Interestingly, severing the crossed pathway did not affect the slower adaptation, suggesting that the rapid and slower time constants may have different mechanisms. This finding is consistent with other work measuring sound-evoked BM responses with electrical stimulation of MOCR fibers in anesthetized guinea pig, that found responses with short time constants (∼50 ms) and much longer time constants (∼10 s) with MOCR activation, and that these BM responses differed in their phases at characteristic frequency (CF), which suggests two separate forms of MOCR inhibition (Cooper and Guinan, 2003, 2006b). Overall, Liberman and colleagues' work in cat established two OAE-based methods to estimate efferent time constants: evoking the MOCR ipsilaterally with the primary tones, and contralaterally with noise. These methods have been replicated in multiple studies investigating MOCR function in humans.

Kim et al., (2001) measured the rapid adaptation of the DPOAE amplitude with time, fitted these data with a two-exponential function and found a faster median time constant of 69–70 ms (10–330 ms), and a slower median time constant of 1.5 s (350 ms–5.5 s). Bassim et al. (2003), in a follow-up study to Kim et al., found very similar time constants with their DPOAE adaptation experiments using ipsilateral and bilateral primary tone presentations, and ipsilateral primary tone presentation with contralateral broadband noise elicitation on the DPOAE amplitude. Time constants were estimated from a two-exponential function: their faster time constant was 73 ms (7–350 ms), and a slower median time constant of 2 s (350 ms–8 s). James et al., (2005) measured the change in DPOAE magnitude over time with contralateral broadband elicitation and found that the time constant of the buildup of the MOCR is on the order of 100 ms. Backus and Guinan (2006) measured MOCR time constants by the change of the stimulus-frequency OAEs (SFOAEs) with elicitor duration and found that gain reduction is characterized by three time constants: a faster time constant of ∼60–80 ms, a medium time constant of 290–350 ms, and a slower time constant of 10 s. A recent study using click-evoked OAEs (CEOAEs) estimated MOCR time constants with a paradigm that only uses presentations of click trains, which serve both as the probe and the elicitor (Boothalingam et al., 2021). In that study, they measured the change in the CEOAE level as a function of time, and the data were fitted using a two-term exponential function, similar to those used in the DPOAE adaptation studies (Liberman et al., 1996; Kim et al., 2001; Bassim et al., 2003). They found a faster time constant of approximately ∼210 ms and a slower time constant of 17–24 s. The authors note that estimating their faster time constant (∼210 ms) with only one function may have resulted in a time constant that combined both the fast and medium time constants that were found in Backus and Guinan (2006).

Studies using analysis windows with fine temporal resolution have shown that the MOCR also has a relatively sluggish delay of 25-ms from the onset/offset of the elicitor to the onset/offset of gain reduction (James et al., 2005; Backus and Guinan, 2006). This has not been accounted for in some studies due to the time window used for analysis [for details see Walsh et al. (2010a)]. It should be noted that the studies using sound elicitors [i.e., James et al. (2005) and Backus and Guinan (2006)] used broadband noise to evoke the MOCR, as broadband noise stimuli have been found to be particularly strong elicitors of cochlear gain reduction (Maison et al., 2000; Lilaonitkul and Guinan, 2009a; Wicher and Moore, 2014; DeRoy Milvae and Strickland, 2021). In summary, the literature of OAE-based measures of MOCR gain reduction are generally in good agreement which show that there are multiple time constants, relatively short time constants 60–100 ms for the fast effects, and there is also a much longer time constant for the slow effects of gain reduction. To reiterate, while the time course of gain reduction has been studied to some extent in humans using OAE-based paradigms, the time course of gain reduction in behavior, specifically using broadband elicitors has not been investigated as well.

The current study was designed to be an extension of previous work by Roverud and Strickland (2010, 2014) using broadband elicitors. As stated earlier, broadband elicitors are typically used in OAE-based paradigms to estimate MOCR strength because they are powerful elicitors, and therefore have also become commonly used in psychoacoustic measures of gain reduction. Despite this, the time constants of gain reduction have not been directly measured using broadband elicitors in a psychoacoustic forward masking paradigm. In previous studies from this laboratory using broadband elicitors, a duration of 50 ms has been used, based on the on-frequency tonal elicitor results from Roverud and Strickland (2014), but it is not clear if this is the most effective duration for a broadband elicitor. In the current study, we used forward masking paradigms established in previous studies [e.g., DeRoy Milvae and Strickland (2018) and Salloom and Strickland (2021)]. We measured gain reduction at 4 kHz, the same frequency used by Roverud and Strickland (2010, 2014), and our current study builds on similar experiments they had conducted using tonal elicitors. We used a wider range of elicitor durations (50–800 ms) than those used in the Roverud and Strickland (2010, 2014) studies, which should overlap with the entirety of the buildup of the MOCR (James et al., 2005; Backus and Guinan, 2006).

II. METHODS

The psychoacoustic experiment used forward masking paradigms to measure cochlear gain reduction at 4 kHz as a function of ipsilateral broadband elicitor duration. Data were analyzed in terms of overall magnitude of the signal threshold shift between the conditions where the elicitor was and was not present, and the corresponding time constants.

A. Subjects

1. Audiological testing

Nine subjects (4 male and 5 female) completed the experiments in the current study. Their ages ranged from 19 to 31 years (median = 24 years) at the time of testing. All subjects had normal auditory function, determined through a battery of audiologic measures. All subjects had clinically normal pure tone thresholds [≤15 dB hearing level (HL) at audiometric frequencies between 250 and 8000 Hz]. Distortion product otoacoustic emissions (DPOAEs) were present (Bio-logic system, Natus Medical, Inc., Pleasanton, CA) from 1500 to 8500 Hz (minimum criteria of –6 dB SPL distortion product and 6 dB SNR for 10 of 12 frequencies tested with no consecutive absent responses). Tympanograms (Tympstar, Grason-Stadler, Inc.) were normal (type A), indicating normal middle-ear function. Ipsilateral middle-ear muscle reflex (MEMR) thresholds were measured using white broadband noise elicitors and 226 Hz probe tones. Because the signal was always presented in the subject's right ear in the experiments of the current study, the clinical MEMR thresholds were measured with respect to the probe in the right ear. The clinical MEMR thresholds were measured in dB HL and converted to corresponding dB SPL units for fair comparison to the experimental elicitors used in the current study. To do so, noise levels from the immittance equipment were recorded from a sound level meter attached to a Zwislocki coupler mounted in a KEMAR ear. The noise levels in dB SPL were approximately 8 dB higher than the nominal levels in dB HL. Overall, no subject's MEMR threshold to broadband noise was below 68 dB SPL, a level that is significantly higher than the elicitor level used in the experiments (50 dB SPL). Additionally, a wideband acoustic immittance (WAI) procedure was conducted to estimate each subject's ipsilateral MEMR threshold to further validate that the stimuli used in the current study did not evoke the MEMR in the gain reduction experiments. The procedure and stimuli used closely followed those used in Bharadwaj et al. (2022). The results from the WAI MEMR measures corroborated well with the clinical MEMR thresholds in that no subject's WAI MEMR threshold was below 50 dB forward pressure level (FPL). The procedure for the WAI MEMR measures, as well as each subject's clinical MEMR reflex thresholds and WAI MEMR thresholds can be found in the Appendix. Taken together, we can confidently conclude that the results from the gain reduction experiments in this study exclude any effects from the MEMR.

An additional two potential subjects started the study but did not complete it. One of these potential subjects had tinnitus, which would have been confounding in our tone detection tasks, as subjects are required to detect quiet sounds in the presence of an elicitor and/or masker. The other potential subject could not consistently perform the behavioral tasks and had issues with differentiating the signal from the other sounds, and was thereby discontinued from the study after multiple days of unsuccessful practice sessions. Data from these additional subjects is not reported here. All subjects were paid for their time in the study except for S1, who is the first author. Other subjects were recruited via fliers on the Purdue campus. All research was conducted under a research protocol approved by the Institutional Review Board at Purdue University to safeguard the rights, safety, and well-being of our subjects.

B. Psychoacoustic measures of gain reduction

1. Stimuli

Estimates of gain reduction were made at 4 kHz using two forward masking techniques that rely on the timing of cochlear gain reduction via the MOCR. The technique used to measure gain reduction with the use of short duration maskers (“masker present” conditions) will be explained first. The 4-kHz signals used in the following behavioral experiments were 10-ms sinusoids, including 5-ms cos² onset and offset ramps. This signal duration is longer than the 6 or 8 ms used in some previous studies [e.g., Jennings et al. (2009), Roverud and Strickland (2010), and DeRoy Milvae and Strickland (2018)] to ensure that the spectral spread was within one auditory filter bandwidth (DeRoy Milvae and Strickland, 2018). Threshold was measured for the signal in quiet. Next, the masker levels needed to mask a signal fixed at 5 dB sensation level (SL) were determined. A 5-dB shift in signal threshold was desired so that the signal was fixed on the lower leg (the linear low-input sound level portion) of the cochlear input-output function. Gain reduction is largest in this region of the input-output function (physiologically: Cooper and Guinan, 2006b; psychoacoustically: Krull and Strickland, 2008; Roverud and Strickland, 2010). The masker duration was 20 ms (including 5-ms cos² ramps), which should be too short to elicit the MOCR during the signal presentation (James et al., 2005; Backus and Guinan, 2006).

For gain reduction measurements, the maskers were on-frequency (4 kHz) and off-frequency (0.6 times the signal frequency, 2.4 kHz) sinusoids. The signal level was set at 5 dB SL, and the masker level was adjusted to find the level needed to just mask the signal. By measuring both on- and off-frequency masked threshold, we were able to test whether the effects of the elicitor on signal threshold were consistent with cochlear gain reduction (explained below). To verify that the maskers were equally effective at masking the signal, the on- and off-frequency maskers were fixed at the thresholds determined earlier, and the signal was varied to check that the signal thresholds were raised to 5 dB SL. On- and off-frequency masked signal thresholds were considered similar if the difference between the two conditions was less than 3 dB. If signal threshold with the fixed maskers differed by more than this amount, the masker level was adjusted, and the signal threshold was remeasured until signal thresholds were within 3 dB of one another. Schematics for these conditions are shown in Fig. 2(A) (off-frequency) and Fig. 2(C) (on-frequency). Signal thresholds in this reference condition were then compared to a condition where an elicitor, intended to elicit the MOCR, preceded both the masker and the signal [Fig. 2(B)—off-frequency and Fig. 2(D)—on-frequency, respectively]. The elicitor was a 50-dB SPL pink broadband noise (0.25–10 kHz) which was fixed in duration [50, 65, 100, 200, 400, 800 ms, including 5-ms cos² onset and offset ramps]. The elicitor durations used in the current study cover a broader range than those used in the Roverud and Strickland (2014) study, and should overlap with the entirety of the buildup of the MOCR (James et al., 2005; Backus and Guinan, 2006). Pink noise has a spectrum level that decreases by 3 dB per octave. This elicitor provides a more accurate comparison of gain reduction across frequency, as pink noise will excite auditory filters across the frequency range with approximately equal energy. Additionally, previous studies have found broadband noise stimuli to be particularly effective elicitors of cochlear gain reduction (Maison et al., 2000; Lilaonitkul and Guinan, 2009a; Wicher and Moore, 2014). The pink noise was presented ipsilaterally with respect to the masker and signal. Ipsilateral elicitors have shown to produce significantly larger cochlear gain reduction compared to contralateral elicitors of the same sound pressure level when measured psychoacoustically (Salloom and Strickland, 2021). 50 dB SPL is a relatively low elicitor level compared to many other studies in humans measuring MOCR effects, which typically use 60-dB SPL elicitors. A previous study from this lab (Salloom and Strickland, 2021) found that a 50-dB SPL elicitor was as effective as a 60-dB SPL elicitor at 4 kHz, and it is less likely to elicit the MEMR.

FIG. 2.

(Color online) Schematics of the stimuli used for the masker present method [(A) and (B), off-frequency; (C) and (D), on-frequency] and the masker absent method [(E) and (F)], respectively. On- and off-frequency maskers were always 20 ms, as was the delay between the elicitor and signal in the masker absent condition. The elicitors were presented ipsilaterally with respect to the signal ear. Elicitor durations ranged from 50 to 800 msec, shown by the yellow arrows on the elicitor. The double-headed arrow (red) indicates that the signal was adaptively varied, while the masker was fixed at a level that shifted the signal by 5 dB with no elicitor present.

A reduction in gain by the elicitor is predicted to shift signal threshold more following the off-frequency masker than the on-frequency one (Kawase et al., 2000; Jennings et al., 2009; Yasin et al., 2014). This is because the off-frequency masker is processed linearly at the signal frequency place and thus is not affected by gain reduction, whereas the signal and the on-frequency masker are nearly equally affected by gain reduction. This contrasts with the prediction of temporal integration of the elicitor and masker, also called additivity of masking (Penner and Shiffrin, 1980; Plack and O'Hanlon, 2003; Oxenham and Moore, 1994). With temporal integration, it would be expected that the on- and off-frequency conditions should produce equal shifts in thresholds with the addition of the elicitor. In contrast, in this study, it is proposed that masking by the masker may occur within a temporal window (because the duration is too short for gain reduction to affect the signal) but masking from the elicitor occurs by gain reduction. The shift in signal thresholds with an off-frequency masker following an elicitor were used for the analyses (magnitude and time constants), as they have been interpreted as a change in gain. Therefore, the difference in signal threshold between the off-frequency masked signal with [Fig. 2(B)] and without [Fig. 2(A)] the elicitor will be referred to as the “masker present” gain reduction estimate used in the current study.

A second method was also used to estimate gain reduction. Instead of using an off-frequency masker to fix the signal on the lower leg of the input-output function, quiet threshold of the signal served as the baseline condition [Fig. 2(E)] and was compared to signal threshold with an elicitor and a 20-ms delay between the elicitor offset and signal onset [Fig. 2(F)]. This will be referred to as “masker absent” estimate of gain reduction. Previous studies using masker present vs masker absent conditions found no significant differences in the magnitude of the gain reduction between the two methods (DeRoy Milvae and Strickland, 2018; Salloom and Strickland, 2021) for signal frequencies of 2 and 4 kHz. We used this masker absent condition to determine if it had a similar growth pattern to the off-frequency masked condition, and because it was more similar to the paradigm used by Oxenham and Plack (2000). To better understand gain reduction effects, schematics of our listening conditions showing the assumed underlying cochlear input-output functions are depicted in Fig. 1. In this study, the growth in signal threshold with elicitor duration was measured for the masker present and absent conditions. From these thresholds, gain reduction was measured in terms of magnitude and time constants.

Figure 1 is a schematic representation of how a reduction in gain by an elicitor would affect the signal (and masker, if present) for each listening condition [off-frequency—panels (A) and (B); on-frequency—panels (C) and (D); masker absent—panels (E) and (F)]. For the conditions without an elicitor, there are some assumptions that should be addressed to better understand the schematic. First, it is assumed that the listener makes use of an auditory filter with a center frequency at or close to the signal frequency. Second, only the components of the masker passing through this filter have any effect in masking the signal. Third, signal threshold is assumed to correspond to a certain signal-to-masker ratio at the output of the auditory filter. These assumptions are foundational to our work and are based on the power spectrum model of forward masking [see Fletcher (1940) and Patterson and Moore (1986)].

The input-output functions for the signal frequency (4 kHz in this study) are shown by the solid line in each panel. These input-output functions approximate the growth of excitation curves measured physiologically on the basilar membrane [e.g., Ruggero et al. (1997)]. For each listening condition, baseline conditions (i.e., no elicitor) are shown in the left panels while corresponding elicitor present conditions are shown in the right panels. Both the on- and off-frequency conditions have the signal at 5 dB SL, while the no masker condition tracks the signal at quiet threshold. Note that quiet threshold is indicated by a thick gray dashed line in all panels, and that the signal is elevated by 5 dB above the line for the masker present conditions, but not for the masker absent condition. Note in these panels the double-headed (yellow) arrow on the y axis (output) shows the threshold signal-to-masker ratio, which is assumed to be constant across conditions (Patterson and Moore, 1986).

Panel [A] shows the off-frequency masked condition where the signal is detected at a constant signal-to-masker ratio, which is the difference between the signal and masker levels at the output of the filter, shown by the yellow double-headed arrow between the output levels for the signal and the masker. Note that the off-frequency masker function is a straight line and is placed further away to represent that it is nearly an octave below the signal frequency. When the elicitor is added to this condition [panel (B)], gain will be decreased, shown by the dashed line directly below the solid line, and the signal will be affected (shown by the downward arrow and italicized “s”), while the off-frequency masker will not. This is because the off-frequency masker is processed linearly at the signal frequency place, and it is not affected by gain reduction. Because the signal output decreases, the input signal level must be increased in order to re-establish the original threshold signal-to-masker ratio, shown by the rightward arrow on the x axis. Overall, this will lead to an increase in the signal threshold following an elicitor.

For the on-frequency masked conditions, shown in panel [C], the signal and the masker share a common frequency, so they are both on the solid line. The signal is also at 5 dB SL, and a constant signal-to-masker ratio is established at the output of the auditory filter for detection shown by the double arrow at the output. The addition of the elicitor, in panel [D], shows that gain reduction does occur, as indicated by the thin black dashed line. However, it affects the signal and masker equally, resulting in no change in the signal-to-masker ratio at the output of the filter and thus no change in signal threshold. The difference in processing between the off- [panels (A) and (B)] and the on-frequency [panels (C) and (D)] maskers is the basis of the gain reduction experiments in this study and is not consistent with excitatory masking. If masking by the elicitor were excitatory, an equal increase in signal threshold would be seen in the on-frequency and off-frequency masker conditions.

Last, for the masker absent condition, in panel [E], the signal is detected at quiet threshold. In panel [F], the elicitor reduces the gain of the signal, indicated by the dashed line. A downward arrow and italicized ‘s’ indicate that the overall output level of the signal is lower and is not detectable at this level. In order to achieve the output level needed for detection, the input of the signal needs to be increased, indicated by the rightward arrow on the x axis. Gain reduction was measured as a function of elicitor duration in the off-frequency masked and no masker conditions. As mentioned in the current section [psychoacoustic stimuli], there was always 20 ms between the offset of the elicitor and the onset of the signal, whether a masker was present [Fig 3(A)] or absent [Fig 3(B)]. Note that the schematic masker in the masker present condition in Fig. 3(A) represents either the on- or off-frequency masked condition. Therefore, t is the total duration from the onset of the elicitor to the onset of the signal minus the onset delay in ms. The onset delay was estimated to be 20 ms. This means t is equal to the elicitor duration for the psychoacoustic data [depicted in Fig. 3 (bottom panel)]. That is, each data point in the time constant estimation corresponded to a total duration of the elicitor [50, 65, 100, 200, 400, and 800 ms] when the corrections for onset delay of the MOCR and the duration of the masker/gap were accounted for, and thus are the same as the values used in the magnitude estimations shown later.

FIG. 3.

(Color online) A schematic of the masker present (A) and masker absent (B) stimuli used, and an accompanying estimation of the time course of gain reduction (bottom). The masker present configuration in panel (A) represents either on-frequency or off-frequency masked conditions and is depicted as “masker” in the legend. To account for the onset delay (bottom), the total time of elicitation starts 20 ms after elicitor onset and ends at the onset of the signal [Δt on/on – 20 (ms)] (see Fig. 6). Because the onset delay of the MOCR and the duration of the masker/gap are accounted for in our time constant estimations, t is equal to the elicitor duration for each data point.

2. Procedure

All psychoacoustic measures were completed in a double-walled sound-attenuating booth. Stimuli were generated with custom matlab software (Bidelman et al., 2015) with a Lynx TWO-B sound card (Lynx Studio Technology, Inc., Costa Mesa, CA). The stimuli were then passed through a headphone buffer (TDT HB6, Tucker-Davis Technologies, Alachua, FL) and delivered to one ear through an Etymotic ER-2 (Etymotic Research, Inc., Elk Grove Village, IL) insert earphone. The subjects had insert earphones in both ears. The insert earphones have a flat frequency response at the eardrum from 250 to 8000 Hz. High pass noise (from 1.2 times the signal frequency to 10 kHz) was used to reduce the possibility of off-frequency listening (Nelson et al., 2001) for all parts of the experiment except during quiet threshold measurements. The high pass noise began 50 ms before the onset of the stimuli and ended 50 ms after the signal offset and was 50 dB spectrum level below the signal level.

All psychoacoustic measurements utilized a three-interval forced-choice (3IFC) task using a matlab GUI, in which only one of the choices contained the signal. Each interval was visually marked on the computer screen, and intervals were separated by 500 ms of silence. Subjects could use either a mouse or the keyboard to indicate which interval contained the signal. Visual feedback was given for correct and incorrect responses. Signal and masker levels were adjusted to estimate a response threshold of 70.7% correct (Levitt, 1971). For signal threshold measures (quiet thresholds and measures of gain reduction), if the subject chose correctly over two consecutive trials, the level of the signal decreased, while an incorrect response would cause the level of the signal to increase (two down, one up). For masking thresholds, if the subject chose correctly over two consecutive trials, the level of the masker increased, while an incorrect response would cause the level of the masker to decrease (two up, one down). The step size was 5 dB for the first four reversals and then decreased to 2 dB for the remaining reversals. The last eight reversals were averaged to produce a final threshold for each run.

Subjects had approximately 1 h of training before data collection began in order to help them understand the task. These training tasks involved listening to tones in quiet, tracking masker thresholds for a fixed-level tone using on- and off-frequency maskers, and tracking signal threshold with fixed level on- or off-frequency maskers, all of which are standard tasks in the current study. Learning was evaluated by comparing thresholds in these conditions, where thresholds were considered similar if they were within 3 dB of each other, and typically each task was completed 3–5 times. If the thresholds in the practice session varied by more than 3 dB consistently, the subject would either receive more training on that specific task until the thresholds were within 3 dB of each other, or they were asked to come back for another session which included training. Most subjects had consistent thresholds after the initial practice session. After training, each session of data collection was 1–1.5 h to prevent attentional fatigue. Each condition was tested at least twice per session, and thresholds are an average of the last two thresholds recorded for that condition. These final thresholds served as the data reported in the figures and the statistical analysis. Runs with a standard deviation (SD) greater than 5 dB were discarded from the overall averages and repeated if necessary. Data from each subject were collected for a minimum of one session for each psychoacoustic task, and additional sessions were conducted if large variability or learning effects occurred. The order of presentation of conditions was interleaved across subjects. All statistical and post hoc analyses of gain reduction were calculated with IBM SPSS 28 statistical software.

Before any statistical test was conducted, all of the data sets were tested for the assumption of normality in SPSS using the Shapiro–Wilk test of normal distribution and the corresponding normal Q-Q plots. With this test, any subset of data tested for this assumption with a p-value equal to or greater than 0.05 would meet the criterion to assume normal distribution. A statistical outlier in this case was defined as any value that lies outside 1.5 times the interquartile range (i.e., above or below the 75th and 25th percentiles, respectively). All statistical tests in the current study met the assumption of normality.

III. RESULTS

A. Threshold shifts with elicitor duration

The individual and average threshold shifts as a function of elicitor duration for the masker present and masker absent conditions can be seen in Figs. 4 and 5, respectively. Cells are labeled by subject number. The dashed line in each panel represents the baseline condition, which is signal threshold with a masker for the masker present conditions, and quiet threshold for the signal for the no-masker condition, respectively. This allows us to compare the shifts in threshold by the elicitor for all conditions within the same figure. Quiet thresholds were recorded to find the masker levels that shifted the signal by 5 dB in the on- and off-frequency masking thresholds tasks (i.e., the masker present baseline), and served as the baseline threshold for the masker absent task. For a given subject, these thresholds were highly consistent across trials, and were always within 2 dB of each other in the overall average. Additionally, each subjects' signal threshold for the fixed on- and off-frequency masked conditions (no elicitor) were within 3 dB of one another, thereby confirming equivalent masking of the signal. These thresholds, as well as quiet thresholds, and masker thresholds for the signal at 5 dB SL are reported for each subject in Table I. Table I also provides each subject's total gain at the signal frequency by taking the difference between the on-frequency and off-frequency masker thresholds for the signal at 5 dB SL. The overall qualitative pattern, for both the individual and group data is that threshold shifts with an off-frequency masker (square symbol) are larger than threshold shifts with an on-frequency masker (diamond symbol), which is consistent with gain reduction and not excitatory masking. The off-frequency masked conditions produce a growth function that is similar in magnitude to the masker absent conditions (square symbol). Both of these observations were tested statistically in the section below.

FIG. 4.

(Color online) Signal threshold shifts as a function of elicitor duration for individual subjects at 4 kHz. Symbols indicate listening condition, while the horizontal dashed lines indicate the reference condition with no elicitor. Baseline for the masker present conditions is signal threshold with a fixed level masker that shifted the signal by 5 dB, and is quiet threshold for the signal for the no-masker condition. Standard deviation (SD) of signal thresholds is indicated by the error bars.

FIG. 5.

(Color online) Average signal threshold shifts (N = 9) as a function of elicitor duration from the data in Fig. 4 (and identically labeled). SEM of signal thresholds is indicated by the error bars.

TABLE I.

Individual subject signal, masker, and masked thresholds for the conditions tested in this study. Quiet thresholds served as the baseline threshold for the masker absent task and were used to find the on- and off-frequency masker levels that shifted the signal by 5 dB. These on- and off-frequency maskers were then fixed in level, and the signal thresholds for both on- and off-frequency conditions are reported here (i.e., the masker present baseline). Note that each subject's on- and off-frequency masked signal thresholds were always within 3 dB of one another, thereby providing equivalent masking on the signal. Each subject's gain estimate at the signal frequency was derived by taking the difference in masked threshold between the off-frequency and on-frequency masked condition when the signal was fixed at 5 dB SL.

Subjects	Quiet thresholds	On-frequency masker thresholds	Off-frequency masker thresholds	On-frequency signal thresholds	Off-frequency signal thresholds	Gain
S1	24.6	23.0	75.0	28.8	28.8	52.0
S2	21.9	19.0	76.0	28.6	29.4	57.0
S3	19.2	22.0	75.9	21.9	22.4	53.9
S4	23.6	24.6	64.7	25.6	24.5	40.1
S5	20.8	17.0	75.0	26.1	26.1	58.0
S6	16.8	21.0	70.0	22.3	22.9	49.0
S7	20.4	20.0	75.0	26.4	25.4	55.0
S8	14.8	18.3	69.7	20.5	19.8	51.4
S9	18.9	17.2	68.7	24.1	23.9	51.5
Average	20.1	20.2	72.2	24.9	24.8	52.2

A two-way 3 × 6 repeated measures ANOVA was conducted to determine if the mean signal threshold shifts (dependent variable) significantly varied with the independent variables masker type (off-frequency, on-frequency, and no-masker) and elicitor duration (50, 65, 100, 200, 400, and 800 ms), as well as to measure any potential interaction effects between independent variables. The results can be seen in Table II.

TABLE II.

Results of the 3 × 6 repeated measures ANOVAs on the magnitude of gain reduction as a function of masker type (off-frequency, on-frequency, masker absent) and elicitor duration (50, 65, 100, 200, 400, 800 ms). Both main effects of masker type and elicitor duration on gain reduction were statistically significant, however, no significant interaction between the two was found. Note that the degrees of freedom in the F-values for elicitor duration are not integers, because sphericity could not be assumed for these data. The more conservative Greenhouse-Geisser critical F-value was reported instead, which helped correct for violation of sphericity. Effect sizes for the ANOVA output are indicated by partial-eta (η²) squared values.

Masker type
F-statistic	p-value	Partial-eta squared (η2)	Paired comparisons	p-value
F(2,16) = 39.635	p < 0.001a	0.832	Off-freq vs On-freq	p < 0.001a
			No-M vs On-freq	p < 0.001a
			Off-freq vs No-M	p = 0.116
Elicitor duration
F-statistic	p-value	Partial-eta squared (η2)	—	—
F(3.014,24.116) = 60.339	p < 0.001a	0.883	—	—
Masker type * elicitor duration interaction
F-statistic	p-value	Partial-eta squared (η 2 )
F(10,80) = 0.832	p = 0.599	0.094

^a Significance.

The analysis indicated that there was a significant main effect of masker type. Bonferroni corrections revealed that threshold shifts for the off-frequency masked conditions were significantly different from those for the on-frequency masked conditions, with an average difference of 6.39 dB (SD = 0.40 dB) between the two conditions across elicitor durations. This result is consistent with gain reduction, and is not consistent with excitatory masking (Salloom and Strickland, 2021). A significant difference was also observed between the masker absent and on-frequency masked conditions, with an average difference of 4.26 dB (SD = 0.33 dB) between the conditions across elicitor durations. However, there was no significant difference between the off-frequency and masker absent conditions, which is consistent with previous work using a fixed-duration 50-ms elicitor (DeRoy Milvae and Strickland, 2018; Salloom and Strickland, 2021), suggesting that both conditions give an estimate of gain reduction. While not significant, the off-frequency condition produced larger shifts compared to the masker absent conditions, with an average difference in threshold shifts of 2.12 dB (SD = 0.54 dB). While small, this difference between threshold shifts may be due to the duration of stimulation of the MOCR, as the off-frequency masked condition may have a slightly longer elicitation time due to the presence of the masker compared to the masker absent condition (see Fig. 3). Next, a significant main effect of elicitor duration was found, indicating that gain reduction magnitude estimates generally increase with elicitor duration. For both the off-frequency and masker absent conditions, the average threshold shifts increased by about 1 dB or less with doubling of the duration of the elicitor. Overall, the off-frequency masked condition produced threshold shifts that ranged from 7.0 to 11.18 dB with elicitor duration, and masker absent condition produced threshold shifts ranging from 4.65 to 9.32 dB with elicitor duration. No interaction was found between the main effects (masker type and elicitor duration).

Signal threshold shifts in the on-frequency condition followed the same pattern with elicitor duration as in the off-frequency and no masker conditions, but with a smaller magnitude. While it is anticipated that the on-frequency masker and signal are affected approximately equally by the elicitor (i.e., since the reduction of gain affects them equally at CF), resulting in little-to-no overall threshold change with the addition of the precursor [see Fig. 1, panels (C) and (D)], it is possible that the level of the signal is reduced below audibility with increased precursor duration. This is considered further in Sec. IV.

B. Time constants of gain reduction

The next step was to estimate time constants for gain reduction from the data. While all three conditions reflect gain reduction, in the on-frequency masker condition many of the data points are not significantly different from the baseline. In addition, the magnitude of gain reduction is estimated by the off-frequency and masker absent conditions, as explained in Sec. III A. Therefore, these conditions were used to estimate the time constants. An inverse exponential function was fit to individual and group data from the previous section, and an overall time constant (τ) and a corresponding variance accounted for value were estimated. The formula for this function is, Y(t) = Y_max(1 – e^−t/τ) which estimates the time at which approximately 63% of the maximal effect of the growth function is achieved. In this formula, Y represents the event or response of interest, t represents elapsed time, and τ represents the time constant. Time constants estimated with this function have previously been used in human behavioral and physiological measures of overshoot (Walsh et al., 2010b), an effect that has been posited to relate to gain reduction via the MOCR [e.g., Strickland (2001)]. With respect to our psychoacoustic experiments, Y is the signal threshold shift between the signal with and without the elicitor condition (i.e., gain reduction). Previous studies have found that after the onset and the offset of the elicitor, there is a delay of approximately 20–25 ms before there is an effect on OAEs or perception (James et al., 2005; Backus and Guinan, 2006; Roverud and Strickland, 2014), and this was used in designing the psychoacoustic stimuli.

In order for individual time constants to be included in the group averaged time constant, at least 60% of the variance in the data needed to be accounted for in the individual fit. Individual and averaged data with fits can be found in Figs. 6 and 7, respectively. In addition to the data points in the time constant estimation, a point was added at zero on the abscissa and ordinate to represent the baseline condition as well as to ground the model. The symbols are closed for off-frequency masked conditions and open for masker absent conditions, and the fitted curve is solid for off-frequency masked and dashed for masker absent conditions. Individual time constants are shown in Table III, and range from 29 to 172 ms. This range is similar to the range of time constants for individual subjects reported from the behavioral overshoot tasks in the Walsh et al. (2010b) study (19.5–141.8 ms). The average time constant for the off-frequency masked conditions was 46.29 ± 4.72 ms (R² = 0.91), and for the no masker conditions was 78.19 ± 12.55 ms (R² = 0.97). Despite the range of time constant values for the individual data, all of the time constants were fairly short, where the average buildup effect is near-maximal within approximately 200 ms of elicitor activation (Fig. 7). Last, the exponential function fit the individual and the average behavioral data quite well, with over 90% of the variance accounted for in each fit.

FIG. 6.

(Color online) Signal threshold shifts as a function of precursor duration for individual subjects. These data are the same as in Fig. 4. Filled symbols are the off-frequency masked data, and open symbols are the no masker data. The solid and dashed curves are the exponential curve fits for the off-frequency masked and no masker conditions, respectively. SD of signal thresholds is indicated by the error bars.

FIG. 7.

(Color online) Average (N = 9) fitted signal threshold shifts as a function of elicitor duration, estimated from the data in Fig. 6 (also identically labeled). SEM of signal thresholds is indicated by the error bars.

TABLE III.

Individual and average time constants and corresponding variance accounted for (R²) when fitting the exponential function for each listening condition. τ units are in ms, and the ± error of the average time constants is the SEM of the individual time constants.

	4 kHz off-frequency		4 kHz masker absent
Subjects	τ	R2	τ	R2
S1	73.40	0.96	78.80	0.98
S2	61.45	0.87	53.59	0.99
S3	33.75	0.97	56.77	0.87
S4	48.68	0.79	171.90	0.98
S5	33.60	0.86	67.56	0.96
S6	29.20	0.88	60.42	0.98
S7	43.17	0.96	56.73	0.96
S8	49.45	0.93	96.40	0.98
S9	43.93	0.96	61.54	0.97
Average	46.29 ± 4.72	0.91	78.19 ± 12.55	0.96

Not all growth functions increase monotonically. In some individual cases there was an oscillating effect where the functions slightly increase and decrease as a function of elicitor duration (off-freq: S4, S5, S6, S7, S8; no masker: S9). For many of these individual functions, the maximal (Y_max) or near maximal threshold shifts occurred for the 200-ms elicitor duration. This oscillating effect has been documented before with on-frequency tonal elicitors (Roverud and Strickland, 2014) and noise elicitors (Oxenham and Plack, 2000). This has been modeled as the elicitor turning down the gain at the signal frequency place and decreasing its own effectiveness as the elicitor duration is increased (Roverud and Strickland, 2014). An elicitor with reduced effectiveness can lead to an improvement in signal threshold (e.g., Fig. 6; S8, off-frequency condition). Consistent with this interpretation, this oscillation is also evident with elicitor duration in some OAE data (James et al., 2005).

IV. DISCUSSION

A. Comparison to previous physiological and psychoacoustic findings

The current study investigated the effect of broadband elicitor duration on a psychoacoustic measure of gain reduction (possibly by the MOCR) in adults with normal hearing. Prior to the current study, the temporal properties of a broadband elicitor on psychoacoustic measures of gain reduction had not been directly tested in a forward masking paradigm. The overall time constants were 46 ± 4.72 ms for the off-frequency masked condition, and 78 ± 12.55 ms for the masker absent condition. The average buildup effect is maximal within approximately 200 ms of elicitor activation. An exponential curve fit the data well both for individual and group data, with most R² values above 90%. Despite the apparent difference between the two time constants measured, they are generally within the range of the fast time constants of the MOCR time course measured from other studies in humans, and will be compared to OAE and psychoacoustic measures of gain reduction in more detail below. It should be reiterated that the use of a relatively low-level elicitor (50 dB SPL) in our psychoacoustic experiments and measuring WAI MEMR thresholds rules out MEMR effects on the responses. Additionally, by using on- and off-frequency maskers, we were able to show that these effects are consistent with predictions of gain reduction, and not additivity of masking.

Our data are highly consistent with the fast time constants of the MOCR build-up estimated from various OAE based paradigms. By measuring DPOAE adaptation (Kim et al., 2001), or DPOAE adaptation and the change in the OAE level with an elicitor (Bassim et al., 2003), Kim et al. (2001), and Bassim et al. (2003) both found fast MOCR time constants around ∼70 ms. Backus and Guinan (2006), also had a similar fast time constants from their SFOAE method, which ranged from 60 to 80 ms, using broadband elicitors. One caveat in comparing our time constants to those found in other studies is that only a few shorter duration elicitors were used in this study (50, 65, 100 ms) making it unlikely to measure a finer-step buildup effect which may result in multiple time constants (Kim et al., 2001; Bassim et al., 2003; Backus and Guinan, 2006). For the same reason, we cannot compare our time constants to the longer time constants (>1 s) that were estimated in those OAE studies since only a few longer elicitor durations were used (400 and 800 ms). While those studies were able to measure multiple time constants due to the nature of their techniques, analyses, and stimuli used, our results likely captured the fast time constants that were found in those OAE studies. Similar to the OAE studies using elicitors (James et al., 2005; Backus and Guinan, 2006), we found that the buildup (i.e., threshold shifts) of the MOCR continued to grow with elicitor duration (Fig. 5), albeit 1 dB or less for each doubling of the elicitor duration.

As stated earlier, this study was designed to be an extension of previous psychoacoustic work by Roverud and Strickland (2010, 2014), which used tonal elicitors to study the time course of gain reduction. Roverud and Strickland (2010, 2014) found that for most subjects, the on-frequency elicitor produced the maximal threshold shift for elicitors of 50 ms, and then the thresholds either plateaued or oscillated with increased elicitor duration. However, for off-frequency elicitors, threshold continued to increase with elicitor durations up to 150 ms. In Roverud and Strickland (2014) time constants were fit to the data as part of a model that also included a temporal integration window and gain reduction. The time constants ranged from approximately 28 to 76 ms. The model was able to predict the oscillation in gain reduction for on-frequency maskers for durations longer than 50 ms. From these findings, multiple studies of gain reduction used 50-ms elicitor durations, including for broadband elicitors, as it was thought that this duration would maximally elicit gain reduction just as it did for on-frequency tonal elicitors (DeRoy Milvae and Strickland, 2018; Hegland and Strickland, 2018; Salloom and Strickland, 2021; DeRoy Milvae and Strickland, 2021). However, the results of the current study indicate that those studies may have underestimated the maximal amount of gain reduction by approximately 2.39 dB, based on the difference between gain reduction estimates for the 200-ms and 50-ms elicitor conditions.

Several factors make it difficult to directly compare results with the current study to the Roverud and Strickland (2010, 2014) studies. First, elicitor durations in those studies ranged from 5 to 100 ms (Roverud and Strickland, 2010), and from 10 to 150 ms (Roverud and Strickland, 2014), while in the present study there was an overall broader range of elicitor durations used, ranging from 50 to 800 ms. The difference in range of elicitors used between studies may be a factor in the time constant estimation since the Roverud and Strickland studies did not use similarly long duration elicitors, and the off-frequency data in the Roverud and Strickland study did not always reach a plateau. Another important distinction between the two earlier studies and the current one is that the current study estimated time constants (63% of Y_max) which are an extrapolated estimate from the curve fitting, while the other studies reported the maximal effect of the on-frequency tonal elicitor (the 50-ms elicitor). In the current study, the average threshold shifts kept increasing with elicitor duration, which is somewhat comparable to the off-frequency elicitor condition in the Roverud and Strickland (2014) study, but at a much slower growth rate (∼1 dB or less per doubling of elicitor duration). While this pattern varied across subjects, it appears that the duration needed to produce maximum or near-maximum gain reduction in the present study is between the time needed for on- and off-frequency elicitors in the Roverud and Strickland (2014) study based on our estimated time constants and individual subject growth patterns. This is consistent with a recent study (DeRoy Milvae and Strickland, 2021), which found that broadband elicitors had only slightly larger effects on signal threshold compared to on-frequency elicitors as a function of elicitor level when the elicitor duration was fixed in duration (50-ms) and the broadband elicitor level was calculated as the energy that should pass through a filter centered at the signal frequency. Since filter bandwidth may be widening (Jennings and Strickland, 2012) and suppression of central components of the noise by surrounding components maybe decreasing (Hegland and Strickland, 2018) with gain reduction, it might be expected that the most effective duration would be longer for a noise than for a tone elicitor. This is likely why some psychoacoustic studies, such as Yasin et al. (2014), used long duration broadband noise elicitors (500 ms) when they measured gain reduction effects (also at 4 kHz), although they did not systematically test for elicitor duration effects. Overall, the data in the current study are consistent with gain reduction and show that the maximal effects of gain reduction occur for elicitor durations of 200 ms or less for ipsilateral broadband stimulation. This is relatively short compared to the continuous or long duration elicitors that have typically been used to activate the MOCR in the psychoacoustic and physiological literature. Future studies of gain reduction using broadband elicitors may fully activate the MOCR with 150–200-ms durations, which could potentially save substantial data acquisition time.

B. On-frequency threshold shifts

While the data in the current study are consistent with gain reduction, possibly by the MOCR, there were threshold shifts for the on-frequency conditions by the elicitor for some individuals and the average data (Figs. 4 and 5). Schematic cochlear input-output functions were used to explain the predicted effects of the elicitor on the conditions tested in the current study which would be consistent with gain reduction (Fig. 1). It was reasoned that there should be an equal effect of the elicitor on the signal and the on-frequency masker since they share the same frequency, and their gain is reduced at the signal frequency place. For changes in gain of 5 dB or less, this would result in no overall change in the SNR at the output of the cochlear filter, and thus no threshold shift of the signal. However, with more gain change, it is possible that the output level of the on-frequency masker and/or signal may be reduced to below quiet threshold. Because of this, at longer elicitor durations, the on-frequency condition becomes more like the masker absent conditions in Figs. 1(E) and 1(F). One way that this may be avoided in future work is to increase the sensation level beyond 5 dB SL, as the reduction in gain by the elicitor may not reduce the level and effectiveness of the on-frequency masker below audibility. While the on-frequency masked conditions did produce threshold shifts for longer duration elicitors, they were significantly smaller than the off-frequency masked conditions by an average of 6.33 dB for each elicitor duration, as reported in Sec. III. Therefore, the results in this study are still consistent with gain reduction, and not excitatory masking.

C. Real world listening conditions and gain reduction

The current study used forward masking techniques designed to estimate a change in cochlear gain which avoids the effects of suppression. Also, the measures in this study estimated changes in the signal level near threshold, where gain reduction increases the signal threshold. However, in realistic listening scenarios where the MOCR may be useful, such as in a crowded area or noisy restaurant, it is likely that gain reduction would occur throughout the signal of interest (i.e., speech) and the background noise, and that suppression would also play a role. The benefits of gain reduction in these environments would occur at higher sound levels (suprathreshold) than those used in this study and would help to increase the slope of the input-output function thereby increasing detection of the signal. In these listening scenarios, the background noise is likely continuous, and therefore the MOCR system is likely to be fully active and adjusting cochlear gain with time. Data from this study and others support that the MOCR system can adjust the dynamic range on relatively short time scales (Roverud and Strickland, 2010, 2014), as well as on longer time scales (Kim et al., 2001; Bassim et al., 2003; Backus and Guinan, 2006; Boothalingam et al., 2021). This is supported by the fact that we found large gain reduction estimates for even the shortest elicitor duration of 50 ms (see Fig. 5), and similar to the OAE based studies, these effects continued to grow with elicitor duration (James et al., 2005; Backus and Guinan, 2006). As stated in the Introduction, there is evidence from sound-evoked responses of the BM recordings in mammals that show faster (∼50 ms) and slower (∼10 s) time constants with electrical stimulation of the MOCR system, each with distinct BM phase responses at CF (e.g., Cooper and Guinan, 2006b). Additionally, recordings from mammal auditory nerve with electrical MOCR fiber stimulation have also shown fast (tens of milliseconds) and slow forms (tens of seconds) of MOCR inhibition (Sridhar et al., 1995; Sridhar et al., 1997), very similar to the fast and slow time constants from the OAE and BM aforementioned studies. Taken together, the MOCR reduces cochlear gain motion on faster and slower time scales. More work is needed to understand if these fast and slow efferent inhibitory mechanisms serve differential roles on cochlear processing, and how each mechanism may influence perception.

Care should be taken when comparing behavioral gain reduction data to physiological (OAEs or mammalian neural responses) data, as the physiological methods have not been a predictor of performance in behavioral measures of gain reduction (see Marrufo-Pérez et al., 2021; Jennings, 2021). Also, it is not clear what part of the cochlear response OAEs reflect (Guinan, 2018), and there may be inherent differences between these measures even if they are reflecting the same mechanism (i.e., MOCR gain reduction) (Jennings, 2021). Therefore, it is important to use behavioral, neural, and OAE measures in order to fully understand gain reduction, which should be considered in future studies. Overall, with broadband sound elicitation, the MOCR system appears to be nearly fully active within 200 ms and continues to build up with continued sound stimulation until fully active. Therefore, the MOCR efferent system can adjust the dynamic range on fast and slower time scales accordingly, which may facilitate the ability for humans to hear in many different and challenging listening environments.

D. Conclusions

• (1)The statistical analysis showed that the main effects of masking type and elicitor duration were significant, but there was no significant interaction between the two main effects. These effects are consistent with gain reduction, and not additivity of masking.
• (2)Gain reduction time constants for ipsilateral broadband elicitors were relatively short; approximately ∼46 ms for the off-frequency masked condition and ∼78 ms for the masker absent condition.
• (3)Future studies using forward masking paradigms to study gain reduction should use 150–200 ms broadband elicitors to maximally or near-maximally stimulate the MOCR. This is longer than the 50-ms duration found using on-frequency tonal elicitors.

APPENDIX: WIDEBAND ACOUSTIC IMMITTANCE (WAI) MEMR MEASURES

1. Stimuli and procedure

As mentioned in Sec. II A, clinical acoustic MEMR thresholds were measured as part of the standard audiological battery for testing subjects. These thresholds were measured using a broadband elicitor and a 226 Hz probe at peak tympanometric static pressure. The goal was to ensure that the MEMR reflex was not responsible for the effects of an elicitor in the current study. Typical clinical immittance measures may underestimate the MEMR by ∼14 dB when compared to WAI MEMR thresholds (Feeney and Keefe, 2001; Feeney et al., 2017). Additionally, WAI studies have also shown the MEMR could affect higher probe frequencies [Feeney et al. (2017), see their Fig. 1]. Therefore, in addition to clinical tympanometry, ipsilateral WAI measurements were made in each subject's test ear (right ear). WAI measures used click probes to assess middle-ear absorbance over a wide range of frequencies (0.25–8.0 kHz) at ambient pressure. MEMR thresholds were measured for each subject using a WAI paradigm adapted from Keefe et al. (2017), which is virtually identical to the paradigm used by Bharadwaj et al. (2022). A 90-dB peak-equivalent SPL (peSPL) click with a flat incident power spectrum in the 0–10 kHz range was used to measure the acoustic immittance properties of the ear-canal and middle-ear system. Each WAI MEMR measurement trial consisted of a series of seven clicks alternating with a 120-ms ipsilateral elicitor (including 5-ms cos² onset and offset ramps) [see Bharadwaj (2022), their Fig. S-3A]. The elicitor was a pink broadband noise (0.5–8 kHz), and ranged from 34 to 88 dB FPL. The gap between the peak of the click and the onset of the noise elicitor (0 voltage point) was approximately 28 ms, and the gap between the offset of the noise (0 voltage point) and the next click was approximately 14 ms, for a total of approximately 42 ms between elicitor presentations (e.g., the click window). This trial structure was used for each elicitor level and the level was repeated 32 times with an inter-trial interval of 1.5 s to allow the MEMR to relax back to baseline levels. For each elicitor level, the immittance measured using clicks two through seven in the sequence were averaged together and the change relative to the first click was calculated as the WAI MEMR metric. The dB change in ear-canal pressure induced by the MEMR was quantified as a function of frequency to yield a pattern of alternating negative and positive peaks at different frequencies. Additional calibrations which help reduce extraneous variance were leveraged in these estimations (fully described in this Appendix). The dB change in ear-canal pressure was added to the dB change in ear-canal conductance to yield a pattern of dB change in absorbance power [see Bharadwaj et al., 2022, Supplementary Fig. 3b-d)]. The absolute value of this dB change in absorbed power was averaged over 0.5 to 2 kHz to yield a single number per elicitor level (Fig. 8). A thresholding procedure was used to reject artifactual trials before averaging. Figure 8 shows ipsilateral WAI MEMR data from a single subject to illustrate how WAI MEMR thresholds were estimated. The top panel shows the shifts in power absorbance (ΔA) across frequency as a function of ipsilateral elicitor level, relative to the absorbance of the click alone. The bottom panel shows the shift in absorbed power as a function of elicitor level over a 500-2000 Hz frequency band, derived from the data in the top panel. A change in absorbance power greater than 0.1 dB (dashed line) was chosen as the threshold for significant MEMR activation (WAI MEMR threshold). This subject's ipsilateral WAI MEMR threshold is approximately 60 dB FPL, shown in the bottom panel. This estimation is made clear in the top panel by viewing the tracing corresponding to the 58 dB FPL elicitor level (faint blue) which is right below MEMR threshold, and the tracing corresponding to the 64 dB FPL elicitor level (green tracing) which is right above MEMR threshold. WAI MEMR thresholds were collected for all subjects in the study. The total duration of this experiment was approximately 16–17 min for a single subject, and data were collected prior to any psychoacoustic experiments.

FIG. 8.

(Color online) WAI data with an ipsilateral elicitor from a single subject (S1). Top: Change in absorbed power across frequency. The tracing for the 64 dB FPL elicitor shows noticeable middle ear absorbance changes (green tracing), while the tracing for the 58 dB FPL elicitor shows negligible changes in middle ear absorbance (faint blue). Bottom: Change in absorbed power in a frequency band ranging from 500 to 2000 Hz. The level at which the fitted function crossed 0.1 dB is considered the WAI MEMR threshold in the current study. WAI MEMR threshold for this individual is approximately 60 dB FPL.

The setup for WAI measures was identical to the measurements in Bharadwaj et al. (2022). WAI measures were made with an ER-10X Extended-Bandwidth Acoustic Probe System (Etymotic Research, Elk Grove Village, IL) which utilizes integrated FPL systems, allowing for accurate in-ear calibrations up to ∼20 kHz in human subjects. This system allowed for probe stimuli (clicks) and ipsilateral MEMR-eliciting stimuli to be presented from separate speakers to limit interchannel interactions and distortions, and a microphone to measure sound pressure near the ear canal. The WAI stimuli were generated with two Tucker-Davis Technologies (TDT) RZ6 programmable DSP processors, and these signals were sent to and amplified by two separate TDT HB7 headphone drivers, which then fed to the ER-10X system and probe.

Before any WAI measurements were collected from a subject, the ER-10X probe needed to be calibrated by determining the Thevenin-equivalent impedance and pressure characteristics of the sound sources (Allen, 1986). These measurements are necessary to later determine the acoustic impedance at a location in the ear canal. The Thevenin-equivalent source and impedance for the click probe were estimated by measuring the acoustic response at the ER-10X microphone when the eartip was coupled to loads whose acoustic impedance values can be approximated using theoretical calculations (closed brass tube cavities of 8 mm diameter and five different lengths, which are the “ER-10X calibrator”). The estimates were refined until the so-called “calibration error” (a dimensionless energy ratio scaled by a factor of 10 000 and averaged over 2–8 kHz) was minimized. Error values of < 1 are typically considered good quality calibration (Neely and Liu, 1994); we consistently obtained errors within a range from 0.01 to 0.04. With the probe properties calibrated, the same click stimulus was then used to estimate the immittance properties of each subject's ear for each insertion of the probe tip. Because WAI measures rely on ear canal probes that are tightly sealed in an individual's ear canal, probe tips that are not fully secure can cause air-leaks or poor fitting that compromise the validity of the measurements and calibrations. Air-leaks can cause changes of absorbance that increase with the degree of the leak, especially at lower frequencies where air leaks reliably cause an increase in absorbance (Groon et al., 2015). Therefore, the criteria of low-frequency (0.2 kHz) absorbance that was used to detect air leaks was less than 29%, and the admittance phase was greater than 44° but no greater than 90°, as established by Groon et al. (2015).

2. Clinical and WAI MEMR thresholds

Table IV shows ipsilateral MEMR thresholds measured clinically and using WAI for each subject. These separate measures are reported to validate that the MEMR was not evoked by the elicitor during the psychoacoustic experiment. As expected from previous research, WAI thresholds are lower than clinically measured thresholds. No subject had an MEMR threshold below 50 dB SPL (clinically measured) or 50 dB FPL (WAI measure). The results from the independent measures indicate that the MEMR was highly unlikely to have been activated by the elicitors used in the psychoacoustic experiments.

TABLE IV.

Individual clinical and WAI MEMR thresholds. In the middle column, MEMR thresholds (dB SPL) measured clinically with a 226-Hz probe and a white broadband noise elicitor are shown. In the right column, and MEMR thresholds measured using WAI (dB FPL) are shown. All thresholds were measured from right ear.

Clinical and WAI measured MEMR thresholds
Subject	Clinical (dB SPL)	WAI (dB FPL)
S1	68	67
S2	83	60
S3	83	57
S4	83	60
S5	83	80
S6	68	52
S7	93	60
S8	78	65
S9	88	74