Medial olivocochlear reflex effects on amplitude growth functions of long- and short-latency components of click-evoked otoacoustic emissions in humans

Abstract

Functional outcomes of medial olivocochlear reflex (MOCR) activation, such as improved hearing in background noise and protection from noise damage, involve moderate to high sound levels. Previous noninvasive measurements of MOCR in humans focused primarily on otoacoustic emissions (OAEs) evoked at low sound levels. Interpreting MOCR effects on OAEs at higher levels is complicated by the possibility of the middle-ear muscle reflex and by components of OAEs arising from different locations along the length of the cochlear spiral. We overcame these issues by presenting click stimuli at a very slow rate and by time-frequency windowing the resulting click-evoked (CE)OAEs into short-latency (SL) and long-latency (LL) components. We characterized the effects of MOCR on CEOAE components using multiple measures to more comprehensively assess these effects throughout much of the dynamic range of hearing. These measures included CEOAE amplitude attenuation, equivalent input attenuation, phase, and slope of growth functions. Results show that MOCR effects are smaller on SL components than LL components, consistent with SL components being generated slightly basal of the characteristic frequency region. Amplitude attenuation measures showed the largest effects at the lowest stimulus levels, but slope change and equivalent input attenuation measures did not decrease at higher stimulus levels. These latter measures are less commonly reported and may provide insight into the variability in listening performance and noise susceptibility seen across individuals.

NEW & NOTEWORTHY The auditory efferent system, operating at moderate to high sound levels, may improve hearing in background noise and provide protection from noise damage. We used otoacoustic emissions to measure these efferent effects across a wide range of sound levels and identified level-dependent and independent effects. Previous reports have focused on level-dependent measures. The level-independent effects identified here may provide new insights into the functional relevance of auditory efferent activity in humans.

INTRODUCTION

Cochlear processing of sound is modified by the medial olivocochlear reflex (MOCR), which reduces cochlear amplifier gain by decreasing the motility of the outer hair cells (1). The functional purpose of the MOCR is not fully understood, but it is thought to improve hearing in background noise by reducing the gain of the cochlear amplifier to extend the dynamic range of auditory nerve fibers (see Refs. 1 and 2 for reviews). The MOCR has also been shown to have a protective effect against noise damage (3). Both of these effects are operative at moderate to high sound levels.

In humans, MOCR effects on cochlear mechanics can be investigated noninvasively using otoacoustic emissions (OAEs). OAE-based assays of MOCR are susceptible to contamination by the middle-ear muscle reflex (MEMR), which is evoked by similar acoustic stimuli but at slightly higher levels. As a consequence, the majority of human studies have necessarily focused on measuring MOCR effects on OAEs evoked by low stimulus levels. Conveniently, the largest MOCR-mediated changes in OAE amplitude have also been observed at low stimulus levels, where cochlear amplifier gain is the greatest. However, given that some important MOCR effects are operative at higher sound levels, it is essential to understand its effects at higher stimulus levels. Click-evoked (CE)OAEs may be useful for probing MOCR effects at these higher levels, because by leaving sufficiently long time intervals between presentations of acoustic clicks, relatively high stimulus levels can be used without the stimulus itself activating the MOCR or MEMR.

Varying the amplitude of a click stimulus in the ear canal creates complex interactions that occur between active and passive vibrations of cochlear structures, spread of excitation along sensory epithelia, and the generation regions and generation mechanisms of CEOAEs. The excitation pattern for low- and moderate-level pure tones is a tall and broad peak that is localized to a given cochlear characteristic frequency place that mediates generation of stimulus-frequency (SF)OAEs (4, 5). CEOAEs and SFOAEs appear to be generated by the same mechanism (6). The tall and broad peak excitation results from cochlear amplification, with amplification generated from distributed locations up to 2 octaves basal of the peak (4, 7). Although amplification can exist basal of the CF place, the effects accumulate as the traveling wave moves apically, such that the largest effects are observed near CF place (see Ref. 4). Thus, the effects of the MOCR will vary among CEOAE components if the components originate from different cochlear regions. MOCR effects on shorter latency components, in particular, indirectly may suggest limits on how far basal these components are generated with reference to the CF place.

As click amplitude increases, additional CEOAE energy is observed with shorter time delays (8–10). Such earlier-occurring portions of CEOAE waveforms have been referred to as “short latency” (SL) components, whereas later-occurring portions of CEOAE waveforms having latencies consistent with generation near CF place have been referred to as “long latency” (LL) components. SL components grow less compressively than LL components (8, 9, 11). One interpretation of the latency and growth attributes of SL components is that SL components originate basal of CF, where motion measurements have shorter latencies and response growth is more linear. This has implications for measurement of MOCR effects as a function of stimulus level: as higher stimulus levels are used, the resulting CEOAE waveforms as a whole contain a greater proportion of SL components. This early energy should show smaller MOCR-related changes (see Ref. 12).

We used time-frequency windowing to separate CEOAEs into SL and LL components. We characterized the effects of MOCR on each component across a 45 dB range of click stimulus levels. If SL CEOAEs are generated basal to the CF region but from regions which still receive significant cochlear amplification, then they should show smaller but still significant MOCR-induced changes. We tested this hypothesis and also provided a detailed description of the effect of MOCR on SL and LL CEOAE growth functions without activating the MEMR.

METHODS

Participants

Inclusion criteria.

This study measured MOCR from young adults with normal hearing. Ear canals were required to be free of any cerumen accumulation and eardrums were required to appear normal via visual otoscopic inspection. Middle ear function (type A, with peak pressure between −100 and +10 daPa, and static admittance between 0.2 and 1.8 mL) and ear canal volume (between 0.5 and 1.8 mL) were required to be normal as assessed by 226-Hz tympanometry. Pure-tone air conduction behavioral thresholds were required to be ≤15 dB HL at octave frequencies from 0.25 to 8 kHz, bilaterally. CEOAEs were measured using a screening protocol similar to the test paradigm used in this study, and participants were required to have CEOAE signal-to-noise ratios (SNRs) ≥8 dB. Additionally, eligible participants reported no history of the following: known or suspected hearing loss, difficulty hearing in quiet and noisy environments, ear surgery (excluding ventilation tubes as a child), significant noise exposure without the use of hearing protection, tinnitus, dizziness, or use of ototoxic medications.

Study cohort.

Participants were recruited primarily from the undergraduate student body of the University of Iowa. Thirty-five individuals (24 females, mean = 21.3 yr, SD = 2.7 yr, range = 18 to 30 yr) met the inclusion criteria, as well as demonstrated the satisfactory stimulus level stability across the recording period (described under Stimulus levels below). Of all the hearing thresholds screened (35 participants × 2 ears × 6 frequencies = 420), the vast majority (97.6%) passed at 10 dB HL, whereas the remaining 2.4% passed at 15 dB HL. The ear from which CEOAEs were recorded was pseudorandomly assigned to achieve balance between left and right ears. In the final cohort, 18 (51.4%) right ears were tested. The research protocol was approved by the University of Iowa Institutional Review Board, and written informed consent was obtained from all participants.

Equipment

Testing took place inside a double-walled sound-treated room. Hearing screenings were performed using standard clinical equipment, which included a tympanometer (Madsen OTOflex 100, GN Otometrics, Schaumburg, IL) and a clinical audiometer (GSI-61, Grason-Stadler, Eden Prairie, MN). Presentation of test stimuli (clicks and noise) and recording of CEOAEs was made using a personal computer running custom software (ARLas; 13). Stimuli were generated in MATLAB (The MathWorks, Natick, MA), digitized at 96 kHz, presented to a 24-bit sound card (RME: Fireface 802, Haimhausen, Germany), and routed to an acoustic probe system (ER-10X; Etymotic Research, Elk Grove Village, IL). Ear canal pressure measurements were made using the acoustic probe system connected to the 24-bit sound card.

Test Paradigm

Stimulus levels.

Clicks and broadband noise had bandwidths extending from 1 to 7.5 kHz. Ear acoustics vary across individuals, resulting in sound pressure levels at the tympanic membrane that can vary considerably from those measured by the probe microphone positioned near the entrance of the ear canal. Calibration in terms of forward pressure level (FPL) reduces this source of variability, ensuring a more uniform stimulus across listeners. Clicks were calibrated to have a flat FPL amplitude spectrum at the eardrum (14–18). Average standard deviation of stimulus level within the frequency passband of the clicks was 0.46 dB, and the maximum deviation from the mean level in any participant at any frequency was 1.2 dB. Clicks were presented at five stimulus levels spaced 9 dB apart. The click root-mean-squared (RMS) amplitudes were 35, 44, 53, 62, and 71 dB FPL. The corresponding level per cycle values (i.e., FFT amplitude with reference to 1-Hz bin width) were 38.8 dB lower. Expressed in SPL (instead of FPL), all values were ∼2 dB higher. Expressed in sensation level, the clicks ranged from approximately 9 to 54 dB SL. Throughout the rest of this paper, click levels are referred to using their nominal peak-to-peak levels of 55, 64, 73, 82, and 91 dB ppFPL.

The MOCR activator, broadband noise, was also calibrated to have flat FPL at the eardrum, and this was achieved for the contralateral noise with an exactness similar to that described for the clicks. An error in the computer program resulted in applying each participant’s contralateral in situ calibration to the ipsilateral ear as well, and this error was not found until data collection was completed. As a result, the ipsilateral noise was not flat (average standard deviation of level across frequency = 5.13 dB vs. 0.41 dB for contralateral noise), though the overall levels were similar (mean absolute difference between ipsilateral and contralateral levels = 1.76 dB, SD = 1.18). We believe this error had a negligible impact on the findings of this study. The presented noise levels had a nominal RMS amplitude of 49 dB FPL (52 dB SPL), which corresponded to ∼45 dB SL.

To maintain these stimulus levels throughout the experiment, three strategies were employed. First, modified earmuffs (of the type typically used for hearing protection) were used to hold the ER10X probe assembly securely in place. The cords connected to the probes were suspended from the ceiling of the sound booth to reduce weight on the probe assembly, as well as reduce acoustic noise associated with contacting participants’ torsos. Participants were seated in a comfortable reclining chair with good head and neck support. This strategy minimized probe movement and slippage during the experiment. Second, data were acquired in 1.6-min blocks (described under Presentation below). Participants were instructed to sit as quietly and as still as possible during each recording block. At the end of each recording block, a dialog box was displayed on a computer screen, asking participants to swallow and then press a button to continue. Participants were encouraged to make any necessary movements or noise during this time between recordings. This strategy helped ensure optimal recording SNR and minimized middle ear pressure changes across the experiment. It also helped regulate participant arousal state, helping keep participants awake and alert. Third, in situ calibration was performed in 10-min intervals. Each new calibration recalculated ear canal forward and reverse pressure levels and was used to calibrate stimulus levels in the next block. This minimized slow drifts in stimulus level across the experiment. The combination of these strategies resulted in stable stimulus levels. Across 74 min of recording, the largest standard deviation of the clicks from their mean levels in any participant was 1.8 dB. The average standard deviation across all participants was 0.43 dB.

Presentation.

Click rate has an important effect on observed MOCR-related changes in CEOAEs. Because the MOCR is activated by sound, the acoustic probe itself may activate the MOCR (19, 20). If this happens, then subsequent additional elicitation using the acoustic activator can be expected to result in less MOCR-induced change than would have occurred had the activator been presented with a probe that did not cause any activation. By using sufficiently slow click rates, CEOAE measurement paradigms may be designed to avoid the problem of the probe partially activating the MOCR. The rate may be chosen based on the known MOCR onset and offset delays, which are ∼25 ms, independent of elicitor level (21). If sufficient time is allowed between successive stimulus presentations, a click probe may be presented and the evoked CEOAE measured in the subsequent 20 ms, before any activation of the MOCR by the click probe itself takes place. Additionally, the potential effect of clicks presented concurrently with the activator should be considered. Each click presentation is a transient increase in the ongoing activator level, potentially causing a brief increase in MOCR activation. Two MOCR time courses, “fast” and “slow,” have been described (22). Fast effects have been reported to be on the order of 100–200 ms (21, 23–26). By using slow click presentation rates of 5/s (200 ms interstimulus intervals), the influence of the clicks on the fast and slow MOCR time course would be expected to be reduced to a negligible level. Extending the logic applied above, the click and CEOAE would be measured before any transient increase in MOCR caused by the click itself, and any transient increase in MOCR activation caused by the click would be expected to decay before the next click presentation. This was also anticipated to hold true for middle-ear muscle reflexes (MEMR), which have similar time constants (27, 28).

Pilot data showed that spacing the clicks 200 ms apart yielded different results than the 50-ms spacing employed by many OAE-based MOCR studies, and that results obtained using 400-ms spacing did not differ from those obtained using 200-ms spacing. Accordingly, clicks were presented at a rate of 5/s (200 ms interstimulus interval) to obtain the experimental data. We hereafter refer to each recorded 200-ms interval as a buffer. Each buffer contained the click stimulus and the CEOAEs. The number of click presentations was varied as a function of click level to obtain an approximately equal SNR at each level. The number of presentations was arrived at based on theoretical expectations about the effect of synchronous averaging, expected compressive growth of CEOAE amplitude, and pilot data. Presentation order of click levels was interleaved to avoid order effects. The order of presentation level was also chosen to spread the higher levels apart in time from each other, thereby minimizing any cumulative effect of repeated, high-level stimulus presentations. The lowest three levels and the highest three levels of clicks were arranged into two stimulus trains, as shown in Fig. 1A. The level 73 dB ppFPL was present in both trains. The total number of click presentations (and therefore the number of recorded buffers) in each condition (with and without noise) for the five stimulus levels (from lowest to highest) was 5,568, 696, 5,742, 168, and 42.

Figure 1.

Test paradigm. A: illustration of a click train. Clicks were spaced 200 ms apart. Trains consisted of clicks at three stimulus levels, spaced as shown. A “low-level train” contained levels of 55, 64, and 73 dB ppFPL, whereas a “high-level train” contained levels of 73, 82, and 91 dB pp FPL. The illustration is of one click train but schematizes both low- and high-level trains. An 800 ms of silence preceded the onset of the clicks, and 600 ms of silence followed the offset (i.e., a 1.4-s intertrain interval). Each train was 16 s long and contained 64 low-, eight medium-, and two high-level clicks. Note that click levels of 73 dB ppFPL were presented in both low-level and high-level trains, occupying either the highest or lowest click level in the train, respectively. During post hoc data analysis, the 73 dB ppFPL responses from both trains were combined after ensuring that there were no systematic differences between the two (mean absolute difference = 0.72 dB, SD = 0.56). B: stimulus block. Six click trains were grouped together to form a block. Broadband acoustic noise was interleaved throughout the block, presented simultaneously with every other click train, as shown. Clicks and noise were presented through separate loudspeakers in the ipsilateral probe assembly, and noise was presented through a single loudspeaker in the contralateral probe assembly. Each block was 96 s long. C: full recording set. A total of 36 blocks were recorded from each participant. Of the 36 blocks in the experiment, 29 were low-level click trains and seven were high-level click trains. D: inset showing the ramping on and off of ipsilateral noise during the with-noise conditions. Ipsilateral noise was gated off 10 ms prior to each click presentation and then gated back on 20 ms afterward. Vertical dashed lines indicate the temporal locations of click presentations. In this panel, time is shown relative to the temporal location of the first dashed vertical line. ppFPL, peak-to-peak forward pressure level.

Acoustic noise, interleaved across blocks (Fig. 1B), was presented bilaterally to activate the MOCR. The noise was presented for 1.4 s before the presentation of clicks with noise to allow time for the MOCR (the “fast” MOCR effects; Ref. 21) to activate before recording. A silent period of 1.4 s was also presented before the presentation of clicks without noise. Contralateral noise was presented continuously during each 16-s with-noise condition. During this 16-s period, ipsilateral noise was gated off 10 ms before each click presentation and then gated back on 20 ms afterward (Fig. 1D). This allowed sufficient time for the click and CEOAE to be recorded without acoustic interference. The ipsilateral noise was on for a total of 13.8 s (86%) of each 16-s with-noise condition. The temporal gating meant that the ipsilateral noise was effectively modulated at ∼5 Hz. Previous research suggests that the MOCR acts as a low-pass system with a cutoff frequency of ∼1 Hz (29, 30). The slow modulation of the ipsilateral noise in the current experiment was expected to have a negligible effect. A total of 36 stimulus blocks (Fig. 1C) were presented to each participant.

Post Hoc Signal Processing

Recordings were sorted by click level and condition (with and without the noise activator). Waveforms were bandpass filtered using Kaiser-based finite impulse response (FIR) filters with cutoff frequencies of 0.75 and 8 kHz. The filtered waveforms were corrected for filter group delay. A quartile-based artifact rejection algorithm was applied to the energy in each waveform (8), after which all remaining waveforms at each level/condition combination were subaveraged into a matrix of 30 buffers.

Three temporal windows were applied to the buffers to isolate different aspects of the recorded sound pressures. Time zero was defined as the location of the largest peak in the recorded click stimulus waveform. The Stimulus Window was centered at time zero and extended from −2 to 2 ms. The ends of this window were ramped on and off using raised cosine ramps of 0.5-ms duration. The CEOAE Window started at time zero and ended at 18 ms. The first 2 ms were zeroed, followed by a 2-ms onset ramp. The offset was also ramped with a 2-ms raised cosine ramp. The SSOAE Window, which was applied only to the buffers with no MOCR activator, started at time 150 ms, ended at 170 ms, and included 2 ms raised cosine ramps on both ends. Additional signal processing specific to each analysis window is described below.

Stimulus window.

Simultaneous monitoring of MEMR and MOCR is critical for ruling out MEMR contribution to changes in OAEs. Activation of the MEMR stiffens the ossicular chain, causing a change in middle ear impedance, which alters the amount of reflected sound pressure as a function of frequency. Frequencies around 1–2 kHz often show a decrease in reflectance amplitude, whereas frequencies around 2–3 kHz show an increase (see, e.g., Refs. 28 and 31). The precise frequencies where increases and decreases occur is variable across individuals. For this reason, the presence of MEMR should be evaluated in relatively narrow frequency bands to ensure that increases in one frequency band are not cancelled by decreases in another. In the present study, waveforms in the Stimulus Window were examined for the presence of MEMR in 1/6-octave frequency bands using statistical resampling methods described by Mertes and Goodman (32). None of the participants tested positive for MEMR in any of the frequency bands tested (1–3.1 kHz in 1/6-octave steps) at any click level. All changes in CEOAEs found in subsequent analyses were therefore considered to be due to MOCR, with no contribution from MEMR.

SSOAE window.

Spontaneous otoacoustic emissions (SOAEs) and synchronous-spontaneous (SS)OAEs are often present in individuals with good hearing thresholds (33–36). SSOAEs are CEOAEs that persist for longer than expected, operationally defined as having significant energy lasting longer than 20 ms after the click stimulus. SSOAEs are typically found in the same narrow frequency bands as spontaneous (S)OAEs (37). Because the potential effects on the CEOAEs of interest (those in the CEOAE window) are the same, we did not differentiate between SSOAEs and SOAEs, and hereafter we refer to them simply as SSOAEs.

Waveforms in the SSOAE Window of each participant were examined, and these emissions were identified using the coherence synchrony measure described by Keefe (35). Frequencies at which SSOAE energy was present were flagged so that their influence on CEOAE results could be minimized through the use of weighting functions, described in the next section. Of 35 participants, 21 (60%) were identified as having one or more SSOAE. This percentage is lower than the 85% and 72% reported by Keefe (35) and Sisto et al. (33), respectively. A likely reason for our lower percentage is the much longer time interval between clicks (200 ms in the current study vs. 42 ms or 80 ms in the Keefe and Sisto et al. studies).

SSOAEs contain energy extending into the following stimulus window. If large SSOAEs were to show a change in amplitude, this could potentially be mistaken as evidence for MEMR. Although this is possible on theoretical grounds, the fact that none of our participants had MEMR suggests that this did not occur in practice. Whether or not the presence of SSOAEs significantly influences MOCR test results may depend on what is being tested. Data suggest no systematic effect of SSOAEs on CEOAE amplitude changes (36). In contrast, group delay and phase measurements are more sensitive (35).

CEOAE window.

Each CEOAE waveform was decomposed using a bank of gammatone filters, time windowed, and then recombined back into a single waveform. The filters were based on models of human auditory filters following Glasberg and Moore (38). Although a gammatone model was used for the present report, the results were not strongly dependent on the particular filter bank used, and similar results were obtained using a bank of FIR filters and a wavelet analysis (39). The filter bank used in this study consisted of eight filters with center frequencies spaced in 1/4th octave steps from 1 to 3.4 kHz. This frequency range was chosen based on previous studies showing that MOCR effects on OAEs decline at higher frequencies (40–42).

The outputs of each filter were time windowed to include SL or LL components. Temporal cutoff values defining SL and LL components were determined in the following way: CEOAEs evoked by the lowest click level from all 35 participants were examined. The envelopes of the filtered CEOAEs were amplitude normalized and averaged to obtain mean CEOAE delays for each frequency. Upper and lower bounds of expected LL delays were taken as the times at which the averaged envelope amplitudes were reduced to 70.7% of the maximum (half power). This processing removed any “very long latency” CEOAE components that would include multiple internal reflections, as well as some energy associated with SSOAEs and SOAEs. The upper bounds for SL components were chosen to avoid any stimulus ringing (verified using an IEC-711 artificial ear). The use of nonlinear CEOAE extraction paradigms could have allowed the inclusion of earlier SL energy but at the loss of linear-growing portions of the emissions. The cutoff values obtained in this manner were similar to those reported by Mertes and Goodman (43), which in turn were based on data from Shera et al. (44).

After time windowing, the filtered waveforms were added back together to yield two composite waveforms. The composite waveform SL included only short-latency energy from 1 to 3.4 kHz. The composite waveform LL included only long-latency energy from 1 to 3.4 kHz. An example of the time-frequency processing described above is shown in Fig. 2. The processing was applied to the CEOAEs obtained at each click level and condition.

Figure 2.

Separation of CEOAE waveforms into short latency (SL) and long latency (LL) components. Data are shown for two click levels from one representative participant. A: CEOAE waveforms evoked by the lowest click level after gammatone filtering. Y-axis values are the center frequency of the filters. Solid, red sloping lines indicate the start and the end of the waveforms retained for analysis, as a function of frequency. Dashed sloping red lines denote the boundary between SL and LL. B: CEOAE waveforms evoked by the second-highest click level after filtering. Layout is the same as in A, except that the waveforms were scaled down in amplitude for visual clarity. C: waveforms obtained by summing the filtered, time-windowed waveforms from the lowest click level (shown in A). D: waveforms obtained by summing the time-windowed, filtered waveforms from the second-highest click level (shown in B). CEOAE, click-evoked otoacoustic emission.

Magnitude and Phase Calculations

Let vector p[n] denote a discrete-time CEOAE pressure waveform at sample n, p of length N samples. The complex analytic signal s[n] is defined using the unit imaginary number j and the discrete Hilbert transform p_h[n] of p[n], by

$$s\left[n\right]=p\left[n\right]+j{p}_h\left[n\right].$$ (1)

For waveforms that are characterized by a narrow band of frequencies at any given instant of time (such as the time-frequency windowed CEOAEs in this study), the analytic signal provides a straightforward calculation of instantaneous amplitude (waveform envelope) and phase,

$$a\left[n\right]=\sqrt{{\mathrm{R}\mathrm{e}\left(s\left[n\right]\right)}^2+{\mathrm{I}\mathrm{m}\left(s\left[n\right]\right)}^2}$$ (2)

$$\phi \left[n\right]={\tan }^{-1}\left(\mathrm{I}\mathrm{m}\left(s\left[n\right]\right)/\mathrm{R}\mathrm{e}\left(s\left[n\right]\right)\right).$$ (3)

The instantaneous frequency of the waveform is minus the first derivative of unwrapped phase with respect to time, $f=-\left(\frac{1}{2\pi }\right)d\phi /dt$. Phase unwrapping can be problematic when CEOAEs are contaminated by noise. We calculated instantaneous frequency without unwrapping phase (35),

$$f\left[n\right]=\frac{1}{2\pi }\left(\frac{U\left(dV/dt\right)-V\left(dU/dt\right)}{U^2+V^2} \right),$$ (4)

where U = Re(e^jφ) and V = Im(e^jφ). In Eq. 4, variables U and V are functions of time n, but this explicit dependency has been omitted for clarity. The lower-case letter d represents differentiation, which was implemented as a discrete first-order difference (Matlab’s gradient function). A weighted smoothing spline with weights equal to a²[n] was used to obtain the final estimate of frequency as a function of time (f[n]).

In the calculations that follow, a tilde (∼) accent represents quantities computed from CEOAE measurements obtained with a noise activator, and no accent represents quantities computed from CEOAE measurements obtained in quiet. A frequency vector F[k] was created, k representing frequencies from 1 to 3 kHz spaced in equal, 1-Hz steps. Using a spline interpolation function and the appropriate estimate of frequency as a function of time (f[n] or $\tilde{f}\left[n\right]$), instantaneous amplitude as a function of time was mapped onto the frequency vector F: a[n] → A[k] and ã[n] → Ã[k]. The MOCR-induced change in CEOAE amplitude as a function of frequency was defined as Δ_A[k] = Ã[k]/A[k]. The vector of amplitude ratios Δ_A was reduced to a scalar value by taking a weighted average across frequency,

$${\delta }_A= {\displaystyle \sum }_{k=1}^K\left({\text{Δ}}_A\left[k\right]\circ W\left[k\right]\right)/{\displaystyle \sum }_{k=1}^{K}W\left[k\right],$$ (5)

where the symbol ° indicates component-wise vector multiplication and W is a weighting function. The weighting function was the square of the amplitude vector obtained in quiet (A²), modified by setting the values to zero at frequencies that were previously identified as having SSOAEs present. The amplitude vectors A and à were similarly reduced to scalar values.

Phase changes were computed in a similar manner to amplitude, but with modifications to account for their circular distribution. Phase differences were first calculated as a function of time, ${\text{Δ}}_{\text{φ}}\left[n\right]=\tilde{\phi }\left[n\right]-\text{φ}\left[n\right]$. Using spline interpolation and a joint estimate of frequency as a function of time (f[n] and $\tilde{f}\left[n\right]$), instantaneous phase difference as a function of time was mapped onto the frequency vector F: Δ_φ[n] → Δ_Φ[k]. The vector Δ_Φ was reduced to a scalar value δ_Φ by taking a weighted circular mean across frequency,

$${\delta }_{\mathrm{\Phi }}= arg\left({\sum }_{k=1}^{K}W\left[k\right]e^{\left(j{\mathrm{\Delta }}_{\mathrm{\Phi }}\left[k\right]\right)}\right).$$ (6)

The amplitude and phase values obtained in this way were similar in overall form to those obtained using conventional Discrete Fourier Transform (DFT) techniques. The technique described here is advantageous for nonstationary signals like CEOAEs, because it does not integrate across the length of the entire waveform to calculate the amplitude and phase of each frequency component. This provides an SNR improvement by not integrating noise across times at which the frequency of interest is not actually present (see also Ref. 45). The method described here resulted in higher amplitude estimates for shorter duration, higher frequency CEOAE components than conventional DFT, leading later to a more balanced averaging of MOCR effects across frequency. After this averaging, our CEOAE measurements were still weighted slightly towards the lower frequencies of the included range. If the “center frequency” after averaging is taken to be the mean (across participants) of the energy-weighted mean frequency of each participant, then the center frequency was 1,807.9 Hz (SD = 333.8 Hz). A uniform distribution (i.e., equal weighting across all included frequencies, but with frequencies spaced on a log₂ axis), would have yielded a center frequency of ∼1,961 Hz (SD = 684 Hz).

Combining Magnitude and Phase Changes

Much of the previous work on human MOC reflex effects has not explicitly considered phase information. From both basic science and clinical perspectives, it is important to understand phase effects. As described by Mertes and Goodman (32), relative changes in amplitude (δ_A) and phase (δ_Φ) can be combined into a single complex value $\delta ={\delta }_A{e}^{j{\delta }_{\text{Φ}}}$. Expression in this form provides a conceptual and mathematical link to the “total quantity of change” metric common the MOCR literature. Total quantity of change is a real value that includes both amplitude and phase, while δ is a complex value. When δ is plotted as a point on the complex plane, total quantity of change is the length of a vector that has been “turned” so that it is pointing to δ not from the origin, but from the point [1,0] on the unit circle (see Ref. 32, supplemental digital content 4). For this reason, Mertes and Goodman called this metric “turned delta” and expressed it as $ƍ=\left||\delta -1|\right|$. Total quantity of change is more commonly calculated in the frequency domain as:

$$ƍ=\sqrt{\sum |{\left|\stackrel{\sim }{P}-P|\right|}^2}/\sqrt{\sum {\left|\left|P\right|\right|}^2},$$ (7)

where the uppercase P denotes the CEOAE complex spectrum (obtained by DFT), and the summation sign is understood to extend across the frequency interval of interest (see, e.g., Refs. 46 and 47). If the waveform is bandpass filtered to contain only the frequencies of interest, the same value can be calculated in the time domain:

$$ƍ=\sqrt{\sum {\left(\stackrel{\sim }{p}-p\right)}^2}/\sqrt{\sum p^2},$$ (8)

where lowercase $\tilde{p}$ and p denote the CEOAE waveform measured with and without noise, and the summation sign is understood to extend across the time interval of interest. The total quantity of change metric is interpreted as the proportion (or multiplied by 100, the percentage) change due to MOCR.

An important property of $ƍ$ is that it does not discriminate between the direction of change. Magnitude increases and decreases, as well as phase leads and lags are all made positive by squaring operations, so that changes in both directions are additive and always yield a positive result. Consequently, total quantity of change must always be equal to (in the case where all phase changes are precisely zero) or greater than changes in amplitude alone (δ_A). In this study, we calculated total quantity of change as $ƍ=\left||{\delta }_A{e}^{j{\delta }_{\mathrm{\Phi }}}-1|\right|$, so as to include the effects of the processing described in the previous section. This also allowed comparison of the relative contributions of magnitude alone, |δ_A − 1|, and phase alone $\left|e^{j{\delta }_{\text{Φ}}}-1\right|$.

Equivalent Input Attenuation

MOCR-induced changes in amplitudes are most often quantified as the amount by which OAE amplitude was reduced at a given stimulus level. This change is usually expressed as a percentage change relative to the amplitude measured without a noise activator (or equivalently, as a decibel value). We refer to these measures as “output” attenuation, because they describe how much the cochlear output (the CEOAE in this case) was reduced. MOCR-induced changes in amplitude can also be quantified as the amount by which the stimulus level must be lowered when presented alone (i.e., without a noise activator) to achieve the same OAE amplitude obtained when presented with noise (reviewed in Refs. 2, 48, and 49). This “equivalent input attenuation” can also be expressed as a relative percent change.

To calculate equivalent input attenuation, estimates of amplitude growth are needed across a range of stimulus levels. Consistent with previous reports, we observed that our CEOAEs evoked by low-level stimuli tended to grow approximately linearly, followed by a region of compressive growth (10, 50, 51). Growth patterns at the highest stimulus level tested were variable across participants, with some growth becoming more linear and some remaining compressive. Both linear and compressive growth can be seen across traveling wave motion measurements with stimulus level (52). Given these growth characteristics, a smoothing spline was chosen as an appropriate fitting function.

We obtained amplitude growth functions by fitting CEOAE amplitude as a function of stimulus level. Separate fits were obtained for SL and LL components with and without noise (four separate fits for each participant). The fits were calculated using a smoothing spline, weighted by the SNR at each stimulus level. To calculate equivalent input attenuation at the lower stimulus levels used, the growth functions must be extrapolated to lower levels than were tested. In our data set, the lowest stimulus level tested (55 dB ppFPL) was ∼10 dB SL, making linear growth a reasonable assumption for extrapolating to stimulus levels that would have been at or below threshold. Curve fitting examples for two participants are shown in Fig. 3.

Figure 3.

A–D: examples of curve fitting to obtain amplitude growth functions. Each row shows data from one participant. Left and right columns show short latency (SL) and long latency (LL) responses, respectively. Red circles show CEOAE amplitudes obtained without noise activators. Blue triangles show amplitudes obtained with noise activators. Red and blue solid lines show the obtained amplitude growth functions. Dash-dot black lines show the noise floors. For visual comparison to linear growth, dashed gray lines toward the upper left of each panel show unity slopes. In D, circled lower-case letters and black arrows illustrate amplitude output attenuation (b-a) and equivalent input attenuation (c-b) for a single stimulus level. The fitting routine was able to adequately fit the variety of amplitude growth functions seen across participants. CEOAE, click-evoked otoacoustic emission; ppFPL, peak-to-peak forward pressure level.

Figure 3D illustrates calculation of amplitude output attenuation and equivalent input attenuation for a single stimulus level. The red circle nearest lowercase letter a is CEOAE amplitude obtained without noise, whereas the blue triangle nearest lowercase letter b is CEOAE amplitude obtained with noise. The length of the black, downward pointing arrow indicates the amount by which the amplitude (output) was reduced with noise. Since amplitude is given here in decibels, output attenuation, the percentage by which the output was reduced, would be calculated as 100(1 − 10^{((b − a)/20)}), with a and b here referring to CEOAE amplitude values on the y-axis. The length of the black arrow pointing leftward from lowercase letter b to where it intersects the red fitted growth function line near lowercase letter c indicates equivalent input attenuation, the amount by which the stimulus, presented without noise, would have to be reduced to yield the same CEOAE amplitude obtained with noise. Equivalent input attenuation would be calculated as 100(1 − 10^{((c − b)/20)}), with b and c here referring to stimulus levels on the x-axis. Graphically, amplitude output attenuation expresses the amount of shift on the y-axis, while equivalent input attenuation expresses the amount of shift on the x-axis.

An important property of equivalent input attenuation is that when the growth functions are both linear, input attenuation is equal to output attenuation. When the growth curves are compressive, input attenuation is greater that output attenuation, and this effect increases with the amount of compression. It should be kept in mind that for portions of curves representing highly compressive growth (typically at high stimulus levels), equivalent input attenuation may be less meaningful or even potentially misleading. For example, highly compressive portions of curves may not intersect within the range of biologically plausible stimulus levels, implying that stimulus level in the with-noise condition could never increase enough to overcome the effects of MOCR at higher presentation levels. Small amounts of “rollover” in the magnitude growth functions at the highest stimulus levels would also preclude simple interpretation. In summary, measures of equivalent input attenuation are probably most useful when considering low- to mid-level stimuli.

RESULTS

A criterion of SNR ≥ 8 dB was required for measurements to be included in the reported results. This criterion was applied separately for each subject at each stimulus level. For a participant’s data to be included at any given level, their measurements with and without the noise activator for SL and LL components all had to meet the SNR criterion. This allowed valid paired comparisons to be made between different measurements obtained at any given stimulus level. Included participants were not required to meet the SNR criterion at all stimulus levels, because including only participants who had adequate SNR in all measurement conditions too severely reduced the number of data points available for analysis. A result of the SNR criterion was a variable number of participants at each stimulus level. These are shown at the bottom of Fig. 4, A and B, but apply to all reported results. SNRs of included data points were similar across stimulus level, and averaged 18.6 dB for LL components measured without noise (SD = 4.46, min = 9.2, max = 30.1) and 14.8 dB for SL components measured without noise (SD = 3.04, min = 8.4, max = 22.9).

Figure 4.

Effects of MOCR on CEOAE amplitude growth. Top: box plots show CEOAE amplitude as a function of stimulus level for short latency (SL, A) and long latency (LL, B) components. Amplitudes obtained without noise are plotted offset to the left in red, and amplitudes obtained with noise are plotted offset to the right in blue. Box plots are overlaid with curves showing median amplitude growth functions. For visual comparison to linear growth, dashed gray lines toward the upper left of the panels show unity slopes. The number of participants included at each stimulus level is shown by numbers enclosed in parentheses along the bottom of A and B. Middle: C and D show slopes of CEOAE amplitude growth functions. Layout is similar to the top row. Bottom: E shows changes in slope (slope obtained with noise activator minus slope obtained without noise activator) for SL (black) and LL (gray) components as a function of stimulus level. Positive values on the y-axis indicate that slope increased (became less compressive) with MOCR activation. F compares changes in slope between SL and LL components (the differences between the pair of black and gray boxplots at each stimulus level). Positive values on the y-axis indicate that MOCR activation caused a larger slope change in LL components than SL components. These data show that, on average, SL components grew more linearly than LL components with click level. MOCR activation caused growth functions to become more linear for both SL and LL components. CEOAE, click-evoked otoacoustic emission; MOCR, medial olivocochlear reflex; ppFPL, peak-to-peak forward pressure level.

CEOAE Amplitude Growth Functions

The growth of CEOAE amplitude as a function of level is shown in Fig. 4, A and B, with corresponding slopes of the growth functions shown in Fig. 4, C and D. For both SL (A and C) and LL (B and D) components, growth was closer to linear at low stimulus levels and became more compressive as stimulus level increased. Growth curves were steeper for SL components than LL components. For SL components obtained without a noise activator, the steepest and shallowest median slopes were 0.85 and 0.48, respectively. For LL components obtained without a noise activator, the steepest and shallowest median slopes were 0.69 and 0.15, respectively.

The overall effect of MOCR activation (blue lines in Fig. 4, A–D) was to reduce the amplitudes and make amplitude growth less compressive. The differences between slopes obtained with a noise activator and slopes obtained without noise are shown in Fig. 4E. Wilcoxon Signed-Rank tests were conducted to compare the difference in slope obtained with and without the noise activator at each stimulus level. The statistical significance of these comparisons was assessed at a nominal α = 0.05 level, corrected using the false discovery rate method for performing multiple comparisons (53, 54). For SL waveforms (black box plots in Fig. 4E), slopes were significantly steeper in the presence of the noise activator for the middle three stimulus levels (all P ≤ 0.002), but not for the lowest and highest stimulus level (P = 0.41 and P = 0.48, respectively).

For the LL waveforms (gray box plots in Fig. 4E), slopes were significantly steeper in the presence of the noise activator, except at the lowest level (P = 0.056; all other levels P ≤ 0.0003). A visual comparison of SL versus LL slope changes in Fig. 4E (compare pairs of black and gray boxplots) suggests that for all but the highest stimulus level, SL an LL components showed similar amounts of slope change. These differences (LL slope change minus SL slope change) are plotted in Fig. 4F. Wilcoxon Signed-Rank tests showed a significant difference between LL an SL at the highest stimulus level (P = 0.038). The differences at the other stimulus levels failed to show significant differences (all P ≥ 0.37).

Combined across SL and LL components at the three middle stimulus levels, the median slope change was 0.11 (Mean = 0.12, SD = 0.03). The relative size of this nearly constant slope difference varied as a function of stimulus level. For the steepest median LL slope obtained without noise (0.69), this represented a MOCR-mediated increase of 16%. For the shallowest median LL slope (0.15), this represented a MOCR-mediated increase of 73%. Overall, the information provided here is the first detailed description of SL and LL CEOAE amplitude growth functions in humans and shows that MOCR activation resulted in significantly steeper growth for both components across a wide range of stimulus levels.

Amplitude Output Attenuation and Equivalent Input Attenuation

The percent attenuation of CEOAE amplitude output (black boxplots) and equivalent input (gray boxplots) as a function of stimulus level is shown in Fig. 5. For both SL (Fig. 5A) and LL (Fig. 5B) components, amplitude output attenuation was largest at the lowest stimulus level and decreased as stimulus level increased. A similar overall pattern was seen with equivalent input attenuation for SL components. However, for LL components, equivalent input attenuation remained more constant, increasing slightly with stimulus level (Fig. 5B, gray boxplots). For SL components, median attenuation values (at lowest and highest stimulus levels, respectively) ranged from 40% to 12.5% for output attenuation and from 48.1% to 29% for equivalent input attenuation. For LL components, median attenuation values ranged from 43.6% to 17.7% for output attenuation and from 57.9% to 69.2% for equivalent input attenuation.

Figure 5.

Top: percentage attenuation in CEOAE output amplitude (black) and equivalent input (gray) as a function of stimulus level for short latency (SL, A) and long latency (LL, B) waveforms. Black boxplots show percent amplitude output attenuation at each stimulus level. Gray boxplots show equivalent input attenuation. For visual clarity, boxplots are offset to the left (output) and to the right (equivalent input) of stimulus level. Positive values on the y-axes indicate that CEOAE amplitude decreased with MOCR activation, with larger values indicating larger reductions. Bottom: differences in attenuation at long latency versus short latency components (LL − SL). C shows differences in amplitude output attenuation. Positive values indicate that LL attenuation was greater than SL attenuation. D shows differences in equivalent input attenuation, with positive values again indicating larger attenuation for LL components. MOCR effects on LL components were larger than on SL components, and equivalent input attenuation was larger than output attenuation. CEOAE, click-evoked otoacoustic emission; MOCR, medial olivocochlear reflex; ppFPL, peak-to-peak forward pressure level.

Equivalent input attenuation was greater than output attenuation at every stimulus level for both SL and LL components (black vs. gray paired boxplots in Fig. 5, A and B), and these differences were significant by Wilcoxon Signed-Rank tests (all P ≤ 0.001). For reporting the size of differences between two quantities that are percentages (e.g., two different measures of percent attenuation), we follow a convention of giving percentage point differences, followed by the percent difference, calculated as 100(|n₁ − n₂|/(n₁ + n₂)/2), in parentheses. For LL components, the size of the difference between output attenuation and equivalent input attenuation increased monotonically with stimulus level, with median differences of 14.2 percentage points (28.0 percent difference) at the lowest stimulus level and 51.4 percentage points (118.3 percent difference) at the highest stimulus level. For SL components, the differences were smaller, ranging from 8 to 16.6 percentage points (18.2 to 79.8 percent difference).

The amount of attenuation measured for LL versus SL components is shown for amplitude output in Fig. 5C. LL components showed larger median amounts of attenuation than SL components at the lowest four stimulus levels. These differences were significant by Wilcoxon Signed-Rank tests for stimulus levels of 55, 73, and 82 dB ppFPL (all P ≤ 0.03), but failed to meet the criterion for statistical significance at 64 dB ppFPL (P = 0.16). At 64 dB ppFPL, the median LL-SL difference was positive and of similar size as the surrounding levels, but for unknown reasons the variability was larger. Averaged across the four lowest stimulus levels, the median LL-SS difference in output attenuation was 10.5 percentage points (30.8 percent difference).

The amount of attenuation measured for LL versus SL components is shown for equivalent input in Fig. 5D. For equivalent input attenuation, LL components showed larger amounts of attenuation than SL components, and these differences were significant by Wilcoxon Signed-Rank tests (all P ≤ 0.02). The increasing sizes of the differences with stimulus level is consistent with the more compressive growth functions of LL components than SL components shown in Fig. 4B. Together, these results demonstrate that LL components showed larger amounts of attenuation than SL components for both output and input measures of attenuation.

Phase-Related Changes

As noted in the methods section, phase measurements can be sensitive to the presence of SSOAEs. Our methods sought to minimize unwanted effects by excluding frequencies containing SSOAEs from the averaged responses. MOCR-induced changes in CEOAE phase are shown in the top row of Fig. 6. Phase changes from the subset (40%) of our participants without any SSOAEs (pink boxplots) are plotted next to phase changes from the full cohort (red boxplots). The data from the full cohort appear similar to the subset, suggesting that our control measures were successful and the group phase results are largely unaffected by SSOAEs. Hereafter, we refer only to phase data from the full cohort.

Figure 6.

Top: changes in CEOAE phase a function of stimulus level for short latency (SL, A) and long latency (LL, B) components. Positive values on the y-axis indicate a phase lead in the presence of MOCR activation. Data from participants with no SSOAEs are shown by pink boxplots; data from the full cohort are shown by red boxplots. Bottom: percentage reductions in CEOAE amplitude only (black), phase only (red), and total quantity (both amplitude and phase; gray) as a function of stimulus level for SL (C) and LL (D) components. Boxplots shown with horizontal offsets for visual clarity. Including phase with amplitude increases the calculated size of MOCR effects. LL components showed larger attenuation than on SL components. CEOAE, click-evoked otoacoustic emission; MOCR, medial olivocochlear reflex; ppFPL, peak-to-peak forward pressure level; SSOAEs, synchronous-spontaneous otoacoustic emissions.

MOCR activation resulted in a phase lead for both SL (Fig. 6A) and LL (Fig. 6B) components. All phase data were evaluated statistically using a test of median direction for circular data given by Fisher (55). The phase differences were unimodally distributed, so that a unique median phase angle could be determined in each case. For odd sample sizes, the median phase angle was the data point dividing the sample into two equal halves. For even sample sizes, median phase angle was taken as the midpoint between the two middle data points. A 95% confidence interval around the median phase angle was calculated, and the median phase was considered statistically significant if the confidence interval did not encompass zero. The MOCR-induced phase leads for SL and LL components (Fig. 6, A and B) were all significantly different from zero (P < 0.01). Across stimulus level, SL components showed median phase leads ranging from 0.01 to 0.02 cycles, and LL components showed median phase leads ranging from 0.04 to 0.07 cycles. LL components had significantly larger median phase leads than SL components at the middle three stimulus levels (P < 0.01) but not at the lowest and highest stimulus levels.

In MOCR studies, phase changes are commonly combined with amplitude changes into a single value, sometimes referred to as “total quantity of change.” Total quantity of change as a relative percent attenuation is plotted in Fig. 6, C and D (gray boxplots). The separate contributions of amplitude (black boxplots) and phase (red boxplots) to the total quantity of change metric are also plotted. Note that the amplitude contributions to total quantity of change in Fig. 6, C and D (black boxplots) are calculated as 100 |δ_A − 1|, which is different from the expression in Fig. 5, A and B, where the directionality of amplitude increases and decreases was preserved using 100(1 − δ_A).

Total quantity attenuation was larger than either amplitude or phase attenuation alone. The differences between total quantity attenuation and the amplitude-only contribution to attenuation were all statistically significant by Wilcoxon Signed-Rank tests (all P ≤ 0.004). Visually, the relative contribution of amplitude and phase to the total quantity was similar, but somewhat variable across stimulus level. The inclusion of phase with amplitude resulted in median increases in attenuation of 3.6 to 7.6 percentage points (12 to 27.9 percent difference) for SL components, and median increases of 6.7 to 12.8 percentage points (18.1 to 38 percent difference) for LL components. Total quantity attenuation was larger for LL components than for SL components at all stimulus levels by Wilcoxon Signed-Rank tests (all P < 0.01). On average, LL components showed 13.1 percentage points more attenuation (37.3 percent difference) than SL components. Compared with magnitude output attenuation measures, MOCR effects were larger when assessed with total quality measures that included phase. LL components showed more total quantity attenuation than SL components.

DISCUSSION

MOCR Effects on SL versus LL Components

We examined the effects of MOCR on SL and LL CEOAE components. The LL components contained energy generated predominately near CF place, where the effects of cochlear amplification are expected to be greatest. In contrast, the SL components contained earlier-occurring CEOAE energy that was presumably generated more basally. Across a range of stimulus levels, MOCR effects on SL and LL components were examined in terms of amplitude growth, slope changes, amplitude output attenuation, equivalent input attenuation, phase, and total quantity of change. Taken together, our results show that MOCR has larger effects on LL components than on SL components.

For a given stimulus frequency, MOCR effects are understood generally to be largest 1) near CF place and 2) at low stimulus levels. These findings have been reported for a variety of physiological measures, including basilar membrane motion, inner hair cell receptor potentials, and auditory nerve firing (see Refs. 1 and 2 for reviews), as well as for CEOAEs (56, 57). Such data are consistent with the hypothesis that MOCR activation reduces the gain of the cochlear amplifier by a fixed proportion, and with the hypothesis that cochlear amplifier gain is greatest near CF place. The CEOAE data in the present study are consistent with these observations and hypotheses, and with the hypothesis that LL CEOAE components are generated near CF place: the largest reductions in CEOAE amplitude were seen for LL components at the lowest stimulus levels tested (Figs. 5B and 6D).

The presence of smaller, but significant, MOCR effects on SL CEOAE components (Figs. 4, A and C; 5A; and 6, A and C) is consistent with studies suggesting that coherently-reflected OAE energy (e.g., CEOAEs and SFOAEs) contains level-dependent components coming from regions somewhat basal of CF place (9, 10, 58, 59). Our data are consistent with the idea of SL CEOAE energy being comprised mostly of sources (coherent reflection or distortion) that are basal to CF, but still within the amplification region. Goodman and colleagues (4) provided evidence that such energy does not come from “far basal” regions, at least in the guinea pig, and this is consistent with measurements in humans (60). If SL components were generated in the “far basal” region, there would presumably be no cochlear amplification for the MOCR to alter, and therefore we would expect to see no changes in CEOAE magnitude or phase when the MOCR was activated. In the present study, CEOAEs composed of SL energy showed clear and statistically significant MOCR-related changes, suggesting that they were generated in region(s) where there is active cochlear gain which could be reduced by MOCR activation. As such, our data are consistent with the idea of SL CEOAE energy being comprised mostly of reflection sources that are “near basal,” not “far basal,” relative to CF place.

MOCR Effects on CEOAE Growth Functions

MOCR-induced changes in CEOAEs can be described by change in slope of the growth function, amplitude output attenuation (y-axis change), equivalent input attenuation (x-axis change), and total quantity of change (magnitude and phase combined). These four measures focus on different aspects of change induced by the same underlying process (reduction of cochlear gain by MOCR). The following summary describes the overall effects of MOCR activation on CEOAEs observed in this study. MOCR activation caused the slope of the growth function to become more linear. The amount of slope change was similar across stimulus levels. The slope change was related to the pattern of amplitude changes observed on the x- and y-axes considered together. Amplitude output attenuation (y-axis change) was largest at low stimulus levels and decreased as stimulus level increased. For LL components, equivalent input attenuation (x-axis change) showed the opposite pattern of growth, being smallest at low stimulus levels and increasing with stimulus level, albeit to a smaller extent. For SL components, equivalent input attenuation showed the same growth pattern as amplitude output attenuation (a reduction with stimulus level). Presumably this difference in growth patterns was due to the more linear growth of SL components without noise, relative to the more compressive growth of LL components. As noted in the methods section, when growth functions are close to linear (SL components), shifts on the x and y axes are similar.

Output attenuation in the LL components was still roughly 2 dB (20%) even at the highest stimulus levels. It is curious why this should be, since cochlear amplification is presumably minimal at high stimulus levels. One reason may be that the highest stimulus level of 91 dB ppFPL was only 54 dB SL and 35 dB SPL when expressed as level per 1-Hz FFT bin width. From this perspective, some cochlear amplification could still be expected even at the highest level used and could therefore show an MOCR-mediated change. It is possible that at high stimulus levels the MOCR could also have caused changes in emissions by altering cochlear reflection via some morphological change in outer hair cells rather than by direct reduction of cochlear amplification (61), though this is speculative.

From a signal processing perspective, passing an amplitude modulated signal through a compressive nonlinearity results in a reduction of modulation depth, since higher level amplitudes receive less gain than lower amplitudes. Thus, it is possible that the less compressive growth functions observed with MOCR activation might indirectly indicate a more pronounced cochlear representation of amplitude fluctuations, which in some conditions could be beneficial to listening tasks involving envelope processing in background noise. Any perceptual effect might be larger at moderate to high input levels, since the growth slope at the lowest levels is less compressive to start with. (Although this is difficult to know, since much of the animal work in this area has focused on MOCR effects on rate-place encoding in the auditory nerve as opposed to temporal envelope encoding.)

Changes in AGFs due to MOCR can be thought of as shifts along the y-axis (amplitude reduction) or along the x-axis (effective attenuation). These are both descriptions of the same underlying phenomenon (change in cochlear gain), but from different viewpoints. When amplitude growth slope is unity, the x-axis and y-axis shifts are identical by definition. At higher stimulus levels, the more compressive the amplitude growth, the larger the values of equivalent input attenuation. Characterization of attenuation on either the x- or y-axis may be useful, depending on the specific question being addressed. For example, Wilson et al. (62) found that differences between MOCR-induced CEOAE amplitude and effective attenuation shifts can be diagnostic for children with autism and hyperacusis. It seems likely that amplitude output attenuation measured at low stimulus levels is be related to issues such as hearing in noisy backgrounds. It is less clear, however, how such low-level results are directly related to high-sound-level issues, such as protecting against noise damage.

Combined Effects CEOAE Amplitude and Phase

A few previous studies have presented MOCR-induced amplitude and phase changes separately (12, 32, 63). The phase changes reported in the current study are similar to those reported in previous work: roughly 0.01–0.07 periods of phase lead on average. The size of the effect does not appear to be substantially different in the present study (with bilateral noise activators) compared with contralateral-only noise activators. It is worth noting that the size of MOCR-induced changes in delay may appear much larger when calculated using standard group delay measures than when using phase delay expressed as an equivalent time delay (see Ref. 12).

When phase changes across levels are compared, our results do not demonstrate an increase in the amount of phase change at high levels, such as that reported by Cooper and Guinan (64). They showed large phase changes at high levels when measuring basilar membrane vibration in anesthetized guinea pigs and chinchillas. Cooper and Guinan suggested this might be due to MOCR-mediated changes in the stiffness of the cochlear partition which, while effecting little change in the basilar membrane motion, caused large changes in phase. Although we did not observe similar phase changes, overlaying and aligning amplitude reduction curves (their Fig. 2A and our Fig. 4B) suggests that our stimulus amplitudes may not have been high enough to see any such effect.

Phase changes have more commonly been combined with amplitude changes into a single value, sometimes referred to as “total quantity of change.” As described in the methods section, the way in which phase is included, mathematically, means that total quantity of change must always be greater than or equal to amplitude change alone. In practice, the phase difference between CEOAEs obtained with and without a noise activator is never precisely zero, so total quantity of change will always be larger. In our study, percent attenuation expressed by total quantity of change was larger than amplitude output attenuation by ∼10 percentage points (29 percent difference).

Although it is gratifying to see larger MOCR effects, this combined quantity is potentially problematic. On theoretical grounds, if the action of MOCR is to reduce cochlear amplification, then the traveling wave pattern on the basilar membrane should become less peaked and have shorter delay. Correspondingly, CEOAEs would be expected to have smaller amplitudes, shorter group delays, and corresponding phase leads (see Ref. 12). Although there are some reports of OAE changes in the opposite directions, such results are likely due to two-source interactions of DPOAE components or issues related to the shifting of fine structure notches in reflection OAEs, as opposed to increases in amplifier gain. Total quantity of change does not discriminate between magnitude increases and decreases, nor between phase leads and lags. It is currently unclear what it means to treat such changes as the same.

If the cochlea acts as essentially as a minimum phase filter (65, 66), and if the action of the MOCR is to simply reduce cochlear gain, then group delay should correspondingly decrease and phase should increase in a predicable way with amplitude changes. In this case, including phase as a measurement of MOCR strength should provide corroborating evidence that increases certainty of the estimate, as opposed to adding to the measurement in a way that increases the apparent size of the effect. On the other hand, if the cochlea does not act as a minimum phase filter (67), then phase changes are of interest, but it is unclear how they should be mathematically combined with amplitude in a meaningful way. For example, it is not clear how a relatively large phase change (without a correspondingly large amplitude change) relates to the “unmasking” of transient inputs to the auditory nerve and improved signal detection in background noise. It is also not clear how such phase changes alone would relate to protective effects against loud sounds.

A final caution in regards to total quantity of change is that phase measurements are sensitive to noise and to the presence of SSOAEs. If care is not taken, spurious changes in phase can alter the apparent size of MOCR-related attenuation of CEOAEs. This problem may be exacerbated in participants with strong MOCR, since activation of the reflex turns down the CEOAE level, reducing SNR. In the present data set (not shown), we observed that in a few participants (typically with lower SNRs), the inclusion of phase changes led to a substantial increase in the size of their calculated attenuation, relative to the rest of the participants. Although this did not substantially alter the group results, it has implications when applied to individuals.

Comparison with Previous Studies

A few seminal studies previously investigated the effects of click stimulus level on MOCR-mediated AGF changes in humans (40, 56, 57, 68, 69). These studies reported some similar general findings to the current study, namely, that the largest CEOAE output amplitude attenuation is found at low stimulus levels, and that the slope of amplitude growth functions tends to increase with MOCR activation. The effects reported in the current study are larger than those reported in these previous studies. We observed median amplitude-only (phase not included) output attenuation values of ∼44% (−5 dB), in contrast to early studies which reported average attenuation values of were roughly 16% (−1.5 dB). Veuillet et al. (57) reported small but statistically significant slope changes of 0.03, compared with median slope change of 0.11 in the current study.

These differences in effect size may be due to several methodological differences, including the use of bilateral versus contralateral noise, the use of long interstimulus intervals, and the use of time-frequency windowing. The use of bilateral acoustic activators may result in greater MOCR activation (70–73). Few studies have examined this effect with CEOAEs, and the most directly comparable previous study would appear to be that of Boothalingam et al. (72), who reported an average, bilaterally activated MOC shift in CEOAEs of approximately −3.2 dB (31% attenuation). With regard to click spacing, early pilot data collected in our laboratory suggested that 200-ms click spacing yielded different results than 50-ms spacing. However, finding the closest possible spacing was not a focus of the study, and we simply chose to error on the side of caution. The extended click spacing used in the present study was unlikely to have affected the results for low stimulus levels (where the largest amplitude reductions were seen). Rather, extended click spacing enabled higher click levels to be used without activating either the MOCR or the MEMR. Given the more linear growth of SL CEOAE components with stimulus level, the use of time-frequency windowing also had the largest effects at higher stimulus levels.

From a practical measurement standpoint, it is of interest whether the techniques described in this study provided increase in the size of measured MOCR activity, relative to previous work generally. Given the large number of variables that differ across studies, this is a difficult question to answer. Differences in stimuli, calibration methods, sample sizes, test paradigms, participant demographics, SNR criteria, number of averages, and method of calculation all contribute to the variability seen. A comprehensive review of the literature is beyond the scope of this paper. However, we estimate that in general, previous studies using CEOAEs to measure MOCR have reported average amplitude attenuation of between −1.5 and −2.5 dB (16% to 25% attenuation). When comparing the size of MOCR effects across studies, it is important to consider which type of attenuation is reported (output versus input), which quantity of change is reported (amplitude only versus total amplitude plus phase), and whether the MOCR activator was contralateral or bilateral. More recent studies (47, 74) tend to report total quantity of change (magnitude and phase combined) output attenuation values, which are larger than many early studies, but are still slightly smaller than those reported here. Studies using bilateral MOCR activators (72, 73) also show slightly smaller changes than those reported here. Our findings of median amplitude output attenuation of 44% (−5 dB) and total quantity of change of 56% (−7 dB) are larger than most previous reports.

Conclusions

Here, we characterized the effects of MOCR on CEOAEs across a 45-dB range of click-stimulus levels. We used a slow click rate to avoid middle-ear muscle effects and time-frequency windowing to measure MOCR effects on SL and LL CEOAE components. We described the effect of MOCR on CEOAEs in terms of amplitude output attenuation, equivalent input attenuation, phase change, total quantity of attenuation, and change in the slope of amplitude growth. SL CEOAE components showed smaller amounts of MOCR-induced attenuation than LL components, consistent with SL origins being “near basal” to the CF place. Amplitude output attenuation and total quantity of attenuation showed the largest effects at the lowest stimulus levels. In contrast, slope and phase changes were relatively constant across stimulus level. Equivalent input attenuation decreased with level for SL components but increased with level for LL components. Measurements that remained robust at higher stimulus levels, such as those shown here, may be useful for understanding individual variability, and for predicting listening abilities in background noise and susceptibility to noise damage.

GRANTS

This work was supported by the Department of Communication Sciences and Disorders at the University of Iowa (S. S. Goodman), New Century Scholars Grant from the American Speech-Language Hearing Foundation (S. Boothalingam), and R01 DC014997 (J. T. Lichtenhan).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

S.S.G. conceived and designed research; S.S.G. performed experiments; S.S.G. and S.B. analyzed data; S.S.G., S.B., and J.T.L. interpreted results of experiments; S.S.G. prepared figures; S.S.G. drafted manuscript; S.S.G., S.B., and J.T.L. edited and revised manuscript; S.S.G., S.B., and J.T.L. approved final version of manuscript.