Associations between the medial olivocochlear reflex, middle-ear muscle reflex, and sentence-in-noise recognition using steady and pulsed noise elicitors

The middle-ear muscle reflex (MEMR) and medial olivocochlear reflex (MOCR) modify peripheral auditory function, which may reduce masking and improve speech-in-noise (SIN) recognition. Previous work and our pilot data suggest that the two reflexes respond differently to static versus dynamic noise elicitors. However, little is known about how the two reflexes work in tandem to contribute to SIN recognition. We hypothesized that SIN recognition would be significantly correlated with the strength of the MEMR and with the strength of the MOCR. Additionally, we hypothesized that SIN recognition would be best when both reflexes were activated. A total of 43 healthy, normal-hearing adults met the inclusion/exclusion criteria (35 females, age range: 19 – 29 years). MEMR strength was assessed using wideband absorbance. MOCR strength was assessed using transient-evoked otoacoustic emissions. SIN recognition was assessed using a modified version of the QuickSIN. All measurements were made with and without two types of contralateral noise elicitors (steady and pulsed) at two levels (50 and 65 dB SPL). Steady noise was used to primarily elicit the MOCR and pulsed noise was used to elicit both reflexes. Two baseline conditions without a contralateral elicitor were also obtained. Results revealed differences in how the MEMR and MOCR responded to elicitor type and level. Contrary to hypotheses, SIN recognition was not significantly improved in the presence of any contralateral elicitors relative to the baseline conditions. Additionally, there were no significant correlations between MEMR strength and SIN recognition, or between MOCR strength and SIN recognition. MEMR and MOCR strength were significantly correlated for pulsed noise elicitors but not steady noise elicitors. Results suggest no association between SIN recognition and the MEMR or MOCR, at least as measured and analyzed in this study. SIN recognition may have been influenced by factors not accounted for in this study, such as contextual cues, warranting further study.


Introduction
Difficulty hearing speech in noise is the most frequently reported issue for individuals with hearing loss (Woods et al., 2015).Hundreds of millions of individuals worldwide are affected by hearing loss (Davis and Hoffman, 2019), and untreated hearing loss is associated with negative outcomes such as decreased quality of life, poorer productivity in the workplace, and dementia (Ciorba et al., 2012;Dalton et al., 2003;Lin et al., 2011;Nachtegaal et al., 2012).Understanding the physiologic underpinnings of hearing in noise is an important step toward improving the diagnosis and treatment of these difficulties.
The auditory system has two reflexes that may work in tandem to improve hearing in noise (Liberman and Guinan, 1998;Pang and Guinan, 1997).One reflex is the middle-ear muscle reflex (MEMR), in which brainstem nuclei innervate the facial nerve to contract the stapedius muscle and alter the impedance characteristics of the middle ear in response to sound (reviewed in Mukerji et al., 2010).The MEMR can be measured indirectly as the change in middle ear admittance (typically using a 0.226-kHz probe tone) in response to an ipsilateral or contralateral reflex elicitor (Schairer et al., 2013).Another reflex is the medial olivocochlear reflex (MOCR), in which brainstem neurons innervate outer hair cells to reduce cochlear amplifier gain (reviewed in Lopez-Poveda, 2018).The MOCR can be measured indirectly as the change in outer hair cell responses [e.g., transient-evoked otoacoustic emissions (TEOAEs; Kemp, 1978) and the cochlear microphonic (Jennings and Aviles, 2023)] in response to an ipsilateral or contralateral reflex elicitor (Collet et al., 1990).
The MEMR may contribute to speech-in-noise (SIN) recognition.When activated by sound elicitors, the MEMR causes contraction of the stapedius muscle that reduces transmission of lower-frequency sounds (below 1-1.5 kHz) through the middle ear system (e.g., Borg, 1968;Wiggers, 1937).As with the MOCR, the MEMR can also be indirectly measured using otoacoustic emissions (OAEs) and the cochlear microphonic (Jennings and Aviles, 2023;Shehorn et al., 2020).Because low-frequency noise can mask higher-frequency sounds via the upward spread of excitation (Miller, 1947;Wegel and Lane, 1924), a reduction in low-frequency transmission of sound should provide an improvement in SIN recognition (Stevens and Davis, 1938).This has been supported by a study of experimental contraction of the stapedius muscle, where masking was reduced as assessed using measurements of auditory nerve fiber responses (Pang and Guinan, 1997).
In humans, one method to infer the contribution of the MEMR is to compare SIN recognition in patients surgically treated for middle ear disorders such as otosclerosis.Placement of an ossicular prosthesis has the potential to restore MEMR function, whereas surgical severing of the stapedius tendon can eliminate the MEMR.For example, patients who underwent stapedectomy that spared the stapedius muscle demonstrated an improvement in monosyllabic word recognition in noise (Lidén et al., 1964).Additionally, individuals who underwent stapedotomy that spared the stapedius muscle demonstrated significantly better word recognition in noise scores compared to patients whose stapedius tendon was severed (Colletti et al., 1988).More recently, Aiken et al. (2013) found that patients with a severed stapedius muscle demonstrated different patterns in SIN recognition across signal-to-noise ratios (SNRs) compared to a control group.In the patient group, performance decreased with decreasing SNR with a shallow slope.In contrast, the control group showed minimal changes in performance as the SNR decreased from +5 to − 5 dB SNR, but then performance decreased steeply at poorer SNRs.The results support the potential benefit of the MEMR for SIN recognition in healthy individuals, but the benefit may depend on the SNR.
The contributions of the MEMR to SIN recognition have also been examined through correlational analyses.Harkrider and Smith (2005) investigated the association between MEMR thresholds and word recognition in noise.The MEMR was measured using a 0.226-kHz probe tone.The MEMR elicitor was broadband noise (spectrum not specified) that was presented ipsilaterally and contralaterally from 50 up to 110 dB HL.Word recognition in noise was measured using NU-6 words presented at 55 dB HL in the presence of multitalker babble with an SNR of +0 dB.The authors found a significant negative correlation between ipsilateral MEMR thresholds and word recognition in noise (i.e., lower MEMR thresholds were associated with better word recognition in noise performance).Shehorn et al. (2020) examined correlations between MEMR strength and SIN recognition in adults with and without self-reported hearing in noise difficulties.MEMR strength was assessed using wideband acoustic immittance (assessment of middle ear function across a broad range of frequencies).Speech stimuli consisted of monosyllabic words presented at three intensity levels in the presence of speech-shaped noise.The authors found that SIN recognition was significantly positively correlated with MEMR strength, but only for the highest intensity level.Similarly, Mepani et al. (2020) also found that MEMR strength was positively correlated with SIN recognition using monosyllabic words.It should be noted that the studies of Shehorn et al. (2020) and Mepani et al. (2020) utilized the MEMR as a potential index of cochlear synaptopathy (for a review, see Bharadwaj et al., 2019), whereas the current study did not examine cochlear synaptopathy.
In addition to the MEMR, the MOCR may also contribute to SIN recognition.When activated by sound elicitors, the MOCR hyperpolarizes the outer hair cells and increases the conductance of the outer hair cells (reviewed in Lopez-Poveda, 2018).These effects can improve auditory nerve encoding of transient sounds in the presence of background noise (Kawase et al., 1993;Winslow and Sachs, 1987).Lesions of the olivocochlear bundle in animals can result in poorer intensity discrimination in noise (May and McQuone, 1995) and poorer phoneme discrimination in noise (Dewson, 1968).These findings have resulted in the hypothesis that the MOCR benefits SIN recognition in humans.
Support for this hypothesis has come from correlational studies of MOCR strength and SIN recognition (e.g., Giraud et al., 1997;Kumar and Vanaja, 2004;Mishra and Lutman, 2014).MOCR strength is typically measured using a contralateral inhibition paradigm (Collet et al., 1990), in which the change in TEOAE amplitude is assessed in the presence of a contralateral MOCR elicitor.Some work has suggested that the relationship between the MOCR and SIN recognition depends on factors such as the SNR used for the SIN task or the specific quantification of performance on the SIN task (e.g., percent correct, threshold of the psychometric function, and information transmitted for speech features) (Abdala et al., 2014;Mertes et al., 2018Mertes et al., , 2019)).However, the relationship between MOCR strength and SIN recognition remains unclear because other studies have reported no significant correlations (Gafoor and Uppunda, 2023b;Mertes and Stutz, 2023;Wagner et al., 2008;Yashaswini and Maruthy, 2019) or negative correlations between the two measures (de Boer et al., 2012;DeRoy Milvae et al., 2021;Karunarathne et al., 2018).The discrepancy in findings indicates that more work is needed to understand the conditions in which the MOCR does or does not contribute to SIN recognition.
In experimental measurements of the MOCR, care is typically taken to avoid eliciting the MEMR because the MEMR can also reduce TEOAE amplitudes, making it difficult to determine the extent to which the MOCR contributed to the change in TEOAE amplitude (Guinan et al., 2003).MEMR activation can be detected by assessing changes in TEOAE stimulus amplitude in response to the elicitor, although some studies attempt to avoid MEMR activation by presenting the elicitor at a level below the threshold of MEMR activation (reviewed in Mertes, 2020).Although experimental isolation of MEMR and MOCR effects are desirable for studying basic properties of each reflex, this approach does not allow for determining the combined contribution of the reflexes in realistic listening situations.The MOCR has a lower threshold for activation than the MEMR (Marks and Siegel, 2017).However, MEMR measurements made using wideband acoustic immittance (e.g., Schairer et al., 2022) have demonstrated MEMR thresholds that may be activated at sound levels that may be encountered in typical listening situations (Wu and Bentler, 2012).Additionally, the MOCR appears to respond similarly to different temporal properties of contralateral noise (Boothalingam et al., 2014;Kalaiah et al., 2017;Mertes, 2018), whereas the MEMR appears to respond differentially to modulated and unmodulated noises (unpublished data obtained in our laboratory).
A previous study examined the combined contribution of the MEMR and MOCR for signal-in-noise detection in humans (Karunarathne et al., 2018).Participants were normal-hearing young adults who underwent contralateral inhibition of TEOAEs and a behavioral task involving detection of a 2-kHz alarm sound in the presence of pink noise.For the contralateral inhibition measurements, a high-level broadband noise elicitor was used that presumably activated both the MEMR and MOCR.Alarm detection thresholds were significantly negatively correlated with the magnitude of contralateral inhibition.The authors concluded that the results suggest that stronger MEMR and MOCR reflexes result in degraded, rather than improved, signal-in-noise detection, at least for negative SNRs (i.e., the signal level is lower than the noise level; reviewed in Jennings, 2021).It should be noted that this study did not directly measure MEMR using immittance, but rather inferred that the contralateral noise level was sufficiently high to evoke the MEMR.Additionally, the noise level during the behavioral task was relatively high (80 dBA) and the detection thresholds were obtained at negative SNRs.
Although a significant body of literature has examined associations between the MOCR and SIN recognition (and to a lesser extent, associations between the MEMR and SIN recognition), there is a gap in knowledge about how elicitation of both reflexes contributes to SIN recognition.Understanding the combined contribution of the two reflexes is important because data on thresholds for the MEMR and MOCR (Feeney et al., 2017;Marks and Siegel, 2017) suggest that both reflexes would be activated in response to commonly-encountered noise levels (Flamme et al., 2012).The goal of the current study was to determine the contribution of the MEMR and MOCR to SIN recognition obtained at average conversational levels as an initial step toward understanding the combined contribution of the MEMR and MOCR to SIN recognition.The MEMR was assessed using wideband absorbance, and the MOCR was assessed using contralateral inhibition of TEOAEs.As a novel approach, we utilized contralateral elicitors that were intended to primarily activate the MOCR or that would activate both the MOCR and MEMR.It is well-established that steady-state broadband noise (hereafter referred to as "steady noise") is an efficient activator of the MOCR (Lilaonitkul and Guinan, 2009;Maison et al., 2000;Velenovsky and Glattke, 2002).The MOCR also responds to amplitude-modulated noises, with studies showing equivalent MOCR activation compared to steady noise (Boothalingam et al., 2014;Mertes, 2018;Mishra and Biswal, 2019, but see also Maison et al., 1997Maison et al., , 1999)).In contrast, the MEMR appears to be preferentially activated by pulsed noises compared to steady noise (Lilly, 1964).Unpublished data collected in our laboratory supports this notion.
For our measure of SIN recognition, we utilized the QuickSIN (Killion et al., 2004) because of its increased ecological validity over stimuli such as monosyllabic words and because a previous study found that performance on the QuickSIN was correlated with MOCR strength (Bidelman and Bhagat, 2015).Due to the correlational nature of the current study, we adapted a modified version of the QuickSIN (Grant et al., 2020) to ensure sufficient across-subject variability on the SIN recognition task in a group of young adults with normal hearing.
For the MEMR and MOCR measures, we hypothesized that there would be a significant main effect of level, where a 65 dB SPL elicitor would yield significantly larger reflex magnitude than a 50 dB SPL elicitor, based on previous work (Hood et al., 1996;Keefe et al., 2017;Veuillet et al., 1991;Wojtczak et al., 2017).For the MEMR, we hypothesized that there would be a significant main effect of elicitor type, where pulsed noise would elicit significantly larger MEMR magnitude than steady noise.This hypothesis is based on previous work (Lilly, 1964;Sesterhenn and Breuninger, 1976) as well as unpublished data obtained in our lab.Conversely, for the MOCR, we hypothesized that there would be no significant main effect of elicitor type based on our previous work (Mertes, 2018).For SIN recognition, we hypothesized that there would be a significant interaction between elicitor type and elicitor level, where the pulsed elicitor would yield a significantly better score than the steady elicitor, and that this difference in score would be significantly larger for an elicitor level of 65 dB SPL compared to 50 dB SPL.The rationale is that the pulsed elicitor presented at 65 dB SPL would elicit both the MEMR and MOCR more strongly compared to the other elicitor conditions, and thus there would be more antimasking that would benefit SIN recognition for the pulsed elicitor presented at 65 dB SPL.Finally, we hypothesized that SIN recognition would be positively correlated with the strength of the MEMR (Shehorn et al., 2020) and with the strength of the MOCR (Bidelman and Bhagat, 2015;Mertes et al., 2019).To the best of our knowledge, this is the first study that examined the contribution of both the MEMR and MOCR to sentence-in-noise recognition.

Participants
The research protocol was approved by the Institutional Review Board of the University of Illinois Urbana-Champaign (IRB #22239).Individuals were recruited from the university campus through flyers, in-class announcements, electronic newsletters, university website postings, and word of mouth.Written informed consent was obtained from all participants prior to enrollment.Participants received their choice of monetary compensation or extra credit for an approved university course.
Our target sample was healthy, young adults with normal hearing.Inclusion criteria consisted of the following: ages 18-30 years; English as a first language; non-smoker; and passing a laboratory hearing screening.Because the MOC reflex was measured in the right ear (described in Section 2.6), participants were also required to be righthanded to maximize the MOC reflex strength (Khalfa et al., 1998).A pass on the hearing screening was defined as the following: an unremarkable otoscopic examination; type A 0.226-kHz tympanograms bilaterally [tympanometric peak pressure between − 100 to +50 daPa, static compliance between 0.2 to 1.8 mL, and an equivalent ear canal volume between 0.6 to 2.5 cc (Mertes, 2018)]; ipsilateral acoustic reflex thresholds ≤90 dB HL bilaterally in response to a pulsed noise elicitor stimulus (Silman et al., 1984); detection of pure-tone air-conduction stimuli at 20 dB HL for octave frequencies from 0.25 to 8 kHz and interoctave frequencies of 3 and 6 kHz bilaterally; word recognition scores ≥90% bilaterally for NU-6 word lists presented at 45 dB HL; and measurable TEOAEs in the right ear (SNR >6 dB from 1 to 4 kHz) in response to 2000 sweeps of 65 dB pSPL clicks (Mertes, 2018).Exclusion criteria consisted of self-report of any of the following: hearing loss; tinnitus; hyperacusis; otalgia; otorrhea; aural fullness; history of ear surgery; ototoxic medication; exposure to sound within the past six months that caused hearing loss, tinnitus, and/or aural fullness; family history of permanent childhood hearing loss; vertigo; speech/language disorder; neurologic disorder; diabetes; hypertension; and excessive discomfort with the experimental stimuli, earphones, and/or testing environment.Finally, participants were excluded if they were unable to comply with the procedures.
A total of 71 individuals were enrolled (56 females).Twenty-eight individuals were excluded due to not meeting inclusion/exclusion criteria, inability to complete all procedures, or missing data.Therefore, data are reported for 43 individuals (35 females).It should be noted that we attempted to recruit an equal number of males and females.The median age was 21 years (range = 19-29 years).

Equipment
The experiment was conducted in a single-walled sound-treated booth meeting standards for maximum permissible ambient noise levels for ears covered and uncovered (ANSI, 1999).Participants were seated in a recliner during testing.Pure-tone air-conduction screenings, word recognition testing, and the QuickSIN were conducted using an Audio-Star Pro audiometer (Grason-Stadler, Inc., Eden Prairie, MN, USA) and an IP30 insert earphone (RadioEar, Middelfart, Denmark).Tympanograms and ipsilateral acoustic reflexes were obtained using a Titan tympanometer (Interacoustics, Middelfart, Denmark).
Wideband absorbance was measured using the Titan tympanometer and Titan Suite software ver.3.6.1 with a wideband tympanometry research license (Interacoustics).TEOAEs were measured using an ER-10C probe microphone system (Etymōtic Research, Elk Grove Village, IL, USA) with +20 dB preamplifier gain that interfaced with a Babyface Pro 24-bit USB audio interface (RME, Haimhausen, Germany).Software for measuring TEOAEs consisted of the Auditory Research Lab auditory software (ARLas) program (Goodman, 2020) and MATLAB ver.2021b (The MathWorks, Inc., Natick, MA, USA).The sampling rate was set to 44.1 kHz for stimulus generation, playback, and recording.Contralateral elicitors were presented using an ER-2 insert earphone (Etymōtic Research) that interfaced with the Babyface Pro.
Contralateral elicitors were calibrated in an AEC202 2-cc coupler (Larson Davis, Depew, NY, USA) using an 824S type I sound level meter (Larson Davis).An AEC204 occluded-ear simulator (Larson Davis) was used to verify absence of measurable TEOAE responses (i.e., SNR <6 dB from 1 to 4 kHz).Routine calibration consisted of the following verifications: ER-10C microphone sensitivity within tolerance; stimulus levels for TEOAEs and contralateral elicitors within tolerance; absence of measurable TEOAEs in a coupler; absence of measurable acoustic reflexes and low absorbance in a coupler; and equivalent ear canal volume within tolerance.The audiometer and tympanometer were also professionally calibrated annually.

Contralateral elicitors
Two contralateral acoustic stimulation (CAS) types were utilized in this study: steady and pulsed noise.The steady noise was generated in MATLAB as Gaussian noise that was lowpass filtered at 16 kHz using a Hann window with a filter order of 256.The noise waveforms were ramped on and off using 10-ms Hann windows.The pulsed noise was derived from the steady noise by zeroing-out every other 50-ms segment of the waveform (Mertes, 2022), thus having a rate of 0.01 kHz and a 50% duty cycle.A new instance of the noise was generated for each presentation to avoid potential adaptation to the noise (Felty et al., 2009).The CAS was presented at two levels: 50 and 65 dB SPL.Hereafter, these will be referred to as the noise conditions Steady 50 dB, Steady 65 dB, Pulsed 50 dB, and Pulsed 65 dB.
A lower level of 50 dB SPL was chosen to elicit primarily MOCR activity, whereas a higher level of 65 dB SPL was chosen to elicit both MOCR and MEMR activity (Guinan et al., 2003;Schairer et al., 2022).We acknowledge that higher levels would elicit stronger MOCR and MEMR activity, but we considered the listening comfort of our participants given that the contralateral elicitors were presented across the duration of the 2.5-h experiment.Additionally, unpublished data obtained in our lab demonstrated that pulsed noise presented at 65 dB SPL elicited the MEMR in most young adults with normal hearing.

General protocol
After obtaining informed consent, participants underwent the hearing screening.Eligible participants then underwent three experimental procedures: MEMR, MOCR, and SIN recognition.The order of experimental procedures was counterbalanced across participants.A short break was provided after the screening and between each experimental procedure (approximately 5 min and 2 min, respectively).Participants were also allowed to take additional breaks as needed.These additional breaks were not documented.The experimenter instructed participants verbally using a script to ensure consistent instructions across participants.
During the measurements, participants were seated in a recliner and were instructed to remain still and refrain from coughing and swallowing as much as possible during the MEMR and MOCR measurements.For the MEMR and MOCR measurements, the probe (Titan or ER-10C) was placed in the right ear and the ER-2 insert earphone to deliver CAS was placed in the left ear.For the SIN recognition measurements, the IP30 earphone to deliver speech and ipsilateral noise was placed in the right ear and the ER-2 insert earphone to deliver CAS was placed in the left ear.The probe cable of the ER-10C was suspended from the ceiling to reduce vibratory noise.During the MOCR and MEMR measurements, participants viewed a silent, closed-captioned video of their choice on an AirPad 2 tablet computer (Apple, Cupertino, CA, USA) to keep them awake and alert.The experimenters monitored the participants from outside the booth using a camcorder and talkback microphone connected to the audiometer.
A schematic of the experimental protocol is shown in Fig. 1.For each procedure, there were 12 runs.The number of runs was selected based on the total number of lists in the QuickSIN corpus.Runs 1-2 were made without CAS to establish an initial baseline (Baseline 1) and runs 3-4 were made without CAS to establish a second baseline (Baseline 2).These baselines were established to determine when CAS caused a significant change in wideband absorbance, TEOAE amplitude, or Quick-SIN score in the noise conditions.Runs 5-12 consisted of the noise conditions, where the order of noise condition was counterbalanced across participants using a balanced Latin square design.Prior to each run, a spoken recording was played through the insert earphone to inform the participant of the status of the experiment (e.g., "This is measurement 1 of 12.").
Study visits lasted a maximum of 2.5 h.Four participants who did not complete all procedures in one visit were invited to return within 60 days to complete the remaining procedures.At the return visit, these participants were first screened using otoscopy and tympanometry to ensure clear ear canals and normal middle ear function.

MEMR measurement and analysis
Details regarding measurement of wideband absorbance using the Titan system are described in Liu et al. (2008).In brief, wideband absorbance was measured by presenting 32 sweeps of broadband clicks at a level of 100 dB peSPL (~65 dB nHL) and a rate of 21.5/s.The stimulus energy in the ear canal was measured with a probe microphone and analyzed in the frequency domain.Wideband absorbance values ranged from 0 to 1, where 1 indicated that all stimulus energy at that frequency was absorbed by the middle ear.Wideband absorbance was measured from 0.226 to 8 kHz.
The MEMR measurements were based on those described recently by our group (Mertes, 2022).Wideband absorbance was measured at ambient pressure to be consistent with the MOCR measurements that were also made at ambient pressure.A single measurement of the MEMR was obtained by presenting CAS, measuring wideband absorbance, then turning off the CAS.Due to software limitations, the experimenter manually turned the CAS off and on using MATLAB. 1 The results of each measurement were automatically exported to MATLAB for offline analysis.After a run was completed, a 5-s pause was implemented in MATLAB to allow for offset of the MEMR prior to the next run.
Absorbance data were extracted following the manufacturer's guidelines (Interacoustics, 2022).Each exported file contained a 175 × 32 matrix of reflectance values, where 175 represents the reflectance across frequency and 32 represents the total number of sweeps.Absorbance was computed as A = 1 − R × R * , where R is the frequency-dependent complex reflectance and R * is the complex conjugate.Absorbance values less than 0 were set to 0 (Sanford et al., 2009) The MEMR was analyzed using methods adapted from previous work (Keefe et al., 2017;Mepani et al., 2020;Mertes, 2022;Wojtczak et al., 2017).Absorbance signal and noise floor curves were obtained by placing odd-and even-numbered sweeps into two separate matrices and computing a mean waveform for each matrix.The signal was computed by adding the two mean curves and dividing by 2. Conversely, the noise floor curve was computed by subtracting the two mean curves and dividing by 2. Visual inspection revealed sufficiently low noise floor curves.Hereafter, "absorbance curve" refers to the signal curve.mate the duration of an MEMR run and the time delay between manually turning on the CAS in MATLAB and starting the absorbance measurement in the Titan Suite software.Timestamps were obtained in MATLAB using the "tic", "toc", and "waitforbuttonpress" functions.Two sets of 12 runs were obtained while the Titan probe was placed in a 2-cc coupler.Across the 24 runs, the median run duration was 6.9 s (range = 6.7-7.2 s) and the median delay was 1. 4 s (range = 1.3-1.6 s).Absorbance curves were then averaged across pairs (e.g., runs 1 and 2, runs 3 and 4, etc.), resulting in six absorbance curves for each participant.Fig. 2A shows the absorbance curves for each condition for one participant as an example.It can be seen that the two curves in the baseline conditions (gray dashed curve and black solid curve) overlap closely.The red and blue curves represent absorbance curves obtained in the Steady noise and Pulsed noise conditions, respectively.Data quality was assessed using a combination of visual inspection and objective criteria.A participant's absorbance data were excluded if there was lack of agreement between the two baseline curves as determined using visual inspection.Data were also excluded if there was a probable acoustic leak during the measurement, defined as a mean absorbance value >0.3 from 0.226 to 0.385 kHz, or if the mean absorbance across any five adjacent frequencies was 0 (Sun et al., 2023).This resulted in five participants being excluded.
To quantify MEMR strength, the change in absorbance (ΔAbsorbance) was computed by taking each absorbance curve (starting with the second baseline curve) and subtracting it from the first baseline curve.Negative values indicated a decrease in absorbance relative to Baseline 1. Fig. 2B shows the difference curves for one participant.It can be seen that the contralateral elicitors tended to decrease absorbance below 1 kHz and increase absorbance above 1 kHz.To account for these decreases and increases in absorbance across frequency, the absolute value of the change in absorbance (|ΔAbsorbance|) was computed.These data were then reduced to a single value ("MEMR strength") by summing the |ΔAbsorbance| values from 0.5 to 4 kHz.This method is akin to the MEMR strength metric reported in previous work (Jennings and Dominguez, 2022;Mepani et al., 2020;Wojtczak et al., 2017).

MOCR measurement and analysis
MOCR stimulus and measurement parameters were adapted from those described in Mertes (2018).Stimuli consisted of clicks with a duration of 90.7 μs (four samples).Stimuli were presented at a target level of 65 dB pSPL and at a rate of 21.5/s.These parameters were chosen to elicit TEOAEs with sufficient amplitudes while avoiding click-induced elicitation of the ipsilateral MEMR and ipsilateral MOCR which would confound the interpretation of the results (Boothalingam and Purcell, 2015;Guinan, 2006).Before each TEOAE measurement, the stimuli were calibrated in-situ by presenting a series of 100 clicks in the ear canal, highpass filtering the recording using an infinite impulse response filter (stopband = 0.075 kHz, passband = 0.13 kHz, order = 20), rejecting artifacts [sweeps with a root-mean-square (RMS) amplitude exceeding 1.5 times the interquartile range (IQR) across all sweeps within a participant], and measuring the peak SPL of the mean stimulus waveform.The peak SPL of the stimulus was adjusted by digitally scaling the amplitude of the stimulus waveform, presenting another 100 clicks, and measuring the corresponding peak SPL again.The in-ear calibration was considered sufficient when the measured peak SPL was within ±0.25 dB of the target level.
A single run consisted of 10 interleaves of clicks with and without CAS. 2 Each interleave consisted of 8000 ms of clicks in the right ear (without CAS), 500 ms of CAS in the left ear, 8000 ms of clicks in the right ear and 8000 ms of CAS in the left ear (with CAS), and 500 ms of silence.The 500-ms segments of CAS only and silence were used to ensure full onset and offset of the MOCR (Backus and Guinan, 2006), but these segments were discarded from the analysis.Each run consisted of 1720 individual clicks with and without CAS (8000 ms × 21.5 clicks/s × 10 interleaves). 3After each run, the recordings were reshaped into two matrices corresponding to the waveforms obtained with and without CAS.The individual waveforms were aligned in time so that time 0 corresponded to the peak of the stimulus.
Because MEMR activation can increase middle ear impedance and confound MOCR measurements (Guinan et al., 2003), we implemented a check to detect the presence of MEMR activation in the MOCR recordings.We computed critical differences in ear-canal stimulus amplitude obtained from test-retest measurements in the absence of CAS (Mertes, 2020).If the change in ear-canal stimulus amplitude in the presence of CAS exceeds the 95% critical difference, this is interpreted as probable MEMR activation.In order to detect MEMR activation, three sets of stimuli were isolated from the recording for each participant: stimuli within the first 8000-ms interleave without CAS ("no-elicitor 1"), stimuli within the first 8000-ms interleave with CAS ("elicitor 1"), and stimuli within the second 8000-ms interleave without CAS ("no-elicitor 2").The click stimuli were time windowed from 0.2 ms prior to the stimulus peak and ended 1.12 ms after the stimulus peak.The stimulus waveforms were then detrended using the 'detrend.m'function in MATLAB.Artifact rejection was then performed by discarding stimulus waveforms whose RMS amplitude exceeded 1.5 times the IQR for that participant.The mean stimulus waveform was computed for all conditions, and the RMS amplitude of each mean waveform was computed.Participants who had >10% of the stimulus waveforms rejected were excluded from the calculation of the critical differences.Stimulus amplitudes were expressed in dB SPL.The standard error of measurement The fifth row shows the stimuli for the SIN recognition measurements, where the list number refers to the QuickSIN.It should be noted that the order of the noise conditions was counterbalanced and the order of the QuickSIN list numbers was randomized across participants, so this figure demonstrates one possible order.
2 It must be noted that a single run for the MOCR measurement was considerably longer than a single run for the MEMR measurement. 1,3Because of the low-level amplitude of TEOAEs, longer averaging times were required for the MOCR measurement to ensure a sufficient SNR compared to the MEMR measurement that was obtained from high-level clicks with a higher SNR.Because of the longer data collection times for the MOCR measurements, interleaving of CAS was used to reduce potential effects of drift in TEOAE amplitude during the course of a measurement (Mertes & Goodman, 2016).
3 After the experiment was completed, a simulation was conducted to estimate the duration of an MOCR run and the time delay between the end of one run and the in-situ calibration of the next run.Timestamps were obtained in MATLAB using the "tic" and "toc" functions.One set of 12 runs was obtained while the ER-10C probe was placed in a 2-cc coupler.Across the 12 runs, the median run duration was 209.5 s (range = 208.8-210.0s) and the median delay was 13.8 s (range = 12.0-18.8s).
was calculated as , where SD is the standard deviation across the no-elicitor 1 and 2 conditions, and r xx is the correlation coefficient between the no-elicitor 1 and 2 conditions.The 95% critical difference was computed as . Separate critical differences were computed for each elicitor condition and are shown in Table 1.If the difference in stimulus amplitude between the no-elicitor 1 and elicitor 1 conditions exceeded the critical difference for a given elicitor type and elicitor level condition, this was interpreted as probable MEMR activation and the case was flagged.
TEOAE analysis examined the entire set of recordings in matrices of waveforms obtained with and without CAS.After windowing each waveform to begin at the stimulus peak (time 0), the first 4 ms was then zeroed-out to reduce stimulus artifact.Waveforms were ramped on and off from 4 to 6.5 ms and from 17.5 to 20.0 ms using a Hann window.Waveforms were then bandpass filtered with a finite impulse response filter (Hann filter, cutoff = 0.5-4 kHz, order = 128), and the zeroing-out and ramping was repeated.Artifact rejection was performed by discarding TEOAE waveforms whose RMS amplitude exceeded 1.5 times the IQR for that participant.TEOAE signal and noise floor amplitudes were obtained by placing odd-and even-numbered waveforms into two separate matrices and computing a mean waveform for each matrix.The signal amplitude was computed by adding the two mean waveforms, dividing by 2, and computing the RMS amplitude of the resulting waveform.Conversely, the noise floor amplitude was computed by subtracting the two mean waveforms, dividing by 2, and computing the RMS amplitude of the resulting waveform.Fig. 3 shows an example of the mean TEOAE waveforms for each condition for one participant.
MOCR strength was quantified as the decibel difference in mean TEOAE waveform amplitude obtained with and without CAS for each condition.Positive MOCR strength values indicated that the amplitude decreased with CAS.The SNR obtained without CAS was required to be ≥12 dB to be included in the analysis.In Fig. 3, the MOCR strength values are displayed in the lower right corner of each panel.

SIN recognition measurement and analysis
SIN recognition was measured in the right ear using the QuickSIN, which consists of recordings of an adult female speaking sentences in the presence of multi-talker babble noise.The test contains practice lists and 12 test lists each containing six sentences.Each sentence contains five keywords.The test materials were stored digitally on the AudioStar Pro audiometer.
Standard administration of the QuickSIN involves changing the SNR at each sentence, beginning with +25 dB and ending with +0 dB in 5-dB steps.However, these SNRs result in limited variability across normalhearing participants (Tai and Husain, 2018).To increase across-subject variability, we adapted a modified QuickSIN described by Grant et al. (2020).We utilized the test tracks that contained speech and noise on separate channels (tracks 24-35).Speech was fixed at 45 dB HL (approximately 60 dB SPL).Within a list, the SNR for each sentence changed from +5 to +0 dB in 1-dB steps. 4,5 Participants were first instructed verbally using a script that was adapted from the QuickSIN manual.To familiarize participants with the task, two practice lists were administered (the practice lists were not repeated in the experiment).Then, 12 runs were administered for the experiment (see Fig. 1).A run consisted of one test list.The order of test lists was randomly selected without replacement for each participant.
For each list, participants were scored on the percentage of keywords correctly repeated back.Each list had a total of 30 keywords.The participant's responses were scored in real-time by one rater.We also audio recorded the participants' verbal responses, which were scored offline by a second rater who was not present during the testing.To help ensure consistency across raters, each rater used detailed scoring instructions.We utilized the instructions in the QuickSIN manual.We also added additional instructions to handle various situations not addressed in the manual.As with the MEMR and MOCR measurements, scores were averaged across pairs (e.g., runs 1 and 2, runs 3 and 4, etc.), resulting in  4 During the course of the experiment, we discovered an error in the QuickSIN recording as stored on the AudioStar Pro audiometer for Track 32, List 9, Sentence 1: "Take shelter in this tent but keep still."For this sentence, the speech and noise were mixed together rather than stored on separate channels as intended.A listening check revealed that the SNR was likely higher than the intended SNR of +5 dB.Visual inspection of results suggested that scores on sentence 9 were similar to the other sentences (see Supplementary Fig. S3), therefore results for this sentence were included in the analysis. 5Due to experimenter error, sentences for one list were mistakenly presented in a different order for 10 participants compared to the remaining participants.Additionally, sentences for two lists were mistakenly presented in a different order for an additional one participant.The specific list(s) varied across participants.Because the SNRs for these sentences were presented in the correct order (+5 to +0 dB), we included the results for these participants.
six SIN recognition scores that corresponded to the two baseline conditions and four noise conditions.

Statistical analyses
Statistical analyses were conducted using SPSS ver.29.0.1.0(IBM Corp., Armonk, NY, USA).Parametric tests were utilized when assumptions were met, otherwise non-parametric equivalent tests were incorporated.The significance level was set to 0.05 for all tests.
Corrections for multiple comparisons were incorporated when applicable.

MEMR
Group mean ΔAbsorbance curves are shown in Fig. 4A.In the Baseline 2 condition, there was a decrease in mean absorbance below approximately 3 kHz and an increase in mean absorbance above approximately 3 kHz, relative to the Baseline 1 condition.A post-hoc analysis compared absorbance changes between runs in the baseline conditions, as in Sun (2016).The change in absorbance was larger between runs 1 and 2 than between runs 3 and 4 (see Supplementary Fig. S1).
Contralateral elicitors decreased absorbance below approximately 1.4 kHz and increased absorbance in the higher frequencies.Group mean |ΔAbsorbance| curves are shown in Fig. 4B.The largest value of |ΔAbsorbance| was near 0.75 kHz.
The distributions of MEMR strength values across noise conditions are shown in Fig. 4C using box and whisker plots.Low MEMR strength values were found in the Baseline 2 condition.For a given noise type, median MEMR strength increased with increasing noise level, as expected.For a given noise level, larger median MEMR strength values were found for pulsed versus steady noise.This can more clearly be  shown in Fig. 4D suggest that there is an interaction between noise type and level because there is a larger change in median MEMR strength as the noise level changes from 50 to 65 dB SPL for pulsed noise compared to steady noise.

MOCR
Four participants did not meet the 12 dB minimum SNR criterion in one or more of the without CAS conditions, so their data were excluded from the analysis.Therefore, data are reported for 39 participants.The number of participants that showed evidence of probable MEMR activation during the MOCR recordings ranged from one to five, depending on the elicitor condition.Data from these participants were only flagged but not excluded from the analyses.These results are discussed in more detail in Section 3.4.
Distributions of TEOAE signal and noise floor amplitudes are shown in Fig. 5A (see Supplementary Fig. 2 for mean signal and noise floor spectra).Paired t-tests revealed that the signal amplitudes obtained with and without CAS were significantly different for all noise conditions (p < 0.05) but not for the two baseline conditions (p > 0.05).Wilcoxon signed-rank tests revealed that the noise floor amplitudes obtained with CAS were significantly higher than those obtained without CAS for the four noise conditions (p < 0.05) but not for the two baseline conditions (p > 0.05).However, Fig. 5B demonstrates that the SNRs obtained with and without CAS were sufficiently high in all noise and baseline conditions, as evidenced by median SNRs of at least 18 dB and 25th percentiles >12 dB.As with MEMR strength, the distributions of MOCR strength values across noise conditions are shown as box and whisker plots in Fig. 5C, and the median and IQR for the noise conditions are shown in Fig. 5D.Low MOCR strength values were found in the two baseline conditions (median values of − 0.01 dB).Median MOCR strength increased with increasing noise level.Larger median MOCR strength values were found for steady noise versus pulsed noise (the opposite pattern was seen for MEMR strength).A Friedman test revealed that there was a significant difference in MOCR strength between conditions, χ 2 (5) = 156.333,p < 0.001.Pairwise comparisons are shown in Table 3. MOCR strength was not significantly different between the following conditions: Baseline 1 and Baseline 2; Steady 50 dB and Pulsed 50 dB; Steady 50 dB and Pulsed 65 dB; and Steady 65 dB and Pulsed 65 dB.It should be noted that the unadjusted p-value for the comparison of Steady 50 dB and Pulsed 65 dB conditions was significant, and the adjusted p-value approached significance.All other pairwise comparisons were significantly different.Although the Friedman test does not test for interactions, the parallel lines in Fig. 5D suggest that there was no interaction between noise type and noise level.

SIN recognition
SIN recognition scores were obtained in 43 participants.Inter-rater reliability (real-time versus offline scoring) was assessed for the first four lists of each participant by computing Spearman rank correlation coefficients and intraclass correlation coefficients.Scores between raters were significantly correlated (r S = 0.97, p < 0.05).The intraclass correlation coefficient was 0.996, which would be considered excellent reliability (Koo and Li, 2016).Hereafter, scores will be reported for those obtained for real-time scoring.
The distribution of scores across conditions is shown in Fig. 6A.For a given condition, there was a considerable range of scores across participants.In the baseline conditions, one participant reached ceiling effects.In all conditions, floor effects were avoided.Median scores across conditions ranged from 53.33 to 55%.A Friedman test revealed that there was a significant difference in score across conditions, χ 2 (5) = 20.076,p < 0.001.Pairwise comparisons are shown in Table 4. Score was significantly higher for the Steady 50 dB condition compared to the Pulsed 65 dB condition, and was significantly higher for the Steady 65 dB condition compared to the Pulsed 65 dB condition.These results were contrary to our hypothesis.
It was also of interest to examine the change in score relative to baseline.Distributions of the changes in score (ΔScore) are shown in Fig. 6B.Median differences across conditions ranged from 0 to 1.67%.For a given condition, changes were in the positive and negative direction, with some changes being as large as 51.67%.Results for individual participants are shown in Fig. 6C.Two participants showed large decrements in score relative to Baseline 1.One participant consistently showed decrements, including in Baseline 2. Another participant only showed large decrements in the noise conditions.In contrast, the median change in score ranged from − 1.67 to 5.00% depending on the condition.
Score as a function of list number was also examined to determine if there were any list effects.Results are shown in Supplementary Fig. S3.
Each panel shows the score as a function of list number for a given elicitor condition.Within a given condition, individual data (unfilled circles) and median scores (red circles) showed some variability across list number.Because list order was randomized for each participant, the number of participants who received a given list in a given condition varied (ranging from 2 to 12 participants).Due to the low numbers of participants, we did not perform statistical analyses of score across list number.However, visual inspection did not reveal any systematic differences in score across list when looking at all conditions.
Score as a function of run number was also examined to determine if there were any learning effects (increase in score across run) or fatigue effects (decrease in score across run).Results are shown in Supplementary Fig. 4. A Friedman test revealed that there was a significant  difference in score between runs, χ 2 (11) = 32.618,p < 0.001.However, all pairwise comparisons were not significant after correcting for multiple comparisons (adjusted p > 0.05 in all cases).Visual inspection of the results presented in Supplementary Fig. 4 did not reveal any systematic changes in score across run.

Associations between MEMR, MOCR, and SIN recognition
Correlations between MEMR strength, MOCR strength, and SIN recognition were examined only for participants whose data were included for all three measurements (35 participants).The following correlations were examined: MEMR strength and MOCR strength; MEMR strength and SIN recognition; and MOCR strength and SIN recognition.To reduce the number of comparisons, we only examined correlations for results obtained in the same CAS condition (e.g., correlations between MEMR and MOCR strength for the Steady 50 dB condition).Due to the presence of outliers and/or non-normality of data, two-tailed Spearman rank correlations were computed.Statistical results are shown in Table 5.All scatterplots are shown in Fig. 7.The only significant correlations were between MEMR strength and MOCR strength for the Pulsed 50 dB condition and for the Pulsed 65 dB condition.
One explanation for the significant correlations between MEMR and MOCR strength for pulsed elicitors is that the changes in TEOAE amplitude were driven by MEMR activation.We indirectly examined this possibility through the scatter plots of MOCR strength as a function of MEMR strength (top row of Fig. 7).Participants without and with probable MEMR activation during the MOCR recording are plotted as unfilled circles and filled red circles, respectively.There did not appear to be a relationship between MEMR strength as assessed using wideband absorbance and probable MEMR activation as assessed using changes in ear-canal stimulus amplitude during the TEOAE measurement.This can be seen most clearly in the panel for the Pulsed 65 dB condition.Some participants with probable MEMR activation (red circles) had relatively low MEMR strength (~4).In contrast, some participants did not exhibit probable MEMR activation (unfilled circles) despite having relatively high MEMR strength (~9-10).
We also reanalyzed the TEOAE data by bandpass filtering the waveforms from 1.5 to 4 kHz in an attempt to avoid frequencies that may have been affected by the MEMR (results are shown in Supplementary Fig. S5).Compared to the 0.5-4 kHz filtering condition (Fig. 5), TEOAE signal and noise floor amplitudes decreased somewhat but SNRs were similar.Additionally, compared to the 0.5-4 kHz filtering condition, MOCR strength decreased across the noise conditions, with the largest decreases seen for the Pulsed 50 dB and Pulsed 65 dB elicitors.Correlations between MEMR strength and MOCR strength were no longer significant for the Pulsed 50 dB condition (adjusted p > 0.05).Correlations between MOCR strength and SIN recognition scores remained non-significant after bandpass filtering from 1.5 to 4 kHz (adjusted p > 0.05 in all cases).

MEMR results
The changes in absorbance between baseline conditions were consistent with those reported by Sun (2016), where eight repeated measurements of absorbance were obtained in adults with normal hearing.The measurements were made without probe replacement.Some significant changes in absorbance were found between the first and second measurement, but not between the seventh and eighth measurements.The author discussed that this initial change in absorbance may be due to a preconditioning effect (Gaihede, 1996), where the first measurement of absorbance changes the properties of the middle ear system.Interestingly, the direction of the absorbance changes we observed (decreases below 3 kHz) was opposite of those reported by Sun (2016).The reason for the discrepancy in results is unclear and may warrant further study.
In the presence of contralateral elicitors, we found a decrease in wideband absorbance below approximately 1.4 kHz and a small increase in absorbance above 1.4 kHz (see Fig. 4A).These results are consistent with a number of studies that have measured the MEMR using wideband absorbance (Feeney et al., 2017;Keefe et al., 2010;Mepani et al., 2020;Mertes, 2022;Schairer et al., 2022;Wojtczak et al., 2017).Additionally,  MEMR strength in all noise conditions was significantly higher compared to the baseline condition, suggesting that the MEMR was elicited in the presence of the elicitors, even despite the potential preconditioning effect.We hypothesized that MEMR strength would be significantly higher for the 65 dB SPL elicitors compared to the 50 dB SPL elicitors, and that MEMR strength would be significantly higher for the pulsed noise compared to the steady noise.The only results that were consistent with our hypotheses were the significantly larger MEMR strength for the Pulsed 65 dB condition compared to the Pulsed 50 dB condition, and the significantly larger MEMR strength for the Pulsed 65 dB condition compared to the Steady 65 dB condition.It is unclear why there was no significant difference in MEMR strength between the Steady 50 dB and Steady 65 dB conditions, given the well-established effect of stimulus level on the strength of the MEMR.It is possible that these stimulus levels were not sufficiently high to elicit MEMR activity.Additionally, the lack of significant difference in MEMR strength between the Steady 50 dB and Pulsed 50 dB conditions was unexpected.MEMR strength in the 50 dB SPL conditions was significantly larger relative to baseline, suggesting that these 50 dB SPL elicitors activated the MEMR.Incorporating a wider range of elicitor levels would provide better resolution of the MEMR growth functions and allow for a determination of an individual's MEMR threshold.
As an initial investigation, our a priori hypothesis was that there would be no interaction between elicitor type and level.However, we found that the slope of the change in MEMR strength with increasing elicitor level appeared to be shallower for the steady elicitor compared to the pulsed elicitor.Including a wider range of levels and/or measuring in smaller than 15-dB steps would provide a more comprehensive characterization of the slopes.

MOCR results
In the current study, the TEOAE signal amplitudes, noise floor amplitudes, and MOCR strength values were satisfactory and consistent with our previous work (e.g., Mertes, 2018;Mertes and Stutz, 2023).As expected, the presence of CAS reduced TEOAE amplitude in nearly all cases (97.4%).Additionally, MOCR strength was significantly larger in all noise conditions compared to the baseline conditions, suggesting that the MOCR was elicited.In the noise conditions, median MOCR strength values ranged from 0.92 to 2.90 dB, consistent with our previous study that incorporated different elicitor types and levels (Mertes, 2018).Additionally, these median values are broadly consistent with various studies of MOCR strength in normal-hearing adults (De Ceulaer et al., 2001;Riga et al., 2016).
We hypothesized that MOCR strength would be significantly higher for the 65 dB SPL elicitors compared to the 50 dB SPL elicitors.Our results were consistent with this hypothesis, which is also consistent with the well-established effect of elicitor level on MOCR strength (Hood et al., 1996;Veuillet et al., 1991).Additionally, we hypothesized that MOCR strength would not be significantly different for the steady versus pulsed noises at a given stimulus level.Our results were also consistent with this hypothesis and with our previous findings (Mertes, 2018).The current results extend our previous findings because the current study utilized a slower pulse rate (0.01 kHz) compared to our previous study (0.1 kHz).These results suggest that the MOCR may be relatively insensitive to the temporal properties of the elicitor stimuli, consistent with recent studies (Boothalingam et al., 2014;Kalaiah et al., 2017; but see also Maison et al., 1997Maison et al., , 1999)).

SIN recognition results
Because our goal was to examine associations between SIN recognition and the MEMR and MOCR, it was important to maximize the across-subject variability in SIN recognition scores while also avoiding floor and ceiling effects.Previous results demonstrate that normalhearing adults reach ceiling performance on the QuickSIN for SNRs of +10 dB and higher (Tai and Husain, 2018).Recently, a modified QuickSIN has been incorporated that utilized SNRs of +10 to +0 dB (Grant et al., 2020).We made further modifications to the QuickSIN by incorporating more challenging SNRs (+5 to +0 dB).Our results demonstrated that floor effects were avoided in all conditions, and ceiling effects were generally avoided, with the exception of one participant who neared ceiling performance for the two baseline conditions.
Contrary to our hypothesis, score was significantly lower in the Pulsed 65 dB condition compared to score in the Steady 50 and 65 dB conditions.We hypothesized that score would be significantly higher in the Pulsed 65 dB condition because this condition would be most likely to activate both the MEMR and MOCR and provide the most antimasking compared to other conditions.
Because we only assessed the effect of the contralateral elicitors on score, we can only speculate about other effects of the contralateral noise.A recent review has highlighted a variety of potentially deleterious effects that noise can have on speech perception, such as increased distractibility and increased cognitive demands (Francis, 2022).It is possible that the pulsed noise presented at 65 dB SPL was more perceptually distracting or annoying compared to the steady noises, resulting in differences in score.Interestingly, there was no significant difference in score between the Pulsed 50 dB condition and the other elicitor conditions, which may suggest that low-level pulsed noise has fewer deleterious effects on score than higher-level pulsed noise.It is also possible that the antimasking provided by the contralateral elicitors may have been counteracted by deleterious cognitive effects on performance introduced by the elicitors, and that the balance between these two depends in part on the elicitor type and level.Parsing out these effects of contralateral elicitors would be important to better understand how the MEMR and MOCR benefit or do not benefit SIN recognition.
Although some subjects showed changes in SIN recognition, there were both improvements and decrements in performance for a given condition, rather than a systematic change in performance across participants.From this study, the source of these changes cannot be determined but we can speculate on potential reasons.Our analyses in Supplementary Fig. S3 suggest that there were not appreciable differences in performance across list when viewed across all elicitor conditions.This finding is expected because list equivalency has been established for the QuickSIN (Killion et al., 2004).Improvements in performance may have been due to antimasking afforded by the MEMR and/or MOCR.Although our analysis suggests that there were no systematic learning effects (Supplementary Fig. S4), improvements may also have been due to learning or training effects at the individual level.We attempted to minimize these effects by introducing two practice lists.It is also possible that some individual participants became more motivated as the task progressed, although it is unclear why this would be the case.In contrast, decrements in performance may be explained by decreases in motivation and/or attention with the task over the course of the measurement.Introduction of the contralateral noise may have been perceptually annoying or distracting to these participants, thus increasing task difficulty.We did not ask participants about their subjective perceptions of the stimuli, but this could be incorporated into future work.

MEMR and MOCR
Our investigation of correlations between the MEMR and MOCR was exploratory in nature.Boothalingam et al. (2021) found no significant correlations between MEMR and MOCR strength from 1 to 4 kHz after excluding outliers.It should be noted that they used clicks to evoke the two reflexes.Marks and Siegel (2017) measured the MEMR and MOCR in the presence of contralateral noise in a group of normal-hearing young adults.They found that the MOCR threshold was significantly lower compared to the MEMR threshold, and that the two thresholds were not significantly correlated.We did not measure threshold for the MEMR or MOCR, so direct comparisons between the current results and those of Marks and Siegel (2017) may not be warranted.
We found that the correlation between MEMR and MOCR strength depended on the elicitor type.A potential explanation is that the pulsed noise was a more effective activator of the MEMR compared to the steady noise.Therefore, measurements of the MOCR using pulsed noise may have included contributions of the MEMR, resulting in significant correlations.
It should be noted that our check of the MEMR using changes in earcanal stimulus amplitude obtained during the MOCR measurements did not show any apparent relationships with the MEMR results obtained using wideband absorbance.For example, the participants with the largest MEMR strength as measured using wideband absorbance did not always show probable MEMR as assessed using changes in ear-canal stimulus amplitude (see top row of Fig. 7).One interpretation is that the ear-canal stimulus amplitude method is less sensitive to detecting MEMR compared to wideband absorbance (Mertes, 2022).However, the higher stimulus level for wideband absorbance may have evoked the ipsilateral MEMR, and introducing the contralateral elicitor may have resulted in binaural facilitation of the MEMR to a greater degree than during the TEOAE measurement (Kawase et al., 1997;Simmons, 1965).The relative sensitivity of the two methods to MEMR detection warrants further study and would benefit from concurrent measurement using the same stimuli and equipment.
Given that the contralateral elicitors evoked the MEMR as measured using wideband absorbance, it is important to consider whether the MOCR measurements were affected by MEMR activation.Generally, the contralateral elicitors caused a reduction in absorbance below 1.4 kHz and an increase in absorbance in the higher frequencies (Fig. 4A), which is broadly consistent with previous results (Boothalingam and Goodman, 2021;Keefe et al., 2017).In contrast, the contralateral elicitors caused a reduction in TEOAE amplitude from 0.5 to 4 kHz (Supplementary Fig. S2).It is possible that the MEMR contributed to the decrease in TEOAE amplitude for frequencies <1.4 kHz, especially in the Pulsed 50 dB and Pulsed 65 dB conditions (compare Fig. 5D to Supplementary Fig. S5D).This suggests that the unfilled circles shown in Fig. 7 do not necessarily indicate absence of MEMR activation.However, the extent of the contribution of the MEMR cannot be determined from the current results.Additionally, it appears less likely that the MEMR contributed to the decrease in TEOAE amplitude above 1.4 kHz, given that the direction of the change in absorbance was opposite of what was observed for the change in TEOAE amplitude.Even after reanalyzing the TEOAE data bandpass filtered from 1.5 to 4 kHz to avoid frequencies affected by MEMR activation, we still found that there was no significant correlation between MOCR strength and SIN recognition.This suggests that the MEMR and MOCR were not associated with SIN recognition as measured and analyzed in this study.

MEMR and SIN recognition
The lack of significant correlations between MEMR strength and SIN recognition was inconsistent with our hypotheses.MEMR activation acts as a highpass filter that should reduce the upward spread of masking (Borg, 1968;Pang and Guinan, 1997).A role of the MEMR in SIN recognition has been supported through studies of patients with compromised MEMR function due to conditions such as otosclerosis or stapedectomy without preservation of the stapedius tendon (Aiken et al., 2013;Colletti et al., 1988;Lidén et al., 1964).
Another source of evidence is studies that have demonstrated significant positive correlations between MEMR strength and word recognition in noise when the words were presented at average conversational levels (Harkrider and Smith, 2005;Mepani et al., 2020) and at high intensity levels (Shehorn et al., 2020).One plausible explanation for these results is that individuals with stronger MEMR have increased antimasking effects that benefit their SIN recognition.
In contrast, Karunarathne et al. (2018) found that the magnitude of the MEMR was inversely related to signal-in-noise detection.This finding suggests that a stronger MEMR may increase, rather than decrease, difficulty with detecting an alarm in the presence of noise.However, as described above in Section 1, their study inferred MEMR activation through contralateral inhibition of TEOAE measures rather than through a dedicated measure of the MEMR.Additionally, their use of non-speech stimuli concentrated at 2 kHz would complicate direct comparisons to results obtained using speech stimuli.
Interestingly, Mepani et al. (2020) found significant correlations between MEMR strength and word recognition in noise but not sentence recognition in noise.The authors speculated that the presence of contextual cues inherent in the sentence materials of the QuickSIN may have contributed substantially to participants' performance, regardless of their MEMR strength.There were differences between their study and ours in terms of administration of the modified QuickSIN as well as measurement and analysis of MEMR strength.Additionally, Mepani et al. (2020) examined correlations between SIN recognition and MEMR strength obtained at the highest elicitor (95 dB SPL) whereas we examined correlations for MEMR strength obtained for the same elicitor level used for the MEMR and QuickSIN measurements.Despite these differences, the similar results between studies suggest that the lack of correlation between MEMR strength and sentence-in-noise recognition is a robust finding.
Despite the finding of no significant correlations between MEMR strength and SIN recognition, it is still possible that the MEMR contributes to sentence-in-noise recognition, but its relative contribution to an individual's task performance may depend on the presence of other factors.For example, the ability to utilize contextual information present within the SIN recognition materials may benefit performance on the task (Elliott, 1995;Kalikow et al., 1977).The QuickSIN has been considered to have limited contextual cues (DiNino et al., 2022;Wilson et al., 2007), but the effects of context in the current study cannot be ruled out.Cognitive factors such as working memory, motivation, and attention can also contribute to performance (reviewed in Akeroyd, 2008;Dryden et al., 2017;Eckert et al., 2016).Future work could measure and potentially control for these factors to aid in determining the contribution of the MEMR to SIN recognition.

MOCR and SIN recognition
As with the MEMR, the lack of significant correlations between MOCR strength and SIN recognition was inconsistent with our hypotheses.Our hypotheses derived from converging lines of evidence, including animal work on antimasking (Kawase et al., 1993;Winslow and Sachs, 1987), effects of olivocochlear lesions on auditory perception in noise (Dewson, 1968;May and McQuone, 1995), models of SIN recognition that include olivocochlear feedback (Brown et al., 2010;Clark et al., 2012), and correlational studies of the MOCR strength and SIN recognition in humans (Abdala et al., 2014;Giraud et al., 1997;Kumar and Vanaja, 2004;Maruthy et al., 2017;Mertes et al., 2018Mertes et al., , 2019;;Mishra and Lutman, 2014;Tokgoz-Yilmaz et al., 2013;Yilmaz et al., 2007).Despite this evidence, other studies in humans have found no significant associations (Gafoor and Uppunda, 2023b;Harkrider and Smith, 2005;Mertes and Stutz, 2023;Mukari and Mamat, 2008;Stuart and Butler, 2012;Yashaswini and Maruthy, 2019) or negative associations (Karunarathne et al., 2018).Potential explanations for the discrepancies in findings include differences in participants, measurement and analysis of the MOCR, and/or measurement and analysis of SIN recognition across studies (Gafoor and Uppunda, 2023b;Mertes et al., 2018;Mishra and Lutman, 2014).A recent meta-analysis found that MOCR strength explained a negligible amount of the variance in SIN recognition (<1%), so the authors concluded that the literature does not support a role for the MOCR (as assessed using contralateral inhibition of OAEs) for SIN recognition (Gafoor and Uppunda, 2023a).The authors also conducted separate meta-analyses on studies that were similar in methodology, but this did not reveal a relationship between MOCR strength and SIN recognition.
As with the MEMR, it is possible that cognitive factors unrelated to the MOCR contributed substantially to participants' performance on the QuickSIN.It has also been posited that assessing the MOCR in a passive measurement separate from SIN recognition may poorly characterize MOCR activity as it occurs during a listening task (de Boer et al., 2012).MOCR activity can be influenced by attention through corticofugal projections from the auditory cortex and inferior colliculus (reviewed in Lauer et al., 2022).To account for this, the MOCR could be measured during the listening task, but this poses methodological challenges such as ensuring that OAEs can be elicited while not being masked by speech or noise stimuli.Two studies have used contralateral elicitors that combined speech and noise, and participants had to complete a task such as categorizing the word (Garinis et al., 2011) or determining the sex of the speaker (Smith and Cone, 2015).It is possible that these measurement paradigms could be extended to a word or sentence recognition task.Alternatively, spontaneous OAEs, which can be elicited without an external stimulus, have been used to measure MOCR activity during a SIN recognition task (Marian et al., 2018), but spontaneous OAEs can be absent despite normal hearing (Sisto et al., 2001).Development of paradigms that allow for measurement of evoked OAEs could provide insights into the functioning of the MOCR during typical listening situations.

Other potential efferent effects 4.5.1. Ipsilateral MEMR and MOCR
The focus of this study was on the contralateral MEMR and MOCR, where a reflex elicitor was introduced to the contralateral ear and the effects of reflex activation were measured in the ipsilateral ear.Specifically, effects of reflex activation were quantified as the change in absorbance or TEOAE amplitude without the contralateral elicitor (i.e., the baseline conditions) versus with the contralateral elicitor.It must be acknowledged that the ipsilateral stimuli (100 dB peSPL clicks for the MEMR and 65 dB pSPL clicks for the MOCR) may have activated the ipsilateral MEMR and MOCR (Guinan et al., 2003).If so, the reflexes would already be partially activated in the baseline condition and the change in absorbance or TEOAE amplitude due to the contralateral elicitor could be reduced.
It has been established that the likelihood of ipsilateral MEMR and MOCR activation increases as the ipsilateral stimulus bandwidth, level, and rate increases (Boothalingam and Goodman, 2021;Boothalingam and Purcell, 2015;Guinan et al., 2003;Rawool, 1995).Johnsen and Terkildsen (1980) reported that a click rate of 16/s (the closest rate to the current study) yielded a mean MEMR threshold of 101.8 dB peSPL with a range of 83-113 dB peSPL.It should be noted that the authors did not report the probe tone frequency.In a series of studies, a click rate of 50/s was found to yield MEMR thresholds of approximately 95-98 dB peSPL in young adults with normal hearing (Rawool, 1995(Rawool, , 1996a(Rawool, , 1996b)).These studies did not examine rates below 50/s, but extrapolating the results in Fig. 1 of Rawool (1995) to 21.5/s (the rate used in the current study) suggests that the mean MEMR threshold would be higher than 100 dB peSPL (the level used in the current study).However, this series of studies used a 0.226-kHz probe tone, so the extent to which the MEMR was activated at higher frequencies is not known.Recently, Boothalingam and Goodman (2021) demonstrated that clicks presented at 95 dB ppSPL with a rate of 62.5/s elicited the MEMR in 100% of young adults with normal hearing when the MEMR was measured from 1 to 3.2 kHz.These results could suggest that the ipsilateral MEMR was evoked during the absorbance measurements in the current study.However, there were differences in stimulus rate and level between their study and the current study.Additionally, they presented clicks bilaterally whereas the current study presented clicks ipsilaterally.Therefore, it is difficult to estimate the percentage of participants in the current study who had ipsilateral MEMR activation, but it appears possible that the ipsilateral MEMR was evoked in at least some participants.Future work could verify if the ipsilateral MEMR is activated using methods reported in Boothalingam and Goodman (2021).Alternatively, the ipsilateral click level and/or rate could be reduced below the ipsilateral MEMR threshold.For example, Goodman et al. (2021) recently reported that none of their normal-hearing adult participants exhibited MEMR activation in response to clicks presented at a rate of 5/s and a level of 91 dB peak-to-peak forward pressure level (ppFPL).
The ipsilateral click stimuli for the MOCR measurements also have the potential to activate the ipsilateral MOCR.Veuillet et al. (1991) reported that for click stimuli presented at 30 dB sensation level (SL), rates <50.3/s(reported as an inter-stimulus interval of >19.9 ms) did not elicit significant MOCR activity.Boothalingam and Purcell (2015) found that for click stimuli presented at 55 dB peSPL, rates <31.25/s did not elicit significant MOCR activity.Comparisons between these studies and the current study are complicated by differences in stimulus level, and Veuillet et al. (1991) did not report the stimulus level in dB peSPL.However, it seems likely that our relatively low level of 65 dB pSPL and slow rate of 21.5/s minimized elicitation of the ipsilateral MOCR.As described above for the MEMR, it would be preferable to either directly measure ipsilateral MOCR activation for a given stimulus or select stimulus parameters that would avoid ipsilateral MOCR activation in all participants.
An additional consideration is that the ipsilateral noise presented during the QuickSIN may have activated the ipsilateral MEMR and/or MOCR to benefit SIN recognition, and so the introduction of the contralateral elicitors may have yielded little or no improvement in SIN recognition score.This could be one explanation for the lack of significant improvement in score for the noise conditions relative to the baseline conditions.The influence of ipsilateral MEMR and MOCR could be limited in a future study by using low levels of ipsilateral speech and noise that would fall below the reflex threshold.

MEMR and MOCR slow effects
Human data has established that the MEMR has an onset latency in the range of 50-80 ms, reaches its maximum amplitude in the range of 200-1000 ms after the presentation of the elicitor, and returns to its baseline amplitude in the range of 500-1400 ms after the elicitor is turned off (Boothalingam and Goodman, 2021;Borg, 1982;Church and Cudahy, 1984;Gorga and Stelmachowicz, 1983;Hung and Dallos, 1972).Recently, Jennings and Dominguez (2022) reported an additional "slow" effect of the MEMR that increased in amplitude over the course of an 11.6-min measurement that involved high-level clicks and noise bursts to elicit the compound action potential.Slow effects could influence the measured MEMR strength, which could complicate comparisons of MEMR strength across elicitor conditions.Jennings and Dominguez (2022) noted that their interstimulus interval of 273 ms was likely too brief to allow for offset of the MEMR between adjacent stimuli, thus leading to the slow effects over the course of the measurement.In our study, we introduced a 5-s pause between runs in the MEMR measurements.It is possible that this pause was long enough to allow for complete offset of any MEMR slow effects prior to the next measurement.Obtaining additional runs without CAS at the end of an MEMR measurement could allow for assessment of any systematic changes in MEMR strength over the course of the experiment.Therefore, we cannot rule out the possibility that MEMR slow effects were present in the current study.
The MOCR also operates on different time scales, often referred to as "fast effects" with a time course of ranging from 70 to 210 ms and "slow effects" with a time course greater than 10 s (Backus and Guinan, 2006;Boothalingam et al., 2021).Based on these values, it is likely that MOCR slow effects were present during a single run because the median run duration was 209.5 s. 3 It is not known if the time between adjacent runs (median = 13.8 s) 3 was sufficient to allow for slow effects in a given run to decay prior to the next run.As with the MEMR measurement, adding an additional measurement without CAS at the end could allow for assessment of systematic changes in MOCR strength over the course of the experiment.It should be noted that one study in humans demonstrated that MOCR slow effects are very minimal in amplitude, especially in comparison to the MOCR fast effects (Zhao and Dhar, 2011).Although MOCR slow effects may have been present in the current study, it is possible that they had minimal influence on the measured MOCR strength.

Limitations and future directions
Our inclusion criterion of ipsilateral acoustic reflex thresholds ≤90 dB HL bilaterally was intended to exclude participants with potential retrocochlear pathology.Six participants were excluded based on this criterion.However, this may have reduced the variability in MEMR results across participants.Future work on the MOCR and MEMR could use alternative methods, such as the auditory brainstem response (Godey et al., 1998), to exclude participants with potential retrocochlear pathology while limiting the effect on inter-subject variability on the MEMR.
Our detection of probable MEMR activation during the MOCR measurements was based on methods described in Mertes (2020).It should be noted that one goal of this method was to develop a relatively quick check of MEMR activation, so only one interleave was used.Therefore, this method may not have detected slow MEMR effects (Jennings and Dominguez, 2022).Future investigations could determine the extent to which this methodology can detect slow MEMR effects.
By including the two baseline conditions at the beginning of a measurement, this would presumably minimize potential effects of response amplitude drift due to changes in middle ear pressure, probe slippage, or other factors (Backus, 2007;Goodman et al., 2013).Minimizing drift in the baseline conditions was important because we wanted to determine if there were significant differences between baseline conditions (e.g., row 1 of Table 3), and the presence of drift would complicate this determination.However, this approach would not minimize drift across the duration of the experiment.A preferable approach would be to flank each noise condition with a baseline/control condition (i.e., without CAS) because the results for a noise condition could be compared to the baseline/control condition obtained immediately prior.For the current study, this would require a total of 18 runs ([5 baseline conditions + 4 noise conditions] × 2 replications).However, the QuickSIN corpus only consists of 12 test lists, and repeating a list could potentially compromise the validity of the test due to learning effects.Utilizing a larger corpus in a future study could minimize these issues and allow for a larger number of runs to be obtained.
Equipment limitations prevented simultaneous measurement of the MEMR and MOCR using the same stimuli and equipment, but this could be incorporated in future studies.Additionally, equipment limitations required that the CAS be turned on and off manually for the MEMR measurements, which likely introduced variability in onset of MEMR activation within and across participants.Results of the simulation described in Footnote 1 revealed a minimum delay of 1.25 s between turning on the CAS and running the absorbance measurement. 1 Previous work has found that the MEMR reaches its maximum amplitude by 1 s or less (Church and Cudahy, 1984;Gorga and Stelmachowicz, 1983;Hung and Dallos, 1972).However, it would be preferable to automate the presentation of the elicitor and absorbance measurement (e.g., as in Keefe et al., 2017) to eliminate this source of variability.
We used two elicitor levels to provide some information about level effects and to ensure that all testing could be completed within a single 2.5-h test session.However, this did not allow for a detailed investigation into level effects.Additionally, our pulse rate of 0.01 kHz for the pulsed elicitor was based on pilot data, but it is possible that other pulse rates may be more effective elicitors of the MEMR.A systematic investigation of different pulse rates could provide a more complete characterization of MOCR and MEMR activation.As described above, we cannot rule out the possibility of MEMR and MOCR slow effects, which may have affected the correlations and/or effect sizes.Future work could account for slow effects by measuring them (e.g., Jennings and Dominguez, 2022;Zhao and Dhar, 2011) or choosing experimental parameters that avoid them.Because the QuickSIN contains some contextual cues, future work could consider utilizing other SIN tests, such as digits in noise, which reduce the contextual cues and presumably the working memory needed to perform the task compared to a sentence task.The current study focused on a relatively homogenous group of young adults with normal hearing.It is possible that the contributions of the MEMR and MOCR to SIN recognition may be more apparent when comparing results with individuals who report difficulties with hearing in noise and/or who have hearing loss, warranting further study.
Finally, our results suggest that pulsed noise elicitors are potent activators of both the MEMR and MOCR.There may be advantages to using such stimuli if the goal is to evoke both reflexes.For example, in the "Post-Talk Q&A" section of Goodman et al. (2018), it was discussed that measuring both reflexes simultaneously (rather than separating out the individual effects of each) could potentially be useful for applications such as predicting an individual's susceptibility to noise-induced hearing loss.Future studies could investigate the potential clinical utility of using pulsed noise elicitors for such applications.

Fig. 1 .
Fig. 1.Schematic of the general protocol.The top row shows the elicitor condition name.The second row shows the run number.Curly brackets indicate that the results for those runs were averaged together for a given elicitor condition.The third row shows the CAS type and level.CAS-indicates that no contralateral noise was presented.The fourth row shows the stimuli for the MEMR and MOCR measurements.Red and blue font indicate the stimuli for the right and left ears, respectively.The fifth row shows the stimuli for the SIN recognition measurements, where the list number refers to the QuickSIN.It should be noted that the order of the noise conditions was counterbalanced and the order of the QuickSIN list numbers was randomized across participants, so this figure demonstrates one possible order.

Fig. 2 .
Fig. 2. (A) Mean absorbance curves obtained from one representative participant.(B) Change in absorbance relative to baseline for the same participant shown in panel A. Dashed line represents no change in absorbance.Figure format is based on Fig. 5 of Keefe et al. (2017) and Fig. 2 of Wojtczak et al. (2017).(C) Absolute value of the change in absorbance for the same participant shown in panel A. Figure format is based on Fig. 1 of Mertes (2022).

Fig. 3 .
Fig. 3. Mean TEOAE waveforms in each condition for the same participant shown in Fig. 2. Each row shows results for a different elicitor type.Within each panel, the gray waveform was obtained without CAS (CAS-) and the black waveform was obtained with CAS (CAS+).Note that in the baseline conditions, the gray and black waveforms were both obtained when no CAS was presented.MOCR strength values are shown in the lower right corner of each panel.

Fig. 4 .
Fig. 4. (A) Mean ΔAbsorbance curves computed across participants.Format is identical to Fig. 2B, with the exception of shaded regions denoting ±1 SEM.(B) Mean |ΔAbsorbance| curves computed across participants.Format is identical to Fig. 2C, with the exception of shaded regions denoting ±1 SEM.(C) Distributions of MEMR strength in each condition.Horizontal red lines are the medians.Boxes encompass the 25th and 75th percentiles.Whiskers extend to the smallest and largest values not considered outliers.Red cross symbols denote outliers.(D) Median MEMR strength for the noise conditions.Errors bars extend to the 25th and 75th percentiles.

Fig. 5 .
Fig. 5. (A) Median TEOAE signal amplitudes (circles) and noise floor amplitudes (squares) for results obtained without and with CAS (CAS-and CAS+, respectively).Error bars extend to the 25th and 75th percentiles.(B) Median TEOAE SNRs for results obtained without and with CAS.Error bars extend to the 25th and 75th percentiles.(C) Distributions of MOCR strength in each condition.Figure format is identical to Fig. 4C.(D) Median MOCR strength for the noise conditions.Figure format is identical to Fig. 4D.
Fig. 5. (A) Median TEOAE signal amplitudes (circles) and noise floor amplitudes (squares) for results obtained without and with CAS (CAS-and CAS+, respectively).Error bars extend to the 25th and 75th percentiles.(B) Median TEOAE SNRs for results obtained without and with CAS.Error bars extend to the 25th and 75th percentiles.(C) Distributions of MOCR strength in each condition.Figure format is identical to Fig. 4C.(D) Median MOCR strength for the noise conditions.Figure format is identical to Fig. 4D.
Fig. 5. (A) Median TEOAE signal amplitudes (circles) and noise floor amplitudes (squares) for results obtained without and with CAS (CAS-and CAS+, respectively).Error bars extend to the 25th and 75th percentiles.(B) Median TEOAE SNRs for results obtained without and with CAS.Error bars extend to the 25th and 75th percentiles.(C) Distributions of MOCR strength in each condition.Figure format is identical to Fig. 4C.(D) Median MOCR strength for the noise conditions.Figure format is identical to Fig. 4D.

Fig. 6 .
Fig. 6. (A) Distributions of SIN recognition scores in each condition.Figure format is identical to Fig. 4C.(B) Distributions of change in SIN recognition scores in each condition, relative to Baseline 1. Positive values indicate increases in score.Figure format is identical to Fig. 4C.(C) Change in SIN recognition score for individual subjects.

Fig. 7 .
Fig. 7. Scatter plots between MEMR strength and MOCR strength (top row), MEMR strength and SIN recognition (middle row), and MOCR strength and SIN recognition (bottom row) for each condition.The unfilled circles in each panel are individual participant data (n = 35).The gray diagonal line in each panel represents a 1:1 correspondence between the two variables.The red filled circles in the top row represent participants with probable MEMR based on changes in earcanal stimulus amplitude (see text for details).

Table 1
95% critical differences in ear-canal stimulus amplitude for each elicitor type and elicitor level condition.

Table 2
Results of pairwise comparisons of MEMR strength across conditions.4D,which plots the medians and IQRs for the noise conditions.A Friedman test revealed that there was a significant difference in MEMR strength between conditions, χ 2 (4) = 106.716,p < 0.001.Pairwise comparisons are shown in Table 2.All pairwise comparisons were significant with the exception of the following: Steady 50 dB and Steady 65 dB; Steady 50 dB and Pulsed 50 dB; and Steady 65 dB and Pulsed 50 dB.Although the Friedman test does not test for interactions, the results

Table 3
Results of pairwise comparisons of MOCR strength across conditions.

Table 5
Associations between MEMR and MOCR strength, MEMR strength and SIN recognition, and MOCR strength and SIN recognition.Spearman correlation coefficients (r s ) and the associated p-values and adjusted p-values are shown.

Table 4
Results of pairwise comparisons of SIN recognition scores conditions.