The role of the medial olivocochlear reflex in psychophysical masking and intensity resolution in humans: a review

Abstract

This review addresses the putative role of the medial olivocochlear (MOC) reflex in psychophysical masking and intensity resolution in humans. A framework for interpreting psychophysical results in terms of the expected influence of the MOC reflex is introduced. This framework is used to review the effects of a precursor or contralateral acoustic stimulation on 1) simultaneous masking of brief tones, 2) behavioral estimates of cochlear gain and frequency resolution in forward masking, 3) the buildup and decay of forward masking, and 4) measures of intensity resolution. Support, or lack thereof, for a role of the MOC reflex in psychophysical perception is discussed in terms of studies on estimates of MOC strength from otoacoustic emissions and the effects of resection of the olivocochlear bundle in patients with vestibular neurectomy. Novel, innovative approaches are needed to resolve the dissatisfying conclusion that current results are unable to definitively confirm or refute the role of the MOC reflex in masking and intensity resolution.

INTRODUCTION

Hearing is mediated by signal processing of sound energy from a series of biological systems in the peripheral and central auditory systems, including the cochlea. A hallmark of cochlear signal processing is nonlinear amplification of basilar membrane (BM) motion by outer hair cells (OHCs) (1). This amplification results in sensitivity to soft sounds and exquisite frequency resolution, and loss of this amplification is a primary cause of cochlear hearing loss (HL) (2). Amplification by OHCs is modulated by efferent input from the medial olivocochlear (MOC) reflex (3, 4). Research in laboratory animals has inspired the development of hypotheses on the functional role of the MOC reflex in human hearing. These hypothesized roles include protection from intense sound exposure (5), facilitation of selective attention (6), early development of inner hair cells (IHCs) (7), prevention of cochlear aging (8), and reduction of masking from background noise (9). The hypothesis that the MOC reflex reduces masking in human hearing has been the focus of intense research for decades. The motivation for this research is that a reduction in masking by the MOC reflex may explain robust speech perception in adults with normal hearing (NH) and account for the hearing difficulties experienced by adults with cochlear HL when listening in noisy backgrounds (10).

This article reviews relevant literature on the relationship between the MOC reflex and psychophysical masking in adult humans, as studied by behavioral and noninvasive physiological methods such as otoacoustic emissions (OAEs), and the effects of surgical resection of the MOC bundle for patients undergoing vestibular neurectomy. Discussion of this relationship is extended to similar psychophysical tasks such as intensity discrimination, amplitude modulation (AM) detection, and other basic auditory percepts. Although numerous studies on the relationship between the MOC reflex and speech perception exist (e.g., Refs. 11–14), review of such studies is beyond the scope of this article. First, a brief description of the anatomy of the MOC system (section: brief review of the anatomy of the moc system) is provided, followed by a description of the effects of the MOC reflex on basilar membrane (section: effect of the moc system on basilar membrane mechanics) and auditory nerve responses (section: effect of the moc system on auditory nerve responses). Sections titled a single-channel framework for moc effects in psychophysics and approaches to studying the moc reflex with psychophysics review approaches to studying the MOC reflex with psychophysics under a common framework. Finally, sections titled effects of cutting the olivocochlear bundle on psychophysical performance and associations among perceptual and otoacoustic measures of moc reflex function cover studies that attempt to relate anatomical (i.e., severed MOC bundle) and physiological (e.g., contralateral suppression of OAEs) findings of MOC function to perceptual results in the same human subjects. The primary topic of this review is the psychophysical correlates of the MOC reflex. The lateral olivocochlear (LOC) system is not reviewed for several reasons, including 1) the LOC system is poorly understood because of limitations in measuring from and stimulating the thin, unmyelinated neurons of this system (15); 2) most or all sound-evoked olivocochlear efferent effects have been shown to originate from the MOC system (16); 3) LOC effects operate on a long timescale (several minutes) and therefore have been implicated in protecting against acoustic trauma (17, 18) and balancing binaural inputs to the central auditory system (3) rather than in reducing masking; and 4) there is not a clear framework for how the LOC system influences masking; however, this system may be modulated by attention when detecting unexpected tones (19). Several excellent reviews of the olivocochlear system provide greater depth on topics such as olivocochlear anatomy (20), olivocochlear physiology (3), efferent-related suppression of OAEs (4), and a broad overview of olivocochlear anatomy, physiology, and related perceptual findings (15).

BRIEF REVIEW OF THE ANATOMY OF THE MOC SYSTEM

MOC neurons are located within the first major binaural processing center in the brain stem known as the superior olivary complex (SOC) (21). Primary afferent nuclei of the SOC are the medial superior olive (MSO) and the lateral superior olive, which process interaural time and intensity differences, respectively (22). Cell bodies of MOC neurons are located in the periolivary region surrounding the MSO (23) and receive afferent and efferent inputs (24). Afferent inputs originate bilaterally from the posterior-ventral and anterior-ventral cochlear nucleus via the small cell cap (25), which is a primary target of auditory nerve (AN) fibers with low and medium spontaneous rates (26). Efferent inputs to MOC neurons originate bilaterally from the thalamus and inferior colliculus (IC) (20, 27), and the auditory cortex (28).

The primary targets of MOC axons are the cochlear OHCs. These axons form part of the vestibular portion of the VIIIth cranial nerve before entering the cochlea. MOC axons emerge from the SOC to form two bundles that take distinct paths toward the OHCs, called the “crossed” and “uncrossed” MOC bundles, where the term “crossed” indicates whether the axons cross the midline in route to the OHCs. Specifically, axons of the crossed MOC bundle travel across the midline to innervate the contralateral OHCs, whereas axons of the uncrossed MOC bundle innervate the ipsilateral OHCs. Afferent and efferent fibers innervating MOC neurons form a sound-evoked feedback loop known as the MOC reflex (29). This reflex has ipsilateral and contralateral pathways. The ipsilateral pathway originates from afferent fibers that respond to stimulation of the ipsilateral cochlea. These fibers cross the midline to synapse on crossed MOC bundle neurons, which in turn send axons back to the OHCs of the ipsilateral cochlea. Thus, the neural pathway for the ipsilateral MOC reflex crosses the midline twice, once via afferent fibers projecting to crossed MOC bundle neurons and again via MOC bundle axons projecting to OHCs. The contralateral pathway originates from afferent fibers that respond to stimulation of the contralateral cochlea. These fibers cross the midline to synapse on uncrossed MOC bundle neurons, which send axons to OHCs located in the ipsilateral cochlea. Differences in innervation pattern, density, and ear of stimulation between crossed and uncrossed MOC bundles have led to the hypothesis that the ipsilateral and contralateral MOC reflex pathways may differ in their functional roles for hearing (30). In addition to the sound-induced reflexive action of the MOC system, this system is modulated by descending inputs from the auditory cortex. For example, Dragicevic et al. (31) showed that electrical microstimulation of the auditory cortex increased and decreased the amplitude of the compound action potential and cochlear microphonic in a frequency- and level-dependent manner. These cortical effects on MOC function may contribute to diminishing the perception of tinnitus (31) and modulating perception in visual (32) or auditory (33) selective attention tasks.

EFFECT OF THE MOC SYSTEM ON BASILAR MEMBRANE MECHANICS

A long-standing theory of cochlear mechanics assumes that cochlear responses are a combination of “passive” and “active” mechanics (34). Passive mechanics emerge from the physical/metabolically inert aspects of cochlear anatomy that respond to a driving force in vivo and postmortem. Active mechanics originate from the metabolically sensitive cochlear amplifier, which is based on OHC motility (35) and powered by active transport mechanisms of the stria vascularis (36). This active transport maintains the endocochlear potential, which supplies current to drive OHC motility (37). The mechanical response of the cochlea is often studied by measuring the displacement or velocity of the BM (38). Here the general term “basilar membrane response” refers to either of these quantities. In healthy animals, responses of the BM in the base of the cochlea to tones presented at the characteristic frequency (CF) are dominated by active mechanics at low sound levels (e.g., <50 dB SPL). At high sound levels (e.g., >80 dB SPL) and for tones presented well below or above the CF (i.e., “off-frequency” tones), the contribution of active mechanics diminishes and BM responses are primarily a result of passive mechanics (39). In the cochlear apex, active mechanics contribute to BM responses to tones at CF and, to some extent, tones above and below CF (40). The shift from primarily active mechanics at low sound levels to primarily passive mechanics at high sound levels results from a progressive decrease in cochlear amplifier gain with increasing sound level. This compressive nonlinearity is evident when BM responses are measured as a function of sound level for tones presented at CF in the base and apex of the cochlea. A plot of BM response (output) versus sound level (input) illustrates the BM input/output (I/O) function, which is characterized by linear response growth at low levels and compressive response growth at moderate to high levels. Figure 1A displays an example BM I/O function from the base of the chinchilla cochlea measured by Ruggero et al. (41). The contribution of active mechanics to BM responses is evident when comparing this function premortem (Fig. 1A, dashed blue line) and postmortem (Fig. 1A, solid red line). The velocity of the BM measured postmortem is greatly reduced at low to moderate sound levels (e.g., ∼30-dB reduction at 60 dB SPL) and minimally decreased at high sound levels (e.g., ∼4-dB reduction at 90 dB SPL). This level-dependent loss of cochlear amplifier gain postmortem is accompanied by an appreciable shift in threshold and linear rather than nonlinear response growth (Fig. 1A, red solid line).

Figure 1.

The effects of hearing loss (A) and medial olivocochlear (MOC) stimulation (B) on basilar membrane (BM) responses to tones presented at the characteristic frequency. Blue dashed lines show BM input/output (I/O) functions for a healthy chinchilla (A) and guinea pig (B). Red/green solid lines show BM I/O functions measured postmortem (A) and after electrical stimulation of the MOC bundle (B). Data are replotted from Ruggero et al. (41) and Cooper and Guinan (46) in A and B, respectively.

The primary effect of MOC stimulation is to decrease the contribution of active mechanics by inhibiting OHC motility and thereby reducing BM gain (42). The effect of MOC stimulation on BM gain has led to the hypothesis that the MOC bundle acts as an automatic gain control that regulates cochlear output (43). MOC bundle stimulation decreases cochlear responses for low-/moderate-level tones presented at or near CF (44, 45). Figure 1B displays BM I/O functions from the base of the cochlea in a guinea pig, as measured by Cooper and Guinan (46). Here, tones were presented at CF and I/O functions were measured with (Fig. 1B, solid green line) or without (Fig. 1B, dashed blue line) electrical stimulation of the MOC bundle. This stimulation attenuated BM displacement by ∼12 dB for low-level inputs (e.g., 15 dB SPL), whereas little to no attenuation is observed for levels higher than 60 dB SPL. Stimulation of the MOC bundle has little to no effect on BM responses to tones presented well below or well above the CF, resulting in an elevated tip of tuning curves measured from BM responses (45). This elevation may broaden (46) or sharpen (44) frequency resolution depending on the displacement/velocity criterion for measuring the tuning curve (e.g., 10 nm) or the quality factor (Q) calculation (e.g., Q10 vs. Q20). In contrast, MOC bundle stimulation in cats consistently decreases frequency resolution (Q20) for most AN fibers with CFs above 2 kHz, regardless of spontaneous rate (47).

Until recently, cochlear mechanics was studied primarily by measuring responses of the BM. Developments in heterodyne low-coherence interferometry (see Ref. 48) have allowed physiologists to measure the mechanics of the reticular lamina (RL). The proximity of the RL to the mechanical-electrical transduction (MET) channels of the cochlear hair cells results in a closer approximation of the drive placed on cochlear hair cells for RL responses than for BM responses. Measurements of RL vibration contradict the notion that active mechanics operate via a local feedback loop and influence only on-frequency stimulation (i.e., at and surrounding the CF; Ref. 49). Instead, RL measurements are consistent with active mechanics contributing to on- and off-frequency stimulation. Guinan (18) reviewed literature on the MOC bundle in light of recent findings on RL mechanics. He noted that MOC bundle stimulation results in a reduction in firing rate of single-unit AN fibers and that this reduction may occur at high sound levels and for off-frequency stimulation (50, 51). Although this finding contradicts theories of BM mechanics, MOC bundle effects on RL motion are expected for high sound levels and off-frequency stimulation when considering the contribution of active processing to RL mechanics. Currently, this expectation remains untested, as the effects of MOC bundle stimulation on RL mechanics have not been directly studied; thus, this review interprets much of the relevant literature from a BM-mechanics perspective. This interpretation will be updated when more is known about how RL mechanics are influenced by MOC bundle stimulation.

EFFECT OF THE MOC SYSTEM ON AUDITORY NERVE RESPONSES

The encoding of sound by AN fibers is often quantified by firing rate and/or spike timing patterns (e.g., Ref. 52). Research on the influence of the MOC reflex on AN responses has primarily centered on rate-based coding (e.g., Ref. 53). For tones presented at CF and in a quiet background, the firing rate of AN fibers increases with increasing tone level. This firing rate versus tone level relationship (i.e., rate-level function) consists of three segments (Fig. 2B, black dotted lines). The lower-level segment has a low firing rate (spontaneous rate) and a shallow slope and represents AN firing to tones below and near firing rate threshold. The high-level segment occurs 30–60 dB above threshold (54, 55) and has relatively higher firing rates (saturation rate) and a shallow slope. This segment is characterized by hard saturation (very shallow slope) for high- and medium-spontaneous rate fibers and sloping saturation (a more gentle slope) in low-spontaneous rate fibers (56, 57). The midlevel segment has a steep slope and is delineated by firing rates between spontaneous activity and the saturation rate. This segment (“dynamic segment”) encompasses firing rates where the AN fiber is most sensitive to changes in input sound pressure and defines the dynamic range of rate-based coding for the fiber (58, 59).

Figure 2.

Antimasking of auditory nerve (AN) fiber responses obtained from cat (A) or from computational model simulations of the cat AN (B). Blue dashed lines show rate-level functions for tones presented at the characteristic frequency while in the presence of broadband noise (Baseline). Green solid lines show rate-level functions measured identically to the Baseline condition except with the presentation of continuous contralateral acoustic stimulation (CAS, broadband noise) intended to evoke the medial olivocochlear reflex. For reference, rate-level functions measured in the absence of noise were simulated (black dotted lines in B). The spontaneous activity was subtracted from the original rate-level functions, resulting in a plot of the driven rate. Data in A are replotted from Kawase et al. (9), and simulations in B are recreated from Chintanpalli et al. (62).

The effects of background noise on AN responses to tones can be assessed by presenting a constant-level noise and measuring the firing rate as a function of tone level (60). The resulting rate-level function is systematically changed from that of tones presented in a quiet background (Fig. 2B, compare black dotted and blue dashed lines). This change is characterized by 1) an elevation in the low-level segment of the rate-level function, as the noise rather than spontaneous activity drives the fiber; 2) an increase in firing rate threshold; and 3) a decrease in the saturation rate. The latter is likely due to firing rate adaptation from the noise, which occurs as the result of synaptic events such as the depletion of neurotransmitter. In other words, the noise reduces the AN fiber response to the tone by occupying synaptic resources that would otherwise be available for tones presented in isolation. The primary effect of background noise on AN responses to tones is a reduction in the dynamic range and slope of the midlevel segment of the rate-level function (i.e., dynamic segment), thus making the AN fiber less sensitive to changes in sound pressure. This loss of sensitivity is thought to degrade the rate-based coding of tones presented in noise (9) and that of amplitude-modulated sounds (52).

Antimasking of Auditory Nerve Responses to Tones

Stimulation of the MOC reflex via electric shock or contralateral acoustic stimulation (CAS) has two main effects on the response of AN fibers to tones presented in background noise (3). First, the AN fiber becomes less sensitive to background noise, resulting in a drop in firing rate for the lower-level segment of the rate-level function, because of a decrease in cochlear amplifier gain. Second, this decrease in firing rate results in less neural adaptation induced by the background noise, resulting in an increase in firing rate for the high-level saturation segment of the rate-level function. The net effect is to restore the dynamic range of AN fibers and thereby improve the rate-based coding of transient and amplitude-modulated sounds. This restoration has been referred to as “antimasking” in studies of MOC function in laboratory animals (61). Although stimulation of the MOC reflex restores the dynamic range of AN fibers, this stimulation does not restore the firing rate threshold. In other words, the firing rate threshold for tones presented in noise remains higher than that for tones presented in quiet when the MOC reflex is stimulated (53). The higher firing rate threshold for tones in noise compared with tones in quiet is evident by a rightward shift in the rate-level function. This shift moves the dynamic segment of the rate-level function to higher input sound pressure levels, enabling the neuron to encode these higher sound pressure levels without becoming saturated (62).

Antimasking of an AN fiber in cats is illustrated in Fig. 2A, which shows rate-level functions for a tone (50 ms, 8,900 Hz) presented in a constant-level, ipsilateral broadband noise with (Fig. 2A, solid green line) or without (Fig. 2A, dashed blue line) CAS (9). Here, the broadband noise CAS was designed to elicit the MOC reflex of anesthetized or decerebrate cats while controlling for middle ear muscle activity via resection of these muscles in both ears. The authors reported that CAS was slightly stronger in decerebrate cats than anesthetized cats for some CFs, consistent with the dampening effects of anesthesia on MOC activity (see Ref. 63). Despite anesthesia, CAS appears to evoke the MOC reflex, albeit more weakly than in awake animals, resulting in significant suppression of OAEs (64) and gross neural potentials (65). Figure 2B shows computational model simulations (62) of this antimasking effect, where the MOC reflex was simulated by reducing the gain of model OHCs by 10 dB (Fig. 2B, solid green line). Simulation of the baseline rate-level function in a quiet background (Fig. 2B, dotted black line) is provided for reference. These simulations capture the major effect of ipsilateral background noise on the rate-level function (i.e., increased low-level firing rate and decreased high-level firing rate relative to the baseline rate-level function) and the antimasking effect of eliciting the MOC reflex.

Rate- and Temporal-Based Codes for Masking

The hypothesis that the MOC reflex reduces masking (3) is inconsistent with the fact that MOC stimulation does not restore firing rate threshold for AN fibers. In this context, the term “masking” is more general than the psychophysical definition. In psychophysical terms, masking refers to an increase in detection threshold of a probe by a competing sound. In terms of electrophysiological measurements (i.e., single-fiber or gross-potential responses), masking is the deterioration of rate-based coding of sound level for a probe presented in background noise, and antimasking is the improvement of such coding as the result of MOC reflex stimulation. Thus, efferent-based release from masking is expected for probes that are assumed to be detected or discriminated by a change in AN firing rate. Models of psychophysical phenomena such as pitch perception (66), loudness (67), masking (68), and frequency selectivity (69) often rely on a rate-based code, suggesting that MOC stimulation may have a broad impact on auditory perception. Despite this, direct physiological evidence of MOC antimasking of AN responses is limited to the detection or discrimination of short sounds in background noise (9, 70). The potential effects of MOC stimulation on other psychophysical tasks such as pitch discrimination, loudness judgments, and frequency selectivity have not been directly studied in AN recordings or are limited to isolated studies (e.g., Ref. 71). Moreover, invasive AN recordings are rare in humans and are limited to patients undergoing neurosurgery (e.g., Ref. 72); thus, effects of MOC stimulation on human hearing must be inferred from perceptual or noninvasive techniques, such as OAEs, as discussed in approaches to studying the moc reflex with psychophysics and associations among perceptual and otoacoustic measures of moc reflex function.

MOC stimulation may also improve temporal-based encoding of the acoustic stimulus. An example of this improvement was provided by Carney (73), who presented a conceptual framework for the temporal coding of vowels based on the profile of across-CF phase locking to the fundamental frequency (F0). This framework relies on the fact that neurons near formants phase lock to a harmonic near the formant frequency rather than to F0 (i.e., synchrony capture; Ref. 74), whereas neurons between formants phase lock strongly to F0. This across-CF profile of F0 phase locking (“fluctuation profile”) is enhanced in the IC and may serve as a neurophysiological correlate of vowel identification. Although the role of the MOC reflex in this vowel-coding hypothesis is intriguing, data are limited to support this hypothesis or other MOC-related hypotheses based on temporal coding. This article reviews the psychophysical literature that addresses the hypothesis that the MOC reflex improves rate-based coding, resulting in antimasking effects for psychophysical masking, intensity discrimination, and AM detection.

A Note about Dynamic Range Adaptation

The AN exhibits dynamic range adaptation in addition to classic firing rate adaptation (58, 59). Dynamic range adaptation for neurons of the auditory system is characterized by a rightward shift in the rate-level function so that the dynamic segment approaches the mean sound level of the acoustic environment (75). Dynamic range adaptation has been observed in neurons in the auditory periphery (58, 59) and auditory midbrain (75, 76) and in electroencephalographic (EEG) recordings sensitive to responses from the human auditory cortex (77). Several mechanisms may contribute to the dynamic range adaptation of the auditory system including the middle ear muscle reflex, the MOC reflex, and synaptic/neural processes (78). Despite this, the term “dynamic range adaptation” typically refers to the latter mechanism (synaptic/neural processes). AN responses in anesthetized animals maintained in an areflexic state (58) and computational model simulations (79) are consistent with dynamic range adaptation in the AN emerging from a synaptic/neural mechanism. Dynamic range adaptation is an important mechanism to consider when assessing the effects of the MOC reflex on psychophysical performance, because, like a reduction in cochlear gain, this mechanism results in a rightward shift in the AN rate-level function and an improvement in rate-based coding of sound intensity (9, 58, 75). Finally, recent experiments have suggested that some psychophysical phenomena attributed to the MOC reflex appear to be more consistent with dynamic range adaptation (80–82).

A SINGLE-CHANNEL FRAMEWORK FOR MOC EFFECTS IN PSYCHOPHYSICS

The Power Spectrum Model of Masking

In psychophysical masking tasks, subjects detect a probe presented with a competing sound called a masker. The purpose of the task is to estimate masking threshold, which is defined as the stimulus level required to detect the probe some percentage correct (e.g., 70.7%). The process for detecting the probe in the presence of a masker is often assumed to follow the power spectrum model of masking (68). Here, power is represented by the firing rate of an ensemble of AN fibers tuned to frequencies surrounding the probe frequency. This ensemble of fibers forms the neurophysiological basis of the “auditory filter,” which is a psychophysical construct representing the frequency resolution of hearing (83). In other words, the power spectrum model of masking is based on the power passing through a single auditory filter (84), also known as a “critical band” (85), or “equivalent rectangular bandwidth” (ERB) (86). The decision variable for the model is the probe-to-masker power ratio [i.e., signal-to-noise ratio (SNR)] at the output of the auditory filter. The central assumption of this model is that percent correct at threshold (e.g., 70.7%) corresponds to a constant effective SNR at the output of the auditory filter.

Early applications of the power spectrum model of masking included linear auditory filters (e.g., Ref. 87). These applications were successful at accounting for many aspects of simultaneous masking for detecting tones in broadband noise (88). Later implementations of the model included amplitude compression (89), which is often assumed to simulate the compression of human BM responses (90). This nonlinearity was necessary to predict empirical masking thresholds on the growth, decay, and additivity of forward masking, as well as the growth of off-frequency simultaneous masking (91), and is often not needed to predict data from listeners with hearing impairment (92). Forward masking is the elevation of probe detection thresholds by a preceding acoustic stimulus. As discussed below, the decay of forward masking has a time constant in the range of that of the MOC reflex (93). Further modifications to the power spectrum model of masking included a sliding temporal integration window at the output of the auditory filter (94), which is needed to account for backward masking and the decay of forward masking. Backward masking is the elevation in probe detection threshold by a stimulus occurring after the offset of the probe. The decision variable of these later implementations of the power spectrum model is the effective SNR after cochlear compression and temporal integration (95). Although inclusion of a temporal integration window successfully predicts the effects of masker or probe duration, or masker-probe delay, this success does not indicate that the ear integrates auditory responses. Alternatives to integration, such as neural adaptation (96), and multiple looks (97)—i.e., the expectation that the auditory system collects several samples (or looks) and statistically evaluates information from these samples—are equally successful at simulating the temporal aspects of auditory processing and may be more physiologically realistic. Moreover, these alternatives are mathematically similar to temporal integration despite being conceptually disparate (98).

The term “effective level” (or “effective SNR”) has taken on several definitions as the power spectrum model of masking has been updated and refined. Originally, the term referred to the level of a stimulus (i.e., probe or masker) after linear band-pass filtering (84, 87, 99–101). All variants of the term “effective level” express the output of a stimulus after processing by the auditory system. The common stages of processing are 1) band-pass filtering, 2) cochlear nonlinearity (compression, suppression), and 3) temporal integration/adaptation/multiple looks. “Effective level” can refer to the input or output of these stages. For example, the term has been used to describe the input to the cochlea after accounting for filtering by headphones and the middle ear (102), the output of an auditory filter after simulation of compression (103, 104) and/or suppression (67, 105), and the output of an auditory filter after simulation of cochlear nonlinearity and temporal integration (106, 107). Synonyms for effective level have emerged, such as dB/ERB, and “internal representation. ” The term “effective level” is not used in this review. Instead, a qualifier is included to indicate the output of the processing stage under consideration. For example, “postfilter level” refers to the output after band-pass filtering by the auditory filter. Similarly, “postcochlear level” refers to the output after band-pass filtering, compression, and suppression, and “postintegration level” refers to these same stages plus temporal integration. A similar naming convention is used to describe the SNRs at various processing stages (i.e., “postfilter SNR,” “postcochlear SNR,” postintegration SNR”).

Alternatives to the power spectrum model of masking exist and are able to account for a wide array of empirical data on psychophysical masking. For example, the model developed by Dau et al. (108) and refined by Dau, Jepsen, and colleagues (109, 110) accounts for simultaneous masking, forward masking, intensity discrimination, and AM detection in adults with normal and impaired hearing. Central to this model is the inclusion of a bank of modulation filters that appears after band-pass filtering, compression, and adaptation. Thus, models of this type are often referred to as “modulation filterbank” models (e.g., Ref. 111). Unlike common implementations of the power spectrum model of masking, modulation filterbank models may include a template-matching criterion rather than a criterion SNR to determine masking threshold (e.g., Ref. 112). Although some versions of the modulation filterbank model adopt assumptions of the power spectrum model (e.g., the “envelope power spectrum model,” Ref. 113), modulation filterbank models have not been evaluated in the context of the MOC reflex. This review focuses on the power spectrum model of masking, given that many psychophysical studies of the MOC reflex are based on this model (114–120). Despite the history and success of the power spectrum model, this model is challenged by several findings that cannot be explained by energy-based detection such as profile analysis (121), the effects of roving masker intensity (122–124). Such challenges to the power spectrum model are addressed elsewhere (125–127).

Estimating the Cochlear I/O Function from Masking Results

The contribution of compression to the success of the power spectrum model has led to the development of psychophysical techniques to indirectly study BM compression in humans with NH and HL (90, 106). Given that psychophysical measurements of BM compression provide an estimate of cochlear gain, recent experiments have adapted these techniques to study the MOC reflex (93, 103, 128–130). A common objective of these psychophysical techniques is to compare psychophysically derived estimates of BM compression in humans with BM measurements in laboratory animals. BM responses in animals are observed for tones presented in isolation; thus such responses do not involve two-tone suppression. Accordingly, common psychophysical techniques for estimating BM compression avoid suppression by measuring forward, rather than simultaneous, masking thresholds.

Various techniques for estimating human BM compression have developed over a 20-yr period. This development has been shaped by the introduction and evaluation of several assumptions to consider when using these techniques to study the MOC reflex. Growth of forward masking (GFM) (90) and temporal masking curves (TMCs) (106) were among the first psychophysical techniques developed to estimate BM compression and are associated with the following assumptions:

1. Thresholds are based on the power spectrum model of masking (i.e., the probe is detected at a constant postcochlear SNR at the output of an auditory filter centered on the probe frequency).

2. The response of off-frequency stimuli grows linearly through this filter, consistent with BM responses to off-CF tones. This off-frequency linearity provides a “linear reference” or “yardstick” (104).

3. BM nonlinearity is static or time invariant.

The GFM and TMC techniques differ in their approach to limit off-frequency listening (131), which occurs when the probe is detected through an auditory filter centered away from the probe frequency and violates assumption 1 above. GFM limits off-frequency listening by presenting background noise to mask the spread of probe energy to off-frequency channels, whereas the TMC technique maintains the probe at a low level to limit this spread. TMCs include the additional assumption that the decay of masking is equivalent for on- and off-frequency maskers, which may not be valid at high sound levels (119). The additivity of forward masking (AFM) (104) technique was developed after assumption 2 (off-frequency linearity) was evaluated and found to be invalid for low-frequency probes (132, 133). The AFM technique avoids this assumption but adopts the additional assumption that the postcochlear levels of two or more maskers add linearly. This assumption of additivity may not hold when these maskers differ in frequency (129, 134, 135).

A variant of the GFM technique involving short maskers (e.g., 20–40 ms; Refs. 103, 129) and the fixed-duration masking curve (FDMC) technique (120, 130) were developed as assumption 3 (static BM nonlinearity) may be valid only for a brief period (e.g., 25 ms) following acoustic stimulation. After this period, responses of the peripheral auditory system are expected to adjust as a result of several dynamic processes such as neural adaptation, middle ear muscle activity, MOC activity, and dynamic range adaptation (78). TMCs violate assumption 3 (static BM nonlinearity) given that masker durations are typically longer than 25 ms, and the masker-probe delay is often as great as 40–80 ms. This violation limits the utility of the TMC technique for studying the MOC reflex, as a baseline condition without potential MOC effects is difficult to establish. A workaround is to measure TMCs in the presence and absence of contralateral noise (128), as discussed in Effects of Precursors/Contralateral Stimulation on Psychophysical Estimates of Cochlear Gain.

A Framework for MOC Effects in Psychophysics

The putative relationship between psychophysical masking and cochlear nonlinearity has led to the hypothesis that masking thresholds may reveal a reduction in cochlear gain from the MOC reflex (136). A common study design for assessing MOC effects is to compare masking thresholds for conditions where MOC activity is expected to be relatively weaker (i.e., baseline condition) versus relatively stronger. For example, masking thresholds may be compared with and without a sound presented before the masker, called a “precursor.” Here the precursor is expected to elicit the ipsilateral MOC reflex, resulting in stronger MOC activity compared with the baseline condition without a precursor. For the baseline condition, detection thresholds are measured for a short probe (e.g., 5 ms) and masker (e.g., 20 ms) where the combined duration of the probe and masker is on the order of the onset latency of the MOC reflex (∼25 ms; Ref. 137). This short combined duration ensures that detection of the probe occurs before the reflex reaches full strength. In the precursor condition, a relatively longer precursor (e.g., 200 ms) precedes the short masker and probe and serves to elicit the MOC reflex so that the probe is detected when ipsilateral MOC reflex strength is greater than that of the baseline condition. An alternative design is to compare masking thresholds with and without a simultaneously presented contralateral noise. The contralateral noise is assumed to elicit the contralateral MOC reflex; thus comparison of detection thresholds with versus without contralateral noise is assumed to reveal the effects of the contralateral reflex.

The power spectrum model predicts that the effect of the MOC reflex on masking thresholds depends strongly on the postfilter SNR at threshold and the I/O function slopes for the probe and masker. This dependence is illustrated in Fig. 3 for on- and off-frequency tonal maskers. The framework that interprets masking thresholds under the assumption of the power spectrum model and the expected effects of MOC feedback has been called the “Guinan/Strickland interpretation” (138). This name acknowledges the synthesis of a wealth of physiological data in terms of antimasking by Guinan (3, 4, 139) and the modification of the power spectrum model of masking to include the effects of the MOC reflex by Strickland (103, 117, 140). Although Strickland quantitatively tested this framework, von Klitzing and Kohlrausch (141) provided the original description of the framework. Here the framework is referred to as the “MOC/power spectrum model. ”

Figure 3.

Illustration of the medial olivocochlear (MOC)/power spectrum model and the effect of signal-to-noise ratio (SNR, rows) and off-frequency stimulation (columns) on model predictions. Dashed blue lines and solid green lines represent basilar membrane (BM) input/output (I/O) functions to tones presented at the characteristic frequency (CF) before and after stimulation of the MOC reflex. Black dotted lines show responses to a tone presented well below CF. Stimulation of the MOC reflex is assumed to result in a reduction in gain for tones presented at CF and no change for tones presented off-frequency. Markers labeled P (probe) and M (masker) represent coordinates of the postfilter input (x-axis) and postcochlear output (y-axis), and the vertical arrows represent the probe-to-masker SNR. The probe-to-masker SNR is compared for conditions with (solid green lines) and without (dashed blue lines) stimulation of the MOC reflex. For on-frequency probes and maskers (left), predicted probe-to-masker SNR increases, remains constant, or decreases with MOC stimulation when the Baseline SNR is positive (top), zero (middle), and negative (bottom), respectively. In contrast, for off-frequency maskers (right) probe-to-masker SNR is predicted to decrease with MOC stimulation regardless of the Baseline SNR.

For on-frequency maskers (Fig. 3, left), a reduction in cochlear gain is expected to improve, worsen, or have no effect on masking thresholds for large positive (e.g., >5–10 dB; Fig. 3, top), large negative (e.g., <5–10 dB, Fig. 3, bottom), or near-0 dB (Fig. 3, middle) postfilter SNRs, respectively. The acoustic SNR (input) and postcochlear SNR (output) share the same direction (i.e., positive or negative) and differ only by the effect of cochlear filtering and compression. For large positive postfilter SNRs, the moderate-level probe produces a higher postfilter level than the lower-level, on-frequency masker (Fig. 3, top left, dashed lines). When the MOC reflex is elicited, cochlear gain is decreased, resulting in a greater reduction in the masker’s, compared to the probe’s, postcochlear level (Fig. 3, top left, solid lines). This decrease improves the postcochlear SNR, resulting in suprathreshold detectability performance (e.g., 95% correct; Fig. 3, top left, arrows). To obtain threshold detectability performance (e.g., 70.7% correct), the experimenter must increase the level of the masker or decrease the level of the probe. For large negative postfilter SNRs, the moderate-level, on-frequency masker produces a larger postfilter level than the lower-level probe (Fig. 3, bottom left, dashed lines). A reduction in gain decreases the probe’s postcochlear level more than that of the masker, resulting in a poorer postcochlear SNR (Fig. 3, bottom left, solid lines) and thus requires the masker level to decrease or the probe level to increase to reach threshold. For postfilter SNRs near 0 dB, the on-frequency masker and probe produce roughly the same postfilter level. A reduction in gain decreases the postcochlear level of the probe and masker equally, resulting in no change in the postcochlear SNR and therefore no change in masking threshold (Fig. 3, middle left).

For off-frequency maskers (Fig. 3, right), cochlear amplification is assumed to be applied to the probe but not to the masker for the auditory filter centered on the probe frequency. Thus, a reduction in cochlear gain decreases the postcochlear level of the probe, while having no effect on the masker. This reduction in gain results in a poorer postcochlear SNR relative to the baseline condition, regardless of the postfilter SNR, and requires the probe level to increase or the masker level to decrease to obtain threshold.

The framework for on-frequency maskers and probes presented at SNRs > 0 dB (Fig. 3, left, top) has been extended to intensity discrimination (142, 143) and AM detection (82, 144–146) tasks. For intensity discrimination, the postfilter SNR is defined by the intensity of the pedestal (I) plus that of the increment (ΔI) divided by the intensity of the pedestal (e.g., [I + ΔI]/I) at threshold. Similarly, for AM detection, the postfilter SNR is determined by the ratio of AM peaks to AM valleys (i.e., peak-to-valley ratio) at threshold. The MOC/power spectrum model predicts little to no MOC effect when intensity discrimination thresholds are low (i.e., small increment or peak-to-valley ratio at threshold) and larger MOC effects when thresholds are poor (i.e., large increment or peak-to-valley ratio at threshold). Indeed, Wojtczak et al. (82) found that improvements in AM detection in the presence of a precursor (i.e., “AM unmasking”) were not consistently observed unless AM thresholds in the baseline condition were poor (i.e., large peak-to-valley ratio). Thus, their experimental design adjusted the level of a simultaneous masker to ensure AM detection thresholds, expressed as 20 × log(m), were high (between −6 and −2 dB).

The MOC/power spectrum model provides a framework for reviewing data from the masking literature in terms of MOC physiology. The following review includes data from human subjects unless otherwise mentioned. The MOC/power spectrum model is based on auditory nerve responses at the output of a single CF measured in laboratory animals. A central assumption of the model is that human auditory nerve responses will be qualitatively similar to those from laboratory animals. This assumption remains unvalidated, as single-unit recordings in humans are not available; however, several studies have shown similarities between gross potentials, such as the compound action potential, measured from humans and laboratory animals (e.g., Ref. 147). Although these similarities support the assumptions of the MOC/power spectrum model, interspecies differences in gross potentials remain (e.g., Ref. 148) and should be considered when interpreting results in terms of this model.

APPROACHES TO STUDYING THE MOC REFLEX WITH PSYCHOPHYSICS

Temporal Effects in Simultaneous Masking

A simultaneous masking phenomenon called “the temporal effect” or “overshoot” is among the first psychophysical results to be interpreted in terms of the MOC reflex (136). The temporal effect refers to an elevation in detection threshold for a short tonal probe (e.g., <20 ms) presented near the onset (onset condition) of a simultaneous masker compared to the same probe presented some 100 ms or so later (steady-state condition; Ref. 149). The elevation in threshold for probes presented with short, gated maskers compared to continuous maskers is also considered a temporal effect (e.g., Ref. 150). Zwicker (151) used the term “overshoot” to describe the elevation in threshold for probes presented at the onset of the masker. In this review, the terms “temporal effect” and “overshoot” are used synonymously to refer to an elevation in probe detection in noise for probes preceded by silence compared to those preceded by acoustic stimulation (i.e., a precursor). The precursor may be a separate stimulus or simply the forward fringe of a continuous or gated masker. This section considers temporal effects in simultaneous masking; however, similar effects have been reported for forward masking (e.g., Ref. 103), and AM detection (e.g., Ref. 144). Other phenomena, such as masker enhancement (152) or vowel enhancement (153), may be considered temporal effects; however, such phenomena are discussed elsewhere (e.g., Ref. 154).

Temporal effects in simultaneous masking measured from adults with NH and HL (10) are illustrated in Fig. 4. Masker level at threshold was measured for short probes (6 ms) at two probe frequencies (2,000 Hz, 4,000 Hz); Fig. 4 shows the data for the 4,000-Hz probe. The probe was delayed by 2 ms (onset condition; Fig. 4, dashed blue lines) or by 198 ms (steady-state condition; Fig. 4, solid green lines) from the onset of a 400-ms broadband masker. Masker level at threshold was measured for five probe levels (50, 60, 70, 80, 90 dB SPL) in adults with NH and four probe levels (70, 80, 90, 100 dB SPL) in adults with cochlear HL. For this experiment, greater sensitivity to the probe is indicated by a higher masker level at threshold. In other words, subjects with the greatest sensitivity can tolerate a higher masker level at threshold compared with less sensitive subjects. A clear temporal effect is observed for subjects with NH, as indicated by higher masker levels at threshold in the steady-state condition (198-ms delay) compared with the onset condition (2-ms delay; Fig. 4A, squares). The magnitude of this temporal effect grows with probe level and slightly decreases at the highest level (Fig. 4C, squares). This nonmonotonic relationship between the temporal effect and stimulus level has been clearly documented in previous studies that included higher probe/masker levels (155) than those tested by Jennings et al. (10) and may be related to a steepening of the BM I/O function slope at high levels (140). Subjects with cochlear HL have smaller temporal effects than subjects with NH, as demonstrated in Fig. 4C (triangles) and previous studies on temporary (156) and permanent (157, 158) cochlear HL.

Figure 4.

Average temporal effects in simultaneous masking measured from adults with normal hearing (NH, 17 subjects) and cochlear hearing loss (HL, 15 subjects). A: growth of simultaneous masking for a short (6 ms), 4,000-Hz tone presented 2 ms (dashed lines) or 198 ms (solid lines) from the onset of a broadband noise masker. Squares and triangles show data from adults with NH and HL, respectively. B: the postfilter signal-to-noise ratio (SNR) at threshold as a function of probe level for probes presented 2 ms and 198 ms from the masker’s onset in adults with NH and HL. C: the temporal effect as a function of probe level for adults with NH and HL, where the temporal effect is defined as the difference between masker level at thresholds for probes presented 198 ms and 2 ms from the masker’s onset. Data are replotted from Jennings et al. (10).

Given the short duration of the probe used to study temporal effects, the postfilter SNRs tend to be much larger than for masking studies with longer probes. Figure 4B shows the postfilter SNRs from Jennings et al. (10), which were computed from the masker level passing through an ERB centered on 4 kHz (456 Hz; Ref. 69). In the onset condition (Fig. 4B, dashed blue lines), these SNRs range from 22 to 31 dB and from 16 to 23 dB in subjects with NH and cochlear HL, respectively. Postfilter SNRs in the steady-state condition (Fig. 4B, solid lines) range from 17 to 20 dB in subjects with NH and from 15 to 16 dB in subjects with cochlear HL. For comparison, postfilter SNRs calculated from Reed and Bilger (Ref. 159; their Fig. 4), who measured masking thresholds for longer (300 ms) probes in broadband continuous noise, were between −6 and −4 dB when the probe was 4,000 Hz. The negative postfilter SNRs from Reed and Bilger (159) are consistent with similar studies on simultaneous masking with long probes presented in broadband noise (e.g., Ref. 87). Small or negative postfilter SNRs at threshold for long versus short tonal probes in noise are likely responsible for the absence of the temporal effect for probes longer than 10 ms (151).

Temporal effects in simultaneous masking were originally interpreted to emerge from AN firing rate adaptation. This interpretation is based on the findings of Smith and Zwislocki (160), who measured AN responses to short intensity increments positioned within a tonal pedestal. They found that the firing rate to an increment was independent of the location of the increment within the longer pedestal. This constant response to the increment occurred despite firing rate adaptation to the pedestal; thus, the effective increment-to-pedestal SNR improves as the AN adapts. This improvement predicts 3- to 5-dB better probe thresholds in the steady-state condition compared with the onset condition, which is qualitatively consistent with temporal effects in simultaneous masking. As discussed by Bacon and Healy (161), firing rate adaptation cannot explain all aspects of the temporal effect in simultaneous masking. For example, adaptation does not account for the magnitude of these temporal effects, which can exceed 20 dB, or their dependence on masker spectrum and level (162). This dependence is characterized by larger temporal effects for off- versus on-frequency maskers and the nonmonotonic growth of the temporal effect with increasing broadband masker level. Furthermore, firing rate adaptation is inconsistent with the finding that temporal effects in simultaneous masking can be evoked by contralateral precursors (161, 163, 164). More recently, the hypothesis that firing rate adaptation accounts for temporal effects in simultaneous masking was quantitatively evaluated with a computational model of the auditory periphery (115). These model simulations revealed that classic firing rate adaptation and dynamic range adaptation could not predict temporal effects in simultaneous masking for listeners with NH or cochlear HL; however, such effects were predicted when simulations included the influence of the MOC reflex.

Temporal effects in simultaneous masking appear to be closely related to the status of the cochlear amplifier, as illustrated by comparing temporal effects from subjects with NH and cochlear HL (Fig. 4). Compared to subjects with NH, subjects with cochlear HL have better detection thresholds for the onset condition and similar thresholds for the steady-state condition. In other words, temporal effects are small or absent in these subjects because hearing loss improves their sensitivity to probes presented near the masker’s onset. This better-than-normal sensitivity is consistent with the effects of cochlear HL on the BM I/O function (165) and the finding that temporal effects are associated with large positive postfilter SNRs at threshold (Fig. 4B). The degree of compression applied to the probe and masker determines the SNR at the output of the cochlea (i.e., postcochlear SNR). For a moderate- to high-level probe and a large postfilter SNR (e.g., 25 dB) the postcochlear SNR will be appreciably reduced in subjects with NH. For example, a 5:1 compression ratio will reduce the 25-dB postfilter SNR to a 5-dB postcochlear SNR. In contrast, the BM I/O function growth at these moderate to high levels approaches linearity for subjects with cochlear HL. Thus, a postfilter SNR of 25 dB is not appreciably influenced by compression, resulting in a favorable postcochlear SNR and better detection thresholds than those for subjects with NH. Further support for this interpretation comes from studies on temporary hearing loss, where sensitivity of subjects with NH temporarily improves in the onset condition when cochlear amplifier gain is compromised by aspirin ingestion (166) or noise exposure (156).

von Klitzing and Kohlrausch (141) suggested that the temporal effect is mediated by an increase in postcochlear SNR during the forward fringe of the masker. They proposed that the MOC reflex might facilitate this increase in SNR. Under the assumptions of the MOC/power spectrum model, the acoustic stimulation immediately preceding the probe determines the status of cochlear amplifier gain. When silence (>350–500 ms) precedes the probe (onset condition), the auditory periphery recovers from residual AN adaptation (e.g., Ref. 167, in chinchilla) and MOC feedback (e.g., Ref. 137, in human). Thus, cochlear amplifier gain is high, and the BM I/O function is compressive. When a precursor precedes the probe (steady-state condition), the MOC reflex is elicited, cochlear gain is reduced, and the BM I/O function approaches linearity, thus improving postcochlear SNR. This approach to linear response growth results in similar postfilter SNRs among subjects with NH and cochlear HL (Fig. 4B, compare filled squares and triangles) and accounts for the roughly equal detection thresholds in the steady-state condition among subjects with NH and HL (Fig. 4A, compare filled squares and triangles).

In addition to accounting for the effects of cochlear HL, the MOC/power spectrum model accounts for the dependence of temporal effects on masker level (117) and contralateral precursors (161, 163, 164) and may be adapted to account for the effects of masker spectrum (114, 140). The success is largely due to the model’s prediction that temporal effects will be largest when the postfilter SNR is positive, as discussed in A Framework for MOC Effects in Psychophysics. These findings support the notion that temporal effects in simultaneous masking are consistent with the MOC reflex and that other dynamic processes, such as AN adaptation and dynamic range adaptation, play a secondary role (115).

Some aspects of the temporal effect in simultaneous masking are not well understood and cannot be explained by a reduction in cochlear gain. For example, narrowband noise (<1 ERB) maskers and precursors result in negative temporal effects, where detection thresholds are poorer in the steady-state condition compared with the onset condition (125, 162, 168–170). Jennings et al. (170) showed that such maskers and precursors produce a modulation-masking effect (171). In other words, the inherent fluctuations of the narrowband noise masker mask the probe when the average period of these fluctuations is similar to the duration of the probe. Such an effect likely occurs central to the cochlea and obscures an interpretation of temporal effect data based on the MOC reflex. Finally, mixed results have been reported on the relationship between temporal effects in simultaneous masking and suppression of OAEs. Keefe et al. (116) reported no relationship, whereas Walsh et al. (138) reported similar temporal effects measured with OAEs and perception. These results are discussed further in associations among perceptual and otoacoustic measures of moc reflex function.

In summary, temporal effects in simultaneous masking reveal that sensitivity to short tones improves with preceding acoustic stimulation (i.e., precursor) and that this improvement is larger in adults with NH than adults with cochlear HL. The dependence of the temporal effect on temporary/permanent HL, level, contralateral stimulation, and masker spectrum is consistent with the MOC/power spectrum model and suggests that the MOC reflex may play a primary role in temporal effects in simultaneous masking. Moreover, model predictions show that simulation of the MOC reflex is essential to predicting the temporal effect in adults with NH and HL, whereas other mechanisms (classic firing rate and dynamic range adaptation) may play a secondary role. Despite these findings, there are aspects of the temporal effect that are inconsistent with the MOC reflex, including potential effects of modulation masking and mixed findings from OAE studies.

Effects of Precursors/Contralateral Stimulation on Psychophysical Estimates of Cochlear Gain

Several studies have tested the hypothesis that eliciting the MOC reflex with sound will decrease psychophysical estimates of cochlear gain. In such studies, the BM I/O function is estimated with GFM, TMC, or FMDC techniques and comparisons are made between thresholds obtained with and without a precursor (103, 129, 130) or with and without contralateral noise (128). In a pioneering study, Krull and Strickland (103) used GFM to determine the effects of a precursor on psychophysical estimates of the BM I/O function in three adults with NH. The additional forward masking introduced by the precursor was used as a proxy for studying a reduction in cochlear gain from MOC feedback. The probe and precursor were 4,000-Hz tones, and the off-frequency masker was a 2,800-Hz tone. Two precursor durations (40, 160 ms) and levels (50, 60 dB SPL) were tested. The combined duration of the masker and probe was limited to 46 ms (40-ms masker, 6-ms probe), which is slightly longer than the onset delay of the MOC reflex (25 ms). Thus, MOC feedback was assumed to have greater effect on the estimated BM I/O function measured with a precursor than that measured without a precursor. Off-frequency GFM without a precursor approached linearity (1 dB/dB) for lower probe levels and was shallow (<1 dB/dB) at moderate to high levels. Estimated BM I/O functions were obtained by fitting off-frequency GFM data with a piecewise linear function that quantified gain, compression, and the intersection (i.e., breakpoint) between linear/compressive growth. When the precursor was absent, compression and the breakpoint varied among subjects from 0.27 to 0.39 dB/dB and from 44 to 66 dB SPL, respectively. GFM measured with a precursor consistently reduced the gain by 9–20 dB and increased the breakpoint by a similar magnitude. For two of the three subjects, the compression slope increased in the precursor conditions. These precursor effects were larger for the 60-dB SPL than 50-dB SPL precursor, whereas the effect of precursor duration (40 or 160 ms) was negligible. This reduction in psychophysically derived gain from the presentation of a precursor is consistent with a reduction in cochlear gain observed when the MOC reflex is evoked by electric shock in laboratory animals (e.g., Ref. 45, in guinea pig). Moreover, larger reductions in gain with higher precursor levels are consistent with the finding that AN responses decrease when eliciting the MOC reflex with CAS and this effect increases with increasing level of CAS (Ref. 172, in cat). The MOC/power spectrum model predicts that a reduction in cochlear gain will decrease the postcochlear level of the probe but not that of the off-frequency masker. Thus, the masker level at threshold is predicted to decrease when a precursor is present. Under the assumption that the response growth of the off-frequency masker approaches linearity through the auditory filter centered on the probe frequency, this decrease in masker level equals the decrease in gain applied to the probe.

The results of Krull and Strickland (103) have been replicated and expanded in several studies to show that ipsilateral precursors decrease psychophysical estimates of cochlear gain. These studies are based on the premise that the precursor decreases the cochlear gain applied to the probe and does not affect the postcochlear level of the off-frequency masker (i.e., Fig. 3, right). These studies included parametric manipulations of precursor level (173), frequency (174), duration (93), laterality (118), and the silent interval between the precursor and masker (i.e., precursor-masker delay; Ref. 130). For example, Yasin et al. (130) measured FDMCs to determine the effects of precursor level and precursor-masker delay on estimated BM I/O functions. The probe was a 4,000-Hz tone presented at 10 dB SL and was preceded by on-frequency (4,000 Hz) or off-frequency (1,800 Hz) maskers. The combined duration of the masker and probe was restricted to 25 ms to limit the effects of MOC stimulation by the masker and probe in the baseline (i.e., no precursor) condition. On- and off-frequency FDMCs were constructed by measuring masker level at threshold for several masker/probe durations. Gain was defined as the difference between on- and off-frequency masker levels at threshold for the shortest duration signal (6 ms). BM I/O functions were derived by plotting the FDMC for the off-frequency masker as a function of that for the on-frequency masker. These I/O functions were fit with a polynomial model to estimate the average compression slope between input levels of 50 to 70 dB SPL. Independent variables were precursor level (20, 40, 60, or 80 dB SPL) and the length of the silent interval between the precursor and masker (0, 50, 100, or 200 ms). The authors hypothesized that the effects of the precursor would increase with increasing precursor level and decrease with increasing precursor-masker delay, consistent with the level dependence and time course of the MOC reflex. For the smallest precursor-masker delay, average estimates of gain decreased by 20 dB as the precursor level increased from 20 to 80 dB SPL. This decrease in gain was accompanied by an increase in the compression slope from 0.45 dB/dB to 0.85 dB/dB. The magnitude of this decrease in gain and increase in compression slope was smaller for longer precursor-masker delays, consistent with recovery from residual MOC feedback during the silent interval between the precursor and masker. The effect of an ipsilateral precursor on average psychophysical estimates of the BM I/O function from NH adults (130) is shown in Fig. 5A. Here, the baseline I/O function without a precursor (Fig. 5A, dashed blue line) is compared with an I/O function with a precursor (Fig. 5A, solid green line), where the precursor level was 60 dB SPL and the precursor-masker delay was 0 ms. Model fits to these I/O functions show that the precursor reduced estimated cochlear gain by 22.9 dB and increased the compression slope from 0.21 dB/dB to 0.61 dB/dB.

Figure 5.

Presentation of an ipsilateral precursor or contralateral acoustic stimulation (CAS) decreases psychophysical estimates of cochlear gain in adults with normal hearing (NH). A: basilar membrane (BM) input/output (I/O) functions for tones presented at the characteristic frequency (CF) derived from fixed-duration masking curves (FDMCs) under conditions with (solid green line) and without (dashed blue line) an ipsilateral precursor and averaged from 6 adults with NH. FDMCs were measured for a 4,000-Hz probe in the presence of a 1,800-Hz or 4,000-Hz masker. The precursor was a 60-dB SPL, 500-ms narrowband noise centered on the probe frequency and presented immediately prior to the masker’s onset (0-ms precursor-masker delay). B: BM I/O functions derived from temporal masking curves (TMCs) in the presence (solid green line) or absence (dashed blue line) of CAS. TMCs were measured for a 2,000-Hz, 10-ms probe presented with 1,220-Hz or 2000-Hz maskers and averaged from 12 adults with NH. For A and B, psychophysical estimates of cochlear gain decrease in the presence of the precursor or CAS relative to baseline. C: a reduction in estimates of cochlear gain (G) measured from growth of forward masking as a function of precursor duration for on (filled green squares)- and off (open green squares)-frequency precursors and averaged from 5 adults with NH. Estimates of gain were obtained for a 6-ms, 4,000-Hz probe presented with a 20-ms, 2,400-Hz forward masker. The on-frequency (4,000 Hz) and off-frequency (2,400 Hz) precursors were presented immediately prior to the masker’s onset and were presented at 60 and 95 dB SPL, respectively. Gain decreases nonmonotonically with precursor duration for on-frequency precursors, consistent with a reduction in the postcochlear output for the probe and on-frequency masker. Data in A, B, and C are from Yasin et al. (130), Fletcher et al. (128), and Roverud and Strickland (176), respectively.

Fletcher et al. (128) estimated I/O functions by using TMCs measured in the presence or absence of contralateral noise to determine the extent to which eliciting the contralateral MOC reflex influenced psychophysical estimates of cochlear compression and gain. The probe was a 10-ms, 2,000-Hz tone presented 0, 5, 10, 15, 20, or 25 ms after a short (30 ms) on- or off-frequency masker. Masker level at threshold was measured for a 10-dB SL probe in the presence or absence of a continuous 54-dB SPL broadband contralateral noise. The primary effect of the noise was to increase the effectiveness of on- and off-frequency maskers, resulting in a decrease in masker level at threshold compared with the baseline TMC. For on-frequency maskers, this decrease was largest when the postfilter SNR was negative (masker-probe delays > 15 ms) and absent when the postfilter SNR was near zero (masker-probe delays < 15 ms). This finding is consistent with the MOC/power spectrum model, as a reduction in gain for negative postfilter SNRs is expected to decrease the postcochlear level of the probe more than that of the masker. TMCs were fit with a model that quantified gain, compression, BM passive processing, decay of masker effectiveness, and detection efficiency. Although results were marked by individual differences, gain was significantly reduced for TMCs measured with contralateral noise compared with baseline TMCs (Fig. 5B, compare dashed blue and solid green lines). This finding is consistent with contralateral noise eliciting the MOC reflex and resulting in relatively stronger MOC effects than in the baseline condition.

Roverud and Strickland (93) developed an abbreviated method for measuring the effect of a precursor on psychophysical estimates of cochlear gain and tested this method in six adults with NH. This method was used to test whether the effects of precursor duration and precursor-masker delay on masking thresholds were consistent with the time course of the MOC reflex. Similar to Krull and Strickland (103), they measured the additional forward masking produced by an on-frequency precursor when presented with a short (20 ms) off-frequency masker. Rather than derive the entire BM I/O function, they identified a single point that fell on the lower-level linear portion of this function. This point varied among subjects and corresponded to probe and off-frequency masker levels ranging from 20 to 40 dB SPL and from 65 to 80 dB SPL, respectively. For the precursor conditions, the authors held the masker constant at the level associated with this single point and measured the probe threshold shift induced by the precursor. The authors used the term “temporal effect” to refer to this shift and interpreted the shift as a reduction in cochlear gain. This temporal effect in forward masking was measured for several precursor durations and precursor-masker delays to test the hypothesis that the magnitude of the forward-masking temporal effect would follow the time course of the MOC reflex. Traditional models of forward masking predict a monotonic decay in masking with increasing delay between the precursor and probe (96). For long precursors (50 ms, 100 ms), this monotonic decay was observed; however, for short precursors (10 ms, 20 ms), thresholds worsened (larger forward-masking temporal effect) and then improved with increasing precursor delay. A model that incorporated the time course of the MOC reflex, as measured from OAEs in humans (175), successfully accounted for this nonmonotonic decay of forward masking. The success of the model originates in the expectation that a short precursor will elicit the MOC reflex and that MOC strength will increase after the offset of the precursor before decaying thereafter. If the probe is presented during this putative increase in MOC strength, detection thresholds will worsen rather than improve.

In a later study, Roverud and Strickland (176) extended their findings to include the effects of precursor duration for on- and off-frequency precursors in five adults with NH. The precursor-masker delay was held constant at 0 ms, and thresholds were measured for a series of precursor durations from 10 to 150 ms. They assumed that 1) MOC reflex strength would build over the time course of the precursor, resulting in a larger reduction in gain for longer precursors than for shorter precursors; 2) a reduction in gain elicited by the precursor would, in fact, decrease the postcochlear level of the precursor itself, for on-frequency but not off-frequency precursors; and 3) this decrease in postcochlear level would be greater for longer than for shorter on-frequency precursors. Under these assumptions, they expected a monotonic increase in probe threshold with increasing off-frequency precursor duration. For on-frequency precursors, they expected an increase and then decrease in probe thresholds due to a reduction in the postcochlear level of the precursor for longer, but not shorter, precursor durations. In other words, a reduction in the postcochlear level of the precursor was expected to release the probe from masking and result in lower detection thresholds for longer precursor durations. Consistent with their assumptions, Roverud and Strickland (176) reported monotonic and nonmonotonic effects of precursor duration for off- and on-frequency precursors, respectively (Fig. 5C). Simulations revealed that such nonmonotonicities were predicted by a forward masking model that included MOC-related gain reduction compared with a traditional model of forward masking.

Forward masking from ipsilateral precursors can be interpreted as evidence for the additivity of forward masking rather than a reduction in cochlear gain via the MOC reflex (104, 177, 178). Under the additivity interpretation, the BM I/O function is static regardless of previous acoustic stimulation, and the postcochlear levels of the masker and precursor add linearly. Accordingly, detection thresholds for the probe increase in the presence of the combined masker and precursor compared with thresholds for the masker or precursor in isolation. This increase amounts to 3 dB when the precursor and masker are equally effective, and the BM growth approaches linearity. In contrast, this increase can be substantially greater (e.g., 10 dB) when the probe is subject to BM compression. Jennings et al. (129) showed that data previously interpreted in terms of additivity of masking (178) could be accounted for by a reduction in cochlear gain. Moreover, predicted precursor effects based on the MOC/power spectrum model were more consistent with data from human listeners than those based on additivity of masking for off-frequency maskers (129, 134, 135). Despite this, a reduction in gain via the MOC reflex is inconsistent with the finding that psychophysical estimates of cochlear gain from AFM do not decrease with increasing precursor duration (177). This led Plack and Arifianto (177) to conclude that the effect of a precursor on masking thresholds is consistent with a combination of additivity of masking and gain reduction mechanisms.

In summary, the presence of an ipsilateral precursor or contralateral noise reduces psychophysical estimates of cochlear gain and compression in a manner consistent with the time course of the MOC reflex. Although the effect of the precursor is consistent with traditional models of forward masking (including additivity of forward masking) in some respects, the effects of masker frequency, precursor duration, and precursor-masker delay are inconsistent with such models unless simulations are modified to include a reduction in cochlear gain.

Effects of Precursors/Contralateral Stimulation on Psychophysical Estimates of Frequency Resolution

Jennings and Strickland (135) expanded on a preliminary experiment (129) to study the effects of an on-frequency precursor on psychophysical frequency resolution for several probe levels using psychophysical tuning curves (PTCs) and the notched-noise method (i.e., “notched-noise tuning curves,” NNTCs). The 6-ms, 4,000-Hz probe was presented after a 20-ms masker, which was preceded by silence or by a 4,000-Hz, 100-ms precursor. The authors reported a significant broadening of frequency selectivity for PTCs and NNTCs measured with a precursor compared to without a precursor. This is illustrated in Fig. 6 by comparing PTCs (Fig. 6B) and auditory filter shapes derived from NNTC data (Fig. 6A) obtained with (solid green lines) and without (dashed blue lines) the precursor. This broadening of tuning with a precursor was qualitatively predicted by a model that simulated the effects of the MOC reflex, whereas no such broadening was predicted by a traditional model of forward masking. The precursor resulted in lower masker levels at threshold, and this precursor effect was larger for off- than for on-frequency maskers, thus explaining the broadened auditory filter shapes. These results are consistent with the MOC/power spectrum model, which predicts that a reduction in cochlear gain will influence the probe and on-frequency masker equally, resulting in little change in the postcochlear SNR, whereas the off-frequency masker is not predicted to be influenced by this reduction in gain, thus resulting in a poorer postcochlear SNR for the precursor compared with the no-precursor condition.

Figure 6.

Presentation of an ipsilateral precursor or contralateral acoustic stimulation (CAS) broadens estimates of frequency resolution for high-frequency probes (e.g., 4,000 Hz; A, B, D) and sharpens these estimates for low-frequency probes (e.g., 500 Hz; C) in adults with normal hearing. Auditory filter estimates from notched-noise tuning characteristics (3 subjects; A) and data from psychophysical tuning curves (PTCs) (2 subjects; B) were obtained from forward masking of a short (6 ms) 40-dB SPL, 4,000-Hz probe with (solid green lines) and without (dashed blue lines) an on-frequency precursor. Similar PTCs for an example subject obtained in simultaneous masking with and without broadband CAS are shown in C and D for 500-Hz and 4,000-Hz probes, respectively. Data from A and B are from Jennings and Strickland (135). Data from C and D are from Vinay and Moore (182).

Several studies have shown that contralateral noise alters PTCs in a manner consistent with MOC feedback (179–183). Such studies often quantify changes in tuning by comparing PTCs obtained without contralateral stimulation (baseline PTC) to those obtained in the presence of continuous or gated contralateral noise. For higher-frequency probes (>1,000 Hz; Fig. 6D) contralateral noise decreases masker level at threshold, compared with the baseline PTC, for off-frequency maskers but not for on-frequency maskers in forward (180) and simultaneous (180–182) masking. In other words, contralateral noise increases the effectiveness of off-frequency maskers. The overall effect of contralateral noise is to broaden the PTC bandwidth for these higher-frequency probes, which is consistent with the reduction in the sensitivity of the tip of higher-CF AN tuning curves with MOC bundle stimulation in cats (47).

For lower-frequency probes (<1,000 Hz) contralateral noise results in sharper or broader PTCs than the baseline PTC for simultaneous (Fig. 6C; Ref. 182) and forward (179) masking, respectively. In simultaneous masking, the increased sharpness is due to an increase in masker level at threshold for off-frequency maskers. In other words, contralateral noise results in less effective off-frequency simultaneous maskers. Aguilar et al. (179) attributed the discrepancy in the effect of contralateral noise on PTCs in simultaneous and forward masking to the influence of cochlear suppression. They reasoned that stimulation of the MOC reflex may reduce the effects of suppression on PTCs measured with simultaneous masking. Specifically, such a reduction in suppression may release the probe from suppressive masking from off-frequency maskers, resulting in higher masker levels at threshold and sharper PTCs.

In summary, on-frequency ipsilateral precursors or contralateral noise influences psychophysical estimates of frequency resolution in a manner consistent with eliciting the MOC reflex. This influence includes the expected decrease and increase in frequency resolution for higher and lower probe frequencies, respectively, and interactions with the effects of cochlear suppression. Consistent with the MOC/power spectrum model, the effects of an ipsilateral precursor or contralateral noise on estimates of frequency resolution are greatest for off-frequency maskers and negligible for on-frequency maskers. Specifically, a reduction in cochlear gain decreases the postcochlear level of the probe and on-frequency masker equally, resulting in little change in masking thresholds at the tip of the tuning curve. Conversely, relatively larger changes in threshold are observed for the tail of the tuning curve, where the MOC reflex is expected to decrease the postcochlear level of the probe but not that of the off-frequency masker.

Effects of Precursors/Contralateral Stimulation on Psychophysical Estimates of Intensity Resolution

Intensity discrimination of short (e.g., <50 ms) tones presented after a forward masker is poor at moderate levels and improves at higher and lower levels, resulting in the so-called “midlevel hump” (184). A similar deterioration of intensity discrimination at moderate levels is observed for short probes presented with gated maskers, which is similar to the onset condition of the temporal effect in simultaneous masking and has been referred to as the “severe departure from Weber’s law” (185, 186). Poor intensity discrimination at moderate levels for conditions associated with the midlevel hump and the severe departure from Weber’s law may be a product of cochlear compression (187), which is hypothesized to limit the increase in postcochlear level of the increment at moderate, but not low, levels. Roverud and Strickland (143) tested this hypothesis by measuring intensity discrimination thresholds under conditions associated with the midlevel hump and compared these thresholds with BM I/O functions derived from GFM. Intensity discrimination was measured with a 30-ms, 6,000-Hz pedestal for pedestal levels from 20 to 80 dB SPL. As expected, intensity discrimination thresholds for all but one subject increased from low to moderate levels and then decreased, resulting in a midlevel hump. The authors found a significant correlation between intensity discrimination thresholds and the slope of the BM I/O function in 7 of 10 NH subjects. This correlation and the results of a follow-up experiment led Roverud and Strickland (143) to conclude that cochlear compression is a primary contributor to the midlevel hump.

In a follow-up study, Roverud and Strickland (142) measured intensity discrimination thresholds in the presence of ipsilateral, contralateral, and bilateral broadband noise to test whether a reduction in cochlear gain via the MOC reflex improved intensity discrimination thresholds for pedestal levels at the midlevel hump. The 50-ms noise was temporally centered on the 6,000-Hz, 30-ms pedestal and was presented with or without a 100-ms forward fringe. The authors hypothesized that the MOC reflex would exert greater influence on intensity discrimination thresholds in conditions with the forward fringe compared to those with no fringe based on the relatively sluggish onset of the reflex (103). The level of the pedestal was held constant at the level associated with the highest intensity discrimination thresholds in quiet (i.e., a level within the midlevel hump) for each subject. Intensity discrimination thresholds for this pedestal level were measured for several ipsilateral, contralateral, and bilateral noise spectrum levels between −10 and 25 dB SPL for noise conditions with and without the forward fringe. For bilateral noise conditions, intensity discrimination was measured for several interaural noise level differences (contralateral noise level RE: ipsilateral noise level) from −5 to 15 dB. Intensity discrimination thresholds measured in the presence of ipsilateral noise were lower than those measured in quiet for four of eight subjects, regardless of whether the noise did or did not include the forward fringe. The authors interpreted this improvement as mediated primarily by cochlear suppression and secondarily by the MOC reflex. A similar improvement was observed for all subjects when bilateral noise was presented with a forward fringe, compared to no fringe, for an SNR of ∼12 dB. These improvements occurred despite an increase in detection thresholds for the pedestal in the presence of ipsilateral noise. The contralateral noise levels associated with improved intensity discrimination (compared with quiet and ipsilateral noise conditions) were unique for each subject, as was the magnitude of the improvement, illustrating marked individual differences in intensity discrimination of short tones in background noise. For half of the subjects (4 of 8), the addition of contralateral noise improved intensity discrimination thresholds beyond that observed for ipsilateral noise alone when the forward fringe was present. No such improvement was observed when the fringe was absent, which led the authors to conclude that the MOC reflex may improve intensity discrimination in bilateral noise when the pedestal is sufficiently delayed (e.g., 100 ms) from the onset of the noise. Although not mentioned by the authors, dynamic range adaptation may also contribute to the improvements in intensity discrimination they observed for pedestals in ipsilateral and bilateral noise when the forward fringe was present. Specifically, stimulation by the forward fringe of the masker may result in a rightward shift in rate-level functions of AN fibers (Ref. 58, in cat) and central auditory neurons (i.e., IC; Ref. 75, in guinea pig) as these neurons adapt to stimulus statistics. Such a shift in dynamic range will improve the neural coding of the intensity increment by decreasing the influence of neural saturation, as has been suggested for experiments in AM detection (80, 82).

AM detection is another method for studying the intensity resolution of the auditory system, and intensity-related cues are thought to facilitate detection of AM, particularly for low modulation rates (188). Interestingly, adults with cochlear HL may have lower (i.e., better) AM detection thresholds than adults with NH. This finding is evident when comparing AM detection thresholds between normal and impaired ears of subjects with unilateral HL (189, 190) and among groups of adults with normal and impaired hearing (191). This better-than-normal AM detection in adults with cochlear HL is expected to result from reduced cochlear compression to the extent that AM detection is mediated by rate-based cues within an auditory filter centered on the carrier frequency. Specifically, an increase in the slope of the BM I/O function from hearing loss increases the postcochlear AM depth relative to that expected for NH, as shown by comparing Fig. 7, A and B. This increase in BM I/O function slope increases the sensitivity of the auditory system to AM and thereby results in lower AM detection in subjects with cochlear HL compared to subjects with NH. Almishaal et al. (144) hypothesized that eliciting the MOC reflex with an ipsilateral precursor would improve AM detection thresholds in adults with NH, given that a reduction in cochlear gain increases the slope of the BM I/O function. This hypothesis is illustrated in Fig. 7, A and C, where the postcochlear AM depth (vertical double arrow and horizontal lines) is shown for an auditory filter centered on the carrier frequency for experimental conditions with (Fig. 7C) and without (Fig. 7A) a precursor. The authors measured AM detection thresholds for a 50-ms, 5,000-Hz carrier modulated at 20 Hz for several carrier levels from 50 to 85 dB SPL. The carrier was presented with low-level, notched noise to limit off-frequency listening, and AM detection thresholds were obtained with or without a 200-ms, 40-dB SPL notched-noise precursor. AM detection thresholds increased and decreased as a function of carrier level in the no-precursor condition, consistent with the midlevel hump observed in intensity discrimination studies (e.g., Ref. 185). In contrast, AM detection thresholds were lower (i.e., better) in the precursor condition for moderate to high carrier levels, resulting in improved AM detection from the precursor (i.e., “precursor benefit”) and thereby decreasing the midlevel hump. The authors concluded that the precursor benefit at moderate carrier levels, and the absence of such a benefit for low-level carriers, was consistent with a reduction in cochlear gain via the MOC reflex. A reduction in cochlear gain is expected to increase the BM I/O function slope at moderate levels where compression exists but not at lower levels where responses grow linearly. The precursor benefit was observed for moderate carrier levels, where thresholds were ∼20 × log10(m) = −10 dB, which is appreciably higher than thresholds observed with longer carriers (e.g., 800 ms), where thresholds approach −30 dB for low modulation rates (<200 Hz; Ref. 192). High AM detection thresholds are associated with high peak-to-valley ratios (or SNRs), calculated as SNR = 20 × log10[(m + 1)/(m − 1)] (193). The peak-to-valley ratio for 20 × log10(m) = −10 dB is ∼6 dB SNR. Although the results of Almishaal et al. (144) are consistent with the MOC reflex, the authors concluded that dynamic range adaptation of peripheral (i.e., AN) and central (e.g., IC) neurons were also consistent with their results.

Figure 7.

Predictions of the medial olivocochlear (MOC)/power spectrum model for amplitude modulation (AM) detection in the presence (C and D) and absence (A and B) of the precursor in adults with normal hearing (NH; A and C) and hearing loss (HL; B and D). Dashed blue lines represent the expected basilar membrane (BM) input/output (I/O) function in the absence of the precursor (i.e., without MOC stimulation). Solid green lines represent the same BM I/O function after MOC stimulation from an ipsilateral precursor or contralateral acoustic stimulation. The width of the gray shaded area on the x- and y-axes represents the input and output modulation (mod.) depths, respectively. The output modulation depth is also shown by the peak-to-valley difference of the waveform insets and the associated double arrows. The model predicts a larger output modulation depth in adults with HL compared to those with NH in the no-precursor condition because of limited outer hair cell gain in adults with HL (compare A and B). Moreover, the model predicts that the presentation of a precursor will improve the output modulation depth in adults with NH and to a lesser extent in adults with HL. Illustrations are modified from Jennings et al. (145).

Jennings et al. (145) modified the experiment of Almishaal et al. (144) to include older (>60 yr) adults with cochlear HL. This modification involved 1) decreasing the carrier frequency to 2,000 Hz (instead of 5,000 Hz) to ensure that no more than mild to moderate HL was present at the probe frequency and 2) increasing the modulation frequency to 40 Hz (instead of 20 Hz), as pilot data revealed that some older adults could not reliably perform the task when the modulation frequency was 20 Hz. The authors hypothesized that adults with cochlear HL would have better AM detection thresholds and a smaller precursor benefit compared to adults with NH. The latter hypothesis was based on the assumption that dysfunctional OHCs in adults with cochlear HL would limit the extent to which the MOC reflex could adjust OHC gain (Fig. 7D). Consistent with these hypotheses, adults with cochlear HL had lower AM detection thresholds for low to moderate carrier levels and a smaller precursor benefit compared to adults with NH (Fig. 8). These better-than-normal AM detection thresholds are reminiscent of better detection thresholds in the onset condition of temporal effects in simultaneous masking for adults with HL compared to adults with NH (see Temporal Effects in Simultaneous Masking).

Figure 8.

Amplitude modulation (AM) detection thresholds (20 × log10[m]) improve in the presence of an ipsilateral precursor. Thresholds were measured as a function of carrier level and in the presence or absence of a notched-noise precursor in adults with normal hearing (NH; A) and hearing loss (HL; B). The 50-ms, 2,000-Hz carrier was modulated at 40 Hz and was immediately preceded by the 40-dB SPL precursor. Data are from Jennings et al. (145).

The observation of a precursor benefit in AM detection in adults with NH has been replicated and expanded to include contralateral/bilateral precursors (146), the measurement of OAEs (82), and comparisons with adults with cochlear implants (CIs) (80). Marrufo-Pérez et al. (146) measured AM detection thresholds for 50-ms, 1,500-Hz carriers presented near the onset or 100 ms before the offset of a 60-dB SPL broadband noise in adults with NH. These temporal locations of the AM probe within the masker are similar to the onset and steady-state conditions of temporal effects in simultaneous masking. They hypothesized that AM detection would improve in the steady-state condition relative to the onset condition, resulting in AM unmasking, and proposed the MOC reflex and dynamic range adaptation as potential mechanisms of this improvement. Consistent with this hypothesis, the authors reported lower (better) AM detection thresholds in the steady-state condition compared with the onset condition for 70-dB SPL and 25-dB SL AM probes presented in ipsilateral, contralateral, and bilateral noises. They concluded that their results were broadly consistent with the results of Almishaal et al. (144) and supported the potential role of the MOC reflex in AM unmasking and acknowledged that dynamic range adaptation may also contribute to this effect. Despite this, they noted that detection thresholds for the unmodulated carrier presented in contralateral noise improved in the steady-state condition compared with the onset condition, which is inconsistent with the effects of the MOC reflex on cat AN responses to tones presented in quiet (172, 194). The authors concluded that such improvements in AM detection in contralateral noise for the steady-state compared with the onset condition were consistent with “transient masking” (195). Bacon and Moore (195) defined transient masking as the process whereby “the transient response to the masker may reduce the usefulness of the transient response to the signal, making the signal more difficult to detect.” Here the “transient response to the masker” applies to the entire response of a short, gated masker or the onset response of a longer masker. In other words, detection of a short probe or a short burst of AM presented at the onset of a masker is impaired by the transient response to the masker’s onset. Mechanisms of transient masking are assumed to originate from central auditory processing pathways involved in stream segregation and attention (169, 195, 196), and this assumption highlights the role of brain mechanisms in the perception of masked sounds.

A puzzling finding that challenges the role of transient masking in the results of Marrufo-Pérez et al. (146) and studies on the temporal effect in simultaneous masking (168, 169) is the better-than-normal detection thresholds in the onset condition for adults with cochlear HL (145, 157). It is unclear why cochlear hearing loss would reduce a masking effect (i.e., transient masking) that is assumed to originate from binaural/cortical processing. Moreover, Jennings et al. (10) noted that transient masking is inconsistent with differences in growth of masking slopes for the onset and steady-state conditions of the temporal effect in adults with NH. Nevertheless, the extent to which peripheral and central auditory pathways contribute to temporal effects in masking and AM detection is poorly understood, and this knowledge gap tempers conclusions about the role of specific mechanisms on perceptual performance. Experiments centered on informational masking (e.g., Ref. 197) and modulation masking (e.g., Ref. 198) provide convincing evidence that processes beyond the cochlea strongly influence detection thresholds. When experiments are designed to limit informational masking (e.g., Ref. 199), the resulting masking thresholds are determined by peripheral processing (i.e., energetic masking), and such studies form the basis for behavioral estimates of cochlear frequency selectivity (e.g., Ref. 200), compression (e.g., Ref. 90), suppression (e.g., Ref. 201), and gain (e.g., Ref. 106).

To further assess the influence of the MOC reflex and dynamic range adaptation on AM unmasking, Marrufo-Pérez et al. (80) measured AM unmasking in 10 adults with CIs (2 unilaterally implanted, 8 bilaterally implanted) and compared these results with the NH listeners from Marrufo-Pérez et al. (146). They reasoned that direct stimulation of the AN via a CI results in perception that bypasses the effects of the MOC reflex on OHC function, particularly in patients with bilateral CIs; thus, this population serves as a model for deefferentation (202, 203). Similarly, direct stimulation of the AN likely retains the effects of dynamic range adaptation in AN fibers and neurons in the central auditory nervous system. Thus, the authors hypothesized that AM unmasking would be small or absent in adults with CIs to the extent that such unmasking is mediated by the MOC reflex. AM detection was measured using the same stimuli and procedures as Marrufo-Pérez et al. (146), except that the levels of the AM carrier and maskers were adjusted to accommodate listening through CIs. AM detection improved in the steady-state condition compared to the onset condition in most subjects with CI. This improvement was ∼4 dB and was independent of the laterality of the masker. The magnitude of AM unmasking was statistically equivalent among listeners with CI and those with NH. These results show that AM unmasking is observed in subjects with CIs where the MOC reflex is bypassed. The authors conclude that classic firing rate adaptation, dynamic range adaptation, and modulation-detection interference (204, 205) are consistent with AM unmasking in adults with CIs. Modulation detection interference is possible because CI processing may force adults with CIs to rely on cues that differ from those used by subjects with NH. For example, CI processing removes temporal fine structure cues and may blur spectral differences between the envelope of the AM probe and that of the masker. The authors conclude that AM unmasking is possible without the MOC reflex; however, they acknowledge that this finding does not rule out the potential role of the MOC reflex in AM unmasking observed in subjects with NH. They noted that differences in stimulus paradigms among studies (144–146) and the lack of AM unmasking in subjects with cochlear HL (145) prevented them from drawing strong conclusions from their data in adults with CIs about the role of the MOC reflex in AM unmasking in adults with acoustic hearing.

In summary, intensity discrimination and detection of AM applied to short tonal carriers improves when the carrier is preceded by ipsilateral, contralateral, and bilateral noise, resulting in unmasking. Such improvements are apparent in adults with normal hearing and patients with CIs and are markedly decreased or abolished in adults with cochlear HL. The effects of HL and carrier level on AM unmasking are consistent with the MOC reflex. Despite this consistency, AM unmasking in patients with CIs and the discrepancies between AM unmasking and OAE measures of MOC activity (Ref. 82; discussed in associations among perceptual and otoacoustic measures of moc reflex function) suggest that other mechanisms, such as dynamic range adaptation, may be involved. Better-than-normal AM detection in adults with HL (145) and AM unmasking from a precursor in adults with NH (82, 144, 146) are consistent with a unifying framework based on the effects of cochlear hearing loss and the MOC reflex on OHC gain (i.e., the MOC/power spectrum model). The ability of this framework to account for precursor effects and the effects of hearing loss in AM detection and temporal effects in simultaneous masking is a clear strength of the framework. However, this framework does not account for AM unmasking in patients with CIs. The strength of alternative frameworks/hypotheses based on dynamic range adaptation (e.g., Refs. 80, 82) depends, in part, on the ability to account for effects of hearing loss, in addition to precursor-related effects. Currently, it is unclear how dynamic range adaptation may account for smaller or absent AM unmasking in adults with cochlear HL. The extent to which temporal effects in masking and AM unmasking are influenced by central processes associated with transient and informational masking influences is unclear and requires further research.

EFFECTS OF CUTTING THE OLIVOCOCHLEAR BUNDLE ON PSYCHOPHYSICAL PERFORMANCE

In many species, the MOC bundle travels to the inner ear with the vestibular portion of the VIIIth nerve (21, 206). Vestibular neurectomy, a medical procedure involving surgical resection of vestibular afferents, may coincidently sever MOC fibers. In cases of intractable vertigo from Meniere’s disease or other pathologies, patients may elect to undergo vestibular neurectomy to mitigate vestibular symptoms (207). This procedure is done for the ear producing the offending vestibular symptoms, thus leaving patients with “operated” and “unoperated” ears. Hearing in the operated ear is typically preserved in patients with vestibular neurectomy (208). In theory, experiments involving such patients are uniquely suited for studying the effects of the MOC reflex on perception by comparing results from the unoperated ear (MOC bundle intact) with those from the operated ear (MOC bundle severed) or comparing results before versus after surgery.

Vestibular neurectomy in humans is expected to have a similar effect on auditory perception as cutting the MOC bundle of laboratory animals, where abnormalities in intensity discrimination (70), central masking (209), and vowel identification in noise (210) have been demonstrated. Results from a wide range of psychophysical tasks have been reported for human patients with vestibular neurectomy (211, 212). Many of these tasks center on evaluating the intensity resolution of the auditory system and include loudness growth, loudness adaptation, loudness summation, intensity discrimination, temporal effects in masking, forward masking, and tone detection in ipsilateral, contralateral, or bilateral noise. In addition to evaluating intensity resolution, tasks involving frequency discrimination, pitch, temporal gap discrimination, laterality judgments, informational masking, manipulation of attention, and speech perception have been reported. Scharf et al. (211) evaluated the influence of vestibular neurectomy on a wide array of psychophysical tasks in 16 subjects with Meniere’s disease. Four of these subjects had “near-normal” hearing, and the remaining had varying degrees of hearing loss. Not all subjects participated in all tasks. Vestibular neurectomy had little influence or produced mixed results on critical ratios measured with ipsilateral and bilateral noise (15 subjects), temporal effects in simultaneous masking (2 subjects), frequency selectivity measured with notched-noise maskers (7 subjects), intensity discrimination (4 subjects), loudness adaptation (7 subjects), and lateralization (2 subjects). A similar null effect was reported by Morand-Villeneuve et al. (213), where loudness ratings did not differ among operated and unoperated ears in six subjects with vestibular neurectomy. In contrast, Scharf et al. (211) reported a consistent effect of vestibular neurectomy for a task involving the detection of unexpected tones, using the “probe-signal” method (214, 215). This method, as employed by Scharf et al. (211), involved presenting a low-level (e.g., 5 dB SL) cue tone before two or more observation intervals. A randomly selected observation interval contained the probe presented just above threshold (e.g., 1 dB SL), whereas the other intervals contained silence. The probe was the same frequency (expected signal) or a slightly lower/higher frequency (unexpected signal) than the cue tone. Consistent with a previous case study (216), detection of unexpected tones, compared with expected tones, improved in the operated ear of subjects with vestibular neurectomy compared with the unoperated ear or the operated ear before surgery. These results led Scharf et al. (211) to conclude that resection of the olivocochlear bundle has little effect on performance in simple auditory tasks. They reasoned that more complex auditory tasks may be needed to reveal effects (if any) of the MOC reflex on perception and that the primary function of the reflex may be to facilitate development of the auditory system in infants and children. Auditory development requires that attention be directed to certain sounds and away from others, and the MOC reflex may facilitate this directed attention (217).

Some effects (or lack thereof) of vestibular neurectomy in humans are inconsistent with lesioning the MOC bundle in laboratory animals. For example, the lack of an effect of vestibular neurectomy on intensity discrimination in humans (211, 216) is inconsistent with animal studies where deficits in intensity discrimination after cutting the olivocochlear bundle were observed (70). Moreover, it is unclear from studies in animals why lesioning the olivocochlear bundle would improve gap discrimination, as was reported for a patient with vestibular neurectomy (216). Finally, the slightly improved frequency discrimination in four subjects with vestibular neurectomy (211) is inconsistent with olivocochlear bundle lesions having no effect on frequency discrimination in lesioned cats (71).

Zeng et al. (212) extended the work of Scharf et al. (211) by studying the effect of vestibular neurectomy on speech perception and several additional psychophysical tasks. These tasks included loudness matching, temporal effects in simultaneous masking, intensity discrimination of suprathreshold probes in the onset and steady-state conditions of the temporal effect in simultaneous masking, and intensity discrimination of a short probe presented after a forward masker. Subjects were six patients who had undergone vestibular neurectomy for various causes of vertigo (including Meniere’s disease), vestibular neuronitis, and head trauma. Three of these subjects had hearing thresholds better than 30 dB HL for frequencies from 250 to 8,000 Hz, and the remaining subjects had various degrees of hearing loss. In one subject (RA), the intensity of a 2,000-Hz tone in the operated ear was compared to that of a similar tone in the unoperated ear during a loudness-matching task. This task was completed before and after surgery. Before surgery, the intensity needed to match the loudness of the 2,000-Hz tone was slightly larger in the operated ear than the unoperated ear, consistent with higher audiometric thresholds for the probe frequency in the operated ear. After surgery the intensity needed to match loudness between ears decreased in the operated ear relative to the preoperation results, despite a 10-dB increase in audiometric thresholds for 2,000-Hz tones postoperatively. This suggests that vestibular neurectomy may have sensitized the operated ear to loud sounds, consistent with the putative role of a dysfunctional MOC system in hyperacusis (e.g., Ref. 218). In five subjects, temporal effects in simultaneous masking were smaller in the operated ear compared with the unoperated ear. Smaller temporal effects were due to the operated ear having lower thresholds in the onset condition (2 subjects) and higher thresholds in the steady-state condition (3 subjects) compared with the unoperated ear. Lower thresholds in the onset condition are consistent with increased hearing loss after surgery; however, poorer thresholds in the steady-state condition are better explained by dysfunction of the MOC reflex. Thus, for roughly half of the subjects with vestibular neurectomy, smaller temporal effects in the operated versus unoperated ear are consistent with the hypothesized role of the MOC reflex on temporal effects in simultaneous masking (136). A similar perceptual deficit for the operated versus unoperated ear was observed for intensity discrimination thresholds of suprathreshold probes presented in the onset and steady-state conditions of the temporal effect. Here probes were presented at 80–85 dB SPL, which corresponds to 2–20 dB SL, depending on the subject. Intensity discrimination in both onset and steady-state conditions was uniformly poorer in the operated ear compared with the unoperated ear. These poorer thresholds are consistent with a small postcochlear SNR between the pedestal and intensity-incremented pedestal due to cochlear compression in the operated ear. The better thresholds in the unoperated ear are consistent with a reduction in compression, and thus a larger postcochlear SNR, as a result of eliciting the MOC reflex. Finally, Zeng et al. (212) measured intensity discrimination of a short (25 ms), 1,000-Hz pedestal in the presence of a 100-ms, 1,000-Hz forward masker for pedestal levels of 40–90 dB SPL. The time delay between the masker offset and pedestal onset was 100 ms. Intensity discrimination under such conditions is poorer at moderate levels (e.g., 60 dB SPL) compared with lower or higher levels, resulting in a midlevel hump (186, 219, 220). Two subjects with vestibular neurectomy were tested, and both exhibited poorer intensity discrimination for moderate-level pedestals and a larger midlevel hump in the operated compared with the unoperated ear. Zeng et al. (212) interpreted this finding as evidence for weak/absent efferent activity in the operated ear, which is consistent with the MOC/power spectrum model. In the context of this model, the forward masker evokes the MOC reflex and reduces cochlear gain and compression, thereby improving the postcochlear SNR between the pedestal and increment. This reduction in gain has the largest effect at moderate levels where cochlear compression is strong (e.g., Ref. 41). Such a reduction in gain is not expected in the operated ear due to resection of auditory efferents during vestibular neurectomy, thus resulting in relatively higher intensity discrimination thresholds for the operated compared with unoperated ear.

In summary, studies on patients with vestibular neurectomy have yet to provide a clear description of the role of the MOC reflex on human hearing. In some cases, these studies contradict one another, such as in revealing the effect of vestibular neurectomy on temporal effects in simultaneous masking (cf. Refs. 211, 212). Moreover, these studies may be inconsistent with psychophysical effects observed in laboratory animals with severed olivocochlear bundles, where deficiencies in intensity discrimination have been demonstrated. Despite advantages of controlling for the MOC reflex by comparing operated and unoperated ears (or operated ears pre- vs. postsurgery), vestibular neurectomy is associated with some disadvantages. For example, most patients who undergo vestibular neurectomy have hearing loss, and this loss may worsen after surgery. The presence of hearing loss is expected to dilute any effects of the MOC reflex when OHCs are dysfunctional (10). Similarly, studies on the effects of vestibular neurectomy on perception were conducted before the MOC/power spectrum model was fully developed; thus, such studies did not consider the influence of postfilter SNR or the frequency relationship between probe and masker when designing perceptual experiments. There is some evidence that vestibular neurectomy may not sever the MOC bundle in humans (221), and even so the effects of resecting the bundle may induce neuroplastic changes that compensate for any perceptual deficits (15). Finally, a caveat to consider when interpreting the effects of vestibular neurectomy on perception in patients with Meniere’s disease is that the comparison between operated and unoperated ears may be confounded by the unknown effect of the disease on the MOC reflex. Thus, null effects may be due to abnormal MOC function in both operated and unoperated ears to the extent that the unoperated ear is affected by Meniere’s disease. Further research is needed to assess the effect of Meniere’s disease on MOC function.

ASSOCIATIONS AMONG PERCEPTUAL AND OTOACOUSTIC MEASURES OF MOC REFLEX FUNCTION

Otoacoustic emissions (OAEs) are low-level sounds emitted by the ear spontaneously or in response to acoustic stimulation (222, 223). OAEs are thought to originate from nonlinear cochlear processes that depend on the status of the cochlear amplifier (224). Partial loss of cochlear amplifier gain, as is often the case with cochlear HL, is associated with diminished or absent OAEs (225). OAEs have been used as a noninvasive proxy for studying how cochlear amplifier gain is modulated by the MOC reflex (e.g., Ref. 226). Studies have included OAEs evoked by transients, such as an acoustic click [i.e., transient-evoked OAEs (TEOAEs), e.g., Ref. 227], pairs of closely spaced tones that result in cochlear distortion products [i.e., distortion-product OAEs (DPOAEs), e.g., Ref. 228], and single-frequency tones [i.e., stimulus-frequency OAEs (SFOAEs), e.g., Ref. 229]. Here, the MOC reflex is studied by 1) observing the influence of a contralateral sound elicitor on the amplitude, frequency, and phase of OAEs generated in the ipsilateral (probe) ear (e.g., Ref. 227) or 2) monitoring time-varying changes in OAE amplitude, frequency, and phase in the probe ear over the duration of the ipsilateral probe stimulus (e.g., Ref. 230). Contralateral and ipsilateral elicitors are expected to evoke a response from the uncrossed and crossed olivocochlear bundles, respectively. Thus, OAEs offer a noninvasive assay of MOC function. In addition to automatic/reflexive stimulation by contralateral sound, the MOC system is modulated by corticofugal pathways, which results in suppression of transient OAEs (231) or suppression/enhancement of distortion-product OAEs (Ref. 232, in gerbil).

Several studies on the relationship between the MOC reflex and auditory perception have adopted a design that compares MOC function measured from OAEs with the effects of contralateral noise/ipsilateral precursors on psychophysical performance [e.g., Ref. 233 (contralateral noise) and Ref. 116 (precursors)]. This design centers on the expectation that both OAE generation and psychophysical performance depend on the status of the cochlear amplifier. Moreover, if perception depends on MOC function, OAEs and psychophysical performance are expected to covary when the MOC reflex is elicited. This covariance may be evaluated using correlation (e.g., Ref. 234) or by observing how OAE assays of MOC function and psychophysical performance do or do not codepend on manipulations of independent variables such as masker-probe delay (116), temporal envelope fluctuations (82, 235), and the presence/absence of auditory processing disorder (236).

Correlations between OAE Suppression and Simultaneous Masking

Several studies have assessed the relationship between detection thresholds in simultaneous masking and contralateral suppression of OAEs (233, 234, 237, 238). Micheyl et al. (233) measured detection thresholds for a 100-ms tone complex temporally centered in a 400-ms, ipsilateral broadband noise presented with or without contralateral noise in 25 young, normal-hearing adults. They measured contralateral suppression of OAEs in the same adults for a shorter version (3 ms) of the tone-complex probe presented in quiet. Correlation analysis revealed that detection thresholds in the presence of contralateral noise were proportional to the magnitude of OAE suppression. In other words, subjects with the largest OAE suppression (i.e., strongest MOC reflex based on OAE measures) had the highest (poorest) detection thresholds. Similarly, stronger OAE suppression was associated with a larger shift in threshold with, compared to without, the presence of contralateral noise. Karunarathne et al. (238) reported a similar correlation between contralateral suppression of click-evoked OAEs and detection thresholds of a broadband “alarm sound” in background noise. Namely, stronger OAE suppression was associated with poorer detection thresholds for the alarm sound. Micheyl and Collet (234) used a design similar to that of Micheyl et al. (233), except they used a sinusoidal rather than a tone-complex probe. Subjects were divided into two groups based on whether they completed the experiment with contralateral noise first (contrafirst group) or second (contrasecond group). All correlations between OAE and detection threshold measures were insignificant for the 1,000-Hz probe; however, detection thresholds for 2,000-Hz probes in contralateral noise were negatively correlated with the magnitude of contralateral suppression of click-evoked OAEs. In other words, subjects with the strongest OAE suppression had the lowest (best) detection thresholds. Contralaterally induced shifts in detection threshold for 2,000-Hz probes were also negatively correlated with OAE suppression, where subjects with strong OAE suppression had the smallest threshold shifts; however, this correlation was only observed in the contrafirst group. Similar findings were reported by Bhagat and Carter (237), who studied the relationship between DPOAE I/O functions in the presence and absence of contralateral noise and tone-in-noise detection thresholds. A significant correlation revealed that larger shifts in DPOAE compression breakpoints were associated with lower (better) detection thresholds for a 1,000-Hz tone presented in broadband noise. No such correlation existed when the probe was 2,000 Hz.

Correlations between OAE Suppression and Forward Masking, Intensity Discrimination, and Loudness

The correlation between contralateral suppression of OAEs and forward masking, intensity discrimination, and loudness has also been studied. Kawase et al. (180) measured PTCs in the presence and absence of contralateral noise in six subjects with NH. The probe was a 20-ms, 2,000-Hz tone burst presented during (simultaneous masking) or after (forward masking) a 505-ms, narrowband-noise masker. The masker was centered on 500, 750, 1,000, 1,500, 1,800, 2,000, and 3,000 Hz and was gated with or without contralateral white noise. Contralateral noise increased the effectiveness of off-frequency maskers, resulting in lower masker levels at detection threshold. This reduction in forward masker level at threshold was positively correlated with contralateral suppression of OAEs. In other words, subjects with stronger suppression of DPOAEs were more susceptible to forward masking than subjects with weaker suppression. Fletcher et al. (128) observed a similar increase in off-frequency masker effectiveness when TMCs were measured with contralateral noise present versus absent. Despite the use of a similar probe frequency (2,000 Hz) as Kawase et al. (180), the effect of contralateral noise on detection thresholds in forward masking was not significantly correlated with suppression of OAEs. The significant (180) and not significant (128) correlations between forward masking thresholds with/without contralateral noise and suppression of OAEs may originate from differences in study design. For example, Fletcher et al. (128) assessed the correlation between suppression of click-evoked OAEs and the difference in gain estimated from TMCs measured with and without contralateral noise. In contrast, Kawase et al. (180) evaluated the correlation between suppression of DPOAEs and the average reduction in low-frequency (750, 1,000 Hz) masker level for detection thresholds measured in the presence and absence of contralateral noise.

Intensity discrimination, which may be mediated by similar mechanisms as auditory masking (e.g., Ref. 186), is affected by the presentation of contralateral noise. Micheyl et al. (239) measured intensity discrimination thresholds and suppression of click-evoked OAEs in 20 subjects with NH. Suppression of OAEs was measured for several click levels, resulting in OAE I/O functions. These I/O functions were well accounted for by a line, and the change of slope and intercept of this line due to presentation of contralateral noise was used to quantify OAE suppression. Intensity discrimination thresholds were measured for a 200-ms, 1,000-Hz pedestal in quiet, in ipsilateral noise, in contralateral noise, and in ipsilateral-plus-contralateral noise. Broadband noise was presented at 10 dB and 35 dB SL to the ipsilateral and contralateral ears, respectively. Intensity discrimination thresholds were lower (better) when noise was presented to both ears rather than just the ipsilateral ear. The improvement in intensity discrimination in binaural versus monaural noise was significantly correlated with changes in the slope and intercept of OAE I/O functions when contralateral noise was presented. The direction of these correlations was consistent with the interpretation that subjects with strong OAE suppression were more sensitive to changes in intensity than subjects with weaker OAE suppression. A follow-up study from the same group (213) assessed the correlation between suppression of OAEs and measurements of loudness, including loudness growth, loudness integration, and loudness summation. Subjects were 22 adults with NH and 6 adults who underwent vestibular neurectomy and had various degrees of hearing loss. Although the authors reported significant contralateral suppression of 1,000-Hz and 2,000-Hz tone burst OAEs in normal-hearing adults, contralateral noise had little to no effect on measures of loudness. Moreover, subjects with vestibular neurectomy did not differ in their loudness judgments between operated and unoperated ears. Finally, the correlations between OAE suppression and any measure of loudness (growth, integration, summation) were insignificant.

Summary of Correlation-Based Studies on OAE Suppression and Perception

The studies discussed above reveal an inconsistent correlation between suppression of OAEs and psychophysical performance, as measured by masking (simultaneous and forward), intensity discrimination, and loudness tasks. The inconsistency is marked by disparate conclusions regarding the role of the MOC reflex in perception, where MOC reflex strength (as measured via suppression of OAEs) is not associated with perceptual performance (e.g., Refs. 128, 213) or is associated with better (234, 237, 239) or worse (180, 233, 238) perceptual performance.

The MOC/power spectrum model predicts that the effect of the MOC reflex on perceptual performance will depend on the postfilter SNR. Specifically, the model predicts that performance will improve (perceptual benefit), deteriorate (perceptual detriment), and remain unchanged when the reflex is elicited for SNRs that are positive, negative, and zero, respectively. Consistent with this prediction, many OAE studies that reported a perceptual benefit of the MOC reflex had positive postfilter SNRs (237, 239), whereas those that reported a perceptual detriment had negative postfilter SNRs (233, 238). These findings are tabulated in Table 1. Despite this, the relationship between OAE suppression and perceptual performance is not as predicted based on postfilter SNR for other studies (128). Thus, the interpretation of perceptual/OAE-correlation studies may be improved by considering the expected effects of postfilter SNR; however, the effects of SNR alone are not sufficient to unify all perceptual/OAE studies under a single framework.

Table 1.

Summary of relationship between otoacoustic emission measures of MOC reflex function and SNR at threshold for perceptual masking/intensity discrimination tasks

Article	Task	SNR, dB	Correlation
Micheyl et al. (233)	Simultaneous masking	likely <0	Negative
Micheyl and Collet (234)	Simultaneous masking	−2 to +4	Negative
Micheyl et al. (239)	Intensity discrimination	+30	Positive
Kawase et al. (180)	Simultaneous/forward masking	+10 to +15	Positive
Bhagat and Carter (237)	Simultaneous masking	−3.5 to +10	Negative
Fletcher et al. (128)	Forward masking	−40 to 0	Not significant
Karunarathne et al. (238)	Simultaneous masking	−10 to −20	Negative

Summary of the relationship between otoacoustic emission measures of medial olivocochlear (MOC) reflex function and signal-to-noise ratio (SNR) at threshold for perceptual masking/intensity discrimination tasks. Studies with large positive SNRs (bold text) are associated with positive correlations, whereas studies with negative or small SNRs are associated with negative correlations. SNRs were approximated by estimating the power passing through the auditory filter centered on the probe frequency according to the equivalent rectangular bandwidth estimates of Glasberg and Moore (69).

Temporal Effects in Simultaneous Masking, AM Unmasking, and OAE Assays of MOC Function

The role of the MOC reflex in perception may be studied by observing the degree to which OAE assays of MOC function covary with perceptual results when the independent variables of the study are manipulated (82, 116, 138, 235). Keefe et al. (116) compared SFOAEs and temporal effects in simultaneous masking using similar OAE/perceptual stimuli and procedures in 14 normal-hearing listeners. This comparison was made under the assumptions of the MOC/power spectrum model and the notion that SFOAEs are a “non-invasive correlate to behavioral tone-in-noise tests” (116). The primary outcome variables were the psychophysical probe and OAE levels defining detection threshold with a maximum likelihood procedure. The probe for the psychophysical experiments was a 20-ms, 4,000-Hz tone burst that was delayed by 1 or 200 ms from the onset of a 400-ms, notched-noise masker. The same probe and masker were used in OAE measurements, along with a suppressor stimulus (4,160 Hz, 30 ms) that was used to extract the SFOAE distortion waveform. The SFOAE distortion waveform is the nonlinear residual resulting from the subtraction of the ear canal pressure measured with simultaneous presentation of the probe, suppressor, and noise (p12) from the sum of probe-in-noise (p1) and suppressor-only (p2) waveforms (see Ref. 116 for details). Probe thresholds for both techniques (i.e., psychophysical, OAE) were measured for the following masker spectrum levels: −14, −2, 10, 22, 34, 46 dB SPL/Hz. The authors reasoned that OAE and psychophysical thresholds should covary with the temporal position of the probe stimulus to the extent that these thresholds share a common mechanism (i.e., MOC reflex). Specifically, they hypothesized that OAE and psychophysical thresholds would be lower (better) when the probe was presented near the temporal center of the masker compared with the masker’s onset. As expected, growth of simultaneous masking was linear (1 dB/dB) for probes presented at the temporal center of the masker and slightly steeper for probes presented near the masker’s onset (155). This outcome resulted in a psychophysical temporal effect that increased from ∼5 dB to 16 dB when the masker spectrum level was raised from −14 dB SPL/Hz to 22 dB SPL/Hz, consistent with previous studies (e.g., Ref. 117). In contrast, growth of OAE masking was very shallow (slope = 0.26 dB/dB), and OAE thresholds showed very little dependence on the temporal position of the probe within the masker. In other words, temporal effects were absent for OAE masking thresholds. The lack of agreement between OAE and psychophysical masking thresholds led the authors to conclude that efferent function may not be the dominant mechanism underlying temporal effects in simultaneous masking or that efferent effects in masking act on neural rather than on OHC responses. Alternatively, the authors suggested that their OAE and threshold measurement procedures were not sensitive to detecting the effect of the MOC reflex on temporal effects in simultaneous masking.

The design of Keefe et al. (116) hinges on the assumption that psychophysical and OAE masking share one or more common mechanisms for stimuli designed to elicit the temporal effect. The stark differences between the growth of masking slopes and the magnitude of the temporal effects among OAE and psychophysical techniques are consistent with independent masking mechanisms for these techniques. The study in Keefe et al. (116) was designed to test the assumptions of the MOC/power spectrum model for psychophysical and OAE masking data. Although models of psychophysical masking have been developed and refined over decades (107, 108), models on the mechanisms of OAE masking and how these mechanisms relate to perception have not been established. Such a model would specify how the MOC reflex influences the ear canal sound pressure of the probe tone, suppressor tone, and noise, all of which are needed to estimate SNR and predict OAE masking thresholds. Moreover, this model would specify how the observed OAE results relate to the signal and masker responses at the output of the AN, which mediates perception. The lack of such a model presents a challenge to interpreting the results from Keefe et al. (116).

Walsh et al. (138) developed a similar nonlinear SFOAE technique to study the relationship between temporal effects in simultaneous masking and an OAE measure of the MOC reflex. This technique was developed concurrently with and independently from the work of Keefe et al. (116). Temporal effects were measured in seven adults with NH with a 10-ms, 4,000-Hz probe presented at several temporal positions (5–385 ms) from the onset of a 400-ms, noise masker. Masker level at threshold was measured for a 60-dB SPL probe presented with several spectral configurations of the masker (broadband, low pass, band pass, high pass). The authors used these measurements to reveal the time course of the temporal effect for maskers with differing spectral configurations. A probe (10-ms, 4,000-Hz tone bursts) and masker (400-ms broadband noise) similar to the psychophysical experiment were used to obtain SFOAEs for several temporal positions of the probe within the masker. The authors used a three-interval (triplet) technique to extract the nonlinear residual SFOAE (nSFOAE). The first and second intervals contained the tone-plus-noise presented individually from receivers 1 and 2. The third interval included the tone-plus-noise stimulus presented simultaneously through both receivers. The nSFOAE is the difference between the linear sum of intervals 1 and 2 and the response of interval 3. The authors hypothesized that the magnitude of nSFOAE as a function of the OAE probe’s position within the masker would covary with the time course of the temporal effect (i.e., the development of masking) measured psychophysically, consistent with a common MOC-based mechanism. Masker level at threshold increased with increasing probe delay for broadband and low-pass maskers. This increase in threshold resulted in a temporal effect that increased in magnitude as a function of probe delay with a time constant of ∼65 ms, consistent with previous studies (151, 240, 241). When averaged across subjects, the temporal effects for probes presented at the onset (5-ms delay) versus the temporal center (200-ms delay) of the masker were 6–7 dB for broadband and low-pass maskers. Similar temporal effects for high-pass and band-pass maskers were small (3.0 dB) or absent (0 dB), respectively. nSFOAE magnitude increased as a function of probe position with a time constant of 72.0 ms, similar to the psychophysical temporal effect. These changes in nSFOAE magnitude were observed for broadband and low-pass noises but not for high-pass and band-pass noises. Thus, the effect of masker spectrum on the temporal effect and nSFOAE magnitude as a function of probe position is qualitatively similar. Despite similarities between the temporal effect and the nSFOAE assay from Walsh et al. (138), the authors caution against the strict interpretation that the temporal effect is an obligatory result of changes in cochlear gain over the duration of the masker, due to inconsistent results among a subset of the subjects. For example, temporal effects in simultaneous masking were generally absent in one subject (KW), despite this subject having typical nSFOAE results. Similarly, two subjects showed large temporal effects with high-pass noise but showed no change in nSFOAE for these noises. Finally, the authors proposed a sophisticated explanation to interpret the paradoxical finding that nSFOAE magnitudes increase over the duration of broadband and low-pass maskers, consistent with an increase in cochlear nonlinearity. The paradox is that cochlear nonlinearity is expected to decrease over the duration of the masker from MOC feedback. The authors present a model based on BM I/O functions and the assumption that cochlear gain during the third presentation of the triplet (tone-plus-noise in both earphones) is lower than that of the first and second presentations (tone-plus-noise in individual earphones). Although their model explains why nSFOAE magnitude increased with probe position, the extent to which these model assumptions are valid is unclear.

Wojtczak and colleagues studied the relationship between an SFOAE assay of MOC function and perception of AM sounds, including Schroeder phase complexes (235) and AM tones (82). Wojtczak et al. (82) tested whether AM unmasking from an ipsilateral precursor was accompanied by an MOC reflex-induced change in cochlear amplifier gain at the probe frequency, as measured from SFOAEs. This approach was based on the assumption that the precursor reduced the gain of the cochlear amplifier and improved AM detection, as predicted by the MOC/power spectrum model. AM unmasking was measured for a 40-ms tonal carrier presented near the onset of a 100-ms noise masker with a bandwidth of two octaves. The 1,000- or 6,000-Hz carrier was sinusoidally amplitude modulated at 50 Hz and presented after silence or one of three precursors. The 400-ms precursor was either a two-octave band noise, a harmonic complex, or a tone with the same frequency as the probe. AM detection was measured for 40-, 60-, and 80-dB SPL probes. The masker level ensured that AM detection thresholds [expressed as 20 × log(m)] were in the range of −6 to −2 dB in the no-precursor condition. AM unmasking was defined as the improvement in AM detection threshold in the precursor condition relative to that measured in the no-precursor condition. AM unmasking was largest (∼8 dB) with the noise precursor compared to other precursors and was independent of the frequency and level of the probe. Conversely, the tone-complex masker resulted in smaller AM unmasking (∼4 dB), and this unmasking was greatest for the 80-dB SPL, 1,000-Hz probe. AM unmasking was not observed for the tonal precursor.

Wojtczak et al. (82) reasoned that a reduction in cochlear amplifier gain from the precursor would be accompanied by a decrease in SFOAE magnitude to the extent that the MOC reflex was responsible for AM unmasking. The change in SFOAE (ΔSFOAE) during the presentation of a 1,000-ms, two-octave band noise elicitor was measured with a heterodyne technique (137, 230). The frequency of the 40-dB SPL OAE probe tone was set at a peak within the OAE fine structure near 1,000 or 6,000 Hz. ΔSFOAE for the continuously presented OAE probe was calculated over a 6-s period that included the 1,000-ms presentation of the noise elicitor. This calculation involved a point-by-point subtraction of the complex-valued heterodyned OAE signal from a predefined baseline. This baseline was defined as the average heterodyned signal within a 400-ms window that ended 100 ms before the onset of the elicitor. The authors evaluated ΔSFOAE during four analysis windows that allowed for estimates of the noise floor, the effects of cochlear suppression, the combined effects of cochlear suppression and the MOC reflex, and the decay of the MOC reflex. The magnitude of the 1,000-Hz ΔSFOAE increased nearly instantaneously at the onset of low-level elicitors (30–40 dB SPL), consistent with cochlear suppression. For higher-level elicitors (50–60 dB), changes in the magnitude of the 1,000-Hz ΔSFOAE over time included a slower component in addition to this instantaneous component. The time course of this slower component was consistent with the time constant of the MOC reflex. The authors obtained response distributions for the estimates of the noise floor, the effects of suppression, and the effects of the MOC reflex. A significant effect of suppression was observed for all subjects for 1,000- and 6,000-Hz probes at all levels. Significant effects of the MOC reflex were observed for the 1,000-Hz ΔSFOAE at 50- and 60-dB SPL noise levels. Similar MOC effects were not significant for the 6,000-Hz probe. The distribution for the combined suppression-plus-MOC reflex effect was not significantly different from that of the suppression effect alone. The authors interpreted these OAE results as evidence that the MOC reflex could not have contributed to perceptual AM unmasking. Moreover, they reasoned that the independence of perceptual AM unmasking on probe level was inconsistent with the level-dependent effects expected from the MOC reflex. This reasoning is expressed in the following statement: “However, because the cochlea applies little or no gain for 80-dB SPL tones, it seems doubtful that the large AM unmasking effect was due to the MOCR [MOC reflex]” (82). This statement does not consider that a reduction in cochlear gain via the MOC reflex will influence the dips of the AM probe and the lower-level masker, which were 15–30 dB below the AM probe level. The cochlea does apply appreciable gain to these lower levels, and this gain may be adjusted by the MOC reflex to produce AM unmasking. Wojtczak et al. (82) proposed alternative mechanisms to account for the AM unmasking observed in their experiment including perceptual grouping (242), MOC feedback to the ventral cochlear nucleus (Ref. 243, in mouse), and dynamic range adaptation (e.g., Ref. 75, in guinea pig).

In summary, the role of the MOC reflex in psychophysical detection has not been clearly revealed by experiments that compare OAE assays of MOC function with perceptual results that appear consistent with MOC feedback. A survey of these studies reveals contradictory results and interpretations about the degree to which OAE findings support a functional role of the MOC reflex in perception. A major limitation of these studies is the lack of the development and validation of a conceptual framework that relates MOC-induced changes in OAEs to changes in AN firing rates, which mediate perception. Such a framework will likely be complex given recent results from McFadden et al. (244), who showed that OAEs (click evoked, distortion product, spontaneous) are poor predictors of perception for a large battery of psychophysical tests in a sample of 75 highly trained subjects. This battery included measurement of absolute thresholds, temporal effects in simultaneous masking, forward masking, psychophysical two-tone suppression, and psychophysical frequency resolution. They reported that the relationship between psychophysical performance and any variety of OAE was generally weak for both sexes and within any race category. This finding led the authors to conclude that mechanisms of OAE generation may not be systematically related to the detection of an acoustic probe in quiet or in the presence of a simultaneous or forward masker. By extension, it is likely that similar conclusions hold for the comparison between MOC-related changes in OAEs and the effects of the MOC reflex on psychophysical detection. Alternatively, psychophysical perception and OAE generation may share a common mechanism; however, this mechanism may be modulated by attention, resulting in uncorrelated findings between passive (e.g., OAEs) and active (e.g., psychophysics) listening tasks (15). The complexity of drawing correlations between contralateral suppression of OAEs and psychophysical performance was recently emphasized by Marrufo-Pérez et al. (245), who listed several uncontrolled factors that may dilute such correlations, including 1) limitations in exploring the stimulus parameter space, 2) nonperipheral perceptual masking, 3) attention, 4) the influence of the middle ear muscle (MEM) reflex, 5) limitations in suppressing the reflection component of DPOAEs, 6) standing waves in the ear canal contributing to inaccurate estimates of stimulus level, 7) poor test-retest repeatability for tone-detection thresholds, 8) naturally occurring individual differences in MOC reflex strength among adults with normal hearing, and 9) the possibility that suppression of OAEs may not exclusively originate from a reduction in cochlear gain.

Findings from McFadden et al. (244) are limited to the OAE types and analyses used in their study and do not indicate that all types/analyses of OAEs are poor predictors of psychophysical performance. Indeed, measurements of SFOAE group delay are consistent with psychophysical estimates of frequency resolution derived from notched-noise masking experiments (200, 246). Similarly, Dewey and Dhar (247) reported a striking similarity between fine-frequency measurements of behavioral detection thresholds in quiet (i.e., threshold microstructure) and the SFOAE microstructure. Finally, DPOAE I/O functions, suppression tuning curves, and suppression masking patterns are qualitatively consistent with psychophysical estimates of compression (e.g., Ref. 248), frequency selectivity (e.g., Ref. 249), and upward spread of excitation (250), respectively [see review by Johnson et al. (251)].

McFadden et al. (244) suggested that measures of auditory function based on neural responses (e.g., compound action potential, auditory brain stem response) might provide a stronger relationship with perceptual performance than OAE measures. This may be due, in part, to the finding that OAEs underestimate the effect of the MOC reflex on neural responses in cats (252). Several recent studies describe potential noninvasive neural assays of the MOC reflex (253–258). Future research is needed to determine the extent to which these neural assays support a role of the MOC reflex in psychophysical perception.

SUMMARY AND CONCLUSIONS

The relationship between the MOC reflex and psychophysical performance has been extensively studied over the past 30+ yr. The effects of an ipsilateral precursor or CAS on temporal effects in simultaneous masking, psychophysical estimates of cochlear gain and frequency selectivity, and intensity resolution are broadly consistent with a reduction in cochlear amplifier gain, which is the primary effect of the MOC reflex. Moreover, models that simulate a reduction in cochlear gain have been successful at accounting for such effects. Precursor effects on masking thresholds and AM detection are consistent with a framework (MOC/power spectrum model) that predicts that a reduction in cochlear gain will influence psychophysical performance for large positive or negative SNRs, or for conditions involving an off-frequency masker. Despite these findings, studies involving patients with vestibular neurectomy have yielded inconsistent results on the influence of the MOC reflex on psychophysical perception. Such studies may be limited by the potential that the MOC bundle is not cut during surgery, the possibility of neuroplastic changes in the auditory system after surgery, and the fact that many patients with vestibular neurectomy have co-occurring hearing loss. Similarly, the relationship between psychophysical perception and the MOC reflex has not been clearly revealed by studies that measured MOC reflex strength via OAEs. This lack of clarity may be due to the possibility that OAE generators are weakly related to perception or that a conceptual framework that relates OAEs to the neural signals that mediate hearing has not been developed. Finally, OAEs and perception are often measured during passive and active listening, respectively, which suggests that the influence of attention (if any) on the MOC reflex is not equivalent among measures.

Currently, the putative role of the MOC reflex on masking and intensity resolution in humans cannot be definitively confirmed or refuted because of the lack of strong, consistent support from studies on vestibular neurectomy and OAEs. This conclusion is dissatisfying and points to the need for the development of novel, innovative approaches for studying the MOC reflex in humans. These approaches may include measures that 1) control for the effects of attention, 2) provide and confirm a theoretical framework on the relationship between OAEs and AN coding, and 3) directly measure the effects of the MOC reflex on cochlear/AN activity via noninvasive auditory evoked potentials. Moreover, these approaches must simultaneously address whether other mechanisms, such as dynamic range adaptation, classic firing rate adaptation, and the middle ear muscle reflex, account for the apparent reduction in gain from psychophysical studies on masking and intensity resolution. Although studies on OAEs have been careful to account for the effects of the middle ear muscle reflex in studies of the MOC system (229, 259, 260), many perceptual studies have not included the same degree of rigor. Importantly, these novel approaches must show that the MOC reflex, an alternative mechanism, or some combination of mechanisms accounts for 1) the influence of cochlear hearing loss on AM unmasking and temporal effects in simultaneous masking, 2) the finding that precursor effects are largest at nonzero SNRs and for off-frequency masking conditions, 3) the nonmonotonic effects of precursor duration in forward masking, 4) the changes in frequency resolution caused by the presence of an ipsilateral precursor or CAS, and 5) AM unmasking in patients with CIs. The ideal result of these novel approaches would be to clearly delineate the contribution of each adaptive mechanism (i.e., MOC reflex, classic firing rate/dynamic range adaptation, middle ear muscle reflex) and the contribution of central processing (e.g., attention, transient/informational masking) to psychophysical measures of masking and intensity resolution. Experimental design considerations for accounting for the effects of informational versus energetic masking may include comparison of subjects with NH and HL and employing methods that limit informational masking (e.g., Ref. 261). Finally, approaches for evaluating the MOC reflex must account for the large variability in MOC reflex strength expected among human subjects based on research in laboratory animals (32, 262). These individual differences in MOC reflex strength are in addition to individual differences in psychophysical estimates of detection efficiency, cochlear compression, and cochlear gain, (e.g., Ref. 263) suggesting that fully accounting for these differences in psychophysical studies of the MOC reflex constitutes a significant challenge. Rigorous documentation of outer hair cell function via OAEs and audiometric thresholds for standard and extended high-frequency tones is a first step to limiting individual differences that are unrelated to individual differences in MOC reflex strength.

Despite compelling results in laboratory animals, the role of the MOC reflex in human communication is poorly understood because of a knowledge gap on how the physiology of the human MOC reflex relates to hearing. Continued research on the role of the MOC reflex in hearing is critical because such research has the potential to establish the human MOC reflex as playing a role in the relative ease or difficulty of hearing in noisy backgrounds in adults with normal hearing and hearing loss. A clear knowledge of the role of the MOC reflex in hearing is expected to open horizons for models of human hearing, the creation of audiological diagnostic tests of the MOC reflex, and the development of MOC reflex-inspired hearing devices, which will ultimately improve communication and quality of life for adults with hearing loss.

GRANTS

This work was supported by Grant K23 DC-014752 from NIH/NIDCD.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

S.G.J. interpreted results of experiments; prepared figures; drafted manuscript; edited and revised manuscript; and approved final version of manuscript.