Psychophysical changes in temporal processing in chinchillas with noise-induced hearing loss: A literature review

The prevalence of noise-induced hearing loss (NIHL) has steadily increased and recent surveys suggest roughly 25% of adults in the United States between the ages of 20 and 69 have features indicative of NIHL (Carroll et al., 2017). The most common symptom of NIHL is a shift in auditory thresholds. However, an important, but often overlooked, symptom of NIHL is the degraded ability to resolve temporal and spectral fluctuations in supra-threshold acoustic signals. These abilities are critical for understanding speech, especially in difficult listening conditions (Giraudi et al., 1980; Fitzgibbons and Wightman, 1982; Dubno et al., 1984; Zeng et al., 1999), highlighting the importance of audiometric testing beyond the standard audiogram.

Ideally, all studies investigating temporal processing and NIHL would incorporate neurophysiological responses, psychophysical data, and an anatomical assessment. This is nearly impossible in humans, especially when trying to control for the effects of aging and varying degrees of individual lifetime noise exposure. Animal models such as the chinchilla have been invaluable in the study of noise-induced temporal processing deficits for several reasons: (1) the chinchilla is audiometrically similar to humans, (2) chinchillas can be readily trained in behavioral tasks, (3) the shape and size of its skull allows for relatively easy access to the middle and inner ear, and (4) the life span of the chinchilla is approximately 20 yr, making them good subjects for tracking the long-term effects of noise exposure (Miller, 1970; Clark, 1991). However, one of the downsides of using chinchillas to study noise-induced hearing deficits is that chinchillas are significantly more susceptible to noise damage than humans, and also appear to be among the most noise-susceptible of all the commonly used laboratory animals (Dobie and Humes, 2017).

By employing rigorous psychophysical methods, researchers can reliably quantify a participant's subjective perception of a physical event while controlling for the subject's internal response biases (Green and Swets, 1966; Wolfe et al., 2009). An extension of auditory psychophysics is animal psychoacoustics, or the study of sound perception in nonhuman animals (Fay, 1994). The main difference between human and animal psychoacoustic studies is that animals have to be behaviorally trained to respond to stimuli, i.e., press a lever when they detect a sound, whereas humans can simply be instructed to perform a given task (Harrison, 1992; Fay, 1994; Long, 1994). There are a number of behavioral techniques used to train and test animals in psychoacoustic experiments. Training methods are generally broken down into two broad categories of either positive or negative conditioning. Both types of training methods have worked well in chinchilla psychoacoustic experiments although different species might respond better to one technique over the other. For instance, positive reinforcement, using food as a reward, was used to obtain reliable hearing thresholds before and after noise exposure, frequency modulation detection, and rate modulation detection in the chinchilla (Clark et al., 1974; Clark and Bohne, 1978; Long and Clark, 1984), while negative reinforcement using a conditioned shock avoidance technique has also been widely used in chinchilla psychoacoustic research. Using these standard behavioral conditioning techniques, combined with tested psychophysical methods for stimulus presentation and data analysis, we have elucidated several noise-induced deficits in temporal processing in a series of studies using the chinchilla psychoacoustic model.

To accurately examine auditory temporal resolution, it is first necessary to establish hearing thresholds for the stimuli used in subsequent temporal resolution experiments. Preliminary threshold measurements are taken by presenting noise stimuli at specified intensities above and below the subject's detection threshold to determine if hearing is normal or to identify the degree of hearing loss caused by noise exposure or other ototraumatic agents. A highly effective behavioral method for assessing hearing thresholds in chinchillas is the shock-avoidance conditioning technique that can be performed using a traditional shuttle-box apparatus (Miller, 1970) or, alternatively, a method we developed that employs a yolk-like restraint (Blakeslee et al., 1978; Giraudi et al., 1980; Giraudi-Perry et al., 1982). A major advantage of the yolk-like restraint is that it holds the animal's head at a constant position in the calibrated sound field which allows for greater control over stimulus presentation. When the chinchilla correctly detects the acoustic signal, it registers a response by making a slight upward motion of its head within a pre-defined post-response time window. A correct response to the stimulus, or “Hit,” is reinforced by turning on a light and avoidance of an electric shock. If the chinchilla fails to respond to the stimulus, a brief electric shock is turned on in combination with a loud buzzer to indicate a “Miss” (for details see Blakeslee et al., 1978). Initially, the animals were trained to detect 500 ms noise bursts in order to avoid a brief electric shock delivered to the tail. Once the animals were reliably responding to noise bursts, the psychophysical Method of Limits was used to determine noise burst thresholds. The noise burst thresholds were then used to establish the sensation level (dB SL) of the noise stimuli used for subsequent gap detection studies. Afterwards, tone bursts (500 ms) were introduced and once the chinchillas were responding reliably, the Method of Limits was used to determine thresholds over a broad frequency range.

Temporal resolution refers to the minimum amount of time needed to segregate, or resolve, acoustic events (Giraudi-Perry et al., 1982). Compared to other sensory systems, the auditory system has extraordinarily good temporal acuity. Most behaviorally meaningful stimuli, such as animal vocalizations and human speech, fluctuate rapidly in amplitude and frequency over time. A listener's ability to perceive these rapid and complex changes in acoustic signals depends on the auditory system's ability to resolve and segregate the changes in the temporal characteristics of sounds (Long, 1994). One of the simplest and most widely used measures of auditory temporal resolution is gap detection. In a typical gap detection task, the listener is trained to detect a brief silent interval, or gap, in an otherwise continuous acoustic signal, typically a broadband noise. As the duration of the gap decreases, the amplitude fluctuations in the signal can no longer be resolved or detected, producing the gap detection threshold or minimum detectable silent interval (Giraudi et al., 1980). Since the gap detection task utilizes simple acoustic stimuli rather than complex vocalizations or speech signals, it is one of the most widely used tests of auditory temporal resolution in both human and animal studies, and in both normal-hearing and hearing-impaired listeners (Salvi et al., 1986). Importantly, altered gap detection performance in hearing-impaired listeners is correlated with poor speech perception even when accounting for the effects of hearing loss, suggesting that the gap detection task is a robust measure of auditory temporal resolution (Fitzgibbons and Wightman, 1982; Tyler et al., 1982; Moore et al., 1992). While gap detection has been widely used to assess cochlear dysfunction, it has also been employed to test for central auditory processing deficits (Musiek et al., 2005; Fuente and McPherson, 2007; Gleich and Strutz, 2011).

After determining noise burst and pure-tone thresholds, chinchillas were trained to detect silent gaps in a wideband noise carrier. The stimuli were presented using the psychophysical Method of Constant Stimuli, which involves the random presentation of a broad range of gap durations (1, 2, 3, 4, 5, 7.5, 10, 30, and 50 ms) presented at a fixed intensity. Because the stimuli are presented in a random order, it reduces the risk that the animals are guessing or anticipating the next stimulus thereby reducing response bias and ensuring the animal's behavior is under stimulus control (Klink et al., 2006). Giraudi et al. (1980) was the first to obtain reliable gap detection thresholds in the chinchilla. In order to compare the data to human listeners, chinchilla gap detection thresholds were measured with two different white noise carriers, one low-pass filtered at 6 kHz and the other at 10 kHz. Gap thresholds were measured at 20, 30, 40, 50, 60, and 70 dB SL [intensities from 23 to 77 dB sound pressure level (SPL)]. Gap thresholds were approximately 3 ms for noise carriers presented from 30 to 70 dB SL, but gradually increased to 6 ms at 20 dB SL (Giraudi et al., 1980). Gap thresholds were slightly larger for noise low-passed filtered at 6 kHz than 10 kHz. These results from normal-hearing chinchillas are comparable to those obtained in humans (Plomp, 1964), birds (Dooling et al., 1978; Okanoya and Dooling, 1990), gerbils (Wagner et al., 2003), rats (Ison, 1982; Syka et al., 2002), and mice (Heffner and Heffner, 2001; Radziwon et al., 2009). Gap detection thresholds for humans and chinchillas are displayed in the schematic in Fig. 1 to illustrate the similarity among listeners with a similar frequency range of hearing.

A major advantage of using the chinchilla to identify deficits in psychophysical measures of temporal resolution is that the amount and type of NIHL can be systematically manipulated to simulate various types of NIHL in humans. Equally important, it allows for a comparison between pre- and post-exposure measures within the same subject (Giraudi-Perry et al., 1982). In contrast, human studies often rely on group comparisons, which introduce more threshold variability that can greatly hamper the interpretation of the data (Salvi and Arehole, 1985). Animal models also afford investigators greater control over the bandwidth and degree of NIHL, and inter-subject variability such as aging, genotype, and previous exposure history. Moreover, the behavioral data can be correlated with neurophysiological and anatomical data within the same subject (Giraudi et al., 1980; Zhang et al., 1990).

To determine the effects of noise exposure on temporal resolution in the chinchilla, Giraudi-Perry et al. (1982) obtained pre- and post-noise exposure audiograms and gap detection thresholds. Following pre-exposure testing, each animal was exposed to octave-band noise centered at 500 Hz at four successively higher intensities starting at 75 dB SPL and ending at 100 dB SPL (i.e., 75, 85, 95, and 100 dB SPL). During the first 18 h of each noise exposure, hearing thresholds progressively increased, but after approximately 24 h, hearing thresholds reached a plateau or asymptote, which remained stable for many months as long as the animals remained in the same noise exposure. The difference between the pre-exposure thresholds and thresholds measured during the plateau is referred to as an asymptotic threshold shift (ATS). Audiogram and gap detection testing began after the animals had been in each noise level for at least 2 days, insuring that thresholds had reached a predictable and stable ATS that affected nearly all frequencies similarly (i.e., a flat hearing loss) (Carder and Miller, 1972; Saunders et al., 1977; Salvi et al., 1978). An advantage of using this type of long-duration noise exposure instead of an acute exposure is that the resulting hearing loss, or ATS, is very stable over a long period of time unlike short-duration noise exposures where the immediate effects generally dissipate over hours or days (see Sec. IX). This is crucial when conducting animal psychoacoustic experiments because it can take days or weeks to obtain all of the relevant behavioral data. During an ATS exposure, the animals can be removed from the noise for a short period of time each day for testing and then returned to the noise to maintain the same level of ATS. Using this type of exposure, we have measured thresholds before, during, and after each exposure to insure maintenance of a stable ATS during the exposure, and permanent threshold shifts (PTSs) following the exposure (Clark, 1991).

Pure-tone, noise burst, and gap thresholds were obtained during the ATS associated with each level of noise exposure. The amount of ATS for each chinchilla was approximately 15, 30, 45, and 50 dB for the 75, 85, 90, and 100 dB SPL noise exposures, respectively. Due to the threshold shifts, the silent gaps in the noise would be perceptually less distinct, making them more difficult to detect. Mild acoustic trauma resulting in approximately 15 dB of threshold shift did not impact gap detection thresholds, despite a hearing loss that would ostensibly reduce the perceptual depth of the gap. However, higher level noise exposures inducing threshold shifts of 40 dB or greater resulted in increased gap detection thresholds. These deficits persisted even after compensating for hearing loss by comparing the gap thresholds at an equivalent SL (Fitzgibbons and Wightman, 1982; Giraudi-Perry et al., 1982). These results suggest that gap detection thresholds remain stable with mild NIHL, but PTSs of 40 dB or greater significantly increase gap detection thresholds in the chinchilla. The noise-induced increases in gap detection were most noticeable at 20–30 dB SL, suggesting that low to moderate SL sounds are the most useful for detecting temporal processing deficits. Furthermore, the gap detection thresholds were still abnormally long even after correcting for hearing loss. As in human psychophysical studies, it appears that NIHL can impact temporal resolution in the chinchilla beyond a simple change in hearing threshold (Giraudi-Perry et al., 1982). Increasing the sound intensity to correct for hearing loss does not restore temporal resolution, illustrating one of the shortcomings of hearing aid amplification.

The interpretation of temporal processing deficits based on psychophysical experiments from hearing impaired patients is complicated by our lack of understanding of the pathophysiological changes in both the peripheral and central auditory pathway. Temporal processing is often conceptualized as a two-stage filtering process consisting of a sharply tuned filter in the cochlea with a short time constant, mainly involved in temporal resolution, and a second filter located centrally with a longer time constant that mainly contributes to temporal summation (integration) (Duifhuis, 1973). As the filters become broader, their response times become shorter. Because NIHL leads to a broadening of the cochlear filters (broader neural tuning curves) (Salvi et al., 1979), one would expect that cochlear damage might lead to an improvement in temporal resolution. However, just the opposite occurs (larger gap thresholds) suggesting that sharp neural tuning, generally ascribed to an active process in the outer hair cells, is not critically important for temporal acuity. An alternative view is that the temporal fluctuations in the envelope of a signal are affected by compression of the basilar membrane input/output function. In a healthy ear, compression flattens the temporal envelope because more gain is applied to the lower level than higher level components of the signal (Horwitz et al., 2011). Since hearing loss, typically associated with outer hair cell damage, leads to a loss in basilar membrane compression, hearing impaired listeners could potentially have better than normal gap thresholds under certain conditions. However, gap detection deficits have been observed in carboplatin-treated chinchillas with intact outer hair cells, robust otoacoustic emissions, and normal thresholds in quiet, but poor detection thresholds in noise (Lobarinas, 2006; Lobarinas et al., 2013, 2016; Salvi et al., 2016). The latter results suggest that impaired temporal resolution may result from disrupted neural adaptation or, more specifically, recovery from adaptation (Harris and Dallos, 1979; Salvi et al., 1986; Relkin and Turner, 1988) which could occur at the afferent synapse beneath the inner hair cells (Kujawa and Liberman, 2015; Liberman and Kujawa, 2017).

Previous studies in humans indicate that gap detection thresholds are shorter (better) at higher frequencies (Fitzgibbons and Wightman, 1982). These results suggest that the high frequencies may play a more important role in temporal resolution than low frequencies. Because the high-frequency region is generally more susceptible to hearing loss from noise and other ototraumatic agents, gap detection thresholds may be especially susceptible to high-frequency hearing loss. To determine the effects of high-frequency hearing loss on gap detection, chinchillas were exposed to noise bands initially centered in high-frequencies that incrementally broadened to include lower frequencies (Salvi and Arehole, 1985). In the sequential exposures, the high-frequency cutoff of the noise remained at 20 kHz, but the low-frequency cutoff was lowered in octave steps (16, 8, 4, 2, and 1 kHz). The intensity of each noise exposure was held constant at 93 dB SPL overall and the animals were tested after being exposed to the noise for at least 48 h to ensure that a stable ATS was established when the gap-thresholds were measured. Pure-tone, noise burst, and gap detection thresholds were obtained pre-noise exposure and during each of the subsequent noise exposures. Behavioral measures of hearing impairment were evaluated with a conditioned shock avoidance paradigm and the thresholds to noise bursts, tone bursts, and gap stimuli were evaluated with an adaptive tracking procedure. As expected, the noise exposures resulted in a 50–60 dB hearing loss that progressed to the low-frequency region as the low-frequency cutoff of the noise was lowered. The noise burst thresholds showed only a small increase (<5 dB) during the first three noise exposures (16–20, 8–20, and 4–20 kHz; 93 dB SPL) suggesting that the audibility of the noise carrier remained unchanged. However, despite this slight change in noise burst thresholds, gap detection thresholds nevertheless increased even though audiograms were normal below 8 kHz (Salvi et al., 1986). The schematic in Fig. 2 shows a visual representation of the expansion of the noise exposure band and the corresponding change in gap detection thresholds. The results indicate that temporal resolution depends not only on the detectability of the noise carrier, but more importantly, on the audibility of the high-frequency components of the test stimuli (Salvi and Arehole, 1985). These results are similar to those seen in humans with high-frequency sensorineural hearing loss (Florentine and Buus, 1984), and suggest that temporal resolution largely depends on the audibility of the high-frequency energy of the noise carrier (Fitzgibbons, 1983; Salvi et al., 1986; Radziwon et al., 2009). These results suggest that noise exposure affecting the ultra-high frequencies may be particularly detrimental to temporal processing. Therefore, an individual with significant hearing loss above 8000 Hz would likely have more difficulty detecting the temporal fluctuations in speech sounds than a person with relatively normal hearing above 8000 Hz (Badri et al., 2011; Monson et al., 2014; Vlaming et al., 2014).

To examine the neural mechanisms of gap detection in the chinchilla, Zhang et al. (1990) recorded the single-unit discharge patterns of auditory nerve fibers that were stimulated with 1 to 10 ms gaps embedded in a broadband noise carrier. The response to the gaps were quantified by the modulation index (maximum firing rate after the gap minus the minimum firing rate after the gap divided by average rate before the gap). The modulation index increased with increasing gap duration and stimulus level. To a large extent, the modulation of auditory nerve firing rates in response to the gaps mirrored the behavioral data except that the neural data were slightly less impacted by the level of the noise carrier (Zhang et al., 1990).

Most natural sounds such as speech and other vocalizations contain rapid amplitude fluctuations occurring in different spectral regions. Since these amplitude fluctuations provide important cues needed to decode and interpret speech, it is important to determine how effective the auditory system is at detecting the amplitude fluctuations that occur at different rates (frequencies). One comprehensive method of assessing temporal resolution involves measuring the threshold for detecting sinusoidally amplitude modulated (SAM) noise. In this paradigm, a listener is presented with a broadband noise carrier presented at a fixed intensity. The envelope of the broadband noise is then increased and decreased (i.e., amplitude modulated, AM) in a sinusoidal fashion at a specified modulation frequency, or rate. The SAM noise threshold is defined as the smallest depth of AM that is detectable. SAM noise thresholds are measured over a broad range of AM frequencies and used to construct a transfer function that shows the modulation threshold as a function of the AM frequency. Modulation depth is expressed as the percent modulation of the carrier noise: Percent modulation depth = [(maximum–minimum level)/average level] × 100%]. Modulation depth can also be expressed in decibels: dB modulation depth = 20 log (% modulation depth/100%, a 1% modulation depth equals −40 dB, 10% modulation equals −20 dB, and 100% modulation depth equals 0 dB). Detection of amplitude modulation with noise carriers is best at low modulation frequencies (low percent modulation). Amplitude modulation thresholds are fairly constant at low frequencies, but gradually increase (larger percent modulation) above some critical frequency. By systematically changing the rate and depth of modulation, an amplitude-modulation transfer function can be constructed showing the modulation threshold expressed in dB or percent modulation as a function of the modulation frequency. The SAM noise transfer function summarizes the auditory system's sensitivity to different rates of amplitude modulation. The transfer function resembles a low-pass filter with a low-frequency plateau region followed by a gradual decline in AM sensitivity beyond the high-pass cutoff frequency (Viemeister, 1977; Salvi et al., 1982).

SAM noise thresholds were obtained from four monaural chinchillas using the aforementioned conditioned shock avoidance paradigm with stimuli presented according to the method of constant stimuli (Giraudi et al., 1980). SAM noise thresholds were measured using two broadband noise carriers; in one case the noise was low-pass filtered at 20 kHz and presented at an overall level of 73 or 53 dB SPL and in the second case, the noise was low-pass filtered at 10 kHz and presented at 72 or 52 dB SPL. In both cases, the noise was AM at 12 different modulation frequencies ranging from 2 to 4096 Hz in octave steps. The SAM transfer function for the chinchilla resembled a low-pass filter with a low-frequency plateau and gradual decline in sensitivity at the high frequencies. Similar to human listeners (Viemeister, 1979), the chinchilla's sensitivity to amplitude modulation remained constant for modulation rates between 2 and 32 Hz; the modulation threshold in that range was approximately 9% (roughly −20 dB). As the rate of modulation increased beyond 32 Hz, amplitude modulation detection thresholds declined at a relatively constant rate of 1.9 dB/octave out to 2048 Hz, after which thresholds plateaued at a modulation depth of 43% (approximately −7 dB) (Salvi et al., 1982). Interestingly, the amplitude modulation transfer functions were nearly identical across both noise bandwidths and both stimulus intensities, suggesting that the level and bandwidth of the noise carriers used in the study had little effect on amplitude modulation detection (Salvi et al., 1982). Two human listeners tested with the same equipment used to evaluate the chinchillas produced amplitude modulation transfer functions that were qualitatively similar to the chinchilla; however, there were several quantitative differences. Low-frequency SAM thresholds for humans were approximately 4% (−28 dB) compared to 9% (−20 dB) for chinchillas, i.e., humans performed somewhat better than chinchillas which may represent another potential limitation of the chinchilla model of temporal resolution. Above 32 Hz, the SAM transfer function for humans declined at a rate of 3.3 dB/octave and reached a plateau around 512 Hz; this contrasts with a 1.9 dB/octave rate of decline and 2048 Hz plateau for chinchillas.

To determine how a progressively increasing high-frequency hearing loss affected the detection of SAM noise, chinchillas were exposed to a series of intense noises in which the low-frequency cutoff of the noise decreased from 8, 4, 2, and then to 1 kHz. Each of these noise exposures lasted 1 week. Hearing tests were conducted during each week after the animals had been in the noise long enough to establish an ATS (i.e., >16 h). During each exposure week, the animals were removed from the noise for 1 h per day to obtain thresholds for tone bursts, noise bursts, and thresholds for SAM noise; these data were compared to pre-exposure values.

During the first week, the chinchillas were exposed to an octave-band noise centered at 8 kHz and presented at 90 dB SPL. During the second week, a second octave-band noise was added to the first noise; the second octave-band noise was centered at 4 kHz and presented at 86 dB SPL. The overall level of the noise exposure remained at 90 dB SPL but the additional 4 kHz component was 86 dB SPL. For the third and fourth weeks, octave-band noises centered at 2 kHz (90 dB SPL) and then 1 kHz (90 dB SPL) were added to the preceding noise exposures. As in the previous SAM noise study, the animals were trained using a conditioned shock avoidance paradigm and the test stimuli were presented using an adaptive tracking procedure where the intensity of the pure-tone and noise burst stimuli was decreased following correct detection of the stimuli and increased following a miss. Likewise, the depth of modulation was decreased following correct detection of the AM stimuli and increased following a miss (Giraudi-Perry et al., 1982; Henderson et al., 1984).

Exposure to the 8 kHz octave-band noise induced approximately 57 dB of ATS at the high frequencies but very little loss at or below 4 kHz. The addition of the 4, 2, and 1 kHz octave bands of noise caused a progressive decrease in the low-frequency boundary of the hearing loss to approximately 2, 1, and 0.5 kHz, respectively. In addition to inducing a high-frequency hearing loss, this series of noise exposures caused a reduction in the SL of the noise carriers used in the AM detection experiment. The 8 kHz octave-band noise reduced the SL of the noise carriers from 49 to 34 dB SL; this caused a modest increase in AM thresholds at the intermediate modulation frequencies (∼32 to 1024 Hz). The results obtained when the 4 and 2 kHz octave-band noises were added to the original noise were nearly identical to those obtained with the 8 kHz octave-band noise (i.e., similar results in terms of SL of the noise carrier and SAM noise transfer function). However, the 1 kHz octave-band noise exposure caused a severe reduction in the SL of the noise carrier so that the noise was now presented at roughly 10 dB SL. This resulted in a dramatic increase in AM detection thresholds across the entire range of modulation frequencies. To compensate for the hearing loss, the level of the noise carriers was increased by 23 dB SPL so that the noise carrier was now presented at 33 dB SL. This restored the SAM noise transfer function so that thresholds were comparable to that seen following the first three noise exposures (Henderson et al., 1984).

Afterwards, the chinchillas were removed from the noise and allowed to recover for 6 months to test for the effects of permanent hearing loss on AM detection. Following recovery, the animals had a slight PTS of 15 dB for the noise stimulus. The post-recovery thresholds for SAM noise were worse than pre-exposure values at AM frequencies from 2 to 512 Hz. Interestingly, when the level of the noise carrier was increased 15 dB to correct for the PTS, AM detection thresholds were still larger (worse) than pre-exposure values. As in the gap detection experiments, these results suggest that persistent damage to the high-frequency regions of the cochlea play an outsized role in auditory temporal resolution, and demonstrate the importance of evaluating auditory function beyond the standard audiogram (Fitzgibbons and Wightman, 1982; Fitzgibbons, 1983; Henderson et al., 1984).

Although SAM broadband noise can be used to provide a fairly comprehensive view of temporal processing in normal hearing individuals, it may be less than optimal for identifying the temporal processing deficits in individuals with hearing loss confined to specific frequency regions. This limitation could be overcome by applying AM modulation only to a specific spectral band (e.g., 3–6 kHz) within the broadband noise carrier. In this case, a SAM transfer function could be derived for a specific region in the audiogram. By shifting the AM component to different spectral regions, frequency-specific SAM transfer functions could be measured in both normal-hearing as well as hearing-impaired regions of the audiogram.

The chinchilla, whose audiogram is similar to humans, has been widely used in anatomical, electrophysiological, and perceptual studies of hearing loss induced by noise and other ototraumatic agents (Carder and Miller, 1972; Hamernik et al., 1984; Sinex et al., 1987; Morest et al., 1997; Bohne and Harding, 2000). Numerous psychophysical studies performed with normal-hearing chinchillas suggest that its temporal acuity is similar to humans and other mammals (Viemeister, 1979; Fitzgibbons and Wightman, 1982; Wagner et al., 2003; Radziwon et al., 2009). Therefore, the chinchilla serves as a useful animal model for obtaining detailed information regarding the temporal processing deficits that arise from various forms of hearing loss. Performing behavioral psychophysical experiments is extremely time-consuming. Thus, it is important to develop a strategy to gather as much data from each noise-exposed animal as possible. A highly effective method for optimizing data collection involves the use of long-duration (>16 h) noise exposures that induce an ATS. Because the hearing loss remains stable during ATS, the animals can be removed from the noise for a short period of time each day to identify the auditory processing deficits that occur during noise exposure and then returned to the noise to maintain that same level of ATS. Once the psychophysical measurements have been completed for a specific set of noise parameters, the intensity or bandwidth of the noise can be increased to establish a new, but more severe hearing loss or one affecting a broader range of frequencies. Once a new ATS is established, the psychophysical measurements can be repeated. By obtaining the same psychophysical measures before, during each stage of ATS, and after recovery from ATS, the investigator is able to gather a large body of psychophysical data from the same subject allowing for precise assessment of hearing loss and auditory processing deficits.

Most animal studies of NIHL typically utilize exposure durations lasting only a few minutes or hours in contrast to the long-duration noise exposure used in studies of ATS. Depending on the intensity of the short- or long-duration noise exposure, as well as other factors, hearing thresholds might recover completely, in which case the exposure would be classified as one causing only temporary threshold shifts (TTS). In other cases, these types of exposures can cause a compound threshold shift (CTS), one in which hearing thresholds are elevated to a significant degree immediately post-exposure, but which only partially recover (Chen et al., 2014). Thus, CTS exposures can be thought of as consisting of two parts, one that recovers (TTS) and one that does not (PTS). It is unclear from the literature, if the suprathreshold processing deficits associated with noise exposures that cause only TTS either from long or short duration exposures, are the same or different from exposures that cause PTS. One approach that might be used to answer this question would be to measure the suprathreshold processing deficits associated with a 1 week ATS exposure that causes a 30 dB threshold shift that completely recovers after the noise is turned off (i.e., a TTS exposure). These results could then be compared to suprathreshold processing deficits associated with a very long duration ATS exposure that results in the same 30 dB PTS. In this comparison, suprathreshold processing deficits are compared with the same amount of threshold shift, 30 dB of TTS vs 30 dB of PTS. However, such a comparison based on the same amount of threshold shift, usually linked to outer hair cell damage (Ryan and Dallos, 1975), may not accurately reflect other cochlear pathologies related to damage to inner hair cells and auditory nerve fibers (Salvi et al., 2016). Therefore, it seems likely that suprathreshold processing deficits are not only related to the amount of TTS and PTS, but also to other pathophysiological changes in the cochlea and central auditory pathway that are not reflected in the pure tone audiogram.

Human studies of hearing loss have largely focused on changes in the pure-tone audiogram, long considered the “gold standard” for quantifying temporary and permanent hearing impairments. The widely used “clinical audiogram” is designed to measure changes in hearing sensitivity in the speech frequencies from 250 to 8000 Hz, but ignores a large portion of the auditory spectrum between 8000 and 22 000 Hz. Frequencies within the clinical audiogram are crucial for processing speech sounds, which rapidly fluctuate in amplitude and frequency. The consonants in speech, located mainly in the upper frequency range, are relatively brief and less intense than lower frequency vowels, which are more intense and longer in duration. While neurons tuned to frequencies above 8000 Hz are less sensitive to low-frequency speech sounds, these high-frequency neurons do respond to temporal fluctuations if the low-frequency speech sounds are presented at high intensities (Kiang and Moxon, 1974; Horwitz et al., 2002). Under these conditions, ultra-high frequency neurons likely contribute to the temporal segregation of sounds that aide in speech perception. This interpretation is consistent with our chinchilla gap detection and amplitude modulation results that show impairments in temporal resolution when the hearing loss is largely restricted to frequencies above 8000 Hz. Therefore, one would expect that hearing loss in the extended high-frequency region, which degrades temporal resolution in the chinchilla, most likely does so in humans as well. However, few clinicians or psychophysicists test for hearing loss in the extended high-frequency range. This “hidden hearing loss” in the extended high-frequency region may explain why some individuals with clinically normal hearing have difficulty understanding speech or develop other hearing problems such as tinnitus (Kim et al., 2011; Babbage et al., 2017; Salvi et al., 2018).

The predominant source of information in speech, music, and environmental sounds is contained in the temporal and spectral changes in these complex acoustic stimuli. Acoustic signals that change rapidly over time convey more information than those that change slowly. The ability of the auditory system to follow the rapid temporal fluctuations in speech and music is constrained by the processing speed of the sensory hair cells and auditory nerve fibers located at the “front end” of the auditory system. Phase locking of auditory nerve fibers to individual cycles of a sine wave is extremely accurate below 1000 Hz (1 ms period), but some phase locking can occur up to several thousand Hz (Rose et al., 1967; Cai and Geisler, 1996). The ability of the auditory system is not only determined by the processing speed of each individual unit, but also by the number of nerve fibers capable of responding to the stimulus. Roughly 15 auditory nerve fibers contact a single inner hair cell. Nerve fibers in a state of absolute refractory would be prevented from responding to a rapid change in stimulus intensity whereas those in a state of relative refractoriness would be more likely to respond faithfully to every amplitude fluctuation in the stimulus.

Our studies of gap detection in the auditory nerve fibers of the chinchilla (Zhang et al., 1990) and other studies of forward masking (Harris and Dallos, 1979; Salvi et al., 1986), suggest that many aspects of temporal resolution can be accounted for at the level of cochlea; however, higher levels of the auditory pathway likely contribute to normal or impaired temporal processing. Recent studies have shown that intense sound exposures can preferentially damage the synapse that links the inner hair cells to the type I neurons, a condition referred to as synaptopathy (Kujawa and Liberman, 2015; Liberman and Kujawa, 2017). There is growing awareness that damage to the inner hair cells, the synapse at the base of the inner hair cells or type I neurons that contact the inner hair cells can lead to a form of hidden hearing loss in which the pure-tone audiogram and otoacoustic emissions are within the normal range but hearing in noise is nonetheless perturbed (Kujawa and Liberman, 2015; Liberman and Kujawa, 2017). Because the synapse at the base of the inner hair cells plays a critical role in the timing of spike discharges in auditory nerve fibers, many recent studies have attempted to identify temporal processing deficits in humans and animal models of synaptopathic hearing loss (Mehraei et al., 2016; Shi et al., 2016; Song et al., 2016; Paul et al., 2017).

Carboplatin, a second generation platinum drug, is much less ototoxic than cisplatin in humans and many other species; however, when administered to chinchillas it selectively damages the inner hair cells and their associated type I auditory nerve fibers (Ding et al., 1999, 2012). Carboplatin-induced inner hair cell lesions are unusual in that they cause a relatively uniform lesion along the length of the cochlea. Because the outer hair cells remain intact after carboplatin treatment, otoacoustic emissions remain normal as they do in noise-induced synaptopathy; however, the gross neural output of the cochlea reflected in the amplitude of the compound action potential is reduced in proportion to the amount of inner hair cell damage (Salvi et al., 2016). Despite the massive loss of inner hair cells and auditory nerve fibers, the chinchilla's post-carboplatin audiogram remains normal until the lesion exceeds ∼80% (Lobarinas et al., 2013). These results provide a stunning example of how insensitive the pure-tone audiogram is at detecting lesions involving the inner hair cells and auditory nerve fibers. Clinicians have long known that some individuals with so-called normal hearing have considerable difficulty hearing in noisy situations (Salvi et al., 2018). Because carboplatin damages many auditory nerve fibers, very few non-refractory neurons would be available to detect a tone presented in background noise. Indeed, when carboplatin-treated chinchillas were evaluated on their ability to detect a tone in a noisy background, they showed significant deficits despite having normal pure-tone thresholds in quiet (Lobarinas et al., 2016). Following the same reasoning, if carboplatin-treated chinchillas have fewer inner hair cells and nerve fibers available to respond to a rapid change in the acoustic environment, then one would predict that their gap detection performance would also be impaired. Indeed, when carboplatin-treated chinchillas were tested on their ability to detect silent gaps in a continuous noise, their gap detection thresholds were much larger than normal even though they presented with normal pure-tone audiograms (Lobarinas, 2006).

Normal-hearing listeners have remarkably good temporal resolution allowing them to rapidly decode the influx of acoustic information in a speech stream. The inability of hearing-impaired listeners to follow these rapid amplitude fluctuations results in the loss of information that impairs speech recognition and language processing. Because its hearing range is similar to humans, the chinchilla has become an important animal model in identifying the temporal processing deficits caused by noise- or drug-induced damage to the cochlea. The studies described in this review illustrate how within-subject experimental designs can be used to precisely characterize deficits in auditory temporal resolution.