Age-related Changes in Neural Coding of Envelope Cues: Peripheral Declines and Central Compensation

Abstract

Aging listeners often experience difficulties in perceiving temporally complex acoustic cues in noisy environments. These difficulties likely have neurophysiological contributors from various levels of auditory processing. Cochlear synapses between inner hair cells and auditory nerve fibers exhibit a progressive decline with age which is not reflected in the threshold audiogram. The functional consequences of this loss for the coding of suprathreshold sound remain poorly understood. Recent studies suggest that cochlear synaptopathy results in degraded representations of temporal envelope cues at the earliest levels of the auditory pathway. Central nuclei downstream of the auditory nerve exhibit a compensatory plasticity in response to this deafferentation, in the form of altered gain. This results in a modulation frequency selective increase in the representation of envelope cues at the level of the auditory midbrain and cortex. These changes may be shaped by mechanisms such as decreased inhibitory neurotransmission occurring with age across various central auditory nuclei. Altered representations of the differing temporal components of speech due to these interactions between multiple levels of the auditory pathway may contribute to the age-related difficulties hearing speech in noisy environments.

Aging listeners often experience difficulties in processing speech and other temporally complex sounds, particularly in noisy environments. These difficulties can arise due to changes occurring at several levels of the auditory pathway from the ear to the brain, with or without threshold sensitivity losses. In this review, we consider the role of altered representations of the temporal properties of complex sounds in the peripheral and central auditory pathways, and their contributions to functional declines in hearing with age.

SUPRATHRESHOLD DECLINES IN AUDITORY TEMPORAL PROCESSING WITH AGE

In the auditory periphery, cochlear hair cell damage or loss has long been considered a hallmark of age-related sensorineural hearing loss. In particular, outer hair cell damage is often associated with poorer hearing thresholds (Dallos and Harris, 1978), assessed clinically by the behavioral pure tone audiogram. Age-related hearing loss defined as declines in threshold sensitivity is exceedingly common; estimates suggest that over 60% of individuals over 70 years of age have a hearing loss significant enough to interfere with communication (Lin et al., 2011a). Left untreated, hearing loss decreases the quality of life and has also been associated with other age-related comorbidities such as cognitive impairment, dementia or Alzheimer’s disease (Gates et al., 2011; Lin and Albert, 2014; Dawes et al., 2015; Fischer et al., 2016; Deal et al., 2017; Wei et al., 2017). Patterns of loss in hearing sensitivity have been associated with underlying cochlear pathologies due to age, both using histological studies of human temporal bones (Schuknecht and Gacek, 1993) and informed by experiments in animal models (Dubno et al., 2013). However, increasing evidence shows that the overt loss of threshold sensitivity measured by the audiogram fails to capture critical aspects of functional hearing declines that older adults experience (He et al., 1998, Hind et al., 2011, Tremblay et al., 2015). Even when matched for good audiometric thresholds, older listeners show performance declines on tasks that require the processing of timing cues in sounds (Pichora-Fuller and Souza, 2001, Frisina and Frisina, 1997).

The auditory system processes behaviorally relevant sound stimuli across multiple time scales. Word and syllabic rates captured by the fluctuations of the speech envelope are <50 Hz, the periodicity envelope which conveys speaker identity and emotion is between 50 and 500 Hz and the rapidly changing temporal fine structure required for hearing speech in noise when envelope cues are degraded occurs at rates >500 Hz to a few thousand Hz (Bregman, 1990; Rosen, 1992). Effective representations of these temporal features are critical for real world communication, which often occurs in a background of other talkers or environmental noise (Mattys et al., 2012). Older and even middle-aged adults experience the most difficulties in processing these complex spectro-temporal cues necessary for communication, especially in challenging acoustic environments (Walton, 2010; Ruggles et al., 2012; Fullgrabe et al., 2015). The presence of multiple speakers, background noise or reverberation degrade the temporal regularities found in speech, and cause deleterious effects on speech perception (Best et al., 2009, 2010; Anderson et al., 2011; Ruggles et al., 2011), even when thresholds remain well preserved (Frisina and Frisina 1997; Gordon-Salant and Fitzgibbons 2001; Goupell, et al. 2017; Jaekel, et al. 2018). Even without threshold evidence for peripheral involvement, such declines in hearing function with age must nevertheless consider neural changes in the peripheral and central auditory pathway.

PROGRESSIVE COCHLEAR DEAFFERENTATION WITH AGING

Recent studies have revealed what may be an important peripheral contributor to these temporal processing declines. Synapses between inner hair cells (IHCs) and auditory nerve fibers (ANFs) are highly vulnerable to both aging and noise exposure (Fig. 1A, B, for review see Kujawa and Liberman, 2015). Normal aging results in a progressive decline in the number of these synapses (Sergeyenko et al., 2013). Synaptic loss precedes loss of hair cells (Fig. 1C) and changes in hearing thresholds. Loss of ANFs and their cell bodies in the spiral ganglion follows with a delay, but the functional connections/communication between affected fibers and their target IHCs are lost with the synapses. Age-related cochlear synaptopathy can be exacerbated by early exposure to noise that causes synaptopathy but no permanent threshold shifts (Fernandez et al., 2015), suggesting that these two etiologies yielding synapse loss may be targeting the same neuronal subpopulations. Initially reported in mouse (Kujawa and Liberman, 2009), this cochlear deafferentation with aging and/or noise exposure is widespread in various mammalian species, including guinea pigs (Lin et al., 2011b), rats (Mohrle et al., 2016), gerbils (Gleich et al., 2016), chinchillas (Hickox et al., 2017), rhesus macaques (Valero et al., 2017) and in post-mortem human temporal bones (Viana et al., 2015; Wu et al., 2019). Since subtotal synaptopathy goes largely undetected by the threshold audiogram, it has been termed as a ‘hidden’ hearing loss (Schaette and McAlpine, 2011).

Fig. 1.

Progressive cochlear synaptopathy with aging is accompanied by degraded representation of envelope cues in the early auditory pathway. (A) Schematic cross section showing three of the ~20 auditory nerve fibers (ANFs) making synaptic contact with an IHC. Presynaptic ribbons and postsynaptic receptor patches are also schematized. The x–y–z axis shows the viewing angle for the confocal x–y projections shown for example IHCs in (B), where immunostaining reveals the juxtaposition of pre-synaptic ribbons (red) and post-synaptic receptor patches (green). Inset shows magnified example images of presynaptic ribbons (red) with apposing post-synaptic glutamate receptor patches (green). (C) Mean (±SEMs) percent survival of cochlear synapses (green line), inner hair cells (gray solid line) and outer hair cells (gray dashed line), relative to 16-week-old animals, at 30 kHz. (D) Mean (±SEM) EFR amplitudes at 1024 Hz AM measured as a function of sound level across the various age groups at 30-kHz frequency region. Dashed lines indicate responses below the noise floor. (E) Mean (±SEM) EFR amplitudes at equal sensation levels (SL) of 0 dB (threshold) to 30 dB. (F) Correlation between EFR amplitudes from panel E at 30-dB SL (dashed box) and the number of remaining synapses/IHC across all the age groups. (G) Mean (±SEM) ABR wave 5: wave 1 ratios across age at 80-dB SPL. Figure modified from Parthasarathy and Kujawa, 2018).

Various lines of evidence have suggested that the vulnerable subpopulations of neurons in the auditory nerve may be those with low spontaneous rates (low-SR) of firing. Losses in these low-SR neurons with aging, noise exposure or ouabain-induced neuropathy have been studied using single-neuron recordings from the auditory nerve (Schmiedt et al., 1996; Furman et al., 2013; Bourien et al., 2014) as well as immunohistological analysis of the spatial organization of these synapses along the modiolar-pillar axis of the inner hair cell (Liberman et al., 2011, 2015; Yin et al., 2014;). While the exact mechanism by which the preferential loss of low-SR fibers may affect auditory processing is still under debate (Carney, 2018), these neurons have higher thresholds, show a greater preference for synchronized firing at moderate to high sound levels and are resistant to the effects of background noise, properties that should aid suprathreshold sound processing in real-world listening situations (Costalupes et al., 1984; Joris and Yin, 1992).

Multiple physiological measures have shown promise as non-invasive assays of cochlear synaptopathy. These include the wave 1 of the auditory brainstem response (ABR) (Kujawa and Liberman, 2009; Sergeyenko et al., 2013), the middle ear muscle reflex (MEMR) (Valero et al., 2016; Valero et al., 2018) and envelope following responses (EFRs; Shaheen et al., 2015; Parthasarathy and Kujawa, 2018). Additionally, MEMRs, which have low-SR neurons as primary drivers (Liberman and Kiang, 1984; Kobler et al., 1992), and EFRs, which use stimuli that may more effectively activate low-SR neurons (Joris and Yin, 1992), may be useful in probing functional consequences of the targeted loss of these neurons in aging and noise-exposed individuals. While these measures have been shown to be reliable indicators of synaptopathy in animal models, the results from human subject groups have been mixed, with some studies showing results that could be consistent with the presence of an underlying synaptopathy (Konrad-Martin et al., 2012; Stamper and Johnson, 2015; Liberman et al., 2016; Mehraei et al., 2016; Wojtczak et al., 2017; Mehraei et al., 2017; Bramhall et al., 2017, 2018; Valderrama et al 2018), some mixed outcomes (Grose et al., 2017) and others finding no evidence of its presence (Grinn et al., 2017, Guest et al., 2017; Prendergast et al., 2017a,b; Fulbright et al., 2017). Age-graded declines in auditory nerve fiber contacts with IHCs documented in human temporal bones (Viana et al., 2015; Wu et al., 2019) have closely paralleled the age-graded declines in mouse (Sergeyenko et al., 2013); however, the possibility that functional consequences of this loss are less robust in humans than those suggested by the animal data cannot currently be ruled out. Additional contributors to the discrepancy may include uncertainties associated with estimating lifetime noise exposure using self-report questionnaires, variability in the nature of the exposures experienced by the various studied subjects, which ranges from military arms fire to single exposures using personal music players, discrepancies in measuring ABRs and EFRs in humans due to variability in head sizes and noise floors, limitations of time which make testing multiple sound levels and frequencies impractical and, of course, the inherently greater variability in highly outbred species, like humans. The presence of outer hair cell loss is another confounding factor, especially when studying age-related cochlear synaptopathy. However, in animal models, synaptopathy due to age is progressive throughout the lifespan, and begins well before frank outer hair cell loss. Therefore, testing middle aged or older listeners with normal audiograms might be beneficial in isolating the effects of synaptopathy. Future work in optimizing stimulus parameters, recording techniques and assessment of noise exposure history are needed to help resolve these inconsistencies. Although the diagnosis of underlying synaptopathy is in itself an important question to be addressed, the next step is a greater understanding of the functional consequences of this synaptopathy on the neural coding of sounds.

ASSESSING THE FUNCTIONAL CONSEQUENCES OF AGE-RELATED SYNAPTOPATHY ON THE NEURAL CODING OF SOUNDS

Cochlear synaptopathy is hypothesized to affect sound processing through the degradation of complex temporal cues. One theory of stochastic undersampling posits that partial deafferentation would reduce the “sampling rate” with which neurons of the auditory nerve represent sound (Marmel et al., 2015). According to this view, synaptopathy would contribute to the degraded temporal resolution seen in older listeners as temporally complex stimuli would require a greater “resolution” to effectively represent all their acoustic features compared to simple stimuli (Lopez-Poveda and Barrios, 2013; Lopez-Poveda, 2014).

To test these theories in human clinical populations and in animal models requires non-invasive measures that can represent the spectro-temporal complexities present in relevant real world stimuli. It is in this context that the EFRs provide valuable, additional information. Since EFRs are a faithful representation of the stimulus envelope, they can provide a snapshot of auditory processing to temporally complex stimuli. EFRs are population responses evoked by the synchronized responses of groups of neurons along the auditory pathway and provide complementary information to the ABRs evoked by brief phasic stimuli (Parthasarathy et al., 2014). Neurons of the auditory pathway represent the temporal aspects of sounds in their firing patterns with different degrees of fidelity (Joris et al., 2004), and it is the ensemble activity of these neurons that is thought to be captured at the scalp by the EFRs. EFRs with generators in the auditory nerve may be more sensitive to noise-induced cochlear synaptopathy than the ABR wave 1 (Shaheen et al., 2015), since the low-SR fibers thought to be preferentially affected by synaptopathy contribute to the sustained neural responses to longer, temporally complex sounds (Joris and Yin, 1992). EFRs elicited to speech sounds, such as a consonant–syllable complex, or to sinusoidally amplitude-modulated tones have been used successfully to probe auditory processing in various human clinical populations, including aging listeners (Tremblay et al., 2003; Clinard et al., 2010; Krishnan and Agrawal, 2010; Anderson et al., 2012; Clinard and Tremblay, 2013; Bidelman et al., 2014; Ananthakrishnan et al., 2016), and in animal studies where they can be recorded along with more invasive measurements and correlated with underlying histopathology (Zhong et al., 2014; Shaheen et al., 2015; Herrmann et al., 2017; Parthasarathy et al., 2018). Hence the EFR shows promise in bridging the gap between human studies and animal models in understanding the temporal processing deficits caused by age-related cochlear synaptopathy.

For understanding the progression and interaction of peripheral and central changes with aging, it would be advantageous to dissociate neural generators. Within this context of spatial specificity, one limitation of the EFR is the lack of spatial resolution compared to metrics like the ABR, which have clearly defined peaks whose neural generators have been more or less localized. The longer duration of stimuli in the EFRs causes some overlap between the responses evoked from different generators with differing response latencies. However, this limitation can be minimized by a few approaches. The first is stimulus selection. The upper limit of phase-locking, or the degree to which neurons faithfully represent the temporal regularities in sound in their spike timings, decreases along the ascending auditory pathway (Joris et al., 2004). Although exact boundaries differ by species, neurons in the auditory nerve can phase lock up to a few thousand hertz, neurons in the inferior colliculus in the midbrain can only phase lock to a few hundred hertz, and neurons in the auditory cortex typically reach their limit by one hundred hertz (see Joris et al., 2004 for review). While using speech sounds to evoke EFRs provides real-world relevance, the use of sinusoidally amplitude-modulated tones can provide better localization of neural generators by fine tuning the stimulus temporal characteristics. EFRs elicited by slower AM frequencies (~40 Hz) are thought to be primarily cortical in origin (Herdman et al., 2002; Kuwada et al., 2002; Ross et al., 2003), whereas EFRs elicited by AM frequencies around 110 Hz are thought to be subcortical, reflecting thalamic, midbrain and brainstem activity (Picton et al., 2003; Parthasarathy and Bartlett, 2012). Recent studies have extended these observations using faster AM frequencies around 800–1000 Hz, with generators from the auditory nerve, as confirmed by the application of ouabain, a neurotoxin, to the round window of mice. Ouabain can be applied at doses that eliminate neural responses without affecting outer hair cell based DPOAE thresholds or amplitudes (Lang et al., 2011; Yuan et al., 2014). Ouabain application suggests that EFRs to modulation frequencies centered around 1000 Hz have putative generators in the auditory nerve (Shaheen et al., 2015; Parthasarathy and Kujawa, 2018) and that faster EFRs to ~4000 Hz AM may even be used to probe hair cell based responses to temporally modulated stimuli (Parthasarathy and Kujawa, 2018). These measurements can serve to confirm that EFRs can be obtained from the earliest neural generators of the auditory pathway.

The second approach to achieving better spatial resolution with EFRs is to use simultaneous recording from multiple channels (Galbraith et al., 2001; Galbraith et al., 2006; Parthasarathy and Bartlett, 2012; Bidelman, 2015; King et al., 2016). Using different electrode montages results in each channel primarily capturing a different electrical dipole, and the difference in the spatial geometry of the dipoles can emphasize rostral versus caudal generators along the auditory pathway (Parthasarathy and Bartlett, 2012; Parthasarathy et al., 2018). Multichannel recordings can also be leveraged to obtain better SNRs of the recordings, by modeling the various dipoles at the scalp and deconstructing them using a complex PCA (Bharadwaj and Shinn-Cunningham, 2014). These approaches can be useful especially when using speech-like stimuli, where the stimulus properties cannot be manipulated to emphasize different generators without compromising on different temporal cues essential for speech comprehension.

TEMPORAL PROCESSING DEFICITS WITH AGE MEASURED USING ENVELOPE FOLLOWING RESPONSES

Temporal processing probed using EFRs reveal widespread deficits along the auditory pathway with age. Degradations in the representation of envelope cues begin at the earliest neural generators with age (Fig. 1D). These deficits begin prior to any changes in hearing thresholds, decline progressively with age and persist at suprathreshold sound levels (Fig. 1E). EFR amplitudes are strongly correlated with the remaining number of cochlear synapses (Fig. 1F). This suggests that age-related cochlear synaptopathy is associated with decreased representation of envelope cues. Aging also results in a decreased dynamic range for representation of level and amplitude modulation depth at these early neural generators (Parthasarathy and Kujawa, 2018). The use of such high modulation frequencies is technically challenging in humans due to the interactions among filter widths, cochlear frequencies and modulation frequencies. Moreover the response amplitudes at faster modulation frequencies also show a much steeper roll off. However, studies which have measured tonal phase-locking to carrier frequencies around 1000 Hz show greater deficits in response amplitudes in aged listeners (Clinard et al., 2010; Marmel et al., 2013).

Degradations in envelope processing persist when using slower modulation frequencies (~100–500 Hz) with generators in the midbrain (Kiren et al., 1994; Chandrasekaran and Kraus, 2010; Herdman et al., 2002). Results from studies using speech tokens to evoke EFRs can also be generally interpreted to contain responses arising primarily from similar subcortical sources, since the F0 envelope periodicities used in these tokens correspond to similar modulation frequencies (Rosen, 1992; Bidelman, 2018). Multiple studies have shown that aging decreases the strength of phase-locking to envelope cues (Anderson et al., 2012; Clinard and Tremblay, 2013; Schoof and Rosen, 2016). EFRs elicited using amplitude modulation frequencies in similar ranges also show a decrease with age (Parthasarathy and Bartlett, 2012). Hence, declines in the integrity of temporal processing seen at the level of the auditory nerve can also be observed along the ascending auditory pathway with age.

However, some evidence suggests that these changes are often minimal when the envelope cues are salient, such as in quiet, with slow modulation frequencies (<50 Hz) or with large modulation depths (Parthasarathy et al., 2010; Parthasarathy and Bartlett, 2011). The wave 1:5 ratio, a marker for central compensation, increases with age (Fig. 1G, Sergeyenko et al., 2013; Parthasarathy and Kujawa, 2018). Compensating for loss of hearing thresholds by increasing sound level does not result in equal response amplitudes in both ABRs and EFRs (Lai et al., 2017). ABR and EFR amplitudes are correlated in young but not aged animals, suggesting a decoupling between the neural mechanisms producing phasic onset responses that constitute the ABR and the sustained responses that constitute the EFR (Parthasarathy et al., 2014). Matching ABR wave 1 amplitudes between young and aged animals results in increased response amplitudes of the EFRs with age at high modulation depths (Lai et al., 2017). Deficits are also exacerbated with the addition of background noise, lower modulation depths or multiple stimuli (Parthasarathy et al., 2010, 2016; Parthasarathy and Bartlett, 2011). These results suggest some form of compensatory activity emerging at the level of the brainstem or midbrain that may restore evoked potential responses in quiet and for strong envelope cues but not for complex listening conditions and degraded envelope cues.

The compensatory plasticity seen at the level of the midbrain is stronger at cortical levels, especially in quiet. EFR amplitudes for responses arising primarily from cortical generators are less affected by aging (Bidelman et al., 2014; Goossens et al., 2016). However, these compensatory mechanisms do not seem sufficient to minimize age-related declines in the presence of signal degradations such as decreased modulation depths (Dimitrijevic et al., 2016) and the presence of multiple speakers (Presacco et al., 2016a), or for adaptation to more complex stimulus statistics (Herrmann et al., 2018). There is also evidence from human literature of compensatory gain at the cortical level, and an altered interaction between the auditory nuclei of the midbrain and the cortex (Bidelman et al., 2014; Presacco et al., 2016b; Valderrama et al., 2018). This could, in part, explain why many studies find only subtle differences in temporal processing with age (Schoof and Rosen, 2016; Paraouty and Lorenzi, 2017), and even when they do find an age-related degradation of temporal processing using both behavioral and neurophysiological measures, they are often not correlated (Clinard et al., 2010; Schoof and Rosen, 2016). These results could be explained by compensatory plasticity occurring at the higher levels of the auditory pathway and causing a divergence between brainstem encoding of sound features and the eventual perception of sound.

CELLULAR MECHANISMS UNDERLYING COMPENSATORY PLASTICITY WITH AGE

The specific neural mechanisms by which the aging brain encodes the temporal regularities in sound, especially at suprathreshold sound levels, is an ongoing field of study that will be critical in understanding the neural basis for difficulties understanding speech in noise. Changes in the interaction between multiple levels of processing after the auditory nerve and the mechanisms underlying these changes are not yet fully understood. Within the cochlear nuclei, auditory nerve projections terminate onto a variety of excitatory and inhibitory cell types that each process their inputs differently and have different spectro-temporal profiles (Trussell, 1999; Yu and Young, 2000), all of which change with age (e.g. Schatteman et al., 2008; Xie, 2016; Xie and Manis, 2017). Cochlear synaptopathy and the silencing of auditory neurons that results may affect suprathreshold coding by degrading the representation of temporal cues being sent to the rest of the auditory pathway (Xie, 2016; Goodman et al., 2018; Parthasarathy and Kujawa, 2018). This would occur across cochlear nucleus neurons tuned to a wide range of frequencies due to the contributions of the low-frequency tails of higher-frequency neurons (Parthasarathy et al., 2016; Lai and Bartlett 2018). Auditory nuclei further downstream may show compensatory plasticity in an attempt to reestablish the homeo-static balance between excitation and inhibition. This compensatory plasticity has been observed in other forms of sensorineural hearing loss and after neurotoxic ouabain, where a decrease in afferent outflow from the periphery results in increased activity in the auditory midbrain and cortex (Kotak and Sanes, 2003; Kotak et al., 2003, 2005; Barsz et al., 2007; Chambers et al., 2016a, b). This helps compensate auditory responses to simple stimuli but comes at the cost of altered temporal processing to complex sounds (Chambers et al., 2016a). These compensatory changes are, in part, mediated by cortical inhibitory circuits which are thought to decrease with a decrease in peripheral inputs (Resnik and Polley, 2017). Inhibitory neurotransmitters GABA and glycine serve to shape responses to temporally complex sounds in the central auditory pathway including the cochlear nucleus (Backoff et al., 1999; Keine et al., 2016), inferior colliculus (Palombi and Caspary, 1996a; Caspary et al., 2002), auditory thalamus (Cai and Caspary, 2015) and cortex (Wang et al., 2002; Razak and Fuzessery, 2009, 2010; Gaucher et al., 2013), increasing their selectivity for acoustic features in stimuli like envelope cues, frequency modulation and preferred frequency. Hence the decrease in inhibition may serve to decrease the selectivity of neurons to the acoustic features in sounds, and thereby decrease the fidelity of neural coding.

Evidence for this compensatory plasticity with age is seen as early as the cochlear nucleus, where there is a decrease in inhibitory glycine binding sites (Wang et al., 2009). This decrease in inhibitory neurotransmission is accompanied by age-related temporal processing deficits in the cochlear nucleus; for example, decreased selectivity for AM frequencies and decreased phase-locking to AM sounds especially evident at lower modulation depths (Caspary et al., 2005; Schatteman et al., 2008). At the inferior colliculus, there is a decrease in GABAergic neurotransmission with age (Raza et al., 1994; Burianova et al., 2009; Rabang et al., 2012). This is accompanied by a shift in the temporal response properties of the neurons toward lower modulation frequencies and decreased responses to rapid modulation rates. However, differences in the responses of aged IC neurons to simple monaural and binaural stimuli and slow modulation frequencies are minimal (Willott et al., 1988a,b, 1991; Palombi and Caspary, 1996a,b; Rabang et al. 2012; Herrmann et al. 2017), and changes emerge only when using faster modulation rates or more complex stimuli like speech envelopes. Comparing local field potentials, as a marker for synaptic inputs to IC neurons (Logothetis et al., 2001), and the spiking activity of the IC neurons shows that the inputs coming in to the IC are degraded with age. However, there is a relative increase in activity in the aged IC neurons that is selective to slower modulation frequencies (Herrmann et al., 2017). Similar experiments using speech-like envelopes indicate a selective increase in representation of periodicity envelope cues, relative to the synaptic inputs, in the neurons of the aged IC (Fig. 2; Parthasarathy et al., 2018). The decreasing inhibitory neurotransmission is further observed in the auditory forebrain, with decreased GABA_A receptor density and currents in the thalamus (Richardson et al., 2013b) and in the auditory cortex (Caspary et al., 2013; Stebbings et al., 2016). This is accompanied by an increase in spontaneous firing rate indicative of an altered balance between excitation and inhibition, but physiological changes with age, such as a decrease in synchronization to some amplitude-modulated sounds, are subtle under simple conditions and emerge only with degraded spectro-temporal cues (Richardson et al., 2013a; Engle and Recanzone, 2013; Overton and Recanzone, 2016; Cai et al., 2016; Aushana et al., 2018). These studies indicate that the results seen at the population level using scalp electrodes are mostly corroborated by intra-IC LFPs and partially corroborated at the single-unit level.

Fig. 2.

Responses from the inferior colliculus show evidence for magnified representation of envelope cues with age (A) Amplitude spectra derived from fast Fourier transforms of scalp-recorded envelope following responses to a modified speech token. Boxplots show coefficients from the cross-correlation for different frequency bands. (B) Spectrum of normalized vector strength for local field potentials (LFPs) recorded from the inferior colliculus of young and aged F-344 rats around the periodicity envelope frequency (105–115 Hz). (C) Spectrum of normalized vector strength of multi-unit activity recorded from the inferior colliculus of young and aged Fischer-344 rats. *p < 0.05. (Figure modified from Parthasarathy et al., 2018).

Understanding the neural mechanisms underlying altered temporal processing with age ultimately will require tracking the changes in identified neuronal subpopulations as a function of age, and comparing them to far-field evoked potential recordings to see how changes observed at the cellular level translate to population responses in humans and animal models. In this process, computational modeling can help to understand how the underlying cellular mechanisms translate to age-related changes in population responses that are seen at the scalp. Phenomenological models can recreate physiological responses observed in humans and animal models (Zilany et al., 2009, 2014; Verhulst et al., 2015; Saremi et al., 2016), and are especially useful in considering multiple nuclei and frequencies contributing to suprathreshold coding observed using evoked potentials (Nelson and Carney, 2004; Verhulst et al., 2018). These models can help isolate the potential contributions of the various components, such as threshold elevations, outer hair cell loss and synaptopathy, in carefully controlled ways that are harder to do in aging humans and to a lesser extent other animal models (Bharadwaj et al., 2014, Verhulst et al., 2015, 2016; Parthasarathy et al., 2016). Biophysical models can help us further understand the cellular basis for age-related changes in sound representation (Rabang and Bartlett, 2011; Rabang et al., 2012; Manis and Campagnola 2018), though optimal constraining of parameters is required when dealing with such models to avoid overfitting (Oleksiak et al., 2011; Vayrynen et al., 2016; Coventry et al., 2017). Titrating the various excitatory and inhibitory components to recreate neuronal responses can help us understand the cellular basis for the changes occurring due to aging at multiple stages of the auditory pathway.

Finally, a task as complex as speech perception is dependent on various extra-sensory processes including attention, motivation, cognitive abilities, working memory and listening effort (Clayton et al., 2016; Eckert et al., 2016; Hornsby et al., 2016; DeCaro et al., 2016). A detailed consideration of these studies is beyond the scope of this review (but see Peelle, 2018 for review, and Pichora-Fuller et al., 2016 for a framework for effortful listening). However, the effects of sensory changes, peripheral and central, on higher order executive functions, and vice versa are critical components in studying age-related hearing loss (see Peelle and Wingfield, 2016 for review). Degradations in the sensory coding of auditory stimuli with age may be compensated by increased listening effort (Ayasse et al., 2017), though this may come at the cost of mental fatigue (Hornsby et al., 2016; Moore et al., 2017). A true understanding of age-related deficits with speech communication will require a consideration of all these higher order executive functions, and their interactions with sensory processing within the auditory pathway.

The effects of age on hearing can manifest at multiple levels of the auditory pathway. Progressive cochlear synaptopathy is an important peripheral contributor to age-related declines in hearing function and is associated with decreased representation of temporal envelope cues at the earliest regions of the auditory pathway. These deficits persist at suprathreshold sound levels and are independent of changes in hearing thresholds. More central auditory regions may exhibit compensatory plasticity due to this reduced peripheral drive, mediated in part by the decrease in inhibitory neurotransmission. These compensatory changes have differential effects on the various components of speech, with aberrant enhancement of periodicity envelope cues, and a decrease in representation of temporal fine structure. A deeper understanding of these compensatory changes as well as the contributions of top-down, auditory and non-auditory processes will be required to fully understand difficulties in speech comprehension with age.

Abbreviations:

ABR

auditory brainstem response

ANFs

auditory nerve fibers

EFRs

envelope following responses

IHCs

inner hair cells

low-SR

low spontaneous rates

MEMR

middle ear muscle reflex